JP2006310504A - Method of manufacturing semiconductor device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve the operation control of an apparatus automatically by cooperating with a manufacture executive system (MES) by detecting the faults of a manufacturing apparatus. <P>SOLUTION: A method of manufacturing a semiconductor device includes a step of saving data arranged per wafer unit in a format defined in advance, a step of detecting the faults to the data, a step of informing the MES of the detected faults, and a step of predicting the maintenance time from the tendency of statistical data. With this configuration, the automatic control of the apparatus can be performed on the occurrence of the faults generated in the apparatus, and the maintenance can be execute based on the device state. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a method of manufacturing a semiconductor device that maximizes the operation of the device by quantifying the processing state of the semiconductor device with data and quickly detecting an abnormality.

As a conventional technique, a method for collecting data from a semiconductor device and a method for detecting abnormal behavior of collected data will be described.
Patent Document 1 describes that data is collected from a device and an abnormality or a deterioration state of the semiconductor device is detected. Specifically, analysis processing is performed on all process data obtained from each data processing apparatus, and the analysis result and process data centralized monitoring processing, processing for reflecting the analysis / statistic result in the recipe, and the like are performed. Thereby, the abnormality and deterioration state of a processing apparatus are discovered in detail and early.
Japanese Patent Laid-Open No. 11-16797

  However, in Patent Document 1, the same data collection device is attached to a plurality of devices, and the format of collected data depends on the data collection device. In view of the current efforts of each device manufacturer, data collection devices provided by multiple device manufacturers have been introduced in one factory / line, and the data acquired by each data collection device can be managed in an integrated manner. It had the problem that it was not possible. In addition, there is a problem that even if an abnormality is detected, the operation of the apparatus cannot be controlled automatically in cooperation with a MES (Manufacturing Execution System).

  According to a first aspect of the present invention, there is provided a method of manufacturing a semiconductor device, wherein the manufacturing execution system instructs the semiconductor device to process the semiconductor device, collects data from the semiconductor device being processed at high speed in real time, A process of determining abnormality of the collected data in real time, a process of transmitting the abnormality of the collected data to a manufacturing execution system, a process of organizing the collected data into a data group in units of wafers at the timing of wafer processing completion, A step of storing data arranged in units of wafers in a predefined format, a step of calculating statistical values from the data arranged in units of wafers, and a step of saving the calculated statistical values in a predefined format And an abnormality determination process for the statistical value, and the statistical value abnormality is transmitted to the manufacturing execution system. A step, a step of calculating the time of optimal maintenance of the semiconductor device from the tendency of the statistical value, characterized in that the maintenance time that the calculated and a step of notifying the maintenance management system.

  According to a second aspect of the present invention, there is provided a method for manufacturing a semiconductor device according to the first aspect, wherein in the step of transmitting abnormality of data collected at high speed in real time from a semiconductor device being processed, a combination of a recipe and a product type is provided to the manufacturing execution system. An abnormality is notified in units.

  According to a third aspect of the present invention, there is provided a method for manufacturing a semiconductor device according to the first aspect, wherein in the step of notifying the statistical value abnormality, the abnormality is notified to the manufacturing execution system in units of combination of recipe and product type. .

  The method of manufacturing a semiconductor device according to claim 4 of the present invention is the method for manufacturing a semiconductor device according to claim 1, wherein the manufacturing execution system to which an abnormality of data collected in real time from the semiconductor device being processed is transmitted is a combination of a recipe and a product type. It is characterized by determining whether or not lot processing is possible and tracking the lot.

  The method of manufacturing a semiconductor device according to claim 5 of the present invention is the method for manufacturing a semiconductor device according to claim 1, wherein the manufacturing execution system to which the abnormality of the statistical value is transmitted determines whether or not lot processing is possible by a combination of a recipe and a product type. It is characterized by tracking.

(1) In view of the production status, it is possible to carry out optimal maintenance timing, and it is possible to maximize the operation of the apparatus.
(2) The method of detecting an abnormality in the slurry supply flow rate can detect an abnormality in accordance with the process processing mode of the apparatus as well as the conventional simple setting of upper and lower limit values (threshold values).

(3) Both abnormal detection of sudden fluctuations in the slurry flow rate and abnormal detection of changes over time (deterioration) can be realized.
(4) The detected abnormality can be automatically notified to the MES, and the operation of the apparatus can be controlled without human judgment.

Hereinafter, a method of manufacturing a semiconductor device according to the present invention will be described based on specific examples shown in FIGS.
(First embodiment)
1 to 7 show a first embodiment for carrying out the present invention.

  FIG. 1 is a system configuration diagram when implementing the present invention. A CMP (Chemical Mechanical Polishing) apparatus 101, a data collection apparatus 102 for collecting data in real time from the CMP apparatus 101, and the data acquired in real time are shown. On the other hand, data (waveform) behavior abnormality determination device 103 that determines abnormality in real time, data primary arrangement device 104 that arranges data in units of wafers from data acquired by the data collection device, and data that is arranged in units of wafers A primary data storage device 105 that stores data, a statistical value calculation device 106 that calculates a statistical value from the stored data, a statistical data storage device 107 that stores the calculated statistical value, and the stored statistics Optimal from the statistical value abnormality determining device 108 for determining abnormal values and the tendency of the stored statistical values Composed of the maintenance management system 111 for managing maintenance of MES (manufacturing execution system) 110 and the semiconductor device for performing the operation control of the lot progress management and the semiconductor device in the maintenance timing calculation unit 109 and the semiconductor factory to calculate the maintenance time.

  In this embodiment, the data collection device 102, the data (waveform) behavior abnormality determination device 103, the data primary organization device 104, the primary data storage device 105, the statistical value calculation device 106, the statistical value data storage device 107, the statistical value abnormality determination The device 108 and the maintenance time calculation device 109 may be expressed using a computer including a central processing unit, storage means, input means such as a keyboard, and display means such as a display.

  The data collection device 102 collects data in real time at a data collection period of 1 Hz or more during the polishing process of the CMP device 101. Data items to be collected are CMP process parameters such as polishing head load, polishing head rotation speed, platen rotation speed, slurry flow rate, etc., and device status information and recipe step information that are parameters indicating process processing status. PJ (Process Job) information, CJ (Control Job) information, and recipe. The PJ information and CJ information mentioned here are standards for semiconductor manufacturing equipment defined by the SEMI standard (Semiconductor Equipment and Materials Institute Standard).

  As a data collection method, S1F3 (device state variable inquiry) and S1F4 (device state variable reply), which are Stream / Functions defined in SECS (Semiconductor Device Communication Standard), may be used, and S6F11, S6F9, etc. The processing information may be acquired using event information collection. In addition to data collection by communication as described above, data may be collected by voltage values from an analog data output port owned by the apparatus, or data may be obtained via a sequencer.

  The data (waveform) behavior abnormality determination device 103 automatically determines whether or not the data shows a normal behavior with respect to the data collected in real time by the data collection device 102.

  FIG. 2 is a graph showing the relationship among the head load 201, slurry flow rate 202, and recipe step 203 in the CMP apparatus. In the case of FIG. 2, the polishing load is controlled for each recipe step 203, step 1 is a head load application step, step 2 is a head load stabilization step, and step 3 is a head load application stop and the polishing process is completed. It is. On the other hand, the behavior of the slurry flow rate 202 and the transition of the recipe step 203 are not synchronized.

  FIG. 3 shows the abnormality determination standard 304 set for the head load 201a, the slurry flow rate 202a, the recipe step 203a, and the slurry flow rate 202a, and the virtual step 305 virtually created. Since the behavior of the slurry flow rate 202a and the transition of the recipe step 203a are not synchronized as described in the explanation of FIG. 2, if the abnormality detection standard 304 is set for the entire recipe step 2, the slurry flow rate 202a of the recipe step 2 is set. Since the slurry flow rate 202a deviates from the abnormality detection standard 304 at the rising edge, it is detected as abnormal. In this case, the behavior of the slurry flow rate 202a is detected as abnormal although the behavior is correct. Therefore, correct automatic abnormality detection cannot be performed unless a virtual step synchronized with the behavior of the slurry flow rate 202a is created. Therefore, automatic abnormality detection is performed by creating a virtual step 305 and enabling the abnormality determination standard 304 only for the section of the virtual step 2-2 in the virtual step 305.

  This is a definition of the virtual step 305. This is a virtual step 2 after a fixed elapsed time (15 seconds in the case of FIG. 3) elapses from the change point where the recipe step 203a changes from 1 to 2. -1 to virtual step 2-2, for example, the amount of change of the slurry flow rate 203a in the past 3 seconds is calculated as needed, and the virtual step is the point that the amount of change becomes 100 ml / min or more. You may make a transition from 2-1 to virtual step 2-2.

  Further, in order for the data (waveform) behavior abnormality determining device 103 to perform abnormality determination in more detail, it is necessary to determine abnormality for each type of processing wafer. This is because the permissible parameter variation differs for each product type. In the present invention, PJ (Process Job) information, which is processing information that can be acquired from a normal semiconductor device, CJ (Control Job) information, and using recipe as key information, the MES is inquired about the product information, thereby acquiring the product information. Anomaly judgment is performed with the anomaly detection standard to be set for the product type.

  When the data (waveform) behavior abnormality determination device 103 detects an abnormality, it has an interface for reporting to the MES, and notifies the MES by associating the recipe and the product type information with the parameter where the abnormality is detected. By notifying an abnormality by associating the recipe with the product type, an instruction (processing stop instruction) to the apparatus is (1) immediate stop (next wafer processing stop) (2) all subsequent processes stop (3) all subsequent processes It is possible to divide the level into recipes and varieties as in the case of stopping processing of a specific recipe / variety. This abnormality notification method and the device operation control method in MES are particularly effective in semiconductor factories that produce a wide variety of products in small quantities. Factory production is possible by controlling the device in consideration of the status of the product in process. Efficiency can be maximized.

  The data primary organizing device 104 organizes data collected in real time in units of wafers at the end of wafer processing and stores the data in the primary data storage device 105. The organization here is to create a data file in a predefined format.

  FIG. 4 shows an example of a data format created for each wafer. In the header part, lot ID (LotID), carrier ID (CarrierID), slot ID (SlotID), wafer ID (WaferID), recipe name (RecipeName), PJ , CJ, product type ID (ProductID), and process ID (ProcessID). The data section (Data) has a time stamp, recipe step, and parameter values as a column structure, and each row has a structure of data for each time stamp.

  The statistical value calculation device 106 calculates a statistical value (for example, maximum value, minimum value, average value, standard deviation) of each parameter from the data file created in the primary data storage device 105 for each wafer. The calculation of the statistical value of each parameter is based on the calculation of the statistical value for each recipe step. However, as described above, as shown in FIGS. 2 and 3, not all process parameters are controlled step by step. By calculating the statistical value of the parameter using the virtual step 305 described with reference to FIG. 3, it is possible to correctly quantify the process state of the apparatus.

  The statistical value data storage device 107 stores the statistical values calculated by the statistical value calculation device 106 in a predefined format. FIG. 5 is an example of a data table storing statistical values. In the case of FIG. 5, data is stored in a relational database, and the table structure of the relational database includes apparatus ID (EquipmentID), chamber ID (ChamberID), parameter name (ParameterName), PJ, CJ, recipe name (RecipeName), Product ID (ProductID), Process ID (ProcessID), Time (TiMEStamp), Step number (Step), Maximum parameter value (Max), Minimum parameter value (Min), Average parameter value (Ave), Standard parameter It has a structure consisting of deviation (Std).

  The statistical value abnormality determination device 108 performs abnormality detection and trend management on the statistical values stored in the statistical value data storage device 107. Abnormality detection and trend management here refers to the abnormality judgment criteria of the new JIS control charts (JIS Z 9021 (1998) “Shoehart control chart” and JIS Z 9020 (1999) “control charts—general guidelines”). Yes. Moreover, you may perform abnormality determination by the Hotelling T2 statistical analysis. Similar to the data (waveform) behavior abnormality determination device 103 described above, the statistical value abnormality determination device 108 has an interface for reporting to the MES, and associates the recipe and the product type information with the parameter in which the abnormality has been detected. Notice. By notifying the abnormality with a combination of recipe and product type, an instruction (processing stop instruction) to the apparatus is (1) immediate stop (next wafer processing stop) (2) all processing stop after the next lot (3) next lot Thereafter, the level can be divided into recipes and varieties as if a specific recipe / variety is stopped. This abnormality notification method and the device operation control method in MES are particularly effective in semiconductor factories that produce a wide variety of products in small quantities. Factory production is possible by controlling the device in consideration of the status of the product in process. Efficiency can be maximized.

The maintenance time calculation device 109 calculates an optimal maintenance time from the trend of the statistical values stored in the statistical value data storage device 107.
FIG. 6 is a trend graph of the average value of the slurry flow rate for each wafer processing, with the date and time on the X axis and the average value of the slurry flow rate on the Y axis. The plot of the average value 601 of the slurry flow rate during wafer processing is linear. The intersection 604 of the approximated straight line 602 and the lower limit control limit line 603 of the slurry flow rate average value is the maintenance essential time. In the case of FIG. 6, the approximate expression of the approximate straight line 603 is Y = −0.0146X + 206.33.
Thus, it can be predicted that the lower limit management limit line 603 (190 ml / min) will be reached when about 510 sheets are processed. The maintenance time calculation device 109 notifies the maintenance management system 111 of the information of 510 sheets, and the maintenance management system that receives this information cooperates with the MES to consider the status of the product in progress in the factory, and the maintenance schedule of the device. The factory production efficiency can be maximized by adjusting. Further, in the case of FIG. 6, linear approximation is used in calculating the maintenance execution time, but logarithmic approximation, polynomial approximation, power approximation, exponential approximation, and moving average may be used as the approximation method.

Next, an apparatus abnormality detection method and an abnormality notification method at the time of abnormality detection according to the present invention will be described.
FIG. 7 is a flowchart showing a series of flows of each apparatus shown in FIG.
First, process processing is started in the CMP apparatus 101 (S701), and the data collection apparatus 102 collects data from the CMP apparatus 101 in real time (S702). Next, it is determined whether or not the timing is an abnormality detection target (S703). Basically, abnormality determination is performed for each recipe step, but abnormality detection may be performed based on a virtual step virtually created as described above. In the case of the abnormality detection target timing, the data (waveform) behavior abnormality determination device 103 performs abnormality determination on the data collected by the data collection device 102 in real time (S704). If an abnormality is detected, the data (waveform) behavior abnormality determination device 103 notifies the MES of the abnormality (S705). When notifying the MES 110 of an abnormality, the information on the recipe and the product type is associated with the parameter in which the abnormality is detected. S702 to S706 are repeated until the abnormality detection target section ends (S707). Data collection is performed while the CMP apparatus 101 is in process even after the abnormality detection target section ends (S708). When the processing is completed, the data collected by the data primary organizing apparatus 104 is collected for each wafer to create a data file, and the data file is stored in the primary data storage apparatus 105 (S709). Next, the statistical value calculation device 106 calculates a statistical value from the data file stored in the primary data storage device 105 and stores it in the statistical value data storage device 107 (S710). The calculation of the statistical value of each parameter is based on the calculation of the statistical value for each recipe step. However, as shown in FIGS. 2 and 3, not all process parameters are controlled step by step. The statistical value of the parameter may be calculated using the virtual step 305 described with reference to FIG. Next, the statistical value abnormality determination device 108 performs the statistical value abnormality determination stored in the statistical data storage device 107 (S711). For the abnormality judgment of statistical values, the abnormality judgment standard of the new JIS chart, the Hotelling T2 statistical analysis, etc. are used. If an abnormality is detected in S711, the MES 110 is notified (S712). When notifying the MES 110 of an abnormality, the information on the recipe and the product type is associated with the parameter in which the abnormality is detected. Further, the maintenance necessary timing is calculated from the trend of the statistical values stored in the statistical data storage device 107 (S714), and the maintenance management system 111 is notified.

(Second Embodiment)
Next, in the first embodiment, when a real-time abnormality determination is not possible, or when an abnormality detection is possible but a data collection device that does not have an interface for abnormality notification to the MES is attached to the semiconductor device An embodiment of the present invention will be described with reference to FIGS.

  FIG. 8 is a system configuration diagram when implementing the present invention. The CMP apparatus 101a, the data collection apparatus 102a that collects data from the CMP apparatus 101a in real time, and the data in wafer units from the data acquired by the data collection apparatus. Data primary arrangement device 104a to be arranged, primary data storage device 105a to save the data arranged in units of wafers, and data (waveform) behavior abnormality determination device 103a to determine abnormality with respect to the primary data storage device 105a data A statistical value calculation device 106a for calculating a statistical value from data stored in the primary data storage device 105a, a statistical value data storage device 107a for storing the calculated statistical value, and the stored statistical value Statistical value abnormality determination device 108a for determining abnormality, and tendency of the stored statistical value Composed of the maintenance management system 111a for managing the maintenance of the MES 110 'and the semiconductor device for performing the operation control of the lot progress management and the semiconductor device in the maintenance timing calculation unit 109a and the semiconductor factory for calculating an et optimum maintenance timing.

  The difference from FIG. 1, which is the configuration described in the first embodiment, is that the data (waveform) behavior abnormality determining device 103a determines an abnormality with respect to the data stored in the primary data storage device 105a. is there. When the data collection device 102a cannot be connected to the data (waveform) behavior abnormality determination device 103 shown in FIG. 1 in which an abnormality is detected in real time, the data stored in the primary data storage device 105a is processed. To detect abnormalities in the waveform data. Even if the data collection device outputs data in any format, the data format is uniquely defined by storing it in the primary data storage device 105a in accordance with the format defined in advance by the data primary organization device 104a. Therefore, the data (waveform) behavior abnormality determination device 103a can detect abnormality. Further, according to the configuration of the present invention, the data (waveform) behavior abnormality determination device 103a only needs to support one data format.

  It goes without saying that the statistical value calculation device 106a, the statistical value data storage device 107a, the statistical value abnormality determination device 108a, and the maintenance time calculation device 109a perform the same functions as described in the first embodiment.

Next, an apparatus abnormality detection method and an abnormality notification method at the time of abnormality detection according to the present invention will be described.
FIG. 9 is a flowchart showing a series of flows of each apparatus shown in FIG.
First, process processing is started in the CMP apparatus 101a (S901), and the data collection apparatus 102a collects data from the CMP apparatus 101a in real time (S902). At the end of wafer processing (S903), the data primary organizing apparatus 104a creates a data file by collecting the collected data in units of wafers, and stores the data in a predefined format in the primary data storage apparatus 105a (S904). Next, the data (waveform) behavior abnormality determination device 103a identifies the section of the waveform data abnormality detection target from the data file (S905). The anomaly detection target section is basically defined for the recipe step, but anomaly detection may be performed based on a virtual step virtually created as described in FIG. In the abnormality detection section identified in S905, the data (waveform) behavior abnormality determination device 103a performs abnormality detection (S906), and when there is an abnormality, notifies the MES 110a of the abnormality. When notifying the MES 110a of an abnormality, the recipe and type information are associated with the parameter where the abnormality is detected and notified.

  S910, S911, S912, S914, and S915 correspond to S710, S711, S712, S714, and S715 of FIG. 7 described in the first embodiment, respectively, and perform the same function, thereby obtaining statistical data. It is obvious to detect anomalies.

  It is useful as a system for semiconductor device abnormality detection and device control / maintenance after abnormality detection.

Configuration diagram according to the first embodiment of the present invention Relationship diagram of recipe step, slurry flow rate and head load in CMP equipment Relationship diagram of virtual step, slurry flow rate and head load in the present invention Data format diagram created for each wafer in the present invention Data table diagram storing statistical values in the present invention The conceptual diagram which calculates the maintenance execution required time from the tendency of the statistical value data in this invention The flowchart which shows a series of flows of the manufacturing method of the semiconductor device as described in 1st Embodiment in this invention. Configuration diagram according to the second embodiment of the present invention The flowchart which shows a series of flows of the manufacturing method of the semiconductor device as described in 2nd Embodiment in this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 CMP apparatus 102 Data collection apparatus 103 Data (waveform) behavior abnormality determination apparatus 104 Data primary arrangement apparatus 105 Primary data storage apparatus 106 Statistical value calculation apparatus 107 Statistical value data storage apparatus 108 Statistical value abnormality determination apparatus 109 Maintenance time calculation apparatus 110 MES (Manufacturing execution system)
111 Maintenance management system

Claims (5)

A process in which the manufacturing execution system instructs the semiconductor device to perform the process;
A process of collecting data at high speed in real time from a semiconductor device being processed;
Performing an abnormality determination in real time of the collected data;
Transmitting the collected data abnormality to the manufacturing execution system;
Organizing the collected data into a wafer-by-wafer data group at the end of wafer processing;
A process of storing data organized in wafer units in a predefined format;
Calculating statistical values from the data arranged in units of wafers;
Storing the calculated statistical values in a predefined format;
Performing an abnormality determination on the statistical value;
Telling the manufacturing execution system of the abnormality of the statistical value;
A step of calculating an optimal maintenance time of the semiconductor device from the trend of the statistical values;
And a step of notifying the maintenance management system of the calculated maintenance time. 2. The manufacturing of a semiconductor device according to claim 1, wherein in the step of notifying the abnormality of the data collected at high speed in real time from the semiconductor device being processed, the abnormality is notified to the manufacturing execution system for each combination of recipe and product type. Method. 2. The method of manufacturing a semiconductor device according to claim 1, wherein in the step of notifying the abnormality of the statistical value, the abnormality is notified to the manufacturing execution system in units of combinations of recipes and varieties. A manufacturing execution system that is informed of abnormalities in data collected at high speed in real time from a semiconductor device that is in process processing determines whether or not lot processing is possible by a combination of a recipe and a product type, and performs tracking of the lot. Item 14. A method for manufacturing a semiconductor device according to Item 1. 2. The method of manufacturing a semiconductor device according to claim 1, wherein the manufacturing execution system to which the abnormality of the statistical value is transmitted determines whether or not lot processing is possible based on a combination of a recipe and a product type, and performs tracking of the lot.

Images