Emergent p-wave Kondo Coupling in Multi-Orbital Bands with Mirror Symmetry Breaking Jun Won Rhim1,∗ and Jung Hoon Han2,3 1School of Physics, Korea Institute for Advanced Study, Seoul 130-722, Korea 2Department of Physics and BK21 Physics Research Division, Sungkyunkwan University, Suwon 440-746, Korea 3Asia Pacific Center for Theoretical Physics, Pohang, Gyeongbuk 790-784, Korea (Dated: January 17, 2013) We examine Kondo effect in the periodic Anderson model for which the conduction band is of 3 multi-orbital character and subject to mirror symmetry breaking field imposed externally. Taking 1 p-orbital-based toy model for analysis, we find the Kondo pairing symmetry of p-wave character 0 emerges self-consistently over some regions of parameter space and filling factor even though only 2 the on-site Kondo hybridization is assumed in the microscopic Hamiltonian. The band structure n in the Kondo-hybridized phase becomes nematic, with only two-fold symmetry, due to the p-wave a Kondocoupling. Thereducedsymmetry shouldbereadily observable inspectroscopic ortransport J measurements for heavy fermion system in a multilayer environment such as successfully grown 6 recently. 1 PACSnumbers: 71.27.+a,73.20.-r,75.30.Mb ] l e - Introduction.-State-of-the-arttechniquetogrowheavy r st fermion compounds with atomic thickness precision has 2D multi-orbital Kondo px, py γ pz . beenachievedinrecentyears[1,2]. Heavyfermionphase lattice system t a andsuperconductivitywereidentifiedinCeCoIn5layerof Surface m lessthanonenanometerthicknessgrowninthis manner, z - prompting speculations of exotic new phases in atomi- y d cally thin heterostructure [2]. Apart from the novelty of Bulk x n Bulk reduced dimensionality, the alternate stacking of heavy o c fermion layer with normal metal YbCoIn5 layer realized [ inthegrowthprocessislikelytogiverisetonewphysical FIG. 1: (Color online) Schematic picture of Kondo lattice phenomena by virtue of the mirror symmetry breaking modelsubjecttomirrorsymmetrybreakingatthesurface or 1 v (MSB) at the interface. Aspects ofnon-centrosymmetric interface. Hopping processes between, for instance, pz and 4 superconductivity have been exhaustively studied since px,y otherwise forbidden in the bulk become possible due to thesymmetrybreakingfieldas shownin theupperright car- 6 its original discovery in the heavy-fermion compound toon. 5 CePt Si [3]. The heavy fermion multilayer structure, 3 3 on the other hand, differs from the conventional non- . 1 centrosymmetric material in that the violation of sym- 0 electron dynamics. In this new picture Rashba interac- metryismacroscopicallyimposed(notmicroscopicallyat 3 tion becomes an electrostatic effect with its magnitude thelevelofcrystalstructure)and,withrecentlyavailable 1 dictated by the degree of orbital asymmetry about the : techniques such as electrolyte gating, controllable. Due v mirror plane (see Fig. 1) [5–7]. to the thin layer thickness, such MSB is likely to infect i X the entire heavy fermion band structure and therefore In fact Kondo lattice behavior takes place in an orbital-rich environment with the extended p- or d- r the nature of emerging Kondo and/or superconducting a orbital states forming the conduction bands while the pairing as well. Some consequences of the MSB in the localized moments arise from f-orbitals [8, 9]. Using heavyfermionmultilayerhavebeenstudiedtheoretically our recent experience with the Rashba phenomena in by Maruyama et al. quite recently [4], where the main non-Kondo materials as a guide, we present a faithful focus was on the novel superconducting pairing symme- modeling of Kondo lattice behavior where multi-orbital triesandotherthermodynamicpropertiesinducedbythe degrees of freedom in a mirror symmetry-broken en- phenomenological Rashba term. vironment are explicitly embedded in the conduction IthasbeencustomaryinthepasttoregardtheRashba band Hamiltonian. The use of Rashba interaction as a interaction as a relativistic effect manifesting itself in a purely phenomenological addition, as is often done in mirrorsymmetry-brokenenvironment. Severalrecentpa- the current literature, is avoided. Our approach differs pers[5–7],however,haveshownthatmulti-orbitaldegen- as well from the large-N generalization of the one-band eracy, in addition to MSB, is an essential pre-requisite Kondo lattice model, where N identical copies with the for the manifestation of strong Rashba coupling in the same orbital symmetry are adopted as a way to justify 2 0.1 0.12 the mean-field approximation [9]. In reality the MSB field - usually the electric field acting perpendicular to III III the interface or the surface - influences orbitals which are extended along the field direction more than those Txy which extend in the orthogonal direction. Conventional TB0.05 K 0.06 k II one-band or large-N approach assuming featureless II orbitals will not capture this effect properly. Tz K 0(a) I I 0(b) 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 1 1.2 The model.- Choosing the degenerate p-orbital states γ γ hopping on a square lattice as a toy example, we write down the Hamiltonian [6, 7] FIG. 2: (Color online) Phase diagrams with (a) α = 0 (no SOC), nc = 2 and (b) α = 0.5 (SOC), nc = 1.5. Phase I denotesallthreep-orbitalshybridized(hx,hy,hz 6=0). Phase σ σ σ II represents only px, py orbitals hybridized (hxσ 6=0,hyσ 6=0, hz = 0). Phase III is trivial with no Kondo hybridization σ HMSB =Xkλσ(cid:16)ǫλkp†λkσpλkσ+3γisλp†zkσpλkσ+h.c.(cid:17) e(ahnnλσcdo=Tunx0tye).r.PtIawnroa(mad)e,itsetarisnscsuthsoeKwdontnobdoyobattearmiendpbevoreatrhttuipcrheasalsalearrbdoeiwaleg,drawmbeysmaTarKyze + µffk†σfkσ (1) (V1,V2K,JK)=(2,0.65,3). Fermisurfacestructuresatthefour Xkσ points indicated by stars are shown in Fig. 3. forthedispersivep-orbital(pλkσ)bandsandasinglepair of localized Kramers doublet (fkσ) with on-site energy J bλ µ . The orbital index λ ∈ {x,y,z} spans the three p- H = K σ¯σp† f −hλp† f +h.c. f K 2 (cid:18) 2 λiσ iσ¯ σ λiσ iσ (cid:19) orbital states of the conduction band and σ = +(−) is Xiλσ the spin up(down). One can show, employing the two J J + K Γpf† f + K Γfp† p . (2) Slater-Koster parameters V1, V2 for σ- and π-bonding 2 σ iσ¯ iσ 2 σ λiσ¯ λiσ of p-orbitals, that ǫλ = −2(V + V )c + 2V (c + c ) Xiσ Xiλσ k 1 2 λ 2 x y with cx(y) = coskx(y), sx(y) = sinkx(y) and cz = sz = 0 In addition to the Kondo hybridization parameters for the assumed two-dimensional bands. Mirror symme- bλσσ′ = hfi†σpλiσ′i and hλσ = bλσσ + bλσ¯σ¯/2, we also try breaking is reflected in the second term of Eq. (1), allow same-orbital spin-flip processes encoded by proportionalto the MSB parmaeter γ [7]. They give the Γp = hp† p i and Γf = hf† f i. The spin-flip σ λ λiσ λiσ¯ σ iσ¯ iσ hybridizationofpz-orbitalwithpx-andpy-orbitals;these paramePters Γp, Γf, and bλ can be nonzero due to the σ σ σσ¯ areprocessesnormallyforbiddenbysymmetryifamirror spin-orbit interaction. We restrict ourselves however to planeexisted. Thiskindofsymmetrybreakingisrealized non-magnetic, singlet Kondo coupling in the mean-field naturallyonthesurfaceofabulkmaterialorbyapplying scheme and avoid the possibility of Kondo pairing be- vertical electric field to the planar system as illustrated tween remote sites [10], which is another known avenue in Fig. 1. to obtain non-s-wave Kondo pairing. The fact that our model can exhibit the p-wave Kondo phase despite the Secondly the Kondo hybridization of local and itin- on-site-only Kondo hybridization Hamiltonian H em- erant spins follows from the Heisenberg-type exchange K phasizes the emergent nature of unconventional pairing. interaction H = J S ·S , with spin-1/2 oper- K K i,λ iλ if As demonstrated in Refs. [5–7], Rashba interaction is ators made out of pλP-orbital (Siλ) and f-orbital (Sif), another emergent property of the band that can be respectively. Finally H = (α/2) L ·σ involving SOC i i i derived from the underlying microscopic Hamiltonian the p-orbital angular momentum opPerator (L = 1) Li consisting only of H and H . andthespin-1/2operator(1/2)σ ateachsiteigivesthe MSB SOC i spin-orbitcoupling(SOC)withinthe p-orbitalbands. In Phase diagram.- All the hybridization parameters can typicalheavyfermionmatter suchasCeCoIn the heavy 5 be determined self-consistently, along with µ to ensure element In playing a main role in the conduction band f h f† f i = 1, and the overall chemical potential µ, structurewillalsocarrysubstantialamountofspin-orbit σ iσ iσ energy α. foPr various input band structure parameters V1,V2,γ, electron filling factor n = h p† p i, at various c λσ λiσ λiσ ThetotalHamiltonianH =HMSB+HSOC+HKcanbe temperatures T. Generic phasPe diagrams resulting from handledwithintheframeworkofrenormalizedmean-field ourmodelHamiltonianareplottedinFig. 2inthe space theory [10–12] which decomposes the Kondo exchange of temperature T and MSB parameter γ with and with- interaction as (σ =−σ) out the SOC. We observe two distinct Kondo-coupled 3 π M With a proper choice of tight binding parameters and filling factors we can identify cases in the self-consistent band structure wherein the two lowest-energy spin-orbit ky 0 Γ X coupledp-orbitalbands andthe two Kramersdoublet f- electron bands occur in the vicinity of the Fermi energy whileallothersarefurtherremovedfromit. Insuchcases −π(a) (b) onecandescribethe Kondophysicswithin the projected M Hilbert space of those four bands associated with the π following operators: k 0 6 y c˜nk = vλnkσpλkσ and f˜kσ =fkσ+ unkσc˜nk. (3) Xλσ nX=3 −π(c) (d) Here, n=1−6 labels the six p-orbitalbands of H + −π 0 π −π 0 π H , the state at k being denoted |vni = c˜† |0Mi SwBith k k SOC k nk x x the coefficients vn . The localized f-electron operator λkσ ismodifiedthroughKondohybridizationwiththehigher FIG. 3: (Color online) Band dispersions at the Fermi level energybands, unkσ =hvkn|HK|fkσi/(µf−εcnk), n=3−6, for thevarious phases marked byred stars (⋆) in Fig. 2. (a), where εc is the energy of n-th band of H +H nk MSB SOC (IIbI)inanFdig(.c)2(aar)ertehspeeFcteirvmeliy,suarnfdac(eds)ocfortrheesppohnadssestoI,thIeI satnadr and|fkσi=fk†σ|0i. TheeffectiveHamiltonianwithinthe in Fig. 2(b). The reduced symmetry in the Kondo-coupled reducedHilbert space of {|vk1i,|vk2i,|f˜k+i,|f˜k−i}, where phases(a), (b),and(d) duetothep-waveKondopairing are |f˜kσi=f˜k†σ|0i, is apparent against the (kx,ky) axes in red dashed lines. Heff = εcnkc˜†nkc˜nk+ εfσkf˜k†σf˜kσ phases labeled I in Fig. 2, where all three p-orbitals nX=1,2 Xσ honyblyridthizee iwni-tphlafn-eorpbxi,taply(horxσb,ihtyσal,shzσhy6=br0id)i,zean(dhzσII =whe0r)e. +n=X1,2,σ(cid:16)Vknσc˜†nkf˜kσ+h.c.(cid:17). (4) The high-temperature non-Kondo phase with hλ = 0 σ is labeled III. Both transitions, I→II and II→III, are The f-level dispersion εf and the effective Kondo cou- σk second-order. Phase I becomes unstable and gives way pling Vknσ =hvkn|HK|fkσi are given by to phase II as showninFig. 2(b) asthe strengthofSOC is increased. Although the addition of SOC complicates the microscopic Hamiltonian, fortunately, many of the εf =µ + 6 |hvkn|HK|fkσi|2, hybridization parameters were found to be vanishing σk f nX=3 µf−εcnk Γp = Γf = bz = bx = by = 0 even with finite SOC. σ σ σσ σσ¯ σσ¯ J 1 Features of the phase diagram are dependent on the Vknσ = 2K (cid:18)2bzσσ¯vznkσ¯ −hσx¯vxnkσ−hyσ¯vynkσ(cid:19). (5) fillingfactorn ofp-electrons. PhaseIisobservedwithin c a small range of nc and, as indicated in Fig. 2(a), may Some properties of the self-consistently obtained hy- even exhibit some re-entrant behavior with increasing bridization parameters of Kondo-coupled phases such as γ. While the details regarding the extent of each Kondo Γp = Γf = bz = bx = by = 0 are used to derive σ σ σσ σσ¯ σσ¯ phase or the Kondo temperature are sensitive to the the above effective Hamiltonian. The energy dispersion occupation nc, the effect of SOC on them was found to obtainedfromHeff matchesthefulldispersionalmostex- be much less significant by comparison. actly. CharacteristicsoftheKondo-coupledbandscanbe discussed faithfully now in terms of the effective Hamil- EffectiveHamiltonianforp-waveKondopairing.-Most tonian, H . eff interestingly, the two Kondo-coupledphases I and II ex- The effective Kondo coupling Vnσ appearing in H k eff hibit the band dispersion with the loss of symmetry un- has the explicit momentum dependence despite the fact der the reflection (k ,k ) → (−k ,k ) or (k ,k ) → that the original Hamiltonian only contained the on-site x y x y x y (k ,−k ), as well as the 90◦ rotation, as shown in Fig. Kondointeraction. The symmetry ofthe Kondopairing, x y 3. Reduction in the band structure symmetry only oc- self-consistently determined to be p-wave as illustrated curs in the Kondo phase implying that it can only be in Fig. 4, also dictates that of the hybridized bands, the consequence of p-wave Kondo coupling. Below we as shown in Fig. 3. The symmetry propertes of Vnσ, k show rigorously how such unconventional pairing occurs in turn, follow from those of the wave functions vn as λkσ for MSB-infected bands. outlined in Table I. 4 k wave function the M point meet the conditions assumed in deriving (kx,ky) {vxnk+,vynk+,vznk+,vxnk−,vynk−,vznk−} the effective Hamiltonian Heff. Here one can show vλnkσ (−kx,ky) {−vxnk∗+,vynk∗+,−vznk∗+,−vxnk∗−,vynk∗−,−vznk∗−} near k = (π,π) is vx1,yk+ ≈ −3γsx,y/2(V1 + V2) and (kx,−ky) {vxnk∗+,−vynk∗+,−vznk∗+,−vxnk∗−,vynk∗−,vznk∗−} vx1,yk− = 0, and vice versa for n = 2. Turning off SOC, (−kx,−ky) {−vxnk+,−vynk+,vznk+,vxnk−,vynk−,−vznk−} all spin-mixing mean field parameters bzσσ¯ vanish. We canconcludeonlytwoequivalentgroundstateswithreal- (−ky,kx) {−vynk+,vxnk+,vznk+,−ivynk−,ivxnk−,ivznk−} valued Kondo parameters hx = hy or hx = −hy are σ σ σ σ TABLE I: Relation between eigenstates of HMSB+HSOC at possible, giving out several symmetry-related momenta. vλnkσ in the first row refers to the amplitude of pλ-orbital with spin σ for the 3iJ hxγ eigenstate at k = (kx,ky); σ = +(−) corresponds to spin Vk ≡Vk1+ =Vk2− = K (sinkx±sinky) (6) 4(V +V ) up(down). Eigenstates at other momenta are related to this 1 2 one as shown in thesubsequentrows. and V1− =V2+ =0 for hx =±hy, respectively. The ef- k k σ σ fective Kondo coupling indeed has the p-wave symmetry (a) α=0 (b) α=0.5 withanodealongthek =+k ork =−k line. Figure π 0 .1 π 0 .3 x y x y 4(b) shows the plot of |V1+| for the case of hx = −hy k σ σ where the node extends along k = −k . It is also seen x y 0.2 thatp-waveKondocouplingisnotpredicatedonSOCat 0 ky0 all. Instead, the p-wave Kondo coupling is strictly a 0.1 consequence of nonzero MSB, γ 6= 0. When γ = 0, Hamiltonian matrices of H + H at (k ,k ), MSB SOC x y −π−π 0 π 0 −π−π 0 π 0 (−kx,ky), (kx,−ky) and (−kx,−ky) are equal to each kx kx other [13]. As a result, those momenta have identical eigenvectors and, according to Eq. (5), the same FIG. 4: (Color online) Contour plots of |Vk1+|, the absolute Kondo coupling Vknσ. In addition the p-wave character value of the effective Kondo hybridization, in the Brillouin arises only when both hx and hy are nonzero. For σ σ zone. (a) α=0 (no SOC) and (b) α=0.5. phases I and II such condition is indeed achieved with |hx| = |hy|. The p-wave Kondo states are favored away σ σ from half-filling regime such as considered in the phase Transformation rules for the wave functions are de- diagram of Fig. 2. At half filling nc = 3 without SOC, duced by examining whether the Hamiltonians at a pair on the other hand, we often find the ground phase with of symmetry-related momenta can transform into each only hz non-vanishing. In this case, the effective Kondo σ other by some kind of unitary or anti-unitary opera- coupling is inthe formVknσ =−JKhzσ¯vznkσ/2≈iJKhσz¯/2, tors such as Kσ ⊗ R (π), KI ⊗ R (π), σ ⊗ R (π) whichiss-waveasintheordinaryheavyfermionmatter. z x y z z and e−i34πDz1/2(3π/2)⊗Rz(π/2). We denote Rλ(θ) for the SO(3) rotation matrix, λ = x,y,z, D1/2(θ) for the Discussion.- As the angstrom-scale deposition tech- z niqueofthinfilmsdevelopsintomaturity,theroleofsur- SU(2) rotation matrix, and K for the complex conjuga- facestateswithbrokenmirrorsymmetrytakesongreater tion. From the Table and Eq. (5) one can show that practical significance. Such MSB field effect leads to in- Vnσ = −σVnσ , and therefore a unitary transfor- −kx,−ky kx,ky terestingconsequenceswhenactingonamulti-orbitalen- mation σ ⊗I between H (k) and H (−k) exists, en- z eff eff vironment such as the emergence of Rashba effect [6, 7], suringtheequivalenceoftheenergiesbetweenkand−k. orbital-dependentspintransfertorque[14],andasargued On the other hand Vnσ , Vnσ and Vnσ are not −kx,ky kx,−ky −ky,kx inthis work,ofun-conventionalKondopairing. Conven- relatedtoVnσ inanywayaccordingtotheTableI. For kx,ky tional one-band Kondo model with the phenomenolog- example,Vnσ =J (−bz vn∗ /2+hxvn∗ −hyvn∗ )/2 −kx,ky K σσ¯ zkσ¯ σ¯ xkσ σ¯ ykσ ical Rashba term could very well miss this feature due does not even have the same magnitude as Vkn,σ in gen- to the inadequate treatment of the multi-orbital char- eralbecause the wave function vλnkσ are complex-valued. acter. The asymmetric band structure we found in the Referring to Eq. (5), neither the reflection about the x- Kondo-coupled phase should be readily detected in the orthey-axisnorthe90◦rotationink-spaceisguaranteed spectroscopic study such as ARPES, or in the transport to yield unitary-equivalent Heff. measurements through their directional responses. Al- It is further possible to find an explicit expression of though proposals of non-s-wave Kondo pairing are al- the effective Kondo coupling to prove its p-wave nature ready available [10–12], they tend to trace its origin to for models without SOC. Given the parameters such as thenon-localcharacterofKondocouplingbetweenlocal- V = 2, V = 0.65, γ = 0.3 and the filling of conduc- ized and itinerant moments [10], or the orbital symme- 1 2 tion electrons n = 1.5, the band dispersions around try of the f-orbitallocalized moments [12]. By contrast, c 5 thep-waveKondohybridizationemerginginourscenario [4] Daisuke Maruyama, Manfred Sigrist, and Youichi is a generic consequence of the macroscopic breaking of Yanase, J. Phys.Soc. Jpn.81, 034702 (2012). mirror symmetry imposed by the geometry of the het- [5] L. Petersen and P. Hedeg˚ard, Surf.Sci. 459, 49 (2000). [6] Seung Ryong Park, Choong H. Kim, Jaejun Yu, Jung erostructureitself. Existingtheoryofheavyfermionhet- HoonHan,andChangyoungKim,Phys.Rev.Lett.107, erostructure [4] focused on the nature of unconventional 156803 (2011). superconductivity resulting from Rashba effect, but did [7] Jin-Hong Park, Choong H. Kim, Jun-Won Rhim, and not address the nature of Kondo pairing itself. Jung Hoon Han, Phys.Rev.B 85, 195401 (2012). J. H. H. is supported by NRF grant (No. 2010- [8] Takahiro Maehira, Takashi Hotta, Kazuo Ueda, and 0008529, 2011-0015631) and acknowledges discussion Akira Hasegawa, J. Phys. Soc. Jpn. 72, 854 (2003). with Yuji Matsuda that inspired this project. [9] D. L. Cox, Exotic Kondo Effects in Metals: Magnetic Ions in a Crystalline Electric Field and Tunnelling Cen- tres (Taylor & Francis, London,1999). [10] PouyanGhaemiandT.Senthil,Phys.Rev.B75,144412 (2007). ∗ Electronic address: [email protected] [11] Heidrun Weber and Matthias Vojta, Phys. Rev. B 77, 125118 (2008). [1] H.Shishido,T.Shibauchi,K.Yasu,T.Kato,H.Kontani, [12] PouyanGhaemi,T.Senthil,andP.Coleman,Phys.Rev. T.Terashima,andY.Matsuda,Science327,980(2010). B 77, 245108 (2008). [2] Y.Mizukami,H.Shishido,T.Shibauchi,M.Shimozawa, [13] One can show that this statement is consistent with Ta- S. Yasumoto, D.Watanabe, M. Yamashita, H. Ikeda, T. bleIafterallowingforthedoubledegeneracyofthestate Terashima, H. Kontani, and Y. Matsuda, Nat. Phys. 7, duetovanishingγ andtheU(1)gaugefreedom ofeigen- 849 (2011). vectors. [3] E.Bauer,G.Hilscher,H.Michor,C.Paul,E.W.Scheidt, [14] Jin-Hong Park, Choong H. Kim, Hyun-Woo Lee, and A. Gribanov, Y. Seropegin, H. No¨el, M. Sigrist, and P. Jung Hoon Han, Phys.Rev.B 87, 041301(R) (2013). Rogl, Phys. Rev.Lett. 92 027003 (2004).