JP2005033171A - Electron projection lithography method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an electron projection lithography method which appropriately corrects variations in the optimum dose caused by the proximity effect and has improved line width control. <P>SOLUTION: A reference bias is determined on condition that the dose (reference dose, reference dose level) on the surface of a sensitive substrate, which is assumed to be able to obtain a predetermined line width with no mask bias correction, has a margin, which is an indicator of variations in the line width caused by unexpected variations in the dose, of 5 nm/μC/cm<SP>2</SP>or less. On the basis of the reference bias, the amount of mask bias for each element of a pattern is calculated, and correction is made to give dimensional change (reshaping) to each element. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an electron beam exposure method used for lithography of semiconductor elements and the like. In particular, the present invention relates to an exposure method with improved line width controllability that is affected by the proximity effect.

  In recent years, development of fine pattern formation technology after the 0.1 μm rule has been actively conducted. Among them, an EB reduction projection exposure method using an electron beam is attracting attention as a method having a throughput capable of dealing with mass production of memory, and is being commercialized.

  One of the features of the EB reduction projection exposure method is that an electron beam with a very high acceleration having an acceleration voltage of 100 kV is used. Thus, by accelerating the electron beam, a very fine resist pattern of 0.1 μm or less can be formed. This is because the electron beam irradiated to the resist is scattered in the polymer of the resist and causes forward scattering in which the trajectory gradually expands. However, when the electron beam is accelerated, the forward scattering diameter decreases. Therefore, it is advantageous for pattern miniaturization.

  By the way, a Coulomb interaction, which is a repulsion between electrons, acts greatly on a portion where the electron beam converges most in the optical system, for example, on a crossover portion or an imaging plane on the wafer. This Coulomb interaction increases the amount of beam blur, resulting in degradation of resolution. However, when the electron beam is accelerated, the Coulomb interaction is reduced, the amount of beam blur is reduced, and the resolution can be improved.

  However, the following problems also occur with the acceleration of electron beams. When an electron beam is incident on the resist, most of the electrons that reach the substrate surface while scattering in the resist enter the substrate as they are and repeat the scattering until the energy becomes zero. The scattering radius becomes larger as the electron beam is accelerated, and is considered to reach about 50 μm when the acceleration voltage is 100 kV. Then, a certain percentage of the electrons that have entered the substrate return to the substrate surface again while repeating scattering. These electrons accumulate in the resist and form a resist image. The electrons returning to the substrate surface are called backscattered electrons. The backscattered electrons sensitize the resist and cause a dose variation. This is called a proximity effect phenomenon. In the EB projection exposure method using a high acceleration electron beam, it is important to correct the variation of the optimum dose due to the proximity effect.

  As a method for correcting the proximity effect, a reshaping correction method and a ghost correction method are considered to be promising. Note that, since the shot size of the variable shaping exposure method is quite small, such as several μm, a method of optimizing the exposure dose for each shot has been put into practical use. However, the EB reduction projection has a large exposure area of 250 μm square. This method cannot be used with an exposure apparatus.

  The reshape correction method is a method of correcting pattern dimensions and shapes on a reticle so that image formation on a wafer has a desired line width and shape. For example, in the case of a pattern such as a so-called DRAM memory gate in which the lines are arranged at 100 μm squares or more at the same interval, the line width increases as the center pattern increases under the exposure conditions in which the peripheral pattern can be transferred to a predetermined line width. Will go. In order to correct this tendency, the line width of the pattern at the center of the reticle may be narrowed (minus correction) by the amount that becomes thicker by exposure.

  By the way, in order to form a finer pattern, the exposure apparatus needs to be adjusted to reduce the amount of beam blur. Further, it is required that the controllability of each performance of the exposure apparatus is balanced and each performance has a sufficient margin. The same applies to the proximity effect correction, and it is desirable that the tolerance of the dimension to be corrected in the reticle pattern is large.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to provide an exposure method that improves the line width controllability by appropriately correcting the variation of the optimum dose amount due to the proximity effect. .

In order to solve the above-described problems, an electron beam projection exposure method of the present invention comprises: forming a device pattern to be transferred on a sensitive substrate on a reticle; illuminating the reticle with an electron beam; and passing an electron beam that has passed through the reticle. Is an exposure method that projects the pattern onto the sensitive substrate to transfer the pattern, and performs mask bias correction including proximity effect correction by giving a dimensional change (reshape) to each element of the reticle pattern in advance, Here, the dose amount (reference dose amount, reference dose level) to the sensitive substrate surface, which is assumed to be able to realize the expected line width with zero mask bias correction, is the change in the correction line width change with respect to unexpected dose amount fluctuations. It is characterized in that the tolerance, which is an index, is set to a condition of 5 nm / μC / cm ² or less.

  By increasing the tolerance of the exposure line width with respect to the dose variation, the required amount of change in the mask bias amount is small even if the parameters of the exposure apparatus fluctuate and the dose varies. Therefore, even if the dose varies, it is possible to perform mask bias correction that can obtain a line width substantially as designed on the sensitive substrate.

Here, if the tolerance is considered as a change ratio of the line width change amount (mask bias amount) to the exposure dose change degree in the relationship between the exposure dose amount and the line width change amount, the tolerance is large. Indicates that this rate of change is small. That is, even if the exposure dose varies greatly, the change amount of the line width change amount is small. In other words, if the margin is considered as the slope of the curve of the graph showing the relationship between the exposure dose and the line width change amount, a large margin means that the slope of this graph is small, and the slope of The absolute value is 5 nm / μC / cm ² or less.

  In the present invention, simulation or test exposure of an appropriate mask bias correction amount for securing a target line width is performed under various conditions in which the dose amount is varied at a typical pattern density of a pattern to be transferred, and the change in the dose amount is controlled. Exposure may be performed under conditions where the change in the mask bias correction amount is small.

It is possible to reliably find a condition with a large margin, which is a condition in which the change in the mask bias correction amount with respect to the change in the dose amount is small (a condition in which the line width change with respect to the dose amount variation is small).
In the present invention, it is preferable that the exposure dose is set higher than the reference exposure dose, and negative mask bias correction is performed accordingly. In other words, the mask bias correction reference should be closer to the isolated pattern. When the exposure dose is set higher than the reference exposure dose (reference dose exposure level) (overdose), the mask bias correction margin is improved.

  As is apparent from the above description, according to the present invention, proximity effect correction by the reshape correction method can be performed with high accuracy. For this reason, not only the 100 nm node but also the establishment of lithography technology as the core of semiconductor device manufacturing technology after the 65 nm node can be expected.

Hereinafter, it demonstrates, referring drawings.
First, the outline of the divided transfer type electron beam exposure method will be described.
FIG. 3 is a diagram showing an outline of the imaging relationship and the control system in the entire optical system of the divided transfer type electron beam exposure apparatus.

  The electron gun 1 arranged at the uppermost stream of the optical system emits an electron beam downward. Below the electron gun 1, two-stage condenser lenses 2 and 3 are provided. The electron beam is converged by these condenser lenses 2 and 3 and crossed over the blanking opening 7. O. Is imaged.

  A rectangular opening 4 is provided below the second-stage condenser lens 3. This rectangular aperture (illumination beam shaping aperture) 4 allows only the illumination beam that illuminates one subfield (pattern small area that becomes one unit of exposure) of the reticle (mask) 10 to pass therethrough. The image of the opening 4 is formed on the reticle 10 by the lens 9.

  A blanking deflector 5 is disposed below the beam shaping opening 4. The deflector 5 deflects the illumination beam when necessary so as to be applied to the non-opening portion of the blanking opening 7 so that the beam does not hit the reticle 10.

  An illumination beam deflector (main deflector of the illumination optical system) 8 is disposed below the blanking opening 7. The deflector 8 scans the illumination beam mainly in the horizontal direction (X direction) in FIG. 3 to illuminate each subfield of the reticle 10 in the field of view of the illumination optical system. An illumination lens 9 is disposed below the deflector 8. The illumination lens 9 images the beam shaping aperture 4 on the reticle 10.

  The reticle 10 actually extends in the plane perpendicular to the optical axis (XY plane). A pattern (chip pattern) forming one semiconductor device chip as a whole is formed on the reticle 10. Of course, a pattern forming one semiconductor device chip may be divided and arranged on a plurality of reticles. The pattern on the reticle 10 is divided into a number of subfields.

  The reticle 10 is mounted on a movable reticle stage 11, and by moving the reticle 10 in a direction perpendicular to the optical axis (XY direction), each reticle on the reticle spreads over a wider range than the field of view of the illumination optical system. The field can be illuminated.

The reticle stage 11 is provided with a position detector 12 using a laser interferometer, so that the position of the reticle stage 11 can be accurately grasped in real time.
Below the reticle 10, projection lenses 15 and 19 and a main deflector (image position adjusting deflector) 16 of the projection optical system are provided. The electron beam that has passed through one subfield of the reticle 10 is imaged at a predetermined position on the sensitive substrate (wafer) 23 by the projection lenses 15 and 19 and the main deflector 16. An appropriate resist is coated on the wafer 23, the electron beam dose is given to the resist, and the pattern on the reticle is reduced and transferred onto the wafer 23.

  A crossover C.D. is formed at a point where the space between the reticle 10 and the wafer 23 is internally divided by a reduction ratio. O. The contrast opening 18 is provided at the crossover position. The contrast opening 18 blocks the electron beam scattered by the non-pattern part of the reticle 10 from reaching the wafer 23.

  The wafer 23 is placed on a wafer stage 24 that can move in the XY directions via an electrostatic chuck (not shown). By synchronously scanning the reticle stage 11 and the wafer stage 24 in opposite directions, each part in the chip pattern extending beyond the field of view of the projection optical system can be sequentially exposed. The wafer stage 24 is also equipped with a position detector 25 similar to that of the reticle stage 11 described above.

  A backscattered electron detector 22 is disposed immediately above the wafer 23. The reflected electron detector 22 detects the amount of electrons reflected by the exposed surface of the wafer 23 or the mark on the stage. For example, the mark on the wafer 23 is scanned with a beam that has passed through the mark pattern on the reticle 10, and the reflected electrons from the mark at that time are detected, whereby the relative positional relationship between the reticle 10 and the wafer 23 and the electron beam (beam) ).

  The lenses 2, 3, 9, 15, 19 and the deflectors 5, 8, 16 are controlled by the controller 31 via the coil power control units 2a, 3a, 9a, 15a, 19a and 5a, 8a, 16a. Controlled. The reticle stage 11 and the wafer stage 24 are also controlled by the controller 31 via the stage controllers 11a and 24a. The stage position detectors 12 and 25 send signals to the controller 31 via interfaces 12a and 25a including amplifiers and A / D converters. The backscattered electron detector 22 also sends a signal to the controller 31 via the same interface 22a.

  The controller 31 grasps the stage position control error and corrects the error by the main deflector 16. Thereby, it is possible to control so that the reduced image of the subfield on the reticle 10 is matched with the target position on the wafer 23. Then, the subfield images are joined on the wafer 23, and the entire chip pattern on the reticle 10 is transferred onto the wafer 23.

Next, the reshape correction method according to the embodiment of the present invention will be described.
First, the bias amount in each part is calculated from the pattern arrangement. The bias amount is a background dose due to backscattered electrons generated by the proximity effect. In general, the higher the pattern density, the greater the bias amount.

  Next, a line width correction amount (mask bias amount) is calculated from the calculated bias amount. This process is a part of a process called data conversion. At this time, it is necessary to determine a portion to be a reference bias and determine a mask bias amount of each portion with respect to this portion. The reference bias can be determined according to various concepts. For example, assuming that there is no proximity effect (bias amount = zero) as a reference, the pattern line width increases as the influence of the proximity effect increases. Therefore, the pattern dimensions on the reticle are designed in the area where the influence of the proximity effect is large. Narrower than the value (minus bias correction). Assuming that the portion having the largest proximity effect is used as a reference, the pattern line width is narrow at the portion where the proximity effect is small, and the portion at which there is no proximity effect is the smallest. Therefore, the pattern dimension on the reticle is made thicker than the design value as the proximity effect is smaller (plus bias correction). Note that the optimum exposure dose at this time (plus bias correction) may be smaller than the optimum exposure dose at minus bias correction.

  By the way, when correcting the mask bias, it is necessary to secure an appropriate margin (tolerance) for each of the exposure performance parameters. That is, even when the parameter fluctuates carelessly, it is necessary to determine the reference bias so that the fluctuation does not affect the mask bias correction as much as possible. Here, the parameters to be particularly considered are the exposure dose amount and the beam blur amount.

Next, the relationship between the exposure dose amount, the beam blur amount (beam blur), and the mask bias amount will be described.
FIG. 1 is a graph showing the relationship between the exposure dose and the mask bias amount. FIG. 1A shows a case where the pattern density is about 50%, and FIG. 1B shows a case where the pattern density is about 0%. .

In each graph, the horizontal axis represents the exposure dose (μC / cm ² ), and the vertical axis represents the mask bias amount (× 100 nm). In the graph, ◆ indicates a beam blur of 15 nm, ▲ indicates a beam blur of 30 nm, and ■ indicates a beam blur of 50 nm.

In the following description, patterns that are mixed in a region where the pattern density is 0 to 50% will be described.
As can be seen from each graph, the mask bias amount greatly depends on the exposure dose amount even when the pattern density is 50% and the proximity effect is large, or when the pattern density is 0% and the proximity effect is not present. Specifically, the mask bias amount decreases from plus to minus as the exposure dose increases. The bias amount (minus bias amount) tends to increase as the pattern density increases.

  Next, the curves connecting the points are compared when the mask bias amount is positive and negative. In both graphs, the slope of the curve connecting the points tends to decrease (sleep) as the mask bias amount becomes negative. The case where the mask bias amount is negative is a case where the exposure dose amount is increased in this example.

As can be seen from the graph, this tendency is greater when the pattern density is 0% (the graph in FIG. 1B).
Here, the tendency that the slope of the curve connecting the points becomes smaller (sleeps) will be described.

The slope of the curve connecting the points indicates the rate of change in the mask bias amount with respect to the degree of change in the exposure dose amount. When the inclination is small (sleeps), the mask bias amount does not change much even if the exposure dose changes. In other words, even if the exposure dose fluctuates due to some influence (exposure device variation, blanking time accuracy, deviation of predicted BGD due to proximity effect, etc.), the line width of the resist pattern on the wafer changes. Indicates less. For example, from the graph of FIG. 1B, when the exposure dose is 15 μC / cm ² , the mask bias amount is about −10 nm, but even if the exposure dose varies and increases to ²⁰ μC / cm ^2. The mask bias amount is about −20 nm, and the difference is 10 nm.

The fact that the change in the mask bias amount is small even when the exposure dose varies is referred to as “the mask bias correction margin (tolerance) is improved”.
Each graph shows a case where the beam blur is changed in three stages (15 nm, 30 nm, and 50 nm), but it can be said that there is almost no difference due to the beam blur in any graph.

From the above, it can be said that the condition for improving the mask bias correction margin is a condition that the exposure dose is large irrespective of the beam blur.
Next, the reference bias will be described.

  As described above, the reference bias indicates a portion of the pattern in which the mask bias is zero (no bias is applied). The dose amount for obtaining the target line width on the wafer in the reference bias portion is referred to as a reference dose amount (reference dose level). For example, in a L / S pattern with a line width design value of 100 nm, assuming a reduction ratio of 1/4, the pattern on a reticle with a line width of 400 nm is the dose when the line width is exactly 100 nm on the wafer. .

Referring to the graph in the figure, the reference dose (reference dose level) is about 9 μC / cm ² in the pattern with the pattern density of 50% in FIG. If the exposure dose is the reference dose (about 9 μC / cm ² ), the mask bias needs to be about +40 nm when the pattern density in the same reticle is 0%. On the other hand, in the pattern having a pattern density of 0% in FIG. 1B, the reference dose (reference dose level) is about 14 μC / cm ² . If the exposure dose is the reference dose (about 14 μC / cm ² ), the mask bias needs to be about −30 nm when the pattern density in the same reticle is 50%.

  As can be seen from each graph, when the exposure dose is set higher than the reference dose (reference dose level) (when the dose is over), the mask bias correction margin is large (the curve connecting the points). Slant of sleep).

As shown in the graph of FIG. 1A, when the exposure dose is 11 μC / cm ² or more, the mask bias correction margin is relatively large (the slope of the curve connecting the points falls), but 10 μC / In the case of cm ² or less, it can be seen that the mask bias correction margin is reduced (the inclination of the curve connecting the points is increased).

Therefore, when the graphs of FIGS. 1A and 1B are compared, the portion having a high mask bias correction margin is the pattern having the pattern density of 0% in FIG. However, in consideration of expanding the pattern selection range, the mask bias correction margin is relatively large even when the pattern density in FIG. 1A is 50% and the exposure dose is 11 μC / cm ² or more. Become.

The specific margin (tolerance) is the slope of the line connecting the points {−0.4 when the exposure dose is 11 μC / cm ² or more when the pattern density in FIG. − (− 0.2)} × 100 nm / (15-11) μC / cm ² ) to 5 nm / μC / cm ² or less is preferable.

The above will be described together with a flowchart.
FIG. 2 is a flowchart of the reshape correction method according to the embodiment of the present invention.
First, in S1, the bias amount of each part of the pattern is calculated. Next, the reference bias is determined in order to calculate the mask bias amount. A portion surrounded by a broken line in the figure is an operation for determining the reference bias. First, in S2, simulation or test exposure is performed with a pattern having a typical pattern density, and the relationship between the exposure dose amount and the mask bias amount is obtained. From the obtained relationship, a pattern portion having a mask bias correction margin of 5 nm / μC / cm ² or less is selected in S3, and the same portion is set as a reference bias in S4.

In S5, the mask bias amount of each part of the pattern is calculated with respect to the set reference bias. Based on the calculation result, a reticle pattern is created in S6.
In the above example, the case of a pattern in which patterns with a density of 0 to 50% are mixed has been described. However, simulation and test exposure are similarly performed when the pattern density range is wider or narrower. Can be sought.

  An example in which an L / S array is formed by simulating a general-purpose memory gate pattern of a 100 nm node will be described. The pattern area in this array is 250 μm × 200 μm, and a line (L) and a space (S) having a width of 100 nm are arranged in this area.

  First, proximity effect correction calculation is performed to calculate the mask bias correction amount of the pattern on the reticle. At this time, the above pattern is input to data conversion software to calculate the mask bias correction amount. At this time, the EID function was approximated by a double Gaussian function. The EID parameters were βf = 7 nm, βb = 31 μm, and η = 0.4. Here, βf is the spread radius of forward scattered electrons, βb is the spread radius of backscattered electrons, and η is the backscattered electron intensity / forward scattered electron intensity.

  The density of the pattern is 0 to 50%. The portion where the pattern density is near 0% is the outermost peripheral portion of the pattern region, and the portion where the pattern density is 50% is a portion located at least 200 μm inside from the four sides (outermost peripheral portion) of the pattern region. is there. Then, the reference bias is set to a portion where the pattern density is 25%, and the mask bias correction is corrected equally for both plus and minus. As a result, -20% mask bias correction is performed at a pattern density of 50%, and + 20% mask bias correction is performed at a pattern density of 0%. Based on this data, the pattern dimensions on the reticle were determined.

The reticle is manufactured by using a membrane made of Si having a thickness of 2 μm and a strut holding the membrane as a basic structure, and using the Si membrane as an electron scatterer, a pattern corrected to the above pattern dimensions as an opening pattern. More specifically, first, a window frame-like pattern serving as a subfield membrane, minor struts, and main struts is formed on the lower surface of the SOI wafer by a photolithography process. Thereafter, the silicon portion on the lower surface was etched to SiO ₂ by a dry etching method. Then, the SiO ₂ layer was peeled off by fluorine treatment or the like to prepare a membrane blank.

  Thereafter, a resist is applied to the upper surface of the membrane blank, and the pattern determined by the above-described mask bias correction is drawn by an EB drawing apparatus. The pattern area of the pattern on the reticle is 1000 μm × 800 μm. Further, the dimension of the subfield membrane is 1.13 mm × 1.13 mm. After forming a resist pattern, the pattern was transferred to silicon membrane blanks by dry etching to complete a reticle.

  The completed reticle was exposed using an EPL experimental column having an on-axis reduction transfer optical system. The main specifications of the experimental lens barrel were an electron beam acceleration voltage of 100 kV, a reduction ratio of 4 times, and a batch exposure area of 0.25 mm × 0.25 mm.

  As the exposure substrate, a 3-inch or 4-inch Si wafer can be used. First, after a resist was applied on a 3-inch wafer with a thickness of 0.3 μm, a pre-bake treatment was performed and the resist was transferred into an exposure apparatus. As the resist, a chemically amplified negative resist manufactured by Tokyo Ohka Kogyo Co., Ltd. was used.

As described above, in this embodiment, the portion where the pattern density is 25% is used as the reference bias. In this case, the exposure dose is 11 μC / cm ² .
After exposure with an exposure apparatus, the substrate was PEB and developed, and image evaluation was performed with a CD-SEM machine. As a result, it was confirmed that an L / S pattern having a uniform line width of 100 nm was formed over the entire exposed region. Therefore, it was found that the proximity effect could be corrected with high accuracy by the correction method of the present invention.

FIG. 1A is a graph showing a relationship between an exposure dose amount and a mask bias amount. FIG. 1A shows a case where the pattern density is 50%, and FIG. 1B shows a case where the pattern density is 0%. It is a flowchart of the reshape correction method which concerns on embodiment of this invention. It is a figure which shows the image formation relationship in the whole optical system of the electron beam exposure apparatus of a division transfer system, and the outline | summary of a control system.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Electron gun 2, 3 Condenser lens 4 Rectangular aperture 5 Blanking deflector 7 Blanking aperture 8 Deflector 9 Lens 10 Reticle 11 Reticle stage 12 Position detector 15, 19 Projection lens 16 Main deflector 18 Contrast aperture 22 Reflection electron detection 23 Wafer 24 Wafer stage 31 Controller

Claims (3)

A device pattern to be transferred onto the sensitive substrate is formed on the reticle,
The reticle is illuminated with an electron beam,
An exposure method for transferring the pattern by projecting an electron beam that has passed through the reticle onto the sensitive substrate,
Perform mask bias correction including proximity effect correction by giving a dimensional change (reshape) to each element of the reticle pattern in advance.
Here, the dose amount (reference dose amount, reference dose level) to the sensitive substrate surface, which is assumed to be able to realize the expected line width with zero mask bias correction, is the change in the correction line width change with respect to unexpected dose amount fluctuations. An electron beam projection exposure method characterized in that the tolerance as an index is set to a condition of 5 nm / μC / cm ² or less. Perform simulation or test exposure of the appropriate mask bias correction amount to secure the target line width under various conditions with varying dose amounts at the typical pattern density of the pattern to be transferred,
2. The electron beam projection exposure method according to claim 1, wherein the exposure is performed under a condition in which the change in the mask bias correction amount with respect to the change in the dose amount is small. 2. The electron beam projection exposure method according to claim 1, wherein the exposure dose is set higher than the reference exposure dose (reference dose level), and negative mask bias correction is performed accordingly.

Images