1 Spin-charge separation: From one hole to finite doping Z.Y. Weng, D.N. Sheng, and C.S. Ting a aTexas Center for Superconductivity, University of Houston, Houston, TX 77204, U.S.A. In the presence of nonlocal phase shift effects, a quasiparticle can remain topologically stable even in a spin- charge separation state due to the confinement effect introduced by the phase shifts at finite doping. True deconfinement only happens in the zero-doping limit where a bare hole can lose its integrity and decay into holon and spinon elementary excitations. The Fermi surface structure is completely different in these two cases, from a large band-structure-like one to four Fermi points in one-hole case, and we argue that the so-called underdoped regime actually corresponds to a situation in between. Spin-charge separation idea [1] has become a operator is expressed as follows very basic concept in understanding the doped ˆ c = h†b eiΘiσ (1) Mott-Hubbard insulator models related to high- iσ i iσ † Tc cuprates. It states that the system has two in- where holon hi and spinon biσ are both bosonic dependent elementary excitations, neutral spinon fields, satisfying the no-double-occupancy con- † † and spinless holon, respectively, as opposed to a straint hi hi + σ biσbiσ = 1. These spinon and single elementary excitation in conventional met- holon fields describeP the elementary excitations in als — the quasiparticle that carries both spin the mean-field state [3] which is controlled by a and charge quantum numbers. In literature, it single bosonic RVB order parameter ∆s and re- has been usually assumed that the quasiparti- duces to the Schwinger-boson mean-field state [4] cle excitation would no longer be stable in such in the limit of half-filling. a spin-charge separation state and should decay The key feature in the bosonization formulation iΘˆ i iΘstring into more elementary spinon and holon excita- (1) is the phase shift field e iσ ( σ) e iσ tions once being created, e.g., by injecting a bare string 1 b h ≡ − b with Θiσ 2 Φi σΦi , in which Φi = hole into the system as illustrated in Fig. 1(a). ≡ b − h h l=6 i θi(l) α αnlα 1 , Φi = l=6 i θi(l)nl , But in this work we will show an example that − Pand θi(l) =P Imln(zi zl). It isP a vortex-like the fate of the quasiparticle can be drastically dif- phase field which precisely− keeps track of the sin- ferent from what has been previously perceived. gular phases (signs) involved in the doped t J Specifically, based on a consistent spin-charge model known as the phase string effect [2]. − separation theory we find that a quasiparticle ex- The decomposition (1) implies that in order for citation does not simply break up into a pair of a quasiparticle created by ciσ to decay into holon- free holon and spinon and disappear in the metal- spinon elementary excitations, a phase shift Θˆ iσ lic regime. On the contrary, it remains topologi- must take place in the background. Since Θˆ iσ is cally stable in a form of spinon-holon bound state basically an infinite-body operator in the thermo- whose symmetry is fundamentally different from dynamic limit, the dissolution of a quasiparticle arXiv:cond-mat/0101411v1 [cond-mat.str-el] 26 Jan 2001 a pair of simple spinon and holon as sketched in would take virtually infinite time to realize under Fig. 1(b). a local perturbation which means that a quasi- Such a spin-charge separation theory [2,3] is particle will practically remain a stable and inde- based on a generalized slave-particle representa- pendent excitation even in a background of holon- tion of the t J model in which the electron (hole) iΘˆ − spin sea. In fact, it can be shown [5] that e iσ , due to its vorticities, introduces a new “angular- momentum-like” quantum number to the quasi- 2 involves an infinite body of them. Upon detailed examination, it is found [5] that the holon field † i b σ h 2 Φi −i 2 Φi hi has to be bound to e while biσ to e ˆ (a) in (1). But while eiΘiσ is well-defined, these two phase factors are not single-valued by themselves except in the zero doping limit. Consequently, at finite doping the holon and spinon constituents are generally confined together due to their bind- ˆ ing to eiΘiσ . Involving infinite-body holons and spinons, the quasiparticle excitation has to be treated more carefully before applying the mean-field approx- imation. Using the exact Hamiltonian to de- (b) termine the equation of motion i∂tciσ(t) = − [Ht−J ,ciσ(t)] and then introducing the mean-field order parameter ∆s in the high order terms, one finds [5] Figure 1. Quasiparticle deconfinement (a) and confinement (b) due to the phase shift field i∂tciσ(t) teff clσ + µciσ − ≈ l=NNX(i) 3 s i ˆ † † iσAh J∆ e Θiσ h b e il + ... (3) −8 i l−σ particle such that the excitation is orthogonal to l=NNX(i) h i the ground state as well as those states of elemen- t tary holon and spinon excitations. Therefore, the where teff = 2 (1 + δ), µ is the chemical poten- h h phase shift field not only plays a role leading to tial. Ail is the difference of Φi /2 at site i and a non-Fermi liquid (with the quasiparticle weight l. Note that without the “scattering” term, (3) Z = 0) as conjectured [1,6] by Anderson on a gen- would give rise to a bare quasiparticle kinetic en- eral ground, but also attaches a distinct quantum ergy ǫk = 2teff (cos kx + cos ky). The “scat- − number to the quasiparticle to ensure its integrity tering” effect in the second line of (3) actually in the metallic phase. corresponds to the decaying process of the quasi- To further understand the phase shift effect, let particle into holon and spinon constituents. But ′ iΘˆ iΘˆ iσ us construct Ψ e iσ ΨG where ΨG is the the presence of e prevents a real decay since | i ≡ | i | i ground state of Ht−J . Then by computing its such a vortex field would cost a logarithmically energy cost one can show [5] divergent energy if being left alone as discussed ′ ′ above. Nonetheless, at Ψ Ht−J Ψ ΨG Ht−J ΨG ln L (2) h | | i − h | | i∝ E > E + E (4) which diverges logarithmically with the sam- quasiparticle holon spinon ple length scale L. But the quasiparticle state a strong signature of spin-charge separation is ex- ciσ ΨG should cost only a finite energy relative pected to show up in the spectral function as the | i to the ground-state energy. It thus imples that logarithmic potential is not important in short- the holon and spinon constituents on the r.h.s. range, high-energy regime. of (1) can no longer behave as mean-field free el- In the ground state, the bosonic holons are † ementary excitations: They have to absorb the Bose condensed with hi = h0 √δ and the d- ˆ h i ∼ effect of the vortex-like phase shift eiΘiσ to re- wave superconducting order parameter ∆SC = 0 sult in the finiteness of the quasiparticle energy. [3]. Based on (3), one can find a coherent con-6 Clearly here one has to go beyond the mean-field tribution from the “scattering” term. The de- treatment of individual holon and spinon as ciσ composition (1) may be rewritten in two parts: 3 ′ iΘˆ ciσ = h aiσ + c where aiσ biσe iσ and where Tc = 0. Namely 0 iσ ≡ ′ † iΘˆ iσ † † ciσ = (: hi :)biσ e with : hi : hi h0. Cor- 2∆k(T = 0) respondingly, the single-electron≡ propagator− may (7) Tc → ∞ be expressed as at δ 0, whereas the BCS theory predicts a 2 ′ → Ge = h0Ga + Ge, (5) constant. The quasiparticle thus gains a “coherent” part where one has [5] h0a which should behave similarly to the con- 2 2 ventional quasiparticle in the BCS theory as it 2 k 2 uk vk h0Ga( ,ω) h0 + . (6) does not further decay at Ek < Es. In this ∼ ω Ek ω + Ek − sense, the quasiparticle partially restores its co- 2 2 herence in the superconducting state. Such a co- with Ek = (ǫk µ) + ∆k , ∆k = SC − | | herent part will disappear as a result of vanish- 3 ∆k+q p 4 J q Γq h2 in which Γq = cos qx +cos qy. ing superfluid density. The total spectral func- 0 P And like in the BCS theory, u2 = [1 + (ǫ tion as the imaginary part of Ge can be written k k k 2 k ′ k 2 − as Ae( ,ω) = h0Aa( ,ω)+ Ae( ,ω) so that at µ)/Ek]/2 and vk = [1 (ǫk µ)/Ek]/2. − − h0 0, even though ∆k does not scale with h0, The large “Fermi surface” is defined by ǫk = µ → 2 the superconducting coherent part h Aa vanishes and ∆k then represents the energy gap opened 0 at the Fermi surface which is d-wave-like as illus- altogether, with Ae reduces to the normal part A′ at T >T . trated in Fig. 2 by the “V” shape lines. Note e c Destruction of Fermi surface: Deconfinement that in the ground state, there also exists an in- of quasiparticle. The existence of a large Fermi dependent discrete spinon excitation level [3] at surface, coinciding with the non-interacting band- Es δJ which leads to Eg = 2Es 41meV (if J∼ 100meV ) magnetic peak at ∼δ 0.14.