The data in this Figure is split into 7 datasets, which are the highest fields in the overall figure dataset
The dataset names (corresponding to the given panel) are

a: 'central_internal_cooperativity_collated_data'
b: 'non_axially_centred_collated_data'
c: 'transverse_displaced_small_collated_data'
d: 'transverse_displaced_large_collated_data'
e: 'two_axial_emitters_collated_data'
f: 'emitters_on_line_example_collated_data'
g: 'emitters_on_line_example_middle_collated_data'

within each dataset and panel, the data fields and plots were as follows:
Data fields:
nn_basis: The basis sizes in which the optimisation was performed (2D array). Units: dimensionless
nn_opt: The number of parameters (basis state coefficients) optimised. For panel b), this is nn_opt complex parameters, but for all other cases there are nn_opt real parameters. Units: dimensionless
convergence_matrix: The expected value from the target mode for the metric (C_{int}, C_{int}^{min}, or C_{int}/L_{y,eff}^{-1} depending upon dataset). Units: either dimensionless, or metres for C_{int}/L_{y,eff}^{-1}
fundamental_metric_expectation: The expected value of the metric (C_{int}, C_{int}^{min}, or C_{int}/L_{y,eff}^{-1} depending upon dataset) for the case of spherical mirrors. Units: either dimensionless, or metres for C_{int}/L_{y,eff}^{-1}

Figure plots:
Heat map: convergence_matrix (heat) vs nn_basis (x array) and nn_opt (y array). No data when nn_opt > nn_basis
Black line on colorbar: fundamental_metric_expectation
Cyan line on heat map and colorbar: The '90% converged' value of fundamental_metric_expectation + 0.9 * (max(convergence_matrix) - fundamental_metric_expectation)
