viscosity. Here we notice that the flow velocity through the filter is proportional to the pressure if we ignore the relatively unimportant changes of viscosity and temperature with altitude. This means that if one wants to sample at a constant rate, one would have to change the velocity of the sampler as the vehicle descends through the atmosphere in order to achieve a constant rate, In general, as one descends through the atmosphere one would expect first to have the free molecular equations apply, and then as one got to lower altitudes where the mean free path was less, one would expect to move into the region where the Poiseuille flow equations are appropriate. Neither expression describes the flow through the capillary very well for a considerable range about the region when the air molecule mean free path is comparable to the capillary wall. gradual transition between the two. The experimental evidence is scanty, but it indicates that there is a For the rest of the discussion, I shall assume that we have chosen a filter which has a sufficiently small pore that one can describe the flow through the filter as always being in the free molecular region, that is, that there is a constant-flow velocity through the system, While I shall on one oc- casion talk about a case where this assumption would not be justified, I shall then make the assumption that we have somehow managed to alter the sampler velocity to achieve a constant-flow velocity through the system. Figure 5 gives the diffusion and impaction criterion equations. We have derived the diffusion equation by assuming, somewhat inconsistently, that the capillary is no longer circular but square. a substantial simplification of the mathematics. This assumption allows One then sets up the boundary condition that the distribution of contaminant is uniform across a cross section of the tube as the flow enters it, and that the concentration of contaminant then drops to zero at the edge of the flow, since material there is condensed on the wall. After a couple of pages of juggling, one comes out with this expression which describes the collection efficiency. My symbols have the same meaning as they did in the flow equation, except that C now stands for debris concentration, while D is the diffusion coefficient. For the impaction criterion, I have assumed that my capillaries were not truly straight, but have sufficient curvature that a particle moving straight would strike the wall after traveling four radii. This is quite arbitrary, but should describe in an approximate fashion the impaction effectiveness on the filter. Impaction is deemed effective for values of K greater than one. Figure 6 summarizes the result of combined diffusion and impaction collection for a reasonable flow situation, To obtain these curves, I have considered pore sizes 10 microns and 100 microns in diameter. I have also assumed that the ratio of the length of the capillary to the radius of the capillary is 10, which gives us that value for K. For a sampler velocity we use 1.34 x 107 em/sec, which is an effective average for the system Mr. Matejka described, For the value of the product (AF) we chose 7.5. correspond to the filter being 75 percent open with a fold factor of ten. This would, for example, Turning to the flow equation we find for the 10-micron filter that the flow may be calculated using the free molecular description, and the average normal flow v is 19,5 percent of the sampler velocity. The 100-micron pore filter flow would display a transi- tional behavior, but we assume that the sampler velocity during the descent is altered to achieve the same value of Vv. Applying the diffusion and impaction criterion, we then obtain these performance curves for the two pore diameters, Diffusion collection is effective for particles smaller than the limit given by that curve, while impaction is effective for particles having diameters longer than the limit curve for that phenomena. We see that for a 10-micron pore diameter, the impaction and diffusion collection processes overlap throughout the entire range from 100 to 250 kilofeet. On the other hand, for a 100-micronfilter, there is an overlap from 150 kilofeet upward, but there is a gap between the two from this altitude downward. This suggests that we are limited to fairly small pore sizes in our filters if we are to rely on these collection mechanisms, 142