Handbook of Magnetic Materials, Volume 5 Elsevier, 1990 Edited by: E.P Wohlfarth and K.H.J. Buschow ISBN: 978-0-444-87477-1
by kmno4
PREFACE TO VOLUME 5 This Handbook on the Properties of Magnetically Ordered Substances, Ferromagnetic Materials, has a dual purpose. As a textbook, it is intended to help those who wish to be introduced to a given topic in the field of magnetism without the need to read the vast amount of literature published. As a work of reference, it is intended for scientists active in magnetism research. To this dual purpose, the volumes of the Handbook are composed of topical review articles written by leading authorities. In each of these articles an extensive description is given in graphical as well as in tabular form, much emphasis being placed on the discussion of the experimental material in the framework of physics, chemistry and materials science. The original aim of Peter Wohlfarth was to combine new developments in magnetism with the achievements of earlier compilations of monographs, to produce a worthy successor to Bozorth's classical and monumental book Ferromagnetism. It is mainly for this reason that Ferromagnetic Materials was chosen as title for the Handbook series, although the latter aims at giving a more complete cross-section of magnetism than Bozorth's book. Here one has to realize that many of the present specialized areas of magnetism were non-existent when Bozorth's book was first published. Furthermore, a comprehensible description of the properties of many magnetically ordered materials can hardly be given without considering, for instance, narrow-band phenomena, crystal-field effects or the results of band-structure calculations. For this reason, Peter Wohlfarth and I considered it desirable that the Handbook series be composed of articles that would allow the readers to orient themselves more broadly in the field of magnetism, taking the risk that the title of the Handbook series might be slightly misleading. Publication of the Volumes of the Handbook cannot be considered as having proceeded by a process of continuous growth. Volumes 1 and 2 first appeared in 1980, and were followed by Volume 3 in 1982. All three Volumes have been reprinted in the meantime and many of the articles contained in them have frequently been quoted in the scientific literature as providing an authoritative description of the achievements made in a given subfield of magnetism. In the
vi
PREFACE TO VOLUME 5
early 1980's the interest of the scientific community in magnetism declined considerably. Many a scientist had moved to greener pastures and, in a way, this had hampered completion of Volume 4, published in 1988. However, at the time that I joined Peter Wohlfarth as co-editor of the Handbook series early in 1986, we both had the impression that there was a pronounced increase in interest from the scientific as well as from the technological side, to the extent that one could speak of a revival of magnetism. Topics like the Kondo effect, spin glasses and valence fluctuations had kept the fires burning, and a considerable proliferation in research effort occurred in other areas like amorphous magnetism, permanent magnets and magneto-optics. The magnetism of thin films and multi-layers and heavy-electron systems became novel topics. Improvements in band-structure calculations and the possibility of applying these calculations to more complicated structures started to play an important role in the understanding of many features of magnetism. All these facts made it necessary that we again pose the question 'what is magnetism and where does it go?' The outcome of such considerations was that we eventually decided to include chapters in the Handbook series that were able to provide the readership with an insight into modern trends in magnetism and new achievements in this area. Several such topics have already been mentioned above. Other topics of this kind dealt with the increased activity and investigations of the magnetism of magnetic superconductors and with investigations of the magnetic properties of hydrides. We also felt that there should be an account of the progress that had been made in the understanding of first-order magnetic processes and of quadrupolar interactions in 4f systems and their role in magnetic ordering and in magneto-elastic effects. The magnetism of strongly enhanced itinerant alloys and compounds and the magnetism of Invar alloys already had a long-standing tradition. However new achievements were made in this area, and these were often obtained with the more sophisticated experimental techniques available nowadays. This also made it necessary to include chapters covering the progress made in these latter fields, and many of these latter topics are contained in this volume. It will be clear that we had to abandon the ideal of restricting the Handbook series to four volumes. The large number of topics will require several more volumes, the more so since the revival of magnetism is still going on. This is, for instance, manifest from the extraordinarily large number of contributions submitted for presentation at the International Conference on Magnetism, held in Paris in July 1988. The number of contributions (about 1300) exceeds by far the number of contributions (about 900) presented at the preceding ICM Conference held in San Francisco in 1985. This conference has also focused the attention on several new fields of interest such as the magnetism of high-temperature superconducting compounds and dilute magnetic semiconductors while it stressed the intimate connection existing between fundamental magnetism and high-density magnetic recording. Also these topics will be accommodated in future volumes of this Handbook series. The task to provide the readership with novel trends and achievements in magnetism would have been extremely difficult without the professionalism of the
PREFACE 'TO VOLUME 5
vii
North-Holland Physics Division of Elsevier Science Publishers and I would like to thank Peter de Ch~tel, Anita de Waard, Bas van der Hoek and Willem jan Maas for the great help and expertise. K.H.J. Buschow Philips Research Laboratories
CONTENTS Preface to V o l u m e 5
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
- . .
v ix
C o n t e n t s of V o l u m e s 1 - 4 . . . . . . . . . . . . . . . . . . . . . . . . . . .
xi
List of c o n t r i b u t o r s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xiii
1. Q u a d r u p o l a r I n t e r a c t i o n s a n d M a g n e t o - e l a s t i c Effects in R a r e - e a r t h IntermetaUic C o m p o u n d s P. M O R I N a n d D. S C H M I T T . . . . . . . . . . . . . . . . . . . . . . 2. M a g n e t o - o p t i c a l Spectroscopy of f - e l e c t r o n Systems W. R E I M a n d J. S C H O E N E S . . . . . . . . . . . . . . . . . . . . . . 3. INVAR: M o m e n t - v o l u m e Instabilities in T r a n s i t i o n Metals a n d Alloys E.F. WASSERMAN . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. Strongly E n h a n c e d I t i n e r a n t I n t e r m e t a l l i c s a n d Alloys R E . B R O M M E R and J.J.M. F R A N S E . . . . . . . . . . . . . . . . . 5. F i r s t - o r d e r M a g n e t i c Processes G. A S T I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. M a g n e t i c S u p e r c o n d u c t o r s O. F I S C H E R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1 133 237 323 397 465
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
551
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . .
577
........
Materials I n d e x . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix
585
CONTENTS OF VOLUMES 1-4 Volume 1 1. 2. 3. 4. 5. 6. 7.
Iron, Cobalt and Nickel, b y E . P . Wohlfarth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dilute Transition Metal Alloys: Spin Glasses, b y J . A . M y d o s h a n d G . J . N i e u w e n h u y s . . . Rare Earth Metals and Alloys, b y S. L e g v o l d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rare Earth Compounds, b y K . H . J . B u s c h o w . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Actinide Elements and Compounds, b y W. T r z e b i a t o w s k i . . . . . . . . . . . . . . . . . . . . . . . . . . Amorphous Ferromagnets, by F . E . L u b o r s k y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Magnetostrictive Rare Earth-Fe 2 Compounds, by A . E . Clark . . . . . . . . . . . . . . . . . . . . . . .
1
71 183 297 415 451 531
Volume 2 1. 2. 3. 4. 5, 6. 7. 8,
Ferromagnetic Insulators: Garnets, by M . A . Gilleo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Soft Magnetic Metallic Materials, b y G . Y . Chin a n d J . H . W e r n i c k . . . . . . . . . . . . . . . . Ferrites for Non-Microwave Applications, b y P.I. Slick . . . . . . . . . . . . . . . . . . . . . . . . . . . . Microwave Ferrites, b y J. Nicolas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Crystalline Films for Bubbles, b y A . H . E s c h e n f e l d e r . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Amorphous Films for Bubbles, b y A . H . E s c h e n f e l d e r . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Recording Materia!s, b y G. Bate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ferromagnetic Liquids, b y S.W. Charles a n d J. P o p p l e w e l l . . . . . . . . . . . . . . . . . . . . . . . . .
1 ~. .
55 189 243 297 345 381 509
Volume 3 1. Magnetism and Magnetic Materials: Historical Developments and Present Role in Industry and Technology, b y U. E n z . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Permanent Magnets; Theory, b y H. Zijlstra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. The Structure and Properties of Alnico Permanent Magnet Alloys, by R . A . M c C u r r i e . . . 4. Oxide Spinels, b y S. K r u p i 6 k a a n d P. N o v 6 k . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. Fundamental Properties of Hexagonal Ferrites with Magnetoplumbite Structure, by H. K o j i m a . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6. 7. 8. 9.
Properties of Ferroxplana-Type Hexagonal Ferrites, b y M . S u g i m o t o . . . . . . . . . . . . . . . . . Hard Ferrites and Plastoferrites, b y H. Stiiblein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sulphospinels, by R . P . van Stapele . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Transport Properties of Ferromagnets, by I . A . C a m p b e l l a n d A . Fert . . . . . . . . . . . . . . . . xi
1
37 107 189 305 393 441 603 747
xii
CONTENTS OF VOLUMES 1-4
Volume 4 1. 2. 3. 4. 5.
Permanent Magnet Materials Based on 3d-rich Ternary Compounds, by K . H . J . Buschow Rare Earth-Cobalt Permanent Magnets, by K.J. Strnat . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ferromagnetic Transition Metal Intermetallic Compounds, by J.G. Booth . . . . . . . . . . . . . Intermetallic Compounds of Actinides, by V. Sechovsk~ and L. Havela . . . . . . . . . . . . . . . Magneto-optical Properties of Alloys and lntermetallic Compounds, by K . H . J . Buschow.
1
131 211 309 493
chapter 1 QUADRUPOLAR INTERACTIONS AND MAGNETO-ELASTIC EFFECTS IN RARE EARTH INTERMETALLIC COMPOUNDS
P. MORIN and D. SCHMITT Laboratoire Louis N@el* C.N.R.S. B.P. 166 X 38042 Grenoble Cedex France
*Associated with Universit6 Joseph Fourier, Grenoble
Ferromagnetic Materials, Vol. 5 Edited by K.H.J. Buschow and E.P. Wohlfartht © Elsevier Science Publishers B.V., 1990
CONTENTS 1. I~ t r o d u c t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Fc rl~ alL, m . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. T h e c t b i c H a n n i l t c n i a n . . . . . . . . . . . . . . . . . . . . . 2.1.1. G e n e r a l e ~ p l e s s i o n . . . . . . . . . . . . . . . . . . . . 2.1.2. H a r m o n i c d e s c r i p t i o n of m a g n e t o s t r i c t i o n . . . . . . . . . . . . . 2.1.3. T h e f e t a l q ~ a & u p o l a r H a m i l ( o n i a n . . . . . . . . . . . . . . . 2.2. T r e a t m e n t of the c u b i c H a m i l t c n i a n . . . . . . . . . . . . . . . . . 2.2.1. F e r t m b a t f o n t h e o r y . . . . . . . . . . . . . . . . . . . . 2.2.2. C e n e r a l i z a t i o n of l h e s u s c e p t i b i l i t y f o r m a l i s m . . . . . . . . . . . . 2.2.3. B e h a v i o u r of l h e ~ a r i c u s s u s c e p t i b i l i t i e s . . . . . . . . . . . . . 2.2.4. M a g n e t i c a n d q u a & u p o l a r t r a n s i l i o n s . . . . . . . . . . . . . . 2.3. E x t e n s i o n to l o w e r s y m m e t r i e s ( h e x a g o n a l a n d t e t r a g o n a l ) . . . . . . . . . 2.3.1. T h e H ~ m i l t o n i ~ n . . . . . . . . . . . . . . . . . . . . . . 2.3.2. H a r m o n : ' c d e s c r i p t i o n of 1Tagnetostr c t i c n . . . . . . . . . . . . . 2.3.3. P e l t u r b a t i o n tl;eo~y . . . . . . . . . . . . . . . . . . . . 2.3.4. C o m p a r i s o n w i t h c u b i c s y m m e t r y . . . . . . . . . . . . . . . . 2.4. A p e c u l i a r c a . e : the semi-classical t r e a l m e n t . . . . . . . . . . . . . . 2.4.1. T h e p a r a m ~ g n e l i c lzhase . . . . . . . . . . . . . . . . . . . 2.4.2. T h e c r d e r e d p h a s e . . . . . . . . . . . . . . . . . . . . . 3. E x p e r i m e n t a l e v i d e n c e . . . . . . . . . . . . . . . . . . . . . . . 3.1. Q u a d l u p o l a r o r d e r i n g s . . . . . . . . . . . . . . . . . . . . . 3.1.1. F e r i c q u a d r ~ p o l a r older:'ng . . . . . . . . . . . . . . . . . . 3.1.1.1. T m Z n . . . . . . . . . . . . . . . . . . . . . . 3.1.1.2. T m C d . . . . . . . . . . . . . . . . . . . . . . 3.1.1.3. C e A g . . . . . . . . . . . . . . . . . . . . . . 3.1.2. A n t : f e r l c q u a d ~ p o / ~ r o l d e r i n g . . . . . . . . . . . . . . . . 3.1.2.1. C e B 6 . . . . . . . . . . . . . . . . . . . . . . 3.1.2.2. T m G a 3 . . . . . . . . . . . . . . . . . . . . . 3.1.2.3. P I P b 3 . . . . . . . . . . . . . . . . . . . . . . 3.1.3. S t r u c t u r a l l r a n s i t i o n s p o s s i b l y of q u a d r u p o l a r o r i g i n . . . . . . . . . 3.1.3.1. P r C u z . . . . . . . . . . . . . . . . . . . . . . 3.1.3.2. U P d 3 . . . . . . . . . . . . . . . . . . . . . . 3.2. N a t u r e of q u a d r u p o l a r a n d m a g n e t i c t r a n s i t i o n s . . . . . . . . . . . . . 3.2.1. Q u a d r u p o l a r t l ~ n s t i o n s . . . . . . . . . . . . . . . . . . . 3.2.2. M a g n e t i c o r d e r i n g in t h e p r e s e n c e of q u a d r u p o l a r i n t e r a c t i o n s . . . . . . . . . . . . . . . . . 3.2.2.1. F i r s t - o r d e r t r a n s i t i o n s a n d tricriti.cality
5 6 7 7 11 12 12 13 16 17 20 24 24 27 28 30 30 30 32 34 34 34 35 36 36 38 38 39 40 41 41 42 43 43 43 44
QUADRUPOLAR
E F F E C T S IN R A R E E A R T H I N T E R M E T A L L I C S
3.2.2.2. Second-order transition in PrMg 2 due to negative qua drupol a r interactions . . . . . . . . . . . . . . . . . . . . . . 3.2.3. Magnetic ordering in the q u a d r u p o l a r phase . . . . . . . . . . . . 3.2.3.1. C e A g , CeB 6 . . . . . . . . . . . . . . . . . . . 3.2.3.2. T m Z n . . . . . . . . . . . . . . . . . . . . . . 3.2.4. Presence of additional couplings . . . . . . . . . . . . . . . . 3.2.4.1. DySb and rare earth monopnictides . . . . . . . . . . . . 3.2.4.2. The RA12 series . . . . . . . . . . . . . . . . . . 3.3. D e t e r m i n a t i o n of the q u a d r u p o l a r p a r a m e t e r s from susceptibility techniques . 3.3.1, Elastic constants . . . . . . . . . . . . . . . . . . . . . 3.3.2, Parastriction . . . . . . . . . . . . . . . . . . . . . . 3.3.3. Third-order p a r a m a g n e t i c susceptibility . . . . . . . . . . . . . 3.4. Effects on the magnetostriction . . . . . . . . . . . . . . . . . . 3.4.1. Magnetostriction of dilute compounds . . . . . . . . . . . . . . 3.4.2. Magnetostriction in the ordered phase . . . . . . . . . . . . . . 3.5. Effects on the magnetization processes . . . . . . . . . . . . . . . . 3.5.1. P a r a m a g n e t i c phase . . . . . . . . . . . . . . . . . . . . 3.5.2. O r d e r e d phase . . . . . . . . . . . . . . . . . . . . . . 3.5.3. Effect on the direction of easy magnetization . . . . . . . . . . . 3.6. Effects on the magnetic structures . . . . . . . . . . . . . . . . . 3.6.1. Multi-axial structures . . . . . . . . . . . . . . . . . . . 3.6.2. I n c o m m e n s u r a t e magnetic structures . . . . . . . . . . . . . . 3.7. Effects on magnetic excitations . . . . . . . . . . . . . . . . . . 3.8. Two-ion anisotropic magneto-elasticity . . . . . . . . . . . . . . . . 3.9. Isotropic magneto-elasticity and pressure effects . . . . . . . . . . . . . 3.9.1. Two-ion contribution . . . . . . . . . . . . . . . . . . . 3.9.2. Single-ion contribution . . . . . . . . . . . . . . . . . . . 3.9.3. Pressure effects . . . . . . . . . . . . . . . . . . . . . 3.10. Magneto-elasticity in the presence of lattice instability . . . . . . . . . . . 4. Magneto-elastic and pair interaction coefficients in rare earth intermetallic series . . . . 4A. CsCl-type structure compounds . . . . . . . . . . . . . . . . . . . 4.1.1. R Z n series . . . . . . . . . . . . . . . . . . . . . . . 4.1.2. R C d series . . . . . . . . . . . . . . . . . . . . . . . 4.1.3. R A g series . . . . . . . . . . . . . . . . . . . . . . . 4.1.4, R C u series . . . . . . . . . . . . . . . . . . . . . . . 4.1.5, R M g series . . . . . . . . . . . . . . . . . . . . . . . 4.2. NaCl-type structure compouods . . . . . . . . . . . . . . . . . . . 4.3. AuCu3-type structure compounds . . . . . . . . . . . . . . . . . . 4.4. Cubic Laves phase compounds . . . . . . . . . . . . . . . . . . . 4.4.1. RA12 compounds . . . . . . . . . . . . . . . . . . . . . 4.4.2. R M 2 compounds (M = Mg, Ni, Co) . . . . . . . . . . . . . . . 4.4.3. R F e 2 compounds . . . . . . . . . . . . . . . . . . . . . 4.5. Hexagonal CaCus-type structure compounds . . . . . . . . . . . . . . 4.5.1. RNi 5 compounds . . . . . . . . . . . . . . . . . . . . . 4.5.2. RCo 5 compounds . . . . . . . . . . . . . . . . . . . . . 4.6. Dilute rare earth systems . . . . . . . . . . . . . . . . . . . . . 4.6.1. R a r e earths diluted in cubic noble-metal hosts . . . . . . . . . . . 4.6.2. R a r e earths diluted in cubic pnictides . . . . . . . . . . . . . . 4.6.3. Other dilute systems . . . . . . . . . . . . . . . . . . . . 4.7. Miscellaneous compounds . . . . . . . . . . . . . . . . . . . . 5. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. G e n e r a l analysis . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1. Coherency of the various determinations . . . . . . . . . . . . .
3
49 49 50 50 50 50 52 53 54 56 57 61 61 64 67 67 69 70 72 72 75 77 79 79 80 80 81 82 83 83 84 85 85 90 92 93 95 95 95 98 100 102 102 102 106 106 107 110 111 113 113 113
4
P. M O R I N and D. SCHMITI"
5.1.2. Comparison of the magneto-elastic coefficients . . . . . . . . . . . 5.1.3. Comparison of the quadrupolar pair-interaction coefficients . . . . . . . 5.2. Origin of the quadrupolar interactions . . . . . . . . . . . . . . . . 5.2.1. The one-ion magneto-elastic coupling . . . . . . . . . . . . . . 5.2.2. Quadrupolar pair interactions . . . . . . . . . . . . . . . . . 5.3. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.1. Symmetrized Stevens operators for the cubic symmetry . . . . . . . . . . A.2. Fourth- and sixth-rank one-ion magneto-elastic Hamiltonian ~ME~ . . . . . . . A.3. Perturbation theory: the tetragonal mode 3' . . . . . . . . . . . . . . A.4. Perturbation theory: the trigonal mode e . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .
114 116 117 117 119 120 122 122 123 124 125 126
1. Introduction
The study of multipolar interactions concerning the unfilled 4f shell has known a great development in rare earth (R) intermetallic compounds. It involves the complex problems of both the crystalline electric field (CEF) acting on a 4f ion and the pair interactions between different ions. The electron-lattice coupling can drive very characteristic magneto-elastic features, in particular in elastic constants and in magnetostriction, through symmetry lowering modes which depend on the 4f electronic states. According to the Jahn-Teller theorem, the magneto-elastic modulation of the CEF corresponds to a gain of energy larger than the elastic energy lost at the thermodynamic equilibrium. These properties have been extensively studied in the past in many systems exhibiting an orbital degeneracy as reported for cubic spinels (Englman and Halperin 1970, Englman 1972) in particular mixed chromites (Kataoka and Kanamori 1972). Gehring and Gehring (1975) and Melcher (1976) have provided the literature with very complete reviews of the cooperative Jahn-TeUer transitions occurring in rare earth insulators within the tetragonal zircon structure. All these reviews show the fundamental importance of studying elastic constants. Indeed they couple, as thermodynamical variables, the strain to the ultrasonic stress in the presence of a quadrupolar moment, describing the orbital character of the 4f shell. Another important feature is the large success met by the mean-field approximation for describing thoroughly the structural transitions. More recently, very large symmetry lowerings, which are mainly of one-ion origin have been observed to occur at the magnetic ordering in rare earth intermetallics (L6vy 1969, Morin et al. 1979). In the example of RFe 2 compounds which are ordered at room temperature, they are large enough to motivate industrial applications as magnetostrictive transducers which are able to compete with ceramics (Clark 1980). In all the series, they are associated with large softenings of elastic constants in the non-ordered phase (Mullen et al. 1974, Moran et al. 1973). From a theoretical point of view, they have required the development of a quantum treatment which takes into account all the characteristics of the 4f wave functions, as will be discussed in the present chapter. However, in R intermetaUics an additional coupling has been revealed, mediated from one site to the others by the conduction electrons. This quadrupolar pair interaction may dominate the magneto-elastic coupling and drive a
6
P. MORIN and D. SCHMITI"
quadrupolar ordering in the paramagnetic phase, as occurs in TmZn and TmCd (Morin et al. 1978b). According to the sign of the quadrupolar pair interactions, ferro- or antiferroquadrupolar orderings have been observed. The symmetry lowering is then, when it exists, a consequence of the quadrupolar symmetry through the magneto-elastic coupling. Such compounds are not stricto sensu Jahn-Teller compounds, because then the magneto-elastic coupling is the driving mechanism. In the presence of dominating bilinear interactions, which is the majority of cases, quadrupolar interactions act on the minimization of the total free energy, and magnetic properties cannot be understood without them in several series (Giraud et al. 1985). For different experimental techniques (Ltithi 1980a), the magneto-elasticity remains the most appropriate probe for determining the quadrupolar pair interactions. This chapter is organized as follows. First, the quantum Hamiltonian is built, which describes both the one- and two-ion, spin and quadrupolar couplings in systems of cubic, tetragonal and hexagonal symmetries. The power of the susceptibility formalism, which allows independent determinations of the various coefficients, is emphasized (section 2). The main experimental evidence of quadrupolar orderings is presented in section 3, before giving the quadrupolar consequences on the magnetic properties. The experimental techniques which are based on the susceptibility formalism of section 2 are illustrated. Section 4 gives a review of the magneto-elastic and pair interaction coefficients which have been determined in the rare earth intermetaUics. Due to the anisotropic character of the magneto-elasticity, studies on single crystals are of fundamental importance and these are the only ones considered here. Section 5 gives an analysis of all these results from both a macroscopic and a microscopic point of view.
2. Formalism
The existence of non-quenched orbital moments in rare earth ions is the source of their anisotropic magnetic properties. A classical description of these properties is no longer valid. In a quantum formalism, the 4f shell is represented by wavefunctions expanded in the IJ, M:) basis. The appropriate Hamiltonian, acting on these wave functions, depends on the point symmetry of the rare earth site under consideration. In addition to the crystalline electric field (CEF) Hamiltonian, any kind of interactions acting on the rare earths or between them (Heisenberg coupling, magneto-elasticity, quadrupolar pair interactions . . . . ) must be considered. In the literature, most of the experimental results reported on the magnetoelastic properties concerns rare earth compounds with cubic symmetry, because for these compounds the number of parameters required for analyzing these properties is the smallest. In the following two sections, we extensively develop the formalism and the treatment of magneto-elasticity and quadrupolar interactions for cubic symmetry. The case of lower symmetries (hexagonal and tetragon-
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H I N T E R M E T A L L I C S
7
al) will be discussed more briefly in section 2.3, while the connection with the classical formalism will be made in section 2.4. 2.1. The cubic Hamiltonian 2.1. i. General expression
The Hamiltonian appropriate for cubic symmetry includes several terms. By using the operator-equivalent method (Stevens 1952) they are written as: ffb°= ~CEF + ~°Z + ~B + ~OQ+ '~ME1 + ~ME2 "l- (Eel + E B + EQ q- EME2) .
(I)
The different terms represent successively the crystal field term, the Zeeman coupling, the Heisenberg-type bilinear interaction, the two-ion quadrupolar term, the magneto-elastic coupling, the elastic energy, and bilinear and quadrupolar corrective energy terms. Each term is described in detail below. The crystalline electric field Hamiltonian: Y(CEF According to different authors, this Hamiltonian takes one of the following forms. The x, y and z axes have been chosen as the four-fold axes in the cubic crystal: ~ C E F = B4(O°4 + 5 0 4 )
-]- B6(0°6 - 2104)
= A4 ( r 4 ) ~ 1 ( 0 40+ 5 0 4 ) + A 6 ( r 6 ) y 1 ( O ~ - 2 1 0 4 = __
Wx (004 -~- 5 0 ~ ) +
F4
W(1 - - I X l )
F6
0
)
(2)
4
( 0 6 -- 2 1 0 6 ) .
The O~'s are the Stevens operators in the formulation of Hutchings (1964). 13I and 7i are the 4th- and 6th-order Stevens coefficients, the value of which depends on the rare earth. The CEF parameters A4
and A6(r 6> originate in the electric field acting on the 4f shell, which arises from the surrounding point charges (Hutchings 1964) or from the conduction electrons (Schmitt 1979). In the last formulation, W represents a scale factor, and x the relative proportion of 4th and 6th order CEF terms (Lea et al. 1962). The Zeeman Hamiltonian: Y(z ~ z = --gJtzB H" J ,
(3)
represents the Zeeman coupling between the 4f magnetic moment M = gj/x B( J ) and the internal field H (applied magnetic field corrected for demagnetization effects). The bilinear Heisenberg-type Hamiltonian: ~ ~fB = -- g J IxBnM " J ,
(4)
8
P. MORIN and D. SCHMITT
represents the Heisenberg-type bilinear interaction, written within the molecular field approximation (MFA), n being the bilinear exchange parameter. An alternative definition for this parameter is 0", with: O* n
=
-C
3kBO* =
(5)
g :2t z B2 J ( J + 1)
where C is the Curie constant. The two-ion quadrupolar Hamiltonian: ~ o In a form analogous to the Heisenberg-type bilinear coupling between magnetic dipoles, a two-ion coupling between 4f quadrupoles is possible (see Schmitt and Levy 1985 and references therein), for which evidence has been found in several rare earth intermetallic compounds (Levy 1973, Levy et al. 1979). According to symmetry considerations, the corresponding Hamiltonian, again in the MFA, may be written as: ~Q = -KY((O°)
O0 + 3 < 0 2 ) O 2 ) - Ke(Pxy + < P y z > G z + ( P z x ) P z x ) •
(6) In this expression, 0~, 022 and the Pi:'s are the second-order (quadrupolar) Stevens operators (Hutchings 1964, Sivardi6re 1975) (see table 1), and K ~ and K ~ are two-ion quadrupolar parameters corresponding to the two linear combinations of quadrupolar operators, which are invariant to all the cubic symmetry operations. As will be discussed below, they are associated with the two anisotropic normal strain modes in cases with cubic symmetry, namely the tetragonal (3/) and trigonal (e) modes. TABLE 1 Expression of the quadrupolar operators. Operator
Expression
0°2
3JZz - J(J + 1)
o~
J~ - s,2 = ~(s~++ :_)
exy
l(JxJy + JyJx) = ~- (J+ - j2_)
--i
Pyz
e zx ~_
1
0 2
2
~(Y~L + L :,) ½(LL+LL)
e
The magneto-elastic Hamiltonian: ggME The magneto-elastic properties of rare earth compounds have been extensively studied since the 1960s (Callen and Callen 1965, Du Tr6molet de Lacheisserie 1970, Du Tr6molet de Lacheisserie et al. 1978). The analysis of these properties employs one-ion and/or two-ion magneto-elastic Hamiltonians, it is usually
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
9
limited to the first order in the strain (harmonic approximation) and contains terms up to the sixth rank for the Stevens operators. Few studies have been made in connection with second-order terms in the strain (anharmonic coupling) (Du Tr6molet de Lacheisserie et al. 1978, Rouchy and Du Tr6molet de Lacheisserie 1979, Wang and Lfithi 1977b) and this second-order magneto-elasticity will be not considered in this chapter. The one-ion magneto-elastic Hamiltonian corresponds to the direct coupling between the deformations of the lattice and the 4f shell. It can be considered as the strain derivative of the crystal field Hamiltonian. The two-ion magnetoelasticity is related to the modification of the two-ion magnetic interactions by the strains. The one-ion magneto-elastic H a m i l t o n i a n : ~fME1
Group theory gives us the correct procedure for constructing this Hamiltonian. First, one writes linear combinations of the first-order strain components eq which transform according to the three irreducible representations F1, F3 and F5 or a, y, e (see table 2) (Du Tr6molet de Lacheisserie 1970). Second, one considers the linear combinations of the Stevens operators of rank I (l = 2, 4 or 6) which transform also according to the irreducible representations of the cubic group (see appendix 1). Finally, one obtains all the possible invariants by summing the direct products of the different combinations which belong to the same representation. A magneto-elastic coefficient then is associated with each invariant. For l = 2 , two magneto-elastic terms are obtained, coming from the two representations /"3(7) and Fs(e), WMEa(I = 2) = Y ~ m ( l = 2) + aq~°hEl(l 2 ) , -----
(7)
with ~IEI(/
= 2)
e
=
--
~, 2 X/3e202),
B~(e~O ° +
Y(MEI(I = 2) = - B
8
e
e
(e,Pxy + e2Py ~ + e3Pzx ) .
(8)
(9)
Thus there are only two normal strain modes associated with symmetry lowerings of the cubic cell, one for a tetragonal and the other for a trigonal strain. TABLE 2 Normalized, syrnmetrizedstrains for the cubic symmetry. Representation
Straincomponents
r~
E° = ~
1"5
E1 = %¢/2 x y '
1
(Ex~+ ~, + ~zz)
E2 = " ~ Eyz'
10
P. MORIN and D. SCHMITT
The l = 4 and l = 6 one-ion magneto-elastic Hamiltonians have also been given (Creuzet and Campbell 1981, Niksch et al. 1982). Their full expressions are given in appendix 2.
The two-ion magneto-elastic Hamiltonian: ~ME2 Although any kind of two-ion interaction can a priori depend on the strains, only the bilinear coupling will be considered here. The same theoretical considerations as above allow us to write the corresponding Hamiltonian as the sum of an isotropic term and two anisotropic ones: ~ME2
=
o~ ~ME2
+
ffb'O~vlE2 "[- ~ M E 2
(10)
"
The first term, (11)
~ME2 = -D~e~ ( J ) " J ,
is related to the volume dependence of the isotropic bilinear exchange interactions. On the contrary, the two other terms,
Yg~E2 =
-Dr[e~'(2(Jz)J~
)~ME2
-De[eel( ( Jx)Jy + ( Jy) Jx) + 1=:2(( J,)Jz + ( Jz)Jy)
-
(L)L
- (4)4)
+ v3e~((L)L
- (4)4)],
(12) =
+ e3((Jz)Jx + (Jx)Jz)],
(13)
are associated with the appearance of anisotropic bilinear interactions under a strain. As for the bilinear coupling itself [eq. (4)] the MFA has been used for this two-ion magneto-elasticity.
The elastic energy: Fez The elastic contribution to the free energy may be expressed as a function of strains and symmetrized elastic constants. In cubic symmetry, the three normal strain modes a, y, e appear in the expression of the elastic energy 1 .,-,al a x 2 1 3' Y 2 Eo, = ~,~ot~ ) + ~c0[(~,) + ( ~ D 21+ 1 Co[(e~) ~ ~ 2 + (e~) 2 + (e~)2].
(14)
The C~ are background elastic constants without magnetic interactions; they are related to the conventional constants by a 0 C O ~-- C 1 1 +
0
.
2C12 ,
C~
:
0 C11 -
0 C12
. ,
s
0
C O-= 2C44
.
(15)
The corrective energy terms E~, EQ, EMe 2 These terms arise from the fact that each rare earth pair is counted twice in the MFA treatment of two-ion magnetic couplings. This occurs for the bilinear [eq. (4)], the quadrupolar [eq. (6)] interactions, and for the two-ion magneto-elastic
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
11
terms as well [eqs. (11)-(13)]: E B = ½nM 2 ,
(16)
Eo=½K'((o°)Z+3(OZ2>z)+½K~(
(17)
EME2 = ½D"e~<J>2 + ½DV[e~(2<jz>2_ <Jx>2_ <jy>2)
+ V~E~(<jx>2__ <jy>2)]_~_De[E~<Jx><jy>q_E;<Jy><Jz> ~- e;<Jz) <J~>] .
(18)
2.1.2. Harmonic description of magnetostriction Starting from the above Hamiltonian, the actual magnetostriction, i.e. the equilibrium values for the various strains e~'s, may be obtained by minimizing the free energy F with respect to each of them: OF/Oe ~= O. That leads to the following general expressions for the 6 symmetrized strains for cubic symmetry ~, 0 + 5 0 4 > + B(6) Ea = ~1 [B(4)
./
1
+ ½V,~<j>2]
y,1
+ ½D~(2(jz)2_ <jr>2_ (jr>2)], =
(19)
(20)
'Y T,2 1 [~/-~B.V+ B~4)+ B(6)(O 6 >
- -
+ ½V~D~((j~>2_ <jy)2)],
(21)
~. l~l [()el,l\ ~2 81 = ~1 [Be+ B(4) + u(6)\~ 6 / -}-B(6)
+ m~(Jx)(Jy>], 1
E2 : ~ 0
(22)
[Be(PYz>+ B (~4 ) < O 4 ' 2 >
l:?a:l / /-'}el,2 ~ e2 062,2> + u(6)\"6 / -~- B(6)<
+D~(J,)(Jz>],
(23)
1 ~ el O;1,3 ]~,52 [ /~e2,3 k e 3 = ~ g [ B ~ ( P ~ > + B(4) -4- B(6)< > + u(6)\,j6 / Co
+ D~(L)
(24)
The macroscopic change of length of a magnetic lattice with cubic symmetry may then be written as a-
8l 1 1 l - V~ e~ + ~
v
2
2
e1(2/3z -/3x - / 3 2 ) + ~
+ V~(e~/3~/3y + e~/3,/3 z + 23/3~/3x).
1
~
2
e2(/3" - / 3 2 ) (25)
12
P. MORIN and D. SCHMITF
To obtain a reliable description of the various magnetic and elastic properties, the expectation values of the Stevens operators must be treated in a self-consistent manner. In particular, this yields the actual temperature dependence of the various strains [eqs. (19)-(25)] in the paramagnetic as well as ordered phases.
2.1.3. The total quadrupolar Hamiltonian By including the preceding equilibrium values of the strains in the starting Hamiltonian, one can write it in a form which does not explicitly depend on the e~'s, but only on the magneto-elastic coefficients and on the expectation values of the various Stevens operators. In most of the cases reported in the literature, one-ion magneto-elasticity involving the quadrupolar operators appears as the dominant contribution. This justifies focusing our attention on it in the following. The relative importance of other terms from an experimental point of view will be discussed later, in particular in sections 3.8 and 3.9. Keeping only the B " contribution leads to an expression for NMEI(1 = 2) which appears to be formally undistinguishable from the two-ion quadrupolar Hamiltonian No; both terms can be grouped together in the total quadrupolar Hamiltonian NOT fft°QT= -G'Y[(O~)O°2 + 3 ( 0 2 2 ) 0 2] - G [( p xy)Pxy ~- (Pyz)Pyz -I- (P~x)P~x] (26) where the new total quadrupolar parameters G ~ are defined as G ~ - (B~) 2 -
-
cg
- - -
C;
~IE ~- K ~ '
,
(27)
+ K ~ = GME + K ~ .
(28)
+
K "~ =
G
E
Equation (26) provides an alternative expression for the quadrupolar couplings, which may be used instead of eqs. (6), (8) and (9), every time that the e "'s are not explicitly needed.
2.2. Treatment of the cubic Hamiltonian The treatment of the Hamiltonian described above can be carried out in two ways: i) First, by diagonalizing the full Hamiltonian in a self-consistent manner. This allows the description of all the magnetization and magnetostriction processes, in particular in the ordered phase or under large external stresses (magnetic field or uniaxial stresses) as shown in section 3. ii) The second possibility is to apply perturbation theory to the above Hamiltonian in order to describe the magnetic and magneto-elastic properties in the
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H INTERMETALLICS
13
non-ordered phase and under small external stress. There are two advantages working in the cubic paramagnetic phase. First, analytical expressions can be obtained for the various susceptibilities connecting the various variables. Second, any strain mode can be systematically studied by applying external stress along the appropriate direction, whilst in the ordered phase only the spontaneous state may be usually achieved because of the strong anisotropy effects. In this section, the susceptibility formalism is developed by applying perturbation theory to the above Hamiltonian which is limited to second-rank quadrupolar terms. An analytical expression will be derived for the free energy, defining the various susceptibilities which appear in the analysis of the magnetization, parastriction and elastic constants measurements. Finally, the behaviour of these susceptibilities and their role in describing phase transitions will be described. 2.2.1. Perturbation theory
The zeroth-order Hamiltonian Y(0 consists of the CEF term [eq. (2)], the diagonalization of which provides the zeroth-order energies and CEF wave functions. According to the direction of external stress, the two normal cubic strain modes may be fully decoupled. Therefore they are considered separately. Tetragonal m o d e y
For a magnetic field applied along a four-fold axis, e.g. z, the only non-zero terms are M = g H a a 3 ( J z ) , Q = ( 0 2°) and e~. The perturbating Hamiltonian is then written as: YU1 = - g j l - % ( H + n M ) J z - (B3,e~ + K3,Q)O°2 + ~-~'0~,,'~3,to3,,= )~ + ½ n M : + -½K3,Q2 .
(29) Perturbation theory has to be carried out up to the second order for e~ and Q and up to the fourth order for H and M in order to derive analytical expression of the generalized Gibbs free energy F = - k B T l n Z. Here Z = T r e x p ( - / 3 Y ( ) is the partition function, k B the Boltzman constant and/3 = 1/(k B T ) . The details of the calculation may be found in appendix 3. The free energy expansion takes the form F3,= FVo - ½Xo(H + n M ) : -
~X3,tt~ 1 ,,,3, ea3, + K3,Q) 2
1. (3)(H + n M ) 4 - (2)tB3,e3,, + K 3 , Q ) ( H + n M ) 2 - zx3, \ •
- - , , I [ 3,
+ ~a"-'o,,ele~'t^3,'2)+ l n M 2 + 1K3,Q2 + . . . .
(30)
This expression includes four CEF single-ion susceptibilities which are defined without any magnetic or quadrupolar interactions, and depend only on the non-perturbed CEF state, i.e. on the cubic CEF level spacings and on the matrix elements of Jz and O ° between cubic CEF wavefunctions (see appendix 3). X0 is the usual (first-order) magnetic susceptibility, X3, is a quadrupolar strain susceptibility, X3, - (2) is a quadrupolar field-susceptibility which couples the quadrupolar
14
P. M O R I N
a n d D. S C H M I T T
operator O ° to the magnetic field, xz, - (3) is the third-order magnetic susceptibility which is related to the initial curvature of the magnetization curves and, finally, F 0 is the zeroth-order free energy corresponding to ~fcEF alone. The equilibrium values for M, e~ and Q are found by minimizing the generalized free energy F3, OF v
.
.
.
OM
OF3,
.
Oe~
OF3,
.
.
0.
oO
(31)
This yields a system of three coupled equations, the solution of which leads to the following relations M=
(3)11 u3 , XMH + XM
(32)
.
B ~'
~=Q ~ C~
(33)
'
Q = xoH 2 ,
(34)
XO
(35)
with XM -- (1 -- nXo) '
["~
1
X(M 3)=
(1
- nx0) 4
+ 203,
xv
(x~) ~ ] 1 - GrX3,J '
(36)
X(2) 3'
Xo
9
(1 - ~ X o ) 2 ( 1 - 6 % )
(37)
Note that we find for the equilibrium value of e~ the same expression as that found without recourse to perturbation theory [eq. (20)]. In the limit of zero field, OF3,/OQ = 0 leads to the strain dependence of the quadrupolar operator: B3,X3, -
Q = 1-
-
?
(38)
K:'X3, e l '
and the expression for the corresponding elastic constant can be written as: d2F 3, C3,= d(e~)2 = C~
( B y) 2Xv (1 - K~'X~,) '
(39)
or
Cy
C--~ = (1 - Grxr)/(1 - K~Xr).
(40)
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
15
Trigonal m o d e e
When the magnetic field is applied along a three-fold axis, e.g. [111], the new symmetry (trigonal symmetry) implies that the only non-zero values involved are 1 M ( J x ) = <Jy> : <Jz> - V ~ -gJ~£B
,
(41)
( P x y ) = (Pyz) = ( P z x ) = P ,
(42)
and e 2~ =
E 1 ~
e 3~ ~
e"
(43)
The perturbating Hamiltonian takes the following form 1
~1 =
V ~ gJld"B(H "[- lq'm)(Jx -[- Jy -Jr Jz) 3 ~-~e[
ex2
- (B~e ~ + K P)(Pxy + Pyz + Pzx) + ~ C o t e ) + ½nM2 + 3 K ' p 2
.
(44)
For the calculations it is more convenient to rotate the coordinate axes so that [111] is the new z-axis (see appendix 4). In this new axes system, the same perturbation theory may be carried out to obtain the trigonal free energy F ". Then applying an inverse rotation allows us to express the latter in the initial coordinate system with four-fold symmetry F ~ = Fo _ ½Xo( H + n M ) 2 - ~ x9~ ( B
~e + K ~ P ) 2
- 3X~2)(Bee" + K ~ P ) ( H + r t M ) 2 - 1X(e3)(n + n M ) 4 3~e," + ~t~ol, e 8 \)2 + ½ n M 2 + 3K~p2
(45)
Here X0 is the same as in F r [eq. (30)] because of the isotropy of the (first-order) magnetic susceptibility, but the other CEF susceptibilities X~, )~z) and x~" (3) differ from the corresponding X~i) 's (see appendix 4). This yields the anisotropy of the properties related to the quadrupolar variables. The equilibrium conditions for M, e ~ and P are OF ~
OF ~
OF t
OM
Oe ~
3P
-0,
(46)
whence the following expressions .
(3) Lr3
M = XMH + X M I,
e = ~BP ,E Co
p
=
XpH 2
,
(47) (48) (49)
16
P. MORIN and D. SCHMITT
with XM as in eq. (35), F
1
X(M3 ) -
[)(~3)
4
+6
(1 - n X o )
1 - 3G~x, ] '
X(2) Xe = (1
-
nx0)2(1
(51)
3G~x~)
-
(50)
In the limit of zero field, OF~/OP leads to the strain dependence of the trigonal quadrupolar operators, 3B~x~ e ~ P = 1 - 3K~x, '
(52)
and the corresponding elastic constant, 1 d2F"
c . . 3 . d(e~) . .e
Co
C~ CO
•
3(B~)ZX~ 1 - 3K~x, '
(53)
or
1 - 3G~x~ ~ 1 - 3K X,
(54)
2.2.2. Generalization of the susceptibility formalism Symmetry considerations allow us to easily extend the preceding expressions for a magnetic field in any direction. For the third-order magnetic susceptibility, symmetry considerations show that there are two independent parameters in the expansion of the magnetization to order H 3 (Birss 1964), the general expressions are
Mx
x (3))I-I (I-Iy 2 + I-I z) ,
.~ (3) /At3 + . (3) Llr3
(3)
2
(3)
2
/ I ( M y I I Y ._[._ 1
(3) - XM~,)H,(H~ + H~Z), (3XM~
(3)/_/3 m z = X M , , / I I z .Jr- i
. (3) -- XMr)Hz(Hx + H y2) . (3XM~
.
.
(55)
. (3) and XM~ - (3) are the total third-order magnetic susceptibilities associated Here xa~:, with the four-fold and three-fold axes, respectively [eqs. (36) and (50)]. This allows us to predict the H 3 dependence of the magnetization for any direction; e.g., applying the magnetic field along any two-fold direction leads to the following relation
X~)l = ZXM~I-(3)+ ~,~,,(3) M~3 .
(56)
In a similar way, symmetry considerations lead to only two independent
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H I N T E R M E T A L L I C S
17
parameters in the expansion of the quadrupolar susceptibilities to order H 2 ( 0 2 0 ) ~ ~XQ(2H~ 1 2 (O~)__
- H~2 - H y2 ) ,
1 2 ~XQ(Hx-H2y),
(57)
= 3XeI-I~t-IJ" (ij = xy, yz, zx) , where X~ and Xe are given by eqs. (37) and (51). These latter relations are useful in the expression for the parastriction, i.e. for the relative change of length A [eq. (25)] for a paramagnetic crystal in an applied magnetic field (see section 3.3.3). 2.2.3. Behaviour o f the various susceptibilities From the expressions of the CEF susceptibilities defined above, it is obvious that the crystal field plays an important role in their temperature dependence, in particular at low temperature where only a few CEF levels are populated. First-order magnetic susceptibility: Xo It is well-known that X0 varies as 1 / T at high temperatures, and that deviations from this law may occur at low temperatures, according to two typical behaviours: (i) with a magnetic CEF ground state, X0 diverges at 0 K; (ii) with a non-magnetic CEF level as ground state, a van Vleck-type behaviour is observed (see inset of fig. la). Third-order magnetic susceptibilities: X~ " (3) Because of the numerous terms in the expression of x (3)~ (/x = y, e), a more complex behaviour may be expected. While the high-temperature variation always follows a 1 / T 3 law, the behaviour of X- ~(3) at low temperature depends on the nature of the ground state. With a magnetic CEF ground state, X- ~(3) diverges at 0 K (fig. la). In the case of a non-magnetic singlet ground state, a van Vleck-type behaviour is observed (fig. lb). A very particular situation arises with a non-magnetic doublet ground state (fig. lc): if the van Vleck-type behaviour is observed for the trigonal susceptibility (H II[111]), a positive divergence occurs for the tetragonal one (HII[001]), due to the presence of mixed Curie-van Vleck terms in the expression of A~y - (3) • Thus, in addition to their anisotropy, the different (3) . (3) low-temperature behaviours oi Xr and x~ give valuable information about the CEF, in particular on the nature of the ground state. Strain susceptibilities: X~ The quadrupolar strain susceptibilities X~ give information about the quadrupolar character of the lowest CEF levels in the same way as X0 gives information about the magnetic character. Thus, for a singlet (fig. 2b) or a Kramers doublet (fig. 2d) ground state, there is no intrinsic quadrupolar moment and the low-temperature behaviour is of van Vleck-type. On the other hand, for a magnetic level as a
18
P. M O R I N and D. SCHMITI"
10
c° "~ I0
0
."
¢ .. /" --
0
20
40
0
q
20
TEMPERATURE
2ooi
-10
[.;t,)
--
40
(K)
Fig. 1. Typical temperature behaviour of the third-order magnetic susceptibility x~ - (3) for the two modes 3' (tctragonal) and e (trigonal) of cubic symmetry, and with a magnetic triplet (a)~ a singlet (b) and a non magnetic doublet (c) as ground state (CEF levels spacings are indicated). Inset of (a): corresponding temperature behaviour of the reciprocal susceptibility 1/Xo (dashed line: behaviour without CEF). (a)
'
--
G
- -
F,
'
r;'~ 1
80
J=6
~0
10
v
i
!
(e)
-
-
r,,
i
(d)
~ --
~5
J=6
20C
"
\
j =~
r;' r;" 100
10(
i
100
TEMPERATURE
'
lb0
(K)
Fig. 2. Typical temperature behaviour of the strain susceptibility X~ for both modes y and e, with a magnetic triplet (a), a singlet (b), a nonmagnetic, quadrupolar doublet (c) and a magnetic, nonquadrupolar Kramers doublet (d) as ground state. C E F levels spacings are indicated.
QUADRUPOLAR
E F F E C T S IN R A R E E A R T H I N T E R M E T A L L I C S
19
triplet/,~t) (fig. 2a) or a quartet Fs, there is an intrinsic quadrupolar moment and the X,'S diverge at 0 K. Finally, the non-magnetic doublet F 3 (fig. 2c) is quadrupolarly active for the 3' mode but inactive for the e one, leading to a Curie and a van Vleck behaviour for X~ and X~, respectively. In all cases, the high-temperature variation of X, follows a 1/T law, as does X0-
Quadrupolar field-susceptibilities: X ~) The low-temperature variation of quadrupolar field susceptibilities depends on both the magnetic and quadrupolar character of the ground state. Thus a pure van Vleck-type behaviour is expected only for a non-magnetic and non-quadrupolar level, i.e. a singlet (fig. 3b). For a Kramers doublet, the presence of an intrinsic magnetic moment leads to the divergence of x~, - (2) (fig- 3d). For the non-magnetic doublet F 3 (fig. 3c), the van Vleck-type behaviour is observed only for the e mode for which there is no corresponding intrinsic quadrupolar moment. For the other cases, a divergence occurs at low temperatures. A very special case arises for the triplet F~ t) (fig. 3a), where the competition between the various components of the level leads to a change of sign of x- r(2) as a function of temperature: this leads to the so-called "reverse magnetostriction effect" (see section 3.4.1). Also, at high temperature, a 1/T 2 law is expected.
(a) 20
- -
r+
_ _
F'~ 1
8
100
\
--->
I --
r,,+++
--r]
10
1
----[-i" J=B
--Q J= 6
P,i
\ (b)
0
T
0 v
i
-1(
i
(c)
- -
r,
--15.
~>~
- -
F~2)
J=6
2°°~T
F~
40
t/ I\
\\+
100
TEMPERATURE
'++q
,.,
L = ,-+,,
20
100
( K)
Fig. 3. Typical t e m p e r a t u r e behaviour of the quadrupolar field-susceptibility X~ ) for both m o d e s 3' and e, with the same ground states as in fig. 2.
20
P. MORIN and D. SCHMITT
2.2.4. Magnetic and q u a d r u p o l a r transitions
In the presence of quadrupolar interactions, rare earth systems are characterized by up to five quadrupolar order parameters (Sivardirre et al. 1973) in addition to the usual magnetic order parameter M = g J P ~ ( J z ) . For cubic symmetry, there are only two independent order parameters, namely Q = (O °) and P = ( P i i ) , associated with the two quadrupolar parameters G r and G ~. In most cases, the bilinear interactions are dominant and drive one single magnetic phase transition, since the quadrupoles necessarily follow the dipoles as they magnetically order. Nevertheless, the quadrupolar interactions may influence the nature of this transition (see section 3.2). In a few cases, the quadrupolar interactions are dominant and drive a quadrupolar ordering without any magnetic dipolar ordering (see section 3.1). Several authors have investigated the phase diagrams associated with the presence of magnetic and quadrupolar interactions, with the frequent restriction of considering effective spins, i.e. by taking into account only the low lying levels. The cases S = 1 (Chen and Levy 1973), S = 3 (Sivardirre and Blume 1972) and S = ~ (Chen and Levy 1971) have been treated. Also the case of the F3-F5 magnetic system (S = 2) has been considered, explicitly including the related crystal field effects (Ray and Sivardi~re 1978). These calculations were all performed within the MFA. Numerous phase transitions have been found to occur, depending on the strength and the sign of the various parameters. Uniaxial ferro- (FQ) and biaxial antiferroquadrupolar (AFQ) orderings have been described (Sivardirre et al. 1973). More complex situations have been calculated by simultaneously considering the quadrupolar interactions belonging to both modes y and e, e.g., a ferriquadrupolar phase associated with the Q component followed by a FQ phase related to the P component (Chen and Levy 1973). In all the cases, the nature of the phase transitions was discussed and triple or tricritical points were found. While the existence of the various magnetic and quadrupolar phase transitions can be qualitatively described within the effective spin formalism, accounting for the real situations found experimentally requires one to consider the actual CEF level schemes, since the precise nature of the ground state as well as possible effects of excited levels via van Vleck-type matrix elements may noticeably change the physical properties. This can be done by using the susceptibility formalism developed above (Morin and Schmitt 1983a). Indeed, the expression for the free energy derived in section 2.2.1 can be used to investigate the possible phase transitions by taking H = 0. Equations (30) and (45) are then written as F v = F~ + ½n(1 - n X o ) M 2 - in4-,~,(3),~ar4 ~,~ + ½G~'(1- G r x r ) Q 2 - n 2r,'y t 1 X r(2) Q M 2 ,
F ~ = F 0 + ½n(1 - n x o ) M 2 - zin4"a t(3)hAr4 1,1 + 3G~(1-3G~x~)P 2 - 3n2G~2)pM
z"
(58)
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
21
According to Landau theory, magnetic (tD) and quadrupolar (to_ and te) critical temperatures are defined by canceling the coefficient of the M 2, Qz or p2 terms in the expansion of F 1 - nXo(to) = 0,
1 -- G~Xr(tQ) = O,
1 - 3G~x~(tp) = 0.
(59)
The actual phase diagram as well as the nature of the transitions is determined by the relative magnitudes of tD, tQ and t e and by the signs and relative magnitudes of the other coefficients. The occurrence of a quadrupolar phase transition is a particularly interesting situation in cases with cubic symmetry where no quadrupolar component can spontaneously exist. In such a case (to_> t D for example) the transition is generally first order because the next term in the Q expansion of the free energy is of odd parity (Q3). The ordered value of Q is positive or negative depending on the sign of the corresponding coefficient. This coefficient is a second-order strain susceptibility [X(s2) in the notation of Morin and Schmitt (1983a)] and involves quadrupolar matrix elements between the various CEF levels. Note that the quadrupolar transition may be second order if Xs" (2) = 0 and the next term in the Q expansion of the free energy is positive. This occurs when the ground state is a quadrupolar doublet (F3 doublet, /"4, F5 triplets or F8 quartet) (Kanamori 1960). Examples of pure quadrupolar phases are shown in fig. 4 for both modes 3' and e (phases a 0 and ae, respectively). In the presence of dominant bilinear interactions (t D > tQ and tp), the magnetic transition is first- or second order, according to the sign of the next term in the M expansion of the free energy, namely x~ -(3) (/x = y, e). As seen in a previous section, xg -(3) is usually negative, whence a second-order magnetic transition. However, with a F3 doublet as ground state, xz, -(3) becomes positive at low temperature (see fig. lc), according to the characteristics of the low-lying CEF levels, leading to a first-order transition (see fig. 5). The limit between the two situations, i.e. xv - ( 3 ) = 0 , corresponds to the tricritical point C in the phase diagram. In the presence of competing bilinear and quadrupolar interactions, the situa-tion is more complex. First, various magnetic and Q or P quadrupolar phases may exist according to the relative strength of the associated coupling parameters n, G ~, G ~ (see fig. 4). Note in particular the possibility for two types of magnetic phases: (i) the phases b 0 or b e correspond to magnetic order where the ground state, coming from the F~ 1) CEF triplet, has mainly a Mj = 5 component; (ii) the phase co_ may be considered as a magnetic phase induced on the non-magnetic level, which is the ground state in the quadrupolar phase ao, and which has a dominant M : = 6 component. Another interesting feature of competing bilinear and quadrupolar couplings is the modification of the nature of the magnetic transition by the quadrupolar interactions. Indeed, the character of the phase transition is now determined by the sign o f -X M,'O (3) = X~)In=0 (see Morin and Schmitt 1983a). According to eqs. (36)
22
P. M O R I N and D. S C H M I T T
a) 5, / ~ ~ '
b) (]*=0 ~ aKp ~ 8'=2K
5 •
O'=4K/
W 0..
5"
,
)-
0'~=6K
o'=gK
bQ ,fQ 0
10
G I (mY,)
bp 0
100
G2(m K)
Fig. 4. Magnetic and quadrupolar phase diagrams for various bilinear coefficients 0* as a function of the quadrupolar parameters G ~ (=-G 1, left part) or G ~ (-=G 2, right part) for a T m 3+ ion (J = 6). a o, ap = pure quadrupolar phases; bo, bp = magnetic phases; Co: induced magnetic phase (from Morin et al. 1987c).
and (50), a positive quadrupolar parameter may drive the transition to be . (3) is larger than ]X~)[ (see fig. 6). Also first-order if its contribution to xM,0 tricritical points may be observed. The opposite situation may also occur, where a .(3) leads to a first-order transition, but a negative quadrupolar positive value of x~ . (3) changes it into a second-order one. Several examples of this contribution to XM,O will be discussed below (see section 3.2.2). Note that most of the experimental first-order transitions may be explained through the present model and do not require us to invoke critical fluctuations (see K6tzler 1984). In conclusion, the existence of bilinear and 3'- and e-type quadrupolar interactions gives rise to a variety of situations. Analyzing the complex related phase diagrams as well as the nature of the magnetic and quadrupolar transitions is qualitatively achieved within effective spin models, but requires a more realistic formalism in order to take the exact nature of the CEF levels involved into account correctly, and to better understand the experimental situations. This can easily be done through the present susceptibility formalism.
Q U A D R U P O L A R EFFECTS IN R A R E EARTH INTERMETALLICS
J=4 W=-5.SK x=.6B
(El 10
23
I-1 ~, ~ ? [ _ _ F 3
[001]
0 4c; 3O -
~
" -TD=tD
2C 1C I
0
/
I
20 TEMPERATURE(K)
40
Fig. 5. Temperature variation of xr - (a) (___X~3)) calculated for a Pr 3+ ion for the C E F level scheme indicated (upper part). Magnetic phase diagram (lower part); the dot-dashed line is the second-order magnetic transition line if neglecting X~3); the dashed line corresponds to a second-order magnetic transition, the full line (TD) to a first-order one; C is a tricritical point (from Morin and Schmitt 1983b).
~"
10 - -
:~
r 2
f2)
5
G1=13mK
--
r3
I 0 .-.
/
_'
10~-
/ I," 0
.
." T.-t IT.,
/
.~.-~_" . -
,
/
J=6 vv = u . ~ 3
n
x=- .34!
, , [oo ] 5 10 15 TEMPERATURE ( K )
20
Fig. 6. Temperature variation of X~)0 calculated for a Tm 3+ ion for the CEF level scheme indicated, for two G r (~G1) values (upper part). Lower part: magnetic phase diagram for G r = 13 inK; the solid (dashed) line represents a first- (second-) order transition; the tricritical point C occurs when XM,O - (3) vanishes; the dot-dashed line corresponds to a second-order transition which would occur if G ~= 0; T o is the quadrupolar ordering temperature, and t r the magnetic ordering temperature within the quadrupolar phase; T o and t o correspond to the first- and second-order magnetic transition, respectively (from Morin and Schmitt 1983a).
24
P. M O R I N
and D. SCHMITT
2.3. Extension to lower symmetries (hexagonal and tetragonal) The magneto-elastic and, more generally, quadrupolar properties have been studied less thoroughly in rare earth intermetallic compounds with a symmetry lower than cubic, e.g. in hexagonal and tetragonal compounds, because of the greater number of parameters involved in the CEF Hamiltonian as well as in the quadrupolar couplings. Only a few systems have been investigated such as metallic Pr (Hendy et al. 1979), dilute yttrium based rare earth compounds (Pureur et al. 1985), and more recently the compound PrNi 5 (Barthem et al. 1988). In this section, the formalism describing the magneto-elastic and quadrupolar properties of the rare earth compounds with hexagonal and tetragonal symmetries is developed. The full Hamiltonian is given, including all the relevant terms for both the one-ion magneto-elastic coupling and the two-ion quadrupolar interactions. Then analogies and differences with respect to the cubic symmetry are emphasized. 2.3.1. The Hamiltonian The general Hamiltonian appropriate for hexagonal and tetragonal symmetries takes the same form as eq. (1). However, the detailed expressions for most terms are different from those of cubic symmetry and are given below (see Morin et al. 1988). The CEF Hamiltonian The CEF Hamiltonian may be written, in terms of the Stevens operators (Hutchings 1964) and in the axes system where the z-axis is parallel to the c-axis of the hexagonal or tetragonal unit cell, as, ~Hex CEF =
~a'et
CEF =
0
0
0
0
0
0
6
(60)
6
B2 0 2 + B404 + B606 + B606 ,
0 0
0 0
4 4
0 0
4 4
B2 02 + B 404 + B404 + B606 q- B606 •
(61)
The bilinear Hamiltonian For hexagonal and tetragonal symmetries, the existence of one privileged direction, namely [001], dictates that there are two invariants in the full bilinear exchange coupling. Their expression is the same in both symmetries and takes the following form within the molecular field approximation ~ 1 = _ ( gjlJ.B)2nC~l ( j ) j , 2
~ B 2= --(g, lxB) n
~2
( 2 ( J z ) J z - (Jx)Jx - ( J y ) J y ) .
(62) (63)
The first term corresponds to the usual isotropic Heisenberg-type bilinear coupling, while the second term is an anisotropic bilinear coupling, forbidden in cubic symmetry, which leads to effective n coefficients which are different parallel and perpendicular to the c-axis in hexagonal and tetragonal symmetry. An experimental way to estimate the magnitude of this a2-term, would be to investigate the
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H I N T E R M E T A L L I C S
25
anisotropy of the magnetic susceptibility in the gadolinium compounds where there is, in principle, no other source of anisotropy. The two-ion quadrupolar Hamiltonian In hexagonal and tetragonal compounds, the O ° quadrupolar component is already ordered (i.e. (O2°) ¢ 0) by the crystal field, but that does not exclude the possible existence of quadrupolar pair interactions, as in cubic symmetry. However, due to the lower symmetry, more than two quadrupolar parameters are involved, and the corresponding Hamiltonian may be written, in the MFA, as a function of the linear combinations of products of second-order Stevens operators, which are invariant under the symmetry operations of the hexagonal or tetragonal point groups: ~Hex
o
=-K
a
0
0
2
(O2)O 2-KÈ[(O2)O2+4(Pxy)Pxy ]
- K~[(Pyz)Py~ + ( P z ~ ) P ~ ] , ~o Tet
~o
0
=
-
K~(02)02
0
7 -
-- K ~ [ ( P y z ) P y z
K
+
2
(64)
2
( 0 2 ) 0 z - K (Pxy)Pxy
(P~)P~].
(65)
The magneto-elastic Hamiltonian As in cubic symmetry, the one-ion magneto-elastic couplings may be considered as the strain derivatives of the crystal field term itself. Within the same harmonic approximation, and limiting ourselves to the second-rank (quadrupolar) terms, the full one-ion magneto-elastic Hamiltonian is written as:
~Hex ME1 = __(B,~le~l
~-
B,~2e,~2)O0 _ B ~(elO ~ 2z + 2e2Pxy ~ ) _ B¢(e~pyz + eIpzx) ,
(66) ~ TME1 et =_(B,~lal
+ B.aa2)
0 2 o --
B,s,O~_Baeapxy
--
B~(s~pyz+e2Pzx).
(67) In these expressions the e~'s are the symmetrized strains (Du Tr6molet de Lacheisserie 1970) (see table 3) and the B"'s the magneto-elastic coefficients associated with the corresponding normal strain modes. Note that there are two independent fully symmetric a-terms, related to the two deformations which maintain the initial symmetry, namely the volume al-strain and the axial a2-strain, the latter one is associated only with a change of the ratio c/a. For the other terms, a symmetry lowering towards orthorhombic or even lower symmetry Occurs.
The two-ion magneto-elasticity is related to the modification of the two-ion magnetic interactions by the strains. In this section, we restrict ourselves to the two a-type bilinear couplings. For each one, two strain modes must be considered; they are associated with volume and axial strain dependences of bilinear interactions. For hexagonal and tetragonal symmetries, the corresponding two-ion
26
P. M O R I N and D. SCHMITI" TABLE 3
Normalized, symmetrized strains for the hexagonal and tetragonal symmetry. Hexagonal
Tetragonal
Expression
Representation
Strain
Representation
Strain
5
~°~
/'1
~°~
-~v~(~ + ~. + ~z~)
5
~°~
5
~°~
~ ( ~ - ½(% + ~x))
5
4
r4
d
v~ %
F6
eg
F5
s;
V ~ G~
magneto-elastic Hamiltonians take an identical form, and may be written in the M F A as o~1
~ME2
~--- t'r~otl otl D ~ 2 e ~2 --~LIalE At- al
gME2a2 = --IZDaXa2 E ~
+
)(j)j
(68)
D~2 J e~2)(2( z a 2L )
-
(L)L
- (Jy)Jy).
(69)
The elastic energy The elastic contribution to the free energy may be expressed as a function of the strains s" and of the appropriate symmetrized elastic constants C~" ( D u Tr6molet de Lacheisserie 1970). For hexagonal and tetragonal symmetries, they are given by:
EHeX : 1 C 0 1 ( G a l ) 2 q_ CO12Eo~IEa2 + 1 C 0 2 ( E a 2 ) 2 el
+ ~Co[(~1)2 + (~)21 + ½Coq(~)2 + (~)21, ETet e, = 1
Co,(eal)2 +
(70)
C012EotlEo~2 + ~ 1t ~..~ao 2 ,ks -ot2\2 ) l z~/
6\2
1
e
e 2
+ ½G(~) 2+ : ~ o ~ ) + ~Co[(<) + ( 4 ) q -
(71)
The background elastic constants without magnetic interactions Co~ are given in table 4.
TABLE 4
Symmetrized elastic constants in hexagonal and tetragonal symmetry.
Hexagonal
Tetragonal
Expression
C al
C ~1
½(2Cn +2C~2 +4Ca3 + C33)
C~2 C~2
C~12 Co2
C~ C~
C~ C~
_
C ~
-~V~(G~ + c,~ - c . - G~) 3'(C n + C ~ 2 - 4 C ~ + 2 G 3 ) C n - C12 2G~ 2(2"66
QUADRUPOLAR
E F F E C T S IN R A R E E A R T H I N T E R M E T A L L I C S
27
The two-ion corrective energy terms As in cubic symmetry, the molecular field treatment of the two-ion interactions requires the usage of corrective terms in the deduced single-ion free energy. According to the two-ion couplings considered above, three corrective terms have to be taken into account, corresponding to the bilinear, the quadrupolar and the two-ion magneto-elastic Hamiltonian, respectively.
Hexagonal symmetry EB'= ln~lM2,
(72)
]IA2 2 = ~1 n,~2[,-) t .... z - Mx2 - My) ,
E~2
. 1KO: < o O > 2
EHeX
o
2 2
+ ½K"[
(73) 1 ' ~K [<eyz> 2 "~ 2 ],
+4
(74)
j~?Hex l,~cd ~1 _}_'rja2Ea2)<j>2_}_ l z r ~ l ~1 a2Ea2 ) I,/.)28 + O,~ 2 ~ME2 = 2[,Ual 8 ~ctl x
[2(L) 2- (L) 2 -
<4>2]
(75)
.
Tetragonal symmetry E~ 1 and E~ ~ are the same as for hexagonal symmetry (eqs. 72 and 73), and 1 y 2\2 rTset=½K~(Oz°)Z+~K (O2/ + ~1K 8 (Pxy) 2 + ½ K * [ ( P y z ) 2 + ( P z x ) 2] (76)
ESet ME2 is the same as for hexagonal symmetry [eq. (75)].
2.3.2. Harmonic description of magnetostriction The equilibrium values of the strains eU's are obtained by minimizing the total free energy with respect to the strains. This leads to relations analogous to those for cubic symmetry, but they involve appropriately symmetrized coefficients.
Hexagonal symmetry E~I
=
[(BalCo2
+
- B
l g FI ot l['~ae2
X [ColCo
2 --
a2
~12
0
l_/rlalt,'~2
oe2 ,~12
C O ) ( O 2 ) + 2\~.-al~0 -- D~Co ) ( J ) a2
:Co~12 )(22 - < L
-D
>2
2
- <4)2)]
(Co'2)21-1 ,
(77)
e ~2 =[(B~2Co I _ B~1C012)( O ° ) + !gD,~2C,~1 2, =1 o - D ~ul I C ou12) ( J ) 2 1 //-,i ~ 2 / - ~ 1 + 2~,"~2~o -X [ColCo
B~
ea = ~
B ~
(022),
e~ = C'-~o(Pyz>,
al 412 Da2Co )(2(L)2 - (L) 2 -
2 - (Co12)2] -1 ,
~
2B ~
e 2= Co" (P~y),
(jy)2)] (78)
(79)
B ~
e~2= C---~o(Vzx)"
(80)
28
P. M O R I N and D. SCHMITT
Tetragonal symmetry e a l and e oe2 are the same as for hexagonal symmetry [eqs. (77) and (78)], and BT
e ~=
-
C~ (O~)
(81)
-
B ~
8 = ~g ( P x y ) , Co
(82)
B *
B ~
e~ = -~o (PYz)'
e2 = ~
(Pzx).
(83)
In the same way as for cubic symmetry, including these equilibrium values in the starting Hamiltonian allows us to express this Hamiltonian in a form which depends only on magneto-elastic coefficients. For example, considering only one-ion magneto-elasticity leads to t h e following total quadrupolar Hamiltonian ff(Hex QT = _ Goe(O2°) O ° - G~[(O~) 0 2 + 4(Pxy)Pxy]
-- G¢[(Pyz)Pyz + (Pzx)Pzx], ~fTet a 0 ~ Q T : -- G < 0 2 ) 0 2
0
2 2 __ G T ( 0 2 ) 0 2
(84) a a
-
-
(Pxy)exy
-- Ge[(Pyz)Pyz + (Pzx)Pzx],
(85)
where the total quadrupolar coefficients G ~ have contributions from both the one-ion magneto-elasticity and the quadrupolar pair interaction:
Hexagonal a2 (B a l ) 2 C O - 2B'~IB~2Co 12 + (B~2)2C01
G~ =
G ~, _
C01C02
(B~)
2
C----~"- +
K ~,
_
~
(C012) 2
: GME +
K~
;
+ R °e = G M E
(/*
=
e, ~'),
"{- K a ,
(86)
(87)
Tetragonal G oe is the same as above (eq. 86), and G~-
(B'~) 2
+ K ~ = C M~E + K ~ ;
=
(88)
2.3.3. Perturbation theory An analysis of the magnetic and magneto-elastic properties of hexagonal and tetragonal compounds within the paramagnetic phases may be done by using perturbation theory. The full development of the susceptibility formalism is treated elsewhere (Morin et al. 1988) and only the main results are reported below.
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
29
Among the different modes, only the a- and e-modes in hexagonal symmetry (a-, y- and 6-modes in tetragonal symmetry) can easily be investigated by applying a magnetic field along the main symmetry directions, namely the [001], [100] and possibly the [110] directions. The quadrupolar components involved are ( O ° ) , ( 0 2) and possibly (Pxy)" Various associated susceptibilities may then be defined, namely strain susceptibilities X, and quadrupolar field-susceptibilities X~ - (2) (/z = a, e . . . . ). The only difference with respect to cubic symmetry is the explicit presence of a spontaneous quadrupolar moment (O°)0 in the expressions for X~, and Xtz -(2) • Also, the field dependence of the magnetization and quadrupole moments may be determined. For the quadrupolar components, relations analogous to those for cubic symmetry are found for hexagonal symmetry =
-
(89)
(9o)
2 ,
with ((2)
X~ = (1 - nx0)2(1 - G " X g ) '
(91)
and similar relations for tetragonal symmetry. For the magnetization, the following expressions are found M=
XMH--
4- . (3) L/3 X M 11
,
(92)
with X0 XM
--
,
(93)
1 - nXo
and X~ ) -
G a (x~(2),~2 ) . (2)-~2 q] [ (/l(e I |X~ 3) + 2 +2G ~ . (1 -- rtX0) 4 t_ 1 ~ G-~Xa 1 --- G- ~ 1
(94)
In these relations, all the susceptibilities assume different numerical values according to the direction of the applied magnetic field, although their expressions are formally equivalent. This is due to the different CEF wave functions determined by the field direction. This behaviour is well-known for the first-order magnetic susceptibility which is different for fields parallel and perpendicular to the c-axis in hexagonal and tetragonal compounds. For the third-order magnetic susceptibility, it turns out that only the G" contribution is present when the field is applied along the c-axis, while both G ~ and G ~ terms must be considered when the field is within the basal plane. Note that the y- and 6-modes in cases with
30
P. MORIN and D. SCHMITT
tetragonal symmetry may be separately investigated by applying a magnetic field along the [100] and [110] directions. Finally, only the isotropic bilinear exchange coefficient n ~1 has been considered here and is denoted n. 2.3.4. Comparison with cubic symmetry The main difference between hexagonal and tetragonal symmetries compared to cubic ones comes from the fact that the quadrupolar component (O~) is already ordered (i.e. ( 0 2°) # 0 ) by the crystal field. This leads to several consequences for the magnetic and magneto-elastic properties. First of all, from eqs. (77) and (78), there exists a spontaneous quadrupolar contribution to the volume (through e ~1) and to the ratio c/a (through e ~2) of the hexagonal and tetragonal unit cell. This feature manifests itself through an anomalous thermal expansion with regard to that of the corresponding compound with non-magnetic lanthanum or yttrium. This has been observed, e.g., in Pr (L/ithi and Ott 1980) and PrNi 5 (Barthem et al. 1988). Second, the a-term in YgOT[eqs. (84) and (85)] gives an additional quadrupolar contribution to the pure CEF second-order Hamiltonian as soon as either B ~1, B ~2 or K ~ is present. As for (O2°), this contribution is temperature dependent; this is a possible explanation for a thermal variation for the total effective second-order crystal field parameter B°2. As a consequence, the CEF level spacing will be temperature dependent, and it is worth taking this contribution explicitly into account besides the pure CEF one, at least in the case of strong quadrupolar coupling. The presence of a quadrupolar moment (020) does not exclude the possible existence of the corresponding quadrupolar pair coupling K ~. However, a quadrupolar phase transition involving this coupling is no longer allowed. This behaviour can be compared to the effect of high rank (fourth- and sixth-rank) pair interactions in cubic compounds which lead to apparent temperature-dependent fourth- and sixth-order CEF parameters (Morin and Williamson 1984). On the other hand, quadrupolar transitions involving the other couplings as K ~, K ~ (K r, K s, K" in tetragonal compounds) are quite possible and have been observed in several rare earth insulators (Gehring and Gehring 1975). 2.4. A peculiar case: the semi-classical treatment
In the case when bilinear interactions overwhelm the CEF coupling, the magnetic ordering temperature is large in comparison with the total CEF spacing. In the paramagnetic phase, the CEF levels may be considered as equally populated, while in the ordered phase, each of the CEF wave functions is "purified" by the exchange field to its dominant IJz) component, at least in the limit of low temperatures. Both situations may then be discussed within a semi-classical analysis. 2.4.1. The paramagnetic phase For vanishing CEF effect, the 4f multiplet is (2J + 1)-fold degenerate and the expressions of the various one-ion susceptibilities previously introduced become
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS C Xo = T
(95)
CQ T'
X~X(2)
(96)
2
2
CQ
(97)
= gjtx B 6 T 2 , C (3)
X(3) =
X(21 s
31
~--
r 3 ,
(98)
C~ ) T2 ,
(99)
where C is the usual Curie constant [see eq. (5)], and CQ, C ~ ) are quadrupolar constants and C (3) is the third-order Curie constant
CQ =
J(J + 1)(2J - 1)(2J + 3) 5k B ,
(lOO)
C(o2) = J(J + 1)(2J - 1)(2J + 3 ) ( 2 J - 3 ) ( 2 J + 5) 70k 2 C(3)
4 4 J(J+l)(2Jz+2J+1) = -gjtx B 90k 3
(101)
(102)
In the same way as C and C (3) are the coefficients of the first two terms in the H~ T expansion of the Brillouin function, CQ and C ~ ) are the coefficients of the first two terms in the s / T expansion of the response function of Q in the Brillouin-Boltzmann statistics 2
e C~ ) e Q ~ Ca -~ + ~5 +'".
(103)
Note that all the C's are isotropic. After renormalization by the bilinear and quadrupolar interactions, the actual susceptibilities . are
C XM -- T - O *
'
2 2 Co.T XQ = gllXB 6 ( T - 0 * ) 2 ( T - 0~) ' C(3)T
(i-U)
[1-
4
L
(lO4)
(2J-1)(ZJ+3)
?-2
75;-T
-T - -' O ~ J '
32
P. M O R I N and D. S C H M I T F
and the actual elastic constant is given by
C
T-O~
C0
T - 0~: "
(lo5)
This equation is very close to that given by Kataoka and Kanamori (1972). O~ = CQG and 0~ = CQK are defined by analogy with 0", as functions of the total and two-ion isotropic quadrupolar parameters G and K. The temperature dependences in eq. (104) have been extensively described by Morin and Schmitt (1981b) and Morin et al. (1980b) at high temperature as well as close to the magnetic ordering temperature. Less attention has been paid to the nature of the quadrupolar transition. It is easy to verify that for J = ½there is no quadrupole moment. For J = ~, C~2) is zero and the next term in the e expansion of Q is negative, the transition is then second-order. In all the other cases, C~ ) is positive and the transition is first order.
2.4.2. The ordered phase Several discussions of the energy in the ordered phase have been given in the past in particular by Clark in volume 1 of these handbooks (Clark 1980). Only those features that are necessary for describing properties encountered in the following sections will be derived here from our basic Hamiltonian. Following Callen and Callen (1963, 1965), the temperature dependences of the average value of the different operators O~' present in eq. (1) are expressed as functions of the normalized 4f magnetic moment, m, through modified hyperbolic Bessel functions. Their angular dependences are written as functions of the 4f magnetic moment cosines, ai, by using the harmonic polynomials. ( 0 O) = j ( j _
l ~
~)Is/2(~
-1
2
(m))(3a 3 - 1),
( 02> = J(J - 1)Is/2(~/~-1(m))(ot
2 -
ce2),
(Pij) = J ( J - ~) 1 ~1s / 2 f(~( -mi ) ) ( a i % ) (O04+504)=_40J(j -- ½ ) ( J _ l ) ( j _ 3
(i,j=1,2,3) 2 ~ 22 + ) I"9 / 2 ( ~ ' - 1 ( m ) ) ( O l l O
{O~ + 2 1 0 6) = 1 8 4 8 J ( J - ½ ) ( J - 1 ) ( J ^
-1
(106)
2
2
3)(j_2)(j_ 2
2
2
cycl. - ½)
~) 2
2
2
2
x I13,2(5g (m))[c~,c~2a 3 _ 1(12ct3 + c~3a a + a,t~2) +
~1
.
Du Tr6molet de Lacheisserie et al. (1978) have studied in detail the classical free energy of a cubic ferromagnet. The classical magnetostriction constants and anisotropy constants are found to be 0 --- 0 J ( J - ½)I5/2(5¢-1(rn)),
A100(T) = 2
Cll
-- C12
B ~
AHa(T ) - 3v~cO4 J ( J - ½)_15/2(~ ~(m)),
(107)
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
33
K I ( T ) = - 4 J ( J - ½) ( J - 1)(J - 23-)[10B419/2(Gt?-'(m))
+ 42(J - 2)(J -
G(T) =
5 )B6~13/2(~ '
-'(m))],
~)B6?~l~/Z(iLf-l(m)).
1848Y(Y- 1 ) ( j _ 1 ) ( J - k ) ( J - 2 ) ( J -
(108)
The K 1 anisotropy constant receives two contributions from 4th and 6th rank polynomials. Owing to their different temperature dependences (fig. 7) they may be the causes of a change in the easy magnetization direction in the ordered phase. It is also well-known that the magnetostriction contributes to the magnetocrystalline energy, e.g. AK 1
=
9
-- ~(Cll
0
0 2 _ C,2)Aioo(T )
-I-
0 2 79 C44~,,11(T)
(109)
.
This magnetostriction contribution may be an appreciable part of the intrinsic K 1 constant: it reaches ~ of K 1 in TbFe 2 at room temperature [see table 8 in Clark (1980)]. In addition, due to different temperature variations, K 1 may change of
I
I
I
I
I
I
I
I
Is/2 10"1 19/2
10-2
113/2 10"3
1.0
.8
.6
.4
.2
0
m
Fig. 7. Normalized hyperbolic Bessel functions J5~2, 19/2 and I13/2 versus the normalized magnetization (from Clark 1980).
34
P. MORIN and D. SCHMITI"
sign. This situation occurs, e.g. in hexagonal RC% compounds as revealed by the variation of the lattice parameters and elastic constants (se e section 4). This classical description of magnetostriction and magnetocrystalline anisotropy gives a phenomenological connection between their temperature variation and that of the 4f magnetic moment. The details of the magnetic couplings leading to this observed magnetization itself are not of too great importance. It is useful in complex situations, as e.g., for rare earths dissolved in a magnetic metal (Fe, Co, M n , . . . ) . Indeed, at least at low temperatures, the magneto-elastic coefficient, in particular for the symmetry associated with the spontaneous state, may be determined by using the assumption that the 3d magneto-elastic contributions are negligible. However, both the necessary separation of the 4f and 3d magnetic moment and the existence of many sublattices limit the insight in the 4f magnetism. Obviously it would be unrealistic to search for quadrupolar pair interactions, when they are overwhelmed by bilinear ones. 3. Experimental evidence
3.1. Quadrupolar orderings To observe quadrupolar orderings it is necessary to find rare earth systems where the bilinear interactions are weak compared to the quadrupolar interactions. Favourable situations exist at each end of the lanthanides series. Indeed, first, the bilinear interactions are at their minimum for the series of compounds under consideration and, second, both the orbital moment L (Tm a+ : J = 6, L = 5; Ce 3+ : J = ~, L = 3; pr3+: J = 4 , L = 5 ) and the Stevens coefficient as, which deeply control the quadrupolar coefficients (Morin and Schmitt 1981a, Schmitt and Levy 1985), are large. In other cases, magnetic and structural transitions coincide (section 3.2.2). Quadrupolar orderings have been reported to occur in TmZn (Morin et al. 1978b), TmCd (Al6onard and Morin 1979), CeAg (Ushizaka et al. 1984, Morin 1988), PrPb a (Bucher et al. 1972a), TmGa 3 (Czopnik et al. 1985), CeB 6 (Effantin et al. 1985) and PrCu2 (Ott et al. 1977a). The first three of them are clearly ferroquadrupolar (FQ), the next three antiferroquadrupolar (AFQ) and the actual situation in the last one is not yet fully elucidated. UPd a which exhibits a structural transition may be relevant to the present review owing to its localized 5f electrons (3H 4 multiplet) (Andres et al. 1978). Structural transitions have been also observed in the RB 4 series, however, their origin has not been fully understood (Will et al. 1986).
3.1.1. Ferroquadrupolar ordering As for the ferromagnetic ordering (FM), the FQ ordering is characterized by a q = 0 propagation vector and is expected to be closely described by the singlesublattice formalism (see section 2.2).
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H INTERMETALLICS
35
3.1.1.1. TmZn. Cubic compounds within the CsCl-type structure are characterized by large quadrupolar interactions, TmZn being the archetype compound (Giraud et al. 1985). The temperature variation of the specific heat (fig. 8) reveals two well-defined transitions, close to each other (Morin et al. 1978b). TmZn undergoes a FQ ordering at T e = 8.55K, where ( O °) is the order-parameter. The transition is clearly first-order. An associated tetragonal strain takes place simultaneously due to the B y magneto-elastic coefficient. A FM ordering occurs in the quadrupolar phase at T c = 8.12 K, with the spins lying along the four-fold axis. The spontaneous tetragonal strain is reinforced by the bilinear interactions and reaches c/a - 1 = -8%0 at 1.5 K. From the jumps in the electrical resistivity at the two transitions, the electronic scatterings of quadrupolar and magnetic origins are of similar magnitude. Numerous properties investigated in the paramagnetic phase of TmZn have been consistently analyzed with a total quadrupolar coefficient G v= 25 mK; the dominant contribution to G v comes from the pair interactions (Givord et al. 1983). This G ~ value describes the temperature T e and no additional coupling (anharmonicity, higher order terms . . . . ) is needed. The FQ ordering lifts the degeneracy of the F~ 1) cubic ground state and splits it into a doublet and a singlet; the non-magnetic singlet is the ground state in the quadrupolar phase (structure a o in fig. 4 of section 2.2.3). Bilinear interactions are strong enough to induce a magnetic moment in this phase. Substituting L u 3+ for Tm 3+ in Tml_xLuxZn or Cu for Zn in TmCuxZnl_ x decreases the bilinear interactions quicker than the quadrupolar ones and the spin system remains paramagnetic: only the FQ ordering is observed (fig. 9) (Morin and Schmitt 1986). Note that in this case the splitting of the F~ 1) cubic triplet in the tetragonal paramagnetic phase has been observed by inelastic neutron scattering (Morin et al. 1981). 60
!
1
i
50 Tm Zn
o
E 40
30
20
l
T
r,, sss
T 8.12K= .
K , , ~ .
I
1
I
7
8
9
TEMPERATURE
( K )
Fig. 8. Specific heat of TmZn; T c is the magnetic ordering temperature, T O the quadrupolar ordering temperature (from Morin et al. 1978b).
36
P. MORIN and D. SCHMITT
10
% TmCu
v
Zn 1-c c
p
LU mr
2 ¢rUJ O.
;* 2 5 ~
:E
--.
hi
-_
.
,
/
f--
-5
20
/
oG 1 //
0.5
,,
15
)
C
Fig. 9. Phase diagram for TmCUl_cZnc; P: paramagnetic, M: modulated, AF: antiferromagnetic, F: ferromagnetic, Q: quadrupolar phases; lines are guides for the eye. Variation of the quadrupolar parameter G r (=G~) and bilinear exchange parameter 0* with c (from Morin and Schmitt 1986).
3.1.1.2. TmCd. Historically, the first FQ ordering in a rare earth intermetallic was observed in this compound at T o = 3.16 K. It was initially analyzed with a F3 cubic ground state, and the first-order character of the transition was assumed to be driven by anharmonic elastic effects (L/ithi et al. 1973b). However, RZn and RCd isomorphous compounds exhibit very similar properties, in particular the CEF (Alronard and Morin 1985). As a matter of fact, there is a close parallel between TmCd and TmZn and all the TmCd properties are fully explained with a G ~= 12 rnK total quadrupolar coefficient without the need of any other coupling (Alronard and Morin 1979). As in TmZn, pair interactions are the dominant contribution to the quadrupolar coupling and the ground state is also the non-magnetic singlet coming from the F~ 1) cubic triplet. Here, however, the spin system remains paramagnetic owing to the weakness of the bilinear interactions. As in TmZn, electrical resistivity measurements give clear evidence for the vanishing of the quadrupolar scattering below T o (fig. 10). 3.1.1.3. CeAg. CeAg is the only example of a FQ ordering for a Kramers ion. It
also belongs to the family of CsCl-type compounds. The cubic ground state is the F8 quartet, well-isolated from the F7 excited level (A = 266 K) (Schmitt et al. 1978). It exhibits a cubic-tetragonal transition at T o = 15.85 K with (O2°) as the order parameter (fig. 11). A FM transition occurs at T c = 5.2 K, the c axis of the tetragonal cell is the easy magnetization direction (Morin 1988). The c / a - 1 strain is very large (1.7% at 1.5 K). A strong dependence of the low-temperature properties on the metallurgical conditions is observed, which is related to the very large magneto-elastic coupling.
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H I N T E R M E T A L L I C S
37
5.6 I
,
,p..."
E 5.4
de_' dT
1
°
I.-te
E
W °
5.0
•"°
0
°
°¢,°°
""
I
i
2
½
t
~
n
I
4 i
4
3
TEM PE RATURE(K) Fig. 10. Low-temperature variation of the resistivity and its derivative (inset) in TmCd (from A16onard and Morin 1985).
u
30 CeAg
_~2o E
0
5
10
1~i
20
T (K) Fig. 11. Specific heat of CeAg. Anomalies at T c = 5.5 K and TQ = 15.5 K correspond to the ferromagnetic and the quadrupolar orderings, respectively (from Ushizaka et al. 1984).
The study of the quadrupolar properties in the cubic phase leads to G r = 150 inK. However, this value would drive a FQ ordering at only 10 K instead of T o = 15.85 K (Morin 1988). This difference is consistent with the weak softening of the Cll - C12 elastic mode (Takke et al. 1981) and reveals the existence of an additional term, i.e. the anharmonic elastic coupling of the Cll - C12 mode with a zone-boundary phonon, which softens at the M point of the Brillouin zone (Knorr et al. 1980). This last mechanism is associated with the incipient instability of the CsCl-type lattice for light rare earth-silver compounds which is of electronic origin (see section 3.10).
38
E MORIN and D. SCHMITI"
3.1.2. Antiferroquadrupolar ordering Several types of AFQ structures may exist. First, the lifting of the degeneracy of the cubic ground state may lead to wave functions different from one site to the other (Sivardi~re and Blume 1972). If the absolute values of the quadrupolar moments are different for the sublattices, that corresponds to a ferriquadrupolar ordering. Second, the ground state can be the same on each site, but the local z-axis may change, although belonging to a given star of crystallographic directions: the corresponding quadrupolar arrangement may be called multi-axial (Sivardi6re 1975). Lastly, more than one propagation vector may be necessary to describe the AFQ structure. In all the cases, magnetic structures with many sublattices may be favoured (see section 3.6). Whatever the actual structure may be, a multi-sublattice formalism is needed. The previous single-ion treatment may be used only for describing the occurrence of a magnetic or quadrupolar moment within a given sublattice. To describe the AFQ as well as AFM properties, no reliable information can be expected from static measurements in these intermetallics. In particular, the macroscopic symmetry may remain cubic. Only microscopic probes, such as neutron diffraction, are effective in revealing the existence of more than one magnetic sublattice.
3.1.2.1.
C e B 6. This cubic compound exhibits very complex properties due to the interplay of localized and delocalized 4f behaviours. The temperature variation of the specific heat reveals two anomalies at T o = 3.3 K and TN = 2.4 K (Peysson et al. 1986). Neutron diffraction experiments have shown that the compound remains paramagnetic down to 2.4 K (Horn et al. 1981). Magneto-elastic effects are present above TQ, although the C44 and Cll - C1~ elastic modes softens by only 2% (L/ithi et al. 1984). As the F8 ground state is well isolated (A = 540 K) (Zirngiebl et al. 1984), a quasi-full softening is expected in the case of a FQ ordering. On the other hand, the temperature variation of C44 and Cll - C12 may be analyzed with negative quadrupolar pair interactions. An AFQ ordering is then probable below TQ. This is confirmed by a neutron diffraction experiment on a monocrystalline sample in a magnetic field (Effantin et al. 1985). Indeed, two kinds of Ce ions have been observed which differ in zero field only by their magnetic susceptibility. The corresponding superstructure lines are described by a [½ ½ ½] propagation vector. The AFQ ordering then consists of an alternating sequence of (111) planes with different quadrupolar moments (fig. 12). Below TN, the spontaneous biaxial spin structure reflects the existence of the two sublattices driven by AFQ ordering. The magnetic field phase diagram reveals the interplay between localized and delocalized 4f magnetism as demonstrated by the anomalous increase of the AFQ phase under a magnetic field (fig. 13). In zero field, T O and the AFQ phase are reduced by the Kondo coupling but with an increasing magnetic field, the Kondo state is progressively suppressed and the AFQ ordering can manifest itself at a higher temperature.
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H INTERMETALLICS
Magnetic ph<,.,,,,,
modulated
Anti ferro
~:[~/+ ,/, ,/2]
,
pho,.,,
:
39
quadrupolar
"~:[~]
z
(b)
(a)
Fig. 12. (a) Biaxial magnetic structure in CeB6 described by the propagation vectors indicated. (b) Antiferroquadrupolar structure described by the k o propagation vector; open and full circles correspond to the two different Ce-quadrupole sublattices (from Effantin et al. 1985).
' A- ,~ ' 'o ~,;too~j "-= ~6 *-I~//[111:]
20
t,....~ It ",
140
/
"iT
160
J/
120 100 80
0
+oIl,.,,,..=,,,,,,.+o,. q o!
60
pl ,+,,°,.o., V ,oI+,+..,.+.°,.o/,.
II
+~"'<""'i
0 1 " " @ 0 2
4
.
17 I~L_.
i
TernperaLure(K)
I 8
l
,o
=
0 10
Fig. 13. Magnetic phase diagram of CeB 6. Phases III and II are drawn in fig. 12 (from Effantin et al. 1985).
T m G a 3. This AuCu3-type cubic compound also exhibits two transitions in the temperature dependence of the specific heat, but very close to each other (TQ = 4.29 and Tr~ = 4.26 K) (fig. 14) (Czopnik et al. 1985). Here again, studies in the paramagnetic phase lead us to introduce quadrupolar interactions, but they 3.1.2.2.
40
P. MORIN and D. S C H M I T r
rN 60
50
I 40 30 20 10 o 4.1
4.2
4.3 T(K)
4.4 -
Fig. 14. Specific heat anomalies in TmGa 3. Arrows indicate the antiferromagnetic (TN) and antiferroquadrupolar (TQ) transitions (from Czopnik et al. 1985).
are too weak to induce a FQ ordering at T e (Morin et al. 1987c). In addition the spontaneous spin structure determined by neutron diffraction on a single crystal is multi-axial (see fig. 53 in section 3.6). This indicates there are different (111) directions as local z-axes at the 4f sites stabilized by AFQ interactions of trigonal symmetry (Morin et al. 1987d). Thus the quadrupolar ordering is of AFQ-type. As for the ferromagnetism in TmZn, the conditions of induced antiferromagnetism on the singlet ground state cause the temperature range of the paramagnetic AFQ phase to be very narrow. P r P b 3 . Another AuCu3-type compound, PrPb3, exhibits a non-magnetic transition at TQ = 0.35 K corresponding to the lifting of the F3 ground state degeneracy (fig. 15) (Bucher et al. 1972a). No spontaneous macroscopic strain has been detected in the quadrupolar phase. This may be due either to its weakness in the case of a FQ ordering or to its nullity in the case of an AFQ ordering. The softening of the C44 and Cll - C12 ultrasonic is far from complete and influenced by high-order magneto-elastic terms (Niksch et al. 1982). They may be analyzed by positive as well as negative quadrupolar interactions. However, both the parastriction and the third-order magnetic susceptibility need negative quadrupolar coefficients to be described (Morin et al. 1982). An AFQ ordering is then highly probable, and ( O °) is the order parameter. In conclusion, it appears that the case of AFQ ordering is not as well understood as the FQ ordering. This is due to experimental difficulties as well as theoretical complications. Indeed, none of the previously mentioned three compounds is a completely favourable candidate for investigating AFQ properties. In CeB 6 the quantitative analysis is complicated by the Kondo behaviour of the 3.1.2.3.
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
41
0.55"K 0.7
0.6
PrPb 3
0.5
~
0.4
0.~
0.2
0.1
0
I
I
0.5
1.0
T(*K)
Fig. 15. Molar specific heat CJ R of PrPb 3 between 0.04 and 1.0 K (from Bucher et al. 1972a).
cerium ions. In TmGa 3 the temperature range of the AFQ phase is much too narrow. PrPb 3 orders at 0.35 K, a temperature characteristic of very weak AFQ interactions, which may easily be broken by the application of a magnetic field. From a theoretical point of view, the existence of many inter- and intralattice pair coefficients complicates the analysis of the data.
3.1.3. Structural transitions possibly of quadrupolar origin A few other structural transitions have been reported in the literature, which may be evidence for quadrupolar interactions. However, the actual situation is not yet fully determined, often because the crystal electric field is not known.
3.1.3.1. PrCu 2. This orthorhombic compound exhibits a second order structural transition at 7.5 K which has been observed in many different experiments, in particular by specific heat (Wun and Phillips 1974). From elastic constant measurements, the primary order parameter was found to be the ez, strain (Ott et al. 1977a). This agrees with neutron diffraction data which have shown that below 7.5 K there is a sizeable splitting of the (101) nuclear reflections corresponding to an angle/3 between the a- and c-axes of the orthorhombic cell that is no longer equal to 90° (Kjems et al. 1978) (fig. 16). A FQ ordering involving the (Pz~) quadrupolar component is likely. Unfortunately, there does not exist a determination of the CEF; thus a quantitative analysis of the observed properties is not possible.
42
P. MORIN and D. SCHMITI"
PrCu z
T=B.OK
T=4.75 g 1.0
0.5 ,
,
-0.01
0 i
~
~-I
0.01-0.01
0
0.01
k" ="
~ ~OR'¢
0.0
i-
4.0
L
~ I.v
=
A
6.0 8.0 TEMPERATURE (K)
10.0
Fig. 16. The upper part shows high resolution scans through the (101) Bragg peak below and above the structural transition temperature in PrCu2 at T = 7.5 K. In the lower part the observed splitting (A) is plotted versus the temperature. The full line represents a mean field calculation (from Kjems et al. 1978). 3.1.3.2. U P d 3. Contrarily to most of the metallic actinides, the hexagonal compound U P d 3 (P63/mmc) exhibits properties characteristic of localized magnetism associated with a 5f 2 configuration (3H 4 ground-state multiplet) (Andres et al. 1978). In a way reminiscent of Pr metal, two U sites exist with hexagonal and nearly cubic symmetries. T w o C E F transitions have been observed by inelastic neutron scattering on a single crystal. They have been attributed to an excitation process from a singlet ground state for both sites (Murray and Buyers 1980). In addition, U P d 3 exhibits a well-defined transition at 7.5 K, strongly depending on the metallurgical status of the sample (fig. 17). N o change of the hexagonal
4
t2
3 B~
-
ou -~-
,
0
2
4
6
4
8
10
12
14
o~_I
0
16
T(K)
Fig. 17. Specific heat of a UPd 3 monocrystalline sample minus that of ThPd 3 used as background
(open circles). The solid line (right-hand scale) is a plot of dp/dT. The slight difference in peak position of the two curves is caused by inaccuracies in the temperature scale of the resistivity measurement (from Andres et al. 1978).
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
43
symmetry was observed from X-ray diffraction experiments on polycrystalline samples (Andres et al. 1978, Zaplinski et al. 1980). However, the increase of the c/a ratio observed at low temperatures might be associated with U atoms becoming more prolate ( ( O ° 2 ) < 0 ) at the quasi-cubic sites, according to an internal ferroquadrupolar process. This assumption might be checked by a full determination of the crystallographic structure on both sides of the transition by using a single crystal.
3.2. Nature of quadrupolar and magnetic transitions 3.2.1. Quadrupolar transitions As discussed in section 2.2.4, in cubic symmetry the transition at the quadrupolar ordering is usually first-order due to the existence of van Vleck matrix elements from the ground state. Indeed for the 6 cases (TmZn, TmCd, CeAg, CeB6, TmGa3, PrPb3) presented in section 3.1, the so called Xs(2~ susceptibility (Morin and Schmitt 1983a), which determines the order of the quadrupolar transition is calculated to be non zero. Thus the observed first-order quadrupolar transitions in the Tm compounds are well predicted by the present formalism. However a quantitative description of the specific heat anomalies is difficult in TmGa 3 and TmZn due to the vicinity of the magnetic and quadrupolar transitions. In TmCd, the measurements are not accurate enough, in particular in the low temperature phase. In PrPb3, the quadrupolar ordering temperature, T~ = 0 . 3 5 K , is small in comparison with the F3 - F4 spacing. The corresponding van Vleck terms do not affect the ordering process too drastically and the ( O °) jump is calculated to be weak in agreement with the temperature dependence of the specific heat (fig. 15). The same feature is also observed in CeB 6 and CeAg where the F8 - F7 splitting reaches 540 and 260 K respectively. For instance in CeAg the jump of (O2°) is calculated to reach only about 10% of its maximum value. The existence of anharmonic elastic terms in this compound makes it impossible to obtain a quantitative analysis of the T~ transition within the present formalism.
3.2.2. Magnetic ordering in the presence of quadrupolar interactions The nature of the magnetic transition in ferro- and antiferromagnetic cubic compounds has motivated a great number of studies, which can be classified according to whether they are based on an MFA analysis or renormalization group (RG) theory (K6tzler 1984). In both cases, the (isotropic) Heisenberg exchange is the starting point. The two approaches differ in the additional couplings which also have to be considered. The MFA model presented above, is in the line of models previously developed for transition metals as MnO, MnS (Lines and Jones 1965). However, instead of exchange striction, it considers for rare earth compounds quadrupolar pair and magneto-elastic interactions. The resulting coefficients of the free energy expansion are expressed in terms of single-ion susceptibilities and determining the phenomenological parameters for the different couplings then leads to the description of the magnetic transition (see section 2.2.4).
44
P. MORIN and D. SCHMITI"
On the other hand, the RG approach is based on symmetry considerations and includes critical fluctuations to predict the order of the magnetic transition. Within this hypothesis, continuous magnetic transitions become discontinuous ones according to the sign of the M4-term in the Landau expansion of the free energy. Note that the RG theory usually considers an incompressible lattice and nonquantized magnetic moments. The hypothesis of an incompressible lattice is in strong opposition with the evidences of large symmetry lowerings in cubic rare earth compounds. Due to the assumption of classical moments, the RG theory cannot predict the occurrence of a first-order transition associated with the positive X:, - (3) value in the case of a F3 non-magnetic ground state (see section 2.2.4 and Blume 1966). The aim of this section is to show through examples from the literature that the nature as well as the temperature of the magnetic transition may be predicted as soon as the quadrupolar magneto-elastic and pair interactions are reliably known (sections 3.2.2.1 and 3.2.2.2). Of course more complex situations may occur due to additional contributions from spin (and quadrupolar) fluctuations (section 3.2.3).
3.2.2.1. First-order transitions and tricriticality TmCu and TmAg A good example of the effect of the quadrupolar interactions on the magnetic transition is given by TmCu. This compound orders at TN = 7 . 7 K in a modulated antiferromagnetic structure, with all the moments parallel to the [001] axis. The transition is first-order (fig. 18). Various experiments performed in the paramagnetic state reliably provide the tetragonal quadrupolar parameter G r = 11 mK
1oo TmCu
-~ so
5 i
+I
i
10
,~-'"'"
.......
i
" "
.-
S
5
i
(b)
f" .,.." ,,, ~ ." " I 6
I
I
+
7
8
g
TEMPERATURE
(K)
Fig. 18. Temperature variation of the specific heat Cp (a) and of the entropy S (b) in TmCu. A n incommensurate antiferromagnetic structure is established at T N = 7.7 K and transforms at 6.7 K into a commensurate one (from Morin and Schmitt 1980).
Q U A D R U P O L A R EFFECTS IN R A R E EARTH INTERMETALLICS
45
(Jaussaud et al. 1980). Just above TN the CEF third-order paramagnetic susceptibility Xz, - (3) is negative; this would lead to a second-order magnetic transition. However, the quadrupolar contribution to the total third-order magnetic susceptibility is positive and large enough to change the sign of X~)0, in the vicinity of TN, as seen experimentally. Consequently, the magnetic dipolar transition is expected to have a first-order character, as effectively observed (fig. 19a).
2
'
I'0.5
'
I
'
i -.
i
I
_m°!
,/ 3.5,,"
-~
T
ii..................
W =1~2K x -=-.42
'
1~o
'
f J" j/J
W=i.05K x=-.4.7
:.:" I
0
"
/ (~
.-"'' ( /"" T~
0/
0
I
I
10
TEMPERATURE(K) Fig. 19. (Top): temperature dependence of X~ ) observed for the tetragonal symmetry in TmCu (a) and TmAg (b). Full and dotted lines describe, for the CEF parameters W and x indicated, the cases with and without quadrupolar interactions, as indicated by the numerical values of the G v coefficient. (Bottom): the deduced magnetic phase diagram for both compounds, n is the bilinear coefficient (n N for an antiferromagnet). Dotted lines: second-order magnetic transition without quadrupolar interactions; dashed lines and full lines denote respectively: second- and first-order magnetic transition in the presence of quadrupolar transitions. C is a tricritical point (from Giraud and Morin 1986).
The isomorphous compound TmAg exhibits very similar CEF and bilinear interactions. It orders antiferromagnetically at T N =- 9.5 K, the transition is second order (Morin and Schmitt 1982a). Quadrupolar interactions within the tetragonal symmetry are weaker than in TmCu. They are characterized by G r = 4 m K (Giraud and Morin 1986). X~ ) is observed to be negative above 10 K, but is calculated to change sign at 6.2 K (fig. 19b). This defines a tricritical point between second-order transition (T > 6.2 K) and first-order ones (T < 6.2 K): the magnetic phase diagram predicts a second-order transition for a temperature of 9.5 K in agreement with the observation. An interesting feature is the large dependence of the tricritical temperature on quadrupolar interactions. Indeed, if G ~ reaches 5 mK it would shift the tricritical point up to 12.5 K, and the magnetic transition would be first-order for the same bilinear coefficient. Owing to the quadrupolar interactions, TmAg appears to be very close to tricriticality.
46
P. MOR1N and D. S C H M I T r
CeZn and CeMg Among other CsCl-type compounds, the CeX present a large set of different properties. Indeed whereas CeAg undergoes two well-separated magnetic and quadrupolar transitions (section 3.1.1.3), CeZn and CeMg order antiferromagnetically at T N = 30 and 19.5 K, respectively, the transition being first-order (Pierre et al. 1981, 1984). Contrary to the case of CeAg, the lattice stability is large; as in CeAg the Kondo-type behaviour for the Ce ion is not important. The tetragonal symmetry lowering below T N reaches record values ( c / a - 1 = 1.7 × 10 -2 and 1.3 × 10 -2 in CeZn and CeMg, respectively). Fitting magnetic properties in the ordered phase leads to G r = 220 and 50 mK in CeZn and CeMg, respectively. Note that these values have not been confirmed by measurements in the cubic paramagnetic phase. In CeZn the corresponding third-order magnetic susceptibility has been calculated to be positive in the vicinity of the N6el temperature, in agreement with the first-order character observed (fig. 20). In CeMg, the value G r = 50mK is a little bit too small: a value of about 75 mK would be large enough and more relevant.
/~-~=264 K
i
I'-~- ~=190KI
~
G.~=5OrnKI
CeZn
'75mK
1, ~ OOmK
\
0
I
/
20 TEM
I
PERATURE(K)
I
40
Fig. 20. Temperature dependence of the third-order magnetic susceptibility /~M,O - (3~ for CeZn and CeMg in the presence of quadrupolar interactions characterized by the G ~ values indicated.
erB 6 The same features are found again in P r B 6 , a compound isomorphous t o C e B 6. Here bilinear interactions are larger than the quadrupolar ones and a first-order antiferromagnetic transition is observed at 6.9 K (McCarthy et al. 1980). This first-order character seems to be stabilized by the same type of AFQ interactions as in the quadrupolar phase of C e B 6 ; indeed at low temperature the magnetic structure is described by the same propagation vectors (Burlet et al. 1988) and at high temperature large softenings of elastic constants are observed (Tamaki et al. 1985). By using quadrupolar coefficients deduced from f e B 6 (L/ithi et al. 1984) and NdB 6 (Tamaki et al. 1985) calculations within the present formalism show that the third-order magnetic susceptibility associated with the trigonal symmetry is positive. Measurements would allow to confirm this analysis.
QUADRUPOLAR
E F F E C T S IN R A R E E A R T H I N T E R M E T A L L I C S
47
Tb pnictides Equi-atomic compounds within the NaCl-type structure have been extensively studied (Hulliger 1978). They are the earliest intermetallics observed to exhibit large one-ion magneto-elastic couplings (L6vy 1969, Mullen et al. 1974). In addition, from numerous studies, they have been shown to exhibit a more complex behaviour, they are at the meeting point of different couplings (section 3.2.4). However, a careful analysis of the antiferromagnetic TbX has shown that quadrupolar interactions unambiguously play a dominant role (K6tzler 1984). Indeed the AF transition is first order in TbP (T N = 7.1 K) and TbAs (T N = 12.5K) and second-order in TbSb (TN= 15.5K), TbBi (TN=17.5K), TbS (TN = 45 K), TbSe (TN = 49 K) and TbTe (T N = 51 K). All the compounds exhibit trigonal spontaneous distortions (Hulliger and Stucki 1978b). Figure 21 shows the resulting jumps of the magnetization and the susceptibility at the first-order transition in TbP compared to the behaviours observed in TbBi. In the t
__°_,--°_
i
TbP
MS(r) Ms(O)
i
i
°~Ooo~_-~ "~
(TN=7.1K )
.8 ! f
1.0
°~,
,
TbBi
\
IT, =17.5K)
~,
\
.6 .4
0.0
\
.2
o
0.4
Xo(T) Xo(T~)
0.6
0.8 T/TN
1.0
1.0
"'~ TbBi (TN :l?.SK)
~.0
0.5
.6
°
.8
a
1.0
3.5
..
0 (TN =7.1 K)
b o,
I
z
~
I
z,
~
I
6
~
T/TN
I
8
Fig. 21. Comparison between: (a) the spontaneous sublattice moments, and (b) the magnetic susceptibility for TbP and TbBi. Full lines are calculated within a formalism similar to that of section 2
(from Koetzler 1984).
48
P. MORIN and D. SCHMIT]"
paramagnetic phase, there are large magneto-elastic contributions to the elastic constants in TbP (Bucher et al. 1976) and TbSb (K6tzler 1984). The quadrupolar interactions are characterized by G" = 61, 32, 22 and 22 mK in TbP, TbAs, TbSb and TbBi, respectively (K6tzler 1984). These G ~ values lead to the temperature variations of X~u,o -(3) shown in fig. 22. Obviously the antiferromagnetic ordering occurs in TbP and TbAs at temperatures TN, for which xM,0 - (3) is positive; this explains the first-order transitions. From an experimental point of view, positive values of X}~) have been observed in TbP for the trigonal symmetry (fig. 23). For TbSb and TbBi, a second-order transition is predicted, as is experimentally
-/4
i
~32mK
I m
,
,~N
22m½
-10-4O
=
10 TEMPERATURE(K)
20
Fig. 22. Temperature dependence of the third-order magnetic susceptibility XM,O - (3) calculated in the presence of quadrupolar interactions in TbP (G ~= 61 mK), TbAs (G " = 32 mK) and TbSb, TbBi (G ~= 22 mK). W and x define the level scheme common to the four compounds. The first-order (second-order) transition at T N in TbP and TbAs (TbSb and TbBi) are driven by quadrupolar interactions.
0.00 B
,
,
X2)
~X =6inK
,
,
TbP ~ l~[I.]
o.oo/-, i i ~ 0.002 0.0
8
10
12
lt,,
T[K]
16
Fig. 23. Temperature dependence of the third-order magnetic susceptibility in TbP. Full lines are calculated with the indicated values for the trigonal quadrupolar coefficient Y( = 12G" (from Raffius and Koetzler 1983).
Q U A D R U P O L A R EFFECTS IN R A R E EARTH INTERMETALLICS
49
observed. Note, however, that the critical point C strongly depends on G ~ and that TbSb and TbBi are not far from tricriticality of a quadrupolar origin.
3.2.2.2. Second-order transition in PrMge due to negative quadrupolar interactions. This compound orders ferromagnetically at T c = 10 K through a secondorder transition (Loidl et al. 1981). The ground state is a F 3 doublet. As mentioned in section 2.2.4, this CEF configuration gives rise to a positive value for Xz, - (3) in the vicinity of T c. This should produce by itself a first-order magnetic transition. On the other hand, the G ~ value suggested by Loidl et al. (1981) for explaining the temperature dependence of the spontaneous magnetic moment, i.e., G ~= -13.3 mK, leads to a negative total third-order magnetic susceptibility )¢(3) M,0" This now, accounts for the second-order character of the transition (fig. 24). I
.J.lo..
3
I
I
i
l
I
%%0 "
#2 PrMg21 0
5
T[K]
t~t 10
Fig. 24. Spontaneous magnetic moment, versus temperature in PrMg z. The dotted line is the Brillouin curve for J = 4. The full and broken lines give the result of a mean-field calculation with and without the quadrupolar interactions G ~= -13.3 mK (from Loidl et al. 1981).
Note that in the ferromagnetic phase, the ( O ° ) quadrupolar components are ferroquadrupolarly aligned by the dominant ferromagnetic interactions, whereas the negative quadrupolar interactions would prefer an AFQ arrangement and are thus frustrated.
3.2.3. Magnetic ordering in the quadrupolar phase The last case corresponds to a dipolar phase transition inside the quadrupolar phase (T < To). Below T c the quadrupoles are ordered and the crystal field is no longer cubic, e.g. it may have tetragonal symmetry. Therefore the various susceptibilities (X0, x~ " (3) , • • .) have new values (X~, x"~ T(3) . . . . ) according to the new tetragonal CEF level scheme. Note that this level scheme is now temperature dependent, like the ordered quadrupolar moment. The discussion of the nature of the dipolar transition is the same as in section 2.2.4, where /1-C(3) M,0 is replaced by )(T(3) 7 . Depending on the sign of X~ (3), the dipolar phase transition has a first- or second-order character.
50
P. MORIN and D. SCHMIqT
3.2.3.1. CeAg, f e B 6. In the case of CeAg, the ferromagnetic transition is second-order in the tetragonal phase (fig. 11) in accordance with theory. Indeed, . T(3) the x~ susceptibility is calculated to be negative as expected for a Kramers ion through the divergence of negative Curie-type terms associated with the magnetic doublet ground state. The same conclusion is also valid for CeB 6. 3.2.3.2. TmZn. In TmCd, bilinear interactions in the tetragonal phase are insufficient to induce a magnetic moment on the singlet ground state. On the contrary, in TmZn bilinear interactions generate a ferromagnetic ordering (T c = 8.12 K) slightly below the quadrupolar ordering (T o = 8.55 K). Calculations lead to a critical situation for Tc: the nature of the magnetic transition seems to be very sensitive to the values of the various parameters. For instance, T c is calculated to be of first order for G ~ = 25 inK, but would be of second-order with G v = 28 mK, as it experimentally appears. Consequently, the situation of TmZn with regard to the ferromagnetic transition seems to be not far from a tricritical point. 3.2.4. Presence of additional couplings In the compounds presented in section 3.2.2.1, the quadrupolar and Heisenberg bilinear interactions are dominant; any other interaction remains undetected; nonetheless, the competition between various couplings may be more balanced in other series such as rare earth pnictides and dialuminides. Two-ion magnetoelasticity and/or fluctuations may play a larger role in the minimization of the free energy. 3.2.4.1. DySb and rare earth monopnictides. This compound appeared in the past as an archetype for a first-order magnetic transition driven by quadrupolar interactions (Bucher et al. 1972b). It antiferromagnetically orders at TN = 9.5 K with a four-fold axis of easy magnetization. Associated with the magnetic structure, the lattice undergoes a structural symmetry lowering, mainly tetragonal, but with a small monoclinic component (L6vy 1969, Felcher et al. 1973) (fig. 25). This implies a magnetic moment not strictly parallel to the [001] axis. The existence of quadrupolar interactions has been unambiguously proved by the strong softening of C l l - C12 (Moran et al. 1973, Levy 1973); the value of G ~= 1 mK deduced is large enough to explain the first-order character of the joint spins and quadrupoles orderings, the third-order magnetic susceptibility is calculated to be positive (fig. 26). However, the analysis of the magnetization processes (Kouvel and Brun 1980, Everett and Streit 1979) and of the third-order magnetic susceptibility (A16onard et al. 1984b) in terms of only quadrupolar interactions fails (fig. 26). Indeed data are positive only in a very short range above TN and cannot be described by a unique G ~ value. Thus, additional spin interactions are present especially for the tetragonal symmetry. The most commonly proposed coupling is based on anisotropic bilinear interactions (Trammell 1963, Kim and Levy 1982, Jensen et al. 1980). It was in
Q U A D R U P O L A R EFFECTS IN RARE EARTH INTERMETALLICS
.~;
,.y-e-V--L~c~-
Oa
h
....+, + , .
-I
l
~
-2-c
(t~tragonal)
-3-E
51
-4-__~n
n
,',
(cubique)
l° c/a
Oy P,
0.997 0.996
= Dy S b
0.995
-5--
Temperature
T° K
Fig. 25. Temperature dependence of the relative change of the tetragonal cell parameters in the ordered phase of DyP (two samples), DyAs and DySb; a 0 is the cubic cell parameter (from L6vy 1969).
a =
i
i
"
\
i
W=.2BK
i
~
x =-.82
GI~0"lmK 0
\.,~e3 1 ,,p o
=b
I
t
~
'
DI(kO~PB/ 3 0 ~
10
I
~
'
[0 0 1]
20 IEMPERATURE(K)
' W=.28K x =-.82 1~=-6.5K GI=1mK
30
Fig. 26. Temperature variation of the third-order magnetic susceptibility for H[I[001 ] in DySb. (a) tentative fits are drawn with the quadrupolar model alone, G 1 = G r = I mK being the value obtained from the softening of the Cu - C12 normal mode. (b) Fits here include, in addition to G: = i mK, the two-ion magneto-elastic coupling characterized by D1 = D r (from A16onard et al. 1984b).
52
P. MORIN and D. SCHMITT
particular postulated that rare earth pnictides, where the nearest-neighbour exchange is dominated by direct exchange and superexchange, are excellent candidates for large two-ion magneto-elastic effects (Lacueva and Levy 1985). From an experimental point of view, anisotropic bilinear couplings have been observed in the magnetic excitation spectrum of TbP (Loidl et al. 1979) and PrSb (Vettier et al. 1977). They have been also used for explaining the different antiferromagnetic structures found according to the temperature in GdX pnictides. Associated with these structures, different two-ion magneto-elastic strains have been also detected (Hulliger and Stucki 1978a, Hulliger and Siegrist 1979). For the experimental conditions which are used to study the third-order magnetic susceptibility , anisotropic bilinear interactions occur through their strain derivative, i.e., the two-ion magneto-elastic coupling, which may be described within the MFA formalism (Aldonard et al. 1984b). For DySb this coupling, characterized by a D 1 coefficient in fig. 26, is opposite to the quadrupolar interactions and improves the fit down to 12 K. However, a second-order transition would be expected. The same short-comings are also observed on the temperature dependence of the reciprocal susceptibility under an uniaxial [001] stress (Morin et al. 1985a). Even if present, two-ion magneto-elasticity, considered in the MFA, is not able to describe the features observed close to T~. On the other hand, a large number of studies has shown that NaCl-type compounds exhibit fluctuations through their critical behaviours close to the N6el temperature (Taub and Parente 1975, Hfilg et al. 1985). These critical fluctuations appear to be favoured in a deformable lattice (Taub and Williamson 1973). In DySb, as in HoSb, neutron diffraction measurements have shown an anomalously large magnetic scattering above T N. In the same type of experiments, the paramagnetic correlations in ErSb are characterized by a strong two-ion anisotropy (Knorr et al. 1983). In conclusion, the nature of the magnetic transition in numerous monopnictides seems to result from a well-balanced competition between different couplings. For many cases, it is difficult to determine the driving mechanism as e.g., in HoSb, which is very close to tricriticality (Jensen et al. 1980). The determination of quadrupolar interactions, coherent in CsCl-type compounds, cannot be achieved in these series. In addition the large dependence of anisotropic magnetic interactions on metallurgical stresses may change the nature of the transition in a given sample as shown in Er compounds (Hulliger and Natterer 1973). As in DySb the difficulty in properly considering fluctuations and anisotropic bilinear couplings may prevent one to fully understand the first-order transitions in DyBi (Hulliger 1980), DyP, DyAs (L6vy 1969), ErBi and ErSb (Hulliger and Natterer 1973) even if quadrupolar interactions are obviously present as proved by the temperature dependence of elastic modes. The R A l 2 series. RA12 compounds (C14 structure) appear to pose a similar problem as was observed in pnictides (del Moral et al. 1987). DyA12 orders ferromagnetically at T c --61.4K through a second-order process. The third-order magnetic susceptibility along the [001] easy magnetization direction is 3.2.4.2.
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H INTERMETALLICS
-3 x 10"5~ /oB/KO¢3
I Gt=OrnK
I
70
80
53
mK i
-2 <3 -1
0
I 60
90
T(K]
Fig. 27. Temperature dependence of the third-order magnetic susceptibility A.... = X~ ) for HI[[001 ] in
DyAI2. Full lines are calculated with a bilinear coefficient n = 17.5 kOe//xB and the values indicated for the quadrupolar parameter G1 = G ~ (from Del Moral et al. 1987). negative in the entire paramagnetic range and cannot be described with a unique G ~ coefficient (fig. 27). However, the data close to T c do not indicate critical behaviour. The present range (8-10 mK) of G ~ is different from values deduced from magnetization measurements in the ordered phase, i.e. - 2 m K (Rossignol 1980). The same failure for describing the third-order susceptibility is found again in ErA12 (del Moral et al. 1987). Data go from positive values at high temperature to negatiye ones close to T c which countermands a first-order transition. The 2 possible G ~ values deduced are larger than in DyA12 in violation of the o~j variation law across the rare earth series (Schmitt and Levy 1985) (see section 5). All these features may be the signature of additional couplings and an improved knowledge would need systematic studies in particular of ultrasonic properties.
3.3. Determination of the quadrupolar parameters from susceptibility techniques Different ways of determining quadrupolar parameters are directly related to the various susceptibilities described in section 2. Indeed, these susceptibilities connect the magnetic variables on one hand and external stresses on the other hand. Thus, (i) the strain susceptibility represents the response of the quadrupolar moments to the corresponding stress; (ii) the quadrupolar field-susceptibility connects the quadrupolar moments to the magnetic field; (iii) the third-order magnetic susceptibility couples the magnetic moment, more exactly the initial curvature of the magnetization curves, and the magnetic field. Each of these three
54
P. M O R I N and D. S C H M I T T
susceptibilities may be acquired by a different experimental technique, namely measurement of the elastic constants, the parastriction and the low-field magnetization, respectively; these techniques are described below. 3.3.1. Elastic constants
As seen in section 2.2, the existence of a magneto-elastic coupling leads to the softening of the elastic constants of appropriate symmetry when the temperature is lowered, and this softening is enhanced by positive two-ion quadrupolar interactions. The experimental technique generally used is the ultrasonic velocity measurement, and several geometrical configurations are considered in order to observe the various ultrasonic propagation modes. In general, the normal (symmetrized) elastic constants are not directly accessible; they have to be obtained from a linear combination of those which are measured. The observation of a softening of elastic constants due to magneto-elastic coupling in the intermetallic rare earth compounds was realized in the early 1970s, and extensively studied in particular in cubic rare-earth antimonides (see figs. 28 and 29). Several behaviours are observed, according to the strength of the quadrupolar (magneto-elastic) coupling as well as the nature of the CEF ground state. The temperature variation of the elastic constants in LaSb is taken as the background for the other compounds. Large anomalies are often detected at the magnetic ordering temperatures and below; they are due to domain-wall stress effects. However, the most spectacular effect remains the softening above the ordering
I.C
0.8
0.6
O~
°il I
'
8b
'
I
J~o
'
z~,o
TEMPERATURE(°K)
'
3~o
Fig. 28. Change in C o = C ~ = (Cal - C12) as a function of temperature in DySb. A fit of the data leads to (g0) 2 = GEE = 1 . 4 m K and I o = K v = - 0 . 4 m K and thus G ~ = 1 mK (from Moran et al. 1973).
Q U A D R U P O L A R EFFECTS IN R A R E EARTH INTERMETALLICS
14.0~t3.9~
55
TmSb
13.8 ~:
13.6 r ~
13.5 Ci1-C12 • I
1
5
.
2
el
~
~2.70 ee o
- • • ee~ e
~:__.z6.sz6.6z6.4z6.~
o
°
C44 ee
"
I
I
Ioo
200
300
T (°K)
Fig. 29. Temperature dependence of the symmetrized modes Clt - C12 = C 7 and C44 = ½ C e for TmSb. Full lines are calculated with only magneto-elastic couplings in the presence of CEF (from Mullen et al. 1974).
temperature, e.g., this softening reaches about 60% of the room-temperature value of C r at 10 K in DySb and occurs over a wide range of temperature (see fig. 28). A good description of the softening of the C ~ elastic constant requires the introduction of a two-ion quadrupolar interaction in addition to the one-ion magneto-elastic coupling (Levy 1973, Moran et al. 1973). The cubic TmX series (X = Zn, Cd, Cu, Ag, Mg) is another series which is particularly spectacular with regard to the softening of the C v elastic constant. For example in TmZn and TmCu (fig. 30) the softening reaches more than 50% of the room temperature value just above the ordering temperature, indicating very strong quadrupolar interactions (Liithi et al. 1979, Jaussaud et al. 1980). In the opposite limit, the softening is much less pronounced in TmAg and TmMg (Giraud et al. 1986, Giraud and Morin 1986). It is worth noting that the analysis of the temperature variation of the elastic constants may give information about the crystal field in addition to the magnitude of the quadrupolar parameters. Indeed, according to the nature and spacing of the low-lying CEF levels, anomalies may occur at low temperature in the strain susceptibilities, thus in the elastic constants (see section 2.2.3). This is particularly emphasized in singlet ground-state systems as in PrSb (Liithi et al. 1973a) and TmSb (see fig. 29) for both C ~ and C ~ modes.
56
P. M O R I N and D. S C H M I T T
I
I
I
l
l
C~(YCu) •/ ' 1 / " .~.I
/
,t~
/.
E
c" :½(c,-c,2
?
o
I
O,l 0
T
t..j
i
- -
experiment fit with G1 = l l . 5 m K R1 = 7,0 mK
J
I
l
100
I
I
2OO
300
T(K)
Fig. 30. T e m p e r a t u r e d e p e n d e n c e of the C' = ½C r = ½(C n - Ct2 ) m o d e in T m C u ; the full line is a theoretical fit with G 1 = G ~ and K 1 = K ~ ( f r o m J a u s s a u d et al. 1980).
3.3.2. Parastriction Parastriction constitutes a very suitable technique for investigating magnetoelastic coupling. It consists of measuring changes in the length of a sample in an applied magnetic field in the paramagnetic phase (Morin et al. 1978c, 1980b). It differs from the usual magnetostriction by the fact that only the low-field variation of the latter is analyzed, namely its H2-dependence. Indeed, as shown above (section 2.2.1), within the limits of validity of the perturbation theory, the coefficient of this HZ-variation is related to the quadrupolar field-susceptibility Xo (or Xp) which depends only on the crystal field and on the magnetic and magneto-elastic parameters. In addition, in agreement with the classical hightemperature approximation (Callen and Callen 1965), a linearization of this susceptibility may be achieved by plotting ) ( Q 1 / 2 a s a function of the temperature. Deviations from this linear behaviour may occur at low temperature due to CEF effects in the same way it does for the magnetic susceptibility. Experimentally the measurement of the relative change of length A = 8l/l is performed successively along and perpendicular to the direction of the applied magnetic fields; one eliminates the volume effects e ~ by considering the difference All - A~. As a consequence, for a magnetic field along a four-fold or three-fold direction, and according to the expression for 81/l [eq. (25)], one obtains respectively
(All- Al)v =
~93_B~ 2 C---~oXQH2'
3 B+ (All- A±)+ = V'2 C O x e H 2 '
(110)
for the two tetragonal and trigonal strain modes. It is worth noting that the same
Q U A D R U P O L A R EFFECTS IN R A R E EARTH INTERMETALLICS
57
relations are found, by considering a fixed measurement direction and by applying successively the magnetic field parallel and perpendicular to that direction. This latter geometry is habitually used by experimentalists for the sake of simplicity. As mentioned above it is useful to plot HI(I All - A± I) 1/2 in order to observe the linear high-temperature b~haviour. A good example is provided by the compound TmZn where the quadrupolar effects are particularly strong (fig. 31). According to eq. 110, the high-temperature slopes give the magneto-elastic coefficients B v and B', as the background elastic constants are known from other measurements. ~I~
~
I
0
,
v
o, o / / '
I
O•
."
I
I
1
Fig. 31. Parastriction of TmZn, measured for the tetragonal and trigonal symmetry lowering modes. Curves are calculated within the susceptibility formalism (G ~= Ga) (from Morin et al. 1980a).
In addition, large deviations from this linear behaviour are observed for the tetragonal mode below 100 K due to CEF effects and to a strong G v parameter. On the other hand the anisotropy of the parastriction is particularly obvious in this compound, showing that measuring the magnetostriction on polycrystals is quite meaningless in these anisotropic systems. Note also that the parastriction gives the sign of the magneto-elastic coefficients, while only its absolute value may be obtained from elastic constant measurements. This technique has been successfully used in several series of rare earth compounds, in particular in RZn compounds where the coefficients associated with both the tetragonal and trigonal strain modes have been obtained for the heavy rare earths (figs. 32 and 33).
3.3.3. Third-order paramagnetic susceptibility The third-order paramagnetic susceptibility has been introduced as a new method for studying quadrupolar interactions in rare-earth compounds (Morin and Schmitt 1979, 1981b). It consists of a detailed analysis of the magnetization induced by an external field in the paramagnetic phase. More precisely it
58
P. MORIN and D. SCHMITF =
i
1S
Er Zn/y
/.
H . [ oo]
.
..~
.
•
,/ TmZn
%s DyZn
"r
20o
too
300
(K)
TEMPERATURE
Fig. 32. Temperature variations of g/I;,, All 1/2 for tetragonal strain in RZn. The full lines are calculated dependencies (from Morin et al. 1980b). -
H //
8 ~o
1o
v
[111]
TmZn DyZn
5
TbZn
"I-
i
100
TEMPERATURE
200
300
(K)
Fig. 33. Temperature variation of H/l All- A±I 1/2 (rhombohedral strain) in RZn. The full lines are calculated with the susceptibility formalism (from Morin et al. 1980b).
corresponds to the second term (in H 3) of th e development of the magnetization as a function of the magnetic field, this term depends on the total quadrupolar interactions in addition to the crystal field (see section 2.2.1). An experimental way to investigate the magnetization curves is to plot M / H versus H2; this provides both the first-order magnetic susceptibility XM, i.e. the
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
59
isotropic null field value, and the third-order one X~ ) , i.e. the anisotropic slope of the linear low-field range. This is well-illustrated by the compound TmCd, where the anisotropic character of X~ ) is particularly obvious (see fig. 34). It is worth noting that the linear part of these curves diminishes as the temperature is decreased, since the higher-order terms (in H s , . . . ) become relatively more and more important. An alternative experimental way, although less direct, is to consider the Arrott's plots M 2 versus H/M. In low field, the variation is linear; the intersection with the x-axis yields (X~)) --1 while the third-order magnetic susceptibility is related to the slope dM2/d(H/M) through the relation
X(~ )
-(XM)4/(dM2/d(H/M)) .
=
The variation of the third-order magnetic susceptibility may then be drawn as a function of the temperature and analyzed with the above formalism (see fig. 35). As an other example, in fig. 36 we show the temperature dependence of X~ ) for the two main symmetry directions in PrPb 3. The anisotropic character of X~ ) is obvious from these curves again, as well as its dependence on the bilinear and quadrupolar interactions. The third-order magnetic susceptibility also explains the paramagnetic anisotropy of the magnetization observed in TbxLal_xA12 for a constant magnetic field within a given plane (Hoenig 1980). The technique of third-order magnetic susceptibility has also been applied to the tetragonal compound TmPO 4 which presents noticeable non-linear terms in the low-field magnetization within the basal plane (Andronenko et al. 1983). Another experimental way to extract X~ ) is to perform torque anisotropy 120
i
"
~
"- "
115
[001]
6 11o ~,
"
"
"
"
,
[111]
TmCd
' ~ 105 EI 77
"
~ - - - - ' -
"" ~ ~
""
76
K
[OO]l
75 74 .....
0
i 100
2
Hi
I ~ 200
I 300
400
(k(De2 )
Fig. 34. Experimentalvalues of M / H versus H 2 in TmCd (from Morin and Schmitt 1982c).
60
P, MORIN and D. SCHMITT
//// ///"
&"
!// II
-4
:o,5 K
,:o
/
×
~
5
,
=-o.34
i
10 15 20 TEMPERATURE ( K )
i
25
30
Fig. 35. Experimental and calculated variations with temperature of X~ ) along the [001] axis in TmCd system. Note the dependence of the coefficient G1 = G ~ on dilution (from Morin and Schmitt 1981b).
1 e*=-9K GI=O '. -
tI
i
l
H//[001]
~o O ~.~:E ~<
I
I
H//[111]
'
]
\", i L
i
PrPb3 W=-.65K x=.6
\
1¢---9K JG .3K
i . . . . . . . . . ;_:.=_._.....
K ............2 : 2 : ~ _-'r='- .....
'G =, --.9. -15mK -/ , //--GI="0' ' - I . IG2= ~ O*---09K \
,e*-gK ;.2K O,=G2__O// //
//
/
~_ /
// /
-1
I
0
5
I
I
10 5 TEMPERATURE (K)
/I
10
15
Fig. 36. Temperature variation of the third-order susceptibility along [001] and [111] in PrPb 3. Dots are experimental points. The dashed lines correspond to behaviours without quadrupolar interactions, but show the effects of the bilinear interactions (0" = - 9 K). The full lines show quadrupolar effects (G v= G1, G ~ = G2) (from Morin et al. 1982). m e a s u r e m e n t s which, in p r i n c i p l e , allow us to d e t e r m i n e p a r a m a g n e t i c susceptibility t e n s o r s o f f o u r t h (~- X ~ )) a n d e v e n sixth r a n k s a n d has b e e n a p p l i e d to R A I 2 c o m p o u n d s ( D e l M o r a l et al. 1987). A n a l t e r n a t e a p p r o a c h to a s c e r t a i n i n g t h e q u a d r u p o l a r effects o n t h e m a g n e t i z a t i o n is b a s e d o n t h e n o n - l i n e a r i t y o f t h e e x c h a n g e field Hexch as a f u n c t i o n o f M
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
61
(Kouvel and Brun 1980) Hex~h = AM
+ A'M 3 .
(111)
However, this procedure provides a )t' value which is temperature dependent, since it is related to the true quadrupolar parameter G" through the relation )t, =
2G u'(X~)) 2 (X0)4( 1 _ G~X# )
(112)
This ,V parameter is therefore connected with the quadrupolar part of X}~); its temperature variation has indeed been observed (Abu-Aljarayesh et al. 1986a)
3.4. Effects on the magnetostriction While parastriction allows us to describe the initial magnetostriction in the paramagnetic phase, i.e. the first term in its field development, a full diagonalization of the Hamiltonian describes the whole magnetostriction curve, in the ordered or non-ordered state, provided the magneto-elastic couplings and all the two-ion interactions are included.
3.4.1. Magnetostriction of dilute compounds The main advantage in investigating dilute magnetic compounds is to be free, at least to a large degree, of two-ion interactions. Indeed it is these interactions that make the analysis difficult. If the effect of the dilution on the macroscopic measured strain is correctly taken into account, one can determine the microscopic one-ion magneto-elastic coupling. Usually authors simply assume that the macroscopic deformation is proportional to the magnetic impurities concentration (Nieuwenhuys et al. 1982b, Pureur et al. 1985). A systematic study of the forced magnetostriction has been carried out for the cubic noble metals compounds (silver, gold) containing magnetic rare earth impurities (Creuzet and Campbell 1981, Campbell and Creuzet 1985). The magnetostriction follows the second-order Stevens coefficient a I (see figs. 37 and 38), indicating that a single-ion model is pertinent, and that the second-rank magneto-elasticity is preponderant in these systems. Nevertheless, a fourth-rank magneto-elastic coupling has also been determined, its contribution is about ten times smaller than the second-rank one. In addition to the magneto-elastic coefficients, the magnetostriction curves are strongly affected by the nature of the low-lying CEF levels. This allows one to choose between possible levels schemes through the shape as well as the temperature variation of the magnetostriction curves (fig. 39). Another spectacular effect related to the crystal field concerns the so-called "reverse" magnetostriction, i.e. the sign reversal of the magnetostriction between the low and high temperatures. This occurs in CexLal_xSb, SmxLaa_xSb (Nieuwenhuys et al. 1982a), and CexLal_~Bi (H~ifner et al. 1983) compounds.
62
P. MORIN and D. SCHMITF 14,
y
lu-,,-~nt
12
10' 8 6
~
14. •106x6L "~Ag Tb
Ag Tb
t
12.
10, 6
A
4 2 0
2
~
2 -4
--6 -8
kG
(a)
0
AgEr
-2
X,~Tm \
-t, -6 -e
~
o
H.kG 0AgEr AgTm (b)
Fig. 37. Parallel magnetostriction data for a field applied along the [001] axis of silver-0.5% rare earth single crystals at 1.4 K: (a) experimental data, (b) calculated neglecting the e~ contribution (from Campbell et al. 1979).
6, ' 106 x"~t
.
/
6-
/AgTb
4.
AgDy
4 / - -
A_gHo
4u , 4u , 6,0 ,H.kG 0[ ~
,
H,kG
-2 -4. •
(a)
-q
fo)
Fig. 38. Parallel magnetostriction data for a field applied along the [111] axis of silver-0.5% rare earth single crystals at 1.4 K: (a) experimental, (b) calculated neglecting the e~ contribution (from Campbell et al. 1979).
For example, the change of sign of the measured strain takes place at 2 K in LaBiCe and at 8 K in LaSbCe (see fig. 40). This can be explained by a simple crystal field theory, which, in some particular cases, leads to an expectation value of the quadrupolar operator O ° which is not always positive. For example, in Ce0.15La0.85Sb, the quadrupolar m o m e n t induced by an external magnetic field on the F 7 doublet ground state is negative at low temperatures. It changes its sign either for very large magnetic fields (about 40 T at 1.5 K) or above 8 K for any field, namely when the influence of the excited F 8 quartet becomes important. This change of sign of the initial magnetostriction curvature is accounted for by the formalism developed in section 2 and corresponds to the change of sign of the quadrupolar field susceptibility x~ - (2) (see fig. 41). Similar reverse magnetostriction
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H INTERMETALLICS
~-110-6) 20
40
60
t
t
i
63
H(kG)
_1
_2 ~2.8 t
_3
Fig. 39. [001] parallel magnetostriction in ErA g (concentration 0.5%) corrected for the volume contribution; circles are experimental data, solid lines calculated. Note the anomalous variation with temperature which is characteristic of the CEF level scheme (from Creuzet and Campbell 1982b).
LaBi:Ce (lO°/o)
6 4
-5T
2
/I 3T
O -2
_1 ~-
-4
o
~ ~\
3 2/
O
LaSb:Ce (15%)
4.8T k,,..~ 4 T
5
10 15 T , (K)
20
Fig. 40. Comparison of the magnetostriction of (a) LaBi : Ce (upper panel) and (b) LaSb : Ce (lower panel) as a function of temperature at various fields. As clearly seen the "reverse magnetostriction" occurs at a lower temperature for (a) than for (b). Solid lines represent a fit to theory (from H/ifner et al. 1983).
64
P. M O R I N and D. SCHMITT r
50
-50
o
LaSb:Ce
LoBi:Ce
/ /
40o
•>• 0
-400
lo
T (K)
20
Fig. 41. Temperature dependences of Xy(~) , the quadrupolar field-susceptibility, calculated in the presence of only the CEF for LaBi : Ce (a), LaSb : Ce (b), LuCd : Tm (c) and LuZn : Tm (d). In these four cases the CEF level scheme drives the so-called reverse magnetostriction.
effects may be expected in other situations, e.g., with a F5 triplet as ground state, as in TmCd or TmZn, for an appropriate dilution.
3.4.2. Magnetostriction in the ordered phase The magnetostriction observed in the ordered phase of magnetic compounds is generally much larger than that in the paramagnetic phase. This is due to the large influence of the exchange field which tends to saturate both the magnetic and quadrupolar moments. The symmetry of the cell distortion is directly related to the direction of the magnetic moments, e.g., in ferromagnetic compounds. In cubic antiferromagnets, the observation of a spontaneous strain may be very helpful in determining the actual magnetic structure (see section 3.6). A quantitative analysis remains more difficult than for a non-ordered state, because of two-ion effects. Depending on the strength of the bilinear exchange coupling, it has been done using either a semi-classical (see section 2.4) or a quantum treatment (see section 2.1.2). In the case where the bilinear interactions are big, large magnetostrictions have been measured at room temP6erature, as e.g. in TbFe 2 where a record magnetostriction of Am = 2400 × 10- was reported (see fig. 42). This motivated numerous investigations because of the interest of technological applications. Still higher strains have been measured at low temperature in the RFe 2 series (see section 4). Huge magnetostriction has also been observed at low temperature in numerous series where the alloyed metal is nonmagnetic, in particular within the NaC1- and CsCl-type structures (L6vy 1969, Morin et al. 1977). Large tetragonal strains of the order of 10 - 2 have been measured; e.g., c / a - 1 =--1.4% in PrZn at 1.5K (Morin and Pierre 1975). In TbZn a change of easy direction within the ordered
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H I N T E R M E T A L L I C S
I
I
I Tb Fe2
'
L
]
65
|
2,000I
~.poly [hi
1,000 t Y
0
V
I
I
I
I
I
5
10
15 H [kOel
20
25
F
Fig. 42. Magnetostriction of TbFe z at room temperature for a single crystal (Aiq curve) and a polycrystal taken for two different orientations (a and b) (from Clark et al. 1974).
10 -2
10- 3
IO-~
Lu
TbZn
1
io-s
10 -6
u~o.s 10" ;' 0 10 - 8
.....
i
100 i
100
200 )
I
200 TEMPERATURE OQ
300
Fig. 43. Thermal variations of the spontaneous strain e 3 = e~ above the temperature at which there is a change in direction of the easy axis and the e 3 strain induced by a 1 kOe field in the paramagnetic state for TbZn ((3, strain-gauge data; Q, capacitance dilatometer data; full curves are calculated variations). Inset: experimental variation of the volume anomaly e v = X/3 e" (from Morin et al. 1977).
66
P. MORIN and D. S C H M I T r -2 10
10.3
10-4
10~ 5 U3
TbZn
10-6
TC
H: lkCe 10- 7
\ ,
10- 8
,
~"
q.
100 200 TEMPERATURE (K)
300
Fig. 44. V a r i a t i o n s w i t h t e m p e r a t u r e o f the spontaneous strain sxz = e~/~/2 b e l o w the temperature where the easy axis changes, and of the exz strain induced by a 1 kOe field in the paramagnetic range for ThZn ( 0 , strain-gauge data; @, capacitance dilatometer data; full lines are calculated variations) (from Morin et al. 1977). ,
Tm Zn
2
~
,
T - 107 K 1
0
-- '?1 _..
01i°, E
9 10 11 TEMPERATURE (K)
-4
o
2
,
6
;
APPLIED FIELD
;o
12
(kOe)
Fig. 45. Experimental variations of the strains parallel (All) and perpendicular ( A l ) to the four-fold direction of the magnetic field at 10.7 K in TmZn. Full and dashed lines are relative to fast and slow rates of variation of the field. The inset shows the temperature dependence of the critical field H c (from Morin et al. 1978b).
QUADRUPOLAR
EFFECTS IN RARE EARTH INTERMETALLICS
67
range occurs; this allows one to follow the temperature variation of the strain associated with the [001] and [101] moment axes (figs. 43 and 44). Finally, in TmZn one transition appears on the magnetostriction curves at temperatures above the Curie temperature (see fig. 45). This behaviour, also observed on the magnetization curves (see section 3.5), can be explained by the appearance of a strained ferromagnetic state above a critical field within the paramagnetic phase, arising from strong quadrupolar interactions.
3.5. Effects on the magnetization processes 3.5.1. Paramagnetic phase Perturbation theory has shown that quadrupolar interactions strongly influence the initial curvature of the magnetization as a function of field in the paramagnetic phase, i.e., the third-order paramagnetic susceptibility (section 2.2). In the same way, quadrupolar couplings influence the higher-order terms, i.e. they have an effect on the whole magnetization process. A diagonalization of the full Hamiltonian allows us to take these couplings into account particularly at high fields. As an example, the magnetization is predicted to be largest along a three-fold axis in [111]
(ooi] I
6
G,=9mK
.... +.:;,~ I
.0 ~+~+- +.+.+-16K
,
/ . +] ~o+- ~o+2OK I
4
-/'/+/4,/*/+°I o/'/+ 4,/ /0 I / /4,/ / ] 0 I/I /4" 4,0 /
d'i
2
o~ +~
•..'+25K,
/n
-
W=1.4 K x =-.42
•
• 4,o011/ +4,0 • 1+4,0 • o+~Oll u
+.,.
I
~'."
jEP=_ 3K
J
"'"
x =-42
~ ,
4 I-
W=I.4K E)w:I 3K ,
" 9K
+ ~...
..-'.~"
G.=gr~--I "11 -----
•
• +~/@/4,/0/
..,+0o., •
--
4~~
e~ e 0 i / @ .:+...-..
04"@0
0
.-1;:;. / 4+
|
e/o/../•
i+tOll t ~+ ll . .
~.-
I
G2=-80 mK
//+."
/."
/
.di.".
-
®// m II
,g .'"
0
I;::"
,
0
SO
/ 100
0 H(k~)
W=I.4K
? ." .e • •
x =-.42
!
50
it~=. 3 K 100
Fig. 46. Magnetization curves along the directions [001] and [111] for TmCu at various temperatures; lines are theoretical fits with the p a r a m e t e r s indicated ( G 1 = G ~, G 2 = G ~) ( f r o m J a u s s a u d et al. 1980).
68
P. MORIN and D. SCHMITF
TmCu in the absence of quadrupolar interactions; by using appropriate values of the parameters G ~ and G ~ one can restore the four-fold axis as the direction where the magnetization is maximum, as observed experimentally (see fig. 46). The magnetization curves show a more spectacular behaviour in case of huge quadrupolar couplings, as in TmZn (Morin et al. 1978b). In this compound, the quadrupolar coupling is strong enough to induce a transition from the paramagnetic to a ferromagnetic state in a wide range of temperatures above the quadrupolar ordering (see fig. 47). This transition occurs for a critical field which depends on the strength of the quadrupolar interactions and is well-described within the MFA. The critical field is also present along the directions of hard
.~
1K 111
0
20
(b
40
20
60
INTERNAL FIELD
80
100
120
(kO~)
Fig. 47. Magnetization curves of TmZn along the [001] axis at different temperatures; the insets show the temperature dependence of the critical field He; (a) experimental, (b) calculated (from Morin et al. 1978b).
QUADRUPOLAR
E F F E C T S IN R A R E E A R T H I N T E R M E T A L L I C S
69
magnetization, although it is larger than along the easy direction, i.e. [001]. For example, along the [101] direction, the magnetic moment is first aligned with the external field, then rotates towards the [001] axis when the component of the field along the [001] direction reaches the associated critical value; finally it comes back toward the initial [101] axis when the field is large enough.
3.5.2. Ordered phase In the ordered state, quadrupolar interactions play an important role in the free energy balance between the various crystallographic axes. In particular, when the field is applied along a direction of hard magnetization in a ferromagnetic compound, the magnetic moment rotates from the easy axis towards the field. The area between the corresponding curve and that extrapolated from the high field behaviour is related to the anisotropy of energy between both directions. This energy difference is strongly influenced by the quadrupolar couplings, as for example in ErZn (see fig. 48). This behaviour is enhanced when the free energies corresponding to two moment directions cross each other at a given temperature T R within the ordered phase. This leads to a first-order phase transition in which the direction of easy magnetization changes. This phenomenon is driven by entropy contributions induced by the anisotropic splitting of the CEF level scheme under the exchange field (see fig. 49). There are many examples in rare earth compounds, especially
6
5F//
b >
, 1mK-
~_o4~.4
* [101]
~I Z
=(I 1 I)
3 I
'
'
'
l
O N FLU Z L~ :E
,
Z/
~ 10
i:I \ 1 TEMIPERATUREI(K) .
4 0
;o
~o
INTERNAL
30 FIELD
40
50
(kOe)
Fig. 48. Magnetization curves in E r Z n along the three principal cubic directions at (a) 4.2 K and (b) 10 K; the inset shows the temperature dependence of the critical field; the points are the experimental data along [101] (O), [111] (zX) and [001] ( 0 ) ; the lines are the theoretical fits (broken curves: without quadrupolar terms; full curves: with G ~ = G 1 = 1.5 m K and G ~ = G 2 = 11 m K terms (from Morin and Schmitt 1978).
70
P. MORIN and D. SCHMITT - 250
-50P
///
HoZn 12
-
// /- ~" " i l l
v
>.-
L9
"Tl
-15OF
-26G
0~ W
Z W
/
W W
_.__~[i01]
//
----
// . . . .
G2:0
/
rr
b_
.... -270 0
G1 : - 4 mK G2= 10 mK
/ / ~[101] i 20
GI=0
i 40
TEMPERATURE
-2sot-
0
(K)
80
[lOl]
[111]
Fig. 49. (Left-hand part) calculated temperature variation of the free energy along [101] and [ l l l ] for HoZn, without quadrupolar terms (broken curves) and with G ~ and G ~ contributions (full curves); (right-hand part) arrangement of the lowest CEF levels in HoZn at T R = 25 K according to whether the moment lies along the [101] or [1ll] direction and taking into account the quadrupolar coefficients G ~= - 4 inK, G ~= 10 mK (from Morin and Schmitt 1978).
with Nd, Tb and Ho (see, e.g., Gignoux et al. 1975, Sankar et al. 1977, A16onard and Morin 1985). In each compound, the two-fold axis is the direction of easy magnetization at low temperature. In HoZn for example, the critical temperature T R is calculated to be 48K with tile CEF alone (Morin and Schmitt 1978). Introducing appropriate quadrupolar parameters G r and G ~ shifts the free energy curves and allows to reduce T R to 25 K in agreement with the experimental data (fig. 49). The effects of quadrupolar interactions on the magnetization processes in antiferromagnets are more difficult to analyze due to the existence of several sublattices (Jensen et al. 1980). On the other hand, they are more relevant for the stability of the actual magnetic structures. This will be discussed in the next section (3.6).
3.5.3. Effect on the direction of easy magnetization In the previous section, the dominant CEF determined the direction of easy magnetization and the quadrupolar interactions determined the exact temperature TR, for the change of the easy axis. In this section, we discuss some cases where the quadrupolar energies are larger than the CEF anisotropy. Good examples are provided by the ferromagnets DyCd and DyZn which have the CsCl-type structure (T c - - 8 0 and 140K, respectively). In both cases, a four-fold axis is observed as the direction of easy magnetization and a large tetragonal spontaneous strain occurs in the ordered phase (0.8% at 4.2K in DyZn) (Morin et al. 1977). Their cubic level schemes are given in fig. 50a together with those of the antiferromagnets DyCu and DyAg. Their anisotropic splitting by bilinear interactions always favours the three-fold axis: the CEF
Q U A D R U P O L A R EFFECTS IN RARE EARTH INTERMETALLICS
zOO
71
100
I
r(3) DyAgi~--~. <..8 ~
!~-Y~.-~-~.~ /
200
c11"~i
®
.....
o
I
@easyaxis [I11]~k~O011
-200
2s~
W=-.57K \ 6
~
(-.}7)
-400
¢_-60K
',
(.9
\
T:4.2K J
(-5)
0 0
-.5
X
Fig. 50. (a) CEF level scheme in dysprosium compounds having the CsCl-type structure. E is the energy in Kelvin, W is a scale factor, the value of which is given in parentheses for each compound, x is the ratio between fourth- and sixth-order CEF terms; (b) x-dependence of the strain susceptibilities, X2 (=Xr) and X~ (=X~), for a given W at low temperature; (c) easy axis at low temperature according to x and G1 ( = G ~) for given values of W and 0". The full line is mainly related to the case of DyAg and DyCu; for DyZn and DyCd, the critical line would lie lower (from Al6onard et al. 1984a).
anisotropy between three-fold and four-fold axes is too large in DyAg and DyCu to be reversed by the tetragonal quadrupolar energy; the three-fold axes remain the easy ones and the (negative) trigonal quadrupolar interactions stabilize the muti-axial spin arrangement (section 3.6.1). In DyZn and DyCd, tetragonal quadrupolar interactions have the same order
8
-J - ~ m ] ~ ~ ~ _ _ .1 t~o~. . . . . . . . . . . -o-
7 "
i
~
" o
.
o
n
o
o[~OO1~
-
- -
2
GI=ImK ,G2=-22inK
N
ErNi 2
Z
. . . . . GI=O ,G2=0 T=1.5 K
:E'
o
2'o
go
go
8'0
'60
1~,o
INTERNAL MAGNETIC FIELD (kOe)
Fig. 51. ErNi2: magnetization variation at 1.5 K as a function of internal field. ©, • and [] are experimental values. Full lines are the variations calculated with 0* = 7 K, W = - 1.0 K, x = - 0.35, G1 (=G ~) = 1 mK and G 2 ( = G ~) = - 2 2 m K . Dashed lines are the variations calculated without quadrupolar terms (from Gignoux and Givord 1983).
72
E MORIN and D. SCHMITr
of magnitude as in DyCu and DyAg, but are more efficient due to the groundstate wave functions (see fig. 50b). They are then able to make the four-fold axis the easy direction above a G r critical value depending on the exact level scheme. This is realized for D y Z n as well as for DyCd (A16onard and Morin 1985). These two compounds represent the border-line case of quadrupolar effects on the temperature T R for the change of the axis of easy magnetization in the ordered phase (section 3.5.2); also in the paramagnetic state the quadrupolar effects are noticed through anisotropic magnetization processes. Similar features may be found also in other series specially when the C E F anisotropy is weak. This is, e.g., the case in ferromagnetic ErNi 2 which exhibits a [001] easy magnetization axis (fig. 51). The magnetization processes along the three main cubic directions are characterized by a crossing between the [001] magnetization curve and the [101] and [111] ones. Their description, as in D y Z n and DyCd, is impossible without quadrupolar interactions stabilizing the [001] axis as the easy one, instead of the [111] axis. 3.6. Effects on the magnetic structures 3. 6.1. Multi-axial structures In addition to their effects on the magnetic transition, the quadrupolar interactions also contribute to the determination of the magnetic structure through the minimization of the free energy. It is well-known that in cubic symmetry the spin structures that are described by one or more propagation vectors, q, which belong to the same star have the same bilinear energy and lead to identical powder neutron diffraction spectra (Wintemberger and Chamard-Bois 1972). The presence of several propagation vectors determine a multi-axial arrangement for spins (see, e.g., fig. 52). The knowledge of the C E F allows one to partly remove this ambiguity.
(17 FI O )
(a)
(b)
@=90 °
(c)
(d) Fig. 52. Possible configurationsfor the ( ½, 1,0)-type antiferromagneticstructure with 0 = 90°; (a) and (d) collinear; (b), (c) and (e) bi-axis; (f) multi-axis. All these structures lead to the same powder neutron diffraction spectrum as that observed in PrAg (from Morin and Schmitt 1982b).
Q U A D R U P O L A R EFFECTS IN R A R E EARTH INTERMETALLICS
73
The magneto-elasticity can determine the actual structure. For instance with a four-fold easy axis, a large ( c / a - 1) magnetostriction may be a clear signature for a collinear arrangement as observed in many CsCl-type compounds (Morin et al. 1978a, Schmitt et al. 1978). On the contrary, the existence of collinear structures with three-fold easy axes has not been demonstrated for CsCl-type rare earth intermetallics. However, there is plentiful evidence for this in the NaCl-type series, in particular in Tb compounds (L6vy 1969, Hulliger and Stucki 1978b). In the presence of negative quadrupolar interactions, multi-axial spin structures are stabilized. This occurs within the trigonal symmetry in CsCl-type compounds as DyAg, DyCu (A16onard et al. 1984a) and NdZn (Morin and de Combarieu 1975) and in the AuCu3-type compound TmGa 3. In the latter compound for instance, powder neutron diffraction spectra have been indexed by using a [ ½ ½0] propagation vector and a [001] spin direction and the hypothesis of a collinear spin arrangement (Morin et al. 1987a). In spite of the sizeable magneto-elastic coupling found in the non-ordered phase (Morin et al. 1987c), no spontaneous strain was observed, making this collinear structure unlikely. The actual spin structure was determined by neutron diffraction under a magnetic field applied along the three main axes of a single crystal (Morin et al. 1987d). No change of magnetic intensities was detected when cooling the sample in different applied fields. This indicates the sample is spontaneously single domain; consequently, the actual structure is a triple-q structure which preserves the cubic symmetry (see fig. 53). This solution was confirmed by determining the different structures revealed by magnetization processes. In particular, along the Jill] direction, the quadrupolar structure and energy are unchanged at the lower critical field, He1, only the spins being reversed (fig. 54), the Zeeman energy corresponds mainly to this change of bilinear energy. At the high critical field, He2, the Zeeman energy has
•
• HC2
-,-10
I
I
2
3 4 5 TEMPERATURE (K)
I
TI
I
I
6
Fig. 53. Magnetic phase diagram for a magnetic field along a three-fold axis in TmGa 3 (from Morin et al. 1987d).
74
P. MORIN and D. SCHMITT
oooo0O0,°~° o°°° O
T=
1,5°4, " o •
o: .%
•
• •
A~A~
•
°"
°~'°t
0"£
•
"5K
Ao*
"d •
oOO oO4°,°° o • o Jo o ~Ae
rn ,~1.¸ v
oJ
•
.%" 2.o ,e •
o
8° o ~ oo
,0
2'o
3'0
20
H (kOe)
10
6'0
70
Fig. 54. Magnetization processes for a magnetic field along [111] in TmG% (from Morin et al. 1987d).
to compensate both the bilinear and quadrupolar energies as the AFQ arrangement is broken at this field. The existence of multi-axial spin structures in rare earth antiferromagnets with high symmetry appears to be a common consequence of AFQ interactions. In series other than the CsC1- and AuCu3-type ones, multi-axial structures also exist, e.g., as in rare earth hexaborides. In the AFQ phase of CeB6, the local anisotropy drives the spins to order in a multi-axial arrangement described by two propagation vectors [¼ 1 ½] and [1 _ ] ½] whereas the AFQ structure itself is described by [½ ½ ½] (fig. 12). The magnetic moments of the two cerium AFQ sublattices order along two mutually perpendicular directions (Effantin et al. 1985). The low-temperature magnetic structure of PrB 6 was initially described by a single [I ~ ½] propagation vector from powder neutron diffraction spectra (McCarthy et al. 1980). Recent experiments on a monocrystalline sample have shown that the structure is in fact double q, described, as in CeB6, by [¼ ¼ ½] and [ ~ - ¼ ½] and thus probably results from the effects of AFQ interactions (Burlet et al. 1988). Another complex situation is observed in the NaCl-type compound, HoP (Fischer et al. 1985). At Tc = 5.4 K it undergoes a second-order transition to a ferromagnetic state. At TF = 4.8 K there is a first-order transition to a flopside state with two Ho sublattices consisting of alternate ferromagnetic (111) planes. From one plane to the next one, the 4f moments change from one four-fold axis to another. Due to the antiferro- and ferromagnetic components, this lowtemperature structure is biaxial (fig. 55). The spontaneous and field-dependent properties were analyzed using the cubic model with two sublattices (Kim and Levy 1982). This is justified by the presence of six low-lying CEF levels nearly degenerate and well-separated from the excited ones. Bilinear and quadrupolar pair interactions were limited to nearest and next-nearest neighbours. The
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H INTERMETALLICS
75
lO-3Jn
10-
5-
10
5.OK I
I
I
I
~1 i
i
5
i
o
71 ""-------
3'0
eo
'
9'0
i
1~o
~'ec*~
Fig. 55. Temperature dependence of neutron diffraction pattern of a HoP powder sample. At 4.2 K the indexing corresponds to the magnetic unit cell with lattice constant a m = 2a. Insets indicate corresponding magnetic structures (from Fischer et al. 1985).
corresponding parameters were deduced from the magnetic excitations measured in the flopside phase by inelastic neutron scattering. Bilinear interactions are positive, in accordance with the ferromagnetic tendency, tetragonal quadrupolar ones are negative in accordance with the change of the z-axis from one plane to another. Thus the negative quadrupolar interactions stabilize the flopside structure. 3.6.2. Incommensurate magnetic structures The nature of bilinear interactions in rare earth intermetallic systems very often leads to modulated magnetic structures (Koehler 1972, Rossat-Mignod 1979). This corresponds to a maximum value of the Fourier transform of the bilinear interactions for an incommensurate wave vector. These structures are either
76
P. MORIN and D. SCHMITT
phase modulated or, when the magnetic moment is confined along a given direction by the anisotropy, amplitude modulated. When considering only bilinear interactions, other couplings being equal, the modulated structures can change when decreasing the temperature below the N6el transition in three ways: (a) For a non-magnetic ground-state (non-Kramers ions, Kondo systems) the modulated structure can remain stable down to 0 K as observed in HoNi0.sCu0. 5 (Gignoux et al. 1977) and in CeA12 (Barbara et al. 1980). (b) For a magnetic ground state (Kramers ions) entropy effects drive the structure to move progressively towards an antiphase structure with constant magnetic moments as in metallic Tm (Koehler et al. 1962). (c) A first-order transition to a commensurate structure (constant magnetic moments) occurs at a relatively high reduced temperature T]TN; e.g., T I T N = 0.87 in TmCu (Morin and Schmitt 1980), 0.82 in DyAg (Kaneko et al. 1987). In the presence of quadrupolar interactions, some modifications are expected. In particular for the first case (a), the nonmagnetic ground state can be quadrupolar (F3 doublet) and the same entropy effects exist for quadrupoles as for spins: consequently, the magnetic structure also has to change as in cases (b) or (c). For case (b), the tendency to a commensurate quadrupolar structure facilitates the squaring-up of a modulated spin arrangement. Unfortunately there is no clear evidence of these two situations up to now. In case (c) the temperature of the first-order commensurate transition Tt appears to be strongly dependent on the large quadrupolar interactions. In TmCu, Tt = 6.7K is close to the temperature To_= 5 K, at which it would ferroquadrupolarly order in the absence of bilinear interactions. Below Tt, both the magnetic and quadrupolar structures are collinear with tetragonal symmetry (Morin and Schmitt 1980). In DyAg, a modulated structure is also observed close to TN (Kaneko et al. 1987). However, the commensurate triple-q structure with spins along the three-fold axes is rapidly stabilized by the trigonal AFQ interactions through a first-order transition at Tt = 46.5 K. The influence of quadrupolar interactions on the commensurate-incommensurate transition appears to be a rather general feature and can be observed in other series such as RB 6 (Effantin 1985, McCarthy et al. 1980) and in RGa 3 (Morin et al. 1987a,d). From a purely magneto-elastic point of view, it is worth noting that a static strain wave is associated with the incommensurate spin wave. Its modulation is one half that of the spins because of the time-reversibility of the quadrupoles. It manifests itself through lattice satellite reflections at wave vectors ~'m~= 2~'n~, where r m is the magnetic propagation vector. They have been observed in metallic Cr (Tsunoda et al. 1974, Pynn et al. 1976), and in rare earth metals such as Ho (Keating 1969, Bohr et al. 1986) and Er (Gibbs et al. 1986). In the incommensurate helicoidal phase of these metals, 2%-satellites have been observed with X-ray diffraction and their intensity was related to the aspherical 4f charge density changing its orientation from plane to plane (fig. 56). In addition, % varies with temperature, the spin structure locks-in at commensurate values of % because it gains some magneto-elastic energy. These lock-ins are also observed in Dy
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS I
I
I
j/~
I
77
I
ERBIUM(OOQ) 2 Tm
47.5K ,;'/ .~".~..
20O o I00 C
49 K.,..-~
~[l~-
2~ ~o× p__. z o
_
,95K..~
505K
---
.... J
51.6K~ 52K .--
k__
k____
i
4/71 200 55K I00 I I I I ~] 0 2.55 2.56 2.57 2.58 2.59 (unitsof c*) m
Fig. 56. Temperature variation of the nuclear scattering at 2Tm in Er near the % = ~ lock-in transition (% is the magnetic propagation vector) (from Gibbs et al. 1986). ( G r e e n o u g h et al. 1981), they appear to be m o r e frequent in phase m o d u l a t e d structfires (existence of a quasi-isotropic plane) than in amplitude m o d u l a t e d structures (easy axis).
3.7. Effects on magnetic excitations T h e effects of quadrupolar interactions o n the e l e m e n t a r y excitations of m a g n e t i c systems w e r e first studied, by using effective spin m o d e l s , at the s a m e time as the
78
P. MORIN and D. SCHMITT
quadrupolar Hamiltonian itself. Cases with S = 1 (Chiu-Tsao et al. 1975, Sivardi~re 1975), S = 3 (Chiu-Tsao and Levy 1976), and S => 3 systems (Sivardibre 1976) were successively investigated. Two types of effects were considered in these calculations, namely: (i) the modification of the usual (spin waves) magnetic excitations by the quadrupolar couplings, and (ii) the occurrence of pure quadrupolar excitations corresponding to two-spin (AM = 2) deviations. The first effect consists simply of a variation of the size of the spin waves dispersion, at least in the ferromagnetic state. In the quadrupolar phase, however, an additional change of shape occurs, the dispersion curves becoming linear in q when q--~ 0. More exciting is the possibility of two-spin (AM = 2) excitations (quadrons) with the same type of dispersion as the AM = 1 magnetic excitations, but for different energies. Such quadrupolar excitations can be measured by inelastic neutron scattering experiments (Chiu-Tsao and Levy 1976, Levy and Trammell 1977). However, to our knowledge, they have never been observed experimentally. In more realistic calculations, the influence of the CEF must be taken into account by considering the whole level scheme, i.e., the actual CEF wave functions. Elementary excitations have then been obtained through a Green function (Sablik and Wang 1978, 1979) or a generalized susceptibility formalism (Morin et al. 1980c) with similar results. Attention was mainly focused on the influence of quadrupolar couplings on the magnetic (spin wave) excitations, which led to the same general conclusions as the effective spin models. In the magnetic phase, quadrupolar interactions modify the average energy of the dispersion curves, through the single-ion Hamiltonian, as well as their shape, through the Fourier transform of the two-ion quadrupolar couplings. The excitations are then dipolar and quadrupolar mixed excitations. The first effect has been widely used, as described in the literature, for adjusting the energy of the excited CEF levels with the average value of the dispersion curves (Furrer et al. 1977, Pierre et al. 1984). The second effect is difficult to observe; it has been invoked in NdSb in order to explain the shape of the experimentally observed excitation spectra (Sablik and Wang 1985). Surprisingly in the quadrupolar phase, the quadrupolar coupling which induces the static quadrupolar ordering does not produce any specific dispersion in addition to tha(arising from the bilinear coupling; it only modifies the overall amplitude of the dispersion curves. This has been clearly demonstrated in TmxLul_xZn, in which a 10% dilution of Tm by Lu suppresses the ferromagnetic ordering existing in TmZn and provides a pure quadrupolar phase below TQ = 5.8 K (Morin et al. 1981). In that phase, the total amplitude of the longitudinal excitation reaches 16 K, while the corresponding value is only 10K in the ferromagnetic phase of TmZn. The only expected effect in the quadrupolar phase is the modification of the transverse excitations by the second (trigonal) quadrupolar interaction. Unfortunately the latter is too weak in Tm0.9Lu0.1Zn to give rise to a noticeable effect.
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
79
3.8. Two-ion anisotropic magneto-elasticity The magneto-elastic modulation of the pair interactions has been neglected in all the preceding sections. This may be justified by analyzing Gd compounds in the absence of any single-ion terms. Usually two-ion magneto-elasticity is only associated with the bilinear interactions. Its modification occurs through either the isotropic exchange striction (see section 3.9) or symmetry lowering modes which introduce anisotropic interactions. The corresponding equilibrium strains have been given in section 2 [eqs. (19)-(24)]. In the CsCl-type compound GdZn, a spontaneous tetragonal strain has been observed below T c = 270 K (Morin et al. 1977, Rouchy et al. 1981). Its zeroKelvin value is e~ = -4.5 x 10 -4, a value one order of magnitude smaller than observed in other RZn (see section 4). The two-ion strain is expected to vary roughly as the De Gennes factor across a series and to take its maximum value for the Gd compound. Thus its contribution to the tetragonal strain rapidly decreases towards each end of the rare earth sequence; e.g., it is only half a percent of the observed value in TmZn. The e-strain has been estimated to be two orders of magnitude smaller than the y-strain. In the Laves phase compound GdAI2, the opposite situation is encountered: e~ is negligible and the e-strain reaches e ~= 1.5 x 10 -5 at 0 K (Du Tr6molet de Lacheisserie 1988). As for RZn, the spontaneous strain is one order of magnitude smaller in GdZn than in other compounds (section 4). The ratio reaches 40 between~strains in TbFe 2 and GdFe 2 polycrystalline samples (Abbundi and Clark 1978). In hexagonal RCo 5 also anisotropic strains are noticeable in the presence of a rare earth orbital moment, but weak in YCo s (Andreev et al. 1985a). In NaCl-type Gd compounds, a spontaneous trigonal strain is observed (Hulliger and Stucki 1978a). The e ~ values observed at low temperatures are 8 x 10 -4, 12 x 10 -4 and 17 x 10 -4 in GdAs, GdSb and GdBi, respectively. For the same trigonal symmetry they are -37, -28 and -33 x 10 -4 in TbAs, TbSb and TbBi (Hulliger and Stucki 1978b). The ratio is then not as large as in the preceding examples, in particular for antimonides and bismuthides. This noticeable two-ion anisotropic magneto-elasticity seems to be a general feature of NaCl-type compounds as observed experimentally and expected theoretically (section 3.2.4). Although the two-ion symmetry lowering modes are usually negligible in comparison with the one-ion modes, the latter example shows that it is necessary to study the Gd compound in a given series systematically.
3.9. lsotropic magneto-elasticity and pressure effects The e ~ volume strain may be obtained from a combination of parallel and perpendicular changes-of length Ix/3 e ~ = AlL+ 2h. from eq. (25)]. Equation (19) shows that one- and two-ion magneto-elastic terms coexist in a given compound. Only the two-ion term is present in Gd compounds. As they are expected to
80
P. MORIN and D. SCHMI'IT
decrease from the Gd value according to the De Gennes law, this allows to evaluate the one-ion contribution in other compounds. 3.9.1. Two-ion contribution In addition to the change of length method, the volume strain may be deduced from measurements of the dependence of the magnetic transition on hydrostatic pressure. For vanishing CEF effects, a MFA treatment proposes an equivalent expression for e ~ as a function of the reduced magnetization o- (Bean and Rodbell 1962): e~-
3 J 2 J+l
ATe 2 NkB -Aff ~r '
where o" is the normalized magnetization. Both methods usually give compatible determinations in Gd compounds. In GdAI2, e.g., the isotropic strain at low temperatures is found to reach - 1 . 7 5 x 10 -3 from change of length measurements (Du Tr6molet de Lacheisserie 1988) and -1.85 × 10 -3 from pressure data (Jaakkola and Hanninen 1980). In GdZn, at T = 0 K a value of +5 x 10 -4 is obtained (Morin et al. 1977, Hiraoka 1974). In GdAg, at T = 0 K a large value, e ~ = -3.8 × 10 -3 may be deduced from ATN/AP = 1.17 K/kbar (Yoshida et al. 1987). This is characteristic of a strong dependence of the magnetic properties on the actual electronic structure, which is close to instability (see section 3.10). 3.9.2. Single-ion contribution For rare earths other than Gd, the isotropic strain also has a single-ion term describing the bulk modulation of the CEF. As the 4th and 6th order CEF terms are of comparable magnitude in rare earth (R) intermetallics, the two magnetoelastic coefficients B(4 ) and B(6 ) have to be considered [eq. (19)]. By comparing results on GdZn with those on TbZn and DyZn, e ~ has been found to be negative in the latter two ( - 3 × 10 -4 and -13 × 10 -4, respectively). This indicates that the CEF part is opposite to that of the exchange and at least twice as large in magnitude (Morin et al. 1977). With regard to the single-ion contribution, thermal expansion appears as the best experimental probe. It may be discussed in the paramagnetic phase within a susceptibility formalism (Morin and Williamson 1984) as well as in terms of Gr/ineisen parameters (Ott and Lfithi 1977). Thus, thanks to thermal expansion results, the pressure dependence of the magnetic susceptibility has been explained in TmSb by a modification of the single-ion susceptibility rather than by a change of the bilinear interactions (Ott and L/ithi 1977). In another singlet ground state system, PrSb, inelastic neutron scattering under pressure (Vettier et al. 1977) and thermal expansion results (Ott 1977) have consistently shown the CEF spacing to decrease when the volume decreases.
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
81
3.9.3. Pressure effects Except for the previous examples of the magnetic susceptibility or the ordering temperature, the literature is rather scarce about pressure effects in rare earth intermetallics. Indeed the energies associated with pressure remain weak in comparison with the intrinsic couplings. Hydrostatic or uniaxial stress seems to be able to have dramatic effects only in the case of magnetic systems which are subcritical. For instance, hydrostatic pressure has been shown to drive an antiferromagnetic ordering in PrSb through a softening of the X point magnetic exciton, the zero-pressure value of which is determined by the intrinsic anisotropic bilinear interactions (McWhan et al. 1979). More relevant to the present review is the case of another single ground state system, Pr (Jensen 1979). The bilinear interactions between rare earths at hexagonal sites are just below the critical value for an antiferromagnetic ordering. This makes the spin system very sensitive to various influences, in particular to magneto-elastic effects with e-symmetry. They have been show to influence deeply the magnetic excitons ( H o u m a n n et al. 1979). As the B ~ magneto-elastic coefficient was deduced from ultrasonic measurements (Palmer and Jensen 1978), it was possible to calculate the e" critical value that is needed to modify the level spacing enough so as to drive the spin system to be critical; an uniaxial stress of 770 bar applied along the a-axis of the hexagonal cell was found to be necessary. Experimentally antiferromagnetic ordering was observed below 7.5 K just under about 800 bar (see fig. 57), thereby verifying the above prediction.
1.0 -
/---r-X
N
"1" l->-
,,
n.I..iJ Z W
/
~,.¢/ t2<
,'
v
......
S-TO
.
0.5 ['-- ~.a~""',v" ~ l r . ~ '''r
0
0 F
0.1
0.2
4/ ,,p"
,'/
,'/ ,,,,//
EL/
0.3
0.4
WAVEVECTOR
0.5 M
Fig. 57. Dispersion relations for magnetic excitons propagating on the hexagonal sites in Pr, at 5.2 K, at room pressure (circles) and under an uniaxial pressure of 800 bar (squares). TO and LO branches correspond to transverse and longitudinal optical modes; the TA and LA transverse and longitudinal acoustic modes were not examined under stress. An asterisk (,) denotes the magnetic satellite position in the ordered phase (from McEwen et al. 1978).
82
P. MORIN and D. SCHMITT
3.10. Magneto-elasticity in the presence of lattice instability In all of the preceding sections, magneto-elasticity was acting on an otherwise stable lattice: As for the lattice parameter, the background elastic constants were assumed to vary weakly across a series as well as to be weakly dependent on the temperature. This situation is the norm, but there are some exceptions. For instance, the CsCl-type structure appears less stable than other crystallographic arrangements such as the MgCu 2- or NaCl-type ones. Indeed some series are incomplete (Iandelli and Palenzona 1979) with compounds exhibiting martensitic transformations (Gefen and Rosen 1981, Al6onard and Morin 1984, Sekizawa et al. 1981). For example, RAg compounds are cubic at all temperatures, and an incipient lattice instability is easily evidenced by hydrostatic pressures at temperatures which drastically depend on the pressure (Kurisu 1987). Chemical pressure obtained by substituting Ag by In produces analogous effects (Ihrig et al. 1975, Fujiwara et al. 1983). The martensitic transformation has been shown to occur mainly through a phonon softening at the M point of the Brillouin zone in LaAgxIn~_ x (Knorr et al. 1980, Abu-Aljarayesh et al. 1986b) rather than through a band Jahn-Teller coupling developing at the F point. Through an anharmonic coupling with the F-point acoustic phonon, this M-point phonon softening drives a partial softening of the lattice elastic constant. A full understanding of this lattice instability requires, in the presence of a strong electron-phonon coupling, dynamical models considering the full dielectric response function of the electronic system. The starting point must be the determination of the actual band structure as has been done for transition metals (Dacorogna et al. 1982). The coexistence of magneto-elasticity and lattice instability makes it difficult to determine the predominant coupling from the temperature dependence of the elastic constants. For instance, in light rare earth-silver compounds, the tempera-
1,4
=
=
I
~" E
(PrAg)o ...................
- ................................................
~c, 1 . 0 ~
.....
* ............
L
o
v
a ,..~ -6 -
---~~--CeAg v/'
,~''~ I
I
I
I
lOO 200 TEMPERATURE (K)
I
300
Fig. 58. Temperature variation of ½Cv= ½(Cll - C1~) for LaAg, CeAg and PrAg. (PrAg)o is the lattice behaviour after correction for the magneto-elastic contribution (from Giraud et al. 1983).
Q U A D R U P O E A R EFFECTS IN RARE EARTH INTERMETALLICS
83
~.5
t t
/,--"
\:~ci,_~
R=Tm Dy
.9
L
Pr Ce La
411
1.2
ionic radius {~,)
Fig. 59. Variation of the three normal elastic constants at room temperature as a function of the ionic radius in the RAg series, normalized to their value in TmAg (from Morin 1988).
ture variation of Cll - C12 in the LaAg compounds is obviously anomalous and cannot be used directly as background in a magnetic compound (fig. 58). Across the series, the instability tendency progressively vanishes when proceeding to heavier rare earth compounds (fig. 59). As discussed from a theoretical point of view for the orthochromites (Kataoka and Kanamori 1972), the anharmonicity involved with the lattice instability manifests itself in the occurrence of the quadrupolar ordering. In CeAg, where the quadrupolar interactions have been determined by means of parastriction and third-order magnetic susceptibility, both the temperature and the discontinuity of the order parameter are increased, e.g., the quadrupolar ordering would occur around 10 K with only quadrupolar interactions, but this temperature is increased to 15.85 K due to the presence of the incipient lattice instability (Morin 1988). 4. Magneto-elastic and pair interaction coefficients in rare earth intermetallic series
4.1. CsCl-type structure compounds CsCl-type rare earth intermetallics RX have been extensively studied due to the availability of their single crystals. Their very simple structure allows a deep insight in the 4f magnetism; analyses are possible according to the alloyed non-magnetic metal (X = Mg, Zn, Cd, Cu, A g , . . . ) . They have revealed large quadrupolar interactions which usually prevent any analysis which would only consider CEF and bilinear interactions of the magnetic data. In particular large spontaneous strains exist with tetragonal symmetry; they reach record values ( c / a - 1 = 1.4% in PrZn and -0.8% in TmZn). Important elastic constant softenings are observed in the paramagnetic range. FQ orderings were discovered in TmZn and TmCd (section 3). The following subsections outline the numerous effects of quadrupolar interactions on the magnetic properties in each series. Quadrupolar interactions are largest in tetragonal symmetry. They are of a ferroquadrupolar type. In trigonal symmetry, they are negative and very often stabilize multi-axial spin structures when the three-fold axis is the direction of easy magnetization. Systematic studies of the TmX have shown how very relevant the (spin and) quadrupolar Hamiltonian is within the MFA. In particular the
84
E MORIN and D. SCHMITT
absence (or the weak value) of fluctuations leads to determinations of G r and G" from third-order magnetic susceptibility measurements in close agreement with those obtained from magneto-elasticity data. However, there are particular situations where the quadrupolar coupling may exist with additional terms (as in the case of the lattice instability for CeAg).
4.1.1. RZn series Light rare earth compounds are antiferromagnetic with a (½,0, 0) propagation vector. From SmZn to TmZn, a ferromagnetic behaviour is observed (table 5). The CEF parameters A4(r4), A6(r 6) are found to be negative by inelastic neutron spectroscopy, without exhibiting an erratic variation (Giraud et al. 1986). When the four-fold axis is the easy magnetization direction, a large spontaneous tetragonal strain is observed. The various magneto-elastic coefficients were determined by different experimental techniques. Note that when the temperature range investigated by parastriction is reduced (high Tc values), only a relationship between G ~ and B ~ is obtained. Some differences between B ~ values deduced from parastriction and spontaneous strain measurements may be noticed, in particular in TbZn, DyZn and HoZn. They are partly explained by an overestimation of the two-ion contribution to the spontaneous strain. Across the series, the coefficient B ~, normalized by the Stevens coefficient exhibits an average value of about -2050 K with the ErZn value slightly remote (see section 5.1). In trigonal symmetry, reliable values of B ~ are achieved in TbZn and HoZn from the spontaneous trigonal strain. In other compounds, only the parastriction may be efficient. However, as the tetragonal symmetry lowering mode is highly favoured in CsCl-type compounds, a very accurate orientation of the magnetic field along [111] is necessary in order to avoid the presence of any tetragonal contribution to the change of length measured. Here again the parastriction leads only to a relation between values of B e and G ~. B e values obtained with both techniques are in good agreement in TbZn and HoZn. The ratio B ~/aj obviously varies across the heavy rare earth compounds; in TbZn, it even Changes sign. The magneto-elastic contribution to G ~ is less than 1 mK for the trigonal symmetry, but reaches 5.2, 9 and 120 mK for tetragonal symmetry in TmZn, PrZn and CeZn, respectively. However, in RZn, the quadrupolar interactions originate mainly from the pair coupling, as proved by the ratio K~/G~jE of 1.6 and 4 which is found in ErZn and TmZn, respectively. They are systematically positive in tetragonal symmetry, negative in trigonal symmetry. In TmZn, they dominate the bilinear interactions, and induce a quadrupolar ordering in the paramagnetic phase. Depending on their sign, they may stabilize either collinear antiferromagnetic structures as in CeZn and PrZn or multi-axial ones as in NdZn. In other compounds, none of the magnetic properties can be described without the quadrupolar interactions as shown for the four-fold axis of easy magnetization in DyZn, the magnetization processes in HoZn and ErZn, the actual temperature for the change of easy axis in NdZn and HoZn and the first-order transition in CeZn (see section 3).
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
85
4.1.2. RCd series This series has been less extensively studied than the RZn one, due to the large neutron absorption cross section of cadmium, which in particular forbids a direct determination of the CEF level scheme by neutron spectroscopy. However, the properties appear to be very close to those of the RZn series from a magnetic point of view. One of its peculiarities rests in the properties of the lattice: La, Ce and Pr compounds exhibit a well-defined martensitic-type transition, which is revealed by electrical resistivity or susceptibility measurements, at 61, 107 and 125 K, respectively (A16onard and Morin 1984). At 216 K CeCd undergoes an additional transition which is as well-defined as the previous one (Fujii et al. 1985). Compounds from Nd to Tm are ferromagnetic, CeCd and PrCd being antiferromagnetic (A16onard and Morin 1985). CeCd was also found to be ferromagnetic (Fujii et al. 1985). The easy magnetization axes and the magnetic moments observed in ferromagnets are closely reminiscent of the corresponding results in RZn. For instance, all of the properties studied in TmCd, in particular the quadrupolar properties, can be described with a CEF level scheme very close to that in TmZn. Assuming the CEF parameters as constant, quadrupolar coefficients may be obtained through the full description of the magnetization properties. This is true in particular for the four-fold easy direction in DyCd as well as in DyZn, and for the temperature of change of easy direction in NdCd as well as in NdZn. On the contrary, observing [111] axes as easy direction in CeCd and PrCd clearly reveals a deep modification of all the one- and two-ion terms driven by the lattice instability. In RCd ferromagnets, quadrupolar interactions are present having large analogies with the RZn case as proved by the comparison for a given rare earth (tables 5 and 6). However, from the thorough study of TmCd, both the pair-interaction coupling and the magneto-elasticity are smaller than in TmZn in tetragonal symmetry, the ratio Kr/G~E, however, is twice as large. This is emphasized by the T o value in both Tm compounds. In trigonal symmetry, the coefficients G ~ are negative as in RZn. 4.1.3. RAg series This series exhibits some specific behaviours. Bilinear interactions favour incommensurate magnetic structures. One notes, however, a tendency to ferromagnetism at the beginning of the series through a ferromagnetic component in PrAg and a ferromagnetic moment in CeAg. The observed properties strongly depend on the preparation. In an unannealed sample, the magnetic structure is systematically commensurate with the lattice periodicity. Removing the metallurgical stresses by annealing the sample restores the intrinsic nature of the bilinear interactions as it was demonstrated in HoAg, ErAg and TmAg (Nereson 1973a,b, Morin and Schmitt 1982a). Another important peculiarity rests in the incipient tendency to lattice instability, although all the RAg are stable within the CsCl-type structure in the absence of external stress (section 3.10). Under hydrostatic or chemical pressure, the instability towards a tetragonal-symmetry based structure (Maetz et al. 1980) is
"0 0
%-..~ N
+1
¢)
I ,io
a
¢) ,io
d
Z
I
v e,
+1 ¢) o
8 Lxl
,9 v
$
Z
2
~v
¢) "0
.= ¢) o.~
~o
86
Z Qo ÷1
~
~
÷1 ~
÷1
Jr
I/
÷
A
~ . ~
87
"0
Z
e~
0
I
¢,)
z
~
,-.-i
0
v ~
~.~A [,.
t¢3
cq
v ~ ~ A
05
©
~
o~ ~
o
0 n~
N
6
88
c~
0
0
¢~ T-..~ ,~.,, " c l r O -
.~, o
I
II
I
÷1
~
o'~
,-~ 0
o
~Av
II ~
o
I
II
0 "0 o"
~'~ 9 A V
"0
C~
oo
•
E~ ~
oo
# _ ~ - ~-~
.
Z ¸
E
0
0
ol
~ . ~~ .. o ~
~ A
~
~~ °o° °~~ ~ ° ~ o ~ ~o o
~-
o~
o
..=
o
==~o O
o
8~
O
o= ,o
90
P. MORIN and D. SCHMITF
revealed by a martensitic transition at a strongly pressure-dependent temperature (Kurisu 1987). Its driving mechanism consists of a softening of the TA phonon at the M point on the Brillouin zone (Knorr et al. 1980) and induces anharmonic contributions to the elastic constant. In the absence of an external stress, the incipient lattice instability takes the opportunity of the quadrupolar ordering in CeAg to manifest itself through both an increase of the quadrupolar ordering temperature (see section 3.1.1) and a spontaneous tetragonal strain ( c / a - 1 = 1.9%) which is larger than expected from the one-ion magneto-elasticity only. In RAg, the tendency to instability is particularly clear in LaAg, CeAg and, to a smaller extent, in PrAg (see section 3.10). The tetragonal magneto-elastic parameters, normalized by the second-order Stevens coefficient, are of the same order of magnitude as in RZn. Their average value is around -2000 K/at., the CeAg value being somewhat smaller (table 7). The trigonal symmetry coefficient observed in PrAg and TmAg leads to positive BE/aj ratios (4400 and 2200 K/at., respectively). As in RZn and RCd, the pair interactions are positive in tetragonal symmetry, negative in trigonal symmetry. For example, in TmAg, they seem to be smaller than in RZn. Their effects are, however, numerous on the magnetic properties. The commensurate-incommensurate transition at 9.5 K in ErAg may be influenced by quadrupolar interactions as discussed for DyAg (section 3.6.2). The N6el temperature in TbAg is the highest one known with a first-order character, which agrees with the G v value determined in the paramagnetic phase. In PrAg, a neutron diffraction experiment with polarization analysis on a single crystal has, at least partly, removed the ambiguities among the possible structures of fig. 52 and has shown that a double-q structure (structure b or c) is very likely in agreement with the existence of rather negative quadrupolar interactions (Givord et al. 1985). 4.1.4. RCu series CsCl-type structure compounds range from GdCu to LuCu; equiatomic alloys with light rare earth crystallize within the FeB structure (Buschow 1980). Lattice instability develops itself in boundary compounds: GdCu undergoes a martensitictype transition with a very large hysteresis (Gefen and Rosen 1981) (table 8). In TbCu the martensitic transition seems to be triggered by the antiferromagnetic ordering (Pierre and Hennion 1982). All the other heavy rare earth compounds are stable down to 0 K. Antiferromagnetic interactions are observed across the series, they are characterized by a (1, ½,0) propagation vector. However, ( ½ - r, ½,0) modulated structures have been observed in TmCu and ErCu close to the N6el temperature in a way reminiscent of RAg compounds. Magneto-elastic coefficients are available for TmCu and DyCu in tetragonal symmetry. In TbCu, a B ~ value may be achieved from the spontaneous strain by assuming a maximum value for ( O °) due to the large exchange field at 4.2K. The resulting ratios Br/aj are -2300, -2450 and -2800 K/at. in Tb-, Dy- and TmCu, respectively: the same sign and order of magnitude are observed as in other CsCl-type series. In trigonal symmetry, B ~ is found to be positive in TmCu as in other TmX.
QUADRUPOLAR
E F F E C T S IN R A R E E A R T H I N T E R M E T A L L I C S
91
"O
~Av
8
PI
o
o
~AV~ O
o
u~ II ~]~
0 ..~ N ..~ o
~4 tt~
8~
,,6
~"
V A A
O
,,6 II A
m M'~
~.~
V
w
I
~e +
E
E ~
rq ,,6
,--Z e., o
Z
~D
92
P. MORIN and D. SCHMITT
Quadrupolar pair interactions are also present. They are positive in tetragonal symmetry although they are a little weaker than in RZn; they dominate the magneto-elastic coupling. The total quadrupolar coefficient G r would drive a quadrupolar ordering at only T o = 5 K in TmCu. Its value in ErCu, deduced from the magnetic excitation spectrum, is lower than in ErZn. The tendency to negative quadrupolar pair interactions in trigonal symmetry is also established in RCu. As discussed in section 3.6.1, they determine the multi-axial structure of DyCu, the CEF being responsible of the three-fold direction of easy magnetization as in DyAg and TmGa 3 (section 3.6). 4.1.5. R M g series
Light rare earth compounds are antiferromagnets. In heavy rare earth compounds, the spin structure is either canted, surprisingly also in GdMg (Kim and Levy 1982), or more complex, because they mix ferro- and antiferromagnetic tendencies (A16onard et al. 1976). From a quadrupolar point of view, spontaneous strains are observed in CeMg and PrMg (table 9). From the extensive study of TmMg (Giraud et al. 1986), it was found that the magneto-elastic and pair quadrupolar interactions are weak in heavy RMg compounds (section 5.1). Then an analogy appears between quadrupolar and antiferromagnetic interactions: they are strong and well-established at the beginning of the series and they rapidly decrease in heavy rare earth compounds.
TABLE 9 Quadrupolar coefficients in the RMg series. Also indicated are the N6el temperature, the antiferromagnetic propagation vector Q, and the spontaneous strain at low temperature. RMg compound
Tr~ Q Easy axis c/a - 1 (%) B" (K) G ~ (mK) B ~ (K) a " (mK)
CeMg
PrMg
TmMg [5]
19.5 [11 1st order (~,0,0) [001] [2] 1.3 -+ 0.2 [2] 98 [6] (50, 150) [11
45 [3]
<1.3
(~,0,0) [001] [4] 1.4 -+ 0.2 [4] 30 [61
-5.4 (0, 3) (14, 21) (-100,-200)
References: [1] Pierre et al. (1984), from magnetic excitation studies. [2] Schmitt et al. (1978). [3] A16onard et al. (1976). [4] Morin et al. (1978a). [5] Giraud et al. (1986) from measurements of parastriction, elastic constants and X~ )[6] From c / a - 1 data, using (O °) ='8 (CeMg) and 28 (PrMg).
Q U A D R U P O L A R EFFECTS IN RARE EARTH INTERMETALLICS
93
4.2. NaCl-type structure compounds Rare earth pnictides (rare earth compounds with group V-A elements) crystallize within the cubic NaCl-type structure. They have been extensively studied from a magnetic point of view (Hulliger 1978). Their semi-metallic behaviour leads to RKKY-type bilinear interactions driving rather low magnetic transition temperatures (T < 60 K). The low-temperature phase is antiferromagnetic of type-I or -II with some exceptions for Ce compounds. The small size of nitrogen ions is probably the origin of ferromagnetism in some nitrides. Associated with the semi-metallic behaviour, the strength of the magnetic interactions strongly depends on the stoichiometric conditions as may be deduced from the spread in the T s values listed in tables 10 and 11. The CEF parameters are very similar for compounds of a given rare earth ion and the dependence across the rare earth series follows, more or less fortuitously, point-charge estimates. However, a clear exception is the case of Ce compounds (Birgeneau et al. 1973). One of the consequences of the positive A4(r 4) parameter and of the weak A6(r 6) parameters, is the great number of nonmagnetic singlet ground states for Pr, Tb and Tm ions; over- (sub-) critical conditions are usually observed in Tb (Pr and Tm) compounds. The magnetoelastic coupling provides clear evidence for the actual magnetic structure, through the tetragonal or trigonal spontaneous strain. The number of multi-axial structures appears to be very small contrarily to the situation in CsCl-type compounds. Numerous compounds order at a first-order transition, in particular those with Tb, Dy and Er ions. In connection with the low-temperature range, this is explained by positive values for the third-order susceptibility driven by quadrupolar interactions in both tetragonal symmetry (Levy 1973) and trigonal symmetry (Koetzler 1984). In tetragonal symmetry, the magneto-elastic quadrupolar term has been observed to be dominant (Mullen et al. 1974, L6vy 1969). The B ~ values have signs opposite to those in the CsCl-type series and they roughly follow the aj variation with a magnitude slightly larger than in the CsCl-type structure (Liithi 1980b). This seems also to be valid for the coefficient in trigonal symmetry. As discussed in section 3.2.4, the anisotropic character of the bilinear interactions as well as the existence of an associated two-ion magneto-elastic coupling has been shown theoretically and experimentally. The occurrence of a sizeable trigonal spontaneous strain in Gd compounds, in particular in GdBi (HuUiger and Stucki 1978a), complicates the analysis of the magnetic properties, such as, e.g., the magnetic transition. Anisotropic bilinear interactions seem to have a dominant role in Ce pnictides due to hybridization effects (Cooper 1982, Takahashi et al. 1982). According to the De Gennes law, two-ion magneto-elastic effects are expected to be noticeable in the middle of the rare earth sequence, e.g., in Tb and Dy compounds. In DySb below TN, the actual direction for the magnetic moment is tilted from the c-axis of the tetragonal cell; the resulting monoclinic cell underlines the presence of the trigonal two-ion magneto-elasticity (Felcher et al. 1973). Above TN, in tetragonal symmetry, the parastriction also reveals the competition between one- and two-ion magneto-elastic terms (Al6onard et al.
~im~ ~ I I ~
÷l
~
ir~
I
0
Z ~
~
~
÷
~
or ~I~
÷~
~
~J~
•
~0
~
F~
z
p~ t~
p~
c~ Tm~
p~ Z
94
~J
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
95
1984b). As for Ho compounds, HoBi exhibits a positive tetragonal strain (Hulliger et al. 1983) which is opposite to the negative one expected on the basis of a dominant one-ion contribution as effectively observed in HoAs; this may be also a signature of this competition.
4.3. AuCu3-type structure compounds Numerous RX 3 intermetallics crystallize with the cubic AuCu3-type structure, in particular for X = Pb, In, T1, Pd, Sn and Ga. Only a small number of studies has been devoted to magneto-elastic and quadrupolar interactions. The non-magnetic transition observed at 0.37 K in PrPb 3 has been analyzed to be driven by an AFQ ordering (see section 3.1.2). This behaviour agrees with a study of TmGa 3 which shows that AFQ interactions drive quadrupoles to order slightly above TN and stabilize the multi-axial spin structure below T N (sections 3.1.2 and 3.6.1). A quadrupolar ordering was also claimed to occur in the paramagnetic phase of Stain 3 and SmSn3; a noticeable softening of the C44elastic constant is observed at high temperatures (Kasaya et al. 1985). The tendency to negative quadrupolar interactions seems to be general in AuCu3-type compounds and responsible for the multi-axial magnetic structures observed. From a purely magneto-elastic point of view, PrPb3, TmGa 3 and CePb 3 (Nikl et al. 1987, Morin et al. 1987b) are the only AuCu3-type compounds studied. The spontaneous strain is null in TmGa3, as the magnetic structure is cubic. It is unobserved in CePb3, owing to the low N6el temperature (TN = 1.1 K) and the non-quadrupolar F7 ground state, and in PrPb3, due to the low value of TQ and the AFQ structure. The magneto-elastic coefficients and their ratios to aj are given in table 12. In both symmetries, the ratios B"/aj are positive; a good coherency is achieved for the tetragonal mode with an average magnitude which is coherent with values in preceding series. Some scattering is observed for the trigonal mode; however, values for PrPb 3 and TmGa 3 are noticeably stronger than in CsC1- and NaCl-type compounds.
4.4. Cubic Laves phase compounds 4.4.1. RAI 2 compounds The RAI 2 compounds have been thoroughly studied for many years with regard to, first, the crystal field and the magnetic properties and, second, the magnetoelastic coupling. All the available experimental results are reported in table 13. Most of them are based on magnetostriction measurements in the ferromagnetic range. That explains why some values seem out of range, as e.g., B" for PrA12 or B r for TbAI2, because they have been obtained from measurements along an axis of hard magnetization in compounds with a large anisotropy. The high value of B ~ for CeA12 has been related to a coupling with optical phonons at the F-point (Liithi and Lingner 1979). All the other experimental values are rather consistent with each other, in particular in NdA12 where the B r value has been obtained by elastic constants, X-ray and magnetostriction measurements. Recently, some
©
Im i-
~4 0
0
~ "ma
o~ ~ ~ <
m<~ 0
0
~o I A
~ < ~
~:~
~ o o
~
~
. ~
x
~
°
~
oo i'~ oo
I
•.~ ~
.~
©
x
z~
Z
96
~
~
x
c~ I 0
oo
oo
E~ I
I
o~°g
~
0
97
98
P. MORIN and D. SCHMITF TABLE 12 Magnetic and quadrupolar ordering temperatures and magnetoelastic coefficients from elastic constant, parastriction and g~ ) experiments in AuCu3-type compounds. AuCu3-type compound CePb3 [1] TN (K) To (K) B r (K) B~/oq (K) K" (inK) G"MZ(mK) a" (inK) B" (K) B~/a~ (K) K" (mK) GME (mK) G ~ (mK)
PrPb3 [2]
TmGa3 [31
0.35 [5] -41 1950 -13 8 -5 -190 [11 9050 -200 7 -200
4.26 [6] 4.29 17 1700 (8,14) 0.5 (8,14) 71 7000 20 10 30
1.1 [4] -80 1400
-190 3350
References: [1] Morin et al. (1987b). [2] Morin et al. (1982). [3] Morin et al. (1987c). [4] Lin et al. (1985). [5] Bucher et al. (1972a). [6] Czopnik et al. (1985). quadrupolar p a r a m e t e r s G v have been obtained by the torque method, which allows to extract the third-order paramagnetic susceptibility x- e(3) in a less direct way than from magnetization m e a s u r e m e n t s (Del Moral et al. 1987). As for the sign of magneto-elastic coefficients, they appear to follow well that of the Stevens factor aj. The ratios BY/aj and BE/aj range around 1000 K and - 3 0 0 0 K, respectively, with a few exceptions (see section 5.1). F r o m values of G r and G rME, the existence of two-ion quadrupolar coupling is suggested. This is in disagreement with elastic constants fits, which did not require considering this coupling.
4.4.2. R M 2 compounds (M = Mg, Ni, Co) For the R M 2 cubic Laves phase compounds where M is not A1 (section 4.4.1) or Fe (section 4.4.3), quadrupolar data are rather scarce. The PrM 2 compounds (M = Pt, Rh, Ru, It) have been investigated by magnetization, neutron diffraction, and specific heat m e a s u r e m e n t s (Greidanus et al. 1983a,b). A systematic overestimation of the ferromagnetic m o m e n t by a calculation including C E F and bilinear exchange interactions, with regard to the observed spontaneous magnetization, was explained by the presence of negative quadrupolar interactions. The same is true in PrMg2, where a quantitative analysis of the t e m p e r a t u r e variation of the ferromagnetic m o m e n t provided the quadrupolar p a r a m e t e r G r = - 1 3 . 3 m K (see table 14). F u r t h e r m o r e , this negative quadrupolar coupling leads
.i
+I
+I
e~
=
Z c~
~
i ~
I
¢J
8
c~c~ I i
~
~°~'~
Z ~
~
"~.~o~ °°
~ ~ ~oo 8
< •
e.
o Z
I
I
I
÷I
S
~
~4~
0
+I I
4~ ~ +I
~:~o~
4
0
i ~°
+I
.o e.
-g X
X
~
99
~
100
P. M O R I N and D. SCHMITT T A B L E 14 Quadrupolar coefficients in some R M 2 compounds. R M 2 compound PrMg 2 [1]
ErNi 2 [2]
B ~ (K) B ~ (K) G ~' (inK) G ~ (mK)
TbCo 2 [3]
HoCo 2 [4]
~0 79 -13.3
1 -22
-2.7 -11.7
References: [1] Loidl et al. (1981). [2] Gignoux and Givord (1983), from magnetization measurements. [3] Gignoux et al. (1979), from X-ray measurements. [4] Castets et al. (1982), from magnetic excitation studies.
to the disappearance of the first-order transition which is expected without it (see section 3.2.2). Some magneto-elastic results are also available in the RNi 2 series, where nickel is not magnetic. The compounds order ferromagnetically at temperatures roughly three times smaller than the corresponding temperatures in the RAI 2 series. At low temperatures, the directions of easy magnetization are the same in RNi 2 and RAI 2 compounds (R = Tb, Dy, Ho); the spontaneous distortions are comparable in both series, except for HoNi 2 where the strain is noticeably larger than in HoA12 (e ~ = -1.1 x 10 -3 in TbNi2, c/a - 1 = - 2 x 10 - 3 and -1.5 x 10 -3 in DyNi 2 and HoNi2, respectively) (Markosyan 1980, 1981). The behaviour of ErNi 2 differs considerably from that of ERA12, since its direction of easy magnetization is [100] instead of [111]. This has been well accounted for by introducing the parameters G ~ and G ~ (see table 14). In the RCo 2 series, cobalt is magnetic and contributes mainly to the volume magnetostriction. A positive trigonal magneto-elastic coefficient B ~ associated with the rare earth has been determined in TbCo 2 (see table 14). This leads to a positive trigonal magnetostriction e ~= 3.1 x 10 .3 at 4.2K which is opposite in sign to that measured in TbNi 2 and TbAI 2. This feature has been related to environmental effects in this type of cubic structure (Gignoux et al. 1979). In HoCo 2, a consistent description of the magnetic excitations as well as the spin rotation temperature requires introducing negative quadrupolar terms in the Hamiltonian (table 14), indicating the probable existence of a contribution of two-ion origin.
4.4.3.
RFe 2
compounds
The magneto-elastic properties of cubic ferromagnetic Laves phase compounds RFe 2 are relevant for the classical analysis, owing to their high magnetic transition temperature (around 700 K in TbFe2). Indeed, the Fe magnetization induces an exchange field large enough to purify fully the 4f magnetic moment towards its maximum value at 0 K. Their magnetostriction has been studied, in particular at
Q U A D R U P O L A R EFFECTS IN R A R E EARTH INTERMETALLICS
101
room temperature, for industrial applications. The large set of available data has been collected in the first volume of this Handbook by Clark (1980). Here, only characteristic features and analyses are considered again. Comparing the strength of the tetragonal and trigonal symmetry-lowering modes leads to a conclusion opposite to that in CsCl-compounds. Indeed, in RF% the trigonal mode is many times larger than the tetragonal mode (fig. 60). TbFe2, the basic compound for applications, is the paradigm of the series. The temperature variation of the Aln trigonal spontaneous strain has been described using eq. (107). As pedagogically explained by Clark (1980), an important step is to determine the 4f magnetization at any temperature by means of neutron diffraction or magnetization measurements. In this latter case, the Fe magnetization (about 1.64/~) is deduced from M6ssbauer spectroscopic results. Noticeable uncertainties may be present at high temperatures; however, the saturation of the 4f moment observed at low temperatures allows one to derive reliable values of B ~ through the relation Be=
3v~ o%,,(o)
J(2J-
1)
5000
4000
-
'
'
'
'
I
.
.
.
.
~ j I ~ Tb Fez
3000
I
'
^
'
'
'
I
(£l(m))
-
~'111
(a)
2000
40"
7
0 -40 -80 -120 -160 -200
L
100
i
i
]
200
n
i
u
I
300
T (*K) Fig. 60. Temperature dependence of the spontaneous strain in TbF% (trigonal symmetry) and in DyF% (tetragonal symmetry). The An1 variation is classically described according to the Tb magnetic moment. For the DyF% data, the full line is only a guide for the eye (from Clark et al. 1977).
102
P. M O R I N and D. SCHMITI" T A B L E 15 Curie temperature, easy magnetization axis, trigonal spontaneous strain at 300 and OK, trigonal magneto-elastic coefficient B e and its ratio to the Stevens coefficient c~j in RFe 2 compounds [data are from tables 5 and 6 of Clark (1980)]. RFe z compound
T c (K) Easy axis hal 1 (300 K) × 1 0 6 )tll 1 (0 K) x 106 B" (K) B'/a: (K)
TbFe 2
DyFe 2
HoFez
ErFe 2
TmFe 2
SmFe 2
697 [111] 2460 4400 94 -9400
635 [001] 1260 3320 a 44 -7070
606 [001] 200 770 a 9 -4100
590 ' [1111 -300 -1760" -24 -9300
560 [111] -210 -3500 -75 -7500
676 -2100
Note: "estimated from the room-temperature value.
Across the series, B" is observed to vary as the second-order Stevens coefficient a: in sign and, with some erratic fluctuations as in other series, also in order of magnitude (table 15). The same technique, based on eq. (107), fails to describe the tetragonal spontaneous strain in DyFe 2 (fig. 60) and HoFe 2 (Abbundi et al. 1979). Indeed, the change of sign of hl00(T) in DyFe 2 and the general temperature variation in HoFe 2 are both unexpected within this model. Whereas the hi00 sign at 4.2 K follows the sign of aj in both compounds, its absolute value is about ten times smaller in DyFe 2 than in HoFe 2, in full disagreement with the ratio of the aj's. Cullen and Clark (1977) have developed an atomic model that shows that internal distortions drive the large external trigonal strain mode whereas the hi00 strain may be sorted out because of the high tetrahedral symmetry of the rare earth site.
4.5. Hexagonal CaCus-type structure compounds Among the series of hexagonal 3d-4f intermetallic compounds, only few RNi 5 and RCo 5 compounds have been investigated from a magneto-elastic point of view; nickel is not magnetic in the first series, but cobalt is in the second one.
4.5.1. RNi 5 compounds Whereas the RNi 5 compounds have been studied for many years with regard to their magnetic properties and CEF effects, only few results are available about their magneto-elastic properties. A first study concerned the thermal expansion of PrNi 5 (L/ithi and Ott 1980) which evidenced crystal field effects, and showed the existence of the two a-mode magneto-elastic coefficients. However, the latter were not determined because the elastic constants were not known at that time. Using2the values of C" from Barthem et al. (1988), we are able to calculate B ~1 and B from these early measurements (see table 16). These values have the same sign, but are 2-3 times smaller than more recent values determined by magnetostriction measurements. In addition, the latter provide an estimation of
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
103
TABLE 16 Quadrupolar coefficientsin RNi5 compounds. RNis compounds B'~1 (K) B ~z (K) B ~ (K) G ~ (inK) G ~ (mK)
PrNi5 [11
PrNi5 [21
TmNi5 [3]
-7.1 28.6
-22 68.6 <15 -10 -20
(-500,-100) (-300,-150) (-1500,-1000)
References: [1] L/ithi and Ott (1980), from thermal expansion measurements. [2] Barthem et al. (1988), from magnetostriction measurements. [3] Barthem et al. (1989), from magnetostriction measurements. the total quadrupolar parameters G ~ and GL By contrast, Barthem et al. (1988) deduce antiferroquadrupolar (negative) interactions between Pr ions in PrNi 5 (K ~ - - 1 5 mK, K ~ - - 2 0 i n K ) . Note that a magneto-elastic coupling was also deduced from inelastic neutron scattering in PrNi 5 (Aksenov et al. 1983), where it should contribute through the hybridization of magnetic excitons and phonons. However, no accurate determination could be obtained. A recent magnetostriction study of TmNi 5 also provided one- and two-ion magneto-elastic coefficients (Barthem et al. 1989). Particularly due to the CEF configuration of this compound, the one-ion and two-ion magneto-elasticities could be separated. More precisely, the axial strain dependence of the isotropic exchange interaction ( a 2 - m o d e ) leads to a strongly anisotropic anomaly in the thermal expansion along a and c-axes below the ferromagnetic ordering temperature T c = 4.5 K, while the magnetostriction is more sensitive to the one-ion magneto-elasticity. However, the deduced B " values are quite large with regard to those in PrNi 5 and do not scale with aj. This behaviour is still to be explained. 4.5.2. R C o 5 c o m p o u n d s
Just as the cubic RFe 2 compounds, RCo 5 compounds are typical cases for applying the classical formalism, as the Co-sublattice induces very large bilinear interactions: the Curie temperatures are around 1000 K (e.g., 940 K in YCos). It is therefore difficult to study paramagnetic properties or to determine the CEF for the rare earth. A main feature is the huge magnetocrystalline anisotropy even in the absence of a magnetic rare earth as observed in YCo 5 (Alameda et al. 1983). From a magneto-elastic point of view, large anomalies have been observed in the temperature variation of the lattice parameters (Andreev et al. 1983, 1985b). It is difficult to observe magneto-elastic properties other than the spontaneous ones except in several compounds which exhibit a change in the direction of easy magnetization, and thus a temperature range of reduced anisotropy. Indeed, as classically explained in hexagonal Co (Barnier et al. 1962) and Gd (Klimker and Rosen 1973), this p h e n o m e n o n is driven by the change of sign at a given temperature of the K 1 anisotropy constant, which phenomenologically includes
104
P. MORIN and D. SCHMITI"
magneto-elastic contributions. This occurs for compounds with a negative Stevens coefficient aj for the rare earth (PrCo 5, NdCo 5, TbCo 5, DyCo 5 and HoCos). The corresponding temperature regions range from 13 K (in TbCos) to 130 K (in HoCos); the rotation of the magnetization induces clear anomalies in both the lattice parameters (Andreev et al. 1983) (fig. 61) and the ultrasonic modes (Deryagin et al. 1985) (fig. 62). It has been shown by X-ray studies that the magneto-elastic coefficient B ~ is small. On the other hand, the relative changes Aa/a and Ac/c of the lattice
v.gs~ --
1*~t~f,
,, -4 v, A3
.!'""I
q.g#Z~
~
85 1
"
Oq. g
q.960
~ I ~ J'l'~+"~
390
350
i
l-I
qJo T K
Fig. 61. Temperature dependence of the lattice parameters a and c, and of the volume V of the unit-cell in the temperature range of the change of easy axis in TbCo 5 (from Andreev et al. 1983).
z+
r,~r~r2
"E i
t
;'¢.2 I0.8
5.2 ~ 0
r
i
200
t r ~"w
I
o00T.~
Fig. 62. Temperature dependence of the elastic constants in the temperature range of the change of easy axis in DyCo 5 (from Deryagin et al. 1985).
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H I N T E R M E T A L L I C S
105
parameters at the average rotation temperature between the high-temperature (magnetization along the c-axis) and low-temperature (magnetization in the basal plane) phases indicate contributions of the two a-modes (see section 2.3). Temperature dependent magneto-elastic coefficients have been introduced (Deryagin et al. 1985),
V~
2(Aa 11 + 3 a
AC'~c~,
B12-
.
2 ( Aa
AC~c .
B~2=
AC'~c ~ 2 ( A a 2 (Aa X/3 2 - - + a c ! t 2 + 3 a
AC'}c ~
2 --a + c /
c /
c/
12 ,
22"
Experimentally, the values of (2Aa/a + Ac/c) have been observed to be very small. The 0 K values of B12 and B22 have been deduced assuming a temperature variation according to the Callen and Callen function is/z[sg-l(m)]. Note that some uncertainties may arise from inaccurate knowledge of the rare earth's magnetic moment, especially for the light rare earths (Andreev et al. 1983). The B~2(0 ) and B~2(0 ) are related to the B ~ and B "2 magneto-elastic coefficients of the present formalism through, BI:(O ) = 2B '~1
B~:(O) = 2V'-6 B "e
It is then possible to follow their variations through the entire series of rare earths. Table 17 collects the values for the different parameters in the series. The T A B L E 17 Magnetostriction constants at OK, h~2(0), h~2(0), associated to the relative change of lattice parameters at the rotation temperature, aa/a and Ac/c, elastic constants C~2 = 2X/2-7-gC~12, C~z = 2C ~2, B ~ and B ~2 magneto-elastic coefficients and their ratio to the Stevens coefficient a~ in RCo 5 compounds. RCo s compounds
h~2(0) x 103 [11 A~2(0) x 103 C~2 (101° Pa) C22 (1010Pa) B "1 (K) B "2 (K)
B~l/aj (K)
B"Z/a: (K)
PrCo s
NdCo s
SmCo 5
TbCo 5
DyCo 5
HoCo5
ErCo 5
TmCo 5
-1.1 2,2 1,6 a 20 a -6.6 -39 314 1860
-0.8 1.7 1.6 [2] 21.2 [2] -5.1 -25 793 3890
1.3 -2.6 1.6 = 22 a 24.6 143 597 3460
-1.8 3.5 1.7 [3] 23.3 [3] -5.5 -31.2 544 3090
-1.7 3.4 1.9 [31 25 [3] -3.7 -20.5 583 3230
-0.7 1.2 2a 27 ~ -1.2 -6.9 540 3100
0.6 -1.2 2" 27 a 1.4 7.6 551 3000
1.5 -3 2a 27" 5.5 30.8 544 3050
Note: aExtrapolated values. References: [1] Andreev et al. (1983). [2] Deryagin et al. (1984), [3] Deryagin et al. (1985).
106
R MORIN and D. SCHMITr
a f d e p e n d e n c e of B ~1 and B ~2 is very closely observed. The result indicates that the a - m o d e s depend on the rare earth and not on the cobalt. This conclusion is confirmed by the study of Y C % by X-ray diffraction. In this c o m p o u n d , the cobalt sublattice appears as contributing only through a two-ion magneto-elastic coupling, modifying the c/a ratio (Andreev et al. 1985a). The partial softening observed on the different elastic constants m a y be expected to be described using the coefficients B ~1, B ~a previously determined. Note that the softening observed on C44 during the rotation also indicates a sizeable magneto-elastic coefficient B ;. Some significant results might be provided by magnetostriction m e a s u r e m e n t s in high magnetic fields for the various symmetries in the rotation t e m p e r a t u r e range.
4.6. Dilute rare earth systems The magnetization of concentrated rare earth compounds generally includes oneand two-ion terms and m a y be strongly influenced by magnetic interactions. Dilute systems should, in principle, be easier to understand by limiting the magneto-elastic effects to one-ion contributions only. A dilution in the 1% range seems to be a satisfactory compromise between experimental sensitivity and residual two-ion effects. Most of the available studies concern rare earths diluted in noble-metal hosts, as Ag, A u and Cu.
4.6.1. Rare earths diluted in cubic noble-metal hosts The noble metals crystallize within the face centered cubic structure, and, because of the low solubility of rare earths in them, concentrations of rare earth impurities smaller than 1% were usually taken for the experiments. Silver-rare earth dilute compounds. A systematic study of heavy rare earths diluted in silver was carried out by magnetostriction m e a s u r e m e n t s on single crystals (Creuzet and Campbell 1981) (see table 18). Varying the cubic C E F TABLE 18 Magneto-elastic coefficients in the AgR series. All results were taken from Creuzet and Campbell (1982a) and from Creuzet (1982) and refer to magnetostriction measurements (rare earth concentration c = 0.5 at.%), except where indicated otherwise. AgR (R =) B r (K) B ~ (K) B(r4) (mK) B~4) (raN)
Tb
Dy
Ho
Er
Tm
2.72 3.9 [21 31.1 -6.3 -67
1.71 0.9 [21 12.1 1.9 45
0.60 3.5 [21 1.18 5.3 9
-1.24 <-0.39 [1] -7 -3.5 -29
-2.47 -3.3 [21 -16.1 -6.0 -200
References: [1] Garifullin et al. (1985) by EPR bulk, c = 0.03%. [2] Nieuwenhuys et al. (1982b) [alternate analysis of results of Creuzet and Campbell (1982a) with another set of CEF parameters].
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
107
parameters may affect the result of the magneto-elastic coefficients. However, the best agreement was achieved by using a CEF level scheme which also describes other results such as susceptibility measurements. They correspond to the same CEF parameters Aa(r 4) =--18 K and A6(r 6) = 7.6K across the series (except for Tm where A6(r 6) -= 6.5 K). These results have also been compared with estimates of the magneto-elastic coefficients obtained by EPR measurements (Pela et al. 1981). The EPR results appear to be systematically underestimated in experiments performed on strained thin films and, to a lesser extent, on strained bulk samples (Campbell and Creuzet 1985). Magnetostriction measurements were analyzed by considering second- and fourth-order one-ion magneto-elastic terms, the latter always contributing for about 10% to the total magnetostriction. The experimental values appear consistent along the series, and follow roughly the appropriate ~j or /3j Stevens coefficients. Typical values are, By oq
_
U~4)_
V~
3
V~
3
B ~
2V2 V2 ~ -270 K ,
U~4)
/3j - 16----~ V4 ~ - 4 0 K ,
3
5
O/j - W~ V 2 - -2100 K ~
5
/3j - 4V2 V4 ~ - 7 5 0 K ,
for all the dilute AgR compounds, the V~'s are defined by Campbell and Creuzet (1985). This coherency is reminiscent of the constant CEF parameters across the series. The isotropic one-ion magneto-elasticity was also investigated from these magnetostriction measurements (Creuzet 1982). The uncertainty in the coefficients is largely due to the weakness of the volume effects. Typical values for the AgR series are,
n(4)_ /3j
Weft
1
16V2 V4 - -120 K ,
n(6)
- -yj -
V3
1
64 V6~ -t- 135 K
Gold- and copper-rare earth dilute compounds. Compared to the AgR-series, rather less results are available on the A..__~uR and CuR dilute compounds. Systematic comparisons are therefore difficult. Nevertheless, the main result is the change of sign of B e and/or B y in AuR and C__uuRcompounds (see table 19), compared to AgR compounds. On the other hand, the volume magnetostriction appears to be still weaker in these compounds than in A__ggRones (Creuzet 1982). 4.6.2. Rare earths diluted in cubic pnictides A systematic study of the tetragonal magneto-elasticity was performed in the cubic NaCl-type LaSbR series by [001] magnetostriction measurements (Nieuwenhuys et al. 1982a). As in the A___ggRsystem, B y follows the sign of the Stevens factor % well (table 20). In comparison to corresponding concentrate
108
P. M O R I N and D. SCHMITT
T A B L E 19 Magneto-elastic coefficients in the A__._quRand C__._~uRseries. All the results are from Creuzet and Campbell (1982b,c) and Creuzet (1982); they were obtained by magnetostriction measurements (rare earth concentration e = 0.5 at.%), except when indicated otherwise. The C E F parameters used are A 4 ( r 4) = - 3 0 K, A 6 ( r 6) = 6 K for the A._.._uuRseries, A 4 ( r 4) = - 8 6 K, A 6 ( r 6) = 12 K for CuEr compound. Compound A..._~uTb B y (K)
1.7
B" (K)
-20
B~4~ (mK) B~4 ~ (mK)
12 -33
A..._~uEr [1]
AuYb
-0.8
-8.3 - 8 . 7 [11 89 44 [1] -17 1200
4.5
C__._~uEr[2] 0.44 -6.5
References: [I] Pela et al. (1981), from E P R measurements on films (c = 1%). [2] Garifullin et al. (1985), from E P R measurements on bulk samples (c = 0.03%).
compounds (section 4.2), these parameters appear to be quite comparable in magnitude, except in LaSbSm. The value obtained by magnetostriction in LaSbDy is consistent with that deduced from magnetization measurements. However, the latter may be considered to be more reliable than the former, because the magnetization reflects the lattice distortion in the immediate vicinity of the rare earth ions, while the magnetostriction cumulates the distortion effects far away from the rare earths. Indeed, it appears closer to the value in DySb than that obtained by magnetostriction. The most original feature in the series is the observation of a "reverse" magnetostriction in LaSbCe and LaSbSm, i.e., a change of sign of the distortion as a function of temperature, which is explained by the sign reversal of the expectation value ( O ° ) (see section 3.4.1). Magnetostriction was also observed for S-state Gd ions diluted in LaSb and LaBi (Hfifner et al. 1983). This has been explained within a phenomenological model which includes admixture effects of the first-excited 6p multiplet. However, corresponding magnetostriction values reach ~ l / l - - 8 x 10 -6 at 1.5 K in a 5 T field for 10%-dilute compounds. These values, scaled for the concentration, are only 1/10 of that observed in GdZn, where the distortion was interpreted as the manifestation of a two-ion magneto-elasticity (Morin et al. 1977). According to section 3.2.4, it could then be appropriate to consider the latter term explicitly. The magnetostriction measurements on LaSbR compounds were extended to the LaBiR series in order to check, in particular, the reverse magnetostriction in Ce and Sm based compounds (H/ifner et al. 1983). Although the CEF itself is smaller in bismuthides than in antimonides, the magneto-elastic coupling seems to be larger (see table 21).
QUADRUPOLAR
EFFECTS IN RARE EARTH INTERMETALLICS
i ~ ,...,
< II
~
~
~9
.~ o=
o C, t"q
Z
I
~
.I
cq
O
r.)
109
110
P. MORIN and D. SCHMITT TABLE 21 Crystal field parameters W, x and magneto-elastic coefficient B y in the diluted LaBiR series (c is the rare earth atomic concentration), from H~fner et al. (1983), from magnetostriction measurements. Rare-earth
W (K) x B y (K) c (%)
Ce
Tb
Er
1.91 1 -65.3 10
-0.45 -0.96 -43.7 3
0.20 0.79 10.0 3
4.6.3. Other dilute systems Few other cubic dilute rare earth systems have been studied with regard to magneto-elasticity. Only on Tb based compounds has there been a determination of magneto-elastic coupling parameters (see table 22). Creuzet et al. (1982) determined both coefficients B y and B e by magnetostriction measurements in cubic CsCl-type Tb0.01La0.99Ag compound. The B y value is roughly eight times smaller than in the concentrated TbAg compound (see table 7), which could be related to the incipient instability of the LaAg matrix (section 3.10). The same is valid for the value of G ~. On the other hand, the B e value seems to be doubtful, since it would lead to an unreasonable value of G~E, i.e., 1 K. Some studies have also been carried out on dilute systems having a lower symmetry, namely the hexagonal YR and MgR compounds. The three secondorder magneto-elastic coefficients B or1, Ba2 and B ~ were determined in YTb, __YDy and __YErcompounds (see table 23). A change of sign of all coefficients occurs for TABLE 22 Crystal field parameters W, x and magneto-elastic coefficients in some diluted rare-earth systems. Compound LaAgTb
w (K) x B ~ (K) B E (K) G y (mK) G" (mK)
0.5 -1 3.3 [1] - - 3 2 0 [1] - 1 [2]
LaA12Tb
0.27 0.9
0 [31 - 3 0 [31
References: [1] Creuzet et al. (1982), from magnetostriction measurements (c = 1%). [2] Hoenig et al. (1980), from magnetization (SQUID) measurements (c = 2%). [3] Hoenig (1980), from magnetization (SQUID) measurements (c = 1 to 6%).
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
111
TABLE 23 Magneto-elastic coefficientsin hexagonal YR system "(fromPureur et al. 1985), using magnetostriction measurements (c = 1 to 2%). Compound B "' (K)
B~2 (K) Be (K)
YTb
YDy
YEr
6.1 18.9 14.8
9.4 8.5 20.4
-1.9
-4.4 -8.4
the compound YEr, due to the Stevens factor aj. In addition, the ratio B ~ / a j is roughly constant throughout the series, B ' ~ / a j -~ V'3 B'~2/a, ~ ~
V2(/'a,1)/Olj
~
--850 K ,
V2(F~,2)/a s ~ -1600 K ,
B~/aj ==-~X/-2 Vz(F~,) / a j ~ -3200 K .
These values lead to distortions which are quite consistent, after correction for the concentration, with those observed in pure rare earth metals (Pureur et al. 1985). In a few cases, the introduction of a fourth-order magneto-elastic coefficient slightly improved the fits. In spite of the low concentration, spin-glass effects have been observed at low temperatures, in particular with Tb. Lastly, magnetostriction for YGd compound (c = 1%) reaches ~c/c ~ 10 -6 for a magnetic field of 6 T along the c axis, which is again of the same order of magnitude as in LaSbGd and LaBiGd (see above). Magnetostriction and thermal expansion data are also available for the heavy rare earths dilute in a single-crystal Mg host (Bijvoet et al. 1980, De Jong et al. 1980, 1982) (concentration about 0.1%). Here, also the magnetostriction agrees with the sign of aj. Unfortunately, no complete analysis was performed in order to determine the actual magneto-elastic coefficients. 4.7. Miscellaneous c o m p o u n d s
The universal character of the magneto-elasticity is emphasized in the literature through accurately conceived data on individual compounds, however, without thorough and systematic analyses across a given series. One can find, e.g., magneto-elastic effects in RCo 3 (Pourarian and Tajabor 1980) for which magnetostriction data on single crystals are available. In other series, like rare earth metals and rare earth borides few experimental results have been quantitatively analyzed. Rare earth metals. Among the rare earth metals, dhcp Pr has been widely studied over the last decade, because of the magneto-elastic effects on the criticality of antiferromagnetic interactions; the strain mode involved in these effects is the e mode (section 3.9). A magneto-elastic coefficient B ~ = 72 K acting
112
P. MORIN and D. SCHMITr
on the hexagonal Pr sites, i.e., G ~ = 76 mK, leads to a consistent description of various experiments, such as elastic constant measurements, magnetostriction, magnetic excitations and the occurrence of an antiferromagnetic ordering under an uniaxial stress (Houmann et al. 1979). This e-mode magneto-elastic coupling was also suggested to explain the field dependence of the thermal expansion (Liithi and Ott 1980). Without any magnetic field, these thermal expansion experiments have been described with the a-mode magneto-elastic coefficients of the hexagonal symmetry B al (-----~V/-3Bv) = - 4 6 K and B a2 (--=V~ B3) -~ 35.5 K. On the other hand, a detailed study of the magneto-elasticity was carried out by magnetostriction measurements (Hendy et al. 1979). A complete analysis was made in terms of one-ion and two-ion magneto-elastic coupling. It was found that one-ion a2-strain coupling is the dominant mechanism between the lattice and the 4f ion. Unfortunately no value for B ~ or D ~ was extracted from the experiments; only the h ~ coefficients relative to both cubic and hexagonal sites of Pr have been obtained. When taking the values of elastic constants from Greiner et al. (1973), the following evaluations can be made for the Pr ions on hexagonal sites: B "2 = 2 2 . 6 K , B ~ = 4 2 K , nal = o a~1 l = D~I a2 ~ 0. These values have the same order of magnitude as those from Houmann et al. (1979) in which the contribution arising from the cubic sites was included. Rare earth borides. Among the six families of binary rare earth borides (Etourneau 1985), RB 4 and R B 6 have been the ones most studied. Their crystallographic and magnetic properties have been shown to depend drastically on the nature and rate of impurities and on the metallurgical conditions. R B 6 compounds. RB 6 compounds all crystallize within a CsCl-type cubic arrangement of B octahedra and R ions. The main part of the literature on this series is devoted to light R compounds. Magneto-elastic and quadrupolar properties have been studied mainly in CeB 6 and negative quadrupolar interactions appear to be the origin of the properties observed (section 3.1.2). The tetragonal magneto-elastic coefficient [gr3[ = 50 K leads to [B~/,~j[ -- 875 K and the trigonal one = 135K to IB'/a,I = 3300K (L/ithi et al. 1984). PrB 6 exhibits properties reminiscent of those of CeB 6 in particular regarding its magnetic structure (section 3.6). The first-order transition (section 3.2.2) may result from bilinear and quadrupolar couplings of equivalent strength, as the bilinear exchange is larger than in CeB 6. Sizeable effects on the elastic constants are observed in PrB 6 and in NdB 6 (Tamaki et al. 1985). Unfortunately no quantitative analysis is available at the present time. In addition, the first-order Nrel transition in GdB 6 sets a puzzling problem which cannot be analyzed in terms of pure quadrupolar interactions. RB 4 compounds. The rare earth tetraborides have tetragonal crystal symmetry (Will et al. 1981). The rare earth atoms are confined in boron cages with an mm local symmetry (4 g). The compounds are antiferromagnetic with the exception of PrB 4 (Buschow and Creighton 1972). They exhibit some magneto-elastic features, especially ErB 4 which undergoes a tetragonal-orthorhombic transition at the Nrel temperature. At low temperature, the lattice symmetry (Pbam) agrees with the
[grsI
Q U A D R U P O L A R EFFECTS IN RARE EARTH INTERMETALLICS
113
magnetic one (Pb'am). However, the most interesting compound might be TbB 4 for which the structural transition occurs in the paramagnetic range at 80 K, far above the N6el temperature (T N = 43 K). Here again, lattice (Pbam) and magnetic (Pbam') symmetries are closely related. In both compounds the structural symmetry lowerings have been observed by X-ray diffraction on polycrystalline samples and are rather weak (Will et al. 1986). To the author's knowledge there are, at present, neither experimental determinations for the CEF nor for the magneto-elastic coupling. It is worth noting that the mm rare earth symmetry induces non-zero (O~) and (022) quadrupolar values. In addition, the easy magnetization directions do not vary across the series in the same way as the second-order Stevens coefficient, revealing that the anisotropy is not only of single-ion origin (Etourneau 1985). 5. Discussion
5.1. General analysis 5.1.1. Coherency of the various determinations To prove the relevancy of the basic Hamiltonian, coherent determinations of the quadrupolar coefficients in independent experimental ways are needed. For example, table 24 collects the various values obtained for the quadrupolar parameters in tetragonal symmetry for TmX compounds within the CsCl-type structure. A very good agreement is observed. In particular, the parastriction removes the ambiguity about the sign of the magneto-elastic coefficient which remains after the elastic constant analysis. The consistency between determinations obtained by direct (elastic constants) and indirect (parastriction and third-order susceptibility) ways, i.e., without and with the presence of an applied magnetic field, demonstrates the relevancy of the quadrupolar Hamiltonian and the weakness of the two-ion magneto-elasticity in this family of TmX compounds. However, we have to keep in mind that the latter contribution may be more significant for large Land6 factors (gj). Higher order one-ion magneto-elasticity as well as anharmonic elastic terms appear to be negligible here. From table 24, the one-ion magneto-elasticity of compounds with 4d elements (Ag, Cd) is observed to be weaker than that of compounds with 3d elements (Cu, Zn). The eccentric case of TmMg is consistent with the absence of d electrons on the Mg-ions (see section 5.2). The same behaviour is observed for the pairinteraction coefficient, which is at its maximum for II-B elements, in particular for zinc. In most cases, the pair interaction contribution to the total coefficient G ~ dominates the magneto-elastic one. TmMg is the only TmX compound within the CsCl-type structure not to be strongly influenced by tetragonal quadrupolar interactions. The microscopic origin of these different coefficients is analyzed in section 5.2. In the presence of cubic level schemes close to each other, this series provides us with clear experimental evidence of quadrupolar effects. Large softenings of
P. MORIN and D. SCHMITT
114
TABLE 24 Quadrupolar coefficients in TmX compounds within the CsCl-type structure. Compound
Coefficient
TmZn
TQ [1] Cll - C12 [2] Parastriction [2] X~ ~ [31
-+29 -31
20
r e [41 c n - c12 [4] Parastriction [5] x~ ~ [6]
-+12.7 -14
11.2
TmCd
B ~ (K)
K s (mK)
G ~ (mK)
Kr/~ ~ ~ M E
25 25 27 25 12.5 12.5 12 13
TmCu
Cn - C12 [7] Parastriction [7] X~ ) [31
-+30 -27
7
11.5 11 10.5
TmAg(8~
Cll - CI2
---20 - 20.3
2
4.4 4 3.5
-+5.4 -5.4
(0, 3) (0, 3) (0, 3)
(0, 3) (0, 3) (0, 3)
1.75 Parastriction X~ ) TmMg(9)
CI1 C12 Parastriction X~ ) --
References: [1] Morin et al. (1978b). [2] Givord et al. (1983). [3] Morin and Schmitt (1981b). [4] A16onard and Morin (1979). [5] Morin et al. (1978c). [6] Morin and Schmitt (1982c). [7] Jaussaud et al. (1980). [8] Giraud and Morin (1986). [9] Giraud et al. (1986). elastic constants exist, with t h e r m a l variations strongly d e p e n d i n g on the competition b e t w e e n the two q u a d r u p o l a r couplings, as shown in T m Z n and T m C d [fig. 1 in Liithi et al. (1979)]. Similar selective characteristics are o b s e r v e d in parastriction ( M o r i n et al. 1978c) and third-order m a g n e t i c susceptibility ( M o r i n and Schmitt 1981b) experiments.
5.1.2. Comparison o f the magneto-elastic coefficients T h e various magneto-elastic coefficients p r e s e n t e d in section 4 for cubic c o m p o u n d s are collected in figs. 63 and 64, after normalization to aj. T h e first c o m m e n t c o n c e r n s the Order of m a g n i t u d e which is relatively i n d e p e n d e n t o f the rare earth across a given series. This is reminiscent o f the b e h a v i o u r of 4th and 6th o r d e r C E F p a r a m e t e r s within a series. A relatively g o o d a g r e e m e n t is o b s e r v e d for the B~/aj ratio b e t w e e n concentrated R S b and dilute L a S b R c o m p o u n d s . This indicates, in particular, that the
Q U A D R U P O L A R EFFECTS IN RARE EARTH INTERMETALLICS i
I
I
5
I
i
i
i
I
RLoSb/ '~
\ /~-~Sb
,,
o. /..o
0
ePb3,,.," • C-3
"
o0
#0%
115
+
~
~
~
.0.-'%o. TmGo3I
÷
RA[2~
+
F
RMg
l'\
.
.
.
el
-5
RZn__
.
RAg
....... •
-
-
TmCd
RC-u". . . . . . . . ">'~,~~.>,...-:~~"
i
I
I
I
i
t
i
i
Ce
Pr
Nd
Tb
Dy
Ho
Er
Tm
Fig. 63. Magneto-elastic coefficient B y for the tetragonal symmetry normalized by the second-order Stevens coefficient in various series of rare earth cubic intermetallic compounds.
I
]
:PrPb 3 ,
L
TmGg3
.~ • PrAg lCePb3
TmMg.
,,t
0
Ce
o/-
Pr
Nd
Tb
Oy
Ho
Er
Tm
Fig. 64. Magneto-elastic coefficient B ~ for the trigonal symmetry normalized by the second-order Stevens coefficient in various series of rare earth cubic intermetallic compounds.
macroscopic strain measured in dilute samples is, in a first approximation, linearly related to the local deformation through the concentration of active sites. For a given series, the two magneto-elastic coefficients are often of opposite sign, as expected from point charge estimates as will be discussed in section 5.2. The only exceptions are the cases of AuCu3-type compounds and RAg dilute alloys. For a given type of structure, values have been observed to be similar in a first approximation. This is in particular true for the CsCl-type structure. However, the positive sign of B ~/as for TbZn is in opposition with the other determinations in the RZn series. This might be the signature of a two-ion contribution although it has been observed nowhere else for this symmetry. Note in addition the small average value of B~/a: in RZn.
116
P. M O R I N and D. SCHMITT
Among the different series, the values observed for the B~/aj's are of the same order of magnitude. This is reminiscent of the CEF parameters which usually correspond to total level spacings of a few hundred Kelvin. In the same way as peculiar surroundings or densities of electronic states drive small or large CEF spacings, as in RBe13 (Vigneron et al. 1980) and RB 6 (Zirngiebl et al. 1984), respectively, the magneto-elastic coefficients may also vary from one series to another. The actual spontaneous strain depends not only on the B" values, but also on other parameters. The CEF usually determines the axis of easy magnetization, then the symmetry lowering mode. The pair interactions state the purification of the CEF wave functions, then the various expectation values. For this reason, RFe 2 compounds develop "giant" strains at room temperature although their low-temperature values are in the same range as for other series. On the other hand, the spontaneous strain, which would be large in the case of a collinear spin structure, fully vanishes in the case of a multi-axial spin structure, preserving cubic symmetry as in DyCu. The magneto-elastic effects are of similar magnitude in rare earth intermetallics and insulators. For instance, in TbVO4, DyVO 4 and TmVO4, the vanadates which undergo a Jahn-Teller tetragonal-orthorhombic transition (Gehring and Gehring 1975), the low-temperature spontaneous strains are e ~= 12.7 × 10 -3, e ~ ' = 5.9 × 10 -3 and e ~= 3 × 10 -3, respectively. These strains are reminiscent of those in intermetallics. Unfortunately, whereas determinations of magneto-elastic coefficients in metals are numerous in cubic symmetry and scarce in tetragonal symmetry, the situation is the opposite in insulators; thus comparing the magneto-elastic coefficients is difficult. However, from the magneto-elastic contributions G~aE = 6 /[~1 and GME = /X2 determined from the ultrasonic velocity softenings (Melcher 1976), the following values IB~I--S7K, In l--80 and 22K are deduced in DyWO4, TbVO4, TmWO4, respectively. These magnitudes are similar to those observed in intermetallics.
5.1.3. Comparison of the quadrupolar pair-interaction coefficients There are fewer systematic studies for the pair interactions than there are for the magneto-elastic coupling. However, from table 24, their existence in TmX is well established and a very good coherency is observed in tetragonal symmetry. Within the CsCl-type structure, the tetragonal pair interactions dominate the magnetoelastic coupling and are the main origin of the ferroquadrupolar ordering. They are always positive and particularly important as soon as the aj coefficient is large. On the contrary, from the determinations in the paramagnetic phase, the trigonal pair interactions are negative and overwhelm the weak magneto-elasticity. For NaCl-type compounds, the tendency is less clear (fig. 65): K ~ is negative in DySb, but seems to be positive in HoSb and null in TmSb, in which no pair interactions seem to be present (Wang and Liithi 1977b). Tetragonal pair interactions are weaker than the magneto-elastic coupling in RSb pnictides. For the trigonal symmetry, the study of the TbX system indicates K ~ to be positive
Q U A D R U P O L A R EFFECTS IN R A R E EARTH INTERMETALLICS
20
-t
80
-Ag
117
.Ag -Go 3.~. '-Cd
-Cd
II
-'Cu
-Cu°
-Sb E 0 X)
- cd "zo ;Ag
"-P%
I
Ce
I
Pr
-Sb
-sbD
I
I
I
Tb Dy Ho Er
I
Tm
Fig. 65. Quadrupolar pair coefficient K ~ for the tetragonal symmetry in some rare earth cubic intermetallics (0: CsCl-type, +: AuCu3-type, [~: NaCl-type compounds).
and larger than G ~E. However, the total quadrupolar interactions remain smaller than the bilinear coupling and only modify the nature of the N6el transition. In insulators, such as the vanadates, the pair-interaction coefficients are fundamentally different. From the analysis of elastic constants (Melcher 1976) the coefficients K ~= h 1 and K ~= A2 are found to be negative and their absolute values are smaller than the magneto-elastic contribution (K ~= -0.35 mK in D y g O 4 , K ~ = -0.77 and - 1 1 mK in TmVO 4 and TbVO4, respectively). In these Jahn-Teller compounds, the structural transition is thus driven by the magnetoelastic coupling, whereas in the intermetallics the pair interactions drive the quadrupolar ordering and then, through the magneto-elastic coupling, the structural transition (Levy et al. 1979). In insulators the pair interactions are mediated by phonons (Gehring and Gehring 1975), and the h coefficient is a so-called "phonon exchange" parameter. It integrates the effective coupling of all the phonons to the Jahn-Teller ions for a given symmetry. The K/GME ratios are very close to --1/3, the theoretical expectation in the absence of any optic phonon mode contribution (Harley et al. 1972). On the other hand, the positive K values observed in intermetallics can originate neither from a k = 0 optical phonon in the NaC1- and CsCl-type structures, nor from the negative phonon self-energy term (Levy et al. 1979). In section 5.2 it will be shown that pair interactions in intermetallics are mainly mediated by conduction electrons.
5.2. Origin of the quadrupolar interactions 5.2.1. The one-ion magneto-elastic coupling From the comparison between the expressions for the cubic crystal field Hamiltonian [eq. (2)] and those for the one-ion magneto-elastic coupling [eq. (7)], it becomes obvious that the magneto-elastic coefficients are the strain derivatives of
118
P. MORIN and D. SCHMITT
the CEF parameters. For example, the second-order CEF parameter B ° vanishes in cubic symmetry, but its derivative with regard to a tetragonal distortion does not, B ~ , =-=
OB °
0~
"
The same is valid for a trigonal deformation, as well as for the higher-rank magneto-elastic coefficients, or if we consider crystals with lower symmetry, e.g., hexagonal or tetragonal ones. Therefore, the question of the origin of the magneto-elastic coupling is identical to the problem of the origin of the crystal field itself. In rare earth intermetallic compounds, it has been shown that the CEF parameters originate mainly from two contributions, namely the point charges and the contributions of the conduction electrons (Schmitt 1979). In cubic systems, these latter have been found to be preponderant, due to the strong anisotropic orbital character of the conduction electrons, as shown, e.g., by augmented-plane-wave band calculations (Belakhovsky and Ray 1975). In systems with hexagonal or tetragonal symmetries, the point-charge contribution seems to play a role which is relatively more important than in cubic systems, at least for the second-order CEF parameters. Then, calculating the magneto-elastic coefficients requires a precise knowledge of how these two CEF contributions depend on a given strain. First, in the presence of a weak strain, the localized charges are displaced from their previous non-strained arrangement, generating a new potential having the new symmetry of the lattice and proportional to the strain. This potential gives the point-charge contribution to the magneto-elastic coefficients. On the other hand, and to the first order of perturbation, this new potential modifies the conduction band. In particular, it lifts the orbital degeneracy of the conduction electron states in the vicinity of the Fermi level. According to Fermi statistics, this leads to a redistribution of the conduction electrons in the band which gives rise to a new CEF contribution, i.e., the band contribution to the magneto-elastic coefficients. Both contributions have been thoroughly investigated in cubic rare earth intermetallic compounds (Morin and Schmitt 1981a). In particular, the expressions for the direct and exchange Coulomb contributions of each type of conduction electrons (p, d, or f character) have been derived. It has been found that the magneto-elastic modifications of the cubic fourth- and sixth-order CEF parameters are negligible and that the second-order magneto-elastic coefficients result from the competition between numerous antagonistic contributions, which are of the same order of magnitude as the experimental values. Within the hypothesis of constant effective charges and a rigid-band model, all contributions are expected to vary as aj, ~j, yj for the second-, fourth- and sixth-order magneto-elastic coefficients, respectively. Furthermore, the pointcharge estimation leads to the constant ratio B ~ / B ~ = -4/V'-3 between the trigonal
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
119
and tetragonal second-order coefficients (see Creuzet 1982); this is not true for the band contribution. As shown in the previous section, this relation is far from being satisfied experimentally for a given compound, and this failure clearly indicates the importance of the conduction electrons. Nevertheless, the difference in sign between both parameters in most of the series as well as their overall magnitude (figs. 63 and 64) suggests a driving role for the point-charge contribution, contrarily to the case of the cubic CEF parameters. In realistic systems, because of the various contributions involved, the magnetoelastic coefficients are thus liable to vary significantly from one compound to another, especially if they belong to different series. This variation may occur via a shift of the values of the effective point charges. For the conduction-band contribution, the variation may be related to the modification of the partial densities of states at the Fermi level, which can be important when the Fermi level lies near a peak of the density of states. The importance of the d electrons on the rare earth site has also been suggested in dilute rare earth alloys (Campbell and Creuzet 1985). In these systems, the magneto-elastic coupling constants were evaluated within an indirect electronlattice interaction model, and using the nearly free electron approximation (Kozarzewski et al. 1985). The general features of the results obtained in this calculation are similar to those described above. The effect of screening of the magneto-elastic coupling by the conduction electrons has been found to significantly reduce the point-charge estimate of the parameters in Laves-phase rare earth intermetallics (Del Moral 1983). As a conclusion, ab initio evaluations of the one-ion magneto-elastic coefficients appear to be difficult to perform. As for the CEF parameters, they originate from numerous antagonistic contributions, and obtaining realistic values seems rather difficult. Qualitatively, their order of magnitude is well accounted for by point-charge and/or conduction electron contributions, and they follow the Stevens factor throughout a given series.
5.2.2. Quadrupolar pair interactions In metallic systems, conduction electrons have been shown to play an essential role in the coupling between magnetic ions, by propagating the local interactions between them and the localized magnetic electrons. This indirect exchange mechanism has been widely invoked in magnetic materials and is known as the Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction. Most of the studies has been focused on bilinear-type indirect coupling, which, in general, is the dominant term. Much less attention was paid to the origin of quadrupolar-type interactions. In rare earth and transition metal insulators, direct coupling between electric multipole moments of the two atoms and indirect coupling through the lattice-ion interaction have both been investigated (Birgeneau et al. 1969). However, in rare earth intermetallic compounds with the NaC1- and CsCl-type structures, both couplings have been proved to be negligible and therefore cannot be the dominant mechanism for the quadrupolar pair coupling (Levy et al. 1979): the
120
P. MORIN and D. SCHMITF
primary source of such a coupling is the indirect Coulomb and exchange interaction via the conduction electrons. The coupling of quadrupoles via the conduction band is analogous to the RKKY interaction quoted above, which involves the spins at different sites through the local s - f exchange interaction. For quadrupoles, it is the direct and exchange part of the full 4f-conduction electron (k-f) Coulomb interaction that contributes to the couplings K ". The corresponding effective couplings were calculated in second-order perturbation theory and by using plane-wave states for the conduction electrons (Teitelbaum and Levy 1976). A more realistic ab initio calculation of bilinear and quadrupolar pair coupling was performed later in the cubic rare earth intermetallic compound DyZn, by using conduction-electron wave-functions and energy bands obtained by a selfconsistent augmented-plane-wave method (Schmitt and Levy 1984, 1985). Spin and orbital effects in the complete k - f interaction were taken into account. The results underline the dominant role played by the d conduction electrons in the bilinear and quadrupolar couplings. In this calculation, the Fourier transform of the bilinear 0*(q) and quadrupolar Kr(q) and K~(q) interactions were found to be of the correct order of magnitude. Their different q-dependence was explained by the behaviour of specific matrix elements of the k - f interaction: the magnitude of these matrix elements is directly related to the symmetry of the two k-states involved. In particular, the ferromagnetism and the ferroquadrupolar tendencies (y-mode) of DyZn are associated with the strong eg-character of the conduction electrons near the Fermi level. On the contrary, the antiferroquadrupolar tendency of the trigonal coupling K~(q) is related to the small amount of t2g-type conduction electrons near the Fermi energy. The dependence of the bilinear and quadrupolar couplings on the Fermi energy was also investigated. Only part of the coupling, i.e. the intraband terms at q = 0, follows the partial densities of states at the Fermi level, which are associated with the type of electrons (% o r t2g ) from which they arise. The other terms, in particular the interband ones, are not directly correlated with the Fermi energy but depend on details of the entire energy band structure, including energy bands well above the Fermi level, which contribute for a noticeable amount. At last, the orbital interactions are the main contributions to the quadrupolar couplings K" so that the latter are expected to vary as otj. 2 That explains the importance of quadrupolar effects at the beginning and end of the rare earth series, where weak spin contributions (small De Gennes factor) allow the quadrupolar coupling to be better revealed. As for the one-ion magneto-elastic coefficients, ab initio evaluations of the two-ion quadrupolar parameters are very difficult to perform and would be necessary for each compound under consideration. 5.3. Conclusion
The aim of this chapter has been to present an extensive overview of the effect of the magneto-elastic coupling and of the quadrupolar interactions in rare earth intermetallic compounds. A complete Hamiltonian has been detailed, which
Q U A D R U P O L A R EFFECTS IN RARE EARTH INTERMETALLICS
121
includes all the relevant terms describing the one-ion magneto-elasticity as well as the two-ion quadrupolar coupling, for all the possible normal strain modes in cubic, hexagonal and tetragonal symmetries. The formalism for using this Hamiltonian has been developed, in particular that related to a perturbation theory. The various effects of the magneto-elastic coupling and of the two-ion quadrupolar interactions on the magnetic and elastic properties have been discussed, with a particular emphasis on the properties related to the perturbation methods: elastic constants, parastriction and third-order magnetic susceptibility. All these experimental manifestations appear as reliable ways to determine the quadrupolar parameters. Due to the strong anisotropic character of the magneto-elastic properties, using single crystals for experiments is quite essential. The magneto-elastic and quadrupolar data available in the literature have been summarized for a great number of rare earth series, within the present formalism, in order to make a comparison between them easier. Some series, especially those having the cubic CsC1- and NaCl-type structures, have been extensively studied in the last decades. Several others have been investigated much less, and some of them would be worth to be studied more thoroughly, in particular when single crystals become available. In most of the experimental results, the second-order one-ion magneto-elastic coefficients are concerned. The ratios B~'/%, in general, present a regular variation throughout a given series, but their magnitude and their sign can drastically change from one series to another. Nevertheless, their absolute value always remains in the 0-4000 K range. These differences show that, even if the point-charge contribution to the magneto-elastic coefficient is important, the role of the conduction electron is far from being negligible in these metallic compounds. An accurate ab initio estimate appears, however, difficult to obtain, due to the great number of different antagonistic contributions. The two-ion quadrupolar interactions have been studied less thoroughly in the rare earth intermetallic compounds. A consistent determination of the corresponding parameters has been successfully carried out in those series where the effects are strong, in particular those where a quadrupolar ordering occurs. Through their contribution to the free energy, the quadrupolar pair-interactions may strongly influence the character of the magnetic transitions, as well as, when they are negative, the stability of multi-axial spin structures. At last the dominant role of the conduction electrons has been emphasized in these systems, contrarily to the case of the rare earth insulators where the coupling between 4f quadrupoles is mainly due to the phonons. The existence of magneto-elastic and quadrupolar couplings then appear quite general in rare earth interrnetallic compounds. They need to be included in order to quantitatively understand the magnetic and elastic properties of these systems.
Acknowledgement It is a great pleasure to thank Professor P.M. L~vy from New York University for very fruitful discussions and a critical reading of the manuscript.
122
P. MORIN and D. SCHMITI"
Appendix I. Symmetrized Stevens operators for the cubic symmetry To completely express the magneto-elasticity of rank l (l = 2, 4, 6), all 21 + 1 Stevens operators are needed. In the present chapter we use the operators defined by Hutchings (1964), namely the quadrupolar operators of table 1 (l = 2) and the O / (m = 0 to l) and O'S(s) (m = 1 to l) for l = 4 and 6 (see table VIII of Hutchings 1964). The linear combinations of these operators, which transform according to the irreducible representations of the cubic symmetry group, read as, Second-rank 5"
O~ ' 1 = 002'
o l '2 = v ~ o i .
5"
Oi '1 = P~y,
° 2 '2 = Pr= ,
0 2 '3 = Pz=.
Fourth-rank ot 0 4 = 0 0 "~- 5 0 ~
FI:
r3:0~4
.
0~4'2 = -4X/-3 0 ] .
'1 = 004 -- 7 0 4 '
6,2
1
3
5:
064 '1 = O ~ ( S ) ,
0 4 = 0 4 -- 0 4 ,
5:
o i '1= o~4(s),
01,2 = --lOi(s
8,3
0 4 =
__ 0 1 ( S ) __ O I ( S )
) q- 7: O 43( s ) ,
0 4 , 3 = - - ~1 1O 4 -- 1 0 43.
Sixth-rank 5:O6
a
= O1--210~"
r~: 06~-- o l - 06 5:
O~ '1~- O 1 71-3 0 ~ 01,1 _-- 0 4 ( S ) ,
5:
01,2=
=
~5 r ~ 0 6 _2[ _
- - ~10 6I
~/'~
06 6 .
-- 5
~ 0 6 "31 - T11/',15 v6,
Y11~.5/ IJ6(S)", .
011,2 = ~ 01 16 ( s ) - 9
3 )-{- ~'~ O I ( s ) , ~06(S
1 9 3 33 [-15 0 6 1 3 = ~1 O 6-[~O 6 q- T t / 6 •
0621 = OI(S),
5:
O~,2
0~6,3 = 1o~6(s)_ ~5 O 63( s ) 06~, 1 = O I ( S ) ,
5:
,
3 1 0 6 ~ 3 = -~0 6-
5 3 ) + IOI(s), O12,2 = I O ~ ( s ) ~'- ~06(S 5 3 3 5 ~0 6 + ~0 6 •
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS
123
Appendix 2. Fourth- and sixth-rank one-ion magneto-elastic Hamiltonian ~Mm Starting from the expressions in table 2 and appendix 1, three terms may be derived for l = 4: one isotropic term which keeps the cubic symmetry, and two anisotropic terms as for l = 2, O~MEI(I = 4) = O~MEI(I = 4) + Yg~m(l = 4) + Y(MEI(I = 4 ) , with ~x ~MEI(I
m --
4) =
~ -B(4)8
a
0 4 ,
--~(4)\e, ltJ4
~/~MEI(I e = 4)
At- /:,2tJ4
) ,
= - - /nJ ( 4. .). . ['Ell . . J 4 1 ..}_ E 2e,,~e I J 4 , 2 + E 3. .[.J. 4 3,) .
The case l = 6 is identical to l = 4, except that, for the e mode, there exist two independent sets of linear combinations of the Stevens operators. Therefore, ~MEI(I = 6) = ffffMEl(l=6) + ffff~ciEl(l=6)+ ffffMEX(/ el e2 a = 6) + ~MEI(/ = 6), with oz
o~
~1;
~MEI(I = 6)
= -B(6)8
F3:
Y(~E,(I
:
F5:
fft~MEI(/ el = 6):
F5:
g Me2E I ( I = 6)
:
6)
ot
o~
0 6 ,
- - D, (~6r) ~, E Ir~r,1 I J 6 + ~_r,-~r,2x 2~6 J , ... 1,1 ..[_ 8 2 0 ; 1 , 2 _]_ E 3e,-~el,3x, - - Dh e( 61 ) .I'EIIA'6 IJ 6 ) ,
= - - / ~ ( 6 ) .j E . .l.l J 6
2,1 + 406
,2
+ E;O;2,3)
,
All the above expressions are referred to axes with four-fold symmetry, i.e., with z-axis parallel to the [001] direction. In some cases, however, it may be useful to consider the z-axis along a [111] direction, e.g., when a magnetic field is applied along that direction. In that case, symmetry considerations imply that the tetragonal strain e r should vanish and that the trigonal strains e" are all identical, e 7 = e 2 = e; = e'. Therefore, the corresponding trigonal magneto-elastic Hamiltonians each involve a single linear combination of symmetrized operators which may be rewritten in the new axes system (according to the convention of Hutchings 1964) as, I=2:
t'xy+ e,z+ ezx-, o ,
1=4:
0 4,1+ 0~,2 + 0 4,3~ ~(O 1 4o +7V'2 O ] ) ,
l=6:
O ; 1 , 1 + O ; 1 , 2 _[_ O;1,3..._> _ _ ~ ( 0 6 1 0 __ ~ / ~ 0 6 3
O62,1 + 062,2 + O6Z,3__> -- T~ 1 ( O 60 --
_ 110~)
2 5 X / 2 O 63 +
130~)
124
P. MORIN and D. SCHMITr
Appendix 3. Perturbation theory: the tetragonal mode Y First we have to define the zeroth-order energies E~ and the eigenvectors [ik) corresponding to the CEF Hamiltonian ~fCEF alone, ~CEF [ik) = E i lik)
(k
= 1,...,
m,).
(Cl)
In each degenerate subspace i, the m~ independent eigenfunctions lik) have to be adapted to the perturbation Y('~ [eq. (29)]. This can be achieved easily by choosing the following expansion in the IJ, M) basis [for the complete expressions, see Lea et al. (1962)], [ik) = aiklJ,1
M ) + a~2lJ,
M
- 4) +
a~klJ, M
- 8) + " "
.
(C2)
A perturbation theory carried out to second-order in e~ and fourth order in H allows one to obtain analytic expressions of the perturbed energies Ezk,
4 Eik = E i + 2
n=l
~ikl~(n)+ ' ' "
(C3)
"
Then, the partition function Z, Z = Z e -t~E'k ,
(C4)
i,k
can be calculated, where/3 = 1 / ( k a T ) , where k B is the Boltzmann constant and T the temperature. One obtains, Z=
ZcEFZ
1 r~ ,BVeV1 + KVQ): .... [I + ½/3Xo(H+ nM)2 + :PXvt
+ PXv'~(2)'B~ert1 + K r Q ) ( H + n M ) 2
-1-1/3(X(3) -t- 1/3(X0)2)(n--}-/,/M) 4-1-.. "l,
(C5)
where ZCE v is relative to the Ei's and Z .... to the elastic energy and to the corrective energy terms [see eqs. (14), (16), (17)]. Finally, this leads to the expression of the total free energy F ~ = - k BT In Z given in eq. (30). . (2) and x~ - (3) are given The expressions of the four CEF susceptibilities X0, X~, x~ by,
x0 = g . . B
i z.12
Il Ok,j, 2 i,k
l
+ 1 ~B T
I/ k,,kl 2 , \ I ;~.i~l )
'
(c6)
(C7)
Q U A D R U P O L A R EFFECTS IN R A R E E A R T H INTERMETALLICS t,(2)
,
2 2
= gylxB
125
Jik,jlQjl,J'l'JY'l',ik q- 2Qik,jlJjl,j'l'Jy'l',ik fi jei,l £ (U i -- E i ) ( E i -- Ej,) j'~i,l' (
Ei,k
I4~,~,l O,~.,~ + 2Oik,jl41,ikJik,ik (
- ]¢i,I 2
7--V
1
1
)
(c8)
+ 2(kBT)2 [J~k,ikl Q,k,ik , __ ,y ,•(3)
1 4 4 ( Jik,jlJjl,j'l'Jj'l',j'T'Jj'T',ik 2k~r (x°)= + g'~"£i,k f' - 4 /#i,, £ (E,--~---E-7')--~i----Ef) j'#i,l' j"¢i,l" 2
2
IJ,k,j,I IJ,k,/','~ + 2J,'k,/,J/t,/','J/',',ikJik,ik(
+2 ~
/~,,,
(E,- g)(E,- E/,)
2
E,
j' #i,l'
2
IJ....~12l.l,,.,jll 2
-2 ,,i,,E
((E i - Ej)2
~ii:-~])
i
+ 6k3T 3 IJik,ik
i4)
2 q- ( E i _
E/)kuT
,
+©I (C9)
where
Jik,/, =
(C10)
a.~,/, = ,
(Cll)
and
are the matrix elements of Jz and O ~ in the CEF levels. For each CEF sublevel, 1 exp(-/3Ei) f / - ZCEF exp(-/3E~) = ~ exp(-/3E~)
(C12)
i,k
is the Boltzmann population factor. Appendix 4. Perturbation theory: the trigonal mode e
When the magnetic field H is applied along a three-fold axis, e.g., [111], it is more convenient to first rotate the coordinate axes so that [111] is the new z-axis. Then the two combinations of operators in eq. (44) may be rewritten as:
L+~+L--.V~L,
126
P. MORIN and D. SCHMITT
and P x y + P y z q-
1
0
P~x --> ~02 "
The perturbation Y(] takes the new form:
Y({'= --g,p.B(H + n M ) ] z - ½(B~e ~ + K~P)O ° 3 ,.-~e; e.,2 + ~t..oke ) + ½nM2 + ~K*P 2.
(D1)
In each subspace i, the new zeroth-order eigenfunctions Ilk }' (corresponding to ! the appropriate YgCEF) have to be adapted to the above Hamiltonian. This condition is always fulfilled if these basis functions are chosen so that AM = 3 in the IY, M) expansion: 2 [ik )' = bill J, M ) + b~klJ, M - 3) + b~klj, M _ 6 )
+ ..- .
(D2)
The same perturbation theory as in appendix 3 may then be carried out, leading to the following expression for the total free energy F ~, 1 ! F ~ = Foe -- ~Xo( H + nM)2 -- ~XvkD 1 ..... e +K~P) 2 1 (2)' [n~ -~X~, ~D e e + K ~ P ) ( H + n M ) 2 - - ~i a, . y (3)'(H + nM)4
+ ~3 o,,.~e,, te
e,,2
) + l n M 2 + ~K~P 2,
(D3)
¢
where the new CEF susceptibilities X~ are related to the operators Jz and O ° in the new system of axes. Their expressions are identical to those given in appendix 3, but their values differ from the corresponding unprimed susceptibilities, due to the new CEF eigenfunctions lik)'. Returning to the initial system of four-fold axes and defining new trigonal susceptibilities we are led to the expression of eq. (45) for the free energy, where: Xo
:
,
Xo
X~
=
1
,
~X~,
. (2)
X~
1.
(2)'
= ga~,
/~3)
. (3)'
= ,'(v
"
(D4)
References Abbundi, R., and A.E. Clark, 1978, J. Appl. Phys. 49, 1969. Abbundi, R., A.E. Clark and N.C. Koon, 1979, J. Appl. Phys. 50, 1671. Abell, J.S., P. Hendy, E.W. Lee and K. AI Rawi, 1979, J. Phys. C 12, 3551. Abell, J.S., A. Del Moral, R.M. Ibarra and E.W. Lee, 1983, J. Phys. C 16, 769. Abu-Aljarayesh, I., J.S. Kouvel and T.O. Brun, 1986a, J. Magn. & Magn. Mater. 54-57, 512.
Abu-Aljarayesh, I., J.S. Kouvel and T.O. Brun, 1986b, Phys. Rev. B 34, 240. Aksenov, V.L., E.A. Goremychkin, E. Mfihle, T. Frauenheim and W. Bfihrer, 1983, Physica B 120, 310. Alameda, J.M., D. Givord, R. Lemaire and Q. Lu, 1983, J. Magn. & Magn. Mater. 31,191. Al6onard, R., and P. Morin, 1979, Phys. Rev. B 19, 3868. A16onard, R., and P. Morin, 1984, J. Magn. & Magn. Mater. 42, 151.
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS A16onard, R., and R Morin, 1985, J. Magn. & Magn. Mater. 50, 128. Al6onard, R., P. Morin, J. Pierre and D. Schmitt, 1976, J. Phys. F 6, 1361. Al6onard, R., P. Morin and J. Rouchy, 1984a, J. Magn. & Magn. Mater. 46, 233. A16onard, R., P. Morin, D. Schmitt and F. Hulliger, 1984b, J. Phys. F 14, 2689. A16onard, R., P. Morin and J. Rouchy, 1988, J. Phys. (France) Colloq. 49, C8-367. Andreev, A.V., A.V. Deryagin and S.M. Zadworkin, 1983, Sov. Phys.-JETP 58, 566. Andreev, A.V., A.V. Deryagin and S.M. Zadworkin, 1985a, Phys. Met. & Metallogr. (USSR) 59, 116. Andreev, A.V., A.V. Deryagin, S.M. Zadworkin and G.M. Kvashnin, 1985b, Sov. Phys.Solid State 27, 1985. Andres, K., D. Davidov, P. Dernier, F. Hsu, W.A. Reed and G.J. Nieuwenhuys, 1978, Solid State Commun. 28, 405. Andronenko, S.I., A.N. Bazhan, I.A. Bondar, V.A. Ioffe and B.Z, Malkin, 1983, Sov. Phys.-Solid State 25, 239. Assmus, W., R. Takke, R. Sommer and B. L/ithi, 1978, J. Phys. C 11, L575. Barbara, B., D.K. Ray, M.F. Rossignol and F. Sayetat, 1977a, Solid State Commun. 21, 513. Barbara, B., M.F. Rossignol and M. Uehara, 1977b, Physica B 86-88, 183. Barbara, B., M.F. Rossignol, J.X. Boucherle and C. Vettier, 1980, Phys. Rev. Lett. 45, 938. Barnier, Y., R. Pauthenet and G. Rimet, 1962, Le Cobalt 15, 1. Barthem, V.M.T.S., D. Gignoux, A. NaitSaada, D. Schmitt and G. Creuzet, 1988, Phys. Rev. B 37, 1733. Barthem, V.M.T.S., D. Gignoux, D. Schmitt and G. Creuzet, 1989, J. Magn. & Magn. Mater. 78, 56. Bean, C.P., and D.S. Rodbell, 1962, Phys. Rev. 126, 104. Belakhovsky, M., and D.K. Ray, 1975, Phys. Rev. B 12, 3956. Bijvoet, J., M.H. De Jong H. H61scher and P.F. de Chatel, 1980, in: Crystalline Electric Field and Structural Effects in f-electron Systems, eds J.E. Crow, R.P. Guertin and T.W. Mihalisin (Plenum, New York) p. 83. Birgeneau, R.J., M.T. Hutchings, J.M. Baker and J.D. Riley, 1969, J. Appl. Phys. 40, 1070.
127
Birgeneau, R.J., E. Bucher, J.P. Maita, L. Passell and K.C. Tuberfield, 1973, Phys. Rev. B 8, 5345. Birss, R.R., 1964, in: Symmetry and Magnetism, ed. E.P. Wohlfarth (North-Holland, Amsterdam) ch. 2. Bloch, J.M., and D. Davidov, 1982, Phys. Rev. B 26, 3631. Blume, M., 1966, Phys. Rev. 141, 517. Bohr, J., D. Gibbs, D.E. Moncton and K.L. D'Amico, 1986, Physica A 140, 349. Bucher, E., K. Andres, A.C. Gossard and J.P. Maita, 1972a, Conf. on Low Temperature Physics, LT-13, 2, 322. Bucher, E., R.J. Birgeneau, J.P. Malta, G.P. Felcher and T.O. Brun, 1972b, Phys. Rev. Lett. 28, 746. Bucher, E., J.P. Maita, G.W. Hull, L.D. Longinotti, B. Lfithi and P.S. Wang, 1976, Z. Phys. B 25, 41. Burlet, P., J.M. Effantin, J. Rossat-Mignod, S. Kunii and T. Kasuya, 1988, J. Phys. (France) Colloq. 49, C8-459. Buschow, K.H.J., 1980, in: Handbook of Ferromagnetic Materials, Vol. 1, ed. E.P. Wohlfarth (North-Holland, Amsterdam) p. 297. Buschow, K.H.J., and J.H.N. Creygthon, 1972, J. Chem. Phys. 57, 3910. Buschow, K.H.J., J.P. de Jong, H.W. Zandbergen and B. van Laar, 1975, J..Appl. Phys. 46, 1352. Cable, J.W., W.C. Koehler and E.O. Wollan, 1964, Phys. Rev. 136, A240. Callen, E.R., and H.B. Callen, 1963, Phys. Rev. 129, 578. Callen, E.R., and H.B. Callen, 1965, Phys. Rev. 139, A455. Campbell, I.A., and G. Creuzet, 1985, J. Phys. F 15, 2559. Campbell, I.A., G. Creuzet and J. Sanchez, 1979, Phys. Rev. Lett. 43, 234. Castets, A., D. Gignoux and B. Hennion, 1982, Phys. Rev. B 25, 337. Chen, H.H., and P.M. Levy, 1971, Phys. Rev. Lett. 27, 1383. Chen, H.H., and P.M. Levy, 1973, Phys. Rev. B 7, 4267. Chiu-Tsao, S.T., P.M. Levy and C. Paulson, 1975, Phys. Rev. B 12, 1819. Chiu-Tsao, S.T., and P.M. Levy, 1976, Phys. Rev. B 13, 3046. Clark, A.E., 1980, in: Handbook of Ferromagnetic Materials, Vol. 1, ed. E.P. Wohlfarth (North-Holland, Amsterdam) p. 533.
128
P. MORIN and D. SCHMITT
Clark, A.E., J.R. Cullen and K. Sato, 1974, AIP Conf. Proc. 24, 670. Clark, A.E., R. Abbundi, H.T. Savage and O.D. McMasters, 1977, Physica B 86-88, 73. Cooper, B.R., 1982, J. Magn. & Magn. Mater. 29, 230. Creuzet, G., 1982, Thesis (University of ParisSud). Creuzet, G., and I.A. Campbell, 1981, Phys. Rev. B 23, 3375. Creuzet, G., and I.A. Campbell, 1982a, J. Phys. (France) 43, 809. Creuzet, G., and I.A. Campbell, 1982b, J. Appl. Phys. 53, 8104. Creuzet, G., and I.A. Campbell, 1982c, J. Magn. & Magn. Mater. 27, 221. Creuzet, G., I.A. Campbell and H.E. Hoenig, 1982, Solid State Commun. 44, 733. Cullen, J.R., and A.E. Clark, 1977, Phys. Rev. B 15, 4510. Czopnik, A., N. Iliew, B. Stalinski, H. Madge, C. Bazan and R. Pott, 1985, Physica B 130, 262. Dacorogna, M., J. Ashkenazi and M. Peter, 1982, Phys. Rev. B 26, 1527. de Jong, M.H., H. H61scher, P.F. de Chatel and J. Bijvoet, 1980, J. Magn. & Magn. Mater. 15-18, 17. de Jong, M.H., J. Bijvoet and P.F. de Chatel, 1982, in: Crystalline Electric Field and Structural Effects in f-electron Systems, eds R.P. Guertin, W. Suski and Z. Zolnierek (Plenum, New York) p. 261. Del Moral, A., 1983, J. Phys. C 16, 4637. Del Moral, A., J.I. Arnaudas, M.R. Ibarra, J.S. AbeU and E.W. Lee, 1986, J. Phys. C 19, 579. Del Moral, A., M.R. Ibarra, J.S. Abell and J.F.D. Montenegro, 1987, Phys. Rev. B 35, 6800. Deryagin, A.V., G.M. Kvashnin and A.M. Kapitonov, 1984, Phys. Met. & Metalloved. 57, 53. Deryagin, A.V., G.M. Kvashnin and A.M. Kapitonov, 1985, Sov. Phys.-Solid State 27, 155. Du Tr6molet de Lacheisserie, E., 1970, Ann. Phys. 5, 267. Du Tr6molet de Lacheisserie, E., 1988, J. Magn. & Magn. Mater. 73, 289. Du Tr6molet de Lacheisserie, E., P. Morin and J. Rouchy, 1978, Ann. Phys. (France) 3, 479.
Effantin, J.M., 1985, Thesis (University of Grenoble), unpublished. Effantin, J.M., J. Rossat-Mignod, P. Buffet, H. Bartholin, S. Kunii and T. Kasuya, 1985, J. Magn. & Magn. Mater. 47-48, 145. Englman, R., 1972, The Jahn-Teller Effect in Molecules and Crystals (Wiley, London). Englman, R., and B. Halperin, 1970, Phys. Rev. B 2, 75. Etourneau, J., 1985, J. Less-Common Met. 110, 267. Everett, G.E., and P. Streit, 1979, J. Magn. & Magn. Mater. 12, 277. Felcher, G.P., T.O. Brun, R.J. Gambino and M. Kuznietz, 1973, Phys. Rev. B 8, 260. Fischer, P., A. Furrer, E. Kaldis, D. Kim, J.K. Kjems and P.M. Levy, 1985, Phys. Rev. B 31, 456. Fujii, H., T. Kitai, Y. Uwatoko and T. Okamoto, 1985, J. Magn. & Magn. Mater. 52, 428. Fujiwara, H., H. Kadomatsu and M. Kurisu, 1983, J. Magn. & Magn. Mater. 31-34, 189. Furrer, A., W.J.L. Buyers, R.M. Nicklow and O. Vogt, 1976, Phys. Rev. B 14, 179. Furrer, A., W.J.L. Buyers, R.M. Nicklow and O. Vogt, 1977, Physica B 86-88, 105. Furrer, A., W. Hfilg, H. Heer and O. Vogt, 1979, J. Appl. Phys. 50, 2040. Garifullin, I.A., T.O. Farzan, G.G. Khaliullin and E.F. Kukovitsky, 1985, J. Phys. F 15, 979. Gefen, Y., and M. Rosen, 1981, J. Phys. & Chem. Solids 42, 857. Gehring, G.A., and K.A. Gehring, 1975, Rep. Prog. Phys. 38, 1. Gibbs, D., J. Bohr, J.D. Axe, D.E. Moncton and K.L. D'Amico, 1986, Phys. Rev. B 34, 8182. Gignoux, D., and F. Givord, 1983, J. Magn. & Magn. Mater. 31-34, 217. Gignoux, D., F. Givord and R. Lemaire, 1975, Phys. Rev. B 12, 3878. Gignoux, D., J.C. Gomez-Sal, R. Lemaire and A. De Combarieu, 1977, Solid State Commun. 21, 637. Gignoux, D., F. Givord, R. Perrier de la Bathie and F. Sayetat, 1979, J. Phys. F 9, 763. Giraud, M., and P. Morin, 1986, J. Magn. & Magn. Mater. 58, 135. Giraud, M., P. Morin, J. Rouchy, D. Schmitt and E. Du Tr6molet de Lacheisserie, 1983, J. Magn. & Magn. Mater. 37, 83.
QUADRUPOLAR EFFECTS IN RARE EARTH INTERMETALLICS Giraud, M., P. Morin and D. Schmitt, 1985, J. Magn. & Magn. Mater. 52, 41. Giraud, M., P. Morin, J. Rouchy and D. Schmitt, 1986, J. Magn. & Magn. Mater. 59, 255. Givord, D., P. Morin and D. Schmitt, 1983, J. Magn. & Magn. Mater. 40, 121. Givord, D., P. Morin and D. Schmitt, 1985, J. Appl. Phys. 57, 2127. Godet, M., and H.G. Purwins, 1976, Helv. Phys. Acta 49, 821. Godet, M., and E. Walker, 1978, Helv. Phys. Acta 51, 178. Greenough, R.D., G.N. Blackie and S.B. Palmer, 1981, J. Phys. C 14, 9. Greidanus, F.J., L.J. de Jongh, W.J. Huiskamp, P. Fischer and A. Furrer, 1983a, Physica B 119, 215. Greidanus, F.J., G.J. Nieuwenhuys, L.J. de Jongh, W.J. Huiskamp and H.W. Capel, 1983b, Physica B 119, 228. Greiner, J.D., R.J. Schiltz, J.J. Tonnies, F.H. Spedding and J.F. Smith, 1973, J. Appl. Phys. 44, 3862. Hfifner, H.U., D. Davidov and G.J. Nieuwenbuys, 1983, J. Magn. & Magn. Mater. 38, 45. Hfilg, B., A. Furrer and O. Vogt, 1985, Phys. Rev. Lett. 54, 1388. Harley, R.T., W. Hayes and S.R.P. Smith, 1972, J. Phys. C 5, 1501. Heer, H., A. Furrer, W. H/ilg and O. Vogt, 1979, J. Phys. C 12, 5207. Hendy, P., K.M. AI Rawi, E.W. Lee and D. Melville, 1979, J. Phys. F 9, 2121. Hiraoka, T., 1974, J. Phys. Soc. Jpn. 37, 1238. Hoenig, H.E., 1980, J. Magn. & Magn. Mater. 15-18, 19. Hoenig, H.E., R. Voitmann and W. Assmus, 1980, in: Crystalline Electric Field and Structural Effects in f-electron Systems, eds J.E. Crow, R.P. Guertin and T.W. Mihalisin (Plenum, New York) p. 187. Horn, S., F. Steglich, M. Loewenhaupt, H. Scheuer, W. Felsch and K. Winzer, 1981, Z. Phys. B 42, 125. Houmann, J.G., B.D. Rainford, J. Jensen and A.R. Mackintosh, 1979, Phys. Rev. B 20, 1105. Hulliger, F., 1978, J. Magn. & Magn. Mater. 8, 183. Hulliger, F., 1980, J. Magn. & Magn. Mater. 15, 1243. Hulliger, F., and B. Natterer, 1973, Solid State Commun. 13, 221.
129
Hulliger, F., and T. Siegrist, 1979, Z. Phys. B 35, 81. Hulliger, F., and F. Stucki, 1978a, in: Rare Earths and Actinides (Institute of Physics, London) p. 92. Hulliger, F., and F. Stucki, 1978b, Z. Phys. B. 31, 391. HuUiger, F., M. Landolt, H.R. Ott and R. Schmelczer, 1975, J. Low Temp. Phys. 20, 269. Hulliger, F., H.R. Ott and T. Siegrist, 1983, J. Less-Common Met. 94, 270. Hutchings, M.T., 1964, Solid State Phys. 16, 227. Iandelli, A., and A. Palenzona, 1979, in: Handbook on the Physics and Chemistry of the Rare Earths, Vol. 1, eds K.A. Gschneidner Jr and L. Eyring (North-Holland, Amsterdam) p. 1. Ibarra, M.R., E.W. Lee, A. Del Moral and J.S. Abell, 1986, J. Magn. & Magn. Mater. 5457, 882. Ihrig, H., D.T. Vigren, J. K/ibler and S. Methfessel, 1975, Z. Phys. B 22, 231. Jaakkola, S.M., and M.K. Hanninen, 1980, Solid State Commun. 36, 275. Jaussaud, C., P. Morin and D. Schmitt, 1980, J. Magn. & Magn. Mater. 22, 98. Jensen, J., 1979, J. Phys. (France) 40, C5-1. Jensen, J., N.H. Andersen and O. Vogt, 1980, J. Phys. C 13, 213. Kanamori, J., 1960, J. Appl. Phys. 31, 145. Kaneko, T., H. Yoshida, M. Ohashi and S. Abe, 1987, J. Magn. & Magn. Mater. 70, 277. Kasaya, M., B. Liu, M. Sera, T. Kasuya and D. Endoh, 1985, J. Magn. & Magn. Mater. 52, 289. Kataoka, M., and J. Kanamori, 1972, J. Phys. Soc. Jpn. 32, 113. Keating, D.T., 1969, Phys. Rev. 178, 732. Kim, D., and P.M. Levy, 1982, J. Magn. & Magn. Mater. 27, 257. Kjems, J.K., H.R. Ott, S.M. Shapiro and K. Andres, 1978, J. Phys. (France) 39, C61010. Klimker, H., and M. Rosen, 1973, Phys. Rev. B 7, 2054. Knorr, K., B. Renker, W. Assmus, B. L/ithi, R. Takke and H.J. Lauter, 1980, Z. Phys. B 39, 151. Knorr, K., A. Loidl and C. Vettier, 1983, Phys. Rev. B 27, 1769. Koehler, W.C., 1972, in: Magnetic Properties
130
P. MORIN and D. SCHMITT
of Rare Earth Metals, ed. R.J. Elliott (Plenum, New York) p. 81. Koehler, W.C., J.W. Cable, E.O. Wollan and M.K. Wilkinson, 1962, Phys. Rev. 126, 1672. K6tzler, J., 1984, Z. Phys. B 55, 119. K6tzler, J., and G. Raffius, 1980, in: Crystalline Electric Field and Structural Effects in f-electron Systems, eds J.E. Crow, R.P. Guertin and T.W. Mihalisin (Plenum, New York) p. 117. Kouvel, J.S., and T.O. Brun, !980, Phys. Rev. B 22, 2428. Kozarzewski, B., J. Kurzyk and J. Deniszcyk, 1985, Phys. Status Solidi B 129, 143. Kurisu, M., 1987, J. Phys. Soc. Jpn. 56, 4064. Lacueva, G., and P.M. Levy, 1985, Phys. Rev. B 31, 650. Lea, K.R., M.J.M. Leask and W.P. Wolf, 1962, J. Phys. & Chem. Solids 23, 1381. L6vy, F., 1969, Phys. Kondens. Mater. 10, 85. Levy, P.M., 1973, J. Phys. C 6, 3545. Levy, P.M., and G.T. Trammell, 1977, J. Phys. (USA) C 10, 1303. Levy, P.M., P. Morin and D. Schmitt, 1979, Phys. Rev. Lett. 42, 1417. Lin, C.L., J. Peter, J. Crow, T. Mihalisin, J. Brooks, A.I. Abou-Aly and G.R. Steward, 1985, Phys. Rev. Lett. 54, 2541. Lines, M.E., and E.D. Jones, 1965, Phys. Rev. A 139, 1313. Lingner, C., and B. L/ithi, 1983, J. Magn. & Magn. Mater. 36, 86. Loidl, A., K. Knorr, J.K. Kjems and B. Lfithi, 1979, Z. Phys. B 35, 253. Loidl, A., K. Knorr, M. Mullner and K.H.J. Buschow, 1981, J. Appl. Phys. 52, 1433. L/ithi, B., 1980a, in: Dynamical Properties of Solids, eds G.K. Horton and A.A. Maradudin (North-Holland, Amsterdam) p. 247. L/ithi, B., 1980b, J. Magn. & Magn. Mater. 15-18, 1. L/ithi, B., and C. Lingner, 1979, Z. Phys. B 34, 157. L/ithi, B., and H.R. Ott, 1980, Solid State Commun. 33, 717. Lfithi, B., M.E. Mullen and E. Bucher, 1973a, Phys. Rev. Lett. 31, 95. L/ithi, B., M.E. Mullen, K. Andres, E. Bucher and J.P. Malta, 1973b, Phys. Rev. B 8, 2639. L/ithi, B., R. Sommer and P. Morin, 1979, "J. Magn. & Magn. Mater. 13, 198. L/ithi, B., S. Blumenroder, B. Hillebrands, E.
Zirngiebl and G. Guntherodt, 1984, Z. Phys. B 58, 31. Maetz, J., M. Mullner, H. Jex, W. Assmus and R. Takke, 1980, Z. Phys. B 37, 39. Markosyan, A.S., 1980, Sov. Phys.-Solid State 22, 2023. Markosyan, A.S., 1981, Sov. Phys.-Solid State 23, 670. McCarthy, C.M., C.W. Thompson, R.J. Graves, H.W. White, Z. Fisk and H.R. Ott, 1980, Solid State Commun. 36, 861. McEwen, K.A., W.G. Stirling and C. Vettier, 1978, Phys. Rev. Lett. 41, 343. McWhan, D.B., C. Vettier, R. Youngblood and G. Shirane, 1979, Phys. Rev. B 20, 4612. Melcher, R.L., 1976, in: Physical Acoustics XII, eds W.P. Mason and R.N. Thurston (Academic Press, New York) p. 1. Moran, T.J., R.L. Thomas, P.M. Levy and H.H. Chen, 1973, Phys. Rev. B 7, 3238. Morin, P., 1988, J. Magn. & Magn. Mater. 71, 151. Morin, P., and A. de Combarieu, 1975, Solid State Commun. 17, 975. Morin, P., and J. Pierre, 1975, Phys. Status Solidi A 30, 549. Morin, P., and D. Schmitt, 1978, J. Phys. F 8, 951. Morin, P., and D. Schmitt, 1979, Phys. Lett. A 73, 67. Morin, P., and D. Schmitt, 1980, J. Magn. & Magn. Mater. 21, 243. Morin, P., and D. Schmitt, 1981a, Phys. Rev. B 23, 2278. Morin, P., and D. Schmitt, 1981b, Phys. Rev. B 23, 5936. Morin, P., and D. Schmitt, 1982a, J. Magn. & Magn. Mater. 28, 188. Morin, P., and D. Schmitt, 1982b, Phys. Rev. B 26, 3891. Morin, P., and D. Schmitt, 1982c, The RareEarths in Modern Science and Technology, eds G.J. McCarthy, H.B. Silber and J.J. Rhyne (Plenum, New York) p. 419. Morin, P., and D. Schmitt, 1983a, Phys. Rev. B 27, 4412. Morin, P., and D. Schmitt, 1983b, J. Magn. & Magn. Mater. 31-34, 1059. Morin, P., and D. Schmitt, 1986, J. Magn. & Magn. Mater. 54-57, 463. Morin, P., and S.J. Williamson, 1984, Phys. Rev. B 29, 1425. Morin, P., J. Pierre and J. Chaussy, 1974, Phys. Status Solidi A 24, 425. Morin, P., A. Waintal and B. Liithi, 1976, in:
Q U A D R U P O L A R EFFECTS IN R A R E EARTH INTERMETALLICS Proc. 12th Rare Earth Research Conf., Vail, Colorado, USA, ed C.E. Lundin (US Bureau of Mines, Denver) p. 95. Morin, P., J. Rouchy and E. Du Tr6molet de Lacheisserie, 1977, Phys. Rev. B 16, 3182. Morin, P., J. Pierre, D. Schmitt and A.P. Murani, 1978a, Solid State Commun. 25, 265. Morin, P., J. Rouchy and D. Schmitt, 1978b, Phys. Rev. B 17, 3684. Morin, P., D. Schmitt and E. Du Tr6molet de Lacheisserie, 1978c, Phys. Lett. A 69, 217. Morin, P., J. Rouchy, D. Schmitt and E. Du Tr6molet de Lacheisserie, 1979, J. Phys. (France) 40, C5-101.. Morin, P., D. Schmitt and E. Du Tr6molet de Lacheisserie, 1980a, J. Magn. & Magn. Mater. 15-18, 601. Morin, P., D. Schmitt and E. Du Tr6molet de Lacheisserie, 1980b, Phys. Rev. B 21, 1742. Morin, P., D. Schmitt, C. Vettier and J. RossatMignod, 1980c, J. Phys. F 10, 1575. Morin, P., D. Schmitt and C. Vettier, 1981, J. Phys. F 11, 1487. Morin, P., D. Schmitt and E. Du Tr6molet de Lacheisserie, 1982, J. Magn. & Magn. Mater. 30, 257. Morin, P., D. Osterman, S.J. Williamson and D. Schmitt, 1985a, Physica B 130, 544. Morin, P., D. Schmitt and C. Vettier, 1985b, J. Phys. (France) 46, 39. Morin, P., M. Giraud, P.L. Regnault, E. Roudaut and A. Czopnik, 1987a, J. Magn. & Magn. Mater. 66, 345. Morin, P., J. Rouchy and G. Creuzet, 1987b, J. Magn. & Magn. Mater. 69, 99. Morin, P., J. Rouchy, M. Giraud and A. Czopnik, 1987c, J. Magn. & Magn. Mater. 67, 95. Morin, P., M. Giraud, P. Burlet and A. Czopnik, 1987d, J. Magn. & Magn. Mater. 68, 107. Morin, P., J. Rouchy and D. Schrnitt, 1988, Phys. Rev. B 37, 5401. Mullen, M.M., B. Lfithi, P.S. Wang, E. Bucher, L.D. Longinotti, J.P. Maita and H.R. Ott, 1974, Phys. Rev. B 10, 186. Murray, A.F., and W.J.L. Buyers, 1980, in: Crystalline Electric Field and Structural Effects in f-electron Systems, eds J.E. Crow, R.P. Guertin and T.W. Mihalisin (Plenum, New York) p. 257. Nereson, N., 1973a, AIP Conf. Proc. 10, 669. Nereson, N., 1973b, J. Appl. Phys. 44, 4727. Nieuwenhuys, G.J., D. Davidov and H.U. Hfifner, 1982a, Phys. Rev. Lett. 49, 1202.
131
Nieuwenhuys, G.J., D. Davidov, H.U. Hfifner and J.M. Bloch, 1982b, Solid State Commun. 43, 51. Nikl, D., I. Kouroudis, W. Assmus, B. Ltithi, G. Bruls and U. Welp, 1987, Phys. Rev. B 35, 6864. Niksch, MI, W. Assmus, B. Liithi, H.R. Ott and J.K. Kjems, 1982, Helv. Phys. Acta 55, 688. Ott, H.R., 1977, in: High Pressure and Low Temperature Physics, eds C.W. Chu and J.A. Woolam (Plenum, New York) p. 205. Ott, H.R., and B. Lfithi, 1977, Z. Phys. B 28, 141. Ott, H.R., K. Andres, P.S. Wang, Y.H. Wong and B. Liithi, 1977a, in: Crystal Field Effects in Metals and Alloys, ed. A. Furrer (Plenum Press, New York) p. 84. Ott, H.R., B. L/ithi and P.S. Wang, 1977b, in: Valence Instability and Related Narrow Band Phenomena, ed. Parks (Plenum, New York) p. 289. Palmer, S.B., and J. Jensen, 1978, J. Phys. C 11, 2465. Pela, C.A., J.F. Suassuna, G.E. Barbefis and C. Rettori, 1981, Phys. Rev. B 23, 3149. Peysson, Y., C. Ayache, J. Rossat-Mignod, S. Kunii and T. Kasuya, 1986, J. Phys. (France) 47, 113. Pierre, J., 1970a, in: Proc. Colloq. Int. du CNRS: Les Elements de Terres Rares, p. 65. Pierre, J., 1970b, Thesis (University of Grenoble) A.O. CNRS 2680, unpublished. Pierre, J., and B. Hennion, 1982, in: Crystalline Electric Field Effects in f-electron Magnetism, eds R.P. Guertin, W. Suski and Z. Zolnierek (Plenum, New York) p. 275. Pierre, J., P. Morin, D. Schmitt and B. Hennion, 1978, J. Phys. (France) 39, 793.. Pierre, J., A.P. Murani and R.M. Galera, 1981, J. Phys. F 11, 679. Pierre, J., R.M. Galera and J. Bouillot, 1984, J. Magn. & Magn. Mater. 42, 139. Pourarian, F., and N. Tajabor, 1980, Phys. Status Solidi A 61,537. Pureur, P., G. Creuzet and A. Fert, 1985, J. Magn. & Magn. Mater. 53, 121. Pynn, R., W. Press, S.M. Shapiro and S.A. Werner, 1976, Phys. Rev. B 13, 295. Raffius, G., and J. K6tzler, 1983, Phys. Lett. A 93, 423. Ray, D.K., and J. Sivardi~re, 1978, Phys. Rev. B 18, 1401. Rossat-Mignod, J., 1979, J. Phys. (France) 40, C5-95..
132
P. MORIN and D. SCHMITT
Rossat-Mignod, J., E Burlet, J. Villain, H. Bartholin, Wang Tcheng-Si, D. Florence and O. Vogt, 1977, Phys. Rev. B 16, 440. Rossignol, M.F., 1980, Thesis (University of Grenoble), unpublished. Rouchy, J., and E. Du Tr6molet de Lacheisserie, 1979, Z. Phys. B 36, 67. Rouchy, J., P. Morin and E. Du Tr6molet de Lacheisserie, 1981, J. Magn. & Magn. Mater. 23, 59. Sablik, M.J., and Y.L. Wang, 1978, J. Appl. Phys. 49, 1419. Sablik, M.J., and Y.L. Wang, 1979, Phys. Rev. B 19, 2729. Sablik, M.J., and Y.L. Wang, 1985, J. Appl. Phys. 57, 3758. Sankar, S.G., S.K. Malik, V.U.S. Rao and W.E. Wallace, 1977, in: Crystal Field Effects in Metals and Alloys, ed. A. Furrer (Plenum, New York) p. 153. Sehmitt, D., 1979, J. Phys. F 9, 1745, 1759. Schmitt, D., and P.M. Levy, 1984, Phys. Rev. B 29, 2850. Schmitt, D., and P.M. Levy, 1985, J. Magn. & Magn. Mater. 49, 15. Schmitt, D., P. Morin and J. Pierre, 1978, J. Magn. & Magn. Mater. 8, 249. Sekizawa, K., H. Chihara and K. Hasukochi, 1981, J. Phys. Soc. Jpn. 50, 3467. Sivardi6re, J., 1975, J. Magn. & Magn. Mater. 1, 23. Sivardibre, J., 1976, J. Magn. & Magn. Mater. 1, 183. Sivardi~re, J., and M. Blume, 1972, Phys. Rev. B 5, 1126. Sivardi6re, J., A.N. Berker and M. Wortis, 1973, Phys. Rev. B 7, 343. Stanley, H.B., J.S. Abell, M.R. Ibarra, E.W. Lee, O. Moze and B.D. Rainford, 1985, Physica B 130, 280. Stevens, K.W.H., 1952, Proc. Phys. Soc. A 65, 209. Takahashi, H., K. Takegahara, A. Yanase and T. Kasuya, 1982, in: Valence Instabilities, eds P. Wachter and H. Boppart (NorthHolland, Amsterdam) p. 379. Takke, R., N. Dolezal, W. Assmus and B. Liithi, 1981, J. Magn. & Magn. Mater. 23, 247. Tamaki, A., T. Goto, M. Yoshizawa, T. Fujimura, S. Kunii and T. Kasuya, 1985, J. Magn. & Magn. Mater. 52, 257.
Taub, H., and C.B.R. Parente, 1975, Solid State Commun. 16, 857. Taub, H., and S.J. Williamson, 1973, Solid State Commun. 13, 1021. Teitelbaum, H.H., and P.M. Levy, 1976, Phys. Rev. B 14, 3058. Trammell, G.T., 1963, Phys. Rev. 131, 932. Tsunoda, Y., M. Mori, M. Kunitomi, Y. Teraoka and J. Kanamori, 1974, Solid State Commun. 14, 287. Ushizaka, H., S. Murayama, Y. Miyako and Y. Tazuke, 1984, J. Phys. Soc. Jpn. 53, 1136. Vettier, C., D.B. McWhan, E.I. Blount and G. Shirane, 1977, Phys. Rev. Lett. 39, 1028. Vigneron, F., M. Bonnet and R. Kahn, 1980, in: Crystalline Electric Field and Structural Effects in f-electron Systems, eds J.E. Crow, R.P. Guertin and T.W. Mihalisin (Plenum, New York) p. 513. Walline, R.E., and W.E. Wallace, 1964, J. Chem. Phys. 41, 3285. Walline, R.E., and W.E. Wallace, 1965, J. Chem. Phys. 42, 604. Wang, ES., and B. L/ithi, 1977a, Physica B 86-88, 107. Wang, ES., and B. Lfithi, 1977b, Phys. Rev. B 15, 2718. Will, G., W. Sch/ifer, F. Pfeiffer, F. Elf and J. Etourneau, 1981, J. Less-Common Met. 82, 349. Will, G., Z. Heiba, W. Sch/ifer and E. Jansen, 1986, AIP Conf. Proc. 140, 130. Wintemberger, M., and R. Chamard-Bois, 1972, Acta Crystallogr. A 28, 341. Wintemberger, M., R. Chamard-Bois, M. Belakhovsky and J. Pierre, 1971, Phys. Status Solidi 48, 705. Wun, M., and N.E. Phillips, 1974, Phys. Lett. A 50, 195. Yoshida, H., S. Abe, T. Kaneko and K. Kamigaki, 1987, J. Magn. & Magn. Mater. 70, 275. Zaplinski, E, D. Meschede, D. Plumacher, W. Schlabitz and H. Schneider, 1980, in: Crystalline Electric Field and Structural Effects in f-electron Systems, eds J.E. Crow, R.P. Guertin and T.W. Mihalisin (Plenum, New York) p. 295. Zirugiebl, E., B. Hillebrands, S. Blumenroder, G. G/intherodt, M. Loewenhaupt, J.M. Carpenter, K. Winzer and Z. Fisk, 1984, Phys. Rev. B 30, 4052.
chapter 2 MAGNETO-OPTICAL SPECTROSCOPY OF f-ELECTRON SYSTEMS
W. REIM Siemens AG Zentrale Forschung und Entwicklung, ZFE ME TPH 11 8520 Erlangen F.R.G. and
J. SCHOENES ETH ZEirich Laboratorium FEtr Festk&rperphysik 8093 Z(Jrich Switzerland
Ferromagnetic Materials, Vol, 5 Edited by K.H.J. Buschow and E.P. Wohlfarth1" © Elsevier Science Publishers B.V., 1990
CONTENTS 1. l ~ t r o d u c t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. ~[LeoJy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Ph e n o r L e n e l o g y . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1. De flnition c f t h e o p t i c a l f u n c t i o n s . . . . . . . . . . . . . . . 2.1.2. F a r a d a y a n d K e r r effects . . . . . . . . . . . . . . . . . . 2.2. Microscol:ic Ir o d e l s to c a l c u l a t e ~ . . . . . . . . . . . . . . . . . . 2.2.1. Inlerle a n d t r a n s i t i o n s . . . . . . . . . . . . . . . . . . . . 2.2.2. Ce ndu c t i o n - e l e c l r o n e ffecls . . . . . . . . . . . . . . . . . . 2.3. T h e H a l l effect a n d the K r a m e r s - K r o n i g e q u a t i o n s . . . . . . . . . . . . 3. E x p e r i m e n t a l c o n s i & r a t i o n s . . . . . . . . . . . . . . . . . . . . . 3.1. T e c h n i q u e s . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. S a m p l e s . . . . . . . . . . . . . . . . . . . . . . . . . . 4. L f g h t l ~ t e e a r t h s . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. N a C l - t y p e c e r i u m , n e o d y m i u m a n d s a m a r i u m c o m p o u n d s . . . . . . . . . . 4.2. ~IhsP4-type c e m p c u n d s . . . . . . . . . . . . . . . . . . . . . 4.3. I n t e r m e t a l l i c c o m p o u n d s . . . . . . . . . . . . . . . . . . . . . 5. Half-filled .,hell i are e a l t h s . . . . . . . . . . . . . . . . . . . . . . 5.1. N a C l - t y p e e u r e p i u m c e m p c u n d s . . . . . . . . . . . . . . . . . . 5.2. D o p e d N a C l - t y F e c o m p e u n d s . . . . . . . . . . . . . . . . . . . 5.3. M iscell~ n e e u s e u r o p i u m c o m p e u n d s . . . . . . . . . . . . . . . . . 5.4. G a d o l i n : ' u m . . . . . . . . . . . . . . . . . . . . . . . . . 5.5. l n t e r m e t a l l i c gadoliniu m c omF oun ds . . . . . . . . . . . . . . . . . 6. H e a ~ y r a t e e a r l h s . . . . . . . . . . . . . . . . . . . . . . . . . 6.1. N a C l - t y p e c o m p o u n d s . . . . . . . . . . . . . . . . . . . . . . 6.2. H e a ~ y r a l e e a r t h m e t a l s . . . . . . . . . . . . . . . . . . . . . 6.3. M i s c e l l a n e o u s c o m p o u n d s c o n t a i n i n g h e a v y r a r e e a r t h s . . . . . . . . . . . 7. U r a n i u m ~omF o u ~ d s . . . . . . . . . . . . . . . . . . . . . . . . 7.1. U r a n i u m d i o x i d e . . . . . . . . . . . . . . . . . . . . . . . 7.2. B i n a r y N a C l - t y p e c o m p o u n d s . . . . . . . . . . . . . . . . . . . 7.3. P s e u d o - b : n a l y N a C l - t y p e c omit o u r ds . . . . . . . . . . . . . . . . . 7.4. qTh3P4 s~Iucture cc mlcounc's . . . . . . . . . . . . . . . . . . . . 7.5. T e r n a r y c o m p o u n d s a n d i n t e r m e t a l l i c s . . . . . . . . . . . . . . . . 8. C o n c l u s i c n ~ n d outloc k . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .
134
135 137 139 139 141 147 150 153 156 157 157 160 162 162 169 172 175 176 186 190 192 194 196 196 202 203 206 206 208 215 222 225 230 231
1. Introduction
Magneto-optical phenomena have been known for more than a century. It was M. Faraday who reported in 1846 (Faraday 1846) the first genuine magneto-optical effect, namely the rotation of the polarization plane of linearly polarized light on transmission through a lead-borosilicate glass to which a magnetic field was applied parallel to the light propagation direction (fig. la). Since then, magnetooptics has gone through several highs. The discovery of what today we call Faraday rotation was a clear proof of the interaction of electromagnetic radiation with magnetized matter. With the experiments of Zeeman (1896, 1897) and the classical electron theory of Lorentz, magneto-optics allowed the determination of the sign of the electron charge and of the ratio e/rn. The developments of quantum mechanics prompted a large number of magneto-optical experiments on gases as well as on diamagnetic and paramagnetic solids to test the implements of the new atomic theory (see, e.g., Schiitz 1936). Microwave absorption on magnetized semiconductors opened the field of magneto-optical determination of band parameters in classical semiconductors like Ge, Si or InSb (Dresselhaus et al. 1955). The sixties brought a first wave of magneto-optical investigations of magnetic semiconductors which, in the following two decades, led to a wider recognition that magneto-optics used as spectroscopy is a very powerful tool to study the electronic structure of magnetic materials. The actual increasing research efforts in magneto-optics rely on one hand, on this basic physical interest and, on the other hand, on the potential for using magneto-optics for erasable optical memories. The present review focuses on the spectroscopic aspect of magneto-optics with the goal to relate the experimentally determined magneto-optical spectra to electronic transitions and, thus, to provide directions for searching for improved magneto-optical materials. We shall not discuss results of Raman scattering, microwave absorption or other resonance effects, which, in a wider sense of the word, are also magneto-optical effects. Instead, we shall mostly review Faraday spectra and polar magneto-optical Kerr effect spectra, and a few times, Voigt effect and equatorial and longitudinal Kerr effect spectra. This selection reflects the different interest which the various magneto-optical effects have received in the past 25 years. As will become clear in the following, the qualitative difference between these various effects is the different proportionality of the magneto135
136
W. REIM and J. SCHOENES
a) FARADAY ROTATION
8
o;I
BE--Re (~)= ~-~ (n+- n-) DF=-Im ( { ) = ~cl (k+- k_) h) MO KERR EFFECT
Polar
Longitudinal
Equatorial
N
n_-n+ 8pK= Im ~÷~__~----~ N N_-N.
PK = R e
~+~'-"1
Fig. 1. Scheme of the Faraday effect (a) and the various magneto-optical Kerr effects (h). O~Kand ~PK are the polar Kerr rotation and ellipticity, respectively, which in the text of this chapter are named OK and eK, for simplicity,if+ and ~ are the complex refractive indices for right and left hand circularly
polarized light, respectively (after Schoenes 1981). optical signal to the magnetic field. In the Faraday configuration, for which the field direction is parallel to the light propagation direction, the linearly polarized wave can be decomposed into a right- and a left-hand circularly polarized wave (fig. la). The rotation of the plane of polarization OF indicates a phase shift R e ( l ~ ) between the two partial waves. The Faraday rotation is, therefore, a magnetic circular birefringence. If the material absorbs light, the difference of the absorption indices for the two circularly polarized waves, k+ - k , gives rise to magnetic circular dichroism DF, which manifests itself as elliptically polarized light. Both effects are, to lowest order, linear functions of the magnetic field. In the Voigt (or Cotton-Mouton) configuration the external magnetic field is perpendicular to the light propagation direction. The difference in the refractive indices for fight polarized parallel and perpendicular to the magnetic field nil - n± gives rise to transverse or linear magnetic birefringence. The difference in the absorption indices k f f - k± leads to transverse or linear dichroism. Both effects are, to lowest order, quadratic functions of the magnetic field. As a consequence, these effects are generally smaller than the circular effects. In addition, they
MAGNETO-OPTICAL SPECTROSCOPY
137
cannot be used to determine the magnetization direction of domains and, therefore, they have attracted much less attention than the circular effects. If the polarized light is reflected from a magnetized solid, one speaks of the magneto-optical Kerr effect (Kerr 1877). The various orientations of the magnetic field relative to the plane of incidence and the direction of propagation allows to distinguish the polar, the longitudinal and the equatorial magneto-optical Kerr effect (fig. lb). In the polar Kerr effect, the light propagation direction is parallel (or antiparallel) to the field direction. This is similar to the situation of the Faraday effect in transmission and, indeed, the polar Kerr effect is, to lowest order, a linear function of the applied field. Yet, the polar Kerr rotation 0K and the polar Kerr ellipticity eK are not linear functions of the complex refractive indices rT+ and tT_ and the usual Kramers-Kronig relation does not hold between 0K and eK (Kahn et al. 1969). Up to this point we have always assumed the magnetic field as to be the driving force for the magneto-optical effects. This is correct, as long as the temperature is not too low and the field is not too high, for diamagnetic materials to come in the region of Landau quantization and for paramagnetic materials to reach paramagnetic saturation. Also, for magnetically ordered materials the field is to be replaced by the magnetization. The last remark which we would like to make in this introduction, regards the choice of the data to be included in this chapter. We have tried to be as complete as possible for the magneto-optical spectra of magnetically ordered systems containing as magnetic ions f-elements only. Problems arise for those materials which contain also transition elements other than the actinide or lanthanide elements. Examples are transition metal-rare earth intermetallics, many garnets and rare earth-ferrites. Magneto-optics of transition metal-rare earth intermetallics are widely discussed in chapter 5 of vol. 4 of this handbook and here we have only included data in direct connection with pure f-electron systems or very recent results. Garnets and ferrites have not been included to keep the article to a tractable size and because the signals related to the rare earth are in most cases only marginal. Albeit we did a computer search, we may have overlooked valuable data for this review. We hope that the authors of these overlooked publications, should there be any, will accept our apologies beforehand, and will understand our difficulties the more they themselves search in the labyrinth of terms for magneto-optical effects and the infinite number of compounds that all f-elements can possibly form. 2. Theory The analysis of the magneto-optical activity of a substance is generally carried out in two stages; first, the measured quantities like the Kerr rotation 0~: and ellipticity eK are related to the dielectric tensor g: or the conductivity tensor ~. In a second step, these functions are then interpreted in terms of band-structure features or may even be calculated directly from relativistic and spin-polarized band-structure calculations.
138
W. REIM and J. SCHOENES
The object of this theoretical section is to provide a self-consistent description of important interrelations between the different optical and magneto-optical functions. In section 2.1.1 and fig. 2 we give a sign convention on, which all graphs in this chapter will be based. For Fe in the visible spectral region, a negative Kerr rotation and a positive Faraday rotation are in agreement with this definition of the sign. The Faraday rotation of quartz glass in the visible region has a positive sign, as it is usually found in literature (Gmelin 1959, Landolt-B6rnstein 1950, Robinson 1964). In section 2.1, we derive the equations connecting the measured quantities Kerr effect or Faraday effect with the elements of the conductivity tensor & or the dielectric tensor g:, which describe the optical and the magnetooptical material properties. Section 2.2 deals with the interpretation of the frequency dependence of & in terms of the spin-polarized band structure of the material, and frequently used concepts will be reviewed. In the same section, we discuss one possible way of large Kerr effect engineering, namely by a splitting of the plasma edge of a metal for different light polarizations. The importance of the
a)
~
Z
Kerr-ellipticity X
6F
c)
Y
Faraday-ellipticity Y Fig. 2. Scheme to illustrate the sign convention for the Faraday and Kerr effects. (a) Electric field vector E at a fixed time t o for right ( + ) hand circularly polarized light (rcp). The incoming beam travels along the + z direction parallel to the applied field B. Looking always along the propagation direction of the light, E describes a clockwise rotation at fixed time t o for the incoming and the transmitted beams but a counter-clockwise rotation for the reflected beam. (b) Time dependence of E+ and E at fixed position z 0 after reflection from the sample for a material with dominant absorption of rcp-radiation, i.e., R+ > R _ and consequently [E+[ > [E I. The Kerr ellipticity is positive [eq. (2.20)]. (c) Time dependence of E+ and E_ at fixed position z 0 after transmission through the sample with dominant absorption of rcp-radiation, i.e., k+ > k and, consequently, IE÷ [ < [E 1. The Faraday ellipticity is positive [eq. (2.40)]. Note the different sense of rotation of the E vector for rcp-light in transmission and reflection.
M A G N E T O - O P T I C A L SPECTROSCOPY
139
normal optical constants for Kerr-effect enhancement will be critically discussed. Finally, section 2.3 shows the relation of non-optical measurements like the Hall effect to magneto-optics and how these additional data may be used efficiently. Also, the Kramers-Kronig integral equations and their application to magnetooptics will be discussed.
2.1. Phenomenology 2.1.1. Definition of the optical functions A circularly polarized wave propagating along the z-axis is described by the electrical field vector E+ = Re[E 0 exp{i(wt - 2rrff+Z/Ao)}(e x +-ie,)],
(2.1)
where upper ( + ) and lower ( - ) signs refer to right- and left hand circular polarization, A0 is the wavelength in vacuum, rT± is the complex index of refraction for the two states of polarization and e~ and ey are unit vectors in x- and y-direction, respectively. Looking in + z direction of a right-handed cartesian coordinate system (direction of propagation), at a fixed place z0, this definition of right (+) circular polarization describes a counterclockwise rotation of the electric field vector with time which is equivalent to a clockwise rotation of E along the direction of propagation at a fixed time to (see fig. 2). To obtain positive values for the absorptive quantities the choice of sign for the time factor in the exponent directly implies the definition ~ , = n± - i k ± ,
(2.2)
for the complex refractive index where n and k are the real indices of refraction and absorption, respectively. The magneto-optical effects are associated with the influence of a magnetic perturbation (either applied field H or intrinsic magnetization M) on the conductivity tensor & or the dielectric tensor g: of a medium. These two tensors are not independent but describe the same phenomena and are related by i47r ~ e]j = 6u - - - - o'u, ¢O
(2.3)
where 6u = 1 for i = j and 6u = 0 for the off-diagonal elements i # j. With the magnetization along the z-direction and applying general symmetry arguments (Roth 1964, Agranovich and Ginzburg 1966), we define the conductivity tensor as
(
O'xy
(2.4)
o and the dielectric tensor g: according to (2.3). All elements of ~ are complex
140
W. R E I M and J. S C H O E N E S
functions O-ij = O'lij "~-
i°'2i/•
(2.5)
By. decomposition of & into a Hermitian and an anti-Hermitian contribution (LeGall and Jamet 1971) it can be shown that the absorptive parts of the diagonal and the off-diagonal elements are O-ixx and %xy, respectively. Because ~ / / ( M ) = ~ j i ( - M ) the diagonal and the off-diagonal tensor elements are even and odd functions of the magnetization, respectively (Roth 1964). The given conductivity tensor applies to isotropic media, to cubic structures with crystallographic axes along x, y, z and to hexagonal and tetragonal structures with the c-axis along z. For crystals with lower symmetry, both the tensor and its relation to the optical properties become more complex (Emslic and Aronson 1983) and natural birefringence often inhibits magneto-optical measurements. The dielectric response of the medium is described by the Maxwell equations: V × E -
1 dB c dt'
(2.6)
V × H = - c 1 [ 47rJ+ d-~tl ,
(2.7)
and V - B = 0 and V - D = 0 in a conducting medium. Neglecting higher orders in the moment densities of the medium in the presence of applied fields, we write D = E + 4~-P and H = B - 4~-M for the new field quantities, electric displacement D and magnetic field H (compared to the averaged electric field E and magnetic induction B). Further, we neglect nonlinear effects in the material response and use P=oLIE , M = X 1 H and the Ohm law J=CrlE with the real tensors polarizability % and Xa and conductivity cr 1. Finally, we define the real dielectric function as e 1 = ~ + 4~-%, and the real permeability P.1 = 8 + 47rxi, with 8 defined as in eq. (2.3). Using these definitions in eqs. (2.6) and (2.7) and with the time dependence + itot as defined in eq. (2.1), we obtain, 1.
V x E = - - l~ttH, C
(2.8)
V×H=
(2.9)
1 [4¢rtr 1 ] dE c k iw + el dt "
Now, we define the complex dielectric function • 4~r g: = e 1 - 1 - - 0"1 ,
(2.10)
(O
which is equivalent to the definition of a complex conductivity (Tr=
{1¢ 1
-[- i o ~ a l
= Cr~ + i OJ(el -- 1 )
4rr
(2.11)
MAGNETO-OPTICAL
SPECTROSCOPY
141
Equation (2.11) together with eq. (2.7) finally allows a generalization of the Ohm law,
io) ( i o )
lq--~-~el E :
./tot =
~+~--~8 E,
(2.12)
with the total current Jtot = (1/4~-)D + J. In general, besides on frequency, the optical functions will also depend on the wavevector g which points along the propagation direction and whose length is complex and equal to 2~rrT/A0. However, as long as the wavelength of the radiation is large compared to atomic distances, which is certainly the case for photon energies of several eV as used for standard magneto-optical measurements, the wavevector dependence can be neglected. Applying the Maxwell equations eqs. (2.6) and (2.7) on an electromagnetic field with E and H defined according to (2.1) one obtains n~z =
~gx~,
(2.13)
which connects the complex index of refraction with the diagonal element of gz. It has sometimes been argued that in ferromagnets the linear relation M = R1H does not hold and/x ~ 1. We follow the argumentation of Landau and Lifshitz (1960) and Pershan (1967) that at optical frequencies one should take /x = 1 even for ferromagnets. With/x = 1, we obtain 6,1x x
=
n 2 -- k 2 ,
(2.14) (2.15)
e2x x = 2 n k ,
and ~xx is defined according to eq. (2.3). The optical reflectivity is given by the Fresnel equations (Born 1964). At normal incidence, the amplitudes r+ and r_ for right and left circular polarizations are r_+ =
Ir_+ I e ia÷ -
n_+
-
ik+
-
1
n_+ - ik_+ + 1 '
(2.16)
which leads to the well-known formula for the reflected intensity R+=r~= 2.1.2.
(n+ - 1) 2 + k 2 ( n + + l ) 2+kz+
(2.17)
F a r a d a y a n d K e r r effects
In reflection, the interaction of light with a magnetized medium is traditionally decomposed into three fundamental effects, namely the transverse (=equatorial), the longitudinal (=meridional) and the polar Kerr effect. These effects differ only in the orientation of the magnetization relative to the plane of incidence and relative to the surface of the medium (fig. lb). All effects are described by the
142
W. REIM and J. SCHOENES
same optical tensor & or ~:, however, only the polar effect and a specific equatorial configuration give linear and the other configurations can give terms linear as well as quadratic in magnetization (Pershan 1967, Hunt 1967). A theoretical formulation of these three effects extended to multi-layer thin films has been given by Smith (1965) and a general treatment of the Kerr effect for arbitrary orientation but restricting to effects linear in the magnetization was published by Hunt (1967). In this review, however, we will restrict to the polar Kerr effect, cbK = OK-- ie K ,
(2.18)
because most experimental work to be reported has been done in this configuration. In this geometry, the magnetization is perpendicular to the surface and parallel to the direction of light propagation, i.e., in + z direction (fig. 2). The eigenmodes are right and left circular polarizations and ~bK is expressed in terms of the Fresnel coefficients (2.16), without any approximations as OK = -
eK -
½(a+ - a _ ) ,
(2.19)
Ir+l-lr-I
ir+l + ir_l •
(2.20)
It should be noted that after reflection, the direction of light propagation is reversed, traveling now along - z . According to eq. (2.1) this implies for right ( + ) circular polarized light a counterclockwise rotation both at fixed time and at fixed position looking along the light propagation direction. The sign of the Kerr rotation has been chosen such, that the ellipticity ei~ will be positive at a photon energy dominated by absorption of right circular polarization (selection rule (Am = +1, see section 2.2) as shown in fig. 2 and accordingly the rotation will be positive at lower energies and negative at higher energies. One should note at this point that there will be no Kerr rotation for a non-absorbing material while there may still be some Kerr ellipticity. This is easily seen rewriting eq. (2.16) as phase and amplitude zl±=arctan
(
2+~-~+ n__+
)
1 ' --
~/(n 2 + kZ+ - 1) 2 + (2k+) 2 Ir±l = (n_+ + 1) 2 + k~
221, (2.22)
As we will see later, this behavior is opposite to the Faraday effect just by the fact that the phase shift, which always leads to magneto-optical rotation, in reflection is proportional to the absorption index k, eq. (2.21), while in transmission it is induced by the refractive index n, eq. (2.39). In order to relate ~bi~and & one makes the ansatz 4~K = 0 K - - i e K
= +i
r + -- r
r++r_
(2.23)
MAGNETO-OPTICAL SPECTROSCOPY
143
After some algebra one finds, that compared to the exact formulas given by eqs. (2.19) and (2.20), the following approximations have been made in eq. (2.23): sin(A+ - A _ ) = ( A + - A_), cos(A+ -- A_)~--I and (r+ - r _ ) 2 ~ 2 r + r _ . Even for Kerr rotations of several degrees, these approximations lead only to minor errors in the determination of ~. Substituting the Fresnel equation, eq. (2.16), into eq. (2.23) we find the following relation between the complex indices of refraction and the Kerr effect, OK = - I m
---~1
eK = - R e ( ~+ - ff
) )
(2.24)
'
(2.25)
Now we use eqs. (2.14), (2.15) and (2.3) and write ~+ . (n_+ . . ik_+) . 2.
1
i 4 7 r o-+ _ ,
(2.26)
o9
where c7_+can be related to ~,x and ~ y using eq. (2.12), °'+- = ~ x x
-T- i O ' x y
(2.27)
.
It is straightforward to relate rT_+ and the conductivity tensor by inserting eqs. (2.27) and (2.26) into eqs. (2.24) and (2.25), --2
-2
87J" ~
n + -- n _ =
n+--n-
-- --
o9
_
rT+rT_ - 1
Oxy
(2.28)
,
~xy
E~xy
i6.~r7 = i(~'xx - 1 ) g - ~ x
"
(2.29)
In here we have used the approximation r7 = ½(if+ + rT_). Taking the real and imaginary part according to eqs. (2.24) and (2.25) yields,
EK ~
4 ~ [ Ao'lxy - BOr2xy ] o9 [ A2-+ B 2 ] '
--
oK=-,;- L ~ + B 2
J'
(2.30) (2.31)
with the coefficients A = (O'2xxn - O'lxxk ) 4"rr = (n 3 - 3 n k 2 - n), O9
B = (o'l~n + o'z~k ) 4"rr = ( - k 3 + 3 n 2 k - k ) . O9
(2.32) (2.33)
144
W. R E I M and J. S C H O E N E S
It is obvious, that the coefficients A and B in eqs. (2.30) and (2.31) are complicated and non-trivial functions of the indices of refraction n and absorption k. Figure 3 gives the signed logarithm of A and B as function of k for different values of n. It becomes clear, that only at singular points and for k < 1 and n >> 1 one term dominates but, in general, it is not possible to neglect one of the prefactors in eqs. (2.30) and (2.31). This makes clear that it is absolutely necessary in the Kerr configuration to measure all four quantities OK, eK, n and k to obtain reliable information on CT~yand hence to be able to relate magnetooptics to spin-polarized band-structure features what certainly is the ultimate goal. The final formulas to calculate ~Txyfrom these measurements are obtained from eqs. (2.30) and (2.31) to be o) O'lxy =
4~r [BOK + AeK] '
(2.34)
O'2xy
60 4--~ [ A O K
(2.35)
=
-- B e K ] "
The equations for the complex Faraday effect (i.e., ~bF = 0v - i e v ) can be derived quite a bit easier and more direct. Linear polarized light incident on a medium at z = 0 may be written as a superposition of two right and left circular polarized waves of equal amplitude according to eq. (2.1) E = ½E0[(e x + i e y ) exp{i(wt - 27rff+Z/Ao) } + (e x - iCy) exp{i(wt - 27rr7 z/A0)}].
(2.36)
This equation can be written as E = ½ E 0 exp{i(o~t- 2~ffZ/Ao))[e x cos½ c~+ ey sin½ c5"],
(2.37)
x Ioglxl 2
0
-2 -3 0
5
10
k
15
Fig. 3. Coefficients of 0~ and e K in the calculation of ~ y [eqs. (2.32) and (2.33)] as a function of n and k (after Reim 1986). Solid lines: x = 3 n 2 k - k 3 - k ; dashed lines: x = n 3 - 3 n k 2 - n. Note the logarithmic scale of the vertical axis and the inclusion of the sign of x.
MAGNETO-OPTICAL SPECTROSCOPY
145
where we have defined the complex phase = (iT+ - ff_)21rz A0
(2.38)
Now it becomes obvious (see fig. 2) that a wave which is linearly polarized at z = 0 will be elliptically polarized after propagating a distance l o with the major axis of the ellipse rotated through the angle tol° n ( i f + - iT_))=-~-c ( + - n _ ) ,
0F = Re(½ c~)= R e ( - ~
(2.39)
and the ellipticity, defined as the ratio of the minor to the major axis of the ellipse, will be / / ~'l 0 / tol 0 e F = - t a n h ( I m ( l c ~ ) ) = -tanh~Im~-~--(if+ - if_))) = t a n h ~ (k+- k)) = wl° (k+ - k ). 2c
(2.40)
In eq. (2.40) we have used the approximation tanh(x) = x, neglecting terms of the order of x 3 and higher. Within this approximation the Faraday ellipticity equals the magnetic circular dichroisrn. Equations (2.39) and (2.40) clarify the interchanged roles of n and k in the Faraday effect as compared to the Kerr effect [eqs. (2.19)-(2.22)]. While the Kerr rotation OK is proportional to the absorption index k, the Faraday rotation 0v is determined by the index of refraction n. To derive the expressions between Faraday effect and the conductivity tensor, we start with eq. (2.28), insert eq. (2.2) and calculate the real and imaginary part 8~r [(n2+ - n2) - (k 2 - k2_)] - 2i[k+n+ - k _ n _ ] = - ~ (O'lxy "Jr-iO-2xy ) .
(2.41)
With the average values n ½(n+ + n ) and k = l(k+ + k_) and the differences An=n+-n andAk=k+-k we w r i t e e q . (2.41) as 8~2[n An - k Ak] = - - - O-lxy ,
(2.42)
8~2[k An + n Ak] = - - trzxy,
(2.43)
to
tO
The combination of eqs. (2.42) and (2.43) leads to expressions of An and Ak as function of ~Ty and finally using eqs. (2.39) and (2.40) the Faraday effect as function of G~y is given by 2~'1o [ rt°'lxy OF
C
[
2 to
e F - - -c
~+
-
-
ktr2,y ] k2
,
+ n 2xy ] k
n2 + k2
J "
(2.44) (2.45)
t46
W. REIM and J. SCHOENES
The final equations used to calculate ~xy from the measured functions n, k, OF and e F are: C
o'lxy --
2,0.io [nOv - keF],
(2.46)
¢
o'2xy = q - ~ l o
[koF + n e F ] .
(2.47)
Equations (2.44), (2.45) and (2.30), (2.31) express the measured magnetooptical effects in transmission and reflection in terms of the conductivity fir, which is the quantity describing the material property for the interaction of radiation with a magnetized medium. It is now possible to compare the size of the magneto-optical effects in reflection and transmission in general for a given medium characterized by the complex tensor components ~'xx and cT~y. In view of applications, it is most interesting to compare the size of the Faraday effect, 27rd °'e + o'2~y IthFI = V~F + e 2 - , - c -- n + k -~ '
(2.48)
of a thin film with thickness d with the bulk Kerr effect 4~
t / O.lxy 2 _1_ IT2xy 2
o~ ~
A 2 + B -~
(2.49)
'
of the same material, .I
FI .
I~KI
. ood
2c
.A/-~-~ .B2 ~ n2 + k2
2 " r r d Vo- x
2 _ -27rd
+ O'2xx -
c
I xxl .
(2.50)
c
Now, the thickness d of the film has to be chosen properly, to be able to compare the two effects. We adjust d such that the transmitted intensity, including reflection losses but neglecting interference effects (Kessler 1963),
.:.0Ii+
., exp(
with the absorption coefficient K = 2tok/c, equals the bulk reflectivity R of eq. (2.17). This leads to the final result
[~bld -
wk log [1 + (k/n)2](1 - R) z
I~xl
1 ( R ) k 2 1) 2 + (2nk) 2 4k log [1 + (k/n)2](1- R) 2 ~ / ( n 2 -
(2.52)
which shows that the ratio of Faraday and Kerr effect is independent of the
MAGNETO-OPTICAL SPECTROSCOPY I
I
147
I
a) "5 tu
3
Index of Refraction n
o
1
2
3
1
2
3
4
5
I
I
6
7
I n d e x of A b s o r p t i o n k
Fig. 4. (a) Ratio of the total Faraday and Kerr signals according to eq. (2.52) for a material with optical constants n and k. Compared is the Faraday rotation of a thin film with a thickness such that the transmittance equals the reflectivity of the bulk material. For large absorption the transmitted intensity is always lower than the reflected intensity and I~FI/I~K[ = 0. (b) As fig. 4a, but excluding reflectivitylosses in the Faraday configuration,i.e., the transmitted intensity has been calculated using ! = I 0 exp (-Kd) instead of eq. (2.51). off-diagonal element O'xy and only depends on the optical constants n and k. In fig. 4a, I~bFl/l~l is plotted versus k for different values of n. For materials with low absorption k < 1, the Faraday effect exceeds the Kerr effect by a large factor because the film can be chosen quite thick to obtain equal intensities in transmission and reflection. For larger k, on the other hand, the absorption is high which leads to very thin films in transmission and causes large reflectivity losses in the Faraday geometry. Consequently, the magnitudes of Faraday and Kerr effects become similar and the Kerr effect finally dominates. However, if the reflectivity losses in the Faraday geometry are neglected, the Faraday effect is the better choice for almost all combinations of the optical constants, as shown in fig. 4b. This fact is used in optical storage applications via quadrilayer structures (Bell and Spong 1978) or by directly chosing the double Faraday rotation instead of the Kerr rotation (Martens and Voermans 1984).
2.2. Microscopic models to calculate The basis of magneto-optical effects in solids is, of course, the occurrence of transitions between the magnetically quantized electronic states. For not magnetically ordered materials, Schlitz (1936) has reviewed the atomistic interpretation of the Faraday rotation using the Z e e m a n effect. The applied magnetic field splits the atomic levels of the initial and final states characterized by the m o m e n t u m
148
W. REIM and J. SCHOENES
quantum numbers S-L-J into sublevels characterized by the now non-degenerate magnetic quantum numbers m. The selection rules for right and left circular polarized light require Am = + 1, respectively. Thus the resonance frequency will, in general, be different for the two circular polarizations leading to the so-called diamagnetic Faraday effect, which is characterized by a line shape as shown in fig. 5b for the most simple magneto-optically active transition between the atomic 2S1/2(S 1, L = 0, J = ½) and the 2p1/2(S = ½, L = 1, J = ½) states. This contribution will be independent of magnetization and temperature, in a first approximation. A strongly different, so-called paramagnetic, line shape occurs if the occupation of the Zeeman-split ground states is different. This leads to a temperature dependent term, the so-called paramagnetic Faraday rotation line shape (fig. 5c) which is proportional to the magnetization. An extensive review of the magneto-optics of nonmagnetic materials has been given by Buckingham and Stephens (Buckingham and Stephens 1966, Stephens 1970). It should be noted that from the sign of the Faraday-rotation it is not possible to decide whether the material is diamagnetic or paramagnetic because, in contrast to the magnetic properties which are determined by the occupied states only, for magneto-optics the quantum numbers of the final states are of equal importance. It has been shown by several authors (Hulme 1932, Argyres 1955, Cooper 1965, Voloshinskii and Bolotin 1964, Bennett and Stern 1965, Erskine and Stern 1973, Roth 1964) that the several orders of magnitude larger magneto-optical effects observed for magnetically ordered materials can not be explained by the Zeeman effect but need the introduction of spin-orbit coupling. The influence of spinorbit coupling has been considered as a perturbation of either the initial (Argyres =
n ,/'i
m = + 1/2
A 1'
I
LCP
m=-
i'--~ n_-n+
1/2
/1
RCP
2s~,2J
/~n+
n /'i
m =+ I / 2
/
.... -l".:n_ .
.
.
.
i
i
.
OJ m =-
LOP
/,/ f~l--
RCP
I, .1 -
a)
I/2
-
~
ta,;
~°
Fig. 5. Dispersion of the Faraday rotation for a 2SI/2----~2p1/2 transition in an applied field which removes the m = -4-½degeneracy. (a) Energy level scheme. (b) Equal intensity transitions leading to a diamagnetic line shape. (c) Unequal intensity transitions (i.e., different occupation of the m = + ½ ground states) with paramagnetic line shape.
MAGNETO-OPTICAL SPECTROSCOPY
149
1955) or the final (Hulme 1932) state energies of an interband transition, or as a modification of the initial state wavefunctions (Kittel 1951), which is of importance if the ground-state orbital momentum is quenched. It has been pointed out that, in general, magneto-optical effects will be proportional to ($i X r E ) " Pi ,
(2.53)
i
which is the spin-orbit energy of an electron with spin s and momentum p moving through the electrical field - r V inside the medium and finally summed over all electrons of the medium. This vector sum indicates that the magneto-optical effects are proportional to the product of magnetization Es i and spin-orbit energy. In nonmagnets or at temperatures above the magnetic ordering temperature, there will be as many electrons with up-spin as there are with down-spin if no external field is applied and Es i will be zero. With an applied field, on the other hand, the electrons in the vicinity of the Fermi surface have a nonzero average value of the spin due to Pauli paramagnetism and we have Es i proportional to the applied field. It is obvious now that, in general, the magneto-optical effects will follow M(T, H , . . . ) and the interchanging roles of M and H are clear. To illustrate these proportionalities, fig. 6 shows the determination of the magnetic phase diagram M(T, H) of UAs (Reim et al. 1984a). In fig. 7, the field dependence of the Faraday rotation of a EuSe thin film is displayed (Schoenes and Wachter 1977). In this case, of a material with no crystalline anisotropy (S ground state), shape anisotropy plays the dominant role because of the large
UAs ( 100 ) hoJ = 1,6eV -2
-a
J
. 1.5
J J J
B IT]
~__ ~_ ~ J _J ~ J _ ~
n." -1 I J
~
-0.5
~ ~ --
0
i 0
t 50
I
J 100
I
I0.0 9.75 9.5 9.25 9.0 8.5 8.0 7.5 7.0 6.5 5.5 4.5 3.5 3.0 2.0 I
150
Temperature [K] Fig. 6. Temperature dependence of the magneto-optical Kerr rotation of a cleaved UAs single crystal for applied fields up to 10 T (after Reim et al. 1984b). Different ferro and antiferromagnetic phases are observed in agreement with the magnetic phase-diagram determined by neutron scattering (RossatMignod et al. 1982).
150
W. REIM and J. SCHOENES
22 r~ T
g
"~o~ = 2 . 3 5 e V
r4
T =2.5K
,o t?_
/ Applied Field Ha E T 3 i
i
i
i
i
=
i
Fig. 7. Field dependence of the specific Faraday rotation Ov/d of a EuSe thin film (after Schoenes and Wachter 1977). The Faraday rotation is proportional to the applied field H~, until H a exceeds the demagnetizing field 47rMs.
demagnetizing factor of 47r for a thin film. For applied fields H < 4~-M s, OF will be proportional to H and will saturate around 47rM s. Based on the concept of proportionality to spin-polarization and spin-orbit coupling, Bennet and Stern (1965) and Erskine and Stern (1973) have developed expressions for ~xy which allow quantitative calculations for interband and intraband transitions.
2.2.1. Interband transitions The expression for the absorptive part of the off-diagonal conductivity o-2xycan be written as the sum of separate contributions from spin-up and spin-down electronic states
O'2xy(to) = O'2xy i, (tO) -Jr-O'2xy~,(tO).
(2.54)
The contribution for transitions between spin-up states a and /3 is given by (Bennet and Stern 1965)
2"n'2e2 ~] tr2xy ~ (tO) = 4hm2VtO ~p
[1(/3 t 1~+ t 1(3/ t )[2__ 1(/3 t 1~'- t [a t )12]
× 6(tO~t3 t - tO),
(2.55)
where the operators ~'-+ = Wx --- i~y are linear combinations of the kinetic momentum operator defined by
(h),
¢r = p + ~
× IFV(r),
(2.56)
M A G N E T O - O P T I C A L SPECTROSCOPY
151
where p is the momentum operator and s × VV represents the spin-orbit contribution. Within this model, spin-flip transitions are neglected although the spinorbit contribution to ~- could couple spin-up to spin-down states. In fact, spin-flip transitions seem to give the dominant contribution to the magneto-optical effects of Fe, Co and Ni, as has been shown by direct determination of & from spin-polarized band-structure calculations (Tawil and Callaway 1973, Callaway and Wang 1973). For a given interband transition a---~b, Erskine and Stern (1973) introduced the concept of the total weight, (2.57)
< 2 y>a- b = fa b 1 2xyl dw
of that transition as the total magneto-optic absorption due to transitions between all sublevels a and/3 of the states a and b at all energies. With the approximation that the only effect of spin-orbit coupling is the splitting of degenerate energy levels, eq. (2.56) can be written as ~r ± = itom(x +_iy) and using eq. (2.55), the total weight is given by
27r2ne2
( O'2xy) a--~b = n-----h--- ~
[to + l(x + iy) l 2 - oJ2nl(x
-
iy),~t~12l%.n,~n¢,
(2.58)
where a and/3 now may be viewed as the spin-orbit split sublevels of a and b, respectively. The joint spin-polarization o) is defined as
n~ 1' nt3 t -n~ $ n~ $ o-j= n~ 1' nt~ 1' +n~ $ nt3 $ '
(2.59)
where n~ and nt~ are the occupied initial and the unoccupied final states, respectively. If the atomic wavefunctions of all states a and /3 are known, (tr2xy) a~b may be calculated. However, up to this stage the concept applies only to atomic states. In solids, where in general the initial and final state are energy bands, some further assumptions have to be included. First, we assume that due to strong exchange forces the spin-orbit coupling lifts the degeneracy of the orbital quantum n u m b e r and the spin-orbit split atomic energy levels will be labeled by rnt. This assumption, of course, holds only in the limit of strong magnetization, i.e., T < T c. Second, we assume that the oscillator strength of the atomic transition a---~ b is unchanged in the solid but is spread out in energy uniformly over the bandwidth E B. With these restrictions in mind, we may use spherical harmonics YT' for the orbital part multiplied by a radial contribution u(r) for the initial and final state wavefunctior/s, and write
Z ([( YT'~ ~]x + iy] YT' )]2 _ ]( y7,~1 ]x - iy] YT' ) ]2)Rib, a¢¢
(2.60)
152
w. REIM and J. SCHOENES
where
R ab
is the radial overlap integral
Rab = fa,b UaUbr2 dr .
(2.61)
Figure 8a,b summarizes the final treatment of a transition between two band states. According to the assumption of spreading the spin-orbit levels over the band, one expects the Am = +1 and Am = - 1 transitions being shifted by the spin-orbit energy Eso. The total weight will show to be proportional to (Eso/EB)O) (Schoenes 1975). It is often helpful to compare the off-diagonal weight (o'2xy)a_,b with the corresponding diagonal weight (osx x )a---~bwhich is defined in an analogous way by ( °rlxx) a---~b--
2 ~2ne 2 h %b](blrla) [2nanb ,
(2.62)
where r is the dipole operator. While all optical transitions contribute to eq. (2.62), only those states with a net spin-polarization add to eq. (2.58) which is an elegant way to separate, e.g., magnetic f-states from non-magnetic p-states. It should be noted that, although unpolarized, exchange split states may contribute to the magneto-optical response (Erskine and Stern 1973, Schoenes 1975). This fact demonstrates the limits of the total weight concept and is only a manifestation
Im(Txy /
,
~\,j"
x
,XiU -"
t ~J
c)
a) Im~xy
Im°x I ~i, /""""~ b)
d) Fig. 8. Schemes of signal versus frequency w to calculate ~2~yfor band to band transitions (after Erskine and Stern 1973, Schoenes 1975). (a) Absorption of right- and left-hand circularly polarized light shifted relative to each other by the spin-orbit energy, shown for one spin-direction only. EBand Eso indicate the bandwidth and the spin-orbit energy, respectively. (b) Resulting off-diagonal conductivitytr2xydisplayed separately for both spin directions. The magneto-opticaleffect will vanish, unless the contributing electronic states are spin-polarized. (c) Situation for unpolarized but exchange split states. (d) Resulting magneto-optical signal ~2xy.
MAGNETO-OPTICAL SPECTROSCOPY
153
of eqs. (2.53) and (2.55) showing that the spin polarization of states at a given binding energy is of importance, and not the integrated value. Figure 8c,d suggests how this type of transition may be integrated into the present concept. The S-shaped off-diagonal absorption bands in o-2~r for up-spins and down-spins are shifted relative to each other by the exchange energy. The difference between these two spectra is the quantity which is to be compared to experimental results. If the initial state of, e.g., a fn___>fn-1 d transition is localized, atomic coupling effects of the fn-~d final state configuration have been shown to be important (Schoenes 1975, Reim and Schoenes 1981). Then, atomic coupling schemes have to be used to describe the structures of the magneto-optical spectra. Depending on the relative magnitude of the Coulomb and the spin-orbit interactions, either L - S or ]-j coupling, or an intermediate scheme, is appropriate.
2.2.2. Conduction-electron effects It has been shown (Erskine and Stern 1973, Reim et al. 1984a, Doniach 1966) that intraband transitions may contribute to o-2xy with two different frequency dependences, namely o-2xy~ 0) -1 and O-2xy~ w -3. The theory gives a proportionality of cTxyto the conduction-electron concentration N via the plasma frequency
2
~Op =
4~rNe2 m*
'
(2.63)
where e is the electron charge and m* is the effective mass of the electron, and to the spin polarization n~ - n ~ O-¢o.d- - , n t + n~
(2.64)
where n t and n+ are the numbers of conduction electrons parallel and antiparallel to the total moment. In addition, proportional to the strength of spin-orbit coupling via the macroscopic dipole moment P0 and the product of electron velocity ev o (Reim et al. 1984a)
O'xy(0"))~-- ~
O"cond -- ~-~2 ~_ ( y q_ io9)2 +
with spin moment a second term is ratio of maximum charge and Fermi
--eVo 1 - •2 + (,y + i o ) ) 2 j ]
.
(2:65)
Here, y( = 1/r) and J2 are the relaxation and the skew-scattering frequency, respectively. The absorptive part of the free-electron contribution then calculates to 2 Or2xy((O) = " ~ O'eond (~-~2 ..~ , 2 -- 032)2 "Jr 4"yaw 2 r 0 [
(.o'y(a 2 --~ V 2 -t- 60 2)
11
eVo (0£ +--~--;-5-)~ +-~y2oo2 j j .
(2.66)
154
W. R E I M a n d J. S C H O E N E S
It is important to note that the sign of the two terms is opposite. The decisive number for the dominance of the first or the second term in eq. (2.66) at a frequency to is the ratio ( le~0[) / ( ~ 2 2 ) .
(2.67)
This ratio can be estimated using the relation Po/eV o = (ma2tOso)/(htOa~b) where a is the lattice constant and htOsn and htOa~ b a r e typical spin-orbit and interband energies, respectively (Stern 1965). The first term in brackets in eq. (2.66) with a OVzxy~ tO-3 frequency dependence in the high-frequency limit, gives the same frequency dependence as the classical Drude treatment of free electrons. This term dominates if the damping is small enough to allow a coupled plasma frequency tOp well below the lowest energy interband transition. It may be viewed as a splitting of the plasma edge for the two circular polarizations (different tOp for right and left circular polarization) and may lead to very large magneto-optical effects if the plasma edge in the optical reflectivity is steep. In the classical Drude treatment, one starts with the equation of motion d2r dr m*~-~ + m*y dt
e dr c dt × H = e E .
(2.68)
Using the time dependence e i¢ot, the dielectric constant for (-+) circular polarizations 47rNer+ to2 = 1+ ~ - 1 + to(_to + tO~+ iv) '
(2.69)
is obtained with the cyclotron frequency to~ = eH/m*c. Using eqs. (2.3), (2.26) and (2.27), we derive ~
Orxy--
tO
~
(~+ -- E_)--
2 to P
47r (ito
toc +
,}/)2 _[_0)2
"
(2.70)
Formally, this equation is equal to the first term of eq. (2.65) except for the conduction-electron spin-polarization prefactor O'cond in eq. (2.65) if toc is identified with the skew-scattering frequency J2. Figure 9 displays the experimental Kerr effect of paramagnetic silver for an applied field of 1 T (Schnatterly 1969), which is a typical example of a Kerr signal due to plasma-edge splitting. All details of the energy dependence of 0K and eK can be successfully modelled using eq. (2.69) and eqs. (2.19)-(2.22) with the input parameters tOp = 9.2 eV, y = 0.04 eV, to~ = 0.11 meV at 1 T, and an unpolarized contribution e b from interband transitions as derived experimentally (Ehrenreich and Phillipp 1962). The reason for the relatively large Kerr effect of several
MAGNETO-OPTICAL SPECTROSCOPY T -o
4
155
r
Silver B=IT
_o
-E K
'~-2
5.5
4
4.5
Photon Energy [eV]
Fig. 9. Polar Kerr rotation OKand ellipticitye~ for paramagnetic silver in an applied field of 1 T (in part after Schnatterly 1969). The screened plasma frequency of silver is 3.8 eV. The Kerr effect originates from the plasma-edge splitting for (±)-circularly polarized light. 10 -3 degrees for a paramagnet is the fact that both A and B in eqs. (2.30) and (2.31) are small and B changes sign where eK peaks. As expected, O'xy is three to four orders of magnitude smaller than in typical ferromagnets. Obviously, large Kerr effects may be generated if toc is enhanced relative to top either by reducing tOp or by increasing tOe. The former concept leads to a Kerr effect of up to 60 ° at room temperature, e.g., for n-type InSb with a plasma frequency of tOp * = 52 meV (Palik et al. 1962). The second way towards large rotations by an increase of the magnitude of the plasma-edge splitting using a ferromagnetic material has only recently been realized (Reim et al. 1984a) and will be discussed in detail in section 5.3. Recently, it has been claimed (Feil and Haas 1987), that a resonance-like enhancement of the Kerr effect may be expected at photon energies where e l x x equals one, as suggested by eq. (2.29). A detailed investigation of the influence of the normal optical constants on the Kerr effect (Schoenes and Reim 1988) has shown, however, that the magnitude of A and B in eqs. (2.30) and (2.31) is of major importance. For a typical ferromagnet, A and B vary by up to two orders of magnitude between 0.5 and 5 eV which indicates that both A, B and tT~y strongly affect the Kerr effect and have both to be treated in a proper way to allow any predictions. Multi-layered metallic thin films consisting of one material with small optical constants and a second material with the desired magnetic properties are one way to modify the coefficients A and B defined in eqs. (2.32) and (2.33) and also to increase OK via eq. (2.31). First results, which are promising, show an increase in Kerr rotation by as much as a factor of 4 without substantial decrease of reflectivity (Katayama et al. 1988, Reim and Weller 1988, 1989).
156
W. REIM and J. SCHOENES
2.3. The Hall effect and the Kramers-Kronig equations The energy range for magneto-optical measurements is limited both on the lowand on the high-energy side. However, very often a low-energy extrapolation to zero energy is necessary to be able to separate interband from intraband transitions and to deduce a model band structure near the Fermi energy. This extrapolation can be done quite accurately for O'lxy by using the Hall effect (Doniach 1966). At to = 0, both o-2,~ and o-2,yvanish as can be shown by symmetry arguments (S.D. Smith 1969). The shape of o-2~yat low energies may be obtained using the Kramers-Kronig integral equations between o-l~y and o-2xy. In general, the electrical resisitivity is defined as the inverse of the conductivity, I~" = ~ - 1 ,
(2.71)
where both & and ~ are tensors. By calculating the inverse tensor of ~ and restricting to terms up to first order in magnetization, one finds for the diagonal element the well-known equation 1
O-lxx(to __ O) = --,
(2.72)
P
and for the off-diagonal element
O'lxy(to
= 0) --
0H2 ,
(2.73)
P where p and PH are the DC resistivity and the Hall resistivity, respectively. If the resistivity is given in units o f / 2 cm, the conversion factor to %xy i n units of s-1 is 9 x 1011. The real and the imaginary part of 6"xy obey the well-known Kramers-Kronig equations (Bennet and Stern 1965) ____2 p f0 ~ to,~--_ tot 2 °'2xy(to') d w ' ,
Cr2xy(to) = 2to P
f0
1
tot2 __ ('02 fflxy(to') do)',
(2.74) (2.75)
where P denotes the principal value of the integral. Although this integral extends from zero to infinity, contributions at to' far off a given frequency to are small due to the local character of the integrand and to changes in sign of tTxy. This fact makes eqs. (2.74) and (2.75) very useful for extrapolating O'xy below the lowenergy limit E 1 of the measurement. For this purpose, an extrapolation of Orlxy below E 1 and towards O-lxy(W= 0) given by eq. (2.73) is chosen in such a way, that eq. (2.75) reproduces the measured trZxy in the energy range above El, and vice versa. We have found that for most of the materials reviewed in this work,
MAGNETO-OPTICAL SPECTROSCOPY
157
the calculated ~xy in the energy range from E 1 to E 1 + 2 eV depends quite strongly on the chosen extrapolation, thus, giving reliable information o n O'xy between E 1 and zero energy. It should be noted that also the Faraday rotation and ellipticity are connected by a Kramers-Kronig integral similar to eqs. (2.74) and (2.75) (Smith 1976), which allows a fairly accurate determination of the complex Faraday effect if only one part has been measured (Schoenes 1975). However, there is no such equation for the Kerr effect (Kahn et al. 1969). A way to gain additional information in spectroscopy is the use of sum rules for the optical constants (Altarelli and Smith 1973). Here we only want to mention the relation ~ 47r (9
2 0"2ij((9)d(9 = -27r o-lq((9 = 0),
(2.76)
between the zero-frequency magnitude of O-lq (i.e., resistivity or Hall effect) and the frequency dependence of o-2q. This is an additional way to check an extrapolation of ~r and to obtain information on optical transitions not covered by the experimental energy range. Recently, an extra contribution to the Faraday rotation besides intraband and interband terms has been proposed arising from the Hall effect (Hartmann and McGuire 1983). This result was deduced from an experimentally observed proportionality of OF to the Hall angle (OH = arctan(pH/p ) = pH/p) for a selected series of similar amorphous alloys. However, in the light of the basic theory given above, the Faraday rotation at a given photon energy will always be proportional to OH for different materials if the conduction-electron contribution dominates. This holds as long as the energy dependence of the conduction-electron contribution t o ~xy is similar for the compared materials and their optical properties are not too different. Faraday rotation and Hall effect are manifestations of the same material property but at different frequencies and the proportionality may be deduced from eqs. (2.46) and (2.73) introducing a factor f((9, ~-) to take care of the frequency dependence, OH/p = 5 × 109f((9, ~') (nO F -- k e F ) .
(2.77)
3. Experimental considerations
3.1. Techniques To determine the four quantities ~xx a n d O'xy at a photon energy h(9, in general four independent measurements are necessary. In the special case when the value of~r at only a few discrete wavelengths of laser lines is of interest, the complex tensor may be determined by the measurement of the reflectance change, AI = I+ - I_ I+ + I _ '
(3.1)
of p-polarized light over a wide range of angles of incidence (Corke et al. 1982).
158
w. REIM and J. SCHOENES
Here, I+ and I_ are the reflected intensities for the two directions of in-plane magnetization, respectively. The advantage is the need of only one relatively simple experimental set-up which, in addition, is easily automated. However, if the investigation is devoted to examine ~r over a large energy range, say 0.5-6 eV, this intensity method is limited by signal to noise ratio and one is restricted to methods specialized to either optics or magneto-optics. To determine ~Txxthe simplest way, in general, is an absorption measurement. However, the reduction of the data to obtain the absorption coefficient K from the ratio of the transmitted to incident intensity can be much harder than eq. (2.51) suggests. The problems arise from reflectivity losses and interference effects (Kessler 1963, Heavens 1965). Furthermore, intraband transitions and fundamental interband transitions lead to values for K in the order of 105106 cm -1. According to eq. (2.51), the thickness of the samples therefore should be less than 200 nm. Thus, thin films are needed. For bulk samples like single crystals, reflection measurements have to be used to examine ~xx. Two procedures exist: in the first category, two measurements are performed at each energy, either two reflectivities different by polarization or angle of incidence or the direct measurement of amplitude r and phase A in eq. (2.16) (Muller 1969, Kolb 1972). In the second category, only one normal incidence reflectivity measurement is performed at each photon energy and the Kramers-Kronig integral R(o~') A(w) = _tOTrp f o l n~--'~ -~-~- dto',
(3.2)
is used to calculate the phase A from the reflected intensity R (Smith and Manogue 1981). The real and imaginary part of the dielectric function are then obtained with the relations, e~x~
( 1 - - R) 2 - 4 R sin 2 A (1 + R - 2V'-R cos A)2 '
(3.3)
'E2xx
4(1 - R)X/-R sin A (1 + R - 2X/-R cos/1) 2
(3.4)
The advantage of this second type of reflectivity measurement is the use of unpolarized light and normal incidence. The main disadvantage comes from the need to extend the measurement over an energy range as wide as possible. It has turned out that the photon energy range needed to make safe use of eq. (3.2), extends roughly from 10 meV to 25 eV, thus covering phonons, plasmons, interband and intraband transitions. For most of the rare earth and actinide compounds to be reported on in sections 4-7, the normal optical constants have been determined by this method and will be used to calculate ~ y from the magnetooptical measurements. To the problem of generating, focussing and detecting monochromatic radiation of various energies, magneto-optics adds the task of polarizing it and detecting very small changes in the degree o f polarization. Figure 10 illustrates an ex-
MAGNETO-OPTICAL SPECTROSCOPY
Polarizing Monochromator Prism ~
He- Cryostat
so,,,
~_-~
o-;;:o,
Superconducting Phase Faraday Magnet Shifter Cell Amplifier
A
....... ~
159
Light
~
Source
Oetec,or Analyzing Prism
Preamp.
m i_ t Audio Osc.
Lock - in Amplifier
Fig. 10. Experimental set-up for polar Kerr effect measurements in the 0.5-5.7 eV photon energy range (after Reim 1986). A typical measurement of a 0.5-5 eV spectrum with less than 0.01 ° unaccuracy takes about 60 min. with a fully automated device (Weller and Reim 1988).
perimental set-up (Reim 1986) which is, among others (Krinchik and Artem'ev 1968, Erskine and Stern 1973, Badoz et al. 1977, van Engen 1983), very well-suited to high sensitivity measurements of Kerr rotation OK and ellipticity e K in the energy range 0.5-5.7 eV at variable temperature and high magnetic fields. The light from halogen, xenon or deuterium sources is polarized with an air-gap Glan Thomson prism. Several samples and one reference mirror are mounted in a gas-flow cryostat allowing temperatures in the range 1.5-320 K and magnetic fields up to 10 T generated by a split coil superconducting magnet. The angle of incidence is kept below 5° to assure complete detection of the polar Kerr component (Metzger et al. 1965). The reflected light is polarization modulated with a Faraday modulator with either a heavy glass or a quartz rod for energies below and above 2.5eV, respectively. Using the Jones matrix formalism (Theocaris and Gdoutos 1979) it can be shown, that by setting the analyzer to 90°-6 relative to the polarizer one obtains at the detector an intensity I = ~I0[(1 - J0(2~b0) cos 2(0K - 6)) + 2Ja(2~b0) sin 2(0K - 6) × sin cot . . . .
cos oJt. • - ] ,
(3.5)
for a Kerr rotation OK and a modulation angle ~b = ~b0 sin tot, where Jn(X) is the nth order Bessel function. Thus, the difference of the analyzer settings for vanishing oJ-signal for light reflected from the reference mirror and from the sample, respectively, gives the Kerr rotation OK of the sample corrected for all other rotations due to windows and optical elements in magnetic stray fields. Both gold and platinum are suitable as a reference mirror, showing a Kerr effect of less than 0.002 over the whole spectral region in a field of 1 T. The Kerr ellipticity e K
160
W. REIM and J. SCHOENES
is measured in the same way as the rotation, but with a Soleil-Babinet compensator at 41-Aretardation mounted between sample and modulator. An orientation with the fast axis of the compensator parallel to the major axis of the light ellipse results in linearly polarized light (Senarmont principle) rotated by e K (Theocaris and Gdoutos 1979). It should be noted that for accurate measurements, the Kerr rotation of the sample has to be compensated for by setting the polarizer to -Oi~(hoo). The absolute error of this method depends on the amount of birefringence in the optical components in the light path and the accuracy in the parallel adjustment of reference mirror and sample. Values of less than 0.001 ° are achievable in the 1-4 eV energy range and complete automatization is possible with a typical measuring time of 60 min for a 0.5-5 eV energy scan (Weller and Reim 1988). The same experimental techniques can also be used for magneto-optical measurements in the Faraday geometry. However, due to the much lower transmitted intensity in the spectral region of fundamental absorption compared to measurements in reflection, the accessible energy range will, in general, be reduced.
3.2. Samples As discussed in section 3.1, absorption and Faraday rotation measurements most often require samples in the form of thin films. With the use of modern sputtering techniques or feedback-loop controlled multielectron-gun evaporation systems, thin film preparation of multi-component compounds is relatively easy (Gambino and McGuire 1986). Most magneto-optical data of rare earth (RE) transition metal (TM) compounds and of the RE chalcogenides, both in reflection and in transmission, have therefore been obtained on thin films using the advantages of high composition flexibility and perfectly flat surfaces. One serious problem related with Faraday rotation measurements on thin films is birefringence in the substrate induced by different coefficients of thermal expansion of film and substrate or by a large magnetostriction of the magnetooptical film (McGuire et al. 1987). A second major problem stems from corrosion of the magneto-optical film if the preparation chamber is not built for in situ measurements, which, of course, is the most elegant way to obtain a well defined surface (Erskine and Stern 1973). If some portion of the material is corroded, the film acts like a multi-layer system both for Kerr and Faraday measurements, but with unknown values of the optical constants and thicknesses. In general, this leads to faulty results. Performing the Kerr rotation measurement through the substrate, as it is frequently done to avoid this difficulty, generates a new problem, because the magnitude of the Kerr effect depends on the optical constants of the surrounding material (Choe 1989) as can easily be seen using the appropriate Fresnel equation in eq. (2.16) and calculating OK and e K via eqs. (2.19) and (2.20). For R E - T M compounds on quartz substrates e.g., the measured Kerr rotation through the substrate is, in general, enhanced by about a factor 1.3 in the visible frequency region. If a laser with high correlation length
MAGNETO-OPTICAL SPECTROSCOPY
161
like the HeNe laser is used as the light source, interference effects within the substrate cause additional alteration of the measured values. Hence, in situ measurements as has been done for Gd (Erskine and Stern 1973), or a careful transfer of the sample into the measuring equipment under inert gas atmosphere are necessary for corrosion sensitive materials. Similar corrosion problems arise if single crystals are used, which was the case for the optical and magneto-optical measurements on most actinide compounds and some RE-chalcogenides. However, single crystals have the advantage that very clean and optically flat surfaces can be prepared by cleaving. Therefore, crystals of these materials have been cleaved in the ultrahigh vacuum of the reflectivity equipment (Schoenes 1980b) or liquid helium in the cryostat for magneto-optical measurements (Reim 1986), whenever it turned out to be necessary. If bulk samples have to be polished to obtain an optically flat surface, besides the mentioned corrosion problems, the uppermost 100nm are mechanically altered by the polishing grains. In reflectivity measurements, this leads to strong scattering losses at wavelengths <400 nm, and consequently to a reduced reflectivity in the UV (Kfing and Schoenes 1980). In magneto-optical measurements, the energy dependence of OK and eK has been found to be, in general, only a little affected (Reim 1985). However, for materials where the electronic or the magnetic structure are determined by a delicate balance between different energy terms, like in intermediate-valence compounds or materials with a complex magnetic phase diagram like UAs, polishing leads to dramatic changes of the critical properties in the uppermost 100 nm. As a typical example, fig. 11 displays the temperature dependence of the Kerr rotation for cleaved and polished UAs single crystals (Reim et al. 1984b) showing a complete wash-out of the magnetic phases for the polished surface.
UAs (loo) • cleaved A polished B = 100 kG E=1.6
o=
eV
~-.5 t hID
Y
I
I
I
t
I
50
100
150
200
250
Ternperoture (K) Fig. 11. The influence of mechanical polishing on the magnetic phases of single crystalline U A s in the uppermost 100 nm, as probed by the magneto-optical Kerr effect (after Reim et al. 1984b).
162
W. REIM and J. SCHOENES
4. Light rare earths
The study of the rare earth or lanthanide (Ln) series is not obstructed by radioactivity, scarcity or short life times as it is the case for most of its 5f homologues. Nevertheless, many white spots in our knowledge of the magnetooptical properties of lanthanides still exist. Disregarding here data for single lanthanide ions diluted in large-gap insulators like CaF 2 (for examples see: Shen and Bloembergen 1964, Weakliem et al. 1970), which are nonmagnetic and therefore not of central interest for this chapter, the reasons for the limited number of studied rare earths can be summarized in the two following points: (i) In materials with crystal-field splitting effects, lanthanide ions with an even number of f-electrons will generally have a nonmagnetic ground state. As a consequence, many high symmetry structures with Pr 3+, Tb 3÷, Ho 3+ and Tm 3÷ are nonmagnetic down to very low temperatures. (ii) The binding energy of the 4fn states in the lanthanide series varies over many electronvolts (Lang et al. 1981). Since in most cases only dipole-allowed transitions will lead to large magneto-optical signals that can be resolved in a metallic sample, the 4fn---~4fn-15d transition energy should preferably be in the easily accessible photon energy range, say 0.5 to 5 eV. This excludes, for most magneto-optical measurements, Sm 3+ and Gd 3+ and puts materials on the basis of Nd 3÷, Dy 3÷ and Er 3+ near the high-energy limit. From these arguments, it becomes clear that the majority of magneto-optical investigations is dealing with compounds and alloys of (a) the light rare earth ions Ce 3+ and Nd 3+, (b) the half filled shell ion Eu 2+ and (c) the heavy rare earths Tm 2+ and Yb 3+. Exceptions are Kerr effect measurements using synchrotron radiation (Erskine 1976, 1977) which allows probing of deep lying 4f states as in Gd 3+. Other exceptions are the thulium monochalcogenides with trivalent thulium. In these compounds one observes large magneto-optical effects due to intraband transitions (Reim et al. 1984a). After the recent discovery of similar contributions in some UTe samples (Reim 1986), in NdS (Br/indle et al. 1988) and in Nd3S 4 (Schoenes et al. 1988), we anticipate that this origin of magneto-optical signals will occur in other magnetic and metallic lanthanide compounds, as well.
4.1. NaCl-type cerium, neodymium and samarium compounds As most lanthanides, cerium, neodymium and samarium form face centered cubic binary chalcogenides with the V I A elements of the periodic table of the elements (S, Se, Te) and fcc pnictides with the V A elements (N, P, As, Sb, Bi). Table 1 collects the lattice constant a, the type of magnetic order, the ordering temperature, Tc or TN, the ordered magnetic moment/Zora and the crystal-field splitting A between the ground state and the excited states. For the Ce compounds the 4fI(:Fs/2) free-ion state splits into a F7 doublet and a F8 quartet, the F7 state in
MAGNETO-OPTICAL SPECTROSCOPY
163
TABLE 1 Crystallographic and magnetic data for cerium and neodymium monochalcogenides and monopnictides, a is the lattice constant of the NaCl-type compounds. The ordered moment /~ordis taken from neutron scattering, except for neodymium chalcogenides where the data are from magnetization measurements in fields of 200 kOe. The crystal-field excitation energies ~ [K] are also preferentially taken from inelastic neutron scattering. For the Ce compounds the energy is between the ground state F~ and the excited state F8. For the Nd compound the ground state is F~2~and the excited states are F6 and F~1~. a (/k)
Magnetic Structure
CeS CeSe CeTe
5.778 [1] 5.992 [1] 6.361
AF-II [21 AF-II [31 A F - I I [3]
CeN [6] CeP [7] CeAs [7] CeSb [7]
5.018 5.932 6.078 6.412
CeBi [7]
6.487
no order AF-I AF-I 6AF phases ( r < TN/2 : AF-IA) AF-I (T < TN/2 : AF-IA)
NdS [81 NdSe [8] NdTe [8]
5.695 5.911 6.284
NdN [9] NdP [9] NdAs [9] NdSb [9] NdBi [9]
5.131 5.826 5.958 6.31 6.41
Tc,N (K) 8 5.4 2.2
/Xord (p~)
A (K)
0.57 [21 0.58 -+ 0.05 [31 0.3 -+ 0.1 [31
142 -+ 6 [41 100 [11 32 [5]
8 8 16
0.83 0.85 2.06
172 159 37
25
2.1
AF-II AF-II AF-II
9.3 10.5 10.6
2.24 (at 200 kOe) 2.34 (at 200 kOe) 2.58 (at 200 kOe)
232 208 150
F AF-I AF-I AF-I AF-I
27 11 13 16 25
3.1 1.83 2.18 2.98 3.14
56,467 34, 170 30,157 18, 103 37, 105
8
References: [1] Hulliger et al. (1978). [2] Schobinger-Papamantellos et al. (1974). [3] Ott ct al. (1979). [4] Furrer (1987). [5] Rossat-Mignod et al. (1985). [6] Kaldis et al. (1982). [7] H~ilgand Furrer (1986). [8] Hulliger et al. (1975). [9] Hulliger (1979).
m o s t cases b e i n g t h e g r o u n d state. T h e e x c e p t i o n is C e N , w h i c h a p p e a r s to h a v e an u n s t a b l e v a l e n c e ( B a e r a n d Z f i r c h e r 1977) a n d for which, to o u r k n o w l e d g e , n o crystal-field e x c i t a t i o n s h a v e b e e n r e p o r t e d . T h e 4f3(419/2) f r e e - i o n g r o u n d state of N d 3+ splits in an o c t a h e d r a l crystal field into states with s y m m e t r y Fs(1), F 6 a n d F8(z). T h e two crystal-field v a l u e s given for t h e n e o d y m i u m p n i c t i d e s c o r r e s p o n d to t h e s e p a r a t i o n o f F 6 a n d F8(2), r e s p e c t i v e l y , f r o m t h e g r o u n d s t a t e F8(1). W i t h t h e e x c e p t i o n o f t h e n i t r i d e s , all c e r i u m a n d n e o d y m i u m chalc o g e n i d e s a n d p n i c t i d e s o r d e r a n t i f e r r o m a g n e t i c a l l y in z e r o field. C e N s h o w s n o o r d e r d o w n to 1.5 K , w h i l e N d N o r d e r s f e r r o m a g n e t i c a l l y . T h e p a r t i c u l a r situat i o n of t h e n i t r i d e s follows f r o m t h e i r small l a t t i c e c o n s t a n t , giving a L n - L n s e p a r a t i o n v e r y s i m i l a r to t h a t in t h e p u r e L n - m e t a l s . T h e small b i n d i n g e n e r g y o f t h e 4f 1 s t a t e in C e 3+ results in a s t r o n g h y b r i d i z a t i o n with o t h e r states, a r e d u c t i o n of t h e m a g n e t i c m o m e n t a n d t h e a b s e n c e o f m a g n e t i c o r d e r . O n t h e o t h e r h a n d , in n e o d y m i u m t h e 4f 3 b i n d i n g e n e r g y is of t h e o r d e r 4 - 5 eV ( L a n g et al. 1981)
164
W. R E I M and J. S C H O E N E S
leaving also in NdN the 4f3 state localized, and allowing the occurrence of ferromagnetic order, due to the dominance of direct exchange to the nearest neighbors over the indirect exchange via the anions to the next-nearest neighbors. The antiferromagnetic ordering in the large majority of these compounds is not very appealing for magneto-optical studies. Yet, in large enough magnetic fields substantial magnetic moments can be induced and the puzzling question of the nature and the binding energy of the 4f state in cerium compounds has triggered an investigation of CeSbxTel_ x with x = 0, 0.75 and 1 (Reim et al. 1986). Figure 12 shows the Kerr rotation OK and the Kerr ellipticity eK for these three compounds in the energy range from 0.5 to 5 eV. At a measurement temperature of 2 K and an applied field of 5 T, CeSb is saturated with a moment of 2.06 p~/Ce (Rossat-Mignod et al. 1985) while CeSb0.ysTe0.25 and CeTe are in an intermediate spin state with, e.g., 0.71/xB/Ce in the latter material (Hulliger et al. 1978). The most striking feature is the value of the maximum Kerr rotation in CeSb. 0~ reaches 14° at the low-energy limit of the measurement which was 0.5 eV. This is the largest ever reported rotation angle for a single reflection in any material. From a Kramers-Kronig transformation of O'2xy and vice versa (see section 2.3), one can anticipate that the maximum rotation will exceed 15° near 0.4 eV. A second interesting feature is the shift in energy of the observed structures from 2eV in CeTe to 1 eV in CeSb0.vsTe0.25, t o ~ 0 . 5 eV in CeSb. Although, as a general rule, assignments to electronic transitions should not be made on the basis of OK and EK spectra, but on the basis of Orlxy and Or2xy spectra, this shift points to a decrease in energy of the responsible transition. This idea also gains support from the fact that if one normalizes the maximum rotation of CeTe and CeSb0.75Te0.25 to the magnetic moment in CeSb, the former two materials would display a maximum rotation of ca. 10°, which is not too far from the 15° occurring in CeSb. Figure 13 displays the real and imaginary parts of the off-diagonal conductivity of CeSb and CeTe as computed from the above Kerr spectra and the optical constants. Below 0.5 eV the spectra have been extrapolated in such a way that the Kramers-Kronig-transformed spectra fit the experimental data above 0.5 eV as 14
O K
CeSb B=5T T = 2K
OK
3 2
"510
CeSboz5Teo.zs B=6T T=2K
CeTe B= 5 T T=2K
OK
1
~n ,
k
8
0
6
-1
#_4
', -~K
-2
~x 2 E
',~EK ~T-7"-
,
.
,
,
ir ii d ii ii g
80
1
2
3
4
5
6
1
2
3
4
Photon Energy
5
6
I
I
I
5
4
5
6
[eV]
Fig. 12. Polar magneto-optical Kerr rotation OK and ellipticity e~ for [100]-oriented cleaved single crystals of CeSb, CeSbo.75Te0.25 and CeTe (after Reim et al. 1986).
MAGNETO-OPTICAL SPECTROSCOPY
165
4
o
CeTe
;
T=2K B=5T
*** z.× ,
02xy
o
2
* o a!
l
0 '(0
7D - 2
!
~n
', ~ 1 - ,
o
,
~t I
3
a
+~o 1 ~ y
~-4
CeSb
+ ÷
T=2K B=5T
-6 -8 i
i
i
2
3
4
6
5 Photon
0 Energy
1
2
(eV)
Fig. 13. Absorptive part ~rzxyand dispersive part O'lxy of the off-diagonal conductivity of CeSb and CeTe (in part after Schoenes 1987). well as the extrapolation of the c o m p l e m e n t a r y spectrum. O n e recognizes that O'lxy of CeSb peaks at 0.4 eV w h e r e also the diagonal conductivity (fig. 14) shows a p r o n o u n c e d shoulder. A f t e r substracting the free-electron contribution or elxx from the total o-ixx the remaining bound-electron-transition part exhibits a maximum at 0.4eV, which has the theoretically expected weight for an f~---~d ~
CeSb
~xx
o v
b'-I
0
0.5
10
1.5
2_0
Photon Energy (eV)
Fig. 14. Absorptive part o'1=~ of the diagonal conductivity of CeSb(+ + + +) at room temperature and f b a decomposition into intra- (crlxx) and interband (~rl~x) contributions (after Schoenes and Reim 1985).
166
W. REIM and J. SCHOENES
transition with localized f-electrons (Schoenes and Reim 1985). It is thus concluded that the fl ~ d ~ transition occurs in CeSb at 0.4 eV and that it gives raise to a so-called diamagnetic line shape (see also section 2.2), i.e., a dispersive behavior in the absorptive part of O"xy(=O'2xy)and an absorptive behavior in the dispersive part of tTxy (=o-lxy). As we will see throughout this chapter, the diamagnetic line shape is the common energy dependence for the f---~d transition in lanthanide and actinide compounds. It is the result of at least two subtransitions with inverse dominances of absorption for right- and left-hand circularly polarized light. At 1.4eV, both o-ix~ and O'e~y show a maximum indicating a transition with a so-called paramagnetic line shape. Presumably, this is the onset of transitions from the valence band, which at its top has strong admixture of d-character and also some admixture of f-character. Coming to CeTe, fig. 13 shows that the large negative peak in Orlxy is shifted to 2 eV. At the same energy, o-1~~ displays a barely resolved structure which is, however, of the correct size for an f---~d transition. A similar result is found in CeS with an excitation energy of 2.6 eV (Schoenes 1984). Photoemission gives binding energies of 2.4 eV in CeS and 2.5 eV in CeTe (Gudat et al. 1981). Thus for Ce-Chalcogenides optics, magneto-optics and photoemission come to the corresponding conclusion that the f-state is localized with a binding energy in the order of 2 eV. The situation is definitely more complicated in the cerium monopnictides. In CeP, CeAs, and CeBi, valence-band photoemission displays two maxima, one near 0.5 and one near 3 eV binding energy, which have been attributed to 4f electron excitation on the basis of their intensity dependence of the incidentphoton energy (Franciosi et al. 1981). Many controversial interpretations have been put forward to explain this double structure and the historically interested reader may consult the proceedings of the various conferences on valence fluctuations held in Rochester (1976), Santa Barbara (1981), Z/irich (1982), K61n (1984) and Sendai (1985). We will not repeat here the various arguments favoring one or the other proposal, but only mention the two lines along which the double structure is interpreted today. The first model, which we may call the localized or many-body model, starts from a localized 4f level interacting with the electrons of a conduction band. A Kondo resonance is formed at the Fermi energy in addition to the bare 4f level 2-3 eV below E F. The relative intensities of the peaks in the photo-emission spectra are determined by the ratio of the hybridization width and the energy of the level (Gunnarson and Sch6nhammer 1983, Sch6nhammer and Gunnarson 1987). In the second model, which we may call the itinerant model, one starts from a band-structure calculation, and the f-states lie at the Fermi energy. This is the ground state of the hybridized system and it is taken to correspond to a well-screened f hole, i.e., a final fl state (H/ifner and Steiner 1982). The second peak is obtained by a calculation in which an f hole is created at a central atom in a larger unit cell and only conduction electrons are allowed to relax and screen the 4f hole (Norman et al. 1985). Thus, for large enough hybridization both models come to the conclusion that the peak near the Fermi energy corresponds to the ground state and the high binding-energy peak is merely a result of the photoemission process. It is then interesting to note that the
MAGNETO-OPTICAL SPECTROSCOPY
167
magneto-optical spectra of CeSb exhibit only one 4f peak and that its energy of 0.4 eV is an excellent measure of the ground-state energy of the 4f state. The second trivalent rare earth element in the series of NaCl-structured lanthanide compounds which is suitable for magneto-optical studies is neodymium. Compounds and alloys of this element have received more attention recently, when it was more widely recognized that the addition of terbium or dysprosium to iron or cobalt reduces the magneto-optical signals (see section 6). To help understanding the magneto-optical response of Nd 3+, the polar Kerr effect has been determined in a concentrated, but simple, magnetic system like NdS (Br/indle et al. 1989). Figure 15 displays the Kerr rotation and ellipticity spectra of a [100] cleaved NdS single crystal at 15 K in a field of 100 kOe. The rotation spectrum can be decomposed into two parts. In the neighborhood of 3 eV one observes a diamagnetic line shape very similar to that found for the first time in TmS and TmSe by Reim et al. (1984a), and which will be discussed in section 6. The energy of this structure corresponds to the plasma energy in NdS and the size of the rotation is similar to that found in TmS. Therefore, the interpretation of the structure follows the same lines as for TINS, for which quantitative fits of OK, eK, trlxy and Or2xy have been performed. These are based on the classical description of the optical and magneto-optical properties, as the sum of Lorentz oscillators for the interband transitions and a Drude term for the intraband transitions. It results in structures with the observed line shapes for OK and eK near the coupled plasma energy. However, to fit the large rotation and ellipticity quantitatively, the cyclotron resonance frequency entering the classical formula has to be enhanced by two orders of magnitude. This can be accounted
4
NdS (100)
3 "v o
T = 15K H = 100kOe
2
-OK
,..I.-
HJ
1 0 - - ~ . ~
O 13_ X
f
-1
JI
-2
E o q)
I
;J -3 -4
,
0
i
,
i
i
i
,
i
i
i
1 2 3 4 5 Photon Energy (eV)
Fig. 15. Polar magneto-optical Kerr rotation OK and ellipticity eK for a [100]-oriented cleaved single crystal of NdS at 15 K and in an applied field of 100 kOe (after Brfindle et al. 1989).
168
W. R E I M and J. S C H O E N E S
for by a microscopic theory for the scattering of conduction electrons in a magnetic material (Reim et al. 1984a). The second structure in fig. 15 is a broader peak near 4 eV which, as a preliminary computation of trxy shows, persists in the latter spectra and presents all the characteristics of an f--->d transition. The last light rare earth to be discussed in this section is Sm 2÷. Suryanarayanan et al. (1972) have co-evaporated Sm and Te onto heated substrates of pyrex, vycor or CaF: to produce thin films of SmTe. The compound formed is nonmagnetic, as is expected from the 7F0 ground state of Sm 2÷. It is also nonconducting, in agreement with the simplest ionic model. Samarium monochalcogenides attracted large interest in the beginning of the seventies, due to the discovery of pressure-induced semiconductor-metal transitions (Jayaraman et al. 1970). In the metallic phase, samarium was shown to be in a valance state intermediate between Sm 2÷ and Sm 3+ (see also section 6.1). The magnetic circular dichroism spectrum reproduced in fig. 16 (Suryanarayanan et al. 1972) is at normal pressure, i.e., for samarium in an integer divalent state. In the studied photon energy range from 2.5 to 4 eV, one observes two S-shaped structures which have corresponding shoulders in the absorption spectra. From a detailed analysis of the temperature dependence of the circular dichroism peaks, the authors derive that the responsible transitions are 4f6(7F0)-->4fS(6H)5deg and 4f6(TF0)--->4fS(6F)5deg. The corresponding two transitions into the 5d t2g subbands are assigned to two absorption peaks at 1.2 and 2.0 eV. These assignments have been corroborated by an investigation of the variation of the absorptionpeak energies as a function of pressure and temperature in the related compound SmS (Batlogg et al. 1974). Of course, an interesting question is whether Sm 2+ is a suitable ion for obtaining large magneto-optical signals. Although the data shown in fig. 16 are not normalized, a comparison with similar data for EuSe (Ferr6 1974) allows the estimation that the magneto-optical response per Gauss applied field is a factor 1000 smaller than in EuSe, which, as we will see in Section 5.1, is an excellent
t
~'~ 2
t
.....~oo K
~--~0 .......
- -
1.5~
.............
~0
111 a -2
/
f_J
~
/- - ~
11
-J
).57-o_ 0
PHOTON ENERGY (eV)
Fig. 16. Optical density K d [eq. (2.51)] and magnetic circular dichroism AD of a SmTe thin film vapor-deposited onto C a F 2. T h e M C D was m e a s u r e d for various temperatures in a field of 35 k O e (after Suryanarayanan et al. 1972).
MAGNETO-OPTICAL SPECTROSCOPY
169
magneto-optical material. In the spirit of what we said at the beginning of this section, the weak magneto-optical response for a nonmagnetic ground state is expected and, therefore, compounds with Sm 2÷ have not been investigated much further.
4.2. Th3P4-type compounds Many lanthanides also form compounds of the Th3P4 structure and its anti-type, the Gd4Bi 3 structure with the V A and V I A elements. It is also well-known that the normal Th3P4 structure is stable with holes in the 8-coordinated sites, leading to compositions between L n a S 4 and L n 2 S 3 (Hulliger 1979). Thus, neodymium forms semiconducting Nd2S 3 (Henderson et al. 1970) as well as the metallic N d 3 S 4 (Schobinger-Papamantellos et al. 1977) with the same cubic T h 3 P 4 s t r u c t u r e . N d a S 4 is especially suited for magneto-optical studies of Nd 3+, because it orders ferromagnetically at (47 --- 1) K (Schobinger-Papamantellos et al. 1977). Its ordered moment at 4.2K has been determined to be (2.08_+ 0.06)/zB. This reduced value, as compared to the free-ion value of gJ = 3.27/z B is ascribed to crystal-field splitting. Figure 17 shows the polar Kerr rotation and ellipticity measured on a single crystal of N d 3 S 4 held at 10K and exposed to an external field of 40kOe (Schoenes et al. 1988). One observes two maxima in the rotation 0K(tO), one very narrow peak reaching a rotation of 2.3 ° is located at 0.8 eV, and a second broader peak with 0.5 ° rotation is centered at 4 eV. To discuss the origin of these peaks and those in the ellipticity eK(tO) we need to know the off-diagonal conductivity.
Nd~S 4 -eK
B =4 T T =1OK
80 E
0
-1
I
I
I
I
I
1 2 3 4 5 Photon Energy [eV] Fig, 17. Polar magneto-optical Kerr rotation OK and ellipticity %: for a polished single crystal of Nd3S 4 at 10 K and in an applied field of 40 kOe (after Schoenes et al. 1988).
W. REIM and J. SCHOENES
170 100 r
80
Nd3$4
60
20
I
0
I
I
2
I
I
I
I
I
I
I
4 6 8 Photon Energy [eV]
I
I
10
I
12
Fig. 18. Near-normal incidence reflectivity of Nd3S 4 at room temperature. The surface is a polished [100] plane (after Studer and Schoenes 1977).
Figure 18 displays the near-normal incidence reflectivity of the same material (Studer and Schoenes 1977) which is needed to compute the diagonal and, together with 0K and eK, also the off-diagonal conductivity. The reflectivity indicates that Nd3S 4 is a metal with a plasma edge of about 1 eV. The dielectric function and the energy loss which have been calculated by Kramers-Kronig transformation of R(to) give for the coupled plasmon energy htop a value of 0.75 eV. This agrees with the low-energy structure in OK and eK, suggesting that the latter are to be associated with the coupled plasmon. A close comparison with TmS, TmSe and Tm0.sEu0.sSe (section 6.1), for which a similar effect has been observed for the first time (Reim et al. 1984a), shows that -O K has a positive maximum at tOp * in N d 3 S 4 and Tmo.sEuo.sSe and a negative extremum (minimum) at tOp in TInS and TmSe. This difference can be explained by the expected difference in the spin polarization of the conduction electrons. For Nd 3÷, L = 6, S = ~ and J = 9, thus the spin moment of the f-electrons is antiparallel to the total moment. The spin moment of the conduction electrons couples parallel with the 4f-spin moment leading to a negative spin-polarization of the conduction electrons relative to the total moment. In Tm 3+, with a more than half-filled 4f shell, L, S and J are parallel, so that the conduction electrons will have a positive spinpolarization. The case of Tm0.sEu0.sSe is complicated by two antiferromagnetically coupled sublattices, of which the Eu 2+ sublattice has the larger moment, and therefore determines the net magnetization direction. Figure 19 exhibits the off-diagonal conductivity spectra of N d 3 S 4. The dominant structures now occur in the 4 eV range, where OK showed a second pronounced peak. These structures are assigned to an f3___~f~5d transition. From fig. 19 one derives that for this transition Orlxy has an absorptive and Or2xy has a dispersive line shape. The transition energy is then ~4 eV, and is determined by the extremum in
MAGNETO-OPTICAL SPECTROSCOPY
171
t\
5
t
o ',
Nd3 $4 B= 4 T T =1OK
,/ / I °-zxY
0
._>
/
e0
!
e-
o -5
121 I
o -10
1
2
3
4
5
Photon Energy [eV] Fig. 19. Real part o-lxyand imaginary part ~r2xyof the off-diagonal conductivity of Nd3S" at 10 K and in an applied field of 40 kOe (after Schoenes et al. 1988).
o-lxy. This agrees reasonably well with a binding energy of the 4f3 level in Nd metal of 4.6 eV (Lang et al. 1981) and with the assignment in NdS (section 4.1). Below 3 eV two further, but much smaller, peaks appear in O-2xy. The peak at 0.8 eV falls together with oJ~ and might be an artifact, produced by the multiplication of strongly varying functions in this energy range. The peak near 2 eV, on the other hand, has its counterpart in O'lx~ and should, therefore, be considered real. Its assignment to an electronic transition is not as straightforward as for the p e a k at 4 eV. However, two explanations can be offered: (i) a transition of d-electrons into empty f-states. Such transitions occur in uranium compounds and they have the same positive sign. Also, the energy of 2.2eV is compatible with the energy of 1.8 eV for the 4f 4 peak observed by bremsstrahlung isochromat spectroscopy in Nd metal (Lang et al. 1981). Against this interpretation speaks the observation of a peak at similar energy in o-1~~ of La3S 4 (Studer and Schoenes 1977), for which the 4f 1 state is expected at only 5.5 eV. Also, the intensity is much larger than expected for a transition into localized f-states considering the results for uranium monochalcogenides (section 7.2). (ii) a transition from the top of the p-band into empty d-states. This interpretation is compatible with the Orlxx(¢-O) data for La3S 4 and may also give some basis for the observed sample-, i.e., stoichiometry, dependence of the intensity of the peak in O-lx~(w). The relative importance of the signal in O'2xycompared to 0-1~~ necessitates f-character to be admixed to the involved p-states in Nd3S 4.
172
W. REIM and J. SCHOENES
4.3. Intermetallic compounds The potential of thermomagnetic writing and magneto-optical read-out has prompted quite early investigations of intermetallic compounds and alloys of rare earths and 3d transition elements. Thus, after Nesbitt et al. (1962) reported a Kerr rotation of 2 ° in SmCos, several RCo s intermetallic compound films with R either Y, Sm, Gd or Nd, were studied by Stoffel and Strnat (1965) and Stoffel (1968). They found that the off-diagonal component of the complex permittivity tensor generally decreases with increasing rare earth content and they did not confirm the large Kerr rotation of SmCo 5. More recent work has been performed by Gambino and McGuire (1985, 1986), Suzuki et al. (1987), Choe et al. (1987), Choe (1989) and Weller and Reim (1989a,b). While the two first authors, as many others, quote only values for a single laser wavelength and find a maximum Kerr rotation of 0.59 ° at 1.96 eV in Pr0.2Fe0.4Co0. 4 (Gambino and McGuire 1986), the latter have investigated the spectral dependence. Figure 20 reproduces the Kerr rotation spectra for polycrystalline Co and various compositions of amorphous
5 i 0.5
Energy 5 I
4 I ~-Co
Thin
(eV)
2 I
Film
/
°~o~
~
~
_
o
0.4
/
0,5
FY 19C0.81
. .....~,~o
---m:T
0.2 o?..ol.dSo o
-~
0
"(29C°.71 I = I
,i
i,..
0.4 0..5 0.2
I
I
I
I
I
I
Nd.t9Co.81
o 0.5
"5 n,-
= I
I
"~-:~---o . . . ... . . .. . . . ~-~
Nd.24C° 76 J
~.......~-~" o- - ~ 2 " - - . __ Ce.Zl C°-79 N d.29C0.71 -"~--=---o --_._~__ .
.
.
.
.
.
" .
.
.
.
0.1 0
I 300
I
I I I = I 400 500 600 Wavelength (nm)
I 700
Fig. 20. Absolute value of the polar Kerr rotation at room temperature and an unspecified field for various compositions of amorphous Y-Co, Ce-Co, Nd-Co and polycrystalline Co films (after Choe et al. 1987).
MAGNETO-OPTICAL SPECTROSCOPY
173
L n - C o films, with Ln being Ce and Nd (Choe et al. 1987). For comparison, also shown, is the effect of alloying Y instead of Ce or Nd. One recognizes, that except for lower Nd-concentrations, the general trend is a decrease of the Kerr rotation at room temperature. The main reason for this behavior is presumably the decrease of the magnetic moment at room temperature. Choe et al. (1987) find that for T = 300 K the magnetization decreases by a factor of approximately 2 when the Ce, Nd of Y content increases from 15 to 30 at%. Nearly the same reduction is found for the Kerr rotation at 633 nm (=1.96 eV). Therefore, if one normalizes for every sample the Kerr rotation by its value at long wavelengths, one can obtain an idea of the spectral contributions of the rare earth atoms to the total Kerr effect. Figure 21 (Choe et al. 1987) displays such normalized spectra, which should now also be compared with the pure Co spectrum (fig. 20). Apparently, Nd leads to an increase of the Kerr rotation near 4 eV. This is the same energy at which Nd3S 4 shows a maximum (fig. 17), which, from an analysis of the diagonal and off-diagonal conductivity, could be assigned to an f3___~f2d transition. Because such an intra-cation transition is not expected to be very much dependent on the compound formed, we suggest that also in NdxCol_ x the peak near 4 eV is due to an f3__, f2d transition (Choe 1989, Weller and Reim 1989a,b). The Kerr effect of alloys of two light rare earths with iron is shown in fig. 22 (Suzuki et al. 1987). As in NdxCOl_ x an increase of the rotation with decreasing wavelength is observed between 800 and approximately 400 nm. For still shorter wavelengths, Suzuki et al. find a zero crossing, which, however, has not been observed by others (Weller and Reim 1989a,b). Very recently, the wavelength dependence of the Faraday rotation has been studied for a variety of amorphous Nd-Co, Nd-Fe and N d - A u thin films (McGuire et al. 1987). Figure 23 shows the results for the Nd-Co system Energy 5 1.5 - I
4 I
/ ~ . ~
__
_
E
5 I
(eV)
~
2 I
Nd.29 Co.71
".
Nd,24Co,76
~.~'~'- - ~--"~'~--~...-..~ "" -. j Nd 19 Co 81 " ' ~ . ' ~- Ce.28Co.72
•
0
'~
I
~00
.
I
/
I
400
I
I
I
500
Wavelength
I
600
I
I
TO0
(nm)
Fig. 21. Kerr rotation spectra for various amorphous Y-Co, Ce-Co and Nd-Co films, normalized by their Kerr rotation values at 700 nm (after Choe et al. 1987). The temperature is 300 K, the field has not been indicated.
174
W. R E I M and J. S C H O E N E S I
~ 0.1 v
I
I
I
I
I
I
o
~ -o.i ,~ -0.2
I \
-0.3 _~ -0.4
g
Nd38 Fe62
\~
~
I
t
Pr4oFe60 I
I
I
I
I
400 600 800 Wovelength X (nm) Fig. 22. Polar Kerr rotation at room temperature and fields of 7 kOe for amorphous films of NdsaFe62 and Pr40Fe60 (after Suzuki et al. 1987).
I
I
/ 0 0~,0 0
-I-++~
~
I
I+
I I NdloC090 0 Nd15C085 • Nd25C075 • NdsoC%0 0 Nd75C025 &&A&
¢ ÷,÷o/ \,-
4
",,.....
% ~ 3
L,t j~'/ •
,,~
/
÷,+ 'o OOooo+++o +*'...... +,***
O ++
~ 2 o Z
\
l d
\
T=20K B=32kG 0 0.0
I 0.5
I 1.0
I 1.5
I 2.0
I 2.5
I 3.0
I 3.5
4.0
Photon Energy {eV)
Fig. 23. Faraday rotation for amorphous Nd-Co films at 20 K and a field of 32 kOe (after McGuire al. 1987, and Reim and Gambino 1986)•
et
(McGuire et al. 1987, Reim and Gambino 1986)• With increasing Nd content, one observes a strong decrease of the rotation at low energies and the disappearence of the peak at 1 eV, which is present for 10 and 15 at.% Nd. On the high-energy side of the spectrum, one finds with increasing Nd concentration first the development of a peak, which then for higher Nd concentrations, shifts from about 3 eV for 25 at.% Nd to 1.5 eV for 75 at.% Nd. McGuire et al. (1987) assign this peak to either a 4f--->5d transition or a transition from below E F to an empty
MAGNETO-OPTICAL SPECTROSCOPY
175
4f state above E v. The energy shift of the peak with Nd concentration is ascribed to a shift of some 4f levels relative to E v. In the light of the Kerr data on N d - C o , NdS and Nd3S 4 and these off-diagonal conductivity spectra of Nd3S4, we would rather expect to find signs of 4f 3~ 5d transitions near 4 eV, which, unfortunately, is outside the range of the Faraday rotation measurements. A peak at 2 eV appears to be compatible with a 5d ~ 4f 4 transition in Nd. When alloying with Co, conduction electrons are transferred from Nd to Co, lowering E v and increasing the 5d---~4f4 transition energy. A shift of 1 eV is conceivable, considering the width of the occupied conduction-band part of approximately 2 eV in Nd (Lang et al. 1981) and observed shifts of the same order of magnitude in the systems C e - P d , T h - P d and U - P d (Schoenes and Andres 1982). This assumption, however, remains speculative as long as the off-diagonal conductivity spectra are unknown. The same is also true for the Faraday rotation data on N d - F e and N d - A u . Figure 24 displays the spectra for the latter system which is the simpler one, because it only contains one type of magnetic ion. One observes one peak near 2 eV, which roughly scales with the Nd content and which slowly shifts to higher energies when the A u content increases from 40 to 60 at%. Both effects are compatible with the assumption of a transition of conduction electrons into an empty 4f 4 state, but a definitive assignment has to await a computation of ~r y(O~) or ~:xy(o~). i
i
i
i
i
i
e,leeee
\ % 3
,"%
i
•
Nd4oAU6o
•
Nd60Au40
\
%
1 T=IOK B=32kG
0 0.0
I
I
I
I
I
I
I
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
P h o t o n Energy (eV)
Fig. 24. Faraday rotation for amorphous Nd-Au films at 10 K and a field of 32 kOe (after McGuire et al. 1987). 5. Half-filled shell rare earths
Although this section deals only with the two ions Eu 2+ and Gd 3+, it is certainly the richest in physics of the three sections on lanthanides. This results merely from the unique role that the europium monochalcogenides play for our basic understanding of magnetism of localized electron systems, as well as the many
176
W. REIM and J. SCHOENES
different magneto-optical effects which these materials display in a unprecedented clarity and magnitude.
5.1. NaCl-type europium compounds Table 2 collects some of the crystallographic, optical and magnetic data of the europium monochalcogenides (EuX, X = O, S, Se, Te). The Eu-monochalcogenides are model systems for isotropic Heisenberg exchange. The ground state is the spherically symmetric spin-only state 8S7/2. The nearest-neighbor exchange integral J1 is positive for all EuX-compounds, but decreases with increasing lattice constant. The next-nearest exchange integral -/2 is positive in EuO and negative in EuS, EuSe and EuTe. The two positive exchange terms in EuO and the net dominance of J1 over J2 in EuS lead to ferromagnetic ordering in these two materials. In EuTe, the negative J2 clearly dominates the positive J1 and antiferromagnetic order of the MnO type occurs below 9.6 K. In EuSe the absolute values of J1 and J2 are similar and the different sign of both terms results in a complex magnetic phase diagram. In zero field, EuSe enters an antiferromagnetic phase below 4.6 K. This phase consists of ferromagnetic (111) planes which are stacked in a + + - - sequence (Griessen et al. 1971). We call this phase AF-IIA in analogy to the AF-IA phase in which ferromagnetic (100) planes are stacked in a + + - - sequence and to the AF-II phase in which (111) planes are stacked in a + - + sequence. Below 2.8 K, EuSe consists of a mixture of an antiferromagnetic phase with order +-+and a ferrimagnetic phase with a stacking + + - + + (Fischer et al. 1969). The ratio of ~3 between the two phases determines the observed spontaneous magnetic moment of about 1.7/.~/Eu. Finally below 1.8 K, an MnO type, AF-II ordering occurs in zero field. Already, for fields of a few hundred Oe, ferrimagnetic phases are induced. Fields above 2 kOe lead to a ferromagnetic-like alignment of the moments (Griessen et al. 1971). In EuTe, the critical field for spin alignment is approximately 7 0 k O e (Oliveira et al. 1971, Schoenes and Wachter 1977). TABLE 2 Lattice constant a, magnetic structure, ordering temperature and exchangeparameters for europiumchalcogenides. The last column lists the gap energies at 300 K
EuO EuS EuSe
a (•) 5.141 [1] 5.968 [1] 6.195 [1]
Magnetic Structure F F AF-IIA ferri AF-II AF-II
T~,c (K) 69.33 [2] 16.57 [3] 4.60 [1] 2.8 [4] 1.8 [4] 9.58 [I]
J1/kB(K)
J2/kB(K)
0.55 [5] 0.22 [6] 0.11 [5]
0.15 [5] -0.10 [6] -0.09 [5]
Eo (eV) at 300 K 1.12 [7] 1.65 [7] 1.80 [7]
EuTe 6.598 [1] 0.06 [5] -0.20 [5] 2.00 [7] References: [1] Schwob (1969). [2] Kornblit et al. (1973). [3] Berkner (1975). [4] Griessenet al. (1971). [5] Zinn (1976). [6] Bohn et al. (1981). [7] Wachter (1972).
MAGNETO-OPTICAL SPECTROSCOPY
177
The energy gaPs as determined from re-emission and reflectivity measurements (Wachter 1972) increase with increasing lattice constant. This is contrary to the expectation for a p (anion)--~ d (cation) transition with increasing covalency, and points to an f--~d transition. As we will see, magneto-optical spectroscopy corroborates this suggestion. The first magneto-optical spectra have been published by scientists from IBM (Greiner and Fan 1966, Suits et al. 1966, A h n and Suits 1967). Although they found very large signals, the quality of some of the samples has been improved since then, and even larger signals and more structured spectra have been obtained. In addition the spectral range has been considerably extended and we will concentrate our discussion on these more recent investigations. Systematic studies of the Faraday rotation and the magnetic circular dichroism of EuS, EuSe and EuTe films have been published by Schoenes (1975) and Ferr6 (1974), respectively. Figure 25 shows the absorption and the Faraday rotation spectra as determined on polycrystalline EuS films in a field of 11.5 kOe for temperatures below as well as above T c (=16 K) (Schoenes 1975). At 8 K and 11.5 kOe the film is not magnetically saturated. Nevertheless, the Faraday rotation reaches a maximum value at 4.3 eV as high as 1.5 x 106 deg/cm. The presentation of the absorption and the Faraday rotation on the same energy scale demonstrates very nicely the larger resolution of magneto-optical spectroscopy and allows a rather straightforward assignment of the structures to electronic transitions. In fact, as
i
i
I
i
EuS I ~4 V 0
.-~' 500K 6K
---
/y z//_
~
_ H = 11,5 kOe 5
:~ 2 /
"\\~ ,
2
5
4
~
- - 22K
5
15
gin0
-.-
H
~o o ~
8K
= 11,5
- -
2
kOe
-
5 4 Photon Energy (eV)
~.~,:-~-r,.---
5
Fig. 25. Absorption (top) and Faraday rotation (bottom) of a polycrystallineEuS film at various temperatures and an applied field of 11.5 kOe (after Schoenes 1975).
178
W. R E I M and J. S C H O E N E S
we will argue quantitatively shortly below, magneto-optics discriminates p--~d against f--~ d transitions, while in absorption p--~ d transitions are giving the larger signals. Thus, fig. 25 indicates that f-~ d transitions occur near 2.1 and 4.3 eV, while p--* d transitions start around 3.5 eV. For substrate temperatures exceeding 500°C, EuS, EuSe and EuTe form (100) textured films on fused quartz (Schoenes and Wachter 1977). The magnetization and the Faraday rotation of these films saturates at 3 K for fields of about 30 kOe [EuS, EuSe (see fig. 7)] and 80 kOe (EuTe). Figure 26 shows the Faraday rotation of a (100) textured EuS film for magnetic saturation (Schoenes 1979, 1987). Also shown is the Faraday ellipticity e F or magnetic circular dichroism as has been computed from a Kramers-Kronig inversion of the Faraday rotation spectrum. Note that the sign of this eF spectrum has been reversed relative to the original to conform with the sign convention used in this chapter (see section 2.1.2). One recognizes a maximum rotation of 2.5 × 106 deg/cm which is, together with similar values for EuSe and EuTe (Schoenes 1979), the largest Faraday rotation of any material. The spectral dependence of the magnetic circular dichroism computed from the Faraday rotation can be compared with that measured by Ferr6 (1974) for films deposited on CaF 2 (fig. 27). The agreement is quite good, and is even better if Faraday measurements at similar small fields as for the circular dichroism are used (Schoenes 1975). Also shown in fig. 27 is the linear magnetic dichroism and the optical density of the used film. Faraday rotation spectra for EuSe and EuTe films are displayed in fig. 28 (Schoenes 1975). Note again that these spectra are not for magnetic saturation and that, at magnetic saturation, the rotation near 2.4 eV reaches 2.2 × 106 deg/ cm in both compounds (Schoenes and Wachter 1977). The similarities between
~2
EuS T = 2.5K H = 100 kOe
"u
%
,,',, /',,
1
/, I
I
I
~
',j
e
L
0
I
V
- - a . _ _
2 3 Phofon Energy (eV)
J
4
5
Fig. 26. Specific Faraday rotation OF and magnetic circular dichroism e F of a (100)-textured E u S film at magnetic saturation, e F is c o m p u t e d from 0F using the K r a m e r s - K r o n i g relation (after Schoenes 1979, 1987).
MAGNETO-OPTICAL SPECTROSCOPY x 10 -2 _
,..~ 2_ t_ '
x 10:
t~o~
o
112KK
.-~ a 3,5
i~,,,' - - ' ~
-,
-1
7K
179
I
c~
A
H~O,72T
H~O,O44T -
s,4 K
xlO -2-
0~5
-2 2 ¢'~
~
~10"3(~ ',
1 --
~ __._17,5K ~ ..... 2 6 K
--6
K
2oK
3
--
2
cn Q
o
i.............. ~
o
=~
H=0,0171 !IB J / --1 _ ~L,'e"C ~
.1
H : 0,14T
0,8 .
6 K
-.
_._2o~
-0,6
"~
(:3 b a;
I "
c i
- 0,4 _ 0,2
;
0 "Z-.
o 2
3
4
i 5
Photon Energy (eV) Fig. 27. Linear magnetic dichroism (top), circular magnetic dichroism (middle) and absorption (bottom) of an EuS film vapor-deposited onto CaF z. Note that the spectra are not normalized by the field nor by the film thickness and that the sign of the magnetic circular dichroism is opposite to the definition in this chapter (after Ferr6 1974).
the three compounds are quite obvious. The Faraday rotation spectra are dominated by two sets of structures which come closer to each other in energy when going from EuS via EuSe to EuTe. This reflects the decrease of the crystal-field splitting of the 5d final states with increasing lattice constant. Each set of structures consists of three or four single structures, which can be interpreted in the framework of atomic coupling schemes between the excited 5d-electron and the remaining 4f6 state (Ferr6 1974, Schoenes 1975). Before we do so, let us discuss Kerr effect measurements on single crystals which undoubtedly show the best spectral resolution. Figure 29 displays the longitudinal double Kerr rotation of EuO and EuS single crystals in an applied field of 2.5 kOe and temperatures of approximately 12 and 8 K, respectively (Greiner and Fan 1966). The photon energy range extends over the low-energy group of structures in EuS, but lacks one of the structures in EuO. The saturation field for the longitudinal effect is smaller than for the polar effect, due to the much smaller demagnetizing factor, and the authors claim that fields less than 2 kOe were sufficient to saturate the samples. The maximum single rotation is 1.3 ° in EuO and 0.8 ° in EuS. Considerably larger rotation angles have been reported for polar Kerr measurements.
180
W. REIM and J. SCHOENES 15
J
I
I
I
EuSe
A
10.
-
-
o
_
4,5K
-
,2 K
-
If",\
i
ii_,
i_.).,_;p_,,,
_
u'~-5 ,..2 "7 E
i
i
I
I
o v ~.
2
5
4
5
i
I
I
I
2
6
---9 K
EuTe i".
_
--15
u. 1
K
i H=ll,5kOe
i
o
__f! /
-t
~
l
i
i
~'~kq I'
-2
I
I
2
i "
VI 5
I
3 4 Photon Energy (eV)
6
Fig. 28. Faraday rotation of polycrystalline films of EuSe (top) and EuTe (bottom) at various temperatures and an applied field of 11.5 kOe (after Schoenes 1975). 3.0
2.0 ILl ILl n-" £.9 IJ.I
o
st
t
/
~l~EuO
1.0
4i
o
£
'
EuS
I J ~'X'I ~
~ i
~ l
on,.- -I.0 n~ n,, lull
v -2.0
-3.0
\
I
I
400
i
I
t
500
I
600
i
I
700
i
I
800
.,,'
i i i
900
I000
WAVELENGTH (m/.~) Fig. 29. Longitudinal double Kerr rotation of EuO and EuS single crystals in an applied field of 2.5 kOe and temperatures of 12 K (EuO) and 8 K (EuS) (after Greiner and Fan 1966).
MAGNETO-OPTICAL SPECTROSCOPY 8
I
i
i
/~°K
~6
L~
i
4
i
EuO
t] a}""--'~"~'K /i/ '~i
EKI'
6
H=40kOe T=|OK
I
I
o.
-8~
EuSe
A
4
H = 40 kOe
i] 1
2
,I
T1.~i = K A
y--",.-.,,"
-~0 E g-4
181
",./ V 1
2
5
4
5
I
2
3
4
5
Pholon Energy (eV)
Fig. 30. Polar Kerr rotation Ox and ellipticity eK of EuO (left) and EuSe (right) for [100]-cleaved single crystals at 10 K and 40 kOe (after Wang et al. 1986, Hiisser 1983, Schoenes and Kaldis 1987). Figure 30 shows results of such measurements extending also over large enough energy ranges to include both sets of structures for E u O (Wang et al. 1986) and EuSe (Hfisser 1983, Schoenes and Kaldis 1987). While a temperature of 10 K is sufficiently low to saturate the magnetization of E u O in a field of 40 kOe, this is not the case for EuSe. Figure 31 shows the effect of varying temperature on the polar Kerr rotation of EuSe (Hfisser 1983). Besides an increase of the maximum Kerr rotation up to +9.7 ° at 2.4 eV if the temperature is decreased to 2 K, one observes shifts and splittings of the structures. The most prominent shift occurs on the low-energy side, known as the magnetic red shift of the absorption edge (Wachter 1972). From fig. 31 one derives a shift of 0.19 eV for the lowest energy peak compared to a shift of 0.13 eV which has been derived for the absorption I0
r /~
5
/ /t~ ~ 4 - 0
$-5
T= 2K
T 1"-'EuSe kOe
tO
~0 ~8 hi
i~'l~ '
I
--- T='% l / -I0
1.75 2.0 2,25 2.5 Photon Energy (eV) Fig. 31. Polar Kerr rotation of EuSe for various temperatures and a field of 40 kOe (after H/isser 1983, Reim 1986, and Schoenes and Kaldis 1987).
182
W. REIM and J. SCHOENES
edge by re-emission measurements on powders (Wachter 1972). From either the Faraday rotation and the circular dichroism spectra (Schoenes 1975) or the Kerr rotation and ellipticity spectra (Schoenes 1987, Schoenes and Reim 1985), the off-diagonal conductivity spectra have been computed. Figure 32 shows the results of the latter for EuO and EuSe. One ascertains that the Faraday rotation of EuSe is very similar to -O-xxy, as expected for insulating materials. The similarities between 0K and O'2xy , o n one hand, and eK and Orlxy , o n the other hand, are less pronounced, due to the fact that the reflectivity is a function of both n and k (see section 2.1.2). The off-diagonal conductivity spectra have been used to make detailed assignments of electronic transitions to the various structures (Schoenes 1975, 1987). Figure 33 shows the atomic coupling schemes drawn to scale with the experimental data of fig. 32. Following a proposal of Kasuya (1972), the various interaction terms of the coupled 4f65d I final state are considered in order of decreasing magnitude. The largest term is the crystal-field splitting of the 5d states, which is found to be 3.7, 2.2, 1.9 and 1.7 eV in the series oxyde, sulfide, selenide and telluride. The next term is the exchange term between the excited 5d electron with SO-- i and the remaining 4f 6 state with Sgp = 3. This term amounts to 0.59 eV in EuO, and 0.43 eV in EuSe, the variation going parallel with the decreasing red shift of the absorption edge. The third term included in fig. 33 is the spin-orbit splitting between the total spin of the 4f65d 1 states and the angular momentum of the 4f 6 state. The resulting J terms have been drawn to obey the rule that two consecutive levels, J and J + 1, are separated by ifJ, where ~ is a spin-orbit interaction constant. The vertical arrows indicate the allowed transitions and the energies given are those of the peaks of O'2xy in fig. 32. For electric dipole-allowed transitions, the selection rules are AS = 0 and AJ = 0, - 1 . For spin-flip transitions, the selection rules are AS = - 1 and AJ = 0. The latter transitions become partly allowed, due to the admixture of states with ST = 7. The fit gives 55 meV in EuO and 46 meV in EuSe for the spin-orbit coupling constant ff (Schoenes 1987).
4f
2I-
0
~
o
~', I-
~t
~:)-1
-2~
-2 i! 1/
H= 40kOe T= 10 K
I! il
-ff
--'"
.:Xo~o~
l
,
I ~ il,
17, !
E,,Se
i"
,
/~,,
, h ,,,,
!i II ,t,"i
',J
"
/ q' "1
I/1ti '
'"J
.... V!i
ii
T: 1OK
',I ~,
-5 I
2
5
I
4
Pholon Energy
2
3
4
5
(eV)
Fig. 32. Real part O'lxyand imaginary part o'2xyof the off-diagonal conductivity of EuO (left) and EuSe (right), as computed from the data in fig. 30 and the optical constants (after Wang et al. 1986, and Schoenes and Reim 1985).
MAGNETO-OPTICAL SPECTROSCOPY 10Dq
183
ST=Ssd+S4f6d=S'T+I~4f6 (Fa} .
.
.
.
11/2
.
9/2
~
ST=Ssd+S4f6
J=ST+L4f6
(Fj)
9/2
7/2 5/2 5/2 I/2
5d "~
IIII
5d ~ 4f 6
10 Dq
5/2 1112 312 I12
. . . . ~(~--------T
/--
ls/~
5d 1 4f 6
1/2:
\ J .....
9/2 7/2
~ ST: 5
TI/2
S :
.
.
.
.
~,~
,,=
11/~ 5/z
------~3/2 9/2
~
---
-.,~
~;~
I/2
9/2 I/2 7/2
EuSe 1/2
4f 7
(8S7/2)
4 f ? (8S7/2)
Fig. 33. Atomic coupling schemes for EuO and EuSe and observed 4fv---~4f65d transitions and energies (after Wang et al. 1986, and Schoenes and Reim 1985).
From the earlier Faraday rotation measurements on films (Schoenes 1975), ff is derived to be 51 meV in EuS, 4 1 meV in EuSe and 32meV in EuTe. The single-ion value in t h e 4 f 6 ( 7 F ] ) configuration of Eu 3+ diluted in LaC13 is ~'io, = 30 meV. (Dieke and Crosswhite 1963). Thus, the europium chalcogenides constitute a quite particular class of materials. The initial state has no spin-orbit splitting and is unfavorable for obtaining large magneto-optical signals. On the other hand, it is 100% spin polarized with the largest possible pure spin moment. The necessary spin-orbit splitting is contributed by the excited state, making these compounds nice examples to show that magneto-optics is also an excited state spectroscopy. Another related final-state effect has been reported for the antiferromagnetic compounds EuSe and EuTe. Figure 34 shows the Faraday rotation of EuTe for various fields and temperatures in the energy range of the 4f7--->4f65d t2g transition (Schoenes 1979). For the lower field of 20kOe one observes with decreasing temperature the growth of a-peak P3 at 2.75 eV. This growth is opposite to the decrease of magnetization of EuTe for T < TN = 9.6 K and for H smaller than the critical field for spin alignment Her = 70 kOe. The same peak P3 disappears again if the field exceeds the critical field, as can be seen from the spectra at 100kOe. Hence, the size of this peak is intimately related to the antiferromagnetic order, and it has been shown that for T < TN the magnetic circular dichroism spectrum can be decomposed into two similar shaped parts,
184
W. R E I M and J. S C H O E N E S
2o
2.6 5
- :\
EuTe 9K~.~
b) H = IOOkOe
I0
o
L~ o
.......~.J// ". 7\ ~ "
-IC
P,
Pz
a) H=2OkOe
\
P~
/
K
&
-2
P~
2Ill
2J2 213
2.4
2!5
2!6
2!7
2:8
Photon Energy [eVl Fig. 34. Faraday rotation of a [100]-textured film of EuTe for fields of 20 and 100 kOe and various temperatures above and below T N = 9.6 K (after Schoenes 1979).
indicating the splitting of the 5d t2g sub-band into two further sub-bands. The latter splitting is attributed to the formation of the antiferromagnetic superlattice. Since 2.75 eV is also the energy of the structure which has been assigned to a spin flip transition in the atomic model, it is interesting to search for a relation between the atomic and band models. Schoenes (1975) has pointed out that the superlattice splitting of the conduction band corresponds to the exchange splitting between the two possible spin configurations of the final 4f65d state. In contrast to the exchange splitting in a ferromagnet, the superlattice splitting in the antiferromagnet occurs only at the superlattice zone-boundaries and both sub-bands contain spin-up and spin-down states. A last result from the Faraday rotation measurements of EuX compounds which we would like to mention here is the decomposition of o-2xy(~O) into single components (Schoenes 1975). Figure 8 gives a scheme of the expected single components in a band-like model. For an f t --~ d t transition from a completely spin-polarized f-state, a o~2xy(W) spectrum in the form of a lying S is expected (fig. 8b). For p--~ d transitions from unpolarized p-states into exchange split d-states a double-peak structure as shown in fig. 8d is expected. Figure 35 shows a decomposition of the experimentally determined o-2xy(~o) spectrum of EuS into two f--~ d like and two p--~ d like components. The integration of the positive part of a single component over the photon energies is half of the weight (o-2xy> as defined in eq. (2.57). The obtained experimental values have been compared with a computation using eq. (2.58) and atomic f and d wave functions. Good
MAGNETO-OPTICAL SPECTROSCOPY
185
1 1.80x I029sec-2
I 2.2 ~ 1029sec'2 ~
,~ 0.65~i029sec-a
0.9 x I0 29 sec-2
.~-./~///A ~
o2, Io",ee'
o
C I -4
3
4
5
eV
6
Photon Energy (eV) Fig. 35. Decomposition of the absorptive part tr2~yof the off-diagonal conductivity of an EuS film in components as expected for f-->d and p-->d transitions (see fig. 8). The numbers are the integral values of the positive parts of -~r:xe.
agreement is obtained for the f--> d t2g and the f--->d eg transition, corroborating also from a quantitative point of view the assignment of the dominant magnetooptical signal near 2 and 4.2eV to transitions of the f electrons into the crystal-field split d t2g and d eg sub-bands. Besides these classical magneto-optical measurements, various other magnetooptical effects have been studied in pure EuX-compounds. We mention here magneto-reflectance (Feinleib et al. 1969, Pidgeon et al. 1969, Scouler et al. 1969, G/intherodt 1974) thermo-reflectance (Mitani and Koda 1973), electro-reflectance (L6fgren et al. 1974) and magneto-absorption (Busch et al. 1964, Wachter 1972, Freiser et al. 1968, Schoenes 1975, 1979, Schoenes and Nolting 1978, Mitani et al. 1975, Llinares et al. 1973, Ferr6 1974). All these experiments can be interpreted within the same model of the electronic structure of the europium chalcogenides, but they often give less complete information. So we abstain from presenting them here in more detail, except for a beautiful example of magneto-reflectance. Figure 36 shows the data obtained by Scouler et al. (1969) for single crystals of EuS and EuSe. If we take the difference of the reflectance spectra at 1.5 K for right- and left-hand circularly polarized light of EuSe we obtain a spectrum which is quite similar to the Kerr-ellipticity spectrum at 2 K and a same field of 40 kOe (fig. 30). However, one notes also that the shoulder at 2.3 eV in the Kerr spectrum is not reproduced and that at energies around 4 eV the resolution of the differential Kerr effect is much higher. The last measurements which we would like to mention, although only shortly, because they are single-wavelength data, are Faraday rotation measurements to study critical exponents near the Curie temperature of EuO (Huang et al. 1974, Huang and Ho 1975) and EuS (Berkner 1975).
186
W. REIM and J. SCHOENES '
' I EuS 1.5*K H, 4 0 kOe
1.4
I
'
I
'
E2
(-~ 1.2
1.0' A
.= 0.8 ,
~'0.6 .Q
~
CrL
/
O"R
0.4
bJ U Z I~ 1.4 uJ
,
~'x
~--~
,
I
I
]
I
I
I
'
I
'
I
i
I
i
EuSe
1.5° K
H = 40 kOe
EI
1.2
E2
1.0 0.8 0.6
L
0.4 1.0
I
I
2.0
I
i
3.0
I 4D
, 5.0
PHOTON ENERGY (eV)
Fig. 36. Reflectance spectra on cleaved single crystals of EuS (top) and EuSe (bottom) for right- and left-hand circularly polarized light. The field is 40 kOe and the temperature is 1.5 K (after Scouler et al. 1969). Right- and left-hand circularly polarized light is interchanged compared to our definition.
5.2. Doped NaCl-type compounds The very large magneto-optical effects of the EuX-compounds (GdX compounds have not been studied so far and should have smaller effects due to their antiferromagnetic order and the large f---~d transition energies) has lead to considerable efforts to try to increase the rather low ordering temperatures to above room temperature. Unfortunately, these efforts were not successful to that measure, but interesting findings regarding the dependence of the magnetic exchange and the magneto-optical signal on the dopants have been made. Since europium is divalent in the EuX-compounds, the substitution of europium by trivalent rare earths will create "free" 5d conduction electrons which can enhance the ferromagnetic coupling via an indirect exchange mechanism of the Ruderman-Kittel type. This possibility has been realized as early as 1964 (Holtzberg et al. 1964) and, at the beginning, has prompted mostly magnetic and transport studies (von Molnar and Methfessel 1967, von Molnar and Shafer 1970, Methfessel and Mattis 1968). The trivalent rare earth commonly used is Gd 3÷, due to its 4fT-configuration being identical to Eu e÷. Tc of up to 135 K have been reported for EuO doped with 3.4 at.% Gd 3+ (Sharer and McGuire 1968). Yet, it has been realized that the local 4f spin of the dopant contributes little to an
MAGNETO-OPTICAL SPECTROSCOPY
187
enhancement of the ordering temperature, but on the contrary, may couple antiparallel to the Eu 2+ spin (McGuire and Holtzberg 1971). Therefore, efforts have been made to increase the free-carrier concentration by replacing the divalent anion by a monovalent ion like I + (Vitins and Wachter 1973). A third class of dopants and composite layers are transitions metals and particularly iron (Lee and Suits 1971, Almasi and Ahn 1970). The effect of Gd 3+ doping on the magneto-optical properties of EuO single crystals has been investigated by Kaldis et al. (1971) and Schoenes and Wachter (1974). A remarkable result of their absorption measurements was the discovery that the magnetic red shift of the absorption edge is substantially reduced upon doping. As fig. 37 shows, the total red shift of pure EuO amounts to AE = 0.26 eV and it follows a spin correlation function (dashed curve). For a free-carrier concentration at room temperature of 6 × 1020 cm -3 the red shift has decreased to AE = 0.12 eV, although the ordering temperature has increased from 69 to 115 K. Various interpretations have been put forward to explain this effect. Schoenes and Wachter (1974) have argued on empirical grounds that the simultaneous increase of T c and decrease of AE indicates a sub-linear decrease with increasing free-carrier concentration of the exchange Hex = - 2 ~'n Jn( r - R n ) s " S . where s is the spin of a conduction electron, S~ are the spins of neighboring Eu 2÷ ions, and J . ( r - R ~ ) is the distance-dependent exchange constant between electron and ion 0.9
~" "~'~• --°~o~o~L
I S0mpe
~ \ >~1.0 - - ' ~ t - ' ~
--'---•--
Nfree c0rrier [i0~9cr~3J
0 2
< 0.1 6.6
g_ 1.1
..... -
/am(T)d-r'\ " ~ . . ~ . , ° ~ o ~ - ~ - - - Y-
Urn(O)[~m(T)dT%,~I.~,=.~,___i. -
_
~
1.2 -- am= mogn.pertofthe exponsioncoeffieie~ 0 50
'
100
~
•
~
_
',,
_
150 200 Temperoture (K)
Fig. 37. Temperature dependence of the absorption edge of pure EuO and EuO doped with various amounts of gadolinium. The given carrier concentration N~r~ carrier is the conduction-electron concentration determined at room temperature by a combination of optical and magneto-optical data. The photon-energy values given here are for a constant absorption coefficient (~0), which otherwise may vary from sample to sample. The dashed line shows a fit of the shift of the absorption edge with the spin-correlation function Um (after Schoenes and Wachter 1974).
188
w. REIM and J. SCHOENES
spins. Nolting and O16s (1981), using a spectral density approach to compute the temperature dependence of the quasi-particle density of states for the conduction band, argued that the red shift depends on the band width and that the temperature-dependent filling of the conduction band causes the red shift to decrease. Schoenes and Wachter (1974) have also used Faraday rotation, absorption and reflectivity measurements to derive the three transport parameters of the free carriers, i.e., their concentration, mobility and effective mass, as a function of temperature. With this electrode-free method they could show that, for not too large doping, the activation energy of the Gd donors decreases on magnetic ordering while the mobility increases. Similar measurements have also been performed for EuTe doped with iodine (Vitins and Wachter 1973). The doping is found to introduce ferromagnetic clusters which give rise to a large wavelength-independent Faraday rotation. Figure 38 shows the field dependence of this wavelength-independent rotation and
I 3000
--~.
y
EUTeoI: T=8°K
]
2000
g o
o I000
06
o.u.
0
I
2
4
6
'.
4
,;
8
Applied Field
H
I0 I>-
WOe
/~
Applied
Field H
I
i:>
,e
Fig. 38. (top) Field dependence of the wavelength-independent Faraday rotation for two EuTe crystals doped with iodine at 8 K. (bottom) Field dependence of the magnetization at the same temperature of 8 K for the same doped samples and for pure EuTc (after Vitins and Wachter 1973).
MAGNETO-OPTICAL SPECTROSCOPY
189
compares it with the magnetization of the same two doped samples and a pure EuTe single crystal. One recognizes the rapid saturation of the frequencyindependent rotation, while the magnetization should only saturate above 70 kOe. An analogous violation of the proportionality to the magnetization is also observed for the temperature dependence of the wavelength-independent Faraday rotation. The authors suggest that the wavelength-independent Faraday rotation is due to magnetic resonance at lower frequency. The effect of doping with iron has been studied on Euo films by Ahn (1970) and McGuire et al. (1971) and on EuS films by Ahn and Sharer (1971). The Curie temperature was found to increase to 200 K for EuO with 8% Fe by weight and to 45 K for EuS with 16% Fe by weight. The magnetization decreases somewhat more than expected from the dilution of the Eu 2+ lattice. The f---~d t2g transition energy is shifted rather little as compared to pure films, while the absorption coefficient is enhanced by a broad background, originating from iron. Figure 39 displays the Faraday rotation measurements of McGuire et al. (1971) which were performed at 6 K in a magnetic field of approximately 20 kOe. The maximum rotation value is depressed from 8 × 105 deg/cm for pure EuO to 6 × 105 deg/cm for the iron-doped sample. The magneto-optical hysteresis loops of Fe-doped EuO are reported to be square-like for temperatures extending almost to the Curie points (Ahn 1970). The physical origin of the important increase of T c on doping with Fe is not completely clear, because it considerably exceeds the effect of adding free electrons. Suggestions are that either the presence of Fe ° around grain boundaries causes a kind of exchange coupling, as has been observed between films of permalloy and EuO (Almasi and Ahn 1970), or that some of the Fe 2+ ions present form (EUO)l_x(FeO)x, where FeO has a N~el temperature of 190 K. The chemical stability of EuO films doped with iron and europium has been investigated by Street (1972). The effect of Eu, Ag and Cu on the ferromagnetic exchange in EuO films has also been studied by Lee and Suits (1971). For the latter two nonmagnetic dopants, they find Curie temperatures of approximately
tO X 10 5
i
,6K,
-5'
4,000
6,ooo
I
i
~ooo
i~ooo
WAVELENGTH
Fig. 39. Faraday rotation for pure E u O a n d doped E u O f i l m a t 6 K a n d 20 kOe (a~erMcGuire et al. 1971).
190
W. R E I M and J. S C H O E N E S
150 K. Yet, the Faraday rotation as function of temperature increases only slowly down to approximately 80 K and accelerates its growth below this temperature. The last type of composite films containing Eu-chalcogenides which we would like to mention are those with enhanced Kerr effect. Ahn (1968) and Suits and Lee (1971) have reported data for EuO films deposited on mirror substrates and find a large increase of the Kerr rotation by closely matching the antireflection condition. Thus, Suits and Lee (1971) find a polar Kerr rotation of 70 ° for an EuO film deposited on a silver film (fig. 40). Yet, the reflectivity at the wavelength of maximum Kerr rotation is nearly zero, reducing again the effective figure of merit for the composite film. Another approach has also been taken by Ahn (1969) who optimized the thickness of the protective Eu20 3 layer on top of the magneto-optically active EuO film so as to enhance the apparent Kerr rotation. Of course, again the reflectivity of the composite film is depressed. Q)
6O OK
~6 "~ 40
/
• Reflectivity
~ 2O
\\ .
//----\/
,
~,. i I I
2O
o O_
-~o
-a •" ,
E o
0.6
/
,
",.
,
~ ""?-"i
0.8 1.0 1.2 WQvelength (/zrn)
o
,.
OS
I4
Fig. 40. Polar Kerr rotation, polar Kerr ellipticity, and reflectivity for an EuO film on top of a silver film on glass. The field is 20 kOe and the temperature is 5 K (after Suits and Lee 1971).
5.3. Miscellaneous europium compounds The large rotatory power of E u 2+ diluted in C a F 2 (Shen and Bloembergen 1964) has led many investigators to search for europium-containing compounds with specific properties. Although not all these compounds do show magnetic order in the temperature range studied, some of them are strongly paramagnetic and the moments might be paramagnetically aligned in a large enough magnetic field. There may then result magneto-optical signals per europium ion which are comparable to those of the europium chalcogenides and we will include a discussion of these materials in the following. The room-temperature Faraday rotation of EuF 2 has been determined by Suits et al. (1966). EuF 2 crystallizes in the cubic C a F 2 s t r u c t u r e and appears to be paramagnetic down to 0 K (Lee et al. 1965). The use of single crystals did limit the spectral range of the Faraday rotation measurement to the transparent region of E u F 2. This limit is given by the onset of 4f--->5d eg transitions. Figure 41 shows the absorption and the Faraday rotation of EuF 2 close to this absorption edge (Suits et al. 1966). From a plot of the reciprocal Verdet constant versus A2, the
MAGNETO-OPTICAL SPECTROSCOPY
-6
191
3O
"r=
tll 20
,'1"
8 _g
¢9
o]
400
5;0
xc.~l
~o
700
o
Fig. 41. Room-temperature Verdet constant (solid line) and absorption (dotted line) of a EuFz single crystal. The absorption is corrected for reflection losses (after Suits et al. 1966).
effective wavelength of the atomic transition causing the rotation and a wavelength-independent proportionality factor are obtained. Both values confirm that the observed Faraday rotation is due to an 4f7--~4f65d eg transition. Similar investigations have also been performed for glasses containing large amounts of Eu 2+ (Suits et al. 1966, Schoenes et al. 1979b). While the former measurements were restricted to room temperature, Schoenes et al. performed measurements down to 1.5 K in fields up to 100 kOe. Under these conditions paramagnetic saturation is reached and rotations of nearly 6000 deg/cm are reached at 0.8 p~m for a sample containing 12 at.% Eu 2+ (fig. 42). At room temperature, . the same sample at 0.8 txm has a Verdet constant of 3.5 x 10 -3 deg cm - 1 0 e -1 (Schoenes et al. 1979b). If we normalize this value to 33 at.% to simulate the Eu 2+ concentration in EuF2, we obtain V-- 0.58 min cm -I Oe -1, which is in very nice agreement with the data shown in fig. 41 for EuF 2. Besides amorphous silicate glasses, Eu 2+ forms the crystalline silicates EuSiO3, Eu2SiO 4 and Eu3SiO 5 (Sharer et al. 1963). The latter two order ferromagnetically near 9 K (Kaldis et al. 1974). Kaldis et al. reported for Eu3SiO 5 a magnetic red shift of 20 meV. Faraday rotation measurements appear to have only been performed for two wavelengths in Eu2SiO 4 (Shafer et al. 1963). At room temperature, the Verdet constant is similar to that in EuF 2. Europium has also been incorporated in various ferric oxide compounds (Kahn et al. 1969). In their comprehensive study, Kahn et al. (1969) show that the effect of europium and many other rare earths on the magneto-optical properties of these materials are only minor at room temperature. For europium, the main reason is that in orthoferrites and garnets europium should be trivalent. The negligible effects of some other rare earths can be understood by their small sublattice magnetization at room temperature. It has been demonstrated, that Eu enters divalent also in amorphous E u - F e C o alloys (Weller and Reim 1989a,b). However, no major contribution to the Kerr effect has been detected at room temperature, which is understood by the large optical constants of the intermetallic alloy and by weak exchange interaction.
192
W. REIM and J. SCHOENES
r~
"o t.a
._g
-o E
y_
o~
20
40 Magnetic
60
80
IO0
Field r'kOe]
2+ 3+ • Fig. 42. Field dependence of the Faraday rotation at 0.8 p~m of the glass Eu0.i2Eu0.02S1o.3100.55 for various temperatures between 1.5 and 300 K (after Schoenes et al. 1979b).
5.4. Gadolinium Regarding its magneto-optical properties, gadolinium is undoubtedly the best investigated rare earth metal. Gadolinium orders ferromagnetically at 293 K. As early as 1963, Lambeck et al. (1963) reported Faraday rotation measurements on vapour-deposited thin films of Gd. They found a negative rotation increasing by a factor of approximately 2 between A = 4000 and 6500 ~ . A t magnetic saturation, which Lambeck et al. reached by applying fields up to 30 kOe to samples cooled down to 93 K, the authors observed a rotation of - 3 . 2 5 × 105 deg/cm at A = 5890 A. Kerr effect measurements on thin films of Gd were first reported by Erskine and Stern (1973), who also presented a thorough analysis of the data. Although these authors used a set-up in which the magnetization direction is in the plane of the sample (longitudinal Kerr effect), Erksine and Stern separated that part of the Kerr effect, which is a linear function of the magnetization, by using a 30 ° angle of incidence and reversing the field. The obtained spectra are displayed in fig. 43. We note the small values compared with the spectra of the europium monochalcogenides. This indicates that the structures are hardly due to f---~d transitions and, in fact, Erskine and Stern assign the two prominent peaks near 2 and 4 eV to p--~ d transitions, the double structure being a consequence of a covalency splitting of the 5d band into two major sub-bands. To detect f---~d
MAGNETO-OPTICAL SPECTROSCOPY I0
I
I
i
I
193
I0 TATION
7 ,q
~5 I-
~ 4
q, Y
o,
T
T
I
T
~'
T
3
4
ti
5
ENERGY (eV)
Fig. 43. Magneto-optical Kerr rotation and ellipticity of polycrystalline Gd films at 105 K and a field of 2 kOe (after Erskine and Stern 1973).
transitions in Gd, the spectral range of the measurements has to be extended beyond the quartz transmission limit of 6 eV. Erskine (1976; 1977) performed such pioneering work at the Synchrotron Radiation Center at Stoughton, Wisconsin. Figure 44 shows the resulting too-2ry(W) spectrum including, as dots, the data from the former measurements in the quartz transmission range. Also shown is the absorptive part of the diagonal conductivity, tr~x. One recognizes that the two peaks at 2 and 4 eV, which we have already discussed before, lie in an energy 5.0
I
4.0
~
,~ 3.0
!
~_o z.o 'B
1.0
I
I
I
t
3.0~
.
_(2
-~' -'~U
2.5~ mm~
~L~ -2.C
/&5
(2] WCrxy
"
(i) / ~ ' - * " - ' " " " °-xx
"~
,~
~
~
,~
~-.0~ c* vff [.5
,,.o
ENERGY (eV)
Fig. 44. The product of the absorptive part of the off-diagonal conductivity tr2~yand the frequency to for polycrystalline Gd films. The dashed horizontal line is an estimate of the free-electron contribution. At the bottom the absorptive part of the diagonal conductivity tr~ is shown (after Erskine 1976, 1977).
194
W. REIM and J. SCHOENES
range which is dominated by free electrons. These free electrons give for o) >> 7 a o9-1 contribution to o-2xy. In a o)O-2xyplot, this flee-electron contribution is energy independent. The broken horizontal line in fig. 44 is the flee-electron contribution in Gd, as estimated by Erskine and Stern (1973). Near 6 and 8 eV, two additional and prominent structures appear, which are assigned to f--~ d transitions. Arguments regarding the sign and the weight of the peaks are presented favoring such an assignment (Erskine 1976). Indeed, a comparison of the line shape and the sign of the peaks with those found in the europium chalcogenides for f---~d transitions shows strong similarities. The more subtle question is why there are also two peaks for the f--~ d transitions in Gd. In the europium chalcogenides the two peaks reflected the crystal-field splitting of the final 5d state. Erskine and Stern (1973) have used a similar final-state splitting to explain the two peaks at 2 and 4 eV in terms of p---~d transitions. Yet, for the two peaks near 6 and 8 eV, Erskine (1976, 1977) discards such a final state splitting in favor of a model assuming two different screening mechanisms for the hole created in the optical excitation process. Erskine compares the energies of the two peaks with XPS data (Hedrn et al. 1972, Lang et al. 1981) where one finds 4f emission approximately 8 eV below the Fermi energy. It is then tempting to assign the 8 eV peaks in both spectroscopies to a same type of partial screening of the core hole, while the 6 eV peak in magneto-optics is ascribed to a fully screened 4f hole. As can be seen, the problem is related to what we have already discussed for cerium pnictides (section 4.1). Also, these different screening mechanisms were assumed to explain a peak separation of +2 eV. Yet, contrary to the cerium pnictides in which magnetooptics displayed one peak and photoemission two peaks, in Gd we encounter one peak in photoemission and two peaks in magneto-optics. On the other hand, we know that for the cerium chalcogenides, in which the 4f binding energy is larger than in the cerium pnictides, both magneto-optics and photoemission show only one peak. Thus, we favor an interpretation of the two peaks near 6 and 8 eV in o-2xy(W) of Gd in terms of a 5d band splitting, the difference in threshold energies for the optical and photoemission process being due to complete and only partial screening, respectively.
5.5. Intermetallic gadolinium compounds The magneto-optical properties of rare earth 3d-transition element alloys are presented in great detail in the chapter by Buschow (chapter 5 of vol. 4 of this handbook). Nevertheless, we would like to mention here, in connection with our discussion of Gd, Kerr rotation measurements for various Gd-containing intermetallic compounds. Figure 45 shows the double Kerr rotation at room temperature for bulk polycrystalline samples of Co, Gd2Co17, GdCos, GdCo 3 and GdCo 2 (Buschow a n d van Engen 1984). From the magnetic properties of Gd-Co, it follows that the moments of the Gd and Co sub-lattices couple antiparallel. For Gd2Co17 and GdCos, this results in a dominance of the Co-sublattice magnetization and for GdCo 3 and GdCo 2 in a dominance of the Gd-sublattice magnetization. To facilitate a comparison, Buschow and van Engen (1984) have, therefore,
MAGNETO-OPTICAL SPECTROSCOPY
195
-0.40 -0.20 f -0"10 t
o.o;l
GdCo3
.~ 0.o~,
~,~ 0.06
i
i
i
i
i
i
r
i
I - 0.20 -0.15
-0.25 -0.26 -0.46 -0.6£ i
i
i
i
-0.40 -0. 60
-0.80
~'0
q
2'0 '
i
~0
i
i
40
i
5.0
energy [eV)
Fig. 45. Double polar Kerr rotation of various polycrystaUine bulk Gd-Co compounds at room temperature and a field of 11.5 kOe. Note the reversing of the positive axis for GdCo2 and GdCo3. The vertical bars correspond to the peak position in Gd (after Buschow and van Engen 1984). reversed the direction of positive Kerr rotation for GdCo 3 and G d C o 2. The two vertical bars indicate the energy of the Kerr rotation peaks observed by Erskine and Stern (1973) in Gd (fig. 43). The rather good agreement of the structures in GdCo 2 with these two energies suggests an interpretation of these two peaks as arising from gadolinium. Yet, one should realize that Kerr rotation spectra of sublattices are, in general, not directly additive, this being contrary to the absorptive parts of the off-diagonal conductivity. Also, because the two structures near 2 and 4 eV in Gd are assigned to transitions between p- and d-band states, band-structures effects will play an important role in contrast, e.g., to the N d - C o system (section 4.3), in which the Nd derived structure at 4 eV is related to a transition of localized atomic-like 4f-states. The photon-energy dependence of the Kerr rotation and ellipticity of amorphous G d - C o and G d - F e films has been studied by Sato and Togami (1983). In their measurements through the glass substrate, they find maximum rotations near 0.9 eV of 0.4 ° in G d - C o and 0.58 ° in G d - F e . From a qualitative comparison with exy(OJ) of Co, Fe and Gd, Sato and Togami (1983) conclude that the magnetooptical effect of amorphous Gd-transition metal films in the visible and near infrared region essentially stems from the magneto-optical polarizability of the transition metal atom. Hansen and Urner-Wille (1979) have investigated the
196
W. REIM and J. SCHOENES
effect of alloying Bi to Gd-Fe. For the wavelength of the H e - N e laser (i.e. 633 nm), they found an increase of the Kerr rotation from 0.29 ° for a Gd0.26Feo.74 sample to 0.41 ° for a sample of (Gd0.26Fe0.74)0.89Bio.11 . Van Engelen and Buschow (1987) have investigated Gd/Fe14B and Gd2Co14B. Maximum Kerr rotations of -0.34 ° and -0.29 ° are reported for the Fe and Co compound, respectively. The Kerr rotation spectra of the Gd compounds are found to be quite similar to the corresponding La compounds, while for Nd2Fe14B a net increase of the Kerr rgtation is observed near 4 eV. 6. Heavy rare earths
Except for thulium chalcogenides, rather few magneto-optical spectra have been collected for the rare earth metals beyond gadolinium and their simple compounds. This is rather unfortunate, since alloys of these heavy rare earths with 3d-transition elements have received much attention and the knowledge from the simple systems could provide important hints for the interpretation of these data. As before, we will start with the discussion of the "simple" NaCl-type compounds, to move then to the pure metals, which are already complicated by the presence of free electrons and strong band-structure effects, and to end with results for a few selected miscellaneous compounds.
6.1. NaCl-type compounds The thulium monochalcogenides have raised much interest in the past due to the occurrence of intermediate valence in some of them. That is: TmS is an integral-valent metal with Tm 3÷, TmTe is an integral-valent semiconductor with 2+ Tm , and TmSe is an intermediate-valent metal with a valence between two and three. In TmSe the energies of the 4f 13 and 4f125d1 states are degenerate, leading to a ground state which is a linear combination of these two states. The unique feature of thulium compounds compared with intermediate-valent materials containing cerium, samarium or europium is that both ground-state configurations of thulium have a non-zero magnetic moment. This leads to magnetic ordering in the thulium monochalcogenides at low temperatures and promises that magnetooptics can probe these two groundstates. The Kerr rotation and ellipticity of pseudobinary TmSe0.32Te0.68 and TmSe is shown in fig. 46 (Schoenes et al. 1985). TmSe0.32Te0.68 has a lattice constant of 6.204 A and is semiconducting while TmSe is an intermediate-valent metal with a lattice constant that is 0.5 .~ smaller. To elucidate the physical property causing the very different magneto-optical spectra of these two materials, fig. 46 also exhibits the Kerr rotation and eUipticity for two more thulium monochalcogenides. We recognize that integral-valent metallic TmS (Reim et al. 1984a) and intermediate-valent TmSe0.87Te0.13 (H/isser 1983) have very similar spectra to intermediate-valent TmSe. It has therefore been concluded (Reim et al. 1984a, Schoenes et al. 1985) that the differentiating feature in these compounds is whether the material is semiconducting or metallic and not whether it is integral
MAGNETO-OPTICAL SPECTROSCOPY
0.5
b)
2t
197
'\
TmS~
-OK. 11"
0.25
"
\
0
-0.25
.
B =4T , =
-0.5
~o -£K
E
8
V
-0,75 I
2
4
3
5
Photon Energy (eV)
c)
d)
+'+ +;
-B K
:\
TmS
++ ÷ i
°"
v "5 a_
T1 2K B=4T
,,"~÷ t +o TmSe87Te13 o
° ~ °~++ -E K * *
'l
~o -E K
T=IOK B=4T
o
;
i 1
i 2
'i~l 3
h 4
i 5
Photon Energy (eV) Fig. 46. Polar magneto-optical Kerr rotation Or: and ellipticity e K of cleaved single crystals of TmSe0.32Te0.68(a), TmSe (b), TmSe0.87Te0.i3 (c) and TInS (d) (after Schoenes et al. 1985, Reim et al.
1984b, Hiisser 1983).
or intermediate-valent, nor whether it is a pseudobinary or a true binary compound. Figure 47 shows the real part of the diagonal conductivity and the complex off-diagonal conductivity of TmSe0.3zTe0.68 as a function of photon energy. If one compares the structures of o'2xy with those of crlxx one finds an agreement in their respective energies, if the peak at 0.5 eV is of paramagnetic line shape and the peaks in O-ixx at 1.7 and 2.5eV give rise to diamagnetic line shapes, i.e., correspond to extrema in o-lxy. Figure 48 displays the energy-level scheme for the 4f13---~ 4f~25d transition, which has been proposed by Schoenes et al. (1985) to account for the magneto-optical data of TmSe0.3zTe0.68. In contrast to the europium chalcogenides, the spin-orbit splitting of the final states (3H and 3F) of thulium is comparable in size to the electrostatic repulsion in the f shell. Thus, a clear hierarchy of the various interaction terms is not present and the 3H and SF terms strongly mix. The last column couples the angular momentum of the fa2 state with J = 3 for the 5d electron. Dipole-allowed transitions require Aj = 0 or - 1 . At low temperatures, one expects for J increasing by one a positive peak in O"2xy and for J decreasing by one a negative peak. These transitions and their
198
W. REIM and J. SCHOENES
TmSe32Te68/
o
to
O
d
O
,'' ,"f' ,o Q O I
2
3
4
TmS%2T%8
O-2*y
I I I
5
B =4T T =,0K
I
b) I
2
4
5
Photon Energy (eV)
Fig. 47. (a) Real part of the diagonal conductivity Glx. of TmSe0.3=Te0.58 and TmSe. (b) Real part Glxy and imaginary part cr2xyof the off-diagonaJ conductivity of TmSe0.32Te0.68 at 10 K and a field of 40 kOe (after Schoenes et al. 1985).
eg4f 12
5/2 7/2 9/2 11/2 7/2
1Dip
~Fs/~
5/2 5/2 1/2 9/2 7/2 5 / 2 9/2 5 / 2
7/2
11/2 9/2
11/2 5/2 15/2 11/2
7/2 7/2 5/2
9/2
1512 11/2 15/2
912
4f13 (
2F7/ 2
+
)
++
i
,n,.q ~ ~t..Qo O- ~ eJcJ c, b+ o
Fig. 48. Atomic coupling scheme for TmSeo.32Te0.68 and observed 4f13--o4f125d transitions and energies (after Schoenes et al. 1985).
MAGNETO-OPTICAL SPECTROSCOPY
199
expected signs and energies are also given in the last column. Transitions with •J = 0 have similar transition probabilities for right and left circularly polarized light and are omitted for clarity. The agreement with experiment is satisfying and indicates once more that the coupling of the excited 5d electron to the 4f"-1 state can not be neglected in the optical absorption process in rare earth chalcogenides and pnictides. Figure 49 shows the real (open circles) and the imaginary (solid circles) part of the off-diagonal conductivity for TInS (Reim et al. 1984a) and TmSe (Schoenes 1987). The large peaks in OK and ei~ at 3 and 2.5eV in TInS and TmSe, respectively, have transformed to small structures at the same energies in OSxyand o-2xy. On the other hand, the off-diagonal conductivity shows very large values at low energies with a marked extremum in o-2xynear 0.5 eV. Reim et al. (1984a) have given quantitative evidence that this behavior, observed for the first time in the metallic thulium monochalcogenides, is associated with the free carriers. In particular, they could show that the structures in OK and e K correspond to a splitting of the reflectivity curves near the plasma edge for right and left circularly polarized light of the order of 30 to 60 meV compared to the 0.4 meV that one expects from the cyclotron resonance frequency at 4 T. This is a strong indication that the dominant term is not the direct action of the field on the free carriers, but that the local moments have to interact with the free carriers. Consequently, skew scattering has been inferred and the full lines in fig. 49 for TmS correspond to fits with the first term of eq. (2.65). Another interesting question is, of course, what has happened with the f13__~fla5d transitions in intermediate-valent TmSe? Unfortunately, the presently
b
8
it
TmS
~xy
T 24KT °11 I B:= ~o[I/
~'2xy
~
1
2
3
~
~
4
5
I
T=2K
B=4T
-30 1
2
3
4
5
Photon Energy [eV] Fig. 49. Real part ~rlxy and imaginary part Or2xy of the off-diagonal conductivity of TmS and TmSe at 2 K and a field of 40 kOe. The full lines are fits with a skew scattering model (after Reim et al. 1984a, Schoenes et al. 1985).
200
W. R E I M a n d J. S C H O E N E S
available data cannot give a definite answer. Figure 49 shows some structures in the 2-5 eV energy range which are not explained by the skew scattering model for free electrons• If we assume a valence of 2.5 for Tm in TmSe, as may be derived from the Curie constant (Boppart and Wachter 1984), and we normalize the data obtained for TmSe0.32Te0.68 (fig• 47) by the factor ½, one anticipates peaks in O'xy of the order of 1013 S-1. This corresponds exactly to the order of magnitude of the deviation of the measured spectra from the free-electron fit in the 2-5 eV energy range (fig. 49). However, similar structures are also present in TmS which has no occupied 4f 13 state. In order to possibly find an answer to the above question, H/isser (1983) and Reim and Wachter (1985) have also investigated Tm0 sEu0.sSe. This material is so close in lattice constant to the semiconductor-metal transition that the metallic state can be induced at the surface by polishing the crystal (Batlogg 1981). Thus, a cleaved single crystal will be semiconducting and a polished one will have a transformed intermediate-valent layer which is thick enough that the crystal, in an optical reflectivity measurement, behaves like a bulk metallic sample. Figure 50 illustrates this difference in a plot of the real part of the optical conductivity for a cleaved and a polished crystal. Figure 51 shows the complex polar Kerr effect for the same two samples. Near 2 and 4 eV we find in both samples the 4f7--~ 4f65d t2g and - e g transitions of europium (see for comparison fig. 30). Outside these energy ranges the two spectra are quite different. For the cleaved Tm0 sEu0.sSe crystal, one recognizes below 2 eV the same structures as for TmSe0 32Te0 68 (fig. • • • " " • 46), Le., 4f 1 3 ~ 4f 1 2 5d transmons. The spectrum below 2 eV of the pohshed Tm0.sEu0.sSe sample, on the other hand, resembles the spectra for TmS and TmSe (fig. 46) with a reversed sign and a shift to 1 eV. This part of the spectrum is 3
I
I
I
I
I
Tm.5Eu.sSe T = 293 K polished --cleaved u' )
"~'N
2
0
i
v x
l
b* 0/
J
0 0
. ,f--I /
I
I
I
J
1
2
5
4
5
Photon Energy (eV) Fig. 50. R e a l p a r t of the optical conductivity Olx x at r o o m t e m p e r a t u r e of a cleaved a n d a polished Tm0.sEu0.sSe single crystal (after H/isser 1983).
MAGNETO-OPTICAL SPECTROSCOPY
2
Tm.sEUsSe
°° b," +t o
:o:o
t8
~. LU i
o
T=IOK,
:°°
o
o
.
~
Tm.5EUsSe
+
cleaved
* ~
polished
+o~
T=IOK, B=4T
+ ~÷
B=4T
201
~ o
?
~ Y :
-6 O_
Eo -I
~o
+~*-o
0
i
s
og
4
; Photon
Energy
(eV)
Fig. 51. Polar magneto-optical Kerr rotation OK and ellipticity e~ of a cleaved (left) and a polished (right) Tm0.sEu0,sSe single crystal at 10 K and a field of 40 kOe (after Hiisser 1983).
therefore assigned to the free carriers, which have a coupled plasma frequency of I eV and which are polarized antiparallel to the net moment of the sample. This antiparallel polarization is a result of an antiferromagnetic coupling of the moments of the Tm and Eu sublattices. The Tm-derived conduction electrons are polarized parallel to the Tm 4f electrons and antiparallel to the dominant Eu 4f electrons. Thus, again free electrons hinder an unambiguous identification of 4f13---* 4f125d transitions in an intermediate-valent thulium compound, and we do not know, at present, how much the f - d and/or f - p hybridization reduces the spin-orbit coupling and how much this affects the magneto-optical signals from f---> d transitions. The second class of heavy rare earth chalcogenides which have been studied by magneto-optical methods are the ytterbium-monochalcogenides (Suryanarayanan et al. 1970, 1974). Suryanarayanan et al. prepared thin films of YbSe (1970) and YbTe (1974) and measured the optical absorption as well as the magnetic circular dichroism (MCD). Figure 52 shows the results of these measurements for a 1000 A thick film of YbTe deposited onto CaF 2 by co-evaporation of the elements Yb and Te (Suryanarayanan et al. 1974). The magnetic field for the MCD measurements was 37.5 kOe. Absorption measurements down to wavelengths of 10 p.m do not show additional absorption bands, indicating that the peak near 2 eV corresponds to the lowest interband transition. Thus, YbTe is a semiconductor (or insulator) with divalent Yb and a completely filled 4f shell. Consequently, YbTe and YbSe are diamagnetic. The temperature dependence of the MCD spectra of the two compounds has been analyzed in terms of diamagnetic and paramagnetic line shapes, and these authors conclude that the relatively weak increase of the MCD peak intensity with decreasing temperature is a narrowing effect of the line and not a population effect of the ground-state. This indicates that the ground state is in fact diamagnetic. The numerous lines have, therefore, to be associated with splittings of the excited 4f135d state. Figure 52 shows also on
202
W. REIM and J. SCHOENES Photon Energy (eV)
2
5
I
4 5 l I
I
I 2.5
5/z ] 3
o
ci
%
~
- - 3&
I
-10
+ •
/52
I /--(
,-
---- 300 K
YbTe
4f155dJ
/ I /
/
/ I
Hcf(d) I
I
I
I
6
5
4
5
Wovelength
--Z 4v4
C
t
rl~ t2gil 7/2 /--
I
-
v3eg L , 7 , ~ - -
i
~.
--IOK
,>
Hso(f) Hso(d)
2
(10 3 ~ )
Fig. 52. (left) Magnetic circular dichroism AD and optical density D of a 1000~ thick film of YbTe at 300 and 10 K. Note that for AD at 300 K the scale on the right is valid and for 10 K the scale on the left is valid. The field is 37.5 kOe. (right) Atomic coupling scheme for YbTe (after Suryanarayanan et al. 1974). the right-hand side the coupling scheme proposed by Suryanarayanan et al. (1974) for YbTe. The crystal-field splitting of the 5d states is estimated in a point-charge model to be approximately 1 eV. The spin-orbit splitting of the 4 ) 3 state into 2F7/2 and 2F5/2 is taken from free-ion spectra to be approximately 1.2 eV. It results in four levels A, B, C and D, of which B and C are nearly degenerate. The peak at approximately 2 e V is assigned to the transition 4f14---~4f13(ZF-.~5de ", //z/ g denoted A, the ~roup of peaks near 3eV is assigned to 14 13 2 13 2 transition s from 4f into the final states 4f ( FT/2)5d t2g (C) and 4f ( Fs/z)5d t2g (B) and the peak at 4.24 eV is assigned to a transition into the 4f13(=Fs/2)5d eg final state. Besides this crude weak-coupling model between 4f 13 and 5d 1 states, these authors also give on the right-hand side of fig. 52 the excited levels in a more sophisticated coupling scheme, but the agreement with the experimental data is rather worse. 6.2. Heavy rare earth metals
Magneto-optical data for the heavy rare earth metals are extremely scarce. To our knowledge, only terbium and dysprosium have been studied, but, even for these two materials, little information is available. Thus, Kerr spectra have not been published and we have to content with a plot of the absorptive component of the off-diagonal conductivity multiplied with the photon energy (fig. 53) (Erskine 1975). As has been pointed out already in the discussion of Gd (section 5.4) this kind of plot eliminates the 1/to energy dependence of the free-carrier contribution to o-2xy. Thus, the spectra should reflect mostly interband transitions. Very little
MAGNETO-OPTICAL SPECTROSCOPY i
I
I
I
203
I
2.0 Gd t'M i U
1.0 t%l
o
)/
0
'
b~
3
Dy
-I.0 0
I
I
[
[
I
1
2
3
4
5
Energy (eV) Fig. 53. The product of the absorptive part of the off-diagonalconductivityo-z~yand the frequencyco for terbium, dysprosiumand gadolinium.The terfiperature and field have not been specified (after Erskine 1975). has been said about how these spectra are to be interpreted, except that the interpretation of the Tb and Dy spectra is complicated by the expected presence of transitions involving 4f-electrons. Judging from the very fiat diagonal conductivity spectrum of Dy (Erskine 1975), it appears, in fact, that s o p and d<--->p transitions dominate the diagonal and f <-->d transitions dominate the off-diagonal conductivity. This then, would indicate that the peak near 4 eV is possibly an f--> d transition in fair agreement with a binding energy of 3.8 eV determined by X-ray photo-electron spectroscopy (Lange et al. 1981).
6.3. Miscellaneous compounds containing heavy rare earths Due to the potential applicability for thermomagnetic writing, alloys based on (Tb,Gd)x (Fe,CO)l_ x have received much attention in the past ten years and the chapter by Buschow (chapter 5 of vol. 4 of this handbook) is reviewing these results in full. Here, we limit ourselves to the presentation of a few selected data, which might be explained on the basis of the previous discussion of the magnetooptical properties of elemental heavy rare-earths or simple heavy rare earths compounds. Choe et al. (1987) have measured the Kerr rotation of G d - C o and Tb-Co films at room temperature for wavelengths ranging from 250 to 700 nm. Guided by the observation that the Kerr rotation at long wavelength is almost the same for a same concentration of different rare earths, Choe et al. (1987) have normalized the Kerr rotation spectra by their respective values at 700 nm and compared it with Y-Co. Figure 54 shows then, that with increasing Tb-content the normalized
204
W. REIM and J. SCHOENES Energy
(eV)
5
4
3
2
I
I
I
I
,
E ¢::
r--°°,,1.0 ~
Gd35Co.65
Gd.26C°.74 ~-Gd19 Co 81
\/
.
.
~
3E
~- 0.5 '~
/ /" zY .25 .75 tL--~---~t Tb19 C081 i ii • •
Cb
i
/ OL,-,
s'~ff~Tb.29 C° 71 ii ,;~'
!! ~
500
Tb35C°65 I
400
I
I
500 Wovelength
I
I
600
,
I
I
700
(nm)
Fig. 54. Kerr rotation spectra for various amorphous Gd-Co and Tb-Co films, normalized by their Kerr rotation values at 700 nm. The temperatures is room temperature and the field is 13 kOe (after Choe et al. 1987). Kerr rotation becomes smaller at shorter wavelengths, while with increasing Gd content the normalized Kerr rotation increases at the shortest wavelengths. From the occurrence of magnetic compensation in the G d - C o and T b - C o systems one concludes that the two sublattices couple antiparallel in both systems. Looking then at fig. 53, which shows positive totr2xy values for Gd as well as for Tb, the different influence of these two rare earths in Co is not obvious, and this indicates once more the difficulty to discuss Kerr rotation spectra directly. As speculation we offer the following explanation. The o)Or2xy spectra are superpositions of free-carrier and interband contributions. An estimate of the former contribution in Gd (see fig. 44) gives a value of 1029 s -2. Because the diagonal optical conductivities of Tb and Gd are quite similar (Erskine 1975), one may also anticipate a similar free-electron contribution to the off-diagonal conductivity of the two metals, which would then lead to interband transitions of opposite sign for Gd and Tb. The system T b - F e has been investigated by Connell (1986), who also offers a microscopic model. In this model, the diagonal and off-diagonal conductivity below 3 eV are dominated by Fe d-states with, however, some negative contributions to Or2xy between 2 and 3 eV that are attributed to d--~ f transitions from Tb. A partial substitution of Tb by Gd, Nd or Sm is found to enhance the Kerr signal in agreement with the offered explanation. From a spectroscopic investigation of various R E - F e C o amorphous films Weller and Reim (1989a,b) conclude that f--~d transitions may contribute at maximum some tenths of a degree to the Kerr effect due to the large values of the optical constants in intermetallic alloys compared to NaC1 structure compounds. The polar Kerr effect of some polycrystalline R E - F e 2 compounds has been measured by Katayama and Hasegawa (1981). As fig. 55 shows the compounds where R E is either R E = Er, H o , Dy or Tb have a sign change in the rotation
MAGNETO-OPTICAL SPECTROSCOPY I
"-
I0
I
i
I
205
I
- ,f'\\./~
tO
•4= l:l "5
0
tr
x,, - I 0
/Y
',
ErFep_
\"'~ "HOFe2 I '--k,,.,//L---.-_ ~DyFez
H : I 2 kOe T= :500 K
0
a_ - 2 0 I
300
I
I
500 Wovelengih
I
i
700 (nm)
Fig. 55. Polar Kerr rotation spectra for polycrystalline bulk samples of GdFe 2, TbFe2, DyFe z, H o F e and ErFe2 at room temperature and a field of 12 kOe (after Katayama and Hasegawa 1981).
2
spectrum between 350 and 500nm, while GdFe 2 always stays positive. The rotation is largest in T b F e 2 , but does not exceed 0.4 °. The equatorial Kerr effect of nearly all R E - F e z Laves phases with heavy rare earths has been reported by Mukimov et al. (1985) and Shapirov et al. (1986). In contrast to the preceding works, these authors have reduced their data to obtain the off-diagonal conductivity for their polycrystalline samples. Figure 56 displays the absorptive and dispersive parts of the off-diagonal conductivity for GdFez, TbFe 2 and ErFe z (Shapirov et al. 1986). The differences are noted to be larger than in F e l T R E 2 compounds, which contain less rare earth, and they are attributed to the hybridization of the f electrons with the conduction electrons. Most of these interpretations are vague or very qualitative. The occurrence of contributions of equal importance from itinerant and localized states at the same energy renders assignments on the basis of simple energy-level schemes very difficult if not impossible. What would be needed here, are band-structure calculations that can handle itinerant and localized states on an equal footing, to then allow a computation of the diagonal and off-diagonal conductivity spectra. There is a long way to go to understand or even predict the magneto-optical properties of rare earths-transition metal intermetallic compounds. In order not to close this paragraph on a pessimistic note, we would like to mention that some rare earths have also been considered for devices such as optical isolators. Besides Eu 2+, Tb 3+ is very promising. Its Faraday rotation in a glass has been investigated, e.g., by Davis and Bunch (1984) at the H e - N e laser wavelength of 6328 ,~. For an unspecified concentration of Tb 3+ in the glass, a specific room-temperature Faraday rotation of 0.004°cm - 1 0 e -1 is quoted. At liquid-helium temperature the Verdet constant is approximately 0.3 ° cm - 1 0 e -1. This is nearly the same value as we have discussed for the E u 2+ glass (section 5.3) at 8000 A and allows a rotation of 45 ° for a i cm sample with application of only 150 Oz. As origin of this large rotation we anticipate a 4fs---~4fY5d transition because the 4f 8 state is about 2 eV below E v.
206
W. REIM and J. SCHOENES I
i
I
•
" f ' ~ ~
E z 0
! / / l//
/"-- !JiI
I I I
I
-72
~~
i
I -70
f"-.,~
•
/il
I
I
/ "~)
--
-
-
-
-
-
GdFe2 TbFea
ErFe2
JJ 7
E(or) Fig. 56. Real part trlxyand imaginary part tr2xr of the off-diagonal conductivity of GdFe 2, TbFe 2 and ErFe 2 polycrystalline bulk samples at room temperature. The field is cited to be 40e! probably 4 kOe (after Sharipov et al. 1986).
7. Uranium compounds Actinide compounds a p p e a r to be a promising field for magneto-optical research due to the fact that the 5f" states m a y be located near the Fermi energy. Regarding the magnitude of the magneto-optical effects c o m p a r e d to rare earth materials, an e n h a n c e m e n t due to the larger s p i n - o r b i t energy can be expected and was in part experimentally verified. A m a j o r drawback is the high radioactivity of most actinide elements and only compounds using depleted uranium have found their way into the laboratories on a b r o a d scale. The configuration of the outer electrons in uranium is 5f36d17s 2. In compounds, different ground-state configurations have been found, e.g., the idealized ground states 5f 2, 5f 3 and 5f36d I in UO2, U P and US, respectively. H o w e v e r , hybridization effects in the latter two metallic samples cause a m o r e general description in terms of 5f3-X6d x and 5f3-X6d 1÷~ states to be appropriate. With the chalcogens X = S, Se, Te and pnictogens Z = P, As, Sb compositions with atomic ratio 1 : 1, 1 : 2 and 3 : 4 are known. Most of the magneto-optical efforts up to now have b e e n focussed onto these materials.
7.1. Uranium dioxide U r a n i u m dioxide, UO2, was one of the first uranium compounds to be investigated intensively because of its use as nuclear fuel. As early as 1959 ( A c k e r m a n n
MAGNETO-OPTICAL SPECTROSCOPY
207
et al. 1959), stoichiometric thin films have been prepared, and the absorption coefficient K was determined showing that UO 2 is semiconducting with an energy gap of about 2 eV. However, only in the early eighties, the electronic structure was completely resolved by optical (Schoenes 1978, 1980b), magneto-optical (Reim and Schoenes 1981) and photoemission (Baer and Schoenes 1980) spectroscopies. Figure 57 displays the derived energy-level scheme consisting of an oxygen 2p valence band, an uranium 6d conduction band with a crystal-field splitting of 2.8 eV between the 5d t2g and 5d eg sub-bands and a localized 5f2 state. Thus, the energy-level scheme looks very similar to the one of EuO (section 5.1), and large magneto-optical effects can be expected. However, UO 2 orders antiferromagnetically at TN = 33 K (Leask et al. 1963), and with the knowledge of the sublattice moment of 1.73/x B from neutron scattering (Faber and Lander, 1976) it was estimated that an applied field of about 190T would be necessary for magnetic