DE2700252A1 - Testing of structures on surfaces of electrical modules - has structures scanned and tested for observance of salient factors, e.g. spacing, angles, straightness - Google Patents

Abstract

The structure testing method is applicable to structures on surfaces of electrical modules or to auxiliary means for production of the subassemblies. The existing actual structure is scanned and intermediately stored. The structure so obtained is tested for observance of specified constructional rules relating to the structure building, such as syntactic connection between adjacent picture elements. Deviations from the rules indicating structure faults are detected. A practical example is afforded by a mask structure consisting of black regions, in front of a white background. The factors to be tested are accuracy of corner angles, straightness of edges and min. separation of the black regions which can be isolated from each other or interdependent.

Description

Procedure for checking defined structures

The invention relates to a method for testing defined structures on the surfaces of electrical assemblies or for the creation of these assemblies necessary tools. For example, for the production of integrated semiconductor circuits are mask sets for the photolithographic production of Ohc: surface structures necessary. These surface structures contain, for example, the geometric Course of openings in the oxide layer on the semiconductor surface, Contact points or connecting lines. As usually on a semiconductor wafer numerous identical components or integrated 'chaltunyen are generated, repeat the structures of a mask in adjacent sub-areas. The dimensions in semiconductor technology are extremely small, so that the structures in the mask sets in many cases may only deviate from the target structures by less than 1 / um. The mask sets must therefore, after their creation, on compliance with the tolerances and on others Manufacturing defects are checked. The masks in use must for early detection of damage continuously monitored so that cost-driving failure causes can be quickly eliminated.

For checking and checking mask sets or other structures Several methods are already known for small dimensions. With a well-known Method (Western Eleetric, The Engineer Vol XVII, No. 3, July 73, pp. 38-47) by subtracting the actual structures from the target structures in the correct phase Using spatial filters, errors in the actual structure are filtered out. These are aberration-free Lenses required, since every optical path difference does not make the display available Error means. Image field and resolution of this known method are thus through the optics are limited. The resolution limit is currently at a few microns, so this known methods cannot be used for very fine structures.

In another known method (IEEE Transactions on Electron Devices Vol ED-22, No. 7, July 75, pp. 487-495) two adjacent structures, which should be identical to each other, scanned by a split laser beam The generated signals are continuously compared with one another, with signal deviations a Display errors in one of the two structures being scanned. In this procedure with a two-beam comparator, two to be compared are always identical StruL. doors required. This results in considerable adjustment problems. Furthermore is also an optic is required here that limits the resolution and is also system-related May contain sources of error.

The object of the invention is to provide a test and control method specify, in which a comparison of the structure to be checked with its target structure or with another structure created according to exactly the same rules is no longer required. This eliminates all adjustment problems. Furthermore should It is no longer necessary to identify the test item, i.e. its target structure to pretend. This object is achieved according to the invention in that in a method of the type described above, the respective prevailing actual structure is scanned and it is stored that the structure thus determined is based on the predefined, for the Construction of the structure of applicable general construction rules (syntactic context the structural elements) is checked for compliance with these rules and that the Control deviations indicating structural errors can be determined.

With this method, it is no longer necessary to use the target structure of the test object to be determined, instead it is only checked whether the structure meets the general construction regulations that apply when realizing the mask structure were applied.

The design rules must be appropriate for mask testing Form to be kept ready. Design rules are, for example, the minimum dimensions the occurring structures or the distances between the structures, the orientation the structure edges and the type of angles occurring in the edge corners. Gives way the structure to be checked depends on these construction rules, must be a structural error are present. It is almost impossible that accidental errors result in a correct Mask another mask is created that still fulfills all construction rules.

The method according to the invention allows an automatic mask test, being the presence of a template (master) or a repeated structure is not assumed and the high resolution of a scanner with only a beam can be fully utilized.

The structures to be tested are preferably with a laser or scanned with an electron beam.

The detector can be arranged for transmission or reflection be. The scanned image is produced with the accuracy required Resolution broken down into individual pixels. The pixels are in the form of black and white information digitized.

This creates an extremely large amount of information that is only could be processed with difficulties in a data processing system. It is therefore it is necessary and sensible to keep this amount of information to the minimum required to reduce. This is done by means of local operations in a sequentially arranged, multi-level data processing chain, at the output of which a residual information is stc1lt, which reproduces the actual course of the structure sufficiently clearly and which can now be checked for compliance with the general design rules.

The data processing system required for this is for example a microprocessor.

With the help of a first operator, the scanned and digitized Black-and-white image denoise to eliminate the noise errors caused by the scanning system to eliminate. For this purpose, certain defective pixels are removed from the raster image converted after querying the neighboring pixels. With the help of a second local The operators are preferably the black and wet edges of the recorded structure determined, and the between two edges, made of black Bold information content existing in pixels is deleted.

Finally, with the help of a third local operator, the conspicuous Points of the overall structure are determined, which then in a data processing system are to be evaluated. Such noticeable points are preferably the corners of the black and hot edges, which in turn by querying the pixels adjacent to corner points for paints and type to be determined - The downstream data processing system has therefore only to check whether the course of the edges for each existing corner corresponding opposite corner is available. Thus, in the data processing system based on the specified and programmed design rules, the consistency or mismatch of vertices and pool types with these rules is found.

in the procedure according to the invention, only the permissible structures are used defined by design rules, so all not by the design rules covered structures are impermissible and are therefore shown as errors. The construction rules result from the respective construction principle for the structures, in particular the basic construction rules of the so-called Pay attention to "pattern generators" are. With masks for integrated As general design rules, semiconductor circuits are preferably the Minimum distances and the edge orientations in two mutually perpendicular Directions. However, it is also possible to use inclined edges or the course of roundings to be defined.

The structures to be checked are preferably axially parallel scanned to one of the occurring Kanterich lungs. The resolution in pixels is chosen so that a minimum distance of 3 to 8 pixels summarizes.

The invention and its further advantageous embodiment are intended in the following will be explained in more detail using a Ausführungsbei game.

FIG. 1 shows a small section from a mask structure shown. The structures consist of black areas against a white background. These areas can be isolated or contiguous from one another. To the general Design rules of the structure shown include the minimum clearances between the individual black areas and between the adjacent edges of a black Area. In the structure shown, the connecting lines run between each two vertices vertically or horizontally. Also this orientation of the edges belongs to the defined construction rules. It can also be defined, for example be that in the structures as it is shown in Figure 1, only right Angles occur in the corners. In the case of a mask, however, the permissible design rules are used expressly agreed. Any structure that is not in inserts these construction rules.

The mask image is made with a scanning electron beam, for example scanned. the test mask is checked in sections, for example. It snows appropriately. to arrange the test sections overlapping to avoid any to get annoying edge defects.

There are raster images of image sections of the to be checked Mask. Figure 2 shows such a section, which shows a faulty mask structure here contains. At the selected resolution, one picture element size corresponds approximately to one Quarter of the occurring minimum distance.

The representation of Figure 2 was made with a to the data processing system connected high-speed printer. The black pixels were in Shape of printed rectangles reproduced while the white Pixels are not shown. In Figure 2 it can be seen that the right below sciiräg Structure running to the top left forms a mask error caused by the data processing system should be determined automatically.

Contains raster images like the one in Figure 2 - especially for complex ones Masks - contain a lot of information that can only be obtained through computationally intensive processing operations can be evaluated. The procedure according to the invention separates the entire processing process in two steps, which are implemented with different means and jis.-total allow a cheap realization with relatively simple means. Included the first step is the preparation of the image information in order to obtain more characteristic ones Structural element: e and the second step the logical check (syntactic Control) the mutual relations between the ascertained and with theirs Coordinates assign structural elements.

It is important that the structural elements are determined here only Requires viewing of very small image sections. This is done using local Operations. in the associated processing window is sequentially over the raster image of the mask section to be checked is moved. This task specifies the use a special one Image processing hardware close to that according to the invention is implemented by a processing chain (pipeline) of local operators.

In contrast, the actual mask check requires syntactic Control of the mutual relationships between the structural elements is a processing plant of higher logical complexity. A suitable data processing unit is used for this purpose - e.g. I3. a microprocessor is used.

The required data processing chain (pipeline) is in the figure 3 shown. A first raster image is provided by the scanner, which is preferably is stored in a shift register. This raster image corresponds to, for example the picture shown in Figure 2. The raster image is created using a local Operator 1 is scanned or interrogated step by step in order to obtain the by the scanning method to eliminate conditional noise error. The local operator 1 always asks about this from pixels adjacent to a specific pixel and then decides whether the queried pixel retains its information or in the opposite in formation is converted. The noisy raster image output by the local operator 1, in your isolated black image elements against a white background or Individual holes in the black area have been eliminated and the black and white edges have been smoothed, is now checked with the help of a second local operator.

This local operator also queries neighboring ones step by step Image points and thus effects an edge determination. This means that with the local Operator 2 all black pixels that lie between the edges of a black Area are to be deleted.

With the help of a downstream third local operator 3 finally the corner points dtr occurring edges and the corner type are determined. The remaining one, however, to identify the existing structure sufficient residual information is transferred to a memory 4 and stored in a microprocessor 5 evaluated. This microprocessor 5 is given the general design rules, that were used when building the checked mask. By comparison of the data fed into the microprocessor using these design rules Deviations that correspond to structural filters are determined and output.

FIG. 4 shows how the raster image delivered by the scanner is initially denoise. In the six boxes shown in the figure is the respectively checked image point characterized by a small arrow. With the help of the local operator, each pixel is checked by checking the neighboring Pixels decided whether this pixel is converted. The top two Images show two white pixels, those of three black pixels each are surrounded and the white image points are each in a row of black image points lie. Then it is clear that the white pixel is due to a noise error and must be converted into a black pixel.

The two middle images are so-called spots, so black pixels in a white environment. These isolated black pixels are then converted into white pixels, as shown.

Fills of rounded corners are shown in the two lower boxes. The black individual pixels falling into the white image field and lying at the corner are recognized as such and converted into white pixels for corner straightening.

FIG. 5 shows how the edges are determined. For this every pixel is checked to see whether it has black pixels on all sides is surrounded.

If this is the case, this black pixel turns into a white one converted. After this operation, only the image edges remain black Lines left.

Reliable mask structures usually have black and white edges, which between each two corner points are uid either in a horizontal or straight line run in the vertical direction. Right angles are formed at the corner points, which are provided with defined minimum leg lengths. That after the edge determination The remaining image must therefore have right corners in four different corners Orientations can occur. These orientations form the corner type. The corners and the corner types are determined with the help of a third local operator, the one defining tone Neighborhood area scanned from te image points, determined. The corner pixel itself can be black or white. The four possible corner types are shown in FIG. 6 s incl shown. They are labeled A, B, C and D.

In FIG. 7, the raster image is shown after it has been de-noised and the edges are determined. The corners determined from this raster image and corner types are saved and evaluated in the microprocessor.

The microprocessor systematically checks whether the Corner points connected to one another by straight, vertical or horizontal edge lines can be. Here it can be defined, for example, that the edges have a Must have a minimum length and that within the edge course a jump by a row of pixels is permitted. Lines recognized as acceptable and Corner points are deleted, so finally only the impermissible edges and the edges emanating from impermissible corners are output by the data processing system will. Such a raster image, which then only contains structural errors, is shown in FIG. You can clearly see that from the bottom right to Structural errors at the top left are retained in full.

The data processing chain is preferably sent to l-iciltqeL-ite and / or connected to high-speed printer. So it is possible to display an error with or without reproducing its surroundings optically for visual classification or as Print out the protocol. The print protocol is particularly suitable for statistical Evaluation and determination of the defect density.

It has been shown that with the method according to the invention, masks or other structures can be checked fully automatically. All image operations From image preparation to corner searches, local operations can be used that only cover a small neighborhood area, for example respectively 5 x 5 or 3 x 3 pixels. These operations are expedient by special Coupling circuits are implemented which interconnect the memory for input and output images associate. It is crucial that only generally valid ones are used to check the masks Non-construction rules for these masks, which thus become the syntax of a structural language must be specified.

Claims (20)

P a t e n t a n s p r ü c h e 1) Procedure for checking more defined Structures on the surfaces of electrical components or those used to manufacture them Construction units required aids, characterized in that each existing lL <L structure is scanned and temporarily stored that the so Determined structure on the basis of a given, general one that applies to the structure of the structure Construction rules (syntactic connection of neighboring tile elements) compliance with these rules is checked, and that the structural errors are indicative Control deviations are determined. 2) Method according to claim 1, characterized in that the to structures under test can be scanned with a LASER or an electron beam and the scanned image at the resolution appropriate to the accuracy required is broken down into pixels. 3) Method according to claim 1 or 2, characterized in that the scanning takes place in transmission or in reflection (incident light). 4) Method according to one of claims 1 - 3, characterized in that that mask structures for integrated Semiconductor circuits, photoresist structures, Conductor structures or component structures on semiconductor bodies or the for Manufacture required aids are checked. 5) Method according to claim 2 or 3, characterized in that the Pixels are diqi talized in a black-and-white information that the Amount of information by means of local operations in a sequentially arranged, multi-level data processing chain reduced to the absolutely necessary minimum and the reduced residual information reflecting the actual course of the structure in a data processing system for compliance with the general construction rules is checked. 6) Method according to claim 5, characterized in that as a data processing system a microprocessor is used. 7) Method according to claim 5, characterized in that with the help a first operator de-noises the scanned and digitized black-and-white image is made by converting certain image points edge errors from the raster image can be filtered out by interrogating neighboring pixels. 8) Method according to one of claims 5-7, characterized in that that with the help of a second local operator the black and white edges of the recorded Structure and the image information content between two edges are determined is deleted. 9) Method according to Ansprucii 8, characterized in that for determining of the black-and-white edges are all between the black pixels of the edges black pixels are also deleted. 10) Method according to one of the preceding claims, characterized in that that with the help of a third local operator the conspicuous points of the overall structure determined and passed on to the data processing system. 11) Method according to claim 10, characterized in that with the help of the third local operator, the corners of the black-and-white edges by querying the the pixels adjacent to the corner points are determined that to the downstream Data processing system only the position of the corner points and the respective corner type is passed on and that in the data processing system based the predefined and programmed construction rules the conformity or Nonconformity of corner points and corner types with these rules found will. 12) Method according to one of the preceding claims, thereby marked that only the permissible construction rules are defined and thus a 1-le is not allowed for structures not covered by the construction rules @ n are. 13) Method according to claim 12, characterized in that the design rules the minimum distances occurring in the structures, the edge orientations and the angles occurring at corner points are defined. 14) Method according to claim 13, characterized in that the design rules to the respective design principle used to manufacture the structures Devices to be adapted. 15) Method according to claim 13 or 14, characterized in that as general design rules structured accordingly Masks for integrated circuits only show the minimum distances and the edge orientations can be defined in two mutually perpendicular directions. 16) Method according to one of the preceding claims, characterized in that that the structures are scanned axially parallel to one of the edge directions occurring will. 17) Method according to one of the preceding claims, characterized in that that the resolution in pixels is chosen so that a minimum distance at least Corresponds to 3 pixels. 18) Method according to claim 15, characterized in that a minimum distance 3 - 8 pixels corresponds. 19) Method according to one of the preceding claims, characterized in that that uses hardwired operators to perform local operations through each of which is a small neighborhood area from Interrogated pixels on the overall image and checked for certain criteria and that in this way the overall picture is gradually incorporated. 20) Method according to one of the preceding claims, characterized in that that the data processing chain is designed so that the image of an error without or with its respective environment for visual classification on a viewing device can be reproduced optically or to determine the defect density or as Control document can be printed out.