THE INTERNATIONAL SERIES or MONOGRAPHS ON PHYSICS GENERAL EDITORS N. F. MOTT E. C. BULLARD D. H. WILKINSONCONTENTS PART I: ELEMENTARY THEORIES I. ATOMIC FORCES , 1, Theoretical Considerations * . . . . 2. Tonic Radi . . . * 9. Heurlatio Rapreasions for Lateles Bacegios . * . Il, LATTICE VIBRATIONS 4. SS ee 5. Vibrations of « Distornle Chain . . 4. Frequency Spectrum off Laitioe Vibrations and Speciiic Heats . 7. Long Lattice Vibrations in the Optical Branches * 4. Infra-red Deepen ‘and the Retardation Effect on lattes 8. Atami Theory of Long Opti! Vibaions and Ina Di persion 10, Experimental pny of Indo-sed Dispersion ly lonie ‘Crystale IIL ELASTICITY AND STABILITY Ti. Homegeneous Deformation and the Elastic Constants ‘12. Mechanical Stability of Simple Lattion: . . . . 13. Relative Stability and Polymorphiem . . : . PART II: GENERAL THEORIES IV. QUANTUM MECHANICAL FOUNDATION M. —————— . . . 1S. Normal Coordinates =. . * “ Biatiatical Mechanica of Byeterce of Oscillators . nm, Statistical Mechanics of @ Molecular System under External Forces 18, Static Polarisability and Polarisability im Variable Fields 19. ‘The Rayleigh and Raman Scattering of Light . 7 20. Placaek's Approximation . 21. Expansion of the Opiat Parameters ond te Coiatn o Optical Effects Vv. THE METHOD OF LONG WAVES 22. The Geometry of Perfect Latuices . 23. The Infinite Lattice Model and General Jnvariance Relations MM. Lattin Waves . 4. Failure of the Method « Momgenmna Deformation - the Method of Long Waves . 24. Long Acoustic Vibrations e . =, ‘The Elastic Constante for Non-ionle Crystale . 2%. Equilitrium Conditions (vanishing aed ~ Panter In variance Relations. .Lg? LATTICE VIBRATIONS Ly (note that E,,, is irroiational): 8, = bw, = —od wy, nthe f= amet sa eli om till beeen ceil iain ices when free charges are present. For any particular problem the corre- sponding of motion for the charges can be readily written down with the help of the electric field (7.23). We notice that the solenoidal motions of the lattice are not affected by the presence of the free charges. On the one hand (7.24) is free of both w;, and p(x, f), and, on the other hand, as the electric field (7.23) is inde- pendent of w,, the latter does not enter the equations of motion of the charges. (7.23) and (7.25) constitute the basic equations for the classical treatment of the motion of charges in the lattice; (7.24) merely gives iso to free transverse vibrations which aro uncoupled to cither the longitudinal vibrations or the charges. To cite a simple example, one may, for instance, readily estimate the energy loss by a charged particle to the lattice from (7.25), if one con- siders the particle approximately as having a constant velocity. 5. Infra-red dispersion and the retardation effect on lattice vibrations$ (65. wicweenpes Sheng om pp ver" The electrostatic method used In the last section ls equivalent to describing the electric interaction between the ions by tho Coulomb force. Since the actual interaction does not instantaneously but propagates with the finite velocity of light, the method is only approxi- the mate. Inarigorous treatment, one should irrotational condition on E and introduce besides (7.9) the remaining Maxwell oquations Veer @ V.H = 0, (8.1) AB = —1H, (8.2) VAH = L(+ 4eP). (6.3) In using these equations of the clectromagnetic theory, we do more than take account of the retardation of the interaction forces. Consider t Cf. H. Friblich amd H. Felner, BRA. Report LIT 14 (0481. whew a ben ecvurate equation than (7.25) has been werd; see, however, ible. L/T 22 TAC aang, Neue 16,1 i Frahop.S A, 20, EE(NY ree a ee a A ee eet,= LATTICE VIBRATIONS m58 the case when the lattice motion is completely suppressed by holding the jons fixed in the configuration w = 0. (7.1), the equation of motion for the ions, is then to be discarded, and (7.2) reduces to —s. P=t,E = eoUe, (a) written alternatively aa D = E+4eP = «E, (8.5) the relation shows that the lattice merely functions as a normal refractive medium with the refractive index «.. Therefore, in this case, the lattice equations used in conjunction with the equations give rise to transverse light waves with the constant phase velocity ¢/ve.. In Fig. 18.0 these light waves and the lattice vibrations obtained in the last section are represented on the same (w,t) plot by dotted lines. These two groups of solutions both represent certain approximate solutions of the joint set of equations (7.1), (7.2) (lattice equations) and (7.9), (8.1), (8.2), (8.3) (electromagnetic equations). In one case, we effectively regard the ions as infinitely heavy and obtain then the radiative waves (the light waves) without lattice motion; in the other ‘case, wo ignore the retardation effect and obtain the lattice vibrations discussed in the last section, which contain no radiative energy. The two groups of diametrically opposite solutions together display the full ground covered by the six equations. In the rigorous treatment, the above approximate solutions are inter- mixed. As we shall see, the transverse lattice vibrations (without retardation) mix with the radiative waves (the light waves) and together they become the observed optical waves (infra-red dispersion). Of these, the waves immediately below the dispersion frequency ay are actually predominantly mechanical oscillations of the lattice, only a small fraction of thelr energy being radiative. If the small admixture of radiative energy is ignored, these wares reduce to the transverse lattice vibrations discussed in the last section. These vibrational modes thus play the dual role of being optical waves as well as lattice vibrations; for these vibra- tons, the electrostatic approximation is adequate. They are, as we shall find, essentially transverse vibrations with phase velocities emall com- pared with ¢/ve,. For the transverse lattice vibrations which were found in the last section as having phase velocities larger than ¢/v¥«., (or wave- length 2 10~* om.), the electrostatic method is invalid. ‘The admixture of radiative energy is in fact so large, that they can hardly be considered as lattice vibrations.mye LATTICE VIBRATIONS o mrs if. i Rees « = oHflogpe Veo. Lhe. Optical waves amd lattice vibrations, Jive wae a FFE ee fats be oat “4 *% == bine) Yrs. 18h. Percentage mechanical energy im tho transverse: modes. We shall also see that in the rigorous theory, we obtain the same Jongitudinal vibrations as in the last section. In other words, the retardation of the Coulomb force haa no effect on the longitudinal vibrations of the lattice,a2 LATTICE VIBRATIONS 1198 Let ua now proceed to consider the underlying mathematics. When solutions the obvious trial w= W,) r= Remit Eu are substituted, the system of equations (7.1), (7. 3). (7-9), (6.1). (8.3), (8.3) reduces to the following: —otw = 6 Why E, (8.6) P = b, w+5,,E, . (8.1) k.(E+40P) = 0, (8.8) k.H = 0, (8.9) kAE = oH, (8.10) kAH = —F(E+4eP). (a1) To reach the physical conclusions outlined above, it is important that we shoukl take care to obtain the complete solutions. In the first place, we notice that, in contrast to the clectrostatic treat- ment, the electric field can in no case vanish identically. For if E were to vanish, it follows from (8.10) that the magnetic field H vanishes. On account of (8.11), P also vanishes. Likewise w vanishes bocause of (8.7) and the fact that E = P = 0, Thus vanishing E leads only to the trivial case that E = H = P = w= 0. Let us now write (8.6) as we pe (8.12) ‘When this relation is used (8.7) reduces to P= {aorta +e (#13) With the help of (8.13) wé find that (8.8) can be written as (Bf + Ant + “Pate, =o. (an) ‘This equation admits two alternative possibilities, namely, olther Case A: 1+4eby+ 27eube, 0, (8.15) saa LATTICE VIBRATIONS » which also implies that (cf. (8.13) E+} isP = 0; (8.18) or Case B: k.E = 0, but as E does not vanish, we must have in this case Elk (8.17) The discussions for the above two alternative types of solution in relation to the remaining equations (8.0), (8.10), (8.11) have to be conducted ‘Conaider first the case A. Owing to (6.16) the equation (8.11) reduces kAH = 0, (8.18) ‘This requires that H either vanishes or is parallel to k; (8.9), on the other "iad It follows H=6. (8.19) —— (3.10) remains to be considered, which becomes in the present kKAE = 0. (8.20) Aa E does not vanish, E must be parallel to k. Apart from being parallel to k, E is otherwise arbitrary, Onco E is chosen, wand P are given by (8.12) and (8.13). The solutions under case A may thus be surmmarized as follows: the vectors are all longi- tudinal, namely, wIPUE Ik: and the frequencies are given by the solution of (8.15): ot mba ne = (Fo (<0, which, we note, is independent of k. We recognize that the solutions are identical with the longitudinal lattice vibrations (the irrotational solu- tions) obtained in the last section by the electrostatic method. The retardation has thus no effect on the longitudinal lattice vibrations. (Consider next the case B. Since in this case E is porpendicular to k, it follows from (8.10) that k, E, H form a right-handed system of orthogonal vectors in the given order, and their scalar magnitudes satisfy the following relation: bE =H. (8.21)“ LATTICE VIBRATIONS m8 (8.9) ie now automatically fulfilled. ‘The only equation still to be con- sidered ia (8.11), which reduces to the scalar equation BH = = (B-+40P), (8.22) After H and P are eliminated with (821) and (813), the equation’ “(= eer Sate Since Z does not vanish, we have a tate Patty = (8.23) Apart from being perpendicular to k, E is arbitrary, Thus to each k there are two independent solutions corresponding to the two possible independent choices for E (e.g. any two mutually perpendicular rectors, both perpendicular to k); the corresponding values of w and P are determined by (8.12) and (8.13). To summarize, all the solutions in case B are transverse, namely, k, (wIPJE), H, are mutually perpendicular. The frequencies, given by the solutions of (8.23), are in this case functions of k. For a given wave-number &, (8.23) gives two solutions, each being, aa noted above, doubly degenerato (Le. having two independent modes). In Fig. 18 all the solutions are represented on the (w,k) plot by solid curves. Remembering that the refractive index is equal to the ratio of the vacuum light velocity ¢ to the phase velocity of an optical wave, we find that (8.23) is identical with the dispersion formula for optical waves which we have already mentioned in the last section. Since the solutions which we have obtained are complete, it follows that all the transverse modes are optical wares. In general, when two oscillatory systems are coupled, the vibrational modes of the coupled system are mixtures of the original vibrations of the two systems. If the frequencies of the original systema are close to one another, their mutual perturbation is strong, and in the coupled modes both aystems take comparable parts. If the original systems have equal frequencies, their interaction is described as being in resonance. On the other hand, if the original frequencies are very different, the mutual perturbation is small and the coupled modes are essentially likeInge LATTICE VIBRATIONS ad the original oscillation of ono system, with a small admixture of the other, or vice versa. In Fig. 18a the transverse solutions (solid curves) represent uae Gate Gat a aac ae es solutions (dotted lines) of the same wave-number k. The interaction point marked as O corresponds to resonance between the two groups of approximate solutions. The resonance effect tails off towards the right, where the upper and lower branches of the solid curves approach the dotted lines. The underlying physical interpretation is clear: towards the right of the resonance point, the frequency of the upper branch becomes so high that the ions cannot take part in the vibrations to an appreciable extent owing to their large inertia; hence the approximate solutions corresponding to fixed ions provide a good approximation. Here the lower branch represents essentially’ lattice vibrations with phase velocities small compared with ¢/ve,; by comparison, the propa- gation velocity of the interaction foree ia so large as to make the retarda- {ion effect inappreciable. These modes are therefore closely approxi- mated by the lattice vibration calculated with unretarded forces. ‘Using the expression for the energy density, to be established presently, we can calculate the relative proportions of the radiative energy and mechanical lattice vibrational energy in the transverse modes. Fig. 186 gives the proportion of mechanical lattice vibrational energy for both branches of the transverse modes. We notice in particular that, towards: the right of the resonance point, the lower branch of the optical waves approaches practically puro mechanical lattice vibrations and tho upper branch, on the other hand, tends to purely radiative energy, both in agreement with the above interpretation. Thero are no modes resembling the pure mechanical transverse lattice vibrations for waves near or to the left of the resonance polnt (£ < (oyve/e) ~ 10 em.~") as given by the electrostatic mothod; for in this region (wave-lengths > 10-* em.) all modes are seen to be mixtures of radiative and mechanical energios in comparable proportions. ‘The essentially mechanical nature of the transverse modes immediately below the dispersion frequency o, has an important consequence on their behaviour as optical waves. The mechanical vibrations conserve energy only in the approximation where terms of third and higher order in the potential energy are ignored (lincarity of the lattice equation (7.1). In reality, each mechanical mode is coupled to the others through the higher-order terms and there is thus a small energy leakage from the mechanical vibrations. In the optical waves near «, the mechanical energy density is so large in relation to the radiative energy flux that the* LATTICE VIBRATIONS 48 small energy lox by the mechanical vibrations absorbs an abnormally lange fraction of the energy Bux, resulting in a strong damping of the waves, Tho experimental aspect of this phenomenon will be cliscussed in § 10. We now proceed to derive the energy density. We shall do so for the more general caso when free charges are present; the result oan then be used to construct the Hamiltonian for charges moving in a lattice, Let the charges be represented by ¢, (i — 1, 2,...,m) and their position vectors by x, In the presence of the charges, the electromagnetic equations (7.9), (8.1), (8.2), and (8.3) become V.(E+4eP) = 4 Febx—x)), (8.24) v.H=0, (8.25) vAE= —!H, (8.26) AH = 1(B+taP +42 Fat tx—x), (8.27) where 3(x—x,) is the three-dimensional Dirac delta-function. (8.24) is equivalent to (7.20) used in the last section; the delta-function gives the charge denaity duo to a unit charge concentrated at one point. Similarly, ¢,%,3(x—x,) in (8.27) is equivalent to the more conventional form pv. Using (8.26) and (8.27), wo can write e e qe AH) - qq (H-VAE—E.VAH) =a (OU A+E E+E.P+ F et, Bape—xy}. (0.28) Integrating both sides over an arbitrary volume, we obtain [leenm) == f (Lat+e.t)4b.P+ 5,8). B8e—x,] dr, (6.29) rs = where the volume integral on the left-hand side has been converted into a surface integral over the bounding surface by Green's theorem. We recognize that (¢/4e)(E AH) is the Poynting vector for electromagnetic energy flux; thus the left-hand side of (8.29) represents the rate at which electromagnetic encrgy flows out of the volume. (It is clear from the Jattion equations (7.1), (7.2) that different volume elements are coupledge LATTICE VIBRATIONS " in the macroscopic theory only through their electric interaction; thus the left-hand side of (8.20) also represents the rate of total energy out- flow from tho volume.) Tho last term on the right-hand sido of (8.20) can be written as — Fak Eley, (8.30) where the summation is extended over the charges contained in the ‘volume considered. Using the equations of motion for the charges: . m2, = eElx,), (a0) we find that ‘ = Fle tD — ty Ry ey, Ela). (8.32) The last term in (8.20) is thus equal to the rate of decrease of the kinetic energy of the charges within the volume. Hence it follows from the energy conservation requirement and (8.29) that E+E.) +E.P (8.93) must be equal to the rate of change of the energy density (including the kinetic energy of the lattice particles as well as all the electromagnetic and potential energies in the system). If we donote the onergy density by U, then dU /dt must be equal to the above expression. The following 1 U = jw'— pb, wb, E— jog E+E. P+ (EH) (a4) fulfils this condition. For differentiating (8.34) with respect to time, we find that = W.0—b,, wo —5,, W.E—b,,w. 2-5, 2.648. P+E. P+ +2 (B.2+H.M. (8.35) ‘The last three terms are equal to (8.33). The rest of the terms can be written as follows: W(W—b,, Wb, E)-+8(P—b,,W—by EB). ‘This expression vanishes because of (7.1), (7.2), and (7.3). Hence apart from an arbitrary constant (equivalent to an arbitrary choice of the zero for measuring the potential energy) (8.34) provides the requisite expres- sion for the energy density. When P is eliminated by the use of (7.2) and eat n"= LATTICE VIBRATIONS noe the b-cosfficients are expressed in terms of ay. «,, «.. the energy density assumes the very simple form: Um {tah +t (eg B+ HY. (8.38) For a transverse mode, the second term representa radiative energy and the first term the mechanical lattice vibration energy. The curves in Fig. 186 are calculated by substituting the transverse solution in the Pespective expreasiona ¢ ‘The Hamiltonian of the system is obtained by integrating / over space and adding the kinetic energies of the charges, namely, H= imate f {io+st0 +5 (oB+H™) dr. 837 The Hamiltonian includes aleo the energy of electromagnetic radiation. For problems where radiation effects are negligible, we can use the electrostatic approximation. H may then be put equal to ero, and the electric fold E eliminated with tho help of (7.23). If furthermore woe write w = w+, we find thet the Hamiltonian can be written as Hm Zim ttt | (i todw D+ et ott +2 — "sat wt—eaf 2{*— 2)]to, Boat gt Bite dr, (6.38) where we have made use of the fact that the integral of the scalar product between a solenoidal vector (e.g. w, or W,) and an irrotationa! veotor (og. w, W, or E,,.) vanishes identically. The last term lin (8.28) can be transformed as follows: E,,. is, we remember, the electric field due to the charges, if they were in vacuum, it is well known in electrostatics that je ia, apart from the electrostatic self-energies of the individual charges, equal to the Coulomb interaction between the charges: 1 “ay mom Therefore if we omit the electrostatic self-energies of the charges from the Hamiltonian, the last term in (8.38) can be replaced by the Coulomb interaction divided by «.. Furthermore, since ag is equal 10 «. ef /¢, (soe t For details are K. Huang, Pres. Rey. Ser. A, 208, 262 (1061), 4xii CONTENTS 9. Central Foress . 30. Coulom> Field ina Lattice Rina’ Method an los Pa «Dipl at or. au. Aceustic Vibrations in Tonle Lattlors (rigid ion medel) a 32. The Elastic and! Piesockeetrie Constants and the Dicketrie Tensor 33. Phenomenological Discussion of the Dispersion Formula fer Complex Lattices 2. Lang Ot! Virion Tenis Late gt om mode ‘34, Polarisable Jona . Vi. THE FREE EXEROY 34. The Specification of Finite Strains . 7. Phenemenclogieal Diarumsion of the Pree Raergy of « Lattior 34. Normal Coordinates of a Latticn . 39, Normalization of Physical Pommetem, Seition Hae and Expanson Methods. * . 4, The Free Energy . . « * ° 42. The Static (Non-vibeating) Lattice . . . . #2. The Taw . * . . . . . VIL THE OPTICAL EFFECTS 44. The Micromopie Theory of Dispermon 44. The Local Treatment of Optical Effrota . * 44. The Fifect of the Anharmonie Potential on Dispenion . * 47. The Dispersion Formula with Damping . * . * 4a, The Effect of the Second-onler Electric Moment . . 40, The First. and Second-order Raman Effecta =. . 50. The Tiiliculn Compononts of the Thermal Seattoring of Light . APPENDIXES 1. Some Common Lattice Structures (p. 1) . ‘Tl. Madelung’s Energy (p. 3) . IIL. Evaluation of Simple Lattice Sama (p. 23) « IV. The te the Viton Syn with the hap of ¥. Energy Density in Ionic Crystals (p. #3) - #8 a3 TT aE: z ee sn VIL The Adiabatic Approximation (p. 170) VIII, Elimination of the Electronic Motion (p.1T%) — . . ws EX, Desilo Petrtinn ond Optiet Rotation tp 8 . 7 ‘X. Recent Publications . . . . * INDEX z - = = - =Inge LATTICE VIBRATIONS ” (7.17)) the two terms in (8.38) with wf can be combined, giving simply jew}. The Hamiltonian (in the electrostatic approximation) can thus be written as 1 d= Finals. D eit + [ QOME+ et wN+ LOWE wh) de — -a[d(t- 1} J W,-Eyaedr. (8.39) ‘We have seen earlior that when the lattice particles aro held fixed at wm 0, the lattice behaves as a normal dielectric with the dielectric constant «,; the first two terms in (8.39) reprosent the Hamiltonian of the charges insuch a medium. The third term representathe Hamiltonian for the lattice in the absence of free charges. The fourth term gives the interaction Letween the system of charges and the lattice motions. Once more wo notice that only the irrotational lattice motion is coupled to the charges present; this is true, however, only in the clectrostatic approxi- mation. For quantum-mechanical applications it is necessary to describe the Jattice motion in terms of normal coordinates. Since the transverse vibrations do not interact with the charges, and are thus of no interest in the present connexion, we shall ignore w, in the following considera- tions, Let us considera finite volume V containing Lx Lx L = N lattice cells by imposing the corresponding periodic boundary condition. Then the normal coordinates can be introduced as follows (for general methods see §. 38). Express w, as the Fourier series wim Jp > oivneuree, (8.49) were owing 10 he pra dion, y is restricted to the following ¥= Tb thbt+A.b?, (8.41) Ay hy, belng integers in the range O< AM < Lt (= 1,2,3. (8.42) Tho polarization veetor, y/ly|. in (840) follows dircetly from the irrotational character of w,. For w, to be real, we must have Oy) = —O*(—y). (8.43) Substituting (8.40) in the Hamiltonian (8.39) and carrying out theod LATTICE VIBRATIONS mg0 ‘volume integral over the volume V, we get (ignoring w,): 1 1 H=-¥ Wm § Digs +3 2, MO Cora) — ~afafe-2) Zan f(g) nwenon ‘The real normal coordinates q,(y), gy) are introduced by writling Ot) = J latr)+ indy!) = —2%(—y) (8.45) {im the case y == 0, Q(0) is the normal coordinate iteelf; for the sake of simplicity in the following we shall not represent explicitly terme in Q(0)), where y can now be restricted to the values lying on one side of an arbitrary plane through the origin of the y-space so that only one of a pair y, —y is counted. In terms of the normal coordinates, the Hamil- tonian then becomes Hm Eomtts3 > peg th > >, Morten + +ALL DMZ f(@) names oe + Sean f [[F)- Basin] (8.40) This is the form in which the Hamiltonian can be used directly for quantum-mechanical considerations. 9. Atomic theory of long optical vibrations and infra-red disper+ sion Having discussed the long optical vibrations and the infra-red disper- sion on a phenomenological basis with (7.1) and (7.2), we ran now establish the atomic theory of these phenomena simply by ceducing Uhese equations from the general atomic theory. For the present purposes, we have to take into attount a feature hitherto ignored in (1 and 3, namely, that the ions are polarixable by electric felis, As explained earlier in § 3, this effect can be ignored when the ions oceupy positions of symmetry, such as in many simple crystals in the equilibrium state (or subject only to clastic «trains). This is no longer permissible if cither a macroscopic electric fickd is plated in the phenomenological equations. ‘The polarization of the fonsigo LATTICE VIBRATIONS ron can be described as follows, When an ion is subject to an electric field E its eleetron cloud is deformed and acquires an electric dipole moment p. If the field is not excessively high the induced dipole moment is propor- tional to the field, ane, (o) where a is the atomic polarizability of tho jon. « is « constant as long as the fick docs not vary appreciably within « period of the clectronie motion in the jon, To derive the phenomenological equations, we have to find, for given wand E, the forces acting on the jons and the dipole moment per unit volume, ‘The dipole moment is due partly to the displacements of the jonie charges and partly to the induced electric moments on the bons, When an ionic chargo Ze is displaced by wu, the net effect is as though a charge —Z+ has been placed at the undisplaced position of the ion to annihilate tho original charge and a fresh charge Ze has been created at the displaced position. The displacement is thus equivalent to tho addition of a dipole with the moment Zeu. ‘The induced electric moment onan jon, on the othor hand, depends on tho field acting upon the ion. In practice, this ficld is never altogether uniform as implied in the formula (9.1); in the following considerations wo shall always take the value of the fied at the centre of the jon. This field, which acts upon an jon, will be designated as the effective field. It is important to realize that the effective field is different from the electric field. ‘The latter is not simply the total field, which imeludes the field of the ion under consideration; it is the total field averaged over the space occupied by a lattice cell. The effective field on ‘an jon, on tho other hand, is the total field with the contribution of the jon itself excluded and the value, as wo havo said, is taken at the centre of the jon. Our first task is to calculate the value of the effective field. Tho discrepancy between the effective fickd on an jon and the macro- scopic field in the same neighbourhood is clearly due only to the con- tributions to these fields of the matter in the close neighbourhood of the jon. For, in the first place, the charge sources farther away are naturally irrelevant for the exclusion of the field due to the ion itself. In the second place, the fields due to such distant sources do not vary appre- ciably in a neighbourhood of the dimensions of a lattice cell, as long as the sources are located at distances away which are large compared with the Initice constant a; thus the average value and the value at a specific point need not be distinguished. Hence let ua draw « sphere around the ion under discussion with a radius & large compared with aez LATTICE VIBRATIONS 11,99 and proceed to calculate the difference between the effective and the macroscopic ficlds due to the matter contained within the For convenience, we take an intermediate value for R such that it is large compared with a, whilo the macroscopic quantities w, E, P do not ‘vary appreciably over the region of the aphere, ‘This is possible, because the lattice constant a is to be regarded as an infinitesimal in a macro- scopic treaiment. Macroscopically speaking, the matter within the sphere is a uniformly polarized sphere. It is woll known in electrostatics that the uniform polarization P is equivalent to a surface charge distribu- tion with a density everywhere equal to the component of P along the outward normal to the surface. From the surfaco charge the macroscopic field at tho centre of the sphere can be easily ealeulated. It is obvious from the symmetry that the resultant field must be parallel to P; thus in considering the contributions due to different surface elements we need only consider tho component in this direction. If we introduce the polar angles ¢, 4 with the polar axis in the direction of P, the contribution due to a surface element A" sin @déd¢ is clearly (fs ape cont = —P cos? sin #dédg. (0.2) Integration over the whole spherical surface gives immediately the macroscopic field due to the matter contained in the sphere as de -5?: (9.3) Next wo have to consider microscopically the effective field on the central jon due to all the other ons in the sphere. We are concemed, of course, only with the field arising from the displacements of the ions and their induced moments (the original field duc to the undisplaced and unpolarized ions vanishes for the highly symmotrie structures under consideration); we may thus imagine each ion site being occupied by a dipole equal to the sum of the displacement dipole Zew and the induced dipole p. Owing to the assumed uniformity over the spherical region, all the dipoles on either the positive or negative on sites are identical among themselves, Unlike the macroscopic fiekl, which depends only on tho macroscopic polarization, the effective field will depend on the detailed arrangement of the dipoles. Here we are concerned only with lattions in which every jon is in « surrounding with tetrahedral symmetry, the meaning of which we shall presontly explain. Por such Jatticos, as we shall find, the effective field due to the dipoles in the sphere vanishes identically.tye LATTICE VIBRATIONS 103 ‘The tetrahedral symmetry “about an jon implios that there oxist Cartesian axes XYZ through the jon such that if there is an jon site with the coordinates (¢,6,¢), there must be identical ion sites (i.o. sites occupied by the samie kind of ion) at points (—a,—b,c), (a, —b,—<«), (—a,b,—c) and the eight more points with coordinates obtained from the above by cyclic permutations of a, b, ¢. (These points are generated from (a,b,c) by tho tetrahedral group of operations about the origin which are rotations by + about the axes X, Y, Z and rotations cither way of 22/3 about the cube diagonals +X = +) =— +42. Tetrahedral symmetry of structuro means that the geometrical arrangement is unaltered, if the structure is rigidly subjected to the tetra- hedral operations. The simultancous existence of the group of twelve points listed above clearly follows directly from this definition.) A dipole Pp at x,(a, b,c) gives rise to a field at the origin equal to mete 4 Let x,,i = 2, 3,..., 12 denote the other eleven points which are occupied by identical dipoles; the contribution to the effective field due to the group of twelve identical dipoles can thus be written as ~ Zidet* 2 hate ‘Wo notice that all twelve points are at the same distance (a"+5*+c%)' from the origin. Thus when the explicit coordinates of the points are substituted, one obtains: 3 — Bare rapt wre a(ap,-+bp,+¢p,)—a(—ap,—bp,+ep,)+-alap,—bp,—ep,)— —a(—ap,+bp,—ep,)+cyclic pormutations of a, b, ¢ x | MaPit Opa mahi =a. Seren) t the Laplacian operator in the coordinates xii); snes 4 (x(4)|) in only « function of the radial magnitude [x(i)| we find imme- on writing the Laplacian operators in polar coordinates that the above relation reduces simply to & (tx + igen} = ak As all the negative ions are equidistant from the origin, [x(i)| = r,. 1, boing the nearest neighbour distance, all terms in the summation are equal, and wo have M ‘Si a Fens 2900}, (9.28) In §3 wo have seen that if only overlap potentials between nearest ealeulating & from the potentials thus determined to compare with (9.21), an equivalent and more direct procedure can be followed, namely, by expressing & directly in terms of the compressibility with the help of (9,25). To show that this is possible let us rewrite some of the equations ‘used in § J with the overlap potential expressed as ¢.as we have done here, Thus the energy per cell (3.10) is rewritten as u(r) a —4 4m, (0.26) where the first torm, we remember, is the Madelung onergy. ‘The equili- brium condition becomes now on (F), 7 fr 4eve. (27) Moreover, remembering that de/y = Jdr/r, we have for the compressi- FAB SGI, A) tar,me LATTICE VIBRATIONS mn where we have mado use of the equilibrium condition (dujdr), = 0; this relation clearly corresponds to (3.14) in § 3. On eliminating the Madelung term in (9,28) with the help of (9.27) we find that u 1 MPH goty 4 28) j~ Haroon Comparing this relation with (0.25), we obtain the following simple relation between the constant £ and compressibility 6: Our intention is to test the validity of (0.21). In view of (0.29), we can do so by verifying the relation(-G. a pagetia tom = bees 1 MAA +2 1 a = which is obtained by eliminating & between (9.21) and (9.29). The relation contains only measurable quantities and relates essentially the com- pressibility and the dispersion frequency.f To indicate to what extent it is fulfilled in actual cases, the ratios of the value of # calculated from (9.30) to the observed value are given in Table 20 for a number of crystals. near the bottom of the table. It is, however, difficult to ascertain the predominant cause of error in individual casos. { Mistorwaily the first connexion between the infre-red freqecmcy wo amd the i a : i i : i : F : 3wa LATTICE VIBRATIONS mye To discuss tho microscopic expression (0.18) for b,,, let us form the ratio of the directly observable value as given by (7.7) to (9.18): sane ise) micro f e n polarinabilties with (9.20) and write the ee om of). (9.31) Tanie 20+ Compressibility calculated from the Dispersion Frequency end Distortion Factor 4 POSE PRERREEE=FSEERES Stitiitiiiisitit ttt The values of for a number of crystals calculated on the basis of (9.31) aro given in Table 20, Their deviations from tho value unity are quite pronounced for the lighter alkali halides, which should approximate most closely the idea) ionic model. One cause for the deviations has been particularly emphasized by Szigeti,t namely, the mutual distortions of neighbouring tons owing to their overlap. This explanation of the discrepancy is favoured by the fact that it will not affect (0.19) or (0.30), which, as we have seen, are reasonably well fulfilled; this we shall see in due course. Moreover, it isa priori clear that some such distortion must 1 Votoms are taken from Bi. Selgnti, Pron, Me. Sve. A. 304, 61 (1960;. nee aP% 7, 10 4 £ Mia oe, reg a af Ved, er iieia ts apes ~ge LATTICE VIBRATIONS 13 accompany the overlap between ions, which is responsible for the overlap forces. Thus we shall investigate how such distortions would modify the microscopic expressions for the b-coefiicients and in what manner they aro related to the deviation of s from unity. Since distortions are merely rearrangements of the charge distribution, the most important effect can be deseribed by the dipole moment associated with the rearrangement. We shall consider only the distortions occurring between nearest neighbours. Let ua denote by m(r) the distor- tion dipole moment between a positive and a negative ion; mir) is a function of their separation r and the sign is chosen such that m is counted positive if the moment is directed from the negative towards the positive ion. Exactly as we have done for the overlap forces (p. 109), we may again consider a positive ion at the origin surrounded by its neighbours at x(i). When the positive ion is in an arbitrary displaced position x, the total distortion dipole moment surrounding it can clearly mens J mcmw-xu[- (FS), (9.32) where the vectors —({(x(i)—x)/(|x{i)—x|)} are unit vectors indicating the directions of the various individual dipoles, which, according to our sign convention, are directed inwards towards the positive ion at x. ‘We can immediately reduce our problem to the same considerations as in the caso of the overlap forces, if we introduce the integral function # of ~ar), momly, ¥(r) = — mir. (9.33) Tt ia easily verified that the a-component of the distortion moment (9.32) can be written in terms of the function ¢ as _ Se Hm0-x0. (9.34) This is identical with the expression (9.23) for the total overlap force on the ion, apart from the replacement of ¢ by ¥. ‘Therefore we can write the distortion moment as 9, (9.35) in analogy with (0.22), where g is given by = F{rera+ Hee| - —Flneas B, (9.36) which is a direct transcription of (9.25) with ¢ now replaced by ¢. ¢ is thus determined jointly by the distortion moment m(r,) already present in the equilibrium state between neighbours and its rate of change with the lon separation. =_—aT '