Three-body scattering hypervolume of two-component fermions in three dimensions

Jiansen Zhang,1, ∗ Zipeng Wang,1 and Shina Tan1, †

1
International Center for Quantum Materials, Peking University, Beijing 100871, China
(Dated: January 10, 2025)
We study the zero-energy collision of three fermions, two of which are in the spin-down (↓)
state and one of which is in the spin-up (↑) state. Assuming that the two-body and the three-
body interactions have a finite range, we find a parameter, D, called the three-body scattering
hypervolume. We study the three-body wave function asymptotically when three fermions are
far apart or one spin-↑ (spin-↓) fermion and one pair, formed by the other two fermions, are far
apart, and derive three asymptotic expansions of the wave function. The three-body scattering
arXiv:2501.05194v1 [cond-mat.quant-gas] 9 Jan 2025

hypervolume appears in the coefficients of such expansions at the order of B −5 , where B is the
hyperradius of the triangle formed by the three fermions. When the interactions are weak, we
calculate D approximately using the Born expansion. We also analyze the energy shift of three such
fermions in a large periodic cube due to D and generalize this result to the many-fermion system. D
also determines the three-body recombination rate in two-component Fermi gases, and we calculate
the three-body recombination rate in terms of D and the density and temperature of the gas.

I. INTRODUCTION scattering wave function out of the range of interaction.

For example, the wave function for the two-body zero-
Both theoretically and experimentally, the research of energy s-wave scattering is
the two-component Fermi system is at the center of many a0
research fields for a long time, such as nuclear physics and ϕ(s) = 1 − (1)
s
quantum field theory [1–4], and ultracold atomic physics
[5–8], etc.. Especially, taking advantage of the technolo- for s > re , where re is the range of interaction. Similarly,
gies for cooling atoms and molecules [9–13] and tuning we can define the three-body scattering hypervolume by
the interactions by Feshbach resonances [14–17], people expanding the three-body wave function asymptotically.
can realize the quantum degeneracy, BCS-BEC crossover In this paper, we study the wave function for three two-
[18–22], Fermi-Hubbard model, etc. in trapped two- component fermions, two of which are in the spin-↓ state
component Fermi gases. The experimental progresses in and one of which is in the spin-↑ state, with zero collision
ultracold Fermi gases promoted a large amount of theo- energy and with total orbital angular momentum L = 1.
retical works [23–27]. We find three expansions for the three-body wave func-
In the dilute ultracold Fermi gases, neutral atoms col- tion: one 111-expansion and two 21-expansions. The 111-
lide at very low energies, and so their de Brogile wave- expansion characterizes the asymptotic behavior when
lengths are much larger than the range of the interac- three fermions are far away from each other, the first
tion. Under this condition, the interaction can be ap- kind 21-expansion characterizes the asymptotic behavior
proximately characterized by a single parameter, i.e., the when two spin-↓ fermions are kept at a fixed distance
s-wave scattering length a0 . (For identical fermions, the and the spin-↑ fermion is far away from the two, and
s-wave scattering is forbidden due to the Fermi statis- the second kind 21-expansion characterizes the asymp-
tics and the dominant effective parameter is the p-wave totic behavior when one spin-↑ fermion and one spin-↓
scattering volume ã1 .) Apart from the two-body scat- fermion are kept at a fixed distance and another spin-↓
terings, three-body scatterings are also common. Simi- fermion is far away from the two. The three-body scat-
larly, one can define a parameter that characterizes the tering hypervolume D appears at the order of B −5 in the
low-energy three-body scattering processes. Such a pa- 111-expansion. The dimension of D is length to the sixth
rameter, which is usually called the three-body scatter- power.
ing hypervolume and denoted by D, was first defined In Sec. II, we define the two-body special functions and
for the three-body scattering of identical bosons in three derive the 111-expansion and the 21-expansions. When
dimensions (3D) [28] and has been generalized to distin- the two-body interactions support bound states, we show
guishable particles in 3D [29], to identical spin-polarized that D acquires a negative imaginary part related to the
fermions in 3D [30], in two dimensions (2D) [31] and in probability amplitudes of the three-body recombination
one dimension (1D) [32] and to identical bosons in 2D processes. In Sec. III, we study the three-body wave func-
[33]. tion using Born approximation when the interaction po-
One way to define the two-body scattering parameter tentials are weak. We derive an approximate formula for
is to determine the asymptotic behavior of the two-body D in terms of the weak potentials. In Sec. IV, we con-
sider three two-component fermions in a large periodic
cubic box and calculate the shift of the energy due to a
∗ Jiansenzhang prc@stu.pku.edu.cn nonzero D. We then generalize the result to the N -body
† shinatan@pku.edu.cn system, where N = N↑ + N↓ , and Nσ is the number of
 2

spin-σ fermions for σ =↑, ↓. In Sec. V, we suppose that s2 , and s3 approach infinity simultaneously. We assume
the interaction potentials can support two-body bound that Ψ0 scales as B p at large B. Ψ0 should satisfy the
states and calculate the three-body recombination rates Laplace equation (∇21 + ∇22 + ∇23 )Ψ0 = 0. For dilute
of two-component Fermi gases in terms of the three-body ultracold gases, the most important scattering channel
scattering hypervolumes. requires the minimum value of p, because the smaller p
is, the more likely for the three particles to come into the
range of interaction. We find that the minimum value of p
II. ASYMPTOTIC BEHAVIOR OF THE for this system is pmin = 1, and the Ψ0 for this minimum
THREE-BODY WAVE FUNCTION value of p can take three linearly independent forms: sx ,
sy , and sz . We will first focus on the three-body wave
We consider three two-component fermions with the function with
same mass mF , and two of them are in the spin-↓ state
and labeled as particles 1 and 2, one of them is in the Ψ0 = sz , (6)
spin-↑ state, and labeled as particle 3. Let ri be the
position vector of the ith particle. We also define the This leading order term has total orbital angular mo-
Jacobi coordinates si and Ri , the hyperradius B, and mentum quantum number L = 1 and magnetic quantum
the hyperangles Θi [34, 35]: number M = 0. There are two other linearly independent
1
leading order terms that also √ scale as B and√they can
si ≡ rj − rk , (2a) be written as −(sx + isy )/ 2 and (sx − isy )/ 2. These
three leading order wave functions can be distinguished
Ri ≡ ri − (rj + rk )/2, (2b)
r by the magnetic quantum number M = 0, ±1 along the z
q
3 direction. In the following, we will study the wave func-
B ≡ (s1 + s2 + s3 )/2 = Ri2 + s2i ,
2 2 2 (2c) tion for the state |L = 1, M = 0⟩, and the other two wave
4
2Ri functions can be generated by applying the ladder opera-
Θi ≡ arctan √ , (2d) tor J± on the state |L = 1, M = 0⟩. Here J± = Jx ± iJy ,
3si
and Jx and Jy are the projections of the total orbital an-
where (i, j, k) is any even permutation of (1, 2, 3). One gular momentum operator in the x and the y directions,
can show that si = √23 B cos Θi , and Ri = B sin Θi . For respectively. We have
later convenience, we define short-hand notations 
∂ ∂
Jx = iℏ sin ϕsi + cot θsi cos ϕsi
s ≡ s3 , (3a) ∂θsi ∂ϕsi

R ≡ R3 , (3b) ∂ ∂
+ sin ϕRi + cot θRi cos ϕRi , (7)
∂θRi ∂ϕRi
Θ ≡ Θ3 . (3c) 
∂ ∂
Jy = iℏ − cos ϕsi + cot θsi sin ϕsi
In the center-of-mass (COM) reference frame, ∂θsi ∂ϕsi
the three-body zero-energy scattering wave function ∂ ∂

Ψ(r1 , r2 , r3 ) satisfies the Schrödinger equation − cos ϕRi + cot θRi sin ϕRi , (8)
∂θRi ∂ϕRi
3 
ℏ2

X
2 where θsi , ϕsi , θRi , and ϕRi are spherical coordinates
− ∇ + Vi (si ) Ψ + U (s1 , s2 , s3 )Ψ = 0, (4)
i=1
2mF i such that

where ℏ is the reduced Plank constant, Vi (si ) is the two- six = si sin θsi cos ϕsi , (9a)
body interaction potential between the jth and the kth siy = si sin θsi sin ϕsi , (9b)
particles and V1 = V2 , and U (s1 , s2 , s3 ) is the three-body siz = si cos θsi , (9c)
interaction potential. We assume that all the interaction Rix = Ri sin θRi cos ϕRi , (9d)
potentials are invariant under overall translations, rota-
tions, and finite-ranged [when si > re , Vi (si ) = 0; when Riy = Ri sin θRi sin ϕRi , (9e)
s1 > re or s2 > re or s3 > re , U (s1 , s2 , s3 ) = 0]. Since we Riz = Ri cos θRi . (9f)
study the problem in the COM frame, the wave function
is translationally invariant, i.e., One can also′
study those Ψ’s with leading order Ψ0 ’s
scaling as B p , where p′ > pmin . For example, if we set
Ψ(r1 , r2 , r3 ) = Ψ(r1 + δr, r2 + δr, r3 + δr) (5) p′ = 2, then there are nine linearly independent leading
order terms, one of which has L = 0 and M = 0, i.e.,
for any displacement δr. s · R, three of which have L = 1 and M √ = 0, ±1, i.e.,
To uniquely determine the wave function Ψ, we also (s × R)z and ∓[(s × R)x ± i(s × R)y ]/ 2, and five of
need to specify the asymptotic behavior of Ψ when all which have L = 2 and M = 0, ±1, ±2, i.e., (sx +isy )(Rx +
three particles are far apart from each other. Let Ψ0 be iRy ), (sx + isy )Rz + sz (Rx + iRy ), −s · R + 3sz Rz , (sx −
the leading order term in the expansion of Ψ when s1 , isy )Rz + sz (Rx − iRy ), and (sx − isy )(Rx − iRy ). These
 3

wave functions are usually less important than the wave Solving Eqs. (16), we find that if s1 > re we have
functions with p = 1 for dilute ultracold gases, and will
not be studied in this paper. r
sl1

(l,m) (2l − 1)!!al 4π
ϕ (s1 ) = − Y m (ŝ1 ),
(2l + 1)!! sl+1
1
2l + 1 l
A. Two-body special functions
(17a)
" #
l+2 l
Before we study the three-body wave function, we in- s (2l − 3)!!al a l rl s1
f (l,m) (s1 ) = − 1
+ +
troduce the two-body special functions [28] ϕ(l,m) (s1 ), 2(2l + 3)!! 2sl−1
1
2(2l + 1)!!
f (l,m) (s1 ), g (l,m) (s1 ), . . . , and ϕ̃(l,m) (s), f˜(l,m) (s), r
g̃ (l,m) (s), . . . that will be used in the 21-expansions. 4π
× Y m (ŝ1 ), (17b)
Consider the scattering between one spin-↑ fermion 2l + 1 l
"
and one spin-↓ fermion in the COM reference frame with sl+4 (2l − 5)!!al al rl sl+2
collision energy E = ℏ2 k 2 /(2µ), orbital angular momen- g (l,m) (s1 ) = 1
− + 1
8(2l + 5)!! 8sl−3
1
4(2l + 3)!!
tum quantum number l and magnetic quantum number r
al rl′ sl1

m along the z direction, where µ = mF /2 is the reduced 4π
mass. The wave function ψ (l,m) (s1 ) can be separated as − Y m (ŝ1 ), (17c)
24(2l + 1)!! 2l + 1 l
r
(l,m) 4π
ψ (s1 ) = u(s1 ) Y m (ŝ1 ), (10)
2l + 1 l where we have chosen a particular overall amplitude for
where Ylm (ŝ1 ) is the spherical harmonics [36]. u(s1 ) sat- ϕ(l,m) (s1 ) in accordance with Ref. [28], and fixed the def-
isfies the radial Schrödinger equation inition of f (l,m) (s1 ) by requiring that f (l,m) (s1 ) does not
contain the term ∝ s−l−1 1 at s1 > re (if it does, one can
d2 u
 
2 du 2 2µV1 (s1 ) l(l + 1) redefine it by adding a suitable coefficient times ϕ(l,m) (s1 )
+ + k − − u = 0. (11)
ds21 s1 ds1 ℏ2 s21 to the definition of f (l,m) (s1 ) to remove such a term),
Since the interaction potential V1 (s1 ) is finite-ranged, we and similarly fixed the definition of g (l,m) (s1 ) by requir-
can find the analytical solution for u(s1 ) outside of the ing that it does not contain the term ∝ s−l−1 1 at s1 > re .
range of interaction: The parameter al is the l-wave scattering “length” al-
though its dimension is length raised to the power of
u(s1 ) = α(l, k) [jl (ks1 ) cot δl (k) − yl (ks1 )] , s1 > re , (2l + 1). rl is the l-wave effective range, and rl′ is the
(12) l-wave shape parameter. The two-body special functions
where α(l, k) is an arbitrary coefficient that determines at s < re and the parameters al , rl , rl′ depend on the
the overall amplitude of the wave function, jl and yl are details of the interaction potential V1 (s1 ). Substituting
the spherical Bessel functions of the first and the second Eqs. (17) into Eq. (15) and comparing the result with
kind respectively, and δl (k) is the l-wave scattering phase Eq. (10) and Eq. (12), one can deduce that δl (k) satisfies
shift. the effective range expansion [37–40]
In the COM frame, the Schrödinger equation for the
collision of particles 2 and 3 with energy ℏ2 k 2 /(2µ) and
specified orbital angular momentum quantum numbers k 2l+1 cot δl (k) = −a−1 2 ′ 4 6
l + rl k /2! + rl k /4! + O(k ) (18)
may be written as
e 1 ψ (l,m) (s1 ) = k 2 ψ (l,m) (s1 ),
H (13) and α(l, k) = −al k l+1 .
where Similarly, one can define the two-body special func-
2µ tions ϕ̃(l,m) (s), f˜(l,m) (s), and g̃ (l,m) (s) for the collision
He1 ≡ −∇2s1
+ 2 V1 (s1 ) (14)
ℏ of particles 1 and 2 which are identical spin-↓ fermions.
is proportional to the two-body Hamiltonian in the COM They satisfy
frame. For low energy scatterings, we can expand the
wave function in powers of k 2 [28]: e 3 ϕ̃(l,m) (s) = 0,
H (19a)
(l,m) (l,m) 2 (l,m) 4 (l,m)
ψ (s1 ) = ϕ (s1 ) + k f (s1 ) + k g
(s1 ) + . . . , e 3 f˜(l,m) (s) = ϕ̃(l,m) (s),
H (19b)
(15)
and the two-body special functions ϕ(l,m) (s1 ), f (l,m) (s1 ), H
e 3 g̃ (l,m) ˜(l,m)
(s) = f (s), . . . , (19c)
and g (l,m) (s1 ) are independent of k and satisfy [28]
e 1 ϕ(l,m) (s1 ) = 0,
H (16a) where
e 1 f (l,m) (s1 ) = ϕ(l,m) (s1 ),
H (16b)
e 1 g (l,m) (s1 ) = f (l,m) (s1 ), . . . .
H (16c) e 3 = −∇2s + 2µ V3 (s).
H (20)
ℏ2
 4

(−q)
At s > re we have the following explicit formulas: where S1 scales as R1−q , and this expansion is called
r the second kind 21-expansion. The leading order term in
sl

(l,m) (2l − 1)!!ãl 4π (1)
this expansion, S1 (R1 , s1 ), scales as R11 because Ψ0 =
ϕ̃ (s) = − Y m (ŝ),
(2l + 1)!! sl+1 2l + 1 l sz = −s1z /2 + R1z , which contains a term ∝ R11 and a
(21a) term ∝ R10 .
l+2 l T (−p) satisfies
 
˜ (l,m) s (2l − 3)!!ãl ãl r̃l s
f (s) = − + +
2(2l + 3)!! 2sl−1 2(2l + 1)!!
− ∇21 + ∇22 + ∇23 T (−p) = 0


r (25)
4π
× Y m (ŝ), (21b)
2l + 1 l
when the pairwise distances s1 , s2 , and s are all nonzero.
sl+4 ãl r̃l sl+2

(2l − 5)!!ãl When particle 3 is far apart form particles 1 and 2, the
g̃ (l,m) (s) = − +
8(2l + 5)!! 8sl−3 4(2l + 3)!! Schrödinger equation (4) is simplified as
r
ãl r̃l′ sl

4π
− Y m (ŝ), (21c) 
ℏ2 2 3ℏ2 2

24(2l + 1)!! 2l + 1 l − ∇ − ∇ + V3 (s) Ψ = 0. (26)
mF s 4mF R
where ãl is the l-wave scattering “length” between two
spin-↓ fermions, r̃l is the l-wave effective range between (−q)
Thus S3 satisfies
two spin-↓ fermions, and r̃l′ is the l-wave shape parameter
between two spin-↓ fermions.
Note that in the two-body special functions ϕ(l,m) (s1 ), e 3 S (0) = 0,
H (27a)
3
f (l,m) (s1 ), and g (l,m) (s1 ), l can be any nonnegative in- e 3 S (−1)
H 3 = 0, (27b)
teger. But in the functions ϕ̃(l,m) (s), f˜(l,m) (s), and
3
g̃ (l,m) (s), l must be a positive odd integer because of e 3 S (−q)
H 3
(−q+2)
= ∇2R S3 , q ≥ 2. (27c)
Fermi statistics. 4
(−q)
Similarly, S1 satisfies:
B. The 111-expansion and the 21-expansions
e 1 S (1) = 0,
H (28a)
1
When three particles are far apart from each other,
He 1 S (0) = 0, (28b)
which means that the pairwise inter-particle distances 1
s1 , s2 , s3 all go to infinity for any fixed ratio s1 : s2 : s3 , (−q) 3 (−q+2)
e1S
H = ∇2R1 S1 , q ≥ 1. (28c)
one can expand Ψ in powers of B −1 : 1
4
∞
Furthermore, one can expand T (−p) as T (−p) =
X
Ψ= T (−p) (r1 , r2 , r3 ), (22) P (i,−p−i)
p=−1 i t3 when R ≫ s > re and expand T (−p) as
(i,−p−i) (i,j)
T (−p) =
P
i t1 when R1 ≫ s1 > re , where t3
where T (−p) scales as B −p . Equation (22) is called the (i,j)
scales as Ri sj , and t1 scales as R1i sj1 . Also, one can ex-
111-expansion. The leading order term in this expan- (−q) (−q) P (−q,j)
sion, T (1) (r1 , r2 , r3 ), is equal to the Ψ0 defined in Eq. (6). pand S3 as S3 = j t3 when R ≫ s > re and
(−q) (−q) P (−q,j)
When particle 1 and particle 2 are kept at a fixed dis- expand S1 as S1 = j t1 when R1 ≫ s1 > re .
tance s and particle 3 is far away from them, one can
expand Ψ in powers of R−1 : The points at which t3
(i,j)
̸= 0 are shown on the (i, j)
∞
plane in Fig. 1(a). The straight line with slope −1 and
X (−q) with vertical intercept −p represents the expansion of
Ψ= S3 (R, s), (23) (−p)
q=0 T3 and the vertical line i = −q represents the ex-
(−q) (i,j)
pansion of S3 . Similarly, the nonzero t1 terms are
(−q)
where S3 scales as R−q , and this expansion is called shown in Fig. 1(b), in which the vertical line i = −q
(−q)
the first kind 21-expansion. The leading order term in represents the expansion of S1 .
(0)
this expansion, S3 (R, s), scales as R0 because Ψ0 is We derive the 111-expansion and the 21-expansions in
independent of R. When particle 2 and particle 3 are the following order: first determine T (1) , and then deter-
kept at a fixed distance s1 and particle 1 is far away (0) (1)
mine S3 and S1 , and then determine T (0) , and then
from them, one can expand Ψ in powers of R1−1 : (0)
determine S (−1) and S1 , and so on. More details are
∞ shown in Appendix A.
(−q)
X
Ψ= S1 (R1 , s1 ), (24) The resultant 111-expansion is
q=−1
 5

(−𝟓) (−4) (−3) (−2) (−1) 𝒋 (0) 𝒋

𝓢3 𝓢3 𝓢3 𝓢3 𝓢3 𝓢3 (−𝟓) (−𝟒) (−𝟑) (−𝟐) (−𝟏) (𝟎) (1)
𝓢𝟏 𝓢𝟏 𝓢𝟏 𝓢𝟏 𝓢𝟏 𝓢𝟏 𝓢𝟏
𝟓 𝟓

𝟒 𝟒

𝟑 𝟑

𝟐 𝟐

𝓣(𝟏) 𝒯 (1)
𝟏
𝟏
𝒯 (0)
𝓣(𝟎)
−𝟓 −𝟒 −𝟑 −𝟐 −𝟏 𝟎 𝟏 𝒊
−𝟓 −𝟒 −𝟑 −𝟐 −𝟏 𝟎 𝒊
(−𝟏)
𝓣
𝓣(−𝟏) −𝟏
−𝟏
𝓣(−𝟐)
𝓣(−𝟐) −𝟐
−𝟐
𝓣(−𝟓) 𝓣(−𝟒) 𝓣(−𝟑)
(−𝟓) (−𝟒) (−𝟑) −𝟑
𝓣 𝓣 𝓣
(i,j) (i,j)
(a) Diagram for t3 (b) Diagram for t1

(i,j) (i,j) (i,j)

FIG. 1. Diagram of the points representing t3 and t1 on the (i, j) plane. Thick dots represent the points at which t3 ̸= 0
(i,j)
or t1 ̸= 0.

Ψ = T (1) + T (0) + T (−1) + T (−2) + T (−3) + T (−4) + T (−5) + O(B −6 lnn B), (29)

where n is a nonnegative integer. We have

T (1) = sz , (30)
a0 R1z a0 R2z
T (0) =− + , (31)
s1 s2
a20 R1z
 
R2z
T (−1) =− 2
(cot Θ1 − Θ1 (1 + cot Θ1 )) − 2
(cot Θ2 − Θ2 (1 + cot Θ2 )) , (32)
π s1 R1 s2 R2
2ωa30 R1z
   
3ã1 sz 3a1 s1z s2z R2z
T (−2) =− + + 3 + (1 − Θ1 cot Θ1 ) − (1 − Θ2 cot Θ2 ) , (33)
s3 2 s31 s2 π s1 R12 s2 R22

3
X β−3,i 
T (−3) = −Riz Ri · si (sin 2Θi − 3 cot Θi − Θi + 3Θi cot2 Θi ) + Ri2 siz (3Θi + Θi cot2 Θi − cot Θi + 2 sin 2Θi )

3 3
Ri si
i=1
    
2a0 ω3 R1z 1 R2z 1
+ Θ 1 − sin 2Θ1 − Θ 2 − sin 2Θ2 , (34)
π R13 s1 2 R23 s2 2

3
2ω ′ √
   
(−4)
X β−4,i 2 2 4 R1z R2z
T = siz sin Θi (−12ω cos Θi + 9 3 − 6π) + 2 Riz Ri · si sin Θi − a0 ω4 − 4 , (35)
i=1
Ri2 s3i Ri B 4 s1 B s2
 6
√ 2  
(−5) 3 3D 2a0 Ω5 X i−1 Riz 5 b
T = − 3 6 sz + (−1) 24 sin Θi cos Θi ln + η−4 (Θi )
8π B 9π i=1 Ri5 si Ri
2
" 2 1
#
X 9a 0 a 2 2 X 1,0
√ X 1,0
+ (−1)i−1 2 3 η−2 (Θi ) C3,m;2,−m Y3m (R̂i )Y2−m (ŝi ) + 2 6η̃−2 (Θi ) C1,m;2,−m Y1m (R̂i )Y2−m (ŝi )
i=1
R i si 3 m=−2 m=−1
3
(   )
X β−5,i 3 3 3 b 1 3 3Riz
+ siz 256ωa0 sin Θi cos Θi ln + ℓi (Θi − sin 4Θi ) + 16ωa0 η−3 (Θi ) − (Ri · si )ℓi η̃−3 (Θi ) ,
i=1
Ri3 s3i Ri 4 Ri2
(36)

where D is the three-body scattering hypervolume, which appears at the order of B −5 and has the dimension of length
to the sixth power, ClJ,M
1 ,m1 ;l2 ,m2
= ⟨l1 , m1 ; l2 , m2 |J, M ⟩ is the Clebsch-Gordan coefficient, and

√
ω = 3 − π/3, (37)
√
ω ′ = − 15 3 + 8π, (38)
√
ω3 = 3ã1 + 3a1 /2 + (7 3/6 − − 4/π)ωa30 3a20 r0 /8, (39)
h  √  i
ω4 = 3ωa30 r0 /2 − 4ωω3 a0 + 15 3 − 2π (a0 ã1 + a0 a1 /2) /π, (40)
a2  √ 
Ω5 = 0 −15 3 + 2π (ã1 + a1 /2) − a0 ω4 /2, (41)
2π
β−3,1 = β−3,2 = −3a0 a1 /2π, (42)
β−3,3 = 3a0 ã1 /π, (43)
β−4,1 = β−4,2 = −a20 a1 /2π, (44)
β−4,3 = a20 ã1 /π, (45)
2
β−5,1 = β−5,2 = −a1 /8π , (46)
2
β−5,3 = ã1 /4π , (47)
√
ℓ1 = ℓ2 = 72πã1 + 9π(−4 + a0 r1 )a1 + 16ωa30 (−16 + 3 3π), (48)
√
ℓ3 = 36πa1 + 9πa0 ã1 r̃1 + 16ωa30 (−16 + 3 3π), (49)

1
η̃−2 (Θ) = (−12Θ − 3 sin 2Θ + 3 sin 4Θ + sin 6Θ) , (50)
12
1 
30 cot Θ + 6Θ(8 − 5 csc2 Θ) − 13 sin 2Θ − 2 sin 4Θ + sin 6Θ ,

η−2 (Θ) = (51)
12
1
η̃−3 (Θ) = (12Θ − 3 sin 2Θ − 3 sin 4Θ + sin 6Θ) , (52)
12
2
η−3 (Θ) = sin Θ 2Θ (2 cos 2Θ + cos 4Θ) + 5 sin 2Θ + 3 sin 4Θ + 16 sin Θ cos3 Θ (−γ + ln sin Θ) ,
 
(53)
5 3 2
η−4 (Θ) = 24 sin Θ cos Θ ln sin Θ − sin Θ(2 cos Θ + cos 3Θ) + 3Θ sin Θ(2 cos 2Θ − cos 4Θ), (54)

γ = 0.57721566 · · · is Euler’s constant, and b > 0 is an arbitrary length scale upon which D usually depends, namely
D = Db . If one changes b, one should also change D in such a way that the three-body wave function is not affected:

8π 3 h 2  √   √  i b
Db − Db′ = a0 4ω 7 3 − 24/π a30 − 18ωa20 r0 − 6 3 3 + 2π (2ã1 + a1 ) ln ′ . (55)
9 b

If a0 ̸= 0, T (−5) contains terms that depend on B like B −5 ln B for fixed hyperangles. But if a0 = 0, such terms
vanish and D does not depend on b.
The first kind 21-expansion reads

(0) (−1) (−2) (−3) (−4) (−5)

Ψ = S3 + S3 + S3 + S3 + S3 + S3 + O(R−6 lnn3 R), (56)
 7

where n3 is a nonnegative integer,

(0)
S3 = c01,0 ϕ̃(1,0) (s), (57)
1
(−1)
X
S3 = c−1
1,m ϕ̃
(1,m)
(s), (58)
m=−1
1
(−2)
X
S3 = c−2
1,m ϕ̃
(1,m)
(s), (59)
m=−1
1 l
(−3)
X X X
S3 = c̃−1 ˜(1,m) (s) + c−3 (l,m)
1,m f l,m ϕ̃ (s), (60)
m=−1 l=1,3 m=−l
1 l
(−4)
X X X
S3 = c̃−2 ˜(1,m) (s) + c−4 (l,m)
1,m f l,m ϕ̃ (s), (61)
m=−1 l=1,3 m=−l
l l
(−5)
X X X X
S3 = c̃−3 ˜(l,m) (s) c−5 (l,m)
l,m f + l,m ϕ̃ (s), (62)
l=1,3 m=−l l=1,3,5 m=−l

and
c01,0 = 3, (63a)
√ a0 1,0 √ a0 1,0
c−1
1,m = 2π C1,m;2,−m Y2−m (R̂) − δm,0 10 π C1,0;0,0 Y00 (R̂), (63b)
√ R R
4 2ω ′ a20 1,0 2ω ′′ a2 1,0
−2
c1,m = √ 2 C1,m;2,−m Y2−m (R̂) + δm,0 √ 02 C1,0;0,0 Y00 (R̂), (63c)
3 πR 3 πR
√
ω ′′ = 3 3 + 2π, (63d)

where δi,j is the Kronecker delta symbol. The coefficients c−i −i

l,m and c̃l,m are functions of R, and they satisfy

3 2 −i
c̃−i
l,m = ∇ c . (64)
4 R l,m
When R → ∞, c−i i −i
l,m scales as 1/R and c̃l,m scales as 1/R
i+2
. The coefficients c−i
l,m are shown in Eqs. (63) and
Eqs. (A39–A54).
The second kind 21-expansion reads
(1) (0) (−1) (−2) (−3) (−4) (−5)
Ψ = S1 + S1 + S1 + S1 + S1 + S1 + S1 + O(R1−6 lnn1 R1 ), (65)
where n1 is a nonnegative integer,
(1)
S1 = d10,0 ϕ(0,0) (s1 ), (66)
(0)
S1 = d00,0 ϕ(0,0) (s1 ) + d01,0 ϕ(1,0) (s1 ), (67)
1
(−1)
X
S1 = d−1
0,0 ϕ
(0,0)
(s1 ) + d−1
1,m ϕ
(1,m)
(s1 ), (68)
m=−1
2 l
(−2)
X X
S1 = d˜00,0 f (0,0) (s1 ) + d−2
l,m ϕ
(l,m)
(s1 ), (69)
l=0 m=−l
1 l 3 l
(−3)
X X X X
S1 = d˜−1
l,m f
(l,m)
(s1 ) + d−3
l,m ϕ
(l,m)
(s1 ), (70)
l=0 m=−l l=0 m=−l
2 l 4 l
(−4)
X X X X
S1 = d˜−2
l,m f
(l,m)
(s1 ) + d−4
l,m ϕ
(l,m)
(s1 ), (71)
l=0 m=−l l=0 m=−l
3 l 5 l
(−5) ˜ (0,0) X X X X
S1 = d˜−1
0,0 g (s 1 ) + ˜
d −3 (l,m)
l,m f (s1 ) + d−5
l,m ϕ
(l,m)
(s1 ), (72)
l=0 m=−l l=0 m=−l
 8

and

d10,0 = R1z , (73a)

1 R1z
d00,0 = − a0 , (73b)
2 R1
3
d01,0 = − , (73c)
2
−1 2ωa20 R1z
d0,0 = − , (73d)
π R12
√ √
−1 2πa0 1,0 −m 5 πa0 1,0
d1,m = − C Y (R̂1 ) + δm,0 C1,0;0,0 Y00 (R̂1 ). (73e)
2R1 1,m;2,−m 2 R1

˜−i ˜
˜−i (1)
The coefficients d−i
l,m , dl,m and dl,m are functions of R1 ,
where hl is the spherical Hankel function of the first
and they satisfy kind, κilν > 0 is the binding wave number such that the
bound pair has binding energy ℏ2 κ2ilν /mF , and uilν (si )
3 2 −i is the radial part of the wave function for the bound pair
d˜−i
l,m = ∇R1 dl,m , (74)
4 and satisfies the Schrödinger equation and the normal-
˜ 3 ization condition
dl,m = ∇2R1 d˜−i
˜−i
l,m . (75)
4
ℏ2 2 ℏ2 2
 
˜−i − ∇si + Vi (si ) + κilν uilν (si )Ylm (ŝi ) = 0,
When R1 → ∞, d−i i
l,m scales as 1/R1 , dl,m scales as mF mF
˜ (78)
1/R1i+2 , and d˜−i i+4 −i
l,m scales as 1/R1 . The coefficients dl,m
are shown in Eqs. (73) and Eqs. (A55–A77).
Note that the 21-expansions Eq. (56) and Eq. (65) are Z ∞
δνν ′
applicable only when the interactions do not support any s2 u∗ilν (s)uilν ′ (s)ds = . (79)
0 κ5ilν
two-body bound states. If the interactions can support
one or more two-body bound states, then three-body re- Because of the restrictions from the Fermi statistics and
combination will occur. In this case, two fermions may the parity of the leading order term, l must be odd when
form a bound state and release the binding energy, and i = 3, and c1ll′ ν = (−1)l+1 c2ll′ ν and u1lν = u2lν .
the bound pair and the free fermion fly apart from each
Since the outgoing wave contributes a positive prob-
other with total kinetic energy equal to the released bind-
ability flux toward the outside of a large hypersurface
ing energy. Thus the 21-expansions should be modified
(defined as the surface with a constant and large value of
as [41]
B in the three-body configuration space), D must gain
3 a negative imaginary part to satisfy the conservation of
(m) probability [41, 42]. From this conservation of probabil-
X X 1,0
Ψ=Ψ
e+ cill′ ν Cl,m;l ′ ,−m φill′ ν (si , Ri ), (76)
i=1 ν,l,l′ ,m
ity, we derive
√
3 3 X |cill′ ν |2
where Ψ e is equal to the right hand side of Eq. (56) when Im(D) = − , (80)
s is fixed and R is large, and equal to the right hand side 8 ′
κ6ilν
ill ν
(m)
of Eq. (65) when s1 is fixed and R1 is large. φill′ ν (si , Ri )
specifies that a bound pair and the remaining fermion where Im(D) is the imaginary part of D, and the sum-
fly apart from each other, where l, m, and ν are the or- mation is over all the dimer states. In Sec. V, we will
bital angular momentum quantum number, the magnetic discover the relation between Im(D) and the three-body
quantum number and the vibrational quantum number of recombination rate.
the bound pair formed by particles j and k, respectively,
and l′ is the orbital angular momentum quantum number
for the relative motion of the bound pair and particle i. III. THE BORN APPROXIMATION
cill′ ν is a coefficient. If |l − l′ | ≥ 2 or l = l′ = 0, we define
cill′ ν = 0. We find If the interaction potentials are weak, we can expand
(m)
D in powers of the potentials. We first express the three-
φill′ ν (si , Ri ) = uilν (si )Ylm (ŝi ) body wave function as a Born series for weak potentials:
 
l′ +1 (1) 2
×i hl′ √ κilν Ri Yl−m
′ (R̂i ), (77)  2
3 Ψ = Ψ0 + GVΨ
b 0 + GV
b Ψ0 + . . . , (81)
 9
P
where V = U (s1 , s2 , s3 )+ i Vi (si ) is the total interaction A. One spin-↑ fermion and two spin-↓ fermions in a
b = −H
operator, G b −1 is the Green’s operator, and H b 0 is cubic box
0
the three-body kinetic energy operator. Starting from
Eq. (6), we derive For simplicity, we assume that all the two-body inter-
actions are fine-tuned such that the two-body scattering
1 α3,3 1 α3,1 1 α3,2 α1,1 phase shifts are zero at small collision energies, but the
GVΨ 0 = − sz + s1z 3 + s2z 3 − R1z
b
3 s3 6 s1 6 s2 s1 three-body hypervolume D ̸= 0. The particles are la-
√ beled in the same way as in Sec. II. We place them in a
α1,2 3 sz
+ R2z − Λ + O(B −6 ) (82) large periodic cubic box with volume Ω. If these fermions
s2 π B6 have no interactions, we may consider an energy eigen-
at large B, where state in which fermions 1, 2, and 3 have momenta ℏk1 ,
ℏk2 , and ℏk3 respectively, with normalized wave function
mF ∞ ′n+1
Z
αn,i ≡ 2 s Vi (s′ )ds′ , (83) eik3 ·r3 eik1 ·r1 eik1 ·r2
ℏ 0 Ψk1 ,k2 ,k3 (r1 , r2 , r3 ) = √ ik ·r ik ·r . (86)
mF
Z ∞ Z ∞ Z π 2Ω3/2 e 2 1 e 2 2
Λ≡ 2 ds′ s′4 dR′ R′2 dθ′ sin θ′ U (s′ , R′ ),
ℏ 0 0 0 We define the momenta ℏkc , ℏp and ℏq in the COM
(84) frame such that

where θ′ is the angle between s′ and R′ , and U (s, R) ≡ 1 1

k1 = kc + q + p, (87)
U (s1 , s2 , s3 ). Comparing these results with the 111- 3 2
expansion, we find the expansions of a0 , a1 , ã1 and D 1 1
k2 = kc + q − p, (88)
to leading order in powers of the potential: 3 2
1
k3 = kc − q, (89)
a0 = α1,1 = α1,2 , (85a) 3
1 1
a1 = α3,1 = α3,2 , (85b) where ℏkc is the total momentum. We then factorize the
9 9 wave function as
1
ã1 = α3,3 , (85c) 1
9 Ψk1 ,k2 ,k3 (r1 , r2 , r3 ) = √ eikc ·Rc Φq,p , (90)
8π 2 Ω
D= Λ. (85d)
3
where Rc = (r1 + r2 + r3 )/3, and
The details of the derivation can be found in Appendix
B. 1
Φq,p = √ e−iq·R3 eip·s3 − e−ip·s3


(91)
For any particular two-body potentials Vi (si ), such 2Ω
as the square-well potential and the Gaussian potential,
one can calculate a0 , a1 and ã1 by solving the two-body is the wave function for the relative motion of the three
Schrödinger equation and verify that the results are con- particles. Consider a state with k1 ∼ k2 ∼ k3 ∼ k ≪
sistent with Eqs. (85) if the potentials are weak. From 1/re , in which the de Broglie wave length of each fermion
Eq. (85d) we know that D linearly depends on U if U is of the order of λ = 2π/k ≫ re . When we introduce
is weak. One can find more precise approximate formu- short-range interactions with range re , the wave function
las for the two-body scattering parameters and D if the is only slightly modified at B ≫ re . When the hyper-
wave function is expanded to the second order in the radius B satisfies re ≪ B ≪ λ, one can expand Φq,p
Born series. When the interactions are not weak, one as
can solve the three-body Schrödinger equation numer- √
2
ically and match the resultant wave function with the Φq,p ≃ ip · s3 . (92)
Ω
asymptotic expansions in Eq. (29) or Eq. (65) to extract
the numerical value of D. We then introduce a nonzero D adiabatically, and Φq,p
is changed to
√ √ !
IV. SHIFTS OF ENERGY DUE TO D 2 3 3D
Φq,p ≃ ip · s3 1 − 3 6 (93)
Ω 8π B
In this section we first consider one spin-↑ fermion
and two spin-↓ fermions in a periodic box and study the for re ≪ B ≪ λ.
energy shifts of this system caused by the three-body Φq,p satisfies the Schrödinger equation,
scattering hypervolume. We then derive the thermody-
namic properties of two-component Fermi gases due to ℏ2 2
the nonzero three-body scattering hypervolume. − ∇ Φq,p = EΦq,p , (94)
mF ρ
 10

outside the range of interaction, where E√is the energy 1. Adiabatic shifts of energy in the thermodynamic limit
of the relative motion, and ρ = (s, 2R/ 3). Suppose
that the box size Ω1/3 is large, and consider two different Note that in such a system, there are four different
interactions that yield two different three-body scatter- scattering hypervolumes, namely, D↑↑↑ , D↑↑↓ , D↓↓↑ , and
ing hypervolumes D1 and D2 . We get the following two D↓↓↓ , where the subscript σσ ′ σ ′′ means that the three-
equations body scattering hypervolume is defined to describe the
ℏ2 2 scattering between three particles with spins σ, σ ′ , and
− ∇ Φ1 = E 1 Φ1 , (95a) σ ′′ . D↑↑↑ and D↓↓↓ have been studied in Ref. [30]. If
mF ρ
we introduce these four nonzero scattering hypervolumes
ℏ2 2 adiabatically, the energy shift at the first order in the
− ∇ Φ2 = E 2 Φ2 , (95b)
mF ρ scattering hypervolumes is equal to the sum of the con-
where the subscripts q, p have been suppressed for sim- tributions from all the triples of fermions with spins ↑↑↑,
plicity. Multiplying both sides of Eq. (95a) by Φ∗2 , mul- ↑↑↓, ↓↓↑, or ↓↓↓. We find
tiplying both sides of Eq. (95b) by Φ∗1 and taking the X 1 D
complex conjugates of both sides, subtracting the two ∆E = E ↑↑↓ nk ↓ nk2 ↑ nk3 ↑
2! k1 k2 k3 1
resultant equations, and taking the integral over ρ for k1 k2 k3

ρ > ρ0 , we find 1 D↓↓↑

+ E nk ↑ nk2 ↓ nk3 ↓
Z 2! k1 k2 k3 1
(E1 − E2 ) d6 ρ Φ1 Φ∗2 1 D
ρ>ρ0 + Ek1↑↑↑k2 k3 nk1 ↑ nk2 ↑ nk3 ↑
3!
ℏ2
Z 
=− d6 ρ Φ∗2 ∇2ρ Φ1 − Φ1 ∇2ρ Φ∗2 ,

1 D↓↓↓
mF + Ek1 k2 k3 nk1 ↓ nk2 ↓ nk3 ↓ , (100)
ρ>ρ0 3!
(96)
−1
where nkσ = 1 + eβ(ϵk −µσ ) is the Fermi-Dirac dis-
where ρ0 is a length scale such that re ≪ ρ0 ≪ λ. Note tribution function for spin-σ fermions, β = 1/kB T , T is
that Φ1 ≃ Φ2 in the region ρ > ρ0 , and the volume of the the temperature of the system, ϵk = ℏ2 k 2 /2mF is the
region ρ < ρ0 is small and can be neglected on the left kinetic energy of a spin-σ fermion with momentum ℏk,
hand side of Eq. (96), so the left-hand side of Eq. (96) and µσ is the chemical potential of spin-σ fermions. The
becomes energy shifts due to D↑↑↑ and D↓↓↓ have been studied
Z
8 8 in Ref. [30]. In the thermodynamic limit, the summa-
√ (E1 − E2 ) d3 s d3 R |Φ|2 ≃ √ (E1 − E2 ) . tion
3 3 ρ>ρ0 3 3 P over Rk can3
be replaced by the integral over k, i.e.,
3
(97) k → Ω d k/(2π) . We carry out the integral and find
We then carry out the integral on the right-hand side of √ 2 X
πℏ 8 e5/2

µ
eσ /Teσ

Eq. (96) by applying Gauss’s divergence theorem and get ∆E = − N σ̄ Dσσ σ̄ k F σ Tσ Li 5/2 −e
64π 4 mF σ
ℏ2 (D2 − D1 )
E2 − E1 =
2
(k1 − k2 ) . (98) ℏ2 X  
+ Nσ Dσσσ kF10σ Teσ5 Li25/2 −eµeσ /Tσ ,
e
2mF Ω2 3
256π mF σ
This result implies that the energy shift Ek1 ,k2 ,k3 of such (101)
a three-body system due to the three-body scattering
hypervolume is where σ̄ = ↓ when σ = ↑, and σ̄ = ↑ when σ = ↓, Teσ =
T /TF σ , µ
eσ = µσ /ϵF σ , and Liν (z) is the polylogarithm
ℏ2 D 2
Ek1 ,k2 ,k3 = (k1 − k2 ) . (99) function of order ν. The chemical potential µσ can be
2mF Ω2 determined by the number of fermions, i.e.,
d3 k
Z
1
B. Energy shifts of many two-component fermions Nσ = Ω , (102)
(2π)3 1 + eβ(ϵk −µσ )
and the thermodynamic consequences
which leads to the following equation:
√
Now one can generalize Eq. (99) to N fermions in the 3 π e3/2  
1=− Tσ Li3/2 −eµeσ /Tσ . (103)
e
large periodic box with volume Ω. Suppose there are Nσ 4
spin-σ fermions, where σ =↑, ↓, and N = N↑ + N↓ . The In the low temperature limit, T ≪ TF σ , we get
number density of the spin-σ fermions is nσ = Nσ /Ω,
X  ℏ2 Nσ̄ Dσσσ̄ k 8  1 π 2 
and we assume the density is low enough such that the ∆E(T ) = Fσ
+ 2
T + O(Tσ )
e e 4
average interparticle distance n−1/3 ≡ (n↑ + n↓ )−1/3 ≫ σ
24π 4 mF 5 12 σ
re . For later use, we define the Fermi wave number kF σ ≡
ℏ2 Nσ Dσσσ kF10σ 1 π 2 e2
 
(6π 2 nσ )1/3 , the Fermi energy ϵF σ ≡ ℏ2 kF2 σ /2mF , and + + T + O(Te 4
) .
the Fermi temperature TF σ ≡ ϵF σ /kB , where kB is the 180π 4 mF 5 6 σ σ

Boltzmann constant. (104)

 11

8
In particular, at zero temperature we get
1.04

ℏ2 X 1.02

∆E(0) = 4
Nσ̄ Dσσσ̄ kF8 σ 6
1.00
120π mF σ 0.98

0.96
2 0.00 0.02 0.04 0.06 0.08 0.10
ℏ X
Nσ Dσσσ kF10σ .

ΔE(T)/ΔE(0)
+ (105)
900π 4 mF σ
4
ΔE-adiabatic

ΔE-isothermal
If we consider the effects of Dσσσ̄ only, the energy shift
is 2

ℏ2 1 π 2 e2
X  
8
∆E(T ) = N σ̄ Dσσσ̄ k F σ + T + O(Teσ4 )
24π 4 mF σ 5 12 σ 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
(106) T/ T F↓

at T ≪ TF σ .
In the intermediate temperature regime, TF σ ≪ T ≪ FIG. 2. The shifts of energy caused by the adiabatic (red
Te , we have line) or isothermal (blue dashed line) introduction of D↓↓↑ as
functions of temperature.
ℏ2 X
∆E(T ) = 4
Nσ̄ Dσσσ̄ kF8 σ
48π mF σ
 
1 temperature limit, T ≪ TF σ ,
× Teσ + √ Teσ−1/2 + O(Teσ−2 )
3 2π
2
1 π 2 e2
ℏ2 N↑ D↓↓↑ 8
ℏ X  
+ 4
Nσ Dσσσ kF10σ ′
∆E (T ) = − 4
T + O(T↓ ) .
k e
72π mF σ 5 12 ↓
24π 4 mF F ↓

1 e2 1 e1/2
 (111)
−1
× T + √ T + O(Tσ ) ,
e (107) In the intermediate temperature regime, TF σ ≪ T ≪ Te ,
2 σ 3 2π σ
ℏ2
where Te ≡ ℏ2 N↑ D↓↓↑ 8
 
2mF re2 kB . The energy shift due to Dσσσ̄ is 1
′
∆E (T ) = k √ Te↓−1/2 + O(Te↓−2 ) .
48π 4 mF F ↓ 2 2π
ℏ2 X (112)
∆E(T ) = 4
Nσ̄ Dσσσ̄ kF8 σ
48π mF σ We plot the shifts of energy as functions of temperature
in Fig. 2.
 
1
× Teσ + √ Teσ−1/2 + O(Teσ−2 ) (108)
3 2π
at TF σ ≪ T ≪ Te .

V. INELASTIC SCATTERINGS AND

2. Isothermal shifts of energy in the thermodynamic limit THREE-BODY RECOMBINATION RATE

For the sake of simplicity, we now assume that only In Sec. II, we mentioned that the three-body scatter-
D↓↓↑ is nonzero, but D↑↑↓ , D↑↑↑ , and D↓↓↓ are zero. ing hypervolume is real if the scattering is purely elastic,
If such an interaction is introduced adiabatically, the but it will gain a negative imaginary part if the scat-
change of temperature will be tering is inelastic. In the latter case, the interactions
  can support one or more two-body bound states, and the
∂∆E
∆T = , (109) three-body recombination will occur. For most ultracold
∂S N,Ω atomic gases, the three-body recombination is possible,
since most ultracold atoms have two-body bound states,
and the temperature will increase (if D↓↓↑ > 0) or de- and it will cause atom loss of the ultracold gas.
crease (if D↓↓↑ < 0). Thus if D↓↓↑ is introduced isother-
mally, the shift of energy ∆E ′ should be approximately When the scattering hypervolume becomes complex,
  the energy shift also gains a negative imaginary part.
∂ Within a short time ∆t, the probability that no three-
∆E ′ = ∆E − C∆T = 1 − T ∆E, (110) body recombination occurs is exp(−2| Im E|∆t/ℏ) ≃ 1 −
∂T
2| Im(E)|∆t/ℏ, and the probability of one recombination
where C is the heat capacity of the non-interacting two- is 2| Im(E)|∆t/ℏ. After one recombination event, three
component Fermi gases at constant volume. In the low atoms will escape from the trap, thus the change of the
 12

number of spin-σ atoms in a short time dt is VI. SUMMARY AND DISCUSSION


dt X 1
dNσ = − 2 Im EkD1σkσ̄2σ̄k3 nk1 σ nk2 σ̄ nk3 σ̄ We studied the three-body problem for two-component
ℏ 2! fermions (one spin-↑ fermion and two spin-↓ fermions)
k1 k2 k3
1 in 3D and derived three asymptotic expansions for the
+2 Im EkD1σσ
k2 k3 nk1 σ nk2 σ nk3 σ̄
σ̄
wave function Ψ. The scattering energy is assumed to be
2!  zero and the angular momentum quantum number L is
1
+3 Im EkD1σσσ
k2 k3 n k1 σ k2 σ k3 σ .
n n (113) assumed to be 1. We defined a new three-body scattering
3! hypervolume D, which appears at the order of B −5 in
the 111-expansion and has the dimension of length to
This leads to the sixth power. The three-body scattering hypervolume
dnσ (1) (2) (3)
plays an important role in the low energy physics of three
= −L3σ nσ n2σ̄ − L3σ n2σ nσ̄ − L3σ n3σ , (114) or more fermions.
dt
We also derived an asymptotic formula for D by using
(1) (2) (3)
where L3σ , L3σ , L3σ are the three-body recombination the Born expansion when the interactions are weak. If
rate constants and they can be expressed as the interactions are strong, one may numerically solve the
three-body Schrödinger equation and extract the numer-
√ ical value of D by matching the numerical wave function
(1) 9 πℏ e5/2  
L3σ = − Tσ̄ Li5/2 −eµeσ̄ /Tσ̄ |Im Dσσ̄σ̄ | kF2 σ̄ ,
e
8mF and the asymptotic expansions we have derived.
(115a) In the remaining part of this paper, we considered
√ these three two-component fermions in a large box with
(2) 9 πℏ e5/2  
periodic boundary condition imposed on the wave func-
L3σ = − T Li5/2 −eµeσ /Tσ |Im Dσσσ̄ | kF2 σ ,
e
4mF σ tion, and found the shifts of the energy eigenvalues due to
(115b) a nonzero D. We then studied the dilute two-component
(3) 27πℏ e5 2   Fermi gases and derived the shifts of their energies due
L3σ = T Li −eµeσ /Tσ |Im Dσσσ | kF4 σ . (115c)
e
32mF σ 5/2 to the four three-body scattering hypervolumes (D↑↑↑ ,
D↑↑↓ , D↓↓↑ , and D↓↓↓ ).
(1) If the interactions can support one or more two-body
L3σ depends on the density nσ̄ and temperature T , and
(2) (3) bound states, then the three-body recombination will oc-
L3σ and L3σ depend on the density nσ and temperature
cur and D must gain a negative imaginary part to satisfy
T.
the conservation of probability. We derived formulas for
In the low temperature limit, T ≪ TF σ , we have (1) (2)
the three-body recombination rate constants L3σ , L3σ
(3)
5π 2 e2
 
(1) 3 ℏ and L3σ in terms of the imaginary parts of the three-
L3σ ≃ |Im Dσσ̄σ̄ | kF2 σ̄ 1 + Tσ̄ , (116a) body scattering hypervolumes, the temperature and the
5 mF 12
densities of the two components of the Fermi gas.
5π 2 e2
 
(2) 6 ℏ
L3σ ≃ |Im Dσσσ̄ | kF2 σ 1 + Tσ , (116b)
5 mF 12
5π 2 e2
 
(3) 6 ℏ 4 Appendix A: DERIVATION OF THE
L3σ ≃ |Im Dσσσ | kF σ 1 + T . (116c)
25 mF 6 σ 111-EXPANSION AND THE 21-EXPANSIONS

In the intermediate temperature regime, TF σ ≪ T ≪ Te , We study the expansion of the scattering wave function
we have for three two-component fermions (2 spin-↓ fermions and
3 1 spin-↑ fermion). We suppose that the collision energy
(1)
L3σ ≃ |Im Dσσ̄σ̄ | (kB T ) , (117a) is zero and the orbital angular momentum L = 1. The
ℏ magnetic quantum number M can be −1, 0 or 1, and
(2) 6
L3σ ≃ |Im Dσσσ̄ | (kB T ) , (117b) here we derive these expansions for the state of M = 0.
ℏ The three-body wave function can be expanded in
(3) mF 2
L3σ ≃ 6 3 |Im Dσσσ | (kB T ) . (117c) three forms:
ℏ
∞
X
Equations (115c), (116c), and (117c) were first derived Ψ01 = T (−p) (r1 , r2 , r3 ), (A1a)
in Ref. [30]. p=−1
According to the three-body threshold laws, the three- ∞
(1) (2) (−q)
X
body recombination rate constants L3σ and L3σ are pre- Ψ01 = S3 (R, s), (A1b)
(3)
dicted to be proportional to T 1 [43, 44], and L3σ is pre- q=0
2 ∞
dicted to be proportional to T [43, 44]. Experimentally, X (−q)
these predictions are confirmed in 6 Li gases [45–47]. Our Ψ01 = S1 (R1 , s1 ), (A1c)
results Eqs. (117) are consistent with these results. q=−1
 13

(−q) (0)
where T (−p) scales as B −p , S3 scales as R−q , and Expanding S3 at s > re , we find
(−q) −q
S1 scales as R1 . If re < s ≪ R, the wave function (0,0)
has a double expansion: t3 = 0, (A12a)
(0,−1)
X (i,j) t3 = 0, (A12b)
Ψ01 = t3 , (A2)
(0,−2) 3ã1
i,j t3 =− sz , (A12c)
s3
and if re < s1 ≪ R1 , the wave function can be expanded t3
(0,j)
= 0, j ≤ −3. (A12d)
as
(1)
X (i,j) Step 2b. S1 satisfies Eq. (28a) and when s1 > re we
Ψ01 = t1 , (A3)
have
i,j
(1) (1,0) (1,j)
X
(i,j) (i,j)
S1 = t1 + t1 . (A13)
where t3 scales as R s , and i j
scales as t1 R1i sj1 . j≤−1
Step 1. The leading-order term in the 111-expansion
is So we deduce that
s1z (1)
X
T (1) = sz = − + R1z , (A4) S1 = d1l,m ϕ(l,m) (s1 ), (A14)
2 l,m

which indicates that where l must be equal to 0 for a reason similar to what
(0,1) we discussed in Step 2a. We expand Eq. (A14) to the
t3 = sz , (A5a) order s01 at s1 > re and find
(i,j)
t3 = 0, if i + j = 1 and j ̸= 1, (A5b) (1,0)
t1 = d10,0 . (A15)
and
Comparing this result with Eq. (A6a), we find
(1,0)
t1 = R1z , (A6a)
s1z d10,0 = R1z . (A16)
(0,1)
t1 =− , (A6b)
2 So we have
(i,j)
t1 = 0, if i + j = 1, but i ̸= 1 and j ̸= 1. (A6c) (1)
S1 = R1z ϕ(0,0) (s1 ). (A17)
(2) (3) (4) (i,j) (i,j)
Since T , T , T , . . . are zero, we get = = t3 t1 Expanding S1
(1)
at s1 > re , we find
0 if i + j ≥ 2.
(0)
Step 2a. Since S3 satisfies Eq. (27a) and when s3 > re (1,−1) a0
t1 =− R1z , (A18a)
we have s1
(1,j)
(0) (0,1)
X (0,j) t1 = 0, j ≤ −2. (A18b)
S3 = t3 + t3 , (A7)
j≤0 Step 3. At re < s1 ≪ R1 , we expand T (0) as
we deduce that T (0) = t1
(1,−1)
+ t1
(0,0)
+ t1
(−1,1)
+ t1
(−2,2)
+ ...
(0)
X
S3 = c0l,m ϕ̃(l,m) (s). (A8) = t1
(1,−1)
+ O(s01 ). (A19)
l,m
−(∇2s1 + 3∇2R1 /4)T (0) should be a linear combination
(l,m)
Note that l must be equal to 1, because ϕ̃ (s) contains of the Dirac δ function of s1 and perhaps some partial
a term proportional to sl at s > re . Expanding Eq. (A8) derivatives of such a δ function when s1 ≪ R1 , since
to the order s1 and using Eq. (17), we find T (0) should satisfy the free Schrödinger equation outside
r of the range of the interaction. Note that
(0,1)
X
0 s 4π m
t3 = c1,m Y (ŝ). (A9) −∇2s1 t1
(1,−1)
= −4πa0 R1z δ(s1 ), (A20)
m
3 3 1
so we have
Comparing this result with Eq. (A5a), we find  
3 2
c01,0 = 3, (A10a) − ∇s1 + ∇R1 T (0) = −4πa0 R1z δ(s1 )
2
(A21)
4
c01,±1 = 0. (A10b)
if s2 ̸= 0 and s ̸= 0. We solve the above equation and
Thus we have find
a0 (0)
S3
(0)
= 3ϕ̃(1,0) (s). (A11) T (0) = − R1z + T2 , (A22)
s1
 14

(0)
where T2 is a term that satisfies 0. So the full T (0) (which is anti-symmetric under the

(0)
2 3 2
− ∇s1 + 4 ∇R1 T2 = 0, possibly except at s2 = 0 or interchange of particles 1 and 2) should be
s = 0. At s ≪ R, we expand T (0) as
a0 a0
T (0) = t3
(0,0)
+ t3
(−1,1) (−2,2)
+ t3 + t3
(−3,3)
+ ... T (0) = − R1z + R2z . (A24)
s1 s2
(0,0)
= t3 + O(s1 ), (A23)
(i,j)
which indicates that −(∇2s + 34 ∇2R )T (0) vanishes at s = If s ≪ R, we expand T (0) as
P
i+j=0 t3 and find

 
(−1,1) a0 Rz Rs 1
t3 =− s 3 + 3sz , (A25a)
2 R R
(−2,2) (−4,4)
t3 = t3 = 0, (A25b)
"
#
2
(−3,3) a0 3 Rz Rs 3R − R5Rs2 2
− 3Rs2
t3 = s 7
+ 3s2 sz , (A25c)
16 R R5
"
#
4 2 2 4
(−5,5) a0 5 Rz Rs 15R − 70R Rs + 63Rs 4 3R4 − 30R2 Rs2 + 35Rs4
t3 =− s + 3s sz , ..., (A25d)
256 R11 R9

(i,j)
where Rs ≡ R · ŝ. If s1 ≪ R1 , we expand T (0) as
P
i+j=0 t1 and find

(0,0) a0 R1z
t1 =− , (A26a)
2 R
 1 
(−1,1) a0 R1z R1s 1
t1 = s1 + 3s 1z , (A26b)
4 R13 R1
"
#
(−2,2) a0 2 R1z R12 − 3R1s 2
R1s
t1 = s − 6s1 s1z 3 , (A26c)
16 1 R15 R1
"
#
(−3,3) a0 3 R1z R1s 3R12 − 5R1s 2
R 2
− 3R 2
t1 =− s + 3s21 s1z 1 5 1s , (A26d)
32 1 R17 R1
"

#
4 2 2 4
(−4,4) a0 4 R1z 3R1 − 30R1 R1s + 35R1s 3 R1s 3R12 − 5R1s 2
t1 =− s − 12s1 s1z , (A26e)
256 1 R19 R17
"
#
4 2 2 4 4 2 2 4
(−5,5) a0 R 1z R 1s 15R 1 − 70R 1 R 1s + 63R 1s 3R − 30R R + 35R
t1 = s5 + 3s41 s1z 1 1 1s 1s
, ..., (A26f)
512 1 R111 R19

where R1s ≡ R1 · ŝ1 . re , we get

(−1)
Step 4a. We expand S3 at s > re as
1 r
(−1,1)
X s 4π m
t3 = c−1 Y (ŝ). (A29)
(−1) (−1,1) (−1,j) 1,m
3 1
X
S3 = t3 + t3 . (A27) m=−1
3
j≤0
Comparing this result with Eq. (A25a), we find
(−1)
S3 satisfies Eq. (27b), thus we have √
3 2a0 Rz (Rx + iRy )
c−1
1,−1 =− , (A30a)
S3
(−1)
=
X
c−1 (l,m) 4 R3
l,m ϕ̃ (s), (A28)
3a0 3R2 + Rz2
l,m c−1
1,0 =− , (A30b)
√2 R3
where l must be equal to 1 because Eq. (A28) must be 3 2a0 Rz (Rx − iRy )
c−1
1,1 = . (A30c)
compatible with Eq. (A27). Expanding Eq. (A28) at s > 4 R3
 15

One can re-express these results in terms of the Clebsch-

(−1)
Gordan coefficients and get Eq. (63b). Expanding S3 √ √ !
at s > re , we get 2 2πJ3 9 πa0 ã1 r̃1 1,0
c−3
1,m = − √ C1,m;2,−m Y2−m (R̂)
R3 2 2R3
(−1,0)
t3 = 0, (A31a) √ J3 + 3K3 1,0
(−1,−1) − δm,0 2 π C1,0;0,0 Y00 (R̂), (A39)
t3 = 0, (A31b) R3
3a0 ã1

1 Rz Rs sz 1

9a1 2ωa30  √ 
(−1,−2) J3 = − + 16 − 3 3π , (A40)
t3 = +3 3 , (A31c) 2 π
2 s2 R3 s R
(−1,j) 3a1 2ωa30  √ 
t3 = 0, j ≤ −3. (A31d) K3 = − 6 − 3π , (A41)
2√ π
Step 4b. We expand S1
(0) 5 21πa0 1,0
at s1 > re as c−3
3,m = 3
C3,m;4,−m Y4−m (R̂)
2R√
(0) (0,1) (0,0)
X (0,j)
S1 = t1 + t1 + t1 . (A32) 39 7πa0 1,0
− C3,m;2,−m Y2−m (R̂), (A42)
j≤−1 2R3
(0)
S1 satisfies Eq. (28b), thus we have √
2 2  √ 
1,0
(0)
X c−4 = − √ 3 3ξ4 + 8πζ4 C1,m;2,−m Y2−m (R̂)
S1 = d0l,m ϕ(l,m) (s1 ), (A33) 1,m
3 πR 4

l,m 1 √ 
1,0
+ δm,0 √ 4 3ξ4′ + πζ4′ C1,0;0,0 Y00 (R̂),
2 πR
where l must be 0 or 1, because Eq. (A33) must be com-
(A43)
patible with Eq. (A32). Expanding Eq. (A33) at s1 > re ,
we have ξ4 = 6a0 ã1 + 57a0 a1 + 20a0 ω3 − 15a20 ã1 r̃1 , (A44)
1 r ζ4 = − 15a0 a1 − 4a0 ω3 + 3a20 ã1 r̃1 , (A45)
(0,1)
X
0 s1 4π m
t1 = d1,m Y (ŝ1 ), (A34) ξ4′ = 8a0 ω3 − 42a0 a1 − 30a0 ã1 + 3a20 ã1 r̃1 , (A46)
m=−1
3 3 1
ζ4′ 3a20 ã1 r̃1


(0,0)
= 2 8a0 ω3 + 30a0 a1 + /3, (A47)
t1 = d00,0 . (A35) √ ′′′ 2
16 21πω a0 1,0
c−4
3,m =− C3,m;4,−m Y4−m (R̂)
Comparing these results with Eq. (A6b) and Eq. (A26a), 3πR 4
√
we find 8 7πa20 1,0
+ 4
C3,m;2,−m Y2−m (R̂), (A48)
a0 R1z √ R
d00,0 =− , (A36)
2 R1 ω ′′′ = 189 3 − 104π, (A49)

√
d01,−1 = 0, (A37a) πω5 1,0
c−5
1,m = − C Y −m (R̂)
3 R5√ 1,m;2,−m 2
d01,0 =− , (A37b) πω5′ 1,0
2 + δm,0 C1,0;0,0 Y00 (R̂), (A50)
R 5
d01,1 = 0. (A37c) √ 2 
2a0 a1 √  √
(0) ω5 = 39 3 − 10π − 2a0 ω4 , (A51)
Expanding S1 at s1 > re , we find 2π
a2 a1  √  18
(0,−1) a20 1 R1z ω5′ = 0 87 3 − 50π − 10a0 ω4 + ωa30 ã1 r̃1 ,
t1 = , (A38a) π π
2 s1 R1 (A52)
3a1 s1z √
(0,−2) 21π
t1 = , (A38b)
h
2 s31 c−5
3,m = − 5
420a1 + 105a0 ã3 r̃3
8R
(0,j)  √  i
t1 = 0, j ≤ −3. (A38c) + 16 315 3 − 1712/π ωa30 C3,m;4,−m 1,0
Y4−m (R̂)
√
One can repeat this procedure step by step, and suc- 30 7ωa30 1,0
(−2) (−1)
cessively determine T (−1) , S3 , S1 , T (−2) , S3 ,
(−3) + √ 5 C3,m;2,−m Y2−m (R̂), (A53)
πR
(−2)
S1 , . . . , T (−5) (−5)
, and S1 . We expand the three-body √
945 11πa0 1,0
wave function order by order in this way, and derive the c−5
5,m = √ C5,m;6,−m Y6−m (R̂)
111-expansion Eq. (29) and the 21-expansions Eq. (56) 8 2R5
√
and Eq. (65). 735 165πa0 1,0
−i − C5,m;4,−m Y4−m (R̂). (A54)
The following is the list of the coefficients cl,m : 8R5
 16
√
−i
The following is the list of the coefficients dl,m : 8 21ω ′′′ a20 1,0
d−4
3,m = √ C3,m;4,−m Y4−m (R̂1 )
√ 3 πR14
2ω3 π 1,0 √
d−2 = − √ 2 C0,0;1,0 Y10 (R̂1 ), (A55) 4 7πa20 1,0
0,0
3R − C3,m;2,−m Y2−m (R̂1 ), (A67)
√ 1′ 2 R14
√
2 2ω a 1,0 105 15πa0 1,0
d−2
1,m = − √ 20 C1,m;2,−m Y2−m (R̂1 ) d−4 =− C4,m;5,−m Y5−m (R̂1 )
3 πR1 4,m
16R14
ω ′′ a2 1,0 √
− δm,0 √ 0 2 C1,0;0,0 Y00 (R̂1 ), (A56) 255 3πa0 1,0
3 πR1 + C4,m;3,−m Y3−m (R̂1 ), (A68)
√ 4R14
3 5πa0 1,0
d−2
2,m =− C2,m;3,−m Y3−m (R̂1 )
4R12
√
3 30πa0 1,0 a0 r0′ ˜˜−1 a0 r0 ˜−3
+ C2,m;1,−m Y1−m (R̂1 ), (A57) d−5
2R12 0,0 = d + d
24 0,0 2 0,0
" √ √
R1z 135 3a0 a2 3 3D
√ + 6 −
2 πω4 1,0 R1 4π 8π 3
d−3
0,0 = √ 3 C0,0;1,0 Y10 (R̂1 ), (A58)
3R1 2a0 Ω5 √ √
  
b
+ 3 + π + 6 3 ln − Ω6 , (A69)
ℓ1 1,0 3π R1
d−3
1,m = √ C1,m;2,−m Y2−m (R̂1 ) √ √
4 2πR13 a1 h  
Ω6 = 2 2π/3 − 5 3/4 ℓ1 − 8ωa30 4 3
4ωa3 1,0 2π
− δm,0 √ 03 C1,0;0,0 Y00 (R̂1 ), (A59) √ √

b
πR1 −3 3γ + π/6 + 3 3 ln
√ R1
8 5a20  √ 
1,0 √  √
d−3 = √ 3 −9 3 + 5π C2,m;3,−m Y3−m (R̂1 ) ã1 h 
2,m
πR1 + 2 2π/3 − 5 3/4 ℓ3 − 8ωa30 4 3
√ π
3 10a20 1,0 √ √

b
− √ 3 C2,m;1,−m Y1−m (R̂1 ), (A60) −3 3γ + π/6 + 3 3 ln , (A70)
πR1 R1
√ √ h
5 21πa0 1,0 2π 2  √ 
d−3
3,m =− C3,m;4,−m Y4−m (R̂1 ) −5
d1,m = a (2ã1 − a1 ) 39 3 − 10π /π
4R13 4R15 0
√ i
39 7πa0 1,0 1,0
− 2a0 ω4 C1,m;2,−m Y2−m (R̂1 )
+ C3,m;2,−m Y2−m (R̂1 ), (A61)
4R13 √ h
√ π  √ 
2 πΩ5 1,0 − δm,0 5 a20 (2ã1 − a1 ) 174 3 − 100π /π
d−4
0,0 = √ 4 C0,0;1,0 Y10 (R̂1 ), (A62) 4R1
3R1 i
√ ′ 2 √ ! + 36ωa30 a1 r1 /π − 20a0 ω4 C1,0;0,01,0
Y00 (R̂1 ),
2ω a0 a1 r1 2 2Ω4 1,0
d−4
1,m = √ 4 − √ 4 C1,m;2,−m Y2−m (R̂1 ) (A71)
πR1 3 πR1
a0 Ω′3 10a0 Ω
e 3 1,0
+ δm,0 √ C 1,0 Y 0 (R̂1 ) (A63) d−5
2,m = √ C2,m;3,−m Y3−m (R̂1 )
12 πR14 1,0;0,0 0 15πR15
√ √
 
10a0 Ωe ′ 1,0
Ω4 = 2ω ′ ω3 a0 + 3 −57 3 + 40π a0 ã1 − √ 5 3 C2,m;1,−m Y1−m (R̂1 ), (A72)
πR1
3 √ 
− −51 3 + 40π a0 a1 , (A64) e 3 = 27 (9a0 a2 r2 − 8ω3 + 84ã1 + 42a1 )
Ω
h2 √  √ i √
Ω′3 = a1 60π − 36 3 − a0 r1 9 3 + 6π − 3π (45a0 a2 r2 − 40ω3 + 396ã1 + 198a1 ) ,
h  √   √ i (A73)
+ 4 ã1 63 3 − 30π − ω3 6 3 + 4π , Ω ′
e = 6ω3 + 9a0 a2 r2 /2
3
(A65) √  √ 
√ − 3 (ã1 + a1 /2) 21 3 − 16π , (A74)
5π 6ωa30 √

√
 
d−4
2,m = 4
√ 82 3 − 45π 21 h  √ 
2R1 3π d−5
3,m = − √ 5 16ωa30 1712 − 315 3π
 a1  45
 16 πR1
1,0
−45 ã1 + + a0 a2 r2 C2,m;3,−m Y3−m (R̂1 ) i
1,0
2 8 − 105π (a0 a3 r3 + 8ã1 − 4a1 ) C3,m;4,−m Y4−m (R̂1 )
√ √
2 30ωa30 1,0 15 7ωa3 1,0
− √ 4 C2,m;1,−m Y1−m (R̂1 ), (A66) − √ 5 0 C3,m;2,−m Y2−m (R̂1 ), (A75)
πR1 πR1
 17
√
8 15a20  √ 
1,0
d−5
4,m = − √ 5 1269 3 − 700π C4,m;5,−m Y5−m (R̂1 )
πR1
√
4 3a20  √ 
1,0
− √ 5 27 3 − 8π C4,m;3,−m Y3−m (R̂1 ),
πR1
(A76)
√
945 11πa0 1,0
d−5
5,m = − √ C5,m;6,−m Y6−m (R̂1 )
16 2R15
√
735 165πa0 1,0
+ C5,m;4,−m Y4−m (R̂1 ). (A77)
16R15

Appendix B: THE BORN APPROXIMATION FOR THE THREE-BODY WAVE FUNCTION

For weak interactions, we expand the three-body wave function as a Born series:
 2
Ψ = Ψ0 + GVΨb 0 + GV
b Ψ0 + . . . , (B1)
P
where V = U (s1 , s2 , s3 ) + i Vi (si ) is the total interaction operator, G b = −H b −1 is the Green’s operator, H
b 0 is the
0
three-body kinetic energy operator, and Ψ0 = sz is the wave function of three free fermions. We assume that Vi (si )
vanishes if si > re and that U (s1 , s2 , s3 ) vanishes if s1 > re or s2 > re or s3 > re .
For the first order term in the Born series, we have
3
X
GVΨ
b 0 = GU Ψ0 +
b GV
b i Ψ0 ,
i=1
Z 3 Z
mF 6 ′ ′ ′ ′ ′ ′ mF X
= d ρ i G (ρi − ρi ) U (s , s
1 2 3 , s )Ψ0 (ρi ) + d6 ρ′i G (ρi − ρ′i ) Vi (s′i )Ψ0 (ρ′i ) , (B2)
ℏ2 ℏ2 i=1
√ √
where ρi = (si , 2Ri / 3) and ρ′i = (s′i , 2R′i / 3) are six dimensional vectors, and G (ρi − ρ′i ) is the Green’s function
in six dimensional space,
1
G (ρi − ρ′i ) = − 3 . (B3)
4π |ρi − ρ′i |4
Defining Ψ1,i = GV
b i Ψ0 , we have
Z
mF 1
Ψ1,i = − d6 ρ′i 3 Vi (s′i )Ψ0 (ρ′i ),
ℏ2 4π |ρi − ρ′i |4
√
Vi (s′i )Ψ0
Z Z
3mF 3 ′ 3 ′
=− d si d Ri 2 . (B4)
8π 3 ℏ2 3
(si − s′ )2 + (Ri − R′ )2

4 i i
Carrying out this integral, we get
   
1 α3,1 (s1 ) α1,1 (s1 )
Ψ1,1 = s1z + ᾱ0,1 (s1 ) − R1z + ᾱ0,1 (s1 ) , (B5a)
6 s31 s1
   
1 α3,2 (s2 ) α1,2 (s2 )
Ψ1,2 = s2z + ᾱ0,2 (s2 ) − R2z + ᾱ0,2 (s2 ) , (B5b)
6 s3 s2
 2 
1 α3,3 (s)
Ψ1,3 = − sz + ᾱ0,3 (s) , (B5c)
3 s3
where
mF si ′ ′n+1
Z
αn,i (si ) ≡ 2 ds s Vi (s′ ), (B6a)
ℏ 0
mF ∞ ′ ′n+1
Z
ᾱn,i (si ) ≡ 2 ds s Vi (s′ ). (B6b)
ℏ si
 18

At si > re ,
Z ∞
mF
αn,i (si ) =αn,i ≡ 2 ds′ s′n+1 Vi (s′ ), (B7)
ℏ 0
ᾱn,i (si ) =0. (B8)

If s1 , s2 , and s3 are all greater than re ,

3
X 1 α3,3 1 α3,1 α1,1 1 α3,2 α1,2
Ψ1,i = − sz 3 + s1z 3 − R1z + s2z 3 + R2z . (B9)
i=1
3 s 6 s1 s1 6 s2 s2

Defining Ψ1,U = GU
b Ψ0 , we have
Z
mF 1
Ψ1,U = − d6 ρ′ 3 U (ρ′ )Ψ0 (ρ′ )
ℏ2 4π |ρ − ρ′ |4
√
U (s′1 , s′2 , s′ )
Z Z
3mF 3 ′ 3 ′ ′
=− d s d R 2 sz . (B10)
8π 3 ℏ2 3

′ 2
(s − s ) + (R − R ) ′ 2
4

1
Since the potential U (s1 , s2 , s3 ) is finite-ranged, the factor 2 in the integrand in Eq. (B10) can be
[ 43 (s−s′ )2 +(R−R′ )2 ]
expanded when s and R go to infinity at a fixed ratio s/R:

1 1 3s · s′ + 4R · R′
2 = + + O(B −6 ). (B11)
B4 B6
3
4 (s − s′ )2 + (R − R′ )2

By symmetry, only the term 3s · s′ /B 6 contributes to the integral. Thus we get

√ Z ∞ Z ∞ Z π Z π Z π Z π
3 3mF s ′ ′ ′ ′4 ′2 ′ ′ ′ ′ ′
Ψ1,U = − ds dR dθ s R sin θ U (s , R ) × dα dγ dβ ′ sin β ′ cos β ′ ŝ′z + O(B −6 )
8π 3 ℏ2 B 6 0 0 0 −π −π 0
√
3 sz
=− Λ + O(B −6 ), (B12)
π B6
where α′ , β ′ and γ ′ are three Euler’s angles, θ′ is the angle between s′ and R′ , and

ŝ′z = ŝ′ · ẑ = sin θz cos ϕz sin β ′ sin α′ − sin θz sin ϕz sin β ′ cos α′ + cos θz cos β ′ , (B13)

and
Z ∞ Z ∞ Z π
mF ′ ′4 ′ ′2
Λ= 2 ds s dR R dθ′ sin θ′ U (s′ , R′ ). (B14)
ℏ 0 0 0

ACKNOWLEDGMENTS

This work was supported by the National Natural Science Foundation of China Grant No. 92365202 and the
National Key R&D Program of China (Grants No. 2019YFA0308403 and No. 2021YFA1400902). We thank Junjie
Liang for discussions.

[1] J. Rotureau, I. Stetcu, B. R. Barrett, M. C. Birse, and trap, Few-Body Syst 54, 635 (2012).
U. van Kolck, Three and four harmonically trapped par- [3] H.-W. Hammer, S. König, and U. van Kolck, Nuclear
ticles in an effective-field-theory framework, Phys. Rev. effective field theory: Status and perspectives, Rev. Mod.
A 82, 032711 (2010). Phys. 92, 025004 (2020).
[2] J. Rotureau, Effective field theory for few particles in a [4] T. Weiss-Attia, M. Schäfer, and B. Bazak, Perturbative
 19

application of next-to-leading order pionless eft for a ≤ superfluid fermionic gases across a feshbach resonance,
3 nuclei in a finite volume, Phys. Rev. D 109, 114515 Phys. Rev. Lett. 97, 180414 (2006).
(2024). [24] T.-L. Ho and S. Zhang, Bose-einstein condensates with
[5] C. N. Gilbreth and Y. Alhassid, Separable effective inter- spin-orbit interaction, Phys. Rev. Lett. 107, 150403
action for the trapped fermi gas: The bec-bcs crossover, (2011).
Phys. Rev. A 85, 033621 (2012). [25] Z.-Y. Shi, S. M. Yoshida, M. M. Parish, and J. Levinsen,
[6] D. Rakshit, K. M. Daily, and D. Blume, Natural and un- Impurity-induced multibody resonances in a bose gas,
natural parity states of small trapped equal-mass two- Phys. Rev. Lett. 121, 243401 (2018).
component fermi gases at unitarity and fourth-order [26] J. Yao, P. Zhang, R. Qi, and H. Zhai, Three-body prob-
virial coefficient, Phys. Rev. A 85, 033634 (2012). lem of bosons near a d-wave resonance, Phys. Rev. A 99,
[7] G. C. Strinati, P. Pieri, G. Röpke, P. Schuck, and M. Ur- 012701 (2019).
ban, The bcs–bec crossover: From ultra-cold fermi gases [27] Y. Chen and X. Cui, Ultrastable super-tonks-girardeau
to nuclear systems, Physics Reports 738, 1 (2018), the gases under weak dipolar interactions, Phys. Rev. Lett.
BCS–BEC crossover: From ultra-cold Fermi gases to nu- 131, 203002 (2023).
clear systems. [28] S. Tan, Three-boson problem at low energy and implica-
[8] A. D. Kerin and A. M. Martin, Quench dynamics of mass- tions for dilute bose-einstein condensates, Phys. Rev. A
imbalanced three-body fermionic systems in a spherical 78, 013636 (2008).
trap, Phys. Rev. A 106, 053310 (2022). [29] Z. Wang and S. Tan, Three-body scattering hypervol-
[9] D. J. Wineland and W. M. Itano, Laser cooling of atoms, ume of particles with unequal masses, Phys. Rev. A 103,
Phys. Rev. A 20, 1521 (1979). 063315 (2021).
[10] J. P. Gordon and A. Ashkin, Motion of atoms in a radi- [30] Z. Wang and S. Tan, Scattering hypervolume of spin-
ation trap, Phys. Rev. A 21, 1606 (1980). polarized fermions, Phys. Rev. A 104, 043319 (2021).
[11] S. Inouye, M. R. Andrews, J. Stenger, H.-J. Miesner, [31] Z. Wang and S. Tan, Scattering hypervolume of fermions
D. M. Stamper-Kurn, and W. Ketterle, Observation of in two dimensions, Phys. Rev. A 106, 023310 (2022).
feshbach resonances in a bose–einstein condensate, Na- [32] Z. Wang and S. Tan, Three-body scattering hypervolume
ture 392, 151 (1998). of identical fermions in one dimension, Phys. Rev. A 108,
[12] P. Courteille, R. S. Freeland, D. J. Heinzen, F. A. van 033306 (2023).
Abeelen, and B. J. Verhaar, Observation of a feshbach [33] J. Liang and S. Tan, Three-body scattering area for par-
resonance in cold atom scattering, Phys. Rev. Lett. 81, ticles with infinite or zero scattering length in two dimen-
69 (1998). sions, Phys. Rev. A 109, 063328 (2024).
[13] M. Greiner, O. Mandel, T. Esslinger, T. W. Hänsch, and [34] E. Nielsen, D. Fedorov, A. Jensen, and E. Garrido,
I. Bloch, Quantum phase transition from a superfluid to The three-body problem with short-range interactions,
a mott insulator in a gas of ultracold atoms, Nature 415, Physics Reports 347, 373 (2001).
39 (2002). [35] E. Braaten and H.-W. Hammer, Universality in few-body
[14] H. Feshbach, Unified theory of nuclear reactions, Annals systems with large scattering length, Physics Reports
of Physics 5, 357 (1958). 428, 259 (2006).
[15] U. Fano, Effects of configuration interaction on intensities [36] D. Varshalovich, D. Varshalovich, A. Moskalev, and
and phase shifts, Phys. Rev. 124, 1866 (1961). V. Khersonskiı̆, Quantum Theory of Angular Momentum:
[16] E. Tiesinga, B. J. Verhaar, and H. T. C. Stoof, Thresh- Irreducible Tensors, Spherical Harmonics, Vector Cou-
old and resonance phenomena in ultracold ground-state pling Coefficients, 3nj Symbols (World Scientific Pub.,
collisions, Phys. Rev. A 47, 4114 (1993). 1988).
[17] C. Chin, R. Grimm, P. Julienne, and E. Tiesinga, Fesh- [37] J. M. Blatt, On the neutron-proton force, Phys. Rev. 74,
bach resonances in ultracold gases, Rev. Mod. Phys. 82, 92 (1948).
1225 (2010). [38] H. A. Bethe, Theory of the effective range in nuclear
[18] C. A. Regal, M. Greiner, and D. S. Jin, Observation of scattering, Phys. Rev. 76, 38 (1949).
resonance condensation of fermionic atom pairs, Phys. [39] J. M. Blatt and J. D. Jackson, On the interpretation of
Rev. Lett. 92, 040403 (2004). neutron-proton scattering data by the schwinger varia-
[19] M. W. Zwierlein, C. A. Stan, C. H. Schunck, S. M. F. tional method, Phys. Rev. 76, 18 (1949).
Raupach, A. J. Kerman, and W. Ketterle, Condensation [40] H.-W. Hammer and D. Lee, Causality and the effective
of pairs of fermionic atoms near a feshbach resonance, range expansion, Annals of Physics 325, 2212 (2010).
Phys. Rev. Lett. 92, 120403 (2004). [41] S. Zhu and S. Tan, Three-body scattering hypervol-
[20] M. Bartenstein, A. Altmeyer, S. Riedl, S. Jochim, umes of particles with short-range interactions (2017),
C. Chin, J. H. Denschlag, and R. Grimm, Collective ex- arXiv:1710.04147 [cond-mat.quant-gas].
citations of a degenerate gas at the bec-bcs crossover, [42] A. Pricoupenko and D. S. Petrov, Three-body interaction
Phys. Rev. Lett. 92, 203201 (2004). near a narrow two-body zero crossing, Phys. Rev. A 100,
[21] T. Bourdel, L. Khaykovich, J. Cubizolles, J. Zhang, 042707 (2019).
F. Chevy, M. Teichmann, L. Tarruell, S. J. J. M. F. [43] B. D. Esry, C. H. Greene, and H. Suno, Threshold laws
Kokkelmans, and C. Salomon, Experimental study of the for three-body recombination, Phys. Rev. A 65, 010705
bec-bcs crossover region in lithium 6, Phys. Rev. Lett. 93, (2001).
050401 (2004). [44] J. P. D’Incao and B. D. Esry, Scattering length scaling
[22] Q. Chen, Z. Wang, R. Boyack, S. Yang, and K. Levin, laws for ultracold three-body collisions, Phys. Rev. Lett.
When superconductivity crosses over: From bcs to bec, 94, 213201 (2005).
Rev. Mod. Phys. 96, 025002 (2024). [45] M. Waseem, J. Yoshida, T. Saito, and T. Mukaiyama,
[23] H. Zhai and T.-L. Ho, Critical rotational frequency for Quantitative analysis of p-wave three-body losses via a
 20

cascade process, Phys. Rev. A 99, 052704 (2019). [47] Y. Ji, G. L. Schumacher, G. G. T. Assump ç ao, J. Chen,
[46] F. c. Top, Y. Margalit, and W. Ketterle, Spin-polarized J. T. Mäkinen, F. J. Vivanco, and N. Navon, Stability of
fermions with p-wave interactions, Phys. Rev. A 104, the repulsive fermi gas with contact interactions, Phys.
043311 (2021). Rev. Lett. 129, 203402 (2022).