Experimental Modules

sci-form organises pre-production algorithms into three tiers based on maturity:

Tier	Stability	API guarantee	Coverage
Alpha	Proof-of-concept	May break between minor versions	Basic integration tests
Beta	Validated	Minor-version bumps only	Regression + reference validation
Experimental (legacy)	Track-by-track	See individual track page	Varies

---

Alpha Modules →

New methods gated by alpha-* feature flags. Functionally correct but API shapes may still change.

ID	Module	Feature flag	Description
A1	alpha::dft	alpha-dft	Kohn-Sham DFT (SVWN / PBE)
A2	alpha::reaxff	alpha-reaxff	ReaxFF reactive force field
A3	alpha::mlff	alpha-mlff	Neural network FF with AEV descriptors
A4	alpha::obara_saika	alpha-obara-saika	Obara-Saika ERIs + Boys function
A5	alpha::cga	alpha-cga	CGA motor algebra dihedral refinement
A6	alpha::gsm	alpha-gsm	Growing String Method
A7	alpha::sdr	alpha-sdr	Semidefinite relaxation embedding
A8	alpha::dynamics	alpha-dynamics-live	Live interactive MD
A9	alpha::imd	alpha-imd	IMD wire-protocol bridge

// WASM: subpath import
import { alpha_compute_dft, alpha_compute_reaxff_gradient } from 'sci-form-wasm/alpha';

# Python: submodule
from sci_form.alpha import dft_calculate, reaxff_gradient

---

Beta Modules →

Validated methods gated by beta-* feature flags. API is stable within minor versions.

ID	Module	Feature flag	Description
B1	beta::kpm	beta-kpm	Kernel Polynomial Method — O(N) DOS
B2	beta::mbh	beta-mbh	Mobile Block Hessian
B3	beta::randnla	beta-randnla	Randomized Nyström EHT — O(N k²)
B4	beta::riemannian	beta-riemannian	Riemannian PSD cone geometry
B5	beta::cpm	beta-cpm	Constant Potential Method

// WASM: subpath import
import { beta_compute_kpm_dos, beta_solve_eht_randnla } from 'sci-form-wasm/beta';

# Python: submodule
from sci_form.beta import kpm_dos, eht_randnla

---

Legacy Experimental Tracks

The older experimental-* feature flags remain available at their existing module paths:

All experimental modules live under sci_form::experimental::* and require explicit feature flags to compile.

Track Summary

Track	Module	Feature Flag	Description
E1	cga	experimental-cga	Conformal Geometric Algebra G(4,1)
E2	rand_nla	experimental-randnla	Randomized NLA for O(N log N) EHT
E3	riemannian	experimental-riemannian	Riemannian L-BFGS for ETKDG
E4	kpm	experimental-kpm	Kernel Polynomial Method — O(N) DOS
E5	eeq	experimental-eeq	Dynamic EEQ Charge Model
E6	alpb	experimental-alpb	Analytical Linearized Poisson-Boltzmann
E7	d4	experimental-d4	DFT-D4 Dispersion Correction
E8	sdr	experimental-sdr	Semidefinite Relaxation Embedding
E9	mbh	experimental-mbh	Mobile Block Hessian
E10	cpm	experimental-cpm	Constant Potential Method
E11	gsm	experimental-gsm	Growing String Method

Groups

Group A — Advanced Geometry

Track	Focus
E1: CGA	Unify rotations/translations via geometric algebra motors
E2: RandNLA	Sub-cubic EHT diagonalization with randomized sketching
E3: Riemannian	PSD-guaranteed embedding via manifold optimization

Group B — Electronic Structure

Track	Focus
E4: KPM	Linear-scaling DOS and density matrix via Chebyshev expansion
E5: EEQ	Geometry-dependent electronegativity equalization charges
E6: ALPB	Analytical implicit solvation with Born radii

Group C — Precision Methods

Track	Focus
E7: D4	Geometry-dependent dispersion with BJ damping
E8: SDR	Convex SDP embedding eliminating the retry loop
E9: MBH	Reduced-DOF vibrational analysis for large molecules
E10: CPM	Electrochemical charge equilibration at fixed potential
E11: GSM	Reaction path and transition state search

---

Enabling Features

Rust

[dependencies]
# Alpha modules
sci-form = { version = "0.11", features = ["alpha-dft", "alpha-reaxff"] }

# Beta modules
sci-form = { version = "0.11", features = ["beta-kpm", "beta-randnla"] }

# Legacy experimental
sci-form = { version = "0.11", features = ["experimental-d4", "experimental-kpm"] }

# Everything + parallel
sci-form = { version = "0.11", features = [
    "alpha-dft", "alpha-reaxff", "alpha-mlff",
    "beta-kpm", "beta-randnla",
    "experimental-d4",
    "parallel"
] }

Run tests

# Alpha tests
cargo test --features alpha-dft,alpha-reaxff,alpha-mlff --lib --test 'alpha_*'

# Beta tests
cargo test --features beta-kpm,beta-mbh,beta-randnla --lib --test 'beta_*'

# Legacy experimental tests
cargo test --features experimental-d4 --test regression -- test_d4

---

Design Principles

1. Isolation — Experimental code never modifies stable modules (src/conformer/, src/eht/, etc.)
2. Feature gates — Each track compiles only when its flag is enabled
3. Independent testing — Each track has its own test suite under tests/experimental/
4. Graduation path — Tracks may be promoted through alpha → beta → stable after validation