Backend
Entropy Stable Element-Local Continuous-Discontinuous Residual Correction Summation-By-Parts Discretization
arXiv cs.NA0 views
arXiv:2610.25388v1 Announce Type: new Abstract: Residual correction provides an efficient alternative to two-point volume flux differencing for enforcing entropy stability in high-order discretizations of hyperbolic partial differential equations. We develop an element-local residual-correction summation-by-parts (SBP) method that combines continuous-SBP (C-SBP) and discontinuous-SBP (D-SBP) coupling on conforming meshes. The construction retains continuous assembly without simultaneous approximation terms (SATs) wherever suitable, restricts SATs to selected discontinuous facets, and permits the coupling to change in time as a foundation for adaptive discretizations. The scalar entropy defect and a graph-based distribution vector acting on conservative variables are computed entirely from element-local data, without neighboring states, auxiliary gradients, or correction SATs. The same correction applies to every facet partition, including the pure C-SBP and D-SBP limits. For fixed partitions, we establish conservation and entropy identities for each element residual, exact global conservation, entropy stability up to an explicit regularization defect, and design-order consistency. Changes in the partition are handled by a mass-weighted orthogonal projection that preserves conservative moments and convex admissible sets, does not increase entropy, and does not amplify error in the broken norm. Numerical experiments in one to three dimensions show that the hybrid method retains the smooth-solution accuracy of its D-SBP endpoint under repeated repartitioning. The nonlinear tests demonstrate robustness comparable to Hadamard-form two-point-flux discretizations, particularly when coupled with artificial dissipation.
Source:https://arxiv.org/abs/2610.25388PreviousImproved algorithms for computing the matrix logarithm based on the double-exponential formulaNextA conservative reduced collision model for the Boltzmann equation
Related posts

Part 2: From Five Stars to One Sentence: Semantics for Custom Compose Controls
0dev.to
A Four-Line Fix for a 9.8: Anatomy of Linux's TCP Handshake Use-After-Free
0dev.to
The Building Is a System: What Software Engineers Can Learn from Construction Site Failures
0dev.to

Part 4: When the Screen Changes Silently: Live Regions, Headings and Focus in Compose
0dev.to
FlatCake: Why Tiered KV Cache Storage Is Holding LLM Serving Back
0dev.to
Why Scrum is Not a Silver Bullet: 5 Structural Drawbacks and Fixes
0dev.to