Searcharxiv⌕ Search

arXiv subjects

Qiqi Rao

Publications and source records attributed to Qiqi Rao.

5 recordsLinked to original sources

Convergence analysis of physics-informed neural networks for subdiffusion

Physics-informed neural networks have emerged as mesh-free solvers for partial differential equations, yet their rigorous convergence theory for time-fractional problems remains limited. We present a convergence analysis of PINNs for linear subdiffusion equations and for the associated backward problem of recovering the initial data from terminal observations. To address the nonlocality and weakly singular kernels of fractional derivatives, we introduce a discretization-based PINN framework built on some popular time-stepping schemes satisfying discrete maximal regularity, including backward Euler, BDF2, L1, and fractional CN schemes. This approach avoids the exact evaluation of fractional derivatives at training points while preserving the stability structure needed for convergence analysis. For the forward problem, we prove that, under the natural fractional maximal-regularity condition $u\in W^{α,p}(0,T;L^2(Ω))\cap L^p(0,T;H^2(Ω)\cap H_0^1(Ω))$, the piecewise linear temporal reconstructions of PINN-loss minimizers converge to the exact solution. This result avoids the higher-order regularity assumptions on the exact solution that are often imposed in existing PINN convergence analyses. We also verify the required discrete maximal regularity and consistency properties for the aforementioned popular time discretizations. We then extend the analysis to the backward subdiffusion problem and establish convergence of the regularized recovery of the initial data from terminal observations, despite its ill-posedness. Numerical experiments for both forward and backward problems support the theoretical results, demonstrate the necessity of regularization, and show that $L^2$-type losses alone are generally insufficient. These results establish a rigorous theoretical basis for discretization-based PINNs for time-fractional diffusion problems.

math.NA↗

Tangential stability and fully discrete convergence of the classical BGN scheme for curve shortening flow

We prove fully discrete convergence of the classical Barrett--Garcke--Nürnberg (BGN) scheme for curve-shortening flow of smooth embedded closed planar curves. The main obstruction is that the mass form controls only normal motion, whereas the tangential motion is determined implicitly by the curvature equation and governs the parametrization. The usual length-decay estimate therefore does not control perturbations of the full position update. We separate temporal and spatial errors through the time-semidiscrete BGN solution. A scalar normal resolvent and exact curvature and length identities yield uniform regularity and first-order time convergence. For the spatial analysis, an adapted normal--tangential norm gives a near-contractive estimate for the linearized update. Gauss--Lobatto cancellations and an exact covariance identity for the assembled nodal normals produce an $H^1$ one-step defect of order $h^{k+1}$, while an exact difference identity controls the nonlinear remainder. For every fixed $k\ge1$, including the original piecewise linear method, we obtain the matched $W^{1,\infty}$ error bound $C(τ+h^k)$ on periodic quasi-uniform meshes with $h\le cτ^2$. All discrete steps are uniquely solvable, and the numerical curves remain regular and embedded. To the best of our knowledge, this is the first fully discrete convergence result for the classical BGN curve-shortening scheme without additional stabilization.

math.NA↗

High-order energy-stable BGN parametric finite element methods for geometric flows

We construct high-order Runge--Kutta extensions of Barrett--Garcke--Nürnberg (BGN) parametric finite element methods for curve-shortening flow and curve diffusion of planar curves and for mean curvature flow and surface diffusion of closed genus-$0$ surfaces. On each time slab $I_m=(t_m,t_{m+1}]$, the continuous equations are posed on the left-endpoint surface $Γ^m=Γ^{t_m}$ through the map $\X^{m,t}:Γ^m\toΓ^t$, whose target is the evolving surface at time $t$. For curves, this formulation follows by harmonic pullback from $Γ^0$ to $Γ^m$. For surfaces, an orientation-preserving harmonic diffeomorphism is posed separately on each time slab, and conformality yields the weight $\frac12|\nabla_{Γ^m}\X^{m,t}|^2$. Evaluating the time-slab equations at the Runge--Kutta internal times and applying mass-lumped parametric finite elements yields systems on the common domain $Γ^m$, with $Γ^{t_m+c_iτ_m}$ as the intermediate target geometry. This internal-time discretization constructs high-order BGN-structured extensions for the four flows. For an algebraically stable tableau with nonnegative weights, every exact stage solution with nondegenerate intermediate configurations satisfies monotone decay of the discrete curve length or surface area for every positive time step for which that solution exists. Radau IIA experiments exhibit energy decay for all four flows and BGN-type mesh redistribution in the curve tests. The Hausdorff self-convergence results exhibit high-order behavior consistent with the corresponding design orders.

math.NA↗

Numerical analysis of 2D Navier--Stokes equations with nonsmooth initial value in the critical space

This paper addresses the numerical solution of the two-dimensional Navier--Stokes (NS) equations with nonsmooth initial data in the $L^2$ space, which is the critical space for the two-dimensional NS equations to be well-posed. In this case, the solutions of the NS equations exhibit certain singularities at $t=0$, e.g., the $H^s$ norm of the solution blows up as $t\rightarrow 0$ when $s>0$. To date, the best convergence result proved in the literature are first-order accuracy in both time and space for the semi-implicit Euler time-stepping scheme and divergence-free finite elements (even high-order finite elements are used), while numerical results demonstrate that second-order convergence in time and space may be achieved. Therefore, there is still a gap between numerical analysis and numerical computation for the NS equations with $L^2$ initial data. The primary challenge to realizing high-order convergence is the insufficient regularity in the solutions due to the rough initial condition and the nonlinearity of the equations. In this work, we propose a fully discrete numerical scheme that utilizes the Taylor--Hood or Stokes-MINI finite element method for spatial discretization and an implicit-explicit Runge--Kutta time-stepping method in conjunction with graded stepsizes. By employing discrete semigroup techniques, sharp regularity estimates, negative norm estimates and the $L^2$ projection onto the divergence-free Raviart--Thomas element space, we prove that the proposed scheme attains second-order convergence in both space and time. Numerical examples are presented to support the theoretical analysis. In particular, the convergence in space is at most second order even higher-order finite elements are used. This shows the sharpness of the convergence order proved in this article.

math.NA↗

Convergence of Arbitrary Lagrangian-Eulerian Second-order Projection Method for the Stokes Equations on an Evolving Domain

The numerical solution of the Stokes equations on an evolving domain with a moving boundary is studied based on the arbitrary Lagrangian-Eulerian finite element method and a second-order projection method along the trajectories of the evolving mesh for decoupling the unknown solutions of velocity and pressure. The error of the semidiscrete arbitrary Lagrangian-Eulerian method is shown to be $O(h^{r+1})$ for the Taylor--Hood finite elements of degree $r\ge 2$, using Nitsche's duality argument adapted to an evolving mesh, by proving that the material derivative and the Stokes--Ritz projection commute up to terms which have optimal-order convergence in the $L^2$ norm. Additionally, the error of the fully discrete finite element method, with a second-order projection method along the trajectories of the evolving mesh, is shown to be $O(\ln(1/τ+1)τ^{2}+\ln(1/h+1)h^{r+1})$ in the discrete $L^\infty(0,T; L^2)$ norm using newly developed energy techniques and backward parabolic duality arguments that are applicable to the Stokes equations with an evolving mesh. To maintain consistency between the notations of the numerical scheme in a moving domain and those in a fixed domain, we introduce the equivalence class of finite element spaces across time levels. Numerical examples are provided to support the theoretical analysis and to illustrate the performance of the method in simulating Navier--Stokes flow in a domain with a rotating propeller.

math.NA↗