1: % CUT HERE
2: %
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: % Start of a3hub.tex
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6: %
7: %\narrowtext
8: %
9:
10: \chapter{Hubbard-Stratonovich transformations}
11: %\markboth{\thechapter: HUBBARD-STRATONOVICH TRANSFORMATIONS}{}
12: \label{chap:ahub}
13:
14: {\it{Our functional bosonization approach is
15: based on two Hubbard-Stratonovich transformations, which are
16: described in detail in this chapter.}}
17:
18: \vspace{7mm}
19:
20: \noindent
21: We start with the imaginary time functional integral formulation of quantum statistical
22: mechanics. This modern approach to the many-body
23: problem has recently been described in excellent
24: textbooks \cite{Popov83,Popov87,Negele88,Kapusta89},
25: so that we can be rather brief here and simply summarize
26: the relevant representations of
27: fermionic correlation functions as Grassmannian functional integrals.
28: We then eliminate the Grassmann fields in favour of collective bosonic fields
29: by means of suitable Hubbard-Stratonovich transformations \cite{Stratonovich57}.
30: These can be viewed as a clever change of variables to collective coordinates in functional
31: integrals, which exhibit the physically most relevant
32: degrees of freedom. The associated Jacobians
33: define the effective actions for the
34: Hubbard-Stratonovich fields.
35: It turns out that the non-perturbative bosonization result
36: for the single-particle Green's function can be obtained
37: with the help of a conventional Hubbard-Stratonovich transformation that
38: involves a space- and time-dependent auxiliary field $\phi^{\alpha}$.
39: This transformation will be discussed in Sect.~\secref{sec:HS1}.
40: On the other hand, for the calculation of the
41: boson representation of the Hamiltonian or the
42: density-density correlation function
43: we need a generalized two-field Hubbard-Stratonovich transformation,
44: which involves besides the $\phi^{\alpha}$-field another
45: bosonic field $\tilde{\rho}^{\alpha}$.
46: Section~\secref{sec:HS2} is devoted to a detailed description of this transformation.
47:
48:
49:
50: \section{Grassmannian functional integrals}
51: %\markboth{\thechapter: HUBBARD-STRATONOVICH TRANSFORMATIONS}
52: %{\thesection: Grassmannian functional integrals}
53:
54: {\it{
55: Fermionic correlation functions
56: can be represented as Grassmannian functional integrals.
57: These representations are particularly convenient
58: for our purpose, because they can be directly
59: manipulated via Hubbard-Stratonovich transformations.}}
60:
61:
62: \vspace{7mm}
63:
64: \noindent
65: The grand canonical partition function ${\cal{Z}}$ of our
66: many-body Hamiltonian defined in Eqs.\ref{eq:Hdef}, \ref{eq:H0def}
67: and \ref{eq:Hintcourse} can be written as
68: an imaginary time (i.e. Euclidean) functional integral
69: over a Grassmann field\index{Grassmann field} $\psi$ \cite{Popov83,Popov87,Negele88,Kapusta89},
70: \begin{equation}
71: \frac{ {\cal{Z}} }{ {\cal{Z}}_{0} }
72: =
73: \frac{
74: \int {\cal{D}} \left\{ \psi \right\}
75: \E^{-S_{\rm mat} \{ \psi \} }
76: }
77: { \int {\cal{D}} \left\{ \psi \right\}
78: \E^{-S_{0} \{ \psi \} } }
79: \; \; \; ,
80: \end{equation}
81: where ${\cal{Z}}_{0}$ is the grand canonical partition function
82: in the absence of interactions, and
83: the Euclidean action $S_{\rm mat} \{ \psi \}$ is
84: given by
85: \begin{eqnarray}
86: S_{\rm mat} \{ \psi \} & = &
87: S_{0} \{ \psi \} + S_{\rm int} \{ \psi \}
88: \; \; \; ,
89: \label{eq:Smatpsidef}
90: \\
91: S_{0} \left\{ \psi \right\}
92: & = &
93: \beta \sum_{ k } \left[ - \I \tilde{\omega}_{n} +
94: \xi_{\vec{k}} \right] \psi^{\dagger}_{k} \psi_{ k }
95: \; \; \; ,
96: \label{eq:S0psidef}
97: \\
98: S_{\rm int} \{ \psi \} & = & \frac{ \beta }{2V} \sum_{{q}} \sum_{ \alpha \alpha^{\prime} }
99: {{f}}_{{q}}^{\alpha \alpha^{\prime}}
100: {\rho}_{ - {q}}^{\alpha} {\rho}_{q}^{\alpha^{\prime}}
101: \label{eq:Sintpsidef}
102: \; \; \; .
103: \end{eqnarray}
104: Here
105: \begin{equation}
106: \rho_{q}^{\alpha} = \sum_{k}
107: \Theta^{\alpha} ( {\vec{k}} )
108: \psi^{\dagger}_{k} \psi_{k+q}
109: \label{eq:rhoalpha}
110: \end{equation}
111: is the Grassmann representation of the
112: sector density operator $\hat{\rho}^{\alpha}_{\vec{q}}$ defined
113: in Eq.\ref{eq:rhoopdef}.
114: Note that the $k$- and $q$-sums in these expressions are over
115: wave-vectors {\it{and}} Matsubara frequencies.
116: Although the Landau parameters\index{Landau interaction parameter} $f^{\alpha \alpha^{\prime}}_{\vec{q}}$
117: that appear in
118: the Hamiltonian $\hat{H}_{\rm int}$ in Eq.\ref{eq:Hintcourse}
119: depend only on the wave-vector ${\vec{q}}$, we have
120: replaced them in Eq.\ref{eq:Sintpsidef}
121: by more general frequency-dependent parameters
122: $f^{\alpha \alpha^{\prime}}_{q} \equiv
123: f^{\alpha \alpha^{\prime}}_{{\vec{q}} , \I \omega_{m}}$.
124: In our functional integral approach
125: the frequency-dependence does not introduce
126: any additional complications.
127: In physical applications the frequency-dependence
128: is due to the fact that the underlying microscopic mechanism responsible for
129: the effective interaction between the electrons
130: is the exchange of some particle with a finite
131: velocity, such as phonons\footnote{
132: The coupled electron-phonon system will be discussed
133: in detail in Chap.~\secref{chap:aph}.}.
134: Moreover, even in the case of electromagnetism the
135: effective interaction becomes frequency-dependent
136: if the corrections of higher order in ${v_{\rm F}}/{c}$ are retained.
137: The static Coulomb potential is just the
138: ${v_{\rm F}}/{c} = 0$ limit.
139: The leading correction
140: is a retarded current-current interaction mediated
141: by the transverse radiation field, which will be
142: discussed in Chap.~\secref{chap:arad}.
143:
144: \vspace{7mm}
145:
146: The time-ordered Matsubara Green's function\index{Green's function!representation as
147: functional integral}
148: defined in Eq.\ref{eq:GMatsubaradef}
149: can be represented as the functional integral average
150: of $\psi_{k} \psi^{\dagger}_{k}$,
151: \begin{equation}
152: G ( k )
153: =
154: - \beta \frac{
155: \int {\cal{D}} \left\{ \psi \right\}
156: \E^{-S_{\rm mat} \{ \psi \} }
157: \psi_{k} \psi^{\dagger}_{k}
158: }
159: { \int {\cal{D}} \left\{ \psi \right\}
160: \E^{-S_{\rm mat} \{ \psi \} } }
161: \label{eq:GmatFourier}
162: \; \; \; .
163: \end{equation}
164: In absence of interactions
165: this reduces to
166: \begin{equation}
167: G_{0} ( k )
168: = - \beta \frac{
169: \int {\cal{D}} \left\{ \psi \right\}
170: \E^{-S_{0} \{ \psi \} }
171: \psi_{k} \psi^{\dagger}_{k}
172: }
173: { \int {\cal{D}} \left\{ \psi \right\}
174: \E^{-S_{0} \{ \psi \} } }
175: = \frac{1 }{ \I \tilde{\omega}_{n} - \xi_{\vec{k}} }
176: \label{eq:G0def2}
177: \; \; \; ,
178: \end{equation}
179: in agreement with Eq.\ref{eq:G0Matdef}.
180: From the Matsubara Green's function we can obtain the real space imaginary time
181: Green's function via Fourier transformation\index{Fourier transformation},
182: \begin{equation}
183: G ( {\vec{r}} , \tau ) = \frac{1}{\beta V}
184: \sum_{k} \E^{ \I ( {\vec{k}} \cdot {\vec{r}} - \tilde{\omega}_{n} \tau )}
185: G ( k )
186: \; \; \; .
187: \label{eq:FTGreal}
188: \end{equation}
189: Defining
190: \begin{equation}
191: \psi ( {\vec{r}} , \tau ) = \frac{1}{ \sqrt{ V } }
192: \sum_{ k } \E^{ \I ( {\vec{k}} \cdot {\vec{r}} - \tilde{\omega}_{n} \tau )}
193: \psi_{ k }
194: \; \; \; ,
195: \end{equation}
196: we can also write
197: \begin{equation}
198: G ( {\vec{r}} - {\vec{r}}^{\prime} , \tau - \tau^{\prime} )
199: =
200: - \frac{
201: \int {\cal{D}} \left\{ \psi \right\}
202: \E^{-S_{\rm mat} \{ \psi \} }
203: \psi ( {\vec{r}} , \tau ) \psi^{\dagger} ( {\vec{r}}^{\prime} , \tau^{\prime} )
204: }
205: { \int {\cal{D}} \left\{ \psi \right\}
206: \E^{-S_{\rm mat} \{ \psi \} } }
207: \label{eq:Gdef}
208: \; \; \; .
209: \end{equation}
210:
211: Two-particle Green's functions can also
212: be represented as functional integral averages.
213: The density-density correlation function\index{density-density correlation
214: function!functional integral representation} defined in Eq.\ref{eq:cordens}
215: can be written as
216: \begin{equation}
217: \Pi ( q ) =
218: \frac{\beta}{ V} \frac{
219: \int {\cal{D}} \left\{ \psi \right\}
220: \E^{-S_{\rm mat} \{\psi \}}
221: \rho_{q } \rho_{-q}
222: }
223: { \int {\cal{D}} \left\{ \psi \right\}
224: \E^{-S_{\rm mat} \{ \psi \}} }
225: \label{eq:cordensfuncgen}
226: \; \; \; ,
227: \end{equation}
228: where the composite Grassmann field corresponding to the Fourier components of the
229: total density is
230: (see Eq.\ref{eq:denstotop})
231: \begin{equation}
232: \rho_{q} =
233: \sum_{\alpha} \rho_{q}^{\alpha}
234: =
235: \sum_{k} \psi^{\dagger}_{k} \psi_{k+q}
236: \; \; \; .
237: \label{eq:rhocomptot}
238: \end{equation}
239: Using Eq.\ref{eq:Thetasum}
240: we may also write
241: \begin{equation}
242: \Pi ( q) = \sum_{\alpha \alpha^{\prime}} \Pi^{\alpha \alpha^{\prime}} ( q )
243: \label{eq:Pitotdecompose}
244: \; \; \; ,
245: \end{equation}
246: where for $q \neq 0$
247: \begin{eqnarray}
248: \Pi^{\alpha \alpha^{\prime} } ( q )
249: & = & \frac{1}{\beta V } \int_{0}^{\beta} \D \tau
250: \int_{0}^{\beta} \D \tau^{\prime}
251: \E^{ - \I \omega_{m} ( \tau - \tau^{\prime}) }
252: \langle {\cal{T}} \left[ \hat{\rho}^{\alpha}_{\vec{q}} ( \tau )
253: \hat{\rho}^{ \alpha^{\prime} }_{ -{\vec{q} } } ( \tau^{\prime} )
254: \right] \rangle
255: \nonumber
256: \\
257: & = &
258: \frac{\beta}{ V} \frac{
259: \int {\cal{D}} \left\{ \psi \right\}
260: \E^{-S_{\rm mat} \{\psi \}}
261: \rho^{\alpha}_{q } \rho^{\alpha^{\prime}}_{-q}
262: }
263: { \int {\cal{D}} \left\{ \psi \right\}
264: \E^{-S_{\rm mat} \{ \psi \} } }
265: \label{eq:cordensloc}
266: \; \; \; .
267: \end{eqnarray}
268: We shall refer to $\Pi ( q )$ as the {\it{global}} or {\it{total}}
269: density-density correlation function, and to $\Pi^{\alpha \alpha^{\prime} } ( q )$ as the
270: {\it{local}} or {\it{sector}} density-density correlation
271: function\index{density-density correlation function!sector}.
272: In the non-interacting limit Eq.\ref{eq:cordensloc} reduces to
273: \begin{equation}
274: \Pi_{0}^{\alpha \alpha^{\prime} }
275: ( q )
276: = - \frac{1}{ {V}} \sum_{\vec{k}}
277: \Theta^{\alpha} ( {\vec{k}} )
278: \Theta^{\alpha^{\prime}} ( {\vec{k}} + {\vec{q}} )
279: \frac{ f ( \xi_{\vec{k+q}} )
280: - f ( \xi_{\vec{k}} ) }
281: { \xi_{\vec{k+q}} - \xi_{\vec{k}} - \I \omega_{m} }
282: \label{eq:Pi0loc}
283: \; \; \; .
284: \end{equation}
285: By relabeling ${\vec{k}} + {\vec{q}} \rightarrow {\vec{k}}$
286: it is easy to see that
287: \begin{equation}
288: \Pi_{0}^{\alpha \alpha^{\prime} } ( q )
289: =
290: \Pi_{0}^{\alpha^{\prime} \alpha } ( - q )
291: \; \; \; .
292: \label{eq:Pi0sym}
293: \end{equation}
294: Substituting Eq.\ref{eq:Pi0loc} into Eq.\ref{eq:Pitotdecompose}
295: and using $\sum_{\alpha} \Theta^{\alpha} ( {\vec{k}} ) = 1$,
296: we recover the non-interacting Lindhard function given in Eq.\ref{eq:Pi0total}.
297: We would like to emphasize that in the above functional integral representations
298: of the correlation
299: functions the precise normalization for the integration measure
300: ${\cal{D}} \{ \psi \}$ is irrelevant, because the measure appears always
301: in the numerator as well as in the denominator.
302:
303:
304:
305:
306:
307: \section{The first Hubbard-Stratonovich transformation}
308: %\markboth{\thechapter: HUBBARD-STRATONOVICH TRANSFORMATIONS}
309: %{\thesection: The first Hubbard-Stratonovich transformation}
310: \label{sec:HS1}
311:
312: {\it{
313: We decouple the two-body interaction between the fermions
314: with the help of a Hubbard-Stratonovich\index{Hubbard-Stratonovich transformation!first
315: (background field $\phi^{\alpha}$)} field $\phi^{\alpha}$.
316: After integrating over the Fermi fields,
317: the single-particle Green's function
318: can then be written as a quenched average
319: with probability distribution given by
320: the effective action of the $\phi^{\alpha}$-field.}}
321:
322:
323: \subsection{Decoupling of the interaction}
324: \label{subsec:HS1sub2}
325:
326: The generalized Landau parameters $f_{q}^{\alpha \alpha^{\prime}}$ in Eq.\ref{eq:Sintpsidef} have units of
327: energy $\times$ volume. Because we would like
328: to work with dimensionless Hubbard-Stratonovich fields,
329: it is useful to introduce the dimensionless\index{Landau interaction parameter!dimensionless
330: ($\tilde{f}_{q}^{\alpha \alpha^{\prime}}$)}
331: Landau parameters
332: \begin{equation}
333: \tilde{f}_{q}^{\alpha \alpha^{\prime}} = \frac{\beta}{V} f_{q}^{\alpha \alpha^{\prime}}
334: \; \; \; .
335: \label{eq:Landaudimless}
336: \end{equation}
337: The interaction part of our Grassmannian action
338: can then be written as
339: \begin{equation}
340: S_{\rm int} \{ \psi \} = \frac{ 1 }{2} \sum_{{q}} \sum_{ \alpha \alpha^{\prime} }
341: {\tilde{f}}_{{q}}^{\alpha \alpha^{\prime}}
342: {\rho}_{ - {q}}^{\alpha} {\rho}_{q}^{\alpha^{\prime}}
343: \label{eq:Sintpsi2}
344: \; \; \; .
345: \end{equation}
346: Using the invariance
347: of the sum in Eq.\ref{eq:Sintpsi2} under simultaneous relabelling
348: $ \alpha \leftrightarrow \alpha^{\prime}$ and $q \rightarrow -q$,
349: it is easy to see that, without loss of generality, we may assume that
350: \begin{equation}
351: {\tilde{f}}_{q}^{ \alpha \alpha^{\prime}} =
352: {\tilde{f}}_{ -q}^{ \alpha^{\prime} \alpha}
353: \label{eq:ftildeherm}
354: \; \; \; ,
355: \end{equation}
356: which is analogous to Eq.\ref{eq:Pi0sym}.
357: We now decouple this action by means of the following Hubbard-Stratonovich transformation
358: involving a dimensionless bosonic auxiliary field $\phi^{\alpha}_{q}$,
359: \begin{eqnarray}
360: \exp \left[ - S_{\rm int} \{ \psi \} \right] & \equiv &
361: \exp \left[
362: - \frac{1}{2} \sum_{q} \sum_{\alpha \alpha^{\prime} }
363: [ \underline{\tilde{f}}_{ {{q}} } ]^{\alpha \alpha^{\prime} }
364: \rho^{\alpha}_{-q} \rho^{\alpha^{\prime}}_{q}
365: \right ]
366: \nonumber
367: \\
368: & &\hspace{-30mm} =
369: \frac{
370: \int {\cal{D}} \left\{ \phi^{\alpha} \right\}
371: \exp \left[ -
372: \frac{1}{2} \sum_{q}
373: \sum_{\alpha \alpha^{\prime} }
374: [ \underline{\tilde{f}}_{ {{q}} }^{-1} ] ^{\alpha \alpha^{\prime} }
375: \phi^{\alpha}_{-q} \phi^{\alpha^{\prime}}_{q}
376: - \I \sum_{q} \sum_{\alpha} \phi^{\alpha}_{-q} \rho_{q}^{\alpha} \right] }
377: {
378: \int {\cal{D}} \left\{ \phi^{\alpha} \right\}
379: \exp \left[ -
380: \frac{1}{2} \sum_{q}
381: \sum_{\alpha \alpha^{\prime} }
382: [ \underline{\tilde{f}}_{ {{q}} }^{-1} ] ^{\alpha \alpha^{\prime} }
383: \phi^{\alpha}_{-q} \phi^{\alpha^{\prime}}_{q} \right] }
384: \label{eq:HStrans1}
385: \; .
386: \end{eqnarray}
387: Here $\underline{ \tilde{f}}_{q}$ is a matrix in the patch indices, with
388: matrix elements given by
389: \begin{equation}
390: [ \underline{ \tilde{f}}_{q}]^{\alpha \alpha^{\prime} }
391: = \tilde{f}_{q}^{\alpha \alpha^{\prime}} = \frac{\beta}{V} f_{q}^{\alpha \alpha^{\prime}}
392: \label{eq:ftildematdef}
393: \; \; \; .
394: \end{equation}
395: Throughout this work we shall use the convention that
396: all underlined quantities are
397: matrices in the patch indices.
398: Eq.\ref{eq:HStrans1} is easily proved by shifting\index{shift transformation} the
399: $\phi^{\alpha}$-field in the numerator of the right-hand side according to
400: \begin{equation}
401: \phi^{\alpha}_{q} \rightarrow \phi^{\alpha}_{q}
402: - \I \sum_{\alpha^{\prime}} [ \underline{\tilde{f}}_{q} ]^{\alpha \alpha^{\prime} }
403: \rho_{q}^{\alpha^{\prime}}
404: \; \; \; ,
405: \label{eq:phialphashift}
406: \end{equation}
407: and using Eq.\ref{eq:ftildeherm}.
408: For later convenience, let us fix the
409: measure for the $\phi^{\alpha}$-integration such that
410: \begin{equation}
411: \int {\cal{D}} \left\{ \phi^{\alpha} \right\}
412: \exp \left[
413: -
414: \frac{1}{2} \sum_{q}
415: \sum_{\alpha \alpha^{\prime} }
416: [ \underline{\tilde{f}}_{ {{q}} }^{-1} ] ^{\alpha \alpha^{\prime} }
417: \phi^{\alpha}_{-q} \phi^{\alpha^{\prime}}_{q}
418: \right]
419: = \prod_{q}
420: \det ( \underline{ \tilde{f}}_{q} )
421: \label{eq:measure1}
422: \; ,
423: \end{equation}
424: where $\det$ denotes the determinant with respect to the patch indices.
425: Note that our complex auxiliary field
426: satisfies $\phi^{\alpha}_{-q} = ( \phi^{\alpha}_{q} )^{\ast}$,
427: because it couples to the Fourier components of the density, which have also this symmetry.
428: Of course, mathematically the $\phi^{\alpha}$-integrals in Eq.\ref{eq:HStrans1} and
429: \ref{eq:measure1} are only well defined if the matrix
430: $\underline{ \tilde{f}}_{q}$ is positive definite.
431: However, Eqs.\ref{eq:phialphashift} and \ref{eq:ftildeherm} are
432: sufficient to proof Eq.\ref{eq:HStrans1} as an algebraic
433: identity, so that we shall use this transformation for intermediate algebraic manipulations
434: even if the matrix $\underline{ \tilde{f}}_{q}$ is not positive definite.
435: Possible infinities due to vanishing (or even negative) eigenvalues
436: of the matrix $\underline{\tilde{f}}_{q}$ cancel between the denominator and numerator of
437: Eq.\ref{eq:HStrans1}.
438: For example, if all matrix elements of a $M \times M$-matrix have the same (non-zero) value,
439: then $M-1$ of its eigenvalues are equal to zero,
440: so that for constant matrices $\underline{\tilde{f}}_{q}$
441: we implicitly assume that the Gaussian integrations in Eq.\ref{eq:HStrans1}
442: have been regularized in some convenient way.
443: Note also that the appearance of
444: $\underline{\tilde{f}}_{{q}}^{-1}$
445: is only an intermediate
446: step in our calculation. The final expressions for physical correlation functions
447: can be written entirely in terms of
448: $\underline{\tilde{f}}_{q}$, and remain finite even if
449: this matrix is not positive definite.
450: Such a rather loose use of mathematics is quite common
451: in statistical field theory,
452: although for mathematicians it is certainly not acceptable.
453: Formally, the appearance of
454: $\underline{\tilde{f}}_{{q}}^{-1}$
455: at intermediate steps can be avoided with the help
456: of the two-field Hubbard-Stratonovich transformation
457: discussed in Sect.~\secref{sec:HS2},
458: see Eq.\ref{eq:HStrans3} below\footnote{
459: Other formal ways to avoid this problem are briefly discussed
460: in the books by Amit \cite[\mbox{p. $24$}]{Amit84}, and by
461: Itzykson and Drouffe \cite[\mbox{p. $ 153$}]{Itzykson89}. On the other hand,
462: Zinn-Justin mentions this problem \cite[\mbox{p. $518$}]{ZinnJustin89},
463: but does not hesitate to perform
464: a transformation of the form \ref{eq:HStrans1} for a general matrix
465: $\underline{\tilde{f}}_{q}$.
466: Moreover, in the book by Negele and Orland \cite[\mbox{p. $198$}]{Negele88} as well as
467: in Parisi's book \cite[\mbox{p. $209$}]{Parisi88}
468: this transformation is used without further comment.
469: I would like to thank Kurt
470: Sch\"{o}nhammer for giving me a copy of his notes with
471: a summary and discussion of the relevant references.}.
472:
473: \subsection{Transformation of the single-particle Green's function}
474:
475: Applying the Hubbard-Stratonovich transformation \ref{eq:HStrans1}
476: to the functional integral representation \ref{eq:GmatFourier}
477: of the single-particle Green's function, we obtain
478: \begin{equation}
479: G ( k )
480: = - \beta \frac{
481: \int {\cal{D}} \left\{ \psi \right\}
482: {\cal{D}} \left\{ \phi^{\alpha} \right\}
483: \E^{- {S} \{ \psi , \phi^{\alpha} \} }
484: \psi_{k} \psi^{\dagger}_{k}
485: }
486: { \int {\cal{D}} \left\{ \psi \right\}
487: {\cal{D}} \left\{ \phi^{\alpha} \right\}
488: \E^{- {S} \{ \psi , \phi^{\alpha} \} }}
489: \label{eq:Gpatchdef}
490: \; \; \; ,
491: \end{equation}
492: where the decoupled action is given by
493: \begin{equation}
494: {S} \{ \psi , \phi^{\alpha} \}
495: = {S}_{0} \left\{ \psi \right\} + {S}_{1} \left\{ \psi , \phi^{\alpha} \right\}
496: + {S}_{2} \left\{ \phi^{\alpha} \right\}
497: \label{eq:Sdecoupdef}
498: \; \; \; ,
499: \end{equation}
500: with
501: \begin{eqnarray}
502: {S}_{1} \{ \psi , \phi^{\alpha} \}
503: & = & \sum_{ q } \sum_{\alpha}
504: \I \rho^{\alpha}_{q} \phi_{ -q}^{\alpha}
505: \; \; \; ,
506: \label{eq:S1decoupdef}
507: \\
508: {S}_{2} \{ {{\phi}}^{\alpha} \} & = &
509: \frac{1}{2} \sum_{q} \sum_{\alpha \alpha^{\prime} }
510: [ \underline{\tilde{f}}_{ {{q}} }^{-1} ]^{ \alpha \alpha^{\prime} }
511: \phi_{-q}^{\alpha} \phi_{q}^{\alpha^{\prime}}
512: \label{eq:S2decoupdef}
513: \; \; \; .
514: \end{eqnarray}
515: Thus, the fermionic two-body interaction has
516: disappeared. Instead, we have the problem of a coupled field theory
517: in which a dynamic bosonic field $\phi^{\alpha}$ is coupled
518: linearly to the fermionic density.
519: The $\phi^{\alpha}$-field mediates the interaction between the
520: fermionic matter in the sense that
521: integration over the $\phi^{\alpha}$-field
522: (i.e. undoing the Hubbard-Stratonovich transformation)
523: generates an effective fermionic two-body interaction.
524: In fact, because all interactions in nature can be viewed as the result
525: of the exchange of some sort of particles, it is more general and fundamental
526: to define the problem of interacting fermions in terms of an action
527: that does not contain the fermionic two-body interaction explicitly, but
528: involves the linear coupling of the fermionic density to another bosonic field.
529: This point of view has been emphasized by
530: Feynman and Hibbs \cite{Feynman65}.
531: We shall come back to the
532: physical meaning of the Hubbard-Stratonovich field $\phi^{\alpha}$
533: in Chap.~\secref{subsec:Definitionofmodel}, where we shall show that
534: for the Maxwell action the $\phi^{\alpha}$-field
535: can be identified physically with the
536: scalar potential of electromagnetism.
537:
538:
539: In a functional integral we
540: have the freedom of performing the integrations
541: in any convenient order.
542: Let us now perform the fermionic
543: integration over the $\psi$-field in Eq.\ref{eq:Gpatchdef} before
544: integrating over the $\phi^{\alpha}$-field.
545: To do this, we write
546: \begin{equation}
547: {S}_{0} \left\{ \psi \right\} +
548: {S}_{1} \left\{ \psi , \phi^{\alpha} \right\}
549: = - \beta \sum_{k k^{\prime} } \psi^{\dagger}_{k}
550: [ {\hat{G}}^{-1}]_{ k k^{\prime} }
551: \psi_{k^{\prime}}
552: \; \; \; ,
553: \label{eq:SmatG}
554: \end{equation}
555: where $\hat{G}^{-1}$ is an infinite matrix in momentum and frequency space, with
556: matrix elements given by the formal Dyson equation
557: \begin{equation}
558: [ \hat{G}^{-1} ]_{k k^{\prime}} =
559: [ \hat{G}^{-1}_{0} ]_{k k^{\prime}} - [ \hat{V} ]_{k k^{\prime}}
560: \label{eq:GhatDyson}
561: \; \; \; .
562: \end{equation}
563: Here $\hat{G}_{0}$ is the non-interacting Matsubara Green's function matrix,
564: \begin{equation}
565: [ \hat{G}_{0} ]_{ k k^{\prime} }
566: = \delta_{k k^{\prime} } G_{0} ( k)
567: \; \; \; , \; \; \;
568: G_{0} ( k ) = \frac{1}{ \I \tilde{\omega}_{n} - \xi_{\vec{k}} }
569: \label{eq:hatG0}
570: \; \; \; ,
571: \end{equation}
572: and the self-energy matrix $\hat{V}$ is defined by
573: \begin{equation}
574: [ \hat{V} ]_{ k k^{\prime} } =
575: \sum_{\alpha}
576: \Theta^{\alpha} ( {\vec{k}} ) V^{\alpha}_{k- k^{\prime}}
577: \; \; \; , \; \; \;
578: V^{\alpha}_{q} =
579: \frac{\I}{\beta}
580: {\phi}^{\alpha}_{q}
581: \label{eq:hatVphi}
582: \; \; \; .
583: \end{equation}
584: Recall that $k$ denotes wave-vector and frequency, so that
585: $\delta_{k k^{\prime} } = \delta_{ {\vec{k}} {\vec{k}}^{\prime}} \delta_{n n^{\prime}}$.
586: Choosing the normalization of the integration measure
587: ${\cal{D}} \{ \psi \}$ suitably, the ``trace-log'' formula \cite{Popov83}\index{trace-log formula}
588: yields
589: \begin{eqnarray}
590: \int
591: {\cal{D}} \left\{ \psi \right\}
592: \exp \left[ {
593: - {S}_{0} \left\{ \psi \right\} - {S}_{1} \left\{ \psi , \phi^{\alpha} \right\}
594: } \right]
595: = \det \hat{G}^{-1}
596: \nonumber
597: \\
598: & & \hspace{-50mm} = \E^{ {\rm Tr} \ln \hat{G}^{-1} }
599: = \E^{ {\rm Tr} \ln \hat{G}_{0}^{-1} }
600: \E^{ {\rm Tr} \ln [ 1 - \hat{G}_{0} \hat{V} ] }
601: \label{eq:Fermiint}
602: \; \; \; ,
603: \\
604: - \beta \int
605: {\cal{D}} \left\{ \psi \right\}
606: \psi_{k} \psi^{\dagger}_{k}
607: \exp \left[ {
608: - {S}_{0} \left\{ \psi \right\} - {S}_{1} \left\{ \psi , \phi^{\alpha} \right\} }
609: \right]
610: \nonumber
611: \\
612: & &
613: \hspace{-50mm} =
614: [ \hat{G} ]_{k k }
615: \E^{{\rm Tr} \ln \hat{G}_{0}^{-1}} \E^{ {\rm Tr} \ln [ 1 - \hat{G}_{0} \hat{V} ] }
616: \label{eq:Fermiint3}
617: \; \; \; .
618: \end{eqnarray}
619: Hence, after integration over the fermions
620: the exact interacting Green's function \ref{eq:Gpatchdef}
621: can be written as a quenched average\index{Green's function!representation
622: as quenched average} of the diagonal element $[ \hat{G} ]_{kk}$,
623: \begin{eqnarray}
624: G( k ) =
625: \int {\cal{D}} \{ \phi^{\alpha} \}
626: {\cal{P}} \{ \phi^{\alpha} \}
627: [ \hat{G} ]_{kk}
628: \equiv
629: \left< [ \hat{G}]_{ k k } \right>_{ S_{\rm eff} }
630: \label{eq:avphi}
631: \; \; \; .
632: \end{eqnarray}
633: Note that
634: $ [ \hat{G}]_{ k k }$ is in general a very complicated
635: functional of the field $\phi^{\alpha}$.
636: The normalized probability distribution\index{probability distribution!background field}
637: ${\cal{P}} \{ \phi^{\alpha} \} $ is
638: \begin{equation}
639: {\cal{P}} \{ \phi^{\alpha} \}
640: =
641: \frac{
642: \E^{ - {S}_{\rm eff} \{ \phi^{\alpha} \} } }
643: {
644: \int {\cal{D}} \left\{ \phi^{\alpha} \right\}
645: \E^{ - {S}_{\rm eff} \{ \phi^{\alpha} \} } }
646: \; \; \; ,
647: \label{eq:probabphidef}
648: \end{equation}
649: where the effective action\index{effective action!for $\phi^{\alpha}$-field}
650: for the $\phi^{\alpha}$-field
651: contains, in addition to the action
652: ${S}_{2} \{ \phi^{\alpha} \} $
653: defined in Eq.\ref{eq:S2decoupdef},
654: a contribution due to the coupling to the electronic degrees of freedom,
655: \begin{equation}
656: {S}_{\rm eff} \{ \phi^{\alpha} \}
657: = {S}_{2} \{ \phi^{\alpha} \}
658: + {S}_{\rm kin} \{ \phi^{\alpha} \}
659: \label{eq:Seffphidef}
660: \; \; \; ,
661: \end{equation}
662: with
663: \begin{equation}
664: {S}_{\rm kin} \{ \phi^{\alpha} \} =
665: - {\rm Tr} \ln [ 1 - \hat{G}_{0} \hat{V} ]
666: \label{eq:Skinphidef}
667: \; \; \; .
668: \end{equation}
669: Note that in Eq.\ref{eq:avphi} one first calculates the Green's function
670: for a frozen configuration of the
671: $\phi^{\alpha}$-field, and then averages the resulting expression over all
672: configurations this field, with the probability distribution given
673: in Eq.\ref{eq:probabphidef}.
674: Such a procedure closely resembles the
675: {\it{background field method}}\index{background field method}, which is well-known in
676: the field theory literature \cite{DeWitt67}.
677: Following this terminology, we shall also
678: refer to our auxiliary field $\phi^{\alpha}_q$
679: as the {\it{background field}}.
680:
681:
682:
683: The above transformations are exact. Of course,
684: in practice it is impossible to calculate the
685: interacting Green's function from Eq.\ref{eq:avphi},
686: because (a) the matrix $\hat{G}^{-1}$ cannot be inverted exactly,
687: (b) the kinetic energy contribution
688: $S_{\rm kin} \{ \phi^{\alpha} \}$ to the effective action
689: of the $\phi^{\alpha}$-field can only be calculated perturbatively,
690: and (c) the probability distribution ${\cal{P}} \{ \phi^{\alpha} \}$
691: in Eq.\ref{eq:probabphidef} is not Gaussian, so that the averaging
692: procedure cannot be carried out exactly.
693: {\it{The amazing fact is now that there exists a physically interesting limit
694: where the difficulties (a), (b) and (c) can all be overcome.}}
695: The above method leads then to a new and non-perturbative
696: approach to the fermionic many-body problem.
697: The detailed description of this method and its application to
698: physical problems is the central topic of this book.
699: The highly non-perturbative character of this approach is evident
700: from the fact that in $d=1$ the
701: well-known bosonization result
702: for the Green's function of the
703: Tomonaga-Luttinger model \cite{Mattis65} can be obtained
704: with this method \cite{Lee88}. This will be explicitly shown
705: in Chap.~\secref{sec:Green1}.
706:
707:
708:
709:
710:
711: \section{The second Hubbard-Stratonovich transformation}
712: %\markboth{\thechapter: HUBBARD-STRATONOVICH TRANSFORMATIONS}
713: %{\thesection: The second Hubbard-Stratonovich transformation}
714: \label{sec:HS2}
715:
716: {\it{
717: In order to introduce collective bosonic density fields, we
718: perform another change of variables in the functional integral
719: by means of a second Hubbard-Stratonovich transformation\index{Hubbard-Stratonovich
720: transformation!second (density field $\tilde{\rho}^{\alpha}$)}.
721: In this way we arrive at the general definition of the bosonized kinetic energy.}}
722:
723:
724: \vspace{7mm}
725:
726:
727: \noindent
728: From Eq.\ref{eq:S1decoupdef} we see that
729: after the first Hubbard-Stratonovich transformation
730: the composite Grassmann field $\rho^{\alpha}$
731: couples linearly to the $\phi^{\alpha}$-field.
732: Evidently the $\phi^{\alpha}$-field is related to the $\rho^{\alpha}$-field
733: in a very similar fashion as
734: the chemical potential is related to the particle number.
735: In other words, the $\phi^{\alpha}$-field is
736: the {\it{conjugate field}} to the sector density $\rho^{\alpha}$.
737: We now use a second Hubbard-Stratonovich transformation
738: to eliminate the composite Grassmann field $\rho^{\alpha}$
739: in favour of a collective bosonic field
740: $\tilde{\rho}^{\alpha}$, which
741: can then be identified physically with the bosonized density fluctuation.
742: This additional transformation is
743: useful for the calculation of quantities that can be written
744: in terms of collective density fluctuations, such as the
745: density-density correlation function or the bosonized
746: Hamiltonian. On the other hand, for the calculation of the {\it{single-particle Green's function}}
747: the first Hubbard-Stratonovich transformation introduced in the previous section
748: is sufficient.
749:
750: \subsection[Transformation of the density-density \mbox{\hspace{20mm}} correlation function]
751: {Transformation of the density-density correlation function}
752:
753: Applying the first Hubbard-Stratonovich transformation
754: \ref{eq:HStrans1} to the Grassmannian functional
755: integral representation \ref{eq:cordensloc}
756: of the sector density-density
757: correlation function\index{density-density correlation function!sector},
758: we obtain
759: \begin{equation}
760: \Pi^{\alpha \alpha^{\prime} } ( q )
761: = \frac{\beta}{V}
762: \frac{
763: \int {\cal{D}} \left\{ \psi \right\}
764: {\cal{D}} \left\{ \phi^{\alpha} \right\}
765: \rho^{\alpha}_{q} \rho^{\alpha^{\prime}}_{-q}
766: \exp \left[ - {S} \left\{ \psi , \phi^{\alpha} \right\} \right]
767: }
768: { \int {\cal{D}} \left\{ \psi \right\}
769: {\cal{D}} \left\{ \phi^{\alpha} \right\}
770: \exp \left[ - {S} \left\{ \psi , \phi^{\alpha} \right\} \right] }
771: \label{eq:PiHS1}
772: \; \; \; .
773: \end{equation}
774: We now decouple the quadratic action $S_{2} \{ \phi^{\alpha} \}$
775: in this expression by means of an integration over another bosonic
776: field $\tilde{\rho}^{\alpha}_{q}$,
777: \begin{eqnarray}
778: \exp \left[ - S_{2} \{ \phi^{\alpha} \} \right] & \equiv &
779: \exp \left[
780: - \frac{1}{2} \sum_{q} \sum_{\alpha \alpha^{\prime} }
781: [ \underline{\tilde{f}}_{ {{q}} }^{-1} ]^{\alpha \alpha^{\prime} }
782: \phi^{\alpha}_{-q} \phi^{\alpha^{\prime}}_{q}
783: \right ]
784: \nonumber
785: \\
786: & & \hspace{-27mm} =
787: \frac{
788: \int {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
789: \exp \left[ -
790: \frac{1}{2} \sum_{q}
791: \sum_{\alpha \alpha^{\prime} }
792: [ \underline{\tilde{f}}_{ {{q}} } ] ^{\alpha \alpha^{\prime} }
793: \tilde{\rho}^{\alpha}_{-q} \tilde{\rho}^{\alpha^{\prime}}_{q}
794: + \I \sum_{q} \sum_{\alpha} \phi^{\alpha}_{-q} \tilde{\rho}_{q}^{\alpha} \right] }
795: {
796: \int {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
797: \exp \left[ -
798: \frac{1}{2} \sum_{q}
799: \sum_{\alpha \alpha^{\prime} }
800: [ \underline{\tilde{f}}_{ {{q}} } ] ^{\alpha \alpha^{\prime} }
801: \tilde{\rho}^{\alpha}_{-q} \tilde{\rho}^{\alpha^{\prime}}_{q} \right] }
802: \label{eq:HStrans2}
803: \; .
804: \end{eqnarray}
805: It is convenient to define the integration measure
806: for the $\tilde{\rho}^{\alpha}$-integral such that
807: \begin{equation}
808: \int {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
809: \exp \left[
810: - \frac{1}{2} \sum_{q} \sum_{\alpha \alpha^{\prime} }
811: [ \underline{\tilde{f}}_{ {{q}} } ]^{\alpha \alpha^{\prime} }
812: \tilde{\rho}^{\alpha}_{-q} \tilde{\rho}^{\alpha^{\prime}}_{q}
813: \right]
814: = \prod_{q} \left[ { \det ( \underline{ \tilde{f} }_{q} )} \right]^{-1}
815: \label{eq:measure2}
816: \; \; \; ,
817: \end{equation}
818: so that with Eq.\ref{eq:measure1} we have
819: \begin{eqnarray}
820: \int {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
821: \int {\cal{D}} \left\{ \phi^{\alpha} \right\}
822: \exp \left[
823: - \frac{1}{2} \sum_{q} \sum_{\alpha \alpha^{\prime} }
824: [ \underline{\tilde{f}}_{ {{q}} } ]^{\alpha \alpha^{\prime} }
825: \tilde{\rho}^{\alpha}_{-q} \tilde{\rho}^{\alpha^{\prime}}_{q}
826: \right.
827: & &
828: \nonumber
829: \\
830: & &
831: \hspace{-40mm}
832: \left.
833: -
834: \frac{1}{2} \sum_{q}
835: \sum_{\alpha \alpha^{\prime} }
836: [ \underline{\tilde{f}}_{ {{q}} }^{-1} ] ^{\alpha \alpha^{\prime} }
837: \phi^{\alpha}_{-q} \phi^{\alpha^{\prime}}_{q}
838: \right]
839: = 1
840: \; \; \; .
841: \label{eq:measure3}
842: \end{eqnarray}
843: Then our two-field decoupling\index{Hubbard-Stratonovich transformation!two-field}
844: of the original fermionic two-body interaction reads
845: \begin{eqnarray}
846: \exp \left[ - S_{\rm int} \{ \psi \} \right] \equiv
847: \exp \left[
848: - \frac{1}{2} \sum_{q} \sum_{\alpha \alpha^{\prime} }
849: [ \underline{\tilde{f}}_{ {{q}} } ]^{\alpha \alpha^{\prime} }
850: \rho^{\alpha}_{-q} \rho^{\alpha^{\prime}}_{q}
851: \right ]
852: =
853: \int {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
854: \int {\cal{D}} \left\{ \phi^{\alpha} \right\}
855: \nonumber
856: \\
857: & & \hspace{-107mm}
858: \times
859: \exp \left[
860: - \frac{1}{2} \sum_{q} \sum_{\alpha \alpha^{\prime} }
861: [ \underline{\tilde{f}}_{ {{q}} } ]^{\alpha \alpha^{\prime} }
862: \tilde{\rho}^{\alpha}_{-q} \tilde{\rho}^{\alpha^{\prime}}_{q}
863: + \I \sum_{q} \sum_{\alpha} \left[ \tilde{{\rho}}_{q}^{\alpha} -
864: {\rho}_{q}^{\alpha} \right] \phi_{-q}^{\alpha} \right]
865: \; .
866: \label{eq:HStrans3}
867: \end{eqnarray}
868: Note that $\rho^{\alpha}_{q} = \sum_{k} \Theta^{\alpha} ( {\vec{k}} )
869: \psi_{k}^{\dagger} \psi_{k+q }$ on the left-hand side
870: of this equation is a composite Grassmann field, while
871: $\tilde{\rho}^{\alpha}_{q}$ on the right-hand side is a complex collective bosonic field.
872: Eq.\ref{eq:HStrans3} can be viewed as a functional generalization of the
873: elementary identity
874: \begin{equation}
875: \E^{-x^2} = \int_{- \infty}^{\infty} \D y \int_{- \infty}^{\infty} \frac{ \D \phi}{2 \pi}
876: \E^{ - y^2 + \I ( y-x) \phi}
877: \label{eq:HS1d}
878: \; \; \; .
879: \end{equation}
880: Let us also point out that the two-field Hubbard-Stratonovich transformation
881: \ref{eq:HStrans3} does not involve the inverse
882: of the matrix $\underline{\tilde{f}}_{q}$, so that
883: it is perfectly well defined for matrices with constant elements.
884:
885: Applying the Hubbard-Stratonovich transformation \ref{eq:HStrans2}
886: to the denominator in Eq.\ref{eq:PiHS1} and integrating over the
887: fermionic $\psi$-field, we obtain with the help of
888: the ``trace-log'' formula \ref{eq:Fermiint},
889: \begin{eqnarray}
890: \int {\cal{D}} \left\{ \psi \right\}
891: {\cal{D}} \left\{ \phi^{\alpha} \right\}
892: \exp \left[
893: - {S} \left\{ \psi , \phi^{\alpha} \right\} \right]
894: =
895: \frac{ \E^{ {\rm Tr} \ln \hat{G}_{0}^{-1} }}
896: { \int {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
897: \E^{ - \tilde{S}_{2} \{ \tilde{\rho}^{\alpha} \} } }
898: & &
899: \nonumber
900: \\
901: & & \hspace{-88mm} \times
902: \int
903: {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
904: {\cal{D}} \left\{ \phi^{\alpha} \right\}
905: \exp \left[ - \tilde{{S}}_{2} \left\{ \tilde{\rho}^{\alpha} \right\}
906: + \I \sum_{q \alpha} \phi^{\alpha}_{-q} \tilde{\rho}^{\alpha}_{q}
907: - S_{\rm kin} \{ \phi^{\alpha} \} \right]
908: \; ,
909: \label{eq:denomint}
910: \end{eqnarray}
911: where the interaction contribution to the effective action of the
912: collective $\tilde{\rho}^{\alpha}$-field is
913: \begin{equation}
914: \tilde{S}_{2} \left\{ \tilde{\rho}^{\alpha} \right\} =
915: \frac{1}{2} \sum_{q} \sum_{\alpha \alpha^{\prime}}
916: [ \underline{\tilde{f}}_{q} ]^{ \alpha \alpha^{\prime} }
917: \tilde{\rho}^{\alpha}_{-q} \tilde{\rho}^{\alpha^{\prime}}_{q}
918: \; \; \; ,
919: \label{eq:Sintrhodef}
920: \end{equation}
921: and the action $ S_{\rm kin} \{ \phi^{\alpha} \}$ is defined in Eq.\ref{eq:Skinphidef}.
922: The relation analogous to Eq.\ref{eq:denomint} for the numerator in Eq.\ref{eq:PiHS1}
923: is
924: \begin{eqnarray}
925: { \int {\cal{D}} \left\{ \psi \right\}
926: {\cal{D}} \left\{ \phi^{\alpha} \right\}
927: \rho^{\alpha}_{q} \rho^{\alpha^{\prime}}_{-q}
928: \exp \left[ - {S} \left\{ \psi , \phi^{\alpha} \right\} \right] }
929: & = &
930: \frac{ \E^{ {\rm Tr} \ln \hat{G}_{0}^{-1} }}
931: { \int {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
932: \E^{ - \tilde{S}_{2} \{ \tilde{\rho}^{\alpha} \} } }
933: \nonumber
934: \\
935: & & \hspace{-72mm} \times
936: \int
937: {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
938: {\cal{D}} \left\{ \phi^{\alpha} \right\}
939: \tilde{\rho}^{\alpha}_{q} \tilde{\rho}^{\alpha^{\prime}}_{-q}
940: \exp \left[ - \tilde{S}_{2} \{ \tilde{\rho}^{\alpha} \}
941: + \I \sum_{q \alpha} \phi^{\alpha}_{-q} \tilde{\rho}^{\alpha}_{q}
942: - S_{\rm kin} \{ \phi^{\alpha} \} \right]
943: \label{eq:enumint}
944: \; .
945: \end{eqnarray}
946: To proof Eq.\ref{eq:enumint}, we introduce the generating functional
947: \begin{equation}
948: {\cal{F}} \{ \tilde{\phi}^{\alpha} \}
949: =
950: \int
951: {\cal{D}} \left\{ \psi \right\}
952: {\cal{D}} \left\{ \phi^{\alpha} \right\}
953: \exp \left[
954: - {S} \left\{ \psi , \phi^{\alpha} \right\}
955: + \I \sum_{q \alpha}
956: \tilde{\phi}^{\alpha}_{-q}
957: \rho^{\alpha}_{q}
958: \right]
959: \; \; \; ,
960: \label{eq:generatingdef}
961: \end{equation}
962: which depends on external bosonic fields
963: $\tilde{\phi}^{\alpha}$ and generates via differentiation (up to
964: a constant factor) the left-hand side of Eq.\ref{eq:enumint},
965: \begin{eqnarray}
966: \left.
967: \frac{ \partial^2 {\cal{F}} \{ \tilde{\phi}^{\alpha} \} }{
968: \partial \tilde{\phi}^{\alpha}_{-q}
969: \partial \tilde{\phi}^{\alpha^{\prime}}_{q} }
970: \right|_{ \tilde{\phi}^{\alpha} = 0 }
971: & = &
972: \I^{2}
973: \int
974: {\cal{D}} \left\{ \psi \right\}
975: {\cal{D}} \left\{ \phi^{\alpha} \right\}
976: {\rho}^{\alpha}_{q} {\rho}^{\alpha^{\prime}}_{-q}
977: \exp \left[ - {S} \left\{ \psi , \phi^{\alpha} \right\} \right]
978: \label{eq:generate}
979: \; \; \; .
980: \end{eqnarray}
981: Applying the Hubbard-Stratonovich transformation \ref{eq:HStrans2}
982: to our generating functional, we obtain
983: \begin{eqnarray}
984: {\cal{ F}} \{ \tilde{\phi}^{\alpha} \}
985: & = &
986: \frac{ 1}
987: { \int {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
988: \E^{ - \tilde{S}_{2} \{ \tilde{\rho}^{\alpha} \} } }
989: \int
990: {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
991: {\cal{D}} \left\{ \phi^{\alpha} \right\}
992: \E^{
993: - \tilde{S}_{2} \left\{ \tilde{\rho}^{\alpha} \right\} }
994: \nonumber
995: \\
996: & & \hspace{-15mm} \times
997: \int
998: {\cal{D}} \left\{ \psi \right\}
999: \exp \left[ - {S}_{0} \{ \psi \}
1000: + \I \sum_{q \alpha} [ \phi^{\alpha}_{-q} \tilde{\rho}^{\alpha}_{q}
1001: - ( \phi^{\alpha}_{-q} - \tilde{\phi}^{\alpha}_{-q} ) \rho^{\alpha}_{q} ]
1002: \right]
1003: \label{eq:generateman}
1004: \; .
1005: \end{eqnarray}
1006: Shifting the integration over the $\phi^{\alpha}$-field according to
1007: $\phi^{\alpha}_{q} \rightarrow \phi^{\alpha}_{q} + \tilde{\phi}^{\alpha}_{q}$,
1008: we replace in the last term of the exponential in Eq.\ref{eq:generateman}
1009: \begin{equation}
1010: \phi^{\alpha}_{-q} \tilde{\rho}^{\alpha}_{q}
1011: - ( \phi^{\alpha}_{-q} - \tilde{\phi}^{\alpha}_{-q} ) \rho^{\alpha}_{q}
1012: \rightarrow
1013: \tilde{\phi}^{\alpha}_{-q} \tilde{\rho}^{\alpha}_{q}
1014: + \phi^{\alpha}_{-q} ( \tilde{\rho}^{\alpha}_{ q} - \rho^{\alpha}_{q} )
1015: \label{eq:shiftreplace}
1016: \; \; \; ,
1017: \end{equation}
1018: so that after the shift the derivatives with respect to the external $\tilde{\phi}^{\alpha}$-field
1019: generate factors of the collective bosonic density field $\tilde{\rho}^{\alpha}$.
1020: Performing now the fermionic integration and taking two derivatives with
1021: respect to the external field, we conclude from Eq.\ref{eq:generateman} that
1022: \begin{eqnarray}
1023: \left.
1024: \frac{ \partial^2 {\cal{F}} \{ \tilde{\phi}^{\alpha} \} }{
1025: \partial \tilde{\phi}^{\alpha}_{-q}
1026: \partial \tilde{\phi}^{\alpha^{\prime}}_{q} }
1027: \right|_{ \tilde{\phi}^{\alpha} = 0 }
1028: & = &
1029: \I^2
1030: \frac{ \E^{ {\rm Tr} \ln \hat{G}_{0}^{-1} }}
1031: { \int {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
1032: \E^{ - \tilde{S}_{2} \{ \tilde{\rho}^{\alpha} \} } }
1033: \int
1034: {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
1035: {\cal{D}} \left\{ \phi^{\alpha} \right\}
1036: \nonumber
1037: \\
1038: & & \hspace{-20mm} \times
1039: \tilde{\rho}^{\alpha}_{q} \tilde{\rho}^{\alpha^{\prime}}_{-q}
1040: \exp \left[ - \tilde{S}_{2} \{ \tilde{\rho}^{\alpha} \}
1041: + \I \sum_{q \alpha} \phi^{\alpha}_{-q} \tilde{\rho}^{\alpha}_{q}
1042: - S_{\rm kin} \{ \phi^{\alpha} \} \right]
1043: \label{eq:generate2}
1044: \; .
1045: \end{eqnarray}
1046: Comparing the right-hand sides of Eqs.\ref{eq:generate} and \ref{eq:generate2},
1047: the validity of Eq.\ref{eq:enumint} is evident.
1048: In summary, with the help of Eqs.\ref{eq:denomint}, \ref{eq:enumint} and
1049: \ref{eq:PiHS1}
1050: the sector density-density correlation function \ref{eq:cordensloc}
1051: can be represented as
1052: \begin{eqnarray}
1053: \Pi^{\alpha \alpha^{\prime} } ( q ) =
1054: \nonumber
1055: \\
1056: & & \hspace{-20mm}
1057: \frac{\beta}{V}
1058: \frac{
1059: \int
1060: {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
1061: \E^{- \tilde{S}_{2} \left\{ \tilde{\rho}^{\alpha} \right\} }
1062: \tilde{\rho}^{\alpha}_{q} \tilde{\rho}^{\alpha^{\prime}}_{-q}
1063: \int {\cal{D}} \left\{ \phi^{\alpha} \right\}
1064: \exp \left[
1065: \I \sum_{q \alpha} \phi^{\alpha}_{-q} \tilde{\rho}^{\alpha}_{q}
1066: - S_{\rm kin} \left\{ \phi^{\alpha} \right\} \right]
1067: }
1068: { \int
1069: {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
1070: \E^{- \tilde{S}_{2} \left\{ \tilde{\rho}^{\alpha} \right\} }
1071: \int {\cal{D}} \left\{ \phi^{\alpha} \right\}
1072: \exp \left[
1073: \I \sum_{q \alpha} \phi^{\alpha}_{-q} \tilde{\rho}^{\alpha}_{q}
1074: - S_{\rm kin} \left\{ \phi^{\alpha} \right\} \right]
1075: }
1076: \; .
1077: \nonumber
1078: \\
1079: \label{eq:Piphirho}
1080: \end{eqnarray}
1081:
1082: \subsection{Definition of the bosonized kinetic energy\index{bosonization!kinetic energy}}
1083:
1084: In complete analogy with Eqs.\ref{eq:avphi}--\ref{eq:Seffphidef},
1085: let us rewrite Eq.\ref{eq:Piphirho} as
1086: \begin{equation}
1087: \Pi^{\alpha \alpha^{\prime} } ( q )
1088: = \frac{\beta}{V}
1089: \int {\cal{D}} \{ \tilde{\rho}^{\alpha} \}
1090: \tilde{{\cal{P}}} \{ \tilde{\rho}^{\alpha} \}
1091: \tilde{\rho}^{\alpha}_{q} \tilde{\rho}^{\alpha^{\prime}}_{-q}
1092: \equiv
1093: \frac{\beta}{V}
1094: \left<
1095: \tilde{\rho}^{\alpha}_{q} \tilde{\rho}^{\alpha^{\prime}}_{-q}
1096: \right>_{ \tilde{S}_{\rm eff} }
1097: \label{eq:avrho}
1098: \; \; \; ,
1099: \end{equation}
1100: where the normalized probability distribution\index{probability distribution!density field}
1101: $\tilde{{\cal{P}}} \{ \tilde{\rho}^{\alpha} \}$
1102: for the collective density field $\tilde{\rho}^{\alpha}$ is
1103: \begin{equation}
1104: \tilde{\cal{P}} \{ \tilde{\rho}^{\alpha} \}
1105: =
1106: \frac{
1107: \E^{ - \tilde{S}_{\rm eff} \{ \tilde{\rho}^{\alpha} \} } }
1108: {
1109: \int {\cal{D}} \left\{ \tilde{\rho}^{\alpha} \right\}
1110: \E^{ - \tilde{S}_{\rm eff} \{ \tilde{\rho}^{\alpha} \} } }
1111: \; \; \; .
1112: \label{eq:probabrhodef}
1113: \end{equation}
1114: The effective action\index{effective action!for $\tilde{\rho}^{\alpha}$-field}
1115: of the $\tilde{\rho}^{\alpha}$-field has again two contributions,
1116: \begin{equation}
1117: \tilde{S}_{\rm eff} \left\{ \tilde{\rho}^{\alpha} \right\}
1118: =
1119: \tilde{S}_{2} \left\{ \tilde{\rho}^{\alpha} \right\}
1120: + \tilde{S}_{\rm kin} \left\{ \tilde{\rho}^{\alpha} \right\}
1121: \label{eq:Seffrhodef}
1122: \; \; \; ,
1123: \end{equation}
1124: with
1125: $\tilde{S}_{2} \left\{ \tilde{\rho}^{\alpha} \right\}$
1126: given in Eq.\ref{eq:Sintrhodef}, and
1127: \begin{equation}
1128: \tilde{S}_{\rm kin} \left\{ \tilde{\rho}^{\alpha} \right\}
1129: = - \ln \left(
1130: \int {\cal{D}} \left\{ \phi^{\alpha} \right\}
1131: \exp \left[ \I \sum_{q \alpha} \phi^{\alpha}_{-q} \tilde{\rho}^{\alpha}_{q}
1132: - S_{\rm kin} \left\{ \phi^{\alpha} \right\} \right] \right)
1133: \label{eq:Skinrhodef}
1134: \; \; \; .
1135: \end{equation}
1136: Note that
1137: $\tilde{S}_{\rm kin} \left\{ \tilde{\rho}^{\alpha} \right\} $ is
1138: related to
1139: $S_{\rm kin} \{ \phi^{\alpha} \}$ via a
1140: {\it{functional Fourier transformation}}\index{Fourier transformation!functional},
1141: while the quadratic action $\tilde{S}_{2} \{ \tilde{\rho}^{\alpha} \}$
1142: is simply obtained from
1143: $S_{\rm int} \{ \psi \}$ in
1144: Eq.\ref{eq:Sintpsidef} by replacing the
1145: composite Grassmann field $\rho^{\alpha}$ by the
1146: collective bosonic field $\tilde{\rho}^{\alpha}$.
1147: In this way the effect of the electron-electron interaction
1148: is taken into account exactly,
1149: while the contribution
1150: $\tilde{S}_{\rm kin} \left\{ \tilde{\rho}^{\alpha} \right\}$
1151: due to the kinetic energy
1152: can in general only be calculated approximately.
1153: In the next chapter
1154: we shall show that
1155: in the limit of long wavelengths and low energies
1156: the effective action
1157: $\tilde{S}_{\rm eff} \{ \tilde{\rho}^{\alpha} \}$
1158: in Eq.\ref{eq:Seffrhodef}
1159: is equivalent with the bosonized Hamiltonian of the interacting Fermi system.
1160: Obviously
1161: $\tilde{S}_{\rm eff} \{ \tilde{\rho}^{\alpha} \}$
1162: is in general {\it{not}} quadratic, so that the
1163: equivalent bosonized Hamiltonian contains terms
1164: describing interactions between the bosons.
1165: However, under certain conditions, which will be
1166: described in detail in Chap.~\secref{sec:closedloop},
1167: $\tilde{S}_{\rm eff} \{ \tilde{\rho}^{\alpha} \}$ can be approximated
1168: by a quadratic form. In this case
1169: bosonization enormously simplifies the many-body problem.
1170: In a sense, the
1171: collective density fields $\tilde{\rho}^{\alpha}$ are the ``correct coordinates''
1172: to parameterize the low-energy excitations of the system.
1173:
1174:
1175: \section{Summary and outlook}
1176:
1177: In this chapter we have used well-known representations of
1178: fermionic correlation functions as Grassmannian functional
1179: integrals and Hubbard-Stratonovich transformations
1180: to eliminate the fermionic degrees of freedom in favour of bosonic ones.
1181: The only new feature of these transformations is that
1182: our Hubbard-Stratonovich fields carry
1183: not only a momentum-frequency label $q$,
1184: but also a label $\alpha$
1185: that refers to the sectors $K^{\alpha}_{\Lambda , \lambda}$ defined
1186: in Chaps.~\secref{subsec:patch} and \secref{sec:sectors}.
1187: Although our manipulations are formally exact,
1188: at this point the reader is perhaps rather skeptical
1189: whether they will turn out to be useful to
1190: obtain truly non-perturbative information about the
1191: interacting many-body system\footnote{As already mentioned, in $d=1$ we have the
1192: ambitious goal to reproduce the exact solution of the Tomonaga-Luttinger model.}.
1193: After all, the use of Hubbard-Stratonovich transformation is
1194: a well-known technique in the theory of strongly
1195: correlated systems \cite{Evenson70,Hertz76,Gomes77,Kotliar86,Schulz90},
1196: and in practice it is very difficult
1197: to go beyond the saddle point approximation.
1198: An important exception is a beautiful
1199: paper by Hertz \cite{Hertz76},
1200: which has inspired the development of our functional bosonization approach.
1201: Hertz
1202: used a Hubbard-Stratonovich transformation to derive
1203: quantum Landau-Ginzburg-Wilson functionals for
1204: interacting Fermi systems,
1205: which form then the basis for a renormalization group analysis.
1206: Our fields $\phi^{\alpha}_q$ are closely related to the
1207: Hubbard-Stratonovich
1208: fields introduced by Hertz; the only difference is that
1209: our fields carry an extra patch index $\alpha$.
1210: As will be shown in Chap.~\secref{chap:agreen}, in this book we shall
1211: be able to {\it{treat the full quantum
1212: dynamics of the Hubbard-Stratonovich field non-perturbatively}} --
1213: we shall neither rely on saddle point approximations,
1214: nor on the naive perturbative calculation of fluctuation corrections
1215: around saddle points!
1216:
1217:
1218:
1219:
1220:
1221:
1222:
1223:
1224:
1225:
1226: %
1227: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1228: % End of a3hub.tex
1229: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1230: %
1231: % CUT HERE
1232: