1: \documentclass[twocolumn]{revtex4}
2: \usepackage[dvips]{graphicx}
3: \begin{document}
4:
5: \title{
6: Impurity-induced transition and impurity-enhanced thermopower
7: in the thermoelectric oxide NaCo$_{2-x}$Cu$_x$O$_4$
8: }
9:
10: \author{I. Terasaki}
11: \email{terra@mn.waseda.ac.jp}
12: \affiliation{
13: Department of Applied Physics, Waseda University, Tokyo 169-8555, Japan\\
14: Precursory Research for Embryonic Science and
15: Technology, Japan Science Technology, Tokyo 108-0075, Japan
16: }
17:
18: \author{I. Tsukada}
19: \affiliation{
20: Central Research Institute for Electric Power Industry,
21: Komae 201-8511, Japan
22: }
23:
24:
25: \author{Y. Iguchi}
26: \affiliation{
27: Department of Applied Physics, Waseda University, Tokyo 169-8555, Japan
28: }
29:
30: \date{\today}
31:
32: \begin{abstract}
33: Various physical quantities are measured and analysed for
34: the Cu-substituted thermoelectric oxide NaCo$_{2-x}$Cu$_x$O$_4$.
35: As was previously known, the substituted Cu enhances the
36: thermoelectric power, while it does not increase the resistivity
37: significantly.
38: The susceptibility and the electron specific-heat are substantially
39: decreased with increasing $x$, which implies that the substituted Cu
40: decreases the effective-mass enhancement.
41: Through a quantitative comparison with the heavy fermion compounds
42: and the valence fluctuation systems, we have found that
43: the Cu substitution effectively increases the coupling between
44: the conduction electron and the magnetic fluctuation.
45: The Cu substitution induces a phase transition at 22 K
46: that is very similar to a spin-density-wave transition.
47: \end{abstract}
48:
49: \pacs{}
50:
51: \maketitle
52:
53: \section{Introduction}
54: Recently layered cobalt oxides have been extensively investigated as a
55: promising candidate for a thermoelectric material.
56: The thermoelectric material is a material that shows large thermopower ($S$),
57: low resistivity ($\rho$) and low thermal conductivity ($\kappa$)
58: \cite{mahan}, and
59: a quantity of $Z \equiv S^2/\rho\kappa$ called the figure of merit
60: characterizes the thermoelectric conversion efficiency.
61: Thermoelectric device can generate electric power from heat through the
62: Seebeck effect, and can pump heat through the Peltier effect.
63: Thus far oxides have been regarded as unsuitable for thermoelectric
64: application because of their poor mobility, but some years ago
65: Terasaki {\it et al.} found that a single crystal of the layered oxide
66: NaCo$_2$O$_4$ exhibits good thermoelectric performance \cite{terra}.
67: Fujita {\it et al.} showed that
68: the dimensionless figure of merit $ZT$ of a NaCo$_2$O$_4$ single crystal
69: exceeds unity at $T=$1000 K \cite{fujita},
70: Ohtaki {\it et al.} \cite{ohtaki} measured $ZT \sim 0.8$ at 1000 K
71: even in the polycrystalline samples of NaCo$_2$O$_4$.
72: Thus this compound is quite promising for
73: thermoelectric power generation at high temperature.
74:
75: Following NaCo$_2$O$_4$, other layered cobalt oxides,
76: Ca-Co-O \cite{li,miyazaki,masset,funahashi},
77: Bi-Sr-Co-O \cite{itoh,itoh2,funahashi2},
78: Tl-Sr-Co-O \cite{hebert}
79: have been found to show good thermoelectric performance.
80: In particular, Funahashi {\it et al}. \cite{funahashi}
81: showed $ZT > 1$ at 1000 K for Ca-Co-O.
82: The most important feature is that
83: the CdI$_2$ type triangular CoO$_2$ block
84: is common to the layered cobalt oxides.
85: We have proposed that the high thermoelectric performance
86: of the layered cobalt oxides cannot be explained
87: by a conventional band picture based on the one-electron approximation,
88: but is understood in terms of the strong electron-electron
89: correlation effects,
90: similarly to the case of heavy-fermion compounds.
91: In fact the material dependence of the thermopower
92: quite resembles that of Ce$M_2X_2$ \cite{terra2}.
93:
94: A prime example for the difficulties of the one-electron picture
95: is observed in the Cu substitution effects in NaCo$_2$O$_4$ \cite{terra3}.
96: The thermopower of NaCo$_{2-x}$Cu$_x$O$_4$ is significantly enhanced,
97: while the resistivity is nearly independent of $x$.
98: This is quite surprising in comparison with
99: normal impurity effects in a metal.
100: The doped impurity acts as a scattering center in usual cases,
101: and does not make a significant change in thermopower,
102: because it is a quantity of the zero-th order of scattering time.
103: Indeed this is what was observed in high-temperature
104: superconductors \cite{tallon}.
105: Importantly, correlation effects can explain
106: the large impurity effect on the thermopower,
107: similarly to the case of dilute magnetic alloys \cite{fischer}.
108: In this paper, we report on measurement of specific heat,
109: susceptibility, Hall coefficient, and transverse magnetoresistance
110: for NaCo$_{2-x}$Cu$_x$O$_4$ polycrystalline samples,
111: and discuss the Cu substitution effects quantitatively.
112:
113:
114: \section{Experimental}
115: Polycrystalline samples of
116: Na$_{1.2}$Co$_{2-x}$Cu$_x$O$_4$
117: ($x$=0, 0.1, 0.2 and 0.3) were prepared
118: through a solid-state reaction.
119: A stoichiometric amount of Na$_2$CO$_3$, Co$_3$O$_4$ and CuO
120: was mixed and calcined at 860$^{\circ}$C for 12 h in air.
121: The product was finely ground, pressed into a pellet,
122: and sintered at 920$^{\circ}$C for 12 h in air.
123: Since Na tends to evaporate during calcination,
124: we added 20 \% excess Na.
125: Namely we expected samples of the nominal
126: composition of Na$_{1.2}$Co$_{2-x}$Cu$_x$O$_4$ to be
127: NaCo$_{2-x}$Cu$_x$O$_4$.
128:
129: The x-ray diffraction (XRD) was measured
130: using a standard diffractometer with Fe K$_{\alpha}$ radiation
131: as an x-ray source in the $\theta -2\theta$ scan mode.
132: The resistivity was measured through a four-terminal method,
133: and the thermopower was measured using a steady-state technique
134: with a typical temperature gradient of 0.5 K/cm.
135: The Hall coefficient ($R_H$) and the transverse magnetoresistance
136: were measured from 15 to 100 K in a closed refrigerator
137: inserted into a room temperature bore of a liquid-He free
138: superconducting magnet.
139: To eliminate the unwanted voltage arising from
140: the misalignment of the voltage pads,
141: the magnetic field was swept from -7 to 7 T
142: with a typical period of 20 min at constant temperatures
143: with a stability of 10 mK.
144: The specific heat was measured using a standard relaxation
145: method with a mechanical heat switch.
146: The mass of the samples used for the measurement
147: is typically 1000 mg and the heat capacity
148: of the samples is always more than two
149: orders of magnitude larger than the addenda heat capacity.
150: The susceptibility was measured with a SQUID susceptometer
151: in a magnetic field of 1 T.
152:
153:
154: \begin{figure}
155: \begin{center}
156: \includegraphics[width=8cm,clip]{f1.eps}
157: \end{center}
158: \caption{The x-ray diffraction patterns of the
159: polycrystalline samples of NaCo$_{2-x}$Cu$_x$O$_4$.
160: The Fe K$_{\alpha}$ is used a an x-ray source.}
161: \label{fig1}
162: \end{figure}
163:
164:
165: \section{Results}
166: Figure 1(a) shows the x-ray diffraction patterns of the prepared samples
167: of NaCo$_{2-x}$Cu$_x$O$_4$.
168: Almost all the peaks are indexed as the $\gamma$ phase
169: \cite{JH,fouassier}, though
170: a small amount (approximately less than 5\%) of
171: unreacted Co$_3$O$_4$ is observed.
172: With increasing Cu content $x$, no additional peak appears,
173: with the patterns unchanged,
174: which shows that Cu is substituted for Co.
175: However, the sample of $x$=0.3 shows a higher background noise,
176: indicating that the crystal quality becomes worse,
177: possibly owing to the limit of solid solution with Cu.
178:
179:
180: Figure 2(a) shows the temperature dependence of the resistivity
181: for NaCo$_{2-x}$Cu$_x$O$_4$.
182: All the samples are metallic down to 4.2 K without any indication
183: of localization.
184: It should be noted that the increased resistance due to
185: the substituted Cu is of the order of 10$\mu\Omega$cm for 1 at.\% Cu,
186: which is anomalously small in the layered
187: transition-metal oxides \cite{fukuzumi}.
188: Another important feature is that the resistivity for the Cu substituted
189: samples show a kink near 22 K as indicated by the dotted line.
190: Since the temperature dependence is steeper below 22 K,
191: the density of states (or the carrier concentration) decreases below 22 K.
192:
193: \begin{figure}[b]
194: \begin{center}
195: \includegraphics[width=8cm,clip]{f2.eps}
196: \end{center}
197: \caption{(a) The resistivity and (b) the thermopower
198: of polycrystalline samples of NaCo$_{2-x}$Cu$_x$O$_4$.}
199: \label{fig2}
200: \end{figure}
201:
202:
203: Figure 2(b) shows the temperature dependence of the thermopower
204: for NaCo$_{2-x}$Cu$_x$O$_4$.
205: Thermopower increases with increasing $x$ with
206: a dip near 22 K and a peak near 10-15 K.
207: As we previously shown in the analysis for Bi-Sr-Co-O \cite{itoh2},
208: a low-temperature thermopower ($S$) of the layered Co oxides
209: is determined by the diffusive term that
210: is proportional to temperature ($T$).
211: Accordingly $S/T$ is an essential parameter
212: similarly to the electron specific-heat coefficient.
213: Thus, the dip, rather than the peak, is a meaningful temperature,
214: which corresponds to the onset of the enhancement
215: in $S/T$ at low temperatures.
216: It should be emphasized that the dip temperature is
217: nearly the same as the kink temperature for the resistivity,
218: which strongly suggests that this temperature is
219: related to a kind of phase transition.
220:
221:
222: \begin{figure}
223: \begin{center}
224: \includegraphics[width=8cm,clip]{f3.eps}
225: \end{center}
226: \caption{(a) The specific heat $C$ and
227: (b) the susceptibility $\chi$
228: of polycrystalline samples of NaCo$_{2-x}$Cu$_x$O$_4$.
229: Note that $C/T$ is plotted in order to emphasize the
230: electron specific heat coefficient $\gamma$.
231: The samples include 5\% of unreacted Co$_3$O$_4$,
232: and the data for 5\% Co$_3$O$_4$ taken from
233: Refs, \cite{LB1,LB2}
234: are plotted by the solid curves.}
235: \label{fig3}
236: \end{figure}
237:
238: Figure 3(a) shows the specific heat $C$ for NaCo$_{2-x}$Cu$_x$O$_4$.
239: In order to emphasize the $T$-linear electron specific heat,
240: we plot $C/T$ as a function of temperature.
241: As shown in Fig. 3(a), the $C/T$ value at 2 K decreases
242: with increasing the Cu content from 0 to 0.2,
243: which means the decrease of the electron specific heat
244: coefficient $\gamma$ with $x$.
245: (For the sample of $x$=0.3, the $C/T$ value increases again,
246: which might be due to an extrinsic origin such as
247: the solid-solution limit of Cu.)
248: Since $\gamma$ is proportional to the density of states
249: and the mass-enhancement factor,
250: the present results indicate that the either or both decrease
251: with the Cu substitution.
252: As for high temperatures, all the data show a peak near 30 K
253: which is the antiferromagnetic transition of the unreacted
254: Co$_3$O$_4$, as shown by the solid curve \cite{LB1}.
255: As mentioned above, the x-ray diffraction patterns reveal
256: less than 5 at.\% of unreacted Co$_3$O$_4$,
257: which is consistent with the peak height of the specific heat at 30 K.
258: We should emphasize here that the existence of Co$_3$O$_4$
259: does not seriously affect the estimation of $\gamma$,
260: because the $C/T$ value for Co$_3$O$_4$ is negligibly small
261: at low temperatures.
262: For $x=$0.2 and 0.3, another peak appears in
263: the specific heat near 22 K,
264: which is close to the kink temperature in $\rho$,
265: and the dip temperature in $S$.
266: We thus conclude that the 22-K anomaly comes from
267: a (2nd order) phase transition.
268:
269:
270:
271: Figure 3(b) shows the susceptibility ($\chi$) of NaCo$_{2-x}$Cu$_x$O$_4$.
272: The substituted Cu also decreases the susceptibility,
273: indicating the decrease of the density of states and/or
274: the mass-enhancement factor.
275: A broad hump near 30 K is due to the antiferromagnetic
276: transition of the unreacted Co$_3$O$_4$,
277: as shown by the solid curve \cite{LB2}.
278: Interestingly, there is no anomaly near 22 K in the susceptibility,
279: suggesting that the transition at 22 K
280: is not the magnetic transition of impurity phases.
281: We further note that the Curie-like contribution
282: is absent in the susceptibility at low temperatures,
283: which shows that magnetic impurities
284: are unlikely to exist other than Co$_3$O$_4$.
285: Quantitatively, the decrease of $\chi$ by Cu
286: is more moderate than that of $C/T$.
287: $C/T$ decreases by a factor of ten from $x=0$ to 0.2,
288: whereas $\chi$ decreases only by a factor of two.
289: This implies that the 22-K transition causes a dramatic reduction
290: of the electron entropy possibly owing to a (pseudo)gap formation,
291: while it does not alter the magnetic excitation at $k=0$.
292: The nature of the 22-K transition will be discussed in the next section.
293:
294: \begin{figure}
295: \begin{center}
296: \includegraphics[width=8cm,clip]{f4.eps}
297: \end{center}
298: \caption{(a) The Hall coefficient $R_H$ and
299: (b) the magnetoresistance $\Delta\rho/\rho$
300: of polycrystalline samples of NaCo$_{2-x}$Cu$_x$O$_4$.}
301: \label{fig4}
302: \end{figure}
303:
304: Figure 4(a) shows the Hall coefficient ($R_H$) of NaCo$_{2-x}$Cu$_x$O$_4$.
305: The sign is negative below 100 K, and the magnitude is
306: as small as 4-6$\times$10$^{-4}$ cm$^3$/C.
307: The Cu substitution does not change the magnitude so much,
308: indicating that the carrier concentration is
309: nearly unchanged.
310: By contrast, it changes the temperature dependence
311: in a complicated way, which implies
312: that plural kinds of carriers are
313: responsible for the electric conduction.
314: The band calculation by Singh \cite{singh}
315: reveals that the two bands of different symmetries cross
316: the Fermi level for NaCo$_2$O$_4$.
317: (See the next section)
318:
319:
320: Contrary to the Hall effect,
321: the magnetoresistance is weakly dependent on the Cu substitution,
322: as shown in Fig. 4(b).
323: By taking a closer look at the $x$ dependence,
324: one can see that the negative magnetoresistance is
325: gradually suppressed by the substituted Cu.
326: This implies that the kink of the resistivity is more or less
327: smeared against magnetic field, which suggests that
328: the magnetic field suppresses the 22-K transition.
329:
330:
331:
332: \section{Discussion}
333: Before going into details,
334: we will begin with a brief review on
335: the electronic states of NaCo$_2$O$_4$.
336: As is well known, the five-fold degenerate $d$ orbitals
337: split into two-fold ($e_g$) and three-fold
338: ($t_{2g}$) degenerate levels in an oxygen octahedron.
339: In the real triangular CoO$_2$ block, the octahedron is compressed
340: along the $c$ direction, and the degeneracy is further quenched,
341: where the lower $t_{2g}$ levels split into $e_g$ and $a_{1g}$ levels.
342: The lower $e_g$ levels correspond to the orbital spread along the
343: CoO$_2$ block to make a relatively broad band,
344: whilst the $a_{1g}$ orbital is spread along the $c$ direction
345: to make a nearly localized narrow band.
346: Since each Co ion is 3.5+ with $(3d)^{5.5}$,
347: the highest occupied orbital is $a_{1g}$ in the atomic limit,
348: and the main part of the Fermi surface consists of
349: the narrow $a_{1g}$ band.
350: In the real band calculation, however,
351: there is significant hybridization
352: between the $a_{1g}$ and $e_g$ levels, and
353: the broader $e_g+a_{1g}$ band touches the Fermi level
354: to make small Fermi surfaces \cite{singh}.
355:
356:
357: We have proposed that the electronic structure of NaCo$_2$O$_4$
358: is similar to that of the Ce-based intermetallics,
359: a prime example of valence-fluctuation/heavy-fermion
360: compounds \cite{terra2}.
361: $\gamma$ and $\chi$ of NaCo$_2$O$_4$ are as large as those of CePd$_3$.
362: In this context, the large thermopower of NaCo$_2$O$_4$
363: is explained in terms of a diffusive contribution of a metal
364: with a heavily enhanced effective mass,
365: and is roughly proportional to $\gamma$.
366: In the heavy fermion compounds, the broad conduction band
367: crosses the Fermi energy, and the narrow (localized) f band
368: is located below the Fermi energy.
369: For NaCo$_2$O$_4$, the broad $e_g+a_{1g}$ band and
370: the narrow $a_{1g}$ band do exist,
371: but both of which cross the Fermi energy
372: to form two kinds of the Fermi surfaces.
373: Thus it is not trivial whether or not the two Fermi surfaces
374: behave heavy-fermion-like in the charge transport.
375: At least we can say that the two Fermi surfaces play different
376: roles, where the Cu substitution induces different effects:
377: $\rho$ is weakly dependent on the Cu content $x$,
378: whereas $S$, $\chi$ and $\gamma$ are strongly dependent on $x$.
379: $S$, $\chi$ and $\gamma$
380: are basically proportional to the density of states in the lowest order,
381: which are determined by the large Fermi surface of the $a_{1g}$ symmetry.
382: In contrast, the carriers on the $e_g+a_{1g}$ band can be highly mobile,
383: because it is spread along the in-plane direction.
384: In short, the $a_{1g}$ and $e_g+a_{1g}$ bands
385: are responsible for the large thermopower
386: and a good electric conduction, respectively.
387:
388: Existence of the $a_{1g}$ and $e_g+a_{1g}$ bands was
389: suggested from the angular dependence of the X-ray absorption
390: spectroscopy experiment \cite{mizokawa, mizokawa2},
391: where the valence bands of NaCo$_2$O$_4$ consist of
392: the $a_{1g}$ and $e_g+a_{1g}$ bands.
393: The valence band of the less conductive Bi-Sr-Co-O
394: is mainly composed only of the $a_{1g}$ band,
395: which is consistent with our speculation that the $e_g+a_{1g}$
396: band is responsible for the metallic nature of NaCo$_2$O$_4$.
397: The large Fermi surface suggested by the band calculation
398: was not seen in the angle photoemission
399: spectra for Bi-Sr-Co-O, which indicates that
400: the band calculation should be modified by additional
401: effects such as the electron-electron or electron-phonon effects.
402:
403:
404: \begin{figure}
405: \begin{center}
406: \includegraphics[width=8cm,clip]{f5.eps}
407: \end{center}
408: \caption{Phase transition at 22 K for the Cu-substituted sample
409: ($x$=0.2). (a) Specific heat, (b) magneto-specific heat
410: $\Delta C(H) \equiv C(H) -C(0)$, and (c) temperature derivative
411: of the resistivity $d\rho/dT$ and the T-linear coefficient of
412: the thermopower $S/T$. The inset shows the magnetic field dependence
413: of the specific heat.}
414: \label{fig5}
415: \end{figure}
416: \subsection{Phase Transition at 22 K}
417: As shown in the previous section,
418: the Cu substitution causes the phase transition at 22 K,
419: which is probed by the jump of the specific heat,
420: the dip in the thermopower, and the kink in the resistivity.
421: Figure \ref{fig5}(a) shows the specific heat for the $x=0.2$ sample
422: (the same data in Fig. \ref{fig3}(a)) as a function of temperature
423: in linear scale in order to see the 22-K anomaly clearly.
424: One thing to point out is that the entropy change
425: of this transition is surprisingly small.
426: As shown in Fig. \ref{fig5}(a), we estimated the entropy change by
427: the area surrounded with $C/T$ and the dotted line,
428: which is approximately 77 mJ/Kmol,
429: corresponding to 0.01$k_B$ per Co.
430: Actually only 5\% of Co$_3$O$_4$ impurity
431: exhibits a specific heat jump of the same order at 30 K.
432:
433: There are two possibilities for the origin of the small entropy change.
434: One is that the 22-K transition is something related to the
435: impurity phase of the order of 1\%.
436: Although we cannot exclude this possibility completely,
437: we will take the other possibility that
438: the small entropy change is an intrinsic nature in bulk,
439: because (i) the field dependence of $C/T$
440: is different between the 22-K transition and
441: the magnetic transition in Co$_3$O$_4$ at 30 K as shown in Fig. \ref{fig5}(b),
442: (ii) a possible impurity phase is a Cu-based magnetic material,
443: which is inconsistent with no anomaly in $\chi$ at 22 K,
444: and (iii) the thermopower and the resistivity systematically
445: change at the same temperature.
446: The most familiar phase transition accompanied by a small entropy change
447: is perhaps superconducting transition.
448: More generally, off-diagonal long range order
449: induces a small entropy change of the order of $Nk_BT/E_F$.
450:
451:
452: Figure \ref{fig5}(c) shows the $T$-linear term of the thermopower ($S/T$)
453: and the temperature derivative of the resistivity $d\rho/d T$,
454: both of which are inversely proportional
455: to the Drude weight \cite{comment}.
456: Their temperature dependences are quite similar to each other,
457: where the magnitude increases up to almost twice below 22 K.
458: This indicates that the Drude weight decreases by 50\% at low
459: temperatures, implying the existence of a (pseudo)gap.
460: As an off-diagonal long-range order with a gapped state,
461: one would think of charge density wave (CDW) or spin density wave (SDW).
462: The calculated Fermi surface \cite{singh}
463: of the $a_{1g}$ band is hexagon-like,
464: which is unstable against CDW or SDW formation with the nesting vector
465: along the $\Gamma$-K direction.
466: We think that the 22-K transition is SDW-like,
467: because CDW is insensitive to magnetic field \cite{coleman}.
468: Actually we can find many similarities between the 22-K transition and SDW
469: transition:
470: The entropy change is observed to be quite small
471: in Cr \cite{CrReview}, YbBiPt \cite{fisk}
472: and (TMTSF)$_2$PF$_6$ \cite{coroneus}.
473: The resistivity shows a hump \cite{CrReview,fisk}, and
474: the thermopower shows a dip at the transition\cite{choi}.
475:
476:
477: It is not surprising that the 22-K transition has little effects on
478: the magnetic susceptibility.
479: Since an SDW state is an antiferromagnetically ordered state,
480: the magnetic excitation is gapless in principle.
481: In fact, the SDW state of Cr exhibits a very tiny (1-2\%) change
482: in the susceptibility at the transition temperature \cite{CrReview},
483: while it causes a clear hump in the resistivity \cite{adachi}.
484: The metallic conduction below 22 K, implies that a part
485: (approximately 50\%) of the Fermi surface remains,
486: which could smears the SDW transition.
487: To clarify the nature of the transition,
488: a local magnetic probe such as NMR or $\mu$SR should be employed.
489:
490:
491: \begin{figure}
492: \begin{center}
493: \includegraphics[width=8cm,clip]{f6.eps}
494: \end{center}
495: \caption{(a) Peltier conductivity $\sigma_P = S/\rho$
496: (b) Hall conductivity $\sigma_{xy} = H R_H/\rho^2$ for
497: the $x=0$ and $x=0.2$ samples
498: (c) the relative change in $\sigma_P$ and $\sigma_{xy}$
499: from $x=0$ to 0.2}
500: \label{fig6}
501: \end{figure}
502: \subsection{Effects on the Hall coefficient and thermopower}
503: Next we will discuss how we can understand the Cu substitution
504: effects on $R_H$ and $S$.
505: We should note here that the sum rules of transport parameters
506: for a multiband system are expressed in the form of conductivities,
507: not in the form of $R_H$ or $S$.
508: Let us denote the conductivities for the $a_{1g}$
509: and $e_g+a_{1g}$ bands as $\sigma^a$ and $\sigma^e$,
510: respectively.
511: The total conductivity $\sigma$ is then written as
512: \begin{equation}
513: \sigma = \sigma^e + \sigma^a
514: \end{equation}
515: Similarly, the total Hall conductivity $\sigma_{xy}$
516: and the total Peltier conductivity $\sigma_P$
517: are written as
518: \begin{eqnarray}
519: \sigma_{xy} &=& \sigma_{xy}^e + \sigma^a_{xy}\\
520: \sigma_{P} &=& \sigma_{P}^e + \sigma^a_{P}
521: \end{eqnarray}
522: where the Peltier conductivity \cite{ong}
523: is defined as
524: $\sigma_P \equiv S \sigma =S/\rho$.
525:
526:
527: Figure \ref{fig6}(a) shows the temperature dependence of
528: $\sigma_P = S/\rho$ for $x=$0 and 0.2.
529: The Cu substitution enhances the Peltier conductivity
530: over the measured temperature range from 4 to 100 K,
531: indicating that the mobility is enhanced by Cu.
532: The enhancement below 22 K is more remarkable
533: in $\sigma_P$ than in $S$,
534: which indicates the mobility is rapidly enhanced
535: below the 22-K transition.
536: Figure \ref{fig6}(b) shows the temperature dependence of
537: $\sigma_{xy} = HR_H/\rho^2$ for $x=$0 and 0.2.
538: The complicated change seen in $R_H$ is converted
539: into a simpler change in $\sigma_{xy}$.
540: Although the Cu substitution effects in $\rho$ is quite small,
541: $1/\rho^2$ term moderates the difference in $R_H$.
542: One can see that $\sigma_{xy}$ is also increased by Cu
543: over the temperature range from 15 to 100 K,
544: as is similar to the case of $\sigma_{P}$.
545:
546: Let us assume that the substituted Cu affects only
547: the $a_{1g}$ band.
548: Then a difference between $x=0$ and 0.2
549: is reduced to a change in $\sigma_P^a$ and $\sigma_{xy}^a$.
550: Figure \ref{fig6}(c) shows
551: $\Delta\sigma_P=\sigma_P(x=0.2)-\sigma_P(x=0)$
552: $\Delta\sigma_{xy}=\sigma_{xy}(x=0.2)-\sigma_{xy}(x=0)$.
553: Most unexpectedly,
554: the change in the Peltier conductivity $\Delta\sigma_P$
555: and the change in the Hall conductivity $\Delta\sigma_{xy}$
556: show nearly the same temperature dependence.
557: In particular, a clear enhancement below 22 K
558: indicates that the phase transition causes an equal
559: impact on $S$ and $R_H$ in the form of
560: the Peltier and Hall conductivities.
561:
562:
563: On the assumption that only the $a_{1g}$ band
564: is modified by Cu, we will consider
565: the change in the $a_{1g}$ band in terms of
566: the carrier concentration $n$, the effective mass $m$,
567: and the scattering time $\tau$.
568: Then $\sigma_P^a$ and $\sigma_{xy}^a$ are roughly expressed as
569: $\sigma_P^a\sim \langle n/m^*\rangle$
570: and $\sigma_{xy}^a\sim \langle \tau/m^*\rangle$, where
571: the average of $\langle\cdots\rangle$
572: is defined as $(4\pi^3)^{-1}\int (v_F)^2\tau \cdots d^3k$.
573: A close similarity between $\Delta\sigma_P$ and
574: $\Delta\sigma_{xy}$ implies that
575: $\Delta\langle \tau \rangle$ and $\Delta\langle n \rangle$
576: are nearly independent of temperature.
577: The $T$-independent $\Delta\langle \tau \rangle$ means
578: the scattering time averaged in the $a_{1g}$ Fermi surface
579: is dominated by impurity scattering,
580: which is consistent with the localized picture
581: of the $a_{1g}$ band.
582: The positive values of $\Delta\sigma_{xy}$ and
583: $\Delta\sigma_P$ suggests that the increase of
584: $\langle 1/m^* \rangle$.
585: This indicates that the mass enhancement
586: is suppressed (the mobility is enhanced) by Cu
587: over the measured temperature range,
588: regardless of the 22-K transition,
589: which is consistent with the decrease in
590: $\gamma$ and $\chi$ by Cu.
591:
592:
593: \begin{figure}
594: \begin{center}
595: \includegraphics[width=8cm,clip]{f7.eps}
596: \end{center}
597: \caption{ (a) Cu dependence of $\chi$, $\gamma$ and S
598: of NaCo$_{2-x}$Cu$_x$O$_4$.
599: (b) Sn dependence of $\chi$, $\gamma$ and S of
600: Ce(Pb$_{1-y}$Sn$_y$)$_3$. }
601: \label{fig7}
602: \end{figure}
603: \subsection{Comparison with Ce based compounds}
604: Based on the heavy-fermion scenario,
605: it seems inconsistent that the substituted Cu causes
606: the decrease in $\gamma$ (Fig. \ref{fig3}) together
607: with the increase in $S$ (Fig. \ref{fig2}).
608: As shown in Fig. \ref{fig7}(a), $\gamma$, $\chi$ and $S$ are plotted as
609: a function of the Cu content $x$.
610: (Note that $\gamma$ was estimated as $C/T$ at 2 K,
611: and $\chi$ was estimated as the raw value of $\chi$ at 2 K.)
612: Although $\gamma$ and $\chi$ decreases with $x$,
613: $S$ significantly increases with $x$,
614: where $S \propto \gamma$ is no longer valid.
615:
616: We should emphasize that the relationship between
617: $\gamma$ and $S$ is complicated also in the Ce-based compounds.
618: Figure \ref{fig7}(b) shows $\gamma$, $\chi$ \cite{lin}
619: and $S$ \cite{sakurai} for Ce(Pb$_{1-y}$Sn$_y$)$_3$
620: as a function of the Sn content $y$.
621: CePb$_3$ is in the heavy fermion regime (low Kondo temperature)
622: and CeSn$_3$ is in the valence fluctuation regime (high Kondo temperature).
623: Thus the solid solution between Pb and Sn changes the material
624: from the heavy-fermion to the valence-fluctuation compound,
625: which is evidenced by the fact that $\chi$ and $\gamma$ monotonically
626: decrease with $y$.
627: On the other hand, $S$ exhibits complicated $y$ dependence.
628: $S$ is negative for $y=0$, increases with $y$ up to 0.4,
629: and eventually decreases from $y$=0.6 to 1.0.
630:
631: This is intuitively understood as follows.
632: When the Kondo temperature is sufficiently low as in the case of CePb$_3$,
633: the RKKY interaction survives at low temperatures, and
634: often causes a magnetic transition.
635: Since the magnetic transition release the entropy of the spin sector,
636: the entropy per carrier (equivalently the diffusive term
637: of the thermopower) would be suppressed
638: against the fluctuation of the magnetic transition.
639: On the other hand, when the Kondo temperature
640: is high enough, the mass enhancement is severely suppressed
641: to give a small thermopower again.
642: Thus the thermopower would take a maximum at an intermediate
643: value of the Kondo temperature.
644: This is indeed what we see in Fig. 7(b),
645: and similar to the general trend of the 20-K
646: thermopower of Ce$M_2X_2$ found
647: by Jaccard {\it et al.}\cite{jaccard}
648:
649:
650: In this context, the NaCo$_2$O$_4$ is located
651: near the heavy-fermion regime,
652: and the substituted Cu caused the decrease in the mass enhancement
653: accompanied by the increase in $S$,
654: which is consistent with the increase in the Peltier and Hall
655: conductivities seen in the previous subsection.
656: Although there is no microscopic relationship between
657: NaCo$_{2-x}$Cu$_x$O$_4$ and Ce(Pb$_{1-y}$Sn$_y$)$_3$,
658: a close resemblance in Fig. 7 suggests that the unsubstituted NaCo$_2$O$_4$
659: corresponds to $y\sim0.2$,
660: while NaCo$_{1.8}$Cu$_{0.2}$O$_4$
661: corresponds to $y\sim 0.4-0.6$.
662: We further note that the Pd substituted NaCo$_2$O$_4$
663: shows {\it negative} thermopower below 50 K,
664: which might correspond to $y< 0.2$ \cite{kitawaki}.
665:
666:
667: \subsection{Order from disorder}
668: Although the microscopic theory for the high thermoelectric performance
669: of NaCo$_2$O$_4$ is still lacking, the following features are established.
670: (1) The mixture of Co$^{3+}$ and Co$^{4+}$ in the low spin state
671: can carry a large entropy of $k_B$log6 \cite{koshibae}.
672: (2) NaCo$_2$O$_4$ shows no structural, electric, and magnetic
673: transitions from 2 to 1000 K \cite{ohtaki,fujita}.
674: (3) From (1)(2), the large entropy cannot be released
675: through phase transitions, and inevitably adhere to
676: the conducting carriers to form a ``heavy-fermion''-like electron.
677:
678: In this respect, NaCo$_2$O$_4$ is very close
679: to the instability for various phase transitions arising
680: from the large entropy per site.
681: The Cu substitution enhances the instability,
682: and eventually causes the SDW-like transition at 22 K.
683: This type of transition is called ``order from disorder''
684: \cite{tsvelik},
685: which has been extensively investigated experimentally
686: as well as theoretically.
687: In other words, instabilities against various phases are competing
688: or disordering in NaCo$_2$O$_4$,
689: and any phase transitions are prohibited down to low temperatures.
690: This does not mean that NaCo$_2$O$_4$ is far from
691: the instability of phase transitions, but rather,
692: is very susceptible to various transitions against various perturbations.
693: In fact, Na$_{1.5}$Co$_2$O$_4$ exhibits a glassy behavior at 3 K
694: due to structure instability of the $\gamma$ phase \cite{takeuchi},
695: and (Bi,Pb)-Sr-Co-O shows a ferromagnetic transition at 4 K
696: due to the lattice misfit \cite{tsukada}.
697:
698: Among various possible transitions, it is not trivial
699: whether an SDW-like state is favored by impurities or not.
700: As an SDW-formation mechanism, we should note here that
701: SDW and CDW are closely related to the nesting of the
702: and the topology of the Fermi surface.
703: They are a property for a metal,
704: and occur when the correlation effect is weak enough
705: to hold one-electron picture based on the band calculation.
706: As often mentioned in the present paper,
707: the experimental results consistently
708: suggest that Cu suppresses the mass enhancement
709: without significant change in the carrier concentration.
710: If so, the decrease in $\chi$ implies that
711: the substituted Cu enhances the screening of
712: the magnetic fluctuation,
713: which might recover the band picture to cause
714: the CDW/SDW instability of the $a_{1g}$ Fermi surface.
715:
716:
717: \section{Summary and Future issues}
718: In this article, we have discussed the Cu substitution effects on the
719: thermoelectric and thermodynamic properties of NaCo$_{2-x}$Cu$_x$O$_4$.
720: The substituted Cu induces a phase transition at 22 K, which is
721: characterized by the kink in the resistivity, the hump in the
722: thermopower, and the jump in the specific heat.
723: We have analysed the nature of the transition, and finally proposed
724: a spin-density-wave-like state as a possible origin,
725: because (i) it accompanies a small entropy change of the order of
726: 10$^{-2} k_B$ per Co, (ii) the transition is sensitive to the
727: magnetic field, (iii) the large Fermi surface of the $a_{1g}$
728: character is gapped.
729: The impurity induced transition is often called ``order from disorder'',
730: which implies that phase transitions are somehow suppressed in
731: the unsubstituted NaCo$_2$O$_4$.
732:
733:
734: Above the transition temperature, the thermoelectric properties
735: are at least qualitatively compared with those of heavy-fermions
736: valence-fluctuation compounds,
737: where mobile holes on the $e_g+a_{1g}$ band and
738: the nearly localized holes of the $a_{1g}$ band
739: correspond to the carrier and the $f$ electrons.
740: In this analogy,
741: the substituted Cu increase the interaction between
742: the $e_g+a_{1g}$ and $a_{1g}$ bands
743: to decrease the effective-mass enhancement.
744:
745: In this article we have reviewed the phenomenology
746: of the Cu substitution effects, but failed to address the
747: microscopic origin and/or the electronic states of
748: the substituted Cu.
749: This is because our experiments were concerned only with the thermodynamic and
750: transport properties of bulk materials.
751: Nonetheless we can say that the scattering cross section will be small for
752: the $d_{x^2-y^2}$ and $d_{z^2}$ levels of the impurity in NaCo$_2$O$_4$,
753: because the valence bands of NaCo$_2$O$_4$ consist of $t_{2g}$.
754: Thus the substituted Cu (possibly divalent \cite{anno})
755: will not increase $\rho$ seriously,
756: because Cu$^{2+}$ has the highest occupied orbital of $d_{x^2-y^2}$
757: that is orthogonal to $t_{2g}$.
758: In addition, strong Jahn-Teller effects of Cu$^{2+}$ may cause local distortion
759: of the CoO$_2$ block, which serves as a kind of chemical pressure
760: to increase $S$ \cite{terra2}.
761: To proceed further, site-selective probes such as NMR,
762: photoemission, and STM/STS should be employed.
763:
764:
765: \section*{Acknowledgments}
766: The authors would like to thank T. Motohashi, H. Yamauchi, N. Murayama,
767: K. Koumoto, and T. Mizokawa for fruitful discussion.
768:
769:
770: \begin{thebibliography}{99}
771: \bibitem{mahan}
772: G. D. Mahan,
773: Solid State Physics 51, 81 (1998).
774: \bibitem{terra}
775: I. Terasaki, Y. Sasago, and K. Uchinokura,
776: Phys. Rev. B56, R12685 (1997)
777: \bibitem{fujita}
778: K. Fujita, T. Mochida, and K. Nakamura,
779: Jpn. J. Appl. Phys. 40, 4644 (2001).
780: \bibitem{ohtaki}
781: M. Ohtaki, Y. Nojiri, and E. Maeda,
782: Proc. 19th Int. Conf. Thermoelectrics (ICT2000),
783: ed. D. M. Rowe, p. 190 (Babrow, Wales, 2000).
784: \bibitem{li}
785: S. Li, R. Funahashi, I. Matsubara, K. Ueno, and H. Yamada,
786: J. Mater. Chem. 9, 1659 (1999).
787: \bibitem{miyazaki}
788: Y. Miyazaki, K. Kudo, M. Akoshima, Y. Ono, Y. Koike, and
789: T. Kajitani,
790: Jpn. J. Appl. Phys. 39, L531 (2000).
791: \bibitem{masset}
792: A. C. Masset, C. Michel, A. Maignan, M. Hervieu, O. Toulemonde,
793: F. Studer, and B. Raveau,
794: Phys. Rev. B62, 166 (2000).
795: \bibitem{funahashi}
796: R. Funahashi, I. Matsubara, H. Ikuta, T. Takeuchi, U. Mizutani,
797: and S. Sodeoka,
798: Jpn. J. Appl. Phys. 39, L1127 (2000).
799: \bibitem{itoh}
800: T. Itoh, T. Kawata, T. Kitajima, and I. Terasaki,
801: Proc. 17th Int. Conf. Thermoelectrics (ICT '98),
802: IEEE, Piscataway, p. 595 (cond-mat/9908039)
803: \bibitem{itoh2}
804: T. Itoh and I. Terasaki,
805: Jpn. J. Appl. Phys. 39, 6658 (2000).
806: \bibitem{funahashi2}
807: R. Funahashi and I. Matsubara,
808: Appl. Phys. Lett. 79, 362 (2001).
809: \bibitem{hebert}
810: S. Hebert, S. Lambert, D. Pelloquin, and A. Maignan,
811: Phys. Rev. B74, 172101 (2001).
812: \bibitem{terra2}
813: I. Terasaki,
814: Mater. Trans. 42, 951 (2001).
815: \bibitem{terra3}
816: I. Terasaki, Y. Ishii, D. Tanaka, K. Takahata,
817: and Y. Iguchi,
818: Jpn. J. Appl. Phys. 40, L65 (2001).
819: \bibitem{tallon}
820: J. L. Tallon, J. R. Cooper, P. S. I. P. N. de Silva,
821: G. V. M. Williams, and J. W. Loram,
822: Phys. Rev. Lett. 75, 4114 (1995).
823: \bibitem{fischer}
824: K. H. Fischer,
825: Z. Phys. B 76, 315 (1989) and references therein.
826: \bibitem{JH}
827: M. von Jansen and R. Hoppe,
828: Z. Anorg. Allg. Chem. 408, 104 (1974).
829: \bibitem{fouassier}
830: C. Fouassier, G. Matejka, J. Reau, and P. Hagenmuller,
831: J. Solid State Phys. 6, 532 (1973).
832: \bibitem{fukuzumi}
833: Y. Fukuzumi, K. Mizuhashi, K. Takenaka, and S. Uchida,
834: Phys. Rev. Lett. 76, 684 (1996).
835: \bibitem{LB1}
836: I. M. Khriplovich, E. V. Kholopov, and I. E. Paukov,
837: J. Chem. Thermodyn. 14, 207 (1982).
838: \bibitem{LB2}
839: P. Cossee,
840: J. Inorg. Nucl. Chem. 8, 483 (1958).
841: \bibitem{singh}
842: D. J. Singh,
843: Phys. Rev. B61, 13397 (2000).
844: \bibitem{mizokawa}
845: T. Mizokawa, L. H. Tjeng, P. G. Steeneken, N. B. Brookes,
846: I. Tsukada, T. Yamamoto, and K. Uchinokura
847: Phys. Rev. B 64, 115104 (2001).
848: \bibitem{mizokawa2}
849: T. Mizokawa, L. H. Tjeng, I. Terasaki, H.-J. Lin, C. T. Chen,
850: Synchrotron Radiation Research Center activity report,
851: Taiwan, China (unpublished).
852: \bibitem{comment}
853: The diffusive part of the thermopower
854: in a quasi-two-dimensional conductor is written as
855: $S/T \propto m/n$, where $n/m$ is the Drude weight
856: [Mandal et al., Phys. Condens. Matter {\bf 8}, 3047 (1996)].
857: For a polycrystalline NaCo$_2$O$_4$, $\rho$ is
858: roughly written as $aT+b$
859: [Kawata et al., Phys. Rev. B 60, 10584 (1999)],
860: and thus $d\rho/dT \propto m/n$.
861: \bibitem{coleman}
862: R. V. Coleman, M. P. Everson, Hao An Lu, A. Johnson
863: and L. M. Falicov,
864: Phys. Rev. B41, 460 (1990).
865: \bibitem{CrReview}
866: E. Fawcett, H. L. Alberts, V. Yu. Galkin, D. R. Noakes, and
867: J. V. Yakhmi,
868: Rev. Mod. Phys. 66, 25 (1996).
869: \bibitem{fisk}
870: R. Movshovich, A. Lacerda, P. C. Canfield, J. D. Thompson, and
871: Z. Fisk,
872: Phys. Rev. Lett. 73, 492 (1994).
873: \bibitem{coroneus}
874: J. Coroneus, B. Alavi and S. E. Brown,
875: Phys. Rev. Lett. 70, 2332 (1993).
876: \bibitem{choi}
877: M. Y. Choi, M. J. Burns, P. M. Chaikin, E. M. Engler
878: and R. L. Greene,
879: Phys. Rev. B31, 3576 (1985).
880: \bibitem{adachi}
881: S. Maki and K. Adachi,
882: J. Phys. Soc. Jpn. 46, 1131 (1979).
883: \bibitem{ong}
884: Z. A. Xu, N. P. Ong, Y. Wang, T. Kakeshita and S. Uchida,
885: Physica C 341-348, 1713 (2000).
886: \bibitem{lin}
887: C. L. Lin, J. E. Crow, P. Schlottmann, and T. Mihalisin,
888: J. Appl. Phys. 61, 4376 (1987).
889: \bibitem{sakurai}
890: J. Sakurai, H. Kamimura, and Y. Komura,
891: J. Mag. Mag. Mater. 76\&77, 287 (1988).
892: \bibitem{jaccard}
893: D. Jaccard, K. Behnia, and J. Sierro
894: Phys. Lett. A163, 475 (1992).
895: \bibitem{kitawaki}
896: R. Kitawaki and I. Terasaki,
897: (in preparation)
898: \bibitem{koshibae}
899: W. Koshibae, K. Tsutsui, and S. Maekawa,
900: Phys. Rev. B 62, 6869 (2000)
901: \bibitem{tsvelik}
902: A. M. Tsuvelik,
903: ``Quantum field theory in condensed matter physics'',
904: (Cambridge University Press, 1995, Cambridge) p.174.
905: \bibitem{takeuchi}
906: T. Takeuchi, M. Matoba, T. Aharen, and M. Itoh,
907: Physica B (in press)
908: \bibitem{tsukada}
909: I. Tsukada, T. Yamamoto, M. Takagi, T. Tsubone, S. Konno,
910: and K. Uchinokura,
911: J. Phys. Soc. Jpn. {\bf 70},834 (2001).
912: \bibitem{anno}
913: H. Anno,
914: Proc. 20th Int. Conf. Thermoelectrics (ICT 2001)
915: \end{thebibliography}
916:
917: \end{document}
918: