US3743423A - Intensity spatial filter having uniformly spaced filter elements - Google Patents

Abstract

A spatial filtering system for inspecting integrated circuit photomasks, and the like. The system includes a spatial filter comprising a matrix-like array of opaque regions on a transparent field. The region-to-region spacing on the filter is uniform and greater than that which results if it is assumed that the lens is perfect and generates an exact Fourier transform at its backfocal plane. As a result, high-frequency spatial information, corresponding to the edges of the photomask features, is suppressed improving the signal-to-noise ratio of the system.

Description

ilnited States Patent 11 1 Heinz et al.

1451 July 3, 1973 INTENSITY SPATIAL FILTER HAVING UNIFORMLY SPACED FILTER ELEMENTS [75] Inventors: Robert Alfred Heinz, Flemington Township, Hunterdon County; Laurence Shrapnel] Watkins, Hopewell Township, Mercer County, both of NJ.

[73] Assignee: Westinghouse Electric Company,

Incorporated, New York, NY.

22 Filed: May 3,1972

21 Appl. No.: 249,985

52 us. C1. 356/71, 350/162 SF, 356/239 51 int. Cl. 00111 21 32, 00215 27/38 [58] Field 6: Search also/ 235,35,

356/71, 168, 200,237, 239; 256/219 CR, 219 DF [56] References Cited UNITED STATES PATENTS 3,414,875 12/1968 Driver et a1. 350/162 SF 3,630,596 12/1971 Watkins 350/162 SF 3,658,420 4/1972 Axelrod 356/71 10/1971 Mathisen 356/71 OTHER PUBLICATIONS Watkins, Proc. of the IEEE, Vol. 57, No. 9, September 1969, Pages 1634-1639. Lohmann et al., Applied Optics, Vol. 7, No. 4, April Primary Examiner-David Schonberg Assistant Examiner-Ronald J. Stern Attorney-W. M. Kain, B. W. Sheffield 57 ABSTRACT A spatial filtering system for inspecting integrated circuit photomasks, and the like. The system includes a spatial filter comprising a matrix-like array of opaque regions on a transparent field. The region-to-region spacing on the filter is uniform and greater than that which results if it is assumed that the lens is perfect and generates an exact Fourier transform at its back-focal plane. As a result, high-frequency spatial information, corresponding to the edges of the photomask features, is suppressed improving the signal-to-noise ratio of the system.

6 Claims, 11 Drawing Figures mzmwm 3 ma 3.741423 sum 2 or 4 PAIENTED JUL 3 I975 SKHQNQ ORDER INTENSITY SPATIAL FILTER HAVING UNIFORMLY SPACED FILTER ELEMENTS BACKGROUND OF THE INVENTION 1. Field of the Invention Broadly speaking, this invention relates to spatial filtering. More particularly, in a preferred embodiment, this invention relates to an improved spatial filtering system which inhibits transmission of substantially all periodic information in the filtered image, thereby significantly improving the signal-to-noise ratio of the systern.

2. Discussion of the Prior Art As is well known, in the manufacture of integrated circuits, and the like, wafers of silicon, or other semiconductor material, are coated with a layer of photoresist and, then, exposed to light through a special photographic plate, known in the industry as a photomask. The exposed photoresist is then developed, in the conventional manner, and unexposed areas of the photoresist are removed, thereby, exposing underlying portions of the silicon wafer. These exposed portions are then subjected to processing steps, such as diffusion, etching, and the like.

A typical IC photomask may comprise a matrix-like array of thousands of nominally identical photomask features, each in itself a complex pattern of lines and other geometric shapes. Such photomasks have heretofore been made by successive photographic reductions from a large, hand-made master pattern, in a step-andrepeat camera, or more recently, by direct exposure of a photographic plate or chromium-coated plate in a computer-controlled electron-beam machine. More recently still, a primary pattern generator (PPG), a computer-controlled, electro-mechanical, laser deflection system, has been successfully employed to manufacture lC photomasks [See Bell System Technical Journal, (Nov. 1970), Vol. 49, No. 9, p. 2,03l-2,076].

However, regardless of the manufacturing process employed, lC photomasks are expensive and time consuming to make. Accordingly, every effort is made to prolong their useful life. Because of the extremely high resolution required with modern IC devices, exposure of photoresist-covered silicon wafers can only be satisfactorily accomplished by a contact-printing process, in which the emulsion side of the photomask is placed in direct physical contact with the wafer. This frequently results in damage to the mask during exposure. Furthermore, pinhole defects may occur during manufacture of the photomask itself, and dust or dirt may settle on the mask during use.

These defects are, of course, very serious, for any wafer exposed to light through a damaged or dirty photomask may yield dozens of defective, or wholly inoperative, IC devices. This situation is further aggravated by the fact that the same photomask is used over and over again. Thus, a given defect on a mask might be responsible for thousands of defective IC devices, a most undesirable situation.

As previously discussed, lC photomasks are too expensive to be discarded after they have been used only a few times. Accordingly, it becomes necessary to carefully inspect each mask after manufacture and also, somewhat less critically, during actual .production. Heretofore, this inspection was done manually by a skilled human operator, with the aid of a microscope. However, because of the complex nature of the geometric pattern in each photomask feature, as well as the fact that each mask contains many thousands of identical features, human error and fatigue have been found to result in the failure to detect significant numbers of defects.

To overcome this problem, a spatial-filtering technique was developed to inspect the photomasks. This technique forms the subject matter of copending U.S. Pat. application, Ser. No. 858,002, filed Sept. 15, 1969, (Watkins Case 1) which application is assigned to the assignee of the instant invention.

As disclosed in said copending application, the photomask to be inspected is illuminated by spatially coherent radiation from a laser and positioned proximate the front-focal plane of a convex lens. In accordance with well-known optical principles, an image will be formed at the rear-focal plane of the lens which corresponds to a Fourier transform of the photomask. That is to say, the image is a composite diffraction pattern whose spatial distribution is the optical product of two components: (1) the interference function of the photomask, comprising a distribution of bright dots of light whose spacing is inversely proportional to the spacing between adjacent features in the mask; and (2) the diffraction pattern of a single feature. Now, as, disclosed in said copending application, if a spatial filter comprising an array of opaque regions on a transparent field, is positioned proximate the back-focal plane of the lens, and if the spacing between the opaque regions corresponds exactly to the spacing between the dots of light in the diffraction pattern, substantially all of the light energy from the laser will be blocked.

However, if the mask is defective in some way, for example, if the mask is scratched, etc., the Fourier transform of the defect will not spatially correspond to the pattern of opaque regions on the filter, and accordingly, some light will succeed in passing through the filter, thereby enabling the scratch or other defect to be easily detected.

The above-described spatial filtering technique has been highly successful in practice. However, certain problems were encountered when an attempt was made to automate the inspection process. For example, in order to eliminate the human factor, a television camera, coupled to a counting device, was positioned'to view the filtered image of the mask. As the camera scanned over the image, the counting device recorded the humber of defects detected, and, if the value so found exceeded some predetermined value, the mask was discarded, or set aside for possible repair.

The system disclosed in copending application, Ser. No. 858,002, (Watkins Case I), assumed, for the sake of simplicity, that the interference function produced by a lens comprises equally spaced dots. In practice this is not exactly so, and the lens generates an interference function in which the dots become progressively further apart by very small increments. Furthermore, the lens may suffer from one or more optical aberrations, such as coma, astigmatism, field curvature, and distortion. The net effect is that, as the light energy impinges on those parts of the filter which lie further and further away from the center of the filter, the opaque regions thereon no longer fully block the light which is coming from the photomask, even in the total absence of defects on the mask. This is so for two reasons: first, the outermost regions are improperly positioned to fully intercept the light from the photomask, even if it were properly focused on the regions. Secondly, because the opaque regions are physically located on a planar sur face, the outermost opaque regions lie increasingly a small distance apart from the true focus of the lens, and hence, in effect, become progressively too small to fully block the light from the photomask. The outermost regions, of course, are intended to intercept the higher spatial frequencies from the photomask and, in practice, the only features on the mask possessing such higher spatial frequencies are the edges of the photomask features.

In prior art systems, where the filtered image was inspected by a human operator, this failure to fully suppress periodic, high frequency, edge information did not prove to be a significant problem. In fact, it was somewhat of an advantage, because the outline of the individual photomask features could be seen very faintly in the background of the image, as viewed by the operator. Thus, the approximate location of the nonperiodic defects which were successfully isolated by the system could be rapidly ascertained. However, in an automated process, this no longer holds true, because a television camera does not have a human operators ability to reason and is unable to discriminate between a true defect and the high-frequency edge information of the photomask features. Thus, in the automated process, the edge information was erroneously counted as a defect, which it is not. An additional problem with the prior art approach of copending application, Ser. No. 858,002, (Watkins Case 1 is that, because of the presence of high-frequency edge information, only a few of the thousands of features on a mask can be inspected at the same time. Now, if an attempt is made to increase the field of view, that is to say, if instead of inspecting only twenty or so of the thousands of features on a given mask it is desired to simultaneously examine several hundred features, the spatial filter must, accordingly, be made with considerably more accuracy.

As one solution to this problem, copending application, R. A. Heinz, et al., Ser. No. 249,983, of even date, discloses a spatial filtering technique employing a spatial filter wherein the spacing between adjacent opaque regions on the filter increases from region-to-region, from the centermost region outward, according to a precise mathematical formula. This spatial-filtering technique has proved highly successful in practice. However, the precision with which the opaque regions must be positioned on the surface of the filter dictates that such a filter be made by the use of a computercontrolled device, such as the primary pattern generator or by an electron-beam machine. This makes the filter expensive and time consuming to produce.

However, not all inspection tasks are so demanding that a mask of this high quality is required. Accordingly, it is an object of this invention to provide a method of spatially filtering an image which suppresses substantially all periodic information in the image, thereby significantly enhancing the signal-to-noise ratio of the image, yet which utilizes a relatively inexpensive spatial filter having a uniform element-to-element spacmg.

It is a further object of this invention to provide a spatial filter of novel construction for practicing the above method.

SUMMARY OF THE INVENTION To attain these, and other objects, a first embodiment of the invention comprises a method of isolating nonperiodic errors in a two-dimensional pattern containing a regular array of nominally identical elements, mutually spaced apart by a predetermined distance along at least one axis. The method comprises the steps of first directing a spatially coherent beam of light at the pattern to diffract the light, and then focusing the diffracted light onto a filter containing a plurality of discrete opaque regions on a transparent field, to spatially modulate the light. The spacing between adjacent regions, along at least one axis of the filter, is uniform and greater than the spacing dictated by the equation:

x nltf/d where,

x the distance from the centermost region (the origin) to the n" region;

A the wavelength of said beam of light;

n the order of the spatial harmonic;

d the step-and-repeat of said regular array of elements;

f the focal length of said focusing lens.

Finally, the spatially modulated light is reimaged to form an image exhibiting the non-periodic errors in the pattern, the filter blocking essentially all periodic information in the image, including the higher spatial frequency components.

For practicing the above method, another embodiment of the invention comprises a spatial filter for filtering the Fourier transform of the image of a workpiece comprising a matrix-like array of nominally identical features. The filter comprises a matrix-like array of opaque regions on a transparent field, the spacing between adjacent regions, along at least one axis of said array, being uniform and greater than the spacing dictated by the equation:

x nhf/d where,

x the distance from the centermost region (the origin) to the n" region;

A the wavelength of the light forming said image;

n the order of the spatial harmonic;

d the step-and-repeat of the workpiece;

f the focal length of the image-forming lens.

By the use of this filter, the above-defined opaque regions inhibit further transmission of substantially all periodic information in the Fourier transform.

The invention and its mode of operation will be more fully understood from the following detailed description, and the following drawing, in which:

DESCRIPTION OF THE DRAWING FIG. I is a partially schematic, isometric view of a I first embodiment of the invention;

FIG. 2 illustrates a typical workpiece of the type which may be inspected by the instant invention;

FIG. 3 shows an enlarged view of a portion of the workpiece shown in FIG. 2;

FIG. 4 depicts the format of the diffraction pattern produced when the workpiece of FIG. 2 is inspected by the apparatus of FIG. 1;

FIG. 5 depicts an illustrative prior art spatial filter;

FIG. 6 is a diagram illustrating the theory underlying the instant invention;

FIG. 7 is a graph showing the spacingof filtering elements on the filter of FIG. 5, as a function of the spatial harmonic;

FIG. 8 depicts the relative orientation of the filtering elements of a prior art filter and the filter according to this invention;

FIG. 9 is a graph showing the spacing of the opaque regions of the filter according to this invention;

FIG. 10 is a diagram illustrating the underlying theory of operation of the filter according to this invention; and

FIG. 11 is a diagram illustrating the operation of the above filter in greater detail.

DETAILED DESCRIPTION OF THE INVENTION FIG. 1 depicts an illustrative embodiment of the invention. As shown, the apparatus comprises a laser 10 which, when connected to a suitable source of energy (not shown), emits a beam of spatially coherent, radiant energy along a longitudinal axis 11. The light from laser 10 is directed through a beam expander 12 comprising a first lens 13 and pinhole 14. The expanded beam is then passed through a collimating lens 16 and finally falls upon the IC photomask 17 to be inspected.

FIGS. 2 and 3 illustrate photomask 17 in greater detail. As shown, the photomask comprises a glass photographic plate 18 having recorded thereon a matrix-like array of nominally identical features 19. As shown in FIG. 3, each feature comprises a complex pattern of opaque areas 21 on a transparent field, the pattern in each feature defining the areas of the photoresistcovered semiconductor wafer which are to beprotected from exposure to the light. It will be noted that all of the edges of the areas 21 in feature 19 are parallel to either the horizontal or to the vertical axes of the mask. By analogy to the orientation of the blocks in a typical city, such a configuration is frequently referred to as Manhattan geometry, although, of course, the invention is not limited to inspecting workpieces having such Manhattan geometry and can inspect, with equal success, other workpiece configurations. It will also be noted that, in FIG. 2, a uniform spacing D is assumed to exist between the center lines of each feature on the mask. It is further assumed that this spacing is the same in both the horizontal and vertical directions, (i.e., D D). Occasionally, a photomask is produced in which the feature-to-feature spacing differs in the horizontal and vertical directions. However, this is easily compensated for in the design of the spatial filter, and the underlying theory of the instant invention applies to both arrangements.

In the drawing, mask 17 is depicted as having a 5 X 5 matrix of features thereon. One skilled in the art will appreciate that this is merely for convenience in illustrating the invention and that an actual photomask may have as many as 40,000 features thereon arranged in a 200 X 200 matrix.

Returning now to FIG. 1', photomask 17 is positioned at the front-focal plane of a second lens 22 which, as previously discussed, will form a Fourier transform of the photomask at the back-focal plane thereof. In accordance with the teachings of copending application, Ser.No. 858,002, (WatkinsCase l), a spatial filter 23 is positioned at the back-focal plane to intercept all periodic information from photomask l7 and to permit all is connected by a lead 26 to a control circuit 27, which includes conventional power supplies, amplifiers, deflection apparatus, etc. A digital-readout device 28 is connected to control circuit 217 by a lead 29 to record the number of defects in photomask 17 which succeed in passing through spatial filter 23 and are detected by camera 25.

FIG. 4 illustrates the pattern which would be seen if a screen were to be positioned at the back-focal plane of lens 22, rather than spatial filter 23. For convenience in drawing, this pattern is shown as a series of black dots on a white field. It will be appreciated that, in actual practice, each of the black dots in FIG. 4 represents a spot of bright light. As seen, the pattern approximates a cross with the spacing between adjacent light dots, in the horizontal direction, being inversely proportional to the feature-to-feature spacing in the horizontal direction in mask 17. Similarly, the spacing between adjacent dots, in the vertical direction is inversely proportional to the feature-to-feature spacing in photomask 17 in the vertical direction. If, as discussed, the feature-to-feature spacing on the mask is uniform, and equal, in both directions, then under the assumptions made in the referenced copending application (Watkins Case I), the dot-to-dot spacing in the diffraction pattern will also be uniform, and equal, in both directions. The large central dot 31 corresponds to the do term of the Fourier transform and, moving to the right, in the horizontal direction, dot 32 corresponds to the first harmonic," or fundamental spatial frequency, (i.e., the step-and-repeat pattern of the mask), dot 33 the second harmonic, and so on.

FIG. 5 depicts a spatial filter of the type disclosed in the above-referenced copending application, (Watkins Case I). This filter comprises an array of opaque regions on a transparent field. This type of filter can be manufactured by the use of any of several known techniques in essentially the same manner that the photomask itself may be manufactured. Considerable success has been obtained by the use of the above-referenced primary pattern generator, and by the use of a stepand-repeat camera. If, as is usually the case, the feature-to-feature spacing on the mask is uniform, and equal, along both the horizontal and vertical axes, then the array of opaque regions in the spatial filter will also be uniform, and equal, along both axes, and will coincide with the location of light spots 31 through 34, etc., in FIG. 4.

While the intensity of light and the size of the spots in the actual diffraction'pattern of FIG. 4 may vary, the

, opaque regions in FIG. 5 are all uniform in size and non-periodic information, such as defects in the photodensity. Of course, the regions must be large enough to block the largest of the light spots shown in FIG. 4.

As previously discussed, the system described in copending application, Ser. No. 858,002 (Watkins Case I) assumed that the lens was perfect and produced equally spaced dots, and this assumption was reasonable for the inspection scheme contemplated by that invention. However, for more critical applications, this assumption is not valid, and the deviation must be taken into consideration. In FIG. 6, a lens 41 is shown positioned so that a diffraction grating 42 is at its frontfocal plane. The diffraction grating has elements spaced apart by a uniform distance d. Typical light rays 43 are shown coming from the diffraction grating at an angle 0 to the horizontal axis, as shown, and are imaged by lens 41 onto the back-focal plane of lens 41.

From basic diffraction theory, it is known that when a plane, collimated beam of light is incident upon an intensity grating, the resulting pattern behind the grating is the superposition of many plane waves, each propagating in a different direction. The angle at which these beams emanate from the grating is a function of the harmonic which they represent, that is:

sin 0 nit/d where,

)t the wavelength of light;

n the order of the harmonic;

d the step-and-repeat of the array;

f the focal length of the Fourier transform lens. Each of these waves is then focused to a spot in the back-focal plane by the Fourier transforming lens 41. The hemispherical surface 44 has been included to aid in computing the location of these images in the plane 45. The location of the light spots on the plane 45 can then be computed from simple geometry:

x =ftan 0 =ftan [sin ("M/d] Since for small angles, i.e., low spatial frequencies, sin 0 5 tan 0 E 0, the above equation reduces to the form which was assumed in copending application, Ser. No. 858,002, (Watkins Case 1), namely,

x nAf/d FIG. 7 is a graph showing the distance from the origin (center) of the opaque filter regions, as a function of the order of the spatial frequency, for the linear equation assumed in the copending aplication, and for the actual equation given in Equation 2 above. It will be observed that for the first few orders, the deviation between the linear graph and the actual, approximately tangential graph is very small, but towards the higher orders, this discrepancy becomes increasingly larger.

The upper half of FIG. 8 depicts the uniform regionto-region spacing employed in prior art spatial filters, corresponding to the linear graph 47 in FIG. 7. Copending application R. A. Heinz, et al., Ser. No. 249,983, of even date, discloses the use of a spatial filter in which the region-to-region spacing is not uniform but, rather, increases according to the tangential-like curve 48 in FIG. 7. Thus, as shown in the lower half of FIG. 8, while the first few opaque regions in the filter disclosed in the copending application, R. A. Heinz, et al., Case I- 1 -4-1 of even date, are at approximately the same position as they would be for the linear case, if one moves outward, to the right, from the center of the filter, the discrepancy between the position of the regions in the linear filter and those in the non-linear filter of the copending application, R. A. Heinz, et al., Case l-l-4-l of even date, becomes increasingly large. Again, it must be emphasized that for clarity, the scale has been greatly exaggerated.

Because the step-and-repeat spacing of typical integrated circuit devices varies from to 120 mils, the typical spacing between the opaque regions on a spatial filter varies from 20 to l20 microns, assuming an HeNe laser and a 100 mm focal length lens. It is, therefore,

essential that the filter be manufactured with the greatest care, and considerable accuracy is required to successively increase the distance between the regions, in accordance with Equation 2. Accordingly, if the spatial filter disclosed in copending application, R. A. Heinz, et al., Case 1-l-4-l of even date, constructed in accordance with Equation 2 and graph 48 of FIG. 7, is substituted for the spatial filter 23 in FIG. 1, the filter will effectively block all repetitive information from the photomask 17, including the edge information, even though the regions are actually positioned on planar surface 45, rather than the actual back plane of lens 22.

From a practical standpoint, the requirements for manufacturing the filter described in copending application, R. A. Heinz, et al., Case 1-1-4-1, of even date, are so demanding, particularly the progressive, but minute, increase in region-toregion spacing, that production can only be satisfactorily accomplished in a computer-controlled device such as the Primary Pattern Generator or in a computer-controlled electronbeam machine. This, of course, makes the filter relatively expensive and time consuming to produce. Fortunately, for less critical applications, where a certain degree of extraneous high-frequency edge information can be tolerated, the filter of the instant invention can be substituted for the filter used in said copending application, R. A. Heinz, et al., Case l-l-4-l, of even date.

As shown in FIG. 9, we have discovered if a line 49 is drawn having a slope slightly steeper than that of line 47, which latter line corresponds to the linear Equation 3 above, the line will intersect the tangential-like curve at the origin and at the point 51. Thus, according to the invention, if a spatial filter is constructed having a uniform region-to-region spacing defined by line 49, rather than line 47 or 48, the opaque filter regions lying to the left of point 51 will be located further from the origin than the theoretical location would dictate, the regions which are clustered about point 51 will be positioned extremely close to the theoretically defined positions, and the regions to the right of 51 will be closer together and closer to the origin than the theoretical location would dictate.

Now, if the opaque regions on the filter are somewhat larger than the spots of light in the diffraction pattern of the workpiece to be examined, the filter regions to the left of point 51 will, nevertheless, intercept the periodic information in the diffraction pattern.

Assume that the point 52 on graph 47 corresponds to the position of that opaque region on filter 23, which no longer satisfactorily intercepts diffracted light from the mask 17. The vertical distance from' this point to the corresponding point 53' on graph 48 represents the deviation x between the actual position of this opaque region and the position that this same order region would have occupied had the regionto-region spacing on the filter been steadily increased, according to graph 48 and Equation 2. In a similar manner, point 54 on graph 49 represents the location of the last opaque region to satisfactorily block diffracted light from pattern 17 in a filter constructed according to this invention. Again x, represents the deviation between the true position of this region and the position that it would have occupied had the region-to-region spacing been steadily increased according to Equation 2. It will be observed that for the same deviation, i.e., when x x,, the location 54 of the last satisfactory opaque region on a filter according to this invention, lies much further to the right than the location 53 on a filter constructed according to the prior art teachings. In other words, even though the region-to-region spacing on the instant filter is uniform, and deviates from the optimum or theoretical spacing, the results obtained by the use of this filter will, nevertheless, be significantly superior to those obtaincd by the use of the filters disclosed in the abovediscussed copending application (Watkins Case 1). The reason for this is that a larger percentage of the opaque filter regions are positioned to effectively block the periodic information, and hence, suppress highfrequency edge information. This is illustrated in more detail in FIG. in which a plurality of opaque filter regions 36 are drawn in alignment with the graph of FIG. 9. For clarity, opaque regions 36 have been assumed to be square. The black regions within each such filter regions are intended to represent the bright spots of light generated by the Fourier transform of photomask 17. As can be seen from the figure, the center opaque filter region is positioned to intercept the dot of light corresponding to the dc. term squarely in the center thereof. This is no surprise because the deviation between graph 48 and graph 49 is zero for the dc term. However, as the order increases, in a direction to the right in FIG. 10, the spots of light are intercepted by the corresponding filter regions more and more to the left of center until at the fourth filter region shown, the lightspot almost fails to be intercepted. Continuing on, however, for higher and higher orders, the light spot begins to move to the right until, at the filter region corresponding to point 51 on the graph (i.e., at the intersection of graphs 48 and 49), the light dot is once more squarely centered in the filter region. Continuing on to still higher orders, the point of interception moves to the right until, as shown, in the last filter region, the light dot completely fails to be intercepted and from then on, the filter will fail to suppress high-frequency edge information, until thelight dot again gets in synchronism with the opaque filter regions.

The apparent movement of the light dot, with respect to the filter regions is shown in FIG. 11. As will be selfevident, the dot appears to move from the center of the opaque region to the extreme left and, then, reverses direction and moves to the right passing once more through the exact center of the opaque filter region. It will be appreciated that FIGS. 9 and 10 are not to scale and are merely illustrative of the operation of the invention. The motion of the light spots shown in FIGS. 9 and 10 actually occurs over several hundred opaque filter regions, rather than the dozen or so shown.

Theoretically, by the. use of the filter according to this invention, nearly twice as many light spots can be successfully intercepted then can be intercepted by the linear prior art filter disclosed in copending application, Ser. No. 858,002, (Watkins Case 1), that is to say,

in a typical filter, approximately 200 or more light reglOnS.

One skilled in the art will appreciate that while the invention has been described with reference tothe inspection of integrated circuit photomasks, it may also be used to inspect any workpiece having optical characteristics approximating those of an optical grating, either transmissive or reflected, e.g., a processed silicon semiconductor slice. For example, the invention has successfully been used to inspect fine metallic grids, and diode array targets, such as those used in the manufacture of Picturephone camera tubes, and the like. Further, if desired, the spatial filter might comprise a matrix of transparentregions on an opaque field rather than a matrix of opaque regions on a transparent field. In this latter event, periodic information would be transmitted, rather than blocked. Of course, the term regions, as used herein, is intended to comprise various shapes, such as circles, squares, triangles, etc. The

actual shape employed is merely a matter of convenience, provided that the corresponding light dot in the diffraction pattern is blocked (or passed). Also, various changes and substitutions may be made to the elements shown, without departing from the spirit and scope of the invention.

Finally, it must again be stressed, that while Manhattan geometry is by far the most common found in integrated circuits, the methods and apparatus of this invention may be used to inspect workpieces having any geometry in their features.

What is claimed is:

l. A method of isolating non-periodic errors in a twodimensional pattern containing a regular array of nominally identical regions, mutually spaced apart by a predetermined distance along at least one axis, which comprises the steps ofi directing a spatially coherent beam of light at the pattern to diffract the light;

focusing the diffracted light on a filter consisting of a plurality of discrete substantially equally sized opaque regions on a transparent field, the spacing between adjacentregions, along at least one axis of the filter being uniform and greater than the spacing dictated by the equation:

where,

x the distance of the n region from the origin, A the wavelength of said beam of light, n the order of the spatial harmonic, d the step-and-repeat distance of said regular array of regions,

f the focal length of said focusing lens,

wherein at least one of said regions is positioned in coincidence. with a location dictated by the equa tion:

where,

x the distance of the n" region from the centermost region (origin); )t the wavelength of said beam of light; n the order of the spatial harmonic. and is greater than one; d. the step-and-repeat distance of said regular array of regions, 5 f the focal length of said focusing lens; to spatially modulate the light; and

reimaging the spatially modulated light to form an image exhibiting the non-periodic errors in the pattern, said filter blocking essentially all periodic information in said image, including higher spatial frequency components.

2. The method according to claim 1 wherein said point of coincidence lies approximately halfway between the centermost filter regions and the edge of said filter.

3. Apparatus for inspecting non-periodic errors in a two-dimensional pattern containing a plurality of nominally identical and regular spaced regions arranged in a planar periodic array, which comprises:

means for directing a spatially coherent beam of light at the plane of the pattern so that the light is diffracted thereby;

a first lens positioned to focus the light diffracted by the pattern;

a planar optical filter consisting of a distribution of discrete substantially equally sized opaque regions on a transparent field, the spacing between adjacent regions, along at least one axis of the filter being uniform and greater than the spacing dictated by the formula:

where,

x the distance of the n" region from the centermost region (origin),

A the wavelength of said beam of light,

n the order of the spatial harmonic,

d the step-and-repeat distance of said regular array of regions,

f the focal length of said focusing lens,

wherein at least one of said regions is positioned in coincidence with a location dictated by the equation:

1: =ftan [sin (nM/d] where,

x the distance of the n" region from the centermost region (origin); A the wavelength of said beam of light; n the order of the spatial harmonic and is greater than one; d the step-and-repeat distance of said regular array of regions; f the focal length of said focusing lens; the filter being positioned at the focal plane of the first lens for spatially modulating the intensity of the light focused thereon by the first lens;

a second lens positioned to reimage the light transmitted by the filter to form a visual image of the non-periodic errors in the pattern; and

means for projecting the visual image onto the image display means.

4. Apparatus according to claim 3 wherein said image display means and said projecting means comprises:

a television camera focused on said visual image;

control means, connected to said camera, for supplying deflection signals and power to said camera, said camera scanning across said visual image to detect said non-periodic errors; and

counting means, connected to the video output of said camera, for counting the number of nonperiodic errors so detected.

5. A spatial filter for filtering the Fourier transform of the image of a workpiece comprising a matrix-like array of nominally identical features, which consists of:

a matrix-like array of discrete substantially equally sized opaque regions on a transparent field, the

spacing between adjacent regions, along at least one axis of said array, being uniform and greater than the spacing dictated by the equation:

where,

x the distance of the n"' region from the centermost region (origin);

A the wavelength of the light forming said image; n the order of the spatial harmonic; d the step-and-repeat distance of the workpiece; f the focal length of the Fourier transform lens;

wherein at least one of said regions is positioned in coicidence with a location dictated by the equation;

where,

x the distance of the 11" region from the centermost region (origin);

A the wavelength of the light forming said image;

n the order of the spatial harmonic and is greater than one;

d the step-and-repeat distance of the workpiece;

f the focal length of the Fourier transform lens; whereby said opaque regions inhibit further transmission of substantially all periodic information in said transform.

6. The filter according to claim 5 wherein said point of coincidence lies approximately halfway between the centermost filter region and the edge of said filter.

. i i i i i

Claims (6)

1. A method of isolating non-periodic errors in a twodimensional pattern containing a regular array of nominally identical regions, mutually spaced apart by a predetermined distance along at least one axis, which comprises the steps of: directing a spatially coherent beam of light at the pattern to diffract the light; focusing the diffracted light on a filter consisting of a plurality of discrete substantially equally sized opaque regions on a transparent field, the spacing between adjacent regions, along at least one axis of the filter being uniform and greater than the spacing dictated by the equation: x n lambda f/d where, x the distance of the nth region from the origin, lambda the wavelength of said beam of light, n the order of the spatial harmonic, d the step-and-repeat distance of said regular array of regions, f the focal length of said focusing lens, wherein at least one of said regions is positioned in coincidence with a location dictated by the equation: x f tan (sin 1 (n lambda )/d) where, x the distance of the nth region from the centermost region (origin); lambda the wavelength of said beam of light; n the order of the spatial harmonic and is greater than one; d the step-and-repeat distance of said regular array of regions; f the focal length of said focusing lens; to spatially modulate the light; and reimaging the spatially modulated light to form an image exhibiting the non-periodic errors in the pattern, said filter blocking essentially all periodic information in said image, including higher spatial frequency components.

2. The method according to claim 1 wherein said point of coincidence lies approximately halfway between the centerMost filter regions and the edge of said filter.

3. Apparatus for inspecting non-periodic errors in a two-dimensional pattern containing a plurality of nominally identical and regular spaced regions arranged in a planar periodic array, which comprises: means for directing a spatially coherent beam of light at the plane of the pattern so that the light is diffracted thereby; a first lens positioned to focus the light diffracted by the pattern; a planar optical filter consisting of a distribution of discrete substantially equally sized opaque regions on a transparent field, the spacing between adjacent regions, along at least one axis of the filter being uniform and greater than the spacing dictated by the formula: x n lambda f/d where, x the distance of the nth region from the centermost region (origin), lambda the wavelength of said beam of light, n the order of the spatial harmonic, d the step-and-repeat distance of said regular array of regions, f the focal length of said focusing lens, wherein at least one of said regions is positioned in coincidence with a location dictated by the equation: x f tan (sin 1 (n lambda )/d) where, x the distance of the nth region from the centermost region (origin); lambda the wavelength of said beam of light; n the order of the spatial harmonic and is greater than one; d the step-and-repeat distance of said regular array of regions; f the focal length of said focusing lens; the filter being positioned at the focal plane of the first lens for spatially modulating the intensity of the light focused thereon by the first lens; a second lens positioned to reimage the light transmitted by the filter to form a visual image of the non-periodic errors in the pattern; and means for projecting the visual image onto the image display means.

4. Apparatus according to claim 3 wherein said image display means and said projecting means comprises: a television camera focused on said visual image; control means, connected to said camera, for supplying deflection signals and power to said camera, said camera scanning across said visual image to detect said non-periodic errors; and counting means, connected to the video output of said camera, for counting the number of non-periodic errors so detected.

5. A spatial filter for filtering the Fourier transform of the image of a workpiece comprising a matrix-like array of nominally identical features, which consists of: a matrix-like array of discrete substantially equally sized opaque regions on a transparent field, the spacing between adjacent regions, along at least one axis of said array, being uniform and greater than the spacing dictated by the equation: x n lambda f/d where, x the distance of the nth region from the centermost region (origin); lambda the wavelength of the light forming said image; n the order of the spatial harmonic; d the step-and-repeat distance of the workpiece; f the focal length of the Fourier transform lens; wherein at least one of said regions is positioned in coicidence with a location dictated by the equation; x f tan (sin 1 (n lambda )/d) where, x the distance of the nth region from the centermost region (origin); lambda the wavelength of the light forming said image; n the order of the spatial harmonic and is greater than one; d the step-and-repeat distance of the workpiece; f the focal length of the Fourier transform lens; whereby said opaque regions inhibit further transmission of substantially all periodic information In said transform.

6. The filter according to claim 5 wherein said point of coincidence lies approximately halfway between the centermost filter region and the edge of said filter.

Images