1: % last correction: Aug 27, 2002 by PK
2: \tolerance = 10000
3: %
4: \documentclass[twocolumn,showpacs,preprintnumbers,prl,amsmath,amssymb,floatfix]{revtex4}
5: %\documentclass[galley,showpacs,preprintnumbers,prl,amsmath,amssymb]{revtex4}
6: %\documentclass[preprint,showpacs,preprintnumbers,prl,amsmath,amssymb,floatfix]{revtex4}
7: %
8: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
9: % for revtex 3:
10: %\documentstyle[aps,epsf,twocolumn]{revtex}
11: %
12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
13: %
14: %\usepackage{dcolumn}% Align table columns on decimal point
15: %\usepackage{bm}% bold math
16: \usepackage{epsfig}
17: %
18: \begin{document}
19: %
20: %\preprint{}
21: %
22: %\draft
23:
24: \title{Exact integral equation for the renormalized Fermi surface}
25: %
26: \author{Sascha Ledowski and Peter Kopietz}
27: %
28: \affiliation{Institut f\"{u}r Theoretische Physik, Universit\"{a}t Frankfurt,
29: Robert-Mayer-Strasse 8, 60054 Frankfurt, Germany}
30: %
31: \date{August 27, 2002}
32: %
33: %
34: \begin{abstract}
35: The true Fermi surface of a
36: fermionic many-body system
37: can be viewed as a fixed point manifold
38: of the renormalization group (RG).
39: Within the framework of the exact functional RG
40: we show that the fixed point condition implies an exact
41: integral equation for the counterterm which
42: is needed for a self-consistent calculation of the
43: Fermi surface.
44: In the simplest approximation, our integral equation
45: reduces to the self-consistent Hartree-Fock equation for the
46: counterterm.
47:
48:
49:
50: \end{abstract}
51: %
52: \pacs{71.10-w, 71-10.Hf, 71.18.+y}
53: %
54: \maketitle
55: %
56:
57: In his authoritative book on interacting Fermi systems
58: Nozi\`{e}res wrote fourty years ago \cite{Nozieres64}:
59: ``In practice, we shall never try to {\it{calculate}} the Fermi surface, which is much too difficult.''
60: What is the reason for this difficulty?
61: Formally, the Fermi surface of an interacting Fermi system
62: is defined as the set of all wavevectors ${\bf{k}}_F$
63: satisfying~\cite{Luttinger60}
64: \begin{equation}
65: \epsilon_{{ \bf{k}}_F } - \mu + \Sigma ( {\bf{k}}_F , i 0 ) = 0
66: \; ,
67: \label{eq:FSdef}
68: \end{equation}
69: where $\epsilon_{\bf{k}}$ is the energy dispersion in the absence of interactions,
70: $\mu$ is the chemical potential, and
71: $\Sigma ( {\bf{k}} , \omega )$ is the exact self-energy of the interacting system \cite{footnotereal}.
72: For simplicity we assume an infinite and
73: spin-rotationally invariant system at zero temperature, so that
74: $\Sigma ( {\bf{k}} , \omega )$ is independent of the spin.
75: Unfortunately,
76: the function $\Sigma ( {\bf{k}}_F , i 0 )$ in Eq. (\ref{eq:FSdef}) is not known a priori,
77: so that the calculation of the true Fermi surface requires
78: the solution of the many-body problem.
79:
80: For weak interactions, one might try to determine
81: the Fermi surface
82: perturbatively by simply calculating
83: $ \Sigma ( {\bf{k}} , i 0 )$ in powers of the interaction and
84: substituting the result into Eq. (\ref{eq:FSdef}).
85: However, in general the perturbation series
86: contains anomalous terms \cite{Kohn60} with
87: unphysical singularities,
88: which are generated because
89: the {\it{ground state}} of the non-interacting
90: system evolves into an {\it{excited state}} of the interacting system
91: when the interaction is adiabatically switched on.
92: As discussed by Nozi\`{e}res \cite{Nozieres64},
93: this artificial level crossing can be avoided by introducing
94: counterterms which are determined by
95: the requirement that
96: the Fermi surface remains fixed as the interaction is adiabatically switched on.
97: This intuitive idea
98: can be implemented perturbatively as follows \cite{Nozieres64,Feldman96}:
99: Suppose we would like to know the true Fermi surface of a system with
100: Hamiltonian
101: ${H} = {H}_0 + {H}_1$, where ${H}_1$ describes some general two-body interaction and
102: the non-interacting part is given by
103: \begin{equation}
104: {H}_0
105: = \sum_{ {\bf{k}} , \sigma} \epsilon_{\bf{k}} c^{\dagger}_{ {\bf{k}} \sigma }
106: c_{ {\bf{k}} \sigma }
107: \; .
108: \label{eq:H0def}
109: \end{equation}
110: Here $c_{ {\bf{k}} \sigma }$ are the usual annihilation operators of
111: fermions with momentum ${\bf{k}}$ and spin $\sigma$.
112: An expansion in powers of $H_1$ leads to Feynman diagrams
113: where vertices corresponding to $H_1$ are connected by
114: propagators
115: $G_0 ( {\bf{k}} , \omega ) = [ \omega - \epsilon_{\bf{k}} + \mu ]^{-1}$.
116: These are singular for $\omega = 0$ and
117: $ \epsilon_{\bf{k}} = \mu$, which is {\it{not}}
118: the true Fermi surface
119: defined in Eq. (\ref{eq:FSdef}).
120: If the perturbative expansion is truncated at a finite order, this leads to
121: the unphysical divergencies mentioned above \cite{Kohn60}. To avoid these, we
122: add the counterterm
123: $\sum_{ {\bf{k}} \sigma} \Sigma ( {\bf{k}}_F , i 0 )
124: c^{\dagger}_{ {\bf{k}} \sigma }
125: c_{ {\bf{k}} \sigma } $ to $H_0$ and subtract it again from $H_1$, writing
126: $H = H_0^{\prime} + H_1^{\prime}$, with
127: \begin{eqnarray}
128: {H}_0^{\prime}
129: = \sum_{ {\bf{k}} \sigma} [ \epsilon_{\bf{k}} +
130: \Sigma ( {\bf{k}}_F , i 0 ) ]
131: c^{\dagger}_{ {\bf{k}} \sigma }
132: c_{ {\bf{k}} \sigma }
133: \; ,
134: \label{eq:H0primedef}
135: \end{eqnarray}
136: and $ H_1^{\prime} = H_1 -
137: \sum_{ {\bf{k}} \sigma} \Sigma ( {\bf{k}}_F , i 0 )
138: c^{\dagger}_{ {\bf{k}} \sigma }
139: c_{ {\bf{k}} \sigma } $.
140: Here ${\bf{k}}_F$ is the wavevector closest to ${\bf{k}}$
141: lying on the Fermi surface, see Fig. \ref{fig:Fermisurface}.
142: %
143: %
144: \begin{figure}[tb]
145: \begin{center}
146: \epsfig{file=fig1.eps,width=70mm}
147: \end{center}
148: \vspace{-2mm}
149: \caption{%
150: Decomposition ${\bf{k}} = \hat{\bf{n}} k_F ( \hat{\bf{n}} ) + {\bf{p}} $
151: of a wavevector ${\bf{k}}$ into a component $ {\bf{k}}_F = \hat{\bf{n}} k_F ( \hat{\bf{n}} )$ on the Fermi surface
152: and a component $ {\bf{p}} $ in the direction of the
153: local Fermi velocity ${\bf{v}}_F$.
154: The thick solid line is a part of the Fermi surface. This construction
155: defines ${\bf{k}}_F$ and $\hat{\bf{n}}$ as a function of ${\bf{k}}$.
156: Here ${\bf{v}}_F = \nabla_{ {\bf{k}}} \epsilon_ { {\bf{k}} } |_{ {\bf{k}} = {\bf{k}}_F } $
157: is defined in terms of the gradient of the free dispersion at the true Fermi surface, so that
158: ${\bf{v}}_F$ is not necessarily perpendicular to the Fermi surface.
159: }
160: \label{fig:Fermisurface}
161: \end{figure}
162: %
163: %
164: Using Eq. (\ref{eq:FSdef}), the
165: corresponding free propagator can then be written as
166: $G_0^{\prime} ( {\bf{k}} , \omega ) = [ \omega - \epsilon_{\bf{k}} +
167: \epsilon_{ {\bf{k}}_F } ]^{-1}$, which by construction is singular on the true Fermi surface.
168: But how do we find the counterterm necessary to calculate $\epsilon_{ {\bf{k}}_F }$?
169: Following the usual strategy adopted in field theory \cite{ZinnJustin89},
170: we may expand the irreducible self-energy associated with
171: the modified interaction $H_1^{\prime}$ perturbatively in powers of $H_1^{\prime}$
172: and require that, order by order in perturbation theory,
173: the corrections vanish when we set $\omega = 0$ and
174: ${\bf{k}} = {\bf{k}}_F$. This renormalized
175: perturbation theory
176: leads to complicated integral equations for
177: $\Sigma ( {\bf{k}}_F , i0 )$, which must be solved
178: numerically \cite{Neumayr02}.
179:
180: A few years ago Anderson \cite{Anderson93}
181: critically discussed
182: a simplified version
183: of the renormalized perturbation theory outlined above,
184: where $\Sigma ( {\bf{k}}_F , i0 )$ is replaced by a constant $\delta \mu $.
185: He correctly pointed out that in general it is not allowed
186: to ignore the momentum-dependence of the counterterms, and
187: %then a constant counterterm $\delta \mu $ is certainly not sufficient
188: %to obtain the renormalized Fermi surface.
189: %Moreover,
190: argued that in two dimensions
191: the effective interaction between
192: electrons with opposite spin and wavevectors on the true Fermi surface depends
193: in a subtle way on the boundary conditions
194: which cannot be adequately
195: taken into account perturbatively.
196: To gain a better understanding of this problem, it should be useful
197: to have an algorithm for calculating the Fermi surface
198: which does not rely on the
199: perturbative expansion of counterterms in powers of the interaction.
200: In this work we show that
201: such an algorithm follows in a straightforward way from
202: the exact functional renormalization group (RG) approach
203: described in Ref.\cite{Kopietz01}.
204:
205:
206: In the past decade several
207: authors have used
208: Wilsonian RG methods to study
209: interacting Fermi systems
210: \cite{Kopietz01,Benfatto90,Polchinski92,Shankar94,Zanchi96,Salmhofer98,Halboth00,Binz02},
211: using different versions of the RG. In particular,
212: the RG transformations used in Refs. \cite{Zanchi96,Salmhofer98,Halboth00,Binz02}
213: exclusively focus on the mode elimination step.
214: Although such a procedure is sufficient
215: if one considers only the one-loop flow of marginal couplings,
216: in general a complete Wilsonian RG transformation consists not only of the
217: mode elimination, but
218: includes also the rescaling of momenta, frequencies and fields \cite{Ma76}.
219: While the field rescaling (i.e. the wavefunction renormalization)
220: becomes only important
221: beyond the one-loop approximation,
222: the RG flow of all relevant and irrelevant couplings
223: is determined in an essential way by the rescaling step
224: even at the one-loop level.
225: We emphasize that the rescaling of momenta and frequencies
226: is more than a trivial mathematical change of variables --
227: it is crucial to detect possible fixed points of the RG and to calculate
228: critical exponents \cite{Ma76}.
229: Moreover, as will be shown below, the length and time rescaling is
230: very helpful for a derivation of the
231: self-consistent Hartree-Fock approximation
232: within the framework of the exact RG.
233: The importance of rescaling in the presence of a Fermi surface has
234: been emphasized by Polchinski \cite{Polchinski92} and by Shankar \cite{Shankar94}.
235: In Ref.\cite{Kopietz01} we have shown how
236: the exact functional RG approach developed previously \cite{Zanchi96,Salmhofer98,Halboth00}
237: can be modified to include the rescaling step.
238: Here, we shall show that the rescaling is crucial
239: to obtain a non-perturbative definition
240: of the Fermi surface as a fixed point manifold
241: of the RG.
242: We obtain
243: an explicit algorithm for calculating the Fermi surface which does not rely on the iterative
244: procedure of fixing the counterterms order by order in perturbation theory.
245: The fixed point property of the Fermi surface has also
246: been emphasized by Ferraz \cite{Ferraz02}, who recently discussed
247: the Fermi surface renormalization in a
248: special two-dimensional system using the
249: field theoretical RG.
250:
251:
252: The exact functional RG can be formulated in terms of an
253: infinite hierarchy of coupled differential equations for the irreducible $2n$-point vertices
254: $\Gamma^{(2n)}_{\xi} ( K_1^{\prime} , \ldots , K_n^{\prime} ; K_n , \ldots , K_1 )$, where
255: $K = ( \sigma , {\bf{k}} , i \omega )$ is a collective label for spin projection $\sigma$,
256: momentum ${\bf{k}}$, Matsubara frequency $i \omega $.
257: Here $\xi$ is an infrared cutoff with units of energy which regularizes the
258: singularity of the free propagator.
259: % \begin{equation}
260: % G_{0, \xi}^{\prime} ( K ) =
261: % \frac{ \Theta_{\gamma} ( \Omega_K -
262: % \xi ) - \Theta_{\gamma} ( \Omega_K - \xi_0 )
263: %}{ i \omega_n - \epsilon_{\bf{k}} + \epsilon_{ {\bf{k}}_F } }
264: % \; .
265: % \label{eq:propdef}
266: % \end{equation}
267: To scale wavevectors toward the Fermi surface, it is convenient
268: to perform a
269: non-linear coordinate transformation in momentum space,
270: ${\bf{k}} = \hat{\bf{n}} k_F ( \hat{\bf{n}} ) + \hat{\bf{v}}_F \xi q / | {\bf{v}}_F |$,
271: and eliminate ${\bf{k}}$ in favor of the dimensionless variable $q$ and the unit vector
272: $\hat{\bf{n}}$.
273: Here
274: $ k_F ( \hat{\bf{n}} )$ is the length of ${\bf{k}}_F$ parameterized by $\hat{\bf{n}} $,
275: and ${\bf{v}}_F = \nabla_{\bf{k}} \epsilon_{\bf{k}} |_{ {\bf{k}}_F}$
276: is the Fermi velocity of the non-interacting
277: system at the true ${\bf{k}}_F$,
278: see Fig. \ref{fig:Fermisurface}.
279: The corresponding unit vector is denoted by
280: $\hat{\bf{v}}_F = {\bf{v}}_F / | {\bf{v}}_F |$.
281: Geometrically, $q = {\bf{v}}_F \cdot {\bf{p}} / \xi =
282: {\bf{v}}_F \cdot ( {\bf{k}} - \hat{\bf{n}} k_F ( \hat{\bf{n}} ) )/ \xi$ measures the
283: distance of a given ${\bf{k}}$ from the Fermi surface in units of the infrared cutoff.
284: We also define rescaled frequencies $\epsilon = \omega / \xi$, and label
285: the degrees of freedom by $Q = ( \sigma , \hat{\bf{n}} , q , i \epsilon )$ instead of
286: $K$. To implement the {\it{scaling toward the Fermi surface}} \cite{Polchinski92}
287: we consider the RG flow of the following rescaled vertices \cite{Kopietz01}
288: \begin{eqnarray}
289: \tilde{\Gamma}_{t}^{ (2n) }
290: ( Q_1^{\prime} , \ldots , Q_n^{\prime} ; Q_n , \ldots , Q_1 ) =
291: & &
292: \nonumber \\
293: & & \hspace{-40mm} \nu_0^{n-1} \xi^{n-2}
294: \left[ Z_t^{\hat{\bf{n}}_{1}^{\prime} }
295: \cdots Z_t^{\hat{\bf{n}}_{n}^{\prime}}
296: Z_t^{\hat{\bf{n}}_{n} } \cdots Z_t^{\hat{\bf{n}}_{1}}
297: \right]^{1/2}
298: \nonumber
299: \\
300: & & \hspace{-40mm} \times
301: \Gamma^{(2n)}_{\xi} ( K_1^{\prime} , \ldots , K_n^{\prime}
302: ; K_n , \ldots ,
303: K_1 )
304: \; .
305: \label{eq:Gammarescale}
306: \end{eqnarray}
307: Here $t = - \ln ( \xi / \xi_0 )$ is a logarithmic flow parameter,
308: $\nu_0 = \int \frac{ d {\bf{k}}}{ (2 \pi )^D} \delta (
309: \epsilon_{ {\bf{k}} } - \epsilon_{ {\bf{k}}_F } ) $ is the
310: density of states (per spin projection) of the non-interacting
311: system at the true Fermi surface, and
312: \begin{equation}
313: Z_t^{\hat{\bf{n}}} = \left[ 1 - \left. \frac{\partial
314: \Sigma_{\xi} ( {\bf{k}}_F , \omega + i 0)}{\partial
315: \omega } \right|_{\omega=0} \right]^{-1}
316: \label{eq:Zdef}
317: \end{equation}
318: is the wavefunction renormalization factor, where
319: $\Sigma_{\xi} ( {\bf{k}} , i \omega )$ is the irreducible self-energy of the system with
320: infrared cutoff $\xi = \xi_0 e^{-t}$. To determine the true
321: Fermi surface self-consistently it turns out to be useful to subtract the counterterm
322: $\Sigma ( {\bf{k}}_F , i0 )$
323: from the irreducible two-point vertex\cite{Kopietz01}, defining
324: \begin{equation}
325: \tilde{\Gamma}_{t}^{(2)}
326: ( Q ) = \frac{Z_t^{\hat{\bf{n}} } }{\xi} \Gamma^{(2)}_{\xi} ( K ) =
327: - \frac{Z_t^{\hat{\bf{n}} } }{\xi}
328: \left[ \Sigma_{\xi} ( K ) - \Sigma ( {\bf{k}}_F , i0 ) \right]
329: \; .
330: \label{eq:Gammatdef}
331: \end{equation}
332: The subtracted two-point vertex satisfies the exact flow equation \cite{Kopietz01}
333: \begin{eqnarray}
334: \partial_t \tilde{\Gamma}_t^{(2)} ( Q ) & = &
335: ( 1 - \eta_t^{\hat{\bf{n}}} - q \partial_q - \epsilon \partial_{\epsilon} )
336: \tilde{\Gamma}_t^{(2)} ( Q )
337: \nonumber
338: \\
339: & - & \int_{Q^{\prime}}
340: \dot{G}_t ( Q^{\prime} )
341: \tilde{\Gamma}_{t}^{(4)}
342: ( Q , Q^{\prime} ; Q^{\prime} , Q )
343: \; ,
344: \label{eq:twopointscale}
345: \end{eqnarray}
346: where
347: $
348: \eta_t^{\hat{\bf{n}}} = - \partial_t \ln Z_t^{\hat{\bf{n}}}$
349: is the flowing anomalous dimension (which vanishes for large $t$ if the system is a Fermi liquid).
350: The integration measure in Eq. (\ref{eq:twopointscale}) is defined by
351: \begin{equation}
352: \int_Q
353: = \sum_{\sigma}
354: \int \frac{ d S_{\hat{\bf{n}} }}{S_D}
355: \int dq J ({\hat{\bf{n}}} , q ) \int \frac{ d \epsilon}{2 \pi}
356: \label{eq:measure}
357: \; \; ,
358: \end{equation}
359: where $dS_{\hat{\bf{n}}}$ is a surface element and $S_D$
360: is the surface area of
361: the unit sphere in $D$ dimensions, and
362: $J ( { \hat{\bf{n}} } , q )$ is the Jacobian
363: associated with the transformation ${\bf{k}} \rightarrow
364: ( \hat{\bf{n}} , q )$ divided by $\nu_0 \xi^2$.
365: The function $ \dot{{G}}_t ( Q )$ in Eq. (\ref{eq:twopointscale}) is defined by
366: \begin{equation}
367: \dot{{G}}_t ( Q ) = \frac{ \delta ( \tilde{\Omega}_Q -1 )}{
368: Z_t^{\hat{\bf{n}}} \left[ i \epsilon -
369: \xi_t^{ \hat{\bf{n}}} ( q ) \right] +
370: \tilde{\Gamma}_t^{(2)} ( Q ) }
371: \; ,
372: \end{equation}
373: where $
374: \xi_t^{ \hat{\bf{n}} } ( q ) =
375: ( \epsilon_{ {\bf{k}}_t } -
376: \epsilon_{ {\bf{k}}_F } ) / {\xi}$, with
377: ${\bf{k}}_t = \hat{\bf{n}} k_F ( \hat{\bf{n}} ) + \hat{\bf{v}}_F \xi_0 e^{-t} q / | {\bf{v}}_F |$.
378: Here
379: $\tilde{\Omega}_Q$ is some function that measures the distance from the Fermi surface, for example
380: $\tilde{\Omega}_Q = |\xi_t^{ \hat{\bf{n}} } ( q ) | \approx |q |$.
381: The flow equation of the rescaled irreducible four-point vertex
382: $\tilde{\Gamma}_{t}^{ (4)}
383: ( Q_1^{\prime} , Q_2^{\prime} ; Q_2 , Q_1 ) $
384: is not explicitly needed in this work; it can be found in
385: Ref.\cite{Kopietz01}.
386:
387: The shape of the Fermi surface of the interacting system
388: is determined by the RG flow of the couplings
389: \begin{equation}
390: \tilde{\mu}_t^{\hat{\bf{n}}} \equiv \tilde{\Gamma}_{t}^{(2)}
391: ( \sigma , \hat{\bf{n}} , q= 0, i \epsilon = i0 )
392: \label{eq:mudef}
393: \; .
394: \end{equation}
395: From Eq. (\ref{eq:twopointscale}) we see
396: that these couplings satisfy the exact flow equation
397: \begin{equation}
398: \partial_t \tilde{\mu}_t^{\hat{\bf{n}}} = ( 1 - \eta_t^{\hat{\bf{n}}} ) \tilde{\mu}_t^{\hat{\bf{n}}}
399: + \dot{\Gamma}_t^{(2)} ( \hat{\bf{n}} )
400: \; ,
401: \label{eq:muexact}
402: \end{equation}
403: with
404: \begin{equation}
405: \dot{\Gamma}_t^{(2)} ( \hat{\bf{n}} ) =
406: - \int_{Q^{\prime}}
407: \dot{G}_t ( Q^{\prime} )
408: \tilde{\Gamma}_{t}^{(4)}
409: ( Q_0 , Q^{\prime} ; Q^{\prime} , Q_0 )
410: \; ,
411: \end{equation}
412: where $Q_0 = ( \sigma , \hat{\bf{n}} , q=0, i \epsilon = i0 )$.
413: Note that in dimensions $D > 1$
414: there are infinitely many couplings $\tilde{\mu}_t^{\hat{\bf{n}}}$, labelled by
415: the unit vector $\hat{\bf{n}}$.
416: Obviously, each $\tilde{\mu}_t^{\hat{\bf{n}}}$ with
417: $ \eta_{\infty}^{ \hat{\bf{n}}} =
418: \lim_{ t \rightarrow \infty } \eta_t^{\hat{\bf{n}}} < 1 $
419: is relevant, so that
420: some fine tuning of the bare couplings $\tilde{\mu}_0^{\hat{\bf{n}}}$
421: is necessary to force $\tilde{\mu}_t^{\hat{\bf{n}}}$ to flow into a fixed point of the RG.
422: Because a finite limit $\tilde{\mu}_\infty^{\hat{\bf{n}}} = \lim_{ t \rightarrow \infty }
423: \tilde{\mu}_t^{\hat{\bf{n}}} $ means that we have found the true Fermi surface of the
424: interacting system \cite{Kopietz01},
425: we conclude that the detailed shape of the Fermi surface
426: is sensitive to the numerical values of the bare couplings of the theory.
427: Polchinski \cite{Polchinski92} pointed out that such a fine tuning of the bare couplings
428: is {\it{unnatural}}, because
429: physical effects which depend on the
430: precise shape of the Fermi surface
431: are not protected against small perturbations.
432: %In other words, the shape of the Fermi surface is not a ``quantum protectorate'' \cite{Anderson00}.
433:
434:
435: To further elucidate the relation between the
436: relevant couplings $\tilde{\mu}_t^{\hat{\bf{n}}}$ and the shape of the Fermi surface,
437: let us explicitly derive from Eq. (\ref{eq:muexact})
438: a self-consistency condition for the Fermi surface.
439: Therefore it is useful to transform Eq. (\ref{eq:muexact}) into an integral equation,
440: \begin{equation}
441: \tilde{\mu}_t^{\hat{\bf{n}} } = e^{ t - \int_0^t d \tau \eta_{\tau}^{\hat{\bf{n}}} }
442: \left[
443: \tilde{\mu}_0^{\hat{\bf{n}} } +
444: \int_{0}^{t} d t^{\prime} e^{ - t^{\prime} + \int_0^{t^{\prime}} d \tau
445: \eta_{\tau}^{\hat{\bf{n}}} }
446: \dot{\Gamma}^{(2)}_{t^{\prime}} ( \hat{\bf{n}} )
447: \right]
448: \; .
449: \label{eq:integral}
450: \end{equation}
451: Suppose now that we have adjusted the bare couplings such that
452: for $t \rightarrow \infty $
453: the flowing couplings $\tilde{\mu}_t^{\hat{\bf{n}} }$ indeed approach finite fixed point values.
454: Assuming that the associated anomalous dimensions
455: $ \eta_{\infty}^{\hat{\bf{n}}} $
456: are smaller than unity\cite{footnoteeta},
457: we conclude from Eq. (\ref{eq:integral}) that
458: the limit $ \tilde{\mu}_\infty^{\hat{\bf{n}} } = \lim_{ t \rightarrow \infty }
459: \tilde{\mu}_t^{\hat{\bf{n}} } $ can only be finite if the initial values
460: $\tilde{\mu}_0^{ \hat{\bf{n}} }$ are chosen such that
461: \begin{eqnarray}
462: \tilde{\mu}_0^{\hat{\bf{n}} } & = & -
463: \int_{0}^{\infty} d t e^{ - t + \int_0^{t} d \tau
464: \eta_{\tau}^{\hat{\bf{n}}} }
465: \dot{\Gamma}^{(2)}_{t} ( \hat{\bf{n}} )
466: \nonumber
467: \\
468: & = & \int_{0}^{\infty} d t e^{ - t + \int_0^{t} d \tau
469: \eta_{\tau}^{\hat{\bf{n}}} }
470: \int_{Q^{\prime}}
471: \dot{G}_t ( Q^{\prime} )
472: \tilde{\Gamma}_{t}^{(4)}
473: ( Q_0 , Q^{\prime} ; Q^{\prime} , Q_0 )
474: \; .
475: \nonumber
476: \\
477: & &
478: \label{eq:integral2}
479: \end{eqnarray}
480: This is an implicit equation for $ \tilde{\mu}_0^{\hat{\bf{n}} } $, relating
481: it to the values of the two-point vertex and the four-point vertex
482: on the entire RG trajectory. Keeping in mind that
483: the right-hand side of Eq. (\ref{eq:integral2})
484: implicitly depends on $\tilde{\mu}^{ {\hat{\bf{n}} }}_t$ and that
485: according to Eq. (\ref{eq:Gammatdef})
486: \begin{equation}
487: \Sigma ( {\bf{k}}_F , i0 ) - \Sigma_{\xi_0} ( {\bf{k}}_F , i0 ) =
488: \xi_0 \tilde{\mu}_0^{\hat{\bf{n}}} / Z_0^{\hat{\bf{n}} }
489: \; ,
490: \label{eq:mugamma}
491: \end{equation}
492: it is obvious that Eq. (\ref{eq:integral2})
493: can be regarded as an integral equation for the
494: counterterm $ \Sigma ( {\bf{k}}_F , i0 )$, the solution of which yields
495: the true shape of the Fermi surface.
496: At this point
497: it is instructive to transform Eq. (\ref{eq:integral2}) back to unrescaled variables,
498: choosing for simplicity the initial conditions
499: $ \Sigma_{\xi_0} ( {\bf{k}}_F , i0 ) = 0$ and $ Z_0^{\hat{\bf{n}}} = 1$.
500: Using the above definitions we find that Eq. (\ref{eq:integral2}) is
501: equivalent with
502: \begin{eqnarray}
503: \Sigma ( {\bf{k}}_F , i0 ) & = & \sum_{\sigma^{\prime} }
504: \int \frac{ d {\bf{k}}^{\prime} }{ (2 \pi )^{D}}
505: \frac{ d \omega^{\prime} }{ 2 \pi }
506: \frac{ \Theta ( \xi_0 - \xi_{ {\bf{k}}^{\prime} } ) }{ i \omega^{\prime} -
507: \epsilon_{ {\bf{k}}^{\prime}}
508: + \mu - \Sigma_{ \xi_{ {\bf{k}}^{\prime} } } ( K^{\prime} ) }
509: \nonumber
510: \\
511: & & \hspace{-17mm} \times
512: \Gamma^{(4)}_{ \xi_{ {\bf{k}}^{\prime} } } ( {\bf{k}}_F , i0 , \sigma ,
513: {\bf{k}}^{\prime}, i \omega^{\prime} , \sigma^{\prime} ;
514: {\bf{k}}^{\prime}, i \omega^{\prime} , \sigma^{\prime} , {\bf{k}}_F , i0 , \sigma )
515: \; ,
516: \label{eq:counterint}
517: \end{eqnarray}
518: where $ \xi_{ {\bf{k}}^{\prime} } = | \epsilon_{ {\bf{k}}^{\prime} } -
519: \epsilon_{ {\bf{k}}_F^{\prime}} | $.
520: Note that the right-hand side of Eq. (\ref{eq:counterint}) involves the flowing self-energy and
521: four-point vertex at the scales $ \xi = \xi_{ {\bf{k}}^{\prime} }$ which depend on the
522: distance from the true Fermi surface.
523: We emphasize that the exact integral equation
524: (\ref{eq:counterint}) and the equivalent rescaled
525: equation (\ref{eq:integral2})
526: fix the counterterm $\Sigma ( {\bf{k}}_F , i0 )$ from the
527: requirement that for $ t \rightarrow \infty$
528: all couplings approach finite fixed point values.
529:
530:
531: If the system is a Luttinger liquid, then the rescaled version (\ref{eq:integral2}) of the
532: self-consistency equation is more convenient than the unrescaled version (\ref{eq:counterint}),
533: because for a Luttinger liquid the marginal part of the four-point vertex vertex
534: $\Gamma^{(4)}_{\xi}$ without wavefunction renormalization factors
535: diverges for $\xi \rightarrow 0$, while the rescaled four-point
536: vertex $\tilde{\Gamma}^{(4)}_t$ defined in
537: Eq.(\ref{eq:Gammarescale})
538: approaches for $t \rightarrow \infty$ a finite limit.
539: In this case the divergence of
540: the unrescaled $\Gamma^{(4)}_{\xi}$
541: is canceled by the vanishing wavefunction
542: renormalization factors at the Luttinger liquid fixed point \cite{Busche02}.
543:
544:
545: Given a solution
546: $ \Sigma ( {\bf{k}}_F , i0 )$
547: of Eq. (\ref{eq:counterint}), we may calculate the compressibility
548: of the system by substituting
549: the result for $ \Sigma ( {\bf{k}}_F , i0 )$ into
550: Eq. (\ref{eq:FSdef}) and solving for
551: $k_F ( \hat{\bf{n}} )$, which implicitly depends
552: on the chemical potential $\mu$.
553: According to the Luttinger theorem \cite{Luttinger60}
554: the density $n ( \mu )$ of the system is determined by
555: the volume enclosed by the Fermi surface,
556: \begin{equation}
557: n ( \mu ) = \sum_{\sigma }
558: \int \frac{ d {\bf{k}} }{ (2 \pi )^{D}} \Theta ( k_F ( \hat{\bf{n}} ) - | {\bf{k}} | )
559: \; ,
560: \label{eq:FSdens}
561: \end{equation}
562: so that we obtain for the compressibility $\chi_n$
563: \begin{equation}
564: n^2 \chi_n =
565: \frac{ \partial n}{\partial \mu } = \sum_{\sigma}
566: \int \frac{ d S_{ {\hat{\bf{n}}} } }{ (2 \pi )^D}
567: k_F^{D-1} ( \hat{\bf{n}} ) \frac{ \partial k_F ( \hat{\bf{n}} ) }{\partial \mu }
568: \; .
569: \label{eq:compres}
570: \end{equation}
571: Note that the compressibility is a
572: functional of the shape of the true Fermi surface
573: of the many-body system, i.e. the compressibility characterizes the RG fixed point.
574: Because the rescaling of momenta and frequencies is crucial to
575: obtain fixed points of the RG \cite{Ma76},
576: the compressibility and other uniform susceptibilities
577: are not so easy to calculate within alternative
578: versions of the RG which
579: omit the rescaling \cite{Salmhofer98,Halboth00,Binz02}.
580:
581:
582: From our point of view, the Fermi surface is a {\it{fixed point}} manifold of the RG,
583: so that it is meaningless to talk about the RG flow of the Fermi surface.
584: However, it is possible to define a ``flowing Fermi surface'' ${\bf{k}}_{F,t}$ via \cite{Kopietz01}
585: \begin{equation}
586: \epsilon_{{ \bf{k}}_{F,t} } - \mu + \Sigma_{ \xi_0 e^{-t}} ( {\bf{k}}_{F,t} , i 0 ) = 0
587: \label{eq:FSdefaux}
588: \; ,
589: \end{equation}
590: which by construction approaches the true Fermi surface
591: for $t \rightarrow \infty$.
592: If we choose the initial conditions at $t=0$ (corresponding to $\xi = \xi_0$) such that
593: $\Sigma_{ \xi_0 } ( {\bf{k}}_{F,0} , i 0 ) = 0$, then
594: ${\bf{k}}_{F,0}$ is the Fermi surface of the non-interacting system at the
595: {\it{same chemical potential}}
596: as the interacting system.
597: This corresponds to a non-interacting system at the
598: density
599: $n_0 ( \mu ) = \sum_{\sigma }
600: \int \frac{ d {\bf{k}} }{ (2 \pi )^{D}} \Theta ( k_{F,0} ( \hat{\bf{n}} ) - | {\bf{k}} | )$,
601: which is in general different from
602: the density $n ( \mu )$
603: of the interacting system given in Eq. (\ref{eq:FSdens}).
604: Note that
605: in Fermi liquid theory one usually works at constant density.
606: However, as discussed by Nozi\`{e}res \cite{Nozieres237},
607: for a self-consistent calculation of the Fermi surface it is more convenient
608: to determine the counterterm at constant chemical potential $\mu$, and
609: calculate the corresponding density afterwards \cite{footnotedens}.
610: The practical advantages of such a procedure have been recognized previously in
611: Ref. \cite{Gonzalez00}. Moreover, in the field-theoretical approach
612: advanced by Ferraz \cite{Ferraz02} the chemical potential is also a RG invariant.
613:
614:
615: In the simplest approximation
616: Eqs. (\ref{eq:integral2}) and (\ref{eq:counterint}) reduce to
617: the Hartree-Fock self-consistency equation for the counterterm
618: $ \Sigma ( {\bf{k}}_F , i0 )$.
619: To see this, we approximate the flowing four-point vertex
620: in Eq. (\ref{eq:counterint}) as follows
621: \begin{eqnarray}
622: \Gamma^{(4)}_{ \xi_{ {\bf{k}}^{\prime} } } ( {\bf{k}}_F , i0 , \sigma ,
623: {\bf{k}}^{\prime}, i \omega^{\prime} , \sigma^{\prime} ;
624: {\bf{k}}^{\prime}, i \omega^{\prime} , \sigma^{\prime} , {\bf{k}}_F , i0 , \sigma )
625: & &
626: \nonumber
627: \\
628: & & \hspace{-72mm} \approx \Gamma^{(4)}_{ \xi = 0 } ( {\bf{k}}_F , i0 , \sigma ,
629: {\bf{k}}^{\prime}_F , i0 , \sigma^{\prime} ;
630: {\bf{k}}^{\prime}_F , i 0 , \sigma^{\prime} , {\bf{k}}_F , i0 , \sigma )
631: \nonumber
632: \\
633: & & \hspace{-72mm} \equiv \Gamma_0^{(4)}
634: ( {\bf{k}}_F, \sigma ; {\bf{k}}_F^{\prime} , \sigma^{\prime} )
635: \; ,
636: \label{eq:Landaudef}
637: \end{eqnarray}
638: i.e. we project all momenta onto the Fermi surface, ignore the
639: frequency dependence, and replace the flowing vertex by its
640: fixed point value.
641: Moreover, at this level of approximation we may also replace
642: $ \Sigma_{ \xi_{ {\bf{k}}^{\prime} } } ( K^{\prime} )
643: \rightarrow \Sigma ( {\bf{k}}_F^{\prime} , i0 )$ on the right-hand side of
644: Eq. (\ref{eq:counterint}). For $\xi_0 \rightarrow \infty$ we then obtain
645: the Hartree-Fock self-consistency equation
646: \begin{eqnarray}
647: \Sigma ( {\bf{k}}_F , i0 ) & = & \sum_{\sigma^{\prime} }
648: \int \frac{ d {\bf{k}}^{\prime} }{ (2 \pi )^{D}}
649: \Gamma_0^{(4)}
650: ( {\bf{k}}_F, \sigma ; {\bf{k}}_F^{\prime} , \sigma^{\prime} )
651: \nonumber
652: \\
653: & \times &
654: \Theta \left( \mu -
655: \epsilon_{ {\bf{k}}^{\prime}} - \Sigma ( {\bf{k}}_F^{\prime} , i0 ) \right)
656: \; .
657: \label{eq:MF}
658: \end{eqnarray}
659: For a given interaction vertex
660: $ \Gamma_0^{(4)}
661: ( {\bf{k}}_F, \sigma ; {\bf{k}}_F^{\prime} , \sigma^{\prime} ) $,
662: Eq. (\ref{eq:MF})
663: can be used to study possible
664: Fermi surface instabilities such as the Pomeranchuk instability \cite{Pomenanchuk58} at
665: the Hartree-Fock level, which should always be
666: the first step before more elaborate methods are used.
667: A trivial generalization of Eq. (\ref{eq:MF}) with a
668: spin-dependent counterterm $\Sigma ( {\bf{k}}_F , i0 , \sigma )$
669: leads to the self-consistent Hartree-Fock equation for spontaneous
670: ferromagnetism, implying the usual Stoner instability
671: for strong enough interactions in the spin-channel.
672: More generally, if we allow for other types of symmetry
673: breaking, within the same approximations as above
674: the exact RG fixed point equation for the
675: two-point vertex can be reduced to
676: the Hartree-Fock self-consistency equation for the corresponding
677: order parameter.
678: Note that no approximation has been made in deriving
679: Eqs. (\ref{eq:integral2}) and (\ref{eq:counterint}), so that
680: these exact fixed point equations (or generalizations thereof for
681: other types of symmetry breaking)
682: can serve as a starting point
683: for a systematic calculation of corrections to the Hartree-Fock approximation.
684:
685:
686: In summary, in this work we have shown how the true Fermi surface of
687: an interacting Fermi system can be defined self-consistently as a fixed point
688: property of the RG. Our main result are the two equivalent
689: integral equations (\ref{eq:integral2}) and (\ref{eq:counterint}),
690: which determine the counterterm $\Sigma ( {\bf{k}}_F , i0 )$
691: necessary to calculate the Fermi surface from Eq. (\ref{eq:FSdef}).
692: In the simplest approximation, Eq. (\ref{eq:counterint}) reduces
693: to the Hartree-Fock self-consistency
694: condition for the counterterm. However, systematic improvements are possible.
695:
696: Very recently Dusuel and Dou\c{c}ot \cite{Dusuel02}
697: presented a detailed analysis of the Fermi surface deformations in
698: quasi one-dimensional electronic systems, using perturbation theory and
699: the RG method.
700: They realized that some ``slight modification'' of the Wilson-Polchinski RG
701: approach is necessary in order to use this approach for a self-consistent
702: calculation of the Fermi surface, but admitted that
703: a practical implementation of such a modification remains to be attempted.
704: We have shown here that
705: the necessary modification of the functional RG
706: used in Refs. \cite{Zanchi96,Salmhofer98,Halboth00,Binz02}
707: is simply the usual rescaling step \cite{Kopietz01},
708: which is an essential part of the
709: ``orthodox'' Wilsonian RG \cite{Ma76}.
710: We conclude that
711: the functional RG approach for interacting Fermi systems
712: in the form advanced in Ref. \cite{Kopietz01}
713: provides an elegant solution to the difficult \cite{Nozieres64} problem
714: of self-consistently constructing the true Fermi surface.
715:
716:
717:
718: We thank Lorenz Bartosch, Tom Busche, Alvaro Ferraz, Volker Meden, Walter Metzner
719: for discussions.
720: This work was partially supported by the DFG via
721: Forschergruppe FOR 412, Project No. KO 1442/5-1.
722:
723: \vspace{-0.5mm}
724: \begin{thebibliography}{99}
725: \vspace{-0.5mm}
726: %
727: \bibitem{Nozieres64}
728: P.\ Nozi\`{e}res, {\it{Theory of Interacting Fermi Systems}},
729: (Benjamin, New York, 1964).
730: %
731: \bibitem{Luttinger60}
732: J. M.Luttinger, Phys. Rev. {\bf{119}}, 1153 (1960).
733: %
734: \bibitem{footnotereal}
735: We assume that $\Sigma ( {\bf{k}}_F , i0 )$ is real, which is always true
736: for Fermi liquids, where the damping of quasiparticles with wavevectors
737: precisely on the Fermi surface vanishes.
738: %Otherwise one should
739: %substitute $\Sigma ( {\bf{k}}_F , i0 ) \rightarrow
740: %{\rm Re} \Sigma ( {\bf{k}}_F , i0 )$ in all expressions involving the
741: %counterterm.
742: %
743: \bibitem{Kohn60}
744: W. Kohn and J. M. Luttinger, Phys. Rev. {\bf{118}}, 41 (1960).
745: %
746: \bibitem{Feldman96}
747: J.Feldman, M. Salmhofer, and E. Trubowitz,
748: J. Stat.Phys. {\bf{84}}, 1209 (1996);
749: J.Feldman, H. Kn\"{o}rrer, M. Salmhofer and E.Trubowitz,
750: {\it{ibid.}} {\bf{94}}, 113 (1999).
751: %
752: \bibitem{ZinnJustin89}
753: See, for example, J. Zinn-Justin, {\it{Quantum Field Theory and Critical Phenomena}},
754: (Clarendon Press, Oxford, 1989).
755: %
756: \bibitem{Neumayr02}
757: A. Neumayr and W.Metzner, cond-mat/0208431.
758: %
759: \bibitem{Anderson93}
760: P. W. Anderson, Phys. Rev. Lett. {\bf{71}}, 1220 (1993)
761: %
762: \bibitem{Kopietz01}
763: P.\ Kopietz and T.\ Busche, Phys.\ Rev.\ B {\bf{64}}, 155101 (2001).
764: %
765: \bibitem{Benfatto90}
766: G. Benfatto and G. Gallavotti, J. Stat. Phys. {\bf{59}}, 541 (1990);
767: Phys. Rev. B {\bf{42}}, 9967 (1990).
768: %
769: \bibitem{Polchinski92}
770: J. Polchinski, {\it{Effective Field Theory and the Fermi Surface}},
771: in {\it{Proceedings of the 1992 Theoretical Advanced Study Institute in
772: Elementary Particle Physics}} (World Scientific, Singapore, 1992).
773: %
774: \bibitem{Shankar94}
775: R. Shankar, Rev. Mod. Phys. {\bf{66}}, 129 (1994).
776: %
777: \bibitem{Zanchi96}
778: D. Zanchi and H. J. Schulz, Phys. Rev. B {\bf{54}}, 9509 (1996);
779: Z. Phys. B {\bf{103}}, 339 (1997); Phys. Rev. B {\bf{61}}, 13609 (2000).
780: %
781: \bibitem{Salmhofer98}
782: M. Salmhofer, {\it{Renormalization}} (Springer, Berlin, 1998);
783: C. Honerkamp, M. Salmhofer, N. Furukawa, and T. M. Rice,
784: Phys. Rev. B {\bf{63}}, 45114 (2001);
785: M. Salmhofer and C. Honerkamp, Prog. Theor. Phys.
786: {\bf{105}}, 1, (2001);
787: C. Honerkamp and M. Salmhofer, Phys. Rev. B {\bf{64}}, 184516 (2001);
788: C. Honerkamp, Ph.D. thesis,
789: ETH Z\"{u}rich, (2000).
790: %
791: \bibitem{Halboth00}
792: C. J. Halboth and W. Metzner, Phys. Rev. B {\bf{61}}, 4364 (2000);
793: Phys. Rev. Lett. {\bf{85}}, 5162 (2001);
794: C. J. Halboth, Ph.D. thesis, Shaker-Verlag, Aachen, (1999).
795: %
796: \bibitem{Binz02}
797: B. Binz, D. Baeriswyl and B. Dou\c{c}ot, Eur. Phys. J. B {\bf 25}, 69 (2002);
798: B. Binz, Ph. D. thesis, cond-mat/0207067.
799: %
800: \bibitem{Ma76}
801: See, for example, S. K. Ma, {\it{Modern Theory of Critical Phenomena}}, (Benjamin/Cummings,
802: Reading, MA, 1976).
803: %
804: \bibitem{Ferraz02}
805: A. Ferraz, cond-mat/0203376.
806: %
807: %\bibitem{Anderson00}
808: %P. W. Anderson, Sciene {\bf{288}}, 480 (2000).
809: %
810: \bibitem{footnoteeta}
811: For $\eta_{\infty}^{\hat{\bf{n}}} > 1$ the
812: coupling $\tilde{\mu}_{t}^{\hat{\bf{n}}}$ becomes irrelevant and
813: approaches a constant value (even without fine tuning) provided the other couplings
814: flow into a fixed point.
815: In this case the gradient of the momentum distribution
816: $ \langle c^{\dagger}_{{\bf{k}} \sigma } c_{ {\bf{k}} \sigma } \rangle $ with respect to ${\bf{k}}$
817: is finite for all ${\bf{k}}$, so that
818: a sharp Fermi surface does not exist.
819: %
820: \bibitem{Busche02}
821: T.\ Busche, L.\ Bartosch, and P.\ Kopietz, cond-mat/0202175,
822: (J. Phys. C: Cond. Mat., in press).
823: %
824: \bibitem{Nozieres237}
825: See the discussion on p. 237 of Ref.\cite{Nozieres64}.
826: %
827: \bibitem{footnotedens}
828: To compare the Fermi surface of the interacting system
829: with the corresponding Fermi surface without interactions at the same density $n$,
830: we should first calculate the density $n ( \mu )$
831: corresponding to the Fermi surface ${\bf{k}}_F$ of the interacting system
832: from Eq. (\ref{eq:FSdens}), and then determine
833: $\mu_0$ such that
834: $n ( \mu ) = \sum_{\sigma }
835: \int \frac{ d {\bf{k}} }{ (2 \pi )^{D}} \Theta ( \mu_0 - \epsilon_{ {\bf{k}}} )$.
836: Given $\mu_0$, we may calculate the Fermi surface of the non-interacting system
837: at the density $n ( \mu )$ from
838: the roots of $ \epsilon_{ {\bf{k}} }= \mu_0 $.
839: %
840: \bibitem{Gonzalez00}
841: J. Gonz\'{a}lez, F. Guinea, and M. A. H. Vozmediano,
842: Phys. Rev. Lett. {\bf{84}}, 4930 (2000).
843: %
844: \bibitem{Pomenanchuk58}
845: I. Ia. Pomeranchuk,
846: Zh. Eksp. Teor. Fiz. {\bf{35}}, 524 (1958)
847: [Sov. Phys. JETP {\bf{8}}, 361 (1958)].
848:
849: %
850: \bibitem{Dusuel02}
851: S. Dusuel and B. Dou\c{c}ot, cond-mat/0208434.
852: %
853:
854: \end{thebibliography}
855:
856:
857:
858: \end{document}
859:
860: