Comments on filtering of defects in integrated circuits.

In 1969 Watkins proposed inspecting IC-photomasks by co­ herent optical filtering. More recently Will and Pennington described an arrangement for spatial frequency filtering without the necessity for any filter alignment, i.e., a system with a wafer (mask) that requires no rotational alignment. The principle is in situ development of the filter plate. It is also said that this system makes it possible to use one subelement of the IC array to generate a filter for the entire array. The arrangement of Will and Pennington is constructed for the detection of defects on the wafer by processing the front surface reflected beam in­ stead of error detection in IC-photomask transparencies. We would like to comment concerning only that part of their pub­ lication that deals with coherent optical techniques. (a) During our research on the detection of flaws in ICphotomasks we tried the technique of in situ development of the exposed filter plate, as mentioned in Refs. 3-5 in order to avoid the realignment of the filter. In our opinion, however, a system that needs a separate filter for each photomask or wafer—and only in this special case is the filter self-aligning—will have only poor chances for its realization in practice. This is also true in the case of using nonphotographic filter materials as, for instance, thermoplastics. The time required to generate separate fil­ ters is too long for rapid automatic inspection. If a single filter is used for a large number of identical photomasks, these have to be aligned relative to the rotational position of the filter in the Fourier plane. Furthermore, we believe that there are more difficult problems in the field of flaw detection as, for example, the use of lenses with a high aperture to detect defects with a size of 0.5-2 μm or the distinction between phase defects and amplitude defects [see also part (c)]. (b) In Ref. 5 we also mentioned that it is possible to con­ struct a filter of a subelement for filtering the entire array. How­ ever, such a filter has only very poor qualities with respect to flaw enhancement. Only the periodic structure of the entire photomask, which is used for filter generation in Refs. 1 and 5, forms stopbands with a density that has the desired shape of a δ-function with high contrast in the spatial frequency domain. (c) During the course of our work on error detection we ex­ perienced some trouble with phase defects caused by the glass substrate and the gelatin of the photomasks. It is a matter of conjecture whether these defects will not exist on wafers that are inspected using the surface reflected beam. Small changes in the thickness of the layers forming the surface relief structure of different wafers (which belong to the same filter) may enhance the signals in the detection plane, although these changes do not necessarily act as electronic defects in the IC array.