Barrier penetration with a finite mesh method

Abstract

A standard way to solve a Schrödinger equation is to discreteize the radial coordinates and apply a numerical method for a differential equation, such as the Runge-Kutta method or the Numerov method. Here I employ a discrete basis formalism based on a finite mesh method as a simpler alternative, with which the numerical computation can be easily implemented by ordinary linear algebra operations. I compare the numerical convergence of the Numerov integration method to the finite mesh method for calculating penetrabilities of a one-dimensional potential barrier, and show that the latter approach has better convergence properties.

I Introduction

In most of physics problems, a Schrödinger equation cannot be solved analytically but has to be solved numerically. For a bound state problem, one may expand wave functions on some finite basis and diagonalize the resultant Hamiltonian matrix. Alternatively, one may discretize the radial coordinates and successively obtain a wave function at the mesh points with e.g., the Runge-Kutta method or the Numerov method koonin .

A yet different method, referred to as a discrete-basis formalism, has been proposed in Ref. FBA18 . In this method, one first forms a Hamiltonian matrix based on discretized radial meshes and solve it with a linear algebra with appropriate boundary conditions. An advantage of this method is that the method is well compatible with a many-body Hamiltonian, in particular in a configuration-interaction formulation BH2022 ; BH2023 ; UH2023 . Notice that the discrete-basis formalism is referred to as a 3D mesh method in the context of nuclear density functional theoryDavies80 ; ev8 ; ev8-2 ; tanimura2015 ; ren2017 ; li2020 .

Even though the discrete-basis formalism has been applied to an induced fission problem FBA18 ; BY2019 ; YBF2020 ; YBF2021 ; BH2023 , its applicability has not yet been clarified, at least for a scattering problem. In this paper, I therefore apply the discrete-basis formalism to a simple one-dimensional barrier penetration problem, and carry out a comparative study of the numerical accuracy. To this end, I shall consider a Gaussian barrier and compare the penetrabilities obtained with the discrete-basis formalism to those with the standard Numerov method.

The paper is organized as follows. In Sec. II, I will detail the discrete-basis formalism for a one-dimensional problem. In Sec. III, I will apply it to a barrier penetration of a one-dimensional Gaussian barrier and discuss the applicability of the discrete-basis formalism. I will then summarize the paper in Sec. IV.

II Discrete-basis formalism for barrier penetration

Consider a one-dimensional system for a particle with mass $m$ under a potential $V(x)$. The Hamiltonian for this system reads,

$$H=-\frac{\hbar^{2}}{2m}\frac{d^{2}}{dx^{2}}+V(x).$$ (1)

I discretize the radial coordinate as $x_{i}=x_{\rm min}+(i-1)\Delta x$, and consider the model space from $x_{1}=x_{\rm min}$ to $x_{N}\equiv x_{\rm max}$. Using the 3-point formula for the kinetic energy in $H$, the Hamiltonian (1) is transformed to a matrix form of

$$H_{ij}=-t\delta_{i,j+1}+(2t+V_{i})\delta_{i,j}-t\delta_{i,j-1},$$ (2)

where $t$ is defined as $t=\frac{\hbar^{2}}{2m(\Delta x)^{2}}$ and $V_{i}\equiv V(x_{i})$. The wave function $\phi_{i}\equiv\phi(x_{i})$ then obeys

$$-t\phi_{0}\delta_{i,1}+\sum_{j=1}^{N}H_{ij}\phi_{j}-t\phi_{N+1}\delta_{i,N}=E\phi_{i}.$$ (3)

In the absence of the potential $V$, the wave function $\phi_{n}^{(0)}$ obeys the equation

$$-t(\phi_{n+1}^{(0)}-2\phi_{n}^{(0)}+\phi_{n-1}^{(0)})=E\phi_{n}^{(0)}.$$ (4)

I consider a free-particle solution given by

$$\phi_{n}^{(0)}\propto e^{-ikn\Delta x}-e^{ikn\Delta x}.$$ (5)

Substituting this into Eq. (4), one finds

$$\cos(k\Delta x)=1-\frac{E}{2t}.$$ (6)

In the presence of the potential $V$, I consider the case where the particle is incident from the left hand side of the potential. Assuming that the potential $V$ almost vanishes at $x_{\rm max}$, the wave function $\phi_{N+1}$ is given by $\phi_{N+1}=e^{ik\Delta x}\phi_{N}$. Substituting this into Eq. (3), one finds

$$\phi_{i}=[(\tilde{H}-E)^{-1}]_{i1}t\phi_{0}\equiv G_{i1}t\phi_{0},$$ (7)

where $\tilde{H}$ is defined as $\tilde{H}_{ij}=H_{ij}-te^{ik\Delta x}\delta_{i,N}\delta_{j,N}$ and the Green’s function $G$ is defined as $G=(\tilde{H}-E)^{-1}$.

Assuming that the potential $V(x)$ is negligible at $x=x_{1}$ and $x_{2}$, the wave functions at these points are given as linear superpositions of $e^{\pm ikn\Delta x}$ with $n=1$ and 2, respectively. I parameterize the coefficients of the linear superpositions in terms of $t$ and the wave function $\phi_{0}$ as,

$$\phi_{1} = (Ae^{ik\Delta x}+Be^{-ik\Delta x})t\phi_{0},$$ (8)

$$\phi_{2} = (Ae^{2ik\Delta x}+Be^{-2ik\Delta x})t\phi_{0}.$$ (9)

This is equivalent to assume

$$G_{11}=Ae^{ik\Delta x}+Be^{-ik\Delta x}.$$ (10)

Substituting Eqs. (8) and (9) into Eq. (3) and using Eq. (6), one finds

$$Ae^{2ik\Delta x}+Be^{-2ik\Delta x}=2\cos(k\Delta x)G_{11}-\frac{1}{t}.$$ (11)

Combining this with Eq. (10), one finds

$$A = \frac{e^{-ik\Delta x}}{e^{ik\Delta x}-e^{-ik\Delta x}}\,(e^{ik\Delta x}G_{11}-1/t),$$ (12)

$$B = -\frac{e^{ik\Delta x}}{e^{ik\Delta x}-e^{-ik\Delta x}}\,(e^{-ik\Delta x}G_{11}-1/t).$$ (13)

Writing the wave function $\phi_{N}$ as $\phi_{N}=G_{N1}t\phi_{0}\equiv Te^{ik\Delta x}t\phi_{0}$, the penetrability $P(E)$ reads,

$$P(E)=\left|\frac{T}{A}\right|^{2}=\left|\frac{2\sin(k\Delta x)G_{N1}}{e^{ik\Delta x}G_{11}-1/t}\right|^{2}.$$ (14)

III Penetrability of a Gaussian barrier

Figure 1: The penetraibilities of a Gaussian barrier given by Eq. (15) with $V_{0}=100$ MeV and $s=$ 2 fm. The mass is set to be $m=29m_{N}$, where $m_{N}$ is the nucleon mass. The upper panel is obtained with the Numerov method (the dashed line) and the discrete-basis formalism (the filled circles) with the mesh size of $\Delta x=0.05$ fm. On the other hand, the lower panel shows the results of the Numerov method (the dashed line), the modified Numerov method (the dotted line), and the discrete-basis formalism (the solid line) with the mesh size of $\Delta x=0.15$ fm.

Let us now numerically evaluate the penetrability for a given potential. For this purpose, I consider a Gaussian potential,

$$V(x)=V_{0}\,e^{-x^{2}/2s^{2}}.$$ (15)

Following Refs. dasso83 ; dasso83b ; HB04 , the parameters are chosen to be $V_{0}=100$ MeV and $s=2$ fm together with $m=29m_{N}$, where $m_{N}$ is the nucleon mass, to mimic the fusion reaction of ${}^{58}$Ni+${}^{58}$Ni. We set $x_{\rm min}=-10$ fm and $x_{\rm max}=10$ fm.

The upper panel of Fig. 1 shows the penetrabilities of the Gaussian barrier obtained with $\Delta x=0.05$ fm. The dashed line and the solid circles denote the results with the standard Numerov method and the discrete-basis formalism, respectively. The value of $\Delta x$ is small enough in this case, and both the method lead to accurate results. The lower panel shows the results with a larger value of $\Delta x$, that is, $\Delta x=0.15$ fm. In this case, the numerical error is significantly large with the Numerov method: the penetrabilities do not reach unity even at energies well above the barrier (see the dashed line). This is the case also with the modified Numverov method ccfull , with which the penetrability even exceeds unity at high energies with a non-monotonic behavior (see the dotted line). In marked contrast, the results with the discrete-basis formalism is rather robust and the penetrabilities are almost the same as the one with $\Delta x=0.05$ fm shown in the upper panel. Notice that the discrete-basis formalism employs the simple 3-point formula for the kinetic energy, while a more sophisticated formula is used in the Numerov and the modified Numerov methods. Yet, it is interesting to notice that the discrete-basis method is numerically more stable than the Numerov and the modified Numerov methods. We point out that $\Delta x$ cannot be taken larger than $\sqrt{2\hbar^{2}/Em}$, though. If $\Delta x$ exceeds this value, the right hand side of Eq. (6) exceeds unity and the wave number $k$ cannot be defined unless it is extended to a complex number.

IV Summary

I examined the applicability of the discrete-basis method for a reaction theory. To this end, I considered barrier penetration of a one-dimensional Gaussian barrier. It was demonstrated that the discrete-basis method provides a more accurate and stable method than the standard Numerov method. This property may be helpful in obtaining numerically stable solutions of coupled-channels equations HOM2022 ; HT2012 .

The discrete-basis formalism has a good connection to a many-body Hamiltonian. As a matter of fact, there have been several applications of this method to microscopic descriptions of induced fission. In such applications, absorbing potentials, or imaginary energies, are introduced to a model Hamiltonian, and the absorbing probability is computed with the so called Datta formula FBA18 . Even though the model setup is somewhat different from a barrier problem in one-dimension, in which there is no absorbing part in the Hamiltonian, the conclusion in this paper would remain the same in the fission problem as well.