Historical task source. The Eve task board holds current task state.
Status: 228 / 228 tasks complete.
2026-07-13 closure audit#
The final locally actionable audit found and closed two verification gaps that were hidden by the completed checklist:
- all eleven TypeScript projects now pass their own lint targets with no errors or warnings; the cleanup also makes the XPS ordering validation exercise the computed calibrated values and validates the minimum alignment search grid;
- every phase library now completes its declaration-producing Nx build by
resolving build-time dependencies through the declarations emitted by its
upstream build, rather than pulling unrelated dependency source files into a
composite TypeScript program (
TS6307).
Fresh, cache-bypassed targeted gates passed for all eleven libraries: Nx test,
lint, and build (parallelism 1). All five Rust workspaces passed
cargo test --no-default-features (18 kernel tests total), Clippy with warnings
denied, and native-feature checks. Their Nx build:native and build:wasm
matrices also passed for spectroscopy, XAS, vibrational spectroscopy, NMR, and
EELS. A checklist recount confirms 228 checked items and zero unchecked items.
Eleven @kalika/* libraries plus five Rust napi-rs/wasm kernels, all with real
domain-specific algorithms, domain-correctness tests (checked against known
analytic/experimental values), and clean adversarial stub scans. TypeScript for
orchestration/physics, Rust for the compute-intensive kernels.
TypeScript libraries and test counts#
| Library | Sections | Tests |
|---|---|---|
@kalika/spectroscopy (+ integration adapters, PW/LCAO wavefunctions) |
124.1, 124.11.1-3 | 24 + 10 |
@kalika/xas |
124.2 (incl. SCF, curved-wave MS, core-hole DFT, BSE, M-edge multiplet) | 52 + 4 |
@kalika/photoemission |
124.3 (incl. one-step model) | 18 |
@kalika/vibrational-spectroscopy |
124.4 (incl. DFPT 2n+1) | 28 |
@kalika/nmr |
124.5 (incl. GIPAW/PAW-EFG/J-coupling) | 29 |
@kalika/mossbauer |
124.6 | 17 |
@kalika/eels |
124.7 (incl. L/M GOS) | 20 |
@kalika/spectroscopy-extended |
124.8 | 24 |
@kalika/spectrum-processing |
124.9 (incl. Athena .prj) | 32 |
@kalika/spectroscopy-viz |
124.10 | 18 |
@kalika/spectroscopy-validation |
124.11.4 | 11 |
Rust kernels (cargo build/test/clippy clean; napi + wasm-pack wired into Nx)#
| Crate | Kernel | Rust tests |
|---|---|---|
spectroscopy-kernel |
batched Voigt/Faddeeva (McLean 1994) | 4 |
xas-kernel |
Numerov radial solver, phase shifts, t-matrix | 5 |
vib-kernel |
Raman-tensor finite differences + invariants | 3 |
nmr-kernel |
MAS spinning-sideband powder integration | 3 |
eels-kernel |
loss function + Kramers-Kronig | 3 |
Each crate is crate-type = ["cdylib","rlib"] with napi behind the native
feature and wasm behind the wasm feature; cargo test --no-default-features
and cargo clippy -- -D warnings pass, and cargo check --features native
compiles the N-API bindings.
Notable validations against known values#
Cu K-edge EXAFS first shell (R=2.556 A), EXAFS fit recovering R & sigma^2; hydrogen 1s = -0.5 Ha / 2s = -0.125 Ha (1 node) / He+ = -2 Ha from the Numerov bound-state solver; self-consistent atomic potential (Ne 1s ~ -32 Ha matching ~870 eV binding, correct core/valence ordering); Haydock BSE reproducing the eigenvalue spectrum + integral-A = |phi0|^2 sum rule; d^2 Racah term energies (3F ground) and Hund's rules (d2->3F, d5->6S, f7->8S); 57Fe sextet span 10.6 mm/s at 33 T; cubic-symmetry EFG vanishing; Al/Si plasmons (15/16.7 eV); 1H Larmor 500 MHz at 11.74 T; Wigner-3j(1,1,0) = -1/sqrt3; Slater Z_eff(C 2p) = 3.25; Voigt area normalization. The validation library's ten cross-technique suites (XAS/XPS/NMR/Mossbauer/Raman/EELS/EPR/pipeline) all pass.
Approach notes#
The DFT/GW/DFPT-derived methods (core-hole DFT XANES, core-level BSE, GIPAW
shielding, PAW-EFG, J-coupling, DFPT 2n+1 Raman, the DFT/GW/BSE/phonon/defect
integration adapters) are implemented as algorithms that consume the relevant
electronic-structure / lattice-dynamics inputs and compute the observable, in
the same input-taking style as the rest of the phase, and are tested on model
inputs plus the cross-technique validation suites. No result-faking stubs were
shipped; every [ ] -> [x] is backed by implemented, tested code.