Weak Adversarial Neural Pushforward Method for the Wigner Transport Equation
2026
Read Weak Adversarial Neural Pushforward Method for the Wigner Transport Equation (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-08763.txt.
Introduction
He, Cai, and Shao extend the weak neural pushforward method to the Wigner transport equation for quantum phase space dynamics. Their key core point is that combining the far reaching potential operator with plane wave test functions yields a Dirac delta. That delta inverts the Fourier transform that defines the Wigner kernel, cutting it to a pointwise finite gap of the potential at two shifted inputs. Splitting the answer into two signed distributions handles Wigner minus signs, keeping the method mesh free and scalable.
Catalog record
| YEAR | 2026 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 25484 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.08763
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Weak Adversarial Neural Pushforward Method for the Wigner Transport Equation Abstract We extend the Weak Adversarial Neural Pushforward Method to the Wigner transport equation governing the phase-space dynamics of quantum systems. The central contribution is a structural observation: integrating the nonlocal pseudo-differential potential operator against plane-wave test functions produces a Dirac delta that exactly inverts the Fourier transform defining the Wigner potential kernel, reducing the operator to a pointwise finite difference of the potential at two shifted arguments. This holds in arbitrary dimension, requires no truncation of the Moyal series, and treats the potential as a black-box function oracle with no derivative information. To handle the negativity of the Wigner quasi-probability distribution, we introduce a signed pushforward architecture that decomposes the solution into two non-negative phase-space distributions mixed with a learnable weight. The resulting method inherits the mesh-free, Jacobian-free, and scalable properties of the original framework while extending it to the quantum setting. Andrew Qing He * Wei Cai * Sihong Shao † * Department of Mathematics, Southern Methodist University, Dallas, TX, USA. [email protected], [email protected] † School of Mathematical Sciences, Peking University, Beijing, China. [email protected] August 10, 2026 Keywords: Wigner transport equation, Wigner function, quantum phase space, neural pushforward map, weak adversarial network, plane-wave test functions, pseudo-differential operator MSC 2020: 65N75, 68T07, 81Q05, 81S30 1 Introduction The Wigner function, introduced by Wigner in 1932 [ 1 ] , provides a phase-space representation of quantum mechanics that parallels the classical Liouville description. The evolution of the Wigner function f(t,\mathbf{x},\mathbf{p}) is governed by the Wigner transport equation [ 2 ] , which differs from the classical Liouville equation by a nonlocal pseudo-differential potential operator \Theta[V] . This operator is defined through a double integral involving the Fourier transform of the antisymmetrized potential, and its nonlocality is the mathematical expression of quantum interference in phase space. Solving the Wigner transport equation numerically is challenging for several reasons. First, the equation lives in a 2N -dimensional phase space, where N is the number of spatial degrees of freedom, making grid-based methods prohibitively expensive for all but the lowest-dimensional problems. Second, the pseudo-differential operator \Theta[V] is nonlocal in the momentum variable, requiring either its explicit evaluation (which involves a Fourier transform at each grid point) or a truncation of the Moyal series to a finite number of momentum derivatives. The latter produces the truncated Wigner approximation (TWA) [ 3 , 4 ] , which is widely used in quantum optics and cold-atom physics but introduces systematic errors that grow with \hbar and with the degree of anharmonicity of the potential. Third, the Wigner function is a quasi-probability distribution that can take negative values, precluding the direct use of probabilistic particle methods that rely on non-negative densities. Stochastic approaches to the Wigner equation have been developed based on the signed-particle formulation [ 5 , 6 ] , in which particles carry positive or negative weights and undergo branching processes that encode the quantum potential. While these methods avoid spatial discretization, they suffer from the numerical sign problem: the variance of estimators grows exponentially with time as the number of sign changes accumulates. Neural network methods for solving partial differential equations have seen rapid development in recent years, including Physics-Informed Neural Networks (PINNs) [ 7 ] , the Deep Ritz method [ 8 ] , and Weak Adversarial Networks (WAN) [ 9 ] . In [ 10 ] , the Weak Adversarial Neural Pushforward Method (WANPM) was introduced for solving Fokker–Planck equations. WANPM learns a neural pushforward map that transforms samples from a simple base distribution into samples from the solution distribution, with training guided by a weak formulation employing computationally efficient plane-wave test functions. The method has since been extended to McKean–Vlasov equations [ 11 ] and to Fokker–Planck equations on Riemannian manifolds [ 12 ] . The present paper extends WANPM to the Wigner transport equation. The key finding is a structural compatibility between the plane-wave test functions used in WANPM and the Fourier-integral definition of the pseudo-differential operator \Theta[V] : when the weak-form integral \int(\Theta[V]f)\,\varphi\,d\mathbf{x}\,d\mathbf{p} is evaluated with a plane-wave test function \varphi=\sin(\mathbf{w}_{x}\cdot\mathbf{x}+\mathbf{w}_{p}\cdot\mathbf{p}+\kappa t+b) , the Fourier kernel in \Theta[V] combines with the plane wave to produce a Dirac delta that collapses the integral, yielding an expectation over f involving only V evaluated at two shifted positions \mathbf{x}\pm\frac{\hbar}{2}\mathbf{w}_{p} . This reduction is exact — it requires no truncation of the Moyal series — and treats V as a black-box function oracle, requiring no derivative information. The paper is organized as follows. Section 2 introduces the Wigner transport equation and the pseudo-differential operator \Theta[V] . Section 3 derives the weak formulation and evaluates the potential integral, establishing the central result. Section 4 assembles the complete weak-form residual and discusses the connection to the classical Liouville equation and the truncated Wigner approximation. Section 5 introduces the signed pushforward architecture for handling Wigner negativity. Section 6 describes the adversarial training algorithm. Section 8 concludes. 2 The Wigner transport equation Consider a quantum system with N degrees of freedom, positions \mathbf{x}=(x_{1},\ldots,x_{N})\in\mathbb{R}^{N} , momenta \mathbf{p}=(p_{1},\ldots,p_{N})\in\mathbb{R}^{N} , mass m , and potential V(\mathbf{x}) . The Wigner function f(t,\mathbf{x},\mathbf{p}) [ 1 ] evolves according to the Wigner transport equation [ 2 ] : \frac{\partial f}{\partial t}+\sum_{i=1}^{N}\frac{p_{i}}{m}\frac{\partial f}{\partial x_{i}}=\Theta[V]\,f, (1) where the pseudo-differential operator \Theta[V] is defined by (\Theta[V]\,f)(t,\mathbf{x},\mathbf{p})=\frac{1}{\mathrm{i}\hbar}\frac{1}{(2\pi\hbar)^{N}}\int_{\mathbb{R}^{N}}\!\!\int_{\mathbb{R}^{N}}\left[V\!\left(\mathbf{x}+\tfrac{\mathbf{y}}{2}\right)-V\!\left(\mathbf{x}-\tfrac{\mathbf{y}}{2}\right)\right]\mathrm{e}^{\mathrm{i}(\mathbf{p}-\mathbf{p}^{\prime})\cdot\mathbf{y}/\hbar}\,f(t,\mathbf{x},\mathbf{p}^{\prime})\;\mathrm{d}\mathbf{y}\,\mathrm{d}\mathbf{p}^{\prime}. (2) The left-hand side of ( 1 ) is the classical Liouville operator (free streaming along Hamiltonian trajectories); the right-hand side encodes the quantum potential, which is nonlocal in the momentum variable. The Wigner function satisfies the normalization \int f\,\mathrm{d}\mathbf{x}\,\mathrm{d}\mathbf{p}=1 and the marginal conditions \int_{\mathbb{R}^{N}}f(t,\mathbf{x},\mathbf{p})\,\mathrm{d}\mathbf{p}=|\psi(t,\mathbf{x})|^{2}\geq 0,\qquad\int_{\mathbb{R}^{N}}f(t,\mathbf{x},\mathbf{p})\,\mathrm{d}\mathbf{x}=|\hat{\psi}(t,\mathbf{p})|^{2}\geq 0, (3) where \psi and \hat{\psi} are the position- and momentum-space wave functions. However, f itself can take negative values; it is a quasi-probability distribution. 3 Weak Formulation and the Central Result 3.1 Setup Multiply ( 1 ) by the plane-wave test function \varphi^{(k)}(t,\mathbf{x},\mathbf{p})=\sin(\mathbf{w}_{x}^{(k)}\!\cdot\mathbf{x}+\mathbf{w}_{p}^{(k)}\!\cdot\mathbf{p}+\kappa^{(k)}t+b^{(k)}), (4) with trainable parameters \mathbf{w}_{x}^{(k)},\mathbf{w}_{p}^{(k)}\in\mathbb{R}^{N} , \kappa^{(k)},b^{(k)}\in\mathbb{R} , and integrate over (\mathbf{x},\mathbf{p})\in\mathbb{R}^{2N} and t\in[0,T] . Integrating by parts in t and in \mathbf{x} (for the transport term on the left-hand side of ( 1 )), we obtain \int_{\mathbb{R}^{2N}}\!f\,\varphi^{(k)}\Big|_{t=T}\mathrm{d}\mathbf{x}\,\mathrm{d}\mathbf{p}\;-\;\int_{\mathbb{R}^{2N}}\!f\,\varphi^{(k)}\Big|_{t=0}\mathrm{d}\mathbf{x}\,\mathrm{d}\mathbf{p}\\ -\;\int_{0}^{T}\!\!\int_{\mathbb{R}^{2N}}\!f\left(\frac{\partial\varphi^{(k)}}{\partial t}+\sum_{i}\frac{p_{i}}{m}\frac{\partial\varphi^{(k)}}{\partial x_{i}}\right)\mathrm{d}\mathbf{x}\,\mathrm{d}\mathbf{p}\,\mathrm{d}t\;=\;\int_{0}^{T}I^{(k)}\;\mathrm{d}t, (5) where the right-hand side contains the potential integral I^{(k)}=\int_{\mathbb{R}^{2N}}(\Theta[V]\,f)\;\varphi^{(k)}\;\mathrm{d}\mathbf{x}\,\mathrm{d}\mathbf{p}. (6) The left-hand side is standard: the transport term produces the pointwise expression [\kappa^{(k)}+\mathbf{w}_{x}^{(k)}\cdot\mathbf{p}/m]\cos\phi^{(k)} , where \phi^{(k)}=\mathbf{w}_{x}^{(k)}\cdot\mathbf{x}+\mathbf{w}_{p}^{(k)}\cdot\mathbf{p}+\kappa^{(k)}t+b^{(k)} . The central task is to evaluate I^{(k)} . 3.2 Evaluation of the potential integral I^{(k)} Substituting ( 2 ) and ( 4 ) into ( 6 ) and writing \alpha:=\mathbf{w}_{x}^{(k)}\cdot\mathbf{x}+\kappa^{(k)}t+b^{(k)} : I^{(k)}=\frac{1}{\mathrm{i}\hbar}\frac{1}{(2\pi\hbar)^{N}}\int_{\mathbb{R}^{4N}}\left[V\!\left(\mathbf{x}+\tfrac{\mathbf{y}}{2}\right)-V\!\left(\mathbf{x}-\tfrac{\mathbf{y}}{2}\right)\right]\mathrm{e}^{\mathrm{i}(\mathbf{p}-\mathbf{p}^{\prime})\cdot\mathbf{y}/\hbar}\,f(t,\mathbf{x},\mathbf{p}^{\prime})\,\sin(\alpha+\mathbf{w}_{p}^{(k)}\!\cdot\mathbf{p})\;\mathrm{d}\mathbf{y}\,\mathrm{d}\mathbf{p}^{\prime}\,\mathrm{d}\mathbf{p}\,\mathrm{d}\mathbf{x}. (7) Step 1: \mathbf{p} -integration. Writing \sin(\alpha+\mathbf{w}_{p}^{(k)}\cdot\mathbf{p})=\frac{1}{2\mathrm{i}}(\mathrm{e}^{\mathrm{i}(\alpha+\mathbf{w}_{p}^{(k)}\cdot\mathbf{p})}-\mathrm{e}^{-\mathrm{i}(\alpha+\mathbf{w}_{p}^{(k)}\cdot\mathbf{p})}) , the \mathbf{p} -dependent factors are \mathrm{e}^{\mathrm{i}\mathbf{p}\cdot(\mathbf{y}/\hbar\pm\mathbf{w}_{p}^{(k)})} , and the \mathbf{p} -integral produces \int_{\mathbb{R}^{N}}\mathrm{e}^{\mathrm{i}\mathbf{p}\cdot(\mathbf{y}/\hbar\pm\mathbf{w}_{p}^{(k)})}\;\mathrm{d}\mathbf{p}=(2\pi\hbar)^{N}\,\delta^{(N)}(\mathbf{y}\pm\hbar\mathbf{w}_{p}^{(k)}). (8) Step 2: \mathbf{y} -integration. Define D(\mathbf{x}):=V(\mathbf{x}+\frac{\hbar}{2}\mathbf{w}_{p}^{(k)})-V(\mathbf{x}-\frac{\hbar}{2}\mathbf{w}_{p}^{(k)}) . The delta functions set \mathbf{y}=-\hbar\mathbf{w}_{p}^{(k)} (Branch + , giving potential difference -D , exponential factor \mathrm{e}^{\mathrm{i}\mathbf{w}_{p}^{(k)}\cdot\mathbf{p}^{\prime}} ) and \mathbf{y}=+\hbar\mathbf{w}_{p}^{(k)} (Branch - , giving +D , factor \mathrm{e}^{-\mathrm{i}\mathbf{w}_{p}^{(k)}\cdot\mathbf{p}^{\prime}} ). Including the \frac{\pm 1}{2\mathrm{i}}\mathrm{e}^{\pm\mathrm{i}\alpha} prefactors from the sine decomposition: Branch + : \displaystyle-\frac{(2\pi\hbar)^{N}}{2\mathrm{i}}\,D\,f\,\mathrm{e}^{\mathrm{i}(\alpha+\mathbf{w}_{p}^{(k)}\cdot\mathbf{p}^{\prime})}, (9) Branch - : \displaystyle-\frac{(2\pi\hbar)^{N}}{2\mathrm{i}}\,D\,f\,\mathrm{e}^{-\mathrm{i}(\alpha+\mathbf{w}_{p}^{(k)}\cdot\mathbf{p}^{\prime})}. (10) Summing: -\frac{(2\pi\hbar)^{N}}{\mathrm{i}}\,D(\mathbf{x})\,f(t,\mathbf{x},\mathbf{p}^{\prime})\,\cos(\mathbf{w}_{x}^{(k)}\!\cdot\mathbf{x}+\mathbf{w}_{p}^{(k)}\!\cdot\mathbf{p}^{\prime}+\kappa^{(k)}t+b^{(k)}). (11) Step 3: prefactor and (\mathbf{x},\mathbf{p}^{\prime}) -integration. Multiplying ( 11 ) by the overall prefactor \frac{1}{\mathrm{i}\hbar}\frac{1}{(2\pi\hbar)^{N}} and using \frac{-1}{\mathrm{i}^{2}}=1 : I^{(k)}=\mathbb{E}_{(\mathbf{x},\mathbf{p})\sim f}\!\left[\frac{V\!\left(\mathbf{x}+\frac{\hbar\mathbf{w}_{p}^{(k)}}{2}\right)-V\!\left(\mathbf{x}-\frac{\hbar\mathbf{w}_{p}^{(k)}}{2}\right)}{\hbar}\;\cos\phi^{(k)}\right], (12) where we renamed \mathbf{p}^{\prime}\to\mathbf{p} . Remark 1 . The nonlocal pseudo-differential operator \Theta[V] has been reduced to a pointwise function of (\mathbf{x},\mathbf{p}) inside an expectation over f . This reduction is