Polaron formation for a non-local electron-phonon coupling: A variational wave-function study C. A. Perroni,1,2 E. Piegari,2 M. Capone,3 and V. Cataudella1,2 1Coherentia-INFM, UdR di Napoli, via Cinthia 80126 Napoli (Na), Italy 2Dipartimento di Fisica, Universit´a di Napoli “Federico II”, Italy 3Enrico Fermi Center, Rome, Italy (February 2, 2008) 4 0 We introduce a variational wave-function to study the polaron formation when the electronic 0 transfer integral depends on the relative displacement between nearest-neighbor sites giving rise 2 to a non-local electron-phonon coupling with optical phonon modes. We characterize the polaron n crossover by analysing ground state properties such as the energy, the electron-lattice correlation a function, the average phonon occupation and the quasiparticle spectral weight. Variational results J are found in good agreement with numerical exact diagonalization of small clusters,and follow the 8 correct perturbative result at weak coupling. We determine the polaronic phase diagram and we find that thetendency towards strong localization is hindered from thepathological sign change of ] the effectivenext-nearest-neighborhopping. l e - r t s . t a I. INTRODUCTION m - d Significantelectron-phonon(el-ph)interactionshavebeenexperimentallydetectedinmanymaterialsofwide inter- n est, like manganites,1 fullerenes,2, carbon nanotubes,3,4 and cuprates.5 In many of these cases, the el-ph interaction o gives rise to polaronic features. A polaronic state results in fact when the electrons are strongly coupled to lattice c distortion,thereforeincreasingtheir effective mass andleadingto astate with lowmobility. Increasingthe el-phcou- [ pling,the spatialextensionofthe lattice deformationdecreases6 andthe polaroncanvaryits size fromlarge tosmall. 1 The single polaronproblem of one electron interacting with the lattice degrees of freedom has been studied in detail, v and allowed us to understand in the detail the physics leading to the formation of polaronic states. In particular it 5 has been shownthat the self-trapping process,which lead to the formationof polarons,is not a phase transition, but 2 1 just a continuous crossoverwith no brokensymmetry.7 In the caseof the Holstein model,8 where quantumvibrations 1 interact locally with the electrons, the crossover from large to small polaron has been extensively studied by several 0 numericaltechniques9–15 andvariationalapproaches.16–18 InparticularallthegroundstatepropertiesoftheHolstein 4 model can be described with great accuracy by a variational approach18 based on a linear superposition of Bloch 0 states that describe weak and strong coupling polaron wave functions. / t The case of non-local interactions, that in general are also present in real materials, is much less understood. The a couplingwithacousticalphononshasbeenstudiedinordertoexplaintheanomaloustransportpropertiesofnon-local m excitations,likesolitonsandpolarons,invarious1Dsystems.19–22 Inparticularthetight-bindingSu-Schrieffer-Heeger d- (SSH)model19 wasintroducedtoexplainthetransportpropertiesofquasione-dimensionalpolymersaspolyacetylene n where the CH monomers form chains of alternating double and single bonds. In this case the localization is due o to a large shrink of two particular bonds and corresponding large hopping integral between the sites. As a result, c the hopping between the two occupied sites and the surrounding ones is reduced resulting in a tendency towards : v localization. i Our purpose here is to examine the single polaron formation in a model where both (Holstein) local and (SSH) X non-localel-phinteractionsarepresent. Duetothecomplexityofthemodel,westarttheanalysisusingaperturbative r approach that, although it can not capture the full multiphononic nature of the polaron, it has proved a remarkably a useful tool in understanding the el-ph physics.12,22 In particular we characterize the ground state properties of the system evaluating the energy, the electron-lattice correlation function and the quasiparticle spectral weight, in order toprovidesignaturesofthe polaronformation. Inthelimitwhenlocalel-phinteractionsaremuchstrongerorweaker than non-local el-ph interactions our model reduces to the standard Holstein model and to the SSH model with a dispersionless phonon spectrum, respectively. Then we start an accurate analysis of the non-local limit case, being thiscasenotyetfullyexamined. Recently,thefullyadiabaticregimeofthismodelhasbeenusedtoexplainchangesin carbon-nanotubelengthasafunctionofchargeinjection.4 Furthermorethenon-localcasehasbeenpreviouslystudied byoneofususingexactdiagonalizationofsmallclustersuptofourlatticesites,whereananomalousopticalabsorption has been identified.12 In particular in Ref. [ 12] it is shown that the strong-coupling solution is characterized by an 1 unphysical sign change of the effective next-nearest-neighbor hopping which is missing when acoustical phonons are considered.21 In this work we improve the previous numerical analysis considering a six-site lattice and introduce a variational wave-function to investigate the thermodynamic limit of the system. The variational approach is based on a linear superposition of Bloch states that provide an excellent description of the lattice deformations on left and right bond ofthepolaron,respectively. Thewave-functioncloselyresemblesavariationalstatepreviouslyproposedforthestudy of the Holstein model18 and for the SSH case it allows to describe polaron features in good agreement with exact numerical diagonalization results. The variational approach recovers the pathological behavior of the effective next- nearest-neighbor hopping pointing out that the non-physical region of parameters always prevents a strong localized solution. We also explicitly show that, when the phonon frequency is not really small, the considered non-local SSH interaction supplies a tendency to localize for the single carrier which can be more effective than the Holstein localization. The scheme of the paper is the following. In Sec. II we present the model and set down the notation. In Sec. III wediscuss perturbativecalculationsshowingthe roleofSSH el-phcoupling withrespectto the Holsteincontribution. Sec. IV is devoted to the presentation of the variational method in the limit of non-local el-ph interactions and its comparison with the exact diagonalization results. Sec. V reports our concluding remarks. II. THE MODEL In extremely generalterms, the interactionbetweenelectronand harmoniclattice deformations is describedby the Hamiltonian p2 x K x = c† t ( x )c + i + i i,j j + e ( x )c† c , (1) H i,σ i,j { k} j,σ 2M 2 i { k} i,σ i,σ i,j,σ i i,j i,σ X X X X wherec† (c )isthe fermioncreation(destruction)operator,σ is thespinindex,t ( x )isthe electronictransfer i,σ i,σ i,j { k} integral for fixed lattice deformations x , M is the ionic mass, K is the spring constant matrix, and e ( x ) is k i,j i k { } { } thelocalenergyoftheelectron. Forsmalldeviationsfromtheequilibriumpositionsofthe latticewecanapproximate t ( x ) and e ( x ) to be linear functions of the lattice displacements x obtaining a general model with el-ph i,j k i k k { } { } { } interactions. In particular, limiting the hopping to nearest-neighbor sites of a linear chain, we make the assumption t ( x )= t+α(x x ) (2) n+1,n k n+1 n { } − − typically employed for the derivation of the el-ph SSH interaction term, and e ( x )=α x (3) i k 1 i { } generallyused in orderto deduce the localel-ph Holstein interaction. If spinless electrons and dispersionlessEinstein phonons are considered, the model becomes = t (c†c +c† c )+ω a†a +H , (4) H − i i+1 i+1 i 0 i i int i i X X where H is int H =gω (c†c +c† c )(a† +a a† a )+g ω c†c (a†+a ), (5) int 0 i i+1 i+1 i i+1 i+1− i − i 1 0 i i i i i i X X with a† (a ) the phonon creation (destruction) operator and ω the quantum of vibrational energy per site. The i i 0 quantity g = α/ 2Mω3 is the SSH coupling that we mainly discuss in this work, while g = α / 2Mω3 is the 0 1 1 0 Holsteinlocalelectron-phononcoupling. Westudythecouplingofasingleelectrontolatticedeformationsusingunits p p such that the lattice spacing a=1 and h¯ =1. III. PERTURBATION THEORY Weak-coupling perturbation theory in the electron-phonon coupling has proved a remarkably useful tool in under- standing the el-ph physics. Besides the obvious ability to describe the weakly interacting regime, the perturbative 2 approach has in fact provided some guidelines to understand the conditions for polaron formation in the Holstein model. More explicitly, the polaron crossover occurs around the coupling value for which the perturbative approach breaks down.12 Here we focus on the case of one electron in a one-dimensional chain. If the el-ph terms are smaller than both the hopping term and the bare phonon term (g,g t˜,1 and t˜= t/ω ), they can be treated as perturbations of the 1 0 ≪ unperturbed Hamiltonian H =H +H . 0 kin ph The second-order correction to the energy of the ground state is given by 1+2t˜ 1+4t˜ 1 ∆E(0)= g2 − g2 , (6) − t˜2p !− 1 1+4t˜ while the perturbative correctionto the free band ε = 2tcos(k) is repoprted in Appendix A. k − We note that for fixed values of the coupling constants g and g the two contributions (SSH-like and Holstein) 1 have different behaviors as functions of the inverse adiabatic ratio t˜. In particular, as shown in Fig. 1, the Holstein contributiontothegroundstateenergyisalwayslower(forg =g )thantheSSHonewhent˜<t˜ witht˜ =4+3√2. 1 w w In other words, when the phonon frequency are not really small, the SSH el-ph coupling is more effective than the Holstein one. The reduced effect of the Holstein el-ph coupling when t˜is small (anti-adiabatic regime) pushes the polaron crossover to larger values of the coupling λ= g2/2ω t as the phonon frequency is increased. Actually, while 0 λ > 1 is the condition for the polaron crossover in the adiabatic regime t˜ 1, in the antiadibatic regime t˜ 1, it ≫ ≪ has been shown that the crossover occurs when α2 = g2/ω2 1, i.e., for λ 1.12,14,15 Recently it has been shown 0 ≃ ≫ that this important role of the degree of adiabaticity is not limited to the single polaron problem, but it also extends to finite dennsities.23 Since polaron formation is not a phase transition and occurs without symmetry breaking, different criteria can be establishedto define the crossovervalues ofthe coupling constantswhichmarkthe polaronicregime. Inthe following we compute some physical quantities which have been often introduced to characterize the polaron crossover. The averagephonon occupation number N = 1 a†a is given by ph Nh i i ii 2λ (1+2t˜P) (1+2t˜) N = 1 +2λ t˜ , (7) ph t˜ " 1+4t˜− # 1 (1+4t˜)3/2 whereλ=g2ω /2tandλ =g2ω /2t. FromEqp. (7)itturnsoutthatthephononnumber,asthe groundstateenergy, 0 1 1 0 is more affectedby the SSH coupling when ω exceeds a givenvalue. In particularfor t˜ 2 (i.e. ω /t 0.5)the SSH 0 0 ≤ ≥ contribution is always higher than the Holstein one. Otherquantitiesofgreatinteresttocharacterizethepolaronformationaretheelectron-latticecorrelationfunctions. In particular we consider the correlation function χ = c†c (a† +a ) between the electronic density on a site i,δ h i i i+δ i+δ i i and the lattice displacement on site i+δ, which measures the entanglement of lattice and electronic degrees of freedom typical of the polaronic state. After a Fourier transformation in the momentum space, at k =0 one has 2g 1 χ = k=0,δ=0 − 1+4t˜ pg g 2t˜+1 χ = (1+2t˜ 1+4t˜) 1 1 k=0,δ=1 −t˜2 − − t˜ 1+4t˜− ! p 1 1+4t˜ 1+2t˜p 1 1+2t˜ 1+2t˜ χ = 2g + 1 2g + 1 . (8) k=0,δ=2 − "t˜ 2t˜3 − 1+4t˜!#− 1" 1+4t˜ 2t˜2 − 1+4t˜!# InFig. 2weplotthecorrelationfunctionatneparest-neighbor(left)pandnext-nearest-neighbpor(right)sitesasfunctions of the inverse adiabatic ratio t˜, for fixed values of the couplings g =g =1. As expected, the value of the correlation 1 function goestozerofor largevalues oft˜,but the behaviorofthe SSH-like contribution(dashedlines)is qualitatively different from that of the Holstein ones (dot-dashed lines). The last quantity we consider is the quasiparticle spectral weight Z(k) = (1 ∂Σ(k,ω) )−1, which measures − ∂ω |ω=ε(k) the renormalization of the electron Green’s function due to the el-ph interaction. The second-order perturbative self-energy Σ(k,ω) is given in Appendix A. Even if the polaronic regime cannot be attained within lowest-order perturbative approach, indications on the beginning of the polaronic crossover can be extracted from the spectral weight expression. In particular the polaron crossover is expected to be associated with a sharp reduction of this quantity as a function of the couplings. The expression of the inverse spectral weight at k =0 is given by 3 2λ (1+2t˜) (1+2t˜) Z(0)−1 =1 1 +2λ t˜ , (9) − t˜ " − 1+4t˜# 1 (1+4t˜)3/2 p while the full momentum dependence of Z(k) is reported in Appendix A. As expected the spectral weight Z(0) is a monotonically increasing function of t˜, for a fixed value of the couplings. It is interesting to note that the reduction of Z(0) due to the SSH-like contribution is more relevant of the Holstein ones for t˜<2, while in the adiabatic limit, i.e. for large value of t˜, it is very small and slow. Strictlyspeaking,perturbativecalculationsonlycorrectlycharacterizethesmallcouplingregime. Inordertoprovide a better insight on the problem of the polaron formation in the systems with non-local interactions, in the following wefocusontheSSHcontributionandsubstantiateourresultsbyanalyticvariationalcalculationsandnumericalexact data. IV. VARIATIONAL APPROACH VS. EXACT DIAGONALIZATION In this section we extend our analysis of the non-local SSH model to the whole range of el-ph couplings using two standard and well grounded techniques, a variational approach and exact diagonalization of small clusters. First we introduce the variational wave function. We consider translation-invariant Bloch states obtained by superposition of localized states centered on different lattice sites.24 These wave-functions have been introduced in order to study the polaron formation within the Holstein model where they are able to fully capture the features of the Holstein polaron.16,18 In this work we extend this kind of wave-functions to the SSH interaction model assuming 1 ψ(i) = eik·n ψ(i)(n) , (10) | k i √N | k i n X where ψ(i)(n) is defined as | k i |ψk(i)(n)i=e[Uk(i)(n)+Uk(i)(n−1)+Uk(i)(n+1)]|0iph φk(i)(m)eik·mc†n+m|0iel, (11) m X with the quantity U(i)(j) given by k g U(i)(j)= [f(i)(q)a eiq·Rj h.c.]. (12) k √N k,j q − q X (i) The phonon distribution function f (q) is chosen as k,j α(i) f(i)(q)= k,j , (13) k,j 1+2t˜β(i) [cos(k) cos(k+q)] k,j − (i) (i) with α and β variationalparameters. In Eq. (11), 0 and 0 denote the phonon and electronvacuum state, k,j k,j | iph | iel respectively, and the variational functions φ(i)(m) are assumed to be k 5 φ(i)(m)= γ(i)(j)δ , (14) k k m,j j=−5 X where γ(i)(j) are variational parameters that take into account the broadening of the electron wave-function up to k fifth neighbors. It is worth to note that traditional variational approaches to the Holstein polaron problem uses the localizedstate(11)whereonlytheon-siteoperatorU(i)(n)isapplied. Thusweintroduceintheexpressionofthetrial k wave-functionthe nearest-neighbor displacement operators U(i)(n+1) and U(i)(n 1), in order to take into account k k − the dependence of the hopping integral on the relative distance between two adjacent ions. ReflectingtheasymmetryoftheSSHcoupling(shrinkingofthebondonwhichtheelectronislocalizedandstretching of the neighboring bonds), we also define two wave-functions that provide a very good description of the lattice deformations on left and right bonds of the polaron. Naturally the left and right directions are relative to the site 4 where the presence of the electron is more probable. Thus in Eq. (11) the apex i=L,R indicates the Left (L) and Right (R) polaron wave-function, respectively. The wave-functions L and R are related as follows f(R)(q)= f(L)(q)<0 k,n − k,n f(R) (q)= f(L) (q)>0 k,n−1 − k,n−1 f(R) (q)= f(L) (q)>0 k,n+1 − k,n+1 (R) (L) φ (m)=φ ( m). (15) k k − AllthevariationalparametersaredeterminedbyminimizingtheexpectationvalueoftheHamiltonian(4)withg =0 1 onthestates(11). Eventhoughthewave-functionsLandRdescribecorrectlythelatticedeformationsoftheleftand right side of the polaron, respectively, the mean values of the Hamiltonian on these states are equal. So the relations (15) can be also viewed as those that leave unchanged the energy functional determined by one wave-function. Thesetwowave-functionscanbe improvedbyincreasingthe extensionofthe phononcontributionsinEq. (11)and of the electron terms in Eq. (14). Furthermore, they are not orthogonal and the off-diagonal matrix elements of the Hamiltonian between these two states are not zero. This allows to determine the ground-state energy by considering as trial state the linear superposition18 of the wave-functions R and L (R) (L) A Φ +B Φ ψ = k| k i k| k i , (16) k | i A2 +B2+2A B S k k k k k where Φ(L) and Φ(R) are the normalized wavep-functions L and R weighted by the coefficients A and B and | k i | k i k k S = Φ(L) Φ(R) (17) k h k | k i is the overlap factor. The wave-function (16) correctly describes the properties of the lattice deformations on both the sides of the polaronand we will find that it is in very good agreementwith the results derivedby the exact diag- onalizations on a chain of 6 sites. Furthermore the variational approach involves a number of variational parameters that does not depend on the length of chain, so it allows to study the thermodynamic limit of the system. The minimization procedure is performed in two steps. First the energies of the left and right wave-functions are separately minimized, then these wave-functions are used in the minimization procedure of the quantity E = k ψ H ψ / ψ ψ with respect to A and B defined in (16).18 Exploiting the equality k k k k k k h | | i h | i ψ(L) H ψ(L) = ψ(R) H ψ(R) =ε , (18) h k | | k i h k | | k i k we obtain ε S E E S ε k k kc kc k k E = − −| − |, (19) k 1 S2 − k where E = Φ(L) H Φ(R) is the off-diagonal matrix element, and A = B . The matrix elements between the kc h k | | k i | k| | k| states ψ(R) and ψ(L) contained in Eq. (19) are reported in Appendix B. k k The total energy functional (19) is minimized with respect to the variational parameters and the optimal ground state energyis plotted inFig. 3 for asix-site lattice andtwo differentvalues ofthe inverseadiabaticparametert˜. We also study the thermodynamic limit and find energy curves very close to those of the finite system. In order to test the validity of our variational approach (VA), we perform exact numerical calculations on small clusters by means of the Lanczos algorithm. We improve the previous exact diagonalization (ED) analysis of the model, investigating small clusters up to six sites.12 As shown in Fig. 3, each variational and exact numerical curve exhibits a kink with increasing the el-ph coupling. We have checked that at these couplings the effective next-nearest-neighbor hopping changessignopening anunphysicalregionofthe parameters. The agreementbetweennumericaldataandvariational approach is very good up to g values close to the unphysical transition. In order to characterize the polaronformation we also analyze the electron-lattice correlationfunction χ defined i,δ in Sec. III. In particular in Fig. 4 we show the behavior of χ as a function of the SSH coupling for δ = 0,1,2 i,δ and t˜= 2.5. As expected, variational results and exact numerical data always recover the perturbative values in the limit of small el-ph coupling. Increasing g the monotonic behavior of the correlation function exhibits a kink, as the ground state energy. In particular the correlation function at next-nearest-neighbor (δ = 2) changes sign as the effective hopping, confirming the pathological behavior. At couplings where the ground state energy and the 5 correlation function show the kink, also the average phonon number is characterized by an anomalous behavior as shown in the bottom right panel of Fig. 4. In order to extract information on the values of g at which polaron crossover begins, before the opening of the unphysical region, we also investigate the behavior of the quasipartcle spectral weight Z(0). We find that increasing the el-ph coupling for fixed values of t˜, the spectral weight starts to drop but it never reaches a really small value before the unphysical sign change of the hopping occurs. Nevertheless we observe distinct signatures of the tendency towards localization, as shown in Fig. 5, where Z(0) is plotted as a function of g for the fixed value t˜=2.5. We conclude our analysis collecting the obtained data in the phase diagram of Fig. 6. It is calculated from the positionofthekinkinthegroundstateenergyobtainedbymeansofthevariationalapproach(diamonds)andtheexact diagonalization (triangles). The agreement between the two methods becomes better moving towards the adiabatic limit. InanalogywiththephasediagramobtainedfortheHolsteinpolaron,18 wealsomarkacrossoverregiondefined as the range of parameters for which Z(0) is less than 0.9. As shown in Fig. 6, we find that the considered SSH model does notpresentanymarkedmixing of electronicand phononicdegreesof freedom,being the stronglycoupled state prevented from the pathology of the model. As far as the fully adiabatic limit ω = 0 is concerned, we verify 0 that the crossover line joins onto the line for the transition to the unphysical region at the critical value λ = 0.25, confirming the discussion in Ref. [ 12]. We finally notice that, as discussed in Ref. [ 12], both the crossover region boundary,andtheinstabilitylineobtainedbyexactdiagonalizationareonlyweaklydependentontheadiabaticratio, andthatλisthe relevantelectron-phononcouplingregardlessthe valueoft˜. ThisisapeculiarityoftheSSHcoupling with respect to the Holstein one, where the polaron crossover moves to large values of λ as the phonon frequency increases.12,14,15,23 V. CONCLUSIONS In this work we discussed the features of one electron non-locally interacting with optical phonons in a discrete chain. We introduced a variational wave function to locate the crossover region for the transition between weak and strong localized polaron solutions. In particular we found that the pathological sign change of the effective next- nearest-neighborhoppingalwaysprecedesastablestronglylocalizedsolution. Suchanunphysicalregionofthemodel parameters does not occur in the case of acoustical phonons being the deformation linked to the particle extension along the entire chain.21 However we have also shown that, for finite values of the adiabaticity parameter, when the phonon frequency is not really small, the non local (SSH) el-ph interaction is more effective than the local (Holstein) one in reducing the mobility of the electron. Then our variational calculations are an interesting starting point to examine the complex problem of the polaron formation in a model where both local and non-local el-ph interactions are present. In particularwe emphasize that the proposedvariationalwavefunction for the SSH limit canbe slightly modified to be suitable for the treatment of the complex case where both interactions are present. Detailed future investigations in this direction are required. Finally we stress that the validity of our variationalresults is supported by an accurate analysis of exact diagonalization data on small clusters. The agreement between VA and ED data is good up to coupling values close to the unphysical region. VI. ACKNOWLEDGEMENTS M.C.acknowledgesthehospitalityandfinancialsupportofthe PhysicsDepartmentofthe UniversityofRome”La Sapienza”, as well as the INFM, UdR Roma 1 and SMC, and Miur Cofin 2001. APPENDIX A Inthe limitofsmallel-phcouplings,the perturbativesecondordercorrection∆E(k) tothe tight-bindingfreeband energy is 1+2t˜cosk sink2 1+4t˜cosk 4t˜2(1 cos2k) ∆E(k)= 4g2ω + − − − 0" 4t˜2 1+4t˜cosk 4t˜2(1 cos2k) − p 4t˜2 # − − 1 p g2ω . (20) − 1 0 1+4t˜cosk 4t˜2(1 cos2k) − − p 6 Moreover,using the bare phonon and electronic Green propagators,the perturbative self-energy reads 4g2ω [sin(k+q) sink]2 g2ω 1 Σ(k,ω)= 0 − + 1 0 . (21) N ω ω ε(k+q)+iδ N ω ω ε(k+q)+iδ 0 0 q − − q − − X X From Eq. (21) we obtain the momentum dependent spectral weight 2λ 4t˜2sin2k(1+2t˜cosk) (1+2t˜cosk) Z(k)−1 =1 1 − t˜ " − (1+4t˜ 4t˜2(1 cos2k))3/2 − 1+4t˜ 4t˜2(1 cos2k)# − − − − (1+2t˜cosk) p +2λ t˜ . (22) 1 (1+4t˜ 4t˜2(1 cos2k))3/2 − − APPENDIX B In this appendix we report the matrix elements between the states ψ(R) > and ψ(L) >. These quantities are | k | k involved in the calculation of the ground-state energy within the variational approach. We find ψ(L) ψ(R) = φ∗(R)( m )φ(R)(m )Z(L−R)(m m ), (23) h k | k i k − 1 k 2 k 1− 2 mX1,m2 where the phonon matrix element Z(L−R)(i j) is defined as k − Zk(L−R)(i−j)=ph h0|e−[Uk(L)(j)+Uk(L)(j−1)+Uk(L)(j+1)]e−[Uk(R)(i)+Uk(R)(i−1)+Uk(R)(i+1)]|0iph. (24) Then we have ψ(L) H ψ(R) = t φ∗(L)(m )φ(R)(m )[eikZ(L−R)(m m +1)e−ikZ(L−R)(m m 1)], h k | kin| k i − k 1 k 2 k 1− 2 k 1− 2− mX1,m2 ψ(L) H ψ(R) = ω φ∗(L)(m )φ(R)(m )[w∗(k)]2Z(L−R)(m m )eiq(m1−m2), (25) h k | ph| k i − 0 k 1 k 2 q k 1− 2 Xq mX1,m2 and (L) (R) ψ H ψ =A +A , (26) h k | int| k i 1 2 with A and A given by 1 2 gω A = 0 φ∗(L)(m )φ(R)(m )w∗(k)eikZ(L−R)(m m +1) eiq(m2−1)(1 eiq)+e−iqm1(e−iq 1) 1 √N k 1 k 2 q k 1− 2 − − q,mX1,m2 h i gω A = 0 φ∗(L)(m )φ(R)(m )w∗(k)e−ikZ(L−R)(m m 1) eiqm2(1 eiq)+e−iq(m1−1)(e−iq 1) . (27) 2 √N k 1 k 2 q k 1− 2− − − q,mX1,m2 h i The quantity ε = ψ(L) H ψ(L) = ψ(R) H ψ(R) is easily derived using the matrix elements given above. k h k | | k i h k | | k i 1J. M. De Teresa, M. R. Ibarra, P. A. Algarabel, C. Ritter, C. Marquina, J. Blasco, J. Garcia, A. del Moral, and Z. Arnold, Nature386, 256 (1997); A.J. Millis, ibid. 392, 147 (1998); M. B. Salamon and M. Jaime, Rev.Mod. Phys. 73, 583 (2001). 2M.Matus, H.Kuzmany,andE. Sohmen,Phys.Rev.Lett. 68, 2822 (1992); K.Harigaya, Phys.Rev.B45, 13676 (1992); B. Friedman, ibid. 45, 1454 (1992); W. M. You, C. L. Wang, F. C. Zhang, and Z.B. Su, ibid. 47, 4765 (1993). 7 3M.Verissimo-Alves,R.B.Capaz, B.Koiller, E.Artacho,andH.Chacham,Phys.Rev.Lett.86,3372 (2001); E.Piegari, V. Cataudella, V.Marigliano Ramaglia, andG.Iadonisiibid.89,049701 (2002); L.M.WoodsandG.D.Mahan,Phys.Rev.B 61, 10651 (2000); H. Suzuraand T. Ando,ibid.65, 235412 (2002). 4YuN. Gartstein, A.A. Zakhidov,and R. H.Baughman, Phys.Rev.Lett. 89, 045503 (2002). 5Guo-Meng-Zhao,M.B.Hunt,H.Keller,andK.A.Muller,Nature385,236(1997);A.Lanzara,P.V.Bogdanov,X.J.Zhou, S. A. Kellar, D. L. Feng, E. D. Lu, T. Yoshida, H. Eisaki, A. Fujimori, K. Kishio, J.-I. Shimoyama, T. Noda, S. Uchida, Z. Hussain, and Z.-X. Shen, ibid. 412, 510 (2001); R. J. McQueeney, J. L. Sarrao, P. G. Pagliuso, P. W. Stephens, and R. Osborn, Phys.Rev.Lett. 87, 77001 (2001). 6A.S. Alexandrov,N. F. Mott, Rep.Prog. Phys. 57, 1197 (1994). 7B. Gerlach and H. Lowen, Phys.Rev. B 35, 4291 (1987); 35, 4297 (1987). 8T. Holstein, Ann.Phys.8, 325 (1959); 8, 343 (1959). 9H.de Raedt and Ad Lagendijk, Phys.Rev.B 27, 6097 (1983); 30, 1671 (1984). 10P. E. Kornilovitch, Phys. Rev.Lett. 81, 5382 (1998). 11E. de Mello and J. Ranninger, Phys. Rev. B 55, 14872 (1997); A. S. Alexandrov, V. V. Kabanov, and D. K. Ray, ibid. 49, 9915 (1994); G. Wellein and H. Fehske,ibid.56, 4513 (1997). 12M. Capone, W. Stephan,and M. Grilli, Phys. Rev.B 56, 4484 (1997). 13S. R.White, Phys.Rev. B 48, 10345 (1993); E. Jeckelmann and S. R.White, ibid. 57, 6376 (1998). 14S. Ciuchi, F. dePasquale, S. Fratini, and D. Feinberg, Phys. Rev.B 56, 4494 (1997). 15M. Capone, S. Ciuchi, and C. Grimaldi, Europhys.Lett. 42, 523 (1998). 16A.H.Romero, D.W. Brown, and K. Lindenberg, Phys.Rev.B 59, 13728 (1999). 17J. Bonca, S. A.Trugman, and I.Batistic, Phys. Rev.B 60, 1633 (1999). 18V.Cataudella, G. De Filippis, and G. Iadonisi, Phys.Rev. B 60, 15163 (1999); 62, 1496 (2000). 19W. P. Su,J. R. Schrieffer,and A.J. Heeger, Phys. Rev.Lett. 42, 1698 (1979); Phys. Rev.B 22, 2099 (1980). 20Y.Lu, Solitons and Polarons in Conducting Polymers, World Scientific, Singapore 1988. 21A.La Magna and R.Pucci, Phys.Rev.B 55, 6296 (1997). 22M. Zoli, Phys. Rev.B 66, 012303 (2002); 67, 195102 (2003). 23M. Capone and S. Ciuchi, Phys. Rev.Lett. 91 186405 (2003). 24Y.Toyozawa, Prog. Theor. Phys. 26, 29 (1961). 8 0 ) 0 ( −1 E ∆ −2 0 5 10 15 20 t/ω 0 FIG. 1. Second order correction to the gound state energy, Eq. (6), as a function of the adiabatic inverse ratio t˜, for g = g1 = 1 (solid line). The dashed line is the SSH-like contribution (g = 1 and g1 = 0), the dot-dashed line is the Holstein contribution (g=0 and g1 =1). 0 0.2 0.1 −0.5 ) ) 0 1 2 = = δ δ 0, −1 0, −0.1 = = k k χ( χ( −0.2 −1.5 −0.3 −2 −0.4 0 5 10 15 20 0 5 10 15 20 t/ω t/ω 0 0 FIG.2. Left: Correlation functions at nearest-neighbor sites (on the left) and next-nearest-neighbor sites (on the right) at k=0as functionsof theinverseadiabatic ratio t˜, for g=g1 =1(solid line). Thedashed lines show theSSH-likecontribution (g=1 and g1=0), thedot-dashed lines the Holstein ones (g=0 and g1=1). −2 −2 t/ω=1 t/ω=2.5 −2.2 0 0 −2.2 −2.4 ) ) 0 0 E( 6 sites ED E( −2.6 P6 Tsites VA −2.4 6 sites VA 6 sites ED PT −2.8 −2.6 −3 0 0.5 1 1.5 0 0.2 0.4 0.6 0.8 1 g g 9 FIG.3. Ground state energy E(0) asa function of theSSHel-phcoupling g for two differentvaluesof theinverseadiabatic ratio t˜=2.5 (left) and t˜=1 (right). Solid and dotted lines are obtained from the variational approach and the Lanczos data for a six-site lattice, respectively; perturbative curves (dot-dashed lines) are plotted for comparison. Symbols mark the kink values of theenergy. 0.4 0 t/ω =2.5 0 0.3 −0.5 0.2 χ00 0.1 χ01 −1 0 −0.1 −1.5 0 0.5 1 1.5 0 0.5 1 1.5 0.4 3 PT 0.2 ED 2 VA χ02 0 Nph 1 −0.2 −0.4 0 0 0.5 1 1.5 0 0.5 1 1.5 FIG. 4. Correlation functions χk=0,δ with δ = 0 (top left), δ = 1 (top right) and δ = 2 (bottom left) as functions of the SSH coupling g for t˜= 2.5. Bottom right: Phonon number vs. g for the same value of t˜. Solid lines are obtained from the variationalapproachinthethermodynamicallimit;dottedlinesshowLanczosdata;perturbativecurvesfromEqs. (8)andEq. (7) with g1=0 (dot-dashed lines) are plotted for comparison. 1 t/ω=2.5 0 ) 0 0.8 ( Z ED VA PT 0.6 0 0.5 1 1.5 g FIG. 5. Spectral weight Z(0) as a function of the SSH el-ph coupling g for t˜= 2.5. The solid line is obtained from the variational approach in the thermodynamical limit; the dotted line shows Lanczos data; the perturbative curve from Eq. (9) with g1=0 (dot-dashedline) is plotted for comparison. 10