PHYSICAL REVIEW D 106, 056025 (2022)

Universal treatment of the reduction for one-loop integrals

in a projective space
Bo Feng ,1,2,3,4,* Jianyu Gong,1,† and Tingfei Li1,‡
1
Zhejiang Institute of Modern Physics, Zhejiang University,
Hangzhou 310027, People’s Republic of China
2
Beijing Computational Science Research Center, Beijing 100084, China
3
Center of Mathematical Science, Zhejiang University, Hangzhou 310027, People’s Republic of China
4
Peng Huanwu Center for Fundamental Theory, Hefei, Anhui 230026, China

(Received 19 May 2022; accepted 13 September 2022; published 30 September 2022)

Recently, a nice work about the understanding of one-loop integrals has been given by Arkani-Hamed
and Yuan [arXiv:1712.09991] using the language of the projective space associated to their Feynman
parametrization. We find this language is also very suitable to deal with the reduction problem of one-loop
integrals with general tensor structures as well as propagators having arbitrary higher powers. In this paper,
we show how to combine Feynman parametrization and embedding formalism to give a universal treatment
of reductions for general one-loop integrals, even including the degenerated cases, such as the vanishing
Gram determinant. Results from this method can be written in a compact and symmetric form.

DOI: 10.1103/PhysRevD.106.056025

I. INTRODUCTION these complicated cases within the same framework as the

ones without higher poles. Recently, combining the uni-
In recent years, we have witnessed enormous progress in
tarity cut and the derivation over mass, the reduction
computing and understanding the analytic structure of
coefficients for higher pole cases can be calculated [30],
scattering amplitude. At the one-loop level, it is well-known
except for the tadpole coefficients.
that a general one-loop integral in the D ¼ 4 − 2ϵ dimension
Recently, we have proposed an improved PV-reduction
can always be reduced to a linear combination of one-loop
method for one-loop integrals [31,32]. The reduction
scalar integrals having no more than five propagators
coefficients can be expressed with the cofactors of the
(master integrals) with reduction coefficients being rational
Gram matrix and have some symmetries. Thus, it is useful
functions of kinetic variables [1–17]. These master integrals
to understand these symmetries appearing in our inter-
at the one-loop level (i.e., tadpoles, bubbles, triangles,
mediate recursion relations and the final results. Notably,
boxes, and pentagons) are well known. Therefore, the main
the analytical structure of one-loop integrals is studied by
problem of one-loop integrals is to calculate the reduction
investigating Feynman parametrization in the projective
coefficients. There are a host of methods to deal with the
space for its compactness and the close relation to geometry
reduction at the integrand level and integral level, such as
[33]. Inspired by the geometric angle, we find it could be
Integration-By-Parts (IBP) [18,19], Passarino-Veltman(PV)
convenient to do reduction for one-loop integrals in
reduction [3], Ossola-Papadopoulos-Pittau(OPP) reduction
projective space. By our study in this paper, one can see
[20–22], and the unitarity cut method [14,17,23–29].
that the symmetry and simplicity of reduction coefficients
Although in practice, we will not meet many situations
are illustrated clearly with the denotations in [33].
where propagators have higher powers, a complete reduc-
Motivated by the work [33], we will develop an alter-
tion method should be able to deal with it. From this point
native method to determine the reduction coefficients of
of view, the IBP method is a complete method since it treats
one-loop integrals in the D ¼ 4 − 2ϵ dimension. The
general tensor integrals with higher poles are related to
*
fengbo@zju.edu.cn integrals En;k ½T in projective space by
†
jianyu_gong@zju.edu.cn Z
‡
tfli@zju.edu.cn hXdn−1 XiT½Xk 
En;k ½T ≡ nþk ; ð1:1Þ
Δ ðXQXÞ 2
Published by the American Physical Society under the terms of
the Creative Commons Attribution 4.0 International license.
Further distribution of this work must maintain attribution to where Δ is a simplex in n-dimensional space, which is
the author(s) and the published article’s title, journal citation, defined by HI X ¼ XI > 0; ∀I ¼ 1;2;…;n. The T is a general
and DOI. Funded by SCOAP3. tensor, which is contracted with k X’s. The homogeneous

2470-0010=2022=106(5)=056025(20) 056025-1 Published by the American Physical Society

BO FENG, JIANYU GONG, and TINGFEI LI PHYS. REV. D 106, 056025 (2022)

coordinate XI is denoted by a square bracket X ¼ Then we put the whole formula into the embedding space
½x1 ∶x2 ∶…∶xn , and two coordinates are equivalent with two higher dimensions by lifting
to each other up to a scaling, i.e., ½x1 ∶x2 ∶…∶xn  ∼
½kx1 ∶kx2 ∶…∶kxn  for any k ≠ 0. The measure in the yμ ⟼ Y M ¼ ðY þ ; Y − ; Y μ Þ ¼ ð1; y2 ; yμ Þ; ð2:4Þ
projective space is given by the differential form,
where we use the light-cone coordinates, i.e., the metric
ϵI1 ;I2 ;…;In ηþ− ¼ η−þ ¼ − 12, ημν ¼ diagðþ; − − −−Þ while all other
hXdn−1 Xi ¼ X dX ∧ dXI3 ∧ … ∧ dX In ;
ðn − 1Þ! I1 I2 entries vanish. For clarity, we will use the capital letters I, J
XQX ¼ QIJ XI X J : ð1:2Þ to denote the components of vectors in the embedding
space, greek letters μ, ν to denote the components of
As pointed out in [33], the integral En;k satisfies a nice Lorentzian vectors, and lowercase letters i, j for the
recursion relation, which will be recalled in Appendix A. external legs. We will also use capital letters Y, X
This property is the key to carry out reduction in this paper. simultaneously to denote vectors in the embedding space
This paper is structured as follows. In Sec. II, we discuss and projective space without ambiguity. Therefore, we can
how to write a general one-loop integral as a sum of simplify the denominator of (2.3) into the inner product of
integrals En;k in projective space. In Sec. III, we derive two vectors in the embedding space, and the quadratic
recursion relations for En;k ½V i ⊗ Lk−i  and dimension expression has been somehow linearized, for example,
recursion relations for nondegenerate Q, then apply them
to the reduction of one-loop integrals. In Sec. IV, we ðy − yi Þ2 ¼ −2Y · Y i : ð2:5Þ
discuss the reduction framework for degenerate Q. In
Sec. V, we show how to obtain the general expression of After defining
reduction coefficient from n-gon tensor integrals to n-gon 2 2 μ
scalar integrals, while general expressions of reduction i ¼ ð1; yi − mi ; yi Þ;
YM ∞ ¼ ð0; 1; 0; …; 0Þ;
YM
coefficients are given in Appendix C. More reduction RM ¼ ð0; 0; Rμ Þ; ð2:6Þ
results are listed in Appendix B.
it is easy to check that (2.3) becomes
II. ONE-LOOP INTEGRALS
Z
IN PROJECTIVE SPACE ðrÞ ð−2Y · Y ∞ Þv−D−r ð2Y · RÞr
I vn ;D ¼ ½dD Y Qn ; ð2:7Þ
j¼1 ð−2Y · Y j Þ
vj
In this paper, we will discuss the reduction of the most
general one-loop integrals, Pn
Z Z where v ¼ vj . The projective space invariant mea-
j¼1
dD k kμ1 kμ2 …kμr dD k kμ1 kμ2 …kμr R D R Dþ2 YδðY 2 Þ
Qn vj ¼ Qn ; sure is given by ½d Y ≡ iπdD=2 vol:GLð1Þ [GL(1) acts as an
2 2 vj
iπ D=2 j¼1 ½ðk − qj Þ − mj 
D=2
j¼1 Dj iπ
overall scaling of the Y coordinates] and the factor
ð2:1Þ ðY · Y ∞ Þv−D−r is necessary for the last expression to be
P genuinely an integral over the projective light cone. Using
where qj ¼ j−1 i¼1 pi . As pointed out in [31], we can the most general Feynman parametrization,
recover the tensor structure by multiplying each index P Z X 
with an auxiliaryPvector Ri;μi . Furthermore, we can combine 1 Γð i mi Þ 1
mn ¼ Q dx1 …dxn δ xi − 1
these Ri to R ¼ ri¼1 ai Ri to simplify the expression (2.1) Am 1 m2
1 A2 …An Γðmi Þ 0 i
to a Lorentzian invariant form, Q mi −1
x
Z × P i P ;
ðrÞ dD k ð2R · kÞr ð i x i A i Þ i mi
I vn ;D ≡ Q : ð2:2Þ
iπ D=2 nj¼1 ððk − qj Þ2 − m2j Þvj
and putting the Feynman parameters into the projective
To recover the result
Q of (2.1), one can expand R and extract space, (2.7) becomes
the coefficient of ri¼1 ai from the auxiliary formula (2.2). Z
With the above explanation, we will focus on the form (2.2) ðrÞ ΓðvÞ
I vn ;D ¼ Q hXdn−1 XiXvn −1
and transform it into projective space as suggested in [33]. Γðvi Þ Δ
First, to make our formulas elegant, we denote yμ ≡ kμ , Z
ð−2Y · Y ∞ Þv−D−r ð2Y · RÞr
yi ≡ qi ; thus, (2.2) becomes × ½dD Y ; ð2:8Þ
ð−2Y · WÞv
Z
ðrÞ dD y ð2R · yÞr
I vn ;D ¼ Q : ð2:3Þ where X ¼ ½x1 ∶x2 ∶…∶xn , which is a vector in aP
different
iπ D=2 nj¼1 ððy − yi Þ2 − m2j Þvj projective (Feynman parametrization) space, W ¼ j xj Y j ,

056025-2
UNIVERSAL TREATMENT OF THE REDUCTION FOR ONE-LOOP … PHYS. REV. D 106, 056025 (2022)

Q
and there is an additional factor Xvn −1 ≡ ni¼1 xvi i −1 ¼ where k can be an arbitrary number and i has the same parity
Qn vi −1 as r due to the power of R2 must be an integer. The expansion
i¼1 ðH i XÞ . Then the Feynman parametrization
R integral
has been written into the compact form Δ hXdn−1 Xi (see coefficients Ckr;i are determined by initial conditions
[33] for more details). We can further simplify (2.8) using the Ck0;i ¼ δi;0 ; Ck1;i ¼ 2kδi;1 and the recursion relation,
common trick,
Ckrþ1;i ¼ ði þ 1ÞCkr;iþ1 þ ð2k − r − i þ 1ÞCkr;i−1 : ð2:14Þ
Z
ðrÞ ΓðvÞ ð−ÞvþDþr ΓðDÞ
I vn ;D ¼Q hXdn−1 XiXvn −1
Γðvi Þ Δ ΓðvÞ From the recursion relation, one can solve Ckr;i for general
 r  v−D−r r, i as
∂ ∂
× RM YM
∞
∂W M
∂W M rþi
Z 2r r!Γðr−iþ1 Y
2 Þ
2
1 Ckr;i ¼ pffiffiffi ðk þ 1 − jÞ
× ½dD Y : ð2:9Þ π i!ðr − iÞ! j¼1
ð−2Y · WÞD
2r r!k!Γðr−iþ1
2 Þ r−i
Up to now, the last integral in (2.9) can be done easily. ¼ pffiffiffi ; ∈ N: ð2:15Þ
One way to solve it is to translate it back to the form (2.3), π i!ðr − iÞ!ðk − rþi
2 Þ!
2
which is1
Plugging (2.13) into (2.12), we get
Z
d yD
1
  2  ðrÞ Γðv − D2 − rÞ Xr
I vn ;D ¼ Q CD=2þr−v ðR2 Þ 2
r−i
iπ D=2 W − y − w2 D
w
W− þ Wþ − W− ð−Þvþr Γðvi Þ i¼0 r;i
ΓðD=2Þ Z
¼ ðW · WÞ−D=2 ; ð2:10Þ hXdn−1 XiXvn −1 ðR · WÞi ðLXÞv−D−r
ð−ÞD ΓðDÞ × Dþr−i : ð2:16Þ
Δ ðW · WÞv− 2
and we have Since the remaining integrals are in the projective space of
Z  r Feynman parameters, we should rewrite W · W; R · W as
ðrÞ ð−Þvþr ΓðD=2Þ vn −1 ∂
I vn ;D ¼ Q hXd XiX
n−1
RM X  X  X
Γðvi Þ Δ ∂W M n n
 v−D−r W·W ¼ xa Y a · xb Y b ¼ xa ðY a · Y b Þxb
∂
× YM ∞ ðW · WÞ−D=2 : ð2:11Þ a¼1 b¼1 a;b
∂W M
¼ XQX;
The action of ðY M ∂ Xn
∞ ∂W M Þ can be done easily after using the fact
R·W ¼ xb R · Y b ¼ V · X;
Y ∞ · Y ∞ ¼ 0, and we get b¼1
Z V ¼ ½R · q1 ∶R · q2 ∶…∶R · qn : ð2:17Þ
ðrÞ ð−Þvþr ΓðD=2Þ Γðv − D2 − rÞ
I vn ;D ¼ Q hXdn−1 XiXvn −1
Γðvi Þ Δ ΓðD=2Þ Then, we get
 r
∂
× ðLXÞv−D−r RM ðW · WÞ−ðv−D=2−rÞ ; Γðv − D2 − rÞ Xr
∂W M ðrÞ
I vn ;D ¼ Q CD=2þr−v ðR2 Þ 2
r−i

ð2:12Þ ð−Þvþr Γðvi Þ i¼0 r;i

Z
P hXdn−1 XiXvn −1 ðVXÞi ðLXÞv−D−r
where we have written ð−2Y ∞ · WÞ ¼ i xi ≡ L · X with × Dþr−i : ð2:18Þ
Δ ðXQXÞv− 2
L ¼ ½1∶1∶…∶1.2 The action of ðRM ∂W∂ M Þ is more compli-
cated. By power counting, we have the general expansion, For general one loop integrals (2.1), one can calculate Q as

X
r 1
ðR · ∂W Þr ðW 2 Þk ¼ Ckr;i ðW 2 Þk− 2 ðR2 Þ 2 ðR · WÞi ; ð2:13Þ
rþi r−i
Qij ¼ ðm2i þ m2j − q2ij Þ; qij ¼ qi − qj : ð2:19Þ
2
i¼0
Now the expression (2.18) is written as the integration over
1
We have used the fact W − ¼ 1 by the Feynman parametriza- the X-projective space. Using the result of [33]
tion. The integration result can be found in the formula (A.44) in Z
the book of Peskin and Schroeder [34]. hXdn−1 XiT½Xk 
2
The reason that we can ignore the action of ðRM ∂W∂ M Þ on En;k ½T ≡ nþk ; ð2:20Þ
ð−2Y ∞ · WÞ is because Y ∞ · R ¼ 0. Δ ðXQXÞ 2

056025-3
BO FENG, JIANYU GONG, and TINGFEI LI PHYS. REV. D 106, 056025 (2022)

where Δ is a simplex in n-dimensional space defined by III. REDUCTION FOR NONDEGENERATE Q

Hi X ¼ Xi > 0; ∀ i ¼ 1; 2; …; n and T is a kth tensor
Having transformed our problem (2.1) to the form (2.23),
contracted with k X’s, the general one loop integral in the
in this section, we will show how to use the tricks of
projective space (2.18) can be written as
integrals in projective space (see [33]) to generate recursion
relations of En;k ½V a ⊗ Lk−a . By applying these recursion
ðrÞ Γðv − D=2 − rÞ X
r
relations iteratively, one can reduce a general one-loop
I vn ;D ¼ Q CD=2þr−v ðR2 Þ 2
r−i

ð−Þvþr Γðvi Þ i¼0 r;i integral to the basis with coefficients written by elegant
expressions. In other words, the reduction can be done
v −1
× En;2v−n−D−rþi ½⊗j H j j ⊗ V i ; ð2:21Þ universally in the new projective space form. As we will
point out, the reduction coefficients will have an interesting
pattern other than the obvious permutation symmetry.3
where for simplicity, we write En;k ½V a ⊗ Lk−a  ≡ En;k ½V a 
Moreover, the reduction process can be carried out in
by neglecting the power of L.
Mathematica automatically.
For the later use, we need to do symmetrization for
v −1
the tensor ⊗j Hj j ⊗ V i. To do so, we can Puse the same
trick as used in (2.1) and (2.2), i.e., Z ≡ v−n A. Recursion relation
i¼1 zi H i and
S ¼ tV þ Z. Thus, (2.21) can be obtained from In this subsection, we derive the recursion relations of
En;k ½V a ⊗ Lk−a . We first consider the case Q is non-
Γðv − D=2 − rÞ X r degenerate. The key equation is the following [see Eq. (4.2)
Q CD=2þr−v ðR2 Þ 2 En;2v−n−D−rþi ½Sv−nþi 
r−i
in [33] ]:
ð−Þvþr
Γðvi Þ i¼0 r;i

ð2:22Þ hXdn−1 XiT½Xk  1 hðQ−1 TÞ½Xk−1 Xdn−2 Xi
¼ dX
ðXQXÞ
nþk
2 nþk−2 nþk−2
ðXQXÞ 2
Q
after taking the coefficients of ti zvn −1 ≡ ti ni¼1 zvi i −1 . k − 1 hXdn−1 XiðtrQ TÞ½Xk−2 
þ ; ð3:1Þ
Taking care of numerical factors, the final expression is nþk−2 ðXQXÞ 2
nþk−2

Xr
i!Γðv − D=2 − rÞ D=2þr−v 2 r−i where dX ¼ dXI ∂X∂ I , trQ T ¼ Q−1
I1 I2 T
I 1 I 2 …I k
. The proof of the
ðrÞ
I vn ;D ¼ C ðR Þ 2 formula can be found in Appendix A. By integrating (3.1),
i¼0
ð−1Þvþr ðv − n þ iÞ! r;i
 we get

× En;2v−n−D−rþi ½Sv−nþi  i v −1 : ð2:23Þ ðbÞ
tz n
En;k ½T ¼ αn;k En−1;k−1 ½ðHb Q−1 TÞ

A special case of (2.23) is that for vn ¼ 1n and r ¼ 0, we þ βn;k En;k−2 ½trQ T; ð3:2Þ
have
where summing over b is implicit and to simplify our
denotations, we have defined
I n;D ¼ ð−1Þn Γðn − D=2ÞEn;n−D ½Ln−D ; ð2:24Þ
1 k−1
αn;k ≡ ; βn;k ≡ : ð3:3Þ
where I n;D is the scalar integral of n propagators in D nþk−2 nþk−2
dimension. Let us give a little explanation for the first term on the right-
Before ending this section, we want to point out that in hand side of (3.2). When integrating a total derivative term,
our discussion, for example in (2.8), the power v − D − r we should choose a patch. For simplicity, we assume
could be positive or negative with the arbitrary choice of r. Xi ¼ 1. Then we get the contribution from the boundary
Since we have kept the dimension D arbitrary, we can take Xb ¼ 0 and Xb ¼ þ∞ for b ≠ i. By the dimensional
D properly (even a negative number) to make v − D − r a regularization, the term with Xb ¼ þ∞ gives zero. For
positive integer to make later discussion legitimate. At the
the term Xb ¼ 0, in hðQ−1 TÞ½Xk−1 Xdn−2 Xi, only when the
end of reduction, we can analytically continue D to the
first index of Q−1 takes the value b, the contribution is
proper dimension. We have checked with several examples
nonzero, which is equivalent to be written as ðHb Q−1 TÞ.
that such a continuation is allowed.
In this paper, we mainly discuss Feynman integrals in
3
D ¼ 4 − 2ϵ-dimension space. At the one-loop level, the In the work [31,32], one can see that the recursion relations,
master integrals are related to En;n−D ½Ln−D , n¼1;2;…;5. for example, the bubble tensors to bubble basis and the triangle
tensor to triangle basis, are similar, except for the boundary
So the main task of one-loop integral reduction is to reduce conditions. This similarity has explained the interesting pattern
general integral En;k ½Sa ⊗ Lk−a  to the basis En≤5;n−D ½Ln−D . we observe in this paper.

056025-4
UNIVERSAL TREATMENT OF THE REDUCTION FOR ONE-LOOP … PHYS. REV. D 106, 056025 (2022)

When we repeatedly do the recursion relation, there will

be a set of Xb setting to zero. Writing the index set as
bj ≡ fb1 ; b2 ; …; bj g with 1 ≤ bi < biþ1 ≤ n, we define
Xðbj Þ to be the new vector in lower-dimensional projective
space obtained by removing these components belonging
to the set bj from the original vector X. With this under-
standing, the meaning of
Z T½Xkðbj Þ hXðbj Þ dn−1−j X ðbj Þ i
ðb Þ
j
En−j;k ½T ≡ n−jþk ð3:4Þ
Δðbj Þ ðXðbj Þ Qðbj Þ Xðbj Þ Þ 2

is clear. Equation (3.4) represents the integral got by

removing propagators belonging to bj .
Now we consider the reduction of En;k ½T with T ¼
V i ⊗ Lk−i as given in (2.23). Since tensor T is contracted
with k X’s, we can symmetrize its last k − 1 indices as
below,

1 X FIG. 1. The reduction process of En¼5;k ½V i¼5 , where the black

V ⊗L
i k−i
→V⊗ σ½V i−1 ⊗ Lk−i ; ð3:5Þ points represent zero terms. The red, orange, and cyan arrows
ðk − 1Þ! σ∈S represent the first, second, and third terms, respectively in (3.6).
k−1

where the σ is the permutation acting on the tensor

scalar basis in a given D-dimensional space, we need to
V i−1 ⊗ Lk−i . By applying (3.1), one has
find the formula to shift the dimension of the scalar basis to
ðbÞ a fixed D.
En;k ½V iþ1  ¼ αn;k ½ðH b VÞEn−1;k−1 ½V i  þ iðVVÞEn;k−2 ½V i−1 
þ ðk − i − 1ÞðVLÞEn;k−2 ½V i ; ð3:6Þ B. Dimension recursion
As we have seen in (2.24), the integral En;k corresponds
where for simplicity, the L tensor part has been omitted. To
to the scalar n-gon diagram in (n − k)-dimensional space.
make our formula more compact, here we have defined
To find dimension recursion relations, we set V ¼ L in
ðABÞ ≡ AQ−1 B. Here, we want to remark on a subtle point. (3.6) and get
In (3.6), the Q is n × n matrix as defined in En;k and
ðbÞ
appears in front of En−1;k−1 . When we try to iteratively use ðbÞ
En;k ¼ αn;k ðHb LÞEn−1;k−1 þ βn;k ðLLÞEn;k−2 : ð3:7Þ
ðbÞ
(3.6) for En−1;k−1, that Q will become the ðn − 1Þ × ðn − 1Þ
matrix QðbÞ . Thus, for later convenience, we define To reduce En;k , where n − k ¼ D þ 2s, s ∈ Z; s ≠ 0, we
ðABÞðaj Þ ¼ Aðaj Þ ðQðaj Þ Þ−1 Bðaj Þ where the matrix ðQðaj Þ Þ−1 can iteratively use (3.7), which is established for the
scalar integrals already. Noticing that in the rhs of (3.7),
is the inverse of the matrix obtained by removing the rows
the first term has the same dimension as the lhs with one
and columns of the index set aj from the original matrix Q.
propagator being removed, while the second term has
As the main result of the whole paper, the recursion
two higher dimensions with the same number of propa-
relation (3.6) plays a crucial role in the reduction of one-
gators. Depending on the sign of s, we can take different
loop integrals. By comparing the power of V, one sees that
manipulations.
it has been reduced from the lhs to rhs. Furthermore, from
(i) s > 0: For this case, we need to reduce an integral in
(2.23), one sees that the power is given by v þ i − n, where
a higher dimension to D dimension, so we solve the
v contains the contribution of higher power of propagators second term in the rhs of (3.7) and get
and i contains the contribution of the tensor numerator;
thus, (3.6) provides the universal reduction of both cases.
ðjÞ
As shown in Fig. 1, after iteratively using (3.6), we get a En;kþ2 − αn;kþ2 ðHj LÞEn−1;kþ1
ðan−n0 Þ En;k ¼ : ð3:8Þ
linear combination of En0 ≤n;k 0 <k , i.e., the scalar integral βn;kþ2 ðLLÞ
I n0 ;D0 in dimension D0 ¼ n0 − k0 . So starting with a general
one-loop integral, one can always reduce it to the scalar It is obvious that such a rewriting (3.8) is legitimate
integrals in different dimensions with coefficients being when and only when ðLLÞ ≠ 0. For ðLLÞ ¼ 0, we
rational functions of external momenta. If we prefer the have

056025-5
BO FENG, JIANYU GONG, and TINGFEI LI PHYS. REV. D 106, 056025 (2022)

ðjÞ ð1Þ
En;k ¼ αn;k ðHj LÞEn−1;k−1 : ð3:9Þ I 2 ¼ −Γð1 − D=2ÞCD=2−1
1;1 E2;2−D ½V
¼ −Γð1 − D=2ÞðD − 2ÞE2;2−D ½V: ð3:14Þ
Both sides have the same dimension, but the rhs of
(3.9) has one less propagator. One well-known
First, we use (3.6) to reduce E2;2−D ½V and get
example of ðLLÞ ¼ 0 is that the bubble with null
external momentum is not a basis anymore, and it is ðiÞ
reduced to two tadpoles. Having established (3.8) E2;2−D ½V ¼ α2;2−D ðH i VÞE1;1−D þ β2;2−D ðVLÞE2;−D ;
and (3.9), we can reduce ðD þ 2sÞ-dimensional ð3:15Þ
integrals to D dimensional iteratively using either
(3.8) or (3.9) depending on if ðLLÞ is zero or not at ðiÞ
where the first term E1;1−D corresponds to
that step.4 D-dimensional scalar tadpoles generated by remov-
(ii) s < 0: For this case, we can use (3.7) directly. ing the ith propagator of bubble I 2 , while the second
ðjÞ term E2;−D corresponds to a (D þ 2)-dimensional
En;k ¼ αn;k ðHj LÞEn−1;k−1 þ βn;k ðLLÞEn;k−2 : ð3:10Þ
bubble. We need to lower the dimension of the
As pointed out before, the first term on the rhs second term further. Here, we assume ðLLÞ ≠ 0, by
corresponds to the scalar integrals in the same using (3.8), we get
dimension but with the j-th propagator has been
ðiÞ
removed and the second term corresponds to the E2;2−D − α2;2−D ðHi LÞE1;1−D
scalar integral in D0 ¼ n − k þ 2 dimension. For the E2;−D ¼ ; ð3:16Þ
β2;2−D ðLLÞ
boundary situation, i.e., n ¼ 1, the first term van-
ishes, and the second term gives a higher dimen- where the two terms in the numerator correspond to
sional scalar basis. Repeating it, we can reduce En;k the D-dimensional bubble and two tadpoles. Plug-
to scalar integrals in D-dimensional space. ging (3.16) and (3.15) to (3.14) and recognizing
them as master integral according to (2.24), we have
C. Examples
ð1Þ
X ðLLÞðH i VÞ−ðHi LÞðVLÞ 2ðVLÞ
To illustrate our method and avoid complicated compu- I2 ¼ − I 2;î þ I2:
tation in general cases, we first consider the reduction of i ðLLÞ ðLLÞ
tensor bubbles, ð3:17Þ
Z D
ðrÞ d l ð2R · lÞr
I2 ¼ v v ; ð3:11Þ So the reduction coefficients are
iπ D=2 D11 D22
ð1Þ ðLLÞðH1 VÞ − ðH1 LÞðVLÞ R · q2
where C2→2;1̂ ¼ − ¼ 2 ;
ðLLÞ q2
D1 ¼ ðl − q1 Þ2 − m21 ; D2 ¼ ðl − q2 Þ2 − m22 : ð3:12Þ ðLLÞðH2 VÞ − ðH2 LÞðVLÞ R·q
ð1Þ
C2→2;2̂ ¼ − ¼ − 2 2;
We set q1 ¼ 0 for simplicity, then ðLLÞ q2
   ð1Þ 2ðVLÞ ðm21 − m22 þ q22 ÞR · q2
1 1 0 C2→2 ¼ ¼ : ð3:18Þ
L¼ ; H1 ¼ ; H2 ¼ ; ðLLÞ q22
1 0 1
 1
m21 2 2
2 ðm1 þ m2 − q2 Þ
2 (ii) Tensor bubble with massless legs
Q¼ 1 2 2 2
: ð3:13Þ One can notice there is a pole of q22 in the reduction
2 ðm1 þ m2 − q2 Þ m22 ð1Þ
coefficients of I 2 , which comes from ðLLÞ,
Here, we give some results for different choices of ranks
and powers to illustrate our idea. 4q22
ðLLÞ ¼ − :
(i) Tensor bubble with primary propagator −2ðm21 þ m22 Þq22 þ ðm21 − m22 Þ2 þ q42
We first consider reducing a rank-1 bubble
ð1Þ ð1Þ ð3:19Þ
I 2 ≡ I f1;1g . Since there are no higher poles, we
just choose S ¼ V; r ¼ 1; v ¼ n ¼ 2 in (2.23), and For q22 ¼ 0, we have ðLLÞ ¼ 0, so we have
we have
ðbÞ
E2;2−D ½V ¼ α2;2−D ðHb VÞE1;1−D þ β2;2−D ðVLÞE2;−D :
4
Please remember that as emphasized under (3.6), at each step,
Q is different. ð3:20Þ

056025-6
UNIVERSAL TREATMENT OF THE REDUCTION FOR ONE-LOOP … PHYS. REV. D 106, 056025 (2022)

Here we need to reduce E2;−D , term E2;2−D corresponds to wanted D-dimensional

ðbÞ
bubble. We need to lift the dimension of the first
ðbÞ α2;−D ðH b LÞE1;1−D term further. By using (3.10), we get
E2;−D ¼ E1;−D−1 ¼ ; ð3:21Þ
β1;1−D ðLLÞðbÞ ðiÞ ðiÞ
E1;3−D ¼ β1;3−D ðLLÞE1;1−D ; ð3:27Þ
where we have used (3.9) and
ðijÞ
ðbÞ where we have used E0;2−D ¼ 0, and the rhs corre-
ðbÞ E1;1−D
E1;−D−1 ¼ : ð3:22Þ sponds to D-dimensional scalar tadpoles. Plugging
β1;1−D ðLLÞðbÞ (3.27) and (3.26) to (3.25) and recognizing them as
master integrals according to (2.24), we have
Using (2.24), we finally get

 1 
ð1Þ;q22 ¼0 1 − D ðVLÞðHb LÞ I v2 ;v¼3 ¼ − ðD − 2ÞðLLÞðiÞ ðH i ZÞ I 2;î
I2 ¼ − ðH b VÞ I 2;b̂ : 4 zv2 −1
D ðLLÞðbÞ 
1 
ð3:23Þ þ ðD − 3ÞðZLÞ I2: ð3:28Þ
2 zv2 −1
Explicitly, we have
There are two configurations,
ð1Þ;q22 ¼0 2ðDm21 − ðD − 2Þm22 ÞR · q2 
I2 ¼− I 2;1̂ 1 
Dðm21 − m22 Þ2 I f2;1g ¼ − ðD − 2ÞðLLÞðiÞ ðHi ZÞ I 2;î
4 z1
4m21 R · q2 
þ I : ð3:24Þ 1 
Dðm21 − m22 Þ2 2;2̂ þ ðD − 3ÞðZLÞ I 2
2 z1
(iii) Scalar bubble with higher poles 1
Then, we consider reducing scalar bubbles I fv1 ;v2 g ¼ − ðD − 2ÞðLLÞðiÞ ðHi H1 ÞI 2;î
4
with higher poles v ¼ v1 þ v2 ¼ 3. Due to there
1
being no tensor structure in the numerator, we just þ ðD − 3ÞðH1 LÞI 2 ;
set S ¼ Z ¼ z1 H1 þ z2 H2 ; r ¼ 0 in (2.23), 2 
1 
I v2 ;v¼3 ¼ −Γð3 − D=2ÞE2;4−D ½Zjzv2 −1 ; ð3:25Þ I f1;2g ¼ − ðD − 2ÞðLLÞðiÞ ðHi ZÞ I 2;î
4 z2

where v2 ¼ f1; 2g; f2; 1g. First, we use (3.6) to 1 
reduce E2;4−D ½Z and get þ ðD − 3ÞðZLÞ I 2
2 z2
ðiÞ
E2;4−D ½Z ¼ α2;4−D ðHi ZÞE1;3−D þβ2;4−D ðZLÞE2;2−D ; 1
¼ − ðD − 2ÞðLLÞðiÞ ðHi H2 ÞI 2;î
4
ð3:26Þ 1
þ ðD − 3ÞðH2 LÞI 2 : ð3:29Þ
ðiÞ
where the first term E1;3−D corresponds to (D − 2)- 2
dimensional scalar tadpoles generated by removing Explicitly, using (3.13), we find the reduction
the ith propagator of bubble I 2 , while the second coefficients are

D−2
Cf2;1g→2;1̂ ¼ ;
ððm1 − m2 Þ2 − q22 Þððm1 þ m2 Þ2 − q22 Þ
ðD − 2Þðm21 þ m22 − q22 Þ
Cf2;1g→2;2̂ ¼ − ;
2m21 ððm1 − m2 Þ2 − q22 Þððm1 þ m2 Þ2 − q22 Þ
ðD − 3Þðm21 − m22 − q22 Þ
Cf2;1g→2 ¼ ; ð3:30Þ
ððm1 − m2 Þ2 − q22 Þððm1 þ m2 Þ2 − q22 Þ
ðD − 2Þðm21 þ m22 − q22 Þ
Cf1;2g→2;1̂ ¼ − ;
2m22 ððm1 − m2 Þ2 − q22 Þððm1 þ m2 Þ2 − q22 Þ
D−2
Cf1;2g→2;2̂ ¼ ;
ððm1 − m2 Þ − q22 Þððm1 þ m2 Þ2 − q22 Þ
2

ðD − 3Þðm21 − m22 þ q22 Þ

Cf1;2g→2 ¼ − : ð3:31Þ
ððm1 − m2 Þ2 − q22 Þððm1 þ m2 Þ2 − q22 Þ

056025-7
BO FENG, JIANYU GONG, and TINGFEI LI PHYS. REV. D 106, 056025 (2022)

(iv) Tensor bubble with higher poles

ð1Þ
At last, we consider a combined case, I v2 ;v¼3 . Here, we need to set S ¼ tV þZ, Z ¼ z1 H1 þ z2 H2 . Setting v ¼ 3,
r ¼ 1, n ¼ 2 in (2.23), we have
ð1Þ
I v2 ;v¼3 ¼ −Γð3 − D=2ÞE2;4−D ½S2 jtzv2 −1 : ð3:32Þ

First, we use (3.6) iteratively to pull out all S’s in the numerator,
ðiÞ
E2;4−D ½S2  ¼ α2;4−D ½ðHi SÞE1;3−D ½S þ ðSSÞE2;2−D þ ð2 − DÞðSLÞE2;2−D ½S
ðiÞ
¼ α2;4−D ½β1;3−D ðHi SÞðSLÞðiÞ E1;1−D þ ðSSÞE2;2−D
ðiÞ
þ ð2 − DÞðSLÞα2;2−D ½ðHi SÞE1;1−D þ ð1 − DÞðSLÞE2;−D : ð3:33Þ

Among the four terms above, we only need to deal with the last term E2;−D since it corresponds to a (D þ 2)-
dimensional bubble, which has been discussed in (3.21). Finally, we have

ðD − 2ÞððLLÞðHi ZÞðZLÞ þ ðLLÞðZLÞðiÞ ðHi ZÞ − ðHi LÞðZLÞ2 Þ
ð1Þ
I v2 ;v¼3 ¼ − 
4ðLLÞ  v −1 I 2;î
tz 2
 
1 ðD − 2ÞðZLÞ2 
þ − ðZZÞ  I2: ð3:34Þ
2 ðLLÞ tzv2 −1

ð1Þ ð1Þ
One can get the reduction results for I f2;1g ; I f1;2g. For example,

ð1Þ ðD − 2Þ
I f2;1g ¼ − ½ðLLÞðH i VÞðH1 LÞ þ ðLLÞðHi H1 ÞðVLÞ þ ðLLÞðVLÞðiÞ ðHi H 1 Þ
4ðLLÞ

ðD − 2ÞðH1 LÞðVLÞ
þ ðLLÞðH 1 LÞðiÞ ðHi VÞ − 2ðHi LÞðVLÞðH 1 LÞI 2;î þ − ðH1 VÞ I 2 ; ð3:35Þ
ðLLÞ

where for simplicity, we will not present explicit expressions for these coefficients.

Note that the reduction coefficients in these examples are the characteristic equation Qξ ¼ 0 always has solutions,
rational functions. For some special masses and momenta and we denote the NQ as the null space spanned by linearly
configurations, denominators can become zero, which leads independent ξ’s.
to several kinds of divergences. Since only ðLLÞ ≡ LQ−1 L
appears in the denominators, all divergences come from the A. Q̃L ≠ 0
Q matrix and its all submatrices, which have det Q ¼ 0 or
When Q is degenerate, the recursion relation (3.1) in the
ðLLÞ ¼ 0. For example, the pole of q22 in (3.18) comes from
last sections breaks down for det Q ¼ 0. Our idea is to
LQ−1 L [see (3.19)]. The divergence of Cf2;1g→2;2̂ is given by consider the tensor structure with one L in the first place and
m21 ððm1 − m2 Þ2 − q22 Þððm1 þ m2 Þ2 − q22 Þ ¼ 0, which is cor- make other (k − 1) indices completely symmetric by sum-
responds to det Qð2Þ ¼ 0 or det Q ¼ 0. One can find the pole ming over all permutations between i V’s and ðk − 1 − iÞ
ðLLÞ ¼ 0 comes from the dimension shifting process (3.8), L’s. Using (A9), we have
which can be addressed by employing (3.7) to reduce En;k to
lower topology. To deal with the divergences coming from En;k ½ðQQ̃LÞ ⊗ V i ⊗ Lk−1−i 
det Q ¼ 0, we need to consider the reduction method for  X
1
degenerate Q elaborated in the next section. ¼ En;k ðQQ̃LÞ ⊗ σ½V i ⊗ Lk−1−i 
ðk − 1Þ! σ∈S
k−1

IV. REDUCTION FOR DEGENERATE Q ðbÞ i
¼ αn;k ðHb Q̃LÞEn−1;k−1 ½V  þ βn;k
i
ðV Q̃LÞEn;k−2 ½V i−1 
k−1
In this section, we generalize our reduction method to
k−1−i
degenerate Q. The basic idea is to generalize the recursion þ ðLQ̃LÞEn;k−2 ½V i  : ð4:1Þ
relation (3.1) to the formula (A9). When Q is degenerate, k−1

056025-8
UNIVERSAL TREATMENT OF THE REDUCTION FOR ONE-LOOP … PHYS. REV. D 106, 056025 (2022)

For a degenerate Q, we can always find a matrix Q̃ ¼ ðjÞ

0 ¼ An;k ½ðH j Q̃VÞEn−1;k−1 ½V i  þ iðV Q̃VÞEn;k−2 ½V i−1 ;
½ξ1 ; ξ2 ; …; ξn , ξi ∈ NQ so that QQ̃ ¼ 0. Then the lhs of
(4.1) vanishes. If Q L ≠ 0, where Q is the adjugate matrix ð4:7Þ
of Q, we can take Q̃ ¼ Q . With the denotation,
and we have
_ ¼ ðAQ̃BÞ;
ðABÞ ð4:2Þ −ðHb_ VÞ ðbÞ
En;k ½V i  ¼ En−1;kþ1 ½V iþ1 : ð4:8Þ
_
ði þ 1ÞðVVÞ
(4.1) becomes
Again, although R appears in the denominator through
αn;k ðHb_LÞEn−1;kþ1 ½V i  þ iβn;k ðVLÞE
_
ðbÞ
n;k ½V 
i−1
_ in the intermediate steps, it will be canceled in the
ðVVÞ
_
þ ðk þ 1 − iÞβn;k ðLLÞE _ is another new
n;k ½V  ¼ 0: ð4:3Þ
i
final reduction coefficients. Similarly, ðVVÞ
type of spurious singularity in our method.
_
Depending on the value of ðLLÞ, we have following the
two cases: C. Dimension recursion
_ ≠ 0, (4.3) can be rewritten as
(1) When ðLLÞ Having reduced to scalar integrals with different dimen-
sions, we want to shift the dimension to a given D.
1
½ðHb_ LÞEn−1;kþ1 ½V i 
jQj¼0 ðbÞ
En;k ½V i  ¼ Depending on various situations, we have
ði − k − 1ÞðLLÞ_ (i) For Q̃L ≠ 0, we can always choose QQ̃ ¼ 0,
_ _ ≠ 0. Then using (4.4) for the case i ¼ 0, the
ðLLÞ
þ iðVLÞE n;k ½V
i−1
; ð4:4Þ
second term vanishes, and we have
where the first term in rhs corresponds to the lower
topologies and the second term has tensor rank −ðHb_ LÞ ðbÞ
En;k ¼ En−1;kþ1 : ð4:9Þ
reduced by one. _
ð1 þ kÞðLLÞ
(2) When ðLLÞ_ ¼ 0, (4.3) can be rewritten as
(ii) For Q̃L ¼ 0, we use (4.8) with i ¼ 1. Then, we shift
V → L þ ϵK with a reference K such that K= ⊥NQ
jQj¼0 −ðHb_ LÞ ðbÞ
En;k ½V i  ¼ En−1;kþ1 ½V iþ1 ; ð4:5Þ and get
_
ði þ 1ÞðVLÞ
−ðH b_KÞ ðbÞ
where, although the tensor rank increased by one on En;k þ ϵEn;k ½K ¼ En−1;kþ1 ½ðL þ ϵKÞ2 :
_
2ϵðKKÞ
the rhs, it belongs to the lower topologies. One
particular thing of (4.5) is that since V depends on ð4:10Þ
the auxiliary R, we will always have ðVLÞ _ ≠ 0.
Another tricky point is that although R appears in the Comparing both sides, especially the ϵ term, we have
denominator in (4.5) in the intermediate steps, it will
be canceled in the final reduction coefficients. Thus, −ðHb_KÞ ðbÞ
they are a new type of spurious singularities, which En;k ¼ En−1;kþ1 ½K: ð4:11Þ
_
ðKKÞ
is tightly related to our method.
We lower the topology in the rhs, so we can reduce it
B. Q̃L = 0 further by using the equation recursively. The depend-
Since Q̃L ¼ 0, every term in (4.3) vanishes. Now we can ence of the choice of K in the intermediate steps will
put V in the first place of the tensor structure and use (A9) vanish in the final reduction coefficients, as shown in
to reach the examples in the next section.

En;k ½ðQQ̃VÞV i  D. Examples

ðjÞ In this section, we illustrate our method for degenerate
¼ An;k ½ðH j Q̃VÞEn−1;k−1 ½V i  þ iðV Q̃VÞEn;k−2 ½V i−1 
Q. To avoid unnecessary and complicated calculations and
þ ðk − i − 1ÞðV Q̃LÞEn;k−2 ½V i : ð4:6Þ compare with the reduction procedures for nondegenerate
Q discussed in the Sec. III C, we focus on bubbles with
Using QQ̃ ¼ Q̃L ¼ 0, (4.6) becomes some special masses and momenta configurations.

056025-9
BO FENG, JIANYU GONG, and TINGFEI LI PHYS. REV. D 106, 056025 (2022)

(i) Q̃L ¼ 0: Massless scalar bubble with equal internal Here, we consider the reduction of the tensor
masses ð1Þ;m1 ¼m2 ;q22 ¼0
bubble I 2 . Recalling the first equation
To check the validity of our method, we first of (3.14),
consider a scalar bubble with m1 ¼ m2 ¼ m and
q22 ¼ 0, defined as ð1Þ
I 2 ¼ −Γð1 − D=2ÞðD − 2ÞE2;2−D ½V; ð4:18Þ
Z
m ¼m2 ;q22 ¼0 d l
D
1
I2 1 ¼ D=2 ðl2 − m2 Þ½ðl − q Þ2 − m2 
; we just need to reduce E2;2−D ½V. Due to Q̃L ¼ 0,
iπ 2
we use (4.8) to get
ð4:12Þ
−1
ðHb_VÞE1;3−D ½V 2 ;
ðbÞ
which can be reduced to two tadpoles.5 E2;2−D ½V ¼ ð4:19Þ
_
2ðVVÞ
We find Q defined in (3.13) degenerates to a
corank-1 matrix,
where the rhs corresponds to integrals of tadpoles
 2  2 with the nondegenerate Q matrix QðbÞ ¼ m2 . We can
m m2 
m −m2
Q¼ ; Q̃ ¼ Q ¼ : use (3.6) to reduce it iteratively,
m2 m2 −m2 m2
ð4:13Þ ðbÞ
E1;3−D ½V 2  ¼ α1;3−D ½ðVVÞðbÞ E1;1−D
ðbÞ

ðbÞ
þð1−DÞðVLÞðbÞ E1;1−D ½V
Since QQ̃ ¼ 0 and Q L ¼ 0, using (2.24), we
ðbÞ
have ¼ α1;3−D ½ðVVÞðbÞ E1;1−D
ðbÞ
I 2 ¼ Γð2 − D=2ÞE2;2−D : ð4:14Þ þð1−DÞðVLÞðbÞ β1;1−D ðVLÞðbÞ E1;−1−D :
ð4:20Þ
One can see the rhs is not irreducible anymore, by
employing (4.11), There are two terms in the last line, we need to
ðbÞ
reduce the second one E1;−1−D since it corresponds
−ðH b_KÞ ðbÞ to the (D þ 2)-dimensional tadpole. Using (3.8) to
E2;2−D ¼ E1;3−D ½K; ð4:15Þ
_
ðKKÞ lower the dimension to D, we get
ðbÞ
where the reference vector K satisfies Q K ≠ 0 and ðbÞ E1;1−D
the rhs corresponds to integrals of tadpole topology E1;−1−D ¼ : ð4:21Þ
β1;1−D ðLLÞðbÞ
acquired by removing one propagator from I 2 . We
then use (3.6) to pull out the K, Plugging (4.20), (4.21), (4.19) into (4.18), we have
ðbÞ ðbÞ
E1;3−D ½K ¼ β1;3−D ðKLÞðbÞ E1;1−D : ð4:16Þ ð1Þ;m1 ¼m2 ;q22 ¼0
I2
Note that the Q matrix in (4.16) is just a number ðHb_VÞðð1 − DÞðVLÞ2ðbÞ þ ðLLÞðbÞ ðVVÞðbÞ Þ
QðbÞ ¼ m2 , so its inverse in ðKLÞðbÞ is just ¼ I 2;b̂ :
_
2ðVVÞðLLÞ
1
Q−1
ðbÞ ¼ m2 . Combining (4.15), (4.16), and (4.14),
ðbÞ

we finally find ð4:22Þ

_ Using
m ¼m ;q2 ¼0 2 − D ðHb KÞðKLÞðbÞ D−2
I2 1 2 2 ¼ I 2;b̂ ¼ I :
2 _
ðKKÞ 2m2 2;1̂ _ ¼ m2 ðR · q Þ2 ;
ðVVÞ 2
ð4:17Þ
ðHb_VÞ ¼ f−m2 R · q2 ;m2 R · q2 g;
One can check that the result above does not depend ðR · q2 Þ2 R · q2
on the choice of K ¼ ða; bÞ as long as a ≠ b. ðVVÞðbÞ ¼ ;0 ; ðVLÞðbÞ ¼ ;0 ;
m2 m2
(ii) Q̃L ¼ 0: Massless tensor bubble with equal internal 1
masses ðLLÞðbÞ ¼ 2 ; ð4:23Þ
m
5
One can check this by using FIRE or direct calculation. we find the reduction relation,

056025-10
UNIVERSAL TREATMENT OF THE REDUCTION FOR ONE-LOOP … PHYS. REV. D 106, 056025 (2022)

ð1Þ;m1 ¼m2 ;q22 ¼0 ðD − 2ÞR · q2 ð1Þ

I 2 ¼ −Γð1 − D=2ÞðD − 2ÞE2;2−D ½V: ð4:31Þ
I2 ¼ I 2;1̂ ; ð4:24Þ
2m2
Choosing Q̃ ¼ Q , due to ðLLÞ_ ¼ LQ L ≠ 0, we
where I 2;1̂ is just the tadpole I 1 ½m with mass m.
(iii) Q̃L ≠ 0: Scalar bubble can use (4.4) to reduce the rhs to
ð1Þ
We then consider the scalar bubble I 2 with
1
degenerate Q but m1 ≠ m2 . Here, we will show that E2;2−D ½V ¼
it can be reduced to two tadpoles using our method. _
ðD − 2ÞðLLÞ
The equation det Q ¼ 0 gives two solutions,
× ½ðH _LÞE _
ðbÞ
b 1;3−D ½V þ ðVLÞE2;2−D ;

q22 ¼ ðm1  m2 Þ2 ; ð4:25Þ ð4:32Þ

which are just the poles in the reduction coefficients where the second term has been discussed in the
of bubbles [see (3.31)].6 Here, we choose last example. We just need to reduce the first term
  ðbÞ
E1;3−D ½V, which corresponds to a tadpole. Using

m22 m1 m2
Q̃ ¼ Q ¼ : ð4:26Þ (3.6), we find
m1 m2 m21
ðbÞ ðbÞ
One can check that Q̃L ≠ 0 for m1 ≠ m2. Due to E1;3−D ½V ¼ β1;3−D ðVLÞðbÞ E1;1−D ; ð4:33Þ
_ ≠ 0, we can use (4.4) to reduce E
ðLLÞ 2;2−D in the
expansion, where the rhs is a D-dimensional scalar tadpole.
Plugging (4.33), (4.32) into (4.31), we get
I 2 ¼ Γð2 − D=2ÞE2;2−D ; ð4:27Þ

and we get ð1Þ;q22 ¼ðm1 m2 Þ2 ðHb_LÞðVLÞðbÞ

I2 ¼ I 2;b̂
_
ðLLÞ
−ðHb_ LÞ ðbÞ
E2;2−D ¼ E1;3−D 2 ðVLÞ_ 2
_
ð3 − DÞðLLÞ q ¼ðm m Þ2
þ I 22 1 2 : ð4:34Þ
_
D − 2 ðLLÞ
ðH _ LÞβ
b 1;3−D ðbÞ
¼ ðLLÞðbÞ E1;1−D : ð4:28Þ
_
ðD − 3ÞðLLÞ Then, we refer to the result (4.29) and find
Plugging (4.28) into (4.27), we have
ð1Þ;q22 ¼ðm1 m2 Þ2 ðHb_LÞðVLÞðbÞ
I2 ¼ I 2;b̂
m ¼m2 ;q22 ¼0 ðD − 2ÞðHb_ LÞðLLÞðbÞ _
ðLLÞ
I2 1 ¼ I 2;b̂ : ð4:29Þ
_
2ðD − 3ÞðLLÞ ðH b_LÞðLLÞðbÞ ðVLÞ
_
þ I 2;b̂ : ð4:35Þ
_ 2
ðD − 3ÞðLLÞ
One can find the explicit expression easily. For
q22 ¼ ðm1 þ m2 Þ2 , the explicit expression is
For q22 ¼ ðm1 þ m2 Þ2, the explicit expression is
m ¼m2 ;q22 ¼0 D−2
I2 1 ¼ I
2ðD − 3Þm2 ðm1 þ m2 Þ 2;1̂ ð1Þ;q22 ¼ðm1 þm2 Þ2 ððD − 2Þm1 þ ðD − 3Þm2 ÞR · q2
I2 ¼ I 2;1̂
D−2 ðD − 3Þm2 ðm1 þ m2 Þ2
þ I : ð4:30Þ
2ðD − 3Þm1 ðm1 þ m2 Þ 2;2̂ R · q2
þ I : ð4:36Þ
ðD − 3Þðm1 þ m2 Þ2 2;2̂
(iv) Q̃L ≠ 0: Tensor bubble
Here, we discuss the reduction of the rank-1
bubble with q22 ¼ ðm1  m2 Þ2 . First, we expand it V. GENERAL EXPRESSION OF Cn→n
ðrÞ
using (2.23)
It seems hard to solve (3.6), but if we only care about the
6 reduction coefficients to the same topology, we can ignore
The condition is the threshold (or the Landau poles) of the
integrals. It appears because the singularity of Landau poles the first term for it contributes only to lower topologies.
becomes higher with higher power of propagators. Thus, it will Then by iteratively using (3.6), keeping only the second
appear in both master integrals and the reduction coefficients. and the third term, one finds that

056025-11
BO FENG, JIANYU GONG, and TINGFEI LI PHYS. REV. D 106, 056025 (2022)

X Qj
i! i−j ðk − i − l þ 1Þ En;n−D−ðr−iÞ ½V i  to En;n−D . After using (3.8) to repeatedly
En;k ½V  ¼
i
ðVVÞ ðVLÞ Ql¼1
2 j

j
j!ði − jÞ!! iþj
2
ðk þ n − 2lÞ
lift k, we have
l¼1
× En;k−i−j þ Lower topology: ð5:1Þ E
En;k−2s ¼ Q n;k
ðLLÞs sl¼1 ðβn;k−2sþ2l Þ
Although the first term has the same topology, with Xs
−αn;k−2sþ2l ðHj LÞ ðjÞ
different choices of i, j, the dimension is different, thus þ Ql E
we need to reduce it further. l¼1
β
p¼1 n;k−2sþ2p ðLLÞ l n−1;k−1−2sþ2l

To simplify our denotation, we define Kþ

n;k;s
X
s Kþ
n;k;s;j ðbÞ
¼ En;k þ ðHb LÞEn−1;k−1−2ðs−jÞ ;
ðLLÞs ðLLÞj
i! ðk − iÞ!ðk þ n − i − j − 2Þ!! j¼1
E n;k;i;j ¼ : ð5:2Þ
j!ði − jÞ!! ðk − i − jÞ!ðk þ n − 2Þ!! ð5:4Þ

Thus, (5.1) becomes where we have defined

X 1
En;k ½V  ¼ i i−j
E n;k;i;j ðVVÞ 2 ðVLÞj En;k−i−j Kþ
n;k;s ¼
Qs ;
l¼1 ðβn;k−2sþ2l Þ
j
−αn;k−2sþ2j
þ Lower topology: ð5:3Þ Kþ
n;k;s;j ¼ Qj : ð5:5Þ
p¼1 βn;k−2sþ2p

Here, we give the computation for tensor reduction with all Since the second term in (5.4) belongs to lower topologies,
propagators power one.7 Thus, we need to reduce all which can be ignored, finally, we have

X i−j
En;n−D−ðr−iÞ ½V i  ¼ E n;n−D−ðr−iÞ;i;j ðVVÞ 2 ðVLÞj En;n−D−ðrþjÞ þ Lower
j
þ
X i−j
Kn;n−D;rþj
¼ E n;n−D−ðr−iÞ;i;j ðVVÞ 2 ðVLÞj rþj En;n−D þ Lower
2

j ðLLÞ 2

þ
X i−j
Kn;n−D;rþj ð−Þn
¼ E n;n−D−ðr−iÞ;i;j ðVVÞ 2 ðVLÞj 2
I n;D þ Lower: ð5:6Þ
ðLLÞ 2 Γðn − D=2Þ
rþj
j

So

ðrÞ
X
r
Γðn − D=2 − rÞ
I n;D ¼ CD=2þr−n ðR2 Þ 2 En;n−D−ðr−iÞ ½V i 
r−i

i¼0
ð−1Þ nþr r;i

þ
X
r
Γðn − D=2 − rÞ X i−j
Kn;n−D;rþj ð−Þn
¼ CD=2þr−n ðR2 Þ 2 E n;n−D−ðr−iÞ;i;j ðVVÞ 2 ðVLÞj I n;D þ Lower
r−i 2

i¼0
ð−1Þnþr r;i
j ðLLÞ 2 Γðn − D=2Þ
rþj

ð−Þr Γðn − D=2 − rÞ Xr X i−j

2 r−i ðVVÞ ðVLÞ

i j
2
¼ CD=2þr−n E n;n−D−ðr−iÞ;i;j Kþ rþj ðR Þ 2 I n;D þ Lower: ð5:7Þ
Γðn − D=2Þ i¼0 j¼0
r;i n;n−D; 2 rþj
ðLLÞ 2

From it, we read out reduction coefficient, with

ðrÞ
X
r X
i
ðrÞ
i−j
ðVVÞ 2 ðVLÞj ð−Þr Γðn − D=2 − rÞ
ðnÞ
¼ cn→n;i;j ðR2 Þ 2 ð5:8Þ
r−i
Cn→n cn→n;i;j ¼
Γðn − D=2Þ
rþj
i¼0 j¼0 ðLLÞ 2

× CD=2þr−n
r;i E n;n−D−ðr−iÞ;i;j Kþ
n;n−D;rþj
; ð5:9Þ
2
7
The general results for arbitrary tensor structure, general
powers as well as the coefficients for lower topologies are given
in Appendix C. where we require r, i, j having the same parity.

056025-12
UNIVERSAL TREATMENT OF THE REDUCTION FOR ONE-LOOP … PHYS. REV. D 106, 056025 (2022)

VI. DISCUSSION the final results too complicated to read. The appearance of
(reduced) Gram matrix has also been observed in our recent
Although the main target of the paper [33] is to under-
work [31,32].
stand general one-loop integrals from a geometric point of
One obvious idea is to generalize the above method to
view, it does contain many important and valuable results.
the reduction of two-loops and higher-loops. Recently,
In this paper, we have elaborated on the method to compute
using the improved PV-reduction method with auxiliary
reduction coefficients for general one-loop integrals.
vectors, we have shown how to do the general tensor
The essential idea of the method is to put the whole one-
loop Feynman integrals in the projective space, in which reduction for two-loop sunset integrals [42]. From the
integrals have compact forms and geometry properties. results in [31,32] and results in this paper, we see that these
Using a vital recursion relation of En;k ½T, one achieves the two methods have some correspondences. In other words,
wanted reduction. The advantage and the most promising they treat the same thing from different but related angles.
point of this method are that we can solve any one-loop Thus, it is natural to ask if we can translate two-loop
integrals with higher poles and tensor structures at the same integrals to a form in the projective space and establish a
time, which is demonstrated by some examples in Secs. III similar recursion relation, possibly using the reduction
and IV. The language of projective space simplifies the results of the sub-one-loop integrals studied in this paper.
reduction process a lot by keeping the elegant contractions
like ðHLÞ; ðLLÞ in these recursion relations without ACKNOWLEDGMENTS
expansion, thus making the whole reduction process a
This work is supported by Qiu-Shi Funding and Chinese
symbolic calculation. Although results listed in the paper
NSF funding under Grants No. 11935013, No. 11947301,
can be obtained by other methods, such as the PV-reduction
and No. 12047502 (Peng Huanwu Center).
and IBP method, the very compact and symmetric analytic
form is the new feature of our method. It is also efficient for
practical calculation.
APPENDIX A: RECURSION RELATIONS
For some programs based on the traditional IBP method,
OF En;k ½T
like FIRE, LiteRed, KIRA, etc. [35–41], they do reduction by
solving linear equations where the determinant of the Gram In this section, we recall the proof of the relation (3.1)
matrix appears with the full-expanded form, which makes given in [33]. Let us calculate directly

 −1 
hQ T½Xk−1 Xdn−2 Xi 1 ðXQdXÞ −1
dX ¼ −ðn þ k − 2Þ hQ T½Xk−1 Xdn−2 Xi þ ðn − 1ÞhQ−1 T½Xk−1 dn−1 Xi
ðXQXÞ
nþk−2
2 ðXQXÞ
nþk−2
2 ðXQXÞ
1
þ ðk − 1ÞXI1 Q−1 T i1 i2 i3 …ik Xi2 Xi3    dXik XI2 ∧ dn−2 XϵI1 I2 …Ik ; ðA1Þ
ðn − 2Þ! I1 i1
where we require T is completely symmetric of the last (k − 1) indices i2 ; i3 ; …ik . For simplicity, we denote
AI ¼ Q−1
Is T
si2 i3 …ik
X i2 X i3    X ik ; BIr ¼ ðQ−1 T½Xk−2 ÞIr : ðA2Þ
Thus, (A1) becomes
 −1 
hQ T½Xk−1 Xdn−2 Xi 1 ðXQdXÞ
dX ¼ nþk−2 −ðn þ k − 2Þ hAXdn−2 Xi þ ðn − 1ÞhAdn−1 Xi þ ðk − 1ÞdXI hBI Xdn−2 Xi :
ðXQXÞ
nþk−2
2 ðXQXÞ 2 ðXQXÞ
ðA3Þ

Using the fact, to simplify the first term in (A3) as

XQA ¼ XQQ−1 T½Xk−1  ¼ T½Xk ; BI I ¼ ðtrQ TÞ½Xk−2 ; XQdXhAXdn−2 Xi ¼ QXdXð−AhXdn−1 Xi þ XhAdn−1 XiÞ
ðA4Þ ¼ −ðXQAÞhXdn−1 Xi

and the Schouten identity, þ ðQXXÞhAdn−1 Xi; ðA6Þ

and the third term in (A3) as
dXI hAXdXn−2 i þ AI hXdn−1 Xi − XI hAdn−1 Xi ¼ 0; ∀ AI
ðA5Þ dXI hBI Xdn−2 Xi ¼ XI hBI dn−1 Xi − BII hXdn−1 Xi; ðA7Þ

056025-13
BO FENG, JIANYU GONG, and TINGFEI LI PHYS. REV. D 106, 056025 (2022)

the (A3) becomes In (A9), the lhs is the first term in rhs of (A8), while the first
 −1 term on the rhs is the boundary contribution of lhs of (A8).
hQ T½Xk−1 Xdn−2 Xi T½Xk hXdn−1 Xi
dX nþk−2 ¼ ðn þ k − 2Þ nþk
ðXQXÞ 2 ðXQXÞ 2 APPENDIX B: MORE RESULTS
− ðk − 1ÞtrQ T½Xk−2 hXdn−1 Xi; 1. Bubbles
ðA8Þ Here list the results for rank from r ¼ 1 to r ¼ 4, where
we set q1 ¼ 0.
which is nothing but the wanted (3.1). (i) r ¼ 1
The derivation above can be generalized to the case
X ðLLÞðH i VÞ−ðHi LÞðVLÞ 2ðVLÞ
det Q ¼ 0, where the Q−1 does not exist. In this case, we ð1Þ
I2 ¼ − I 2;î þ I2:
can replace Q−1 by arbitrary matrix Q̃ in (A1) and repeat i ðLLÞ ðLLÞ
the same derivations to reach the similar expression likes ðB1Þ
(A8), except Q−1 replaced by Q̃. Rearranging (A8), we get
So we have
ðbÞ
En;k ½ðQQ̃TÞ ¼ αn;k En−1;k−1 ½ðHb Q̃TÞ þ βn;k En;k−2 ½trQ̃ T;
ð1Þ ðLLÞðH2 VÞ − ðH2 LÞðVLÞ R·q
ðA9Þ C2→2;2̂ ¼ − ¼− 2 2;
ðLLÞ q2
where ðQQ̃TÞ has the same rank as T, ð1Þ ðLLÞðH1 VÞ − ðH1 LÞðVLÞ R · q2
C2→2;1̂ ¼ − ¼ 2 ;
ðLLÞ q2
ðQQ̃TÞI1 I2 ;…;Ik ¼ QI1 J1 Q̃J1 J2 T J2 ;I2 ;I3 ;…;Ik ; 2ðVQ−1 LÞ ðm21 − m22 þ q22 ÞR · q2
ð1Þ
C2→2 ¼ ¼ : ðB2Þ
ðtrQ̃ TÞI3 ;I4 ;…;Ik ¼ Q̃I1 I2 T I1 I2 …Ik : ðA10Þ ðLQ−1 LÞ q22

(ii) r ¼ 2

ð2Þ 2R2 ðH i LÞ 2DðHi LÞðVLÞ2 2ðHi VÞððVLÞ þ ðVLÞðiÞ Þ 2ðHi LÞð−DðVLÞ2 þ ðVLÞ2 þ ðLLÞðiÞ ðVVÞÞ
C2→2;î ¼ þ − − ;
ðD − 1ÞðLLÞ ðD − 1ÞðLLÞ2 ðLLÞðiÞ ðD − 1ÞðLLÞðiÞ ðLLÞ
ð2Þ 4ðR2 − ðVVÞÞ 4DðVLÞ2
C2→2 ¼ þ : ðB3Þ
ðD − 1ÞðLLÞ ðD − 1ÞðLLÞ2
The exact form is
ð2Þ ðm21 − m22 þ q22 ÞR2 Dðm21 − m22 þ q22 ÞðR · q2 Þ2
C2→2;2̂ ¼ − ;
ðD − 1Þq22 ðD − 1Þq42
ð2Þ R2 ðm21 − m22 − q22 Þ ð−Dm21 þ Dm22 − 3Dq22 þ 4q22 ÞðR · q2 Þ2
C2→2;1̂ ¼ − − ;
ðD − 1Þq22 ðD − 1Þq42
ð2Þ R2 ð−2m21 ðm22 þ q22 Þ þ ðm22 − q22 Þ2 þ m41 Þ ð−2m21 ðDm22 − ðD − 2Þq22 Þ þ Dðm22 − q22 Þ2 þ Dm41 ÞðR · q2 Þ2
C2→2 ¼ − þ :
ðD − 1Þq22 ðD − 1Þq42
(iii) r ¼ 3

ð3Þ
4ðD þ 1ÞðHi VÞððVLÞðiÞ ðVLÞ þ ðVLÞ2 þ ðVLÞ2ðiÞ Þ 4ðH i VÞð−2ðVVÞ − ðVVÞðiÞ þ 3R2 Þ
C2→2;î ¼− −
DðLLÞ2ðiÞ DðLLÞðiÞ
4ðD þ 2ÞðHi LÞðVLÞ3 4ðHi LÞðVLÞðDðVLÞ2 þ ðVLÞ2 − 3ðLLÞðiÞ ðVVÞ þ 3R2 ðLLÞðiÞ Þ
þ þ
ðD − 1ÞðLLÞ3 DðLLÞ2ðiÞ ðLLÞ
4ðH i LÞðVLÞ
þ ðD2 ðVLÞ2 þ DðVLÞ2 − 3DðLLÞðiÞ ðVVÞ − 2ðVLÞ2 þ 3DR2 ðLLÞðiÞ Þ;
ðD − 1ÞDðLLÞðiÞ ðLLÞ2
ð3Þ 24R2 ðVLÞ 8ðD þ 2ÞðVLÞ3 24ðVVÞðVLÞ
C2→2 ¼ 2
þ − : ðB4Þ
ðD − 1ÞðLLÞ ðD − 1ÞðLLÞ3 ðD − 1ÞðLLÞ2

056025-14
UNIVERSAL TREATMENT OF THE REDUCTION FOR ONE-LOOP … PHYS. REV. D 106, 056025 (2022)

The exact form is

ð3Þ ðD þ 2Þðm21 − m22 Þ2 ðR · q2 Þ3 3R2 R · q2

C2→2;2̂ ¼ − þ
ðD − 1Þq62 D−1
ð2DðD þ 2Þm22 − 2ðD2 − 2D þ 4Þm21 ÞðR · q2 Þ3 þ 3Dðm21 − m22 Þ2 R2 R · q2
þ
ðD − 1ÞDq42
R · q2 ð6R2 ððD − 2Þm21 þ Dm22 Þ þ DðD þ 2ÞðR · q2 Þ2 Þ
− ;
ðD − 1ÞDq22
ð3Þ 3R2 ð4ðD − 1Þm22 q22 − Dm41 þ 2Dm22 m21 − Dm42 þ Dq42 ÞR · q2
C2→2;1̂ ¼
ðD − 1ÞDq42
ðR · q2 Þ3
½−4ðD2 þ D − 2Þm22 q22 − 2Dm21 ððD þ 2Þm22 − 2ðD − 1Þq22 Þ
ðD − 1ÞDq62
þ DðD þ 2Þm41 þ DðD þ 2Þm42 þ Dð7D − 10Þq42 ;
ð3Þ ðm21 − m22 þ q22 ÞðR · q2 Þ3
C2→2 ¼ ½−2m21 ððD þ 2Þm22 − ðD − 4Þq22 Þ þ ðD þ 2Þðm22 − q22 Þ2 þ ðD þ 2Þm41 
ðD − 1Þq62
3R2 ðm21 − m22 þ q22 Þð−2m21 ðm22 þ q22 Þ þ ðm22 − q22 Þ2 þ m41 ÞR · q2
− : ðB5Þ
ðD − 1Þq42

(iv) r ¼ 4

ð4Þ 8ðDþ2ÞðH i LÞðVLÞ2 ððD2 þ3D− 4ÞðVLÞ2 þ6DðLLÞðiÞ ðR2 − ðVVÞÞÞ

C2→2;î ¼
ðD− 1ÞDðDþ1ÞðLLÞðiÞ ðLLÞ3
8ðHi LÞ
þ ððD2 þ4Dþ3ÞðVLÞ4 þ3ðLLÞ2ðiÞ ðR2 −ðVVÞÞ2 þ6ðDþ1ÞðLLÞðiÞ ðVLÞ2 ðR2 − ðVVÞÞÞ
DðDþ1ÞðLLÞ3ðiÞ ðLLÞ
8ðHi LÞ
þ ððD3 þ4D2 − D−4ÞðVLÞ4 þ6ðD2 þD−2ÞðLLÞðiÞ ðVLÞ2 ðR2 − ðVVÞÞ
DðD − 1ÞðLLÞ2ðiÞ ðLLÞ2
2

8ðDþ2ÞðDþ4ÞðH i LÞðVLÞ4
þ3DðLLÞ2ðiÞ ðR2 −ðVVÞÞ2 Þ þ
ðD2 − 1ÞðLLÞ4
8ðHi VÞ
− ððVLÞð−5ðVVÞ − ðVVÞðiÞ þ6R2 Þ þ ðVLÞðiÞ ð−3ðVVÞ− 3ðVVÞðiÞ þ6R2 ÞÞ
DðLLÞ2ðiÞ
8ðDþ3ÞððVLÞþðVLÞðiÞ ÞððVLÞ2 þðVLÞ2ðiÞ ÞðHi VÞ
− ;
DðLLÞ3ðiÞ

ð4Þ 96ðDþ2ÞðVLÞ2 ðR2 − ðVVÞÞ 48ðR2 − ðVVÞÞ2 16ðD2 þ6Dþ8ÞðVLÞ4

C2→2 ¼ þ 2 þ : ðB6Þ
ðD2 −1ÞðLLÞ3 ðD − 1ÞðLLÞ2 ðD2 − 1ÞðLLÞ4

2. Triangles
For triangle topology, we have presented results for scalar triangles with higher poles. Here, we present some examples,
including the tensor triangles without higher poles and with higher poles.
For the tensor triangles without higher poles, we have
(i) r ¼ 1

ð1Þ ðHi LÞðVLÞ ð1Þ 2ðVLÞ

C3→3;î ¼ − ðHi VÞ; C3→3 ¼ : ðB7Þ
ðLLÞ ðLLÞ

056025-15
BO FENG, JIANYU GONG, and TINGFEI LI PHYS. REV. D 106, 056025 (2022)

The expression is the same as (B2) for bubble topology. This phenomenon is not an accident and will be persistent to
other topologies.
(ii) r ¼ 2

ð2Þ ðHi LÞðHj LÞðiÞ ðVLÞ2 ðHj LÞðiÞ ðHi VÞððVLÞðiÞ þ ðVLÞÞ
C ¼ − þ ðHi VÞðHj VÞðiÞ þ ði ↔ jÞ;
3→3;b
ij ðLLÞðiÞ ðLLÞ ðLLÞðiÞ

ð2Þ 2ðHi LÞððD − 2ÞðVLÞ2 þ ðLLÞðiÞ ðR2 − ðVVÞÞÞ 2ðD − 1ÞðHi LÞðVLÞ2 2ðHi VÞððVLÞðiÞ þ ðVLÞÞ
C3→3;î ¼ þ ;
ðD − 2ÞðLLÞðiÞ ðLLÞ ðD − 2ÞðLLÞ2 ðLLÞðiÞ
ð2Þ 4ððD − 1ÞðVLÞ2 þ ðLLÞðR2 − ðVVÞÞÞ
C3→3 ¼ : ðB8Þ
ðD − 2ÞðLLÞ2

Again, the last two coefficients are very similar to these given in (B3).
For the tensor with higher poles, we have
(i) v ¼ 4, r ¼ 1

ð1Þ 1
C ¼ ðD − 2ÞðLLÞðijÞ ððHi SÞðHj SÞðiÞ þ ði ↔ jÞÞ;
v3 →3;b
ij 8
 
ð1Þ D − 3 ðHi LÞðSLÞ2
Cv ¼ − ðHi SÞðSLÞðiÞ − ðHi SÞðSLÞ ;
3 →3;î 4 ðLLÞ
ð1Þ ðD − 3ÞðSLÞ2 1
Cv3 →3 ¼ − ðSSÞ: ðB9Þ
2ðLLÞ 2

Here, we have suppressed the process to take the coefficient of tzv3 −1 ,i.e., jtzv3 −1 . The similarity of the above
expression with the one given in (3.25) is obvious.
(ii) v ¼ 4, r ¼ 2

ð2Þ ðD − 2ÞðH i LÞðHj LÞðiÞ ðSLÞ3 ðD − 2ÞðHj LÞðiÞ ðHi SÞððSLÞ2ðiÞ þ ðSLÞðSLÞðiÞ þ ðSLÞ2 Þ
C ¼ −
v3 →3;b
ij 6ðLLÞðiÞ ðLLÞ 6ðLLÞðiÞ
1
þ ðD − 2ÞðH i SÞðHj SÞðiÞ ððSLÞðiÞ þ ðSLÞ þ ðSLÞðijÞ Þ þ ði ↔ jÞ;
6
ð2Þ ðHi LÞðSLÞððD − 2ÞðSLÞ2 þ 3ðLLÞðiÞ ðR2 − ðSSÞÞÞ ðD − 1ÞðHi LÞðSLÞ3
Cv ¼ þ
3 →3;î 3ðLLÞðiÞ ðLLÞ 3ðLLÞ2
ðH i SÞððLLÞðiÞ ððSSÞðiÞ þ 2ðSSÞ − 3R2 Þ − ðD − 2ÞððSLÞ2ðiÞ þ ðSLÞðSLÞðiÞ þ ðSLÞ2 ÞÞ
þ ;
3ðLLÞðiÞ
ð2Þ 2ðD − 1ÞðSLÞ3 2ðSLÞðR2 − ðSSÞÞ
Cv3 →3 ¼ þ : ðB10Þ
3ðLLÞ2 ðLLÞ

Here, when calculating the reduction coefficients, we take the coefficient of zv3 −1 for terms containing one R2 and
one S, while we take the coefficient of t2 zv3 −1 for terms containing three S’s.

3. Boxes
For the box topology, we present three cases: (1) r ¼ 1; v − n ¼ 0, (2) r ¼ 0; v − n ¼ 1, (3) r ¼ 1; v − n ¼ 1. As
pointed out in the triangle topology, the reduction coefficients have some similarities between different topologies. In fact,
the similarity is classified by the pair ðr; v − nÞ as one can check by using the results listed in the Appendix and the main
body of the paper.

056025-16
UNIVERSAL TREATMENT OF THE REDUCTION FOR ONE-LOOP … PHYS. REV. D 106, 056025 (2022)

(i) r ¼ 1; v − n ¼ 0

ð1Þ ðHi LÞðVLÞ ð1Þ 2ðVLÞ

C4→4;î ¼ − ðHi VÞ; C4→4 ¼ : ðB11Þ
ðLLÞ ðLLÞ

(ii) r ¼ 0; v − n ¼ 1

1
C b ¼− ðD − 2ÞðLLÞðijkÞ ððHj LÞðiÞ ðHk LÞðijÞ ðHi ZÞ þ permutations of ðijkÞÞ;
v4 →4;ijk 16
1
C ¼ ðD − 3ÞðLLÞðijÞ ððHj LÞðiÞ ðHi ZÞ þ ði ↔ jÞÞ;
v4 →4;b
ij 8
1
Cv4 →4;î ¼ − ðD − 4ÞðLLÞðiÞ ðHi ZÞ;
4
1
Cv4 →4 ¼ ðD − 5ÞðZLÞ: ðB12Þ
2
(iii) r ¼ 1; v − n ¼ 1

ð1Þ 1
C ¼− ðD − 2ÞðLLÞðijkÞ ððHk LÞðijÞ ðH i SÞðHj SÞðiÞ þ permutations of ðijkÞÞ;
b
v4 →4;ijk 16
ð1Þ 1
C ¼ ðD − 3ÞðLLÞðijÞ ððH i SÞðHj SÞðiÞ þ ði ↔ jÞÞ;
v4 →4;b
ij 8
ð1Þ ðD − 4ÞððLLÞðHi SÞððSLÞðiÞ þ ðSLÞÞ − ðHi LÞðSLÞ2 Þ
Cv ¼− ;
4 →4;î 4ðLLÞ
 2

ð1Þ 1 ðD − 4ÞðSLÞ
Cv4 →4 ¼ − ðSSÞ : ðB13Þ
2 ðLLÞ

4. Pentagons
For pentagon topology, the reduction coefficients are similar, so we just present one example, i.e., scalar pentagon with
higher poles.
(i) v ¼ 6

1
C ¼ ðD − 2ÞðLLÞðijklÞ ððHi LÞðjkÞ ðHj ZÞðHk LÞðjÞ ðH l LÞðijkÞ þ permutations of ðijklÞÞ;
v5 →5;ic
jkl 32
1
C b ¼ − ðD − 3ÞðLLÞðijkÞ ððHi LÞðjÞ ðHk LÞðijÞ ðHj ZÞ þ permutations of ðijkÞÞ;
v5 →5;ijk 16
ð1Þ 1
C ¼ ðD − 4ÞðLLÞðijÞ ððHj LÞðiÞ ðHi ZÞ þ ði ↔ jÞÞ;
5→5;b
ij 8
1
Cv5 →5;î ¼ − ðD − 5ÞðLLÞðiÞ ðHi ZÞ;
4
1
Cv5 →5 ¼ ðD − 6ÞðZLÞ: ðB14Þ
2
One thing we want to point out is that these coefficients have a manifest permutation symmetry. Using these observations,
the expression can be very compact, as shown above.

APPENDIX C: GENERAL EXPRESSION OF REDUCTION COEFFICIENTS

One can solve the recursion relations (3.6) in Sec. III iteratively8 and get general expressions for the final reduction
coefficients. Here, we present only the final results for nondegenerate Q without derivation details. We define

8
For the degenerate case discussed in the Sec. IV, one can do similar computation, although it will be more complicated.

056025-17
BO FENG, JIANYU GONG, and TINGFEI LI PHYS. REV. D 106, 056025 (2022)

Y
s
F n;k;i
i−nv ;s ¼ P i−nv þ1;s
k;i
αn;k−2l ðC1Þ
l¼0

with

i−d;s ¼ ði þ 1 − dÞP i−dþ2;s−1 þ ðk − i þ 1 þ d − 2sÞP i−dþ1;s−1 ;

P k;i k;i k;i
ðC2Þ

i−d;s ¼ 0 for either d < 0 or s < 0. With the initial condition P 0;0 ¼ 1, (C2) can be solved. For simplicity, we
where we set P k;i k;i

define
ðb Þ ðb ;b ;b ;…;bl Þ
l
En−l;k−l ¼ En−l;k−l
1 2 3
; ðWVÞðbl Þ ¼ ðWVÞðb1 ;b2 ;b3 ;…;bl Þ
ðHLÞb̂l ¼ ðHb1 LÞðHb2 LÞðb1 Þ ðHb3 LÞðb1 b2 Þ    ðH bl LÞðbl−1 Þ
ðs;j Þ 1 −j2 −j3 −−jl
ðLLÞðbl Þl ¼ ðLLÞj1 ðLLÞjðb2 1 Þ ðLLÞjðb3 1 ;b2 Þ    ðLLÞjðbl l−1 Þ ðLLÞs−j
ðbl Þ ðC3Þ

and

X
i X
i−1 X
l1 X
i−l 1 −1 X
l1 −1 X
l2 l1X
−l2 −1 X
la−1 −1 X
la la−1X
−la −1
¼ 
fl;j;sga l1 ¼0 j1 ¼0 s ¼⌈i−l1 −1⌉ l2 ¼0 j2 ¼0 s ¼⌈l1 −l2 −1⌉ la ¼0 ja ¼0 s ¼⌈la−1 −la −1⌉
1 2 2 2 a 2

n;k;i n−1;k−1−2s1 ;l1 n−2;k−2−2s1 −2s2 ;l2 1 −2s2 −−2sa−1 ;la−1

F n;k;i
fl;sg ¼ F l1 ;s1 F l2 ;s2 F l3 ;s3    F n−a;k−a−2s
la ;sa
a

Y
a
−1þl −lp −sp 1−l þlp þ2sp la −ja
S l0 ;fl;j;sga ;ba ¼ ðSSÞðbp−1p−1
Þ ðSLÞðbp−1
p−1
Þ ðSSÞðb2a Þ ðSLÞjðba a Þ ; ðC4Þ
p¼1

where specially

X
i X
i
i−j0
¼ ; F n;k;i
fl;sg ¼ 1; ðHSÞ b ¼ 1; S i;fl;j;sg0 ;b0 ¼ ðSSÞ 2 ðSLÞj0 ; E n;k−2js0 j;l0 ;j0 ¼ E n;k;i;j0 : ðC5Þ
0 b0
fl;j;sg0 j0 ¼0

Then, we denote some notations used in dimension shifting as follows:

1 −αn;k−2sþ2j Y
s Y
j
K−n;k;s ¼ Qs ; K−n;k;s;j ¼ Qj ; Kþ
n;k;s ¼ βn;kþ2sþ2−2p ; Kþ
n;k;s;j ¼ αn;kþ2s−2j βn;kþ2sþ2−2p ;
l¼1 ðβn;k−2sþ2l Þ p¼1 βn;k−2sþ2p p¼1 p¼1

ðC6Þ

and

X
s X X
s−j1 s−j 1 −j2 s−j1 −jX
2 −−jl−1
1
K−;b
n;k;2s;l ¼
l
 ðs;j Þ
K−n;k;s;j1 K−n−1;k−1;s−j1 ;j2 K−n−2;k−2;s−j1 −j2 ;j3    K−n−l;k−l;s−j1 −j2 −j3 −−jl
j1 ¼1 j2 ¼1 j3 ¼1 jl ¼1 ðLLÞðbl Þl
X
s X X
s−j1 s−j 1 −j2 s−j1 −jX
2 −−jl−1

Kþ;b
n;k;2s;l ¼
l
 Kþ þ þ þ
n;k;s;j1 Kn−1;k−1;s−j1 ;j2 Kn−2;k−2;s−j1 −j2 ;j3    Kn−l;k−l;s−j1 −j2 −j3 −−jl
j1 ¼0 j2 ¼0 j3 ¼0 jl ¼0

Kbn;k;2s;l
SignðsÞ;b
l
¼ Kn;k;j2sj;l l : ðC7Þ

The final expression for general reduction coefficients is

X X
r ξðn;v;iÞ
X 
ðba Þba2 
ðrÞ
¼ σ 2 r−i 
C i;a1 ;a2 ;fl;j;sg ðR Þ S ξ;fl;j;sga
cn;v;r ;b ðHSÞc ðHLÞ Kn−a11 ;−μ−2js i ; ðC8Þ
2
vn →n;bba a1 1 a1 ba1 c
ðba1 Þba2 a1 j−la1 −ja1 ;a2
a1 þa2 ¼a; i¼0 fl;j;sga t zvn −1
σ∈S½ba  1

056025-18
UNIVERSAL TREATMENT OF THE REDUCTION FOR ONE-LOOP … PHYS. REV. D 106, 056025 (2022)

where we have defined μðn; v; r; iÞ ¼ r − i þ 2n − 2v; ξðn; v; iÞ ¼ v − n þ i, and the dimensionless factor is

ð−Þrþvþnþa1 þa2 Γðv − D=2 − rÞi! n;n−D−μðn;v;r;iÞ;ξðn;v;iÞ

i;a1 ;a2 ;fl;j;sg ¼
cn;v;r
a1
F
Γðn − a1 − a2 − D=2Þðv − n þ iÞ! fl;sga1
E n−a1 ;n−a1 −D−μðn;v;r;iÞ−2jsa j;la ;ja :
1 1 1
ðC9Þ

[1] L. M. Brown and R. P. Feynman, Radiative corrections to [20] G. Ossola, C. G. Papadopoulos, and R. Pittau, Reducing full
Compton scattering, Phys. Rev. 85, 231 (1952). one-loop amplitudes to scalar integrals at the integrand
[2] D. B. Melrose, Reduction of Feynman diagrams, Nuovo level, Nucl. Phys. B763, 147 (2007).
Cimento 40, 181 (1965). [21] G. Ossola, C. G. Papadopoulos, and R. Pittau, Numerical
[3] G. Passarino and M. J. G. Veltman, One loop corrections for evaluation of six-photon amplitudes, J. High Energy Phys.
eþ e− annihilation into μþ μ− in the Weinberg model, Nucl. 07 (2007) 085.
Phys. B160, 151 (1979). [22] R. K. Ellis, W. T. Giele, and Z. Kunszt, A numerical
[4] G. ’t Hooft and M. J. G. Veltman, Scalar one loop integrals, unitarity formalism for evaluating one-loop amplitudes,
Nucl. Phys. B153, 365 (1979). J. High Energy Phys. 03 (2008) 003.
[5] W. L. van Neerven and J. A. M. Vermaseren, Large loop [23] Z. Bern, L. J. Dixon, D. C. Dunbar, and D. A. Kosower, One
integrals, Phys. Lett. 137B, 241 (1984). loop n point gauge theory amplitudes, unitarity and collinear
[6] R. G. Stuart, Algebraic reduction of one loop Feynman limits, Nucl. Phys. B425, 217 (1994).
diagrams to scalar integrals, Comput. Phys. Commun. 48, [24] R. Britto, F. Cachazo, and B. Feng, Generalized unitarity
367 (1988). and one-loop amplitudes in N ¼ 4 super-Yang-Mills, Nucl.
[7] G. J. van Oldenborgh and J. A. M. Vermaseren, New algo- Phys. B725, 275 (2005).
rithms for one loop integrals, Z. Phys. C 46, 425 [25] R. Britto, E. Buchbinder, F. Cachazo, and B. Feng,
(1990). One-loop amplitudes of gluons in SQCD, Phys. Rev. D
[8] Z. Bern, L. J. Dixon, and D. A. Kosower, Dimensionally 72, 065012 (2005).
regulated one loop integrals, Phys. Lett. B 302, 299 (1993); [26] J. M. Campbell, E. W. N. Glover, and D. J. Miller, One loop
Erratum, Phys. Lett. B 318, 649 (1993). tensor integrals in dimensional regularization, Nucl. Phys.
[9] Z. Bern, L. J. Dixon, and D. A. Kosower, Dimensionally B498, 397 (1997).
regulated pentagon integrals, Nucl. Phys. B412, 751 (1994). [27] Z. Bern, L. J. Dixon, and D. A. Kosower, One loop
[10] J. Fleischer, F. Jegerlehner, and O. V. Tarasov, Algebraic amplitudes for eþ e− to four partons, Nucl. Phys. B513, 3
reduction of one loop Feynman graph amplitudes, Nucl. (1998).
Phys. B566, 423 (2000). [28] C. Anastasiou, R. Britto, B. Feng, Z. Kunszt, and P.
[11] T. Binoth, J. P. Guillet, and G. Heinrich, Reduction formal- Mastrolia, Unitarity cuts and reduction to master integrals
ism for dimensionally regulated one loop N point integrals, in d dimensions for one-loop amplitudes, J. High Energy
Nucl. Phys. B572, 361 (2000). Phys. 03 (2007) 111.
[12] A. Denner and S. Dittmaier, Reduction of one loop tensor [29] R. Britto and E. Mirabella, Single cut integration, J. High
five point integrals, Nucl. Phys. B658, 175 (2003). Energy Phys. 01 (2011) 135.
[13] G. Duplancic and B. Nizic, Reduction method for dimen- [30] B. Feng and H. Wang, Reduction of one-loop integrals with
sionally regulated one loop N point Feynman integrals, Eur. higher poles by unitarity cut method, J. High Energy Phys.
Phys. J. C 35, 105 (2004). 07 (2021) 204.
[14] A. Denner and S. Dittmaier, Reduction schemes for one- [31] B. Feng, T. Li, and X. Li, Analytic tadpole coefficients of
loop tensor integrals, Nucl. Phys. B734, 62 (2006). one-loop integrals, J. High Energy Phys. 09 (2021) 081.
[15] R. K. Ellis and G. Zanderighi, Scalar one-loop integrals for [32] C. Hu, T. Li, and X. Li, One-loop Feynman integral
QCD, J. High Energy Phys. 02 (2008) 002. reduction by differential operators, Phys. Rev. D 104,
[16] G. Ossola, C. G. Papadopoulos, and R. Pittau, Reducing full 116014 (2021).
one-loop amplitudes to scalar integrals at the integrand [33] N. Arkani-Hamed and E. Y. Yuan, One-loop integrals
level, Nucl. Phys. B763, 147 (2007). from spherical projections of planes and quadrics, arXiv:
[17] Z. Bern, L. J. Dixon, D. C. Dunbar, and D. A. Kosower, 1712.09991.
Fusing gauge theory tree amplitudes into loop amplitudes, [34] M. E. Peskin and D. V. Schroeder, An Introduction to
Nucl. Phys. B435, 59 (1995). Quantum Field Theory (Addison-Wesley, Reading, USA,
[18] K. G. Chetyrkin and F. V. Tkachov, Integration by parts: The 1995).
algorithm to calculate beta functions in 4 loops, Nucl. Phys. [35] A. V. Smirnov, Algorithm FIRE—Feynman Integral REduction,
B192, 159 (1981). J. High Energy Phys. 10 (2008) 107.
[19] F. V. Tkachov, A theorem on analytical calculability of four [36] A. V. Smirnov, FIRE5: A C++ implementation of Feynman
loop renormalization group functions, Phys. Lett. 100B, 65 Integral REduction, Comput. Phys. Commun. 189, 182
(1981). (2015).

056025-19
BO FENG, JIANYU GONG, and TINGFEI LI PHYS. REV. D 106, 056025 (2022)

[37] R. N. Lee, LiteRed 1.4: A powerful tool for reduction [40] P. Maierhöfer, J. Usovitsch, and P. Uwer, KIRA—A Feynman
of multiloop integrals, J. Phys. Conf. Ser. 523, 012059 integral reduction program, Comput. Phys. Commun. 230, 99
(2014). (2018).
[38] A. V. Smirnov and F. S. Chuharev, FIRE6: Feynman Integral [41] M. Gerlach, F. Herren, and M. Lang, TAPIR: A tool for
REduction with modular arithmetic, Comput. Phys. Commun. topologies, amplitudes, partial fraction decomposition and
247, 106877 (2020). input for reductions, arXiv:2201.05618.
[39] A. von Manteuffel and C. Studerus, REDUZE 2—distributed [42] B. Feng and T. Li, PV-reduction of sunset topology with
Feynman integral reduction, arXiv:1201.4330. auxiliary vector, Commun. Theor. Phys. 74, 095201 (2022).

056025-20