1: \documentclass[12pt]{article}
2: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3: %\usepackage{amsmath}
4: \usepackage{amsmath,amsfonts,amscd}
5: \usepackage{bm}
6: \usepackage{graphicx}
7: \oddsidemargin -20pt
8: \textwidth 6.7in
9: \textheight 43\baselineskip
10: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
11: \arraycolsep=2pt
12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
13: \def\cA{{\cal A}}
14: \def\cD{{\cal D}}
15: \def\cE{{\cal E}}
16: \def\cH{{\cal H}} % Hilbert space
17: \def\cN{{\cal N}}
18: \def\cU{{\cal U}}
19: \def\cH{{\cal H}}
20: \def\cQ{{\cal Q}}
21: \def\cZ{{\cal Z}}
22: \def\Qp{{\cQ_p}}
23: \def\Zp{{\cZ_p}}
24: \newcommand{\R}{\mathbb{R}}
25: \def\Z{{Z\kern-.5em{Z}}}
26: \def\V{{V\kern-.7em{V}}}
27: \def\ie{{\em i.e.\, }}
28: \def\eg{{\em e.g.\, }}
29: \def\etc{{\em etc.\, }}
30: \def\d{\partial}
31: \def\bra{\langle}
32: \def\ket{\rangle}
33: \def\det{{\rm det\,}}
34: \def\lr{{\rm L}^2({\R^d})}
35: \def\Ex{{\rm M}} % Expectation
36: \def\Cov{{\rm R}} % Covariance
37: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
38: \def\bvec#1{{\bm #1}} % bold vectors
39: \def\vv{\bvec{v}} % velocity field
40: \def\vb{\bvec{b}}
41: \def\vz{\bvec{\zeta}} % vorticity field
42: \def\vk{\bvec{k}}
43: \def\vq{\bvec{q}}
44: \def\veta{\bvec{\eta}}
45: \def\vf{\bvec{f}}
46: \def\vx{\bvec{x}}
47: \def\dk#1#2{\frac{ d^{#2}{#1} }{ (2\pi)^{#2} }} % invariant measure in FT
48: \def\da#1#2{\frac{ d{#1}}{{#1}^{{#2}+1}}}
49: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
50: \begin{document}
51: \title{Wavelet based regularization for Euclidean field theory and
52: stochastic quantization}
53: \author{M.V.Altaisky \\
54: \small Joint Institute for Nuclear Research, Dubna, 141980, Russia \\
55: \small and Space Research Institute, Moscow, 117997, Russia \\
56: \small e-mail: altaisky@mx.iki.rssi.ru}
57: \date{}
58: \maketitle
59: \begin{abstract}
60: It is shown that Euclidean field theory with polynomial interaction, can be regularized using the wavelet representation of the fields.
61: The connections between wavelet based regularization and stochastic
62: quantization are considered with $\phi^3$ field theory taken as an example.
63: \end{abstract}
64: \section{Introduction}
65: The intimate connection between quantum field theory and stochastic
66: differential equations has called constant attention for
67: quite a long time \cite{Nelson1985,GJ1981}.
68: Among different aspects of this connection
69: the stochastic quantization, proposed by G.Parisi and Y.Wu \cite{PW1981},
70: is especially important, for it is believed to provide a new approach to
71: quantization of Euclidean fields alternative to path integrals and
72: canonical quantization.
73: The stochastic quantization is also particularly attractive for
74: quantization of gauge theories as it does not require gauge fixing.
75: Besides that, the stochastic quantization procedure being based on
76: random processes, and therefore on the measure theory, may be considered
77: as a measure based formalism for constructing divergence free theory from
78: beginning, instead of artificial regularization of the Riemann integrals in
79: Euclidean space.
80:
81: We know that stochastic processes often posses self-similarity.
82: The renormalization procedure used in quantum field theory is also based on
83: self-similarity. So, it is natural to use for the regularization of field
84: theories the wavelet transform (WT), the decomposition
85: with respect to the representation of the affine group.
86: In this paper two ways of wavelet based regularization are presented.
87: First, the direct substitution
88: of WT of the fields into the action functional leads to
89: a field theory with scale-dependent coupling constants. Second, the
90: WT, being substituted into the Parisi-Wu stochastic quantization
91: scheme \cite{PW1981},
92: provides a stochastic regularization with no extra vertexes introduced into the theory.
93:
94: \section{Scalar field theory on affine group}
95: The scalar field theory with the polynomial interaction $V(\phi)$
96: defined on Euclidean space $\R^d$ is
97: one of the most instructive models any textbook in field theory starts with,
98: see \eg \cite{Ramond1981}. This field theory is defined by the generating
99: functional
100: \begin{equation}
101: W_E[J] = \cN \int \cD\phi \exp\left[-S_E[\phi(x)] + \int d^dx J(x)\phi(x) \right]
102: \label{gf},
103: \end{equation}
104: where $S_E[\phi(x)]$ is Euclidean action, $\cN$ is formal
105: normalization constant. The (connected) Green functions
106: ($m$-point cumulative moments) are evaluated
107: as functional derivatives of the logarithm of generating functional
108: $W_E[J] = e^{-Z_E[J]}$:
109: \begin{equation}
110: G_m(x_1,\ldots x_m) \equiv \bra \phi(x_1)\ldots\phi(x_m)\ket = -
111: \left. \frac{\delta^m}{\delta J(x_1)\ldots\delta J(x_m)}\right|_{J=0}
112: \ln W_E[J].
113: \label{gfm}
114: \end{equation}
115: The generating functional \eqref{gf}
116: describes a quantum field $\phi(x)$
117: with the action
118: \begin{equation}
119: S_E[\phi] =\int_{\R^d} d^d x \left[{
120: \frac{1}{2}(\d_\mu\phi)^2+ \frac{m^2}{2}\phi^2 + \lambda V(\phi)}\right].
121: \label{fe}
122: \end{equation}
123: Alternatively, the theory of quantum
124: field in Euclidean space is equivalent to the theory of classical fluctuating
125: field with the Wiener probability measure
126: $\cD P = e^{-S_E[\phi]}\cD\phi$. In this case
127: $m^2 = |T-T_c|$ is the deviation from critical temperature and
128: $\lambda$ is the fluctuation interaction strength.
129:
130: In the simplest case of a scalar field with the fourth power interaction
131: \begin{equation}
132: S_E[\phi] =\int d^d x \left[{
133: \frac{1}{2}(\d_\mu\phi)^2+ \frac{m^2}{2}\phi^2 + \frac{\lambda}{4!}\phi^4
134: }\right]
135: \label{f4l}
136: \end{equation}
137: the theory is often referred to as the Ginsburg-Landau model for its ferromagnetic
138: applications.
139: The $\phi^3$ theory is also a useful toy model used, for instance, to describe
140: electron-phonon interactions \cite{AGD}.
141:
142: The straightforward way to calculate the Green functions is to factorize the
143: interaction part of generation functional \eqref{gf} in the form
144: \begin{equation}
145: W_E[J] = \exp\left[-V\left(\frac{\delta}{\delta J}\right) \right]
146: \exp\left(-\frac{1}{2}JD^{-1}J \right),
147: \quad D = -\d^2 +m^2
148: \end{equation}
149: The perturbation expansion is then evaluated in $k$-space, where
150: $$
151: {\widehat D}^{-1}(k) = \frac{1}{k^2+m^2}.
152: $$
153: From group theory point of view
154: the reformulation of a field theory from the coordinate representation
155: $\phi(x)$ to the momentum representation $\hat\phi(k)$ by means of
156: Fourier transform
157: $\hat\phi(k)=\int_{\R^d} e^{\imath k x} \phi(x) d^dx$, is only a
158: particular case of decomposition
159: of a function with respect to representation of a Lie group $G$.
160: For the Fourier transform $G$ is just an abelian group of translations,
161: but other groups
162: may be used as well, depending on the physics of a particular problem.
163:
164: For a locally compact Lie group $G$ acting transitively
165: on the Hilbert space $\cH$
166: it is possible to decompose vectors $\phi\in\cH$ with respect to the
167: square-integrable representations $U(g)$ of the group $G$
168: \cite{Carey76,DM1976}:
169: \begin{equation}
170: |\phi\ket = C_\psi^{-1} \int_G |U(g)\psi\ket d\mu(g)\bra\psi U^*(g)|\phi\ket,
171: \label{pu}
172: \end{equation}
173: where $d\mu(g)$ is the left-invariant Haar measure on $G$.
174: The normalization constant $C_\psi$ is determined by the norm of the
175: action of $U(g)$ on the fiducial vector $\psi\!\in\!\cH$, \ie any
176: $\psi\!\in\!\cH$, that satisfies the admissibility condition
177: \begin{equation}
178: C_\psi := \|\psi \|^{-2} \int_{g\in G} |\bra\psi|U(g)\psi \ket|^2 d\mu(g)
179: <\infty,
180: \label{adc}
181: \end{equation}
182: can be used as a basis of wavelet decomposition.
183:
184: In operator notation the decomposition \eqref{pu} is also referred to as
185: a partition of unity with respect to the representation $U(g)$:
186: $$\hat1 = C_\psi^{-1} \int_G |U(g)\psi\ket d\mu(g)\bra\psi U^*(g)|.$$
187:
188: It is apparently possible to generalize the
189: Feynman-Dyson perturbation expansion in a given field theory $S_E(\phi)$
190: by using decomposition \eqref{pu} for non-abelian groups.
191:
192: For definiteness, let us consider the fourth power
193: interaction model with the (Euclidean) action functional
194: \begin{equation}
195: \begin{array}{lcl}
196: S[\phi] &=& \frac{1}{2}\int_{\R^d} \phi(x_1)D(x_1,x_2)\phi(x_2)dx_1dx_2 \\
197: &+& \frac{\lambda}{4!}\int V(x_1,x_2,x_3,x_4)\phi(x_1)\phi(x_2)\phi(x_3)
198: \phi(x_4)dx_1dx_2 dx_3dx_4,
199: \label{Sphi}
200: \end{array}
201: \end{equation}
202: where $D(x_1,x_2)$ is the inverse propagator of the free model ($\lambda\!=\!0$).
203: Using the notation
204: $$U(g)|\psi\ket \equiv |g,\psi\ket, \quad \bra\phi|g,\psi\ket \equiv \phi(g),
205: \quad \bra g_1,\psi|D|g_2,\psi\ket \equiv D(g_1,g_2),
206: $$
207: we can rewrite the generating functional \eqref{gf} of the field theory
208: with action \eqref{Sphi} in the form
209: \begin{equation}
210: \begin{array}{lcl}
211: W_G[J(g)] &=& \int \cD\phi(g) \exp\Bigl(
212: -\frac{1}{2}\int_G \phi(g_1)D(g_1,g_2)\phi(g_2)d\mu(g_1) d\mu(g_2) \\
213: &-&\frac{\lambda_0}{4!}\int_G \tilde V(g_1,g_2,g_3,g_4)
214: \phi(g_1)\phi(g_2)\phi(g_3)\phi(g_4)
215: d\mu(g_1) d\mu(g_2)d\mu(g_3) d\mu(g_4) \\
216: &+& \int_G J(g)\phi(g)d\mu(g)
217: \Bigr),
218: \end{array}
219: \label{gfi4}
220: \end{equation}
221: where $$
222: \begin{array}{lcl}\tilde V(g_1,g_2,g_3,g_4)&=&\int_{\R^d}V(x_1,x_2,x_3,x_4)
223: (U(g_1)\psi(x_1))(U(g_2)\psi(x_2)) \\
224: & & (U(g_3)\psi(x_3))(U(g_4)\psi(x_4))dx_1dx_2dx_3dx_4
225: \end{array}
226: $$
227: is the result of application of the wavelet transform in all arguments
228: of $V$.
229: %$$\tilde\phi(g) := \int\overline{U(g)\psi(x)} \phi(x)dx$$
230: %to $V(x_1,x_2,x_3,x_4)$ in all arguments $x_1,x_2,x_3,x_4$.
231:
232: At this point we ought mention a real-world problem arising from substitution of the functional integration over physical fields by
233: integration over the fields defined on a Lie group $G$. For the abelian group
234: of translations (homeomorphic to $\R^d$) the inverse transform from Euclidean
235: space to Minkovski space by changing $t$ into $\imath\!t$ does not yield any
236: problems with causality, since the chronological ordering of operators is
237: easily imposed. For non-abelian groups it is not clear how to order, or how
238: to commute, the operator-valued fields at different points of the Lie
239: group manifold $[\phi(g_1),\phi(g_2)]=?$. This is an obstacle
240: preventing quantization of gauge theories using wavelets
241: \cite{Federbush1995}.
242: That is why we will further consider only $c$-valued fields, bearing in
243: mind the interpretation of Euclidean field theory in terms of the
244: models of statistical mechanics.
245:
246: Let us turn to the particular case of the affine group that is of our
247: principle interest
248: $$
249: x'=aR(\theta)x+b,\quad R(\theta) \in SO_d, x,x',b \in \R^d
250: .
251: $$
252: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
253: Hereafter we assume the basic wavelet $\psi$ is invariant under
254: $SO_d$ rotations $\psi(\mathbf{x})=\psi(|\mathbf{x}|)$ and drop
255: the angular part of the measure for simplicity.
256: After this simplifying assumption, the left-invariant
257: Haar measure on affine group is
258: $d\mu(a,\vb)=\frac{da d^d\vb}{a^{d+1}}$. The representation $U(g)$ induced
259: by a basic wavelet $\psi(x)$ is
260: \begin{equation}
261: g:\vx'=a\vx+\vb,\quad U(g)\psi(\vx)=a^{-d/2}\psi\left( \frac{\vx-\vb}{a}\right).
262: \label{rag}
263: \end{equation}
264: The (bold) vector notation is dropped where it does not lead to confusion.
265: The last thing we need to construct the generating functional of a
266: field theory on affine group is to substitute wavelet decomposition
267: \begin{equation}
268: \phi(x)= C_\psi^{-1}\int a^{-d/2}\psi\left(\frac{x-b}{a}\right)\phi_a(b)
269: \frac{dad^db}{a^{d+1}},
270: \label{iwt}
271: \end{equation}
272: into Euclidean action $S_E[\phi]$. Here we use normalization
273: for the rotationally invariant wavelets
274: \begin{equation}
275: C_\psi = \int \frac{|\hat\psi(k)|^2}{S_d|k|^d}d^dk,
276: \label{adcf}
277: \end{equation}
278: where the area of the unit sphere in $d$ dimensions $S_d$ has come
279: from rotation symmetry.
280: The wavelet coefficients
281: \begin{equation}
282: \phi_a(b) = \int a^{-d/2}\bar\psi\left(\frac{x-b}{a}\right)\phi(x)d^dx
283: \label{wtl2}
284: \end{equation}
285: represent the snapshot of the field $\phi(x)$ taken at the scale $a$ with
286: the aperture function $\psi(x)$, and will be referred to as scale components
287: of the field $\phi$. See e.g. \cite{Chui} for more details
288: on wavelets.
289:
290: The restriction imposed by the admissibility condition \eqref{adc} on
291: the fiducial vector $\psi$ (the basic wavelet) is rather loose:
292: only the finiteness of the integral $C_\psi$ given by \eqref{adcf} is
293: required.
294: This practically implies only that $\int \psi(x)dx =0$ and that
295: $\psi(x)$ has compact support. For this reason the wavelet transform
296: \eqref{wtl2}
297: can be considered as a microscopic slice of the function $\phi(x)$ taken
298: at a position $b$ and resolution $a$ with ``aperture'' $\psi$. Of course,
299: each particular aperture $\psi(x)$ has its own view, but the physical
300: observables should be independent on it. In practical applications of
301: WT either of the derivatives of the Gaussian
302: $\psi_n(x) = (-1)^n d^n/dx^n e^{-x^2/2}$ are often used, but for the
303: purpose of this present paper only the admissibility condition is important but not
304: the shape of $\psi(x)$.
305:
306: So, for the case of decomposition of a scalar field in $\R^d$
307: with respect to affine group \eqref{rag}, the inverse free field propagator matrix
308: element is
309: \begin{eqnarray*}
310: \bra a_1,b_1;\psi | D | a_2, b_2; \psi\ket = \int d^dx (a_1a_2)^{-\frac{d}{2}}
311: \bar\psi \left( \frac{x-b_1}{a_1} \right) D \psi\left(\frac{x-b_2}{a_2}\right) \\
312: = \int \dk{k}{d} e^{ik(b_1-b_2)} (a_1a_2)^{\frac{d}{2}} \overline{\hat \psi(a_1 k)}
313: (k^2+m^2) \hat \psi(a_2 k) \\
314: \equiv \int \dk{k}{d} e^{ik(b_1-b_2)} D(a_1,a_2,k).
315: \end{eqnarray*}
316: Assuming the homogeneity of the free field in space coordinate, i.e. that
317: matrix elements depend only on the differences $(b_1\!-\!b_2)$ of the
318: positions,
319: but not the positions themselves, we can use $(a,k)$ representation:
320: \begin{eqnarray}
321: \nonumber D(a_1,a_2,k) &=& a_1^{d/2} \overline {\hat \psi(a_1 k)} (k^2+m^2)
322: a_2^{d/2}\hat \psi(a_2 k) \\
323: D^{-1}(a_1,a_2,k) &=& a_1^{d/2} \overline {\hat \psi(a_1 k)}
324: \left( \frac{1}{k^2+m^2} \right)
325: a_2^{d/2}\hat \psi(a_2 k)\\
326: \nonumber d\mu(a,k) &=& \dk{k}{d}\da{a}{d}.
327: \end{eqnarray}
328: Thus, we have the same diagram technique as usually, but with extra
329: ``wavelet'' term $a^{d/2}\hat \psi(a k)$ term on each line and
330: the integration over $d\mu(a,k)$ instead of $dk$.
331:
332: Now, turning back to the coordinate representation \eqref{wtl2}, where
333: $a$ is the resolution (``window width'') and recalling the power law
334: dependence resulted from the Wilson expansion, we can define the
335: $\phi^4$ model on affine group, with the coupling constant
336: dependent on scale. The simplest case of the fourth power interaction
337: of this type is
338: \begin{equation}
339: V_{int}[\phi] =\int \frac{\lambda(a)}{4!} \phi^4_a(b) d\mu(a,b),
340: \quad \lambda(a)\sim a^\nu.
341: \label{vint}
342: \end{equation}
343:
344: The one-loop order contribution to the Green function $G_2$
345: in the theory with interaction \eqref{vint}
346: can be easily evaluated \cite{Alt2001} by integration over a scalar
347: variable $z = ak$:
348: \begin{equation}
349: \int \frac{a^\nu a^d |\hat\psi(ak)|^2 }{k^2+m^2}\dk{k}{d}\da{a}{d}
350: = C_\psi^{(\nu)} \int \dk{k}{d} \frac{k^{-\nu}}{k^2+m^2},
351: \end{equation}
352: where $$C_\psi^{(\nu)}=\int |\hat\psi(z)|^2 z^{\nu-1}dz.$$
353: Therefore, there are no UV divergences for $\nu>d\!-\!2$.
354:
355: However, the positive values of $\nu$ mean that the interaction strengths at large
356: scales and diminishes at small. This is a kind of asymptotically free
357: theory that is hardly appropriate say to magnetic systems. What is required
358: instead is a theory with the interaction vanishing outside a given
359: domain of scales. Such models will be presented in the next section
360: by means of scale-dependent stochastic quantization.
361:
362: \section{Stochastic quantization with wavelets}
363: The method of stochastic quantization
364: first introduced by G.Parisi, and Y.Wu \cite{PW1981} consists in
365: substitution of functional integration by averaging over certain random
366: process.
367: Let $S_E[\phi]$ be the action Euclidean field theory \eqref{fe} in $\R^d$.
368: Then, instead of direct calculation of the Green functions
369: \eqref{gfm} from the generation functional \eqref{gf}, it is possible to
370: introduce a fictitious
371: ``time'' variable $\tau$, make the quantum fields into stochastic fields
372: $\phi(x) \to \phi(x,\tau),x\!\in\!\R^d,\tau\!\in\!\R $ and evaluate the
373: moments
374: $\langle \phi(x_1,\tau_1) \ldots \phi(x_m,\tau_m) \rangle$ by averaging over
375: a random process $\phi(x,\tau,\cdot)$ governed by the Langevin equation
376: \begin{equation}
377: \frac{\d\phi(x,\tau)}{\d\tau} +\frac{\sigma^2}{2}
378: \frac{\delta S}{\delta\phi(x,\tau)} = \eta(x,\tau).
379: \label{le}
380: \end{equation}
381: The gaussian random force $\eta$, that drives the Langevin equation
382: \eqref{le},
383: has zero mean and is $\delta$-correlated in both the $\R^d$ coordinate and
384: the fictitious time:
385: \begin{equation}
386: \langle \eta(x,\tau)\rangle=0, \quad
387: \langle \eta(x,\tau)\eta(x',\tau') \rangle = \sigma^2
388: \delta(x-x')\delta(\tau-\tau').
389: \label{gforce}
390: \end{equation}
391: The physical Green functions are obtained by taking the steady state limit
392: \begin{equation}
393: G(x_1,\ldots,x_m) = \lim_{\tau\to\infty}
394: \langle \phi(x_1,\tau) \ldots \phi(x_m,\tau) \rangle
395: \label{ssl}
396: \end{equation}
397: either in the solution of the Fokker-Planck equation, or in the stochastic
398: perturbation expansion given by stochastic generating functional
399: \begin{equation}
400: W[J(x,\tau)] = \int \cD\phi \cD\eta
401: \exp\left[{
402: \int d^dx d\tau \left( J\phi - \frac{\eta^2}{2\sigma^2}\right)
403: }\right]
404: \delta \left[{
405: \frac{\d\phi}{\d\tau}+\frac{\sigma^2}{2}\frac{\delta S}{\delta\phi}-\eta
406: }\right].
407: \label{cfr}
408: \end{equation}
409: The stochastic quantization procedure has been considered as
410: perspective candidate for the regularization of gauge theories,
411: for it does not require gauge fixing. However
412: $\delta$-correlated Gaussian random force in the Langevin equation still
413: yields singularities in the perturbation theory. For this
414: reason a number of modifications based
415: on the noise regularization
416: $\eta(x,\tau) \to
417: \int dy R_{xy}(\d^2)\eta(y,\tau)$ have been proposed
418: \cite{BGZ1984,BHST1987,IP1988}.
419: The introduction of a colored noise
420: $\bra\eta(x,\tau)\eta(x',\tau')\ket = \delta(x-x')|\tau-\tau'|^n$ instead of
421: $\delta$-correlated one can be also used to avoid UV divergences
422: \cite{BGZ1984,Alfaro1985}. Other methods of regularization
423: in stochastically quantized theories were also considered
424: \cite{NY1983,CKS1988}.
425:
426: In this paper we intend to revive the method of stochastic quantization by
427: applying the continuous wavelet transform to both the fields $\phi(x,\tau)$ and
428: the random force $\eta(x,\tau)$. It will be shown that ultra-violet divergences
429: in so constructed perturbation expansion can be eliminated for some particular
430: choice of the random force correlator taken in wavelet space. Our method is rather
431: general and can be applied to any stochastic systems described by the Langevin
432: equation. The idea of our method was proposed in \cite{AltDAN03} and consists in the
433: following.
434:
435: Instead
436: of the usual space of the random functions
437: $f(x,\cdot)\in(\Omega, \cA, P)$, where
438: $f(x,\omega)\in \lr$ for each given realization $\omega$ of the random process, we go
439: to the multi-scale representation provided by the continuous wavelet
440: transform \eqref{wtl2}:
441: \begin{equation}
442: W_\psi(a,\vb,\cdot) = \int |a|^{-\frac{d}{2}}
443: \overline{\psi\left(\frac{\vx-\vb}{a}\right)}f(\vx,\cdot)d^dx.
444: \label{wtr}
445: \end{equation}
446: Since the structure of divergences and the localization of the solutions
447: are determined by the spatial part of the random
448: force correlator (see \eg \cite{KPZ1986}), the wavelet transform is
449: performed only in the
450: {\em spatial} argument $x\!\in\!\R^d$ of the dynamical variable $\phi(x,\tau)$,
451: but not in its fictitious time argument.
452:
453: The inverse wavelet transform
454: \begin{equation}
455: f(\vx,\cdot) = C_\psi^{-1} \int |a|^{-\frac{d}{2}}
456: \psi\left( \frac{\vx-\vb}{a}\right) W_\psi(a,\vb,\cdot)
457: \frac{dad\vb}{a^{d+1}}
458: \label{fa}
459: \end{equation}
460: reconstructs the common random process as a sum of its scale components,
461: \ie projections onto different resolution spaces.
462:
463: The use of the scale components instead of
464: the original stochastic process provides an extra analytical
465: flexibility of the
466: method: there exist more than one set of random functions
467: $W(a,\vb,\cdot)$ the images of which have
468: coinciding correlation functions in the space of $f(x,\cdot)$.
469: It is easy to check that the random process generated by wavelet
470: coefficients having in $(a,k)$ space the correlation function
471: \begin{equation}
472: \bra \widehat W(a_1,k_1) \widehat W(a_2,k_2)\ket
473: = C_\psi (2\pi)^{d} \delta^d(k_1+k_2)
474: a_1^{d+1} \delta(a_1-a_2) D_0
475: \label{w2}
476: \end{equation}
477: has the same correlation function as white noise has:
478: $$
479: \begin{array}{lcl}
480: \bra f(x_1) f(x_2) \ket &=& D_0 \delta^d(x_1-x_2), \\
481: \bra {\hat f}(k_1) \hat f(k_2)\ket &=& (2\pi)^d D_0 \delta^d(k_1+k_2) \label{gn}\\
482: \nonumber \bra \widehat W_\psi(a_1,k_1) \widehat W_\psi(a_2,k_2)\ket &=&
483: (2\pi)^d D_0 \delta^d(k_1+k_2)
484: (a_1 a_2)^{d/2} \overline{\hat \psi(a_1 k_1) \hat\psi(a_2 k_2)}.
485: \end{array}
486: $$
487: Therefore, starting from a given random process in the space
488: of scale-dependent functions $W(a,\vb,t)$, rather than in a common
489: space of square
490: integrable functions $f(\vx,t)$, we can design a narrow band
491: forcing with no contradictions
492: to other physical constraints. This can be done by applying
493: the requirement $W(a,\vb,t)\!\to\!0$ for all $a$ outside a certain domain
494: of scales $[a_{min},a_{max}]$.
495:
496: Now let us turn to the stochastic quantization of the $\phi^3$ theory with
497: the help of scale-dependent noise designed in the above described way.
498: The Euclidean action of the $\phi^3$ theory is
499: \begin{equation}
500: S_E[\phi(x)]= \int d^d x \left[\frac{1}{2} (\d\phi)^2 + \frac{m^2}{2}\phi^2
501: + \frac{\lambda}{3!}\phi^3 \right].
502: \label{pf3}
503: \end{equation}
504: Therefore the corresponding Langevin equation used for stochastic
505: quantization is written as
506: \begin{equation}
507: \frac{\d\phi(x,\tau)}{\d\tau} + \frac{D_0}{2}\left[
508: -\Delta \phi + m^2\phi + \frac{\lambda}{2!} \phi^2 \right] = \eta(x,\tau),
509: \label{le3}
510: \end{equation}
511: where in common case the $\delta$-correlated random force is applied
512: $$
513: \bra \eta(x,\tau) \eta(x',\tau')\ket = D_0 \delta^d(x-x')\delta(\tau-\tau').
514: $$
515: Following \cite{AltDAN03}, we perform continuous wavelet transform of the
516: fields and forces in the spatial coordinate
517: \begin{equation}
518: \phi(x) = C_\psi^{-1} \int \exp(\imath(\vk\vx-\omega\tau)) a^{\frac{d}{2}}
519: \hat\psi(a\vk) \hat \phi(a,k) \dk{k}{d+1} \da{a}{d},
520: \label{zt}
521: \end{equation}
522: using hereafter apparent $(d\!+\!1)$ dimensional notation $x=(\vx,\tau),k=(\vk,\omega)$.
523: To generalize the scale-dependent force \eqref{w2} we introduce dependence on
524: $k$ in the force correlator
525: \begin{equation}
526: \begin{array}{lcl}
527: \bra \widehat \eta(a_1,k_1) \widehat \eta(a_2,k_2)\ket
528: &=& C_\psi (2\pi)^{d+1} \delta^{d+1}(k_1+k_2)
529: a_1^{d+1} \delta(a_1-a_2) 2D(a_2,\vk_2),\\ \bra \hat\eta(a,k) \ket &=& 0,
530: \end{array}
531: \label{sfnc}
532: \end{equation}
533: which coincides with \eqref{w2} if $D(a,\vk)=const$, and is therefore capable
534: of giving white noise in that limiting case. After substitution of
535: \eqref{zt} into the Langevin equation \eqref{le3} we yield the stochastic
536: integro-differential equation for stochastic fields $\hat\phi(a,k,\cdot)$
537: \begin{equation}
538: \begin{array}{l}
539: (-\imath\omega + \vk^2 + m^2)\hat\phi(a,k) = \hat\eta(a,k)
540: -\frac{\lambda}{2} a^{\frac{d}{2}} \overline{\hat\psi(a\vk)}
541: C_\psi^{-2} \int (a_1 a_2)^\frac{d}{2}\\
542: \hat\psi(a_1\vk_1) \hat\psi(a_2(\vk-\vk_1))
543: \hat \phi(a_1,k_1) \hat \phi(a_2,k-k_1) \dk{k_1}{d+1} \da{a_1}{d}\da{a_2}{d}.
544: \end{array}
545: \label{phi3w}
546: \end{equation}
547:
548: Starting from the zero-th order approximation
549: $\hat \phi_0(a,k) = G_0(k) \hat\eta(a, k)$ with the bare Green function
550: $$G_0(k) = \frac{1}{-\imath\omega + \vk^2 + m^2}$$
551: and iterating the integral equation \eqref{phi3w},
552: in one loop approximation we get the correction to the stochastic
553: Green function
554: \begin{equation}
555: G(k) = G_0(k) + \lambda^2 G_0^2(k) \int \dk{q}{d+1} \Delta(\vq)
556: |G_0(q)|^2 G_0(k-q) + O(\lambda^4),
557: \label{phi3G1}
558: \end{equation}
559: where $\Delta(\vk)$ is the scale averaged correlator \eqref{dak}:
560: \begin{equation}
561: \Delta(\vk) \equiv C_\psi^{-1} \int \frac{da}{a} |\hat\psi(a\vk)|^2 2D(a,\vk)
562: \label{dak}
563: \end{equation}
564: In the same way all other momenta \eqref{gfm} can be evaluated. Thus the
565: common stochastic
566: diagram technique is reproduced with the scale-dependent random force
567: \eqref{sfnc} instead of the standard one.
568: The 1PI diagramms corresponding to the stochastic Green function decomposition
569: \eqref{phi3G1} are shown in Fig.~\ref{gf:pic}.
570: \begin{figure}[h]
571: \centering\includegraphics[width=8cm]{gf.eps}
572: \caption{Diagramm expansion of the stochastic Green function in $\phi^3$-model}
573: \label{gf:pic}
574: \end{figure}
575: Similarly, to stochastic Green function \eqref{phi3G1}, the one-loop
576: contribution to the stochastic pair correlation function can be evaluated in
577: $(a,k)$ space
578: \begin{equation}
579: \bra \hat\phi(a_i,k)\hat\phi(a_f,-k) \ket \equiv C(a_i,a_f,k)
580: = C_0(a_i,a_f,k) + \lambda^2 C_2(a_i,a_f,k)
581: + O(\lambda^4),
582: \label{scf}
583: \end{equation}
584: where $$C_0(a_i,a_f,k)=\Delta(\vk)|G_0(k)|^2 (a_ia_f)^{d/2}
585: \overline{\hat\psi(a_i\vk)\hat\psi(-a_f\vk)}.$$
586: The one-loop contribution to the pair correlator is
587: \begin{eqnarray}
588: C_2(a_i,a_f,k) &=& \frac{1}{2} |G_0(k)|^2 (a_i a_f)^{d/2}
589: \overline{\hat \psi(a_i\vk)\hat \psi(-a_f\vk)} \label{phi3C1} \\
590: \nonumber & & \int \dk{q}{d+1} |G_0(q)|^2|G_0(k-q)|^2
591: \Delta(\vq) \Delta(\vk-\vq).
592: \end{eqnarray}
593: The 1PI diagramms corresponding to the stochastic pair correlator decomposition
594: \eqref{scf} are shown in Fig.~\ref{cf:pic}.
595: \begin{figure}[h]
596: \centering\includegraphics[width=8cm]{cf.eps}
597: \caption{Diagramm expansion of the stochastic pair correlation function in $\phi^3$
598: model}
599: \label{cf:pic}
600: \end{figure}
601: An important type of scale-dependent forcing is that acting on a single
602: scale:
603: \begin{equation}
604: D(a,\vk)=\delta(a-a_0) D(\vk).
605: \label{sb}
606: \end{equation}
607: In some sense this mimics a field theory on a grid with a fixed mesh $a_0$.
608:
609: As an example, let us consider the $\phi^3$ model with the single-scale
610: force \eqref{sb} and the ``Mexican hat'' used as a basic wavelet
611: \begin{equation}
612: \hat \psi(k) = (2\pi)^{d/2} (-\imath \vk)^2 \exp(-\vk^2/2), \quad
613: C_\psi = (2\pi)^d.
614: \label{mh}
615: \end{equation}
616: This gives the effective force correlator
617: \begin{equation}
618: \Delta(\vq) = 2 \frac{(a_0\vq)^4}{a_0} e^{-(a_0\vq)^2}D(\vq)
619: \label{efc2}.
620: \end{equation}
621: The loop integrals taken with this effective force
622: correlator \eqref{efc2} can be easily seen to be free of
623: ultra-violet divergences. The IR divergences also become milder
624: because of the wavelet power factor $(a_0\vq)^4$.
625:
626: In fact, substituting \eqref{efc2} into
627: expressions for one-loop contributions to the stochastic Green
628: function and the correlation functions, \eqref{phi3G1} and \eqref{phi3C1},
629: respectively, we get
630: \begin{eqnarray}
631: G_2(k) &=& G_0^2(k) I_{G2}, \\
632: \nonumber I_{G2} &=& \int \dk{\vq}{d} \Delta(\vq) \int_{-\infty}^\infty
633: \frac{d\Omega}{2\pi} \frac{1}{\Omega^2 + (\vq^2+m^2)^2}
634: \frac{1}{-\imath(\omega-\Omega) + (\vk-\vq)^2+m^2} \\
635: C_2(a_i,a_f,k) &=& \frac{1}{2} |G_0(k)|^2 (a_i a_f)^{d/2}
636: \overline{\hat \psi(a_i\vk)\hat \psi(-a_f\vk)} I_{C2},\\
637: \nonumber I_{C2} &= & \int \dk{\vq}{d} \Delta(\vq) \Delta(\vk-\vq)
638: \int_{-\infty}^\infty \frac{d\Omega}{2\pi} \frac{1}{\Omega^2 + (\vq^2+m^2)^2}
639: \frac{1}{(\omega-\Omega)^2 + [(\vk-\vq)^2+m^2]^2}.
640: \end{eqnarray}
641: For the case of single-scale forcing
642: \eqref{efc2} the exponential factor in $\Delta(\vq)$ will suppress any
643: power divergences comming from the Green functions. In fact, in a stationary
644: limit ($\omega\!\to\!0$), after the integration over the frequency
645: variable $\Omega$, we get
646: \begin{eqnarray}
647: \lim_{\omega\to0} I_{G2} &=&
648: \int \dk{\vq}{d} \Delta (\vq) \frac{1}{2(\vq^2+m^2)}
649: \cdot\frac{1}{\vq^2+(\vk-\vq)^2+2m^2} \\
650: \lim_{\omega\to0} I_{C2} &=&
651: \int \dk{\vq}{d} \Delta (\vq) \frac{1}{2(\vq^2+m^2)((\vk-\vq)^2+m^2)}
652: \cdot\frac{1}{\vq^2+(\vk-\vq)^2+2m^2}.
653: \end{eqnarray}
654: An explicit calculation gives the frequency dependence:
655: \begin{eqnarray*}
656: I_{G2}&=& \frac{1}{2B}\frac{1}{(A-B)-\imath\omega}
657: + \frac{1}{(\omega+\imath A)^2+B^2} \\
658: I_{C2}&=& \frac{1}{2A} \frac{1}{(\omega+\imath A)^2+B^2}
659: + \frac{1}{2B} \frac{1}{(\omega-\imath B)^2+A^2}
660: \end{eqnarray*}
661: where $$A = (\vk-\vq)^2 + m^2, \quad B = \vq^2 + m^2.$$
662:
663: For the $\phi^4$ and higher polynomial models the above described method
664: can be applied in a straightforward way.
665: \section{Conclusion}
666: In this paper, that is extended version of \cite{AltG24}, we have presented
667: a new approach to regularization of field theories based on wavelet
668: decomposition of the fields.
669: The idea of using the affine group in quantum field theory is not
670: new. It is in the basis of the renormalization group method the modern
671: quantum field theory resides on. However, the direct use of the
672: {\em representations} of affine group as a basis in the space of wave
673: functions can provide a new perspectives in construction of a
674: divergence free field theory. From the point of view of functional analysis
675: this is
676: an extension of the space of square integrable functions $\phi(x)$ to
677: the space of scale-dependent functions $\phi_a(x)$, representing
678: a snapshots of the field $\phi$ at a given resolutions $a$. The usage of
679: functions depending on scale suggests that the interaction potential should
680: be scale-dependent too. Such potentials having been de facto already in use
681: in renormalization group technique, can be directly incorporated into quantum
682: field theory from beginning. One of means to do it is the method of
683: scale dependent stochastic quantization proposed in this paper.
684: Although direct construction of the quantum field theory on the affine group
685: presented in the beginning of this paper is also possible.
686: \section*{Acknowledgement}
687: The author has benefited from discussions with Profs. S.Chaturvedi,
688: H.H\"uffel, A.K.Kapoor, V.B.Priezzhev and V.Srinivasan. This work
689: is supported in part by Russian Foundation for Basic Research,
690: project 03-01-00657.
691: discussions.
692:
693: \begin{thebibliography}{20}
694:
695: \bibitem{Nelson1985}
696: E.~Nelson.
697: \newblock {\em Quantum fluctuations}.
698: \newblock Princeton University Press, New Jersey, 1985.
699:
700: \bibitem{GJ1981}
701: J.~Glim and A.~Jaffe.
702: \newblock {\em Quantum physics}.
703: \newblock Springer-Verlag, NY, 1981.
704:
705: \bibitem{PW1981}
706: G.~Parisi and Y.-S. Wu.
707: \newblock Perturbation theory without gauge fixing.
708: \newblock {\em Scientica Sinica}, 24:483, 1981.
709:
710: \bibitem{Ramond1981}
711: P.~Ramond.
712: \newblock {\em Field Theory: A modern Primer}.
713: \newblock Benjamin/Cummings Publishing Company,Inc., Massachussets, 1981.
714:
715: \bibitem{AGD}
716: A.A. Abrikosov, L.P. Gorkov, and I.E. Dzaloshinskij.
717: \newblock {\em Quantum field theory methods in statistical physics}.
718: \newblock Dobrosvet, Moscow, second edition, 1998.
719: \newblock (In Russian).
720:
721: \bibitem{Carey76}
722: A.L. Carey.
723: \newblock Square-integrable representations of non-unimodular groups.
724: \newblock {\em Bull. Austr. Math. Soc.}, 15:1--12, 1976.
725:
726: \bibitem{DM1976}
727: M.~Duflo and C.C. Moore.
728: \newblock On regular representations of nonunimodular locally compact group.
729: \newblock {\em J. Func. Anal.}, 21:209--243, 1976.
730:
731: \bibitem{Federbush1995}
732: P.~Federbush.
733: \newblock A new formulation and regularization of gauge theories using a
734: non-linear wavelet expansion.
735: \newblock {\em Progr. Theor. Phys.}, 94:1135--1146, 1995.
736:
737: \bibitem{Chui}
738: C.K. Chui.
739: \newblock {\em An introduction to wavelets}.
740: \newblock Academic press, inc, 1991.
741:
742: \bibitem{Alt2001}
743: M.V. Altaisky.
744: \newblock $\phi^4$ field theory on a lie group.
745: \newblock In B.G. Sidharth and M.V. Altaisky, editors, {\em Frontiers of
746: fundamental physics 4}, pages 124--128, NY, 2001. Kluver Academic / Plenum
747: Publishers.
748:
749: \bibitem{BGZ1984}
750: J.~Breit, S.~Gupta, and A.~Zaks.
751: \newblock Stochastic quantization and regularization.
752: \newblock {\em Nuclear Physics B}, 233:61--68, 1984.
753:
754: \bibitem{BHST1987}
755: Z.~Bern, M.~Halpern, L.~Sadun, and C.~Taubes.
756: \newblock Continuum regularization of quantum field theory (i). {S}calar
757: prototype.
758: \newblock {\em Nuclear Physics B}, 284:1--91, 1987.
759:
760: \bibitem{IP1988}
761: R.~Iengo and S.~Pugnetti.
762: \newblock Stochastic quantization, non-{M}arkovian reqularization and
763: renormalization.
764: \newblock {\em Nuclear Physics B}, 300, 1988.
765:
766: \bibitem{Alfaro1985}
767: J.~Alfaro.
768: \newblock Stochastic analytic regularization.
769: \newblock {\em Nuclear Physics B}, 253(3/4):464--476, 1985.
770:
771: \bibitem{NY1983}
772: M.~Namiki and Y.~Yamanaka.
773: \newblock Stochastic quantization method in operator formalism.
774: \newblock {\em Progr. Theor. Phys.}, 69(6):1764--1793, 1983.
775:
776: \bibitem{CKS1988}
777: S.~Chaturvedi, A.K. Kapoor, and V.~Srinivasan.
778: \newblock Renormalization of stochastically quantized fields.
779: \newblock {\em Int. J. Mod. Phys. A}, 3(1):163--165, 1988.
780:
781: \bibitem{AltDAN03}
782: M.V. Altaisky.
783: \newblock Langevin equation with scale-dependent noise.
784: \newblock {\em Doklady Physics}, 48(9):478--480, 2003.
785:
786: \bibitem{KPZ1986}
787: M.~Kardar, G.~Parisi, and Y.-C. Zhang.
788: \newblock Dynamic scaling of growing interfaces.
789: \newblock {\em Phys. Rev. Lett.}, 56(9):889--892, 1986.
790:
791: \bibitem{AltG24}
792: M.~Altaisky.
793: \newblock Wavelet based regularization for {E}uclidean field theory.
794: \newblock In J.-P. Gazeau, R.~Kerner, J.-P. Antoine, and S.~Metens, editors,
795: {\em Proc. Int. Conf. Group24: {P}hysical and mathematical aspects of
796: symmetries}, Bristol, 2003. IOP Publishing.
797:
798: \end{thebibliography}
799:
800:
801: \end{document}
802:
803:
804:
805:
806: