Biological Low-Voltage Scanning Electron Microscopy
Biological Low-Voltage Scanning Electron Microscopy Edited by
Heide Schatten Department of Veterinary Pathobiology, University of Missouri-Columbia, Columbia, Missouri, USA
James B. Pawley Department of Zoology, University of Wisconsin, Madison, Wisconsin, USA
Heide Schatten Department of Veterinary Pathobiology University of Missouri-Columbia 1600 E. Rollins Street Columbia, MO 65211 USA
[email protected]
James B. Pawley Department of Zoology University of Wisconsin 250 N. Mills Street Madison, WI 53706 USA
[email protected]
Library of Congress Control Number: 2007931613 ISBN 978-0-387-72970-1
e-ISBN 978-0-387-72972-5
c 2008 Springer Science+Business Media, LLC All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. Printed on acid-free paper. 9 8 7 6 5 4 3 2 1 springer.com
Preface
Ever since its advent in the 1950s, the scanning electron microscope hs offered an image of surface structures that was both uniquely detailed and easily interpretable. Although initially, the resolution of the images it produced did not match that of those made using the transmission electron microscope, this deficit no longer relevant on biological specimens where, for both instruments, image quality of is now limited fundamentally by the fragility of the specimen. High-resolution, low-voltage scanning electron microscopy (LVSEM) is now a powerful tool to study biological structure. Particularly when coupled with novel specimen preparation techniques, it has allowed us to understand in three dimensions objects that previously could only be imaged from serial-sectioned material analyzed with transmission electron microscopy. Because LVSEM provides images with clear topographic contrast from specimens coated with only an extremely thin metal coating, it can provide highresolution images of macromolecular complexes and of structural interactions that are free from the confusion of structural overlap. What is more, it provides this analysis in the context of being able to view the large specimen areas needed to provide context. These capabilities are particularly useful for applications in cellular biology. In addition, specific molecular components on surfaces and internal cell structures can be identified by using colloidal-gold labeling techniques. Particular internal structures can be viewed either by isolating them or by using novel fracturing and sectioning techniques that cause the internal components to occur on the outer surface of the specimen. There, the LVSEM can image them with a degree of topological precision that is often not possible with conventional TEM. Given these tremendous capabilities, it seemed to us both surprising and unfortunate that that LVSEM was not used more often to study biological structure. In a time when interest is extending from the genome to the proteome and when we increasingly want to know not only the existence but also the localization and interactions of specific molecules and molecular complexes, it seemed to us that LVSEM was the ideal modality for answering a myriad of important questions in cellular biology and development. What seemed to be needed was a way to make potential users more aware of LVSEM’s unique and powerful capabilities and also to provide the reader with both meaningful examples from a variety of applications and suitable protocols for preparing specimens. v
vi
Preface
We approached a number of leaders in the field with this idea and received a most enthusiastic response. The topics chosen were selected to be of interest to scientists, technicians, students, teachers, and to all who are interested in expanding their knowledge related to LVSEM. The specific topics covered in this book include highresolution LVSEM applications to cellular biology and detailed specimen preparation techniques for molecular labeling and correlative microscopy, cryoSEM of biological samples, and new developments in LVFESEM instrumentation in x-ray microanalysis at low beam voltage. Biological Low Voltage Scanning Electron Microscopy covers many aspects of specimen preparation and provides specific protocols for practical applications that are commonly not available in research papers. It also gives general as well as detailed insights into the theoretical aspects of LVSEM. The book is intended for a large audience as a reference book on the subject. By providing both theory and practical applications related to imaging biological structures with LVFESEM, we hope that it will fill a gap in the literature. During the editing process of this book, two of our most treasured colleagues, who have advanced the field immensely, passed away. Both the late Dr. Hans Ris and the late Dr. Stanley Erlandsen were passionate about the usefulness of LVSEM to enhance their own research, and as such, they left a wealth of new knowledge, novel techniques, and ideas for new applications for the scientific community. Their contributions are of great value to future scientist, students, technical staff, and many other using LVSEM. The editors are most grateful to all authors who have contributed their superb and unique expertise to this project and shared their insights with the present community interested in microscopy and those who will enter the field in the future. We would like to thank Kathy Lyons, our ever-so-patient editor at Springer. In addition, one of the editors (JP) would also like to thank Bill Feeny, the Zoology Departmental artist, and Kandis Elliot, the Botany Department artist, for their help in preparing the figures. Heide Schatten, Columbia, Missouri USA Jim Pawley, Madison, Wisconsin USA
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
v
List of Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix 1 The Early Development of the Scanning Electron Microscope Dennis McMullan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
2 LVSEM for Biology James B. Pawley . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
3 The Aberration-Corrected SEM David C. Joy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 4 Noise and Its Effects on the Low-Voltage SEM David C. Joy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 5 High-Resolution, Low Voltage, Field-Emission Scanning Electron Microscopy (HRLVFESEM) Applications for Cell Biology and Specimen Preparation Protocols Heide Schatten . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 6 Molecular Labeling for Correlative Microscopy: LM, LVSEM, TEM, EF-TEM and HVEM Ralph Albrecht and Daryl Meyer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 7 Low kV and Video-Rate, Beam-Tilt Stereo for Viewing Live-Time Experiments in the SEM Alan Boyde . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 8 Cryo-SEM of Chemically Fixed Animal Cells Stanley L. Erlandsen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 9 High-Resolution and Low-Voltage SEM of Plant Cells Guy Cox, Peter Vesk, Teresa Dibbayawan, Tobias I. Baskin, and Maret Vesk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229 vii
viii
Contents
10 High-Resolution Cryoscanning Electron Microscopy of Biological Samples Paul Walther . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 11 Developments in Instrumentation for Microanalysis in Low-Voltage Scanning Electron Microscopy Dale E. Newbury . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
List of Contributors
Ralph Albrecht Department of Animal Sciences Department of Pediatrics Division of Pharmaceutical Sciences University of Wisconsin-Madison Madison, Wisconsin, USA Tobias I. Baskin Biology Department University of Massachusetts Amherst, Massachusetts, USA Alan Boyde Biophysics OGD, QMUL Dental Institute London, UK Guy Cox Electron Microscope Unit University of Sydney NSW, Australia Teresa Dibbayawan Electron Microscope Unit University of Sydney NSW, Australia Stanley L. Erlandsen (Deceased) Department of Genetics Cell Biology and Development University of Minnesota Medical School Minneapolis, Minnesota, USA
David C. Joy Science and Engineering Research Facility University of Tennessee Knoxville, Tennessee, USA Dennis McMullan 59 Courtfield Gardens London SW5 0NF, UK Daryl Meyer Department of Animal Sciences University of Wisconsin-Madison Madison, Wisconsin, USA Dale E. Newbury Surface and Microanalysis Science Division National Institute of Standards and Technology Gaithersburg, Maryland, USA James B. Pawley Department of Zoology University of Wisconsin Madison, Wisconsin, USA Heide Schatten Department of Veterinary Pathobiology University of Missouri-Columbia Columbia, Missouri, USA Peter Vesk Lecturer in the School of Botany University of Melbourne Melbourne, AU ix
x
Maret Vesk Electron Microscope Unit University of Sydney NSW, Australia
List of Contributors
Paul Walther Central Electron Microscopy Unit University of Ulm Ulm, Germany
Chapter 1
The Early Development of the Scanning Electron Microscope Dennis McMullan
It has been forty years since the scanning electron microscope (SEM) became a significant instrument in the scientific community. In 1965, the Cambridge Instrument Company in the United Kingdom marketed their Stereoscan 1 SEM, which was followed about 6 months later by JEOL of Japan with the JSM-1. Before 1965, there were about thirty years of intermittent SEM development in Germany, the United States, England, and Japan, although Japanese development was apparently not covered in the published literature. Development began in the 1930s in Germany, and then began towards the end of that decade in the United States. During these early years, there were two different approaches: the first, which had some specific relevance to the low voltage scanning transmission microscope (LVSEM); and the later one that was linked to the transmission electron microscope (TEM) and led to the scanning transmission electron microscope (STEM) and the future form of the current SEM. But first, we must share a few words about the early history of scanning and its application in microscopy.
Invention of Scanning In the 1840s, Alexander Bain, a Scottish clockmaker, invented the principle of dissecting an image by scanning, and he was granted a British patent (Bain 1843) for the first fax machine (McMullan 1990). At the transmitter, a stylus mounted on a pendulum contacts the surface of metal type forming the message, thus closing an electrical circuit. At the receiver, a similar stylus, also on a pendulum, records electrochemically on dampened paper. Following each swing of the pendulums, the type and the recording paper are lowered by one line. The means for starting the pendulums swinging simultaneously and synchronising them magnetically are described in the patent.
Scanning Optical Microscopy The first proposal in print for applying scanning to microscopy was made in Dublin by Edward Synge (1928). The proposal was for a scanning optical microscope, and his goal was to overcome the Abbe limit on resolution by H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
1
2
D. McMullan
what is now called “near-field microscopy”—that is, the production of a very small light probe by collimation through an aperture smaller than the wavelength of the light. Synge was a scientific dilettante who had original ideas in several scientific fields, but did not attempt to put them into practice (McMullan 1990). However, he considered some of the problems that would be encountered with a scanning microscope and he proposed the use of piezo-electric actuators (Synge 1932), which are now used with great success in the scanning tunneling microscope and other probe instruments—including, of course, the near-field optical microscope itself. He envisaged fast scanning of the sample so that a visible image could be displayed on a phosphor screen, and he also pointed out the possibility of contrast expansion to enhance the image from a low contrast sample—probably the first mention of image processing by electronic means (as distinct from photographic).
Charged Particle Beams A proposal for using an electron beam in a scanning instrument was described in German patents by Hugo Stintzing of Giessen University (1929). These patents were concerned with the automatic detection, sizing, and counting of particles using a light beam, or—for those of below-light microscopic size—a beam of electrons. The focusing of electrons was (at that time) unknown to him, as to most others, and he proposed obtaining a small diameter probe using crossed slits. The samples would be either mechanically scanned in the case of a light beam, or one could use electric or magnetic fields to deflect an electron beam. Suitable detectors would be used to detect the transmitted beam that was attenuated by absorption or scattering. The output was to be recorded on a chart recorder so that the linear dimension of a particle could be given by the width of a deflection, and the thickness by the amplitude—the production of a two-dimensional image was not suggested. Stintzing did not apparently attempt the construction of this instrument and there are no drawings accompanying the patent specification. Thomas Mulvey (1962), however, published a block-schematic diagram of Stintzing’s proposal much later.
The Transmission Electron Microscope In the early 1930s, the main center for the development of the electron microscope was the Berlin Technische Hochschule in the laboratory of Professor Matthias where Max Knoll was a research assistant supervising students—including Ernst Ruska, whose subject area was electron optics. Arising from this work, Knoll and Ruska demonstrated the first transmission electron microscope with a magnification of x16. From the very beginning of electron microscopy, the imaging of solid samples was an important goal, particularly as the methods for producing thin samples were not developed until later. The first attempt was by Ruska (1933), with the sample
1 The Early Development of the Scanning Electron Microscope
3
Fig. 1.1 Early TEM image of an oxide replica of etched aluminium (Mahl 1940)
surface normal to the viewing direction and illumination by an electron beam at grazing incidence to the surface. Ruska obtained images of copper and gold surfaces but at a magnification of only x10. A few years later, he made a second attempt (Ruska & Müller 1940) with the same geometry and with only marginally better results. von Borries (1940) was much more successful with his grazing incidence method in the transmission electron microscope (TEM), where the sample surface is at a few degrees both to the viewing direction and to the illuminating beam. A breakthrough in the microscopic imaging of surface topography in the TEM was the 1940s introduction of replicas by H. Mahl (1940), and these set the standard for the next 25 years—although they were tedious to make and could be subject to serious artifacts. An early example is shown in Fig. 1.1.
Electron Beam Scanner Knoll, the co-inventor of the TEM with Ruska, was the first to publish images from solid samples obtained by scanning an electron beam (Knoll 1935). In 1932, very soon after the building of the first TEM at the Berlin Technische Hochschule, he joined the Telefunken Company as the director of research to develop television (TV) camera tubes. There, he designed an electron beam scanner for studying the targets of these tubes. A schematic block diagram is shown in Fig. 1.2 (the sample was mounted at one end of a sealed-off glass tube, and an electron gun was located at the other). The accelerating potential was in the range of 500–4,000 V, and the beam was focused on the surface of the sample and scanned by deflection coils in a raster of 200 lines and 50 frames/s. The current collected by the sample (the difference of the incident and secondary emitted currents) was amplified by a thermionic tube
4
D. McMullan
Fig. 1.2 Schematic diagram of Knoll’s (1935) electron beam scanner; (labels translated by T. Mulvey)
amplifier and intensity-modulated cathode-ray tube that was scanned by deflection coils connected in a series with those on the electron-beam scanner. By changing the ratio of the scan amplitudes, the magnification could be varied. Knoll used mainly unity magnification, but he could increase it to about x10 before the resolution was limited by the diameter of the scanning probe. This apparatus had virtually all the features of an SEM, but Knoll surprisingly (in view of his earlier work on the TEM) did not use additional electron lenses to reduce the size of the probe below 100 µm. The resolution he obtained, however, was entirely adequate for his purpose. The beam current was relatively high—on the order of microamps—and therefore thermionic tubes could be used to amplify the signal current in spite of the fast scan rate. He must have realized that reducing the size of the probe would be counter-productive because there were no suitable high-gain electron detectors in existence at that time. Similar images were produced by others working on the development of TV cameras in the 1930s (e.g., von Ardenne 1985), but Knoll was the only one at the time who looked at samples other than camera tube targets (e.g. silicon iron in Fig. 1.3), and he also elucidated the contrast mechanisms of secondary electron coefficients and topography. The images were true secondary electron images because the electron gun and sample were enclosed in the highly evacuated and baked glass envelope, and there was therefore little or no contamination of the surface. It is only comparatively recently that ultra-high vacuum (UHV) SEMs have been available that can work in this imaging regime. Knoll continued using his electron beam scanner (which he named der Elektronenabtaster) for a number of purposes, including the study of oxide layers on metals (Knoll 1941).
1 The Early Development of the Scanning Electron Microscope
5
Fig. 1.3 Electron-beam scanner image of a silicon iron sheet showing electron channeling contrast; horizontal field width = 50 mm (Knoll 1935)
A few years later, Manfred von Ardenne, also in Berlin, built a very different instrument that was in fact a scanning transmission electron microscope (STEM) that he intended to also use for solid samples. His hope was not realized at the time, but his work established many of the principles used in all future SEMs. This is described further in this chapter, but first we will consider Vladimir Zworykin’s RCA microscope—which actually came after von Ardenne’s but was related much more closely to Knoll’s pioneering work—and also to some aspects of later LVSEMs.
The RCA Scanning Electron Microscope Zworykin, who was director of research at the RCA Laboratories in Camden, N.J., initiated a development program for SEM (Zworykin et al. 1942) in 1938 that continued until about 1942. He was a pioneer of electron-scanned TV camera tubes
6
D. McMullan
dating back to the 1920s, and had also developed the first optical microscope with video output (Zworykin 1934). The development of the SEM was done in parallel with that of a TEM, and by the same staff—in particular, J. Hillier, E.G. Ramberg, A.W.Vance and R.L. Snyder as well as Zworykin himself. Although Zworykin had every microscope paper from Germany translated as soon as it was received (Reisner 1989), he was apparently not influenced by von Ardenne’s work on the STEM/SEM. Instead, he started by repeating Knoll’s beam scanner experiments (in effect) using a monoscope. The monoscope was a pattern-generating, cathode-ray tube that had been invented in Knoll’s department at Telefunken (1935) and further developed by RCA for television use (Burnett 1938). His team then built an SEM based on the monoscope, but with two magnetic lenses to produce a very small focused probe, and a demountable vacuum system so that the sample could be changed (Zworykin et al. 1942). The scan rate was the US TV standard—441 lines and 30 frames/second—and the signal was amplified by a thermionic tube video amplifier. For a signal-to-noise ratio of 10, the signal current had to be 3 × 10−8 A, which could only be reached if the probe diameter was about 1 µm. He then tried to obtain a high current in a smaller probe by using a field emission gun with a single-crystal tungsten point, presumably based on experience with the point projection microscope that had been built in the RCA Laboratories by G.A. Morton and E.G. Ramberg (1939). To reach a sufficiently high vacuum, Zworykin had to return to having the gun and the sample in a glass envelope that had been baked and sealed off. A single magnetic lens was used, and fleeting images were obtained at x8,000 magnification, with scanning at TV rate and a thermionic tube amplifier. Stable images could no doubt have been achieved, but at that time a practical microscope would not have resulted because demountable UHV techniques did not then exist. To overcome the noise problem, Zworykin therefore decided to build an SEM with an efficient electron detector and a slower scan. The detector was the combination of phosphor and photomultiplier that T.E. Everhart and R.F.M. Thornley (1960) used nearly twenty years later in an improved form. To bring the secondary electrons to it, he designed an electrostatic immersion lens that retarded the beam electrons and accelerated the secondaries. Figure 1.4 shows the final electron optical arrangement. Electrostatic lenses were used to produce a demagnified image of the source on the sample that was held at +800 V relative to the grounded gun cathode. The electron beam leaving the gun was accelerated to 10 keV in the intervening electron optics. The secondary electrons returning from the sample were similarly accelerated, and diverged as they passed through the 4th electrostatic lens and hit the phosphor screen with an energy of 9.2 keV. In the first instrument, the scanning was done by electro-mechanically moving the sample relative to the beam using loudspeaker voice-coils and (later) hydraulic actuators—it was only in the final version that magnetic scanning of the beam was employed. The scan time was fixed at 10 min by the facsimile recorder that was used for image recording, and also controlled the microscope scans. There was no provision for a faster scan or the production of a visible image on a TV monitor—this seems strange remembering Zworykin’s TV background, but it may have been because the signal bandwidth was seriously limited by the decay time of
1 The Early Development of the Scanning Electron Microscope
7
Fig. 1.4 The electron optics of the SEM built by Zworykin et al. (1942)
the phosphor, which was a problem important in later work at Cambridge (McMullan 1952). The optimum focus setting was found by maximizing the high frequency components in the video waveform observed on an oscilloscope, a method that was originally proposed by von Ardenne (1938b). Although the intention was to produce contrast by differences in the secondary emission ratio of the surface constituents, and the incident beam energy of 800 eV was chosen with this in mind, contamination of the surface in the rather poor vacuum prevented meaningful compositional contrast from being obtained. Surprisingly, Zworykin did not anticipate this, although he was an experienced vacuum physicist. Only two years earlier, secondary emission measurements and the effects of contamination had been published by Bruining and de Boer (1938). Actually, all of Zworykin’s published micrographs were of etched or abraded samples, and contrast was topographic (Zworykin et al. 1942)—for example, etched brass (see Fig. 1.5). The quality of the recorded images was rather disappointing, and together with the lack of a visible image, must have been a factor in RCA’s decision to discontinue the project. The main reason, however, was undoubtedly the excellent results that were, as mentioned earlier, being obtained with replicas viewed in the
8
D. McMullan
Fig. 1.5 Micrograph of etched brass produced by the SEM of Zworykin et al. (1942)
TEM (Mahl 1940). In any event, all available technical effort had to be directed to the highly successful RCA EMB TEM, which was just then coming into production (Reisner 1989).
Von Ardenne’s Scanning Electron Microscope While RCA was developing its SEM, Manfred von Ardenne (a private consultant who had his own laboratory) was developing the first scanning electron microscope with a submicron probe. In 1936, he was contracted by Siemens & Halske AG to investigate the possibility of using a scanned electron probe to avoid the effects of objective lens chromatic aberration with thick samples in TEM. In the course of this work, he laid the foundations of electron probe microscopy by making and publishing (von Ardenne 1938a,b) a detailed analysis of the design and performance of probe-forming electron optics using magnetic lenses. The analysis covered the limitations on probe diameter due to lens aberrations and the calculation of the current in the probe. He also showed how detectors should be placed for bright-field
1 The Early Development of the Scanning Electron Microscope
9
and dark-field STEM and for imaging a solid sample in a SEM, and considered the effects of beam and amplifier noise on imaging. To fulfill the Siemens contract, von Ardenne built the first scanning transmission electron microscope (STEM) and demonstrated the formation of probes down to 4 nm in diameter. But in the short time available, he was limited to employing existing technology, and because there was no suitable low-noise electronic detector, he used photographic film—consequently there was no immediately visible image. A schematic of the microscope column is shown in Fig. 1.6. A demagnified image of the crossover of the electron gun was focused on the sample with two magnetic lenses, and X-Y deflection coils were mounted just above the second of these. Immediately below the sample, was a drum around which was wrapped the photographic film. The image was recorded by rotating the drum and simultaneously moving it laterally by means of a screw while the currents in the deflection coils were controlled by potentiometers mechanically coupled to the drum mechanism. The intensity of the beam was very low (about 10−13 A) and it was necessary to record the image over a period of about 20 minutes. Because the image was not visible until the film had been developed, focusing could only be accomplished indirectly by using the stationary probe to produce a shadow image of a small area of the sample on a single-crystal Zinc sulfide screen that was observed through an optical microscope and prism system. The recordings were inferior to those from the TEM that was being constructed by Ruska and von Borries at Siemens, and the hoped-for advantages of STEM with thick samples were not realized. Von Ardenne spent a short time trying to use the instrument in the SEM mode on bulk samples, but could only obtain low resolution images because of the detector problem. The sample current was amplified by thermionic tubes and a large probe current was needed. He did not publish any images. In total, von Ardenne worked for less than two years on scanning electron microscopy before concentrating on the development of his universal TEM (von Ardenne 1985). Then, with the start of World War II, he began work on a cyclotron and isotope separators for nuclear energy projects. If he had been able to continue, there is little doubt that he would have built an efficient SEM within a year or two: this is evidenced by a patent (von Ardenne 1937) that included a proposal for double-deflection scanning, two papers (von Ardenne 1938a,b), and a book (von Ardenne 1940). Two of the chapters in the book were on scanning microscopy and were based on the 1938 papers, but included additional material relating to imaging the surfaces of solid samples. Most importantly, he proposed a detector using an electron multiplier with beryllium copper dynodes (see Fig. 1.7) that could be opened to the atmosphere and worked with efficiently under poor vacuum conditions. Measurements of the secondary emitting ratio of beryllium copper and its stability when exposed to the atmosphere were otherwise only first reported in 1942 by I. Matthes (1942) of the AEG Research Institute in Berlin, but von Ardenne was probably aware of this research a year or two before. In his book, von Ardenne also discussed the interaction between the beam electrons and the sample, and suggested that back-scattering would cause a loss of resolution, illustrating this with a diagram that has quite a modern look
10
Fig. 1.6 Cross-section of the column of von Ardenne’s (1938b) STEM
D. McMullan
1 The Early Development of the Scanning Electron Microscope
11
Fig. 1.7 Electron multiplier with beryllium copper dynodes proposed by von Ardenne (1940) as a secondary electron detector for SEM. The drawing shows the first three stages of the multiplier and its position relative to the objective lens and sample
(see Fig. 1.8). He argued that the incident beam electrons produce secondary electrons at or near the surface from an area approximately equal to the beam diameter, and give a high-resolution image (nutzbare Strahlung). The beam electrons penetrate the sample, and a proportion of them is backscattered and reach the surface where they produce further secondaries. These two signals are now generally referred to as SE-I and SE-II, respectively (Drescher et al. 1970, Peters 1982). The backscattered electrons are emitted from an area of diameter comparable to the penetration depth, and the secondaries they produce (schädliche Strahlung) may impair the resolution (he did not, however, consider the case of a sample with small inclusions below the surface). He concluded that good resolution might be obtained either with a very low-energy beam (1 keV), or with one having a high energy (50 keV). In the first case, the backscattered electrons would emerge from an area of the surface a little larger than the incident beam, and the resolution would be unaffected. On the other hand, the secondary electrons produced by the backscatter of a 50-keV beam would affect a much larger area, and would be evenly distributed so that their main effect would be to increase the background (reduce the contrast) rather than affect the resolution. Von Ardenne’s scanning microscope was destroyed in an air raid on Berlin in 1944, and after the war he did not resume his work in electron microscopy but researched in other fields—first in Russia and then in Dresden (in 1955), which was then in East Germany. Additional information about von Ardenne’s scientific work is available in his autobiography (von Ardenne 1972) and by McMullan (1988).
12
D. McMullan
Fig. 1.8 Diagram illustrating von Ardenne’s (1940) discussion of secondary electron imaging of a surface
The Cambridge Scanning Electron Microscopes Apart from a theoretical analysis of resolving power by a French author (Brachet 1946), no other substantial work on SEMs was reported until 1948, although recent research has revealed that some rather primitive experiments were done by A. Léauté (1946) at L’Ecole Polytechnique in Paris during World War II (Hawkes & McMullan 2004). In the 1940s, and for many years after, the feeling among most electron microscopists was that the SEM was not worth further consideration in view of the apparent failure at RCA—if such an experienced team could be that unsuccessful, it seemed very unlikely that anyone else could produce an effective instrument. A notable exception to this general opinion was that of Denis Gabor (1945). It was then that Charles Oatley at the Department of Engineering of the University of Cambridge decided to take another look at the SEM, although he related that “several experts expressed the view that this [the construction of an SEM] would be a complete waste of time” (Oatley et al. 1985). He explained at some length the reasons that brought him to this decision, but the main technological justifications were that “Zworykin and his collaborators had shown that the scanning principle
1 The Early Development of the Scanning Electron Microscope
13
was basically sound and could give useful resolution in the examination of solid surfaces” and “improvements in electronic techniques and components had resulted from work during the war” (Oatley, 1982). He also felt that the RCA detector had a low efficiency and only a small proportion of the secondaries were reaching it, with the result that the images were noisy in spite of the long recording time. Independently of von Ardenne, he proposed to use an electron multiplier with beryllium-copper electrodes (Allen 1947), having been promised one by A.S. Baxter at the Cavendish Laboratory, who was making multipliers of this type (Baxter 1949). The full story of Oatley’s achievement is presented in a recent publication (Breton et al. 2004). I was selected by Oatley to build an SEM as a Ph.D. project—it was a challenging task because electron microscopy was a completely new subject for everyone in the laboratory, although I had had some experience in the radar and television industries, including the development and manufacture of cathode-ray tubes. I first completed a 40 keV electrostatically focused TEM that had been started by another PhD student, K.F. Sander. He abandoned it at an early stage and changed the subject of his research project to electron trajectory plotting (Sander 1951). I converted it to a STEM, and then to an SEM, by the addition of scan coils, an electron multiplier detector, and a long persistence cathode-ray tube monitor (McMullan 1952). It was not apparent how Zworykin’s results might be improved upon. A higherincident beam energy was expected to be beneficial, but it was not clear how image contrast would be formed. As mentioned earlier, Bruining and de Boer (1938) had shown that the secondary emission from a surface is critically dependent on the vacuum conditions, and it was plain that the achievable vacuum would not be good enough for there to be meaningful secondary-electron compositional contrast from a polished sample. Images of surfaces were obtained at grazing incidence and viewing direction (2 deg) in the TEM by Bodo von Borries (1940) and others, and it seemed probable that similar images could be produced in the SEM. I therefore mounted a sample of etched aluminium at a rather larger angle (30 deg; because the backscattered electrons did not have to be focused) and was rewarded by the now commonplace threedimensional appearance that is the hallmark of SEM images and a consequence of their large depth of focus. One of the first images—of etched aluminium—is shown in Fig. 1.9: (a) the direct view image (about 0.9-sec frame period), and (b) a 5-min recording. The beam energy was 16 keV and the resolution about 50 nm, limited by astigmatism in the objective lens and insufficient magnetic shielding. A block diagram of the SEM is shown in Fig. 1.10 (McMullan 1953). There was a relatively fast-scan, long-persistence cathode-ray tube display (405 lines, 1.8 fields/sec interlaced and a 5-min frame scan for photographic recording). Other features included a nonlinear amplifier for gamma control, and beam blanking for DC restoration. Double-deflection scanning coils were added later. The most important differences between this instrument and Zworykin’s were the much higher incident beam energy (15-20 keV), and the contrast produced mainly by scattered electrons from the tilted sample. The mechanism of contrast formation was investigated and shown to be topographic. No attempt was made to collect
14
D. McMullan
Fig. 1.9 One of the early images (etched aluminium surface) produced with SEM1. Angle of incidence of 16 keV electrons 25◦ : (a) visible image, 0.95 frames/s; beam current 1.5 × 10−10 A. (b) 5 min recording; 10−13 A. (McMullan 1952, 1953)
Fig. 1.10 Block schematic of SEM1. (McMullan 1952, 1953)
1 The Early Development of the Scanning Electron Microscope
15
Fig. 1.11 Photograph of SEM1 taken in 1953 when K.C.A Smith took over
low-energy secondaries. In fact, I thought that they would be detrimental because of the inevitable contamination on the surface of the sample. I overlooked the increase in signal that is obtained from the low energy secondaries. I realized that there was another advantage in using a high-energy scanning beam: this was that in principle atomic number contrast was possible using backscattered electrons. An experimental curve of emission ratio (for 20 keV primaries) versus atomic number had recently been published by Palluel (1947), but an attempt at obtaining atomic number contrast failed. Some years later, Oliver Wells (1957) was more successful. The obvious disadvantage of high-beam energies was that the resolution was limited by penetration of the primary electrons. I suggested low-loss electron imaging to minimize this, but was not able to implement it (this was also done many years later by Wells, who published in 1971). One other contrast mechanism that I tried was cathodoluminescence, and I was able to demonstrate that phosphors with too long a decay constant to be used for producing images at a 0.9-sec frame time with a Zworykin-type detector were completely satisfactory when excited at a high-current density by a focused probe (Smith & Oatley 1955). Figure 1.11 is a photograph of the microscope (now named SEM1) taken in 1953 shortly before K.C.A. Smith assumed responsibility for it and turned it into an SEM that could produce images comparable with some of those from modern microscopes.
Further Development of the Microscope Smith introduced many improvements to SEM1, including a stigmator and a tilting sample stage, and he increased the efficiency of the detection system by moving the electron multiplier nearer the sample so that low-energy secondary electrons were
16
D. McMullan
collected, thus increasing the signal current. He showed that metalized insulating samples could be imaged, and he examined a wide variety of samples including germanium point-contact rectifiers and biological specimens. He also built an environmental cell for wet specimens (anticipating the environmental scanning electron microscope, ESEM); this had thin windows to admit the focused beam and allow the scattered electrons to reach the multiplier. Although the results were rather disappointing, it led Oatley to suggest replacing the second window with a shortdecay-time plastic scintillator and photomultiplier, and dispensing with the bulky electron multiplier (Smith 1956). This, in turn, resulted in the development of the Everhart and Thornley (1960) detector. Further SEMs were built in the engineering department at Cambridge: SEM2 (Wells 1957); SEM3 (Smith 1960); SEM4 (Stewart 1962); and SEM5 (Pease & Nixon 1965). All were used on a wide variety of samples and for the development of new techniques. Other important instrumental advances made by Oatley’s group during the remainder of the 1950s through to 1965 included: atomic number contrast (Wells 1957); stereomicroscopy (Wells 1960); voltage contrast (Oatley & Everhart 1957); low-voltage (1–2 keV) SEM (Thornley 1960a); high temperature (1200 ◦ C) imaging of thermionic cathodes in a SEM (Ahmed 1962); high-resolution (10 nm) SEM (Pease & Nixon 1965); etching of surfaces in a SEM by ion bombardment (Stewart 1962); ion etching and microfabrication in SEM (Broers 1965); and microelectronics in SEM (Chang 1966). Most of this work is described in papers by Oatley (1982) and Oatley et al. (1985).
Materials Research In 1955, Smith used SEM1 for the first three applications of a scanning electron microscope in materials research: 1. Smith was visited by F.P. Bowden, head of the Surface Physics Laboratory in the Department of Physical Chemistry at Cambridge University, and J. McAuslen of Imperial Chemical Industries, who brought a sample of silver azide crystals. The thermal decomposition of these was being investigated in their laboratory using a TEM in the reflection mode, but had failed due to the premature ignition of the crystals under the intense illumination that was needed. In the SEM, with the crystals mounted on a small hot-plate, the decomposition could be readily controlled and observed without difficulty (Bowden & McAuslan 1956; McAuslen & Smith 1956; see Fig. 1.12. 2. J.H. Mitchell, controller of research at Ericsson Telephones, Ltd. approached Oatley about their work on the etching of germanium surfaces and the emergence of edge dislocations. They wished to establish whether there were pits or raised areas of sublight-microscopic size. J.W. Allen brought specimens to Cambridge and the micrograph in Fig. 1.13 shows a feature produced by the etchant CP4 (a mixture of acids with a small amount of bromine). (Allen & Smith 1956). 3. The third event occurred when Dr D. Atack—then on sabbatical leave from the Pulp and Paper Research Institute of Canada (PPRIC) where he was director of
1 The Early Development of the Scanning Electron Microscope
17
Fig. 1.12 Partial decomposition of a small needle of silver azide. (Smith 1956)
Fig. 1.13 Crystallographic feature on a germanium surface etched with CP4. (Allen & Smith, 1956)
18
D. McMullan
Fig. 1.14 Application of the SEM to the study of wood fibers: a surface of newsprint. (Smith 1956)
the Applied Physics Division—came and asked if he could try some of his pulp and paper samples in the SEM. In this case, a TEM had also been tried with unsatisfactory results (Page 1958). The experiments were highly successful, as shown in Fig. 1.14 (Smith 1956) and led to PPRIC purchasing a SEM (see next section).
The First Low Voltage Scanning Electron Microscope (LVSEM) Imaging In the present context, the work of R.F.M. Thornley on LVSEM in Oatley’s laboratory (Thornley 1960a) needs to be described in more detail. He investigated some of its possibilities, and by meticulous testing and modification was able to greatly improve the performance of the instrument he had been allocated (SEM2), which had been built and used by Wells (1957). This included eliminating the many sources of interference (mainly 50 Hz) he had identified and making alterations to various components—some quite minor—so that eventually he was able to obtain a probe diameter of 200 nm at 1 keV. This is, of course, ridiculously large by today’s standards, but for then it was a considerable achievement. His modified SEM2 was the first LVSEM (Thornley 1960a) of the modern type (strictly speaking, Zworykin’s (1942) was the first, but the low voltage was irrelevant to the rather poor images it was able to produce).
1 The Early Development of the Scanning Electron Microscope
19
To quote from a paper Thornley presented at the European Regional Conference on Electron Microscopy in Delft in 1960 (Thornley 1960b): “Previous work at Cambridge has used electron accelerating voltages greater than 6 kV, limiting the use of the instrument to the examination of conducting surfaces in order to avoid the formation of charging artefacts. Insulators can be covered with a thin metallic film to overcome this restriction, but deformation while under observation is difficult because the conducting film cracks away from the substrate as shown by Wells (1957). If the beam voltage is reduced until the secondary emission coefficient of the specimen is equal to or greater than unity, the surface potential will be automatically stabilised at that of its surroundings, in this case, at earth potential. Under these conditions, the effective secondary emission coefficient of an insulator is unity, regardless of the angle of incidence of the primary beam and the picture contrast is controlled only by collector modulation. . . . .” “. . . . .[Fig. 1.15], of a fractured ceramic surface, shows that, with a suitable choice of collector position, contrast similar to that expected from oblique viewing at high voltage can be obtained at low voltages, in this case, 1.5 kV. Provided the stability requirements can be met, the ultimate resolution, for voltages above 500 V, is limited by the same factors as in the high voltage case to between 50 and 100 Å as shown by Everhart (1958). It has been found that the reduction in gun brightness at low voltages is largely compensated by the increase in secondary emission, so no change in recording time has been necessary. The micrograph shown was recorded over 2 min, using an instrument originally designed for 25 kV operation, but fitted with a modified gun, permitting operation down to 300 volts. Contrast due to surface films is enhanced at low voltages because differences in secondary emission coefficients are more pronounced and penetration effects are reduced, the range of a 500-V electron in aluminium being roughly 30Å. . . . . .”
Further work on ceramics was reported in a paper with L. Cartz of Morganite Research and Development Ltd. (Thornley & Cartz 1962):
Fig. 1.15 Imaging the surface of a sintering fault in an alumina ceramic with 1.5-keV electrons. (Thornley 1960)
20
D. McMullan “A direct electron-optical method of observing an insulator surface is described and applied to a series of alumina ceramics. The surface of the object requires no previous treatment of any kind, and a resolving power of 2,000 Å has been obtained with a depth of focus of about 50 µm. Different phases and components can be distinguished. Fractured surfaces, a fault region, and polished surfaces of various alumina ceramics are examined.”
Commercial Production of SEMs Following the encouraging results obtained using SEM1 to study wood fibers (Atack & Smith 1956), L.R. Thiesmeyer, director of the Canadian Pulp and Research Institute, ordered a fully engineered microscope (SEM3) from the Cambridge University Engineering Department. This was used for many years in their Montreal Laboratories and was the earliest industrial application of an SEM on a daily basis. Smith developed SEM3 and completed it in 1958: it was the first magnetically focused SEM (Smith 1959). The lower section of the column below the table consisted of a modified Metropolitan Vickers (later AEI)-type EM4 TEM (Page 1954) and contained the electron gun, condenser lens, transmission sample stage, objective lens, and double pole piece projector lens. For scanning operations, the transmission objective and projector were used together in various combinations and powers according to the spot diameter required to provide the first stage of spot demagnification. Immediately above the table, there was a section of the column containing the scanning coils and the objective lens that was of the pin-hole type (Liebmann 1955), with three adjustable apertures. There was a tilting sample stage and the Everhart-Thornley type of secondary electron detector. Thiesmeyer and Atack were among the very few who (at that time) saw the great potential of SEM. Although Oatley’s group had produced and published highquality micrographs from many different samples, there was still considerable resistance to SEM among microscopists. Over several years, Oatley expended much effort in trying to persuade electron microscope manufacturers to market an SEM (Jervis 1971, 1972; Oatley 1982; Breton et al. 2004) but he was only finally successful in 1962 when the Cambridge Instrument Company decided to go ahead with the production of an SEM based on the instruments developed by Oatley’s group. This decision was influenced by A.D.G. Stewart, one of Oatley’s students who agreed to join the company and later played a leading part in the development of the Stereoscan, as the new SEM was named (Stewart & Snelling 1965; Stewart 1985). The prototype Stereoscan (see Fig. 1.16) went to the Dupont Chemical Corporation in the United States in 1964, and in the following year the first two production models were sold to P.R. Thornton at the University of North Wales and to J. Sikorski at Leeds University in the United Kingdom, the third to G. Pfefferkorn at Münster University in Germany, and the fourth to the Central Electricity Laboratories in Leatherhead, United Kingdom. In the words of Professor Sir Charles Oatley (1982), “By this time the Company had launched a publicity campaign and orders began to roll in. An additional batch of twelve microscopes was put in hand; and then a further forty . . . . . . . the scanning microscope had come of age.”
1 The Early Development of the Scanning Electron Microscope
21
Fig. 1.16 The Cambridge Instrument Company’s Stereoscan Mk 1 prototype (Stewart & Snelling 1965)
The first commercial competitor was the Japanese firm, JEOL, who marketed their JSM-1 SEM about six months later and were soon followed by others (McMullan 2004).
Other SEMs up to 1965 and Beyond SEM developments in other laboratories prior to 1965, as evidenced in scientific publications, included the following: • An SEM was built in France by Bernard & Davoine (1957) at the National Institute of Applied Science in Lyon. It had a probe size of the order of 1 µm and was used over a period of years mainly for cathodoluminescence studies. • AEI in the United Kingdom—the major TEM manufacturer at that time— developed an SEM but did not proceed after the first instrument, sold in 1959, turned out to be unsatisfactory (Jervis 1971, 1972). • Wells et al. (1965) built an advanced SEM for semiconductor studies and microfabrication for the Westinghouse Laboratories in Pittsburgh, Pennsylvania, and demonstrated EBIC imaging.
22
D. McMullan
• In the USSR, there was a SEM at Moscow University as early as 1960 (Kushnir et al. 1961). • There were other groups, especially in Japan, who did not publish at the time (Fujita 1986). The year 1965 marked the beginning of the general use of scanning electron microscopy, but there were other developments coming to fruition at the time that had a great importance for low-voltage SEM. These were in ultra-high vacuum technology that enabled vacuum systems that were capable of reaching 10−8 Pa or less to be easily put together using standard components: in particular, the Conflat flange (Wheeler & Carlson 1962), all-metal vacuum valves (Wheeler 1976), and the development of the sputter-ion pump, starting with Hall (1958). Therefore, just about the time that the SEM was becoming generally used with the Stereoscan and the JEOL JSM-1, it was becoming possible to realize Zworykin’s (1942) initial concept of a SEM with a low-voltage scanning beam (800V), a cold field-emission gun cathode, secondary emission contrast from a polished sample, and a fast scan. It seems probable that the success of the Stereoscan-type SEMs had the initial effect that electron microscopists were so occupied in using the new instruments that there was no immediate interest in the possibilities of still more powerful techniques. Therefore, the application of UHV to microscopy was initially confined to the development of the STEM, particularly by Albert Crewe and his coworkers at the Argonne Laboratory (Crewe 1966; Crewe et al. 1968). The first UHV, low-voltage SEM with field-emission gun was described by Welter and Coates (1974), who had earlier collaborated with Crewe.
References Ahmed, H. (1962) “Studies on high current density thermionic cathodes”. PhD Dissertation, University of Cambridge. Allen, J.S. (1947) An improved electron multiplier particle counter. Rev. Sci. Instrum. 19, 739–749. Allen, J.W. and Smith, K.C.A. (1956) Electron microscopy of etched germanium surfaces. J. Electronics 1, 439–443. D. Atack and K. C. A. Smith, (1956) “The scanning electron microscope. A new tool in fiber technology,” Pulp Pap. Mag. Can. (Convention issue) 57, 245–251. Bain, A. (1843) Electric time-pieces and telegraphs. British Patent No 9745, filed 27 May 1843. Baxter, A.S. (1949) “Detection and analysis of low-energy disintegration particles.” Ph.D. Dissertation, University of Cambridge. Bernard, R. and Davoine, F. (1957) The scanning electron microscope (in French). Ann. Univ. Lyon Sci. Sect. B[3] 10, 78–86. Bowden, F.P. and McAuslan, J.H.L. (1956) Slow decomposition of explosive crystals. Nature 178, 408–410. Brachet, C. (1946) Note on the resolution of the scanning electron microscope (in French). Bull. Assoc. Tech. Marit. Aeronaut. 45, 369–378. Breton, B., McMullan, D. and Smith, K.C.A. (eds, 2004). “Sir Charles Oatley and the scanning electron microscope”, Adv. Imaging Electron Phys. 133 (P.W. Hawkes, editor-in-chief), Elsevier Academic Press: San Diego, London. Broers, A.N. (1965) “Selective ion beam etching in the scanning electron microscope”. PhD Dissertation, University of Cambridge.
1 The Early Development of the Scanning Electron Microscope
23
Bruining, H. and de Boer, J.H. (1938) Secondary emission. Physica (Amsterdam) 5, 17–30. Burnett, C.E., (1938) The Monoscope. RCA Review 2, 414–420. Chang, T.H.P. (1967) “Combined micro-miniature processing and microscopy using a scanning electron probe system”. PhD Dissertation, University of Cambridge. Crewe, A.V. (1966) Scanning electron microscopes: is high-resolution possible? Science 154, 729–738. Crewe, A.V., Eggenberger, D.N., Wall, J. and Welter, L.M. (1968) Electron gun using a fieldemission source. Rev. Sci. Instrum. 39, 576–583. Drescher, H., Reimer, L., and Seidel, H. (1970) Back-scattering coefficient and secondary electron yield from 10 - 100 keV electrons in the scanning electron microscope (in German). Z. angew. Phys. 29, 331. Everhart T.E. and Thornley R.F.M. (1960) Wide-band detector for micro-microampere low-energy electron currents. J. Sci. Inst. 37, 246–248 (1960). Fujita, H. (1986) “The History of Electron Microscopes”, 11th International Congress on Electron Microscopy: Kyoto, Japan. pp. 187–193. Gabor, D. (1945) “The Electron Microscope”. Hulton Press: London. Hall, L.D. (1958) Electronic ultra-high vacuum pump. Rev. Sci. Instrum. 29, 367–370. Hawkes, P.W. and McMullan, D. (2004) A forgotten French scanning electron microscope and a forgotten text on electron optics. Proc. Roy. Microsc. Soc. 39, 285–290. Jervis, P. (1971/72) Innovation in electron-optical instruments – two British case histories. Research Policy 1, 174–207. Knoll, M. (1935) Static potential and secondary emission of bodies under electron irradiation (in German). Z. tech. Phys. 16, 467–475. Knoll, M. (1941) Detection of attached oxide layers on iron with the scanning electron microscope (in German). Phys. Z. 42, 120–122. Kushnir,Yu.M., Fetisov, D.V. and Raspletin, K.K. (1961) Scanning electron microscope and X-ray microanalyser. Bull. Acad. Sci. USSR. Phys. Ser. (Engl. Transl) 25, 709–714. Léauté, L.[A.] (1946) Applications of the electron microscope in metallurgy (in French). In “L’Optique Electronique” (L. de Broglie ed.) Editions de la Revue d’Optique Théorique et Expérimental: Paris, 1946, 209–220. Liebmann, G. (1955) The magnetic pinhole electron lens. Proc. Phys. Soc. Ser. B 68, 682–685. Mahl, H. (1940) Supermicroscopic determination of the orientation of single aluminium crystals (in German). Metallwirtschaft 19, 1082–1085. Matthes, I. (1942) Investigation of the secondary electron emission from various alloys (in German).. Z. tech. Phys. 22, 232–236. McAuslan, J.H.L. and Smith, K.C.A. (1956) The direct observation in the scanning electron microscope of reactions, in Electron Microscopy: Proceedings of the Stockholm Conference, Sept. 1956, edited by F. S. Sjostrand and J. Rhodin (Academic, New York 1957), pp. 343–345. McMullan, D.(1952) “Investigations relating to the design of electron microscopes”. Ph.D. Dissertation, Cambridge University. McMullan, D. (1953) An improved scanning electron microscope for opaque specimens. Proc. Inst. Electr. Engrs. 100, Part II, 245–259. McMullan, D. (1988) Von Ardenne and the scanning electron microscope. Proc. Roy. Microsc. Soc. 23, 283–288. McMullan, D. (1990) The prehistory of scanned image microscopy, Part 1: scanned optical microscopes. Proc. Roy. Microsc. Soc. 25, 127–131. McMullan, D. (1995) Scanning electron microscopy 1928 – 1965. Scanning 17, 175–185. McMullan, D. (2004) A history of the scanning electron microscope, 1928 – 1965. Adv. Imaging Electron Phys. 133, 523–545. Morton, G.A. and Ramberg, E.G. (1939) Point projector electron microscope. Phys. Rev. 56, 705. Mulvey, T. (1962) Origins and historical development of the electron microscope. Brit. J. Appl. Phy. 13, 197–207.
24
D. McMullan
Oatley, C.W. (1982) The early history of the scanning electron microscope. J. Appl. Phys. 53, R1-R13. Oatley, C.W. and Everhart, T.E. (1957) The examination of p-n junctions with the scanning electron microscope. J. Electron. 2, 568–570. Oatley, C.W., McMullan, D., and Smith, K.C.A. (1985) The development of the scanning electron microscope. in “The Beginnings of Electron Microscopy” (P.W. Hawkes ed.) Adv. Electronics Electron Phys. Suppl. 16, 443–482. Page, D.H. (1958) Reflexion electron microscopy at high angles. Brit. J. Appl. Phys. 9, 60–67 Page, R.S. (1954) A compact console-type electron microscope. J. Sci. Instrum. 31, 200–205. Palluel, P. (1947) Backscattered components of electron secondary emission from metals (in French) C.R. Acad. Sci. 224, 1492–1494. Pease, R. F. W. and Nixon W.C. (1965) High-resolution scanning electron microscopy. J . Sci. Instrum. 42, 31–35. Peters, K-R. (1982) Generation, collection and properties of an SE-1 enriched signal suitable for high-resolution SEM on bulk specimens, in “Electron Beam Interactions with Solids”, SEM (Inc), Chicago, pp 363–372. Reisner, J.H. (1989) An early history of the electron microscope in the United States. Adv. Electronics Electron Phys. 73, 134–231. Ruska. E. (1933) The electron microscopic imaging of surfaces irradiated with electrons (in German). Z . Phys. 83, 492–497. Ruska E, and Müller, H.O. (1940) Progress on the imaging of electron irradiated surfaces (in German). Z Phys 116, 366–369. Sander, K.F. (1951) “An automatic electron trajectory tracer and contributions to the design of an electrostatic electron microscope”. PhD Dissertation, University of Cambridge. Smith, K. C. A. (1956) “The scanning electron microscope and its fields of application”. Ph.D. Dissertation, University of Cambridge. Smith, K. C. A. (1959) Scanning electron microscopy in pulp and paper research. Pulp Pap. Mag. Can. 60, T366-T371. Smith, K. C. A. (1960) A versatile scanning electron microscope, in The Proceedings of the European Regional Conference in Electron Microscopy, Delft, 29 August–3 September 1960 (Houwink, A.I. and Spit, B.J. eds.; Nederlandse Vereniging voor Elektronenmicroscopie, Delft n.d.) pp. 177–180. Smith, K. C. A. and Oatley, C. W. (1955) The scanning electron microscope and its fields of application. Br. J. Appl. Phys. 6, 391–399. Stewart, A.D.G. (1962) Investigation of the topography of ion bombarded surfaces with a scanning electron microscope, in Electron Microscopy, Fifth International Congress for Electron Microscopy, Philadelphia, Pennsylvania, 29 August – 5 September, 1962 (Breeze, S.S., ed.; Academic Press, New York, 1962) pp. D12-D13. Stewart, A.D.G. (1985) The origins and development of scanning electron microscopy. J. Microsc. 139, 121–127. Stewart, A.D.G. and Snelling, M.A. (1965) A new scanning electron microscope, in Electron Microscopy 1964, Proceedings of the Third European Regional Conference, Prague, 26 August – 3 September 1964 (Titlbach. M. ed.: Publishing House of the Czechoslovak Academy of Sciences: Prague) pp. 55–56. Stintzing, H. (1929) Method and device for automatically assessing, measuring and counting particles of any type, shape and size (in German). German Patents Nos 485155–6. Synge, E.H. (1928) A suggested method for extending microscopic resolution into the ultramicroscopic region. Phil. Mag. 6, 356–362. Synge, E.H.(1932) An application of piezo-electricity to microscopy. Phil. Mag. 13, 297–300. Telefunken A G (1935) Improvements in or relating to cathode-ray tube picture transmitters. British Patent No. 465715 (Convention Date (Germany) Oct. 3 1935). Thornley, R.F.M. (1960) “New applications of the electron microscope”. PhD Dissertaion, Universty of Cambridge. Thornley, R.F.M. (1960) Recent developments in scanning electron microscopy, in The Proceedings of the European Regional Conference in Electron Microscopy, Delft, 29 August–3
1 The Early Development of the Scanning Electron Microscope
25
September 1960 (Houwink, A.I. and Spit, B.J. eds.; Nederlandse Vereniging voor Elektronenmicroscopie, Delft n.d.) pp. 173–176. Thornley R.F.M. and Cartz L. (1962) Direct examination of ceramic surfaces with the scanning electron microscope. J. Am. Ceram. Soc. 45, 425–428. von Ardenne, M.(1937a) Improvements in electron microscopes. British Patent No 511204, convention date (Germany) 18 Feb. von Ardenne, M. (1938a) The scanning electron microscope. Theoretical fundamentals (in German). Z. Phys. 109, 553–572. von Ardenne, M. (1938b) The scanning electron microscope. Practical construction (in German). Z . tech. Phys. 19, 407–416. von Ardenne, M. (1940) “Electron Microscopy” (in German). Springer Verlag: Berlin . von Ardenne, M. (1972) “A Happy Life in Engineering and Research” (in German) Kinder Verlag: Munich and Zurich. von Ardenne, M. (1985) On the history of scanning electron microscopy, the electron microprobe, and early contributions to transmission electron microscopy. in “The Beginnings of Electron Microscopy” (PW Hawkes ed), Adv. Electronics Electron Phys. Suppl. 16, 1–21. von Borries, B. (1940) High-resolution images from the electron microscope used in reflection (in German). Z. Phys. 116, 370–378. Wells, O.C. (1957). “The construction of a scanning electron microscope and its application to the study of fibres”. Ph.D. Dissertation, Cambridge University. Wells O.C. (1960) Correction of errors in stereomicroscopy. Br. J. Appl. Phys. 11, 199–201. Wells, O. C. (1971) Low-loss image for surface scanning electron microscope. Appl. Phys. Lett. 19, 232–235. Wells, O.C., Everhart, T.E., and Matta, R.K.(1965) Automatic positioning of device electrodes using the scanning electron microscope. IEEE. Trans. Electron. Dev. ED-12, 556–563. Welter, L.M. and Coates, V.J. (1974) High-resolution scanning electron microscopy at low accelerating voltages. Proc. 7th Ann. SEM Symposium, IIT Research Institute, Chicago. (O. Johari ed.) pp. 59–66. Wheeler, W.R. (1976) Recent developments in metal-sealed gate valves. J. Vac. Sci. Technol. 13, 503–506. Wheeler, W.R. and Carlson, M.A. (1962) Ultra-high vacuum flanges. Trans. AVS Nat. Vac. Symp. 1961, p. 1309–1318, Pergamon Press: Oxford Zworykin, V.A. (1934) Electric Microscope. 1st Congresso Internazionale di Electroradiobiologia 1, pp 672–686. Zworykin, V.A., Hillier, J., and Snyder, R.L. (1942) A scanning electron microscope. ASTM. Bull. 117, 15–23; (Abstract) Proc. Inst. Radio Engrs. 30, 255. Zworykin, V.A., Morton, G.A., Ramberg, E.G., Hillier, J. and Vance, A.W. (1945) “Electron Optics and the Electron Microscope”, Wiley, New York.
Chapter 2
LVSEM for Biology James B. Pawley
Key words: low voltage, high-resolution, scanning electron microscopy, radiation damage
Introduction1 Two Approaches to Microscopical Imaging Early methods of microscopical imaging involved the use of lenses to focus and magnify the pattern of light transmitted, refracted, or reflected by the specimen. Contrast in the final image depended on the extent to which the features of the specimen absorbed, refracted, or reflected the light. Most early methods of transmission electron microscopy (TEM) also followed this approach in that the pattern of transmitted electrons emerging from the far side of the specimen was focused by appropriate lenses to form the final image. In 1935, however, Irwin Knoll pioneered a new approach where the properties of the specimen were not imaged directly in space but were instead sampled in time by a small beam of electrons that sequentially illuminated one point on the object at a time. The final image was built up from a time-sequence of data, and was displayed by a second electron beam on a cathode ray tube (CRT). The two beams swept in synchrony in a rectangular pattern, or raster, over both the specimen and the CRT. The brightness of the beam in the CRT was made proportional to the intensity of some signal generated by the beam striking the specimen, and the magnification was the ratio of the dimensions of the two rasters (Knoll 1935) (see Fig. 2.1). Although few microscopical sampling methods could be implemented in Knoll’s time, the sampling approach embodied in this second type of microscope has many potential advantages. To produce an image or map, the results of an interaction between the beam and the specimen need only be detectable, rather than focusable (i.e., the electron microscope (EM) specimen no longer has to be thin enough to 1
This chapter relies to a considerable extent on Pawley (1992).
H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
27
28
J. B. Pawley
Fig. 2.1 Knoll’s diagram of a scanning-type microscopical imaging system (Knoll 1935)
transmit the impinging electrons.) In addition, because the image information is carried as a time-varying electronic signal, a variety of analog and digital signal processing procedures can easily be applied to this signal to emphasize the particular aspects of it that are of interest to the viewer (i.e., contrast can be arbitrarily manipulated electronically). Unfortunately, Knoll’s instrument was really more like a video image sensor than a microscope. It operated only at very low magnification and, as discussed in more detail in Chapter 1 (this volume), the performance of a more advanced design by von Ardenne (von Ardenne 1938) was also limited by the capabilities of the electronics of the period. In Knoll’s instrument, the scanning transmitted electron signal was recorded directly on a mechanically-scanned photographic plate, and the complexity of the mechanism needed to scan the plate in synchrony with the beam effectively limited application of the scanning approach to transmission electron microscopy for some time. Scanned electronic imaging was instead applied to the field of television because there the time-sequential nature of the electronic signal that is present between the imaging tube and the display CRT greatly simplifies its widespread dissemination by broadcasting (McMullan 1990).
The Rise of the Modern Surface-imaging SEM In the 1950’s, the scanning approach to electron microscopy was rejuvenated by the Cambridge group under Oatley (Oatley 1972, 1982; McMullan 1953a, 1953b; Wells 1975). This group developed improved detectors for secondary electrons (SE) and backscattered electrons (BSE) (Everhart & Thornley 1960; Wells & Oatley 1959) and also benefited from wartime improvements in the performance of the
2 LVSEM for Biology
29
electronics used for scanning the beam in a rectangular raster and for the electronic display and recording of the detected signals as images. The SEMs developed and used world-wide since that time owe most of their important features to early developments by this group (Oatley et al. 1965).
Electrons as Probes in Scanning Microscopes Compared to light, x-rays, or other elementary particles, electrons are perhaps the ideal excitation source for scanned probe microscopic imaging. Four of their important features include: • Monoenergetic sources of high specific brightness (quanta/cm2 /ster) are readily available and easy to maintain (Oatley 1975; Hainfeld 1977) • The electron wavelength is very short and available lenses can focus electrons into a Gaussian probe as small as 0.5–2 nm in diameter depending on beam voltage • The charge on the electron makes it possible to use electromagnetic fields to scan the probe over the surface of the specimen rapidly and accurately • Energetic electrons striking a solid specimen surface are capable of exciting a wide variety of detectable signals (Everhart et al. 1959; Clarke 1970). The resulting signals include electrons produced by secondary emission (SE) (Everhart et al. 1959), auger emission (AE) (MacDonald 1971; Gerlach & MacDonald 1976), electron channeling (Coates 1969; LeGressus et al. 1983), and backscattering (BSE) (Ball & McCartney 1981), as well as characteristic and continuum x-rays (Duncumb 1957; Newbury et al. 1988; Statham 1988), electron-hole pairs (Breese 1982), light (Jakubowicz 1987) and heat (measured as sound) (Rosencwaig 1982). Because these signals are only elicited from the area of the specimen immediately under the beam, they need only be collected (rather than focused) to produce an image or map of some aspect of the material producing the interaction. The spatial resolution of the map will depend on the size of the probe/specimen interaction volume. Though all of these interactions and others have been used to produce useful SEM images, it is fair to say that images using either the SE or the BSE signal constitute the vast majority of recorded images. The reasons for this are: • SE and BSE can be both produced and detected with high quantum efficiency • On many specimens, the amount of SE signal collected from each point is roughly proportional to the angle between the viewing direction and the normal surface. Such an image conveys a fairly accurate impression of surface topography to the human brain. (Everhart et al. 1959; Wells & Oatley 1959) • In favorable circumstances, much of the SE or BSE signal can be generated from the volume of the specimen immediately adjacent to the beam impact point. Consequently, given a small beam, these signals have the potential for transmitting structural information with high spatial resolution
30
J. B. Pawley
In the case of the SE signal, specimen topography produces variations in the detected signal that closely mimic those that produce changes in the apparent brightness of a macroscopic surface with the same relative shape and illuminated with diffuse light coming from the direction of the SE collector. As a result, an SE image of a rough microscopic specimen can be easily and accurately interpreted in terms of its topographic shape (Hayes 1980). The total BSE signal is a strong function of the density of the specimen under the beam, and BSE images of flat specimens are therefore primarily two-dimensional maps of material density versus position (Boyde 2003; Ferguson et al. 2003). Because of the predominant importance of these signals, the remainder of this chapter will concentrate on those aspects of SEM design and operation that affect the contrast and resolution of the SE and BSE images (Chapter 3 contains a comprehensive discussion of SEM resolution.).
Limitations Associated with the Use of Electrons as the Probing Radiation Electrons are in many ways an ideal excitation source for scanned probe microscopy, but they also have some disadvantages that are not associated with other focusable quanta such as ions (Levi-Setti et al. 1984; Wang et al. 1989), light photons (Pawley 2006), or x-rays (Cheng & Jan 1987; Atwood & Barton 1989). These disadvantages are generally so well-known and accepted, however, that their existence often passes without comment—they are perhaps worth considering here more explicitly. The rationale for this is that success in using the SEM to image complex, organic surfaces will, in large part, depend on the ability of the researcher to avoid or ameliorate the effects of these disadvantages. It is also true that many of these effects vary strongly with beam voltage (Vo ) and, as a result, this parameter becomes the major determinant of SEM performance (Joy 1984, 1985, 1991b; Joy & Pawley 1993). The Characteristics of Electron Lenses All available electron lenses are converging and hence, in practice, the effect of lens aberrations can only be limited by reducing the lens aperture angle, α. The dominant lens aberrations are chromatic and spherical2 . The former produces a blurred spot of diameter: dc = Cc α
V0 V0
(2.1)
where Cc is the chromatic aberration coefficient and V0 is the voltage spread of the beam. Spherical aberration produces a blurring: 2 See Chapter 3, for a description of methods to correct spherical and chromatic aberrations in electron lenses and of the limits on depth-of-focus that then become paramount.
2 LVSEM for Biology
31
ds =
1 Cs α3 2
(2.2)
where Cs is the spherical aberration coefficient. Both Cc and Cs are lengths on the order of the focal length of the electron lens.3 Although the de Broglie wavelengths (λ) of 1–30 keV electrons are short (39 to 7 pm) compared to those of light (400–700 nm), diffraction remains a serious limitation because α must be kept very small to reduce the effects of aberrations. This is especially true for beam energies in the lower part of this energy range. Blurring due to diffraction is: dd = 0.6α/λ
(2.3)
The Intensity of Electron Sources Because electron lenses must be used at small acceptance angles (α = 10−2 −10−3 radians), it is not always easy to provide enough current in the beam to produce a well-defined image in a reasonable length of time. The current density in the focused spot can never be greater than that it is at the source. Early sources consisted of a heated tungsten hairpin and could produce a maximum useable intensity of only about 10 A/cm2 (Oatley 1975; Ohshita et al. 1978) or, assuming perfect optics, about 10−12 A in a 10 nm probe. This current corresponds to 6×106 electrons/sec, and if we imagine an image made up of 1,000 lines, each with 1,000 picture elements (pixels) that are scanned in one second, this corresponds to an average of only 6 electrons/pixel. The actual number is governed by Poisson statistics and is there√ fore 6 ± 6. Clearly, only a very high-contrast specimen can be imaged at all under these conditions, and even then only defined by two or three gray levels. In fact, the contrast of small surface features imaged with SEs is often only a few percent, and the current density in the spot may be reduced well below that at the cathode by practical and theoretical considerations. Under these conditions, small features can only be detected by producing more signal (i.e., 104 –105 quanta/pixel). This can be done either by scanning much more slowly or by using higher brightness electron sources to provide more beam current, thereby improving the statistical accuracy of the brightness measurement in each pixel (Wells 1975, 1978; Pawley 1990, see also Chapter 4) for a more comprehensive discussion of the effect of contrast and statistics on resolution.
The Effects of the Electron Charge Although the fact that the electron has a charge simplifies the process of focusing the probe and of scanning it in a raster, it complicates SEM observations of nonconductive specimens. As the total yield of BSE, plus SE, per beam electron is usually 3
See Chapter 3, for a discussion of aberration correctors as used for SEM.
32
J. B. Pawley
less than 100% at V0 > 5kV, negative charge accumulates within the scanned area of bulk specimens (i.e., those thicker than the electron range). Fields associated with this charge can defocus or deflect the beam and interfere with the collection of low energy (<50eV) secondary electron signal (Pawley 1972; Shaffner & Hearle 1976; Brunner & Schmid, 1986; Joy 1987). Because of their charge, beam electrons also interact with each other as they travel between the source and the specimen. The effect of this interaction is to reduce the effective brightness of the electron source (Boersch 1954; Pfeiffer 1972; Barth et al. 1990), and it is more pronounced when operating at low V0 and when using electron optical designs containing high-current density crossovers. Radiation Damage Produced by the Probing Beam When beam electrons 1–30 keV in energy strike a surface, they interact either with the positive charge of the nuclei or with the negative charge of the electrons in the specimen. Because of the large difference between the mass of the electron and that of the nucleus, and the requirement that individual beam/specimen interactions must conserve both energy and momentum, only the electron-electron collisions involve the transmission of a substantial fraction of the energy of the beam electron to the specimen (inelastic collisions). As a result, the kinetic energy of an electron in the beam is deposited within the specimen by means of a series of inelastic collisions with specimen electrons. Because many of these collisions transfer more than the few eV needed to break molecular bonds, the effect of this energy deposition on the specimen is indistinguishable from that produced by a similar dose of x-rays or any other form of ionizing radiation (Glaeser 1971, 1975; Cosslett 1978; Armstrong et al. 1990). The SI unit for the absorbed dose of ionizing radiation is the grey (1 Gy = 1 joule absorbed/kg = 100 rads). A 1kV beam with a current I0 of 10−12 A deposits V0 I0 = 10−9 watts of power within the specimen. If this is uniformly absorbed over a penetration depth of 10 nm, within a scanned area 1µm on a side, (i.e., ∼100 kx) on a specimen of density 1g/cc, radiation will be deposited at a rate of 105 gy/s (or 107 rads/s). Although at SEM energies (1–30 kv) such an immense dose has almost no effect on crystalline metals, it can produce subtle changes in ionic materials (Hobbs 1979; Diehl et al. 1990; Adams et al. 1990; Humphreys et al. 1990; Yokota et al. 1990) and usually rapidly degrades most organic compounds to a carbon skeleton (Glaeser 1971, 1975). Living specimens can be examined in the SEM only if the region of high-energy deposition is made to coincide with some fairly thick and inert biological structure such as the outer cuticle of an insect (Pease et al. 1966 also referred to by Boyde in Chapter 7, this volume) or the cell wall of a plant (Pease & Nixon 1968). On bulk specimens, there are two aspects of how radiation damage varies with Vo : a near-surface dose and a total dose. Although it is true that, at low V0 , a greater fraction of the energy is deposited within a few nm of the specimen surface, it is also true that as a result more secondary electrons are produced per beam electron. As a result, the near-surface dose needed to produce a SE image of a certain quality is almost independent of Vo .
2 LVSEM for Biology
33
If we take the matter one step further and consider a signal to be not the total number of SE produced when the beam strikes a flat surface at normal incidence, but rather the changes in this signal (contrast) produced by the topographic features of a rough surface, then we are interested in how signal contrast changes with V0 . In the size range of interest, SE topographic contrast is almost always higher at low V0 . This effect can be seen in images of a specimen of solid gold (Au)/palladium (Pd)—a high-density metal (see Fig. 2.2). The change of contrast with V0 is clearly evident. In fact, on this Au/Pd specimen, the reduced contrast at high V0 made it so difficult to focus properly that we had to focus on a piece of charging dust. The contrast from the pit was invisible when the beam was scanned fast. Figure 2.3 shows the surface of a slightly more relevant specimen—the cyst wall of a Giardia parasite—coated with 2–3 nm of platinum (Pt). Although the contrast of the high V0 images has been increased photographically, the low kV images seem better because they are less noisy and one can see that information about the surface of the fibers peaks at somewhere around 2 kV. Because of this relationship between kV and contrast, the near-surface radiation dose needed to produce an image of a given statistical quality (i.e., freedom from statistical noise) will be less at low V0 . Higher image contrast also means that less I0 is required to produce an image of a given quality at low V0 . As a result, not only the near-surface dose, but also the total power deposited (V0 I0 ) to the specimen is reduced. A dramatic example of how much it is reduced can be seen in Fig. 2.4, which shows the first and second members of two stereo pairs made at either 1.5 or 10 kV. The specimen is the cyst wall of Giardia—a protozoan parasite that has proven to be very sensitive to radiation (Erlandsen et al. 1989a, 1989b, 1990b 2004. See also Chapter 8). Although some shrinkage of the cyst wall can be seen even at 1.5 kV, the gross disfiguration present at 10 kV means that, on such specimens, one can only use the lower V0 (Erlandsen et al. 1989a). Although most biological specimens are fortunately not as sensitive as this, deformation on a smaller scale is more the rule than the exception when viewing topographically diverse, lightly coated biological or polymer specimens at high magnification.
Vacuum Requirements Imposed by the Electron Probe Because the mean-free path of electrons in a gaseous environment is short (nm), SEM specimens are almost always maintained in a high-vacuum environment. This makes it difficult to view specimens having high vapor pressures and, as a result, specimens containing liquids such as biological tissue must either be dried or viewed frozen on a low-temperature stage.4 4 The exception is the environmental SEM in which the specimen resides, in a chamber maintained at a fairly high ambient pressure, and is separated from most of the electron optical column by a stage of high-speed differential pumping. Although this technique is very useful for viewing uncoated insulators, it can only operate at fairly low resolution and relatively
34
J. B. Pawley
Fig. 2.2 Effect of V0 on image contrast for solid metal. The metal is an Au/Pd alloy that has been shaped with a microindenting tool and was provided by David Plantz (Materials Science, University of Wisconsin-Madison). Because this early specimen contaminated very rapidly, each image had to be made using a different indentation. When attempting to focus the 20 kV image with a rapid scan raster, the pit was essentially invisible and focus could only be adjusted on the central dust particle that was bright because it charged up. As V0 increases, the reduction in the topographic contrast of small features is evident
2 LVSEM for Biology
35
Fig. 2.3 Effect of V0 on image contrast for a dried biological specimen coated with 2–3 nm Pt and viewed at: (a) 1 kV, (b) 2 kV, (c) 5 kV, (d) 20 kV. The specimen is the cyst wall of Giardia (Erlandsen et al., 1989b). Although the same structures can be recognized in all the images, the only features having high contrast when viewed at higher kV are those that are thin enough for the beam to penetrate and produce significant signal from the lower surface
The vacuum constraint is important because, at the finest level of microscopic imaging, the most severe obstructions to an understanding of native biological structure are often related to the need to prepare specimens so that they will be mechanically stable in the vacuum of the SEM (Kellenberger 1991; Erlandsen et al. 1989a.
high V0 because scattering of the primary electron beam by the gas surrounding the specimen produces a high background signal. For this reason it will not be considered further here (Danilatos 1988).
36
J. B. Pawley
Fig. 2.4 Reduction at low V0 in the effect of radiation damage on a sensitive biological specimen. Upper two images show both members of a stereo pair of an empty, isolated Giardia cyst wall made at 1.5 kV (Erlandsen et al. 1989b). Some shrinkage is evident in the second member of the pair (right), but this is minor compared to the extreme shrinkage evident when the same experiment is performed on a different cyst wall at 10 kV (lower pair). The fact that Giardia cyst walls contain a high proportion of glyco-proteins may explain their high radiation sensitivity. It should also be noted that, in contrast to the intact Giardia shown in Fig. 2.3, these cyst walls are empty and contain no stabilizing internal structures
2 LVSEM for Biology
37
See also Chapters 5, 6, 8 and 10). Indeed, the challenge of preparing biological specimens without artifacts is probably the main limitation to the more widespread use of the SEM in biology.
Beam-induced Surface Contamination The effects produced on the specimen by exposure to the scanned radiation source represented by the electron beam are not limited to the shrinkage and/or volatilization of its constituents by radiation damage. The reverse process is also true. Except in vacuum systems that have been highly baked, the surfaces of solids are always covered with small organic molecules held in place by van der Waals forces. At room temperature, thermal excitation causes these molecules to migrate over the surface. Those excited or ionized by interaction with the beam, however, are likely to react chemically and often polymerize into a surface film that soon obscures the true surface of the specimen (Fourie 1981; Hren 1986). This process is generally referred to as contamination. It can be reduced either by improving the vacuum to decrease the amount of surface-migrating hydrocarbons, or by heating or cooling the specimen. Heating breaks the van der Waals bonds, driving the hydrocarbons from the surface (Ogura et al. 1989a) but it can only be applied to heat-tolerant specimens. Cooling has the effect that the hydrocarbon molecules, although still bound to the surface, have insufficient thermal energy to migrate to the irradiated area (Wall 1980). Figure 2.5 shows an image of the surface of an Au grid held at −160 ◦ C in a Gatan cryotransfer stage. The left contamination raster was deposited by rapidly scanning the surface with a 100 kX raster for 100 seconds at room temperature, while the raster on the right was deposited while the specimen was at −160 ◦ C. The picture frame around the room temperature raster is caused by the polymerization of hydrocarbons the instant they diffuse into the scanned area. The reduction in hydrocarbon contamination when the specimen is irradiated at low temperature is clearly evident.
Lack of Chemical Contrast Mechanisms As noted above, the SE signal from the SEM produces images that are readily interpretable in terms of topography, while the BSE signal is proportional to the local specimen density, and as described in Chapter 11, the detection of characteristic x-rays allows one to map elemental composition. In many disciplines (particularly biology), however, the observer would also like to determine the molecular constitution of the viewed surface as well as the atomic. In biology, classical histology has been useful to the biologist because it uses specific dyes to label particular chemical species within the cell. Aside from some early attempts to localize fluorescent histological markers using the SEM in the cathodoluminescent (CL) mode (Pease & Hayes 1966; Hayes 1980), and the use of procedures that deposit heavy metals that can be localized in BSE images (Becker & Sogard 1979; Vanderburgh et al. 1987), little histochemical work has been done in the SEM. The main reason for this is that
38
J. B. Pawley
Fig. 2.5 Reduction in contamination rate at low specimen temperature. A rapid-scan 100 kx raster was allowed to strike this Cu grid specimen for 100 sec twice, first at room temperature (left square deposit) and then at −120 ◦ C (right square deposit). A micrograph of both deposits was then recorded at 1.5 kV, −120 ◦ C. The picture frame nature of the room-temperature deposit shows molecules polymerized as they entered the irradiated area by surface diffusion. Its absence in the less-visible contamination pattern on the right shows that surface diffusion is not important at low specimen temperature
radiation associated with the electron beam rapidly destroys fluorescent or other organic markers, and this fact can only be seen as a disadvantage in using electrons as probes. Alhough normal histochemistry is difficult in the SEM, immunolocalization techniques can be very sensitive and productive. This is especially true of those techniques in which the selective molecule (antibody, lectin, or messenger) is coupled to a particle of colloidal-gold (Faulk & Taylor 1971; Horisberger & Rosset 1977; Horisberger 1979; Bendayan 1984, 1987; Albrecht & Hodges 1988; Erlandsen et al. 1990a, 1991. See also Chapter 6), because these particles can often be seen by both light microscopy (LM) and electron microscopy (EM) techniques (Albrecht et al. 1989; Goodman et al. 1990). Gold markers as small as 1–2 nm in diameter can be seen in BSE images from the SEM (Erlandsen et al. 1991; Müller & Hermann 1990) making it, in some ways, the preferred method for localizing macromolecular markers of this type. Alhough it is conventional to view these specimens using the BSE signal at relatively high Vo , it is also possible to view them at low V0 using the SE signal (Pawley & Albrecht 1988; see Fig. 2.6).
2 LVSEM for Biology
39
Fig. 2.6 Gold label visibility as a function of V0 . On the left are a series of three SE micrographs of the same platelet, surface labeled with 18 nm Au-fibrinogen, carbon coated for conductivity, and recorded at (a) 1, (b) 2, and (c) 5 kV (fieldwidth = 6.6 µm). The images on the right show the same label particles at higher magnification on an adjacent cell (fieldwidth = 1.3 µm). At 1 kV, only the outer surface of the carbon coat is visible. At 2 kV, the gold core can just be discerned, while at 5kV, the image closely resembles a 20 kV BSE image (Pawley & Albrecht 1988; See also, Chapter 6)
40
J. B. Pawley
The Finite Size of the Interaction Volume As with all microscopic techniques, there has always been much interest in the ultimate spatial resolution of the SEM. The crucial factors affecting SEM resolution were debated by von Ardenne in 1938 and in the discussions following McMullan’s first paper (McMullan 1953a. See also Chapter 1). Obviously, the primary limitation on resolution is the size of the probe and McMullan discussed the factors that limited its diameter at that time. Haine and Gabor pointed out, however, that “electron diffusion” (i.e., scattering within the specimen) would eventually place a more fundamental limit on SEM resolution. In a scanning probe instrument such as the laser confocal fluorescence microscope (Pawley 2006), in which the detected interaction occurs between a 1–2 eV photon from the beam and a single atom or molecule of the specimen, the interaction volume is effectively zero and the spatial resolution is almost entirely defined by the size of the exciting probe. When the probe is a beam of kilovolt electrons striking a low-density bulk specimen, however, a large number of inelastic collisions are required to absorb the energy of each beam electron. Between each of these inelastic collisions, elastic collisions between the beam electrons and the nuclei of atoms within the specimen can produce large changes in trajectory, causing the beam to spread out. As a result, absorbing the energy from each beam electron involves a large number of electron/specimen interactions, occurring over a significant region of space. When the beam diameter is small compared to this volume, the spatial resolution of the produced image is apt to be limited more by the dimensions of the volume within which detectable interactions take place than by the beam diameter. As noted in Chapter 1, this fact was well understood by von Ardenne in the 1940s and K.C.A Smith makes the same point in his thesis, published in 1956: “It would appear desirable to reduce the accelerating voltage until the penetration is of the same order as the size of the detail to be resolved; response from such detail then being greatest.” Because of this complication, there is no standard method of measuring SEM resolution and it has therefore always been an elusive concept (Wells 1974; Catto & Smith 1973). Unlike the TEM, where measurement of the contrast transfer function of a thin-phase object provides at least a standard method and a useful basis for the comparison of instrumental performance (Thon 1965), the search for an ideal, bulk SEM test specimen has been long but generally unsuccessful (Ballard 1972; Watabe et al. 1978; Peters 1989; Postek 1987; Black & Ballard 1982). Most highresolution test specimens used today produce contrast through changes in specimen density or signal collection efficiency rather than through changes in topography (Broers 1974, 1982). As a result, even when they employ Fourier Transform techniques (Dodson & Joy 1990; Wepf & Gross 1990; Wepf et al. 1991), they give only an estimate of beam diameter rather than the size of the smallest topographic feature that might be visible on the surface of a flat, bulk specimen (See Watabe et al. 1978; Peters 1980, 1982, 1985, 1991 for possible exceptions). The discussion that follows refers only to bulk specimens and not to specimens that have been kept so thin that subsurface scattering has been eliminated by virtue of the fact that there is no subsurface (Hermann et al. 1988). In this case, the
2 LVSEM for Biology
41
interaction volume has much the same dimensions as the probe (Joy 1991a). An image of this type is shown in Figs. 2.7a and 2.7b, which show segments of isolated actin filaments adhering to a thin film on an EM grid. This specimen was prepared by Ya Chen by high-pressure freezing, followed by low-temperature freeze drying, cyro-coating with Cr, and observation at 30 kV on a cryotransfer stage in the Hitachi S-900. Figure 2.7c shows the same approach applied to produce a stereo image of virus particles. The problem, then, is that for most detectable beam/specimen interactions, some of the signal is produced by interactions taking place at some distance from the area of the specimen initially struck by the primary beam (Seiler 1976; Hasselbach
(a)
(b)
Fig. 2.7 Thin specimens suitable for observation at high V0 . Specimens that are thin and can be mounted on thin substrates can take advantage of the smaller probe diameter available at high V0 because there is little subsurface material to generate any background noise signal. (a) shows several views of cryoprepared actin filaments extracted and enlarged from a single micrograph originally recorded at 110 kx. (b) shows a stereo pair recorded at 400 kx from a similar specimen. Field width = 160nm. (c) Stereo-pair of cryo-prepared virus particles recorded at 400 kx, 30 kV. Field width = 160 nm. The image on the left was recorded first. All images generously provided by Ya Chen, of the IMR, University of Wisconsin-Madison
42
J. B. Pawley
(c)
Fig. 2.7 (continued)
et al. 1983). The SE signal consists not only of SE produced where the beam strikes the specimen, but also of SE excited by BSE as they re-emerge through the specimen surface or strike the lens pole piece. As a result, the signal from the SE detector often has a significant BSE component (Crewe & Lin 1976; Peters 1979, 1982, 1986a, 1986b; see Fig. 2.6). The momentum of the probe electrons carries them into the specimen, exciting a volume within the specimen (Wight & Zeissler 1999). On a specimen with uniform subsurface structure, the total volume excited is roughly hemispherical, while that in which most of the energy is absorbed is more pear-shaped. The exact shape can only be discussed in terms of an energy-deposition probability distribution that defines the way that the energy disposition rate decreases with radial distance from the beam impact point—a process that will be further complicated by any inhomogeneity in subsurface specimen density (see Fig. 2.6c). The shape and size of this volume have been thoroughly investigated for simple geometries using Monte-Carlo simulation techniques (Kotera et al. 1981; Joy 1985; Catto & Smith 1973; Joy 1984, 1987, 1991a, 1991b; Joy et al. 1982; Murata et al. 1987, Joy & Pawley 1993). Figure 2.8 emphasizes the great reduction in this volume with kV. While (at 1 kV on a Pt specimen) the dimensions of this interac5/3 tion volume may be only a few tens of nm, these dimensions scale with Vo /D (Reimer 1979; Joy 1991a) (where D = density) and may reach hundreds of µm at 30 kV on samples such as dried lung (D = 0.05g/cc). Figure 2.9 shows how the size of the area, from which 85% of the SE emerges, varies with V0 on a bulk specimen. At best, signal derived from interactions remote from the beam impact point on relatively flat specimens produce a background signal that greatly reduces the
2 LVSEM for Biology
43
Fig. 2.8 Monte-Carlo plots of electron scattering in carbon at a nominal density of 1 for V0 = 1.5, 5 and 20 kV. (Generously provided by D. Joy, University of Tennessee.) 1010
Dry biological tissue
Emission area (Å2)
109
Dry biological tissue
108 Amorphous carbon
Amorphous carbon 107
106
Area from which 85% of signal emerges 105
1
3
5 Energy (keV)
7
9
Fig. 2.9 Calculated variation with V0 of the size of the area around the beam from which 85% of the BSE signal escapes, assuming a dried biological specimen (i.e., a carbon matrix with a density 0.2g/cm3 ) and carbon (density 1g/cm3 ). (Generously provided by D. Joy, University of Tennessee.)
44
J. B. Pawley
topographic contrast produced by small surface features. In practice, the lowresolution background signal can be much larger than the near-probe signal on rougher samples. While the average value of the background can be electronically subtracted with black-level control, statistical variations in this signal cannot be subtracted and they therefore tend to obscure the real topographic signal. This problem can only be overcome by counting more events and hence subjecting the specimen to more radiation. As microscope resolution can only be usefully defined in terms of contrast vs spatial frequency, specifying useful topographic resolution as a function of V0 and, in the absence of a suitable test specimen, becomes much more complicated than a simple measurement of beam diameter. When one adds the effects of the radiation damage needed to produce the signal to the specimen, and (possibly) of the effect of adding the conductive coating layer needed to avoid charging on insulators, the estimation of topographic resolution can rapidly become more philosophical than scientific (Pawley & Erlandsen 1989).
Response to These Limitations It bears repeating that success in achieving the ultimate in SEM image quality will largely depend on the extent to which the disadvantages inherent in the use of electrons as the probing radiation source can be overcome. Because both attainable beam diameter and the severity of a number of these deleterious properties vary strongly with Vo , it is probably the most important experimental variable in the quest for ultimate SEM performance. Although the two essential instrumental features needed for high-resolution SEM at low V0 have been available for at least 20 years, they were not available to the first developers of the modern SEM in the 1950’s. As a result, the early widespread acceptance of 10–30 kV as the normal range for SEM operation depends more on the historical development of the field than on a sober scientific analysis. As commercial instruments embodying both of the features necessary to produce a small beam at low V0 have become commercially available only recently, now is perhaps a good time to reevaluate the role of V0 in effective SEM resolution. I will start with a brief historical summary of the development of the SEM5 . This will be followed by a description of the relevant features and performance of modern LVSEM equipment, an analysis of factors that presently limit high-resolution SEM performance, and a brief survey of recent applications in the study of biological specimens. I will conclude with a few comments on the application of cryotechniques and on possible future developments.
5
A much more complete history can be found in Chapter 1 by McMullan.
2 LVSEM for Biology
45
The Evolution of Topographic Imaging with the SEM The SEMs developed by the Oatley group in the 1950s (Oatley 1982) evolved at the same time that other groups in Cambridge were developing electron optical columns for the point-projection x-ray microscope (Cosslett 1954; Nixon 1955) and the scanning x-ray microprobe (Duncumb 1957), after that introduced by Castaing & Guiniert 1949). These latter two instruments operated at 10–30 kV and used hot tungsten filaments—as did the early SEM columns. Aside from pioneering work by R.F.M. Thornley (Thornley 1960; Thornley & Cartz 1962), who used low V0 (1.5–3 kV) primarily to avoid charging on uncoated, frozen, biological specimens, the emphasis was on what we might now call high voltage (10–30 kV) use (Pawley 1997). The reasons for this choice were fairly clear-cut. Initially, resolution was limited more either by mechanical and electronic instability or by lens aberrations than by the size of the interaction volume (Oatley et al. 1965; Pease & Nixon 1965). In any case, the effective size of the interaction volume could be markedly reduced by coating the outer surface of low-density specimens with 20 nm of Au (D = 19.7g/cm3 ), which effectively reduced penetration by about 100x compared to dried, biological specimens (D = 0.2 g/cm3 ). The actual electron-optical performance of the early instruments was so low that the presence of such a thick coating did not obscure any objects that might otherwise have been seen.
Barriers to Operation at Lower Vo As first enumerated by Oatley et al. (1965), the barriers to high-resolution operation at low V0 include: 1) Low-source brightness. The current density in the focused spot, J, is limited by the Langmuir equation (Langmuir 1937): J = J0 (eV0 /kT) + 1 sin2 α =
J0 eV0 α2 kT
(2.4)
where Jo is the current density at the surface of a thermionic cathode, e is the electron charge and kT is the thermal energy of the electrons leaving the cathode surface at temperature T. Because of this relationship, reducing V0 from 20 to 1 kV requires a spot at least 20x larger in area to produce the same beam current. In practice, the required area of the spot is more like 100x larger (or 10x greater in diameter) because most thermal electron guns are not designed to reach theoretical performance at low kV (Oatley 1975; Yamazaki et al. 1984). 2) Chromatic aberration produces a greater defocusing effect. As chromatic blurring, dc , scales with αV0 −1 (see Equation 2.1), the only way to preserve spot size as V0 is reduced, is to also reduce α. This reduction, however, compounds both the lack of brightness and the effect of diffraction (Shao & Crewe 1987).
46
J. B. Pawley
3) Longer electron wavelength. This increases the effect of diffraction. As the V0 is reduced, λ increases and α must be increased to keep dd constant (see Equation 2.3). 4) Susceptibility to stray AC and DC fields. The strength of the magnetic field needed to focus an electron probe with a given focal length, f, varies with V0 0.5 . Consequently, as V0 is reduced, stray AC and DC magnetic fields constitute a proportionately greater fraction of the strength of the lens field, and they therefore produce a relatively greater distortion in the shape or position of the probe (Pawley 1985a; 1987b). In addition, the electron beam at lower V0 is less likely to penetrate (and thereby discharge) electrostatically charged oxide or hydrocarbon contamination layers on the inner surfaces of metallic column components (Anger et al. 1983). Therefore, low V0 operation requires a much higher level of instrumental cleanliness and magnetic shielding for proper operation. Oatley et al., did not list a fifth consideration that became evident only later: Contaminating surface layers of hydrocarbons or other materials are not only deposited somewhat more efficiently at low Vo , they are also more easily visualized at low kV (Brandis et al. 1984; Pawley 1984b; Pawley 1985a). The first three points listed above insure that, with a given electron optical column, a smaller probe can always be made at higher Vo . The other two factors merely add practical complications to the process of actually using the SEM at low Vo . Although the advantages of LVSEM were espoused by some (Kosuge et al. 1970; Boyde 1971; Boyde et al. 1974), the difficulties listed above severely limited the use of SEM at V0 <3 kV (LVSEM) for two decades (Dilly 1980; Pawley & Wall 1982; Pawley 1984b; Pawley & Erlandsen et al. 1989; Volbert 1984).
Electron Optical Developments In the 1970s, improved electron sources utilizing LaB6 Schottky thermionic cathodes (Broers 1974) and cold field-emission (FE) cathodes (Crewe et al. 1968; Crewe 1973; Hainfeld 1977) were first introduced for use in the SEM. In addition, it was found to be possible to collect SE or BSE from a specimen mounted inside the lens field of a TEM-type objective lens (Koike et al. 1971) (Wells et al. 1973; Broers et al. 1975; Broers 1982). This optical layout reduced lens-aberration coefficients by a factor of about 20. Initially, LaB6 sources were more popular than FE sources, as they had much less stringent vacuum requirements (Sewell & Ramachandran 1978, Miyokawa et al. 1988). As commonly used, however, they had only about 10–20x the brightness of the tungsten sources, while the FE was 105 times brighter. FE sources also had a smaller effective energy spread (V0 = +0.15eV (Crewe et al. 1971) versus 2–3 eV for LaB6 (Wells 1975), and this was important when operating at low V0 where chromatic aberration is often the dominant aberration (Bauer & Speidel 1981; Tuggle et al. 1986; Shao & Crewe 1987). This improved LVSEM performance was demonstrated first by Welter & Coates (Welter & Coates 1974) on an FE-SEM with
2 LVSEM for Biology
47
the specimen mounted below the lens and later on a prototype instrument with a quasi-immersion lens by N. Tamura (Tamura et al., 1988). A system for extracting a SE signal from an SEM specimen that was scanned while inside the lens field was first described by Koike for the 100 kV TEM/SEM (Koike et al. 1971) and later by Buchanan (Buchanan 1982) on a dual-stage SEM. This approach is limited to observations on specimens that are both nonmagnetic and small enough to fit into the lens gap. In addition, there is a deleterious interaction (described below) between the beam and the field of the SE detector. The immersion lens, however, has the advantage of reducing Cc and Cs by factors of 10 to 30 while permitting a highly efficient collection of the SE signal, and effectively shielding the specimen area from the effects of stray magnetic fields.
Improvements in SE and BSE Detectors In parallel with these electron-optical developments, were improvements in the SEM detectors themselves—especially those for SE and BSE. The detector used for surface imaging in McMullan’s microscope actually used an electron multiplier with beryllium-copper dynodes mounted below the specimen, and amplified the SE produced when BSE from the specimen struck the lower surface of the objective lens (McMullan 1953a).6 This was soon replaced by the smaller Everhart-Thornley (E-T) SE detector, which was much more convenient, efficient, and versatile (Everhart & Thornley 1960).
Secondary Electron Detectors In the E-T detector, SE from the specimen are first attracted to a +300 V grid mounted to one side of the specimen, before being accelerated and focused onto a scintillator held at +10kV. Here, each electron, which now has energy of ∼10 keV, produces a few hundred photons—some of these travel down a light guide to a photomultiplier tube (PMT). The current output of the PMT becomes the video signal that is used to modulate the intensity of the display CRT. As long as each SE produces enough light at the scintillator to elicit ∼5 photoelectrons from the PMT photocathode, the detector is effectively noise free. In addition, it has a quantum efficiency, Q, of almost 100% and a bandwidth of more than 10 MHz (Wells 1975).7 The one disadvantage of the E-T detector is that it requires a transverse electrostatic field in the region above the specimen through which the probe must pass. This is also true of the variant of the E-T detector developed for the TEM/SEM by Koike et al. (1971). The transverse collection field displaces and distorts the beam (Pawley 1990), particularly at low V0 (see Fig. 2.10).
6 7
A similar method of detecting BSE was reintroduced much later by (Moll et al. 1979). See also Chapter 4 for a more complete discussion of this vital topic.
48
J. B. Pawley
Fig. 2.10 Schematic diagram showing the controlled misalignment that is present when a highresolution SEM is aligned at low-beam voltage to minimize image movement as the objective lens current is varied. Between A and B, the beam is attracted to the SE detector on its way towards the specimen, and between B and C the objective lens field makes the beam parallel to the EO axis but displaced off the axis by distance, d (Pawley 1990). Some modern SEMs now employ a Wiener (ExB) filter that superimposes a weak magnetic field perpendicular to the plane of the paper to reduce this electrostatic displacement effect and simplify instrument alignment (see Fig. 2.13b)
2 LVSEM for Biology
49
Consequently, there has been a sustained effort to develop an SE detector without this defect. This can be done either by superimposing a weak magnetic field perpendicular to the electrostatic collection field—a so-called Weiner filter (Zach & Rose 1986; Schmid, Brunner, & Schmid 1986,1987;Zach 1989; Pawley 1990)— or to develop cylindrically symmetric detectors for secondary and backscattered electrons arranged so that they are centered on the beam axis, such as those used on the Zeiss Gemini line of SEMs (Vermeulen 2004).
Backscattered Electron Detectors The collection field of the E-T detector is not strong enough to attract the more energetic (>50eV) BSE, but if a BSE trajectory happens to intersect the scintillator, signal is produced. Furthermore, if the E-T detector grid is biased to –50V to exclude the SEs, then a pure BSE signal is produced. Because BSEs travel in relatively straight lines, and the E-T scintillator subtends a rather small solid angle, only a small fraction of the BSEs are collected and the signal from those BSEs shows strong line-of-sight shadowing effects. In fact, BSE image contrast from an E-T detector on rough samples is often dominated by this shadow effect rather than by the density contrast often associated with BSE images. To overcome this effect, Wells (Wells & Bremer 1970; Wells 1979) developed scintillator detectors with a wide variety of sizes and geometries to intercept a larger fraction of the BSEs emitted from the specimen. This trend culminated in a detector by Vivian Robinson, in which the beam passed through a small hole in a piece of plastic scintillator, the lower surface of which was formed with a hemispherical depression with the specimen at its center (Robinson 1974). On flat specimens, the contrast revealed by the signal from such a detector was almost entirely caused by changes in density (Ball & McCartney 1981). Because both plastic and inorganic scintillators are insulators, their front surfaces must be covered with a conductive film to prevent surface charging. As a result, they are only sensitive to BSEs with sufficient energy to penetrate the film. In addition, the fluorescent efficiency of the plastic scintillator slowly degrades with prolonged use (Pawley 1974). Efforts to improve the BSE detector involved three alternative approaches—the development of better scintillator materials, thinner conductive coatings, and large-angle BSE detectors employing solid-state diodes or microchannel plate amplifiers (MCP). A BSE striking a silicon detector mounted above the specimen produces one electron-hole pair for each 2.3 eV of energy deposited (Kimoto & Hashimoto 1966). The resulting current can be amplified to produce a signal. While these detectors can, in principle, be very small, most are not in practice (Wolf & Everhart 1969). Larger detectors have greater capacitance and hence higher dark current and lower bandwidth. In addition, the presence of an insensitive dead layer near the surface makes them relatively inefficient at low Vo . Although some later devices have partially overcome this limitation (Walker et al. 1989), the bandwidth and noise characteristics of unbiased solid-state BSE detectors are still inferior to those of the
50
J. B. Pawley
scintillator/PMT systems. As a result, they are primarily used in situations where the installation of a light pipe is inconvenient. An MCP is a disk of glass a few millimeters thick and perforated with a denselypacked, hexagonal array of angled, micrometer-sized holes. All surfaces are covered with a continuous film of resistive dynode material, and a voltage of a few kV is imposed between the top and bottom of the disk. Electrons striking the lower, grounded surface produce secondaries, and these are attracted up the holes toward the positive side. Charge amplification occurs as these electrons strike the walls and elicit additional secondaries. For each incident electron, an amplified pulse can be collected from an electrode mounted above the positive surface of the disk. With a central hole to pass the beam, the whole system can be operated as a flat, symmetrical electron multiplier and, by adjusting the potential of the surface facing the specimen, the MCP can be made sensitive to SE and/or BSE (Venables & Harland 1973; Russell 1984, 1988; Russell & Mancuso 1985; Helbig et al. 1987; Gray et al. 1989; Postek et al. 1990a, 1990b). If the MCP is coupled to a position-sensitive electron detector, an entire electron diffraction pattern can be collected from each point in the raster (Ichinokawa 1990). Because the MCP has no dead layer, BSE with energies below 1 kV can be readily detected, and with only a slight positive bias, it can collect SE with high quantum efficiency and with no transverse field at the specimen. The only disadvantages are high cost, the fact that the signal emerges at a DC voltage 1–2 kV positive to ground (Postek et al. 1990a), and a 3–7 mm thickness that precludes use with most SEM optical systems operating with a short working distance—particularly those in which the specimen is mounted in the lens gap.8 Autrata steadily improved the efficiency and performance of the scintillator/PMT BSE detectors employing radiation-damage-resistant, single-crystal YAG scintillators (Autrata et al. 1978, 1983; Autrata 1989, 1990; Walther et al. 1991; Erlandsen et al. 1990a, 1991). Later models had thinner conductive coatings and special reflective and anti-reflective coatings on the appropriate surfaces of the crystal to help the light emerge from the high-refractive index scintillator and enter the light pipe (see Fig. 2.11a). This increases the efficiency with which light is transmitted to the PMT by almost 4x. As a result, this type of detector now works down to V0 = 1.3 kV (Müller & Hermann 1990). With such a detector, one can easily discriminate between 5 nm and 20 nm gold labels on biological specimens (see Fig. 2.11b). Compared with the MCP detector, it is much smaller and hence can be used to collect the BSE signal while an E-T detector simultaneously collects the SE signal, and this is even possible on SEMs that use immersion lenses (Pawley & Albrecht 1988). Compared to the silicon BSE detectors, the scintillator/PMT detectors generally have lower noise and higher bandwidth. Figure 2.12 shows the performance at different V0 of a modern scintillator, BSE detector mounted on a high-resolution SEM. The signal from the smaller (1 nm) Au particles seems best at about 8 kV. Below this voltage, increased probe diameter reduces sharpness, and hence contrast. Above this voltage, increased signal
8 The axial SE and BSE detectors introduced in the Zeiss Gemini SEM (Vermeulen 2004) avoid some of these problems, but their S/N performance is not easy to assess (see Chapter 4).
2 LVSEM for Biology matted periferal area
51 polished periferal area Al foil wrapping
YAG
diffusion layer
light guide
conductive oxide layer
to PMT
cement antireflecting layer
(a)
(b)
Fig. 2.11 (a) Design details of the improved Autrata YAG/BSE detector (Autrata 1990). The transparent, conductive tin-oxide layer is very thin, reducing BSE absorption losses but preventing surface charging. The diffusion layer scatters light from the far side of the detector towards the light-pipe and the antireflecting layer reduces reflection losses at the interface between the high-refractive index YAG and the fused-quartz light-pipe. (b) Double-labeled, detergent-extracted platelet from human blood, labeled with 5 nm Au-antitubulin and 18 nm Au-fibrinogen imaged both at 20 kV and at 5 kV with BSE in the S-900 (Pawley & Albrecht 1988). These images show that, given a suitable detector, even the smallest colloidal-gold labels can be easily seen at low kV. The arrows indicate overlying biological structures that are seen through at 20kV but not at 5 kV
from the substrate reduces image contrast from small surface features, increasing image noise. Although further improvements are possible, present detectors provide such a high level of performance that the major limitation is the signal loss associated with BSE going through the 1–2 mm hole through which the beam passes. If the scintillator were mounted in the aperture plane as the final, objective aperture, the area of the hole could be significantly reduced. Practical and geometrical constraints make this idea difficult to implement at present, but the SE/BSE detector on the Zeiss Gemini SEM comes close.9
9
The performance of SE and BSE detectors is covered in more detail in Chapter 4.
52
J. B. Pawley
Fig. 2.12 BSE images at 3,4,5,8,20 and 30 kV of the same field of 1 and 5 nm gold particles on a Cr-coated cell originally taken at 800 kx in the S-900, using a cold stage to reduce contamination. The particles appear smaller at high V0 because of the smaller beam diameter but visibility for the smaller particles seems greatest at about 8 kV. (Micrographs generously provided by Paul Walther, IMR, University of Wisconsin-Madison)
The Re-emergence of Low Voltage SEM These instrumental developments occurred against a background of increased interest in the use of the SEM to investigate the ever-smaller features of solidstate integrated circuits (Oatley & Everhart 1957; Thornhill & MacKintosh 1965; Catalano 1976; McMullin 1976; Todokoro et al. 1980, 1983; Menzel & Kubalek 1982; Breese 1982; Hashimoto et al. 1982; Buchanan 1982; Buchanan & Menzel 1984; Menzel & Buchanan 1985; Bennett & Guller 1986; Sugiyama et al. 1986, 1988; Russell 1988; Arnold et al. 1989; Krause et al. 1987, 1989). This interest affected the evolution of the SEM because it soon became evident that the electronic properties of the insulating layers deposited onto the semiconductor devices were rapidly degraded by exposure to the scanning electron beam (Speth & Fang 1965; Szedon & Sandor 1965; Keery et al. 1976; Miyoshi et al. 1982). Damaging effects, such as shifts in the threshold voltage of fieldeffect transistors, were often traceable to the electric field produced by electrons injected into, and trapped within, insulating layers buried in the device (Miyoshi et al. 1982). The only effective remedy was to reduce beam penetration by reducing V0 to 0.5–2 kV. The economic importance of the semiconductor market segment led manufacturers to greatly improve the low V0 performance of their instruments (Pawley 1984a, 1987b). As they succeeded, it became possible to make electron beams at low V0 that were small enough to test the idea that low V0 operations might be the preferred approach for achieving high-resolution topographic imaging
2 LVSEM for Biology
53
(Pawley & Wall 1982; Pawley 1984a, 1984b, 1985a, 1997; Boyes 1984a, 1984b; Hefter 1987, Joy & Pawley 1993). Aside from the developments in electron sources, lenses, vacuum systems, and magnetic shielding mentioned above (Pawley 1985b), there were other approaches to high-resolution LVSEM. These included several early attempts to develop aberration correctors (Frosien et al. 1989; Jones 1989; Zach 1989)10 and the placing of planar specimens in a strongly decelerating electrostatic field. On flat specimens, the latter strategy has the effect of reducing the lens aberration coefficient by a factor of 103 (Polasko et al. 1983). Although promising in some respects, the limitation to fairly planar specimens has (until now) confined use of this approach to very low-beam voltages (∼10–200 v) where surface contrast can be difficult to analyze (Mullerova 2001; Zobacova & Frank 2003). In 1986, an FE source—a short focal-length lens—and a good vacuum system were combined into a single commercial instrument for the first time (Nagatani & Saito 1986a, 1986b; Ohama et al. 1986). Such instruments could demonstrate a 3 nm probe at 1.5 kV, and so made it possible to evaluate the potential of LVSEM for highresolution topographic imaging (Pawley 1984b, 1984a, 1997; Osumi et al. 1989a, 1988b, 1989, 1990, 1995, 2006; Boyes 1984b, 1984a). More recent instruments have pushed performance to 1 nm at 1 kV on in-lens instruments, and to 1.5 nm on microscopes in which the specimen is mounted in a large chamber below the final lens.
Modern Instrumentation for High-Resolution LVSEM Alhough the low V0 performance of almost all types of commercial SEM has improved greatly over the past two decades, as noted above, this article emphasizes the ultimate capabilities of the technique. Therefore, I will confine my attention to those systems using an immersion lens and a FE source—specifically, the Hitachi S-900 (Nagatani & Saito 1986a, 1986b; Nagatani et al. 1990). The JEOL S-890 (Ohama et al. 1986; Kersker et al. 1989) and the ISI DS-130-FE that used a heated, point filament operating in TF (or Schottky) mode (Tuggle & Watson 1984; Tuggle et al. 1986; Orloff 1985; Buchanan 1982; Yamazaki et al. 1989) were similar in performance, but have all now been superseded by more modern instruments such as the Hitachi 4800 and 5500 (see Fig. 2.13). The relative merits of heated vs. cold FE sources have been widely debated at length (Hainfeld 1977; Orloff 1981, 1985; Swanson & Rathkey 1989). On balance, the emission from a TF filament operated with V0 = 1kV is free of the 1–2% temporal instability characteristic of cold FE (Nomura et al. 1973; Saito et al. 1982) and it has the ability to produce a higher total current in fairly large probes. On the other hand, it has somewhat lower brightness and a larger effective Vo . The vacuum requirements for TF are 10–50x less stringent than those for cold FE (10−7 pascal
10
This is covered in more detail in Chapter 4.
54
J. B. Pawley
(a)
Selectable signal detection mode Incident beam BSE -> SE Conversion Plate
Low voltage BSE detector
Concentric YAG BSE detector for high-kV SE detector Ex B Upper polepiece
Specimen
Objective lens
Lower polepiece
Brightfield/STEM aperture STEM detector (b)
Fig. 2.13 (continued)
2 LVSEM for Biology
55
(c)
(d)
Fig. 2.13 Modern field-emission SEM performance. (a) Shows the Hitachi S-5200. The column of the newer S-5500 is entirely contained in a rectangular box that provides shielding from sound and electromagnetic fields. (b) Both instruments offer a variety of SE and BSE detectors optimized for use at low and high beam voltage. A set of coils (ExB) mounted above the specimen, provides a magnetic field perpendicular to the electron-optical axis that compensates for the displacement produced by the electrostatic SE-collection field of the Everhart-Thornley detector. (c) A highmagnification image of a gold-on-solid-carbon substrate made at 1 kV using a Hitachi 5200. (d) An image of a similar specimen made at a beam voltage of 500v using a Hitachi S-4800 SEM. The S-4800 has a normal SEM specimen chamber mounted entirely below the objective lens. (All the images in this figure were generously provided by Hideo Naito and Vinh Van Ngo on behalf of Hitachi Inc, Pleasanton, CA.)
56
J. B. Pawley
vs 5 × 10−9 pascal), but the significance of this difference is somewhat offset by the fact that the TF gun operates hot, and the increased outgassing associated with this fact makes the vacuum conditions for stable TF emission almost as difficult to meet as those for FE. On theoretical grounds, performance should be quite similar for beam currents around 10−9 A, with TF being more suited to higher currents and FE capable of a smaller probe size when smaller currents are sufficient. No critical comparisons of actual, comparable, high-resolution, TF and FE instruments have yet been published, but one difference that may be important is total tip current. TF operates with a total current from the tip (Ig ) of 100–500 µA, while FE usually operates at Ig = 5–10 µA. The charge-charge interactions within the beam that cause increases in the effective V0 (Boersch 1954; Pfeiffer 1972; Barth et al. 1990) are proportional to J2g , where Jg refers to the current density in the first crossover. If the gun is designed and operated in such a way that Jg is higher in the TF system, this will reduce performance at low-kV where V0 has such a strong effect on beam diameter (Shao & Crewe 1987). Aside from different guns, modern, high-resolution FE-SEMs have important differences in the design of their electron optics, as well as that of their vacuum systems, specimen stages, and digital control/display electronics. The author’s experience is limited to the S-900, which was the first to be introduced and the following discussion will reflect this bias. The S-900 at the Integrated Microscopy Resource was the first commercial, inlens, FE-LVSEM to be installed anywhere in the world. Because it permitted one to obtain ∼5x better resolution on biological specimens than had been possible with any previous machines, our early work revealed a number of problems, both in the operation of the instrument and in the preparation of specimens. Although much of this chapter concentrates on these problems, the reader is reminded that almost all of the instrumental problems have now been corrected on later microscopes. Likewise, the improvements in specimen preparation described here and in the other chapters allow one to reach ever closer to the potential of the technique for revealing the three-dimensional intricacies of biological structure.
The SE Performance of Early FE-SEMs at Low Vo The S-900 at the Integrated Microscopy Resource was the first in-lens, cold FE-SEM to be delivered. Consequently, a number of modifications were needed to optimize its LVSEM performance (Pawley 1990). At first, problems with the vacuum system had the effect that a surface viewed at room temperature was often obscured by contamination after a single scan. Figure 2.14 shows two images of the surface of a specimen made by depositing fibrinogen molecules onto a carbon foil attached to an EM grid. Figure 2.14b was recorded first at 100 kx, 1.5 kV, and then Fig. 2.14a was recorded at 50 kx, and the result was magnified x2. Only the top left corner of the field of view is shown, but one can easily identify the three blobs present in both figures and also see how much clearer the image looks away from the area first scanned at 100 kx.
2 LVSEM for Biology
57
Fig. 2.14 Contamination obscures macromolecular structure. These two images were taken of the same area of a specimen made by depositing fibrinogen molecules onto a thin carbon film on a EM grid. After freeze-drying and coating, the specimen was imaged at 100 kx, 1.5 kV in the Hitachi S-900, before the vacuum modifications noted in Fig. 2.16. Immediately afterwards, a second 100 sec scan was recorded at 50 kx. (a) Shows a corner of the second image. The lighter area in the lower right corner is the area covered with contamination during the first scan. (b) Shows the small area of the first 100 kx scan that covered the same area as the box in (a). Although the darker area of (a) provides a noisier image (because the same 100 s exposure time was used to cover a 4x larger area of the specimen), the contrast of the fibrinogen molecules is higher because they have not been so covered with contamination
In terms of the vacuum environment, it is helpful to remember that materials present at a partial pressure of 10−4 Pa will deposit approximately one monolayer per second onto any adjacent solid surface. Even though microscope columns commonly operate at a much higher vacuum, this is not always true of the airlock vacuum. A rough-pumping oil with a vapor pressure of only 10−7 Pa will still coat all airlock surfaces in an hour or two. To solve this problem, we changed over to oil-free vacuum pumping throughout the system, as is shown in Fig. 2.15. The mechanical rough pumps were replaced with oil-free molecular-drag pumps (Danielson 1987), backed by diaphragm pumps, and an additional Gatan anti-contaminator was added. The sliding O-ring that sealed the side-entry stage rod to the inner side of the differentially-pumped airlock region was replaced with an oil-free seal made of spring-loaded teflon. Other changes involved efforts to reduce the effect of internal mains-frequency stray magnetic field (Pawley 1987a) and modifications to the control circuits to simplify the alignment process. As noted above, the side-mounted, Everhart-Thornley SE detector produces a collection field that displaces the probing beam before it reaches the final lens (Zach & Rose 1986) (see Fig. 2.10). As a result, the beam enters the lens field off axis (Pawley 1990). Because the displacement is proportional to Vo , realignment is required more often and this process is assisted if the controls for aligning the effective axis of the stigmator are readily accessible and if all the lens and stigmator currents can be wobbled. Many of these minor changes are now standard features on present commercial instruments. Once they are implemented, an image similar to that in Fig. 2.13b
58
Gatan anticontaminator -180° C
60 l/s
nose piece
AIRLOCK to pump
20 l/s isolation valve
20 l/s PM
T
stage rod
ion guage Original, Viton O-ring replaced with teflon seal. LN2 control tc
buffer 340 l /s turbo
10-4
mBar
MDPI
M D II P
2-stage diaphragm pump
buffer
S900 VACUUM MODIFICATIONS
<75 mBar
J. B. Pawley
Fig. 2.15 Modifications to the original vacuum system of the Hitachi S-900 as implemented on the instrument at the IMR in Madison. The two oil-filled rough pumps were replaced by molecular-drag pumps (Danielson 1987) backed by motor-driven diaphragm pumps. A Gatan double-blade anti-contaminator was added and the o-ring on the inner side of the differentially-pumped airlock was replaced with a spring-loaded, teflon seal so that the outer surface of the specimen rod no longer had to be greased
2 LVSEM for Biology
59
can be rapidly and routinely obtained. The specimen is Pt sputtered onto a solid, degassed carbon rod. Even though the fine contrast reflects only local density differences rather than topography, it does give useful confirmation of a beam diameter of ∼1.6 –3 nm at 1.5 kV (depending on the instrument). This is in accordance with the performance predicted by a more accurate theoretical analysis of the exact interactions between diffraction and chromatic and spherical aberrations (Crewe 1985; Shao & Crewe 1988, 1989). The S-900H employed a lens with lower aberration coefficients and reduced specimen motion (Sato et al. 1990; Osumi et al. 1990) as long as the specimen was in the upper, or UHR, position. These changes permitted a further reduction in spot size of ∼40%. In recent years, both JEOL and Hitachi have introduced more advanced lowvoltage, cold-FE SEMs. These instruments offer low-kV resolution performance that is almost twice as good as that obtained from the S-900 (1.6 vs 3 nm). Other manufacturers have offered heated FE sources that have lower vacuum requirements, and provide greater current stability and higher maximum beam current, at the cost of slightly lower source brightness and greater energy spread. With any of these modern instruments, it now seems safe to say that useful resolution of biological specimens is more likely to be limited by imperfections in specimen preparation (as discussed below) than by instrumental performance.
Other Limitations on LVSEM Performance The results just shown demonstrate that, in a modern high-resolution SEM, a probe diameter of <3 nm can be obtained at any V0 above 1.5 kV. Although the ultimate probe size is always inversely proportional to kV, in biology considerations other than probe size usually limit useful performance at about this level in any case. These practical considerations include factors such as the progressive loss of image contrast caused by the increase in the interaction volume if V0 is increased to reduce probe diameter, and the necessity of coating specimens with more metal to prevent them from charging up when struck by the beam. In the case of biological LVSEM, these considerations also include biophysical factors related to the necessity of maintaining the structural integrity of the specimen in the vacuum of the microscope. This usually requires that the specimen be subjected to chemical fixation, dehydration, and critical-point drying (CPD) (Ris 1985). Freezing followed by freeze-drying (Pawley & Ris 1987) and even the direct observation of the frozen specimen are real alternatives (Pawley et al. 1991). Unfortunately, as will be discussed towards the end of the next section, only the last of these methods has been shown effective for preserving structure to a resolution of less than 3 nm. Of course, it is not surprising that chemical fixation destroys structure at this level: glutaraldehyde does cross-link the free amino groups of polypeptides and amino-lipids (Johnson 1985a, 1985b), thereby inactivating enzymes and eventually killing the cell. This clearly indicates a change in structure at some level. Dehydration is even more damaging, because the shape of any biomacromolecule or a membrane is produced and maintained by its interaction with water.
60
J. B. Pawley
Removing the water destroys the ionic, hydrophobic, and hydrogen bonds that perform this function. Naturally, the conformation of these structures is severely affected. Indeed, Boyde and Maconnachie (1979, 1981) found that 60% volume shrinkage was common when preparing soft tissues for the SEM. Probably the best we can hope for in fixed and dehydrated biological specimens is to preserve the large macromolecules, and the structures assembled from them or covalently attached to them (Ris 1985, 1988, 1990, 1991). In 1985, Hans Ris published the definitive study on CPD, showing the deleterious effects of allowing even very small amounts of water to be present in either the final intermediate liquid (usually ethanol or acetone, but sometimes amyl acetate) or the liquid CO2 transition liquid (Ris 1985). Unlike many who had previously evaluated the performance of CPD for SEM specimens, Ris used the high-voltage electron microscope (HVEM) to view whole-mount cells that had been grown on thin films supported by EM grids. As a result, he was less interested in the appearance of the outer surface of the cell,11 and more interested in its internal structure—particularly the structure of its cytoskeletal elements such as microtubules, actin filaments, and intermediate filaments. In a careful series of experiments on a wide variety of structural proteins (actin, tubulin, keratin, chromatin, etc.), both in purified preparations and as they appear in cells, Ris has able to demonstrate that the trabelular appearance of the cytoskeleton noted in previous HVEM publications (particularly, the so-called thick-thin filaments that more resembled stretched chewing gum than they did rigid structural elements) was not seen if the intermediate and transition liquids were dried with molecular sieve and if the CO2 in the CPD bomb was mechanically agitated during processing. The HVEM stereo images in his paper confirm that, in properly-prepared whole-mount cells, the appearance of the cytoskeleton is very similar to that seen in thin sections—in particular, that the structural elements have a uniform thickness along their length and that clusters of ribosomes are clearly identifiable (often located at filament junctions). This TEM study is relevant to modern LVSEM because the practical resolution of these two types of instruments is now very similar, and the high-resolution structural information from both is more likely to be limited by artifacts associated with fixation, dehydration, and staining than by the imaging performance of the microscopes themselves. Particularly when the SEM is used to study the internal structure of cells, it is important that specimens be prepared using only optimal techniques, and in particular, that every effort be made to ensure that all the water is removed from the intermediate and transitional liquids used for CPD. As saturated lipids lacking a free amino group are not fixed by either glutarladehyde or OsO4 , it is perhaps not surprising that membranes seem to be especially fragile (Langford & Coggeshall 1980; Erlandsen et al. 1989a; Pawley & Erlandsen 1989; Johnson 1985a, 1985b. See also Fig. 2.25 in the Results section.). Although some of the deleterious effects of fixation and dehydration are readily apparent (holes in membranes, etc. [Erlandsen et al. 1989a]), electron diffraction
11 The cell surface could be seen in the HVEM, when necessary, by coating the specimen surface with a thin metal coat.
2 LVSEM for Biology
61
studies on protein crystals show that fixation and dehydration destroy virtually all native structure below 2–3 nm. Three other considerations should be mentioned when discussing useful SEM resolution: coating, contamination, and delocalization.
Coating A practical limitation that effects all insulating specimens is that, unless they are somehow rendered conductive, they charge up when irradiated by the probing electron beam and, if the resulting electric field is allowed to extend into the space above the specimen surface, the SE collection efficiency is affected and the image becomes unstable (Pawley 1972; Shaffner & Hearle 1976; Brunner & Schmid 1986; Joy 1987). Indeed, the thin metal coatings applied to reduce the effects of charging on insulators do not actually prevent injected electrons from being trapped in the bulk of the insulator. Although they do bleed away some of them, their main effect is to shield the subsurface electrostatic field created by this trapped charge from reaching the space above the specimen where it would interfere with secondary electron collection. Dried biological specimens are particularly troublesome, because they are not only extremely good insulators, but their low atomic number and low density (0.2 g/cm3 ) make them inefficient at scattering the electron beam to produce SE or BSE signals. Because these signals represent a departure of negative charge, their low magnitude exacerbates the build-up of negative charge. Consequently, with few exceptions (Welter & Coates 1974; Arro et al. 1981; Osumi et al. 1988b), SEM specimens are normally either coated with heavy metal (Echlin 1991; Adachi et al. 1976; Ingram et al. 1976; Peters, 1979, 1980, 1986a, 1986b; Wildhaber et al. 1985; Winkler et al. 1985; Lindroth & Sundgren 1989; Wepf & Gross 1990; Ogura et al. 1989b) or especially treated with heavy metal compounds (Ohtsuka et al. 1981; Kelley et al. 1973; Munger & Mumaw 1976; Irino et al. 1978; Kubotsu & Ueda 1980; Murphy 1978, 1980; Murakami & Jones 1980; Tanaka 1980, 1981) to increase their density and conductivity before being viewed in the SEM.12 Although such chemical treatments have produced some truly striking results (Tanaka 1990) they usually destroy at least microfilaments, decorate membranes with heavy metal compounds, and are suspected of destroying or at least severely modifying other structural features on the finest scale. As a result, they will not be discussed further here, although they may warrant future study. The necessity of coating means that the question of biologically-relevant resolution becomes inextricably bound to a consideration of coating and coating methods. 12 Although environmental SEM (ESEM) seems inappropriate for high-resolution studies of biological specimens, the technique does have advantages when viewing topographically-complex insulating specimens (an extreme example might be fiberglass batting, which is very difficult to coat effectively.). In this case, the remnant gas in the chamber becomes ionized by the electron beam and provides a source of positive charge capable of neutralizing surface charging. In short, ESEM is another solution to viewing insulators in the SEM.
62
J. B. Pawley
A thin (∼1 nm) metal coating does indeed increase the amount of signal produced near the point of beam impact. Although not continuous, such a film also provides sufficient conductivity to prevent the negative charge trapped below the surface from displacing or defocusing the beam or from interfering with the collection field of the SE detector (Pawley 1972). It is hard to imagine a method for coating a topologically diverse surface in a uniform and continuous manner, however, and there is always a danger that any metal coating will preferentially decorate those surface-active features of the specimen that can serve as nucleating sites (Braten 1978; Winkler et al. 1985; Bachmann et al. 1985; Wepf et al. 1991) while obscuring others. The problem, then, becomes one of optimizing the coating thickness to permit the production of good images while obscuring as little structure as possible. A consideration of the effects of coating and V0 on resolution is simplified by separately considering the two chief purposes of coating: suppressing charging and producing image contrast.
Avoidance of Charging Charge accumulation occurs because, on flat specimens viewed at 5–20 kV, more electrons enter the surface than leave it. On such samples, the sum of the SE and BSE coefficients increases as V0 is reduced from 20 kV, and goes slightly above unity at 2–4 kV. This has led many to believe that charging artifacts should disappear at low Vo . Although this can be true for relatively flat, insulating specimens (such as passivated semiconductors), where the SE and BSE coefficients are fairly constant over the entire sample (Sugiyama et al. 1988), it is less true for geometrically complex biological samples.13 Although (averaged over the entire viewing area) the current absorbed by these materials may be zero, contrast in the SE image is intrinsically a reflection of marked variations in the effective SE coefficient within the scanned area. Consequently, contrast in the SE signal implies local variations in the SE coefficient and therefore local variations in local net charge deposition. In addition to charge accumulation in the area covered by the scanned raster, adjacent parts of rough samples may become charged by scattered electrons that strike surfaces away from the beam impact point. It is the local areas of charge imbalance produced by these mechanisms that produce charging artifacts on rough specimens at low Vo . Paradoxically, at higher SEM voltages (3–15 kV) where, from a straightforward analysis of total electron yield vs voltage, net charge accumulation should be more severe, it is sometimes found that charging artifacts on marginally-coated, topologically complex, biological samples are actually less serious, (Fig. 2.16), especially if a relatively thin layer of biological material is supported on a conductor such as silicon or metal. The simplest explanation for this apparent anomaly is beaminduced conductivity. Energy deposition by the beam produces sufficient free carriers within the specimen to carry small horizontal and vertical currents by drifting in
13
No matter what the beam voltage, the SE coefficient of a deep narrow hole is zero.
2 LVSEM for Biology
63
Fig. 2.16 Three micrographs made at 1, 1.4 and 2 kV, of a specimen of freeze-dried sperm cells. As many of the sperm form a loose matrix suspended above the substrate, the thin Pt coating was insufficiently conductive to prevent charging. However, one can easily see that the SE-signalcollection artifacts are less severe when V0 (NOTE: the 0 should be subscript) was 2 kV than when it was 1 or 1.5 kV. Having more charging at lower beam energy is unusual and may be related to the fact that the higher beam energy produces more SE emission from the far side of the sperm tails, increasing the effective SE coefficient. Alternatively, it may be due to the higher beaminduced conductivity characteristic of the more energetic beam electrons. These images constitute yet another example of the fact that specimen charging on topologically complex surfaces can be very different from that deduced from experiments on polished, flat surfaces
the electric fields produced by small, local charge inhomogeneities (Catalano 1976; McMullin 1976; Jakubowicz 1987; Leamy et al. 1978). As beam-induced conductivity extends only throughout the volume of beam interaction, it is more effective at higher V0 where the interaction volume is much larger and where, in addition, more transient charge carriers are produced by each incoming electron. Even allowing for these exceptions, it is still true that, on most lightly-coated specimens, charging artifacts are less severe at around 1.5 kV (see Fig. 2.17). They are also reduced by high scan speed (Welter & Coates 1974) and this is one of the prime advantages to using a digital image frame store to integrate signal from
64
J. B. Pawley
Fig. 2.17 Surface of diatom lightly (but insufficiently) coated with Pt at 1.5 kV (a) and 5 kV (b), showing reduced charging and increased contrast at 1.5 kV. Coating of spherical objects is complicated by the need for the metal to cover the surfaces on the bottom of the sphere where it contacts the substrate. This problem can be reduced by using substrates that are both conductive and not covered with a thick oxide layer, such as Au-coated glass or Si chips
a number of noisy fast-scan images to produce a final, noise-free image (Yamada et al. 1991; Martin et al. 1985; LeFloch et al. 1987; Morandi et al. 1989). It is clear that fast scanning does not actually reduce charging, it only stabilizes the effect of the charge distribution. If video-rate charging is occurring, one will notice that the average image brightness changes abruptly and then slowly recovers in response to a large-step change in magnification.
Image Contrast Apart from reducing the effects of charging on the image, the higher density of the metal coating produces more intense scattering of the beam at the specimen surface. Uncoated organic specimens really do not give much topographic signal at normal kV (see Fig. 2.18). The effective thickness of a metal coating is its thickness in the direction parallel to the beam. Specimen topography modulates the effective thickness of a uniform metal coating because protuberances catch more metal than hollows. This modulation in thickness in turn affects the amount of scattering and hence produces contrast in the SE or BSE signal. This contrast is referred to as topographic Z contrast (Joy 1991a, 1991b) and it is related to surface angle only to the extent that the latter determines the effective coating thickness. At 1 kV, useful topographic images can be produced from samples that have been coated only with carbon—a material that produces no topographic Z contrast when applied to biolog-
2 LVSEM for Biology
65
Fig. 2.18 Effect of V0 on Topographic contrast when imaging uncoated, low atomic-weight specimens. Trichocysts (fibrous structures extruded from the surface of paramecia) tend to attach and spread on a fractured carbon substrate. While at 4kV, the image of such a specimen resembles that from a coated specimen, contrast drops steadily as V0 and the size of the penetration volume increases. These images emphasize that most of the contrast seen on low-atomic weight specimens viewed at high V0 is due to changes in the effective thickness of the metal coating
ical specimens (Pawley & Albrecht 1988) (see Fig. 2.6), although the conductivity or amorphous carbon is so low that relatively thick coatings are needed to reduce charging. At high V0 , however, image contrast is determined almost entirely by the effective thickness of the metal coating (Broers et al. 1975; Arro et al. 1981; Broers 1982; Joy 1987, 1991b, 1991a) (see Fig. 2.18). A considerable body of work on the effects of different types of coating on high-resolution SEM image contrast was carried out by Peters (Peters 1980, 1982, 1985, 1986a, 1986b, 1988; Peters & Fox 1990). This work led these workers to recommend that, in the SEM, fine details (<5 nm) can best be seen on biological samples if such samples are coated with uniform films of Ta or Cr deposited by Penning sputtering to an estimated thickness of 1–2 nm, and viewed using a V0 of 30–40 kV. Similarly, Bell and coworkers used a TEM/SEM at 160 kV on very thin specimens (culture cells growing on EM grids), coated with Pt, W, and Ta by DC-ion sputtering (Bell et al. 1989; Lindroth et al. 1988; Lindroth & Sundgren 1989).
66
J. B. Pawley
The reasoning was that the very thin coating tended to reduce the production of BSE and hence reduced the emission of secondary electrons related to collisions occurring away from the beam impact point where these BSE emerge. Because the SE signal produced by the scattering of the beam in this thin coating is very small at high V0 , and because there is no way to totally eliminate the diffuse SE background signal produced by the emerging BSE (or more accurately, the statistical noise associated with this diffuse signal [Wells,1975, 1978]), the contrast of small features is very low (approximately 1%). This low contrast can only be overcome by using a large Io to reduce the statistical noise. The images shown in support of this theoretical interpretation are usually of samples with relatively simple topography but they do indeed show small topographic details—the contrast of which may be interpreted as being consistent with the theory (Peters 1988, 1989)—and in later years, this work was carried on by Apkarian (Apkarian et al. 2003). Nonetheless, this theoretical explanation still contains several approximations that may make it less than the final word on the subject, and this is especially true when the theory is applied to SEM samples that have a more complex topography and that are radiation sensitive. The complex topography may make it hard to avoid charging artifacts if the visible parts of the specimen have only a very thin coating and the high I0 needed to overcome the low contrast implies high radiation damage. For instance, it may be possible to imagine depositing a uniform 1–2 nm coat of chromium (Cr) or tantalum (Ta) onto the surface of a topologically simple specimen such as an antibody on the surface of a red blood cell supported by a silicon substrate—but how uniform or stable would such a film be in reality? Even if decoration (Winkler et al. 1985; Waltzthony et al. 1981) were ignored, what is the range of thickness from place to place caused by changes in the local surface angle, by the shadowing of adjacent structures, and even by stochastic considerations applied to the small number of atoms (5–10 thick) needed to make up a 1–2 nm film? These unavoidable and uncontrollable variations in coating thickness would also give rise to topographic Z contrast, and the situation would be much more severe on specimens of high topographic diversity such as those shown later in this chapter. To begin with, it is almost certain that the outer monolayers of any thin film composed of these metals would be oxidized by exposure to air or even to the vacuum of a normal SEM (Joy 1991a). How much actual metal is left after this has happened? Because platinum or tungsten films, deposited similarly, form 1–2 nm microcrystals surrounded by 1 nm spaces, it is hard to see why Cr or Ta films do not do likewise if they are really metals rather than oxides. Of course, oxides can have a high SE coefficient and the metal atoms in them will scatter whether they are part of an oxide or not. Because chromium oxide is only one third as dense as chromium, however, the film may physically be up to three times thicker than expected from the reading on the quartz film-thickness monitor. In addition, conductivity in a metal depends on the overlap of the valence states between the neighboring atoms in a lattice. If this overlap does not occur, either because of oxide formation or because the metal is not crystalline, the electrical conductivity of the film is severely reduced.
2 LVSEM for Biology
67
This is not an exhaustive list of questions that can be raised regarding the highbeam voltage/thin, uniform film analysis and they are raised here not because I plan to answer them (indeed I am unsure whether answers accurate enough to describe the situation on a surface of even modest complexity are possible) but because I feel that, to some extent, such a detailed theoretical consideration of contrast mechanisms in the face of so many unknowns may tend to distract interest from other more important features of the SEM image. There is a danger in deriving the optimum conditions for revealing small surface structures from the analysis of topographically simple surfaces that are, in fact, unrepresentative of the types of specimens that might be most profitably viewed in the SEM. It is not necessarily true that the main goal of high-resolution SEM, at whatever voltage, should be to mimic the capabilities of the best TEM replica techniques to see ever-smaller features, such as the subunits of large proteins or intramembrane particles on relatively flat surfaces (Heuser 1979). In fact, because ionizing radiation damages the structure of biological materials much more than it does the structure of amorphous carbon replicas, and because a fresh, cryofractured surface is far cleaner (in terms of contamination) than any room-temperature SEM sample, it seems unlikely that such an effort will be entirely successful. Even if it were successful, the SEM procedure would probably be far more difficult and timeconsuming than the TEM alternative because the SEM produces high-resolution images much more slowly than the TEM (100 sec/frame in SEM vs 0.1 sec/frame in TEM). On the other hand, there is a case to be made that the three most important characteristics of the SEM as a tool for the study of biological morphology are: • SEM images at both low and high magnification are available to the observer directly, in the form of easily interpretable images of the three-dimensional surfaces of complex solids (Hayes,1973, 1980), especially when they are viewed in stereo • The SEM specimen can be large, continuous, and topologically complex. As a result, microstructures can easily be related to features on a larger scale— something that is seldom possible when viewing either a series of serial sections or the disjointed and fragmentary remains of the freeze-fracture replica process. An example might be the surface of an early stage sea urchin embryo, as seen in Fig. 2.19 (Holy et al. 1991) • Unlike the TEM replica, the surface of the SEM specimen can be treated directly with autoradiographic emulsion (Salpeter et al. 1988) or gold-conjugated labels (Faulk & Taylor 1971; Albrecht & Hodges 1988) to identify the location of specific molecules of biological interest with a spatial and chemical specificity that is limited only by the size of the label and the precision of the specimen preparation procedures. In short, I see the most profitable realm of application for high-resolution SEM as being in the imaging of surfaces such as those shown in the remainder of this chapter.
68
J. B. Pawley
Fig. 2.19 Surface of two cells in a critical-point dried sea urchin embryo. At certain cell stages, the surface of these cells is highly convoluted. It is hard to imagine being able to make a replica of such a surface that could be viewed intact in the TEM. Specimen prepared by Gerard Coffe (IMR University of Wisconsin-Madison). The preparation of this specimen is discussed in more detail in Chapter 5
In our work, we have preferred to use thin films of ion-beam-sputtered Pt or tungsten (W) (Evans & Franks 1981; Franks et al. 1980; Kemmenoe & Bullock 1983) over those made by glow-discharge sputtering or vacuum evaporation for a number of reasons: 1) Compared to diode sputtering, it is performed in a higher vacuum and is therefore less likely to involve the inadvertent deposition of organic oils onto the specimen surface. 2) Because both the current and the voltage of the ion beam can be controlled independently, coating protocols are more reproducible. 3) By using only Pt targets, we were able to avoid the build-up of thick oxide films on their surface when the chamber was opened. As a result, there is no oxide layer to be sputtered off and useful Pt sputtering occurs only a few seconds after the ion beam is turned on. Furthermore, the coatings degraded more slowly than those made with more reactive metals. 4) It is very difficult to ensure that the Ar glow discharge contains no reactive oxygen. Even if room air is carefully pumped from the system with a high-vacuum pump and the chamber then flushed with Ar before coating, the discharge itself involves the deposition of considerable power (2–20 watts). Not only can this power affect the specimen directly, it can also heat chamber components, causing outgassing. The water vapor that is the most common component of such outgassing is rapidly degraded by the discharge to produce various reactive species,
2 LVSEM for Biology
69
including oxygen. The early literature of discharge coating for SEM included many examples involving the destruction of delicate surface features. Although modern equipment includes improvements to reduce these effects (better vacuum, cleaner gases, cooled components etc.), it seems foolish to take unnecessary risks. 5) Thermal evaporation of Pt inevitably heats the specimen. When such films are deposited onto a carbon foil mounted on a TEM grid and imaged directly at 20 kV in BSE (see Fig. 2.20), they are seen to be composed of 1–2 nm particles with 1–2 nm spaces between them. Because a 1nm Pt particle has only four atoms on an edge, a significant reduction in particle size seems unrealistic given relevant forces such as surface tension. On the other hand, even though it is clearly discontinuous, such a film seems to provide sufficient electrical conductivity to prevent charging on most specimens as long as they are mounted on a conductor (such as a chip of polished Si) and observed at low Vo (see Fig. 2.21).
Contamination Because the SE signal is more characteristic of the near-surface region at low V0 , hydrocarbon contamination layers deposited on a lightly-coated specimen are far more visible under these conditions, and are also deposited more efficiently per beam electron (see Fig. 2.22), a fact noted by Thornley as discussed in Chapter 1. Contamination in the SEM is a complex matter (Wall 1980; Fourie 1981; Hren 1986). As discussed earlier, hydrocarbons can originate from an imperfect vacuum environment. They can also come from the specimen itself. Specimen-born contamination consists of that present on the specimen when it is introduced into the microscope, that which it may pick up in an improperly pumped airlock, and that produced by the beam through radiation degradation of the specimen (Wall 1980). The first can be reduced simply by keeping the specimen clean. Biological specimens that have been dried using CPD are rinsed clean of low molecular weight hydrocarbons as they pass through the successive changes of the intermediate and transition fluids that are used as part of the drying process. All that is required is to avoid hydrocarbon deposition during coating, storage, or insertion. This can be done by carrying out coating procedures in a good, oil-free vacuum, by storing specimens in clean, dry, glass or metal containers, and by ensuring that the specimen airlock is pumped by in an oil-free manner (i.e., using a molecular-drag or a turbomolecular pump). While contamination originating from beam-specimen interactions cannot be totally eliminated, it can be reduced by minimizing the amount of electron radiation delivered to the organic part of the specimen. On specimens thick enough to absorb the total energy of the beam, this suggests the use of low V0 . As a practical matter, contamination no longer seems to be a limitation on our use of the modified S-900 at 1.5 kV on CPD specimens.
70
J. B. Pawley
Fig. 2.20 Structure of a 1–2 nm thick, ion-beam-sputtered Pt coating. The structure of this Pt film is clearly visible in this high-V0 BSE image. The background noise caused by subsurface scattering was not a factor because the film was deposited on a thin carbon substrate. This image also shows that one can still easily recognize 5 nm colloidal-gold particles (larger white blobs towards the top of the picture) in the presence of such a coating
2 LVSEM for Biology
71
Fig. 2.21 Contamination rasters deposited on a Pt-coated silicon substrate using rapid scan (left) or slow scan (right) raster speeds and deposited at either 1 kV (center) or 5 kV (lower). Each patch shows a nested set of three, 100-sec rasters made at 100 kx, 50 kx and 25 kx. The upper two rasters were deposited at 1 kV after the specimen had been heated enough to increase surface mobility (making contamination worse) but not enough to out-gas the hydrocarbons from the surface
Fig. 2.22 Specimen rod insert designed to hold in place tissue-culture cells, attached to Si or glass chips, using a BeCu wire spring bearing on the top surface. The advantage of this approach is that the BeCu wire is brought into contact with the coated face of the specimen substrate. If the substrate is glass, this arrangement reduces the chance of charging artifacts because the top surface is likely to have a thicker metal coating than the edges or bottom
72
J. B. Pawley
On freeze-dried specimens viewed at room temperature, the problem is much more severe (Hermann et al. 1988) because this drying procedure leaves in place any low molecular weight molecules present at the time that the specimen was quenched. This is a serious problem because freeze-drying is widely accepted as the process most likely to preserve fine structure in SEM specimens (Wepf & Gross 1990; Boyde & Maconnachie 1979, 1981; Pawley & Ris 1987; Hermann et al. 1988). The only solution is to use a cold stage to keep the freeze-dried specimen cold while it is under observation (see Fig. 2.7). Even better, is the process pioneered by Martin Müller, Roger Wepf, and Paul Walther at the Eidgenössische Technische Hochschule (ETH) in Zurich that starts with prolonged freeze-drying carried out at very low temperature (−120 ◦ C). The specimen is then metal-coated and viewed cold, without ever being warmed above −100 C. This preserves the very fine structural details that depend on the presence of bound water. This was the technique used at the Integrated Microscopy Resource (IMR) by Ya Chen to produce Fig. 2.7.
Delocalization On a more theoretical level, it has been shown that the production of a secondary electron is a delocalized process. Secondary electrons can be excited from a site as much as a nanometer or two from the nearest primary electron trajectory (Isaacson & Langmore,1974). This occurs because every electron in the specimen near the incoming primary experiences a pulse of repulsive force as it passes. Specimen electrons will be ionized as long as the energy of this pulse exceeds their binding energy. As this delocalization is more pronounced at higher voltage, Albert Crewe has stated that the optimal voltage for topographic SEM using the SE image should be chosen as a trade-off between this effect and simple electron optics, and will probably be found at about 5 to 7 kV (Crewe 1985). Although this analysis has been disputed at least with respect to crystalline, metal specimens where phonon scattering is possible (Cowley 1990; Liu & Cowley,1988), it seems likely that delocalization does indeed place a limit on the ultimate resolution obtainable on noncrystalline specimens viewed in the SE mode (Joy 1991a). As BSE signal production is an elastic process, it is not subject to this limitation, so it is possible that the highest resolution imaging may involve low V0 BSE (Joy 1991a).
The Multi-Factor Approach From the foregoing, it seems clear that choosing the optimal V0 for high-resolution SEM of real biological specimens requires considering variables in addition to the ability to produce a small probe. Indeed, because a probe sufficiently small to visualize 3 nm features can be made at any V0 above 1.5 kV, probe size per se is seldom a variable at all. The effect of the other variables can perhaps best be considered with reference to Fig. 2.23, which plots the general variation with V0 of a number of factors that
2 LVSEM for Biology
73
spatial resolution
Factors affecting: Topographic Resolution in the SEM
primary electron range = k(V0)5/3 coating thickness de-localization beam diam = k(V0).5 contamination
0 0
beam voltage = V0
Fig. 2.23 The variation with V0 of factors affecting practical topographic resolution in the SEM. Because these parameters have different slopes, there exists a V0 for which a minimum exists. When the Hitachi S-900 was used to image a variety of specimens, this minimum appears to occur at ∼1.5 kV
affect the final, useful resolution—primary electron range, radiation damage, coating thickness, contamination, and delocalization. No numerical values are shown on the V0 , or resolution scales, because these cannot be specified without both information regarding the geometry of the specimen, a definition of topographic resolution, and information about the radiation sensitivity of the specimen. Although there may be debate regarding the relative magnitudes of the effects diagrammed in Fig. 2.23, the signs and approximate magnitudes of the slopes of these curves are probably reasonable. Such a graph emphasizes the point that the optimal choice of V0 involves a tradeoff between all of these factors. In 1956, K.C.A. Smith made the same point with respect to the first useful SEM: “On the basis of the foregoing considerations, it may be concluded that in order to resolve detail of given dimensions on a particular specimen both the spot size and the accelerating voltage must be reduced to certain critical values. Since reduction of the accelerating voltage increases the minimum obtainable spot size, there will clearly be optimum conditions depending on the nature of the specimen.” (Smith 1956). Because instrumental performance has improved considerably since 1956, a wide variety of S-900 users at the IMR usually found the optimal V0 for SE imaging of lightly-coated, CPD, biological specimens, in the range 1.5–2.5 kV. It was also important that specimens mounted on pieces of glass coverslip or Si wafer were held in position on the specimen rod by a BeCu spring pushing down on the top (i.e., coated) surface of the glass (see Fig. 2.21). Now that instruments of similar and even higher performance are more widely available, others seem to have reached similar conclusions.
74
J. B. Pawley
Biological Applications of High-Resolution LVSEM Early LVSEM results Starting with the introduction of the low-aberration FE-SEMs in 1986, a number of early papers described the results obtained at low kV with these instruments. These applications ranged from specimens that were particularly sensitive to radiation damage, such as the cyst walls of Giardia (see Figs. 2.3 and 2.4; Erlandsen et al. 1989b; see also Chapter 8), to the delicate cell-surface structures found on the surfaces of sea urchin cells at certain stages of embryonic development (see Fig. 2.19). Now that the practical resolution of LVSEM rivals that of thin-section TEM, one of the major challenges facing anyone preparing LVSEM specimens is how to break the cells open in such a way that the fracture surface reveals the intracellular structures of interest. Such structures can be revealed by the freeze-fracture (Haggis & Pawley 1988) or the dry-fracture of tissue culture cells caused by touching the intact cell to the surface of adhesive tape (Ris 1988, 1989; Lim et al. 1987; Ris & Pawley 1989; see Fig. 2.24). Initially, we were surprised to find that as long as it had been critical point dried with great care to avoid the presence of any water in the final intermediate and transition liquids (Ris 1985), and then coated as described earlier, the cell surface often looked very different from that which we had been used to. Figure 2.25 shows the junction of two anastomosing GH3 pituitary cells. These cells had earlier been treated in a ball mill to make them permeable to macromolecules, and then allowed to recover (Martin 1989). The upper cell is assumed to be dead while that below appears the same as an untreated cell. In both cases, most of the lipids seem to have been removed from the membranes and one sees mainly the underlying fibrous cortex of membrane and cytoskeletal proteins.
Fig. 2.24 Dry-fractured tissue-culture cell. This tissue-culture cell was originally critical-point dried and coated with carbon by Hans Ris for whole-mount viewing in the high-voltage TEM. A section of the top surface of the cell has been removed by allowing it to adhere to the sticky side of a piece of pressure-sensitive tape. While being viewed using a dissecting light microscope. When this image is viewed in stereo, one can see the cell membrane coming out of the plane of the page from the left and elements of the cytoskeleton to the right. This image was one of the first made of a biological specimen using S-900 prototype at the Hitachi factory in Naga, Japan in 1986
2 LVSEM for Biology
75
Fig. 2.25 The surface of two GH3 pituitary cells. These cells had earlier been treated in a ball mill to make them permeable to macromolecules (Martin 1989). They were then allowed to recover, fixed, critical-point dried, coated with ion-beam-sputtered Pt and viewed in the Hitachi S-900. The upper cell is assumed to be dead while the lower one appears similar to an untreated cell. In both cases, most of the lipids seem to have been removed from the membranes and one sees mainly the underlying fibrous cortex of membrane and cytoskeletal proteins
Other early specimens include mitotic apparatuses isolated from a sea urchin embryo by Gerard Coffe, (see Fig. 2.26; Thompson-Coffe 1996). The preparation of these specimens is discussed at greater length in Chapter 5), and various types of fecal bacteria found in conjunction with Stan Erlandsen’s studies of Giardia (see Fig. 2.27; Erlandsen et al. 1990b, 1991, 2004; see also Chapter 8). Figure 2.28 shows a specimen in which the pattern of collagen deposition is visible in a pit excavated in the polished surface of a piece of sperm whale tooth dentine by an osteoclast (specimens provided by Dr. Alan Boyde, the pioneer in the use of LVSEM in biology. See also Chapter 7). Although in this specimen the banded nature of the collagen bundles was initially clearly visible, this could no longer be seen in specimens that had been stored in a desiccator for only a few weeks.
Nuclear Pore Complex The cell nucleus, containing the genetic material, is separated from the cytoplasm by a double membrane called the nuclear envelope. Nuclear pores are holes in the nuclear envelope that form where the two membranes fuse. They permit the passage of large macromolecules into, and out of, the nucleus.
76
J. B. Pawley
(a)
(b)
Fig. 2.26 Mitotic apparatuses isolated whole from dividing cells in a sea-urchin embryo. These structures were isolated and prepared by critical-point drying by Gerard Coffe and colleagues at the IMR, University of Wisconsin-Madison (Thompson-Coffe 1996. The preparation of these specimens is discussed in more detail in Chapter 5). One can easily see the difficulties inherent in attempting to cover such a complex surface with a thin metal coating of uniform thickness. By observing it at 1.5 kV, however, the average secondary electron coefficient is increased sufficiently to prevent noticeable charging artifacts even though the metal coat used was very thin and undoubtedly not fully uniform or continuous. Both of the mitotic apparatuses shown have been fractured to permit one to see structures in the vicinity of the centrosome, an area that can be recognized by the fact that many of the major microtubule bundles seem to radiate from it
2 LVSEM for Biology
77
Fig. 2.27 Bacteria from a fecal sample. (a) The important thing about this stereo image is that surface features are visible not only on the surface of the critical-point dried bacterium but also on the surface of the thin filaments suspending it above the substrate. Had the beam voltage beam much higher than 1.5kV, the beam would have penetrated these filaments making them appear uniformly bright. (b) This image emphasizes the LVSEM’s ability to couple high-resolution information about delicate structures within a context provided by low-magnification views. (Both specimens were prepared by Stanley Erlandsen, University of Minnesota.)
78
J. B. Pawley
Fig. 2.28 The pattern of collagen deposition in a pit excavated by an osteoclast. This specimen was provided by Dr. Alan Boyde (University of London, UK), a pioneer in the use of LVSEM in biology (Jones et al. 2004). It was prepared by plating living osteoclasts onto the polished surface of a piece of sperm whale tooth dentine and then removing them before critical-point drying the dentine for observation in the S-900. The image in the lower panel covers the area of the specimen shown in the center of the upper, low-magnification view. Although the banded nature of the collagen bundles is clearly visible in this specimen (lower panel), it was no longer visible in similar specimens that had been stored in a desiccator for a few weeks
2 LVSEM for Biology
79
In recent years, there has been much progress in the understanding of this important transport process (see reviews by Gerace & Burke [1988] and Roderick et al. [2006]). It has been shown that molecules smaller than about 9 nm can diffuse freely into, and out of, the nucleus. On the other hand, larger molecules are transported by an energy-requiring process selectively, unidirectionally, and in a manner controlled by the nuclear pore complex (NPC)—a complicated structure associated with each nuclear pore. Molecules that are transported through the nuclear pore must contain a specific signal sequence that is recognized by a receptor on the NPC, and binds to it, before being moved through the pore. Until the 1980s, it was believed that the NPC was symmetrical and consisted of two rings of spherical particles—one on each side of the nuclear envelope. This view was based on the TEM study of thin sections and isolated, negatively-stained, nuclear pores. In 1989, Hans Ris (Ris 1989, 1990, 1991, 1997; Ris & Malecki 1993) used the LVSEM to study the nuclear membrane from frog and newt oocytes because these nuclei are very large, can be isolated by hand, and, at certain stages, are almost completely covered with NPCs. After isolation, the nuclei were transferred to a glass carrier substrate, torn open with fine forceps, fixed, CPD and coated with a thin layer of Pt by argon ion-beam sputtering. Both the cytoplasmic part and the intranuclear part of the nuclear pore complex were then imaged directly, in stereo, by LVSEM. The cytoplasmic side resembles the image obtained by TEM after negative staining, except that the eight components of the cytoplasmic ring are now not spheres but short rods that are about twice as high as they are wide (see Fig. 2.29). The views of the intranuclear component were totally new (see Figs. 2.30 and 2.31). The inside of the NPC consists of a ring 120 nm in diameter, from which eight thin filaments project into the nuclear space with their ends attached to a smaller ring about 60 nm wide—forming a structure that resembles a fishtrap. This structure had never been imaged before, perhaps because it is too large to be contained in a single TEM thin section and, in negatively-stained preparations of isolated nuclear membranes, only the cytoplasmic side that attaches to the formvar film is enclosed in the negative stain. A layer of negative stain thick enough to include the entire nuclear pore complex is too thick to be penetrated by the 100 kV beam in the TEM (Ris 1997; Ris & Malecki 1993. See Chapter 5 for more information regarding the NPC and its optimal preparation for observation in the LVSEM). The structural features of the NPC are perfectly matched to the following capabilities of the high-resolution-LVSEM: • It is an important structure made of a number of polypeptides, each of which could, in principle, be labeled with colloidal-gold markers (Walther et al. 2001, 2002) • It is small enough to be beyond the capabilities of the normal SEM • It is large enough to be difficult to reconstruct from cryo-TEM images (Aebi et al. 1990; Roderick 2006) • Many of its structural features are still evident after being coated with the 1–2 nm discontinuous, ion-beam-sputtered Pt coating referred to earlier • Interpretation of the images is greatly facilitated by being able to view thousands of NPCs in the context of other, lower-magnification images covering almost the entire nuclear envelope
80
J. B. Pawley
Fig. 2.29 Cytoplasmic surface of the nuclear envelope, showing the outer component of the nuclear-pore complex (NPC) This image shows the outer surface of the nuclear envelope of the oocyte of the newt Notophthalmus viridescens. Under a dissecting microscope, Hans Ris removed the nucleus in low-salt buffer using fine forceps, punctured it while it was attached to a chip of carbon-coated coverglass. It was then fixed in 2% glutaraldehyde in LSB at pH 7.4 and containing 0.2% tannic acid, post-fixed in 0.1% osmium-tetroxide, washed in distilled water and then prepared for LVSEM by critical-point drying and ion-beam-sputtered Pt coating (Ris 1991, 1997). The nuclear envelope has become folded, allowing one to view the NPC structure from all angles and see that the eight subunits surrounding the hole are cylinders, not spheres as had been previously believed
Fig. 2.30 Nuclear surface of the nuclear envelope of the frog Xenopus Lavis, showing the “inside” of the nuclear-pore complex (NPC). This specimen was prepared and imaged by Hans Ris using techniques similar to those described for Fig. 2.29. It shows the fish-trap structure formed of eight fibers that connect the nuclear pore structure to a much smaller ring. Some of the rings show additional fibers leading from the far side of this small ring (Ris 1991, 1997). Field width = 810 nm. (Reprinted from Ris 1990 with kind permission of IOP Publishing Ltd. Bristol, UK.)
2 LVSEM for Biology
81
(a)
(b)
(c)
Fig. 2.31 Nuclear-pore complex (NPC) of the Xenopus oocyte prepared and imaged by Hans Ris (IMR, Madison) using procedures similar to those described for Fig. 2.29 and showing views from both the outside (a) and the inside (b and c) of the nuclear envelope. To reveal some of the elements of the lamin network, (c) shows a specimen that has been extracted with a solution of 0.1% Triton X100 detergent and 1% glutaraldehyde in LSB, prior to post fixation. Field width = 788 nm. (Reprinted from Ris 1990 with kind permission of IOP Publishing Ltd. Bristol, UK
82
J. B. Pawley
• Low kV observation is sufficiently benign to permit one to collect both members of a stereo-pair without creating so much damage that the specimen becomes distorted and the stereo pair becomes misleading LVSEM studies of this structure are still in progress, but it seems safe to say that the results so far have already produced a serious reevaluation of the earlier TEM and STEM results (Aebi et al. 1990; Ris 1991; Kiseleva et al. 2004; Roderick 2006). At the time of their discovery, the structural details of the NPC were probably smaller, by an order of magnitude, than any other biological structures in which SEM results had caused a similar reevaluation of previous TEM work.
Cryotechniques There is probably general agreement that, if ice crystal artifacts can be avoided, cryotechniques represent the preferred method of stabilizing biological tissues before viewing them in the vacuum of the electron microscope (Lepault et al. 1991). The problems associated with avoiding ice crystals have been widely studied (Robards & Sleytr 1985; Read & Jeffree 1991; Echlin 1992; Müller 1992; see also Journal of Microscopy 160-3 (1990) and 161-1 and 2 (1991), and Chapters 8 and 10). Initially, reliable methods did not exist for freezing specimens more than 10–15 µm thick (Sitte et al. 1987), but in the 1980s, the equipment for subjecting the specimen to a pressure of ∼2100 bar as it is frozen became more reliable (Studer et al. 1989) and more widely available. As a result, some specimens as large as 500 µm thick can now be frozen without cryoprotectants. As such, specimens are judged to be almost free from artifacts. Two groups (at the IMR in Madison, Wisconsin and under Martin Müller at the ETH in Zurich) took the lead in investigating the possibility of viewing the surfaces of such specimens directly, on a cold stage in the Hitachi S-900 (Malecki & Ris 1991; Malecki & Walther 1991, Müller 1992, Müller & Hermann 1990). Earlier attempts to use cryo-SEM to mimic the results obtainable from freezefracture (Echlin 1971; Pawley & Norton 1978; Pawley et al. 1978, 1980) were unsuccessful at high-resolution because of: • • • •
Insufficient image contrast at the high V0 used Contamination of the coated fracture surface by, condensable vapors Mechanical instability in available cold stages Gross radiation damage to the specimen (Talmon 1984), a situation exacerbated by low image contrast
The improvements in instrumentation discussed above permit a major reduction (∼100x) in the total radiation required to make an image, and the new microscopes also have much cleaner vacuum systems so contamination is no longer a serious problem (Walther et al. 1990a, 1990b). Finally, the side-entry eucentric goniometer stages employed on these instruments can accept high-stability, cryotransfer stage rods of the type normally used for TEM electron crystallography (Chiu
2 LVSEM for Biology
83
et al. 1988, 2002; Downing 1991; Li et al. 2002). This combination of equipment was employed in the study of biological tissue early on (Müller 1992; Pawley et al. 1991; Herter et al. 1991). Early examples are included here and more recent ones in Chapters 8 and 10. Initially, the absence of a conventional method of coating the frozen fracture surface meant that our first cryo-studies were necessarily made on uncoated fracture surface using the SE mode at about 1.5 kV to reduce charging. More recently, we have moved to a slightly higher voltage (2.4–4 kV) so that we could obtain a useable image using the Autrata BSE detector discussed earlier. Although serious charging of the specimen surface still affects the BSE image, resulting in a beam that may be defocused or deflected, the unstable variations in image brightness that plagued the SE image can largely be avoided (see Fig. 2.32). Figure 2.33 is an image of a sea urchin egg that has been high-pressure frozen, fractured, and observed on a Gatan 636 cold stage in the S-900 without further treatment (i.e., no chemical fixation, staining, or coating). The surface has been slightly freeze-etched at −100 ◦ C for about five minutes to remove some of the ice matrix, revealing the structures that are visible in the micrograph. However, the rectangular hole in the fracture surface shown in Fig. 2.33 makes clear the importance of radiation damage. Because this damage is proportional to the current density, raising the magnification by a factor of 10 for final focusing increases the local dose by 100x. If, on the other hand, one focuses and adjusts the stigmators on the full image during a slow scan,14 one can avoid this localized damage and produce a BSE image such as Fig. 2.34. Once equipment became available for cryo-coating cold specimens that had been fractured after being mounted on a cryotransfer stage, it was possible to see how closely the LVSEM could replicate the image available from freeze-fracture TEM. One such result is shown in Fig. 2.35. This shows not only the expected hexagonal pattern of intramembrane particles on the p-face of a fractured, starved yeast cell, but it also shows the central dot in the image of each particle, as seen in the best TEM results (Walther et al. 1990a). How far can this technique be pushed? Clearly the ultimate limit is imposed by radiation damage to the underlying icy matrix, especially if one accepts the need to record stereo pairs. This damage can be reduced by doing everything right. In particular, one must focus on a part of the fracture surface to the side of the area where the final slow-scan images will be recorded, and likewise one should use nearby structures for aligning the second member of the stereo pair. Using these precautions, one can record a stereo image such as that in Fig. 2.36 that shows a freeze-fractured mitochondria recorded at 5 kV and 400 kx. To compensate for the distortion caused by vertical (with respect to the image plane) stage motion during the two 100 sec exposures, and to facilitate comfortable viewing of this stereo pair, the magnification of the right image had to be increased by about 5%. 14 This will only work if one has carefully centered the objective aperture so that the image does not move when the focus control is adjusted. Likewise, the action of the stigmator coils must be very precisely electronically centered so that their adjustment produces no displacement of the image (Pawley 1990).
84
J. B. Pawley
Fig. 2.32 SE (top) and BSE (bottom) image made at V0 = 2 kV of an uncoated, unstained, biological, freeze-fracture surface. Although charging artifacts in the form of slight deflections or defocusing of the beam are visible in some parts of the image, the brightness defects that make the SE image unusable are absent in the BSE image. (Micrograph generously provided by Paul Walther, IMR, University of Wisconsin-Madison; see also Chapter 10)
2 LVSEM for Biology
85
Fig. 2.33 Radiation damage of the surface of an untreated cryo-fractured blastula stage sea urchin embryo. The square hole visible near the center was etched into the fracture surface when the focusing raster damaged the specimen sufficiently to vaporize the water and organic constituents located there. The embryo was frozen by plunging into supercooled ethane, fractured with a cold razor blade in liquid nitrogen and cryo-transferred into the SEM using a Gatan cold stage. The sample was imaged frozen-hydrated, at a temperature of −140 ◦ C at 2.6 kV, using the Autrata BSE signal. (Micrograph generously provided by Paul Walther, IMR, University of Wisconsin-Madison; see also Chapter 10)
86
J. B. Pawley
Fig. 2.34 BSE image at low V0 of a blastula-stage sea urchin embryo that has been high-pressure frozen, fractured, freeze-etched and viewed directly at −120 ◦ C on a cold stage in the Hitachi S-900. The contrast in this image is remarkable in view of the lack of any significant density variations in the specimen. The contrast seems to be mainly collection contrast: BSE emerging from the surfaces of holes in the specimen have less chance to reach the large-angle, Autrata detector mounted symmetrically above the specimen (Autrata 1990). (Micrograph generously provided by Paul Walther, IMR, University of Wisconsin-Madison; see also Chapter 10)
2 LVSEM for Biology
87
Other errors visible in the stereo pair may reflect changes in the structure of the specimen caused by radiation damage. Because neither of these factors would have limited performance when imaging a metal-shadowed carbon replica in the TEM, high-magnification stereo imaging of fracture faces is probably not the ideal use for cryo-LVSEM.
Fig. 2.35 A freeze-fractured surface of starved yeast, viewed frozen-hydrated on the cryo-transfer stage of the Hitachi S-900. The boxed image has been enlarged to show the dot in the center of each of the hexagonally-arranged intramembranous particles. This structure is well known from images of metal-shadowed carbon replicas viewed in the TEM. (Micrograph generously provided by Paul Walther, IMR, University of Wisconsin-Madison)
88
J. B. Pawley
Fig. 2.36 Stereo pair image recorded at 2kV of the freeze-fractured surface of a frozen-hydrated mitochondrion, made using minimal-dose techniques. The original images were recorded at 70 kx. To compensate for the distortion caused by vertical (with respect to the image plane) stage drift during the two, 100 sec exposures, and to facilitate comfortable viewing of this stereo pair, the magnification of the right image had to be increased by about 5%. Other errors visible in the stereo pair may reflect changes in the structure of the specimen caused by radiation damage. (Micrograph generously provided by Ya Chen, IMR, University of Wisconsin-Madison)
The freeze-fracture, thaw-fix technique developed by Geoffrey Haggis (Haggis 1987; Haggis & Pawley 1989) is a useful alternative method because the vitreous water that is so sensitive to electron radiation (Talmon 1984) is no longer present when the specimen is imaged. In this procedure, the freeze-fractured specimen is removed from the vacuum and dropped into fixative. It is then processed by CPD, coated with 1–2 nm of ion-beam-sputtered Pt, and viewed at ambient temperature. Figure 2.37 shows the surface of a fractured 3T3 cell prepared in this way. Because the thaw-fix step allows some of the material below the fracture surface to wash away, this technique produces specimens that allow one to see much farther into the cell than does normal freeze-fracture. Another application in which cryo-LVSEM has proven useful relates to the specimens normally involved in high-resolution TEM electron crystallography. In this technique, two-dimensional crystalline arrays of biomacromolecules are deposited onto a carbon film, flash frozen, and imaged at very low electron dose in a cryo-TEM (Glaeser 1971, 1975; Chiu et al. 1985; Downing 1991; Li et al. 2002). The technique takes advantage of the spatial redundancy of the crystalline array to calculate the three-dimensional structure of a single average subunit while subjecting each subunit to a radiation dose so small that little damage is done in the process. To succeed, it is important that the crystal be deposited on a planar surface having a
2 LVSEM for Biology
89
known orientation to the TEM beam. However, a problem can arise if a carbon film that is planar when attached to a copper EM grid at room temperature becomes crinkled as the grid is cooled to 100 ◦ K for cryo-TEM (Fig. 2.38a-c). Because the shape of the film is essentially invisible in normal TEM or SEM, this crinkling process went unobserved until such a grid was cooled down in a cryo-LVSEM (Booy and Pawley, 1993). The solution was to use grids made out of a material, such as Pt, having an expansion coefficient that more closely matches that of amorphous carbon (Fig. 2.38d-f).
Recent Developments One of the prime advantages of SEM over previous types of electron and light microscopes has always been its ability to make informative stereo images (Wergin & Pawley 1980). To make such images, it is necessary to have a specimen containing three-dimensional structure mounted on a tiltable stage in an imaging system characterized by considerable depth-of-focus. The SEM meets all these criteria, and a number of stereo images have been included in this and in other chapters to emphasize the advantages of stereo imaging. Indeed, as was first pointed
Fig. 2.37 Stereo pair of a 3T3 cell prepared by fresh-freezing in propane, freeze-fracture and thawing into fixative. The specimen was then critical point dried and coated with ion-beam-sputtered Pt by G. Haggis (Agriculture Canada, Ottawa) and viewed in the S-900 at the IMR. Visible are the nucleus (top), nuclear membrane, and peri-nuclear space. The open nature of the preparation comes from it having been thawed into the fixative, a process that permits some constituents to wash away, revealing others below that are more firmly attached. Field width = 2.8 µm
90
J. B. Pawley
Fig. 2.38 Cryo-crinkling that occurs when carbon substrate films supported by Cu EM grids are cooled down. Carbon-coated 400-mesh copper grid, examined at 1.5 kV and 40 ◦ tilt as it is cooled from room temperature (a), to −170 ◦ C (b), and then re-warmed to room temperature (c). The grid is covered with patches purple membrane protein, embedded in glucose prepared by flotation from mica by Ken Downing (LBL, Berkeley). One can easily see the wrinkles that develop as the specimen is cooled. However, when a similar film was supported by a Pt grid (d, e and f), there is little change in the wrinkle pattern as the temperature changes (Booy & Pawley 1993)
out by Boyde in the 1970s, one can be confident in understanding the surface structure of a complex SEM specimen only when one views the results in stereo (Boyde et al. 1974). One way to encourage stereo imaging is to make it easier for the microscopist to view the image in stereo in real time while scanning the specimen (Pawley 1978. An expanded discussion of live-time stereo imaging is provided in Chapter 7 by Boyde—the first person to develop such a system). This can be accomplished by using special scan coils to control the angle at which the electron beam arrives at the specimen surface (see Fig. 2.39a). By applying a square wave signal to these coils, the arrival angle can be made to oscillate plus and minus a few degrees from the electron-optical axis. If the square wave is synchronized with either the line-scan or frame-scan signal of the microscope, alternate lines or frames will correspond to the view that would be seen by the right or left eye when viewing a stereo pair. If
2 LVSEM for Biology
91
the video signal is then synchronously fed to two different displays, so arranged that each eye sees only one screen and the images seem to overlap, the viewer will perceive a stereo image (see Fig. 2.39b). We installed such a system on the Hitachi S-900 at the IMR. Because the imaging beam must pass through the objective lens both off the EO-axis and at angle to it, it is subjected to additional aberrations—such as coma and field curvature—that are not usually important in electron microscopy. As a result, the optical performance is compromised to some extent, although this can be minimized by carefully centering the final aperture and incorporating additional circuitry so that the stigmator controls can be adjusted independently for the right and left views. Although the two images shown in Fig. 2.40 had to be recorded using a slow vertical scan, these images are very similar to those that would have been observed by the operator, with the exception that the live image would have been somewhat noisier. Using a V0 of
(a)
Fig. 2.39 (continued)
92
J. B. Pawley
(b)
Fig. 2.39 Schematic diagram of the electron optical components used to implement beam-tilt stereo. (a) Beam-tilt gun-alignment coils located below the FE gun can be used to tilt the beam as it goes through the normal scanning coils. This has the effect that the angle at which the scanned raster approaches the surface of the specimen can be adjusted electronically. By applying a squarewave signal to the tilt coils, so that the raster-approach angle is tilted plus or minus a fixed amount, and one can sequentially record the two images needed to make up a stereo pair. b. If the video circuitry of the two video displays is modified so that the image data collected at each beam tilt is only conveyed to the appropriate screen. Using a mirror stereoscope to superimpose the images on the two screen images, a stereo pair image can be viewed in real time by an operator. (A more complete discussion of beam-tilt stereo can be found in Chapter 7)
Fig. 2.40 Real-time stereo image of a tissue-culture cell, recorded using the system diagrammed in Fig. 2.39. Although these slow-scan images are less noisy than are the fast-scan images used when actually viewing the specimen in real time, because the fast-scan image persists on the retina for some time, they do give a fair idea of the available image quality
2 LVSEM for Biology
93
Fig. 2.41 High magnification real-time stereo image of a tissue-culture cell. Same as Fig. 2.40, but at higher magnification to show the loss of image quality caused by the beam traversing the magnetic field of the final lens displaced from the optical axis and at an angle to it
2.5 kV, we found that this system could provide a useful, real-time, stereo image up to 20 kx (see Fig. 2.41). Although the mirror-stereo scope viewer shown in Fig. 2.39b has the disadvantage that the image can be viewed by only a single observer who must place his or her head in a specific location, it does provide better optical separation between the two images than is available from anaglyph (red-green) systems, which alternately show each image in a different color on a two-dimensional color display viewed through red-green glasses (Wergin & Pawley 1980). On more recent microscopes that utilize digital image memories to collect the data, it would be straight forward to display the two images on a standard stereo display that combines LCD elements in front of the screen with polarizing elements in front of each eye to ensure that each eye sees only the appropriate image.
Outlook It is hoped that this chapter has successfully made the case that a modern, lowaberration FE-SEM used at low V0 is probably the ideal method for viewing the rough surfaces of biological objects. Now that such SEMs have become more widely available, it seems likely that they will be used increasingly for this purpose. Two factors are holding back this process—the high cost of the equipment and the need for great care in the proper preparation of the specimen. Regarding the first problem, it is perhaps worth remembering that the second FE-SEM offered by the Hitachi Corporation operated at 2–4 kV, was enclosed in a small cube that sat on a desktop, and cost only $29,000 (USD). Although one must admit that dollars were worth more in the 1970s, it is clear that no physical laws must be violated to
94
J. B. Pawley
produce a useful low-V0 FE-SEM for considerably less than the $800k that is now commonly asked. For more detailed coverage of the optimal methods of specimen preparation, the reader is referred to Chapters 5, 6, 8, 9 and 10. Acknowledgment Almost all of the results I have included in this review were derived either from specimens generously provided by others or from micrographs taken by them. I would like to thank Drs. Stan Erlandsen (U. Minnesota) for Figs. 2.3, 2.4, and 2.27; David Plantz (U. Wisconsin) for Fig. 2.2; David Joy (U. Tennessee) for Figs. 2.8 and 2.9; Ralph Albrecht (U. Wisconsin-Madison) for Figs. 2.6, 2.11, and 2.14; Geoffrey Haggis, (Agriculture Canada, Ottawa) for Fig. 2.37; Paul Walther (U. Uhlm) for Figs. 2.12, 2.32, 2.33, 2.34, and 2.35; Girard Coffe, for Figs. 2.19 and 2.26; Hideo Naito (Hitachi Scientific Equipment, California) for Fig. 2.13; Alan Boyde (U. London) for Fig. 2.28; Ya Chen (U. Wisconsin-Madison) for Figs. 2.7, and 2.36; Hans Ris (U. Wisconsin-Madison) for Figs. 2.17, 2.29, 2.30 and 2.31; Steve Goodman (U. Wisconsin-Madison) for Fig. 2.20. In addition, I would also like to thank David Joy, Hans Ris, and Paul Walther for their comments on the manuscript, Charles Thomas for contending with the typing and managing the references, and Ya Chen for helping with the photography. The work was supported by NIH Grant DRR-570 to the Madison Integrated Microscopy Laboratory and by NSF Grant 9103081 to develop a prototype, commercial LVSEM.
References Adachi K, Houjou K, Katoh M, Kanaya K (1976) High-resolution shadowing for electron microscopy by sputter deposition, Ultramicrosc. 2:17–29 Adams WW, Price G, Krause SJ (1990) Imaging of polymer single crystals in low voltage, highresolution scanning electron microscopy, Proc. XIIth ICEM Mtg. 1106–1107 Aebi U, Jarnik M, Reichelt R, Engel A (1990) Structural analysis of the nuclear pore complex by conventional and scanning transmission electron microscopy, (CTEM/STEM), EMSA Bull. 20 2:69–76 Albrecht RM and Hodges GM (1988) Biotechnology and Bioapplications of Colloidal-Gold, SMI International: Chicago, IL Albrecht RM, Simmons SR, Prudent JR, Erickson CM (1988) High-resolution SEM of colloidalgold labels, Proc. EMSA, 46:214–215 Albrecht RM, Goodman SL, Simmons SR (1989) Distribution and movement of membrane associated platelet glycoproteins: use of colloidal-gold with correlative video enhanced light microscopy, low-voltage high-resolution scanning electron microscopy and high voltage transmission electron microscopy, Am.J. Anat. 185:149–164 Anger K, Lischke B, Sturm M (1983) Material surfaces for electron-optical equipment, Scanning 5:39–44 Apkarian RP, Wright ER, Seredyuk VA, Eustis S, Lyon LA, Conticello VP, Menger FM (2003) Inlens cryo-high-resolution scanning electron microscopy: methodologies for molecular imaging of self-assembled organic hydrogels. Microsc. Microanal. 9-4:286–295. Armstrong DA, Luo S, Joy DC (1990) Re-examining mechanisms of radiation damage in organic specimens, Proc. ICEM Mtg. 12:812–813 Arnold WH, Singh B, Phan K (1989) Linewidth metrology requirements for submicron lithography, Solid State Tech. 32-4:139–145 Arro E, Collins VP, Brunk UT (1981) High-resolution SEM of cultured cells: Preparatory procedures, Scanning Electron Microsc. II, 159–168 Atwood D and Barton R (1989) Proc. of the Workshop on X-ray Microimaging for the Life Sciences, Berkeley, CA, Lawrence Berkeley Lab, University of California-Berkeley, Lawrence Berkeley Laboratory, Berkeley, California, LBL Report No. 27660
2 LVSEM for Biology
95
Autrata R, Schauert P, Kvapil JS, Kvapil J (1978) A single crystal of YAG - new fast scintillator in SEM, J. Phys.E: Sci.Instrum. II, 707–708 Autrata R, Schauer P, Kvapil JS, Kvapil J (1983) Single-crystal aluminates−new generation of scintillators for scanning electron microscopes and transparent screens for electron optical devices, Scanning Electron Microsc. II, 489–500 Autrata R (1989) Backscattered electron imaging using single crystal scintillator detectors, Scanning Microsc. 3:739–763 Autrata R (1990) New configurations of single-crystal scintillator detectors in SEM, Proc. XIIth ICEM Mtg, 376–377 Bachmann L, Becker R, Leupold G, Barth M, Guckenberger R, Baumeister W (1985) Decoration and shadowing of freeze-eched catalase crystals, Ultramicrosc. 16:305–320 Ball MD and McCartney DG (1981) The measurement of atomic number and composition in a SEM using backscattered detectors, J. Microsc. 124:57–68 Ballard DB (1972) Comparison and evaluation of specimens for resolution standards, Scanning Electron Microsc. 1972, 121–128 Barth JE, Jansen GH, Kruit P (1990) Low voltage diffraction limited probe current: Limits due to e-e interactions in intermediate cross-over, Proc. XIIth ICEM Mtg. 394–395 Bauer B, Speidel R (1981) Influence of energy on spread of field-emitted electrons on resolution in the scanning transmission electron microscope (STEM), Ultramicrosc. 6:281–286 Becker RP and Sogard M,(1979) Visualization of subsurface structures in cells and tissues by backscattered electron imaging, Scanning Electron Microsc. 1979, II, 835–870 Bell PB, Lindroth M, Fredriksson BA (1989) Preparation of cytoskeletons for high-resolution scanning and scanning transmission electron microscopy, Scanning Microsc. Supplement 3, 117–135 Bendayan M (1984) Protein-A gold electron microscopic immunocytochemistry: methods, applications and limitations, J. Elec.Microsc.Tech. 1:243–270 Bendayan M (1987) Introduction of the protein-G gold complex for high-resolution immunocytochemistry, J. Elec.Microsc.Tech. 6:7–13 Bennett MH and Guller GE (1986) In process inspection and metrology of semiconductor wafers with the use of an automated low voltage SEM, Microbeam Anal. 21:649–652 Black DR and Ballard DB (1982) Sputter coated carbon specimens for SEM performance testing, Proc.EMSA 40:750–751 Boersch H (1954) Experimentele bestimmung der energieverteilung in thermisch ausgloesten elektronen strahlen, Z.Phys. 139:139 Booy FP and Pawley JB (1993) Cryo-crinkling: what happens to carbon films on copper grids at low temperature, Ultramicroscopy, 43:273–280 Boyde A (1971) A review of problems of interpretation of the SEM image with special regard to methods of specimen preparation, Scanning Electron Microsc. 1971, 1–8 Boyde A, Jones SJ, Pawley JB (1974) Some practical applications of real-time TV speed stereo SEM in hard tissue research, Scanning Electron Microsc. III:109–115 Boyde A, Maconnachie E (1979) Volume changes during preparation of mouse embryonic tissue for scanning electron microscopy, Scanning 2:149–163 Boyde A, Maconnachie E (1981) Morphological correlations with dimensional change during SEM specimen preparation, Scanning Electron Microsc. IV:27–34 Alan Boyde, (2003), The real response of bone to exercise, J. Anat. 203:173 Boyes ED (1984a) High-resolution, low voltage scanning electron microscopy (LVSEM). Inst.Phys.Conf.Ser. 68:485–488 Boyes ED (1984b) High-resolution at low voltage: The SEM philosopher’s stone? Proc.EMSA 42:446–450 Brandis EK, DeStafeno J, Flitch R, Landengerger R (1984) Low voltage SEM, auger, and XPS of surface contaminants, Proc. EMSA 42:458–459 Braten T (1978) High-resolution scanning electron microscopy in biology: artifacts caused by the nature and mode of application of the coating material, J. Microsc. 113:53–59 Breese JF (1982) Quantitative investigations in semiconductor devices by electron beam induced current mode: A review, Scanning Electron Microsc. IV, 1487–1500
96
J. B. Pawley
Broers AN (1974) Recent advances in SEM with lanthanum hexaboride cathodes, Scanning Electron Microsc. 10–18 Broers AN, Panessa BJ, Gennaro JF (1975) High-resolution scanning electron microscopy of bacteriophage 3D and T4, Science 189:637–639 Broers AN (1982) Resolution in surface scanning electron microscopy of bulk materials, Ultramicrosc. 8:137–144 Brunner M, Schmid R (1986) Charging effects in low voltage scanning electron microscope metrology, Scanning Electron Microsc. II, 377–382 Brunner M, Schmid R (1987) Characteristics of an electric/magnetic quadrupole detector for low voltage scanning electron microscopy, Scanning Microsc. IV:1501–1506 Buchanan R (1982) New SEM lens gives sharpest micrographs yet, Industrial Res.& Dev. Aug., 92–95 Buchanan R, Menzel E (1984) Some recent development in low voltage E beam testing of IC’s, Proc.EMSA 42:460–464 Castaing R, Guiniert A (1949) Application of electron probes to metallographic analysis (in French), Proc. ICEM Mtg. 1:60–63 Catalano JF (1976) SEM on charge injection semiconductor devices, Scanning Electron Microsc. 1976; I:521–528 Catto CJD, Smith KCA (1973) Resolution limits in the surface scanning electron microscope, J. Microsc. 98:417–435 Cheng PC and Jan GJ (1987) X-ray Microscopy, Heidelberg, Germany: Springer-Verlag, 415pp. Chiu W, Downing KH, Hobbs LW, Shuman H, Talmon Y (1988) The EMSA committee on cryoelectron microscopy technology, EMSA Bulletin 18-1:16–25 Chiu W, Baker ML, Jiang W, Zhou ZH, (2002) Deriving folds of macromolecular complexes through electron cryomicroscopy and bioinformatics approaches. Curr Opin Struct Biol. 122:263–9. Clarke DR (1970) Review: Image contrast in the scanning electron microscope, J. Material Sci. 5:689–708 Coates DG (1969) Pseudo-kikuchi oritntation analysis in the scanning electron microscope, Scanning Electron Microsc. 29–40 Cosslett VE (1954) X-ray microscopy, Med.Biol.Illus. 4:95–103 Cosslett VE (1978) Radiation damage in the high-resolution electron microscopy of biological materials: A review, J. Microsc. 113-2:113–129 Cowley JM (1990) High-resolution scanning electron microscopy of surfaces, Proc. XIIth ICEM Mtg. 296–297 Crewe AV (1973) Production of electron probes using a field emission source, In: Progress in Optics XI, ed. Wolf, E : North Holland, 225–246 Crewe AV (1985) Towards the ultimate scanning electron microscope, Scanning Electron Microsc. 1985; II:467–472 Crewe AV, Eggenberger DN, Wall J, Welter LM (1968) Electron gun using field emission sources, Rev.Sci.Inst. 39:576–583 Crewe AV, Isaacson M, Johnson D (1971) A high-resolution electron spectrometer for use in transmission scanning electron microscopy, Rev.Sci.Inst. 42-4:441–420 Crewe AV, Lin PSD (1976) The use of backscattered electrons for image purposes in a scanning electron microscope, Ultramicrosc. 1:231–238 Danielson P (1987) Rough-pumping without oil contamination, Res. & Dev. 29-4:97–104 Danilatos GD (1988) Foundations of environmental scanning electron microscopy, In: Advances in Electronics and Electron Physics, ed. Hawkes, P NY: Academic Press, 109 Diehl P, McCartney MR, Smith DJ (1990) Effects of electron irradiation on alkaline earth fluorides, Proc. XIIth ICEM Mtg. 4:794–795 Dilly PN (1980) Enhanced contrast of cilia using low accelerating voltages as an aid to low power survey and counting, Scanning 3:283–284 Dodson TA, Joy DC (1990) Fast fourier transform techniques for measuring SEM resolution. Proc. XIIth ICEM Mtg. 406–407
2 LVSEM for Biology
97
Downing KH (1991) Spot-scan imaging in transmission electron microscopy, Science, 251(4989):53-59 Duncumb P (1957) In: X-ray Microscopy and Microradiography, ed. Cosslett, VE, Engstrom, A and Pattee, HH, NY: Academic Press, p. 435. Echlin P (1971) The examination of biological material at low temperatures, Scanning Electron Microsc. 1:225–232 Echlin P (1991) Recent advances in specimen coating techniques for electron microscopy, Scanning Electron Microscopy, 1:225–232 Echlin P (1992) Low-Temperature Microscopy and Analysis. Plenum Press: New York and London. Erlandsen SL, Gould PR, Frethem C, Wells CL, Pawley JB, Hamilton DW (1989a) Membrane fixation for high-resolution low voltage SEM: Studies on Giardia, rat spermatozoa, and mouse macrophages. Scanning 11:169–175 Erlandsen SL, Bemrick WJ, Pawley J (1989b) High-resolution electron microscopic evidence for the filamentous structure of the cyst wall in Giardia muris and Giardia duodenalis. J. Parasitology 75:787–797 Erlandsen SL, Frethem C, Autrata R (1990a) Workshop on high-resolution immunocytochemistry of cell surfaces using field emission SEM, J. Histochem.Cytochem. 38:1779–1780 Erlandsen SL, Sherlock LA, Bemrick WJ (1990b) The detection of Giardia muris and Giardia lamblia cysts by immunofluorescence in animal tissues and fecal samples subjected to cycles of freezing and thawing. J. Parasitology 76:267–271 Erlandsen SL, Bemrick WJ, Schupp DE, Shields JM, Jarroll EL, Sauch JF, Pawley JB (1991) High-resolution immunogold localization of Giardia cyst wall antigens using field emission SEM sixth secondary and backscatter electron imaging, J. Histochem.Cytochem. 38:625–632 Erlandsen SL, Russo AP, Turner JN (2004), Evidence for Adhesive Activity of the Ventrolateral Flange in Giardia lamblia, J. Eukaryot. Microbiol. 51-1:73–80 Evans AC, Franks J (1981) Specimen coating for high-resolution scanning electron microscopy, Scanning 4:169–174 Everhart TE, Thornley RFM (1960) Wide-band detector for micro-ampere low-energy electron current, J. Sci.Inst. 37:246–248 Everhart TE, Wells OC, Oatley CW (1959) Factors affecting contrast and resolution in the scanning electron microscope, J. Elec.Cont. 7:97–111 Faulk WP, Taylor GM (1971) An immuno colloid method for the electron microscope, Immunocytochem, 8:1081–1083 Ferguson VL, Bushby AJ, and Boyde A (2003) Nanomechanical properties and mineral concentration in articular calcified cartilage and subchondral bone, J. Anat. 203:191–202 Fourie JT, (1981) Electric effects in contamination and electron beam etching, Scanning Electron Microsc. 1981; I:155–162 Franks J, Clay CS, Peace GW (1980) Ion beam thin film deposition, Scanning Electron Microsc. I:155–162 Frosien J, Plies E, Anger K (1989) Compound magnetic and electrostatic lenses for low voltage applications. J. Vac.Sci.Technol. B7 6:1874 Gerace L, Burke B (1988) Functional organization of the nuclear envelope, Ann.Rev.Biochem. 4:335–374 Gerlach RL, MacDonald NC (1976) Recent advances in scanning auger instrumentation, Scanning Electron Microsc. I:199–206 Glaeser RM (1971) Limitations to significant information in biological electron microscopy as a result of radiation damage. J. Ultrastruct.Res. 36:466 Glaeser RM (1975) Radiation damage and biological electron microscopy, InPhysical Aspects of Electron Microscopy and Microbeam Analysis, ed. Siegel, BM and Beaman, DR. NY: Wiley and Sons, 205–230 Goodman SL, Park K, and Albrecht RM (1990) A correlative approach to colloidal-gold labeling with video enhanced light microscopy, low voltage scanning electron microscopy and high voltage electron microscopy, In: Colloidal-Gold: Methods and Applications, ed. Hayat, MA, Van Nostrand Reinhold, NY, 369–409
98
J. B. Pawley
Gray J, Corey D, Ellis G, Sokol R (1989) Microchannel plate-based detection systems for scanning electron microscopy, Proc.EMSA 47:762–767 Haggis GH (1987) Freeze-fracture of cell nuclei for high-resolution SEM and deep-etch TEM, Proc. Electron Microsc. Soc. Am. 45:560–564 Haggis GH, Pawley JB (1988) Freeze-fracture of 3T3 cells for high-resolution scanning electron microscopy. J. Microsc. 150:211–218 Hainfeld J (1977) Understanding and using field emission sources, Scanning Electron Microsc. I:591–604 Hashimoto N, Todokoro H, Fukuhara S, Senoo K (1982) Process characterization of MOS devides by scannnig electron microscopy with 0.5-1kV electrons, Jpn. J. Appl. Phys. I (Japan) 21:199–203 Hasselbach J, Reike U, Straub M (1983) An imaging secondary electron detector for the scanning electron microscope, Scanning Electron Microsc. II:467–478 Hayes TL (1973) Scanning Electron Microscopy, In: Advanced Techniques in Biological Electron Microscopy, ed. Koehler, JK, Heidelberg: Springer-Verlag, 154–209 Hayes TL (1980) Biophysical aspects of scanning electron microscopy, Scanning Electron Microsc. I:1–10 Hefter J (1987) Morphological characterizations of materials using low voltage scanning electron microcsopy, Scanning Microsc. 1:13–21 Helbig JF, Rydgren RD, Kotorman L (1987) Channel plate detection in low energy scanning electron microscopy, Scanning Microsc. IV:1491–1499 Hermann R, Pawley J, Nagatani T, Müller M (1988) Double-axis rotary shadowing for highresolution scanning electron microscopy, Scanning Electron Microsc. II:1215–1230 Herter P, Tresp G, Hentschel H, Zierold K, Walther P (1991) High-resolution SEM of frozenhydrated and freeze substituted kidney tissue, J. Microsc. 161-2:375–385 Heuser JE (1979) Quick-freeze, deep-etch preparation of samples for 3D electron microscopy, Trends Biochem.Sci. 6:64–68 Hobbs LW (1979) Radiation effects in analysis of inorganic specimens by TEM, In: Introduction to Analytical Electron Microscopy, ed. Hren, JJ, Goldstein, JI and Joy, DC, NY: Plenum Press, 437–480 Holy J, Simerly C, Paddock S, Schatten G (1991) Three-dimensional imaging of fertilization and early development, J. EM Tech. 17:384–400 Horisberger M, Rosset J (1977) Colloidal-gold: a useful maker for transmission and scanning electron microscopy, J. Histochem.Cytochem. 25:295–305 Horisberger M (1979) Evaluation of colloidal-gold as a cytochemical marker for transmission and scanning electron microscopy, Biol.Cellulaire 36:253–58 Hren J (1986), Barriers to AEM: Contamination and etching, In: Principles of Analytical Microscopy, eds. Joy, D, Romig, AD and Goldstein, J, NY-London: Plenum Press, 353–375 Humphreys CJ, Bullough TJ, Devenish RW, Maher DM, Turner PS (1990) The interaction of electron beams with solids - some new effects, Proc. XIIth ICEM Mtg. 4:788–789 Ichinokawa T (1990) Scanning low energy electron diffraction microscopy combined with scanning tunnling microscopy, Proc. XIIth ICEM Mtg. 303–303 Ingram P, Morosoff N, Pope L, Allen F, Tisher C (1976) Some camparisons of the techniques of sputter (coating) and evaporative coating for scanning electron microscopy, Scanning Electron Microsc. I:75–82 Irino S, Murakami T, Fujita T, Nagatani T, Kaneshige T (1978) Microdissection of tannin-osmium impregnated specimens in the scanning electron microscope: Demonstration of arterial terminals in human spleen, Scanning Electron Microsc. I:111–116 Isaacson M, Langmore JP (1974) Determination of the non-localization of the inelastic scattering of electrons by electron microscopy, Optik 41:92–96 Jakubowicz A (1987) Theory of electron beam induced current and cathodoluminescence contrasts from structural defects of semiconductor crystals: Steady-state and time-resolved, Scanning Microsc. 1-2:515–533 Johnson TJA (1985a) Aldehyde fixatives: Quantification of acid-producing reactions, J. EM Tech. 2:129–138
2 LVSEM for Biology
99
Johnson TJA (1985b) Glutaraldehyde fixation chemistry: A scheme for rapid crosslinking and evidence for rapid oxygen consumption, In: Science of Biological Specimen Preparation, ed. Johari, O, AMF O’hare (Chicago) IL: Scanning Electron Microscopy, Inc., 51–62 Jones AV (1989) High-resolution at low voltage: A new approach, Proc. EMSA 47:76–77 Jones SJ, Boyde A, and Ali NN (2004)The resorption of biological and non-biological substrates by cultured avian and mammalian osteoclasts, Anat. and Embryol, 170-3:247–256 Joy DC, Newbury DE, Myklebust RL (1982) The role of fast secondary electrons in degrading spatial resolution in the analytical electron microscope, J. Microsc. 128-II: RP1-RP2 Joy DC (1984) Resolution in low voltage SEM, Proc. EMSA 42:444–445 Joy DC (1985) Resolution in low voltage scanning electron microscopy, J. Microsc. 1403:283–292 Joy DC (1987) A note on charging in low voltage SEM, Microbeam Anal. 22:83–86 Joy DC (1991a) Contrast in high-resolution scanning electron microscope images, J. Microsc. 161-2:343–355 Joy DC (1991b) LVSEM, In: Electron Microscopy and Microanalysis 1987, ed. Brown, LM, Institute of Physics Conference Series, 175–180 Joy DC and Pawley JB (1993) high-resolution Scanning Electron Microscopy, Ultramicrosc., 47:80–100 Keery WJ, Leedy KO, Galloway KF (1976) Electron beam effects on microelectronic devices, Scanning Electron Microsc. I:507–514 Kellenberger E (1991) The potential of cryofixation and freeze substitution: Observations and theoretical considerations, J. Microsc. 161, II:183–203 Kelley RO, Dekker RA, Bluemink JG (1973) Ligand-mediated osmium binding: Its applications in coating biological specimens for SEM, J. Ultrastr.Res. 45:254–258 Kemmenoe BH, Bullock GR (1983) Structure analysis of sputter-coated and ion-beam sputtercoated films: A comparative study, J. Microsc. 132-2:153–163 Kersker M, Neilsen C, Otsuji H, Miyokawa T, Nakagawa S (1989) The JSM-890 ultra highresolution scanning electron microscope, Proc. EMSA 47:88–89 Kimoto S and Hashimoto H (1966) Stereoscopic observation in scanning microscopy using multiple detectors, In: The Electron Microprobe, Proc. Symp. held in Washington, D.C., October, 1964, eds. Heinrich, KFJ and Wittry, DB, New York, NY: John Wiley and Sons, 480–489 Kiseleva E, Drummond SP, Goldberg MW, Rutherford SA, Bagley S, Allen TD and Wilson KL (2004). A stable actin-dependent filament network links nuclear pore complexes to nucleoli and Cajal bodies. J Cell Sci; 117: 2481–2490. Knoll M (1935) Aufladepotentiel und sekundaremission electronenbestrahlter korper, Z.Phys. 16:467–475 Koike H, Ueno K, Suzuki M (1971) Scanning device combined with conventional electron microscope, Proc.EMSA 29:28–29 Kosuge T, Hashimoto H, Sato M and Komoto S (1970) Quality of the secondary electron image at low accelerating voltage, In: Microscopie Electronique, ed. Favard, P. Paris: Society Francaise de Microscopie Electronique, 201–202 Kotera M, Muarat K, Nagami K (1981) Monte Carlo simulation of a 1-10 keV electron scattering on a gold target, J. Appl.Phys. 52-2:997–1003 Krause SJ, Maracas GN, Varhue WJ, Joy DC (1989) Low voltage, high-resolution scanning electron microscopy of semiconductors, Proc. EMSA 47:82–83 Kubotsu A, Ueda M (1980) A new conductive treatment of the specimen for scanning electron microscopy, J. Electron Microsc. 29-1:45–53 Langford LA, Coggeshall RE (1980) The use of potassium ferricyanide in nural fixation, The Anatomical Record 197:297–303 Langmuir DB (1937) Theoretical limitations of cathode-ray tubes, Proc. IRE 24-8:977–991 Leamy HJ, Kimerling LC, Ferris SD (1978) Electron beam induced current, Scanning Electron Microsc. 1978, I:717–726 LeFloch H, Franceschi JL, Gourand T, Launay P (1987) Digital image acquisition in scanning electron microscopy, Scanning 9:26–30
100
J. B. Pawley
LeGressus C, Durand JP, Massignon D, Deacon OL (1983) Electron channelling effect on secondary electron image contrast, Scanning Electron Microsc. II:537–542 Lepault J, Erk I, Nicolas G, Ranck JL (1991) Time resolved cryo-electron microscopy of vitrified muscular components, J. Microsc. 161-1:47–59 Levi-Setti R, Wang YL, Crow G (1984) High spatial resolution SIMS with the UC-HRL scanning ion microprobe, J. Phys. (Paris) 45, C9-197-C9-205 Li H, DeRosier DJ, Nicholson WV, Nogales E, and Downing KH (2002) Microtubule structure at 8 Å resolution. Structure 10:1317–1328 Lim SS, Ris H, Schnasse B (1987) Pigment granules in goldfish xanthophores are attached to intermediate filaments, J. Cell Biol. 105, 37a Lindroth M, Bell PB, Fredriksson BA (1988) Comparison of the effects of critical point drying and freeze-drying on cytoskeletons and microtubules, J. Microsc. 151-2:103–114 Lindroth M, Sundgren JE (1989) Ion-beam-sputtered and magnetron-sputtered thin films on cytoskeletons: A high-resolution TEM study, Scanning 11:243–253 Liu J, Cowley JM (1988) High-resolution secondary electron imaging in a scanning transmission electron microscopy instrument, Scanning Microsc. 2-1:65–81 MacDonald NC (1971) Auger electron spectroscopy for scanning electron microscopy Scanning Electron MicroscI:89–96 Malecki M, Ris H (1991) Preparation of cell suspensions for ultrastructural studies, Scanning 131:82–83 Malecki M, Walther P (1991) High pressure freezing of cell aggregates for LVSEM, Scanning 13-1:68–69 Martin JP, Jenkinson G, Bulgin D (1985) Quantitative scanning electron microscopy using integrated digital image store for on-line image analysis, Scanning 7:239-242 Martin TF (1989) Cell cracking: permeabilizing cells to macromolecular probes. In Methods in Enzymology, Academic Press, NY. 168:225–233 McMullan D (1953a) An improved scanning electron microscope for opaque specimens, Proc. IEE 100-2:245–259 McMullan D (1953b) The scanning electron microscope and the electron-optical examination of surfaces, Electron. Eng.(England) 25:46–50 McMullan D (1990) The prehistory of scanned image microscopy, Part 2: The scanning electron microscope, Proc. Roy. Microsc. Soc. 25:189–194 McMullin PG (1976) Quality evaluation of the GaAs-AlGaAs heterostructure wafers using the electron beam induced current technique, Scanning Electron Microsc. I:543–550 Menzel E, Buchanan R (1985) Some recent developments in low voltage E-beam testing of ICs, J. Microsc. 140-3:331–349 Menzel E, Kubalek E (1982) Fundamentals of electron beam testing of integrated circuits, Scanning 5:103–122 Miyokawa T, Norioka S, Goto S (1988) Development of a conical anode FE gun for low voltage SEM, Proc. EMSA 46:978–979 Miyoshi M, Isikawa M, Okumura K (1982) Effects of electron beam testing on the short channel metal oxide semiconductor characteristics, Scanning Electron Microsc. 1982, IV: 1507–1514 Moll SH, Healey F, Sullivan B, Johnson W (1979) Further development of the converted backscattered electron detector, Scanning Electron Microsc. II:149–154 Morandi C, Vanzi M, Bianco F, Neri R (1989) A PC-AT-based system for the acquisition of SEM images, Scanning 11:81–85 Mullerova I (2001) Imaging of specimens at optimized low and very low energies in scanning electron microscopes.Scanning. 23-6:379–94. Müller M (1992) The Integrating Power of Cryofixation Based Electron Microscopy in Biology. Acta Microscopica. 1:37–44. Müller M, Hermann R (1990) Towards high-resolution SEM of biological materials, Proc. XIIth ICEM Mtg. 4–5 Munger B, Mumaw V (1976) Specimen preparation for SEM study of cells and cell organelles in uncoated preparations, Scanning Electron Microsc. I:275–280
2 LVSEM for Biology
101
Murakami T, Jones AL (1980) Conductive staining of biological specimens for non-coated scanning electron microscopy: Double coat staining by tannin-osmium and osmiumthiocarbohydrazide-osmium methods, Scanning Electron Microsc. I:221–226 Murata K, Kawata H, Nagami K (1987) Electron scattering of in low voltage scanning electron microscopy targets, Scann.Microsc.Suppl. I:83–91 Murphy J (1978) Non-coating techniques to render biological specimens conductive, Scanning Electron Microsc. II:175–194 Murphy J (1980) Non-coating techniques to render biological specimens conductive: 1980 update, Scanning Electron Microsc. I:209–220 Nagatani T and Saito S (1986a) Instrumentation for ultra high-resolution scanning electron microscopy, In: Electron Microscopy 1986, ed. Imura T, 2101–2108 Nagatani T, Saito S (1986b) Instrumentation for ultra high-resolution scanning electron microscopy, Proc. XIIth ICEM Mtg. 2101–2104 Nagatani T, Sato M, Osumi M (1990) Development of an ultra high-resolution low voltage (LV) SEM with an optimized "in-lens" design, Proc. XIIth ICEM Mtg. 388–389 Newbury DE, Marinenko RB, Brught DS, Myklebust RL (1988) Computer-aided imaging: Quantitative compositional mapping with the electron probe microanalyzer, Scanning 10: 213–255 Nixon WC (1955) Improved resolution with x-ray projection microscope, Nature 175:1078–1079 Nomura S, Komoda T, Kameryo T, Nakaizumi V (1973) Stable field emission gun with an electronic feedback system, Scanning Electron Microsc. 65–72 Oatley CW, Everhart TE (1957) The examination of p-n junctions with the scanning electron microscope, J. Electronics II, 6:568–570 Oatley CW, Nixon WC and Pease RFW (1965) Scanning Electron Microscopy, In: Advances in Electronics and Electron Physics, New York, NY: Academic Press, 181–247 Oatley CW (1972) The Scanning Electron Microscope, Part I: The Instrument, Cambridge, UK:The University Press, 1–194 Oatley CW (1975) The tungsten filament gun in the scanning electron microscope, J. Phys. E.: Sci. Inst. 8:1–5 Oatley CW (1982) The early history of the scanning electron microscope, J. Appl. Phys. 53-2: R1-R13 Ogura K, Ono A, Kersker MM,(1989a) Reduction in contamination using a specimen heating holder in an ultrahigh-resolution SEM, Proc.EMSA 47:724–725 Ogura K, Adachi S, Satoh T, Watebe T, Kersher MM (1989b) Magnetron sputter coating for ultra high-resolution scanning electron microscopy (Simultaneous coating of platinum and tungsten using a magnetron sputter coater), Proc. EMSA 47:80–81 Ohama J, Ono A, Harada Y, Gotoh S (1986) An ultra high-resolution SEM equipped with strongly excited objective lens and field emission gun, Proc. XIth ICEM Mtg. 373–374 Ohshita A, Shimoyana H, Maruse S (1978) Brightness in the hot cathode electron gun at high emission densities, J. Electron Microsc. 27-4:253–257 Ohtsuka A, Murakami T, Irino S, Jones AL (1981) Mounting of biological microsamples on protein coats for TaOTO non-coated scanning electron microscopy, Scanning Electron Microsc. II:83–86 Orloff J (1981) A comparison of lanthanum hexaboride, cold field emission and thermal field emission electron guns for low voltage scanning electron microscopy, Scanning Electron Microsc. II:83–86 Orloff J (1985) Thermal field emission for low voltage scanning electron microscopy, J. Microsc. 140, III:303–311 Osumi M, Yamada N, Nagatani T (1988a) High-resolution low voltage SEM of cell wall regeneration of yeast Saccharomyces pombe protoplasts, Proc.EMSA 46:208–209 Osumi M, Baba M, Naito N, Taki A, Yamada N, Nagatani T (1988b) High-resolution low voltage scanning electron microscopy of uncoated yeast cells fixed by the freeze-substitution method, J. Electron Microsc. 37-1:17–30 Osumi M, Yamada N, Kobori H, Taki A, Naito N, Baba M, Nagatani T (1989) Cell wall formation in regenerating protoplasts of Schizosaccharomyces pombe: Study by high-resolution low voltage scanning electron microscopy, J. Electron Microsc. 38-6:437–468
102
J. B. Pawley
Osumi M, Yamada N, Kobori H (1990) Biological application of ultrahigh-resolution low voltage scanning electron microscope, S-900LV: Ultrastructure of glucal fibrils of the reverting protoplast in fission yeast, Hitachi Instrument News, Electron Microscopy Edition 19: 38–39 Osumi M, Yamada N, Yaguchi H, Kobori H, Takashi Nagatani T, and Sato M (1995) Ultrahighresolution Low-voltage SEM Reveals Ultrastructure of the Glucan Network Formation from Fission Yeast Protoplast, J. Elect. Microsc. 44:198–206 Osumi M, Konomi M, Sugawara T, Takagi T, and Baba M (2006) High-pressure freezing is a powerful tool for visualization of Schizosaccharomyces pombe cells: ultra-low temperature and low-voltage scanning electron microscopy and immunoelectron microscopy, J. Electron Microsc. 55-2: 75–88. Pawley JB (1972) Charging artifacts in the scanning electron microscope. Scanning Electron Microsc. I:153–160 Pawley JB (1974) Performance of SEM scintillator materials, Scanning Electron Microsc. 27–34 Pawley JB (1978) Design and performance of presently available TV-rate stereo SEM systems. Scanning Electron Microscopy. I:l57–66 Pawley JB (1984a) SEM at low beam voltage, Proc. EMSA 42:440–444 Pawley JB (1984b) Low voltage scanning electron microscopy, J. Microsc. 136:45–68 Pawley JB (1985a) Low voltage scanning electron microscopy in electron optical systems, Scanning Electron Microsc. 253–272 Pawley JB (1985b) Strategy for locating and eliminating sources of main frequency magnetic stray field, Scanning 7:43–46 Pawley JB (1987a) Use of pseudo-stereo techniques to detect stray field in the SEM, Scanning 9-3:134–136 Pawley JB (1987b) Low voltage scanning electron microscopy, Microbeam Anal. 22:83–86 Pawley JB (1990) Practical Aspects of high-resolution LVSEM, Scanning, 12:247–252, Pawley JB (1992) LVSEM for high-resolution Topographic and Density Contrast Imaging. in Advances in Electronics and Electron Physics,ed. Hawkes PW and Kazan, B, Academic Press, New York 83:203–274, Pawley JB (1997) Development of Field-emission Scanning Electron Microscopy for Imaging Biological Surfaces, Scanning, 19-5:324–336 Pawley JB, ed. (2006) Handbook of Biological Confocal Microscopy, 3rd edition, Springer/Plenum NY Pawley J, Hayes TL, Hook G (1978) Preliminary studies of coated complementary freeze-fractured yeast membranes viewed directly in the SEM, Scanning Electron Microsc. II:683–690 Pawley JB, Norton JT (1978) A chamber attached to the SEM for fracturing and coating frozen biological specimens, J. Microsc. 112:169–182 Pawley JB, Hook G, Hayes TL, Lai C (1980) Direct scanning electron microscopy of frozenhydrated yeast, Scanning 3-3:219–226 Pawley JB, Wall J (1982) A low voltage SEM optimized for high-resolution topographical imaging, Proc. EUREM 1:383–384 Pawley JB, Ris H (1987) Structure of the cytoplasmic filament system in freeze-dried whole mounts viewed by HVEM, J. Microsc. 13:319–332 Pawley JB, Albrecht RM (1988) Imaging colloidal-gold labels in LVSEM, Scanning 10: 184–189 Pawley JB, Erlandsen SL (1989) The case for low voltage high-resolution scanning electron microscopy of biological specimens, Scann.Microsc.Suppl. 3:163–178 Pawley JB, Walther P, Shih SJ, Malecki M (1991) Early results using high-resolution, low voltage, low temperature SEM, J. Microsc. 162-2:327–335 Pease RFW, Hayes TL (1966) Scanning electron microscopy of biological material, Nature 210:1049 Pease RFW, Nixon WC (1965) High-resolution SEM, J. Sci.Instrum. 42:31–35 Pease RFW, Hayes TL, Camp AS, Amer NM (1966) Electron microscopy of living insects, Science 154:1185–1186 Pease RFW, Nixon WC (1968) EM of sprouting seeds, Proc.EMSA 26:88–89
2 LVSEM for Biology
103
Peters KR (1979) Scanning electron microscopy at macromolecular resolution in low energy mode on biological specimens coated with ultra thin metal films, Scanning Electron Microsc. II:133–148 Peters KR (1980) Penning sputtering of ultra thin metal films for high-resolution electron microscopy, Scanning Electron Microsc. I:143–154 Peters KR (1982) Conditions required for high quality high magnification images in secondary electron scanning electron microscopy, Scanning Electron Microsc. IV:1359–1372 Peters KR (1985) Working at higher magnifications in scanning electron microscopy with secondary and backscattered electrons on metal coated biological specimens and imaging macromolecular cell membrane structures, Scanning Electron Microsc. IV:1519–1544 Peters KR (1986a) Rationale for the application of thin, continuous metal films in high magnification electron microscopy, J. Microsc. 142:25–34 Peters KR (1986b) Metal coating thickness and image quality in scanning electron microscopy, Proc. EMSA 44:664–667 Peters KR (1988) Current state of biological high-resolution scanning electron microscopy, Proc. EMSA 46:180–181 Peters KR (1989) Ultra high-resolution SEM at high voltage images individual Fab fragments applied as molecular label to cell surface receptors, Proc. EMSA 47:71–72 Peters KR( 1991) Scanning electron microscopy: Contrast at high magnification, In: Microbeam Analysis 1984, ed. Romig AD and Goldstein JJ, 77–80 Peters KR, Fox MD (1990) Ultra-high-resolution cinematic digital 3D imaging of the cell surface by field emission scanning electron microscopy, Proc. XIIth ICEM Mtg. 12–13 Pfeiffer HC (1972) Basic limitations of probe forming systems due to electron-electron interactions, Scanning Electron Microsc. 113–120 Polasko KJ, Yau YW, Pease RFW (1983) Low energy electron beam lithography, Optical Eng. 22:195–198 Postek MT (1987) Resolution and measurement in the scanning electron microscope, Proc.EMSA 45:534–535 Postek MT, Keery WJ, Frederick NV (1990a) Development of a low-profile high-efficiency microchannel-plate detector system for SEM imaging and metrology, Scanning/90 Abst. FACMS Inc., 53 Postek MT, Keery WJ, Frederick NV (1990b) Low-profile microchannel-plate electron detector system for SEM, Proc. XIIth ICEM Mtg. 378–379 Read NC, Jeffree CE (1991) Low temperature scanning electron microscopy in biology, J. Microsc. 161-I:47 Reimer L (1979) Electron-specimen interactions, Scanning Electron Microsc. II:111–124 Ris H (1985) The cytoplasmic filament system in critical point dried whole mounts and plastic-embedded sections, J. Cell Biol. 100:1474–1487 Ris H (1988) Application of LVSEM in the analysis of complex intracellular structures, Proc. EMSA 46:212–213 Ris H (1989) Three-dimensional imaging of cell ultrastructure with high-resolution low voltage SEM, Inst.Phys.Conf.Ser. 98 (Chp. 16), 657–662 Ris H (1990) Application of low voltage, high-resolution SEM in the study of complex intracellular structures, Proc. XIIth ICEM Mtg. 18–19 Ris H (1991) The three-dimensional structure of the nuclear pore complex as seen by high voltage electron microscopy and high-resolution low voltage scanning electron microscopy, EMSA Bull. 21-1:54–56 Ris H (1997) High-resolution field-emission scanning electron microscopy of nuclear pore complex. Scanning 19:368–375 Ris H and Pawley JB (1989) Analysis of complex three-dimensional structures involved in dynamic processes by high voltage electron microscopy and low voltage high-resolution scanning electron microscopy, In: Microscopy of Subcellular Dynamics, ed. Pattner, H Boca Raton, FL: CRC Press, 309–323 Ris H, Malecki M (1993) High-resolution field-emission scanning electron-microscope imaging of internal cell structures after epon extraction from sections—a new approach to correlative ultrastructural and immunocytochemical studies. J Struct Biol 111:148–157
104
J. B. Pawley
Robards AW and Sleytr UB (1985) Low Temperature Methods in Biological Electron Microscopy, Amsterdam: Elsevier Robinson VNE (1974) The construction and uses of an efficent backscattered electron detector for SEM, J. Phys.E: Sci.Instrum. 7:650–652 Roderick Y, Lim H, Aebi U, Stoffler D (2006) From the trap to the basket: getting to the bottom of the nuclear pore complex, Chromosoma 115:15–26 Rosencwaig A (1982) Thermal wave imaging, Science 218:223–228 Studer D, Michel M, Müller M, (1989) High-pressure freezing comes of age. Scanning Microsc 3, Suppl 3:253–268 Russell PE (1984) Microchannel plates as specialized scanning electron microscopy detectors, Scanning Electron Microsc. 197–200 Russell PE (1988) Low voltage SEM for metrology and inspection, Microbeam Anal. 23:463–465 Russell PE, Mancuso JF (1985) Microchannel plate detector for low voltage scanning electron microscopes, J. Microsc. 140 III:323–330 Saito S, Nakaizumi Y, Mori H, Nagatani T (1982) A field emission SEM controlled by microprocessor, EMI I, (Deutsch Gessellschaft fur Electronenmikroscopy e.V.) 379, 380 Salpeter MM, Marchaterre M, Harris R (1988) Distribution of extrajunctional acetylcholine receptors on a vertebrate muscle: Evaluated by using a scanning electron microscope autoradiographic procedure, J. Cell Biol. 106:2087–2093 Sato M, Nakaizumi Y, Yamada M, Nagatani T (1990) Development of a low accelerating voltage SEM (S-900H), Hitachi Instrument News, Electron Microscopy Edition 19:45–49 Schmid R, Brunner M, (1986) Design and application of a quadrupole detector for low voltage scanning electron microscopy, Scanning 8-6:294–299 Seiler H (1976) Determination of the "information depth" in the SEM, Scanning Electron Microsc. I:9–16 Sewell PB, Ramachandran KN (1978) Grid aperture contamination in electron guns using directly heated lanthanum hexaboride sources, Scanning Electron Microsc. I:221–232 Shaffner TH, Hearle JWS (1976) Recent advances in understanding specimen charging, Scanning Electron Microsc. I:61–70 Shao Z, Crewe AV (1987) Chromatic aberration effects in small electron probes, Ultramicrosc. 23:169 Shao Z, Crewe AV (1988) A study on the optimization of aperture in an aberrated probe forming system, Optik, 79-3:105–110 Shao Z, Crewe AV (1989) On the resolution of the low-energy reflection microscope based on wave electron optics, Ultramicrosc. 31:199 Sitte H, Edelman L and Neumann K (1987) Cryofixation without pretreatment at ambient pressure. In: Cryotechniques in Biological Electron Microscopy, eds. Steinbrecht, RA and Zierold, K, Berlin: Springer, 87–113 Smith KCA (1956) The scanning electron microscope and its fields of application, PhD thesis, Engineering School, Cambridge University, UK.137–138 Speth AJ, Fang FF (1965) Effects of low energy electron irradiation on Si-insulated gate FETs, Appl.Phys.Let. 7:6 Statham PJ (1988) Pitfalls and advances in quantitative elemental mapping, Scanning 10: 245–252 Studer D, Michel M and Müller M (1989) High pressure freezing comes of age, Scanning Microscopy, Suppl. 3, 1989: The Science of Biological Specimen Preparation for Microscopy and Microanalysis, eds. Albrecht RM and Ornberg RL, Chicago (AMF O’Hare), IL: Scanning Microscopy Intl., 253–269 Sugiyama N, Ikeda S, Uchikawa Y (1986) Low voltage SEM inspection of micro electronic devices, J. Electron Microsc. (Japan) 35 1:9–18 Sugiyama N, Ikeda S, Uchikawa Y (1988) SEM voltage contrast mechanism of passivated devices, Scanning 10-1:3–8 Swanson LW, Rathkey DS (1989) A comparison of Schottky emission and cold field emission cathodes, Proc.EMSA 47:90–91 Szedon JR, Sandor JE (1965) The effect of low energy electron irradiation of metal-oxidesemiconductor structures, Appl.Phys.Let. 6-9:181–182
2 LVSEM for Biology
105
Talmon Y (1984) Radiation damage to organic inclusions in ice, Ultramicrosc. 14:305–316 Tamura N, Saito H, Ohyama J, Aihara R, Kabaya A, (1988) Field emission SEM using strongly excited objective lens, Proc.EMSA 68:69–70 Tanaka K (1980) Scanning electron microscopy of intracellular structures, In: International Review of Cytology, NY: Academic Press, 97–115 Tanaka K (1981) Demonstration of intracellular structures by high-resolution scanning electron microscopy, Scanning Electron Microsc. II:1–8 Tanaka K (1990) High-resolution scanning electron microscopy in biology, Proc. XIIth ICEM Mtg. 14–15 Thompson-Coffe C, Coffe G, Schatten H, Mazia D, and Schatten G (1996) Cold-Treated Centrosome: Isolation of the Centrosomes from Mitotic Sea Urchin Eggs, Production of an Anticentrosomal Antibody, and Novel Ultrastructural Imaging. Cell Motil. Cytoskel. 33: 197–207. Thon F (1965) Z.Naturforsch 20a:154 Thornhill JW, MacKintosh IM (1965) Application of the scanning electron microscope to semiconductor device structures, Microelectronics and Reliability (GB) 4:96–100 Thornley RFM (1960) Recent developments in scanning electron microscopy, Proc.EUREM 173–176 Thornley RFM, Cartz L (1962) Direct examination of ceramic surfaces with the scanning electron microscope, J. Am.Ceram.Soc. 45:425–428 Todokoro H, Fukuhara S, Sakitani Y (1980) Low acceleration SEM, Proc.EMSA 38:70–71 Todokoro H, Fukuhara S, Komoda T (1983) Stroboscopic scanning electron microscopy with 1 keV electrons, Scanning Electron Microsc. II:561–568 Tuggle DW, Watson SG (1984) A low voltage field emission column with a Schottky emitter, Proc.EMSA 42:454–457 Tuggle DW, Swanson LW, Gesley MA (1986) Current density distribution in a chromatically limited electron probe, J. Vac.Sci.Tech. 4-1:131–134 Vanderburgh DJ, Ackerley CA, Lynn DH, Anderson RC (1987) The use of silver nitrate staining and backscattered electron imaging to visualize nematode sensory structures, Scanning Microsc. 1-IV: 1881–1886 Venables JA, Harland CJ,(1973) Electron backscattering patterns - A new technique for obtaining crystallographic information in the SEM, Phil.Mag. 27:1193–1200 Vermeulen JP (2004) 12 Years Zeiss Gemini FESEM Technology. Imaging & Microsc. Spring. 34–35 Volbert B (1984) Low voltage scanning electron microscopy and its applications, Electron Opt.Rep. 31:44–53 von Ardenne M (1938) The scanning electron microscope: Practical construction (in German), Z.Phys. 19:407–416 Walker CGH, Prutton M, Dee JC, ElGomati MM, Cowham MJ (1989) An ultra high vacuum compatible backscattered electron detector, Inst.Phys.Conf.Ser. 98 (Chpt. 12), 555–558 Wall JS (1980) Contamination in the SETM at ultra high vacuum, Scanning Electron Microsc. I:99–106 Walther P, Hentschel J, Herter P, Müller T, Zierold K (1990a) Imaging of intramembranous particles in frozen-hyrated cells (Saccharomyces cerevisiae) by high-resolution cryo SEM, Scanning 12:300–307 Walther P, Herter P, Hentschel J, Hentschel H (1990b) High-resolution scanning electron microscopy of kidney tissue using cryo-techniques, Proc. XIIth ICEM Mtg. 8–9 Walther P, Autrata R, Chen Y, Pawley JB (1991) Backscattered electron imaging for highresolution surface SEM with a new type YAG detector, Scanning Microsc. 5: 301–310 Walther TC, Fornerod M, Pickersgill H, Goldberg M, Allen TD and Mattaj IW (2001) The nucleoporin Nup153 is required for nuclear pore basket formation, nuclear pore complex anchoring and import of a subset of nuclear proteins The EMBO Journal 20: 5703–5714 Walther TC, Pickersgill HS, Cordes VC, Goldberg MW, Allen TD, Mattaj IW, Fomerod M (2002) The cytoplasmic filaments of the nuclear pore complex are dispensable for selective nuclear protein import, J. Cell Biol, 158-1:63–77
106
J. B. Pawley
Waltzthony D, Moor H, Gross H (1981) Ice crystals specifically decorate hydrophilic sites on freeze-fractured model membranes, Ultramicrosc. 6:259–266 Wang YL, Raval A, Levi-Setti R (1989) Dendritic oxide growth on the surface of liquid gallium, Scanning Microsc. 3 III:731–737 Watabe T, Hoshino T, Harada Y (1978) The visibility of individual ferritin particles in a scanning electron microscope with a field emission gun, Ultramicrosc. 3:19–27 Wells OC (1974) Resolution of the topographic image in the SEM, Scanning Electron Microsc. 1–8 Wells OC (1975) Scanning Electron Microscopy, New York, NY:McGraw Hill Wells OC (1978) Note on signal-to-noise ratio (SNR) in the scanning electron microscope, Scanning Electron Microsc. I:99–302 Wells OC (1979) Effects of collector take-off angle and energy filtering on the BSE image in the SEM, Scanning 2:199–216 Wells OC, Oatley CW (1959) Factors affecting contrast and resolution in the SEM, J. Electron Control 7:97–111 Wells OC, Bremer CG (1970) Collector turret for scanning electron microscope, Rev.Sci.Inst. 41:1034–1037 Wells OC, Broers AN, Bremer CG (1973) Method for examining solid specimens with improved resolution in the scanning electron microscope (SEM), Appl.Phys.Let. 23-6:353–355 Welter LM, Coates VJ, (1974) High-resolution scanning electron microscopy at low accelerating voltages, Scanning Electron Microsc. 59–66 Wepf, R, Gross,H (1990) Pr/Ir/C, A powerful coating material for high-resolution SEM, Proc. XIIth ICEM Mtg. 6–7 Wepf R, Amrein M, Burkli U, Gross H (1991) Platinum-iridium-carbon, a high-resolution shadowing material for TEM, STM and SEM of biological macromolecular structures, J. Microsc. 163:51–65 Wergin WP and Pawley JB (1980) Recording and projecting stereo pairs of scanning electron micrographs, Scanning Electron Microscopy I:239–249 Wight SA, and Zeissler CJ (1999) Direct Measurement of Electron Beam Scattering in the Environmental Scanning Electron Microscope Using Phosphor Imaging Plates, Scanning, 22: 167–172 Wildhaber I, Gross H, Moor H (1985) Comparitive studies of very thin shadowing films produced by atom beam sputtering and electron beam evaporation, Ultramicrosc. 16:312–330 Winkler H, Wildhaber I, Gross H, (1985) Decoration effects on the surface of a regular protein layer, Ultramicrosc. 16:331–339 Wolf ED, Everhart TE (1969) Annular diode detector for high angular resolution pseudo-kikuchi patterns, Scanning Electron Microsc. 41–44 Yamada S, Ito T, Gouhara K, Uchikawa Y (1991) Electron count imaging in SEM, Scanning 13:165–171 Yamazaki S, Kawawoto H, Saburi K, Naktasuka H, Buchanan R (1984) Improvement in SEM gun brightness at low-kV using an intermediate extraction electrode, Scanning Electron Microsc. I:23–28 Yamazaki S, Sato T, Aota S, Buchanan R (1989) Dual stage SEM with thermal field-emission gun, Proc.EMSA 47:94–95 Yokota Y, Hashimoto H, Yamaguchi T (1990) Electron radiation damage of natural zeolites at room and low temperature, Proc. XIIth ICEM Mtg. 4:808–809 Zach J, Rose H (1986) Efficient detection of secondary electrons in low voltage scanning electron microscopy, Scanning 8-6:285–293 Zach J (1989) Design of a high-resolution low voltage scanning electron microscope, Optik 83-1: 30–40 Zobacova J, and Frank L (2003) Specimen charging and detection of signal from non-conductors in a cathode lens-equipped scanning electron microscope. Scanning 25-3:150–156.
Chapter 3
The Aberration-Corrected SEM David C. Joy
Introduction Professor Albert V Crewe once famously remarked that “imaging in the SEM is like looking at the world through the bottom of a beer bottle.” This comment aptly recognizes the fact that the lenses in an SEM are very far from perfect and that as a result, they drastically restrict the imaging potential of the instrument. This situation arises because all electron optical lenses intrinsically suffer from aberrations that degrade their performance. While it has long been a goal of microscope designers to eliminate these aberrations and so enhance the imaging performance of the SEM, it is only within the last few years that viable techniques to correct aberrations have become commercially available (Zach & Haider 1995; Krivanek et al. 1999). The aim of this chapter is to examine not so much how aberration correction is performed, but rather to investigate what the advantages and disadvantages of an aberration-corrected SEM (ACSEM) might be for the microscopist.
Aberrations in SEM Optics The imaging performance of an SEM is determined by the minimum size of the electron probe that it can produce and by the beam current in that probe. Unfortunately, the spot size cannot be made arbitrarily small, and the current cannot be set as high as might be desirable because of the electron-optical aberrations of the lenses and the limited brightness of the sources. In an idealized simple SEM (see Fig. 3.1), the diameter dg of the focused beam at the specimen would be equal to M.s where M is the demagnification of the objective lens and s is the effective source size of the electron gun. With such an arrangement, dg could be made as small as desired by increasing the amount of demagnification, and the current profile across the electron probe would, like that of the electron source, be Gaussian (i.e., bell-shaped) in form. The current in the probe depends on the brightness β (amps/cm2 /steradian) of the electron source. The brightness of the focused beam is equal to the brightness of the source, because brightness is conserved (Goldstein et al. 2003), so calculating H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
107
108
D. C. Joy
Fig. 3.1 Schematic ray diagram of a basic SEM
the current density (amps/cm2 ) per unit solid angle (steradian) at the focal point from the geometry and equating this to the gun brightness β, we obtain β=
1 IB 4I B . = 2 2 2 πdg2 /4 π α 2 π α dg
(3.1)
This brightness equation shows that for a given spot diameter dg, the amount of current Ib in the probe would depend on α2 because β is constant. The probe current could therefore be raised as much as desired by increasing α up to the point where χ(= α.M) and the cone of illumination accepted from the source was equal to the total angle αmax over which the gun emitted electrons. This idealized situation is never encountered in practice because there are many effects that degrade the performance of the lenses. The most important of these are listed below in Table 3.1, which classifies aberrations by their order (i.e., the symmetry of the error that they introduce into the image) and gives their common name (Hawkes & Kasper 1985): After correction of lower order aberrations, the second order chromatic and fifth order spherical aberration, together with residual coma and third and fourth order astigmatism, still remain and will become the most significant contributions.
Order 1 2 3
Table 3.1 Definition of Orders of Aberration Name of Aberration Defocus Chromatic Spherical
In addition, 1st and 3rd order astigmatism and 2nd order axial coma can be corrected by stigmators and controlled by precision alignment of the column.
3 The Aberration-Corrected SEM
109
Spherical Aberration The effect of spherical aberration is shown schematically in Fig. 3.2. In the ideal lens of Fig. 3.1, all of the electrons leaving the source point s are brought to a focus at the beam spot dg . In any real microscope, the rays of electrons diverging from the object and traveling at small angle to the axis of the lens arrive at the Gaussian focus point Fg . Electrons traveling through the lens at a high angle to the axis, however, are brought to a focus at points closer to the lens, as shown. In this situation, a point object no longer gives a unique point image, but rather many image points stretched out along the optic axis. The nearest approximation to a best image is now the focal position where the different overlapping cones of electrons have the smallest overall radius. This location is known as the disc of least confusion (DOLC) and occurs somewhere between the lens and the Gaussian focus point. If the object (in this case, the electron source) is of zero size, then the diameter of the disc of least confusion dsph is dsph =
1 .Cs α 3 2
(3.2)
where α is the maximum convergence angle for any ray traveling through the lens, and Cs is the spherical aberration coefficient. Cs is a constant of the lens, has the physical units of a length, and to a first approximation is equal in magnitude to the working distance (WD) of the lens. The effect of spherical aberration (see Fig. 3.3) is to distort the beam profile by making it narrower at the center but broader at the outside edges. The minimum beam size varies as α3 (see Equation 3.2) while the current in the probe varies as α2 (see Equation 3.1), so attempting to increase the current in the beam by raising α inevitably and rapidly leads to a much enlarged probe size. The existence of spherical aberration, therefore, forces the user to compromise between the ultimate spot size and the amount of current contained in the beam. Glass lenses in an
Fig. 3.2 The origin of spherical aberration
110
D. C. Joy
Fig. 3.3 The profile of a beam dominated by spherical aberration
optical microscope also suffer from spherical aberration, but concave and convex lenses have aberrations that are opposite in sign (i.e., negative or positive) and so compound lenses formed by combining concave and convex elements can have zero net aberration. Unfortunately it was shown by Otto Scherzer (1936) that not only do all electron optical lenses of conventional design exhibit spherical aberration, but also that the aberration is always of the same sign, so combining such lenses always increases the aberration. As a result, until recently, the only practical method for reducing Cs was to design the objective lens to have a very short focal length—i.e., put the specimen as close to the objective lens as possible—because this minimizes Cs, as noted above. For example, in the Hitachi S5000 FEGSEM, the Cs , Cc coefficients of the advanced immersion lens (focal length ∼200 micrometers) are of the order of 50 micrometers, and so are comparable with the values displayed by some aberration-corrected designs. Such a level of performance, however, results from a restrictive lens geometry that limits both sample size and the available tilt and translation. It is not, therefore, a general solution to the problem.
Chromatic Aberration Chromatic aberration is the error caused because the focal length of the objective lens in the SEM varies with the energy of the electrons passing though it—it is longer for higher energies and shorter for lower energies (see Fig. 3.4). The beam of electrons emerging from the gun is never monoenergetic, but instead has some energy width E whose magnitude depends on the nature and the operating temper-
3 The Aberration-Corrected SEM
111
Fig. 3.4 The origin of chromatic aberration
ature of the source. For example, a conventional tungsten hairpin thermionic emitter that runs at 2800 ◦ K has a E of about 2eV, while a Schottky field emitter that runs at about 1700 ◦ K has a E of about 1eV, and a cold field emitter that runs at 300 ◦ K has a E of only about 0.3eV. The beam of electrons passing through the lens is therefore dispersed depending on its energy spread, giving a range of focal positions along the axis of the lens. The minimum diameter, or DOLC, of the beam dChr is then given as
dChr = Cc .α
E E0
(3.3)
where Cc is the chromatic aberration coefficient of the lens. Like Cs , this has the dimensions of a length and magnitude that is again about equal to the WD of the lens. Chromatic aberration becomes important at low-beam energies because it varies as 1/E0 , and so the effect—which usually can be neglected at 30 keV—becomes dominant in the low energy range of 2 keV or below. For example, if E is 2eV, E0 is 1keV, Cc is 1cm, and α is 10 millirads, then the chromatic aberration disc is 200 nm in diameter. The effect of chromatic aberration on the beam profile (see Fig. 3.5) is to redistribute current—which should be on the center axis of the beam—to the outside edge of the beam by forming a skirt. This skirt increases the diameter of the probe, which lowers the resolution and reduces the signal contrast in the image. Although all lenses suffer from chromatic aberration, optical lenses can be corrected (made achromatic) by making compound lenses of different glasses. This option is not available for electron-optical lenses, so chromatic aberration is always present to degrade the beam in the SEM.
112
D. C. Joy
Fig. 3.5 The profile of a beam dominated by chromatic aberration
Diffraction Diffraction is not an aberration of the lens, but (because electrons are waves as well as particles) an inescapable consequence of passing an electron beam through a lens. If we look at the beam profile at the focus of even a perfect lens, then the profile has the form shown in Fig. 3.6—a central maximum surrounded by several additional peaks of lower intensity. This Airy disc, named after the scientist who first discussed it, is formed by interferences between electrons traveling along the axis of the lens and electrons scattered at the edge of the lens. The width of the beam ddiff between the first order zeros (gaps) in the Airey disc is given as ddi f f = 0.61
λ α
(3.4)
where λ is the wavelength of the electron and α is the convergence angle of the beam. Equation 3.4 is also a particular example of the Abbe. diffraction limit of an optical system. For low-energy electrons 1.26 λ = √ nm E0
(3.5)
where E0 is the beam energy in eV so that the wavelength (at 1keV and below) is a significant fraction of a nanometer, and the diffraction disc (for a typical convergence angle of 10 mrads) reaches several nanometers in magnitude. If the spherical
3 The Aberration-Corrected SEM
113
Fig. 3.6 The beam profile produced by diffraction. The diffracted beam width is the value between the first zero locations in the intensity profile
and chromatic aberration contributions can be reduced, then the convergence angle can be increased and the effect of diffraction becomes less significant. If diffraction could be controlled, however, it would be possible to control the depth of field of the image, which is often an important property of the instrument.
Optimizing Spot Size in the Presence of Aberrations An initial estimate of the effective probe size, d, which results from all of the aberration effects discussed earlier can be made by adding the aberration components in quadrature 2 λ 2 1 E 2 3 + + Cc α + dg2 Cs α d = 0.61 α 2 E0 2
(3.6)
Without spherical and chromatic aberrations, and diffraction, d would just be equal to dg —the demagnified image (M.s from Fig. 3.1) of the electron source. With aberrations, the probe size is always larger than dg and varies in a complex way with α and E0 . Figure. 3.7 shows how this works for a low voltage SEM with typical lens properties and operating conditions—a WD of 4 mm, a Cs value of 5 mm, a Cc value of 5 mm, a beam energy of 1keV, a E spread of 0.7 eV, and dg = 1.5 nm. The effect of each of the different aberrations is shown separately, together with the final
114
D. C. Joy
Fig. 3.7 The magnitude of spherical, chromatic, and diffraction aberrations and the total probe size estimated from quadrature addition for an uncorrected SEM (see details in text) operating at 1 keV
probe size estimated by quadrature (see Equation 3.6) and the relative magnitude of the current in the probe. Because α is varied from 1 milliradian to 20 milliradians, d initially falls from a value of about 40 nm, reaching a minimum of about 18 nm when α is 4 mrads, and then rapidly increasing to become greater than 80 nm when α is 20 mrads. The reasons for this behavior are clear from the different contributions visible on the graph. At small convergence angles, the beam is dominated by the diffraction effect that increases as 1/α. This can only be minimized by increasing the angle, α, which also increases the current in the probe (∼α2 ). But as α is increased from 1 millirad, the combined effects of spherical and chromatic aberration rapidly increase in magnitude, so the quadrature probe size goes through a minimum and then begins to rise sharply. The optimum operating condition—the one that gives the smallest probe size and the highest current simultaneously—is at the bottom of the dip in the profile. That position is often called the diffraction-limited condition, because when employed it places the SEM in the same condition as an optimally adjusted optical microscope in which the size of the smallest visible object will be equal to the diffraction limit of Equation 3.4. While this provides a relatively small spot size, high-image resolution is achieved at the expense of limiting the beam current IB that is available. This trade-off between resolution and signal-to-noise ratio is unwelcome because signal-to-noise, which varies as (IB )1/2 , determines the quality and quantity of information in the image and we always want the highest possible beam current. Combining Equations 3.1 and 3.6 gives
3 The Aberration-Corrected SEM
2 βπ 2 α 2 1 λ 2 E 2 2 3 IB = − 0.61 − Cc α . d − Cs α 4 2 α E0
115
(3.7)
showing that aberrations not only make the spot size larger (see Equation 3.6), but also reduce the current in the probe—first, by diverting electrons from the probe (see Equation 3.7), and second, by restricting the magnitude of α to minimize the unwanted contributions to probe size. Clearly, the key to improving SEM performance is finding some way to modify the objective lens so that the aberrations can be reduced or eliminated.
Aberration Correctors Almost from the day in 1937 when Otto Scherzer showed that electron-optical lenses displayed positive-definite aberrations, attempts have been made to design and construct devices to correct such aberrations. In a later paper (Scherzer 1947), it was shown that the problem of inherent aberrations did not extend to lenses that were either not cylindrically symmetrical or in which the electron energy or lens excitation varied with time. In the decades since Scherzer’s papers, there has been a large amount of work to develop devices that can exploit these exceptions and reduce or eliminate the aberrations of conventional lenses. Although not often recognized as such, the stigmator (Hillier and Ramberg 1947) now fitted to every electron microscope is the earliest example of an aberration corrector. The stigmator is a weak quadrupole lens that is used to cancel the geometric (Type 3) stigmatic error of a cylindrical lens, such as the objective lens. Deltrap (1964) actually demonstrated a corrector consisting of quadrupole and octapole elements that resulted in zero net spherical and chromatic aberration. Unfortunately, his system was difficult to align because he did not have access to computers to automate and control the procedure, was not stable over long periods of time, and had a high level of uncorrected astigmatism. Joy and Newbury (1972) were able to show experimentally that a time-dependent correction for Cs —again proposed by Otto Scherzer (1947)—could also be successfully implemented in a SEM. Their application was for a specialized type of operation, however, and could not readily be extended to normal imaging conditions. At the time this book was written (early 2006), two commercial systems from CEOS (Zack and Haider 1995) and NION (Krivanek et al. 1999) were available. These devices both rely on a combination of multipole magnetic lenses, which when placed in the optical path of the objective lens can be used to control one of the major aberrations: • Quad-Pole (4) element lens can be used to correct first order chromatic aberration • Octa-Pole (8) element lens can be used to correct third order spherical aberration • Hexa-Pole (6) element lens can be used to correct the aberrations caused by the other multipoles
116
D. C. Joy
Correctors based on electrostatic—rather than electromagnetic—electron optics have also been employed. The ingenious cathode lens configuration of Mullerova & Frank (1993) that functions by rapidly slowing down the electron beam in an abrupt electrostatic field gradient within the magnetic field of the lens, also effectively acts as a corrector. If the initial energy of the beam is E0 , the final energy of the beam after deceleration is EFinal , and the distance over which deceleration is accomplished is L, then the effective aberration coefficients of the combined arrangement satisfy the relationship Cc = −Cs = L.(EFinal /E0 )
(3.8)
showing that at first approximation, the aberration coefficients of the original lens have no effect on performance. By keeping L small and EFinal much lower than E0 , impressively low spherical and chromatic aberrations can be obtained. This arrangement, although not often recognized as such, is the most widely used aberration corrector because it appears in several different commercial SEMs and has made possible the extension of low-voltage scanning microscopy to the ultra-low energies of 100eV or less. Because aberration correction is a rapidly evolving field that requires a detailed knowledge of electron optics, and whose analysis involves the use of advanced mathematical concepts, it will not be treated in any detail here. Instead we will concentrate on asking more basic questions, such as • What kind of aberration correction is best for low-voltage SEM? • How much will the performance of the SEM be improved by aberration correction? • Are there any drawbacks to correcting any or all of the (third order) aberrations?
What Kind of Aberration Correction is Needed? There are a number of options available when deciding how to apply aberration corrections to a low-voltage SEM. Basically these would be: • Reduce spherical aberrations only • Reduce chromatic aberrations only • Reduce both Cs and Cc by a dual-purpose corrector element such as the cathode lens configuration or a combination of multipole lenses • Applying some form of correction to reduce the magnitude of the diffraction limit • All of the above Clearly, the more corrections that are applied, the more elaborate the system becomes, and both cost and complexity get out of hand. The choice also depends on what the aim of the aberration correction is defined to be. Is the goal to produce a smaller spot size in the beam, to provide more beam current, or to improve (or at
3 The Aberration-Corrected SEM
117
least control more effectively) some other parameter of the imaging performance? A reasonable starting point to answer such questions is to investigate which of the possible corrections (Cs , Cc , or diffraction) gives the biggest benefit in terms of the current available and the minimum probe size that can be produced. An estimate of the results can again be made from Equation 3.9: d2 =
2 2 λ 1 E 2 + + dg2 Cs α 3 + Cc α α 2 E0
(3.9)
If, for example, Cc could be reduced to zero, then the optimum aperture αopt and the corresponding minimum probe size dmin would be, respectively: −1/4 1/4
αopt = Cs
λ
and
1/4
dmin = Cs λ3/4
(3.10)
Alternatively, if Cs could be reduced to zero, then the optimum aperture αopt and the corresponding minimum probe size dmin would be: αopt
E −1/2 1/2 = Cc λ E0
and
dmin
E 1/2 1/2 = 1.4 Cc λ E0
(3.11)
If, for the sake of example, we assume that in the first case Cs was 5 mm, and in the second case that Cc was 5 mm, then at 1 keV, αopt is about 5 mrads, and dmin is 0.75 nm. On the other hand, in the second case where Cs is zero and Cc is taken to be 5 mm, then, using a Schottky emitter with an energy width E = 0.7eV, the corresponding aperture angle αopt is 0.6 mrads and dmin is 9.5 nm. Although the relative magnitude of the apertures and probe sizes vary with the assumptions that are made, for a wide range of plausible assumptions, it is always true that in SEMs operating below about 3 keV, chromatic aberration degrades the probe size and beam current more seriously than does the spherical aberration. In seeking to improve LVSEM performance, therefore, the first priority must be to minimize chromatic aberration. Direct correction of Cc is a promising approach, because the optical elements for accomplishing this can likely be shared with those of the Cs correctors. Because αopt and dmin vary as (Cc .E/E0 )1/2 (as compared 1/4 with the functional Cs variation for spherical aberration), even incremental improvements in the magnitudes of either Cc or E produce useful benefits. Thus, the monochromators that have been developed by several groups and can produce electron beams with an energy spread of below 100 meV provide a marked reduction in the effects of chromatic aberration. Monochromators, however, reduce the brightness of the electron source by a considerable fraction (30–90%) and so reduce the signal-to-noise ratio of the image. A further alternative would be to consider unconventional electron sources. One example would be a photocathode (or negative affinity) emitter, because these intrinsically have an energy width of the order of 50 meV while still producing relatively high brightness and current. Although no such source has yet been demonstrated operating in a high-resolution
118
D. C. Joy
mode, experience in the semiconductor industry with defect review tools using such sources seems very promising. To illustrate the benefits of controlling chromatic aberration, Fig. 3.8 plots the spherical, chromatic, and diffraction components of the probe size, together with the quadrature value and the relative beam current, for a system operating at 1 keV with a Schottky source, and in which the chromatic aberration coefficient Cc has been reduced to 50 micrometers while keeping Cs still at 5 mm. The comparison with Fig. 3.7 is instructive. With chromatic aberration effectively eliminated, the minimum probe size now falls to about 5 nm with the diffraction minimum occurring at a convergence angle of about 10 mrads. The spot size-limited resolution has therefore improved by a factor of almost 4x, and the current into that probe has increased by more than 6x. It is also worth noting that the minimum in the quadrature probe size plot, as a function of the convergence angle, is both broader and flatter than when chromatic aberration is a factor, thus giving the operator a wider range of close-to-optimum operating conditions. If the Schottky source with its relatively wide energy spread of 0.7 eV was to be replaced by a negative affinity photo-cathode source with an energy spread of 0.05 eV, then a comparable level of performance could be achieved even with a Cc coefficient approaching 1 mm. Once chromatic aberration is controlled, it becomes worthwhile to try and minimize spherical aberration using one of the approaches outlined earlier. Figure 3.9
Fig. 3.8 The magnitude of spherical, chromatic, and diffraction aberrations and the total probe size estimated from quadrature addition for an aberration-corrected SEM (see details in text) operating at 1 keV
3 The Aberration-Corrected SEM
119
Fig. 3.9 Comparison of images recorded (a) without and (b) with aberration correction recorded on a JEOL 7555 LSI Inspection CD-SEM at a working distance of 4 mm, using a Schottky electron source, and a beam energy of 1 keV. (Images used by permission of JEOL USA Ltd.)
shows how reducing both chromatic and spherical aberration will affect the electronoptical performance of the SEM. Once again, the SEM is assumed to be operating at 1 keV with a Schottky emitter, and now both Cs and Cc are taken to be 50 micrometers in value. Noting that the plot is now on an expanded vertical scale, it can be seen that the minimum probe size is now about 2 nm and is achieved at a convergence angle of 30 mrads. In fact, the probe size is almost constant, from 15 to 45 millirads, allowing the probe current and the depth of field to be controlled by selection of the aperture without penalizing the resolution. The quadrature approximation is probably pessimistic, however, because the various contributions to the probe size are not all Gaussian in form. A more realistic assessment can be made from a ray-tracing analysis (Cliff and Kenway 1985) that uses a computer model of the optical properties of the SEM column to track electron trajectories as they travel from the gun, through the lenses, and onto the specimen surface. In this method, the resolution is then calculated by scanning this model probe across an imaginary edge and determining the distance over which the signal rises from 10% to 90% of the maximum, which is the standard definition for SEM resolution (Joy 1972). The 10%–90% probe resolution for the uncorrected system at α = 4 mrads is found to be close to 5 nm—as compared to the quadrature estimate of 18 nm—while for the fully corrected case, the 10%–90% figure is 1.8 nm (at α = 25 mrad). Ray tracing, therefore, also predicts a significant enhancement in the imaging resolution of the system as the result of aberration correction,
120
D. C. Joy
giving numbers that are generally close to, but more optimistic than, those of the simple quadrature model. A detailed analysis of the ray traces confirms that a key contribution of this improvement comes from the elimination of the chromatic aberration skirt that degrades the rise-time of the probe across the edge. In summary, as the effect of chromatic aberration is reduced, the optimum aperture angle becomes larger. If spherical aberration is also reduced, this trend is enhanced and the aperture angle becomes still larger while the minimum probe size becomes still smaller. As an additional benefit, the rise in the optimum aperture also means that the additional contribution from diffraction (which varies as 1/α) automatically becomes smaller. Diffraction correction is therefore not going to be a necessary step towards achieving the best performance even at the lowest energies.
How would an ACSEM perform? The discussion so far has demonstrated that an SEM equipped with a corrector for chromatic aberration, or for both chromatic and spherical aberration, would—for the same electron source operating at the same energy—put more current into a smaller spot size than an uncorrected instrument. The expectation, therefore, would be that an ACSEM would produce images displaying better resolution and a higher signalto-noise ratio. Experimental proof of this statement is, however, not readily available at this time because only a limited number of ACSEMs exist and few results from these machines have been published. One example (Honda & Takashima 2004) shows a gold-on-carbon sample imaged in secondary electron mode on a JEOL 7555 LSI Inspection CD-SEM at a working distance of 4 mm, using a Schottky electron source and a beam energy of 1 keV. Figure 3.10(a) shows the image without—and Fig. 3.10(b) the image with—aberration correction. At the WD of 4 mm in the uncorrected case, the objective lens has a Cs of 7.5 mm and a Cc of 3.6 mm, giving an optimum beam-convergence angle of about 8 milliradians for a resolution of 5 nm at 1 keV. If the first order chromatic and third order spherical aberrations are completely compensated by the corrector, then the resolution, limited by the second order chromatic and fifth order spherical aberrations, is about 1.5 nm for a beam-convergence angle of about 40 milliradians. Because the micrographs were not recorded at high magnification, an assessment of the resolution is not easy, but the corrected image appears to visually show more detail and has better-delineated edges. In addition, gold particles in the corrected image show a clear edge bright line outlining each metal grain. The width and visibility of this edge line—produced by the high-resolution SE1 component of the secondary signal—is a good guide to the resolution achieved, so its presence in the corrected image and its absence in the uncorrected case is a good indication of enhanced performance. The apparent signal-to-noise ratio of the corrected instrument also appears to be better than that of the uncorrected tool, although it is not certain that the micrographs were recorded under comparable conditions. It is evident that assessing the benefit (or otherwise) of an ACSEM will require the use of well-characterized samples and the application of methods that can analyze images to yield reliable quantitative data on the resolution
3 The Aberration-Corrected SEM
121
Fig. 3.10 “(a) -image of gold on carbon without correction at 1keV (b) corresponding image with aberration correction”. Images are courtesy JEOL (USA) Inc
and signal-to-noise ratio. Computer programs for this purpose have recently been described (e.g., Joy 2002) but their application is limited by the lack of suitable and reproducible test samples at the nanometer scale. While a comparison of the apparent resolution—and signal-to-noise ratio—of corrected and uncorrected images is valuable, it is not the most sensitive test of the true benefits provided. A more complex but revealing view of the change in performance that occurs when aberration correction is implemented can be obtained by looking at the optical transfer function (OTF) of the SEM under different conditions. While this type of analysis is widely applied to high-resolution TEMs, it is not often used for SEMs, although it is potentially a valuable tool (Maulny & Fanget 2000). This approach applies the ideas of Fourier imaging to the SEM. The microscope is considered as a channel, through which information is being passed. In a perfect SEM, the Fourier power spectrum of the signal output would be identical to the Fourier power spectrum of the input—i.e., the specimen under observation. In any real SEM, the inputs and outputs are not identical because higher spatial frequencies (that is, the signal components from rapidly changing features on the sample surface) are transferred less efficiently by the microscope than low frequencies. At any spatial frequency ϖ (units of nm−1 ), the ratio between the modulus of the output Fourier spectrum and the input Fourier spectrum is the OTF (ϖ). For perfect transfer, the OTF is equal to unity, but in any real system, the OTF is less than unity and falls with increasing spatial frequency. Ultimately, at some high enough frequency, the OTF becomes so small that the incoming signal is lost in the noise of instrument and no further information is transferred. The spatial frequency ϖmax at which this occurs is the Fourier resolution limit of the microscope (Joy 2002) and represents the useful imaging limit of the microscope. In an ideal SEM, OTF(ϖ) would be unity from ϖ = 0 to ϖ = ϖmax , and then drop to zero. In real instruments, however, the OTF has a complex variation with beam voltage that implies that different spatialfrequency components in the image are passed with different efficiency, resulting in a distortion of the image detail. Figure 3.11 compares the computed OTF (Sato & Orloff 1992) of an uncorrected SEM—assumed for the purposes of this calculation to have Cs , Cc = 3 mm, and to
122
D. C. Joy
Fig. 3.11 OTF of uncorrected and corrected machine
be operating at 1 keV using a Schottky field emission source with E = 0.7eV—with the identical but corrected system in which Cs and Cc have both been reduced to 50 micrometers. The uncorrected system has an OTF that falls rapidly as the spatial frequency is increased, and becomes less than 0.01 (1% transmission) at a spatial frequency o corresponding to a resolution cut-off of about 2.4 nm—below the resolution of most SEM lens controllers. By comparison, the OTF of the aberration corrected tool has a higher cut-off frequency, corresponding to a resolution limit approaching 1 nm and significantly enhanced information transfer at all lower frequencies, as well as a factor of 10x for 2 nm size detail. An ACSEM thus provides superior ultimate resolution and enhanced performance at all lower spatial frequencies, making the attainment of optimized imaging simpler and more reproducible even for less skilled operators. Although this single example of data is hardly a definitive proof of the benefits of aberration correction in the SEM, it is good evidence that it can provide important enhancements to the performance. The biggest potential benefit of aberration correction is that it decouples the performance of the instrument from the geometrical constraints of the objective lens. As discussed earlier, the aberration coefficients of a lens scale with its focal length (working distance), so a high-performance lens must have a small focal length. This is achieved by building a lens with a small gap between the upper and lower pole-pieces—an arrangement that seriously restricts the size of samples that can be inserted and the amounts of specimen motion and tilt that is possible. When aberration compensation is applied, high performance will
3 The Aberration-Corrected SEM
123
no longer be synonymous with a short working distance and it will become possible to achieve competitive levels of performance while providing sufficient space for large samples, multiple detectors, and ancillary devices such as heaters or coolers for environmental microscopy, or mechanical probes for surface interactions. This, of course, assumes that external disturbances such as electro-magnetic fields and vibrations are negligible or properly shielded against, which may be a challenging task. The promise of a new freedom for lens and specimen chamber design, however, is a major reason for moving to an aberration-corrected SEM.
Will There be Problems Associated with Using an ACSEM? The undoubted benefits of aberration correction must be weighed against some potential problems. The most significant of these is the degradation of the depth of field (DoF) of the image (Goldstein et al. 2003)—the vertical distance over which a given feature appears to be in focus. The high DoF of a conventional SEM is often considered to be one of its most valuable attributes, because it permits the viewing of surfaces with large amounts of topography. The DoF is usually defined as Depth of Field =
Pixel Size α
(3.12)
where α is the beam-convergence angle, and the pixel size is the effective width (in nm) of the pixels displaying the image. Consider an uncorrected but high-quality SEM with a standard digital display of ∼1000 × 1000 pixels. When operating at its maximum useful magnification of 200 kX or so, the pixel size is of the order of 1 nm—a figure that is comparable with the anticipated image resolution of 1-2 nm α—which must be kept small (as discussed earlier) to limit aberrations that will be of the order of 10 milliradians, giving a DoF of 100 nm, which is adequate to successfully image all but the most highly topographic surfaces. In an aberration-corrected SEM (by comparison), the maximum useful magnification will be higher—for example, x1,000,000—resulting in a pixel size of 0.2 nm, and the optimum value of α will be increased to about 40 milliradians. The DoF is now just 5 nm, so traversing a specimen will require constant readjustment of the focus, and imaging anything but an almost atomically flat surface will result in a visible loss of image detail. This problem can be partially circumvented by storing a through-focal series of images and using special purpose software to reconstruct a high DoF image, but this is time-consuming and exposes the specimen to a much-increased electron dose. When fully aberration corrected (e.g., Fig. 3.9), however, the probe size is substantially constant over a wide range of α values, allowing the DoF to be increased by a factor of 2-3x without seriously impacting the imaging performance, although at the expense of reduced current. Nevertheless, the fall in the DoF will change the manner in which the ACSEM is used, and may limit its effectiveness in applications such as the critical dimen-
124
D. C. Joy
sion (CD) metrology of semiconductor devices—where DoF is of major importance (Sato & Mizuno 2000). A second consideration is the need for the operator to be able to quickly and accurately adjust the aberration correctors so that the performance is actually improved rather than made worse. Aberration correctors on TEMs and STEMs can be adjusted precisely by forming a Ronchigram (Cowley 1986)—a special type of defocused diffraction pattern that can be obtained from electron transparent specimens—and optimizing its symmetry and other characteristics. This procedure can be automated and carried out to the necessary degree of accuracy within one or two minutes. Depending on the stability of the microscope and of the corrector, this procedure may need to be repeated every hour or so because improper aberration correction is worse than applying no correction at all. When imaging a bulk (i.e., non-electron transparent) sample in the SEM, however, nothing comparable to a Ronchigram is available, so aligning and optimizing the correctors will be more challenging. The principle to be followed is illustrated in Fig. 3.12, which shows how Joy & Newbury (1973) adjusted a spherical aberration corrector in an SEM. The sample used in Fig. 3.12(a) was a copper mesh grid with an 8-micrometer repeat spacing. The Cs value of the lens was about 30 mm (Cambridge Stereoscan Mk II SEM, vintage 1970). To adjust the system, the incident electron beam was set to rock its angle of incidence about the optic axis of the microscope. If Cs was to be zero, then (see Fig. 3.2, for example) all of the incident beams, at whatever angle to the axis, would reach the same focal point on the axis and the beam would move laterally but would simply rock about the Gaussian focus of the lens. The resultant image would then have infinite magnification because the area scanned on the sample would be zero. If Cs is not zero, however, then as the incident angle is varied, the point at which the beam cuts the optic axis will change and the beam will always be scanning a finite-sized area producing an image of the type shown in Fig. 3.12(b). If the strength of the Cs corrector is adjusted as the magnitude of Cs falls, the area scanned by the beam decreases and the apparent magnification of the image becomes higher as seen in Figs. 3.12(c) and 3.12(d). At the optimum adjustment (e) —where Cs is minimized—the magnification reaches a maximum value with the beam traversing an area less than about 4 micrometers in width as it rocks through an angle of about 5 degrees, corresponding to a corrected Cs value of about 8 mm. Still further adjustment reverses the sign of Cs (barrel distortion changes to pincushion distortion) and the apparent image magnification falls rapidly in Fig. 3.12(f) as the spherical aberration increases once more. This type of procedure could be applied to the optimization of a modern Cs corrector, if a suitably miniaturized grid structure could be provided. Because the technique requires the ability to rock the beam, however, it cannot be readily applied to arbitrary specimens and cannot aid with any adjustment for chromatic aberration, so its utility is limited. Honda & Takashima (2003) describe a setup method that relies on an initial step in which the excitation of the objective lens and of the corrector are set to predetermined theoretical values close to the anticipated optimum value. Next, Cc is corrected by observing the edge sharpness of a small circular feature as the beam energy is swept about its nominal value. When Cc is properly compensated, the focus of the beam, and hence the sharpness of the object, will
3 The Aberration-Corrected SEM
125
Fig. 3.12 Temporal correction of Cs (a) SEM image of 1 micron square grid, (b)-(f) successive beam rocking images of the grid recorded as the Cs correction is varied. The correction is optimized in 3.12(e). From Joy and Newbuy (1972)
not vary with the oscillation of the beam energy. Cs is adjusted by again observing a circular feature while oscillating the objective lens through focus. As long as Cs is not compensated, the intensity profile of the probe will resemble a volcano with enhanced edges but a depressed central region. This results in an image in which features are doubled away from best focus. Because the Cs corrector is adjusted towards its optimum value, the feature doubling disappears on both sides of correct focus. Finally hexa-pole correctors are applied to minimize the residual aberrations due to coma, and 3- and 4-fold astigmatism. Because these various steps interact with each other, several cycles of refinement are needed to get the best result, and
126
D. C. Joy
so the process is carried out under computer control. Unlike the Ronchigram case, where an analysis of the image provides numerical data on the residual aberration coefficients, the SEM procedure seems less quantified, so the skill and experience of the operator will be a consideration in the quality of the final result. For many users, a final (and possibly fatal) problem with aberration correction in the SEM is the cost involved. At the time this book was written (early 2006), aberration correction technology added about $500,000 to the cost of an instrument, placing such microscopes into the rarefied atmosphere of machines costing over $1 million in total. It must also be expected that, at least initially, service costs, up-time, and general reliability will also be impacted adversely. The judgement as to whether or not this extra expense and potential inconvenience is warranted can only be made by the user based on their needs and expectations. Eventually, the incremental cost will start to seem less worrisome, the benefits will seem more valuable, and this technology will become a standard part of the specification of any high-performance SEM. Such a move will, however, be aided by designs in which the corrector can be turned off completely, restoring the instrument to a conventional mode of use and permitting operators to choose how to configure the microscope for their own particular problems and levels of expertise.
Conclusions Because aberrations seriously impact the performance of the SEM, the advent of aberration correctors will have a major effect on the design, application, and performance of these instruments. Particularly in low-voltage scanning microcopy, where aberrations dominate the performance that is presently available, the acceptance and application of aberration-corrected instruments will lead to a substantial advance in state-of-the-art imaging. Acknowledgment The author is grateful to JEOL USA for permission to use Fig. 3.9, and to Charlie Nielson and William Powell of JEOL for providing helpful documentation. Thanks also to Drs. Harald Rose, Edgar Volkl, Michael O’Keefe, and Larry Allard for valuable discussion and insights.
References Cowley JM (1986) J.Elect. Micros. Tech., 3, 25–28 Deltrap JHM (1964), Ph.D.Thesis, University of Cambridge Goldstein J, Newbury DE, Joy DC, Lyman C, Echlin PE, Lifshin E, Sawyer L, and Michael J (2003) “Scanning Electron Microscopy and X-ray Microanalysis”, 3rd Edition, Kluwer Academic/ Plenum Publishers), 689pp Hawkes PW and Kasper E (1985) Principles of Electron Optics, Vol 1, London: Academic Press, p. 297–302 Hillier J and Ramberg EG (1947) J.Appl.Phys. 18:48–71
3 The Aberration-Corrected SEM
127
Honda K and Takashima S (2003) JEOL News 38:36–40 Joy DC (1974) Proc. 7th SEM Symposium, ed. O Johari, (IITRI:Chicago), p 327 Joy DC (2002) J. of Microscopy, 208:24–34 Joy DC and Newbury DE (1972) J. Mat. Sci., 7, p 715 Kenway PB and Cliff G (1984) Inst. Phys. Conf. Ser. 68:33–36 Krivanek OL, Delby N, and Lupini AR (1999) Ultramicroscopy 78:161–180 Maulny A and Fanget GL (2000) in Metrology 2000 edited by D Herr, Proc. SPIE Conf. on Micro-lithography, 3332:71–80 Mullerova I and Frank L (1993) Scanning 15:193–199 Sato M and Mizuno F (2000) J.Vac.Sci.Technol B 18:3047–51 Sato M and Orloff J (1992) Ultramicroscopy 41:181–192 Scherzer OL (1936) Zeitchrift Physik 101:593–603 Scherzer OL (1947) Optik 2:114–142, Zach J and Haider M (1995) Nucl. Instrum. and Methods in Physics Research (Section A), 363:316–325
Chapter 4
Noise and Its Effects on the Low-Voltage SEM David C. Joy
Introduction Noise is the single most important limiting factor in scanning electron microscopy. Because of the presence of noise, we are forced to operate the SEM to maximize the available beam current and the beam dose (current × time) at the expense of degraded image resolution, increased charging, and more sample damage. Recent developments in high-performance electron guns, aberration correctors, and lenses are all part of an attempt to attain control of the noise while still achieving ever higher levels of resolution. In this chapter, we will examine noise in the SEM, its origin and properties, its measurement, and how the properties of the detectors used for the collection of secondary emission (SE) electrons and backscatter electrons (BSE) signals affect the noise.
The Origin of Noise A typical dictionary definition (i.e., http://dictionary.reference.com) describes noise as something “that is loud, unpleasant, unexpected, or undesired; a disturbance that obscures or reduces the clarity of a signal.” The idea is clearly that noise is an unwanted addition to the environment, and is something from the outside world that intrudes on our inner senses. From this perspective, noise—because it is external— can be reduced or eliminated by closing the window, turning down the TV, or some other simple action. Unfortunately, noise in the SEM does not fit this definition. Figure 4.1 shows a micrograph of a highly polished piece of silicon wafer. This digitally recorded image was taken with a carefully stabilized beam current, and because the silicon itself has no surface detail or structure of any kind, we would expect that the signal level at every pixel in the image would be identical to that at every other. While this appears to be true to the human eye, we find a significant spread of brightness between the pixels if we take a histogram of the image (as seen in Fig. 4.2). While there is an obvious average value for the image level, all but a small fraction of the pixels are either brighter or darker than this average value, and we can quantify the magnitude of this variation by calculating the standard deviation of the histogram. If we equate the average value of the histogram with the signal H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
129
130
D. C. Joy
Fig. 4.1 Secondary electron image of a piece of polished silicon wafer
level S, then the standard deviation of the histogram represents the uncertainty—or the noise N—in the image. This uncertainty, or noise, in the scanning electron microscope is the evidence of the fluctuations that occur in the signal level observed from a particular pixel in the image, even under conditions where the incident beam, the sample, and the recording conditions are kept constant. This randomness is a result of the fact that electron production from the gun, and electron interactions with the specimen, are
Fig. 4.2 Histogram of the digital image in figure (1)
4 Noise and Its Effects on the Low-Voltage SEM
131
statistical in nature and consequently are different for every individual pixel. Thus, far from being an external disturbance imposed on the signal, noise in the SEM— often called shot noise or Johnson noise—is thus an integral property of the signal and can never be eliminated or turned off. The best that can be achieved is to maximize the signal-to-noise ratio (SNR). Noise might also be added into the image from other sources. The detector that collects the electron signal, for example, may be internally generating its own unwanted contribution and the electronic circuits that amplify and digitize the signal as it passes through the SEM could also add additional noise. On a modern SEM that is properly operated and maintained, however, measurements show (Joy et al. 1996) that these additional contributions are significantly lower in magnitude than the inherent statistical noise of the signal, and therefore will not need to be considered further.
The Statistics of Noise To be able to properly understand noise, it is first necessary to have some mathematical description of its behavior. For convenience, most textbooks (e.g., Goldstein et. al 2003) have followed the early classic work of Everhart [1970] and describe noise as being Gaussian in nature. Because the number of electrons collected from each pixel in an image is small—typically 20 or less—, however, the mathematical distribution would be expected to be Poisson in type, because Poisson statistics deal with discrete events occurring in fixed intervals of time or space. In addition, it has been suggested that the noise distribution of the SE signal might be even more complex than either Gaussian or Poisson, because the statistics of both the incident and the backscattered electrons contribute to the resulting statistics of the emitted secondaries (Morton & Mitchel 1948; Baiker 1960; Fillipov 1966; Seiler 1983). Because noise significantly affects the predicted performance of an SEM, it is important to test this assumption. In addition, there is increasing interest in simulating SEM images for various purposes, and the need to incorporate realistic noise in such cases raises the question as to its exact description. Poisson statistics apply when counting a number of objects or events—such as electrons—in a certain time period. These events must be random—that is, there must be no correlation between successive events. If the average number of electrons counted in a given time period is s, then the probability P(M) of observing M electrons in that time period is P(m) = exp
−S
SM . M!
(4.1)
If the mean number of electrons collected from a pixel representing the observed signal level S is M, then the standard deviation in that number—i.e., the noise—is M1/2 , and the SNR is then M/ M1/2 , and thus M1/2 .
132
D. C. Joy
When S becomes large, the Poisson distribution of Equation 4.1 tends to the form (Q − S)2 1 . exp − P(Q) = √ 2S 2π S
(4.2)
which is a function with the familiar Gaussian bell shape—a maximum value at Q = S and a standard deviation of S1/2 . The Gaussian distribution, however, strictly applies only when the outcome of the experiment can take fractional values and not to inherently quantized events such as the counting of electrons.
Measuring Secondary Electron Noise Early attempts to study SE noise—for example the work of Filippov [1966] and Pawley [1974]—had to rely on indirect measurements of the signal because digital recording and analysis techniques were not available. On a modern SEM, however, such measurements can now be performed very easily by using the digital imaging and storage capability that is provided and then analyzing the data with a suitable software package such as IMAGE Java (a public domain program that can be downloaded from http://rsb.info.nih.gov/ij/). The experiment consists of digitally recording secondary electron images from featureless surfaces such as clean, polished, silicon wafers. These images are also deliberately defocused to eliminate any contrast from residual surface detail. By varying the beam currents from a few picoamps (pA) to a few nanoamps (nA), and by choosing pixel dwell times varying from microseconds to milliseconds, the mean number of secondary electrons generated per pixel can be varied from less than 1 up to a maximum of 500 or more. The actual average number of secondaries NSE emitted from a sample under some beam currents can be found with the formula: NSE = 6.IB .τ.δ
(4.3)
where IB is the beam current in picoamps, τ is the pixel dwell time in microseconds, and δ is the secondary electron yield coefficient for the sample at this beam energy. Using the brightness and contrast controls of the SEM, the average signal level is adjusted to lie approximately in the center of the dynamic range of the system. The digital micrographs can then viewed in IMAGE Java, or any modern imageprocessing program, and processed to generate a histogram of the intensity distribution. The histogram itself can then be analyzed by putting the numerical data into a spreadsheet and attempting to fit both a Gaussian (see Equation 4.2) and a Poisson distribution (see Equation 4.1) to the data. In this way, the quality of fit and any deviations between Gaussian and Poisson behavior can be observed and quantified. Figure 4.3 shows the analysis of data recorded with a beam current of 1 pA and with a pixel time chosen so that the calculated number of secondary electrons emitted per pixel was about 0.8 (using the SE yield data for silicon obtained from the database at http://pciserver.bio.utk.edu/metrology). Using the Brightness and Con-
4 Noise and Its Effects on the Low-Voltage SEM
133
Fig. 4.3 Histogram of image data recorded with an average of 0.8 secondary electrons emitted per pixel showing the Poisson and Gaussian fits to the data
trast controls of the SEM, the signal was adjusted to span the dynamic range of image brightness from zero to 255. It is observed that the experimental histogram is well-fit by a Poisson distribution with a mean value s of 0.8. What is perhaps unexpected is that the quantization of the data—i.e., the brightness steps representing the collection of 0, 1 or 2 secondaries per pixel—is clearly evident, proving that the Everhart-Thornley SE detector in a SEM is capable of counting individual electrons (and also confirming that noise generated in the detector or processing electronics is not an important consideration). The Gaussian distribution, on the other hand, cannot in any way be fit to this data. The SNR measured from the histogram—i.e., the ratio of the mean signal count M to the standard deviation of the distribution—is 1.2, in this case, which is actually slightly better than would have been predicted (i.e., if s = 0.8, then the SNR s1/2 is 0.9). Figure 4.4 shows the corresponding data when the dwell time is increased by a factor of four, so the average SE count at each pixel is now 3.2. As before, the Poisson expression (see Equation 4.1) with s equal to 3.2 accurately fits the experimental data and the quantization (8 steps) of the image brightness is easily discernible. For this higher number of counts, the Poisson distribution is already beginning to look more symmetrical and, as shown, a Gaussian fit (see Equation 4.2) centered on a signal level of 98 with a standard deviation adjusted to match the distribution also represents a plausible fit for the data. Thus, it is clear that even when only a few electrons per pixel are being counted, a Gaussian profile can be fit to the data. The
134
D. C. Joy
Fig. 4.4 Histogram of image data recorded with an average of 3.2 secondary electrons emitted per pixel showing the Poisson and Gaussian fits to the data
numerical parameters of the Gaussian, however, are not a true physical representation of signal because it is still strongly quantized. It is also interesting to note that the measured SNR is now 2.4, which again is slightly higher than the expected value of 1.8. Finally (see Fig. 4.5), with the dwell time increased still more to give an average signal of 6.5 SE per pixel, it shows that both a Poisson distribution with s = 6.5, and a Gaussian of appropriate mean value and standard deviation fit very well, even though some quantization of the pixel brightness values is still visible.
What do These Results Mean? On the evidence of experiments such as those shown in the previous section, it is fair to conclude that as the electron count per pixel increases, the differences between the Poisson and Gaussian representations of the signal become too small to be significant and either representation can be used as convenient. Therefore, although the signal really is Poisson in form because it is a count of fundamental particles, it can, for practical purposes, be more conveniently treated as a Gaussian. The fact that this is so, and that the SNR also tracks the value expected for a Gaussian distribution, shows that the statistics of secondary electron emission are not adversely affected by the fact that both the incident and the backscatter electron signals—and their associated statistical fluctuations—contribute to the effect. In fact, the observation that the
4 Noise and Its Effects on the Low-Voltage SEM
135
Fig. 4.5 Histogram of image data recorded with an average of 6.4 secondary electrons emitted per pixel showing the Poisson and Gaussian fits to the data
SNR is actually slighter better than the value expected for the actual electron count per pixel might suggest that the secondary electrons that are emitted are not totally independent but may have some limited degree of correlation between them, but it also possible that bandwidth of the electronics leading to the digitizer is insufficient to respond to the large pixel-to-pixel excursions characteristic of noisy images. This would cause the digitized signal to have a lower variance than the actual signal.
How to Measure Signal-to-Noise Ratio (SNR) Once the nature of electron noise has been established, it is useful to have some way of measuring the SNR of an SEM image. Provided that the specimen has no detail of its own, the SNR can be determined directly from the histogram, provided that a little care is taken. If an image like that in Fig. 4.1 is recorded to give a histogram similar to that shown in Fig. 4.2, then it is evident that the position of the peak (i.e., the signal level) and the width of the distribution (i.e., the standard deviation representing the noise) will both be affected by the settings of the Brightness and Contrast controls of the SEM. The brightness control raises or lowers the DC background on which the signal is riding, while the contrast control adjusts the range over which the signal swings from maximum to minimum. To get a meaningful measurement
136
D. C. Joy
of the SNR, it is necessary to properly account for these variations. The procedure for this is straightforward and only requires that the microscope be equipped with a Faraday cup so the incident beam current striking the sample can be determined. With the SEM operating at some known beam current IB1 , the Brightness and Contrast controls are adjusted to give an image for which the histogram falls in the lower portion of the intensity scale. The peak position S1 of the histogram is then noted. The beam current is now increased to some new value IB2 , without touching either the Brightness or the Contrast controls, and the new position of the peak S2 is determined. If the signal is now so large that it causes the signal level to clip (i.e, exceed the maximum allowed value), then the Brightness and Contrast controls must be readjusted and the experiment repeated until clipping no longer occurs. Plotting the peak position against the beam current then gives a straight line (see Figure 4.6), the intercept or “zero offset” of which measures by how much the Brightness and Contrast controls have shifted the DC level. The peak values can then be corrected for this offset and the signal to noise ratio calculated. The peak values can then be corrected for this offset and the signal to noise ratio calculated. The corrected signal values are then S1 –S, S2 –S . . . and the corresponding SNRs are (S1 –S)/standard deviation 1 and so on. The beam can be blanked for one or two seconds during the scan by offsetting the objective aperture, thereby inserting a peak in the histogram representing the true signal zero value.
Fig. 4.6 Plot showing the variation of histogram peak position with beam current and the adjustment for the brightness and contrast controls
4 Noise and Its Effects on the Low-Voltage SEM
137
It is often of interest to be able to measure the SNR of an image that contains detail where it is either not possible or not desirable to take several measurements at different beam currents—for example, when studying a previously recorded digital image. This can be done using a method introduced by Frank and El Ali [1975], in which the variance and co-variance of either two sequentially recorded images of the same area (Erasmus et al. 1980), or of alternate pairs of lines in a single copy of the image (Oho et al. 2000; Joy 2002), are calculated. If the variances of the first and second images are var(1) and var(2), respectively, and the covariance between images 1 and 2 is cov(1,2), then cov(1, 2) Rn = √ var(1).var(2)
and the S N R =
Rn 1 − Rn
(4.4)
This algorithm has been found to give good results on a wide range of actual SEM images. The computing time required is only a second or so, and the obtained SNR values are reproducible and reliable. Note, however, that this method only works well on micrographs that contain real image detail. When applied to the empty, noise-only, images of the type shown in Fig. 4.1, the results are not a good estimate of the true SNR. For an interesting alternative approach to SNR measurements from a single image, see Thong et al. [2001]. With the high speed computation that is now available, it would be a welcome advance to see a real-time readout of the SNR provided on an SEM to help the operator properly judge and optimize imaging conditions.
The Effect of the Detector Detectors are the part of the signal collection chain in the SEM where the SNR is at its lowest. This is because detectors do not generally collect 100% of the signal leaving the specimen. Therefore, because the SNR depends on the electron count, the value will always be at its minimum at this stage of the signal detection process. The effectiveness of the detector is quantified by the parameter called the detector quantum efficiency (DQE), which is defined from the relationship
S N
2 = D Q E. exp
S N
2 (4.5) theor y
where (S/N)exp is the measured SNR of the image coming from the detector and (S/N)theory is the expected SNR for the beam-operating conditions and sample that are being employed. The DQE of a detector will be influenced both by noise generated within the detector itself, and by the geometric collection efficiency of the detector. In modern SEMs, the inherent noise of the detector is generally not large enough to be noticeable. Therefore, in the situation where we can assume that the detector is itself noise free but collects only some fraction ε of the total available signal, Nε electrons actually reach the detector only if N electrons are
138
D. C. Joy
generated per pixel at the sample only. The SNR at the specimen (i.e., (S/N)theory ) is then N1/2 as before, while the SNR at the detector (i.e., (S/N)exp ) will be (Nε)1/2 . Applying Equation 4.5 then shows that DQE = ε, so the quantum efficiency in this instance measures the geometrical fraction of the signal intercepted by the detector. If the detector response to the incoming electrons depends on their energy or their trajectory, or if the detector itself generates noise, then the DQE value will be always be lower than the simple geometric efficiency. This is typically the case for devices such as backscatter, x-ray, and photon (cathodo-luminescence) detectors. The DQE of any SEM detector can readily be measured using simple techniques based on those discussed above (Joy et al. 1996). The basic requirements are again a featureless sample of a clean material, such as silicon, whose secondary electron yield δ and backscatter electron yield η are known at one or more beam energies; the detector to be measured; a knowledge of the pixel dwell time τ for the scan speed in use; and the ability to measure the incident beam current of the SEM. Step 1: With the microscope operating at some convenient energy—typically 10–20 keV—and at a known beam current IB1 —measured from a Faraday cup or a similar arrangement—an image of a region of the silicon wafer is recorded at a known pixel dwell time τ (in case of doubt about this parameter, its value can be sufficiently well-estimated by dividing the time required to store one frame of the image by the number of pixels in the image). The Brightness and Contrast controls on the SEM are adjusted to place this signal in the lower half of the signal range between black level and peak white. Step 2: The incident beam current is now increased to a value IB2 —by using a larger aperture or raising the filament temperature—and a second image is recorded at the same scan rate and with the same brightness and contrast settings as before. Step 3: As discussed in the previous section, and illustrated in Fig. 4.6, the histograms of these two images are then obtained, and the positions S1 and S2 of the two signal peaks are determined. S1 and S2 are then plotted against the beam current, as in Fig. 4.6, to determine the DC offset S. The beam can also be temporarily blanked by the aperture to determine true DC zero. Step 4: Select one of the images ( e.g., number 2) and compute the peak signal = S2 –S, the standard deviation SD2 , and hence the experimental SNR (S/N)exp = (S2 –S)/SD2 from the histogram. Step 5: Now, calculate the theoretical SNR (S/N)theory using Equation 4.3. Here, the beam current is IB2 and the pixel dwell time is τ. If the specimen is silicon, the SE yield δ at 10 keV is 0.21 and the SE yield at 20 keV is 0.10 at normal incidence. Noting the beam current in units of picoamps and the dwell time in microseconds, and inserting the appropriate SE yield for the beam energy in use as NSE , the number of SE produced per pixel is calculated from Equation 4.3. The theoretical SNR (S/N)theory is then NSE 1/2 . Step 6: Calculate the DQE of the detector by inserting (S/N)exp and (S/N)theory into Equation 4.5. This procedure can be simply applied to any SEM detector to obtain information about the efficiencies of its various detector systems. Table 4.1 shows typical data for
4 Noise and Its Effects on the Low-Voltage SEM
139
Table 4.1 Experimental DQE value for typical SEM SE detectors SEM Type DQE Maker S4500 S4500 S4500 XL30 S360
TTL ET MCP/SE ET ET
0.76 0.18 0.058 0.17 0.017
Hitachi Hitachi Hitachi FEI Cambridge
various kinds of secondary electron detectors on microscopes produced by several different manufacturers. This list is by no means comprehensive, and the actual values will fluctuate even between different instruments of the same type, but it provides interesting data about the relative performance levels that can be achieved. It should be noted, for example, that in the Hitachi S4500 the standard EverhartThornley (E-T) detector beneath the lens and close to the specimen has a DQE of 0.18 but that the through-the-lens (TTL), or upper, SE detector has a DQE of 0.76. The magnetic field of the lens is obviously much more effective at sweeping electrons into the TTL detector than is the electrostatic field that guides secondaries towards the E-T detector. However, because both of these detectors are operating at the same time together, they succeed in capturing between them a total of 0.18 + 0.76 (about 94%) of the total available SE signal. By comparison, a microchannel plate (MCP) detector biased for SE use has a DQE of only 0.058, even though it is installed close to, and right above, the specimen. The DQE of an E-T secondary detector, averaged over data from a dozen different SEMs from five manufacturers, has been found to be between 0.1 and 0.2 (although the value shown of 0.017 measured from an old Cambridge Stereoscan shows that much lower values can be encountered in some cases, especially if the instrument has been so heavily used that the scintillator has been degraded [Pawley 1974]). This, of course, reflects the fact that the E-T detector is of limited efficiency by design because it is placed asymmetrically with respect to the specimen and it is this arrangement that produces the characteristic detector efficiency contrast effects in the SE image (Goldstein et. al 2003). If the analysis is made from a low-magnification micrograph, then by performing the calculation on just a small region of the image (e.g., 16×16 pixels), the spatial variation of the DQE can be mapped to show how the system responds to the position from which SE are emitted. At low magnifications, the DQE of an E-T detector may vary by 50% across the field of view. Detectors placed above the specimen typically show less variation, but degradation towards the edges of the field of view will still occur.
The DQE of Digital Detectors At present, all SEM detectors operate in an analog mode in which the rate of the incoming electrons striking the detector is treated as representing a current. The output voltage from the detector is then proportional to, or an analog of, the electron
140
D. C. Joy
current. It is possible and often more desirable, however, to detect electrons by a digital, or counting, procedure in which each detected electron contributes one additional count to the received signal recorded for the pixel of interest (Yamada, et al, 1991). Because even a low-imaging current (e.g., 10–100 picoamps) represents a count rate of 106−107 electrons/sec—a figure that is at the upper limit of most readily available counting systems—digital detection has not yet been featured in any SEMs. With advances in electronics, however, it seems certain that digital detection will play a role in future instruments because it offers some advantages. The geometric efficiency contribution to the DQE of a digital system is the same as for an analog detector, and in most cases this implies that both analog and digital systems will have the same value. However, the digital system is processing a train of pulses of varying magnitudes representing both electrons of different energies and any contribution of noise from the detector itself. It can be shown (Herrman 1983) that, in this case, the DQE of the detector, excluding geometrical efficiency, has the value DQE =
1 var h 1+ 2 h¯
(4.6)
where var h is the variance of the pulse height distribution, and -h its mean value. An important option in a digital system is the ability to set a threshold level below which all pulses are rejected, because this can reduce or eliminate noise and other spurious contributions to the signal. If the low-level discrimination is set so that only some fraction η of the incoming pulse train is accepted, then Equation 4.6 reduces to (Herrman 1983) DQE = ε
(4.7)
The Importance of DQE Measurements made on similar detectors in different SEM columns, but made by the same manufacturer (e.g., successive updates of the same basic microscope), often show wide variations in DQE, suggesting that little or no attention is paid to testing or optimizing detectors during design and development. This situation will only be improved if users begin to measure the DQE of their detector systems and insist on each detector meeting an acceptable value as part of the purchase specification of the instrument. Compared to the cost of improving the electron source or the lenses, the cost of improving the detector DQE (and hence the SNR of the image) is modest, while the improvements in imaging and perceived performance are worthwhile and long-lasting. In addition, the loss in SNR caused by the degraded DQE has to be recovered by increases in beam current (which worsens the resolution) or recording time (which may cause increased contamination and sample damage).
4 Noise and Its Effects on the Low-Voltage SEM
141
The DQE of backscattered electron detectors is measured in the same way as that described previously, except that in Step 5, Equation 4.3 will be modified to have the form NBSE = 6.IB .τ.η
(4.8)
where η is now the backscattering coefficient for the target material. For example, if a silicon wafer is again used as the specimen, then η is 0.17 at both 10 and 20 keV. For illustration, Table 4.2 compares the DQE of four different BSE detectors installed at various times in the same microscope (a Hitachi S3500 VPSEM): a MCP detector biased for BSE observation; an E-T detector negatively biased to function as a BSE detector; a Centaurus detector (GW Electrons: Atlanta, GA), and a Robinson scintillator BSE detector (Hitachi High Technology: Pleasanton, CA). It is evident that the DQE values for BSE detectors are significantly lower than those for SE detectors, despite the fact the BSE detectors are usually designed to subtend as large a solid angle as possible at the sample and therefore intercept a high fraction of the emitted signal. This discrepancy occurs because the DQE of a BSE detector—unlike that of a SE detector—is not equal to its geometric collection efficiency, but is further modified by its response to the energy spectrum of the backscattered electrons. For example, many scintillator or solid state BSE detectors have little or no sensitivity for electrons less than perhaps 5 keV in energy. As a result, the DQE at around 10 keV will be much lower than the corresponding value at 30 keV, where the average energy of the BSE—about 50% of the incident electron energy—will also be increased. The magnitude of this effect can be gauged by comparing the data for the Robinson detector at 10 and 30 keV. At 10 keV, the DQE is less than 0.05 and must be rated as poor, but at 30 keV, the DQE is 0.3, which is actually higher than that of most SE detectors. DQE data has also been obtained on other detector systems. For example, the gaseous secondary electron detectors (GSED) used in variable pressure SEMs have been found to typically display a DQE in the range 0.1 to 0.3, depending on parameters such as the gas pressure and the applied electric field—and so, are comparable to conventional scintillator SE detectors. By comparison, a conventional energy dispersive x-ray detector (Joy et al. 1996) had a DQE of 0.0007. That value can be compared to a computed geometrical efficiency of 0.0017 for the same detector. The difference between these two DQE values illustrates the effects of processing a digital pulse train of varying pulse height, as shown in Equation 4.6.
Table 4.2 Experimental DQE values for typical. SEM BSE detectors Type DQE MCP/BSE ET in BS mode Robinson (10keV) Robinson (30keV) Centaurus (10keV)
0.058 0.001 0.043 0.3 0.09
142
D. C. Joy
The Effect of SNR on Microscope Performance We have seen how noise arises in the SEM image, its characteristics, and how it is affected by the properties of the detector system fitted to the instrument. Finally, it is useful to examine what effect noise—or more properly, the SNR—has on the performance of the microscope. The standard answer to this is provided by a discussion of the Rose criterion (Rose 1948; Goldstein et. al 2003). This says, that if some feature in the image has a contrast level (or visibility) C—defined as S/Smax , where Smax is the maximum signal intensity across a feature and S is the corresponding change in signal level across that feature compared to the background—then the inequality C>
5 (S/N )
(4.9)
where S/N is the SNR must be satisfied for that feature to be said to be detectable above the noise. The Rose criterion was deduced from experimental studies by observers looking at noisy images and therefore is an empirical result rather than a fundamental prediction. Nevertheless, it has been found useful in practice and has been widely applied as a guide for predicting SEM imaging performance. More recent computer simulation studies (Bright et al. 1998) have confirmed the essential validity of the result under more general and reproducible conditions than those of the original study. The Rose criterion shows that the ability of the image to reveal low-level detail is determined by the achieved SNR. Because typical contrast levels associated with topographic contrast in the SEM lie in the 10–30% (C=0.1 to 0.3) range, applying the Rose criterion suggests that an SNR of the order of 20 or more is necessary to successfully record such features in the image. For a given scan speed, this determines the incident beam current that is required. Because the beam current IB itself is a sensitive function of the probe size and other electron-optical properties of the microscope—varying as I B = 1.88β
d 8/3 2/3
(4.10)
Cs
where β is the brightness (amp/cm2 /str) of the electron source at the operating energy, d is the effective probe size of the beam, and Cs is the spherical aberration coefficient of the objective lens of the SEM—improvements to the SNR on a given microscope can only be achieved by increasing the recording time or using a larger spot size and therefore degrading the resolution. The SNR therefore directly determines the envelope of operating parameters for the microscope. The Rose criterion was developed by the analysis of data from experiments involving human observers, and its predictions are thus dependent on the way in which our brains process images and noise. In particular, it is commonplace to note that once we know what the detail in an image should look like, our brains are very
4 Noise and Its Effects on the Low-Voltage SEM
143
skilled at discovering it to be present (“I’ll see it when I believe it”). Consequently, knowing what to look for in a field of view probably relaxes the beam current (or SNR) conditions necessary for a human to see it. This ambiguity can be removed by choosing to treat the SEM as a channel for the transfer of information and applying communication theory to it (Shannon 1948). A detailed analysis of SEM imaging—incorporating a model of SE emission (Simon 1970; Joy and Bunn (1989) —proposes that the information capacity B (bits/pixel) of this SEM signal channel can be written in the form
B = loge υ 1/2 −
exp(−υ) er f (υ) + υ 1/2 υ2
(4.11)
where υ = Smax /(SNR)2 with SNR erf is the error function, and Smax is a normalizing constant. Examination of this expression shows that to obtain a 3-bit/pixel image (i.e., one with eight statistically distinguishable gray levels), requires an SNR of about 12:1. This is a value considerably more optimistic than would be predicted by the Rose model, where the ability to see 12% contrast (i.e., a change of one gray level in eight) would require an SNR of 60:1 from Equation 4.9. This discrepancy in the estimates of the SNR required to guarantee a given level of contrast visibility leads to significantly different predictions of the SEM performance that could be achieved. To increase the SNR from 12:1 to 60:1, while maintaining the same scan speed, requires raising the beam current by a factor of 25x. Applying Equation 4.10 for a given gun and lens system shows that achieving this increase in beam current requires increasing the probe size by a factor of almost 3x, which will most probably lead to a degradation in image resolution. Therefore, more work is needed to decide how reliable either of these predictions are. SNR measurements—using the methods discussed earlier—and visual assessments of typical SEM images suggest that the predictions of the Simon analysis are generally more realistic than those of the Rose model. For example, Fig. 4.5 in this chapter shows data displaying about 2.5 bits/pixel (six gray levels of quantization) for a SNR of 2.6, which is consistent with the prediction of Equation 4.11. In any case, it is certain that the SNR plays a major role in defining the quality of imaging that can be achieved from a given electron-optical system. Consequently, the ability to measure this parameter of the image, and to enhance it by designing detectors of improved DQE, is an essential step towards improving the scanning microscope. Because the reduction in SNR that results from a failure to optimize the DQE of a given collector significantly degrades the overall performance of the SEM, detector design should therefore receive the same level of care and financial support that is currently given to other components in the microscope column. Acknowledgment The author is grateful to Drs. Gian Lorusso, Ira Rosenberg, Edgar Volkl and Bonnie Smithson for helpful discussions and insights on noise and detectors.
144
D. C. Joy
References Baiker JA (1960) IEEE Trans. Nucl. Sci. 7:74 Bright et al. (1998) J. of Microscopy, 189:25 Erasmus SJ, Holburn DM, Smith KCA (1980) Instr. Phys. Conf. Ser. 52, p73–76 Everhart TE (1970) Cambridge Stereoscan Colloquium. Engis Corp; Chicago, 2:1–6 Filippov YA (1966) Soviet Physics-Solid State, 8:701–705 Frank J, El Ali L (1975) Nature 256, 376–379 Goldstein J, Newbury DE, Joy DC, Lyman C, Echlin P, Lifshin E, Sawyer L, and Michael J (2003), Scanning Electron Microscopy and X-ray Microanalysis, 3r d Edition, Kluewer Academic/Plenum Publishers 689 Herrman K-H (1983) in “Quantitative Electron Microscopy”, Proc. 25th Scottish Universities Summer School in Physics, (Scottish Universities Summer School: Aberdeen, UK), ed, J N Chapman and A J Craven, 119 Joy DC (2002) J. of Microscopy 208:4–34 Joy DC and Bunn RD (1989) Proceeding 47th Annual Meeting EMSA, ed, G W Bailey. San Francisco Press: San Francisco, 47:84–85 Joy DC, Joy CS, and Bunn RD (1996) Scanning 18:181–186 Morton GA and Mitchel JA (1948) RCA Review 9:632, Oho E, Asai N, Itoh S (2000) J. Electron Microscopy 49:761–763 Pawley J (1974) Scanning Electron Microscopy/1974, Proc. 7th Annual SEM Symposium. ed. O Johari, IITRI: Chicago, 27–34 Rose A (1948) Advan. Electron. Electron Physics 1:131 Seiler H (1983) Secondary Electrons in the SEM, J. App. Phys., 54:R1 Shannon C (1948) Bell Telephone System Tech. Publ. Monograph B-1598 Simon R (1970) J. Appl. Physics, 41:4632 Thong THL, Sim KS, Phang JCH (2001) Scanning 23:328–336 Yamada S, Ito T, Gouhara K, Uchikawa Y (1991) Scanning, 13:165–171
Chapter 5
High-Resolution, Low Voltage, Field-Emission Scanning Electron Microscopy (HRLVFESEM) Applications for Cell Biology and Specimen Preparation Protocols Heide Schatten
Abstract High resolution, low-voltage, field-emission scanning electron microscopy (HRLVFESEM), coupled with newly-developed specimen preparation protocols has permitted novel and powerful applications for cell biology, as it allows detailed insights into small biological features due to the minimal coating requirements. It allows resolution at levels that previously could only be achieved with transmission electron microscopy (TEM) and it has made it possible to generate three-dimensional images of structures as well as interactions with macromolecular complexes without confusion of structural overlap which allows clear interpretation, particularly when stereo-imaging is applied. In this chapter several examples are presented using three specific preparation techniques that can be applied to a variety of different specimens with specimen-specific modifications. The three separate sections in this chapter are 1) Visualization of sub-membranous cytoskeletal features using cytoskeleton stabilization and membrane extraction protocols. Two examples of different specimens in this section are (1.1) subpellicular cytoskeletal structures of the apicomplexan parasite Toxoplasma gondii, and (1.2) submembraneous actin cytoskeleton in osteocytes; 2) Visualization of whole mounts and isolated cell structures. The examples presented in this section include (2.1) Visualization of isolated nuclear envelope; (2.2) visualization of isolated mitotic spindles; (2.3) visualization of centrosomes; and 3) Visualization of resin-extracted de-embedded thick sections. De-embedding of thick-sectioned biological material is a unique approach to examine the interior of cells and tissue. Serial sections can be obtained and the entire cell or tissue can be evaluated by 3-D reconstructions of thick-sectioned material. In addition, the area of interest can be tilted and viewed from various angles for accurate interpretation. The power of this unique method is demonstrated in three applications focused on: (3.1) Nuclear pore complexes (Ris and Malecki, 1993); (3.2) Muscle fibers (Ris and Malecki, 1993); (3.3) Toxoplasma parasite internal structures. All methods presented in this chapter provide unique approaches to visualize and analyze delicate biological structure.
H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
145
146
Heide Schatten
Introduction High-resolution low-voltage scanning electron microscopy with field-emission guns has allowed new experimental approaches and excitingly novel applications for cell biology. It provides topographical contrast of small features that are only minimally coated while charging is substantially decreased as a result of lowered beam energy. This is important for viewing samples with minimal depth. Improvement in instrumentation, coupled with new preparation techniques, has provided insights into biological structure that previously could only be achieved with transmission electron microscopy (TEM). The methods described in this chapter have proven superior to TEM in many aspects, because of the ability to generate three-dimensional images of structures as well as of interactions with macromolecular complexes without confusion of structural overlap. Such clarity in interpretation, available from LVFESEM, particularly when stereo-imaging is applied, can not be achieved with TEM or any other imaging method available so far. In this chapter, three specific preparation techniques will be presented that can be applied to a variety of different specimens with specimen-specific modifications. The preparation techniques have been developed with various collaborators and have been used with stunning results by the late Hans Ris, a pioneer in combining novel preparation techniques with advances in cell biology. Separate sections in this chapter are devoted to 1) Visualization of sub-membranous cytoskeletal features using cytoskeleton-stabilization and membrane-extraction protocols; 2) Visualization of whole mounts and isolated cell structures; and 3) Visualization of resin-extracted, de-embedded thick sections.
General Methods and Materials The following general protocols can be applied to various specimens with reliable and consistent results. Modifications are required when specific structures of interest are under investigation such as the use of tannic acid or modification in the postfixation concentration of osmium when cytoskeletal components are analyzed. The following protocols are modified from Ris [1985]. Fixation 1 employs different fixation mixtures without tannic acid. a. 2 % glutaraldehyde in 0.1M PIPES (or HEPES) buffer at pH 7.0 for 30 min followed by either 0.05 %, 0.1 % or 1 % OsO4 at room temperature for 10 min, or without OsO4 . b. 2 % paraformaldehyde, 1.25 % glutaraldehyde in 0.1M phosphate buffer at pH 7.0 for 10 min. c. 2 % glutaraldehyde in 0.1M HEPES buffer at pH 7.0 containing 0.05 % saponin for 30 min. Fixation 2 employs the different fixation mixtures with tannic acid. a. 2 % glutaraldehyde in 0.1M HEPES buffer at pH 7.0 containing either 0.2 % or 2 % tannic acid for 30 min followed by either 0.05 % or 0.1 % OsO4 at room temperature for 10 min, or without OsO4 .
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
147
b. 2 % glutaraldehyde in 0.1M HEPES buffer at pH 7.0 containing 0.05 % saponin and 0.2 % tannic acid for 30 min followed by 0.1 % OsO4 for 10 min. c. 1 % glutaraldehyde, 50mM KCl, 50 mM sodium phosphate buffer at pH 7.0, 5mM MgCl2 , 4 % tannic acid. Dehydration follows the typical dehydration steps from 30 % EtOH or acetone, to 100 % ethanol or acetone. It is extremely important to remove all traces of water from the final EtOH by adding water-absorbing components such as Molecular sieves to the final ethanol or acetone steps. Traces of water will create artifacts (for details please see Ris, 1985). Critical point drying: Careful critical point drying (CPD) is absolutely essential to avoid artifacts (Ris, 1985). Critical point drying was introduced to avoid producing surface tension artifacts during specimen drying. This is achieved by first placing the specimen in a pressure chamber immersed in a suitable transition liquid, usually liquid CO2 . At a critical point, when a specific temperature and pressure is reached, the transition liquid is converted into a gas. Critical point drying is now employed routinely by using commercially-available critical point drying devices. CPD is used to remove all of the ethanol or acetone by thoroughly flushing the chamber with liquid CO2 under pressure. The transition from liquid CO2 to gaseous CO2 is achieved by raising the temperature beyond the critical temperature and then slowly allowing the now-gaseous CO2 to escape. This prevents any liquid-gas interphase passing through the specimen. To avoid distortions of the biological structure by incomplete removal of EtOH or acetone, it is important to attach a water-absorbing molecular-seive filter in the line between the bomb and the CO2 tank. Such filters are now routinely supplied by manufacturers such as the Tousimis Research Corporation (Rockville, MD). Since the specimen chambers tend to trap residual EtOH or acetone such traces need to be removed by agitating the specimen holder while replacing the specimen with CO2 . Please read the manufacturer’s instructions carefully for accurate critical point drying. The critical-point-dried samples should be stored immediately in a desiccator that contains phosphorous pentoxide. Coating: In most cases light coating is required. For our samples, we routinely use high-vacuum, ion-beam sputter coating with 1nm Pt but modifications may be needed depending on the material under investigation (please see other chapters in this book). Analysis: Critical-point dried and coated material is imaged with a Hitachi-S-900 FESEM or similar instrumentation operating at 1.5 kV accelerating voltage. 1) Visualization of sub-membranous cytoskeletal features using cytoskeleton stabilization and membrane extraction protocols Two examples of different specimen are presented here, the (1.1) subpellicular cytoskeletal structures of the apicomplexan parasite Toxoplasma gondii, and the (1.2) submembraneous actin cytoskeleton in osteocytes. For both, the cytoskeleton stabilization methods introduced by Schliwa (1981) have been used in combination with minimal membrane extraction. Both examples have revealed for the first time, structural features that have not been possible to visualize with other methods.
148
Heide Schatten
Toxoplasma: The protozoan parasite Toxoplasma gondii (T. gondii) represents a large group of apicomplexan parasites (that include Plasmodium falciparum, the pathogen for malaria, and Cryptosporidium parvum that causes outbreaks of cryptosporidiosis) with a highly unusual cytoskeletal system that is crucial for locomotion, for host cell invasion, and for cell division. The unconventional apicomplexan cytoskeletal system differs in many aspects from that in well-studied mammalian cells (Shaw et al., 2000; Striepen et al., 2000). In T. gondii, actin is thought to be organized parallel to the plasma membrane in the space between the inner membrane complex (IMC) and plasma membrane (Pinder et al., 1998). The pellicle’s microtubule cytoskeleton is a complex of fibers that originate from an apical polar ring (reviewed by Duber et al., 1998) and is described in the classic electron microscopy studies by Aikawa et al. , (1977). Twenty-two single microtubules project backward in a counterclockwise helical course from the polar ring, extending approximately 2/3 of the length of the cell body. These cytoskeletal fibers interact with other cytoskeletal components and with the parasite’s membrane system. Because actin dynamics are rapid, it has been difficult to capture actin filaments in a native state. As actin could not be visualized by conventional electron microscopy, it was thought that it exists primarily in the globular form. By using our membrane-extraction and cytoskeleton-stabilization protocols, it was possible to visualize these fixation-sensitive actin-like fibers for the first time (Schatten et al., 2003).
Materials and Methods Parasites are prepared for LVFESEM analysis according to our previously described methods (Schatten and Ris, 2002, 2004; Schatten et al. , 2003). The virulent RH strain of T. gondii is used (Sabin, 1941) and tachyzoites are grown in primary human foreskin fibroblasts (HFF cells). Cells are inoculated at moderate density (ca. 2000 cells/cm2 ) into tissue culture flasks or plates in modified Eagle’s medium (MEM; ca 0.3-0.4ml/cm2 ) containing 10 % heat-inactivated fetal bovine serum (HFF medium) and incubated at 37◦ C in a humidified CO2 incubator. Parasite cells are collected at various time points after invasion into the fibroblast cells. Actin-depolymerization experiments, to prove the existence of actin-like fibers, employ the microfilament inhibitor cytochalasin D (Sigma, St. Louis), used in our experiments at 20µM in medium for 1h (Dobrowolski and Sibley, Dobrowolski). The remaining cytoskeleton is then analyzed for structural differences in cytoskeletal organization. Material used for LVFESEM is prepared by extracting cells with 0.15 % Triton X-100 in cytoskeleton-stabilization buffer (PHEM; Schliwa, 1981) consisting of 60mM Pipes, 25 mM Hepes, 10mM ethylene glycol bis (beta-aminoethyl ether)N,N,N’, N-tetra acetic acid, 2 mM MgCl2 , pH 6.9 (Schatten and Ris 2002, 2004; Schatten et al., 2003). The stabilization protocol will not induce artifacts but it will stabilize the components in their natural state without the risk of depolymerizing
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
149
the cytoskeleton during fixation and subsequent processing. Samples are fixed for 30 min in 2 % glutaraldehyde in 0.1 M Hepes buffer, pH 7.4 with 0.05 % saponin and 0.2 % tannic acid. After rinses with 0.1 M Hepes cells are postfixed in 0.1 % OsO4 for 10 minutes (Schatten and Ris, 2002, 2004; Schatten et al. , 2003). The osmium concentration is kept low to preserve the fixation-sensitive actin fibers. Incubation in 1 % uranyl acetate follows for 10 min before dehydration in ascending ethanol series, critical point drying and ion-beam sputter coating with 1nm Pt. Analysis was performed at the Integrated Microscopy Resource in Madison, WI, with the Hitachi S-900 FESEM operating at 1.5 kV. Figures 5.1 and 5.2 are examples of results obtained with the methods described above.
Osteocytes: Materials and Methods Preparation of primary cultures of chick osteocytes is described in detail in a recent paper by Kamioka et al., (2006). Briefly, cells are attached to a glass carrier coated with poly-D-lysine and fibronectin and the cell membrane is removed with Triton X-100 and cytoskeleton stabilization buffer (Schliwa, 1981) as described above for
Fig. 5.1 Parasite attached to the surface of a host cell, in the process of penetration. The cell membrane was removed with 0.15 % Triton X-100 in cytoskeleton-preserving buffer (Schliwa et al., 1981). We can see a network of fibers 7nm thick (arrows) that are equal in size to actin fibers in the HFF host cell (arrowheads). It has been known that a Toxoplasma-specific actin is present below the cell membrane, in the anterior region of the parasite. This is the first time that such an actin-sized network has been imaged (Bar = 200nm). Reprinted with permission from Schatten et al., (2003).
150
Heide Schatten
Fig. 5.2 Parasite attached to a host cell as in Figure 1. After removing the cell membrane with Triton X-100, the cells were treated with cytochalasin D. Fixation and postfixation are the same as in Figure 5.1. The network of 7 nm fibers has disappeared, supporting the conclusion that they consist of actin. Below the actin layer, we now see the subpellical cytoskeleton. There are several longitudinally arranged 20-nm fibers (arrows) in the position where 22 subpellicular microtubules are known to exist, originating from the apical polar ring at the tip of the cell. Associated with the microtubules are fibers about 10 nm thick (arrowheads), which can be seen both above and below the microtubules in stereo images. The diameter of the fibers suggests that they are intermediatelike filaments (Bar = 200nm). Reprinted with permission from Schatten et al., (2003).
T. gondii, followed by dehydration and subsequent processing as described in subsection 1.1.
Visualization of submembranous actin network Figure 5.3 shows the actin cytoskeleton of an osteocyte isolated from chick embryos in culture after Triton X-100 treatment. Seen here are bundles of actin filaments radiating from central focal points to the cell surface and become cores of the cell projections that connect to adjacent osteocytes. 2) Visualization of whole mounts and isolated cell structures Various methods for SEM imaging have made use of isolation of internal cell structures as well as other preparation methods that are summarized in Ris and Malecki (1993). Delicate isolated cell structures and macromolecular complexes have routinely been imaged with TEM-negative staining and, more recently, also with immunofluorescence microscopy. However, while extremely powerful in many aspects, negative staining or immunofluorescence microscopy preparations most often do not reveal sufficient structural detail. For negative staining, a variety of different protocols are frequently necessary to determine the true structure; for immunofluorescence microscopy, only analysis of predetermined labeling of cellular
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
151
Fig. 5.3 Chick osteocyte attached to glass coated with polylysine-D and fibronectin. The cell membrane was removed with Triton X-100 in cytoskeleton-preserving buffer (Schliwa et al., 1981). We can see a network of actin filaments radiating from central focal points to the cell surface and becoming cores of the cell projections that connect to adjacent osteocytes. (Bar = 200nm). Reprinted with permission from Ris [1998].
components is possible. HRLVFESEM allows visualization and imaging of isolated cellular components and their structural details, as only minimal coating is required. Visualization of macromolecular complexes and their interaction with other cellular components allows more specific analysis of intracellular interactions. In this section we show different examples that demonstrate the unique usefulness of LVFESEM for revealing structural details owed to low coating. The examples presented
152
Heide Schatten
here are (2.1) Visualization of isolated nuclear envelope; (2.2) visualization of isolated mitotic spindles; (2.3) visualization of centrosomes.
Whole mounts of isolated nuclear envelopes from frog oocytes. The nuclear pore complex (NPC) is a complex macromolecular structure containing nanometer-range components (Ris, 1991; Jarnik and Aebi, 1991; Ris and Malecki, 1993). The preparation and isolation protocol of NPCs is described in detail by Ris and Malecki (1993). Briefly, nuclei from mature oocytes are obtained by release into low-salt buffer (LSB: 10mMHepes buffer, pH 7.4, 1 mM KCl, 0.5 mM MgCl2 ), cleaned of yolk and transferred onto a 7×17-mm piece of mica (Malecki and Ris, 1991) that had been coated with 3-aminopropyltriethoxysilane (Aldrich Chemical Co, Milwaukee, WI). After attachment, the nucleus is slit open with fine glass needles to remove the nucleoplasm. Then the mica with the nuclei is immersed into fixative (1-2 % glutaraldehyde and 0.2 % tannic acid in LSB). After washes in LSB, the preparation is fixed in 0.1-1 % OsO4 for 10 minutes, followed by changes in distilled water. A whole-mount image of an isolated nucleus is shown in Figure 5.4, clearly displaying the surface of the nuclear envelope that is covered with many cytoplasmic components of NPCs,. The intranuclear surface in the upper area of Figure 5.4 reveals fishtrap-like structures at the intranuclear surface.
Isolated mitotic spindles The isolation of mitotic spindles has been a major advance in cellular biology and has allowed detailed studies to decode the machinery for cell division and to study components that play a role in chromosome and spindle pole organization on cellular and molecular levels. The generation of poly- and monoclonal antibodies has made it possible to characterize the specific structures and macromolecular complexes that are involved in cellular dynamics during mitosis and include cytoskeletal components with the majority being microtubules and their associated proteins, microtubule-motor proteins, enzymes and numerous others (Harris et al. , 1980; Harris, 1986; Bestor and Schatten, Bestor; Hertzler and Clark 1993; Salmon, 1982; Schatten et al., 1986; Sluder et al., 1989; Staiber, 1994; Gard, 1991; Lee et al., 2000; Meng and Wolf, 1997). However, many of the biochemical and molecular findings have not yet been assigned to specific structural entities. HRLVFESEM has made it possible to visualize the interactions of molecular components and analyze trafficking along microtubules. The following provides one example of numerous others on the ultrastructure of the isolated mitotic apparatus. Seen here are microtubules emanating from a central core. The microtubules are decorated with components that are still not clearly identified. Seen here are also vesicles that traffic along microtubules. ImmunoFESEM will allow characterization of structural components
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
153
Fig. 5.4 Stereoimage of the nuclear envelope (NE) isolated from an oocyte of the newt Notophthalamus viridescens. The nucleus was dissected into low-salt buffer (LSB), transferred onto a cover-glass chip, extracted with 0.1 % Triton X-100 in LSB, opened with fine glass needles, fixed in 2 % glutaraldehyde and 0.2 % tannic acid in LSB, postfixed in 0.1 aqueous osmium tetroxide, dehydrated with ethanol, and critical point dried. The cytoplasmic surface of the NE is seen on the lower right, closely packed with nuclear pore complexes (NPCs). In the upper half, we look down on a structure consisting of an annulus with eight cylindrical particles. A central particle is visible in most NPCs. In the lower half, we look down on the intranuclear surface of the NE. The fishtraps, representing the intranuclear segments of the NPCs, are clearly visible. Between the fishtraps, are scattered remnants of the lamin fibers, disupted during swelling of the NE in LSB. Bar, 200 nm. Reprinted with permission from Ris and Malecki, (1993).
that play a role in mitotic dynamics. For isolation and fixation procedures please see Schatten et al. (2000). Figure 5.5 shows an FESEM stereo-image of an isolated mitotic spindle from a sea-urchin egg.
Isolated centrosome material: sample preparation and fixation Centrosomes are the primary microtubule-organizing centers that undergo molecular and structural reorganizations throughout the cell cycle. Pioneering studies on centrosomes had been performed over 100 years ago by Theodor Boveri [1901] in sea-urchin eggs providing a wealth of information on centrosome structure based on iron hematoxylin as the primary light microscopy staining technique. Boveri showed that sperm centrosomes are introduced into the egg after fertilization and reorganize to form the mitotic apparatus during cell division. A renaissance of research on centrosomes has emerged over the past decade with advances in immunofluorescence and molecular methods (Schatten et al. , 1987; 1992). It has been recognized that centrosomes undergo cell cycle-dependent structural changes that follow a pattern of compaction and decompaction analogous to chromosomes that undergo cycles of condensation and decondensation throughout the cell cycle. Centrosome abnormali-
154
Heide Schatten
Fig. 5.5 LVFESEM image of the aster from mitotic spindle, showing microtubules emanating from a central core. The microtubules are decorated with components that are still not clearly identified. Also seen here are vesicles that traffic along the microtubules. Bars as indicated at the lower right. Reprinted with permission from Thompson-Coffe et al. (1996).
ties can be causes for infertility and for diseases of various nature, including cancer. For recent reviews on centrosomes please see Manandhar et al., (2006) and Sun and Schatten (2000a-c). While immunofluorescence microscopy has generated the majority of new data on centrosomes, TEM ultrastructural studies have not revealed much information, as centrosomes display an osmiophilic amorphous appearance. By employing HRLVFESEM on isolated centrosomes, it has been possible to assign a structure to centrosome material (Thompson-Coffe et al., 1996; Schatten et al., 2000).
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
155
Material and Methods: The centriole-centrosome complex is isolated by lysis in a cold extraction buffer using protocols that we have employed previously (Thompson-Coffe et al. , 1996; described in Schatten et al. , 2000). Briefly, centrosomes are isolated from extracted cells by lysis in a cold extraction buffer containing 1 M glycerol medium, 10 mM piperazine-N,N -bis(2-ethane-sulfonic acid, PIPES, Sigma), 1 mM MgSO4 , 1 mM ethylene glycol-bis-(B-aminoethyl ether)N,N,N ,N -tetraacetic acid (EGTA, Sigma), pH 6.8, with 0.1 % Triton X-100, 1 mM DL-dithiothreitol, 100 μg/ml soybean trypsin inhibitor, and 1 mM phenyl methyl sulfonyl fluoride (PMSF). Cells are lysed by vortex mixing, and the isolated centrosome complex is collected by centrifugation at 3000 rpm for 30 min in a swinging-bucket rotor, resuspended in 2 M glycerol medium (as for 1 M, with increased glycerol), underlaid with 3 M glycerol medium (as for 1 M, with increased glycerol), and recentrifuged at 3,000 rpm an additional 2-3 times. The final pellet is resuspended in 50 % glycerol, 10 mM PIPES, 1 mM MgSO4 , 1 mM EGTA, 1 mM PMSF, pH 6.8. The isolated centrosomes are then fixed, dehydrated, CPD, coated and analyzed with HRFESEM (Thompson-Coffe et al. , 1996; Schatten et al. , 2000; Schatten and Chakrabarti, 2004). Direct visualization of centrosomes using HRLVFESEM is shown in Figure 5.6. Seen here is a centrosome that had been isolated from mitotic sea-urchin eggs, treated with cold (0◦ C) for 24 hours to compact centrosomal material and image the cold-aggregated centrosome structure. Particle structures of 1-2μm are embedded within a structural matrix. 3) Visualization of resin-extracted de-embedded thick sections De-embedding of thick-sectioned biological material is a unique approach to examine the interior of cells and tissue. By combining TEM preparation techniques with three-dimensional analysis of thick-sectioned material from which resin has been removed, remarkably detailed insights can be obtained of internal cellular structures and their relationship with other cellular components. The method described here can be applied to a variety of different cells and tissues and is not harsh on cellular structures which had been a problem with previous resin extraction methods. In addition, antigenicity of proteins is retained which allows labeling of thick-sectioned material in which interior cell structures are brought to the surface, therefore overcoming penetration problems of immunogold labeling. The thick-sectioned, de-embedded material imaged with HRLVFESEM also provides excellent contrast and depth-of-field for stereo-imaging which greatly facilitates interpretation and analysis. The low beam voltage reduces the need for sample coating and allows detailed topographical contrast with little radiation damage. The method has an advantage over other techniques, such as freeze-fracturing, which depends on random fracture planes through the cell, since serial sections can be obtained and the entire cell or tissue can be evaluated by 3-D reconstructions of thick-sectioned material, In addition, the area of interest can be tilted and viewed from various angles for accurate interpretation.
156
Heide Schatten
Fig. 5.6 LVFESEM image of centrosomes isolated from cold-treated mitotic sea-urchin eggs displaying 1-2 μm repeating subunits embedded within a structural matrix. Magnification displayed on lower right of micrographs. Reprinted with permission from Thompson-Coffe et al. (1996).
The method presented here is based on an approach introduced by Iwadare et al., (1990) using a modified potassium-methoxide/crown-ether complex. A crown-ether complex epoxy resin removal kit is now commercially available (Polysciences Inc, Warrington, PA). Previous resin-extraction methods had also been used with success (Winborn, 1976; Erlandsen et al. , 1973, 1979; Baskin et al. , 1979; Rodning et al. , 1980) but required elaborate protocols using sodium or potassium methoxide for removing Epon, which is harsher than the method introduced by Iwadare et al.
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
157
[1990]. Ris and Malecki (1993) were the first to apply the modified potassiummethoxide/crown-ether complex method to analyze the nuclear pore complex (NPC) and insect striated muscle after chemical fixation or cryo-immobilization of samples that had been processed and embedded in Epon 812. The method was then applied to other research applications. The three examples presented here are all the result of collaborations with the late Hans Ris. Presented here are highlights from previously published papers. All applications using the modified potassium-methoxide/crownether complex in combination with HRLVFESEM were analyzed with the Hitachi S-900 instrument in Madison, WI. Visualization of small delicate structures in these de-embedded thick sections is possible because of the minimal coating requirements when imaging with HRLVFESEM. The power of this unique method is demonstrated in the following 3 applications: (3.1) Nuclear pore complexes (Ris and Malecki, 1993); (3.2) Muscle fibers (Ris, 1993); (3.3) Toxoplasma parasite internal structures. As with all EM methods, extreme care must be used to avoid potential artifacts, as described in section 1 of this chapter. The need for careful critical point dryingto avoid water of ethanol residues (Ris 1985) is re-emphasized. This is important for all EM specimen preparations.
Materials and Methods: Plastic embedding and preparation of thick sections. Samples are embedded in Epon as routinely performed for TEM. Epon is used in a softer preparation mixture than is used for ultrathin sections. Semi-thick sections of 200 nm are cut at the slowest possible speed. Sections are picked up with a copper-wire loop and transferred onto coverglass chips coated with carbon and made sticky with 3aminopropyltriethoxysilane (Aldrich Chemical Company, Milwaukee, WI) (Ris and Malecki, 1993). For Epon resin extraction the Polysciences EpoxyResin Removal Kit (Polysciences Inc.) is used. Sections are then critical-point dried as described in section 1 and coated with a thin layer (ca. 1nm) of Pt deposited by Argon ion-beam sputtering. FESEM: The de-embedded critical point-dried sections are then viewed with a Hitachi-S-900FESEM, or similar instrument, operating at 1.5 kV accelerating voltage.
Application 1: Nuclear pore complex (NPC). The tissues described in following applications 1 and 2 were rapidly frozen by plunging into melting ethane (Malecki, 1992) and transferred to a Balzer’s FSD010 freeze-substitution apparatus. The nuclear envelopes were freeze-substituted in acetone with 0.5 % osmium tetroxide for 2 days at −80◦ C and 1 day at −20◦ C. The preparation was slowly warmed to room temperature, washed three times with acetone, infiltrated overnight in an acetone-Epon mixture (1:1) and inverted over
158
Heide Schatten
the final Epon mixtures in a silicone rubber flat embedding mold. Polymerization is achieved overnight at 37◦ C, 24 hrs at 50◦ C, and 24 hrs at 60◦ C. The samples are stripped off the Epon block and sections 200–300 nm thick are placed on 5×10mm carbon-coated glass chips that had been previously coated with 3aminopropyltriethoxysilane. Epon extraction from sections is achieved by holding the glass chip vertically in 5 ml of the Polysciences Epoxy-removal crown-ether complex for 4 min. After washing in distilled water, the preparation was dehydrated through an ethanol series (50, 70, 80, 95, and 100 %) and critical point dried. The preparations were imaged with the Hitachi S-900 SEM operating at 1.5 kV. If needed, a thin layer of Pt (ca. 1 nm) is deposited by argon-ion beam sputtering (Kemmenoe and Bullock, 1983; Franks et al., 1980). Figures 5.7 and 5.8 illustrate the information gained from applying the described technique. The whole-mount image of the surface of the isolated nucleus in Figure 5.1 is compared with a de-embedded thick-section of the isolated nucleus in Figure 5.7. As described above, the whole-mount preparation (Figure 5.1) clearly displays the surface of a nuclear envelope that is covered with the cytoplasmic components of many NPCs. The intranuclear surface in the upper area of Figure 5.1 reveals fishtrap-like structures at the intranuclear surface. A comparable image obtained with the de-embedding method confirms the structural details seen in the whole-mount preparation and in the tangential section (Figure 5.7) additionally shows the eight subunits of the nuclear pore arranged in a circle of about 120 nm in diameter (arrow). A central particle is also seen. In Figure 5.7b, the 120-nm ring is associated with eight fibers projecting upward and terminating in a 50-nm ring (arrow). The NPCs are seen here in side view which reveals more information than would be available with thin-section TEM. Further detail of the fishtrap-like structure is revealed in cross-sections shown in Figure 5.8. The figure legends are taken from the original paper (Ris and Malecki, 1993) and describe the new obervations in detail.
Application 2: Insect flight muscle. Figures 5.9 and 5.10 illustrate the results from thick-sectioned, de-embedded material from Drosophila flight muscle and show the complex three-dimensional organization of myosin, actin, and myosin cross-bridges. The results are compared with whole mounts obtained with high voltage TEM (HVEM). The preparation of flight muscle from Drosophila virilis followed protocols for dissecting the muscle in Karnovsky’s fixative (0.1 M Hepes buffer, pH 7.4, 2 % paraformaldehyde, 1.25 % glutaraldehyde) and transfer to fresh fixative for 30 min. Following rinses in 0.1M Hepes buffer, the muscle tissues were transferred to 1 % osmium tetroxide for 10 min, rinsed in distilled water, stained with 1 % uranyl acetate for 10 min, dehydrated in ethanol, and embedded in Epon 812. Sectioning and resin-extraction followed the protocols described above for nuclear pore complexes. Figure 5.9a shows a stereo-image of a longitudinally-sectioned (200 nm) myofibril, sectioned parallel to the myofilaments. The Z-line is clearly seen with the
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
159
Fig. 5.7 (a) Stereoimage of a tangential section (250 nm thick) through an isolated NE of a Xenopus laevis oocyte, after Epon extraction. This image shows the cytoplasmic surface with many NPCs. The NPCs have the same appearance as in whole mounts of the NE (Fig. 5.1). Bar, 200 nm. (b) Stereo image of a tangential section through an isolated NE from a Xenopus laevis oocyte. The intranuclear surface is exposed here. Several polar views of the fishtraps are visible (arrows), resembling the fishtraps seen in whole mounts (see Fig. 5.1). Of special interest are the side views of fishtraps (black arrowheads), which can be obtained only with the method described in this article. Bar, 200 nm. Reprinted with permission from Ris and Malecki (1993).
attached actin filaments. A 250-nm section imaged with HVEM is shown for comparison in Figure 5.9b. Cross-sectioned myofibrils are shown in Figure 5.10 in which HRLVFESEM of the de-embedded material (Figure 5.10a) is compared to a HVEM of sectioned material in Figure 5.10b. In the de-embedded preparation, the regular hexagonal organization of the myofilaments with cross-bridges of thin filaments
160
Heide Schatten
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
161
between them (Figure 5.10a) is seen while the cross-bridges are hardly visible in the comparable HVEM image (Figure 5.10b).
Application 3: Internal structures of the parasite Toxoplasma gondii. The following application is a new approach to our Toxoplasma project described in section 1 of this chapter. Several methods have been employed to study hostcell invasion and interactions within the parasitophorous vacuole (PV) and all have contributed excellent information. Conventional TEM (Aikawa et al. , 1977), SEM (Klainer et al. , 1973), and freeze-etch replication (Sulzer et al., 1974) electron microscopy have provided insights into various aspects of parasite-host cell interactions, but the de-embedding technique has proven most suitable to study parasitehost cell interactions in the PV. It provided a new approach to look at internal structures and the three-dimensional organization of specific structures and their interactions with other cellular components.
Materials and Methods As described in section 1, human fibroblasts and Toxoplasma parasites are cultured on coverslips and fixed at various time points after parasite invasion (Schatten and Ris, 2002; Schatten et al. , 2003) each for 30 min in 1 % glutaraldehyde in 0.1M Hepes buffer, pH 7.4 containing 0.05 % saponin and 0.2 % tannic. Washes in 0.1M Hepes buffer are followed by postfixation in buffered 0.1 % OsO4 for 10 minutes, staining with 1 % uranyl acetate, dehydration on EtOH, and embedding in Epon 812. Semi-thick sections (200 nm) are cut orthogonally to the cell substrate and ribbons are collected on cover-glass strips that will fit into the Hitachi-S-900 stage. Epon is extracted from the sections with the Polysciences Epoxy-removal crown-ether complex. After washes in distilled water and dehydrations in ethanol, these sections
Fig. 5.8 (a) Stereo image of a cross-section (250 nm thick) through an isolated NE from an oocyte of Xenopus laevis. A fishtrap in profile is visible at the arrow. Bar, 200 nm. (b) Stereo image of a cross-section (250 nm thick) through an isolated NE from a Xenopus laevis oocyte. A side view of a fishtrap is seen at the arrows. The tips of the arrows point to the small ring at the top of the fishtrap. A cable-like structure is attached to this ring. The nature of this ring is better seen in whole mounts. Bar, 200 nm. (c) Whole mount of an isolated oocyte nucleus from Xenopus laevis with the intranuclear surface exposed. The nucleus was isolated in LSB, attached to a glass chip, extracted with 0.1 % Triton-X-100 and 1 % glutaraldehyde in LSB, postfixed with 1 % aqueous osmium tetroxide, and critical point dried. The tops of adjacent fishtraps are connected to each other through a complex cable system. Bar, 200 nm. Reprinted with permission from Ris and Malecki (1993).
162
Heide Schatten
Fig. 5.9 (a) Stereo image of a longitudinal section (250 nm thick) through a thoracic flight muscle from Drosophila virilis. The Z line is seen in the lower left corner. At the arrow the section is properly oriented to show two adjacent thick filaments and an actin filament between them. Several periodic cross-bridges connecting thick and thin filaments are visible. Bar, 200 nm. (b) Longitudinal section (250 nm thick) through a myofibril from Drosophila thoracic flight muscle photographed with the Madison HVEM at 1 MeV. By careful adjustment of the tilt and rotation, the section was oriented so that myofilaments in the Z axis overlapped. The arrangement of thick and thin filaments is clearly shown, but the cross-bridges are barely visible. Bar, 200 nm. Reprinted with permission from Ris and Malecki (1993).
are critical point dried and coated with a thin layer of Pt (about 1 nm) which is deposited by argon-ion beam sputtering. Sections are imaged with a Hitachi S-900 FESEM operating at 1.5kV. Conventional TEM of thin-sectioned material is shown for comparison.
Parasite-host cell interactions in the parasitophorous vacuole: During host-cell invasion, Toxoplasma forms a specialized membrane-surrounded compartment, called the parasitophorous vacuole (PV) consisting of fibrous and
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
163
Fig. 5.10 (a) De-embedded samples: Stereo image of a cross-section (250 nm thick) through a thoracic flight muscle from Drosophila virilis. The hexagonal arrangement of thick and thin filaments is clearly visible (compare with the high voltage TEM image in b). The tubular structure of the thick filaments is also noticed in the SEM picture of extracted sections (i.e., above the arrow). Where thin filaments are oriented orthogonally to the section plane, they are clearly visible as a bright dot between adjacent thick filaments (e.g., arrowheads). Individual myosin bridges between thick and thin filaments stand out sharply. Viewed in stereo, looking down into the section, one can see the helical arrangement of cross-bridges along a thick filament (e.g., above arrows). Bar, 200 nm. (b) HVEM samples: Cross-section (250 nm thick) through thoracic flight muscle from Drosophila, photographed with HVEM at 1 MeV. The section was oriented so that it was orthogonal to the imaging beam. The regular arrangement of thick and thin filaments is clearly preserved, but the individual myosin cross-bridges are barely visible. Bar, 200 nm. Reprinted with permission from Ris and Malecki (1993).
164
Heide Schatten
tubular material that undergoes time-dependent changes. The parasitophorous vacuole membrane (PVM) is formed by components of the host cell as well as the parasite. While residing in the parasitophorous vacuole, the parasite needs to interact with the host cell to extract nutrients for survival, growth and multiplication and to modify the PVM for its protection. The mechanisms and structural components that allow communication between the vacuole and the host cell plasma membrane are not fully understood. Transmission electron microscopy had identified a nanotubular protein-rich network (Sibley, 1995; Sibley et al. , 1986; Mercier et al. , 1998, 2002) but thin-section TEM was not able to fully identify the complex interactions between the parasite and PVM, as this mode of cell imaging only allows imaging of a small slice of the entire vacuole containing the parasite. By using our de-embedding methods, we were able to visualize, in three dimensions, continuous structural
Fig. 5.11 Conventional transmission electron micrograph of thin-sectioned Toxoplasma. Seen here are three parasites within the PVM surrounded by tubular and fibrous material that is only seen fragmented because thin sections only reveal part of the tubular and vesicular structures.
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
165
interactions between the host cell membrane and the parasite within the PV, consisting of fibrous and tubular material that undergoes time-dependent changes. Figures 5.11, 5.12, 5.13 and 5.14 are comparisons of TEM images (Figure 5.11) and HRLVFESEM images of de-embedded thick sections (Figures 5.12,5.13,5.14) of parasites in the PV. The complex tubular and fibrous network connecting the parasite with the PVM, is easily discernable in Figures 5.12 and 5.13 while only short fragments are seen in Figure 5.11. The conventional TEM image in Figure 5.11 shows the parasite’s cell organelles as beautifully described by Aikawa et al., (1977) and include mitochondria, nucleus, apicoplast, ER, and 3 secretory systems: the micronemes, eight rhoptries and dense granules that are involved in PV formation (reviewed by Schwartzmann and Saffer, 1992). Micronemes (Carruthers and Sibley, Carruthers; Carruthers et al. , 1999) and rhoptries (Nichols et al. , 1983) are the first secretory systems to be involved in the formation of the PV during parasite-host cell invasion. The TEM image (Figure 5.11) depicts the most prominent secretory structures (rhoptries) within the parasite that are seen in one dimension. The rhoptries are filled with heavy OsO4 stain which makes structural analysis or immunolocalization of internal rhoptry proteins difficult. De-embedded material shows in 3-D some rhoptry associations with cytoplasmic components and structural features that line the rhoptry interior (Figure 5.12). These images demonstrate that the de-embedding technique offers new possibilities to identify components within the rhoptries that are not visible in conventional TEM a result of heavy osmication.
Fig. 5.12 High-resolution field emission scanning electron micrograph of Epon de-embedded section. Oblique section of a parasite in the process of host-cell invasion. The three-dimensional imaging of Toxoplasma gondii-host cell membrane interactions reveals numerous bridges. A fibrous network connects the parasite membrane with the membrane of the host cell. Rhoptries connected by fine bridges to the parasite’s cytoplasm are seen within the parasite.
166
Heide Schatten
Fig. 5.13 High resolution field emission scanning electron micrograph of Epon de-embedded section. The micrograph shows an oblique section of a parasite in the process of host cell invasion. The arrow points to newly-resolved structures on the parasite surface, some of which contact the host-cell surface during invasion.
Fig. 5.14 High resolution field emission scanning electron micrograph of Epon de-embedded section. The micrograph displays a parasite inside the parasitophorous vacuole with material that connects the parasite to the host-cell membrane. Rhoptries and dense granules are depicted within the parasite.
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
167
References Aikawa, M., Komata, Y., Asai, T., and Midorikawa, O. (1977). Transmission and scanning electron microscopy of host cell entry by Toxoplasma gondii. Am. J. of Pathology 87, 285–296. Baskin, D.G., Erlandsen, S.L., and Parsons, J.A. (1979). Influence of hydrogen peroxide or alcoholic sodium hydroxide on the immunocytochemical detection of growth hormone and prolactin after osmium fixation. J. Histochem. Cytochem. 27:1290–1292. Bestor, T. H. and Schatten, G. (1981) Anti-tubulin immunofluorescence microscopy of microtubules present during the pronuclear movements of sea urchin fertilization. Dev. Biol. 88, 80–91. Boveri, T. (1901). Zellen-Studien IV: Ueber die Natur derCentrosomen. Jena. Zeitschr. Naturwiss. 35:1–220. Carruthers, V.B., V.B., Giddings, O.K., and Sibley, (1999). Secretion of micronemal proteins is associated with toxoplasma invasion of host cells. Cell Microbiol. 1:225–235. Carruthers, V.B. and Sibley, (1997). Sequential protein secretion from three distinct organelles of Toxoplasma gondii accomnpanies invasion of human fibroblasts. Eur. J Cell Biol. 73: 114–123. Dobrowolski, J.M., and Sibley, L.D. (1996). Toxoplasma invasion of mammalian cells is powered by the actin cytoskeleton of the parasite. Cell 84, 933–939. Dubey, J.P., Lindsay, D.S., and Speer, C.A. (1998). Structures of Toxoplasma gondii tachyzoites, bradyzoites, and sporozoites and biology and development of tissue cysts. Clinical Microbiol. Review 11(2), 267–299. Erlandsen, S.L., Thomas, G.W., and Wendelschaefer, G. (1973). A simple technique for correlating SEM with TEM on biological tissue originally embedded in epoxy resin for TEM. Scanning Electron Microsc. 350–356. Erlandsen, S.L., Parsons, J.A., and Rodning, C.B. (1979). Technical parameters of immunostaining of osmicated tissue in epoxy sections. J. Histochem. Cytochem. 27:1286–1289. Franks, J., Clay, C.S., and Peace, G.W. (1980). Ion beam thin film deposition. Scanning Electron Microsc. 1:155–162. Gard, D. L. (1991) Organization, nucleation, and acetylation of microtubules in Xenopus laevis oocytes: a study by confocal immunofluorescence microscopy. Dev. Biol. 143, 346–362. Harris, P., Osborn, M., and Weber, K. (1980) A spiral array of microtubules in the fertilized sea urchin egg cortex examined by indirect immunofluorescence and electron microscopy. Exp. Cell Res. 126, 227–236. Harris, P. J. (1986) Cytology and immunocytochemistry. Methods Cell Biol. 27, 243–262. Hertzler, P. L. and Clark, W. H., Jr. (1993) The late events of fertilisation in the penaeoidean shrimp Sicyonia ingentis. Zygote 4, 287–296. Iwadare, T., Harada, E., Yoshino, S., and Arai, T. (1990). A solution for removal of resin from epoxy sections. Stain Technol. 65:205–209. Jarnick, M., and Aebi, U. (1991). Toward a more complete 3-D structure of the nuclear pore complex. J. Struct. Biol. 107:291–308. Kamioka, H., Ishihara, Y., Ris, H., Murshid, S.A., Sugawara, Y., Takano-Yamamoto, T., and Lim, S-S. (2006). Primary cultures of chick osteocytes retain functional gap junctions between osteocytes, and between osteocytes and osteoblasts. Microsc. Microanal., in press. Kemmenoe, B.H., and Bullock, G.R. (1983). Structure analysis of sputter-coated and ion-beam sputter-coated films: A comparative study. J. Microsc. 132:153–163. Klainer, A.S., Krahenbuhl, J.L., and Remington, J.S. (1973). Scanning electron microscopy of Toxoplasma gondii. J. Gen. Microbiol. 75:111–118. Lee, J., Miyano, T., and Moor, R. M. (2000) Spindle formation and dynamics of gamma-tubulin and nuclear mitotic apparatus protein distribution during meiosis in pig and mouse oocytes. Biol. Reprod. 5, 1184–1192. Malecki, M., and Ris, H. (1991). Preparation of cell suspensions for ultrastructural studies. Scanning 13:82–83.
168
Heide Schatten
Malecki, M. (1992). Light microscopy of living cells correlative to high-voltage electron microscopy and low-voltage scanning electron microscopy of cell cryo-whole mounts. Proc. Annu. Meet. Electron Microsc. Soc. Am. 50:566–567. Malecki, M., and Ris, H. (1992). Surface topography and intracellular organization of human cells in suspension as revealed by scanning electron microscopy. Scanning 14:76–85. Manandhar, G., Schatten, H., and Sutovsky, P. (2005). Centrosome Reduction during Gametogenesis and its Significance. Biol. Repro., 72:2–13 Meng, L. and Wolf, D. P. (1997) Sperm-induced oocyte activation in the rhesus monkey: nuclear and cytoplasmic changes following intracytoplasmic sperm injection. Hum. Reprod. 5, 1062–1068. Mercier, C.M., Cesbron-Delauw, M.F., and Sibley, L.D. (1998). The amphipathic alpha helices of the Toxoplasma protein GRA2 mediate post-secretory membrane association. J. Cell Sci. 111:2171–2180. Mercier, C.M., Dubremetz, J-F., Rauscher, B., Lecordier, L., Sibley, L.D. and Cesborn-Delauw, MF. (2002). Biogenesis of nanotubular network in Toxoplasma parasitophorous vacuole induced by parasite proteins. Mol. Biol. Cell 13:2397–2409. Nichols, B.A., Chiappino, M.L., and O-Connor, G.R. (1983). Secretion of the rhoptries of Toxoplasma gondii during host-cell invasion. J. Ultrastr. Res. 83:85–98. Pinder, J.C., Fowler, R.E., Dluzewski, A.R., Bannister, L.H., Lavin, F.M., Mitchell, G.H., Wilson, R.J., and Gratzer, W.B. (1998). Acto-myosin motor in the merozoite of the malaria parasite, Plasmodium falciparum: implications for red cell invasion (in process citation). J. Cell Sci. 111, 1831–1839. Ris, H. (1985). The cytoplasmic filament system in critical point-dried whole mounts and plasticembedded sections. J. Cell Biol.100:1474–1487. Ris, H. (1991). The three-dimensional structure of the nuclear pore complex as seen by high voltage electron microscopy and high resolution and low voltage scanning electron microscopy. Electron Microsc. Soc. Am. Bull. 21:54–56. Ris, H., and Malecki, M. (1993). High resolution field emission scanning electron microscope imaging of internal cell structures after Epon extraction from sections: a new approach to correlative ultrastructural and immunocytochemical studies. J. Struct. Biol. 11, 148–157. Ris, H. (1998). Low voltage field emission Sem: Tool for structural cell biology. ICEM14, Electron Microscopy IV, 439–440. Rodning, C.B., Erlandsen, S.L., Coulter, H.D., and Wilson, I.D. (1980). Immunohistochemical localization of IgA antigens in sections embedded in epoxy resins. J. Histochem. Cytochem. 28:199–205. Sabin, A.B. (1941). Toxoplasmic encephalitis in children. J. Am. Med. Assoc. 116, 801–807. Salmon, E. D. (1982) Mitotic spindles isolated from sea urchin eggs with EGTA lysis buffers. Methods Cell Biol. 25, 69–105. Schatten, H., Schatten, G., Mazia, D., et al. (1986) Behavior of centrosomes during fertilization and cell division in mouse oocytes and sea urchin eggs. Proc. Natl. Acad. Sci. USA 83, 105–109. Schatten, H., Walter, M., Mazia, D., Biessmann, H., Paweletz, N., Coffe, G., and Schatten, G. (1987). Centrosome Detection in Sea Urchin Eggs with a Monoclonal Antibody Against Drosophila Intermediate Filament Proteins: Characterization of Stages of the Division Cycle of Centrosomes. Proc. Natl. Acad. Sci. USA 84, 8488–8492. Schatten, H., Walter, M., Biessmann, H., and Schatten, G. (1992). Activation of Maternal Centrosomes in Unfertilized Sea Urchin Eggs. Cell Motil. Cytoskel. 23, 61–70. Schatten, H., Hueser, C.N., and Chakrabarti, A. (2000). From fertilization to cancer: The role of centrosomes in the union and separation of genomic material. Microscopy Research and Technique 49:420–427. Schatten, H., and Ris, H. (2002). Unconventional specimen preparation techniques using high resolution low voltage field emission scanning electron microscopy to study cell motility, host cell invasion, and internal structures in Toxoplasma gondii. Microscopy and Microanalysis 8: 94–103.
5 HRLVFESEM Applications for Cell Biology and Specimen Preparation Protocols
169
Schatten, H., Sibley. D., and Ris, H. (2003). Structural evidence for actin filaments in Toxoplasma gondii using high resolution low voltage field emission scanning electron microscopy. Microscopy and Microanalysis 9:330–335. Schatten, H., and Ris, H. (2004). Three-dimensional Imaging of Toxoplasma gondii- Host Cell Membrane Interactions. Microscopy and Microanalysis, 10:580–585. Schatten, H., and Chakrabarti, A. (2004). Detection of centrosome structure in fertilized and artificially activated sea urchin eggs using immunofluorescence microscopy and isolation of centrosomes followed by structural characterization with field emission scanning electron microscopy. In: Methods in Molecular Biology, vol. 253: Germ Cell Protocols: Vol 1 Sperm and Oocyte Analysis. Edited by H. Schatten, Humana Press Inc, Totowa, NJ, p. 151–164. Schliwa, M., van Blerkom, Porter, K. (1981). Stabilization of the cytoplasmic ground substance in detergent-opened cells and a structural and biochemical analysis of its composition. Proc Natl Acad Sci USA 78, 4329–4333. Schwartzmann, J.D., and Saffer, L.D. (1992). How Toxoplasma gondii gets into and out of host cells. Subcell. Biochem. 18:333–364. Shaw, M.K., Compton, H.L., Roos, D.S., and Tilney, L.G. (2000). Microtubules, but not actin filaments, drive daughter cell budding and cell division in Toxoplasma gondii. J. Cell Sci. 113: 1241–1254. Sibley, L.D., Krahenbuhl, J.L., Adams, G.M.W., and Weidner, E. (1986). Toxoplasma modifies macrophage phagosomes by secretion of a vesicular network rich in surface proteins. J. Cell Biol. 103:867–874. Sibley, L.D., (1995). Invasion of vertebrate cells by Toxoplasma gondii. Trends Cell Biol. 5, 129–132. Sluder, G., Miller, F. J., Lewis, K., et al. (1989) Centrosome inheritance in starfish zygotes: selective loss of the maternal centrosome after fertilization. Dev. Biol. 131, 567–579. Staiber W. (1994) Immunofluorescence study of spindle microtubule arrangements during differential gonial mitosis of Acricotopus lucidus (Diptera, Chironomidae). Cell Struct. Funct. 19, 97–101. Striepen, B., Crawford, M.J., Shaw, M.K., Tilney, L.G., Seeber, F., and Roos, D.S. (2000). The plastid of Toxoplasma gondii is divided by association with the centrosomes. Sun, Q-Y., and Schatten, H. (2006a). Centrosome inheritance after fertilization and nuclear transfer in mammals. In: Somatic Cell Nuclear Transfer, ed: Peter Sutovsky, Landes Bioscience, in press. Sun, Q-Y., and Schatten, H. (2006b). Regulation of dynamic events by microfilaments during oocyte maturation and fertilization. Reproduction 131:193–205. Sun, Q-Y., and Schatten, H. (2006c). Multiple roles of NuMA in vertebrate cells: review of an intriguing multifunctional protein. Frontiers in Bioscience 11:1137–1146. Thompson-Coffe, C., Coffe, G., Schatten, H., Mazia, D., and Schatten, G. (1996). Cold-Treated Centrosome: Isolation of the Centrosomes from Mitotic Sea Urchin Eggs, Production of an Anticentrosomal Antibody, and Novel Ultrastructural Imaging. Cell Motil. Cytoskel. 33: 197–207. Winborn, W.B. (1976). Removal of resins from specimens for scanning electron microscopy, in M.A. Hayat, (Ed.), Principles and Techniques of Scanning Electron Microsocpy, Vol. 5, pp. 21–35.
Chapter 6
Molecular Labeling for Correlative Microscopy: LM, LVSEM, TEM, EF-TEM and HVEM Ralph Albrecht and Daryl Meyer
Introduction As described in detail elsewhere in this book, the development of field-emission sources has made it possible to form an intense beam of low-voltage (low-V0 ) electrons with a small beam diameter. This permits effective scanning electron microscope (SEM) imaging at low accelerating voltage (LVSEM). Good contrast, reduced radiation damage, and less specimen charging are all associated with the use of low V0 (Joy 1989); (Pawley & Erlandsen 1989), see also Chapter 3). Furthermore, such an imaging system has many advantages for imaging low-contrast biological specimens labeled with colloidal-metal particles (now often called colloidal-metal nanoparticles or simply nanoparticles). At low kV, (V0 ∼1.3 kV), very thinly coated (1 to 2 nm.), or completely uncoated, specimens can be observed without charging as long as they are relatively flat.1 By limiting beam penetration in such low-density biological material (carbon), the use of low V0 permits one to image such surfaces at high-resolution. When metal nanoparticle labels are imaged with V0 = 1.5 kV, what is actually visualized is the surface of the protein (i.e., the antibody or ligand, or the active fragment of the antibody or ligand) covering the metal particle. When viewed at highresolution, this allows one to actually analyze the material linking the label particle to the surface structures of the specimen (Lai 1992; Goodman & Albrecht 1996). As V0 increases, the electron beam penetrates through the protein coat to the metal core. Thus, simply by increasing V0 to 3 or 5 kV, the secondary electron (SE) images of metal particles as small as 2 to 3 nm begin to show up as substantially brighter than the surrounding biological material, and individual particles can be readily identified and confirmed (Pawley & Albrecht 1988; Simmons & Albrecht 1989. See also Fig. 2.6). With more heavily-coated specimens or where charging is a problem, highresolution, backscattered-electron imaging (BSE) can be effectively employed. Current BSE detectors function with V0 as low as 2 kV, but improved BSE detector
1
Charging can remain an issue on many topologically complex specimens.
H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
171
172
R. Albrecht, D. Meyer
performance and smaller beam diameter (resolution) are apparent at slightly higher kVs, where a greater fraction of the signal comes from scattering in the metal. Classically, colloidal-gold labels were viewed at accelerating voltages in the 10–20 kV range. Thus, on uncoated or thinly-coated specimens, the high atomic number of the gold produces an SE or BSE signal that is much stronger than that from the surrounding organic material. Thick coatings of carbon or metal diminish the ability to detect Z-contrast, particularly at low accelerating voltages. Small particles (<8 nm) attached to conjugated molecules can sometimes be identified by shape alone; however, this is not reliable for specimens in which other small spherical molecular species are present. To positively identify small labels that are beneath the cell surface or are covered by the labeling antibody or ligand molecules, one must use V0 >5 kV, whether one is using SE or BSE. For uncoated or thinly-coated small (<5 nm) particles, the brightness of the metal core increases as V0 increases to about 5 kV. The same increase in beam penetration that permits one to generate signal from the subsurface metal, however, also decreases the contrast of fine surface details on biological specimens (Albrecht et al. 1993). As noted previously, when uncoated specimens are viewed at low kV, often the antibody or ligand coating the individual colloidal metal particles is imaged with no indication, other than shape, that a colloidal particle resides within. As V0 increases, scattering in the heavy metal reveals the presence of the metal particle. It is often very useful to compare the low-kV SEM images (for high surface resolution) to the higher kV images (for particle identification). If V0 increases to 10 or 20 kV, one begins to see particles inside, or even under, cells. When this occurs, particles on the top surface can be separated from subsurface particles by using stereo-pair images. In cells originally prepared with a thin metal coating for correlative LVSEM, followed by whole-mount transmission electron microscopy (TEM), heavy-metalstained intracellular cytoskeletal elements and membrane structures become readily visible in SEM images made at higher kV (∼15–20 kV), using either the SE or BSE signal (Albrecht et al. 1993). As discussed below, heavy-metal colloidal particles can be produced in different sizes and shapes and using different metals. Although the image brightness of all the heavy metal particles initially increases with V0 , distinctive particle shapes can be seen more readily at higher V0 , where proportionally more signal is obtained from the particles themselves and less from the surrounding biological tissue. While one can use an SEM or a STEM equipped with an EDX detector to discriminate particles based on their elemental composition, the low-signal levels available from conventional instrumentation generally limit such studies to particles larger than 10 nm. On the other hand, by permitting higher beam currents (and therefore higher data rates,) future SEMs with aberration-corrected lenses, improved detectors, and more sophisticated software may soon permit the automatic imaging and identification of small particles, based on elemental composition, at practical rates of data acquisition and accuracy (See also Chapter 3 and 11.). Finally, cells prepared on thin substrates and appropriately stained, can also be observed using TEM, intermediate-voltage TEM (IVEM), or high-voltage TEM (HVEM) for high-resolution studies of particle distribution relative to internal
6 Molecular Labeling for Correlative Microscopy
173
ultrastructure. As described in the next section, particles can also be differentiated on the basis of composition using electron energy loss (EELS) imaging (Meyer et al. 1996).
General Properties of Molecular Labels for LVSEM Historically, specific molecules present in biological SEM samples were labeled at lower resolution using various synthetic and biological particles that were easy to identify because of their unusual composition, shape, or size. These particles were coupled to antibodies or ligands to provide an unambiguously identifiable label. The particles used included microspheres of polystyrene, silica, or polyacrylamide, as well as various viruses such as phages, phage tails, TMV, and bushy stunt virus. Early on, even red blood cells were conjugated to antibodies to serve as markers (Albrecht & Wetzel 1979). Alternatively, enzymes were located using histochemistry to deposit precipitates that could be imaged either directly or after being stained with heavy-metal compounds. Peroxidases, esterases, and phosphatases were often localized in this manner (Wetzel & Albrecht 1989). Most of the approaches involving particles were limited in spatial resolution and quantitative accuracy, while the metal precipitates were often fairly large and the labeling systems often difficult to control.
Properties of Labels Relative to LVSEM (Shape/Size and Composition) Shape/size As the LVSEM improved, high spatial resolution labeling became possible, utilizing colloidal-gold spheres from 3 nm to 100 nm. Particles larger than 5 nm can be easily identified by their regular round shape, particularly when they are attached to surfaces. Although coating for LVSEM is minimal, and is in some cases completely unnecessary, any coating degrades the ability to visualize the smaller labels—when they are subsurface or embedded in antibody or ligands, they can be difficult to detect at very low kV (∼1.5 kV). For multiple labeling or colocalization, the classic approach was to use particles of different sizes. To avoid ambiguity, the size classes had to be sufficiently different not to overlap. Unfortunately, no more than two distinguishable label sizes are available in the molecular or submolecular spatial resolution range. Once outside this range, labeling resolution becomes supramolecular. Where epitope density is sufficiently high, epitope-masking can occur. In this case, the label on one type of epitope masks nearby epitopes of another type—a problem that will clearly be more severe on larger particles. Consequently,
174
R. Albrecht, D. Meyer
quantitative or semiquantitative studies are restricted to the smallest particle sizes because, in these cases, a single, monovalent antibody active fragment (Fab, or active ligand fragment) is bound to a single, colloidal particle. This gives a particle to epitope ratio of 1:1 and each label corresponds to one detected epitope. Because larger particles have greater surface area, they may contain more than one active site, and therefore each one can label either one or more than one surface epitope (depending on particle size and epitope density). The larger the particle is, the greater this problem becomes. As before, using particles of different sizes has the problem of epitope-masking, but apart from this, one can never be sure that the ratio of label to epitope is 1:1 for particles larger than 3 to 5 nm. One solution is to produce small labels with differing shapes. In addition to spheres, particles can be made in various geometric shapes—such as pyramids—as well as with irregular or popcorn surfaces. Other heavy metals such as palladium, platinum, or silver can be used to produce characteristic spherical, irregular, and faceted particles2 (see Figs. 6.1 (see color plate 1) and 6.2 below). The use of particles having different shapes permits the use of multiple particles of similar size, minimizing the steric problems discussed above. We have also utilized a methodology that produces particles with extremely small size variation. This allows the use of more than two classes of very small particles simultaneously—for example, 3 nm, 5 nm, and 7 nm. Although this keeps the labeling generally within the molecular resolution limits, and minimizes masking, the amount of time required to accurately measure the size and/or shape of such small particles when they are actually adherent to labeled cells or tissues makes this approach impractical at present (Meyer & Albrecht 2002). While smaller sizes are necessary for high spatial resolution, they also are preferred for better penetration of the label into tissues. In this context, one must remember that the important variable is the overall size of the label—i.e., the particle plus the labeling molecule, such as an antibody, antibody fragment, ligand, or ligand fragment. Simply decreasing the size of the colloidal particle will only reduce steric hindrance if doing so reduces the overall probe size. With smaller particles, this may involve reducing both the particle size and the size of the labeling molecule—for example, using a Fab fragment instead of a whole antibody molecule. Methods that use smaller particles as a nucleating agent for the deposition of additional silver (Danscher 1981; Stierhof et al. 1991; Scopsi 1989) or gold (Hainfeld et al. 1999) permit one to increase particle size, making each particle easier to visualize in the SEM or even by light microscopy. Such chemical enhancement procedures permit one to use small particles to enhance penetration or to label densely-packed epitopes, and still view the particles easily in the SEM. However, one must use careful technique to assure uniform particle growth. 2 Metals such as copper, iron, and nickel can also be prepared and then visualized in LVSEM with somewhat less contrast. These elements are more reactive, particularly in biological environments, and stabilization and conjugation to active label molecules, such as antibodies, is somewhat more difficult. Good results may require a core-shell approach, where the core metal particle is first coated with a thin shell of a more stable metal such as gold, with surface properties that are more conducive to the stable attachment of protein labels (Bleher et al. 2004; Bleher et al. 2007).
6 Molecular Labeling for Correlative Microscopy
175
Fig. 6.2 (a) Examples of various faceted colloidal labels including pyramids, hexagons, and cubes. (line = 100 nm). (b) Example of multiple labeling on a cell surface with particles of differing shape, size, and composition. (c) Portion of a platelet surface showing labels of four distinct shapes. (1) = 18 nm colloidal-gold is conjugated to mouse anti-human P-selectin a marker of platelet activation. (2) = popcorn or umbonate 18 nm colloidal palladium conjugated to human factor X, a subunit of the thrombinase complex. (3) = hexagonal 15 nm colloidal palladium conjugated to sheep anti-human factor V IgG, demonstrating location of endogenous human factor V, another subunit of the thrombinase complex. (4) = 5 nm colloidal-gold particles conjugated to exogenous human factor V. Line = 100 nm
Multiple labeling is often compromised by chemical enhancement because enhancement tends to be somewhat irregular, masking any small differences in the size of the original nucleating particles. Enhancement of small particles of different shapes generally results in particles that are more or less round, depending on the level of enhancement. The relative enhancement that is possible with different types of metal remains to be delineated.
176
R. Albrecht, D. Meyer
Composition Although, historically, most of the labels used in high-resolution SEM studies have been colloidal-gold, as noted earlier, it is possible to produce colloidal particles from a number of other elements including platinum (Pt), silver (Ag), palladium (Pd), ruthenium (Ru), rhodium (Rh), and iron (Fe). It is generally desirable to synthesize particles from those heavy metals that are nonreactive in biological systems. For metals other than gold (Au) or Pd, the conjugation of identifier molecules is not always straightforward. These problems can sometimes be avoided by constructing core-shell particles, in which a core of a desired metal such as Pt or Fe is surrounded by a thin but continuous shell of Au. As noted previously, this provides a practical solution to the difficulties dealing with protein conjugation and the stabilityreactivity of some metals in a biological environment (Bleher et al. 2004; 2007).
Correlative Studies For correlative studies, specimens are usually first viewed by light microscopy and then prepared either as whole-mount cells on thin substrate films supported by EM grids or as semithin sections. They can then be examined either uncoated or lightly coated (usually 1 nm Pt) by LVSEM, and then by TEM (Sims et al. 2006; Meyer 2006; Bleher 2005; Wetzel & Albrecht 1989).3 Energy-filtering TEM (EFTEM) imaging can be used to unambiguously identify very small particles based on their elemental composition as revealed by their energy-loss spectrum imaging (ESI). This is obviously very useful for molecular and submolecular labeling, whenever the simultaneous detection or colocalization of multiple epitopes or binding sites is required. This sequence of observations also allows one to image the surface structure of large numbers of particles first using LVSEM, and then clearly and rapidly identify their composition using EF-TEM (Koeck and Leonard 1996; Bleher 2005; Meyer 2006; Bleher et al. 2007).
Correlative Labeling of Living Specimens for LM and EM It is often useful to label living cells or tissue with colloidal labels to track particle movement in the light microscope. Subsequently, one may wish either to know the relationship of the labeled species to the structural elements and surface features of the cell, or to understand their relationship to other labeled species at levels of resolution requiring SEM, TEM, or both. 3 Please note that electron irradiation in the SEM, and even more severely in the TEM, destroys protein structure and hence can modify both the epitope/binding site presence and its availability. Thus, immunolabeling or ligand labeling of cells after they have been viewed by EM is very unreliable.
6 Molecular Labeling for Correlative Microscopy
177
Several approaches to correlative labeling are suitable for samples to be viewed by LM followed by LVSEM. The most obvious is to combine fluorescent labels with those using colloidal metal particles. Although one can purchase secondary antibodies that are said to be conjugated to both fluorescent markers and to 15 to 30 nm colloidal-gold particles, placing metal particles of any size in close proximity to a fluorescent marker produces nearly 100% quenching of the fluorescence signal. This is true across a range of excitation and emission wavelengths and with a wide variety of different fluorescent dyes (Kandela and Albrecht, 2007). These preparations often appear to work only because, during manufacture, not all of the fluorescent antibody molecules are conjugated to gold particles, and those not so coupled can still fluoresce. This can be readily demonstrated simply by ultracentrifugation. When the heavy-metal conjugates have been removed, a substantial amount of non-gold-conjugated antibody remains in the supernatant, and this fluoresces normally. The gold-conjugated antibody at the bottom of the tube shows no fluorescence even though each label molecule is attached to both the gold and a fluorescent dye. As a result, when using such a mixed conjugate label, some epitopes are labeled with the combined gold/fluorescent antibody and are visible in EM (but invisible in the LM), while other epitopes are labeled with only the fluorescent antibody. This can give the false impression that direct, correlative labeling has occurred, and although such labeling may be useful is some circumstances, it should only be used when epitopes are plentiful, the two conjugates have similar affinities, a high labeling efficiency is not required from the particle-conjugated antibody, and quantitative or even semiquantitative results are not important (Meyer 2006: Kandela 2004). For correlative labeling with more precision, we have investigated a double antibody-labeling approach where the primary antibody is conjugated with the heavy-metal particle and a second antibody is then coupled to a fluorescent dye. Coupling the primary antibody—or active antibody fragment—to the colloidal particle insures that the label visible in the LVSEM is located as closely as possibleto the actual position of the epitope. Having the fluorescent dye on the second antibody does not diminish the resolution attainable with the light optics, but it does substantially reduce the quenching caused by the metal particle (see Fig. 6.3). With a 5-nm metal particle coupled to a fluorescent dye by a second antibody, quenching is only ∼50%. With a 15 nm particle, the quenching is approximately 70%, but this is generally well within the sensitivity of the CCD cameras currently in use on fluorescence microscopes. If one must reduce quenching further, the fluorescent dye can be coupled to a third antibody—a procedure that reduces quenching to less than 10%, again without compromising LM spatial resolution (Kandela et al. 2004, 2005,2007; Kandela and Albrecht 2007).
Imaging Colloidal Labels in the LM Interference-based light microscopy—either video-enhanced differential-interference contrast (DIC) or asymmetric illumination interference contrast (AIC)—can be used to image and track either individual colloidal metal particles or groups of
178
R. Albrecht, D. Meyer
Fig. 6.3 Model system utilizing colloidal-gold/anti-fibrinogen antibody conjugates to label a fibrinogen monolayer on a flat substrate. Y-axis is particle density (particles/um2 ) while the X-axis is minutes of labeling time. Solid lines show the theoretical labeling intensity, while data points are actual particle concentrations. Upper lines are for particle concentrations of 3.9×1012 particles/ml in the labeling solution and the lower lines are for 3.66×1011 particles/ml. The upper lines diverge as labeling reaches a maximum and steric hindrance prevents additional labels from reaching the surface and binding. Maximal labeling occurs at about 50 to 60 minutes. The lower lines do not diverge, as surface labeling reaches only about 10% by one hour, hence no hindrance is occurring. At lower concentrations maximal labeling may take days to even weeks (Park & Simmons 1988)
them (Simmons et al. 1990). In the LM, we have imaged individual particles a small as 10 nm conjugated to ligands specific for cell-surface receptors. Although the shape of the particles is not resolved in the photon-based system, their inflateddiffraction image can be seen and one can track the centroid of each diffraction image with nm accuracy (Gelles, et al., 1988). The diffraction image appears dark with gray, somewhat indistinct edges and is roughly the size of the resolution limit of the microscope. Although these images can be readily tracked in living cell systems, they are substantially larger than the particles themselves so particle numbers must be limited to insure only one particle is being followed. Groups of even smaller particles (∼8 nm) produce dark areas that can also be tracked. The position of the particles relative to structural elements inside the cell, or on its surface, can then be determined by LVSEM and/or TEM/EF-TEM following fixation, dehydration and drying (see Figs. 6.4, 6.5, and 6.6 (see color plate 2)). Unfortunately, as particles of different sizes and shapes appear similar in the LM, one can only track one type of label at a time in real time. However, subsequent LVSEM or TEM can be used to identify the exact same particle(s) tracked in the living preparation and to differentiate one particle type from another at this point.
6 Molecular Labeling for Correlative Microscopy
179
Fig. 6.4 (a) A time-series of DIC images, showing initial binding of Au-conjugated fibrinogen (black) on the surface membrane over the subjacent peripheral web and outer filamentous zone of the cytoskeleton of a fully-spread, substrate-adherent human blood platelet. This label binds to the integrin receptor for fibrinogen on platelets. Over several minutes, bound labels are transported over the surface, toward the platelet center. They come to rest, still on the membrane surface, but now overlying the inner filamentous zone. The platelet is then fixed, stained with osmium and uranyl-actetate, and dried by the critical-point procedure for subsequent LVSEM and HVEM. Bar = 1 µm. (b) Surface image of the same uncoated platelet as Figure 6.4a, viewed in LM at 1kV accelerating voltage in a modified Hitachi S-900 SEM. It shows the platelet surface and the fibrinogen-covered individual gold particles in detail. (c) SEM at 5 kV, still in the SE mode. The increased beam penetration clearly demonstrates the location of the gold particles (bright spots) relative to stained internal cytoskeletal structures. Surface structures are less apparent Bars = 1 µm. (d) Shows an SEM image of the same platelet as seen in (a) and (b), using a beam voltage of 20 kV and somewhat higher magnification. This demonstrates the relationship between the labels on the membrane surface and the underlying cytoskeletal organization. The higher accelerating voltage results in greater beam penetration so that labels can now be seen in the outer filamentous zone area. These were not visible at lower kV and represent particles on the bottom, substrate adherent, surface of the platelet. Bar = 0.25 µm. (e) An HVEM of the same platelet as seen in a, b, c, and d. Legend: P = peripheral web, OF = outer filamentous zone, IF = inner filamentous zone, G = granulomere, mt = microtubules, dark arrowheads = margin of inner filamentous zone, white arrowheads in Fig. 6.3 point to labels trapped under the platelet. In (d) and (e), Bars = 0.25 µm
180
R. Albrecht, D. Meyer
Fig. 6.4 (continued) (f) HVEM stereo pair of a whole-mount of the same platelet as seen in (a) through (e), which offers a clear view of the platelet cytoskeleton. The gold labels (black spots) are clearly seen in relationship to the subjacent platelet cytoskeleton. Bar = 1 µm
Quantum Dots Several years ago, we investigated quantum dot particles for use in LM/EM correlative labeling. These particles ranged in size from 3 to 10 nm, and were composed of semiconductors such as silicon (Si), or CdSe surrounded by ZnS. They fluoresce with high efficiency when excited by a wide range of visible wavelengths.4 Because the wavelength of the light emitted by the quantum dot is directly proportional to its diameter, it seemed likely that one could use multiple labels that could be distinguished in both the LM (on the basis of emission wavelength) and the
4 Including far-red wavelengths, a factor that assists the penetration of excitation photons deep into tissue.
6 Molecular Labeling for Correlative Microscopy
181
Fig. 6.5 Shows how DIC imaging can be used to track the movement of individual Au-label particles across the surface of a fully-spread platelet. (a) shows the DIC image at the conclusion of particle movement. Black areas are individual particles or particle groups, or platelet granules. (b) shows the tracking data following individual particles as they move centripetally across the platelet surface. (c) shows the same platelet as seen in (a) and (b) but following fixation, staining, and dehydration, and imaged using the LVSEM at 1.5 kV accelerating voltage. Arrows point to individual particles tracked in (b). (d) is a stereo-pair HVEM image of the same platelet. Arrows indicated tracked particles at the time of fixation and demonstrates the position of the particle labels relative to internal structure. Legend: P = peripheral web, OF = outer filamentous zone, IF = inner filamentous zone, G = granulomere zone, m = microfilament bundles. Arrows point to individual labels also seen in Fig. 6.7c. The asterisk indicates a particle tracked in Fig. 6.7b and also seen in Figs. 6.7c and 6.7d. Labels fixed in transit generally still appear on the membrane surface over the outer filamentous zone while labels that have completed their movement are generally seen over the inner filamentous zone. Bar = 1.0 um
182
R. Albrecht, D. Meyer
Fig. 6.5 (continued)
TEM/SEM (on the basis of size or elemental composition). Because quantum-dot fluorescence is very resistant to fading, repeated imaging and even back-and-forth imaging between EM and LM might be possible. Our goal was to use quantum dots of different elemental composition and different sizes within the molecular-labeling range to produce labels that could be conjugated to antibodies or active fragments, imaged in the LM by fluorescence (and possibly also using interference contrast), and then visualized in the LVSEM by shape and at higher V0 by either SE or BSE imaging. Finally, ESI could be used to differentiate particles based on their size and elemental composition. Unfortunately, we found that several issues still needed to be addressed. The solubility and reactive nature of the dots required that they be coated with several materials to make them both stable and nontoxic in a biological environment and also to permit antibodies or ligands to be effectively coupled to them. Once this was accomplished, the final size of the dot, plus the coatings, was on the order of 20 to 40 nm—a fact that compromised both molecular or submolecular spatial resolution and a 1:1 ratio of label to epitope (Alivisatos et al. 2005). The relatively small size of the actual dots, their density, and the atomic number of the elements comprising them also made it difficult to differentiate them using either SE or BSE detection modes in the SEM, no matter what V0 was used. Even in standard TEM, quantum dots can be difficult to visualize and often are nearly impossible to recognize in conventionally-stained sections. In addition, ESI imaging in an EF-TEM was insufficiently sensitive to allow one to characterize such small, low-density particles on the basis of the small differences present in their metallic composition. Improvements in the techniques are occurring, however, and as the sensitivity of the instrumentation increases, headway beyond our initial assessments is now possible (Nisman et al. 2004; Deerinck et al. 2005).
6 Molecular Labeling for Correlative Microscopy
183
In conclusion, in the same way that dyes having differing excitation and emission, wavelengths permit simultaneous multiple labeling and colocalization in light microscopy, colloidal metal labels of differing shape and composition permit multiple labeling and colocalization studies at molecular and submolecular levels of spatial resolution in the LVSEM, and permit correlative LM/SEM/TEM studies (Kandela et al. 2007a; 2007b).
Preparation and Imaging of Biological Specimens Specific methodologies are needed to prepare biological specimens for correlative imaging using LM, LVSEM, TEM, and even atomic force microscopy (AFM). Proper preparation allows one to see labeled cellular structures and individual molecules within, or on, cells using multiple microscopic modes—each of which can view exactly the same cell or molecule.
General Preparation of Specimens for LVSEM and Correlative High-Resolution Imaging Because LVSEM imaging permits high-resolution with minimal coating, membrane integrity and associated membrane structures can be viewed clearly. It is important, therefore, to insure that membrane and nonmembrane structures are preserved as faithfully as possible with respect to the living state. Other chapters (particularly Chapter 5) deal with preparative methodology, so we will not repeat that information here other than to list in the next section, a typical procedure employed in our studies to minimize extraction artifacts and preserve membrane integrity.
Correlative Imaging Starting with LM of Living cells, Then Preparing for LVSEM, TEM and Dry AFM5 Cells adhered to a suitable substrate are initially fixed while being observed with the LM, or immediately after their removal from the LM stage. For subsequent correlative SEM, FESEM, TEM, EF-TEM, HVEM and dry AFM, cells are fixed for 30 min in 1% glutaraldehyde in 0.1 M HEPES buffer, pH 7.3, containing 0.2% tannic acid and 0.5% saponin. After a HEPES buffer rinse, cells are post-fixed in HEPES-buffered 0.05% OsO4 Maupin-Szamair & Pollard (1978) for 15 min in 5
(Goodman et al. 1991; Eppell et al. 1995; Albrecht Meyer 2002)
184
R. Albrecht, D. Meyer
the dark, rinsed in distilled water, and stained with 1% uranyl acetate for 15 min (Loftus & Albrecht 1983, 1984; Loftus et al. 1984). Cells are then dehydrated using a graded series of ethanol to 100% ethanol that has been dried over recentlyregenerated, Type 3A (0.3 nm exclusion pore diameter) molecular sieve and are then dried by the critical-point procedure using CO2 as the transitional fluid. The criticalpoint apparatus is equipped with an in-line, molecular-sieve water-vapor trap, and a hydrophobic filter. Dried samples are stored over molecular sieve or in vacuo. The presence of trace amounts of water during critical-point drying (CPD), either from incomplete dehydration or as a contaminant in the CO2 , may lead to artifactually flocculated microtrabecular-like structures (Ris 1985). Alternatively, ultra-rapid freezing and freeze drying can also be used to dehydrate and dry cells in preparation for EM. This may eliminate the need for chemical fixation and substantially reduces shrinkage associated with drying by the criticalpoint procedure (Albrecht & MacKenzie 1975). It is also possible to use various freeze-substitution procedures, as well as direct examination of small samples frozen by high-pressure freezing or plunged into slush nitrogen, and observed using a cryostage (see also Chapters 8 and 10).
General Comments on the Generation of Colloidal Metal Particles The utmost care must be taken in the production of colloidal metal sols. To obtain reproducible results, particular attention must be paid to using glassware and other vessels that are unblemished and properly cleaned, and to insure that all reagents and solutions are of the highest purity. Scratches and other blemishes can serve as unintended sites of particle nucleation, adversely affecting the particle size distribution. Glassware should also be washed thoroughly with a nonionic detergent followed by copious rinsing with double-distilled H2 O (ddH2 O) to remove any dirt that might otherwise nucleate particles and any ionic species that might cause the colloidal suspensions to flocculate by thinning the ionic double layer surrounding each particle. This double layer maintains particle stability by the interaction of mutually repulsive electrostatic forces. If the electrostatic double layer becomes too weak, Brownian motion will cause particles to approach one another closely enough for van der Waal’s and London dispersion forces to bind them irrevocably together, forming aggregates of ever-increasing size. Eventually these aggregates settle out of the suspension. As the effectiveness of ions in collapsing the charge double layer is proportional to the sixth power of their valence, even very small amounts of a divalent or trivalent cation can induce particle aggregation. For similar reasons, reagents should be of a pure grade and all solutions should be filtered prior to use. While synthesis of colloidal nanoparticles does require attention to detail, we hope this does not discourage the reader from making his/her own colloidal labels. For additional information and instructions, including a half-hour video on preparation and conjugation of colloidal-gold nanoparticles, see our Biological and Biomaterials Preparation, Imaging, and Characterization web page at the University of
6 Molecular Labeling for Correlative Microscopy
185
Wisconsin’s Department of Animal Sciences (http://www.ansci.wisc.edu/facstaff/ Faculty/pages/albrecht/albrecht_web/Programs/microscopy/colloid.html).
Particle Size Colloidal-gold (cAu) particles having a range of average diameters may be obtained by reducing dilute solutions of tetrachloroauric acid with any of a number of agents, each of which will yield a specific particle-size range (Handley 1991). Reduction with white phosphorus produces 3 nm or 5 nm cAu particles, while sodium citrate is commonly used to synthesize particles with diameters ranging from 16 nm to several hundred nm. Particles of intermediate sizes can be obtained by reduction with a combination of sodium citrate and tannic acid. The color of the colloid varies with size (Fig. 6.1c, see Color plate 1).
3 nm cAu Some of the gold sols that Michael Faraday made using white phosphorus that he famously characterized in 1857 are still extant, and are located at the Faraday Museum of the Royal Society in London (Faraday 1857). His method was subsequently modified by Turkevich et al. (1951) and later by Roth (1982) To generate these small particles, the following procedure can be used: the pH of a 100 ml solution of 0.01% tetrachloroauric acid is adjusted slightly alkaline of neutrality, then brought to a boil. Using a magnetic stir bar to insure rapid mixing, 0.5 ml of a saturated solution of white phosphorus in diethyl ether is rapidly injected. The color changes immediately to an orange-red, indicative of the sol formation. The sol is boiled an additional 10 to 30 minutes to ensure that the reaction has come to completion. The saturated solution of white phosphorus in diethyl ether is prepared by adding finely-diced shavings of phosphorus to a few mls of ether, followed by purging of the container with nitrogen or argon, and stirring with a magnetic stir bar until a cloudy suspension develops. The suspended particles are allowed to settle overnight before the saturated solution is used in the reaction. A saturated solution of copper sulfate is kept on hand to neutralize excess phosphorous and clean up any spills. CAUTION: White phosphorus must always stored under water and will spontaneously combust if left in air. Particle preparation is best carried out in a chemical hood and—as for all particle synthesis—gloves, eye protection, and appropriate laboratory clothing should be used.
3–17 nm cAu Slot and Geuze (1985) modified a technique introduced earlier by Mühlpfordt (1982), in which a mixture of tannic acid and sodium citrate was used as the reducing agent. By varying the concentration of the tannic acid relative to the citrate
186
R. Albrecht, D. Meyer
concentration, one could produce monodisperse cAu sols having particles of discreet diameters ranging from 3 to 17 nm. To make 100 ml of the sol, 1 ml of a 1% solution of tetrachloroauric acid is added to 79 ml ddH2 O and heated to 60 ◦ C. The reducing solution is prepared by adding together 4 ml, 1% trisodium citrate, 0 to 5 ml 1% tannic acid, a volume of 25 mM potassium bicarbonate equal to that of the tannic acid used, and sufficient ddH2 O to bring the total volume to 20 ml. The reducing solution is then warmed to 60 ◦ C, after which it is added quickly to the gold solution with rapid mixing provided by a magnetic stir bar. Upon development of a reddish color, the timing of which varies from instantaneous to nearly an hour depending upon the concentration of the tannic acid, the sol is brought to a boil and then permitted to cool. The particle size is inversely related to the tannic acid concentration.
>16 nm cAu Suspensions of cAu particles larger than 16 nm are prepared as described by Frens (1973). The procedure is analogous to the tannic acid method previously described, in that the strength of the reducing agent used determines the particle size. In this case, however, tannic acid is not added and sodium citrate acts as the sole reducing agent. Bring 200 ml of a 0.01% solution of tetrachloroauric acid to a boil, along with a stock solution of 1% trisodium citrate. Up to 4 ml of the citrate is then injected rapidly into the gold-containing flask with vigorous mixing. The color changes from clear to light purple to black as the auric ions are reduced to elemental gold atoms that then begin to nucleate. As additional metal condenses around the nuclei, the color again changes to various shades of red, ranging from clear, bright ruby for the smaller particle sizes, to violet for the intermediate sizes, and finally to an opaque rust with a bluish Tyndall effect for the largest particles. The sol is then refluxed an additional 20 minutes.
Colloids of Other Noble Metals The preparation of colloidal particles composed of other noble metals is similar to that of gold colloids in that a salt of the metal is chemically reduced in an aqueous medium. Sols of different mean particle diameter may be produced by using different reducing agents. Methods analogous to Fren’s citrate reduction technique, or to Slot’s and Geuze’s citrate/tannic acid technique, in which the size of cAu particles is controlled by the concentration of the reducing agent used, have proven largely inadequate for synthesizing sols composed of other noble metal elements over the same range of sizes. We have developed a procedure that utilizes differing concentrations of sodium ascorbate for the synthesis of cPd particles having mean diameters ranging from
6 Molecular Labeling for Correlative Microscopy
187
Table 6.1 Synthesis of colloidal palladium Sol
[Na Ascorbate] (%)
Average Diameter (nm)
¯ n (%) σn /X
1 2 3 4
0.2 0.05 0.01 0.005
6.6 ± 0.07 7.9 ± 0.09 10.1 ± 0.10 11.7 ± 0.17
14.7 17.8 15.0 21.5
6.6 to 11.7 nm, and having very small size variances (Meyer & Albrecht 2003: see Table 6.1). Apart from this particular method, a wholly different approach is usually necessary. One such approach is the Keimmethode (literally translated from the German, seed method) developed around the turn of the previous century by the Nobel laureate, Richard Zsigmondy (1906). Zsigmondy’s method used very small cAu particles, prepared using Faraday’s white-phosphorus reduction technique, as nuclei around which additional gold could be condensed during a second step that used another reducing agent. In this way, larger particles could be produced, and the size was dependent largely upon the ratio of the number of nuclei to the total amount of gold salt. Some years later Voigt and Heumann applied Zsigmondy’s work to the synthesis of uniformly sized cAg particles (Voigt and Heumann, 1928). We too have found this method satisfactory for the synthesis of cPt and cPd particles of graded sizes (Meyer & Albrecht 2002). See Table 6.2. Not only is Zsigmondy’s method applicable to the synthesis of monometallic colloidal particles of different sizes, but it is also useful for the synthesis of multimetallic, core-shell particles in which a small colloidal particle of one metal serves as the core around which a shell (or shells) of one or more other metals can be condensed (e.g., Henglein 2000; Hodak et al. 2001). We have used the multimetallic, core-shell approach to make gold-coated platinum particles, which are easier to conjugate to antibody ligands than are bare platinum particles (Bleher 2005; Bleher et al. 2007), as well as gold-coated magnetite particles that can be used to label cells. Once attached to a target cell, the magnetite can be inductively heated and used to produce membrane pores or to specifically kill the labeled cells without
Sol Nuc. Sol 1 2 3 4 5
Table 6.2 Synthesis of cPt using the nucleating sol procedure Volume Nucleating Sol Added (ml) Average Particle Diameter (nm) – 10.0 5.0 2.5 1.0 0.5
2.5 ± 0.03 5.5 ± 0.06 7.1 ± 0.06 9.0 ± 0.09 11.6 ± 0.12 13.4 ± 0.16
188
R. Albrecht, D. Meyer
damaging adjacent unlabeled cells. Neither the particles, nor the magnetic field, are harmful by themselves (Kaiser et al. 2007; Kandela 2006).
Fabricating Colloids of Different Shapes Popcorn cPd Solutions of 10% L-ascorbic acid and 1% dipotassium tetrachloropalladate are chilled in an ice water bath. To make 500 ml cPd sol, 4.1 ml of the Pd salt solution is added to the ascorbic acid with vigorous agitation. The colloid begins to develop within a minute, the flask is returned to ice, and the reaction is allowed to come to completion overnight. The mean particle diameter is roughly 18 nm (see Figs. 6.1a and 6.1b).
Faceted cPd To make 100 ml of the sol, 200 µl of 20% trisodium citrate and 820 µl 1% dipotassium tetrachloropalladate are combined in 94 ml ddH2 O. Once the contents of the flask have reached a rapid boil, 5 ml of 6.4% ascorbic acid is added with vigorous mixing. The colloid forms instantaneously and the flask is refluxed for 20 min. Although the particles described above provide considerable variety, we note that the current emphasis on nanotechnology is giving rise to a variety of new synthetic methodologies for the synthesis of various nanoparticles, and many or these may prove useful for labeling in the future (see Figs. 6.2a and 6.2b).
Conjugating Particles to Markers (Antibodies, Active Antibody Fragments, Ligands or Ligand Fragments) Proteinaceous ligands form the most stable conjugates with colloidal metal particles by hydrophobic interactions when the pH of the sol is slightly basic to the pI of the ligand, and the amount of protein available for conjugation is just sufficient to form a monolayer around each particle. To find this concentration—also referred to as the minimum protecting amount (MPA)—empirically, a concentration isotherm is performed. Small, incremental volumes of the ligand are added to each test tube of a dilution series, covering a range of perhaps two- to ten-fold. To these ligands, one then adds a constant volume of the sol, whose pH has been adjusted slightly basic to the pI of the ligand. Following gentle agitation, a saturated sodium-chloride solution equivalent to one-tenth the volume of the conjugate is added to each tube to test the stability of the conjugates.
6 Molecular Labeling for Correlative Microscopy
189
With unconjugated particles, the sudden influx of cations compresses the diffuse counter-ion electric double layer surrounding each particle. Because it is this electric double layer that is responsible for the stability of lyophobe sols, adding the salt destabilizes the sol, producing a color change because the light scattering of the aggregated particles changes as they grow in size. In the case of cAu18, the color shifts from red to blue and is apparent within a second or two. The change in color of substantially smaller cAu2–5 particles may develop over the course of several hours, presumably because the van der Waals forces between smaller particles are weaker. Ligands conjugated to colloidal metal particles form a physical barrier that prevents the particles from diffusing close enough to one another for aggregation to occur (Vervey & Overbeek 1948). The MPA is defined as the lowest concentration of ligand that prevents the color shift noted in the previous paragraph. If the pI of the ligand is unknown, it can be approximated in a similar manner. In this case, a constant volume of cAu at various pHs is added to tubes, each containing the same amount of ligand, and the stability of the conjugates is assessed following the addition of salt. Once the MPA and optimal pH have been determined, ligands and particles may be conjugated in bulk. Often, a stabilizing agent such as a small amount of PEG (a dimer of 20,000 MW) or bovine serum albumin (BSA) equivalent to 0.01% (w/v) may also be added to the conjugates of the larger nanoparticles (10 nm diameter and larger) to enhance stability. It is thought that the smaller molecules cover any surface of the nanoparticle not attached to the ligand or antibody, and also stabilize the ligand or antibody protein. The conjugates are then lightly pelleted by ultracentrifugation, not only to concentrate them for storage, but also to remove any excess ligand and/or stabilizer left behind in the supernatant. This step also tends to remove aggregates, which tend to form hard pellets that adhere to the walls of the tube. The supernatant is removed and discarded, and the pellet resuspended in medium or a buffer compatible with the tissue to be labeled. The relative particle concentration may be determined spectrophotometrically, and will—along with other variables such as ambient temperature, particle size, viscosity and the ionic strength of the medium, probe-target affinity, and steric hindrance— affect the amount of time necessary for labeling (Park 1988; 1989). A cautionary word is necessary here. Particle concentration is of particular concern. Both commercially-supplied particles and those that we may synthesize are often stored at particle densities in the range of 1012 particles/ml. Fig. 6.3 shows data from a model-labeling system employing a high-affinity antibody or ligand. While final particle concentrations of ∼5×1012 particles/ml result in complete labeling after about 60 min (theoretical and tested), a 1:10 dilution to 5×1011 particles/ml results in only ∼10 to 15% labeling after 60 min, and the time needed for complete labeling may be hours, or even days. The general shape of this concentration/labeling time relationship is true for all sizes of particles. Thus, to obtain a high labeling efficiency, the first concern should be particle concentration. Once one has a particle concentration sufficient to permit maximal theoretical labeling in the time available, one can address other considerations such as antibody affinity and steric interference.
190
R. Albrecht, D. Meyer
Methodology for Correlative Labeling for LM, LVSEM, and TEM Interference Imaging for Particle Identification and Tracking in LM Followed by Prep for High-Resolution SEM and/or TEM Imaging Various possibilities exist for tracking ligand- or antibody-conjugated colloidal particles on, or in, living cells. Although the colloidal particles used for labeling are an order of magnitude or more smaller than the Abbe limit, we have found that particles as small as 10 nm can be tracked using interference LM imaging of their inflated-diffraction image (Simmons 1990; Albrecht 1993; Goodman, 1996). Either DIC or AIC (Kachar 1985) imaging can be effective in conjunction with highresolution/sensitivity video imaging with suitable contrast enhancement. Because the inflated-diffraction image of any one particle appears much larger than the particle itself, care must be exercised that the number of labels being tracked at any time is small enough that particle-particle interference is minimized. It also must be kept in mind, however, that if one wishes to simultaneously label a number of receptor or antigenic sites, particularly on living cells, a higher concentration of label must be employed. Particle concentrations in the 2×1013 to 5×1013 /ml can work well with short labeling time (5 min or less), after which the unbound label is washed from the system. Generally, an inverted LM with a rectified DIC optical system is the most effective imaging system for particle tracking (see Figs. 6.4 a, b, c, and d). Fig. 6.5 shows how DIC imaging can be used to track the movement of individual Au-label particles across the surface of a fully-spread platelet. A sensitive CCD camera is also required. It is also possible to combine particle distribution/tracking studies with various fluorescent indicator dyes so that particle tracking can be combined with simultaneous measurements of calcium flux or cell pH (See Color Plate 2). Fig. 6.6b show images that provide a measure of the free internal [Ca++ ] as the particles move (Sims et al. 2006). Once particles are tracked in living cells, the cells can be prepared for correlative LVSEM and TEM or HVEM. The exact same particles tracked in LM can be viewed both in TEM or EF-TEM to identify particle composition and/or shape as well as their relationship to internal ultrastructure (as seen with TEM) or surface structure (as seen by SEM; see Figure 6.6c). If one wished to use correlative fluorescence imaging rather than DIC images for tracking individual particles, one can use a fluorescent second antibody or an antiligand antibody as described next.
Correlative Fluorescence LM and Heavy Metal Nanoparticle SEM/TEM As discussed previously, direct conjugation of both a fluorescent molecule and a colloidal heavy metal particle to the same identifier molecule (i.e., antibody, ligand, or
6 Molecular Labeling for Correlative Microscopy
191
active fragments of antibody or ligand) quenches the fluorescent signal. This occurs with all particles bigger than 3 nm and with all fluors across the UV, visible, and red excitation wavelengths (Kandela et al. 2003, 2004, 2007a; Kandela and Albrecht 2007). For correlative imaging, the primary antibody, typically a high-affinity monoclonal or affinity-purified polyclonal, is conjugated directly to the colloidal particle of choice. The second antibody is often an affinity-purified donkey or goat antiprimary antibody that is conjugated to the fluorescent dye molecule such as a dye from the Cy or Alexa-Fluor series. Preconjugated, affinity-purified, antisecondary antibodies can usually be purchased from a number of commercial venders. Using this technique, one can label a variety of sectioned, frozen-sectioned, or whole-mount preparations. Substrates vary. We often use finder grids supporting films of polyvinyl formal (Formvar), or polystyrene if subsequent observation includes LM, FESEM and TEM, or EF-TEM. Where TEM is not part of the correlative series, cells can be prepared on optically-clearable filter membranes or segments of coverslips. If prefixation is employed, samples are generally washed twice in a 0.05 M HEPES buffer for 2 to 5 min, followed by incubation in 0.05 M of Glycine in HEPES to neutralize reactive aldehyde groups. Nonspecific binding sites can be blocked with 0.1% BSA in HEPES, or some other suitable blocking agent, for 15 min. Primary antibodies are diluted in HEPES and the incubation performed for one hour at room temperature. Unbound antibodies are removed by washing five times for 3 min with HEPES buffer. In all cases, primary antibodies were detected with secondary antibodies in HEPES conjugated against one of the Cy or Alexa Fluor dyes. After washing, cells on substrates in Petri dishes with a coverslip built into the bottom, are covered with HEPES and observed with a suitable fluorescence LM with illumination and excitation/emission filters specific for each of the fluorescent dyes. For subsequent correlative studies involving detection of the colloidal-gold particles in LVSEM or EF-TEM, specimens on grids are prepared as described in the previous paragraphs for correlative SEM and TEM. Cells on grids can be post-fixed on drops of 2.5% glutaraldehyde in HEPES for 15 min and rinsed two times briefly with ddH2 O. Frozen sections can be air dried after covering with a thin layer of 2% methylcellulose in H2 O. Whole-mount specimens can be very rapidly frozen by plunging them into slush nitrogen or using high-pressure freezing and then freeze dried (Albrecht 1975). It is also possible to observe samples in the frozen state if a cryocoater, transfer device, and cryostage are available (see Chapter 10). Freeze substitution methodology is another option. Finally, samples can be dehydrated via a conventional alcohol series and dried by the critical-point procedure. No coating is required for EF-TEM or other TEM studies, while a very thin, 1-nm, ion beamsputtered coating of platinum may be useful for LVSEM observation. Figs. 6.4b through 6.4f show examples of LVSEM images of cells mounted on thin films that are made without any coating. Fig. 6.7 (see color plate 3) shows how particles can be categorized based on elemental composition as determined by EF-TEM, and Fig. 6.8 shows a doublelabeled specimen in which the double labels can be recognized and discriminated in both the fluorescence LM images and in the TEM and the EF-TEM images. In
192
R. Albrecht, D. Meyer
Fig. 6.8 Platelet sections were labeled with rabbit anti-alpha actinin-cAu6 (colloidal-gold 6 nm) and labeled with second (antirabbit) antibody conjugated to Cy2 fluorescent dye (green). The identical platelets were also labeled with mouse anti-actin-cPd6 (colloidal palladium, 6 nm) and then with anti-mouse second antibody, conjugated and to Cy3 fluorescent dye (red). (a) Bright, generally spherical spots indicate concentrations of alpha actinin. (b) The same field as (a), showing a more uniform distribution of actin and lack of discrete focal concentrations. Bar = 2.5 um. (c) and d) are zero-loss images of the platelet seen in the center right of (a) and (b). In the lower magnification image (c), the left (red) arrow points to the area of the bright dot showing the strong alpha-actinin staining, seen in (a). The right (blue) arrow points to several labels just to the right of the bright area which stained positive for actin. Bar = 400 nm. (d) shows an enlarged area of (c). Numerous 6 nm spherical labels are seen concentrated in the area of the fluorescent dot in (a). Individual 6 nm labels are seen singly or in small clusters elsewhere in the platelet cytoplasm. Bar = 100 nm. (e) is a zero loss image of the area containing the clustered 6 nm labels (seen in 6-8d), which corresponds to the Cy2 labeling for the alpha-actinin in (a). (f) and (g) show the results of ESI imaging, as in Fig.6.7. (f) is an ESI image showing the presence of gold, while (g) is the image for Pd. Nearly all particles in the densely-labeled area are colloidal-gold, thus indicating the presence of alpha-actinin molecules; no Pd signal is found in this area. (h), (i), and (j) show the zero loss, ESI analysis for gold, and the ESI analysis for palladium (respectively). No ESI signal for Au is seen in this area, but the 6 nm particles (mostly spherical but with some faceted particles) in this area are positive for palladium indicating the presence of actin molecules (Kandela et al. 2007a)
6 Molecular Labeling for Correlative Microscopy
193
Fig. 6.8 (continued)
this example, the two different EM labels can be discriminated both on the basis of shape and on the basis of their elemental composition (Kandela et al. 2007a).
Conclusion In the 1960s and 70s, scanning and transmission electron microscopy were the preeminent microscopical methods used in cell biology. Since then, new methods of interference and fluorescence contrast light microscopy—especially those that permit imaging in three dimensions—have become increasingly dominant. There are several reasons for this change. Light microscopy can be used to view specimens that are alive and interacting with their environment, and thanks to the ubiquitous availability of fluorescent dyes having great sensitivity and specificity, LM permits the observer to follow the motion of particular molecules, and to do so in three dimensions. Finally, LM specimen preparation is relatively easy compared to EM. On the other hand, the LM has severe limitations in terms of spatial resolution, and the ability to localize labels on the scale of nanometers can only be approached by indirect techniques such as fluorescence resonance energy transfer (FRET). This is the spatial scale best approached using various types of electron microscopes. The approaches described here allow the researcher to avoid having to choose between LM and EM and, instead, choose to use both on the exact same specimen. This process allows one to take advantage of the best features of each method. It is hoped that this chapter will encourage the reader to explore various combinations of these techniques.
194
R. Albrecht, D. Meyer
Acknowledgment The authors wish to acknowledge Dr. Reiner Bleher, Dr. Irawati Kandela, and Mr. Joe Heintz for their contributions to the preparation of this chapter.
References Albrecht RM, Simmons SR, and Pawley JB (1993) Correlative video-enhanced light microscopy, high voltage transmission electron microscopy, and field emission scanning electron microscopy for the localization of colloidal-gold labels. In: Immunocytochemistry: A Practical Approach. J.E. Beesley, ed. Oxford University Press, Oxford. pp 151–176. Albrecht RM, and MacKenzie AP (1975) Cultured and Free Living Cells. In Principles and Techniques for Scanning Electron Microscopy: Biological Applications. M.A. Hayat ed. Van Nostrand Reinhold, NY, 3:109–153 Albrecht RM and Meyer DA (2002) All that glitters is not gold: Approaches to labeling for EM. Microscopy and Microanlysis 8, suppl.2:194–195. Albrecht RM and Wetzel B (1979) Ancillary methods for biological SEM. Scan. Elec. Microsc. Vol. III:203–222. Alivisatos AP, Gu W, and Larabell C (2005) Quantum dots as cellular probes. Annual Review of Biomedical Engineering 7(1), 55–76. Bleher R, Meyer DA, and Albrecht RM (2004) Multiple labeling for EM using colloidal particles of gold, palladium, and platinum as markers. Microscopy and Microanalysis 10, Suppl.2:158–159. Bleher R, Meyer DA, and Albrecht RM (2005) High-resolution multiple labeling for immunoEM applying metal colloids and energy filtering transmission electron microscopy (EFTEM). Microscopy and Microanalysis 11, Suppl.-2:1100–1101. Bleher R, Kandela I, Meyer DA, and Albrecht RM (2007). Immuno-EM using colloidal metal nanoparticles and electron spectroscopic imaging for co-localization at high spatial resolution. J. Micros. In Press. Danscher G (1981) Localization of gold in biological tissue: a photochemical method for light and electron microscopy. Histochemistry 71:81 Deerinck TJ, Giepmans BNG, and Ellisman MH (2005) Quantum Dots as Cellular Probes for Light and Electron Microscopy. Microscopy and Microanalysis 11, Suppl.-2:914–915. Eppell SJ, Simmons SR, Albrecht RM, and Marchant RE (1995) Cell surface receptors and proteins on platelet membranes imaged by scanning force microscopy using immunogold contrast enhancement. Biophysical Journal 68:671–680. Faraday M (1857) Experimental relations of gold (and other metals) to light. Philosophical Transactions of the Royal Society of London. 147:145–181. Frens G (1973) Nature Physical Science. 241:20–22 Gelles J, Schnapp BJ, and Sheetz MP (1988) Tracking kinesin-driven movements with nanometrescale precision, Nature 331:450–453. Goodman SL, Park K, and Albrecht RM (1991) A correlative approach to colloidal-gold labeling with video-enhanced light microscopy, low-voltage scanning electron microscopy, and highvoltage electron microscopy. In: Colloidal-Gold: Principles, Methods, and Applications. Vol. 3. M.A. Hayat., ed. Academic Press, Inc., San Diego. pp 369–409. Goodman SL, and Albrecht RM (1996) Cell Surface Interactions: the Blood Platelet as a Paradigm. Ch 15 Interfacial Phenomena and Bioproducts, J.L. Brash and P.W. Wojciechowski (eds.), Marcel Dekker Inc., New York, NY publishers, pp. 485–506. Hainfeld JF, Powell RD, Stein JK, Hacker GW, Hauser-Kronberger C, Cheung ALM, and Schofer C (1999) Gold-based autometallography; Proc. 57th Ann. Mtg., Micros. Soc. Amer.; G. W. Bailey, W. G. Jerome, S. McKernan, J. F. Mansfield, and R. L. Price (Eds.); SpringerVerlag: New York, NY; 486–487. Handley DA (1989) The development and application of colloidal-gold as a microscopic probe. In: Colloidal-Gold: Principles, Methods, and Applications. Vol. 1. M.A. Hayat., ed. Academic Press, Inc., San Diego. pp 13–32.
6 Molecular Labeling for Correlative Microscopy
195
Henglein A (2000) Colloidal palladium nanoparticles: reduction of Pd (II) by H2 ; Pdcore Aushell Agshell particles. The Journal of Physical Chemistry B. 104(29):6683–6685. Hodak JH, Henglein A, and Harland GV(2001) Tuning the spectral and temporal response in PtAu core-shell nanoparticles. Journal of Chemical Physics. 114(6):2760–2765. Joy DC (1989) Control of charging in low-voltage SEM. Scanning 11-1:1–4. Kaiser M, Heintz J, Kandela I, and Albrecht RM (2007) Tumor cell death induced by membrane melting via Inductively Heated Core/Shell Nanoparticles. Microscopy and Microanalysis, 13, supp-2:18–19. Kandela IK, Bleher R, and Albrecht RM (2004), Correlative Labeling Studies in Light and Electron Microscopy. Micros & Microanaly 10, suppl-2:1212-1213. 2004. Kandela IK, Bleher R, and Albrecht RM (2005) Correlative Immunolabeling on Etched Epon Samples. Microscopy and Microanalysis 11, supp-2:1098–1099. Kandela IK and Albrecht RM (2007) Fluorescence quenching by colloidal heavy metal nanoparticles: Implications for correlative fluorescence and electron microscopy studies. Scanning 21:152–161. Kandela IK, Bleher R, and ALbrecht RM (2007a) Multiple correlative immunolabeling for light and electron microscopy using fluorphores and colloidal metal particles. J. Histochem. Cytochem. In Press. Kandela IK, Bleher R, and Albrecht RM (2007b) Immunolabeing for correlative light and electron microscopy on ultrathin cryosections. Micros. and Microanal. In Press. Kandela IK (2006) Development of Metal Nanoparticle Immunoconjugates for Correlative Labeling in Light and Electron Microscopy and as Active Targeted Delivery Systems. PhD Thesis, University of Wisconsin, Madison, WI Kachar B (1985) Asymmetric Illumination Contrast: A Method of Image Formation for Video Light Microscopy. Science 227:766–768. Koeck PJB and Leonard KR (1996) Improved Immuno Double Labeling for Cell and Structural Biology. Micron 27:57–165. Lai QJ, Simmons SR, Albrecht RM, and Cooper SL (1992) Protein and cell adhesion to block polymer microdomains, Transactions of the Society for Biomaterials. Lai QJ (1992) The role of proteins and cellular interactions on the blood compatibility of polymers and other artificial surfaces. PhD Thesis, University of Wisconsin, Madison, WI Meyer DA and Albrecht RM (2002) Size selective synthesis of colloidal platinum nanoparticles for use as high-resolution EM labels. Microscopy and Microanalysis 8, Suppl.-2: 124–125. Meyer DA and Albrecht RM (2003), Sodium ascorbate method for the synthesis of colloidal palladium particles of different sizes. Microscopy and Microanalysis 9-Suppl.-2:1190–1191. Meyer DA, Bleher R, Kandela IK, Oliver JA, and Albrecht RM (2006) The development of alternative markers for transmission electron microscopy and correlative transmission electron and light microscopies. Microscopy and Microanalysis 12, Suppl.-2:32–33. Mühlpfordt H (1982) The preparation of colloidal-gold particles using tannic acid as an additional reducing agent. Experientia 38:1127–1128. Nisman R, Dellaire G, Ren Y, Li R, and Bazett-Jones DP (2004) Application of Quantum Dots as Probes for Correlative Fluorescence, Conventional, and Energy-filtered Transmission Electron Microscopy. J. Histochem. Cytochem. 52–1:13–18. Park K, Simmons SR, and Albrecht RM (1988) Surface characterization of biomaterials by immunogold staining - Quantitative analysis. In Biotechnology and Bioapplications of Colloidal-Gold Albrecht RM and Hodges GM eds. Scanning Microscopy International, Chicago IL. pp 41–52. Park K, Park H, and Albrecht RM (1989), Factors affecting the staining with colloidal-gold. In: Colloidal-Gold: Principles, Methods, and Applications. Vol. 1. M.A. Hayat., ed. Academic Press, Inc., San Diego. pp 489–518. Pawley JB,Albrecht RM (1988), Imaging colloidal-gold labels in LVSEM, Scanning 10:184–189 Pawley JB and Erlandsen SL (1989) The case for low voltage high-resolution scanning electron microscopy of biological samples. Scanning Microscopy 3, Suppl-3:163–178.
196
R. Albrecht, D. Meyer
Roth J, (1982) The preparation of protein A-gold complexes with 3 nm and 15 nm gold particles and their use in labelling multiple antigens on ultra-thin sections. Histochemical Journal. 14–5:791–801. Scopsi L (1989) Silver Enhanced Colloidal-Gold Method.,In: Colloidal-Gold: Principles, Methods, and Applications. Vol. 1. M.A. Hayat., ed. Academic Press, Inc., San Diego. pp 251–295. Simmons SR and Albrecht RM (1989) Probe size and bound label conformation in colloidal goldligand labels and gold-immunolabels. Scanning Microscopy 3-Suppl. 3:27–34. Simmons SR, Pawley JB, and Albrecht RM(1990) Optimizing parameters for correlative immunogold localization by video-enhanced light microscopy, high-voltage transmission electron microscopy, and field emission scanning electron microscopy. J. Histochem. Cytochem. 38:1781–1785. Sims P, Albrecht RM, Pawley JB, Centonze V, Deerinck T, and Hardin J (2006) When Light Microscope Resolution Is Not Enough: Correlational Light Microscopy and Electron Microscopy, in Handbook of Biological Confocal Microscopy, 3r d edition, (Pawley, JB, ed.), Springer, NY, 49:846–860. Slot JW and Geuze HJ (1985) A new method of preparing gold probes for multiple-labeling cytochemistry. European Journal of Cell Biology. 38–1:87–93. Stierhof Y-D, Humbel BM, and Schwartz H (1991) J. Electron. Micros. Tech 17: 336. Turkevich J, Stevenson PC, and Hillier J (1951) A study of the nucleation and growth processes in the synthesis of colloidal-gold. Discussions of the Faraday Society. 11:55–75. Verwey EJW and Overbeek JThG (1948) Theory of the Stability of Lyophobic Colloids. Elsevier Publishing Co., Inc. New York. Voigt J, and Heumann J (1928) Die herstellung schutzkolloidfreier, gleichteiliger silberhydrosole. Zeitschrift für Anorganische Chemie. 169(1–3):140–150. Wetzel BK and Albrecht RM (1989) The Evolution of Correlative Techniques for Electron Microscopy – An Overview In: The Science of Biological Specimen Preparation. R.M. Albrecht and R.L. Ornberg (eds). Scanning Microscopy Supplement 3:1–6. Zsigmondy R (1906) Zeitschrift für Physikalische Chemie. 56:65.
Chapter 7
Low kV and Video-Rate, Beam-Tilt Stereo for Viewing Live-Time Experiments in the SEM Alan Boyde
Introduction Micro-Imaging Surfaces That Are Undergoing Experimental Procedures Almost all of our visual impressions of the world around us involve images of surfaces. Indeed, one of the great advantages of the scanning electron microscope (SEM) over previous types of microscope is that it forms images of surfaces that are easily interpreted by our neuro-visual system. This system has been conditioned from childhood to decode three-dimensional shapes from light intensity cues that mimic quite closely the relation between viewing angle and intensity that modulate the secondary electron (SE) signal in the SEM. Because the SEM provides such images at much higher resolution and with great depth of field than is available from our unaided eye, it is the ideal method for observing the course of experiments involving the cutting or dissection of specimens on a microscopic scale. This advantage is complemented by the large volume around the SEM specimen available for installing and operating the tools needed to cut or dissect the specimen. On the other hand, actually carrying out such experiments requires more threedimensional information than one can obtain from the standard, monocular view. Fortunately, one can use deflection coils to change the angle at which the imaging raster approaches the specimen surface and so make real-time stereo images of the experimental region. Techniques for doing so are discussed below in the context of their historical development. Although the mechanical disruption of the surface of metal-coated insulators will cause charging artifacts, this can be reduced to a manageable level by a combination of high-speed scanning and operating at low kV.
H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
197
198
A. Boyde
Static Three-dimensional Recording and Viewing The correct appreciation of unknown sample morphology in the SEM requires threedimensional visualization. In the early days of SEM, this was achieved via the recording of stereoscopic pairs of images that were later analyzed with an appropriate viewing system, e.g., a mirror stereoscope (Wells 1957; Boyde 1970a), polarized light or anaglyph (Boyde 1971a,b). As the three-dimensional image could only be seen in stereo after printing the photographic negatives, there was an inevitable delay between observation in the SEM and appreciation of what was there. Initially, images were recorded on 35mm film, and a whole 36-exposure roll had to be recorded and processed just to understand the actual three-dimensional morphology of the surface properly. While tedious, this procedure was not entirely without merit, however, because if the tilt angle difference between the stereo-pair images was accurately controlled (Boyde 1970b; Houghton et al. 1971), the visual impression could also be formally analyzed by stereophotogrammetry to reconstruct and measure all the XYZ coordinates of any features recorded in both members of the stereoscopic pair (Boyde 1970a, 1973, 1974). Somewhat later, the introduction of Polaroid materials reduced the time between recording the data and seeing stereo images to minutes, rather than hours, but was expensive. More recently still, the availability of multiple digital-image memories allows us to record tilt-series images at small-tilt-angle increments and then view them as a movie sequence to let motion parallax convey the depth information (Boyde 2003). In the time frame that I am addressing, a movie sequence could only be achieved by recording a series of images scanned at video rate on videotape, but even this did not provide real-time three-dimensional information.
Video-Rate LV SEM The realization of video-rate beam scanning in the SEM occurred in the late 1960s. We acquired such a TV scanning system in 1968 and recognized a constellation of problems and solutions. On the one hand, the need to acquire in 40ms an image that had traditionally been acquired in 40 seconds implied a dramatic increase in the secondary emission (SE) signal from the Everhart-Thornley biased-scintillator collector. This implied using substantially higher beam currents to generate the signal. Fortunately, it was discovered that, because the high-scan speed effectively spreads the beam over the entire imaged field, it did far less damage than would have occurred if the same beam power were concentrated on one point at a time, as in normal, slow-scan image acquisition. In addition, rapid scanning drastically reduced the problem generally referred to as “charging”1 and the remaining effects could be effectively eliminated by working at low accelerating voltage, around 2–3 kV.
1
In fact, the problem of changes in the brightness of the SE image that are not associated with the geometry or composition of the specimen is often caused by periodic discharging of the sample surface. Such abrupt changes in surface charge distort the electrostatic SE collection field,
7 Viewing Live-Time Experiments in the SEM
199
Damp-Sample SEM In nature, every biological sample is wet. By retaining a fraction of this water in the sample, its conductivity is sufficient for practical SEM purposes—but how can one work with damp samples in the SEM vacuum? At that time, we were working with the original, classical Cambridge Instrument Company “Stereoscan” SEMs. In these instruments, the electron-optical column was isolated from the sample chamber so that the two spaces could be pumped separately. In normal operation, one was prevented from switching on the filament and the high voltage until a predefined vacuum had been reached in both column and specimen chamber. A small aperture at the bottom of the column maintained a differential pressure between column and chamber after the two were reconnected by opening a lifting-and-sliding column-isolation valve. Our mode of operation was to disconnect the wire to the vacuum gauge from the sample chamber well before the normal operating vacuum had been achieved, so that we could then study damp and outgassing biological samples—and this was especially simple in TV mode. This work was the first practical application of wet or environmental SEM, which might better be termed bad- or poor-vacuum SEM. Thus low voltage combined with TV rate scanning and bad vacuum permitted SEM image acquisition—without the need for photography—from wet, or at least imperfectly dried, biological samples.
Video Image Recording Circa 1970 Image recording was achieved using one-inch videotape recorders (originally an Ampex machine with ten inch tape reels, and later an International Video Corporation (IVC) machine with eight inch reels). Both gave one hour of recording on a reel. The tape was expensive by our standards at the time, and older experimental records had to be deleted to record new material. We had no second recorder of our own and, therefore, no in-house means for editing the tape. In any event, we did not want to risk the loss of quality associated with rerecording and preferred to use primary recordings for demonstration and lecture purposes. Showing our results at meetings posed some real problems. To do so involved transporting the extremely heavy recorders with large TV monitors. We certainly could not demonstrate our achievements at long distance, for example at the Illinois Institute of Technology Research Institute (IITRI) series of SEM meetings in Chicago, the center of SEM information exchange at that time. There was also the problem of incompatibility between the US and European video standards. In essence, we could say what we had seen, but we could not show it.
causing uninterpretable linear brightness variations. Charging effects are discussed in more detail in Chapter 2.
200
A. Boyde
The Driving Force for Three-dimensional Video-Rate LV SEM In the early 1970s, we had a considerable interest in understanding the mechanical failure mechanisms operating in human dental enamel. In repairing the ravages of dental disease, the clinical approach adopted was to cut away the rotten (“carious”) tissue back to sound enamel and dentine, prior to the placement of a restoration. Enamel has a remarkable anisotropic structure that functions very well when the tissue is more or less intact and is supported by the underlying dentine. Exactly what happened when the support of nearby enamel and dentine was removed was not known. We, therefore, decided to use wet LV SEM to study cavity-margin preparation procedures, with experiments configured to fit into the SEM specimen chamber. We began with monoscopic video-rate SEM. To cut things in the SEM specimen chamber—which this work involved—meant to contrive situations where the field of operations was centered in the field of view in the SEM, and where one had the ability to move a cutting tool relative to the object to be cut. This was achieved in two ways. JB Pawley designed a robust micromanipulator that fitted through the left-hand-side port of the Stereoscan specimen chamber (see Fig. 7.1 (see color plate 4)), and this could be used to move a tool, either to cut the specimen while moving, or to center the tool on the electron-optical axis so the sample itself could be moved against the tool (Pawley & Boyde 1975). It does not take long to realize that the latter is generally the better approach: The focus remains constant, and it appears that the ”camera” is ”panning” with the dynamic cutting interaction (see Fig. 7.2 (see color plate 5)). We, therefore, used robust tool posts clamped to the base of the specimen stage to fasten the tool in the position in three-dimensional space corresponding to the center of the SEM field of view (see Fig. 7.3). An alignment jig enabled us to find this position when the specimen stage was removed from the SEM (see Fig. 7.4 (see color plate 6)).
Fig. 7.3 Stereoscan 1 specimen stage with tool post carrying a dental chisel clamped to its base. A tooth specimen was screwed to the moveable sample holder and the XYZ stage controls used to move the sample with respect to the stationary knife
7 Viewing Live-Time Experiments in the SEM
201
The use of fixed cutting tools required extreme care in judging the threedimensional position of the instrument with respect to the sample. Enamel is by far the hardest biological tissue and diamond knives are very expensive. The need to be able to control and visualize such cutting experiments drove us to develop a system for real-time three-dimensional SEM (Boyde et al. 1972). This system also proved to be hugely beneficial with all other SEM samples in that it allowed us to simply see and understand three-dimensional surface features in real time without any need for comprehensive photography. Once the stereo system was operating, there was a dramatic reduction in image recording, most of which had been done simply to be able to understand the three-dimensional image.
The Scheme for Video-Rate Three-dimensional SEM The scheme that we chose was to deflect the electron beam within the column— rather than within the specimen chamber as was later done in Canada by Eric Chatfield et al. [1974]. To do this required replacing the original scan-coil assembly inside the bore of the final (third) condenser lens (i.e., the objective lens) of the Stereoscan with a new one that provided four sets of additional coils. These coils provided both the right and left stereo-tilt and the vertical-overlap alignment deflections in addition to the original coils already installed for scanning the imaging raster.2 As our system meant that the beam would transit the objective lens off axis, it also required that the final aperture be moved from its original position below the scan coils to one above it, and also that it be provided with a precise centration mechanism. The stereo-tilt coils were arranged as upper and lower X-Y deflectors. The upper X coils deflected the raster off axis, either to the left or the right, and then the lower X coils deflected it back towards the axis again so that both left and right image rasters converged at the focus plane on the sample. The upper and lower Y coils were used for vertical alignment of the two images: Effectively, they insured that the right-view raster scanned the same area on the specimen as the left-view raster. The raster-scan coils operated as normal. A control box permitted the choice of no deflection, hold left, hold right, or alternate left and right views. This last mode could be made to switch the stereo tilt on alternate lines, alternate interlace fields or alternate full-frames (see Fig. 7.5 (see color plate 7)). Control knobs shifted the two images in the horizontal (parallax) and vertical directions for alignment. The parallax adjustment shifted with working
2
Chatfield et al. [1974] placed their stereo deflecting device between the final pole-piece and the sample, a choice that greatly reduced the open space above the specimen. The three commercial video-rate stereo SEM systems available in 1975 are reviewed in Pawley [1975].
202
A. Boyde
distance (see Fig. 7.6). The destination of the video signal was switched to the right or left display in synchronism with the beam deflection. The ”black box” used to write the video information to tape could also read it back so that the left and right channel video information could be properly decoded. The first display arrangement used two monochrome video monitors and a mirror stereoscope, so that the left and right eye views were displayed on the corresponding monitors: Only the operator/observer could see the three-dimensional image (see Fig. 7.7 (see color plate 8)). A later, and more satisfactory, arrangement used a professional color video monitor with the left-eye video signal sent to the red channel, and the right-eye signal sent either to the green or to both green and blue channels. The resulting anaglyph image was viewed through red and green (blue-green) filter spectacles. Figure 7.8 (see color plate 9) shows all the components needed to produce, observe and record live-time stereo SEM movies and Fig. 7.9 shows the relation between the stereo-tilt angle, beam kV, and system tilt setting. Although this anaglyph system allowed a number of people could see the live three-dimensional image at the same time, there were still extreme difficulties in giving public demonstrations. Our Rank-Cintel professional color video monitor was a two-man lift, as was the Ampex recorder. This problem was only overcome by converting the videotapes to 16mm movies (see Fig. 7.10 (see color plate 10)).
X-shift R
G
R G
G
R control setting
working distance G
R
R G
R
G
Fig. 7.6 Diagram showing the effect of the x-alignment control of UCL real-time stereo TV SEM system. This control is used to assure that the left and right rasters overlap as they strike the specimen (top). This adjustment must be changed if the specimen working distance varies (bottom)
7 Viewing Live-Time Experiments in the SEM
203
20
20¡ –4 mm zero –4mm 6.0 mm 11.0 mm 15.0 mm 18.5 mm
10 18 17 16 15
15¡
14
angle, α, degrees
13 12 11 10¡
10 9 8 7 6 5
5¡
4 3 2 1
1.5 2.8
5.1
9.6
14.2
19.1
0 0
1
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 18 kV
Fig. 7.9 Calibration curves for the stereo beam tilt angle difference obtained in the alternate-line mode, as a function of working distance from the pole piece and the accelerating potential
Experiments Cutting and Dissecting in the SEM The tools used most productively were diamond ultra-microtome knives, tungsten carbide dental chisels and tungsten microneedles. Diamond ultra-microtome knives were modified by removing the water-retaining ”boat,” which was a part of the commercial product, and then used to study cutting processes in dental enamel, dentine and bone (see Figs. 7.2 and 7.11). Tungsten carbide cutting instruments, here referred to as ”chisels,” were specially fabricated for us from dental bur stock by Beavers Dental Products Corp., Morrisburg, Ontario, Canada, and used for cutting enamel to simulate clinical cutting situations. Dental enamel has a complex anisotropy, and all clinical and experimental approaches were thoroughly investigated. The dynamic three-dimensional videotape recordings were presented at the Association of Teachers of Conservative Dentistry meeting in the United Kingdom in 1972 (Boyde 1973, 1976, 1978, 1990).
204
A. Boyde
Fig. 7.11 Image recorded using slow scan using the same stereo deflection used for dynamic threedimensional viewing and showing the cutting of a human tooth enamel “cavity margin” with a tungsten-carbide “chisel” using the experimental arrangement shown in Fig. 7.3
Microneedles were fabricated by electrolytic etching of tungsten wire: These were employed in microdissection of soft tissue samples, especially: (i) Dissecting freeze-dried or critical-point dried osteoblasts on formative bone surfaces (osteoid) to determine the relationship between the cell elongation and the orientation of the collagen that they had just deposited (see Fig. 7.12: Boyde, Jones & Pawley 1974; Jones, Boyde & Pawley 1975; Jones & Boyde 1976); (ii) Dissecting freeze-fractured and freeze-dried cartilage growth plates to remove one cell at a time to permit energy dispersive x-ray (EDX) elemental analysis of potassium concentrations within these cells (Boyde & Shapiro 1980; Shapiro & Boyde 1984); (iii) Dissecting conventionally-fixed and either freeze-dried or critical-point dried chick embryos (in collaboration with Mary Bancroft and Ruth Bellairs); (iv) Dissecting osmium-thiocarbohydrazide post-fixed mouse embryos to study early developmental stages (see Fig. 7.13: Tamarin & Boyde 1976, 1977); (v) Dissecting a variety of soft tissues fixed only in osmium in a boric acid – borate buffer system and then critical-point dried prior to micro-dissection (e.g., liver, Boyde, 1974; Boyde et al. 1974; Boyde 1975; Fig. 7.14); (vi) Dissecting micro-corrosion casts of kidney vasculature (Fig. 7.15. Specimen prepared by Janice Nowell, University of California, Davis); (vii) Dissecting the rat incisor enamel organ to reveal the organization of secretory ameloblasts making decussating enamel prisms in the inner zone enamel and the distribution of the interameloblastic space compartment and its
7 Viewing Live-Time Experiments in the SEM
205
a.
Fig. 7.12 Osteoblasts on the endocranial surface of neonatal rat calvarium were freeze dried or critical-point dried in situ. Using three-dimensional SEM, individual osteoblasts were dissected off, or bent back, to reveal the orientation of the underlying collagen. This proved to be mostly parallel with axis of elongation of the osteoblasts in the plane of the surface. Field width 85 microns. a) Sequence of three stereo-pairs extracted from a compressed digital video file made from the original analog video data. When viewed in stereo one can easily see the dissecting needle (top center) first touching and then moving an osteoblast. This dislodged cell then charges up and “pithballs” away from the field of view. Its absence can be verified in the bottom stereo image.
206
b.
A. Boyde
c.
54
16 10 3
2
2
0-15 16-30 31-45 46-60 61-75 76-90
Fig. 7.12 b) Slow-scan image taken of the collagen matrix underlying a (now removed) cell. Although made with one stereo tilt still applied, the image was recorded in slow scan to obtain better image quality and show collagen orientation. c) Histogram of alignment with respect to underlying collagen in 15-degree classes for 37 rat (white) and 50 monkey (black bars) osteoblasts
connections with spaces in the papillary layer during maturation (see Fig. 7.16: Boyde & Reith 1977). Barber and Emerson [1978] also exploited tungsten microneedles in threedimensional dissection in the SEM, employing the in-chamber deflection system invented by Chatfield et al. [1974].
Fig. 7.13 Mouse embryo (late 8-day or early 9-day), prepared by the OTO fixation, and critical-point dried, undergoing micro-dissection in the SEM using real-time three-dimensional visualisation
7 Viewing Live-Time Experiments in the SEM
207
Fig. 7.14 a) Ox liver, fixed with OsO4 using boric acid borate buffer, critical point dried, and micro-dissected in the SEM under three-dimensional visualization. b) Slow scan image of nearby area, showing bisected bile canaliculi running along the faces of isolated hepatocytes. Field width 40 microns
Observation of Prepared SEM Samples With Complex three-dimensional Surfaces Direct view three-dimensional SEM proved to be very useful in unravelling a variety of complex three-dimensional systems. For example, we studied the morphology of the developing enamel surface in many primate species, including man, gorilla, chimpanzee, gibbons and old- and new-world monkeys. Here the specimens were prepared by wet dissection of the enamel organ to remove the formative ameloblast cells from the forming enamel surface. By rotating and tilting the sample under
208
A. Boyde
Fig. 7.15 Stereo images of plastic-injected, micro-corrosion casts of kidney vasculature prepared by Janice Nowell, (University of California, Davis). The stereo images were extracted from video sequences originally recorded onto videotape and later converted to compressed video. Upper pair recorded at nominal magnification of x50, lower at x100
three-dimensional observation, we could determine both the direction of the floor of the ameloblastic pit (Tomes’ process) and the orientation of the underlying prismatic unit. We developed a method for determining the depth of the pits caused by etching the dental enamel with different acids. This depended upon visualising the angle at which the wall of the pit obscured vision of its floor, a process that was greatly aided by three-dimensional vision (Boyde, Jones & Reynolds 1978). We also studied capillary blood vessel casts, prepared by Kees Hodde, Amsterdam, of brain, inner ear, enamel organ and bone growth plates in the rat. With David Lim, we studied the guinea pig inner ear, and with Janice Nowell, normal and diseased human lung sample (see Fig. 7.17; Pawley & Nowell 1973). We showed that one could make direct, three-dimensional measurements by using a calibrated stage z-motion control to move any sample with respect to a fixed, “floating-mark cursor” of crossed lines introduced electronically into the red and green video images (Boyde 1975; Holburn & Smith 1979). Thus we could rapidly make precise and accurate measurements of the depths of osteoclastic resorption pits and ameloblastic pits, both items of significance to us in our researches at that time.
7 Viewing Live-Time Experiments in the SEM
209
Fig. 7.16 Osmium perfusion fixed, critical-point dried rat incisor ameloblasts dissected using realtime three-dimensional SEM, showing non-parallel arrangement of alternating transverse rows of secretory ameloblasts, from which it can be inferred that these cells move past each other in rows whilst secreting the enamel which contains decussating (mutually crossing over at 90 degrees) prisms. Field width 40 microns
Observation of Live, Self Moving, Insects and Arthropods Insects and arthropods—protected within their sealed exoskeletons—are normally capable of surviving punishing temperatures and desiccating conditions. We found that the small, red spiders that inhabited the laboratory made convenient experimental objects when their backs were glued to a normal specimen stub. We could observe the movements of limbs, antennae and mouth parts, and later free the experimental subjects who had survived the SEM vacuum conditions. Somewhat later, Ruth Bellairs arrived with a deep interest in a much larger Indian millipede (insect) species. Again we made spectacular stereo movies of these animals, each of which survived and could be revived after SEM study (Fig. 7.18 (see color plate 11)). Most remarkably, we even saw living flea parasites on the limbs of the millipedes (“Little fleas have lesser fleas upon their backs to bite them”).
Freeze Drying in the SEM We also studied the process of ice sublimation from frozen samples in the SEM, using a liquid-nitrogen-cooled, temperature-control stage. The detail was so good that, during the drying process, we could visualize the formation of projecting peaks
210
Fig. 7.17
A. Boyde
7 Viewing Live-Time Experiments in the SEM
211
of ice outside the cells being studied, and then see these small pieces flake off completely (see Fig. 7.19 a-c). We could also examine the process of “super drying,” a process in which the firmly-bound “structural water” is lost after the freeze sublimation process is complete and the specimen is substantially heated (to 200 ◦ C) (see Fig. 7.19).
Temporal Resolution To measure the time resolution of our system, we imaged the workings of a fine ladies’ watch at 1 kX to make it look like Big Ben.
Discussion Our system was unique in its time, and our patent (Boyde et al. 1972) was infringed by several manufacturers when they later realized its worth (and also realized that we could not afford to defend it!). As such systems have now disappeared from the market place, why should we have hopes of a renaissance of real-time threedimensional SEM today? On the optimistic side, we can name several important evolutionary changes:• Perhaps the most important single point is that we are now able to display almost any computer output data at meetings or on the web. • The field-emission guns now common on SEMs allow one to obtain a large enough current for low-kV video-rate imaging without incurring any major loss in resolution (see also Chapter 2). • SEMs now are glorified computers, a factor that simplifies the process of properly recording, sorting and displaying the appropriate views to the proper eyes. • With motorized stages, we can program stereo fly-through views of real threedimensional objects. • Now that we are reaching the post-genomics era, the understanding of structure, and of whole-systems biology, has once again become at least partly respectable: back to structure!
Fig. 7.17 (Continued) Normal and emphysematous lung tissue, prepared by Janice Nowell, (University of California, Davis). On can easily see the tremendous loss of gas-exchange surface characteristic of the diseased state. The stereo images were extracted from video sequences originally recorded onto videotape and later converted to compressed video. Upper two pairs, normal lung; lower two pairs, emphysematous lung tissue. In each set, the upper member was recorded at 50x, and the lower member at 500x, nominal magnification
212
A. Boyde
Fig. 7.19 (a,b,c) Mono-stills showing stages in the freeze drying of K4 cells grown on stainless steel support and held at a temperature control stage at –100 ◦ C. d) A small part of the field after the specimen had been heated to 200 ◦ C to insure that all bound water was removed
• Beam-defining apertures are nowadays placed well above the final lens, and gunalignment coils can be used to impart stereo tilts (See also end of Chapters). • ESEMs today are more suited to wet and dirty work, and their small field of view and limited working distance suffices for most three-dimensional SEM.
Therefore, there is a very good chance that LV three-dimensional environmental SEM may be ready for rebirth, heralding a return of the practice of using the SEM to observe experiments in biology. Acknowledgment The conception of the Stereoscan as an experimental environment in which so many things could be changed at will was a pure marvel! I thank ADG (Gary) Stewart for this remarkable foresight. Particular thanks are due to many colleagues who contributed to the studies described here. Peter Howell, Jim Pawley, Denis Cook and Jay Morgan helped with the concept and the construction and maintenance of the equipment. Dalibor Pospíil recently gave immense help and encouragement as well as providing the means for rescuing the archival video recordings and transcribing them into modern computer format (Boyde & Pospíil 2005). Jim Pawley also helped greatly by encouraging me to write this chapter and then by editing it.
7 Viewing Live-Time Experiments in the SEM
213
References Barber VC, Emerson CJ (1978) Techniques ultilizing real time stereo scanning electron microscopy in the microdissection of biological tissues. J Microsc 115:119–125. Boyde A (1970a) Practical methods, application and problems of 3-D analysis of scanning electron microscopic images. Scanning Electron Microscopy 1970, 105–112; IITRI Chicago. Boyde A (1970b) Calibration of the specimen stage of the Stereoscan. Beitr. Elektronenmikroskop. Direktabb. Oberfl. (Munster) 3:403:410. Boyde A (1971a) Recording anaglyph stereopairs in the scanning electron microscope and some other uses of colour images in the scanning electron microscope. Beitr. Elektronenmikr. Direktabb. Oberfl. (Munster), 4:443–452. Boyde A (1971b) Direct recording of stereoscopic images with the scanning electron microscope by the anaglyph colour technique. Medical and Biological Illustration 21:130–133. Boyde A (1973) Dynamic TV and real-time stereo TV studies of cutting enamel dentine and bone in the scanning electron microscope. Proc Conf on SEM: Systems and Applications, Newcastle, Institute of Physics - Bristol, pp15. Boyde A (1973) Quantitative photogrammetric analysis and qualitative stereoscopic analysis of scanning electron microscope images. J Microsc 98:452–471. Boyde A (1974) Real-time stereo TV speed scanning electron microscopy. Beitr. elektronenmikroskop. Direktabb. Oberfl. (Munster) 7:221–230. Boyde A (1974) Three-dimensional aspects of SEM images. In: Scanning electron microscopy Chap. 11 (OC Wells Ed), McGraw-Hill, New York, pp 277–336. Boyde A (1975) A method for the preparation of cell surfaces hidden within bulk tissue for examination in the scanning electron microscope. Scanning Electron microscopy 1975, 295–304. Boyde A (1975) Measurement of specimen height difference and beam tilt angle in anaglyph realtime stereo TV SEM systems. Scanning Electron Microscopy 1975, 189–198. Boyde A (1976) Enamel structure and cavity margins. Operative Dentistry 1:13–28. Boyde A (1978) Cutting teeth in the scanning electron microscope. Scanning 1:157–165. Boyde A (1990) Physical effects of clinical procedures on the hard dental tissues. In: The Dentition and Dental Care, (RJ Elderton Ed) Heinemann, London, pp325–347. Boyde A (2003) Improved digital SEM of cancellous bone: scanning direction of detection, through focus for in-focus and sample orientation. J Anat. 2003 202:183–194. Boyde A, Pospíil D (2005) Getting ahead with real time three-dimensional SEM. Proceedings of The Anatomical Society of Great Britain and Ireland, Oxford Jan 2005, J Anat 206:494. ISSN 0021–8782 Boyde A, Cook AD, Morgan JE (1972) Scanning electron microscope display method and apparatus. UK Patent No 1393881. Boyde A, Jones SJ, Pawley JB (1974) Some practical applications of real-time TV speed stereo scanning electron microscopy in hard tissue research. Scanning Electron Microscopy (IITRI) 1974, 109–116. Boyde A, Jones SJ, Reynolds PS (1978) Quantitative and qualitative studies of enamel etching with acid and EDTA. Scanning Electron Microscopy 1978, II:991–1002. Boyde A, Jones S.J. and Pawley J.B. Some practical applications of real-time TV-speed stereo SEM in hard tissue research. Scanning Electron Microscopy 1974, III:109–116 (1974). Boyde A, Reith EJ (1977) Scanning electron microscopy of rat maturation ameloblasts. Cell Tiss. Res. 178:221–228. Boyde A, Shapiro IM (1980) Energy dispersive X-ray elemental analysis of isolated epiphyseal growth plate chondrocyte fragments. Histochem. 69:85–94. Chatfield EJ, More J, Nielsen VH (1974) Stereoscopic scanning electron microscopy at TV scan rates. Scanning Electron Microscopy 1974, 117–124. Holburn DM, Smith KCA (1979) On-line topographic analysis in the SEM. Scanning Electron microscopy 1979, II:47–52.
214
A. Boyde
Houghton AH et al. (1971) A new goniometer stage for the Cambridge Stereoscan designed and measured performance. Beitr. Elektronenmikr. Direktabb. Oberfl. (Munster), 4:429–441. Jones SJ, Boyde A (1976) Is there a relationship between osteoblasts and collagen orientation in bone? Israel J Med. Sci. 12:98–107. Jones SJ, Boyde A, Pawley JB (1975) Osteoblasts and collagen orientation. Calcif. Tissue Res. 159:73–80. Pawley JB (1978) Design and performance of presently available TV-rate stereo SEM systems, Scanning Electron Microscopy-1978 I:l57–66 Pawley JB, Boyde A (1975) A robust micromanipulator for the scanning electron microscope. J Microsc 103:265–270. Pawley JB and Nowell JA (1973) Microdissection of biological SEM samples for further study in the TEM, Scanning Electron Microscopy 1973 III:333–340 Shapiro IM, Boyde A (1984) Microdissection-elemental analysis of the mineralising growth cartilage of the normal and rachitic chick. Metab. Bone. Dis. Rel. Res. 5:317–326. Tamarin A, Boyde A (1976) Three-dimensional anatomy of the 8-day mouse conceptus: a study by scanning electron microscopy. J Embryol. Exp. Morp. 36:575–596. Tamarin A, Boyde A (1977) Facial and visceral arch development in the mouse embryo: a study by scanning electron microscopy. J Anat. 124:563–580. Wells OC (1957) The construction of a scanning electron microscope and its application to the study of fibres. PhD Thesis, University of Cambridge.
Chapter 8
Cryo-SEM of Chemically Fixed Animal Cells Stanley L. Erlandsen
Introduction In the last several decades, many improvements have been made that increase the resolution of the scanning electron microscope (SEM), and commercial instruments are now available that enable resolution in the nanometer range on biological specimens. Paralleling this increase in instrument resolution has been the need to develop better techniques to preserve the physiological structure of cells and tissues. Interpretation of biological structure at high-resolution requires high fidelity of preservation and also requires the avoidance of artifacts created through preparatory procedures. Early SEM studies of cell surfaces were limited by conventional preparative methods, such as chemical dehydration and the use of critical-point drying to prepare specimens for the vacuum of the SEM. Removal of water from the sample under these conditions produced a substantial shrinkage of cellular dimensions (Alan Boyde and Maconnachie 1979) and also a collapse of molecular structure (Edward Kellenberger 1987). In a model system measuring volume changes in mouse embryonic tissue, Alan Boyde and Maconnachie (1979) demonstrated that shrinkage of tissue volume up to ∼30% occurred with ethanol dehydration, and additional shrinkage occurred in the critical-point drying due to changes in the dielectric constants of ethanol and CO2 . In this study, freeze-drying produced the best retention of volume (85%) as compared to critical-point drying (41%) or air drying from a volatile solvent (12%). The key to optimal structure preservation is the manner in which bulk or extracellular water is removed from the specimen, leaving the bound water to maintain structural integrity of macromolecules. Standard dehydration techniques cause bound water to be removed and large surface molecules to undergo molecular collapse. This topic has been discussed in detail by Edward Kellenberger (1987) and others in a treatise on cryo-immobilization (Steinbrecht & Zierold 1987). In the early 1990’s, numerous reports supported the use of controlled freeze-drying (Walther et al. 1989; Kusamichi et al 1990; Herman & Müller 1991; Wepf et al. 1994). Normal freeze-drying, where the sample is absolutely dried to remove all water, results in loss of molecular structure, whereas, if water is removed by controlled freeze-drying under high vacuum conditions (usually <1×10−6 Torr) and temperature (colder than −85 ◦ C), the vapor pressure of water in ice will permit
H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
215
216
S. L. Erlandsen
its sublimation, and the subsequent removal of extracellular ice results in exposure of the original surface of the specimen (cells, macromolecules). When controlled freeze-drying is used together with high-resolution cryo-coating of heavy metals (Cr, Pt, W, W/Ir) and cryo-FESEM, it has been possible to demonstrate resolution in the range of 2–3 nm on semi-periodic test samples (Ya Chen et al. 1995 a,b; Victoria Centonze & Chen 1997; Chen et al. 1997; Hermann and Müller 1991; Wepf et al. 1994).
Prefixation Before Cryo-immobilization for Cryo-SEM? Martin Müller (1990) pointed out that for dynamic processes to be arrested in a life-like state, specimens should be cryo-immobilized rapidly by plunge freezing or by slower freezing methods, such as high pressure freezing, where the physical properties of water are altered. Cryo-immobilization methods allow specimen immobilization in milliseconds, rather than the seconds to minutes required for chemical fixation. If specimens are to be examined by transmission electron microscopy (TEM), then the processing of specimens cryo-immobilized in complex buffers/fixatives by freeze-substitution in acetone, methanol or other solvents would remove these buffer salts or other macromolecules prior to embedding and sectioning. However, for cryo-SEM study, the cells to be studied must be able to withstand rinsing in distilled water prior to freezing. Controlled freeze-drying is used to sublimate away extracellular ice and expose the extracellular cell surface or macromolecule. To accomplish this, the specimen must be fixed enough to resist rinsing with distilled water without producing new artifacts. If soluble salts are not rinsed away with distilled water, they will form a contaminant particulate layer on the exposed surface when the ice layer (water) is removed by sublimation. While the use of a prefixation step can assist in stabilization of cell structure for morphological or immunological studies, the slow rate of chemical fixation compared to rapid cryo-immobilization precludes its use for studying cellular kinetic processes in cells that are completed in milliseconds, or in the accurate immobilization of soluble diffusible substances. Because macromolecules on the extracellular surface of cells mediate cellular interactions through signaling or adhesion, methods using chemical fixation have been developed to investigate their structure and distribution by cryo-SEM. Stabilization of cells with fixatives containing low concentrations of aldehydes has been used to investigate cell and molecular structure and to localize cell adhesion molecules on the cell surface (Stanley Erlandsen et al 2001).
Can Conventional Fixation Preserve Surface Topography of Isolated Macromolecules? Effect of Fixation and Critical-point Drying A familiar test specimen in studies on cryo-immobilization has been the polyheads of T4 phage (Hermann and Müller 1991), and the S layer of a number of bacteria (Wepf et al. 1994) because these samples form semi-periodic structures that have
8 Cryo-SEM of Chemically Fixed Animal Cells
217
Fig. 8.1 Cryo-TEM of polyheads of T4. A. Polyheads fixed conventionally and critical-point dried show the overall shape of the polyhead, but no detail of the polygonal subunits is visible. B. Polyheads cryo-immobilized and partially freeze-dried. Subunits in polyhead are easily discernible. (from Hermann and Müller 1991) Bar = 100 nm
been well-characterized and studied by TEM, SEM, and scanning probe microscopy (SPM). Optimum structure information on the capsomers of the T4 phage polyheads (∼8.3 nm overall diameter composed of 6 subunits of 3 nm diameter, and a centerto-center capsomer spacing of 13 nm) was obtained by freeze-drying and metal shadowing at low temperature, followed by cryo-TEM at 30 keV (see Fig. 8.1). If the polyhead sample was freeze-substituted in acetone containing 2% osmium, and freeze-dried from acetone, then the substructure of the capsomers in the polyhead could be easily identified. However, if the polyhead sample was chemically fixed, dehydrated in ethanol and then critical-point dried, there was complete loss of polyhead substructure (see Fig 8.1A). The authors did note, although it was not illustrated, that glutaraldehyde fixation itself did not seem to destroy polyhead substructure as it was visible after air-drying of aldehyde fixed samples, but not in unfixed air-dried samples or fixed CPD specimens.
Stabilization of Macromolecular Structure with Aldehyde Fixation Chemical fixation with low concentrations (0.1–1.0%) of glutaraldehyde can be used to stabilize macromolecules prior to cryo-immobilization. Ya Chen et al. (1995 a,b) and Victoria Centonze (1997) used chemical fixation of macromolecules to facilitate the examination of the molecular structure of actin, fibronectin and individual reovirus particles. These macromolecular samples were adsorbed onto glow- discharged grids and then fixed for 10–15 minutes with 0.1% glutaraldehyde,
218
S. L. Erlandsen
Fig. 8.2 Cryo-SEM of actin filament reconstituted in vitro and fixed in dilute glutaraldehyde. Specimen was coated with chromium and image was collected in secondary electron mode. Observe helical twist of two polypeptide chains in filament and 5 nm subunits. (Courtesy of Dr. Ya Chen) Bar = 60 nm
and in some cases immunolabeled with colloidal-gold before fixing again, but with 1% glutaraldehyde. The aldehyde fixation stabilized these macromolecules before they were rinsed with distilled water to wash away buffer salts prior to cryimmobilization. This distilled water rinse permitted sublimation of extracellular ice to expose the surface-bound macromolecules before cryo-coating with metal and cryo-SEM (Fig. 8.2). Despite the use of fixation, the biological resolution obtained by cryo-SEM in these studies was on the order of 2–3 nm.
Chemical Fixation of Cells Prior to Cryo-Immobilization Fixation and Cryo-immobilization of Yeast In several yeast species, the extracellular cell wall is covered with a dense layer of brush-like fibrillar material that extends outward about 100 nm (Tokunaga et al. 1986; Walther et al. 1988; Osumi et al. 1988). The cell coat of fibrils is not preserved by conventional chemical fixation and dehydration; however, it is easily identified by cryo-immobilization followed by cryocoating and cryo-SEM (Fig. 8.3). Initially, Walther et al. (1988) assumed that conventional dehydration of the yeast cells extracted the fibrilllar material, giving the extracellular surface a smooth homogenous surface. Kusamichi et al (1990) performed an interesting experiment on the stability of this fibrillar layer and demonstrated that detection of the macromolecular structure was dependent on how water was removed from the specimen. In Figures 8.3A and 8.3B, the surface morphology of conventionally fixed/dehydrated and critical-point dried yeast is compared to the surface structure of yeast cryo-immobilized by freezing in liquid propane cooled by liquid nitrogen. Conventionally fixed/dehydrated and critical-point dried yeast lacked
8 Cryo-SEM of Chemically Fixed Animal Cells
219
Fig. 8.3 Comparison by LVSEM of the extracellular surface of Candida albicans prepared by conventional glutaraldehyde-osmium fixation and critical-point drying (panel A) as compared to cryo-immobilization (Panel B). Except for presence of budding scars, the surface of yeast cells in A appear smooth, whereas in Panel B, cryofixed cells can be seen to possess a cell wall composed of closely packed fibrils. Panel C shows a higher magnification of cell wall of critical-point dried cells lacking fibrillar appearance. If samples of conventionally-fixed and critical-point dried cells are rehydrated in water, then cryo-immobilized, the fibrillar component is preserved and can be easily seen. Bar = 1 µm for A & B and 0.5 µm for C & D. (with permission from Kusamichi et al, 1990)
220
S. L. Erlandsen
A
B
Fig. 8.4 Images of Candida albicans fixed in glutaraldehyde (A) or with glutaraldehyde-osmium (B) and then cryo-immobilized for cryo-SEM. Note easily resolved fibrillar appearance of cell wall indicating that conventional fixation is compatible with preservation of fibrils. Bar = 0.5 µm. (with permission from Kusamichi et al, 1990)
8 Cryo-SEM of Chemically Fixed Animal Cells
221
fibrillar material, although bud scars were easily discerned. On the other hand the cryo-immobilized samples examined by cryo-SEM clearly show fibrillar material on the extracellular surface. These authors demonstrated that the fibrillar coat could be detected by cryo-SEM after cryo-immobilization from water, 30% ethanol, 50% ethanol, but not after exposure to greater than 80% ethanol. In addition, cryo-immobilization of yeast fixed with glutaraldehyde alone or with glutaraldehyde followed with postfixation in osmium, as seen in Figs. 8.4A and 8.4B, revealed distinct fibrils on the extracellular surface. These results point out that chemical fixation of cells with either glutaraldehyde, or both glutaraldehyde and osmium, does not by itself produce artifactual collapse of cell surface macromolecules. Another important contribution by Kusamichi et al (1990) was the demonstration that if critical-point dried yeast cells (see Fig. 8.3C) seen to lack the fibrillar cell coat were re-immersed in water and then cryo-immobilized, the fibrillar cell coat was again distinctly visible (see Fig. 8.3D). In other words, the use of critical-point drying on yeast caused molecular collapse of these fibrils, but not removal of the fibrillar coat, and rehydration of the critical-point dried yeast followed by cryo-immobilization allowed the macromolecular structure of the fibrils to be re-established. These experiments strongly support the theory of Edward Kellenberger (1987) involving how water is removed from biological specimens, a process of critical importance in maintaining macromolecular structure. These experiments also point out that macromolecular structure is not always destroyed by chemical fixation and can be often be re-established, if necessary, by rehydration.
Chemical Fixation of Human Platelets Cryomethods together with cryo-SEM can be used to detect individual extracellular macromolecules on the surface of bulk samples. A model system was developed by Stanley Erlandsen et al (2001) in which three molecules (displaying unique topographical shapes in the glycocalyx on the extracellular surface of human platelets) were immunolabeled with colloidal-gold markers, then visualized with high-resolution in-lens cryo-SEM. Illustrated in Fig. 8.5 (see color plate 12) are the three cell-adhesion molecules in the glycocalyx that were selected for detection, including the rod-shaped P-selectin (CD62P) molecule, the GPI-IX complex (CD42a/CD42bα,bβ) consisting of a linear array of polypeptide chains, and the knob-like shaped integrin, GPIIbIIIa(CD41/CD61). Human platelets were spread on sapphire discs and spread cells were stabilized with 0.01% glutaraldehyde prior to immunolabeling with 10 nm colloidal-gold and post-fixation with 3% glutaraldehyde, and in some cases, a second post-fixation in 2% osmium (data not shown). After stabilization by fixation, platelet samples were then rinsed in distilled water before plunge freezing into propane chilled with liquid nitrogen. After partial freezedrying at –85 ◦ C, the double-coating method of Walther et al. (1995) was used and samples were cryocoated by evaporation of 2 nm of TaW at 45◦ by electronbeam deposition followed by 7–10 nm of carbon at 90◦ (See also Chapter 10).
222
S. L. Erlandsen
In-lens cryo-SEM was carried out at 10 keV and the TaW coating was visualized by high-resolution BSE imaging using an Autrata-modified YAG detector. The double coating method of Walther et al. (1995) was selected since the use of high keV with BSE imaging results in higher resolution than secondary imaging due to the reduction in the size of the electron probe and an apparent reduction in contamination (high energy of BSE makes contamination transparent); however, it should be noted that this technique requires a high quality BSE detector, such as the Autrata-modified YAG detector for visualization of colloidal-gold markers at high magnifications. Spread platelets appeared as flattened cells on the sapphire disc (see Fig. 8.6A) and at high magnification the individual cell adhesion molecules could be clearly defined and also measured using the STERECON method described by Michael Marko et al. (1996). Rod-shaped P-selectin molecules appear as dimers on the extracellular surface, and 10nm colloidal-gold particles could be seen projecting from the amino terminal lectin domain (Fig. 8.6B). Stereo-pairs clearly demonstrated the projection of the P-selectin dimers, ∼48 nm above the plasmalemma surface, which is similar to the length of this molecule as reported by other methods (Ushiyama et al. 1994). The GPIbα component of the GPI-IX complex (consisting of 5 polypeptide chains) was seen projecting ∼37 nm above the cell surface (reported values of about 43 nm for extracellular domain [Fox et al. 1988; Kieffer & Phillips 1990; Lopez, 1994]) as a linear array with 10 nm colloidal-gold particles attached to the amino terminus of the two GPIbα chains (compare Fig. 8.6C). The integrin, GPIIbIIIa, was seen as knob-like protrusions on the cell surface ranging from ∼17–22 nm which compared favorably with reported values of 20–22 nm (Kieffer & Phillips 1990; see Fig. 8.6D). The use of dilute glutaraldehyde for stabilizing spread human platelets facilitated cell stabilization and immunolabeling of the extracellular surface prior to cryo-immobilization and permitted identification of three distinctly-shaped cell adhesion molecules in the glycocalyx. Not only did prefixation of platelets enable the immunolocalization of these cell adhesion molecules prior to cryo-immobilization, but values for cell adhesion molecule height in this study were remarkably similar to those obtained by other methods (Erlandsen et al. 2001). These results demonstrate that careful use of prefixation can be used to investigate immunological identification, distribution and dimensional topography of macromolecules by cryo-SEM at high-resolution.
Chemical Fixation of Microorganisms Chemical fixation with glutaraldehyde has been used successfully to prefix microorganisms prior to cryo-immobilization in either liquid nitrogen slush at −210 ◦ C or by high pressure freezing (Erlandsen et al. 2003). The prefixation step again facilitated rinsing of prokaryotic cells (the bacteria Proteus mirabilus and Enterococcus faecalis) and the eukaryotic protozoan, Giardia lamblia. Cryosamples were sublimated at −90 ◦ C to remove extracellular ice and then cryocoated with either PtTa
8 Cryo-SEM of Chemically Fixed Animal Cells
223
Fig. 8.6 Cryo-SEM of spread human platelets lightly fixed in glutaraldehyde and immunolabeled for cell adhesion molecules before cryo-immobilization. A and B show spread platelet immunolabeled for P-selectin. A shows entire platelet while B is a high magnification showing dimeric Pselectin molecules projection above the surface of the membrane. Two colloidal-gold markers can be seen on the amino terminus of each P-selectin dimer. C indicates the linear array of molecules (GPIb has gold label) in the GPI-IX complex and D reveals the gold labeling of individual integrin molecules (GPIIbIIIa). Bar = I µm for A, 130 nm for B, and 100 nm for C and D. (with permission from Stanley Erlandsen et al 2001)
or Pt followed by carbon using the double coating method of Walther et al. (1995). High-resolution imaging of P. mirabilus revealed a sponge-like glycocalyx or capsule, and the presence of periodic structure in the bacterial flagella, presumably the flagellin subunits (see Fig. 8.7).
224
S. L. Erlandsen
Fig. 8.7 Cryo-SEM of glutaraldehyde-osmium fixed Proteus mirabilus. Colloidal-gold labeling for surface antigens was detected using backscatter electron imaging with a YAG detector. Glycocalyx of microbe can be seen as a meshwork of filaments, and several bacterial flagella are also apparent. Bar = 200 nm. (from Erlandsen et al. 2003)
Potential for Investigating Extra- and Intracellular Structure in Chemical Fixed Cryo-samples Cryo-SEM has been successfully applied to investigate isolated macromolecules as well as the extracellular surfaces of many different cell types. Recently two methods have been reported that may extend cryo-SEM studies by permitting direct observation of intracellular structure. Cryotomy of cryo-immobilized samples has been used in examination of both plant and animal cells (Nusse & van Aelst 1999; Paul Walther & Martin Müller 1999). Cryosectioning is accomplished using diamond knives at low temperature, but rather than examining the surface of the cryosection, which often shows knife marks or compression, this technique entails examining the surface of the tissue block by cryo-SEM. Nusse & van Aelst (1999) reviewed the application of cryoplanning of botanical samples and described the advantages, compared to freeze-fracture, for obtaining flat surfaces for morphometric studies or x-ray analysis. Walther and Müller (1999) obtained high-resolution structural information from high-pressure frozen animal and plant cells after ultracryotomy with a diamond knife. The best results were obtained with adequately frozen microcrystalline samples, and intracellular structures were observed, including membrane systems and cytoskeletal components, i.e., microtubules. Although the authors
8 Cryo-SEM of Chemically Fixed Animal Cells
225
Fig. 8.8 Images of yeast and bacteria prepared in a cryo-dual-beam FIB/SEM. A is ion-milled surface of yeast pellet showing several yeast cells. B is the surface of yeast membrane revealing the presence of 10 nm particles despite imaging at 2 keV. C is a cross-section of the microbe Bactobacillus, and examination of ion-milled surface reveals presence of mesosomes and tight packing of ribosomes (not shown). A = 1 µm: B = 200 nm; C = 250 nm. (with permission from Mulders 2003; G.I.T. Imaging & Microscopy 2:8-10)
226
S. L. Erlandsen
worked with unfixed cryo-immobilized samples, it seems reasonable to speculate that chemical prefixation by perfusion (to stabilize complex or difficult-to-access animal tissues such as brain or intestinal tract) could be used to permit dissection of selected areas that could then be cryo-immobilized by high-pressure freezing or other freezing methods to obtain microcrystalline samples for cryosectioning. A second method for obtaining intracellular structural information in cryosamples is the use of a cryo-dual-beam instrument incorporating both focusing electron (SEM) and focusing ion beam (FIB) columns. Mulders (2003) recently reported the use of cryo-SEM/FIB on biological samples, including yeast, bacteria, and gut epithelial cells. Intracellular structure detected included cell membranes and nuclei in yeast (Fig. 8.8A); however, the magnifications published were not sufficient to evaluate the presence of the fimbria on the extracellular surface of the yeast cell wall. Higher resolution SEM of the yeast membrane revealed 10 nm particles in membrane surfaces exposed by ion-milling (see Fig. 8.8B), despite the fact that the in-lens field emission SEM was operated at 2 keV, a voltage where beam diameter may exceed 2 nm, and hence provide less resolution than would be possible using higher keV, backscatter electron imaging in the doublecoating method (Walther et al. 1995). Nonetheless, the potential of dual-beam FIB/SEM can be seen in Fig. 8.8C, where ion milling of the bacterium Bactobacillus, revealed mesosomal membrane and cytoplasm containing densely packed ribosomes. Theoretically, this bacterium could be serial cryosectioned. The use of cryo-FIB permits a micro-surgical approach for investigating both surface and intracellular structure. At its optimum, it may eventually permit three-dimensional or tomographic structural analysis by cryo-FIB/SEM, although to date, very few cryo-FIB/SEM studies on chemically-fixed cells or tissue exist due to limited access to appropriate instrumentation.
Summary Chemical fixation of cells provides an opportunity to work with osmotically sensitive cells and obtain results that cannot be achieved any other way. The most important artifact to control is chemical dehydration, which is now known to be the major cause of molecular collapse. Using model systems such as yeast, it has been demonstrated that the removal of bound water during dehydration in 80–90 % ethanol produces collapse, but rehydration of samples in water followed by cryo-immobilization can restore the surface fimbrae of yeast back to their normal appearance. Prefixation of cells prior to freezing permits immunolocalization of surface molecules and correlation with surface topography. CryoSEM of chemically fixed cells and tissues should continue to provide exciting results in the future, and when combined with high-resolution light microscopic methods (live cell imaging, confocal or 2-photon microscopy), or with cryo-dualbeam FIB/SEM, should yield more information on extracellular and intracellular structure.
8 Cryo-SEM of Chemically Fixed Animal Cells
227
References Books Echlin P (1992) Low temperature microscopy and analysis. Plenum, New York Severs NJ, Shoton DM (eds) (1995) Rapid freezing, freeze fracture, and deep etching. Wiley-Liss, New York Steinbrecht RA, Zierold K (eds) (1987) Cryotechniques in biological electron microscopy. Springer-Verlag, New York Articles Boyde A, Maconnachie A (1979) Volume changes during preparation of mouse embryonic tissue for scanning electron microscopy. Scanning 2:149–163 Centonze VE, Chen Y (1995) Visualization of individual reovirus particles by low-temperature high-resolution scanning electron microscopy. J Struct Biol 115:215–225 Chen Y et al (1995) Imaging of cytoskeletal elements by low-temperature high-resolution scanning electron microscopy. J Micros 179:67–76 Chen Y et al (1997) High-resolution cryo-scanning electron microscopy study of the macromolecular structure of fibronectin fibrils. Scanning 19:349–355 Erlandsen S et al (2003) High-resolution cryo-FESEM of microbial surfaces. Micros Microanal 9:1–6 Erlandsen SL et al (2001) High-resolution cryo-FESEM and detection of individual cell adhesion molecules by stereo imaging in the glycocalyx of human platelets: Immunogold localization of P-selectin (CD62P), integrin GpIIb/IIIa (CD41/CD61), and GpI-IX (CD42a,b). J. Histochem Cytochem 49: 809–819 Erlandsen SL et al (2001) Cryo field emission SEM (FESEM). BioTechniques 31:300–305 Fox JEB et al (1988) Structure of the glycoprotein 1b-IX complex from platelet membranes. J Biol Chem 263:4882–4890 Hermann R, Müller M (1991) Prerequisites of high-resolution scanning electron microscopy. Scanning Microscopy 5:653–664 Kellenberger E (1987) The response of biological macromolecules and supramolecular structure in the physics of specimen cryopreparation. In: Steinbrecht RA, Zierold K (eds) Cryotechniques in biological electron microscopy. Springer, Berlin, pp. 35–63 Kieffer N, Phillips DR (1990) Platelet membrane glycoprotein:functions in cellular interactions. Annu Res Cell Biol :329–357 Kusamichi M et al (1990) Influence of surrounding media on preservation of cell ultrastructure of Candida albicans revealed by low temperature scanning electron microscopy. J Electron Micrsc 39:477–486 Lopez JA (1994) The platelet glycoprotein 1b-IX complex. Blood Coagul Fibrinol 5: 97–119 Marko M, Leith A (1996) STERECON-three dimensional reconstructions from stereoscope contouring. J Struct Biol 116:93–98 Mulders H (2003) The use of a SEM/FIB dual beam applied to biological samples. GIT Imag Microsc 2:8–10 Müller M (1992) The integrating power of cryofixation-based electron microscopy in biology. Acta Microscopia 1:37–44 Nusse J, Van Aelsi AC (1999) Cryo-planing for cryo-scanning electron microscopy. Scanning 21:372–378 Osumi M et al (1988) High-resolution low voltage scanning electron microscopy of uncoated yeast cells fixed by the freeze-substitution method. J Electron Microsc 37:17–30 Pawley JB, SL Erlandsen (1989) The case for low voltage high-resolution scanning electron microscopy of biological samples. In: Albrecht RM, Orenberg RM (eds) (1988) The science of biological specimen preparation for microscopy and microanalysis. Scanning Microscopy International, Chicago, pp 163–173
228
S. L. Erlandsen
Steinbrecht RA, Zierold K (eds) (1987) Cryotechniques in biological electron microscopy. Springer, Berlin Tokunaga M et al (1986) Ultrastructure of outermost layer of cell wall of Candida albicans observed in rapid-freezing technique. J Electron Microsc 35:237–246 Ushiyama S et al (1994) Structural and functional characterization of monomeric soluble P-selectin and comparison with membrane P-selectin. J Biol Chem 268:15229–15237 Walther P et al (1988) Morphological organization of glycoprotein containing cell surface structures in yeast. J Ultrastruct Mol Struct Res 101:123–136 Walther P et al (1989) Imaging of intramembranous particles in frozen-hydrated cells (Saccharomyces cerevisiae) by high-resolution cryo SEM. Scanning 12:300–307 Walther P, M Müller (1997) Double-layer coating for field emission cryo-scanning electron microscopy—present state and applications. Scanning 19:343–348 Walther P, Müller M (1999) Biological ultrastructure as revealed by high-resolution cryo-SEM of blockfaces after cryo-sectioning. J Microsc 196:279–287 Walther P et al (1995) Double layer coating for high-resolution low temperature SEM. J Microsc 179: 229–237 Walther P, Müller M (1999) Biological ultrastructure as revealed by high-resolution cryo-SEM of blockfaces after cryo-sectioning. J Microsc 196(3):279–287 Wepf R (1994) High-resolution SEM of biological macromolecular complexes. In Bailey GW, Garrett-Read AJ (eds) Proceedings of the 82nd meeting of EMSA. San Francisco Press, San Francisco, pp 1026–1027
Chapter 9
High-Resolution and Low-Voltage SEM of Plant Cells Guy Cox, Peter Vesk, Teresa Dibbayawan, Tobias I. Baskin, and Maret Vesk
Introduction: The Plant Cell Plants share most of their biochemical pathways and organelles with animal cells. The most obvious difference is the presence in plants of the photosynthetic carbon fixation pathway and its associated organelle the chloroplast. Chloroplasts, carbon fixation and oxygen evolution are found only in leaves and other green photosynthetic tissues, but chloroplast derivatives, plastids, are found fulfilling other functions (storage, pigmentation, geoperception) in most other plant tissues. Otherwise plant intracellular organelles and cytoskeletal components—mitochondria, Golgi, lysosomes, endoplasmic reticulum, ribosomes, microtubules and actin microfilaments—are mostly recognizable and familiar to animal-cell biologists. There is, however, a fundamental difference in the way the plant cell is constructed and organized. Plant cells are bounded by a rigid polysaccharide wall, typically with a simple polygonal shape. These cells are often very large, up to millimeters in length in stems and roots and hundreds of micrometers in other tissues. Most of the content of these huge cells is a large vacuole, with the cytoplasm in fully mature cells forming a thin layer often no more than a few nanometers thick inside the wall. The dividing cells in developing tissues and organs are much smaller and are filled with cytoplasm; it is in the process of differentiation that the cells expand and elongate. The vacuoles are much more than empty space. While in woody tissues the mechanical strength of the cell walls is the main source of structural strength and rigidity for the organ, the majority of plant tissues rely on turgor pressure of the vacuole sap maintaining the cell wall in tension to achieve rigidity. This in turn means that cell-to-cell connections (plasmodesmata) are quite different from the structures fulfilling similar functions in animal cells. The regularity and large size of plant cells made them much easier for early microscopists to recognize than animal cells, and the early development of cell theory was very largely based on observations from plants. The term “cell” was coined by Robert Hooke in 1665 to describe the structure he saw in cork (bark, a dead tissue) but cells were also portrayed clearly in his image of a living nettle leaf (Hooke 1665). The nucleus was also discovered (by Robert Brown) in plant tissue—orchid leaf—some 150 years later (Brown 1831). A few years later Matthias Schleiden [1838] developed the first formal statement of a unified theory of growth,
H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
229
230
Cox et al.
development and organogenesis based on cell theory, again entirely in the plant kingdom. Only then was the concept extended to animal cells by Schwann [1839]. These structural features, which were so advantageous to the early microscopists, create major problems once microscopy reaches into the ultrastructural domain. The tough cell wall is not only a barrier to reagents, it is also difficult to section or fracture and can be very difficult to remove chemically. The vacuole is so large that in a random view of a differentiated cell one is likely to see very little apart from vacuole sap (in a section or frozen fracture) or vast expanses of tonoplast (the bounding membrane of the vacuole) in whole cell preparations. The vacuole of a fully differentiated plant cell is so large and low in solute content that no freezing technique currently available can vitrify it, so ice crystal formation is massive and potentially damaging in any cryopreparation. Almost more significant than any of these considerations is the vital importance of turgor to the structural integrity of the cell. Any change in the cell’s osmotic environment is potentially disruptive, but in particular excessive ionic concentrations external to the cell will remove turgor and lead to plasmolysis, where the cytoplasm strips away from the wall, and its natural arrangement is entirely lost. Fixation prevents these effects (provided they have not already occurred) but also removes much of the mechanical strength of the tissue, so that it is easily damaged in handling. Plant tissues as a whole contain air, not liquid, between the cells. Almost any treatment will involve filling these air spaces with fluid, again impacting on the natural state of the cells. The external surface of most plant tissues is protected by a highly water- and chemical-resistant cuticle, and only dissection will allow fixation or other chemical treatment. However, the cells are so large that many will be damaged in the process. Sub millimeter blocks of tissue are commonly recommended for electron microscope fixation – but since differentiated plant cells can be millimeters in length, such a block would contain no intact cells! Ways of overcoming these problems are varied. Individual components can be isolated from the cell, so that the problem is shifted to the isolation technique. The walls can be removed from some cell types by a combination of enzyme treatment and mechanical disruption, leaving functional, living protoplasts (which must, however, be maintained in a very carefully controlled osmotic environment). Plasmolysis can be used as a tool to expose inner wall layers or the external surface of the plasma membrane. Where we do need to look at intact cells, undifferentiated cells contain small or no vacuoles, and though these cells will only enable us to answer general questions, they do contain a full population of organelles. Where we do need to study differentiated cells, we have to accept that patience is a virtue, since we will often search many fields before finding useful exposures of cytoplasm. The topic of preparing plant cells for field-emission SEM was reviewed extensively by Vesk et al. [2000], and this chapter draws heavily on that review.
Whole Plant Cells and High-resolution SEM (HRSEM) The fundamental consideration in any specimen preparation technique for imaging components of the cytoplasm (organelles or cytoskeleton) at high-resolution in SEM is selective removal of unwanted material. This must include the cytosol in both
9 High-Resolution and Low-Voltage SEM of Plant Cells
231
plant and animal cells, but in plants we also may need to remove the tonoplast, since with the vacuole content removed this will effectively cover the cytoplasm. Two basic specimen preparation protocols have been devised to achieve these goals. Both protocols depend on (i) freeze fracturing to reveal areas of interest and (ii) selective removal of cytosol. These are illustrated in Table 9.1. The fact that two protocols are necessary emphasizes that there is no unique solution. Osmic maceration (left side of the table) extracts cytosol while giving both rigidity and electron density to membranes. This can give excellent images of organelles but cytoskeletal elements are removed or obscured. Detergent permeabilization (right column) removes membranes to a greater or lesser extent, revealing the cytoskeleton, while typically giving poorer imaging of organelles.
Table 9.1 The two basic protocols OSMIUM MACERATION
DETERGENT PERMEABILIZATION
FIX 0.5% paraformaldehyde + 0.5% glutaraldehyde/phosphate buffer pH 7.2, 1 h WASH 2–3 × 5 min same buffer POSTFIX 1% OsO4 /buffer, 1 hWASH dH2 O—several rinses
0.04% saponin in MTS buffer + taxol WASH MTS buffer (+sucrose) FIX 0.5% glutaraldehyde + 0.5% paraformaldehyde/buffer (+ sucrose) Buffer (+ sucrose), 15 min POSTFIX 0.5% buffered OsO4 –min WASH Buffer—15
CRYOPROTECT 25% DMSO, 50% DMSO 30 min each FREEZE & FRACTURE Liquid nitrogen—razor blade & hammer THAW 50% DMSO WASH Buffer—several rinses POSTFIX 1% OsO4 /buffer, 1 h MACERATE 0.1% OsO4 /buffer >48 h POSTFIX 1% OsO4 /buffer, 1 h WASH dH2 0 - several rinses STAIN 2% aq. tannic acid, 1 h WASH dH2 0
CONDUCTIVE STAIN 0.5% aq. OsO4 , 15 min
DEHYDRATE Graded ethanol series CRITICAL-POINT DRY SPUTTER COAT Pt or Cr
232
Cox et al.
Osmic Maceration This approach is based on the protocols of Tanaka and Fukudome [1991] and Barnes [1992]. Bulk tissue is cut into pieces ca. 0.5 mm x 5 mm in buffer and transferred into freshly prepared fixative; for marine algae, sucrose is added to the initial fixative to give a concentration of 0.6 M to preserve correct osmolarity. For unicellular algae, the fixative can be added to the cell cultures and the cells centrifuged to form a loose pellet, which is then processed in the centrifuge tube until freezing. After thorough washing to remove all traces of OsO4 and cryoprotection, the specimens are simply plunge-frozen in liquid nitrogen. To simplify the freezing and fracturing process, a number of tissue slices can be packed together into a gelatin capsule in 50% DMSO as described by Koga et al. [1992]. Fracturing is carried out with a precooled razor blade and hammer on a brass plate cooled with liquid nitrogen, as described in detail by Barnes [1992]. Unicellular algae require more drastic fracturing since the cells are both separate and resistant. The pellet is spread on a piece of aluminium foil, which is folded, plunge-frozen in liquid nitrogen and fractured by repeated hammer blows (Vesk et al. 2000). After thawing, the samples are post-fixed and macerated in repeated changes of buffered osmium tetroxide. It is important that the solutions are changed whenever
Fig. 9.1 One chloroplast in a maize (Zea mays) mesophyll cell, after osmic maceration. Granal and stromal thylakoids are visible, with spherical lipid droplets (plastoglobuli). A strand of ER is seen lower left. JEOL JSM 6000 FESEM, secondary electron detector, at 5 kV. Scale bar = 1 µm
9 High-Resolution and Low-Voltage SEM of Plant Cells
233
discolouration of the OsO4 is observed. Different workers have used maceration times of up to a week, but the primary aldehyde fixation has a substantial effect on the maceration time required. If a “standard” TEM fixation of 2–3 % glutaraldehyde is used, the time needed for sufficient extraction of cytosol is greatly extended, whereas a dilute aldehyde prefixation followed by OsO4 can make extraction easier (Lea et al. 1992). Vesk et al. [2000] found that a relatively weak primary fixative of 0.5% glutaraldehyde + 0.5% paraformaldehyde in 0.05M sodium phosphate buffer reduced the time of maceration needed for plant tissue to 48–72 hours at room temperature. The final stage is provision of a conductive stain by mordanting in 2% tannic acid and final staining in 1% OsO4 . After ethanol dehydration, samples are critical point dried from CO2 . In animal tissues, careful control of maceration time can result in some degree of cytoskeletal preservation (Tanaka and Fukudome 1991), but to the best of our knowledge this has not been attained with plant cells. Preservation of organelles, however, can be excellent. Figure 9.1 presents a relatively low-magnification image, taken at 5 kV, of the chloroplast of a mesophyll cell in maize. The granal and stromal thylakoids are clearly visible, as well as lipid droplets (plastoglobuli). Starch is not formed in these chloroplasts. Figure 9.2 shows a maize bundle sheath chloroplast, which has no grana. This enables a clearer view of the fine details of the thylakoid membranes, which are seen in extensive surface views.
Fig. 9.2 Maize (Zea mays) chloroplast (above), microbodies, ER and tonoplast (bottom) after osmic maceration, at high magnification. JEOL JSM 6000 FESEM, secondary electron detector, at 5 kV. Scale bar = 100 nm
234
Cox et al.
Detergent Extraction The detergent extraction technique (originally consisting of saponin permeabilization followed by a brief conductive staining in OsO4 , quick freezing, fracturing and freeze drying in a freeze-etch apparatus) was developed to image cortical microtubules in onion-root tip cells (Vesk et al. 1994). The method was later modified to include a light aldehyde fixation step after the permeabilization and prior to freezing and fracturing manually, and critical-point rather than freeze drying (Vesk et al. 1996). The later protocol, as shown in Table 9.1, gives fewer problems with charging in the SEM, and also makes it possible to carry out immunogold labelling; it has since become the starting point for schedules for many diverse tissues. Only minor changes (mainly to the saponin permeabilization) were needed to prepare such varied specimens as onion-root tips, Chara internodal cells and the unicellular alga Chlamydomonas (Vesk et al. 2000). Fowke et al. [1999] have used the same basic protocol, with slight modifications, to image, and immunogold label, microtubules in highly vacuolated suspensor cells in white spruce somatic embryos. Figures 9.3 and 9.4 show the interaction between actin microfilament bundles, ER and coated vesicles in a Chara cell. The goal of modifying the detergent extraction to permit better preservation of membranous organelles while still maintaining the excellent retention and exposure
Fig. 9.3 Chara corallina prepared by digitonin extraction, showing cables of actin filaments passing through fenestrae in the ER. Clathrin-coated vesicles, close to the plasma membrane, are visible through the fenestrae, and one is clearly linked by two bridges to the actin bundle (large arrow, and Fig 9.4). Links between actin cables and ER are also frequently seen (small arrow). Imaged on a JEOL JSM 6000 FESEM, secondary electron detector, at 5 kV. Scale bar = 1 µm
9 High-Resolution and Low-Voltage SEM of Plant Cells
235
Fig. 9.4 Higher magnification image of coated (left) and uncoated (right) vesicles “walking” along an actin cable. The left vesicle is also seen in Fig 9.3. Imaged on a JEOL JSM 6000 FESEM, secondary electron detector, at 5 kV. Scale bar = 100 nm
of the cytoskeleton is appealing but has proved elusive. The use of digitonin rather than saponin has shown some success in this regard with animal cells (Temkin et al. 1993). This technique has shown some promise on young spinach leaves, with chloroplasts and microtubules present though the thylakoid membranes of the chloroplasts are somewhat swollen. However, in most cells very much poorer results were obtained, with few fractures through organelles and extensive vesiculation of both the chloroplast and mitochondrial envelopes (Vesk et al. 2000). Poor diffusion of digitonin through plant tissue seems the likely culprit, and there does remain the possibility that further technical development could give improved results.
Disruption and Isolation An alternative approach is to disrupt cells while selectively preserving or extracting the structures of interest. One such approach depends on the fact that many cytoskeletal components are stable in the absence of calcium. When living organs are sectioned in a buffer containing a calcium chelator and buffered to pH 7.0, such as a typical PIPES/magnesium/EGTA buffer, most of the contents of the cell are lost, but microtubules and some plasma membrane remain over the cell walls (Tian et al. 2004). Figure 9.5 shows an example of this. The image shows a wide area of membrane overlaid by several cortical microtubules, which in turn are partially overlaid by membranous elements that are probably endoplasmic reticulum. The full arrow indicates a structure that shows the geometrical lattice of a clathrin-coated
236
Cox et al.
Fig. 9.5 Disrupted hypocotyl of Nicotiana tobacum (tobacco) seedling, sectioned in a microtubule stabilizing buffer. For the left-hand panel, the sample was dehydrated, critical-point dried, sputter coated with platinum (nominal depth of 2 nm) on an EMI-Tech Turbo 575X unit, and imaged. For the right-hand panels, the sample was incubated in a mouse monoclonal anti-α-tubulin serum followed by goat-anti-mouse IgG conjugated to 10 nm gold, post-fixed in glutaraldehyde, dehydrated, critical-point dried, coated with carbon to a nominal depth of 10 nm by discharge-evaporated graphite in an EMI-Tech Turbo 950 unit, and imaged. The left-hand and upper-right panels are SE images taken with the upper detector, the middle-right panel is a BSE image, taken with the pseudo in-lens, YAG detector supplied by the manufacturer. The lower-right panel is a manual overlay of the SE and BSE images. Imaging done on a Hitachi 4700 S FESEM at an accelerating voltage of 5 kV. Scale bars = 500 nm (left) and 150 nm (right)
pit and, hence, is presumably a forming endosome. The arrowhead points to what appears to be a polysome on the surface of the presumptive ER. Many bridge-like points of apparent attachment are visible running between the microtubules and the plasma membrane. Such a preparation is readily combined with immunological techniques. For example, to confirm that the long structures are microtubules, a section was incubated in an antibody against α-tubulin followed by a secondary antibody conjugated with 10-nm gold. The sample was post fixed to retain the antibodies. Note that while post fixation is commonly done with osmium tetroxide, we find that glutaraldehyde is preferable because osmium tetroxide solutions may form precipitates of similar size to the gold particles. After dehydration and critical-point drying, samples were coated with carbon. In general, the carbon coat does not provide as high contrast with the SE detector as does the platinum coat, but long, fibrous elements are clear in the SE image (Fig. 9.5, upper right). In the BSE channel, the gold conjugates are well contrasted, and the overlay indicates abundant α-tubulin label on the long elements, identifying them as microtubules. Chlamydomonas reinhardtii is a unicellular green alga, which has become the organism of choice for experimental plant-cell biology. As a green alga it has a photosynthetic system essentially equivalent to that of higher plants, but it has a
9 High-Resolution and Low-Voltage SEM of Plant Cells
237
small and very well-characterized genome. Many specific mutant strains are available. The Chlamydomonas flagellar apparatus has been shown in TEM studies to be structurally complex and associated with an elaborate microtubular cytoskeletal system (Ringo 1967; Cavalier-Smith 1974). Confocal microscopy shows that there is a further cytoskeletal system of centrin (caltractin) linking the complex to the nucleus (Wright et al. 1985; Cox et al. 1997). One convenient mutant available lacks a cell wall and hence is much more susceptible to lysis than the wild type. The two flagella will adhere readily to poly-l-lysine coated films, and if the cell is then ruptured by mild detergent treatment or osmotic shock, the flagellar system and its associated cytoskeleton (often with the nucleus in place as well) will remain attached to the film. This can be critical-point dried and coated, giving very high-resolution images of structures that are barely discernible in the TEM (see Fig. 9.6) (Dibbayawan & Cox 1999). Since pre-treatment is minimal these samples are also very well suited to immunolabelling, which can help identify the different cytoskeletal components. The three-dimensional nature of these images is extraordinary, and to enable this to be fully appreciated, Figure 9.6 is presented as a stereoscopic pair for viewing with a simple pocket stereoscope. It is also possible to view it without mechanical aids, by diverging the eyes, though this takes some practice. Algae of the family Characeae have extraordinarily large ‘giant’ internodal cells. These are multinucleate and measure centimeters in length and millimeters in diameter. These form the ‘stems’ of the alga and are separated by nodes of normalsized cells where whorls of branches form. In spite of their unusual construction, characean algae are regarded as close to the evolutionary lineage that led to land plants. In Chara the cytoplasm of the giant cells is separated into two zones, one static, containing the chloroplasts, and the other streaming, with the junction delineated by bundles of actin filaments. This has made Chara a popular organism for studying cytoplasmic streaming (Williamson 1993). Cutting the end from a giant internodal cell allows the vacuolar sap to drain away. Gentle pressure with forceps sleeved in Teflon will then extrude the cytoplasm.
Fig. 9.6 Chlamydomonas flagellar base, stereo pair. The doublet microtubules of the flagellum and the complex structures associated with the basal bodies are demonstrated in these micrographs. Scale bar = 100 nm. SE image, 15 kV
238
Cox et al.
Fig. 9.7 Several actin cables associated ith chloroplasts in extruded Chara cytoplasm. Small, unidentified organelles and possible fragments of ER are also present. Bar: 1 µm; JEOL 6000F, 10 kV
Vesk et al. [2000] extruded the cytoplasm in this way into a drop of buffered stabilizing fixative (0.1% glutaraldehyde + 0.5% paraformaldehyde in 4 mM EGTA, 25 mM KCl, 4 mM MgCl2 , 5 mM Tris, 200 mM sucrose, pH 7.6) on a glass or Thermanox plastic coverslip. After a brief (5 minute) incubation, samples were washed in PBS for 10 min, briefly post-fixed in 0.1% OsO4 in PBS (also 5 min), washed, dehydrated and critical-point dried. The main features observed in extruded Chara cytoplasm are rows of chloroplasts attached to actin cables, and these images reveal a close relationship between the actin cables, presumed ER and the chloroplast envelope (see Fig. 9.7). There is an extensive ER network of anastomosing tubes and cisternae with actin bundles weaving through the network, possibly implying that the ER is involved as well as the actin cables in mediating cytoplasmic streaming. The relatively robust actin cables that run along rows of chloroplasts have been imaged before in a conventional SEM by McLean and Juniper [1993] and by Reichelt et al. [1995], but the FESEM images show much higher resolution.
Plasmolysis and Protoplasts A major topic of interest in plant-cell biology is the association between the cell and its polysaccharide wall. To image cell walls with SEM, the plasma membrane needs
9 High-Resolution and Low-Voltage SEM of Plant Cells
239
Fig. 9.8 Inflorescence stem of Arabidopsis thaliana. The higher plant cell wall imaged in surface view with FESEM. The predominantly fibrous texture is evident, as are some encrusting and possibly cross-linking material. Platinum coating and imaging as for Fig. 9.5
to be removed. For samples with large cells, such as cucumber hypocotyls or the inflorescence stems of Arabidopsis thaliana (Fig. 9.8), cutting the organ in a buffer, such as PBS, that does not stabilize the cytoskeleton suffices to remove the majority of cytoplasm and plasma membrane (Marga et al. 2005). The remaining material is then fixed with a mixture of paraformaldehyde and glutaraldehyde, dehydrated, and critical-point dried. Most components of the cell wall are unlikely to be removed by these treatments; therefore, the imaged material reflects the entire cell wall architecture, rather than only cellulose. This can give excellent images of the wall (Baskin 2006), but since the plasma membrane has been lost there is no information preserved about the relationship between the wall and the plasma membrane, its interface with the cell. To visualize both cell-wall and plasma membrane, alternative techniques are needed. Furthermore, in Arabidopsis roots, a popular model system for examining the cell wall, cutting the roots in buffer is not sufficient to remove the plasma membrane, which clings tenaciously to the wall even when all cytoplasm is lost. One way to overcome these problems is to take advantage of the osmotic sensitivity of plant cells. When exposed to a hypertonic medium the protoplast shrinks and separates from the cell wall in plasmolysis. Cables of cytoplasm—Hechtian strands—remain, running between the protoplast and adhesion sites on the cell walls.
240
Cox et al.
Plasmolyzed cells can be embedded in polyethylene glycol (PEG) immediately prior to fixation (Wiedemeier et al. 2002). Plasmolysis pulls the membrane away from the cell wall, allowing it to be imaged at high-resolution. The PEG can be removed after sectioning, and the cells critical-point dried. Figure 9.9 shows a low magnification image of a section prepared in this way. In some cells, the rounded
Fig. 9.9 Root of Arabidopsis thaliana. Sample was dehydrated, critical-point dried, sputter coated with platinum (nominal depth of 2 nm), and imaged on a Hitachi 4700 FESEM with the upper, secondary electron detector
9 High-Resolution and Low-Voltage SEM of Plant Cells
241
Fig. 9.10 The developing cell wall over regenerating Mougeotia protoplasts after (a) 4 hours and (b) 6 hours of regeneration. JEOL 6000F, 5 kV. Scale bars 100 nm
242
Cox et al.
cytoplast remains and, interestingly so do some of the Hechtian strands, normally observed only with light microscopy. An alternative approach is to use protoplasts—plant cells from which the cell wall has been removed by enzymatic digestion of cellulose. These cells are extremely fragile but can be maintained in a medium containing a suitable osmoticum (typically sorbitol). Protoplasts will immediately begin forming a new cell wall, and within 24 hours will typically have a dense coating of cellulose microfibrils. They are, therefore, ideal for studying the deposition of cellulose at the plasma membrane from the exterior. Furthermore, since they are so fragile, bursting the protoplasts will reveal interior views often showing the plasma membrane and its associated cortical microtubules. These microtubules are widely regarded as being involved in the determination of cellulose microfibril alignment. Whiffen et al. [2002] studied this process in the filamentous alga Mougeotia. This is a unicellular alga, so the cells are cylindrical, and growth and organization are strongly polar. Protoplasts become spherical since, as in all green plants, the wall is responsible for maintaining cell shape. However, polarity is restored rapidly, with polar organization of microtubule foci visible with 4 hours (Galway & Hardham 1986). The original studies on this system were carried out by Marchant and colleagues (Marchant & Fowke 1977; Marchant 1978, 1979; Marchant & Hines 1979). In addition to transmission electron microscopy they used SEM, pioneering many of the techniques subsequently used in HRSEM work. However the relatively poor resolution available from SEMs of that period placed limitations on the level of detail that could be discerned In the HRSEM (Whiffen et al. 2002), the first, solitary, cellulose microfibrils are visible after two hours of regeneration. After four hours, a sparse meshwork covers the cell (see Fig. 9.10a), and after six hours there is a tightly woven cellulose mat surrounding the regenerating cell, but still no evidence of any preferred microfibril orientation (Fig. 9.10b). To observe the inner surface of the plasma membrane, cells were allowed to settle on plastic coverslips coated with polyethylenimine and then burst by osmotic shock in a microtubule-stabilizing buffer. Microtubules were not seen on plasma membrane fragments for the first three hours of regeneration, but scattered ones (confirmed by immuno-gold labeling) were seen by four hours. These showed no particular orientation, and even after six and seven hours, most microtubules seemed to be randomly arranged, though the beginnings of parallel arrays were occasionally seen. In many respects these results only confirmed what had previously been seen with TEM, but the ability to obtain images at high-resolution of the entire regenerating cell wall and relate it to cell shape adds a new dimension to these studies.
Conclusions High-resolution scanning electron microscopy has a lot to offer in the study of the structure and function of plant cells. Because of the rigid architecture of the plant cell, three-dimensional relationships are critical to cell organization at all stages
9 High-Resolution and Low-Voltage SEM of Plant Cells
243
of the cell cycle in a way that does not apply to most animal cells. The handful of examples presented in this chapter illustrates this point, and many other topics in plant cell biology await the opportunity to be probed in this way. The major barrier to further progress is the hugely time-consuming nature of the work. Even successful and reliable sample preparation techniques will typically require hours of searching in the microscope for useful fields—not because the preparation has failed, but because plant tissues intrinsically contain large expanses of “nothing”— vacuoles and intercellular spaces. Notably absent from this brief review are any studies using high pressure freezing (HPF) to vitrify extended samples. While most plant tissues cannot be vitrified to a depth of several cell layers, as is possible with animal material, high-pressure freezing has nevertheless been the basis of many of the most exciting recent studies of plant cell structure and function in the transmission electron microscope (Staehelin et al. 1990. See also Chapter 8 and 10). Many of these have involved very laborious three-dimensional reconstruction using TEM tomography techniques (Otegui & Staehelin 2004). There is every reason to believe that HPF could be equally beneficial for direct visualization of three-dimensional structure in the HRSEM, perhaps via freeze-fracture and deep etching. Or perhaps by some other technical breakthrough—the techniques in general use date back 15 years or more, and the time must be ripe for a novel approach. Acknowledgment Figures 9.5, 9.8 and 9.9 were obtained at the University of Missouri-Columbia’s Electron Microscopy Core Facility. The other figures were taken at the Electron Microscope Unit, University of Sydney, and we thank the staff there, especially Tony Romeo, for technical assistance. This research was partially supported by grants from the University of Sydney and the Australian Research Council. Work on the plant cell in T.I.B.’s laboratory is supported by the US Department of Energy (Award No. 03ER15421), which does not constitute endorsement by that department of the views expressed herein. We thank Brian Gunning for his continuing encouragement and enthusiasm for this work.
References Barnes SH (1992) Ultrastructural imaging of freeze-fractured plant cells in the scanning electron microscope. Microsc Res Tech 22:160–169 Baskin TI (2006) Imaging the primary cell wall. In: Hiyashi T (ed) The Science and Lore of the Plant Cell Wall. BrownWalker, Boca Raton, pp. 11–22 Brown R (1833) On the Organs and Mode of Fecundation in Orchidaceae and Asclepiaceae. Transactions of the Linnean Society of London 16:685–745 Cavalier-Smith T (1974) Basal body and flagellar development during vegetative cell cycle and the sexual cycle of Chlamydomonas reinhardtii. J Cell Sci 16:529–556 Cox G et al (1997) Of plants that swim—three-dimensional architecture of the flagellar root apparatus as revealed by confocal microscopy. Cell Vision 4:245–246 Dibbayawan TP, Cox G (1998) Flagellar root apparatus of Chlamydomonas: an immunocytochemical study using confocal microscopy and FESEM. In: Calderón Benavides HA, Jóse Yacamán M (eds) Electron Microscopy 1988: Proceedings of the 14th International Congress on Electron Microscopy, Cancun, Mexico, Vol IV. Institute of Physics Publishing, Bristol and Philadelphia, 729–730
244
Cox et al.
Fowke L et al (1999) Combined immunofluorescence and FESEM study of plasma membraneassociated organelles in highly vacuolated suspensor cells of white spruce somatic embryos. Cell Biol Internat (in press) Hooke R (1665) Micrographia. Royal Society, London Koga H et al (1992) Application of an osmium-maceration technique to observe plant-microbe interfaces of Italian ryegrass and crown rust fungi by scanning electron microscopy. Can J Bot 70:438–442 Lea PJ et al (1992) Chemical extraction of the cytosol using osmium tetroxide for high-resolution scanning electron microscopy. Microsc Res Tech 22:185–193 McLean B, Juniper BE (1993) The arrangement of actin bundles and chloroplasts in the nodal regions of Characean internodal cells. Eur J Phycol 28:33–37 Marchant HJ (1978) Microtubules associated with the plasma membrane isolated from the green alga Mougeotia. Experimental Cell Research 115:25–30 Marchant HJ (1979) Microtubules, cell wall deposition and the determination of plant cell shape. Nature 278:167–168 Marchant HJ, Fowke LC (1977) Preparation, culture and regeneration of protoplasts from filamentous green algae. Canadian Journal of Botany 55:3080–3086 Marchant HJ, Hines ER (1979) The role of microtubules and cell wall deposition in the elongation of regenerating protoplasts of Mougeotia. Experimental Cell Research 115:25–30 Marga F et al (2005) Cell wall extension results in the coordinate separation of parallel microfibrils: Evidence from scanning electron microscopy and atomic force microscopy. Plant Journal 43:181–190 Otegui MS, Staehelin LA (2004) Electron tomographic analysis of post-meiotic cytokinesis during pollen development in Arabidopsis thaliana. Planta 218:501–515 Reichelt S et al (1995) Visualization of immunogold-labelled cytoskeletal proteins by scanning electron microscopy. Eur J Cell Biol 67:89–93 Ringo DL (1967) Flagellar motion and fine structure of the flagellar apparatus in Chlamydomonas. J Cell Biol 33:543–571 Staehelin A et al. (1990) Macromolecular differentiation of Golgi stacks in root tips of Arabidopsis and Nicotiana seedlings visualised by high pressure frozen and freeze-substituted samples. Protoplasma 157:75–91 Schleiden M (1838) Beiträge zur Phytogenesis. Müllers Archiv, pp 137–176 Schwann T (1839) Mikroskopische untersuchungen über die ubereinstimmung in der struktur und dem wachstume der tiere und pflanzen. Re-edited by F. Hünseler in 1910. Engelmann, Leipzig Tanaka K, Fukudome H (1991) Three-dimensional organization of the Golgi complex observed by scanning electron microscopy. J Electron Microsc Tech 17:15–23 Tian GW et al (2004) The higher plant cortical microtubule array analyzed in vitro in the presence of the cell wall. Cell Motility and the Cytoskeleton 57:26–36 Vesk M et al (2000) Field emission scanning electron microscopy of plant cells. Protoplasma 210:138–155 Vesk PA et al (1994) Imaging of plant microtubules with high-resolution scanning electron microscopy. Protoplasma 182:71–74 Vesk PA et al (1996) Field emission scanning electron microscopy of microtubule arrays in higher plant cells. Protoplasma 195:168–182 Wiedemeier AMD et al (2002) Mutant alleles of arabidopsis RADIALLY SWOLLEN 4 and RSW7 reduce growth anisotropy without altering the transverse orientation of cortical microtubules or cellulose microfibrils. Development 129:4821–4830 Whiffen, LK et al (2002) High-resolution microscopy of regenerating Mougeotia (Chlorophyceae) protoplasts. European Journal of Phycology 37:339–347 Williamson RE (1993) Organelle movements. Ann Rev Plant Physiol 44:181–202 Wright RL et al (1985) A nucleus-basal body connector in Chlamydomonas reinhardtii that may function in basal body localization or segregation. J Cell Biol 101:1903–1912
Chapter 10
High-Resolution Cryoscanning Electron Microscopy of Biological Samples Paul Walther
Abstract Cryoscanning electron microscopy of fast frozen samples is the most direct approach for imaging aqueous organic material at nanometer scale resolution. It circumvents artifact formation caused by chemical fixation and drying. Cryofixation is preferentially done by high-pressure freezing because it allows for fixation of native samples up to 200 µm thick and 2 mm wide with minimal or no ice crystal damage. Because in the SEM only the surface of the sample (natural or artificial) is visible, water-covered biological surfaces must be made accessible to the electron beam by either freeze substitution, critical-point drying or freeze drying. If one wishes to view the inner part of a frozen sample, it must be opened by fracturing or by cryosectioning and then be coated with an electrically-conductive layer both to prevent charging and to produce contrast at the sample surface. Here we describe a method developed specifically for frozen-hydrated bulk samples: doublelayer coating. In this approach, the frozen sample is coated in a manner similar to that used to produce a TEM-replica, e.g., with a shadow of platinum (at an angle of 45◦ ) followed by an additional layer of carbon. The sample is then cryotransferred to an SEM equipped with a cold stage and imaged using the material-dependent backscattered electron signal (BSE), which shows the platinum distribution. With this method, charging artifacts and the effects of beam damage are significantly reduced compared with secondary electron imaging of conventionally-coated frozen samples. Although currently the resolution of the replica technique cannot be surpassed, double compared to layer coating greatly facilitates the processing of brittle rapidly frozen samples, because in contrast to TEM, no removal of the biological material from the replica is necessary. This makes cryo-SEM especially suitable for the analysis of high-pressure frozen samples. For small samples, such as macromolecular complexes or viral particles, the best resolution was achieved by a uniform coating with about 1.5 nm of tungsten and high-resolution imaging with secondary electrons. This method provides an extremely good signal-to-noise ratio and allows for macromolecular resolution without image averaging and has, therefore, a high potential for imaging of single events, such as actin-membraneinteractions.
H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
245
246
P. Walther
Introduction Scanning electron microscopy (SEM) has the potential to visualize structures from macroscopic resolution down to the molecular level. It is, therefore, able to fill the enormous gap between molecular biological imaging techniques such as x-ray diffraction and electron diffraction (in the range of Å) and light microscopical techniques, such as confocal laser microscopy (in the range of µm). This wide range of magnification allows one to study macromolecules in the complex context of cells, organelles or small organisms. The resolving power of commercial scanning electron microscopes has been considerably improved during the last twenty years due to the use of field-emission cathodes as electron sources (Crewe et al. 1968) and to improved lens design (reviewed by Pawley 1992). Modern high-resolution SEMs have a primary beam diameter of less than 1 nm (Nagatani et al. 1987). Resolution, however, does not solely depend on primary beam diameter, but is also a function of complex beamspecimen interactions, and depends on the nature of the specimen itself. Specimen preparation is, therefore, the key to making full use of the performance of modern high-resolution SEMs.
Cryoelectron Microscopy of Biological Samples Specimen preparation is necessary to make the structures of interest visible in the electron microscopes. Which preparative procedure is most suitable depends upon the question to be answered and technical limitations of the instrumentation (e.g., vacuum and electron irradiation in the SEM). In addition, preparative artifacts must be prevented if the sample is to be imaged in a state as close as possible to the living one. Müller (1992) formulated a “hierarchy of biological specimen preparation”: Specimen preparation starts with sampling (excision of tissue, harvesting of cells), followed by immobilization and fixation (e. g. freezing) and follow up procedures, and ends with microscopy, image interpretation and analysis. The final result of this chain of procedures can never be better than the first step, and at best, each step may only preserve the information density retained by the previous one. In this work, a specimen preparation concept is reviewed and further developed that enables us to investigate bulk samples, arrested in a defined physiological state, at nm scale resolution. It features the use of cryopreparation methods (reviewed by Ryan and Knoll 1994) whereby the native, aqueous sample is fast frozen and afterwards prepared in the frozen state, where it behaves like a solid sample. Cryopreparation can overcome limitations of classical preparation methods based on chemical fixation, followed by dehydration with organic solvents and drying. This work is based on a long tradition of cryoelectron microscopy. The potential advantages of cryopreparation methods for electron microscopy were appreciated as early as the 1950s (e.g., Fernandez-Moran 1960). In the early 1960s, however, ice crystal
10 High-Resolution Cryoscanning Electron Microscopy of Biological Samples
247
formation during freezing was recognized to be a major barrier to good preservation. Although, as a compromise, ice crystal formation could be prevented by adding a cryoprotective solution such as glycerol, because this could only be done after chemical prefixation and before freezing, it made the technique useless for arresting a physiological state by fast freezing. Later, techniques were developed that allowed for fast freezing of small samples from the native state without the use of cryoprotectants. These included quenchfreezing, freezing on a cold metal block and LN2 jet-freezing (reviewed by Bachmann 1987). However, because water both has a very high heat of fusion and is a poor conductor, it was only the development of high-pressure freezing (HPF) (Moor and Riehle 1968) that finally allowed cryofixation of samples more than a few µms thick. HPF permitted good results on samples up to about 200 µm thick and 2 mm in diameter. Another milestone was the invention of the "freeze-etch replica" technique, pioneered by Steere (1957) and then developed into a widely applied routine method by Moor & Mühlethaler (1963). In this technique, the cryofracture face is coated first with heavy metal and then carbon. After thawing the specimen, the metal-carbon replica was stripped off, cleaned and then viewed in the transmission electron microscope (TEM) (reviewed by Moor 1990). Thornley was the first to look directly at a frozen bulk sample mounted on a cold stage in the cryo-SEM (1960), a method further developed by Echlin (1971) and others (reviewed by Echlin 1992). It was, however, limited by the low topographic resolution capabilities of the SEMs then available. In the 1980s SEMs employing field-emission cathodes that permitted smaller primary beam diameters became commercially available and it was realized that one could only make use of this improved instrumental performance by a better understanding of contrast mechanisms (Seiler 1967; Reimer 1978; Peters 1982; Joy and Pawley 1990, 1992), beam-specimen interactions (Pawley and Erlandsen 1989) and by improving surface coatings (Peters 1986; Hermann & Müller 1991). In the next section, steps in the preparation of samples for high-resolution cryoSEM are described and analyzed. Cryofixation is followed by processing of the frozen samples to make the structures of interest amenable to being imaged with an electron beam. The next preparation steps (surface coating and imaging) are closely related and will be described together. Finally, we will discuss beam-specimen interactions in the cryo-SEM.
Specimen Preparation for The Cryo-SEM Cryofixation An aqueous sample can be converted into the solid state by freezing. During freezing, however, water tends to form crystals that damage the biological structure by both mechanical disruption and because the “mother liquor” remaining is hyperosmolar. The faster the freezing, the lower the tendency of the water to form crystals
248
P. Walther
and the less time available for the crystals to grow (reviewed by Bachmann and Meyer 1987). Unfortunately, ice has a poor thermal conductivity, limiting the freezing speed inside any biological sample. Therefore, even with an indefinitely high freezing speed at the surface, the depth of the sample that is frozen with no or minimal ice-crystal damage is limited to about 5 µm. As noted above, ice-crystal formation can be avoided by adding chemical freezing-protectants, such as glycerol, but as these additives disturb the physiological state of the sample, they should be avoided. The simultaneous application of a pressure of about 2000 bar in high-pressure freezing (reviewed by Müller and Moor 1984; Moor and Riehle 1968) enables one to freeze native samples up to a thickness of about 200µm and a diameter of 2 mm without the formation of “visible” ice crystals. The pressure reduces the formation of ice crystals by about a factor of 10 (Sartori et al. 1993). A number of publications make clear that the structure of high-pressure frozen samples more-closely mimics that of the in-vivo situation than occurs in with chemical fixation (Shimoni and Müller 1998; Szczesny et al. 1996; Kaneko and Walther 1995; and others). The state of the art in high-pressure freezing has been summarized in 11 articles of a special edition of the Journal of Microscopy (Oxford 212, Part 1, 2003). Highpressure freezing is now generally acknowledged as the gold standard for the preparation of cells and tissue for electron microscopy (Hohenberg et al. 2003). High-pressure freezing requires special equipment: At present there are three brands of high-pressure freezing machines on the market: The Bal-Tec HPM (BalTec, Principality of Liechtenstein) has been the first commercially available system on the market. The Wohlwend HPF Compact 01 (Wohlwend GmbH, Sennwald Switzerland, distributed by TECHNOTRADE International, Inc.: Manchester, New Hampshire) is based on a similar technical system and has some interesting extra features that will be discussed later. The Leica EM-Pact (Leica Microsystems: Vienna) is based on a different technical system. For cryo-SEM, high-pressure freezing is the ideal cryofixation tool because it allows one to freeze relatively large samples. The geometry of sample and sample holder should be optimized for freezing but should also take into consideration follow-up preparation steps, such as fracturing, sectioning and freeze drying. These steps require stable a mechanical attachment between the sample and the support.
Processing of the Frozen Samples SEM enables the investigation of bulk samples, but only structures available to the electron beam (i.e., natural or artificial surfaces) can be imaged. In biology, however, the structures of interest are often not at the sample surface, but are either inside of cells and tissue or on cell surfaces that are covered by water. To access the structures of interest, either the overlaying aqueous solution must be removed or the sample must be opened by cryofracturing or cryosectioning. This can be achieved by a wide variety of preparation methods for SEM and TEM.
10 High-Resolution Cryoscanning Electron Microscopy of Biological Samples
249
Investigation of Natural Surfaces Frozen-Hydrated Surfaces The easiest preparation can be performed, when the biological surface is naturally surrounded by gas (e.g., a plant leaf surface that is exposed to air). In this case, freezing is usually less critical, as long as no high-resolution information is required, because the surface can be directly exposed to the cryogen and therefore frozen relatively well, even if the bulk part of the sample experiences severe damage. After freezing, the samples can be immediately cryo-coated and observed in the cryo-SEM without any need for further preparation (reviewed by Echlin 1992). An example of this approach is given in Fig. 10.1.
Freeze-Dried Surfaces However, most biological surfaces are normally immersed in liquid medium. To observe these cell surfaces, this surrounding medium must be removed after the initial freezing step. Air drying would be an inappropriate approach because the surface tension at the liquid-gas interface destroys fine structural details. During freeze drying, however, the water sublimes from the frozen and therefore rigid sample, preventing the surface effects of air drying. Because only volatile components sublime during freeze drying, it is only useful if the sample is frozen in pure water. Some
Fig. 10.1 Example of a bulk surface that in vivo exposed to gas. Figure 10.1a shows a rust mite (Aculus schlechtendali) on an apple leaf. Figure 10.1b: The higher magnification of the encircled area shows the holes produced by the mite’s stylet during feeding. (Spieser et al. 1997). The living sample was frozen in liquid nitrogen, cryosputter coated and imaged in the frozen-hydrated state (For technical details see the Materials and Methods section)
250
P. Walther
Fig. 10.2 Surface structures on the cell wall of the yeast cell Saccharomyces cerevisiae. They are covered with fimbriae that are visible after partial freeze drying in the SEM (Fig. 10.2a). The same fimbriae structure is also visible after freeze substitution, embedding and thin sectioning (Walther et al. 1988). The fimbirae are no longer visible when the cells are dehydrated with ethanol and critical-point dried (Fig. 10.2b)
samples, such as yeast, can withstand distilled water. It turns out that the final appearance of the yeast cell wall is influenced by the preparation technique used. When prepared by partial freeze drying, hair-like structures are visible on the cell wall surface on bulk SEM samples (see Fig. 10.2a). Similar hair-like surface structures were also observed after freeze substitution, embedding and thin-sectioning (Walther and Ziegler 2002). However, in a comparative study where cells were dehydrated with ethanol and critical-point dried, they appeared smooth (see Fig. 10.2b), indicating that the surface structure was destroyed by room-temperature dehydration. It is now evident that these hair-like structures are mainly glycoproteins that are influenced by the physiological state of the cell (Walther et al. 1988).
Fracture Faces The easiest way to look inside a frozen sample is to cryofracture it. In this field, a lot of work has already been done since the TEM replica technique was introduced by Steere in 1957 and later improved by Moor & Mühlethaler (1963). In this approach, the surface of the fractured sample is coated with a heavy metal layer (e.g., platinum) to create contrast, and this is followed by a thicker carbon layer for mechanical stability. Afterwards, all the organic components are removed from the replica by cleaning it in bleach or sulfuric acid, making it beam transparent so that it can be imaged in the TEM. A limitation of this very successful method is that, because biological samples frozen from the native state are very brittle, the
10 High-Resolution Cryoscanning Electron Microscopy of Biological Samples
251
replica tends to fall into small pieces during cleaning. This problem is circumvented in cryo-SEM because, as the bulk specimen itself can be imaged rather than just the beam-transparent replica, the replica-cleaning process is unnecessary (Read & Jeffree 1991; Echlin 1992). As a result, one can image freeze-fractured surfaces that would be difficult to replicate, such as the very brittle high-pressure frozen samples (see Figs. 10.3 (see color plate 13) and 10.4).
Fig. 10.4 High-pressure-frozen and cryofractured pancreas tissue imaged in the SEM with backscattered electrons. The arrows point to porous structures in the ER membranes, which were described earlier as fenestrae (see Results). Figure 10.4a: Perpendicular fracture to the ER membranes in the middle of the image, and fracture parallel to the membranes at the right hand side. Figures 10.4b and 10.4c: Higher magnification of a similar area. The cytosol is filled with ribosomes and some filamentous structures. The different fracture faces of the ER membranes are exposed. The smooth extraplasmic fracture face (EF) is facing the ER lumen; the rough protoplasmic fracture face (PF) is facing the cytosol
252
P. Walther
Surface Coating and Imaging the Samples in the Cryo-SEM The problems of imaging in the cryo-SEM (contrast formation and beam damage) are closely related to the problems of surface coating. They will, therefore, be discussed together. The goal is to visualize as much detail as possible in well-preserved structures.
Physical and Instrumental Parameters That Limit Spatial Resolution Joy and Pawley (1992) defined the limitations of spatial resolution in the SEM as follows: “The spatial resolution of the scanning electron microscope is limited by at least three factors: the diameter of the electron probe, the size and shape of the beam/specimen interaction volume with the solid for the mode of imaging employed and the Poisson statistics of the detected signal.”
The Diameter of the Primary Beam as a Function of the Accelerating Voltage The diameter of the probe is the first resolution-limiting factor of all scanning type microscopes. Field-emission SEMs of the in-lens type provide the smallest available probe sizes of less than 1 nm at Vo = 30 kV . Most of the studies in this work were performed with the Hitachi S-900 and S-5200 in-lens SEM, which provide a beam diameter of less than 1 nm at 30 kV and about 1 nm at Vo = 10 kV (Nagatani et al. 1987). At lower primary accelerating voltages, however, the diameter increases and reaches about 3 nm at 1 kV (Nagatani et al. 1987).
Electron Diffusion in the Sample as a Function of the Accelerating Voltage. Whereas the diameter of the primary beam depends almost exclusively on the instrument used, the other two resolution limiting factors (electron diffusion and statistics) are much more complex and are mainly a function of the specimen itself. The interaction volume of the primary beam in the sample increases with increasing Vo because electrons with higher energy penetrate deeper into the sample (reviewed by Pawley 1992). The interaction volume affects the spatial resolution because the electrons that finally form the image originate not only from the impact point of the primary beam, but from any part of the larger interaction volume that approaches an external surface. Because resolution is also heavily influenced by the properties of the sample surface, the effect that this penetration volume has on spatial resolution is complex, and is highly dependent on the surface coating, as discussed by many authors (Seiler 1967; Peters 1982, 1986; Joy and Pawley 1992). The work described here sometimes uses the backscattered electron signal (see Figs. 10.1, 10.3, and 10.4), which has a number of advantages over the secondary electron signal: Because backscattered electrons have much higher energy, the image is less affected by unpredictable specimen charging. Contrast for-
10 High-Resolution Cryoscanning Electron Microscopy of Biological Samples
253
mation is, therefore, easier to understand than with secondary electrons. The method presented here requires a very sensitive detector for backscattered electrons. We used a YAG-type (cerium-dopedyttrium-aluminum-garnet) detector as described by Autrata et al. (1992). This detector is very sensitive down to accelerating voltages of 3 kV. The sensitivity, however, drops considerably below 3 kV (Walther et al. 1991). At lower Vo the electrons do not penetrate so far into the sample; consequently, the signal is produced closer to the sample surface. On a coated sample, therefore, a better signal-to-noise ratio is obtained at lower Vo . However, accelerating voltages below 3 kV were not useful because of low BSE detector sensitivity. On the other hand, at low Vo the diameter of the primary beam is increased, which also reduces the resolution. In conclusion, at low magnifications, when the diameter of the primary beam is not the factor that limits resolution, one gets more contrast and less radiation damage at lower Vo (e.g. 3 to 10 kV), whereas at very high magnifications (above 100 000 x) the diameter of the primary beam is limiting and, therefore, a higher resolution is obtained at high voltage (e.g., 30 kV; see Fig. 10.5). For many applications an accelerating voltage of 10 kV was found to be useful.
Uncoated Samples The most direct approach would be to look at uncoated samples (see Figs. 10.3d and 10.5a), thereby preventing artifacts related to surface coating. Uncoated biological samples are electrical insulators; therefore, they tend to charge in the electron beam. This tendency is reduced at low accelerating voltages because proportionally more backscattered and secondary electrons are produced, and an equilibrium between arriving and departing electrons can be approached on the sample surface. In addition, at low V0 , the penetration depth of the primary beam is reduced, as explained above, which means that the image-forming signal is produced closer to the impact point of the primary beam. However, it turned out that the limiting factor is not so much the diameter of the primary beam as the poor contrast and signal production on an uncoated biological sample. Although in many cases low magnification images of uncoated samples look relatively good and are comparable to images of coated samples, at magnifications above x10,000 the image quality becomes very poor. This is because charging can only be neutralized when averaged over a large area but not on every detail at high magnification, and local charging still makes image interpretation difficult. The problem of charging on uncoated biological specimens can be greatly reduced by using the backscattered electron signal, as this signal is less sensitive to the small local electrostatic fields caused by sample charging (Walther et al. 1991). This method works well at low magnifications. At higher magnification, however, resolution is very limited: Intramembranous particles, for example, are not visible, probably because even at the low V0 of 3 kV, the extraction depth of the backscattered electrons in biological samples is still considerably larger than the size of such particles (about 10 nm) (Seiler 1962). Unfortunately, below 3 kV the backscattered electrons do not have enough energy to produce sufficient signal in scintillator/PMT detectors.
254
P. Walther
Fig. 10.5 High-resolution cryo-SEM of samples imaged with secondary electrons. Figure 10.5a represents a tip from an atomic force microscope covered with tobacco mosaic viruses. The sample was not metal coated. On this special sample, the best results regarding resolution and contrast were obtained at low kV (1.5 kV). All other samples of Fig. 10.5 were plunge-frozen, partially freeze dried at a temperature of 177 K and cryo-coated with about 1 nm of tungsten by electron-beam evaporation. These samples were imaged with the secondary electron signal at 173K. Figures 10.5b and 10.5c represent tobacco mosaic virus. The optical diffraction pattern of the encircled area (Fig. 10.5d) shows diffraction spots at 2.2 nm corresponding with the periodic protein coat. This is the best resolution achieved with an SEM in our hands. Figure 10.5d shows two freeze-dried and tungsten coated actin filaments. Actin subunits are directly visible on this image without image processing or averaging. Similar results have also been shown by Wepf et al. (1994). We believe that this approach is powerful for the study of non-periodic structures such as actin-membrane interactions. Figures 10.5f and 10.5g: Adenovirus at medium and high magnification. Some of the fibers with the terminal knobs are attached to the carbon grid and are therefore well retained
10 High-Resolution Cryoscanning Electron Microscopy of Biological Samples
255
There are special occasions in which particular material combinations favor the use of uncoated samples. Figure 10.5a shows such a specimen, the tip of an atomic force microscope, coated with tobacco mosaic viruses in order to investigate the elastic module of the protein envelope. However, this technique is generally not considered to be suitable for imaging uncoated biology samples at high-resolution.
Coated Samples The ideal coating strategy is to cover the specimen surface with a thin, homogenous metal layer to improve electrical conductivity. This reduces charging; however, because the metal has a much higher density than the biological material, it also localizes signal generation closer to the point at which the primary beam strikes the surface of the sample (Peters 1986; Hermann & Müller 1991). A further improvement can be achieved by optimizing the coating layer. To replicate the surface profile as nearly as possible, the grain size must be as fine as possible, such as those that characterize tungsten and chromium films (Peters 1986; see Fig. 10.5). When it comes to relatively large bulk samples, thin coatings have two limitations: The conductivity of a thin layer (e.g., 1 to 3 nm of platinum) is usually insufficient when it must be applied over a large area (e.g., 1 mm). In addition, as is known from cryo-TEM and cryo-STEM, frozen-hydrated samples are very sensitive to the mass loss caused by electron beam irradiation (Zierold 1983, 1986). Such mass loss is a function of the electron dose used (Heide 1984), and this can be calculated from the beam current, the exposure time and the exposed area. The beam current used in this study was 1−5×10−11 A. Scanning an electron beam of 2×10−11 A over an area of 1 µm2 (corresponding to a primary magnification of 100k x) for 40 s deposits an electron dose of 5×103 electrons per nm2 . According to Heide (1984), this induces a mass loss from the sample surface of about 2.6 nm thickness. Fortunately, this problem can be partially overcome by using the double-layer coating (Walther and Hentschel 1989). This method is similar to that used for making TEM replicas (Moor & Mühlethaler 1963). The sample is first coated with a thin layer of heavy metal (platinum-carbon or tungsten with an average thickness of 2 to 3 nm). This layer is then backed up with a 5 to 10 nm carbon layer to improve mechanical stability and enhance electrical conductivity. However, when such a double-layer-coated sample is imaged with secondary electrons, the contrast is mainly formed by the real physical surface of the sample, i.e., the overlying 5 to 10 nm carbon coat (Walther et al. 1995), which hides small surface structures. However, if the backscattered electron signal is used (see Figs. 10.1, 10.3, and 10.4) the resulting signal is mostly a function of Z. The electron beam striking a double-layer coated sample passes through the low-atomicnumber carbon layer with minimal scattering, and most of the contrast is produced by scattering events that occur at the heavy metal layer that is in close contact with the biological structures of interest. Although the electron beam may penetrate even deeper into the sample, little of the scattering that occurs there produces detectable contrast, and, in addition, as it is caused by the homogenous bulk of the sample it
256
P. Walther
is almost random and only adds statistical noise to the image. The most remarkable effect of the additional carbon layer is that it enhances mechanical stability, making the specimen surface almost immune to the effect of mass loss (Walther et al. 1995; Walther and Müller 1997). Although the backscattered electron images produced in this way are noisier than the best secondary electron images (see Fig. 10.5) the method greatly facilitates work at the microscope. Because the samples are less sensitive to the electron beam, low-dose techniques are not required, and frozen-hydrated samples can be imaged almost as easily as conventional ones, a feature that may account for the fact that the technique has been used by several other authors (Hermann & Müller 1997; Dreher et al. 1997; Meile et al. 1997; Watzke & Dieschbourg 1994; Herter et al. 1991; Njisse et al. 2004; Erlandsen et al. 1997; Wild et al. 2005). The source of this reduced sensitivity is most probably related to the following two effects. First, the carbon layer may act as a rigid mechanical structure that keeps the platinum granules in place, even if the underlying aqueous structures are destroyed. Second, the carbon layer may act as a barrier that prevents material from leaving the sample (Zierold, personal communication). This effect would explain the reduced mass loss. However, when it comes to very high-resolution imaging (above x200 k), it was found that the best resolution is achieved by applying a very thin layer of tungsten (average thickness in the order of 1 nm) that is then imaged with the secondary electron image (Hermann et al. 1988; Hermann & Müller 1991). Figure 10.5 gives some examples of this approach. It is assumed that in this case the signal-to-noise ratio is better when using the secondary electron signal compared to the backscattered electron signal, and this becomes resolution limiting at very high magnifications. In addition, the 10 nm carbon coat used for double-layer coating might cause some slight spreading of the primary beam that becomes relevant at very high magnifications above 200k x.
Materials and Methods Figure 10.1: Mites on Apple Leaves Apple leaf samples with the rust mite Acari Eriophyiidae on it were punched out with an ophthalmologic punch and frozen by plunging into liquid nitrogen. The frozen samples were mounted onto a Gatan cryoholder and cryosputter coated in a MED 020 (Bal-Tec, Princ. of Liechtenstein) with 7 nm of platinum at a temperature of 140 K. The samples on the cryoholder were cryotransferred in liquid nitrogen to the Hitachi S-900 in-lens field-emission scanning electron microscope. The images of the frozen-hydrated samples were recorded digitally with a Gatan Digi-Scan interface at a temperature of 140 K and an accelerating voltage of 10 kV using the backscattered electron signal as described by Walther and Müller (1997) (Spieser et al. 1998).
10 High-Resolution Cryoscanning Electron Microscopy of Biological Samples
257
Figures 10.2 a and b: Surface of Freeze Dried (Fig. 10.2a) and Critical-Point Dried (Fig. 10.2b) Yeast Figure 10.2a: The yeast cells (Saccharomyces cerevisiae) were washed with distilled water and pelleted. A 400 mesh gold grid was dipped into the loose yeast pellet and placed between two, low-mass copper platelets (Müller and Moor 1984). Rapid freezing was then achieved by the propane-jet technique (Müller et al. 1980) without chemical pretreatment. Freeze fracturing and freeze drying for the SEM were performed in a Bal-Tec BAF 300 freeze etching instrument. The rapidly frozen specimen sandwiches were inserted into a special holder (Müller and Moor 1984) and fractured at 168 K at a pressure of 10−7 mbar. The samples were then warmed to 173 K and dried for 35 min. The sample was then coated by electron-beam evaporation with 5 nm of carbon and then 5 nm of platinum-carbon. After warming to room temperature the sample was examined at room temperature in a Hitachi S-700 below-the-lens field-emission SEM using the secondary electron signal (Walther et al. 1988). Figure 10.2b: Critical-point dried yeast (Saccharomyces cerevisiae). The cells were chemically fixed with 2% glutharaldehyde in cacodylate buffer, followed by 1% osmium tetroxide in cacodylate buffer. Afterwards the samples were dehydrated in a graded series of ethanol and critical-point dried using carbon dioxide (Anderson 1951). The samples were then coated as described for Fig. 10.2a.
Figures 10.3a, 10.3b, and 10.3c A 300-mesh copper EM grid was dipped into a pellet of commercially- available baker’s yeast cells (Saccharomyces cerevisiae) kept in the stationary growth phase and mounted between the flat sides of two aluminum planchettes (Engineering Office M. Wohlwend GmbH, CH-9466 Sennwald, Switzerland). These sandwiches were frozen with the new high-pressure freezer Wohlwend Compact HPF 01 (TECHNOTRADE International, Inc.: Manchester, New Hampshire) without any cryoprotectants and without ethanol (Ethanol is usually used to synchronize pressure and temperature drops in the pressure chamber of a high-pressure freezer but, as an option, it can be switched off in the new HPF compact 01 without affecting performance due to increased synchronization). The frozen sandwiches were mounted on a Bastacky holder. After transfer to the BAF 300 freeze-etching device (Bal-Tec: Principality of Liechtenstein), the temperature of the sample stage was raised to 155 K (vacuum about 2×10−7 mbar). The sandwiches were cracked with a steel knife and then double-layer coated, as described by Walther and Hentschel (1989) and Walther et al. (1995), by electron beam evaporation with 3 nm of platinum-carbon from an angle of 45◦ and about 6 nm of carbon perpendicularly. The additional carbon coat increases electrical conductivity (prevents charging) and even more importantly, reduces the effects of beam damage (Walther et al. 1995). After coating, the freeze-etching device was vented with dry nitrogen and the Bastacky-holder
258
P. Walther
holding the sample quickly removed and immersed in liquid nitrogen, where it was mounted onto the Gatan cryoholder 626 (Gatan, Inc.: Pleasanton, California). Under liquid nitrogen (cryoshield of the Gatan holder closed), the sample was transferred to the SEM and quickly inserted. After about 5 min the cryoshield was opened. Specimens were investigated at a temperature of 148 K in the Hitachi S-5200, inlens field-emission SEM, (Hitachi: Tokyo, Japan). The beam current was about 1–3 times 10−11 A. The primary accelerating voltage (Vo) was 10 kV. Imaging was performed by collecting the backscattered electron (BSE) signal. The contrast is mainly formed by the underlying platinum layer, which stays in contact with the biological structure of interest. When the secondary electron signal was used, one would image the surface of the overlaying, relatively thick carbon coat and small structural details are lost (Walther and Hentschel 1989, Walther & Müller 1995, 1999). BSEs were directly recorded with the built in YAG-BSE detector and (indirectly) with the converted BSE signal, the so-called composite rich image. In order to improve the signal-to-noise ratio, the YAG-BSE image and the composite rich image were superimposed (Since frozen-hydrated samples are sensitive to the electron beam, it is important to keep the irradiation dose as low as possible). The images were recorded digitally with a resolution of 1280 × 960 pixels. The digital images were processed with Adobe Photoshop and Microsoft Power Point software. To reduce the apparent noise, we used the Adobe Photoshop Gaussian blur filter with a radius of 0.3 pixel. No other image processing apart from brightness and contrast correction was performed.
Figure 10.3d: Observation of Uncoated Yeast Fracture Faces A pellet of commercially available baker’s yeast cells (Saccharomyces cerevisiae) in the stationary growth phase was mounted on gold planchettes (Bal-Tec: Principality of Liechtenstein) and frozen by plunging it into ethane cooled by liquid nitrogen. The frozen pellet was fractured in liquid nitrogen with a razor blade and the planchettes were mounted on a Gatan cryostage. The samples were directly transferred into a Hitachi S-900 SEM without coating and imaged at a low accelerating voltage of 2.6 kV using the backscattered electron signal recorded with a sensitive annular YAG-detector (Walther et al. 1991).
Figure 10.4. High-Pressure Freezing, Freeze-Fracturing, and Cryosectioning Of Pancreas Tissue Small pieces (200 or 400 µm thick and about 1 mm in length) were cut with a razor blade from the pancreas of anaesthetized Wistar rats (100–150 g) and placed in aluminum planchettes (diameter 3 mm with a central cavity of 2 mm in diameter and 200 µm depth; Engineering Office M. Wohlwend GmbH, CH-9466 Sennwald, Switzerland). The cavities between the tissue and the aluminum planchettes
10 High-Resolution Cryoscanning Electron Microscopy of Biological Samples
259
were filled with hexadecene (Studer et al. 1989). The planchette thick was covered with another planchette, so that a cavity 200 or 400 µm was formed, and then high-pressure frozen with the standard holder in a Wohlwend high-pressure freezer as described previously for Fig. 10.3. The frozen tissue pieces were carefully removed from the planchettes under liquid nitrogen with the help of a scalpel, and then clamped into a special holder (Bastacky et al. 1995) in between two indium foils to improve mechanical, thermal and electrical contact. Stable mounting of the frozen sample is a most crucial preparation step for successful cryosectioning (Walther & Müller 1999; Walther 2003a and 2003b). After transfer to the BAF 300 freeze-etching device (Bal-Tec: Principality of Liechtenstein), the temperature of the sample stage was raised to 155 K, (vacuum about 2×10−7 mbar). The samples were cryofractured with a steel knife and then double-layer coated, transferred to the cryo-SEM and imaged as described previously for Fig. 10.3.
Figure 10.5a Tobacco mosaic viruses were attached to the tip of an atomic force microscope (AFM). This sample was not coated with a metal in order to perform AFM force measurements after SEM observation. This sample was imaged at a low primary beam voltage of 1.5 kV to prevent charge-up.
Figure 10.5b to 10.5g Tobacco mosaic virus, actin filaments and Adenovirus particles were adsorbed on a carbon film on a 400 mesh EM grid. The water was partially removed by blotting and the samples were plunge frozen in ethane supercooled with liquid nitrogen. The samples were then mounted on a Bastacky holder and transferred to the BAF 300 freeze-etching device (Bal-Tec: Principality of Liechtenstein). The temperature of the sample stage was raised to 178 K (vacuum about 2 ×10−7 mbar), and the samples were freeze-dried for 20 min. Then the samples were rotary coated with 1.5 nm tungsten at an angle of 45◦ (Before starting the experiment some tungsten was evaporated in order to remove tungsten oxide). After coating, the holder with the samples was mounted on the Gatan stage, cryotransferred to the Hitachi S-5200 SEM and imaged at a temperature of 178 K and an accelerating voltage of 30 kV recording the secondary electron signal. Acknowledgment I thank my supervisors Martin Müller (Zurich), the late Karl Zierold (Dortmund), and Jim Pawley (Madison) for teaching me the miracles of cryoelectron microscopy. I am also very grateful to the late Stan Erlandsen (Minneapolis) because he supported my work many times by letting me work in his laboratories and by inviting me to several conferences in the United States.
260
P. Walther
References Anderson TF (1951) Techniques for the preservation of three-dimensional structures in preparing specimens for the electron microscope. Trans NY Acad Sci. 13:130–133 Autrata R et al (1992) An efficient single crystal BSE detector in SEM. Scanning 14:127–135 Bachmann L, Meyer E (1987) Physics of water and ice: Implications for cryo-fixation. In: Steinbrecht RA, Zierold K (eds) Cryotechniques in Biological Electron Microscopy. Springer, Heidelberg, pp 3–34 Bastacky Jet al (1995) A specimen holder for high-resolution low-temperature scanning electron microscopy. Microsc Res Tech 32:457–458 Crewe AV et al (1968) Electron gun using a field emission source. Rev Sci Instrum 39:576–583 Dreher F et al (1996) Interaction of a lecithin microemulsion gel with human stratum corneum and its effect on transdermal transport. J of Controlled Release 45:131–140 Echlin, P (1971) The examination of biological material at low temperatures. In: Johari O (ed) (1971/1) Scanning Electron Microscopy, ITTRI, Chicago, pp 225–232 Echlin P (1992) Low-Temperature Microscopy and Analysis. Plenum, New York Erlandsen SL et al (1997) High-resolution backscattered electron detection of cell surface molecules on human platelets using the double-layer coating method and cryo-field emission scanning electron microscopy. Scanning 19:356–360 Fernandez-Moran H (1960) Low temperature preparation techniques for electron microscopy of biological specimens based on rapid freezing with helium II. Ann NY Acad Sci 85:689–713 Gross H et al (1978) Freeze fracturing in ultrahigh vacuum at -196◦ . C J Cell Biol 76:712–728 Heide HG (1984) Observation on ice layers. Ultramicroscopy 14:271–278 Hermann R et al (1988) Double-axis rotary shadowing for high-resolution scanning electron microscopy. Scanning Microsc 2:1215–1230 Hermann R, Müller M (1991) High-resolution biological scanning electron microscopy: A comparative study of low temperature metal coating techniques. J Electron Microsc Techn 18:440–449 Hermann R, Müller M (1997) Limits in high-resolution scanning electron microscopy: Natural surfaces? Scanning 19:337–342 Herter P et al (1991) High-resolution scanning electron microscopy of inner surfaces and fracture faces of kidney tissue using cryo-preparation methods. J Microsc 161:375–385 Hohenberg HH et al (2003) Foreword, special issue on high pressure freezing. J Microsc 212:1–2 Joy DC, Pawley JB (1992) High-resolution scanning electron microscopy. Ultramicrosc 47:80–100 Kaneko Y, Walther P (1995) Comparison of germinating pea leaves prepared by highpressure freezing—freeze substitution and conventional chemical fixation. J Electron Microsc 44:104–109 Kühlbrandt W et al (2002) Regulation of the neurospora H+ -ATPase. Science 297:1692–1696. Meile L et al (1997) Bifidobacterium lactis sp. nov., a moderately oxygen tolerant species isolated from fermented milk. System Appl Microbiol 20:57–64 Moor H, Mühlethaler K (1963) Fine structure in frozen-etched yeast cells. J Cell Biol 17:609–628 Moor H, Riehle U (1968) Snap-freezing under high pressure: a new fixation technique for freezeetching. Proc 4th European Reg. Conf. of Electron Microscopy 2:33–34 Moor H (1990) Cryo-techniques and related methods. In: Günter JR (ed) History of electron microscopy in Switzerland. Birkhäuser Verlag. Basel, pp 191–195. Müller M et al (1980) Freezing in a propane jet and its application in freeze-fracturing. Mikroskopie 36:129–140 Müller M, Moor H (1984) Cryo-fixation of suspensions and tissues by propane jet freezing and high-pressure freezing. Proc 42nd Ann Meet Electron Microsc Soc Am:6–9. Müller M (1992) The Integrating Power of Cryo-fixation Based Electron Microscopy in Biology. Acta Microscopica 1:37–44 Nagatani T et al (1987) Development of an ultra high-resolution scanning electron microscope by means of a field-emission source and in-lens system. Scanning Microsc 1:901–909 Nijsse J et al (2004) Cold-induced imbibition damage of lettuce embryos: A study using cryoscanning electron microscopy. Seed Science Research 14:117–126
10 High-Resolution Cryoscanning Electron Microscopy of Biological Samples
261
Pawley JB, Erlandsen SL (1989) The case for low voltage high-resolution scanning electron micrsocopy of biological samples. Scanning Microsc 3:163–178 Pawley J B (1990) Practical aspects of high-resolution LVSEM. Scanning 12:247–252 Pawley JB (1992) LVSEM for high-resolution topographic and density contrast imaging. Adv in El and El Physics 83:203–275 Peters KR (1982) Conditions required for high quality high magnification images in secondary electron-I scanning electron microscopy. Scanning Electron Microsc 4:1359–1372. Peters KR (1986) Rationale for the application of thin, continuous metal films in high magnification electron microscopy. J Microsc 142:25–34 Read ND, Jeffree CE (1991) Low-temperature scanning electron microscopy in biology. J Microsc 161:59–72 Reimer L (1978) Scanning electron microscopy—present state and trends. Scanning 1:3–16 Ryan KP, Knoll G (1994) Time-resolved cryo-fixation methods for the study of dynamic cellular events by electron microscopy: a review. Scanning Microsc 8:259–288 Sartori N et al (1993) Vitrification depth can be increased more than 10-fold by high-pressure freezing. J Microsc 172:55–61 Seiler H (1967) Einige aktuelle probleme der sekundärelektronenemisison. Z angew Phys 22:249–263 Shimoni E, Müller M (1998) On optimising high pressure freezing: from heat transfer theory to a new micro-biopsy device. J Microsc 192:236–247 Spieser F et al (1998) Impact of Aculus schlechtendali (NALEPA) feeding on apple leaf gas exchange and leaf color associated with changes in leaf tissue. (Environ Entomol 27:1149–1165 Steere RL (1957) Electron microscopy of structural detail in frozen biological specimens. J Biophys Biochem Cytol 3:45–60 Studer D et al (1989) High-pressure freezing comes of age. Scanning Microsc 3, Suppl 3: 253–268 Szczesny PJ et al (1996) Light damage in rod outer segments: The effects of fixation on ultrastructural alterations. Current Eye Res 15:807–814 Thornley RFM (1960) Recent developments in scanning electron microscopy. Proc EUREM 173–176 Walther P et al (1988) Morphological organization of glycoprotein containing cell surface structures in yeast. J Ultrastruct Molecular Struct Res 101:123–136 Walther P, Hentschel J (1989) Improved representation of cell surface structures by freeze substitution and backscattered electron imaging. Scanning Microsc 3, Supplement 3: 201–211 Walther P et al (1991) Backscattered electron imaging for high-resolution surface SEM with a new type YAG-detector. Scanning Microsc 5:301–310 Walther P et al (1995) Double layer coating for high-resolution low temperature SEM. J Microsc 179:229–237 Walther P, Müller M (1997) Double layer coating for field emission cryo-SEM—present state and applications. Scanning 19:343–348 Walther P and Müller M (1999) Biological ultrastructure as revealed by high-resolution cryo-SEM of blockfaces after cryo-sectioning. J Microsc 196(3):279–287 Walther P, Ziegler A (2002) Freeze substitution of high-pressure frozen samples: the visibility of biological membranes is improved when the substitution medium contains water. J Microsc 208:3–10 Walther P (2003a) Recent progress in freeze fracturing of high-pressure frozen samples. J Microsc 212:34–43 Walther P (2003) Cryo-fracturing and cryo-planning for in-lens cryo-SEM, using a newly designed diamond knife. Microscopy and Microanalysis 9:279–285. Watzke H, Dieschbourg C (1994) Novel silica-biopolymer nanocomposites: The silica sol—gel progress in biopolymer organogels. Adv in Colloid and Interface Science 50:1–14 Wepf R et al (1994) High-resolution SEM of biological macromolecular complexes. In: Bailey GW, Garratt-Reed AJ (eds) Proc. 52 Ann. Meet Microsc. Soc. Amer., San Francisco Press, San Francisco, pp. 1026–1027
262
P. Walther
Wild P et al (2005) Impairment of nuclear pores in bovine herpesvirus 1 infected mdbk cells. J Virol 79:1071–1083 Zierold K (1983) X-ray microanalysis of frozen hydrated specimen. Scanning Electron Microsc 2: 809–825 Zierold K (1986) Preparation of cryo-sections for biological microanalysis. In: Müller M et al (eds) (1985) The science of biological specimen preparation. SEM Inc., AMF O’Hare. 119–127.
Chapter 11
Developments in Instrumentation for Microanalysis in Low-Voltage Scanning Electron Microscopy Dale E. Newbury
Abstract X-ray microanalysis performed in the low-voltage scanning electron microscope, where low voltage is generally considered to involve operation with an incident beam energy of 5 keV or lower, offers significant improvements in the lateral and depth spatial resolution compared to conventional beam energy operation of 10 keV and higher. This improvement must be offset against the limitations that the low-beam energy imposes upon analytical x-ray spectrometry. The lowered primary excitation restricts the atomic shells that can be excited, generally to a value below 90% of the beam energy. This situation forces the analyst to choose L and M shell x-rays, where under conventional excitation conditions the K and L shells would be chosen because of their much higher x-ray (fluorescence) yields. The consequence is that the analyst must contend with poorer limits of detection, and some elements normally measured may become practically inaccessible at concentrations below 0.1 mass fraction, or even higher. New developments in x-ray spectrometry techniques, including optic-augmented wavelength-dispersive spectrometry and microcalorimeter energy-dispersive x-ray spectrometry, will improve the analysis situation. Key words: biological analysis, electron probe microanalysis, microanalysis, scanning electron microscopy, spectrometry, x-rays
Introduction X-ray microanalysis performed in the scanning electron microscope (SEM) is a powerful technique for characterization of elemental distributions on a microscopic scale in biological specimens and has a long history of applications (Echlin & Galle 1976; Echlin 1984; Echlin 1992; Goldstein et al. 2003). Because of the interest in fine-scale biological microstructures, especially at interfaces and membranes, improvement in the lateral and depth spatial resolution has been a research objective from the earliest activity in the field. Dramatic advances in SEM instrumentation performance in recent years have made possible effective imaging operation at beam energies well below those that have been conventionally used. This so-called H. Schatten, J. B. Pawley (eds.), Biological Low-Voltage Scanning C Springer 2008 Electron Microscopy.
263
264
D. E. Newbury
low-voltage SEM or LVSEM operation also opens the possibility of low beam energy x-ray microanalysis, and this capability has been successfully applied to biological problems by several researchers (Echlin 2002). LVSEM microanalysis is, however, subject to many restrictions not normally encountered in conventional electron beam x-ray microanalysis. In particular, there are aspects of the x-ray measurements in LVSEM x-ray microanalysis that suffer limitations due to the performance of the x-ray spectrometers that are commonly available. New developments in x-ray spectrometry will impact these limitations and permit more effective application of the LVSEM to biological systems.
Improved Spatial Resolution: The Rationale for LVSEM X-Ray Microanalysis X-Ray Production and Spatial Resolution in Conventional Beam Energy Analysis Conventional electron beam x-ray microanalysis in the scanning electron microscope (SEM) is considered to span the beam energy range from approximately 10 keV to 30 keV (Goldstein et al. 2003). The upper energy limit is set by the accelerating voltage range of the SEM. The practical minimum depends upon the excitation function for x-ray production. Production of x-rays by electron excitation requires that the incident beam energy, E0 , prior to any energy loss in the target, exceeds the critical ionization energy, Ec , for the atomic species of interest. The ratio of beam energy to critical ionization energy is referred to as the overvoltage, U: U = E0 /Ec
(11.1)
Thus U must exceed unity for a given species in order for any characteristic x-ray production to occur. For a particular element and ionization edge, the intensity, Ich , of characteristic x-ray production varies strongly with U: Ich ∼ (U − 1)n
(11.2)
where the exponent, n, depends on species and atomic shell, 1.3 ≤ n ≤ 1.7. When the x-ray production from different elements is compared, or else from different ionization edges for the same element, an additional factor, the fluorescence yield, ω, which is the fraction of ionization events that give rise to x-ray photon emission, as opposed to Auger electron emission, must be considered. In general, when different elements are compared with photons of similar energy from different atomic shells, ωK >> ωL > ωM . Electron-excited x-ray spectra also contain a significant x-ray background under all characteristic peaks. This background is the continuous x-radiation created during deceleration of the beam electron within the Coulombic field of the atoms. The
11 Developments in Instrumentation for Microanalysis in LVSEM
265
energy lost from the beam electron in one of these deceleration events is converted into an x-ray photon whose energy can take on a continuum of values from the lower threshold (arbitrarily 100 eV) up to the incident beam energy E0 . The total energy carried by an incident electron can be converted to a photon in a single event, creating the Duane-Hunt limit. The intensity of the x-ray continuum generated at a particular energy Eν follows the relation: Icm ∼ Z(E0 − Eν )/Eν
(11.3a)
Considering the continuum that creates the background directly under a characteristic peak, then Eν = Ech . Although Ech < Ec for a given element, we can approximate U = E0 /Ec as E0 / Eν . With this approximation, Equation (11.3a) can be rewritten as: Icm ∼ Z(E0 − Eν )/Eν ∼ Z(U − 1)
(11.3b)
The spectral peak-to-background ratio can then be obtained from Equations (11.2) and (11.3b): P/B = Ich /Icm = (U − 1)n /Z(U − 1) = (1/Z)(U − 1)n−1
(11.4)
For a typical value of n = 1.5, the P/B thus increases slowly as the 0.5 power of (U-1). Critical aspects of analytical x-ray spectrometry depend strongly on P/B and thus upon U. For example, our ability to detect the presence of elements at various concentration levels depends strongly upon U. For this discussion, the concentration, C, categories major, minor, and trace will be defined as: Major: C > 0.1 mass fraction Minor: 0.01 ≤ C ≤ 0.1 Trace: C < 0.01 While it is possible to detect major constituents as long as the overvoltage U > 1, providing sufficient time is spent to accumulate characteristic peak counts above the background, U = 1.1 represents a practical minimum for the acceptable overvoltage to achieve results when long-time expenditures are possible (>1,000 seconds). For a more reasonable expenditure of time (e.g., 100 seconds), in conventional analytical practice it is generally desirable to select E0 so that U ≥ 2 for the highest value of Ec in the suite of elements to be studied. For measurement of minor and trace elements, a high overvoltage is necessary to achieve adequate P/B for detectability because of the reduced exponent given by Equation (11.4). Thus, careful attention must be paid to the choice of beam energy and overvoltage when developing the analytical strategy to solve a particular problem. Ec for a given shell, e.g., the K-shell, depends upon atomic number, ranging from EK = 54.75 eV for Li, the lowest atomic number element that produces an x-ray, to EK = 115,600 eV for U. The high values of EK
266
D. E. Newbury Table 11.1 X-ray shell choices for conventional beam energy x-ray microanalysis K-shell Z = 4 Be (EK = 116 eV) to Z = 36 Kr (EK = 14323 eV) L-shell Z = 23 V (EL = 512 eV) to Z = 83 Bi (EL = 13424 eV) M-shell Z = 58 Ce (EM = 883 eV) to Z = 92 U (EM = 3551 eV)
that occur for intermediate and high atomic number elements force the analyst to choose a lower energy shell, such as the L- or M-shell, for those elements to give access to a value of Ec that can be excited with the available beam energy of the SEM (typically a maximum of E0 = 30 keV). Table 11.1 presents the elemental shell assignments chosen for conventional analytical practice to measure the naturally occurring elements with a minimum overvoltage of U = 2 (beam energy selected at the maximum of the conventional operational range 10 ≤ E0 ≤ 30 keV). Note that this selection of shells gives a comfortable range of overlap since several elements can be measured with two shells: e.g., Cu from the K and L shells and Ta from the L and M shells, if the beam energy is 20 keV or higher. Figure 11.1 depicts a Periodic Table shaded to indicate typical shell choices for conventional microanalysis conditions. The sampling volume of electron-excited x-ray spectrometry depends upon the electron range within the target. The electron range is both composition and beam energy dependent (Kanaya & Okayama 1972): R(nm) = 27.6(A/Z0.89 ρ)E0 1.67
(11.5a)
Semiconductor EDS (129 eV resolution at MnKα) E0 ≥ 20 keV
K-shell
K&L
U ≥ 2 (Ec ≤ 10 keV)
L-shell
L&M
M-shell Not detectable
H Li
He
Be
B
C
N
O
F
Ne
Na
Mg
Al
Si
P
S
Cl
Ar
K
Ca
Sc
Ti
V
Cr
Mn
Fe
Co
Ni
Cu
Zn
Ga
Ge
As
Se
Br
Kr
Rb
Sr
Y
Zr
Nb
Mo
Tc
Ru
Rh
Pd
Ag
Cd
In
Sn
Sb
Te
I
Xe
Cs
Ba
La
Hf
Ta
W
Re
Os
Ir
Pt
Au
Hg
Tl
Pb
Bi
Po
At
Rn
Fr
Ra
Ac Ce
Pr
Nd
Pm
Sm
Eu
Gd
Tb
Dy
Ho
Er
Tm
Yb
Lu
Th
Pa
U
Np
Pu
Am
Cm
Bk
Cf
Es
Fm
Md
No
Lr
Fig. 11.1 Periodic Table showing typical choices of atomic shells for operation in the beam energy range for “conventional” microanalysis, E0 = 10 keV – 30 keV
11 Developments in Instrumentation for Microanalysis in LVSEM
267
where A is the atomic weight (g/mole), Z is the atomic number, ρ is the density (g/cm3 ), and E0 is the incident beam energy (keV). Energy loss occurs along the electron range, so that calculation of the volume of x-ray production for a particular atomic species involves a correction to Equation (11.3) to reduce the range to account for the cessation of x-ray production when the beam electron loses energy down to the value of Ec : Rx (nm) = 27.6(A/Z0.89 ρ)(E0 1.67 − Ec 1.67 )
(11.5b)
Table 11.2 contains values of Rx for various atom species with K-edge ionization energies ranging from 7.11 keV (Fe) to 0.69 keV (F) present as dilute constituents (C < 0.01 mass fraction) in a carbon matrix. For the conventional beam energy range, the x-ray production range is always in excess of 1 µm except for the hardest radiation considered, FeK, at the lowest beam energy, 10 keV, where the range decreases to approximately 0.7 µm. While these ranges give a measure of the depth of penetration of the beam into the target, Monte Carlo electron trajectory simulation suggests that the lateral spread of the beam is of a similar value. A crude but useful estimate of the sampling volume is to consider that volume to have the form of a hemisphere centered on the beam impact point whose radius is the range. Of course, within this sampling volume the local density of x-ray production varies greatly with position relative to the beam impact point, but the hemispherical sampling volume gives a conservative estimate of the spatial resolution of analysis. Since we are often interested in the partitioning of an element between two structures, an estimate of the x-ray sampling volume is useful in estimating the likelihood of inadvertent sampling of a second, nearby structure while interrogating the first. Examination of Table 11.2 suggests that, for most elements of interest in biological matrices, electron-excited x-ray spectrometry is restricted to super-micrometer spatial resolution when the beam energy is selected in the conventional range, 10 – 30 keV.
X-Ray Production and Spatial Resolution in Low Beam Energy Microanalysis The strong beam energy dependence of the electron and x-ray ranges, which follow an exponent of 1.67 in Equations (11.3) and (11.4), leads to the key rationale for low voltage microanalysis: The lateral and depth spatial resolution can be greatly Table 11.2 Conventional beam energy microanalysis: total electron range and x-ray production range for various atomic species present as traces (C < 0.01 mass fraction) in a matrix of carbon (ρ = 2 g/cm3 ) E0 (keV)
Range
FeK
CaK
ClK
PK
NaK
FK
30 25 20 15 10
9.85 µm 7.26 µm 5.00 µm 3.09 µm 1.57 µm
8.96 µm 6.37 µm 4.11 µm 2.20 µm 0.68 µm
9.43 µm 6.85 µm 4.59 µm 2.68 µm 1.16 µm
9.66 µm 7.07 µm 4.81 µm 2.90 µm 1.38 µm
9.73 µm 7.14 µm 4.88 µm 2.97 µm 1.06 µm
9.81 µm 7.22 µm 4.96 µm 3.06 µm 1.53 µm
9.83 µm 7.24 µm 4.98 µm 3.08 µm 1.55 µm
268
D. E. Newbury
reduced by selecting the beam energy at or below 5 keV. The upper bound of 5 keV for the low-beam energy range is based upon the observation that this is the lowest beam energy for which a useful characteristic x-ray peak can be detected for all of the naturally occurring elements of the Periodic Table (further discussion below). The electron and x-ray ranges for the same elements in a carbon matrix as measured in the low beam energy range are listed in Table 11.3. The decrease in the various values of the range in the low beam energy regime compared to the conventional beam energy regime is quite striking. For example, the x-ray production range for ClK is 4.8 µm (4800 nm) at 20 keV and only 304 nm at 5 keV. The reduction in the sampling volume (and mass) follows the cube of the linear dimension, so that the analytical volume and mass are reduced by a factor of approximately 4000 by reducing the beam energy from 20 keV to 5 keV. Analytical electron microscopy (AEM) with x-ray spectrometry performed at high beam energy, E0 ≥100 keV, is capable of even better lateral resolution, by a factor of at least ten, in a low atomic number matrix like carbon, but AEM requires that the specimen be prepared as a thin section, 100 nm or thinner (Joy et al. 1986). Low beam energy analysis in the SEM has the great advantage of being able to work directly with the as-received sample, or at most requiring only one prepared surface so that specimen thickness is not an issue. LVSEM-microanalysis has the added advantage that large lateral areas, at least a square centimeter in size or larger, can be readily accessed with typical SEM stage movements. The development of the variable pressure/environmental SEM has further improved the compatibility of biological specimens with the vacuum environment of the SEM
Other Advantages of Low Beam Energy Microanalysis While the great improvement in spatial resolution obtained with low-beam-energy microanalysis compared to the conventional analysis strategy is the prime rationale for its utilization, there are additional positive factors. The x-ray spectrum measured under conventional beam energy conditions is subject to significant modification due to the physics of x-ray generation and propagation. The spectrum of x-rays that emerges from a thick target is substantially different from the ideal spectrum Table 11.3 Low beam energy microanalysis: total electron range and x-ray production range for various atomic species present as traces (C < 0.01 mass fraction) in a matrix of carbon (ρ = 2 g/cm3 ) E0 (keV)
Range
FeL
CaL
ClK
PK
NaK
FK
5 4 3 2 1
494 nm 340 nm 211 nm 107 nm 34 nm
475 nm 321 nm 191 nm 88 nm 15 nm∗
488 nm 334 nm 205 nm 101 nm 28 nm
304 nm 151 nm∗ nd∗∗ nd nd
374 nm 221 nm 91 nm∗ nd nd
456 nm 302 nm 172 nm 69 nm∗ nd
476 nm 322 nm 192 nm 89 nm 16 nm∗
∗ Overvoltage ∗∗ nd,
below U = 2 but above U = 1.25 not detectable due to U < 1.
11 Developments in Instrumentation for Microanalysis in LVSEM
269
generated within the target, i.e., the spectrum that only incorporates the physics of generation of characteristic x-rays and the electron-induced bremsstrahlung (braking radiation) x-rays. The measured x-ray intensity is subject to effects of electron backscattering, electron energy loss, x-ray absorption and x-ray fluorescence, which are collectively referred to as “matrix effects,” since the exact values of the effects for a given photon energy depend upon the exact composition of the target matrix. The absorption, atomic number and secondary fluorescence correction factors derived from these physical effects form the basis of quantitative electron probe microanalysis, in which the composition of the unknown target is determined relative to standards of known composition (Goldstein et al. 2003). The standards can be as simple as a pure element or a stoichiometric compound, which is especially useful for those elements that either are not solid at room temperature in a vacuum, such as chlorine, or which are unstable under electron beam bombardment, such as sulfur. Operation in the low beam energy microanalysis regime has a significant impact on reducing the magnitude of these matrix correction factors.
Minimized X-ray Absorption The as-generated spectrum (E0 = 20 keV) for a carbon target containing 0.005 mass fraction of each of the elements listed in Table 11.2 is shown in Fig. 11.2 along with the spectrum after propagation through the specimen. X-rays are subject to photoelectric absorption within the target: I/I0 = exp−(µ/ρ)ρs
(11.6)
where I0 is the x-ray intensity generated at a particular location in the target, I is the intensity that emerges at the surface after traveling along a path s, (µ/ρ) is the mass absorption coefficient, and ρ is the density (g/cm3 ). Generally, the mass absorption coefficients increase as photon energies decrease for a given matrix. Sharp increases in the mass absorption coefficient occur for photon energies just above the critical ionization energy for an atomic species, which can be seen as sharp steps in the intensity of the continuum background (bremsstrahlung) x-rays. Table 11.2 indicates that for E0 = 20 keV in a carbon matrix, the x-rays are produced in a distribution that extends to depths exceeding 4 µm, leading to the strong absorption effects seen in Figs. 11.2a and 11.2b, which show simulations of the x-ray spectrum as generated within the specimen and upon exiting the specimen after absorption. Despite the low atomic number of the sample matrix, which is predominantly carbon (97%) in this simulation, the absorption of the low energy photons is so severe that the generated and emitted x-ray intensities do not converge until a photon energy of approximately 5 keV; above this value, the spectrum is not significantly affected by x-ray absorption. In the low beam energy regime, the electron range and x-ray production ranges are much shorter, by a factor of 10 or more, and consequently, the x-ray absorption paths are reduced proportionally. Since the absorption function follows an
270
D. E. Newbury (a)
Intensity
70,000
FeL 0
0
1.0
2.0
3.0
4.0 5.0 6.0 7.0 Photon Energy (keV)
8.0
9.0 10.0
(b)
Intensity
5000
FeL
0 0
1. 0
2 .0
3. 0
4.0 5.0 6.0 7.0 Photon Energy (ke V)
8.0
9.0
10.0
Fig. 11.2 Simulation of the x-ray spectrum with an incident beam energy of 20 keV for a target containing 0.005 mass fraction each of F, Na, P, Cl, Ca, and Fe in C (balance = 0.97 mass fraction) as generated within the specimen (line trace) and upon exiting the specimen after absorption (filled): (a) vertical axis scaled to CK peak; (b) expanded vertical scale. The generated and emitted x-ray intensities do not converge until a photon energy of approximately 5 keV
exponential dependence on distance, this reduction in absorption path length has a dramatic effect on the spectrum, as can be seen in Figs. 11.3a and 11.3b, where the generated and emitted spectra are compared at E0 = 5 keV. The generated and emitted spectra converge above a photon energy of approximately 1.8 keV, so that, with respect to absorption, the measured spectrum is unaffected from 1.8 keV to 5 keV, which is a much reduced absorption situation compared to analysis under conventional beam energy conditions in Fig. 11.2. Because x-ray absorption is minimized, secondary x-ray fluorescence, which is initiated by x-ray absorption, becomes negligible in the low voltage analysis regime, so the fluorescence matrix correction factor is essentially unity.
11 Developments in Instrumentation for Microanalysis in LVSEM
271
(a)
Intensity
35000
0 0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
3.5
4.0
4.5
5.0
Photon Energy (keV) (b)
Intensity
4000
FeL
0 0
0.5
1.0
1.5
2.0
2.5
3 .0
Photon Energy (keV)
Fig. 11.3 Simulation of the x-ray spectrum with an incident beam energy of 5 keV for a target containing 0.005 mass fraction each of F, Na, P, Cl, Ca, and Fe in C (balance = 0.97 mass fraction) as generated within the specimen (line trace) and upon exiting the specimen after absorption (filled): (a) vertical axis scaled to CK peak; (b) expanded vertical scale. The generated and emitted x-ray intensities converge at a photon energy of approximately 1.8 keV
Electron Scattering Effects Electron scattering affects x-ray production and spatial resolution through two mechanisms: (1) Elastic scattering causes electrons to deviate from their initial incident trajectory, and a significant fraction undergoes sufficient collisions to escape the specimen as backscattered electrons, reducing the total possible x-ray production by carrying off energy that could have made additional x-ray events; and (2) inelastic scattering processes (in addition to the inner shell ionization that
272
D. E. Newbury
initiates x-ray emission) that collectively reduce the energy of the beam electrons. Both of these effects are strongly dependent on the beam energy and the atomic number (composition) of the target and form the basis of the “atomic number” correction in quantitative electron probe microanalysis. In the low-beam-energy regime, the effects of electron scattering become relatively less significant because of the reduced overvoltage, so that although backscattering still occurs, most of the possible x-ray production is actually created in the specimen. When LVSEM x-ray production from a particular element, e.g., Mg, dispersed in a light atomic number matrix that consists mostly of carbon, is compared to x-ray production from a standard that consists of the same element in pure form, the atomic number correction factor, which includes the ratio of the efficiencies of production in the sample and standard, is close to unity.
Table 11.4 X-ray shell choices for low beam energy x-ray microanalysis (E0 = 5 keV and U > 1.1) K-shell Z = 4 Be (EK = 116 eV) to Z = 21 Sc (EK = 4496 eV) L-shell Z = 22 Ti (EL = 454 eV) to Z = 53 I (EL = 4559 eV) M-shell Z = 54 Xe (EM = 672 eV) to Z = 92 U (EM = 3551 eV)
Semiconductor EDS (129 eV resolution at MnKα) E0 = 5 keV
K-shell
U0 ≥ 1.1(Ec ≤ 1.5 keV)
L-shell M-shell
Not detectable
H
He
L i Be
B
C
N
O
F
Ne
Al
Si
P
S
Cl
Ar
Cu
Zn Ga
Ge
As
Se
Br
Kr
Na
Mg
K
Ca
Sc
Ti
V
Cr
Rb
Sr
Y
Zr
Nb
Mo Tc
Ru Rh
Pd
Ag
Cd
In
Sn
Sb
Te
I
Xe
Cs
Ba
La
Hf
Ta
W
Re
Os
Pt
Au
Hg Tl
Pb
Bi
Po
At
Rn
Fr
R a Ac Ce
Pr
Nd
P m Sm E u
Th
Pa
U
Mn
Fe
Np
C o Ni
Ir
Pu
Gd Tb
Dy
Ho Er
T m Yb L u
Am C m Bk
Cf
E s F m M d No
Lr
Fig. 11.4 Periodic Table showing choice of atomic shells available for microanalysis in the lowbeam-energy range, E0 = 5 keV and U = 1.1
11 Developments in Instrumentation for Microanalysis in LVSEM
273
General Limitations of Low Beam Energy Microanalysis The choice of a beam energy of 5 keV or less obviously limits the span of photon energies of characteristic x-rays that can be excited. The value 5 keV is taken as the beginning of the low-beam energy range based on the observation that this is the lowest beam energy for which the elements of the entire Periodic Table (except for H and He) can be excited and detected, at least when present in the specimen as a major constituent (C > 0.1 mass fraction). Even this analytical strategy for shell choice for low-beam-energy microanalysis is substantially compromised from that listed above for the conventional beam energy range because U = 1.1 must be accepted as a minimum overvoltage. With E0 = 5 keV, the shell choices for U > 1.1 are as follows in Table 11.4, and are shown graphically in Fig. 11.4: (a)
(b)
(c)
Fig. 11.5 Appearance of K, L, and M peaks excited with low overvoltage from elements at the upper end of the shell range for low beam energy analysis: (a) scandium, U = 1.1 for Sc K-shell; (b) KI, U = 1.1 for I L-shell; U = 1.39 for K K-shell; (c) uranium, U = 1.41 for U M-shell; all at E0 = 5 keV
274
D. E. Newbury
While this strategy can, in principle, reach all naturally occurring elements, it must be realized that elements on the upper end of each range are very poorly excited since U is just above the threshold. Figure 11.5 contains examples of EDS spectra for elements at the upper end of each range with E0 = 5 keV (Sc K-shell, U = 1.11; I L-shell in KI, U = 1.10; and U M-shell, U = 1.41). The peak-to-background is seen to be low for these elements, considering that Sc and U are measured as pure elements. For the compound KI, I is present at 0.76 mass fraction and yet the I Lα peak is very low compared to the KKα, due partially to the lower overvoltage for the iodine (U = 1.39 for KKab while U = 1.097 for ILIIIab ). A second physical factor also impacts the KI spectrum. Many elements that would normally be measured with K- or L-shell x-rays in the conventional-beam-energy range must instead be analyzed with L- and M-shell x-rays in the low-beam-energy regime. For a given atomic shell, the fluorescence yield ω, which is the fraction of ionization events that yield x-rays, is a strong increasing function of photon energy and the shell. In general ωK >> ωL > ωM for photons of similar energy. The lower fluorescence yield of the L-shell compared to the K-shell has a strong effect on lowering the iodine intensity compared to the potassium intensity in the KI spectrum of Fig. 11.5b. The practical consequence of this fluorescence behavior on low voltage analysis is that the intensity that can be measured for many intermediate and high atomic number elements relative to the continuum background will be much lower than that obtained in conventional analysis. The real impact of the x-ray generation physics becomes more
Semiconductor EDS (129 eV resolution at MnKα) E0 = 2.5 keV
K-shell
U ≥ 1.1(Ec ≤ 2.25 keV)
L-shell M-shell Not detectable
H
He
L i Be Na
Mg
K
Ca
Sc
Ti
V
Cr
Rb
Sr
Y
Zr
Nb
Cs
Ba
La
Hf
Fr
Ra
Ac
Mn
B
C
N
O
F
Ne
Al
Si
P
S
C l Ar
Fe
C o Ni
Cu
Zn
Ga
Ge
As
Se
Br
Kr
Mo Tc
Ru
Rh
Pd
Ag
Cd
In
Sn
Sb
Te
I
Xe
Ta
W
Re
Os
Ir
Pt
Au
Hg
Tl
Pb
Bi
Po
At
Rn
Ce
Pr
Nd
P m Sm E u
Gd
Tb
Dy
Ho
Er
T m Yb
Lu
Th
Pa
U
A m C m Bk
Cf
E s F m M d No
Lr
Np
Pu
Fig. 11.6 Periodic Table showing choice of atomic shells available for microanalysis in the low beam energy range, E0 = 2.5 keV and U = 1.1. Note significant loss of elements that can be effectively measured
11 Developments in Instrumentation for Microanalysis in LVSEM
275
obvious as the beam energy is reduced below 5 keV, whereupon large sections of the Periodic Table become effectively inaccessible. There is simply no shell with adequate fluorescence yield available. As shown schematically in Fig. 11.6, at E0 = 2.5 keV many elements are not detectable at any concentration. If minor and trace level concentrations were considered, even more elements would have to be rated inaccessible.
Current X-ray Spectrometry Capabilities for LVSEM Microanalysis With existing technology, the analyst has two choices for performing analytical x-ray spectrometry in the low beam energy regime: (1) the semiconductor energy dispersive spectrometer (EDS) and (2) the wavelength dispersive spectrometer (WDS).
EDS The simplest, easiest to use, and by far most common form of x-ray spectrometry implemented in the SEM is the energy dispersive x-ray spectrometer, which is based upon a semiconductor detector. Usually the EDS is made of silicon (Si-EDS), although germanium detectors (Ge-EDS) are also available. The popularity of the EDS for all SEM applications, including low-beam-energy analysis, arises principally because the energy dispersive aspect of its operation permits the analyst to record the complete x-ray spectrum at every location sampled. The photon detection mechanism of the Si-EDS is based upon photoelectric absorption within the Si crystal, followed by inelastic scattering of the photoelectron, creating charge carriers proportional in number to the original photon energy. While this detection process is serial in time for single photons, it is effectively parallel in photon energy. That is, the entire energy range of photon production can be continuously monitored with the Si-EDS, from a threshold of about 100 eV to the photon energy equal the incident electron beam energy E0 , which is the upper limit (Duane-Hunt limit) of the x-ray bremsstrahlung (braking radiation) or continuum. Si-EDS does suffer from several drawbacks in its capacity to measure the x-ray spectrum. The principal limitation of Si-EDS is relatively poor energy resolution when compared to the natural peak width. Measured at a photon energy of 5890 eV (MnKα), the optimum resolution is approximately 125–130 eV (depending on detector material and size), where resolution is defined as the full peak width at half the maximum peak intensity (FWHM), which can be compared with a “natural” peak width of approximately 1.5 eV for MnKα. This substantial peak broadening is a consequence of the limited number of charge carriers (electron-hole pairs) generated by inelastic scattering of the
276
D. E. Newbury (a) 1000
Resolution, Peak Width (eV)
Large Area Si-EDS 100 High Resolution Si-EDS LiF
10 PET
TAP
Kα1–Kα2
1
0.1 0
2000
4000
6000
8000
10000
Photon energy eV
(b) 1000
Resolution, Peak Width (eV)
Large Area Si-EDS 100 High Resolution Si-EDS
10 PET
TAP 1
Kα1–Kα2 0.1 0
Fig. 11.7
500
1000 Photon energy (eV)
1500
2000
11 Developments in Instrumentation for Microanalysis in LVSEM
277
photoelectron that is ejected from a silicon atom following the absorption of an x-ray photon. The number of charge carriers (electrons or holes) liberated is approximately: ne,h = Ep /3.6 eV
(11.7)
where Ep is the photon energy. For MnKα at 5890 eV, this gives an average √ of 1636 charges. One sigma (σ for counting or Poisson statistics is equal to n, where n is the number of counts) of this number is 40.4, or about 2.47%. The FWHM of a Gaussian peak is 2.355σ, which thus predicts a value of E/E of 5.8% based upon the simple counting statistics, while the observed optimum FWHM of the Si-EDS is about 129 eV/5890 eV = 2.2%. The resolution depends on the photon energy, as plotted in Fig. 11.7 for a small cross-sectional area detector (10 mm2 , referred to as high-resolution) and for a large cross sectional area (60 mm2 , “large solid angle”) detector. Some individual values of the FWHM are listed in Table 11.5 for a small area, high-resolution (129 eV at MnKα) Si-EDS. The impact of the Si-EDS detection process upon the ideal spectrum of Fig. 11.2 is shown in Figs. 11.8a (linear intensity scale) and 11.8b (logarithmic intensity scale), where the narrow, high peaks in the generated spectrum are substantially broadened and reduced relative to the x-ray continuum background. Actual performance of Si-EDS in the low photon energy region is illustrated in Fig. 11.9, which shows the PKα and PKβ region of the spectrum of Fluorapatite. The resolution of the Si-EDS is inadequate to separate the PKα and PKβ so that a composite peak is observed. There are several negative consequences of this significant broadening that the characteristic x-ray peak undergoes due to the poor resolution of the Si-EDS measurement process:
“Lost” Elements As noted in the discussion of shell selection strategy, the analyst seeking to work in the low-beam-energy regime must deal with the low relative intensity of characteristic peaks due to the low overvoltage and the restriction to shells with low fluorescence yield. This combination of factors results in an inherently low P/B (as-generated) for many elements, and the peak broadening action of the EDS further lowers the measured P/B, all of which acts to further restrict access
Fig. 11.7 (Continued) Resolution (FWHM) versus photon energy for various types of x-ray spectrometers. (a) 0 – 10 keV; (b) 0 – 2 keV. Key: Solid lines: Si-EDS large area (150 eV at MnKα) and “high” resolution (129 eV at MnKα); dashed lines: WDS for various diffractors, LiF, TAP (thallium acid phthalate), PET (pentaerythritol); open squares, diamonds, circles: WDS with synthetic multilayer diffractors; x: first generation NIST microcalorimeter EDS, with analog processing; +: second generation NIST microcalorimeter EDS, optimized for low photon energy, with analog processing; filled triangles: second generation NIST microcalorimeter EDS, broad range version (at MnKα) and low photon energy version (at Al Kα)with optimized digital processing; pointcentered squares: Kα1 width (FWHM) for various elements
278
D. E. Newbury Table 11.5 Comparison of Si-EDS and WDS characteristics Si-EDS WDS (diffractor)
1. Resolution at MnK α(5890 eV) at CaKα (3691 eV) at PKα (2015 eV) at AlKα(1487 eV) at MgKα (1254 eV) at OK (523 eV) at CK (282 eV) 2. Peak Interference 3. Limit of detection (Mass fraction) 4. Photon energy range
5. Photon energy coverage 6. Time constant
7. Typical beam current 8.∗ Count rate E0 = 5 keV Pure element equivalent Ca5 (PO4 )3 F 9. Best application 10. Elemental Mapping ∗
129 eV
12 eV (LiF)
102 eV 83 eV 76 eV 71 eV 60 eV 63 eV Frequent 0.001 to 0.01
12.4 eV (PET) 2.4 eV (PET) 7.5 eV (TAP) 6.2 eV (TAP) 30 eV (LDE1) 11.7 eV (LDEC) Rare 10−5 to 10−4
100 eV to E0 (E0 ≤ 30 keV)
100 eV – 12 keV with 6 diffractors: LDEB, LDEC, LDE1 TAP, PET, LiF ∼resolution 1 µs
continuous entire range 50 µs (129 eV best resolution;3 kHz) 5 µs (30 kHz count rate;178 eV) 1 nA OK 1.78 × 105 c/s/nA/sr FK 1.26 × 105 c/s/nA/sr PK 8.78 × 104 c/s/nA/sr CaKα 7.65 × 103 c/s/nA/sr Qualitative analysis Major, some minor
100 nA 3.60 × 07 × 102 c/s/nA 9.07 × 102 c/s/nA Trace analysis Major, minor and trace
Note difference in dimensions
to elements. Figures 11.4 and 11.6 show the practical situation for Si-EDS at (a) E0 = 5 keV and (b) E0 = 2.5 keV considering all elements present as major constituents (C > 0.1 mass fraction). While generalizations such as those embodied in Figs. 11.4 and 11.6 can be made, the actual detectability situation must be examined for each element of interest. Consider the case of the biologically important element calcium. When the incident-beam energy is reduced below the K-edge energy of calcium, EK = 4.04 keV , the analyst must choose the Ca L-shell with ELIII = 0.349 keV . The CaLα peak at 0.341 keV is certainly within the measurable energy range of EDS, which can readily measure CK at 0.282 keV. However, the fluorescence yield is so low that CaLα is not detectable with EDS on a practical basis. Figure 11.10 shows the CaL region of a spectrum of the mineral Fluorapatite recorded at E0 = 5 keV for the extreme case of 3,000 seconds at 40% deadtime, resulting in a peak channel intensity of 510,000 counts for OK. The spectrum is also shown after multiple linear least squares peak stripping of the CK and OK. No peak structure for CaLα can be discerned in the continuum background despite the extremely high spectral statistics, which are certainly at the limit of a practical measurement dose. For EDS measurements, the CaK-shell is the only practical choice, which restricts E0 to a minimum of 4.4 keV to achieve a minimum value of U = 1.1.
11 Developments in Instrumentation for Microanalysis in LVSEM
279
(a)
(b)
Fig. 11.8 Effect of EDS detector broadening. Simulation of the x-ray spectrum with an incident beam energy of 5 keV for a target of 0.005 mass fraction each of F, Na, P, Cl, Ca, and Fe in C (0.97 mass fraction): upon exiting the specimen after absorption and after the detector broadening: (a) linear intensity scale and (b) logarithmic intensity scale
Peak Interference There are numerous practical examples of peak interference that are encountered in Si-EDS spectrometry beyond the intrafamily interference situations that occur for elements such as Na, Mg, Al, Si, and P, for which the Kα-Kβ pair is unresolved, and Cr, Mn, Fe, Co, Ni, Cu, Zn, etc. for which the L-family peaks are unresolved. Examples of significant interelement interferences include VL, CrL with OK, SK with PbM, MnL, FeL with FK, ZnL with NaK, AsL with MgK, BrL with AlK, etc. When conventional beam energy analysis is employed, it is usually possible to identify these interference situations by examining peaks arising from other shells, e.g., if PbM is suspected of interfering with SK, then the presence of lead can be confirmed by exciting PbL with a beam energy above 13 keV (PbLIII excitation
280
D. E. Newbury
Fig. 11.9 P K region of the x-ray spectrum of Fluorapatite, Ca5 (PO4 )3 F, as measured with a SiEDS (resolution 129 eV at MnKα) and with WDS (PET diffractor; resolution approximately 9 eV at a photon energy of 2 keV). Note the complete separation of the PKα and PKβ peaks with WDS, and the observation of the satellite peak on the shoulder of the PKα peak. E0 = 5 keV
edge). This possibility is lost with low beam energy microanalysis because of the maximum beam energy of 5 keV. Instead, a peak deconvolution procedure must be applied assuming that one of the interfering peaks is present, and the spectrum residuals studied for evidence of the other peak. Another practical problem with Si-EDS is the impact that the elements C and O have on the interference situation. This interference is obvious in the case of CaL shown in Fig. 11.10, but the region impacted by the C and O peak spans the low
Fig. 11.10 Fluorapatite, Ca5 (PO4 )3 F, as measured for 3,000 seconds at 40% deadtime (8 nA) and E0 = 5 keV (OK peak channel = 510,000 counts). Dose = 24,000 nA-s. Background under the peaks is shown after multiple linear least-squares peak stripping of CK and OK from references measured on pure carbon and SiO2 Note that no peak structures for CaLα and CaLl can be seen at the appropriate locations marked
11 Developments in Instrumentation for Microanalysis in LVSEM
281
Fig. 11.11 Interference of CK and OK peaks with L-shell and M-shell elements
photon energy peaks of many other elements. Figure 11.11 shows the position of the L- and M- peaks of these elements relative to CK and OK as measured with a 129-eV Si-EDS. Moreover, although the fluorescence yields of CK and OK are low (ωK = 0.0035 for C and 0.0090 for O) compared to K-shell yields of higher atomic number elements (ωK = 0.45 for Cu), the K-shell yields are much higher than the L- and M- shell fluorescence yields in this energy range (ωL = 0.00047 for K and 0.00067 for Ca; ωM = 0.00033 for Sr and 0.00066 for Ag), so that carbon and oxygen, even when present at low concentrations or as surface layers, tend to dominate this region of the spectrum.
Deteriorated Limits of Detection The mere spreading out of the characteristic photons over a wider energy range due to the Si-EDS measurement process would have no particular consequences upon detection sensitivity except for the fact that the characteristic peak spreads out over the x-ray continuum of all photon energies. The characteristic x-ray peaks are situated upon this continuum background, and any attempt at a quantitative measurement must proceed from a separation of the two spectral components so that an accurate measurement of the characteristic x-ray intensity is obtained. An estimate of the background under the peak must be subtracted from the total intensity. The natural statistical variance in this continuum background forms the eventual limit to the recognition of the characteristic peak, and thus defines the limit of detection. As the resolution becomes poorer, more background radiation is incorporated in any measurement of a characteristic peak, and therefore the variance of this background is greater, giving a poorer limit of detection.
282
D. E. Newbury
For a general estimate of the concentration limit of detection, CDL , from the spectrum of a pure element, the background NB,DL within the energy window that defines the peak integral for the desired trace element is first determined. The intensity ratio of unknown to pure element standard, or k-value, that corresponds to the detection limit is calculated for the condition that the intensity of the unknown has fallen to 3 times the standard deviation of the background counts, Iunk = 3 NB,DL 1/2 : kDL = Iunk /Istd = 3NB,DL 1/2 /NS
(11.8)
where NS = NP,S – NB,S is the pure element standard intensity, corrected for background. The corresponding CDL is then found with the appropriate ZAF matrix correction factors for the trace element(s) of interest calculated as part of the analytical procedure: CDL = kDL (ZAF)i
(11.9)
For low beam energy analysis, the ZAF factors generally converge toward unity, so that the assumption CDL = kDL is generally adequate for estimating detection. Applying this procedure to the problem of detecting elements in the spectrum of Fluorapatite measured at E0 = 5 keV shown in Fig. 11.12, the limit of detection is 0.002 mass fraction for MgK and 0.012 mass fraction for KKα, with a dose of 100s at approximately 12% deadtime (yielding an OK peak channel intensity of 4900 counts with Si-EDS performance of 135 eV resolution at MnKα). There is no peak interference for MgK and KKα in this situation. Determining the limit of detection when peak interference occurs requires deconvolution of the interfering peak. The variance due to the high peak channels increases the background variance after peak
Fig. 11.12 Fluorapatite, Ca5 (PO4 )3 F, as measured for 500 seconds at 12% deadtime and E0 = 5 keV (OK peak channel = 24,490 counts). The limit of detection for MgKα, β is calculated to be 0.002 mass fraction for this favorable case of no peak interference and high dose. For KKα, the limit of detection is 0.012 mass fraction due to lower overvoltage
11 Developments in Instrumentation for Microanalysis in LVSEM
283
removal, so that the CDL is degraded. The detection limit for potassium is poorer than that for magnesium because of the reduced overvoltage for the potassium K-shell relative to the magnesium K-shell. CDL depends on measurement conditions, particularly overvoltage, and the specimen composition, but generally for Si-EDS, CDL will be in the range 0.001 to 0.01 mass fraction unless very long spectrum measurement times are used (500s and higher). The behavior of the peak-to-continuum background as U is reduced is shown in Fig. 11.13 for a pure silicon target. From the measured P/B values and the peak counting rate of these spectra, the CDL can be calculated. Figure 11.14a shows a plot of CDL for a silicon matrix determined from these experimental measurements as a function of beam energy for the low-beam-energy regime. As the beam energy is decreased, the limit of detection moves from the trace constituent region (a)
(b)
Fig. 11.13 Appearance of a pure element K-shell peak (SiKα, β) as a function of overvoltage: (a) Linear intensity scale. Note that the spectrum for lowest overvoltage, U = 1.09 at E0 = 2 keV, does not appear to show a peak with the scaling used; (b) logarithmic intensity scale which reveals the SiKα, β peak at U = 1.09
284
D. E. Newbury (a) Silicon Overvoltage Study
5 keV 100
1
P/B
Minor
10
Trace
0.1
0.01
CMDL (100s)
P/B
Major
0.001 CMDL (100s)
1
0
5
10 U = E0 / Ec
0.0001 20
15
(b) Silicon Overvoltage Study
5 keV 1
Major
CMDL (mass fraction)
0.1
Minor
CMDL (100s) CMDL (200s) CMDL (1000s)
0.01
Trace
0.001
0.0001
0
5
10 U = E0 / Ec
15
20
11 Developments in Instrumentation for Microanalysis in LVSEM
285
(C < 0.01 mass fraction) into the minor constituent region (0.01 ≤ C ≤ 0.1). Extending the measurement time can reduce CDL , as shown in Fig. 11.14b, but other factors, such as possible alteration of the specimen due to radiation damage and instability due to instrumental drift may occur with increased electron dose and prevent the use of long dwell times.
WDS Physical Basis and Limitations Wavelength-dispersive x-ray spectrometry (WDS) historically preceded Si-EDS as the x-ray measurement technology for electron beam systems. The WDS is based upon Bragg diffraction of the x-rays, which are produced as a virtual point source at the beam impact on the specimen, which has micrometer source dimensions compared to the centimeter dimensions of the diffractometer. By establishing the proper geometric arrangement for the specimen x-ray source, the diffraction crystal, and the detector, x-ray diffraction will produce a scattering maximum for a crystal with atomic plane spacing, d, and for a particular x-ray wavelength, λ, at the Bragg angle, θB , nλ = 2d sin θB
(11.10)
where n is an integer giving the order of the reflection. For a given wavelength, the intensity generally decreases as n increases. The WDS is much sharper in energy resolution than the Si-EDS because the diffraction process changes efficiency rapidly with a small angular change in crystal orientation relative to the x-ray source (the beam impact on the specimen). The resolution depends upon the crystal chosen and the photon energy, but it is generally about 2 to 15 eV FWHM for the commonly-used diffractors needed to span the photon energy range of interest, as illustrated in Fig. 11.7. The performance of the WDS in resolving the PKα and PKβ peaks is shown in Fig. 11.9, where the resolution is sufficient to easily separate these peaks, as well as to reveal a satellite peak on the shoulder of the PKα peak. The WDS as mounted on an electron beam column consists of a high precision mechanical platform that can move the diffractor and detector along a focusing circle relative to a fixed x-ray source (beam striking the specimen) with tolerances of milliradians in angular position and micrometers in spatial position. In fact, in order to present the specimen at a location that is reproducible within micrometers, a fixed optical microscope with shallow depth of focus is usually used to define
Fig. 11.14 Limits of detection from the conventional-beam-energy range into the low-beamenergy regime: (a) P/B (measured) and CMDL for a silicon matrix for 100 s spectrum measurement at 40% deadtime as a function of overvoltage. (b) CMDL for a silicon matrix for 100 s, 200 s, and 1000 s spectrum measurement at 40% deadtime as a function of overvoltage
286
D. E. Newbury
the position of optimum WDS transmission. The specimen is moved mechanically along the z-axis (i.e., parallel to the optic axis of the SEM) to bring it into focus in the fixed optical microscope, thus consistently establishing a position with a reproducibility of a few micrometers. As an alternative, the SEM focus can be used to determine WDS focus, but because the SEM normally operates with such a large depth of focus compared to an optical microscope, the necessary precision in specimen positioning is not adequate for highly precise WDS focusing, where a defocus of 10 µm can cause a significant loss in spectrometer efficiency, depending on the exact spectrometer optical design being used. The diffraction process acts to limit the instantaneous transmission of x-rays through the spectrometer to a narrow band that is essentially a fraction of the resolution, or typically a few eV at most. Thus, the WDS is a serial spectrometer in photon wavelength (or equivalently, photon energy) in which the vast majority of the spectrum emitted from the specimen is wasted at any instant. To view the peak shape, or an extended portion of the spectrum, the WDS must be mechanically scanned by moving the diffractor and the detector on an ideal focusing circle so that Equation (11.8) is continuously satisfied for a succession of wavelengths. The geometry of the spectrometer constrains the wavelength range that can be diffracted from a crystal with a specific d-spacing. To cover the wavelength range that corresponds to x-rays with energies from 100 eV to 5 keV for low beam energy microanalysis requires at least six different diffractors, where the photon energy range of each diffractor is shown in Fig. 11.7. WDS spectrometers are typically constructed to accommodate two or four diffractors on a turret. Thus, to obtain a spectrum by WDS that covers the complete energy range, each crystal must be scanned in succession and changed, a process that is typically automated. However, the absolute efficiency of a WDS is low because of the small solid angle of collection (compared to the typical Si-EDS placement) and the losses in the diffraction process, which lead to requirements for high beam currents (>100 nA) and long dwell times per spectrometer step. Thus, a scan of the complete photon energy range of a diffractor might require 100 s to 1000 s at beam currents as high as 500 nA. Once the peaks of interest have been identified by scanning, peak intensity measurements can be performed much more rapidly by sequentially moving the spectrometer to on-peak and off-peak (background) positions and if necessary, the beam current can be reduced. As an example of the increased sensitivity of WDS, consider the problem of detecting the CaL peaks that were undetectable with EDS under high dose conditions following multiple linear least squares fitting of the interfering CK and OK peaks. Figure shows the results of a WDS scan with a synthetic layered material diffractor where the CaLα peak is just detectable above background. The improved P/B performance of the WDS is critical to detecting the CaLα peak. The
Fig. 11.15 WDS scans of Fluorapatite, Ca5 (PO4 )3 F, with LDE1 synthetic layered material diffractor; (a) Broad scan, showing the OK peak and CaL region, E0 = 5 keV, iB = 10 nA, 1 s per channel, 2000 channels. (b) Narrow scan showing the CaL region with a dose 100× greater, E0 = 5 keV, iB = 200 nA, 20 s per channel, 500 channels. Dose = 2,000,000 nA-s
11 Developments in Instrumentation for Microanalysis in LVSEM
287
(a) 140
E0 = 5 keV
120
iB = 10 nA 1 s per channel
100 Counts per second
2000 channels 80 Fluorapatite, Ca5(PO4)3F
60
40 CaLα
CaLl
20
0 100
120
140
160
180
200
Spectrometer Position (mm) (b) Apatite (CaL) 160 CaLα E0 = 5 keV iB = 200 nA
140
Counts per second
20 s per channel 500 channels
120
100
CaLl
80
60 Fluorapatite, Ca5(PO4)3F
40 150
160
170 180 Spectrometer Position (mm)
190
200
288
D. E. Newbury
dose necessary for WDS detection, however, actually exceeds that used for the failed EDS attempt shown in Fig. 11.10 by a factor of approximately 80. This is nevertheless a reasonable comparison because the WDS is capable of exploiting the high dose-rate regime while the EDS cannot. WDS can operate with very high beam current (200 nA vs. 8 nA for EDS) by using the action of the diffraction process to accept only those x-rays in the narrow region of interest and to exclude most of the x-rays which would contribute to detector deadtime and paralyze the EDS. Scanning the full CaLα peak in Fluorapatite along with the adjacent background requires such a high electron dose that, for biological specimens such scanned measurements may be impractical due to radiation damage limits. In actual quantitative analysis measurement practice, the WDS would be addressed under computer control to the peak position and then to nearby background positions to gather the necessary information to measure the peak above background, which can be accomplished with a much lower total dose. Thus, if the specimen can withstand the high beam current, the measurement of CaLα by peak jumping WDS would be a possibility. Such high beam currents are generally not consistent with high-resolution, low-voltage SEM performance, however.
Combined Si-EDS and WDS: The complementarity of WDS and Si-EDS Because of the large time penalty required to record a full WDS spectrum, it is not common practice to perform a full qualitative analysis with WDS at every specimen location analyzed. The Si-EDS, however, is extremely well suited for qualitative analysis, since it inevitably measures the entire excited x-ray spectrum at every location. This is just one aspect that illustrates the complementary nature of the Si-EDS and WDS spectrometries. As shown in Table 11.5, on many points critical to efficient and successful analysis, the strengths and weaknesses of the Si-EDS and the WDS actually complement each other. This complementarity has led to the development of an EPMA equipment configuration that combines Si-EDS and WDS capabilities, often with multiple WDS spectrometers. Such an EDS/WDS system is supported by computer-aided analysis software to optimize the data collection and processing and to combine EDS and WDS measurements. Extensive analytical procedures can be executed under completely automatic control. For example, good analytical strategy for an EPMA with both Si-EDS and WDS uses the Si-EDS spectrum for qualitative analysis at every location being analyzed, at least for major and minor constituents. The WDS can then be addressed to measure the elemental peaks that present special problems such as interference from a nearby major constituent peak, or if an element of special interest is anticipated to be near the Si-EDS limit of detection. Finally, to perform a quantitative analysis, the peak intensities measured with Si-EDS for major constituents can be combined with those measured by WDS, typically assigned to minor and trace elements.
11 Developments in Instrumentation for Microanalysis in LVSEM
289
Alternatively, for maximum data quality, the WDS can be used to measure peak intensities for all constituents identified during the EDS qualitative analysis phase. Combined EDS-WDS systems have been highly successful for electron beam x-ray microanalysis in the conventional-beam energy range where high-beam current (10 nA – 500 nA) can be readily delivered so that reasonable counting rates can be achieved with WDS. With the large beam currents necessary for reasonable WDS counting rates, it is usually necessary to withdraw the EDS to reduce the solid angle so as to reach acceptable counting rates. For low beam energy microanalysis, the EDS has a great advantage because of the very large solid angle that can be obtained by placing a large (30 mm2 or larger) detector in close proximity (e.g., 1 cm) to the beam impact on the specimen. Such a detector arrangement would yield a solid angle of 0.3 sr, which represents about 5% of the total solid angle above the target. With this collection efficiency, it is possible to obtain useful EDS spectra in reasonable integration times, e.g., 100 – 500 s, with the 100 pA to 1 nA beam current carried by the focused beam of the LVSEM. When large beam diameters can be used at the severe expense of imaging resolution, it is possible under low voltage conditions to obtain beam currents in the 10 nA – 50 nA range with the conventional tungsten thermionic source, which makes WDS practical. The SEM spatial resolution under microanalysis conditions can be improved with a higher brightness source, but such a source must also be capable of delivering high total current as well. Sources based upon lanthanum hexaboride provide an improvement of a factor of five in brightness, but recently the thermal field emission gun has been widely adopted because of its even greater brightness, total current capability and stability.
Recent Instrumentation Developments for LVSEM Microanalysis Three recent and continuing developments in x-ray spectrometry instrumentation should impact positively upon low-beam-energy microanalysis performed in the high-resolution, low beam current SEM. These developments improve the resolution, the solid angle of collection and/or the limiting count rate.
The Silicon Drift Detector: Improving Solid Angle and Counting Speed The new class of semiconductor-based energy-dispersive x-ray spectrometer known as the silicon drift detectors (SDD) employs the same x-ray detection physics as the Si(Li)-EDS, but the structure of the SDD represents a radical departure from the conventional monolithic Si(Li)-EDS design in several aspects, as shown schematically in Fig. 11.16:
290
D. E. Newbury
Silicon Drift Detector (SDD) X-rays 300 μm
SDDs are thin
SDD Backsurface SDDs have a complex back surface electrode structure.
Ring electrodes Resistor bridge
Central anode, 80 µm diameter
Area 5 mm2 to 100 mm2
The anode of an SDD is ~ 0.005 mm2 for a 50 mm2 detector, about 1/10,000 the area of EDS
Fig. 11.16 Schematic diagram of a silicon drift detector (SDD) listing main differences with conventional Si-EDS
Thickness The conventional Si-EDS is constructed from a thick crystal of silicon, typically 3 mm thick, whereas the SDD is based upon a thin silicon crystal wafer, typically 300 – 400 µm in thickness. This change in thickness reduces the collection path for deposited charge by an order of magnitude. Of course, the reduced thickness of the SDD lowers the efficiency for high energy photons compared to the conventional Si(Li)-EDS due to penetration, but for 14 keV photons, which corresponds to the upper energy limit for L-shell peaks of naturally occurring elements (ULα = 13.6 keV), the SDD (300 µm thick wafer) still retains about 50% efficiency, which is adequate for practical spectrometric applications, and thicker wafer detectors (400 µm) to extend the efficiency range are possible.
Applied Potential Distribution The conventional Si-EDS has uniform electrodes on the entrance and exit surfaces, while on the exit surface, the SDD has a complex pattern of nested electrode rings with a resistor bridge across the rings that permits application of a stepped potential distribution. This applied potential creates a lateral as well as a transverse field pattern that produces a collection channel through the wafer thickness, which is tilted toward the central collection anode. Thus, charge deposited because of the capture of a photon anywhere in the cross sectional area of the SDD, which is typically 50 mm2 and can be 100 mm2 , is efficiently transported to the very small central anode.
11 Developments in Instrumentation for Microanalysis in LVSEM
291
Electrodes The Si(Li)-EDS anode occupies the entire surface of the detector, while the SDD anode is a disk less than 100 µm in diameter, about 1/10000 the area of the anode of a 50 mm2 detector. This greatly reduces the noise contribution that results from the anode-silicon interface. With this noise term effectively eliminated from the detector noise budget, the SDD can be operated at higher temperature than the Si(Li)-EDS.
Operating Temperature The Si(Li)-EDS is usually operated with liquid nitrogen cooling, while the SDD operating temperature is about –20 ◦ C to –40 ◦ C, which can be achieved with Peltier electrical cooling. An airflow or water-cooling system exhausts the heat from the Peltier cooler. The reduced cooling demands and the relative simplicity of the cooling system make it easier to accommodate the SDD on electron beam instruments, and novel x-ray detector configurations become possible.
Resolution at MnKα For an equivalent detector area, the SDD has been demonstrated to actually achieve superior resolution: a resolution of 127 eV for an SDD at an operating temperature of −20 ◦ C compared to 129 eV for Si-EDS at an operating temperature of −190 ◦ C for equivalent area, 10 mm2 detectors. For a 50 mm2 detector, a resolution of 134 eV can be obtained, compared to approximately 142 eV for Si(Li).
Detector Counting Rate For an equivalent detector resolution, the SDD is capable of higher counting rates, by a factor of five to ten, when short time constants are used. A limiting output count rate greater than 500 kHz has been demonstrated. Measured input count rate vs output count rate response at several time constants for a 50 mm2 detector is shown in Fig. 11.17.
Low-energy Photon Detection The entrance electrode of the SDD can be made extremely thin, which increases the detection efficiency for low-energy x-rays. A spectrum for manganese (MnLα = 0.636 keV) is shown in Fig. 11.18a, demonstrating the high MnLα/MnKα ratio that can be achieved. Note that the high MnLα/MnKα ratio is not achieved by losing MnKα x-rays due to penetration through the thin SDD. At the energy of MnKα, 5.89 keV, the efficiency of the SDD is unity. Penetration through the SDD first
292
D. E. Newbury
Fig. 11.17 Output count rate and deadtime measured as a function of beam current on an Mn target at E0 = 20 keV for a silicon drift detector operating with several peaking time constants: 8 µs (resolution 134 eV at MnKα), 1 µs (163 eV), 500 ns (188 eV), and 250 ns (217 eV)
begins at approximately 8 keV. SDD spectra of the transition elements from Mn to Cu are superimposed in Fig. 11.18b.
Detector Area and Detector Arrays Individual SDD detectors can readily be made with a large active area. SDD detectors with an area of 50 mm2 are routinely produced, and detectors as large as 100 mm2 have been demonstrated. Moreover, arrays of detectors that can occupy very large solid angles, approaching π steradians, have been successfully assembled and operated to merge the individual detector spectra into a single output spectrum. For application of the SDD to low-beam-energy microanalysis, properties (4), (5), (7) and (8) (as listed in Table 11.5) are most attractive. The restriction to low-beam currents that results under high spatial resolution, low-beam-energy microanalysis operation means that the high counting speed of the SDD is not likely to be tested. However, the capability to create individual SDD detectors with large areas and SDD detector arrays should be very helpful. Because the SDDs can be cooled much more simply than Si-EDS, it should be possible to accommodate the SDD in close proximity to the specimen, which when combined with the
11 Developments in Instrumentation for Microanalysis in LVSEM
293
Cu
15000
SDD TC = 4 µs E0 = 20 keV
Intensity (counts)
Ni Co
Cr Mn Fe
Fe Mn
Co Ni Cu
Cr
0 0
1.0
2.0
3.0
4.0 5.0 6.0 7.0 Photon Energy (keV)
8.0
9.0
10.0
Fig. 11.18 (a) Silicon drift detector spectrum of manganese at E0 = 15 keV and a 4 µs peaking time constant showing excellent sensitivity for the MnL peak family. (b) SDD spectra for Cr, Mn, Fe, Co. Ni, and Cu showing L- and K-families
large detector area, will yield a solid angle of x-ray collection that is much larger than the conventional Si-EDS. Since the total dose delivered in low-beam-energy microanalysis is usually small compared to conventional beam conditions, more efficient collection of the x-rays with a larger solid angle detector will improve all aspects of analytical x-ray spectrometry, especially detection sensitivity. For those specimens that can sustain higher beam currents, the SDD makes it possible to capture x-ray spectrum images in which the complete energy-dispersive x-ray
294
D. E. Newbury
spectrum is recorded at each pixel of a scan. Such a procedure captures all possible elemental information about the specimen region within the performance limitations imposed by the electron beam dose and the spectrometer characteristics.
X-ray Optics-Augmented WDS Improvements to WDS: optic-enhanced WDS As noted above, the WDS has the spectral resolution to solve most problems related to peak interference and detection of peaks with low P/B that are encountered in either conventional or low-beam-energy microanalysis. Unfortunately, the inherent inefficiency of the WDS severely limits operation with the low beam currents available with the LVSEM. To increase the efficiency, a new class of WDS spectrometer has been demonstrated that incorporates polycapillary x-ray optics to substantially increase the solid angle of collection. X-ray steering is achieved by means of the high degree of reflection of x-rays that approach a smooth glass surface below the critical angle (typically milliradians) for scattering. A large area to increase the scattering below the critical angle is created through the use of polycapillary optics (Kumakhov 1990, 1998). Each channel of the optic is a single thin-walled glass capillary, many of which are bundled to form close-packed structures and then heated to bond the polycapillaries. The uniform bundle can be further heated and drawn to create a tapered optic that can, for example, collect x-rays over a large solid angle and then gradually bring those x-rays into a long parallel section for transport and presentation to the next optical component. Because the solid angle for reflection is inversely proportional to photon energy, polycapillary optics are most efficient for low energy photons. To increase the effective solid angle of WDS (geometric collection limits), a two-zone polycapillary optic has been incorporated to more efficiently couple the x-ray source at the specimen to the diffractor. As shown schematically in Fig. 11.19, the first zone of the polycapillary is a conical section that couples the x-ray emission collected over a large solid angle to the second section that parallelizes and transports the beam to a flat crystal diffractor, which is followed by a large detector. An example of a spectrum of MnF2 , also containing oxygen, is shown in Fig. 11.20, where the MnLα (637 eV) is well separated from the FK (677 eV) and the OK (525 eV). Various performance measures of the opticaugmented WDS for several elements measured with K-shell x-rays are listed in Table 11.6. The rapid developments in positionally-sensitive detectors (PSD) of high spatial resolution may make possible another interesting variation of the WDS. Charles Fiori et al. (1991) proposed a scheme to make use of an aspect of the focusing properties of the WDS to simultaneously detect a sufficiently wide photon energy range E to permit imaging a characteristic x-ray peak and the adjacent background. Their proposal noted that in a conventional focusing WDS, the geometry requires the x-ray source (beam-excited region of the specimen with micrometerdimensions), the diffractor and the detector to be placed on the focusing circle
11 Developments in Instrumentation for Microanalysis in LVSEM
295
Fig. 11.19 Schematic diagram of a polycapillary optic coupling the x-ray source to a flat diffractor in a wavelength dispersive x-ray spectrometer. In an accurate representation, the capillaries in the conical section would be tapered to a convergence at the source
(Rowland circle), so that convergence occurs for x-rays of a very narrow energy band at the detector. By moving the source and detector off the focusing circle, a much broader range of photon energies, with each increment dE diffracting from a different location on the diffractor, is brought to the focus at the conjugate point. By intercepting these converging rays before the conjugate point with a planar
Fig. 11.20 Spectrum of MnF2 measured with an optic-enhanced WDS. (Example courtesy Parallax Research, Inc.)
296
D. E. Newbury
Table 11.6 Measured Performance of an Optic-Augmented WDS (Parallax Research, Inc. LEXS) Measured LEXS Performance Element
Energy(eV)
Counts/sec/nA
P/B
Resolution (eV)
Sensitivity(ppm)
Be B C N(BN) O(SiO2 ) Mg Al Si
108 183 277 392 525 1254 1487 1740
350 3500 5750 416 375 600 500 400
40 50 >100 40 80 400 300 300
8 18 18.6 16 17 14 19 24
100 30 14 130 60 18 25 25
detector, a range of photon energy, E, could be simultaneously imaged on a linear detector device, although with reduced efficiency because only a small portion of the diffractor satisfies the Bragg equation for any paw photon energy. Being able to image the full width of an x-ray peak would be of special value for low photon energy peaks (Ech < 2 keV) because these peaks are subject to “chemical effects” from the differing energy levels of electrons involved in bonding. The peak position and shape can be strongly affected by chemical effects, making it necessary in conventional WDS practice to scan and integrate the peak to obtain an accurate measure of the intensity. This procedure necessarily incurs a large time penalty and inefficiency due to the loss of x-rays that are not diffracted.
Microcalorimetry The x-ray microcalorimeter is a radically different approach to x-ray spectrometry that is especially promising for LVSEM/microanalysis Wollman er al., (1997, 2000). The detection physics process of the microcalorimeter consists of measuring the temperature rise when a single x-ray photon is absorbed in a metal target, as illustrated schematically in Fig. 11.21a. Maximizing the response to the capture of a photon and minimizing the time needed to return to the baseline state both depend on minimizing the heat capacity of the detector, which depends on physical parameters such as detector volume and material parameters such as the specific heat. By operating at approximately 100 mK, a suitably low value of the heat capacity can be obtained with a metal absorber such as gold or bismuth. The temperature rise of the absorber can be measured by several different techniques, of which a leading example is a circuit incorporating a transition edge sensor (TES), illustrated schematically in Fig. 11.21b. The TES is a binary metal foil (e.g., Cu-Mo) whose layer thicknesses can be manipulated to adjust the superconducting transition temperature. The measurement circuit consists of a current source, an inductor, the metal (e.g., Bi) x-ray absorber, and the TES, arranged so that heat from the absorber must pass through the TES to reach a low-temperature reservoir. When this circuit is cooled by an external refrigeration circuit, the overall circuit resistance falls, and Joule heating I2 /R increases. When the TES reaches its superconducting
11 Developments in Instrumentation for Microanalysis in LVSEM
297
(a) X-ray Temperature
Thermometer ΔE
C
G Thermal Conductance
Heat Capacity
τ=
∆E C
C G
Time
(b) Electrothermal Feedback Transition-Edge Sensor Microcalorimeter I V V
SQUID
X-ray absorber
Electrical Circuit
2
Pjoule V R R
Thermal conductance
TES Thermometer Psink
Stable equilibrium V2 = Psink R
Thermal Circuit
Heat Sink
Fig. 11.21 (a) Principle of operation of the x-ray microcalorimeter, showing the capture and the time history of the resulting temperature pulse in the metal absorber. (b) Electrical and thermal circuit for a transition edge sensor (TES) temperature measuring system for the x-ray microcalorimeter
temperature the circuit establishes electrical and thermal equilibrium. When an x-ray is now absorbed by this quiescent circuit, the extra heat deposited is automatically compensated by a decrease in the internal current flowing in the circuit to balance the Joule heating. The changing current induces a changing magnetic field in the inductor, which is measured by a superconducting quantum interference device (SQUID). The energy of the x-ray photon is thus measured as the time integral of the magnetic field. The microcalorimeter measurement process is inherently energy dispersive. The resolution performance of the microcalorimeter EDS as a function of photon energy is shown in Fig. 11.7, where it is compared with Si-EDS and WDS with various diffractors. The demonstrated resolution of the microcalorimeter detector is in the range 2 to 15 eV, which makes it comparable to WDS over most of the photon energy range used for analysis. For photon energies below approximately 700 eV, the microcalorimeter resolution is actually substantially better than the resolution achieved with WDS using the layered synthetic material (LSM) diffractors. Like the Si-EDS, the microcalorimeter EDS process is subject to paralyzable deadtime with a limiting count rate between 500/s to 1,000/s.
298
D. E. Newbury
Fig. 11.22 Microcalorimeter EDS x-ray spectrum of PbS (galena) showing separation of the SKα from the PbMα and Mβ peaks
The microcalorimeter thus combines the energy dispersive character of the semiconductor EDS with the energy resolution of the EDS. An example of the microcalorimeter spectrum of PbS, where the SK and PbM shell x-rays interfere with conventional Si-EDS, is shown in Fig. 11.22. The performance characteristics of the microcalorimeter, listed in comparison with WDS and Si-EDS in Table 11.5, make it especially attractive for low energy x-ray microanalysis in the LVSEM. The full analytical x-ray range can be continuously monitored and the entire spectrum excited in the LVSEM is continuously available. Figures 11.23a and b show an LVSEM microanalysis example of the extreme interference case of NK, OK and TiL, where despite the high-resolution of the microcalorimeter, interference still occurs between NK and TiL, so that the contributions of each element must be determined from a series of measurements on TiN, Ti metal, and Ti oxide. A strong chemical effect on the shape of the TiLα peak is observed in the TiN spectrum in Fig. 11.23a when compared to Ti metal or the Ti-oxide spectra, Fig. 11.23b. The principal limitations are the small physical size of the detector, about 0.25 mm2 , and need to cool the detector and cryoelectronics near absolute zero. The small size of the microcalorimeter detector can be compensated by the use of a polycapillary x-ray optic to couple the electron-excited x-ray source to the detector. An improvement in the detector solid angle by a factor of 300 has been demonstrated with a polycapillary optic consisting of a divergent section at the x-ray source and a convergent section at the detector. For the first commercial presentation of an x-ray microcalorimeter intended for use on an SEM, the necessary cooling system has been realized as a combination of a liquid helium refrigerator to 4 K and an adiabatic demagnetization refrigerator from 4 K to 50 mK (low temperature reservoir), with the detector operating at 100 mK. While the microcalorimeter-EDS is in its initial phase of practical investigation as an x-ray spectrometry tool for the LVSEM, the future of this technique is
11 Developments in Instrumentation for Microanalysis in LVSEM
299
(a) 300 µcal EDS
N Kα
Ti
Ti Lα1,2
TiN
Counts
200
Ti Lβ3,4 100
Ti Lβ1
Ti Ll Ti Lη
0 400
440 480 Energy (eV)
520
(b) 300
µcal EDS Ti TiOx
Ti Lα1,2
O Kα
Counts
200
100
Ti Ll Ti Lη
Ti Lβ1
Ti Lβ3,4
0 400
440 480 Energy (eV)
520
Fig. 11.23 X-ray microcalorimeter spectra of (a) titanium metal (scraped under inert gas) and TiN and (b) titanium metal (scraped under inert gas) and titanium oxide. Note the interference of TiLl and N K despite high-resolution (2 eV FWHM at AlKα). Also, note the severe shape change in the TiLa peak in the TiN compared with the Ti metal and the titanium oxide
extremely promising. In particular, the possibility exists that an array (n x m) of microcalorimeter detectors can be made. In this way, the limitations imposed by the small size of the individual microcalorimeter detectors can be overcome by effectively increasing the total detector area to n∗ m, and the total output count rate, less than 1 kHz with a single detector, can be increased by the same factor to the range of 10 kHz to 100 kHz, and perhaps even higher. Such a detector would equal or substantially improve upon the resolution of a conventional WDS while providing the energy dispersive function and a count rate similar to the conventional WDS for a single energy channel.
300
D. E. Newbury
An Application of Low Voltage Microanalysis: Particles Particle analysis provides an example of broad interest in the biological and physical sciences as well as in engineering fields where low voltage microanalysis can significantly benefit by incorporating new spectrum measurement technology. Under conventional-beam energy conditions, the x-ray microanalysis of microscopic particles becomes especially difficult when the particle dimensions approach the electron and/or x-ray ranges, i.e., linear dimensions below 10 micrometers (Goldstein et al. 2003). Electrons can penetrate through such particles, directly affecting the x-ray generation function for the particle but also indirectly affecting the measured x-ray spectrum through the remote generation of x-rays from the substrate and nearby features. Further, x-ray absorption can be profoundly modified by particle geometry effects, especially for low-energy photons. High energy photons generated under conventional beam energy conditions, especially the x-ray continuum, create secondary fluorescence almost entirely outside the particle dimensions, and this contributes to the measured spectrum, which can be an important problem when minor and trace-level constituents are important. Theoretical studies as well as careful experimentation have established the value of low-voltage microanalysis for application to particle studies. As illustrated in Figs. 11.24 and 11.25 from the work of Small (2002), the sharp reduction in the size of the interaction volume with the reduction in beam energy makes it possible to treat particles much more like bulk targets. Under LVSEM conditions, even simple normalization of the analytical total can produce quantitative results with substantially reduced relative error distributions compared to analysis in the conventional beam energy regime. Moreover, not only can the direct excitation at low-beam energy be constrained to occur within the particle dimensions, but electrons that escape after scattering within the particle have such reduced overvoltage that remote excitation is also diminished. Finally, because of lower photon energy, the continuum and characteristic x-ray photons produced at low-beam energy are less likely to escape the particle and cause remote secondary fluorescence. These are real advantages, but the limitations imposed on x-ray measurements by the physics
(a)
(b)
Fig. 11.24 Monte Carlo plots for the interaction of a 20 kV electron beam with a 2 µm K-411 particle. (a) Electron trajectories. (b) Mg Kα x-ray generation. (from Small 2002)
11 Developments in Instrumentation for Microanalysis in LVSEM
301
Fig. 11.25 Monte Carlo plots for the interaction of a 5 kV electron beam with a 2 µm K-411 particle.( a) Electron trajectories. (b) Mg Kα x-ray generation (from Small 2002)
of low beam energy excitation, as well as the measurement limitations of EDS (resolution) and conventional WDS (efficiency), conspire to restrict practical utility. Considering the instrumentation limitations, the microcalorimeter EDS has demonstrated particular utility for LVSEM particle analysis (Wollman et al. 2000). Figures 11.26 and 11.27 show examples of nanoscale particles on a silicon substrate excited with a beam energy of 5 keV. Although even under LVSEM conditions there is scattering into the silicon substrate, the resolution of the microcalorimeter EDS is sufficient to separate the Al K-family and the W M-family from the Si K-family contributions from the substrate, as shown in Fig. 11.26. In the case of the titanium
Fig. 11.26 Microcalorimeter EDS spectrum of a 300 nm particle of tungsten on a silicon substrate. Note detection of W M family and separation from SiKα. E0 = 5 keV (Wollman et al., 2000)
302
D. E. Newbury
microcalorimeter EDS
Counts
150
OKα
0.3 µm TiO2 particle on Si 1.8 keV beam voltage 0.53 nA beam current 400 s live time 406 s real time
100
30 s–1 input count rate 29 s–1 output count rate 1% dead time
50
CKα Ti Lα,β Ti L,η
Ti Lβ3,4
0 200
300
400 Energy (eV)
500
600
Fig. 11.27 Microcalorimeter EDS spectrum of a 300 nm particle of titanium on a silicon substrate. Note detection of TiL family and separation from OK. E0 = 5 keV
particle in Fig. 11.27, the resolution of the microcalorimeter EDS is sufficient to permit detection of the Ti L-family peaks despite their low yield (ωL = 0.0016) from the much more intense OK peak (ωK = 0.0090). These examples illustrate the utility of high performance spectrometry for the solution of real problems.
Summary Low voltage SEM/microanalysis can be a powerful tool for the biological microanalyst by providing improved spatial resolution and reduced matrix effects on the measured x-ray intensities. There are inevitable limits imposed by the physics of x-ray generation with low beam energy upon the elemental coverage and detection sensitivity that can be achieved. Optimal performance of the x-ray spectrometer is critical to achieving the best results. For most situations, the conventional EDS is the most effective choice because of its large solid angle, which helps to compensate for the low intensity of the low-voltage electron beam. When the resolution of conventional EDS is inadequate, WDS can solve the spectrometry problem, but the inherent inefficiency of WDS generally limits its application because of the need for high beam current. Augmenting the WDS with x-ray capillary optics can improve the collection efficiency to permit effective operation with lower beam current. The development of the SDD should permit improved geometric collection efficiency over the conventional EDS and a factor of ten greater count rate for the same resolution. The microcalorimeter EDS is the most promising spectrometry technology for
11 Developments in Instrumentation for Microanalysis in LVSEM
303
the future, combining as it does WDS resolution with energy-dispersive operation. An array of microcalorimeters with large collection angle and an output count rate of 10 kHz to 100 kHz would be extraordinarily powerful for all aspects of biological microanalysis under LVSEM operating conditions.
References Echlin P (ed) (1984) Analysis of biological and organic surfaces. Wiley, New York Echlin P (ed) (1992) Low temperature microscopy and analysis. Plenum, New York Echlin P (2002) Low voltage energy dispersive quantitative X-ray microanalysis of inorganic light elements in bulk frozen hydrated biological specimens. Microsc & Microanalysis 8:120 Echlin P and Galle P (eds) (1976) Biological microanalysis. Societe Francaise de Microscopie Electronique, Paris Goldstein JI et al (2003) Scanning electron microscopy and x-ray microanalysis, 3rd edition. Kluwer, New York Joy DC et al (eds) (1986) Principles of analytical electron microscopy. Plenum, New York Kanaya K, Okayama S (1972) J Phys D: Apll Phys, Penetration and energy-loss theory of electrons in solid targets, 5:43–58 Kumakhov MA (1990) Nucl Instrum Methods Phys Res B, Channeling of photons and new X-ray optics, 48:283–286 Kumakhov MA (1998) Adv X-ray Anal, Development of X-ray Polycapillary Optics, 41:214 Small JA (2002) J Res NIST, The Analysis of Particles at Low Accelerating Voltages (< 10kV) With Energy Dispersive X-Ray Spectroscopy (EDS), 107:555–566 Wollman DA et al (1997) J M, High-resolution, energy-dispersive microcalorimeter spectrometer for X-ray microanalysis, J. Micros 188:196–223 Wollman DA et al (2000) Microcalorimeter energy-dispersive spectrometry using a low voltage scanning electron microscope. J Microscopy 199:37–44
Index
Aberrations, electron-optical, 30, 59, 110–112, 122 chromatic, see Chromatic aberration interactions, 59, 116 higher order, 120–122 reduction by use of a retarding lens, 46, 53 spherical, see Spherical aberration Aberration-corrected SEM (ACSEM), 53, 107–108, 172 aberration correctors, 53, 115–116, 120–127 commercial systems, 115 cost, 126 chromatic aberration, 110–112 depth of field, 113–115 depth-of-field, limited, 123–125 diffraction, 112–113 optical transfer function (OTF), 121–122 optimizing spot size, 113–115 options in, 116–120 problems associated with, 123 results, 121 spherical aberration, 109–110 Aberration correctors, 53, 115–116 ACSEM, 120–122 alignment of, 115, 124–125 options in, 116–120 Ronchigram adjustment, 124 ACSEM, see Aberration-corrected SEM (ACSEM) Actin filaments colloidal gold labeling of, 192 critical-point drying, 60, 148–149 cryo-preparation of, 41, 216–218, 247, 258, 263 images of, 42, 149, 157, 162–163, 218, 234–235, 237, 239, 258 insect flight muscle, 158–160 plant cells, 233–235, 237, 239
sub-membranous network, 150–152 Toxoplasma gondii, 147–149 Aldehyde fixation, 59, 80–81, 146, 217–219 cytoskeleton, 148–149, 237 insect flight muscle, 158 human platelets, 183, 221–223 maceration, in plant tissue, 233 microorganisms, 222–224 neutralization with glycine, 191 nuclear-pore complex, 80–81, 152 plants, 230–231 post fixation, 236 protocols, 146, 149, 152, 161, 183 stabilization prior to cryofixation, 217–218, 221, 260 Toxoplasma Gondii, 148, 163 yeast, 218–219, 261 See also Glutaraldehyde Alignment, electron-optical, 48, 57, 92 aberration correctors, 115, 124–125 stereo imaging SEM, 92 stigmator, 108 Ampex video recorder, 202 Anti-contaminator, 57–58 Antibody labeling, 38, 66, 171–174, 177, 183, 187 See also Colloidal-metal labeling, and Correlative LM/LVSEM/TEM conjugation, 177–178 Fab fragments, 188–189 fluorescent metal particles, 177 methods, 190–191 plants, 236 Apicomplexan parasites, 148 Arabidopsis thaliana, 239–241 Artifacts, 37, 60, 147, 218, 250 charging, 62–64, 66, 71, 76, 84, 197, 247 critical-point drying (CPD), 147, 157, 254 decoration, 66, 257 extraction, 183
305
306 Artifacts (cont.) freeze drying, 216 freezing, 82, 224, 226, 230, 249–253 incomplete dehydration, 59, 147, 149–150, 157, 161, 178, 181, 184, 216, 218, 226 microtrabeculae, 60, 184, 226 membrane extraction, 183 replicas, 3 specimen preparation, 60, 147 stabilization, see Stabilization Astigmatism, 13, 108, 115, 125 of aberration correctors, 115, 125 higher order, 108, 125 in real-time stereo imaging, 91 Astigmator, alignment, 57, 83, 91, 115 Auger emission (AE), 29, 268 Background Bremsstrahlung, 273, 279, 268–269, 278 control, action of, 135 effect of, 142 in energy-filtering TEM, Color plate, 3 in environmental SEM, 35 from subsurface scattering, 11, 41–44, 66, 70 in x-ray microanalysis, 268–269, 273, 278–279, 281–287, 290, 292, 298 Backscattered electron detectors, 28, 48–51, 54, 222, 224, 236, 257, 262 conductive coating, 49, 51 detector quantum efficiency, 141, 255–260, 262 measurement, 141 performance, 142 shadowing effect, 49 Backscattered electrons (BSE), 11, 13, 28–31, 37–39, 42–43, 46–52, 54–55, 61–62, 64, 66, 69–70, 72, 83–86, 131, 141, 171–172, 182, 222, 236, 255, 260, Color plate, 13 atomic-number contrast, 15–16, 30, 49, 171, 247, 259 coefficient, 62, 141 collection contrast, 11, 49 density contrast, 30, 49, 171, 231, 259 detector, 49, 50–51, 54, 222, 224, 236, 257, 262 detector quantum efficiency (DQE), 139, 141, 255–260, 262 double-layer coating, 221–223, 225, 247, 259, 261–263
Index imaging, 11, 13–14, 37–39, 49, 51–52, 70, 83–86, 221–223, 225, 236, 247, 253, 257, 259–261, Color plate 13 noise, 131, 260 range, 42 reduces charging, 83–86 reduces S/N, 131 reduces spatial resolution, 11 reduces x-ray signal, 275 signal, 15 signal/noise, 131 statistics, 131 uncoated specimens, 51–52, 257 Bacteria, 75, 77, 216, 222–226 Bain, Alexander, 6 Bandwidth, detector, 6, 15, 46, 49, 50, 135 secondary electron, 6, 15, 46, 50 backscattered electron, 49, 135 phosphor limit, 15 Barrel distortion, of electron lenses, 124 Beam-induced, conductivity, 63 Beam-induced surface contamination, 37, 57 Biological specimens, see Specimens, biological Boric-acid/borate buffer, 204, 207 BSE, see Backscattered electrons (BSE) Buffers, 80, 146–150, 183, 189, 239 boric-acid/borate, 204, 207 cacodylate, 261 cold extraction, 155 cytoskeleton-preserving, 147, 149, 151, 183, 236–237, 243 low-salt, 80, 152, 153 PBS, 238, 239 plant cells, 231, 233 PIPES/HEPES, 146, 158, 161, 183 pre-cryo buffers, 216–218, 221–222, 226 Cacodylate buffer, 261 Cambridge SEMs, 12–15 Carbon replica specimens, 3, 7, 67, 87, 161, 216–217, 247, 251, 254–255, 259 Cathode-ray tube, 4, 6, 13, 27–28, 47 Cathodoluminescence, 15, 21, 37, 138 Cellular interactions, 146, 151–152, 161–165 Centriole-centrosome complex, 155 Centrosomes, 153–155 Charged particle beams, 2, 29 Charging, insulating specimen artifacts, 19, 31, 33–34, 61, 63–64, 146, 171, 245, 257, 259, 261 avoidance, 62, 71 beam-induced, conductivity, 63 of BSE detector, 49, 51 causes, 29, 31, 61–64, 66
Index coating, see Coating of column, worse at low-V0 , 46 distortion produced by, 32, 83–84, 198 examples, 63–64, 84 reduced in BSE imaging, 84, 86, 247, 257 reduced at low-V0 , 45, 129, 146, 171, 263 scan-speed effects, 198–199 specimen mounting, 71 theory, 61–64, 66 uncoated specimens, 33, 45, 64–65, 83–84, 171–172, 176, 179, 255, 257, 260, Color plate 3 when metal coat disrupted, 197 voltage contrast, 16 Chemical dehydration, 215 Chemical fixation, 217 cytoskeleton, 148–149, 237 glutaraldehyde, see Glutaraldehyde of human platelets, 183, 221–223 of insect flight muscle, 158 maceration, in plant tissue, 233 of microorganisms, 222–224 neutralization with glycine, 191 of nuclear-pore complex, 80–81, 152 osmium, see Osmuim-tetroxide, (OsO4 ) of plants, 230–231 post fixation, 236 stabilization prior to cryofixation, 217–218, 222 of Toxoplasma Gondii, 148, 163 of yeast, 218–219, 261 See also Fixation Chisels, dental, for cutting in SEM, 203–204, 206 Chloroplasts, 229, 232–233, 235, 237, 238, Color plate 13 Chromatic aberration, 30, 110–112, 118 coefficient, 30, 111, 118 correction, 110–112 energy spread, 46, 53, 56, 59, 110, 117–118 Coating, 44–45, 147, 61–74, 147, 155, 157–158, 173, 191, 216, 240, 249–250, 256, 259–260 anti-reflecting, 50 artifacts, 66, 69, 257 carbon, 39, 57, 65, 172, 174, 172, 221–223, 236, 247, 251, 254, 259 charging, 66, 73 chromium, 65–66, 218, 255 of colloidal particles, 172, 182 conductivity, 69, 199, 259–261 contrast, 64–65, 256–257 cryo-coating, 41, 83, 216, 218, 221, 253, 256–260, 263 cryosputter, 253, 260
307 decoration, 66, 257 double-layer coating, 221–223, 226, 249–250, 256, 259–260 effective thickness, 64 glow-discharge, 68–69 ideal, 259 ion-beam sputtering, 59, 68, 75, 79, 88–89, 147, 162, 191, 226 limitations caused by, 59, 61–74 for LVSEM, 61–74, 155, 173 oxide formation, 66 Penning sputtering, 65 of plants, 231, 236, 241 platinum (Pt), 63–64, 68, 75, 79–80, 149, 162, 216, 240 protocols, 68, 157, 261 reduces beam penetration, 45 of scintillator detectors, 49, 51 structure, 62, 65, 70, 79 tantalum/tungsten (Ta/W), 66, 222 for TEM, 191 theory, 61–73, 259–260 thickness, effective, 64, 260 topographic-Z contrast, 64 tungsten, 66, 68, 247, 258–260, 263 vacuum required, 69 Colloidal-metal labeling, 51–52, 171–193 antibody, see Antibody labeling for correlative imaging, 38, 66, 171–174, 183, 187 double labeling, 51, 177, 191 EF-TEM for elemental analysis, 190–193 epitope density, 173–174, 178, 182 fabricating different shapes, 188 faceted cPd particle, 174–175, 188, 192 generation of, 184–185 gold labels, 39, 50–51, 155, 172, 180, 223, 236, 242 ionic double layer, 184, 189 lectins, 38, 51, 175 with LM of living cells, 179, 181–183, 190 particle size, 185–188 plant specimen, 236 popcorn cPd particles, 174–175, 188, Color plate 1 stabilizing colloids, 174, 176, 189 tannic acid, 183, 185–186 using different metals, 186–188 Composite rich image, 262 Conductive staining, 231, 234 Conductivity, electrical, specimen, 31, 39, 44, 49–50, 51, 61–66, 69, 199, 247, 259 beam-induced, 62–63 conductive staining, 60, 204–206, 234 See also Charging and Coating
308 Conductivity, thermal, ice, 252 Contamination, 7, 34, 57, 69, 71, 82 avoidance by double-layer coating, 222, 254, 259 beam-induced, 37, 57 of column, worse at low-V0 , 46 reduces contrast, 140 reduction using cryo, 38, 52 surface-diffusion, 37–38, 57, 67, 71 worse at low V0 , 15, 46, 71, 73 theory, 37 vacuum modifications, 4, 58 Continuum x-rays, 29, 269, 273, 278–279, 282–285, 300 peak-to-continuum, 287 Contrast, image, 44, 136–139, 172, 236, 247, 251, 256–257, 262 adjustment, 134–135, 190 affected by V0 , 33–5 atomic number contrast, 13, 15–16, 40, 172 backscattered electrons, 49, 59, 172 See also Z-contrast cathodoluminescence, 15, 21, 37, 138 chanelling contrast, 5 and charging, 29, 31, 61–64, 66 chemical contrast, lack of, 37–8 collection contrast, 40 and contamination, 46, 57, 140 control, action of, 135 density contrast, 7, 13, 30, 42, 61, 86, 269 differential interference contrast, 177, 179, 181–182, 190, Color plate 2 double-layer coating, 254, 259 effective coating thickness, 61–62, 64–65, 172, 259, 262 expansion, 2, 28 fluorescence, LM, 38, 193 interaction volume, 59 See also Penetration, electron mechanical disruption of, 197, 205 and resolution, 11, 30, 44, 111, 142–143 secondary electron, 7, 32, 50–51, 64–65, 251 and signal-to-noise, 31, 33–35, 57 of small features, 31, 34, 44, 52, 66, 125, 146, 157, 262 in TEM, 40 theory, 4, 31, 33, 66–69 topographic, 7, 13, 19, 31, 33, 40, 44, 64–65, 142–143, 146, 155 topographic-Z contrast, 64 uncoated specimens, 33, 45, 64–65, 83–84, 171–172, 176, 179, 255, 257, 260, Color plate 3 very-low V0 SEM, 53
Index voltage contrast, 16 See also Charging visibility, 142–143 Z-contrast, 13, 15–16, 40, 172 Correlative LM/LVSEM/TEM, 172, 176–177, 190–193 fluorescent labels w/metal particles, 190–192 quantum dots, 180–182 specimen preparation, 183–184, 190–191 CPD, see Critical-point drying (CPD) Critical-point drying (CPD), 147, 149, 184, 191, 215–219, 233–234 ameloblasts, 209 artifacts, 59–60, 215–219 bacteria, 75, 77, 216, 222–226 bound water, 72 centriole/centrosome, 155 cleans specimen, 69, 74 of cryo-specimens, osteoblasts, 88, 205–206 dissection of, 204–206, 207, 209 GH3 cell, 75 importance of proper dehydration, 59–60, 147–148, 150, 152, 157, 184, 254 liver, 207 lung, 210 mitotic apparatus, 76, 154 mouse embryo, 206 nuclear-pore complex, 79–80, 153, 158, 160–161 plants, 231, 233–234, 236–237, 239, 241 platelets, human, 39, 51, 175, 179–181, 190–192 sea-urchin, embryo, 68 sperm-whale dentine, 78 tissue culture cell, 74, 89 Toxoplama gondii, 161 yeast, 219–221, 261 Cryo-SEM, 82–89, 250–251 artifacts, ice-crystal, 82, 224, 226, 230, 249–253 cryo-immobilization, 215–226 cryo-planing, 224 fracture faces, 254–255 fixed animal cells, 215–216, 224–226 focused-ion beam cutting (cryoFIB), 225–226 freeze-dried surfaces, 86, 88, 253–254 freeze-substitution, 157, 184, 191, 216, 247, 254 frozen-hydrated surfaces, 86, 88, 253 high-pressure freezing, 191, 224, 236, 243, 247, 251–252, 255, 261, 263 ice, thermal conductivity, 252
Index ice-crystal artifacts, 82, 224, 226, 230, 249–253 micro-crystalline ice, 224, 226 pre-fixation, 216 processing of frozen samples, 252 protocols, 260–263 specimen preparation, 251–252 stabilization with aldehydes, 217–218 surface coating, 256–260 thaw-fix, 88–89 thickness, properly frozen, 247, 252 Cryofixation, 215–226, 247, 251–252 Cryosputter, 253, 260 Cytoskeleton artifacts (trabeculae), 60, 184 plants, 231, 235, 237, 239 protocols, 147 results, 74, 149–151, 179–180 stabilizing buffer, 145–149, 217, 236–237, 245 See also Microtubules and Actin Cryotomy of frozen samples, 224 Cyst wall of Giardia, 33 Damage, see Specimen damage De Broglie wavelengths, 31 De-embedding, crown-ether, 155–158 Deflection, beam, 2–4, 9, 197 double-deflection scanning, 9, 13 for live-time stereo, 90–93, 198, 201–202, 206 Dehydration, 59, 147, 149–150, 161, 178, 181, 216, 218 artifact, “microtrabeculae”, 60, 184, 226 bound water, 72, 212, 215 crown-ether de-embedding, 161 incomplete, causes artifacts, 59–60, 147–148, 150, 152, 157, 184, 236, 250, 254 limitations, 60–61, 215, 218, 236, 254 plants, 233, 236 protocols, 147, 149, 161, 177 Delocalization, of SE production, 72 limit on high-resolution, 74–75 Density contrast, 30, 42, 61, 86, 259 Dental enamel, 203 Der Elektronenabtaster, 4 Detector quantum efficiency (DQE), 137–140, 143 digital detectors, 139–140 importance of, 140–141 table, 139, 141 Detectors, electron backscattered electrons (BSE), 28, 48–51, 54, 141, 222, 224, 236, 257
309 collection field, effect, 48–49, 139 DQE, 137–141, 143 Everhart–Thornley SE detector, 20, 47, 55, 57, 133, 139, 198 improvements in, 47 performance, measurement, 132 secondary electrons, 20, 47, 55, 57, 133, 139, 198 tables of, 139, 141 Detectors, x-ray microcalorimeter, 296–301 SiLi (EDS), 279–288 silicon drift detector (SDD), 293–298 wavelength dispersive (WDS), 289–292 x-ray optics augmented WDS, 298–300 Detergent extraction, 51, 81, 231, 234–235, 237 Triton X-100, 148–151, 153, 155, 161 Diamond ultra-microtome knives, 203 Diffraction, 31, 45–46, 59–60, 112–114, 117–118 correction, 120 in light microscopy, 178, 190, Color plate 2 limit, 116 pattern, 50, 124, 258 wavelength-dispersive x-ray detector, 289–292 x-ray, 250 Diffraction-limited condition, 114 Diffusion, surface, contamination, 38 Disc of least confusion (DOLC), 109 Disruption/isolation, plant cells, 235–237 Distortion, 32, 47–48, 82–84, 88 of beam, by collection field, 47–48 of beam, by spherical aberration, 109 of SE collection field, by charging, 32, 83–84, 198 of image, barrel/pincushion, 124 of image, caused by charging, 32, 83–84 of specimen, improper drying, 147, 215–219 of specimen, radiation damage, 82, 88 Double-layer coating, 221–223, 225, 247, 257–259, 261, Color plate 13 advantages, 260 DQE, see Detector quantum efficiency (DQE) tables of, 139, 141 EDS, see Energy-dispersive spectrometer (EDS) Electron, 29–30 Electron beam advantages as imaging probe, 30 current density, 15, 31, 45, 56, 83 induced conductivity, 62–63 limitations, 30–36
310 Electron beam (cont.) penetration, 11, 257, 271, 294–295 range, 32, 73, 270–273 scanner, 3–5 See also Electron source Electron lenses, characteristics of, 30–31, 122 convergence angle, 109, 112–113, 118–120, 123 distortion of, 124 magnetic, 6, 8–9, 20, 46 multipole, 115–116 retarding, 46, 53 Electron microscope development of, 2 scanning approach to, 4, 28–29 See also Scanning electron microscope (SEM), Low-voltage scanning electron microscope, and Transmission electron microscope Electron range, 19, 32, 42–43, 73, 270–273 See also Penetration Electron sources, 31–32, 45, 56, 83, 108, 117–120 Boersch effect, 32, 56 brightness, 32, 45–46 current density, 15, 31, 45, 56, 83, 108 effective source size, 107 electron-electron interactions, 56 energy spread, 46, 53, 56, 59, 110, 117–118 field-emission (FE) sources, 22, 46, 53, 56, 74, 94, 146, 171, 250–251, 256, 260 Langmuir equation, 45 lanthanum hexaboride (LaB6 ), 46, 293 monochromator, 117 Schottky sources, 46, 53, 56, 111, 117–120, 122 source size, 31, 107 thermal-field (TF), sources, 53, 56 tungsten hairpin, thermal, 31, 45, 111, 293 Enamel, dental, 201 Energy-dispersive spectrometer (EDS), 279–281 detector arrays, 296–298 detector counting rate, 295 energy resolution at MnKα , 281–282, 295 limits of detection, 285–289 “Lost” elements, 281–283 low-energy photon detection, 295–296 operating temperature, 295 peak interference, 283–285 spatial resolution, 268, 271–272, 275, 293 Energy-filtering TEM (EF-TEM), 176, 178 results, 193, Color plate 3
Index Environmental SEM, 33, 61, 123, 199–200, 207, 212, 272 early history, 16 Everhart, Thomas E., 6, 16, 131 Everhart-Thornley detector, 20, 47, 55, 57, 133, 139, 198 Faceted cPd particle, 174–175, 188, 192 Fax machine, prototype, 1 Field-emission (FE) sources, 22, 46, 53, 56, 74, 94, 146, 171, 250–251, 256, 260 early FE-SEM, 46, 56, 74 modern microscopes, 55, 59, 110, 157, 235 outlook, 93 results, 55 for video imaging, 211 Fixation, 59, 216–222, 236, 247–248 artifacts, 59–61, 217, 250, 252 buffers, see Buffers correlative LN/LVSEM/TEM, 178–181, 191 glutaraldehyde, see Glutaraldehyde Karnovsky’s, 158 maceration, 233 paraformaldehyde, 146, 158, 231, 233, 238, 239 osmium tetroxide, see Osmium tetroxide (OsO4 ) osmium-thiocarbohydrazide (OTO), 204, 206 perfusion fixation of rat, 209 plant cells, 230, 233, 236 polyethylene-glycol, 189, 240 post fixation, 81, 146, 161, 236 prefixation for cryo, 191, 216–222, 226, 233, 237, 251 protocols, 146–153, 157–158 tannic acid, 80, 146–147, 149, 152–153, 161, 183, 185, 231, 233 thaw-fix, 88–89 Focused-ion beam cutting (cryoFIB), 225–226 Fracture, to reveal internal structure carbon, 65 ceramic, 19–20 cleanliness, 67, 82 coating, 251, 262 disadvantages, 155 dry fracture, 74, 76 fracture faces, 254 freeze-fracture, 67, 74, 82–88, 155, 230–233, 235, 249, 253, 259, Color plate 13 osmotic shock, 237, 242 performed in SEM, 204 of plants, 224, 230–233, 235
Index protocol, 261–263 thaw-fix, 89 uncoated, 45, 83–84, 255, 257, 260, Color plate 3 Freeze drying, 41, 57, 59, 72, 184, 209–212, 215–217, 234, 247, 252, 253–254, 261 low-temperature, 258 in SEM, 41, 57, 209–211, 247, 252–254, 261 Freeze-fracture, 67, 74, 82–88, 155, 231–233, 234, 249, 253, 259, Color plate 13 mimicked by cryo-LVSEM, 84–87, 89, 219–220, 225, 255, 261 replicas, 67, 161, 247, 251, 254–255, 259 Freeze-substitution, 157, 184, 191, 216, 247, 254 Frog oocytes, nuclear-pore complex, 152 Giardia, parasites, 33, 35–36, 74–75, 222 Glutaraldehyde fixative, 59, 146–147, 149, 152, 217–226 correlative LM/LVSEM/TEM, 183, 191 flight muscle, 158 Giardia, 222 membrane, 60 microorganisms, 224–226 neutralization with glycine, 191 nuclear-pore complex, 80–81, 153 plants, 231, 233, 236–239 platelets, 183, 221–223 protocols, 146–147, 149, 152 Toxoplasma, 161 yeast, 218–221 Gold labels, 39, 50–51, 155, 172, 180, 223, 234, 242 GPI-IX complex, 221–222 High-pressure freezing, 191, 224, 236, 243, 247, 251–250, 255, 261, 263 of pancreas tissue, 262–263 High-voltage electron microscopy (HVEM), 60, 158–159, 161–163, 180–182, 190–193 correlative microscopy, 190–193 specimen preparation, 172 Human platelets, 39, 51, 175, 179–181, 190–192, 221–223 fixation of, 221–223 Ice crystal damage, 82, 224, 226, 230, 249–253 Image recording, video, 199 Immunolocalization, 38, 150, 152–154, 176, 216, 218, 221, 234, 236–237
311 correlative, see Correlative LM/LVSEM/TEM fluorescence microscopy, 150, 153–154 gold labeling, 155 human platelets, 221, 223 plants, 234 See also Colloidal metal labeling Insect flight muscle, 158–161 Insects and arthropods in situ experiments, 204–206 live observations, 209–211 charging when metal coat disrupted, 197 osmium-thiocarbohydrazide (OTO), 204, 206 Interaction volume, electron, 19, 40–41, 43, 45, 59, 130, 260, 271 See also Penetration Isolated organelles centrosomal material, 153–161 frog oocytes, nuclear envelope, 152 mitotic spindles, 152–153 Johnson noise, 131 Karnovsky’s, fixative, 158 Knoll, Max, 2, 3–4 diagram scanning microscope, 3, 28 Langmuir equation, 45 Lanthanum hexaboride (LaB6 ) cathodes, 46, 293 Laser confocal fluorescence microscope, 40 Lectin labels, 38, 51, 175 Lens aberrations, 30–31 Light microscopy (LM), 27, 29, 89, 153, 174, 176–177 colloidal gold specimens, 38 confocal, 40, 226, 250 correlative microscopy, 176–177, 183, 193, 226, 240 fluorescence, 37–38, 40, 150, 153–154, 177, 180, 182, 190–193, Color plate 2 quantum dot, 180–182 scanning, 2, 30 wavelength, 31 Low beam energy, see Low-voltage Low-voltage scanning electron microscope (LVSEM), see Low-voltage SEM Low-voltage microanalysis advantages of, 272–273 applications, 300–302 capabilities, 279–281 combined Si-EDS/WDS, 292–293 detector arrays, 296–298
312 Low-voltage microanalysis (cont.) detector counting rate, 295 for discriminating metal-particle labels, 172 electron scattering effects, 275–276 energy-dispersive spectrometry, (EDS), see Energy-dispersive spectrometry instrumentation, 267–268 limitations of, 277–279 limits of detection, 285–289 low-energy photon detection, 295–296 “Lost” elements, 281–283 microcalorimetry, 296–299 minimized x-ray absorption, 273–275 operating temperature, 295 particles, 300–302 peak interference, 283–286 physical basis/limitations, 289–292 rationale for, 268–271 recent instrumentation, 294 resolution at MnKα , 295 spatial resolution, improved, 272–273 silicon-drift detector (SDD), 293–294 wavelength-dispersive detector, 289–292 x-ray optics-augmented WDS, 296–298 x-ray production/spatial resolution, 271–272 Low-voltage SEM (LVSEM), 59–61 barriers to operation at low V0 , 45–46 beam-induced surface contamination, 37 colloidal metal labels, see Colloidal metal labels composition contrast, 13, 15–16, 40, 172, 176 correlative studies, 176–183 electrons as probing radiation, 29–30 first instrument, 19, 45 history, 18–20, 268–269 imaging, 45–46 instrumentation for high-resolution, 53–56 interaction volume, 19, 40–43, 45, 59, 130, 260, 271 limitations, 59–61 performance of early FE-SEMs, 56–59 re-emergence, 52–53 shape/size, 173–175 specimen preparation, see Specimen preparation stereo SEM, real-time, 89–93, 198–209 Thornley R.F.M., 6, 16, 18, 45, 69, 251 vacuum requirements, 33–36, 46, 53, 56–59, 66, 272 Low-voltage SEM of animal cells actin cytoskeleton, 147–148 ameloblasts, 209 bacteria, 75, 77, 216, 222–226
Index centriole/centrosome, 153–161 of cryo-specimens, osteoblasts, 88, 205–206 dissection of, 204–207, 209 GH3 cell, 75 liver, 207 lung, 210 mitochondria, 83, 88, 165, 235 mitotic apparatus, 76, 152–154 mouse embryo, 206 nuclear-pore complex, 79–80, 152–153, 158, 160–161 osteocytes, 149–151 plants, 231, 233–234, 236–237, 239, 241 platelets, human, 39, 51, 175, 179–181, 190–192, 221–223 sea-urchin, embryo, 68 sperm-whale dentine, 78 tissue culture cell, 74, 89 Toxoplama gondii, 148–150, 161–162 yeast, 219–221, 261 Low-voltage SEM of plant cells, 229–230 Arabidopsis thaliana, 239–241 Chara, 234, 237, 238 clathrin-coated vesicles, 234–235 coated vesicles, 235 detergent extraction, 231, 234–235, 237 diffusion of fixatives, 235 disruption and isolation, 235–237 handling, 230 HRSEM, 230–232 osmic maceration, 232–233 osmotic shock, 237, 242 plasmolysis and protoplasts, 238–240 Low-voltage SEM microanalysis, see Low-voltage microanalysis Maceration, plants, 232–233 Markers, see Colloidal metal labeling Measuring secondary electron S/N, 132–137 Microcalorimetry, x-ray detector, 296–299 resolution, 297, 298–299 Microscopy development, 15–16 rise of surface imaging, 28–29 two approaches, 27–28 Microtubules, 60, 76, 148, 150, 153–154, 169, 234–236, 237 cortical, plants, 241–242 pellicle cytoskeleton, 148 stabilizing buffer, 147, 149, 151, 183, 236, 242 Mitochondria, 83, 88, 165, 229, 235 Mitotic spindles, 76, 152–154 Monoscope, 6
Index Mouse embryo, 206 Multipole electron lenses, 115–116
313
Optical transfer function (OTF), 121–122 Optimizing spot size, 113–115 Osmic maceration, 232–233 Osmium-thiocarbohydrazide (OTO) conductive staining, 60, 204, 206, 234 density effects, 231 Osmium tetroxide (OsO4 ) fixative, 60, 146, 232–233 boric acid buffer, 207 for colloidal-gold labeling, 224 freeze-substitution, 217 maceration, plants, 231–233 membrane fixation, 60 nuclear-pore complex, 80–81, 153, 157–158, 161 osmium-thiocarbohydrazide (OTO), 204, 206, 232 perfusion fixation of rat, 209 plants, 231–232, 234, 237 platelets, 179, 221 Proteus mirabilus, 224 protocols, 146–147, 149, 152, 161, 165, 179, 183 yeast, 219–221, 261 Osteocytes, 149
parasitophorous vacuole, (PV), 162–163 Toxoplasma gondii, 146, 147–149, 157, 161–166 Particles, 171 contrast in BSE, 52 counting, early, 2 intramembrane, 67, 83, 87, 257 metal coating, 69–70 SE image of, 120 virus, 41, 217, 247, 263 x-ray microanalysis, 300–302 See also Colloidal metal particles Pellicle microtubule cytoskeleton, 148 Penetration, electron beam, 11, 15, 19, 32, 40, 42, 52, 57, 65, 139, 256–257, 271–273, 277 for BSE imaging, 139, 79 effects spatial resolution, 11, 15, 19, 32, 40, 52, 65, 256–257, 271 of electron beam, 11, 257, 271 electron range, 271 of fixative, 155 of gold labels, 174, 180 reduced by heavy-metal coating, 45, 171 in x-ray microanalysis, 294–295 Pincushion distortion, of electron lenses, 125 Plant cells, 229–243 Arabidopsis thaliana, 239–241 Chara, 234, 237, 238 chloroplasts, 229, 232–233, 235, 237, 238 coated vesicles, 235 clathrin-coated vesicles, 234–235 detergent extraction, 231, 234–235, 237 diffusion of fixatives, 235 disruption and isolation, 235–237 handling, 230 osmic maceration, 232–233 osmotic shock, 237, 242 plasmolysis and protoplasts, 238–243 Plasmolysis and protoplasts, 238–243 Poisson statistics, 31, 131–135, 256, 281 Polaroid materials, 198 Polyethylene-glycol, preservative, 189, 240 Popcorn cPd particles, 174–175, 188 Protoplasts, 230, 238–242 PV, see Parasitophorous vacuole (PV)
Paraformaldehyde, fixative, 146, 158, 231, 233, 238, 239 Parasites, 148–150, 163 on fleas, 209 Giardia, 33, 35–36, 74–75, 222 interactions, 162–166
Quantum dot, 180–182 Quantum efficiency, 137–138 detector, 29, 47, 50, 137–138 See also Detector quantum efficiency (DQE)
Nanoparticles, see Colloidal metal particles Near-field microscopy, 2 Noise, in the SEM signals, 129–143 detector, effect of, 137–139 DQE, importance of, 140–141 DQE, of digital detectors, 139–140 effect on microscope performance, 142–143 measuring noise in SE signal, 132–135 measuring signal-to-noise ratio, 135–137 origin of, 129–131 Poisson statistics, 31, 131–135, 256, 281 statistics of, 6, 50, 114, 117, 120–121, 131, 135–138, 142, 247, 257, 260, 262 NPC, see Nuclear pore complex (NPC) Nuclear envelope, 75, 79–81, 145, 152–153, 157–158 Nuclear pore complex (NPC), 79–81, 152, 153, 157–158, 160–161
314 Radiation damage, of specimen, 16, 32–33, 44, 73, 82–88, 129, 155, 171, 252–253, 259 RCA scanning electron microscope, 5–8 Resin-extracted thick sections, 155–166 Resolution, spatial, of LVSEM, 29, 31, 40, 43–44, 52, 105–115, 120–123 aberrations, 30–31, 107–111 See also Aberrations and Aberration corrector affected by electron diffusion, 40, 254 beam current limit, 108, 142 See also Electron source, brightness, and Visibility in biology, 59, 61, 72–73 coating, 61–69 contamination, 69–72 delocalization, 72, 74–75 detectors, 47, 50 diffraction, 31, 112 early SEMs, 9, 11, 13, 15–16, 19 electron optics, 46 electron sources, 56 interaction volume, 19, 40–41, 43, 45, 59, 130, 260, 271 light microscope, 12 multi-factorial, 72–73 optical transfer function (OTF), 121–122 optimizing performance, 113–115 probe diameter, 6, 8, 18, 41, 50, 59115, 142 of small features, 31, 34, 44, 52, 66, 125, 146, 157, 262 source brightness, 45 spatial frequency, 7, 44, 121–122 stray magnetic fields, 46–47, 57 visibility, see Visibility versus TEM, 67 in x-ray microanalysis, 272–273 Ris, Hans, 59–60, 74, 79–82, 146, 150, 152, 157 micrographs, 74, 79–82, 159–163 modern instruments, 53–56, 120 muscle structure, 162–163 nuclear-pore complex, 79–82, 159–160 results, 55, 121 SEM specimen preparation, 59–60, 150, 152, 157, 184 Ronchigram, for correcting aberrations, 124 Rose Criterion, 142–143 Ruska, Ernst, 2–3, 9 Scanned electronic imaging and television, 28 Scanning approach to microscopy, 28–29 invention of, 1
Index Scanning electron microscope (SEM) history, 1 Cambridge commercial SEMs, 12–15 charged particle beams, 2 commercial production of, 20–21 cutting and dissecting in, 203–205 damp-sample SEM, 199 early history, 21–22 electron beam scanner, 3–5 electron optics of, 7 electrons as probes in, 29 evolution, 45–47 freeze drying in, 41, 57, 209–211, 247, 252–254, 261 live, arthropods, 209–211 low V0 performance, 56–59 micro-imaging surfaces, 197 noise performance, 129–143 prototype, SEM1, 14–16 prototype, SEM2 (LVSEM), 16, 18, 20 prototype, SEM3, commercial, 16, 20 RCA early SEM, 5–8 scanned imaging, invention of, 1 static 2-dimensional viewing, 198 temporal resolution, 211 three-dimensional surfaces, 207–209 video-rate LVSEM, 198, 200–201 video-rate stereo SEM, 201–203 Von Ardenne’s SEM, 8–12 See also Low-voltage SEM Scanning optical microscopy, 1–2 Scanning transmission electron microscope (STEM), 5, 8–12, 22 detector layout, 54 Scanning-type microscope, 28–29 Bain, Alexander, 6 double deflection, 13 Knoll’s diagram of, 28 scanning system, 2–4, 9, 29 Stintzing, H., 2 Scherzer, Otto, 110, 115 and spherical aberration, 110 Schottky emitter, 46, 53, 111, 117–120, 122 energy spread, 117–118 Scintillator-PMT detector, 198 SE and BSE detectors, see Detectors Sea urchin, 67–68, 74–76, 83–87, 153, 155–156 Secondary electron emission (SE), 29–30, 132–134 coefficient, 4, 19, 62, 66, 132 coefficient, effective, 19, 62–63, 76 detectors, 20, 46–49, 55, 57, 133, 137, 139, 198 silicon, 138
Index SEM, see Scanning electron microscope (SEM) Shielding, magnetic, 13, 46–47, 53, 55, 123 Shot noise, 131 See also Statistical noise Signal-to-noise ratio (SNR), 6, 50, 114, 117, 120–121, 131, 135–138, 142, 247, 257, 260, 262 early SEM, 6 limits microscope performance, 135–137, 142–143 Rose criterion, 142–143 visibility, 39, 52, 142–143 Silicon drift detectors (SDD), 293–294 thickness, 294 Source, electron, see Electron sources Spatial frequency, of images, 7, 44, 121–122 Specimens, biological actin cytoskeleton, 147–148 ameloblasts, 209 bacteria, 75, 77, 216, 222–226 centriole/centrosome, 153–161 of cryo-specimens, osteoblasts, 88, 205–206 dissection of, 204–206, 207, 209 GH3 cell, 75 liver, 207 lung, 210 mitochondria, 83, 88, 165, 235 mitotic apparatus, 76, 152–154 mouse embryo, 206 nuclear-pore complex, 79–80, 152–153, 158, 160–161 osteocytes, 149–151 plants, 231, 233–234, 236–237, 239, 241 platelets, human, 39, 51, 175, 179–181, 190–192, 221–223 sea-urchin, embryo, 68 sperm-whale dentine, 78 tissue culture cell, 74, 89 Toxoplasma gondii, 145, 147–150, 161, 165 viruses, 41–42, 173, 258–259, 263 yeast, 219–221, 261 Specimen charging, see Charging Specimen damage, 32, 44, 73, 82–88, 129, 155, 171, 259 advantage of LVSEM, 82, 198 beam current density, 83 caused by fixation, 61 caused by glow-discharge coating, 68–69 contamination, see Contamination critical-point drying (CPD) cryo-techniques, 82, 252–253 dehydration, 142 depth effect, 32
315 example images, 36, 83, 85, 88 fluorescence, LM, 38 freezing artifacts, 82, 224, 226, 249–253 and image contrast, 66, 140 ionizing radiation, 32 mass loss, 259 microtrabeculae, 60, 184 radiation damage, 32–36, 44, 66–67, 73, 82–88, 129, 155, 171, 250, 257, 259, 288 reduced at low V0 , 66–67, 82–88, 257 reduced by double-layer coating, 260–261 and scan speed, 198 SEM compared to TEM, 67 theory, 32, 37, 67 tin x-ray microanalysis, 289, 292 Specimen preparation, animal cells antibody staining, see Immunolocalization and Colloidal metal labeling bound water, 72, 212, 215, 250–251 coating, see Coating CPD, see Critical-point drying (CPD) correlative methods, 183–184, 190–193 de-embedded TEM specimens, 155–166 dehydration, see Dehydration detergent extraction, see Detergent extraction effect of fixation and CPD, 216–217 fixation, see Fixation focused-ion beam cutting, 225 freeze drying, see Freeze drying nuclear-pore complex, 152, 157–159 stages of, 250–251 substrates, see Substrates thaw-fix, following cryo, 88 theory, 250–251 whole-mounts, 60, 74, 150, 152, 158–163, 172, 176, 180, 190–191, 230 Specimen preparation, plant cells, see Plant cells Spherical aberration, 31, 109–110 coefficient, 31, 109–110, 142 Scherzer’s rule, 110, 115 Stabilizing before cryofixation, 217–218, 221–222, 226 Stabilizing buffer, cytoskeleton, 145–149, 217, 236–237, 242 Stabilizing colloids, 174, 176, 189 Statistical noise in SEM signal, 31, 131, 134, 256, 281 Poisson noise, 31, 131–135, 256, 281 STERECON method, 220–222
316 Stereo SEM, 16, 41–42, 67, 200–209 high-resolution, 41–42 real-time, 89–93, 200–209 results, 92–93 X-alignment control of, 202 Stereo-tilt coils, 201 Stigmator, 15, 57 alignment, 57, 83, 91, 115, 125 for real-time stereo imaging, 91 Stintzing, H., 2 Stray field, magnetic, 46–47, 57 mains-frequency, 57 Substrates, specimen carbon films, 89–90 carbon grid, 258, 263 chip holder, 71 copper platelet, for cryo, 261 cryo-specimens, 89–90 metal-coated glass, 80, 157–158 sapphire, 221 silicon chip, 62, 66, 71, 129–130 wetting, 217 Surface coating, see Coating Synge, Edward, 1–2 Tables Conventional E0 microanalysis: electron range and x-ray production range for trace species in a carbon matrix, 271 Definition of Orders of Aberration, 10 DQE values for typical SEM BSE detectors, 141 DQE values for typical SEM SE detectors, 139 Low E0 microanalysis: total electron range and x-ray production range for various trace species in a carbon matrix, 272 Measured Performance of an OpticAugmented WDS (Parallax Research, Inc. LEXS), 296 Periodic Table showing choice of atomic shells available for microanalysis at E0 = 2.5 keV, 278 Synthesis of colloidal palladium, 187 Synthesis of cPt w/nucleating sol procedure, 187 The two basic plant tissue protocols, 231 X-ray shell choices for conventional beam energy x-ray microanalysis, 270 X-ray shell choices for low beam energy x-ray microanalysis (E0 =5 keV), 276
Index Comparison of Si-EDS and WDS characteristics, 282 Test specimens carbon, simulation, 32, 43 coating analysis, 70 gold-on-carbon, 59, 120–121 polished silicon, 129–130, 132, 141 T4 polyhead, 217 Pt on carbon, 55, 59, 65 for x-ray microanalysis, 271, 273, 284 TEM, see Transmission electron microscope Thaliana, 239–241 Thermal (tungsten) electron sources, 37 Thermal-field (TF), electron sources, 53, 56 energy spread, 117–118 Thick specimen, definition, 32 Thick-thin filaments, 60 Thornley, R.F.M., first LVSEM, 6, 16, 18, 45, 69, 251 Everhart–Thornley SE detector, 20, 47, 55, 57, 133, 139, 198 first LVSEM paper, 19, 45 Three-dimensional recording/viewing, 198, 207–209 Topographic imaging, 45–46 coding for shape, 30, 33–35 collection contrast, 40 contrast, see Contrast evolution of, 45–46 first images, 3, 7, 13 at low-V0 , 53 modulates metal coating thickness, 64 of small features, 34, 44, 52, 66, 125, 146, 157, 262 See also Resolution Topographic z contrast, 64 Toxoplasma gondii, 145, 147–150, 161, 165 cytoskeleton, 147–148, 150 internal structures, 161, 165 Transmission electron microscope (TEM), 3–4, 6, 18–20, 27 aberration correction, 121, 124 in biology, compared to SEM, 146, 162–165 cold stages, 83 collecting secondary electrons, 46–47 contrast transfer function measurement, 40 conversion to SEM, 20 correlative studies, 176, 190, 193 cryo-specimen flatness, 89–90 damage, 82, 88, 177 electron crystallography, 82, 88 energy-filtering TEM (EF-TEM), 176, 178, 190, 193, Color plate 3 fixation, plant cells, 233
Index freeze-fracture replicas, 251, 254 grazing incidence images, 13 high-voltage electron microscopy (HVEM), 60, 158–159, 161–163, 180–182, 190–193 imaging speed, 67 limitations, 147, 154 of nuclear-pore complex, 80–82 scanning transmission, 5–6, 9–10 specimen damage, 259 specimen preparation, 155, 157 specimen substrate, flatness, 89–90 surface replica specimens, 3, 7, 67, 87, 161, 216–217, 247, 251, 254–255, 259 TEM/SEM, 65, 182 test specimens, 217 whole-mount specimens, 60, 172, 176, 180, 190 Tungsten hairpin thermal electron source, 31, 45, 111, 293 Ultra-high vacuum (UHV) SEM, 4, 6, 22 sputter-ion pump, 22 Uncoated specimens, 33, 45, 64–65, 83–84, 171–172, 176, 179, 255, 257, 260, Color plate 3 yeast fracture faces, 83–84, 262 sea urchin, 83–86 Vacuum requirements, 33–36, 46, 53, 56–59, 66, 272 coating, 68–69 demountable, 6 environmental SEM, 33, 61, 123, 199–200, 207, 212, 272 for field-emission sources, 7, 53 living arthropods, 209 low-temperature freeze drying, 215, 261, 263 molecular drag pump, 57–58, 69 for Schottky TF sources, 56 for specimen coating, 68, 147 sputter-ion pump, 22 surface analysis, 13, 66, 69 system, 58 “wet-SEM”, 199 UHV SEMs, 4, 22, 37 See also Contamination Van der Waals forces, contamination, 37 Very-low V0 SEM, 53 Vibration, 123 sound shielding, 55 Video-rate image recording, 199
317 Video-rate LVSEM, 198 Viruses, 41, 258 adenovirus, 258, 263 as labels, 173 reovirus, 42 tobacco-mosaic virus, (TMV), 258–259, 263 Visibility, 39, 52, 108, 142–143 Rose criterion, 142–143 signal-to-noise ratio, 6, 50, 114, 117, 120–121, 131, 135–138, 142, 247, 257, 260, 262 Voltage contrast, 16 Von Ardenne, Manfred, 4–8, 13 early STEM, 8–12 resolution limit, 12, 40 Wavelength dispersive spectrometer (WDS) combined Si-EDS and, 292–293 physical basis and limitations, 289–292 resolution, 285, 286 x-ray optics-augmented, 296–298 WDS, see Wavelength dispersive spectrometer (WDS) Weiner filter, SE detection, 49, 55 Whole-mount specimens, 60, 74, 150, 152, 158–163, 172, 176, 180, 190–191, 230 correlative studies, 172, 176 dry-fractured cell, 74 frog oocyte nuclear membranes, 150, 152, 158 mitotic apparatus, 75–76, 152–156 plant protoplasts, 230 platelets, 39, 51, 175, 179–181, 190–192, 221–223 X-ray microanalysis detectors, see Detectors, x-ray of particles, 278, 300–302 peak-to-background, for particles, 278 See also Low-voltage microanalysis X-ray optics-augmented WDS, 296–298 YAG BSE detector, 50–51, 54, 222, 224, 236, 258 Yeast, 219–221, 261 cryo-immobilization of, 218–219 Z-contrast, 49, 59, 172 Zworykin, 5–8 early SEM, 7