Flux-Corrected Diagonal Frog: second order and positivity at all time steps

Battle Count: 50

The paper has direct relevance to quantitative finance through the Dupire forward equation for option-implied transition densities. The Fokker-Planck equation is the governing PDE for local volatility calibration, and the nonnegativity requirement corresponds to the absence of butterfly arbitrage. The FCDF schemes provide a positivity-preserving, second-order accurate method for computing these densities at all time steps, which is essential for robust calibration of local volatility surfaces. However, the paper is primarily a numerical methods contribution rather than a trading strategy or financial model paper. The financial application is mentioned in the introduction but not developed as a primary focus.

The method involves multiple sophisticated components: (1) splitting the second-order operator into M-matrix core and antidiffusive correction, (2) Picard iteration with Zalesak-type flux limiting within implicit banded solves, (3) defect-corrected time stepping with combined flux clamping, (4) active-set reformulation with semismooth Newton iteration, (5) coverage logic switching between nonlinear and linear branches based on computable thresholds. Each component requires careful implementation of banded linear algebra, flux assembly, limiter budget computation, and pattern identification. The active-set solver adds complementarity problem structure. However, all operations are O(n) per sweep, and the banded structure is preserved throughout.

The paper provides detailed mathematical formulations, algorithm descriptions, and numerical experiments with specific parameters (OU benchmark, advection-dominated benchmarks). However, no code repository is mentioned. The method relies on companion papers [Itkin, 2026] and [Itkin and Kazbek, 2026] for full context. All proofs are provided in appendices. Numerical parameters are specified but implementation details of the banded solver and active-set algorithm would require additional work to reproduce.