The roughness defined for a sample can be turned into real-space height maps, useful e.g. for direct visual comparison with atomic-force-microscopy (AFM) images or as a sanity check on the chosen autocorrelation, cross-correlation, and height-distribution parameters.
RoughnessMap(n_pts_x, n_pts_y, Lx, Ly, sample, i_layer[, seed])
generates one random rough surface, iteratively adjusted to reproduce the given spectrum and height statistics. Algorithm by Pérez-Ràfols & Almqvist, Tribology International 131, 591-604 (2019).
For multilayers, rough interfaces can also be generated together:
maps = ba.RoughnessMap.generateInterfaceMaps(
n_pts_x, n_pts_y, Lx, Ly, sample, interface_indices[, seed])
The result is a NumPy array with shape
(len(interface_indices), n_pts_y, n_pts_x). For example,
interface_indices=[1, 2] returns the maps for interfaces 1 and 2 in
that order.
Unlike two independent RoughnessMap(...).generate() calls, the
stack generator uses the roughness cross-correlation model when one is
defined. The lowest rough interface is generated by the regular
single-interface IAAFT algorithm. Interfaces above it are then built
bottom-up from the nearest generated rough interface below: each
Fourier component is split into the inherited part prescribed by the
adjacent cross-spectrum and an orthogonal independent residual with
the remaining auto-spectrum. This makes replicated long-wavelength
features visible while keeping the one-interface generator unchanged.
The requested interface indices only select maps from the result; the
whole rough stack is generated internally. Therefore
generateInterfaceMaps(..., [i], seed) is not generally equivalent to
an independent RoughnessMap(..., i, seed).generate() call for an
upper interface. The height distribution is adjusted by IAAFT only for
the lowest generated rough interface. Upper interfaces are constrained
by the target auto-spectrum and adjacent cross-spectrum; their height
statistics are therefore an approximation inherited from the spectral
construction.
