arXiv:cond-mat/9803086 v1 8 Mar 1998 inaotatatbemdl nti ae,w promul- we paper, a this of In existence the model. gate tractable expan- a an there facilitates about which bluntly: sion parameter more small obvious it no put is To strongly-coupled, regime. a in non-perturbative apparently are namely, materials obstacle, these serious that a of because fruition reached not odcosi 1986 in conductors aeil’rc hs iga n hnmnlg have phenomenology focus. and into come diagram phase these advanced, rich have materials’ precision sample experimental As and activity. stimulated quality which theoretical event and experimental an intense physics, matter condensed of oeawihw iht xli.Tepaedarmas diagram phe- temperature phase the The of function review explain. a to briefly wish us we let which nomena consequences, its and tion of calculations controlled quantities. for measurable This way experimentally diagram. the phase the paves of description region exotic particularly a of lyblee ob rprinlt h ffciehl con- hole effective the each to in gener- proportional centration is be but to material, believed to ally material from varies doping the r nuaigatfroant As eo h N´eel the below (AFs) temperature antiferromagnets insulating are izugLna n odnter tnadfrcon- (SCs). same for the superconductors standard by theory ventional London described and adequately Ginzburg-Landau is super- the phase purposes conducting phenomenological many For occurs. nFg .Bs nesodadi aycssvr well very cases many ( in undoped and the understood are Best characterized 1. Fig. in n o eprtrs( temperatures low and h icvr ftecpaehg-eprtr super- high-temperature cuprate the of discovery The eoepugn nooreeei fti uldescrip- dual this of exegesis our into plunging Before oa iudTer ftePed-a hs fHigh- of Phase Pseudo-Gap the of Theory Liquid Nodal iiga h omrdwv nodes, d-wave former the at living iea ffciefil hoyfrti hs.Teter scom is theory The phase. 1-dim this + for 2 theory a field use effective We an rive quasiparticles). d-wave (whic the photoemission from electron and inherited transport in e.g. denced, ucpiiiy o-rva hra odciiy n tw and conductivity, thermal non-trivial susceptibility, eaaetendllqi rmvrosodrdphases. ordered various from liquid zero-temper nodal the the discuss c separate also including We phenomena, i interesting superconductivity. of theory d-wave panoply field a effective of our analysis bet and naturally antiferromagnet, interpolates liquid insulating nodal The theory. Landau odcos nteasneo muiis hs nld pow u include in these state pseudo-gap impurities, the of of absence explanation the an In as conductors. numer it suggest has It to us superconductor. d-wave a quantum-disordering T N eitoueadsuythe study and introduce We 2 tmdrt oig (0 dopings moderate At . nttt o hoeia hsc,Uiest fCaliforn of University Physics, Theoretical for Institute .INTRODUCTION I. oee,mn hoeia ffrshave efforts theoretical many However, CuO 1 a aese ntercn history recent the in watershed a was weakly-coupled 2 T < T ae)i niae schematically indicated is layer) T n doping and enBlns ate .A ihr n htnNayak Chetan and Fisher, A. P. Matthew Balents, Leon c x ( x 3 )mtras which materials, 0) = ) superconductivity )), n motn dis- important One da’description ‘dual’ x oa liquid nodal tentr of nature (the . 1 ekycoupled weakly ∼ < x ∼ < Jn ,2004) 9, (June 0 oe eotmeaueqatmpaeotie by obtained phase quantum zero-temperature novel a , . 3) 1 oteflcutn ag edo ulGinzburg- dual a of field gauge fluctuating the to T uecnutra ucino doping of function a as superconductor nthe in ture ubro euiu n opligexperiments, compelling after and accepted beautiful generally of now number and a years many for pected high-T the between tinction C eo h ihtmeauedte ie( line dotted high-temperature the Below SC. ed ogpesqaiatce eiiga h ornodes four wavefunction. the pair at symmetry the residing d-wave of quasiparticles zero-temperature gapless particular, to a In leads in ingredient description. crucial quantum a is symmetry their nFg ,agersle hteiso (ARPES), photoemission angle-resolved 1, Fig. in transport, nry(ige)lcrncadsi ere ffreedom. of degrees spin and (single-)electronic energy measurements -adoepril hregp,telte evi- latter the gaps, charge one-particle and o- i.1 ceai hs iga fahigh-temperature a of diagram phase Schematic 1: Fig. eetsuishv eeldpzln behavior puzzling revealed have studies Recent AF T xiispoone orodanisotropy fourfold pronounced exhibits h d entedwv uecnutradthe and superconductor d-wave the ween . ag reig niermgeim and antiferromagnetism, ordering, harge oefleog opri detailed a permit to enough powerful s x tr unu hs rniin which transitions phase quantum ature a at abr,C 93106-4030 CA Barbara, Santa ia, rsdo gapless of prised 2 rlwmgei re,a order, magnetic er-law ninldaiytasomto ode- to transformation duality ensional − u eakbepoete hc lead which properties remarkable ous underdoped 8 y droe ihtmeauesuper- high-temperature nderdoped 2 NMR, ( Pseudo-Gap 2 d-wave niaeadaai euto flow- of reduction dramatic a indicate 9 , 10 T*(x) arn ymty hspairing This symmetry. pairing ) pia conductivity, optical T neutral einbtenteA and AF the between region c Superconductors c n ovninlSs sus- SCs, conventional and ia particles Dirac DSC T lna spin -linear x 11 n tempera- and , 12 n other and T ∗ 4 ( , 5 x X 6 )) , is 7 Furthermore, the ability of ARPES measurements to re- spin-ful) quanta we call nodons. In the dual variables, solve wavevector dependence exposes an angular varia- the fundamental Lagrangian of our model is tion similar to that of d-wave quasiparticles in the SCing † z x phase6,7. This portion of the phase diagram is commonly = ψ1[i∂t vF τ i∂x v∆τ i∂y]ψ1 + (1 2, x y) L κµ − −2 ↔ ↔ called the pseudo-gap region. The ultimate nature of the + (∂µ iaµ)Φ VΦ( Φ ) corresponding underlying quantum ground state is an in- 2 | − | − | | −1 1 2 2 triguing theoretical puzzle, and a principal subject of this +aµǫµνλ∂ν (Aλ κ Jλ)+ (ej b ) − λ 2κ − paper. 0 +2λJ0 + N . (1.1) To proceed, we look to the experiments for guidance. L They indicate three types of excitations which are im- In the sections which follow, we will elucidate the physics portant below the dotted line in Fig. 1: the ordering of this Lagrangian in some detail, so we restrict our- fields related to antiferromagnetism and superconductiv- selves, in this introduction, to a whirlwind tour. The † ity, and d-wave quasiparticles near the four nodes. Con- fields ψj , ψj , j = 1, 2 are the nodon creation and anni- spicuously absent from this list are electrons and holes at hilation operators at the two antipodal pairs of nodes. an ordinary Fermi-surface. The physics of this omission Jλ is a bilinear in the nodon operators which has an in- is that pairing occurs (due to unspecified strong interac- terpretation in the d-wave superconducting phase as the tion physics) at the high energy T ∗(x). Given these in- quasiparticle electrical 3-current. Φ is the complex scalar gredients, one natural theoretical strategy is to attempt field representing the vortices, and aµ is the gauge field to approach the pseudo-gap state by increasing x from which is dual to the phase of the superconducting or- the AF at half-filling. Many researchers have already der parameter. The term proportional to λ describes the attempted this approach, but it remains inconclusive. effects of particle/hole asymmetries, and the N term de- We, instead, tackle the pseudo-gap state from the scribes the coupling of the nodons to antiferromagnetism,L right, literally. To do this, we must contemplate which we will return to presently. quantum-disordering the d-wave superconductor. For A remarkable result of calculations with Eq. 1.1 is that simplicity, we will assume for the moment a purely two- gapless nodons survive the quantum disordering of the dimensional model of a single CuO2 plane. We imagine SC! The nodons are like the smile of the Cheshire cat: that pairing establishes a local superconducting d-wave the dx2−y2 order parameter is gone, but the nodes re- order parameter ∆(~x, t) = ∆ eiϕ, where ~x is the two- main. The consequent Nodal Liquid (NL) described by dimensional coordinate and|t |is time. The experimen- Eq. 1.1 is a distinct and novel zero-temperature quan- tal properties of the pseudo-gap state can be interpreted tum phase with a number of fascinating properties. For as an indication that ∆ is large in this region, so that simplicity, consider first a hypothetical NL phase at half- quantum fluctuations of| the| phase of the order-parameter filling in the absence of impurities. The possibility of AF ϕ must be responsible for the lack of off-diagonal long- ordering will be included later via N . We hypothesize range-order ( ∆ = 0), even as T 0. The important that antiferromagnetism might beL avoided and the NL long-distanceh dephasingi is accomplished→ by vortex loops achieved in a half-filled square lattice model by adjusting and lines, around which ϕ winds by 2π. To destroy an attractive nearest-neighbor interaction and second- the long-range correlations in ∆, we must± unbind vortex neighbor electron hopping amplitude. The NL is a nomi- loops of arbitrarily large size, just as vortex-antivortex nally insulating state, with non-zero gaps ∆1 and ∆2 for pairs unbind above the Kosterlitz-Thouless transition adding both individual electrons/holes and Cooper pairs, temperature in a two-dimensional superfluid. To imple- respectively. Gapless nodons with anisotropic ballistic ment this unbinding, it is extremely helpful to use the dispersion (ω k), however, persist, and can carry both 2 + 1-dimensional duality13,14 relating an XY-model for spin and current.∼ With particle/hole symmetry (λ = 0 ∆ and a Ginzburg-Landau model with complex field Φ in Eq. 1.1), we expect power law ( 1/ x 4) spin cor- (“disorder parameter”) interacting with a gauge field a. relations at (0, 0), (π, π), (π, 0), and∼ symmetry-related| | The duality interchanges Cooper pairs and vortices, so points in the Brillouin zone. Scaling arguments lead us that the desired quantum disordered d-wave state is the to expect a weak dissipative dynamic contribution to the ordered (condensed) phase, Φ = 0, of the Ginzburg- conductivity which, in the presence of particle/hole sym- Landau theory. The imaginary-timeh i 6 effective action for metry, would vary as Re σ(ω,T = 0) ω6. Nodons ∼ 2 this dual theory is nothing but the Ginzburg-Landau free also contribute a quadratic specific heat Cnodon aT . energy functional for a three-dimensional superconductor Despite the similarity of the nodons to d-wave quasipar-∼ at finite-temperature. ticles, the single-particle spectral function is predicted to The fate of the d-wave quasiparticles in this construc- show a gap at the Fermi energy in the NL. We expect, tion requires particular care, since these are strongly cou- however, this gap to be strongly angle-dependent: of or- pled to the fluctuating pair field. In Sec. II A, we show der the pairing scale T ∗ away from ( π/2, π/2) and ∗ ± ± how the quasiparticles can be treated by extracting the reduced to ∆1 T at these special points. A parti- U(1) phase from the bare electron operators. Once the cle/hole asymmetric≪ NL should exhibit similar behavior, phase is extracted, one is left with a set of gauge-invariant but with singularities shifted from ( π/2, π/2) in mo- fermion operators which create electrically neutral (but mentum space, and an even smaller± contribution± to the

2 low-frequency dissipative conductivity which, in the ab- the AF, both single- and two-particle gaps are non-zero, sence of umklapp scattering and impurities must come but the spin gap vanishes due to low frequency magnons. from interactions with phonons. In the d-wave superconductor, the two-particle gap van- Consider next doping the NL. Naively, this can be ishes, since the pairs have condensed, but spin and single- modeled via an increase in the in-plane hole concentra- particle gaps are “almost” non-zero (which we call “0+”), tion, although the actual transfer of charge to the CuO2 since only the quasiparticles carry these quantum num- layers may be not be complete. In the grand canonical bers, and their density of states vanishes with the energy. ensemble, charge is added by increasing µ = A0/2. In Passing from the SC to the NL, ∆2 changes from zero to + the dual theory (see Eq. 1.1), this chemical potential acts non-zero, and ∆1 changes from 0 to a true gap. The like an external magnetic field h = 2µ in the Ginzburg- transition to the AF occurs then simply by developing a Landau theory (but, unlike a magnetic field, it of course non-zero staggered magnetization. does not break time-reversal invariance). For small µ, the This transition and other magnetic physics is discussed system remains in the Meissner phase, and no dual flux in Sec. IV, using the N´eel Lagrangian density penetrates – i.e. no charge is added to the system within the charge gap. Following the analogy with Ginzburg- Kµ 2 N = ∂µN VN ( N ) Landau theory,3 we expect that the nature of doping de- L 2 | | − | | † z y † pends upon the Abrikosov parameter κv = λv/ξv, where +g[N ψ τ σ σψ + h.c.], (1.2) · λv and ξv are effective dual penetration and coherence lengths, respectively. For κv > 1/√2, type II doping where g measures the strength of the coupling between occurs, and the “field” penetrates∼ first for µ > µc1 in the N´eel order parameter and the nodons. Eq. 1.2 can a dual flux lattice. Dual flux tubes are in fact Cooper be obtained by introducing a 2kF density-density interac- pairs, so this is a paired Wigner crystal (PWC) state. tions between the nodons and decoupling the antipodal We show in Sec. III that gapless nodons survive the dop- terms with the N´eel vector N. Let us once again con- ing and coexist with the PWC. We generally expect the sider first the case of half-filling with particle/hole sym- displacements of the Wigner crystal to be pinned either metry. For sufficiently strong interaction g, or when the by the periodic lattice potential or disorder (either is ef- quadratic coefficient rN in VN (N) is negative, one obtains fective when arbitrarily weak), so that this phase remains an AF phase with N = 0. In this phase the nodons insulating. As doping increases from zero, the character- develop a gap and low-energyh i 6 spin quanta are carried en- istic nodon momenta shift further from ( π/2, π/2). tirely by spin waves. Depending upon the “mass” rΦ of ± ± Continued doping to µ > µc2 leads to another transi- VΦ, this is either a simple AF or AF order coexisting with tion into the “normal” state of the dual theory, which is a d-wave SC. Decreasing g or increasing rN destroys the nothing but the d-wave SC. Neglecting disorder, and with long-range AF order and liberates the nodons. This inter- weak lattice effects, quantum fluctuations are expected to esting phase transition is discussed in Sec. IV. Increasing drive this 2+1-dimensional flux lattice melting transition rΦ results in a further transition to the d-wave super- weakly first order. For κ < 1/√2, one has instead type I conductor, which we believe is in the three-dimensional doping at a single “critical∼ field” µc. This is a first order inverted-XY universality class. Tuning rΦ = rN = 0 de- transition, accompanied by a jump in the hole concentra- scribes a multicritical point connecting directly the AF tion from 0 to xc at µ = µc. In the canonical ensemble and d-wave SC phases. with fixed 0 0. Agreement with experi- the theoretical point of view, as it offers a compelling ment then requires that g be sufficiently strong to induce interpolation between the undoped AF and the d-wave AF order. As the chemical potential µ is increased above superconductor. Consider the following three important the charge gap to induce holes into the system, the hole energies : the single-particle gap, ∆1, the minimum en- density or dual “internal field” becomes non-zero. From ergy required to add a charge e and spin s =1/2 to the Eq. 1.1, this creates an effectively larger particle/hole system; the two-particle gap,± ∆ , the minimum energy asymmetry λ λ x. This presents a competition. 2 eff − ∼ to add charge 2e and spin s = 0; and the spin gap, ∆s, By ordering the N´eel vector the system can create a gap the energy required± to add spin s = 1 but no charge. In for the nodons and reduce their kinetic energy. However,

3 ′ at finite λeff the nodon Dirac point would prefer to move ∆ (~q)= V (k,q,k ) c ′ c ′ . (2.4) k h −k +q↓ k ↑i away from ( π/2, π/2), which reduces the gain in ki- k′ netic energy.± As x±increases, therefore, we may expect X 16 to drive transitions from the AF to AF/NL and pure BCS theory can be implemented in terms of the spa- NL phases. Of course, there are in fact many different tially varying pair field, obtained via Fourier transfor- scenarios for type I and type II doping, small or large in- mation, ∆k(~x). The self-consistent gap equation is usu- trinsic λ, etc. These are discussed in Sec. IV. Once mag- ally solved for a spatially uniform order parameter, with ∆ ∆ (~x). Singlet pairing implies ∆ = ∆ , andin a netism has been discussed, we conclude with a summary k ≡ k k −k of the main points of the paper, open issues, relations to d-wave superconductor ∆k has four zero’s or nodes as ~k other work, and a brief discussion of experimental impli- varies around the Fermi surface. Our strategy will be to cations. Finally, two appendices include technical details obtain an effective field theory which has a local d wave − of microscopic and ǫ-expansion calculations. gap function, determined by strong coupling physics be- low some length scale Λ−1 of say 5-10 lattice spacings, but which can fluctuate quantum mechanically on longer II. D-WAVE AND DUALITY spatial scales. These longer length scale quantum fluc- tuations will be responsible for quantum disordering the A. Model and Symmetries d-wave superconductor, and will allow us to access a new phase – the nodal liquid. As we shall see, an important role is played by the d wave quasiparticles, which survive Consider a tight binding model of electrons hopping the quantum disordering.− To implement this approach, on a square lattice, with a local Hamiltonian satisfying we first briefly recapitulate the properties of quasiparti- certain general symmetries. We will assume the system is cles in the d wave superconductor. both U(1) and SU(2) invariant (i.e. we neglect spin-orbit − coupling), and has time-reversal, reflection, and four-fold rotational symmetry. Sometimes it will also be conve- B. Quasiparticles nient to specialize to models which possess an additional discrete particle/hole symmetry. We denote lattice elec- † With spatially uniform d-wave order given by ∆k, tron creation and annihilation operators as cα(~x) and c (~x), where ~x is the two-dimensional coordinate in the the effective Hamiltonian for the quasiparticles is H = α H + H , with H the kinetic energy and frame with x = x1 and y = x2 parallel to the a and b 0 1 0 crystalline axes (i.e. to the Cu–O bonds). Here α is a † † ∗ spin label. In momentum space the kinetic energy takes H1 = [∆kck↑c−k↓ + ∆kc−k↓ck↑]. (2.5) the usual form, Xk Since ∆k = ∆−k for singlet pairing, it is natural to break H = ǫ c† c , (2.1) 0 k kα kα sums into positive and negative ky. To do so, consider a kα X four component fermion field, Υaα(~k), at each wavevector and at this stage we allow for general electron interac- with ky > 0 positive: tions: Υ ck↑ † † 11 ′ ′ ′ † Hint = V (k,q,k )c−k+q,αckβ c−k +q,βck ,α. (2.2) Υ21 c−k↓ ′ Υ (~k)= =   . (2.6) k,q,k aα  Υ  X 12 ck↓    Υ22   c†  A discrete particle/hole transformation is implemented    − −k↑  by   In the second column vector the minus sign has been p/h introduced so that Υaα transforms like a spinor un- c (~x) ei~π·~xc† (~x), (2.3) α −→ α der an SU(2) rotation, i.e. as Υaα UαβΥaβ. Here U = exp(iθ σ) is a global spin rotation→ with Pauli ma- with ~π = (π, π). Many common models (e.g. the Hub- · trices σαβ. bard and t-J) are invariant under a particle/hole trans- In these variables, the quasiparticle Hamiltonian be- formation at half-filling. Invariance of the kinetic energy comes implies that ǫk = ǫk+~π, a form valid with near neigh- − ′ bor hopping. However, a second neighbor hopping term † z + − ∗ Hqp = Υ (~k)[τ ǫk + τ ∆k + τ ∆ ]Υ(~k), (2.7) violates particle/hole symmetry. k k As discussed in the introduction, we wish to describe X the physics below the relatively strong d-wave pairing where the prime on the summation denotes over ky pos- scale T ∗, in order to approach the pseudo-gap phase from itive, only, and we have introduced a vector of Pauli ma- the superconducting side. To do so, we imagine introduc- trices, ~τab acting in the particle/hole subspace. Also, we ing a d-wave order parameter are employing the notation τ ± = (τ x iτ y)/2. ±

4 and the particle/hole symmetry breaking term,

† q y z λ = λΨj τ Ψj. (2.11) q x H

K The quasiparticle Hamiltonian takes the form of (four) 2 K1 Dirac equations, and can be readily diagonalized giving a dispersion relation for the first pair of nodes,

E (~q)= (v q + λ)2 + ∆˜ 2q2, (2.12) 1 ± F x | | y b q and a similar expression with qx and qy exchanged for a the second pair. Notice that non-zero λ indeed shifts the positions of the nodes. Fig. 2: The wavevectors K~ i, qx, qy in relation to the a,b axes. The dotted line represents the putative Fermi surface. C. Quantum Fluctuations

With approximate particle/hole symmetry, the d-wave Up to this point, we have taken a spatially constant nodes are located near the special wavevectors K~ j, with ± gap function ∆(˜ ~x). To disorder the d-wave supercon- K~ = (π/2,π/2) and K~ = ( π/2,π/2). Since our aim 1 2 ductor it is necessary to allow for quantum fluctuations is to obtain an effective description− at low energies and of this order parameter. It is tempting to uniformly long lengthscales, it is sufficient to focus on the gapless suppress the complex order parameter, and simply put modes near these points, integrating out the electrons far ∆˜ = 0. But doing so recovers the conventional metallic away in the Brillouin zone. It is then convenient to intro- state with a Fermi surface. Our task is trickier, since we duce two continuum fields Ψ , one for each pair of nodes, j are searching for an intermediate phase, which has strong expanded around K~ , K~ : ± 1 ± 2 local d wave pairing (which destroys the Fermi surface) but with− longer length scale quantum fluctuations de- Ψ (~q)=Υ (K~ + ~q). (2.8) jaα aα j stroying the superconducting phase coherence. Our task Here, the wavevectors ~q are assumed to be small, within a is similar to the problem of describing the hexatic phase 17 circle of radius Λ around the origin. With this definition, in a classical two-dimensional triangular solid, which is the particle/hole transformation is extremely simple, intermediate between the crystalline and liquid phases. Guided by this example and the principle of pairing be- Ψ Ψ†. (2.9) low T ∗, we want to fix the magnitude of the complex pair → field, and introduce fluctuations of its phase. always For this reason it is convenient to define the con- Pursuant to this goal, we write tinuum fields Ψ around K~ , and account for deviations ± j of the node momenta from these values by a particle/hole iϕ ∆˜ v∆e , (2.13) symmetry-breaking parameter λ. → Once we have restricted attention to the momenta near where v∆ is real and ϕ can be interpreted as the phase of the nodes, it is legitimate to linearize in the quasiparticle the complex superconducting order parameter. The BCS Hamiltonian. The resulting continuum theory is more gap equation has a degenerate manifold of solutions, for conveniently written in coordinates perpendicular and arbitrary phase ϕ. This degeneracy is responsible for the parallel to the Fermi surface, so we perform the rota- Goldstone modes, wherein ϕ varies slowly in both space tion via x (x y)/√2 and y (x + y)/√2, cor- → − → and time. Our goal is to obtain an effective theory for respondingly transforming the momenta qx and qy (see the space and time dependence of ϕ, similar in spirit to Fig. 2). Linearizing near the nodes, we put ǫK1+q = vF qx the non-linear sigma models “derived” for localization.18 ˜ where vF is the Fermi velocity and ∆K1+q = ∆qy, where Specifically, we focus on spatial variations of ϕ(~x) on ∆˜ has dimensions of a velocity. An identical lineariza- scales longer than Λ−1. Since ϕ can vary spatially, care tion is possible around the second pair of nodes, except is needed in introducing it into the quasiparticle Hamil- with qx qy. Upon Fourier transforming back into real tonian: ↔ space, Ψj(~q) Ψj(~x), we arrive at a compact form for → + + iϕ/2 iϕ/2 the Hamiltonian density of the quasiparticle excitations ∆˜ τ i∂y v∆τ e (i∂y)e . (2.14) in the d-wave superconductor: = + with → Hqp HΨ Hλ This symmetric form leads to an hermitian Hamiltonian, =Ψ†[v τ zi∂ + (∆˜ τ + + ∆˜ ∗τ −)i∂ ]Ψ HΨ 1 F x y 1 physical currents, and respects the symmetries of the +(1 2; x y), (2.10) problem. A careful derivation of Eq. 2.14 is given in ↔ ↔

5 † z x Appendix A. With this prescription, the quasiparticle = ψ [i∂ v τ i∂ v∆τ i∂ ]ψ Lψ 1 t − F x − y 1 Hamiltonian becomes +(1 2, x y), (2.20) ↔ ↔ † z s isϕ/2 isϕ/2 = Ψ [v τ i∂ + v∆τ e (i∂ )e ]Ψ Hqp 1 F x y 1 interacting with the phase of the order-parameter: s=± X +(1 2; x y). (2.15) = ∂ ϕJ . (2.21) ↔ ↔ Lint µ µ

Since ϕ can also fluctuate with time, it will convenient to Here the electrical 3-current Jµ is given by consider the time dependence via a Lagrangian formula- 1 tion. The Lagrangian density is J = ψ†τ zψ , (2.22) 0 2 j j =Ψ†i∂ Ψ H . (2.16) Lqp j t j − qp vF † The appropriate Lagrangian for the phase of the d- Jj = ψ ψ . (2.23) 2 j j wave order parameter is simply Because the transformation in Eq. 2.18 is local, identical 1 2 ϕ = κµ(∂µϕ) , (2.17) expressions hold for these currents in terms of the quasi- L 2 particle fields, Ψ. The form of the particle/hole asymme- where the Greek index µ runs over time and two spatial try term remains the same in terms of the nodon fields: coordinates: µ =0, 1, 2= t,x,y. Here κ0 is equal to the † z 2 λ = λψj τ ψj . (2.24) compressibility of the condensate and κj = vc κ0 (for L j =1, 2= x, y) with v the superfluid sound velocity.− We c It is instructive to re-express the components of the expect that the pair compressibility κ is approximately 0 currents J back in terms of the original electron opera- one half the electron compressibility of the original elec- µ tors. One finds tron model – in the absence of interactions. If the pairing is electronic in origin, one expects that the scale for the 1 † † J0 = (c c + c c ), (2.25) “charge velocity” vc is the Fermi velocity. 2 Kj Kj −Kj −Kj As discussed in the introduction, treatment of quan- tum phase fluctuations is complicated by the mixing of (with an implicit spin summation) which corresponds particle and hole variables via the complex gap function. physically to the total electron density living at the To isolate the uncertain charge, we therefore perform a nodes, in units of the Cooper pair charge. Similarly, change of variables, defining a new set of fermion fields vF † † ψj via Jj = (c c c c ) (2.26) 2 Kj Kj − −Kj −Kj ψ = exp( iϕτ z/2)Ψ . (2.18) j − j corresponds to the current carried by the electrons at the nodes. Thus, Jµ can be correctly interpreted as the In the superconducting phase, and in the absence of quasiparticles three-current. quantum flucutations of the order-paramater phase, one To complete the description of a quantum mechani- can set ϕ = 0, and these new fermions are simply the cally fluctuating order parameter phase interacting with d-wave quasiparticles. However, when the field ϕ is dy- the gapless fermionic excitations at the nodes, we min- namical and fluctuates strongly this change of variables imally couple to an external electromagnetic field, Aµ. is non-trivial. In particular, the new fermion fields ψ are Since the nodon fermions are neutral, the only cou- electrically neutral, invariant under a global charge U(1) pling is to the order-parameter phase, via the substitu- transformation. As we shall see, when the d-wave super- tion ∂µϕ ∂µϕ Aµ. For simplicity, here and in the conductivity is quantum disordered, these new fields will rest of the→ paper,− we have set the Cooper pair charge play a fundamental role, describing low energy gapless ex- 2e = 1. The final Lagrangian then takes the form citations, centered at the former nodes. For this reason, = + + + , with we refer to these fermions as nodons. For completeness, L Lϕ Lψ Lint Lλ we quote the symmetry properties of the nodon field un- 1 = κ (∂ ϕ A )2, (2.27) der a particle/hole transformation. Since ϕ ϕ, one Lϕ 2 µ µ − µ has simply → −

ψ ψ†. (2.19) = (∂ ϕ A )J , (2.28) → Lint µ − µ µ

The full Lagrangian in the d-wave superconductor, and ψ still given by Eq. 2.20. = + , can be conveniently re-expressed in terms TheL time component of the electromagnetic field is pro- L Lϕ Lqp of these nodon fields since qp = ψ + int + λ with a portional to the chemical potential µ, i.e. A0 =2µ. For free nodon piece, L L L L electrons at half-filling one has µ = 0. Doping can be

6 κµ 2 achieved by changing µ. Long-ranged Coulomb interac- v = (∂µ iaµ)Φ VΦ( Φ ), (2.33) tions could be readily incorporated at this stage by treat- L 2 | − | − | | ing A0 as a dynamical field and adding a term to the La- where Φ is a (dimensionless) complex field for the 2 v grangian of the form, coul = (1/2)(∂jA0) . The spatial vortices. The vortex 3-current, following from j = L µ components of the electromagnetic field, Aj , have been ∂ v/∂aµ, is included to keep track of the current operator. In par- L jv = κ Im[Φ∗(∂ ia )Φ]. (2.34) ticular, the total electrical 3-current is obtained by dif- µ µ µ − µ ferentiating the Lagrangian, i.e. J tot = ∂ /∂A , which µ L µ For small Φ (appropriate close to a second order transi- gives | | 2 tion) one can expand the potential as, VΦ(X)= rΦX + 4 J tot = κ (∂ ϕ A )+ J . (2.29) uΦX . The remaining piece of the dual Lagrangian is µ µ µ − µ µ 1 2 2 −1 Here the first terms are the Cooper pair 3-current, and a = (ej b )+ aµǫµνλ∂ν (Aλ κλ Jλ), (2.35) L 2κ0 − − the second the quasiparticles current. The equation of motion for the phase of the order-parameter, ∂ /∂ϕ = 0, with dual “magnetic” and “electric” fields: b = ǫij ∂iaj tot L −1 implies the continuity equation ∂µJµ = 0. and ej = vc (∂j a0 ∂0aj). It is straightforward to verify that the dual− Lagrangian has the desired prop- erty that Eq. 2.32 follows from the equation of motion ∂ /∂a = 0. D. Duality LD µ

To quantum-disorder the d-wave superconductor, one III. NODAL LIQUID PHASE must allow for vortices in the pair-field phase, ϕ. We do this using field-theoretic duality, as described, e.g. in In this section we employ the dual representation of Ref. 13. To this end we introduce a vortex 3-current, jv, µ the d-wave superconductor to analyze the quantum dis- which satisfies, ordered phase - a new phase of matter which we refer to as jv = ǫ ∂ ∂ ϕ. (2.30) a nodal liquid. The dual representation comprises a com- µ µνλ ν λ plex vortex field, which is minimally coupled to a gauge field, as well as a set of neutral nodon fermions. Without In the presence of vortices, ϕ is multi-valued, ∂µϕ is not v the nodons and in imaginary time, the dual Lagrangian curl-free, and jµ is non-vanishing. In the desired dual representation, the vortices become the quantized parti- is formally equivalent to a classical three-dimensional su- cles, rather than the Cooper pairs. However, even in the perconductor at finite temperature, coupled to a fluctu- dual representation one still needs to conserve the total ating electromagnetic field. To disorder the d-wave su- electrical charge. This can be achieved by expressing the perconductor, we must order the dual “superconductor” total electrical 3-current as a curl, – that is, condense the vortices. The nature of the result- ing phase will depend sensitively on doping, since upon tot doping, the dual “superconductor” starts seeing an ap- Jµ = ǫµνλ∂ν aλ, (2.31) plied “magnetic field”. Below, we first consider the sim- where we have introduced a “fictitious” dynamical gauge pler case of half-filling. We then turn to the doped case, field, aµ. Upon combining Eqs. 2.29-2.31, one can elimi- where two scenarios are possible depending on whether 3 nate the pair-field phase, ϕ, and relate aµ to the vortices: the dual “superconductor” is Type I or Type II.

v −1 −1 j = ǫµνλ∂ν [κ ǫλαβ∂αaβ + Aλ κ Jλ], (2.32) µ λ − λ A. Half-filling where Jµ is the quasiparticle 3-current defined earlier in Eqs. 2.22-2.23. Specialize first to the case of electrons at half-filling, A dual description is obtained by constructing a La- with particle-hole symmetry. In the dual representation, v grangian, D, depending on aµ, Jµ and jµ, whose equa- the “magnetic field”, b, is equal to the deviation of the to- tion of motion,L obtained by differentiating with respect to tal electron density from half-filling. Thus at half-filling a , leads to the above equation. To assure that the vor- b = 0 and the dual Ginzburg-Landau theory is in zero µ h i tex 3-current is conserved, it is useful to introduce a com- applied field. The quantum disordered phase corresponds plex field, Φ, which can be viewed as a vortex destruc- to condensing the vortices, setting Φ =Φ0 = 0. In this h i 6 tion operator. Since a vortex acquires a 2π phase upon dual Meissner phase, the vortex Lagrangian becomes encircling a Cooper pair, the vortex field should be min- 1 2 2 imally coupled to aµ. The appropriate dual Lagrangian v = κµΦoaµ. (3.1) can be conveniently decomposed as = + + , L 2 LD Lψ Lv La where ψ is given in Eq. 2.20. The vortex piece has the It is then possible to integrate out the field aµ. The full Ginzburg-LandauL form,3 Lagrangian in the nodal liquid phase is then

7 2 1 ǫ0 2 B 2 † σ = + A I + E + O (∂J) , (3.2) Sq = c c . (3.4) nl ψ µ µ j 2 k+q k L L 2 − 2µ0 k   X where we have introduced the physical magnetic and elec- At low energies in the nodal liquid phase one can focus tric fields: B = ǫ ∂ A and E = ∂ A ∂ A . The last ~ ~ ij i j j j 0 − t j on momenta near the nodes: k = Kj. The electron two terms describe a dielectric, with magnetic perme- operators near the nodes can be rewritten± in terms of 2 2 −1 ability µ0 = κ0Φ0 and dielectric constant ǫ0 = (µ0vc ) , the nodon operators, and one finds that back in real with the sound velocity entering, rather than the speed space the long-wavelength piece of the spin operator, of light. The external electromagnetic field is coupled to S(~x)= q exp(i~q ~x)Sq, is simply the 3-current Iµ, which can be expressed as a bi-linear of · the nodon fermions as, P 1 S(~x)= ψ† (~x)σψ (~x). (3.5) 2 ja ja ǫ0 2 Iµ = 2 2 [κν ∂ν Jµ κµ∂µ(∂ν Jν )]. (3.3) κ0vc − Spin correlation and response functions can then be com- puted from the free nodon theory. For example, the uni- Notice that this 3-current is automatically conserved: form spin susceptibility is given by ∂µIµ = 0. The order (∂J)2 terms which we have not written out ∞ explicitly are quartic in the fermion fields, and also in- χ = dE( ∂f/∂E)ρn(E), (3.6) 0 − volve two derivatives. Since describes Dirac fermions Z Lψ in 2 + 1 space-time dimensions, these quartic fermion where the nodon density of states is ρn(E) = terms are highly irrelevant, and rapidly vanish under a (const)E/vF v∆, and f(E) is a Fermi function. One finds rescaling transformation. Thus, in the absence of exter- χ T/vF v∆. There are also low energy spin excita- nal electromagnetic fields, the description of the nodal tions∼ at wavevectors which span between two different liquid phase is exceedingly simple. It consists of four nodes. The associated spin operators can readily be be neutral Dirac fermion fields – two spin polarizations re-expressed in terms of the nodon fields. For example, (α =1, 2) for each of the two pairs of nodes. the staggered magnetization operator, S~π, is Despite the free fermion description, the nodal liquid non-trivial 1 † y y † phase is highly when re-expressed in terms of S~π = ψ (τ σσ )ψ + h.c. . (3.7) the underlying electron operators. Indeed, the ψ fermion 2   operators are built from the quasiparticle operators Ψ in Notice that this operator is actually “anomalous” in the d-wave superconductor, but are electrically neutral, terms of the conserved nodon charge. We will return due to the “gauge transformation” in Eq. 2.18. More- to the effects of finite wavevector magnetic fluctuations over, in the nodal liquid phase, the Cooper pairs are and ordering in Section IV. not superconducting, but rather in a dielectric Mott- 19 In addition to carrying spin, the nodons carry energy, insulating phase, immobilized by their commensurabil- and so will contribute to the thermal transport. At finite ity with the underlying crystal lattice. Although the ψ temperature, Umklapp scattering processes (or impuri- fermions are electrically neutral, they do carry a new ties) give a finite thermal conductivity; in their absence conserved “charge”. In fact, there are four new con- the nodon thermal conductivity is infinite. served charges, since the Lagrangian is invariant under the global transformations ψ eiθjα ψ for arbi- jaα → jaα trary constant phases, θjα, with j = 1, 2 and α = 1, 2. 2. Charge response We refer to the ψ fields as nodon operators, their quanta as nodons, and the associated conserved quantities as The electrical charge properties in the nodal liquid “nodon charges”. As seen from Eq. 2.20, these con- phase are, however, somewhat trickier. Imagine chang- served charges are related to the quasiparticle current, † ing the chemical potential away from µ = 0. In terms of since Jj = (vF /2)ψj ψj . However, in the nodal liquid the dual vortex “superconductor” this corresponds to ap- phase the electrical current operator is Iµ, not Jµ, since plying an external “magnetic” field, due to the coupling Iµ = ∂ nl/∂Aµ. = 2µb. The vortices, however, are in the “Meissner” L Lµ − phase, and for µ µc the applied field will be screened out, maintaining the≤ internal field at b = 0. That is, the 1. Spin response electron density will be pinned at half-filling, until the chemical potential passes through the Mott gap for the The spin response functions in the nodal liquid are insulating system of Bosons (Cooper pairs).19 rather straightforward, since the electron spin operators Despite the presence of a charge gap, there are low en- have a simple representation in terms of the nodons. In ergy current fluctuations in the nodal liquid. Indeed, in particular, consider the spin operator for small momen- this phase the electrical current operator is Iµ, which is tum, bi-linear in terms of the nodon fermions, ψ. One might

8 imagine employing this current operator to compute the in a Mott insulating phase. This indicates a gap in the electrical conductivity in the nodal liquid. For this, one electron spectral function at the Fermi energy, of order ∗ requires computing a two-point correlator of the elec- ∆1 µc at the nodes. If µc is small relative to T , the trical current operator at zero wavevector (say in the corresponding∼ gap will show strong four-fold anisotropy x direction) I (q =0)=(ǫ /κ v2)∂2J (q = 0). But no- in momentum space, varying from of order T ∗ down to − x 0 o s t x tice that Jx(q = 0) is proportional to a globally conserved of order µc near the nodon wave-vectors. In the discus- † sion section, we comment briefly on how such a d-wave nodon charge, since Jx(~x) = (vF /2)ψ1ψ1. Thus, when the nodon number is conserved one has Ix(q = 0) = 0, pseudo-gap feature is likely enhanced when the NL is and the nodons do not contribute to the electrical con- doped in the presence of impurities. ductivity. (There will of course be a response at finite fre- quencies in the imaginary part of the conductivity from the Mott-insulating phase of the Cooper pairs.) B. Doping the Nodal Liquid When impurities or Umklapp scattering is present, however, the nodon number is no longer conserved, and We now consider the effects of doping charge into the the nodons presumably will contribute to the real part of nodal liquid phase. In a grand canonical ensemble this is the electrical conductivity, at least at finite frequencies. achieved by changing the chemical potential, µ = A0/2. Specifically, the umklapp scattering term with momen- In the dual Ginzburg-Landau description of the vortices, tum transfer 2~π is given by a chemical potential acts as an applied dual field, as seen from Eq. 2.35, since = uǫABCDψ ψ ψ ψ + h.c., (3.8) Lumklapp jA jB jC jD µ = 2µb. (3.13) where the composite index A runs over 1, 2, 3, 4 corre- L − sponding to aα = 11, 12, 21, 22. By power counting, this The dual magnetic field, b = ǫij ∂iaj , is the total electric term is irrelevant by one power of frequency. Hence, it charge in units of 2e. For a hole doping with concentra- 2 enters scaling forms in the combination u = uω. tion x, one has b = x/2a , with a0 the crystal lattice eff h i 0 According to the Kubo formula, constant. Provided the applied dual field, 2µ, is smaller than the critical field (2µc) of the Ginzburg-Landau the- 1 σ(ω) I (q =0,ω)I (q =0, ω) ory, the dual superconductor stays in the Meissner phase ∼ ω h x x − i – which is the nodal liquid phase at half-filling. But for 3 ω Jx(q =0,ω)Jx(q =0, ω) . (3.9) µ µ dual flux will penetrate the Ginzburg-Landau ∼ h − i c superconductor,≥ which corresponds to doping the nodal From scaling, we expect the latter correlation function to liquid phase. The form of the dual flux penetration will vary as ω. However, as we noted above, it actually van- depend critically on whether the dual Ginzburg-Landau ishes in the absence of Umklapp scattering; therefore it is theory is Type I or Type II. Within a mean-field treat- determined by the correction to scaling, which, naively, ment this is determined by the ratio of the dual penetra- is of the form: tion length, λv, to the dual coherence length, ξv (where 3 2 2 6 the subscript v denotes vortices). In particular, Type σ(ω) ω ω ueff u ω . (3.10) ∼ · · ∼ II behavior is expected if λv/ξv 1/√2, and Type I be- ≥ Finally, it is instructive to consider the behavior of havior otherwise. In the Ginzburg-Landau description λv the electron Green’s function, which can be accessed in determines the size of a dual flux tube, which is simply photo-emission and tunneling experiments. The electron a Cooper pair. We thus expect that λv will be roughly operator c (~x) can be conveniently decomposed in terms equal to the superconducting coherence length, ξ, which α ˚ of the nodon operators by focusing on momenta near the is perhaps 15 20A in the cuprates. On the other hand, ξ is the size of− the “vortex-core” in the dual vortex field, nodes. For example, near the node at K~ one can write, v j and presumably can be no smaller than the microscopic crystal lattice spacing, ξ 3 5A˚. This reasoning sug- iK~ j ·~x iϕ(~x)/2 v cα(~x)= e e ψj1α(~x)+ ... (3.11) ≥ − gest that λv/ξv is probably close to unity, so that either Type I or Type II behavior might be possible - and could In the nodal liquid phase, the electron Green’s function, be material dependent. We first consider such Type II G(~x, t)= c†(~x, t)c(~0, 0) factorizes as, h i doping, returning below to the case of a Type I Ginzburg- ~ ~ Landau theory. G(~x, t)= eiKj ·~x e−iϕ(~x,t)/2eiϕ(0,0)/2 ψ† (~x, t)ψ (~0, 0) . h ih j1α j1α i (3.12) 1. Type II Behavior Although the nodon correlator is a power law, falling off as x −2 and t−2, one expects the correlator over exponen- The phase diagram of a clean three-dimensional type tials| | of the pair field phase to fall off exponentially in the II superconductor is well understood.3 Above the lower nodal liquid, since the Cooper pairs (Bosons) are locked critical field, Hc1, flux tubes penetrate, and form an

9 Abrikosov flux lattice - usually triangular. As the applied b(~x) = n(~x)= n0 + δn(~x). Here, the mean pair density h i 2 field increases the flux tubes start overlapping, when their is simply n0 = x/2a0 for doping x, and δn(~x) has the pe- separation is closer than the penetration length. Upon riodicity of the Wigner crystal. This “background” field approaching the upper critical field Hc2 their cores start couples to the nodons, and from Eq. 2.35 leads to a term overlapping, the Abrikosov flux lattice disappears, and of the form, the superconductivity is destroyed. Mean field theory 1 predicts a second order transition at Hc2, but with ther- −1 † z b = b(~x) κ0 J0 = n(~x)ψj τ ψj . (3.14) mal fluctuations one expects this to become weakly first L h i 2κ0 order.20 This weak first order transition separates a flux- What is the effect of this term on the nodons? Consider lattice phase from a non-superconducting flux-liquid. first the spatially constant piece, proportional to n . This These results hold equally well for our dual Ginzburg- 0 term can be absorbed into , which contains a term of Landau superconductor, except that now the direction ψ † z L parallel to the applied field is actually imaginary time. the form, ψj vF qj τ ψj , and leads to a momentum space Moreover, the Ginzburg-Landau order parameter de- shift of the nodes, with qj qj + (n0/2vF κ0). Since → scribes quantum vortices, and the penetrating flux tubes the pair compressibility κ0 ∂n0/∂µ, the shift satisfies ∼ are Cooper pairs. Upon doping the nodal liquid with vF δq δµ, as expected from the change of the area en- ∼ µ > µc1, charge is added to the 2d system, which corre- closed by the Fermi surface upon doping. sponds to the penetration of dual magnetic flux. In this The spatially varying background density, δn(~x), dual transcription, the resulting Abrikosov flux-lattice which has the periodicity of the underlying Wigner crys- phase is a Wigner crystal of Cooper pairs, with one tal, causes a mixing between nodon states at momen- Cooper pair per real space unit cell of the lattice. We tum differing by Wigner crystal reciprocal lattice vectors. denote this paired Wigner crystal phase by PWC. Upon When the reciprocal lattice vectors are larger than the further doping, one passes via a weak first order transi- momentum cutoff, Λ, of the nodons, the δn(~x) term can- tion (at µ = µc2) into the dual flux-liquid phase. In this not scatter within the low energy nodon theory, and can phase the lattice of Cooper pairs has melted, and they be dropped. At lower pair densities, it becomes necessary are free to condense - this is the d-wave superconductor. over a range of length scales to retain the new periodicity This latter transition should occur when the spacing be- by working with nodon Bloch states, rather than plane tween dual flux tubes becomes roughly comparable to the waves. (Since only linear derivatives enter into the nodon coherence length, ξv. Experimentally, superconductivity Lagrangian, Bloch wavefunctions in the lowest band can typically sets in for x = 0.1, which corresponds to one be readily constructed.) In any event, at length scales Cooper pair for every 20 or so Cu atoms, and a mean larger than the Wigner crystal lattice spacing, the form pair separation of 4-5a0. This again suggests that λv/ξv of the effective theory of the nodons is identical to that is probably of order one. at half-filling. In the Cooper pair Wigner crystal phase, translational We thereby arrive at a description of a rather remark- symmetry is spontaneously broken. However, in a real able new phase of matter. A Paired Wigner Crystal material the Wigner crystal will have a preferred loca- (PWC) of doped Cooper pairs co-exists with neutral tion, determined by impurities and perhaps crystal fields, gapless fermionic excitations – the nodons. In this co- which will tend to pin and immobilize the Wigner crys- existing phase, which we denote as PWC/NL, low en- tal. The resulting phase should be an electrical insu- ergy spin and thermal properties will be dominated by lator. Moreover, in two-dimensions even weak impuri- the nodons. The behavior will be qualitatively similar to ties will smooth the weakly first order transition between that in the undoped nodal liquid phase. We propose that the Wigner crystal and superconducting phases. (In the this PWC/NL phase is present in the pseudo-gap region absence of impurities, long-ranged Coulomb interactions of the high Tc cuprates. preclude phase separation, so a mixed phase would result - see Type I behavior below.) A striking and unusual feature of the PWC phase, is 2. Type I Behavior that it co-exists with the nodal liquid, as we now ar- gue. With a weak (commensurate) pinning potential In a classical Type I superconductor, the applied field 3 present, the Wigner crystal phase is a dielectric. The is expelled until the critical Hc is exceeded. At this point charge response is thus essentially the same as that of there is a first order phase transition from the Meissner the undoped phase at half-filling, except with a modi- phase with all the flux expelled, to a normal metal phase fied dielectric constant and magnetic permeability. Thus in which (essentially) all the field penetrates. If a thin even with doping, it is possible to “integrate out” the film type I superconductor is placed in a perpendicular charge fluctuations described by the fields aµ, and ar- field, screening currents are unable to expel all the flux, rive at the nodal liquid Lagrangian Eq. 3.2, except with and a “mixed” or “intermediate” state occurs. In this different values of ǫ0 and µ0. The only complication is mixed phase, regions of superconductivity co-exist with that there will be a background frozen in charge den- normal metallic regions. In some cases the superconduct- sity, from the Wigner crystal of Cooper pairs, so that ing regions form stripes, but generally the lowest energy

10 configurations are determined in large part by material uniform and staggered magnetization. Focusing on the imperfections, and tend to be “history” dependent. N´eel ordering, we imagine integrating out M0 to obtain If our dual Ginzburg-Landau theory describing quan- the Lagrangian tized vortices is of type I, then similar properties are ex- pected. Specifically, as the chemical potential increases, 1 2 = Kµ ∂µN VN ( N )+ gN S~π, (4.2) the dual field – which is the Cooper pair density – re- L 2 | | − | | · mains at zero until a critical chemical potential µc is where K = K, K = K = v2K, with v the spin- reached. At this point there is a first order phase tran- 0 1 2 − s s sition, between the nodal liquid phase at half-filling, and wave velocity in the AF. Here S~π is the spin operator at momentum ~π, expressed as a bi-linear in terms of the a d-wave superconductor at finite doping, xc. At fixed electron operators as in Section III. The staggered mag- doping x < xc, phase separation is impeded by long- ranged Coulomb interactions between the Cooper pairs. netization operator can be readily re-expressed in terms The system will break apart into co-existing “micro- of the nodons as, phases” of nodal liquid and d-wave superconductivity. 1 The configuration of the co-existing “micro-phases” will S = ψ†τ yσσyψ† + h.c. , (4.3) ~π 2 be determined by a complicated competition between the   Coulomb energy and the (positive) energy of the domain given earlier in Eq. 3.7. Near any phase transitions, walls. In practice, impurities will also probably play a and for most phenomenological purposes, it is sufficient very important role. This doping scenario is similar to to take a simple form for the potential: VN ( N ) = 15 2 4 | | that envisaged by Emery and Kivelson, who have ex- rN N + uN N . The parameter rN controls the pres- tensively discussed the possibility of phase separation as a ence| or| absence| | of AF order. In mean-field theory, and mechanism for high Tc superconductivity. Unfortunately, neglecting for the moment the nodon coupling g, the with the transition being strongly first order in this case, ground state passes from long-range to short-range AF the associated physics is rather non-universal. order as rN is tuned from negative to positive. A precise ′ determination of rN , uN , and g in terms of t, t , U, x, etc. is the province of a more microscopic theory. IV. ANTIFERROMAGNETISM IN THE NODAL We note that, in principle, Eq. 4.2 allows for the pos- LIQUID sibility of incommensurate spin-density-wave ordering at wavevectors other than (π, π), which would correspond A. Effective Action for Antiferromagnetism to a state which spontaneously develops a spatially peri- odic expectation value for N. We find such ordering un- We now turn to the low-doping region of the phase likely within this model, however, and have therefore not diagram of Fig. 1. Retaining the Nodal Liquid as the included terms responsible for locking in possible higher- underlying description of the low-energy fermionic de- order commensurate magnetic wavevectors. Since incom- grees of freedom, we consider antiferromagnetic order- mensurate order seems not to be realized experimentally ing. In principle, this can arise in two ways. Antiferro- at low doping, we hope this omission is unimportant. magnetism could stem from interactions between nodons, i.e. physics below the scale T ∗. This can be modeled in principle by including simple inter-nodon interactions. B. Magnetism and Phases at Half-Filling The experimental coincidence of T ∗ with the N´eel tem- perature and magnon bandwidth J suggests that such Once we have coupled in the N´eel order parameter a separation of scales is not valid. Instead, antiferro- field, we can describe magnetic phases, in addition to magnetic correlations may exist already at (high) en- the nodal liquid and d-wave superconducting phases of ∗ ergies comparable to T . This sort of local AF am- earlier sections. Here we first focus on the situation at plitude could be captured by decoupling spin-spin (or, half-filling, where our effective field theory already de- e.g. on-site Hubbard) interactions in a microscopic model scibes a number of magnetic and non-magnetic ground with a Hubbard-Stratonovich transformation. Such a states. It will be useful to further specialize initially to decoupling introduces a conjugate field M, which inter- models with particle-hole symmetry, returning later to acts with the electron operators via a term of the form the half-filled but particle-hole asymmetric case below. † 1 σ HM = ~x M(~x) cα 2 αβcβ. Integrating out high-energy degrees of freedom· generates an effective action for M. P We expect dominant ordering tendencies at momentum 1. Particle-hole symmetric case ~π = (π, π), and so decompose

M M0 + exp(i~π ~x)N, (4.1) The full effective Lagrangian has two order parameter ∼ · fields, the N´eel order parameter, N, and the vortex com- where N and M0 are slowly-varying. The fields M0 and plex field, Φ, which is minimally coupled to a gauge field, N have the physical interpretation of the coarse-grained aµ. The N´eel order parameter is directly coupled to the

11 nodons, whereas the vortex field only sees the nodons in- in the j = 1 sector, and an identical form with qx and directly via the gauge field. Ordering of the two fields is qy interchanged for the other pair of nodes (j = 2). determined by the coefficients of the quadratic terms in In all nodon sectors there is a non-zero gap, equal to the Lagrangian, namely rN and rΦ. It will be convenient gN0. In the lower left quadrant of Fig. 3a, with rΦ nega- to plot the phase diagram at half-filling in the rN rΦ tive, this corresponds to the usual N´eel antiferromagnet. plane. The phase diagram with particle-hole symmetry− With the nodons gapped out, the only low energy exci- (λ = 0) is shown in Fig. 3a. Here we briefly discuss each tations are the spin waves of the antiferromagnet. For of the four phases. rΦ positive, in the lower right quadrant, antiferromag- ~ netism co-exists with d-wave superconductivity. In the r r~ N N d-wave superconductor, the nodons become equivalent to the d-wave quasiparticles, so that the d-wave state in this NL dSC NL dSC quadrant is rather unusual. In particular, it is a d-wave δr~ superconductor with a full single particle gap, and an ab- rφ N r φ 2 sence of nodal quasiparticles. We have denoted this with AF AF/dSC AF/NL AF/dSC λ g a subscript g – for gap – in the phase of the lower right quadrant. AF/dSC AF g Before turning to the effects of particle-hole asymme- tries, it is interesting to briefly discuss the nature of the (a) λ=0 (b) λ>0 phase transitions between the four phases in Fig. 3a. Consider first the vertical phase boundary, separating Fig. 3: Phase diagrams at half-filling for the particle- the superconducting from non-superconducting phases. hole (a) symmetric and (b) asymmetric cases. In (b), For rN negative there are no gapless nodons, and the both horizontal phase boundaries shift downward with magnet is ordered in both phases. At the transition, increasing particle/hole asymmetry parameter λ, dimin- rΦ = 0, where superconductivity develops, we can em- ishing the domain of the AF phase. ploy the dual Ginzburg-Landau theory. Equivalently, we can return to the original representation (before duality)

Consider first rN large and positive, so that N´eel order in terms of the pair field phase, ϕ, which (in the ab- 22 is not present. For rΦ negative, the vortices will con- sence of long-ranged Coulomb interactions ) is simply dense (the Meissner phase in the dual Ginzburg-Landau the classical three-dimensional XY model. The result- theory) leaving the nodons as the only low energy gap- ing transition is in the classical 3d-XY universality class. less excitations. This is the nodal liquid phase. As rΦ For rN positive, we have to worry about the presence of changes sign, the vortices will disorder, entering the “nor- gapless nodons, which might effect the nature of the su- mal” phase of the dual Ginzburg-Landau theory. This perconducting transition at rΦ = 0. However, since the phase corresponds to the d-wave superconductor, shown dual Ginzburg-Landau field Φ is only indirectly coupled in the upper right quadrant of Fig. 3a. The d-wave super- to the nodons, via the gauge field aµ, we expect that the conducting phase can be obtained in a microscopic lattice transition will still be in the 3d-XY universality class. In- model even at half-filling, by appropriately choosing the deed, power counting in d spatial dimensions about the electron interaction terms.21 Gaussian fixed point, reveals that the coupling term (of † When rN is large and negative, the model magneti- the form ∂aψ ψ) has scaling dimension (3d+1)/2. Being greater than d + 1 for all d 1, and much greater near cally orders into the antiferromagnetic N´eel phase. With ≥ particle-hole symmetry the antipodal nodes are separated the upper-critical dimension (duc = 3), one expects this by the N´eel ordering wavevector, (π, π), so the nodons are coupling to be strongly irrelevant in the two-dimensional “nested”. This opens a gap in the nodon spectrum, as case of interest. we now demonstrate. With either sign of rΦ the gauge Consider next the phase transition into the antiferro- field aµ can be integrated out, generating irrelevant four magnetic state, upon crossing the horizontal axis. For fermion nodon interaction terms, which can be dropped. either sign of rΦ, since the vortices and gauge field aµ Then, upon putting N = N0yˆ into the effective La- can be integrated out generating irrelevant four-nodon grangian, we arrive ath thei following quadratic Lagrangian interactions, the transition is described by the N´eel order- for the nodon fields, ing field coupled to the nodon bi-linear, with Lagrangian = + . This is an interesting field theory de- L Lψ LN † y † scribing Dirac fermions anomalously coupled to a fluc- nodon = ψ + gN0(ψ τ ψ + h.c.). (4.4) L L j j tuating O(3) field. Power counting about the Gaussian fixed point (free nodons and r = u = 0) reveals that This model can be readliy diagonalized with an appro- N N the coupling term is relevant in two dimensions, but if priate Bogoliubov transformation, giving energy eigen- the model is suitably generalized to higher dimensions values, becomes marginal in d = 3. This suggests an attack near four space-time dimensions, working perturbatively for 2 2 2 E1(~q)= (vF qx) + (v∆qy) + (gN0) , (4.5) small ǫ = 4 (d + 1). A complication is that there are ± − q 12 in fact three independent velocities, two for the nodons Lagrangian to arrive at a quadratic nodon Lagrangian (vF , v∆) and one for the N´eel spin field (vs). In Appendix of the form nodon + λ, with nodon given in Eq. 4.4. B we first consider the special case in which all three ve- Once again aL BogoliubovL transformationL diagonalizes the locities are set equal. This model is then a bona fide rel- quadratic form, and with non-zero λ the energy eigenval- ativistic field theory, with Lorentz invariance. A leading ues (in the j = 1 nodon sector) satisfy, E4 2AE2+B =0 order perturbative renormalization group in ǫ reveals the with, − existence of a new non-trivial fixed point. The associated 2 2 2 2 critical properties and exponents are briefly discussed in A = (gN0) + (vF qx) + (v∆qy) + λ , (4.8) Appendix B. Remarkably, as we also show in Appendix B, this relativistic fixed point is in fact linearly stable to and small deviations in the three velocities. A microscopic 2 2 B = [A 2λ ]+(2λv∆q ) . (4.9) model with different velocities scales into a relativistic − y form near criticality. As further discussed in Appendix 23 If gN0 > λ, there is no solution at E = 0, so that there is B, this model is closely related to a remarkable model a gap in the nodon spectrum. The resulting phases - an- which has has a non-trivial fixed point in D = 2 + 1 with tiferromagnet for r < 0 and co-existing gapped d-wave an exact (N = 1) space-time supersymmetry.24 Super- Φ superconductivity with antiferomagnetism for rΦ > 0 - symmetry is powerful enough to determine several exact occur for large negative r , where N is large (given by critical exponents in D = 3,23 which serve as a useful N 0 N = ( r )/2u within mean-field theory). But for check on our ǫ expansion results. 0 N N smaller r− , when gN < λ, zero energy solutions do It is finally worth mentioning the effects of the coupling N 0 exist, andp| there| are gapless nodon states present! This to the nodons, on the location of the antiferromagnetic leads to the two new phases present in Fig. 3b. ordering transition. With r positive one can safely in- N Specifically, for r negative, the new phase exhibits tegrate out the nodons. This leads to a supression of r , Φ N gapless nodon excitations co-existing with long-range an- with a “renormalized value”,r ˜ , given to leading order N tiferromagnetic order. The nodons are now incommensu- by, rate with the magnetic order (at ~π), since the zero energy 2 4 nodon state for λ>gN0 occurs at a shifted wavevector: r˜ = r (cΛ/v∆)g + O(g ). (4.6) N N − q = v−1 λ2 (gN )2 =0, (4.10) Here Λ is the high momentum cutoff on the nodons, and c x F − 0 6 a cut-off dependent constant of order one. The coupling and q = 0. This interestingp new phase, which we de- to the nodons thus tends to enhance the magnetic or- y note as AF/NL, exhibits gapless incommensurate mag- dering. Within mean-field theory, the antiferromagnetic netic fluctuations co-existing with the ~π magnons of the ordering transition will take place when the “renormal- N´eel state. ized” coefficientr ˜N = 0, as indicated in Fig. 3a. For positive rΦ, where d-wave superconductivity is present, the gapless nodon excitations are simply d-wave quasiparticles, incommensurate with the N´eel order. In 2. Effects of Particle-hole asymmetry the new phase, the incommensurate gapless d-wave su- perconductor co-exists with antiferromagnetic order, as Staying at half-filling, we next consider the effects depicted in Fig. 3b of particle-hole asymmetries, generated for example by It is worth emphasizing that with particle-hole asym- a second neighbor electron hopping term in a micro- metry present (ie. non-zero λ) the region of antiferromag- scopic square lattice model. As discussed in Section II, a netic order in the phase diagram is diminished, relative particle-hole asymmetry generates an additional term in to the case with λ = 0, as depicted in Fig. 3b. This the nodon Lagrangian of the form, occurs because incommensurate nodons are less effective at renormalizing rN . Specifically, upon integrating out = λψ†τ zψ . (4.7) Lλ j j the nodons with non-zero λ one finds the same form as in Eq. 4.6, but with c cF (λ/vF Λ), where F (X) is a In the absence of coupling to the N´eel field, this sim- monotonically decreasing→ function of X with F (0) = 1. ply causes a momentum space shift in the positions of This leads to a downward shift in the horizontal phase the nodes, by an amount δq = λ/vF . As we shall see, boundary in Fig. 3b by an amount, this has a profound effect when the nodons are coupled 2 to the N´eel order parameter, since such a shift destroys δr˜N = [1 F (λ/vF Λ)](cΛ/v∆)g . (4.11) the “nesting” of the nodons. Indeed, this leads to two − additional phases at half-filling, as shown in the phase Physically, particle-hole asymmetries, such as a second diagram for λ = 0 in Fig. 3b. neighbor hopping term, tend to frustrate and weaken the Since both additional6 phases are antiferromagnetically antiferromagnetism at ~π. With non-zero λ the nodes are ordered, we once again integrate out Φ and aµ (valid shifted off commensurablilty, and the magnetism is in- for non-zero r ), and put N = N yˆ into the effective deed weakened. This effect leads to a natural mechanism Φ h i 0

13 for the destruction of antiferromagnetism upon doping, two alternative phases in undoped cuprates. Since by as we describe in the next section. assumption we are varying only µ, and µ couples indi- Finally, we briefly discuss the nature of the phase tran- rectly to magnetism via the corresponding dual “flux” sitions between the six phases present at half-filling with in Eq. 2.35, the AF order and nodons are effected only particle-hole asymmetry. Arguments as in the previous once charge is added to the system. What is the effect subsection, suggest that the vertical phase boundary sep- of such charge doping on the AF? From Eqs. 2.24,2.35, arating the three superconducting from the three non- we see that the added charges act simply to increase the superconducting phases (at rΦ = 0) should, as before, effective particle/hole asymmetry of the nodons. In par- be in the universality class of the classical 3d-XY model. ticular, Since the nodons are incommensurate at the upper hor- x izontal phase boundary where antiferromagnetism first λeff = λ + . (4.13) 4κ a2 appears, they will decouple from the critical magnetic 0 0 fluctutations. Both of these two transitions (for posi- Eq. 4.13 is an extremely useful result. Using it, we can tive and negative rΦ) should thus be in the universality simply trade the doping x for an effective particle/hole class of the classical 3d Heisenburg model. At the lower asymmetry to determine the fate of the system from the horizontal phase boundaries, the gapless nodons disap- phase diagrams at half-filling, Fig. 3. pear. The critical properties are correctly described by the quadratic nodon Lagrangian, considered above. In AF particular, for λ> gN0, one can linearize for small mo- (a) AF/PWC AF/PWC/NL PWC/NL dSC mentum around the shifted zero energy nodes by putting 0 xc2 x δq = q v−1 λ2 (gN )2, which gives, x x − F − 0 p 1 AF/NL 2 2 (b) AF/PWC/NL PWC/NL gN0 2 2 dSC E = 1 (vF δqx) + v∆qy . (4.12) x ± − λ 0 xc2 "   ! # As the transition is approached, the velocity along the AF x-direction – i.e. perpendicular to the Fermi surface – (c) AF + dSC dSC vanishes and the nodons become quasi-1D. 0 xc x

Fig. 4: Possible phase diagrams as a function of dop- C. Doping the Antiferromagnet ing for the type II (a and b) and type I (c) scenarios. We tentatively identify the PWC/NL phase, which ex- We are now in a position to extend our understand- hibits neither magnetism or superconductivity, with the ing of doping to include AF order at half-filling, in ac- pseudogap state of the underdoped high-Tc materials. cord with experimental observations. In all high-Tc ma- terials, doping is actually achieved by chemical substitu- Consider then first the type II doping scenario. The tion/depletion of atoms between the CuO2 layers. While charge behavior is similar to that obtained when dop- it is generally believed that this process transfers charge ing the NL, Sec. III. Upon increasing µ from zero, the to the CuO2 planes, this charge transfer is not necessarily dual “flux” is first expelled from the sample, and the proportional to the chemical doping, defined as the frac- system remains undoped. Charge first enters above the tion of atoms substituted or removed. To simplify this dual “lower critical field”, µ > µc1, forming a Paired discussion, however, we shall assume that chemical dop- Wigner Crystal (PWC) with density x(µ) due to long- ing indeed corresponds to hole doping, and consider the range Coulomb interactions. Since x is small at this phase diagram as a function of hole concentration x. A point, the crystal coexists with the AF, so the actual further assumption concerns the degree of particle/hole phase for small x is an AF/PWC. As x increases, so does asymmetry. Since the composition is changing with dop- λeff , unbinding the nodons into the AF/PWC/NL. This ing, the parameter λ should in general be a function of can be understood from the evolution of the phase dia- x (in fact, we expect the asymmetry to increase with x). grams at half-filling as a function of λ, as shown in Fig. 3. To simplify the discussion, we shall further assume that As x increases further, the NL and AF become increas- any explicit dependence of λ on x is weak, and therefore ingly incommensurate, and the energy gain from their treat the effects of doping solely through the chemical coupling is eventually reduced sufficiently to destroy the potential µ. AF order in a transition to a PWC/NL phase - again At half-filling, the system sustains long-range AF or- a feature of the phase diagrams at half-filling. Finally, der. In general, we expect a non-zero particle/hole asym- when x xc2, the upper critical field is reached and metry, so that this could correspond to either the AF or the crystal≥ melts into the dSC phase. This progression AF/NL state, the latter occuring if λ is sufficiently large. is shown schematically in Fig. 4a. An alternate type II We do not believe current experiments distinguish the doping scenario, shown in Fig. 4b, is that the system is an

14 AF/NL at half-filling, in which case the phase diagram is rubric, which is expressed in (1.1), is the central theme of unchanged except for the absence of the AF/PWC phase. Zhang’s SO(5) theory.32 However, there is a direct tran- Another possibility is type I doping. Because this sition from AF to dSC in the SO(5) theory, whereas the involves a strong first-order transition in the absence Nodal Liquid intervenes in our theory. There is a further of Coulomb interactions, the mixed (micro-phase sepa- important distinction, namely, that our theory focuses rated) state could occur as a coexistence between a num- on the zero-temperature quantum phase transitions of ber of different phases. The simplest phase diagram in- the high-Tc materials. This is one reason why our theory cludes only coexistence between the AF and pure dSC accords primary importance to the low-energy fermionic as in Fig. 4c. Because the physics of the mixed state is degrees of freedom. Finally, our prediction of phase sepa- highly non-universal, we do not discuss it further here. ration at the dSC transition in the type I scenario as well as our interpretation of T ∗ echoes the ideas of Emery and Kivelson.15 V. DISCUSSION There are a number of important issues which we have not addressed in this paper. By restricting our atten- tion to the region underneath the dashed line in Fig. 1, The main result of this paper is the Lagrangian, we have skirted one of the most controversial questions Eqs. 1.1-1.2, which describes the Nodal Liquid phase, in this field: what mechanism drives pair formation at its interaction with external electromagnetic fields, and this scale? Presumably this physics must be understood transitions between it and the antiferromagnetic (AF) for progress to be made on the part of the phase dia- and superconducting (dSC) phases. This Lagrangian fol- gram above and to the right of the dashed line. This lows directly from disordering the d-wave superconduc- would require an investigation complementary (but per- tor. It implies that, in the underdoped region, low-energy haps orthogonal in spirit) to ours. Also, our discussion fermionic degrees of freedom are located solely at four of transport was necessarily incomplete because finite- isolated (Dirac) points in the Brillouin zone – a hypothe- temperature transport can be particularly subtle (for re- sis which is strongly supported by ARPES6,7, NMR9,10, cent examples of this, see Ref. 33) and the transport prop- optical conductivity11,12, and other experiments2. The erties of the Nodal Liquid deserve a thorough exposition main consequences of Eqs. 1.1-1.2 are (1) the predic- of their own, which we defer. Moreover, a pure sample tion of a new zero-temperature phase, the Nodal Liq- would melt at a finite temperature phase transition, al- uid, which interpolates between the AF and dSC phases; though this transition would be rounded by impurities. (2) a quantitative description of charge and spin dynam- The melting temperature is expected to vanish upon ap- ics in this phase; (3) specific predictions for the critical proaching either zero doping or the PWC/NL to dSC behavior at the AF and dSC ordering transitions; and, quantum phase transition, and therefore has maximum above all, (4) a coherent weak-coupling framework – with at some intermediate x in the underdoped regime. the Nodal Liquid as its foundation – for understanding The effects of disorder are also quite subtle, and war- the underdoped side of the high-Tc phase diagram. We rant a full and separate treatment. Nevertheless, a few tentatively identify the coexisting Paired Wigner Crys- comments are germane to this discussion. The first, and tal/Nodal Liquid (PWC/NL) phase as the pseudo-gap most basic, is that disorder plays a significant role in the state in a hypothetical disorder-free underdoped cuprate. physics of the cuprates. Even in YBa2Cu3O7−δ, which In the real materials, however, disorder will play a role, is believed to be cleaner than, say, La2−xSrxCuO4, as we briefly discuss below. doping cannot help but introduce disorder. According Our description of this part of the phase diagram en- to standard arguments,34 first-order phase transitions joys kinship with several other approaches. The Nodal will be driven second-order by arbitrarily weak disor- Liquid phase is reminiscent of the π-flux state25 and the der in two dimensions. In particular, we expect the SU(2) MFT staggered-flux state,26 and is the d-wave PWC/NL dSC transition to be second order with im- analog of the short-range27 resonating valence bond spin- purities present.→ Moreover, based on the irrelevance (in liquid state.28 These states also involve neutral Dirac the technical sense) of the coupling between the nodons fermions interacting with a gauge field, but the cou- and the superconducting phase in the clean case, we pling to electromagnetic fields, the coupling between the suspect this transition may be in the same universality fermions and the gauge fields, the bosonic charged de- class as the superconductor–insulator transition. This grees of freedom, and the evolution with doping are all could explain the experiments of Fukuzumi, et al. on 29,30 35 rather different in the Nodal Liquid. Several authors YBa2Cu3−yZnyO7−δ. The disorder will also have an have conjectured that the lightly-doped 3-leg Hubbard effect on the phases themselves. For instance, transport ladder might serve as a paradigm for the underdoped and spin dynamics in the Nodal Liquid will be influ- cuprates and Furukawa and Rice31 have tried to substan- enced by disorder. Finally, we note that disorder will tiate these claims with weak-coupling RG calculations on transform the PWC into a Bose Glass (BG). One con- partially nested Fermi liquids. The Nodal Liquid con- sequence would be power-law suppression (rather than cretely realizes the attractive features of this proposal. a hard gap) of the low frequency electron spectral func- The basic idea of bringing AF and dSC under the same tion at the nodes in the BG/NL phase because the BG is

15 compressible.19 Having adopted a panoramic view in the angular brackets ~x~x′ under the two sums indicate sums preceeding section, we can afford, in closing, to narrow over all nearest-neighborh i pairs of sites. our focus to the BG/NL (and the PWC/NL from which To derive an effective field theory, it is convenient to it descends) because it is our candidate for the T = 0 use an imaginary time path integral formulation. In this pseudo-gap phase: a phase without a Fermi surface or case one studies the partition function Z = Tre−H/T , long-range order but possessing low-energy fermionic ex- where T is the temperature. It can be represented using citations centered about four points in the Brillouin zone. Grassman coherent states as

Z = [dc][dc]e−S, (A3) ACKNOWLEDGMENTS Z where the Euclidean action is We are extremely grateful to Doug Scalapino for nu- merous helpful conversations. MPAF would like to thank S = dτ cα(~x)∂τ cα(~x)+ H[c, c] . (A4) Steve Girvin for illuminating discussions on d-wave quasi- Z ( ~x ) particles. This work has been supported by the National X Science Foundation under grants No. PHY94-07194, We consider here only T = 0, for which the τ integration DMR94-00142 and DMR95-28578. domain is infinite. The possibility of dSC and AF phases can be entertained by decoupling the above action using Hubbard-Stratonovich transformations. One finds that APPENDIX A: MICROSCOPIC APPROACH Z = [dc][dc][dM][d∆][d∆]e−S1 , (A5) Z In this appendix, we describe techniques to derive the with S = dτ[ c (~x)∂ c (~x)+ H ]. The effective effective field theory from some specific microscopic mod- 1 ~x α τ α eff Hamiltonian can be decomposed into Heff = Hqp+HM + els. As our purpose is primarily phenomenological, we R P H∆, with H = H + H , and will consider one of the simplest models which develops qp 0 int antiferromagnetism and d-wave superconductivity. This H = t c† (~x)c (~x′) + h.c. + µ n(~x) (A6) is a square lattice extended Hubbard model with nearest- 0 − α α h~x~x′i ~x neighbor hopping t, on-site electron-electron repulsion U X   X and nearest neighbor attraction V (we emphasize that Hint = M(~x) S(~x) this model is not realistic, but is chosen for illustrative − · X~x purposes), i.e. αβ ′ αβ ′ + ∆~x~x cβ(~x )cα(~x) + ∆~x~x′ cα(~x)cβ(~x) , (A7) ′ † † ′ h~x~x i H[c ,c]= t cα(~x)cα(~x ) + h.c. +˜µ n(~x) X h i − ′ 3 h~x~x i ~x H = M(~x) 2, (A8) X   X M 8U | | +U [n(~x)]2 V n(~x)n(~x′). (A1) ~x − X h~x~x′i 1 αβ αβ H∆ = ∆ ′ ∆ ′ . (A9) X V ~x~x ~x~x h~x~x′i † X Here n(~x)= cα(~x)cα(~x), andµ ˜ is the microscopic chem- ical potential, and we have neglected to include terms Eqs. A6-A9 form a basis for studying the original ex- (e.g. a second-neighbor hopping t′) which break parti- tended Hubbard model. Following the philosophy of cle/hole symmetry at half-filling. Also, for simplicity we Sec. II A , we imagine integrating out high-frequency measure distances in units of the lattice spacing. The modes in the functional integral to arrive at an effec- tendency toward AF and dSC states can brought out by tive field theory for the low-lying degrees of freedom. In using the identities [n(~x)]2 = 4/3[S(~x)]2 +2n(~x) and the process, one will generate dynamics for the order pa- ′ † † ′ ′ − ′ rameter ∆ and the magnetization M. For the most part, n(~x)n(~x ) = cα(~x)cβ(~x )cβ(~x )cα(~x), for ~x = ~x . The Hamiltonian can be rewritten as 6 symmetry considerations require the corresponding La- grangians to take the forms given in Sec. II A and Sec. IV, H[c†,c]= t c† (~x)c (~x′) + h.c. + µ n(~x) so we choose not to complicate the presentation by ex- − α α h~x~x′i ~x plicitly performing these integrations (e.g. diagrammat- X   X ically). 4 2 † † ′ ′ U S(~x) V cα(~x)cβ(~x )cβ(~x )cα(~x), (A2) One subtle point in the analysis of Sec. II A, how- −3 | | − ′ X~x hX~x~x i ever, does warrant a more careful treatment. This is the coupling of the nodons to the superconducting phase- where we defined a shifted chemical potential µ, and have gradient, from which follows the expressions for the neglected an unimportant constant. As usual, the lattice quasiparticle current, Eqs. 2.22-2.23. In Sec. IIA, we de- 1 † σ spin operator is defined by S(~x)= 2 cα(~x) αβcβ(~x). The rive these using the “symmetric” prescription of Eq. 2.14.

16 We now show that the currents are indeed obtained cor- As expected, the θ field is massive, and can be integrated rectly using this prescription. out. Equivalently, one minimizes 2 with respect to θ. αβ H ′ This process restores isotropy and gives We first specialize to the case of singlet pairing, ∆~x~x = ′ ∆(~x, ~x )(δα↑δβ↓ δα↓δβ↑). Since ∆ lives on the bonds, it γ − (∂ ϕ)2. (A15) is convenient to associate two such fields with each site H2 → 8 j in the square lattice, i.e. This is just the spatial component of the superfluid La- 2 ∆1(~x) ∆(~x, ~x +ˆe1), (A10) grangian, Eq. 2.17, with γ = vc κ0. The corresponding ≡ time component can be obtained similarly. ∆2(~x) ∆(~x, ~x +ˆe2), (A11) ≡ Finally, consider the coupling of of the phase ϕ to the wheree ˆ1, eˆ2 are unit vectors along the a and b axes of the fermionic quasi-particle operators. To study this, we square lattice, respectively. Note that at this point, we take for simplicity µ = M = 0. Using the definitions have specified no particular relation between ∆1 and ∆2, in Eqs. A10-A11, the coupling term in Eq. A7 can be so that the model has the potential both for d-wave and s- rewritten wave pairing. To distinguish them, we must consider the form of the effective action for ∆, ∆ generated upon in- † † Hint = ∆j (~x) c↑(~x)c↓(~x +ˆej) + h.c. , (A16) tegrating out the high-energy modes. By symmetry, the ( − ↑↔↓ ) j,~x h i simplest local allowed additional term on the lattice is X a sum of U(1)-invariant two-bond products around each where the sum includes all lattice sites and j =1, 2. To lattice site, which can be written as S = dτ H , with ease comparison with Sec. II A, we have returned now to 2 2 † a Hamiltonian formalism, replacing cα by cα. We are now γ˜ R in a position to take the continuum limit. In this case, it H = ∆ (~x)∆ (~x)+ ∆ (~x)∆ (~x eˆ ) 2 8 1 2 2 1 − 1 suffices to neglect the massive θ mode, and simply take ~x  iϕ X ∆1 = ∆2 = ∆0e . For agreement with Sec. IIA, we − define v∆ = 2√2∆0, or ∆1 = ∆2 = ∆˜ /2√2. In addi- +∆1(~x eˆ1)∆2(~x eˆ2)+ ∆2(~x eˆ2)∆1(~x)+c.c. . − − − tion, we take the continuum limit− of the electron fields,  (A12) using the decompositions † † x+y x+y † x−y x−y Of course, the actual quadratic ∆∆ interaction terms c Ψ111i Ψ122( i) +Ψ211( i) Ψ222i , ↑ ∼ − − − − generated from the high-energy degrees of freedom will c† Ψ† ix+y +Ψ ( i)x+y +Ψ† ( i)x−y +Ψ ix−y, be much more complex. However, since the general form ↓ ∼ 112 121 − 212 − 221 of the long-wavelength effective action is dictated by and the hermitian conjugates of these equations. In- symmetry, this example suffices for illustrative purposes. serting these into Eq. A16, gradient-expanding the Ψ Writing ∆ = ∆ eiφj , Eq. A12 becomes fields, and rotating 45 degrees to x y coordinates along j 0 − the (π, π) and ( π, π) directions, one obtains Hint = γ 2 − H = cos(φ (~x) φ (~x)) + cos(φ (~x) φ (~x eˆ )) d x int, with 2 4 1 − 2 2 − 1 − 1 H ~x  R ˜ X ∆ † + † + int = Ψ1τ i∂yΨ1 (i∂yΨ1)τ Ψ1 + h.c. + cos(φ1(~x eˆ1) φ2(~x eˆ2)) + cos(φ2(~x eˆ2) φ1(~x)) , H " 2 − # − − − − −    +(1 2, x y). (A17) (A13) ↔ ↔ ˜ with γ = (∆ )2γ˜. We assume γ > 0, in which case this in- This form is identical to the ∆ term in Eq. 2.10 when 0 ˜ teraction favors a relative phase difference of φ φ = π, the order parameter ∆ is constant, but the symmetric 1 − 2 i.e. d 2 2 order. placement of derivatives is important in the presence of x −y ˜ iϕ We now turn to the fluctuations around the uniform phase gradients. In particular, now let ∆ = v∆e and dSC state. To do so, we let φ = ϕ, φ = ϕ + θ + π. integrate by parts to transfer the derivative in the second 1 2 † ˜ The phase ϕ is precisely the order parameter phase in- term from the Ψ to the ∆Ψ combination. Then, using troduced in Sec. II A. The other variable θ represents an- the operator identity other branch of massive fluctuations around the d-wave 1 eiϕi∂ + i∂ eiϕ = eiϕ/2i∂ eiϕ/2, (A18) state. We can thus assume θ 1, and that ϕ is slowly 2 y y y varying, i.e. ∂ ϕ 1. Under≪ this assumption, we can j one obtains
take the continuum≪ limit and replace the positional sum 2 † s isϕ/2 isϕ/2 in Eq. A13 by an integration. This gives H2 = d x 2, = Ψ [v∆τ e (i∂ )e ]Ψ + (1 2; x y). H Hint 1 y 1 ↔ ↔ with s=± R X θ2 1 1 θ (A19) = γ + (∂ ϕ)2 ∂ ϕ∂ ϕ + (∂ ∂ )ϕ . H2 2 4 j − 4 x y 2 x − y   Eq. A19 is identical to the symmetrized form of the (A14) phase-quasiparticle interaction hypothesized in Eq. 2.15.

17 APPENDIX B: RENORMALIZATION GROUP if we were truly interested in the vicinity of four dimen- ANALYSIS sions, but our choice probably makes more sense given that we are eventually concerned with ǫ 1. In any → In this appendix, we present some details of the RG case, the difference is fairly trivial: factors of 2 would be calculations for the transition at half-filling with parti- replaced by factors of 4 in traces over the 4D γ-matrices. cle/hole symmetry between the nodal liquid phase and With our convention, we obtain the following one-loop the antiferromagnet. As discussed in Section IV, this flow equations: transition is described by the same field theory as the du phase transition between the d-wave superconducting = ǫu 44u2 32λu + 32λ2 phase and the phase with co-existing antiferromagent and dl − − dλ gapped d-wave order. A striking feature of this transi- = ǫλ 20λ2 tion, which we access perturbatively in an ǫ =4 (d + 1) dl − − drN 1 expansion, is that it is Lorentz-invariant. Although the rN = (2 20u 16λ)rN model possesses three independent velocities, the differ- dl ≡ ν − − daj ences between these scale to zero at the critical point. = 12λaj, (B4) dl − The full Lagrangian is given by, 2 † z x where λ = g . N and ψ have anomalous dimensions = ψ1 (i∂t vF τ i∂x v∆τ i∂y) ψ1 8λ and 3λ, respectively. These flow equtions have a fixed L † − z − x + ψ (i∂t v∆τ i∂x vF τ i∂y) ψ2 √ 2 − − point at O(ǫ): λ = ǫ/20, u = ǫ 371/440, rN = 0, aj = 0. 1 2 2 2 2 This fixed point has the following interesting features: + Kµ ∂µN rN N u N 2 | | − − (1) ‘Relativistic Invariance.’ Since the velocity differ- + gN [ψ†τ yσσyψ† + c.c.], (B1) ·
ences scale to zero according to (B4), all physically mea- surable quantities are a function of x2+y2 t2 in the units where we have suppressed the spin subscripts α, β on the which we have chosen, or, upon restoring− the velocities, σ Pauli matrices, and the particle/hole subscripts a,b on 2 2 2 2 2 (x + y ) vs t . the τ Pauli matrices. Here K0 = K and Kj = v K, for − s (2) An antiferromagnetic correlation length which di- j = 1, 2. Notice that this model has three independent− verges as velocities, vF , v∆ and vs. ∗ ∗ Now, we can rescale the N field to set K = 1, and c −1/(2−20u −16λ ) ξ rN rN (B5) rescale ~x to set vs = 1. The Lagrangian can then be ∼ | − | rewritten as: as the transition is approached. = ψ† (i∂ τ zi∂ τ xi∂ ) ψ (3) Critical correlation functions with the following L jα t − x − y jα 1 1 2 power-law decays + (∂ N)2 (∂ N)2 r N2 u N2 2 t − 2 j − N − N † y y † δij + g [ψ τ σσ ψ + c.c.]
Ni(x, t)Nj (0, 0) 1 ∗ † · z x h i∼ ~x2 t2 2 −16λ ψ1 (a1τ i∂x + a2τ i∂y) ψ1 | − | − † z x † 1 ψ (a2τ i∂x + a1τ i∂y) ψ2 (B2) ψ (x, t)ψ(0, 0) ∗ (B6) − 2 h i∼ ~x2 t2 1−6λ | − | where As a check on the reliability of the ǫ-expansion, we vF 23 a1 = 1, consider the following related model : vs − v∆ † z x a2 = 1. (B3) = ψ (i∂ τ i∂ τ i∂ ) ψ v t x y s − L 1 − 1 − + ∂ Φ 2 ∂ Φ 2 u Φ 4 + g[ψ†τ yψ† +c.c.], (B7) Power counting about the Gaussian theory reveals that 2| t | − 2| j | − | | both u and g2 are relevant in D = 2+1 = 3 space-time di- mensions, but become marginal in D = 4. To implement where ψ is now a single two-component spinor, and Φ is a a perturbative RG calculation thus requires continung complex field. Using the same ǫ-expansion procedure, we ∗ ∗ the model above D = 3. This is a little tricky, due to the find a fixed point at λ = u = ǫ/6; at this fixed point, Pauli-matrix algebra. One approach is to dimensionally the fields Φ and ψ have anomalous dimension ǫ/6. It is continue the loop integrals, but leave the 3D Pauli-matrix a remarkable and fortunate fact that this model exhibits algebra unchanged. This turns out to be equivalent to in- N = 1 supersymmetry. As a result, the existence of a troducing into the Dirac equations an extra Pauli matrix, fixed point is guaranteed and the scaling dimensions of ψ τ y, multiplied by another (3rd) spatial dimension. Alter- and Φ can be determined exactly, in agreement with the natively, one could replace the two-dimensional Pauli ma- ǫ expansion. trices by the 4 dimensional γ-matrices, appropriate for 4D spinors. This− latter procedure would be more correct

18 1 J. G. Bednorz and K. A. M¨uller, Z. Phys. B 64, 189 (1986). (1989). 2 For a recent review, see M. B. Maple, cond-mat/9802202 15 S. A. Kivelson and V. J. Emery, Synthetic Metals 80, 151 (unpublished). (1996); V.J. Emery and S.A. Kivelson, Phys. Rev. Lett. 3 Introduction to Superconductivity by M. Tinkham, Mc- 74, 3253 (1995); and references therein. Graw Hill (New York) 1996. 16 See Theory of Superconductivity by J. Schrieffer, Benjamin- 4 D. A. Wollman, D. J. Van Harlingen, J. Giapintzakis, and Cummings (1983). D. M. Ginsberg, Phys. Rev. Lett. 74, 797 (1995). 17 D.R. Nelson and B.I. Halperin, Phys. Rev. B19, 2457 5 J. R. Kirtley, C. C. Tsuei, J. Z. Sun, C. C. Chi, L. S. Yu- (1979). Jahnes, A. Gupta, M. Rupp, and M. B. Ketchen, Nature 18 P.A. Lee and T.V. Ramakrishnan, Rev. Mod. Phys. 57, 373, 225 (1995). 287 (1985). 6 A. G. Loeser, Z.-X. Shen, D. S. Dessau, D. S. Marshall, C. 19 M.P.A. Fisher, P.B. Weichman, G. Grinstein and D.S. H. Park, P. Fournier, and A. Kapitulnik, Science 273, 325 Fisher, Phys. Rev. B40, 546 (1989). (1996); and references therein. 20 D.R. Nelson, Phys. Rev. B60, 1973 (1988). 7 H. Ding, T. Yokoya, J. C. Campuzano, T. Takahashi, M. 21 F. F. Assaad, M Imada, and D. J. Scalapino, Phys. Rev. B Randeria, M. R. Norman, T. Mochiku, K. Kadowaki, and 56, 15001 (1997). J. Giapintzakis, Nature 382, 51 (1996); and references 22 For a discussion of Coulomb effects see M.P.A. Fisher and therein. G. Grinstein, Phys. Rev. Lett. 60, 208 (1988). 8 B. Batlogg, H. Y. Hwang, H. Takagi, R. J. Cava, H. L. Rao, 23 N. Seiberg, private communication (1997). and J. Kwo, Physica 235-240C, 130 (1994); and references 24 See Supersymmetry and Supergravity by J. Wess and J. therein. Bagger, Princeton University Press, Princeton (1983). 9 W. W. Warren, Jr., R. E. Walstedt, G. F. Brennert, R. J. 25 I. Affleck and J.B. Marston, Phys. Rev. B 37, 3774 (1988); Cava, R. Tycko, R. F. Bell, and G. Dabbagh, Phys. Rev. Phys. Rev. B 39, 11538 (1989). Lett. 62, 1193 (1989) 26 X.G. Wen and P.A. Lee, Phys. Rev. Lett. 76, 503 (1996). 10 M. Takigawa, A. P. Reyes, P. C. Hammel, J. D. Thompson, 27 S.A. Kivelson, D.S. Rokhsar, J.P. Sethna, Phys. Rev. B 35, R. H. Heffner, Z. Fisk, and K. C. Ott, Phys. Rev. B43, 247 8865 (1987). (1991). 28 P.W. Anderson, Science 235, 1196 (1987). 11 C. C. Homes, T. Timusk, R. Liang, D. A. Bonn, and W. 29 T.M. Rice, S. Haas, M. Sigrist, and F.C. Zhang, Phys. Rev. N. Hardy, Phys. Rev. Lett. 71, 1645 (1993). B56, 14655 (1997). 12 A. V. Puchkov, P. Fournier, D. N. Basov, T. Timusk, A. 30 S.R. White and D.J. Scalapino, Phys. Rev. B57, 3031 Kapitulnik, and N. N. Kolesnikov, Phys. Rev. Lett. 77, (1998). 3212 (1996). 31 N. Furukawa and T.M. Rice, cond-mat/9803045. 13 M. Peskin, Ann. Phys. 113, 122 (1978); P.O. Thomas and 32 S.C. Zhang, Science 275, 1089 (1997). M. Stone, Nucl. Phys. B144, 513 (1978); X.G. Wen and A. 33 K. Damle and S. Sachdev, Physical Review B56, 8714 Zee, Int. J. Mod. Phys. B 4, 437 (1990). (1997). 14 C. Dasgupta and B.I. Halperin, Phys. Rev. Lett. 47, 1556 34 Y. Imry and S.-K. Ma, Phys. Rev. Lett. 35, 1399 (1975). (1981); M.P.A. Fisher and D.H. Lee, Phys. Rev. B39, 2756 35 Y. Fukuzumi, et al. Phys. Rev. Lett. 76, 684 (1996).

19