JPH05253797A - Judging system for abnormality on line - Google Patents

Abstract

PURPOSE:To immediately judge abnormality of equipment concerning manufacture of a thin film product by three algorism, namely, by comparing the specification value with the working value of a product, by computing the estimated working value from the setting values of respective equipments parameters and comparing it with the working measured value, or by judging the working value to repeat increase and decrease over a certain number of times. CONSTITUTION:This line abnormality judging system is constituted of a computer for judgement 1, a past recorded data base 6 for equipments parameters, and a size measured value data base 7, an estimated working value computing formula is prepared, an estimated working value is computed from the setting value of equipments by the use of the formula, and if there is difference over a certain quantity at comparing the estimated value with the working measured value of the product, abnormality of the equipments is judged.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a line abnormality judging method, and in particular, an abnormality of a manufacturing facility in manufacturing a thin film product such as a semiconductor facility is determined by (1) a difference between a standard value and a processed value, (2) Using three algorithms, the difference between the predicted value and the processed value, and (3) the tendency of the processed value, line abnormalities are monitored in real time to detect equipment abnormalities at an early stage or predict abnormalities in advance to detect defects. The present invention relates to a technique for reducing the amount.

[0002]

2. Description of the Related Art There is a method of measuring the quantity related to abnormality or failure of manufacturing equipment, obtaining the quantity as an index of abnormality or failure from those values, and judging the state of equipment by this, "measurement and control" Vol.25, No.10 (October 1986), pages 3-10, "Current Status and Future of Equipment Diagnosis Technology," discusses by Toshio Sada and Shozo Takada.

In the case of this method, an abnormality of individual equipment,
It is necessary to know in advance the value of equipment that is clearly associated with a failure. Furthermore, it is necessary to attach a sensor or the like that is not directly related to the operation of the equipment as a means for detecting the value of the equipment. However, in a manufacturing facility that performs fine processing such as semiconductor products, the sensors themselves may interfere with the normal operation of the facility and cause defects.

[0004]

The miniaturization of semiconductor products has been advancing year by year, and as a result, the precision of manufacturing equipment itself has become insufficient. In other words, the processing size is 1 micron (0.0
In the manufacture of semiconductor products having a size of 01 mm or less, the processing values of the products often fail because the processing values do not meet the standard even if the equipment parameters slightly deviate from the set values. Therefore, it is difficult for the above-mentioned conventional technique to detect an abnormality in the semiconductor equipment, and it is impossible to detect a slight abnormality in the equipment. Therefore, when an abnormality is detected by this method, a considerable failure has already occurred, and as a result, a large number of defects occur and the damage becomes great. Moreover, the equipment had to be stopped for a long time for the repair, which greatly hindered the production.

An object of the present invention is to detect abnormalities in a line and equipment quickly by combining a plurality of abnormality determination methods in order to monitor the state of equipment for performing such fine processing, and reduce defects. is there.

[0006]

[Means for Solving the Problems] In order to achieve the above object, a database of standard values of product works, a database for collecting set values of equipment, and actual measured values of products processed under the set values are provided. Have a database to collect,
A means for analyzing those data and calculating a prediction algorithm for predicting the processed value of the product from the set value of the equipment is provided. Then, (1) a means for determining that the actually measured value exceeds the standard value of the product, (2) when the set value of the equipment is input, the predicted processed value of the product is calculated from the prediction algorithm, and then the actually measured value of the product is calculated. Means for measuring and comparing the data, and (3) means for determining whether the processed value has continuously increased or decreased a fixed number of times are provided.

[0007]

According to the present invention, when acquiring the information about the abnormality of the equipment, the standard value of the product is stored, and the input value of each setting parameter of the equipment is stored, and a plurality of types of parameters obtained from a fixed amount of data in advance are stored. A predicted algorithm for the actual measured value of the product work is calculated from the measured actual processing data, and the predicted processed value of the product is calculated using the algorithm from the set value of the equipment parameter. After that, the actual measurement value of the product is compared with the calculated predicted processing value, and if a difference of a certain amount or more occurs between the two, an abnormality of the equipment is determined. Also, if the actual measurement value exceeds the standard value, the equipment abnormality is determined. Further, the abnormality is determined even when the actual measurement value increases or decreases a certain number of times in a row.

[0008]

EXAMPLES Examples of the present invention will be described below. FIG. 1 shows a processing flow of an embodiment of the determination method of the present invention.
In this embodiment, as an example of a prediction algorithm, a statistical method such as multiple regression analysis for calculating a prediction value from some parameters will be taken up and a prediction formula thereof will be described.

First, set values of some parameters of the equipment are collected as measurement data. Next, each parameter is substituted into the predicted processing value calculation formula 13, and the processing value of the product predicted from the set value of the parameter is calculated. And as soon as the processing of the product is completed, the actual measurement value of the product is measured,
First, it is determined whether the measured value exceeds the standard value of the product. Next, the actual measurement value of the product is compared with the above-mentioned predicted value, and if the difference between the two is a certain amount or more, it is determined that the processing was not performed according to the set value, and the equipment is warned. In this way, as soon as the processing is completed, the equipment is judged to be abnormal, and even if it is within the standard value, the difference between the predicted value and the measured value is judged to be abnormal if it exceeds a certain amount, so it is possible to detect the abnormality quickly. It is possible to prevent the creation of defects.

FIG. 2 shows, as an embodiment, a hardware configuration when this system is applied to equipment for a photolithography process of semiconductor manufacturing equipment. The system of this embodiment comprises a judgment computer 1 for making an abnormality judgment, a photolithography process exposure equipment 8, a computer 2 for collecting setting parameter values of the equipment, a developing equipment 9 and a computer 3 for collecting setting parameter values of the equipment. Also, the inspection equipment 11 for measuring the actual measurement value of the processed product and the computer 12 for collecting the measurement value are provided.
0 is connected, and the result is displayed on the CRT 5 by operating the keyboard 4.

The judgment computer 1 has an equipment parameter actual result database 6 and a dimension actual measurement value database 7. By inputting a product name, the number of lots, a process name, each equipment parameter actual value and a period from the keyboard 4, the judgment computer 1 Formula 13 for calculating the predicted dimension value of the product by analyzing the data is created. FIG. 3 shows a calculation flow of the predicted processing value calculation formula 13. In this way, the predicted machining value calculation formula 13 serving as a reference for determining the equipment abnormality is created.

As the equipment parameter, for example, the exposure equipment 8
The exposure energy (x1), exposure time (x2), resist film thickness (x3), resist temperature (x
4) etc. are mentioned. Further, regarding the developing equipment 9, there are a developing time (z1), a resist film thickness (z2), a resist temperature (z3) and the like. FIG. 4 shows a predicted processed value calculation formula 13 calculated from these parameters using a statistical method such as multiple regression analysis. By using such a calculation formula, when the product is put into the equipment and the equipment parameters are set, the dimension value predicted from the value is calculated and displayed on the CRT 5.

FIG. 5 shows the equipment parameter result database 6
2 shows a data table and a data table of the dimension measurement value database 7. These data tables can be displayed on the CRT 5 by operating the keyboard 4 of the judgment computer 1.

Next, a method of determining abnormality will be described. First, a semiconductor wafer is put into the exposure equipment 8 and each equipment parameter is set. The computer 2 first checks whether the set parameters are correct, and if they are correct, they are sent to the judgment computer 1 via the communication line 10. Therefore, each set value is substituted into the calculation formula, and the predicted dimension value (L2) is calculated and displayed on the CRT 5. After that, when the processing of the product is completed, the actual processing dimension (L1) of the product is measured by the dimension inspection equipment 11. Then, the measured value (L1) is first judged whether or not it is within the standard value of the dimension. If it is out of the standard value, a warning is given at this point that the equipment is abnormal. If it is within the standard value, then the measured value (L1) and the predicted value (L2) are compared, and if the difference between them is within a certain amount, it is normal, and if it exceeds a certain amount, the dimension predicted from the set value. It is judged that it was not processed according to the value and warns that some trouble has occurred in the equipment. If this is expressed by an equation, | L2-L1 |
If ≧ a, it is determined to be abnormal. (However, a is a fixed amount.) That is, according to this method, if there is a difference of a fixed amount or more between the predicted value and the measured value even if the measured value of the processed product is within the standard value, it is determined that the equipment is abnormal. Even if the product is a non-defective product, it is possible to detect abnormalities in the equipment that processed it at an early stage, and it is possible to prevent the occurrence of a large number of defects due to the abnormality in the equipment at an early stage by quickly performing maintenance on the equipment.

FIG. 6 shows a graph of measurement results of dimensions. The horizontal axis represents time and the vertical axis represents the dimension processing value. The actual measurement value 16 is plotted against the dimension standard value 17, and if the standard value 17 is exceeded, it is determined to be abnormal.

Next, FIG. 7 shows a graph comparing the predicted value and the measured value of the dimension. Calculation formula 1 even within the dimension standard value 17
If the difference 19 between the predicted value 18 and the measured value 16 calculated from 3 is a certain amount or more, it is determined that the machining has not been performed according to the normal set value, and it is determined as abnormal. These screens can be displayed on the CRT 5 by the judgment computer 1.
Further, when it is determined that there is an abnormality, the determination computer 1 can inform the operator of the abnormality by sounding a buzzer or changing the color display on the CRT 5.

FIG. 8 shows the data of the predicted value and the measured value of the dimension C.
The table displayed on the RT screen is shown. Product name HD from the figure
4010 lot no. The measured value is 1.102, which is within the management standard of the measured value with respect to the predicted value of 1.132, but the difference from the predicted value is 0.0 when the fixed amount a = 0.025.
Since it is 30 or more than a certain amount, it is judged that the processing was not performed according to the set value, and the abnormality of the first exposure equipment is warned and the next product introduction is stopped. As a result, it is possible to minimize the occurrence of defects due to equipment abnormalities, take immediate countermeasures for equipment, and reduce downtime.

Next, the rewriting of the prediction formula will be described. The accuracy of the prediction formula itself is important for the accurate detection of equipment anomalies by this method. That is, if the reliability of the prediction formula is low, the abnormality of the equipment warned based on it cannot be trusted, and in an extreme example, there is a possibility that a false warning may be tampered with and the normal equipment may be changed into an abnormal state. Therefore, it is necessary to judge the validity of the prediction formula itself. FIG. 9 shows a flow chart for determining the accuracy of the prediction formula.

Next, an example of a method for determining the validity of the prediction formula will be described with reference to FIG. First, the product is thrown in to set the equipment, and when the difference between the predicted value (L2) and the actually measured value (L1) from each setting parameter becomes a certain amount or more, the equipment warning is given. At that time, the judgment computer 1 Refer to the database that created the prediction formula. If the data used to create the prediction formula is older than 3 months, the warning of the prediction formula itself is warned. Then, measurement is performed, and if the predicted value and the measured value are greatly different from each other three times in a row, the calibration of the prediction formula is instructed. Then, search each database for each parameter setting value and measurement data for the last one month,
Recalculate a new prediction formula. In this way, the prediction formula itself can be determined and the accuracy can be improved, and an abnormality in the equipment can be accurately warned.

Another example of determination of abnormality will be described. FIG. 10 shows a graph of measurement results of dimensions. Judgment computer 1
Has an algorithm for determining that the measured value 16 of the dimension is abnormal if it is continuously increased or decreased by a predetermined number.

Here, an example in which it is determined to be abnormal if the number of times of increase increases three times or more is shown. Machining actual value 16 is dimension standard value 17
Even within the range, since the processing value has increased three times in a row,
It is judged that if the machining is continued as it is, the standard value 17 will be exceeded, an abnormality is judged, and an instruction is given to perform maintenance of the equipment.

Further, the function of the judgment computer 1 can be installed in the calculator 2 or the calculator 3, so that the judgment computer 1 can be omitted.

The above-described embodiments of the present invention are merely examples, and other equipment for semiconductors, for example, film forming equipment such as CVD, and implantation equipment can also be used. Other T
It can also be applied to mass production equipment for thin film products such as FT (thin film transistor equipment).

[0024]

As described above, in the present embodiment, the abnormality of the equipment is (1) Whether the dimensional machining value of the product exceeds the standard value. (2) The predicted value and the dimensional machining predicted from the setting parameters of the equipment. Difference in value (3) Judgment is made by three algorithms of whether or not the dimension processing value of the product has continuously increased or decreased a certain number of times.

As a result, the abnormality of the equipment which cannot be detected by the usual algorithm (1) is detected at an early stage,
By performing maintenance quickly, it is possible to minimize the occurrence of product defects due to equipment abnormalities, which contributes to improved productivity by reducing material costs and downtime for equipment repairs. it can.

[Brief description of drawings]

FIG. 1 is a process flow diagram of an abnormality determination method of the present invention.

FIG. 2 is a hardware configuration diagram when the present invention is applied to a photolithography process which is one of manufacturing processes of semiconductor products.

FIG. 3 is a flowchart for calculating a processing value prediction formula necessary for determining an abnormality.

FIG. 4 is a diagram showing an example of a predictive machining value calculation formula of a dimension created from each equipment parameter of hot storage equipment.

FIG. 5 is a diagram showing a data table of data registered in the equipment parameter actual result database 6 and the measured dimension value database 7.

FIG. 6 is a diagram showing an example of determination of abnormality by comparing a machining value and a standard value as a time series plot of dimension machining values.

FIG. 7 is a diagram showing an example of determination of abnormality by comparing a difference between a processed value and a predicted value as a time series plot of dimension processed values.

FIG. 8 is a diagram showing a display example of a screen for determining a facility abnormality by comparing the dimension prediction value and the actual measurement value data.

FIG. 9 is a processing flow chart for determining the accuracy of a predicted machining value calculation formula.

FIG. 10 is a diagram showing, as a time-series plot of dimension processing values, an example of determining an abnormality by continuously increasing the processing values a fixed number of times.

[Explanation of symbols]

1 ... Judgment computer, 2 ... Data collection computer, 3 ... Data collection computer, 4 ... Keyboard, 5 ... CRT, 6 ... Facility parameter record database, 7 ... Actual dimension database, 8 ... Exposure facility, 9 ... Development Equipment, 10 ... communication line, 11 ... dimension inspection equipment, 12 ... data collection computer,
13 ... Predicted machining value calculation formula, 14 ... Equipment parameter data table, 15 ... Dimension measured value data table, 16 ... Machining measured value, 17 ... Dimension standard value, 18 ... Dimension predicted value, 19
... difference between predicted value and actual measurement value, 20 ... abnormality determination data table, 21 ... actual measurement value that has been increased three times in a row.

Claims (9)

[Claims] Claims: 1. Three determination algorithms for the difference between the standard value of a product work and the actual measurement value of machining and the predicted machining value predicted from the set value of equipment and the difference between the actual machining value after machining and the increase / decrease tendency of the actual machining value. Based on this, a line abnormality determination method characterized by determining equipment abnormality. 2. The line abnormality determination method according to claim 1, wherein an abnormality of the line is determined by an arbitrary one of the three determination algorithms or a combination of two determination algorithms. 3. A line abnormality characterized in that in the judgment method according to claim 1, a preset standard value is compared with an actual machining value of a product work, and if the actual machining value exceeds the standard value, it is determined to be abnormal. Judgment method. 4. The judgment method according to claim 1, wherein the predicted machining value of the product predicted from the set value of the equipment is compared with the actual measurement value after machining, and if a difference of a certain amount or more occurs, it is judged as abnormal. A line abnormality determination method characterized in that 5. The line abnormality determination method according to claim 1, wherein when the actual measurement value of the product workpiece continues to increase or decrease by a preset number, it is determined to be abnormal. 6. A database for storing set values for equipment, a database for processed values for products, and those data are analyzed,
A line anomaly judgment method characterized by being composed of a judgment equipment for judging equipment abnormalities. 7. The line abnormality determination method according to claim 6, wherein a predictive algorithm for predicting a processed value of a product is created from a set value of a fixed amount of equipment and a processed value of a product. 8. Using the prediction algorithm obtained in claim 7, the predicted machining value predicted from the set value is compared with the actual machining value of the product, and if a difference of a certain amount or more occurs, it is determined that the equipment is abnormal. A line abnormality determination method characterized by: 9. With respect to the prediction algorithm obtained in claim 7, the accuracy of the prediction algorithm itself is determined when a fixed amount of set value and actual machining value data is accumulated, and the prediction algorithm is calibrated if necessary. A line abnormality determination method characterized in that