3.2. Surface microstructure The surface structures of the samples assessed in this study are exemplified inFig. 2. Note the differences with respect to the surface roughness. The inlet material(Fig. 2A) contains also a fraction of residual fibres that were not fibrillated duringthe homogenization process. The major part of the fibre fraction has been removedby a simple fractionation step (see also Tanaka et al., 2012). The nanopaper madewith the accept fraction is thus smoother (Fig. 2B), compared to the reject fraction,where most of the residual fibres have been collected (Fig. 2C).The differences between the 3 assessed samples are clearly confirmed by aquantification of the surface roughness at various scales (Fig. 3). As expected, thebottom side of a given nanopaper is smoother than the top side. This is due to theresidual fibres, which have a major effect on the top side. There are minor differ-ences between the three assessed fractions, when comparing the bottom sides ofthe nanopapers (Fig. 3A). Note that the bottom side roughness of the nanopapers issimilar to the roughness quantified on the petri dish surface, approximately 0.05 mat the assessed wavelengths. This confirms the conformation of the bottom side ofthe nanopapers to the surface of the petri dishes.The roughness of the top surfaces differs significantly (Fig. 3B). The majorincrease with respect to the roughness is for wavelengths larger than 80 m, whichis a clear indication of the occurrence of residual fibres. The accept fraction is rela-tively smooth over all the assessed wavelengths, indicating a successful removal ofmicrometre-sized structures. Hence, Fig. 3B indicates two important characteriza-tion aspects; (i) fractionation can be applied to reduce the fraction of residual fibresand (ii) LP has sufficient resolution to detect the micro-roughness of nanopapersurfaces.