lindhard

A clean-room, MIT-licensed Rust engine for Monte Carlo transport of ions and electrons in matter. It computes implantation range profiles, damage, sputtering and backscatter for ions, and energy deposition for low-energy electrons. SRIM/TRIM covers the same problem class; lindhard uses none of its code or data.

Status: early, pre-release (0.0.1). Most of milestone M0, the amorphous ion core, has landed: materials and the element table, deterministic per-ion random streams and a parallel driver, screened potentials and the scattering integral, electronic stopping, the 1D layered BCA engine with full recoil cascades, range, damage and sputter/backscatter tallies, and the CLI. Validation levels 1 (analytic checks, in CI) and 2 (code-to-code oracles) are in place, and level 3 has its first experimental dataset (B in amorphous Si; P and As remain open under #51). Known deviations are reported, not hidden: with the default inputs the computed B ranges run long against that measurement, and the Ar → Cu sputter yield sits about 2x below RustBCA's. The investigation in #61 found that the default engine sends too much energy to electronic loss at low energy, because nuclear loss is cut off at p_max while electronic loss is not. #64 added TRIDYN-style weak collisions as an opt-in (physics.weak_collisions), which brings the cascade's electronic share into line with the LNST partition but lowers the sputter yield further, so the gap to RustBCA remains open under #61. See docs/validation.md, WORK_PLAN.md for milestones and docs/architecture.md for the design. The electron engine (M1) is in progress: Mott elastic and dielectric-function inelastic cross sections, event-by-event transport with secondaries, surface barriers and insulator losses, the electron tallies and a CLI [electron] run mode have landed; PSF extraction and validation against published data are under way. Crystalline targets (M2) are in progress: cubic lattices, neighbor search, crystal-aware flight, and Debye thermal displacements have landed; hexagonal-lattice work is awaiting merge.

Quick start

cargo run -p lindhard-cli -- check examples/b_5keV_si.toml
cargo run --release -p lindhard-cli -- run examples/b_5keV_si.toml --out out/b_5keV_si

run writes summary.json and CSV profiles into the output directory. The input schema and output layout are in docs/cli.md; more inputs are in examples/.

Why another code

Some open codes already handle parts of this problem well. The gaps are in license and scope:

• SRIM/TRIM is closed source, Windows-only and non-commercial. It treats every target as amorphous and as fixed, so the target never changes with dose.
• Crystalline implant simulation is not available in any open, maintained code. That covers channeling under tilt and twist, screen oxides, and damage that builds up to amorphization. Today it means IMSIL, Crystal-TRIM, MARLOWE or commercial TCAD.
• Ion codes that are open are either copyleft (RustBCA, iradina, IM3D) or use stopping tables taken from SRIM (OpenTRIM).
• Low-energy electron transport (eV–50 keV, for e-beam lithography proximity functions, SEM and EBIC) has one permissive open code, Nebula, and it has stalled. The reference physics lives inside Geant4 or AGPL codes.

lindhard aims to be permissive, embeddable, fast, reproducible and validated:

• MIT license, with no copyleft dependency anywhere in the build.
• A library first; the CLI and the Python bindings sit on top of it (WASM later).
• Every ion runs on its own seeded random stream, so results do not depend on the thread count.
• Every model is checked against analytic limits, independent codes and published measurements (docs/validation.md).

docs/prior-art.md is the survey this project starts from. docs/history.md traces the method back to the 1947 Los Alamos neutron histories.

The name honours Jens Lindhard. His LSS theory is the standard theory of ion range, he did the foundational work on channeling, and his dielectric function is the starting point for inelastic electron scattering. Between them those cover both halves of this project.

Layout

Path	What
lindhard/	The library: materials, potentials, stopping, transport, tallies
lindhard-cli/	lindhard binary: TOML in, JSON/CSV out
lindhard-py/	lindhard Python package (pyo3 + maturin): NumPy arrays out, same TOML schema
examples/	Example CLI inputs
book/	Source of the lindhard book (physics manual and user guide), built with mdBook
tools/	Developer tools: profiling helpers
validation/	Validation harness: oracle runner, experimental datasets, results
docs/	Design, physics and project documentation (index below)

Documentation

• The lindhard book (source in book/): the physics manual, one page per model, and the user guide, including a first run from start to finish.
• docs/architecture.md: design and module plan.
• docs/cli.md: the lindhard command, input schema and outputs.
• docs/stopping-models.md: electronic stopping models, their validity ranges and sources.
• docs/validation.md: validation plan and current results.
• docs/benchmarks.md: how performance is measured, and first numbers.
• docs/data-provenance.md: where every dataset comes from.
• docs/python.md: the Python bindings.
• docs/README.md: the full documentation index.
• docs/prior-art.md and docs/history.md: the survey and the lineage.

Contributing

Read CONTRIBUTING.md first. It sets out the clean-room rule: what may be learned from other codes, what may be ported, and what may never enter this tree.

License

MIT. See LICENSE.