jaxvacua.flux_utils.map_to_fd

Contents

jaxvacua.flux_utils.map_to_fd#

map_to_fd(model, moduli, tau, fluxes, enabled=False)#

Optionally map (moduli, tau, fluxes) to the \(\text{SL}(2,\mathbb{Z}) \times\) monodromy fundamental domain via FluxEFT.map_to_fd(), handling the numpy/JAX boundary so call sites in eager Python loops stay clean.

Details

The real mathematical work (\(\text{SL}(2,\mathbb{Z})\) action on \(\tau\) + monodromy shifts on the moduli axions + boundary snap) is delegated to model.map_to_fd(). This helper adds three pieces of bookkeeping on top:

  1. Opt-out short-circuit: if enabled=False the inputs are returned unchanged. Lets call sites be written without an outer if block.

  2. numpy ↔ JAX marshalling: inputs are cast to JAX (jnp.asarray, complex); outputs come back as numpy / Python scalars (np.asarray with int32 fluxes, complex tau). Intended for eager-Python sampling loops that consume the result on the numpy side.

  3. Output dtype contract: fluxes is returned as int32 (they represent integer flux quanta).

Parameters:

  • model (Any) – Finder instance providing map_to_fd(). Passed explicitly so the helper works with any FluxEFT subclass.

  • moduli (Any) – Complex structure moduli.

  • tau (Any) – Axio-dilaton.

  • fluxes (Any) – Flux vector.

  • enabled (bool) – If False, the helper is a no-op and returns the inputs verbatim. If True, the FD mapping is applied. Defaults to False.

Returns:

  • Tuple[Array, complex, Array] – (moduli_fd, tau_fd, fluxes_fd).

  • When ``enabled=True`` these are numpy arrays / Python ``complex``;

  • when ``enabled=False`` they are the original input objects unchanged.

Return type:

Tuple[Any, Any, Any]

Contents