JP2000003857A - Manufacture of aligner and semiconductor element - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to achieve highly-accurate energy-amount control, by adjusting the intensity of pulse light based on the information updated based on the output of a photodetector. SOLUTION: A laser beam B0 emitted from a pulse laser light source 1 is inputted into a beam expander 14. The beam expander 14 emits the beam as a laser beam LB1. Then, the beam enters a light decreasing part 15 and is attenuated. The attenuated beam LB1' is inputted into an interference fringe decreasing part 17. The interference fringe decreasing part 17 moves the interference fringe on a reticle by changing the incident angle of the beam LB1' into a fly-eye lens 19 and thereby smoothing the fringe. In other words, the uniformity of illuminace is enhanced. Then, the beam passed through the interference fringe decreasing part 17 is returned with a mirror 18 and inputted into the fly-eye lens 19. Therefore, the incident angle of the beam LB1' at the incident surface at the fly-eye lens 19 is changed from time to time.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to control of the amount of irradiation energy to a sensitive object, and more particularly to control of the total amount of irradiation energy when the amount of energy from a pulse oscillation type energy source changes with time. For example, the present invention relates to an energy amount control device suitable for controlling an exposure amount of an exposure device using a pulse laser such as an excimer as exposure light.

[0002]

2. Description of the Related Art In an exposure apparatus (stepper, aligner, etc.) using a pulse laser as a light source used in a lithography process for manufacturing a semiconductor device, a laser beam generally has a variation of about ± 10% for each pulse. On the exposure surface (reticle or wafer), a regular interference pattern due to the coherence of the laser light, as well as scratches in the illumination optical system,
Irregular interference patterns (speckles) in which a large number of light beams having different phases caused by dust, surface defects, and the like are superimposed, may cause uneven illuminance on the exposed surface.

[0003] The above two interference patterns, particularly regular interference patterns, have a significant effect on the control of the pattern line width in the photolithography process of manufacturing semiconductor devices. Therefore, for example, a regular interference pattern or speckle (hereinafter collectively referred to as an interference pattern) is smoothed by a method equivalent to the method disclosed in JP-A-59-226317 or JP-A-1-259533. Is also considered. The interference pattern disclosed in the above publication is smoothed (incoherent) by moving a laser beam one-dimensionally or two-dimensionally (raster scan) at a fixed cycle by a vibrating mirror (galvano mirror, polygon mirror) to spatially coherence. This is to reduce energy consumption. In other words, the interference pattern is moved on the reticle by changing the incident angle of the laser beam to the illuminance equalizing means (optical integrator) for each pulse, and finally the interference pattern is smoothed, in other words, the illuminance This is to improve uniformity. At this time, a plurality of pulses are emitted in synchronization with one-dimensional or two-dimensional scanning by the vibrating mirror. Usually, the oscillation pulse width of an excimer laser is 2
It is extremely short, about 0 nsec. Even if the vibrating mirror is vibrated at about 10 Hz, one pulse of the excimer laser is
During the oscillation period of the oscillating mirror, it behaves as if it were stopped.

Therefore, in order to achieve the smoothing of the interference pattern by a plurality of pulses (uniform illuminance) and the desired control of the amount of exposure, (1) the energy amount per pulse is kept substantially constant during the exposure. (2) It is important to obtain the target exposure amount with an integral multiple of the half cycle of the mirror vibration in one-dimensional or two-dimensional scanning by the vibrating mirror.

[0005]

By the way, the laser light has a variation of about ± 10% for each pulse, and there is a short-term and long-term decrease in the laser density. This phenomenon of reduction in laser density is particularly remarkable in a gas laser, and an active medium (for example, K
(RF, XeCl, etc.), the output decreases with the deterioration of the mixed gas. In general, when the relationship between the applied voltage to the pulse laser light source and the energy amount of a pulse emitted under the applied voltage (hereinafter simply referred to as oscillation energy amount) changes over time, for example, the gas laser If the output of the laser light source decreases with the deterioration of the mixed gas of the active medium, a desired oscillation energy amount can be obtained even if the applied voltage corresponding to the pulse energy amount to be emitted next is determined from the above relationship. I couldn't do that. Generally, a laser light source has a constant applied voltage mode and a constant energy amount mode. In particular, when the output decreases due to gas deterioration in the constant energy mode, two electrodes inside the chamber are conventionally based on the output of the monitor unit that receives a part of the laser beam and detects the energy amount. The device is designed to gradually increase the applied voltage during the period so as to reduce the output reduction. However, in the energy control as described above, a constant oscillation energy amount is always obtained, but it is not possible to obtain a desired oscillation energy amount (that is, an applied voltage corresponding to the oscillation energy amount) over the entire range of a predetermined energy amount control range. In other words, there is a problem that accurate fine adjustment of the oscillation energy amount cannot be performed for each pulse, and high-accuracy energy amount control cannot be achieved.

The present invention has been made in view of the above points, and even if the relationship between the applied voltage and the oscillation energy in the energy generation source changes with time, highly accurate energy amount control can be performed by adjusting the applied voltage. The aim is to obtain an energy quantity control device that can be achieved.

[0007]

According to the present invention, there is provided an exposure apparatus having a light source for emitting pulsed light, the voltage applied to the light source and the light being emitted under the applied voltage. An operation for storing information relating to the relationship with the intensity of the pulsed light, and a photodetector for receiving a part of the pulsed light, and for updating the information stored in the storage based on the output of the photodetector Means for adjusting the intensity of the pulsed light based on the updated information.

In the present invention, data relating to a voltage applied to an energy generation source such as a pulse light source (or an actual charging voltage at the time of energy oscillation) and an oscillation energy amount are fetched for each unit of pulse number or unit time and stored in advance. The relational expression between the applied voltage and the oscillation energy amount stored in the means is sequentially updated by calculation. For this reason, even if the relationship between the applied voltage or the charging voltage and the oscillation energy amount changes over time, the relational expression expressing the change is appropriately updated according to the change over time, so that good energy amount control is always achieved. It becomes possible.

[0009]

FIG. 1 is a block diagram showing a schematic configuration of an energy control apparatus according to a first embodiment of the present invention. According to a predetermined oscillating condition input from the output device 8, the irradiation target 3 is irradiated with pulse energy oscillated from the pulse oscillation type energy generation source 1.

In FIG. 1, an applied voltage control unit 11 controls a high-voltage discharge voltage (corresponding to an applied voltage) of a pulse oscillation type energy generation source 1 and corresponds to an energy amount of pulse energy to be irradiated next. By applying the applied voltage to the energy generating source 1, the energy amount is adjusted for each pulse. The trigger control unit 10
After a predetermined charging time required by the energy generation source 1 has elapsed, an external trigger pulse is sent to the energy generation source 1 to control its oscillation (number of pulses, oscillation interval, etc.). Here, both the trigger control unit 10 and the applied voltage control unit 11
It operates according to a predetermined command output from a main control system 9 (described later). The pulse energy oscillated from the energy generation source 1 may be any coherent laser light, non-coherent pulse light, or pulse energy other than light such as an electron beam.

The energy beam EB emitted from the energy source 1 is split by the beam splitter 2, and most of the energy beam EB passes through the beam and irradiates the non-irradiated object 3. Meanwhile, beam splitter 2
A part of the energy beam EB reflected by the laser beam enters the energy monitor element 4 (for example, a pyroelectric power meter or a PIN photodiode in the case of an excimer laser). A signal corresponding to the energy amount of each pulse is accurately output. The output signal from the monitor element 4 is input to an energy amount monitor 5, where it is converted into an energy amount for each pulse. Therefore, the monitor element 4 and the energy amount monitoring unit 5 constitute an energy amount measuring means of the present invention, and are associated with the energy beam irradiated on the irradiation target 3 in a predetermined relationship, that is, the optical performance of the beam splitter 2. The energy amount of the energy beam uniquely determined by is measured.

The actually measured value measured by the energy monitor 5 (the value may be any value corresponding to the actually measured energy, and need not be the energy itself) is sent to the calculator 6 where each pulse is While the amount of energy for each is sequentially accumulated, the arithmetic unit calculates the voltage applied to the energy source 1 and the amount of oscillation energy under the applied voltage or the actual amount of energy given to the irradiation target 3 (so-called dose). The information on the relationship with the amount is updated (details will be described later). Note that the relationship between the actual energy amount of the energy beam and the sensitivity of the monitor element 4 is determined in advance by the power meter for the monitor element 4 and stored in the memory 7. Also, instead of calculating the integrated energy amount by the arithmetic unit 6,
The energy amount monitor unit 5 may sequentially accumulate the energy amount for each pulse. In this case, the oscillation energy amount and the accumulated energy amount for each pulse are output to the calculator 6.

The arithmetic unit 6 includes arithmetic means for updating the relationship between the applied voltage and the amount of oscillation energy (or dose) according to the present invention. The information stored in the memory 7 and the applied voltage during the oscillation of the previous pulse , And the oscillating energy amount from the energy amount monitoring unit 5, and based on these data, updates the information on the relationship between the applied voltage and the oscillating energy amount according to a predetermined calculation process, and stores it in the memory 7 (details will be described later). . Further, the arithmetic unit 6 includes means for determining an applied voltage in the present invention, and the energy amount monitoring unit 5
The energy amount of the energy beam to be irradiated next is obtained in accordance with a predetermined energy amount control logic (to be described later) based on the integrated energy amount obtained by sequentially integrating the oscillation energy amounts detected by the above for each pulse. Based on the updated relationship between the applied voltage and the oscillation energy amount, the computing unit 6 determines the applied voltage corresponding to the pulse energy amount to be irradiated next by calculation, and sends it to the main control system 9.

The main control system 9 outputs the applied voltage and the oscillation trigger pulse command signal determined by the arithmetic unit 6 to each of the applied voltage control 11 and the trigger control unit 10. The input / output device 8 receives various parameters required for oscillation from the operator, and notifies the operator of the energy amount of the final pulse as necessary. The memory 7 stores parameters (constants), tables, and the like necessary for the oscillating operation, various calculations, and the like input from the input / output device 8.

In FIG. 1, a charging voltage (uniquely corresponding to the applied voltage applied to the energy generating source 1) between the two electrodes during energy oscillation in the energy generating source 1 is detected. The means (charging voltage measuring means of the present invention) is not shown, and the present embodiment will describe a case where there is no charging voltage measuring means. In the case where the charging voltage measuring means is provided in the energy generation source 1, the measurement result by the charging voltage measuring means is taken into the calculating means 6, and information on the relationship between the charging voltage and the oscillation energy amount is updated instead of the applied voltage. Therefore, the description is omitted here.

Next, a method of updating the information relating to the relationship between the applied voltage and the oscillation energy amount, which is the center of the present invention, will be described in detail. FIG. 2 is a diagram showing an example of the relationship between the applied voltage and the energy amount of the emission pulse, and has a quadratic function here. In FIG. 2, the black circles are actual data, and the actual battle is a curve obtained by fitting the data with a quadratic function by the least squares method. Therefore, fitting of a quadratic function by the least squares method will be described.
In this embodiment, the least squares method is used for fitting a quadratic function. However, other than the least squares method described below, for example, a method of maximizing the maximum deviation from data may be adopted. Therefore, if the applied voltage is V and the pulse energy amount is P, the model function is expressed by the following equation.

[0017]

(Equation 1)

Here, a, b, and c are unknowns determined by the least squares method. Next, each data regarding the actual applied voltage and the pulse energy amount is set as Vi and Pi. The subscript i represents the newness of the data, i = 1 is the latest data,
Hereinafter, past data becomes larger as i becomes larger. Further, the evaluation function E is defined as follows.

[0019]

(Equation 2)

Here, γi is a weight (described later) for the data (Vi, Pi). Now, in order to obtain the unknowns a, b, and c that minimize the evaluation function E, the following three-dimensional simultaneous equations may be solved.

[0021]

(Equation 3)

That is, the unknowns a, b, and c can be obtained from the following equations by solving the simultaneous three-dimensional equations according to the theory of the least squares method.

[0023]

(Equation 4)

However, B in the formula (3) is as follows.

[0025]

(Equation 5)

Incidentally, γi is a weight for the data (Vi, Pi). As described above, the relationship between the applied voltage V and the oscillation energy amount P gradually changes in accordance with the change over time (for a gas laser, the deterioration of the mixed gas inside the chamber). Therefore, the weight γi is calculated as γ1>γ2>
By setting γ3..., it is necessary to make the weight of the latest data the heaviest and to gradually reduce the weight as the data becomes older. Therefore, the weight γi is defined as in the following equation.

[0027]

(Equation 6)

Here, ε is a constant satisfying 0 <ε <1, and specifically, is a value determined in advance in accordance with the rate of change with time of the relationship between the applied voltage V and the oscillation energy amount P. . For example, if ε = 0.995, the contribution (weight) to the evaluation function E of the data (Vi, Pi) about the oscillation pulse before about 920 shots is about 1% compared to the latest pulse. Therefore, when the value of the constant ε approaches 1, as is apparent from the above equation (3), a large number of data (Vi, Pi)
), The unknowns a to c are determined, so that it is possible to prevent a decrease in the detection accuracy of the unknowns a to c due to a variation (about ± 10%) in the energy amount for each pulse. However, if the relationship between the applied voltage V and the oscillation energy amount P fluctuates more than the above-mentioned fluctuation due to any cause, there is a large number of past data (Vi, Pi), so that the fluctuation is not delayed. Follow unknowns a to c
Can be difficult to calculate. Conversely, the value of the constant ε is set to 0
, The unknowns a to c are determined from a small amount of data (Vi, Pi), and the followability to the fluctuation of the above relationship is improved, but the influence of the variation of the energy amount for each pulse is great. become. Therefore, in practice, it is desirable to determine the value of the constant ε by experiment in consideration of the balance between the two according to the stability of the energy generation source 1 and the like.

Next, each element of the matrix on the right side of the above equation (3) is determined as follows.

[0030]

(Equation 7)

Then, the applied voltage V and the oscillation energy amount P
The above elements, that is, the parameters C1 to C5 and D1 to D3 are stored in the memory 7 as information relating to the relationship. When new data (Vi, Pi), that is, the latest data (V1, P1) is obtained, the arithmetic unit 6 stores the data in the memory 7
, The parameters C1 to C5 and D1 to D3 are fetched, and the parameters C2 to C5 and D1 to D3 are calculated by the following recurrence formula except for the constant C1 (C1 = 1 / (1-.epsilon.2)).
Shall be updated. As a result, by applying the recurrence formula, the operation time required for updating the information on the relationship between the applied voltage V and the oscillation energy amount P can be shortened, and the arithmetic unit 6 newly stores the updated information in the memory 7. Will do.

[0032]

(Equation 8)

By solving the above-described equation (1) for the applied voltage V and employing only the solution satisfying V ≧ 0, the following equation is obtained.

[0034]

(Equation 9)

Therefore, the parameter C1 updated in the above (6) is appropriately updated every unit pulse number or every fixed time.
Equation (3) is calculated from C5, D1 to D3 to obtain an unknown a
When c is obtained, the relational expression (Equation (1)) between the applied voltage V and the oscillation energy amount P is updated, and furthermore, the oscillation of the next pulse determined based on the predetermined energy amount control logic by Expression (7). This makes it possible to determine the applied voltage V corresponding to the energy amount P.

Next, an example of the energy amount control logic of this embodiment will be described. The target integrated energy amount to be given to the irradiated object 3 by irradiation of a plurality of pulses is S0, and the required energy amount control accuracy is A0 (where 0 <A
0 <1) The variation in the amount of oscillation energy between pulses at the same applied voltage is ΔP (for example, ΔP / P = 10%
If the oscillation energy amount corresponding to the maximum applied voltage Vmax in the above equation (1) is set to Pmax (maximum value), the energy amount P1 of the first pulse energy is obtained.
Is determined from the following equation.

[0037]

(Equation 10)

Further, assuming that the energy amount of the k-th pulse is Pk, the integrated energy amount (measured value) Ij up to the j-th pulse energy is expressed by the following equation.

[0039]

[Equation 11]

Accordingly, the (j + 1) -th pulse energy amount Pj + 1 is determined by the following equation. In addition, min
(Σ, ρ) indicates that the smaller value of σ and ρ is adopted.

[0041]

(Equation 12)

Although the energy control logic has a large dynamic range of the energy control (that is, the adjustment range of the applied voltage V), if the target integrated energy S0 is not so large as compared with the maximum pulse energy Pmax, the energy control logic becomes very small. There is an advantage that a desired irradiation energy amount can be obtained by the number of pulses. Next, the operation of this embodiment will be described with reference to FIG. FIG. 3 is a schematic flowchart illustrating an example of the operation of the present embodiment. In step 100, it is determined whether or not the parameters C1 to C5 and D1 to D3, which are information on the relationship between the applied voltage V and the oscillation energy amount P stored in the memory 7, are to be initialized. This may be determined by the operator or automatically performed according to a predetermined program. In the energy control device shown in FIG. 1, the computing unit 6 actually determines whether or not to initialize based on the oscillation stop time of the energy generation source 1 and the like, and the energy generation source 1 continues to oscillate. ,
Further, even if the oscillation is stopped, when the stop time is short and the parameters C1 to C5 and D1 to D3 in the memory 7 are sufficiently reliable, the process proceeds to step 103. On the other hand, when initialization is necessary, for example, when the energy generation source 1 is started up or its oscillation has been stopped for a long time, the process proceeds to step 101, where the applied voltage V is sequentially changed, and the value of each voltage value is changed. The pulse energy amount P for each pulse oscillated by (1) and (2) is fetched from the energy amount monitoring unit 5, and the relationship shown in FIG. The measurement operation in this step is completed in a short time, and the output of the pulse energy does not decrease due to a change with time during the operation. The operation here is programmed in advance and performed automatically. In the next step 102, the arithmetic unit 6 executes step 1
Based on the data (Vi, Pi) obtained in step 01, the parameters C1 to C5 and D1 to D3 are calculated from the above equation (6) and stored in the memory 7.

In the next step 103, the calculator 6 sets the target integrated energy S0 given from the input / output device 8.
And the energy amount control accuracy A0 stored in the memory 7,
Variation in pulse energy between pulses (ΔP / P),
And the maximum applied voltage Vmax. The variation (ΔP / P) may be a value obtained from the actual data (Vi, Pi) obtained in the previous step 101. The computing unit 6 determines the first pulse energy amount P1 according to the equation (8), and calculates the equation (7) from the unknowns a to c obtained by the equation (3). The applied voltage V1 corresponding to the pulse energy amount P1 is calculated (step 104). Thereafter, the main control system 9 sends a command signal corresponding to the calculation result of the calculator 6 to the applied voltage control unit 1.
1 and the trigger control unit 10, and after the application of the applied voltage to the energy generation source 1 by the applied voltage control unit 11 is completed, charging is completed after a certain period of time.
Sends a trigger pulse to the energy source 1 (step 105). As a result, the energy source 1 emits the energy beam EB with the variation of the pulse energy amount suppressed within (ΔP / P).

Next, the energy amount monitoring unit 5 measures the oscillation energy amount using the monitor element 4, and the calculator 6 sequentially integrates the energy amount for each pulse to calculate the integrated energy amount I0 (step 106). . Then, in step 107, the calculator 6 sets the integrated energy amount I0
It is determined whether or not (actually measured value) satisfies the energy control accuracy A0. That is, it is determined whether or not the integrated energy amount I0 satisfies the following equation (11). If so, the sequence of one energy irradiation on the irradiation target 3 is completed.

[0045]

(Equation 13)

On the other hand, if the above expression (11) is not satisfied, the routine proceeds to step 108, where the data (V
i, Pi) and the parameter C
1 to C5 and D1 to D3 are updated. In this embodiment, 1
A sequence for updating parameters for each pulse is employed. Then, in the next step 109, the arithmetic unit 6
Is to determine the second pulse energy P2 according to the equation (10) and calculate the applied voltage V2 corresponding to the pulse energy P2 from the equation (7).
Return to 05. In step 105, the main control system 9 supplies a command signal to the applied voltage control unit 11 and the trigger control 10 based on the applied voltage V2 in the same manner as the above operation, and further calculates the integrated energy amount I0 by the calculator 6 (step 106). In step 107, it is determined whether or not the integrated energy amount I0 satisfies the above equation (11). If not satisfied, the process proceeds to step 108 again, and the above (11)
Steps 105 to 109 are repeatedly executed until the expression is satisfied, and the energy irradiation operation ends when the expression (11) is satisfied.

Therefore, by repeatedly performing the above operation, even if the relationship between the applied voltage V and the oscillation energy amount P changes with time during the energy irradiation operation, the relational expression expressing the change (formula (3) above) ) Is appropriately updated according to the change over time, so that it is possible to always achieve good energy amount control accuracy. Next, a second embodiment of the present invention will be described with reference to FIGS. FIG. 4 is a plan view showing a schematic configuration of the energy control apparatus according to the present embodiment. In this embodiment, a pulse laser light source that emits pulsed light in the far ultraviolet region is used as the energy generation source 1, and the reticle R Wafer W at a predetermined exposure amount
2 shows a configuration applied to a stepper for transferring data to a printer. FIG.
In FIG. 7, members having the same functions and functions as those of the first embodiment (FIG. 1) are denoted by the same reference numerals.

In FIG. 4, the trigger control unit 10 and the applied voltage control unit 11 send a trigger pulse to the pulse laser light source 1 based on a command from the main control system 9 as described in the first embodiment. It sends an applied voltage corresponding to the energy amount of the pulse light to be irradiated. The pulse laser light source 1 has a band narrowing wavelength stabilizing mechanism composed of an etalon, a dispersing element, and the like at a part between two resonance mirrors disposed at both ends of a laser tube. As a laser light source. Further, a high-voltage discharge is generated between two electrodes provided in parallel along the optical axis of the laser light, so that a deep ultraviolet light having a wavelength such as to expose the resist layer, for example, a KrF excimer laser light (wavelength 2
48 nm).

The laser beam LB0 emitted from the pulse laser light source 1 has a rectangular section corresponding to the arrangement of the two electrodes, that is, a rectangular section having an aspect ratio of about 1/2 to 1/5. . Therefore, the laser beam LB0 is incident on a beam expander 14 (beam cross-sectional shape conversion optical system) in which two sets (irregularities) of cylindrical lenses are combined, and the beam expander 14 emits the laser beam LB0
Of the laser beam LB1 whose beam cross section has been converted into a substantially square beam.

Emission beam LB from expander 14
1 is incident on the light reduction section 15, where the beam light amount (energy) is from 0% (complete transmission) to 100% (complete light shielding)
Between a continuous and stepwise decay. The dimming control unit 12 sends a predetermined driving command to the dimming unit 15 to control the dimming rate (or transmittance). Darkening part 1
The extinction ratio (or transmittance) of 5 is determined by controlling the number of pulses Nsp required to smooth an interference pattern generated on the reticle R or the wafer W and the integrated light amount given to the wafer W to a desired exposure amount control accuracy. Ne required for control
The number of pulses Nex required for the actual exposure determined from
p and the appropriate exposure.

Here, for example, assuming that the light reduction efficiency of the light reduction unit 15 is set in six discrete steps, the light reduction efficiency is selected based on the pulse number Nexp and the proper exposure amount before the start of the exposure. It does not change to another value during the exposure of one shot. In other words, the dimming unit 15 always reduces the light amounts of all the pulse lights by a predetermined amount unless there is a change in the exposure condition for the wafer W (for example, an appropriate exposure amount per shot according to the sensitivity characteristic of the resist). A light amount adjusting mechanism which uniformly attenuates at a light rate and has a relatively low response speed (switching speed of the reduction efficiency) may be used.

The light reduction unit 15 suitable for use in this embodiment.
For example, a method in which six types of mesh filters having different attenuation rates (transmittances) are attached to a turret plate and the turret plate is rotated is adopted. FIG. 5 shows an example of the structure of the rotating turret plate 16 and the six types of mesh filters 16a to 16f. The filter 16a is a mere opening (transparent) portion and has an attenuation rate of 0% (that is, a transmittance of 10%).
0%). Each filter 16a-16f
Are arranged at about six positions along a circle centered on the rotation axis of the rotating turret plate 16 at an interval of about 60 ° so that any one filter is located in the optical path of the substantially square beam LB1 from the expander 2. Is configured.

FIG. 6 shows the rotary turret plate 1 shown in FIG.
6 shows the relationship between the rotation amount and the transmittance. Here, the amount of rotation when the filter 16a is located in the optical path of the beam LB1 is set to zero, and the rotating turret plate 16 is rotated counterclockwise in FIG. In FIG. 6, the rotating turret plate 16 is set at about 60 °.
When rotated by (π / 3), the beam LB at a predetermined rate
1 is dimmed. The rotation amount is 2π (360 ° or 0 °).
In the case of (°), the transmittance is 100% because the filter 16a is selected. Here, as the dimming element attached to the rotating turret plate 16, a dielectric mirror having a different transmittance may be used other than the mesh filter. In addition, two sets of rotating turret plates 16 are provided so as to be relatively rotatable at regular intervals, and for example, the transmittance of the dimming element of the first rotating turret plate is set to 100%, 90%, 80%,
70%, 60%, and 50%, and the transmittance of the dimming element of the second rotating turret plate is 100%, 40%, 30%, and 20%.
%, 10%, and 5%, the combination of the two
A total of 36 transmittances can be realized.

Incidentally, a system in which a predetermined rectangular aperture and a zoom lens system are combined as the dimming section 15 to continuously dimm the light by changing the zoom ratio and the aperture diameter, two glass plates (quartz, etc.) Were held substantially parallel at predetermined intervals,
A method of rotating a so-called etalon, a method of relatively moving two phase gratings or light and dark gratings, or a method of rotating a polarizing plate when linearly polarized laser light is used as exposure light may be used. Absent.

Returning to the description of FIG. 4, the substantially parallel beam LB1 ', which has undergone a predetermined attenuation in the light reduction unit 15, enters an interference pattern reduction unit 17 for smoothing an interference pattern. The interference pattern reduction unit 17 has a vibrating mirror (galvanometer mirror, polygon mirror, or the like) that vibrates one-dimensionally (or two-dimensionally) by an actuator (a piezo element or the like), and the beam LB to the fly-eye lens 19 for each pulse.
By changing the incident angle of 1 ', the interference pattern is moved one-dimensionally (or two-dimensionally) on the reticle and finally smoothed, in other words, the illuminance uniformity is improved.

The beam LB1 'that has passed through the interference pattern reduction unit 17 is one-dimensional (or two-dimensional) at a small angle.
After being turned into an oscillating beam, the beam is turned back by the mirror 18 and enters a fly-eye lens 19 as an optical integrator. Accordingly, the angle of incidence of the beam LB1 'incident on the fly-eye lens 19 at the plane of incidence on the fly-eye lens 19 changes every moment. Here, the fly-eye lens 19 is formed by bundling a plurality of rod-shaped element lenses, and the secondary light source images (here, the partial light beams of the beam LB1 ′) are condensed at the exit end by the number of element lenses. Spots) are formed.

FIG. 7 is a schematic explanatory view showing the relationship between the incident beam of the fly-eye lens 19 and the secondary light source image (spot light) according to the principle disclosed in Japanese Patent Application Laid-Open No. 59-226317. . Each rod lens 19a of the fly-eye lens 19 is a quadrangular prism of quartz glass having convex spherical surfaces formed at both ends. When the beam LBb (parallel light beam) is incident on the fly-eye lens 19 in parallel with the optical axis AX, the exit end of each rod lens 19a of the fly-eye lens 19, or the position where a predetermined amount of the light exits from the exit end into the air, Spot light S
Pb condenses. Although this spot light SPb is shown for only one rod lens in FIG. 7, it is actually formed on all the emission sides of the rod lens irradiated with the beam LBb. Moreover, each spot SPb is focused on the beam LBb substantially at the center of the exit surface of the rod lens. On the other hand, when the parallel beam LBc inclined rightward with respect to the optical axis AX enters the fly-eye lens 19, it is collected as a spot light SPc on the left side of the exit surface of each rod lens 19a. Similarly, the parallel beam LBa inclined leftward with respect to the optical axis AX is collected as spot light SPa on the right side of the exit surface of the rod lens 19a. Therefore,
Due to the one-dimensional vibration of the parallel beam LB1 'by the interference pattern reduction unit 13, all of the plurality of spot lights generated on the exit side of the fly-eye lens 19
9 (optical axis AX) in one direction.

As described above, most of the plurality of beams LB2 forming each spot light formed on the exit side of the fly-eye lens 19 are transmitted through the beam splitter 2 and incident on the condenser lens CL as shown in FIG. Thereafter, they are superimposed on the reticle R, respectively. As a result, the reticle R is illuminated with a substantially uniform illuminance distribution, and the pattern of the reticle R is changed by the projection lens PL onto the wafer W (the illuminated object 3 in the first embodiment) mounted on a stage (not shown).
Is transferred to the resist layer at a predetermined exposure amount. At this time, a plurality of spot lights formed at the exit end of the fly-eye lens 19 are re-imaged on at least the pupil (entrance pupil) Ep of the image (wafer) side telecentric projection lens PL, and a so-called Koehler illumination system is configured. You.

As described above, the interference pattern reduction unit 17 vibrates the beam incident on the fly-eye lens 19 to move the interference fringes generated on the reticle surface or wafer surface by a very small amount. The purpose is to smooth the light and dark fringes transferred to the resist layer, thereby reducing the visibility of interference fringes. In the present embodiment, the fly-eye lens 1 is used to smooth the interference pattern.
The laser beam incident on the laser beam 9 is vibrated. Alternatively, for example, a rotating turret plate diffuser may be rotated in synchronization with light emission of the pulse plate.

Next, a part of the beam LB2 split by the beam splitter 2 is condensed on the light receiving surface of the light receiving element 4 'by the condensing optical system 20. The light receiving element 4 'accurately outputs a photoelectric signal corresponding to the light amount (light intensity) of each pulse of the beam LB2, and is constituted by a PIN photodiode or the like having sufficient sensitivity in the ultraviolet region. The photoelectric signal output from the light receiving element 4 'is input to a light quantity monitor 5', and is converted into an actual light quantity for each pulse by the light quantity monitor 5 '. Therefore, the light receiving element 4 'and the light amount monitor 5' constitute an energy amount measuring means of the present invention, and the energy amount thus measured is sent to the calculator 6, where the light amount for each pulse is sequentially integrated. On the other hand, in the arithmetic unit 6, the measured value (light amount) is used for controlling the exposure amount, updating information relating to the relationship between the applied voltage of the pulse laser light source 1 and the oscillation energy amount, and the trigger pulse oscillated from the trigger control unit 10. Is the basic data of the oscillation control for each shot. Note that the relationship between the actual light amount of the laser beam irradiated on the wafer W by the power meter and the sensitivity of the light receiving element 4 'is determined in advance for the light receiving element 4', and the two are correlated in a fixed relationship. Is stored in Also,
The light quantity monitor unit 5 'sequentially integrates the light quantity for each pulse, and calculates the accumulated light quantity and the light quantity for each previous pulse by a calculator 6
May be output.

The calculator 6 updates the information on the relationship between the applied voltage (or charging voltage) and the oscillation energy (or dose) in the present invention, and updates the information on the pulse energy to be irradiated next. Means for determining a corresponding applied voltage, the role of which is as described in the first embodiment. However, this embodiment is different from the first embodiment in that the interference pattern is smoothed (illuminance uniform). Therefore, the energy control logic in the arithmetic unit 6 in this embodiment is different from that in the first embodiment, and will be described later in detail. Although the charging voltage measuring means is not shown in the present embodiment as in the first embodiment, the arithmetic unit 6 updates information on the relationship between the charging voltage and the oscillation energy amount instead of the applied voltage. Needless to say, such a configuration may be adopted.

The main control system 9 outputs the command of the applied voltage and the oscillation trigger pulse determined by the arithmetic unit 6 to each of the applied voltage control 11 and the trigger control unit 10, and also controls the dimming degree according to various calculation results. And a predetermined command signal relating to interference pattern control is sent to each control unit to control the overall operation of the stepper. Further, it outputs the final integrated light amount (total exposure amount to the wafer W) to the input / output device 8 as necessary.
The input / output device 8 is a man-machine interface between the operator and the main body of the stepper. The input / output device 8 receives various parameters required for exposure from the operator and notifies the operator of the operation state of the stepper.

The memory 7 stores parameters (constants) and tables required for the exposure operation and various calculations input from the input / output device 8 or the sensitivity characteristics of the light receiving element 4 '. In particular, in the present embodiment, while the beam LB1 'oscillates by a half cycle by the interference pattern reduction unit 17, information for determining the minimum number of pulses (Nvib to be described later) necessary for smoothing an excellent interference pattern is provided. It is remembered. Here, the half cycle of the beam refers to the order of spot light SPa → SPb → SPc (or reverse) in FIG.
Corresponds to tilting the beam by the vibration angle α ° in the order of LBa → LBb → LBc (or vice versa). Note that the actual amount of tilt of the oscillating mirror is α from the double angle theorem.
° / 2.

The arithmetic unit 6 has a pulse number Nsp necessary for smoothing the interference pattern stored in advance in the memory 7 and a necessary exposure amount control accuracy for one-shot exposure. Based on the pulse number Ne and the data on the appropriate road light amount S per shot according to the sensitivity characteristic of the resist, the light reduction rate β of the light reduction unit 15 and the average light amount value per pulse (PAV · β) described later. ) Is calculated. The arithmetic unit 6 calculates a target integrated light amount to be given to the wafer W when each pulse is irradiated with the average light amount value, and then calculates the target integrated light amount and the actual measurement amount transmitted from the light amount monitoring unit 5 ′. The difference D from the actual integrated light amount obtained by sequentially integrating the values (energy amounts) is calculated. Then, an applied voltage of the pulse laser light source 1 is calculated based on the difference D, and information on the applied voltage is stored in the main control system 9.
Output to

In other words, the arithmetic unit 6 obtains the light quantity of the pulse light to be irradiated next from the average light quantity value (PAV · β) based on the difference D, and sends it to the main control system 9. Then, based on this correction value, the applied voltage control unit 11 controls the applied voltage of the pulse laser light source 1, that is, corrects the applied voltage by an amount corresponding to the correction value and gives it to the pulse laser light source 1. . Incidentally, the voltage applied to the pulse laser light source 1 and the light quantity (pulse energy) of the emission pulse thereof
An example of the relationship is as shown in FIG.

The main control system 9 includes the pulse laser light source 1
The driving signal is output to the interference pattern control unit 13 so that the pulse emission of the light beam and the deflection angle of the beam by the interference pattern reduction unit 17 are synchronized. Note that this synchronization follows the output of a detector that monitors the beam deflection angle with high precision, and sends an oscillation start and stop signal to the trigger control unit 10 so that the pulse laser light source 1 triggers pulse emission. You may make it output.

Here, the aforementioned dimming section 15 is not limited to the position shown in FIG. 4, but also between the pulse laser light source 1 and the expander 14, or of the resonator mirror inside the pulse laser light source 1. The same effect can be obtained even if it is interposed.
Further, in the case where the above-described method of causing the beam to vibrate at a small angle by the interference pattern reduction unit 17 is not adopted, it may be inserted between the interference pattern reduction unit 17 and the fly-eye lens 19. However, in any case, the dimming unit 15 needs to be placed in a stage before the laser light is incident on the fly-eye lens 19. This is because a dimming element such as a mesh filter often causes deterioration in illuminance uniformity in a beam cross section, and it is necessary to eliminate this by the fly-eye lens 19.

Next, the minimum number of pulses Nsp required to smooth the interference pattern (uniform illuminance) will be described. The smoothing of the interference pattern is disclosed, for example, in Japanese Patent Application Laid-Open No. 1-257327. Therefore, a brief description will be given here. In the above publication, when an optical integrator, particularly an illumination optical system having a fly-eye lens is employed, a plurality of pulsed lights are irradiated while moving an interference pattern formed on a reticle (or a wafer) within a certain range. Therefore, when performing the smoothing, the minimum number of pulses to be irradiated during the movement of the interference pattern by one pitch is limited to a predetermined value in advance, and the pulse light of the number equal to or more than the minimum number of pulses must be irradiated. It uses the principle that it must be.

Now, as shown in FIG. 7, the interference pattern is generated when the spot lights formed by the respective rod lenses of the fly-eye lens 19 interfere with each other. At this time, a case where only the spot lights of two rod lenses adjacent to each other interfere with each other, or a case where three spot lights in the arrangement direction of the rod lenses interfere with each other may be used. What is necessary is just to consider a case where a number of spot lights interfere with each other.

Therefore, theoretically, the minimum necessary for smoothing an excellent interference pattern depends on the number of spot lights which interfere with each other among the numbers of the rod lenses constituting the fly-eye lens 19 in the arrangement direction. Pulse number Nsp, and the minimum pulse number Nv to be irradiated during a half cycle of the beam vibration by the vibrating mirror required for one pitch movement of the interference pattern
ib (Nvib is an arbitrary integer satisfying Nvib ≧ Nsp) is also determined.

For example, when only two adjacent spot lights interfere with each other, the intensity distribution of the interference pattern is mathematically and theoretically a simple sine wave. In order to smooth this interference pattern, two pulses, one pulse before and after shifting the phase difference between the two spot lights by π (movement of a half cycle of the interference pattern), are required. . In general, when n spot lights interfere with each other, theoretically, it is possible to irradiate one pulse while moving the interference pattern by 1 / n period, and to smooth it with a total of n pulses. .

Here, considering the intensity distribution in one direction (for example, Y direction) of an interference pattern generated when one pulse is emitted on the reticle, generally, bright stripes and dark stripes are formed at a predetermined pitch Yp in the Y direction. Are alternately arranged. However, depending on the configuration of the fly-eye lens in two stages, etc., in addition to the basic component of the pitch Yp determined by the arrangement pitch of the rod lenses of the fly-eye lens, the wavelength of the laser beam, and the like, a weak interference fringe whose intensity changes at a finer pitch is generated. It may appear overlapping.
Therefore, in practice, complete smoothing is rarely achieved under the above conditions, and the optimum value of the minimum pulse number Nsp required for smoothing the interference pattern needs to be determined by experiments or the like.

In the present embodiment, based on the minimum pulse number Nsp obtained by experiment, the angle change of the vibrating mirror (not shown) and the light emission pulse interval (frequency) are set so as to satisfy the relationship of N · ΔY ≧ Yp. Set. Then, the interference pattern is sequentially shifted in the Y direction by a minute amount ΔY (ΔY <Yp) on the resist layer for each emission of one or more pulses,
At the completion of the exposure, the interference pattern is smoothed (illuminance uniformity) to obtain a substantially constant illuminance distribution including a minute amount of ripple so as not to affect the accuracy.

Considering the conditions required for uniform illuminance, the following two conditions are given. (1) Within a half cycle of the mirror vibration (a period in which the beam vibration angle changes from 0 ° to α °), pulse light emission of a certain number Nsp or more is performed substantially uniformly. Here, the pulse number Ns
p is the visibility of the interference pattern
y), and the greater the visibility, the greater the value of Nsp. Further, the pulse number Nsp is determined in advance by an experiment such as trial printing, and the numerical value differs between apparatuses having different optical systems. Therefore, N
When pulse light emission of a number smaller than sp is almost equally distributed within a half cycle of a beam vibration angle change (0 ° → α °),
In terms of uniformity of illuminance (illumination unevenness on the image plane), the accuracy does not fall within a desired accuracy.

(2) 1 at a predetermined angle of the vibrating mirror
The average exposure energy per pulse should be substantially constant for any angle within the beam oscillation range (0 ° → α °). The second condition is that the total number of pulses Nexp actually required for one-shot exposure is an integral multiple of the number of pulses Nvib in a half cycle of the beam vibration, and the first pulse light is transmitted through the mirror vibration (0 This is achieved by emitting light at the maximum angle (for example, the angle 0 ° at which the beam LBa in FIG. 4 is obtained) of (° to α ° / 2). When the pulse number Nexp is an integral multiple of the number of pulses in one cycle of the mirror vibration (0 ° to α ° / 2), the first pulse light is transmitted by the mirror vibration (0 ° to α).
Light emission may be started at an arbitrary angle of (° / 2).

From the above, if the average exposure energy per pulse is adjusted to determine the optimum number of pulses so that the above two conditions (1) and (2) are simultaneously satisfied, the illuminance uniformity can be improved. And light exposure control can be made extremely efficient at the same time. Further, if not only the exposure energy but also the oscillation period (oscillation speed) of the oscillation mirror is changed, the number of exposure pulses is not increased more than necessary, which is advantageous in throughput.

The voltage applied to the pulse laser light source 1 is set to a value slightly smaller than the maximum applied voltage Vmax to the pulse laser light source 1 in consideration of the variation in exposure energy for each oscillation pulse. In one-shot exposure, after the first pulse light is emitted under the above set value, the main control system 9 calculates the applied voltage by the computing unit 6 for each of the second and subsequent pulses. The voltage is controlled sequentially.
Therefore, an exposure energy control range by controlling the applied voltage of the pulse laser light source 1 will be described next. The average exposure energy per pulse is PAV (under the dimming rate of the dimming unit 15 is 1), and the variation between pulses of this exposure energy is ΔPAV, and pulse emission is performed N times in one shot exposure. Then, the variation of the actual integrated light amount I with respect to the target integrated light amount PAV · N (appropriate exposure amount S, S = PAV · N) is expressed by the following equation (12).

[0078]

[Equation 14]

As is apparent from the above equation (12), the control ratio of the amount of light by the applied voltage control is 0 <ΔPAV / PAV <
1. Assuming that the number of pulses N is a relatively large integer, あ る 程度 1 ±
(ΔPAV / PAV) / (1−ΔPAV / PAV)}. Therefore, the maximum value of this control ratio ｛1 / (1−ΔPAV / PA)
V) In order to prevent｝ from exceeding the maximum value of the exposure energy control range, the average exposure energy control value set before the exposure is changed to the maximum value of the control range (1−ΔP
(AV / PAV) times or less.

Actually, a desired exposure energy is obtained based on the relationship shown in FIG. 2 such that the average exposure energy PAV is (1−ΔPAV / PAV) times or less of the maximum output of the pulse laser light source 1. What is necessary is just to set the voltage applied to the pulse laser light source 1 (discharge voltage between electrodes). For example, in the case of an excimer laser, (ΔPAV / PAV) is usually about 10%, so that if the exposure energy at the maximum applied voltage of the pulse laser light source 1 is 10 mJ / cm2, the exposure energy PAV will be 9 mJ / cm2 or less. May be set to the applied voltage. In practice, the applied voltage should be set so that the exposure energy PAV is, for example, 5 mJ / cm2 or less in consideration of the phenomenon of laser density decrease (output decrease due to deterioration of the mixed gas) and the life of optical components. Is desirable.

In the present embodiment, the light reduction section 15 is adjusted according to the exposure conditions (type of resist, appropriate exposure amount, etc.) on the wafer W.
, The exposure energy of the second and subsequent pulsed light is finely adjusted only to correct an error in the integrated light amount due to variations between pulses. Therefore, it is not necessary to greatly change the inter-electrode discharge voltage in the pulse laser light source 1, and the dynamic range of the applied voltage control unit 11 can be reduced. Therefore, the relationship between the applied voltage and the exposure energy in one-shot exposure needs to use only a small part of the graph shown in FIG. 2, and the relationship can be sufficiently approximated by a linear expression.

Next, a brief description will be given of the pulse number Ne required to achieve a desired exposure amount control accuracy A (A = 1−I / PAV · N) in one-shot exposure. Since the exposure amount control in the present embodiment adjusts the exposure energy for each pulse and makes the actual integrated light amount I substantially equal to the target integrated light amount PAV · N, the final error of the integrated light amount is obtained. Is the variation in the exposure energy of the final pulse light. Therefore, the variation of the exposure energy must be within the tolerance of the exposure amount control accuracy. That is, the exposure energy PAV must be set to a small value, and the number of pulses N required for one-shot exposure (N = S /
PAV) must be a relatively large number. From this, in the above equation (12), (ΔPAV / PAV) N can be regarded as zero, and if each side is divided by PAV · N to arrange, the exposure amount control accuracy A is expressed as the following equation. .

[0083]

(Equation 15)

Here, when the exposure amount control accuracy A becomes the maximum allowable error in the equation (13), ie,

[0085]

(Equation 16)

, The number of pulses N required to achieve the exposure amount control accuracy becomes the smallest. Thus, the pulse number Ne is expressed by the following equation (15).

[0087]

[Equation 17]

Accordingly, if exposure is performed with at least the pulse number Ne equal to or more than the pulse number Ne represented by the above equation (15),
The final integrated light quantity I is the target integrated light quantity PAV · N,
Control accuracy of ± A (for example, in the case of 1%, A = 0.01) is guaranteed. Next, a method for determining the number Nexp of exposure pulses for one shot will be described. In general, the number of exposure pulses Nexp is Nexp = iNT (S / PAV)
Becomes Note that iNT (ω) indicates that the decimal portion of the real value ω is rounded up and converted to an integer value.

Now, the pulse light oscillated from the pulse laser light source 1 has a predetermined dimming rate β (0 ≦ β)
≦ 1), the light is uniformly attenuated and irradiated onto the reticle R. For this reason, the number of exposure pulses Nexp is required to satisfy the following conditional expression (16).

[0090]

(Equation 18)

In order to achieve the exposure amount control accuracy A described above, the following expression (17) must be satisfied.

[0092]

[Equation 19]

Further, in order to smooth the interference pattern, the number of exposure pulses Nexp must be an integral multiple of the number of pulses Nvib in a half cycle of the vibrating mirror. For this reason, the number of exposure pulses Nexp is expressed by the following equation (18).

[0094]

(Equation 20)

Therefore, the dimming rate β of the dimming section 15 is (15)
From equation (17),

[0096]

(Equation 21)

Is represented by The integer m is (15),
From the expressions (17) and (18), it is expressed by the following expression (20).

[0098]

(Equation 22)

Further, since the extinction ratio β is 1 or less,
From equations (16) and (18), they are expressed as follows.

[0100]

(Equation 23)

As described above, in this embodiment, first, the extinction ratio of the extinction section 15 is determined so as to satisfy the expression (19), that is, the filter of the rotary turret plate 16 is selected. Next, the pulse number Nexp calculated from the equation (16) under the extinction ratio of the selected filter is (1).
It is checked whether the expressions (7) and (18) are satisfied.
If not satisfied, a filter that satisfies the expression (19) and has a smaller extinction ratio is selected so that the number of exposure pulses Nexp satisfies the expressions (17) and (18). When the number of exposure pulses Nexp is determined in this manner, (20) and (2)
It suffices to determine m and Nvib so as to simultaneously satisfy the expression (1).

As an example, the variation ΔPAV / PAV between the pulses of the average exposure energy is reduced by 10% (ΔPAV / PAV =
0.1), assuming that the exposure amount control accuracy A is 1% (A = 0.01), the number of pulses Ne is 12 from equation (15). On the other hand, when the dimming rate β of the dimming section 15 is 1, the average exposure energy PAV is 2 mJ / cm2, the appropriate exposure amount S is 80 mJ / cm2, and the number of pulses Nsp required for smoothing the interference pattern is 50 pulses. Then, the exposure pulse number Nexp becomes 40 pulses from the equation (16), but this pulse number Nexp does not satisfy the equation (18). Therefore, the dimming rate of the dimming section 15 is set to be smaller than 1, and based on this dimming rate, that is, the average exposure energy PAV.beta.
The number of exposure pulses Nexp calculated from the expression 6) is (18)
Try to satisfy the formula.

Here, when it is possible to continuously set the dimming rate β of the dimming unit 15, Ne = 12 and Nsp = 50 pulses, and therefore, based on the equations (20) and (21). m, Nvib are set. At this time, the combination of (m, Nvib) is, for example, (1,50), (1,60), (2,
100), but here, the number of exposure pulses Nexp (Nexp = m
M = 1, Nvib = 5 to minimize Nvib)
The exposure pulse number Nexp is set to 50 by setting to 0. If exposure is performed with Nexp = 50, optimization of the exposure amount and smoothing of the interference pattern can be performed with the minimum number of pulses Nexp. As a result, the dimming rate β of the dimming unit 15 is set to 0.80 according to the expression (16). Further, ΔPAV is ± 0.2 mJ from ΔPAV / PAV = ± 10%.
/ Cm 2, the variation ΔP of the average light amount PAV · β
AV · β is ± 0.160 mJ / cm 2. Accordingly, since the variation of the average light amount of the final pulse light, that is, the error of the final integrated exposure amount can be considered to be about ± 0.160 mJ / cm 2, the exposure amount control accuracy (1%) is sufficiently achieved. You can see that

On the other hand, when the setting of the dimming rate β of the dimming unit 15 is discontinuous (when the dimming unit 15 is a rotary turret plate or the like), first, the rotary turret plate 1 satisfying the expression (19) is used.
6 is selected, and whether or not the pulse number Nexp calculated from the expression (16) under the extinction ratio (for example, β = 0.5) of the filter satisfies the expression (18) Check if. Here, in order to minimize the above Nexp, a filter satisfying the equation (19) is selected from the filter having the largest dimming rate. β
= 0.50 (that is, PAV.beta. = 1 mJ / cm @ 2), Nexp = 80 pulses, and (17), (1)
8) will be satisfied. Thus, the pulse number Ne
Once xp is determined, since Nexp = 80, it is only necessary to determine m and Nvib that simultaneously satisfy the expressions (20) and (21). Here, m = 1 and Nvib = 80.

As is apparent from the above description, when the setting of the dimming rate of the dimming unit 15 is discontinuous, the dimming rate cannot always be set to the optimum value obtained from the calculation.
The number of pulses Nexp becomes large as compared with the case where continuous setting is possible, which may be disadvantageous in throughput. For this reason,
As the light reduction section 15, a light reduction rate can be set continuously, or a light reduction rate can be set finely even if the light reduction rate is discontinuous (for example, two rotating turret plates are used). It is desirable to use a combination thereof.

Next, a method of updating the information relating to the relationship between the applied voltage and the oscillation energy amount, which are central in this embodiment, will be described in detail. As described above, in this embodiment, since the dynamic range of the exposure energy control by the applied voltage control is small, a part of the relationship shown in FIG. 2 is used. Therefore, the relationship between the applied voltage and the amount of oscillation energy can be sufficiently approximated by a linear function. Therefore, when a function is applied to the relationship between the applied voltage and the amount of oscillation energy, it is sufficient to proceed with the linear function. Therefore, the model function in the present embodiment is expressed as follows.

[0107]

(Equation 24)

Here, s and t are parameters obtained by the least square method. In this embodiment, the evaluation function E
Is expressed by the following equation in consideration of the weight γi given to the data (Vi, Pi).

[0109]

(Equation 25)

Now, the unknowns s, which minimize the evaluation function E,
t is, according to the theory of least squares,

[0111]

(Equation 26)

By solving the system of two simultaneous equations,
It can be obtained from the following equation.

[0113]

[Equation 27]

However,

[0115]

[Equation 28]

Here, just as in the first embodiment, the weight γi for the data is set as in equation (4), and
4) Each element of the matrix on the right side of the equation is expressed as the following equation.

[0117]

(Equation 29)

At this time, the above-mentioned elements, that is, the parameters G1 to G3, H1, and H2 are stored in the memory 7 as information relating to the relationship between the applied voltage and the oscillation energy amount. Then, new data (Vi, Pi), that is, the latest data (V
When (1, P1) is obtained, the arithmetic unit 6 fetches the parameters G1 to G3, H1, and H2 from the memory 7, and except for the constant G1 (G1 = 1 / (1-.epsilon.2)), the following gradually increases. It is assumed that G2, G3 and H1, H2 are updated using the formulas. As a result, by applying the recurrence formula, the operation time required for updating the information on the relationship between the applied voltage V and the oscillation energy amount P can be shortened, and the arithmetic unit 6 newly stores the updated information in the memory 7. Will do.

[0119]

[Equation 30]

Accordingly, G2, G3, H1,
H2 is updated, and the equation (24) is calculated, and the unknowns s,
When t is obtained, the relational expression (Equation (22)) between the applied voltage V and the oscillation energy P is updated, and the applied voltage V corresponding to the oscillation energy P of the pulse light to be irradiated next is changed to (2
It can be obtained by the equation 2).

Next, the operation of this embodiment will be described with reference to FIG. FIG. 8 is a schematic flowchart showing an example of the operation of the present embodiment. First, in step 200, the arithmetic unit 6 sets the parameter G stored in the memory 7
It is determined whether or not 1 to G3, H1, and H2 are to be initialized. If the parameters G1 to G3, H1, and H2 are sufficiently reliable, the process proceeds to step 203. On the other hand, when initialization is necessary, for example, when the pulse laser light source 1 is started or its oscillation is stopped for a long time, or immediately after the gas exchange of the active medium in the chamber is performed, the process proceeds to step 201, and , The applied voltage V is changed in several ways within the range of use (the range where the relational expression (22) can be sufficiently approximated by a linear function), and the light amount of each pulse oscillated under each voltage value is monitored. Take in from section 5 '. Thereafter, in step 202, the arithmetic unit 6 executes step 20.
Based on the data (Vi, Pi) obtained in (1), (2)
6) Calculate the parameters G1 to G3 and H1 and H2 from the equations and store them in the memory 7.

In the next step 203, the arithmetic unit 6 calculates the target integrated light amount S from the input / output device 8, the exposure amount control accuracy A, the variation between pulses of the average exposure energy (ΔPAV / PAV), the maximum applied voltage from the memory 7. Vmax and the number of pulses Nsp required for smoothing the interference pattern are fetched.
Of course, the variation between the pulses of the average exposure energy (ΔP
(AV / PAV) may be a value obtained from the data (Vi, Pi) in step 201 as in the first embodiment. Next, the arithmetic unit 6 calculates the extinction rate β (the average light amount P
AV · β), that is, the number of exposure pulses Nexp under the dimming rate of the filter of the rotating turret plate 16 and the selected filter, and the number of pulses Nvi in the half cycle of the vibrating mirror.
b and an integer m are calculated (step 204). continue,
In step 205, the rotary turret plate 16 is rotated to set the dimming rate of the dimming unit 15 to beta, and in step 206, the applied voltage V for obtaining the average exposure energy PAV is calculated from equation (22). Here, the average exposure energy PAV is the exposure energy PAVmax oscillated from the pulse laser light source 1 under the maximum applied voltage Vmax.
It is desirable to set it to (1-ΔPAV / PAV) times or less of (maximum value).

In the next step 207, the pulse counter n and the integrated light amount I are set to Nexp and zero, respectively. In the following step 208, the main control system 9
Determines whether the pulse counter n is zero, and if not, proceeds to the next step 209. In step 209, after the applied voltage of the pulse laser light source 1 is set by the applied voltage control unit 11, a trigger pulse is sent from the trigger control unit 10 to the pulse laser light source 1 to emit one pulse. In the following step 210, a value PAV corresponding to the actual light amount of the pulse light oscillated by the light receiving element 4 '
a is detected and sent to the computing unit 6 via the light amount monitoring unit 5 '. Next, in step 211, the calculator 6 calculates the integrated light amount I, sets the integrated light amount to I = I + PAVa, and sets the pulse counter to (Nexp-1).

Next, at step 212, the average light amount value PAV · β determined at the previous step 204 according to equation (27).
And the difference D between the target integrated light amount and the actual integrated light amount I obtained by the calculator 6 are obtained.

[0125]

(Equation 31)

Subsequently, in step 213, based on the difference D, the light amount PAV ′ of the pulse light to be irradiated next is calculated as follows.
It is obtained by correcting from the average light amount PAV · β determined in step 204 by the equation (28).

[0127]

(Equation 32)

In the next step 214, the arithmetic unit 6 calculates the data of the previous pulse, that is, the applied voltage V applied to the pulse laser light source 1 in step 209 and the light amount PAVA for one pulse detected in step 210. , The parameters G1 to G3, H1, H2 are updated from the above equation (26) and stored in the memory 7. And the next step 2
At 15, the computing unit 6 calculates the light amount P by the equation (22).
An applied voltage corresponding to AV ′ is calculated, and the process returns to step 208. In this step 208, it is determined whether or not the pulse counter n is zero as in the operation described above. If not zero, the process proceeds to step 209 and steps 209 to 21
After performing the same operation as described above in step 5, the process returns to step 208 again, and repeats steps 209 to 215 until the pulse counter n becomes zero.
The exposure operation is terminated at the time point when becomes zero.

Next, the state of exposure control in the apparatus according to the present embodiment will be described with reference to FIG. FIG. 9 is a graph showing the relationship between the number of pulses and the integrated exposure amount when one shot is exposed. Here, the case where exposure is completed by eight pulses is shown. For simplicity of explanation, it is assumed that the dimming rate β of the dimming unit 15 is set to 1. In FIG. 9, a straight line indicated by a two-dot chain line indicates a target value of the integrated exposure amount to be given by the pulse light having the average light amount value PAV · β determined in step 204. Therefore, in the present invention, the light amount is adjusted for each of the second and subsequent pulses so that the exposure amount is controlled in accordance with the target value.

Now, assuming that the first pulse light is emitted with an exposure amount of PAV1 as a target, and that the actually detected exposure amount after emission is PAV1 ', the second pulse light is emitted at the target amount of PAV1'. The difference between the exposure amount 2PAV1 and PAV1 '(2PAV1
-PAV1 ') = PAV2, and light emission is performed. At this time, the applied voltage controller 11 only needs to apply an applied voltage corresponding to the light amount PAV2 to the pulse laser light source 1. Similarly, assuming that the actually measured value of the second pulse light amount is PAV2 ', the third pulse light is (3
Light emission is performed with the light amount set as PAV1-PAV1 '-PAV2') = PAV3.

Therefore, by repeating the above operation, the exposure is completed at the eighth pulse while the deviation from the target line of the two-dot chain line is small. At this time, the final exposure control accuracy (variation of the measured value with respect to the appropriate exposure) is
It is clear that the light amount error (variation) of the eighth pulse light is caused. As described above, in the second embodiment of the present invention, when exposing one shot with a plurality of pulsed lights, the target value of the integrated exposure amount up to the (i + 1) -th shot and the past value (up to the i-th shot) are obtained. The light quantity of the (i + 1) th pulse light to be irradiated next is set based on the difference from the integrated exposure amount. However, if there is any tendency in the variation method for each pulse, the ratio between the target value for each unit pulse and the actually measured value for each unit pulse is averaged over a plurality of past pulses, and the target value is calculated based on this ratio. A new target value may be set by dividing by the average value.

In the second embodiment, the case where the pulse energy oscillated from the exposure illumination light source is coherent laser light has been described. However, the light source of the exposure apparatus emits incoherent pulse light. In the case where the pulse energy other than light such as an electron beam is emitted, there is no need to consider the reduction (smoothing) of the interference pulse. Therefore, the minimum number of pulses (the number of pulses necessary to control the variation in the actual integrated exposure amount given to the wafer W by the irradiation of the plurality of pulse lights with the desired exposure amount control accuracy with respect to the target appropriate path light amount). (Corresponding to the pulse number Ne of the present embodiment) is determined from the fluctuation range of the pulse energy per pulse and the exposure amount control accuracy, and the target value of the pulse energy is determined for each pulse based on the pulse number and the optimum exposure amount. Should be set. That is, the error of the final integrated exposure amount with respect to the proper exposure amount is determined by the light amount error of the final pulse light, so that the variation of the final pulse falls within the allowable error of the exposure amount control accuracy (15). )), The number of pulses is determined, and the average amount of energy per pulse may be set according to the expression (16).

Further, in the exposure apparatus of the above embodiment (FIG. 4), the operation in which the applied voltage control for the pulse laser light source 1 is effective (necessary) is performed in order to obtain information on wafer exposure and the relationship between the applied voltage and the exposure energy. (Step 201 in FIG. 8). Usually, the oscillation of the pulse laser light source 1 in the exposure apparatus
L and a detection reference position (mark detection center position) of an off-axis alignment system that exclusively detects only alignment marks on the wafer W, and a projection position (exposure position) of a projected image of the reticle R. It is also required when measuring the distance, the so-called baseline. Here, the operation of measuring a baseline in an exposure apparatus using a pulse laser as a light source is disclosed in, for example, Japanese Patent Application Laid-Open No. 64-10105, and thus the description is abbreviated, but a plurality of reflecting members (or optical fibers) are used. By illuminating the reference mark provided on the wafer stage from below (inside the wafer stage) with the illumination light transmitted from the pulse laser light source,
The position of the wafer stage when the alignment mark of the reticle R matches the projected image of the reference mark on the wafer stage is detected. Further, the position of the wafer stage when the center position of the mark detection and the reference mark on the wafer stage coincide with each other is obtained using an alignment system, and the interval between these two points is set as the baseline. For this reason,
At the time of the baseline measurement as described above, it is not always necessary to control the applied voltage to the pulse laser light source 1. Therefore, the oscillation mode of the pulsed laser light source 1 can be switched between the applied voltage control mode and the constant energy amount mode (or the constant applied voltage mode) in accordance with the operation mode of the exposure apparatus, so that wafer exposure and data collection in step 201 are performed. It is desirable to use the constant energy mode in operations other than the above.

As described above, in the first and second embodiments of the present invention, information relating to the relationship between the applied voltage (or charging voltage) and the amount of oscillation energy (or dose) is described in (5).
Or as matrix elements shown in equation (25), each element, that is, the parameters C1 to C5, D1 to D3, or G1
Instead of G3, H1, and H2, for example, when updating the unknowns a to c or s, t at the time of initializing the above-mentioned relation (step 102 or 202) (step 108,
Or 214) the unknowns a ′ to c ′ or the ratio of s ′, t ′ may be used as the information, and the correction ratio of the applied voltage necessary to obtain a predetermined pulse energy to the applied voltage at the time of initialization of the above relationship As a function of the pulse energy, and this function may be updated as appropriate.

In the first and second embodiments of the present invention, a case has been described in which information on the relationship between the applied voltage (or charging voltage) and the amount of oscillation energy (or dose) is updated for each pulse. If the pulse oscillation frequency is high and the operation time for each pulse cannot be sufficiently obtained, the information may be updated every unit pulse (for example, every 5 pulses) or every certain time. Absent. At this time, it is desirable to set the update interval of the information to a time sufficiently shorter than the temporal change of the relationship between the two. In the case of the above method, a method (thinning method) that uses only data for each unit pulse number or for every fixed time and does not use other data is expected to be the least computational and effective.

[0136]

As described above, in the present invention, the applied voltage (or charging voltage) and the oscillation energy amount (or dose amount) are appropriately taken in every unit number of pulses or every unit time, and are calculated. We decided to update the relational expression between them. Therefore, the above relational expression is updated successively according to the temporal change in the relation between the applied voltage and the oscillation energy amount, so that good energy amount control is always possible.

Also, when the energy generation source is a pulse laser and the partial gas exchange of the laser or the like is performed along with the exposure operation, the temporal change in the relationship between the applied voltage and the oscillation energy (that is, in the chamber). (Reduction in output due to deterioration of the gas mixture). In particular, in an exposure apparatus, it is convenient for almost completely smoothing the final interference pattern (uniform illuminance).
Even if there is a decrease in the output of the pulse laser light source, etc., one shot exposure should be performed with the minimum necessary number of pulses while optimizing the exposure amount and smoothing the interference pattern with higher accuracy than before. Is possible, and productivity can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of an energy amount control device according to a first embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of a relationship between an applied voltage in an energy generation source and an output (pulse energy) under the applied voltage.

FIG. 3 is a flowchart showing an example of the operation of the energy amount control device according to the first embodiment shown in FIG. 1;

FIG. 4 is a plan view showing a schematic configuration of a stepper provided with an energy amount control device according to a second embodiment of the present invention.

FIG. 5 is a configuration diagram showing an example of a rotating turret plate suitable for being applied to a dimming unit.

FIG. 6 is a diagram showing a relationship between a rotation amount of a dimming element and transmittance when dimming is performed by the rotating turret plate shown in FIG. 5;

FIG. 7 shows a beam incident on an optical integrator (fly-eye lens) and its secondary light source image (spot light).
FIG.

FIG. 8 is a flowchart showing an example of an operation of energy amount control according to the second embodiment shown in FIG. 4;

FIG. 9 is a graph showing a state of exposure amount control according to the second embodiment shown in FIG. 4;

[Explanation of symbols]

1 ... pulse oscillation type energy generation source, 5 ... energy amount monitor unit, 5 '... light amount monitor unit, 6 ...
Arithmetic unit, 7: memory, 8: input / output device, 9 ...
Main control system, 10: trigger control unit, 11: applied voltage control unit, 12: dimming control unit, 13: interference pattern control unit, 15: dimming unit, 16 ..Rotating turret plate, 17 ... interference pattern reduction unit, 19 ...
Fly-eye lens, R: reticle, PL: projection lens, Ep: incident value, W: wafer

────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] June 7, 1999 (1999.6.7)

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] Name of invention

[Correction method] Change

[Correction contents]

Patent application title: Exposure apparatus and method for manufacturing semiconductor element

[Procedure amendment 2]

[Document name to be amended] Statement

[Correction target item name] Claims

[Correction method] Change

[Correction contents]

[Claims]

Claims (1)

[Claims] 1. An exposure apparatus having a light source for emitting a pulse light and transferring a mask pattern onto a substrate, comprising: a voltage applied to the light source; and an intensity of the pulse light emitted under the applied voltage. A storage unit for storing information relating to the relationship with, and a calculation unit for updating information stored in the storage unit based on an output of the photodetector, the photodetector receiving a part of the pulsed light. An exposure apparatus, comprising: adjusting the intensity of the pulse light based on the updated information.