1: % Thesis chapter on functional Renormalization Group
2:
3:
4: \chapter{Functional RG technique: a short overview}
5: \label{sec:rg}
6: \vspace{2cm}
7: \textit{We review the functional renormalization-group approach for
8: interacting Fermi systems in the 1PI version.
9: Introducing an infrared cutoff $\Lam$ in the free propagator and
10: differentiating the effective action with respect to $\Lam$,
11: an exact hierarchy of differential flow equations for
12: the 1PI vertex functions is derived, describing
13: the gradual evolution from the microscopic model Hamiltonian
14: to the effective action as a function
15: of the continuously decreasing energy cutoff.
16: We briefly discuss the relation to alternative formulations of the fRG
17: approach.}
18: \vspace{.75cm}
19:
20:
21: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
22:
23:
24: \section{Introduction}
25: \label{sec:rg:intro}
26:
27:
28: The renormalization-group is a powerful method
29: in the study of low-dimensional Fermi systems, providing in particular
30: a systematic and unbiased method to study
31: competing instabilities and entangled infrared
32: singularities at weak coupling.
33: Early renormalization-group approaches for one-dimensional systems,
34: combined with exact solutions of fixed-point
35: models, have been a major source of physical insight \cite{Sol79,Gia03}.
36: From the renormalization-group point of view, the existence of the
37: Luttinger liquid requires the cancellation
38: of contributions to the flow of the two-particle vertex
39: to all orders \cite{Metzner:9701012}.
40: This is a one-dimensional phenomenon, in
41: higher dimensions the interactions in general diverge and
42: the flow in the fermionic variables breaks down indicating a possible
43: opening of a gap in the fermionic excitation spectrum.
44:
45:
46: Wilson's renormalization-group approach \cite{Wil71,WK74} of successive
47: integration of degrees of freedom with different energy scales
48: determines the evolution of the bare action of the system, given by the
49: microscopic Hamiltonian, to the final effective action,
50: from which all physical quantities can be extracted.
51: The hierarchy of coupled differential flow equations
52: for the Green or vertex functions describing the
53: full functional evolution of the effective
54: action has been first implemented for bosonic field theories in the context
55: of critical phenomena \cite{WH73,Pol84,Wet93}.
56: The intuition of the relevance of fRG methods
57: for interacting Fermi systems followed in the 1990s
58: \cite{BG90,FT90,Sh91,Sh94}, together with important rigorous work
59: \cite{Salmhofer:1999}.
60: The infinite hierarchy of flow equations
61: can be solved exactly only in special cases, for
62: instance the Luttinger model \cite{Schuetz04}.
63: Truncations however preserve the
64: successive handling of energy scales and the consequent treatment of
65: infrared singularities,
66: characteristic of a renormalization-group treatment.
67:
68:
69: There are several variants of the fRG flow equations. The
70: flow equations for the connected amputated Green functions
71: correspond to the Polchinski scheme, first derived in
72: Ref.~\cite{Pol84,KKS92}.
73: The expansion of the connected amputated Green functions in
74: 1PI vertex functions led to the respective flow equations
75: \cite{WH73,Wei76},
76: subsequently derived from the Legendre transform of the generating
77: functional \cite{Wet93,Mor94,SH01}.
78: The Wick-ordered scheme is obtained from the Polchinski scheme by
79: expanding the generating functional of the connected amputated Green
80: functions in Wick-ordered polynomials
81: \cite{Wie88,Salmhofer:9706188,Salmhofer:1999}.
82: Important applications of the fRG in condensed-matter physics include the
83: two-dimensional Hubbard model using the Polchinski scheme
84: \cite{Zanchi:9703189,Zanchi:9812303}, the Wick-ordered scheme
85: \cite{HM00} and also the 1PI scheme \cite{HSFR01}.
86: In the context of classical disordered systems a fRG approach
87: is necessary to overcome the problem of dimensional reduction
88: \cite{Wie03}.
89: One-dimensional impurity problems and Luttinger-liquid physics are
90: most conveniently investigated in the 1PI scheme, as self-energy
91: contributions are included to all orders.
92:
93:
94: In the following the hierarchy of differential flow equations
95: for the 1PI vertex functions is derived,
96: which is obtained by differentiating
97: the corresponding generating functional with respect to an infrared
98: cutoff introduced in the free propagator
99: \cite{Salmhofer:9706188}.
100:
101:
102: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
103:
104:
105: \section{Generating functional}
106: \label{sec:rg:form}
107:
108:
109: We consider a system of interacting fermions
110: with single-particle propagator of the noninteracting system
111: $G_0$.
112: The properties of the system are determined by the action
113: %
114: \begin{equation}
115: \label{eq:act}
116: S[\psi,\psib] = (\psib, G_0^{-1} \psi) - V[\psi,\psib] \; ,
117: \end{equation}
118: %
119: where $\psib$ and $\psi$ are Grassmann variables associated with
120: creation and annihilation operators,
121: and $V[\psi,\psib]$ is an arbitrary many-body interaction.
122: Here we introduced the short-hand notation
123: $(\psib,G_0^{-1} \psi) = \sum_{K,K'} \psib_K [G_0^{-1}]_{K,K'}\psi_{K'}$, where
124: $K$ contains the Matsubara
125: frequency in addition to the single-particle quantum numbers and
126: $\sum_K$ stands for summation over the discrete indices and integrals
127: over the continuous ones.
128:
129:
130: All connected Green functions are obtained from
131: the generating functional \cite{NO87}
132: defined by
133: %
134: \begin{align}
135: \label{eq:gen}
136: e^{-\cG[\eta,\etab]} & = \frac{1}{\cZ_0}
137: \int\! d\psi d\psib \,
138: e^{S[\psi,\psib]} \, e^{-(\psib,\eta) -(\etab,\psi)}
139: \nonumber \\[2mm]
140: & =\int\! d\mu_Q[\psi,\psib] \,
141: e^{-V[\psi,\psib]} \, e^{-(\psib,\eta) -(\etab,\psi)} \; ,
142: \end{align}
143: %
144: with Grassmann source terms $\eta$ and $\etab$.
145: The normalized Gaussian measure with covariance $Q=G_0^{-1}$
146: %
147: \begin{equation}
148: \label{eq:norm}
149: d\mu_Q[\psi,\psib] = \frac{1}{\cZ_0}\,
150: d\psi d\psib \, e^{(\psib,Q\psi)}
151: \end{equation}
152: %
153: includes the exponential of the
154: quadratic part of the action
155: and the noninteracting partition function $\cZ_0$, such that
156: $\int \! d\mu_Q[\psi,\psib]=1$.
157: The generating functional for the connected Green functions
158: is related to the partition function of the physical system with
159: action (\ref{eq:act}) by
160: %
161: \begin{equation}
162: \cG[\eta,\etab]= - {\rm ln} \, \cZ[\eta,\etab] \; .
163: \end{equation}
164: %
165: In the noninteracting case $V[\psi,\psib] = 0$, and the Gaussian
166: integral
167: %
168: \begin{equation}
169: \int\! d\mu_Q[\psi,\psib] \, e^{-(\psib,\eta) -(\etab,\psi)} =
170: e^{-(\etab,G_0\eta)}
171: \end{equation}
172: %
173: implies that $\cG[\eta,\etab] = (\etab,G_0\eta)$.
174:
175:
176: The connected $m$-particle Green functions are given by
177: the derivatives of the generating functional $\cG[\eta,\etab]$
178: with respect to the source terms at $\eta=\etab=0$
179: %
180: \begin{align}
181: G_m(K'_1,\dots,K'_m;K_1,\dots,K_m) &=
182: (-1)^m \bra \psi_{K'_1} \dots \psi_{K'_m}
183: \psib_{K_m} \dots \psib_{K_1} \ket_c
184: \nonumber \\[2mm]
185: & =\, \left.
186: \frac{\partial^m}{\partial\eta_{K'_1} \dots \partial\eta_{K'_m}}\,
187: \frac{\partial^m}{\partial\etab_{K_m} \dots \partial\etab_{K_1}}\,
188: \cG[\eta,\etab] \right|_{\eta = \etab = 0} \; ,
189: \end{align}
190: %
191: where $\bra \dots \ket_c$ is the connected average of the
192: product of Grassmann variables between the brackets.
193:
194:
195: The connected amputated Green functions are generated by the
196: {\em effective interaction} $\cV[\chi,\chib]$ defined by
197: %
198: \begin{equation}
199: \label{eq:effint}
200: e^{-\cV[\chi,\chib]} =
201: \int\! d\mu_Q[\psi,\psib] \,
202: e^{-V[\psi+\chi,\psib+\chib]} \; .
203: \end{equation}
204: %
205: The substitution $\chi = G_0 \eta$ and $\chib = G_0^T \etab$,
206: where $G_0^T$ is the transposed propagator,
207: relates $\cV[\chi,\chib]$ to the functional $\cG[\eta,\etab]$ by
208: %
209: \begin{equation}
210: \cV[\chi,\chib] = \cG[\eta,\etab]-(\etab,G_0 \eta) \; .
211: \end{equation}
212: %
213: The functional derivatives of $\cV[\chi,\chib]$ generate
214: connected Green functions divided by
215: $G_0 (K_1) \dots G_0 (K_m) \, G_0 (K'_1) \dots G_0 (K'_m)$,
216: that is, propagators amputated from external
217: legs in the corresponding Feynman diagrams.
218: The term $(\etab, G_0 \eta)$ cancels the noninteracting part of
219: $\cG[\eta,\etab]$ such that $\cV[\chi,\chib] = 0$ for
220: $V[\psi,\psib] = 0$.
221: Hence, the noninteracting propagator is subtracted from the
222: one-particle Green function generated by $\cV[\chi,\chib]$.
223:
224:
225: The generating functional $\Gam[\phi,\phib]$ for the {\em 1PI} vertex
226: functions $\gamma_m$ is derived from the Legendre transform of
227: $\cG[\eta,\etab]$ by
228: %
229: \begin{equation}
230: \label{eq:onepii}
231: \Gam[\phi,\phib] +(\phib,Q\phi)=
232: \cG[\eta,\etab] + (\phib,\eta) -(\etab,\phi) \;,
233: \end{equation}
234: %
235: with
236: %
237: \begin{align}
238: \label{eq:help1}
239: \phi &= \frac{\partial \cG }{\partial \etab} & \eta &= \frac{\partial \Gam}{\partial \phib} +Q \phi& \nonumber\\[2mm]
240: \phib & = \frac{\partial \cG }{\partial \eta} &\etab &= \frac{\partial \Gam}{\partial \phi} -Q^T \phib &
241: \end{align}
242: %
243: and
244: %
245: \begin{equation}
246: \label{eq:help2}
247: \frac{\delta^2
248: \cG}{\delta\eta\,\delta\bar\eta} = \Big(\frac{\delta^2
249: \Gam}{\delta\phi\,\delta\bar\phi} +Q \Big)^{-1}\; .
250: \end{equation}
251: %
252: For the special case without interaction
253: $\cG[\eta,\etab]=(\etab,G_0\eta)$ leads to $\Gam[\phi,\phib]=0$.
254:
255:
256: The choice of the appropriate generating functional
257: for a convenient formulation of a renormalization-group approach
258: depends on the physical problem under investigation.
259: For a detailed description of the different schemes we refer to
260: Ref.~\cite{EnssThesis};
261: here we will concentrate on the 1PI version of the fRG.
262:
263:
264: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
265:
266:
267: \section[RG differential flow equation for $\Gam$]{RG differential flow
268: equation for $\mathbf{\Gam}$}
269: \label{sec:rg:flow}
270:
271:
272: In this section we briefly review the general renormalization-group setup,
273: introduced as a transformation that leaves the generating functional for
274: the correlation functions invariant,
275: and concentrate subsequently on the derivation of a continuous
276: renormalization-group equation for the 1PI functions, following the
277: derivation in the context of interacting Fermi systems in Ref.~\cite{SH01}.
278:
279: The addition principle for Gaussian fields implies that for the
280: decomposition $G_0=G^<_0+G^>_0$ the corresponding Gaussian measure factorizes as
281: %
282: \begin{align}
283: e^{-\cW[\Phi,\Phib]}&=\int\! d\mu_{Q} [\Psi,\Psib] e^{-\cY[\Psi+\Phi,\Psib+\Phib]}
284: \nonumber \\[2mm]
285: &=\int\! d\mu_{Q^<}[\Psi_<,\Psib_<]
286: \int\! d\mu_{Q^>} [\Psi_>,\Psib_>] e^{-\cY[\Psi_<+\Psi_>+\Phi,\Psib_<+\Psib_>+\Phib]} \; ,
287: \end{align}
288: %
289: with $\Psi = \Psi_<+\Psi_>$. The generating functional $\cW$
290: corresponds to $\cW=\cV$ for the particular choice $\cY=V$.
291: This leads to the semigroup law of the renormalization group
292: %
293: \begin{equation}
294: \label{eq:law}
295: e^{-\cW[\Phi,\Phib]}=\int\! d\mu_{Q^<}[\Psi',\Psib'] e^{-\cW_>[\Psi'+\Phi,\Psib'+\Phib]} \; ,
296: \end{equation}
297: %
298: where in $\cW_>=\cW(Q^>,\cY)$ the fields with propagator $Q^>$ have been
299: integrated out.
300: The semigroup law implies that the system $(Q,\cY)$ under analysis is
301: exactly equivalent to the system $(Q^<,\cW(Q^>,\cY))$.
302: In the present case $Q^>$ is a covariance with infrared
303: cutoff $\Lam$, and $Q^<$ has support only for fields with energies smaller
304: than $\Lam$.
305:
306:
307: Set up in this way, the renormalization group is simply a symmetry of
308: the generating functional $\cW(Q,\cY)$.
309: In differential form $\cW(Q,\cY)$ is independent of $\Lam$, that is,
310: %
311: \begin{equation}
312: \frac{\partial}{\partial \Lam} \cW(Q,\cY) = 0 \; .
313: \end{equation}
314: %
315: Inserting the right-hand side of Eq.~(\ref{eq:law}) leads to the
316: flow equation describing
317: the gradual evolution from $\cY$ to the effective functional $\cW$
318: as a function of the continuously decreasing energy cutoff $\Lam$.
319: $\cW_>$ is an infinite power series in the fields;
320: the quadratic and quartic terms correspond to the self-energy and
321: the effective interaction, higher order terms are however always
322: present and the convergence of the infinite series is a nontrivial
323: problem \cite{SH01}.
324:
325:
326: In the following the differential equation for the generating
327: functional $\Gam^{\Lam}[\phi,\phib]$ of the 1PI functions,
328: starting point for the hierarchy of differential flow equations,
329: is derived.
330: Introducing an infrared cutoff at an energy scale $\Lam > 0$ in the
331: bare propagator leads to a $\Lam$-dependent generating functional for
332: the connected Green functions defined by
333: %
334: \begin{equation}
335: \label{eq:glambda1}
336: e^{-\G^\Lambda[\eta,\bar\eta]}
337: = \int \dmu{Q^\Lambda}{\psi}\, e^{-V[\psi,\bar\psi]}\,
338: e^{-\inner{\bar\psi}{\eta}-\inner{\bar\eta}{\psi}} \; .
339: \end{equation}
340: %
341: The original functional is recovered in the limit $\Lam \to 0$.
342: Similarly the functional $\Gam^{\Lam}[\phi,\phib]$ generating the 1PI
343: vertex functions is constructed with $G_0^{\Lam}$ replacing $G_0$
344: in Eqs.~(\ref{eq:onepii} - \ref{eq:help2}).
345: Differentiating the above Eq.~(\ref{eq:glambda1}) for
346: $\cG^{\Lam}[\eta,\bar\eta]$ with respect to $\Lam$
347: yields
348: %
349: \begin{align}
350: -\frac{\partial \cG^\Lambda[\eta,\bar\eta]}{\partial \Lam}\, e^{-\G^\Lambda[\eta,\bar\eta]}
351: & = - \Tr\,(G_0^\Lambda {\dot Q^\Lambda}
352: ) \,e^{-\G^\Lambda[\eta,\bar\eta]} \nonumber \\[2mm]
353: & \quad\, + \int \dmu{Q^\Lambda}{\psi} \,
354: (\bar\psi,{\dot Q^\Lambda}\psi) \, e^{-V_0[\psi,\bar\psi]} \,
355: e^{-(\bar\psi,\eta)-\inner{\bar\eta}{\psi}} \nonumber\\[2mm]
356: & = -\big[ \Tr \,(G_0^\Lambda {\dot Q^\Lambda}
357: ) +\Laplace{\dot
358: Q^\Lambda} \big] e^{-\G^\Lambda[\eta,\bar\eta]} \;,
359: \end{align}
360: %
361: where the first term comes from the derivative of the normalization
362: factor (\ref{eq:norm}), and $\Tr$ denotes the sum over all space-time
363: indices. The functional Laplace operator $\Laplace{Q}$ is defined as
364: %
365: \begin{align}
366: \Laplace{Q}
367: = \inner{\ddd\eta}{Q\ddd{\bar\eta}}
368: = \sum_K \ddd{\eta_K}\, Q_K\, \ddd{\bar\eta_K} \;.
369: \end{align}
370: %
371: The flow of $\G^\Lambda[\eta,\bar\eta]$ is then
372: %
373: \begin{align}
374: \label{eq:gflow}
375: \frac{\partial \G^\Lambda[\eta,\bar\eta]}{\partial \Lam}
376: = \Tr \,(G_0^\Lambda {\dot Q^\Lambda} )
377: -\Tr\left[\dot Q^\Lambda\,
378: \frac{\delta^2 \G^\Lambda[\eta,\bar\eta]}
379: {\delta\eta\, \delta\bar\eta}\right]
380: +\inner{\frac{\delta\G^\Lambda[\eta,\bar\eta]}{\delta\eta}} {\dot
381: Q^\Lambda \frac{\delta\G^\Lambda[\eta,\bar\eta]}{\delta\bar\eta}}\;.
382: \end{align}
383: %
384: Using the Legendre transform (\ref{eq:onepii}) the derivative of
385: $\Gam^{\Lam}[\phi,\phib]$ reads
386: %
387: \begin{align}
388: \frac{\partial \Gamma^\Lambda[\phi,\bar\phi]}{\partial \Lam}
389: & = \frac{\partial \G^\Lambda[\eta,\bar\eta]}{\partial \Lam}
390: - (\bar\phi,\dot Q^\Lambda\phi) \nonumber\\[2mm]
391: & = \Bigl(\frac{\delta\G^\Lambda[\eta,\bar\eta]}{\delta\eta},
392: \dot Q^\Lambda \frac{\delta\G^\Lambda[\eta,\bar\eta]}{\delta\bar\eta}\Bigr)
393: \nonumber\\[2mm]
394: &\quad\,-\Tr \,\Bigl[ {\dot Q^\Lambda}
395: \Bigl(\frac{\delta^2 \G^\Lambda[\eta,\bar\eta]} {\delta\eta\,\delta\bar\eta}
396: -G_0^\Lambda\Bigr)\Bigr] - (\bar\phi,{\dot Q^\Lambda}\phi) \nonumber\\[2mm]
397: & =
398: -\Tr \Bigl[ \dot Q^\Lambda
399: \Bigl( \Bigl(\frac{\delta^2
400: \Gamma^\Lambda[\phi,\bar\phi]}{\delta\phi\,\delta\bar\phi}
401: + Q^\Lambda \Bigr)^{-1} - G_0^\Lambda \Bigr)\Bigr] \;,
402: \end{align}
403: %
404: leading to the exact renormalization-group equation
405: %
406: \begin{align}
407: \label{rge}
408: \frac{\partial}{\partial \Lam} \Gamma^\Lambda[\phi,\bar\phi] =
409: \Tr \Bigl[G_0^\Lambda \frac{\partial (G_0^\Lambda)^{-1}}{\partial \Lam}
410: \Bigr]-
411: \Tr \Bigl[\Bigl(\frac{\delta^2
412: \Gamma^\Lambda[\phi,\bar\phi]}{\delta\phi\,\delta\bar\phi}
413: + (G_0^\Lambda)^{-1}\Bigr)^{-1} \frac{\partial (G_0^\Lambda)^{-1}}{\partial \Lam} \Bigr] \; .
414: \end{align}
415: %
416: With the initial condition
417: %
418: \begin{equation}
419: \Gam^{\Lam_0}[\phi,\phib] = V[\phi,\phib]
420: \end{equation}
421: %
422: Eq.~(\ref{rge}) determines the flow of $\Gam^{\Lam}$ uniquely for
423: all $\Lam < \Lam_0$.
424:
425:
426: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
427:
428:
429: \section{Expansion in the fields and exact hierarchy of flow equations}
430: \label{sec:rg:exp}
431:
432:
433: The renormalization-group equations for the 1PI $m$-particle vertex functions
434: $\gamma_m^{\Lam}$ are derived by expanding $\Gam^{\Lam}[\phi,\phib]$ in
435: Eq.~(\ref{rge}) as a power series in the fields.
436: The coefficients in the expansion of $\Gam^{\Lam}[\phi,\phib]$ determine
437: $\gamma_m^{\Lam}$ by
438: %
439: \begin{align}
440: \label{eq:gam}
441: \Gamma^{\Lam}[\phi,\bar\phi] = \sum_{m=0}^\infty \frac{1}{(m!)^2}
442: \sum_{K_1\dots K_m} \sum_{K_1'\dots K_m'}
443: \gamma_m^{\Lam}(K_1',\dots,K_m';K_1,\dots,K_m)
444: \prod_{j=1}^m \bar\phi_{K_j'} \phi_{K_j} \; .
445: \end{align}
446: %
447: Due to the antisymmetry properties of the Grassmann variables
448: only antisymmetric vertex functions contribute.
449:
450:
451: Similarly the second derivative on the right-hand side of Eq.~(\ref{rge})
452: can be expanded.
453: Separating the $\phi$-independent part corresponding to the self-energy
454: yields
455: %
456: \begin{align}
457: \frac{\delta^2
458: \Gamma^\Lambda[\phi,\bar\phi]}{\delta\phi\,\delta\bar\phi}=
459: \frac{\delta^2
460: \Gamma^\Lambda[\phi,\bar\phi]}{\delta\phi\,\delta\bar\phi} \at_{\phi=\bar\phi=0}
461: +\tilde{\Gamma}^\Lambda[\phi,\bar\phi] = -\Sg ^\Lambda
462: +\tilde{\Gamma}^\Lambda[\phi,\bar\phi] \; ,
463: \end{align}
464: %
465: where the remaining functional $\tilde{\Gamma}^\Lambda[\phi,\bar\phi]$
466: is defined by Eq.~(\ref{eq:gam}) with indices $m$ and $j$ starting from
467: $2$.
468: The second term on the right-hand side of the flow equation
469: (\ref{rge}) then reads
470: %
471: \begin{align}
472: \Bigl(\frac{\delta^2 \Gamma^\Lambda[\phi,\bar\phi]}
473: {\delta\phi\,\delta\bar\phi} + (G_0^\Lambda)^{-1} \Bigr)^{-1}
474: &= \bigl( \tilde\Gamma^\Lambda + (G^\Lambda)^{-1} \bigr)^{-1}
475: \nonumber\\[2mm]
476: &= G^\Lambda\, \sum_{l=0}^{\infty} (-1)^l \big(\tilde\Gamma^\Lambda \,G^\Lambda\big)^l\; ,
477: \end{align}
478: %
479: with the full propagator $G^\Lambda$ defined via the Dyson equation
480: $(G^\Lambda)^{-1} = (G_0^\Lambda)^{-1} - \Sigma^\Lambda$.
481: Introducing the {\em single-scale propagator} $S^\Lambda$ as
482: %
483: \begin{equation}
484: \label{eq:ssp}
485: S^\Lambda = G^\Lambda\,\frac{\partial (G_0^\Lambda)^{-1}}{\partial \Lam} \, G^\Lambda
486: \end{equation}
487: %
488: the differential equation (\ref{rge}) for $\Gam^\Lambda$ is
489: %
490: \begin{align}
491: \label{eq:gammaflow3}
492: \frac{\partial}{\partial \Lam} \Gamma^\Lambda &=
493: \Tr \Bigl[G_0^\Lambda \frac{\partial (G_0^\Lambda)^{-1}}{\partial \Lam}
494: \Bigr]+
495: \sum_{l=0}^{\infty} (-1)^{l+1} \Tr \Big[G^\Lambda \frac{\partial}{\partial \Lam} [G_0^\Lambda]^{-1} \big(G^\Lambda\, \tilde\Gamma^\Lambda \big)^l \Big]
496: \nonumber \\[2mm]
497: &=
498: \Tr \Bigl[\bigl(G_0^\Lambda-G^{\Lam} \bigl)\frac{\partial (G_0^\Lambda)^{-1}}{\partial \Lam}
499: \Bigr]+
500: \sum_{l=0}^{\infty} (-1)^l \Tr \Big[S^\Lambda \tilde\Gamma^\Lambda\,\big(G^\Lambda\, \tilde\Gamma^\Lambda \big)^l \Big] \; .
501: \end{align}
502: %
503: The first term corresponds to a vacuum energy not
504: entering the correlation functions, while
505: the second one contains one-loop diagrams with $(l+1)$
506: vertices $\tilde\Gamma^\Lambda$ connected by one single-scale
507: propagator $S^\Lambda$ and $l$ full propagators $G^\Lambda$.
508: The term linear in $\tilde\Gamma^\Lambda$ generates self-energy
509: corrections.
510:
511:
512: Inserting the components $\gamma_m^\Lambda$ on the left-hand side and
513: the components $\tilde\gamma_m^\Lambda$ on
514: the right-hand side of the flow equation (\ref{eq:gammaflow3}) for
515: $\Gamma^\Lambda$ we obtain a system of
516: equations for $\gamma_m^\Lambda$.
517: In a graphical representation the equations for $m \leq 3$ are
518: %
519: \begin{align}
520: \label{eq:hiera}
521: \parbox{23mm}{\unitlength=1mm\fmfframe(3,8)(2,8){
522: \begin{fmfgraph*}(17,17)
523: \fmfv{d.sh=circle,d.f=full,d.size=1mm}{i1}
524: \fmfforce{(9mm,17mm)}{i1}
525: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{v}
526: \fmfleft{i}
527: \fmfright{o}
528: \fmf{plain}{i,v,o}
529: \end{fmfgraph*}}}
530: &\! =\!
531: \parbox{18mm}{\unitlength=1mm\fmfframe(2,10)(2,10){
532: \begin{fmfgraph*}(17,17)
533: \fmfsurroundn{e}{8}
534: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{v}
535: \fmfleft{i}\fmfright{o}\fmf{phantom}{i,v,o}\fmffreeze
536: \fmf{plain}{v,e6}
537: \fmf{plain}{v,e8}
538: \fmf{slplaino,right=100,tension=0.4,label=$S^\Lambda$}{v,v}
539: \end{fmfgraph*}}}
540: \end{align}
541: %
542: \begin{align}
543: \label{eq:hierb}
544: \parbox{23mm}{\unitlength=1mm\fmfframe(3,8)(2,8){
545: \begin{fmfgraph*}(17,17)
546: \fmfv{d.sh=circle,d.f=full,d.size=1mm}{i1}
547: \fmfforce{(9mm,17mm)}{i1}
548: \fmfsurroundn{e}{8}
549: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{v}
550: \fmfleft{i}\fmfright{o}\fmf{phantom}{i,v,o}\fmffreeze
551: \fmf{plain}{v,e2}
552: \fmf{plain}{v,e4}
553: \fmf{plain}{v,e6}
554: \fmf{plain}{v,e8}
555: \end{fmfgraph*}}}
556: & \!=\!
557: \parbox{23mm}{\unitlength=1mm\fmfframe(3,10)(2,10){
558: \begin{fmfgraph*}(17,17)
559: \fmfsurroundn{e}{6}
560: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{v}
561: \fmfleft{i}\fmfright{o}\fmf{phantom}{i,v,o}\fmffreeze
562: \fmf{plain}{v,e1}
563: \fmf{plain}{v,e4}
564: \fmf{plain}{v,e5}
565: \fmf{plain}{v,e6}
566: \fmf{slplaino,right=100,tension=0.4,label=$S^\Lambda$}{v,v}
567: \end{fmfgraph*}}}
568: +
569: \parbox{47mm}{\unitlength=1mm\fmfframe(-4,6)(1,6){
570: \begin{fmfgraph*}(47,17)
571: \fmfleftn{l}{3}
572: \fmfrightn{r}{3}
573: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{vl}
574: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{vr}
575: \fmffixed{(.55w,0)}{vl,vr}
576: \fmfleft{i}\fmfright{o}\fmf{phantom}{i,vl,vr,o}\fmffreeze
577: \fmf{plain}{vl,l1}
578: \fmf{plain}{vl,l3}
579: \fmf{plain}{vr,r1}
580: \fmf{plain}{vr,r3}
581: \fmf{slplain,left=0.5,label=$S^\Lambda$}{vl,vr}
582: \fmf{plain,right=0.5,label.side=left,label=$G^\Lambda$}{vl,vr}
583: \end{fmfgraph*}}}
584: \end{align}
585: %
586: \begin{align}
587: \label{eq:hierc}
588: \parbox{23mm}{\unitlength=1mm\fmfframe(3,8)(2,8){
589: \begin{fmfgraph*}(17,17)
590: \fmfv{d.sh=circle,d.f=full,d.size=1mm}{i1}
591: \fmfforce{(9mm,17mm)}{i1}
592: \fmfsurroundn{e}{12}
593: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{v}
594: \fmfleft{i}\fmfright{o}\fmf{phantom}{i,v,o}\fmffreeze
595: \fmf{plain}{v,e1}
596: \fmf{plain}{v,e3}
597: \fmf{plain}{v,e5}
598: \fmf{plain}{v,e7}
599: \fmf{plain}{v,e9}
600: \fmf{plain}{v,e11}
601: \end{fmfgraph*}}}
602: &\! =\!
603: \parbox{23mm}{\unitlength=1mm\fmfframe(3,10)(2,10){
604: \begin{fmfgraph*}(17,17)
605: \fmfsurroundn{e}{10}
606: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{v}
607: \fmfleft{i}\fmfright{o}\fmf{phantom}{i,v,o}\fmffreeze
608: \fmf{plain}{v,e1}
609: \fmf{plain}{v,e6}
610: \fmf{plain}{v,e7}
611: \fmf{plain}{v,e8}
612: \fmf{plain}{v,e9}
613: \fmf{plain}{v,e10}
614: \fmf{slplaino,right=100,tension=0.4,label=$S^\Lambda$}{v,v}
615: \end{fmfgraph*}}}
616: +
617: \parbox{47mm}{\unitlength=1mm\fmfframe(-4,6)(1,6){
618: \begin{fmfgraph*}(47,17)
619: \fmfleftn{l}{3}
620: \fmfrightn{r}{4}
621: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{vl}
622: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{vr}
623: \fmffixed{(.55w,0)}{vl,vr}
624: \fmfleft{i}\fmfright{o}\fmf{phantom}{i,vl,vr,o}\fmffreeze
625: \fmf{plain}{vl,l1}
626: \fmf{plain}{vl,l3}
627: \fmf{plain}{vr,r1}
628: \fmf{plain}{vr,r2}
629: \fmf{plain}{vr,r3}
630: \fmf{plain}{vr,r4}
631: \fmf{slplain,left=0.5,label=$S^\Lambda$}{vl,vr}
632: \fmf{plain,right=0.5,label.side=left,label=$G^\Lambda$}{vl,vr}
633: \end{fmfgraph*}}}
634: \!\!\!\!\! + \;\;\,
635: \parbox{30mm}{\unitlength=1mm\fmfframe(-3,0)(-2,0){
636: \begin{fmfgraph*}(33,33)
637: \fmfsurroundn{e}{24}
638: \fmfv{d.sh=circle,d.f=30,d.si=10mm}{vl,vr,vu}
639: \fmffixed{(.5w,0)}{vl,vr}
640: \fmffixed{(.25w,.43h)}{vl,vu}
641: \fmf{plain}{vu,e6}
642: \fmf{plain}{vu,e8}
643: \fmf{plain}{vl,e14}
644: \fmf{plain}{vl,e16}
645: \fmf{plain}{vr,e22}
646: \fmf{plain}{vr,e24}
647: \fmf{slplain,left=0.5,label=$S^\Lambda$}{vl,vu}
648: \fmf{plain,left=0.5,label=$G^\Lambda$}{vu,vr}
649: \fmf{plain,left=0.5,label=$G^\Lambda$}{vr,vl}
650: \end{fmfgraph*}}}
651: \end{align}
652: %
653: The initial conditions for the
654: vertex functions at $\Lam = \infty$ are given by the bare
655: interactions of the system. In particular, the flow of the two-particle
656: vertex starts from the antisymmetrized bare two-particle interaction
657: while $m$-particle vertices of higher order vanish at $\Lam = \infty$,
658: in the absence of bare $m$-body interactions with $m>2$.
659:
660:
661: Note that the right-hand side of the equation for $\gamma_m$
662: contains $\gamma_{m+1}$.
663: The infinite system of differential equations contains
664: only one-loop terms in every equation, as the differential formulation
665: of Eq.~(\ref{rge}) contains only a single trace,
666: and for the 1PI scheme no tree terms appear.
667: The infinite hierarchy produces the full Green functions,
668: generating graphs with an arbitrary
669: number of loops; truncations amount to a partial inclusion of
670: higher order contibutions generated during the flow,
671: where the internal lines contain only modes above the cutoff scale $\Lam$.
672: Consequences of symmetries are discussed in Ref.~\cite{SH01}.
673:
674:
675: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
676:
677:
678: \section{Comparison to other RG schemes}
679: \label{sec:rg:alt}
680:
681:
682: Infrared divergencies arising in the context of perturbative expansions or
683: in proximity of phase transitions can alternatively be regularized by
684: temperature, a weak coupling strength or a finite system size.
685: In the fRG approach the cutoff scale is introduced only in the quadratic
686: part of the bare action, and the regularization is implemented with respect
687: to energy scales.
688: The temperature and interaction flows are derived in Refs.~\cite{HS01,HRAE04}
689: respectively, a pedagogic introduction is given in Ref.~\cite{EnssThesis}.
690: The renormalization-group equations describe the
691: flow of the correlation functions as the cutoff scale is lowered.
692: The choice of the basis set for the correlation functions determines a
693: particular scheme.
694:
695:
696: In addition to the 1PI scheme described previously,
697: the various generating functionals introduced in Sec.~\ref{sec:rg:form}
698: correspond to different schemes.
699: Starting point for the {\em Polchinski scheme} is the effective interaction,
700: generating functional of the connected amputated Green functions.
701: The flow equation for $\cV^{\Lam}[\chi,\chib]$ is
702: derived by replacing $Q$ by $Q^{\Lam}$ in Eq.~(\ref{eq:effint}) and
703: taking the derivative with respect to $\Lam$.
704: An expansion in powers of $\chi$
705: and $\chib$ of the functional $\cV^{\Lam}[\chi,\chib]$ in the
706: renormalization-group equation leads to Polchinski's flow
707: equations for amputated connected Green functions \cite{Pol84,KKS92}, with
708: a similar structure as for the connected Green functions.
709: The connected amputated Green functions are the expansion coefficients
710: of the generating functional $\cV^{\Lam}[\chi,\chib]$ in terms of monomials
711: of the source fields $\chi$ and $\bar\chi$.
712: Alternatively, one can also expand $\cV^{\Lam}[\chi,\chib]$ with respect to
713: {\em Wick-ordered} polynomials, leading to the Wick-ordered
714: Green functions as expansion coefficients
715: \cite{Wie88,Salmhofer:9706188,Salmhofer:1999}.
716: The flow equations are characterized by a bilinear structure in
717: the vertices on the right-hand side connected by bare $\Lambda$-dependent
718: propagators.
719: The Wick ordering also implies that except for the differentiated propagator
720: the internal lines are supported below scale $\Lambda$ instead of above it.
721: Thus, for a momentum cutoff only momenta close to the Fermi surface
722: contribute at low cutoff scale $\Lam$.
723: This justifies a parametrization of the coupling
724: functions by projecting onto the Fermi surface \cite{HM00}.
725: Self-energy corrections are however most conveniently taken into
726: account in the 1PI formalism with full propagators on the internal lines.
727:
728:
729: In an exact treatment all schemes are equivalent,
730: differences arise with truncations of the infinite hierarchy of flow
731: equations.
732: While the full hierarchy of flow equations leads to the correct solution
733: to all orders in perturbation theory independently of the scheme, in the
734: computation of the lowest orders a particular scheme might be more
735: suitable than others, depending on the considered physical problem and
736: properties.
737: An important point for the choice concerns the possibility of an efficient
738: parametrization of the effective interactions by a manageable number of variables.
739:
740:
741: Continuous symmetries in the bare
742: action lead to conservation laws and Ward identities
743: relating Green and response functions, as a consequence of the Noether
744: theorem.
745: These are generally not preserved for the truncated flow equations,
746: in contrast to the solution of the infinite flow-equation hierarchy,
747: as shown in detail in Ref.~\cite{EnssThesis}.
748: For a gauge-invariant construction however, as for
749: the temperature-flow scheme, the Ward identities between Green and
750: response functions are satisfied exactly despite truncations.
751: The related property of self-consistency is satisfied by
752: construction in conserving approximations
753: \cite{BK61}, but generally violated
754: in truncated fRG flows.
755: However, in the one-dimensional lattice models for Luttinger
756: liquids, the truncated fRG is nevertheless surprisingly successful and
757: self-consistency does not appear to play an important role.
758:
759:
760:
761:
762: %%% Local Variables:
763: %%% mode: latex
764: %%% TeX-master: "thesis"
765: %%% End:
766: