Content uploaded by M. V. Altaisky
Author content
All content in this area was uploaded by M. V. Altaisky on Jan 07, 2014
Content may be subject to copyright.
arXiv:hep-th/0305167v2 30 Oct 2003
Wavelet based regularization for Euclidean
field theory
∗
M V Alta isky
Joint Institute for Nuclear Research, Dubna, 141980, Russia;
and Space Research Institute, Moscow, 117997, Russia;
e-mail: altaisky@mx.iki.rssi.ru
Sep 20, 2002
Abstract
It is shown that Euclidean field theory with polynomial interac-
tion, can be regularized using the wavelet representation of the fields.
The conn ections between wavelet based regularization and stochastic
quantization are considered.
1 Introduction
The connections between quantum field theory and stochastic differential
equations have been calling constant attention for quite a long time [1, 2].
We know that stochastic processes of t en posses self-similarity. The renor-
malization procedure used in quantum field theory is also based on the self-
similarity. So, it is natural to use for the regularization of field theories
the wavelet transform (WT), the decomposition with respect to the repre-
sentation of the affine group. In this paper two ways of regularization are
presented. First, the direct substitution of WT of t he fields into the action
functional leads to a field theory with scale-dependent coupling constants.
∗
Proc. GROUP 24: Physical and Mathematical Aspects of Symmetries, edited by J-P.
Gazeau, R. Kerner, J-P. Antoine, S. Metens, J-Y. Thibon; IOP Publishing, Bristol, 2003
1
Second, the WT, being substituted into the Pa risi-Wu stochastic quantiza-
tion scheme [3], provides a stochastic regularization with no extra vertexes
introduced into the theory.
2 Scalar field theory on affin e group
The Euclidea n field t heory is defined o n R
d
by the generating functional
W
E
[J] = N
Z
Dφ exp
−S[φ(x)] +
Z
d
d
xJ(x)φ(x)
, (1)
where S[φ] is the Euclidean action. In the simplest case of a scala r field with
the fourth power interaction
S[φ] =
Z
d
d
x
1
2
(∂
µ
φ)
2
+
m
2
2
φ
2
+
λ
4!
φ
4
. (2)
The φ
4
theory is of t en referred to as the Ginsburg-Landau model for its
ferromagnetic applications. The φ
3
theory is also a useful model.
The perturbation expansion generated by the functional (1) is usually
evaluated in k-space. The reformulation of the theory from the coordinate
(x) to momentum (k) representation is a particula r case of decomposition of
a function with respect to the representation of a Lie group G. G : x
′
=x+b
for the case of Fourier transform, but o t her groups may be used as well. For
a locally compact Lie gr oup G acting transitively on the Hilbert space H it
is possible to decompose state vectors with respect to the representations of
G [8, 9]
|φi = C
−1
ψ
Z
G
U(g)|ψidµ(g)hψ|U(g)|φi. (3)
The constant C
ψ
is determined by the norm of the action of U(g) on the
fiducial vector ψ ∈ H, C
ψ
= k ψk
−2
R
g∈G
|hψ|U(g)| ψi|
2
dµ(g); dµ(g) is the
left-invariant measure on G.
Using decomposition (3), it is possible to define a field theory on a non-
abelian Lie group. Let us consider the fourth power interaction model
Z
V (x
1
, x
2
, x
3
, x
4
)φ(x
1
)φ(x
2
)φ(x
3
)φ(x
4
)dx
1
dx
2
dx
3
dx
4
.
2
Using the notation U(g)| ψi ≡ |g, ψi, hφ|g, ψi ≡ φ(g), hg
1
, ψ|D|g
2
, ψi ≡
D(g
1
, g
2
), we can rewrite the generating functional (1) in the form
W
G
[J] =
Z
Dφ(g) exp
−
1
2
Z
G
φ(g
1
)D(g
1
, g
2
)φ(g
2
)dµ(g
1
)dµ(g
2
)
−
λ
4!
Z
G
V (g
1
, g
2
, g
3
, g
4
)φ(g
1
)φ(g
2
)φ(g
3
)φ(g
4
)dµ(g
1
)dµ(g
2
)dµ(g
3
)dµ(g
4
)
+
Z
G
J(g)φ(g)dµ(g)
, (4)
where V (g
1
, g
2
, g
3
, g
4
) is the result of the transform φ(g) :=
R
U(g)ψ(x)φ(x)dx
applied to V (x
1
, x
2
, x
3
, x
4
) in all arguments x
1
, x
2
, x
3
, x
4
.
Let us turn to the particular case of the affine group.
x
′
= ax + b, U(g)ψ(x) = a
−d/2
ψ((x − b)/a)), x, x
′
, b ∈ R
d
. (5)
The scalar field φ(x) in the action S[φ] can be written in the form of wavelet
decomposition
φ(x) = C
−1
ψ
R
1
a
d/2
ψ
x−b
a
φ
a
(b)
dadb
a
d+1
,
φ
a
(b) =
R
1
a
d/2
¯
ψ
x−b
a
φ(x)d
d
x.
(6)
In the scale-momentum (a, k) representation the matrix element of the free
field inverse pro pagator ha
1
, b
1
; ψ|D|a
2
, b
2
; ψi =
R
d
d
k
(2π)
d
e
ik(b
1
−b
2
)
D(a
1
, a
2
, k)
has the form
D(a
1
, a
2
, k) = a
d/2
1
ˆ
ψ(a
1
k)(k
2
+ m
2
)a
d/2
2
ˆ
ψ(a
2
k). (7)
The field theory (4) with the propagator D
−1
(a
1
, a
2
, k) gives standard Feyn-
man diagram technique, but with extra wavelet factor a
d/2
ˆ
ψ(ak) on each line
and the integrations over the measure dµ(a, k) =
d
d
k
(2π)
d
da
a
d+1
instead of
d
d
k
(2π)
d
.
Recalling the power law dependence of the coupling constants on the
cutoff momentum resulting from the Wilson expansion, we can define a scalar
field model on the affine group, with the coupling constant dep endent on
scale. Say, the fourth power interaction can be written as
V [φ] =
Z
λ(a)
4!
φ
4
a
(b)dµ(a, b), λ(a) ∼ a
ν
. (8)
3
The one-loop order contribution to the Green function G
2
in the theory
with interaction (8) can be evaluated [11] by integ ration over z = ak:
Z
a
ν
a
d
|
ˆ
ψ(ak)|
2
k
2
+ m
2
d
d
k
(2π)
d
da
a
d+1
=
Z
d
d
k
(2π)
d
C
(ν)
ψ
k
−ν
k
2
+ m
2
, C
(ν)
ψ
=
Z
|
ˆ
ψ(z)|
2
dz
z
1−ν
. (9 )
Therefore, there are no UV divergences for ν > d − 2. This is a kind of
asymptotically free theory which is hardly appropriate, say, to spin systems.
What is required to get a finite theory is an interaction vanishing outside a
given domain of scales. Such model is presented in the next section by means
of the stochastic quantization framework.
3 Stochastic quantization with wave lets
Let us remind the basic ideas of the stochastic quantization [3, 14, 12, 13]. Let
S[φ] be an action of the field φ(x). Instead of calculation of the physical Green
functions, it is possible to introduce the “extra-time” variable τ: φ(x) →
φ(x, τ) and evaluate the moments hφ(x
1
, τ
1
) . . . φ(x
m
, τ
m
)i
η
by averaging over
the random process φ(x, τ, ·) governed by the Langevin equation with the
Gaussian random force
˙
φ(x, τ)+
σ
2
2
δS
δφ(x, τ)
= η(x, τ), hη(x, τ)η(x
′
, τ
′
)i = σ
2
δ(x − x
′
)δ(τ − τ
′
). (1 0)
The physical G reen functions are then obtained by taking the limit τ
1
=
. . . = τ
m
= T → ∞.
The stochastic quantizatio n procedure has been considered as a perspec-
tive candidate fo r the regularization of gauge theories, for it respects local
gauge symmetries in a natural way. However a δ-correlated Gaussian random
noise in the Langevin equation still yields sharp singularities in the pertur-
bation theory. For this reason a number of modifications based on the noise
regularization η(x, τ) →
R
dyR
xy
(∂
2
)η(y, τ) have been proposed [4, 7, 6].
In this paper, following [15], we start with the random processes defined
directly in wavelet space. The use of the wavelet coefficients instead of the
original stochastic processes provides an extra analytical flexibility of the
method: t here exist more than one set of random functions W (a, b, ·) the
images of which have coinciding correlation functions. It is easy to check
that the random process generated by wavelet coefficients with the correlation
4
function h
c
W (a
1
, k
1
)
c
W (a
2
, k
2
)i = C
ψ
(2π)
d
δ
d
(k
1
+k
2
)a
d+1
1
δ(a
1
−a
2
)D
0
has the
same correlation function as the white noise has [17].
As an example, let us consider the Kardar-Parisi-Zhang equation [16]:
˙
Z − ν∆Z =
λ
2
(∇Z)
2
+ η. (11)
Substitution of wavelet transform
Z(x) = C
−1
ψ
Z
exp(ı(kx − k
0
t))a
d
2
ˆ
ψ(ak)
ˆ
Z(a, k)
d
d+1
k
(2π)
d+1
da
a
d+1
into (11), with the random f orce of the form
hbη(a
1
, k
1
)bη(a
2
, k
2
)i = C
ψ
(2π)
d+1
δ
d+1
(k
1
+ k
2
)a
d+1
1
δ(a
1
− a
2
)D(a
2
, k
2
), (12)
leads to the integral equation
(−ıω + νk
2
)
ˆ
Z(a, k) = η(a, k) −
λ
2
a
d
2
ˆ
ψ(ak)C
−2
ψ
R
(a
1
a
2
)
d
2
ˆ
ψ(a
1
k
1
)
ˆ
ψ(a
2
(k − k
1
))
k
1
(k − k
1
)
ˆ
Z(a
1
, k
1
)
ˆ
Z(a
2
, k − k
1
)
d
d+1
k
1
(2π)
d+1
da
1
a
1
d+1
da
2
a
2
d+1
.
From this the one-loop contribution the stochastic Green function follows:
G(k) = G
0
(k) − λ
2
G
2
0
(k)
R
d
d+1
k
1
(2π)
d+1
∆(k
1
)
k
1
(k − k
1
)|G
0
(k
1
)|
2
kk
1
G
0
(k − k
1
) + O(λ
4
),
(13)
where G
−1
0
(k) = −ıω + νk
2
. The difference from the standard approach [16]
is in the appearance of the effective force correlator
∆(k) ≡ C
−1
ψ
Z
da
a
|
ˆ
ψ(ak)|
2
D(a, k), (14)
which has the meaning of the effective force averaged over all scales.
Let us consider a single-band forcing [17] D(a, k) = δ(a − a
0
)D(k) and
the “Mexican hat” wavelet
ˆ
ψ(k) = (2π)
d/2
(−ık)
2
exp(−k
2
/2). In the leading
order in small parameter x = |k|/|k
1
| ≪ 1, the contribution to the stochastic
Green function is:
G(k) = G
0
(k) + λ
2
G
2
0
(k)
S
d
(2π)
d
a
3
0
k
2
ν
2
d − 2
8d
Z
∞
0
D(q)e
−a
2
0
q
2
q
d+1
dq + O(λ
4
).
5
4 Langevin equation for the φ
3
theory with
scale-depend ent noise
Let us apply the same scale-dependent noise (12) to the Langevin equation
for φ
3
theory. The standard procedure of the sto chastic quantization then
comes from the Langevin equation [6]
˙
φ(x, τ) +
−∆φ + m
2
φ +
λ
2!
φ
2
= η(x, τ). (15)
Applying the wavelet transform to this equation, we get:
(−ıω + k
2
+ m
2
)
ˆ
φ(a, k) = ˆη(a, k) −
λ
2
a
d
2
ˆ
ψ(ak)C
−2
ψ
R
(a
1
a
2
)
d
2
ˆ
ψ(a
1
k
1
)
ˆ
ψ(a
2
(k − k
1
))
ˆ
φ(a
1
, k
1
)
ˆ
φ(a
2
, k − k
1
)
d
d+1
k
1
(2π)
d+1
da
1
a
1
d+1
da
2
a
2
d+1
.
(16)
Iterating the integral equation (16), we yield the correction to the stochastic
Green function
G(k) = G
0
(k) + λ
2
G
2
0
(k)
Z
d
d+1
q
(2π)
d+1
∆(q)|G
0
(q)|
2
G
0
(k − q) + O(λ
4
). (17)
The analytical expressions for the stochastic Green functions, can be
obtained in the ω → 0 limit. As an example, we take the D(a, q) =
δ(a − a
0
)D(q) for φ
3
theory and the Mexican hat wavelet. The one loop con-
tribution to the stochastic Green function G(k) = G
0
(k) + G
2
0
λ
2
I
2
3
+ O(λ
4
)
is
lim
ω→0
I
2
3
=
Z
d
d
q
(2π)
d
∆(q)
1
2(q
2
+ m
2
)
·
1
q
2
+ (k − q)
2
+ 2m
2
.
The same procedure can be applied in higher loops. As it can be seen,
for constant o r compactly supported D(q) all integrals are finite due to the
exponential factor coming fro m wavelet ψ.
Acknowlegement
The author is grateful to Profs. H.H¨uffel and V.B.Priezzhev for useful dis-
cussions.
6
References
[1] E. Nelson. Quantum fluctuations. Princeton University Press, New Jer-
sey, 1985.
[2] J. Glim and A. Jaffe. Quantum physics. Springer-Verlag, NY, 1981.
[3] G. Parisi and Y.-S. Wu. Scientica Sinica, 24:483,1981.
[4] J. Breit, S. Gupta, and A. Zaks. Nuclear Physics, B233:61–68, 1984.
[5] Z. Bern, M. Halpern, L. Sadun, and C. Taubes. Physics Letters,
B165:151–156, 1985.
[6] R. Iengo and S. Pugnetti. Nuclear Physics , B300 :1 28–142, 1988.
[7] Z. Bern, M. Halpern, L. Sadun, and C. Taubes. Nuclear Physic s , B284:1–
91, 1987.
[8] A. Carey. Bull. Austr. Math. Soc., 15:1–12, 1976.
[9] M. Duflo and C. Moore. J. Func. Anal., 21:209–243, 1976.
[10] C. Chui. An i ntroduction to wavelets. Academic press, inc, 1991.
[11] M. Altaisky. In B.G. Sidharth and M.V. Altaisky, editors, Frontiers of
fundamental physics 4, pages 124–128, NY, 2001. Kluver Academic /
Plenum Publishers.
[12] P. Damgaard and H. H¨uffel. Physics Reports. 152:227-248, 1987.
[13] M. Namiki. Stochastic quan tiza tion . Springer, Heidelberg, 1992.
[14] J. Zinn-Justin. Nuclear Physics B, 275(FS17):135–159, 1986.
[15] M.V.Altaisky. European Journal of Physics B, 8(4 ) :6 13-617, 1999.
[16] M. Kardar, G. Parisi, and Y.-C. Zhang. Phys. Rev. Lett., 56(9):889–892,
1986.
[17] M. Altaisky. Doklady Physics, 48(9):478-480, 2003.
7