RAPID COMMUNICATIONS

PHYSICAL REVIEW B VOLUME 62, NUMBER 10 1 SEPTEMBER 2000-II

Thermal conductivity and specific heat of the linear chain cuprate Sr2 CuO3 :
Evidence for thermal transport via spinons
A. V. Sologubenko, E. Felder, K. Giannò, and H. R. Ott
Laboratorium für Festkörperphysik, ETH Hönggerberg, CH-8093 Zürich, Switzerland

A. Vietkine and A. Revcolevschi

Laboratoire de Physico-Chimie des Solides, Université Paris-Sud, 91405 Orsay, France
共Received 5 May 2000兲
We report measurements of the specific heat and the thermal conductivity of the model Heisenberg spin-1/2
chain cuprate Sr2 CuO3 at low temperatures. In addition to a nearly isotropic phonon heat transport, we find a
quasi-one-dimensional excess thermal conductivity along the chain direction, most likely associated with spin
excitations 共spinons兲. The spinon energy current is limited mainly by scattering on defects and phonons.
Analyzing the specific heat data, the intrachain magnetic exchange J/k B is estimated to be ⯝2650 K.

There is a considerable theoretical interest in one- and phonons. We find no evidence for a mutual scattering
dimensional 共1D兲 Heisenberg spin-1/2 systems because they between spin excitations and hence it seems to be absent or
exhibit a number of properties that are entirely dominated by at least negligibly small, in agreement with theoretical pre-
quantum-mechanical behavior and have no analogues in dictions for integrable models.
three-dimensional systems. In particular, it has been shown The specimens used in these experiments were cut from a
that the Heisenberg S⫽1/2 chain represents an integrable single crystal that had been grown by the traveling solvent
system characterized by a macroscopic number of conserva- floating zone method. The details of crystal growth and
tion laws.1 One important conserved quantity is the energy structural characterization are described elsewhere.10 For
current,1,2 implying an ideal 共infinite兲 thermal conductivity thermal transport measurements, three rectangular-bar-
along the chains at nonzero temperatures, if perturbations shaped samples of typical dimensions 2.5⫻1⫻1 mm3 with
from impurities, phonons, or an interchain coupling, which the longest dimension parallel to either the a, b, or c axis
were prepared. Two additional samples, #1 and #2, cut from
always lead to nonintegrable models, are negligible. It is
the same piece, were used for specific heat measurements.
an open question, to what extent a real material may
The thermal conductivity was measured using a conventional
be regarded as an ideal integrable system. Probably, the
steady-state method as described in Ref. 3. A standard relax-
most obvious evidence for the predicted anomalous heat ation technique was employed for the specific heat measure-
transport is the recent observation of an unusually high ments. The magnetic susceptibility ␹ was measured with a
quasi-1D magnon thermal conductivity in the series commercial SQUID magnetometer.
(Sr,Ca,La) 14Cu24O41 . 3,4 The structure of these materials Small amounts of excess oxygen are known to be present
contains two building blocks with 1D character, namely in as-grown crystals of Sr2 CuO3 , giving rise to a Curie-
CuO2 chains and Cu2 O3 ladders, both oriented along the Weiss term in the temperature dependence of the magnetic
same direction. Unfortunately, the dimerization within the susceptibility6 due to uncompensated Cu S⫽1/2 spins. In
chains and a non-negligible interchain interaction in this sys- order to study the influence of excess oxygen, we annealed
tem complicate the analysis of the observed features in terms sample #2 at 870 °C for 72 h under argon atmosphere, as
of an integrable model. described in Ref. 8. From the results of our measurements of
In this work, we have searched for anomalies in the ther- ␹ (T), the ratio of the number of residual spin-1/2 impurities
mal transport of Sr2 CuO3 , which is often considered as the to the total number of Cu ions was estimated to be 1.8
best physical realization of the 1D Heisenberg S⫽1/2 model. ⫻10⫺4 for the unannealed sample #1 and 6⫻10⫺5 for the
The crystal structure of Sr2 CuO3 contains chains formed by annealed sample #2.
CuO4 squares sharing oxygen corners.5 The chains run along The results of the specific heat (C p ) measurements in the
the b axis and, as shown in the inset of Fig. 2, the CuO4 temperature range between 1.5 and 22 K are presented in
squares lie in the ab plane. The intrachain exchange interac- Fig. 1 as a plot of C p /T versus T 2 . The solid lines in Fig. 1
tion between neighboring Cu2⫹ ions connected via 180° are fits to the data above 4 K using the approximation
Cu-O-Cu bonds, measured as J/k B , is between 2150 and
3000 K.6–9 The ratio k B T N /J, where T N is the 3D Néel tem- C p ⫽ ␥ T⫹ ␤ T 3 ⫹ ␦ T 5 . 共1兲
perature, is as small as 2⫻10⫺3 , reflecting an extremely
small ratio J ⬘ /J, J ⬘ representing the interchain interaction. The parameter values are ␥ ⫽2.12⫻10 ⫺3 J mole⫺1
Our observations indicate an excess thermal conductivity K⫺2 , ␤ ⫽1.359⫻10 ⫺4 J mole⫺1 K⫺4 , and ␦ ⫽1.310
along the chain direction, provided by quasi-1D spin excita- ⫻10⫺7 J mole⫺1 K⫺6 for sample #1, and ␥ ⫽2.06
tions 共spinons兲. According to our analysis presented below, ⫻10⫺3 J mole⫺1 K⫺2 , ␤ ⫽1.531⫻10 ⫺4 J mole⫺1K⫺4 , and
its magnitude is limited by scattering of spinons on defects ␦ ⫽1.310⫻10⫺7 J mole⫺1 K⫺6 for sample #2.

0163-1829/2000/62共10兲/6108共4兲/$15.00 PRB 62 R6108 ©2000 The American Physical Society

 RAPID COMMUNICATIONS

PRB 62 THERMAL CONDUCTIVITY AND SPECIFIC HEAT OF . . . R6109

FIG. 1. Specific heat of Sr2 CuO3 as C p /T versus T 2 . The solid FIG. 2. Temperature dependence of the thermal conductivity of
lines represent the fit to Eq. 共1兲. The inset shows the data at low Sr2 CuO3 along the a, b, and c axes. ␬ s is the calculated spinon
temperatures. The value of T N denoted by an arrow is from Ref. 18. thermal conductivity along the b axis. The solid line is the estimated
sum of spinon and phonon thermal conductivities assuming that the
The sum ␤ T 3 ⫹ ␦ T 5 in Eq. 共1兲 is a common low-temperature spinon mean free path is equal to the distance between bond defects
approximation for the lattice specific heat. The fit values of 共see text兲. The schematic crystal structure is shown in the inset.
the parameter ␤ result in values of the Debye temperature of
⌰ D ⫽441⫾10 K and 424⫾10 K for samples #1 and #2,
respectively. Since Sr2 CuO3 is an insulator, the linear in T SrCuO2 , containing similar spin-1/2 chains but assembled
contribution is not due to itinerant electrons but is ascribed to pairwise in arrays of zigzag chains.21 The more likely possi-
spin degrees of freedom. bility that is consistent with our observations is that our
The elementary excitations of a 1D Heisenberg spin-1/2 anomalies reflect the AFM transition reported in Refs. 19
antiferromagnetic system are not S⫽1 magnons but S⫽1/2 and 20, but now shifted to lower temperatures because of a
topological excitations,11 now commonly called ‘‘spinons.’’ smaller number of impurities in the samples. It has been
The applicability of the spinon model has been demonstrated shown that nonmagnetic impurities interrupting the spin-1/2
experimentally for several S⫽1/2 chain systems, including chains enhance staggered spin-spin correlations,22 a common
Sr2 CuO3 .12,13 The corresponding specific heat at TⰆJ/k B is feature of various low-dimensional Heisenberg spin
given by14–16 systems.23 A convincing manifestation of this feature is the
stabilization of the long-range AFM order by replacing Cu2⫹
2Nk B2 with nonmagnetic ions in the spin-ladder system SrCu2 O3
C s⫽ T, 共2兲 共Ref. 24兲 and the spin-Peierls system CuGeO3 共Refs. 25,26兲.
3J
Recently, a field-induced staggered magnetization near im-
where N is the number of magnetic ions in the system. The purities was also observed in Sr2 CuO3 by NMR
fit values of the parameter ␥ give J/k B ⫽2620⫾100 K and measurements.27 Therefore, it seems quite likely that the ex-
2690⫾100 K for samples #1 and #2, respectively. This may perimentally observed Néel temperatures are always en-
be compared with J/k B ⯝2200 K deduced from magnetic hanced via the influence of impurities and exceed the value
susceptibility data8,9 and a somewhat larger value of J/k B that is given by the ratio J ⬘ /J itself. Indeed, for Sr2 CuO3 ,
⫽2850 K, as obtained from the analysis of an optical ab- calculations considering only dipolar interchain coupling6
sorption spectrum.7,17 The cited susceptibility measurements yield a value of T N as low as 0.028 K.
covered a wide temperature range between 5 and 800 K, The results of the thermal conductivity ␬ along the a, b,
whereas the absorption spectrum presented in Ref. 7 has and c axes are presented in Fig. 2. For both directions per-
been recorded at low temperatures 共32 K兲. Hence the dis- pendicular to the chain direction, ␬ (T) shows a peak at
crepancy between the values of J may be ascribed to its T max⬃20 K and a decrease with increasing temperature,
possible decrease with increasing temperature.18 Our rather tending to a T ⫺1 variation above T⭓200 K. This behavior
large low-temperature values of J are compatible with this is typical for phonon thermal transport.28 The thermal con-
suggestion. ductivity along the chain direction, ␬ b , exhibits the same
The inset of Fig. 1 reveals anomalies of the specific heat temperature dependence at T⭐T max but obviously not so at
with onsets below approximately 3.5 K for both samples, higher temperatures. We suggest that this difference is
indicating some sort of phase transition. For the annealed caused by an additional quasi-1D heat transport along the
sample the anomaly is shifted to lower temperatures with chain direction, provided by spin excitations. For insulators
respect to the peak for the as-grown sample. Both anomalies the phonon-phonon scattering mechanism leads to ␬
occur at lower temperatures than the Néel temperatures T N ⬀T ⫺n (n⬃1) at T⭓⌰ D . For layered structures, such as
found by ␮ SR 共Ref. 19兲 (4.15⬍T N ⬍6 K) and neutron Sr2 CuO3 , where the layers are perpendicular to the a axis,
scattering20 (T N ⫽5.4 K) measurements, respectively. One one may expect that in this temperature region the ratio be-
possible explanation for this disagreement is that, besides the tween an in-plane and the out-of-plane phonon conductivity
transition to an antiferromagnetically 共AFM兲 ordered state at is larger than the anisotropy of ␬ (T) along two different
T N ⫽5.4 K, not reflected in C p (T), there is another transi- in-plane directions.29 Since the difference between ␬ c 共in-
tion at T c ⬍3.5 K. Recently, two subsequent magnetic phase plane兲 and ␬ a 共out-of-plane兲 is very small, the difference
transitions at T c1 ⫽5.0 K and T c2 ⫽1.5 K were observed for between ␬ b and ␬ c 共both in-plane兲 should even be smaller.
RAPID COMMUNICATIONS

R6110 A. V. SOLOGUBENKO et al. PRB 62

⫺1
a sum 兺 l s,i of the inverse mean free paths produced by each
scattering mechanism. It turns out that

l s⫺1 ⫽AT exp共 ⫺T * /T 兲 ⫹L ⫺1 , 共4兲

with the parameters A⫽8.2⫻105 m⫺1 K⫺1 , T * ⫽186 K,

and L⫽7.86⫻10⫺8 m, reproduces our results in the whole
temperature region between 50 and 300 K with high accu-
racy, as may be seen in Fig. 3. If the first term on the right-
hand side of Eq. 共4兲 is to represent Umklapp processes,
spinon-spinon scattering may be ruled out because this case
would require T * to be of the order of J/k B ⬃2600 K. Since
T * is close to ⌰ D /2⫽220 K, spinon-phonon Umklapp pro-
FIG. 3. Spinon mean free path for Sr2 CuO3 . The solid line is a cesses are the most likely choice for this contribution.
fit to Eq. 共4兲. The solid squares represent the distance between two The second and constant term in Eq. 共4兲 is attributed to
neighboring bond defects 共from Ref. 30兲 and the dashed line is a scattering of spin excitations on defects interrupting the
polynomial fit to these data. Cu-O chains. In that case the value of the parameter L is a
measure for the mean distance between the defects. Surpris-
This argument is, of course, not valid for the temperature ingly, this distance is much shorter than the average distance
region near and below T max , where the influence of sample of 2⫻104 Å between S⫽1/2 impurities, estimated from the
boundaries and various defects is important and, therefore, Curie-Weiss-type term of the magnetic susceptibility. A
the behavior of ␬ (T) is, to some extent, sample dependent. similar discrepancy has also been encountered in the inter-
Our fitting of ␬ c (T) and ␬ a (T) of Sr2 CuO3 , employing the pretation of NMR measurements on Sr2 CuO3 .27 Recently, in
Debye model of phonon thermal conductivity in a similar order to explain some peculiar features of NMR spectra of
way as described in Ref. 3, has shown that the influence of Sr2 CuO3 , Boucher and Takigawa32 introduced the concept of
boundary scattering on the thermal conductivity is negligible mobile ‘‘bond-defects,’’ which are not related to interstitial
at T⭓50 K. The same calculation indicates that an enhanced excess oxygen. The calculated mean distance between two
phonon contribution to ␬ b below 50 K may mostly be traced neighboring bond defects, consistent with the NMR data be-
back to an enhanced phonon mean free path which is limited tween 20 and 60 K,32 is shown in Fig. 3. We note a reason-
by boundary and defect scattering. Based on all these argu- ably good overlap with our data of l s at T⫽60 K. The
ments presented above, we make the crucial assumption that model of Boucher and Takigawa predicts that at lower tem-
the phonon thermal conductivity at T⭓50 K is almost iso- peratures the interaction between defects is important and the
tropic. number of bond defects decreases with increasing tempera-
In order to single out the heat transport due to spin exci- ture. If our parameter L in Eq. 共4兲 indeed represents the dis-
tations, the phonon thermal conductivity, averaged over the a tance between bond defects, it ought to be temperature de-
and c directions at T⭓50 K, was subtracted from the experi- pendent below 50 K. Unfortunately, a quantitative check of
mental data of ␬ b . The resulting spin part ␬ s is also plotted this conjecture is difficult because of the uncertain subtrac-
in Fig. 2. In a first approximation, valid for any system of tion of the phonon background at T⭐50 K. However, the
quasiparticles, the 1D thermal conductivity is given by the idea that the bond defects are the main source of spinon
simple kinetic expression ␬ s ⫽C s v s l s where C s is the spe- scattering at low temperatures is, at least qualitatively, con-
cific heat, v s is the velocity, and l s is the mean free path of sistent with the temperature dependence and the anisotropy
the spin excitations. The velocity of spinons is11 v s of the thermal conductivity also below 50 K. To demonstrate
⫽Ja ␲ /2ប, where a is the distance between the spins along this, we calculated the total thermal conductivity ( ␬ ph⫹ ␬ s )
the chain direction. Since TⰆJ/k B still holds, Eq. 共2兲 for the along the b axis at temperatures between 20 and 60 K as-
specific heat is valid, and thus the thermal conductivity of suming, first that ␬ ph is isotropic also in this temperature
spinons is given by the simple equation range and equal to the average of ␬ c and ␬ a and, second that
l s is equal to the distance between neighboring bond defects
given in Ref. 32 共the dashed line in Fig. 3兲. The resulting
k B2 ␲
␬ s ⫽N s a l s T, 共3兲 temperature dependence of ␬ b shown by the solid line in Fig.
3ប 2 is in qualitative agreement with the experiment.33
Concluding this paper, we return to the question whether
where N s is the number of spins per unit volume. Equation the present results are relevant vis à vis integrable models.
共3兲 has been shown to also be valid for 1D magnon We argue that a sizeable quasi-1D thermal transport medi-
systems.30 ated by spin excitations does exist in Sr2 CuO3 . Its magnitude
We calculated l s using Eq. 共3兲 and taking into account is not exceptional but the scatterers, i.e., defects and
small 共maximum 6.5% at 300 K兲 deviations31 of C s (T) from phonons, limiting the mean free path of spin excitations are
linearity. The calculated spinon mean free path is shown in extrinsic to the magnetic system. Our analysis indicates the
Fig. 3. It tends to reach a constant value at low temperatures absence or negligibly small influence of spinon-spinon scat-
and decreases with increasing temperature. Assuming that tering on the thermal conductivity, in agreement with the
the different scattering mechanisms act independently, the predictions made for integrable models.1 Our results imply
inverse total mean free path of spinons l s⫺1 may be written as that a dissipationless energy current, expected for systems
 RAPID COMMUNICATIONS

PRB 62 THERMAL CONDUCTIVITY AND SPECIFIC HEAT OF . . . R6111

that fulfill the assumptions of the integrable Heisenberg 1D be considered as an excellent realization of a 1D S⫽1/2
S⫽1/2 model, is not robust if perturbations like defects and Heisenberg antiferromagnet.
lattice excitations interfere. The very high value of the mag- We acknowledge useful discussions with X. Zotos and F.
netic exchange interaction within the chains and the low tem- Naef. This work was financially supported in part by the
perature magnetic phase transition, identified via our specific Schweizerische Nationalfonds zur Förderung der Wissen-
heat measurements, confirm that Sr2 CuO3 may, nevertheless, schaftlichen Forschung.

1 18
X. Zotos, F. Naef, and P. Prelovšek, Phys. Rev. B 55, 11 029 M. Takigawa, O. A. Starykh, A. W. Sandvik, and R. R. P. Singh,
共1997兲. Phys. Rev. B 56, 13 681 共1997兲.
2
Th. Niemeijer and H. A. W. van Vianen, Phys. Lett. 34A, 401 19
A. Keren et al., Phys. Rev. B 48, 12 926 共1993兲.
共1971兲. 20
K. M. Kojima et al., Phys. Rev. Lett. 78, 1787 共1997兲.
3
A. V. Sologubenko, K. Giannò, H. R. Ott, U. Ammerahl, and A. 21
I. A. Zaliznyak et al., Phys. Rev. Lett. 83, 5370 共1999兲.
Revcolevschi, Phys. Rev. Lett. 84, 2714 共2000兲. 22
S. Eggert and I. Affleck, Phys. Rev. B 46, 10 866 共1992兲; Phys.
4
K. Kudo et al., J. Low Temp. Phys. 117, 1689 共1999兲. Rev. Lett. 75, 934 共1995兲.
5
Chr. L. Teske and Hk. Muller-Büschbaum, Z. Anorg. Allg. 23
M. Laukamp et al., Phys. Rev. B 57, 10 755 共1998兲.
Chem. 371, 325 共1969兲. 24
M. Azuma et al., Phys. Rev. B 55, 8658 共1997兲.
6
T. Ami et al., Phys. Rev. B 51, 5994 共1995兲. 25
S. B. Oseroff et al., Phys. Rev. Lett. 74, 1450 共1995兲.
7
H. Suzuura et al., Phys. Rev. Lett. 76, 2579 共1996兲.
8
26
M. Hase et al., Physica B 215, 164 共1995兲.
N. Motoyama, H. Eisaki, and S. Uchida, Phys. Rev. Lett. 76, 27
M. Takigawa, N. Motoyama, H. Eisaki, and S. Uchida, Phys.
3212 共1996兲.
Rev. B 55, 14 129 共1997兲.
9
D. C. Johnston, Acta Phys. Pol. A 91, 181 共1997兲.
10
28
R. Berman, Thermal Conduction in Solids 共Clarendon, Oxford,
A. Revcolevschi, A. Vietkine, and H. Moudden, Physica C 282-
1976兲.
287, 493 共1997兲.
11
L. D. Faddeev and L. A. Takhtajan, Phys. Lett. 85A, 375 共1981兲.
29
S. Y. Ren and J. D. Dow, Phys. Rev. B 25, 3750 共1982兲.
12
O. V. Misochko et al., Phys. Rev. B 53, 14 733 共1996兲.
30
D. L. Huber, Prog. Theor. Phys. 39, 1170 共1968兲.
13
H. Fujisawa et al., Phys. Rev. B 59, 7358 共1999兲.
31
D. C. Johnston et al., Phys. Rev. B 61, 9558 共2000兲.
14
M. Takahashi, Prog. Theor. Phys. 50, 1519 共1973兲.
32
J. P. Boucher and M. Takigawa, Phys. Rev. B 62, 367 共2000兲.
33
15
F. D. M. Haldane, Phys. Rev. Lett. 66, 1529 共1991兲. Note that the possibility of an enhanced low-temperature phonon
16
W. McRae and O. P. Sushkov, Phys. Rev. B 58, 62 共1998兲. mean free path along the b axis is not taken into account in this
17
J. Lorenzana and R. Eder, Phys. Rev. B 55, 3358 共1997兲. calculation.