What solprop can do#

solprop solves forward pulse-propagation problems in waveguides and free space. The main model is the unidirectional pulse propagation equation (UPPE). For a single guided envelope with specified dispersion and nonlinearity, the generalised nonlinear Schrödinger equation (GNLSE) provides a convenient description. An envelope calculation is not automatically a GNLSE calculation: the spatial representation and the response model matter too.

What question does a propagation calculation answer?#

You specify a field at an entrance plane and ask how it evolves forward through a specified medium. Outputs can include the spectrum after broadening, pulse power and duration, energy in a chosen ultraviolet band, modal content, a beam’s transverse fluence, and plasma density. Parameter scans repeat that question for different inputs or structures. The calculation predicts the consequences of a chosen physical model; it does not infer an experimental geometry or material response from the desired result.

A useful workflow has four parts:

Choose the propagation equation, spatial description and material responses.
Set the input pulse or beams, geometry, grids and accuracy controls.
Run, saving enough fields and statistics for the observables of interest.
Interpret those observables and vary the numerical resolution and model assumptions that could affect the conclusion.

The RDW lesson follows all four parts. The detailed simulation setup chapters explain each choice.

Pulses and propagation#

A pulse can be described by its real electric field or by a complex envelope around a reference frequency. The field description resolves the optical carrier; the envelope description can reduce the sampling needed for a narrowband pulse. The choice also determines how carrier-sensitive responses are represented. The grid chapter explains the choices, sampling, and Fourier conventions.

Inputs include Gaussian and sech pulses, measured or custom pulses, and coherent sums of pulses with different wavelengths, delays, and phases. Spatial runs add a transverse beam profile. See input pulses and polarisation.

The UPPE combines forward linear propagation with a nonlinear source. It can carry a real field or a complex envelope, with a guided-mode expansion or a spatial grid. The GNLSE route instead describes a single envelope with prescribed dispersion coefficients and nonlinear strength. The N = 3 soliton lesson uses its cubic limit to explain compression and recurrence. In either route, an option being available does not mean that every response is enabled or valid for every geometry.

The input may have spectral phase, delay, chirp, multiple coherent components, or one or two transverse polarisation components where supported. General vector modal guides use their electric and magnetic mode fields to reconstruct physical observables. Conventions matter when importing a spectrum: solprop’s user-facing spectra use the optics Fourier sign, while custom array launches state the internal transform they require. See grids and Fourier conventions and pulses before combining data from another code.

Waveguides and free space#

For guided propagation, solprop offers effective-area models, resolved capillary mode sets, and general modal guides with electric and magnetic mode fields. Guide models include gas-filled capillaries, antiresonant fibres, step-index and graded-index fibres, rectangular and planar guides, and supplied modal data. Tapers and pressure gradients describe changes along the propagation direction. The waveguide guide explains the assumptions and supported combinations.

For free-space propagation, radial grids describe cylindrically symmetric beams, while Cartesian grids resolve both transverse coordinates. A line grid represents a field uniform in the other transverse direction. These descriptions have different computational costs, symmetries, and transverse measures; use choosing a spatial description to decide between them. Free-space energy, fluence and intensity are the physical axial Poynting flux in the medium, so a launch energy is what an energy meter there would read, and a radial gas pressure gradient is admitted under a certificate that bounds the reflection the forward model omits (energy and flux in a medium).

Description	Typical calculation	Main modelling choice
Effective-area single mode	A fibre envelope or mode-averaged hollow guide	Assume a fixed profile; no modal transfer or local transverse response
Resolved capillary modes	Few-cycle compression, RDW emission, Raman or plasma in gas	Retain enough modes and enough transverse quadrature to resolve their combined field
General modal guide	Solid-fibre or supplied vector modes	Use the guide’s mode fields and physical Poynting observables
Radial spatial grid	Symmetric focusing and an axial filament	Assume cylindrical symmetry; test radial resolution and domain radius
Cartesian spatial grid	Crossing beams or asymmetric propagation	Resolve both transverse axes and any phase gradients
Line or planar description	Propagation with one invariant or guided transverse direction	Track the correct transverse measure, including quantities per unit width

The radial filament and crossing beams show two different reasons to resolve space directly. The examples gallery adds antiresonant, solid, rectangular and planar guide demonstrations.

Material responses#

Material dispersion, nonlinear coefficients, Raman responses, and ionisation-rate models come from the optmat material library. The available responses include instantaneous and frequency-dependent electronic nonlinearity, third-harmonic generation, rotational and vibrational Raman, photoionisation and plasma, and shot noise. Gas mixtures can carry several species with their own responses.

Availability depends on the grid, geometry, and response combination. In particular, cycle-averaged envelope plasma has different assumptions from carrier-resolved plasma, and not every Raman model supports vector fields. The materials and response chapters state these limits and explain how to select a model and inspect its provenance.

The material selection is part of the scientific input. A run records the models it uses so that results can be traced to their dispersion laws, nonlinear coefficients, rate models and validity ranges. Gas pressure and temperature affect density and therefore both linear and nonlinear response. A warning about a material fit’s range should lead to an interpretation check, not merely to a smaller integration step. See materials and gas mixtures.

The instantaneous Kerr response describes intensity-dependent refraction; delayed Raman response describes molecular or material memory; photoionisation and plasma add a changing electron population and its optical response. These mechanisms can coexist and influence each other. The MI ensemble and comb-to-continuum calculation illustrate how different responses lead to different routes to broadband light.

An optional microscopic helium TDSE response is intended for research configurations with explicit calibration and convergence checks. Its manual explains its current scope and limitations; it should not be treated as a generally qualified replacement for the rate-based plasma models.

Results, workflows, and execution#

Runs can save fields and statistics in memory or to HDF5. Processing functions recover spectra, temporal traces, energies, widths, and other observables; plotting functions produce propagation maps and slices. Design tools estimate useful length scales and phase matching, while offline optics and compressors apply spectral transfer functions. Parameter scans organise repeated calculations.

Start with results and interpretation, then design and workflows. The examples show self-compression, dispersive-wave emission, supercontinuum generation, Raman dynamics, mode conversion, and beam propagation.

The NumPy reference and Numba CPU backends are included with solprop. The separate native backend provides compiled CPU and GPU execution. Choose a backend using the supported combinations and the backend guide; the installed native build determines whether it executes on a CPU or GPU.

Scope and limits#

The propagation direction is privileged. These are forward models, so reflected waves and a general bidirectional Maxwell problem are outside their scope. A chosen spatial symmetry excludes the corresponding symmetry-breaking dynamics: a radial calculation cannot predict azimuthal filament breakup, and a one-mode calculation cannot predict transfer into unrepresented modes. Some guide models are asymptotic or idealised and require dimensions large compared with wavelength. Material data and response approximations have their own independent limits.

Numerical accuracy also has several independent controls. A small local step error does not establish adequate temporal bandwidth, transverse sampling, quadrature, modal basis size or boundary distance. Conserved energy alone does not prove a correct pulse shape. Begin with the observable you intend to trust, then use the accuracy chapters to identify the refinements that can change it.

Backend support is documented by combination of geometry, grid and response. A faster backend does not introduce a different physical model, and an unsupported combination is refused rather than silently substituted. Research features such as the optional TDSE response require the additional calibration and convergence work stated in their own chapters. Qualification reports are evidence for specified cases, not universal accuracy guarantees.

Continue with choosing a spatial description, then use the first lesson to put these choices together.