ebook img

Zinc Impurities in the 2D Hubbard model PDF

4 Pages·0.13 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Zinc Impurities in the 2D Hubbard model

Zinc Impurities in the 2D Hubbard model Th. A. Maier and M. Jarrell1 1University of Cincinnati, Cincinnati OH 45221, USA We study the two-dimensional Hubbard model with nonmagnetic Zn impurities modeled by bi- 2 narydiagonal disorder using QuantumMonteCarlo within the DynamicalCluster Approximation. 0 With increasing Zn content we find a strong suppression of d-wave superconductivity concomitant 0 toareductionofantiferromagnetic spinfluctuations. Tc vanisheslinearlywithZnimpurityconcen- 2 tration. ThespinsusceptibilitychangesfrompseudogaptoCurie-Weisslikebehaviorindicatingthe existence of free magnetic moments in the Zn doped system. We interpret these results within the n RVBpicture. a J 3 Introduction Chemical substitution in high-T super- proximation (CPA) [8]. The CPA shares the same mi- c conductors provides a powerful probe into the complex croscopic definition [9] as its equivalent for correlated ] n nature of both the superconducting and normal state of clean systems, the Dynamical Mean Field Approxima- o these materials. Experiments substituting different im- tion (DMFA) [10]. Both approachesmap the lattice sys- c purities for Cu show that nonmagnetic impurities (Zn, tem onto an effective impurity problem embedded in a - r Al) are just as effective in suppressing superconductiv- host that represents the remaining degrees of freedom. p ity as magnetic dopants (Ni, Fe) [1]. Based on Ander- This single-site approximation neglects interference ef- u s son’s theorem [2] these results are taken as strong evi- fects of the scattering off different impurity sites (cross- t. dence for an unconventional pairing state described by ingdiagrams)andcorrelationeffectsbecomepurelylocal. a an anisotropic order parameter with nodes on the Fermi Therefore,it inhibits a transitionto a state describedby m surface. Indeed, recently the characteristicfourfoldsym- a nonlocal (d-wave) order parameter and thus is not ap- - d metry of the dx2−y2-wave order parameter was observed propriate for our investigations here. n in the spatialvariationofthe localdensity ofstates near The Dynamical Cluster Approximation (DCA) [9, 11, o theZnimpurityusingscanningtunnelingmicroscopy[3]. 12, 13] systematically incorporates nonlocal corrections c [ Nuclear magnetic resonance experiments show that to the CPA/DMFA by mapping the lattice system onto even a nonmagnetic impurity substituted for Cu induces an embedded periodic cluster of size Nc. For Nc =1 the 1 aneffectivemagneticmomentresidingontheneighboring DCA is equivalentto the CPA/DMFAand by increasing v 7 Cu sites [4]. In addition, the bulk susceptibility in im- theclustersizeNc thelength-scaleofpossibledynamical 3 purity substituted underdoped cuprate superconductors correlationscanbegraduallyincreasedwhiletheDCAso- 0 shows Curie-Weiss like behavior irrespective of the mag- lution remains in the thermodynamic limit. In the clean 1 netic structure of the dopants [1]. This indicates the ex- limit the DCA applied to the Hubbard model has been 0 istence of free magnetic moments in the impurity doped shownto describe the essentiallow energyphysics of the 2 CuO planes. Theformationofthesemomentswithsub- cuprates [14, 15, 16]: It captures the antiferromagnetic 0 2 / stitution of nonmagnetic impurities is explained to arise phase near half filling and the transition to a supercon- t a from breaking singlets[1, 5] in an antiferromagnetically ductingphasewithdx2−y2-waveorderparameteratfinite m correlated host by removing Cu spins. doping. In the normal state it exhibits non-Fermi liquid - In this letter we focus on the suppression of supercon- behavior in form of a pseudogap in the density of states d andasuppressionofspinexcitationsatlowtemperatures ductivity by Zn impurities and the change of the bulk n in the underdoped regime. o magnetic susceptibility to a Curie-Weiss like behavior c in underdoped systems. Despite a wide variety of the- In this letter we study the two-dimensional(2D) Hub- v: oretical studies addressing these issues, a complete un- bardmodelincludingapotentialscatteringtermaccord- i derstanding of the effects of Zn doping has not been ing to the chemistry of Zn impurities using the DCA. X achieved. These approaches are mostly based on the de- We will show that potential scattering by Zn impuri- ar scription of impurities embedded in a BCS host [6] or tiesstronglysuppressessuperconductivityaswellasspin ongeneralizationsoftheAbrikosov-Gorkovequationsfor fluctuationsandchangesthemagneticsusceptibility toa nonmagnetic impurities in unconventional superconduc- Curie-Weiss like temperature dependence in the under- tors using phenomenological pairing interactions [7]. In doped region. ordertocaptureeffectslikemomentformationbysubsti- Formalism The correlated electrons in a CuO plane 2 tution of nonmagnetic impurities, it seems necessary to doped with Zn impurities are described by the 2D Hub- describe correlations and the scattering from impurities bard model with diagonal disorder on the same footing within a microscopic approach. The most widely applied technique developed to de- H =−t X c†iσcjσ +UXni↑ni↓+Xǫiniσ, (1) scribe disordered systems is the Coherent Potential Ap- <ij>,σ i iσ 2 where we used standard notation. The disorder induced QMC result for the cluster Green function Gc = ǫ1,...,ǫN,ij bytheZnsitesoccursinthelocalorbitalenergiesǫ which Gc [U,G ]thusdependsontheparticulardisorder i ij ǫ1,...,ǫN,ij are independent quenched random variables distributed configuration {ǫ }. The disorder averaged cluster Green i according to some specified distribution P(ǫ ,...,ǫ )= function is then obtained from the individual results by 1 N N P(ǫ ) where N is the number of lattice sites. For a Qi=1 i Gc =hGc i, (6) concentration x of Zn impurities we use the binary alloy ij ǫ1,...,ǫN,ij distribution where the average < ... >= N dǫ P(ǫ )(...) is to R Qi=1 i i P(ǫ )=xδ(ǫ +V/2)+(1−x)δ(ǫ −V/2), (2) be taken for a system of N sites. Note that in principal i i i c QMCcalculationsforallpossibledisorderconfigurations where V is the energy difference between the Cu and Zn {ǫ } have to be carried through. However, contributions i 3dorbitals. Forsimplicityweusesite-independentvalues from configurations with m Zn impurities on the cluster for the hopping integralt and the Coulomb repulsion U. (m ≤ Nc) are weighted by a factor of xm(1−x)(Nc−m) The microscopic derivation of the DCA algorithm was to the integral Eq.(6). Therefore it seems reasonable for discussed in detail for correlated systems in [12, 16] and small concentrations x ≪ 1 and cluster sizes N to con- c for disordered systems in [9]. The DCA has a simple sideronlythoseconfigurationswithnoneoronlyasingle physical interpretation for systems where the intersite Zn impurity on the cluster and to neglect configurations correlations have only short spatial range. The corre- withmorethanoneZnimpurity. Thenthedisorderaver- sponding self-energy may then be calculated on a coarse agedclusterGreenfunctionGc isobtainedasaweighted ij grid of Nc =LDc selected K points only, where Lc is the sum from the two configurations as linear dimension of the cluster of K points. Knowledge ofthemomentumdependenceonafinergridmaybedis- Gc =xN Gc +(1−xN )Gc , (7) ij c 1,ij c 0,ij carded to reduce the complexity of the problem. To this endthefirstBrillouinzoneisdividedintoN cellsofsize where Gc denotes the QMC result for the cluster c 1/0,ij (2π/L )D around the cluster momenta K. The propa- Greenfunction for the configurationwith one or zeroZn c gators used to form the self-energy are coarse grainedor impuritiesonthecluster,respectively. Theprefactorsfol- averagedoverthemomentaK+k˜surroundingthecluster lowwith(1−x)Nc ≈1−xNc,xNc(1−x)Nc−1 ≈xNc for momentum K x << 1 in linear approximation in the impurity concen- tration x. The disorder averaged cluster Green function N G¯(K,iωn) = c XG(K+k˜,iωn) Gcij is then transformed back to cluster reciprocal space N to calculate a new estimate of the cluster self-energy k˜ N 1 = c X . (3) Σ(K)=G(K)−1−Gc(K)−1, (8) N iω −ǫ −Σ(K,iω ) k˜ n K+k˜ n which is then used to repeat the steps starting from In Eq.(3) we make use of the fact that the approxima- Eq.(3) until convergence is reached. tion of the lattice self-energy Σ(k) by the cluster self- Results Westudythesuperconductinginstabilityand energy Σ(K) optimizes the free energy [12, 13]. Σ(K) magnetic properties of the Hamiltonian Eq.(1)in the in- summarizes the single-particle effects of the interaction termediatecouplingregimeat5%holedoping(n=0.95) term (second term) and the disorder term (third term) for different Zn impurity concentrations x. We perform oftheHamiltonian,Eq.(1). ThedispersionǫK+k˜ denotes calculations for Nc = 4, the smallest cluster size that thespectrumofthekineticpart(firstterm)ofEq.(1). In allows for a transition to a dx2−y2-wave superconduct- order to avoid overcounting of self-energy diagrams on ing state while preserving the lattice translational and the cluster, the cluster excluded propagator point group symmetries. We set the hopping integral t = 0.25eV and the Coulomb repulsion U = W = 2eV, G(K)−1 =G¯(K)−1+Σ(K) (4) where W = 8t is the bandwidth of the non-interacting system. We choose V = 5eV > W +U to simulate the is used as bare propagator for the cluster problem. To nonmagnetic closed shell (d10) configuration of the im- diagonalize the disorder part it is convenient to per- purity site. form a Fourier transform to cluster real space G = 1/NcPKeiK·(Xi−Xj)G(K) where the disorder is diiajgo- tenWdeedsesa-rwchavede,fopr- asunpderdc-ownadvuectoivrditeyr wpiatrhams-ewtearvse., eAxs- nal. The inverse bare cluster propagatorfor a particular we found for the clean system (x=0) [14, 16], only the disorder configuration {ǫ } is then given by i dx2−y2-wave pair field susceptibility for zero center of [G−1 ] =[G−1] −ǫ δ , (5) mass momentum diverges at low temperatures. Fig.1 ǫ1,...,ǫN ij ij i ij illustrates the inverse pair field susceptibility P−1 as a d and is used to initialize a QMC simulation to calcu- function of temperature T for the impurity concentra- late the effects of the Coulomb interaction U. The tions x = 0 (circles), x = 0.05 (squares) and x = 0.10 3 15 5 x=0.00 0.02 χ(T)=χ -χ T+C/(T-Θ) x=0.05 c 4 0 1 x=0.10 T 0.01 10 3 0 1 0.05 0.1 -Pd x χ P -1∝ (T-T)γ 2 d c 5 x=0.00, T=0.020, γ=0.72 c x=0.05, T=0.012, γ=0.82 1 c x=0.10, T=0.002, γ=0.86 c 00 0.05 0.1 0.15 0.2 00 0.05 0.1 0.15 0.2 T T FIG. 1: The inverse pair-field susceptibility versus tempera- FIG.2: The bulkmagneticsusceptibility versustemperature turewhenNc =4,U =W =2eV,V =5eV at5%dopingfor when Nc = 4, U = W = 2eV, V = 5eV at 5% doping for impurity concentrations x = 0 (circles), x = 0.05 (squares) impurity concentrations x = 0 (circles), x = 0.05 (squares) andx=0.10(diamonds). Thesolidlinesrepresentfitstothe and x = 0.10 (diamonds). The solid lines are guides to the function Pd−1 ∝(T −Tc)γ. Inset: Critical temperature Tc as eye. The dashed line represents a fit to the function χ(T)= a function of impurity concentration x. χ0−χ1T +C/(T −Θ). (diamonds). The corresponding critical temperature T χ(T) shows Curie-Weiss like behavior; i.e. it can be fit c is then calculated by extrapolating P−1(T) to zero us- by the function χ(T) = χ − χ T + C/(T − Θ), with d 0 1 ing the function P−1 ∝ (T −T )γ (represented by the Θ≈0. This clearly indicates the formation of free mag- d c solid lines in Fig.1). T is rapidly suppressed, and the netic moments due to the breaking of singlet bonds by c critical exponent γ increases,with increasing Zn content the substitution of nonmagnetic impurities. x. Consistent with experiments [1] this falloff of Tc is The inverse antiferromagnetic spin susceptibility linear in x resulting in a critical impurity concentration χ (T) for the same parameter set as in Fig.2 is illus- AF xc ≈ 0.10, beyond which the instability to the super- trated in Fig.3. For the clean system (x=0) 5% hole conducting phase disappears. Compared to experiments doping marks the criticaldoping, beyond whichno tran- (xc <∼0.05) our calculation predicts a result about twice sitiontoanantiferromagneticstatecanbefound[14]. As as high for the critical doping. However,considering the canbe seenfromthe extrapolationof the circlesin Fig.3 simplifieddescriptionoftheproblemusinga2DHubbard the corresponding Ne´el temperature T = 0. As the N model with the complexity of the Zn impurities reduced concentration of impurities is increased we notice that to a model of diagonal disorder, we expect agreement for T >∼ 0.05 χAF decreases with x. However at lower only on a qualitative level. temperatures (T <∼0.05)χ−AF1 for finite impurity concen- In order to study the correlation between the suppres- tration x falls below the curve of the clean system and sionofsuperconductivityandpossiblemomentformation eventually goes through zero indicating a transition to by Zn substitution we study the bulk (q = (0,0)) and the antiferromagnetic state. Therefore it can be inferred the antiferromagnetic (q = (π,π)) spin susceptibilities. that a finite concentration of impurities, i.e. spin vacan- Fig.2 shows the bulk magnetic susceptibility χ(T) as a cies,acts to enhance antiferromagneticspincorrelations. function of temperature for the same impurity concen- Ourresultsforthesiteandtimeaveragedlocalmagnetic trations x used in Fig.1. As we demonstrated in [14], moment Tχii shown in the inset to Fig.3 support this the clean system exhibits an anomaly in χ(T) at a char- conjecture: At low temperatures this quantity increases acteristic temperature T⋆, below which spin excitations with impurity concentration x indicating that local mo- are suppressed and χ(T) falls with decreasing tempera- ments get stabilized. It is important to note that recent ture. T⋆ thusmarkstheonsetofsingletformationdueto NMR measurements on YBCO [17] show a similar en- strong short-ranged antiferromagnetic spin correlations. hancementofantiferromagneticspincorrelationsaround In agreement with experiments [1], the substitution of Zn impurities at low temperatures. impurities leadsto adramaticchangeofthe lowtemper- Similar results were also obtained in [18, 19] where an ature behavior of the magnetic spin susceptibility. De- antiferromagneticHeisenberg model with a spin vacancy spite the nonmagnetic nature of the impurity scattering wasstudied. Inthelatterreport,theenhancementofan- term in our simulation, χ(T) displays an upturn at low tiferromagneticspincorrelationswasascribedtopruning temperatures as x is increased. Moreover at x = 0.10, ofsingletsbythenonmagneticimpurityinanresonating- 4 00..44 tions that mediate pairing get suppressed. Consistent with experiments, T decreases linearly with impurity c 0.3 concentration. With increasing Zn content we further 00..33 χii find a change of the magnetic susceptibility to Curie- T Weiss like behavior indicative of the existence of free 0.2 1 magnetic moments. -F00..22 χA 0.02 0.04 0.1 χAF-1∝(T-TN)γ ThOeunrornemsualtgsnceatincbimepuunrditeiresstobordeawkiltohcianltshinegRleVtsBapnidcttuhrues. T generate unpaired spins. A concomitant suppression of 00..11 x=0.00, γ=1.53 x=0.05, γ=1.42 antiferromagnetic spin fluctuations that mediate pairing x=0.10, γ=1.35 results from the pruning of RVB states [19]. Conse- quently,superconductivitygetsstronglysuppressedwith 00 00 0.05 0.1 0.15 0.2 increasing Zn content. T Acknowledgements We acknowledge useful conversa- tions with W. Putikka, D.J. Scalapino, Y. Wang and M. FIG. 3: The inverse antiferromagnetic (q =(π,π)) spin sus- Vojta. This work was supported by NSF grant DMR- ceptibility versus temperature when Nc =4, U =W =2eV, V = 5eV at 5% doping for impurity concentrations x = 0 0073308. This research was supported in part by NSF (circles), x = 0.05 (squares) and x = 0.10 (diamonds). The cooperative agreement ACI-9619020 through computing solid lines represent fits to the function χ−AF1 ∝ (T −TN)γ. resources provided by the National Partnership for Ad- Inset: The site and time averaged magnetic moment versus vanced Computational Infrastructure at the Pittsburgh temperature; spin vacancies reduce fluctuations. The solid Supercomputer Center. lines are guides to theeye. valence-bond picture (RVB). In the RVB picture, the short-ranged order in the doped model is described as a [1] G. Xiao, et al., Phys.Rev.B 42, 8752 (1990). collectionofnearest-neighborsingletbondswhichfluctu- [2] P.W. Anderson, J. Phys. Chem. Solids 11, 25 (1959). ateasafunctionoftimeandspace. WhenaZnimpurity [3] S.H. Pan, et al., Nature403, 746 (2000). is introduced into the system, the singlet fluctuations on [4] H. Alloul, et al., Phys. Rev. Lett. 67, 3140 (1991); A.V. adjacent sites are suppressed since these sites now have Mahajan, et al.,Phys. Rev.Lett. 72, 3100 (1994). one less neighbor to form short-ranged bonds with. The [5] A.M. Finkelstein, et al.,Physica C 168, 370 (1990). [6] T. Xiang and J.M. Wheatley, Phys. Rev. B 51, 11721 suppression of singlet fluctuations on the sites adjacent (1995); G. Haran´ and A.D.S. Nagi, Phys. Rev. B 58, to the Zn dopant also enhances moment formation on 12441 (1998). these sites. [7] R.J.Radtke,etal.,Phys.Rev.B48653(1993);S.Haas, Wefindfurtherevidenceofthesuppressionofspinfluc- et al.,Phys.Rev. B 56, 5108 (1997). tuations by Zn impurities in the decrease of the anti- [8] D.W.Taylor,Phys.Rev.1561017(1967);P.SovenPhys. ferromagnetic exponent γ towards the mean-field result Rev.156809(1967);P.L.LeathandB.Goodman,Phys. of one with increasing impurity content x. Concomi- Rev. 148 968 (1966). [9] M. Jarrell and H.R. Krishnamurthy, Phys. Rev. B, 63 tantwiththisbehaviorthecorrespondingexponentγ for 125102 (2001). the pair-field susceptibility (cf. Fig.1) increases towards [10] W. Metzner and D. Vollhardt, Phys. Rev. Lett. 62, 324 one. Fromthisobservationweconcludethatthesuppres- (1989); E. Mu¨ller-Hartmann: Z. Phys. B 74, 507–512 sionofsuperconductivityinZndopedsystemsoriginates (1989);T.Pruschke,etal.,Adv.inPhys.42,187(1995); in the suppression of antiferromagnetic spin fluctuations A. Georges, et al.,Rev.Mod. Phys.68, 13 (1996). that mediate pairing. [11] M.H. Hettler et al., Phys.Rev.B 58, 7475 (1998). Summary Inthisletterwehaveusedthe2DHubbard [12] M.H. Hettler, et al.,Phys. Rev.B 61, 12739 (2000). [13] Th. Maier, et al.,Eur. Phys. J. B 13, 613 (2000). model with binary diagonaldisorder to study the effects [14] M. Jarrell, et al.,EuroPhy.Lett., 56 563 (2001). of nonmagnetic (Zn) impurities on high-temperature su- [15] Th. Maier, et al.,Phys. Rev.Lett. 85, 1524 (2000). perconductivity. WefindthatZnimpuritiesstronglysup- [16] M. Jarrell, et al.,Phys.Rev. B 64, 195130 (2001). press superconductivity. As evidenced by the increasing [17] M.-J. Julien et al.,Phys.Rev. Lett.84, 3422 (2000). mean-field character of the pairfield and antiferromag- [18] N. Bulut, et al.,Phys. Rev.Lett. 62, 2192 (1989). netic susceptibilities with Zn substitution, spin fluctua- [19] G. B. Martins, et al.,Phys. Rev.Lett. 78, 3563 (1997).

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.