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