tpt-science

Simulation and modeling substrate for the TPT Solutions science verticals: differential equations, structured-grid PDE, multi-scale simulation orchestration, probabilistic programming, rigid-body physics, quantum simulation, tomographic imaging, and astrodynamics.

Dual-licensed under MIT or Apache-2.0.

Partially published to crates.io. The 18 simulation crates are on crates.io at 0.1.0. The umbrella tpt-sci, tpt-sci-units, and the tooling/binding crates (tpt-sci-cli, tpt-sci-wasm, tpt-sci-py) are not yet published and are consumed as git or path dependencies. There is no v0.1.0 git tag yet. APIs are unstable until the first tagged release.

Status

This repo is scaffolded from tpt-rust-map/template/. Build order and crate scope follow todo.md and spec.txt. Per tpt-rust-map/TODO.md, each crate's own flagged-needs-audit / flagged-deferred status is resolved before that crate is implemented — Phase 1 (scaffolding) does not require those audits.

Crate	Domain	Status	Wrap / build	Notes
tpt-sci-ode	differential-equations	implemented	build from scratch	from-scratch ODE/DAE engine (Tsit45, TR-BDF2, ESDIRK34, variable-order BDF); dense + sparse LU; forward sensitivities; f32/f64; optional Cranelift JIT (cranelift-jit); diffsol is a dev-only oracle
tpt-sci-grid	pde	implemented	build from scratch	1-D/2-D/3-D finite-difference grids; dense and sparse (CSR) Laplacians; depends on tpt-math-linalg
tpt-sci-sim-core	simulation	implemented	build from scratch	multi-scale orchestration, ODE and diffusion sub-models, cross-scale coupling, checkpoints; depends on tpt-sci-ode, tpt-sci-grid
tpt-sci-ppl	probabilistic-programming	implemented	build from scratch (NUTS)	consolidates tpt-augur; nuts-rs wrap dropped; R-hat, ESS, divergence-rate diagnostics via Trace
tpt-sci-image	imaging	implemented	build from scratch	2-D parallel-beam CT, 3-D volume CT, and cone-beam FDK; depends on tpt-math-signal-fft, tpt-math-linalg
tpt-sci-physics-rigid	physics	implemented	build from scratch	rigid bodies with quaternion orientation, Coulomb friction, sweep-and-prune broad phase; rapier disqualified (ADR 0007)
tpt-sci-quantum	quantum	implemented	build from scratch	state-vector (≤20 qubits) and density-matrix paths with Kraus noise channels; QuantRS2 disqualified (ADR 0007)
tpt-sci-astro	astrodynamics	implemented	build from scratch	Keplerian two-body; J2/J4 secular rates; atmospheric drag; third-body (Kozai-Lidov); solar radiation pressure
tpt-sci-reaction-network	systems-biology	implemented	build from scratch	Catalyst.jl-style DSL → mass-action ODE; exact SSA, tau-leaping/CLE, minimal SBML reader, conservation laws; SDE/jump out of scope
tpt-sci-md	molecular-dynamics	implemented	build from scratch	Lennard-Jones and EAM; direct-sum Ewald; SHAKE/RATTLE constraints; cell lists; RDF; depends on tpt-math-linalg
tpt-sci-dft-classical	materials	implemented	wrap feos (feos-dft)	PC-SAFT classical/soft-matter DFT via feos; from-scratch square-gradient solver in 1-D and 3-D; MIT OR Apache-2.0
tpt-sci-kinetics	chemical-kinetics	implemented	build from scratch	depends on tpt-sci-reaction-network; Arrhenius, Langmuir–Hinshelwood (multi-site), Eley–Rideal, coverage-dependent activation
tpt-sci-chemoinformatics	cheminformatics / QSPR	implemented (v0.1, unpublished)	build from scratch	SMILES parser, molecular descriptors, Morgan fingerprints, ridge + Tanimoto-kNN QSPR; TRPA1/TRPV1 model needs user-supplied curated data (demo set is illustrative only)
tpt-sci-cfd-core	fluid-dynamics	implemented	build from scratch	depends on tpt-math-linalg; incompressible Navier–Stokes with explicit and SIMPLE schemes, k-ε/k-ω SST, unstructured FVM/FEM
tpt-sci-hemodynamics	biomechanics	implemented	build from scratch	depends on tpt-sci-cfd-core, tpt-sci-ode; 1-D compliant vessel network, Womersley flow, Windkessel 0-D/1-D/3-D coupling
tpt-sci-electrophys	biomechanics	implemented	build from scratch	depends on tpt-sci-ode, tpt-sci-grid; Hodgkin–Huxley and Ten Tusscher ionic models, anisotropic tensor diffusion, bidomain
tpt-sci-climate	earth-science	implemented	build from scratch	depends on tpt-sci-ode, tpt-math-linalg; 0-D EBM, multi-band and correlated-k radiative transfer, 3-D tracer transport, primitive-equation GCM
tpt-sci-ocean	earth-science	implemented	build from scratch	depends on tpt-sci-cfd-core; 2-D shallow water, 3-D z-level ocean with non-hydrostatic projection, nudging, EnKF, 3D-Var-lite
tpt-sci-dft-electronic	materials	implemented	build from scratch	Kohn–Sham in 1-D and 3-D real space; LDA and PBE; norm-conserving local pseudopotentials; periodic k-point band structure (1-D)
tpt-sci	umbrella	implemented	re-export	one dependency; every domain crate behind a feature (default: ode, grid, sim-core; full: all)
tpt-sci-units	shared	implemented	build from scratch	typed SI quantities and physical constants (gas constant, Stefan–Boltzmann, Earth's μ) used by kinetics, climate and astro
tpt-sci-cli	tooling	implemented	build from scratch	tpt-sci command: runs reaction-network JSON specs and writes CSV
tpt-sci-wasm	bindings	implemented	wasm-bindgen	WebAssembly exports (run_spec, EBM, CT, lid-driven cavity) plus a browser playground
tpt-sci-py	bindings	implemented	PyO3 / maturin	Python bindings; excluded from the Cargo workspace so CI does not link libpython

Choose your crate

Start from the question you are trying to answer, then follow the crate.

If you want to…	Start with	Then
Solve an ODE / DAE system	tpt-sci-ode	tpt-sci-reaction-network if the system is a chemical or compartmental network
Model a reaction network (mass-action, SSA, tau-leaping, SBML)	tpt-sci-reaction-network	tpt-sci-ode for deterministic runs
Solve a PDE on a regular grid	tpt-sci-grid	tpt-sci-sim-core to couple it to ODE sub-models
Run several models at different scales together	tpt-sci-sim-core	tpt-sci-ode, tpt-sci-grid
Fit parameters to data (Bayesian, NUTS)	tpt-sci-ppl	tpt-sci-ode for ODE-based likelihoods
Reconstruct images from projections (CT)	tpt-sci-image	tpt-sci-grid for the underlying geometry
Simulate rigid bodies and collisions	tpt-sci-physics-rigid	tpt-sci-astro for orbital motion
Simulate qubits or noisy quantum circuits	tpt-sci-quantum	—
Orbital mechanics and perturbations	tpt-sci-astro	tpt-sci-ode for custom force models
Molecular dynamics of a particle system	tpt-sci-md	tpt-sci-dft-classical for interfacial density profiles
Classical / soft-matter DFT (adsorption, surface tension)	tpt-sci-dft-classical	tpt-sci-kinetics for surface reactions
Surface or reactor kinetics	tpt-sci-kinetics	tpt-sci-reaction-network for the underlying CRN
Incompressible flow	tpt-sci-cfd-core	tpt-sci-ocean for geophysical flows
Blood flow in vessel networks	tpt-sci-hemodynamics	tpt-sci-cfd-core for 3-D coupling
Cardiac / neural excitation	tpt-sci-electrophys	tpt-sci-hemodynamics for coupled physiology
Climate or atmospheric chemistry	tpt-sci-climate	tpt-sci-ocean for coupled ocean–atmosphere work
Ocean circulation or data assimilation	tpt-sci-ocean	tpt-sci-cfd-core for the solver substrate
Electronic structure (Kohn–Sham)	tpt-sci-dft-electronic	tpt-sci-md for classical atomistic work

Suggested tutorial path

1. Foundations: tpt-sci-ode (van_der_pol) → tpt-sci-grid (diffusion).
2. Coupling: tpt-sci-reaction-network (sir) → tpt-sci-sim-core (multi_scale_cookbook).
3. Inference: tpt-sci-ppl (posterior).
4. Applied domains: pick one of tpt-sci-image, tpt-sci-astro, or tpt-sci-cfd-core; each crate's README lists its own scope and limits.

The per-crate examples/ directories are the runnable reference; the Examples table below lists them.

no_std posture

All crates in this pillar are std-only by design. The tpt-math dense linear-algebra substrate and tpt-sci-ode (which implements its own ODE engine from scratch) are themselves std crates, so every simulation/physics/quantum crate transitively requires std. None currently target no_std.

Per ADR 0001, no_std is opted into per-crate, never forced workspace-wide: the CI no_std job is an intentional no-op until a crate is explicitly confirmed no_std and built with -p <crate> against thumbv6m-none-eabi. See CHANGELOG.md for the per-crate audit table.

Depending on the crates

The 18 simulation crates are on crates.io at 0.1.0, so add them directly:

[dependencies]
tpt-sci-ode = "0.1"
tpt-sci-grid = "0.1"

The umbrella crate tpt-sci is not on crates.io yet. Depend on it through git, pinned to a commit (no release tag exists yet):

[dependencies]
tpt-sci = { git = "https://github.com/tpt-solutions/tpt-science", rev = "<commit>", features = ["full"] }

Once v0.1.0 is cut from master (steps in RELEASE.md), tag = "v0.1.0" can replace rev. The tpt-math-* substrate is resolved from crates.io, so no sibling checkout is needed.

Building

This workspace depends on the tpt-math substrate (tpt-math-numeric, tpt-math-linalg, tpt-math-signal-fft), resolved as regular versioned dependencies from crates.io — no sibling checkout needed:

git clone https://github.com/tpt-solutions/tpt-science.git
cd tpt-science
cargo build --workspace
cargo test --workspace

• Edition: 2024.
• MSRV: 1.85 (pinned via rust-version in [workspace.package]).
• Toolchain: fixed by rust-toolchain.toml via rustup.

Quickstart

use tpt_sci_ode::{OdeProblem, Method};

// dy/dt = -y  ->  y(t) = y0 * exp(-t)
let prob = OdeProblem::new(|_t, y, dydt| dydt[0] = -y[0], vec![1.0], 0.0).unwrap();
let y = prob.solve(Method::Bdf, 1.0).unwrap();
assert!((y[0] - std::f64::consts::E.recip()).abs() < 1e-6);

See the crate READMEs for the full API: tpt-sci-ode, tpt-sci-grid, tpt-sci-sim-core, tpt-sci-ppl, tpt-sci-image, tpt-sci-physics-rigid, tpt-sci-quantum, tpt-sci-astro, tpt-sci-reaction-network, tpt-sci-md, tpt-sci-dft-classical, tpt-sci-kinetics, tpt-sci-cfd-core, tpt-sci-hemodynamics, tpt-sci-electrophys, tpt-sci-climate, tpt-sci-ocean, tpt-sci-dft-electronic.

Examples

Every simulation crate ships a runnable examples/ program that is meatier than the README snippet — a good first place to see the API in action:

Crate	Example	Demonstrates
tpt-sci-ode	van_der_pol	Van der Pol limit cycle + dense trajectory (solve_dense)
tpt-sci-quantum	bell_ghz	Bell/GHZ states + measurement statistics
tpt-sci-reaction-network	sir	Full SIR run, peak-infected tracking
tpt-sci-grid	diffusion	1-D diffusion of a Gaussian bump on a uniform grid
tpt-sci-sim-core	multi_scale_cookbook	SIR → diffusion cross-scale coupling (composes reaction-network + grid)
tpt-sci-ppl	posterior	NUTS posterior + R-hat / ESS / divergence diagnostics
tpt-sci-image	reconstruction	Parallel-beam CT of a phantom (FBP)
tpt-sci-physics-rigid	collision	Spheres under gravity + walls, elastic collisions
tpt-sci-astro	propagation	LEO propagation + J2 RAAN regression
tpt-sci-md	lennard_jones	LJ fluid: velocity-Verlet trajectory + RDF
tpt-sci-dft-classical	adsorption	PC-SAFT density-profile solve (wrapped feos)
tpt-sci-kinetics	reactor	Arrhenius + Langmuir–Hinshelwood, ODE via tpt-sci-ode
tpt-sci-cfd-core	cavity	Lid-driven cavity, divergence-free projection
tpt-sci-hemodynamics	arterial_segment	Compliant vessel area/flow over a cardiac cycle
tpt-sci-electrophys	ap_wave	Action-potential propagation across a tissue sheet
tpt-sci-climate	warming	0-D EBM CO₂-doubling equilibrium warming
tpt-sci-ocean	shallow_water	Shallow-water gravity-wave propagation
tpt-sci-dft-electronic	harmonic_atom	1-D Kohn–Sham solve of a two-electron well

sim-core also ships decay_coupled (ODE↔diffusion coupling + checkpoint). Run any of them with, e.g.:

cargo run --example van_der_pol -p tpt-sci-ode

Benchmarking & coverage

• Benchmarks. criterion benches live in crates/*/benches/: tpt-sci-ode (solve), tpt-sci-grid (laplacian), tpt-sci-quantum (apply_gate), and tpt-sci-image (radon) — each of the 18 simulation crates ships a representative Criterion suite. Run a single crate with cargo bench -p <crate>, or the whole workspace with cargo bench --workspace. The CI benches job runs a shortened measurement so the suite stays fast: cargo bench --workspace --benches -- --warm-up-time 0.5 --measurement-time 1 --sample-size 10.
• Coverage. The CI coverage job produces an lcov report via cargo-llvm-cov and uploads it as a build artifact.
• Docs. The CI doc job builds docs with RUSTDOCFLAGS=-D warnings, catching rustdoc warnings and broken intra-doc links (doctests are exercised by cargo test).

See RELEASE.md for the (currently dormant) publish/versioning process.

Changelogs

Each crate keeps its own CHANGELOG.md (initial v0.1.0 entry plus an [Unreleased] section): see e.g. crates/tpt-sci-ode/CHANGELOG.md. The workspace-level CHANGELOG.md tracks cross-cutting changes.

License

Licensed under either of MIT or Apache-2.0 at your option.

Copyright (c) 2026 TPT Solutions.