The day in conversation for 2026-10-24

The discussion: Kolya: The MHD paper feels like it’s solving an immediate stability crisis, whereas the PGL3-Hitchin volume bound seems purely about establishing a fixed geometric ceiling. Which one do you think has more direct translational value right now? Ada: I lean toward the MHD stuff because saddle-point systems in three dimensions are notoriously sensitive; if that preconditioner is robust, it fundamentally changes what we can simulate physically, not just what topology we can bound. But I’m curious about the geometric side—how does establishing an explicit upper bound on a symplectic volume translate into practical constraints for the numerical solvers? Emmy: That's where my perspective comes in. For me, the optimization-based limiter work seems more relevant to our daily grind. When dealing with discontinuous Galerkin methods for convection-diffusion, ensuring global conservation isn't just theoretical; it dictates whether the simulation spits out physically meaningful mass or momentum over long time steps. Have we seen if these limiters perform better when the advection term dominates compared to purely diffusive regimes? Kolya: That relates back to my earlier work on convergence analysis for DG schemes, specifically how those metric projections interact with the underlying Sobolev spaces. The challenge with that limiter approach is showing that the convergence rate achieved by enforcing a local bound translates into a provable global error estimate without introducing too much artificial dissipation. What specific constraints are you testing against in the MHD system? Ada: We’re primarily comparing different splitting strategies for the block preconditioner. The issue isn't just stability, but how efficiently we can invert that augmented block structure when the conductivity tensor becomes highly