The whole surface area.
Every public symbol in sweep. Click any equation to see its
constructor, fields, and example usage. All pages render directly from
Python docstrings via mkdocstrings. Equation-class counts follow
sweep.equations.list_equations(): aliases, the DAS
facade and the AcousticAniso factory are not counted.
Acoustic
8 classesScalar acoustic wave equations: constant density, variable density, visco-acoustic. Most-used family; default for FWI.
AcousticDefault. Second-order in time, PML boundaries.
Acoustic1stFirst-order velocity-stress formulation.
Acoustic3D3-D extension. Larger memory footprint.
AcousticVRZ · AcousticVRZ3DVariable density via vertical impedance perturbation.
AcousticLSRTMLSRTM-aware acoustic; mp reflectivity perturbation field.
AcousticLSRTM3D3-D extension of LSRTM-aware acoustic.
ViscoAcousticNearly constant-Q visco-acoustic (Zhu & Harris 2014).
Elastic
2 classesParticle velocities and stresses. Required for converted-wave physics.
Anisotropic
4 classesTTI elastic media with rotated symmetry axis. Required when tilted shales / layering matter.
ElasticTTIRotated staggered-grid TTI elastic.
ElasticTTISGAxis-aligned staggered-grid TTI elastic.
ElasticTTISG3D3-D axis-aligned staggered-grid TTI elastic; 21 Bond-rotated stiffnesses.
ElasticTTI2ndDisplacement-based TTI elastic (Oh et al. 2020).
DAS family
4 classes + facadeDistributed-acoustic-sensing strain / strain-rate operators on top of the elastic wavefield. Two formulations, Zhao and Mu, each 2D + 3D, behind the DAS facade (method="zhao" or "mu"). DASElastic / DASElastic3D and DASModeler are aliases.
Other equations
9 classes + factoryCurvilinear-grid topography, vector reflectivity (VRR) and the anisotropic (VTI / TTI) acoustic family, with its AcousticAniso factory.
Operators
3 classesSpatial-derivative operators the equations are built on; rarely instantiated directly.
Propagators
2 classes + optionsWrap any equation into a callable solver. Equation-agnostic; same API for every physics family above.
PropTorchPyTorch backend. Eager + compiled CUDA paths.
PropJaxJAX backend. Functional, jit'd, pmap-friendly.
Full · BoundarySaving · CkptGradient-memory strategies, plus the eager / CUDA option blocks and their defaults.
Parallel
2 classesSplit one model across GPUs — a tile per rank, a halo exchange per step. Shot-parallel DDP needs nothing from here.