betaFeedback

WorksPhysics and Astronomy

Nonlocalization of singular potentials in quantum dynamics

Abstract

Nonlocal modeling has drawn more and more attention and becomes steadily more powerful in scientific computing. In this paper, we demonstrate the superiority of a first-principle nonlocal model – Wigner function – in treating singular potentials which are often used to model the interaction between point charges in quantum science. The nonlocal nature of the Wigner equation is fully exploited to convert the singular potential into the Wigner kernel with weak or even no singularity, and thus highly accurate numerical approximations are achievable, which are hardly designed when the singular potential is taken into account in the local Schrödinger equation. The Dirac delta function, the logarithmic, and the inverse power potentials are considered. Numerically converged Wigner functions under all these singular potentials are obtained with an operator splitting spectral method, and display many interesting quantum behaviors as well.

2020 Mathematics Subject Classification: 81S30; 45K05; 35Q40; 65M70; 35S05

Background and motivation

It has been shown that the point charge description of electrons usually agrees well with the experimental results,[1] where the interaction between them is dominated by the Coulomb potential — a typical singular potential in quantum science.[2, 3, 4, 5, 6] Such Coulomb interaction has found various applications in physics[2, 7] and chemistry.[1, 8, 9] Apart from that, there exist some other singular potentials to describe the interactions arising from scattering problems,[10] the short-range interactions in condensed matter,[11, 12] Dirac monopole in the magnetic field[13] and etc. The logarithmic potential is also adopted to measure the entropy density in the study of two-phase flow.[14]

Directly plugging a singular potential into the Schrödinger equation

iℏ∂∂tψ(x,t)=−ℏ22m∇x2ψ(x,t)+V(x)ψ(x,t),(1)\mathrm{i}\hbar\frac{\partial}{\partial t}\psi(x,t)=-\frac{\hbar^2}{2m}\nabla_x^2\psi(x,t)+V(x)\psi(x,t), \tag*{(1)}

and then seeking the numerical solutions runs into a problematic situation, where ψ(x,t)\psi(x,t) gives the wavefunction, mm and ℏ\hbar signify the mass of the particle and the reduced Planck constant, respectively. Let’s take the Dirac delta function potential as an example:

V(x)=Hδ(x)(2)V(x)=H\delta(x) \tag*{(2)}

with a power of HH (the influence of potential) and the Dirac delta function (see Figure 1):

The Dirac delta function potential and the Wigner kernelThe Dirac delta function potential and the Wigner kernel

Figure 1. (a) The Dirac delta function potential (2). (b) The Wigner kernel (8).

δ(x)={+∞x=0,0x≠0,∫Rδ(x) dx=1.(3)\delta(x)=\begin{cases} +\infty& x=0,\\ 0 & x\ne0 \end{cases}, \qquad\int_{\mathbb{R}}\delta(x)\,\mathrm{d}x=1. \tag*{(3)}

This Dirac delta function potential, which diverges at x=0x=0, is often adopted to model an infinite well and barrier.[15] Obviously, there is no way for the finite difference method to find a suitable approximation to the function (3) and then to the equation Figure 1. The Dirac delta function potential and its Wigner kernel with power H=1H = 1. It can be intuitively seen that the singular potential is transformed into a non-singular Wigner kernel.

(1) equipped with the singular potential (2). Recourses to the Galerkin method inevitably sacrifices the accuracy or convergence order, which has been already pointed out by [16, 17] in solving the elliptical boundary value problem with the Dirac delta function source −Δu(x)=δ(x)-\Delta u(x) = \delta(x).

In this paper, we adopt a nonlocalization approach based on the integral formulation to deal with the singular potential situation. Specifically, we turns to the Wigner function [18]

f(x,k,t)=∫Rψ(x+y2,t)ψ†(x−y2,t)exp⁡(−iky) dy,(4)f(x,k,t) = \int_{\mathbb{R}} \psi\left(x+\frac{y}{2},t\right)\psi^{\dagger}\left(x-\frac{y}{2},t\right)\exp(-iky)\,\mathrm{d}y, \tag*{(4)}

and its governing equation

∂∂tf(x,k,t)+ℏkm∇xf(x,k,t)=ΘV[f](x,k,t),(5)\frac{\partial}{\partial t}f(x,k,t) + \frac{\hbar k}{m}\nabla_x f(x,k,t) = \Theta_V[f](x,k,t), \tag*{(5)}

both of which are defined in phase space (x,k)(x,k) with xx being the position and kk the wavenumber. Starting from the definition (4) where the Wigner function is calculated from the density matrix ψ(x,t)ψ†(x,t)\psi(x,t)\psi^{\dagger}(x,t) by changing to center-of-mass coordinates followed by a Fourier transform, the Wigner equation (5) can be derived from the Schrödinger equation (1) in a straightforward manner. The nonlocalization of singular potentials is embodied in the pseudo-differential operator

ΘV[f](x,k,t)=∫RVw(x,k−k′)f(x,k′,t) dk′,(6)\Theta_V[f](x,k,t) = \int_{\mathbb{R}} V_w(x,k-k')f(x,k',t)\,\mathrm{d}k', \tag*{(6)}
Vw(x,k)=12πiℏ∫Rexp⁡(−iky)[V(x+y2)−V(x−y2)] dy,(7)V_w(x,k) = \frac{1}{2\pi i\hbar}\int_{\mathbb{R}}\exp(-iky)\left[V\left(x+\frac{y}{2}\right)-V\left(x-\frac{y}{2}\right)\right]\,\mathrm{d}y, \tag*{(7)}

and all the information of potential V(x)V(x) is contained in the Wigner kernel Vw(x,k)V_w(x,k). Substituting the Dirac delta function potential (2) into Eq. (7) leads to

Vw(x,k)=2Hπℏsin⁡(2xk),(8)V_w(x,k) = \frac{2H}{\pi\hbar}\sin(2xk), \tag*{(8)}

the plot of which is displayed in Figure 1. It can be readily observed there that the Wigner kernel Vw(x,k)V_{\mathrm{w}}(x,k) is no longer singular and thus we have a chance to seek highly accurate numerical solutions to the Wigner equation (5) with singular potentials. That is, the point singularity in V(x)V(x) is distributed over the whole space with the nonlocal action of pseudo-differential operator, thereby alleviating or even eliminating the singularity. After obtaining the Wigner function (4), the average of a quantum operator A^\hat{A} can be expressed as

⟨A^⟩t=∬R×RA(x,k)f(x,k,t) dxdk,(9)\langle\hat{A}\rangle_t = \iint_{\mathbb{R}\times\mathbb{R}} A(x,k)f(x,k,t)\,\mathrm{d}x\mathrm{d}k, \tag*{(9)}

where A(x,k)A(x,k) gives the corresponding classical function in phase space. In other words, the Wigner function formulation is fully equivalent to the wavefunction formulation for quantum mechanics.[19] Generally speaking, nonlocal models may offer more explanations for phenomena that involve possible singularities including the interaction with singular potentials and occurring at a distance.[20]

By exploiting the intrinsic nonlocal nature of the Wigner function approach, we are able to obtain highly accurate numerical approximations to observable quantities in quantum dynamics with singular potentials with the help of spectral methods and operator splitting techniques. For demonstration purposes, this work focuses on the singular potentials the Wigner kernels of which have analytical forms, like the Dirac delta function potential. Otherwise, other extra numerical techniques should be adopted, for example, the truncated kernel method.[21, 22] It should be noted that there exist few high precision numerical simulations of the Wigner equation under singular potentials except for a recent attempt to numerically solve the Wigner-Coulomb system[23] as well as some qualitative analysis results.[24, 25]

The rest of the paper is organized as follows. Section 2 presents the numerical results for the Dirac delta function potential and an comparison with the finite size model. Scattering of the Fermi-Dirac distribution in 4-D phase space is shown as well. Extensions to other three types of singular potentials are given in Section 3. Finally, conclusions and discussions are drawn in Section 4.

Quantum dynamics in a Dirac delta function potential

After truncating the kk-space into \mathcal{K}=[k_{\min},k_{\max}],{citep}26` a Fourier spectral approximation with N_k$ terms to the Wigner function f(x,k,t)f(x,k,t) reads

f(x,k,t)≈fNk(x,k,t)=∑ν=−Nk/2+1Nk/2αν(x,t)ψν(k),(10)f(x,k,t) \approx f_{N_k}(x,k,t)=\sum_{\nu=-N_k/2+1}^{N_k/2}\alpha_\nu(x,t)\psi_\nu(k), \tag*{(10)}

where ψν(k)=e2πiν(k−kmin⁡)/Lk\psi_\nu(k)=e^{2\pi i\nu(k-k_{\min})/L_k} with Lk=kmax⁡−kmin⁡L_k=k_{\max}-k_{\min} gives the basis. Then the pseudo-differential term (6) can be approximated as follows

ΘV[f](x,k,t)≈ΘVT[fNk](x,k,t)=∑ν=−Nk/2+1Nk/2cν(x)αν(x,t)ψν(k),(11)\Theta_V[f](x,k,t) \approx\Theta_V^{T}[f_{N_k}](x,k,t)=\sum_{\nu=-N_k/2+1}^{N_k/2}c_\nu(x)\alpha_\nu(x,t)\psi_\nu(k), \tag*{(11)}
cν(x)=∫K′Vw(x,k′)e−2πiνk′/Lk dk′,K′=[−Lk,Lk].(12)c_\nu(x)=\int_{\mathcal{K}'}V_{\mathrm{w}}(x,k')e^{-2\pi i\nu k'/L_k}\,\mathrm{d}k',\qquad\mathcal{K}'=[-L_k,L_k]. \tag*{(12)}

For the situation of the Dirac delta function potential (2), plugging Eq. (8) into Eq. (12) leads to the following close formula

cν(x)=2Hiπℏ(sin⁡(ων+(x)Lk)ων+(x)−sin⁡(ων−(x)Lk)ων−(x)),(13)c_\nu(x)=\frac{2Hi}{\pi\hbar}\left(\frac{\sin(\omega_\nu^{+}(x)L_k)}{\omega_\nu^{+}(x)}-\frac{\sin(\omega_\nu^{-}(x)L_k)}{\omega_\nu^{-}(x)}\right), \tag*{(13)}

where ων±(x)=2x±2πLkν\omega_{\nu}^{\pm}(x)=2x\pm\frac{2\pi}{L_k}\nu and the limits must be used when ων±(x)=0\omega_{\nu}^{\pm}(x)=0. It can be easily observed in Eq. (11) that only cν(x)c_{\nu}(x) involves the singular potential (2), through the non-singular Wigner kernel (8), thereby implying that the formula (13) treats the singularity with high accuracy, where the numerical errors only come from the truncation of kk-space and the spectral approximation of f(x,k,t)f(x,k,t). After that, we adopt the Chebyshev spectral element method with inflow boundary conditions [27] in xx-space and the fourth-order operator splitting technique [28] in tt-direction to determine the remaining unknowns αν(x,t)\alpha_{\nu}(x,t). For simplicity, we use the same MM collocation points in all QQ cells in xx-space. Moreover, the above numerical method can be readily extended to 4-D and higher-dimensional scenarios in a dimension-by-dimension manner by using the tensor product of 2-D basis functions.

The L2L^2-error

ϵ2(t)=[∬X×K(F(x,k,t)−fref(x,k,t))2 dxdk]1/2(14)\epsilon_2(t)=\left[\iint_{\mathcal{X}\times\mathcal{K}}\left(F(x,k,t)-f^{\mathrm{ref}}(x,k,t)\right)^2\,\mathrm{d}x\mathrm{d}k\right]^{1/2} \tag*{(14)}

and L∞L^\infty-error

ϵ∞(t)=max⁡(x,k)∈X×K{∣F(x,k,t)−fref(x,k,t)∣}(15)\epsilon_\infty(t)=\max_{(x,k)\in\mathcal{X}\times\mathcal{K}}\left\{\left|F(x,k,t)-f^{\mathrm{ref}}(x,k,t)\right|\right\} \tag*{(15)}

are used to analyze the convergence of the errors, where X:=[XL,XR]\mathcal{X}:=[X_L,X_R] is the computational domain in xx-space, F(x,k,t)F(x,k,t) represents the numerical solution, and the numerical solution obtained on the finest mesh is taken as the reference fref(x,k,t)f^{\mathrm{ref}}(x,k,t). For convenience, the above errors in Eqs. (14) and (15) are numerically calculated on the following uniform mesh

(xi,kj)=((i−1/2)(XR−XL),(j−1/2)(kmax⁡−kmin⁡))/Num,i,j=1,…,Num,(16)(x_i,k_j)=\left((i-1/2)(X_R-X_L),(j-1/2)(k_{\max}-k_{\min})\right)/N_{\mathrm{um}},\quad i,j=1,\ldots,N_{\mathrm{um}}, \tag*{(16)}

where NumN_{\mathrm{um}} denotes the mesh size.

2-D scattering of Gaussian wave packet

As stated in,[26, 27] the Gaussian wave packet

f(x,k,0)=1πe−(x−x0)22σ2−2σ2(k−k0)2(17)f(x,k,0)=\frac{1}{\pi}e^{-\frac{(x-x^0)^2}{2\sigma^2}-2\sigma^2(k-k^0)^2} \tag*{(17)}

is usually adopted as the initial function to test the convergence rate as well as to investigate the quantum tunneling, where (x0,k0)(x^0,k^0) gives the initial center position and σ\sigma is the minimum position spread. We will simulate its quantum scattering in the Dirac delta function potential, which has never been reported before in the literature. For the purpose of testing only, we set ℏ=1\hbar=1 eV ⋅\cdot fs, m=1m=1 eV ⋅\cdot fs2⋅^2\cdot nm−2^{-2}, x0=−10x^0=-10 nm, k0=2k^0=2 nm−1^{-1}, and σ=2\sigma=2 nm. The computational domain is chosen as [XL,XR]=[−30[X_L,X_R]=[-30 nm, 3030 nm] which is divided evenly into Q=20Q=20 cells, −kmin⁡=kmax⁡=π-k_{\min}=k_{\max}=\pi nm−1^{-1}, and the quantum evolution with a fixed time step Δt=0.01\Delta t=0.01 fs is stopped at tfin=10t_{\mathrm{fin}}=10 fs.

The first test tries H=1H=1 in Eq. (2) and the resulting Wigner functions at t=2.5,5,7.5,10t=2.5,5,7.5,10 fs are displayed in Figure 2, which is obtained on the finest mesh we have tried: Nk=512N_k=512, M=55M=55. It is clearly observed there that the wave packet still goes partially through the barrier through the barrier whose height is infinite. That is a clear manifestation of quantum tunneling effect in the Dirac delta function potential,[29] reflecting a fundamental difference between quantum world and macroscopic world. A possible explanation is that the width of the Dirac delta function barrier tends to be extremely small, very close to 0 in particular, albeit the infinitely large height. To study the convergence rate with respect to NkN_k, the number of collocation points in each xx-cell is fixed to be M=55M=55. Similarly, when studying the convergence rate with respect to MM, the number of collocation points in kk-space is fixed to be Nk=512N_k=512. Figure 3 displays the spectral convergence of ϵ2(10)\epsilon_2(10) and ϵ∞(10)\epsilon_\infty(10) against NkN_k or MM where we have set Num=600N_{\mathrm{um}}=600 in Eq. (16). That is, the numerical results in Figure 2 are numerically converged.

Wigner function at $t=2.5$ fs

(a)

Wigner function at $t=5$ fs

(b)

Wigner function at $t=7.5$ fs

(c)

Wigner function at $t=10$ fs

(d)

Figure 2. The Wigner functions at different instants: A Gaussian wave packet runs through the Dirac delta function barrier Eq. (2) with power H=1H=1. Quantum tunneling effect can be clearly observed.

Spectral convergence with respect to $N_k$ and $M$ during the scattering of a Gaussian wave packet

Figure 3. Spectral convergence with respect to NkN_k (left) and MM (right) during the scattering of a Gaussian wave packet in the Dirac delta function potential Eq. (2) with H=1H=1.

To be more specific, the uncertainty

σx(t) σp(t)=⟨(x^−⟨x^⟩t)2⟩t⟨(p^−⟨p^⟩t)2⟩t(18)\sigma_x(t)\,\sigma_p(t)=\sqrt{\left\langle\left(\hat{x}-\langle\hat{x}\rangle_t\right)^2\right\rangle_t}\sqrt{\left\langle\left(\hat{p}-\langle\hat{p}\rangle_t\right)^2\right\rangle_t} \tag*{(18)}

is adopted to measure the nonlocality, and its numerical value is still obtained on the uniform mesh given in Eq. (16), where p=ℏkp=\hbar k is the momentum. We show in Table 1 that the numerical results with different mesh sizes and find that the numerically converged value for σx(10)σp(10)\sigma_x(10)\sigma_p(10) is 20.8510 for H=1H=1.

NkN_kσx(10)σp(10)\sigma_x(10)\sigma_p(10)MMσx(10)σp(10)\sigma_x(10)\sigma_p(10)NumN_{um}σx(10)σp(10)\sigma_x(10)\sigma_p(10)
3219.96162121.016910020.8355
6420.89382620.868220020.8467
12820.85703120.840630020.8510
25620.85293620.850240020.8510
30020.85244120.851545020.8510
40020.85154520.850950020.8510
50020.85105120.851055020.8510
51220.85105520.851060020.8510

Table 1. Numerical values for σx(10)σp(10)\sigma_x(10)\sigma_p(10) with respect to increasing NkN_k, MM and NumN_{um} and the converged value is 20.8510.

In addition to the uncertainty, we continue to use the partial mass

Pr(t)=∫[0,XR]×Kf(x,k,t) dxdk(19)P_r(t)=\int_{[0,X_R]\times\mathcal{K}} f(x,k,t)\,\mathrm{d}x\mathrm{d}k \tag*{(19)}

for investigating the tunneling effect for ten different powers: H=±0.5H=\pm0.5, ±1\pm1, ±1.5\pm1.5, ±2\pm2, and ±2.5\pm2.5. PrP_r can also be regarded as the tunneling rate in view of that the total mass equals to one. Figure 4 and Figure 5 show the tunneling rates and uncertainties for potential barriers H>0H>0 and potential wells with H<0H<0, respectively. The curves in Figure 4 exhibit the deceleration of Pr(t)P_r(t) as HH increases, and uncertainty peaking at H=1H=1. It can be further found that the moment when the uncertainty reaches the maximum coincides with that the tunneling rate reaches to about 0.5. At this moment, the variances accumulate the most and it is difficult to observed the position and momentum of the wavepacket simultaneously. Moreover, when the power is high enough, i.e., H≥1.5H\ge1.5, there are significant fluctuations of σx(t)σp(t)\sigma_x(t)\sigma_p(t) and Pr(t)P_r(t). That could be comprehended as follows. The influence of the power HH leads to the high oscillations in the Wigner function at the center x=0x=0 nm. Therefore, these two observables fluctuate during the wave packet interacting with the barrier. When it comes to the wells with negative power, it can be observed in Figure 5 that the trend of the tunneling rate Pr(t)P_r(t) is opposite to that of the uncertainty σx(t)σp(t)\sigma_x(t)\sigma_p(t): Pr(t)P_r(t) decreases and σx(t)σp(t)\sigma_x(t)\sigma_p(t) increases as ∣H∣|H| increases.

Tunneling rates and uncertainties for the Dirac delta function barriers

Figure 4. (a) Tunneling rate. (b) Uncertainty. Tunneling rates and uncertainties for the Dirac delta function barriers with powers H=0.5H = 0.5, 11, 1.51.5, 22 and 2.52.5. The moment when the uncertainty reaches the maximum coincides with that the tunneling rate reaches about 50%.

Tunneling rates and uncertainties for the Dirac delta function well

Figure 5. (a) Tunneling rate. (b) Uncertainty. Tunneling rates and uncertainties for the Dirac delta function well with powers H=−0.5H = -0.5, −1-1, −1.5-1.5, −2-2 and −2.5-2.5.

Finite size effect

The point charge causes singularity because it has no size. In view of this, one may use a finite size model to avoid the singularity. The Gaussian function with size of aa, denoted by Va(x)V_a(x), is usually used to mimic the point charge model,[1, 30] the validity of which relies on the following limit

δ(x)=lim⁡a→0+0Va(x)=lim⁡a→0+012πaexp⁡(−x22a2).(20)\delta(x) = \lim_{a \to0+0} V_a(x) = \lim_{a \to0+0} \frac{1}{\sqrt{2\pi a}}\exp\left(-\frac{x^2}{2a^2}\right). \tag*{(20)}

However, we would like to point out that there is a huge gap between the quantum behavior caused by the point charge δ(x)\delta(x) and finite-size charge Va(x)V_a(x).

aa nm1050.50.10.011E-31E-41E-160
σx(10) σp(10)\sigma_x(10)\ \sigma_p(10)0.98111.06631.22721.00410.80270.80050.80030.800320.8510

Table 2. Uncertainties for different sizes at tfin=10t_{\mathrm{fin}} = 10 fs. a=0a = 0 nm signifies the Dirac delta function barrier. The uncertainty of the Dirac delta function barrier δ(x)\delta(x) is much larger than that of the Gaussian barrier of any finite size Va(x)V_a(x).

Table 2 displays the numerically converged uncertainties at tfin=10t_{\mathrm{fin}} = 10 fs for decreasing sizes. When a=10−16a = 10^{-16} nm, σx(10) σp(10)\sigma_x(10)\ \sigma_p(10) is about 0.8003, which is far less that 20.8510 caused by δ(x)\delta(x). In fact, as the size gradually becomes smaller, the uncertainty first grows to 1.2272, then gradually decreases, and finally stays around 0.8003. Such apparent discrepancy can be also observed from the Wigner functions at the final instant in Figure 6. Compared the Wigner function under Va(x)V_a(x) with a=0.5a = 0.5 nm in Figure 6a, it is obvious that the Wigner function under δ(x)\delta(x) in Figure 6b reaches the extrema around x=0x = 0 nm, where the singular point exactly locates at, and the oscillation between positive and negative values is more vicious. More specifically, at the final instant, the average position and moment are (⟨x^⟩10,⟨p^⟩10)=(9.7732,1.9840)(\langle\hat{x}\rangle_{10},\langle\hat{p}\rangle_{10})=(9.7732,1.9840) under the Gaussian barrier, but alter to (⟨x^⟩10,⟨p^⟩10)=(1.1313,0.0047)(\langle\hat{x}\rangle_{10},\langle\hat{p}\rangle_{10})=(1.1313,0.0047) under δ(x)\delta(x). Figure 6c provides the tunneling rate Pr(t)P_r(t). It reaches almost to 1 as expected under the Gaussian barrier, which is consistent with the weak presence of negative part of the Wigner function in Figure 6a. By contrast, it is manifested in Figure 6d that the same wave packet can just partially pass through the Dirac delta function barrier and its uncertainty increases significantly, which should result from the infinite height of the potential.

Wigner function for $V_a(x)$ with $a = 0.5$ nm

(a) Va(x)V_a(x) with a=0.5a = 0.5 nm.

Wigner function for $\delta(x)$

(b) δ(x)\delta(x).

Tunneling rate

(c) Tunneling rate.

Uncertainty

(d) Uncertainty.

Figure 6. Different quantum behavior caused by the Gaussian barrier with finite size Va(x)V_a(x) and Dirac delta function barrier δ(x)\delta(x). The Wigner functions at tfin=10t_{\mathrm{fin}} = 10 fs are displayed in (a) and (b), the tunneling rate in (c), and the uncertainty in (d). Both tunneling rates and uncertainties are obviously different.

In a word, our numerical experiments suggests an essential difference between the singular potential and its regularized counterpart as already shown in investigating nuclear magnetic shielding.[30] No matter how small the size of Gaussian barrier we choose, it is still a smooth and local potential. The Dirac delta function potential, on the contrary, an ideal model widely used to simulate the point charge field source of quantum chemical reactions, has some essentially difference from such regularized one in studying quantum phenomena.

4-D scattering of Fermi-Dirac distribution

In view of the high dimensionality of phase space, the foregoing treatment of singular potentials using the Wigner function approach can be also extended to high-dimensional scenarios. This section is devoted to scattering of the Fermi-Dirac distribution in 4-D phase space under a singular potential. Specifically, we adopt the following position-independent 2-D Fermi-Dirac distribution function[31, 32] as the initial data for the Wigner equation (5)

f(x1,x2,k1,k2,0)=2mkBTπℏ∫0∞11+exp⁡(y2+(ℏk1)2+(ℏk2)22m−EFkBT) dy(21)f(x_1,x_2,k_1,k_2,0)=\frac{\sqrt{2mk_BT}}{\pi\hbar}\int_0^\infty\frac{1}{1+\exp\left(y^2+\frac{(\hbar k_1)^2+(\hbar k_2)^2}{2m}-E_F\over k_BT\right)}\,\mathrm{d}y \tag*{(21)}

where m=0.067mem=0.067m_e, me=5.68562966 eV⋅fs2⋅nm−2m_e=5.68562966\ \mathrm{eV}\cdot\mathrm{fs}^2\cdot\mathrm{nm}^{-2}, ℏ=0.658211899 eV⋅fs\hbar=0.658211899\ \mathrm{eV}\cdot\mathrm{fs}, kB=8.61734279×10−5 eV⋅K−1k_B=8.61734279\times10^{-5}\ \mathrm{eV}\cdot\mathrm{K}^{-1}, TT is taken as 300 K, and EF=0.1E_F=0.1 eV signifies the Fermi energy. Meanwhile, we choose an annular singular potential V(x1,x2)=∑i=18δ(x1−d1i)δ(x2−d2i)V(x_1,x_2)=\sum_{i=1}^{8}\delta(x_1-d_1^i)\delta(x_2-d_2^i), where all singular points, numbered 1 to 8 in an anti-clockwise direction with the right-most one being the first, are evenly distributed in a circle with radius equal to 2 nm and (d1i,d2i)(d_1^i,d_2^i) gives the position of the ii-th singular point.

4-D scattering of Fermi-Dirac distribution

The computational domain is X×X×K×K\mathcal{X} \times\mathcal{X} \times\mathcal{K} \times\mathcal{K} with X=[−10 nm,10 nm]\mathcal{X} = [-10\ \mathrm{nm}, 10\ \mathrm{nm}] and K=[−π nm−1,π nm−1]\mathcal{K} = [-\pi\ \mathrm{nm}^{-1}, \pi\ \mathrm{nm}^{-1}]. We use the same NkN_k collocation points for all K\mathcal{K}, the same number of elements Q=5Q = 5 and MM collocation points in each element for all X\mathcal{X}. The quantum dynamics is evolved to tfin=2.5t_{\mathrm{fin}} = 2.5 fs with time step Δt=0.01\Delta t = 0.01 fs. We measure the errors of spatial marginal distribution of the Wigner function,

Fsm(x1,x2,t)=∬K×Kf(x1,x2,k1,k2,t) dk1dk2,(22)F_{\mathrm{sm}}(x_1,x_2,t) = \iint_{\mathcal{K}\times\mathcal{K}} f(x_1,x_2,k_1,k_2,t)\,\mathrm{d}k_1\mathrm{d}k_2, \tag*{(22)}

in a similar way to calculating Eqs. (14) and (15). Figure 7 presents the errors against NkN_k and MM after fixing Num=400N_{um}=400 in Eq. (16) and the spectral convergence is evident again. The spatial marginal distributions on the finest mesh at different instants are displayed in Figure 8 where the corresponding constant distributions in the free space, Fsmfree(x1,x2,t)≡0.05384F^{free}_{sm}(x_1,x_2,t) \equiv0.05384, have been subtracted. It can be observed there that the Fermi-Dirac distribution first reacts strongly to the eight singular points in the circle during which many small oscillations are produced in the central area of the circle, and then gradually expands to the surroundings. Obviously, such expanding is blocked by the eight Dirac delta function potentials and the resulting interference forms 12 branches outside the circle: 4 big branches lie in the main directions, north, east, south and west, respectively, and the remaining 8 small branches are equally distributed between them, where six lines that are determined by pairs of singular points: (1, 7), (2, 6), (3, 5), (1, 3), (4, 8) and (5, 7) serve as the boundaries of branches. Inside the circle, the interference pattern shows a clear square structure that is also shaped by the same six lines. At the final time t=2.5t=2.5 fs, a basin structure emerges: The spatial marginal distribution inside the circle is reduced to less than FsmfreeF^{free}_{sm} (see Figure 8(i)), and that above FsmfreeF^{free}_{sm} is all outside the circle (see Figure 8(h)).

Errors of the spatial marginal distribution against $N_k$ and $M$

Figure 7. 4-D scattering of Fermi-Dirac distribution: Errors of the spatial marginal distribution against NkN_k (left) and MM (right).

Eight-panel spatial marginal distributions and contour plot

Figure 8. 4-D scattering of Fermi-Dirac distribution: Spatial marginal distributions in Eq. (22) subtracted by the corresponding constant distributions in the free space. Eight Dirac delta function potentials give eight singular points (small white dots), which are numbered 1 to 8 in an anti-clockwise direction with the right-most one being the first and evenly distributed in a circle with radius equal to 2 nm. In (i), we plot the contour inside the circle at t=2.5t = 2.5 fs with 10 equally spaced contour lines from −0.01216-0.01216 to 0.0040870.004087.

Extensions to other singular potentials

In this section, we devote ourselves into using the Wigner function approach to the following singular potentials:

  • The logarithmic potential

V(x)=Hlog⁡(x)(23)V(x) = H\log(x) \tag*{(23)}

which is naturally related to the entropy expression;[14]

  • The inverse power potential for α∈(0,1)\alpha\in(0,1)

V(x)=H∣x∣−α(24)V(x) = H|x|^{-\alpha} \tag*{(24)}

which can be found in various quantum mechanical models;[5, 33, 34]

  • The inverse square potential

V(x)=H∣x∣−2(25)V(x) = H|x|^{-2} \tag*{(25)}

which has strong singularity at x=0x=0 and is extensively used in high-energy scattering studies.[34]

The logarithmic potential

Plugging Eq. (23) into Eq. (7) yields the corresponding Wigner kernel

Vw(x,k)=−Hℏsin⁡(2xk)∣k∣,(26)V_w(x,k) = -\frac{H}{\hbar}\frac{\sin(2xk)}{|k|}, \tag*{(26)}

and then substituting it in Eq. (12) leads to

ℏ2Hicν(x)=∫0Lksin⁡(2xk′)sin⁡(ν~k′)k′ dk′:=(∫0ε+∫−εLk)gν(x,k′) dk′,(27)\frac{\hbar}{2Hi}c_\nu(x) = \int_{0}^{L_k}\frac{\sin(2xk')\sin(\widetilde{\nu}k')}{k'}\,\mathrm{d}k' := \left(\int_{0}^{\varepsilon}+\int_{-\varepsilon}^{L_k}\right)g_\nu(x,k')\,\mathrm{d}k', \tag*{(27)}

where ε\varepsilon is a prescribed small parameter and ν~=2πν/Lk\widetilde{\nu}=2\pi\nu/L_k. Using the Taylor expansion in (0,ε)(0,\varepsilon) gives

∫0εgν(x,k′) dk′=ν~xε2−(13ν~x3+112ν~3x)ε4+O(ε7),\int_{0}^{\varepsilon}g_\nu(x,k')\,\mathrm{d}k' = \widetilde{\nu}x\varepsilon^2-\left(\frac{1}{3}\widetilde{\nu}x^3+\frac{1}{12}\widetilde{\nu}^3x\right)\varepsilon^4+O(\varepsilon^7),

and with the help of the cosine integral function Ci⁡(x)=−∫x+∞cos⁡tt dt\operatorname{Ci}(x)=-\int_x^{+\infty}\frac{\cos t}{t}\,\mathrm{d}t, we have

∫εLkgν(x,k′) dk′=Ci⁡(∣ων+(x)∣ε)−Ci⁡(∣ων−(x)∣ε)−Ci⁡(∣ων+(x)∣Lk)+Ci⁡(∣ων−(x)∣Lk)2\int_{\varepsilon}^{L_k}g_\nu(x,k')\,\mathrm{d}k'=\frac{\operatorname{Ci}(|\omega_\nu^+(x)|\varepsilon)-\operatorname{Ci}(|\omega_\nu^-(x)|\varepsilon)-\operatorname{Ci}(|\omega_\nu^+(x)|L_k)+\operatorname{Ci}(|\omega_\nu^-(x)|L_k)}{2}

Accordingly, the expansion coefficient cν(x)c_\nu(x) given in an oscillatory improper integral (27) can be approximated to the machine resolution by choosing ε=1E−5\varepsilon=1\mathrm{E}{-5}. Other parameters are set to be: H=1H=1, −XL=XR=30-X_L=X_R=30 nm, −kmin⁡=kmax⁡=π-k_{\min}=k_{\max}=\pi nm−1^{-1}, Nk=512N_k=512, Q=20Q=20, M=55M=55, Δt=0.01\Delta t=0.01 fs, and Num=600N_{\mathrm{um}}=600. We use an initial Gaussian wave packet close to the origin by setting x0=−1x_0=-1 nm and k0=0.5k_0=0.5 nm−1^{-1} in Eq. (17). Numerical convergence tests are given in Figure 9 and show clearly the spectral accuracy. The left column of Figure 10 plots the numerically converged Wigner functions at t=2.5t=2.5, 5, 7.5, 10 fs obtained on the finest mesh. It can be observed there that, the wave packet is attracted by the logarithmic potential (23), and keeps moving around the singular point, during which many small oscillations appear around the origin along with the singularity.

Numerical convergence plots against $N_k$ and $M$

Figure 9. The logarithmic potential: Numerical convergence against NkN_k (left) and MM (right) at tfin=10t_{\mathrm{fin}}=10 fs.

Figure 10. The logarithmic potential: The Wigner functions corresponding to the analytical Wigner kernel (26) (left), to the “approximate” Wigner kernel (28) (middle) and their spatial marginal distributions (right).

It has been shown that the Poisson summation formula can be used to approximate the Wigner kernel (7) well (yζ=ζΔyy_\zeta=\zeta\Delta y and Δy\Delta y being the spacing):

Vw(x,k)≈Δy2πiℏ∑ζ=−∞+∞[V(x+yζ2)−V(x−yζ2)]e−ikyζ(28)V_w(x,k)\approx\frac{\Delta y}{2\pi i\hbar}\sum_{\zeta=-\infty}^{+\infty}\left[V\left(x+\frac{y_\zeta}{2}\right)-V\left(x-\frac{y_\zeta}{2}\right)\right]e^{-iky_\zeta} \tag*{(28)}

when the external potential V(x)V(x) is smooth and localized,[26] but fails when taking the Coulomb interaction into account.[23] Here we would like to confirm this failure using the logarithmic potential. After using the “approximate” Wigner kernel (28) to replace the analytical one (26) and keeping all other settings unchanged, we rerun the simulations in the left column of Figure 10 and display the resulting Wigner functions in the middle column of Figure 10, which shows some obvious discrepancy. Compared with the reference solutions, the Wigner functions obtained with Eq. (28) display much more severe oscillations with higher peaks and deeper valleys, and the corresponding spatial marginal distributions show some spurious oscillations and non-physical negative values (see the right column of Figure 10).

The inverse power potential

According to the Fourier transform of the inverse power function with α∈(0,1)\alpha\in(0,1): F[∣x∣α](ξ)=−2sin⁡(π2(α−1))Γ(α)∣ξ∣−α\mathcal{F}[|x|^{\alpha}](\xi)=-2\sin\left(\frac{\pi}{2}(\alpha-1)\right)\Gamma(\alpha)|\xi|^{-\alpha}, the Wigner kernel of Eq. (24) reads

Vw(x,k)=Hsin⁡(π2α)Γ(α)22−απℏsin⁡(2xk)∣k∣α(29)V_w(x,k)=\frac{H\sin\left(\frac{\pi}{2}\alpha\right)\Gamma(\alpha)2^{2-\alpha}}{\pi\hbar}\frac{\sin(2xk)}{|k|^{\alpha}} \tag*{(29)}

The inverse power potential

where Γ(x)\Gamma(x) gives the Gamma function. Combining Eqs. (29) and (12) together yields

cν(x)=2Hisin⁡(π2α)Γ(α)22−απℏ∫0Lksin⁡(2xk)sin⁡(2πνk/Lk)kα dk,(30)c_{\nu}(x)=\frac{2H i\sin\left(\frac{\pi}{2}\alpha\right)\Gamma(\alpha)2^{2-\alpha}}{\pi\hbar}\int_{0}^{L_k}\frac{\sin(2xk)\sin(2\pi\nu k/L_k)}{k^{\alpha}}\,\mathrm{d}k, \tag*{(30)}

which can be efficiently approximated to the machine accuracy with the help of the Generalized Hypergeometric function 1F2((1−α)/2;1/2,(3−α)/2;x){}_1F_2((1-\alpha)/2; 1/2,(3-\alpha)/2; x).{3} For example, when α=1/2\alpha=1/2, that Generalized Hypergeometric function reduces to the Fresnel function C(x)=∫0xcos⁡(πt2/2) dtC(x)=\int_0^x\cos(\pi t^2/2)\,\mathrm{d}t. We adopt the same parameters as in § 3.1 to simulate the scattering between the Gauss wave packet and the inverse power potential. Figure 11 shows the spectral convergence with respect to NkN_k and MM again, and Figure 12 the Wigner functions at four different instants. We are able to observe there that the wave packet partially passes through the inverse power potential barrier; and the negative Wigner function, sandwiched between two scattered wave packets in opposite directions, strongly implies the uncertainty principle around the singularity. Figure 13 further plots the effect of power α\alpha on the tunneling rate Pr(t)P_r(t) in Eq. (19) and uncertainty σx(t)σp(t)\sigma_x(t)\sigma_p(t) in Eq. (18). It is evident that, the tunneling rate gradually increases as α\alpha rises, which reflects that the width of the potential becomes steadily smaller. Since Pr(t)P_r(t) never rises over 0.5 throughout the scattering, the uncertainty σx(t)σp(t)\sigma_x(t)\sigma_p(t) shows a mounting tendency whilst Pr(t)P_r(t) ascends, which is incurred by the shape change of the potential, namely the increase of α\alpha.

Numerical convergence plots for the inverse power potential

Figure 11. *The inverse power potential: Numerical convergence with respect to NkN_k (left) and MM (right) at tfin=10t_{\mathrm{fin}} = 10 fs.

Wigner function at $t = 2.5$ fs

(a) (a) t=2.5t = 2.5 fs.

Wigner function at $t = 5$ fs

(b) (b) t=5t = 5 fs.

Wigner function at $t = 7.5$ fs

(c) (c) t=7.5t = 7.5 fs.

Wigner function at $t = 10$ fs

(d) (d) t=10t = 10 fs.

Figure 12. *The inverse power potential: The Wigner functions at different instants. We set H=1H = 1 and α=1/2\alpha= 1/2 in Eq. (24).

Tunneling rate and uncertainty plots

Figure 13. The inverse power potential: Tunneling rates and uncertainties. It is clearly seen that the uncertainty σx(t)σp(t)\sigma_x(t)\sigma_p(t) shows a mounting tendency whilst Pr(t)P_r(t) ascends.

The inverse square potential

The Wigner kernel of the inverse square potential (3.3) is

Vw(x,k)=−4Hℏ∣k∣sin⁡(2xk)(31)V_w(x,k)=-\frac{4H}{\hbar}|k|\sin(2xk) \tag*{(31)}

and plugging it into Eq. (12), we are able to get a close form for cν(x)c_\nu(x) like Eq. (13). The parameters are: [XL,XR]=[−30 nm,30 nm][X_L,X_R]=[-30\ \mathrm{nm},30\ \mathrm{nm}], [kmin⁡,kmax⁡]=[−π nm−1,π nm−1][k_{\min},k_{\max}]=[-\pi\ \mathrm{nm}^{-1},\pi\ \mathrm{nm}^{-1}], Nk=512N_k=512, Q=40Q=40, M=55M=55, Num=600N_{\mathrm{um}}=600, Δt=0.005 fs\Delta t=0.005\ \mathrm{fs}, H=1H=1, x0=−5 nmx^0=-5\ \mathrm{nm}, k0=1 nm−1k^0=1\ \mathrm{nm}^{-1}. Figure 14 verifies the spectral convergence against both NkN_k and MM. Figure 15 displays the quantum dynamics where the Gaussian wave packet is almost totally reflected after hitting the singular barrier. The tunneling rate Pr(t)P_r(t) are 0.01158, 0.009115, 0.006731 and 0.002025 at t=2t = 2, 4, 5, and 8 fs, respectively, indicating transparently that it is difficult for the wave packet to pass through the barrier due to the strong singularity of the inverse power potential. This is very different from the scattering shown in § 3.2 with the inverse power potential which has much weaker singularity. Moreover, severe oscillations of positive and negative Wigner functions clearly appear around the origin, which accords with the summary that the quantum behavior near the singularities is difficult to be measured.

Spectral convergence against $N_k$ and $M$ for the inverse square potential

Figure 14. The inverse square potential: Spectral convergence against NkN_k (left) and MM (right).

Wigner functions at different instants for the inverse square potential

Figure 15. The inverse square potential: The Wigner functions at different instants. The wave packet is almost totally reflected by the singular barrier.

Conclusions

With the help of the Wigner function approach for quantum mechanics — a first-principle nonlocal model, we perform highly accurate numerical simulations of quantum dynamics under singular potentials, during which the nonlocal characteristic of Wigner function contributes to the attenuation of singularity of the potentials. Numerically converged Wigner functions under the Dirac delta function, the logarithmic, and the inverse power potentials are obtained with an operator splitting spectral method. Many interesting quantum behaviors are also revealed during the scattering under these singular potentials. It should be noted that all existing Wigner simulations truncate the nonlocal integral in kk-space, but the effect of such truncation on long-time simulations of quantum dynamics is hardly estimated in advance. Instead, motivated by recently proposed adaptive technologies on unbounded domains,[36, 37, 38] we are developing numerical methods to solve the Wigner equation without truncating the nonlocal kk-integral.

Acknowledgements. This research was supported by the National Key R&D Program of China (Nos. 2020AAA0105200, 2022YFA1005102) and the National Natural Science Foundation of China (Nos. 12288101, 11822102). SS is partially supported by Beijing Academy of Artificial Intelligence (BAAI).

References

  1. [1]L. Visscher and K. G. Dyall. Dirac–Fock atomic electronic structure calculations using different nuclear charge distributions. Atomic Data and Nuclear Data Tables, 67:207, 1997.DOI
  2. [2]A. K. Roy. Studies on some singular potentials in quantum mechanics. International Journal of Quantum Chemistry, 104:861, 2005.DOI
  3. [3]K. M. Case. Singular potentials. Physical Review, 80:797, 1950.DOI
  4. [4]A. M. Perelomov and V. S. Popov. “Fall to the center” in quantum mechanics. Theoretical and Mathematical Physics, 4:664, 1970.DOI
  5. [5]K. Meetz. Singular potentials in nonrelativistic quantum mechanics. Il Nuovo Cimento (1955-1965), 34:690, 1964.DOI
  6. [6]W. M. Frank, D. J. Land, and R. M. Spector. Singular potentials. Reviews of Modern Physics, 43:36, 1971.
  7. [7]S. Keraani. Wigner measures dynamics in a Coulomb potential. Journal of Mathematical Physics, 46:063512, 2005.DOI
  8. [8]M. Cinal. Highly accurate numerical solution of Hartree–Fock equation with pseudospectral method for closed-shell atoms. Journal of Mathematical Chemistry, 58:1571, 2020.DOI
  9. [9]G. W. Wei. Discrete singular convolution for the solution of the Fokker–Planck equation. Journal of Chemical Physics, 110:8930, 1999.DOI
  10. [10]G. Esposito. Scattering from singular potentials in quantum mechanics. Journal of Physics A: Mathematical and General, 31:9493, 1998.DOI
  11. [11]S. Tan. Energetics of a strongly correlated Fermi gas. Annals of Physics, 323:2952, 2008.
  12. [12]M. F. Gusson, A. O. O. Gonçalves, R. O. Francisco, R. G. Furtado, J. C. Fabris, and J. A. Nogueira. Dirac δ-function potential in quasiposition representation of a minimal-length scenario. European Physical Journal C, 78:1, 2018.DOI
  13. [13]M. W. Ray, E. Ruokokoski, S. Kandel, M. Möttönen, and D. S. Hall. Observation of Dirac monopoles in a synthetic magnetic field. Nature, 505:657, 2014.DOI
  14. [14]C. G. Gal, A. Giorgini, and M. Grasselli. The nonlocal Cahn–Hilliard equation with singular potential: well-posedness, regularity and strict separation property. Journal of Differential Equations, 263:5253, 2017.
  15. [15]M. Belloni and R. W. Robinett. The infinite well and Dirac delta function potentials as pedagogical, mathematical and physical models in quantum mechanics. Physics Reports, 540:25, 2014.DOI
  16. [16]R. Araya, E. Behrens, and R. Rodríguez. A posteriori error estimates for elliptic problems with Dirac delta source terms. Numerische Mathematik, 105:193, 2006.DOI
  17. [17]R. Scott. Finite element convergence for singular data. Numerische Mathematik, 21:317, 1973.DOI
  18. [18]E. Wigner. On the quantum correction for thermodynamic equilibrium. Physical Review, 40:749, 1932.
  19. [19]T. L. Curtright, D. B. Fairlie, and C. K. Zachos. A concise treatise on quantum mechanics in phase space. World Scientific Publishing Company, 2013.DOI
  20. [20]M. D’Elia, Q. Du, C. Glusa, M. Gunzburger, X. Tian, and Z. Zhou. Numerical methods for nonlocal and fractional models. Acta Numerica, 29:1124, 2020.
  21. [21]L. Greengard, S. Jiang, and Y. Zhang. The anisotropic truncated kernel method for convolution with free-space Green’s functions. SIAM Journal on Scientific Computing, 40:A3733, 2018.DOI
  22. [22]F. Vico, L. Greengard, and M. Ferrando. Fast convolution with free-space Green’s functions. Journal of Computational Physics, 323:2191, 2016.DOI
  23. [23]Y. Xiong, Y. Zhang, and S. Shao. A characteristic-spectral-mixed scheme for six-dimensional Wigner-Coulomb dynamics. arXiv:2205.02380, 2022.arxiv.org/abs/2205.02380
  24. [24]B. Li and J. Shen. The Wigner–(Poisson)–Fokker–Planck equation with singular potential. Applicable Analysis, 96:563, 2017.DOI
  25. [25]Iliševič, D. and Turnšek, A. Stability of the Wigner equation–a singular case. Journal of Mathematical Analysis and Applications, 429:273, 2015.
  26. [26]Y. Xiong, Z. Chen, and S. Shao. An advective-spectral-mixed method for time-dependent many-body Wigner simulations. SIAM Journal on Scientific Computing, 38:B491, 2016.arxiv.org/abs/1602.08853
  27. [27]S. Shao, T. Lu, and W. Cai. Adaptive conservative cell average spectral element methods for transient Wigner equation in quantum transport. Communications in Computational Physics, 9:711, 2011.DOI
  28. [28]Z. Chen, S. Shao, and W. Cai. A high order efficient numerical method for 4-D Wigner equation of quantum double-slit interferences. Journal of Computational Physics, 396, 2019.DOI
  29. [29]H. Grabert and U. Weiss. Quantum tunneling rates for asymmetric double-well systems with Ohmic dissipation. Physical Review Letters, 54:1605, 1985.pubmed.ncbi.nlm.nih.gov/10031086
  30. [30]Y. Xiao, W. Liu, L. Cheng, and D. Peng. Four-component relativistic theory for nuclear magnetic shielding constants: Critical assessments of different approaches. Journal of Chemical Physics, 126:214101, 2007.pubmed.ncbi.nlm.nih.gov/17567184
  31. [31]Z. Chen, H. Jiang, and S. Shao. A higher-order accurate operator splitting spectral method for the Wigner-Poisson system. Journal of Computational Electronics, 21:756, 2022.
  32. [32]A. J. MacLeod. Algorithm 779: Fermi-Dirac functions of order-1/2, 1/2, 3/2, 5/2. ACM Transactions on Mathematical Software (TOMS), 24:1-12, 1998.DOI
  33. [33]R. Serber. Scaling law for high-energy elastic scattering. Physical Review Letters, 13:32, 1964.DOI
  34. [34]G. Tiktokpoulos. High-energy large-angle scattering by singular potentials. Physical Review, 138:B1550, 1965.
  35. [36]M. Xia, S. Shao, and T. Chou. A frequency-dependent p-adaptive technique for spectral methods. Journal of Computational Physics, 446:110627, 2021.arxiv.org/abs/2010.02008
  36. [37]M. Xia, S. Shao, and T. Chou. Efficient scaling and moving techniques for spectral methods in unbounded domains. SIAM Journal on Scientific Computing, 43:A3244-A3268, 2021.arxiv.org/abs/2009.13170
  37. [38]T. Chou, S. Shao, and M. Xia. Adaptive Hermite spectral methods in unbounded domains. Applied Numerical Mathematics, 183:201-220, 2023.

Paper details

vlnc:W_934qr4ph
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