Tip convolution artefact is one of the most important error sources in SPM. As the SPM tip is never ideal (like delta function) we often observe a certain degree of image distortion due to this effect. We can even see some SPM tips imaged on the surface scan while sharp features are present on the surface.
We can fortunately simulate and/or correct the tip efects using algorithms of dilation and/or erosion, respectively. These algorithms were published by Villarubia (see [1]).
For studying the tip influence on the data we need to know tip geometry firts. In general, the geometry of the SPM tip can be determined in these ways:
Within Gwyddion, we can use the first and the last approach of the mentioned ones. Using tip modelling ( → → ) most of the tips with simple geometries can be simulated. This way of tip geometry specification can be very efficient namely when we need to check only certainty map of perform tip convolution simulation.
To obtain more detailed (and more realistic) tip structure blind tip estimation algorithm can be used ( → → ).
Blind tip estimation algorithm is an extension of the well-known fact that on some surface data we can see images of certain parts of tip directly. The algortihm iterates over all the surface data and at each point tries to refine each tip point according to steepest slope in the direction between concrete tip point and tip apex.
We can use two modification of this algorithm within Gwyddion: partial tip estimation that uses only limited number of highest points on the image and full tip estimation taht uses full image (and is much slower therefore). Within Gwyddion tip estimation module we can use also partial tip estimation results as starting point for full estimation. This shlould improve the full tip estimation algorithm speed.
SPM tips obtained from data of previous figure using blind estimation algorithm.
When we know tip geometry, we can use tip convolution (dilation) algorithm to simulate data acquisition process. For doing this use Dilation module ( → → ). This can be in particular useful when working with data being result of some numerical modelling (see e.g. [2]).
Note this alrgorithms (as well as the following two) requires compatible scan and tip data, i.e. the physical dimensions of a scan pixel and of a tip image pixels have to be equal. This relation is automatically guaranteed for tips obtained by blind estimate when used on the same data (or data with an identical measure). If you obtained the tip image other means, you may need to resample it.
The opposite of the tip convolution is surface reconstruction (erosion) that can be used to correct partially the tip influence on image data. For doing this, use Surface Reconstruction function ( → → ). Of course, the data corresponding to point in image not touched by tip (e. g. pores) cannot be reconstructed as there is no information about these points.
As it can be seen, the most problematic parts of SPM image are data points, where tip did not touch the surface in a single point, mut in multiple points. There is a loss of information in these points. Certainty map algorithm can mark point where surface was probably touched in a single point.
Certainty map obtained from standard grating. Note that the modelled tip parameters were taken from datasheet here for illustration purposes. (left) – sample, (right) – sample with marked certainty map.
Certainty map algortihm can be therefore used to mark data in the SPM image that are corrupted by tip convolution in an irreversible way. For SPM data analysis on surfaces with large slopes it is important to check always presence of these points. Within Gwyddion you can use Ceratinty Map function for creating these maps (
→ → ).[1] J. S. Villarubia, J. Res. Natl. Inst. Stand. Technol. 102 (1997) 425.
[2] P. Klapetek, I. Ohlídal, Ultramicroscopy, 94 (19-29), 2003