Method and system for classifying an integrated circut for optical proximity correction

[0001] This application is related to co-pending patent application Ser. No. ______, entitled Method and System for Hierarchy Driven Chip Covering for OPC Purposes, filed on the same day and assigned to the same Assignee as the present application.

[0002] The present invention relates to integrated circuit chip design, and more particularly to performing optical proximity correction on an integrated circuit design.

[0003] The minimum feature sizes of integrated circuits (ICs) have been shrinking for years. Commensurate with this size reduction, various process limitations have made IC fabrication more difficult. One area of fabrication technology in which such limitations have appeared is photolithography.

[0004] Photolithography involves selectively exposing regions of a resist coated silicon wafer to a radiation pattern, and then developing the exposed resist in order to selectively protect regions of wafer layers.

[0005] An integral component of photolithographic apparatus is a reticle which includes a pattern corresponding to features at one layer in an IC design. The reticle typically includes a transparent glass plate covered with a patterned light blocking material such as chromium. The reticle is placed between a radiation source producing radiation of a pre-selected wavelength and a focusing lens which may form part of a stepper apparatus. Placed beneath the stepper is a resist covered silicon wafer. When the radiation from the radiation source is directed onto the reticle, light passes through the glass (regions not having chromium patterns) and projects onto the resist covered silicon wafer. In this manner, an image of the reticle is transferred to the resist.

[0006] As light passes through the reticle, it is refracted and scattered by the chromium edges. This causes the projected image to exhibit some rounding and other optical distortion. While such effects pose relatively little difficulty in layouts with large feature sizes (e.g., layouts with critical dimensions above about 1 micron), they cannot be ignored in layouts having features smaller than about 1 micron. The problems become especially pronounced in IC designs having feature sizes near the wavelength of light used in the photolithographic process.

[0007] To remedy this problem, a reticle correction technique known as optical proximity correction (OPC) has been developed. Optical proximity correction involves adding dark regions to and/or subtracting dark regions from a reticle design at locations chosen to overcome the distorting effects of diffraction and scattering. Typically, OPC is performed on a digital representation of a desired IC pattern. First, the digital pattern is evaluated with software to identify regions where optical distortion will result. Then the optical proximity correction is applied to compensate for the distortion. The resulting pattern is ultimately transferred to the reticle glass.

[0008] OPC, as now practiced, involves modifying a digital representation of a reticle design. The modification is performed by a computer such as workstation having appropriate software for performing OPC. Points separated by less than the critical dimension on the design are evaluated in sequence and corrected as necessary. Evaluation of each point requires analysis of surrounding features in two-dimensions to determine whether problematic diffraction effects are likely. If so, an appropriate correction (serif or segment removal, for example) is performed.

[0009] A problem with using OPC when performing a full mask design correction is that a substantial commitment must be made in terms of time and computing power in order to optically correct the integrated circuit design. For example, a moderately complex integrated circuit design may require at least a few days to correct with OPC even when the OPC algorithm runs on the fastest modern workstations.

[0010] The computational expense of OPC can be understood by recognizing that the correction often involves adding multiple small serifs to corners of design features and removing from, adding to, or displacing lateral sections of lines. The modifications are made only after evaluating an initial pattern with very fine granuiaritytypically evaluating potential correction points separated by no more than about 0.02 micrometers (using a 0.25 micron critical dimension technology). Note that a typical reticle design may include about 50-100 million rectangles of average size 0.5 by 0.5 micrometers. The features in and surrounding each rectangle must be evaluated in two dimensions. For example, a decision as to whether a point under consideration should be corrected may be made only after a five-by-five grid of surrounding points is first evaluated. This process is very time consuming and requires significant computational resources because enormous amounts of data must be evaluated in order to perform OPC on a single IC chip.

[0011] Accordingly what is needed is a system and method for efficiently and accurately performing OPC on an IC chip. The present invention addresses such a need.

[0012] The present invention provides a method and system for performing OPC on an IC chip by classifying the IC chip into groups of congruent tasks. The system and method of the present invention includes dividing the IC chip into a plurality of local task regions, identifying congruent local task regions, classifying congruent local task regions into corresponding groups, and performing OPC for each group of congruent local task regions.

[0013] By identifying and grouping congruent local task regions in the IC chip, according to the method and system disclosed herein, only one OPC procedure (e.g., evaluation and correction) needs to be performed per group of congruent local task regions. The amount of data to be evaluated and the number of corrections performed is greatly reduced because OPC is not performed on repetitive portions of the IC chip design, thereby resulting in significant savings in computing resources and time.

[0014] FIG. 1 is a block diagram illustrating an environment in which the present invention may operate.

[0015] FIG. 2 is a flowchart illustrating a process for performing OPC classification in accordance with a preferred embodiment of the present invention.

[0016] FIG. 3 is a diagram illustrating two exemplary feature types.

[0017] FIG. 4 is a block diagram illustrating an exemplary local task region and its associated region of optical influence.

[0018] FIG. 5 is a block diagram illustrating an exemplary extended task region with intersecting feature types.

[0019] FIG. 6 is a flowchart illustrating a process for generating the characteristic vectors according to a preferred embodiment of the present invention.

[0020] FIG. 7 is a flowchart illustrating a process for normalizing the characteristic vectors according to a preferred embodiment of the present invention.

[0021] FIG. 8 is a block diagram illustrating the extended task region of FIG. 5 after it has been manipulated.

[0022] FIG. 9 is a flowchart illustrating a process for classifying the characteristic vectors according to a preferred embodiment of the present invention.

[0023] The present invention relates to a method and system for performing OPC on an IC chip design. The following description is presented to enable one of ordinary skill in the art to make and use the invention and is provided in the context of a patent application and its requirements. Various modifications to the preferred embodiments and the generic principles and features described herein will be readily apparent to those skilled in the art. Thus, the present invention is not intended to be limited to the embodiment shown but is to be accorded the widest scope consistent with the principles and features described herein.

[0024] As discussed above, OPC is a technique that corrects a mask design for potential light diffraction effects. In conventional systems, a mask design is fully created and is then corrected using an OPC technique. Some examples of OPC techniques are described in Otto, et al., Automate Optical Proximity Correction-A Rules-based Approach, SPIE Optical/Laser Microlithography VII, March 1994, and Helmsen et al. 3D Lithography Cases for Exploring Technology Solutions and Benchmarking Simulators, SPIE: Optical/Laser Microlithography VI, pp. 382-394, March 1994, that are incorporated herein in their entirety. As stated above, performing OPC on the entire integrated circuit design requires substantial time and computational resources because large amounts of data corresponding to millions of minute features in the mask design must be evaluated and corrected.

[0025] The method and system of the present invention is directed to reducing the amount of data evaluated for OPC by consolidating identical groups of data. An IC chip design typically is made up of a number of elements or building blocks, which are repeated throughout the chip. These building blocks are hierarchical and referred to as cells, macrocells, or cores, depending on the hierarchical position of the element. For instance, a cell might represent a simple flipflop or SRAM circuit, while a macrocell might represent a more complex element (e.g., a large adder), which incorporates several cells. The lowest level building blocks are single polygons, also called features.

[0026] While an IC chip typically includes millions of features, it generally includes a relatively small number of macrocells. Because macrocells are repeatedly utilized throughout the IC chip design, congruent (i.e., same or similar) feature arrangements are inherent.

[0027] According to the present invention, the IC chip design is divided into local task regions. Congruent, i.e., identical or similar, local task regions are then identified and grouped. By grouping congruent local task regions, the data from only one representative local task region in a group needs to be evaluated and corrected. This greatly reduces the amount of data evaluated for OPC, which in turn, saves time and computing resources. The resulting correction is applicable to all the local task regions in that group.

[0028] The present invention is implemented as a software program, which in a preferred embodiment, operates on an engineering workstation or other type of computer system 102, as shown in FIG. 1. Referring now to FIG. 1, which is a block diagram illustrating an environment in which the present invention may operate, workstation 102 includes a microprocessor 104 that is coupled to a non-volatile storage device 106, e.g., a conventional hard disk storage device. The hard disk storage 106 includes a cell library 128. The cell library 128 is a physical database representing a collection of cells and can include any combination of design elements, e.g., feature types, macrocells, cores, etc. The microprocessor 104 and the hard disk storage device 106 are coupled to a volatile storage device 110, e.g. a random access memory (RAM). A monitor 112, is coupled to the workstation 102 for displaying a user interface.

[0029] The RAM 110 stores, during operation, an operating system (O/S) 114, e.g., UNIX, a conventional input/output (I/O) module 116, an IC chip design 118, an optical proximity correction (OPC) module 120, and an OPC classification module 122. The IC chip design 118 is submitted by a designer (not shown) via the input/output (I/O) module 116. Preferably, the IC chip design 118 is described in some format, e.g., GDS II, which allows the OPC classification module 122 to interpret the design. The OPC classification module 122 is coupled to the OPC module 120, and in accordance with the present invention, classifies and submits data from IC chip design 118 to the OPC module 120 for correction.

[0030] The function of the OPC classification module 122 is discussed in conjunction with the flowchart in FIG. 2, which illustrates a process 200 for performing IC chip classification in accordance with a preferred embodiment of the present invention. Referring to FIGS. 1 and 2, the process begins in step 202 when the OPC classification module 122 identifies all possible feature types stored in the cell library 128 and used in the IC chip design 118. For each feature type, an index is assigned and an origin point is chosen, as shown in FIG. 3.

[0031] FIG. 3 is a diagram illustrating two exemplary features (300, 302) that have been indexed 1 and 2 respectively. The origin point is chosen to be the lower left corner of the feature bounding box, represented by an x in FIG. 3.

[0032] Referring again to FIGS. 1 and 2, once all possible feature types have been indexed, the classification module 122 divides the IC chip design 118 into a matrix of local task regions, in step 204. In a preferred embodiment, the matrix is a square matrix, each square having a size, S, whereby S is a function of the complexity of the correction procedure. Typically, the size of the local task region is in the range of 5-10 , where is the wavelength of the light used to expose the photoresist. Preferably, S is a multiple of the IC chip design's layout grid size, which is the minimal grid used for macrocell placement during design.

[0033] Each local task region represents an area of the IC chip which will be subjected to an OPC procedure, i.e., the feature line(s) in the local task region will be subject to correction. Due to the nature of OPC and photolithography, any correction performed within the local task region may have an effect on feature lines in the area surrounding the local task region. This affected area is known as a region of optical influence.

[0034] FIG. 4 is a block diagram illustrating a local task region 502 having a size S, and its surrounding region of optical influence 504 having a size d. The region of optical influence, d, is approximately 2 . The area defined by the local task region 502 and its surrounding region of optical influence 504 is referred to as an extended task region 500. Accordingly, when correcting the feature line(s) in the local task region 502, one must also be mindful of the feature line(s) in the extended task region 500.

[0035] Referring again to FIGS. 1 and 2, once the IC chip design 118 has been divided into local task regions (502), the classification module 122 generates a characteristic vector, C(L), for each local task region, L, in the matrix in step 206. The characteristic vector describes the features and their relative orientations and locations within the local task region 502. In a preferred embodiment, the characteristic vector is:

[0036] C(L)(n, (i₁, f₁, a₁, x₁, y₁), . . . (i_n, f_n, a_n, x_n, y_n))

[0037] where n is the number of features intersecting the extended task region 500 for L. Elements (i_k, f_k, a_k, x_k, y_k) describe the k^th feature intersecting the extended task region 500, where i_k is the index of the feature type; f_k and a_k represent vertical flip and rotation angle, and x_k and y_k represent feature origin coordinates relative to the center of the local task region 502.

[0038] FIG. 5, for example, represents an extended task region 500, through which the features 1 (300) and 2 (302) (illustrated in FIG. 3) intersect. Feature 1 (300) has been flipped once along its x-axis and then rotated counterclockwise 90 degrees from its normal or original position (shown in FIG. 3). Feature 2 (302) is in its normal position, i.e., zero flip and zero rotation. If S1600, the resulting characteristic vector for the local task region 502 will be:

[0039] C(L₅₀₂)(2, (2, 0, 0, 1000, 1800), (1, 1, 90, 1400, 400))

[0040] Thus, for every local task region 502, a characteristic vector will be generated in step 206.

[0041] In a preferred embodiment of the present invention, the classification module 122 generates each characteristic vector by traversing the IC chip design 118, e.g., in GDS II format. As is well known to those skilled in the art, the IC chip design 118 is described hierarchically through levels of increased granularity. At the lowest level of the hierarchy are individual feature instances.

[0042] FIG. 6 is a flowchart illustrating a process 600 for generating the characteristic vectors according to the preferred embodiment of the present invention. The process begins, at step 601, by initializing all of the characteristic vectors to zero values. Then, in step 602, the classification module 122 traverses the hierarchical description of the IC chip to identify a first feature instance (e.g., feature 1 300 in FIG. 5). In step 604, the absolute location of the feature instance (300) is determined within the IC chip via the hierarchical description, as is well known to those skilled in the art.

[0043] Next, the classification module 122 identifies the extended task regions (e.g., 500 in FIG. 5) intersecting the feature instance (300) in step 606. Once the extended task regions 500 are identified, so too are the local task regions 502 affected by the feature instance 300. In step 608, the classification module 122 then updates the characteristic vectors associated with these local task regions 502 to include the description of the feature instance 300.

[0044] In step 610, it is determined whether more feature instances exist by, once again, traversing the hierarchical layout description. If more feature instances exist, the next feature instance (e.g., feature 2 302 in FIG. 5) is identified in step 612, and the above described process steps are repeated for each feature instance. When the entire hierarchy has been traversed and all feature instances have been analyzed, all characteristic vectors for the local task regions 502 have been updated.

[0045] Referring again to FIG. 2, once the classification module 122 has generated the characteristic vectors in step 206, they must be placed in a uniform format that allows them to be compared to one another properly. Accordingly, in step 208, the characteristic vectors are normalized.

[0046] FIG. 7 is a flowchart illustrating process 700 for normalizing a characteristic vector in accordance with a preferred embodiment of the present invention. The normalizing process is a two step process which begins after the classification module 122 identifies a first local task region 502, via step 702. In the first step, the associated characteristic vector is sorted in increasing order by feature index in step 704. So, for example, the characteristic vector for the local task region 502 represented in FIG. 5 will be, after sorting:

[0047] C(L₅₀₂)(2, (1, 1, 90, 1400, 400), (2, 0, 0, 1000, 1800))

[0048] Next, in step 706, the extended task region 500 for the local task region 502 is manipulated, so that the first feature instance 300 is in its normal position, i.e., the vector parameters f_i0 and a_i0. Referring again to the extended task region 500 in FIG. 5, in order to place feature 1 (300) in its normal position, the extended task region 500 can be flipped along its y-axis and rotated counterclockwise 270 degrees. The resulting extended task region 500 after this manipulation is presented in FIG. 8.

[0049] FIG. 8 is a diagram illustrating an extended task region 500 after normalization. As is shown, feature 1 (300) is now in its normal position and feature 2 (302) has been flipped once and rotated 90 degrees counterclockwise. Thus, the resulting normalized characteristic vector for the local task region 502 is:

[0050] C(L₅₀₂)(2, (1, 0, 0, 400, 1400), (2, 1, 90, 1800, 1000))

[0051] Referring again to FIG. 7, in step 708, it is determined whether more local task regions 502 exist. If so, the next local task region 502 is identified (step 710) and its characteristic vector normalized according to the process described above.

[0052] Referring again to FIGS. 1 and 2, after all characteristic vectors have been normalized in step 208, they are compared to determine which local task regions 502 are congruent, via step 210. Local task regions 502 are congruent if their normalized characteristic vectors are the same. In step 212, the classification module 122 groups or classifies congruent local task regions 502, and generates a list of different local task regions 126 in step 214. In step 216, the list 126 is submitted to the OPC module 120 for correction. Thus, instead of evaluating and correcting the entire IC chip design, the OPC module 120 only evaluates and corrects the local task regions 502 on the list. The amount of data analyzed by the OPC module 120 is greatly reduced, thereby saving time and computing resources.

[0053] By comparing and classifying congruent characteristic vectors, the amount of data submitted to the OPC module for correction is greatly reduced. Nevertheless, sorting and classifying the characteristic vectors can be a cumbersome task because these vectors are typically long. Accordingly, in a preferred embodiment, a hashing unit 124 (FIG. 1) in the OPC classification module 122 hashes each normalized characteristic vector to produce a corresponding hash value. Comparing these hash values significantly simplifies the sorting and classification process.

[0054] FIG. 9 is a flowchart illustrating a process 800 for sorting and classifying the characteristic vectors to produce the list 126 submitted to the OPC module 120, according to a preferred embodiment of the present invention. In step 802, the hash unit 124 selects two random regular bit matrices A and B of size NN, where N is a constant, e.g., N32. Then, each characteristic vector is treated as a sequence of bits (step 804), divided into m groups of N, such that:

[0055] C(L)(X₁, . . . , X_m)

[0056] where X₁, . . . , X_m are bit vectors of size N. Note that the characteristic vector is simply presented in a different way and that no manipulation or calculation is involved.

[0057] Next, in step 806, the hashing unit 124 calculates a hash value, H(L), for each characteristic vector according to the following iterative procedure:

[0058] H₁X₁

[0059] H₂AX H₁B

[0060] H₃AH₂B

[0061] . . .

[0062] H_mAH_m1BH(L)

[0063] The resulting hash value, H(L), is a short, constant sized vector (e.g., 32 bits), which is practical for sorting and classifying the characteristic vectors. In step 808, sorting and classification is performed by hash value. Because there is a small, but finite, chance that non-congruent local task regions will have the same hash value, an additional sub-classification by characteristic vector is performed in step 810. Thus, at the end of this process, congruent local task regions 502 are classified without having to compare normalized characteristic vectors to one another.

[0064] A method and system for classifying an IC chip into congruent local task regions for performing OPC correction has been disclosed. By identifying and grouping congruent local task regions in the IC chip, only one OPC procedure (e.g., evaluation and correction) needs to be performed per group of congruent local task regions. The amount of data to be evaluated and the number of corrections performed is greatly reduced because OPC is not performed on repetitive portions of the IC chip design, thereby resulting in significant savings in computing resources and time.

[0065] The present invention has been described in accordance with the embodiments shown, and one of ordinary skill in the art will readily recognize that there could be variations to the embodiments, and any variations would be within the spirit and scope of the present invention. Accordingly, many modifications may be made by one of ordinary skill in the art without departing from the spirit and scope of the appended claims.