1: %\documentstyle[preprint,aps]{revtex}

2: \documentstyle[eqsecnum,aps]{revtex}

3: \def\btt#1{{\tt$\backslash$#1}}

4: \begin{document}

5: \draft \preprint{HEP/123-qed}

6: \title{Path Integral of the Holstein Model with a $\phi^4$ on site

7: potential}

8: \author{ Marco Zoli }

9: \address{Istituto Nazionale Fisica della Materia - Dipartimento di Fisica

10: \\ Universit\'a di Camerino, 62032, Italy. - marco.zoli@unicam.it}

11: 

12: \date{\today}

13: 

14: \maketitle

15: \begin{abstract}

16: We derive the path integral of the semiclassical, one dimensional

17: anharmonic Holstein model assuming that the electron motion takes

18: place in a bath of non linear oscillators with quartic on site

19: hard (and soft) potentials. The interplay between {\it e-ph}

20: coupling and anharmonic force constant is analysed both in the

21: adiabatic and antiadiabatic regime. In the latter we find much

22: larger anharmonic features on the thermodynamic properties of low

23: energy oscillators. Soft on site potentials generate attractive

24: centres at large amplitude oscillator paths and contribute to the

25: anomalous shape of the {\it heat capacity over temperature} ratio

26: in the intermediate to low $T$ range. This anharmonic lattice

27: effect is superimposed to the purely electronic contribution

28: associated to a temperature dependent hopping with variable range

29: inducing local disorder in the system.

30: \end{abstract}

31: 

32: \pacs{PACS: 31.15.Kb, 71.38.-k, 63.20.Ry }

33: 

34: \narrowtext

35: \section*{I. Introduction}

36: 

37: There is at present a large interest on the effects of a strong

38: electron-phonon coupling in a number of systems ranging from dimer

39: molecular junctions \cite{kaat} to carbon nanotubes \cite{mah},

40: from organic molecular crystals \cite{han}, to DNA \cite{alexan}

41: and cuprate superconductors \cite{sch,lan}. Several theoretical

42: studies have focussed on the interplay between e-ph coupling and

43: non linearities in the framework of the Holstein model

44: \cite{chris,voulga} investigating the phase diagram both in the

45: adiabatic \cite{zieg} and the antiadiabatic regime. Mainly in the

46: latter anharmonic effects are believed to be large \cite{pietro}

47: thus offering a picture to explain the high $T_c$ of binary alloys

48: such as superconducting $MgB_2$ with a small Fermi energy and

49: sizeable {\it e-ph} coupling \cite{cava}.

50: 

51: The path integral formalism provides a powerful method to study

52: quantum systems in which a particle is non linearly coupled to the

53: environment \cite{fehi,kleinert,eck}. A previous path integral

54: analysis \cite{io5} has pointed out how the phonon dispersion,

55: that has to be taken into account in the computation of the ground

56: state properties of the Holstein Hamiltonian

57: \cite{holst1,raedt2,io98}, induces non local {\it e-ph}

58: correlations which renormalize downwards the effective coupling

59: and ultimately broaden the size of the polaronic quasiparticle.

60: This explains why the polaron mass in a dispersive Holstein model

61: \cite{atin} is lighter than in a dispersionless model

62: \cite{alekor}. Also the thermodynamics of the Holstein Hamiltonian

63: can be computed within a dispersive model which accounts for the

64: lattice structure \cite{io5}.

65: 

66: The Holstein diatomic molecular model was originally cast

67: \cite{holst} in the form of a discrete non linear Schr\"odinger

68: equation for electrons whose probability amplitude at a molecular

69: site depends on the interatomic vibration coordinates. The non

70: linearities are tuned by the {\it e-ph} coupling \cite{kenkre}

71: whose strength drives the crossover between a large and a small

72: polaron state according to the degree of adiabaticity and the

73: dimensionality of the system \cite{trug}.

74: 

75: In the Holstein model, the phonon thermodynamics is not affected

76: by {\it e-ph} induced anharmonicities \cite{io4}. This follows

77: from the fact that the Holstein perturbing source current is local

78: in time and it does not depend on the electron path coordinate. As

79: a consequence, in the total partition function, electron and

80: lattice degrees of freedom are disentangled and the latter can be

81: integrated out analytically as far as a harmonic lattice model is

82: assumed. However non linearities may arise in Holstein like

83: systems also by virtue of on site potentials dependent on the

84: lattice structure and, in principle, independent of the {\it e-ph}

85: coupling. We are thus led to investigate the thermodynamics of the

86: anharmonic Holstein model with a quartic on site potential which

87: may be repulsive or partly attractive according to the sign of the

88: force constant and the amplitude of the lattice displacement

89: paths. The path integral approach permits to monitor the physical

90: properties for any value of the coupling strengths \cite{feynman}.

91: The presence of a $\phi^4$ potential may in turn affect also the

92: {\it e-ph} interactions and sinergically interfere on the

93: equilibrium properties of the system. This is the focus of the

94: present paper. Section II presents the Hamiltonian model while the

95: path integral method is briefly described in Section III. The

96: derivation of the total partition function of the system is

97: presented in Section IV and Section V contains the discussion of

98: the physical results. The conclusions are drawn in Section VI.

99: 

100: \section*{II. The Anharmonic Holstein Model}

101: 

102: We consider the one dimensional anharmonic Holstein Hamiltonian

103: consisting of: i) one electron hopping term, ii) an interaction

104: which couples the electronic density ($f_{\bf l}^{\dag}f_{\bf l}$)

105: to the lattice displacement $u_l$ at the {l}-site, iii) a bath of

106: $N$ identical dispersionless anharmonic oscillators with mass $M$

107: and frequency $\omega$ :

108: 

109: \begin{eqnarray}

110: & &H =\, H^e + H^{e-ph} + H^{ph} \nonumber

111: \\

112: & &H^e =\, - t \sum_{<{l, m}>} f_{l}^{\dag} f_{m} \nonumber

113: \\

114: & &H^{e-ph}=\,  {\bar g} \sum_{l=1}^Nu_l f^{\dag}_l f_{l} \,

115: \nonumber

116: \\

117: & &H^{ph}=\, {M \over 2} \sum_{l=1}^N \bigl(\dot{u}_l^2 + \omega^2

118: u_l^2 \bigr) + {{\delta} \over 4} \sum_{l=1}^N u_l^4  \, \nonumber

119: \\ & &{\bar g}=\,g \sqrt{2M \omega}

120: \end{eqnarray}

121: 

122: the sum $<,>$ is over $z$ nearest neighbors and $t$ is the tight

123: binding overlap integral.  $g$ is the {\it e-ph} coupling in units

124: of $\hbar \omega$. Choosing the atomic mass $M$ of order $10^4$

125: times the electron mass, we get ${\bar g} \simeq \, 1.1456 \times

126: g \sqrt{2 \hbar \omega}{}\, \bigl[ meV \AA^{-1} \bigr]$ where

127: $\hbar \omega$ is given in $meV$. $\delta$, in units $meV

128: \AA^{-4}$, controls the strength of the non linearities and

129: determines whether the on site potential $V(u_l)=\, {M}\omega^2

130: u_l^2/2 + {{\delta}}u_l^4/4$ is hard ($\delta > 0$) or soft

131: ($\delta < 0$). In the latter case, $V(u_l)$ attains the maximum

132: at $u_l^2=\,M\omega^2/|\delta|$ and the inflection point occurs at

133: $u_l^2=\,M\omega^2/3|\delta|$. The condition $|\delta|u_l^2 > 2

134: M\omega^2$ yields an attractive on site potential. Then, the range

135: of the atomic path amplitudes generating attractive scattering

136: centres depends on the value of the anharmonic force constant. For

137: $|\delta| > 2 M\omega^2$ the potential becomes attractive for a

138: portion of large amplitude atomic paths while small amplitude

139: paths weigh the repulsive range.

140: 

141: In the following computation of the electron path integral coupled

142: to the anharmonic oscillator, after setting the potential

143: parameters, we select at any temperature the class of atomic paths

144: which mainly contribute to the euclidean action. As the

145: distribution of the path amplitudes has a cutoff on the scale of

146: the lattice constant, say i.e. $u_l^2 < 1 \AA^2$, the on site

147: potentials are always bound from below also in the attractive

148: cases.

149: 

150: \section*{III. The Path Integral Method}

151: 

152: The Holstein Hamiltonian in (1) can be mapped onto the time scale

153: according to space-time mapping techniques extensively described

154: in previous works \cite{io5,hamann,io3} and hereafter outlined.

155: 

156: Defining ${x}(\tau)$ and ${y}(\tau')$ as the electron coordinates

157: at the ${l}$ and ${m}$ lattice sites respectively, $H^e$ in (1)

158: transforms into

159: 

160: \begin{equation}

161: H^e(\tau,\tau')=\, -{t} \bigl( f^{\dag}({x}(\tau)) f({y}(\tau')) +

162: f^{\dag}({y}(\tau')) f({x}(\tau)) \bigr)

163: \end{equation}

164: 

165: where $\tau$ and $\tau'$ are continuous variables $\bigl( \in [0,

166: \beta]\bigr)$, with $\beta$ being the inverse temperature. After

167: setting $\tau'=\,0$, ${y}(0) \equiv 0$ and taking the thermal

168: averages for the electron operators over the ground state of the

169: Hamiltonian one gets the average electron hopping energy per

170: lattice site:

171: 

172: \begin{eqnarray}

173: & &h^e(x(\tau)) \equiv {{<H^e(\tau)>} \over N}= \, - {t}\Bigl(G[-{

174: x}(\tau), -\tau ] + G[{x}(\tau), \tau ]\Bigr)\, \nonumber \\

175: \end{eqnarray}

176: 

177: where, $G[{x}(\tau), \tau ]$ is the electron propagator at finite

178: temperature.

179: 

180: By treating the lattice displacements in (1) as $\tau -$ dependent

181: classical variables, $u_l \rightarrow u(\tau)$, we obtain from

182: $H^{e-ph}$ in (1) the averaged {\it e-ph} energy per lattice site

183: which is identified as the perturbing source current $j(\tau)$ in

184: the path integral method:

185: 

186: \begin{eqnarray}

187: j(\tau) \equiv {{<H^{e-ph}(\tau)>} \over N}= \,{\bar g} u(\tau)

188: \end{eqnarray}

189: 

190: As the Hamiltonian model assumes a set of identical oscillators we

191: study the path integral for the electron coupled to a single

192: anharmonic oscillator of the bath. The path integral reads:

193: 

194: \begin{eqnarray}

195: & &<{x}(\beta)|{x}(0)> =\, \int D{x}(\tau) exp\Biggl[-

196: \int_0^{\beta}d\tau E(x(\tau))\Biggr] \nonumber

197: \\ &\times& \int Du(\tau) exp\Biggl[- \int_0^{\beta}

198: d\tau O(u(\tau)) \Biggr] \,\nonumber

199: \\

200: & &E(x(\tau))=\,{{m_e} \over 2} \dot{{x}}^2(\tau) + h^e(x(\tau))

201: \,\nonumber

202: \\

203: & &O(u(\tau))=\, {M \over 2} \bigl( \dot{u}^2(\tau) + \omega^2

204: u^2(\tau) \bigr) + {{\delta} \over 4} u^4(\tau) +

205: j(\tau)\,\nonumber

206: \\

207: \end{eqnarray}

208: 

209: where the kinetic term ($m_e$ is the electron mass) is normalized

210: by the functional measure of integration over the electron paths.

211: 

212: Since the electron hopping does not induce a shift of the

213: oscillator coordinate the Holstein {\it e-ph} interactions are

214: local in time and, in the semiclassical treatment, the source

215: current $j(\tau)$ is independent of the electron path. As a

216: consequence oscillator and electron coordinates appear

217: disentangled in (5) while the coupling occurs through the

218: parameter $\bar{g}$.

219: 

220: \section*{IV. The Partition Function}

221: 

222: The quantum statistical partition function $Z_T$ is derived by

223: integrating (5) after imposing periodicity conditions, $\beta$ is

224: the period, both on the electron and oscillator paths:

225: 

226: \begin{eqnarray}

227: Z_T &=&\,\int dx <{x}(\beta)|{x}(0)> =\, Z_{el} \times Z_{osc}

228: \nonumber

229: \\

230: Z_{el} &=&\, \oint Dx(\tau) exp \Bigl[- \int_0^{\beta}d\tau

231: E(x(\tau)) \Bigr] \nonumber

232: \\

233: Z_{osc} &=&\, \oint Du(\tau) exp\Bigl[- \int_0^{\beta} d\tau

234: O(u(\tau)) \Bigr]

235: \,\nonumber \\

236: \end{eqnarray}

237: 

238: where  $\oint Dx(\tau)$ and $\oint Du(\tau)$ are the functional

239: measures of integration.

240: 

241: The electronic contribution $Z_{el}$ is computed by expanding the

242: paths in Fourier components

243: 

244: \begin{eqnarray}

245: x(\tau)&=&\,x_o + \sum_{m=1}^{M_F}\bigl( r_m \cos(\nu_m \tau) +

246: s_m \sin(\nu_m \tau) \bigr)\, \nonumber

247: \\

248: &&r_m =\,2\Re x_m \, \nonumber

249: \\

250: &&s_m =\, -2\Im x_m \, \nonumber

251: \\

252: &&\nu_m =\,2m\pi/\beta \, \nonumber

253: \\

254: \end{eqnarray}

255: 

256: 

257: and taking the following measure of integration

258: 

259: \begin{eqnarray}

260: && \oint Dx(\tau) \equiv {{\sqrt{2}} \over {(2

261: \lambda_{m_e})^{(2M_F+1)}}} \int_{-\infty}^{\infty}{dx_o} \,

262: \nonumber

263: \\

264: &\times&

265: \prod_{m=1}^{M_F}  (2\pi m)^2 \int_{-\infty}^{\infty} dr_m

266: \int_{-\infty}^{\infty} ds_m \, \nonumber

267: \\

268: & &\lambda_{m_e}=\,\sqrt{\pi \hbar^2 \beta/m_e}\, \nonumber

269: \\

270: \end{eqnarray}

271: 

272: The cutoffs over the Fourier coefficient integrations have to

273: ensure proper normalization of the kinetic term in absence of

274: hopping processes. Thus $Z_{el}$ transforms into

275: 

276: \begin{eqnarray}

277: Z_{el}&\simeq& \,{{\sqrt{2}} \over {(2 \lambda_{m_e})^{(2M_F+1)}}}

278: \int_{-\Lambda/2}^{\Lambda/2}{d{x}_o} \prod_{m=1}^{M_F}

279: \int_{-\Lambda}^{\Lambda} d{r}_m \int_{-\Lambda}^{\Lambda} d{s}_m

280: \, \nonumber

281: \\ &\times& \exp\biggl(- {{\pi^{3}} \over \lambda^2_{m_e}} \sum_{m=1}^{M_F}

282: m^2({r}^2_m + {s}^2_m) - \int_0^{\beta}d\tau h^e(x(\tau)) \biggr)

283: \, \nonumber

284: \\

285: \end{eqnarray}

286: 

287: with $\Lambda \propto \lambda_{m_e}$ \cite{io3} indicating that

288: large amplitude electron paths have to be selected at low

289: temperatures where the quantum effects are larger. Two Fourier

290: components, $M_F=\,2$, suffice to attain stable results as

291: $h^e(x(\tau))$ depends smoothly on the electron path. The hopping

292: term accounts for the deviation from the Gaussian behavior.

293: Numerical analysis shows that, for any choice of path parameters,

294: $h^e(x(\tau))$ decreases by decreasing the temperature but its

295: overall contribution to the electron action is substantial also at

296: low $T$.

297: 

298: Let's focus now on the anharmonic oscillator term $Z_{osc}$. The

299: oscillator path is expanded in $N_F$ Fourier components

300: 

301: \begin{eqnarray}

302: u(\tau)=\,u_o + \sum_{n=1}^{N_F}\bigl( a_n \cos(\omega_n \tau) +

303: b_n \sin(\omega_n \tau) \bigr)\, \nonumber

304: \\

305: \end{eqnarray}

306: 

307: with Matsubara frequencies $\omega_n=\,2n\pi/\beta$ and

308: coefficients $a_n\equiv \Re u_n$, $b_n\equiv -\Im u_n$ satisfying

309: the conditions $a_n =\,a_{-n}$ and  $b_n =\,-b_{-n}$. The latter

310: are consistent with the choice of real paths and simplify the

311: following $\tau$ integration of the on site potential.

312: 

313: Note that the periodicity property, $u(\tau) =\,u(\tau + \beta)$,

314: would be fulfilled also taking the very $a_n$ coefficients in (10)

315: \cite{fey}. However such a choice would not permit to fit with

316: accuracy the harmonic oscillator partition function which is known

317: exactly: $Z_h =\,\bigl[2 \sinh(\beta\omega/2)\bigr]^{-1}$. Infact,

318: in the path integral method, the harmonic partition function

319: $Z^{PI}_h$ reads

320: 

321: \begin{eqnarray}

322: Z^{PI}_h =\, \oint Du(\tau) exp\Biggl[- \int_0^{\beta} d\tau

323: \Bigl[ {M \over 2} \bigl( \dot{u}^2(\tau) + \omega^2 u^2(\tau)

324: \bigr) \Bigr] \Biggr]\, \nonumber

325: \\

326: \end{eqnarray}

327: 

328: and taking the functional measure

329: 

330: \begin{eqnarray}

331: \oint Du(\tau) \equiv & & {{\sqrt{2}} \over {(2

332: \lambda_M)^{(2N_F+1)}}} \int_{-\infty}^{\infty}{du_o} \,

333: \nonumber

334: \\

335: &\times& \prod_{n=1}^{N_F}  (2\pi n)^2 \int_{-\infty}^{\infty}

336: da_n \int_{-\infty}^{\infty} db_n \, \nonumber

337: \\

338: \end{eqnarray}

339: 

340: with $\lambda_M=\,\sqrt{\pi \hbar^2 \beta/M}$, one gets from

341: (10)-(12):

342: 

343: \begin{eqnarray}

344: Z^{PI}_h=\, {1 \over {\beta \omega}} \prod_{n=1}^{N_F}

345: {{(2n\pi)^2} \over {{(2n\pi)^2 + (\beta \omega)^2}}} \, \nonumber

346: \\

347: \end{eqnarray}

348: 

349: Instead, dropping the $b_n$ terms in (10) and (12), one would get

350: the square root of the product series in (13) which does not yield

351: a reliable fit of $Z_h$ even for large $N_F$. Note that at high

352: $T$, the condition $2n\pi \gg \beta \omega$ is fulfilled for small

353: integers $n$, hence the main contribution to $Z^{PI}_h$ is given

354: by the ${1 /{\beta \omega}}$ factor which stems from the

355: $\int{du_o}$ in (12). This is consistent with the expectation that

356: high $T$ paths are well approximated by their $\beta$-averaged

357: value $u_o$ whereas fluctuation effects become increasingly

358: relevant towards the low $T$ regime in which $N_F$ rapidly grows.

359: $N_F$ clearly varies also with the oscillator energy while the

360: shape of $N_F(\omega,T)$ may differ according to the harmonic

361: function ($Z_h$, harmonic free energy or specific heat) one

362: chooses to fit.

363: 

364: The anharmonic partition function $Z_{osc}$ in (6) can be worked

365: out analytically using (10),(12) and performing the time

366: integration of the oscillator functional $O(u(\tau))$. This

367: permits to get an insight into the role of the non linear terms.

368: The lenghty calculation yields:

369: 

370: 

371: \begin{eqnarray}

372: &&Z_{osc}=\, {{\sqrt{2}} \over {(2 \lambda_M)^{(2N_F+1)}}}

373: \int_{-\infty}^{\infty}{du_o} \exp\bigl( -\beta \bar g u_o -

374: \kappa u_o^2 -\beta\delta u_o^4/4 \bigr)\, \nonumber \\ && \times

375: \prod_{n=1}^{N_F} (2\pi n)^2 \int_{-\infty}^{\infty} da_n

376: \exp\Biggl[-\sum_{n=1}^{N_F} \Bigl[ (\gamma_n + 3\beta\delta

377: u_o^2/4)a_n^2  \nonumber

378: \\

379: &&+ {{\beta\delta u_o} \over 2} \sum_{m=1}^{N_F}c({n,m}) +

380: {{\beta\delta} \over {16}} \sum_{m,p=1}^{N_F}d({n,m,p}) \Bigr]

381: \Biggr]\, \nonumber \\ && \times \int_{-\infty}^{\infty} db_n

382: \exp\Biggl[-\sum_{n=1}^{N_F} \Bigl[ (\gamma_n + 3\beta\delta

383: u_o^2/4)b_n^2  \nonumber

384: \\

385: &&+ {{3\beta\delta u_o} \over 2} \sum_{m=1}^{N_F}e({n,m}) +

386: {{\beta\delta} \over {16}} \sum_{m,p=1}^{N_F}\bigl({{6}}f({n,m,p})

387: - g({n,m,p}) \bigr) \Bigr] \Biggr]\, \nonumber \\ \nonumber

388: \\

389: &&\kappa=\,\pi

390: (\beta\omega)^2/2\lambda_M^2 \, \nonumber

391: \\

392: && \gamma_n=\, \pi \bigl((2\pi n)^2 + (\beta\omega)^2 \bigr)/4

393: \lambda_M^2 \, \nonumber

394: \\

395: && c({n,m})=\,a_n a_m (a_{n+m}+a_{n-m}) \nonumber

396: \\

397: && d({n,m,p})=\,a_n a_m a_p

398: (a_{n+m+p}+a_{n-m+p}+a_{p-n-m}+a_{p-n+m}) \nonumber

399: \\

400: && e({n,m})=\,a_n b_m (b_{n+m}- b_{n-m}) \nonumber

401: \\

402: && f({n,m,p})=\,a_n a_m b_p

403: (b_{n+m+p}+b_{n-m+p}+b_{p-n-m}+b_{p-n+m}) \nonumber

404: \\

405: && g({n,m,p})=\,b_n b_m b_p

406: (b_{n+m+p}-b_{n-m+p}+b_{p-n-m}-b_{p-n+m}) \nonumber

407: \\

408: \end{eqnarray}

409: 

410: In $c({n,m})$ the coefficients $a_{j}$ $(j=\, n+m, n-m)$ vanish

411: if: $j \leq 0$ or $j > N_F$. In $d({n,m,p})$, $a_k \neq 0 \,\,

412: (k=\,n+m+p, \,\, n-m+p, \,\, p-n-m, \,\, p-n+m)

413: \Longleftrightarrow 1 \leq k \leq N_F$. Analogous conditions hold

414: for the coefficients $b_{j\, (k)}$ in the $e({n,m})$, $f({n,m,p})$

415: and $g({n,m,p})$ functions.

416: 

417: Note that the effective {\it e-ph} coupling $\bar{g}$ is

418: associated only to the $\tau$-independent component $u_o$, that is

419: to the $\beta$-averaged displacement path \cite{io5}. This follows

420: from the fact that the perturbing source current is non retarded

421: in the Holstein model as a consequence of the local nature of the

422: e-ph interaction.

423: 

424: The quartic potential induces a strong mixing of the Fourier

425: components of the path that highly complicates the numerical

426: problem. Thus the value of $N_F$ appears to be crucial in the

427: computation. We determine $N_F(\omega,T)$ by fitting (with an

428: accuracy of $2 \cdot 10^{-2}$) the exact harmonic free energy

429: ($F_h=\,-\ln(Z_h)/\beta$), through the path integral harmonic free

430: energy $F_h^{PI}$ obtained from (13). As an example, for the

431: oscillator with $\omega=\,20meV$, $N_F(T=\,10K)=\,59$ and

432: $N_F(T=\,200K)=\,8$.

433: 

434: Inspection of (14) offers the key to perform reliable path

435: integrations according to the sign of the $\phi^4$ potential. At

436: high temperature, a large contribution to $Z_{osc}$ is expected to

437: come from the paths having $u_o$ which maximizes $\exp\bigl(\kappa

438: f(u_o)\bigr)$ with $f(u_o)=\, -(a u_o + u_o^2 + b u_o^4), \, a

439: \equiv \,\beta \bar g/\kappa; \,\, b \equiv\,\beta\delta/4\kappa$.

440: In general, we find that for a hard (soft) potential, the

441: $du_o$-integration has to be carried out along the $u_o < 0 \,(u_o

442: > 0)$ axis, with cutoff $|u_o|_{max} \sim 0.6/\sqrt{\kappa}$. This

443: permits to include the set of paths which mainly contribute to the

444: euclidean oscillator action. On the Fourier coefficients

445: integrals, $\int da_n \,\int db_n$, we set the cutoffs

446: $|a_n|_{max}, \, |b_n|_{max} \sim 0.6/\sqrt{\gamma_n}$ both for

447: hard and soft potentials thus achieving numerical convergence and

448: correct computation of the Gaussian integrals once the non

449: linearities are switched off. It turns out, see the definitions in

450: (14), that the cutoffs on the oscillator path integration are

451: increasing functions of temperature ($\propto \sqrt{T}$)

452: consistently with the physical expectations of large amplitude

453: displacements at high $T$. As the path displacements encounter an

454: upper limit due to the cutoffs, $u(\tau) \leq |u_o|_{max} +

455: 2\sum_{n=1}^{N_F}|a_n|_{max}$, the distribution of on site

456: potentials has a lower limit also in the case of soft and

457: attractive non linearities. This avoids numerical divergences and

458: makes the problem physically meaningful.

459: 

460: 

461: \section*{V. Results}

462: 

463: We test the relevance of the non linearities on the equilibrium

464: thermodynamics of the system and present the calculation for the

465: heat capacity in the intermediate to low temperature range.

466: 

467: Figures 1 show the behavior of a low energy ($\omega=\,20meV$)

468: oscillator without ($g=\,0$) and with ($g=\,2,4$) coupling to the

469: electronic subsystem in the adiabatic regime: $t/ \omega =\,5$.

470: Figs.1(a),(b) assume an anharmonic potential with positive quartic

471: force constant $\delta = 10^3, 10^4 meV \AA^{-4}$ respectively.

472: Also the harmonic heat capacity is reported on for comparison. The

473: hard potential lifts the free energy over the harmonic values with

474: a more pronounced enhancement at increasing $T$ and for larger

475: $\delta$. The effect on the free energy second derivative is

476: however scarce and essentially consists, see Fig.1(a), in a slight

477: increase ( decrease) of the heat capacity at low (high)

478: temperatures. The reduction of the heat capacity (with respect to

479: the harmonic plot) at intermediate and high $T$ is more evident in

480: Fig.1(b) where the quartic force constant is larger. This result

481: is in accordance with diagrammatic perturbative treatments of the

482: anharmonic crystals which predict negative contributions to the

483: constant volume specific heat arising from positive force

484: constants in the quartic potential \cite{martin,io90}. The {\it

485: e-ph} coupling strength also tends to decrease the oscillator heat

486: capacity but the overall effect is small as the insets in the

487: figures show: infact the electronic term dominates the total heat

488: capacity $C$ and the oscillator contribution is not

489: distinguishable in the plots of the $C$ over temperature ratios

490: versus $T$. The low temperature upturn is due to the large

491: electron energy term in (9) and precisely ascribable to the

492: feature of the variable range (on the $\tau$ scale) of the

493: electron hopping, captured by the path integral formalism

494: \cite{io3}.

495: 

496: The cases of a soft on site potential are reported on in Fig.1(c)

497: with $\delta = -10^3 meV \AA^{-4}$ and Fig.1(d) with $\delta =

498: -10^4 meV \AA^{-4}$. The characteristic potential parameter is

499: $M\omega^2/|\delta| \simeq 0.5$ and $0.05 \AA^2$, respectively.

500: The potential $V(u(\tau))=\, {M}\omega^2 u(\tau)^2/2 +

501: {{\delta}}u(\tau)^4/4$ is attractive for those paths such that

502: $u(\tau)^2 > 2M\omega^2/|\delta|$.

503: 

504: At any $T$, we integrate over a distribution of time dependent

505: potentials. Thus, at a given lattice site, the electron may

506: experience an attractive or repulsive scattering centre according

507: to the size of the atomic path. As an example, at $T=\,200K$, we

508: get a maximum path $u_{max}$ such that $u_{max}^2 \sim 0.15

509: \AA^2$. This guarantees that $V(u(\tau))$ is generally repulsive

510: in Fig.1(c) and attractive for a broad class of paths in Fig.1(d).

511: 

512: The effects on the oscillator heat capacity are twofold: i) the

513: soft potential enhances the heat capacity mainly in the low $T$

514: range with respect to the hard potential and this feature is much

515: more pronounced in Fig.1(d); ii) the trend of the {\it e-ph}

516: coupling is opposite to that observed in Figs.1(a),(b): now by

517: increasing the $g$ values one gets higher heat capacities although

518: the size of this effect is small on the scale of the electronic

519: terms in the adiabatic regime as revealed by the $C/T$ plots in

520: the insets.

521: 

522: Let's come to the antiadiabatic regime ($t/ \omega =\,0.5$)

523: discussed in Figures 2, where a low harmonic energy ($\omega

524: =\,10meV$) is assumed to emphasize the size of the anharmonicity

525: together with a very narrow electron band. A hard on site

526: potential with $\delta = 10^3 meV \AA^{-4}$ is taken in Fig.2(a):

527: the shape of the oscillator heat capacity signals the effects of

528: the non linearities which flatten the curve at intermediate $T$

529: and enhance the heat capacity also at low $T$ with respect to the

530: corresponding case of Fig.1(a). As the Debye temperature is now

531: smaller (than in Fig.1(a)) the hard anharmonicity decreases the

532: constant volume heat capacity with respect to the harmonic plot

533: over a broader temperature range.

534: 

535: The {\it e-ph} coupling plays a minor role also in antiadiabatic

536: conditions. The anharmonic contribution is visible in the total

537: heat capacity as the inset makes evident although the dominant

538: electronic feature persists in the low $T$ upturn of $C/T$. In

539: Fig.2(b), we assume $g =\,2$ and consider two cases of soft

540: potential: the oscillator anharmonicity becomes relevant and such

541: to modify the shape of the anomalous upturn in the total heat

542: capacity. An enhancement of the $C/T$ values is observed at

543: intermediate and low $T$ and, in the case of the largest

544: $|\delta|$ generating a soft attractive potential, the oscillator

545: heat capacity $C_{osc}$ yields an upturn in $C_{osc}/T$

546: independently of the electronic term.

547: 

548: \section*{VI. Conclusions}

549: 

550: We have studied the path integral of the one dimensional non

551: linear Holstein model in which a set of dispersionless oscillators

552: provides the environment for the electron. The model is

553: semiclassical as the lattice displacements are treated classically

554: while the electron operators are thermally averaged over the

555: ground state Hamiltonian. The {\it e-ph} coupling of the model is

556: local and generates a perturbing current which linearly depends on

557: the oscillator path amplitude $u(\tau)$ where $\tau$ is the time

558: (or inverse temperature) of the Matsubara Green functions

559: formalism. The anharmonicity on the lattice site is modelled

560: through a $\phi^4$ potential that may result attractive for a set

561: of displacement paths in the case of a negative quartic force

562: constant (soft potential). We have derived the path integral of

563: the interacting system and computed the total partition function

564: selecting, as a function of the temperature ($T \leq 200K$), both

565: the electron and oscillator paths which yield the largest

566: contribution to the action. While quantum electron paths have

567: increasing amplitudes at decreasing temperatures, the atomic

568: displacements are growing functions of $T$. This relevant physical

569: feature is accounted for in our model as the cutoffs on the

570: electron path integration are proportional to the electron thermal

571: wavelength whereas,  on the atomic path integration, we find

572: cutoffs proportional to $\sqrt{T}$.

573: 

574: The oscillator partition function includes the effect of the

575: coupling to the electron subsystem but the on site anharmonicities

576: play a major role mainly when the potential is soft and the

577: harmonic energy is low. Among the thermodynamic properties we have

578: chosen to present the heat capacity $C$ in view of the upturn in

579: the $C/T$ behavior due to the low $T$ electron hopping tuned by

580: the value of the overlap integral in the Hamiltonian model. The

581: computation is highly time consuming especially because of the

582: strong mixing of the path Fourier components generated by the non

583: linear potential.

584: 

585: In general we find that: i) in the case of a hard quartic

586: potential, switching on the {\it e-ph} coupling leads to lower the

587: free energy and its second derivative, ii) when the quartic

588: potential is soft, {\it e-ph} coupling and anharmonicity act

589: sinergically enhancing the thermodynamic functions.

590: 

591: These results can be understood on general physical grounds.

592: Infact, a hard quartic potential shifts the characteristic phonon

593: frequency upwards thus hardening the spectrum and broadening the

594: size of the quasiparticle. If the {\it e-ph} coupling is enhanced

595: (independently of the on site anharmonicity) the magnitude of the

596: source which causes the lattice distortion becomes larger. This

597: further hardens the vibrations and leads to decrease the

598: anharmonic heat capacity over the whole temperature range.

599: Instead, when the quartic potential is soft the phonon frequency

600: is lowered and the oscillator potential well is more flexible. In

601: this case, larger {\it e-ph} coupling strengths favour the

602: self-trapping of quasiparticles with heavier effective masses.

603: This is physically equivalent to soften the phonon spectrum and

604: enhance the heat capacity.

605: 

606: The electron contribution to the heat capacity is dominant in the

607: adiabatic regime whereas antiadiabatic systems are expected to

608: present significant anharmonic corrections. Infact, in the

609: antiadiabatic regime the quasiparticle is a small size object on

610: the lattice scale and the electron energy associated with the

611: overlap integral is small. Thus, this regime proposes a physical

612: picture in which the electron hardly hops from site to site and

613: its effective mass becomes heavier on the scale of the atomic

614: mass. But a potential well generated by "lighter oscillators" is

615: more sensitive to on site anharmonic effects.

616: 

617: In particular, soft potentials increase the heat capacities over

618: the harmonic values and reinforce the upturn in the $C/T$ versus

619: $T$ plots when the on site anharmonicity is such to produce

620: attractive potentials for a set of lattice displacement paths.

621: Since the path amplitudes are larger at high $T$, soft attractive

622: potentials induce rapidly increasing phonon heat capacities at

623: growing $T$ as shown in Fig.2(b).

624: 

625: Thus our path integral investigation and the thermodynamical

626: results point to a complex role of the lattice anharmonicities in

627: the one dimensional Holstein model and suggest that on site

628: potentials may be experienced as attractive or repulsive according

629: to the temperature and the amplitude of the atomic path. Such

630: potentials may provide scattering centres generating local

631: disorder whose effect on the system thermodynamics is superimposed

632: to the disorder induced by the hopping of electrons with variable

633: range.

634: 

635: 

636: \begin{figure} \vspace*{18truecm}

637: \caption{(Color online) Anharmonic Oscillator Heat Capacities

638: versus temperature for three values of {\it e-ph} coupling $g$ and

639: oscillator energy $\omega =\,20meV$. (a) Hard potential with force

640: constant $\delta = 10^3 meV \AA^{-4}$; (b) $\delta = 10^4 meV

641: \AA^{-4}$; (c) soft potential with $\delta = -10^3 meV \AA^{-4}$;

642: (d) $\delta = -10^4 meV \AA^{-4}$. The harmonic heat capacity is

643: plotted in (a) and (b) for comparison. The insets show the {\it

644: Total (electronic plus anharmonic oscillator) Heat Capacity over

645: temperature} ratios in the adiabatic regime $t/ \omega =\,5$.}

646: \end{figure}

647: 

648: \begin{figure} \vspace*{12truecm}

649: \caption{(Color online) Anharmonic Oscillator Heat Capacities

650: versus temperature in the antiadiabatic regime $t/ \omega =\,0.5$

651: and oscillator energy $\omega =\,10meV$. The harmonic heat

652: capacity is also plotted. (a) Hard potential force constant

653: $\delta = 10^3 meV \AA^{-4}$ with three values of {\it e-ph}

654: coupling $g$. (b) Two soft potential force constants at fixed {\it

655: e-ph} coupling. The insets show the {\it Total (electronic plus

656: anharmonic oscillator) Heat Capacity over temperature} ratios. The

657: electronic contribution is plotted separately for comparison. }

658: \end{figure}

659: 

660: 

661: 

662: 

663: \begin{references}

664: \bibitem{kaat}

665: G.A. Kaat, K. Flensberg, cond-mat/0411173

666: \bibitem{mah}

667: L.M.Woods, G.D.Mahan, Phys.\ Rev.\ B {\bf 61}, 10651 (2000).

668: \bibitem{han}

669: K.Hannewald, V.M.Stojanovi\'{c}, J.M.T.Schellekens, P.A.Bobbert,

670: G.Kresse, J.Hafner, Phys.\ Rev.\ B {\bf 69}, 075211 (2004).

671: \bibitem{alexan}

672: S.S. Alexandre, E. Artacho, J.M. Soler, H. Chacham, Phys.\ Rev.\

673: Lett. {\bf 91}, 108105 (2003).

674: \bibitem{sch}

675: T. Schneider, R. Khasanov, K. Conder, H. Keller, J. Phys. Condens.

676: Matter {\bf 15}, L763 (2003).

677: \bibitem{lan}

678: A. Lanzara, P.V. Bogdanov, X.J. Zhou, S.A. Kellar, D.L. Feng, E.D.

679: Lu, T. Yoshida, H. Eisaki, A. Fujimori, K. Kishio, J.-I.

680: Shimoyama, T. Noda, S. Uchida, Z. Hussain, Z.-X.Shen, Nature {\bf

681: 412}, 510 (2001).

682: \bibitem{chris}

683: Y. Zolotaryuk, P.L. Christiansen, J.J. Rasmussen, Phys.\ Rev.\ B

684: {\bf 58}, 14305 (1998).

685: \bibitem{voulga}

686: N.K. Voulgarakis, G.P. Tsironis, Phys.\ Rev.\ B {\bf 63}, 14302

687: (2001).

688: \bibitem{zieg}

689: D.Schneider, K.Ziegler, cond-mat/0504688.

690: \bibitem{pietro}

691: L.Boeri, G.B.Bachelet, E.Cappelluti, L.Pietronero, Phys.\ Rev.\ B

692: {\bf 65}, 216501 (2002).

693: \bibitem{cava}

694: T.Yildirim, O. G\"ulseren, J. W. Lynn, C. M. Brown, T. J. Udovic,

695: Q. Huang, N. Rogado, K. A. Regan, M. A. Hayward, J. S. Slusky, T.

696: He, M. K. Haas, P. Khalifah, K. Inumaru, and R. J. Cava, Phys.\

697: Rev.\ Lett. {\bf 87}, 037001 (2001).

698: \bibitem{fehi}

699: R.P.Feynman, A.R.Hibbs, {\it Quantum Mechanics and Path Integrals}

700: Mc Graw-Hill, New York (1965).

701: \bibitem{kleinert}

702: H.Kleinert, {\it Path Integrals in Quantum Mechanics, Statistics

703: and Polymer Physycs} World Scientific Publishing, Singapore

704: (1995).

705: \bibitem{eck}

706: U.Eckern, M.J.Gruber, P.Schwab, cond-mat/0503369.

707: \bibitem{io5}

708: M.Zoli, Phys.\ Rev.\ B {\bf 71}, 184308 (2005).

709: \bibitem{holst1}

710: T.Holstein, Ann.\ Phys.\ (N.Y.)  {\bf 8}, 343 (1959).

711: \bibitem{raedt2}

712: H.De Raedt, A.Lagendijk, Phys.\ Rev.\ B {\bf 30}, 1671 (1984).

713: \bibitem{io98}

714: M.Zoli, Phys.\ Rev.\ B {\bf 57}, 10555 (1998).

715: \bibitem{atin}

716: M.Zoli, A.N.Das, J.Phys.: Cond. Matter {\bf 16}, 3597 (2004).

717: \bibitem{alekor}

718: A.S.Alexandrov, P.E.Kornilovitch, Phys.\ Rev.\ Lett. {\bf 82}, 807

719: (1999).

720: \bibitem{holst}

721: T.Holstein, Ann.\ Phys.\ (N.Y.)  {\bf 8}, 325 (1959).

722: \bibitem{kenkre}

723: V.M.Kenkre, H.-L.Wu, I.Howard,  Phys.\ Rev.\ B {\bf 51}, 15841

724: (1995).

725: \bibitem{trug}

726: C.Li-Ku, S.A.Trugman, J.Bon\v{c}a, Phys.\ Rev.\ B  {\bf 65},

727: 174306 (2002).

728: \bibitem{io4}

729: M.Zoli, Phys.\ Rev.\ B {\bf 70}, 184301 (2004).

730: \bibitem{feynman}

731: R.P.Feynman, Phys. Rev. {\bf 97}, 660 (1955).

732: \bibitem{hamann}

733: D.R.Hamann, Phys.\ Rev.\ B {\bf 2}, 1373 (1970).

734: \bibitem{io3}

735: M.Zoli, Phys.\ Rev.\ B {\bf 67}, 195102 (2003).

736: \bibitem{fey}

737: R.P.Feynman, {\it Statistical Mechanics: A Set of Lectures}

738: W.A.Benjamin, Inc. (1972).

739: \bibitem{martin}

740: D.L.Martin, Can.J.Phys. {\bf 65}, 1104 (1987).

741: \bibitem{io90}

742: M.Zoli, Phys.\ Rev.\ B {\bf 41}, 7497  (1990).

743: \end{references}

744: 

745: \end{document}

746: