cond-mat0605437/tj.tex
1: %LATEX FILE
2: 
3: \documentclass[aps,prb,twocolumn,superscriptaddress]{revtex4}
4: \usepackage{graphicx}
5: \usepackage{bm}
6: \usepackage{epsfig} 
7: \usepackage{amsfonts}
8: \usepackage{amsmath}
9: 
10: \allowdisplaybreaks[1] 
11: 
12: \newcommand*{\be}{\begin{equation}}
13: \newcommand*{\ee}{\end{equation}}
14: \newcommand*{\bea}{\begin{eqnarray}}
15: \newcommand*{\eea}{\end{eqnarray}}
16: \newcommand*{\vs}[1]{\vspace{#1 mm}}
17: \newcommand*{\sd}{^{\dagger}}
18: \newcommand{\up}{\uparrow}
19: \newcommand{\down}{\downarrow}
20: \def\abs#1{| #1 |}
21: \def\bra#1{\langle #1|}
22: \def\ket#1{|#1\rangle}
23: \def\p{\partial}
24: 
25: \def\ii{\'{\i}}
26: 
27: 
28: \begin{document}
29: 
30: \title{The short-range correlations of a doped Mott insulator}
31: 
32: \author{Tiago C. Ribeiro}
33: \affiliation{Department of Physics, University of California, Berkeley, California 94720, USA}
34: \affiliation{Material Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA}
35: %\email{tribeiro@lbl.gov}
36: %authors from another affiliations
37: 
38: \date{\today}
39: 
40: \begin{abstract}
41: This paper presents numerical studies of the single hole $tt't''J$ model 
42: that address the interplay between the kinetic energy of itinerant 
43: electrons and the exchange energy of local moments as of interest 
44: to doped Mott insulators.
45: Due to this interplay, two different spin correlations coexist around
46: a mobile vacancy.
47: These \textit{local} correlations provide an effective two-band picture
48: that explains the two-band structure observed in various theoretical and 
49: experimental studies, the doping dependence of the momentum space
50: anisotropic pseudogap phenomena and the asymmetry between hole and 
51: electron doped cuprates.
52: \end{abstract}
53: 
54: %\pacs{74.72.-h,74.25.Jb,71.10.Fd}% PACS, the Physics and Astronomy
55:                              % Classification Scheme.
56: 
57: \maketitle
58: 
59: 
60: 
61: \section{\label{sec:intro}Introduction} 
62: 
63: The evolution between the weakly correlated Fermi metal and the strongly 
64: coupled Mott insulator is a major and long-standing problem in the
65: field of condensed matter physics. \cite{M4916, A5902, H6338}
66: It concerns a vast list of material compounds \cite{IF9839} where the 
67: local character of $d$ and $f$ orbitals enhances the electron effective 
68: mass together with the role of the electron-electron Coulomb repulsive 
69: interaction.
70: The resulting competition between the small kinetic energy and the 
71: strong interaction may lead itinerant electrons in the metallic
72: state to form local moments in the Mott insulating state. \cite{M4916}
73: In this paper, I focus on the interplay between such itinerant and 
74: localized electrons.
75: 
76: The generalized-$tJ$ model explicitly embodies the above interplay.
77: Indeed, ``$t$'' stands for the kinetic energy term of itinerant charge 
78: carriers and ``$J$'' stands for the interaction term between localized 
79: spins.
80: The intricacy of this model follows from the mutual frustration 
81: between these two terms. 
82: Specifically, the $J > 0$ term favors a staggered moment spin background 
83: that constrains the motion of vacancies, while the $t$ term moves electrons
84: around and, thus, reshuffles and destroys the underlying 
85: antiferromagnetic (AF) spin pattern.
86: The two-dimensional (2D) $tt't''J$ model, which this paper addresses, 
87: is especially interesting because the compromise between the 
88: spin exchange and hole kinetic energies is particularly subtle in 
89: the parameter regime of interest to real materials, such as the 
90: high-temperature superconducting cuprates. \cite{A8796,D9463,LN0617}
91: 
92: Following the recent improvement in computational resources, experimental 
93: resolution and sample quality, various non-trivial results have illuminated 
94: our understanding of 2D doped Mott insulators.
95: It is exciting to note that many of these results are consistently
96: obtained by different theoretical approaches and by experiments.
97: For instance, a variety of numerical and analytical studies show that
98: the electronic spectrum below the chemical potential has a robust
99: two-band structure and that changing the electron density 
100: redistributes the spectral weight in a momentum dependent way.
101: \cite{PH9544,MH9545,DZ0016,GE0036,TS0009,T0417,RW0501,KK0614,PB0402,CC0502}
102: Similar conclusions apply to the two dispersive features displayed by
103: angle-resolved photoemission spectroscopy (ARPES) on the cuprates.
104: \cite{RS0301,KS0318,YZ0301,SR0402,DS0373}
105: Since the electron dynamics in strongly correlated systems follows
106: from the local environment around the carriers, \cite{SHORTRANGE}
107: %\footnote{
108: %This statement finds support in various theoretical and experimental
109: %results. 
110: %For instance:
111: %\textit{(i)} analytical methods show that AF correlations suppress 
112: %\textit{inter}-sublattice hopping \cite{KL8980,RV8893,MH9117,RH9808}  
113: %while certain spin liquid correlations suppress 
114: %\textit{intra}-sublattice processes instead; \cite{RW0301}
115: %\textit{(ii)} exact diagonalization, \cite{MH9545} quantum Monte Carlo 
116: %\cite{PH9544}, cluster perturbation theory \cite{ST0401} and 
117: %cellular dynamical mean-field theory \cite{KK0614,KK0606} show how electronic 
118: %spectral features keep track of the underlying local correlations;
119: %\textit{(iii)} the ubiquituous flat bands observed in the cuprates 
120: %around $(\pi,0)$ \cite{KS9498} follow from the local spin correlations
121: %as well; \cite{DN9428,RW0531}
122: %\textit{(iv)} AF correlations renormalize the cuprate nodal 
123: %dispersion width down to $\approx 2.2J$. 
124: %\cite{RS0301,KS0318,SR0402,DS0373,KW9845,TY0403,D9463,TS0009}
125: %}
126: the above two-band structure reflects the presence of two local 
127: correlations, which arise due to the interplay between itinerant 
128: electrons and local moments. 
129: 
130: This paper explores the microscopic origin of the above short-range
131: correlations and, thus, of the aforementioned
132: two-band structure that appears in both theory and experiments.
133: Specifically, in Sec. \ref{sec:local} the exact diagonalization and 
134: the self-consistent Born approximation techniques are employed
135: to study the single hole problem in the $tt't''J$ model.
136: I show that the interplay between the ``$t$'' and ``$J$'' terms 
137: of the Hamiltonian translates into the coexistence of two different types 
138: of spin correlations around the vacancy --
139: one type is driven by the kinetic energy term and the other by the 
140: exchange energy term.
141: These short-range correlations, which follow from purely
142: local energetic considerations and whose properties are studied in
143: Sec. \ref{sec:properties}, underlie a diverse set of non-trivial 
144: and, by now, well established properties of 2D doped Mott 
145: insulators.
146: These include the doping dependence of the pseudogap dispersion and of 
147: the pseudogap momentum space spectral weight distribution, 
148: \cite{DS0373,ZY0401,YZ0301,RS0301,KF0517,TY0403,AR0201} as 
149: well as the asymmetry between the hole and electron doped regimes of the 
150: cuprate compounds (Sec. \ref{sec:anisotropy}). 
151: \cite{DS0373,KW9845,YZ0301,RS0301,AR0201,TM9496,T0417,RW0501,KK0614}
152: 
153: 
154: 
155: \section{\label{sec:local}Two local correlations}
156: 
157: \subsection{\label{subsec:model}The model system}
158: 
159: The single hole 2D $tt't''J$ Hamiltonian is
160: \be
161: H_{tt't''J} = - \sum_{\langle ij \rangle, \sigma} t_{ij} \left(
162: \widetilde{c}_{i,\sigma}\sd \widetilde{c}_{j,\sigma} + H.c.\right) +
163: \sum_{\langle ij \rangle} J_{ij} \bm{S}_i.\bm{S}_j
164: \label{eq:Htj}
165: \ee
166: where $\widetilde{c}_{i,\sigma}$ is the constrained electron 
167: operator $\widetilde{c}_{i,\sigma} = c_{i,\sigma} (1-n_{i,-\sigma})$.
168: $t_{ij}$ equals $t$, $t'$ and $t''$ for first, second and third 
169: nearest-neighbor (NN) sites respectively and vanishes otherwise.
170: The exchange interaction only involves NN spins for 
171: which $J_{ij}=J$.
172: In this paper $J \in [0.2 , 0.8]$ (units are set so that $t=1$),
173: which includes the experimentally relevant regime $J\approx0.4$.
174: The calculations are not extended down to $J=0$ because, in that limit,
175: the hole is subjected to the Nagaoka instability \cite{N6692} and, thus,
176: the physics for $J\approx0$ is specific to such a regime and is not 
177: relevant to materials like the cuprates. \cite{WA0111}
178: The calculations for $J>0.8$ do not change the argument below nor
179: the consequent conclusions.
180: 
181: As mentioned in Sec. \ref{sec:intro}, this paper addresses the
182: electron dynamics as probed by the electron spectral function.
183: Its focus does not lie in the full details of the
184: spectral function line shape, but rather on the fact that the 
185: electronic spectrum displays two separate dispersive features below 
186: the Fermi level.
187: This work also concerns the momentum distribution of electron 
188: spectral weight, which displays distinct behavior in separate 
189: regions of the Brillouin zone, namely the regions around 
190: $(\pi/2,\pi/2)$ and $(\pi,0)$.
191: The above facts encode short time and short length scale physics.
192: Hence, within the above context, it is relevant to study 
193: small lattice systems and,
194: unless otherwise stated, all the results below come from the exact 
195: diagonalization of $H_{tt't''J}$ on a $4 \times 4$ lattice.
196: The exact diagonalization analysis is further substantiated by results
197: from the self-consistent Born approximation approach to the 
198: spinless-fermion Schwinger-boson representation of the $tJ$ model 
199: \cite{RV8893,MH9117,RH9808} on a $16 \times 16$ lattice.
200: 
201: 
202: \subsection{\label{subsec:onehole}One-hole states}
203: 
204: There exist two extreme limits where the interplay between 
205: itinerant electrons and local moments occurs, namely the one where:
206: \textit{(i)} most electrons are itinerant and the corresponding
207: Fermi energy is the highest energy scale in the problem;
208: \textit{(ii)} most electrons form local moments and the system
209: reduces to a lattice of spins with a few mobile vacancies.
210: The former case is captured by the well understood Kondo model, which
211: addresses how the Fermi sea accommodates the presence of a local moment.
212: \cite{W7573} 
213: The second case, which is of interest close to the Mott insulator 
214: transition, differs from the standard Kondo lattice problem since the 
215: spin-spin interaction is larger than the itinerant electrons' Fermi 
216: energy. \cite{RW0674}
217: In this case, it is rather convenient to consider how the spin background 
218: adjusts to the presence of a hole. 
219: 
220: Hence, in what follows, one studies the lowest energy configurations 
221: of the spin background around a single vacancy.
222: In particular, one considers the lowest energy single hole state
223: $\ket{\psi_{\bm{k}},J,t't,''}$ for each momentum $\bm{k}$, where
224: $J$, $t'$ and $t''$ label the model parameters that define
225: the corresponding Hamiltonian $H_{tt't''J}$. 
226: This state can fall into two categories -- it either has zero
227: or non-zero quasiparticle spectral weight
228: $\left| \bra{\psi_{\bm{k}},J,t',t''} \widetilde{c}_{\bm{k},\sigma} 
229: \ket{\text{HF GS}} \right|^2$, where $\ket{\text{HF GS}}$ denotes the 
230: groundstate of the half-filled system.
231: For all $\bm{k}$, $t'$ and $t''$ there exists a certain
232: $J_c(\bm{k},t',t'')$ such that $\ket{\psi_{\bm{k}},J,t',t''}$
233: has zero quasiparticle spectral weight if and only if 
234: $J \leq J_c(\bm{k},t',t'')$.
235: The intuition behind this result is that for $J/t \gg 1$ the large
236: spin stiffness renders the spin background robust to the hole motion,
237: while for small enough $J/t$ the soft AF spin configuration is 
238: dramatically modified by the doped hole
239: (the Nagaoka instability \cite{N6692} perfectly illustrates this last 
240: case).
241: If $J \leq J_c(\bm{k},t',t'')$ one denotes 
242: $\ket{\psi_{\bm{k}},J,t't,''}$ by $\ket{\widetilde{U}_{\bm{k}},J,t',t''}$
243: (hence, by definition, $\ket{\widetilde{U}_{\bm{k}},J,t',t''}$
244: has vanishing quasiparticle spectral weight).
245: If, instead, $J > J_c(\bm{k},t',t'')$ the single hole state
246: $\ket{\psi_{\bm{k}},J,t't,''}$ can be approximately recast
247: as
248: \begin{align}
249: \ket{\psi_{\bm{k}},J,t',t''} \ &\cong \ q(\bm{k},J,t',t'') 
250: \ket{Q_{\bm{k}},t',t''} + \notag \\
251: & \quad \quad + u(\bm{k},J,t',t'') \ket{U_{\bm{k}},t',t''}
252: \label{eq:decomp}
253: \end{align}
254: where $\ket{Q_{\bm{k}},t',t''}$ and $\ket{U_{\bm{k}},t',t''}$ 
255: are orthonormal states (to be defined below) that do \textit{not} 
256: depend on $J$, \cite{Utilde}
257: %\footnote{
258: %Note the difference between the kets $\ket{\widetilde{U}_{\bm{k}},J,t',t''}$
259: %and $\ket{U_{\bm{k}},t',t''}$.
260: %The former is $J$-\textit{dependent} and is defined for 
261: %$J < J_c(\bm{k},t',t'')$.
262: %The latter is $J$-\textit{independent} and is defined for 
263: %$J > J_c(\bm{k},t',t'')$.
264: %Similar notation is used for both states since they have 
265: %similar properties in the parameter regime of interest to 
266: %the cuprates (see Sec. \ref{sec:properties}).
267: %}
268: while $q(\bm{k},J,t',t'')$ and $u(\bm{k},J,t',t'')$
269: are $J$-dependent coefficients that obey the normalization condition
270: $\left|q(\bm{k},J,t',t'')\right|^2 +
271: \left|u(\bm{k},J,t',t'')\right|^2 = 1$.
272: Eq. \eqref{eq:decomp}, which applies in a large range of 
273: $J$ values, is a major result in this paper.
274: It implies that, in a large interval of values of 
275: $J > J_c(\bm{k},t',t'')$, the eigenstates 
276: $\ket{\psi_{\bm{k}},J,t't,''}$ define a line parameterized by $J$ 
277: which approximately lies in a 2D plane in the single hole $tt't''J$ 
278: model Hilbert space.
279: The physical content of this statement, 
280: together with evidence supporting Eq. \eqref{eq:decomp}, are 
281: presented below.
282: 
283: \begin{table}
284: \begin{ruledtabular}
285: \begin{tabular*}{\hsize}{cc|cccccc}
286: %\hline
287: & J & 0.3 & 0.4 & 0.5 & 0.6 & 0.7 & 0.8 \\
288: \hline
289: & $(\tfrac{\pi}{2},\tfrac{\pi}{2})$ & 0.9994 & 0.9994 & 0.9998 & 1 & 0.9998 & 0.9990 \\ 
290: $t'=0$ & $(\pi,0)$ & 0.9994 & 0.9994 & 0.9998 & 1 & 0.9998 & 0.9990 \\
291: $t''=0$ & $(\pi,\tfrac{\pi}{2})$ & 0.9972 & 0.9977 & 0.9993 & 1 & 0.9992 & 0.9970 \\
292: ED & $(\tfrac{\pi}{2},0)$ & 0.9975 & 0.9980 & 0.9994 & 1 & 0.9994 & 0.9977 \\
293: & $(0,0)$ & 0.9946 & 0.9923 & 0.9963 & 1 & 0.9938 & 0.9766 \\ 
294: \hline
295: & $(\tfrac{\pi}{2},\tfrac{\pi}{2})$ & 0.9996 & 0.9996 & 0.9998 & 1 & 0.9998 & 0.9990 \\ 
296: $t'=0$ & $(\pi,0)$ & 0.9994 & 0.9994 & 0.9998 & 1 & 0.9997 & 0.9986 \\ 
297: $t''=0$ & $(\pi,\tfrac{\pi}{2})$ & 0.9989 & 0.9988 & 0.9996 & 1 & 0.9995 & 0.9978 \\
298: SCBA & $(\tfrac{\pi}{2},0)$ & 0.9989 & 0.9988 & 0.9996 & 1 & 0.9995 & 0.9978 \\
299: & $(0,0)$ & 0.9016 & 0.9005 & 0.9766 & 1 & 0.9842 & 0.9488 \\ 
300: \hline
301:  & $(\tfrac{\pi}{2},\tfrac{\pi}{2})$ & 0.9994 & 0.9994 & 0.9998 & 1 & 0.9998 & 0.9990 \\ 
302: $t'=-0.2$ & $(\pi,0)$ & -- & 1  & 0.9998 & 0.9997 & 0.9999 & 1 \\
303: $t''=0.1$ & $(\pi,\tfrac{\pi}{2})$ & 0.9936 & 0.9952 & 0.9986 & 1 & 0.9987 & 0.9950 \\
304: ED & $(\tfrac{\pi}{2},0)$ & 0.9907 & 0.9943 & 0.9986 & 1 & 0.9988 & 0.9957 \\
305: & $(0,0)$ & 0.9880 & 0.9856 & 0.9943 & 1 & 0.9940 & 0.9807 \\ 
306: %\hline
307: \end{tabular*}
308: \end{ruledtabular}
309: \caption
310: {\label{tab:check2d} Square of the overlap of 
311: $\ket{\psi_{\bm{k}},J,t',t''}$ with the Hilbert space 
312: $\{ \ket{\psi_{\bm{k}},J=0.2,t',t''}, \ket{\psi_{\bm{k}},J=0.6,t',t''} \}$
313: for different $J$ and $\bm{k}$. 
314: Both exact diagonalization (ED) and self-consistent Born approximation (SCBA) 
315: results are shown for $t',t''=0$.
316: Exact diagonalization results are also shown for $t'=-0.2$, $t''=0.1$.
317: For $t'=-0.2$, $t''=0.1$ and $\bm{k}=(\pi,0)$ the Hilbert space 
318: $\{\ket{\psi_{\bm{k}},J=0.4,t',t''}, \ket{\psi_{\bm{k}},J=0.8,t',t''} \}$
319: is used instead.}
320: \end{table}
321: 
322: If $t',t''=0$ then $J_c(\bm{k},t'=0,t''=0) < 0.2$ for all $\bm{k}$ in 
323: Table \ref{tab:check2d}.
324: This table shows that for all $J \in [0.2 , 0.8]$, as well as 
325: for all depicted momenta $\bm{k}$,
326: the states $\ket{\psi_{\bm{k}},J,t'=0,t''=0}$
327: have almost unit overlap with the 2D Hilbert space
328: $\{\ket{\psi_{\bm{k}},J=0.2,t=0',t''=0}, 
329: \ket{\psi_{\bm{k}},J=0.6,t'=0,t''=0} \}$.
330: This conclusion is further substantiated by the self-consistent Born 
331: approximation technique on a $16\times 16$ lattice (see
332: Table \ref{tab:check2d}), thus showing that 
333: the above result is not specific to the $4\times 4$ lattice used in 
334: the exact diagonalization calculation. \cite{CHOICE}
335: %\footnote{
336: %The above choice of $J=0.2$ and $J=0.6$ could have been made different.
337: %However, any reasonable set of two values of $J$ in the range
338: %$[0.2 , 0.8]$ would lead to similar conclusions.
339: %}
340: A very similar observation holds when $t',t'' \neq 0$, as 
341: Table \ref{tab:check2d} illustrates for $t'=-0.2$, $t''=0.1$.
342: The only difference between the above $t',t''=0$ and $t'=-0.2$, $t''=0.1$
343: cases is that $J_c(\bm{k}=(\pi,0),t'=0, t''=0) < 0.2$
344: while $0.3 < J_c(\bm{k}=(\pi,0),t'=-0.2, t''=0.1) < 0.4$.
345: Since the approximate equality in Eq. \eqref{eq:decomp} only applies for
346: $J > J_c(\bm{k},t',t'')$, in Table \ref{tab:check2d} one uses the 2D 
347: Hilbert space $\{\ket{\psi_{\bm{k}=(\pi,0)},J=0.4,t'=-0.2, t''=0.1}, 
348: \ket{\psi_{\bm{k}=(\pi,0)},J=0.8,t'=-0.2, t''=0.1} \}$
349: to illustrate that Eq. \eqref{eq:decomp} also applies when 
350: $\bm{k}=(\pi,0)$ and $t'=-0.2$, $t''=0.1$.
351: 
352: The above numerical results show that Eq. \eqref{eq:decomp} 
353: is a very good approximation for a wide range of 
354: values of the exchange coupling $J$. \cite{WIDERANGE}
355: %\footnote{
356: %The width of the interval of $J$ values displayed in 
357: %Table \ref{tab:check2d}, namely $[0.2 , 0.8]$, is of the 
358: %order of the maximum $J$ value.
359: %}
360: However, one is still free to choose any orthonormal pair of states 
361: $\ket{Q_{\bm{k}},t',t''}$ and $\ket{U_{\bm{k}},t',t''}$
362: in the 2D Hilbert space used as a reference.
363: A physically sensible choice comes from requiring
364: $q(\bm{k},J,t',t'')$ to monotonously increase with $J/t$  
365: [thus $u(\bm{k},J,t',t'')$ monotonously decreases with $J/t$].
366: Since cranking up $J$ enhances the quasiparticle features of doped 
367: carriers, \cite{MH9117} the above condition is automatically satisfied 
368: if $\ket{U_{\bm{k}},t',t''}$ has vanishing quasiparticle spectral weight. 
369: This prescription uniquely determines Q states 
370: ($\ket{Q_{\bm{k}},t',t''}$) and U states ($\ket{U_{\bm{k}},t',t''}$)
371: which, one should note, are not eigenstates 
372: of $H_{tt't''J}$. \cite{SINGLEJ}
373: %\footnote{
374: %As it follows from the approximate equality in Eq. \eqref{eq:decomp}, 
375: %the states $\ket{Q_{\bm{k}},t',t''}$ and $\ket{U_{\bm{k}},t',t''}$ 
376: %are relatively insensitive to variations of the two values
377: %of $J$ used to define the above mentioned 2D plane in the 
378: %single hole $tt't''J$ model Hilbert space.
379: %In the limit where these two values of $J$ become infinitesimally close
380: %to each other, $\ket{Q_{\bm{k}},t',t''}$ and $\ket{U_{\bm{k}},t',t''}$
381: %are constructed from the state vectors
382: %$\ket{\psi_{\bm{k}},J,t',t''}$ and $\p \ket{\psi_{\bm{k}},J,t',t''}/ \p J$.
383: %Therefore, one can determine a pair of Q and U states
384: %for every single value of $J > J_c(\bm{k},t',t'')$.
385: %}
386: The above construction implies that Q states bear the electron-like
387: properties of the true eigenstates $\ket{\psi_{\bm{k}},J,t't,''}$ and,
388: indeed, for all values of $t'$ and $t''$ used throughout this paper
389: one has that
390: $0.5 \lesssim
391: \frac{|\bra{Q_{\bm{k}}} \widetilde{c}_{\bm{k},\sigma} \ket{\text{HF GS}}|^2}
392: {|\bra{\text{HF GS}} \widetilde{c}_{\bm{k},\sigma}\sd 
393: \widetilde{c}_{\bm{k},\sigma} \ket{\text{HF GS}}|} \lesssim 0.8$
394: in the momentum space region around the $(\pi,0)-(0,\pi)$ 
395: line. \cite{MBZB}
396: %\footnote{
397: %This is the region where the lowest eigenstates lie and whose
398: %properties one is more concerned about.
399: %}
400: 
401: \begin{figure}
402: \includegraphics[width=0.48\textwidth]{spin_back.eps}
403: \caption{\label{fig:spin}
404: (a) $S_{\bm{k}}^Q(\bm{i})$ and (b) $S_{\bm{k}}^U(\bm{i})$ 
405: where $\bm{i}$ is the distance to the vacancy (black square at the 
406: center), $\bm{k}=(\tfrac{\pi}{2},\tfrac{\pi}{2})$ and $t',t''=0$.
407: Different $\bm{k}$, $t'$ and $t''$ lead to qualitatively similar 
408: conclusions, as seen in Fig. \ref{fig:staggered}.
409: }
410: \end{figure}
411: 
412: The previous argument clarifies how the spin background adjusts to 
413: the presence of a moving hole.
414: Specifically, spins show two different types of correlations -- one 
415: type is enhanced upon increasing $J/t$ and the other becomes more 
416: pronounced when $J/t$ is reduced.
417: By definition, Q and U states capture these correlations and,
418: not surprisingly, they display distinct physical properties.
419: Simply based on the above energetic considerations, one expects
420: the former states to retain the AF correlations of the 
421: undoped system, while the doping induced spin correlations in U states 
422: facilitate hole hopping.
423: The analysis in Sec. \ref{subsec:r_and_k} confirms
424: this microscopic picture.
425: In principle, a similar construction 
426: applies to models other than the 2D $tt't''J$ model.
427: The significant fact about this model is that, for experimentally 
428: relevant parameters, the overlap of both Q and U states with 
429: $\ket{\psi_{\bm{k}},J,t',t''}$ is large and exhibits a 
430: considerable momentum dependence (see Sec. \ref{sec:anisotropy}). \cite{3D}
431: %\footnote{
432: %The above formal construction to derive Q and U states can be applied,
433: %for instance, to the 3D $tt't''J$ model.
434: %However, in that case, U states are expected to be less relevant
435: %since AF correlations, which are more robust in 3D,
436: %lead to strong quasiparticle properties throughout the Brillouin zone.
437: %}
438: 
439: 
440: 
441: \section{\label{sec:properties}Properties of the local correlations}
442: 
443: The construction in Sec. \ref{subsec:onehole} identifies two
444: different spin configurations that coexist around a mobile
445: vacancy.
446: It also provides a recipe to separately obtain these
447: configurations and, thus, to study their properties. 
448: 
449: 
450: \subsection{\label{subsec:r_and_k}Real and momentum space properties}
451: 
452: \begin{figure}
453: \includegraphics[width=0.48\textwidth]{Rspace.eps}
454: \caption
455: {\label{fig:staggered}
456: $\widetilde{S}_{\bm{k}}^Q(\nu)$ (left panels) 
457: and $\widetilde{S}_{\bm{k}}^U(\nu)$ (right panels) 
458: for different momenta $\bm{k}$.
459: (a) and (b) $t'=-0.3, t''=0.2$. 
460: (c) and (d) $t',t''=0$.
461: (e) and (f) $t'=0.3, t''=-0.2$. 
462: }
463: \end{figure}
464: 
465: First, consider the average spin density pattern around the hole
466: \be
467: S_{\bm{k}}^Y(\bm{i})\equiv\bra{Y_{\bm{k}}} \sum_j S_{j+i}^z 
468: \widetilde{c}_{j,-1/2} \widetilde{c}_{j,-1/2}\sd \ket{Y_{\bm{k}}} 
469: \label{eq:spin_density}
470: \ee
471: for both $Y=Q$ and $Y=U$.
472: Fig. \ref{fig:spin} illustrates how different the spin background
473: is in Q and U states for 
474: $\bm{k}=(\tfrac{\pi}{2},\tfrac{\pi}{2})$ and $t',t''=0$.
475: The former preserve an evident staggered pattern while the latter
476: display an almost uniform distribution of the spin-1/2 introduced
477: in the system upon doping.
478: To show that this picture remains valid for other values of $\bm{k}$, 
479: $t'$ and $t''$, take the average of the staggered magnetization over 
480: the hole's $\nu^{th}$ NN sites: 
481: $\widetilde{S}_{\bm{k}}^Y(\nu) \equiv 
482: - \langle(-)^{i_x+i_y}S_{\bm{k}}^Y(\bm{i})\rangle_{\nu} - 
483: \tfrac{1}{N-1}\tfrac{1}{2}$
484: (here $Y=Q,U$). \cite{AVERAGE_S}
485: %\footnote{
486: %The average magnetization
487: %$\tfrac{1}{N-1}\sum_j \langle S_j^z \rangle = \tfrac{1}{N-1}\tfrac{1}{2}$, 
488: %where $N$ is the number of lattice sites, is subtracted to reduce 
489: %finite size effects.
490: %}
491: Figs. \ref{fig:staggered}(a), \ref{fig:staggered}(c) and 
492: \ref{fig:staggered}(e) show that for different $\bm{k}$, $t'$ and $t''$
493: the doped hole in Q states coexists with the staggered 
494: spin pattern inherited from the undoped system.
495: This state of affairs is in sharp contrast with the results for
496: U states, where the AF spin pattern of the undoped system is destroyed 
497: and the staggered magnetization around the hole 
498: is close zero and even negative
499: [Figs. \ref{fig:staggered}(b), \ref{fig:staggered}(d) and 
500: \ref{fig:staggered}(f)].
501: One can check that a similar conclusion holds for
502: $\widetilde{\text{U}}$ states [these are the energy 
503: eigenfunctions $\ket{\psi_{\bm{k}},J,t',t''}$ when 
504: $J<J_c({\bm{k}},t',t'')$].
505: 
506: \begin{table}
507: \begin{ruledtabular}
508: \begin{tabular*}{\hsize}{c|cc|cc|}
509: %\hline
510: & \multicolumn{2}{c|}{$t'=-0.3; t''=0.2$} &
511: \multicolumn{2}{c|}{$t'=0.3; t''=-0.2$} \\
512: %\hline
513: $\bm{k}$ & $(\pi,0)$ & $(0,0)$ & $(\tfrac{\pi}{2},\tfrac{\pi}{2})$ & $(0,0)$ \\
514: %\hline
515: $\bm{q}=(0,0)$ & 0.0329 & 0.0343 & 0.0164 & 0.0328 \\
516: %\hline
517: $\bm{q}=(\pi,\pi)$ & 0.5437 & 0.6051 & 0.5293 & 0.5141 \\
518: %\hline
519: \end{tabular*}
520: \end{ruledtabular}
521: \caption
522: {\label{tab:MD} 
523: $\sum_{\bm{q}}' n_{\bm{k}}^{\widetilde{U}}(\bm{q},+\tfrac{1}{2})\equiv
524: \sum_{\bm{q}}'\bra{\widetilde{U}_{\bm{k}}}
525: \widetilde{c}_{-\bm{q},-\tfrac{1}{2}} 
526: \widetilde{c}_{-\bm{q},-\tfrac{1}{2}}\sd  \ket{\widetilde{U}_{\bm{k}}}$.
527: $\bm{q}=(0,0)$ results involve sum over $\bm{q}=(0,0)$,
528: $\bm{q}=(\pm\tfrac{\pi}{2},0)$ and $\bm{q}=(0,\pm\tfrac{\pi}{2})$.
529: $\bm{q}=(\pi,\pi)$ results involve sum over $\bm{q}=(\pi,\pi)$,
530: $\bm{q}=(\pm\tfrac{\pi}{2},\pi)$ and $\bm{q}=(\pi,\pm\tfrac{\pi}{2})$.
531: The model parameters used are relevant to both hole doped cuprates
532: ($J=0.4$, $t'=-0.3$, $t''=0.2$) and electron doped cuprates
533: ($J=0.4$, $t'=0.3$, $t''=-0.2$). \cite{TM0017}
534: }
535: \end{table}
536: 
537: In order to complement the above real space picture, 
538: one also considers the hole momentum distribution function
539: \be 
540: n_{\bm{k}}^Y(\bm{q},\sigma)\equiv\bra{Y_{\bm{k}}} 
541: \widetilde{c}_{-\bm{q},-\sigma} \widetilde{c}_{-\bm{q},-\sigma}\sd  
542: \ket{Y_{\bm{k}}}
543: \label{eq:hole_momentum}
544: \ee
545: for $Y=Q,U$.
546: Since Q states bear an electron-like character, the hole momentum 
547: distribution function
548: $n_{\bm{k}}^Q(\bm{q},+\tfrac{1}{2})$ is peaked at 
549: $\bm{q}=\bm{k}$. 
550: A smaller peak is also observed at $\bm{q}=\bm{k}+(\pi,\pi)$ due 
551: to the strong AF correlations. \cite{EO9541}
552: In U states, the hole strongly interacts with the surrounding spins 
553: and, as a result, the hole momentum distribution function 
554: $n_{\bm{k}}^U(\bm{q},+\tfrac{1}{2})$ peaks around 
555: $\bm{q}=(\pi,\pi)$ for all momenta $\bm{k}$ [Fig. \ref{fig:bands} (a)].
556: Table \ref{tab:MD} illustrates that the hole density in
557: $\widetilde{\text{U}}$ states also peaks around $(\pi,\pi)$ 
558: independently of the momentum $\bm{k}$.
559: 
560: The above results confirm that Q states capture the AF 
561: correlations that persist around the vacancy away from
562: half-filling.
563: This is expected since these states have a well defined 
564: quasiparticle character.
565: A remarkably different picture holds for the U and 
566: $\widetilde{\text{U}}$ states, whose quasiparticle spectral weight
567: vanishes.
568: As the above spin density results indicate, the spin correlations
569: in these states spread the extra $S^z=\tfrac{1}{2}$ away from the vacancy.
570: The resulting loss of spin exchange energy is accompanied by a gain in 
571: the hole kinetic energy, as it follows from
572: the hole momentum distribution results which support that, in these
573: states, the hole always lies around the bare band bottom 
574: [which is located at $(\pi,\pi)$]. 
575: This evidence resembles predictions from spin-charge separation scenarios. 
576: Indeed, within the slave-boson \cite{LN9221,WL9603,LN0617} 
577: and the doped-carrier frameworks, \cite{RW0501,RW0674}
578: the electron decays into a charged spinless boson, which condenses at 
579: $(\pi,\pi)$, and a spin-1/2 chargeless fermion, which 
580: carries the remaining momentum. 
581: The above calculations determine the equal time correlations probed by the
582: quantities in Eq. \eqref{eq:spin_density} and Eq. \eqref{eq:hole_momentum}
583: in a small lattice 
584: and, thus, cannot prove the existence (or lack thereof) of true 
585: spin-charge separation.
586: Still, they support that, in U and $\widetilde{\text{U}}$ 
587: states, the lattice spins screen the hole in conformity with 
588: short-range aspects of spin-charge separation phenomenology.
589: 
590: \begin{figure}
591: \includegraphics[width=0.48\textwidth]{bands_merged_Kspace.eps}
592: \caption
593: {\label{fig:bands}
594: (a) $\sum_{\bm{q}}' n_{\bm{k}}^U(\bm{q},+\tfrac{1}{2})$.
595: Empty symbols involve sum over $\bm{q}=(\pi,\pi)$,
596: $\bm{q}=(\pm\tfrac{\pi}{2},\pi)$ and $\bm{q}=(\pi,\pm\tfrac{\pi}{2})$.
597: Full symbols involve sum over $\bm{q}=(0,0)$,
598: $\bm{q}=(\pm\tfrac{\pi}{2},0)$ and $\bm{q}=(0,\pm\tfrac{\pi}{2})$.
599: (b)-(d) Dispersion relations for $\ket{Q_{\bm{k}}}$
600: (full line), $\ket{U_{\bm{k}}}$ (dashed line) and $\ket{\psi_{\bm{k}}}$
601: (dash-dot line). Upper, middle and lower set of dispersions are obtained
602: for $J$ equal to $0.2$, $0.4$ and $0.7$ respectively.
603: ($\circ$) indicates the best energy obtained by a linear combination of
604: $\ket{Q_{\bm{k}}}$ and $\ket{U_{\bm{k}}}$ when $J<J_c(\bm{k},t',t'')$ 
605: (in which case $\ket{\psi_{\bm{k}}} = \ket{\widetilde{U}_{\bm{k}},J,t',t''}$).
606: }
607: \end{figure}
608: 
609: 
610: \subsection{\label{subsec:renorm}Effect on electron dynamics}
611: 
612: The different hopping terms in the $tt't''J$ model Hamiltonian 
613: [Eq. \eqref{eq:Htj}] move electrons between first, second and third 
614: NN sites under the no-double-occupancy constraint.
615: These processes may or may not be restrained by the surrounding spin 
616: correlations. \cite{PROBE}
617: %\footnote{ 
618: %This is the reason why electronic spectral properties provide 
619: %an indirect probe of the local correlations \cite{SHORTRANGE}.
620: %}
621: For instance, NN hopping is frustrated by the two-sublattice 
622: structure of AF correlations.
623: Intra-sublattice hopping processes are, however, consistent
624: with the staggered pattern of AF correlations which, thus, do
625: not strongly renormalize $t'$ and $t''$. \cite{RENORM}
626: %\footnote{
627: %As shown in Ref. \onlinecite{TM0017}, the hole dispersion in
628: %the $tt't''J$ model along $(\pi,0)-(0,\pi)$ is controlled by 
629: %$t' \approx -2t''$ and is a factor $\sim 2-3$ smaller than
630: %expected from the bare parameters.
631: %This is to be contrasted with the nodal dispersion width which
632: %goes from the bare value $8t$ down to $2.2J$ due to AF spin
633: %correlations. \cite{D9463,TM0017,KL8980,MH9117,DS0373}
634: %In the parameter regime of interest to the cuprates this corresponds
635: %to a renormalization by a factor of $\sim 10$.
636: %}
637: 
638: \begin{table}
639: \begin{ruledtabular}
640: \begin{tabular*}{\hsize}{ccccccc}
641: %\hline
642: $t'$ & $t''$ & $\Delta E^{\psi}$ & $\Delta E^Q$ & $\Delta E^U$ &
643: $W_{\bm{k'}}^Q$ & $W_{\bm{k''}}^Q$ 
644: \\
645: \hline
646: -0.3 & 0.2 & 0.69 & 1.43 & 0.22 & 0 & 0.75 \\
647: %\hline
648: -0.2 & 0.1 & 0.56 & 0.92 & 0.14 & 0.45 & 0.72 \\
649: %\hline
650: %-0.1 & 0.05 & 0.30 & 0.33 & 0.30 & 0.63 & 0.69 \\
651: %\hline
652: 0 & 0 & 0 & 0 & 0 & 0.66 & 0.66 \\
653: %\hline
654: %0.1 & -0.05 & -0.35 & -0.37 & -0.09 & 0.71 & 0.65 \\
655: %\hline
656: 0.2 & -0.1 & -0.75 & -1.08 & -0.10 & 0.76 & 0.50 \\
657: %\hline
658: 0.3 & -0.2 & -0.80 & -2.33 & -0.29 & 0.82 & 0 \\
659: %\hline
660: \end{tabular*}
661: \end{ruledtabular}
662: \caption{\label{tab:pseudogap}
663: $\Delta E^Q$, $\Delta E^U$, $\Delta E^{\psi}$ and
664: $W_{\bm{k}}^Q$ with $\bm{k}=\bm{k'} \equiv (\pi,0)$
665: and $\bm{k}=\bm{k''} \equiv (\tfrac{\pi}{2},\tfrac{\pi}{2})$ for several 
666: $t'$ and $t''$ and $J=0.4$.}
667: \end{table}
668: 
669: The way the spin correlations in U states renormalize $t$, $t'$ and $t''$
670: is strikingly different though.
671: Firstly, these correlations are induced as a way to enhance NN hopping.
672: Secondly, they heavily renormalize $t'$ and $t''$.
673: To establish the latter fact, consider the hole dispersion in Q states
674: $E_{\bm{k}}^Q\equiv\bra{Q_{\bm{k}}}H_{tt't''J}\ket{Q_{\bm{k}}}$ and
675: the hole dispersion in U states
676: $E_{\bm{k}}^U\equiv\bra{U_{\bm{k}}}H_{tt't''J}\ket{U_{\bm{k}}}$.
677: Table \ref{tab:pseudogap} displays how the dispersion width between
678: $(\pi,0)$ and $(\pi/2,\pi/2)$ changes with $t'$ and $t''$  for both 
679: Q states ($\Delta E^Q\equiv E_{(\pi,0)}^Q-E_{(\pi/2,\pi/2)}^Q$) and U 
680: states ($\Delta E^U\equiv E_{(\pi,0)}^U-E_{(\pi/2,\pi/2)}^U$). \cite{FLAT}
681: %\footnote{
682: %The fact that this width vanishes for $t',t''=0$ is an artifact of
683: %the $4\times4$ lattice. \cite{dagot}
684: %In larger systems, the dispersion width along $(\pi,0) - (\pi/2,\pi/2)$
685: %for $t',t''=0$ is $\approx 20\%$ of the nodal dispersion width. 
686: %\cite{MH9117,DN9428}
687: %}
688: Indeed, the effect of $t'$ and $t''$ on $\Delta E^Q$ is almost 
689: one order of magnitude larger than on $\Delta E^U$.
690: 
691: Interestingly, certain spin liquid correlations discussed in the 
692: context of the $tJ$ model strongly inhibit coherent intra-sublattice 
693: hopping. \cite{RW0301}
694: This fact, together with the above results, further supports that 
695: spin correlations in U states resemble spin liquid correlations at 
696: short length scales.
697: 
698: 
699: 
700: \section{\label{sec:anisotropy}Momentum space anisotropy}
701: 
702: In the cuprates' renowned pseudogap metallic regime, 
703: the low energy physics is determined by the states around the
704: $(\pi,0)-(0,\pi)$ line. \cite{DS0373}
705: However, there is a clear distinction between the nodal 
706: [$\vec{k}=(\pm \pi/2,\pm \pi/2)$] and the antinodal 
707: [$\vec{k}=(\pi,0),\, (0,\pi)$] regions.
708: Specifically, ARPES detects an energy difference between the 
709: single-electron spectral features around $(\pi/2,\pi/2)$ and $(\pi,0)$
710: (whence the term ``pseudogap'') \cite{DS0373,DY9651} and,
711: in addition, a strong suppression of the electronic character of 
712: excitations is observed in the pseudogap region.
713: These phenomena occur in both hole and electron doped compounds
714: with a crucial difference:
715: in the former, low energy quasiparticles appear close to 
716: $(\pi/2,\pi/2)$ but not around $(\pi,0)$; \cite{ZY0401,YZ0301,RS0301,KF0517}
717: in electron doped materials, both the pseudogap and the excitations
718: with little electron-like character are pushed toward the
719: zone diagonal. \cite{AR0201,KB0322,HP0111} 
720: 
721: This phenomenology is reproduced by the generalized-$tJ$ model, where
722: it stems from the role of the intra-sublattice hopping parameters 
723: $t'$ and $t''$. 
724: Indeed, for $t',t'' = 0$, the quasiparticle states at $(\pi/2,\pi/2)$ 
725: and $(\pi,0)$ have both comparable energies and spectral weight 
726: intensities. \cite{MH9117,DN9428}
727: On the other hand, non-zero $t'$ and $t''$ fit the experimentally 
728: observed dispersion width along $(\pi,0) - (\pi/2,\pi/2)$.
729: \cite{GV9466,NV9576,KW9845,T0417}
730: These intra-sublattice hopping parameters further lead to pseudogap 
731: states with modified spin background correlations 
732: \cite{TS0009,LL0301,ME9916} and, thus, with small spectral weight. 
733: \cite{KW9845,TM0017,T0417,RW0501,KK0614,CC0502} 
734: As to the difference between the hole and electron doped regimes,
735: it simply follows from the change in the sign of $t'$ and $t''$. 
736: \cite{TM9496,GV9466,KW9845,T0417,RW0501,KK0614,CC0502}
737: 
738: The main message of this paper is that the results obtained in 
739: Sec. \ref{sec:local} and Sec. \ref{sec:properties} provide a 
740: microscopic two-band picture that rationalizes the above 
741: generalized-$tJ$ model behavior.
742: This picture embodies the effect of the Q and U states' short-range 
743: correlations, which underlies the momentum anisotropic pseudogap 
744: behavior, as well as its dependence on the electronic density 
745: (see below).
746: 
747: 
748: \subsection{\label{subsec:twoband}Two-band picture}
749: 
750: Sec. \ref{subsec:onehole} identifies the two distinct spin correlations 
751: that dress the vacancy in low energy single-hole states.
752: The static properties of these low energy states then follow from 
753: the reduced two-band Hamiltonian \cite{SINGLEJ}
754: \be
755: H_{reduced,\bm{k}} \equiv \left[ \begin{array}{cc}
756: \bra{Q_{\bm{k}}}H_{tt't''J}\ket{Q_{\bm{k}}} &
757: \bra{Q_{\bm{k}}}H_{tt't''J}\ket{U_{\bm{k}}} \\
758: \bra{U_{\bm{k}}}H_{tt't''J}\ket{Q_{\bm{k}}} &
759: \bra{U_{\bm{k}}}H_{tt't''J}\ket{U_{\bm{k}}}
760: \end{array}\right]
761: \label{eq:Hreduced}
762: \ee
763: This Hamiltonian yields the two spectral dispersions
764: observed both by ARPES data \cite{RS0301,KS0318,YZ0301,SR0402}
765: and by various theoretical studies of the related $tJ$ and Hubbard 
766: models. \cite{PH9544,MH9545,DZ0016,GE0036,RW0501,KK0614}
767: It also determines the hybridization of Q and U states and, thus, 
768: the spectral weight distribution throughout momentum space. 
769: Therefore, $H_{reduced,\bm{k}}$ must capture the aforementioned
770: role of $t'$ and $t''$ in the pseudogap phenomenology.
771: 
772: Interestingly, the above role of $t'$ and $t''$ can be discussed 
773: only in terms of the dispersions $E_{\bm{k}}^Q$ and $E_{\bm{k}}^U$.
774: To see this, note that $t'$ and $t''$ strongly affect the dispersion 
775: of a hole surrounded by the AF correlations in Q states 
776: [in fact, $\Delta E^Q = A (-4t'+8t'')$, 
777: where the renormalization factor $A\gtrsim 1/2$].
778: Therefore, $t' < 0$ and $t'' > 0$ increase the energy 
779: $E_{(\pi,0)}^Q$ and decrease $E_{(\pi/2,\pi/2)}^Q$. 
780: Consequently, 
781: $E_{(\pi,0)}^{\psi}\equiv\bra{\psi_{(\pi,0)}}H_{tt't''J}\ket{\psi_{(\pi,0)}}$
782: also increases and 
783: $E_{(\pi/2,\pi/2)}^{\psi}\equiv
784: \bra{\psi_{(\pi/2,\pi/2)}}H_{tt't''J}\ket{\psi_{(\pi/2,\pi/2)}}$
785: also decreases and, hence, a pseudogap 
786: $\Delta E^{\psi}\equiv E_{(\pi,0)}^{\psi}-E_{(\pi/2,\pi/2)}^{\psi}$
787: opens at $(\pi,0)$.
788: Intra-sublattice hopping is, however, strongly frustrated in U states and,
789: thus, $t' < 0$ and $t'' > 0$ increase the energy difference 
790: $E_{(\pi,0)}^Q - E_{(\pi,0)}^U$ while reducing
791: $E_{(\pi/2,\pi/2)}^Q - E_{(\pi/2,\pi/2)}^U$
792: [Figs. \ref{fig:bands}(b)-\ref{fig:bands}(c)].
793: This impacts the extent to which $\ket{Q_{\bm{k}}}$
794: and $\ket{U_{\bm{k}}}$ hybridize, reducing 
795: $W_{(\pi,0)}^Q\equiv |\langle \psi_{(\pi,0)}\ket{Q_{(\pi,0)}}|^2$ and
796: enlarging 
797: $W_{(\pi/2,\pi/2)}^Q\equiv |\langle \psi_{(\pi/2,\pi/2)}
798: \ket{Q_{(\pi/2,\pi/2)}}|^2$ (Table \ref{tab:pseudogap}).
799: For $J=0.4, t'=-0.3, t''= 0.2$ 
800: the energy $E_{(\pi,0)}^Q$ is so large that the minimum energy obtained 
801: by a linear combination of $\ket{Q_{(\pi,0)}}$ and $\ket{U_{(\pi,0)}}$ 
802: becomes higher than that of a different state $\ket{\widetilde{U}_{(\pi,0)}}$
803: with vanishing quasiparticle spectral weight [Fig. \ref{fig:bands}(b)]. 
804: As a result, $W_{(\pi,0)}^Q = 0$.
805: At the same time $W_{(\pi/2,\pi/2)}^Q = 0.75$, so that a sharp difference
806: is encountered between the electron-like character of nodal and 
807: antinodal states. \cite{ME9916,TS0009,LL0301}
808: 
809: The change between the cuprates' hole and electron doped regimes
810: amounts to a change in the sign of $t'$ and $t''$, 
811: in which case $\Delta E^Q = A (-4t'+8t'')$ changes sign as well.
812: The above argument then still applies, with the roles of 
813: momenta $(\pi,0)$ and $(\pi/2,\pi/2)$ interchanged 
814: [Fig. \ref{fig:bands}(d) and Table \ref{tab:pseudogap}].
815: 
816: 
817: \subsection{\label{subsec:doping}Doping dependence}
818: 
819: The above calculation and the ensuing arguments concern a single 
820: hole surrounded by a spin background and, thus, do not have to
821: straightforwardly apply in the presence of a finite hole density.
822: Interestingly, though, a large body of evidence suggests that
823: single-hole physics is relevant away from half-filling.
824: Indeed, quantum Monte Carlo, \cite{PH9544,GE0036} exact 
825: diagonalization \cite{MH9545} and cellular dynamical mean-field 
826: theory \cite{KK0614} studies show that the two-band structure
827: identified in the half-filled spectral function below the chemical 
828: potential remains almost unaffected upon hole doping, whose main 
829: effect is to transfer spectral weight between the pre-existing 
830: bands in a momentum dependent manner.
831: This behavior is expected as long as short-range AF correlations are 
832: present. \cite{PH9544,MH9545,GE0036}
833: Since calculations on the $U/t=8$ Hubbard model \cite{DM9582} find 
834: that such correlations persist around the vacancy up to the 
835: hole density $x=0.25$, the above two-band picture may 
836: apply in a wide doping range.
837: Cuprate ARPES data also displays the two dispersive
838: features throughout a large portion of the phase diagram,
839: \cite{RS0301,YZ0301,KS0318,SR0402,DS0373} hence, it
840: complies with the aforementioned theoretical expectations.
841: 
842: It is well known that the pseudogap phenomenology weakens upon 
843: increasing the dopant density.
844: Hence, the pseudogap magnitude diminishes away from half-filling,
845: \cite{DS0373,TY0403} as does the difference in the electron-like 
846: character of nodal and antinodal excitations. 
847: \cite{RS0301,YZ0301,ZY0401,KF0517,AR0201}
848: This experimental evidence is captured by the naive extension
849: of the above two-band picture to the finite hole density case.
850: Indeed, Sec. \ref{subsec:twoband} shows that the momentum
851: space anisotropic behavior follows from the effect of $t'$ and
852: $t''$ in the dynamics of holes surrounded by short-range AF 
853: correlations.
854: Upon doping, these correlations are gradually replaced by the 
855: doping induced correlations which prevail in U states.
856: Since the latter strongly renormalize $t'$ and $t''$, the 
857: differentiation between the $(\pi/2,\pi/2)$ and $(\pi,0)$ regions
858: is also gradually depleted.
859: 
860: Refs. \onlinecite{RW0501,RW0674} develop a new mean-field 
861: approach to the $tt't''J$ model that embodies the
862: above two-band picture in the presence of a finite hole density.
863: It explicitly captures the interplay between the mobile holes
864: and the above two different spin correlations and correctly
865: describes the microscopic electron dynamics in the 2D doped Mott
866: insulator.
867: This assertion is attested by the successful comparison to
868: other theoretical approaches and especially to a vast portfolio
869: of non-trivial cuprate ARPES and tunneling conductance data.
870: The latter include the aforementioned nodal-antinodal dichotomy,
871: the Fermi arcs, the peak-dip-hump structure, the kink and the extended
872: flat regions close to $(\pi,0)$ in the electron dispersion and the 
873: large diversity of tunneling spectra. \cite{RW0501,RW0531}
874: 
875: 
876: 
877: \section{\label{sec:conclusion}Conclusions}
878: 
879: In this paper, I numerically study how a single mobile hole is 
880: dressed by the encircling spins within the $tt't''J$ model context. 
881: Purely local energetic arguments decide whether a staggered
882: moment configuration or a spin configuration reminiscent
883: of spin liquid physics prevails around the vacancy.
884: In the experimentally relevant parameter regime, the competition
885: between the two spin correlations is very subtle and can be
886: particularly sensitive to the hole momentum.
887: Consequently, the electron spectral properties can be extremely 
888: momentum dependent, displaying a pseudogap and distinct quasiparticle
889: properties in the nodal and antinodal regions, as observed in 
890: both hole and electron doped cuprate compounds.
891: \cite{DS0373,ZY0401,YZ0301,RS0301,KF0517,TY0403,AR0201,KW9845}
892: AF short-range correlations are gradually depleted upon doping and,
893: thus, the above differentiation between the nodal and antinodal 
894: regions is expected to disappear further away from half-filling,
895: in agreement with the phenomenology of high-T$_c$ superconductors. 
896: \cite{DS0373,TY0403,ZY0401,KF0517,AR0201}
897: 
898: The above considerations agree with prior work 
899: \cite{PH9544,KK0614,SP0302,ST0401,RW0531,KK0606}
900: substantiating that the pseudogap and the resulting momentum space 
901: anisotropy follow from the local interaction between the doped 
902: carriers and the short-range spin correlations that strive close 
903: to the Mott insulating transition.
904: This paper goes a step further and provides an exact scheme to 
905: determine the local spin correlations that dress moving carriers.
906: In particular, it shows that the interplay between the itinerant
907: doped carrier and the surrounding local moments translates into
908: the coexistence of two different local correlations, namely the 
909: staggered moment correlations already present in the undoped system 
910: and a different type of correlations induced upon carrier doping.
911: The latter correlations are shown to be responsible for short-range 
912: phenomenology characteristic of spin-charge separated states and
913: to have a peculiar impact in the electron dynamics, specifically, 
914: they strongly renormalize $t'$ and $t''$.
915: 
916: One way to optimize both the hole kinetic energy and the spin 
917: exchange energy in doped Mott insulators is to spatially 
918: separate charge and spin degrees of freedom into, say, stripe-like 
919: configurations. \cite{ZG8991}
920: The above calculation suggests an alternative scenario:
921: the quantum superposition of two types of local states which 
922: separately enhance the $J$ and $t$ terms of the Hamiltonian.
923: These two states have a drastically different effect on the electron 
924: dynamics and provide a simple two-band microscopic picture 
925: of doped Mott insulators.
926: %which reproduces a variety of experimental data. \cite{RW0501,RW0531}
927: In this picture, the vacancy is, at times, surrounded by 
928: staggered moments while, at other instants, spin liquid correlations 
929: take over in order to facilitate the vacancy's motion.
930: In the pseudogap momentum space region, the kinetic energy of a 
931: vacancy surrounded by a local AF spin configuration increases with 
932: $|t'|$ and $|t''|$, thus tilting the balance in favor of the above 
933: spin liquid correlations (whose energy is less sensitive to
934: $t'$ and $t''$).
935: As a result, in the pseudogap regime, the way the spin background 
936: dresses the hole at $(\pi/2,\pi/2)$ differs from the way it dresses 
937: the hole at $(\pi,0)$.
938: This fact is in consonance with previous numerical evidence for the
939: approximate decoupling of spin and charge degrees of freedom in
940: the pseudogap states.
941: \cite{ME9916,TS0009,LL0301}
942: 
943: Finally, I remark that the above two-band picture provides the basis 
944: to develop new approximate schemes to describe doped Mott insulators.
945: The mean-field theory developed in Refs. \onlinecite{RW0501,RW0674}
946: constitutes one such example.
947: Remarkably, this approach reproduces a variety of experimental data. 
948: \cite{RW0501,RW0531}
949: 
950: 
951: 
952: 
953: \begin{acknowledgments}
954: This work was partially supported by the Funda\c c\~ao 
955: Calouste Gulbenkian Grant No. 58119 (Portugal), NSF Grant No. DMR-01-23156,
956: NSF-MRSEC Grant No. DMR-02-13282 and by the DOE Grant No. DE-AC02-05CH11231.
957: \end{acknowledgments}
958: 
959: 
960: \bibliography{hightc}
961: 
962: \end{document}
963: 
964: 
965: 
966: 
967: