JP2002182366A - Pattern forming method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To correct the dependence on the crude density of line widths depending upon the distances between the patterns of fine processing patterns using a phase shift. SOLUTION: One layer is formed by plural times of exposure including high- resolution exposure using plural fine patterns (gate finger parts) and plural phase shift patterns (shifters) which are respectively arranged on both sides in the fine line width direction of the fine patterns and negate the interference of light by the phase difference of the light passing the same and ordinary exposure exclusive of the fine pattern points. In forming the mask patterns to be used for the ordinary exposure, the sizes of the light shielding patterns (gate electrodes) on a mask for the ordinary exposure superposed on the positions corresponding to the fine patterns on the mask for the high-resolution exposure are changed in the direction of reducing the line width difference after the resolution which occurs according to the crude density of the fine patterns. If, for example, the gates of the dense patterns are connected, the dependence of the line widths on the crude density is corrected.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern generation method suitable for forming a fine pattern on a wafer substrate from a transmitted light of a photomask using an exposure apparatus in a process of manufacturing a highly integrated semiconductor device. About the method.

[0002]

2. Description of the Related Art A photomask used in a manufacturing process of a semiconductor device has a structure in which a light-shielding film is formed on a glass substrate.
The light-shielding pattern on the photomask is projected and exposed on the photoresist applied to the wafer surface. A light-shielding pattern on a photomask is obtained by converting designed CAD data into data for a drawing apparatus and patterning the data faithfully, and transferring the photomask pattern onto a wafer with high accuracy is performed by a semiconductor photomask. This is a lithography process.

In a photolithography process in a semiconductor device manufacturing process, a high resolution exceeding a resolution limit determined by a wavelength of light is required because a pattern near an exposure wavelength is required to be formed. Therefore, in recent years, a phase shift method has been used as a lithography technique capable of forming a fine pattern having a wavelength equal to or shorter than the exposure wavelength.

[0004] The principle of the spatial frequency modulation type phase shift method will be described below. FIG. 16 is a diagram showing the principle of the phase shift method in comparison with the conventional method, and FIG. 17 is a diagram showing Fourier spectra of the phase shift method and the conventional method. Here, it is assumed that the period of the light intensity transmittance takes a value near the resolution limit in a one-dimensional periodic pattern of 1 / ν ₀ . Since the light transmission target is a mask pattern close to the resolution limit, the amplitude of the light transmitted through the photomask can be approximated as follows when only the sinusoidal fundamental frequency components are considered.

[0005]

[Formula 1] Conventional mask: T (x) = | cos2πν ₀ x | (1-1) Phase shift mask: T (x) = cos2πν ₀ x (1-2)

In the conventional method shown in FIG. 16A, light transmitted through a photomask forms an angle θ (sin θ = ν ₀ λ) with the _0th -order diffracted light that travels straight along the optical axis. The light is separated into ± first-order diffracted light and enters the projection lens.
On the other hand, as shown in FIG.
The light transmitted through the phase shift mask is separated into ± first-order diffracted light at an angle of θ / 2 with respect to the optical axis, and is incident on the projection lens. In each case, only the diffracted light transmitted through the inside of the projection lens contributes to image formation. The Fourier spectrum formed on the pupil plane of the projection lens is expressed as follows from the Fourier transforms of Expressions (1-1) and (1-2).

[0007]

## EQU2 ## Conventional mask: F (ν) = (4 / π) ｛δ (ν) / 2 + [δ (ν + ν ₀ ) + δ (ν-ν ₀ )] / 2 + ‥‥｝ (2-1) phase shift mask: F (ν) = (1/2 ) _{[δ (ν + ν 0/2} ) + δ (ν-ν 0/2) ] ... (2-2)

As shown in FIG. 17, in a conventional transmission type mask that does not consider the phase, spectral components exist at ν = 0 and ± ν ₀ , whereas in a phase shift mask, ν = ±
It exists only in the ν _0/2. That is, the basic spectrum obtained from the phase shift mask is at half the position of the spectrum obtained from the transmission mask. This is equivalent to halving the spatial frequency of the mask pattern. The stepper projection lens acts as a low-pass filter that transmits only spatial frequency components smaller than the natural frequency νc (= NA / λ). Spatial frequency ν of the fine pattern to be transferred
Considering the case where ₀ is νc <ν ₀ ≦ 2νc, in the case of a transmission type mask, the spectral component of ± ν ₀ does not pass, so that image contrast cannot be obtained. On the other hand, to form a pattern image on the image plane because the phase shift mask transmits a basic spectrum ν = ± ν _0/2. This is a feature of the spatial frequency modulation type phase shift mask that has the greatest effect of improving the resolution.

The basic layout structure of the spatial frequency modulation type phase shift mask is, as shown in FIG.
The shifters are arranged so that the openings of the mask alternately have opposite phases, using two phases of 180 degrees and 180 degrees. As shown in FIG. 16B, there are a shifter in which the shifter is laminated with a film different from the light-shielding film, and a shifter in which the glass surface is etched to have the function of the shifter. This shifter arrangement corresponds to a two-color problem in which a two-dimensional map is painted separately, and shifter phases must be alternately arranged. Therefore, in a complicated wiring layout, inconsistency in arrangement (phase inconsistency) in which the shifter phase cannot be reversed alternately cannot be avoided in principle.

As the first prior art, Oi et al.
of Design of Phase-Shifting Masks Utilizing a Com
pactor, JJAP Vol.33 (1994) No.12B, pp 6774-6778], the phase inconsistency using layout compression (compression) in the phase shift method using the phase difference of transmitted light between patterns to be formed. We discuss methods to avoid.

FIG. 18 is a diagram showing the principle of a so-called Levenson-type phase shift using a negative resist premised in the first prior art. In the photomask shown in FIG. 18, the chromium opening provided at the pattern formation location has
180 degree phase shifter (glass etching groove)
The grooves are alternately arranged with no phase grooves. Therefore, in this phase shift method, a phase difference of light transmitted through the pattern itself is used, and a negative resist in which a pattern transfer portion remains without being dissolved in a developing solution is essential. Practical negative resists are difficult to obtain, and the pattern layout itself must be a pattern that produces a phase difference. Therefore, it is effective for simple lines and spaces, but is not effective for isolated patterns. Therefore, the first conventional technique has a drawback that, even if phase inconsistency can be avoided, there are many restrictions on the resist material and pattern formation on which it is based.

In recent years, studies have been made on a plurality of types of phase shift exposure using a mask in which light passing through both sides of a fine pattern to be formed is provided with a phase difference, and normal exposure for removing unnecessary patterns caused thereby. This is a technique based on a single exposure (hereinafter, a phase shift multiplex exposure method). This second prior art is described in "Automatic Generati
on of Phase-Shifting mask Patterns Using Shifter-E
dge Lines, MICRO-AND-NANO-ENGINEERING'97. FIG. 19 is a view showing the principle of a Levenson-type phase shift using a positive resist as the phase shift multiplex exposure method. The phase shift mask (FIG. 19A) shown in FIG. 19 has a fine pattern (chrome pattern).
Are alternately arranged with a 180-degree phase shifter (glass etching groove) and without a 0-degree phase groove. Since a positive resist is used, the periphery is covered with chrome. Then, in the second exposure (normal exposure), the surrounding wiring (not shown) is resolved while shielding the fine pattern using the photomask shown in FIG.

[0013]

In the second prior art based on the phase shift multiplex exposure method, in a fine pattern formed by a phase shift pattern, the density of the formed pattern depends on the distance between the patterns. May cause width variation. In order to reduce this, it has been conventionally considered to change the line width of the fine pattern according to the distance between the patterns, but in a fine pattern in which the control in the photomask manufacturing process is at a limit, the The correction result may be out of the line width limit in the photomask manufacturing process, and manufacturing may not be possible.

The present invention has been made in view of the above-mentioned circumstances, and an object of the present invention is to design a phase shift mask in which a line width depending on a distance between patterns of a fine processing pattern using a phase shift. An object of the present invention is to provide a method of generating a mask pattern for multiple exposure, which corrects the density dependency.

[0015]

In order to achieve the above object, a pattern generating method according to the present invention comprises a plurality of fine patterns, and a plurality of fine patterns, which are arranged on both sides in the fine line width direction of the fine patterns, respectively. This is a method for generating a pattern to be used when one layer is formed by a plurality of exposures including a high resolution exposure using a plurality of phase shift patterns for canceling light interference by a phase difference and a normal exposure except for a fine pattern portion. In generating the mask pattern used for the normal exposure, the size of the light-shielding pattern on the normal exposure mask superimposed on a position corresponding to the fine pattern on the high-resolution exposure mask is determined by the density of the fine pattern. Is changed in a direction to reduce the line width difference after the resolution, which is generated according to.

While the reduction of the line width difference depending on the fineness of the fine pattern has been limited by the correction using only the conventional phase shift mask, this pattern generation method has a limitation of two.
In the subsequent normal exposure, the line width difference after resolution caused by the density of the fine pattern can be reduced or corrected. In general, the fine pattern of the phase shift mask has already been miniaturized to the limit of photomask production.
There is often room for size change in the normal exposure mask after the first time. In this method, the line width difference between fine patterns can be reduced or corrected while avoiding restrictions from the photomask manufacturing limit.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention is a method for forming a fine pattern below the resolution limit using a spatial frequency modulation type phase shift method, and the type of the fine pattern is not limited. A typical pattern desired to be below the image limit is a gate pattern of a MOS transistor.

Hereinafter, after describing the MOS transistor structure, an embodiment of a pattern generation method according to the present invention will be described in detail with reference to the drawings, taking as an example the ultra-high resolution exposure of the gate pattern.

FIG. 1 is a sectional structural view of a general MOS integrated circuit. In FIG. 1, reference numeral 100 denotes a MOS.
An integrated circuit; 101, a semiconductor substrate such as a p-type silicon wafer; 102, a source / drain impurity region into which n-type impurities or the like are introduced at a high concentration; 104, p-type impurities or the like to prevent formation of a parasitic transistor; The channel stopper introduced to the concentration, 106 is LOC as an element isolation layer
OS and 108 indicate a gate oxide film, and 110 indicates a gate electrode made of n-type polysilicon or the like. In such a MOS integrated circuit, the highest line width controllability is required when the gate electrode 110 is processed. That is, the gate pattern width of the transistor (generally, the gate length Lg) is
The transistor characteristics (the driving capability such as the gate threshold voltage Vth and the mutual conductance gm) are determined, and the line width variation directly affects the characteristic variation. Therefore, the line width control of the gate electrode pattern is the most important in MOS transistor formation.

FIG. 2 is a block diagram showing a schematic configuration example of the pattern generation device. The pattern generation device 1 of the illustrated example includes an input unit 2, a phase shift pattern generation unit 4, a normal mask pattern generation unit 6, a pattern change unit 8, and an output unit 10.
And the control unit 12. The phase shift pattern generation unit 4 further includes a graphic operation unit 14, a graphic processing unit 1,
6. It has a pattern search means 18, a phase setting means 20, and a phase difference confirmation means 22. Among them, the figure calculation means 14 and the figure processing means 16 correspond to “arrangement determination means”,
Further, the pattern search means 18, the phase setting means 20, and the phase difference confirmation means 22 correspond to "phase determination means".
The graphic calculation means 14 corresponds to "pattern extraction means" or "pattern synthesis means", and the graphic processing means 16 corresponds to "unit pattern generation means".

The functions of these components will be described in detail in the following description of the procedure.
And 16 determine the shapes of the plurality of phase shift patterns respectively arranged on both sides in the fine width direction of the plurality of fine patterns based on the positional relationship between the plurality of fine patterns. Also, for example, three means 18,
The phase determining means constituted by 20 and 22 determines the phase of the plurality of phase shift patterns based on the positional relationship of the plurality of fine patterns, and the phase difference between the pair of phase shift patterns close to each other with the fine patterns interposed therebetween. Is set to 180 degrees. In this example, the “fine pattern” corresponds to FIG.
Of the light-shielding pattern which forms the gate electrode 110 of FIG. 19, and the openings of the light-shielding pattern arranged with a phase difference of 180 degrees on both sides in the gate length direction of the light-shielding pattern (180-degree phase opening and 0-degree phase opening in FIG. 19) Is the “phase shift pattern”
Corresponds to.

Next, a first method of generating a phase shift pattern will be described with reference to the flowchart of FIG. This generation method can be implemented by, for example, a pattern generation device 1 as shown in FIG. Here, the control unit 1 in the pattern generation device 1
Under the control of 2, each means executes the processing.
In step 1 of FIG. 3, a unit pattern having a width sufficient for the diffracted lights transmitted on both sides of the gate electrode pattern to cancel each other out and serving as an element constituting the desired phase shift pattern is generated.

FIG. 4 is a flowchart showing an example of a procedure for generating the unit pattern. FIG. 5 shows the MOS of this example.
FIG. 7 is a plan view showing a process of generating a unit pattern in the integrated circuit. In the method illustrated in FIG. 4, first, FIG.
After the input unit 2 inputs CAD data, step ST
At 11, a fine pattern is extracted from the CAD data. This fine pattern is a reference for determining the position and length of the unit pattern generation (the size orthogonal to the fine line width to be resolved). The extraction of the fine pattern is performed by CAD from the input unit 2.
This is executed by the graphic calculation means 14 of FIG. 2 based on the data. More specifically, in the MOS integrated circuit of this example, a portion having a fine line width that determines the transistor characteristics is an overlapping portion between the gate polysilicon layer and an element region other than the LOCOS. Thus, the gate pattern Pa and the element region pattern Pb are read. Then, as shown in FIG. 5B, the element shape patterns read from the CAD data are superimposed on each other, and a graphic product process (AND process) is performed to extract the fine gate portion Pc (FIG. 5).
(C)).

Next, in step ST12 of FIG. 4, the extracted fine gate portion Pc is resized twice. This resizing is performed by the graphic processing means 16 in FIG. More specifically, as shown in FIG. 5D, the gate width Wg of the fine gate portion Pc is increased by a predetermined length on both sides in the gate width direction in order to secure an alignment margin during multiple exposure. As shown in FIG. 5E, the fine gate portion Pd after the enlargement in the gate width direction is resized on both sides in the gate length direction. As a result, a final fine gate pattern Pe with the original gate length Lg0 reduced to Lg 'is obtained. The gate length Lg ′ after this resizing is one of the parameters that most affect the line width for resolving the gate pattern.

Then, in step ST13 of FIG. 4, a unit pattern forming a phase shift pattern is generated based on the resized fine gate pattern Pe.
This pattern generation is executed by the graphic processing means 16 in FIG. Specifically, as shown in FIG. 5 (f), on both sides of the fine gate pattern Pe in the fine line width direction,
It has a width not less than the predetermined width L sufficient for the phase shift and has a width of 2 along the entire longitudinal side (length Wg ′) of the fine gate pattern Pe.
One unit pattern Pf is generated. The shape of the unit pattern Pf is arbitrary as long as the above requirements are satisfied.
The unit pattern Pf is generated in a rectangular shape shown in FIG. 4F, and the unit pattern generation process shown in FIG. 4 is completed.

Next, the processing procedure returns to the flow of FIG. 3, and in step ST2, merging by OR processing is performed. This process is executed by the graphic calculation means 14 in FIG. 2, and a desired phase shift pattern is obtained. FIG. 6 shows an example of the obtained phase shift pattern. In the illustrated example, the phase shift patterns P1, P5
And P6 are unit patterns Pf11 and Pf42, respectively.
Alternatively, Pf51 has a single configuration, but the phase shift patterns P2 to P4 are configured from the graphic sum of a plurality of unit patterns. The phase shift pattern P2 is
One unit pattern Pf1 of the fine gate pattern Pe1
2 and the pattern sum of the other unit pattern Pf21 of the fine gate pattern Pe2. Similarly, the phase shift pattern P4 is composed of one unit pattern Pf32 of the fine gate pattern Pe3 and the other unit of the fine gate pattern Pe4. It is composed of the figure sum with the pattern Pf41. Further, the phase shift pattern P4 is calculated from the figure sum of three unit patterns, that is, one unit pattern Pf22 of the fine gate pattern Pe2, the other unit pattern Pf31 of the fine gate pattern Pe3, and one unit pattern Pf52 of the fine gate pattern Pe5. It is configured. In the formation of the phase shift pattern formed by the sum of these figures, the width of the unit pattern is adjusted so that a minute gap which is a problem in the photomask manufacturing process is not formed between the unit patterns.

Subsequently, based on the information sent from the input unit 2 in FIG. 2, for example, the pattern search means 18 determines the initial search position and its phase (step ST3), and in the next step ST4 , Search for an adjacent pattern having the minimum line width. For example, in the example of FIG.
Focusing on No. 3, first, this is set as the first search pattern, and it is determined whether the phase is 0 degree or 180 degrees. Then, the pattern search means 18 detects L1 to L3 within a predetermined fine line width close to the pattern of interest P3, and detects the adjacent patterns P2, P4 close within the fine line width.
And P6 are detected.

In the next step ST5, the phase of each of the detected phase shift patterns P2, P4 and P6 is set to the phase difference π from the pattern of interest. This phase setting is executed by the phase setting means 20 of FIG. For example, a preset pattern of interest P
Assuming that the phase of No. 3 is 0 degree, the adjacent patterns P2 and P4
And P6 are all 180 degrees in phase. In the next step ST6, for example, the pattern search means 18 in FIG. 2 uses the proximity pattern for which the phase has been determined as a new pattern of interest, and checks whether or not there is another proximity pattern in the new pattern of interest. If there is a proximity pattern, the flow returns to step ST4, and the detection of the proximity pattern (step ST4), the phase determination (step ST5), and the presence / absence confirmation of the proximity pattern (step ST6) are repeated for the new pattern of interest. In the example of FIG. 6, the remaining patterns P1 and P5 are also searched by these series of processes, and each is set to the 0-degree phase, and two types of phase distribution as shown in FIG. 7 are determined. When it is determined that there is no proximity pattern, the flow proceeds to the next step ST7.

In step ST7, for example, it is checked by the phase check means 22 in FIG. 2 whether or not there is a phase undetermined pattern. In the example of FIG. 6, it is determined here that there is no undecided pattern, and in a succeeding step ST8, it is checked whether or not the phase difference is 180 degrees different on both sides of the fine line width of each fine gate pattern. “OK”), the phase shift pattern and its phase distribution information are output from the output unit 10 in FIG. 2 and sent to the next EB data converter. In step ST7, when there is a phase undecided pattern, the processing flow returns to before step ST3, and the processing is reset from the resetting of the search position. As a result, the phase determination is repeated in the above-described procedure until there is no longer a phase undetermined pattern. On the other hand, if it is determined in the next step ST8 that there is a phase inconsistency ("NOT OK"), the process proceeds to step S8.
Proceed to T9.

In step ST9, the phase shift pattern is changed to eliminate the phase inconsistency. This processing is executed by the pattern changing unit 8 in FIG. Specifically, first, the arrangement area size of the two phase shift patterns at the phase inconsistency is checked, and it is determined whether or not at least one of them has a space margin for adding (dividing into two) the phase shift pattern. . That is, the width of the two phase shift patterns in the gate length direction is checked, and it is checked whether the width is at least twice the minimum width L required for obtaining the phase shift effect. In the case of two times or more, the phase inconsistency is resolved by adding a phase shift pattern having a phase difference of π to that location. Although not shown in the flow chart of FIG. 3 in particular, a large-size portion where a pattern can be added is detected in the chain search path where the phase inconsistency has occurred, and the phase inconsistency is resolved there.
The phase setting can be performed again by the above procedure.

Even if such a phase inconsistency remedy is taken,
If the phase inconsistency cannot be resolved, in the present embodiment, if the phase inconsistency, or another location having less influence on transistor characteristics and the like exists in the same chain search path as the phase inconsistency, at that location, Correction (layout correction) between the fine gate patterns is performed to enable close arrangement of a phase shift pattern pair having a phase difference of π. The layout correction may be a manual correction or an automatic correction using a layout compactor or the like. In the case of using a layout compactor, after adding restrictions at the above-mentioned correction location, for example, the constraint of the correction location is reflected in the design rule of the compactor, and two phase shifts having a phase difference of 180 degrees are performed at the correction location. Compress the entire layout so that pattern placement is possible. At this time, for example, restrictions are set in advance on the corrected portion so as not to apply compression, so that design rules such as a surrounding pattern and a margin for matching are reduced, and the size of a necessary phase shift pattern is also reduced. Since the size of the phase shifter with this restriction is not reduced,
Space is required for adding a pattern. In the above description, if the location where the phase contradiction occurs is not a fine pattern, the phase contradiction may be left as it is because it is erased by the second exposure. Naturally, the phase inconsistency can be eliminated in the process of expanding the entire layout in the same manner.

After that, if the phase inconsistency can be resolved by adding a phase shift pattern, the result (the phase shift pattern after the change and its phase distribution information) is output from the output unit 10 in FIG. The pattern generation ends.

When a phase shift pattern is added and the gap portion is resolved, the light-shielding pattern in the second normal exposure of the multiple exposure is corrected if necessary. Note that the normal mask pattern generation unit 6 in FIG.
A protection pattern that is superimposed on the fine gate with a size slightly larger in consideration of the alignment margin is also generated.
Since the detailed contents will be described later, the description is omitted here.

In the first generation method of the phase shift pattern, the shape and phase of the phase shift pattern are not determined at the cell design stage, but the shape and phase of the phase shift pattern are determined based on the position and shape of the fine gate pattern. Are determined sequentially and sequentially. For this reason, once the shape and arrangement of the gate pattern are determined, the phase shift pattern and its phase can be automatically determined, and the pattern design of the phase shift pattern can be performed at high speed. In particular, since the phase determination is performed by a chain search method, feedback of this correction (phase change) is easy. For example, when a phase inconsistency occurs, a pattern is added at a location where there is room for resolving the phase inconsistency, or when a layout change is required to increase the gate pattern layout interval. The subsequent phase difference confirmation and phase reset are smooth. Even if these methods cannot resolve the phase inconsistency, the phase inconsistency can always be finally resolved by layout compression or expansion. Further, since the shape and phase of the phase shift pattern are obtained based on the positional relationship of the fine pattern, the phase shift pattern can be closely arranged with a phase difference of 180 degrees in any direction in a two-dimensional plane. Therefore, the fine patterns need not be parallel, and there are fewer restrictions than in the past. As described above, this method of generating a phase shift pattern has the advantage that the restrictions on the pattern generation are smaller and the efficiency is higher than before.

FIG. 8 is a block diagram showing another configuration of the pattern generation device, and FIG. 9 is a flowchart showing a second generation method of the phase shift pattern.

Compared to the case shown in FIG. 2, the pattern generating device 30 is different from the pattern data searching device 18 and the phase difference checking device 22 in that the graph data generating device 3
2. It has a graph loop confirming means 34 and a graph data changing means 36. Here, a storage means 40 is added. The functions of these means 32, 36, 38 and 40 will be described later. Other configurations are the same as those described above. However, the flow of the pattern data between the pattern changing unit 8, the phase setting unit 20, and the output unit 10 is slightly different from the first method.

In this pattern generation method, steps up to step ST2 in FIG. 9 are the same as those in FIG. 3 described above, and the flow chart in FIG. 4 is applied as it is. Step S in FIG.
At T3, graph data is created by the graph data generating means 32 of FIG.

FIG. 10 is a diagram showing elements of graph data and an example of a data configuration. Graph data is composed of two elements called nodes and edges. The node is information indicating each formed phase shift pattern, and in the example of FIG. 10A, the phase shift patterns P1, P2,
,..., P5 correspond to nodes,. The edge is information indicating a positional relationship between the respective phase shift patterns, for example, information indicating that a distance in the gate length direction is a fine line width. In the example of FIG.
Phase shift patterns P1 and P2, P2 and P3, P3
Edges E12, E23, E34, and E35 indicate that the distance between P4 and P3 and the distance between P3 and P5 are fine line widths, respectively. These information are stored in the storage unit 4 shown in FIG.
0, but the specific storage structure (database) includes, for example, a list of nodes (phase shift pattern identification information), a list of edges (information of a pair of adjacent phase shift patterns with a fine line width), There are three types of databases: a pattern edge list for each node (information on patterns close to each phase shift pattern). FIG.
0 (c) is a diagram representing a pattern edge list for each node as an example of a database. When the CAD design is performed on a cell basis, these three types of databases are provided for each cell type.

In the generation of such graph data, when determining an edge, edge detection can be achieved by sequentially examining the relationship between adjacent patterns by a method similar to the pattern search in FIG. FIG. 11A shows two examples of a graph data (graph graphic) of the graph data expressed by the nodes and the edges, and two corresponding patterns. As described above, when the graph data is used, a plurality of patterns located at a distance apart by the minimum line width of the fine gate pattern can be collectively formed into a graphic.

In the next step ST4, for example, the graph data generation means 32 determines whether or not the generated graph graphic is a closed loop. FIG. 11 (a)
If it is not a closed loop as in the example of FIG. 7, the flow proceeds to step ST5, and each phase is determined by the phase setting means 20 in the same manner as in FIG. 3 (FIG. 11B), and its phase shift pattern and phase Information is output from the output unit 10.

On the other hand, if the generated graph graphic is a closed loop, the flow proceeds to step ST6, and it is checked whether the number of closed loop nodes is an even number (even number loop) or an odd number (odd number loop). This loop confirmation is executed by the graph loop confirmation means 34 of FIG. FIG.
12A illustrates the case of an even loop, and FIG. 12B illustrates the case of an odd loop. In an even-numbered loop as shown in FIG. 12A, phase inconsistency cannot occur in principle, so each phase is set as in the first embodiment (step ST5), and the result is output. On the other hand, in the odd loop shown in FIG. 12B, it is apparent that a phase inconsistency occurs.
7, and the graph data is changed by the graph data changing means 36 in FIG. Then, the next step ST8
After the pattern change similar to that in FIG. 3, that is, the layout change or the layout compression (expansion is possible) of the gate pattern is performed, the result is output through the phase setting (step ST5).

Note that the graph change portion in the previous step ST7 is basically arbitrary, but if this is done, the phase inconsistency can not always be solved successfully by a single layout change or compression (or expansion). Therefore, as a modification, if the width of each pattern in the gate length direction is included in the node of the previous graph data in addition to the identification information of each phase shift pattern, the width is large enough to add a phase shift pattern. It is desirable because it is easy to identify whether or not the graph is suitable for changing the graph. Thus, when changing the pattern,
With the simple addition of the phase shift pattern, the phase can be immediately determined without phase inconsistency. Also, even when a pattern change is necessary, based on the width information in the gate length direction of each pattern,
The movement amount of the gate pattern or the layout compression ratio (or expansion ratio) can be assumed in advance, and the pattern can be changed only once.

FIG. 12 (c) shows an example of a changed graph figure and pattern in which the phase contradiction is eliminated by any of the above methods. The change of the pattern of the second normal exposure is performed in the same manner as the above case.

The method of generating the second phase shift pattern has an advantage that the phase can be determined on a graph in addition to the same effect as the first method. Further, once the graph data is created, a chain search of the pattern is unnecessary, and the number of repetitions of the phase difference confirmation and the phase setting can be reduced. In other words, in the first method, it is necessary to actually set the phase in the pattern search, change the pattern when a phase contradiction occurs, confirm the phase difference, determine the phase again, and repeat this. According to the second method, it is possible to easily know whether or not there is a phase inconsistency by simply determining whether the related graph data group (graph graphic) forms a closed loop or, in the case of a closed loop, whether the number of nodes is even or odd. In addition, there is an advantage that the efficiency can be further improved because a pattern change or the like can be dealt with immediately. This graph data is simple because it requires minimal information for determining the phase. By storing it for each pattern and updating it every time a pattern is changed, the time and effort required to create and change the graph data is reduced. It is much smaller than usefulness (efficiency of pattern generation).

Next, a description will be given of the reduction in line width variation due to the density of the fine pattern in the second normal exposure. This method may be combined with any of the above first and second phase shift pattern generation methods. FIG.
FIGS. 13A and 13B are diagrams for explaining line width variations due to the density of the fine pattern. FIG. 13A is a plan view of the pattern having the density of the fine pattern, and FIG. 13B is a phase shift pattern of the second embodiment. 9 is a graph showing a light intensity simulation result of gate formation by a generation method. FIG. 13 (a)
Are the element region pattern, the gate electrode pattern and 18
Each pattern of the 0-degree phase shifter is superposed, and a gate pattern with a large space (sparse pattern) in the element region 1 and a gate pattern with a small space (dense pattern) in the element region 2 are formed. The horizontal axis of the graph in FIG. 13B is the focus depth, and the vertical axis is the gate fine line width of the light transfer image.

From the simulation result of FIG. 13B, it is found that the gate line width after resolution is large between the sparse pattern and the dense pattern (> 20 n) in a wide focus depth range.
m) It can be seen that they are different. As a result, the fine line width of the gate tends to be thick in a sparse pattern, and to correct the line width difference depending on the sparseness of the gate interval, the gate line width on the current photomask on the sparse pattern side should be corrected. You need to make it thin. However, the fine line width on the photomask of the fine pattern formed by the phase shift pattern is already a narrow mask line width to the limit of the photomask manufacturing process. It is not preferable in the narrowing direction.

Therefore, by changing the line width of the gate light-shielding pattern on the normal exposure mask which is superimposed on the gate resolution pattern (resist latent image) by the phase shift exposure in the normal exposure, the unevenness of the gate line width is reduced. to correct. The light-shielding pattern on the normal exposure mask is usually
It is formed slightly thicker in the fine line width direction in consideration of misalignment of the mask with the gate resolution pattern. For this reason, the light shielding pattern has a design rule that is not so strict as compared with the exposure wavelength. Therefore, the line width of the light-shielding pattern can be changed within a range that does not cause a problem in the photomask manufacturing process.

FIG. 14 shows a pattern plan view and a light intensity simulation result obtained by actually changing the line width of the light-shielding pattern and correcting the line width difference depending on the density of the gate intervals. The line width difference depending on the fineness of the arrangement interval of the fine gate pattern can be corrected in the direction of increasing the width of the light-shielding pattern used for the second normal exposure, and in the example of FIG.
The light-shielding patterns in the dense pattern were all connected in the width direction. From the light intensity simulation results using such a changed light-shielding pattern, it is clear that the line width of the fine pattern does not depend on the pattern interval, and the same light-shielding pattern is generated regardless of the conventional pattern density. It has been found that the line width difference depending on the density of the gate interval can be corrected to a satisfactory level. In addition, since the mask pattern and the minimum line width do not need to be changed, it is not necessary to make the line width standard in photomask production strict.

Finally, a method of generating a light-shielding pattern used in the second and subsequent normal exposures will be described in the case where fine gate patterns are sparse and dense. The generation of the light-shielding pattern for normal exposure is basically performed by the normal mask pattern generation unit 6 in FIG. 2 or FIG. 8 unless otherwise specified. FIG.
5 is a pattern plan view in each generation process of the light-shielding pattern for normal exposure.

First, in FIG. 15 (a), a figure product process (AND process) of the gate electrode pattern and the element region pattern is performed to extract a fine gate pattern in the same manner as in the generation of the phase shift pattern. I do. This pattern extraction is executed by having the same function as that of the graphic operation unit 14 in FIGS. 2 and 8 in the normal pattern generation unit 6 or by using the graphic operation unit 14 in common. Next, in FIG. 15B, the extracted pattern is subjected to predetermined resizing on both sides in the fine width direction (gate length direction). This resize is
This is performed by having the same function as that of the graphic processing unit 16 in FIGS. 2 and 8 in the normal pattern generation unit 6 or by using the graphic processing unit 16 in common. Then, in FIG.
By the same means as in the AND processing of FIG. 15A, a graphic product processing (OR processing) of the resized fine pattern and the gate electrode pattern is performed to generate a light shielding pattern.

Normally, the generation of the light-shielding pattern is completed up to this point. In FIG. 15D, the light-shielding pattern is partially widened. This process is performed, for example, through a process of specifying a dense pattern by pattern search and a process of widening the dense pattern portion. The specification of the dense pattern is executed by having the same function as that of the pattern search unit 18 in FIG. 2 in the normal pattern generation unit 6 or by using the pattern search unit 18 together. The following is detected as a dense pattern, and the predetermined width or more is detected as a sparse pattern. Then, the width of the dense pattern portion is expanded by a predetermined amount inside the space between the patterns of the dense pattern portion detected by such a method. In this example, the width is increased until the interval between adjacent dense patterns is eliminated. As a result, FIG.
The changed light-shielding pattern shown in (e) is obtained. FIG. 15F shows the changed light shielding pattern in relation to the phase shifter.

In the above method, FIGS. 13B and 14
As is clear from the comparison of (b), by changing the size of the light-shielding pattern for normal exposure in the space between the fine gate patterns in the pattern separation direction, the solution generated according to the density of the phase shift pattern in the fine line width direction. The fine line width difference after image can be reduced or corrected. This pattern correction is performed in the second line where the line width standard is not as strict as that of the phase shift mask.
Since it is performed using the normal exposure pattern after the first time, there is an advantage that there is a pattern correction margin equal to or larger than the fine line width difference, and the fine line width difference is easily corrected.

[0053]

According to the pattern generation method of the present invention, the resolution of the fine pattern of the high resolution exposure in accordance with the fineness of the fine pattern in the fine line width direction is corrected by correcting the light-shielding pattern for normal exposure. The width difference can be reduced or corrected without being restricted by the resolution limit.

[Brief description of the drawings]

FIG. 1 is a sectional structural view of a general MOS integrated circuit.

FIG. 2 is a block diagram illustrating a schematic configuration example of a pattern generation device according to an embodiment of the present invention.

FIG. 3 is a flowchart of a pattern generation method according to the embodiment of the present invention.

FIG. 4 is a flowchart illustrating an example of a procedure of a unit pattern generation step in FIG. 3;

FIG. 5 is a pattern diagram showing a process of generating a unit pattern using a MOS integrated circuit as an example.

FIG. 6 is a pattern diagram showing an example of an obtained phase shift pattern.

FIG. 7 is a pattern diagram showing an example of phase setting for the pattern of FIG. 6;

FIG. 8 is a block diagram showing another configuration of the pattern generation device according to the embodiment of the present invention.

FIG. 9 is a flowchart showing another method for generating a phase shift pattern according to the embodiment of the present invention.

FIG. 10 is a diagram showing elements of graph data and a data configuration example.

FIG. 11 is a diagram of a pattern and a graph showing two examples when there is no closed loop.

FIG. 12 is a diagram of patterns and graphs illustrating the case where the closed loop is an even loop and the case of an odd loop.

FIG. 13 is a diagram showing an example of a pattern with density and a simulation result of light intensity in order to explain line width variation due to density of a fine pattern.

FIG. 14 is a diagram showing a pattern and a light intensity simulation result when the line width of the light-shielding pattern is actually changed to correct the line width difference depending on the density of gate intervals.

FIG. 15 is a pattern plan view in each generation process of a light-shielding pattern for normal exposure.

FIG. 16 is a diagram showing the principle of the phase shift method in comparison with the conventional method.

FIG. 17 is a diagram showing Fourier spectra of transmitted light in the phase shift method and the conventional method.

FIG. 18 is a diagram illustrating the principle of a so-called Levenson-type phase shift using a negative resist premised in the first related art.

FIG. 19 is a diagram illustrating the principle of a Levenson-type phase shift using a negative resist as a phase shift multiplex exposure method according to a second conventional technique.

[Explanation of symbols]

1, 30: pattern generation device, 2: input unit, 4: phase shift pattern generation unit, 6: normal mask pattern generation unit,
8: pattern change unit, 10: output unit, 12: control unit, 1
4 ... Graphic calculation means, 16 ... Graphic processing means, 18 ... Pattern search means, 20 ... Phase setting means, 22 ... Phase difference confirmation means, 32 ... Graph data generation means, 34 ... Graph loop confirmation means, 36 ... Graph data change means.

Claims (1)

[Claims] A high-resolution exposure using a plurality of fine patterns and a plurality of phase shift patterns arranged on both sides of the fine patterns in the fine line width direction and canceling out light interference by a phase difference of transmitted light, A method for generating a pattern used when forming one layer by multiple exposures including normal exposure other than the fine pattern portion, and when generating a mask pattern used for the normal exposure, the method on the high resolution exposure mask A pattern generation method for changing the size of a light-shielding pattern on a mask for normal exposure superimposed on a position corresponding to a fine pattern in a direction to reduce a line width difference after resolution generated according to the density of the fine pattern.