DE2616716A1 - Image fault detection system - uses optical filter in plane of spatial frequency diffraction spectrum blocking correct spectrum - Google Patents

Abstract

A method of detecting faults in image patterns, e.g. in integrated circuit masking, uses optical filtering. It is designed to isolate statistical errors and to simplify fault detection by reducing production complexity. The spatial frequency diffraction spectrum of the image is stopped with a flat laminar amplitude filter (4) in preferred directions defined by the preferred directions of the desired image structure. Faults in the image cause light to be transmitted. Signals due to residual light which cannot normally be filtered are separated from the error signals using a threshold detector. Light detection is by television or photoelectrical methods.

Description

Optical filter method for the detection of defects in image

patterns The invention relates to a method for detection of errors in image patterns, e.g. IC masks, by means of optical filtering.

Examples of such image patterns are primarily IC masks, which play an important role in the manufacture of integrated circuits. failure in these masks lead to carnival-like end products. It is therefore necessary to continuously check the masks and to sort out defective masks.

A purely visual inspection is very time consuming and tedious, as every the numerous individual images of a mask can be examined with the aid of a microscope Got to. In recent years, therefore, several procedures have been introduced to facilitate the Mask testing has become known, e.g. Watkins, L.S .: Proc. IEEE 57 (1969) pp. 1634-1639.

This method works with an amplitude filter in the spatial frequency level within a two lens imaging optical system. With the The individual diffraction orders of the periodic mask structure are amplitude filters faded out by means of opaque covers and the diffraction light of the Errors essentially let through. The defects appear as light spots on dark The underground made clearly visible. This makes it essential to detect errors improved.

However, the production of the amplitude filter for IC masks can still cannot be regarded as satisfactorily resolved. The best results are achieved at best with synthetically manufactured filters. For this purpose the diffraction spectra calculated or measured and with this data then the filters with the help of a Photo repeater manufactured.

In the case of integrated circuits, the current development leads to larger ones Individual samples (area around 10 mm2 and larger) and even smaller struc- tures which already have errors from 2 to 3 / in order to be faulty or unusable Products.

The distance between the diffraction orders lies in these large individual patterns at the location of the amplitude filter iin1um range (approx. 501um), and the production of Amplitude filters according to the type described prepare disproportionately large measuring and manufacturing difficulties. That poses a big problem for that Filters required glass carriers. The defect detection of defects from 2 to 3 / um Size requires high-resolution and highly corrected special lenses, their imaging properties can be significantly worsened by a glass carrier in the beam path. Without carrier however, a corresponding filter can hardly be implemented.

The object of the invention is to provide a method with which the statistical errors can be separated from the structure of the image patterns and the complicated manufacturing process for the filters, which has hitherto been a practical one Use of this procedure stood in the way of avoiding so that the error detection is made much easier.

This task is solved in that the diffraction spectrum of the image pattern in the spatial frequency plane with the help of a flat and carrierless, outside of the image pattern held amplitude filter hidden in preferred directions is, d.ie given by the preferred directions of the target structure of the image pattern and the light resulting from the defects is transmitted to a substantial extent will.

The extensive dimming of the spatial frequency spectrum (instead of the point-like masking of all diffraction orders) can therefore be achieved with the IC masks perform successfully in terms of error detection because the target structures have distinct preferred directions in the masks. So is most of the structures limited by border lines offset by 900, which is why a spectrum with only two pronounced ones Preferred directions arise. In addition to the straight edge lines, there are usually only a few additional structures with e.g. inclined edge lines below 300 and 450, their diffraction spectra can also be faded out over a large area. The directions of the edge lines of mask defects, on the other hand, are of a statistical nature, as a result of which their spatial frequency components fall only for small sections of the edge lines with the blocking areas of the amplitude filters together.

In order to ensure sufficient fault detection, an extensive amplitude filter in itself has an extended pass band in have as many directions as possible. So there is a compromise on that two opposing demands for a large blocking effect for the target structure and high permeability to find statistical errors.

In practical use it turns out that e.g. the corners of the mask structure, that have slightly rounded edge lines cannot be completely suppressed. These corner point signals but are still significantly below those of actual error signals. With a threshold detector these two signals however, a distinction can be made, so that an automatic error detection method can be implemented with little effort can be developed. Through the extensive filtering of the spatial frequency spectrum there is also no need for the extremely necessary in the known Watkins process precise angle adjustment of the mask, so that the diffraction spectrum and the individual Covers of the amplitude filter are brought to cover.

The invention is to be explained in more detail using drawings.

They show: FIG. 1 the basic structure of an optical spatial frequency filtering, 2 shows a detail from a mask structure with straight and inclined edge line delimitation, 3 shows a section from a diffraction spectrum of structures with straight and inclined edge line delimitation (schematic representation), FIG. 4 shows an exemplary embodiment a planar amplitude filter and FIG. 5 a schematic, one-dimensional Section through a filtered mask image with errors and one incomplete blocked target structure (corner points).

With reference to FIG. 1, the mode of operation of the optical filter is intended procedure are described in more detail. The image template 1, which can also be used, for example, step-by-step or continuously can be moved is illuminated with a monochromatic light wave 2. These Light wave is focused by the lens 3 in its focal plane 4. That on the original picture Diffracted light generates the spatial frequency spectrum 5 in the focal plane of the lens 6 in its rear focal plane 7 all rays become image points again merged. The actual filtering process takes place in level 4, where the optical Informat.ion about the structure of the mask contained in the diffraction spectrum is. For error detection, the diffraction spectrum (diffraction light) 5 of the target structure the mask with the help of opaque covers (amplitude filters) if possible be largely blocked, while the diffraction light from errors the filter as possible can happen undisturbed. The defects are then shown as light areas in the image plane 7 clearly visible on a darkened surface and can then be detected by a photo detector 7 'or by a television camera.

FIG. 2 shows a section 8 from a mask structure with straight lines and 450 sloping margins. This results in a diffraction spectrum 9, as shown schematically in a greatly simplified manner in FIG. 3. The spectrum can can be blocked with the flat amplitude filter 10 shown in dashed lines.

Fig. 4 shows a possible amplitude filter for this Beumungsa spectrum. The filter consists of simple black paper strips 11 attached to a frame 12 can be worn. A disruptive carrier material is no longer necessary.

Fig. 5 shows the detectability of an error from the not completely suppressible target structure. 13 and 14 represent linear sections from a mask structure, 15 represents an error. Along the dashed line Line 16 is to take a closer look at the intensity profile of the filtered image. With 17 the remaining background brightness is displayed, 18 and 19 are the Corner points of the target structure 13 and 14, which are caused by the two-dimensional amplitude filter (Fig. 4) cannot be completely suppressed. The intensity profile 20 of the error 15 is well above that of the corner points. Experimentally, a 2 to 3 fold Value measured. With the help of a threshold value 21, the structure of the target caused signals are suppressed, so that only error 15 (signal 20) is detected.

Patent claims: L e r s e i t e

Claims (6)

Patent claim: ¼) Method for the detection of defects in image patterns (e.g. IG masks) by means of optical filtering, characterized in that the diffraction spectrum the image pattern in the spatial frequency plane with the help of a flat and carrierless, The amplitude filter held outside the image pattern is masked out in preferred directions which are given by the preferred directions of the target structure of the image pattern, and a substantial part of the light resulting from the defects is transmitted will. 2. The method according to claim 1, characterized in that not filterable Residual signals of the target structure of the mask by means of a threshold value detector from the error signals to be detected are separated. 3. The method according to claim 1 and 2, characterized in that the Error detection takes place with the help of ferusehtechnischer means. 4. The method according to claim 1 or one of the following, characterized in that that the error detection takes place with the help of photoelectric detectors. 5. The method according to claim 1 or one of the following, characterized in, that the original image moves step by step or continuously to detect the error will. 6. The method according to claim 5, characterized in that at the same time several original images can be detected in parallel.