thunder4d.inputfile

Standard THUNDER4D input file (JSON): one file describes one simulation completely (laser, gas, cell, mirrors, grid, solver, run), so that a design exported by the design helper, a GUI session or a script can be re-run from the file alone.

import thunder4d as th

cfg = th.load_input("cell.json")        # defaults filled in, checked
b = th.run_input(cfg)                   # build + run; b.sim holds the results
th.save_input(cfg, "copy.json")

Units are in the key names (_J, _fs, _nm, _m, _bar, _K). Tables (measured spectrum, mirror reflectivity and GDD curves) are stored inline as {"wavelength_nm": [...], "value": [...]}. Unknown keys are rejected (a typo would otherwise be ignored silently), except in the free-form design and notes blocks. Every missing required field is listed in one error message.

Block	Keys (required in bold)
laser	energy_J, duration_fs (transform-limited FWHM), wavelength_nm, shape ("gaussian" | "measured"), spectrum, gdd_fs2, tod_fs3
beam	injection ("linear" | "nonlinear" Kerr-matched | "waist"), w0_m
medium	gas, pressure_bar, temperature_K, n2_1bar (None: thunder4d.materials)
cell	roc_m, separation_m, n_passes, herriott (information: round_trips, k, branch)
mirrors	reflectivity and gdd_fs2 (scalar or table, per bounce), tod_fs3, diameter_m, coating (information)
grid	preset ("coarse" | "medium" | "fine"), box_factor, points_per_waist, time_window_factor (override the preset), lambda_min_nm, lambda_max_nm, precision ("single" | "double"), device ("auto" | "cpu" | "gpu"), fft_backend ("auto" | "scipy" | "fftw")
solver	phi_max, gouy_step, self_steepening, kernels ("auto" | "numba" | "numpy")
run	record_every, near_field, record_mirror
design	free form: provenance written by the design helper (predictions, margins)

thunder4d.inputfile.default_input(**blocks)

A complete input with the defaults; keyword blocks are merged in, e.g. default_input(laser={"energy_J": 1e-3, "duration_fs": 200}, medium={"gas": "Ar", ...}).

thunder4d.inputfile.load_input(path_or_dict)

Read a file (or take a dict), fill in the defaults and check it (ValueError listing every problem, KeyError for an unknown key).

thunder4d.inputfile.save_input(cfg, path)

Write an input, defaults filled in, as indented JSON.

thunder4d.inputfile.build_simulation(cfg, verbose=False)

Input -> object with cell, grid, pulse and sim (a LensFrameSimulation ready to run), plus w0_injected_m.

thunder4d.inputfile.run_input(cfg, verbose=True, max_seconds=None)

build_simulation() then sim.run(); returns the built object.

thunder4d.inputfile.input_problems(cfg)

Every problem of a merged input, as a list of (path, message), where path is the key at fault ("cell.separation_m"); empty when the input is valid. load_input() raises with these messages.

thunder4d.inputfile.describe_input(cfg)

Quick checks of an input before launching it, in milliseconds (neither the pulse nor the solver is built): linear eigenmode (w0_m, w_mirror_m, zR_m, gouy_per_pass_rad), grid size (grid_shape, field_MB), Gaussian estimates of the peak power, of p_over_pcr and of the peak fluence on the mirrors (mirror_fluence_J_cm2).

thunder4d.inputfile.FIELDS

How each key is shown and edited (label, displayed unit, kind of editor, help, conditions), as {"block.key": Field}. The GUI builder is generated from it, and a test checks that every key of the input file has an entry: a new key gets its Field in the same change.