Selecting and inspecting material data#

Start with the recommended material models, then change a model only when the measurement or spectral range calls for it. Record both the material data and the response approximation. This chapter shows how to select, inspect and supply data; material laws explains their units, density scaling and validity.

Materials from optmat#

optmat is a curated library of optical material data: every model states its sources, its qualified range, its reference state and a status, and carries an identity whose SHA-256 digest changes whenever its numbers can. solprop takes its gases’ linear index, equation of state and transport properties, and its solids’ complex indices, from optmat; it keeps the propagation physics (linear density scaling, the mixture rule, the flow problem, the mode equations). It also takes each gas’s electronic Kerr coefficient, at the run’s carrier, fused silica’s nonlinear response — its total \(n_2\), Raman fraction and Raman shape — from optmat’s silica bundle, each gas’s ionisation rate (rates from optmat) and a gas fill’s Raman response (Raman responses from optmat). An omitted selection is MaterialSelection(), which takes optmat’s recommended models. The material integration proposal records how the data are mapped into solprop.

Selecting gases#

A MaterialSelection names the models. Its linear field is "recommended" (optmat’s recommended model of each gas), an optmat model name ("borzsonyi_2008/sellmeier_Ar", from optmat.gas.models.registry), or an optmat model object. solprop.physdata.species() and the mixture constructors take it, and without one take MaterialSelection():

from solprop.physdata import Mixture, MaterialSelection, species

argon = species("Ar")  # fit/Ar, eos/Ar, transport/Ar, and its
# hyperpolarizability/Ar, evaluated at a run's carrier
fill = Mixture.fill({"He": 0.5, "Ar": 0.5}, 3.0)
air = Mixture.air(1.0)  # CIPM-2007 "dry air (N2, O2, Ar)"
other = species("Ar", selection=MaterialSelection(linear="borzsonyi_2008/sellmeier_Ar"))

The species then carries an OptmatSusceptibility and an OptmatGas. A supplied Species in a mixture is taken as it is and never matched to a built-in gas by its name, so a custom species keeps its own law beside optmat-selected ones.

electronic and carrier_wavelength are resolved as the next three sections describe: "recommended" and "carrier-kerr" (one policy) take the Kerr coefficient at the carrier, and the dispersive policies "gas-polynomial" and "symmetric-kerr" take optmat’s whole hyperpolarisability model (below). raman selects a gas response in a gas-filled guide or a silica bundle for the GNLSE; the permitted objects differ between those contexts. ionisation names the run’s rate models: plasma=True takes optmat’s rate for each species, its plasma loss pays that rate’s own energy per event (energy_loss="peak-weighted"), and a custom species needs an explicit rate; see rates from optmat. raman names a gas fill’s Raman response and raman_shift its density-shift approximation (Raman responses from optmat).

The electronic Kerr coefficient at the carrier#

optmat fits each gas’s electronic second hyperpolarisability \(\gamma_T(\omega)\), with its dispersion, in the Taylor convention \(p^{(3)} = \gamma_T E^3/6\). In solprop’s \(P^{(3)} = \varepsilon_0\chi^{(3)}E^3\) the per-particle coefficient is therefore

\[ \gamma^{(3)}(\lambda_c) = \frac{\gamma_{T,\mathrm{Kerr}}(\lambda_c)}{6\varepsilon_0}, \qquad \chi^{(3)}(z) = \sum_s N_s(z)\,\gamma^{(3)}_s(\lambda_c), \]

which optmat’s public gamma3("kerr", wavelength) returns. A built-in species carries the model, an OptmatKerr, not a number: its gamma3 is None, and kerr_coefficient() evaluates the model at a carrier \(\lambda_c\). A run’s carrier is its grid’s reference wavelength, which the drivers build at lambda0; the selection’s carrier_wavelength may be left unset, and if it is given it must agree with that, or the run is refused — one carrier sets every optmat species’ coefficient. The Kerr fits hold from 400 nm to 20 µm and warn once per run outside.

This is the carrier-kerr approximation (SEP 3, section 4.3): one frozen coefficient per species and run, used for every process the cubic term drives — self- and cross-phase modulation and, with thg=True, the third harmonic. It is not broadband four-wave mixing or process-correct third-harmonic generation, and the carrier is not tracked as the spectrum moves (the dispersive policies below are the alternative). optmat’s distinct third-harmonic coefficient measures the cost: the run’s manifest records, for each optmat population, the Kerr coefficient used, the third-harmonic one at the same fundamental and their ratio (1.18 for argon at 800 nm), or no ratio where optmat’s third-harmonic evaluation leaves its fitted data (N₂O at 800 nm). No Raman amplitude is multiplied by the electronic coefficient, and no local-field factor is added.

The nonlinear index keeps the exact conversion with the linear index \(n_0\), \(n_2 = 3\chi^{(3)}/(4\varepsilon_0 c n_0^2)\): optmat’s own gas n2 takes \(n_0 = 1\) (1.6 % high for 30 bar of argon) and is never imported. chi3(..., wavelength=) and n2(..., wavelength=) give the selected values (or take the carrier from selection=’s carrier_wavelength); a Mixture evaluates its populations at a carrier with gamma3_path(wavelength), which solprop.tools.soliton_params() passes its wavelength.

from solprop.physdata import MaterialSelection, chi3, n2, species

selection = MaterialSelection()
argon = species("Ar", selection=selection)
argon.kerr_coefficient(800e-9)  # 1.476e-51 m^5/V^2
chi3("Ar", pressure=3.0, selection=selection, wavelength=800e-9)
n2("Ar", pressure=3.0, wavelength=800e-9, selection=selection)

The dispersive electronic policies#

MaterialSelection(electronic="gas-polynomial") and "symmetric-kerr", from SEP 6, evaluate optmat’s whole fit, \(\gamma_T=\gamma_0[1+AS_2+BS_2^2+B'S_4]\) over each quartet’s four signed vacuum wavenumbers, at the absolute frequencies of every quartet the cubic couples, instead of one coefficient: the polynomial itself, or a symmetric factorisation of its Kerr diagonal. Each built-in gas resolves its fit with optmat’s support record of it, the fitted frequency sums and the resonance levels, as a DispersiveKerrModel in SI. From the record solprop derives the gas’s leg bound and band edge; every leg beyond the bound is held there, the run’s band is labelled supported, bounded or best effort, and a run is never refused for the frequencies its field occupies. solprop writes none of those numbers down: a changed record changes the run’s identity. carrier_wavelength is not used by the electronic response under these policies — they have no carrier — but still sets the Raman pump dispersion and the ionisation rate.

The policy is the run’s: under it every optmat gas of a capillary run, its mixture route and its scans gets a DispersiveKerrResponse, one per population. A species without optmat’s model (a supplied one) is refused by name rather than given a borrowed curve, a constant KerrResponse declared beside the policy is refused, and so is the GNLSE route, whose coefficient is constant. On the general modal route a region declaring its gas fill (GasKerr) takes the policy. The manifest’s electronic entries record each population’s model and support digests, leg bound, band edges and regions in place of the carrier coefficient. The responses run on numpy_ref, numba_cpu and the native backend, on its CPU build and its qualified device builds. The physics, the two evaluators, the regions and how to choose between the policies are in a frequency-dependent electronic response.

Fused silica’s nonlinear response#

optmat’s silica bundle (SilicaBundle) holds fused silica’s total long-pulse \(n_2(\lambda)\) (the Kerr fit kerr/SiO2), a Raman shape \(h_R\) of unit integral and that shape’s own Raman fraction \(f_R\), so that the three cannot be mismatched. The electronic part is \((1 - f_R)n_2\) and the delayed part \(f_R n_2\), convolved with \(h_R\).

bundle	shape	\(f_R\)	used for
`"default"`	Hollenbeck and Cantrell (2002), paired with the Kerr fit	0.2188	the scalar GNLSE; copolarised only
`"tensor"`	Lin and Agrawal (2006), with Hellwarth’s isotropic \(a\) and anisotropic \(b\), \(a + b = h_R\)	0.245	a vector silica response

Both hold the same total \(n_2\), so a scalar and a vector run agree for long pulses; the vector split’s electronic part is 3.3 % smaller and its Raman part 12 % larger. prop_gnlse takes the bundle’s \(f_R\) for an omitted raman_fraction with Raman on, its sampled \(h_R\) as the response, and n2="recommended" as its \(n_2\) at the carrier (GNLSE). A material region’s tensor response takes the tensor bundle as SilicaTensorResponse(parameters=SilicaBundle.resolve(tensor=True)): its isotropic channel is \(h_0 = a + b/3\) and its rank-two channel \(h_2 = \tfrac23 b\), which the seam weights \((1, 0)\) and \((\tfrac14, \tfrac34)\), so a copolarised probe sees \(a + b\) and a cross-polarised one \(a\); bundle_response() builds the same two terms as a RamanResponse. A tensor response of the copolarised-only default bundle is refused by name. A general modal guide’s SilicaTensorResponse() is this tensor bundle by default, and the bundle carries an identity into the manifest and the guide’s record. Lin and Agrawal’s own time constants and weights are given explicitly, as SilicaTensorResponse(parameters=LinAgrawalRaman(...)) (LinAgrawalRaman), and are the caller’s; the two kernels agree to \(4\times10^{-16}\).

optmat’s silica raman_coefficient and electronic_susceptibility include its index \(n_0^2 \approx 2.1\): \(K = 2n_0^2\varepsilon_0 c f_R n_2\) and \(\chi_e = \tfrac43 n_0^2\varepsilon_0 c(1 - f_R)n_2\). solprop’s GNLSE calibration and its vacuum-normalised amplitudes take \(n_0 = 1\) in both (conventions §4), so only \(n_2\), \(f_R\) and the normalised \(h_R\) are imported; a test records the \(n_0^2\) ratio.

Selecting a run’s materials#

The drivers take the same selection as materials=: prop_capillary, prop_waveguide, prop_gnlse, prop_free, prop_radial and prop_planar, the Simulation.from_* constructors behind them, and solprop.sol’s re-exports; an omitted one is MaterialSelection(). What the driver resolves itself comes from optmat; what the caller built is taken as it is.

driver	resolved under the selection
`prop_capillary`	the gas by name, a fraction mapping’s names, `"air"` and `Gas.Air` (the composition `"dry air (N2, O2)"`), and the default cladding of a numbered mode set, which becomes optmat’s recommended silica `fit/SiO2`
`prop_waveguide`	the gas, the fraction mapping and `"air"`; the supplied modes keep their claddings. A provider’s material regions are its own; a region’s `SilicaTensorResponse()` is optmat’s tensor silica bundle by default
`prop_free`, `prop_radial`	a `SpatialMedium` naming a built-in gas becomes the one-population mixture of that gas under the selection, at the same pressure and temperature; a supplied susceptibility is kept beside optmat’s equation of state. The declared responses are the caller’s: write `KerrResponse(gamma3("Ar", wavelength=lambda0))` for optmat’s coefficient at the carrier. An optmat rate’s energy per event is bound to the input’s peak field, as on the capillary route
`prop_planar`	the slab modes’ fill of a built-in gas with its default susceptibility, which takes the selection’s single species at the slab’s own pressure and temperature (no mixture is needed): its law, and its equation of state for the density. The manifest records the species and its dilute-scaling check, as for a mixture’s population. A slab declared without a cladding takes optmat’s recommended silica. A supplied susceptibility, cladding or response is the caller’s. An optmat rate’s energy per event is bound to the input’s peak field, as on the capillary route. Gas mixtures in a planar guide remain unsupported, with or without a selection
`prop_gnlse`	fused silica’s response from optmat’s silica bundle: the Raman fraction an omitted `raman_fraction` takes with Raman on, the Raman shape, and `n2="recommended"`; the fibre’s dispersion and a given `gamma` or `n2` are the caller’s

With Raman on, a capillary or waveguide fill resolved under the selection takes optmat’s gas Raman response: one response for the fill’s whole composition, built only from the collision partners optmat curates, and refused by name for any other mixture (Raman responses from optmat).

A supplied mixture, mode set, cladding, glass or susceptibility is never rebuilt or matched to a built-in by its name.

from solprop.driver import Simulation

simulation = Simulation.from_capillary(
    50e-6,
    4e-4,
    "Ar",
    1.0,
    lambda0=800e-9,
    wavelength_limits=(300e-9, 3e-6),
    trange=100e-15,
    fwhm=30e-15,
    energy=1e-6,
    plasma=False,
    n_r=8,
    n_save=2,
)
manifest = simulation.material_manifest()
sorted(record.name for record in manifest.records)
# ['eos/Ar', 'fit/Ar', 'fit/SiO2', 'hyperpolarizability/Ar', 'transport/Ar']

Each run is one evaluation context: its construction, its translation, its record and its run share one log of what has been warned about, so a range, status or dilute-scaling warning is raised once per run, however many modes or stages evaluate the material. An adapter also evaluates each frequency array once: a translation reads a material at every sample of its tables along z, and a repeated array takes the stored index or polarisability, bitwise the same, with the same warnings and records. The run writes its material manifest into its output (output schema 8): every optmat model it resolved, its species, compositions and solids with their axes, the policies in force (the carrier-kerr law and its carrier, or a dispersive law with its regions; a GNLSE run’s held-back Raman fraction and where it came from), the dilute-scaling error over its band, the bands outside a model’s qualified intervals, each population’s carrier Kerr coefficient with its third-harmonic discrepancy or its dispersive model’s record, and the citations of its other material objects (output). A resume, or the reuse of a completed scan point, is refused when any of its identities, policies or the optmat version differs, naming the entry that changed. A scan resolves the selection at each point, from the point’s own objects, so no point depends on what was resolved before it.

Units, reference state and operating state#

optmat works in SI throughout: pressures in pascals, vacuum wavelengths in metres. At solprop’s interface pressures stay in bar and are converted by exactly \(10^5\); an angular frequency becomes \(\lambda = 2\pi c/\omega\), and \(\omega = 0\) is optmat’s static limit \(\lambda = \infty\). Derivatives are taken at fixed state, with \(\mathrm{d}/\mathrm{d}\omega = -(\lambda/\omega)\,\mathrm{d}/\mathrm{d}\lambda\) inside optmat, which returns them in \(\omega\) directly.

A source’s reference state is where its data were measured or reduced to; it is recorded with the model and is not the simulation’s operating state. optmat’s gas models give the per-particle polarisability \(\alpha(\omega)\), which is a dilute, zero-density quantity: there is no reference density to divide out, and reference_state records zero pressure and the temperature the source refers it to. The operating state is the fill’s own pressure and temperature, 293.15 K unless given.

The gas index at linear density scaling#

The susceptibility is solprop’s existing linear law with optmat’s refractive polarisability (which already includes the magnetic term),

\[\chi(\omega; z) = \sum_s \frac{N_s(z)\,\alpha_s(\omega)}{\varepsilon_0},\qquad n - 1 = \frac{\chi}{\sqrt{1+\chi}+1},\]

with optmat’s exact \(\partial_\omega\alpha\). This is the dilute limit of optmat’s own constitutive law, Lorentz–Lorenz with a refractivity virial, and differs from it by a relative amount of order \(N\alpha/(3\varepsilon_0)\). At 800 nm and 293.15 K:

gas	1 bar	10 bar	30 bar
He	−0.002 %	−0.016 %	−0.047 %
Ar	−0.019 %	−0.19 %	−0.57 %
Xe	−0.053 %	−0.55 %	−1.9 %
SF₆	−0.055 %	−0.62 %	condensed

A fill with an optmat-selected gas therefore measures this error at each gas’s largest density along the fill, over its band, when its spectrum is sampled (dilute_scaling_checks()), and records it in the run’s manifest whatever its size. Above DILUTE_SCALING_THRESHOLD, 1 % relative in \(n-1\), it raises one DiluteScalingWarning per gas, with the value; the run keeps the dilute law. At 800 nm and 293.15 K argon and nitrogen pass the bound near 52 and 53 bar, krypton near 31 bar, xenon near 17 bar and SF₆ near 15 bar. A finite-density law is SEP 11’s.

Gas.Air is optmat’s named composition "dry air (N2, O2, Ar)": the mole-weighted sum of its constituents’ polarisabilities, with its own identity and derivative. OptmatSusceptibility.resolve(Gas.Air, "dry air (N2, O2)") is the argon-free composition. Mixture.air takes its fractions from the same named compositions, renormalised over the constituents they keep.

Density and transport#

OptmatGas is optmat’s real-gas law \(N = p/(Zk_\mathrm{B}T)\) and its exact inverse, on the gas’s phase domain: below saturation, or above the critical temperature. A condensed state is refused, with the gas, the state and its saturation pressure named: SF₆ at 30 bar and 293.15 K is refused because SF₆ condenses at 21.0 bar there. Zero pressure is zero density, and back, without calling optmat. Gas.Air takes the density of CoolProp’s pseudo-pure dry air, which includes the unlike-pair interactions, and that fluid’s molar mass, 0.0289655 kg/mol, for every mass density, sound speed and mean free path; the constituents’ mean, 2 × 10⁻⁴ lower, enters only the viscosity’s Wilke rule. A mixture keeps the partial-pressure rule: each population’s density is its pure gas’s at its partial pressure. The two differ for air: the partial-pressure populations of Mixture.air are \(1.9\times10^{-4}\) less dense than the pseudo-pure air at 1 bar, \(1.8\times10^{-3}\) at 10 bar, while their polarisabilities are identical.

For a fill of optmat-selected species, solprop.gasflow takes optmat’s transport properties (optmat.gas.transport): the zero-density viscosity, the ideal-gas heat capacity and the molar mass, each over optmat’s own temperature range, the real-gas sound speed of the equation of state, and optmat’s Wilke rule for a mixture (air resolved into its constituents). The mean free path, the flow law and the regime checks are solprop’s. A custom species without an OptmatGas equation of state has no transport properties, and a flow fill containing one is refused.

Solids#

The solid factories take model=: None (the default) or "recommended" takes optmat’s recommended model; an optmat model name or object takes that one. Each returns an OptmatGlass, usable wherever a glass is. Two adapters of one model compare equal, and hash alike, when they compute the same index — the same model digest under the same transparent approximation and bound — so two default walls or cores built separately are one material.

call	optmat model	range	notes
`fused_silica()`	`fit/SiO2`	24.8 nm–125 µm	absorbing, causal; 20 °C bulk grade
`calcium_fluoride()`	`daimon_2002/sellmeier_20C`	138 nm–2.33 µm	measured in nitrogen, made absolute
`magnesium_fluoride(axis=...)`	`dodge_1984/sellmeier_o`, `_e`	0.20–7.04 µm	axis required
`barium_fluoride()`	`malitson_1964/sellmeier`	0.27–10.3 µm	room air, made absolute
`sapphire(axis=...)`	`weber_2003/sellmeier_o`, `_e`	0.22–5.0 µm	axis required
`bk7()`	`schott_2023/sellmeier`	0.31–2.33 µm	lossless
`silver()`	`table/Ag`	4.1 nm–12.4 µm	absorbing table
`aluminium()`	`table/Al`	0.79 nm–200 µm	absorbing table

glass(name, axis=...) reaches the same models by name; so do the offline optics and the compressors, which name their glasses.

Named models. model= selects any other optmat model of the material, among them the published Sellmeier formulas, made absolute (see below):

factory	optmat model
`fused_silica`	`malitson_1965/sellmeier`; `franta_2016/udm_lithosil_q2`
`calcium_fluoride`	`daimon_2002/sellmeier_20C`; `malitson_1963/sellmeier`
`magnesium_fluoride`	`dodge_1984/sellmeier_o`
`bk7`	`schott_2023/sellmeier`
`sapphire`	`weber_2003/sellmeier_o` (Malitson and Dodge’s ordinary-ray formula, as Weber prints it)
`barium_fluoride`	`malitson_1964/sellmeier`

Absolute index. Every optmat index is relative to vacuum. The published Sellmeier formulas of CaF₂, MgF₂, BaF₂, sapphire and N-BK7 were measured relative to air (or nitrogen), and optmat multiplies by the reference gas’s index; taking such a formula as an absolute index would put it 3.7–5.6 × 10⁻⁴ low.
Sign. optmat’s index is \(n + i\kappa\) with \(\kappa \ge 0\) for loss, solprop’s convention, and is never conjugated.
Birefringence. MgF₂ and sapphire are uniaxial, and an optmat model is per ray: a window or wall names its ray, axis="o" or axis="e", and the recommended model without an axis is refused by name — magnesium_fluoride(), sapphire() and glass("MgF2") included, a deliberate breaking change. A named model is per ray too, and refuses the other axis.
Metals are tables with their interpolant’s exact derivative. Outside their range they hold their end values, with a warning.
Lossless routes. optmat’s silica carries a small \(\kappa\) even in its transparent window, and a route that solves only a lossless problem cannot hold it: the step-index modes (StepIndexMode, core and cladding), a general modal guide’s material laws (IndexMaterial, which the step-index and graded-index providers solve with) and the step-index vector fields. Such a route takes an optmat index through its transparent approximation automatically and never refuses it for absorption (lossless_index()): the real part of the index and of its derivative, a recorded approximation within a declared domain. Every lossless route of one adapter takes one transparent counterpart, so the modes of one guide still share their material. Dropping an absorption coefficient \(4\pi\kappa/\lambda\) above TRANSPARENT_ABSORPTION_BOUND (100 m⁻¹, which silica’s window from 200 nm to 2.5 µm passes and its 170 nm edge and 2.73 µm band fail) over the band the run evaluates warns once per run (MaterialRangeWarning, naming the material, the largest dropped absorption and its wavelength). The run’s log records what the counterpart dropped in that run, which the run’s manifest carries with the bound (transparency); the counterpart, shared by the runs of one source, also keeps what it has dropped over its life. material.as_transparent(absorption_bound=...) declares the approximation with another bound. A complex index of solprop’s own keeps each route’s existing handling: the step-index modes take its real part as documented, while the general providers and the vector fields refuse it. A hollow guide’s wall, and bulk propagation, take the complex index as it is.

Qualified intervals, statuses and warnings#

A model holds on its qualified intervals of vacuum wavelength. The atoms, H₂, D₂, N₂ and O₂ have one, which reaches the static limit; the infrared-active molecules have windows between their absorption bands (CH₄ five, N₂O four, SF₆ one up to 5 µm). A band outside them is still evaluated, and raises one MaterialRangeWarning per material and evaluation context (a fill’s spectral sampling, say). It names the parts of the band outside — below the shortest qualified wavelength, in a gap between two intervals with the gap’s reason (absorption, width, representation), or beyond the longest — and quotes optmat’s own warning. The same holds for the solids, whose ranges are one interval each.

Every model has a status: validated (a default with a stated uncertainty), provisional or untested (a stopgap, or nothing to test it against), comparison (a curated alternative), or user (built from a caller’s parameters). A provisional or untested model warns once when used, MaterialStatusWarning, with optmat’s reason; among the defaults that is CH₄’s viscosity in a gas-flow fill and SF₆’s ionisation rate, and N₂O’s Raman response where raman="provisional" admits it. comparison and user models were chosen by the caller and are recorded without a warning.

optmat’s errors become MaterialError, naming the material, the quantity, the state or band asked for, and the model.

Records#

Every optmat-backed object exposes material_records: a tuple of MaterialRecord, each the model’s name, quantity, material, digest, status, optmat version and any overrides. A Species, a Mixture (each model once), an OptmatSusceptibility, an OptmatGas, an OptmatKerr, an OptmatGlass, a SilicaBundle (the bundle, its Kerr fit and its shape) and a SampledShape all carry them, and a run records them in its material manifest. A mixture built by Mixture.air(selection=...) also records the name of the optmat composition its fractions came from (composition).

Quick experiments#

A modified material needs no change to optmat. optmat’s own overrides keep the model’s identity honest: the result has status user, its changes in modified and the curated model’s digest in base.

from optmat.gas.models.registry import default_model
from optmat.solid.models.formulas import FormulaModel
from optmat.solid.models.registry import default_model as solid_model

from solprop.physdata import MaterialSelection, OptmatGlass, calcium_fluoride, species

# a curated gas model with a narrower qualified range
narrow = default_model("Ar").with_parameters(intervals=((120e-9, float("inf")),))
argon = species("Ar", selection=MaterialSelection(linear=narrow))

# a curated solid with modified Sellmeier terms
curated = solid_model("CaF2")
terms = dict(curated.terms) | {"B": tuple(1.001 * b for b in curated.terms["B"])}
window = calcium_fluoride(model=curated.with_parameters(terms=terms))

# a caller's own Sellmeier formula, as an absolute index
mine = OptmatGlass.from_model(
    FormulaModel.sellmeier(
        "SiO2",
        {"B": (0.6961663, 0.4079426, 0.8974794), "C": (0.004679, 0.013512, 97.934)},
        (210e-9, 3.71e-6),
        citation="Malitson (1965)",
    )
)

A caller’s table of measured \(n\) and \(k\) is made the same way, with optmat’s TabulatedModel.from_arrays passed to OptmatGlass.from_model; that is the one constructor of a table from arrays, and solprop adds none to its TabulatedGlass. solprop’s own supplied types — SellmeierGlass, TabulatedGlass, a custom species with a TabulatedSusceptibility — keep working beside them, and carry their own citations.

Limitations#

The gas law is linear density scaling, with the error above declared and measured, not removed.
solprop requires optmat 0.3.0 or later: the noble gases’ recommended ionisation rates take the frequency-dependent Coulomb factor from 0.3.0, which run records and their digests assume; a gas’s Raman status is read without building its response, which 0.2.1 could not do; D₂’s viscosity is a smooth correlation from 0.2.2, so a D₂ temperature path over its whole range has a viscosity series; Gas.Air takes its molar mass from the equation of state, which 0.2.0 did not expose; and 0.1.0’s density inverse (a pressure from a density) failed for Ar, N₂ and CH₄ below about 255 K, 284 K and 256 K. A temperature at an equation’s lowest temperature, its triple point, is refused.
One grade and temperature per solid; N-BK7 is lossless; no OH, dopant or temperature absorption model.
Gas mixtures in a planar guide remain unsupported: a slab’s fill is one species.
By default the Kerr coefficient is one number per species and run, at the carrier (carrier-kerr); the third harmonic is driven by the same number. The dispersive policies are opt-in and gases only; they run on numpy_ref, numba_cpu and the native backend (backends); beyond a gas’s band edge they are best effort, with no resonant response (limitations).
A free-space run’s declared KerrResponse is the caller’s, coefficient included; gamma3(gas, wavelength=lambda0) gives optmat’s.
The design tools take the same materials= (design tools), so a design and a run under one selection agree.
Objects that describe a gas by name outside a driver — a mode’s own gas= fill, a SpatialMedium evaluated on its own, and the accessors gamma, gamma3, chi3, n2, density and refractive_index — take optmat’s recommended model, as MaterialSelection() does.

This page documents the functionality proposed and implemented by SEP 3; its dispersive electronic policies are SEP 6’s.