JP2006013178A - Method and system for exposure - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve matching deviation precision in an actual exposure stage by applying a statistical model and predicting a matching deviation correction value. <P>SOLUTION: An exposure method has a calculating means (S13) of calculating a 1st matching deviation value (optimum matching deviation value) from existent data, a predicting means (S14) of performing calculation processing for the calculated 1st matching deviation value with the statistical model and predicting a 2nd matching deviation correction value (matching deviation value used for this lot) to be used in an exposure processing stage, and an exposure processing means (S16) of performing the exposure processing stage by using the predicted 2nd matching deviation correction value. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an exposure method and an exposure system in an electronic component manufacturing process.

  In the lithography process of electronic parts, especially semiconductor manufacturing processes, the overlay accuracy required for pattern formation is ensured, and furthermore, photoresist coating is performed with a minimum test exposure (advance exposure) to determine exposure conditions. There is a demand for maintaining the processing capabilities of the developing device, the exposure device, and the misalignment inspection device to the maximum.

  Usually, in the lithography process of a newly introduced product, for example, the first lot is subjected to advance exposure, and the pattern misalignment error in the misalignment inspection apparatus is changed to shift component, wafer magnification component, wafer rotation component, shot magnification. The least square fitting of the component, shot rotation component, etc. is input to the exposure apparatus for correction, and then the actual exposure process is executed to form a pattern.

  For subsequent lots in the same lithography process, for example, the exposure process is executed based on the optimum misalignment correction value, and the preceding exposure is omitted. In such a lithography process, a method for accurately calculating a misalignment correction value using existing data has been developed, and for example, a method for automating advance exposure insertion and removing an abnormal value has been reported (for example, Patent Documents). 1).

However, as semiconductor devices are further miniaturized, it is an important issue to accurately predict misalignment correction values and feed them back to subsequent lots.
P2001-257159 (page 13, FIG. 4)

  An object of the present invention is to provide an exposure method and an exposure system used in a manufacturing process of an electronic component, and an object of the invention is to predict a misalignment correction value and improve a misalignment accuracy of a pattern obtained by exposure processing.

  According to a first aspect of the present invention, as an exposure method, a calculation unit that calculates a first misalignment correction value from existing data, and the calculated first misalignment correction value are calculated using a statistical model, A predicting unit that predicts a second misalignment correction value used in the exposure processing step; and an exposure processing unit that executes the exposure processing step using the predicted second misalignment correction value. To do.

  According to a second aspect of the present invention, as an exposure method, calculation means for calculating a first misalignment correction value from existing data, and the calculated first misalignment correction value as a statistical model Prediction means that performs calculation processing using a regression model and an autoregressive moving average model, respectively, and predicts a second misalignment correction value used in the exposure processing step, and the predicted autoregressive model and autoregressive moving average model The exposure processing step is executed using a selection unit that selects a second misalignment correction value with high accuracy among the second misalignment correction values and the selected second misalignment correction value. Exposure processing means.

  According to a third aspect of the present invention, as an exposure system, a data server that stores data, a first misalignment correction value is calculated from the data stored in the data server, and the calculated first A misalignment correction value is calculated using a statistical model, and a system control unit for predicting a second misalignment correction value used in the exposure processing step and the exposure processing step are executed using the second misalignment correction value. It has an exposure processing apparatus and a misalignment inspection apparatus that measures misalignment values of the lot subjected to exposure processing.

  According to the present invention, by applying a statistical model, it is possible to predict a misalignment correction value and improve misalignment accuracy in an actual exposure process.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram showing the exposure system of this embodiment. In this embodiment, a silicon substrate is used as a semiconductor substrate, and in order to manufacture the same product, the exposure system is operated with a so-called lot, which is a group of, for example, about 20 silicon substrates as a group.

  The data server 13 includes a lot history database 14 that manages information such as manufacturing process and history of product lots, and information such as misalignment correction values input to the exposure device 18, and product lots measured by the misalignment inspection device 20. Are composed of two blocks of a misalignment inspection database 15 for managing information such as a misalignment value and a measurement coordinate value in overlay inspection.

  On the other hand, the system control unit 10 includes a data transfer unit 12 and a calculation unit 11. The data transfer unit 12 exchanges data with the exposure device 18 of the exposure device group 17, the misalignment inspection device group 20 of the misalignment inspection device group 19, the data server 13, and the arithmetic unit 11 in the system control unit 10. Is called.

  In addition, the calculation unit 11 calculates the optimum misalignment correction value with the existing data that is the first misalignment correction value based on the past lot data or the data of each semiconductor substrate constituting the lot, and uses the obtained existing data. This lot, which is the second misalignment correction value, using statistical methods such as a moving average (MA) model, an autoregressive (AR) model, and an autoregressive moving average (ARMA) model based on the optimal misalignment correction value of , Misalignment correction value setting, determination of the best statistical model, instruction to the exposure device 18, the misalignment inspection device 20, the data server 13, and the data transfer unit 12 are executed.

  FIG. 2 is a flowchart showing the exposure procedure in this embodiment. First, the process starts (S11), and the lot information of the main lot entering the exposure process is recognized (S11). For example, a lot ID number of a so-called lot box in which the silicon substrates in the lot are stored together is read by a barcode reader or the like.

  The read lot ID number is sent to the calculation unit 11 via the data transfer unit 12, and in response to an instruction from the calculation unit 11, the ID number is referred to and the following existing data is searched from the lot history database 15 ( S12). For example, the misalignment inspection data is searched for lots having the same product, the same process, and the same exposure apparatus used for the exposure process.

  This history data includes the product name, process name, exposure apparatus when exposing the lower layer, its exposure conditions, matters, exposure date and time, and the like. As a result of this search, the optimum misalignment correction value of the corresponding lot is read from the existing data of the corresponding lot, for example, to all the corresponding lots in the time series and read out to the calculation unit 10.

  The optimum misalignment correction value in the existing data which is the first misalignment correction value is the misalignment correction value set when the lot is exposed and the misalignment correction value obtained from the actually measured misalignment amount. And determine the difference (S13).

  Here, the misalignment correction value is calculated as follows. That is, the actual misalignment error measured by the misalignment inspection apparatus is corrected by least square fitting using, for example, a shift component, a wafer magnification component, a wafer rotation component, a shot magnification component, a shot rotation component, and the like.

In calculating the misalignment correction value obtained from the actually measured misalignment amount, the following equations (1) and (2) are used.

  The kn terms (for example, n = 1 to 20) and the Kn terms (for example, n = 1 to 20) correspond to the misalignment correction values obtained from the actually measured misalignment amounts. The kn term and the Kn term are set to predetermined values depending on the product and the process. In this embodiment, for example, the 0th order term is a shift component, and the 1st order term is a magnification component and a rotation component. Equation (1) is used to calculate a shot component which is an exposure region unit in a semiconductor substrate, and equation (2) is used to calculate a semiconductor substrate (Wafer) component.

  The optimum misalignment correction value in the existing data is calculated using the misalignment correction value obtained from the measured misalignment amount as described above and the misalignment correction value set at the time of exposure.

  Next, based on the optimum misalignment correction value in the existing data, the misalignment in this lot is predicted by a statistical model, and the misalignment correction value to be applied to this lot as the second misalignment correction value is calculated ( S14). By using this method, it is possible to improve the accuracy of the misalignment correction value applied to the present lot, and to reduce the re-exposure of the exposure process.

  The misalignment correction value for this lot is calculated using the following MA model, AR model, and ARMA model as statistical models.

First, calculation with the MA model is shown by equation (3). The optimum misalignment correction values for existing m lots in the same process (j-1 lot, j-2 lot, j-3 lot,..., Jm lot) are y _j-1 , y _j-2,. .., y _jm Further, if the misalignment correction value in this lot input to the exposure apparatus is y _j , the following equation is obtained.

Next, calculation in the AR model is shown. In the AR model, the misalignment correction value for this lot is calculated by fitting an equation for an existing p lot as shown below. First, according to the p-order model, the optimum misalignment correction values for the existing q lots (p lot,... Q lot) in the same process are set as x ₁ , x _{2 ,.} When the misalignment correction value in the input lot is x _j , the following equations (4) and (5) are obtained.

y _j is the difference between the misalignment correction value xj in this lot and the average of the optimum misalignment correction values for the existing lots. That is, by obtaining yj, xj can be calculated using equation (4). In order to obtain yj, a p-order AR model of the following equation (6) is assumed.

The residual sum of squares Qp is set as shown in equation (7), a ₁ , a ₂ ,..., A _p are obtained so that Qp is minimized, and the determinant of equation (8) is solved.

Substituting a ₁ , a ₂ ,..., A _p obtained by the equation (8) into the equation (6) to obtain y _j . Substituting this y _j into the equation (5), x _j is calculated.

Next, calculation by the ARMA model is shown. In the ARMA model, the misalignment correction value for this lot is calculated as follows by assuming a formula for the existing p lot and using that formula. First, the optimal misalignment correction values of existing q lots (p lot,... Q lot) in the same process are set to x ₁ , x _{2 ,.} If the misalignment correction value in this lot is x _j , the following equations (9) and (10) are obtained.

y _j is the difference between the misalignment correction value x _j in this lot and the average of the optimum misalignment correction value of the existing lots minute. That is, by obtaining y _j , x _j can be calculated using equation (9). In order to obtain y _j , a p-order ARMA model shown in the following equation (11) is assumed.

The residual sum of squares Q _p is set as in equation (12), a _p1 , a _p2 ,..., A _pp are obtained so that this Q _p is minimized, and the determinant of equation (13) is solved.

(13) determined by the formula _{a p1, a p2,} ···, substituted in the a _pp (11) equation to determine the y _j. Substituting this y _j into the equation (10), x _j is calculated.

  As described above, in the case of the MA model, m specified in advance is input for each product and process, and in the case of the AR model and ARMA model, p and q specified by the product and process are input and used for this lot. A misalignment correction value to be calculated is calculated (S14).

  Next, the misalignment correction value used for the main lot determined as described above is input to the exposure apparatus as data (S15).

  Subsequently, an exposure process is executed for this lot (S16). For example, this lot configured with 20 to 25 silicon substrates as a unit is introduced into the exposure apparatus. The silicon substrates introduced into the exposure apparatus are exposed one by one. The exposure to the silicon substrate is performed using one chip or two chips in the silicon substrate as one shot unit.

  After the exposure is completed, for example, a silicon substrate is extracted from the lot, the silicon substrate is introduced into a misalignment inspection apparatus, and misalignment measurement is performed (S17).

  Next, the misalignment measurement data is transmitted to the calculation unit 11 via the data delivery unit 12, and the calculation unit 11 compares the obtained misalignment measurement data with a predetermined standard of this lot, and enters the standard. It is determined whether it is out of specification or not (S18). Regardless of the standards, the obtained misalignment measurement data is sent to and stored in the lot inspection database. The data within the standard is stored in a predetermined file for use in calculating the misalignment correction value of the subsequent lot (S20). On the other hand, the data outside the standard is used for analyzing the cause, for example, It is stored in another file (S19), and the processing of this lot ends (S21).

  In this embodiment, by applying a statistical model, it is possible to predict misalignment correction values and improve misalignment accuracy in the actual exposure process.

  In addition, it is possible to reduce the necessity of performing the preceding process for each lot, and to reduce the occurrence of redoing due to a large misalignment, thereby preventing a reduction in productivity.

  FIG. 3 is a flowchart showing an exposure procedure in this embodiment. The present embodiment basically takes the same procedure as that of the first embodiment. The difference from the flowchart in the first embodiment is that a step of selecting an optimum model from the AR model and the ARMA model is added.

  The exposure system in the present embodiment is the same as the exposure system 1 shown in FIG. 1 in the first embodiment 1, and the description thereof is omitted.

  As shown in FIG. 3, the exposure procedure starts first (S30), and the lot information of the lot that enters the exposure process is recognized (S31). For example, a lot ID number of a so-called lot box in which the silicon substrates in the lot are stored together is read by a barcode reader or the like.

  The read lot ID number is sent to the calculation unit 11 via the data transfer unit 12, and in response to an instruction from the calculation unit 11, the ID number is referred to and the following existing data is searched from the lot history database 15 ( S32). For example, the misalignment inspection data is searched for lots having the same product, the same process, and the same exposure apparatus used for the exposure process.

  This history data includes the product name, process name, exposure apparatus when exposing the lower layer, its exposure conditions, matters, exposure date and time, and the like. As a result of this search, the optimum misalignment correction value of the corresponding lot is read from the existing data of the corresponding lot, for example, for all the corresponding lots, by reading the time series in the calculation unit 11.

  The optimum misalignment correction value in the existing data which is the first misalignment correction value is the misalignment correction value set when the lot is exposed and the misalignment correction value obtained from the actually measured misalignment amount. And determine the difference (S33).

  Here, the misalignment correction value is calculated as follows. That is, the actual misalignment error measured by the misalignment inspection apparatus is corrected by least square fitting using, for example, a shift component, a wafer magnification component, a wafer rotation component, a shot magnification component, a shot rotation component, and the like.

In calculating the misalignment correction value obtained from the actually measured misalignment amount, the following equations (1) and (2) are used.

  The kn terms (for example, n = 1 to 20) and the Kn terms (for example, n = 1 to 20) correspond to the misalignment correction values obtained from the actually measured misalignment amounts. The kn term and the Kn term are set to predetermined values depending on the product and the process. In this embodiment, for example, the 0th order term is a shift component, and the 1st order term is a magnification component and a rotation component. Equation (1) is used to calculate a shot component which is an exposure region unit in a semiconductor substrate, and equation (2) is used to calculate a semiconductor substrate (Wafer) component.

  The optimum misalignment correction value in the existing data is calculated using the misalignment correction value obtained from the measured misalignment amount as described above and the misalignment correction value set at the time of exposure.

  Next, based on the optimum misalignment correction value in the existing data, the misalignment in this lot is predicted by a statistical model, and the misalignment correction value to be applied to this lot as the second misalignment correction value is calculated. At this time, each misalignment correction value is calculated using the AR model and the ARMA model as statistical models (S34). Calculation formulas based on the respective models are as shown in the formulas (3) to (13) used in the first embodiment.

  First, p is set in the AR model and the ARMA model. For example, p is set from 1 to 50, and 50 formulas are set for each of the AR model and the ARMA model. At this time, q is, for example, all the existing lots from the time when the same product is manufactured and started to the current lot.

The AR model is expressed by the following equation (5).

The ARMA model is expressed by the following equation (10).

Next, an information amount reference value is obtained in each of the above-described 100 equations using equation (14).

  Among the 100 formulas set, p is selected that minimizes the information amount reference value, and the misalignment correction value is calculated in formulas (5) and (10).

  A model with good accuracy is selected from the obtained misalignment correction values, and set as a misalignment correction value to be applied to this lot (S35).

  Next, the misalignment correction value used for the main lot determined as described above is input to the exposure apparatus as data (S36).

  Subsequently, an exposure process is executed for this lot (S37). For example, this lot configured with 20 to 25 silicon substrates as a unit is introduced into the exposure apparatus. The silicon substrates introduced into the exposure apparatus are exposed one by one. The exposure to the silicon substrate is performed using one chip or two chips in the silicon substrate as one shot unit.

  After the exposure is completed, for example, a silicon substrate is extracted from the lot, the silicon substrate is introduced into a misalignment inspection apparatus, and misalignment measurement is performed (S38).

  Next, the misalignment measurement data is transmitted to the calculation unit 11 via the data delivery unit 12, and the calculation unit 11 compares the obtained misalignment measurement data with a predetermined standard of this lot, and enters the standard. It is determined whether it is out of specification or not (S39). Regardless of the standards, the obtained misalignment measurement data is sent to and stored in the lot inspection database. The data within the standard is stored in a predetermined file for use in calculating the misalignment correction value of the subsequent lot (S41). On the other hand, the data outside the standard is used when analyzing the cause, for example, It is stored in another file (S40), and the processing of this lot is finished (S42).

  In this embodiment, by applying a statistical model and selecting an optimal model from among the statistical models, a more accurate misalignment correction value is predicted, thereby improving misalignment accuracy in the actual exposure process. It becomes possible to make it.

  In addition, it is possible to reduce the necessity of performing the preceding process for each lot, and to reduce the occurrence of redoing due to a large misalignment, thereby preventing a reduction in productivity.

  FIG. 4 is a flowchart showing the exposure procedure in this embodiment. This embodiment basically takes the same procedure as that of the second embodiment. The difference from the flowchart in the second embodiment is that, in the step of selecting the optimum model from the AR model and the ARMA model, p is set so as to minimize the information amount reference value in the second embodiment. On the other hand, in this embodiment, p is set so as to minimize the residual sum of squares.

  The exposure system in the present embodiment is the same as the exposure system 1 shown in FIG. 1 in the first embodiment 1, and the description thereof is omitted.

  As shown in FIG. 4, the exposure procedure starts first (S50), and the lot information, for example, the lot ID number, is recognized for the lot entering the exposure process (S51). The recognized lot ID number is sent to the calculation unit 11 via the data transfer unit 12, and in response to an instruction from the calculation unit 11, the ID number is referred to and existing data is searched from the lot history database 15 (S52). .

  As a result of this search, the optimum misalignment correction value of the corresponding lot is read from the existing data of the corresponding lot, for example, to all the corresponding lots in the time series and read out to the calculation unit 10.

  The optimum misalignment correction value in the existing data which is the first misalignment correction value is the misalignment correction value set when the lot is exposed and the misalignment correction value obtained from the actually measured misalignment amount. The difference is obtained and determined (S53).

  Here, the misalignment correction value is calculated as follows. That is, the actual misalignment error measured by the misalignment inspection apparatus is corrected by least square fitting using, for example, a shift component, a wafer magnification component, a wafer rotation component, a shot magnification component, a shot rotation component, and the like.

In calculating the misalignment correction value obtained from the actually measured misalignment amount, the following equations (1) and (2) are used.

  The kn terms (for example, n = 1 to 20) and the Kn terms (for example, n = 1 to 20) correspond to the misalignment correction values obtained from the actually measured misalignment amounts. The kn term and the Kn term are set to predetermined values depending on the product and the process. In this embodiment, for example, the 0th order term is a shift component, and the 1st order term is a magnification component and a rotation component. Equation (1) is used to calculate a shot component which is an exposure region unit in a semiconductor substrate, and equation (2) is used to calculate a semiconductor substrate (Wafer) component.

  The optimum misalignment correction value in the existing data is calculated using the misalignment correction value obtained from the measured misalignment amount as described above and the misalignment correction value set at the time of exposure.

  Next, based on the optimum misalignment correction value in the existing data, the misalignment in this lot is predicted by a statistical model, and the misalignment correction value to be applied to this lot as the second misalignment correction value is calculated. At this time, each misalignment correction value is calculated using the AR model and the ARMA model as statistical models (S54). Calculation formulas based on the respective models are as shown in the formulas (3) to (13) used in the first embodiment.

  First, p is set in the AR model and the ARMA model. For example, p is set from 1 to 50, and 50 formulas are set for each of the AR model and the ARMA model. At this time, q is, for example, all the existing lots from the time when the same product is manufactured and started to the current lot.

  P that minimizes the residual sum of squares defined by the following formulas (6) and (11) is selected from 100 formulas, and the misalignment correction value is set in formula (5) or formula (10). calculate.

A model with good accuracy is selected from the obtained misalignment correction values and set as a misalignment correction value to be applied to this lot (S55).

  Next, the misalignment correction value used for the main lot determined as described above is input to the exposure apparatus as data (S56). Subsequently, an exposure process is executed for this lot (S57). After the exposure is completed, for example, a silicon substrate is extracted from the lot, the silicon substrate is introduced into a misalignment inspection apparatus, and misalignment measurement is performed (S58).

  Next, the misalignment measurement data is transmitted to the calculation unit 11 via the data delivery unit 12, and the calculation unit 11 compares the obtained misalignment measurement data with a predetermined standard of this lot, and enters the standard. It is determined whether it is out of specification or not (S59). Regardless of the standards, the obtained misalignment measurement data is sent to and stored in the lot inspection database. The data within the standard is stored in a predetermined file for use in calculating the misalignment correction value of the subsequent lot (S61). On the other hand, the data outside the standard is used for analyzing the cause, for example, It is stored in another file (S60), and the processing of this lot is finished (S62).

  An example of the prediction accuracy of the misalignment correction value based on the statistical processing model used in this embodiment is shown in FIG. Both the AR model and the ARMA model are time series models, and their effects are manifested. Although data for 5 lots is used, when the AR model (AR) and the ARMA model (ARMA) are used as compared with the conventional example (current state), the misalignment amount 3 sigma is reduced. Recognize.

  In this embodiment, by applying a statistical model and selecting an optimal model from among the statistical models, a more accurate misalignment correction value is predicted, thereby improving misalignment accuracy in the actual exposure process. It becomes possible to make it.

  In addition, it is possible to reduce the necessity of performing the preceding process for each lot, and to reduce the occurrence of redoing due to a large misalignment, thereby preventing a reduction in productivity.

  FIG. 6 is a flowchart showing an exposure procedure in this embodiment. This embodiment basically takes the same procedure as that of the second embodiment. The difference from the flowchart in the second embodiment is that in the step of selecting the optimum model from the AR model and the ARMA model, p is set so that the information amount reference value is minimized in the second embodiment, Further, in the third embodiment, p is set so as to minimize the residual sum of squares, whereas in this embodiment, the difference between the predicted value and the optimum correction value for the existing lot is taken, and this difference is taken. This is the point where p is set so that the deviation or variance of the above is minimized.

  The exposure system in the present embodiment is the same as the exposure system 1 shown in FIG. 1 in the first embodiment 1, and the description thereof is omitted.

  As shown in FIG. 6, the exposure procedure starts first (S70), and the lot information, for example, the lot ID number, is recognized for the lot entering the exposure process (S71). The recognized lot ID number is sent to the calculation unit 11 via the data transfer unit 12, and in response to an instruction from the calculation unit 11, the ID number is referred to and existing data is searched from the lot history database 15 (S72). .

  As a result of this search, the optimum misalignment correction value of the corresponding lot is read from the existing data of the corresponding lot, for example, for all the corresponding lots, by reading the time series in the calculation unit 11.

  The optimum misalignment correction value in the existing data which is the first misalignment correction value is the misalignment correction value set when the lot is exposed and the misalignment correction value obtained from the actually measured misalignment amount. The difference is obtained and determined (S73).

  Here, the misalignment correction value is calculated as follows. That is, the actual misalignment error measured by the misalignment inspection apparatus is corrected by least square fitting using, for example, a shift component, a wafer magnification component, a wafer rotation component, a shot magnification component, a shot rotation component, and the like.

In calculating the misalignment correction value obtained from the actually measured misalignment amount, the following equations (1) and (2) are used.

  The kn terms (for example, n = 1 to 20) and the Kn terms (for example, n = 1 to 20) correspond to the misalignment correction values obtained from the actually measured misalignment amounts. The kn term and the Kn term are set to predetermined values depending on the product and the process. In this embodiment, for example, the 0th order term is a shift component, and the 1st order term is a magnification component and a rotation component. Equation (1) is used to calculate a shot component which is an exposure region unit in a semiconductor substrate, and equation (2) is used to calculate a semiconductor substrate (Wafer) component.

  The optimum misalignment correction value in the existing data is calculated using the misalignment correction value obtained from the measured misalignment amount as described above and the misalignment correction value set at the time of exposure.

  Next, based on the optimum misalignment correction value in the existing data, the misalignment in this lot is predicted by a statistical model, and the misalignment correction value to be applied to this lot as the second misalignment correction value is calculated. At this time, each misalignment correction value is calculated using the AR model and the ARMA model as statistical models (S74). Calculation formulas based on the respective models are as shown in the formulas (3) to (13) used in the first embodiment.

  First, p is set in the AR model and the ARMA model. For example, p is set from 1 to 50, and 50 formulas are set for each of the AR model and the ARMA model. At this time, q is, for example, all the existing lots from the time when the same product is manufactured and started to the current lot.

  Next, the difference between the predicted value for the existing lot and the optimum correction value is taken, p is set so that the deviation or variance of this difference is minimized, and the misalignment correction is performed in Equations (5) and (10). Calculate the value.

  A model with good accuracy is selected from the obtained misalignment correction values and set as a misalignment correction value to be applied to this lot (S75).

  Next, the misalignment correction value used for the main lot determined as described above is input to the exposure apparatus as data (S76). Subsequently, an exposure process is executed for this lot (S77). After the exposure is completed, for example, a silicon substrate is extracted from the lot, the silicon substrate is introduced into a misalignment inspection apparatus, and misalignment measurement is performed (S78).

  Next, the misalignment measurement data is transmitted to the calculation unit 11 via the data delivery unit 12, and the calculation unit 11 compares the obtained misalignment measurement data with a predetermined standard of this lot, and enters the standard. It is determined whether it is out of specification or not (S79). Regardless of the standards, the obtained misalignment measurement data is sent to and stored in the lot inspection database. The data within the standard is stored in a predetermined file for use in calculating the misalignment correction value of the subsequent lot (S81). On the other hand, the data outside the standard is used for analyzing the cause, for example, It is stored in another file (S80), and the processing of this lot is finished (S82).

  In this embodiment, by applying a statistical model and selecting an optimal model from among the statistical models, a more accurate misalignment correction value is predicted, thereby improving misalignment accuracy in the actual exposure process. It becomes possible to make it.

  In addition, it is possible to reduce the necessity of performing the preceding process for each lot, and to reduce the occurrence of redoing due to a large misalignment, thereby preventing a reduction in productivity.

  Note that the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit of the present invention.

  In the first to fourth embodiments described above, the procedure for calculating the misalignment correction value in units of lots has been described, but the same processing can be performed in units of semiconductor substrates or exposure shots.

  Further, q to be substituted into the equation in the regression model is not limited to all existing lots. For example, when the apparatus constant changes due to exposure apparatus maintenance, apparatus reset, or the like, q may be set to target subsequent lots or semiconductor substrates.

  The role of calculating the optimum misalignment correction value from existing lot data or semiconductor substrate data, and the correction of the next lot or semiconductor substrate using the MA model, AR model, and ARMA model statistical methods from the optimum misalignment correction value The role of calculating the value and the role of determining the best model from the above models are not necessarily performed by the system control unit. For example, it may be performed in the exposure apparatus or in a misalignment apparatus.

  Further, the measurement performed by the misalignment inspection apparatus may use a misalignment measurement function in the exposure apparatus.

  In the above embodiment, the order of p is set to 50. However, it is possible to select the order depending on the number of existing lots. In the AR model, q to be substituted into the equation is set for all existing lots, but is not limited thereto. For example, if the apparatus constant changes due to exposure apparatus maintenance, apparatus reset, or the like, q may be set to target the subsequent lots. In the embodiment, the method for selecting the optimum model from the AR model and the ARMA model is described. However, even if the optimum model is selected from the AR model alone, the optimum model from the ARMA model alone is selected. May be selected.

  In addition, the present invention can be configured as described in the following supplementary notes.

  6. The exposure system according to claim 5, wherein any one of a moving average model, an autoregressive model, and an autoregressive moving average model is used as the statistical model.

  As Supplementary Note 2, a data server for storing data, an optimal misalignment value is calculated from the data, and the calculated optimal misalignment value is calculated using an autoregressive model and an autoregressive moving average model as a statistical model. Each of the calculation processing, the misalignment correction value of the main lot in the exposure process step is predicted, and the high accuracy of the misalignment correction value of the main lot by the predicted autoregressive model and autoregressive moving average model A system control unit that selects a misalignment correction value for the main lot, an exposure apparatus group that performs the exposure process using the misalignment correction value for the selected main lot, and a misalignment of the exposed main lot. An exposure system comprising a misalignment inspection apparatus group for measuring values.

  As Supplementary Note 3, in the calculation process, any one of the order that minimizes the information criterion, the order that minimizes the residual sum of squares, the order that minimizes the deviation of the residue, or the order that minimizes the dispersion of the residue An exposure system characterized by selecting an order.

1 is a block diagram showing an exposure system in a first embodiment of the present invention. 5 is a flowchart showing steps of an exposure method according to the first embodiment of the present invention. The flowchart which shows the step of the exposure method in 2nd Example of this invention. 10 is a flowchart showing steps of an exposure method according to the third embodiment of the present invention. The graph which shows the improvement of the precision of the misalignment correction value in the 3rd Example of this invention. The flowchart which shows the step of the exposure method in the 4th Example of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Exposure system 10 System control part 11 Calculation part 12 Data delivery part 13 Data server 14 Lot inspection database 15 Lot history database 16 Lot storage 17 Exposure apparatus group 18 Exposure apparatus 19 Misalignment inspection apparatus group 20 Misalignment inspection apparatus

Claims (5)

Calculating means for calculating a first misalignment correction value from existing data;
Predicting means for calculating the calculated first misalignment correction value using a statistical model and predicting a second misalignment correction value used in the exposure processing step;
An exposure method comprising: an exposure processing unit that executes the exposure processing step using the predicted second misalignment correction value.   The exposure method according to claim 1, wherein any one of a moving average model, an autoregressive model, and an autoregressive moving average model is used as the statistical model. Calculating means for calculating a first misalignment correction value from existing data;
Prediction for calculating the first misalignment correction value calculated using the autoregressive model and the autoregressive moving average model as statistical models, respectively, and predicting the second misalignment correction value used in the exposure process. Means,
Selecting means for selecting a second misalignment correction value with high accuracy among the second misalignment correction values based on the predicted autoregressive model and autoregressive moving average model;
An exposure method comprising: an exposure processing unit that executes the exposure processing step using the selected second misalignment correction value.   In the calculation process, the order that minimizes the information criterion, the order that minimizes the residual sum of squares, the order that minimizes residue deviation, or the order that minimizes residue dispersion is selected. The exposure method according to claim 3. A data server for storing data;
A first misalignment correction value is calculated from data stored in the data server, the calculated first misalignment correction value is calculated using a statistical model, and a second misalignment used in an exposure processing step. A system controller that predicts a correction value;
An exposure processing apparatus that executes an exposure processing step using the second misalignment correction value;
An exposure system comprising: a misalignment inspection apparatus that measures misalignment values of an exposed lot.

Images