JAXVacua

JAXVacua#

Package architecture#

The package is organised as a layered pipeline. lcs_tree is the data interface; the linear chain periods → css → FluxEFT → FluxVacuaFinder adds physics one layer at a time; orthogonal tools (sampling, flux_bounding, freezer) plug into FluxVacuaFinder; helper modules (cytools_interface, hypergeometric_models, flux_utils) feed the pipeline.

jaxvacua — Package architecture
Step 1  —  Geometry input
lcs_tree
JAX-registered pytree carrying $\kappa_{ijk}$, $a$-matrix, $c_2$, $\chi$, GV / GW invariants and conifold data
from_cytools() · from_dict() · from_file()
Step 2  —  Model construction
periods
Prepotential $\mathcal{F}(X)$, period vector $\Pi(z) = (\mathcal{F}_I,\, X^I)$, Kähler potential $K$, gauge-kinetic matrix $\mathcal{N}_{IJ}$
css  ⊇ periods
Kähler metric $K_{i\bar\jmath}$, ISD matrix $M_{AB}$, non-holomorphic completion of $\mathcal{N}_{IJ}$
FluxEFT  ⊇ css
Superpotential $W$, F-terms $D_I W$, scalar potential $V$, tadpole $N_{\rm flux}$
FluxVacuaFinder  ⊇ FluxEFT
Newton solver, Hessian, mass spectrum
Kähler cone  ·  conifold loci  ·  PFV
Step 3  —  Vacuum search
Step 4  —  Refinement & analysis
Newton refinement  +  mass spectrum
Solve $D_I W = 0 \;\Longrightarrow\; (z^*, \tau^*)$ via FluxVacuaFinder.newton_method_flux_vacua()
Hessian / mass matrix via FluxEFT.hessian and FluxEFT.mass_matrix
I/O & helpers (consumed across the pipeline)
cytools_interface
CYTools → lcs_tree
hypergeometric_models
Closed-form one-modulus families
flux_utils
Flux ↔ PFV conversions

Solid arrows are required code dependencies; dashed arrows indicate “used by” tools that FluxVacuaFinder calls into rather than stages it produces. The orange callout marks lcs_tree as the data hub — every input on-ramp eventually populates it, and every model layer reads from it. Concretely, lcs_tree carries the topological data \((\kappa_{ijk},\, c_2,\, \chi,\, n_\beta^0)\); the chain then constructs the period vector \(\Pi(z) = (\mathcal{F}_I,\, X^I)\), the Kähler potential \(K(z, \bar z)\), the GVW superpotential \(W = \int_X G_3 \wedge \Omega\), and the F-terms \(D_I W = \partial_I W + (\partial_I K)\, W\).

Subpackages#