JAXVacua#
Type IIB flux vacua: complex-structure and axio-dilaton sector, vacuum finding, stability analysis.
JAXVacua is the core flux-vacuum engine of the StringForge ecosystem. It implements the layered pipeline from topological data to flux-vacuum solutions, built natively on JAX with automatic differentiation, just-in-time compilation, and hardware acceleration throughout.
Status#
Public.
What it owns#
jaxvacua.lcs.lcs_tree— JAX-registered pytree carrying topological data (Hodge numbers, intersection numbers, GV invariants, conifold info).jaxvacua.periods.periods— period vector, prepotential, Kähler potential at large complex structure.jaxvacua.css.css— Kähler geometry of the complex-structure sector; gauge-kinetic matrix.jaxvacua.flux_eft.FluxEFT— GVW superpotential, F-term scalar potential, D3-tadpole, SL(2,ℤ) duality.jaxvacua.flux_vacua_finder.FluxVacuaFinder— Newton solver, ISD-biased initial-guess sampling, Hessian / mass spectrum viajax.hessian.jaxvacua.flux_bounding.bounded_fluxes— flux enumeration and stochastic search with cluster parallelisation hooks.jaxvacua.sampling.data_sampler— flux / moduli sampling utilities.jaxvacua.freezer— light-field EFTs after integrating out heavy moduli (e.g. conifold freezing).
The full module-by-module breakdown lives in the architecture page.
What it consumes#
cytools—Polytope,Triangulation,Conefor KS-polytope input.jaxpolylog—jax_polylog_vmapfor the instanton sum.stringforge— for catalog / vacua-vault tooling, end users typically go throughstringforge.lcs_database.LCSDatabaserather than constructing models by hand.
Links#
Full documentation: https://jaxvacua.readthedocs.io/en/latest/
Source: AndreasSchachner/jaxvacua
Reference paper: arXiv:2306.06160 (Dubey, Krippendorf, Schachner — JHEP 12 (2023) 146)