Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Application of Green's function technique to paramagnetic resonance Frank, Barry 1965

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-UBC_1965_A1 F7.pdf [ 3.95MB ]
JSON: 831-1.0085839.json
JSON-LD: 831-1.0085839-ld.json
RDF/XML (Pretty): 831-1.0085839-rdf.xml
RDF/JSON: 831-1.0085839-rdf.json
Turtle: 831-1.0085839-turtle.txt
N-Triples: 831-1.0085839-rdf-ntriples.txt
Original Record: 831-1.0085839-source.json
Full Text

Full Text

The University of B r i t i s h Columbia FACULTY OF GRADUATE STUDIES PROGRAMME OF THE FINAL ORAL EXAMINATION FOR THE DEGREE OF DOCTOR OF PHILOSOPHY of B A R R Y F R A N K B.Sc, M c G i l l University, 1961 M.Sc,,. McGill University, 1962 THURSDAY, JULY 8th, 1965, AT 9:30 A„M 0 IN ROOM 301, HENNINGS BUILDING External Examiner: D„ D. Betts COMMITTEE IN CHARGE Chairman: W. Ho Gage R, Barrie J . A. R, Coope J. M. McMillan P. Ra s t a l l Co F, Schwerdtfeger R. F. Snider Department of Physics University of Alber t a Edmonton, Alberta APPLICATION OF GREEN'S FUNCTION TECHNIQUE TO PARAMAGNETIC RESONANCE ABSTRACT This thesis contains discussions of a number of points which arose when the author was studying the "paramagnetic resonance l i n e shape problem". The so-c a l l e d moment method i s discussed, and a new deriva-t i o n of the moments of the l i n e shape function i s given. Single-spin operators are introduced which simplify the c a l c u l a t i o n of these moments. The Green's function technique,, as applied to t h i s prob-lem, and the decoupling approximations associated with the technique, are looked at from the point of view of r e l i a b i l i t y and complexity. As a test of the r e l i a b i l i t y of any decoupling, a theorem concerning the moments of a l i n e shape a r i s i n g from such a decoupling i s discussed and proved. The Green's function technique i s applied to the case of the one-dimensional Ising model with spin 1/2, where no de-coupling of the hierarchy of Green's function equa-tions i s necessary. A method of c a l c u l a t i n g thermal averages for th i s case, using diff e r e n c e equations, i s given. GRADUATE STUDIES F i e l d of Study; Physics Quantum Theory of S o l i Plasma Physics Advanced Magnetism S t a t i s t i c a l Mechanics Advanced Quantum Mechanics s R. Barrie L. Go de Sobrino M. Bloom Ro Barrie : F. A, Kaempffer APPLICATION OF GREEN'S FUNCTION TECHNIQUE TO PARAMAGNETIC RESONANCE by BARRY FRANK B.Sc., McGil l Univers i ty, 1961 M.Sc , McGil l Univers i ty, 1962 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in the Department of PHYSICS We accept th i s thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA June, 1965 In p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t o f the r e q u i r e m e n t s f o r an advanced degree a t t h e U n i v e r s i t y o f B r i t i s h C o l u m b i a , I a g r e e t h a t t h e L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e and s t u d y . I f u r t h e r a g r e e t h a t p e r -m i s s i o n f o r e x t e n s i v e c o p y i n g o f t h i s t h e s i s f o r s c h o l a r l y p u r p o s e s may be g r a n t e d by t h e Head o f my Department o r by h i s r e p r e s e n t a t i v e s , , I t i s u n d e r s t o o d t h a t c o p y i n g o r p u b l i -c a t i o n o f t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l not be a l l o w e d w i t h o u t my w r i t t e n p e r m i s s i o n . Department o f Physics  The U n i v e r s i t y o f B r i t i s h Columbia Vancouver 8, Canada Date T ^ L 8 ^lQL£ i i ABSTRACT This thesis contains discussions of a number of points which arose when the author was studying the "paramagnetic resonance line shape problem". The so-called moment method is discussed, and a new derivation of the moments of the line shape function is given. Single-spin operators are introduced which simplify the calculation of these moments. The Green's function technique, as applied to this problem, and the decoupling approximations associated with the technique, are looked at from the point of view of reliabil ity and complexity. As a test of the reliability of any decoupling, a theorem concerning the moments of a line shape arising from such a decoupling is discussed and proved. The Green's function technique is applied to the case of the one-dimensional Ising model with spin ht where no decoupling of the hierarchy of Green's function equations is necessary. A method of calculating thermal averages for this case, using difference equations, is given, i i i TABLE OF CONTENTS Page Abstract . . . . . . . . . . i i Table of Contents i i i Acknowledgements . v Chapter I - Introduction and Summary .. .. 1 Chapter II - General Formalism . . 5 Section 1: System Hamiltonian 5 Section 2: General Formalism of Absorption Line Shape 12 Chapter III - Green's Functions and the Ising Model 17 Chapter IV - The Moment Method 22 Section 1: General Discussion . . . . . . . . . . . . . . . . . 22 Section 2: Application to Paramagnetic Resonance Line Shape Problem . . . . 26 Chapter V - The Moment Theorem . .• 35 Section 1: Relation of Moments to Terms in Hierarchy of Equations 35 Section-2: Decoupling Procedures . . . . . . . . . 36 Section:3: Statement and Proof of Moment Theorem 43 Section 4: Summary on Use of Theorem 50 Chapter VI - Resonance Line Shape Problem 51 Section 1: Decoupling Methods 51 Section 2: Solution of Equations 61 Section 3: Complexity Requirement . . 67 Section 4: Application to Case of Completely Non-equidistant Single-Spin Energy Levels 71 Section 5: Summary of Chapter VI . 75 i v Page. Appendix I: Mathematical Appendix to Chapter III . . . . 76 Appendix II: Thermal Averages with Ising Model'Hamiltonian: .' . . . . 78 i Appendix III: Proof of (5-14) from (5-2a) and (5-3) .84 p Appendix IV: Reality of A^ of Equation (5-47) .87 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . 89 V ACKNOWLEDGEMENTS I wish to express my gratitude to Professor R. Barrie for suggesting this problem and for his continued interest and valuable advice throughout the performance of this research. I wish also to thank the National Research Council of Canada for financial help in the form of a Studentship. 1 CHAPTER I Introduction and Summary This thesis contains discussions of a number of points which arose when the author was studying the "paramagnetic resonance line shape problem." This problem is one which has long defied attempts at solution and we do not succeed in solving i t . However, some points arose which i t was thought were, in themselves, worthy of discussion. The motivation for tackling the problem was that with the recently introduced techniques of double-time temperature-dependent Green's functions, some hope of solution might be entertained. Let us f i rst briefly review the problem (details are given in Chapter II). The physical system consists of a crystalline sample containing N identical paramagnetic unit systems or "spins". This system is acted on by a constant magnetic field HQ. As a result the system has a certain number of eigenstates, and transitions between these eigenstates can be induced by a high-frequency oscillating magnetic field H-j(t) = H^  cos wt. The resulting absorption of the high-frequency radiation is referred to as paramagnetic resonance absorption and the problem is to calculate, from f irst principles, the shape of one of the absorption lines. In particular one wishes to know how this shape changes -with temperature. The problem that has been studied is that in which the temperature-dependence of the line shape arises solely from the average occupation numbers of the various energy levels of the system. In the absence of interaction between the unit spin systems, these energy levels would be, except for the lowest and the highest energy levels, highly degenerate. The spin-spin interactions then remove these degeneracies to a certain degree. There exists, therefore, no natural relaxation mechanism that would yield a continuous line shape: the line shape function consists of a series of delta-2 functions. One has basically to carry out a quantum-mechanical calculation to find the energy levels, and then to follow this with a statistical-mechanical treatment to find occupation probabilities. With the inclusion of other thermal effects (e.g. lattice vibrations), one would be led to broadened temperature-dependent energy levels and perhaps a continuous line shape. We do not discuss this aspect of the problem. The Hamiltonian of the system consists of three terms: (a) a spin Hamiltonian which is the sum of the Hamiltonians of the individual paramagnetic units in the presence of the applied constant magnetic field and the crystalline f ie ld; (b) a spin-spin interaction which consists of a dipole-dipole interaction and an exchange interaction; (c) the interaction with the applied oscillating magnetic f ie ld. With no crystalline field present, the energy levels of the individual spins are equally spaced, with spacing "tvtJ0 say. The absorption spectrum then consists of a set of lines at frequencies 0, WQ, 2WQ, 3WQ, . . . and to single out the primary line at WQ one does a truncation of the spin-spin interaction (Van Vleck, 1948). If a crystalline field is present and the effective spin is greater than one-half, then in general the individual spin levels for part (a) of the Hamiltonian will not be equally spaced. One then obtains a set of primary lines (with their associated secondary ones). A truncation of part (c) of the Hamiltonian is then required (Kambe and Usui, 1952) to single out a particular primary line. This latter truncation has been performed (Pryce and Stevens, 1950); McMillan and Opechowski, 1960) using projection operators which pick out eigenstates of the full Hamiltonian of part (a). It is possible to carry out the second truncation (Kambe and Usui) using operators defined in the representation of single-spin eigenfunctions. These latter operators are introduced in Chapter II of this thesis. 3 The authors mentioned above (Van Vleck, Kambe and Usui, McMillan and Opechowski) did not attempt to calculate the line shape but contented themselves with calculating its low moments (f irst and second). This is referred to as the moment method, and l i t t le can be said from these moments concerning the line shape i tself . We show in Chapter IV that our single-spin operators simplify the calculation of these moments. In recent years, the technique of double-time temperature-dependent Green's functions has been used to calculate line shapes directly, and the motivation of this work was to find i f this technique was of value in the paramagnetic resonance line shape problem. As a preliminary exercise, the line shape function for the one-dimensional Ising model with spin H and nearest^neighbour interactions was studied and is discussed in Chapter III. In this case no decoupling of the hierarchy of Green's function equations is necessary. The technique has been applied by Tcmita and Tanaka (1963) to the case in which no crystalline field is present. Their decoupling procedure is designed to make the resulting closed set of equations as tractable as possible. Our attempt at decoupling is designed to be as reliable as possible, but unfortunately gives rise to a set of equations which we have not been able to solve. As a test of the reliabil ity of any decoupling, a theorem concerning the moments of a line shape arising from such a decoupling was used, and is discussed and proved in Chapter V. Decoupling approximations are also discussed in Chapter V. The result of our decoupling is given in Chapter VI and compared with that of Tomita and Tanaka (1963). These decouplings of course lead to a set of delta-functions for the line shape, and Tomita and Tanaka derived a continuous line shape by introducing rather arbitrarily a smearing of delta-functions. This aspect 4 of the problem is also discussed in Chapter VI where the general Ising model is used as an illustrative example. For a treatment of the problem in which the individual spin energy levels are not equidistant, the individual spin operators previously mentioned were used; i t was for this problem that they were originally introduced. This problem proved intractable and the reasons for this are discussed, also in Chapter VI. 5 CHAPTER II General Formalism 1. System Hamiltonian Our physical system consists of a crystal l ine sample containing N identical paramagnetic electronic unit systems, or "spins", possessing magnetic moments. It is assumed that the sample has been placed in a constant magnetic f i e ld \jQ and a high frequency osci l la t ing magnetic f i e ld H-|(t) = H-j cos fe?t) perpendicular to H Q . Each spin is defined by a "spin-Hamiltonian" H(0^ which incorporates the effect of the external constant magnetic f i e ld and of any arbitrary crystal l ine e lect r ic f i e ld that may be present. The osci l la t ing magnetic f i e ld induces transitions between the energy states of the spin system, and experimentally, we find that energy is absorbed from the osci l la t ing f i e l d . The "line shape problem" cons.ists of finding the power absorbed by the spin system as an analytic function of <0 , the frequency of the osc i l la t ing f i e l d . In addition to the interaction of the spins with the two magnetic f ie lds and with the crystal l ine electrostat ic f i e l d , there are weak interactions within the spin system. Were these latter interactions absent, the line^shape would consist only of widely separated delta-function peaks; these "spin-spin interactions" serve to "broaden" the l ines , giving them their shape ( i . e . , closely spaced delta-functions of dif fering strengths occur, and they are not experimentally resolvable). Mention must also be made of interactions between the spin system and the la t t ice . We take i t that the temperature is low enough so that the effects of la t t ice vibrations can always be neglected. The Hamiltonian describing our paramagnetic system may be written 6 H r - - (2-1) where ^ T ^ V ^ ( 2_ 2 ) ^ » ^ (2-3) n 4 ° ' » ^ (2-4) ' 1 4 ^ = 2 Z T 4 - ; ° ^ is the Zeeman, or unperturbed Hamiltonian with i label l ing the (2-5) lat t ice sites a-ZSl W!) 1 s the spinr-spin interaction Hamiltonian il including di interactions including dipole-dipole and isotropic exchange (2-6) - M K c e a o t i s the external perturbation, with M the magnetic moment of the spin system. * (2-3) and (2-4) are weak-interaction assumptions. We assume that 1*^ possesses axial symmetry, the axis being the direction of L"L> , which we choose to be the z -ax is ; and that the eigenstates of 14^ are also eigenstates of the operator corresponding to the z-component of the spin. Before discussing the line-shape function cr or i ts moments, we must modify the Hamiltonian in two ways. The term 04-^  in the Hamiltonian must be "truncated" in order to eliminate the weak "secondary" lines from o~ <A>). The argument for this has been given by Van Vleck (1948). This truncation consists in keeping only that part of T-r^ (we wil l cal l i t 14^  ) which commutes with "H^ ; i . e . f V ^ ] - O (2.8) 7 We further define Then - O (2-10) Also, when a crystal l ine e lect r ic f i e ld is present, and, thus, the energy levels of an unperturbed spin are not in general equidistant, must be modified in such a way as to cause transitions between only two levels of a spin. This must be done in order to ensure that other "primary" lines of comparable intensity to the one under consideration, are not included in the calculations* The necessity for this further truncation was pointed out by A Kambe and Usui (1952). We let represent the modified >f^, and we put a circumflex over i ts associated operators. When no crystal l ine f i e ld is present, and He is in the ( -Z)-direct ion, the Hamiltonian 'W'^c-f. 1 S Qiven by equations (2-9), (2-5), and (2-6) with J d (2-11) or fail. ..f. - t <\ s~s+) + c : j s * s * <*-'»> where C..= A + B., I J y 1 J V Y +±va/lV^4"'^ (2"13) fa is the direction cosine of the line of length r ; joining spin i to 8 spin j r e l a t i v e to the z - a x i s ; A ; . are the exchange i n t e r a c t i o n constants ; g ^ a n d g 1 ( are the g - f a c t o r s perpendicular and p a r a l l e l to the symmetry a x i s ; {!> i s the Bohr magneton; and | . j s (2-14) 0 o - frf I > 0 i s the unperturbed absorpt ion frequency. Where a c r y s t a l l i n e e l e c t r i c f i e l d . i s p resent , the unperturbed spins have, in g e n e r a l , non-requidistant energy l e v e l s . We consider here only those spec ia l s i n g l e - s p i n Hamiltonians which have the property t h a t , i f are the simultaneous e igenstates of Si and 04t with eigenvalues / \ and --Is r e s p e c t i v e l y , i . e . (2-15) then 1 1 1 A (2-17) Equation (2-17) s i g n i f i e s that in the extension from the n o - c r y s t a l l i n e f i e l d case to the c r y s t a l l i n e f i e l d c a s e , the order of the unperturbed energy l e v e l s , which in both cases are l a b e l l e d by the eigenvalues of remains unchanged. The statement about non-equid is tant energy l e v e l s than means that i t i s not genera l ly true that when \ k ^ . The 14^' given by equation (2-12) does not any longer commute with >T = V-L where >K i s def ined by (2 .16) . It must be truncated 9 s t i l l further. The truncation can easily be done i f one defines operators by the relation From definit ion (2-19), equation (2-16), and.the well-known commutation relations sat is f ied by the ordinary ladder operators J>j , the following equalit ies can be derived and are found to be useful: LAl t Sj^J s 7 R A* (2-20) Ax^I' = AlA 'A+i (2-22a) AA4, '> ' S F O - I A'> A'A^I (2-22b) A> 4 v ' = ^ A + I A A A A -H (2-22c) A> At- = £V< A: (2-22D) With the use of equations (2-20) to (2-22), we find for the crystal l ine f i e ld case _ + i V - * (2"23) where means summation over a l l ordered pairs (X"^') # I am indebted to Dr. K. Nishikawa for the suggested use of these operators. 10 satisfying I y H -TJ/ - / A">I ' 1 (2-24) Using (/>$)* = AM (2-25) which follows from definition (2-19), one can verify that i'&Lj^c-p, ^s Hermitian. (The asterisk will always be used to denote the Hermitian conjugate). A Hamiltonian which we consider at some length in a later chapter is that of the Ising model. This Hamiltonian was f irst proposed by Ising (1925) to describe certain ferromagnetic systems. It turned out not to be valid from a physical point of view, in that i t does not describe, with any accuracy, the interactions taking place within any known spin system. However, i t is invaluable from a mathematical point of view, in that i t is a specialization, of equation (2-12)>which is easy to handle. The model Hamiltonian is given by (2-1) , (2 -2) , (.2-5) and (2-6) with ^ = - O 0 S ^ (2-26) ^x,M>-^ r^ ' (2-27) where the cfy are the exchange interaction constants. It is seen that (2-27) is just the general paramagnetic resonance Hamiltonian (2-23) with the A ;j and the C\\ formally put equal to zero and — n respectively. In many instances a further specialization is useful, namely that of putting R£ L, \ nearest neighbours £ - / » (2-28) tj ^ Q otherwise 11 The Hamiltonian with this restriction on the £y will be called simply the "Ising model Hamiltonian". When this restriction is removed, we will specifically refer to the "general" Ising model Hamiltonian. These are the only Hamiltonians that will be considered in this thesis. A It remains now to exhibit ^ x for these various cases. We have, from.equation (2-7), A. (2-29) as the oscillating magnetic field is taken to be in the x-direction. No truncation of M x is needed for the case of equidistant unperturbed energy levels, and so in the absence of a crystalline, electric field (and this includes the Ising model cases), However, for non-equidistant energy levels, i.e. in the presence of a crystalline field, one must use 2. where ^ — means summation over all A such that / A + 1 - T 7 - U ) , (\6^ • (2-31) A -and u), is the unperturbed frequency of the primary line under consideration, not necessarily given by (2-14), but always positive. It is of practical interest to know that (H )M ^4 is a special case °f (H)c,jf, , and that (At)^ c.-f. i s a special case of ( f^\x)c,i^i . To show this, one needs the help of the relations (2-32) X t C ^ d l X (W» (2.33) 12 where in (2-34) i t is assumed, natural ly, that i^, and / ^ ( exist (see equation (2-24)). (2-34) is a consequence of (2-33) and the definit ions of the sets <3i and,6, .. Equation (2-32) follows immediately from the definit ion (2-19) of the <d.J . In developing our formalism for the l ine shape function, then, only the - , i Hamiltonian, describing the crystal l ine f i e ld case need be taken into consideration. 2. General Formalism of Absorption Line Shape. We start from the well-known relation between the absorption constant <T"(iS) and the magnetic suscept ibi l i ty ?£(<S): (T(S) - coyest. O l ^ (u) t (^>0) , (2-35) In the Kubo-type formulation (Kubo, 1957; O'Rourke, 1957) "X (t^) is written t o - ^ ' < [ > j x (t)> < 2 " 3 6 ' where we have used the following notation: Mx : the x-component of the total magnetization operator O f t 1 f t * O f l f t i f l i - i ) , (2-37) 1+ - wi l l henceforth be called the Hamiltonian of the system <<- > * a u (e. -J ; a= u / k a : Boltzmann's constant, (2-38) T : temperature of the system at the time t = - *0 . The following assumptions were involved in these formulas: (1) The external magnetic f i e l d H^(t) has been applied in the x-direct ion. 13 (2) The response is looked at in the x-direct ion. (3) H-j(t) has been applied adiabatically starting from the distant past: lim H^t) =0 . (4) The system was in thermal equilibrium at t = - oo . (5) The Hamiltonian associated with the osc i l la t ing f i e ld is of the form - £ . H , £ ) . (2.39) (6) The o s c i l l a t i n g . f i e l d is suf f ic ient ly weak that i ts effect can be -treated in f i rs j : order perturbation theory. One normally defines the paramagnetic resonance line shape function to be not V~(\S) , as given by (2-35), but g (u) as given by g 6JY* cow/f. Z"*, y.(iS) ( ( l J>o) (2-35a) where the constant is such that ^^j) is normalized to unity. (Cf.McMillan and Opechowski (I960)). It is of advantage to keep only that part of X6*0 as defined by (2-36) which vanishes for |*)<0 • Employing equations (2-8), (2-'9), (2-37), (2-30), (2-21), and (2-31), and the well-known identity c V c * = Z L fcrf (2-40) where [.„ [ F Q fi] fi] f F p" (2-41) one has ft* f t ) . f i W | ! l t * . ^ ( V ' . W E ^ g J ' ^ f l . ) * 14 With the use.of the identity (2-43) which follows from the.cyclic properties of the trace for.any operators A and B < C ^ B ( t ^ > - < t ^ - 0 , * > ' (2-43) j and the relation . ^ M * 7 * ^ ^ ' ( 2"4 4 ) equation u-oo ;eanDe rewritten as TO MOWS: + L ^  , Wt ft ^fS;,^^ - 4 * 4 _J/ W j / ^ / S ^ C ^ I (2.45, First, notice that in the second integral, Sj = £1 ^ M V I may be replaced by ZL (see Appendix I). As the system is definitely quantized along the z-axis, the first and fourth integrals vanish automatically. Moreover, the third integral provides a negligible contribution to % (\S) for u) ^  0; this follows from the weak- -interaction assumption:(2-r3)° i f one puts ^ = O , the integral gives a delta-function peak at u> - - t00 . Inclusion of >rc0 serves to introduce many,other delta-function paaks into the region around - U^> , with very l i t t l e overlap into the region of positive u)- . By a similar argument, the second integral is seen to provide a negligible contribution to in the region of negative 6J .# # This argument is_ equivalent to that of McMillan and Opechowski (1960). They argue that i f >}»> is not too large, < EA implies E^ ,: < E^k except for a negligible number of pairs of states l*,l)tftjk) where >|JW //*,kj = /O and -4 l^tk) (k = 1,2,...^, the degeneracy of E^ ). 15 Thus, i f the magnetic susceptibility is redefined in terms of the second integral alone, i.e. then (2-35) may be rewritten without, the condition " t*>><? ".Further, one always has (2-47) where is any function of u) The function /(<S) as given by (2-46) is closely related to the temperature-dependent double-time Green's function. These Green's functions were first introduced by Bogoliubov et al. (Bogoliubov and Tyablikov, 1959; Zubarev, 1960). The Green's function <t"A I B^. of the operators A and B is a two-branch analytic function of E defined everywhere outside the r»al axis C . IU U t < [ * f t ) , Q > ; X , £ < o By comparison of (2-46) with (2-48) one.can immediately see the relation * (2-49) where we have used only the.identity (2-43). Note that the above Green's function is of the form ^A|A^ (by (2-25)). One can show that the Green's function ^ A l B ^ as defined by (2-48) is analytic both in the upper and lower half-planes and that i t is found from the hierarchy of equations , , v e « f l i ^ = ( 2 _ 5 0 ) 16 Our problem is then to solve this hierarchy with the condition that the solution be analytic in the domain for which the Green's function is defined, that is, outside the real axis. The actual procedure of solving this equation approximately will be discussed in Chapter V. For further details of the Green's function method, we refer to Zubarev's article. 17 CHAPTER III # Green,'s. Functi.Qtis....and--th.e. Isi ng - Mode-1 ... In some simple.cases i t is found that the hierarchy of Green's Function equations (2-50).decouples automatically after,a finite number of steps. One such case is the.problem of finding.the amplitudes and frequencies of the three component the absorption band.of the one-dimensional Ising model, consisting of an arbitrary number N of identical.spins arranged in a ring. It is assumed that each spin has spin quantum number h, and that only nearest-neighbour interactions occur. The Hamiltonian of this system is given by equations (2-26), (2-27), and (2-28), i.e. . A/ .2: 4" r * 14 = - U - cl j^y J k 3 k+i (3_D where h = k)o as equation (2-14) £= exchange interaction constant I and k, label.the lattice sites, with the (N+l) s t site identical with the 1st. The line shape function ^ (iS) is again defined by equation (2-35a), which, through equation (2-49) specialized to the no-crystalline field case, leads to A ( U ) ) O L &W X * tfeta-tg. (3-2) # I am indebted to Professor W. Opechowski for his suggestion that this problem be considered. 18 where c.(s 3 ^ <i^ /i-;^  0-3) The first three Green's function equations are then (£-U W ^ a < ^ > ^ * « t f W ( 3 .4) (3-6) The last term in equation (3-6) arises since (3-7) is a consequence of the relation It is now seen that equations (3-4), (3-5), and (3-6) form a closed set of simultaneous equations which can easily be solved for Cj0js ; that i s , there is no need to write any higher-order equations. Since this state of affairs appeared without .any.decoupling.becoming.necessary, we say that the hierarchy decouples automatically at the tlfird stage. The solution is 1 (3-9 where we have introduced the notation = <S& (3-10) noting that ^ l " ^ » ^ 2 ^ , and 3^ are, independent of i because of translational invariance. The line shape is then given, with the use of (3-2) and (3-3), by = -^[(<»-*<*>)r(>>-K) + (±<p+a<& + *<ii)P(u-^t)] (3_11) which is a sum of three delta-functions at specified frequencies, with amplitudes given by linear combinations of thermal averages. It now remains to calculate these thermal averages. Since several authors (e.g. Huang, 1963) have derived explicit representations for the partition function Z, we will take i t that Z is known, and use Huang's result Z : = ( A 4.) * (/-/ (3-12) 20 where A ± ]/cosk^l) -^'^^k(i^J ( 3 - 1 2 ) We immediately get <1> and <2> by direct differentiation from the formulas ' (3-14) 1 = A/-2: "^S4) (3-15) The thermal average ( 3 ^ is obtained as a linear combination of 0 ^ and ^2^ by the use of the following.general identity (see, e.g., Zubarev, 1960) (3-16) Putting t = t' = 0, A = S. + , and.B..= in equation (3-16), and using the result (3-9), we get - J - <i7 - <y-^<3>. - i (<•> - ^6-> + 4<3>) - p ^ ^ - l ( 3 _ 1 7 ) o r (3-18) where DQ, D-J , and D 2 are simple functions of h and & . In the derivation of (3-17), use was made of the relation (3-19) 21 Equations (3-14), (3-15), and (3-18), when combined with equations (3-12), (3-13), and (3-17), give us, when substituted into equation (3-11), an explicit,representation of the absorption band of the one-dimensional Ising model.. The result corresponds exactly with that found by Y.Y. Lee (1961), who had quite complicated summations to perform. The latter arose from his combinatorial approach to the problem, and necessitated the finding of suitable.generating functions. He simplified the combinatorial problem by specializing to the case in which N = 4P -1 (P=l,2,...) and then considered his results valid for large N. In our approach there is no need for this specialization, as the complicated combinatorial problem is avoided. The advantage of our method lies in the identification of the amplitudes of the resonance lines with easily recognizable thermal averages, and in the ease with which these averages can be calculated, from a knowledge of the partition function, through differentiation and further application of the Green's function technique. The thermal average <3^ cannot be obtained directly from Z; nor can higher-order thermal averages. A direct method of finding thermal averages, using difference equations, is given in Appendix II. 22 CHAPTER IV The Moment Method 1. General Discussion In the problem of the general paramagnetic resonance line shape, unlike that of the one-dimensional Ising Model, the hierarchy of Green's function equations (2-50) does not decouple automatically at an early stage. The solution of this problem by the Green's function technique then involves looking for a decoupling procedure which gives a closed set of equations satisfying the following requirements: i t must be (a) sample enough to be tractable, (b) complicated enough to yield far more than the small number of isolated delta-function peaks obtained for the Ising model, and (c) "faithful" enough so that i t furnishes a line shape function that is resonably close, in some approximation at least, to the experimentally-observed line shape. The difficulties inherent in the search for such a suitable decoupling procedure are representative of those encountered in other approaches to the line shape problem. Consequently, a method has been evolved which enables one to discuss the shape of the resonance lines without actually having to find an analytic expression for the line shape function. Known as the "moment method", this procedure consists in defining a line shape function in a formal way, and in calculating its first few moments in various approximations. It was introduced by Van Vleck (1948), and used by Pryce and Stevens (1950), Kambe and Usui (1952) and McMillan and Opechowski (1960). Kambe and Usui derived quite general exact expressions for the moments, and these were applied by McMillan and Opechowski to the case in which a crystalline electrostatic field is present. In this 23 section, a derivation, different from that of Kambe and Usui, of the expressions for the moments is given. We shall show that i f a normalized line shape function g r t A £ u d ) , defined by with (4-1) satisfies the conditions that A* = B (4-3) and that a LiS) - O u)<0 (4-4) ye,* > * *t h then its p- moment about the o r i g i n , , is given by /y- <f{^KB3>/<f",Bl> p=0,l>2,... (4-5) With the use.of (2-40), (2-43), (4-3), and the identity <£»ct)0 = <[B*,fl*fc\]> ( 4-6 ) we can immediately write, from (4-2), x (4-7) 24 We now make-use of the characteristic function M(t), defined by M(t) = / ^ e.* — ^ B A M (4-8) which possesses the inverse Fourier transform Repeated differentiation of (4-8) with respect to t, i.e. gives, simply, the Taylor coefficients in the expansion of M(t) about t=0, i.e. M(t). • (4-H) Substituting from (4-11) into (4-9), we get - .;jt c.t)f, , , ' ° (4-12) Comparison of (4-12) with the alternate expression for g (iS) given by (4-1) and (4-7) yields = (constant, independent of p) X <C£ (/9 ;'H} P ) B j j ^ , (4-13) The constant is determined by the normalization requirement on g (^) , resulting in the required expression, (4-5), for the moments about the origin. 25 If, further^ ^ can be expressed according to (2-9) as where (2-10) holds, and also [ f K ^ ^ J ~ ^>o?\ (4-14) then <I1^^> (4.15) are the moments about the unperturbed frequency . (We call the Mp "central moments".) For T = O (4-16) Also, by induction i t can be shown, using (4-14) and (2-10), that f i \ f t } p = i i t r < o / " 7 , (4-17) Combining with (4-5), we get t. fc^r«,- £ %»:KCI«W\tf> (4.,8) T 2 0 26 Taking p=0,l,2, successively, we see that formula (4-15) follows. 2. Application to Paramagnetic Resonance Line Shape Problem. For the paramagnetic resonance .line shape problem, one can immediately proceed to calculate moments in various approximations by substituting f\ - « s £ — - X (4 -19) and 1 3 * * * * (4-20) into the expressions (4-15) for the central moments. The expressions (4-19) and (4-20).for A and B come from equation (2-46) for ^C(^). One takes commutators with (equation (2-23)) according to the rules (2-20) and (2-22), and takes traces in the most convenient representation. This procedure is straightforward. McMillan and Opechowski (1960), starting from formulas first derived by Kambe and Usui, were able to cast expressions for the first and second central moments into a form suitable for computation. They then applied their results to the specialized problem that we have been considering. Their only initial assumptions are given by our inequalities (2-3) and (2-4). The procedure for their calculation of the first central moment, M.|, will here be given in outline. They start with H, - _ . ——r- (4-21) 27 where ^ A A M = M+ +M_ is the component of the magnetic moment of the spin system in the direction of the oscillating field, truncated so as to single out a particular primary line. A A M+ and M_ are defined by ' ^ , i l M . I / j , k V l 0 0 * ( 4 . 2 3 ) where /=^ ,0 satisfies the eigenvalue equations (4-25) They then rewrite equation (4-21) in terms of Q and M rather than tr , A A A M, M+ and M_ . This is done with the use of projection operators P^  defined such that (4-26) With the use of the observation that i f "H1"^  is not too large, then E-<A ^  implies ^ ^ t k except for a negligible number of pairs of states W^'O IA 1%) » t n e definitions (4-22) and (4-23) of and M_ are changed to 2 8 Q o'the.-ruj!Se. ( 4 - 2 7 ) ^ . i lA - lAkV / K W M l ^ k ) ; ' 0 oik ( 4 - 2 8 ) Then * = ^ * P A ( « , , where rt+ = ^ N P ° < ( 4 - 3 0 ) A = 2 L ( 4 - 3 ! ) means summation over all values of <?< and A wherein - ^ = U)0 ( 4 - 3 2 ) T 3 Equation ( 4 - 2 1 ) becomes M, = d " ' / S (4-33) " hB = ^ ^ (t^S.MP.M-t^^P^MP.M/ (4.34) McMillan and Opechowski make use of an approximation technique used by Pryce and Stevens, and also by Kambe and Usui. This technique depends on the weak-interaction assumption ( 2 - 3 ) . One expands all exponentials involving ^ as follows 29 provided [ c , ^ ^ ) C , ^ c J ] ^ O ' and k e e p s o n l y the lowest-order terms in the square brackets. Here Cj and Cr, are any operators such that the above provision is- satisfied. We call the approximation which consists in keeping only the first £ terms in the brackets, the th " approximation of McMillan and Opechowski", and denote i t by (MO)^  . Evaluation of formula (4-33) in (M0)j consists then in replacing by b * • The analysis is done by McMillan and Opechowski for the case when the unperturbed spin has R energy values ap ag, ..., a^ (the equivalent of our ) which are all non-degenerate. In this connection, a set is defined such that i f T is in 6>4 , there exists an integer T / such that Q-r' " ~ ^° (4-36) In the first approximation the term Q (4-34) can then easily be reduced to the form |< ft. . . 6 - d - t ^ S L t -S-f lUHMl^kjl . (4-37) In the above form, the calculation of 8 becomes a combinatorial problem, and involves doing summations like + hR- hi where are the unperturbed eigenfunctions of the i t h spin, with eigenvalues Qy. so that T6. 30 i 1 (4-39) The cause of the combinatorial problem is in the degeneracy of the E ^ . The degree of complexity increases when expressions like and those entering the higher moments are considered, and when approximations higher than the first are used. The combinatorial problem is avoided for the special case that we consider by the introduction of the operators defined by (2-19). To see how these operators simplify the calculations for our special case, we show how the more general case of McMillan and Opechowski is simplified by the use of a different kind of projection operator; these projection operators, first introduced by Kambe and Usui (1952), are then related,for the special case, to the operators ' v We define projection operators which project onto eigenstates of ?*• rather than the operators P v which project onto eigenstates of the full Zeeman Hamiltonian, by (4-40) The relations (4-41) and 0 (4-42) follow from the definition (4-40). (4-46) 31 Note that in these equations X no longer labels eigenvalues of , ..Co) but only eigenvalues-of according to equation (4-39). A A One can then show that the following expressions for M+ and M satisfy the definitions (4-27) and (4-28): M. - P/M, p; ( 4 . 4 4 ) where = M j (4-45) One also has ^ - JIS : z = _ p,i ^ p/5 pA;. where ^> means summation over all A j , / i y such that ( 4 " 4 7 ) A One can prove the above statements by considering matrix elements of M+, A —, M_, and (as given above) between the orthonormal product eigenstates \KkK...L) - I^X^\*"*IX»)* (4-48) Here /^^ , for example, is an eigenstate of with eigenvalue Q^ . The product eigenstates (4-48) are themselves eigenstates of <y^{9) = >f; with eigenvalues a^ + a^ +...+ a ^ = , and are linear combinations of the l/rM) i ) defined by (4-24) and (4-25)3for f i x e d . Let us now evaluate (4-49) 32 the first approximation part of .the...denominator of given by equation (4-33) with the use of the PJV . Using (4-43) and (4-44), 3 • (4-50) From (4-41) and translational invariance, s = ifH. i x [b 5 (?; H: p j M- p j (4-51) Let us consider only the first term of (4-51), and expand the trace in the orthonormal product eigenstates (4-48), to) ~rec«^ >,)A,)..,)A/v (Pr:HiPj.MiPj)lK^.-.^)} T 6 . Gi v where a., 4- I - . + -a (4-52) (4-53) One derives B (second term) in a similar fashion to get 6 = A/Z^' d-^OZL- iMM^MP (4-54) 33 where (4-55) and use has been made of equation (4-36). There was no need to solve a combinatorial problem, and result (4-54) agrees with McMillan (1959). AIT other traces can be done in the same manner, even in higher approximations. For the special case when (a) }4 • has axial symmetry; (b) the external L magnetic field tl° is directed along the axis of symmetry (Z-axis); (c) the order of unperturbed energy levels is preserved in the transition to the no-crystalline field case (equation (2-17)); and (d) the relation [S-J^ ~ ° holds, we may write (4-56) M_ = /15a. (4-57) ^ _ c j_ a s 9 i v e n by (2~ 2 3) V * P where X , <&±y , » and Gt are as defined in Chapter II. A direct comparison between equations (4-56), (4-57) and equations (4-43), (4-44) cannot be.made because of the differences in the definitions of G^  and G^* and in the labelling procedures. However, -it may easily be verified that ^ A M+, as given by equation (4-56), conforms to the definition (4-27), and that M , as given by equation (4-57), conforms to definition (4-28). It has been checked, for the specialized problem wherein conditions (a) to (d) (above) hold, that formula (4-15), in the case of an absorption line from the ground state to the first excited state of the unperturbed spin, yields the same final results for the first and second central moments as were obtained by 34 McMillan and Opechowski (1960); the calculations were done in the first two approximations of McMillan and Opechowski, for the cases of effective spin S = % and S = 1. 35 CHAPTER V The Moment Theorem 1. Relation of Moments to Terms in Hierarchy of Equations. It is interesting to take note of the identification between the raw moments (moments about the origin) and the first terms on the R.H.S. of the Green's function hierarchy of equations (2-50). The latter may be written P = 0,1,2,... Comparing with the expression (4-5) for yt^p , one sees that, apart from a factor 3TT the identification is complete. It is assumed that conditions (4-3) and (4-4) are fulfilled. Moreover, i f condition (4-14) holds,the hierarchy of equations (5-1) is easily cast into the form ? I B > £ • £ <a^ ifB7>+i(f>mns% ( 5 . 2 ) p = 0,1 ,2 It is in this form that the Green's function hierarchy of equations will be used in the remainder of this thesis. Apart from the same factor /^A.B]^, one identifies the first terms on the R.H.S. of the latter equations with the central moments, M p (equation (4-15)). One may ask whether this identification can be made directly from equation (5-2) and the relation - r „ (5-3, I—I 36 between the Green's function <^/9)B~/^ and the line shape function ^ proof that this can be given in Appendix III. This identification is, of course, valueless i f one wants only to calculate exact moments; the Green's function technique need not be considered at all for this purpose. However, very often the Green's function technique, coupled with a certain decoupling procedure, is used to derive an approximate analytic expression for the line shape function, and i t is asked what the associated moments are and how closely they approximate the true moments. The integrations involved are often extremely difficult to carry out. A theorem, closely related to the aforementioned identification, can be proved which provides a relatively simple method for finding the approximate moments, and eliminates the necessity for performing any integrations. Before this theorem is stated and proved, however, an outline of decoupling procedures as they are normally followed, will be given. 2. Decoupling Procedures First, rewrite (5-2)'.as follows: e. Gnp ~ ' p +• G» p + | (5-2a) where we have replaced E ^ U)„ by £_ (5-4a) $Cfi^f\3%-by a? (5-4b) and < [ { M f B ] ) > by T" p (5-4c) A method one might follow would be to write the f i r s t q equations as they stand, and in.the q+lst to replace by a linear combination of the q+1 preceding Green's functions, the coefficients (usually thermal averages) being independent of E; i.e. 37 e. £>, - T +• (5-5a2) i i i £ - ~ T " + A (5-5a ) f 1 ^ q We call this "decoupling at the (q+l)st stage". A closed set of linear equations is obtained which must be solved'for C)0 , which is simply related to (to) .. (It will be obvious throughout when Oi0 stands for the exact G 0 and when i t is used to mean the approximate value obtained on solution of the decoupled set of equations.) This method is fouffd" to be unsatisfactory in most cases. Usually G>9 is made up of more elementary Green's functions Q0 such that (5-6) e.g. and a 0 = ^ I L ^ L A ^ { ^ ^ - ^ 1 % (5-7a) Let us write a separate hierarchy of equations for each of the elementary Green's functions Ln 6i) ^ (»0 r - v a. (5-83^ £ U, - ', ^ (5-8a2) ^ 6 ^ = T ^ * ^ (5-8a q + 1) 38 Equations (5-8) represent Tl hierarchies of equations, each hierarchy labelled by a different ^ . They differ from equations (5-2a) in that different A and B have been used in the defining equation (5-2). The T are the V . The U r thermal averages associated with the Green's functions CL (V^o) are not meant to be elementary Green's functions, but are defined by equations (5-8), through the given G0 and equations (5-2). (There is no difficulty (<*) in picking out which terms are to be grouped with the T^ _ and which with the Cv in writing equations (5-8); all the Cv vanish in the limit as C — > °° , whereas the T ^  are independent of £. .) One might want to decouple each hierarchy as follows: where independent of o< (5-9a) Each set (labelled by o4 ) of (5-8) is now closed, and can be solved for Gig Then the sum (5-6) can be taken over all the solutions. The above procedure is not equivalent to that leading to equations (5-5), for i f we sum equations (5-8) over ©4 , we get the set of equations ^ a, - T + ^ ! i I (5-10) Because of condition (5-9a), the set (5-10) is not closed, as was the set (5-5). This method of decoupling is also unsatisfactory in most cases. What lly h, usual appens is that G>, is also made up of elementary Green's functions tf\ Z - (5-1,) 39 and one might want to decompose C^, in terms of the u, ' with different coefficients for each 60^6) pair. So separate equations for are needed. Continuing on in this way, we find that our most general set of equations can be written /* (5-12ai) r 'a. (5-12a2) where each Green's function that appears is elementary. Here Gi , the result of some specified decomposition (5-12b) is some linear combination of all the other elementary Green's functions. Equation (5-12a,) represents T"L equations, one for each value of oi ; equation (5-12a2) represents ( TL TTL* ) equations, where TTl is the number of 3^ -values; and so on. The number of equations in the set (5-12) is equal to the total number n of elementary Green's functions that appear. We call equation (5-12) our "closed set of decoupled equations". For convenience, we relabel all the elementary Green's functions in (5-12) by the single subscript . Equation (5-12) is then rewritten (L G)^ * -r- J>Z_ G, o**!,!,...,* (5-13) 40 where (— are coefficients which may be zero, and are independent of E. The T,^  are the thermal averages associated with the &^  , in accordance with equations (2-50). In order to identify 0)0 , we call S o the set of all values of o< such that A ° = S T . ^ • (5-i4) It is then possible to identify Cn, i £>v ,,, t A| i in terms of the and the Cf . For, summing (5-13) over o<£ «Te , we get e- J E L = e. G 0= 2 1 7 7 -r- 2 2 £ f 6*, (5-15) where the summation convention (repeated indices are summed over all possible values) has been used, and will be used throughout the rest of the chapter. Comparing with (5-2a) for p=0, we get <^6.S« (5-16) a > = f r z * & ( 5 - 1 7 » We also have, s t i l l using equations (5-13), Comparison with (5-2a) for p=l gives 4 I (5-19) G v = X L C ^ U (5"2°) 41 In general, for ^ - ^ w e c a n write (5-21) c+t-So f ' ! f ~ (5-22) r - ^ > — rf" a as up to this point the decomposition (5-12b), which represents our (q+l)st state decoupling, does not affect the original hierarchy of equations (5-2a) For f)>^ » the above identifications ((5-21) and (5-22)) cannot be made. However, defining primed quantities 77 and Q by A ^L£9 (5-23) 7Tr. C ^ T< (5-24) for all i , we can continue equations (5-15) and (5-18) into the infinite set '/ + GL' ail l e- \ = U + Ht*. , » » L <5-25) which is based on the closed set of decoupled equations (5-13). Note that -J2'= 1 - (4 if) (5.26a) and 0>[ ~ . (5-26b) 42 It is of interest to note that one can equivalently define and G e by equations (5-25), (5-26a), (5-26b), and the decoupled set of equations in which one is interested, e.g. (5-5) or (5-13). Equations (5-23) and (5-24), our previous definitons of and Q, , may then be deduced through the use of equations (5-13), (5-21), and (5-22). Let us consider, as an example of this alternate definition, the special case of decoupled equations (5-5). Looking at equation (5-5a ,,), we define / q+i by f | T F * (5-27) and write a Green's function equation for i t , using equations (5-5) to do so: e~ ^+» - e. ^ olt + e. ^ - o (5-28) +- ^ +- d^ 2^ alt A J L I ^  + X* 6 ^ +d<di)G,{ l s 0 _£ = 0 5 where in (5-28) we define d_j, d_2,.to be identically zero. We then define by ' 5+1 - J> 11 (5-29) A N D G<J+SL by ^ (5-30) We next write an equation for 6^  + 3 . to get, in the same way as above. 6 ^ 1 = / + cLdf*) lJi (fereen's functions) (5 and define by / ' = >~ (M.,-h d*<J4)Tj (5-32) 7 * ° and so on. / It is this latter definition of the ^ that is useful in practical applications of the Moment Theorem, which we are now prepared to state. 3. Statement and Proof of Moment Theorem Theorem: The ^ moment M^  associated with the line shape function derived from the closed set of decoupled equations (5-13) is given by H - "Tr'/77' (5-33) for all f? . 44 Corollary I For (q+l)strstage decoupling, the ..first q+1 central moments MQ, M^ , Mq are exact. The proof follows immediately from the definition (5-23) of and the identification (5-21) for .( ^  . Corollary II If one wants a line shape function which gives the first.q+1 moments exactly, one cannot afford to decouple before the (q+1) stage. The proof is analogous to that for Corollary I, with the additional remark that a thermal average cannot in general be expressed exactly in terms of lower-order thermal averages. Proof of Theorem The proof will be part for a simple case f i r s t , and then in full for the general case. Case (a): Closed set of equations given by (5-5). Equations (5-5) form a special case of the set of equations (5-13). For proving this case i t is more instructive to use the language of the special case, which means using the definitions of and G\€ given at the end of Section 2. Since equations (5-5) represent q+1 linear equations, each V K (Oittj) can be written, by Cramer's Rule, as a quotient of determinants, the denominator being a polynomial one degree higher in than the numerator. Barring accidental degeneracies, then, each can be written as ' ~ ^ ~ f * 7 (5-34) p S 3 | C» "* fe» p 45 which results from the decomposition into partial fractions of the quotient solution. are certain coefficients independent of E, some of which f may be zero. The A^ are real as is shown in Appendix I ft. for the general case. The Ep are real, distinct, and independent of t . The reality of the £p follows from the analyticity conditions on the Green's functions. Their independence of i follows from the observation that all the Gp have the same denominators in the quotient of determinants. Substitution of equation (5-34) into equation (5-5) yields P P Zl fe-V£e?ff/ = 17 + ( 5 36> Equating coefficients of results in three conditional equations t > 5 (5-37) £, Aj - K j , 0 * 4 < f ( 5 . 3 3 ) Now, using equations (5-3), (5-34), (5-4a) and the asymptotic formula i • _ _ J PfA* L T ? C £ ) where P denotes the Cauchy principal value, we find 46 + -IT^ V, dole ? r Jul p 1 p (5-41) From (5-38) and (5-39), i f 0 ^ ^ ^ , (5-42) and from (5-37) from (5-26a) (5-43) 47 So, for T ^ ^  , equation (5-41) yields r " ^ ' ° • (5-44) If T= q+1, we have P from (5-38) from (5-39) f- Jj.T^ from (5-37) £ = 0 + i from (5-29) (5-45) and so (5-46) 1* from (5-41), (5-44), and (5-45). The theorem is thus verified for ( = 0,1..., q+1 (equation (5-33)). The proof is not continued for higher values of f for this special case, as the notation becomes quite unwieldy. Case (b): General closed set of equations (5-13). For this case we make use of the definitions (5-23) and (5-24) of \JL and From the theory of linear equations, the solution of equation (5-13) is given as a quotient of determinants, the denominator being a polynomial one degree higher in £. than the numerator. Barring accidental degeneracies then, each Ci. can be written as ^ p 48 which results from the decomposition into partial fractions of the quotient solution. n» the number of elementary Green's functions, is the degree of the polynomial comprising the denominator;. are certain coefficients independent of £• , some of which may be zero, and the are real,, distinct, and independent of U . The reality of the £^ a follows from the condition that the Green's functions be analytic on the upper and lower half-planes. Their independence of <?< follows from the observation that all the Cj*. have the same denominators in the above-mentioned quotient of determinants. Substitution of (5-47) into (5-13) yields where the summation convention is s t i l l in use. Equating coefficients of results in two conditional equations: • * ^ all o4 ' (5-49) It is important for the proof that the A . be shown to be real. This .a is done in Appendix IV. Multiplying equation (5-50) by 1, £p , f ^ ) , . . . successively, ancj( using (5-49), one finds a fl A P 49 From equations. (5-3), (5-4a), (5-4b) for p=0, (5-47), (5-40), (5-14), and the reality of the ^ £ ,' the moments of are given by M; = ^"^ &>ol£*^~i, L 1^  ^ p by (5-51) ' / 7 T ' bY definition (5-29) This completes our proof. 5 0 4. Summary on Use of Theorem.. It may be worthwhile here to summarize this chapter indicating how the theorem would be used. Examples of its use will be given in the following chapter. The line shape function can be expressed in terms of a Green's function G 0 and the equation for Go leads to a hierarchy of equations involving higher-order Green's functions C,^ . This hierarchy is decoupled in some manner leading to a closed set of equations which can be solved for the approximate G)0 \ One could then find the moments of the line shape by suitable integrations with the approximate line shape function. The theorem shows that this integration is not necessary. The approximate moments up to the stage q+1 at which the decoupling was done are just the usual thermal averages appearing in the equations and agree with the exact moments. To find the higher moments of the approximate line shape one continues the closed hierarchy of equations. In this continuation no new Green's functions appear; each time a Green's function appears in an equation i t is decomposed with the rule first used. The thermal averages appearing in these higher equations are the appropriate higher moments of the approximate line shape function. This procedure for finding moments has the advantage that i t does not require one to solve equation (5-13) explicitly; one need specify only the decomposition (5-12b) that is being used. CHAPTER VI Resonance Line..Shape Problem 1. Decoupling Metftods At the beginning of Chapter IV, the properties demanded of a suitable decoupling technique are outlined. These are the tractability, degree of complexity, and faithfulness of the resulting closed set of linear equations. It is, in any specific case, a simple matter to determine whether a set of equations is complex or tractable. It is the question of faithfulness that will be dealt with in this chapter. We postulate a faithfulness criterion: one decoupling procedure is said to be more faithful than another i f the first few moments of its consequent line shape are closer to the true moments in some valid approximation. The phrase "in some valid approximation" is inserted, as expressions for the moments are invariably given in terms of thermal averages; and even the thermal average of ->i has never been given exactly. The criterion is practical, for the moment theorem of the last chapter immediately enables one to write down expressions for the moments which are to be compared with the true moments. In line with this practical approach, only the first few moments are usually asked for, as higher moments become increasingly difficult to calculate in any approximation. Our criterion will now be used to investigate the relative faithfulness of two specific decoupling methods, when these are applied to the problem of finding the paramagnetic resonance line shape function associated with the physical system described by the no-crystalline field Hamiltonian (2-9), where 1^: and are given by equations (2-11) and (2-12a) respectively. One method we shall call the excess-over-random (EOR) method; the other, due to Tomita and Tanaka (1963), which is a partial excess-over-random.method,.we shall refer to as the T-T.method. 52 We specialize our physical system s t i l l further by requiring that i t consist of spins having S = . For this case, from and the well-known commutation relations for spin operators, the following relations can be proved: JV JT- s : ' s r = o The significance of the above relations is that any product of two spin operators at one lattice site can be reduced to a linear combination of single spin operators. The EOR method, which takes full advantage of relations (6-2), proceeds as follows: One rewrites the Green's function equations (2-50) for the "ordinary" Green's functions [k\ = f A / B " ] = A/B^- in terms of comulant Green's functions (A/ = (A/B) which are defined by 01 - ('1 (6-3a) rial * </>dl +- <*>('! +• 0 * 1 (6-3b) etc., where the numerals "1", "2",... represent various operators. After 53 the above decompositions, one stops at a certain order and discards the highest-order cumulant Green's function. For example, in second-stage decoupling, one stops after equation (6-3c) and neglects the cumulant Green's function (l23"|, supposing i t to be small. One of course ver i f ies that the operators "1" and "2" i n , for instance,Cl23l , are associated with different la t t ice s i tes , Otherwise, the product of operators "1." and "2" is to be treated as a single operator "12" in the above decompositions. As an example, the Green's function [S- Si Sj j for the case S = must be changed to -^ p [.Sj^j before any decompositions are made. The T-T method proceeds in the same manner, with the one essential difference that products of operators associated with the same lat t ice site are not treated as a single product operator in the decompositions (6-3). The f irst . two Green's function equations, written for the elementary Green's functions, are - S L C ^ S + l ^ Z (6"4a) 4 Note that we in fact want to find x (6-5) These equations hold for any effective spin S. 54 Second-stage decoupling by the EOR method, for the case S=T/2, consists in making the fol lowing replacements: where ^ A . 5 ( ^ J ^ 1 S independent of i by trans!at ional invariance. When any two of p , i , and j are equal, use of equations (6-2) ensures that no new Green's functions are present, and hence that the resu l t ing set of decoupled equations i s closed. Second-stage decoupling by the T-T method consists in making the replacements + ( < f , V > - v V f ^ (6"7a> ^<s-rr><^i^ (6-7b) for al1 p , j , and i . It may be seen.that in the EOR method certa in cor re la t ions , namely auto-cor re lat ions , are retained in the i r exact fo rm.whi le in the T-T method they are approximated. Thus, i f we take p - i ^ j in equation (6-7a), the replacement * / / £ (6-7c) 55 is effectively performed in the T-f method; in the EOR method, the left-hand side of (6-7c) is retained without approximation. Substituting equations (6-6). and (6-7) into equation (6-4), the EOR-decoupled equations are (6-8a) and of w 2» where we have used the notation (6-9) 56 Other types of seconcUstage decoupling may certainly be tried. One type that seems worthy of consideration is due to Callen (1963). It makes use of the fact that may be expressed in two ways in terms of £ ~ a n d £ • for S = i.e. and f,-*s 't-S^'f* (6"11b) -\' is replaced by taL(?s:-rx)+o-A(±-rrs:) ( , l l c ) 57 after which decoupling proceeds by the usual EOR method. Here &<C is an arbitrary parameter which must be fixed by some suitable physical c r i te r ion . The Callen method wil l not be considered in this thesis , as the calculations become very long and involved. Other possible methods have been looked at and tested on the Ising model Hamiltonian, but they were found to be of far less value than the methods being considered here. From Corollary I to the Moment Theorem, the f i r s t moment, derived from both the EOR- and the T-T-decoupled equations (6-8) and (6-10), is exact, since q = 1. Let us look,then, at the second moment. By use of the procedure suggested by the Moment Theorem, one arrives at the following expressions for M2(T-T) and M2(E0R) : (6-12) M 2 ( EO R) . M 2(T-T>; • • ; , 3 £ m*t«$T> N ) (6-13) The correct expression for M 2 , denoted here by M2(C0RRECT), is derived from equation (4-15) with (K = SZ- S> (6.14a) IB - X , (6-i4b) 58 It is given by ^(CORRECT). = - ^ C ; < f / f y ( a < ^ V * > +<^;*:>-<fs;s,9 ( 6 , 5 ) To compare these formulae for the second central moment, we use the McMillan and Opechowski approximation technique which was described in Chapter IV in the section after equation (4-35). The following relat ions, val id to (M0)2 and for S = are needed for our proposed comparison. (6-16) is the first-approximation part, of » and is independent of i , 59 is independent of i by translational invariance, Substituting the above.expressions into (6-12), (6-13), and (6-15) the expressions for M2 become, in (MO).^  , - ^ { V £ ^ + A A ( B ^ < (6-i 7) in (M0)2 M , (T-f) = - R X £ */ «-/?) - H i r ^ ' / 4 ) ^ ^ and M2(E0R) = M2(C0RRECT) = R.H.S. of (6-18) -f 3 / y 3 (u^m ./^ * +• * ( / ^ ) ] t (6-19) And so the second-approximation part of M2(T-T~) is found to be in error. Moreover, this error is not 0(1/N) of the correct second-approximation part. This may be seen by comparing, for example, the terms and 60 The f i r s t term is not 0(1/N) of the second even though the latter involves an extra summation. This is so because and Cj^ fa l l off as rapidly as l / l ^ - ^ e j 3 with increasing distance between la t t ice s i tes . O'Reil ly and Tsang (1962) have done summations l ike the above for the special case of Calcium Fluoride crysta ls , bearing out this statement. Note, however, the error . is ni l when T = 0, as in this l imit and also when T = °° , as then - O We now look at third centra! moments. With considerable work i t can be shown that, to (M0)^ , M3(E0R) = M3(C0RRECT) a T T ' ° C _ _ ^ - , ) (6-20) 3 5" §8 - A 133 f-trZl%ii l3n}} where the notation A B C D ~ = ZZ. iQuL^itCteb;?" has been used. M3(T-T) is more d i f f i c u l t to compute, even to (MOjj, but i t can be shown to be in error generally by showing i t to be so for the Ising model. For this case M3(CORRECT) = " ^ - J - * 4/^o ) [Kz/~Hfo~) (6-21) and M3(T-T) = M3(C0RRECT) - £^jf ^) (6-22) 61 Again i t is seen that the error is not of 0(1/N), and that i t vanishes only in the l imits T=0 and T = The following table is drawn up in order to highlight the differences in moments calculated by the EOR and T-T methods. The ticks represent areas of agreement with the correct values, the. crosses, disagreement. M l " 2 M (MO^ ( M 0 ) 2 (M0)3 (M0)j (M0)2 ( M 0 ) 2 T-T X X X EOR 1/ X By our c r i t e r ion , i t follows that EOR second-stage decoupling should y ie ld a more fai thful l ine shape function than the method of Tomita and Tanaka. 2. Solution of Equations. When one comes to consider the solution.of the EOR and T-T-decoupled equations (6-8) and (6-10), the t ractabi l i ty requirement assumes extreme importance. The T-T equations turn out to be completely admissible of an analytic solution in Fourier-transformed space, while the EOR equations/ which are almost ident ica l , have some added terms which increase the number of simultaneous equations which must be solved for. Cn0 to something of the order of N. Moreover, the T-T method yields a set of tractable equations for spin values higher than J^, and can easily be extended to third-stage decoupling. The EOR method, on the other hand, becomes prohibit ively complicated when higher spin values and higher-stage decoupling are used. 62 For the reason that T-T second-stage decoupling does not conform sufficiently to the complexity requirement at high temperatures(i.e., temperatures in the paramagnetic range), Tomita and Tanaka extended their calculations to include third-stage decoupling. Now, T-T third-stage decoupling leads to equations which,;according to our criterion, are even more faithful than EOR second-stage decoupling. This follows from Corollary I to the moment theorem, which tells us that the N^, as well as the Mp resulting from a third-stage decoupling, is exact. For the above reasons, no serious attempt will be made to solve the second-stage EOR-decoupled equations other than formally. A simple case will be considered, however, to show that these equations satisfy the complexity requirement, even at high temperatures. The f irst two Fourier-transformed second-stage T-T-:decoupled equations, derived from the spatial equations (6-10) are (6-23a) and (6-23b) where (6-24a) (6-24b) (6-24c) (6-24d) 63 and the notation of Tomita and Tanaka J •4 ,i ! = l-k (6-24g) has been employed. <-J|^  JJ^ .^ I ^ 0\.<^ cj-h ^ and 0-^ ,$ are Fourier transforms of certain linear combinations of thermal averages. Note that G n = '+•1= /fl 0 " [o\-- u k = Equations (6-23) are easily solved for (6-25) where (t\ . The solution is (6-26) f 0 «f-(6-27a) (6-27b) Equation (6-26) does not satisfy the complexity requirement for high temperatures,.as 64 and reduces to a sum of two delta-functions. The f i rst two Fourier-transformed second^stage EOR-decoupled equations, derived from the spatial equations (6-8) are (6-29a) + f('M P(fc,t)(tl+ 0 / / f c < ^ , t M ( u < £ | ] (6-295) where «k,^  - (ft) r_ {c ( U ' - f , t-t - k ) y - v t (8.30., - /^(k-k ' - ; , k-k'), <6-30b) To solve equations (6-29) formally for ^ | , we treat the terms in the square bracket of equation (6-29b) as sufficiently small, compared to (i/VT) {{(k f) s o t n a t a n iteration procedure can be used; this is equivalent to the recognition that the EOR and T-T methods do not diverge widely from one another. Letting (6-31) the second EOR equation, (6-29b), may be written 65 + o M ^ M ^ . k , t<| ( , 3 2 ) or, with the use of the iteration procedure on f j^. 0 +• ' 3 ? where (6-33) (6-34) We note that S~ (k satisfies the "integral equation" ss cud- 3L : + mzL ^  (L> L' ]. (6-35) Similarly, ^ e ( |? ( ^ satisfies the "integral equation" Equations (6-29), solved simultaneously, yield the formal expression = £ - A a d c V - * T V - D C ^ M ^ M ^ " (6-37) 1 66 One would like to solve equations (6-35) and (6-36) exactly for ^ ^ a n d lO respectively. Unfortunately, this cannot readily be done, as the coefficients in these equations do not possess convenient symmetry properties. Let us, however, consider the simple case when Equations (6-35) and (6-36) then have the solutions (6-38) J f i N (6-39) (6-40) Substituting (6-39) and (6-40) into equation (6-37), we get the following expression for : jk ' — - , (6-41) 67 Equation (6-41) satisfies the complexity requirement even in the high-temperature limit for which (6-28) holds. We note that this requirement is also satisfied when equation (6-41) is specialized to the Ising model case, when (see equation The latter statement cannot be made about equation (6-26), the equivalent T-T solution. It will be demonstrated, in the next section, that also 3rd-stage T-T decoupling is not "complex" enough for the Ising model case. 3. Comp 1 exity Requirement We have seen above that at second-stage decoupling the EOR method satisfies the complexity requirement that the line shape consist of many delta-functions. In order to have this requirement satisfied with Tomita-Tanaka decoupling, these authors had to go to a third-stage decoupling. In this section, we shall discuss this point. Using third-stage decoupling, Tomita and Tanaka arrived at a closed set of equations which proved to be quite tractable, having as a solution (6-47)) independent of q (6-42) (6-43) with r 68 where ' ^ a * (^.^ and C. 3^,^') involve Fourier transforms of A^, C.., and of certain linear combinations of thermal averages, and (6-45) Equation (6-43) can, by use of the method of partial fractions, be cast into a form which leads to a sum of delta-functions for the line shape. These must be smeared in some way so that a continuous line shape function results. One can perform this smearing only i f the delta-functions are sufficiently large in number, and their arguments sufficiently dense along the <0-axis. It is this condition that has been referred to as the complexity requirement.# At first sight, Tomita and Tanaka's expression (6-43) for (^j seems to satisfy the complexity requirement. It especially seems to do so for the general paramagnetic case. However, i t demonstrably does not for the case of the general Ising model, at any temperature and for any spin. This will now be shown. # Grant and Strandberg (1964) satisfied this complexity requirement by treating the distribution of spins well-removed from the one of interest as continuous. 69 Recall that for the general Ising model (2-27) A ^ M = o (6-46a) i . e . , f\ 5 O all £ (6-46b) As a consequence, the quantities L0 (fi and L*(k t%t%') involved in equation (6-43) become and e ( j ) a l J 6 -r / u C W = A e independent of q (6-47) i X w O - C ^ f e r t 3 ^ . ^ - <6"48) Then J)1" and rY[?~ (equations (6-44a) and <(6-44b) ) become 3J = - / ^ l(6-49a) V - - ^ o - ^ t v M - r ) 7 6 . 4 9 b ) Now t^fc. » Z l ^ ^ , ^ . 3 t> V(^ (6-50) f ^ i. 3 by translational invariance. So, i f we define Q ^ Cg (IA) C ik) and as follows 70 8 (6-51) C - 3 M s ZlIU we may write (6-52) This yields, via (6-25 ) and (5-40 )., a line shape function which is expressed as a sum of three delta-functions. It is, of course, wrong to extrapolate from these to a continuous line shape function. Now, the general Ising model admits interaction of any spin with all its neighbours, no matter how far away. The most elementary considerations indicate that unless one is dealing with the one-dimensional Ising model with S = %, nearest-neighbour interactions, many more than three delta functions 71 are required to describe the line shape. One must conclude that Tomita and Tanaka's third-stage decoupling,cannot .be.used on the Ising model with any success. 4, Application to .Case of Completely Non'Teqwdlst-ant" Si.rjgle-Spin  Energy Levels-. This chapter so far has considered the case in which the single-spin energy levels are equally spaced. The most fruitful approach to the case of completely non-equidistant energy levels would appear to be through £ the use of the operators introduced in Chapter II. This, however, proves sufficiently complicated, even with T-T decoupling, that i t is unlikely to yield a line shape, and one has to be content with calculations of moments as in Chapter IV. To see this, let us discuss first-stage decoupling and indicate! what would,happen in second-stage decoupling. The statement that the unperturbed energy levels are completely non-equidistant implies that whenever one has V = A (6-53) This is a restriction on the elements of the set 67 defined in Chapter II. The Hamiltonian M can then be written as.(see equation (2-23)) 14= ^-Wf + t ^ [ f y ^ . ^ ' ) * ^ * / (6-54) where is defined by equation (2-16). 72 We will take the unperturbed frequency, W)a ,of the primary line under consideration to be given by (6-55) We then look for in order to find the line shape function (6-56) ft The first Green's function equation is (6-57) + (6-58) where we have used the relation (6-59) 73 which follows from combined with (2-32) and (2-22). T-T decoupling in the first stage consists, here, in performing the decomposition (6-3c) and neglecting higher-order cumulant Green's functions. This is equivalent to making replacements like the following: With this decoupling, equation (6-58) becomes ^* ('^mX ( = a i r + IZ. A i <AJA ^ , > ( / U ! \ - ZZ- A*j <C^ >2- A^^UmX J (6-60) 74 Taking Fourier transforms as in Section 2 and using (6-56) and (6-57), one gets a single delta-function for the line shape: g(iS) = F [ oo - u0 A ( 0 c(o) k' (6-61) where yti^" ^ ^Cl^^ independent of i > _ ^S^y independent of i j In deriving equation (6-61), one had simply to solve one equation for To see the difficulties that arise when second-stage decoupling is attempted, consider the simplest case, that of effective spin S = 1 (with A = -1, say, in (6-55)). One must keep all Green's functions of the form where there are two ) S and one s(L in each Green's function. There are nine of these, seven of which are independent. A separate Green's function equation must be written for each of these seven, a decomposition made, on the right hand side of each equation, of higher-order Green's functions into lower-order ones, according to equations (6-3), and Fourier transforms taken. One then has, optimistically, seven coupled simultaneous linear equations to 75 solve. The latter number increases quickly when higher spins are used. This process is quite unwieldy, and so the analysis has not been pursued any further. S Summary of Chapter VI. We have,,in this chapter, applied.a moment theorem to compare the faithfulness of two decoupling techniques. The results of the comparison are summarized in the table on page 61 . The EOR decoupling proves more faithful than T-T decoupling, but unfortunately, less tractable. It satisfies the complexity requirement at second-stage decoupling, whereas T-T decoupling satisfies this requirement only with a third-stage decoupling. For the Ising model, this complexity is not satisfied even with 3rd-stage T-T decoupling. The argument in favour of Tomita and Tanaka's approach to the general paramagnetic resonance line shape problem is that i t does, at least in the case of equally-spaced unperturbed energy levels, provide an analytic line shape function. This may be more important than its providing correct moments. 7 6 APPENDIX I To show that in (A-l) ^ Z L ^ ' f i m a y b e r e P 1 a c e d b y AJA'+I , Now, where we have expanded the trace in terms of eigenstates of ? ^ with eigenvalues . Let Then Using (2-8), (2-21), and the hermiticity o f 1 4 ^ , we have - £ M / A - ( T ^ K f - (77*i-T') ^ . A (A-5) 77 Therefore But, for A£ ^ , -~T^  = t J e (see equation (2-31)), So CXec,) "<u ( a . 7 ) i . e . , { A 6 6,) is zero unless \ £ Ch Similarly, where From (A-8), (A-7), (A-9), (A-3), and (A-2) (A-9) • xea, APPENDIX II Calculation of Thermal Averages for the One-Diimensional Ising Model using a Difference Equation Technique. As an illustration of the technique, the derivation of the thermal average £ 0 7 will be given. independent of i S ^) where we have taken i = N = ' A where J7)^ 71 j. are eigenstates and eigenvalues of 5 ^ » i«e. i ^ / ^ S 7 ? i / ^ > (A-12) and 71 ^ can assume the values ± j - 1 /l£T~ ; and the following notation is used: (A-13) («,>U - ^ K ^ l ^ ' 4 ' / O l ^ j o e?A , (A-i4) We will also use <*dc* s*hfy (A.15a) (A16a) (A-16c) (A-16d) Performing the summation over 01^ , (A-17) Looking at the first term on the R.H.S. of (A-17), we write i t in the following way: (A-18) Noting that (A-19a) 1L and <^^VTI,.XV^ <^^ t>+ku;,> (A_igb) i t follows that equation (A-18), after i t is summed over tf, ^ » A s ^ » 80 may be written in the form (A-20) where #p and bp satisfy the recursion relations "P* 1 P ' P (A-21a) L., ~ <f#p + ( A . 2 1 b ) with the initial conditions b-j = 0, a-j = 1. Equations (A-21) can easily be cast into the form of a pair of second-order difference equations: (A-22a) (A-22b) Equations (A-22) possess the solutions af - r t Wft u «f (,23a) L p , F ^ W G i f A - Y ( A. 2 4 b ) where F a, F^, G ,.G^  are independent of p and are determined from the initial conditions, and A± are given by equation (3-13). 81 In the calculation of the second term of equation (A-17), one has occasion to look at the equations £ M 5 (A_25a) 4+1 = ^cr* (A-25b) with initial conditions C| = 0, d^  = 1. They possess solutions T F; 6\+V +• Cc ( A . 2 6 a ) c/p » F*/ 4 ^ (A-26b) where F c, Gc, F d, Gd are again determined from the initial conditions. Whatever thermal averages are done, one can always arrange the terms in the appropriate summations in such a way that either (A-21) or (A-25) need be used for each term. A l i s t of the values of the F's and G's follows, where the notation ]f = ]/ U- + (A-27) is used. 82 6w r) From the above table, the following relations can constructed and are found to be useful: (A-28a) (A-28b) Continuing from equation (A-20), performing the final summations, and using equations (A-21), we have (A-29) 2. r^t+l . Similarly, the second term of (A-17), which we call fetii)^ , is given by (A-30) 83 and so - a -y ' (A-3D One similarly finds that the partition function Z is given by as given by Huang (1963). All other thermal averages can be found in the same way, when this is tractable. For example, the thermal average S^9 ^ becomes cumbersometo find by this method, as eleven summations must be done explicitly before the difference equation solution can be applied. 84 APPENDIX III Proof of (4-15) directly from (5-2a) and (5-3) Let tf> , i N ^ j (A-33) (A-34) (A-35) Then Multiplying both sides of equation (5-2a) by fit)> and summing over n from 0 to «Q , one derives the equation which has as solution B J (A-37) A boundary condition is needed to fix t e ?the constant of integration. We make use of the fact that when 7 4 w - 0 , the hierarchy of equations (5-2a) reduces to the single equation e. Gi9 ~ T 9 (A-38) and moreover, " " ' • • (A-39) Substituting (A-39) into (A-37) and using (A-35), C &AE)= ~T~ O-e/"'^') (A_4o) when 1 4 1 0 = 0 . To make (A-40) (A-38), one needs The assumption is then made that (A-41) holds also when 7/") ^ o Now (A-42) 86 where we have used the relation -r Vr) - ~i~(-r) { M 3 , which follows from (A-34) and the reality of the "77, . Equation (A-42) is equivalent to the stage reached in Chapter IV, equation (4-7), where a similar problem was considered. The proof is completed in essentially the same way, i.e., via the definition of a suitable characteristic function. 87 APPENDIX IV Real i ty of A^ of •equa.t.ion (5-4-7-) /3 It is assumed that and C'., are real, as they are invariably simple known combinations of thermal averages. From equation (5-51), 2L ( ^ are real for all c< Let A [ [fC] Then { Real part of A ok Imaginary part of A i.e. r Writing equation (A-45) in matrix form for 0,1,2 n-1, i ,e. (A-44) (A-45) having the values £ f f A (A-46) 88 i t is seen that the " £.• matrix", being a Vandermonde matrix and having all the £p distinct, is non-singular, leading to-.the conclusion that (A-47) all o( . This completes the proof. 89 BIBLIOGRAPHY Bogoliubov, N.N. and Tyablikov, S.V. .1959. Dokl.Akad.Nauk.(USSR), 126, 53; Translation Soviet Phys. D#klady, 4, 604. Call en, H.B. 1963. Phys.Rev.230, 890. Grant, W.J.C. and Strandberg, M.W.P. 1964. Phys.Rev.135, A715. Huang, K. 1963. "Statistical Mechanics" (John Wiley & Sons, Inc., New York). Ising, E. 1925. Z.Physik.3J_, 253. Kambe, K. and Usui, T. 1952. Progr.Theor.Phys.8, 302. Kubo, R. 1957. J.Phys.Soc. (Japan), 12^ , 570. Lee, Y.Y. 1961. Can.J.Phys.39, 1733. McMillan, M. 1959. M.Sc. Thesis, University of British Columbia, Vancouver, B.C. McMillan, M. and Opechowski, W. 1960. Can.J.Phys.38, 1168. O'Reilly, D.E. and Tung Tsang. 1962. Phys.Rev.128, 2639. O'Rourke, R.C. 1957. U.S. Naval Research Laboratory Report No.4975, p.207. Pryce, M^ H.L. and Stevens', K.W.H. 1950. Proc.Phys.Soc.A,63_, 36. Tomita, K. and Tanaka, M. 1963. Progr.Theor.Phys.29_, 528. Van Vleck, J.H. 1948. Phys.Rev.74, 1168. Zubarev, D.N. 1960. Usp. Fiz.Nauk. (USSR) ,71_, 71; translation: Soviet Phys.USPEKHI 3, 320. 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items