Protocol creates first magnonic GKP states in crystals
Ellipsoidal magnetic crystals and cavity-coupled qubits produce phase-space grid states for bosonic error correction.
physics.comp-ph
Computational Physics
All aspects of computational science applied to physics.
sort pith recommended most recent
Nodal DG scheme meets all three MHD numerical requirements
HLL flux with entropy-stable speeds plus locally divergence-free projection achieve positivity, entropy stability and divergence control in
· “A positivity preserving and entropy stable nodal discontinuous Galerkin scheme for ideal MHD”
GPUs give 2x-2.6x speedup for DFT in NAO code
Two-node benchmarks show practical acceleration for both collinear and noncollinear cases over CPU-only runs on 96 cores.
· “GPU Acceleration of Collinear and Noncollinear DFT Using a Numerical Atomic Orbital-Based DFT Code”
Opinions on AI in physics compiled to seed community talks
A discussion summary on how AI is changing research and education offers a prompt for other groups to address local issues.
Amorphous BN monolayers mix two crystallite types
o-B2N2 and o-B4N4 patches coexist stably, with analogous mixed crystallites appearing in LiCl and BCN monolayers.
· “Polymorphic crystallites model for monolayer amorphous materials”
Trapped-ion hardware runs first QLBM with non-uniform advection
Quantum circuits on IonQ processors simulate 3D transport under spatially varying velocities, highlighting readout bottlenecks and new wall-
Azulenoid carbon lattices host e/3 corner charges
C6 symmetry produces higher-order topology and fractionally quantized corner states in two organic 2D structures.
· “Second-order topology in two-dimensional azulenoid kekulene carbon lattices”
Bandit method screens oxide dopants at 14 percent the usual cost
It locates the best band-gap candidate in every trial of a 529-candidate search by replacing most DFT evaluations with surrogates.
Active learning yields largest CHNO explosives database
Surrogate model predicts detonation performance with R² over 0.98 from chemical structures.
· “Active Learning for Generalizable Detonation Performance Prediction of Energetic Materials”
LSTM memory enforces cross-field consistency in electrothermal flow
Depth-recursive mechanism cuts non-physical artifacts and beats baselines in four convective and drag regimes.
First-principles charge map predicts silicon's forbidden reflection
The famous (222) X-ray peak's temperature slide follows directly from computed valence-electron motion.
Madelung boundary term made analytic for any triclinic crystal
Direct summation with closed-form facet potentials recovers exact Madelung constants for low-symmetry lattices.
Spin kernel computation shows warm-dense LSDA mismatch
Long-wavelength limit of the finite-temperature spin XC kernel agrees with LSDA spin stiffness at low T but exposes a residual when warm and
· “Finite-Temperature Spin Exchange-Correlation Kernel of the Uniform Electron Gas”
Constrained neural models clone density functionals self-consistently
Molecular-only training data enables accurate reproduction of lattice constants and bulk moduli across metallic, covalent, ionic, oxide, and
Two-layer nets cut critical slowing down to log scaling
Diffusion models for the O(n) model train in time that grows only logarithmically with size when locality is built into the architecture.
GPU photon transport matches GEANT4 at 313x speedup
Validated sensor by sensor in kiloton-scale liquid argon, the GPU path makes full optical simulation practical.
· “Accelerating Optical Photon Simulation in DUNE with Opticks”
Cubes settle without face-parallel contact
Open-source fluid–contact solver runs 103,823 cubes: shape sets microstructure.
PET enzyme recycling leaves fluffy residue with up to 30x surface
Model predicts a dominant near-particle-sized fluffy cluster plus millions of tiny microparticles, guiding filter-vs-wait choices.
· “Residual semi-crystalline particles released during enzymatic degradation of plastics”
Unequal macroparticle weights skew PIC plasma temperatures
The algorithm's intrinsic collisions thermalize macroparticles, not physical particles, making low-weight species run hot.
Five of 40 lithium antiperovskites are vibrationally stable
Two more, Li3FS and Li3FSe, stabilize under a few percent triaxial compression.
Lattice Boltzmann computes the 3D statistical Euler limit
Randomized vortices converge in Wasserstein distance to a measure solving the Euler statistical hierarchy.
· “Computing statistical Euler limits of the Navier--Stokes equations in three dimensions”
TiFe6Ga6's ground state is a double spin spiral
Neutron scattering should see 135-degree turns between every second Fe layer and 75-degree turns between neighbors.
· “Emergent Noncollinearity and Near-Degenerate Magnetic Superlattices in AT6X6 Kagome Metals”
Magnon coupling supplies iron's missing T² resistivity
First-principles calculation matches the measured 2.2e-5 μΩ·cm·K⁻² coefficient, via phonon-magnon interplay.
72% of deposition hotspots are missed without an eddy-resolving model
In a CT-based airway model, unresolved turbulence spreads deposition by ~20% and hides most top-5% concentration peaks.
Hidden orbital pathways steer hot-carrier relaxation in MXenes
A quantum-kinetic model shows spectroscopically faint channels can still dominate energy flow.
Quantum photocurrents set the shape of MXene terahertz pulses
A multiscale model links engineered three-color laser pulses to THz output through transient photocurrents.
No-supercell route computes spin-phonon coupling from first principles
Downfolding electron-phonon coupling gives real-space exchange derivatives for any magnetic pair, shown on SrMnO3.
· “First-Principles Spin-Lattice Coupling from Downfolded Electron-Phonon Interaction”
Quantum circuits reproduce Landau damping with tunable accuracy
Paper benchmarks Trotter and THRIFT decompositions against exact Hamiltonian evolution, mapping accuracy against circuit depth.
Varying-horizon peridynamics emerges from a variational principle
Dual-horizon equations fall out of the action principle; asynchronous stepping cuts force evaluations by 30%.
No neural net beat this solver
Even a 5.8x cheaper substitute only ties the solve, and a correct safety gate is a slowdown.
Two-timescale memory reaches target loss 55 steps sooner than Muon's record
Swapping only the momentum kernel for fast and slow branches accelerates a frozen LLM training benchmark.
· “A Physical Response-and-Memory Model for Muon Optimization”
This paper proposes a particle method for the Boltzmann equation that samples…
A particle method for the Boltzmann equation samples the finite-time transition of a lifted linear operator, learning the needed Green's…
Gold nanoparticles linked by polymers can compute with heat
Tuning linker flexibility and nanoparticle temperatures makes heat-flow outputs separate two input classes, simulations show.
· “Polymer-Linked Nanoparticle Networks Running on Heat Can Act as Computing Devices”
Computer simulations of a phase-field model show that slow cooling converts a planar…
Slow cooling, but not fast cooling, allows intersecting ferroelectric domain walls to acquire over/under depth ordering in a nondimensional…
· “Mechanistic Phase-Field Modelling of Woven-Domain Formation in Ferroelectric Material”
This paper replaces the expensive matrix-exponential steps of a Fokker-Planck solver with…
A positivity criterion for rational time-stepping maps of eventually positive generators selects Padé(0,2) over Crank-Nicolson and yields a…
Bit-string symmetry cuts molecular grid points by 81%
Algebraic symmetry reduction keeps VCI energies within 1 cm⁻¹ of full grids.
· “General Symmetry-Based Potential Energy Surface Grid Reduction in Normal Coordinates”
Wire-based white matter matches FEM for EEG source imaging
No full head mesh: wire-based white matter matches FEM localization, with live tract views.
· “Volume-Surface-Wire Integral Equations for EEG Source Imaging”
Rare chaotic states get more training weight
A dynamics-aware loss delays autoregressive error collapse on chaotic systems, beating uniform and density baselines.
· “DAW: Dynamics-Aware Weighting for Deep Learning Forecasts of Chaotic Systems”
New SPH scheme holds water perfectly still to 10^-13
The scheme cancels the hydrostatic pressure gradient exactly, so resting water no longer invents spurious currents.
Defect finder works at melting heat with no hot-atom training
Localizes every self-interstitial and never invents a vacancy across FCC, BCC, and HCP iron.
· “Cross-Temperature Defect Identification in Atomistic Simulations via Multi-Level Domain Alignment”
Chiral pitch sets the scale for soliton formation in liquid crystals
A closed-form model reproduces self-focusing and astigmatic nematicons at a fraction of the cost.
Moment nudging cuts PIC plasma errors by up to 100x
A Wasserstein gradient flow turns moment observations into particle corrections for plasma PIC runs.
· “Wasserstein Moment Nudging for Vlasov-Poisson Data Assimilation”
Lower PINN residual picks worse solver in 83% of matched tests
A penalised constraint wins on residual while violating physics; three admissibility checks restore trustworthy ranking.
Neural surrogates win by mixing attention with mesh message passing
RWR interleaves global latent attention with local message passing, winning on flow fields, gradients, and wall shear
· “Read, Write, Relax: Why Neural PDE Surrogates Need Both Global and Local Processing”
Hybrid model maps solenoid field to 0.05 gauss accuracy
A physics-informed neural network cleans up what a harmonic fit misses, reaching the precision Mu2e needs.
Stability math fails to rank vortex memory's noise tolerance
Six vortex codewords hold 2.585 bits per cell; basin geometry, not eigenvalues, sets their write fidelity.
· “Attractor-Basin-Limited Fidelity in Reproducible Multistate Vortex Memory”
Cancel a decay path and a leaky atom emits thousands of photons
Coherent interference kills the off-target decay, so optical readout can collect thousands of photons instead of ten.
· “Increased cyclicity of atomic transitions via coherent interference of decay paths”
Low-rank reversibility cuts adjoint-optimization memory 500x
A few checkpoints replace the whole forward trajectory, keeping gradients accurate for kinetic plasma optimization.
Neural networks now solve quantum magnets and Hubbard models
A best-practice review shows how NQS beat the sign-problem barrier in frustrated and doped systems.
· “Neural quantum states in condensed matter: advances, best practices, and prospects”
No phase transition at J2=-1/4 in the frustrated honeycomb Ising model
At a special coupling the system stays disordered at all temperatures with residual entropy 0.23096093 per site.
· “Meandering stripes in the frustrated J₁-J₂ Ising model on the honeycomb lattice”
Liquid rings shape Schwarz-P space bricks from droplets
A 80-mm prototype matches its computed shape to 2.75 mm; an orbit experiment follows in 2027.
Symmetry checks cut 60% of noise bias in gauge-theory simulation
Small-lattice calibration recovers noiseless condensate values on larger lattices; VQE gains only better energy estimate
· “Symmetry Constrained Quantum Error Mitigation for the Schwinger Model”
Thermal scattering barely changes ptychographic phase and amplitude
Simulated total and elastic-only 4D STEM datasets reconstruct to under 2 percent difference for all atom species.
· “On Transmission Function Amplitude and Phase Recovery in Multislice Electron Ptychography”
Laser-excited lattice motion simulated ~1000x faster
Machine-learned Hamiltonian with occupation-weighted forces reproduces TDDFT coherent phonons in Si and FeSe.
· “Scalable photoexcitation-induced molecular dynamics with machine-learned Hamiltonians”
Higher-fidelity room acoustics improve speech enhancement training
On real measured rooms, hybrid-simulated training data cut word error rates more than image-source data.
Marketron model cut to nine identifiable parameters
One parameter set fits the SPX surface, turning a formerly hidden wedge into a measurable flow-risk premium.
· “Calibrating Inelastic Markets to Options: The Lean Marketron and the Generalized Langevin Equation”
Two pulses 1.2 ns apart turn chaotic gold melt into ordered arrays
The tuned delay suppresses droplet breakup, yielding ~20-nm-wide plasmonic resonances for label-free biosensing.
Resampled moments match Monte Carlo in neural forecasts
A piecewise-linear approximation plus resampling propagates uncertainty through chaotic neural surrogates.
· “Uncertainty propagation in auto-regressive random neural network models”
Modal symmetry suppresses crosstalk in laser-written waveguides
Co-designed elliptical cores and perturbation parity steer 1S and 2Py modes without opening a leakage channel.
· “Symmetry-guided modal control in elliptical femtosecond-laser-written photonic waveguides”
One physics solve recovers the hidden ranking of a surrogate library
A single shared correction matched 99.6% of pairwise preferences with no ground-truth reference
· “Shared Physics Responses Recover Hidden Rankings in Neural Operator Libraries”
A single zero-temperature run gives spectra at all temperatures
Hot bands and thermal broadening follow from one classical path plus a recomputed Gaussian width, in seconds.
ML compression's carbon payoff hinges on the power grid
Per byte processed, the ML compressor BOA emits more CO2 than ZSTD or LZMA, so the green case is conditional.
· “Green BOA: Determining the environmental break-even point for ML-based data compression”
Scattered sound reduces to a sum over resonator modes
Modal overlap integrals reproduce Mie and finite-element acoustic scattering spectra.
· “Resonant state expansion for acoustic resonators. Part II. Scattering problem”
One Gaussian wavepacket yields full resonance Raman spectra
Hagedorn wavepackets turn every vibrational transition into a cheap overlap along a single path, exact in harmonic molecules.
· “Resonance Raman spectroscopy from ab initio Hagedorn wavepacket dynamics”
A known cylinder's modes now solve perturbed acoustic resonators
Uniform, radial, and sectoral changes of density and compressibility become a matrix eigenvalue problem.
· “Resonant state expansion for acoustic resonators. Part I. Eigenvalue problem”
Stamped inputs make PDE surrogates return the wrong physics
Clean error stays low, so standard checks miss a trigger that swaps the physical regime.
Power series matches helium's second virial coefficient from 65 K up
A quantum weight generated by recursion, cut off when its imaginary part passes pi/2, reproduces the data.
Most gold nanocluster crystal structures are minor isomers
At 300 K the clusters interconvert among many shapes, so X-ray snapshots often miss the dominant geometry.
· “Dynamic Ensembles of Phosphine-Stabilized Gold Nanoclusters”
Correlations give Platonic particles their own drag laws
Particle-resolved simulations fit shape- and orientation-aware coefficients over Reynolds numbers 0.1 to 300.
Workflows as programs: compile your analysis
Flow compiles declarative workflows into backend-independent plans, so analyses can outlive their software and keep automatic provenance.
· “FAST-HEP: Compiling Declarative Analysis Workflows for High-Energy Physics and Beyond”
Hankel transforms become matrix products in CGC cross sections
Bessel oscillations are integrated in closed form, cutting a DIS dijet dipole from hours of CPU time to about two minutes on one GPU.
· “Fourier Transforms of Color Glass Condensate Multi-Wilson-Line Correlators via Filon Quadrature”
From 4 to 49: AI chemistry agents now run whole experiments
A hand-built census charts the shift from single calculations to end-to-end in silico experiments.
· “Science Done on a Machine by a Machine: AI Agents in Computational Chemistry”
Solver nails position-dependent mass to machine precision
New open-source code validates the BenDaniel-Duke operator, opening strong-field simulation to heterostructures.
· “Unified Strong-Field Dynamics Simulations from Atoms to Heterostructures”
Coupled-cluster accuracy from one DFT calculation
One network corrects a cheap DFT Hamiltonian and extrapolates seven properties beyond training size.
· “Coupled-cluster molecular properties across the main group that extrapolate beyond training size”
A single damping constant tames RNN quantum-state training
The fix lets minSR natural-gradient optimization beat Adam on 1D spin models using only ~100 samples.
· “Quantum Geometric Tensor Preconditioning for Stable Training of Recurrent Neural Quantum States”
Flow matching gains explicit energies for PDE fields
One read-out formula adds MCMC-corrected sampling, OOD scoring, and posterior inference to a single PDE model.
Tensor hypercontraction reaches chemical accuracy for HF and MP2
Six THC construction variants compared on 150 reactions; the cubic-scaling option costs little accuracy.
· “Benchmarking Least-Squares Tensor Hypercontraction Techniques for Molecular Systems”
Infinite-array theory fails for finite arrays up to 1000 elements
By 1000 elements, wideband arrays still differ from infinite-array predictions; two new predictors estimate the error.
· “Investigation of the Asymptotic Properties of Active Impedance in Large Finite Array Antennas”
Neural screen finds hard-carbon sodium anodes above 800 mAh/g
A surrogate trained on 64 structures ranks 13,000+ and points to low-density, porous carbons as the sweet spot.
Pulsed cathode bursts keep the same virtual-cathode ion drive
Kinetic simulation shows bursty emission only switches the ion-accelerating potential dip on and off.
