Built-in Reference Models

The starkzee.models package provides five independent lineshape models that share a common call signature and serve as benchmarks against StarkZee’s fully coupled solver. They are grouped into tabulated models (reading precomputed databases) and analytical models.

Tabulated Models

Stehlé (MMM) — stehle

The Stehlé model reads precomputed Stark lineshapes from the Model Microfield Method (MMM) database [stehle] for hydrogen Balmer and Paschen transitions over a grid of \(N_e\) and \(T_e\) values. The tables contain the unmagnetized (\(B = 0\)) Stark + fine-structure profile as a function of a reduced detuning \(\Delta\omega/F_0\) (Holtsmark normal field units). The pipeline is:

  1. Bilinear interpolation over the \((N_e, T_e)\) table grid (with a Fortran-style three-point hyperbolic interpolator for the detuning axis).

  2. Thermal Doppler broadening via FFT convolution.

  3. Normal Zeeman splitting applied after the table: the whole Stark+Doppler profile is rigidly copied to \(\omega \pm \omega_L\) (Larmor frequency) with angle-dependent \(\pi/\sigma\) weights:

    \[I(\theta) = \sin^2\!\theta\,I_\pi + \tfrac{1+\cos^2\!\theta}{2}\,(I_{\sigma+} + I_{\sigma-})\]

The Stark and Zeeman effects are therefore treated as separable. This is valid only when the Zeeman splitting \(\mu_B B\) is much smaller than the Stark width. Stehlé covers arbitrary \((n_u, n_l)\) and species with no restriction on \(B\)-field value (it is simply ignored in the table lookup).

Rosato — rosato

The Rosato database [rosato] solves the Stark-Zeeman problem jointly and tabulates the result with \(B\) as an explicit table axis (\(B \in \{0, 1, 2, 2.5, 3, 5\}\) T). The pipeline is:

  1. Interpolation over the \((N_e, T_e, B)\) grid. The \(B\)-interpolation is scaled: before blending two bracketing profiles at \(B_0\) and \(B_1\), each profile’s detuning axis is stretched by \(B_\mathrm{node}/B\). This exploits the fact that Zeeman splitting scales linearly with \(B\), aligning the \(\sigma\) components before interpolating.

  2. Angle dependence from two real tables (parallel and perpendicular to \(\vec{B}\)), blended as \(I = I_\parallel\cos^2\theta + I_\perp\sin^2\theta\). Both tables carry the correct \(\pi/\sigma\) lineshape and width — the angle enters as a physical interpolation, not just an amplitude weight.

  3. Thermal Doppler broadening via FFT convolution.

Rosato is restricted to deuterium Balmer lines and the table coverage \(N_e \in [10^{13}, 10^{16}]\) cm\(^{-3}\), \(T_e \in [0.316, 31.6]\) eV, \(B \le 5\) T.

Analytical Models

Lomanowski — lomanowski

Polynomial fits for the Stark FWHM from Lomanowski et al. [lomanowski]:

\[\Delta\lambda_S = c_{n_u n_l}\,N_e^{a_{n_u n_l}}\,T_e^{-b_{n_u n_l}} \quad [\mathrm{nm}]\]

Coefficients \((a, b, c)\) are tabulated for H/D Balmer and Paschen lines up to \(n_u = 9\). A Voigt profile combines this Stark Lorentzian with a Doppler Gaussian. No magnetic field treatment.

Parameterized Stehlé — stehle_param

An analytical parametric fit to the Stehlé tables, providing a fast closed-form approximation to the field-free Stark profile without reading the NetCDF database.

Voigt — voigt

A pseudo-Voigt profile using the Griem \(\alpha_{12}\) Stark half-width [griem] combined with thermal Doppler broadening. Serves as a quick estimate when only order-of-magnitude Stark width accuracy is required.

Comparison with StarkZee

Comparison with StarkZee

Feature

StarkZee

Stehlé

Rosato

Magnetic field

Full simultaneous diagonalization of \(H = H_A + V_E\)

\(B = 0\) in tables; added as rigid triplet after

\(B\) is a table axis

Fine structure

Yes — spin-orbit, MV, Darwin

No (degenerate hydrogenic)

In tables

Electron broadening

GBK semi-classical, frequency-dependent

Unified theory (more rigorous far wings)

In tables

Ion dynamics

Quasi-static + optional FFM

Static + some ion-dynamics treatment

In tables

Microfield

Hooper screened

Holtsmark/Hooper variant

Holtsmark variant

\(B\)-treatment

Intrinsic to Hamiltonian

External convolution

\(B\)-scaled interpolation

The most important physical distinction at \(B \ne 0\): StarkZee diagonalizes \(H = H_A + V_E\) simultaneously for all field strengths. When \(\mu_B B \sim 3n\,e\,a_0\,F\) (Zeeman and Stark splitting comparable) the eigenstates are genuine Stark-Zeeman hybrids — neither pure Zeeman nor pure Stark states. No post-processing convolution can reproduce this mixing. The separable approximation (Stehlé) and scaled-interpolation approach (Rosato) both become inaccurate in this regime.