JPS61251024A - Exposing apparatus - Google Patents

Abstract

PURPOSE:To improve the alignment precision in addition to the features of global alignment process with high throughput by a method wherein shot position corrected values are measured in terms of the deviation from the printed target position per short of each wafer. CONSTITUTION:A wafer is set up on the first shot position of a wafer chuck and then deflection Cm from the target position is measured as a corrected value Qm, 1 in case of alignment. The second wafer is processed likewise but the mean value Qm, 2 between the corrected value Qm stored in a memory and the second deflection Cm is measured as new corrected value to be stored. Besides, the different (DELTAQm) between the corrected values Qm, 1 and Qm, 2 is also measured. Later the wafers are printed and if the trends of position deviation of patterns printed in the preceding process show resemblances among one another, the fluctuation DELTAQm in corrected values begins to be reduced at the print when several wafers are processed. Moreover when the DELTAQms are checked and detected to be less than specified tolerance T2, the dispersion of position deviations between the target position and the aligned positions can be judged as small.

Description

[Detailed description of the invention] [Field of invention].

The present invention relates to an exposure apparatus suitable for semiconductor manufacturing.
In particular, the present invention relates to an exposure apparatus that achieves both accuracy and throughput.

[Background of the Invention] Conventionally, a device called a stepper is known as an exposure device used in semiconductor manufacturing. This stepper moves the semiconductor wafer step by step under the projection lens.
A pattern image formed on a reticle is reduced using a projection lens, and multiple locations on one wafer are sequentially exposed to light.

Incidentally, recent advances in the degree of integration and pattern miniaturization of semiconductor integrated circuits have been remarkable, and as a result, the overlay accuracy of the reticle and wafer during pattern printing has been reduced to 0.
What used to be about 1μ is now required to be on the order of 0.02μ.

In order to achieve such high accuracy, a so-called TTL-die-pie-die alignment method, in which the reticle and wafer are aligned through a projection lens before each exposure, is effective. However, in this case, there is a disadvantage that throughput decreases. On the other hand, the so-called global
The alignment method is a method in which the entire wafer is aligned only once before exposure of the first shot of one wafer, and thereafter step feeding and exposure are repeated by monitoring only the amount of movement using a laser interferometer, etc. It is. However, although this global alignment method increases throughput, it has the disadvantage that overlay accuracy decreases.

Additionally, as a compromise between the die-by-die alignment method and the global alignment method, there is also a method called zone alignment method, which aligns multiple shots on a wafer at once. In this case, the throughput and overlay accuracy are also intermediate between the die-by-die alignment method and the global alignment method.

[Object of the Invention] In view of the problems in the conventional example described above, the object of the present invention is to
The objective is to achieve both overlay accuracy and throughput in a so-called step-and-repeat type exposure apparatus.

[Summary and Effects of the Invention] In order to achieve the above object, in the present invention, when processing some wafers, the initial wafer is aligned for each shot by a die-by-die alignment method,
Burn. At this time, based on the alignment amount at the position, that is, the deviation from the target printing position according to the step feed l setting value of the printing position, the predicted deviation value at the position, that is, the correction value for step feeding to the position. Calculate. When the number of wafers processed increases and the calculated correction value converges, the process shifts to the global alignment method using this correction value.

In other words, the present invention has the feature of using a self-learning function to shift from the die-by-die alignment method to the global alignment method. If the trends of deviations from the target printing position are similar, throughput can be improved while keeping the overlay error small. In addition, the self-learning function automatically corrects overlay errors caused by the dynamic characteristics of the X-Y stage (even if the model is the same, there are differences between machines), and the number of auto-alignment adjustments is automatically corrected. In addition to reducing the fluctuation of the correction value, even if there are shots for which auto-alignment is not possible, it is possible to align them blindly.

[Explanation of Examples] Hereinafter, the present invention will be explained in detail using the drawings.

FIG. 1 shows the configuration of a semiconductor exposure apparatus according to an embodiment of the present invention. In the figure, 1 is a reticle, 2 is a reticle stage, 3a and 3b are alignment marks of the reticle, 4a and 4b are quick return mirrors, 5 is a field lens, 6 is a projection optical system, and 7 (7x, 7y) is a laser interference Total, 8x, 8y are mirrors for laser length measurement, 9
a, 9b are off-axis microscopes, 10 is a wafer, 11
is a wafer chuck, 12 is a wafer stage, 13a, 1
3b is a relay lens, and the relay lens 13 and quick return mirror 14 constitute a part of a TTL optical system for detecting the relative positional relationship between the reticle 1 and the wafer 10 via the projection optical system 6.

FIG. 2 shows the electrical circuit configuration of the device of FIG.

The device in the figure includes a microprocessor (MPLI) 14,
Read only memory (ROM) 15, random access memory (RAM) 16, reticle stage driver 17
, a wafer stage driver 18, an auto alignment (AA) measurement system 19, and the like. 20 is a keyboard.

M P Ll 14 controls operations such as wafer alignment and step feeding in this apparatus.

The ROM 15 stores a control program for the MPU 14.

The RAM 16 is for temporarily storing various data generated when the MPU 14 executes the above control program, and a part of the RAM 16 includes an alignment correction value table for correcting the step feed amount. Areas such as a global alignment flag for global alignment transition are provided.

The AA measurement system 19 determines the pattern image on the reticle 1 based on the reflection signals from the alignment mark 3 on the reticle 1 and the alignment marks (not shown) for each shot on the wafer 10 detected through the TTL optical system. This is for automatically aligning each shot of the wafer 10.

Next, the operation of the apparatus shown in FIG. 1 will be explained according to the flowchart shown in FIG. When a new wafer is mounted on the wafer chuck 11 by existing means (not shown) in step 31, the wafer is set at the first shot position in step 32. That is, a cough wafer is first put through a mechanical pre-alignment device (not shown) so that the alignment marks 10a, 10b on the wafer 10 are aligned with the microscope 9a,
Roughly, the wafer 10 is placed under the projection lens 6 with an error of ±10 μ or less based on the position detection information of the alignment marks 10a, 10b by the microscopes 119a, 9b. Sent by error. Next, the positions of these shots are detected at two predetermined shot positions on the wafer via the TTL optical system and the AA measurement system 19, and the amount of deviation of the entire wafer is detected from the detection results. In this way, what is detected in two shots is especially the amount of deviation Δθ in the rotational direction of the entire wafer.
This is to measure more accurately. The positional relationship between the shot whose position has been detected and the first shot is known, and the feed amount to the first shot can be determined from this positional relationship and the amount of deviation detected above. Here, in general, rotation errors of the entire wafer are corrected by rotating the wafer chuck 11.

Further, the feeding to the first shot position is performed while measuring the length with the laser interferometer 7.

Next, in step 33, an overlay error (shift) between the printed pattern on the wafer and the reticle image formed by the projection lens 6 at each shot position is detected. In step 34, this deviation is set as D and compared with a predetermined tolerance T1. If it is within the tolerance, proceed to step 37. On the other hand, if it is outside the tolerance, auto-alignment drive is performed in step 35, and after resetting the global alignment flag G in step 36, the process returns to step 33. This auto-alignment drive is performed by the wafer stage 12 in the X-Y direction, and by the reticle stage 2 in the rotational direction θ. In step 37, the global
Check alignment flag G. If global
If the alignment flag G is reset, die-pie-die alignment is currently in progress, and the process proceeds to step 38. On the other hand, if the flag G is set and global alignment is in progress, steps 38 and 39 are executed. Skipping to step 40, printing is performed.

During die-to-die alignment, in step 38, the shot position coordinates after alignment are calculated from the laser interferometer reading and the final overlay error, and the deviation Cl11 of the printed pattern position from the target position in this shot is calculated.
seek. In step 39, the average value of the deviation Cll1 in the m-th shot of each wafer is set as a new correction value QIII.
, p. This new correction value QIII,l) may be calculated from the deviation Cm of the current shot and the correction value Qm,p-1 of this shot up to the previous wafer using the following formula (%). Further, in this step 39, the difference between the old correction value Qm,p-1 and the new correction value Qm,p is calculated as ΔQl. In step 40, the pattern is printed.

In step 41, it is determined whether or not printing has been completed for all shots of one wafer. If the process has not been completed, the process proceeds to step 42; if it has been completed, the process proceeds to step 45. At step 42, the global alignment flag G is examined to determine how to advance the wafer 10 to the next shot. That is, if the global alignment flag G has been reset, alignment is being performed using the die-pie-die alignment method, so step feed is performed to the target position of the next shot in step 43, and the global alignment is performed. When alignment flag G is set and alignment is performed using the global alignment method, the above-mentioned correction value Qm,
Step feed to a position shifted by p. Thereafter, the process returns to step 33 and repeats the above-described process.

Also, when all shots of one wafer have been printed,
The process proceeds to step 45, in which the global alignment flag G is checked, and if the global alignment flag G is set, the process proceeds to step 4.
If the global alignment flag G has been reset, the process proceeds to step 46. Step 4
In step 6, it is checked whether the above-mentioned ΔQIll is within the predetermined tolerance 2 for all shots. Then, if all shots are within the tolerance, a global alignment flag G is set in step 41, and on the other hand, if any shot is outside the tolerance, the process directly proceeds to step 48. In step 48, a check is made to see if printing has been completed for all 10 tons of wafers. If the processing for 10 bits has been completed, the above-mentioned series of processing is completed. If the processing for one lot has not been completed, the process returns to step 31 and the above-described processing is performed on a new wafer.

Next, the processing order starting from the first wafer will be explained.

The wafer is set on the wafer chuck and set at the first shot position. Since the initial wafer does not have a known correction value from the target position, the global alignment flag G
While remaining reset, each shot is aligned and printed using the die-pie-die alignment method. Also,
During alignment, a deviation CI from the target position is determined, and this deviation CI is set as a correction value Qll, 1.

The process for the second wafer is almost the same as for the first wafer, but in step 39, the correction value QIm stored in the memory is
and the average value Ql,2 of the current offset needle position measurement value CIl+ is calculated and stored as a new correction value. Also, the correction value Qm, 1
The difference (ΔQlll) between and Q+n,2 is also calculated. After that, printing is performed, and the same process is performed for all subsequent shots. If the trends in the positional deviation of the patterns printed in the preceding process are similar for each wafer, the variation ΔQIl in the correction value will become smaller after processing several wafers. Therefore, every time one wafer is completed, the ΔQm of all shots is checked in step 46, and if they are all smaller than the predetermined tolerance T2, the variation in the deviation between the target position and the position after alignment between wafers is small. Since this is indicated, the global alignment flag G is set, and the alignment is then shifted to the global alignment method.

In this global alignment method, during step feeding, the amount of movement for moving the wafer from the target position to a position shifted by the above-described calculated correction value is calculated, and based on this, step feeding is performed to perform printing.

[Variation of Implementation Example] In the above-described embodiment, the correction values Q n+, p are calculated as the deviation Cl11 from the target position in the same shot l for each workpiece.
Although the correction value was calculated as the average value of , the correction value may be calculated by weighting according to the number of measured sheets, for example, as in the general formula 0. In this case, if R=1/P, the correction value Q
m and p are the average values of the deviations Cm as in the above embodiment.Also, although the self-learning function in the x-Y axis direction has been described above, the self-learning function in the θ direction can also be provided in the same manner. can.

Note that in the above embodiment, steps 33 and 34, in which it is checked whether the overlay error is within the tolerance at each shot position even after the transition to global alignment, are performed in order to further improve the throughput. This confirmation may be performed every few shots or every few wafers, or may be omitted. Even if there are shots that cannot be auto-aligned due to missing alignment marks or the like, the omitted shots can be aligned by blind hitting.

Also, when alignment is performed using the die-pie-die alignment method, after the printing in step 40, when step feeding to the next shot, the printing target position is sent to the position corrected by the correction value. You can also do this. In this case, the processes of step 42 and step 43 will be deleted.

[Effect of the invention] As described above, according to the present invention, initially
While processing using the alignment method, based on the deviation from the target printing position for each shot of each wafer, we detect the tendency of deviation of each shot (pattern) due to machine differences, etc., as described above, and the tendency of the influence on the wafer due to the wafer process process. It has a self-learning function that calculates a shot position correction value based on this detected position, and after grasping the above-mentioned tendency, it performs global alignment that steps forward to the shot position corrected using this correction value. The global model of high throughput is achieved by
In addition to the features of the alignment method, alignment accuracy can be improved.

[Brief explanation of the drawing]

1 is a schematic diagram of main parts of a semiconductor exposure apparatus according to an embodiment of the present invention; FIG. 2 is a block diagram showing the electric circuit configuration of the apparatus of FIG. 1; and FIG. 3 is a diagram of the apparatus of FIG. 1. 3 is a flowchart for explaining the operation of FIG. 1 reticle, 2 reticle stage, 3a, 3b alignment mark of doubleticle, 4a, 4b: quick return mirror, 5: reduction lens, 6: projection optical system, 7x
, 7y: laser interferometer, 10: wafer, 12: wafer stage, 13: MPtJ, 16: RAM119
:Auto alignment measurement system.

Claims (1)

[Claims] 1. An exposure apparatus that sequentially prints an actual device pattern image on a reticle at each shot position on a semiconductor wafer by step-and-repeat operation, the group having undergone at least one printing process. means for detecting deviations from respective target printing positions of a plurality of pattern images printed on at least the original wafer;
a first calculating means for calculating a printing position correction value for each shot based on the above-mentioned deviations for sequentially processed wafers; a memory means for storing each of the correction values; and a variation of the above-mentioned correction value for each wafer. a step feed control means for step feeding the wafer by setting a position obtained by correcting each of the printing target positions by the correction value as a shot position when at least the variation becomes less than a predetermined value; An exposure apparatus comprising: 2. The means for detecting the deviation from the target burning position is T.
An exposure apparatus according to claim 1, comprising a TL alignment system and a length measuring means for measuring the position of the wafer after alignment.