Prog. Theor. Exp. Phys.

2021, 023B02 (13 pages)

DOI: 10.1093/ptep/ptaa175

The renormalization group and the diffusion

equation
Masami Matsumoto∗ , Gota Tanaka∗ , and Asato Tsuchiya∗
Department of Physics, Shizuoka University 836 Ohya, Suruga-ku, Shizuoka 422-8529, Japan
Graduate School of Science and Technology, Shizuoka University 3-5-1 Johoku, Naka-ku, Hamamatsu 432-
8011, Japan

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

∗
E-mail: matsumoto.masami.16@shizuoka.ac.jp, tanaka.gota.14@shizuoka.ac.jp,
tsuchiya.asato@shizuoka.ac.jp
Received November 19, 2020; Revised November 27, 2020; Accepted November 27, 2020; Published December 4, 2020

...................................................................................................................
We study the relationship between the renormalization group and the diffusion equation. We
consider the exact renormalization group equation for a scalar field that includes an arbitrary
cutoff function and an arbitrary quadratic seed action. As a generalization of the result obtained
by Sonoda and Suzuki, we find that the correlation functions of diffused fields with respect to
the bare action agree with those of bare fields with respect to the effective action, where the
diffused field obeys a generalized diffusion equation determined by the cutoff function and the
seed action and agrees with the bare field at the initial time.
...................................................................................................................
Subject Index B32

1. Introduction
It is recognized that the diffusion equation is connected with the renormalization group, since diffu-
sion can be regarded as a continuum analog of block-spin transformation or coarse-graining. Indeed,
the solution to the diffusion equation in d dimensions,

∂τ ϕ(τ , x) = ∂i2 ϕ(τ , x), (1.1)

with an initial condition ϕ(0, x) = φ(x) is given by


ϕ(τ , x) = d d x K(x, x , τ )φ(x ), (1.2)

where K(x, x , τ ) is the heat kernel,

 
 1 (x − x )2
K(x, x , τ ) = exp − . (1.3)
(4πτ )d/2 4τ
Here, ϕ(τ , x) can be viewed as a coarse-grained field obtained by smearing φ in a ball centered at x
√
with radius τ . On the other hand, the gradient flow equation in gauge theories, which is regarded as
a generalized diffusion equation respecting gauge symmetry, has recently been used to renormalize
composite operators and so on [1–3].
Thus, it is natural to expect that there is relationship between (generalized) diffusion equations
or gradient flow equations and renormalization group equations. Indeed, the relationship has been
studied in Refs. [4–11]. In particular, it was shown in Ref. [11] that the correlation functions of the
diffused fields, Eq. (1.2), with respect to the bare action agree with those of bare fields with respect to
© The Author(s) 2020. Published by Oxford University Press on behalf of the Physical Society of Japan.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/),
which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
Funded by SCOAP3
PTEP 2021, 023B02 M. Matsumoto et al.

the effective action, where the effective action obeys the exact renormalization group (ERG) equation
with a cutoff function and a seed action. This implies that correlation functions of the composite
operators consisting of the diffused fields of Eq. (1.2) are finite, so that the diffused fields can be
used to renormalize the composite operators.
In this paper we study a generalization of the result in Ref. [11]. We consider the ERG equation
for a scalar field with an arbitrary cutoff function and an arbitrary quadratic seed action. We show
that the correlation functions of diffused fields with respect to the bare action agree with those of
bare fields with respect to the effective action obeying the above ERG equation, where the diffused
field obeys a generalized diffusion equation determined by the cutoff function and the seed action,
and agrees with the bare field at the initial τ .

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

We also perform the  expansion using the derivative expansion as a check of validity of the ERG
equation. We reproduce the well-known scaling dimensions of operators around the Wilson–Fisher
fixed point.
Throughout this paper, except for Sect. 5, we work in the momentum space, and introduce the
notation
 
dd p
≡ . (1.4)
p (2π)d

Note that the Fourier transforms of Eqs. (1.1)–(1.3) are, respectively,

(∂τ + p2 )ϕ(τ , p) = 0, (1.5)

ϕ(τ , p) = K(p, τ )φ(p), (1.6)

K(p, τ ) = e−τ p .
2
(1.7)

This paper is organized as follows. In Sect. 2 we give our statement on the relationship between
the ERG equation and a generalized diffusion equation. In Sect. 3, we prove our statement using
functional differential equations for generating functionals of the correlation functions. In Sect. 4 we
provide another proof of our statement by solving the ERG equation using a functional integration
kernel. In Sect. 5 we perform the  expansion using the derivative expansion. We reproduce the well-
known scaling dimensions of operators around the Wilson–Fisher fixed point. Section 6 is devoted
to the conclusion and a discussion. We clarify the reason for restricting ourselves to seed actions that
are quadratic in φ.

2. Relation between renormalization group and diffusion equation

The ERG equation is a functional differential equation that describes nonperturbatively how the
effective action S at the energy scale  changes when  is decreased to  − d. The one for a
scalar field in d dimensions is specified by a cutoff function Ċ (p2 ) and a seed action Ŝ , and takes
the form [12–15]
   
∂ −S [φ] 1 δ2 δ δ Ŝ [φ] −S [φ]
− e =− Ċ (p )
2
e−S [φ] − Ċ (p )
2
e .
∂ 2 p δφ(p)δφ(−p) p δφ(p) δφ(−p)
(2.1)

Ċ is quasi-local and incorporates UV regularization, while the seed action Ŝ is a functional of φ
that has derivative expansion. Here we consider an arbitrary cutoff function Ċ (p2 ) and an arbitrary

2/13
PTEP 2021, 023B02 M. Matsumoto et al.

quadratic seed action that satisfy the above conditions. The coarse-graining procedure is fixed by
the cutoff function and the seed action. We parametrize the seed action as

1
Ŝ [φ] = − Ċ −1 (p2 )χ (p2 )φ(p)φ(−p), (2.2)
2 p 
where χ is an arbitrary regular function. The initial condition for Eq. (2.1) is given at a bare cutoff
scale 0 , and S0 = S=0 is a bare action.
We define the vacuum expectation value with respect to the effective action S as

1
· · ·  = Dφ · · · e−S [φ] , (2.3)

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

Z
with

Z = Dφ e−S [φ] . (2.4)

We show the following relation on the correlation functions:

 n c  n c
 
φ(pa ) = ϕ(τ , pa ) + δn,2 (2π)d δ d (p1 + p2 ) r(, p21 ), (2.5)
a=1  a=1 0

where c stands for the connected part, and

1 1
τ= − . (2.6)
2 20
ϕ(τ , p) is a solution to the differential equation

1 δ Ŝ [ϕ] 1
∂τ ϕ(τ , p) = 2 Ċ (p2 ) = − 2 χ (p2 )ϕ(τ , p), (2.7)
2 δϕ(τ , −p) 2
or, equivalently,

∂ δ Ŝ [ϕ]
− ϕ(τ , p) = Ċ (p2 ) = −χ (p2 )ϕ(τ , p), (2.8)
∂ δϕ(τ , −p)
with the initial condition

ϕ(0, p) = φ(p). (2.9)

r(, p2 ) obeys the differential equation

∂r(, p2 )
 = Ċ (p2 ) + 2χ (p2 )r(, p2 ), (2.10)
∂
and satisfies the initial condition

r(0 , p2 ) = 0. (2.11)

Equations (2.7) or (2.8) can be viewed as a generalized diffusion equation. Equation (2.5) gives a
relation between the ERG equation and the generalized diffusion equation.
For later convenience, we solve Eqs. (2.10) and (2.8). The solution to Eq. (2.10) with the initial
condition in Eq. (2.11) is given by

r(, p2 ) = A−1 2 −2
,0 (p )B,0 (p ),
2
(2.12)

3/13
PTEP 2021, 023B02 M. Matsumoto et al.

with
  0 −1
d
A,0 (p ) = −
2
Ċ (p2 )B
2
 , (p )
2
, (2.13)
  0

 0 
d
B,0 (p ) = exp
2
χ (p ) .

2
(2.14)
 
The solution to Eq. (2.8) with the initial condition in Eq. (2.9) is given by
−1
ϕ(τ , p) = B, 0
(p2 )φ(p). (2.15)

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

We will give a proof of Eq. (2.5) in the next section. Before closing this section, we give an example
of Ċ (p2 ) and χ (p2 ):
2 ζ
Ċ (p2 ) = − K (p2 ) + 2 K (p2 )(1 − K (p2 )), (2.16)
 2 p
with

K (p2 ) = e−p
2 /2
, (2.17)
p2 η
χ (p2 ) = 2 − . (2.18)
2 2
Setting ζ = η in Eq. (2.16) yields an ERG equation studied in Refs. [11,16]. We rescale ϕ as
 η
  2
ϕ (τ , p) = ϕ(τ , p). (2.19)
0
Then, Eq. (2.5) reduces to
 n c   nη   c
 0 2
n
φ(pa ) = ϕ  (τ , pa )

a=1  a=1 0
2 (p2 )
 η 
K 1 − K (p2 ) 0 1 − K0 (p2 )
+ δn,2 (2π)d δ d (p1 + p2 ) − .
p2 K (p2 )  K0 (p2 )
(2.20)

We see from Eq. (2.7) that ϕ  satisfies Eq. (1.5) and is given by Eq. (1.6). Equation (2.20) is nothing
but the relation obtained in Ref. [11].

3. Proof of the relation in Eq. (2.5)

In this section we give a proof of the relation in Eq. (2.5). For this purpose, we define two functionals
of J (p):
 
i J (p)φ(−p) −S [φ]
U [J (·), ] = ln Dφ e p e , (3.1)

V [J (·), ] = V  [J (·), ] + R[J (·), ], (3.2)

with
 

V [J (·), ] = ln Dφ e
i p J (p)ϕ(τ ,−p) e−S0 [φ] , (3.3)

4/13
PTEP 2021, 023B02 M. Matsumoto et al.


i2
R[J (·), ] = r(, p2 )J (p)J (−p). (3.4)
2 p

U is the generating functional for the connected correlation functions of φ with respect to S , while
V  is the one for the connected correlation functions of ϕ with respect to S0 . U and V agree at
 = 0 because ϕ and r(, p2 ) satisfy Eqs. (2.9) and (2.11), respectively:

U [J (·), 0 ] = V [J (·), 0 ]. (3.5)

In the following, we will show that U and V satisfy the same functional differential equation,

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

∂
which is the first order in the  derivative. First, we calculate − ∂ U [J (·), ] as follows:
   
∂ 1 i p J (p)φ(−p) ∂ −S [φ]
− U [J (·), ] = U Dφ e − e
∂ e ∂
   
1 i J (p)φ(−p) 1 δ2
= U Dφ e p − Ċ (p2 ) e−S [φ]
e 2 p δφ(p)δφ(−p)
  
δ δ Ŝ [φ]
− Ċ (p2 ) e−S [φ]
p δφ(p) δφ(−p)
   
i2 i i J (p)φ(−p) δ Ŝ [φ] −S [φ]
=− Ċ (p2 )J (p)J (−p) + U Dφ e p Ċ (p2 )J (p) e
2 p e p δφ(p)
   
i2 i δ Ŝ [φ] 1 δ
=− Ċ (p )J (p)J (−p) + U
2
Ċ (p )J (p)
2
eU
2 p e p δφ(p) i δJ (·)
 
i2 δU
=− Ċ (p )J (p)J (−p) − χ (p2 )J (p)
2
. (3.6)
2 p p δJ (p)

Here, we have used Eq. (2.1) in the  second

 equality and performed partial path integrations in the
third equality. In the fourth, δφ(p) i δJ (·) is obtained by replacing φ(p ) in a functional δδφ(p)
δ Ŝ [φ] 1 δ Ŝ [φ]
with
1 δ
i δJ (−p ) . In the fifth, we used the explicit form of Ŝ , Eq. (2.2).
∂
Next, we calculate − ∂ V [J (·), ] as follows:

∂
− V [J (·), ]
∂
     
i2 ∂r(, p2 ) i i p J (p)ϕ(τ ,−p) ∂
=−  J (p)J (−p) + V  Dφ e J (p) − ϕ(τ , −p) e−S0 [φ]
2 p ∂ e p ∂

i2  
=− Ċ (p2 ) + 2χ (p2 )r(, p2 ) J (p)J (−p)
2 p
  
i i p J (p)ϕ(τ ,−p) δ Ŝ [ϕ] −S [φ]
+ V −R Dφ e J (p)Ċ (p2 ) e 0
e p δϕ(τ , p)

i2  
=− Ċ (p2 ) + 2χ (p2 )r(, p2 ) J (p)J (−p)
2 p
  
i δ Ŝ [φ] 1 δ
+ V −R Ċ (p )J (p) 2
eV −R
e p δφ(p) i δJ (·)

5/13
PTEP 2021, 023B02 M. Matsumoto et al.

 
i2   δ
=− Ċ (p ) + 2χ (p )r(, p ) J (p)J (−p) −
2 2 2
χ (p2 )J (p) (V − R)
2 p p δJ (p)
 
i2 δV
=− Ċ (p2 )J (p)J (−p) − χ (p2 )J (p) . (3.7)
2 p p δJ (p)
Here, we have used Eqs. (2.8) and (2.10) in the second equality, Eq. (2.2) in the fourth equality, and
Eq. (3.4) in the last equality.
We see from Eqs. (3.6) and (3.7) that U and V satisfy the same functional differential equation,
which is the first order in the  derivative. We therefore conclude with Eq. (3.5) that

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

U [J (·), ] = V [J (·), ]. (3.8)

Finally, we calculate the left-hand side of the relations in Eq. (2.5):

 n c 
 1 δn 
φ(pa ) = n U [J (·), ]
i δJ (−p1 ) · · · δJ (−pn ) J =0
a=1 

1 δn 
= n V [J (·), ]
i δJ (−p1 ) · · · δJ (−pn ) J =0

1 δ n 
= n V  [J (·), ] + δn,2 (2π)d δ d (p1 + p2 )r(, p21 )
i δJ (−p1 ) · · · δJ (−pn ) J =0
 n c

= ϕ(τ , pa ) + δn,2 (2π)d δ d (p1 + p2 )r(, p21 ). (3.9)
a=1 0

Thus, we have completed the proof of Eq. (2.5).

4. Functional integration kernel

In this section we show Eq. (2.5) by solving Eq. (2.1) in terms of a functional integration kernel,
which is a functional analog of the heat kernel in Eq. (1.3). We seek a functional integration kernel
K[φ, φ  , , 0 ] that is a solution to the ERG equation in Eq. (2.1) satisfying the initial condition
K[φ, φ  , 0 , 0 ] = [φ − φ  ], where is the delta functional. It is easy to show that K is given by
1
K[φ, φ  , , 0 ] = det[A,
2
B
0 ,0
]
  
1  
× exp − A,0 (p )(B,0 (p )φ(p) − φ (p))(B,0 (p )φ(−p) − φ (−p)) ,
2 2 2
2 p
(4.1)
where A,0 and B,0 are defined in Eqs. (2.13) and (2.14), respectively, and satisfy the initial
conditions

A0 ,0 = ∞, B0 ,0 = 1. (4.2)

Then, Eq. (2.1) is solved in terms of K as



e−S [φ] = Dφ  K[φ, φ  , , 0 ] e−S0 [φ ] . (4.3)

By substituting Eq. (4.3) into the definition of U , Eq. (3.1), performing φ integration, and using
Eqs. (2.12) and (2.15), we can show Eq. (3.8). Thus, we give another proof of the relation in Eq. (2.5).

6/13
PTEP 2021, 023B02 M. Matsumoto et al.

5.  expansion
In this section, as a check of the validity of the ERG equation in Eq. (2.1), we perform the  expansion
in d = 4 −  dimensions using the derivative expansion. We restrict ourselves to the case in which
Ċ (p2 ) and χ (p2 ) are given by Eqs. (2.16) and (2.18), respectively. We will see that the scaling
dimensions of operators around the Wilson–Fisher fixed point are reproduced for arbitrary η and ζ .
The  expansion using the derivative expansion was studied for the Polchinski equation [18], which
can be viewed as an ERG equation, in Ref. [17], and for another specific ERG equation in Ref. [9].
Here we use the procedure performed in Ref. [9].
We set 0 = ∞ so that Eq. (2.6) reduces to τ = 1/2 . We expand Ċ in terms of τ as1



Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

−τ p2 ζ  −τ p2

Ċ (p ) = −e
2
2τ − 2 1 − e
p
 
−τ p2 ζ
= −2τ e 1 − + O(τ ) . (5.1)
2
Then, we rewrite Eq. (2.1) in the coordinate space as
  
η δS [φ]
∂τ S [φ] = −∂ φ(x) − φ(x)
2
x 4τ δφ(x)
 

1 ζ δS [φ] δS [φ] δ 2 S [φ]
− 1− K(x, y, τ ) − , (5.2)
2 2 x,y δφ(x) δφ(y) δφ(x)δφ(y)
 
where x = d d x and K(x, y, τ ) is defined in Eq. (1.3).
In the following, we perform the derivative expansion keeping up to two derivatives. The effective
action is expanded as
 
1  2 
S [φ] = Vτ (φ(x)) + Wτ (φ(x)) ∂φ(x) . (5.3)
x 2
By using the formula

   
K(x, y, τ )f φ(x) g φ(y)
x,y
  
     2    
= f φ(x) g φ(x) − τ ∂φ(x) f  φ(x) g  φ(x) + O(τ 2 ) , (5.4)
x

we obtain, from Eq. (5.2),

       
η  1 ζ 2 1 ζ −d/2  d
∂τ S [φ] = − φVτ − 1− V + 1− (4πτ ) Vτ + Wτ
x 4τ 2 2 τ 2 2 2τ

   
 2 η 1 1 ζ    
+ ∂φ Vτ − φWτ + Wτ − 1− Vτ Wτ + 2Vτ Wτ − τ Vτ2
4τ 2 2 2
  
1 ζ
+ 1− (4πτ )−d/2 Wτ . (5.5)
4 2

2
Here we keep a factor e−τ p as a regularization of the delta function, which includes ambiguity. Indeed,
1
2 2
we have checked that the conclusion is unchanged if e−τ p is replaced by be−τ p with b a constant. A similar
ambiguity in the regularization of the delta function is also seen in Eq. (5.13).

7/13
PTEP 2021, 023B02 M. Matsumoto et al.

On the other hand, from Eq. (5.3) we obtain

  
1  2
∂τ S [φ] = ∂τ Vτ (φ) + ∂φ ∂τ Wτ (φ) . (5.6)
x 2
Comparing Eqs. (5.5) and (5.6) gives rise to
     
η 1 ζ 1 ζ d
∂τ Vτ (φ) = − φVτ − 1− Vτ2 + 1− (4πτ )−d/2 Vτ + Wτ , (5.7)
4τ 2 2 2 2 2τ
   
η 1 ζ    
∂τ Wτ (φ) = 2Vτ − φWτ + Wτ − 1 − Vτ Wτ + 2Vτ Wτ − τ Vτ2
2τ 2 2
 

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

1 ζ
+ 1− (4πτ )−d/2 Wτ . (5.8)
2 2
We normalize the kinetic term in Eq. (5.3) as Wτ (φ) = 1, so that

Wτ (φ) = 0, Wτ (φ) = 0. (5.9)

Then, Eqs. (5.7) and (5.8) reduce to

     
η  1 ζ 2 1 ζ −d/2  d
∂τ Vτ (φ) = − φVτ − 1− V + 1− (4πτ ) Vτ + , (5.10)
4τ 2 2 τ 2 2 2τ
 
η  ζ
∂τ Wτ (φ) = − + ζ Vτ + 1 − τ Vτ2 . (5.11)
2τ 2
We make a change of variable φ → ϕ such that Wτ +δτ = 1:

dϕ(x) = Wτ +δτ (φ) dφ(x). (5.12)

Then, the Jacobian for the change of variable is

  
  
 δφ  η ζ ζ τ
  = 1 − δτ 
aK(x, y, τ ) − + Vτ + 1 − Vτ δ (d) (x − y)
2
 δϕ  4τ 2 2 2
x,y
 
  
−d/2 η ζ  ζ τ 2
= exp −δτ a(4πτ ) − + Vτ + 1 − V , (5.13)
x 4τ 2 2 2 τ

where we have regularized the delta function δ (d) (x − y) by aK(x, y, τ ), with a a constant.
Further, we make quantities dimensionless as

x → τ 1/2 x, ϕ(x) → τ −(d−2)/4 ϕ(x), Vτ → τ −d/2 Vτ . (5.14)

Then, from Eqs. (5.7), (5.13), and (5.14), we obtain

 
    
d d −2  1 ζ 2 B 1  B ζ
τ ∂τ Vτ = Vτ − ϕVτ − 1+ V + 1+ a− ζ Vτ + a 1− V 2
2 4 2 2 τ 2 2 2 2 τ
     ϕ
B d 1 ζ
+ d − aη − ζ − 1− V dϕ  Vτ2 , (5.15)
4 2 2 2 τ 0

where B = (4π)−d/2 .
We expand Vτ in terms of ϕ as
v2 ϕ 2 v4 ϕ 4 v6 ϕ 6 v8 ϕ 8 v10 ϕ 10 v12 ϕ 12 v14 ϕ 14
Vτ = v0 + + + + + + + + ··· . (5.16)
2! 4! 6! 8! 10! 12! 14!

8/13
PTEP 2021, 023B02 M. Matsumoto et al.

By substituting Eq. (5.16) into Eq. (5.15), we obtain

   
  
ζ 2 ζ 3 B 1
τ ∂τ v2 = v2 − 1 + v − 1− v + 1+ a− ζ v4
2 2 2 2 2 2
 
ζ
+ aB 1 − v2 v4 , (5.17)
2
   
  
 ζ ζ 2 B 1
τ ∂τ v4 = v4 − 4 1 + v2 v4 − 6 1 − v v4 + 1+ a− ζ v6
2 2 2 2 2 2
 
ζ
+ aB 1 − (3v42 + v2 v6 ), (5.18)

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

2
 
  
ζ B 1
τ ∂τ v6 = (−1 + )v6 − 1 + (10v4 + 6v2 v6 ) +
2
1+ a− ζ v8
2 2 2
   
ζ ζ
+ aB 1 − (15v4 v6 + v2 v8 ) − 1 − (38v2 v42 + 9v22 v6 ), (5.19)
2 2
   
  
3 ζ B 1
τ ∂τ v8 = −2 +  v8 − 1 + (56v4 v6 + 8v2 v8 ) + 1+ a− ζ v10
2 2 2 2
 
ζ
+ aB 1 − (35v62 + 28v4 v8 + v2 v10 )
2
 
ζ
− 1− (12v22 v8 + 168v43 + 232v2 v4 v6 ), (5.20)
2
 
ζ
τ ∂τ v10 = (−3 + 2)v10 − 1 + (126v62 + 120v4 v8 + 10v2 v10 )
2

    
B 1 ζ
+ 1+ a− ζ v12 + aB 1 − (210v6 v8 + 45v4 v10 + v2 v12 )
2 2 2
 
ζ
− 1− (602v2 v62 + 520v2 v4 v8 + 15v22 v10 + 2556v42 v6 ), (5.21)
2
   
5 ζ
τ ∂τ v12 = −4 +  v12 − 1 + (792v6 v8 + 220v4 v10 + 12v2 v12 )
2 2

    
B 1 ζ
+ 1+ a− ζ v14 + aB 1 − (462v82 + 495v6 v10 + 66v4 v12 + v2 v14 )
2 2 2
 
ζ
− 1− (4104v2 v6 v8 + 980v2 v4 v10 + 18v22 v12 + 19580v4 v62 + 8536v42 v8 ), (5.22)
2

where  = 4 − d.
We look for the Wilson–Fisher fixed point by assuming that

v2∗ = O(), v4∗ = O(), v6∗ = O( 2 ), vn∗ = O( 3 ) for n ≥ 8. (5.23)

We set τ ∂τ vn∗ = 0 in Eqs. (5.17)–(5.19), and expand the right-hand side of Eq. (5.17) up to the first
order in , and those of Eqs. (5.18) and (5.19) up to the second order in . Then, we obtain equations
determining the fixed point as follows:

  
B 1
0 = v2∗ + 1+ a− ζ v4∗ , (5.24)
2 2

9/13
PTEP 2021, 023B02 M. Matsumoto et al.
   
  
 ∗ ζ ∗ ∗ ζ ∗ 2 B 1
0 = v4 − 4 1 + v v + 3aB 1 − (v4 ) + 1+ a− ζ v6∗ , (5.25)
2 2 2 4 2 2 2
 
∗ ζ
0 = −v6 − 10 1 + (v4∗ )2 . (5.26)
2

We solve these equations as

  
 1
v2∗ =− 1+ a− ζ + O( 2 ), (5.27)
6A 2

v4∗ = + O( 2 ), (5.28)

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

3B0 A
 
∗ 10 2 ζ
v6 = − 1+ + O( 3 ), (5.29)
(3B0 A) 2 2

where
 
1 2
A≡ a− ζ + 3aζ − 2a + 2. (5.30)
2

We take a such that A = 0. Note that B = (4π)−d/2 = B0 + O() = (4π)−2 + O().

By substituting vn = vn∗ + δvn into Eqs. (5.17)–(5.22) and expanding the right-hand side of
Eqs. (5.17)–(5.22) up to the first order in  and δvn , we obtain

    
ζ ∗ ζ ∗
τ ∂τ δv2 = 1 − 2 1 + v + aB 1 − v δv2
2 2 2 4

    
B 1 ζ ∗
+ 1+ a− ζ + 2a 1 − v δv4 , (5.31)
2 2 2 2
 
    
ζ ∗  ζ ∗ ζ ∗
τ ∂τ δv4 = −4 1 + v δv2 + −4 1+ v + 6aB 1 − v δv4
2 4 2 2 2 2 4

    
B 1 ζ ∗
+ 1+ a− ζ + 2a 1 − v δv6 , (5.32)
2 2 2 2
 
    
ζ ∗ ζ ∗ ζ ∗
τ ∂τ δv6 = −20 1 + v4 δv4 + −1 +  − 6 1 + v2 + 15aB 1 − v δv6
2 2 2 4

    
B 1 ζ ∗
+ 1+ a− ζ + 2a 1 − v δv8 , (5.33)
2 2 2 2
 
    
ζ ∗ 3 ζ ∗ ζ ∗
τ ∂τ δv8 = −56 1 + v δv6 + −2 +  − 8 1 + v + 28aB 1 − v δv8
2 4 2 2 2 2 4

    
B 1 ζ ∗
+ 1+ a− ζ + 2a 1 − v δv10 , (5.34)
2 2 2 2
 
    
ζ ∗ ζ ∗ ζ ∗
τ ∂τ δv10 = −120 1 + v δv8 + −3 + 2 − 10 1 + v + 45aB 1 − v δv10
2 4 2 2 2 4

    
B 1 ζ ∗
+ 1+ a− ζ + 2a 1 − v δv12 , (5.35)
2 2 2 2
 
    
ζ ∗ 5 ζ ∗ ζ ∗
τ ∂τ δv12 = −220 1 + v4 δv10 + −4 +  − 12 1 + v2 + 66aB 1 − v δv12
2 2 2 2 4

10/13
PTEP 2021, 023B02 M. Matsumoto et al.

    
B 1 ζ ∗
+ 1+ a− ζ + 2a 1 − v δv14 . (5.36)
2 2 2 2
We regard these as a linear transformation for δvn . Then, the eigenvalue equation reads
 95 
− λ5 + λ4 −5 + 5 − 15(2 + ζ )v2∗ + aB(2 − ζ )v4∗
2
 
+ λ3 −5 + 15 − 50(2 + ζ )v2∗ + B(−100 + 270a − 235aζ + 25ζ 2 − 50aζ 2 )v4∗
 199 
+λ 2
5 − 15(2 + ζ )v2∗ + B(−108 − 17a − 2 ∗
aζ + 27ζ − 54aζ )v4
2
2
 

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

+ λ 6 − 17 + 44(2 + ζ )v2∗ + B(112 − 288a + 256aζ − 28ζ 2 + 56aζ 2 )v4∗

− 3 + 12(2 + ζ )v2∗ + B(48 − 36a + 66aζ − 12ζ 2 + 24aζ 2 )v4∗ = 0. (5.37)

Here we have ignored Eq. (5.36) and dropped the term proportional to δv12 in Eq. (5.35). By
substituting Eqs. (5.27) and (5.28) into this equation, we obtain
     
−λ5 − λ4 5 − (45 + 110D) − λ3 5 − (20 + 110D)
6 3
     
+ λ2 5 − (93 + 110D) + λ 6 − (17 + 110D) + 3 = 0, (5.38)
6 3
where
a(2 − ζ ) a(2 − ζ )
D≡ = . (5.39)
A (a − 1/2)ζ 2 + 3aζ − 2a + 2
The solutions to Eq. (5.38) are
1 1 3 19
λ2 = 1 − , λ4 = − , λ6 = −1 − , λ8 = −2 − ,
6 2 2 6

λ10 = −3 + (77 + 110D). (5.40)
6
λ2 –λ8 are the well-known scaling dimensions of operators around the Wilson–Fisher fixed point.
We have checked that λ10 is also obtained correctly if Eq. (5.36) is included.

6. Conclusion and discussion

We have studied the relationship between the renormalization group and the diffusion equation.
We considered the ERG equation for a scalar field that includes an arbitrary cutoff function and
an arbitrary quadratic seed action. As a generalization of the result in Ref. [11], we found that the
correlation functions of diffused fields with respect to the bare action agree with those of bare fields
with respect to the effective action, where the diffused field obeys a generalized diffusion equation
determined by the cutoff function and the seed action and agrees with the bare field at the initial time.
This result is reasonable in that diffusion is associated with coarse-graining and the coarse-graining
procedure is fixed by the cutoff function and the seed action. We performed the  expansion using the
derivative expansion as a check of the validity of the ERG equation. We reproduced the well-known
scaling dimensions of operators around the Wilson–Fisher fixed point.
We comment on a case in which the seed action includes terms that are higher than the second
order in φ. In this case, U satisfies a functional differential equation given by the fourth equality in

11/13
PTEP 2021, 023B02 M. Matsumoto et al.

Eq. (3.6), while V satisfies a functional differential equation given by the third equality in Eq. (3.7).
The latter equation includes terms that mix δV /δJ with δR/δJ , and those terms do not exist in the
former equation. This implies that the relation in Eq. (2.5) does not hold for this case even if r is
modified. We need to generalize Eq. (2.5) in some way.
Finally, we note that Eq. (2.7) is a “gradient flow equation,”

δ S̃
∂τ ϕ(τ , p) = − , (6.1)
δϕ(τ , −p)
with


Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

2
S̃ = χ (p2 )ϕ(τ , p)ϕ(τ , −p). (6.2)
p 4

While the above gradient flow equation is at first sight conceptually different from the one studied
in the context of gauge theories, we hope that our findings in this paper will give some insights into
the construction of an ERG equation with manifest gauge invariance.

Acknowledgements
A.T. was supported in part by Grant-in-Aid for Scientific Research (No. 18K03614) from Japan Society for
the Promotion of Science.

Funding
Open Access funding: SCOAP3 .

References
[1] R. Narayanan and H. Neuberger, J. High Energy Phys. 03, 064 (2006) [arXiv:hep-th/0601210] [Search
INSPIRE].
[2] M. Lüscher, Commun. Math. Phys. 293, 899 (2010) [arXiv:0907.5491 [hep-lat]] [Search INSPIRE].
[3] M. Lüscher, J. High Energy Phys. 1008, 071 (2010); 1403, 092 (2014) [erratum] [arXiv:1006.4518
[hep-lat]] [Search INSPIRE].
[4] A. Kagimura, A. Tomiya and R. Yamamura, [arXiv:1508.04986 [hep-lat]] [Search INSPIRE].
[5] F. Capponi, A. Rago, L. Del Debbio, S. Ehret, and R. Pellegrini, PoS LATTICE 2015, 306 (2016)
[arXiv:1512.02851 [hep-lat]] [Search INSPIRE].
[6] S. Aoki, J. Balog, T. Onogi, and P. Weisz, Prog. Theor. Exp. Phys. 2016, 083B04 (2016)
[arXiv:1605.02413 [hep-th]] [Search INSPIRE].
[7] S. Aoki and S. Yokoyama, Prog. Theor. Exp. Phys. 2018, 031B01 (2018) [arXiv:1707.03982 [hep-th]]
[Search INSPIRE].
[8] H. Makino, O. Morikawa, and H. Suzuki, Prog. Theor. Exp. Phys. 2018, 053B02 (2018)
[arXiv:1802.07897 [hep-th]] [Search INSPIRE].
[9] Y. Abe and M. Fukuma, Prog. Theor. Exp. Phys. 2018, 083B02 (2018) [arXiv:1805.12094 [hep-th]]
[Search INSPIRE].
[10] A. Carosso, A. Hasenfratz, and E. T. Neil, PoS LATTICE 2019, 287 (2019) [arXiv:1912.01766
[hep-lat]] [Search INSPIRE].
[11] H. Sonoda and H. Suzuki, Prog. Theor. Exp. Phys. 2019, 033B05 (2019) [arXiv:1901.05169 [hep-th]]
[Search INSPIRE].
[12] J. I. Latorre and T. R. Morris, J. High Energy Phys. 0011, 004 (2000) [arXiv:hep-th/0008123] [Search
INSPIRE].
[13] S. Arnone, A. Gatti, and T. R. Morris, J. High Energy Phys. 0205, 059 (2002) [arXiv:hep-th/0201237]
[Search INSPIRE].
[14] S. Arnone, T. R. Morris, and O. J. Rosten, Eur. Phys. J. C 50, 467 (2007) [arXiv:hep-th/0507154]
[Search INSPIRE].

12/13
PTEP 2021, 023B02 M. Matsumoto et al.

[15] T. R. Morris, Nucl. Phys. B 573, 97 (2000) [arXiv:hep-th/9910058] [Search INSPIRE].

[16] H. Sonoda, Prog. Theor. Exp. Phys. 2015, 103B01 (2015) [arXiv:1503.08578 [hep-th]] [Search
INSPIRE].
[17] J. O’Dwyer and H. Osborn, Ann. Phys. 323, 1859 (2008) [arXiv:0708.2697 [hep-th]] [Search
INSPIRE].
[18] J. Polchinski, Nucl. Phys. B 231, 269 (1984).

Downloaded from https://academic.oup.com/ptep/article/2021/2/023B02/6020279 by guest on 04 February 2022

13/13