JP2011238769A - Process control device and process control method, control program, readable recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance yield and device performance by optimizing deposition conditions (process parameters).SOLUTION: A process control device comprises process parameter collection means 111 for collecting process parameters containing a series of feature amounts of fluctuation period and stable period during production of products for every production sequence, feature amount calculation means 112 for extracting the feature amount of collected process parameters that is the feature amount of fluctuation period and stable period, quality information collection means 113 for collecting quality information indicating yield and performance of products thus produced, and feature amount determination means 114 for determining the feature amount of process parameters that most contribute to yield and performance by performing correlation analysis using the feature amount of process parameters and the quality information.

Description

  The present invention relates to a process management apparatus for analyzing data in process management for grasping a relation between data widely handled in the industry and extracting a meaningful result for producing an industrially superior result, and the same. The present invention relates to a process management method, a control program for causing a computer to execute each step of the process management method, and a computer-readable recording medium on which the control program is recorded.

  Conventionally, there is a process management method for efficiently extracting a correlation between data that is included in data stored in a computer system but cannot be easily detected at a glance and is buried. Such a process management technique is disclosed in Patent Document 1.

  FIG. 9 is a block diagram illustrating a configuration example of a main part of an abnormality extraction device in a conventional processing process disclosed in Patent Document 1.

  In FIG. 9, a conventional processing process process abnormality extraction device 100 includes an input device 101 for inputting parameters and manufacturing conditions necessary for performing abnormality extraction, quality result information I1, troubles with the manufacturing device / inspection machine, Storage device 102 in which device history information I2, manufacturing history information I3, and analysis result information I4 including information such as maintenance is stored, central processing device 103 that performs abnormality extraction, and data analysis that belongs to central processing device 103 A multi-stage multivariate analysis unit 104 and a data output device 105 for displaying or printing an abnormality extraction result or the like are provided.

  The multi-stage multivariate analysis unit 104 includes a data search unit A for searching for necessary information from the storage device 102, an analysis data generation unit B for generating analysis data from the searched data, and the cause of the abnormality. It has an automatic extraction processing unit C for extraction, a central control unit D for controlling these processes, and a work data storage unit E for storing and transferring work data of each process.

  In the diffusion process of this machining process step, as shown in FIG. 10, the lots L1 to L6 are sequentially sent to the devices DA, DB, DC and manufactured (processed) at the times indicated by the black circles, and the devices DA, DB , DC shall be maintained at the times indicated by maintenance MA, MB and MC, respectively. Therefore, manufacturing history information of lots L1 to L6, maintenance records for the devices DA, DB, and DC, and device history information that is a record of trouble occurrence / release when a trouble TC occurs in the device DC are accumulated. Is done. Further, the yields Y1 to Y6 and the electrical characteristics of the lots L1 to L6 are measured in the inspection process, and recorded as quality result information.

  In such a diffusion step, when the average yield of a certain product type falls below the target value, a process for extracting a factor that has influenced the yield reduction is performed.

  The multistage multivariate analysis means 104 uses the manufacturing history information including trouble and maintenance information related to the manufacturing apparatus and the inspection apparatus, and the quality result information indicating the product yield and the electrical characteristic value to By analyzing the multivariate analysis of the causal relationship in multiple stages, the factors that hinder the yield and device performance are extracted. The method has the feature that the number of variables used in one analysis is fixed, the factors are narrowed down using a known variable increase / decrease method, and the final data analysis is performed only with the items narrowed down in each analysis. .

JP 2002-110493 A

  In the above-described conventional configuration disclosed in Patent Document 1, although narrowing down (ranking) device data that is an abnormal factor for yield and device performance is performed by known multi-stage correlation analysis and multiple regression analysis, Individual differences in equipment (which unit of which equipment was used, how much the average gas flow rate is, and what is the temperature and pressure, etc.), maintenance time, trouble content and occurrence time, etc. Only using device history information, manufacturing history information such as lot (when it was manufactured), and quality result information such as yield and electrical characteristics, it is sensitive to manufacturing variations in multi-layered film generators such as semiconductor manufacturing. There is a problem that it is not possible to sufficiently and accurately narrow down defective factors for the yield and device performance that react to the device.

  That is, in the apparatus history information, for example, individual differences between apparatuses, the process film formation conditions include several hundred management items (process parameters) such as gas flow rate, temperature, and pressure. In the configuration, using only the fluctuation of the stable period as the feature value (average value, maximum, minimum) for the indicated value of each management item is sensitive to manufacturing variations such as multi-layered film generators such as semiconductor manufacturing. There has been a problem that it is not possible to sufficiently and accurately narrow down defective factors for the reaction yield and device performance.

  The present invention solves the above-described conventional problems, and is sensitive to manufacturing variations in a multi-layered film generation apparatus such as a semiconductor manufacturing in a process that requires stable supply management of film forming conditions such as a crystal growth apparatus. A process management apparatus and a process management apparatus capable of narrowing down the cause of failure to the yield and device performance that reacts to the target sufficiently and accurately and improving the yield and device performance by optimizing the film formation conditions (process parameters) It is an object of the present invention to provide a process control method, a control program for causing a computer to execute each step of the data analysis method, and a computer-readable readable recording medium on which the control program is recorded.

  A process management apparatus of the present invention is a process management apparatus that manages a product manufacturing process that includes at least one process parameter in manufacturing conditions of a semiconductor manufacturing apparatus and that is manufactured under manufacturing conditions in which a plurality of manufacturing sequences exist. Process parameters including a series of characteristic quantities of a variation period and a stable period following the process, and a process parameter collecting means for collecting the process parameters for each production sequence, and characteristics of the collected process parameters Feature quantity calculating means for extracting feature quantities of the fluctuation period and the stable period, quality information collecting means for collecting quality information indicating the yield and performance of the manufactured product, and feature quantities of the process parameters And the correlation analysis using the quality information, the pro- gram with the largest contribution to the yield and performance is obtained. Those having a characteristic data determination means for determining a characteristic quantity of the scan parameters, the object is achieved.

  Preferably, the feature amount calculating means in the process management apparatus of the present invention sets a transition period as a transition period until the process parameter reaches a predetermined range of a predetermined instruction value, and the transition period and the predetermined range. And at least the arrival time to the stable period among the rise time of the process parameter, the arrival time to the stable period, and the fall time of the process parameter as the feature amount of the process parameter Extract.

  Further preferably, in the transition period, the feature amount calculation means in the process management apparatus of the present invention calculates the upper limit value, the lower limit value, the median value, the standard deviation, and the upper limit value and the lower limit value within the transition period. At least one of the exceeding number of cycles is extracted as a feature amount of the process parameter.

  Further preferably, the feature amount calculation means in the process management apparatus of the present invention, in the stable period, the upper limit value, the lower limit value, the median value, the standard deviation, and the upper limit value and the lower limit value within the stable period. At least one of the exceeding number of cycles is extracted as a feature amount of the process parameter.

  Further preferably, when the product manufacturing process in the process management apparatus of the present invention is processed by a plurality of processes, the feature amount determination means is characterized by the process parameter having the greatest contribution to the yield and performance in the plurality of processes. Determine the amount.

  Further preferably, in the feature amount determination means in the process management device of the present invention, the product yield information and the influence on the performance from the product yield information as the quality information and the feature amount of the process parameter associated with each product. Multiple regression analysis is used as a method for calculating the degree.

  The process management method of the present invention is a process management method for managing a product manufacturing process that includes at least one process parameter in a manufacturing condition of a semiconductor manufacturing apparatus and that is manufactured under a manufacturing condition in which a plurality of manufacturing sequences exist. A parameter collection means, which is a process parameter including a series of feature quantities of a variation period and a stable period following the production of the product, and collecting the process parameters for each production sequence; and a feature quantity calculation Means for extracting feature quantities of the collected process parameters, which are feature quantities of the fluctuation period and the stable period; and a quality information collection means is a quality indicating the yield and quality of the manufactured product. A step of collecting information, and a feature amount determination means, using the feature amount of the process parameter and the inspection information. By performing analysis, which has a step of determining the characteristic quantity of contribution is the largest process parameters to the yield and performance, the object is achieved.

  Preferably, in the feature quantity calculating step in the process management method of the present invention, the fluctuation period until the process parameter reaches a predetermined range of a predetermined instruction value is a transition period, and the transition period and the predetermined range And at least the arrival time to the stable period among the rise time of the process parameter, the arrival time to the stable period, and the fall time of the process parameter as the feature amount of the process parameter Extract.

  Further preferably, in the transition period, the feature amount calculation step in the process management method of the present invention includes an upper limit value, a lower limit value, a median value, a standard deviation, and the upper limit value and the lower limit value within the transition period. At least one of the exceeding number of cycles is extracted as a feature amount of the process parameter.

  Further preferably, the characteristic amount calculation step in the process management method of the present invention includes, in the stable period, an upper limit value, a lower limit value, a median value, a standard deviation, and the upper limit value and the lower limit value within the stable period. At least one of the exceeding number of cycles is extracted as a feature amount of the process parameter.

  Further preferably, when the product manufacturing process in the process management method of the present invention is processed by a plurality of processes, the step performed by the feature amount determining means is a process having the largest contribution to yield and performance in the plurality of processes. The feature amount of the parameter is determined.

  Further preferably, in the step performed by the feature amount determination means in the process management method of the present invention, the product yield information as the quality information associated with each product and the feature amount of the process parameter to the product yield and performance. The multiple regression analysis is used as a method of calculating the degree of influence.

  The control program according to the present invention describes a processing procedure for causing a computer to execute each step of the process management method according to the present invention, thereby achieving the above object.

  The readable recording medium of the present invention is a computer-readable medium in which the control program of the present invention is stored, whereby the above object is achieved.

  With the above configuration, the operation of the present invention will be described below.

  For example, in the film forming process of the semiconductor manufacturing process, there are hundreds of management items (process parameters) such as gas flow rate, temperature and pressure as the processing conditions. When extracting management items (process parameters) that are abnormal factors that cause product performance abnormalities and yield reduction, it is important that anyone can easily and accurately extract abnormalities according to the actual situation.

  In the present invention, process parameters including a series of feature values of a variation period when manufacturing a product and a subsequent stable period, the process parameter collecting means for collecting the process parameters for each manufacturing sequence, Feature quantity calculation means for extracting feature quantities of process parameters that are variable periods and subsequent stable periods; quality information collection means for collecting quality information indicating the yield and performance of manufactured products; and process And a feature amount determining means for determining the feature amount of the process parameter having the greatest contribution to the yield and performance by performing correlation analysis using the feature amount of the parameter and the quality information.

  As a result, in processes that require stable supply management of film deposition conditions such as crystal growth equipment, not only fluctuations in the stable period but also fluctuations in the transition period until the stable period are considered as feature quantities of the process parameters. Therefore, it is possible to more accurately and accurately narrow down the yield factors and device performance factors that react sensitively to manufacturing variations such as in the production of semiconductor multi-layered films, etc., and optimize film deposition conditions (process parameters). Yield and device performance can be improved.

  As described above, according to the present invention, not only the fluctuation of the stable period but also the fluctuation of the transition period until the stable period is processed in a process that requires stable supply management of the film forming conditions such as the crystal growth apparatus. Considering it as a parameter feature value, it is possible to more accurately and accurately narrow down the yield factors and device performance factors that react sensitively to manufacturing variations such as semiconductor multilayer manufacturing equipment, etc. Yield and device performance can be improved by optimizing parameters.

It is a block diagram which shows the principal part hardware structural example of the process management apparatus in one Embodiment of this invention. It is a flowchart for demonstrating operation | movement of the process management apparatus of FIG. (A) And (b) is a figure which shows a series (1) of a graph from the variation period of a process parameter to a stable period when the horizontal axis is time and the vertical axis is the value of a management item (process parameter). is there. (A) and (b) are a series of graphs (part 2) from the variation period of the process parameter to the stable period and further to the variation period when the horizontal axis is time and the vertical axis is the value of the management item (process parameter). FIG. It is a block diagram which shows schematic structure of the graph drawing apparatus used applying the feature-value determination means of FIG. It is the figure which illustrated the display content displayed on each area | region of the graph drawing template in the graph drawing template part of FIG. It is the figure which illustrated the content of the tied data table which memorize | stored the data used as the object of a graph drawing by the graph drawing apparatus of FIG. It is a flowchart of the graph drawing method by the graph drawing apparatus of FIG. It is a block diagram which shows the principal part structural example of the abnormality extraction apparatus in the conventional process process currently disclosed by patent document 1. FIG. It is a conceptual diagram of the diffusion process used as the abnormality extraction object by the abnormality extraction apparatus of FIG.

  Hereinafter, embodiments of a process management apparatus and a process management method of the present invention will be described in detail with reference to the drawings. In addition, the length, shape, etc. in each figure are not limited to the structure to illustrate from a viewpoint on drawing creation.

  FIG. 1 is a block diagram illustrating a hardware configuration example of a main part of a process management apparatus according to an embodiment of the present invention.

  In FIG. 1, a process management apparatus 10 according to the first embodiment includes a control unit 11 including a CPU (Central Processing Unit) 11 that performs overall control, and a keyboard and a mouse for inputting input commands to the control unit 11. , A touch panel and a pen input device, and an operation unit 12 such as an input device that receives and inputs via a communication network (for example, the Internet or an intranet), and an initial screen, a selection scene, and a control result screen by the control means 11 on the display screen And a display unit 13 for displaying various screens such as an operation input screen, a ROM 14 as a computer-readable readable recording medium storing a control program and its data, and the control program and its data at the time of startup. Thus, a work memory that reads and stores data for each control by the control means 11 And a RAM15 as a storage unit that acts faster as Li.

  The control means 11 is a process parameter for manufacturing a product based on the control program in the ROM 14 and its data, the process parameter collecting means 111 for collecting the process parameter for each manufacturing sequence, and the collected process parameter Feature quantity calculating means 112 for extracting (or calculating) feature quantities, quality information collecting means 113 for collecting quality information (or inspection information) for calculating yield and quality of manufactured products, process parameter By performing correlation analysis using the feature quantity and quality information (or inspection information), each function of the feature quantity determination means 114 that determines the feature quantity of the process parameter having a large contribution to the yield is executed, and semiconductor manufacturing is performed. Including at least one process parameter in the manufacturing conditions of the device, It is adapted to manage production processes produced by the production conditions forming sequence is present.

  The process parameter collection unit 111 is a management item (process parameter) for manufacturing a product, and collects a process parameter for each manufacturing sequence. For example, process film formation conditions include hundreds of management items (process parameters) such as gas flow rate, temperature, and pressure.

  The feature amount calculation unit 112 sets the transition period until the predetermined process parameter reaches the indicated value as a transition period, and distinguishes between the transition period and the stable period after the predetermined process parameter reaches a certain range. Then, each arrival time (for example, how long the gas flow rate has reached a stable range) is calculated (or extracted) as a process parameter feature amount.

  In addition, the feature amount calculation unit 112 has an upper limit value (maximum value), a lower limit value (minimum value), and a median value in the transition period in a transition period until the predetermined process parameter reaches a certain range of the indicated value. The standard deviation and the number of cycles exceeding the upper limit value and the lower limit value are calculated (or extracted) as the process parameter feature amount. In addition, the feature amount calculation unit 112 has an upper limit value (maximum value), a lower limit value (minimum value), a center value in the stable period in a stable period after the predetermined process parameter reaches a certain range of the indicated value. The value, the standard deviation, and the number of cycles exceeding the upper limit value and the lower limit value are calculated (or extracted) as process parameter feature quantities.

  In short, FIGS. 3 (a) and 3 (b), 4 (a) and 4 (b) show the same thing, the horizontal axis is time, and the vertical axis is the value of the management item (process parameter). It is. For example, when the gas flow rate is used as a process parameter, a change in the gas flow rate from the start of flowing a predetermined gas into the chamber containing the wafer to the end of the flow is shown. The value A or less is regarded as “0”. The upper limit value B and lower limit value C of the fluctuation period, the target value D of the stable period, the upper limit value E and the lower limit value F of the stable period, and A to F are set. The feature quantity calculating means 112 calculates (or extracts), for example, a gas flow rise time G, a stable period arrival time H, and a fall time I as process parameters. In addition, the feature quantity calculating means 112 has, for example, the maximum value J and the minimum value K of the fluctuation period of the gas flow rate, the maximum value L and the minimum value M of the stable period, the median value O and the standard of the variable period and the stable period as process parameters. The deviation P is calculated (or extracted). Further, the feature quantity calculation means 112 calculates, for example, the number of times that the upper limit value B and the lower limit value C of the gas flow rate fluctuation period have been exceeded and the number of times that the upper limit value E and the lower limit value F of the stable period have been exceeded as process parameters (or Extract. Each feature affects the device yield and quality.

  The quality information collecting unit 113 collects quality information (inspection information) indicating the yield and performance of the manufactured product. For example, among photoluminescence data obtained by imaging reflected light from an epitaxial growth substrate after irradiating excitation light from above the epitaxial growth substrate obtained by epitaxially growing a compound semiconductor layer on a single crystal substrate, the light wavelength is the compound Since it varies depending on the composition and film thickness of the semiconductor epitaxial film, the data is used to manage the epitaxial layer that affects the composition and film thickness. In addition to the composition and film thickness, the light intensity is related to dislocations in the crystal of the epitaxial layer, stacking faults, contamination, and foreign matter. Therefore, inspection data of “light wavelength” and “light intensity” of the entire substrate can be used as quality information (inspection information). In this case, in the same production lot, inspection data of “light wavelength” and “light intensity” of the entire substrate can be used as each feature amount of process parameters and quality information (inspection information).

  The feature quantity determination unit 114 uses process parameter feature quantities as a plurality of derived management items, for example, “gas flow rate”, and quality information (inspection information) indicating product yield and performance, for example, “light intensity”. The multi-step correlation analysis and multiple regression analysis are performed on the causal relationship between the two types of information, the feature quantities of each process parameter are ranked, and the feature quantities of the process parameters that have the greatest contribution to product yield and quality. And an optimum management value of the manufacturing condition is extracted based on the feature amount of the extracted process parameter.

  The ROM 14 as a readable recording medium may be composed of a hard disk, a formable optical disk, a magneto-optical disk, a magnetic disk, an IC memory, and the like. The control program and its data are stored in the ROM 14, but the control program and its data may be downloaded to the ROM 14 from another readable recording medium or via wireless, wired or the Internet.

  By the process management apparatus 10 having the above-described configuration, when the average yield or performance of a certain product type decreases below the target value in the inspection data (inspection information), the factor (process parameter Processing for extracting (feature amount) will be described.

  FIG. 2 is a flowchart for explaining the operation of the process management apparatus of FIG.

  As shown in FIG. 2, first, in step S1, the process parameter collection unit 111 forms a growth film by gas, for example, as a management item for each manufacturing sequence as fine time-series data from the manufacturing apparatus or measuring instrument. In the case of a single layer (including CVD), a series of process parameters such as a gas flow rate including a fluctuation period and a stable period, temperature and pressure are collected from the device storage unit (ROM 14).

  Next, in step S2, the quality information collection unit 113 collects quality information (inspection information), for example, “light intensity” data for calculating or extracting the yield and performance of the manufactured product. The quality information (inspection information) is inspection information such as light intensity and light wavelength, for example, in the case of reflected light data. If the quality information (inspection information) does not satisfy a predetermined threshold value, the quality information (inspection information) becomes defective (NG), which affects the yield.

  Subsequently, in step S3, the feature amount calculation unit 112 uses the collected process parameters as the management items, and continuously fine data including the processing variation period (transition period) and the stable period for each management item. The feature amount is extracted from. As the feature amount, the feature amount calculation means 112 has been described above, but as shown in FIG. 4B, as the actual measurement value, the process parameter rise time G, stable period arrival time H, fall time I The maximum value J and minimum value K of the fluctuation period, the maximum value L and minimum value M of the stable period, the median value O and the standard deviation P of the stable period, and the number of times the upper limit value B and the lower limit value C have been exceeded And so on.

  For example, in semiconductor manufacturing, there is a film growth technique using a plurality of gases in a multilayer laminated film generation apparatus. From these manufacturing apparatuses, it is possible to output changes in management items (process parameters) exceeding several hundreds indicating the processing state of the manufacturing process as time-series information with fine granularity.

  In the conventional process management method, only the fluctuation of the stable period was used as the feature amount (average value, maximum, minimum, etc.). In this embodiment, the manufacturing process is focused on the change at the time of gas transition in the multilayer film generator. From the fine-grained time-series equipment data collected from equipment and measuring instruments, the maximum value, minimum value, median value, elapsed time, and upper limit value in the fluctuation period (transition period) of each step for each management item (process parameter) By extracting various feature amounts such as the number of times B and the lower limit C are exceeded, the feature amount of the process parameter having the greatest influence is extracted from one management item.

  Here, a description will be given of a method of extracting feature values of the most useful process parameters that have influenced yield and performance degradation from management items (process parameters) collected at the time of manufacturing a certain product.

  4 (a) and 4 (b) express one of hundreds of management items (process parameters). For the transition data expressed in this time series, “setting” “Values A to F” are given, and “feature amounts G to R” are calculated or extracted. The feature values shown in Table 1 are calculated or extracted for each management item (process parameter).

  The feature quantity for each management item (process parameter) extracted by the above method and the inspection information or quality information indicating the yield and performance (quality) of the product are linked for each lot, and preparation for correlation analysis is made.

  After that, in step S4, the feature quantity determination means 114 uses the derived process parameter feature quantities as a plurality of management items, for example, “gas flow rate”, and quality information (inspection information) indicating product yield and performance, for example, “ Using the “light intensity”, the causal relationship between both pieces of information is subjected to a multi-stage correlation analysis, and in step S5, the feature amount determination unit 114 ranks the feature amounts of the process parameters. In step S6, items related to process management are picked up among the feature amounts of the ranked process parameters. If items related to process management are not picked up in each feature quantity (NO), the process returns to step S4. While the items related to process management are picked up in each feature amount (YES), the feature amount of the process parameter that contributes most to the product yield and quality is extracted as the optimum control value of the manufacturing conditions Then, the process proceeds to the next step S7.

  Further, in step S7, the optimum management value of the extracted manufacturing condition is applied to the in-process manufacturing apparatus.

  Therefore, the process management method using the process management apparatus 10 according to the present embodiment is a process parameter when the process parameter collection unit 111 manufactures a product based on the control program in the ROM 14 and its data. A step of collecting the process parameter for each sequence, a step of extracting the feature amount of the collected process parameter by the feature amount calculating unit 112, and an inspection information collecting unit 113 calculating the yield and quality of the manufactured product. The step of collecting quality information (inspection information) to be performed, and the feature amount determination means 114 performs correlation analysis using the feature amount of the process parameter and the inspection information, so that the product yield and the degree of contribution to the quality can be increased. Extract optimal control values of manufacturing conditions as feature values of large process parameters It is intended to execute the steps in a computer system.

  As described above, by performing multi-level correlation analysis on the derived feature quantity for each management item (process parameter), which management item (process parameter) affects the yield and performance of the product. It is possible to rank and feed back to the management parameters of the processing conditions of the manufacturing apparatus in the process.

  That is, by performing a correlation analysis of factors that affect the yield and performance of the product by the feature amount of the process parameter, it can be expressed as a correlation coefficient, a dispersion ratio, a statistic, or a numerical value. By expressing the causal relationship between the yield and device performance impediments as numerical values, it is possible to rank them, and in the past, only the stable period was considered, but the gas fluctuation period (transient period) of the device It can be seen that the characteristic amount such as the peak value of the gas flow rate in (1) greatly affects the yield and performance of the product. As a result, detailed and effective failure factors including not only the stable period but also the variable period can be more sufficiently and accurately narrowed down, and improvement is achieved.

  When the product manufacturing process is processed by a plurality of processes, the step performed by the feature amount determination unit 114 determines the feature amount of the process parameter having the greatest contribution to the yield and performance in the plurality of processes. In addition, in the step performed by the feature amount determination means 114, the product yield information and the feature amount of the process parameter associated with each product are used to calculate the influence on the product yield and performance. Use.

  Here, the following graph drawing apparatus of FIG. 5 can be employed as a technique for performing a multi-stage correlation analysis by associating the extracted feature values of each management item with the product yield and performance.

  That is, the feature amount determination unit 114 includes the feature amounts of the derived process parameters as a plurality of management items, for example, “gas flow rate”, and quality information (inspection information) indicating the yield and performance of the product, for example, “light intensity”. , The multi-step correlation analysis of the causal relationship between the two types of information, ranking the feature quantities of each process parameter, and as the process parameter feature quantity having the greatest contribution to product yield and quality, the manufacturing conditions The optimum management value is extracted, and this feature amount determining means 114 can be applied to the graph drawing apparatus shown in FIG.

  As an example of a technique for performing a multi-stage correlation analysis by linking the extracted feature quantities of each management item, for example, “gas flow rate”, and quality information (inspection information) indicating product yield and performance, for example, “light intensity” The graph drawing apparatus will be described below with reference to FIGS.

  FIG. 5 is a block diagram illustrating a schematic configuration example of a graph drawing apparatus to which the feature amount determination unit 114 of FIG. 1 is applied.

  In FIG. 5, a graph drawing apparatus 1 that constitutes an example of the feature amount determining unit 114 includes a link data table 2 described later, a graph drawing template unit 3 that performs display using a graph drawing template 4 described later, The variable selection operation unit 1f for the operator to specify the analysis target from the multivariate stored in the attached data table 2 and the analysis operation unit 1g for the operator to specify the bivariate on the display screen ing.

  The graph drawing template unit 3 includes data editing means 1a for editing M variables and N variables to be analyzed selected by the variable selection operation unit 1f, and (M × N) statistics (for example, phase variables). Statistic calculation means 1b for calculating (number of relations), data arrangement means 1c for arranging the calculated (M × N) statistics as matrix elements of N rows and M columns, and an analysis operation unit 1g on the display screen. The graph output means 1d for drawing the relationship between the two variables specified by, and the factor analysis means 1e for analyzing the relationship between the multivariates.

  Each unit of the graph drawing template unit 3 is configured by a CPU (Central Processing Unit) that operates according to a predetermined program. The association data table 2 is configured by an external storage device such as a hard disk drive. The variable selection operation unit 1f is configured by an input device such as a mouse or a keyboard. The analysis operation unit 1g is configured by a known touch panel LCD (liquid crystal display device) that performs input by touching a certain part of the display screen with a pen, for example. Of course, the analysis operation unit 1g may be an input method in which a cursor on the display screen is moved with a mouse and a certain location in the display screen is designated by clicking.

  FIG. 6 is a diagram exemplifying display contents displayed in the regions 4a to 4d of the graph drawing template 4 in the graph drawing template unit 3 of FIG.

  As shown in FIG. 6, the graph drawing template unit 3 works as each display area setting unit, and the graph drawing template 4 set on the display screen has the (M × N) statistics in the lower right in this example. A matrix display area 4c for displaying the quantities in a matrix, a first variable name display area 4a on the upper right, a second variable name display area 4b on the lower left, and a graph display area 4d on the upper left, respectively. The areas 4a to 4d are incorporated into one rectangular frame displayed on the same display screen.

  In the matrix display area 4c, the data arrangement function unit 1c and the statistic calculation function unit 1b function as a statistic display processing unit, and (M × N) statistics, in this example, as shown in FIG. Further, “performance” Q1, Q2,..., QM, which are M variables belonging to the first data group 2a, and “factors” F1, F2,..., N variables, which belong to the second data group 2b. , FN (M × N) correlation coefficients are displayed side by side as matrix elements of N rows and M columns. Thereby, the worker can easily select, for example, a correlation coefficient that is large from the displayed correlation coefficients.

  Here, the association data table 2 that stores data to be graph drawn by the graph drawing device 1 will be described.

  In this example, as shown in FIG. 7, the associating data table 2 includes “performance” Q1, as M variables belonging to the first data group 2a for each product lot Lt1, Lt2,. Q2,..., QM and “factors” F1, F2,..., FN as N variables belonging to the second data group 2b are associated (linked) with each other and stored in a predetermined storage unit. ing.

  In an example in the field of semiconductor manufacturing, the data represented by “performance” Q1, Q2,..., QM includes device performance obtained by, for example, wafer testing and electrical characteristics. It corresponds to the quality information (inspection information) indicated, for example, “light intensity”. In addition, the data represented by the “factors” F1, F2,... FN is oxide film thickness, line width, inspection process data, etc., but here, it corresponds to a feature quantity of a process parameter, for example, “gas flow rate”. is doing.

  Each product lot is specified by L “product IDs” Lt1, Lt2,..., LtL as a common identifier 2c. That is, in this association data table 2, “performance” Q1, Q2,..., QM and N as M variables for each row by L identifiers (lot numbers) Lt1, Lt2,. Are associated with "factors" F1, F2, ..., FN.

  Next, returning to FIG. 6, in the first variable name display area 4a, the graph drawing template unit 3 works as a variable name display processing unit, and names of M variables belonging to the first data group 2a. In this example, “performance Q1”, “performance Q2”,..., “Performance QM” are displayed side by side in the row direction corresponding to each column of the matrix display area 4c.

  Similarly, in the second variable name display area 4b, the names of N variables belonging to the second data group 2b, in this example, “Factor F1,” “Factor F2,”..., “Factor FN” They are displayed side by side in the column direction corresponding to each row of the matrix display area 4c. As a result, the operator sees the names of the variables displayed in the first and second variable name display areas 4a and 4b, so that the variables belonging to which data group and which data group belong to the matrix display area 4c. It can be easily recognized whether the statistics between the variables are displayed.

  Further, in the graph display area 4d, the graph output function unit 1d functions as a graph display processing unit, and between the two variables between the variables belonging to the first data group 2a and the variables belonging to the second data group 2b. A graph image representing the relationship is displayed. In this example, the operator uses the analysis operation unit 1g as a matrix element designation unit, and among the matrix elements displayed in the matrix display area 4c on the display screen, the matrix element 3b in the eighth row and the fourth column (that is, the phase element). In accordance with the designation of the relation number “0.55”), the scatter representing the correlation between the performance Q4 corresponding to the column (fourth column) and the factor F8 corresponding to the row (eighth row) FIG. 3c is displayed.

  In this example, the scatter diagram 3c shows the variable “performance Q4” belonging to the first data group 2a as the first coordinate axis and the variable “belonging to the second data group 2b”. The horizontal axis (X axis) as the second coordinate axis representing the factor F8 ”is displayed as a reference. Each point 3d in the scatter diagram 3c represents data of a certain product lot. The vertical axis and the horizontal axis may be displayed interchangeably.

  As described above, in the graph drawing device 1, the graph image corresponding to the matrix element (designated 2) is linked with the analysis operation in which the operator designates the matrix element in the matrix display area 4c by the analysis operation unit 1g. The relationship between the variables) is displayed in the graph display area 4d. According to the input method for designating any one of the matrix elements displayed in the matrix display area 4c by the analysis operation unit 1g described above, the variables belonging to the first data group 2a and the variables belonging to the second data group 2b. Can be efficiently specified simultaneously by one operation (for example, one touch).

  Next, the graph drawing method by this graph drawing apparatus 1 is demonstrated based on the flowchart of FIG. As this premise, it is assumed that the contents of the association data table 2 have been edited in advance by the operator as illustrated in FIG.

  As shown in FIG. 8, when the operation of the graph drawing apparatus 1 starts, first, in step S11, the initial state of only the table area of the graph drawing template 4 is displayed on the display screen.

  Subsequently, in step S12, variable names and matrix elements in the row direction and the column direction are displayed based on the contents of the association data table 2. In this example, “performance Q1”, “performance Q2”,..., “Performance Q10” are displayed in the first variable name display area 4a as names of M (= 10) variables belonging to the first data group 2a. Is displayed. Similarly, “factor F1”, “factor F2”,..., “Factor F10” are displayed in the second variable name display area 4b as the names of N (= 10) variables belonging to the second data group 2b. indicate. Furthermore, in the matrix display area 4c, “performance” Q1, Q2,..., Q10 which are 10 variables belonging to the first data group 2a and “factor” which is 10 variables belonging to the second data group 2b are displayed. (M × N) = (10 × 10) correlation coefficients between F1, F2,..., F10 are displayed side by side as 10 × 10 matrix elements.

  Thereafter, in step S13, the graph drawing template unit 3 determines whether or not there is a matrix element designation using the analysis operation unit 1g by the operator. If a matrix element is specified (YES in step S13), the process proceeds to step S14, a graph focusing on the specified matrix element is calculated, and the graph image is displayed in the graph display area 4d. . In this example, in response to the fact that, for example, the matrix element in the eighth row and the fourth column (that is, the correlation coefficient “0.55”) is specified among the matrix elements displayed in the matrix display area 4c on the display screen. Then, the scatter diagram 3c representing the correlation between the performance Q4 corresponding to the column (fourth column) and the factor F8 corresponding to the row (eighth row) is displayed.

  In this case, the operator simply designates any one of the matrix elements displayed in the matrix display area 4c on the display screen via the analysis operation unit 1g. Correspondingly, a variable belonging to the first data group 2a and a variable belonging to the second data group 2b can be designated at the same time. The relationship between the designated two variables can be displayed in an easy-to-see manner. Therefore, the operator can efficiently extract a desired relationship between two variables from a large amount of data, and can express it clearly on one screen. As a result, ranking by the relationship between the two variables (correlation coefficient in this example) can be easily performed, and the trade-off relationship between the multivariables can be easily analyzed.

  Subsequently, in step S15, the graph drawing template unit 3 determines whether or not there is a variable name designation using the analysis operation unit 1g by the operator as a variable name designation unit. Specifically, whether or not the variable name (in this example, “performance Q4”) displayed in the first variable name display area 4a is designated, or the variable displayed in the second variable name display area 4b It is determined whether or not a name (“Factor F8” in this example) is designated. If no variable name is designated (NO in step S15), the process immediately returns to step S13. On the other hand, if a variable name is specified (YES in step S15), the graph drawing template unit 3 works as an arrangement order processing unit, and rearranges matrix elements for the specified variable name in the order of absolute values of statistics. Specifically, when the variable name displayed in the first variable name display area 4a is designated, N statistics arranged in the column direction with respect to the variable “performance Q4” for which the name is designated have absolute values. The statistics displayed in the matrix display area 4c are rearranged in units of rows so as to be in order (descending order in this example). Alternatively, when the variable name 7b displayed in the second variable name display area 4b is designated, the M statistics arranged in the row direction for the variable “factor F8” for which the name is designated are in the order of absolute values ( The statistics displayed in the matrix display area 4c are rearranged in units of columns so that they are in descending order in this example. As a result, the operator can easily grasp the relationship between the variables of the two groups, in this example, the ranking based on the correlation coefficient, at a glance.

  Therefore, it is possible to easily rank and extract process parameters having a large contribution on the graph drawing apparatus shown in FIG.

  For example, “light intensity” which is quality information is one item of performance belonging to the first variable name display area 4a, and the feature value “maximum value of gas flow rate” is one item of the factor of the second variable name display area 4b. It becomes. By selecting each item name, since it is a numerical value, ranking (rearrangement in ascending order or descending order) is performed, and the most influential factor can be easily extracted.

  As described above, according to the present embodiment, a process parameter collecting unit that collects a process parameter for each manufacturing sequence, which is a process parameter including a series of feature amounts of a variation period and a stable period when a product is manufactured. 111, feature quantity calculation means 112 for extracting the feature quantities of the collected process parameters, which are variable period and stable period, and quality information collection for collecting quality information indicating the yield and performance of the manufactured product Means 113 and feature quantity determining means 114 for determining the feature quantity of the process parameter having the greatest contribution to the yield and performance by performing correlation analysis using the feature quantity and quality information of the process parameter. Yes.

  In this case, the process film formation conditions include hundreds of management items (process parameters) such as gas flow rate, temperature, and pressure. Conventionally, only the fluctuation of the stable period is used as the feature amount (average value, maximum, minimum, etc.) for the indicated value of each management item. On the other hand, in the present embodiment, not only the fluctuation of the stable period but also the change at the time of gas transition (fluctuation period) of the multilayer laminated film production apparatus or the like, the fine particle size collected from the apparatus and the measuring instrument is fine. From the time-series device data, the peak value, elapsed time, number of cycles, etc. in the fluctuation period (transition period) of each step for each management item are extracted as feature amounts, and the feature amount having the greatest influence on one management item It is possible to easily extract feature quantities having a large influence.

  As a result, the factors affecting the product yield and performance can be ranked in detail from the feature amount of a series of multiple management items in the fluctuation period and the subsequent stable period. It is possible to narrow down more sufficiently and accurately the yield factor that reacts sensitively to the manufacturing variation of the generation device and the like, and the cause of failure for the device performance. Therefore, feedback to the processing conditions in the manufacturing process can be realized, and product performance and yield can be improved.

  In addition, as mentioned above, although this invention has been illustrated using preferable embodiment of this invention, this invention should not be limited and limited to this embodiment. It is understood that the scope of the present invention should be construed only by the claims. It is understood that those skilled in the art can implement an equivalent range based on the description of the present invention and the common general technical knowledge from the description of specific preferred embodiments of the present invention. Patents, patent applications, and documents cited herein should be incorporated by reference in their entirety, as if the contents themselves were specifically described herein. Understood.

  The present invention relates to a process management apparatus for analyzing data in process management for grasping a relation between data widely handled in the industry and extracting a meaningful result for producing an industrially superior result, and the same. In the field of a process management method, a control program for causing a computer to execute each step of the process management method, and a computer-readable readable recording medium on which the control program is recorded, film formation conditions such as a crystal growth apparatus are stable. In processes where supply management is required, not only fluctuations in the stable period but also fluctuations up to the stable period are considered as feature values of the process parameters, so it is sensitive to manufacturing variations such as semiconductor multilayer production equipment More fully and accurately narrow down failure factors for reacting yield and device performance Bets are made, it is possible to improve the yield and device performance by optimizing the film forming conditions (process parameters).

10 Process management device 11 Control means (CPU)
111 process parameter collection means 112 feature quantity calculation means 113 quality information collection means 114 feature quantity determination means 12 operation section 13 display section 14 ROM
15 RAM
DESCRIPTION OF SYMBOLS 1 Graph drawing apparatus 2 Linking data table 3 Graph drawing template part 4 Graph drawing template 1f Variable selection operation part 1g Analysis operation part 1a Data editing means 1b Statistics amount calculation means 1c Data arrangement means 1d Graph output means 1e Factor analysis means 4c Matrix Display area 4a First variable name display area 4b Second variable name display area 4d Graph display area

Claims (14)

A process management apparatus that manages a product manufacturing process that includes at least one process parameter in manufacturing conditions of a semiconductor manufacturing apparatus and that is manufactured under manufacturing conditions in which a plurality of manufacturing sequences exist,
A process parameter including a series of feature values of a variation period and a stable period following the manufacture of the product, the process parameter collecting means for collecting the process parameter for each manufacturing sequence, and the collected process parameter Feature amount calculation means for extracting feature amounts of the variation period and the stable period, quality information collection means for collecting quality information indicating the yield and performance of the manufactured product, and features of the process parameters A process management apparatus comprising: a feature amount determination unit that determines a feature amount of a process parameter having the greatest contribution to the yield and performance by performing a correlation analysis using the amount and the quality information.   The feature amount calculating means defines a transition period until the process parameter reaches a predetermined range of a predetermined instruction value as a transition period, and includes a transition period and a stable period after reaching the predetermined range. 2. The process management apparatus according to claim 1, wherein the process management device extracts the process parameter rising time, the time to reach the stable period, and the time to reach the stable period as feature quantities of the process parameter. .   The feature amount calculating means calculates, in the transition period, at least one of an upper limit value, a lower limit value, a median value, a standard deviation, and a cycle number exceeding the upper limit value and the lower limit value within the transition period. The process management apparatus according to claim 2, wherein the process management apparatus is extracted as a feature amount of a process parameter.   In the stable period, the feature amount calculating means calculates at least one of an upper limit value, a lower limit value, a median value, a standard deviation, and the number of cycles exceeding the upper limit value and the lower limit value in the stable period. The process management apparatus according to claim 2, wherein the process management apparatus is extracted as a feature amount of a process parameter.   2. The process according to claim 1, wherein when the product manufacturing process is processed by a plurality of processes, the feature amount determination unit determines a feature amount of a process parameter having the greatest contribution to yield and performance in the plurality of processes. Management device.   The feature quantity determination means uses a multiple regression analysis as a method for calculating the product yield and performance influence from the product yield information as the quality information linked to each product and the feature quantity of the process parameter. Item 2. The process management apparatus according to Item 1. A process management method for managing a product manufacturing process that includes at least one process parameter in a manufacturing condition of a semiconductor manufacturing apparatus and that is manufactured under a manufacturing condition in which a plurality of manufacturing sequences exist,
A process parameter collecting means, which is a process parameter including a series of feature quantities of a variation period and a subsequent stable period when the product is manufactured, and collecting the process parameter for each manufacturing sequence;
A feature quantity calculating means for extracting the feature quantities of the collected process parameters and the feature quantities of the fluctuation period and the stable period;
A quality information collecting means for collecting quality information indicating a yield and quality of the manufactured product;
A step of determining a feature quantity of the process parameter having the greatest contribution to the yield and performance by performing a correlation analysis using the feature quantity of the process parameter and the inspection information. Management method.   In the feature amount calculating step, a fluctuation period until the process parameter reaches a predetermined range of a predetermined instruction value is set as a transition period, and the transition period and a stable period after reaching the predetermined range. The process management method according to claim 7, wherein at least the arrival time to the stable period among the rise time, the arrival time to the stable period, and the fall time of the process parameter is extracted as a feature quantity of the process parameter. .   In the transition period, the feature amount calculating step includes at least one of an upper limit value, a lower limit value, a median value, a standard deviation, and a cycle number exceeding the upper limit value and the lower limit value in the transition period. The process management method according to claim 8, wherein the process management feature is extracted as a feature amount of a process parameter.   In the stable period, the feature amount calculating step calculates at least one of an upper limit value, a lower limit value, a median value, a standard deviation, and a number of cycles exceeding the upper limit value and the lower limit value in the stable period. The process management method according to claim 8, wherein the process management feature is extracted as a feature amount of a process parameter.   8. The feature amount determining unit in the case where the product manufacturing process is processed in a plurality of steps, the step of determining the feature amount of the process parameter having the greatest contribution to yield and performance in the plurality of steps. Described process control method   In the step performed by the feature amount determination means, a method of calculating the influence rate on the product yield and performance from the product yield information as the quality information linked to each product and the feature amount of the process parameter is a multiple regression analysis. The process management method according to claim 7, wherein:   A control program in which a processing procedure for causing a computer to execute each step of the process management method according to claim 7 is described.   A computer-readable storage medium storing the control program according to claim 13.

Images