A proposal for removing 𝝅𝑵 -state contamination from

the nucleon induced pseudoscalar form factor in lattice

QCD

Shoichi Sasaki,𝑎,∗ Yasumichi Aoki,𝑏 Ken-Ichi Ishikawa,𝑐 Yoshinobu Kuramashi,𝑑

Kohei Sato,𝑒 Eigo Shintani,𝑑 Ryutaro Tsuji, 𝑓 Hiromasa Watanabe𝑔 and Takeshi
Yamazakiℎ,𝑑
arXiv:2501.13490v1 [hep-lat] 23 Jan 2025

(PACS Collaboration)
𝑎 Department of Physics, Tohoku University, 980-8578, Sendai, Japan
𝑏 RIKEN Center for Computational Science, 650-0047, Kobe, Japan
𝑐 Core of Research for the Energetic Universe, Graduate School of Advanced Science and Engineering,

Hiroshima University 739-8526, Higashi-Hiroshima, Japan

𝑑 Center for Computational Sciences, University of Tsukuba, 305-8577, Tsukuba, Japan
𝑒 Degree Programs in Pure and Applied Sciences, Graduate School of Science and Technology, University

of Tsukuba, Ibaraki 305-8571, Japan

𝑓 High Energy Accelerator Research Organization (KEK), 305-0801, Tsukuba, Japan
𝑔 Yukawa Institute for Theoretical Physics, Kyoto University, Kyoto 606-8502, Japan
ℎ Institute of pure and Applied Sciences, University of Tsukuba, 305-8571, Tsukuba, Japan

E-mail: [email protected]

In the PACS10 project, the PACS collaboration has generated three sets of the PACS10 gauge
configurations at the physical point with lattice volume larger than (10 fm) 4 and three different
lattice spacings. The isovector nucleon form factors had been already calculated by using two
sets of the PACS10 gauge configurations. In our strategy, the smearing parameters of the nucleon
interpolation operator were highly optimized to eliminate as much as possible the contribution of
excited states in the nucleon two-point function. This strategy was quite successful in calculations
of the electric (𝐺 𝐸 ), magnetic (𝐺 𝑀 ) and axial-vector (𝐹𝐴) form factors, while the induced
pseudoscalar (𝐹𝑃 ) and pseudoscalar (𝐺 𝑃 ) form factors remained strongly affected by residual
contamination of 𝜋𝑁-state contribution. In this work, we propose a simple method to remove
the 𝜋𝑁-state contamination from the 𝐹𝑃 form factor, and then evaluate the induced pseudoscalar
charge 𝑔 ∗𝑃 and the pion-nucleon coupling 𝑔 𝜋 𝑁 𝑁 from existing data in a new analysis. Applying
this method to the 𝐺 𝑃 form factor is also considered with a help of the axial Ward-Takahashi
identity.

The 41st International Symposium on Lattice Field Theory (Lattice 2024)

July 28th - August 3rd, 2024
The University of Liverpool, United Kingdom

∗ Speaker

© Copyright owned by the author(s) under the terms of the Creative Commons
Attribution-NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND 4.0). https://pos.sissa.it/
A proposal for removing 𝜋𝑁-state contamination from the nucleon induced pseudoscalar.... Shoichi Sasaki

1. Introduction

The axial structure of the nucleon is highly connected with the physics of chiral symmetry
and its spontaneous breaking, which ensures the presence of pseudo Nambu-Goldstone particles
such as the pion. This is empirically known as the partially conserved axial-vector current (PCAC)
hypothesis, where the divergence of the axial-vector current is proportional to the pion field.
Applying this idea to the axial-vector matrix element of the nucleon given by

⟨𝑁 ( 𝑝 ′ )| 𝐴 𝛼 (𝑥)|𝑁 ( 𝑝)⟩ = 𝑢 𝑁 ( 𝑝 ′ ) 𝛾 𝛼 𝛾5 𝐹𝐴 (𝑞 2 ) + 𝑖𝑞 𝛼 𝛾5 𝐹𝑃 (𝑞 2 ) 𝑢 𝑁 ( 𝑝)𝑒 𝑖𝑞·𝑥

 
(1)

with 𝑞 = 𝑝 ′ − 𝑝, a specific relation, known as the Goldberger-Treiman (GT) relation [1], is derived
between the axial-vector coupling defined by the axial-vector (𝐹𝐴) form factor at 𝑞 2 = 0 and the
residue of the pion-pole structure in the induced pseudo-scalar (𝐹𝑃 ) form factor. Instead of PCAC,
the axial Ward-Takahashi identity, 𝜕𝛼 𝐴 𝛼 (𝑥) = 2𝑚𝑃(𝑥), leads to the generalized GT relation [2, 3]:

2𝑀 𝑁 𝐹𝐴 (𝑞 2 ) − 𝑞 2 𝐹𝑃 (𝑞 2 ) = 2𝑚𝐺 𝑃 (𝑞 2 ), (2)

which is satisfied among the three nucleon form factors including the pseudoscalar (𝐺 𝑃 ) form factor
defined in the pseudoscalar matrix element of the nucleon as

⟨𝑁 ( 𝑝 ′ )|𝑃(𝑥)|𝑁 ( 𝑝)⟩ = 𝑢 𝑁 ( 𝑝 ′ ) 𝛾5 𝐺 𝑃 (𝑞 2 ) 𝑢 𝑁 ( 𝑝)𝑒 𝑖𝑞· 𝑥 .

 
(3)

In addition, the following pion-pole dominance (PPD) ansätz [4] for 𝐹𝑃 (𝑞 2 ) and 𝐺 𝑃 (𝑞 2 ) at low 𝑞 2 ,

2𝑀 𝑁 𝐹𝐴 (𝑞 2 ) 𝑚 2𝜋
𝐹𝑃PPD (𝑞 2 ) = and 2𝑚𝐺 PPD 2
𝑃 (𝑞 ) = 2𝑀 𝑁 𝐹 𝐴 (𝑞 2
) , (4)
𝑞 2 + 𝑚 2𝜋 𝑞 2 + 𝑚 2𝜋

satisfies the generalized GT relation (2).

Although the axial-structure of the nucleon has been studied extensively in lattice QCD at
the physical point, no results have been obtained that satisfy the generalized GT relation well or
give better accuracy than the PPD model. Indeed, in our previous works [5, 6], both the 𝐹𝑃
and 𝐺 𝑃 form factors are significantly underestimated in the low-𝑞 2 region compared to the PPD
model. This is simply due to strong 𝜋𝑁 excited-state contamination. In this work, we propose a
simple method to remove the 𝜋𝑁-state contamination from the 𝐹𝑃 and 𝐺 𝑃 form factors, and then
evaluate the induced pseudoscalar charge 𝑔 ∗𝑃 = 𝑚 𝜇 𝐹𝐴 (0.88𝑚 2𝜇 ) and the pion-nucleon coupling
2)
𝑔 𝜋 𝑁 𝑁 = lim𝑞2 →∞ (𝑞 2 + 𝑚 2𝜋 ) 𝐹𝑃2𝐹(𝑞𝜋 from existing data in a new analysis.

2. Standard method

The nucleon two-point (2pt) function from the source-time position (denoted 𝑡src ) to the sink-
time position (denoted 𝑡sink ) is defined as

1 n o 1 + 𝛾4
𝐶 𝑁 (𝑡src − 𝑡sink ; 𝒑) = Tr P+ ⟨𝑁 (𝑡sink ; 𝒑)𝑁 (𝑡src ; − 𝒑)⟩ with P+ = , (5)
4 2
where the nucleon operator 𝑁 (𝑡; 𝒒) carrying a three-dimensional momentum 𝒑.

2
A proposal for removing 𝜋𝑁-state contamination from the nucleon induced pseudoscalar.... Shoichi Sasaki

The nucleon form factors are extracted from the nucleon three-point (3pt) function consisting
of the nucleon source and sink operators with a given local current (𝐽) defined as
1 n o
𝐶 𝐽5𝑧 (𝑡; 𝒑 ′ , 𝒑) = Tr P 5𝑧 ⟨𝑁 (𝑡sink ; 𝒑 ′ )𝐽 (𝑡; 𝒒)𝑁 (𝑡src ; − 𝒑) , (6)
4
where the projection operator P 5𝑧 = P+ 𝛾5 𝛾3 is chosen for 𝐽 = 𝐴 𝛼 or 𝑃. We then calculate the
following ratio constructed from an appropriate combination of the 2pt and 3pt functions [7, 8] with
a fixed source-sink separation (𝑡sep ≡ 𝑡sink − 𝑡src ):
√︄
5𝑧
(𝑡; 𝒑 ′ , 𝒑)
′
𝐶 𝐽 𝐶 𝑁 (𝑡sink − 𝑡; 𝒑)𝐶 𝑁 (𝑡 − 𝑡src ; 𝒑 ′ )𝐶 𝑁 (𝑡sink − 𝑡src ; 𝒑 ′ )
R 5𝑧 (𝑡; 𝒑 , 𝒑) = . (7)
𝐽 𝐶 𝑁 (𝑡sink − 𝑡src ; 𝒑 ′ ) 𝐶 𝑁 (𝑡sink − 𝑡; 𝒑 ′ )𝐶 𝑁 (𝑡 − 𝑡src ; 𝒑)𝐶 𝑁 (𝑡sink − 𝑡src ; 𝒑)

Since all 𝑡-dependence due to the contribution of the nucleon ground state can be eliminated in the
ratio (7), the target quantity can be read off from an asymptotic plateau of the ratio R 5𝑧 ′
𝐽 (𝑡; 𝒑 , 𝒑),
being independent of the choice of 𝑡sep , if the condition 𝑡sep /𝑎 ≫ (𝑡 − 𝑡src )/𝑎 ≫ 1 is satisfied.
In this study, we consider only the rest frame of the final state with 𝒑 ′ = 0, which leads to
the condition of 𝒒 = 𝒑 − 𝒑 ′ = 𝒑. Therefore, the squared four-momentum transfer is given by
𝑞 2 = 2𝑀 𝑁 (𝐸 𝑁 (𝒒) − 𝑀 𝑁 ) where 𝑀 𝑁 and 𝐸 𝑁 (𝒒) represent the nucleon mass and energy with the
momentum 𝒒. In this kinematics, we use a simpler notation like R 5𝑧 5𝑧
𝐽 (𝑡; 𝒒) and 𝐶 𝐽 (𝑡; 𝒒).
5𝑧
The ratio R 𝐽 (𝑡; 𝒒) gives the following asymptotic values including the respective form factors
in the asymptotic region [3]:
h i
−1
R 5𝑧
𝐴𝑖 (𝑡; 𝒒) = 𝐾 (𝐸 𝑁 (𝒒) + 𝑀 𝑁 ) e𝐴 (𝑞 2 )𝛿𝑖3 − 𝑞 𝑖 𝑞 3 𝐹
𝐹 e𝑃 (𝑞 2 ) + · · ·, (8)
h i
−1 e
R 5𝑧
𝐴4 (𝑡; 𝒒) = 𝑖𝑞 3 𝐾 𝐹𝐴 (𝑞 2 ) − (𝐸 𝑁 (𝒒) − 𝑀 𝑁 ) 𝐹 e𝑃 (𝑞 2 ) + · · ·, (9)
 
−1 e
R 5𝑧
𝑃 (𝑡; 𝒒) = 𝑖𝑞 3 𝐾 𝐺 𝑃 𝑞
2
+ · · ·, (10)
√︁
with 𝐾 = 2𝐸 𝑁 (𝒒) (𝐸 𝑁 (𝒒) + 𝑀 𝑁 ). The ellipsis denotes excited-state contributions, which are
supposed to be ignored in the case of 𝑡sep /𝑎 ≫ (𝑡 − 𝑡src )/𝑎 ≫ 1. Three target quantities: 𝐹 e𝐴 (𝑞 2 ),
e𝑃 (𝑞 2 ) and 𝐺
𝐹 e𝑃 (𝑞 2 ) 1 can be read off from an asymptotic plateau of the ratio R 5𝑧 (𝑡; 𝒒), being
𝐽
independent of the choice of 𝑡sep . This approach is hereafter referred to as the standard method.

3. Simple subtraction method

In our previous works [5, 6, 9, 10], the 3pt functions involving the 𝐴4 current are not taken
into account for the calculation of the 𝐹𝐴 (𝑞 2 ) and 𝐹𝑃 (𝑞 2 ) form factors. This is simply because,
to the best of our knowledge, the 𝐴4 correlator was found to be statistically very noisy in Ref. [9],
where the time-reversal averaging was performed using both forward and backward propagation in
time for all 3pt functions. However, as pointed out for the first time in Ref. [11], the ratio correlator
R 5𝑧
𝐴4 (𝑡, 𝒒) does not show a plateau, but rather a peculiar behavior that depends linearly on the current
insertion time 𝑡 with a steep negative slope under their kinematic setup. When no time-reversal
averaging is applied in our data, an almost linear 𝑡-dependence is indeed confirmed, giving the same
slope, although the direction is reversed according to the respective kinematics.
1Hereafter, the form factors with and without tilde denote the bare and renormalized ones, then e.g. 𝐹𝐴 = 𝑍 𝐴 𝐹
e𝐴.

3
A proposal for removing 𝜋𝑁-state contamination from the nucleon induced pseudoscalar.... Shoichi Sasaki

Figure 1: Schematic view of the ground-state contribution (A) and two types of the leading 𝜋𝑁 contributions
(B) and (C) for the axial-vector matrix element.

As discussed in Ref. [12], such peculiar time dependence is understood as the leading contri-
bution from the 𝜋𝑁 state in R 5𝑧 𝐴4 (𝑡, 𝒒), arising in the tree diagram of the baryon ChPT. Importantly,
the momentum 𝒒 injected by the axial-vector current is entirely inherited by the pion state, since the
pion in such 𝜋𝑁 state remains in the on-mass shell. The kinematics of the leading 𝜋𝑁 contribution
is therefore restricted to two special cases as depicted in Fig. 1 (B) and (C) [13, 14].
In our previous works [5, 6, 9, 10], the 𝐹𝑃 form factor obtained from R̃ 5𝑧 𝐴𝑖 (𝑡, 𝒒) was indeed
significantly affected by the excited-state contamination, though no such effect was observed for
the 𝐹𝐴 form factor 2. Therefore, we assume that the contributions from the leading 𝜋𝑁 state for
R 5𝑧
𝐴 𝛼 (𝑡, 𝒒) can be described as follows
h i
−1 e
R̃ 5𝑧
𝐴𝑖 (𝑡, 𝒒) ≡ R 5𝑧
𝐴𝑖 (𝑡, 𝒒) − 𝛿 R
𝑖3 𝐴3
5𝑧
(𝑡, 𝒒 0 ) = −𝑞 𝑞
3 𝑖 𝐾 𝐹𝑃 (𝑞 2
) − Δ + (𝑡, 𝑡 sep ; 𝒒) , (11)
h  i
−1
R 5𝑧
𝐴4 (𝑡, 𝒒) = 𝑖𝑞 3 𝐾
e𝐴 (𝑞 2 ) − (𝐸 𝑁 (𝒒) − 𝑀 𝑁 ) 𝐹
𝐹 e𝑃 (𝑞 2 ) + 𝐸 𝜋 (𝒒)Δ− (𝑡, 𝑡sep ; 𝒒) (12)

with 𝒒 0 = (𝑞 1 , 𝑞 2 , 0) satisfying |𝒒 0 | = |𝒒|. The functions Δ± (𝑡, 𝑡sep ; 𝒒) encode the leading 𝜋𝑁
contributions, which provide the residual 𝑡-dependence with a given 𝑡sep .
For the case when the current operator carries the momentum 𝒒, the 𝜋𝑁 contribution can be
expressed by the following form with 𝑡-independent coefficients 𝐵 and 𝐶

Δ± (𝑡, 𝑡sep ; 𝒒) = 𝐵𝑒 −Δ𝐸 (𝒒,−𝒒)𝑡 ± 𝐶𝑒 −Δ𝐸 (0,𝒒) (𝑡sep −𝑡 ) , (13)

where the non-interaction estimates Δ𝐸 (𝒒, 𝒌) = 𝐸 𝜋 ( 𝒌) + 𝐸 𝑁 (𝒒 + 𝒌) − 𝐸 𝑁 (𝒒) may be used.

Therefore, the time derivative of the 𝜋𝑁 contribution Δ± (𝑡, 𝑡sep ; 𝒒) may have the following property:

𝜕4 Δ± (𝑡, 𝑡sep ; 𝒒) = −𝐸 𝜋 (𝒒)Δ∓ (𝑡, 𝑡sep ; 𝒒) + (𝐸 𝑁 (𝒒) − 𝑀 𝑁 )Δ± (𝑡, 𝑡sep ; 𝒒), (14)

which offers us to separate the 𝜋𝑁 contribution Δ± (𝑡, 𝑡sep ; 𝒒) from the 𝐹 e𝑃 form factor using the
5𝑧
time-derivative of the ratio correlator 𝜕4 𝑅 𝐴 𝛼 (𝑡; 𝒒). Hereafter the nucleon energy 𝐸 𝑁 (𝒒) and the
pion energy 𝐸 𝜋 (𝒒) are simply abbreviated by shorthand notations 𝐸 𝑁 and 𝐸 𝜋 , respectively.
The new method for determining 𝐹 e𝑃 (𝑞 2 ), including the time derivative of the 𝐴4 and 𝐴𝑖
correlators, is given by
" #
R̃ 5𝑧
𝐴 (𝑡, 𝒒) 𝐾 𝜕4 R̃ 5𝑧
𝐴 (𝑡, 𝒒) 𝜕4 R 5𝑧
𝐴4 (𝑡, 𝒒)
2
e𝑃 (𝑞 ) = −𝐾
𝐹 𝑖
+ Δ𝐸 𝑁 𝑖
+ (15)
𝑞𝑖 𝑞3 Δ𝐸 2𝑁 − 𝐸 2𝜋 𝑞𝑖 𝑞3 𝑖𝑞 3

2Recent studies solving the generalized eigenvalue problem including 𝜋𝑁 operators also show that the 𝜋𝑁 contribu-
e𝑃 (𝑞 2 ) and 𝐺
tions are strong in 𝐹 e𝑃 (𝑞 2 ), but not in 𝐹
e𝐴 (𝑞 2 ) [15, 16].

4
A proposal for removing 𝜋𝑁-state contamination from the nucleon induced pseudoscalar.... Shoichi Sasaki

√︁
with Δ𝐸 𝑁 ≡ 𝐸 𝑁 − 𝑀 𝑁 and 𝐾 = 2𝐸 𝑁 (𝐸 𝑁 + 𝑀 𝑁 ). The first term corresponds to 𝐹 estd (𝑞 2 )
𝑃
in the standard method. The leading 𝜋𝑁 contributions represented in terms of Δ+ (𝑡, 𝑡sep ; 𝒒) and
Δ− (𝑡, 𝑡sep ; 𝒒) can be completely eliminated by adding the second term in Eq. (15). For the ground
state contribution, namely 𝐹 e𝑃 (𝑞 2 ), Eq. (15) is just a harmless linear combination exploiting the
redundancy in the determination of 𝐹 e𝑃 (𝑞 2 ) from both of 𝐶 5𝑧 (𝑡; 𝒒) and 𝐶 5𝑧 (𝑡; 𝒒). Therefore, if
𝐴4 𝐴𝑖
Eq. (15) successfully shows good plateau behavior, independent of the choice of 𝑡sep , it guarantees
that the ground state contribution can be read accurately without excited-state contamination.
Since the 𝐺 e𝑃 (𝑞 2 ) was also observed to be strongly contaminated from the excited state, similar
e𝑃 (𝑞 2 ), in our previous works [5, 6, 9, 10], we simply assume that
to 𝐹
h i
−1 e
R 5𝑧
𝑃 (𝑡, 𝒒) = 𝑖𝑞 3 𝐾 𝐺 𝑃 (𝑞 2
) − Δ 𝑃 (𝑡, 𝑡 sep ; 𝒒) , (16)

where Δ 𝑃 (𝑡, 𝑡sep ; 𝒒) encodes the leading 𝜋𝑁 state contributions that cause a residual 𝑡-dependence
in R 5𝑧
𝑃 (𝑡; 𝒒). Unlike in the case of the axial-vector currents, only a single correlator cannot remove
the 𝜋𝑁 contribution Δ 𝑃 (𝑡, 𝑡sep ; 𝒒). Instead, it was found that the axial Ward-Takahashi identity is
well satisfied in terms of the 3pt functions of the nucleon in our previous study [6]:

𝑍 𝐴 [𝜕𝛼 𝐶 5𝑧 5𝑧
𝐴 𝛼 (𝑡; 𝒒)] = 2𝑚 PCAC 𝐶 𝑃 (𝑡; 𝒒), (17)

where 𝑚 PCAC corresponds to the bare quark mass which coincides with the value determined from
the pion 2pt functions [6]. Recall that Eq. (17) is satisfied without isolating the ground-state
contribution from the excited-state contributions [6]. Thus, Eq. (17) leads to the following PCAC
relation for the leading 𝜋𝑁 contributions involved in R̃ 5𝑧 5𝑧
𝐴𝑖 (𝑡; 𝒒) and R 𝑃 (𝑡; 𝒒):

𝑀 𝜋2
Δ 𝑃 (𝑡, 𝑡sep ; 𝒒) = 𝑍 𝐴 Δ+ (𝑡, 𝑡sep ; 𝒒), (18)
2𝑚 PCAC

which offers a simple subtraction method for determining the 𝐺 e𝑃 (𝑞 2 ) as below

" #
R 5𝑧
𝑃 (𝑡, 𝒒) 𝑍 𝐴 𝐵 0 𝐾 𝜕4 R̃ 5𝑧
𝐴 (𝑡, 𝒒) 𝜕4 R 5𝑧
𝐴 (𝑡, 𝒒)
2
e𝑃 (𝑞 ) = 𝐾
𝐺 + Δ𝐸 𝑁 𝑖
+ 4
(19)
𝑖𝑞 3 Δ𝐸 2𝑁 − 𝐸 2𝜋 𝑞𝑖 𝑞3 𝑖𝑞 3

𝑀 𝜋2 estd (𝑞 2 ) in the standard method.

with 𝐵0 = 2𝑚PCAC . The first term corresponds to 𝐺 𝑃

4. Numerical results
e𝑃 (𝑞 2 ) and 𝐺
In this study, we reanalyze the data sets generated in Refs. [5, 6] for 𝐹 e𝑃 (𝑞 2 ) using
the new method described in Sec. 3. The two data sets are computed with the first and second

Table 1: Summary of simulation parameters in 2+1 flavor PACS10 ensembles with two different lattice
spacings. See Refs. [6, 17, 18] for further details.

𝛽 𝐿3 × 𝑇 𝜅 𝑢𝑑 𝜅𝑠 𝑐 SW 𝑎 −1 [GeV] 𝑀 𝜋 [GeV] 𝑍 SF
𝐴
1.82 1283 × 128 0.126117 0.124902 1.11 2.3 135 0.9650(68)
2.00 1603 × 160 0.125814 0.124925 1.02 3.1 138 0.9783(21)

5
A proposal for removing 𝜋𝑁-state contamination from the nucleon induced pseudoscalar.... Shoichi Sasaki

PACS10 ensembles, which are two sets of gauge configurations generated in a large volume of about
(10 fm) 4 by the PACS Collaboration with the six stout-smeared O (𝑎) improved Wilson-clover quark
action and Iwasaki gauge action at 𝛽 = 1.82 and 2.00 corresponding to the lattice spacings of 0.09
fm (coarse) and 0.06 fm (fine), respectively [5, 6]. A brief summary of the simulation parameters
is given in Table 1. The simulated pion masses on both lattices are almost at the physical point.

50 70

tsep/a=13 40
tsep/a=13
60
tsep/a=16 Q4 tsep/a=16

Q4
30

tsep/a=19 20 tsep/a=19 50

10 40
-10 -5 0 5 10 -10 -5 0 5 10
80 40 150 60

70 30
140 50
Q1

Q1
Q5

Q5
60 20

130 40
50 10

40 0 120 30
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
60 40 110 60

50 30
100 50
Q2

Q6

Q2

Q6
40 20

90 40
30 10

20 0 80 30
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
50 40 90 50

40 30
80 40
Q3

Q3
Q7

Q7
30 20

70 30
20 10

10 0 60 20
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
(t-tsep/2)/a (t-tsep/2)/a (t-tsep/2)/a (t-tsep/2)/a

Figure 2: The values of 2𝑀 𝑁 𝐹𝑃std (left) and 2𝑀 𝑁 𝐹𝑃 (right) computed using the second PACS10 ensemble
(1604 lattice) with 𝑡sep /𝑎 = 13 (diamonds), 16 (squares) and 19 (circles) for all momentum transfers as
functions of the current insertion time slice 𝑡. In the right panel, the horizontal bands are calculated from the
PPD model (2𝑀 𝑁 𝐹𝑃PPD (𝑞 2 )).

70 110

tsep/a=13 60
tsep/a=13 100

tsep/a=16 tsep/a=16
Q4

Q4

50 90

tsep/a=19 40 tsep/a=19 80

30 70
-10 -5 0 5 10 -10 -5 0 5 10
110 60 230 90

100 50 220 80
Q1

Q1
Q5

Q5

90 40 210 70

80 30 200 60

70 20 190 50
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
90 50 160 80

80 40 150 70
Q2

Q6

Q2

Q6

70 30 140 60

60 20 130 50

50 10 120 40
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
80 50 120 60

70 40 110 50
Q3

Q3
Q7

Q7

60 30 100 40

50 20 90 30

40 10 80 20
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
(t-tsep/2)/a (t-tsep/2)/a (t-tsep/2)/a (t-tsep/2)/a

Figure 3: The values of 𝐺 e𝑃 (right) computed with 𝑡 sep /𝑎 = 13 (diamonds), 16 (squares) and
estd (left) and 𝐺
𝑃
19 (circles) for all momentum transfers as functions of the current insertion time slice 𝑡. In the right panel,
the horizontal bands are calculated from the PPD model (𝐺 ePPD (𝑞 2 )).
𝑃

The 𝐹𝑃 form factor is extracted from Eq. (15) as a function of the current insertion time 𝑡. In
Fig. 2, we compare the 𝑡-dependence and 𝑡sep -dependence of 𝐹𝑃 (𝑞 2 ) obtained by both the standard
(left panel) and the simple subtraction (right panel) methods for the 1604 lattice ensemble. The new
method is really effective in obtaining an asymptotic plateau in all cases of 𝑡sep /𝑎 = {13, 16, 19} for

6
A proposal for removing 𝜋𝑁-state contamination from the nucleon induced pseudoscalar.... Shoichi Sasaki

all 𝑞 2 . Indeed, as shown in the right panel, the 𝑡-dependence is eliminated and the 𝑡sep -dependence
is not visible either. Furthermore, the plateau values are consistent with the PPD model.
As shown in Fig.3, similarly for the 𝐺 𝑃 form factor, the new method eliminates the slight
convex shape associated with the excited-state contribution and yields a plateau behavior consistent
with the PPD model without 𝑡sep dependence as well. In Fig. 4, we plot the 𝑞 2 dependence of
2𝑀 𝑁 𝐹𝑃 (𝑞 2 ) (left panel) and 2𝑚 PCAC 𝐺e𝑃 (𝑞 2 ) (right panel) for all data sets of 1604 and 1284 lattices
together with experimental data points from muon capture [19] and pion-electro production [20].

1.6
160 ~
2MNFPPPD(q), fine 2mPCACGPPPD(q), fine
1.4
Expt. (muon capture) tsep/a=16, coarse
140
Expt. (pion-electroproduction) tsep/a=14, coarse
tsep/a=16, coarse 1.2

2mPCACGP (q2) [GeV]

120 tsep/a=19, fine
tsep/a=14, coarse
tsep/a=16, fine
tsep/a=19, fine 1
100 tsep/a=13, fine
2MNFP(Q )
2

tsep/a=16, fine
tsep/a=13, fine 0.8
80

60 ~ 0.6

40 0.4

20 0.2

0 0
0 0.02 0.04 0.06 0.08 0.1 0.12 0 0.02 0.04 0.06 0.08 0.1 0.12
2 2 2 2
q [GeV ] q [GeV ]

Figure 4: Results of 2𝑀 𝑁 𝐹𝑃 (𝑞 2 ) (left panel) and 2𝑚 PCAC 𝐺

e𝑃 (𝑞 2 ) (right panel) obtained by the new method
2
as a function of 𝑞 . In each panel, the solid curve is given by the PPD model defined in Eq. (4).

Next, 𝑔 ∗𝑃 and 𝑔 𝜋 𝑁 𝑁 are evaluated from the obtained 𝐹𝑃 form factor according to the 𝑞 2
dependence analysis based on the z-expansion method applied to (𝑞 2 + 𝑚 2𝜋 )𝐹𝑃 (𝑞 2 ). As shown in
Fig 5, no 𝑡sep dependences are visible for either case and the discretization error on these quantities
is less than 3-4 %, which is well controlled in our calculations as well as 𝑔 𝐴. More importantly, the
evaluation is possible with much smaller errors than the experimental values for 𝑔 ∗𝑃 and comparable
errors for 𝑔 𝜋 𝑁 𝑁 . Summary plots of our results together with the experimental values and other
lattice QCD results for 𝑔 ∗𝑃 and 𝑔 𝜋 𝑁 𝑁 can be found in Ref. [21].

5. Summary

We have studied nucleon form factors in the axial-vector and pseudo-scalar channels in 2+1
flavor QCD using two sets of the PACS10 configurations at coarse and fine lattice spacings. Our
simulations were carried out in very large spatial volumes, which allow us to access the low 𝑞 2 region,
at the physical point essential for low-energy chiral behavior. The nucleon interpolating operator
has been adopted with well-tuned smearing parameters that guarantee ground-state dominance in
𝐹𝐴 (𝑞 2 ), although the two types of pseudo-scalar form factors, 𝐹𝑃 (𝑞 2 ) and 𝐺 𝑃 (𝑞 2 ), still suffer from
the excited-state contamination. In this study, we thus propose a simple subtraction method for
removing the so-called leading 𝜋𝑁-state contamination induced by the pion-pole structure appears
in 𝐹𝑃 (𝑞 2 ) and 𝐺 𝑃 (𝑞 2 ). The new method achieves the following points: 1) it can use the 3pt-
functions of both spatial and temporal axial-vector currents to determine 𝐹𝑃 (𝑞 2 ), 2) it is applicable
for 𝐺 𝑃 (𝑞 2 ) with a help of the PCAC relation, 3) it eliminates both 𝑡-dependence and 𝑡sep -dependence

7
A proposal for removing 𝜋𝑁-state contamination from the nucleon induced pseudoscalar.... Shoichi Sasaki

11 17
4 4
128 (Coarse lattice) 128 (Coarse lattice)
10.5 4
160 (Fine lattice) 16
4
160 (Fine lattice)
MuCap Expt. (2017): 8.06(48) Isospin-Averaged Expt. 13.36(8)
10

gπNN (renormalized)
g*P (renormalized)

15
9.5
14
9
13
8.5
12
8

7.5 11

0.6 0.8 1 1.2 1.4 1.6 0.6 0.8 1 1.2 1.4 1.6
tsep (fm) tsep (fm)

Figure 5: The source-sink separation (𝑡 sep ) dependence of the renormalized values of 𝑔 ∗𝑃 (left) and 𝑔 𝜋 𝑁 𝑁
(right). In each panel, the horizontal axis gives 𝑡sep in physical units, while the horizontal dashed line with
the gray band denotes the experimental value.

in both 𝐹𝑃 (𝑞 2 ) and 𝐺 𝑃 (𝑞 2 ), 4) it makes results compatible with both the experiment and the PPD
model, and 5) it provides more accurate results of two target quantities, 𝑔 ∗𝑃 and 𝑔 𝜋 𝑁 𝑁 , comparing
the multi-state analysis that was used in other groups [14, 22–24].

Acknowledgments

We would like to thank members of the PACS collaboration for useful discussions. K.-I. I. is
supported in part by MEXT as “Feasibility studies for the next-generation computing infrastruc-
ture". Numerical calculations in this work were performed on Oakforest-PACS in Joint Center
for Advanced High Performance Computing (JCAHPC) and Cygnus and Pegasus in Center for
Computational Sciences at University of Tsukuba under Multidisciplinary Cooperative Research
Program of Center for Computational Sciences, University of Tsukuba, and Wisteria/BDEC-01 in
the Information Technology Center, the University of Tokyo. This research also used computational
resources of the K computer (Project ID: hp1810126) and the Supercomputer Fugaku (Project ID:
hp20018, hp210088, hp230007, hp2320199, hp240207) provided by RIKEN Center for Computa-
tional Science (R-CCS), as well as Oakforest-PACS (Project ID: hp170022, hp180051, hp180072,
hp190025, hp190081, hp200062), Wisteria/BDEC-01 Odyssey (Project ID: hp220050) provided by
the Information Technology Center of the University of Tokyo / JCAHPC. The calculation employed
OpenQCD system(http://luscher.web.cern.ch/luscher/openQCD/). This work is supported by the
JLDG constructed over the SINET5 of NII, This work was also supported in part by Grants-in-Aid
for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology
(Nos. 18K03605, 19H01892, 22K03612, 23H01195, 23K03428, 23K25891) and MEXT as “Pro-
gram for Promoting Researches on the Supercomputer Fugaku” (Search for physics beyond the
standard model using large-scale lattice QCD simulation and development of AI technology toward
next-generation lattice QCD; Grant Number JPMXP1020230409).

8
A proposal for removing 𝜋𝑁-state contamination from the nucleon induced pseudoscalar.... Shoichi Sasaki

References

[1] M.L. Goldberger and S.B. Treiman, Form-factors in Beta decay and muon capture, Phys.
Rev. 111 (1958) 354.

[2] W.I. Weisberger, Unsubtracted Dispersion Relations and the Renormalization of the Weak
Axial Vector Coupling Constants, Phys. Rev. 143 (1966) 1302.

[3] S. Sasaki and T. Yamazaki, Nucleon form factors from quenched lattice QCD with domain
wall fermions, Phys. Rev. D 78 (2008) 014510 [0709.3150].

[4] Y. Nambu, Axial vector current conservation in weak interactions, Phys. Rev. Lett. 4 (1960)
380.

[5] E. Shintani, K.-I. Ishikawa, Y. Kuramashi, S. Sasaki and T. Yamazaki, Nucleon form factors
and root-mean-square radii on a (10.8 fm)4 lattice at the physical point, Phys. Rev. D 99
(2019) 014510 [1811.07292].

[6] R. Tsuji, Y. Aoki, K.-I. Ishikawa, Y. Kuramashi, S. Sasaki, K. Sato et al., Nucleon form
factors in 𝑁 𝑓 =2+1 lattice QCD at the physical point: Finite lattice spacing effect on the
root-mean-square radii, Phys. Rev. D 109 (2024) 094505 [2311.10345].

[7] LHPC, SESAM collaboration, Moments of nucleon generalized parton distributions in

lattice QCD, Phys. Rev. D 68 (2003) 034505 [hep-lat/0304018].

[8] QCDSF collaboration, Nucleon electromagnetic form-factors on the lattice and in chiral
effective field theory, Phys. Rev. D 71 (2005) 034508 [hep-lat/0303019].

[9] PACS collaboration, Nucleon form factors on a large volume lattice near the physical point
in 2+1 flavor QCD, Phys. Rev. D 98 (2018) 074510 [1807.03974].

[10] PACS collaboration, Calculation of the derivative of nucleon form factors in Nf=2+1 lattice
QCD at M𝜋=138 MeV on a (5.5 fm)3 volume, Phys. Rev. D 104 (2021) 074514
[2107.07085].

[11] G.S. Bali, S. Collins, M. Gruber, A. Schäfer, P. Wein and T. Wurm, Solving the PCAC puzzle
for nucleon axial and pseudoscalar form factors, Phys. Lett. B 789 (2019) 666
[1810.05569].

[12] O. Bar, 𝑁 𝜋-state contamination in lattice calculations of the nucleon axial form factors,
Phys. Rev. D 99 (2019) 054506 [1812.09191].

[13] H.B. Meyer, K. Ottnad and T. Schulz, A new method for suppressing excited-state
contaminations on the nucleon form factors, PoS LATTICE2018 (2018) 062 [1811.03360].

[14] RQCD collaboration, Nucleon axial structure from lattice QCD, JHEP 05 (2020) 126
[1911.13150].

9
A proposal for removing 𝜋𝑁-state contamination from the nucleon induced pseudoscalar.... Shoichi Sasaki

[15] L. Barca, G. Bali and S. Collins, Toward N to N𝜋 matrix elements from lattice QCD, Phys.
Rev. D 107 (2023) L051505 [2211.12278].

[16] C. Alexandrou, G. Koutsou, Y. Li, M. Petschlies and F. Pittler, Investigation of pion-nucleon

contributions to nucleon matrix elements, Phys. Rev. D 110 (2024) 094514 [2408.03893].

[17] PACS collaboration, Finite size effect on pseudoscalar meson sector in 2+1 flavor QCD at
the physical point, Phys. Rev. D 99 (2019) 014504 [1807.06237].

[18] PACS collaboration, Finite size effect on vector meson and baryon sectors in 2+1 flavor
QCD at the physical point, Phys. Rev. D 100 (2019) 094502 [1907.10846].

[19] T. Gorringe and H.W. Fearing, Induced Pseudoscalar Coupling of the Proton Weak
Interaction, Rev. Mod. Phys. 76 (2004) 31 [nucl-th/0206039].

[20] S. Choi et al., Axial and pseudoscalar nucleon form-factors from low-energy pion
electroproduction, Phys. Rev. Lett. 71 (1993) 3927.

[21] R. Tsuji, Y. Aoki, K.-I. Ishikawa, Y. Kuramashi, S. Sasaki, K. Sato et al., Studies of nucleon
isovector structure with the PACS10 superfine lattice, in 41st International Symposium on
Lattice Field Theory, 11, 2024 [2411.16784].

[22] Nucleon Matrix Elements (NME) collaboration, Precision nucleon charges and form
factors using (2+1)-flavor lattice QCD, Phys. Rev. D 105 (2022) 054505 [2103.05599].

[23] Precision Neutron Decay Matrix Elements (PNDME) collaboration, Nucleon isovector
axial form factors, Phys. Rev. D 109 (2024) 014503 [2305.11330].

[24] Extended Twisted Mass collaboration, Nucleon axial and pseudoscalar form factors using
twisted-mass fermion ensembles at the physical point, Phys. Rev. D 109 (2024) 034503
[2309.05774].

10