Physica
1lOA (1982) 373407
North-Holland
ON THE DYNAMICS
Publishing
OF BAND
WITH
Uniuersitiit
JAHN-TELLER
SYSTEMS
A15STRUCTURE
Robert Fakultiit fiir Physik,
Co.
Konstanz,
Received Revised
KRAGLER Postfach
5560, D7750
Konstanz,
Germany
10 July 1980 17 June 1981
We construct a thermodynamic model for a system of electronic d-bands coupled to the elastic lattice. For the electronic fluctuations a Debye-type of relaxation, for the lattice displacement a modified elastic equation of motion is assumed. As a result a coupling of certain acoustic and relaxational modes is found leading to a soft mode instability. For non-vanishing external fields displacement and dielectric response functions are derived. The special case for wavevector qjl[l IO] is worked out explicitly. A comparison between the present phenomenological model and a microscopic multiple-band electron-phonon transport theory, recently given by the author, reveals remarkable agreement between both approaches.
1. Introduction
High-temperature superconductors of AlS-structure such as Nb$n and V3Si exhibit a number of anomalous properties’) among which the structural phase transition and the associated softening of the elastic shear mode are of particular interest. The features which have been one of the main objectives of the theories for the martensitic transition of intermetallic AU-compounds, are: (i) The elastic softening on cooling, most pronounced for the modulus i(c,, - c12) which almost vanishes at the transition, whereas the shear modulus cU is only weakly temperature dependent and the bulk modulus f(c,, + 2c12) stays essentially constant”). (ii) The [llO] transverse acoustic (T,A) phonon with [liO] polarization which corresponds to the shear modulus $(c,,- c,?), exhibits a marked softening extending halfway to the Brillouin zone boundary4). (iii) A central peak for the same branch has been observed&) which diverges as the martensitic transition is approached from above. These and other anomalous properties of A15compounds, especially the exceedingly large normal-state specific heat, indicate an anomalous large density of states at the Fermi leve15). Most of the anomalies seem to arise from the presence of three orthogonal sets of linear chains of transition-metal atoms in the A15 structure which give rise to a planar Fermi surface6) and to a narrow peak in the electronic density of states. 0378-437 1/82/0ooooooO DO2.75 0 1982 North-Holland
R. KRAGLER
374
By exploiting to date explain According been
some
the quasi-one-dimensional ihe martensitic
to an earlier attempts”.‘?
which a sublattice ruled showed
band
transition
suggestion ascribing
structure
as being by Anderson
the transition
distortion
is associated.
However,
In addition,
Sham’s
that
only
the FIZ-optic
mode
is weakly
all theories driven’-“).
and Blount”) to a soft optic
out by experiment4b).
assuming a linear coupling between this optic mode the latter one is driven soft first. Recently, the importance of anharmonicities
nearly
electronically
there
have
phonon
with
this mechanism lattice-dynamical
has been theory”“)
temperature-dependent, mode
and,
and the acoustic
for the mode
shear
softening
has
been studied by Achar and Barsch’“). According to their model the shear mode softening arises from a subtle balance between two opposing effects, the softening caused by the electronic instability and the stiffening due to anharmonicities. In spite of this finding, anharmonicities enter, if at all, only through higher order electronic contributions in the known theories of the martensitic transition ‘5.‘6a,2’). The electronic models to date attribute the mode softening either (i) to a direct coupling of electronic degrees of freedom to the lattice dilatation (d-band Jahn-Teller effect at high-symmetry points of the Brillouin zone) or (ii) to a coupling via an optic mode which pairs the transition-metal atoms in the chains (Peierls-like charge-density-wave (CDW) driven instability). The first model based on the band Jahn-Teller effect is due to Labbe and Friedel (LF)‘). There, a ID band structure is assumed for each of the linear chains. The Fermi level is located close to the bottom of one of the empty d-bands (at the F-point in reciprocal space). The 1D density of states exhibits a l/u/E-singularity. The transition is regarded as a band Jahn-Teller effect where the triple degeneracy of the ID bands is removed by a tetragonal distortion which shifts the bands relative to each other. Subsequently, it was demonstrated by Cody, Cohen and Halloran (CCH)‘) constructing an idealized Jahn-Teller model with a step-function density of states
N(E)
- O(E),
as in the 2D case, and with the same degeneracy
splitting
like LF, that the essential point of the LF-model is a rapid variation of the density of states near the Fermi level. This feature, common even to the more refined versions’*“.’ 4 of the LF or CCH model which have been treated since then, is most unlikely for a realistic 3D band structure. In fact, band-structure calculations by Mattheiss22) have revealed that due to interchain coupling the ID character of the bands is wiped out. Moreover, since the assumption of the LF and CCH-model locating the Fermi level in the vicinity of the band edge, is highly artificial, Gorkov”) has considered the martensitic transition to be the result of a Peierls-like CDW-driven transition in the ID chains. The Fermi level is located close to the X-point where there is a band degeneracy
BAND
JAHN-TELLER
SYSTEMS
WITH
AIS-STRUCTURE
375
in the A15structure of the cubic lattice. Again, this X-point degeneracy is lifted by pairing the transition-metal atoms in the chains. There results a lowering of energy as the Peierls gap opens up. This energy gap is proportional to the amplitude of the pairing mode which, in turn, is coupled to the elastic strain, thus leading to a tetragonal distortion of the lattice. Based on the X-point Gorkov-model Bhatt and McMillan’4”) have formulated a dynamical Landau theory where the shear mode instability is caused by a CDW instability through the coupling of the CDW’s to the rlz-optic modes. The highly successful Bhatt-McMillan theory was even further improved to a two-band tight-binding model’“) based on a 3D-band structure. This model exhibits both a X-point Peierls gap (like the Gorkov-model) and a Jahn-Teller degeneracy splitting (like the LF-model) at the M-points and at the rX-saddle points, which are jointly responsible for the electronic instability. Hence, this model is a kind of hybrid model since it incorporates the essential features of the LF and Gorkov-model. Recently, it has been pointed out by Bhatt’&) that the essential results of the Gorkov-Peierls model by Bhatt and McMillan are independent of a specific microscopic mechanism and would likewise follow from a Jahn-Teller type instability too. Most of the above-mentioned theories of the martensitic transition are static ones, only a few onesg~“c,‘3*14a.16) are concerned with the dynamical aspect of the transition. The first proper dynamical band Jahn-Teller model was developed by Pyttep considering the interaction of the elastic strain with electronic density fluctuations. Separating static and dynamic strain contributions in the equations of motions provides a relation between the shift of the band edges and the static distortion. Renormalized acoustic modes are obtained by considering fluctuations about the static quantities. This procedure is, however, based on the assumption that acoustic modes do not induce any electron redistribution which restricts the theory to the collisionless limit. Strictly speaking, Pytte’s theory does, therefore, not apply to the hydrodynamic limit where the renormalization of the acoustic modes is of interest too. In a more recent treatment, in the spirit of Pytte, given by Thomas and the present author”), the elastic shear mode is coupled directly to a CDW of the same symmetry. In this respect the model by Bhatt and McMillan’4”) is more general in that it includes also r12-optic modes. Here, the shear modes are coupled to the CDW’s only by the intermediary of the optic modes. In both theories the shear mode softening originates from the temperature dependence of the coefficient of that term which incorporates the electronic variable to second order. While in ref. 13 the temperature dependence of this quantity which turns out to be the inverse of an effective density of states, is determined empirically, the Bhatt-McMillan model assumes instead the usual
R. KRAGLER
316
linear both
temperature theories
charge
dependence
make
density
equations dissipative
In the
as generalized
terms
associated
function.
characteristic transition. relaxation dominant
fluctuations.
arise
dissipation
of Landau
theory.
use of a Debye-type
This
Bhatt-McMillan
Lagrangian with
mechanism theory
equations
are expressed
involves
a relaxation
mechanism
the
of motion
the CDW’s
function
for the dissipative
As to the dynamical
of relaxation
and diverges
aspect, for
where
by means time
the
dynamical the of a
which
is
at the martensitic
Quite similarly, the model of Thomas and the author introduces a time l/y which characterizes interband relaxation processes in the hydrodynamic limit. As to the equations of motion one
assumes for the electronic equilibrium configuration
fluctuations a relaxation depending on the lattice
towards an instantaneous distortion. The distortion.
on the other hand, satisfies the usual equation of motion with an induced stress depending, in turn, on the electronic variable. In summary, both theories are able to reproduce qualitative features of the observed dispersion of the TrA phonon&) associated with the shear modulus $(c,, - cl?). As the central peak is concerned, however, both theories amount to the same conclusion
that this feature
cannot
be ascribed
to electron
dynamics:
at least
a slower, non-electronic relaxation mechanism seems to be required. It is the purpose of the present paper to extend the simple treatment of ref. 13 in a way such that the full symmetry of the AIS-structure is taken into account. Hence, in the high-temperature phase to which our phenomenological model is restricted, the cubic lattice is coupled to the system of three degenerate bands arising from the three orthogonal chains of transition-metal atoms. potential The paper is organized as follows: In section 2 the thermodynamic is constructed
in terms
of the
elastic
strains
e and
the electronic
d-band
occupation n, making use of symmetry. Coupled equations of motion for the relaxing electronic band system and the elastic lattice are derived in section 3. Section 4 gives the bare response and the secular equation for the coupling between
acoustic
and relaxational
modes.
The generalized
response
obtained
for non-zero external fields is the subject of section 5. It turns out that the exact response matrices are expressible by means of Dyson-like equations in terms of the bare quantities, obtained for the uncoupled system. Explicit calculations of the response matrices are performed in sections 6 and 7, where special attention is paid to the case of wavevector ql/[llO]. For this particular direction the secular equation for the coupled modes is investigated in some detail in section 8. Finally, in the Conclusion the phenomenological model is compared with a microscopic multiple-band transport theory recently given by the author16) revealing remarkable agreement between both theories.
BAND
2. Thermodynamic
JAHN-TELLER
SYSTEMS
WITH
AlS-STRUCTURE
377
potentials
In order to study the dynamics of a band Jahn-Teller system which describes at least some physical aspects of A15compounds in the hightemperature phase, we construct a model which involves the electronic band occupation numbers ni (i = x, y, z) and the elastic strain components iii (i, j = x, y, z) as relevant thermodynamic variables. Henceforth, the thermodynamic state of the system is defined by the set of variables {ni, eij}. The electronic band system is characterized by the occupation numbers of those d-bands only which partake in the dynamics of the present system. In particular, the quantities ni are defined as deviations from the equilibrium occupation of these relevant d-bands. Disregarding s-d scattering, the total number of d-electrons is conserved so that Ei ni = 0. On reasons of a compact notation the three components ni are condensed into n = {ni}. The state of the elastic lattice is characterized by homogeneous distortions which are specified by the six independent components of the strain tensor e. For the formulation of a dynamical theory, however, the acoustic mode amplitudes come into play. Therefore, it will turn out to be more convenient working with the elastic displacement u rather than with the elastic strain e. The latter one can be written as symmetrized gradient of u, i.e. e = Vsu = i[ Vu + ( VU)~].
(2. la)
The appending subscript S on V means that only the symmetric part of the displacement gradient Vu is to be taken, the superscript T denotes transposition. Similarly, the diagonal components of the strain tensor are designated by e D = VD,, = a(v)
- u.
(2. lb)
Here, the subscript D on V denotes the operation of projecting out only the diagonal components of e. The matrix 9(V) is a differential operator with diagonal elements defined by 9(V)ii = di whereas the off-diagonal elements 9(V), (i# j) vanish. This symbolic notation will considerably facilitate the manipulation of the equations of motion and will therefore allow us, to expose the principal steps of the derivation of eqs. (3.5) and (3.11) more clearly. Since the thermodynamic potential which will be constructed in terms of the variables n and e, has to fulfill certain symmetry requirements it is therefore advantageous to start already with symmetry-adapted variables {v,, E,} defined as linear combinations of the corresponding Cartesian quantities. As regards to the underlying symmetry considerations the details are exposed in 15)and will therefore not be given here.
378
R. KRAGLER
e,,Mi
EI =
(exx+ eys+
~2 =
(en
63 =
(2ezz - e, - e,,)/VG,
-
(2.2a)
e,,)NZ
and, similarly, vI
=
v2 =
(n, +
fly
(n, -
n,)/V?,
+
n,)/d,
(2.2b)
v3 = (2n, - n, - n,)/d6. For reasons
of generality
we admit
vi f 0, i.e. allow temporarily
violation
of
the conservation of d-electrons. The connection between symmetry-adapted and Cartesian bases is given by the transformation matrix T(,,j (cu = 1,2.3; j = X, Y, z),
(2.3)
so that
v=T-n, The with
remaining the
advantage potentials
Voigt
E =
r -e'.
off-diagonal
(2.2c) strain
components
eii are chosen
notation,
E, = 2eii (a = 9 - i ~ j) where of the symmetry-adapted coordinates is that can be put into block diagonal form.
in agreement
i # j. The obvious the thermodynamic
We now consider the four contributions to the thermodynamic potential: Firstly, there is an elastic strain energy contribution which is expanded in terms of the elastic strain tensor e. Linear terms vanish due to cubic symmetry. Anharmonic contributions are neglected; for the importance of those higher-order terms on the mode softening see for example Achar and Barschm). In explaining the salient features of the lattice dynamics of A15 compounds in the cubic phase the harmonic approximation will be sufficient. Hence, in terms of the symmetrized strains E = (E,, . . , Ed) we have IFL(e) = 2~ . % . G = i(c,, + 2c,zg
+ j(c,1
- C,z)(&
+ E:)+
&QE:
+ E: + E$.
(2.4)
V is a (6 x 6)-matrix diagonal in the extended basis E, which contains linear combinations of the bare elastic moduli ceP according to symmetry. The elastic matrix c,~ in Voigt notation is connected with the usual fourth rank
BAND
JAHN-TELLER
SYSTEMS
WITH
AlS-STRUCTURE
319
tensor of the elastic moduli due to the mapping c,~ = Cijkl where (Yis related to (ij) by CY= i = j or (Y= 9- i -j (i+ j) and /3 to (kl), respectively. In the case of cubic symmetry with only three independent elastic moduli and cM, the matrix V has three eigenvalues: There is a singlet of Cl13 cl2 Al-symmetry associated with the bulk modulus (c,, +2ci2) where the corresponding eigenvector describes a dilatation l l, a doublet E associated with the shear modulus (cl1 - clJ belonging to uniaxial distortions l2 and l3, and a triplet TT with eigenvalue cM involving the shear strains l4, l5 and E6. Due to the coupling of the elastic lattice to electronic degrees of freedom the bare elastic moduli c,~ will be renormalized and become temperature dependent. The crystal is stable against homogeneous deformations if all eigenvalues of C ,+ are positive. If one of these eigenvalues, however, decreases to zero the crystal may then distort continuously to a new structure, the symmetry of which is determined by the eigenvector of the “soft” eigenvalue. Secondly, there is an electronic contribution of the d-band system FB(v) = ;Y . ii-’
.u
= g-‘(v:
+ v: + I&
(2.5)
where 3 = 21. The bare coefficient Z can be interpreted as static dielectric response function due to &I = -Z&L with the electronic potential I_Lbeing conjugate to the occupation number n. In a microscopic formulation of the theory’&) the quantity Z represents an effective density of states
Z(T) = Ix (- ma4 k
= J( dENr(E)(-
af’/aE),
(2.6)
which depends sensitively on the band structure through Nl(E) and on the temperature T due to the Fermi distribution f:(E). Here, K = (k, I) with I referring to the band index. Indeed, in A15compounds Z = Z(T) increases when the temperature is lowered. Thirdly, we assume a deformation potential coupling between the shallow d-bands and the elastic lattice. To lowest order, the contribution due to the electron-lattice interaction turns out to be of simple structure in the symmetrized basis F&E, v) =
E
-
‘9 * u
(2.7a)
with the coupling matrix V. It is noteworthy that the shear strains ~4, ESand 6, do not couple in our model which means that only electronic fluctuations within each band are accounted for. In analogy to the lattice term, eq. (2.4), it follows from (2.7a)
R. KRAGI.ER
380
F&E,
V) = (%,, + 2!9&,u,
Following
Pytte’),
9,, = -2%,2, shears
we choose
so that
the longitudinal
the coupling
E? and E+ Then
(2.7b)
+ (Y,, ~ %,Z)(~Z~Z+ QV?). deformation
is restricted
the coupling
matrix
potential
to volume
such that
conserving
% can be written
elastic
as
%= 09, where
(2.8)
D = S,, - 9J12is the deformation
potential
which
the band edges due to an applied uniaxial strain. The matrix 9 is diagonal with ?pZZ= Y,, = I being accounts for the fact metry and, henceforth,
where
non-zero
the shift of only.
:‘P,, = 0
that an isotropic deformation E, cannot change symgives rise only to s-d electron redistributions, which
are neglected in the present paper. 9 assumes the following form?: y-1
considers
In terms
of the Cartesian
basis the matrix
.gb.y=s = j-;j
4; is a constant
matrix
(2.9) which
has all elements
equal
Hence, the matrix elements of .9 are ‘9,, = : and convince oneself that 5!*= 9 is an idempotent matrix. Finally, we add an external field term F&r,
to unity.
‘3i, = --- I. It is easy
(2. IO)
v) = v . 5’“’ + G * c?“‘.
where 5’“’ = (5;“‘. . . . &‘Yt) and 6”’ = (US”. . . . . CT?‘) are the external potential and stress components in the symmetrized basis. Because
chemical
of
5”“’= y * P ex’
(3. I I)
there is v - 5’“’ = n * pex’. As regards to the stress that a factor i is omitted tribution
to
will be useful
thermodynamic
3. Equations
potential
components so that Z
oY’ to (T;;X’we follow
for the derivation consists
the convention field con-
-Gex' = e:Ct. The external of the response
matrices
below.
The
of the sum of (2.4) to (2.7) and (2.10).
of motion
In order to study the dynamics of the present system we have to start from the equations of motion for the electronic band occupation n on the one side and the displacement field u on the other side. As to the electronic d-band system we assume that the band occupation numbers n relax towards their instantaneous equilibrium values no(e) belong-
BAND
JAHN-TELLER
ing to a distorted ansatz
which
structure
reference
accounts
SYSTEMS
configuration
for
this
of this rate equation
AIS-STRUCTURE
of the elastic
behavior
is most
WITH
lattice.
is a Debye-type
transparent
381
The simplest
relaxation.
in the symmetrized
The basis
with v and E (3.1) where
the relaxation
matrix
r is chosen
to be of the form
r=+P.
(3.2)
rll = 0, leading to ti, = 0, accounts for the fact that the d-electrons is conserved in our model. In the corresponding
total number of equations for u2
and v~, describing the relaxation of the orthorhombic and tetragonal configurations, rZ2 = Ta = y is assumed for reasons of simplicity. y plays the role of an interband relaxation rate characterizing the electron redistribution between equivalent d-bands. The rate equation (3.1), taken for granted here, can be justified on a microscopic basis’6”,b). Indeed, as has been demonstratedlti), rate equations of this type are deducible from a multipleband Boltzmann equation to which the relaxation-time approximation has been applied, however, modified in a way so that local particle-number conservation is guaranteed. The instantaneous equilibrium value V’(E) which appears in eq. (3.1) is determined
by the requirement
1
* v = - yZ(DP
where the relation The change-over 9-l
rewritten
t
(6F/6v) = 0,giving rise to (3.3)
eq. (3.1) becomes
&1+yB C
with
i.e.
* E - Zg”“‘.
IJO(e) = - ZDB Hence,
of stationarity,
where
* E + 9 . &‘“‘),
9’ = 9 has been used. to the Cartesian basis 9
turns
into
9
(3.4)
is accomplished
according
by multiplication
to eq. (2.9). Then,
eq. (3.4) is
as
$+y9).
with the matrix
n = - yZDiR( V) * u - yZ9 -
prxt,
(3.5)
R. KRAGIXR
382
As the elastic
lattice
u is of the usual the linear
is concerned
the equation
form but with an additional
coupling
between
the elastic
of motion induced
lattice
for the displacement
stress
term arising
and the electron
bands,
from
thus (3.7)
The meaning
of this extra
term is such that the deformation
takes
place
with
respect to a distorted reference configuration depending on the electronic band occupation n. The instantaneous equilibrium value e”(n) is inferred from the condition of stationarity, expression for the divergence V.c:e”(n)= As regards
(SF/Se) = 0. which of the induced stress
-DLRT(V).n-
coupling
* n = g(V)
v.$?D.2
to
the
(3.X) side of eq. (3.8) which originates
(2.7). we made use of the identity
* 9 f n = 9T(
With the help of eq. (2. lb) which strain tensor e to the displacement ing abbreviated notation
V) . n.
(3.9)
relates the diagonal u. and, moreover.
elements ef’ = v,, of the introducing the follow-
V. c : e = V. c : Vsu = (cV@ V) - U. where c is the fourth-rank equation (3.7) can be casted
Taking
the Fourier
transform
quantities are denoted end up with (WI + iys) (pw’1-
(3. IO)
tensor of the into the form
elastic
moduli,
the
cV@ VI * u=mr(v)*n+V~cr”‘.
(p $I-
u +iDCR’(q)
Christoffel
(3.1 I)
of eqs. (3.5) and (3.1 I). where
by the same
symbols
* n - -rZDLB((q) * u = - iyZ9
cq@q).
following
V.(Y”‘.
to the first term on the right-hand
from the electron-lattice
leads
which
the transformed
now depend
- y’“‘.
* n = -icP.q.
of (3.12a, b) it is expedient to rewrite For a further treatment coupled equations in a more succinct fashion by introducing (3 X 3) matrices:
on (q, w), we
(3.12a) (3.12b) this system
of
the following
A(w) = ~7 + iy$
(3.13a)
B(q) = - yZI%Uq).
(3.1%)
BAND
C(q)
JAHN-TELLER
SYSTEMS
WITH
AIS-STRUCTURE
383
(3.13c)
= iD9T(q),
wq,4=pw2+w3qq,
(3.13d)
E = iyZ.2.
(3.13e)
Henceforth,
eqs. (3.12a, b) can be combined to the following matrix equation:
A(w) Wd C(q)
D(q, A
* (:I=
(3.14)
cr~f.e”d)-
4. Bare response
According to eqs. (3.13b) and (3.13~) the matrices explicitly on the deformation potential parameter D, coupling between the elastic lattice and the electronic electronic band system is decoupled from the elastic (3.14) turns out to be simply
B(q) and C(q) depend which brings about the d-bands. For D = 0, the lattice. The solution of
@(q, w) = X0(% w) * 6U(q, WI,
(4. la)
Wq, 0) = E”(q, w) - Wq, ~1,
(4. lb)
where we have redefined the thermodynamic 6p = n,
6d = Xu,
variables (4.2)
and the conjugate fields SD = Cc=‘, &I = (iX)-’ aext - q.
(4.3)
X = q2fJop is a normalization factor with 0, being an appropriate frequency which characterizes the harmonic lattice. The bare response matrices are given by X’(q,o)=-A-‘.E=-Z
(4.4a)
P(q, w) = x*D-’ = (2&p) . (pw21 - q*fi-1.
(4.4b)
For the decoupled subsystem, the frequencies of the pure modes are given as the eigenvalues of the matrices A(w) and D(q, w). According to det A(w) = w(w + i-y)*, the response matrix X0 exhibits relaxation poles. In addition, the origin of all term terms involving the matrix 9 can be traced back to interband processes. Therefore, from the specific form of X0 one may conclude that interband transitions are responsible for the dynamical redistribution of electrons in order to bring the perturbed electronic band
R. KRAGLER
384
system
back
suggestive
into
For the response from
(4.4b).
which
equilibrium.
to interpret that
Since,
for
D = 0 one
X” as the bare dielectric
has
Sp = X0. SU, it is
response.
will be quite similar. matrix 3:” the argument mode frequencies 8’ has poles at the acoustic
are determined,
for example.
as the bare displacement
response
As to the electronic band Debye-type, the eigenvalues
by eq. (4.5b). Thus, of the harmonic
It follows w = tf1,
E:” can be interpreted
lattice.
system which exhibits a relaxation behavior of and eigenvectors of the homogeneous equation
A*n=(wl+iy!I)*n=O are easily
found
to be n “I = (I 11)/\/3,
(R,),
n “) = ( ii2)/\/6, 1 n”) = (I iO)/v2,
(R,).
w,=o: w I.3 = -iy
:
(4.6)
(RI).
with the mode R, corresponds to a uniform The eigenvector n”’ associated band occupation, which is just another way to express the number conservation of d-electrons. The remaining modes Rz and Rx, which are proper relaxation modes, describe tetragonal and orthorhombic band occupations, respectively. As the lattice
is concerned,
the familiar
Christoffel
equation
for an isotropic
solid D. u = (co21 ~ p ‘$A). is recovered for vanishing A = CK @ K only depends elastic
moduli Cjq+
tensor
c
u = 0.
(4.Sb)
external stress on the direction
and contains,
Us”. Here, cosines K
for cubic
the Christoffel matrix = q/y and on the bare
symmetry,
the elements
(C,, - (.jq)K;, (4.7)
(Cl?+
CM)KiKj
(i#
j).
For wave propagation in certain symmetry planes, such as (001) and (Oil). closed-form solutions for the frequencies of the acoustic bulk modes w = t f1, can be given=). Eigenvalues and eigenvectors for certain high-symmetry directions of the cubic crystal are summarized in table I. If the coupled D is non-zero and, henceforth, the matrices 6 and C are present, it renders possible, for vanishing external fields, to find even then closed-form solutions of the homogeneous system of coupled equations (3.14) for high-symmetry directions. As expected, some of the relaxational and the acoustic modes turn out to be mixed, while others are still unaffected by the coupling. This is particularly the case for the relaxation mode R. Whether
BAND
JAHN-TELLER
SYSTEMS
WITH
385
AlS-STRUCTURE
modes are coupled or not depends on the direction of q. Table II lists the coupled modes, the frequencies of which are always determined by a cubic secular equation of the form NA(q, o) = (w + ir)(W*- 0:) + i-yfi&
(4.8a)
= 0,
where (4.8b)
f_xA = a,ZD2q21pR:.
The frequencies 0, occurring in eq. (4.8a) are those of the harmonic lattice given in table I. The prefactor ar in (4.8b) is a numerical factor given in the third column of table II, accounting for the different renormalization of the
TABLE Frequencies
Rh and polarization
e(A)
along high-symmetry
I
of acoustic
directions
bulk
modes for propagation
of a cubic crystal
ci
e(A)
A
(1W
LA
‘I-IA ‘LA
1(010) (001) [1101
R:=~(C,,+C12+2Ckl)q21P
(1 IO)/+2
LA
n: = f (Cl1 -
(I io)hd
TIA
0: = c44q21p
cool)
TzA
R:=f(cll+2c12+4CM)q2/p
(lllph
LA
c12)q21p
11111 TIA n:,,=f(CII-C12+C44)q21~
TzA
TABLE Coupling
of relaxational
The frequencies
and acoustic
of the coupled with the quantity
i
[ t@J
[ttol 11111
Coupled
modes
II
modes
in the high-symmetry
modes are determined or specified
in the fourth
a&
ar
a1.A= 213
2ZD2/3c,,
&@TIA
ar,A= l/2
ZD'/(c,, - cd
Rs@LA
a1.A = l/6
ZD2/3(c,r
aTA = l/3
ZD’/(c,,
R, 69 TzA
(4.8)
column
R,$ LA
R~@TIA
directions.
by a cubic equation
+ Cl2 + 2644)
- c12+ CM)
R. KRAGLER
386
bare acoustic determine of the latter
modes
relaxation ones
decreases. section
A. The roots
the frequencies mode
become
coupled
normal
with
the acoustic
(w = -i-r)
damped
A detailed
of eq. (4.8a) which
of three whereas
discussion
the attenuation
of the secular
are in general
modes.
complex,
Due to the mixing
modes
(w = + 0,)
of the relaxation
equation
the mode
(4.8a) is postponed
to
8.
5. Generalized
response
Now, we extend electronic potential present.
our investigations to the case where external fields, the pex’ and the elastic stress sex’ introduced in eq. (2.10), are of CL”’ and iarrt - q one has to
In order to find solutions n and u in terms invert eq. (3.14) with the result (6
. D-’ . C ~ A)-‘.
. A-‘.
-(D-C
-(B.D
E
6)-l.
C.
(-A-‘.
‘.C-A) E)
‘.E-C’.D -(D-C.A
’
i
‘.B)-‘/
(5.1) The form
of this equation
which
yields
respect to the external fields pext and following dynamical susceptibilities: X(q, w) = (6 -6-l
g(q)
we define = (-
IIX)C’=
response
iaex’. q. suggests
. C - A)-’ . E.
E(q, w) = N2(6 - C . A ’ . 6) Moreover,
the linear
(iDlN)%(q).
u with the
(5.2a) (5.2b)
‘.
an electron-lattice
of n and to introduce
coupling
matrix (5.3)
proportional to the deformation potential parameter D. The physical meaning of the susceptibilities X and e, as given by (5.2). is elucidated, when the coupling is switched off. Then, the matrices 6 and C vanish, and hence X and E simplify to X0 and E”, see eq. (4.4). Since X0 and E” are the bare response matrices, X will be the dielectric response and Z the displacement response modified due to the coupling of the lattice to the relaxing band system. The notation of (4.2) and (4.3) allows to rewrite eq. (5.1) in a succinct form (5.4) The appearance
of the cross-terms
X. g . E” and Z . g’ . X0 in the response
BAND
JAHN-TELLER
SYSTEMS
WITH
AIS-STRUCTURE
387
matrix is a consequence of the electron-lattice coupling incorporated in the coupling matrix g. One easily convinces oneself of the symmetry relation X.g.ELXQ.g*= which guarantees the cross-symmetry of the response matrix. It is evident from eq. (5.4) that there are two contributions to the dielectric response (Q/W),
= X,
(6~/6J)~ = X0 - g - 8,
and, likewise, for the displacement (sd/sJ),
= E:,
(5Sa)
response
(Sd/SU), = E”. g+ . X.
(5.5b)
The subscripts J and U remind for the fact that the corresponding external fields are kept fixed, which means that either the elastic lattice or the electronic band system is assumed to be rigid. In full analogy to the microscopic treatment’@) the off-diagonal contributions to the response can be rewritten in such a way that the fields SJ and 6I.7 are eliminated in favor of 6du = B - 6J and Sp, = X - SU. This allows a very physical interpretation of the response in the coupled electron-lattice system: Henceforth, electronic fluctuations 6p are either caused by a variation of the external potential 6IJ = pext, directly affecting the band occupation, or result, rather indirectly, from an external stress 6J a aext - q. This gives rise to an elastic deformation du according to which the bands are shifted by means of the deformation potential coupling and respond, in turn, with a redistribution of electrons. Thus Sp=X.6LJ+X”.g.6du.
(5.6a)
Similarly, elastic fluctuations 6d are either caused by applying an external stress 6J = aext. q directly to the elastic lattice, or arise, in an indirect way, from an external potential 6U which unbalances the band occupation 6~~. Because of the coupling the lattice reacts with an elastic distortion to the electron redistribution. Therefore, 6d=8.8J+e0.g+.L$,
(5.6b)
Now, the response matrices X and 8, given by eqs. (5.2a) and (5.2b), can be redefined in terms of infinite series of X0 and Z”, respectively, leading to Dyson-like matrix equations. As regards to the dielectric response X, one rewrites eq. (5.2a) into X=[i+(-A-’
.E).C*.C-‘.C]-‘.(_A-‘.E),
(5.7a)
where use has been made of the relation 6 = E * CT which holds because of
3xX
R
KRAGLER
9 * 3 = 3. With (4.4a) and P(q, w) = ~ CT * D_’ * c = g(q). which is interpreted form of a matrix
as an electronic Dyson
E”(q. w) * g’(q), “self-energy”.
(5.8a) eq. (5.7a) is casted
into the
equation
x=x”+xO*P*x.
(5.9a)
The graphical representation of this equation is shown in fig. la. Iteration of X leads to a chain of X0’s, where each X” is connected with neighbouring ones by the intermediary of the quantity P. Strictly speaking. if we compare this equation with our microscopic transport theory”‘), P is not a self-energy but rather an irreducible scattering vertex which involves square of the electron-lattice coupling g and the bare lattice “propagator” the latter being represented by a wavy line, see fig. lb. Obviously, terminology used for the diagrammatic representation is taken from many-body theory. In an analogous fashion, the displacement response can also be rewritten into ~:-[1+o~‘.C.(~A~‘.E),C’l Using
‘.,v”D
eq. (4.4b) and defining
II(q, o) = p.Y-‘C
. (-A~
we end up with a matrix
an elastic
‘. E) . C’ = g+(q).
Dyson
equation
of eqs. (5. IO) and (5. I I) 5, defined
by eq. (5.2b).
‘,
“self-energy”
the E:“, this
(5.7b) by
X”(q. 0).
g(q).
(5.Xb)
of the form
_= = _=(I +E”.rI.E.
(5.Ob)
the graphical representation of which is given in fig. 2a. The quantity II would correspond to the irreducible polarization part involving the bare electronic fluctuation “propagator” X” which is contracted by g, see fig. 2b. Perhaps, the interpretation of X, 2, P and II in terms of propagators and self-energies seems far-fetched at this stage. However. this point will be elucidated in the Conclusion, where the present phenomenological model will be compared with our previous formulation of a multiple-band transport theory”), which amounts to the same results in the hydrodynamic limit.
(0)
Fig.
I. (a) Diagrammatic
m
illustration
terms of X0: (b) the electronic
= )x0(
+ ) XouPyi
of eq. (5.Ya) which gives the exact
“self-energy”
P. eq. (5.8a).
dielectric
re\ponae
X in
BAND
JAHN-TELLER
SYSTEMS
WITH
AIS-STRUCTURE
Fig. 2. (a) Diagrammatic illustration of eq. (5.9b) which gives the exact in terms of B’; (b) the lattice “self-energy” H, eq. (S.8b).
6. Calculation
of the displacement
displacement
(6.1)
of @ are as follows:
= - bqiqi, aq:, @ii
@ii = c -
t
matrix B is defined through eq. self-energy II times X2, see eq.
Q, = (D - C - A-’ - 6) = (pw21 - q2A) - ,yD29tT(q) - 3 (q). elements
response
response E:(q, w)
For Df 0, the exact displacement response (5.2b). Since C - A-’ - I3 is equal to the lattice (S.lOb), we obtain with (5.3) and (4.4a)
The matrix
389
(6.2)
where a =(c,,-c&-?xD2,
(6.3a)
b = (~12 + CM)- fxD2, c = pw2The calculation
c4‘$q2. of a
requires
inversion
of Q, and multiplication
with X2 =
21&p. Hence
(6.4) with matrix
elements
(ii = (C - aqf)(C -
aq:) - b*qiqi,
5ij = Eji = [b2q2k+ b(c - aq:)]qiqi,
(6.5)
where (i, j, k) is understood as a cyclic permutation of the indices (x, y, z). Because the pole structure of the displacement response 5, as given by eq. (6.4), is not easily seen for general wavevector q, we shall concentrate on the special case q = q(llO)/d% Ancillary to eq. (6.3a) we introduce the ab-
R. KRAGI,ER
190
breviations d = bq?/2. with which
e = c ~ u472,
the determinant
(Mb) of @ simplifies
to
det @““) = c(e? ~ d’). The matrix
elements
(6.6)
of 5, eq. (6.5), reduce
[X.X= tyV = ce,
5:: = ez - d’.
[AL= 5y.r= cd.
[.rL= [v: = 0.
to (6.7)
so that
(h.8)
Immediately
accessible
to
physical
interpretation
are
only
certain
linear
combinations of the matrix elements of E”‘“‘, which determine the renormalization of the bare acoustic mode frequencies due to electron-lattice coupling
=
=
20” iy (02 - .n~)+-------w+iy
ml iy-(W*- n:>+----w +iy
#‘O)(q, n,.
(&‘)a)
ZD*q” 6p
(6.9b)
ZD*q” 2P
w) = E:“I’,‘“‘( q, w) = $J+
flz and a3 are the harmonic
frequencies
(69c) of the longitudinal
mode.
of the
transverse shear-modes with polarization along (liO] and [OOI], respectively, propagating in [I IO]-direction. However, due to the coupling between the harmonic lattice and the relaxing electron bands, some of the elastic modes couple to a relaxational mode, as given in table I, and are, thus, modified. The frequencies of these renormalized modes are determined as the poles of E?‘“‘, eq. (6.9a-c). Obviously, the denominators of 5”“” and .\“O’ lead to the same cubic equation (4.8) with &A = l/6 and a r’A = l/2, respectively. The TzA-shear mode, however, does not couple to one of the relaxational modes and remains therefore unchanged. Similar results for Ep’(q, W) would be obtained for
BAND
JAHN-TELLER
SYSTEMS
WITH
AIS-STRUCTURE
391
propagation along other high-symmetry directions, such as [loo] or [Ill]. As expected, the coupling of those modes is in agreement with table II. Henceforth, details will not be given here, since the calculation is analogous to that for q(([llO]. It is of interest to study the displacement response function E$““(q, w) in two limiting cases: (i) In the high-frequency limit o * y, the coupling of the TrA-shear mode with the relaxation mode R2 is negligible (6.10a)
Poles occur at the harmonic mode frequency w = k&(q) found in the high-temperature regime. (ii) In the case of the low-frequency limit w + y the electron band occupation follows the slowly varying lattice distortion adiabatically. Hence, local equilibrium is established and the poles of (6. lob)
now occur at the renormalized frequency fi:(q, 7’) = n:(q) - Ss(q, T), where 6!(q, T) = Z(T)D2q2/2p. In the soft-mode theory the (quasi) harmonic frequency &(q) is commonly denoted by wee whereas w. substitutes the renormalized frequency &(q, T) which goes to zero and is thus called the soft-mode frequency. With the static displacement response function 8$r’o’(q) = -2&/fi:, following from (6.10b), the expression for the displacement response function, see eq. (6.9b), can be casted into a canonical formkx25)
WO’(q,w) = aY’O’(d
w2
_
fi*;
2
!jJr (q 2
o)’
(6.11)
7
This is the dynamic susceptibility of a damped harmonic oscillator with a frequency-dependent damping r2(q, w) = 6:(q, T)/(o - ir) which couples linearly through 6* to a relaxing mode with relaxation time l/y. The free oscillator has the (quasi) harmonic frequency R,(q). Due to the coupling to a “slow variable” which comprises the electronic degrees of freedom, the low-frequency susceptibility a l/n: turns out to be a l/h;. At the structural instability the temperature-dependent renormalized frequency fi2(q, To) vanishes while the harmonic frequency L?,(q) remains finite.
W?
R. KRAGI.ER
7. Calculation
of the dielectric
For the calculation
response X(q, w)
of the dielectric
response
matrix
X it is expedient
to
start from eq. (S.7a) X=[l-XO*P]
‘ax”,
(7.1)
which gives the exact response matrix X in terms of the bare response matrix X” and the electron “self-energy” P. According to eq. (5.8a), P requires the knowledge of the bare displacement response matrix z” which is easily derived
from eq. (6.4)
E”(q, 0) = with matrix
$$!$43%WI,
elements
6: given
(‘7.2) by eq. (6.5), however.
u and h substituted
(7.3a)
&I = (Cl1 - C44), h,, = (C,? + (‘4). The same considerations
hold for det a”. The index
help of eqs. (5.3) and (4.4a) the quantity
x0 * P = (D2x(w;y)/.hw(q)
* mq,
=
4q?5P,
5ij
=
5jt
+
4f5;
+
q:SL
0 denotes
X0. P can be rewritten
D = 0. With the into
w).+?‘(q) = zC(q, w),
with the abbreviation z = D*,Y(w; r)/9 det a” The matrix elements of 5 are defined as lit
by
-4qiqj5~-4qiqk5Yk
and
+
(7.4)
X(W; y) = -- Z iy/(w t iy).
2qjqk‘$$.
(7.5)
where
=
-2qitC
2qf5;
(i. j, k) are cyclic
+
qt[L
permutations
[ 1 - X” . PI-’ = [det( 1 - zg)lm’S(q,
+
sqiqjt1:
~
q&J&I:
-
q,qk[yk,
of (x, y, z). It follows
from eq. (7.4) that (7.6)
WI.
with l9ii = (I -
Zljj)(I
-
i!
-
(7.7)
Z'
(ii, j, k) cyclic) i%,
=
6ji
=
ZJij(
1 - Z{kk)+
Eq. (7.6) multiplied matrix X
Z2c&{jk.
from the right with X0 finally
x(w; Y)
‘(% w, = 3 det(, The matrix
elements
_ zl;) @(q, w).
of 0 = 36 - 42 turn out to be
yields the dielectric
response
(7.8)
BAND JAHN-TELLER
SYSTEMS
WlTH
AlS-STRUCTURE
393
@ii = 26ii - 6ij - i?iky
(i, j, k) cyclic
(7.9)
Oij = 2611 - 93ik- IYii.
Although
9 is symmetric,
this is no longer true for 0
since (6 - 2)‘=
4*6+6*2.
This rather involved expression for the dielectric response matrix X(q, w) is best appreciated by considering again the special case for q = q(llO)/~/z as was already done for the displacement response Z(q, w). First of all, =6”“‘, required for the evaluation of P, is obtained from eq. (6.8) by replacing the expressions for c, d and e by co = p(w2- n:>, do
=
Cc12
+
(7.3b)
c&2/2,
eo= pw2-
(Cl1
+
c44)q2/2,
according to which the determinant
as follows:
fl$)(w2- 0:).
@b”” = p3(02 - flf)(o’-
det
of a0 factorizes
(7.10)
Then, the matrix ~‘““’ reduces to
p3’(q,
w)
=
pq2(w2-
n:>
(7.11)
;; ( -
t 53
with 51= f [w2 -
(Cl1
52= -2[w253 =
[w2
-
(CII
+:
(Cl,
cl*+ +:
c44)(q2/2P)ll
s
c12+
%
c44m2/2pN,
(7.12)
c12)(q2/2P)l.
-
Because of (51
-
52)
=
P (w2
(5,
+
52)
=
t 53 =
-
m
(7.13) the determinant
t cm2
-
m,
of (7 - z{) turns out to be
det( 1 - z~(“~‘)
=[
(02-n:,+&3gq. [(&n3+-L&z3
Introducing
(w’the abbreviation
n:>-(02-
0:)
(7.14)
394
R. KRAGLER
k = z(w’the matrix
0;)p2q2 =
elements
x(w ; y)D2q2 9p(J
- n:)(02
(7.15)
.- n;)’
of ?%(“O’ (q, o) simplify
considerably:
19::‘“’ = 1911’“’= (1 - k<,)( 1 - k12) - i k’& 19$czl;“’ = (1 - k<,)’ - k2,$ (7.16)
S$“) = 19$‘~’ = k&( 1 - k(j) + ; k’& “‘0’ = 19.G with which
~"'0, L1
-
$j'l_lO,= 'L
the matrix
~("0, zv
elements
=
_I
2 &Al
~
kc'+
k52),
for 0”‘“’ can now be calculated
according
to
the prescription of eq. (7.9). However, in analogy to (6.9). we are only interested in certain linear combinations of @I!“’ which allow an immediate physical interpretation of the dielectric response X”“‘. Hence (m’-
n;> + ~
iy
ZD’q’ - __
w+iy
iy w +iy
(w2-fj9+
Finally, ponents
21,
I*
ZD’q’ . (w’~ 6P 1
by multiplication with x(w; r)/3 det(l of the dielectric response X”“’ are
xl”O’(q,w)
xi”O’(q,
=
x!i:‘“;‘o’(q, w) + x:‘,‘O’(q,w)
=
lixb; Y) 1- x(w ; r>(~2q2/6p)/(w2 - 0:)
w) =
X$‘O’(q,0)
~
(co- 0;) ‘. 0;)
‘_
- z&“‘“‘) the
(7.17)
relevant
com-
(7.18a)
x’,‘:“‘(q. 0)
(7.18b)
x(w; Y) = 1- ~(0;
rW2q2/2p)/(~2 - 0:)’
x$“O’(q, w) = X11’O’(q, w) = 2x’,“‘“(q. If we remember the form from eq. (7.4) by making following expressions: Pdq,
w)
=
't3:<'$)= I
w).
(7.18~)
of the electron “self-energy” P”“‘. easily obtained use of the property % . CR= Se, we are led to the
P3(q, w), (7.19)
P2(q, 0) = Q$@$ wMoreover, expression,
with given
02.
LZ,=~,,+&=$, C&=&-&=1 and in eqs. (7.18) can be casted into the form
S,-Sr,=I
the
BAND
JAHN-TELLER
SYSTEMS
WITH
AlS-STRUCTURE
395
(7.20) Eq. (7.20) clearly exhibits an enhancement of the bare dielectric response matrix X’(q, w) = x(w; y)S due to the electron-lattice coupling. Henceforth, the simple relaxation pole of x(w; y) = -Z iy/(w + ir) is modified such that due to the mixing with acoustic modes there occur altogether three poles determined again by the cubic equation (4.8) in section 4. However, when comparing the dielectric response function X’j’“‘(q, w), eq. (7.20) with the corresponding displacement response function a’:“‘(q, w), eq. (6.9~) one feature is noteworthy: as expected, #:“1”“’has a pole at w = +Rj. In contrast to that, the pure relaxation pole w = -iy in X1”” is modified due to its coupling to the LA mode with w = +R1.
8. Coupled modes For the calculation of the renormalized response matrices s and X in sections 6 and 7 we have confined us to q\j[llO]. This choice was motivated by the fact that the softening, as known from the experiments on Nb$Sn3,4), is most pronounced for the shear mode propagating in [llO] direction with polarization along [iio]. Obviously, both the displacement response #‘“‘, eq. (6.9b), and the dielectric response XP’O), eq. (7.18b), can be casted into the form W”‘(q, w) = (0 + iy)/%(q, X$“‘)(q, w) = -iyZ(w’-
w),
02,)/N?(q, w),
(8. la) (&lb)
where the common denominator Nz(q, w) is the polynomial given in eq. (4.8) specified for q1][110] and A = T,A. The zeros of Nt(q, w) = 0 which is a cubic equation in o, determine the frequencies w,(n = 1,2,3) of three coupled modes for given q and CQ.Since the subsequent discussion is focused only on the behavior of the [ 1lO]TiA-mode mixing with the relaxation mode RZ, the index A will be suppressed in the sequel. As to the interband relaxation rate y, introduced in (3.2), this parameter of the present phenomenological model characterizes the tendency of the electronic band system to relax towards equilibrium by means of redistribution processes. Therefore, y cannot be handled as a freely varying parameter such as temperature or wavevector. Only the case of strong interband coupling, i.e. l/y-+0, deserves special attention. In this limit the acoustic modes
1%
R. KRAGl.ER
w,,? = ?&(q,
T), see eq. (6.10b), are undamped
This temperature
dependence
states
Z, eq. (2.6), occurring
looked
at as a measure
relative
deviation
temperature
for
values
equation
w1.3= 2 \/fit
with temperature.
elastic
softening, shear
since
modulus
it determines
the
(c.:, - CT,) at a given
(c,, - c,?) at room temperature. cy = 0 and u = 1 it is easy to find the exact
(4.8a).
pled from the relaxation and the pure relaxation be the high-temperature In the opposite
the
of the renormalized
from its value
For the limiting the secular
but decrease
comes into play because of the effective density of in CY= ZD’/(crr ~ c,?). The quantity (Y itself can be
In the case
CY= 0 the acoustic
mode. Because the bare sound mode wJ = ~ i-y are recovered, limit.
modes
roots
of
are decou-
modes wI.? = -t R?(q) 0 is considered to
ck =
case cy = 1 one determines 7 - y-14 - i y/2,
(8.2)
(02 = 0.
The soft-mode w2 is an indication for a lattice instability in the system occurring at a transition temperature To which is determined through tv(T,J = 1 and defines the stability limit of the high-temperature cubic phase. Whether the modes wI,? are overdamped or not depends on the magnitude of fl?/y. For C&/y c 4 the root wI = -inSly is proportional to yz and could therefore be attributed to a diffusive mode with diffusion constant fi = ~$7, where ri = root ~3 = -ir[ I - (0,/-y)‘] is approximately in(C11~ c,2)/2p. The other dependent of q and hence interpreted as a relaxation mode. For intermediate values of CYail three modes are strongly affected by the coupling and mingle with each other so that a numerical treatment is required. Studying the q-dependence of the coupled modes for 0 < cy < I provides insight as to how the soundwave dispersion is modified due to the deformation potential coupling. For simplicity, a linear dispersion law &(q) = roq is assumed for the bare acoustic modes at 9 = 0. Fig. 3 shows
the motion
of the normal
plane as a function of wavevector q the direction of increasing q varying boundary fqBz = y/u{,. In addition to of the soundwave-like mode w~(RJ~;
modes
in the complex
frequency
where the arrows on the curves indicate between zero and half the Brillouin zone the o-plane also real and imaginary part cu) are given as a function of q. For four
values of the parameter CY,suitably chosen, typical situations are encountered: For cy s 8/9 the damping of the acoustic modes w,,? increases with q whereas the relaxation mode w3 becomes less and less attenuated and moves towards the origin. For cx approaching the value 8/9 the curves of the modes w,,? become pinched for q - y/2u0. Eventually, at (Y= 8/9, the frequencies of all three modes coincide for q3 = y/d/3v0 at w = -iy/3. As regards to the dispersion Re WI the deviation from linearity increases with 0~. A bump develops which becomes more and more pronounced and turns into a cusp touching the
BAND
JAHN-TELLER
SYSTEMS
WITH
AlS-STRUCTURE
391
Rew,
-lkw, l
Rew
k-+-
-Imw
-IRlW
a=1
Rew,
Rew,
t
t
a)819
-Imw,
-Imw, Rew
;T-Re’ + -Imw
a = 8/9
-Imw
a<819
Fig. 3. Motion of the coupled mode frequencies wi (i = 1,2,3) in the complex o-plane for four values of a and 0
398
q-axis
R. KRAGLER
in q3 for cy = 8/9. In contrast
smoothly
with q, exhibiting
a whole
range
for which
of q-values
all three
modes
reappear
teristic
feature
normal being
is clearly
to this remarkable
no dramatic between modes
strongly reflected
change.
are overdamped due
~ Im w, grows occurs
- cu/o,, and q2== y\/3
ql = 2y\/l
damped
behavior
If (Y> g/9, there
until for q > qZ propagating
to mode
in the dispersion.
now
- 2cu/2vo
mixing.
This
charac-
For 0 d q s q, Re W, goes
through a maximum and vanishes again for q = q,. Then, between q, there appears a gap where Re W, = 0. i.e. no propagating modes exist. stability limit (Y= 1 this gap then extends from zero to q,, = y/2vo. The in the w-plane, formed by the mode frequencies o,,? for OS q s
and qZ At the “loop” q,, has
contracted to zero as CY+ 1. A soft-mode w2 = 0 occurs in agreement with eq. (8.2), whereas the other mode W, remains overdamped for 0 s q s q,,. Only above q, damped but propagating modes CD,,? reappear. This behavior is
0.0025 Fig. 4. Dispersion of the soundwavelike value n/y = I corresponds to q = ! q~,.
mode.
Re(~,/y),
as function
of rl/y
‘* q and
(1. The
BAND
JAHN-TELLER
SYSTEMS
WITH
AlS-STRUCTURE
399
clearly demonstrated by Re wI which is different from zero only above the threshold qo. As regards to the damping, - Im wI increases continuously with q and assumes the constant value - iy/2 for q > qo. Figs. 4 and 5 summarize in perspective graphs the dependence of soundwave dispersion Re ol and the damping Im wl, respectively, on wavevector q (0 s q s { qsz) and parameter (Y(0 s (YG 1). The interesting behavior obviously occurs in the region 8/9 s (Ys 1 where the dispersion curves clearly show a gap and the damping curves develop an edge which becomes more and more pronounced as (Y+ 1. It can be concluded from this discussion of the behavior of the coupled modes, that the essential feature of the modified dispersion is a strong deviation from linear behavior over an appreciable region of the Brillouin zone when approaching the stability limit. Concomitant with this deviation there is a strong increase of sound wave damping which vanishes in the high-temperature limit.
Fig. 5. Damping
of the soundwave-like
mode,
- Im(wl/y),
as function
of O/y a q and cx.
R. KRAGLER
9. Conclusion In the foregoing studied
in some
feature
of
coupling
the
there
renormalization
sections detail
model,
it turned
is a mixing of these
the dynamics
by analyzing out
of acoustic modes
of a band
Jahn-Teller
the generalized
response.
that
because
and relaxational
with respect
of
the
system electron-lattice
modes
to temperature,
was
As a salient resulting
wavevector
in a and
relaxation rate. Recently, the same physical system has been treated by the author on a microscopic basis with recourse to multiple-band electron-phonon transport theory’6”.h). In the sequel we refer to these papers as I and II. Although the model differs considerably from the treatment of this microscopic phenomenological one, presented here, there is, nevertheless, far-reaching agreement between the results derived from both models. Therefore, a closer comparison of both models which will be the subject of this section. is well suited. In the phenomenological model the thermodynamic state characterized by the d-band occupation-number deviations
of the system n
and
the
is dis-
placement gradient V,u. In terms of these variables the thermodynamic potential is constructed subject to certain symmetry requirements. As regards to the microscopic model of ref. 16, a Hamiltonian description is used. The operators corresponding to the thermodynamic variables n and u The Wigner operator (-L desare P,,’ = aLa,, and A, = bh + bk, respectively. cribes density fluctuations between different electron bands where the index K = (k, I) comprises both wavevector k and band index I. A, denotes the usual phonon normal coordinate of mode A = (q, j) where j designates the branch index. Similar to the thermodynamic potential defined in section 2 the corresponding Hamiltonian for the electron-phonon system consists of three contributions: one for the phonons in quasi-harmonic approximation, another part for band Anharmonicities account
electrons and the as well as Coulomb
usual electron-phonon coupling term. interaction are not explicitly taken into
in this model.
As in eq. (2.10) external fields Jh and UK,, which couple linearly to A, and p,,,, respectively, serve as a mathematical device to generate with the help of functional derivative methods various correlation functions together with the corresponding integral equations defining them, and, moreover, establish relations for the generalized linear response. The phenomenological equations of motion for the electron band occupation n, eq. (3.S) and for the elastic displacement u, eq. (3. I I), which were the very starting point of our model, have their counterparts in the microscopic model. As the phonons are concerned, this is the equation of motion for the
BAND JAHN-TELLER
non-equilibrium
expectation
SYSTEMS
WITH
AIS-STRUCTURE
401
value ((A)), see ref. 16a ibidem, eq. (I. 2.29)
DO’ - ((A)) = J + g((p)) = j.
(9.1)
Here, j is an effective phonon source field incorporating the feedback of the electron bands with respect to the elastic lattice. DO’ is a differential operator which defines the bare phonon Green’s function Do. It is now a straightfoward procedure to rewrite eq. (9.1) into (3.12b). In order to do so, eq. (9.1) has to be Fourier transformed, multiplied by (2flhp)“*e(X) and subsequently summed over all phonon branches j. e(A) is the polarization vector for the phonon mode A. Then, the resulting equation will coincide with (3.12b) if the following transcriptions are made:
ui(q, ~1 c 2 (2~*f)-“*ei(A)((A,(w))),
(9.2a)
i
ndq, 0) =
c k
VfK(q,w>=
c
with A = (q, j) and K% = (k 2 i dynamical matrix cq@ q with
c&l q =
ci
(9.2b)
N(PK+KJWNh q, I).
Furthermore,
one has to identify
e(A)pR%*(A),
the
(9.3)
and relate the gradient of the external stress tensor sex’ to the phonon source field Jh - iuext(q, o) - q =
2 (2R,p)“*e(A)J*(w).
(9.4)
If for the deformation potential coupling tensor G”,,,,,,,,,occurring in the electron-phonon coupling g, eq. (II. 5.10), the restrictions G” = - 2G12 and G, = 0 are made ancillary to the symmetry requirements which also apply to the tensor of the elastic moduli, then D%(q)
(9.5)
= Ge.mmqm.
Using these definitions, eq. (9.1) can be casted into the following form: (PO21 - cq@ q) * u = x * &
(9.6)
where Ni = - iaext. q -
iDaT
- n.
(9.7)
Obviously, eq. (9.6) is identical with (3.12b). It is noteworthy that inversion of eq. (9.1) provides a relation for the linear response with respect to the
402
R. KRAGLER
effective
phonon
source
field j, eq. (I. 6.4), namely
6((A)) = Do. S.f Thus,
the equation
the linear
side are equivalent.
electrons
the analogous
procedure
In order to end up with a relaxation-type
occupation
a Bethe-Salpeter
(9.1) for ((A)) on the one side and eq. (9.8a) for
on the other
to the band
is not so obvious. d-band
of motion
response
As regards
(9.8a)
numbers equation
n, in the microscopic
model
leading of equation
to (3.12a) for the
we have to start from
(I. 3.3)
Ko = GG + GGIoKo
(9.9)
for the electron density fluctuation propagator K0 = ((ApAp)). Moreover, G = i((p)) is the usual electron Green’s function and I, is the irreducible particle-hole scattering vertex. As shown in lbh), the integral equation determining K0 can be brought into the form of a Peierls-Boltzmann equation (II. 4.1) supposed the ladder approximation is applied to the scattering vertex IO, see (I. 3.10), and linear response theory is used in order to establish a connection between the fluctuation propagator K. and the deviation affr = a((~,+, )) from the equilibrium distribution ft = (pKK), 6((p)) = (- i)Ko - 80, Here,
So
(9.8b)
is an effective
electronic
potential
defined
as
Sti = U + g - ((A)),
(9.10)
which takes into account the feedback of the elastic lattice with respect to the electron bands, similarly to j in the case of phonons. The structure of the resulting Boltzmann equation is even closer to (3.12a) if the collision term is taken within the relaxation-time approximation (II. 4.2) modified
in such a way that local electron
Finally, if we sum the resulting equation n(q, w), and take the long-wavelength relaxation-type of equation: (wl
+ i#?)
number
conservation
is guaranteed.
over k in order to get an equation for limit, we end up with the following
- 66,
- n = -iyZ2
(9.11)
where Sti = peX’+ iIN?
- II.
Here, U = cc’“’ has been used deduced from a Bethe-Salpeter (3.12a). As to the
microscopic
origin
(9.12) together with eqs. (9.2) and (9.5). Eq. (9. I I) equation (9.9), is obviously identical with of the
relaxation
rate
y and
the
effective
BAND
JAHN-TELLER
SYSTEMS
WITH
AlS-STRUCTURE
403
density of states Z both occurring in (9. ll), a comparison with the microscopic model reveals that y is associated with the interband scattering rate ylr, (l+ 1’) arising from the relaxation-time approximation. The interpretation of the quantity Z which emerges in the course of the derivation of the generalized Boltzmann equation is in agreement with eq. (2.6). Up to now we have retraced the fundamental equations of motion (3.12a) and (3.12b) of the phenomenological model on a microscopic basis. Yet, the analogies between both models go further. Concerning the generalized response it turns out that its structure is the same in both models, compare eqs. (II. C.16) and (5.4). Moreover, eqs. (5.5) and (5.6) which refer to the more physical interpretation of the response have their microscopic counterparts in eqs. (II. C.17) to (II. C.19). As to the response matrices X0, X, so and z which are defined through eqs. (4.4) and (5.2), within the microscopic model these quantities are deduced from the propagators Ko, K, Do and D, see eqs. (I. 2.19) and (I. 3.1) by suitable retardation and contraction, for example
Xdq, w) = - i W) =
Bjj'(q,
z. ’ 1g g &,ldkc,
DjjJq,
W +
k'e', qw + is),
iS),
(9.13a) (9.13b)
see eqs. (II. 2.9) and (II. 5.6). The bare displacement response E” is obtained from eq. (9.6) by functional differentiation with respect to J in agreement with the definition of the bare propagator Do = S((A))/SJ. The bare dielectric response X0 is similarly deduced from (9.11) by application of S/Sfi. This procedure concurs with the definition of the bare fluctuation propagator (- i)Ko = S((p))/So, see (9.8b). One easily convinces oneself that the expressions for E” and X0, using eqs. (9.6) and (9.11), are identical with those of eq. (4.4). In contrast to the present phenomenological model, however, the microscopic treatment is not restricted to the hydrodynamic limit o, q * v, -=zy. Eq. (9.11) is only a special case of the more general, q and o dependent kinetic equation deduced from (9.9). Henceforth, the expression for X0 obtained from the generalized Boltzmann equation (II. 4.7) c
t
w
y/se,) q.u,+iy,[(W-4’0.)~+iY~l.SP(q,W)
=-‘q
w-q-i)
(af O,/W [- q - u,l + iy?L] - SU(q, w), K +iy s
where ys = El, ylr and y = i -rrr (I+ I’), consists of two parts according to (II. 4.13). There is a diagonal contribution proportional to 1 due to intraband
404
R. KRAGLER
processes and another to interband transitions. In the hydrodynamic
off-diagonal
contribution
limit the dielectric
proportional
response
function
to 2 attributed (Sp/Sfi)
resulting
from eq. (9.14) X”(q, 0) = w
Ligq2(iQ21
+
iyw 9 (w + ir) + igq2 1
(9. IS)
exhibits an (intraband) di$usion pole w = -igq’ with diffusion constant 9 = 2)zF/3ys, (see ref. Iti) ibidem eq. (3.15a)) for the diagonal part and an (interband) relaxation pole w = -iy for the off-diagonal part. Clearly, the diagonal contribution vanishes in the limit q +O and there is full agreement with eq. (4.4a). Of course, guided by the transport equation (9.14), which results from a microscopic treatment, one could think of generalizing the simple relaxation ansatz for the electron bands by supplementing the left-hand side of eq. (3.1) with a drift term. However, this would only lead to a kinetic equation which is obtained from (9.14) when u, is replaced by ut so that the k summation drops out. The corresponding expression for X0 is of the same structure as (9.15) with the important difference that igq’ is formally replaced by -q +z)~.not being justified at all. The conclusion, one can draw from the foregoing consideration therefore is that the resulting expression (4.4a) of the phenomenological model is just correct for q < ~‘\/5y/vr where the relaxation pole is predominant. However, for larger q, where intraband diffusion takes over, one has to proceed along the lines given in ref. 16c using a more general Boltzmann equation approaches instead of the simple rate equation (3.1). Irrespective of these limitations, the matrix equation (5.9) defining the renormalized response functions X and E in terms of the bare quantities X0 and E”, have microscopic counterparts. Since X is obtained from the fluctuation propagator K by means of the prescription (9.13a), it is near at hand interpreting the iterative equation (5.9a) as the retarded version of the integral equation (I. 3.13) K = Ko + Kol,K.
(9.16a)
The renormalization of KO involves the chain scattering vertex I, = -igDog which considers the contribution of phonon chain diagrams not taken into account in the ladder approximation. Therefore, it is evident from the microscopic derivation that P = g - 2’ - g+, as given by (5.8a), does not correspond to an electron self-energy but rather to a retarded scattering vertex involving phonon chains giving rise to an enhancement of X0. Again, in the hydrodynamic limit, the microscopic expression for X, eq. (II. C.8b), is in agreement with the result (7.20) of the phenomenological model.
BAND
JAHN-TELLER
SYSTEMS
WITH
AIS-STRUCTURE
405
Similar considerations apply for 8, related to the exact phonon propagator D by (9.13b). It is therefore suggestive to interpret eq. (5.9b) as retarded phonon Dyson equation (I. 3.15) D = Do + D,,IZD.
(9.16b)
The renormalization of DO due to the phonon self-energy II = g(--i)&g, eq. (I. 3.16a), only considers the effect of electron density fluctuations according to the underlying physical restrictions of the microscopic model. Thus, making contact with the microscopic treatment, the quantity II = gt - X0 - g can be looked at as a retarded phonon self-energy, see (II. 5.7). Again, in the hydrodynamic limit the expression for E provided by the microscopic model, eq. (II. 5.15b), coincides with the result (6.9b) leading to the same cubic secular equation (4.8) resp. (II. 6.1) which determines the frequencies of the coupled modes. In conclusion, the detailed comparison of the phenomenological model, presented here, with a previous microscopic model, based on multiple-band electron-phonon transport theory, reveals a remarkable similarity in the structure of both theories. In spite of the different starting points the results of both models agree in the hydrodynamic limit, so that both formulations provide equivalent descriptions of a band Jahn-Teller system.
Acknowledgement
The author wants to express his gratitude to Prof. R. Klein for his interest in this work and for critical reading of the manuscript.
References 1) L.R. Testardi, in Physical Acoustics, vol. 10, W.P. Mason and R.N. Thurston, eds. (Academic Press, New York, 1973) p. 193. M. Weger and I.B. Goldberg, in Solid State Physics, vol. 28, H. Ehrenreich, F. Seitz and D. Turnbull, eds. (Academic Press, New York, 1973) p. 1. Yu.A. Izyumov and Z.Z. Kurmaev, Sov. Phys.-Usp. 17 (1974) 356. L.R. Testardi, Rev. Mod. Phys. 47 (1975) 637. 2) L.R. Testardi, T.B. Bateman, W.A. Reed and V.G. Chirba, Phys. Rev. Lett. 15 (1965) 250. L.R. Testardi, R.R. Soden, E.S. Greiner, J.H. Wernick and V.G. Chirba, Phys. Rev. 154 (1967) 399. L.R. Testardi and T.B. Bateman, Phys. Rev. 154 (1967) 402. K.R. Keller and J.J. Hanak, Phys. Lett. 31(1%6) 263; Phys. Rev. 154 (1%7) 628. 3) W. Rehwald, Phys. Lett. 27A (1%8) 287. L.J. Vieland, R.W. Cohen and W. Rehwald, Phys. Rev. Lett. 26 (1971) 373. W. Rehwald, M. Rayl, R.W. Cohen and G.D. Cody, Phys. Rev. B6 (1972) 363.
406
R. KRACI.ER
4) (a) G. Shirane.
J.D. Axe and R.J. Birgeneau. and J.D. Axe.
Phy\.
Rev. B4 (1971) 9957.
(c) G. Shirane
and J.D. Axe,
Phy\.
Rev.
J.D. Axe and G. Shirune.
Phy\.
Rev. B8 (1973)
G. Shirane
Phy\.
Rev. B18 (197X) 3742.
5) A.M.
and J.D. Axe.
Clogston
6) M. Weger.
and V. Jaccarino,
Rev.
1.B. Goidherg 7) J. Lnhht
Mod.
and M. Weger.
Phys. Rev.
J. Physique
J. I,ahbt
S. BariSit.
Solid State (‘ommun. R.W.
versity, R.W.
Cohen.
Cohen.
G.D.
Phy\.
IO) W. Dieterich H. Schuster
Cody
Rev.
2. Physik
Achar
and G.R. Z.
in
R.N.
(h) R.N. (c)
Z. Physik
and O.N.
Bhatt. Bhatt.
?I)
Anderson
llni-
19 (1967) X40. 1094.
IS?.
(1970)
X5: 247 H.J.
Symposium
Rev. H6 (lY73)
3yX4.
137.
346X.
(1971)
203:
Quei\\er. 24 (1973)
38 (1974)
Dorokhoc.
in
rd.
Festktirperprohleme (Pergamon
XIII.
Press/Vicueg.
122: JETP
J. Low Temp.
J. Physique
1,ett. 17 (lY73)
379; JFTP
X30. Phy$. 22 (1976)
1014: Solid State Commun.
McMillan.
_% (1975)
I: JETP
19 (lY7h)
I.ett.
21 (197.5)
1107
1.153.
Phys. Rev. Bl4 (1976)
1007.
1915.
Lee. Phys. Rev. B16 (1977) 42Xx.
Int. Conf.
on Quasi One-Dimensional
and H. Thomas, and E.I.
Blount.
and G.R.
to as II).
B39 (1980) YY.
Phys. Rev. 1,ett. 14 (1965) Blount,
Phys. Rev.
Phys. Rev.
L,ett. 25 (1970)
2172.
219.
166(196X) 1014.
Phys. Rev. B19 (1979) 3761.
Phys. Rev. BlO (1974)
Berlin,
to as I).
201 (referred
2. Physik
and E.I. Barsch,
Conductors.
and B. 1,eontiC. eds. (Springer.
102A (1980)
and J.I*. Birman,
Achar
154.
1725: Phy\.
102A (1980) 22 (referred
Physica
B.W. Battermann
J. Noolandi.
Physic\,
Andrews
938 (197X) 314.
in Proc.
Physica
R. Kragler
20) B.N.N.
Lett.
Status Sol. H76 (1976)
PhyGa,
S. BariSi~. A. BjeliS. J.R. Cooper Phys. 96 p. 234.
Klein
(St.
PhyT. Rev. 817 (1978) 2947.
Physica
(h) R. Kragler,
248 (1971)
Phys. Rev. B16 (1977)
Bhatt and P.A.
19) B.M.
cd\.
3SA (1971) 4X.
Phyx. Rev. 88 (1973)
235
44 (1976)
R.N.
17) P.W.
on I.OH Temperature
McCall,
1973) p. 359 (ihid. $5).
R.N.
18) J. Perel,
Int. Conf.
323.
_34A (1971)
Phy\.
in Proc. Nobel
and H. Thomas,
16) (a) R. Kragler.
(c)
246 (1971)
1,ett. 27 (1971)
State
Bhatt and W.I..
R. Kragler,
BS (1971) Y4l.
and D.M.
Phys. Rev.
Phys,. I,ett.
260: Sov. Phys.-JETP
Ciorkov
IS) R. Kragler.
Rev.
1176: Phys. Rev. B4 (1971)
Phys. I.ett.
Barsch,
Solid
Gorkov,
13) R. Kragler
Finlayson
Z. Physik
Physik
310; Sov. Phys.-JETP 14) (a)
1Ith
in Proc.
D.M.
and I..J. Sham.
MeiBner,
I>ett. 20 (1974) (h) I..P.
IYS. 30 (1969) Y55.
254 (1971) 464.
Oxford/Braunschweig. L.P.
29 (196X)
J. Physique
A40 (1972) 415.
25 (1970)
(a) l..J. Sham, Phys. Rev.
(a)
1621.
Phys. 1,ett. 37A (1971) 409.
(h) J. Noolandi
12)
31 (lY70)
Solids 28 (1967) 2477.
and J.J. Halloran.
Lett.
and P. Fulde.
W. Dieterich.
Advances
Solid\
153. 303, 70X.
9 (I97 I) 1507; Phy\.
and H. Schuster,
W. Dieterich.
1357.
1969) p. 1009.
and W. Klo\e,
W. Dieterich
(c) G.
Chem.
I.ett.
Alien.
and W. Klose,
W. Dieterich
B.N.N.
27 (1966)
l..J. Vielnnd.
1968, J.F.
St. Andrewc.
9) E. Pytte.
1X03.
c’3 (1971,) 1-23.
and F. Cyrot-Lackmann.
K. Sauh and S. BarikiC, Phys. Cody.
3Y7.
158 (1967) 647. 655.
S. BaGSit.
St. Andrew\
9 (1971)
1965.
121 (1961)
Phys. Rev. 172 (1968) 451: J. Physique
I.ahhC.
8) G.D.
27(1971)
175; J. Phys. (‘hem.
J. Physique
S. BariSii- and J. LabhC. J. Phy\. J.
l.ett.
Phys. Rec.
Phys. 36 (1964)
and J. Friedel.
J. Lahht!.
II)
Solid State (‘ommun.
(b) G. Shirane
616.
Dubrovnik
107X.
1979): I>ect. Notes in
BAND
22) L.F. Mattheiss,
Phys.
JAHN-TELLER
SYSTEMS
Rev. 138 (1965) Al 12; Phys.
WITH
AIS-STRUCTURE
407
Rev. B12 (1975) 2161.
23) The same symmetry for the deformation potential coupling matrix B is assumed by Bhatt and Lee, see ref. 14b, ibid. eq. (9). 24) A.G. Every, Phys. Rev. Lett. 42 (1979) 1065. B.A. Auld, Acoustic Fields and Waves in Solids (Wiley, New York, 1973) chap. 7. 25) Y. Yamada, G. Shirane and A. Linz, Phys. Rev. 177 (1969) 848.