In semiconductor wafer manufacturing the tool and process control is one of the main issues for providing high quality and efficiency of the device output of a manufacturing facil- ity. Typically, process control is performed by measuring some characteristics of a group of wafers, e. g. a lot, after the wafers have been processed in a manufacturing tool. One example is measuring a critical dimension of semiconductor wa-f-e-rs in--a-l--t a-f--r-e-xpo-s.r-e-.'The-lot--is-uualy--p-r-oGe-se-d in a batch. Since processing tools and inspection tools are commonly grouped in different parts of a fab, wafers combined in a lot travel through the fab after the batch is finished with processing. Eventually, the wafers of the lot are meas- ured in an inspection tool and the measured characteristics may reveal problems with the processing tool, such as focus, alignment, coating or development problems in the case of wa- fers being processed in a lithographic cluster. These prob- lems can often be fixed by adjusting tool parameters, which are used to process wafers. Accordingly, having adjusted the processing tool setup using the inspection results as an in- put for correcting tool parameters, the next incoming batch or lot of semiconductor wafers can be processed with improved sets of tool parameters, resulting in an increased production yield.

Usually these measurements in an inspection tool are based on dynamic or static sampling. Typically, not every device in a batch is controlled, rather statistical methods are used to pick out a subsample of the lot for the measurement and the results are feeded back for the new incoming batch devices, which are to be processed. Thereby, it is assumed that the next incoming batch devices being manufactured with improved tool parameters behave in the same manner as the previous batch with respect to the tool parameters, which might not always be the case.

Using this conventional method of statistical process control (SPC) several further disadvantages arise. One is that proc- essed wafers of a batch or lot, that have selectively just been measured in an inspection tool indicating problems with the process tool setup, all have to be moved into rework be- cause they are each most probably affected by these problems.

Therefore, free inspection tool capacity as well as the throughput and yield of the process tool are disadvanta- geously reduced.

Furthermore, problems with single devices might not be cov- ered, if, e. g., always just a first wafer in a lot is af- fected by a problem but it is not selected for inspection by statistical sampling. In this case the origin of a paramater drift in a process tool might not be recognized, or tool pa- rameters might be adjusted inappropriately in reaction to a measured parameter drift. Thus, expensive process tool capac- ity is wasted and the time to manufacture a device is disad- vantageously increased.

It is therefore a primary objective of the present invention to accelerate the throughput of lots or batches of semicon- ductor wafers in a process tool, to increase the wafer yield and to save costs in semiconductor wafer manufacturing.

The objective is solved by a method for exposing at least two semiconductor wafers with a pattern, the exposure being per- formed in an exposure tool, an alignment being performed in response to a first set of alignment parameters, which depend on characteristics of each of said wafers, and a second set of alignment parameters, which account for an exposure tool- offset, comprising the steps of providing a first semiconduc- tor wafer to said exposure tool for exposing said wafer, per- forming alignment of said first semiconductor wafer, with de- termining values for said first set of alignment parameters, exposing the first semiconductor wafer using the combination of said first and second sets of alignment parameters, deter- mining values for a set of parameters representing the over- lay accuracy of said first pattern on said first semiconduc- tor wafer, adjusting values of said second set of alignment parameters to correct for an overlay inaccuracy of said pat- tern, exposing a semiconductor wafer including an alignment in response to said adjusted values.

Using the method according to the present invention the over- lay measurement results of a first semiconductor wafer, which can be determined in several ways, are used to change the tool parameter setup for the following or any other second wafer in a batch. In the context of this invention the align- ment is performed using two sets of alignment parameters. The first set of alignment parameters describes the way, in which an individual wafer is aligned and how the grid of exposure fields is set on the wafer. Examples of such parameters are the chip magnification, shifts in x-and y-direction, grid scaling in x-and y-direction, orthogonality of x-and y- direction, chip rotation, etc. These parameters are used for wafer steppers and scanners, whereby in the case of scanners the chip magnification is adjustable in x-and y-direction separately. Such parameters are determined during an align- ment prior to any exposure of a wafer.

The same parameters but with different values build up the second set of alignment parameters, which accounts for the exposure tool-offset. These offset values are each added to or subtracted from the corresponding parameters of the first set of alignment parameters after the values of said first set have been determined. These tool-offset parameters are used to correct for misalignments during the alignment of the semiconductor wafer.

Having determined the values for a set of alignment parame- ters representing the overlay accuracy of the pattern on the first semiconductor wafer, that just has been exposed or is to be exposed in the exposure tool, the quality of the align- ment performance can be checked. Generally, sets of parame- ters representing the overlay accuracy can be related to sets of alignment parameters. Therefore, according to the present invention the second set of tool-offset alignment parameters is adjusted in its values to correct for an overlay inaccu- racy of the pattern that either has been exposed or is to be exposed in case that the overlay data reveal values beyond a tolerance level.

The objective can further be solved using a similar method, where the tool-offset parameters can be adjusted using al- ready at least one semiconductor wafer. With each further wa- fer that is aligned and optionally exposed, the tool-offset parameters can even be further refined. The corrected tool- offset parameters necessarily will lead directly to an im- proved tool-setup, due to which the yield is advantageously increased. Moreover, the method according to the present in- vention may be combined with a method similar to the advanced process control in the sense that the present method may be repeated a few times until there repeatedly is no adjustment of the values of the second set of alignment parameters nec- essary, after which the determination step for the values of the set of parameters representing the overlay accuracy can be skipped. Thus, the application of the method of the pres-

ent invention leads to a situation, where costs can be saved due to a reduced amount of inspection time for wafers after exposure. Furthermore, the increased yield advantageously leads to a decreased throughput time.

In a another aspect of the present invention the case of cal- culating the parameters reflecting the overlay accuracy using a formula and adjusting the second set of alignment parame- ters, i. e. the tool-offset alignment parameters, is each per- formed prior to exposing the semiconductor wafer. This aspect of the method according to the present invention has the ad- vantage, that the wafer is not exposed, until the sets of alignment parameters are arranged with values, that provide sufficient quality for the eventual exposure. Therefore, the yield is still further increased, the throughput time even further decreased and the costs are advantageously reduced.

In a further aspect the step of determining the values for the set of parameters representing the overlay accuracy is considered to be performed by calculating the values from those alignment parameters, that have been determined in the exposure tool alignment, i. e. the first set of alignment pa- rameters. This calculation can be based on the fact, that an overlay parameter measured with an inspection tool can as well be derived or calculated with a mathematical formula comprising a combination of alignment parameters of the first set, each of them being supplied with a coefficient, which depends on the exposure tool. This determination is found to recover a sample of measurements of corresponding overlay pa- rameters to a high degree of accuracy. The coefficients have to be determined once for a complete tool-offset and can then be reused until the next complete setup of the corresponding exposure tool is performed. Typically, this is a duration of weeks or months.

Therefore, the overlay parameters are known immediately after the alignment in the exposure tool when the determination of

the corresponding alignment parameters is finished. This preferentially happens by means of an alignment measurement.

Since the values for the set of overlay parameters are known from the calculation, a measurement with an inspection tool can be skipped resulting in a further reduced wafer through- put time. The second set of alignment parameters, i. e. the tool-offset alignment parameters are then adjusted according to the set of calculated overlay parameters.

In a further aspect the mathematical formular for calculating the values of the set of parameters representing the overlay accuracy is considered to be a linear function with the coef- ficient in front of each alignment parameter being dependent on the exposure tool employed. A high degree of accuracy for recovering the overlay measurement results with a linear function is found, resulting in an easy and straightforward calculation step. Only a few coefficients are involved in this case.

In a further aspect the stopping of the batch queue as a re- sult of values for the set of the parameters representing the overlay accuracy exceeding threshold values corresponding to tolerance levels is considered. After the batch or lot proc- essing is stopped system maintenance can be conducted in re- sponse to a warning signal, that has been issued in this case. After having reset said exposure tool processing of semiconductor wafers can be continued.

In a further aspect a second determination of the values of the set of parameters representing the overlay accuracy of the wafer pattern is considered. A warning signal is issued, when one of the parameters of the set of parameters repre- senting the overlay accuracy of the pattern increases beyond a pre-determined tolerance level. A detailed investigation of the parameters and the reason why the tolerance level has been exceeded, can than be performed by remeasuring values of the overlay parameters, which priorily have been calculated

using the formular from the first set of alignment parame- ters.

In a further aspect it is considered that each of the at least two semiconductor wafers is processed in a next manu- facturing step after exposure without being inspected in an overlay tool.

In a further aspect it is considered, that also the values for the set of tool-offset alignment parameters are calcu- lated from the set of parameters representing the overlay ac- curacy using a linear formula. This feature provides a fast feedback for the alignment procedure from the overlay parame- ter determination.

In a furhter aspect, a neuronal network is built by means of the results of the calculation of the values for the set of parameters representing the overlay accuracy performed by a control unit 100 in comparison with the measurement of the values for the set of parameters representing the overlay ac- curacy in the inspection tool 40, for adjusting the formula for calculating said values for the set of parameters repre- senting the overlay accuracy.

The invention will be better understood by reference to the following description of embodiments of the invention taken in conjunction with the accompanying drawings, wherein figure 1 schematically shows the wafer process flow in li- thography as well as the alignment feedback accord- ing to an embodiment of the present invention, figure 2 displays the same as in figure 1, but for an em- bodiment comprising an offset adjustment prior to exposure according to the present invention,

figure 3 displays a diagram of grid scale data in x-and y- direction measured as a wafer alignment parameter during an alignment for the exposure of DRAM wafers within a period of 3 months versus exposure date, figure 4 shows a diagram with used values for the tool- offset parameter chip magnification according to prior art (diamonds) versus date, overplotted with calculated offset values (squares) for the same pa- rameter according to an embodiment of the present invention, for the exposure of the same DRAM wafers as displayed in figure 3, figure 5 shows a plot of the overlay parameter chip magnifi- cation representing the overlay accuracy of a pat- tern as measured with an KLA overlay inspection tool, with and without applied tool-offset (dia- monds and circles, resp.) according to prior art, and as being calculated (squares) according to an embodiment of the present invention, for the same DRAM wafers as displayed in figure 3, versus date.

The control of the wafer flow through processing steps 20, 30,40,90 according to an embodiment of the method of the present invention is shown in figure 1. There, thick arrows denote the physical wafer flow, while narrow arrows denote a flow of information, either signal and/or the data corre- sponding to the sets of alignment parameters. A semiconductor wafer batch queue 10 is provided for exposure. Prior to being exposed in an exposure step 30 a first semiconductor wafer 1 of the batch undergoes an alignment step 20 on the wafer stage inside the exposure tool 35.

After the wafer has been exposed the wafer alignment parame- ter data, which are measured during the alignment step 20, are examined in a control unit 100. Thereby, using the wafer alignment parameter data it calculates the relevant inspec-

tion overlay data, i. e. values for. the set of overlay parame- ters representing the overlay accuracy of the pattern that is structured on the first semiconductor wafer 1.

For example, the chip magnification expressed as an overlay parameter is calculated using a linear formula chip _ mag (overlay) = 1.24 + 0. 50-scale-x + 0. 50-scale _ y + chip _ mag (align) where chip _ mag (overlay) denotes the chip magnification as a parameter representing the overlay accuracy, chip mag (align) denotes the chip magnification as a wafer alignment parameter as determined during the alignment step, scalex and scale y denote the grid scaling in x-and y-direction, respectively.

The coefficients 1.24,0.50 and 0.50 are exposure tool- dependent and are fixed for the duration of a typical tool setup.

The calculated overlay parameter values are then compared each with parameter dependent tolerance levels according to the overlay accuracy specification provided for the product.

If one the one side the calculated overlay data provide suf- ficient results, i. e. within a tolerance range, the wafer is forwarded to the next process step 90. On the other side if the overlay data reveal calculated values that increase be- yond said tolerance levels, the wafer is forwarded to an in- spection tool 40 for performing a second determination of the overlay parameters, i. e. a measurement. If this measurement actually reveals insufficient overlay quality the wafer is sent into rework 50, while it can be forwarded to the next process step 90 in case the overlay measurement reveals oppo- sitely sufficient overlay accuracy. In the latter case a re- determination of the calculation formula used in control unit 100 for determining the values for the set of overlay parame- ters from the values of the set of wafer alignment parameters might be appropriate. Nevertheless, the new alignment offset

can be determined by control unit 100 according to measure- ment results of the inspection tool 40 as described below.

While a warning signal 105 is issued by the control unit 100 for controlling the further processing of the semiconductor wafer 1 as described above, also adjusted values for the set of tool-offset parameters are determined by control unit 100 for the alignment procedure. These values are to be added to the values of the set of wafer alignment parameters that are determined during the alignment step 20. A second semiconduc- tor wafer 2 is then aligned and exposed in the exposure tool 35 using the adjusted tool-offset parameter values as indi- cated by the dashed line in the exposure unit 30 of figure 1, that correct for a possible misalignment derived by control unit 100 for the first semiconductor wafer 1. The alignment data may also be received by the control unit 100 prior to exposure, but the feedback to the alignment step 20 is only provided for the alignment of the following second semicon- ductor wafer 2. The exposure of the first semiconductor 1 is performed using the originally given values for the set of tool-offset alignment parameters (the first set of alignment parameters).

Another embodiment of the present invention is shown in fig- ure 2. It deviates from the embodiment shown in figure 1 by a closed loop feedback from control unit 100, which performs the wafer alignment parameter determination, the overlay pa- rameter calculation and the tool-offset parameter adjustment already after the alignment 20 prior to wafer being exposed in exposure step 30. Semiconductor wafer 1 is aligned, while values for the alignment parameters are received by control unit 100. As in the previous embodiment control unit 100 then calculates values for the set of overlay parameters repre- senting the overlay accuracy. Because the semiconductor wafer 1 is not yet exposed and still performing the alignment step 20, the tool-offset alignment parameters can be adjusted in- situ in case the overlay accuracy-and thus the tool-offset

- can still be improved for the same first semiconductor wa- fer 1.

After being exposed an overlay control can optionally be is- sued by means of statistical process control (SPC) 101 for performing a statistical verification of the overlay results calculated due to the control unit 100.

A comparison of the method according to the present invention with prior art is indicated in figures 3-5. As an example the tool-offset parameter for chip magnification is derived using wafer alignment parameters like grid scale in x-and y- direction as input for the calculation. The grid scaling in both directions for a measurement period of several wafers during a period 3 months is shown in figure 3. There, it can easily be seen, that e. g. the grid scaling in y-direction is scattering around each of the two distinct values of about 1.3 and 3 ppm.

It has been found, that the altering grid scaling expressed as an alignment parameter has an influence on the chip magni- fication expressed as a parameter representing the overlay accuracy. Therefore, the chip magnification-this time ex- pressed as an alignment parameter-can be provided with an offset to correct for the jumps in the grid scaling data.

In figure 4 the corresponding tool-offsets, that have been used to provide a conventionally optimum overlay accuracy, are displayed as diamonds. They were fixed at five distinct values for the tool-offset alignment parameter chip magnifi- cation, which were: initially 1.3 ppm, then 2 ppm, followed by 1 ppm, followed by 3 ppm and finally 1.9 ppm during the same period of 3 months, and for the same exposure tool and wafers as in figure 3.

In figure 5 the corresponding diamonds reveal the final out- come for the overlay accuracy with respect to chip magnifica- tion. This prior art case is characterized by a large scatter

between +2 ppm and-2 ppm in chip magnification. The dis- tinctly set tool-offset parameter chip magnification (dia- monds in figure 4) provided a linear shift (offset) in figure 5 from the data represented by the circles, which represent the measured chip magnification overlay parameter as being measured without an offset applied during the alignment.

Thus, the distinctly set offset parameters shifted the meas- ured overlay data advantageously into the tolerance window for overlay, which, e. g., may be represented by a range for chip magnification between-1 ppm and +1 ppm. But this hand- set chip magnification offset cannot impede the large scatter that is obvious from figure 5 from the diamonds as well as the circles. In this prior art case, the amount of rework therefore is disadvantageously large.

Using the embodiment of the present invention, e. g. according to figure 2, the tool-offset parameter chip magnification may be calculated individually for each wafer. A linear formula is used, given by chip_ mag (of f set) = 4. 21-0. 44-scale-x-0. 55-scale-y, whereby chip _ mag (Offset) denotes the tool-offset parameter chip magnification, and scalex and scale_y denote the wafer alignment parameter grid scaling in x-and y-direction, re- spectively. The coefficients 4.21,0.44 and 0.55 have been determined previously in the context of a change of the tool setup, e. g. after a maintenance.

The ideal offset calculated using this formula is displayed as a set of squares in figure 4. The data scatter around the conventionally derived tool-offset chip magnification. But because this scatter originates from a correction of the scatter in the chip magnification overlay data, the quality of the overlay results has strongly increased as can be seen by the squares in figure 5. All values for the overlay chip magnification are within the tolerance level of-1 ppm to +1 ppm. Thus, the yield is advantageously increased and the throughput time efficiently reduced. Moreover, free inspec- tion tool capacity is gained, since the overlay control in most cases can be skipped using the method according to the present invention.

List of reference numerals: 1 first semiconductor wafer 2 second semiconductor wafer 10 batch or lot of wafers 20 alignment 30 exposure step 35 exposure tool 40 overlay inspection 50 rework 90 next process step 100 control unit 101 statistical process control 105 warning signal