Вы находитесь на странице: 1из 4

LETTERS

Large enhancement of the thermopower in


NaxCoO2 at high Na doping
MINHYEA LEE1 , LILIANA VICIU2 , LU LI1 , YAYU WANG1 *, M. L. FOO2 , S. WATAUCHI2 ,
R. A. PASCAL JR2 , R. J. CAVA2 AND N. P. ONG1
1

Department of Physics, Princeton University, Princeton, New Jersey 08544, USA


Department of Chemistry, Princeton University, Princeton, New Jersey 08544, USA
*Present address: Department of Physics, University of California, Berkeley, California 94720, USA

Present address: Department of Chemistry, University of Toronto, Toronto, Ontario M5S 3H6, Canada

Present address: Division of Synthesis of Superconducting Materials, Yamanashi University, Yamanashi 400-8511, Japan

e-mail: npo@princeton.edu
2

Published online: 11 June 2006; doi:10.1038/nmat1669

esearch on the oxide perovskites has uncovered electronic


properties that are strikingly enhanced compared with
those in conventional metals. Examples are the high
critical temperatures of the cuprate superconductors and the
colossal magnetoresistance in the manganites. The conducting
layered cobaltate Nax CoO2 exhibits several interesting electronic
phases as the Na content x is varied14 , including water-induced
superconductivity4 and an insulating state3 that is destroyed
by eld5 . Initial measurements1 showed that, in the as-grown
composition, Nax CoO2 has moderately large thermopower S
and conductivity . However, the prospects for thermoelectric
cooling applications faded when the gure of merit Z was found
to be small at this composition (0.6 < x < 0.7). Here we report
that, in the poorly explored high-doping region x > 0.75, S
undergoes an even steeper enhancement. At the critical doping
xp 0.85, Z (at 80 K) reaches values 40 times larger than in
the as-grown crystals. We discuss prospects for low-temperature
thermoelectric applications.
In the large-x region of Nax CoO2 (x > 0.75), progress has been
hampered by diculties in growing single crystals as well as by
phase-separation eects which appear above a threshold doping
known as xp . Powder neutron diraction6 has revealed subtle shifts
in the Na ions as x is increased above 0.75. The doping interval
0.75 < x < xp is a homogeneous phase H2 , in which the thickness
t of the CoO2 layers undergoes a steep increase6 . Above xp , the
neutron results indicate phase separation (the mixed-phase region
is labelled H2 + H3 ). Muon spin rotation7,8 , nuclear magnetic
resonance9 (NMR) and susceptibility9,10 experiments also suggest
mixed phases in highly doped samples. Finally, the limiting phase
H3 (x 1) is inferred from NMR and magnetic susceptibility to
be a non-magnetic insulator with a slight admixture of a residual
metallic phase9,10 .
In a thermopower experiment, the charge current density
J = E produced by the electric eld E is cancelled by the
thermoelectric current density J = (T ) driven by the

temperature gradient T ( is the thermoelectric or Peltier


conductivity). Whereas the thermopower S = / is the quantity
usually reported, we have found that provides a more incisive
probe for sorting out the results in the region H2 + H3 . Results from
10 crystals (of nominal size 400 700 m) with x ranging from 0.71
to 1 reveal that rises steeply to a peak at the threshold doping
xp , with nearly constant (as shown in Fig. 1, samples are labelled
110 in order of increasing x ). In the mixed region above xp both
quantities decrease rapidly.
Crystals of Nax CoO2 grow as either bilayer or trilayer structures
(two and three CoO2 layers per unit cell, respectively)11 . The
in-plane transport quantities are very similar in the two structures.
However, at large x , the three-layer crystals (samples 2 , 9 and 10 )
tend to lock to the commensurate dopings x = 3/4 and 1, which
provide valuable checks.
Figure 1 shows curves of the in-plane resistivity versus T
(temperature) in the three-layer (Fig. 1a) and two-layer (Fig. 1b)
cobaltate. Beginning with the lowest curves (samples 14 in
Fig. 1b), we note that the proles of (T ) are metallic with the
residual resistivity ratio RRR 20, comparable to the as-grown
crystals (0.6 < x < 0.7). Signicantly, in samples 2 6 (regions H2
and slightly beyond xp ), the room-temperature values (300) lie
in the range 12 m cm. As x crosses xp into H2 + H3 (samples
58), (300) rises steeply to 240 m cm (in 10 ), which far
exceeds the Mott limit. Surprisingly, the curves retain a metallic
prole even as x 1 . This implies that, for x > xp , the carriers
strongly segregate into a fraction of the layers f . The observed
(T ) reects the intrinsic resistivity of these conducting layers
inated by the geometric factor 1/f . Moreover, the persistence
of metallic behaviour to 4 K implies that the layers remain
electrically connected down to low T . Thermal activation of
carriers over low barriers, frequently seen in phase-separated
systems, is not observed.
Curves of the thermopower are shown in Fig. 2a. The curve
with the smallest S (sample 1) is typical of as-grown crystals1,2 .
537

nature materials VOL 5 JULY 2006 www.nature.com/naturematerials

2006 Nature Publishing Group

LETTERS
400

a
10*
102

8
300

9*
S (V K1)

(m cm)

101

9*

200

10*

2*
100

100

10 1

2*
1

100

200

T (K)

100

(m cm)

8
7

200

250

300

150
T (K)

200

250

300

CBT
4

1.5
Z (10 3 K 1)

100

150
T (K)

2.0

101

50

300

10 1

6
1.0

0.5
8
3

1
10 2

0
0

100

200

300

T (K)

Figure 1 In-plane resistivity versus T in log-linear scale. a, In the three-layer


and b, in the two-layer cobaltate. The samples, numbered in order of increasing x,
are: sample 1 (x 0.71), 2 (x 0.75), 3 (x 0.80), 4 (x 0.85), 5 (x 0.88), 6
(x 0.89), 7 (x 0.96), 8 (x 0.97), 9 (x 0.99) and 10 (x 1.0). The
asterisks indicate the three-layer crystals in a. From x = 0.71 to xp 0.85, does
not change signicantly (samples 15). Inside the mixed-phase region H2 + H3 ,
however, increases rapidly (610 ). The persistent metallic prole inside H2 + H3
suggests that conducting layers are embedded in an insulating matrix. Scaling of the
curves of is not as satisfactory as for (see below). Contacts with contact
resistance 210  were attached by lightly abrading the crystal surface and
attaching gold wires with silver paint (Du Pont 4922N). The size of the contact pads
(see the Supplementary Information) leads to a total uncertainty in the absolute
values of of 15%.

Across the H2 region (samples 2 , 3, 4), the ST curves show an


increasingly negative curvature that grows into a broad peak of
magnitude 240 V K1 at 130 K in sample 4. In the mixed-phase
region, the peak value rises further to 300350 V K1 (samples
58), before settling to 200 V K1 as x 1 (samples 9 , 10 ).
These large S values far exceed any reasonable extrapolation of
the Sommerfeld expression S = (kB /e)(kB T /F ) (kB is Boltzmanns
constant, e the electron charge and F the Fermi energy). To match
the observed derivative S/T 3 V K2 at low T , we would
need F to equal 30 K, which is unphysically small. The unusually
large S, coexisting with low resistivities (100  cm in samples
46), challenges our understanding of how strong correlation
enhances the Peltier eect. The role of magnetism in the enhanced
thermopower is strongly implied by the observed2 strong eld

50

100

Figure 2 The in-plane thermopower S and the gure of merit Z in Nax CoO2 . a, In
sample 1, S is very similar to that of Terasaki et al.1 . As x increases into phase H2
(2 4), the prole of S develops an increasing bulge near 130 K that grows rapidly
to peak values of 200250 V K1 . In the mixed-phase region, S further increases
to 300350 V K1 (samples 58), before settling down to 228 V K1 in the limit
x 1 (sample 9 ). A striking pattern is that the ST proles (all hole-type) are
nominally similar in shape in the mixed-phase region. The exception is sample 10
in which f 1/200 (see the Supplementary Information). b, The gure of merit
Z = S 2 /, with measured separately (see the Supplementary Information). As x
increases from 0.71 (sample 1), the peak value of Z increases steeply to
1.8 103 K1 in sample 5. At higher x, the peak value of Z falls rapidly because of
the sharp increase in . At 80 K, Z in sample 5 is 40 times larger than that in
sample 1. The dashed line labelled CBT is Z reported17 for CsBi4 Te6 .

dependence of S at x = 0.68 and the presence of weak magnetic


ordering7,8,12,13 in samples with x > 0.75. Several groups2,14,15 have
used the Heikes formula to explain the large S. The role of magnetic
polarons has also been suggested13,16 .
Figure 2b shows plots of the gure of merit Z = S2 / for some
of the samples (the in-plane thermal conductivity is reported
in the Supplementary Information). Near the threshold xp = 0.85
(samples 4, 5), Z rises to a prominent maximum below 100 K. As
mentioned, the peak value of Z at 80 K is 40 times larger than that
in the as-grown composition (compare sample 5 with sample 1).
We discuss Z below. The dimensionless gure of merit ZT is plotted
in the Supplementary Information.
The Peltier conductivity shows nominally similar proles
(Fig. 3). In samples 1 and 2 , shows broad peaks at 100 and
80 K, respectively, but remains fairly small. However, as we enter the

538

nature materials VOL 5 JULY 2006 www.nature.com/naturematerials

2006 Nature Publishing Group

LETTERS
80

200
150

140

(A mK 1)

100
6
4

100

100

200

80

40

60

300

20
H1

50

0
0.6

2*
1

50

2*

H2

H2 + H3

9*,10*
0.7

0.8
x

xp

0.9

20

0
1.0

9*

0
0

40

T (K)

60

9*
50

(A mK 1)

150

2*

(103 1 cm 1)

p
(A mK 1)

120
100

H3

100

150
T (K)

200

250

300

Figure 3 Curves of the Peltier conductivity versus T and scaling of the T


curves in selected samples. As x traverses the region H2 (samples 14), the peak
in (T ) systematically shifts from 100 to 50 K. Because of the shift these curves
cannot be scaled together. However, in the region H2 + H3 (510), the peak in
remains at 50 K. Inset: the curves in samples 510 match the curve in sample 4
when multiplied by the scale factor p (p = 1.3,3.9,5.4,38 and 200 in samples 5,
7, 8, 9 and 10 , respectively). Sample 6 has a cusp of unknown origin that spoils
the scaling. Sample 10 is discussed in the Supplementary Information.

H2 region increases rapidly, showing a sharply peaked prole in


sample 4. In simple metals, (kB2 /e)(T /F ) decreases linearly
to zero as T 0. At high T , tends to saturate to values in the
range 110 A mK1 . Hence, the pronounced peak and large values
of reported here (max 135 A mK1 in sample 4) are strikingly
anomalous, and are probably a consequence of strong correlation.
By tracking Tmax (the peak temperature of ), we see that
versus T in samples 14 (x xp ) evolves dierently from in 510
(in H2 + H3 ). In the rst group, Tmax decreases from 110 to 50 K
(in samples 1 and 4, respectively). The continuous change in Tmax
precludes a simple scaling analysis. In contrast, in the second group,
Tmax remains xed at 50 K. Moreover, by multiplying each curve
by a scale factor p, we may match it to the curve of sample 4 (Fig. 3,
inset). The exception is sample 6, which has a cusp at 36 K (of
unknown origin) that ruins the scaling. Hence, as we traverse the
homogeneous region H2 , the peak value of increases rapidly from
20 A mK1 in 2 to 135 A mK1 in sample 4, but the form of versus
T also changes continuously. However, once we cross the threshold
xp into the mixed phase, the prole is locked to that of sample 4.
The simplest explanation of the scaling is that, in the doping
region H2 + H3 , the carriers segregate into continuous conducting
layers embedded in a nearly insulating matrix. The conducting
layers correspond to doping xp , whereas the insulating matrix is at
x = 1.0. The mean doping xm is then given by
xm = f xp + (1 f ),

(xp < xm < 1)

(1)

Figure 4 Variation of (50 K) and (300 K) versus x in Nax CoO2 at large x. As x


traverses H2 , (50 K, circles) rises sharply to a maximum at xp in correlation with
the increase in the CoO2 layer thickness6 . Because (300 K, triangles) is nearly
constant, the peak in (50 K ) leads to a peak in Z as well. In H2 + H3 , and
(plotted versus xm calculated from p using equation (1)) decrease linearly as
xm 1, implying phase separation of holes into conducting layers. The phase H3 is
insulating. In the trilayer crystals (samples 2 ,9 and 10 ), x is locked to the
commensurate values 3/4 and 1. The x values in samples 1 and 4 were
measured by X-ray diffraction (lines are guides to the eye). The error bars in xm
reect the uncertainty in measuring and S (see the Supplementary Information).

where f is the fraction of layers with doping xp . Additivity of the


Peltier currents implies that the observed is given by

(T ) = f p (T ) + (1 f )1 (T ),

(2)

where subscripts p and 1 refer to quantities evaluated at x =


xp (0.85) and 1, respectively.
The scaling behaviour in the inset of Fig. 3, implies that the
second term in equation (2) is negligible if f is not too small.
The Peltier conductivity 1 (T ), which has a maximum value of
0.85 A mK1 , is observable only when f becomes very small,
as in the case of sample 10 (see the Supplementary Information).
Away from this limit in the mixed-phase region, we may take
f 1/p; this allows us to nd xm using equation (1). For example,
in sample 9 , the scaling of shows that 1 in 40 of the layers
is conducting (xm = 0.996 instead of 1.00), which is consistent
with the value of (300) in Fig. 1a. In addition, we carried
out other checks on the calibration. The trilayer crystals, with
x = 3/4 and 1, provide valuable calibration points. Moreover,
x in samples 1 and 4 was determined from the c -axis lattice
parameter measured by X-ray diraction. These checks support
our calibration.
With the inferred values of x and xm , we next plot the
variation of at 50 K (circles) and the conductivity (triangles)
at 300 K in the phase diagram at large doping (Fig. 4). From
x = 0.71 to xp = 0.85, (50) rises by a factor of 14. With the
modest changes in (300) (triangles) and (see the Supplementary
Information), this results in a very large increase in Z . We
emphasize that these increases occur inside the homogeneous phase
539

nature materials VOL 5 JULY 2006 www.nature.com/naturematerials

2006 Nature Publishing Group

LETTERS
H2 (rather than in the mixed region above xp = 0.85). Inside
H2 , strong repulsion between the holes, weak magnetic ordering,
and geometric frustration associated with hopping on a triangular
lattice all seem to be crucial for producing the steep increase in .
Above xp , in the mixed phase region, both and decrease linearly
to zero. Finally, at x = 1, the system is insulating as inferred9,10
from NMR.
The 40-fold enhancement of Z realized at 80 K greatly improves
the prospects for thermo-electric applications. The peak value
in Fig. 2b (Z 1.8 103 K1 ) is among the highest known
for a hole-type material below 100 K. The promising material
CeBi4 Te6 also has a large Z at 100 K (ref. 17). In comparison,
Z in Nax CoO2 peaks at low temperatures, whereas the curve for
CeBi4 Te6 (curve labelled CBT) falls steeply. The alloy Bi1y Sby has
long been known18 to have an even larger Z at low T (at 80 K,
Z = 7 103 K1 for y 0.12). However, it is electron-like. To
realize its advantages in a Peltier device operating below 100 K,
we need to pair BiSb with a hole-type material with comparable
Z , but the 35-year search has turned up very few candidates. We
calculate that an optimized device with Nax CoO2 (x = xp ) paired
of merit Z np = 2.5 103 K1 at 80
with Bi1y Sby has a device
gure
2
K (Z np = (S n Sp ) /( n n + p p )2 where subscripts n and p
index the two materials). This value is among the highest reported
to date. Further enhancement of Z in Nax CoO2 seems possible if we
degrade .
METHODS
Sodium metal (0.5 g) and tetrahydrofuran (30 ml) in a pyrex tube were warmed
at 100 C in an oil bath6 . Benzophenone (2 g) was added and the solution was
heated until it turned blue due to the formation of the benzophenone ketyl
radical anion. As-grown crystals of Na0.75 CoO2 were placed along a pipette-like
tube and immersed in the blue solution of the sodium radical anions. The
pyrex tube was quickly capped and heated for four days at 100 C. Finally, the
crystals were extracted and washed with tetrahydrofuran and ethanol to
eliminate trace sodium.

Received 15 February 2006; accepted 9 May 2006; published 11 June 2006.


References
1. Terasaki, I., Sasago, Y. & Uchinokura, K. Large thermoelectric power in NaCo2 O4 single crystals. Phys.
Rev. B 56, R12685R12687 (1997).
2. Wang, Y., Rogado, N. S., Cava, R. J. & Ong, N. P. Spin entropy as the likely source of enhanced
thermopower in Nax Co2 O4 . Nature 423, 425428 (2003).
3. Foo, M. L. et al. Charge ordering, commensurability, and metallicity in the phase diagram of the
layered Nax CoO2 . Phys. Rev. Lett. 92, 247001 (2004).
4. Tanaka, K. et al. Superconductivity in two-dimensional CoO2 layers. Nature 422, 5355 (2003).
5. Balicas, L., Abdel-Jawad, M., Hussey, N. E., Chou, F. C. & Lee, P. A. Shubnikov de haas oscillations
and the magnetic-eld-induced suppression of the charge ordered state in Na0.5 CoO2 . Phys. Rev. Lett.
94, 236402 (2005).
6. Huang, Q. et al. Coupling between electronic and structural degrees of freedom in the triangular
lattice conductor Nax CoO2 . Phys. Rev. B 70, 184110 (2004).
7. Mendels, P. et al. Cascade of bulk magnetic phase transitions in Nax CoO2 as studied by muon spin
rotation. Phys. Rev. Lett. 94, 136403 (2005).
8. Sugiyama, J. et al. Dome-shaped magnetic phase diagram of thermoelectric layered cobaltites. Phys.
Rev. Lett. 92, 017602 (2005).
9. Lang, G. et al. Evidence of a single nonmagnetic Co3+ state in the Na1 CoO2 cobaltate. Phys. Rev. B 72,
094404 (2005).
10. de Vaulx, C. et al. Nonmagnetic insulator state in Na1 CoO2 and phase separation of Na vacancies.
Phys. Rev. Lett. 95, 186405 (2005).
11. Viciu, L. et al. Crystal structure and elementary properties of Nax CoO2 (x = 0.32, 0.5, 0.6, 0.75, and
0.92) in the three-layer Na1 CoO2 family. Preprint at <http://arxiv.org/cond-mat/0601587> (2006).
12. Bayrakci, S. P. et al. Bulk antiferromagnetism in Na0.82 CoO2 single crystals. Phys. Rev. B 69,
100410 (2004).
13. Bayrakci, S. P. et al. Magnetic ordering and spin saves in Na0.82 CoO2 . Phys. Rev. Lett. 94,
157205 (2005).
14. Koshibae, W. & Maekawa, S. Eects of spin and orbital degeneracy on the thermopower of strongly
correlated systems. Phys. Rev. Lett. 87, 236603 (2001).
15. Chaikin, P. M. & Beni, G. Thermopower in the correlated hopping regime. Phys. Rev. B 13,
647651 (1976).
16. Bernhard, C. et al. Charge ordering and magnetopolarons in Na0.82 CoO2 . Phys. Rev. Lett. 93,
167003 (2004).
17. Chung, D.-Y. et al. CsBi4 Te6 : A high-performance thermoelectric material for low-temperature
applications. Science 287, 10241027 (2000).
18. Yim, W. M. & Amith, A. Bi-Sb alloys for magneto-thermoelectric and thermomagnetic cooling.
Solid-State Electron. 15, 11411165 (1972).

Acknowledgements
We acknowledge support from the US Oce of Naval Research (Contract N00014-04-1-0057) and by
the National Science Foundation Grant DMR 0213706.
Correspondence and requests for materials should be addressed to N.P.O.
Supplementary Information accompanies this paper on www.nature.com/naturematerials.

Competing nancial interests


The authors declare that they have no competing nancial interests.
Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

540

nature materials VOL 5 JULY 2006 www.nature.com/naturematerials

2006 Nature Publishing Group

Вам также может понравиться