JP5323966B2 - Mask blank substrate, mask blank, photomask, and semiconductor device manufacturing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a mask blank substrate which does not need to simulate flatness of a mask substrate after chucking the mask substrate on a mask stage of an exposure apparatus and which achieves desired flatness after chucking regardless of chuck structure of the exposure apparatus. <P>SOLUTION: A mask blank substrate for a photo mask is chucked on a mask stage of an exposure apparatus. A main surface at a side of the mask blank substrate, where a thin film for forming a transfer pattern is disposed, has flatness of 0.3 &mu;m or less in a 142 mm square region including its central portion and has a convex shape being relatively high at its central portion and relatively low at its peripheral portion. Difference upon fitting, a virtual reference main surface, having a spherical shape with true sphere, to the main surface in a 132 mm square region of a virtual reference substrate is 40 nm or less. <P>COPYRIGHT: (C)2013,JPO&amp;INPIT

Description

  The present invention relates to a mask blank substrate and a mask blank for a photomask used in a photolithography process.

  A photomask is used in a photolithography process of a semiconductor manufacturing process. As the miniaturization of semiconductor devices progresses, the demand for miniaturization in this photolithography process is increasing. In particular, the exposure apparatus using ArF exposure light (193 nm) has been increased in NA in order to cope with miniaturization, and further increased in NA due to the introduction of immersion exposure technology. In order to cope with such a demand for miniaturization and an increase in NA, it is required to increase the flatness of the photomask. In other words, as the pattern line width becomes finer, the allowable amount of displacement of the transferred pattern due to the flatness becomes smaller, and as the NA increases, the focus tolerance in the lithography process becomes smaller. Therefore, the flatness of the main surface of the mask substrate, particularly the pattern forming side (hereinafter, this main surface is simply referred to as the main surface or the substrate main surface) is becoming more important. .

  On the other hand, when the photomask is chucked on the mask stage of the exposure apparatus by a vacuum chuck, it may be greatly deformed at the time of chucking due to compatibility with the mask stage and the vacuum chuck. That is, since product management is conventionally performed with the flatness of the photomask before chucking, even if the main surface shape before chucking is a good product with high flatness, depending on the compatibility with the mask stage and vacuum chuck, When chucked on the mask stage of the exposure apparatus, the photomask may be deformed and the flatness of the photomask may be greatly deteriorated. In particular, the tendency is remarkable in the substrate that has a low symmetry of the shape of the main surface and tends to have a twisted shape. For this reason, it is necessary to consider the flatness when the photomask is chucked on the vacuum chuck. Conventionally, a method for selecting a mask substrate with good flatness after being chucked on a mask stage of an exposure apparatus has been proposed (see, for example, Patent Document 1).

JP 2003-50458 A

  However, according to the conventional method, for each of a plurality of mask substrates (mask blank substrates), information indicating the surface shape of the main surface and flatness information of the main surface before and after chucking on the mask stage of the exposure apparatus are obtained. Obtain information indicating the flatness of the main surface by simulation when the mask substrate is set in the exposure apparatus from the flatness of the main surface of the mask substrate and the structure of the mask chuck of the exposure apparatus. There wasn't. For this reason, conventionally, it has been very troublesome to select a mask substrate having good flatness after being chucked on the mask stage of the exposure apparatus. Further, the structure for chucking the mask substrate on the mask stage differs depending on the exposure apparatus, and it is necessary to select a mask substrate for each exposure apparatus.

  Conventionally, in the polishing process of the substrate, we focused on finishing the flatness of the main surface of the substrate to a higher level, selecting a substrate polished to a high flatness from the polished substrates, and using it as an exposure apparatus to be used. The method of extracting the suitable ones by simulation was taken. However, when polishing to be a substrate having high flatness by a double-side polishing apparatus that polishes a plurality of substrates at the same time, among the substrates polished simultaneously, the number of substrates that reach the target flatness is small, The yield of board production was a problem. Furthermore, as described above, a substrate polished to a high flatness is not necessarily a substrate suitable for an exposure apparatus to be used, and the yield of substrate production is greatly reduced, which is a problem.

  The present invention has been made in view of such points, and it is not necessary to simulate the flatness of the mask substrate after chucking on the mask stage of the exposure apparatus, and the desired post-chucking regardless of the chuck structure of the exposure apparatus. An object of the present invention is to provide a mask blank substrate, a mask blank, and a photomask that can achieve the flatness of the above.

  The mask blank substrate of the present invention is a mask blank substrate for a photomask that is chucked by a mask stage of an exposure apparatus, and the substrate has a virtual reference substrate on the main surface on which a thin film for forming a transfer pattern is provided. The difference when fitting is performed in a region within a 132 mm square including the central portion with respect to the virtual reference main surface is 40 nm or less, and the virtual reference main surface is relatively high in the central portion and relative in the peripheral portion. It is characterized by a convex shape that becomes lower and a spherical shape of a true sphere in a region within a 132 mm square including the central portion.

  In the mask blank substrate of the present invention, the flatness in a 132 mm square region of the virtual reference main surface is preferably 0.3 μm or less. In this case, it is preferable that the radius of curvature of the virtual reference main surface is 14,500,000 mm or more. In the mask blank substrate of the present invention, it is preferable that the substrate has a flatness of 0.3 μm or less in a 142 mm square region including the central portion on the main surface on the side where the thin film for forming the transfer pattern is provided. .

  The mask blank of this invention comprises the said mask blank board | substrate and the thin film formed on the said main surface of the said mask blank board | substrate.

  In the mask blank of the present invention, the thin film is preferably made of a material containing chromium or a material containing molybdenum silicide.

  The photomask of the present invention has a transfer pattern constituted by the thin film of the mask blank.

  In the photomask of the present invention, flatness in a 132 mm square region of the main surface on the side where the thin film is provided when chucked by an exposure apparatus is required for a DRAM half-pitch 32 nm generation photomask. It is preferable that the flatness to be satisfied is satisfied.

  The semiconductor device manufacturing method of the present invention is characterized in that the photomask obtained by the above method is chucked on a mask stage of an exposure apparatus, and the transfer pattern of the photomask is transferred onto the semiconductor substrate by lithography. To do.

  The mask blank substrate of the present invention is a convex shape in which the shape of the virtual reference main surface is relatively high at the central portion and relatively low at the peripheral portion in advance, in a region within 132 mm square of the virtual reference main surface. A virtual reference substrate having a spherical shape, that is, a substrate having an ideal main surface shape is selected, and a 132 mm square including the central portion of the main surface on the side where the thin film of the actually manufactured substrate is provided with respect to the virtual reference main surface. And the difference is 40 nm or less, and the flatness in the region within 142 mm square including the center is 0.3 μm or less is used as the mask blank substrate for the acceptable product. Therefore, it is not necessary to simulate the flatness of the mask substrate after being chucked on the mask stage of the exposure apparatus.

  In addition, the polishing accuracy when manufacturing a mask blank substrate is relaxed compared to the case of manufacturing a very high flatness substrate, and even a substrate that originally could not satisfy very high flatness, A substrate capable of producing a photomask exhibiting sufficient transfer performance with a predetermined exposure apparatus can be shipped as an acceptable product. Therefore, a significant improvement in yield can be achieved. Although it took a lot of time to perform simulation for each manufactured substrate one by one, in the case of the present invention, the manufactured substrate is compared with a virtual reference main surface of a predetermined virtual reference substrate. Thus, the time required for the pass / fail judgment of the substrate can be greatly reduced. Furthermore, by setting a virtual reference main surface of a virtual reference substrate that can be used in common with various chuck-type exposure apparatuses, a photomask that exhibits a predetermined transfer performance can be manufactured regardless of the chuck structure of the exposure apparatus. A mask blank substrate can be supplied.

It is a top view of a substrate main surface direction when a photomask is placed on a chuck stage of an exposure apparatus. It is a figure which shows the shape of the photomask before chucking a photomask on the chuck | zipper stage, (a) is a side view seen from the A direction shown in FIG. 1, (b) is seen from the B direction shown in FIG. FIG. 2A and 2B are views showing the shape of the photomask after the photomask is chucked on the chuck stage, wherein FIG. 2A is a side view seen from the direction A shown in FIG. 1, and FIG. 2B is a view seen from the direction B shown in FIG. FIG. 2 is a contour diagram showing the main surface shape of a substrate to which the present invention is applied, (a) is a contour diagram showing the main surface shape of the substrate before chucking on the chuck stage of the exposure apparatus, and (b) is a chuck of the exposure apparatus. A contour map showing the main surface shape of the substrate after chucking on the stage is shown. It is a top view of the main surface direction of the board | substrate for mask blanks concerning embodiment of this invention. It is sectional drawing which follows Y1-Y1 in Fig.5 (a). It is sectional drawing which follows XY1-XY1 in Fig.5 (a). It is a figure which shows the partial expanded cross section of the board | substrate for mask blanks shown in FIG. It is a figure which shows schematic structure of the sputtering device used when manufacturing the mask blank which concerns on embodiment of this invention. FIG. 4 is a contour diagram showing the main surface shape of a glass substrate manufactured in Example 2. It is a figure which shows the main surface shape of the cross section which follows the XYR1-XYR1 line in the glass substrate shown in FIG. 8, and the cross section along the XYR2-XYR2 line. It is a contour map of the shape of a virtual reference main surface. It is the figure which fitted the virtual reference main surface shown in FIG. 10 to the glass substrate shown in FIG. It is a figure regarding this fitting difference when fitting was performed in FIG. It is the schematic explaining the processing state by the MRF processing method of Example 7, (a) shows front direction sectional drawing, (b) shows side direction sectional drawing.

  In the mask blank substrate of the present invention, the photomask when chucked to the mask stage is more important than manufacturing a substrate having a very high flatness when the main surface is not chucked to the mask stage. It is important to make the main surface on which the transfer pattern is formed flat enough to exhibit sufficient transfer performance when the transfer pattern is transferred by an exposure apparatus.

  Analysis of changes in the shape of the substrate when the photomask was chucked on the mask stage of the exposure apparatus revealed the following. Usually, when the photomask is chucked on the mask stage, the exposure apparatus uses the main surfaces on the two end face sides facing each other as a chuck area.

  A substrate whose main surface has been polished by a polishing apparatus has a strong tendency to have a cross-sectional shape that is basically high in the center and low on the end surface due to the nature of the polishing. Photo produced from a substrate having such a main surface shape The mask also has the same surface shape and is often chucked by the mask stage of the exposure apparatus. FIG. 1 is a plan view when a photomask having such a shape is placed on a chuck stage of an exposure apparatus (a part where the surface of the photomask of the mask stage directly contacts and chucks). FIG. 2A is a side view seen from the A direction (short side direction of the chuck stage) shown in FIG. 1 in a state before the photomask is chucked by the chuck stage. FIG. 2B is a side view seen from the B direction (long side direction of the chuck stage) shown in FIG. 1 in the same state before the photomask is chucked by the chuck stage. As can be seen from FIG. 2A, due to the surface shape of the photomask, both end face sides of the photomask are floated on the short side of the chuck stage. As can be seen from FIG. 2B, due to the surface shape of the photomask, both end face sides of the photomask are in a floating state on the long side of the chuck stage.

  In such a mounting state, when the photomask is chucked on the chuck stage, as shown in FIGS. 3A and 3B, the four end face sides of the floating photomask are pulled by suction, The force which has the effect | action which deform | transforms into a secondary component upwards from four end surface directions is added. That is, the substrate tends to be subjected to a force that causes the main surface to deform into a quadratic curved surface (spherical shape) that protrudes upward from the chuck areas on the four end face sides toward the center.

  FIG. 4 is a diagram showing the shape of the substrate before chucking (before suction) and after chucking (after suction) the substrate to which the present invention is applied to the mask stage of the exposure apparatus, and FIG. It is a figure which shows the shape of the board | substrate before adsorption | suction, FIG.4 (b) is a figure which shows the shape of the board | substrate after adsorption | suction. As can be seen from FIG. 4A, the four corners of the main surface of the substrate are slightly higher than the height of the main surface of the chuck area, and gradually increase toward the center. . That is, a substantially circular contour line is shown on the substrate before adsorption. As can be seen from FIG. 4 (b), the substrate after adsorption shows substantially rectangular contour lines, and the number of contour lines within a 132 mm square is small and the interval is wide. In other words, the shape of the main surface of the substrate after chucking is significantly flatter than that before chucking.

  Considering these tendencies, in the mask blank substrate of the present invention, as an ideal substrate (virtual reference substrate), the shape of the main surface (virtual reference main surface) is relatively high at the central portion and at the peripheral portion. It is assumed that the convex shape is relatively low and has a spherical shape with a height difference of at least 0.3 μm or less in a 132 mm square region of the virtual reference substrate. When the shape of the virtual reference main surface after the photomask using the virtual reference substrate is chucked by the exposure apparatus is simulated, the flatness of the virtual reference main surface is 0.08 μm or less. Then, fitting is performed in a region within a 132 mm square including the central portion of the main surface on the side where the thin film of the substrate actually manufactured by predetermined polishing is applied to the reference curved surface shape of the virtual reference substrate. In which the flatness in a 142 mm square region including the central portion is 0.3 μm or less is used as an acceptable mask blank substrate. As a result, the obtained mask blank substrate can satisfy the flatness required for a DRAM half-pitch (hp) 32 nm generation photomask in a 132 mm square region where a transfer pattern is formed. can get.

  When fitting the virtual reference main surface to the 132 mm square area on the main surface of the actual substrate (real substrate) after polishing, the virtual reference main surface is larger than the main surface of the real substrate at the boundary of the 132 mm square area. It is preferable that the fitting is performed at such a height positional relationship that the height becomes at least high. More preferably, it is preferably performed in a height relationship where the virtual reference main surface coincides with the main surface of the actual substrate as much as possible at the boundary portion of the 132 mm square area.

  Note that the spherical shape of the virtual reference main surface here is not limited to a complete spherical partial shape. Depending on the tendency of the cross-sectional shape of the actual substrate after polishing due to the characteristics of the polishing apparatus used in the polishing process, and the chucking force of the chuck on the mask stage of the exposure apparatus in which the actual substrate is used, However, there is a case where a tendency to apply a force for deforming more strongly than the other pair of end surfaces orthogonal to each other may be increased. In such a case, the shape of the virtual reference main surface may be an elliptical spherical shape.

  By applying the present invention, the polishing accuracy when manufacturing a mask blank substrate is relaxed as compared with the case of manufacturing a substrate with a very high flatness, and furthermore, originally a very high flatness could not be satisfied. Even if it is a substrate, a substrate capable of producing a photomask exhibiting sufficient transfer performance with a predetermined exposure apparatus can be shipped as an acceptable product. Therefore, a significant improvement in yield can be achieved. In addition, in the case of the present invention, fitting to a virtual reference main surface of a predetermined virtual reference substrate is required to perform simulation for each manufactured substrate one by one. The time required for the pass / fail judgment of the substrate can be greatly reduced. Further, by selecting a virtual reference main surface of an ideal virtual reference substrate for each chuck method of the exposure apparatus, and setting a virtual reference main surface of a virtual reference substrate that can be commonly applied from those reference curved surfaces, an exposure apparatus Regardless of the chuck structure, a mask blank substrate capable of producing a photomask exhibiting predetermined transfer performance can be supplied.

  About the manufactured board | substrate, the flatness in the area | region within 142 square mm including the center part of the main surface is accept | permitting only what is 0.3 micrometer or less as an acceptable product, It has a flatness larger than 0.3 micrometer. In the case of a photomask, the amount of deformation when chucked by an exposure apparatus is large, and the positional deviation in the plane direction of the transfer pattern formed on the photomask is large.

  In addition, a product having a difference of 40 nm or less when fitted in a region within a 132 mm square including the central portion of the main surface of the manufactured substrate with respect to the reference curved surface shape of the virtual reference substrate is accepted. The difference at the time of fitting is that when the main surface of the substrate manufactured on the reference curved surface is fitted, the difference when the main surface of the substrate is above the reference curved surface is allowed up to 40 nm. A difference of up to 40 nm is allowed when the main surface is below the reference curved surface.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
5A and 5B are diagrams for explaining a mask blank substrate according to an embodiment of the present invention. FIG. 5A is a plan view, and FIG. 5B is a cross-sectional view taken along line Y1-Y1 in FIG. (C) is a sectional view taken along line XY1-XY1 in (a). Note that the shape shown in FIG. 5B is almost the same as the shape in the cross-sectional view taken along line X1-X1 in FIG. 5A, and the shape shown in FIG. 5C is the line XY2-XY2 in FIG. Is substantially the same as the shape in the cross-sectional view taken along the line. In the mask blank substrate shown in FIG. 5 (a), the main surface on the side where the thin film for forming the transfer pattern is provided has a flatness of 0.3 μm or less in a region within a 142 mm square including the central portion, and at the central portion. It is a convex shape that is relatively high and relatively low at the periphery. In FIG. 5A, the length of one side of the mask blank substrate is Ls (A = 152 mm), the length of one side of the 142 mm square region is Lb (B = 142 mm), and the region of the 132 mm square region is The length of one side is Lp (C = 132 mm). As shown in FIGS. 5B and 5C, the flatness in the 142 mm square region includes the highest portion and the lowest portion of the mask blank substrate 1 in the region. The portion where the difference H is the maximum.

  In addition, the mask blank substrate has a difference of 40 nm or less when the virtual reference main surface of a predetermined virtual reference substrate is fitted to the shape of the main surface. Here, the virtual reference substrate is a convex shape in which the shape of the virtual reference main surface is relatively high at the central portion and relatively low at the peripheral portion, and is spherical in a region within 132 mm square of the virtual reference substrate. It is a shape. More specifically, it refers to a substrate in which the shape of the virtual reference main surface has a flatness of an area within a 132 mm square including the central portion thereof is 0.3 μm or less, and preferably the flatness of an area within a 132 mm square is zero. A substrate having a thickness of 2 μm or less is preferable. In particular, a virtual reference substrate for obtaining a mask blank substrate that can be used in common with various chuck-type exposure apparatuses preferably has a shape in which the virtual reference main surface is defined by a true spherical surface. .

FIG. 6 shows a partially enlarged sectional view of the mask blank substrate 1 shown in FIG. The virtual reference main surface 3 is the main surface shape of the virtual reference substrate, and is in a state when fitted to the main surface 2. And the difference when fitting with the virtual reference main surface 3 in a 132 mm square area (the area indicated by Lp in FIG. 5A) including the central portion of the main surface 2 is D ₁ and D ₂ . is there. D ₁ is the maximum difference (absolute value) of the portions where the substrate main surface 2 is above the virtual reference main surface 3, and D ₂ is the substrate main surface 2 below the virtual reference main surface 3. It is the largest difference (absolute value) of the parts in The larger difference between the differences D ₁ and D ₂ is 40 nm or less.

  The measurement range is preferably an area within a 132 mm square including the central portion of the mask blank substrate. By ensuring the flatness in this range, the fine pattern can be transferred accurately. The difference between the mask blank substrate and the virtual reference substrate whose convex shape is a predetermined reference curved surface and the virtual reference substrate whose convex shape is a predetermined reference curved surface is the wavelength modulation of TTV (plate thickness variation). It was determined as described above by measuring with a wavelength shift interferometer using a laser. This wavelength shift interferometer calculates the difference in height of the measured surface from the interference fringes between the reflected light reflected from the measured surface and the back surface of the mask blank substrate and the measuring machine reference surface (front reference surface). The frequency difference of each interference fringe is detected, and the interference fringe with the measuring machine reference surface (front reference surface) by the reflected light reflected from the measured surface and the back surface of the mask blank substrate is separated, The uneven shape of the measurement surface is measured.

In the present invention, a glass substrate can be used as the mask blank substrate. The glass substrate is not particularly limited as long as it is used as a mask blank. Examples thereof include synthetic quartz glass, soda lime glass, aluminosilicate glass, borosilicate glass, and alkali-free glass. In the case of a glass substrate for EUV mask blanks, in order to suppress distortion of the transferred pattern due to heat during exposure, it is within the range of about 0 ± 1.0 × 10 ⁻⁷ / ° C., more preferably about 0 ± 0. A glass material having a low coefficient of thermal expansion in the range of 3 × 10 ⁻⁷ / ° C. is used. Furthermore, since a large number of films are formed on a glass substrate, an EUV mask blank is made of a highly rigid glass material that can suppress deformation due to film stress. In particular, a glass material having a high Young's modulus of 65 GPa or more is preferable. For example, amorphous glass such as SiO ₂ —TiO ₂ glass or synthetic quartz glass, or crystallized glass on which β-quartz solid solution is deposited is used.

  Such a mask blank substrate can be manufactured through, for example, a rough polishing process, a precision polishing process, and an ultraprecision polishing process.

The shape of the substrate to be manufactured is polished so as to fit the shape of the virtual reference main surface of the virtual reference substrate. When the shape of the reference curved surface to be fitted is, for example, a spherical surface, adjustment is performed in each polishing step so as to approach a curved surface defined by x ² + y ² + z ² = r ² (r: radius of curvature). The curved shape of the virtual reference main surface with a flatness of 0.3 μm or less in an area within 132 mm square is one having a curvature radius r of approximately 14,500,000 mm or more, and is an area within 132 mm square. The curved shape of the virtual reference main surface with a flatness of 0.2 μm or less is that whose radius of curvature r is approximately 21,720,000 mm or more.

  A mask blank can be obtained by forming at least a light-shielding film on the main surface having the convex shape of the mask blank substrate. Examples of the material constituting the light shielding film include chromium or molybdenum silicide. Further, other films, an antireflection film, a semi-transmissive film, and the like may be appropriately formed depending on the use and configuration of the photomask. As the material of the antireflection film, MoSiON, MoSiO, MoSiN, MoSiOC, MoSiOCN, or the like is preferably used, and as the material of the semi-transmissive film, CrO, CrON, or the like is preferably used.

  The light shielding film can be formed by a sputtering method. As the sputtering apparatus, a DC magnetron sputtering apparatus, an RF magnetron sputtering apparatus, or the like can be used. When sputtering the light-shielding film onto the mask blank substrate, it is preferable to form the film by rotating the substrate and disposing the target at a position inclined by a predetermined angle from the rotation axis of the substrate. By such a film formation method, the in-plane variation of the light shielding film can be reduced and the film can be formed uniformly. Since the main surface of the mask blank substrate of the present invention has a convex shape that draws a predetermined curved surface (for example, a secondary curved surface) outward from the center of the substrate as a vertex, this is a highly symmetric substrate. Thus, by forming the light shielding film, a mask blank having high symmetry in the plane can be obtained.

  In the case of film formation by rotating the substrate and placing the target at a position where the sputter target is inclined at a predetermined angle from the rotation axis of the substrate, the in-plane distribution of the phase angle and transmittance is determined between the substrate and the target. It also changes depending on the positional relationship. The positional relationship between the target and the substrate will be described with reference to FIG. The offset distance (the distance between the central axis of the substrate and the straight line passing through the center of the target and parallel to the central axis of the substrate) is adjusted by the area where the distribution of the phase angle and the transmittance should be ensured. Generally, when the area where the distribution should be secured is large, the necessary offset distance becomes large. In the embodiment, in order to realize the phase angle distribution within ± 2 ° and the transmittance distribution within ± 4 ° within the substrate within 132 mm square, the offset distance needs to be about 200 mm to 350 mm, which is a preferable offset. The distance is 240 mm to 280 mm. The target-substrate vertical distance (T / S) varies in the optimum range depending on the offset distance, but in order to achieve a phase angle distribution within ± 2 ° and a transmittance distribution within ± 4 ° within a substrate within 132 mm square. The target-substrate vertical distance (T / S) needs to be about 200 mm to 380 mm, and the preferable T / S is 210 mm to 300 mm. The target inclination angle affects the film formation speed. In order to obtain a large film formation speed, the target inclination angle is suitably from 0 ° to 45 °, and the preferred target inclination angle is from 10 ° to 30 °.

  A photomask can be manufactured by patterning at least the light shielding film by photolithography and etching to provide a transfer pattern. Note that the etchant for etching is appropriately changed according to the material of the film to be etched.

  Next, examples performed for clarifying the effects of the present invention will be described. In the following examples, the case where the mask blank substrate is a glass substrate will be described.

Example 1
As for the shape of the mask blank substrate manufactured in this Example 1, the shape of the virtual reference main surface of the virtual reference substrate is a spherical surface with a radius of curvature r = 14,508,150 mm, within a 132 mm square. A curved surface shape with a flatness of 0.3 μm is formed, and polishing is performed to fit the shape. Specifically, it is manufactured through the following polishing steps.

A predetermined number of sheets are set on a double-side polishing machine on a synthetic quartz glass substrate (152.4 mm x 152.4 mm x 6.35 mm) that has been lapped and chamfered, and roughened under the following polishing conditions: A polishing step was performed. After the rough polishing step, the glass substrate was subjected to ultrasonic cleaning in order to remove the abrasive grains adhering to the glass substrate. The polishing conditions such as the processing pressure, the number of rotations of the upper and lower surface plates, and the polishing time were appropriately adjusted.
Polishing liquid: Cerium oxide (average particle size 2 μm to 3 μm) + water Polishing pad: Hard polisher (urethane pad)

Next, a predetermined number of sheets were set in a double-side polishing apparatus on the glass substrate after rough polishing, and a precision polishing step was performed under the following polishing conditions. After the precision polishing step, the glass substrate was subjected to ultrasonic cleaning in order to remove abrasive grains adhering to the glass substrate. The polishing conditions such as the processing pressure, the number of rotations of the upper and lower surface plates, and the polishing time were appropriately adjusted. The main surface shape on the side of forming the transfer pattern of the glass substrate after the precision polishing step is polished by adjusting various conditions so that the four corners are convex. This is because in the next ultra-precise polishing process, the four corners of the main surface of the substrate are preferentially polished, and this prevents the edge of the four corners from sagging. The flatness within a 142 mm square can be made 0.3 μm or less.
Polishing liquid: Cerium oxide (average particle size 1μm) + water Polishing pad: Soft polisher (suede type)

Subsequently, a predetermined number of sheets were set in a double-side polishing apparatus on the glass substrate after precision polishing, and an ultra-precision polishing process was performed under the following polishing conditions. After the ultraprecision polishing step, the glass substrate was ultrasonically cleaned to remove abrasive grains adhering to the glass substrate. The polishing conditions such as the processing pressure, the number of rotations of the upper and lower surface plates, and the polishing time were appropriately adjusted. This ultra-precision polishing process has a characteristic that four corners are preferentially polished due to the substrate shape being square. The polishing conditions are set so that the flatness within the 142 mm square of the main surface of the substrate does not become larger than 0.3 μm while the surface roughness of the main surface of the substrate is set to a predetermined roughness of 0.4 nm or less. . Thus, the glass substrate according to the present invention was produced.
Polishing liquid: Colloidal silica (average particle size 100 nm) + water Polishing pad: Super soft polisher (suede type)

  With respect to the glass substrate thus obtained, the convex shape of the virtual reference substrate whose shape or convex shape of the glass substrate is a predetermined curved surface is a predetermined curved surface (spherical surface, with a radius of curvature r = 14,508,150 mm, 132 mm. The difference from the virtual reference main surface, which is a curved surface shape with a flatness of 0.3 μm in the region within the corner. As a result, the shape of the obtained glass substrate was a convex shape in which the shape of the main surface on which the thin film forming the transfer pattern was provided was relatively high at the center and relatively low at the periphery. Furthermore, the number of acceptable products that can be used as the mask blank substrate of the present invention is less than 40 nm when fitting with the virtual reference main surface, that is, 95 out of 100 substrates, according to the conventional substrate selection method It was possible to manufacture at a higher yield than (80 out of 100 were acceptable products).

Next, a back surface antireflection layer, a light shielding layer, and a surface antireflection layer were formed in this order on the glass substrate obtained as described above. Specifically, a Cr target is used as a sputtering target, and a mixed gas of Ar, CO ₂ , N ₂ , and He is used as a sputtering gas (gas flow ratio Ar: CO ₂ : N ₂ : He = 24: 29: 12: 35). And a gas pressure of 0.2 Pa, a DC power supply of 1.7 kW, and a CrOCN film having a thickness of 39 nm was formed as a back surface antireflection layer. Next, a Cr target is used as a sputtering target, a mixed gas of Ar, NO, and He is used as a sputtering gas (gas flow ratio Ar: NO: He = 27: 18: 55), a gas pressure is 0.1 Pa, and a DC power supply is used. A CrON film having a thickness of 17 nm was formed as a light shielding layer at 1.7 kW. Next, a Cr target is used as a sputtering target, and a mixed gas of Ar, CO ₂ , N ₂ , and He is used as a sputtering gas (gas flow ratio Ar: CO ₂ : N ₂ : He = 21: 37: 11: 31), A CrOCN film having a thickness of 14 nm was formed as a surface antireflection layer with a gas pressure of 0.2 Pa and a DC power supply of 1.8 kW. The back surface antireflection layer, the light shielding layer, and the surface antireflection layer formed under these conditions had very low stress throughout the light shielding film, and the change in the shape of the substrate could be minimized. In this way, a mask blank was produced.

  A photomask (binary mask) was prepared by patterning the light-shielding film and the antireflection film of the mask blank thus obtained into a predetermined pattern. The obtained photomask was mounted on three exposure apparatuses having different vacuum chuck structures for verification. As an exposure apparatus to be used, the chuck structure of the mask stage is of a type in which the main surfaces on the two opposite end faces of the photomask are chuck areas, and the portion of the chuck area that contacts the photomask is made of a material with low elasticity. Two types are selected for the so-called hard chuck structure formed, and one type is selected for the so-called soft chuck structure in which the chuck structure of the mask stage is formed of a highly elastic material where the photomask contacts. did. Flatness was determined with a wavelength shift interferometer using a wavelength modulated laser. As a result, in any of the exposure apparatuses, the flatness of the photomask after chucking was 0.12 μm or less in all cases, and good transfer performance could be obtained as a DRAM hp32 nm generation photomask.

(Example 2)
In the second embodiment, the shape of the virtual reference main surface of the virtual reference substrate to be aimed is a spherical surface having a curvature radius r = 21,762,225 mm, 132 mm square and a flatness of 0.2 μm. The glass substrate was manufactured by appropriately adjusting the polishing conditions in the same polishing step as in Example 1 in the shape. And about the obtained glass substrate, the convex shape of the virtual reference substrate whose shape or convex shape of the glass substrate is a predetermined curved surface is a predetermined curved surface (spherical surface with a radius of curvature of r = 21,762,225 mm, 132 mm square. In the region, the difference from the virtual reference main surface having a curved surface shape with a flatness of 0.2 μm was determined by a wavelength shift interferometer using a wavelength modulation laser. As a result, the shape of the obtained glass substrate was a convex shape in which the shape of the main surface on which the thin film forming the transfer pattern was provided was relatively high at the center and relatively low at the periphery.

  Next, fitting with the virtual reference main surface was performed on the manufactured glass substrate. FIG. 8 is a contour diagram showing the shape of the main surface of the substrate measured with a wavelength shift interferometer for one of the manufactured glass substrates. Moreover, each main surface shape in the both diagonal lines (XYR1-XYR1 line | wire of FIG. 8, XYR2-XYR2 line | wire) of the glass substrate is shown in FIG. As a result of the measurement, the flatness within a 142 mm square of this glass substrate was 0.19 μm, the flatness within a 132 mm square was also 0.18 μm, and the flatness to be obtained was 0.2 μm or less. FIG. 10 is a contour map showing the shape of the virtual reference main surface for fitting within a 132 mm square of the virtual reference substrate. FIG. 11 shows a cross-sectional shape of the glass substrate shown in FIG. 8 when the virtual reference main surface of FIG. 10 is fitted within a 132 mm square. FIG. 12 shows the difference between the main surface of the glass substrate and the ideal reference main surface shape when this fitting is performed. As shown in FIG. 12, when the fitting is performed, a portion where the height of the virtual reference substrate is higher than the height of the main surface of the glass substrate is represented by a positive numerical value. The part where the height of the surface is higher is represented by a negative value.

  From the results shown in FIG. 12, the fitting difference was 0.0075 μm (7.5 nm) when the value was positive, and −0.0067 μm (6.7 nm) when the value was negative. In addition, even within the entire 132 mm square, the fitting difference was 0.011 (11 nm) at most and 40 nm or less, and it was a highly accurate acceptable product. Similarly, as a result of fitting with the virtual reference main surface for the other manufactured glass, the difference when fitting is 40 nm or less, that is, the acceptable product that can be used as the mask blank substrate of the present invention, It was 90 sheets out of 100 sheets, and it was possible to manufacture with a higher yield than the case of using the conventional substrate selection method (80 sheets out of 100 sheets were acceptable products).

  Next, a back surface antireflection layer, a light shielding layer, and a surface antireflection layer were formed in this order on the glass substrate obtained as described above under the same film forming conditions as in Example 1 to prepare a mask blank. Furthermore, the photomask (binary mask) was produced by patterning the light shielding film and antireflection film of the mask blank obtained in this way into a predetermined pattern. The obtained photomask was mounted on three types of exposure apparatuses having different vacuum chuck structures in the same manner as in Example 1 and verified. Flatness was determined with a wavelength shift interferometer using a wavelength modulated laser. As a result, in any exposure apparatus, the flatness of the photomask after chucking is 0.08 μm or less in all cases, which is sufficient as a DRAM hp22 nm generation photomask as well as a DRAM hp32 nm generation photomask. Good transfer performance could be obtained.

(Example 3)
On the glass substrate prepared in Example 1, a laminated film including a phase shift film, a back surface antireflection film, a light shielding film, and a surface antireflection film was formed. Specifically, a mixed target of Mo and Si (atomic% ratio Mo: Si = 10: 90) is used as a sputtering target, and a mixed gas of Ar, N ₂ , and He is used as a sputtering gas (gas flow ratio Ar: N ₂ : He = 5: 49: 46), a gas pressure of 0.3 Pa, a DC power source power of 2.8 kW, and a MoSiN film as a phase shift film were formed to a thickness of 69 nm. Next, the substrate on which the phase shift film was formed was heat-treated (annealed) at 250 ° C. for 5 minutes.

Next, a light shielding film composed of a back surface antireflection layer, a light shielding layer, and a front surface antireflection layer was formed. Specifically, first, a Cr target is used as a sputtering target, and a mixed gas of Ar, CO ₂ , N ₂ , and He is used as a sputtering gas (gas flow ratio Ar: CO ₂ : N ₂ : He = 22: 39: 6 33), a gas pressure of 0.2 Pa, a DC power supply of 1.7 kW, and a CrOCN film having a thickness of 30 nm was formed as a back-surface antireflection layer. Next, a Cr target is used as a sputtering target, a mixed gas of Ar and N ₂ is used as a sputtering gas (gas flow ratio Ar: N ₂ = 83: 17), a gas pressure is 0.1 Pa, and a DC power supply is 1.7 kW. Then, a CrN film having a thickness of 4 nm was formed as a light shielding layer. Next, a Cr target is used as a sputtering target, and a mixed gas of Ar, CO ₂ , N ₂ , and He is used as a sputtering gas (gas flow ratio Ar: CO ₂ : N ₂ : He = 21: 37: 11: 31), A CrOCN film having a thickness of 14 nm was formed as a surface antireflection layer with a gas pressure of 0.2 Pa and a DC power supply of 1.8 kW. The back surface antireflection layer, the light shielding layer, and the surface antireflection layer formed under these conditions have very low stress in the entire light shielding film, and the phase shift film also has very low stress. Minimized.

  A photomask (phase shift mask) was produced by patterning the light-shielding film and the antireflection film of the mask blank thus obtained into a predetermined pattern. The obtained photomask was mounted on three exposure apparatuses having different vacuum chuck structures in the same manner as in Example 1 and verified. Flatness was determined with a wavelength shift interferometer using a wavelength modulated laser. As a result, in any exposure apparatus, the flatness of the photomask was 0.12 μm or less in all cases, and good transfer performance could be obtained as a DRAM hp32 nm generation photomask.

Example 4
On the glass substrate produced in Example 2, a laminated film composed of a phase shift film having the same structure as that of Example 3, a back surface antireflection film, a light shielding film, and a surface antireflection film was formed. A photomask (phase shift mask) was produced by patterning the light-shielding film and the antireflection film of the mask blank thus obtained into a predetermined pattern. The obtained photomask was mounted on three types of exposure apparatuses having different vacuum chuck structures in the same manner as in Example 1 and verified. Flatness was determined with a wavelength shift interferometer using a wavelength modulated laser. As a result, in any exposure apparatus, the flatness of the photomask after chucking is 0.08 μm or less in any case, and not only the DRAM hp32 nm generation photomask but also the DRAM hp22 nm generation photomask, A sufficiently good transfer performance could be obtained.

(Example 5)
On the glass substrate produced in Example 1, a MoSiON film (back surface antireflection layer), MoSi (light shielding layer), and MoSiON film (antireflection layer) were formed as a light shielding film. Specifically, a target of Mo: Si = 21: 79 (atomic% ratio) is used, and Ar, O ₂ , N ₂ and He are sputtered at a gas pressure of 0.2 Pa (gas flow ratio Ar: NO: N ₂ : He). = 5: 4: 49: 42), the power of the DC power source is 3.0 kW, and a film made of molybdenum, silicon, oxygen, and nitrogen (MoSiON film: the atomic percent ratio of Mo to Si in the film is about 21:79) ) With a film thickness of 7 nm, then using the same target, Ar with a sputtering gas pressure of 0.1 Pa, DC power supply of 2.0 kW, and a film made of molybdenum and silicon (MoSi film: in the film) The atomic percent ratio of Mo to Si is about 21:79) with a film thickness of 35 nm, and then using a target of Mo: Si = 4: 96 (atomic percent ratio), Ar, O ₂ , N ₂ and He Sputtering gas pressure 0. Pa (gas flow ratio _{Ar: NO: N 2: He} = 5: 4: 49: 42) and then, the power of the DC power supply at 3.0 kW, molybdenum, silicon, oxygen, of nitrogen film (MoSiON film: film The atomic percent ratio of Mo to Si was about 4:96) with a film thickness of 10 nm. The total film thickness of the light-shielding film was 52 nm. The back surface antireflection layer, the light shielding layer, and the surface antireflection layer formed under these conditions had very low stress throughout the light shielding film, and the change in the shape of the substrate could be minimized.

  A photomask (binary mask) was prepared by patterning the light-shielding film and the antireflection film of the mask blank thus obtained into a predetermined pattern. The obtained photomask was mounted on three exposure apparatuses having different vacuum chuck structures in the same manner as in Example 1 and verified. As a result, in any exposure apparatus, the flatness of the photomask was 0.12 μm or less in all cases, and good transfer performance was obtained as a DRAM hp32 nm generation photomask.

(Example 6)
On the glass substrate produced in Example 2, a light-shielding film was formed by sequentially laminating a MoSiON film (back surface antireflection layer), MoSi (light-shielding layer), and MoSiON film (antireflection layer) having the same structure as in Example 5. . A photomask (binary mask) was prepared by patterning the light-shielding film and the antireflection film of the mask blank thus obtained into a predetermined pattern. The obtained photomask was mounted on three types of exposure apparatuses having different vacuum chuck structures in the same manner as in Example 1 and verified. Flatness was determined with a wavelength shift interferometer using a wavelength modulated laser. As a result, in any exposure apparatus, the flatness of the photomask after chucking is 0.08 μm or less in all cases, which is sufficient as a DRAM hp22 nm generation photomask as well as a DRAM hp32 nm generation photomask. Good transfer performance could be obtained.

(Example 7)
In Example 1, the main surface of the glass substrate that had been subjected to the ultra-precision polishing step and the ultrasonic cleaning was subjected to local processing by the MRF (Magneto Rheological Finishing) processing method. First, the flatness of the main surface of the glass substrate was measured with a wavelength shift interferometer using a wavelength modulation laser (measurement region: a region within 142 mm square concentric with the center of the substrate). Next, based on the actually measured values, first, it is verified whether the flatness is in a range of 0.3 μm or less in a 142 mm square region on the main surface of the substrate. When the flatness exceeds 0.3 μm, an area having a height exceeding 0.3 μm is identified as an area that requires local machining as viewed from the lowest point, and the required machining amount is calculated. Next, fitting is performed on the reference curved surface of the virtual reference substrate with respect to the region within 132 mm square of the substrate main surface based on the actual measurement value of the substrate main surface. In this case, fitting is performed so that the reference curved surface is not located at a height above a predetermined allowable maximum fitting difference (40 nm) with respect to the main surface of the substrate in a 132 mm square region. Then, with respect to the fitted reference curved surface, a position where the main surface of the substrate is located above a predetermined allowable maximum fitting difference (40 nm) is specified as an area that requires local processing, and a necessary processing amount is calculated. . At this stage, the substrate that is determined not to require local processing is an acceptable product that can be used as the mask blank substrate of the present invention.

  Next, local processing is required, and local processing by the MRF processing method is performed on the glass substrate whose region is specified. The MRF processing method is a method in which polishing abrasives contained in a magnetic fluid are brought into contact with a substrate with the aid of a magnetic field, and the residence time of the contact portion is controlled to locally perform polishing processing. In this polishing process, the longer the convexity of the convex part, the longer the residence time of the contact part by the abrasive grains. Further, the smaller the convexity of the convex part, the shorter the residence time of the contact part by the abrasive grains is controlled.

  FIGS. 13A and 13B are schematic diagrams for explaining a processing state by the MRF processing method, in which FIG. 13A is a front sectional view and FIG. 13B is a side sectional view. In this figure, according to the MRF processing method, a polishing blank (not shown) contained in a magnetic fluid 41 containing iron (not shown) is used as a workpiece for mask blanks with the aid of a magnetic field. 1 is contacted at a high speed, and the dwelling time of the contact portion is controlled to perform local polishing. That is, a mixed liquid of the magnetic fluid 41 and the polishing slurry 42 (magnetic polishing slurry 4) is put into the disc-shaped electromagnet 6 that is rotatably supported, and the tip thereof becomes the polishing spot 5 for local processing and should be removed. The convex portion 13 is in contact with the polishing spot 5. In this manner, the magnetic polishing slurry 4 flows in a substantially two-layer state along the magnetic field on the disk, with a large amount of polishing slurry 42 distributed on the substrate 1 side and a large amount of magnetic fluid 41 distributed on the electromagnet 3 side. . A part of this state is used as a polishing spot 5 for locally polishing, and is brought into contact with the surface of the substrate 1, whereby the convex portion 13 is locally polished and controlled to a flatness of several tens of nm.

  In this MRF processing method, unlike the conventional polishing method, since the polishing spot 5 always flows, there is no deterioration in processing accuracy due to wear or shape change of the processing tool, and it is necessary to press the substrate 1 with a high load. Therefore, there is an advantage that there are few latent scratches and scratches in the surface displacement layer. Further, in the MRF processing method, when the substrate 1 is moved while bringing the polishing spot 5 into contact, the moving speed of the substrate 1 is controlled according to the processing allowance (required processing amount) set for each predetermined region. The removal amount can be easily adjusted.

  As the polishing slurry 42 mixed with the magnetic fluid 41, a slurry in which fine abrasive particles are dispersed in a liquid is used. The abrasive particles are, for example, silicon carbide, aluminum oxide, diamond, cerium oxide, zirconium oxide, manganese oxide, colloidal silica, and the like, and are appropriately selected according to the material of the workpiece, the processed surface roughness, and the like. These abrasive particles are dispersed in a liquid such as water, an acidic solution, or an alkaline solution to form an abrasive slurry 42 and mixed with the magnetic fluid 41.

As a result of fitting between the main surface of the mask blank substrate 1 and the virtual reference main surface, local polishing processing is performed for the calculated required processing amount at a location that is determined to require local polishing processing by the MRF processing method. It was. Next, since the main surface subjected to the local polishing process was rough, double-side polishing was performed using a double-side polishing apparatus only for a short time. Double-side polishing was performed under the following polishing conditions. The polishing conditions such as the processing pressure, the number of rotations of the upper and lower surface plates, and the polishing time were appropriately adjusted.
Polishing liquid: colloidal silica (average particle size 70 nm)
+ Alkaline aqueous solution (NaOH, pH 11)
Polishing pad: Super soft polisher (suede type)

  As a result, the shape of the obtained glass substrate was a convex shape in which the shape of the main surface on which the thin film forming the transfer pattern was provided was relatively high at the center and relatively low at the periphery. Furthermore, the number of acceptable products that can be used as the mask blank substrate of the present invention is 100 out of 100 when the difference from the virtual reference main surface is 40 nm or less, that is, can be manufactured with a very high yield. did it.

  Next, a light-shielding film was formed on the glass substrate obtained as described above in the same manner as in Example 1 to produce a mask blank. Furthermore, the photomask (phase shift mask) was produced by patterning the light shielding film and antireflection film of the obtained mask blank into a predetermined pattern. The obtained photomask was mounted on three exposure apparatuses having different vacuum chuck structures in the same manner as in Example 1 and verified. Flatness was determined with a wavelength shift interferometer using a wavelength modulated laser. As a result, in any exposure apparatus, the flatness of the photomask was 0.12 μm or less in all cases, and good transfer performance could be obtained as a DRAM hp32 nm generation photomask.

(Example 8)
In Example 2, the main surface of the glass substrate that had been subjected to the ultra-precise polishing step and ultrasonic cleaning was subjected to local processing by MRF (Magneto Rheological Finishing) processing method in the same manner as in Example 7. Here, not only the flatness is in the range of 0.3 μm or less in the 142 mm square region of the main surface of the substrate, but the flatness is in the range of 0.2 μm or less in the 132 mm square region. Local processing was performed. As a result, the shape of the obtained glass substrate was a convex shape in which the shape of the main surface on which the thin film forming the transfer pattern was provided was relatively high at the center and relatively low at the periphery. Furthermore, the number of acceptable products that can be used as the mask blank substrate of the present invention is 100 out of 100 when the difference from the virtual reference main surface is 40 nm or less, that is, can be manufactured with a very high yield. did it.

  Next, a light shielding film was formed on the glass substrate in the same manner as in Example 1. A photomask (binary mask) was produced by patterning the light-shielding film of the mask blank thus obtained into a predetermined pattern. The obtained photomask was mounted on three types of exposure apparatuses having different vacuum chuck structures in the same manner as in Example 1 and verified. Flatness was determined with a wavelength shift interferometer using a wavelength modulated laser. As a result, in any exposure apparatus, the flatness of the photomask after chucking is 0.08 μm or less in all cases, which is sufficient as a DRAM hp22 nm generation photomask as well as a DRAM hp32 nm generation photomask. Good transfer performance could be obtained.

  Thus, according to the present invention, the mask blank substrate has a predetermined convex shape such that the shape of the main surface on which the thin film for forming the transfer pattern is provided is relatively high at the center and relatively low at the peripheral edge. Since the difference in height with respect to the virtual reference substrate, which is the virtual reference main surface, is within ± 40 nm, it is not necessary to simulate the flatness of the mask substrate after chucking on the mask stage of the exposure apparatus, and the chuck of the exposure apparatus Regardless of the structure, the desired flatness after chucking can be realized.

  The present invention is not limited to the above embodiment, and can be implemented with appropriate modifications. For example, the material, size, processing procedure, and the like in the above embodiment are merely examples, and various modifications can be made within the scope of the effects of the present invention. In addition, various modifications can be made without departing from the scope of the object of the present invention.

DESCRIPTION OF SYMBOLS 1 Mask blank substrate 2 Main surface 3 Virtual reference main surface 4 Magnetic polishing slurry 5 Polishing spot 6 Electromagnet 13 Convex part 41 Magnetic fluid 42 Polishing slurry

Claims (9)

A mask blank substrate for a photomask chucked by a mask stage of an exposure apparatus,
The substrate has a difference of 40 nm or less when the main surface on the side where the thin film for forming the transfer pattern is provided is fitted to the virtual reference main surface of the virtual reference substrate in a 132 mm square region including the central portion. Yes,
The virtual reference main surface has a convex shape that is relatively high at the central portion and relatively low at the peripheral portion, and has a spherical shape of a true sphere in a region within a 132 mm square including the central portion. Mask blank substrate.   2. The mask blank substrate according to claim 1, wherein flatness in a 132 mm square region of the virtual reference main surface is 0.3 μm or less.   3. The mask blank substrate according to claim 2, wherein a radius of curvature of the virtual reference main surface is 14,500,000 mm or more.   4. The flatness in a 142 mm square region including the central portion on the main surface on the side where the thin film for forming the transfer pattern is provided is 0.3 μm or less. 5. The mask blank substrate as described in 1.   A mask blank comprising: the mask blank substrate according to claim 1; and a thin film formed on the main surface of the mask blank substrate.   6. The mask blank according to claim 5, wherein the thin film is made of a material containing chromium or a material containing molybdenum silicide.   A photomask having a transfer pattern composed of the thin film of the mask blank according to claim 5.   The flatness in the 132 mm square region of the main surface on the side provided with the thin film when chucked by the exposure apparatus satisfies the flatness required for a DRAM half-pitch 32 nm generation photomask. The photomask according to claim 7.   9. A method of manufacturing a semiconductor device, comprising: chucking the photomask according to claim 7 on a mask stage of an exposure apparatus; and transferring a pattern of the photomask onto a semiconductor substrate by lithography.

Images