1: \documentclass[aps,prl,twocolumn]{revtex4}
2: \usepackage{graphicx,epsf}
3:
4: \begin{document}
5:
6: \title {
7: Spontaneous interlayer coherence in bilayer Kondo systems
8: }
9: \author{T. Senthil}
10: \affiliation{Department of Physics, Massachusetts Institute of
11: Technology, Cambridge MA 02139}
12: \author{Matthias Vojta}
13: \affiliation{\mbox{Institut f\"ur Theorie der Kondensierten Materie,
14: Universit\"at Karlsruhe, 76128 Karlsruhe, Germany}}
15:
16: \date{March 8, 2005}
17:
18:
19: \begin{abstract}
20: Bilayer Kondo systems present interesting models to illustrate the competition between the
21: Kondo effect and intermoment exchange.
22: Such bilayers can exhibit two sharply distinct Fermi liquid phases which are distinguished
23: by whether or not the local moments participate in the Fermi sea.
24: We study these phases and the evolution from one to the other upon changing Kondo coupling.
25: We argue that an ordered state with spontaneous interlayer phase coherence
26: generically intervenes between the two Fermi liquids.
27: Such a condensate phase breaks a U(1) symmetry and is bounded by a finite-temperature
28: Kosterlitz-Thouless transition.
29: Based on general arguments and mean-field calculations
30: we investigate the phase diagram and associated quantum phase transitions.
31: \end{abstract}
32: \pacs{}
33:
34: \maketitle
35:
36:
37: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
38:
39: Landau's Fermi liquid theory provides the basis for our understanding of most conventional metals.
40: A particularly interesting application of Fermi liquid theory is to
41: intermetallic compounds containing localized spin moments on $d$ or $f$
42: orbitals and additional bands of conduction electrons.
43: These systems are conveniently modelled by the Kondo lattice
44: Hamiltonian, with exchange interactions between the
45: local moments and the conduction electrons, and possibly
46: additional exchange couplings between the local moments
47: themselves \cite{hewson}.
48: These display metallic ``heavy'' Fermi liquid phases
49: (albeit with large quasiparticle effective masses and other renormalizations)
50: in addition to phases with magnetic or superconducting long-range order \cite{doniach}.
51: The quantum phase transitions between the various phases are currently subject of
52: intense research activities.
53: For instance the transition between the heavy Fermi liquid and a magnetic metal exhibits a
54: number of interesting non-Fermi liquid phenomena
55: which have not yet found a consistent theoretical explanation \cite{piers}.
56:
57: Of particular interest in heavy electron physics is the issue of the topology of the
58: Fermi surface \cite{oshi}, and how it evolves across various phase boundaries.
59: It is believed by some workers in the field that this issue is at the heart of
60: the non-Fermi liquid physics observed near heavy electron critical points \cite{piers,edmft,ffl}.
61: Our understanding of such matters is rather primitive --
62: clearly it would help to build more intuition about such phenomena.
63: In this paper we consider a paradigmatic model that illustrates some of these issues
64: and the associated difficulties.
65:
66: We consider bilayer Kondo systems, consisting of pairs of identical two-dimensional layers with local
67: moments and conduction electrons.
68: We assume that the conduction electrons cannot hop between the two layers.
69: As argued below, such bilayer systems allow
70: for two different paramagnetic Fermi liquid phases.
71: These two phases are sharply distinct: the distinction is in the topology
72: of the Fermi surface.
73: In one case the Fermi surface may loosely be dubbed ``large'' -- physically this happens
74: when the local moments participate in the Fermi surface.
75: In the other case the local moments are not part of the Fermi sea and the Fermi surface
76: may loosely be dubbed ``small''.
77: We emphasize that Luttinger's theorem \cite{oshi} is satisfied
78: in both phases -- however the Fermi surface topology is different,
79: implying that the two phases cannot be smoothly connected to each other.
80: We argue below that a novel ordered phase generically intervenes in between the two phases.
81: This ordered state involves spontaneous development of interlayer coherence despite
82: the absence of any direct single particle hopping.
83: The interlayer coherence may either be in a particle-hole (i.e exciton) condensate or
84: in a particle-particle condensate. The former is analogous to states much discussed in bilayer
85: quantum Hall structures \cite{qhth,qhexp} while the latter is a bilayer superconductor with Cooper pairs shared
86: between the two layers.
87: At finite temperatures above the ordered phase we expect signatures of critical
88: behavior associated to strong Fermi surface fluctuations.
89:
90:
91: To be specific, consider the Hamiltonian:
92: \begin{eqnarray}
93: H &=& -\!\!\sum_{\langle jj'\rangle\sigma\alpha} \!\!
94: t_{jj'} c_{j\sigma\alpha}^{\dagger} c_{j' \sigma\alpha} +
95: \frac{J_K}{2} \sum_{j\sigma\sigma'\alpha} \! \vec{S}_{j\alpha} \!\cdot c^\dagger_{j \sigma\alpha}
96: \vec{\tau}_{\sigma\sigma'} c_{j\sigma'\alpha} \nonumber \\
97: &&+ \sum_{\langle jj'\rangle\alpha} I_{jj'} \vec{S}_{j\alpha} \cdot \vec{S}_{j'\alpha}
98: + \sum_{j} J_\perp \vec{S}_{j1} \cdot \vec{S}_{j2} .
99: \label{h}
100: \end{eqnarray}
101: The lattice sites are labelled with $j$ in each layer, and $\alpha=1,2$ is
102: the layer index. The local moments $\vec{S}_{j\alpha}$ are $S=1/2$ spins, and the
103: conduction electrons $c_{j \sigma\alpha}$ ($\sigma = \uparrow
104: \downarrow$) hop on the sites $j$, $j'$ of some regular lattice in
105: $d$ spatial dimensions ($d=2$ for most what follows)
106: with amplitude $t_{jj'}$.
107: $J_K >0$ are the Kondo exchanges, $I_{jj'}$ denote explicit short-range
108: Heisenberg interactions within each plane of local moments, and
109: $J_\perp$ is the interlayer exchange interaction.
110: A chemical potential for the $c_\sigma$
111: fermions which fixes their density at $\rho_c$ per unit cell of
112: the ground state is implied.
113:
114: The absence of any direct interlayer hopping implies that the number of electrons in each layer
115: is separately conserved. The corresponding symmetry plays an important role in our analysis.
116: In contrast, only the {\em total} spin of both layers is conserved.
117: %Further, the Hamiltonian is invariant under interchange of the two layers.
118:
119:
120: \begin{figure}
121: \epsfxsize=2.7in
122: \centerline{\epsffile{pd_sch.eps}}
123: \caption{
124: Schematic phase diagram of the bilayer Kondo lattice model, see text for details.
125: At $T\!=\!0$ the interlayer coherence phase (COH) breaks a U(1) symmetry; it is
126: bounded by a finite-temperature Kosterlitz-Thouless transition.
127: The dashed lines denote crossovers below which the local moments
128: get screened either by the Kondo effect or by interlayer singlet
129: formation.
130: (The in-plane exchange $I$ is assumed to be zero, and magnetic ordering
131: tendencies due to RKKY interaction, which occur for small $J_K/t$, are ignored.)
132: \vspace*{-10pt}
133: }
134: \label{fig:pd1}
135: \end{figure}
136:
137: For situations where the ground state of the system is a Fermi liquid,
138: it can be characterized by its Fermi volume, i.e., the volume in momentum
139: space enclosed by the Fermi surface, which is defined by the location of
140: poles in the conduction electron Green's function at the Fermi level.
141: The Fermi liquid phase of a single-layer Kondo system is known to
142: have a Fermi volume, ${\cal V}_{FL}$, counting all electrons in the unit cell, i.e.,
143: both conduction electrons and local moments,
144: as determined by the Luttinger theorem \cite{oshi}:
145: \begin{equation}
146: {\cal V}_{FL} = {\cal K}_d \, \rho_T\mbox{(mod 2)}. \label{vfl}
147: \end{equation}
148: Here ${\cal K}_d = (2 \pi)^d /(2 v_0)$ is a phase space factor,
149: $v_0$ is the volume of the unit cell of the ground state, $\rho_T
150: = n_{\ell} + \rho_c$ is the total density of electrons per volume
151: $v_0$, and $n_{\ell}$ (an integer) is the density of local moments
152: per volume $v_0$.
153:
154: For Fermi liquid phases in the present bilayer system, the Fermi volume can be defined separately
155: for both layers as the number of electrons in each layer is separately conserved.
156: But what about Luttinger's theorem?
157: By an extension of the argument in Ref. \onlinecite{oshi}
158: it is easy to see that only the {\em total} volume of the
159: Fermi surface of both layers together is constrained. Specifically we have
160: \begin{equation}
161: {\cal V}^{(1)}_{FL} + {\cal V}^{(2)}_{FL} = {\cal K}_d \, \left(\rho^{(1)}_T + \rho^{(2)}_T\right)~~ \mbox{(mod 2)}.
162: \label{vflbl}
163: \end{equation}
164: Here the superscript $1,2$ is a layer index. Thus individual Fermi volumes of the two layers (though well-defined)
165: are not necessarily required to match the total electron count in their layer.
166:
167:
168: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
169:
170: {\it Phase diagram.}
171: %
172: Let us start with discussing various limits in parameter space.
173: States without magnetic long-range order can be conveniently discussed keeping $I=0$.
174: We also assume the conduction band to be away from perfect particle-hole symmetry
175: to avoid non-generic nesting effects.
176:
177: For large Kondo coupling, $J_K \gg J_\perp$, each local moment $\vec S_{j\alpha}$
178: will be Kondo screened by the conduction electrons in layer $\alpha$.
179: Thus, a heavy Fermi liquid is formed separately in each layer, with a large
180: Fermi volume per layer -- we dub this FL-L phase.
181: The effect of a small non-zero $J_\perp$ is to generate spin correlations between the
182: two Fermi liquids. These will in general lead to innoccuous renormalizations of the
183: Fermi liquid parameters \cite{note1}.
184: For dominating interlayer exchange, $J_\perp \gg J_K$, the local moments form
185: interlayer singlets. For $J_K=0$ the conduction electrons
186: in each layer form a Fermi liquid with a small Fermi volume per layer (FL-S phase).
187: Finite $J_K$ can be treated in perturbation theory, leading to spin-spin couplings
188: of order $J_K^2/J_\perp$ between the two Fermi liquids, which again only lead to unimportant renormalizations
189: of Fermi liquid parameters \cite{burdin}.
190:
191: Note that either Fermi liquid satisfies the generalized Luttinger theorem in Eq. (\ref{vflbl}).
192: Nevertheless the topologies of the Fermi surfaces are quite different.
193: Indeed it is not possible to smoothly go from FL-S to FL-L without a phase transition.
194: To see this note that the interlayer exchange symmetry is preserved in both phases:
195: this symmetry guarantees that the volume (and shape)
196: of the Fermi surfaces in the two layers are equal.
197: As the total Fermi volume is quantized according to Eq. (\ref{vflbl}), the individual
198: Fermi volumes $ {\cal V}^{(1), (2)}_{FL}$ must also be quantized.
199: However they are quantized to two different values in the two phases. In FL-L,
200: \begin{equation}
201: {\cal V}^{(1)}_{FL} = {\cal V}^{(2)}_{FL} = {\cal K}_d (\rho_c + 1)
202: \end{equation}
203: while in FL-S
204: \begin{equation}
205: {\cal V}^{(1)}_{FL} = {\cal V}^{(2)}_{FL} = {\cal K}_d (\rho_c ) \,.
206: \end{equation}
207: Thus the two phases cannot be smoothly connected.
208:
209:
210: As function of $J_K/J_\perp$ the system evolves from FL-S to FL-L which involves
211: a change in the quantized value of the {\em Fermi surface volume per layer}: how does this happen?
212: Besides a direct first-order transition, we show below that intermediate ordered phases which break some symmetry are
213: very natural. An example is an intermediate phase with spontaneous
214: interlayer phase coherence (COH), the order parameter being
215: $\phi = \sum_{j\sigma} \langle c_{j\sigma1}^\dagger c_{j\sigma2}\rangle$.
216: This phase breaks the separate conservation of electron number
217: in the individual layers, while preserving total electron number conservation.
218: An alternate possibility is an interlayer superconductor which
219: breaks the symmetry of total electron number conservation
220: while preserving conservation of the {\em difference} of the electron number in the two layers.
221: In our bilayer system, it does not seem possible to have a direct second-order
222: ``Fermi-volume-changing'' transition between the two Fermi liquids.
223: Paranthetically, we note that finite exchange interaction $I$ can lead to
224: various phase with magnetic long-range order -- these will not be considered
225: further.
226:
227:
228: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
229:
230: \begin{figure}
231: \epsfxsize=2.7in
232: \centerline{\epsffile{phd2.eps}}
233: \caption{
234: Mean-field phase diagram, obtained from Eq. (\ref{mf1})
235: on a bilayer square lattice, with parameters
236: near-neighbor hopping $t=0.6$, Kondo coupling $J_K = 1$,
237: conduction band filling $\rho_c = 0.3$, and
238: a unit cell of one conduction electron site per layer.
239: Thin (thick) lines denote second (first) order transitions.
240: Addtional transitions within the COH phase where a Fermi surface sheet
241: (dis)appears (see text and Fig.~\ref{fig:fsevol}) are not shown.
242: Only the phase boundaries of the COH phase will survive upon
243: inclusion of fluctuations around the saddle point.
244: \vspace*{-10pt}
245: }
246: \label{fig:mfpd}
247: \end{figure}
248:
249: {\it Mean-field theory.}
250: %
251: We use a fermionic auxiliary-particle representation of the local
252: moments:
253: \begin{equation}\label{sf}
254: \vec S_{j\alpha} = \frac{1}{2} \sum_{\sigma\sigma'}
255: f^\dagger_{j \sigma\alpha}
256: \vec{\tau}_{\sigma\sigma'} f_{j\sigma'\alpha}
257: \end{equation}
258: where $f_{j\sigma\alpha}$ describes a spinful fermion destruction
259: operator at site $j$.
260: As usual, mean-field equations are obtained by decoupling the various
261: interaction terms, resulting in the mean-field Hamiltonian \cite{hewson}:
262: \begin{eqnarray}
263: H_{\rm mf} & = & \sum_{k\sigma\alpha} \epsilon_k c^{\dagger}_{k\sigma\alpha} c_{k\sigma\alpha} -
264: \sum_{j\sigma} \chi_j (f^{\dagger}_{j\sigma 1} f_{j\sigma 2} +
265: \mbox{
266: h.c.} )
267: \nonumber\\
268: + && \!\!\!\!
269: \sum_{j\sigma\alpha} \mu_{f\alpha} f^{\dagger}_{j\sigma\alpha}f_{j\sigma\alpha} -
270: \sum_{k\sigma\alpha} b_\alpha
271: (c^{\dagger}_{k\sigma\alpha}
272: f_{k\sigma\alpha} + \mbox{h.c.}) \,.
273: \label{mf1}
274: \end{eqnarray}
275: $\epsilon_k$ is the conduction band dispersion resulting from the $t_{jj'}$,
276: and we have assumed $I=0$.
277: The mean-field parameters are determined by the saddle-point equations
278: \begin{eqnarray}
279: 1 &=& \sum_\sigma \langle f^{\dagger}_{r\sigma\alpha} f_{r\sigma\alpha} \rangle \,,\\
280: 2 b_\alpha & = & J_K \sum_\sigma \langle c^{\dagger}_{j\sigma\alpha} f_{j\sigma\alpha} \rangle\,, \\
281: 2 \chi_j & = & J_\perp \sum_\sigma \langle f^{\dagger}_{j\sigma 1} f_{j\sigma 2} \rangle \,.
282: \end{eqnarray}
283: The above mean-field theory can be justified by generalizing the spin
284: symmetry from SU(2) to SU($N$) and taking the limit $N\to\infty$ \cite{AffMar}.
285: (Very similar results are obtained using a Sp($N$) generalization \cite{tJ}.)
286: Low-energy fluctuations around the mean-field solution are phase
287: fluctuations of the mean-field order parameters, i.e., local U(1) gauge
288: fluctuations.
289:
290: The FL-L phase with Kondo screening is described by non-zero expectation values
291: $\langle c_{j\alpha}^\dagger f_{j\alpha}\rangle$ at the mean-field level.
292: The FL-S phase with interlayer singlet formation of the local moments
293: has non-zero $\langle f_{j1}^\dagger f_{j2}\rangle$.
294: Note that both quantities involve the auxiliary $f$ particles and
295: are thus {\em not} physical observables, i.e., they cannot sustain expectation values
296: once gauge fluctuations are included.
297:
298: Consider now a state where both $\langle f_{j1}^\dagger f_{j2}\rangle$ and
299: $\langle c_{j\alpha}^\dagger f_{j\alpha}\rangle$ are non-zero.
300: Then it follows that $\phi = \langle c_{j\sigma 1}^\dagger c_{j\sigma 2}\rangle \neq 0$ --
301: this gauge-invariant complex quantity {\em is} a physical observable, which breaks
302: a physical U(1) symmetry and characterizes a
303: phase with {\em spontaneous interlayer phase coherence}!
304: For the two-dimensional system under discussion, small finite temperature will
305: lead to quasi-long-range order, which disappears in a finite-temperature
306: Kosterlitz-Thouless transition. No other finite-temperature phase transitions
307: are expected, as there are no broken symmetries in both FL phases.
308: The resulting schematic phase diagram is in Fig.~\ref{fig:pd1}.
309: We have performed a fully self-consistent solution of the mean-field equations
310: for various sets of parameters,
311: and a sample mean-field phase diagram is in Fig.~\ref{fig:mfpd}.
312:
313:
314: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
315:
316: {\it Interlayer coherence phase.}
317: %
318: We now discuss a few properties of the interlayer coherence phase.
319: It is characterized by the presence of a particle-hole (i.e. excitonic)
320: condensate, with the expectation value $\phi = \langle c_{j\sigma1}^\dagger c_{j\sigma2}\rangle$
321: being non-zero, in the absence of explicit interlayer hopping.
322: Thus there will be a spontaneous bilayer splitting in the electronic
323: structure, i.e.,
324: the electrons in both layers will spontaneously form a bonding and an antibonding
325: band -- this feature should be observable in high-resolution photoemission
326: experiments.
327: The COH phase will have an electronic Fermi surface that satisfies the
328: generalized Luttinger theorem in Eq. (\ref{vflbl}).
329: However, this Fermi surface will be shared by electrons in the two layers --
330: in the sense that the quasiparticle states are admixtures of electrons from both
331: layers.
332: Thus the notion of two distinct quantized Fermi surfaces associated with either layer
333: is no longer meaningful.
334: The Fermi surface evolution from FL-L to FL-S is illustrated in Fig.~\ref{fig:fsevol}.
335: The broken U(1) symmetry implies a linearly dispersing collective
336: Goldstone mode with quantum numbers (total) charge 0 and spin 0.
337:
338:
339: \begin{figure}
340: \epsfxsize=3.5in
341: \centerline{\epsffile{fsevol1.eps}}
342: \caption{
343: Schematic evolution of the Fermi surfaces (in two-dimensional momentum space)
344: along the horizontal axis of Fig.~\ref{fig:pd1},
345: for a conduction band filling $\rho_c < 1/2$, and $n_{\ell}=1$.
346: In both the FL-L and FL-S phases there are two identical Fermi surfaces for both
347: layers, with volumes $1+\rho_c$ and $\rho_c$ respectively.
348: Upon entering the COH phase, this degeneracy is lifted, and a spontaneous
349: bilayer splitting develops. Within the COH phase there exist additional transitions
350: where one Fermi surface sheet (dis)appears; the total Fermi volume (mod 2)
351: is always given by $2 \rho_c$.
352: \vspace*{-10pt}
353: }
354: \label{fig:fsevol}
355: \end{figure}
356:
357: The present interlayer coherence phase is the zero-field analogue of the
358: corresponding phase in bilayer quantum Hall systems \cite{qhth,qhexp}.
359: In the quantum Hall context this type of order has also been dubbed
360: pseudospin ferromagnetism, where the pseudospin degree of freedom
361: refers to the layer index of a given electron.
362:
363: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
364:
365: {\it Fluctuation effects.}
366: %
367: As argued above, both Fermi liquid phases are stable w.r.t. fluctuations,
368: their main effect being the restoration of the U(1) gauge symmetry
369: which is broken at the mean-field level; this also implies the absence of any
370: finite-temperature transition above the FL phases.
371:
372: In the interlayer coherence phase amplitude fluctuations of the mean-field
373: parameters will be gapped. An action for the phase fluctuations can be
374: formally obtained by integrating out all fermions from the problem;
375: the result is easily seen to generate the phase mode of the
376: order parameter with a linear dispersion.
377:
378: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
379:
380: {\it Interlayer superconductivity.}
381: %
382: It is possible to imagine other instabilities near the FL-L -- FL-S transition.
383: One is the formation of coherence in the particle-particle sector, i.e.,
384: interlayer superconductivity (with a charge-2 condensate).
385: At the mean-field level, this can be captured by a mixed decoupling, namely
386: the $J_K$ term with a particle-hole field $b$
387: and the $J_\perp$ term with a particle-particle field $\chi$ (or vice versa).
388: Then, the phase with both $b$ and $\chi$ non-zero will have a non-vanishing
389: expectation value $\langle c_{j1} c_{j2}\rangle$, representing
390: interlayer superconducting pairing, i.e., Cooper pairs shared by the two layers,
391: with $s$-wave orbital symmetry.
392: %
393: We note that the inclusion of Coulomb repulsion in the model will
394: favor the interlayer coherence phase over the interlayer superconducting phase.
395:
396: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
397:
398: {\it Discussion.}
399: %
400: We have described the physics of bilayer Kondo lattice systems
401: which contain the competition between the Kondo effect and intermoment exchange.
402: Two distinct Fermi liquid phases can occur upon varying
403: the ratio of the Kondo interaction and the interlayer coupling between the local
404: moments.
405: These Fermi liquids may be loosely distinguished by whether or not the local moments participate
406: in the Fermi surface.
407: A more precise distinction is provided by the topology of their Fermi surfaces
408: as captured by the notion of the quantized value of the Fermi volume {\em per layer}.
409: We have argued that a phase with a form of exotic order,
410: namely spontaneous interlayer coherence in either the particle-hole or the particle-particle channel,
411: generically intervenes between the two Fermi liquids.
412: Experimental observation of the former phase may be possible in
413: angle-resolved photoemission where a spontaneous splitting
414: between the bands of the two layers
415: (or a splitting larger than that predicted by band theory)
416: will be observed.
417:
418: The finite-temperature properties in the transition region between the two phases
419: are particularly interesting.
420: In the region above the phase transition to the ordered phase the behavior is
421: that of a metal with an undecided Fermi volume per layer, i.e., the
422: Fermi surface will be strongly fluctuating.
423: We expect that the properties will be non-Fermi liquid like and possibly similar to that observed
424: in various ``strange metals'' such as near heavy electron magnetic critical points
425: or in the optimally doped cuprates.
426: Interestingly, signatures of fluctuating Fermi surfaces near quantum criticality have been
427: experimentally detected, e.g., in the heavy electron metal YbRh$_2$Si$_2$ \cite{ybrhsi,silke}
428: and in the bilayer ruthenate Sr$_3$Ru$_2$O$_7$ \cite{grigera,hooper}.
429: %
430: Developing a theory of strange metal states is a major challenge in condensed matter physics --
431: we hope that the bilayer systems discussed here can contribute in some way.
432:
433:
434: % DISCUSS MAGNETIC FIELD? FL-S HAS METAMAGNETISM!
435:
436:
437: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
438:
439: % \acknowledgments
440:
441: We thank S.~Sachdev for illuminating discussions.
442: This research is supported by the National Science Foundation
443: grant DMR-0308945 (T.S.) and by the DFG Center for Functional
444: Nanostructures Karlsruhe (M.V.). T.S. also acknowledges funding from the
445: NEC Corporation, the Alfred P. Sloan Foundation, and an award from
446: The Research Corporation.
447:
448: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
449:
450: \vspace*{-10pt}
451: \begin{thebibliography}{}
452: \vspace*{-10pt}
453:
454: \bibitem{hewson} A.~C.~Hewson, {\em The Kondo Problem to
455: Heavy Fermions}, (Cambridge University Press, Cambridge, 1993).
456:
457: \bibitem{doniach} S.~Doniach, Physica B \& C {\bf 91}, 231 (1977).
458:
459: \bibitem{piers} P.~Coleman, C.~P\'epin, Q.~Si, and
460: R.~Ramazashvili, J. Phys: Condens. Matt. {\bf 13}, 723 (2001).
461:
462: \bibitem{oshi} M. Oshikawa, Phys. Rev. Lett. {\bf 84}, 3370 (2000)
463:
464: \bibitem{edmft} Q.~Si, S.~Rabello, K.~Ingersent, and J.~L.~Smith, Nature
465: {\bf 413}, 804 (2001) and \prb {\bf 68}, 115103 (2003).
466:
467: \bibitem{ffl} T. Senthil, S. Sachdev, and M. Vojta, Phys. Rev.
468: Lett. {\bf 90}, 216403 (2003); T. Senthil, M. Vojta, and S.
469: Sachdev, Phys. Rev. B {\bf 69}, 035111 (2004).
470:
471:
472: \bibitem{qhth}
473: {\em Perspectives in Quantum Hall effects},
474: ed. S. Das Sarma and A. Pinczuk
475: (Wiley, New York, 1997).
476:
477: \bibitem{qhexp}
478: %I. B. Spielman, J. P. Eisenstein, L. N. Pfeiffer, and K. W. West,
479: I. B. Spielman {\em et al.},
480: \prl {\bf 84}, 5808 (2000).
481:
482: \bibitem{note1} As is usual for any Fermi liquid, at very low temperatures some kind of ordering, such as superconductivity, is
483: always possible. This will be ignored in our discussion.
484:
485:
486: \bibitem{burdin} A similar observation, albeit in a somewhat different context, was first made in
487: S. Burdin, D.R. Grempel, and A. Georges, Phys. Rev. B {\bf 66}, 045111 (2002).
488:
489: \bibitem{AffMar}
490: I. Affleck and J. B. Marston, Phys. Rev. B {\bf 37}, R3774 (1988).
491:
492: \bibitem{tJ} S.~Sachdev and N.~Read, Int. J. Mod. Phys. B {\bf 5}, 219 (1991).
493:
494: \bibitem{ybrhsi}
495: P. Gegenwart {\em et al.}, Phys. Rev. Lett. {\bf 89}, 056402 (2002).
496:
497: \bibitem{silke}
498: S. Paschen {\em et al.}, Nature {\bf 432}, 881 (2004).
499:
500: \bibitem{grigera}
501: S.~A.~Grigera {\em et al.}, Science {\bf 294}, 329 (2001) and unpublished.
502:
503: \bibitem{hooper}
504: J. Hooper {\em et al.}, \prl {\bf 92}, 257206 (2004).
505:
506:
507: \end{thebibliography}
508:
509: \end{document}
510: