The influence of diffraction on the shape and size of features printed using x-ray proximity printing with a collimated x-ray source (measured beam divergence of approximately 0.2 mrad full width at half-maximum) at mask to wafer gaps of 25 mum and above is described. Three major conclusions can be drawn from the results: (1) Diffraction can distort the shape of a printed feature, and the systematic shape changes observed in resist images can be explained using simple scaling based on Fresnel diffraction; (2) The linewidth change with exposure dose is independent of feature type and size, and depends only on the square root of the mask to wafer gap; and (3) The bias of each printed feature varies when all of the feature types and sizes are printed at the same dose. Resist images of 2.0-0.25 mum contact holes printed at mask to wafer gaps ranging from 25 to 515 mum are presented. The square contact holes on the mask print diamond shaped at a Fresnel number of 2.5. Isolated lines, spaces, and line-space arrays ranging from 1.0 to 0.25 mum were printed in thin PMMA (0. 1 mum thick) resist at gaps ranging from 25 to 80 mum over a series of exposure doses. A 10% increase in exposure dose results in a 16, 20, and 30 nm change in linewidth at gaps of 25, 40, 80 mum, respectively. When the 1.0-0.25 mum lines, spaces, and line-space arrays are printed with a single dose at a 40 mum gap, the bias variation among the features appears to be approximately +/- 20 nm. However, this bias measurement is very sensitive to the metrology technique used, and systematic errors in these absolute measurements may be a significant contributor to the observed bias variation. All the experimental results are compared to calculated aerial images. Exposures in thick PMMA (0.5 mum) show trends similar to those observed in thin resist. However, as the thickness of the resist increases, the dose at which the printed linewidth has zero bias decreases, and approaches the dose required to clear a large open area.