1: \documentclass[twocolumn,showpacs,showkeys,prl]{revtex4}% Physical Review B
2: \usepackage{graphicx}% Include figure files
3: \usepackage{dcolumn} % Align table columns on decimal point
4: %\usepackage{bm} % bold math
5: \setlength{\headsep}{0.8 in}
6: \renewcommand {\baselinestretch} {1}
7: \newcommand{\st}{SrTiO$_3$}
8: \newcommand{\bt}{BaTiO$_3$}
9: \newcommand{\ct}{CaTiO$_3$}
10: \newcommand{\lt}{LaTiO$_3$}
11: \newcommand{\btd}{BaTiO$_{3-\delta}$}
12: \newcommand{\po}{P$_{O_2}$}
13: \newcommand{\s}{$\sigma$}
14: \newcommand{\C}{$^{\circ}$C}
15: \newcommand{\vo}{V$_O$$^{\cdot\cdot}$}
16: \newcommand{\PAS}{Positron Annihilation Spectroscopy}
17: \newcommand{\ESR}{Electron Spin Resonance}
18:
19: \begin{document}
20:
21: %\preprint{APS/123-QED}
22: \title{Spin singlet small bipolarons in Nb-doped \bt}
23: \author{T. Kolodiazhnyi}
24: \email{koloditv@mcmaster.ca}
25: \author{S. C. Wimbush}
26: \affiliation{National Institute for Materials Science, ICYS, 1-1
27: Namiki, Tsukuba, Ibaraki 305-0044, Japan}
28:
29: \date{\today}% It is always \today, today,
30: % but any date may be explicitly specified
31:
32: \begin{abstract}
33:
34: The magnetic susceptibility and electrical resistivity of n-type
35: BaTi$_{1-x}$Nb$_x$O$_3$ have been measured over a wide temperature
36: range. It is found that, for 0 $<$ \emph{x} $<$ 0.2, dopant
37: electrons form immobile spin singlet small bipolarons with binding
38: energy around 110 meV. For \emph{x} = 0.2, a maximum in the
39: electrical resistivity around 15 K indicates a crossover from band
40: to hopping transport of the charge carriers, a phenomenon expected
41: but rarely observed in real polaronic systems.
42:
43: \end{abstract}
44:
45: \pacs{71.38.Mx, 71.38.Ht}% PACS, the Physics and Astronomy
46: % Classification Scheme.
47: \keywords{\bt, bipolarons, magnetic properties, electrical
48: resistivity}
49: %Use showkeys class option if keyword display desired
50: \maketitle
51:
52: %\section[Introduction]{Introduction}
53:
54: The theoretical treatment of itinerant carriers interacting with a
55: strongly polarizable lattice has been conducted for over 50 years,
56: resulting in a variety of different models. Depending on the degree
57: of electron-lattice interaction, these models can be roughly divided
58: into two main groups. The first \cite{pekar:1946,frohlich:1954}
59: treats the electron as a Bloch-wave-like delocalized particle called
60: a Fr\"{o}hlich or large polaron. The second
61: \cite{tjablikov:1952,holstein:1959} assumes an extreme localization
62: of the itinerant electron within one or several lattice sites (i.e.,
63: a small polaron, SP) accompanied by a strong local lattice
64: deformation. It is expected that under certain conditions, polarons
65: can pair to form large/small bipolarons, BP
66: \cite{vinetskii:1957,emin:1989}. Although not without controversy
67: \cite{chakraverty:1998}, several authors suggest \cite{emin:1989}
68: that Bose-Einstein condensation of bipolarons may result in
69: high-\emph{T}$_c$ superconductivity.
70:
71: Evidence of a strong electron-lattice interaction has been
72: accumulating for \emph{A}TiO$_3$, where \emph{A} = Ba or Sr
73: \cite{levstik:2002,lenjer:2002}. These are textbook examples of
74: crystals having a simple perovskite structure and high static
75: dielectric constant furnished by a soft phonon mode. Stoichiometric
76: \emph{A}TiO$_3$ are band gap insulators. Donor doping or reduction
77: in an oxygen-deficient atmosphere results an n-type material with
78: unconventional electronic conductivity where the carriers move
79: within a very soft and deformable lattice. In \st, the overlap
80: between the Ti-Ti 3d t$_{2g}$ orbitals is strong enough to prevent
81: the localization of the itinerant electrons which then form large
82: polarons. As a result, n-type \st\ shows metallic conductivity with
83: an enhanced effective electron mass \emph{m$^*$} of 6\emph{m$_e$} to
84: 14\emph{m$_e$} \cite{gervais:1993,verbist:1992}. It has been argued
85: that, at low temperatures, large polarons in n-type \st\ condense
86: into the large BPs responsible for superconductivity below 0.6 K
87: \cite{verbist:1992}.
88:
89: The lattice constant of cubic \bt\ (\emph{a} = 0.4005 nm) is 2.5$\%$
90: larger than that of \st\ (\emph{a} = 0.3905 nm). This small increase
91: in \emph{a} causes a dramatic difference in the electronic
92: properties of the two compounds. Unlike n-type \st, the conductivity
93: of n-type \bt\ shows insulating behavior with the low conductivity
94: attributed to the thermally activated hopping of strongly localized
95: electrons (i.e., small polarons) between neighboring Ti sites.
96: Gerthsen \emph{et al}.\cite{gerthsen:1965} explained the
97: mid-infrared reflectance of n-type \bt\ in terms of the SP
98: absorption model. In accord with the hopping transport of SPs,
99: Bursian \emph{et al}.\cite{bursian:1972} found that the Seebeck
100: coefficient of weakly-doped n-type \bt\ is temperature independent
101: above 300 K. Direct evidence of the self-trapped Ti$^{3+}$ SP was
102: provided by ESR studies of acceptor depleted \bt\
103: \cite{scharfschwerdt:1996}. Nevertheless, the SP interpretation of
104: the electronic properties of \bt\ is far from uncontroversial. Two
105: major points of disagreement are the similarity in the Hall and
106: drift mobilities ($\mu_{\mathrm{Hall}}$ $\approx$
107: $\mu_{\mathrm{drift}}$ $\approx$ 0.2 cm$^2$/V s at 300 K) which,
108: according to the SP model, should satisfy $\mu_{\mathrm{Hall}}$
109: $\gg$ $\mu_{\mathrm{drift}}$ \cite{ihrig:1978}, and the large
110: difference between the optical and thermal activation energies
111: (\emph{E}$_{\mathrm{op}}$ = 500 -- 600 meV, \emph{E}$_{\mathrm{th}}$
112: = 23 meV) of the proposed SPs, that should satisfy
113: \emph{E}$_{\mathrm{op}}$ = 4\emph{E}$_{\mathrm{th}}$
114: \cite{ihrig:1976}.
115:
116: In our opinion, the polaronic approach to understanding the unusual
117: properties of n-type \bt\ requires that the possibility of electron
118: pairing be taken into account. The coupling of electrons into
119: Anderson-type \emph{on-site} spin singlet Ti$^{2+}$ small bipolarons
120: according to Ti$^{4+}$ + 2\emph{e}$^-$ $\rightarrow$ Ti$^{2+}$ was
121: first postulated by Moizhes and Suprun \cite{moizhes:1984}. Later,
122: the idea of BPs in \bt\ was supported by Lenjer and co-workers
123: \cite{lenjer:2002}, who observed an unexpected increase in the
124: Ti$^{3+}$ ESR signal with temperature. This observation led them to
125: suggest that, due to the high correlation energy that sets the
126: energy cost for a double occupancy of the Ti$^{2+}$ ion, the
127: symmetry of the BP is rather of the \emph{inter-site}, i.e.,
128: Heitler-London type, where the electron pair is localized at two
129: neighboring Ti$^{3+}$ ions according to 2Ti$^{4+}$ + 2\emph{e}$^-$
130: $\rightarrow$ 2Ti$^{3+}$.
131:
132: The central question addressed in this Letter is whether the
133: Ti$^{3+}$ \emph{S} = $\frac{1}{2}$ small polarons in n-type \bt\
134: remain isolated at \emph{T} = 0 K or if they pair into spin singlet
135: \emph{S} = 0 bipolarons. We also report on the physical properties
136: of the BPs in n-type \bt, as revealed by magnetic and electrical
137: measurements.
138:
139: Polycrystalline BaTi$_{1-x}$Nb$_x$O$_3$ samples with 0 $\le$
140: \emph{x} $\le$ 0.2 were prepared by solid state reaction from high
141: purity 99.99+$\%$ BaCO$_3$, TiO$_2$, and Nb$_2$O$_5$. Sintering was
142: performed in forming gas (3$\%$ H$_2$/97$\%$ N$_2$) to ensure
143: electronic rather than cation vacancy compensation of the Nb dopant
144: \cite{lewis:1986}. Phase purity of the ceramics was confirmed by
145: powder x-ray diffraction and energy dispersive x-ray microanalysis
146: with no evidence of secondary phases up to \emph{x} = 0.2. According
147: to x-ray photoelectron spectroscopy measurements, charge
148: compensation in the sintered BaTi$_{1-x}$Nb$_x$O$_3$ samples is
149: accomplished by the formation of Ti$^{3+}$ ions while the Nb ions
150: remain predominantly in the 5+ oxidation state.
151:
152: Assuming strong localization of excess electrons on isolated
153: Ti$^{3+}$ ions down to 0 K (SP ground state), the magnetic
154: susceptibility should decrease with temperature in accordance with
155: the Curie law,
156: \begin{equation}
157: \chi_{\mathrm{para}} = \frac{N_{\mathrm{A}} p^2 \mu_{\mathrm{B}}^2
158: c_{\mathrm{P}}}{3k_{\mathrm{B}}T}, \label{eq1}
159: \end {equation}
160: where \emph{N}$_{\mathrm{A}}$ is the Avogadro number, $p$ is the
161: effective magnetic moment in Bohr magnetons ($\mu_{\mathrm{B}}$),
162: \emph{k}$_{\mathrm{B}}$ is the Boltzmann constant, \emph{T} is the
163: temperature and \emph{c}$_{\mathrm{P}}$ is the concentration of
164: small polarons. Figure \ref{fig1} shows the temperature dependence
165: of the molar magnetic susceptibility, $\chi$$_{\mathrm{M}}$, of the
166: BaTi$_{1-x}$Nb$_x$O$_3$ ceramics in the 2 -- 800 K temperature
167: range. It is obvious that the $\chi$$_{\mathrm{M}}$(\emph{T}) data
168: do not support a simple SP scenario.
169: \begin{figure}[htb]
170: \begin{center}
171: \scalebox{1.1}{\includegraphics [0,0][176,189]{fig1.eps}}
172: \caption{Magnetic susceptibility per mole of Ti of the BaTi$_{1-x}$Nb$_x$O$_3$ ceramics with \emph{x} = 0, 0.002, 0.008, 0.020,
173: 0.100 and 0.200.
174: Solid lines are fits to the data as described in the text.}
175: \label{fig1}
176: \end{center}
177: \end{figure}
178: \begin{figure}[htb]
179: \begin{center}
180: \scalebox{1.1}{\includegraphics*{fig1b.eps}}
181: \caption{Comparison of the magnetic susceptibility of Nb-doped \bt\ with that of \bt\ doped with donors having a stable oxidation state (Y$^{3+}$ and La$^{3+}$).
182: All three curves reveal a magnetic anomaly.}
183: \label{fig1b}
184: \end{center}
185: \end{figure}
186: The observed anomalous increase in $\chi$$_{\mathrm{M}}$ at
187: intermediate temperatures is associated with the breaking up of spin
188: singlet BPs. An overall increase in $\chi$$_{\mathrm{M}}$ with
189: dopant concentration is attributed to the formation of the BP band
190: with associated van Vleck paramagnetism. To fit the data we apply
191: Emin's formalism \cite{emin:1996} for dissociation of small BPs in a
192: weak magnetic field (\emph{g}$\mu_{\mathrm{B}}$\emph{B} $\ll$
193: 2$k_{\mathrm{B}}T$). According to Emin, the concentration of SPs,
194: \emph{c}$_{\mathrm{P}}$, produced by thermal dissociation of BPs is
195: given by
196: \begin{equation}
197: c_{\mathrm{P}} = \frac{1-(1-x)\sqrt{1+ \left[ x(2-x)/(1-x)^2 \right]
198: e^{\varepsilon_{\mathrm{b}}/k_{\mathrm{B}}T}
199: }}{1-e^{\varepsilon_{\mathrm{b}}/k_{\mathrm{B}}T}}, \label{eq2}
200: \end {equation}
201: where \emph{x} is the concentration of the Nb$^{5+}$ dopant and
202: $\varepsilon$$_{\mathrm{b}}$ is the BP binding energy, i.e., the
203: energy required to split the BP into two SPs. The data in
204: Fig.\ref{fig1} were fitted with the general formula
205: \begin{equation}
206: \chi_{\mathrm{M}} = \chi_{\mathrm{dia}} + \chi_{\mathrm{VV}} +
207: \frac{A}{T} + \chi_{\mathrm{para}}, \label{eq3}
208: \end {equation}
209: where $\chi_{\mathrm{dia}}$ = --2.32$\times$10$^{-5}$ cm$^3$/mol is
210: the diamagnetic susceptibility of the undoped, stoichiometric \bt,
211: $\chi_{\mathrm{VV}}$ is the temperature independent van Vleck
212: paramagnetic contribution, and \emph{A} accounts for
213: impurity-related paramagnetism dominant at \emph{T} $<$ 40 K.
214:
215: The van Vleck term is given by
216: \begin{equation}
217: \chi_{\mathrm{VV}}=\frac{2N_{\mathrm{A}} x |\langle l
218: |\mu_{\mathrm{z}}|0\rangle|^2}{\Delta}
219: \label{eq4}
220: \end {equation}
221: where $\langle l |\mu_{\mathrm{z}}|0\rangle$ is a non-diagonal
222: matrix element of the magnetic moment operator connecting the BP
223: ground state 0 with the excited state \emph{l} of energy $\Delta$ =
224: \emph{E$_l$} -- \emph{E$_0$} above the ground state.
225:
226: The results of the fit are shown as solid lines in Fig.\ref{fig1}.
227: The fitting parameters are summarized in Table \ref{table1}. A
228: somewhat larger value of \emph{A} for the \emph{x} = 0.2 sample is
229: attributed to the partial compensation of the Nb dopant by
230: paramagnetic cation vacancies as we approach the solubility limit of
231: Nb in \bt. In agreement with Eq.\ref{eq4}, $\chi_{\mathrm{VV}}$ is
232: found to scale linearly with dopant concentration. Good fits were
233: obtained for \emph{x} = 0, 0.002, 0.1 and 0.2 and somewhat worse
234: fits for \emph{x} = 0.008 and 0.02. We can explain the poor fit at
235: intermediate dopant concentrations since, while fitting the data, we
236: have assumed a constant $\varepsilon$$_{\mathrm{b}}$ and
237: $\chi_{\mathrm{VV}}$ for the entire temperature range. It is known,
238: however, that undoped \bt\ exists in four crystallographic forms
239: \cite{jona:1962}: rhombohedral at \emph{T} $<$ 200 K, orthorhombic
240: at 200 K $<$ \emph{T} $<$ 270 K, tetragonal at 270 K $<$ \emph{T}
241: $<$ 393 K and cubic at 393 K $<$ \emph{T} $<$ 1670 K. Since both
242: $\varepsilon$$_{\mathrm{b}}$ and $\chi_{\mathrm{VV}}$ depend on the
243: crystal symmetry, the use of single values of
244: $\varepsilon$$_{\mathrm{b}}$ and $\chi_{\mathrm{VV}}$ for \emph{x} =
245: 0.008 and 0.02 gives only marginally satisfactory results. At the
246: lowest dopant concentration (\emph{x} = 0.002), 70$\%$ of the BPs
247: are already dissociated into SPs at 200 K, i.e., fully within the
248: rhombohedral phase. As a result, this fit is good in spite of the
249: single values of $\varepsilon$$_{\mathrm{b}}$ and
250: $\chi_{\mathrm{VV}}$ used. Doping with Nb results in a significant
251: lowering of the phase transition temperatures in \bt\
252: \cite{feltz:1977}. Indeed, low-temperature x-ray analysis of the
253: \emph{x} $\ge$ 0.1 samples confirmed that they remain cubic down to
254: at least 20 K. Hence, the use of single values of
255: $\varepsilon$$_{\mathrm{b}}$ and $\chi_{\mathrm{VV}}$ produces a
256: good fit also in these cases. The data show that in the doping range
257: covering two orders of magnitude, the bipolaron binding energy is of
258: the order of 110 meV. For the most highly doped sample (\emph{x} =
259: 0.2), the bipolaron binding energy (172 meV) is found to be rather
260: high. This may be due to the solubility limit of donor atoms and/or
261: BP overcrowding effects.
262: \begin{table}[htb]
263: \begin{center}
264: \caption {\label{table1}Fitting parameters for $\chi_{\mathrm{M}}$
265: of BaTi$_{1-x}$Nb$_x$O$_3$.}
266: \begin{tabular}{lcccc}\hline\hline\\[0.1ex]
267: \emph{x} &$\chi_{\mathrm{VV}}$$\times$10$^5$ & \emph{A}$\times$10$^5$ & $\varepsilon$$_{\mathrm{b}}$ & \emph{p}\\
268: & [cm$^3$/mol]& [cm$^3$K/mol]& [meV]& \\[1ex]
269: \hline\\[0.5ex]
270: 0.000 & 0 & 2.16 & -- & -- \\[0.5ex]
271: 0.002 & 0.02 & 5.73 & 106 & 0.57 \\[0.5ex]
272: 0.008 & 0.47 & 11.73 & 122 & 0.46 \\[0.5ex]
273: 0.020 & 0.73 & 16.64 & 111 & 0.46 \\[0.5ex]
274: 0.100 & 2.68 & 5.96 & 105 & 0.42 \\[0.5ex]
275: 0.200 & 5.69 & 22.5 & 172 & 0.36 \\[0.5ex]
276: \hline
277: \end{tabular}
278: \end{center}
279: \end{table}
280:
281: One might argue, however, that the BP interpretation of the
282: $\chi_{\mathrm{M}}$ anomaly is not convincing enough, and that
283: thermal activation of intrinsic or extrinsic lattice defects may
284: generate similar effects. Let us address this crucial argument in
285: detail. Since no $\chi_{\mathrm{M}}$ anomaly is observed in undoped,
286: stoichiometric \bt, the possibility of the \emph{S} = 0
287: $\rightarrow$ \emph{S} = $\frac{1}{2}$ activation of any supposed
288: extrinsic impurities can be ruled out. According to the point defect
289: chemistry of \bt, the Nb$^{5+}$ dopant is compensated by barium or
290: titanium vacancies at high oxygen partial pressures
291: (\emph{P}$_{\mathrm{O}_2}$) and by electrons at low
292: \emph{P}$_{\mathrm{O}_2}$ \cite{lewis:1986}. Since the samples were
293: prepared at low \emph{P}$_{\mathrm{O}_2}$, the concentration of
294: cation vacancies is negligible. However, even if a small amount of
295: cation vacancies were still present in the heavily doped samples,
296: their thermal activation energy is at least five times higher than
297: $\varepsilon$$_{\mathrm{b}}$ \cite{lewis:1986}. It is quite possible
298: that oxygen vacancies ($\mathrm{V_{O}}$) are present in the samples.
299: According to recent calculations \cite{honnerberg:2000}, these
300: vacancies will form a $[$Ti$^{3+}$--$\mathrm{V_{O}}$--Ti$^{3+}$$]$
301: neutral complex defect. In fact, the
302: $[$Ti$^{3+}$--$\mathrm{V_{O}}$--Ti$^{3+}$$]$ complex is nothing
303: other than the small BP bound to the oxygen vacancy, although in
304: this case the electrons are localized on the Ti 3z$^2$--r$^2$
305: orbitals \cite{solovyev:1000}. Activation of this defect according
306: to $[$Ti$^{3+}$--$\mathrm{V_{O}}$--Ti$^{3+}$$]$ $\rightarrow$
307: $[$Ti$^{3+}$--$\mathrm{V_{O}}$$]$ + Ti$^{3+}$ may indeed cause a
308: $\chi_{\mathrm{M}}$ anomaly similar to that observed in
309: Fig.\ref{fig1}. However, donor doping always decreases the
310: concentration of $\mathrm{V_{O}}$ \cite{lewis:1986}. Hence one would
311: expect a decrease in the $\chi_{\mathrm{M}}$ anomaly with increasing
312: doping, which is opposite to what is seen in our results. Finally,
313: we rule out the possibility that the $\chi_{\mathrm{M}}$ anomaly is
314: associated with a change in the ionization state of Nb$^{5+}$
315: according to Nb$^{4+}$ + Ti$^{4+}$ $\rightarrow$ Nb$^{5+}$ +
316: Ti$^{3+}$ since Fig.\ref{fig1b} shows similar magnetic anomalies
317: occurring when electrons are introduced into \bt\ by donors having a
318: stable oxidation state (Y$^{3+}$ and La$^{3+}$).
319:
320: Recently, magnetic susceptibility measurements have been made on
321: several 3d$^1$ titanates including MgTi$_2$O$_4$, NaTiSi$_2$O$_6$
322: and LiTiSi$_2$O$_6$ \cite{isobe2:2002}. It was found that magnetic
323: anomalies in all the above compounds were associated with the
324: formation of spin-singlet Ti$^{3+}$--Ti$^{3+}$ pairs. In contrast to
325: the above compounds, the formation of BPs in n-type \bt\ is not
326: driven by a change in the crystal symmetry, as revealed in the
327: samples with \emph{x} $\ge$ 0.1. It also appears that pairing of the
328: electrons occurs even for very diluted 3d$^{0+\delta}$ ($\delta$
329: $\ll$ 1) electronic concentrations as in the case of \emph{x} =
330: 0.002.
331: \begin{figure}[htb]
332: \begin{center}
333: \scalebox{0.8}{\includegraphics*{fig2.eps}}
334: \caption{Electrical resistivity of the BaTi$_{1-x}$Nb$_x$O$_3$ ceramics with \emph{x} = 0.002, 0.008, 0.020,
335: 0.100 and 0.200. The inset highlights the crossover from band to hopping conduction in the \emph{x} = 0.200 sample.}
336: \label{fig2}
337: \end{center}
338: \end{figure}
339:
340: In view of the heated debate regarding the possibility of the
341: existence of \emph{mobile} small BPs \cite{chakraverty:1998}, we
342: have also examined the electronic behavior of n-type \bt. According
343: to the resistivity data shown in Fig.\ref{fig2}, the ground state of
344: n-type \bt\ is insulating throughout the 0 $<$ \emph{x} $<$ 0.2
345: doping range \cite{comment1:2005} and therefore comprises immobile
346: small BPs. The total conductivity can be expressed as a sum of
347: polaronic, $\mathrm{\sigma_{P}}$, and bipolaronic,
348: $\mathrm{\sigma_{BP}}$, conductivities
349: \begin{equation}
350: \sigma = \mathrm{\sigma_{P}} + \mathrm{\sigma_{BP}} =
351: ec_{\mathrm{P}} \mu_{\mathrm{P}}
352: +2e\frac{x-c_{\mathrm{P}}}{2}\mu_{\mathrm{BP}}, \label{eq5}
353: \end {equation}
354: where $\mu$$_{\mathrm{P}}$ and $\mu$$_{\mathrm{BP}}$ are the drift
355: mobilities of SPs and BPs, respectively. We assume, in accordance
356: with Ref. \cite{emin:1996}, that for a low doping level the main
357: contribution to $\sigma$ comes from the SPs that are formed by
358: thermal dissociation of the BPs and that $\mathrm{\sigma_{BP}}$ is
359: negligible due to the large effective mass and very low mobility of
360: BPs. Then, in the case of non-adiabatic hopping of SPs
361: \cite{bursian:1971},
362: \begin{equation}
363: \sigma \approx ec_{\mathrm{P}} \mu_{\mathrm{P}} \propto
364: \frac{1}{T^{1.5}}\mathrm{exp} \left[-\frac{\varepsilon_{\mathrm{b}}
365: + E_{\mathrm{th}}}{k_{\mathrm{B}}T} \right]. \label{eq6}
366: \end {equation}
367:
368: Taking \emph{E}$_{\mathrm{th}}$ = 23 meV from Ref. \cite{ihrig:1976}
369: and $\varepsilon$$_{\mathrm{b}}$ from Table \ref{table1}, the total
370: activation energy of conductivity, $\varepsilon$$_{\mathrm{b}}$ +
371: \emph{E}$_{\mathrm{th}}$, should be around 135 $\pm$ 10 meV at high
372: temperatures, with a slight deviation for the \emph{x} = 0.2 sample.
373: Indeed the resistivity data confirm that all but the \emph{x} = 0.2
374: sample show activated conductivity above \emph{T} $\approx$ 155 K
375: with an activation energy of ca. 135 meV as shown in Fig.\ref{fig3}.
376: The strong downturn from Arrhenius behavior for \emph{T} $<$ 155 K
377: is due to an enhancement of the mobility by quantum lattice
378: fluctuations \cite{fratini:2003}.
379: \begin{figure}[htb]
380: \begin{center}
381: \scalebox{1}{\includegraphics{fig3.eps}}
382: \caption{Arrhenius plot of the resistivity of the BaTi$_{1-x}$Nb$_x$O$_3$ ceramics assuming non-adiabatic hopping of the localized carriers.}
383: \label{fig3}
384: \end{center}
385: \end{figure}
386: As the dopant concentration increases, the $\mathrm{\sigma_{BP}}$
387: contribution can no longer be neglected. In fact, the sample with
388: the highest doping level (\emph{x} = 0.2) shows a strong
389: conductivity enhancement at low temperature, with a maximum in
390: $\rho$ at \emph{T} = 15 K as shown in the inset of Fig.\ref{fig2}.
391: According to the SP theory, the resistivity maximum is expected
392: \cite{fratini:2003} but rarely observed in real polaronic systems
393: \cite{schein:1978}. In our case, it is associated with a crossover
394: from the coherent transport of BPs within the bipolaron band below
395: 15 K to the incoherent (i.e., hopping) transport of BPs above 15 K.
396: We should stress here that this phenomenon is not unique to Nb-doped
397: samples and has also been observed in La-doped \bt.
398:
399: Finally, let us compare the thermal dissociation energy of BPs,
400: $\varepsilon$$_{\mathrm{b}}$, obtained in this work with literature
401: data on the infra-red absorption band of n-type \bt\ centered at
402: \emph{E}$_{\mathrm{op}}$ = 500 -- 600 meV
403: \cite{berglund:1967a,bursian:1971}. Optical absorption of BPs
404: involves the splitting of the two electrons and transfer of one of
405: them to the next neighboring site. According to Emin
406: \cite{emin:1993}, this process would require an energy of
407: \emph{E}$_{\mathrm{op}}$ $\approx$ 4$\varepsilon$$_{\mathrm{b}}$.
408: Taking $\varepsilon$$_{\mathrm{b}}$ data for weakly doped samples
409: (\emph{x} $<$ 0.02) from Table \ref{table1}, the optical absorption
410: due to the BPs would produce a band centered at 420 -- 490 meV,
411: which is in reasonable agreement with the literature data. A
412: BP-associated optical absorption is expected to be strongly
413: temperature dependent since the concentration of the BPs decreases
414: with temperature. An alternative source of the mid-infrared
415: absorption band in n-type \bt\ could be the d$_{xy}$ $\rightarrow$
416: d$_{xz}$, d$_{yz}$ orbital (intra-band) excitations. Formation of
417: the SP removes the three-fold t$_{2g}$ orbital degeneracy in \bt\
418: due to the \emph{T}$_2$$\times$\emph{e} Jahn-Teller effect.
419: According to ESR data, the energy associated with the orbital
420: excitations within the t$_{2g}$ band in n-type \bt\ is around 400 --
421: 536 meV \cite{scharfschwerdt:1996a}. In this case, however, the
422: intensity and the energy maximum of the optical absorption should be
423: temperature independent. Since there are no data on the temperature
424: dependence of the mid-infrared band of n-type \bt, the certain
425: origin of this band remains unresolved.
426:
427: In conclusion, as revealed by magnetic susceptibility data,
428: ``itinerant'' electrons in BaTi$_{1-x}$Nb$_x$O$_3$ with 0 $<$
429: \emph{x} $<$ 0.2 form localized pairs with a spin singlet ground
430: state. The binding energy of the BPs ranges from 105 meV to 122 meV
431: depending on the dopant concentration. Electrical resistivity
432: measurements show that, throughout this doping range, the BPs are
433: \emph{immobile}. A crossover from band to hopping transport is
434: manifested by a resistivity maximum at \emph{T} = 15 K in the most
435: heavily doped (\emph{x} = 0.2) sample investigated. This indicates
436: that at rather high doping concentration, small BPs in \bt\ become
437: \emph{mobile}. In a more general perspective, we speculate that \bt\
438: and \st\ play host to the two extreme cases of the bipolaron ground
439: state: the small BP in \bt\ evolves into a large BP in \st. Hence,
440: \bt\ -- \st\ may become a model system for studies of the physics of
441: small and large BPs as it became a model system for studies of soft
442: mode behavior 50 years ago.
443: \begin{acknowledgments}
444:
445: This study was performed using Special Coordination Funds for
446: Promoting Science and Technology from the Ministry of Education,
447: Culture, Sports, Science and Technology of the Japanese Government.
448: The authors are grateful to I. Solovyev, A. Mishchenko and D.
449: Khomskii for helpful comments.
450:
451: \end{acknowledgments}
452:
453: \begin{thebibliography}{99}
454: %\bibitem{zaanen:1985}J. Zaanen, G. A. Sawatzky and J. W. Allen, Phys.
455: %Rev. Lett., \textbf{55}, 418 (1985).
456: \bibitem{pekar:1946}S. I. Pekar, Zh. Eksp. Teor. Fiz. \textbf{16}, 335 (1946).
457: \bibitem{frohlich:1954}H. Fr\"{o}hlich, Adv. Phys. \textbf{3}, 325 (1954).
458: \bibitem{tjablikov:1952}S. V. Tjablikov, Zh. Eksp. Teor. Fiz. \textbf{23}, 381 (1952).
459: \bibitem{holstein:1959}T. Holstein, Ann. Phys. \textbf{8}, 325 (1959).
460: \bibitem{vinetskii:1957}V. L. Vinetskii and M. Sh. Giterman, Zh. Eksp. Teor. Fiz. \textbf{33}, 730
461: (1957), V. L. Vinetskii, \emph{ibid}. \textbf{40}, 1459 (1961).
462: \bibitem{emin:1989}D. Emin and M. S. Hillery, Phys. Rev. B \textbf{39}, 6575
463: (1989). A. Alexandrov and J. Ranninger, \emph{ibid}. \textbf{23},
464: 1796 (1981).
465: \bibitem{chakraverty:1998}B. K. Chakraverty, J. Ranninger and D.
466: Feinberg, Phys. Rev. Lett. \textbf{81}, 433 (1998). A. S.
467: Alexandrov, \emph{ibid.} \textbf{82}, 2620 (1999).
468: \bibitem{levstik:2002} A. Levstik, C. Filipic, V. Bobnar and R.
469: Pirc, Appl. Phys. Lett. \textbf{81}, 4046 (2002).
470: \bibitem{lenjer:2002}S. Lenjer, O. F. Schirmer, H. Hesse and Th. W. Kool, Phys. Rev. B \textbf{66},
471: 165106 (2002).
472: \bibitem{gervais:1993}F. Gervais, J. L. Servoin, A. Baratoff, J. G.
473: Bednorz, G. Binnig, Phys. Rev. B \textbf{47}, 8187 (1993).
474: \bibitem{verbist:1992} G. Verbist, F. M. Peeters and J. T. Devreese,
475: Ferroelectrics \textbf{130}, 27 (1992). N. F. Mott in \emph{Polarons
476: and Bipolarons in High-T$_c$ Superconductors and Related Materials},
477: edited by E. K. H. Salge, A. S. Alexandrov and W. Y. Liang
478: (Cambridge University Press 1995), pp. 2-10.
479: \bibitem{gerthsen:1965} P. Gerthsen, R. Groth, K. H. H\"{a}rdtl, D.
480: Hesse and H. G. Reik, Solid State Comm. \textbf{3}, 165 (1965).
481: \bibitem{bursian:1972}E. V. Bursian, Ya. G. Girshberg and E. N.
482: Starov, Sov. Phys. Solid State \textbf{14}, 872 (1972).
483: \bibitem{scharfschwerdt:1996} R. Scharfschwerdt, O. F. Schirmer, H.
484: Kr\"{o}se and Th. W. Kool, Ferroelectrics \textbf{185}, 9 (1996).
485: \bibitem{ihrig:1978}H. Ihrig and D. Hennings, Phys. Rev. B
486: \textbf{17}, 4593 (1978).
487: \bibitem{ihrig:1976}H. Ihig, J. Phys. C: Solid State Phys. \textbf{9}, 3469
488: (1976).
489: \bibitem{moizhes:1984} B. Y. Moizhes and S. G. Suprun, Sov. Phys. Solid
490: State \textbf{26}, 544 (1984).
491: \bibitem{lewis:1986} G. V. Lewis and C. R. A. Catlow, J. Phys. Chem.
492: Solids \textbf{47}, 89 (1986). J. Daniels and K. H. H\"{a}rdtl,
493: Philips Res. Rep. \textbf{31}, 489 (1976).
494: \bibitem{emin:1996}D. Emin, Phys. Rev. B \textbf{53}, 1260 (1996).
495: \bibitem{jona:1962}F. Jona and G. Shirane, \emph{Ferroelectric Crystals},
496: Pergamon Press, Oxford 1962.
497: \bibitem{feltz:1977} A. Feltz and H. Langbein, Ferroelectrics \textbf{15}, 7 (1977).
498: \bibitem{honnerberg:2000} H. Honnerberg and A. Birkholz, J.
499: Phys.: Condens. Matter \textbf{12}, 8239 (2000).
500: \bibitem{solovyev:1000} I. Solovyev (private communication).
501: \bibitem{isobe2:2002}M. Isobe, E. Ninomiya, A. N. Vasil'ev and Y.
502: Ueda, J. Phys. Soc. Jpn. \textbf{71}, 1423 (2002). M. Isobe and Y.
503: Ueda, \emph{ibid}. \textbf{71}, 1848 (2002).
504: \bibitem{comment1:2005} In accordance
505: with the resistivity data, no electronic contribution to the
506: specific heat has been detected at low temperatures, indicating the
507: absence of itinerant carriers.
508: \bibitem{fratini:2003}S. Fratini and S. Ciuchi, Phys. Rev. Lett.
509: \textbf{91}, 256403 (2003).
510: \bibitem{schein:1978}L. B. Schein, C. B. Duke and A. R. McGhie, Phys.
511: Rev. Lett. \textbf{40}, 197 (1978). P. Kwizera, M. S. Dresselhaus
512: and D. Adler, Phys. Rev. B \textbf{21}, 2328 (1980). G. M. Zhao, V.
513: Smolyaninova, W. Prellier and H. Keller, Phys. Rev. Lett.
514: \textbf{84}, 6086 (2000).
515: \bibitem{berglund:1967a}C. N. Berglund and H. J. Braun, Phys. Rev.
516: \textbf{164}, 790 (1967).
517: \bibitem{bursian:1971} E. V. Bursian, Ya. G. Girshberg and E. N.
518: Starov, phys. stat. sol. b \textbf{46}, 529 (1971).
519: \bibitem{emin:1993}D. Emin, Phys. Rev. B \textbf{48}, 13691 (1993).
520: \bibitem{scharfschwerdt:1996a} R. Scharfschwerdt, A. Mazur, O. F.
521: Schirmer, H. Hesse, and S. Mendricks, Phys. Rev. B \textbf{54},
522: 15284 (1996).
523: \end{thebibliography}
524:
525: \end{document}
526: