1: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
2:
3:
4: \usepackage{graphicx}
5: \usepackage{dcolumn}
6: \usepackage{bm}
7:
8:
9: \newcommand{\mb}[1] {\mbox{\boldmath $#1$}}
10: \newcommand{\bq}{\begin{equation}}
11: \newcommand{\eq}{\end{equation}}
12: \newcommand{\bqa}{\begin{eqnarray}}
13: \newcommand{\eqa}{\end{eqnarray}}
14: \newcommand{\nn}{\nonumber \\}
15: \newcommand{\ij}{\langle i j \rangle}
16:
17:
18: \def\be {\begin{equation}}
19: \def\ee {\end{equation}}
20: \def\bea {\begin{eqnarray}}
21: \def\eea {\end{eqnarray}}
22: \def\bnn {\begin{eqnarray*}}
23: \def\enn {\end{eqnarray*}}
24: \def\dag {\dagger}
25: \def\f {\frac}
26: \def\rg {\rangle}
27: \def\laplaceR {\nabla_{\!\scriptscriptstyle R}^2}
28: \def\laplacer {\nabla_{\!\scriptscriptstyle r}^2}
29: \def\nc {\nonumber \\}
30: \def\gradR {\mb{\nabla}_{\!\scriptscriptstyle R}^{}}
31: \def\gradr {\mb{\nabla}_{\!\scriptscriptstyle r}^{}}
32: \def\bA {{\bf A}}
33: \def\bR {{\bf R}}
34: \def\br {{\bf r}}
35: \begin{document}
36:
37:
38:
39:
40:
41: \title{Heavy-fermion spin liquid in the strong hybridization limit of the finite-U Anderson lattice model}
42: \author{Ki-Seok Kim}
43: \affiliation{ School of Physics, Korea Institute for Advanced
44: Study, Seoul 130-012, Korea }
45: \date{\today}
46:
47: \begin{abstract}
48: Studying the finite-U Anderson lattice model in the strong
49: hybridization limit, we find a heavy-fermion spin liquid phase,
50: where both conduction and localized fermions are strongly
51: hybridized to form heavy fermions but this heavy-fermion phase
52: corresponds to a symmetric Mott insulating state owing to the
53: presence of charge gap, resulting from large Hubbard-U
54: interactions in localized fermions. We show that this
55: heavy-fermion spin liquid phase differs from the "fractionalized"
56: Fermi liquid state, where the latter corresponds to a metallic
57: state with a small Fermi surface of conduction electrons while
58: localized fermions decouple from conduction electrons to form a
59: spin liquid state. We discuss the stability of this anomalous spin
60: liquid phase against antiferromagnetic ordering and gauge
61: fluctuations, in particular, instanton effects associated with
62: confinement of slave particles. Furthermore, we propose a
63: variational wave function to check its existence from the
64: microscopic model.
65: \end{abstract}
66:
67: \pacs{71.10.-w, 71.10.Hf, 71.27.+a, 75.30.Mb}
68:
69: \maketitle
70:
71:
72:
73: \section{Introduction}
74:
75:
76:
77: Spin liquid has been intensively studied, motivated from the
78: resonating-valance-bond (RVB) scenario for anomalous finite
79: temperature physics of the Pseudogap phase in high $T_{c}$
80: cuprates.\cite{RVB_SL} However, such spin liquid physics is
81: fascinating itself because this phase has a nontrivial order
82: called topological or quantum order beyond the description of the
83: Landau-Ginzburg-Wilson paradigm for phase transitions, and its low
84: energy physics is described by a gauge theory, far from the
85: conventional structure of condensed matter theories, allowing
86: spin-fractionalized excitations called
87: spinons.\cite{Wen_Quantum_Order}
88:
89:
90:
91: Recently, several geometrically frustrated insulators are proposed
92: to be genuine symmetric Mott insulating phases, i.e., spin
93: liquids.\cite{Spin_liquids} Although such magnetic frustration can
94: be a strong candidate as the mechanism for the existence of spin
95: liquid, it is still interesting to find another mechanism for its
96: existence. In this paper we propose a heavy-fermion spin liquid
97: phase in the strong hybridization limit of the finite-U Anderson
98: lattice model (ALM).
99:
100:
101:
102: An important thing is how to suppress magnetic ordering.
103: Antiferromagnetic order is well known to occur at half filling in
104: the square lattice due to Fermi-nesting. One way to suppress the
105: antiferromagnetic ordering is to make spin singlets introducing
106: another band-electrons. Furthermore, if the nesting property can
107: be destroyed, such an ordering tendency will be much more
108: suppressed. This motivates us to consider the ALM since it shows a
109: large Fermi surface when conduction electrons are hybridized with
110: localized electrons.
111:
112:
113:
114: In this paper we consider the strong hybridization limit, turning
115: on on-site Hubbard interactions to localized electrons. Two
116: possibilities are expected. One is that strong local interactions
117: weaken hybridization, causing the "fractionalized" Fermi liquid
118: phase, where localized fermions form a spin liquid phase while
119: conduction electrons are in the Fermi liquid
120: state.\cite{Senthil_Kondo} However, Fermi-nesting in the localized
121: fermion-band would result in antiferromagnetic ordering.
122: Magnetically frustrated interactions are required for the
123: fractionalized Fermi liquid to realize. The other is that the
124: hybridization still survives against local interactions, but
125: charge fluctuations are suppressed due to such interactions. Since
126: Fermi-nesting does not exist and Kondo singlets are formed owing
127: to hybridization, antiferromagnetic ordering does not arise, but
128: this corresponds to an insulating state owing to charge gap.
129: Because such a phase contains all symmetries of the original
130: lattice model, it can be identified with a spin liquid state with
131: heavy neutral fermions, i.e., spinons. We propose that this
132: heavy-fermion spin liquid state can emerge in the strong
133: hybridization limit of the finite-U ALM.
134:
135:
136:
137: To describe Mott transition, it is necessary to take on-site
138: Hubbard interactions non-perturbatively. Recent studies on the
139: Hubbard model have revealed that the slave-rotor representation
140: can treat such interactions well, describing the Mott
141: transition.\cite{Kim_Rotor} Applying the U(1) slave-rotor
142: representation to the finite-U ALM, and performing gauge
143: transformation\cite{Kim_Kondo_SB} appropriate in the strong
144: hybridization limit, we find an effective U(1) gauge Lagrangian in
145: terms of renormalized conduction and localized fermions and
146: collective density-fluctuation bosons interacting via U(1)
147: slave-rotor gauge fluctuations. We show that such density
148: fluctuations associated with localized fermions become gapped,
149: increasing the Hubbard interaction-U in the strong hybridization
150: limit. This gapped phase is identified with an insulating phase
151: with all symmetries of the ALM, thus heavy-fermion spin liquid. We
152: discuss the stability of the spin liquid phase against gauge
153: fluctuations, in particular, instanton excitations associated with
154: confinement of slave-particles. Furthermore, we propose a
155: variational wave function to check its existence from not the
156: effective field theory but the microscopic model itself.
157:
158:
159:
160: It should be noted that the heavy-fermion spin liquid phase
161: differs from the fractionalized Fermi liquid
162: state\cite{Senthil_Kondo} because the former is an insulator, thus
163: there is no Fermi surface while the latter is a metal to have a
164: small Fermi surface of conduction electrons. In this respect the
165: present Mott transition in the ALM should be discriminated from
166: Kondo breakdown as the orbital-selective Mott
167: transition\cite{Pepin}. Furthermore, the spin liquid state in the
168: fractionalized Fermi liquid phase would be unstable against
169: antiferromagnetic ordering owing to the presence of Fermi-nesting.
170:
171:
172:
173: It is important to notice that the limit considered in this paper
174: is different from that studied in the context of heavy fermion
175: physics.\cite{Review_theory,Coleman_Review} Quantum phase
176: transitions in heavy fermion compounds are studied in the large-U
177: limit of the ALM, where the Kondo lattice model (KLM) or
178: infinite-U ALM is in the main interest. On the other hand, the
179: present study considers the strong hybridization limit, not
180: seriously taken into account in the KLM or infinite-U ALM context.
181:
182:
183:
184: \section{U(1) slave-rotor representation of the finite-U Anderson lattice model}
185:
186:
187:
188: We start from the finite-U ALM \bqa && L_{ALM} =
189: \sum_{i\sigma}c_{i\sigma}^{\dagger}(\partial_{\tau} -
190: \mu)c_{i\sigma} -
191: t\sum_{\ij\sigma}(c_{i\sigma}^{\dagger}c_{j\sigma} + H.c.) \nn &&
192: - V\sum_{i\sigma}(c_{i\sigma}^{\dagger}d_{i\sigma} + H.c.) +
193: \sum_{i\sigma}d_{i\sigma}^{\dagger}(\partial_{\tau} +
194: \epsilon_{d})d_{i\sigma} \nn && +
195: U\sum_{i}d_{i\uparrow}^{\dagger}d_{i\uparrow}d_{i\downarrow}^{\dagger}d_{i\downarrow}
196: , \eqa where $c_{i\sigma}$ represents conduction electrons with
197: dispersion $\epsilon_{k}^{c} = - 2t (\cos k_{x} + \cos k_{y})$ in
198: two dimensions, and $d_{i\sigma}$ localized electrons with
199: localized level $\epsilon_{d}$.
200:
201:
202:
203: Quantum phase transitions are expected to occur via competition
204: between the hybridization term with $V$ and the local interaction
205: term with $U$. Without the interaction term of $U$ this model is
206: exactly solvable, resulting in hybridization between the
207: conduction and localized bands. Our problem is to study what
208: happens to the hybridization when the local interaction $U$ is
209: turned on.
210:
211:
212:
213: The local interaction term can be decomposed into the charge and
214: spin channels \bqa &&
215: U\sum_{i}d_{i\uparrow}^{\dagger}d_{i\uparrow}d_{i\downarrow}^{\dagger}d_{i\downarrow}
216: =
217: \frac{U}{4}\sum_{i}(\sum_{\sigma}d_{i\sigma}^{\dagger}d_{i\sigma}
218: - 1)^{2} \nn && -
219: \frac{U}{4}\sum_{i}(\sum_{\sigma}\sigma{d}_{i\sigma}^{\dagger}d_{i\sigma})^{2}
220: + \frac{U}{2}\sum_{i\sigma}d_{i\sigma}^{\dagger}d_{i\sigma} -
221: \frac{U}{4}\sum_{i}1 . \eqa In this paper we do not consider the
222: spin channel, justified in the strong hybridization limit where
223: antiferromagnetic ordering is severely suppressed.
224:
225:
226:
227: Recently, we could describe the genuine Mott transition without
228: symmetry breaking, introducing the slave-rotor representation in
229: order to take into account local interactions non-perturbatively
230: in the Hubbard model.\cite{Kim_Rotor} Decomposing the localized
231: electron $d_{i\sigma} = e^{-i\theta_{i}}\eta_{i\sigma}$, one can
232: rewrite the ALM in the following way \bqa && L_{ALM} =
233: \sum_{i\sigma}c_{i\sigma}^{\dagger}(\partial_{\tau} -
234: \mu)c_{i\sigma} -
235: t\sum_{\ij\sigma}(c_{i\sigma}^{\dagger}c_{j\sigma} + H.c.) \nn &&
236: -
237: V\sum_{i\sigma}(c_{i\sigma}^{\dagger}e^{-i\theta_{i}}\eta_{i\sigma}
238: + H.c.) + \sum_{i\sigma}\eta_{i\sigma}^{\dagger}(\partial_{\tau} +
239: \epsilon_{d})\eta_{i\sigma} \nn && + \frac{U}{4}\sum_{i}L_{i}^{2}
240: - i\sum_{i}L_{i}\partial_{\tau}\theta_{i} \nn && +
241: i\sum_{i}\varphi_{i}(L_{i} -
242: [\sum_{\sigma}\eta_{i\sigma}^{\dagger}\eta_{i\sigma} - 1]) , \eqa
243: where $\epsilon_{d}$ is replaced with $\epsilon_{d} - U/2$. It is
244: easy to show that Eq. (3) is exactly the same as Eq. (1) with Eq.
245: (2) after integrating out the $\varphi_{i}$ field with
246: $\eta_{i\sigma} = e^{i\theta_{i}}d_{i\sigma}$. In this expression
247: the electron Hilbert space $|d_{i\sigma}>$ is given by the direct
248: product of the fermion and boson Hilbert spaces
249: $|\eta_{i\sigma}>\bigotimes{|L_{i}>}$ according to the
250: decomposition $d_{i\sigma} = e^{-i\theta_{i}}\eta_{i\sigma}$,
251: where $L_{i}$ represents an electron density at site $i$. It is
252: clear that any decomposition method enlarges the original electron
253: Hilbert space, thus an appropriate constraint associated with the
254: decomposition should be imposed. The Lagrange multiplier field
255: $\varphi_{i}$ expresses the U(1) slave-rotor constraint $L_{i} =
256: \sum_{\sigma}\eta_{i\sigma}^{\dagger}\eta_{i\sigma} - 1$, implying
257: that the fermion and boson Hilbert spaces are not independent,
258: thus the two operators $\eta_{i\sigma}$ and $e^{i\theta_{i}}$
259: also. Then, $e^{-i\theta_{i}}$ is identified with an annihilation
260: operator of an electron charge owing to the constraint $L_{i} =
261: \sum_{\sigma}\eta_{i\sigma}^{\dagger}\eta_{i\sigma} - 1$ and the
262: canonical relation $[L_{i}, \theta_{j}] = -i\delta_{ij}$ imposed
263: by $- iL_{i}\partial_{\tau}\theta_{i}$. In this respect collective
264: density fluctuations associated with localized fermions can be
265: taken into account non-perturbatively in the U(1) slave-rotor
266: representation.
267:
268:
269:
270: \section{Heavy fermion physics: Kondo breakdown as the orbital-selective Mott transition}
271:
272:
273:
274: The problem is how to treat the hybridization term with
275: $e^{-i\theta_{i}}$. One direct way is to integrate out conduction
276: electrons and obtain an effective Lagrangian in terms of localized
277: fermions and their density-fluctuation bosons.\cite{Senthil_Kondo}
278: Integrating out the density field $L_{i}$ and conduction electron
279: field $c_{i\sigma}$, we obtain \bqa && L_{eff} =
280: \sum_{i\sigma}\eta_{i\tau\sigma}^{\dagger}(\partial_{\tau} +
281: \epsilon_{d})\eta_{i\tau\sigma} -
282: i\sum_{i}\varphi_{i\tau}[\sum_{\sigma}{\eta}^{\dagger}_{i\tau\sigma}{\eta}_{i\tau\sigma}
283: - 1] \nn && -
284: \sum_{ij\sigma}\eta_{i\tau\sigma}^{\dagger}e^{i\theta_{i\tau}}\Delta^{c}_{ij,\tau\tau'}e^{-i\theta_{j\tau'}}\eta_{j\tau'\sigma}
285: + \frac{1}{U}\sum_{i}(\partial_{\tau}\theta_{i\tau} -
286: \varphi_{i\tau})^{2} , \nn \eqa where $\Delta^{c}_{ij,\tau\tau'}$
287: is the single particle propagator of the conduction electron,
288: given by \bqa && \Delta^{c}_{q,\omega} = \frac{V^{2}}{i\omega +
289: \mu - \epsilon_{q}^{c}} \nonumber \eqa in the energy-momentum
290: space.
291:
292:
293:
294: Although the above treatment itself is exact so far, it has an
295: important assumption that the identity of conduction electrons is
296: sustained. In other words, feedback effects of localized fermions
297: and density fluctuations to conduction electrons, self-energy
298: corrections of conduction electrons, are not introduced. Since
299: such feedback effects can be induced only via the hybridization
300: coupling term, this treatment may be regarded as the weak
301: hybridization approach, not justified in the strong hybridization
302: limit.\cite{HMM} Alternatively, it may be viewed as the
303: large-U-limit approach owing to small $V/U$, where the infinite-U
304: ALM can be considered or the KLM is taken via virtual charge
305: fluctuations. This type of treatment has been utilized both
306: intensively and extensively, associated with the
307: single-impurity\cite{FG} and heavy fermion
308: physics\cite{Coleman_Review}.
309:
310:
311:
312: Performing the Hubbard-Stratonovich transformation for the
313: non-local hopping term in both time and space, we obtain an
314: effective Lagrangian \bqa && L_{eff} =
315: \sum_{i\sigma}\eta_{i\sigma}^{\dagger}(\partial_{\tau} +
316: \epsilon_{d})\eta_{i\sigma} -
317: i\sum_{i}\varphi_{i}[\sum_{\sigma}{\eta}^{\dagger}_{i\sigma}{\eta}_{i\sigma}
318: - 1] \nn && + \sum_{q} \chi_{q}^{\eta*}(\partial_{\tau} - \mu +
319: \epsilon_{q}^{c})\chi_{q}^{\theta} +
320: \frac{1}{U}\sum_{i}(\partial_{\tau}\theta_{i} - \varphi_{i})^{2}
321: \nn && -
322: V\sum_{ij}[\sum_{\sigma}\eta_{i\sigma}^{\dagger}\chi_{ij}^{\theta}\eta_{j\sigma}
323: + e^{i\theta_{i}}\chi_{ij}^{\eta *}e^{-i\theta_{j}} ] , \eqa where
324: $\chi_{ij}^{\eta}$ and $\chi_{ij}^{\theta}$ are effective hopping
325: parameters determined in the self-consistent analysis \bqa &&
326: \chi_{ij}^{\eta*} = V \langle{e^{i\theta_{i}}(\partial_{\tau} -
327: \mu + \epsilon_{ij}^{c})^{-1}e^{-i\theta_{j}}}\rangle , \nn &&
328: \chi_{ij}^{\theta} = V
329: \langle{\eta_{i\sigma}^{\dagger}(\partial_{\tau} - \mu +
330: \epsilon_{ij}^{c})^{-1}\eta_{j\sigma}}\rangle . \eqa
331:
332:
333:
334: It is not easy to perform the self-consistent analysis because
335: such effective hopping parameters have both frequency and momentum
336: dependencies (nonlocal in time and space). This non-locality may
337: give rise to crucial effects to the heavy fermion physics. Since
338: this is not our main subject, we leave its detailed analysis in an
339: important future problem, and here, discuss what is expected in
340: this treatment.
341:
342:
343:
344: Generally speaking, two phases would be allowed in this
345: approximation, determined by the condensation of slave-rotor
346: bosons. In the large $V/U$ limit collective density-fluctuation
347: bosons become condensed, and heavy-fermion Fermi liquid appears to
348: form a large Fermi surface since $\langle{e^{i\theta_{i}}}\rangle
349: \not= 0$ results in
350: $\langle{c_{i\sigma}^{\dagger}\eta_{i\sigma}}\rangle \not= 0$ as
351: shown in the hybridization term of Eq. (3). In the small $V/U$
352: limit such boson excitations become gapped, and localized fermions
353: decouple from conduction electrons owing to
354: $\langle{c_{i\sigma}^{\dagger}\eta_{i\sigma}}\rangle = 0$, forming
355: a spin liquid state. Such a phase is called the fractionalized
356: Fermi liquid state with a small Fermi surface of conduction
357: electrons.\cite{Senthil_Kondo} As a result, the heavy-fermion
358: Fermi liquid to fractionalized Fermi liquid transition is
359: identified with Kondo breakdown as the orbital-selective Mott
360: transition\cite{Pepin} in the U(1) slave-rotor representation of
361: the finite-U ALM. However, the spin liquid state would be unstable
362: against antiferromagnetic ordering owing to Fermi-nesting in the
363: square lattice.
364:
365:
366:
367: Although one important issue, that is, the abrupt volume-change of
368: the Fermi surface at the heavy-fermion quantum critical
369: point,\cite{YbRh2Si2} can be explained by the Kondo breakdown
370: transition as discussed above, the fact that the antiferromagnetic
371: transition of localized fermions arises simultaneously at the same
372: quantum critical point as the Kondo breakdown transition is still
373: far from our understanding because such a phenomenon is beyond the
374: description of the Landau-Ginzburg-Wilson theoretical framework
375: for phase transitions.\cite{LGW_Why} How to incorporate such an
376: antiferromagnetic transition in the Kondo breakdown one is an
377: important open problem.
378:
379:
380:
381: \section{Strong hybridization approach}
382:
383:
384:
385: In the previous section we have discussed the heavy fermion
386: physics in the "large-U" (more accurately, weak hybridization)
387: limit of the ALM although its mathematical construction is
388: performed at finite-U. In this section we take the strong
389: hybridization approach, considering that the hybridization term is
390: relevant in the renormalization group sense when $U = 0$. We show
391: that this type of approach allows a new possible phase called the
392: heavy-fermion spin liquid in the finite-U ALM.
393:
394:
395:
396: \subsection{Gauge transformation and effective Lagrangian}
397:
398:
399:
400: The relevance of the hybridization coupling in $U = 0$ motivates
401: us to consider the gauge transformation of $c_{i\sigma} =
402: e^{-i\theta_{i}}\psi_{i\sigma}$, where this type of approach has
403: been utilized in the context of quantum disordered
404: superconductors\cite{Quantum_disordered_SC}. Integrating out the
405: density field $L_{i}$ and inserting another slave-rotor
406: decomposition $c_{i\sigma} = e^{-i\theta_{i}}\psi_{i\sigma}$ into
407: Eq. (3), we find the following expression \bqa && Z =
408: \int{D\psi_{i\sigma}D\eta_{i\sigma}D\theta_{i}D\varphi_{i} }
409: \exp\Bigl[ - \int{d\tau} \Bigl\{
410: \sum_{i\sigma}\psi_{i\sigma}^{\dagger}(\partial_{\tau} \nn && -
411: \mu - i\partial_{\tau}\theta_{i})\psi_{i\sigma} -
412: t\sum_{\ij\sigma}(\psi_{i\sigma}^{\dagger}e^{i\theta_{i}}e^{-i\theta_{j}}\psi_{j\sigma}
413: + H.c.) \nn && - V\sum_{i\sigma}(\psi_{i\sigma}^{\dagger}
414: \eta_{i\sigma} + H.c.) +
415: \sum_{i\sigma}\eta_{i\sigma}^{\dagger}(\partial_{\tau} +
416: \epsilon_{d})\eta_{i\sigma} \nn && -
417: i\sum_{i}\varphi_{i}[\sum_{\sigma}{\eta}^{\dagger}_{i\sigma}{\eta}_{i\sigma}
418: - 1] + \frac{1}{U}\sum_{i}(\partial_{\tau}\theta_{i} -
419: \varphi_{i})^{2} \Bigr\} \Bigr] . \eqa
420:
421:
422:
423:
424:
425:
426: Couplings between charge fluctuations and renormalized conduction
427: electrons can be decomposed using the Hubbard-Stratonovich
428: transformation. Following Ref. \cite{Kim_Rotor}, we find the
429: effective Lagrangian from Eq. (7) \bqa && L_{eff} = L_{0} +
430: L_{\psi} + L_{\eta} + L_{V} + L_{\theta} , \nn && L_{0} =
431: \frac{1}{U}\sum_{i}q_{ri}^{2} - \frac{2}{U}\sum_{i}q_{ri}
432: (\varphi_{ri} - p_{i}) + \sum_{\ij}txy , \nn && L_{\psi} =
433: \sum_{i\sigma}\psi_{i\sigma}^{\dagger}(\partial_{\tau} -
434: \mu)\psi_{i\sigma} -
435: i\sum_{i}p_{i}[\sum_{\sigma}\psi_{i\sigma}^{\dagger}
436: \psi_{i\sigma} - 1] \nn && -
437: tx\sum_{\ij\sigma}(\psi_{i\sigma}^{\dagger}e^{-ia_{ij}}\psi_{j\sigma}
438: + H.c.) , \nn && L_{\eta} =
439: \sum_{i\sigma}\eta_{i\sigma}^{\dagger}(\partial_{\tau} +
440: \epsilon_{d})\eta_{i\sigma} -
441: i\sum_{i}(\varphi_{ri}-q_{ri})[\sum_{\sigma}{\eta}^{\dagger}_{i\sigma}{\eta}_{i\sigma}
442: - 1] , \nn && L_{V} = - V\sum_{i\sigma}(\psi_{i\sigma}^{\dagger}
443: \eta_{i\sigma} + H.c.) , \nn && L_{\theta} =
444: \frac{1}{U}\sum_{i}(\partial_{\tau}\theta_{i} - \varphi_{ri})^{2}
445: - 2ty\sum_{\ij}\cos(\theta_{i}-\theta_{j}+a_{ij}) , \eqa where the
446: hopping parameters are represented as $x_{ij} = xe^{ia_{ij}}$ and
447: $y_{ij} = y e^{ia_{ij}}$, and $\varphi_{i} = \varphi_{ri} -
448: q_{ri}$ is used with $q_{ri} = {Uq_{i}}/{2}$. The unidentified
449: parameters can be determined self-consistently in the saddle-point
450: analysis \bqa && p_{i} = \langle
451: \partial_{\tau}\theta_{i}\rangle , ~~~~~ q_{ri} =
452: i\frac{U}{2}\langle\sum_{\sigma}\psi_{i\sigma}^{\dagger}\psi_{i\sigma}
453: - 1\rangle , \nn && x = |\langle e^{-i\theta_{i}}e^{i\theta_{j}}
454: + H.c.\rangle| , ~~~~~ y =
455: |\langle\sum_{\sigma}\psi_{i\sigma}^{\dagger}\psi_{j\sigma} +
456: H.c.\rangle| . \nn \eqa $p_{i}$ and $q_{ri}$ play the role of
457: renormalization for the fermion and boson chemical potentials,
458: respectively.
459:
460:
461:
462: The effective Lagrangian Eq. (8) is the main result of this paper,
463: showing an important feature of the finite-U ALM, that is, the
464: presence of two kinds of relevant interactions. As mentioned
465: before, the hybridization coupling is relevant to give rise to a
466: band-hybridized metal if the local charge-fluctuation energy is
467: not taken into account. On the other hand, an appropriate
468: treatment of local charge fluctuations has been shown to result in
469: the Mott transition from a spin liquid Mott insulator to a Fermi
470: liquid metal at a critical value $U_{c}$ without hybridization,
471: i.e., in the Hubbard model.\cite{Kim_Rotor}
472:
473:
474:
475: Compared to the effective Lagrangian Eq. (5) for the heavy-fermion
476: quantum phase transition, the effective Lagrangian Eq. (8) always
477: allows the band-hybridization while Eq. (5) does not. In Eq. (5)
478: the boson condensation corresponds to the hybridization transition
479: since it gives an effective hybridization coupling constant
480: $V_{eff} = |\langle{e^{-i\theta_{i}}}\rangle|V$. Such a transition
481: is involved with the abrupt volume change of the Fermi surface. On
482: the other hand, in Eq. (8) the renormalized conduction and
483: localized fermions are always hybridized, forming a large Fermi
484: surface of spinons (neutral fermions). Then, the boson
485: condensation in Eq. (8) corresponds to the genuine Mott transition
486: in these heavy fermions.
487:
488:
489:
490: Since the slave-rotor variable has its own dynamics in the
491: saddle-point approximation, given by the U(1) rotor (XY) model,
492: the quantum transition described by condensation of the rotor
493: field occurs when $Dy/U \sim 1$ with the half bandwidth $D$, where
494: $Dy$ is an effective bandwidth for the rotor
495: variable.\cite{Rotor_model} It is important to understand that the
496: ratio $V/U$ between the hybridization and on-site interactions
497: controls the hopping parameter $y$, where the hopping parameter
498: decreases as this ratio decreases. This results in the quantum
499: critical point $(V/U)_{c}$ between the heavy-fermion Fermi liquid
500: and heavy-fermion spin liquid, which differs from the
501: heavy-fermion quantum critical point between the heavy-fermion and
502: fractionalized Fermi liquids.
503:
504:
505:
506: It is interesting to notice that this treatment [Eq. (8)] gives
507: rise to two different energy scales.\cite{Two_scales} One
508: corresponds to the band-hybridization scale $T_K$, and the other
509: is associated with the coherence scale $T_{FL}$, resulting in
510: Fermi liquid. $T_{K}$ is proportional to the hybridization gap $V$
511: in the strong coupling approach, consistent with the single
512: impurity energy scale in the strong coupling limit. On the other
513: hand, $T_{FL}$ is proportional to the effective stiffness
514: parameter $Dy|\langle{e^{i\theta_{i}}}\rangle|^{2}$ controlling
515: the phase coherence of $\theta_{i}$.\cite{TK_TFL}
516:
517:
518:
519: \subsection{Mean-field analysis}
520:
521:
522:
523: For the saddle-point analysis we resort to large $N$
524: generalization replacing $e^{i\theta_{i}}$ with $\phi_{is}$, where
525: $s = 1, ..., N$.\cite{Kim_Kim} Taking the mean-field ansatz of
526: $ip_{i} = p$, $i\varphi_{ri} = \varphi_{r}$, $iq_{ri} = - q_{r}$,
527: and $i \lambda_{i} = \lambda$ where $\lambda_{i}$ introduces the
528: rotor constraint $\sum_{s}|\phi_{is}|^{2} = 1$, we obtain the free
529: energy functional \bqa && F_{MF} = \nn && -
530: \frac{1}{\beta}\sum_{k\sigma}\sum_{\omega_{n}}\ln(i\omega_{n} -
531: E_{k+}) -
532: \frac{1}{\beta}\sum_{k\sigma}\sum_{\omega_{n}}\ln(i\omega_{n} -
533: E_{k-}) \nn && + \frac{1}{\beta}\sum_{ks}\sum_{\nu_n}\ln\Bigl( -
534: \frac{1}{U}[i\nu_{n} + \varphi_{r}]^{2} + y\epsilon_{k}^{\phi} +
535: \lambda \Bigr) \nn && + \sum_{k}\Bigl(Dxy - \frac{2}{U}
536: q_{r}[\varphi_{r} - p + \frac{q_{r}}{2}] + p + \varphi_{r} + q_{r}
537: \nn && - \lambda + \mu [1-\delta]\Bigr) . \eqa Here the
538: renormalized fermion spectrum is given by $E_{k\pm} =
539: [\frac{(x\epsilon_{k}^{\psi}-\mu - p ) + (\epsilon_{d} -
540: \varphi_{r} - q_{r})}{2}] \pm
541: \sqrt{[\frac{(x\epsilon_{k}^{\psi}-\mu - p ) - (\epsilon_{d} -
542: \varphi_{r} - q_{r})}{2}]^{2} + V^{2}}$, and $\delta$ is hole
543: concentration for the conduction band. $\omega_{n}$ ($\nu_{n}$) is
544: the Matzubara frequency for fermions (bosons).
545:
546:
547:
548:
549:
550:
551: Minimizing the free energy Eq. (10) with respect to $\lambda$,
552: $x$, $y$, $q_{r}$, $\varphi_{r}$, $p$, and $\mu$, we obtain the
553: self-consistent mean-field equations. Performing the Matzubara
554: frequency summations and momentum integrals with $\sum_{k} =
555: \frac{1}{2D}\int_{-D}^{D}d\epsilon$ (constant density of
556: states\cite{Kim_Rotor,Kim_Kim}), we find \bqa && 1 =
557: \frac{\sqrt{U(\lambda+Dy)} -
558: \sqrt{U(\lambda+y\epsilon_{\theta})}}{Dy} , ~~~~~
559: \epsilon_{\theta} = \frac{1}{y}\Bigl(\frac{\varphi_{r}^{2}}{U} -
560: \lambda\Bigr) , \nn && x = \frac{(2\lambda - Dy)\sqrt{U(\lambda +
561: Dy)} - (2\lambda -
562: y\epsilon_{\theta})\sqrt{U(\lambda+y\epsilon_{\theta})}}{3(Dy)^2}
563: , \nn && y = \frac{1}{4}\Bigl(1 - \frac{\epsilon_{\psi}^2}{D^2}
564: \Bigr) + \frac{V^2}{2D^2x^2} \Bigl[ \ln \Bigl\{
565: \frac{\sin[\tan^{-1}(\frac{x\epsilon_{\psi}-\mu_r}{2V})]-1}{-\sin[\tan^{-1}(\frac{Dx+\mu_r}{2V}
566: )]-1}\Bigr\} \nn && -
567: \Bigl\{\frac{1}{\sin[\tan^{-1}(\frac{x\epsilon_{\psi}-\mu_r}{2V})]-1}
568: - \frac{1}{-\sin[\tan^{-1}(\frac{Dx+\mu_r}{2V} )]-1} \Bigr\} \nn
569: && - \ln \Bigl\{
570: \frac{\sin[\tan^{-1}(\frac{x\epsilon_{\psi}-\mu_r}{2V})]+1}{-\sin[\tan^{-1}(\frac{Dx+\mu_r}{2V}
571: )]+1}\Bigr\} \nn && -
572: \Bigl\{\frac{1}{\sin[\tan^{-1}(\frac{x\epsilon_{\psi}-\mu_r}{2V})]+1}
573: - \frac{1}{-\sin[\tan^{-1}(\frac{Dx+\mu_r}{2V} )]+1} \Bigr\}
574: \Bigr] \nn && +
575: \frac{\mu_rV}{D^2x^2}\Bigl[\frac{1}{\cos[\tan^{-1}(\frac{x\epsilon_{\psi}-\mu_r}{2V})]}
576: - \frac{1}{\cos[\tan^{-1}(\frac{Dx+\mu_r}{2V} )]} \Bigr] , \nn &&
577: \epsilon_{\psi} = \frac{1}{x}\Bigl( \frac{V^2}{\epsilon_{d} -
578: q_{r}-\varphi_{r}} + \mu + p \Bigr) , \nn &&
579: \frac{2}{U}(\varphi_{r} - p + q_r) - 1 = - \frac{1}{2}\Bigl(1 +
580: \frac{\epsilon_{\psi}}{D}\Bigr) \nn && -
581: \frac{V}{Dx}\Bigl[\frac{1}{\cos[\tan^{-1}(\frac{x\epsilon_{\psi}-\mu_r}{2V})]}
582: - \frac{1}{\cos[\tan^{-1}(\frac{Dx+\mu_r}{2V} )]} \Bigr] , \nn &&
583: \frac{2}{U}(\varphi_{r} - p) = (1 + \frac{\epsilon_{\theta}}{D}) =
584: ( 1 - \frac{\epsilon_{\psi}}{D}) , ~~~~~ q_{r} = - \frac{U}{2}
585: \delta , \nn && \mu_{r} = \mu + p + \epsilon_{d} - q_r -
586: \varphi_r . \eqa
587:
588:
589:
590: Condensation of the density-fluctuation bosons occurs when their
591: excitation gap closes or its effective chemical potential
592: vanishes, i.e., $\varphi_{r} = 0$. This causes $p = 0$, resulting
593: in $\epsilon_{\theta} = - D$ ($\epsilon_{\psi} = D$). We obtain
594: $\lambda = Dy$ at the quantum critical point, thus find the
595: relation $Dy/U = 2$, completely consistent with the previous
596: qualitative analysis for the rotor model.\cite{Rotor_model} The
597: hopping parameter for renormalized conduction fermions is $x =
598: 1/3$ at the quantum critical point. As a result, the quantum
599: critical point $(V/D, U/D)_{c}$ is determined by \bqa &&
600: \frac{2U}{D} = \frac{V^2}{2(D/3)^2} \Bigl[ \ln \Bigl\{
601: \frac{\sin[\tan^{-1}(\frac{D/3-\mu_r}{2V})]-1}{-\sin[\tan^{-1}(\frac{D/3+\mu_r}{2V}
602: )]-1}\Bigr\} \nn && -
603: \Bigl\{\frac{1}{\sin[\tan^{-1}(\frac{D/3-\mu_r}{2V})]-1} -
604: \frac{1}{-\sin[\tan^{-1}(\frac{D/3+\mu_r}{2V} )]-1} \Bigr\} \nn &&
605: - \ln \Bigl\{
606: \frac{\sin[\tan^{-1}(\frac{D/3-\mu_r}{2V})]+1}{-\sin[\tan^{-1}(\frac{D/3+\mu_r}{2V}
607: )]+1}\Bigr\} \nn && -
608: \Bigl\{\frac{1}{\sin[\tan^{-1}(\frac{D/3-\mu_r}{2V})]+1} -
609: \frac{1}{-\sin[\tan^{-1}(\frac{D/3+\mu_r}{2V} )]+1} \Bigr\} \Bigr]
610: \nn && +
611: \frac{\mu_rV}{(D/3)^2}\Bigl[\frac{1}{\cos[\tan^{-1}(\frac{D/3-\mu_r}{2V})]}
612: - \frac{1}{\cos[\tan^{-1}(\frac{D/3+\mu_r}{2V} )]} \Bigr] , \nn &&
613: \delta =
614: \frac{V}{D/3}\Bigl[\frac{1}{\cos[\tan^{-1}(\frac{D/3-\mu_r}{2V})]}
615: - \frac{1}{\cos[\tan^{-1}(\frac{D/3+\mu_r}{2V} )]} \Bigr] , \nn &&
616: \mu_r = \frac{D}{3} - \Bigl( \frac{V^2}{\epsilon_{d} +
617: \frac{U}{2}\delta } - [\epsilon_{d} + \frac{U}{2}\delta ] \Bigr) .
618: \eqa Actually, we find $(0.112,0.044)$ at $\delta = 0.800$ and
619: $\epsilon_{d}/D = - 0.001$, thus $(V/U)_{c} = 2.545$.
620:
621:
622:
623: This quantum phase transition results from gapping of density
624: fluctuations of localized fermions in the presence of strong
625: hybridization with renormalized conduction fermions, thus
626: differing from the Kondo breakdown transition\cite{Senthil_Kondo}
627: in the ALM. In $V/U > (V/U)_{c}$ collective density fluctuations
628: become softened, and the saddle-point equation of $\lambda$ is
629: replaced with $1 = \sqrt{2U/Dy} + Z$, where $Z =
630: |\langle{e^{i\theta_{i}}}\rangle|^{2}$ is the condensation
631: amplitude. Thus, the coherence temperature is given by $T_{FL}
632: \approx (Dy - 2U)^{1/2}$ near the heavy-fermion Mott critical
633: point, and below this temperature the valance-fluctuation-induced
634: heavy-fermion phase arises.
635:
636:
637:
638: \section{Heavy-fermion spin liquid to heavy-fermion Fermi liquid quantum transition: its critical field theory and non-Fermi liquid physics}
639:
640:
641:
642: To investigate low energy physics near the heavy-fermion spin
643: liquid to Fermi liquid quantum critical point, we derive an
644: effective field theory from Eq. (8). From the fermion mean-field
645: Lagrangian $L_{\chi} =
646: \sum_{k\sigma}\Bigl[\chi_{k\sigma+}(\partial_{\tau} +
647: E_{k+})\chi_{k\sigma+} + \chi_{k\sigma-}(\partial_{\tau} +
648: E_{k-})\chi_{k\sigma-}\Bigr]$ where $\psi_{k\sigma} =
649: u_{k}\chi_{k\sigma+} + v_{k}\chi_{k\sigma-}$ and $\eta_{k\sigma} =
650: v_{k}\chi_{k\sigma+} - u_{k}\chi_{k\sigma-}$ with $u_{k} =
651: {V}/{\sqrt{(E_{k+} - \epsilon_{k}^{\psi})^{2} + V^{2}}}$ and
652: $v_{k} = - ({E_{k+} - \epsilon_{k}^{\psi}})/{\sqrt{(E_{k+} -
653: \epsilon_{k}^{\psi})^{2} + V^{2}}}$, we find its critical field
654: theory ${\cal L}_{\chi} = \chi_{r\sigma}^{\dagger}(\partial_{\tau}
655: - \mu_{c})\chi_{r\sigma} + \frac{1}{2m_{\chi}}|(\nabla -
656: i\mathbf{a}_{r})\chi_{r\sigma}|^{2}$. Here the renormalized
657: fermion field $\chi_{r\sigma}$ represents $\chi_{i\sigma-}$ in the
658: continuum limit, and the effective band mass is given by $m_{\chi}
659: = \Bigl({\partial^{2}E_{k-}}/{\partial k^2}\Bigr)^{-1} =
660: {m_{e}}/\Bigl[{\frac{\epsilon_{dr}}{\epsilon_{dr}^2 +
661: V^2}\Bigl(\mu_{c} + \frac{V^2}{\epsilon_{dr}}\Bigr) - 1}\Bigr]$
662: with the "bare" band mass $m_{e} \sim (tx)^{-1}$, renormalized
663: localized level $\epsilon_{dr} = \epsilon_{d} +
664: \frac{U}{2}\delta$, and critical chemical potential $\mu_{c}$.
665: Note that $m_{\chi}$ diverges in the limit of $V \rightarrow
666: \infty$.
667:
668:
669:
670: The effective boson Lagrangian can be written with an
671: electromagnetic field $\mathbf{A}$ as ${\cal L}_{\phi} =
672: [(\partial_{\tau} +
673: i\varphi_{r})\phi_{rs}^{\dagger}][(\partial_{\tau} -
674: i\varphi_{r})\phi_{rs}] + |(\nabla - i\mathbf{a}_{r} -
675: i\mathbf{A})\phi_{rs}|^{2} + m_{\phi}^{2}|\phi_{rs}|^{2} +
676: \frac{u_{\phi}}{2}|\phi_{rs}|^{2}$, where the rotor field
677: $e^{i\theta_{r}}$ is replaced with $\phi_{rs}$, and the rotor
678: constraint is softened via introduction of local interactions
679: $u_{\phi}$. $m_{\phi}$ is an effective mass, given by
680: $m_{\phi}^{2} = (V/U)_{c} - (V/U)$. An important point is that the
681: renormalized boson chemical potential $\varphi_{r}$ becomes zero
682: at the quantum critical point ($m_{\phi}^{2} = 0$), thus the
683: linear time-derivative term vanishes.
684:
685:
686:
687:
688:
689:
690:
691: We find the critical field theory at the heavy-fermion spin liquid
692: quantum critical point \bqa && S_{c} = \int{d\tau}d^2x \Bigl[
693: \chi_{r\sigma}^{\dagger}(\partial_{\tau} - \mu_{c})\chi_{r\sigma}
694: + \frac{1}{2m_{\chi}}|(\nabla -
695: i\mathbf{a}_{r})\chi_{r\sigma}|^{2} \nn && +
696: |\partial_{\tau}\phi_{rs}|^{2} + |(\nabla - i\mathbf{a}_{r} -
697: i\mathbf{A})\phi_{rs}|^{2} + \frac{u_{\phi}}{2}|\phi_{rs}|^{2}
698: \Bigr] \nn && + S_{eff}[\mathbf{a}_r] , \eqa where the critical
699: gauge action is given by $S_{eff}[\mathbf{a}_{r}] =
700: \frac{1}{2}\sum_{q,\omega_{n}}\Bigl(
701: \gamma_{F}\frac{|\omega_{n}|}{q} + \frac{N}{8}q\Bigr)
702: \Bigl(\delta_{ij} - \frac{q_{i}q_{j}}{q^2}\Bigr)a_{ir}a_{jr}$ with
703: the damping strength $\gamma_{F} = k_{F}/\pi$ and boson flavor
704: number $N$.\cite{Kim_Kim} The first term represents dissipative
705: dynamics in gauge fluctuations due to particle-hole excitations of
706: $\chi_{r\sigma}$ fermions near the Fermi surface, and the second
707: term arises from critical boson fluctuations at the quantum
708: critical point.\cite{Kim_Kim} The Maxwell gauge action is omitted
709: since it is irrelevant in the renormalization group sense. Note
710: that the time component of the gauge field mediates a local
711: interaction, thus safely ignored in the low energy
712: limit.\cite{Sigma}
713:
714:
715:
716: In the random phase approximation one finds the following
717: expression for the free energy $F/V = \int\frac{d^2q}{(2\pi)^{2}}
718: \int\frac{d\omega}{2\pi}\coth\Bigl[\frac{\omega}{2T}\Bigr]
719: \tan^{-1}\Bigl[\frac{{\rm Im}D(q,\omega)}{{\rm
720: Re}D(q,\omega)}\Bigr]$, where the gauge kernel is given by
721: $D(q,\omega) = \Bigl( - i \gamma_{F}\frac{\omega}{q} +
722: \frac{N}{8}q\Bigr)^{-1}$ in real frequency at the quantum critical
723: point. This free energy leads to the divergent specific heat
724: coefficient $\gamma = {C}_{V}/T \propto - \ln T$. The dc
725: conductivity is given by the Ioffe-Larkin combination-rule
726: $\sigma_{tot} = {\sigma_{\phi}\sigma_{\chi}}/({\sigma_{\phi}+
727: \sigma_{\chi}})$, assuming gaussian gauge
728: fluctuations.\cite{IL_sigma} In the one loop diagram of
729: gauge-boson exchange one finds ${1}/{\tau_{tr}^{\chi}} \sim
730: T^2$\cite{Kim_Kim} and $1/{\tau_{tr}^{\phi}} \sim T$\cite{Sigma},
731: where $1/{\tau_{tr}^{\chi}}$ ($1/{\tau_{tr}^{\phi}}$) is the
732: scattering rate of the $\chi_{r\sigma}$ fermion ($\phi_{rs}$
733: boson) with temperature $T$. As a result, the total dc
734: conductivity is given by $\sigma_{tot} \sim T^{-1}$ in low
735: temperatures, giving rise to the linear resistivity $\rho_{tot}
736: \sim T$ at the quantum critical point. On the other hand, in $V/U
737: < (V/U)_{c}$ but $|V/U - (V/U)_{c}| \rightarrow 0$ corresponding
738: to the heavy-fermion spin liquid state in the zero temperature
739: limit, the gauge kernel shows the crossover behavior from
740: $D(q,\omega) = \Bigl( - i \gamma_{F}\frac{\omega}{q} +
741: \frac{N}{8}q\Bigr)^{-1}$ to $D(q,\omega) = \Bigl( - i
742: \gamma_{F}\frac{\omega}{q} + \frac{q^2}{g}\Bigr)^{-1}$ with an
743: effective internal gauge-charge $g$ as temperature goes down. Here
744: the crossover energy scale would be an excitation gap of
745: $\phi_{rs}$, given by $|\varphi_{r}|$. Accordingly, the specific
746: heat coefficient exhibits the upturn behavior from $\gamma \sim -
747: \ln T$ to $\gamma \sim T^{-1/3}$. But, the resistivity still
748: behaves as $\rho_{tot} \sim T$ owing to $\rho_{\chi} \sim T^{4/3}$
749: in the $z = 3$ critical field theory,\cite{Sigma} where $z$ is the
750: dynamical exponent.
751:
752:
753:
754: \section{Discussion and summary}
755:
756:
757:
758: \subsection{Stability of the heavy-fermion spin liquid against gauge fluctuations}
759:
760:
761:
762: The stability of the heavy-fermion spin liquid state against gauge
763: fluctuations is an important problem for this phase to realize in
764: the finite-U ALM. It has been shown that the fermion-gauge action
765: in Eq. (13), obtained via integrating out gapped boson
766: excitations, has an infrared stable interacting fixed point with
767: an effective nonzero internal charge when the flavor number or
768: density of fermions is sufficiently large.\cite{IR_FP} Actually,
769: the present author has studied that such a fixed point can arise
770: when the fermion conductivity, given by its current-current
771: correlation function, is sufficiently large.\cite{Kim_SL_FS}
772: Notice that the fermion conductivity is associated with the
773: density of fermions. At this fixed point the critical field theory
774: is characterized by the dynamical exponent $z = 3$, as discussed
775: before.
776:
777:
778:
779: The problem is the stability of this fixed point against instanton
780: excitations resulting from compact gauge fluctuations. A similar
781: situation appears in the compact QED$_{3}$, where Dirac fermions
782: interact via compact U(1) gauge fluctuations. The infrared
783: interacting fixed point in the QED$_{3}$ has been shown to be
784: stable against instanton fluctuations since the scaling dimension
785: of an instanton insertion operator is proportional to the flavor
786: number of massless Dirac fermions, thus irrelevant in the large
787: flavor limit.\cite{Hermele} Following the similar strategy with
788: this case, the present author has shown that the scaling dimension
789: of the instanton operator is proportional to the fermion
790: conductivity analogous to the flavor number of Dirac fermions,
791: thus irrelevant in the large conductivity limit.\cite{Kim_SL_FS}
792: This seems to be natural because the fermion conductivity is
793: associated with screening of gauge interactions.
794:
795:
796:
797: In the heavy-fermion spin liquid phase the fermion conductivity
798: may not be sufficiently large because the strong hybridization
799: makes the hybridized band flat. Intuitively, the existence of the
800: heavy-fermion spin liquid phase is not clear because the strong
801: hybridization coupling is necessary for its existence, but such
802: hybridization prohibits this phase from being stable against gauge
803: fluctuations. Since its stability depends on the fermion
804: conductivity,\cite{Kim_SL_FS,Nagaosa} to determine its presence
805: with gauge interactions is beyond the scope of the present study.
806: However, its existence is allowed in principle.
807:
808:
809:
810: The stability of this phase against antiferromagnetic ordering is
811: confirmed since the Fermi-surface nesting does not exist in the
812: heavy-fermion band. This not only justifies our neglect of spin
813: fluctuations, but also propose new mechanism for the existence of
814: the spin liquid phase.
815:
816:
817:
818: \subsection{A variational wave function}
819:
820:
821:
822: The above discussion motivates us to find the heavy-fermion spin
823: liquid phase from the microscopic model itself, here the finite-U
824: ALM. In this respect we propose a variational wave function for
825: its existence. Since we start from the conduction-valance
826: hybridized state, the following wave function is naturally
827: proposed \bqa && |\Phi\rangle =
828: \Pi_{k\sigma}(c_{k\sigma}^{\dagger} +
829: a(k)d_{k\sigma}^{\dagger})|0\rangle , \eqa where $|0\rangle$ is
830: the vacuum state and $a(k)$ is the variational parameter given by
831: $a(k) = V/[(\epsilon_{k}^{c} - \mu - \epsilon_{d})/2 +
832: \sqrt{(\epsilon_{k}^{c} - \mu - \epsilon_{d})^{2}/4 + V^{2}}]$ in
833: $U = 0$. Turning on on-site interactions to localized electrons,
834: the variational ground state is proposed to
835: be\cite{Valance_HF_GPGW} \bqa |\Psi> && = P_{PG}|\Phi\rangle \nn
836: && = \Pi_{i}(1 - \kappa n_{i\uparrow}^{d}n_{i\downarrow}^{d})
837: \Pi_{k\sigma}(c_{k\sigma}^{\dagger} +
838: a(k)d_{k\sigma}^{\dagger})|0> , \eqa where $P_{PG} = \Pi_{i}(1 -
839: \kappa n_{i\uparrow}^{d}n_{i\downarrow}^{d})$ is the partial
840: Gutzwiller projector, suppressing charge-fluctuation effects, with
841: an interaction-dependent $\kappa$ determined self-consistently as
842: a function of $U$.\cite{Gassamer_SC}
843:
844:
845:
846: Since this ground-state wave function captures both relevant
847: interactions, $V$ and $U$ appropriately, physics included in Eq.
848: (15) is expected to be essentially imposed in the slave-rotor
849: effective gauge theory [Eq. (8)]. When gauge fluctuations are
850: ignored as the saddle-point approximation in Eq. (8), the
851: $|\Phi\rangle$ state is exactly recovered from the fermion
852: Lagrangian $L_{F} = L_{\psi} + L_{\eta} + L_{V}$ because $L_{F}$
853: gives rise to the same $a(k)$ in $|\Phi\rangle$. Note that the
854: gauge transformation $c_{i\sigma} =
855: e^{-i\theta_{i}}\psi_{i\sigma}$ is an essential procedure for
856: obtaining this strong hybridized dynamics. On the other hand, the
857: partial Gutzwiller projection operator $P_{PG}$ is simulated in
858: the slave-rotor representation, $L_{\theta}$ taking into account
859: on-site interactions of localized electrons appropriately. This
860: approach is parallel to that in the doped Mott insulator problem,
861: where the Gutzwiller projected BCS wave function, i.e., the RVB
862: state had been proposed,\cite{RVB} and such a variational ground
863: state was simulated in the context of the gauge
864: theory.\cite{RVB_SL} It will be an interesting project to check
865: whether this ground state wave function can result in both the
866: fractionalized Fermi liquid and heavy-fermion spin liquid.
867:
868:
869:
870: \subsection{Summary}
871:
872:
873:
874: In this paper we have studied the finite-U ALM in the strong
875: hybridization limit. The conventional treatment, integrating out
876: conduction electrons to obtain an effective Lagrangian in terms of
877: localized fermions and collective density-fluctuation bosons,
878: describes the heavy-fermion quantum transition from the
879: heavy-fermion Fermi liquid with a large Fermi surface to the
880: fractionalized Fermi liquid with a small Fermi surface, where the
881: abrupt volume change of the Fermi surface is involved. On the
882: other hand, the strong hybridization approach shows the
883: heavy-fermion spin liquid to heavy-fermion Fermi liquid
884: transition, where this type of Mott transition differs from the
885: orbital-selective one via Kondo breakdown.
886:
887:
888:
889: We propose a schematic phase diagram of the finite-U ALM in Fig.
890: 1, where HF-FL, HF-SL, and F-SL represent the heavy-fermion Fermi
891: liquid, heavy-fermion spin liquid, and fractionalized Fermi
892: liquid, respectively. It is important to notice that this model
893: has two independent parameters scaled by the conduction bandwidth,
894: $(V/D, U/D)$. In the small $U/D$ limit the hybridization coupling
895: is relevant to form a hybridized band, thus the heavy-fermion
896: Fermi liquid arises. In the large $U/D$ limit the Kondo breakdown
897: transition has been shown to occur in the KLM or infinite-U ALM,
898: thus there is a critical $V/D$ separating the fractionalized Fermi
899: liquid ($\langle{c_{\sigma}^{\dagger}\eta_{\sigma}}\rangle = 0$)
900: from the heavy-fermion Fermi liquid
901: ($\langle{c_{\sigma}^{\dagger}\eta_{\sigma}}\rangle \not= 0$). On
902: the other hand, in the large $V/D$ limit the heavy-fermion band is
903: formed first
904: ($\langle{\psi_{\sigma}^{\dagger}\eta_{\sigma}}\rangle \not= 0$),
905: and increasing $U/D$ (marked by the arrow line) is expected to
906: cause the Mott transition ($\langle{e^{-i\theta}}\rangle = 0$
907: $\rightarrow$ $\langle{c_{\sigma}^{\dagger}\eta_{\sigma}}\rangle
908: \approx \langle{e^{-i\theta}}\rangle
909: \langle{\psi_{\sigma}^{\dagger}\eta_{\sigma}}\rangle = 0$) in this
910: heavy-fermion band.
911:
912:
913:
914: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
915: \begin{figure}
916: \includegraphics[width=8cm]{ALM_Phase_Diagram.eps}
917: \caption{Schematic phase diagram of the finite-U Anderson lattice
918: model without antiferromagnetism at zero temperature: HF-FL
919: (heavy-fermion Fermi liquid), HF-SL (heavy-fermion spin liquid),
920: and F-SL (fractionalized Fermi liquid)} \label{Fig. 1}
921: \end{figure}
922: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
923:
924:
925:
926: We proposed the variational ground-state wave function for the
927: finite-U ALM since the stability of the heavy-fermion spin liquid
928: state against compact gauge fluctuations cannot be fully confirmed
929: in the effective field theory approach, thus it is necessary to
930: check its existence from the microscopic model. It will be
931: interesting to observe such a spin liquid state in the strong
932: hybridization and large repulsion limits, where the heavy-fermion
933: band still exists, but this corresponds to an insulator owing to
934: charge gap.
935:
936:
937:
938: We close this paper discussing the possibility of
939: valance-fluctuation-induced superconductivity.\cite{Miyake}
940: Recently, we have generalized the U(1) slave-rotor formulation of
941: the Hubbard model into the SU(2) one, allowing not only local
942: density fluctuations but also pairing excitations.\cite{Kim_Rotor}
943: Both collective charge fluctuations form an SU(2) slave-rotor
944: matrix field, where its off-diagonal components are associated
945: with superconductivity. We believe that the SU(2) slave-rotor
946: decomposition is also available to the finite-U ALM. This
947: superconducting mechanism may explain physics in $PuCoGa_5$, where
948: superconductivity occurs with HF physics at the same
949: time.\cite{Coleman_Review,PuCoGa5} This interesting possibility is
950: under investigation.
951:
952:
953:
954: \begin{thebibliography}{9}
955: \bibitem{RVB_SL} P. A. Lee, N. Nagaosa, and X.-G. Wen, Rev. Mod.
956: Phys. {\bf 78}, 17 (2006), and references therein.
957: \bibitem{Wen_Quantum_Order} X.-G. Wen, Phys. Rev. B {\bf 65}, 165113
958: (2002). See more introductory-level articles in
959: http://dao.mit.edu/~wen/.
960: \bibitem{Spin_liquids} R. Coldea, D. A. Tennant, A. M. Tsvelik,
961: and Z. Tylczynski, Phys. Rev. Lett. {\bf 86}, 1335 (2001); R.
962: Coldea, D. A. Tennant, and Z. Tylczynski, Phys. Rev. B {\bf 68},
963: 134424 (2003); Y. Shimizu, K. Miyagawa, K. Kanoda, M. Maesato, and
964: G. Saito Phys. Rev. Lett. {\bf 91}, 107001 (2003); Y. Kurosaki, Y.
965: Shimizu, K. Miyagawa, K. Kanoda, and G. Saito Phys. Rev. Lett.
966: {\bf 95}, 177001 (2005); A. Kawamoto, Y. Honma, and K. I. Kumagai,
967: Phys. Rev. B {\bf 70}, 060510 (2004); J. S. Helton, K. Matan, M.
968: P. Shores, E. A. Nytko, B. M. Bartlett, Y. Yoshida, Y. Takano, A.
969: Suslov, Y. Qiu, J.-H. Chung, D. G. Nocera, and Y. S. Lee, Phys.
970: Rev. Lett. {\bf 98}, 107204 (2007); P. Mendels, F. Bert, M. A. de
971: Vries, A. Olariu, A. Harrison, F. Duc, J. C. Trombe, J. S. Lord,
972: A. Amato, and C. Baines, Phys. Rev. Lett. {\bf 98}, 077204 (2007);
973: Oren Ofer, Amit Keren, Emily A. Nytko, Matthew P. Shores, Bart M.
974: Bartlett, Daniel G. Nocera, Chris Baines, Alex Amato,
975: cond-mat/0610540.
976: \bibitem{Senthil_Kondo} T. Senthil, Subir Sachdev, and Matthias
977: Vojta, Phys. Rev. Lett. {\bf 90}, 216403 (2003); T. Senthil, M.
978: Vojta, and S. Sachdev, Phys. Rev. B {\bf 69}, 035111 (2004).
979: \bibitem{Kim_Rotor} Ki-Seok Kim, Phys. Rev. Lett. {\bf 97},
980: 136402 (2006); cond-mat/0609415, to be published in Phys. Rev. B;
981: Phys. Rev. B {\bf 74}, 115122 (2006); Phys. Rev. B {\bf 73},
982: 235115 (2006).
983: \bibitem{Kim_Kondo_SB} Ki-Seok Kim, Phys. Rev. B {\bf 71}, 205101
984: (2005).
985: \bibitem{Pepin} C. Pepin, cond-mat/0610846.
986: \bibitem{Review_theory} P. Coleman, C. Pepin, Q. Si, and R.
987: Ramazashvili, J. Phys. Condens. Matter {\bf 13}, R723 (2001).
988: \bibitem{Coleman_Review} P. Coleman, cond-mat/0612006, and
989: references therein.
990: \bibitem{HMM} A similar situation also happens in the magnetic
991: transition of itinerant electrons, described by the
992: Hertz-Moriya-Millis theory [T. Moriya and J. Kawabata, J. Phys.
993: Soc. Jpn. {\bf 34}, 639 (1973); T. Moriya and J. Kawabata, J.
994: Phys. Soc. Jpn. {\bf 35}, 669 (1973); J. A. Hertz, Phys. Rev. B
995: {\bf 14}, 1165 (1976); A. J. Millis, Phys. Rev. B {\bf 48}, 7183
996: (1993)]. Since feedback effects to the electron degrees of freedom
997: by order parameter fluctuations are not taken into account, this
998: Landau-Ginzburg-Wilson-type approach should be regarded as a weak
999: coupling one, not justified in the strong coupling
1000: limit.\cite{Kim_Kim}
1001: \bibitem{FG} In the single impurity Anderson model Florens
1002: and Georges integrate out conduction electrons, and obtain an
1003: effective Lagrangian in terms of localized spinons $\eta_{\sigma}$
1004: and holons $e^{i\theta}$ at the impurity site using the
1005: Hubbard-Stratonovich transformation [S. Florens and A. Georges,
1006: Phys. Rev. B {\bf 66}, 165111 (2002)]. They derive self-consistent
1007: equations for spinon and holon self-energies, which coincide with
1008: the non-crossing approximation framework of the slave-boson
1009: theory.
1010: \bibitem{YbRh2Si2} S. Paschen, T. Luhmann, S. Wirth, P.
1011: Gegenwart, O. Trovarelli, C. Geibel, F. Steglich, P. Coleman, Q.
1012: Si, Nature {\bf 432}, 881 (2004); P. Gegenwart, T. Westerkamp, C.
1013: Krellner, Y. Tokiwa, S. Paschen, C. Geibel, F. Steglich, E.
1014: Abrahams, and Q. Si, Science {\bf 315}, 969 (2007).
1015: \bibitem{LGW_Why} One cautious person may claim that the
1016: slave-particle framework is not appropriate for the
1017: Landau-Ginzburg-Wilson forbidden quantum transition in the
1018: heavy-fermion context since no local gauge-invariant Kondo order
1019: parameter exists, and no symmetry is broken in the Fermi-liquid
1020: ground state while the Fermi liquid description in the
1021: slave-particle context is in the condensation of slave-bosons,
1022: thus breaking the internal U(1) gauge symmetry inevitably. This
1023: statement is only partially correct. This situation reminds us of
1024: the characterization of superconductivity. Usually, we interpret
1025: superconductivity as the condensed phase of Higgs bosons (Cooper
1026: pairs). This identification necessarily breaks the U(1) gauge
1027: symmetry, thus cannot be correct in a rigorous manner. However, we
1028: should admit that this characterization is useful. The way to
1029: understand this identification is to break the gauge symmetry
1030: explicitly, fixing one gauge. Since the gauge symmetry is
1031: explicitly broken via gauge fixing, one can define the local order
1032: parameter such as Cooper pair bosons. This is a physically
1033: appealing way to understand such a boson-condensed phase in a
1034: simple mean-field manner.
1035: \bibitem{Quantum_disordered_SC} T. Senthil and M. P. A. Fisher,
1036: Phys. Rev. B {\bf 62}, 7850 (2000); M. Franz and Z. Tesanovic,
1037: Phys. Rev. Lett. {\bf 87}, 257003 (2001). The coupling term of
1038: $|\Delta|e^{-i\phi}c_{\uparrow}c_{\downarrow}$ between Cooper
1039: pairs and electrons in the context of $d-wave$ superconductivity
1040: plays the same role as the hybridization coupling term of
1041: $Ve^{-i\theta}c_{\sigma}^{\dagger}\eta_{\sigma}$ between
1042: collective density-fluctuation modes and conduction and localized
1043: fermions in the ALM. Here, $|\Delta|$ and $\phi$ are the amplitude
1044: and phase of Cooper pair fields. To solve this coupling term,
1045: several kinds of gauge transformations are introduced. In these
1046: decoupling schemes strong phase fluctuations of Cooper pairs,
1047: arising from the phase-density uncertainty, screen out charge
1048: degrees of freedom of electrons, causing electrically neutral but
1049: spinful electrons called "spinons". As a result, the phase factor
1050: disappears in the Cooper pair-electron coupling term when it is
1051: rewritten in terms of spinons. Instead, this coupling effect
1052: appears as current-current interactions of neutral spinons and
1053: phase fields of Cooper pairs in the kinetic term of electrons. In
1054: this respect the present gauge transformation may be regarded to
1055: follow the spirit of the quantum disordered superconductor.
1056: \bibitem{Rotor_model} Consider the boson-Hubbard model $H =
1057: - t \sum_{\ij}(b_{i}^{\dagger}b_{j} + H.c.) +
1058: \frac{U}{2}\sum_{i}n_{i}^{2}$ with $n_{i} = b_{i}^{\dagger}b_{i}$.
1059: Using $b_{i} = \sqrt{n}e^{i\theta_{i}}$ and integrating out the
1060: $n$ field, one can obtain the Lagrangian of this Hamiltonian, $L =
1061: \frac{1}{2U}\sum_{i}(\partial_{\tau}\theta_{i})^{2} -
1062: 2tn\sum_{\ij}\cos(\theta_{i} - \theta_{j})$, the same as
1063: $L_{\theta}$ of Eq. (8) in the mean-field approximation. It is
1064: clear that this boson Hubbard model shows the QPT from the Mott
1065: insulating phase to the superfluid state at $D/U \sim 1$.
1066: \bibitem{Two_scales} S. Burdin, A. Georges, and D. R. Grempel,
1067: Phys. Rev. Lett. {\bf 85}, 1048 (2000).
1068: \bibitem{TK_TFL} One need not be surprised at $T_{K} \sim V$
1069: (instead of the exponentially small $T_{K}$) in the strong
1070: hybridization limit because this is a usual result of the
1071: mean-field treatment, considering the antiferromagnetic case
1072: (Hartree-Fock calculation) where the magnetization order parameter
1073: is exponentially small in the small-U limit of the Hubbard model
1074: while it is proportional to Hubbard-U in the large-U limit. In the
1075: strong hybridization approach the coherence temperature $T_{FL}$
1076: would be exponentially small, corresponding to the
1077: boson-condensation temperature. On the other hand, in the
1078: conventional heavy-fermion metal-to-metal transition discussed in
1079: section III, the Kondo temperature $T_{K}$ will be exponentially
1080: small since this corresponds to the weak hybridization approach,
1081: where $V_{eff} = |\langle e^{-i\theta}\rangle| V$ is an effective
1082: hybridization coupling constant with the boson-condensation
1083: amplitude $|\langle e^{-i\theta}\rangle|$. Actually, in the
1084: single-impurity Anderson model this was indeed found in the
1085: non-crossing approximation scheme.\cite{FG}
1086: \bibitem{Kim_Kim} Ki-Seok Kim and Mun Dae Kim, Phys. Rev. B {\bf
1087: 75}, 035117 (2007).
1088: \bibitem{Sigma} P. A. Lee and N. Nagaosa, Phys. Rev. B {\bf 46},
1089: 5621 (1992); L. B. Ioffe and G. Kotliar, Phys. Rev. B {\bf 42},
1090: 10348 (1990).
1091: \bibitem{IL_sigma} L. B. Ioffe and A. I. Larkin, Phys. Rev. B {\bf
1092: 39}, 8988 (1989). Expanding the effective action of Eq. (8) to the
1093: gaussian order for the U(1) gauge field $a_{\mu}$ with $\mu = x,
1094: y$, the partition function is given in a highly schematic form
1095: \bqa && Z \approx \int{D\psi_{\sigma}D\eta_{\sigma}D\theta
1096: Da}\exp\Bigl[- \bigl\{S_{\psi}^{(0)} + S_{\eta}^{(0)} +
1097: S_{\theta}^{(0)} \bigr\} \nn && - \Bigl(\frac{\delta
1098: S_{\psi}}{\delta{a}}\Bigl|_{0} + \frac{\delta
1099: S_{\theta}}{\delta{a}}\Bigl|_{0} \Bigr)a -
1100: \frac{1}{2}\frac{\delta^{2}S_{\psi}}{\delta{a}^{2}}\Bigl|_{0}{a}^{2}
1101: \nn && -
1102: \frac{1}{2}\Bigl(\frac{\delta^{2}S_{\theta}}{\delta{a}^{2}}\Bigl|_{0}{a}^{2}
1103: + 2\frac{\delta^{2}S_{\theta}}{\delta{a}\delta{A}}\Bigl|_{0}{a}{A}
1104: + \frac{\delta^{2}S_{\theta}}{\delta{A}^{2}}\Bigl|_{0}{A}^{2}
1105: \Bigr) \Bigr] \nn && = Z_{0}\exp\Bigl[ -
1106: \frac{1}{2}\frac{\delta^{2}S_{\theta}}{\delta{A}^{2}}\Bigl|_{0}{A}^{2}
1107: +
1108: \frac{1}{2}\frac{\Bigl(\frac{\delta^{2}S_{\theta}}{\delta{a}\delta{A}}\Bigl|_{0}\Bigr)^{2}}
1109: {\frac{\delta^{2}S_{\theta}}{\delta{a}^{2}}\Bigl|_{0}+\frac{\delta^{2}S_{\psi}}{\delta{a}^{2}}\Bigl|_{0}}A^{2}\Bigr]
1110: \nonumber \eqa with $Z_{0} =
1111: \int{D\psi_{\sigma}D\eta_{\sigma}D\theta} e^{- (S_{\psi}^{(0)} +
1112: S_{\eta}^{(0)} + S_{\theta}^{(0)})}$, where the super- or sub-
1113: script $(0)$ means $a = 0$ in the saddle-point approximation.
1114: Differentiating the above effective action by an electromagnetic
1115: field $A_{\mu}$ twice, one can obtain the Ioffe-Larkin
1116: conductivity expression, $\sigma_{tot} =
1117: \sigma_{\theta}\sigma_{\psi}/(\sigma_{\theta} + \sigma_{\psi})$,
1118: where $\sigma_{\theta} =
1119: \frac{\delta^{2}S_{\theta}}{\delta{a}\delta{A}}\Bigl|_{0} =
1120: \frac{\delta^{2}S_{\theta}}{\delta{a}^{2}}\Bigl|_{0} =
1121: \frac{\delta^{2}S_{\theta}}{\delta{A}^{2}}\Bigl|_{0}$ and
1122: $\sigma_{\psi} =
1123: \frac{\delta^{2}S_{\psi}}{\delta{a}^{2}}\Bigl|_{0}$. Here, the
1124: saddle-point condition \bqa && \langle \frac{\delta
1125: S_{\psi}}{\delta{a}}\Bigl|_{0} + \frac{\delta
1126: S_{\theta}}{\delta{a}}\Bigl|_{0} \rangle \equiv \langle J_{\psi} +
1127: J_{\theta} \rangle = 0 \nonumber \eqa is imposed in the
1128: Ioffe-Larkin formulation.
1129: \bibitem{IR_FP} S. Chakravarty et al., Phys. Rev. Lett. {\bf 74},
1130: 1423 (1995); J. Gan and E. Wong, Phys. Rev. Lett {\bf 71}, 4226
1131: (1993); C. Nayak and F. Wilczek, Nucl. Phys. B {\bf 417}, 359
1132: (1994); J. Polchinski, Nucl. Phys. B {\bf 422}, 617 (1994).
1133: \bibitem{Kim_SL_FS} Ki-Seok Kim, Phys. Rev. B {\bf 72}, 245106
1134: (2005).
1135: \bibitem{Hermele} M. Hermele, T. Senthil, M. P. A. Fisher,
1136: P. A. Lee, N. Nagaosa, and X.-G. Wen, Phys. Rev. B {\bf 70},
1137: 214437 (2004).
1138: \bibitem{Nagaosa} N. Nagaosa, Phys. Rev. Lett. {\bf 71}, 4210
1139: (1993).
1140: \bibitem{Valance_HF_GPGW} P. Fazekas, Lecture Notes on Electron
1141: Correlation and Magnetism, (World Scientific, Singapore, 1999).
1142: \bibitem{Gassamer_SC} B. A. Bernevig, R. B. Laughlin, and D. I.
1143: Santiago, Phys. Rev. Lett. {\bf 91}, 147003 (2003); Bogdan A.
1144: Bernevig, George Chapline, Robert B. Laughlin, Zaira Nazario, and
1145: David I. Santiago, cond-mat/0312573; F. C. Zhang, Phys. Rev. Lett.
1146: {\bf 90}, 207002 (2003).
1147: \bibitem{RVB} P. W. Anderson, P. A. Lee, M. Randeria, T. M. Rice,
1148: N. Trivedi, F. C. Zhang, J Phys. Condens. Matter {\bf 16} (2004)
1149: R755-R769; A. Paramekanti, M. Randeria, and N. Trivedi, Phys. Rev.
1150: Lett. {\bf 87}, 217002 (2001); M. Randeria, A. Paramekanti, and N.
1151: Trivedi, Phys. Rev. B {\bf 69}, 144509 (2004); A. Paramekanti, M.
1152: Randeria, and N. Trivedi, Phys. Rev. B {\bf 70}, 054504 (2004).
1153: \bibitem{Miyake} S. Watanabe, M. Imada, and K. Miyake, J. Phys.
1154: Soc. Japan 75, 043710 (2006).
1155: \bibitem{PuCoGa5} J. L. Sarrao, L. A. Morales, J. D. Thompson, B.
1156: L. Scoot, G. R. Stewart, F. Wastin, J. Rebizant, P. Boulet, E.
1157: Colineau, and G. H. Lander, Nature {\bf 420}, 297 (2002).
1158: \end{thebibliography}
1159:
1160:
1161:
1162:
1163:
1164:
1165: \end{document}
1166: