An essential problem for these membranes is the high porosity and the different hardness of the 
individual layers, arising from the different sintering temperature. A lot of preparational steps are 
necessary to receive the true microstructure of a specimen size: 6.5 mm in diameter and 1mm in 
height. At first, the whole membranes were impregnated with epoxy resin under vacuum in order to 
stabilise the porous structure for the next preparation steps, cutting and mechanical processes of 
grinding and polishing in a special specimen holder. At last an ion beam etching process was 
performed to receive the true porosity up to the range of nanometers./6/ The controlling of the 
preparation results can be observed through an optical microscope, but it is only possible to show 
the layer structure. To determine the nanopores, the cross section of membranes are to be 
investigated under the high resolution electron microscope (FESEM). For the quantitative 
characterization of nanopores methods for optimizing the contrast enhancement were developed. 
The image analyses were performed by the system LEICA Quantimet 570. The chord lengths 
(lineal analysis) and the intersection circle diameters (area analysis) are determined in relation to 
the x-coordinates (direction of gradient). They represent the basic dates for estimating of the 
gradient of the pore and particle size distributions. This method is based on the calculation of > 
moments of chord length, intersection circle diameter and spherical diameter distributions./4/ The 
conditions consist of the acceptance of the spherical shape of the investigated objects and the 
normalised lognormal distribution (LOOG) of spherical diameters. 
By means of sedimentation of glass spheres in epoxy resin another variant of a particle reinforced 
gradient material was prepared the gradient of the microstructure parameters ( L, D, o, V,, N,) 
and the volume cumulative distribution could be determined for these specimens. The same 
procedure was also used for the characterization of the graded membrane layers. 
Results 
The results of layer investigations are summarised in the following figures. Figure 1 shows the 
cross section through the layer arrangement as a survey. The cross section specimen was ground 
and mechanically polished, adjacent to ion beam milling with angle of 10 degree and argon ions at 
1,5 h. From the bottom to the top we see the support, first and second intermediate layers, the 60 
nm and at last the graded TiO, layer, which are clearly differ from each other because of the 
decreasing pore size. The graded layer consists of (fig. 2) larger particles out of the coarser powder 
fraction with larger holes and the finest particles with hardly detectable interstices . It is supposed Fig 3: UF 
that the epoxy resin in the pores gives rise to this insufficient resolution. Compared to these shown 
images in fracture plane images (fig. 4) pore sizes down to the range of <15 nm can be estimated. 
For the evidence of artefact free preparation of nanoporous layers commercially membranes with 
narrow pore size distribution were used. It was possible to image the finest layer with 200 nm, but 
not resolve the individual particles. In the first experiment with the slope cutting method (45 
degree) a magnification of M = 250.000 : 1 for imaging of the 60 nm layers (fig. 3) could be 
realised. Some particles and sintering necks are shown. The pores between the sintered particles 
have a tubular shape. Very small amounts of closed pores arise from the gaps between the 
individual particles. It is possible to use the InLens detector for the SEM imaging and a short work 
distance. A thin coating with carbon or Au/Pd was indispensable. For the test of the proposed 
statistical methods for calculation the gradient of microstructure parameters glass spheres with a 
definite diameter distribution were sedimented. The results of the lineal analysis show a good 
agreement with the diameter distribution of the used glass spheres (fig. 5-7 and 8). The methods are 
also applied to the gradient layers are shown in figure 2. However, the number of evaluated pores 
were not sufficient enough to get a good statistics. It is necessary to increase the resolution and the 
number of measure fields . 
168