Thermal conductivity of some low-dimensional magnetic
systems
Citation for published version (APA):
Buijs, J. A. H. M. (1983). Thermal conductivity of some low-dimensional magnetic systems. Technische
Hogeschool Eindhoven. https://doi.org/10.6100/IR154977
DOI:
10.6100/IR154977
Document status and date:
Published: 01/01/1983
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be
important differences between the submitted version and the official published version of record. People
interested in the research are advised to contact the author for the final version of the publication, or visit the
DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page
numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne
Take down policy
If you believe that this document breaches copyright please contact us at:
openaccess@tue.nl
THERMAL CONDUCTIVITY OF SOME
LOW-DIMENSIONAL MAGNETIC SYSTEMS
\ '
THERMAL CONDUCTIVITY OF SOME·
LOW-DIMENSIONAL MAGNETIC SYSTEMS
PROEFSCHRIFT
ter verkrijging van de graad van doctor in de technische
wetenschappen aan de Technische Hogeschool Eindhoven, op
gezag van de rector magnificus, prof. dr. S.T.M. Ackermans,
voor een commissie aangewezen door het college van dekanen
in het openbaar te verdedigen op dinsdag 14 juni 1983
te
16.00 uur
door
Johannes Antonius Henricus Maria Buijs
geboren te Gilze-Rijen
1983
Offsetdrukkerij Kanters B.V.,
Alblasserdam .
Dit proefschrift is goedgekeurd door de promotoren
TABLE OF CONTENTS
I INTRODUC'l'ION
II REVIEW OF TBERMAL CONDUCTIVITY IN MAGNETIC INSULATORS
2. 1. Int:roduction
2. 2.
Elementax>y rmgnetia exai tátions
.2. 2. 1.
Magnons
2. 2. 2.
Soliton8
2.3.
Phonon saattering p:roaesses
2.3.1.
Non-magnetia saattering
2.3.2.
Re8onant magnon-phonon interaation
2.;;.;;.
T!Jo magnon - one phonon saattenng
/2.3.4.
Other magnetia saattenng machanisms
2.4.
EarUer e:JJPerimentat !ûOrk
III EXPERIMENTAL METHODS
3.1.
Int-Poduction
3.2. E:JJPe'Pimentai apparatus
3. 3.
T!Jo-step measunng method
IV TBE . .RMAL CONDUC'l'IVITY OF CoBr 2
.6s
2o
4 .1.
Int:roduation
4.2.
E:JJPerimentai resuits on CoB-P
2.8H 2o
1 3 6 13 18 19 24 29 31 39 40 46 5Çl4. 2.1.
Temperature dependenae of the the'l'mat aonàua-
55tivity
4.2.2. Magnetia field ,dependenoe of the the:l'mal aon-
56duativity
4.3.
Interpretation of the e:JJPe'Pimentai result8
4.3.1.
Qualitative interpretation
584.
3.2. Calautations based on the r-esonant magnon:-
62phonon ,interaation
4.3.3.
Cafoutation8 based on the 2 rmgnon-
651 phonon saatte'Ping
.,·
\
4.4. Discuzsion
V THERMAL CONOOCTIVITY OF TMMC AND DMMC 5. 1. I ntrodwtion
67
. 5. 1. 1. Statia properties of TMMC and DMMC 71
5. l.. 2. E:reitatiorw in TMMC and DMMC 74
5.2. E:r:perimentai resuZts on TMMC and DMMC
5.2.1. Terrrperature dependenae of the themrr:t.Z aondw- 79
tivity
5. 2. 2. Magnetia fieZd dependenae of the the1'maZ 81
a<Yniluctivity
5. 3. Inte:rpretation of the e:r:per:imentai resuZts 5.3.1. Introdwtion
5.3.2. Interpretation based on magnorw 5.3.3. Inte:rpretation based on soZitorw 5.4. ConaZusiorw and remarks
APPENDIX A \ APPENDIX B REFERENCES SAMENVATTING 83 84 87 95 97 100 104 109
CHAPTER I INTRODUCTION
The el'ementary excitations in magnetic systems have alreaäy been
studied fora large'number of ye~rs. With respect to the subjects of
investigations one can distinguish between the statie and dynamic properties of magnetic systems.
The first category includes measurements of thermodynamic properties like specific heat, susceptibility, magnetisation and allied quantities. Common resonance techniques like NMR and ESR can aiso be employed. for the study of statie behavior. '!hese resonance techniques, however, can also yield information about the dynamic properties, for instance when relaxation effects are considered. Neutron scattering .and thermal conductivity measurements can likewise serve for the study of dynamic behavior. It is the last mentioned technique that will be of particular
interest in the i~vestigations presented in this thesis.
In thermal conductivity measurements the magnetic system cannot be studied separately. There bas to be a certain amount of magneto-elastic coupling in order to probe the influence of magnetic
excita-tions. An external magnetic field is a powerful tool in examining the
influence of these magnetic excitations on the heat transport. The crucial point in the interpretation of field and temperature dependent thermal conductivity data remains the knowledge of the nature of the
magneto-elastic interactions, which is reflected by the pos~ible
scattering mechanisms of elementary magnetic and latticé excitations. The interest in the dynamic behavior of magnetic excitations bas
increased enormously in the past few years. The discovery of the
occurence of solitons in low-dimensional magnetic systems may be considered as an important factor in this respect. These non-linear basic exci tations can exist together wi th the' well known linear excitations (spinwaves) and manifest themselves primarily in the dynamics of the magnetic system. The solitons are believed to occur only in the paramagneti,c phase of linear chain systems. Therefore we might expect to observe characteristics of both the linear and the non-linear excitations, depending on whether we are measuring in-or outside the three-dimensionally in-ordered phase.
1 1
i
The organisation of this thesis will be as follows. In cllapter II
we will present some introductoeyiconcepts. The basic excitations in
magnetic cfystalline insulators wil,l be outlined and a the.oretical
introduction to some magneto-elastic scattering processes will be
presented. This chapter closes with a brief survey of what bas been reported on the topic of heat transport in magnetic materials so far.
The experimental apparatus will be presented in chapter III, together with the measuring method applied.
In chapter IV we present results of thermal conductivity measure-ments in the compound CoBr
2
.6a
2o.
In this magnetically quasitwo-dimensional system the spinwaves (magnons) are the relevant magnetic i
excitations throughout the investigated field and temperature range
and consequently this system may serve
at
an indication for magnoninfluences on the thermal conductivity. A qualitative interpretation based on spinwave behavior completes this chapter.
The final chapter V will deal with the field and temperature
dependance of the heat transport
ln
two quasi one-dimensionalsubstances ( (CH
3l 4NMnC13 and (CH3)!2NH2Mncl3) in which soliton influences may be expected. Measurements are performed as well in
a~ outside the ordered phase, which allows a discussion of the
CHAPTER II REVIEW OF THE.RMAL CONDUCTIVITY IN MAGNETIC INSULATORS
2.1.
Int~oduationIn condertsed matter heat can be transported by the various
excita-tions which may appear, Since we will confine ourselves to magnetic insulators, contributions from conduction electrons will not be considered. The remaining excitations can be distinguished to occur either in the lattice or in the magnetic system. The lattice excitations are most commonly described by pláne harmonie waves. The
quantum-mechanical excitiation of a lattice vibrational normal mode is called
a phonon. An introduction to the description of collective lattice
vibrations in terms of phonons can be found in any textbook ön solid state physics. For a review of the description of the process of heat transport by phonons we refer to Klemens (1951), Carruthers (1961), Parrott and Stuckes (1975) and references therein.
The idea that also magnetic excitations (more specifically the magnetic counterpart of the phonon, the magnon} could have a contri-bution to the thermal conductivity was proposed by Sato in 1955 and was further investigated in many articles since then. In all investi-gations the existence of interactions between lattice and magnetic excitations is crucial. These interactions arise from magneto-elastic effects which are ,inevitable in any real solid. Without this
magneto-elastic interaction heat cannot be transferred from the lattice
vibrational system to the magnetic system and vice versa. A consequence of this interaction is, however, that magnetic excitations can also
have a negative influence on the phonon heat transport by the induced scattering. In genera! the strength of the mutual interaction
determines.whether magnetic excitations will have a net positive
or nega~ive contribution to the thermal conductivity.
Thè thermal conductivity is formally defined by the equation:
- À·'i/T, (2 .1}
~ 2 +
tempe-... . 1
rature gradient (K/m) and À deno~es the thermal conductivd.ty tensor
(W/mK). We will confine ourselves to cases where the heat flux is
parallel to .the temperature gradient. This leads to a simplification of equatlon 2. 1 :
(2. 2)
where a indicates a certain spatial direction. The macroscopie thermal conductivd.ty coefficient Àa can be related to thé microscopie parameters of the system within the relaxation time approximation (see for instance Carruthers 1961). Ina non-magnetic insulator, where the excitations
can be regarded as a gas of phonons, this approximation leads to the
expression:
+
+ 2 + +
Àa(T)
=
E
c
(q,T) vp,a (q)Tp(q,T)<i,p p . (2. 3)
The wavenumber q of the phonons runs over the entire first Brillouin
zone, while the suffix p denotes .the various acoustic and optical
branches with dispersion relation w (q). The specific heat of a certain
p
mode is 9iven by:
+
C {q,T)
p
The phonon veföcity is defined by:
+ + % +
v (q)
=
v..w; {q)p qp
- 1}2
The remaining and most complicated quantity in equation 2.3 is the
+
(2 .4)
(2.5)
phonon relaxation time T (q,T), which expresses the characteristical p .
time the phonon system needs to recover equilibrium after a disturbance.
The relaxation time can be built up from several independent
contribu-tions by adding the associated relaxation rates:
-1
T
As examples of specïfic scattering processes we mention the crystal boundary scattering, scattering from iatt:ice imperfections and all
kinds of phonon-phonon scattering processes. In the description of
these latter processes the ~il:hharmonic interaction t~rms have to be ·
retained in the hamiltonian under consideration.
In maqnetic insulators we have to include the maqnetic excitations in the model describing the thermal conductivity. In a maqnetic system with wave-like excitations we can also introduce a dispersion
+
re7ation wm(k}. The polarisation suffix m again indicates the possible branches, but is in this case restricted to one for a ferromaqnet and
+
two for a (2-sublattice) antiferromaqnet. The maqnetic field H now
enters the expression for Àa as an external parameter, leading to the extended version of eq. 2.3:
+ 2 + + + C (q,T)v (q)T (q,T,H) P p,a P + + 2 ++ + + + +~ C (k,T,H)v (k,B)< (k,T,B) k,m m m,a . m (2. 7) +
The quantities C and v in this case are formally identical to C
+ m m p
and v , provided the maqnetic excitations have the same boson
,P
description as the phonons. The maqnetic relaxation tiroo 'm as well 'as 'p becoroo fielÇl dependent due to the influence of various
maqneto-elastic scattering processes. Depending on the strenqth of the
spin-lattice interaction one can distinguish several cases (Walton 1972).
For a small but non zero interaction the maqnetic effects on T can
p
be negli;lcted and the two parts of eq. 2. 7 can be oQserved separately. A positive contribution to Àa of the maqnetic excitations may be
. expected in this case (Sanders and Walton 1977a). For large interaction strengths any maqnetic contribution to Àa will be masked by the
enhanced maqnetic scattering of the phonons. The second part of eq.
2. 7 can be omitted in this case and maqnetic scattering will cause
a severe reduction of À compared to the case of the non-maqnetic
system. In all intermediate cases both terms in eq. 2.7 have to be taken into account.
In a real system several methods can be employed to investigate the influence of the magnetic system pn À. One can comparé thermal conducti-vity data.on a magnetic compound with these on a non-magnetic isomorphous system. Assuming the respective lattice systems to behave identical, the difference in magnitude of the observed conductivities is
a
streng indi-cation for thé effect of magnetic excitations. In some cases, however, i t is impossible .to grow a non-(or dia-)magnetic isomorphous crystal and besidés, since À is a delicate function of crystal imperfections the two lattice systems do not have to behave similarly a priori. Therefore frequently insight in the role of spin-lattice scattering processes has been obtained solely from the temperature dependance of À. Much more conclusive evidence can be optained, however, by measuring the magnetic field dependance of À at a certain fixed temperature. In sufficiently high magnetic fields the activation epergy of the itiagnetic excitations will be raised so that they can b.e neglected at lowtemperatures and only phonon conductivity and scattering remains. By comparing the high field thermal conductivity with the zero field data one can conclude whether the magnetic excitations have a positive contribution to the heat transport or merely act as an extra scattering mechanism for the phonons~ In the present investigation we will explore this method. '
In the next section we will pay attention to the magnetic excitations of interest. Apart from linear·excitations (magnons), .we shal.l also consider non-linear excitations ("solitons") since i t has been. realized recently that these can play a dominant rele in the dynamics of certain magnetic systems. In section 2. 3 we shall describe the various phonon scattering processes, with the emphasis on magnon-phonon scattering. In section 2.4 we present a survey of earlier experimental werk, indicating what features may be expected in magnon bearing systems.
2, 2. Elementa:x>y magnetic e:ccitations
2.2.1. Magnons
with a predominating Heisenberg or XY-like exchange interaçtion can be descr±bed by linear spinwave theory (see for instctlice Keffer 1966). The temperature range in which this model is applicable depends, among others, on the spatial dimensionality of the magnetic system. For quasi 1-dimensional systems the validity range of the spinwave
'description extends even above the 3d oraering temperature, due to the well developed correlations in the chain. In 2d and 3d systems the
spin~ave concept is only valid well inside the ordered phase.
The compounds of present interest are considered to be good
physical realizations of the 2d XY model (CoBr
2.6H2
o,
chapter IV) andthe ld Heisenberg model (TMMC and DMMC, chapter V) respectively. As -+
·an example, the deri vation of the spinwave dispersion relations
w
(k)via the Holstein-Primakoff (HP) formalism (1940) for the 2d XY model
will be presented in the remaining part of this section. A small
orthorombic anisotropy will be included by means of anisotropic exchange -+
interac~ions. Results for W(k) in case of the 1d Heisenberg model
including single-ion anisotropy terms will be given in Appendix A. The
spinwave dispersion relations, together with the Base-Einstein dis-tribution function, can provide us with all thermodynamic quantities of the. magnetic system.
We start with the following hamiltonian for a 2-sublattice anti-ferromagnetic system:
(2.8)
m
The exchange constants Ja(a = x,y,z) are all negative and
!Jzl > !Jyl > !Jxl. This condition defines the three principal, magnetic
axes easy (z) , intermediate (y) and hard (x} • The indices il and
m run over "up" and "down" lattice sites respectively. The notation
<!lm> indicates that all nearest neighbour exchange couplings have to
be considered. With respect to the actual situation i~ CoBr
2.6a2
o,
we will restrict ourselves to the appropriate model in this case, i.e.Fig.2.1.
Schematio representation
of the 2d antiferromagnetic
struoture.
a 2d arrangement of spins on two penetrating sublattices (fig. 2.1).
The magnitude of the effective spin is S =
t.
We will first considerthe case where the magnetic field is directed along the easy axis. At
low temperatures three s~ccessive phases will be realized when the
field is raised, the antiferromagnetic phase, the spin-flop phase and
the paramagnetic phase. The derivation of wtk) in this case is
analogous to the more general treatment by Iwashita and Uryû 0975a). The basic step in the HP formalism is the introduction of rising
+ - +
-and lowering operators a.e,, a!l -and, bm' bm via:
s+ -
125
f.e,a~ s+ 12S'b+ f !l - m m msï
=55'.
a~f.e,
s-
m =v'2s'
f b mm-
(2. 9) z + - sz=
+ b+b-st=
s - at at -s m mmwith f.e, =
Á -
(a!ai/2S)1
and fm
=
/t -
(b:b~/2S)
1
•
In linear spinwave,theory one uses the approximation fN,m 0 = 1. Inserting 2.9 (fk,m 0 = 1) in
eq.
2.8 leads to a hamiltonian in terms of spin-deviation operators.The quadratic part of this hamiltonian is:
'(2. 10)
+ +
Next the spinwave creation and an~ihilation operators
aj;
and b; areirttroduced by means of the Fourier transformations:
++ ...
a+ =I~
'*-
+ -ik.ri b+ =I~ ~ b; e-ik.rm9.. N k .~ e m N k . (2 .11) ... + b- =I~ ... + - 2 ~ - ik.r9,,
!1.
b~
eik.rm at =IN ~e m N " k k + ...1In these transformations r9.. and rm denote the respective positions of
the lattice sites and N is the total number of spins. For convenience
/ ...
the vector signs of k in the indices of the magnon operators are omitted. The lattice vectors comprise a complete set:
+ + ...
E e1(k-k'l.r9,,=
!
Ó(k-k')JI, 2 (2. 12)
Inserting 2. 11 in to hamiltonian 2. 10 and applying the above completeness , .
relation leads to:
(2. 13)
with
...
·ït"
+ ... +d Y(k) '>' ei P, where 11 i hb
an
t
P =rJI, - rm runs over a . nearest ne g oursof one lattice site.In the configuration of fig. 2.1
y(k)
can be written as:+ .
y(k)
=
4 cos(ka a/2)cos(~ b/2) (2.14)Hamiltonian H~ {2.13) can be transformed toa diagonal represéntation
by means of a so-called Bogoliubov transformation. 'I'his transformation
is extensively studied by Colpa· (1978) from a mathematica! point of
can be found in Appendix B. The result of the diagonalization is:
(2. 15}
The spinwave spectrum consists of two branches
w
1 (k) andw
2Ck)
forwhich
th~ dispersion relations read as:(2.16)
q =
Ï<E~
+E;}
+E~
- E;
22 21 2 2 l:i
r = [(El +E2) E
3 + (El -E2} f4(El +E2) - E4})
In zero external magnetic field the spinwave spectrum exhibits two
+
e~rgy gaps at k
=
O, which are given by:(2. 17)
tiro2 (0)
Át the ZQne-boundary (y (kmax) = 0) the two branches coincide .• In fig.
2. 2 an e~ample of a zero field magnon dispersion is presented.
The antiferromagnetic phase remains stable for fields up to the spinflop critical field HSF given by:
(2.18)
In the spinflop phase we have to describe the system on another coordinate system. In fig. 2. 3 the new set of coordinates is denoted
by a tilde. The canting angle
a,
for which the linear part of the spinwave hamiltonian vanishes, fellows fromz z y z
sin
e
=
g llsH/-as
(J + J l (2. 19). The dispersion curves are now gi ver by (see appendix B) ~
0 kmax
Fig. 2. 2.
Zero fieid magnon dispersion
reiations for a 2-subiattiae
antiferromagnet.
z,~
Fig.2.3.
Loaation of the quantization
a,xes in the spin-fiop phase.
hw
1,2ckl
= /
(B1:!:
B )2 2 - B2 3 (2. 20)with B1 -8S (J y cos 0 - J sin 2 z . 2 0) + g z ~H z sin e
(-~
- JYs1n2e
+ Jzcos 2 +B2 = 0)sy(k)
~ y 2 z 2 +
B3
c-
+ J sine-
J cos eisy(kl 1Th.e paramagnetic phase (complete alignment of spins) is re;;i.ched for 0
=
îf/2. The torque balance equation 2, 19 provides US the valueof the paramagnetic transition field H,,E:
(2. 21)
In the paramagnetic phase the derivation of magnon dispersion is analogous to the AF-phase and leads to:
-hwl , 2
Ck
l= /
(C + c l 2 - c2with
In fact the two dispersion curves gi ven by eq. 2. 22 can be mapped into a single curve extending over the double first Brillouin zone; The dispersion is consequently identical to that of a ferromagnet. The complete field dependence of the magnon dispersion curves is mainly
....
reflected by the behavior of the. two k
=
0 modes. In fig. 2. 4 a representative behavior of these energy-gaps is sketched, with the magnetic field along the easy direction;Expressions foriiw
1,2Ckl with the field along the intermediate or hard direction can be obtained fr0m the equations 2. 19 - 2. 22 by the following interchange of parameters:
Fig.2.4.
Fieid dependence of the
energy-gaps ûlith the fietd
in the easy direction.
Fig.2.5.
Fieid dependence of the
energy-gapá ûlith the fieid
along the intemediate (full
curve) or hard direction
(braken curve).
1
I ,
+ H// inteniiediate axis + . Hl/ "hard axisThe appropriate g-values
(gt'
and. gx respecti vely) should be used. Thefield dependance of the energy gaps in these two "perpendicular" cases is schematically given in fig.· 2.5 (gz >
gY
> gx). rn this figure HPI and HPH denote the respective paramagnetic transition fields.The main features of the dispersion relations as derived in this section have been verified by experiment (Keffer 1966). Especially the behavior of the energy gaps, whi.<rh can be observed wi th anti ferro- · magnetic resonance techniques seems to be established rather well (see for instance Date and Nagata 1963 and Anders. et al. 1978). Therefore · we may conclude that linear spinwave theory offers a satisfactory description, at least in first approximation, of the magnetic exci tations of systems wi th nearly isotropic interactions' near the groundstate •
2.2.2. Solitons
Since a few years it is realized that in one-dimensional systems also another class of excitations may appear under certain circum-stances. In contrast to spinwaves these excitations are soluti9ns of a non-Unea:P equation of motion of the magnetic system. 'l'hey are denoted by the name "solitary waves" owing to their limited spatial
extent. A certain class of non-linear equations, among which the
sine-Gordon (s-G) equation, possesses solitary wave solutions which preserve identity after colliding. This special kind of solitary waves is called soliton. In 1978 Mikeska showed that the equation of motion describing a linear chain Heisenberg system with a small XY anisotropy
and a magnetic field in the XY plane could be transformed into the
s-G equation. Kjems and Steiner (1978) claimed the first experimental observation of solitons in the ferromagnetic chainlike compound CsNiF
solitons in magnetic systems, which is still g0ing on.
One-dimensional Heisenberg systems with XY symmetry can be described by the following hamiltonian
(2. 23) The single-ion anisotropy term (A > 0) fa.vours alignment of the spins in the "easy" XY plane. The spin-system of quasi one-dimensional (anti) ferromagnets in the para.magnetic phase is very adequately described by this m::>del hamiltonian. It has been demonstrated that in the limit of classical spins, the equation of motion associated to this hamiltonian can be transformed, within a continuum approximation, to a second order differential equation:
(2.24)
The transfo:rmation of 2.23 into the so-called sine-Gordon equation 2.24 is given in detail by Riseborough et al. (1981) for ferromagnets and by Mikeska (1980), Leung et al. (1980) and Maki (1980) for anti-ferromagnets. The angle q:l is defined in fig. 2.6. The "magnon veloc;:ity" c and the "mass parameter" m can be expressed in the parameters of hamiltonian 2.23. The results are comprised in table 2.1. The excita-tion spectrum of a sine-Gordon chain consists of three kinds of solutions, magnoris, kinks and breathers (Currie et al. 1980). The 'latter two are solitons.
c
m b.AF
Fig.2.6. Definition of the angte
lPfor
the F (a)
andAF
(b)ahain.
(<Pin
xy-ptane)F AF 2aS
r1ïJ'
fi 4t-
8IJl
~g
glla~
a 2JS4alJls
Table 2.1<' Parameter8 of
the sine-Gordon equation.
The
magnons
represent small amplitude oscillations {sin (j) Fl:l q>) of the spin system as a whole. The plane wave solutions (q> ,..,, exp.[i (WT ;... kzl] lare characterized by the dispersion relation:
(2. 25)
This expression, which also gives the associated energy hw(k), is a sim-plified case of the.· magnon dispersion that was more generally introduced in the preceding section. Because of the continuum approximation in the derivation of 2.24, relation 2.25 will only be appropriate for small wavenumbers.
The
kirik soliton
solution of eq. 2.24 is given byq>(z,t)
=
4 arctan exp±
ym(z - vt- z0) (2.26)
with
y
The kink velooity v can range from zero to the maximum velocity c. In fig. 2. 7. tP is plotted as a function of ym(z -vt - z l. The anti-kink
0 .
solution (minus sign in eq. 2.~6) is also indicated in this figure. With the definition of the angle q> as shown in fig. 2.6 we conclude that a kink soliton excitation in a ferromagnetic chain corresponds to a full 211' rotation of the spins within a limited distance in the chain direction. In the antiferromagnetic chain, however, the spins rotate over an angle 1l'. This distinction is related to the degeneracy of the antiferromagnetic groundstate. lj'ig. 2.8 shows an illustration of the 1l' and 211'-kink soli ton. For slowly movinq kinks the soli ton lenqth is approximated by 6/m. The energy associated with a kink
lt -4 0
'
. ' 'anti-kink ...._
--2 ym (z-vt -iol 4 Fig.2.7.Graphic repreeentation .of the
kink
andanti-kink eolution
of the eine-Gordon equation.
Fig.2. 8.
2:i-soliton
A kink soliton in
the AF and F ahain.
soliton is given by (see for instance Mikeska 1978, 1980): F. (v) = ~ (0)
K K (2.27)
The kink rest energy EK(O) can be specified by:
F E.K(O)
8(2g~HXJS
3)
l:i (2. 28)AF EK (0) ~HXS (2. 29)
The second type of non-linear solutions of the s-G equation 2. 24 is the
breather:
2 , {tan 6 sin [wB y (t - vz/c - t0} )} 1.p (z, tl = 4 arctan c0sh [my sin e (z - vt - z ) ] 0 (2. 30)· Thè form of a breather is that of a translating evelope (velocity v < c)
'
with an internal oscillation with frequency
w
8 < w0• The breather energy can b~ written as:
(2. 31)
OWing to this particular energy a breather i·s frequently considered as a pulsating kink-antikink bound pair ..
In a classical treatment of the excitations of a system described by the. s-G equation, · the magnons, kinks and breathers together exhaust the excitation spectrum {Currie et al. 1980). Inclusion of quantum effects, however, yields a different view of the excitation spectrum of s-G chains. Takayama and Maki (1980) showed that in fact kinks and
breathers are the elementary (quantum) excitations and magnons corres-pond to the lowest breather modes. Another effect of the incorporation of quantum corrections is the renormalisation of the kink rest energy
(?Jla.Jl;i 1981). Experimentally a reduction of EK(O) with, about 20% is found for as well the s 1 ferromagnet (Kjems and Steiner 1978) as the S = 5/2 antiferromagnet (Boucher et al. 1981).
In the derivation of the s-G equation 2.24 for the angle ~ the spins were confined to the XY plane. Consequently only one of the two
anti-ferromagnetic spinwave branches is obtained. Mikeska (1980) considered the more genera! case which includes out of plane spin motions. Only in the limiting case of zero applied magnetic field the out of plane spin motions could be described by a s-G equation. The energy of the associated "hard" kink solitons is characterized by:
(2. 32)
which is in fact the corresponding "hard" antiferromagnetic magnon energy multiplied by s. In finite magnetic fields the out of plane motions cannot be described by s-G equations. The excitations under these circumstancés will retain some soliton-like features and appropriate theoretical models are being developed.
the elementary excitations of the sine-Gordon chain, the question remains to what extent each individual excitation contributes to the 'thermodynamical propérties of the system. This problem is, however, ·not
completely solved. The magnon contribution to thermodynamic quantities can be detehnined straightforward. Because of the correspondence ·of
the magnons to the lowest energetic breather modes, further breather
contributions to the thermodynamics are left out of consideration. The total density of kinks (plus anti-kinks) is calculated to be (Currie et al. 1980, Leung et al. 1980) :
n K
=
(2. 33)Trullinger and Bishop (1981) have discussed the validity range of this expression for the kink density, ::r'hey conclude that for temperatures
kBT ~ 0,2 EK(O) equation 2.33 adequately describes the kink density.
At higher temperatures kink-kink interactions have to be taken into account. On the low-temperature side the onset of three-dimensional ordering restricts the existence of solitons in real magnetic chain systems. Inside the ordered phase magnons predominantly determine the thermodynamica! behavior. The energy associated with out of plane spin motions (compare eq. 2.32) is generally much larger than the in plane kink energy, so that "hard" solitons play a minor role in the low temperature properties of the magnetic system.
2.3. Phonon saattering pioaesses
2.3.1. Non-magnetia saattering
Pho~on scattering processes in non-magnetic crystals have been
studied extensively ,since they determine the thermal conductivity of
crystalline solids. We refer to the review articles by Klemens (1951)
and carruthers (1961) for a general introduction. Here we shall only give a very brief summary on the most important scattering processes and their influence on the thermal conductivity in dielectric materials.
In general the phonon scattering processes can be divided in two categories. The first category contains phonon scattering by crystal-boundaries and -imperfections with a temperature independent relaxation rate. The second category contains all kinds of phonon-phonon scattering processes which are essentially temperature dependent. The competition between these two categories of scattering processes causes the well known temperatµre dependence of the thermal conductivity of solids with a maximum in the cryogenic region. The limiting low temperature behavior is governed by the boundary scattering with relaxation time:
+ - +
T
8(q)
=
d/v(q) (2.34)In this expression
d
denotes the average crystal dimension. In realcrystals, however, in ternal grain boundaries may cause a mean free
path which is a few orders of magnitude smaller than
d.
The resultinglimiting low temperture conductivity will be proportional to the
specific heat (see eq. 2.3) of the lattice. (À,..., Tn in the n-dimensional
isotropic lattice). Scattering from lattice imperfections {impurities, isotopes, vacancies, dislocations) may result in a slightly different
low temperature behavior of À·. With increasing temperature
phonon-phonon scattering processes gradually become more important, leading to an ultimate decrease of the thermal conductivity. Our attention
will be focussed on the low temperatures (liquid 4He region) since
the most interesting magnetic effects.on À may be expected in that
region.
2.3.2.
Resonant magnon-phonon interaotion.
The total hamiltonian for a combined spin-lattice system can be written as:
(2. 35)
where the spin hamiltonian B5 includes all magneto-èlastic interactions.
as:
L
hw
(q) ( c + c - +.!.. )
+ p q,p q,p 2
q,p
(2.36)
Here c+ and c ~re the phonon creation and annihilation operators
q,p q,p
respectively. The general spin hamiltonian can be written as:
(2.37)
The succes si ve terms denote the gene ral exchange interaction, the single-ion anisotropy and the Zeeiran interaction. Dipole interactions can be included in the first term. Since the exchange and single-ion anisotropy parameters depend on the relative positions of the various ions, it is obvious that the associated energy is modulated by the
+
lattice vibrations. After defining the lattice displacement ufl by
+ +
r + u.0 fl,O x,
(2. 38)
we can expand the spinhamiltonian in the displacemènts as fellows:
(2.39)
The "pure" magnetic hamiltonian
H~
contains the equilibrium parameters'"*
-+ + *-+J 0_(r0 -r ) and D(r0 ) and can be treated according to the spinwave
Mlt N,O m,o N,O
formalism described in section 2.2.1. The interaction hamiltonian
H'
contains first order terms in the,displacements and the appropriate
"*
"*
gradients of JQJll and Dfl. At low temperatures higher-order terms in
the expansion 2.39 can be omitted. The first-order interaction
hamiltonian H' describes one-phon9n processes, as will be apparent
+ e+ ~
lw
(q} p (2.40L(see for instance Akhiezer et al. 1961). In this expressi~n
;;+
is theq,p
unit vector in the direction of the displacement and p and
v
d.enotethe crystal density and volume respectively. Introduction of the spin
d.eviation operators (eq. 2.9} in the interaction .hamiltonian
H'
willlead to terms of first-order in the phonon operators and of arbitrary order iri the magnon operators. These terms correspond to:the distinct processes one magnon - one phonon, 2 magnon - 1 phonon etc. The one
1 '
magnon~one phonon process, also called resonant interaction, manifests itself essentially different than the higher-order processes. The resonant interaction changes the equilibrium properties of i;he basic excitations, while the higher-order processes reflect dissipative scattering mechanisms. Therefore the resonant interaction requires a distinct approach.
The effect of the resonant interaction on the equilibrium properties of a coupled spin-lattice system is visualized as a hybridization of the 'respective magnon and phonon energies. The resulting cµspersion curves are schematically given in fig. 2.9 for the antiferromagnetic system. The process can be described by the approximated interaction
Fig.2.9.
Hyb'l'idization of magnon
(m1,m2J and
phonon (p)
energies as a result of
\
the resonant magnon-phonon
interaation.
! .
hamiltonian (see for instance Boiteux et al. 1972, Laurence and Petitgrand 1973 and White et al. 1965):
+ +
(2.41)
The operators
ak
and$k
are associated with the pure magnetic system(eq. 2.15). The interaction parameter Dk can be specified as Dk D
Ik
where D denotes a typical strength: of the repulsion of the branches.The magneto-elastic modes, as sketched in fig. 2.9, are now obtained
by diagonali~ing the total hamiltonian, which can be written as:
HT =
<{~~)
M[~]
(2.42)with M=
[
h~
(k)li~
(k)~
l
Die
o
hwm2
(klThe thermal conductivity of the coupled system can now be obtained by
performing the swmnation 2. 7 over the magneto-elastic modes
w
1 (k),
w
2 (k) and w3 (k), which correspond :to the eigenvalues of M. The main
problem in this procedure is the calculation of the relaxation rates
of. the magneto-elastic modes. An approximation of these relaxation
rates can be obtained by the method of Kittel (Laurence and Petitgrand
1973, Kittel 1958). This method is based on the presupposition that the relaxation rates of the unperturbed magnon and phonon modes
-1 -1 -1 . .
(Tml' Tm
2' Tp ) are known (see section 2.3.1 for phonon scattering
processes and Dyson (1956) for magnon scattering processes). For the unperturbed modes complex energies are introduced by :
liW
"'"hw
+ iÎi T -1 p p ptw
ml 11wm1 + ih
Tml -1 (2.43}tW
=hw
m2 .m2h
-1 + i Tm2Replacement of the real energies in the total hamiltonian 2.42 by the associated complex ones and following the same diagonalization procedure will lead to the complex magneto-elastic mode energies:
-1
The parameters Tj (j = 1,2,3) provide approximations of the ultimate
relaxation rates.
The method of Kittel, as described above, can be used to calculate all possible influences (positive as well as negative) on the thermal conductivity which may arise from the resonant interaction. In real systems, however, magnons frequently have no positive contribution,to the heat transport. In such èases the calculations of the thermal conductivity (eq. 2.7} can be restricted to the phonon modes while the
resonant interaction is included by the phenomenological exp~ession
for the relaxation rate (Walton et al. 1973, Tiwari and Ram 1980):
-1 T2W2 p
(2.45)
with
wc
the intersection frequency of the unperturbed magnon and phononmodes. In real systems, wit.h phonon velocities much larger than the maghon velocity, the intersection will occur at very small wave-numbers
.and consequently
wc
.can be approximated by wm(O). In anantiferro-magnetwtth two magnon branches, this leads to the approximate relaxation rate due to the resonant interaction:
-1 T (x) res 2 x (2.46)
with x =
Îiw
/k
8T, x1
=
k
1 (0)/k8T and x2=
fiw
2CO)/k8T. The field. \ p
dependance of this process follows from the field dependent behavior of the energy gaps (figs. 2.4 and 2.5).
2. 3. 3. Two magnon - one phonon saattering
The process of two magnon - one phonon scattering is described by
the terms of the interaction hamiltonian
H'
(2.39) which are quadraticin the spin-deviation operators. Expressed on the basis of the non
perturbed magnetic system, the second orde~ interaction hamiltonian
will take the following general form:
(2 .47)
The parameters R. . and
c. .
have to be calculated and will be a function+ + + l.J l.J •
of q, k and k'. The hamiltonian H
2 can be divided in two parts
describing radiation (R) processes (magnon + phonon & magnon) and
' conversion (C) processes (magnon + magnon <1+ phonon) respecti vely" These
processes will contribute to the relaxation rates of as well magnon as phonon modes. Again we shall restrict ourselves to cases in which magnon contributions to the thermal conductivity can be neglected. The two magnon - one phonon scattering in this case provides the simplest dissipa:tive magnon-phonon scattering mechanism which does not effect
equilibrium properties. Hamiltoni~n 2.47 refers to the unperturbed
magnon and phonon modes which implies that the resonant interaction is left out of consider:ation in the present description of the two magnon -one phonon process.
The requirement of conservation of energy and wavevector in two magnon - one phonon scattering processes restricts the contribution of
/
the various terms of hamiltonian 2.47 to the phonon relaxation rate.
Time dependent perturbation theory leads to the foHowing expressions
for the phonon relaxation rate for the two distinct classes of two magnon - one phonon scattering (see for instance Akhiezer et al. 1961, Keffer 1966): -1 + T R . . (q) l.J -1 + T C .. (q) l.J 2'11' ++++1 2 1 + + + I - h 1 R .. (q,k,k+q) n. (k) - n. (k+q)
h
2t
l.J l. J 1 + + + 1 +*
O( W. (k} - W. (k+q) - W (q)) l. J p 27T 1 + + + + 12 + + +'ti2
i
cij (q,-k,k+q) (ni (-kl+
nj (k+ql + '1)*Ö
cw. c-k>
+w.
ck~>
l. J + - w (q)) p (2.48) (2.49)These relaxation rates express the net result of a certain proc:ess, i.e. the difference between the phonon creation and the reverse annihilation process. In real physical systems with phonon velocities much larger than magnon velocities, not all processes can conserve energy and momentum. For instance in a 2-sublattice antiferromagnet only four
processes (and reverse ones) remain. These processes (~
1
, ~11
, c22 and c
21l are depicted in fig. 2.10.
Fig.
2.10.Allowed 2 magnon-1 phonon
proaesses in a typiaal
In actual qalculations of the relaxation rates 2.48 and 2.49 the
magnon spectra 1ll
1{k) and
ro
2Ck)
are assumed to be isotropic in n-+dimensional space. Under this assumption the summation over k can be
+
replaced by an integration over the n-dimensional k-space, leading to
the modified z:elax~tion rates:
(2.50) -1 + I + + + + · 1 2 + + + I ...., + + 1 -1 n-' T C (er) = F Ci. (q,-k ,k +q) {ni {-k ) +n. {k +q) +1) v
1{-k )+vj (k +q) k .
ij ~ n J c c c J c c c c
(2.51)
+
The pre-factor Fn contains all appropriate constants and kc denotes the, wavevector,for which energy is conserved. The only remaining problem
is the càlculation of the matrix ~lements Rij en cij. In the following
we will present an example of this calculation based on the spinwave
treatment of section 2.2.1. In fact this is a generalisation of the
procedure of Ono (1975) and Dixon (1980), who described the uniaxia,l
case.
Adopting the collllllOn simplification to ignore the non-diagon~l
elements of the general spin hamiltonian 2. 37, we start from the
·anisotropic exchange hamiltonian 2.8. The expansion of the exchange
parameters is in first order described by: (see eq, 2.38)
The interaction hamiltonian H1 {eq. 2.39) now becomes:
H' = -2 E
<lm>
(2.52)
+
+Inserting eq. 2.40 for uR, and um and defining:
(2. 54) leads to: + H' = +E H'(q,p) (2.55) q,p with + H' (q,p) (2.56)
The abbre.viation h.c. denotes the hermittian conjugated term. For the AF-phase we now introduce the spin-deviation operators (eq. 2.9) and retain terms displaying them in second order:
+ + +.+
z + - + - } -~ + - iq. rR,
iq.
rm-2A q,p (a0a0 +bb) sw (q)c (e· -e )+h.c.
N N m m p q,p (2.57)
The next step is the transformation to magnon operators via the Fourier
transforms 2.11. The result is:
+ { - - + +++ +- + + +
H 2
1(q,p)
=
L
A [~bk U(k+q,k) + ~bk U(-k+q,-kl]with A q -Ax + Ay q,p q,p A+ -Ax - Ay q q,p q,p Az -2 Az q q,p and (compare 2. 13)
From now on i t will be convenient to use a matrix representatiOn of the interaction hamiltonian 1
....
*
H2(q,p)....
~ ~ k+ql\~ c + h.c. (2.59) k>O q,p with ~=*
(~,bk, + + a_k,b-k) (2.60) and 0 0 0 A U(-k,-k-q} - -+ -+-+ q 1 0 (2.61)Let Ek be the matrix that diagonalizes
H~,
i.e. (Appendix B)(2.62)
with
'
....
Subjecting the interaction hamiltonian H
transformation (~.6) Will now provide US the desired matrix elements Rij ànd.Cij: (2.63) with + + + + + + + +
..:
+ ......
+ + ... R 11 (q,k,k+q) ~12 (q,k,k+q) c11 (q,-k,k+ql c12 (q,-k,k+q) + + + + + + + + + -+ + _,. -+ + + +<!\!,/
~1\1=
Rz 1 (q,k,k+q) R 22 (q,k,k+q) ·. c21 (q,--k,k+q) c22 (q,-k,k+q)*
+ + + +*
+ ... ++...
...
..,...,.....
... + ... c 11 (q,k,-k-ql c12 (q,k,-k-ql R11 (q,-k,--k-q) R12 (q,-k,-k-q}*
+ + + .... c 21 lq,k,-k-ql*
+ + + + c 22 (q,k,-k-ql + + + + R21 (q,-k,--k-q) Rz2 (q, -k, -k-q) + + + .... (2.64)This ultimate form of the interaction hamiltonian is equivalent to the
general hamiltonian 2.47. A numerical calculation of the matrix-elements
of 2.64, together with the approximated expressions 2.50 and 2.51, wi.11
now enable us to determine the 2 magnon - 1 phonon relaxation rates. In
the SF and P phase, as well as the phases with perpendicular field
direction, a straightforward substitution of
J'"
by Aa as describedq,p
abqve, will provide. us the matrix elements 2.64 of the var~ous processes.
2.3.4. OtheP magnetic ecattePing mecha:nismB
After the more elaborate description of two specific magnon-phonon scattering processes, we will now present some brief information on possible other magnetic scattering mechanisms in the compounds of interest. As already pointed out in the beginning of section 2.3.2, all kinds of multi magnon - multi phonon scattering processes may contribute to the phonon relaxation rate. The influence of these higher-order processes will of course strongly depend on temperature. Generally the first-order approximation in the phonon operators will hold for. temperatures small compared to the Debije temperature of
the crystal. This implies that one-phonon processes will dominate
in the liquid helium range. For the magnon operators the situation
is different. It is a well known fact (Keffer 1966) that spinwave~
theory up to second order in the magnon operators only gives satisfactory results in the zero temperature limit. In the inter-pretation of temperature-dependent magnetic properties the inclusion
~f higher-order terms in the magnetic hamiltonian,is essential.
Th,erefore we may expect an increasing contribution of higher-order
scattering processes (3 magnon - 1 phonon etc.) to the phonon relaxation
at'.higher temperatures. In the liquid helium range t.~e influence of
higher-order processes may already be considerable. In thermal
conductivity measurements, no direct evidence for·the influence of higher-order processes has been published thus fä.r. This may, however,
be related to the extremely difficult calculation of the associated
relaxation rates. In ether dynamic properties the rele of multimagnon processes has been established rather well (see, for instance, Nishihara et al. 1975).
Another type of magnetic scattering is the so-called critical scattering which can occur in the immediate vicinity of the ordering tenperature. In this process phonons interact with fluctuations in the magnetic 'system, which are typical for the three- dimensional orde ring phenomenon. The critical scattering is described by Kawaski (1963) and Stern (1965), while also the paper of Dixon (1973) may be consulted.
The critical scattering will manifest itself as a dip in the thermal
conductivity around the ordering temperature. Well inside the ordered phase this mechanism may be ruled out as a noticable scattering source for the phonons.
Finally some remarks will be made op the scattering of phonons by
magnetic solitons •. Since magnons and solitons are elementary excitations of a magnetic system with predomi)nating exchange interactions, which
strongly depend on position, the 1
existence of some kind of soliton-phonon scattering will be evident. One may visualize a soliton-soliton-phonon
scattering by ~oting that a soli ton in the magnetic system is
acconpanied by a local lattice distortion, if a magneto-elastic coupling is present. This lattice distortion acts as a scattering
source for .the phonons. Schöbinger and Jelitto {1981) developed a
simple model for the soliton-phonon interaction in a one-dimensio~al
system. They arrive at the conclusion that a renormalisation of phonon dispersionc arises, in addi tion to all kinds of two phonon -one soliton scattering of the Raman type. These latter processes imply an energy exchange between soliton and phonon modes with the restric-tion that only the kinetic part of the soliton energy is involved. This restriction refiects an essential difference with magnon-phonon scattering •. This distinction is strongly related to the different nature of magnons and (kink-)solitons, as was already pointed out in section
2.2.2. However, it remains to be investigated to what extent the
· conclusion of Schöbinger and Jelitto will be modified by the inclusion of lattice discreteness and quantum effects. We will return to this topic in chapter V •
. · It should be stressed once more that the magnetic scattering processes as described in the preceding sections refer to magnetic systems displaying collective magnetic excitations (Heisenberg and XY-like systems). The investigation of magnetic scattering phenomena in alternative systems (paramagnets, Ising-like systems) is beyond the scope of this thesis.
2.4. Ea:t>lier e:x:perimental
~orkIn this section we will give a review of developments in the field of 1;:hermal conductivity measurements on (anti-) ferromagnetic compounds. Our attention will mainly be airected to those systems exhibiting predominant exchange interactions i.e. systems in which the magnetic excitations in principle may have a positive contribution to the heat transport process. We do not intend to be complete but merely want to show what in genera! can be expected from experiments and what at-tempts have been made to explain the observed behavior. Extensi ve reviews on the same topic have been'given by Sanders and Walton (1977a) .and Dixon (1980). These authors also present rather complete lists
of references on the subject.
1.
with collective magnetic excitations roughly can be divided in two
categories: compounds wi th a posi ti ve magnon contributi.on to À and
'
those in which the scattering of the phonons by magnons predominates.
Oficourse.in any real system both effects will be present, but depending on the magnon-phonon scattering rate the net effect can be
pqsitive or negative. Since Hisano Sato (1955) su~gested that magnons
prob,ably could contribute to the thermal transport, several investiga-tors started to look for experimental evidence. The first results on
MnO (Slack and Newman 1958), on some ferrites (Douthett and Friedberg
1961) and on MnF
2 and CoF2 (Slack 1961) indeed indicated magnetic
effects on À, but positive contributions could not be reported. The
general tendency in these compounds is enhanced scattering of phonons
around the ordering temperature. In 1962 however, Friedberg and Harris
were the first to report rather strong evidence for a substantial lhagnon conductivity in Yttrium Iron Garnet (YIG). Comparable
measure-ments on the same compound appeared shortly afterwards by Lüthi ( 1962)
and Douglas (1963), and in 1973 again by Walton et al. The field and
temperature dependent data of figures 2.11 and 2.12 are taken from
;;:
5
~
.< s.o~--~,----~--~.,,...,.,-_, 2.0-1.0,... 0.5- 0.2-0.1 YIG ••
Q • 0 Q • 0 • •o•
e 0 0 • 0 • 0 x x • 0 0 x • 0 0 Xx X x x x • zero field tAr
> x 40k0e tApH l o Àf- ÀPH 1 1 1 0.2o.s
1.0 TlK> -Fig. 2.11.TheY'mal aond:uativity of
YIG as a function oftemperatUX'e.(Walton et ai.
1973)0.8 0.6 § ~ ~ 0.4 1 1 \ \ ~ \ \ \
·'
\ \ e'\._•',
.
;--... YIG • exp. data
- - theory without interii<:tio - - - theory including "
l•O.nK
Fig.2.12.
• .-- .... _ ..:a:::'"::::l...--..---,,r
02
•
· Fieüldependenae of
the thermalaonductivity of
YIG at 0.77K.(Walton et
al.1973)
0 2 8 10
this last article.
YIG is a three-dimensional ferromagnet. In ferromagnetic compounds application of a magnetic field results in a uniform shift of the magnon branch to hiqher energies. Ina field of 40 kOe*the lowest
magnon mode in YIG is at about 6 K. Therefore the results of À at
40 kOe in Fig. 2.11 may be identified with the phonon thermal conductivity <>pH> and \ , - \ s corresponds to the positive magnon contribution. As can be seen in fig. 2.12 this magnon contribution amounts to about 70%. The field dependent data can reasonably be explained by a theory comprising phonon as well as magnon contributions
3 2
{proportional to T and T respectively) and an additional resonant magnon-phonon interaction {dashed curve in fig. 2.12).
Further evidence for magnon heat transport has been observed in a series of two-dimensional Heisenberg ferromagnets (CnH
2n+lNH3l2CuC14 (CnèuCl). Reports on thermal conductivity in these systems have been published by Gorter et al. (1969), coenen et al. (1976 and 1977) and
*Throughout this thesis the cgs-unit Oersted (Oe) will be used for the magnetic field strength. In vacuum a magnetic field of 1 Oe
-4
àe
Lang et al. (1977). Figures 2.13 and 2.14 are reproquced from thelatter authors. In Fig. 2.13 it can be seen that in
c
1cucl the magnonscarry more than 90% of the heat at certain temperatures. This, is the
largest contribution ever reported. The dip in i,.(T) at Te (H
=
0 kOe) inFig. 2.14 emphasizes the fact that magnon-phonon scattering can not be
neglected. rurthermore we will mention the magnon corttribution
in
EuS ·(Me Collum et al. 1964) ëµid EuO (Martin and Dixon 1972), two
three-dimensional ferromagnets. There is only one .antiferromagnetic system in which magnon contribution is anticipated: Cu{NH
3J4
so
4.a
2o
by Miedema et al. {1964). The evidence, however, is not decisive.Examples of magnetic systems in which the magnon-phonon scattering dominates, are more numerous. In general, three specific scattering
1.0
•
•
c
1cuct
•
•
T•3.46Kê
•
::::
0.5•
s:
';;:•
•
• •
•
•
• • • •
•
0 5 10 15 H/TCkOe/KlFig, 2.13.
Field dependenae of the
the:r>rnal aond:uativit;y of
c1euct.The solid line
Pepresents the sum of
aalaulated magnon and
phonon aontributions.
(de
Langet al.1977)
20 ;;:: E v ~ -< O.l. OJ 0.2 0
Fig.2.14.
Temperature
d.epende~aeof the theY'fflal aonduativity
of C1CuCl at H=O
and60k0e.
0.2 0 5 HKl 10 20 Fi,g.2.15.
ThePmal aonduativity
data of FeCl
2
(open
airales).The
dra:i.unaurve is the result of
a resonant interaation
model.(Laurenae and
Petitgrand 1973)
mechanisms have been applied succesfully, i.e. the critical scattering the resonant interaction and the 2 magnon -1 phonon scattering. Since the second and third process will have special attention in the present
investigation we will first emphasize on these.
The most succesful demonstration of the influence of a resonant magnon-phonon interaction on the thermal conductivity has been reported by Laurence and Petitgrand (1973) for the antiferromagnetic compound
FeC12• The measured temperature dependance and a fit using a
theoreti-cal model for resonant scattering are displayed in Fig. 2.15. The
fact that the anomalous minimum in À (T) at about 1 7 K is reproduced
very well, is rather convincing evidence for the existence of the resonant process. Later reports on the temper at ure dependence in applied magnetic fields (Petitgrand and Laurence 1975) seem to corroborate this interpretation.
Other interpretations based on the resonant process were made on GdCl
3 (Rives et al. 1969) and CoC12.Ga2o (Rives and Bhatia 1974).
These" interpretations, however, are rather questionable as we shall see later on.
The first attempts to incorporate the 2 magnon ,... 1 phonon processes in
the interpretation of thermal conductivity measurements originate
from Nettleton (1964). He tried to explain À(T) in MnF
the theory developed by Upadhyahya and Sinha (1963) for 2 magnon -1 phonon scattering but the agreement was very poor. In -1975 Ono was able to give a satisfactory qualitative explanation of results on GdV0
4 obtained by Metcalfe and Rosenberg (1972) with the aid of a morè detailed theoretical description of the process. The most outstanding reports on this kind of scattering we re gi ven by Dixon and co-workers on the compound MnC1
2,4H2
o.
Measurements of À and a q\.talitative explanation had already been published several times{Ri.ves and Walton 1968, Ri.ves et. al. 1%9, Ri.ves et al. 1970, Ri.ves f971). Only recently Dixon et al, (1980} succeeded in giving a
remarkably good qualitative and quantitative explanation from as well the temperature
as
the field dependence. In the preceding article .the theoretical background was presented in complete detail (Dixón 1980)In the figures 2.16 and 2.17 we reproduce some of their results. Since
MnC12.4H2o is a three-dimensional antiferromagnetic compound the spinwaves disappear above TN, resulting in a negligible field effect. At zero field the magnons cause a reduction of the phonon thermal conductivity of at most 30% while .application of a field results in a much larger scattering at the spinflop transition (Fig. 2.17). The
• 0 k()e
6 40 k()e
Q ; Q S Q L - - - ' - - - ' - - - ' 2 - - - ' J
T <Kl
Fig.2.16. Temperatui>e dependenae 1of the thermal aonduativity
of MnCZ2.4H2
o
divided by T3.Curve A is a fit to the
40 kOe data.Curve B ie a model aalaulation inaluiling
1