1: % "VERSION 1, last correction: November 23, 2006 by PK"
2: \tolerance = 10000
3:
4: %
5: % take this for final format:
6: %\documentclass[twocolumn,showpacs,pra,amsmath,amssymb,floatfix]{revtex4}
7: \documentclass[twocolumn,showpacs,prl,amsmath,amssymb,floatfix]{revtex4}
8: %\documentclass[twocolumn,showpacs,prb,amsmath,amssymb,floatfix,eqsecnum]{revtex4}
9: %
10: % take this this for submission in preprint style
11: %\documentclass[preprint,showpacs,prb,amsmath,amssymb,floatfix,eqsecnum]{revtex4}
12: %\documentclass[preprint,showpacs,prb,amsmath,amssymb,floatfix]{revtex4}
13: %\documentclass[twocolumn,showpacs,superscriptaddress,prb,amsmath,amssymb,floatfix,eqsecnum]{revtex4}
14: %
15: % for one-column latex take this:
16: %\documentclass[galley,showpacs,prl,amsmath,amssymb]{revtex4}
17: %
18: %\usepackage{dcolumn}
19:
20:
21: \usepackage{amsmath,amssymb}
22: \usepackage{bm}
23: \usepackage{epsfig}
24: %\usepackage{psfrag}
25:
26: % set \bd to \bf or \bm
27: \newcommand{\bd}{\bm}
28:
29: \begin{document}
30:
31: \title{
32: Low-density expansion for the two-dimensional electron gas
33: }
34:
35: \author{Francesca Sauli and Peter Kopietz}
36:
37: \affiliation{Institut f\"{u}r Theoretische Physik, Universit\"{a}t
38: Frankfurt, Max-von-Laue Strasse 1, 60438 Frankfurt, Germany}
39:
40: % \date{\today}
41: \date{August 23, 2006}
42:
43:
44:
45: \begin{abstract}
46:
47:
48: We show that in two dimensions ($2D$) a systematic expansion of the self-energy and the
49: effective interaction
50: of the dilute electron gas
51: in powers of the two-body $T$-matrix $T_0$
52: can be generated from the exact hierarchy of
53: functional renormalization group
54: equations for the one-particle irreducible vertices using the
55: chemical potential as flow parameter.
56: Due to the interference of particle-particle and particle-hole
57: channels at order $T_0^2$, in $2D$
58: the ladder approximation
59: for the self-energy is not reliable
60: beyond the leading order in $T_0$.
61: We also discuss two-body scattering in vacuum
62: in arbitrary dimensions from the renormalization
63: group point of view and argue that
64: the singular interaction proposed by
65: Anderson [Phys. Rev. Lett. {\bf{65}}, 2306 (1990)] cannot be ruled
66: out on the basis of the ladder approximation.
67:
68:
69:
70: % We also rewrite the Lippmann-Schwinger equation for $T_0$
71: % as a renormalization group flow equation
72: % and show that in dimensions $D<2$
73: % is has a non-Gaussian fixed point
74: % characterized by a finite $s$-wave phase shift
75: % $\delta_0 = - \pi ( 2 - D )/2$.
76:
77:
78:
79:
80:
81:
82: \end{abstract}
83:
84: \pacs{71.10.Ca, 71.10.Hf, 73.43.Nq}
85:
86:
87: %\preprint{}
88:
89: %\draft
90:
91: \maketitle
92:
93:
94:
95:
96: Calculating the physical properties of strongly interacting electrons
97: in two spatial dimensions ($2D$) remains one of the big challenges of condensed
98: matter theory.
99: Most authors agree that the normal state
100: of interacting electrons in $2D$
101: is a Fermi liquid, at least for weak interactions.
102: There are even rigorous proofs that
103: at weak coupling certain two-dimensional models for interacting
104: electrons (including the repulsive
105: Hubbard model away from half filling) are Fermi liquids above
106: an energy scale which is non-perturbative
107: in the coupling constant \cite{Disertori00}.
108: However, due to a lack of controlled methods,
109: one cannot exclude the possibility that
110: for sufficiently strong interactions
111: Fermi liquid theory breaks down in~$2D$.
112: Anderson has argued repeatedly \cite{Anderson90,Anderson01}
113: that due to non-perturbative effects
114: which are neglected by the usual field-theoretical machinery of many-body theory,
115: the normal state of the two-dimensional Hubbard model
116: is not a Fermi liquid for arbitrary strength of the interaction.
117: This scenario has been criticized early on by
118: Engelbrecht and Randeria \cite{Engelbrecht90},
119: who pointed out that a calculation
120: of the self-energy $\Sigma ( {\bd{k}} , \omega )$
121: within the ladder approximation (LAP)
122: predicts Fermi liquid behavior. However,
123: the LAP has some rather peculiar and most likely unphysical
124: features; in particular, the limiting behavior of $\Sigma ( {\bd{k}} , \omega )$
125: for small frequencies $\omega$ and for wave-vectors ${\bd{k}}$ in the vicinity
126: of the Fermi surface ${\bd{k}}_F$
127: is very sensitive
128: to the order in which the limits $\omega \rightarrow 0$ and
129: ${\bd{k}} \rightarrow {\bd{k}}_F$ are taken \cite{Fukuyama90,Halboth98}.
130:
131:
132: Unfortunately, going beyond the LAP is very difficult, because the
133: particle-hole scattering channels
134: have to be taken into account on the same footing with
135: the particle-particle scattering
136: channel retained within the LAP, i.e.,
137: one has to solve coupled Bethe-Salpeter equations in several
138: channels.
139: In this work we shall take a fresh look at this problem using a
140: formally exact hierarchy for renormalization group (RG) flow
141: equations for the one-particle irreducible vertices~\cite{Kopietz01}.
142:
143:
144:
145: {\it{Functional RG in the spin-singlet channel.}}
146: Starting point of our analysis are the formally exact RG flow equations
147: for the irreducible self-energy $\Sigma_{\Lambda} ( K )$ and the
148: antisymmetrized
149: effective interaction vertex
150: $\Gamma^{(4)}_{\Lambda} ( K_1^{\prime} \sigma_1^{\prime} ,
151: K_2^{\prime} \sigma_2^{\prime} ; K_2 \sigma_2 , K_1 \sigma_1 )$.
152: Here $K = ( {\bd{k}} , i \omega )$ denotes wave-vector and
153: Matsubara frequency, and $\sigma_i $ are the spin projections.
154: $\Lambda$ is a cutoff scale separating the
155: low-energy from the high-energy degrees of freedom, such that
156: for $\Lambda = \Lambda_0$ the low-energy fluctuations are suppressed.
157: For the Hubbard model on a $D$-dimensional hyper-cubic lattice with lattice spacing $a$
158: and on-site repulsion $U$ the bare interaction is
159: $
160: \Gamma^{(4)}_{\Lambda_0} =
161: ( \delta_{ \sigma_1 , \sigma_1^{\prime} } \delta_{ \sigma_2 , \sigma_2^{\prime} }
162: -
163: \delta_{ \sigma_1 , \sigma_2^{\prime} } \delta_{ \sigma_2 , \sigma_1^{\prime} } ) \tilde{U}_0
164: $, where $\tilde{U}_0 = a^D U$.
165: As we iterate the RG, the flowing effective interaction develops a spin-triplet
166: component, so that in general
167: \begin{eqnarray}
168: \Gamma^{(4)}_{\Lambda} ( K_1^{\prime} \sigma_1^{\prime} ,
169: K_2^{\prime} \sigma_2^{\prime} ; K_2 \sigma_2 , K_1 \sigma_1 ) & &
170: \nonumber
171: \\
172: & & \hspace{-51mm} =
173: ( \delta_{ \sigma_1 , \sigma_1^{\prime} } \delta_{ \sigma_2 , \sigma_2^{\prime} }
174: -
175: \delta_{ \sigma_1 , \sigma_2^{\prime} } \delta_{ \sigma_2 , \sigma_1^{\prime} } )
176: \Gamma_{\Lambda}^{s} ( K_1^{\prime} ,
177: K_2^{\prime} ; K_2 , K_1 )
178: \nonumber
179: \\
180: & & \hspace{-50mm} +
181: ( \delta_{ \sigma_1 , \sigma_1^{\prime} } \delta_{ \sigma_2 , \sigma_2^{\prime} }
182: +
183: \delta_{ \sigma_1 , \sigma_2^{\prime} } \delta_{ \sigma_2 , \sigma_1^{\prime} } )
184: \Gamma_{\Lambda}^{t} ( K_1^{\prime} ,
185: K_2^{\prime} ; K_2 , K_1 )
186: \; .
187: \end{eqnarray}
188: The singlet part
189: $\Gamma_{\Lambda}^{s} ( K_1^{\prime} ,
190: K_2^{\prime} ; K_2 , K_1 )$ is symmetric under the exchange $K_1^{\prime}
191: \leftrightarrow K_2^{\prime}$ or $K_1
192: \leftrightarrow K_2$, while the
193: triplet part
194: $\Gamma_{\Lambda}^{t} ( K_1^{\prime} ,
195: K_2^{\prime} ; K_2 , K_1 )$
196: is antisymmetric and therefore vanishes for $K_1^{\prime} = K_2^{\prime}$ or
197: $K_1 = K_2$.
198: The RG flow equation for $\Gamma^{(4)}_{\Lambda}$
199: contains also mixed terms $\Gamma^{t}_{\Lambda} \Gamma^{s}_{\Lambda}$ and
200: a term involving the
201: six-point vertex $\Gamma_\Lambda^{(6)}$.
202: However, $\Gamma^{t}_{\Lambda}$ and $\Gamma_\Lambda^{(6)}$
203: are both irrelevant
204: in the RG sense \cite{Kopietz01}, so that we shall neglect them.
205: Within this approximation, the
206: RG flow of the self-energy is determined by
207: \begin{equation}
208: \partial_{\Lambda} \Sigma_{\Lambda} ( K ) = \int_{ K^{\prime}}
209: \dot{G}_{\Lambda} ( K^{\prime} ) \Gamma_{\Lambda}^{s}
210: ( K , K^{\prime} ; K^{\prime} , K )
211: \; ,
212: \label{eq:flowself}
213: \end{equation}
214: and the effective interaction satisfies at zero temperature
215: \begin{widetext}
216: \begin{eqnarray}
217: \partial_{\Lambda} \Gamma^s_{\Lambda}
218: ( K_1^{\prime} , K_2^{\prime} ; K_2 , K_1 )
219: & = & - \int_K [ \dot{G}_{\Lambda} ( K ) G_{\Lambda} ( K_1 + K_2 - K )
220: + {G}_{\Lambda} ( K ) \dot{G}_{\Lambda} ( K_1 + K_2 - K ) ]
221: \nonumber
222: \\
223: & & \hspace{5mm} \times
224: \Gamma^s_{\Lambda} ( K_1^{\prime} , K_2^{\prime} ; K_1 + K_2 - K , K )
225: \Gamma^s_{\Lambda} ( K, K_1 + K_2 - K ; K_2 , K_1 )
226: \nonumber
227: \\
228: &- & \frac{1}{2} \int_K
229: \Bigl\{ [
230: \dot{G}_{\Lambda} ( K ) G_{\Lambda} ( K + K_1 - K_1^{\prime} )
231: +
232: {G}_{\Lambda} ( K ) \dot{G}_{\Lambda} ( K + K_1 - K_1^{\prime} ) ]
233: \nonumber
234: \\
235: & & \hspace{5mm} \times
236: \Gamma^s_{\Lambda} ( K_1^{\prime} , K + K_1 - K_1^{\prime} ; K , K_1 )
237: \Gamma^s_{\Lambda} ( K_2^{\prime} , K ; K + K_1 - K_1^{\prime} , K_2 )
238: + ( K_1^{\prime} \leftrightarrow K_2^{\prime} )
239: \Bigr\}
240: % \nonumber
241: % \\
242: % & & - \frac{1}{2} \int_K [
243: % \dot{G}_{\Lambda} ( K ) G_{\Lambda} ( K + K_1 - K_2^{\prime} )
244: % +
245: % {G}_{\Lambda} ( K ) \dot{G}_{\Lambda} ( K + K_1 - K_2^{\prime} ) ]
246: % \nonumber
247: % \\
248: % & & \hspace{8mm} \times
249: % \Gamma^s_{\Lambda} ( K_2^{\prime} , K + K_1 - K_2^{\prime} ; K , K_1 )
250: % \Gamma^s_{\Lambda} ( K_1^{\prime} , K ; K + K_1 - K_2^{\prime} , K_2 )
251: \; ,
252: \label{eq:flowvertex}
253: \end{eqnarray}
254: \end{widetext}
255: where
256: $\int_{K} = \int \frac{ d^D k }{ ( 2 \pi )^D} \frac{ d \omega}{ 2 \pi } $, and
257: we have introduced
258: the cutoff-dependent propagator
259: \begin{equation}
260: G_{\Lambda} ( K ) = \Theta_{\Lambda} ( {\bd{k}} ) [ i \omega -
261: \xi_{ \bd{k}} - \Sigma_{\Lambda} ( K ) ]^{-1}
262: \; ,
263: \end{equation}
264: and the single scale propagator
265: \begin{equation}
266: \dot{G}_{\Lambda} ( K ) =
267: [ 1 + G^0_{\Lambda} ( K ) \Sigma_{\Lambda} ( K ) ]^{-2} \partial_{\Lambda} G^0_{\Lambda} ( K )
268: \; .
269: \end{equation}
270: Here $\Theta_{\Lambda} ( {\bd{k}} )$ is some cutoff function which suppresses the
271: low-energy modes, normalized such that $\Theta_0 ( {\bd{k}} ) =1$ and
272: $\Theta_{\infty} ( {\bd{k}} ) =0$.
273: The cutoff-dependent non-interacting propagator is
274: $G_{\Lambda}^0 ( K ) = \Theta_{\Lambda} ( {\bd{k}} ) [ i \omega -
275: \xi_{\bd{k}} ]^{-1} $, with
276: $\xi_{\bd{k}} = \epsilon_{\bd{k}} - \mu$, where
277: $\epsilon_{\bd{k}}$ is the bare energy dispersion and $\mu$ is the chemical potential.
278: The first term on the right-hand side
279: of Eq.~(\ref{eq:flowvertex}) is
280: the contribution from the particle-particle channel, while the other two terms
281: correspond to the particle-hole channels (also called zero-sound channels).
282:
283:
284: {\it{Ladder approximation.}}
285: In the limit of vanishing density only the particle-particle channel contributes
286: to the effective interaction. At low densities,
287: it is then reasonable to calculate the self-energy in LAP,
288: where the contribution of the particle-hole channels
289: to the effective interaction
290: in Eq.~(\ref{eq:flowvertex}) is simply ignored.
291: It is then advantageous to introduce
292: total and relative energy-momenta as independent variables,
293: $P = K_1 + K_2 = K_1^{\prime} + K_2^{\prime}$,
294: $Q = ( K_1 - K_2 )/2$, $Q^{\prime} = (K_1^{\prime} - K_2^{\prime} )/2$,
295: and define
296: \begin{equation}
297: \Gamma^s_{\Lambda} ( \frac{P}{2} + Q^{\prime} , \frac{P}{2} - Q^{\prime} ;
298: \frac{P}{2} -Q , \frac{P}{2} + Q ) = \Gamma_{\Lambda} ( Q^{\prime} , Q ; P )
299: \; .
300: \label{eq:gammaQQP}
301: \end{equation}
302: With this approximation Eq.~(\ref{eq:flowvertex}) reduces to
303: \begin{eqnarray}
304: \partial_{\Lambda} \Gamma_{\Lambda} ( Q^{\prime} , Q ; P )
305: & = & - 2 \int_{ K } \dot{G}_{\Lambda} ( \frac{P}{2} + K )
306: {G}_{\Lambda} ( \frac{P}{2} - K )
307: \nonumber
308: \\
309: & & \times \Gamma_{\Lambda} ( Q^{\prime} , K ; P )
310: \Gamma_{\Lambda} ( K , Q ; P )
311: \; .
312: \label{eq:gammaT}
313: \end{eqnarray}
314: Assuming that the bare interaction is independent of the
315: frequency-part of $Q$ and $Q^{\prime}$, this remains also true for the
316: renormalized interaction, so that we may write
317: $ \Gamma_{\Lambda} ( Q^{\prime} , Q ; P ) =
318: \Gamma_{\Lambda} ( {\bd{q}}^{\prime} , {\bd{q}} ; P )$.
319: If we replace the flowing Green functions in Eq.~(\ref{eq:gammaT})
320: by the non-interacting ones, the frequency sum is easily carried out,
321: \begin{eqnarray}
322: \partial_{\Lambda} \Gamma_{\Lambda} ( {\bd{q}}^{\prime} , {\bd{q}} ; P )
323: & = & \int \frac{ d^D k }{ (2 \pi )^D}
324: \partial_{\Lambda} [ \Theta_{\Lambda} ( \frac{ \bd{p} }{2} + {\bd{k}} )
325: {\Theta}_{\Lambda} ( \frac{ \bd{p} }{2} - {\bd{k}} ) ]
326: \nonumber
327: \\
328: & & \hspace{-27mm} \times
329: \frac{ 1 - f ( \xi_{ \frac{ \bd{p}}{2} + {\bd{k}} } ) - f ( \xi_{ \frac{ \bd{p}}{2} - {\bd{k}} } ) }{
330: i \bar{\omega} - \xi_{ \frac{\bd{p}}{2} + \bd{k} } - \xi_{ \frac{\bd{p}}{2} - \bd{k} } }
331: \Gamma_{\Lambda} ( {\bd{q}}^{\prime} , {\bd{k}} ; P )
332: \Gamma_{\Lambda} ( {\bd{k}} , {\bd{q}} ; P )
333: \; ,
334: \label{eq:gammaT2}
335: \end{eqnarray}
336: where $P = (\bd{p} , i \bar{\omega} )$, and
337: $ f ( \xi ) = \Theta ( - \xi )$ is the Fermi function at zero temperature.
338: Eq.~(\ref{eq:gammaT2}) is equivalent with the
339: Bethe-Salpeter equation for the effective interaction in the
340: particle-particle channel \cite{Fetter71}.
341: We recover the LAP if we assume that
342: $\Gamma_{\Lambda} ( {\bd{q}}^{\prime} , {\bd{q}} ; P ) = \Gamma_{\Lambda} ( P )$
343: is independent of ${\bd{q}}$ and ${\bd{q}}^{\prime}$, in which case
344: Eq.~(\ref{eq:gammaT2}) becomes an ordinary differential equation,
345: $
346: \partial_{\Lambda} \Gamma_{\Lambda} ( P ) = \dot{\chi}_{\Lambda} ( P )
347: \Gamma^2_{\Lambda} ( P )
348: $,
349: where
350: \begin{eqnarray}
351: \dot{\chi}_{\Lambda} ( P ) & = &
352: \int \frac{ d^D k }{ (2 \pi )^D}
353: \partial_{\Lambda} [ \Theta_{\Lambda} ( \frac{ \bd{p} }{2} + {\bd{k}} )
354: {\Theta}_{\Lambda} ( \frac{ \bd{p} }{2} - {\bd{k}} ) ]
355: \nonumber
356: \\
357: & & \times
358: \frac{ 1 - f ( \xi_{ \frac{ \bd{p}}{2} + {\bd{k}} } ) - f ( \xi_{ \frac{ \bd{p}}{2} - {\bd{k}} } ) }{
359: i \bar{\omega} - \xi_{ \frac{\bd{p}}{2} + \bd{k} } - \xi_{ \frac{\bd{p}}{2} - \bd{k} } }
360: \; .
361: \label{eq:dotchidef}
362: \end{eqnarray}
363: The solution of the above differential equation with initial condition
364: $ \Gamma_{\Lambda_0} ( P ) = \tilde{U}_0$
365: yields the usual LAP
366: for the effective interaction,
367: \begin{equation}
368: \Gamma ( P ) = \Gamma_{ \Lambda =0} ( P ) = \tilde{U}_0 [ 1 + \tilde{U}_0 \chi ( P ) ]^{-1}
369: \; ,
370: \label{eq:Tmanybody}
371: \end{equation}
372: where the pair susceptibility
373: $\chi ( P ) = \int_0^{\Lambda_0} d \Lambda \dot{\chi}_{\Lambda} ( P )$
374: is given by
375: \begin{eqnarray}
376: \chi ( P )
377: & = & -
378: \int \frac{ d^D k }{ (2 \pi )^D}
379: \left[ 1 - \Theta_{ \Lambda_0} ( \frac{ \bd{p} }{2} + {\bd{k}} )
380: {\Theta}_{\Lambda_0} ( \frac{ \bd{p} }{2} - {\bd{k}} ) \right]
381: \nonumber
382: \\
383: & & \hspace{16mm} \times
384: \frac{ 1 - f ( \xi_{ \frac{ \bd{p}}{2} + {\bd{k}} } ) - f ( \xi_{ \frac{ \bd{p}}{2} - {\bd{k}} } ) }{
385: i \bar{\omega} - \xi_{ \frac{\bd{p}}{2} + \bd{k} } - \xi_{ \frac{\bd{p}}{2} - \bd{k} } }
386: \; .
387: \label{eq:chires}
388: \end{eqnarray}
389:
390:
391:
392: {\it{Zero density limit.}}
393: In this limit $\mu \rightarrow 0$ and we may approximate
394: $\epsilon_{\bd{k}} = {\bd{k}}^2/(2 m)$, where $m$ is some effective band mass.
395: The Fermi functions $f ( \xi_{ \frac{\bd{p}}{2} \pm {\bd{k}} } ) $ then vanish.
396: Using for simplicity a sharp cutoff in momentum space,
397: $\Theta_{\Lambda} ( {\bd{k}} ) = \Theta ( | {\bd{k}}| - \Lambda )$, we
398: find from Eq.~(\ref{eq:gammaT2}) that
399: the two-body $T$-matrix
400: $
401: T_{\Lambda} ( \bd{q}^{\prime} ,
402: \bd{q} ; i \bar{\omega} ) \equiv \Gamma_{\Lambda} (\bd{q}^{\prime} ,
403: \bd{q} ; {\bd{p}}=0, i \bar{\omega} )
404: $ in vacuum
405: satisfies the flow equation
406: \begin{eqnarray}
407: \partial_{\Lambda} T_{ \Lambda} ( \bd{q}^{\prime} ,
408: \bd{q} , i \bar{\omega} ) & = & - \int \frac{ d^D k}{ ( 2 \pi )^D}
409: \frac{ \delta ( | {\bd{k}} | - \Lambda ) }{ i \bar{\omega} - \Lambda^2/m }
410: \nonumber
411: \\
412: & & \times
413: T_{\Lambda} ( \bd{q}^{\prime} , \bd{k} , i \bar{\omega} )
414: T_{\Lambda} ( \bd{k} , \bd{q}, i \bar{\omega} )
415: \; .
416: \label{eq:LSRG}
417: \end{eqnarray}
418: The initial condition is $ T_{ \Lambda_0} ( \bd{q}^{\prime} ,
419: \bd{q} , i \bar{\omega} ) = \tilde{U}_{ \bd{q}^{\prime} - \bd{q} }$,
420: where $\tilde{U}_{ {\bd{k}} }$ is the Fourier transform of the bare interaction.
421: Eq.~(\ref{eq:LSRG}) is the RG version of the Lippmann-Schwinger
422: equation for the $T$-matrix of
423: elementary scattering theory~\cite{Pethick02}.
424: The low-energy behavior of the $T$-matrix defines the
425: $s$-wave scattering length $a_s$ via
426: $T ( 0,0 , i 0 ) = \gamma_D a_s^{D-2} / m$, where
427: $\gamma_D$ is a numerical constant ($\gamma_3 = 4 \pi$).
428: Introducing the dimensionless coupling constant
429: $u_l = m K_D \Lambda^{D-2} T_{\Lambda} ( 0,0 , i0) = K_D \gamma_D
430: ( a_s \Lambda )^{D-2}$, where $l = \ln ( \Lambda_0 / \Lambda )$ and
431: $K_D = 2^{1-D} \pi^{ -D/2} / \Gamma ( D/2) $ is the surface area of the
432: $D$-dimensional unit sphere divided by $(2 \pi )^D$,
433: we obtain from Eq.~(\ref{eq:LSRG}) the RG flow equation
434: \begin{equation}
435: \partial_l u_l = \epsilon u_l - u_l^2
436: \; , \; \epsilon = 2 - D
437: \; .
438: \label{eq:uflow}
439: \end{equation}
440: The corresponding flow of $u_l$ is shown in Fig.~\ref{fig:uflow}.
441: %
442: %
443: \begin{figure}[tb]
444: \centering
445: % \psfrag{dg2}{$D \geq 2$}
446: % \psfrag{dk2}{$D < 2$}
447: % \psfrag{d2}{$D = 2$}
448: % \psfrag{u}{$u_{\ast}$}
449: \epsfig{file=fig1.eps,width=65mm}
450: \vspace{-4mm}
451: \caption{%
452: (Color online) RG flow of the rescaled zero energy part
453: $u_l = m K_D \Lambda^{D-2} T_{\Lambda} ( 0,0 , i0) $ of the two-body $T$-matrix.
454: }
455: \label{fig:uflow}
456: \end{figure}
457: %
458: For $D > 2$ the RG flow has a stable Gaussian fixed point G
459: at vanishing interaction, and an unstable fixed point UF at finite negative interaction
460: $u_{\ast} = \epsilon < 0$.
461: At this point the zero energy $T$-matrix diverges for $\Lambda \rightarrow 0$,
462: $ T_{\Lambda} ( 0 ,0 , i0 ) \propto u_{\ast} \Lambda^{2 - D} \rightarrow - \infty$,
463: so that UF
464: corresponds to the unitary Fermi gas
465: in the limit of vanishing density~\cite{Ho04}.
466: For $D=2$ both fixed points UF and G merge into a single point
467: UFG at zero interaction.
468: Finally, for $ D < 2$ the unitary fixed point UF
469: has zero interaction,
470: because in this case an infinitesimal attractive interaction leads to a zero energy
471: bound state \cite{Nishida06}.
472: However, there is a new non-Gaussian fixed point NG for
473: finite repulsive interaction $u_{\ast} = \epsilon >0$, which
474: resembles the Wilson-Fisher fixed point
475: below four dimensions \cite{Wilson72}.
476: We now show that this fixed point is characterized by a finite phase shift
477: $\delta_0 (0) = - \pi\epsilon /2$ for $s$-wave scattering at zero energy.
478: Setting again
479: $\tilde{U}_{ \bd{k} } = \tilde{U}_0$,
480: the solution of Eq.~(\ref{eq:LSRG}) is
481: \begin{equation}
482: T_0 ( i \bar{\omega } ) = T_{ \Lambda =0} ( \bd{q}^{\prime} , \bd{q} , i \bar{\omega} )
483: = \tilde{U}_0 [ 1 + \tilde{U}_0 \chi_0 ( i \bar{\omega} ) ]^{-1}
484: \; ,
485: \label{eq:T2}
486: \end{equation}
487: where from Eq.~(\ref{eq:chires}) we obtain
488: \begin{eqnarray}
489: \chi_0 ( i \bar{\omega})
490: & = & -
491: \int \frac{ d^D k }{ (2 \pi )^D}
492: \frac{ \Theta ( \Lambda_0 - | {\bd{k}} | ) }{
493: i \bar{\omega} - \bd{k}^2 /m }
494: \; .
495: \label{eq:chires0}
496: \end{eqnarray}
497: For $D < 2 $ the integral is convergent for $\Lambda_0 \rightarrow \infty$,
498: and we obtain for real frequencies,
499: \begin{eqnarray}
500: \chi_{0} ( \omega + i 0 ) & = & \pi \nu_D ( | \omega | , m_r)
501: \bigl\{ \Theta ( - \omega ) [ \sin ( \pi \epsilon /2 )]^{-1}
502: \nonumber
503: \\
504: & &
505: + \Theta ( \omega ) \left[ \cot ( \pi \epsilon /2 ) + i \right] \bigr\}
506: \; ,
507: \end{eqnarray}
508: where $\nu_D ( \omega , m_r ) = K_D m_r (2 m_r \omega )^{ - \epsilon /2}$
509: is the density of states of a free particle with reduced mass $m_r = m/2$.
510: In $2D$, where the density of states
511: $\nu_2 ( m ) = m / (2 \pi )$ is constant, we have
512: \begin{eqnarray}
513: \chi_{0} ( \omega + i 0 ) = \pi \nu_2 ( m_r) \left[
514: \ln \left( { E_0}/{| \omega | } \right) + i \Theta ( \omega )
515: \right]
516: \; ,
517: \end{eqnarray}
518: where $E_0 = \Lambda_0^2/ m$.
519: The phase shift for $s$-wave scattering in the spin-singlet channel, defined
520: via~\cite{Engelbrecht90,Nozieres85}
521: $ T_0 ( \omega + i 0 ) = - | T_0 ( \omega + i 0 )| e^{ i \delta_0 ( \omega ) }$,
522: is for $\omega > 0$ given by
523: \begin{equation}
524: \delta_0 ( \omega )
525: = - \arctan \bigl\{ \pi g ( \omega ) /[1 + \pi g ( \omega ) \cot( \pi \epsilon /2)]
526: \bigr\} \; ,
527: \end{equation}
528: where
529: $g ( \omega ) = \nu_D ( \omega , m_r ) \tilde{U}_0$ is the relevant dimensionless
530: coupling. Because $g ( 0 ) = \infty$ for $D < 2$ due to the
531: divergence of the density of states, we conclude
532: that $\delta_0 ( 0 ) = - \pi \epsilon /2$ for any finite $\tilde{U}_0 >0$.
533:
534: {\it{Low-density expansion in $2D$.}}
535: If we approximate
536: the effective interaction in Eq.~(\ref{eq:flowself})
537: by $\Gamma ( P )$ given in Eq.~(\ref{eq:Tmanybody})
538: and ignore self-energy corrections to the propagators,
539: we obtain the usual LAP for the self-energy~\cite{Engelbrecht90,Fukuyama90}.
540: However, the LAP is {\it{not}} a systematic low-density expansion.
541: %This
542: %is obvious from the fact that
543: %the effective interaction $\Gamma ( P )$ in Eq.~(\ref{eq:Tmanybody})
544: %depends on $\mu $ in a rather complicated and non-linear manner.
545: In $2D$ the relevant dimensionless
546: low-density parameter is $ \alpha = [ \ln ( \Lambda_0/ k_F ) ]^{-1} $.
547: One should therefore expand the LAP in powers of $\alpha$.
548: The self-energy within LAP up to order $\alpha^2$
549: has been evaluated by
550: Bloom~\cite{Bloom75} using
551: the pseudo-potential method pioneered by Galitskii~\cite{Galitskii58,Fetter71}.
552: In this method one first derives a non-linear
553: integral equation relating the
554: effective interaction in LAP to the on-shell two-body $T$-matrix in vacuum.
555: Iteration of this integral equation generates the desired
556: expansion in powers of $\alpha$.
557:
558: The RG equations (\ref{eq:flowself}) and (\ref{eq:flowvertex})
559: offer an alternative way to generate the low-density expansion in $2D$.
560: The advantage of the RG approach is that the contribution from the
561: particle-hole channels neglected in LAP can be systematically included.
562: The crucial point is that Eqs.~(\ref{eq:flowself}) and (\ref{eq:flowvertex})
563: remain formally valid
564: if we take the derivative with respect to any parameter
565: appearing only in the free propagator \cite{Kopietz01}.
566: In particular, we may choose $\Lambda = \mu$, in which case
567: Eq.~(\ref{eq:flowself}) and (\ref{eq:flowvertex}) describe the evolution
568: of the self-energy and the effective interaction as we change
569: the chemical potential.
570: The propagator in the flow equations should then be replaced by
571: $
572: G_{\mu} ( K ) = [ i \omega - \epsilon_{\bd{k}} + \mu
573: - \Sigma_{\mu} ( K ) ]^{-1}
574: $,
575: and the single scale propagator by
576: $\dot{G}_{\mu} ( K ) = - G^2_{\mu} ( K )$.
577: The flow equation for the self-energy now reads
578: \begin{equation}
579: \partial_{\mu} \Sigma_{\mu} ( K ) = \int_{ P}
580: \dot{G}_{\mu} ( P-K) \Gamma_{\mu}
581: ( \frac{P}{2} -K, \frac{P}{2} -K ; P )
582: \; ,
583: \label{eq:flowself2}
584: \end{equation}
585: and the flow equation for the effective interaction
586: can be obtained by making analogous substitutions
587: in Eq.~(\ref{eq:flowvertex}).
588: Assuming that an expansion in
589: powers of $\mu$ is possible,
590: we may generate this expansion
591: by successive differentiation of
592: Eq.~(\ref{eq:flowself2}) and the analogue of Eq.~(\ref{eq:flowvertex}).
593: For $\tilde{U}_{\bd{k}} = \tilde{U}_0 $ the initial conditions at $\mu=0$ are
594: $\Sigma_{\mu=0} ( K ) =0$ and
595: $\Gamma_{\mu =0}
596: ( \frac{P}{2} -K, \frac{P}{2} -K ; P ) =
597: T_0 ( i \bar{\omega} - \frac{ {\bd{p}}^2 }{ 4 m} ) \equiv T ( P ) $.
598: The leading term in the expansion of the self-energy is
599: \begin{equation}
600: \Sigma_{\mu} ( K ) = \mu \int_{ P}
601: \dot{G}_{0} ( P-K) T ( P ) + O ( \mu^2 )
602: \; ,
603: \label{eq:selfexpansion}
604: \end{equation}
605: and from
606: Eqs.~(\ref{eq:flowvertex}) and (\ref{eq:gammaQQP}) we obtain for the
607: effective interaction
608: \begin{eqnarray}
609: \Gamma_{\mu} ( Q^{\prime} , Q ; P ) & = & T ( P )
610: - \mu T^2 ( P )
611: [ \partial_{\mu} \chi ( P ) ]_{\mu =0}
612: %\int_K \partial_{\Lambda} [ {G}_0 ( K ) G_0 ( P -K )]_{ \Lambda =0}
613: \nonumber
614: \\
615: & & \hspace{-25mm} - \frac{\mu}{2}
616: \int_K \Bigl\{
617: [ \dot{G}_0 ( K ) G_0 ( K + Q + Q^{\prime} )
618: \nonumber
619: \\
620: & & \hspace{-22mm}
621: + {G}_0 ( K ) \dot{G}_0 ( K + Q + Q^{\prime} )
622: ] T ( K + Q + \frac{P}{2} )
623: \nonumber
624: \\
625: & & \hspace{-22mm} \times T ( K + {Q}^{\prime} + \frac{P}{2} )
626: + ( Q^{\prime} \rightarrow - Q^{\prime} )
627: \Bigr\} + O ( \mu^2 )
628: ,
629: \label{eq:vertexexpansion}
630: \end{eqnarray}
631: where the pair susceptibility
632: $\chi ( P ) $
633: at finite $\mu$ is given in Eq.~ (\ref{eq:chires}).
634: The second term on the right-hand side of
635: Eq.~(\ref{eq:vertexexpansion}) is due to the
636: $\mu$-dependence of the effective interaction in LAP, while
637: the last term is the contribution from the particle-hole channels neglected in LAP.
638:
639: Due to the
640: non-analytic $\mu$-dependence of the density of states for $D \neq 2$
641: the above expansion in powers of $\mu$
642: is only possible in $D=2$.
643: To see this more clearly, let us
644: explicitly evaluate the first term on the right-hand side of
645: Eq.~(\ref{eq:selfexpansion}). The result can be written as
646: \begin{equation}
647: \Sigma_{\mu} ( {\bd{k}} , i \omega ) \approx \rho_0 (\mu )
648: T_0 ( i \omega - \frac{\bd{k}^2 }{4m} )
649: \; ,
650: \label{eq:sigmares}
651: \end{equation}
652: where $\rho_0 ( \mu ) = \mu \nu_2 ( m )$.
653: In fact, Eq.~(\ref{eq:sigmares}) is
654: the leading term in the expansion in powers
655: of $T_0$ for arbitrary $D$
656: if we identify $ \rho_0 ( \mu ) \equiv
657: (2/D) \mu \nu_D ( \mu , m ) \propto \mu^{D/2}$
658: with the density (per spin projection) in the absence of interactions.
659: The non-analyticity in $\mu$ for $D \neq 2$ is
660: obvious.
661:
662: If we approximate $T_0$ by the bare interaction $\tilde{U}_0$,
663: then Eq.~(\ref{eq:sigmares}) reduces to
664: the Fock correction to the self-energy.
665: In contrast to the usual
666: LAP~\cite{Engelbrecht90,Fukuyama90},
667: the particle-particle
668: susceptibility is evaluated at vanishing density in Eq.~(\ref{eq:sigmares}).
669: It is instructive to explicitly calculate the
670: corresponding single-particle Green function
671: $G ( {\bd{k}} , i\omega )$ for $ D \leq 2$.
672: The relation between $\mu$ and Fermi momentum $k_F$
673: is then $\frac{k_F^2 }{2m} + \rho_0 ( \mu ) T_0 ( - \frac{k_F^2}{4m}) = \mu$.
674: Expanding $G^{-1} ( {\bd{k}} , i \omega )$
675: for small $\omega$ and $|{\bd{k}} | - k_F$
676: we obtain the Fermi liquid form
677: \begin{equation}
678: G ( {\bd{k}} , i \omega ) \approx Z [ i \omega - (k_F / m_{\ast})
679: ( | {\bd{k}} | - k_F ) ]^{-1}
680: \label{eq:QP}
681: \; ,
682: \end{equation}
683: with quasi-particle residue $Z = 1 - \epsilon^2 g^2 [\epsilon + g ]^{-2} $
684: and effective mass renormalization
685: $m/m_{\ast} = 1 - \frac{1}{2} \epsilon^2 g^2 [ \epsilon + g ]^{-2}$.
686: Here $g = \nu_D ( \mu, m_r ) \tilde{U}_0$, and in $2D$
687: we should replace $\epsilon \rightarrow \alpha = [ \ln ( \Lambda_0/k_F ) ]^{-1}$.
688: For $g \rightarrow \infty$
689: we obtain $Z \approx 1 - \epsilon^2$ and $m/m_{\ast} = 1 - \epsilon^2/2$.
690: This agrees qualitatively with the results by Bloom \cite{Bloom75}, who obtained
691: different numerical coefficients on front of the correction terms,
692: which is due to the fact that his approximation amounts to retaining
693: also the second term on the right-hand side of Eq.~(\ref{eq:vertexexpansion}).
694: However, to order $T_0^2$ the particle-hole contributions
695: to Eq.~(\ref{eq:vertexexpansion})
696: compete with the particle-particle channel retained in Ref.~[\onlinecite{Bloom75}],
697: so that the approximation employed by Bloom is inconsistent.
698: In this point we agree with Anderson \cite{Anderson01}, but for different
699: reasons: there is no mathematical mistake in
700: Bloom's analysis, but in $2D$ the LAP is
701: inconsistent beyond the leading order in $T_0$.
702:
703: In summary, we have reconsidered the low-density expansion for the
704: electron gas in dimensions $D \leq 2$ using functional RG methods.
705: At vanishing density,
706: we have rewritten the Lippmann-Schwinger equation for the
707: two-body $T$-matrix in vacuum
708: as a RG flow equation and have pointed out that for $D<2$ this equation has a
709: non-Gaussian fixed point characterized by a finite $s$-wave phase shift.
710: At low densities in $2D$
711: a systematic expansion of the self-energy and of the effective interaction
712: in powers of the two-body $T$-matrix can be generated
713: with the help of the functional RG flow equations for
714: the irreducible vertices using
715: the chemical potential as flow parameter.
716: We have argued that in $2D$ the LAP is inconsistent beyond the leading order
717: in this expansion because already at order $T_0^2$
718: the contribution from the particle-hole channels competes
719: with the particle-particle channel and gives rise to
720: a complicated momentum- and frequency dependence
721: of the effective interaction which still has to be explored.
722: Because the LAP is not reliable in $2D$, one cannot use the LAP
723: to rule out that the true effective interaction in $2D$
724: exhibits a singular dependence on the relative momenta,
725: as proposed by Anderson \cite{Anderson90,Anderson01}.
726:
727:
728: We thank M. Salmhofer for his comments on
729: rigorous results for two-dimensional Fermi systems.
730:
731:
732:
733: \vspace{-4mm}
734: \begin{thebibliography}{99}
735: \vspace{-4mm}
736: %
737: \bibitem{Disertori00}
738: M. Disertori and V. Rivasseau, Phys. Rev. Lett. {\bf{85}}, 361 (2000);
739: %Comm. Math. Phys. {\bf{215}}, 251 and 291 (2000);
740: W. Pedra and M. Salmhofer, in {\it{Proceedings of the 14th International Congress of Mathematical Physics}}, (Lisbon, 2003);
741: G. Benfatto, A. Giuliani, and V. Mastropietro, cond-mat/0507686.
742: %
743: \bibitem{Anderson90}
744: P. W. Anderson, Phys. Rev. Lett. {\bf{64}}, 1839 (1990);
745: {\it{ibid.}} {\bf{65}}, 2306 (1990); {\it{ibid.}} {\bf{66}}, 3226 (1991);
746: {\it{ibid.}} {\bf{71}}, 1220 (1993).
747: %
748: %\bibitem{Anderson97}
749: %P. W. Anderson, {\it{The Theory of Superconductivity
750: %in the High-Tc Cuprates}}, (Princeton University Press, Princeton, 1997).
751: %
752: \bibitem{Anderson01}
753: P. W. Anderson, cond-mat/0101417.
754: %
755: \bibitem{Engelbrecht90}
756: J. R. Engelbrecht and M. Randeria,
757: Phys. Rev. Lett. {\bf{65}}, 1032 (1990);
758: {\it{ibid.}} {\bf{66}}, 3226 (1991).
759: %
760: \bibitem{Fukuyama90}
761: H. Fukuyama and Y. Hasegawa, Prog. Theor. Phys. Suppl. {\bf{101}}, 441 (1990);
762: H. Fukuyama, Y. Hasegawa, and O. Narikito,
763: J. Phys. Soc. Jpn. {\bf{60}}, 2013 (1991).
764: %
765: \bibitem{Halboth98}
766: C. Halboth and W. Metzner, Phys. Rev. B {\bf{57}}, 8873 (1998).
767: %
768: \bibitem{Kopietz01}
769: P. Kopietz and T. Busche, Phys. Rev. B {\bf{64}}, 155101 (2001);
770: M. Salmhofer and C. Honerkamp, Prog. Theor. Phys. {\bf{105}},
771: 1 (2001).
772: %
773: \bibitem{Fetter71}
774: A. L. Fetter and J. D. Walecka, {\it{Quantum Theory of Many-Particle Systems}},
775: (McGraw-Hill, New York, 1971).
776: %
777: \bibitem{Pethick02}
778: See, for example, C. J. Pethick and H. Smith,
779: {\it{Bose-Einstein Condensation in Dilute Gases,}}
780: (Cambridge University Press, Cambridge, 2002), chapter 5.
781: %J. J. Sakurai, {\it{Modern Quantum Mechanics}},
782: %(Addison-Wesley, Reading, 1994).
783: %
784: \bibitem{Ho04}
785: T. Ho, Phys. Rev. Lett. {\bf{92}}, 090402 (2004).
786: %
787: \bibitem{Nishida06}
788: Y. Nishida and D. T. Son, Phys. Rev. Lett. {\bf{97}}, 050403 (2006);
789: cond-mat/0607835.
790: %
791: \bibitem{Wilson72}
792: K. G. Wilson and M. E. Fisher, Phys. Rev. Lett. {\bf{28}}, 240 (1972).
793: %
794: \bibitem{Nozieres85}
795: P. Nozi\`{e}res and S. Schmitt-Rink, J. Low Temp. Phys. {\bf{59}},
796: 195 (1985).
797: %
798: \bibitem{Bloom75}
799: P. Bloom, Phys. Rev. B {\bf{12}}, 125 (1975).
800: %
801: \bibitem{Galitskii58}
802: V. M. Galitskii, Zh. Eksp. Teor. Fiz. {\bf{34}}, 151 (1958)
803: [Sov. Phys. JETP {\bf{7}}, 104 (1958)].
804:
805: % \bibitem{footnotescatt}
806: % For $D \neq 0$ we may define the $s$-wave scattering length $a_s ( q )$
807: % from the
808: % angular average on resonance
809: % $
810: % \langle T_{ \Lambda =0} ( q \hat{{\bd{n}}} , {\bd{q}};
811: % \frac{q^2}{2m_r} + i 0 ) \rangle_{ \hat{{\bd{n}}}} =
812: % \Omega_D a_s^{D-2} / (2 m_r )
813: % $
814: % where $\hat{\bd{n}}$ is a unit vector
815: % and $\langle \ldots \rangle_{ \hat{\bd{n}} }$ denotes angular average
816: % over the directions of $\hat{\bd{n}}$.
817:
818: \end{thebibliography}
819:
820:
821:
822:
823:
824:
825:
826:
827:
828: \end{document}
829:
830:
831:
832:
833:
834:
835:
836:
837: