Advanced options

For most cases you only need to adjust the Gaussian range parameters nmax, r1, rnmax, (and additionally Nmax, R1, RNmax for three-body systems) in the numerical parameters. If you struggle to reach convergence or have other issues, changing the advanced parameters can help:

• complex_scaling_angle: The complex rotation angle in degrees. Due to the Gaussian basis functions, this is limited to a maximum of 45°. If you deal with broad, or just-above-threshold resonances, you might need a large complex rotation angle to uncover the resonance. For very broad resonances, it might however be practically impossible to resolve it numerically. If you need more precision for narrow resonances, reduce the complex rotation angle.
• complex_range_freq: Frequency of the complex-ranged Gaussian basis functions. A larger value corresponds to faster oscillations. Experience showed that typically a value between 0.5 and 2.0 is advisable, default is at 0.9.
• kmax_theta: Number of angular nodes used for the interpolation of the interaction matrix elements when complex-ranged basis functions are used. The effective Gaussian range is then complex and confined to the sector $|\arg\alpha| \le \arctan\omega$, over which the interpolation is two-dimensional. The setup cost grows in proportion to this number, so kmax_interpol can usually be reduced in compensation. Increase it if you need higher accuracy, in particular for complex_ranged=:both, where the four-fold enlarged basis makes the overlap matrix nearly singular.
• threshold: Cut-off below which eigenvalues of the norm-overlap are discarded when solving the generalized eigenvalue problem. If you experience unphysical, very large negative eigenvalues, try increasing the value of threshold. However, the overall resolution of your calculation will suffer accordingly. The underlying reason is that the non-orthogonal basis becomes almost degenerate and the generalized eigenvalue problem becomes ill-defined. A better solution is in most cases to ensure a larger difference between r1 and rnmax (and/or between R1 and RNmax).
• kmax_threshold: Number of effective Gaussian ranges used for interpolation of the interaction matrix elements. If your interaction potential features many details, try increasing this number.
• lmin, lmax: Make sure you allow for appropriate angular momenta of your subsystems for the desired total angular momentum and parity. Including large values (4 and above) can quickly blow-up your calcuational cost in both time and memory (especially for three-body systems).
• mu0, c_shoulder: Parameters for treating general central potentials in the ISGL method. Default values are 0.08 and 1.6, provided by the original authors of the method. Change only if you are desperate.
• large number of allocations/memory usage: This is due to numerical integration of the matrix elements via QuadGK. If your interaction potential allows for an analytic treatment, consider implementing it yourself and/or file an issue. Currently, only Gaussian interactions are treated analytically, but more will be added in the future.