CN101013271A - Method for correcting layering optical proximity effect - Google Patents

Abstract

The hierarchical optical proximity effect correction method disclosed by the present invention includes the preparation of the pre-correction unit library, the initialization of the mask pattern offset and the steps of fast calculation using a dynamic adjustment algorithm, and proposes a method for correcting the optical proximity effect in the process of correcting the optical proximity effect By making full use of the layout hierarchical structure method, the high-precision reticle preparation process, which is considered to be very expensive due to the large amount of calculations under deep submicron conditions, can greatly reduce the computational complexity with the help of the new algorithm. This new method provides a cell library with pre-corrected results for the IC design process, enabling design engineers to more flexibly and efficiently check the design. The method proposed in the present invention can be used to assist in the manufacture of high-precision integrated circuit reticles, improve the predictability of results in the design of integrated circuits, improve the calculation speed of OPC correction, reduce costs, improve the production yield of integrated circuits and shorten Production cycle.

Description

A Hierarchical Correction Method for Optical Proximity Effect

technical field

The invention relates to a layered optical proximity effect correction method, which is suitable for assisting in the manufacture of high-precision integrated circuit reticles and belongs to the field of integrated circuit computer aided design.

Background technique

When the minimum feature size and spacing of integrated circuits are reduced below the wavelength of the light source used in lithography, due to the inevitable effects of light diffraction and photoresist development and etching, the mask (Mask) pattern and silicon The graphics printed on the wafer will no longer be consistent, and the distortion of IC layout graphics transfer will increase significantly, seriously affecting the production yield of integrated circuits. This phenomenon is called "Optical Proximity Effect (OPE, Optical Proximity) Effects)". Usually, the distortion phenomenon produced by the graphics actually printed on the silicon wafer includes: corner rounding or distortion, line length shortening, line width deviation of sparse and dense lines, line width difference between transparent mask and opaque mask. These distortions can cause as much as 20% deviation of the actual exposure pattern relative to the original layout design pattern, which greatly exceeds the tolerance limit of 10% deviation in industrial lithography. At present, the most advanced lithography technology in the world belongs to this category. Wavelength Lithography". In order to solve various difficulties in the design and manufacture of integrated circuits in the ultra-deep submicron era, and to make the results of lithography best meet the goals of layout design, Resolution Enhancement Technology (RET, Resolution Enhancement Technology) came into being. This technology mainly uses "Optical Proximity Correction (OPC, Optical ProximityCorrection)", "Phase Shift Mask (PSM, Phase Shift Mask)" and "Off-Axis Illumination (OAI, Off Axis Illumination)" and other methods to reduce the optical proximity effect on integration The impact of circuit production yield, and enable existing integrated circuit production equipment to manufacture chips with smaller feature sizes under the same production conditions. The so-called model-based OPC is to correct the lithography results by changing the mask pattern. Its basic method is to establish a simulation model based on the parameters of the lithography equipment and the actual lithography process, and use this model to correct the lithography results. The mask pattern is systematically pre-corrected, so that the degree of nonlinear distortion caused by light diffraction and photoresist exposure, development and etching is reduced. With the continuous reduction of the feature size of integrated circuits, this high-precision Model-based OPC is becoming more and more common in the field of IC manufacturing.

Although the successful application of OPC and some other RET technologies has made the current manufacturing capability of integrated circuits smoothly transition to the deep submicron era, things are not as smooth as imagined. In the practical application of OPC, with the increase of layout data volume and complexity, some problems need to be solved urgently. (1), the hierarchical structure of the layout cannot be fully utilized, and the design flexibility is reduced; (2), the operation time of OPC is too long, even if it is performed in parallel on multiple computers, it will take several days, and the product is designed from The time to launch on the market has also increased, and product competitiveness has weakened; (3), the amount of mask data has grown explosively, easily reaching a size of several GB or even a dozen GB, which makes data transmission very inconvenient and inefficient . With the increasing complexity of the integrated circuit manufacturing process, these problems have a greater impact on yield and manufacturing costs, and the importance of hierarchical processing in the OPC calibration process has become increasingly apparent. This method is also called It is called Cell-based OPC correction technology, which can effectively reduce computing time, reduce data storage volume and improve the flexibility of the entire manufacturing process.

Generally speaking, in order to realize the hierarchical processing method, it is necessary to effectively solve the problem of mutual influence between adjacent graphics in the OPC process. The traditional OPC is carried out for the entire layout. It adopts a strategy of flattening the entire design and then performing graphics processing to ensure the accuracy of the final correction results. This is because when adjacent cells (layout units) are put together, they will affect each other's OPC correction results, and the final OPC result of each graphic will be affected by the graphics within a certain range around it, so because The environment of each Cell instance in the layout is different, resulting in the final OPC results of each Cell instance will be different from each other, so it is difficult to implement hierarchical OPC. Specifically, in the process of OPC processing, if any graphic changes, then due to the influence of its light intensity on the surroundings, the nearby graphics will also change accordingly, and after these nearby graphics change, a little farther away The graphics at the location will change accordingly, which will lead to an effect similar to water waves, which will affect the whole body. Although the degree of this effect will become smaller and smaller as the distance increases, within a certain range, The resulting error cannot be ignored.

This mutual influence between graphics has become the most important obstacle to realize hierarchical OPC. At present, two schemes are mainly adopted in the world to overcome this difficulty. One is to use auxiliary graphics to simulate the environment of the Cell instance in the layout to obtain approximate results. Although this method is feasible in some cases, the accuracy of this method is not satisfactory for the ever-changing layout environment. . Another solution is to use the repeated graphics in the layout. Before doing OPC, find out all the repeated graphics in the layout, and then within a certain range of these graphics, it can be considered that the influence of the external environment is small, or even negligible. In this way, the same OPC correction result can be used for the internal area of repeated graphics to greatly reduce redundant operations. However, although the accuracy of the results of this method is satisfactory, its scope of application is very limited. The area of completely consistent graphics in the layout is not large. If only the area within a certain range of them can save redundant operations, then the performance The improvement is very limited, so at present this method is mainly used in the processing of RAM layout with a large number of repeated graphics.

Contents of the invention

The purpose of the present invention is to propose a hierarchical optical proximity effect correction method, so as to reduce the time for manufacturing high-precision integrated circuit reticles under sub-wavelength lithography conditions, improve the production yield of integrated circuit products and shorten the production cycle.

In order to achieve the above-mentioned purpose, the hierarchical optical proximity effect correction method of the present invention includes the preparation of the pre-correction cell library (Post-OPC Cell Library), the initialization of the mask pattern offset and the fast operation using the dynamic adjustment algorithm, the steps are as follows :

1) Initialization:

Set the simulation model for optical proximity effect correction (Simulation Model),

Photolithographic mask pattern, GDSII input,

The basic parameters of the lithography machine, λ, NA, σ,

where λ is the wavelength of the light source, NA is the numerical aperture of the optical system, and σ is the coherence coefficient of the illumination;

2) Prepare the pre-calibration cell library:

Using the traditional model-based optical proximity effect correction method, each layout unit is pre-corrected, and the pre-corrected graphics are saved back to the cell library. During the pre-correction process, the surrounding environment of each unit is not considered on the correction result. Influence;

3) Graphic segmentation and offset initialization:

First divide each side of the input mask graphic into several segments according to predefined rules, and then initialize the offset of each segment, and replace the corresponding segment offset in the input graphic with the segment offset of the pre-corrected cell library. The offset of the segment, if there is a situation where the two cannot correspond, take the segment with the closest position to set the offset;

4) Dynamically divide three different types of segments in hierarchical OPC operations:

Carry out OPC operation on the initialized mask graph, and iteratively calculate the new offset O ₁ of each edge according to the greedy algorithm, and compare it with the original offset O ₀ after each cycle, according to the set threshold O _t dynamically divides the graph into three different types of segments:

(a) Stable Segment: The division standard is |O ₁ -O ₀ |<O _t

(b) Unstable Segment: The division standard is |O ₁ -O ₀ |≥O _t

(c) Adjacent Segment: Suppose the minimum distance from a certain segment to its surrounding unstable segment is D _m , and the influence range of the unstable segment is D _i , then its division standard is (|O ₁ -O ₀ |< O _t )&&(D _m <D _i );

5) Use dynamic adjustment algorithm for quick correction:

For the unstable segment and propagating segment obtained after the Nth iteration, its new position is recalculated according to the traditional model-based optical proximity effect correction method in the N+1 iteration, and for the For the stable segment, it is no longer necessary to calculate its new position in the N+1 iteration. After each OPC iteration, according to the new offset of each segment, the stable segment and Unstable segment, then according to the position of the unstable segment by the method described in c) in step 4), the propagation segment is redefined, and then enters the next iteration;

6) OPC calibration termination condition

After each OPC iteration, the accuracy of the correction result is calculated according to formula (1),

$\mathrm{<span\; class="notranslate">\; cost</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; EPE</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

$<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; D.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

In the formula, EPE (Edge Placement Error) is the segment position error, x is the sampling point position on each segment, D represents the graphic outline of the design target, W indicates the actual simulated graphic profile, and sums all samples on the input mask graphic click proceed,

If the correction accuracy does not meet the predefined required value Cost ₀ , continue the iteration according to step 5), and terminate the iteration if it meets the requirement, and obtain the corrected mask result.

The present invention adopts a hierarchical correction method, which is mainly used in the iterative process of OPC. After each cycle of OPC correction is completed, the dynamic adjustment algorithm is used to automatically identify and distinguish the affected graphics from the unaffected graphics. , and in the next cycle, only those affected local ranges are dynamically adjusted, and for the unaffected graphics, keep them unchanged, in order to achieve the best results and the fastest speed, and make this method It can be applied to various layout environments without restriction.

The beneficial effects of the present invention are:

The hierarchical optical proximity effect correction of the present invention proposes a brand-new acceleration algorithm, which can effectively improve the speed of optical proximity effect correction, and enable the preparation of high-precision masks that are considered to be expensive due to the large amount of calculations under deep submicron conditions. With the help of the new algorithm, the computational complexity of the process can be greatly reduced. At the same time, this new method provides a cell library with pre-corrected results for the IC design process, enabling design engineers to check the design more flexibly and effectively. The method proposed in the present invention effectively utilizes the hierarchical structure, which can be used to assist in the manufacture of high-precision integrated circuit reticles, improve the predictability of results in the design of integrated circuits, increase the operation speed of OPC correction, reduce costs, and improve integration The production yield of the circuit is improved and the production cycle is shortened.

Description of drawings

Fig. 1 is a flow chart of hierarchical optical proximity effect correction;

Figure 2 is an illustration of offset initialization;

Fig. 3 is the schematic diagram of determining the propagation section by the unstable section;

FIG. 4 is a schematic diagram of the positions of EPE sampling points of each graphic in the layout.

Detailed ways

Further illustrate the present invention below in conjunction with accompanying drawing.

The hierarchical optical proximity effect correction method, the process is shown in Figure 1, including the preparation of the pre-correction cell library (Post-OPC Cell Library), the initialization of the mask pattern offset and the fast calculation using the dynamic adjustment algorithm, the steps are as follows :

1) Initialization:

Set the simulation model for optical proximity effect correction (Simulation Model),

Photolithographic mask pattern, GDSII input,

The basic parameters of the lithography machine, λ, NA, σ,

where λ is the wavelength of the light source, NA is the numerical aperture of the optical system, and σ is the coherence coefficient of the illumination;

2) Prepare the pre-calibration cell library:

Before the formal operation, the traditional model-based optical proximity effect correction method is used to pre-correct each layout cell, and the pre-corrected graphics are saved back to the cell library, and the surroundings of each cell are not considered during the pre-correction process The impact of the environment on its correction results, and data preparation for the acceleration algorithm of hierarchical OPC;

3) Graphic segmentation and offset initialization:

First divide each side of the input mask graphic into several segments according to predefined rules, and then initialize the offset of each segment, specifically, replace the input with the segment offset of the graphic in the pre-corrected cell library The offset of the corresponding segment in the graph. If the two cannot correspond, the segment with the closest position is used to set the offset. Through this step, the initial position of the graph is established for the subsequent fast algorithm.

The above-mentioned graphic segmentation method is determined by the Adaptive Fragmentation rule (adaptive segmentation rule, N.B. Cobb, "Fast Optical and Process Proximity Correction Algorithms for Integrated Circuit Manufacturing", Ph.D. dissertation, University of California at Berkeley, 1998).

As shown in Figure 2, compare the original input graphic CBOD with the OPC pre-corrected graphic C'B'O'A'D' in the unit library, and assign the initial offset value OO to the OB segment according to the comparison result ', assign the offset initial value BB' to the BC segment, assign the offset initial value DD' to the OD segment, and so on, if no matching segment is found, take the offset of the closest segment.

4) Dynamically divide three different types of segments in hierarchical OPC operations:

Carry out OPC operation on the initialized mask graph, and iteratively calculate the new offset O ₁ of each edge according to the greedy algorithm, and compare it with the original offset O ₀ after each cycle, according to the set threshold O _t dynamically divides the graph into three different types of segments:

a) Stable segment (Stable Segment): the segment that has reached the optimal position and is not affected by other graphics; the division standard of the stable segment is: |O ₁ -O ₀ |<O _t

b) Unstable Segment: A segment that needs to be readjusted due to the influence of the surrounding graphics; the division standard of the unstable segment is: |O ₁ -O ₀ |≥O _t

c) Propagation segment (Adjacent Segment): It is not affected in this cycle, but it is within the influence range of the unstable segment, so it may need to be readjusted in subsequent cycles;

Assuming that the minimum distance from a certain segment to its surrounding unstable segment is D _m , and the influence range of the unstable segment is D _i ,

Then the division standard of the propagation segment is: (|O ₁ -O ₀ |<O _t )&&(D _m <D _i );

As shown in Figure 3, EF is an unstable segment obtained from the analysis, and its influence range can be determined according to D _i , as shown in the rectangle ABCD. The stable segment falling into this range needs to be reclassified as a propagation segment.

5) Use dynamic adjustment algorithm for quick correction:

For the unstable segment and propagation segment analyzed after the Nth iteration, in the N+1 iteration, according to the traditional model-based optical proximity effect correction method (N.B.Cobb, "Fast Optical and Process Proximity Correction Algorithms for Integrated Circuit Manufacturing", Ph.D.dissertation, University of California at Berkeley, 1998) recalculate its optimal position, and for the stable segment obtained after the Nth iteration, it is not necessary to calculate its optimal position in the N+1 iteration Excellent location. After each OPC iteration, according to the new offset of each segment, re-divide the stable segment and the unstable segment according to the method described in step 4) a) and b), and then according to the position of the unstable segment according to step 4) c ) The method re-demarcates the propagation segment, and then enters the next iteration.

6) OPC calibration termination condition

After each OPC iteration, the accuracy of the correction result is calculated according to formula (1),

$\mathrm{<span\; class="notranslate">\; cost</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; EPE</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

$<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; D.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

In the formula, EPE (Edge Placement Error) is the segment position error, x is the sampling point position on each segment (as shown in Figure 4), D represents the graphic outline of the design target, W indicates the actual simulated graphic profile, and the summation is the input All sampling points on the mask pattern are performed.

If the correction accuracy does not meet the predefined required value Cost ₀ , continue the iteration according to step 5), and terminate the iteration if it meets the requirement, and obtain the corrected mask result.

Claims (1)

1. the optical proximity correction method of a stratification comprises the preparation of precorrection cell library, the initialization of mask graph side-play amount and utilize dynamic adjustment algorithm to carry out quick computing, and step is as follows:

1) initialization:

Set the realistic model of optical proximity correction,

The mask figure, the GDSII input,

The basic parameter of litho machine, λ, NA, σ,

Wherein, λ is the wavelength of light source, and NA is the numerical aperture of optical system, and σ is the coefficient of coherence of illumination;

2) prepare the precorrection cell library:

Adopt traditional optical proximity correction method, each domain unit is carried out precorrection, the figure after the precorrection is preserved back in the cell library, in the precorrection process, do not consider of the influence of each unit surrounding environment its correction result based on model;

3) figure is cut apart and the side-play amount initialization:

Each the bar limit that to import mask graph earlier is divided into several segments according to predefined rule, again every section side-play amount is carried out initialization, substitute the side-play amount of corresponding section in the tablet pattern with the field offset amount of figure in the precorrection cell library, if run into the situation that the two can't be corresponding, then fetch bit is put immediate section side-play amount is set;

4) dynamically divide three kinds of dissimilar sections in the stratification OPC computing:

Mask graph after the initialization is carried out the OPC computing, and according to the new side-play amount O on every limit of greedy algorithm iterative computation ₁, after each loop ends with former side-play amount O ₀Compare, according to preset threshold O _tFigure is divided into three kinds of dissimilar sections dynamically:

(a) stable section: the criteria for classifying is | O ₁-O ₀|＜O _t

(b) unstable section: the criteria for classifying is | O ₁-O ₀| 〉=O _t

(c) propagation segment: establishing certain section is D to its minor increment of unstable on every side section _m, the coverage of unstable section is D _i, then its criteria for classifying be (| O ₁-O ₀|＜O _t) ﹠amp; ﹠amp; (D _m＜D _i);

5) utilize dynamic adjustment algorithm to proofread and correct fast:

To unstable section and the propagation segment that draws after the N time iteration, in the N+1 time iteration, recomputate its reposition according to traditional optical proximity correction method based on model, to the stable section that draws after the N time iteration, in the N+1 time iteration, need not calculate its reposition again, after each OPC iteration according to the new side-play amount of every section set by step 4) in a), b) described method repartitions stable section and unstable section, and then according to the position of unstable section set by step 4) in c) described method delimit again to propagation segment, enter next iteration afterwards;

6) OPC proofreaies and correct end condition

After each OPC iteration, by formula (1) calculation correction result's degree of accuracy,

$\mathrm{Cost}=\underset{i}{\mathrm{\&Sigma;}}{\left|\mathrm{EPE}\left(x\right)\right|}^2$

$=\underset{i}{\mathrm{\&Sigma;}}{|D\left(x\right)-W\left(x\right)|}^2---\left(1\right)$

EPE is the fragment position error in the formula, and x is the sampling point position on each section, and D represents the graph outline of design object, and W represents the graph outline of actual emulation, and summation is carried out all sampled points on the input mask graph,

If correction accuracy does not satisfy predefined required value Cost ₀, then set by step 5) and continue iteration, if meet the demands then termination of iterations, the mask result after obtaining proofreading and correct.

Images