betaFeedback

WorksPhysics and Astronomy

Scattering Faddeev calculations in the double continuum

Abstract

We use the configuration-space Faddeev formalism to study scattering of three particles in the double continuum where all particles are free. All scattering processes, starting from and ending in both single and double continua, are collected in a unique matrix. We apply our method to the benchmark system of neutron-deuteron scattering.

Introduction

The Faddeev formalism [1] describes the quantum mechanical scattering of three particles. It has been developed to circumvent the failure of the Lippmann-Schwinger equation to give a unique solution to the Schrödinger equation. When working in configuration-space, a unique solution is obtained after imposing proper boundary conditions on the wavefunction. For bound state calculations as well as the scattering of one particle on the bound cluster of the other two, those boundary conditions are well known. For situations where all three particles are free, a few years were needed to carefully derive those boundary conditions [2].

Configuration-space Faddeev scattering calculations in the double continuum are now routinely performed. Notably, we can quote calculations for the three helium atoms system at low energy [3] and importantly the scattering from a neutron or a proton on a deuteron which has since become a benchmark [4] system for a variety of different theoretical methods.

The difficulty in treating the double continuum correctly lie in the fact that the calculated wavefunction contains both the information about two-particle channels and three-body breakup channels. Disentangling those two behaviours from the wavefunction presents some difficulties since those channels are not orthogonal in the usual sense (we note however the elegant method presented in [5] where those two contributions are indeed orthogonalized).

Here, we report on a simple method to extract both the two-body and the three-body contributions from the wavefunction which stems from the fact that those two different contributions are well describe in two different coordinate systems: Jacobi coordinates for the two-body channels and polar coordinates for the three-body breakup channels. We then apply our methods to the neutron-deuteron scattering.

Faddeev formalism

In this work, we are interested in the scattering processes between three particles. Incoming and/or outgoing states of the three particles can be of two types. It can be a state where one of the particle is free with respect to a bound cluster of the other two particles. This state will be refered to as a (1+2)(1+2) or single continuum state. The other type of asymptotic state is one where all three particles are free with respect to each other. We will refer to that state as a (1+1+1)(1+1+1) or double continuum state.

The origin of the terms single or double continuum states is as follow. In a (1+2)(1+2) state, the energy between the free particle and the cluster is unrestricted and can take any value. It forms a continuum. In a (1+1+1)(1+1+1) state, the total energy is also unrestricted and forms a continuum. In that state though, there is another continuous degree of freedom which is how this total energy is partitioned between the internal coordinates of the particles. This partition forms another continuum.

For three particles, there are two internal coordinates which can be taken as the Jacobi vectors x\mathbf{x} and y\mathbf{y}. While the first continuum will be of course the total energy EE, the partition of this total energy among the Jacobi vectors can be parametrized by the ratio y/xy/x of the Jacobi vectors norms which will then constitutes our second continuum.

We use mass-scaled Jacobi coordinates {xi,yi}\{\mathbf{x}_i,\mathbf{y}_i\} and (i,j,k)(i,j,k) is a cyclic permutation of (1,2,3)(1,2,3) (we take ℏ=1\hbar=1 in the following)

xi=τxi(rj−rk)(1)\mathbf{x}_i=\tau_{x_i}\left(\mathbf{r}_j-\mathbf{r}_k\right) \tag*{(1)}
yi=τyi(ri−mjrj+mkrkmj+mk)(2)\mathbf{y}_i=\tau_{y_i}\left(\mathbf{r}_i-\frac{m_j\mathbf{r}_j+m_k\mathbf{r}_k}{m_j+m_k}\right) \tag*{(2)}

where τxi=2μjk\tau_{x_i}=\sqrt{2\mu_{jk}}, τyi=2μi,jk\tau_{y_i}=\sqrt{2\mu_{i,jk}}, μjk=mjmkmj+mk\mu_{jk}=\frac{m_jm_k}{m_j+m_k} and μi,jk=mi(mj+mk)mi+mj+mk\mu_{i,jk}=\frac{m_i(m_j+m_k)}{m_i+m_j+m_k}. The τ\tau factors absorb the masses so that the kinetic energy operator is T^i=−Δxi−Δyi\hat{T}_i=-\Delta_{x_i}-\Delta_{y_i}. We remark that

τxiτyi=2∏nmn∑nmn=2μ3B(3)\tau_{x_i}\tau_{y_i}=2\sqrt{\frac{\prod_n m_n}{\sum_n m_n}}=2\mu_{3B} \tag*{(3)}

where μ3B\mu_{3B} is the three-body reduced mass which is independent of the Jacobi arrangement (more generally, the NN-body reduced mass is μNBN−1=∏m∑m\mu_{NB}^{N-1}=\frac{\prod m}{\sum m}). The Jacobi vector xi\mathbf{x}_i is represented in spherical coordinates consisting of its norm xix_i and spherical angles x^i≡(θxi,φxi)\hat{x}_i\equiv(\theta_{x_i},\varphi_{x_i}). The same goes for yi\mathbf{y}_i. The 6D vector (xi,yi)(\mathbf{x}_i,\mathbf{y}_i) can therefore be written as (xi,yi,x^i,y^i)(x_i,y_i,\hat{x}_i,\hat{y}_i) corresponding to the partition R6=R2⊗R4\mathbb{R}^6=\mathbb{R}^2\otimes\mathbb{R}^4 into two radii and four angles. We introduce the polar coordinates1 (ρ,αi,x^i,y^i)(\rho,\alpha_i,\hat{x}_i,\hat{y}_i) as

ρ2=xi2+yi2(4)\rho^2=x_i^2+y_i^2 \tag*{(4)}
αi=arctan⁡(yi/xi)(5)\alpha_i=\arctan(y_i/x_i) \tag*{(5)}

In the following, we might refer to Jacobi coordinates as cartesian for obvious reasons. Note that the hyperradius ρ\rho is independent of the Jacobi arrangement.

The total wavefunction Ψ\Psi is the solution of the Schrödinger equation (T^+V^−E)Ψ=0(\hat{T}+\hat{V}-E)\Psi=0. The zero for the center of mass energy EE is taken to be at breakup. Therefore, above-breakup scattering means that E≥0E\geq0. The potential term is taken as a sum of pairwise, central potentials i.e. V=V1(x1)+V2(x2)+V3(x3)V=V_1(x_1)+V_2(x_2)+V_3(x_3). The total wavefunction Ψ\Psi is decomposed into three Faddeev components ϕi\phi_i satisfying three coupled equations

Ψ(xi,yi)=ϕ1(x1,y1)+ϕ2(x2,y2)+ϕ3(x3,y3)(6)\Psi(\mathbf{x}_i,\mathbf{y}_i)=\phi_1(\mathbf{x}_1,\mathbf{y}_1)+\phi_2(\mathbf{x}_2,\mathbf{y}_2)+\phi_3(\mathbf{x}_3,\mathbf{y}_3) \tag*{(6)}
(T^i+V^i−E)ϕi(xi,yi)=−V^i[ϕj(xj,yj)+ϕk(xk,yk)](7)\left(\hat{T}_i+\hat{V}_i-E\right)\phi_i(\mathbf{x}_i,\mathbf{y}_i)=-\hat{V}_i\left[\phi_j(\mathbf{x}_j,\mathbf{y}_j)+\phi_k(\mathbf{x}_k,\mathbf{y}_k)\right] \tag*{(7)}

The 4D angular part is solved through partial waves expansion with respect to bipolar spherical harmonics Ya(x^i,y^i)\mathcal{Y}_a(\hat{x}_i,\hat{y}_i) defining radial wavefunctions fi,a(xi,yi)f_{i,a}(x_i,y_i) for component ii in partial wave aa as

ϕi(xi,yi)=∑afi,a(xi,yi)xiyiYa(x^i,y^i)(8)\phi_i(\mathbf{x}_i,\mathbf{y}_i)=\sum_a\frac{f_{i,a}(x_i,y_i)}{x_i y_i}\mathcal{Y}_a(\hat{x}_i,\hat{y}_i) \tag*{(8)}
Ya(x^i,y^i)=[Yℓa(x^i)⊗Yλa(y^i)]LM(9)\mathcal{Y}_a(\hat{x}_i,\hat{y}_i)=\left[Y_{\ell_a}(\hat{x}_i)\otimes Y_{\lambda_a}(\hat{y}_i)\right]^{LM} \tag*{(9)}
ℓ^xi 2Ya(x^i,y^i)=ℓa(ℓa+1)Ya(x^i,y^i)(10)\hat{\ell}_{x_i}^{\,2}\mathcal{Y}_a(\hat{x}_i,\hat{y}_i)=\ell_a(\ell_a+1)\mathcal{Y}_a(\hat{x}_i,\hat{y}_i) \tag*{(10)}
ℓ^yi 2Ya(x^i,y^i)=λa(λa+1)Ya(x^i,y^i)(11)\hat{\ell}_{y_i}^{\,2}\mathcal{Y}_a(\hat{x}_i,\hat{y}_i)=\lambda_a(\lambda_a+1)\mathcal{Y}_a(\hat{x}_i,\hat{y}_i) \tag*{(11)}

For simplicity, we consider partial waves consisting of only orbital angular momentum channels i.e. bipolar spherical harmonics. In general, they can also include spin angular momentum. This will be illustrated later when we will apply our formalism to neutron-deuteron scattering.

The total wavefunction Ψ\Psi can itself be expanded in partial waves as

Ψ(xi,yi)=∑aua(xi,yi)xiyiYa(x^i,y^i)(12)\Psi(\mathbf{x}_i,\mathbf{y}_i)=\sum_a\frac{u_a(x_i,y_i)}{x_i y_i}\mathcal{Y}_a(\hat{x}_i,\hat{y}_i) \tag*{(12)}

defining radial wavefunction ua(xi,yi)u_a(x_i,y_i). Inserting those partial waves expansions in equation (6) and use the orthogonality relation for the bipolar spherical harmonics allows to get an expression for the total wavefunction radial part uau_a in terms of the component radial parts fi,af_{i,a}

ua(xi,yi)=fi,a(xi,yi)+∑b∬dx^idy^ixiyixjyjYa∗(x^i,y^i)fj,b(xj,yj)Yb(x^j,y^j)+∑c∬dx^idy^ixiyixkykYa∗(x^i,y^i)fk,c(xk,yk)Yc(x^k,y^k)(13)\begin{aligned} u_a(x_i,y_i)={}&f_{i,a}(x_i,y_i)+\sum_b\iint d\hat{x}_i d\hat{y}_i\frac{x_i y_i}{x_j y_j}\mathcal{Y}_a^*(\hat{x}_i,\hat{y}_i)f_{j,b}(x_j,y_j)\mathcal{Y}_b(\hat{x}_j,\hat{y}_j)\\ &+\sum_c\iint d\hat{x}_i d\hat{y}_i\frac{x_i y_i}{x_k y_k}\mathcal{Y}_a^*(\hat{x}_i,\hat{y}_i)f_{k,c}(x_k,y_k)\mathcal{Y}_c(\hat{x}_k,\hat{y}_k) \tag*{(13)} \end{aligned}
ua(xi,yi)≡fi,a(xi,yi)+∑bJijab[fj,b](xi,yi)+∑cJijac[fk,c](xi,yi)(14)u_a(x_i,y_i)\equiv f_{i,a}(x_i,y_i)+\sum_b\mathcal{J}_{ij}^{ab}[f_{j,b}](x_i,y_i)+\sum_c\mathcal{J}_{ij}^{ac}[f_{k,c}](x_i,y_i) \tag*{(14)}

where we have introduced the integral transform Jijab\mathcal{J}_{ij}^{ab} of a generic function hh

Jijab[h]=∬dx^idy^ixiyixjyjYa∗(x^i,y^i)h(xj,yj)Yb(x^j,y^j)(15)\mathcal{J}_{ij}^{ab}[h]=\iint\mathrm{d}\hat{x}_i\mathrm{d}\hat{y}_i\frac{x_i y_i}{x_j y_j}Y_a^*(\hat{x}_i,\hat{y}_i)h(x_j,y_j)Y_b(\hat{x}_j,\hat{y}_j) \tag*{(15)}

In equation (14), this integral transform takes in a radial wavefunction for component jj in a partial wave bb and outputs a radial wavefunction for component ii in partial wave aa. We refer to it as Jacobi transform 2. It plays a central role in the Faddeev formalism since it is this term which couples the equations for the different components. It is also the tool which allows to sum up the contribution from the different Faddeev components to get physical observables. We note that in polar coordinates, it is independent of the hyperradius ρ\rho: given a function of αj\alpha_j, it then outputs a function of αi\alpha_i.

Inserting the partial wave expansion in equation (7) leads to 2D coupled integro-differential equations for the radial wavefunctions fi,af_{i,a}

(ti,a−E)fi,a(xi,yi)+∑bvi,abfi,b(xi,yi)=−∑cvi,ac∑j≠iJijcb[fj,b](xi,yi)(16)(t_{i,a}-E)f_{i,a}(x_i,y_i)+\sum_b v_{i,ab}f_{i,b}(x_i,y_i)=-\sum_c v_{i,ac}\sum_{j\ne i}\mathcal{J}_{ij}^{cb}[f_{j,b}](x_i,y_i) \tag*{(16)}

where ti,a=−∂2∂xi2−∂2∂yi2+ℓa(ℓa+1)xi2+λa(λa+1)yi2t_{i,a}=-\frac{\partial^2}{\partial x_i^2}-\frac{\partial^2}{\partial y_i^2}+\frac{\ell_a(\ell_a+1)}{x_i^2}+\frac{\lambda_a(\lambda_a+1)}{y_i^2} and

vi,ab=Jiiab[Vi]=∬dx^idy^iYa∗(x^i,y^i)Vi(xi,yi)Yb(x^i,y^i)(17)v_{i,ab}=\mathcal{J}_{ii}^{ab}[V_i]=\iint\mathrm{d}\hat{x}_i\mathrm{d}\hat{y}_iY_a^*(\hat{x}_i,\hat{y}_i)V_i(x_i,y_i)Y_b(\hat{x}_i,\hat{y}_i) \tag*{(17)}

For central isotropic pairwise potential Vi(xi)V_i(x_i), we obviously have vi,ab=Vi(xi)δabv_{i,ab}=V_i(x_i)\delta_{ab}.

To solve numerically the equation (16), the partial Faddeev components fi,af_{i,a} are expanded onto a basis of cubic Hermite splines ss

fi,a(xi,yi)=∑n,manmi,asn(xi)sm(yi)(18)f_{i,a}(x_i,y_i)=\sum_{n,m}a_{nm}^{i,a}s_n(x_i)s_m(y_i) \tag*{(18)}

The numerical solution of equation (16) consists in finding the expansion coefficients anmi,aa_{nm}^{i,a}. Specifying proper grids in xix_i and yiy_i and use an orthogonal collocation method leads to the creation of the different operators appearing in equation (16): the kinetic energy operator T\mathbb{T}, an indicator operator 1\mathbb{1}, a potential operator V\mathbb{V} and a Jacobi kernel operator K\mathbb{K}. While the kinetic and potential operators are self-explanatory, the indicator operator 1\mathbb{1} simply transforms from expansion coefficients to values at the collocation points (it accounts for the energy EE in equation (16)) and the kernel operator K\mathbb{K} couples the different equations and therefore implements the Jacobi transform. Explicit expressions for the matrix elements of those operators can be found in [6].

A solution of equation (16) involves the resolution of the inhomogeneous linear system (T+V+K−E1)v=−Kχ(\mathbb{T}+\mathbb{V}+\mathbb{K}-E\mathbb{1})v=-\mathbb{K}\chi where the vector χ\chi contains the information about the incoming state as well as enforces the boundary conditions. The solution vector vv contains the expansion coefficients anmi,aa_{nm}^{i,a}. We will go into more details about the numerical methods we use in the following sections.

Asymptotic channels functions

(1+2) asymptotic channels

Below breakup, at E<0E<0, the only scattering processes possible are (1+2)→(1+2)(1+2)\to(1+2). The asymptotic region consists of yi→∞y_i\to\infty while the coordinate xix_i stays bounded by the potential ViV_i. The uncoupled differential equation for the Faddeev component fif_i is

(−∂2∂xi2−∂2∂yi2+ℓ(ℓ+1)xi2+λ(λ+1)yi2+Vi(xi)−E)fi(xi,yi)=0(19)\left(-\frac{\partial^2}{\partial x_i^2}-\frac{\partial^2}{\partial y_i^2}+\frac{\ell(\ell+1)}{x_i^2}+\frac{\lambda(\lambda+1)}{y_i^2}+V_i(x_i)-E\right)f_i(x_i,y_i)=0 \tag*{(19)}

This equation is separable and a general solution is

fi(xi,yi)=∑νcνφν(xi)h(+)(qyi)(20)f_i(x_i,y_i)=\sum_\nu c_\nu\varphi_\nu(x_i)h^{(+)}(q y_i) \tag*{(20)}

where cvc_v are complex coefficients, φv(xi)\varphi_v(x_i) is a bound rovibrational wavefunction for the potential Vi(xi)V_i(x_i) with vibrational number vv, angular momentum ℓ\ell and energy ϵv,ℓ<0\epsilon_{v,\ell}<0 and h(+)(qyi)h^{(+)}(qy_i) is an outgoing Ricatti-Hankel function with momentum qq and angular momentum λ\lambda. Conservation of energy gives

E=ϵv,ℓ+q22μi,jk(21)E=\epsilon_{v,\ell}+\frac{q^2}{2\mu_{i,jk}} \tag*{(21)}

The sum in equation (17) runs over vibrational numbers for which q2>0q^2>0 i.e. over the open channels: indeed, only for real qq does the function h(+)(qyi)h^{(+)}(qy_i) carry probability flux at infinity and will then contribute to the scattering process. The asymptotic channels functions for the (1+2)(1+2) scattering processes are then the functions φv(xi)h(+)(qyi)\varphi_v(x_i)h^{(+)}(qy_i) which we will refer to as ”bound plane waves”. The degrees of liberty in equation (17) are the coefficients cvc_v which will become the relevant scattering matrix elements once the calculated Faddeev component is matched to its asymptotic form.

(1+1+1) asymptotic channels

Above breakup, at E>0E>0, the scattering process (1+2)→(1+1+1)(1+2)\to(1+1+1) i.e. breakup, becomes energetically possible. The asymptotic region now consists of both xi→∞x_i\to\infty and yi→∞y_i\to\infty. It now explores a region where Vi(xi)→0V_i(x_i)\to0 so that the potential can be neglected in the asymptotically uncoupled differential equation for fif_i. This uncoupled differential equation is best solved in polar coordinates and reads

(−∂2∂ρ2−1ρ∂∂ρ−1ρ2∂2∂αi2+ℓ(ℓ+1)ρ2cos⁡2αi+λ(λ+1)ρ2sin⁡2αi−E)fi(ρ,αi)=0(22)\left(-\frac{\partial^2}{\partial\rho^2}-\frac{1}{\rho}\frac{\partial}{\partial\rho}-\frac{1}{\rho^2}\frac{\partial^2}{\partial\alpha_i^2}+\frac{\ell(\ell+1)}{\rho^2\cos^2\alpha_i}+\frac{\lambda(\lambda+1)}{\rho^2\sin^2\alpha_i}-E\right)f_i(\rho,\alpha_i)=0 \tag*{(22)}

Separation of variables posing fi(ρ,αi)=p(ρ)q(αi)f_i(\rho,\alpha_i)=p(\rho)q(\alpha_i) leads to

(−ρ2∂2∂ρ2−ρ∂∂ρ−ρ2E)p(ρ)=−ν2p(ρ)(23)\left(-\rho^2\frac{\partial^2}{\partial\rho^2}-\rho\frac{\partial}{\partial\rho}-\rho^2E\right)p(\rho)=-\nu^2p(\rho) \tag*{(23)}
(−∂2∂αi2+ℓ(ℓ+1)cos⁡2αi+λ(λ+1)sin⁡2αi)q(αi)=ν2q(αi)(24)\left(-\frac{\partial^2}{\partial\alpha_i^2}+\frac{\ell(\ell+1)}{\cos^2\alpha_i}+\frac{\lambda(\lambda+1)}{\sin^2\alpha_i}\right)q(\alpha_i)=\nu^2q(\alpha_i) \tag*{(24)}

The equation (20a) is the standard Bessel differential equation and an equation closely related to (20b) appears in an early paper from Delves [7]. The separation constant ν2\nu^2, eigenvalue of the differential operator −∂2∂αi2+ℓ(ℓ+1)cos⁡2αi+λ(λ+1)sin⁡2αi-\frac{\partial^2}{\partial\alpha_i^2}+\frac{\ell(\ell+1)}{\cos^2\alpha_i}+\frac{\lambda(\lambda+1)}{\sin^2\alpha_i}, takes values ν=ℓ+λ+2(n+1)\nu=\ell+\lambda+2(n+1) with n=0,1,2...n=0,1,2... while the functions q(αi)q(\alpha_i) are the so-called Delves functions Dν(ℓ,λ)D_\nu^{(\ell,\lambda)}

q(αi)=Dν(ℓ,λ)(αi)∝cos⁡ℓ+1(αi)sin⁡λ+1(αi)Pn(λ+1/2,ℓ+1/2)(cos⁡2αi)(25)q(\alpha_i)=D_\nu^{(\ell,\lambda)}(\alpha_i)\propto\cos^{\ell+1}(\alpha_i)\sin^{\lambda+1}(\alpha_i)P_n^{(\lambda+1/2,\ell+1/2)}(\cos2\alpha_i) \tag*{(25)}

where Pn(a,b)P_n^{(a,b)} is a Jacobi polynomial. We note that the Delves functions are eigenfunctions of the Jacobi transform with eigenvalues β\beta i.e.

Jijaa[Dν(ℓa,λa)]=βijaaDν(ℓa,λa)(26)\mathcal{J}_{ij}^{aa}\left[D_\nu^{(\ell_a,\lambda_a)}\right]=\beta_{ij}^{aa}D_\nu^{(\ell_a,\lambda_a)} \tag*{(26)}

More generally, the Jacobi transform Jijab\mathcal{J}_{ij}^{ab} of a Delves function Dν(ℓb,λb)D_\nu^{(\ell_b,\lambda_b)} is proportional to a Delves function Dν′(ℓa,λa)D_{\nu'}^{(\ell_a,\lambda_a)}. A general solution of equation (19) is then

fi(ρ,αi)=∑νcνDν(ℓ,λ)(αi)Hν(+)(kρ)(27)f_i(\rho,\alpha_i)=\sum_\nu c_\nu D_\nu^{(\ell,\lambda)}(\alpha_i)H_\nu^{(+)}(k\rho) \tag*{(27)}

with k=Ek=\sqrt{E} and Hν(+)H_\nu^{(+)} an outgoing Hankel function of order ν\nu. At this point, we can perform an additional manipulation on the functions in equation (23): unlike the previous channels functions in equation (17), the above sum involves outgoing Hankel functions of a unique argument kρk\rho. We can therefore use the asymptotic behavior of the Bessel functions to write everything in terms of, for example, H0(+)H_0^{(+)}

fi(ρ,αi)⟶ρ→∞(∑νcνe−iνπ2Dν(ℓ,λ)(αi))H0(+)(kρ)≡A(αi)H0(+)(kρ)(28)\begin{aligned} f_i(\rho,\alpha_i)\underset{\rho\to\infty}{\longrightarrow}\left(\sum_\nu c_\nu e^{-i\frac{\nu\pi}{2}}D_\nu^{(\ell,\lambda)}(\alpha_i)\right)H_0^{(+)}(k\rho) \\ &\equiv A(\alpha_i)H_0^{(+)}(k\rho) \tag*{(28)} \end{aligned}

In equation (23), the degrees of liberty are the coefficients cνc_\nu while in equation (24) the only degree of liberty is the unknown function A(αi)A(\alpha_i). Equations (23) and (24) are two different interpretations of the fact that a (1+1+1)(1+1+1) asymptotic channel lie in a double continuum. It must then be described either by an (countably) infinite number of scalar scattering matrix elements or by a single function of the continuous variable αi\alpha_i.

Here, we adopt the point of view of equation (24) : the asymptotic channels functions for the (1+1+1)(1+1+1) scattering processes are the functions H0(+)(kρ)H_0^{(+)}(k\rho) which we will refer to as ”cylindrical waves”. The matching of the calculated Faddeev component to its asymptotic form will lead to scattering matrix elements which will be the breakup amplitudes, functions of the variable αi\alpha_i.

Reaching the asymptotic region

Reaching the asymptotic region does not occur on the same footing for (1+2)(1+2) single continuum channels and the (1+1+1)(1+1+1) double continuum channels. For this discussion, we assume that the various pairwise potentials Vi(xi)V_i(x_i) have a finite range bb (this range can be, for instance, the region for which the least bounded two-body wavefunction have significant values) and are short-range (thus we exclude the Coulomb potential which require a particular treatment). We note that the coupling term in the right-hand side of equation (13) involves the potential Vi(xi)V_i(x_i) and the other partial Faddeev components fj(xj,yj)f_j(x_j,y_j) and fk(xk,yk)f_k(x_k,y_k).

The single continuum channel (1+2)(1+2) asymptotic region for the partial component fi(xi,yi)f_i(x_i,y_i) corresponds to xi<bx_i<b and yi→∞y_i\to\infty. In this limit, it is easy to see that both xjx_j and xkx_k tends towards yiy_i. Therefore, both partial components fjf_j and fkf_k will vanish as they will involve two-body wavefunctions in the potential VjV_j and VkV_k evaluated at xj,xk→∞x_j,x_k\to\infty. Therefore, the (1+2)(1+2) asymptotic region is reached for yi≫by_i\gg b. We note that all the Jacobi transforms therefore vanish in equation (11) and that the Faddeev component actually becomes the total wavefunction ϕi(xi,yi)→Ψ(xi,yi)\phi_i(x_i,y_i)\to\Psi(x_i,y_i).

The situation is quite different for the double continuum channel (1+1+1)(1+1+1). Here, all components ff have non-negligible values everywhere in the (x,y)(x,y) plane. It is not obvious at first whether the coupling term even vanishes. A careful analysis [8] shows that the coupling term in this case actually behaves as 1/y3/21/y^{3/2} for y→∞y\to\infty and xx finite. So, the asymptotic region can be eventually reached but much slower than in the single continuum case as it results from interferences between the three Faddeev partial components. Additionally, for a given ymax⁡y_{\max} we can define a critical polar angle αc=arctan⁡(byi/ymax⁡)\alpha_c=\arctan(b y_i/y_{\max}). The effect of the potential will have a lingering effect on the breakup amplitudes for αc<α<π/2\alpha_c<\alpha<\pi/2. This is the region where two of the particles stay close to each other for a very long time. It originates from the fact that very little of the total energy EE has been partitioned in the xx degree of freedom. Only for ymax⁡→∞y_{\max}\to\infty does αc→π/2\alpha_c\to\pi/2. For this reason, calculations done in a finite box {0<x<xmax⁡,0<y<ymax⁡}\{0<x<x_{\max},0<y<y_{\max}\} will usually have breakup amplitudes imprecise near α≈π/2\alpha\approx\pi/2 and appropriate methods [5] are needed to actually retrieve the breakup amplitude at infinity.

Structure of the scattering matrix above breakup

At E>0E>0, the (1+1+1)(1+1+1) channel is open so that those scattering processes are possible

(1+2)→(1+2)elastic (1+2)(1+2)→(1+1+1)breakup(1+1+1)→(1+2)3B recombination(1+1+1)→(1+1+1)elastic (1+1+1)\begin{aligned} (1+2)&\to(1+2) && \text{elastic }(1+2) \\ (1+2)&\to(1+1+1) && \text{breakup} \\ (1+1+1)&\to(1+2) && \text{3B recombination} \\ (1+1+1)&\to(1+1+1) && \text{elastic }(1+1+1) \end{aligned}

In practice, we rather calculate the transition matrix T\mathbf{T}. Schematically, and of we collect all the nn (1+2)(1+2) channels in a first block and all the pp (1+1+1)(1+1+1) channels in a second block, solving the scattering problem amounts at finding a set of N=n+pN=n+p (the number of open channels) linearly independent solutions (f1,f2,…,fN)(f^1,f^2,\ldots,f^N) of the coupled Faddeev equations which behaves asymptotically as

(f1⋮fnfn+1⋮fN)≈(10⋱1i1(α)0⋱ip(α))(φ1(x) j(q1y)J0(kρ)⋮φn(x) j(qny)J0(kρ)1J0(kρ)⋮1J0(kρ))+(T11T12(α)T21T22(α))(φ1(x)h(+)(q1y)⋮φn(x)h(+)(qny)H0(+)(kρ)⋮H0(+)(kρ))(29)\begin{pmatrix} f^1 \\ \vdots\\ f^n \\ \hline f^{n+1} \\ \vdots\\ f^N \end{pmatrix} \approx \begin{pmatrix} 1 & & & 0 \\ & \ddots& & \\ & & 1 & \\ \hline & & & i_1(\alpha) \\ & & & \\ 0 & & & \ddots\\ & & & i_p(\alpha) \end{pmatrix} \begin{pmatrix} \dfrac{\varphi_1(x)\,j(q_1y)}{J_0(k\rho)} \\ \vdots\\ \dfrac{\varphi_n(x)\,j(q_ny)}{J_0(k\rho)} \\ \hline \dfrac{1}{J_0(k\rho)} \\ \vdots\\ \dfrac{1}{J_0(k\rho)} \end{pmatrix} + \begin{pmatrix} T_{11} & T_{12}(\alpha) \\ \hline T_{21} & T_{22}(\alpha) \end{pmatrix} \begin{pmatrix} \varphi_1(x)h^{(+)}(q_1y) \\ \vdots\\ \varphi_n(x)h^{(+)}(q_ny) \\ \hline H_0^{(+)}(k\rho) \\ \vdots\\ H_0^{(+)}(k\rho) \tag*{(29)} \end{pmatrix}

The functions i1(α),…,ip(α)i_1(\alpha),\ldots,i_p(\alpha) characterize the incoming (1+1+1)(1+1+1) channels and the functions jj and J0J_0 are regular (Ricatti)-Bessel functions. The blocks T11T_{11} (elastic) and T21T_{21} (3B recombination) contains scalar complex numbers while the blocks T12T_{12} (breakup) and T22T_{22} (elastic) consist of complex functions of the polar angle α\alpha. Transition and scattering matrices are related via S=1+2iTS=1+2iT. We will illustrate in the next section how to retrieve the usual unitarity and reciprocity properties from the scattering matrix SS.

Application : the neutron-deuteron scattering

We apply the Faddeev formalism to the neutron (nn) scattering on the deuteron (dd). This system has become a benchmark test for various numerical methods [4]. The different scattering processes we will describe are

n+d→n+delastic ndn+d→n+n+pbreakupn+n+p→n+d3B recombinationn+n+p→n+n+pelastic nnp(30)\begin{aligned} n+d &\to n+d && \text{elastic } nd \\ n+d &\to n+n+p && \text{breakup} \\ n+n+p &\to n+d && \text{3B recombination} \\ n+n+p &\to n+n+p && \text{elastic } nnp \tag*{(30)} \end{aligned}

We adopt the same level of approximation as in the benchmarks and consider a system of three identical nucleons. Proton (p) and neutron only differ by their isospin state. Therefore, there is a unique Faddeev component ϕ\phi to be calculated. The nucleon mass mm is set to ℏ2/m=41.47 MeV fm2\hbar^2/m=41.47\ \mathrm{MeV\ fm^2}. We calculate for the doublet system JΠ=1/2+J^\Pi=1/2^+ and consider the following two channels written in a LSLS coupling scheme:

channel aa(ℓa,λa)L(\ell_a,\lambda_a)L(sa,σa)S(s_a,\sigma_a)Stat_aname
1(0,0)0(0,0)0(0,12)12(0,\frac{1}{2})\frac{1}{2}1singlet
2(0,0)0(0,0)0(1,12)12(1,\frac{1}{2})\frac{1}{2}0triplet

Table I. The two channels for the doublet (JΠ=1/2+J^\Pi=1/2^+) system.

In those channels, sas_a (tat_a) refer to the (iso)spin of a pair while σa\sigma_a is the spin of the spectator. In the evaluation of the kernel operator which acts between different Jacobi arrangements, recoupling coefficients of the form ⟨(sjsk)sjk,si,S∣(sisk)sik,sj,S⟩\langle(s_js_k)s_{jk},s_i,S|(s_is_k)s_{ik},s_j,S\rangle have to be considered for both the spins and the isospins. Standard expressions for those recoupling coefficients in terms of 6-jj symbols [9] lead to the following spin wsw_s and isospin wtw_t recoupling matrices in the basis of the two channels above

ws=(−12−3232−12)(31)w_s=\begin{pmatrix} -\frac{1}{2} & -\frac{\sqrt{3}}{2}\\ \frac{\sqrt{3}}{2} & -\frac{1}{2} \end{pmatrix} \tag*{(31)}
wt=(−1232−32−12)(32)w_t=\begin{pmatrix} -\frac{1}{2} & \frac{\sqrt{3}}{2}\\ -\frac{\sqrt{3}}{2} & -\frac{1}{2} \end{pmatrix} \tag*{(32)}

The total recoupling matrix ww is then w=ws⊙wtw=w_s\odot w_t:

w=(14−34−3414)(33)w=\begin{pmatrix} \frac{1}{4} & -\frac{3}{4}\\ -\frac{3}{4} & \frac{1}{4} \end{pmatrix} \tag*{(33)}

where ⊙\odot is the Hadamard (element-wise) product. In practice, once a kernel operator kk has been evaluated with respect to the orbital angular momentum channel (ℓa,λa)L=(0,0)0(\ell_a,\lambda_a)L=(0,0)0 the full operator in the basis of the channels in table (I) is K=w⊗kK=w\otimes k where ⊗\otimes is the Kronecker product.

The pairwise potentials are taken to be sums of Yukawa potentials [4]

V1(x)=1438.72e−3.11xx−513.968e−1.55xx(34)V_1(x)=1438.72\frac{e^{-3.11x}}{x}-513.968\frac{e^{-1.55x}}{x} \tag*{(34)}
V2(x)=1438.72e−3.11xx−626.885e−1.55xx(35)V_2(x)=1438.72\frac{e^{-3.11x}}{x}-626.885\frac{e^{-1.55x}}{x} \tag*{(35)}

with distances expressed in fm and energies in MeV. With these parameters, the deuteron is the bound state of the triplet channel potential V2V_2 and has an energy of Ed=−2.2306E_d = -2.2306 MeV. To make the connection with equation (29), our scattering matrix S\mathbf{S} will be at most of dimension 3×33 \times3 consisting of the unique triplet (1+2)(1 + 2) asymptotic channel and the singlet and triplet (1+1+1)(1 + 1 + 1) channels:

S=(S11S12(α)S13(α)S21S22(α)S23(α)S31S32(α)S33(α))(36)\mathbf{S} = \begin{pmatrix} S_{11} & S_{12}(\alpha) & S_{13}(\alpha) \\ S_{21} & S_{22}(\alpha) & S_{23}(\alpha) \\ S_{31} & S_{32}(\alpha) & S_{33}(\alpha) \end{pmatrix} \tag*{(36)}

Obviously, for E<0E < 0 only the single continuum channel is open and the scattering matrix reduces to S=(S11)\mathbf{S} = (S_{11}).

Finally the off-diagonal blocks of the scattering matrix above relate processes where the reduced mass changes between incoming and outgoing states. Those matrix elements have to be renormalized in order to be able to properly compare incoming and outgoing probability fluxes. It is quite clear that the reduced mass which enters the probability flux for a single continuum (1+2)(1 + 2) channel is μi,jk\mu_{i,jk}, the reduced mass in the yy direction. The reduced mass to be used in a double continuum (1+1+1)(1 + 1 + 1) channel is given by the Jacobian of the transformation between cartesian and polar coordinates [10]. This Jacobian involves the quantity μjkμi,jk\sqrt{\mu_{jk}\mu_{i,jk}} which is precisely the three-body reduced mass μ3B\mu_{3B}. Therefore, we define a reduced mass matrix μ\boldsymbol{\mu} as

μ=(1μi,jk/μ3Bμi,jk/μ3Bμ3B/μi,jk11μ3B/μi,jk11)(37)\boldsymbol{\mu} = \begin{pmatrix} 1 & \sqrt{\mu_{i,jk}/\mu_{3B}} & \sqrt{\mu_{i,jk}/\mu_{3B}} \\ \sqrt{\mu_{3B}/\mu_{i,jk}} & 1 & 1 \\ \sqrt{\mu_{3B}/\mu_{i,jk}} & 1 & 1 \end{pmatrix} \tag*{(37)}

and we renormalized the scattering matrix as S→μ⊙S\mathbf{S} \to\boldsymbol{\mu} \odot\mathbf{S}.

Numerical methods

Calculations can be done either in cartesian or polar coordinates. Both methods have their advantages and drawbacks. Cartesian grids are well adapted to describe the bound plane wave part of the Faddeev components but the kernel operator acts on both xx and yy grids. Polar grids are obviously adapted to describe the cylindrical wave part of the components and the fact that the kernel operator only act on the α\alpha grid simplify its construction. While our previous work [6] was done in cartesian coordinates, here we work in polar coordinates. The grid in ρ\rho is chosen to cover distances of the order of τx−1≈τy−1≈100\tau_x^{-1} \approx\tau_y^{-1} \approx100 fm. It runs up to ρτxτy=140\frac{\rho}{\sqrt{\tau_x\tau_y}} = 140 fm with nρ=500n_\rho= 500 collocation points. When doing calculations in polar coordinates, care must be taken to the grid in α\alpha: a high density of points is needed around α≈π/2\alpha\approx\pi/2 so that the grid correctly sample the deuteron bound state wavefunction. We use the method described in [11] to construct an optimized grid in α\alpha with nα=256n_\alpha= 256 collocation points. Both grids in ρ\rho and α\alpha are non uniform with more points at small ρ\rho and α\alpha near 0 and π/2\pi/2. We have verified that increasing the number of points in both grids did not change significantly our results anymore.

The boundary condition imposed on the α\alpha grid is that the partial radial Faddeev component must vanish at α=0\alpha= 0 and α=π/2\alpha= \pi/2. Operators are first calculated without imposing a boundary condition in ρ\rho. The boundary condition in ρ\rho is imposed in each open channel by replacing the last two cubic spline basis functions by an appropriate linear combination so as to impose a particular logarithmic derivative.

Below breakup, at E<0E < 0, the imposed boundary condition consists of that of an outgoing bound plane wave (see equation (20)). This leads to an inhomogeneous boundary condition where the logarithmic derivative in ρ\rho depends on α\alpha. Above breakup, we impose a boundary condition of an outgoing Hankel function (see equation (28)). This boundary condition does not depend on α\alpha. In this case, we note however that this boundary condition is not adapted to the bound plane wave part of the partial Faddeev component: while the calculated component will consist mostly of an outgoing Hankel function H0(+)(kρ)H_0^{(+)}(k\rho), the bound plane wave part of it will be a superposition of regular and outgoing functions.

A solution is obtained by solving the following linear system

(T+V+K−E1)v=−Kχfor a (1+2) incoming state(38)(\mathcal{T} + \mathcal{V} + \mathcal{K} - E1)v = -\mathcal{K}\chi \qquad\text{for a }(1 + 2)\text{ incoming state} \tag*{(38)}
(T+V+K−E1)v=−(V+K)χfor a (1+1+1) incoming state(39)(\mathcal{T} + \mathcal{V} + \mathcal{K} - E1)v = -(\mathcal{V} + \mathcal{K})\chi \qquad\text{for a }(1 + 1 + 1)\text{ incoming state} \tag*{(39)}

where the vector χ\chi contains expansion coefficients of the appropriate incoming state. As shown in equation (29) this incoming state is either a bound plane wave (a solution of (T+V−E1)(\mathcal{T} + \mathcal{V} - E1)) or a cylindrical wave i(α)H0(+)(kρ)i(\alpha)H_0^{(+)}(k\rho) which is a solution of (T−E1)(\mathcal{T} - E1). The solution vector vv contains the expansion coefficients of the partial Faddeev component f(ρ,α)f(\rho,\alpha). This Faddeev component consists further of two parts fa(ρ,α)f_a(\rho,\alpha) with a=1,2a=1,2 for the two singlet and triplet channels.

We show in Figure 1 typical partial Faddeev components below breakup. Components are shown as a function of unscaled Jacobi distances τx−1x\tau_x^{-1}x and τy−1y\tau_y^{-1}y expressed in fm. At negative energies, the singlet channel is closed. The corresponding partial component is non-negligible only in the inner region where both channels are coupled. It vanishes exponentially elsewhere. The triplet component is seen to behave as a bound plane wave asymptotically, in the region xx finite and y→∞y \to\infty.

Calculated partial Faddeev components for nd scattering below breakup

Figure 1. Calculated partial Faddeev components for ndnd scattering, below breakup, at E=−1E=-1 MeV.

We show in Figure 2 typical partial Faddeev components above breakup. We have chosen E≈7.17E \approx7.17 MeV which corresponds to an energy in the laboratory frame Elab=14.1E_{\mathrm{lab}}=14.1 MeV which is one of the energy for which benchmark data are quoted in [4]. The two energies are related by

Calculated partial Faddeev components for nd scattering above breakup

Figure 2. Calculated partial Faddeev components for ndnd scattering, above breakup, at E=7.17E=7.17 MeV. This energy corresponds to Elab=14.1E_{\mathrm{lab}}=14.1 MeV from [4].

E=Ed+23Elab(40)E = E_d + \frac{2}{3}E_{\mathrm{lab}} \tag*{(40)}

We can clearly see that the singlet component behaves as a cylindrical wave asymptotically. The triplet component have both a bound plane wave and a cylindrical wave behaviour.

In order to extract the bound plane wave part from the partial component fa(ρ,α)f_a(\rho,\alpha), this function is resampled as fa(x,y)f_a(x,y) on a cartesian grid inscribed in the polar grid. We schematically illustrate this process in Figure 3.

Figure 3. Schematics of the resampling process. The Faddeev components are calculated on a polar grid (black grid) as f(ρ,α)f(\rho,\alpha). They are resampled on a cartesian grid (red grid) as f(x,y)f(x,y) in order to extract the bound plane wave part from them.

In its most general form, the function fa(x,y)f_a(x,y) behaves asymptotically as a linear combination of regular φd(x)j(qy)\varphi_d(x)j(qy) and outgoing φd(x)h(+)(qy)\varphi_d(x)h^{(+)}(qy) bound plane waves as well as regular J0(kρ)J_0(k\rho) and outgoing H0(+)(kρ)H_0^{(+)}(k\rho) cylindrical waves. Its projection on the deuteron bound state ⟨φd(x)∣fa(x,y)⟩≡g(y)\langle\varphi_d(x)|f_a(x,y)\rangle\equiv g(y) can therefore be written as

g(y)≈c1j(qy)+c2h(+)(qy)+c3j(y)+c4h(+)(y)(41)g(y) \approx c_1j(qy)+c_2h^{(+)}(qy)+c_3j(y)+c_4h^{(+)}(y) \tag*{(41)}
j(y)=⟨φd(x)∣i(arctan⁡(y/x))J0(kx2+y2)⟩(42)j(y)=\left\langle\varphi_d(x)\left|i(\arctan(y/x))J_0\left(k\sqrt{x^2+y^2}\right)\right.\right\rangle \tag*{(42)}
h(+)(y)=⟨φd(x)∣i(arctan⁡(y/x))H0(+)(kx2+y2)⟩(43)h^{(+)}(y)=\left\langle\varphi_d(x)\left|i(\arctan(y/x))H_0^{(+)}\left(k\sqrt{x^2+y^2}\right)\right.\right\rangle \tag*{(43)}

where the coefficients c1c_1, c2c_2, c3c_3 and c4c_4 are unknowns which we determine through a non linear fit of the function g(y)g(y) over a range of yy values. The two coefficients c1c_1 and c2c_2 will eventually define the elastic scattering amplitude. Since we have imposed a boundary condition pertaining to an outgoing cylindrical wave on the scattered Faddeev component, the coefficient c3c_3 should be zero which we indeed verify through the fit. Note that in the definition of the function h(+)(y)h^{(+)}(y) the unknown breakup amplitude has been replaced by the incoming function i(α)i(\alpha). This approximation is justified a posteriori through the excellent quality of the fit. As a consequence, the coefficient c4c_4 has no physical meaning. Having extracted the coefficients c1c_1 and c2c_2, we can subtract this bound plane wave part from the scattered Faddeev component. In its most general form, this difference behaves asymptotically as

fa(ρ,α)−c1φd(ρcos⁡α)j(qρsin⁡α)−c2φd(ρcos⁡α)h(+)(qρsin⁡α)≈C(α)J0(kρ)+Tij(α)H0(+)(kρ)(44)f_a(\rho,\alpha)-c_1\varphi_d(\rho\cos\alpha)j(q\rho\sin\alpha)-c_2\varphi_d(\rho\cos\alpha)h^{(+)}(q\rho\sin\alpha)\approx C(\alpha)J_0(k\rho)+T_{ij}(\alpha)H_0^{(+)}(k\rho) \tag*{(44)}

where the functions C(α)C(\alpha) and Tij(α)T_{ij}(\alpha) are unknowns which we again determine through a non linear fit at each angle α\alpha. As before, through the application of the boundary condition on the scattered Faddeev component the function C(α)C(\alpha) should be zero which we indeed verify through the fit. Finally, the function Tij(α)T_{ij}(\alpha) is the sought-after breakup amplitude, the relevant T matrix element appearing in eq. (26).

Throughout these fitting procedures, it is important to include in the fit the regular parts and to verify that the corresponding coefficients (c3c_3 and C(α)C(\alpha) above) are indeed zero. This serves as a numerical test on the quality of the calculated Faddeev component.

The total dimension of the operators in equations (34) is N=ncnanαnρN=n_cn_an_\alpha n_\rho where nc=1n_c=1 is the number of Faddeev components, na=2n_a=2 the number of channels, nαn_\alpha the number of collocation points in α\alpha and nρn_\rho the number of collocation points in ρ\rho. In our application, we have N=256,000N=256{,}000 for a density around 0.1%0.1\%; the resolution of the linear system therefore cannot be done by a direct method. We have use the iterative GMRES [12] method in conjunction with a preconditioner consisting of (T−E1)−1(\mathbb{T}-E\mathbb{1})^{-1}. Indeed, the operator (T−E1)(\mathbb{T}-E\mathbb{1}) can be expressed as a sum of two Kronecker products i.e. (T−E1)=A1⊗B1+A2⊗B2(\mathbb{T}-E\mathbb{1})=A_1\otimes B_1+A_2\otimes B_2 where the operators AA are of dimension nρ=500n_\rho=500 and the operators BB of dimension ncnanα=512n_cn_an_\alpha=512. This allows to express the inverse of (T−E1)(\mathbb{T}-E\mathbb{1}) through the evaluation of inverses and diagonalizations of those smaller operators (see for instance [13, 14]). The iterative method with this preconditioner then converges in a few iterations.

Elastic n-d scattering

We start by showing our results for the elastic nn-dd scattering, which correspond to the element S11S_{11} from equation (36), in Figure 4. Calculations are done starting from the deuteron bound state energy EdE_d up to around E≈30E \approx30 MeV. This range is chosen to cover the energies for which benchmark data exists above breakup ; those energies are quoted in terms of the energy in the laboratory frame ElabE_{\mathrm{lab}} (see equation (40)).

Figure 4. Elastic scattering matric element S11S_{11} (black plain and dashed curves) for the doublet JΠ=1/2+J^\Pi= 1/2^+ as a function of the center of mass energy EE. Also shown is the envelope ±∣S11∣\pm|S_{11}| (blue dashed curves). Data below breakup are from [15]. Data above breakup are from [4, 16] and are tagged according to their energy in the laboratory frame (the relation between the two energies is given by eq. (40)).

Below breakup, the benchmark data are taken from reference [15] which uses a theoretical methods similar to ours i.e. configuration-space Faddeev calculations using cubic splines orthogonal collocation.

Above breakup, the benchmark data are taken from references [4, 16] and correspond to a collection of different theoretical methods both in configuration-space and momentum-space. The agreement between our calculations and the various benchmark data, both below and above breakup, is very good. This agreement validates our method of resampling the Faddeev components on a cartesian grid in order to extract the S11S_{11} matrix element from them.

Breakup amplitudes

We then show our results for the breakup amplitudes, which correspond to the matrix elements S12(α)S_{12}(\alpha) and S13(α)S_{13}(\alpha) from equation (36), in Figure 5.

Figure 5. Breakup amplitudes 2πeiπ/42iS12(α)\sqrt{\frac{2}{\pi}\frac{e^{i\pi/4}}{2i}}S_{12}(\alpha) and 2πeiπ/42iS13(α)\sqrt{\frac{2}{\pi}\frac{e^{i\pi/4}}{2i}}S_{13}(\alpha) in the singlet and triplet channels at Elab=14.1E_{lab}=14.1 MeV for the doublet system JΠ=1/2+J^\Pi=1/2^+. Benchmark data are from reference [4].

The benchmark data are taken from reference [4]. In order to facilitate the comparison between our calculations and the benchmark data, our breakup amplitudes are multiplied by 2πeiπ/42i\sqrt{\frac{2}{\pi}\frac{e^{i\pi/4}}{2i}}. Indeed, in reference [4] a matrix element of T\mathbf{T} instead of S\mathbf{S} is presented and this matrix element is expressed with respect to the asymptotic function eikρkρ\frac{e^{ik\rho}}{\sqrt{k\rho}} instead of H0(+)(kρ)H_0^{(+)}(k\rho). The agreement between our breakup amplitudes and the benchmark data is very satisfactory. We have also compared our breakup amplitudes at Elab=4E_{lab}=4 MeV and Elab=42E_{lab}=42 MeV with the same level of agreement. We can see in Figure 5 what we have discussed in section III C: the sharp drop near α≈π/2\alpha\approx\pi/2 results from the fact that this region is not converged given our grid in ρ\rho. One would have to do a calculation for a larger value of ρ\rho in order to have breakup amplitudes converged for all values of α\alpha.

Three-body recombination

Nuclear reactions involving three particles are exceptionally rare so there is no data on the three-body recombination reaction n+n+p→n+dn+n+p \to n+d to the best of our knowledge. Nevertheless, our formalism leads to the full scattering matrix so it calculates the probabilities for this process. According to equation (36) the three-body recombination processes corresponds to the matrix elements S21S_{21} and S31S_{31}. They differ by the two incoming functions i1(α)i_1(\alpha) and i2(α)i_2(\alpha) which characterize the incoming n+n+pn+n+p state (see equation (29)). Following the discussion in section III B, we take for these incoming functions the first two Delves functions Dν(ℓ,λ)(α)D_\nu^{(\ell,\lambda)}(\alpha). In our case where ℓ=λ=0\ell=\lambda=0 we have Dν(0,0)(α)∝sin⁡(2(n+1)α)D_\nu^{(0,0)}(\alpha)\propto\sin(2(n+1)\alpha) with n=0,1,2…n=0,1,2\ldots

The calculated three-body recombination matrix elements S21S_{21} and S31S_{31} as a function of the energy are shown figure (6).

Elastic n-n-p scattering

Finally, we present some results for the elastic n+n+p→n+n+pn+n+p \to n+n+p scattering. The initial n+n+pn+n+p is taken to be describe by the function i1(α)∝sin⁡2αi_1(\alpha)\propto\sin2\alpha (see the discussion above for three body recombination processes). This process corresponds to the matrix elements S22(α)S_{22}(\alpha) and S23(α)S_{23}(\alpha) in equation (36). The calculated elastic n+n+pn+n+p scattering matrix elements S22(α)S_{22}(\alpha) and S23(α)S_{23}(\alpha) for an energy E=0.43E=0.43 MeV are shown figure (7).

Properties of the scattering matrix

The scattering matrix in equation (36) is very compact and describe all processes to and from a single continuum channel or a double continuum one. However, its matrix elements are both regular complex numbers (for processes ending in a single continuum state) and complex functions of the polar angle α\alpha (for processes ending in a double continuum state). In this section, we present how to retrieve usual properties like unitarity and reciprocity from such a scattering matrix. First, we define a matrix of incident functions I\mathcal{I} as

Figure 6. Three-body recombination matrix elements S21S_{21} and S31S_{31} from initial states consisting of the first two Delves functions sin⁡2α\sin2\alpha and sin⁡4α\sin4\alpha respectively. The envelope is also shown in dashed blue lines.

Figure 7. Amplitudes for elastic n+n+p→n+n+pn+n+p\to n+n+p scattering at E=0.43E=0.43 MeV (which corresponds to Elab=4E_{\mathrm{lab}}=4 MeV).

I=(1000i1(α)000i2(α))(45)\mathcal{I} = \begin{pmatrix} 1 & 0 & 0 \\ 0 & i_1(\alpha) & 0 \\ 0 & 0 & i_2(\alpha) \end{pmatrix} \tag*{(45)}

It is simply the matrix in front of the regular functions in equation (29) and collects the functions describing the incident double continuum states.

Then, we define a custom binary operation ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle which will act on the scattering matrix elements. Let γ\gamma, δ\delta be complex numbers and let f(α)f(\alpha), g(α)g(\alpha) be complex functions of the polar angle α\alpha. The binary operation ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle is defined as

⟨γ,δ⟩=γδ(46)\langle\gamma,\delta\rangle= \gamma\delta \tag*{(46)}
⟨γ,g⟩=γ∫dα(g(α)+J[g](α))(47)\langle\gamma,g\rangle= \gamma\int\mathrm{d}\alpha\left(g(\alpha)+\mathcal{J}[g](\alpha)\right) \tag*{(47)}
⟨f,δ⟩=δ∫dαf(α)(48)\langle f,\delta\rangle= \delta\int\mathrm{d}\alpha f(\alpha) \tag*{(48)}
⟨f,g⟩=∫dαf(α)(g(α)+J[g](α))(49)\langle f,g\rangle= \int\mathrm{d}\alpha f(\alpha)\left(g(\alpha)+\mathcal{J}[g](\alpha)\right) \tag*{(49)}

In addition to reducing to the usual multiplication, it performs an integration over all polar angle and importantly implements the Jacobi transform with respect to the second argument (for the sake of clarity, we dropped the subscripts and superscripts of the Jacobi transform ; it has to be understood as a sum over all different components and all partial waves).

Finally, we define a ∗\ast operation between two matrices. This operation corresponds to the usual matrix product (i.e. a dot product between lines of the first matrix and columns of the second one) where the multiplication is replaced by our custom binary operation ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle. Thanks to this ∗\ast operation, we are able to obtain familiar and compact formula both for the unitarity and the reciprocity of the scattering matrix S\mathbf{S}.

Unitarity

The unitarity condition for S\mathbf{S} reads

S∗S†=I∗I(50)\mathbf{S}\ast\mathbf{S}^{\dagger}=\mathcal{I}\ast\mathcal{I} \tag*{(50)}

The above expression equates outgoing probability flux (the left hand side) to incoming flux (the right hand side). Further, we can always choose the functions i(α)i(\alpha) in I\mathcal{I} to have unit incoming flux with respect to ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle to obtain

S∗S†=1(51)\mathbf{S}\ast\mathbf{S}^{\dagger}=1 \tag*{(51)}

We note however that a unit flux with respect to ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle does not mean that the functions i(α)i(\alpha) are normalized in the usual way i.e.

⟨i,i⟩=1⟹⟨i∣i⟩=1(52)\langle i,i\rangle=1\Longrightarrow\langle i\mid i\rangle=1 \tag*{(52)}

because of the contributions from the Jacobi transform.

We can then define a defect from unitarity ηU\eta_U as

ηU=∥1−S∗S†∥(53)\eta_U=\left\lVert1-\mathbf{S}\ast\mathbf{S}^{\dagger}\right\rVert \tag*{(53)}

where ∥⋅∥\lVert\cdot\rVert is a matrix norm. We stress that ηU\eta_U measures not only the conservation of probabilities of the various scattering processes (as the diagonal elements of S∗S†\mathbf{S}\ast\mathbf{S}^{\dagger} being equal to one) but also the degree of orthogonality of the asymptotic channels (as the off-diagonal elements of S∗S†\mathbf{S}\ast\mathbf{S}^{\dagger} being equal to zero).

Reciprocity

In order to verify reciprocity, the breakup amplitudes S12(α)S_{12}(\alpha) and S13(α)S_{13}(\alpha) have to be projected with respect to ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle onto the incoming functions i1(α)i_1(\alpha) and i2(α)i_2(\alpha) in I\mathcal{I}. This can be perform in a compact way using the ∗\ast operation and thus we define

S=I∗ST(54)\mathcal{S}=\mathcal{I}\ast\mathbf{S}^{\mathrm{T}} \tag*{(54)}

and the reciprocity property of S\mathbf{S} means that S\mathcal{S} must be symmetric. We then define a defect from reciprocity as

ηR=∥S−ST∥∥S+ST∥(55)\eta_R=\frac{\left\lVert\mathcal{S}-\mathcal{S}^{\mathrm{T}}\right\rVert}{\left\lVert\mathcal{S}+\mathcal{S}^{\mathrm{T}}\right\rVert} \tag*{(55)}

Figure 8. Defects from unitarity and reciprocity of the calculated scattering matrix as a function of the energy.

We show in figure (8) the defects from unitarity and reciprocity of the calculated scattering matrix as a function of the energy. Several features are worth mentioning. At negative energy where the S\mathbf{S} matrix reduces to the single element S11S_{11} the quality of the calculations is very satisfactory. The square modulus of S11S_{11} only differing from one at a level of 10−5−10−410^{-5}-10^{-4}.

At positive energy, there is an overall degradation in the accuracy of the calculation. Of notice is the poor values of both ηU\eta_U and ηR\eta_R at small positive energies. This is probably due to the fact that our calculation is done for a fixed grid designed to represent distances of the order of 100 fm. Since the asymptotic region in the double continuum channels involves interferences from the cylindrical waves from different Faddeev components, one needs to use a much larger grid in ρ\rho when the wavenumber k=Ek=\sqrt{E} in H0(+)(kρ)H_0^{(+)}(k\rho) is small in order to describe several wavelengths of the cylindrical waves.

Another feature are the apparent oscillations in the defects as a function of the energy (note however that because of the logarithmic scale, those are small in amplitude). This is probably due to the fact that, as we mentioned in section IV A, in imposing a boundary condition adapted to an outgoing cylindrical wave this boundary condition is not adapted to the bound plane wave part of the Faddeev component. As a function of the energy, the applied boundary condition will be periodically rather well adapted to an outgoing bound plane wave though.

In the end, the defects are on a level of 10−3−10−210^{-3}-10^{-2} on the full range of energies which is in fact comparable to our previous calculation [6] and stays satisfactory.

Conclusions

In this work, we have presented a simple method to perform scattering Faddeev calculations in the double continuum based on a resampling of the calculated polar wavefunction on a cartesian grid. We have apply successfully our method on the benchmark neutron-deuteron J=1/2J=1/2 scattering. Note that we have also performed scattering calculations for the quartet J=3/2J=3/2 with the same level of agreement to benchmark data.

We have unify both the single and double continuum channels in a unique scattering matrix and shown how to retrieve its usual properties.

Declaration of interests

The authors do not work for, advise, own shares in, or receive funds from any organisation that could benefit from this article, and have declared no affiliation other than their research organisations.

References

References

  1. [1]LD Fadeev. Scattering theory for a three-particle system. Sov. Phys.-JETP, 12:1014–1019, 1961.
  2. [2]SP Merkuriyev, C Gignoux, and A Laverne. Three-body scattering in configuration space. Annals of physics, 99(1):30–71, 1976.
  3. [3]EA Kolganova, AK Motovilov, and SA Sofianos. Three-body configuration space calculations with hard-core potentials. Journal of Physics B: Atomic, Molecular and Optical Physics, 31(6):1279, 1998.DOI
  4. [4]James Lewis Friar, GL Payne, W Glöckle, D Hüber, and H Witalla. Benchmark solutions for n-d breakup amplitudes. Physical Review C, 51(5):2356, 1995.pubmed.ncbi.nlm.nih.gov/9970316
  5. [5]PA Belov and SL Yakovlev. Asymptotic method for determining the amplitude for three-particle breakup: Neutron-deuteron scattering. Physics of Atomic Nuclei, 76(2):126–138, 2013.
  6. [6]Romain Guérou t. Calculation of bound and continuum states of the ne3 van der waals trimer. Journal of Physics B: Atomic, Molecular and Optical Physics, 58(1):015201, 2024.
  7. [7]LM Delves. Tertiary and general-order collisions (ii). Nuclear Physics, 20:275–308, 1960.
  8. [8]W Glöckle and GL Payne. Boundary conditions for three-body scattering in configuration space. Physical Review C, 45(3):974, 1992.
  9. [9]Alan Robert Edmonds. Angular momentum in quantum mechanics, volume 4. Princeton university press, 1996.
  10. [10]Marc Nadal-Ferret, Ricard Gelabert, Miquel Moreno, and José M Lluch. A method to compute probability current in generic coordinates. The Journal of chemical physics, 134(7), 2011.
  11. [11]Vladimir Roudnev and Michael Cavagnero. Automatic grid construction for few-body quantum-mechanical calculations. Computer Physics Communications, 182(10):2099–2106, 2011.DOI
  12. [12]Youcef Saad and Martin H Schultz. Gmres: A generalized minimal residual algorithm for solving nonsymmetric linear systems. SIAM Journal on scientific and statistical computing, 7(3):856–869, 1986.
  13. [13]NW Schellingerhout, LP Kok, and GD Bosveld. Configuration-space faddeev calculations: Supercomputer accuracy on a personal computer. Physical Review A, 40(10):5568, 1989.pubmed.ncbi.nlm.nih.gov/9901938
  14. [14]Rimantas Lazauskas and Jaume Carbonell. The faddeev–yakubovsky symphony. Few-Body Systems, 60(4):62, 2019.DOI
  15. [15]CR Chen, GL Payne, James Lewis Friar, and Benjamin F Gibson. Low-energy nucleon-deuteron scattering. Physical Review C, 39(4):1261, 1989.pubmed.ncbi.nlm.nih.gov/9955332
  16. [16]James Lewis Friar, Benjamin F Gibson, G Berthold, W Glöckle, Th Cornelius, H Witala, J Haidenbauer, Y Koike, GL Payne, JA Tjon, et al. Benchmark solutions for a model three-nucleon scattering problem. Physical Review C, 42(5):1838, 1990.

Paper details

vlnc:W_betqtsf2
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents