JP2007102207A - Creating and applying variable bias rule in rule-based optical proximity correction for reduced complexity - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an optical proximity correction (OPC) based integrated circuit (IC) design system and method for applying specified correction rules in rule-based OPC, that is more fracture-friendly and generates simpler OPC output, without significantly compromising accuracy. <P>SOLUTION: The optical proximity correction (OPC) based integrated circuit design system and method introduce a variable rule where the rules are specified in terms of multiple correction actions that yield acceptable results, but this category of rules provides more degrees of freedom in actual application so that the rule-based OPC tool can intelligently select the proper valid rule that minimizes the OPC complexity or meets other objectives. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates generally to systems and methods for designing integrated circuits for fabrication by a semiconductor manufacturing process, or transferring between technology nodes of an integrated circuit design, and more particularly, for integrated circuit design layout. The present invention relates to a system and method for performing optical proximity effect correction.

  This application is a method for creating and applying variable bias rules in METHOD FOR CREATING AND APPLYING VARIABLE BIAS RULES IN RULE-BASED OPTICAL PROXIMITY CORRECTION FOR REDUCED COMPLEXITY. And relates to US Provisional Patent Application No. 60 / 619,160, filed October 15, 2004.

  The semiconductor manufacturing industry has continued to develop semiconductor design and fabrication processes, and to develop processes for producing smaller and more compact designs. Semiconductor devices operate at higher speed with less power consumption and less heat generation than devices with larger geometries. Furthermore, the smaller the geometry, the more circuit elements can be built into the silicon chip, thus making the integrated circuit (IC) more complex and the same that can be produced on a single silicon wafer. The number of die copies can be increased. Currently, some single semiconductor chips contain over one billion patterns. As a result, IC design and semiconductor fabrication processes are overly complex because hundreds of processing steps may be involved. The occurrence of errors or small errors in either the design or process steps will require redesign, otherwise it can cause yield loss in the final semiconductor product. Yield can be defined as a comparison to the theoretical number of devices that can be produced assuming that there are no defective devices in the number of functional devices produced by the process.

  Improving time to market and yield is a vital issue in the semiconductor manufacturing industry and has a direct economic impact on the semiconductor industry. That is, shortening time to market and improving yield can be interpreted as being available early and increasing the number of devices that manufacturers can sell.

  The IC design layout consists of a number of geometries in the form of polygons. Polygons are used to build different features consisting of the whole polygon or a part of it that is typically characterized by the required geometric properties such as dimensions. Each feature constitutes one or more shapes that represent one or more edges of the feature. A topology configuration is formed by a combination of adjacent features. This is often referred to as a pattern or structure. Therefore, the IC layout can be regarded as constituting a large number of repeated patterns, and any number of such patterns constitutes a part of the IC layout. These patterns of the IC layout are transferred to the silicon ware mainly by a process called lithography. In the most commonly used lithography process, referred to as photolithography, a mask or reticle having transparent and opaque areas representing structures in one IC layout is illuminated by a light source. Next, the light emitted from the mask is focused on the photoresist layer deposited on the wafer. The wafer is then developed to have the removed resist portions and etched wafer portions to form the desired geometric pattern. Typically, in IC designs with large feature dimensions, the design pattern is transferred to the mask with high accuracy and then transferred to the wafer with a lithography process, commonly referred to as WYSIWYG ("What you see is what you get"). Reach the phenomenon called.

  Due to the ever-decreasing size of features, higher pattern density, and difficulties arising in the development of IC manufacturing equipment, the manufacture of modern IC designs has feature dimensions near and at wavelengths that are below the wavelength of the light source. Faced with major injuries in the area below, as well as associated yield problems. The fact that imaging diffraction is limited in the vicinity of the wavelength and in the sub-wavelength region has caused the disappearance of the previous WYSIWYG framework. With the advent of near-wavelength and sub-wavelength lithography, patterns projected onto the wafer through the lithographic process have become severely distorted. As shown in FIG. 1 (a), typical distortions are line edge displacements that are not rare in line edge shapes, rounded corners exhibited by corner shapes, and line end shortening that occurs in line end shapes. Including.

  In view of the growing gap between near-wavelength and sub-wavelength design manufacturability, the use of optical resolution enhancement techniques (RET) such as optical proximity correction (OPC) is less than 0.18 μm It is popular in many designs and production methods to produce a feature size of. Typically, for example, a design tape-out in the form of GDSII is input to a RET implementation (RET implementation), eg, OPC data conversion, and a new GDSII design tape out is used as a reference GDSII data is generated.

  OPC is a process that modifies the physical layout from the design tape out to correct pattern distortions caused by light diffraction, resist development, etching, and other undesirable effects that occur during the lithography process. . Two types of methods are known for OPC. One is rule-based OPC, where the correction rules are determined in advance and how they vary depending on the shape to be considered, eg some simple measure related to features with width and spacing Specifies whether to correct the geometric shape. There are three main types of rule-based OPC. The first type is a hammerhead, which is applied to the line end shape, as shown in FIG. Another type is serif, which is applied to corners to correct corner rounding, as shown in FIG. The last type is edge bias, which corrects line edge displacement as shown in FIG. Bias correction is a more comprehensive form of correction since more complex corrections are represented by a collection of individual edge biases. For example, serif or hammerhead correction can be viewed as applying individual biases to two or more consecutive edge segments, and in the case of serifs, as shown in FIG. Fill. One polygon may contain many shapes, and as demonstrated in the example shown in FIG. 1 (b), all three types of corrections may be required. Rule-based OPC is usually simple and fast based on geometry. However, rule-based OPC is not very accurate, and therefore more advanced technology nodes such as technology nodes greater than 0.25 μm, or metals or layers with loose accuracy requirements and containing only relatively large geometries. Only used on insignificant layers (less than 0.18μ). Furthermore, the generation of correction rules is not systematic and often requires experience, and as the complexity of the geometry increases, the number of rules increases, and perhaps the complexity of the rules also increases. Is difficult to maintain.

  Another OPC technique is model-based OPC. Model-based OPC uses a lithography model that incorporates the effects of the optical components and the overall layout into the silicon pattern transfer process to simulate and predict the corresponding pattern on the wafer. And calculate and apply the necessary corrections for each geometry shape. The lithography model consists of a) the optical model, and b) the effects of other processes including chemicals, etching, and other factors. Optical models typically operate within a finite range called the proximity range. When simulating an optical image for a position commonly referred to as an evaluation point, only layout geometries that are within the radius of a proximity range centered on the evaluation point are considered, and geometry outside this range is ignored. The calibrated lithography model incorporates the effects of the optical model and other processes. To obtain a calibrated model, the inspection mask is illuminated, a wafer image is formed, measurements are taken, and then the data is fitted. The lithography model is independent of the physical layout.

  Model-based OPC processes typically involve a step called dissection, where polygons can be broken down into edge segments and these can be moved individually to correct each segment. An evaluation point is specified on the segment, and the model is evaluated there to calculate certain main wafer characteristics, such as edge displacement, as shown in FIG. Calculate the required edge movement in the field to minimize the simulated edge displacement. Because of the use of models and correction algorithms, correction does not require correction rules a priori and is usually more accurate than those corrected by the rule-based approach. On the other hand, due to the high granularity of correction, the correction layout obtained by the model-based method is typically more complex than that by the rule-based method. Therefore, in general, model-based OPC is used entirely for critical layers of designs of 0.18 μm or less.

  Following the OPC process, a mask data preparation (MDP) step is performed on the data to break up the data and write the mask using the results. Today, two main writing techniques are used in mask manufacturing. The first technique, called “raster scanning”, scans the entire mask with an electron or light beam and turns it on when the mask is exposed. The second approach, called “vector scan”, exposes the tangible beam at the required coordinates representing the data on the mask where the mask is to be exposed. For a tangible beam exposure tool, the data typically only needs to include a certain set of angles. Typically, these angles are 45 degrees, 90 degrees, and 135 degrees due to shape constraints that can be generated by the exposure tool. In this writing technique, the MDP process is a fracturing that involves taking polygons of complex shapes and breaking them into smaller primitives, typically trapezoids, which can be written by a mask writer. FIG. 2 shows an example of breaking a polygon into a combination of rectangles. The number of primitives after destruction for IC layout is called “shot count” or “figure count”. In a tangible beam exposure tool, the majority of the total mask writing time is directly proportional to the shot count. Longer mask writing times mean longer turnaround times, longer exposure tool wear and wear, and therefore higher mask costs. As a result, the smaller the shot count, the better from the viewpoint of mask cost. On the other hand, OPC typically increases the complexity of the layout geometry, resulting in a dramatic increase in shot count. In FIG. 2, for example, the original polygon 1100 has six vertices, which can be divided into two rectangles 1101 and 1102. After OPC, the resulting polygon 1200 has 28 vertices and is divided into 12 rectangles 1201 to 1212, so it increases 6 times. In addition, two slivers are created. These small slivers result in inaccurate exposure when exposing the mask, which results in reduced dimensional accuracy. If OPC can be simplified and applied without significantly degrading accuracy, the shot count can be reduced and the sliver can be removed. Referring again to FIG. 2, if the OPC can be simplified so that small correction jogs 1001, 1002, and 1003 can be removed, the resulting OPC output 1300 includes 22 vertices and 9 Dividing into rectangles, the shot count is reduced by 25% without generating a sliver.

  Part of the reason why rule-based OPC is more intriguing than model-based OPC is that rule-based OPC has a lower complexity of polygons created than model-based OPC, and therefore has a lower shot count, As a result, the mask cost is lowered. However, even with the rule-based OPC, the obtained OPC output may still be difficult to be divided depending on the environmental configuration. FIG. 2 shows a polygon 1400 having three adjacent features 1410, 1411, and 1412. A typical rule-based OPC using a width-space formula creates the following OPC features: Hammer head 1401, outer serifs 1402, 1403 and 1404, inner serif 1405, and biases 1406, 1407, 1408 and 1409. Bias 1406 and 1407 are different because of the change in spacing caused by geometry 1410, while biases 1408 and 1409 are different because of the change in spacing caused by geometry 1411. Further, due to the presence of the geometry 1412, the serif 1404 differs from the serifs 1402 and 1403. As a result of this OPC, as shown in FIG. 2, rectangles 1501 to 1511 are generated by the division. Sliver 1503 results from a mismatch between the 1406 to 1407 bias jump and the 1408 to 1409 bias jump.

  In this way, the division based on the rule-based OPC is difficult because the “best” correction or bias amount is strictly applied according to the correction rule for obtaining the most accurate result, and there is a change in the surrounding environment. In some cases, it is inevitable to introduce a correction jog. Current techniques for further simplifying the OPC output involve a smoothing step and the non-smooth correction consists of a number of correction jogs within a certain tolerance band, as shown in FIG. 3 (a). As shown in Fig. 4, the surface is flattened so that jogs are eliminated. This technique is ad hoc because it does not provide a way to control the accuracy of the correction. Indeed, after such a change in OPC output, the correction accuracy can be significantly affected. There is a need for a method that simplifies and generates an OPC output without specifying a correction rule and applying these rules to a rule-based OPC that can be easily divided without significantly deteriorating accuracy.

  The present invention aims at this. Various embodiments of the present invention provide numerous advantages over conventional IC design methods and systems.

  An embodiment of an IC design system and method based on optical proximity correction (OPC) according to the present invention provides numerous advantages over conventional design systems and techniques, so that an IC design system for performing OPC according to the present invention. And the method will be more useful to semiconductor manufacturers. In accordance with various embodiments of the present invention, variable bias rules are introduced, such as varying bias ranges or discrete bias values when bias is required on certain types of grids that yield acceptable results. Specify rules for the set. This classification of rules increases the degree of freedom in actual application so that the rule-based OPC tool can intelligently select the appropriate effective rules that minimize the OPC complexity.

  Accordingly, one embodiment of the present invention provides an IC design system and method for creating and applying variable bias rules in rule-based OPC to reduce OPC complexity. According to the present invention, a system for applying rule-based OPC is provided, where each rule is described by a range of dependent variables and a corresponding range of bias values, which is a set of all effective bias values. .., So that the corrected features are within the allowable range specified in advance. In addition, a method of applying rule-based OPC is provided, where each rule is described by a range of dependent variables and a corresponding range of bias values, each of which is corrected so that the corrected features meet the required performance objectives. Preferably, the bias value is selected from a range or set of values so as to minimize the number of additional jogs that are played.

  The foregoing and other objects, aspects and advantages of the invention will become more readily apparent from the following detailed description of various embodiments, which proceeds with reference to the accompanying figures.

To facilitate an understanding of the present invention, various embodiments of the present invention will be described in connection with the accompanying drawings of the drawings. In the figures, like reference numerals indicate like elements.
The invention is particularly applicable to software-based IC design systems implemented on computers for generating IC designs using rule-based optical proximity effect correction (OPC) to correct IC design layouts. It is in this context that various embodiments of the present invention are described. However, the IC design system and method for providing rule-based OPC to reduce complexity according to various embodiments of the present invention can be implemented in hardware as well as other modules or functionality not described herein. It will be appreciated that it can be incorporated, so it has more utility.

  FIG. 4 is a block diagram showing an example in which an IC design system 10 that provides a rule-based OPC for reducing complexity according to an embodiment of the present invention is mounted on a personal computer 12. That is, the personal computer 12 has a display unit 14, which can be a cathode ray tube (CRT), a liquid crystal display or the like, a processing unit 16, and a software application executed by the personal computer and a user performing bidirectional processing. One or more input / output devices 18 are provided. In the illustrated example, the input / output device 18 can include a keyboard 20 and a mouse 22, but can also include other peripheral devices such as a printer, a scanner, and the like. The processing unit 16 may further include a central processing unit (CPU) 24, a permanent storage device 26 such as a hard disk, tape drive, optical disk system, removable disk system, and memory 28. . The CPU 24 can control the persistent storage device 26 and the memory 28. Typically, a software application can be stored persistently in persistent storage 26 and then loaded into memory 28 when the software application is executed by CPU 24. In the illustrated example, the memory 28 can accommodate an IC design tool 30 that performs rule-based OPC. IC design tool 30 may be implemented as one or more software modules executed by CPU 24.

  According to the present invention, the IC design system 10 that provides rule-based OPC for reducing complexity can also be implemented using hardware, such as a client / server system, a web server, a mainframe computer. It can also be implemented on different types of computer systems such as workstations. An implementation example of the IC design system 10 as software will now be described in more detail.

  One embodiment of the present invention performs hybrid OPC for IC design tape out. For example, an IC design tape out can be a GDS or OASIS file, or a file having other formats.

  Various embodiments of the present invention provide a method for specifying correction rules in terms of a set of valid corrections or bias values that can be selected by an actual rule-based OPC processor, rather than in terms of deterministic corrections or bias values. provide. In one embodiment of the present invention, the rule-based OPC processor selects a correction or bias value from the set of available values and applies the correction to minimize the complexity of the OPC output. That is, the rule-based OPC processor applies correction so as to minimize the number of vertices in the OPC output.

The correction rule can be expressed mathematically as R = F (C ₁ , C ₂ ,...), Where C ₁ , C ₂ ,. . . Is a “condition”. Possible conditions are: pattern shape (line or reverse line end, inner or outer corner, jog, etc.), pattern dimensions (width, spacing, height, etc.), circuit characteristics (gate, end cap, contact enclosure, etc.), And any of the foregoing, and a more complex characterization combining all. Rule R itself adds a certain size serif (the serif contains two bias values) that moves the edge outward by a certain amount (bias), a certain size hammer head (hammer 3 heads) A more complex act that combines one or more of these actions (for example, increasing the poly line bias and reducing the diffusion bias by the same amount) Command the action to be performed on the pattern. Complete rule set may also include complex rules with different formulation (designated different c _i, different forms of f, and a different R). The complete rule set can be easily represented in tabular form, where c _i is the “key” column and R is the value column (as in the database).

In the more precise and commonly used form of rules used in rule-based OPC, the form of the correction rule can be expressed as a multivariable function, ie (b ₁ , b ₂ ,..., C ₁ , c ₂ , ) = F (g ₁ , g ₂ ,..., D ₁ , d ₂ ,...), And g _i is the line edge, line end, corner, gate, contact enclosure (contact enclosure), or such as that a large number of design layer Maga upcoming complex boolean or shape specified by the geometrical calculation, a variety of shapes, d _i is the width, spacing, height, enclosure margin or, It is a function of a number of dimensions measured by a complex formula. The output of the rule is composed of one or more bias correction amounts b _i that give correction applied to the pattern shape, and auxiliary parameters c _i that accurately specify the portion of the shape to which the bias is applied. Typically, the shape is pre-selected (eg, an edge or part of an edge) and the rule function is (b ₁ , b ₂ ,...) = F (g ₁ , g ₂ ,..., D ₁ , d ₂ ,. If the shape consists of a single edge segment, the rule function simplifies to b = f (g ₁ , g ₂ ,... D ₁ , d ₂ ,...). Where b is the bias applied to the edge segment. This is a comprehensive representation as other more complex shapes can be decomposed into a collection of single edge segments as described in the prior art section. Expressed in a table format, each rule is g ₁ , g ₂ ,. . . , D ₁ , d ₂ ,. . . Contains a "key" column, and the "value" column contains b. In other words, each rule is a mapping between the value of the dependent variable and the bias value. The smaller the interval, the more accurate the rule, but the larger the rule table. An example of such a rule table is as follows.

  A complete rule to be applied to the IC layout is composed of a large number of rule sets, and different conditions are designated for each rule set. For example, a rule for correcting a poly layer can consist of three rule sets, one applied to the transistor gate (a condition that identifies the applicable poly pattern) and another one is: Applies to poly lines that are at least some distance away from diffusion (another condition that identifies applicable poly patterns), and the rest of the poly pattern (a third condition that identifies applicable poly patterns) To do. These conditions are then used to identify different patterns and use the corresponding rule set when applying rule-based OPC to these patterns.

  Conventional rule-based OPC solutions are based on one deterministic enemy action for each matching condition and typically do not consider the complexity of the OPC output. For example, in a rule where the correction is a bias amount, there is a single bias value associated with the correction rule, as illustrated in the previous example (previous rule table). For the table format, each rule entry has a single correction value. For this reason, as shown in FIG. 5A, when a transition of width or space occurs at an edge, there is a possibility that a jog is induced in the correction output.

According to the present invention, at least one correction rule is characterized by at least two, but possibly more, correction actions on the condition-specified set. In a more general form, such a multivalued rule can be represented by ({R}) = f (C ₁ , C ₂ ,...), Where {R} is the set of all possible corrections. It is. For example, {R} = {R ₁ , R ₂ ,. . . } Is a set of corrective actions R ₁ , R ₂ ,. . . . Means an aggregate of As another example, {R} = [bmin, bmax], where the corrective action is a bias amount that can have any value from bmin to bmax. During the actual correction, one action is selected in order to achieve a preset purpose, typically a reduction in mask complexity.

  Rule determination can be based on simulation or manual. Simulation-based methods are automated methods for obtaining bias. Filed on the same day as this application, MODEL-BASED PATTERN CHARACTERIZATION TO GENERATE RULES FOR RULE-MODEL-BASED HYBRID OPTICAL PROXIMITY CORRECTION The co-pending US patent application entitled, describes a systematic method for generating rules using simulation. The entire contents thereof are included in the present application by quoting here. For example, this approach simulates a pattern with a specified width and spacing and sets a bias value that yields the smallest edge placement error (EPE). If one specifies an EPE tolerance range and can be further tightened as an option, as described in the above patent application, as shown in FIG. There is a range of bias values that can yield a certain EPE. This range is called “variable bias range”. In principle, any bias value falling within this range can be selected and applied by the rule-based OPC processor. In practice, there may be additional constraints (such as a correction grid) that limit this range to several distinct values (eg, all integer values within this range). Within this variable bias range, there is one value that is considered “optimal”. Such optimal bias is usually based on metrics such as EPE (optimal bias yields the minimum absolute value of EPE as shown in FIG. 6), image tilt, and mask error enhancement factor (MEEF). Conventional rule-based OPC is based on a single bias value and is typically one such optimal bias. Embodiments of the present invention introduce a variable bias rule format that allows selection from a bias value within a variable bias range, or a collection of multiple bias values. This increase in the degree of freedom in selecting a rule-based OPC can be applied to minimize the complexity of the OPC result.

  FIG. 5 shows an example comparing the use of a new multi-value rule and a conventional single-value rule when applying rule-based OPC. A pattern having a width w0 has changing neighboring intervals s1 and s2. Assume that s1 and s2 are sufficiently far apart that they fall in the category of different spacing intervals in the rule table. When the single value rule in FIG. 5A is applied, the bias values corresponding to (w0, s1) and (w0, s2) are different, and after applying the rule-based OPC, the result of jogging at the interval transition position is Become. This adds two new vertices to the output. However, even when s1 and s2 are different, the variable bias ranges corresponding to (w0, s1) and (w0, s2) may overlap as shown in FIG. As a result, the rule-based OPC according to one embodiment of the present invention selects a single bias value that falls within both bias ranges. This bias value is valid for both situations. This results in a single bias value without adding any new vertices. For this reason, OPC is simplified.

  FIG. 7 illustrates an example where jogs can be removed after applying rule-based OPC according to an embodiment of the present invention, even if jogs exist in the original geometry before OPC. As shown in FIG. 7, jogging causes transitions in both pattern width and spacing, resulting in two width / spacing pairs (w1, s1) and (w2, s2). This corresponds to two different rules in the rule table. The bias b1 can be selected from [b1min, b1max], the bias b2 can be selected from [b2min, b2max], and the difference between b1 and b2 is the size of the jog as shown in FIG. And applying these biases eliminates jogs. As a result, the OPC output actually removes the four vertices that originally make up the two jogs.

  FIG. 8 shows another example in which a completely smooth output cannot be obtained, but an output that minimizes the figure count is possible. The original feature has a width transition from w1 to w2 and an interval transition from s1 to s2. As a result, three rule entries corresponding to (w1, s1), (w2, s1), and (w2, s2) are obtained, and the bias ranges are [b1min, b1max] and [b2min, b2max], respectively. And [b3min, b3max]. If the three bias ranges overlap and there is a single bias value that falls within all three ranges, jogging can be avoided by selecting and applying such biases. However, in this example, the variable bias ranges [b1min, b1max] and [b3min, b3max] do not overlap, but the bias ranges [b2min, b2max] overlap. Then, considering the change in width caused by the opposing edge, the rule-based OPC according to an embodiment of the present invention can be obtained from the intersection of [b2min, b2max] and [b3min, b3max] and from [b3min, b3max] A bias is selected and applied to the three edges. A jog occurs at the output, but this jog aligns with the opposing jog as shown in FIG. 8 (c), so no additional graphics are created after the split. When using three different bias values for the three width and spacing conditions as in the conventional rule-based OPC (shown by the dashed line in FIG. 8A), as shown in FIG. Additional shapes will be created later.

  FIG. 9 is a flowchart for applying rule-based OPC from a variable rule set that minimizes the complexity of OPC output. Importantly, instead of choosing a rule table that consists of a single corrective action that each rule optimizes performance objectives, ie, minimizes EPE, the rule-based OPC selects each rule. It is possible to use a variable rule set consisting of several corrective actions that achieve the performance objectives, all within tolerances, and to select the corrective action that yields the simplest OPC output. As shown in FIG. 9, in a first step 8001, pattern characteristics (such as width, interval, pattern type, etc.) are extracted for each shape in the layout. Then, in step 8002, the corrective action is looked up in the rule table or calculated according to the rule expression. In step 8003, the different outputs possible with different combinations of corrective action are examined. When the correction action is determined for each shape, some or all of the shapes have more than one correction action. First, in step 8004, a corrective action is determined for these shapes that does not produce a new jog after applying the rule, or removes the jog in the original layout. For the remaining shapes, the correction jog is inevitable no matter how the correction action is selected. In an optional next step 8005, corrective action is determined for the remaining shapes and after applying the rules, the resulting jog aligns with the vertices of the opposite shape (avoids creating additional shapes in the cut). . For the remaining shapes, the selection of the corrective action does not simplify the OPC output, so the optimal corrective action is applied to these shapes as in conventional rule-based OPC. For example, a correction action that applies the smallest absolute value of EPE is applied. As shown in FIG. 9, newly introduced steps 8003 to 8005 utilize selection of a correction rule that generates an OPC output with reduced complexity compared to conventional rule-based OPC.

  One possible change to this method is to reduce the variable bias rule to a single bias rule by rule fusion aimed at reducing correction variability and avoiding jogging. For example, assume that there are four rules from a rule set shown in the following table format.

These four rules can be merged into a single rule as follows.

  This rule illustrates that the same 6 nm bias can be obtained for any feature with a width variation in the range of 300-400 nm, a spacing variation of within 400-600 nm, and a height range of 800-900 nm. To do. Therefore, for features with width or spacing transitions within these ranges, no additional jogs will occur after the rule-based OPC.

  Another embodiment of the invention involves adding a small exception tolerance when matching or applying correction rule conditions to reduce correction complexity. FIG. 10 shows some examples. FIG. 10 (a) shows a pattern with three measured width / interval pairs (w1, s1), (w2, s2), and (w3, s3), where (w1, s1) and (w2 , S2) each have associated rules r1 and r2 (both are ranges as shown by the dashed box). There is no rule associated with separating two edges (w3, s3). When OPC is strictly applied according to the rule, the correction shown in FIG. 10B is obtained, and one jog in the original shape is replaced with two jogs in the corrected shape. However, if the edge portion associated with (w3, s3) is small (ie h is smaller than the pre-specified tolerance), the rule-based OPC ignores this small exception and r2 to include this small segment. Extend the edge to fit. In this case, the rule-based OPC can select the correction so as to remove the jog in the original shape as shown in FIG. This is because the bias ranges for r1 and r2 overlap as shown in FIG. FIG. 10 (d) shows another example where the original geometry 1601 has a small jog 1603 along the edge 1602. The edge 1602 has a width w and an interval s that satisfy the conditions of the rule having the corresponding bias range [bmin, bmax]. The jog has a width w1 and an interval s1, and there is no corresponding rule in the rule table. Therefore, the rule-based OPC correction based on strict rule matching selects a certain bias amount from the range [bmin, bmax], and generates the fraud 1604 shown in FIG. In this case, the original jog still exists and remains uncorrected. However, if the jog size h is small, the rule-based OPC can extend the correction to cover this jog, so that the correction shown in FIG. 10 (f) has a smooth output and no jog. 1605 is obtained. In yet another example shown in FIG. 10 (g), the original graphic 1611 has a width w and no jogs, but due to the adjacent feature 1620, a transition from s to infinity along the feature 1611. Will occur. For the portion of the edge 1612 whose width is w and interval is s, a corresponding rule whose bias range is [bmin, bmax] is obtained, but the remaining portion of the edge 1613 is the width w, infinite interval and small size h. And no matching rule is sought. Therefore, the rule-based OPC correction based on strict rule matching selects a certain bias amount from the range [bmin, bmax], and generates the correction 1614 shown in FIG. This adds jogs that were not in the original features. However, if the size of the mismatched edge portion h is small, the rule-based OPC can extend the correction to cover this portion with the same amount of bias, resulting in a smooth output as shown in FIG. And a jog-free correction 1615 is obtained.

  So far, in selecting the corrective action for each rule, the rule-based OPC has been centered on making a selection so as to minimize the complexity of the OPC output. However, when making a selection, it does not exclude the use of other objectives for the rule-based OPC, but includes those that harmonize with the traditional “optimal” correction action. For example, rule-based OPC can minimize absolute EPE or CD error, maximize image tilt or contrast, maximize dose or defocus latitude, or increase aberration or mask error. A corrective action can be selected based on the minimization of (MEEF). The rule-based OPC can also be selected to select different purposes for different types of patterns, for example, for transistors and gates with minimal EPE and for features with small critical dimensions. Image tilt can be maximized and OPC output complexity can be minimized for features with large critical dimensions.

  Although the foregoing description has been made with reference to specific embodiments of the invention, those skilled in the art will recognize that modifications may be made to these embodiments without departing from the principles and spirit of the invention. I will be. Accordingly, the scope of the invention can only be ascertained with reference to the appended claims.

FIG. 1, consisting of FIGS. 1 (a) to 1 (g), shows pattern distortion from a lithography process and application of optical proximity correction (OPC). FIG. 2 shows an example of splitting polygons into rectangular combinations, and the increase in figure count and sliver incorporation by applying OPC to simple geometries. FIG. 3 consists of FIG. 3 (a) and FIG. 3 (b), consists of a number of correction jogs within the tolerance band with non-flat correction, flattening the OPC output so that there is no jog. Show. FIG. 4 is a block diagram illustrating an example of an IC design system that performs model-based pattern characterization and rule generation, and hybrid OPC, according to one embodiment of the present invention. FIG. 5 consists of FIGS. 5 (a) and 5 (b), comparing a multi-value rule according to various embodiments of the present invention with a conventional single value rule, and a rule-based OPC according to an embodiment of the present invention. An example in which possible outputs are compared when rules are applied using, is shown. FIG. 6 shows a range of bias values that can result in an edge placement error (EPE) within an optionally tightened tolerance range. FIG. 7 illustrates an example where a jog that originally exists in the pre-OPC geometry can be removed after applying rule-based OPC in accordance with various embodiments of the present invention. FIG. 8 is composed of FIGS. 8A to 8C, and shows another example in which an output that suppresses the division count to a minimum is possible although a completely smooth output cannot be obtained. FIG. 9 shows a flow diagram for variable bias rule determination according to one embodiment of the rule-based OPC method of the present invention. FIG. 10 comprises FIG. 10 (a) to FIG. 10 (i), and shows small exception processing when the correction rule condition is matched to reduce the correction complexity, or when the correction rule is applied.

Claims (44)

A correction rule specifying method using a correction rule for rule-based optical proximity effect correction (OPC),
A correction rule specifying method in which a correction rule is specified, each correction rule includes a set of matching conditions and a corresponding correction action set, and at least one correction rule has at least two corresponding correction actions.   The method of claim 1, wherein the match condition is parameterized and specified with respect to a rule table.   The method of claim 1, wherein the corrective action is parameterized and specified in terms of edge movement.   4. The method of claim 3, wherein the at least one correction rule having at least two corresponding correction actions is specified with respect to an amount of edge movement given in a set of discrete numbers.   4. The method of claim 3, wherein the at least one correction rule having at least two corresponding correction actions is specified with respect to an amount of edge movement provided by at least one continuous range.   4. The method according to claim 3, wherein a correction action is designated by two parameters representing edge movement amounts of two edges of a corner so that a result of the correction after application of the correction rule is a serif. .   7. The method of claim 6, wherein the at least one edge displacement is given in a discrete number set.   The method of claim 6, wherein the at least one edge displacement is given in a set of continuous ranges.   The method according to claim 3, wherein a correction action is designated by three parameters representing edge movement amounts of three edges of the line end so that the correction result after application of the correction rule is a hammer head. Method.   The method of claim 9, wherein the at least one edge movement amount is given in a discrete number set.   The method of claim 9, wherein the at least one edge movement amount is given in a set of continuous ranges.   2. The method of claim 1, wherein the at least one correction rule having at least two corresponding correction actions is integrated by the rule-based OPC by selecting one actual correction action from the set of correction actions. The method applied to the layout.   13. The method of claim 12, wherein the selection of correction rules is based on minimizing output complexity.   14. The method of claim 13, wherein the output complexity minimization may be formed by avoiding jogging when applying more than one correction rule to the same polygon edge. Including minimizing the number of vertices.   14. The method of claim 13, wherein the output complexity minimization is formed by removing jogs in the original layout geometry when applying one or more correction rules to multiple edges. A method comprising minimizing the number of possible vertices.   14. The method of claim 13, wherein minimizing the complexity of the output includes aligning vertices formed on the edge with its opposing vertices to avoid the formation of additional fragmented figures. .   14. The method of claim 13, wherein the output complexity minimization allows exceptions to correction rules and expands corrections to exception shapes by rule-based OPC, otherwise caused by exception shapes in the OPC output. Removing the vertices that do.   The method of claim 17, wherein the exceptional shape is a shape having a height or width that does not exceed a pre-specified threshold.   13. The method of claim 12, wherein the selection of a correction rule is based on obtaining a minimum absolute value of edge placement error (EPE) for a corresponding edge.   13. The method of claim 12, wherein the selection of a correction rule is based on obtaining a maximum absolute value of an intensity slope for the corresponding edge.   13. The method of claim 12, wherein the selection of correction rules is based on obtaining an optimal focus latitude for a corresponding edge.   13. The method of claim 12, wherein the selection of a correction rule obtains one of a group of performance metrics consisting of a minimum mask error enhancement factor (MEEF), a minimum sensitivity to aberration, and a maximum image contrast. Based on the method. A rule-based optical proximity correction (OPC) system,
Means for receiving a correction rule comprising a set of matching conditions and at least two corresponding correction actions;
Means for determining an actual correction action by selecting one correction action from the at least two correction actions;
An optical proximity effect correction system.   24. The system of claim 23, wherein the match condition is parameterized and specified with respect to a rule table.   24. The system of claim 23, wherein the corrective action is parameterized and specified in terms of edge movement.   26. The system of claim 25, wherein the correction rule is specified with respect to an amount of edge movement given in a set of discrete numbers.   26. The system of claim 25, wherein the correction rule is specified in terms of an amount of edge movement given by at least one continuous range.   26. The system according to claim 25, wherein a correction action is designated by two parameters representing edge movement amounts of two edges of a corner so that a result of the correction after application of the correction rule is a serif. .   29. The system according to claim 28, wherein the amount of edge movement is given in a set of discrete numbers.   30. The system of claim 28, wherein the edge movement amount is given in a set of continuous ranges.   26. The system according to claim 25, wherein a correction action is designated by three parameters representing edge movement amounts of three edges at a line end so that a result of the correction after application of the correction rule is a hammer head. system.   32. The system of claim 31, wherein the at least one edge movement amount is given in a discrete number set.   32. The system of claim 31, wherein the amount of edge movement is given in a set of continuous ranges.   24. The system of claim 23, further comprising means for receiving an integrated circuit layout, and applying rule-based OPC correction to the integrated circuit layout.   35. The system of claim 34, wherein the selection of correction rules is based on minimizing output complexity.   36. The system of claim 35, wherein the output complexity minimization may be formed by avoiding jogging when applying more than one correction rule to the same polygon edge. A system that includes minimizing the number of vertices.   36. The system of claim 35, wherein the output complexity minimization is formed by removing jogs in the original layout geometry when applying one or more correction rules to multiple edges. A system that includes minimizing the number of possible vertices.   36. The system of claim 35, wherein minimizing the complexity of the output includes aligning vertices formed on the edges with their opposing vertices to avoid the formation of additional fragmented shapes. .   36. The system of claim 35, wherein the output complexity minimization allows exceptions to correction rules and expands corrections to exception shapes by rule-based OPC, otherwise caused by exception shapes in the OPC output. A system that includes eliminating vertices to perform.   40. The system of claim 39, wherein the exceptional shape is a shape having a height or width that does not exceed a pre-specified threshold.   35. The system of claim 34, wherein the selection of a correction rule is based on obtaining a minimum absolute value of edge placement error (EPE) for a corresponding edge.   35. The system of claim 34, wherein the selection of a correction rule is based on obtaining a maximum absolute value of an intensity slope for that edge.   35. The system of claim 34, wherein the selection of correction rules is based on obtaining an optimal focus latitude for a corresponding edge.   35. The system of claim 34, wherein the correction rule selection obtains one of a group of performance metrics consisting of a minimum mask error enhancement factor (MEEF), a minimum sensitivity to aberration, and a maximum image contrast. Based on the system.

Images