Semiconductor device manufacturing system and device diagnostic method

Abstract

To identify and correct a performance difference with a secular change of the same chamber, a performance difference with component exchange or component cleaning or a performance difference between different chambers.SOLUTION: A device diagnosis apparatus for a semiconductor processing device (plasma processing device) uses data of a semiconductor device manufacturing apparatus including a reference chamber to perform arithmetic processing, thereby mapping a feature amount to two or more dimensional axes with a plurality of parameters on the axis and creating reference chamber feature amount map data. The device diagnosis apparatus uses data of a semiconductor device manufacturing apparatus including a correction target chamber to perform arithmetic processing, thereby mapping a feature amount to two or more dimensional axes and creating correction target chamber feature amount data. Based on patterns which appear in the reference chamber feature amount map data and the correction target chamber feature amount map data, the device diagnosis apparatus identifies a performance difference between different chambers, a performance difference with a specular change of the same chamber and a performance difference with component exchange or component cleaning of the same chamber.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an equipment diagnostic device for diagnosing semiconductor processing equipment, a semiconductor manufacturing equipment system, a semiconductor device manufacturing system, and an equipment diagnostic method. In particular, the present invention relates to a state monitoring and prediction system for semiconductor processing equipment (semiconductor device manufacturing equipment) using an equipment diagnostic device, and is applicable to a technology for identifying, or a technology for identifying and correcting, a performance difference due to a change over time in the same processing room (referred to as a chamber), a performance difference due to part replacement or part cleaning in the same chamber, or a performance difference between different chambers, by performing arithmetic processing using data acquired from the semiconductor device manufacturing equipment. [Background technology]

[0002] In a semiconductor device manufacturing apparatus, performance fluctuations of the semiconductor device occur due to factors over time. In addition, replacement of consumable parts of the manufacturing apparatus and cleaning of parts also cause changes in the state of the chamber, resulting in performance fluctuations of the semiconductor device.

[0003] Solutions to the above problems have been proposed in JP-T 2018-533196, JP-A 2014-22695, JP-A 2009-295658, and the like. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Special Publication No. 2018-533196 [Patent Document 2] JP 2009-295658 A [Patent Document 3] JP 2014-22695 A Summary of the Invention [Problem to be solved by the invention]

[0005] JP 2018-533196 A proposes a correction system for chamber-to-chamber differences, but does not mention a specific correction algorithm. JP 2009-295658 A proposes a correction method using plasma emission intensity and plasma density, but does not mention an actual method for identifying chamber differences. JP 2014-22695 A proposes a correction method using plasma emission intensity and high-frequency power source peak-to-peak voltage Vpp, but when performing correction for multiple parameters, it is necessary to simultaneously identify chamber differences for multiple parameters, and there is no mention of this point. In addition, the type of sensor used for correction is based on the empirical rules of an expert, so it cannot be said to be an optimal solution. [Means for solving the problem]

[0006] A system for monitoring and predicting the status of semiconductor device manufacturing equipment, comprising: an equipment diagnostic device for identifying a performance difference between different chambers, a performance difference caused by a change over time in the same chamber, or a performance difference caused by a part replacement or part cleaning in the same chamber; a) in a first semiconductor device manufacturing apparatus or a second semiconductor device manufacturing apparatus serving as a reference chamber, a first step of collecting data on a plasma generation state using various sensors mounted on the first semiconductor device manufacturing apparatus having the reference chamber or the second semiconductor device manufacturing apparatus when generating plasma by changing parameters of a plurality of manufacturing conditions; A second step of applying a calculation process to the collected data to calculate a plurality of representative values ​​of plasma generation conditions under the parameters of the plurality of manufacturing conditions; A third step of performing a calculation process using the plurality of representative values ​​to calculate a feature amount; a fourth step of mapping the feature quantity onto a two- or more-dimensional graph with the parameters of the plurality of manufacturing conditions as axes to generate reference chamber feature quantity map data (D107); b) in the second semiconductor device manufacturing apparatus, the second semiconductor device manufacturing apparatus having a chamber which is to be subjected to correction of a performance difference between the different chambers, or the reference chamber which has changed over time and is to be subjected to correction of a performance difference due to the change over time of the same chamber, or the reference chamber which has changed over time and is to be subjected to correction of a performance difference due to the replacement or cleaning of parts of the same chamber, a fifth step of collecting data on the state of plasma generation using various sensors mounted in the second semiconductor device manufacturing apparatus having the chamber to be corrected or the reference chamber that has changed over time when plasma is generated while parameters of a plurality of manufacturing conditions are changed; A sixth step of applying a calculation process to the collected data to calculate a plurality of representative values ​​of plasma generation conditions under the parameters of the plurality of manufacturing conditions; A seventh step of performing a calculation process using the plurality of representative values ​​to calculate a feature amount; an eighth step of generating correction target chamber feature map data (D112) by mapping the feature amount to a two-dimensional or more-dimensional graph with the parameters of the plurality of manufacturing conditions as axes; c) a ninth step of identifying a performance difference between the different chambers, a performance difference caused by a change over time in the same chamber, or a performance difference caused by a part replacement or part cleaning in the same chamber, based on patterns appearing in the reference chamber feature map data and the correction target chamber feature map data; The state of the second semiconductor device manufacturing equipment is diagnosed by an algorithm including the above. Effect of the Invention

[0007] The above-mentioned means makes it possible to optimize information required for identifying and correcting performance differences due to aging of the same chamber, performance differences due to part replacement or part cleaning of the same chamber, or performance differences between different chambers, and to simultaneously identify and correct differences between chambers in parameters of a plurality of manufacturing conditions. This makes it possible to suppress performance variations of semiconductor devices manufactured by semiconductor device manufacturing equipment and to make the performance of the semiconductor devices uniform. [Brief description of the drawings]

[0008] [Figure 1] 11 is a flowchart showing a case in which the present invention is applied to a plasma processing apparatus diagnostic device that identifies and corrects differences between chambers by utilizing template matching in the field of image processing. [Diagram 2] 1 is a hardware configuration diagram of a semiconductor device manufacturing system including a plasma processing apparatus and a plasma processing apparatus diagnostic apparatus according to a first embodiment. [Diagram 3] FIG. 2 is a functional block diagram of a plasma processing apparatus diagnostic device according to the first embodiment. [Figure 4] FIG. 4 is a diagram showing an example of time-series data of each sensor value acquired from the plasma processing apparatus according to the first embodiment. [Diagram 5] FIG. 4 is a diagram showing an example of a feature map data format according to the present invention. [Figure 6] FIG. 4 is a diagram showing an example of data stored in each cell of feature amount map data according to the present invention. [Figure 7] FIG. 4 is a diagram showing an example of sub-map extraction from correction target chamber feature amount map data according to the first embodiment. [Figure 8] FIG. 11 is a diagram showing an example of a result of performing template matching using the reference chamber feature amount map data and a sub-map according to the first embodiment. [Figure 9] FIG. 2 is an overall configuration diagram of a semiconductor device manufacturing system according to a first embodiment, and is a network configuration diagram of a plasma processing apparatus, a plasma processing apparatus diagnostic apparatus, and a prediction result aggregating apparatus included in the semiconductor device manufacturing system. [Figure 10] 11 is a flowchart showing optimization of features used in template matching according to the second embodiment. [Figure 11] 13 is a graph comparing a correction amount calculated from a pseudo offset and a correction amount identified by template matching according to the second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, the embodiments will be described with reference to the drawings. However, in the following description, the same components are given the same reference numerals and the repeated description may be omitted. Note that the drawings may be shown more diagrammatically than the actual embodiment in order to clarify the description, but they are merely examples and do not limit the interpretation of the present invention. EXAMPLES

[0010] The equipment diagnostic device (302: see FIG. 3) of the first embodiment is related to a status monitoring and prediction system for a semiconductor processing device (semiconductor device manufacturing equipment 201), and is capable of optimizing information required to identify and correct performance differences between different processing chambers (hereinafter referred to as chambers), performance differences due to changes over time in the same chamber, or performance differences due to part replacement or part cleaning in the same chamber, and identifying and correcting differences between chambers in terms of multiple parameters.

[0011] Here, the performance difference between different chambers refers to a performance difference between a reference chamber installed in a semiconductor device manufacturing equipment (first semiconductor device manufacturing equipment) (200) having a chamber that is used as a reference (referred to as a reference chamber) and a chamber that is to be corrected (referred to as a chamber to be corrected or a chamber to be corrected) installed in a semiconductor device manufacturing equipment (second semiconductor device manufacturing equipment) (201) having a chamber to be corrected (referred to as a chamber to be corrected or a chamber to be corrected).

[0012] The performance difference is a difference (performance difference: for example, difference in etching amount, difference in film thickness) in the process results (e.g., etching, film formation) between the process using the reference chamber and the process using the chamber to be corrected. For example, there is a performance difference between the plasma process using the plasma generated in the reference chamber and the plasma process using the plasma generated in the chamber to be corrected. The performance difference between the plasma process using the plasma generated in the reference chamber and the plasma process using the plasma generated in the chamber to be corrected means that a performance difference occurs between a semiconductor device manufactured using a semiconductor device manufacturing apparatus having a reference chamber and a semiconductor device manufactured using a semiconductor device manufacturing apparatus having a chamber to be corrected. In other words, the performance difference between the plasma process using the plasma generated in the reference chamber and the plasma process using the plasma generated in the chamber to be corrected causes a performance fluctuation of the semiconductor device.

[0013] The performance difference due to changes over time in the same chamber refers to the performance difference between a chamber (corresponding to a reference chamber) in a state before changes over time (or in a state in which changes over time are minimal) and a reference chamber (corresponding to a chamber to be corrected) in a state in which changes over time have occurred in a chamber of the second semiconductor device manufacturing equipment (201).

[0014] The performance difference associated with part replacement or part cleaning in the same chamber refers to the performance difference between the chamber of the second semiconductor device manufacturing equipment (201) in a state before the part replacement or part cleaning (corresponding to the reference chamber) and the chamber in a state after the part replacement or part cleaning (corresponding to the chamber to be corrected).

[0015] The equipment diagnostic device (302) according to the embodiment identifies a performance difference between different chambers, a performance difference due to aging of the same chamber, or a performance difference due to part replacement or part cleaning of the same chamber, and corrects the performance of the chamber to be corrected to be equivalent to the performance of a reference chamber by the identified correction amount. The equipment diagnostic device (302) transmits the identified correction amount to the second semiconductor device manufacturing equipment (201).

[0016] As a result, the second semiconductor device manufacturing equipment (201) can generate, for example, plasma corrected by the identified correction amount in the chamber (chamber to be corrected). This makes it possible to suppress fluctuations in the performance of the semiconductor device manufactured by the second semiconductor device manufacturing equipment (201) and to make the performance of the semiconductor device uniform.

[0017] Fig. 1 is a flow chart showing the case where the present invention is applied to a plasma processing apparatus diagnostic device that identifies and corrects differences between chambers using template matching in the image processing field. Fig. 2 is a diagram showing the hardware configuration of a semiconductor device manufacturing apparatus system including a plasma processing apparatus and a plasma processing apparatus diagnostic device according to the first embodiment.

[0018] The semiconductor device manufacturing system (semiconductor manufacturing system) 20 shown in FIG. 2 includes a plasma processing device (PPE) 201 and a computer (CAL) 204. The plasma processing device (PPE) 201 indicates a semiconductor manufacturing device (also called a semiconductor device manufacturing device) used for manufacturing a semiconductor device. The plasma processing device 201 is, for example, a plasma etching device, and is equipped with a chamber (also called a processing chamber, reaction vessel, reaction chamber, or reactor) used for performing plasma processing on a semiconductor substrate. The computer (CAL) 204 has an external IF (EXIF: also called an external interface circuit) 205, a processor (PROC) 206, a memory (MEM) 207, and a storage (STG) 208. In this example, the external IF 205 of the computer 204 is electrically connected to the plasma processing device 201 by metal wiring ML for transmitting and receiving signals of a network wiring such as a LAN cable or a USB cable. The computer 204 receives signal data during plasma processing from various sensors mounted on the plasma processing device 201 via data transfer (DT) 202 from an external IF 205, and transmits the results of calculations by the computer 204 from the external IF 205 to the plasma processing device 201 via correction signal transmission (CST) 203.

[0019] 3 is a diagram showing functional blocks of the plasma processing apparatus diagnostic apparatus according to the first embodiment. A plasma processing apparatus diagnostic apparatus (EDE: equipment diagnostic apparatus, also referred to as diagnostic apparatus) 302 is implemented on a computer 204. The diagnostic apparatus 302 is configured to be realized by the computer 204 when the processor 206 of the computer 204 executes a software program for realizing the function of the diagnostic apparatus 302 stored in the storage 208, for example. The diagnostic apparatus 302 may be configured with hardware circuits for each of the functions (303, 305, 307, 311) described below. The diagnostic apparatus 302 has a data acquisition unit (DAD) 303, a data correction unit (DCU) 305, a device state prediction calculation unit (PCU) 307, and a judgment and control unit (JCU, hereinafter referred to as judgment and control unit) 311.

[0020] The data acquisition unit (DAD) 303 has a function of receiving the data transmitted by the data transfer 202 and storing it as device data (DED) 304 in the storage 208 on the computer 204 .

[0021] The data correction unit (DCU) 305 has a function of converting the equipment data 304 into corrected chamber characteristic map data (CTCMD) 306 .

[0022] The equipment state prediction calculation unit (PCU) 307 has a function of using the correction target chamber feature map data 306 and the reference chamber feature map data (RCMD) 308 to identify a correction amount, based on a pattern appearing in the reference chamber feature map data and the correction target chamber feature map data, by shifting the recipe parameters, which are set values ​​such as microwave intensity, coil current, pressure, and high frequency bias power during plasma processing, from the set values ​​of the correction target chamber to obtain the same results in the plasma processing in the correction target chamber and the reference chamber. The recipe parameter correction amount identification unit (CIU) 309 is a unit that executes calculations to determine how much the recipe parameters differ between the correction target chamber feature map data 306 and the reference chamber feature map data 308, and has a function of storing the calculation result as equipment state prediction calculation result data (CRD) 310 in the storage 208.

[0023] The judgment / control unit (JCU) 311 has a function of creating and transmitting a message to be sent to the plasma processing apparatus 201 from the apparatus state prediction calculation result data 310. The correction amount transmission unit (CAT) 312 has a function of transmitting the created message to the plasma processing apparatus 201 via the external IF 205.

[0024] The flow chart of Fig. 1 shows an algorithm executed by the diagnostic device 302. The operation of the diagnostic device 302 based on the flow chart of Fig. 1 is as follows.

[0025] First, the diagnostic device 302 executes steps S103, S105, and S106 in the reference chamber C101 to generate the reference chamber feature map data D107. The reference chamber feature map data D107 may be generated by the diagnostic device 302 as described above, or may be generated by collecting the reference chamber equipment data D104 by another method, storing it on the storage of a computer other than the computer on which the diagnostic device 302 is implemented, and executing steps S103, S105, and S106 on the computer.

[0026] When it is time to perform chamber correction, the diagnostic device 302 executes steps S108, S110, and S111 in the chamber C102 to be corrected, thereby generating correction target chamber feature amount map data D112.

[0027] The diagnostic equipment 302 executes steps S113, S114, and S115 using the reference chamber feature map data D107 and the correction target chamber feature map data D112 to identify the correction amounts of the recipe parameters (step S116, data D117), and finally transmits a signal to the plasma processing equipment 201 (step S118).

[0028] As a result, diagnostic device 302 transmits the identified correction amount as a signal to plasma processing device 201 via external IF 205. Having received the identified correction amount as a signal, plasma processing device 201 can correct various parameters of the manufacturing conditions for generating plasma in the chamber based on the identified correction amount. This makes it possible to generate plasma in the chamber of plasma processing device 201 in which performance differences between different chambers, performance differences due to changes over time in the same chamber, or performance differences due to part replacement or part cleaning in the same chamber have been corrected. This provides a unique effect of suppressing performance fluctuations in semiconductor devices manufactured by plasma processing device 201 and enabling the performance of semiconductor devices to be made uniform.

[0029] The chamber to be corrected may be a chamber different from the reference chamber, or may be the same chamber as the reference chamber but the state of which has changed over time or due to part replacement or part cleaning.

[0030] Details of each step are as follows:

[0031] S103 (first step): In the reference chamber C101, plasma is generated and plasma processing is performed while changing multiple recipe parameter setting values ​​at specific intervals within a specific value range, and data on the plasma generation status at that time is collected using various sensors mounted in the plasma processing apparatus (semiconductor device manufacturing apparatus 200, 201) and saved as reference chamber apparatus data D104.

[0032] S105 (second step): Using the reference chamber apparatus data D104, calculate a representative value of the sensor value collected during plasma processing at each set value of the multiple recipe parameters changed in step S103. In other words, apply arithmetic processing to the collected data (reference chamber apparatus data D104) to calculate multiple representative values ​​of the plasma generation status under the parameters of the multiple manufacturing conditions.

[0033] S106 (third and fourth steps): Reference chamber feature map data D107 is created from the representative values ​​calculated in step S105, and saved. That is, arithmetic processing is performed using the representative values, and feature values ​​are calculated (third step). Then, the calculated feature values ​​are mapped onto a two- or more-dimensional graph with the parameters of the multiple manufacturing conditions as axes, to create reference chamber feature map data (fourth step).

[0034] S108 (fifth step): In the chamber C102 to be corrected, plasma is generated and plasma processing is performed while changing multiple recipe parameter setting values ​​in a specific value range at specific intervals in the same manner as in S103, and data on the plasma generation status at that time is collected using various sensors mounted in the plasma processing apparatus (semiconductor device manufacturing apparatus) 201 and saved as chamber apparatus data D109 to be corrected.

[0035] S110 (sixth step): Using the chamber apparatus data D109 to be corrected, a representative value of the sensor value collected during plasma processing at each setting value of the multiple recipe parameters changed in S108 is calculated in the same manner as in step S105.

[0036] S111 (seventh and eighth steps): Using the representative values ​​calculated in step S110, correction target chamber feature map data D112 is created and stored in the same manner as in step S106. That is, arithmetic processing is performed using a plurality of representative values ​​to calculate feature values ​​(seventh step). Then, the calculated feature values ​​are mapped onto a two- or more-dimensional graph with the parameters of a plurality of manufacturing conditions as axes to create correction target chamber feature map data D112 (eighth step).

[0037] S113: When the number of data points of the reference chamber feature map data D107 and the correction target chamber feature map data D112 is small, interpolation or extrapolation is performed.

[0038] S114: A partial area of ​​the correction target chamber feature amount map data D112 is extracted as a sub-map.

[0039] S115: Using the submap extracted in step S114 and the reference chamber feature map data, template matching is performed to check to which position on the reference chamber feature map data the submap matches.

[0040] Steps S114 and S115 are looped to perform template matching using multiple sub-maps and obtain multiple results.

[0041] S116: A final correction amount is determined (identified) from the acquired multiple template matching results, and saved as result data (also called correction amount data) D117. That is, in steps S113-S116 (ninth step), performance differences between different chambers, performance differences due to changes over time in the same chamber, or performance differences due to part replacement or part cleaning in the same chamber are identified based on patterns appearing in the reference chamber feature map data and the correction target chamber feature map data. The identified final correction amount is saved as result data D117.

[0042] S118 (tenth step): The contents to be transmitted to the plasma processing apparatus 201 are determined and transmitted.

[0043] FIG. 4 shows an example of time series data of each sensor value acquired from the plasma processing apparatus according to the first embodiment. FIG. 4 shows an example of time series data of each sensor value of the apparatus data 304 acquired in steps S103 and S108. In FIG. 4, the vertical axis indicates the sensor value (VS) and the horizontal axis indicates time (t). Here, a period T401 indicates a transient state region immediately after the recipe parameters are set, a period T402 indicates a steady state region, and periods T403 and T404 indicate regions immediately before the end of the plasma processing. In steps S105 and S110, a representative value is calculated from the time series data as shown in FIG. 4 for each sensor. Here, the representative value may be statistical data such as an average value or standard deviation in a specific time region of the time series data, or a sensor value at a certain time.

[0044] Fig. 5 is a diagram showing an example of a feature amount map data format according to the present invention, and Fig. 6 is a diagram showing an example of data stored in each cell of the feature amount map data according to the present invention.

[0045] FIG. 5 shows three-dimensional feature map data 3DMP as an example of feature map data created in steps S106 and S111. In the feature map data 3DMP, different recipe parameter A (PA), B (PB), and C (PC) setting values ​​are associated with the height direction, width direction, and depth direction, respectively. That is, the recipe parameters A (PA), B (PB), and C (PC) are set as the axes of parameters of a plurality of manufacturing conditions. The height direction of the feature map data 3DMP shown in FIG. 5 indicates the settable range PASR of the recipe parameter A (PA) from the minimum value (MIN_PA) to the maximum value (MAX_PA). The width direction of the feature map data 3DMP indicates the settable range PBSR of the recipe parameter B (PB) from the minimum value (MIN_PB) to the maximum value (MAX_PB). The depth direction of the feature map data 3DMP indicates the settable range PCSR of the recipe parameter C (PC) from the minimum value (MIN_PC) to the maximum value (MAX_PC). Recipe parameters A, B, and C are multiple recipe parameters changed in steps S103 and S108, and for example, recipe parameter A can be microwave intensity, recipe parameter B can be coil current, and recipe parameter C can be pressure. Recipe parameters A, B, and C are not limited to these, and other parameters can also be used.

[0046] Here, one small cube C501 represents plasma processing conditions that are certain set values ​​of recipe parameters A, B, and C. Fig. 6 shows array data A601 held inside one small cube C501 in Fig. 5. In this example, the array data (feature array) A601 is an array of 10 features (a1-a10).

[0047] In steps S106 and S111, the feature amount is calculated using the representative value calculated for each recipe parameter setting condition, and is stored as a feature amount array inside the small cube at the position corresponding to the recipe parameter setting condition on the feature amount map data 3DMP in FIG. 5. Here, the method of calculating the feature amount from the representative value may be to calculate the feature amount by performing arithmetic operations using multiple representative values, or the representative value may be used as the feature amount as it is. In addition, for each element of the feature amount array, a value that has been converted by normalizing using the values ​​in all the small cubes of the feature amount map data 3DMP may be stored in the feature amount array as the final feature amount. For example, the coil voltage, magnetron current, valve opening, etc. are stored as feature amounts in the elements of the feature amount array (a1-a10) in FIG. 6. It is easy to imagine that this process is similar to the process of associating R, G, and B luminance information with each pixel position in an image. This process makes it possible to apply an image processing solution to the chamber matching problem. Furthermore, even if the feature quantity has a tendency to change depending on the recipe parameters A, B, and C, it becomes possible to simultaneously estimate the values ​​of the recipe parameters A, B, and C from the obtained feature quantity values. Here, although three-dimensional feature quantity map data 3DMP is exemplified in this embodiment, the same effect can be obtained if the data is two or more dimensions.

[0048] FIG. 7 is a diagram showing an example of sub-map extraction from the correction target chamber feature map data according to the first embodiment. FIG. 7 shows a two-dimensional map M701 when the feature map data 3DMP of FIG. 5 is viewed from the front, and a two-dimensional map M702 when viewed from above. In step S114, an area B703 surrounded by a thick line in FIG. 7 is extracted as a sub-map from the correction target chamber feature map data D112 having the data format of FIG. 5. As a result, the sub-map B703 becomes 3x3x3 three-dimensional feature map data in which a small cube C704 at the center is the recipe parameter setting value (A0, B0, C0). Here, if the correction target chamber feature map data D112 is subjected to interpolation or extrapolation in S113, a sub-map is extracted from the correction target chamber feature map data after the interpolation or extrapolation process. Furthermore, the size of the submap B703 to be extracted is not limited to 3x3x3, but may be, for example, a three-dimensional map such as 3x4x5, a two-dimensional map such as 3x4x1, or a one-dimensional map such as 1x1x3.

[0049] FIG. 8 is a diagram showing an example of a result of performing template matching using the reference chamber feature map data and the submap according to the first embodiment. FIG. 8 shows a result of performing template matching using the reference chamber feature map data and the submap B703 extracted in step S114. In FIG. 8, the vertical axis indicates a matching index (MI), and the horizontal axis indicates a submap scanning position (SSP) on the reference chamber feature map data. The template matching performed in step S115 scans the submap B703 extracted in step S114 on the reference chamber feature map data D107, calculates the matching index (MI) at each scanning position (SSP), and obtains a result as shown in FIG. 8. Here, when the reference chamber feature map data D107 is subjected to interpolation or extrapolation in S113, the submap extracted from the correction target chamber feature map data after the interpolation or extrapolation process is scanned on the reference chamber feature map data after the interpolation or extrapolation process. In addition, the scanning position P801 where the matching index MI is minimum in FIG. 8 is the point (matching point) where the feature amount array of the submap B703 and the feature amount array at the scanning position P801 of the reference chamber feature amount map data D107 are closest in value.

[0050] 5, the setting values ​​of the recipe parameters A, B, C of the small cube C704 at the center of the submap B703 at the scan position P802 are (A1, B1, C1), and therefore the setting values ​​of the recipe parameters A, B, C of the small cube C704 at the center of the submap B703 at the scan position P802 are (A0, B0, C0 ...A , V B , V C ) is calculated using the following formula 1.

[0051] (V A , V B , V C )=(A0-A1, B0-B1, C0-C1) (Formula 1) Here, the matching index may be an index such as the sum of absolute difference (SAD) or the sum of squared difference (SSD), or may be an index such as normalized cross correlation (NCC) or zero-mean normalized cross correlation (ZNCC).In addition, the method for determining the matching point is the scanning position where the matching index is the minimum value, but a method may be implemented in which, for example, a regression model is used in an area near the minimum value of the matching index to perform fitting and precisely determine the matching point.

[0052] After steps S114 and S115 are completed for one submap (B703), the process returns to step S114 to extract another submap from an area of ​​the chamber C102 to be corrected that is different from the submap B703, and template matching is performed again in step S115 to identify the amount of correction. In this manner, template matching is performed using various (multiple) submaps to identify the amount of correction.

[0053] In step S116, a final correction amount is calculated by averaging the multiple correction amounts obtained by repeatedly performing steps S114 and S115 as described above, and multiple template matching results and correction amount identification results obtained by repeatedly performing steps S114 and S115 are saved as result data D117. Here, only highly reliable matching points may be collected by a process of selecting only results whose matching index values ​​at matching points are equal to or less than a specific threshold value from the multiple template matching results obtained by repeatedly performing steps S114 and S115, and the final correction amount may be calculated by averaging these. Also, although the final correction amount is calculated by averaging here, the final correction amount may be calculated by, for example, performing a median or mode calculation.

[0054] In step S118, the contents of the message to be transmitted to the plasma processing apparatus 201 are determined from the result data D117, and the contents of the message are transmitted from the external IF 205 to the plasma processing apparatus 201. Here, the correction amount calculated in step S116 may be transmitted as a message, or, if any of the absolute values ​​of the correction amounts of the recipe parameters is equal to or greater than a certain threshold, a message may be transmitted with contents encouraging replacement of a part related to that recipe parameter. Also, if highly reliable matching point information cannot be obtained in step S116, a message may be transmitted with contents indicating failure of template matching.

[0055] Fig. 9 is an overall configuration diagram of a semiconductor device manufacturing system according to Example 1, and shows a network configuration diagram of plasma processing apparatuses, plasma processing apparatus diagnostic apparatuses, and prediction result aggregating apparatuses included in the semiconductor device manufacturing system. Fig. 9 shows a network configuration diagram of plasma processing apparatus diagnostic apparatuses (EDEs) PC901, PC902, and PC903 connected to the respective plasma processing apparatuses (PPEs) 201a, 201b, and 201c, and a prediction result aggregating apparatus (FRA) PC905 that aggregates the prediction results of the plurality of plasma processing apparatus diagnostic apparatuses PC901, PC902, and PC903.

[0056] As shown in FIG. 9, the semiconductor device manufacturing system 900 includes semiconductor device manufacturing systems (semiconductor manufacturing systems) 20a, 20b, and 20c, a prediction result aggregation device (FRA) PC905, and a network NW904 to which the semiconductor device manufacturing systems 20a, 20b, and 20c and the prediction result aggregation device PC905 are connected. As described in FIG. 2 and FIG. 3, the semiconductor device manufacturing system 20a includes a plasma processing device (PPE) 201a and a plasma processing device diagnostic device (EDE) PC901. Similarly, the semiconductor device manufacturing system 20b includes a plasma processing device (PPE) 201b and a plasma processing device diagnostic device (EDE) PC902. The semiconductor device manufacturing system 20c includes a plasma processing device (PPE) 201c and a plasma processing device diagnostic device (EDE) PC903.

[0057] Here, the plasma processing apparatuses (201a, 201b, 201c) can correspond to the second semiconductor device manufacturing apparatus, while the plasma processing apparatus (PPE) 200 described later can correspond to the first semiconductor device manufacturing apparatus.

[0058] Each of the plasma processing apparatuses 201a, 201b, and 201c is connected to a plasma processing apparatus diagnostic apparatus PC901, PC902, and PC903, one each. Each of the plasma processing apparatus diagnostic apparatus PC901, PC902, and PC903 is the plasma processing apparatus diagnostic apparatus (EDE) 302 shown in FIG. 3. The plasma processing apparatus diagnostic apparatus PC901, PC902, and PC903 are connected to a network NW904 using communication wiring such as a LAN cable. Furthermore, a prediction result aggregating apparatus PC905 is connected to the network NW904.

[0059] The prediction result aggregating device PC905 aggregates the state prediction results of the corresponding plasma processing devices (201a, 201b, 201c) calculated by the plasma processing device diagnostic devices PC901, PC902, PC903.

[0060] 9, a semiconductor manufacturing equipment system 900 configuring a network as shown in FIG 9 allows an engineer to monitor and compare the states of each plasma processing equipment (201a, 201b, 201c) simply by accessing a prediction result aggregating device PC905. Also, for example, it becomes possible to distribute reference chamber feature map data D107 from the prediction result aggregating device PC905 to each plasma processing equipment diagnostic device PC901, PC902, PC903.

[0061] That is, the reference chamber feature map data D107 is created by the plasma processing apparatus diagnostic device PC910 connected to the plasma processing apparatus (PPE) 200 including the reference chamber C101. The created reference chamber feature map data D107 is configured to be transmitted from the plasma processing apparatus diagnostic device PC910 to the prediction result aggregating device PC905 and stored in the prediction result aggregating device PC905. The plasma processing apparatus diagnostic device PC910 can also be considered to execute steps S103, S105, and S106. Here, it is also possible to collect the reference chamber apparatus data D104 by another method without using the plasma processing apparatus diagnostic device PC910, store the reference chamber apparatus data D104 in a storage on a computer different from the plasma processing apparatus diagnostic device PC910, and execute steps S103, S105, and S106 on the computer different from the plasma processing apparatus diagnostic device PC910 to create the reference chamber feature map data D107. When creating the reference chamber feature map data D107 without using the plasma processing apparatus diagnostic apparatus PC910 in this manner, the reference chamber feature map data D107 does not need to be stored in the prediction result aggregating device PC905 by transmitting it from the plasma processing apparatus diagnostic apparatus PC910 to the prediction result aggregating device PC905, but rather the reference chamber feature map data D107 may be stored in, for example, portable storage, connected to the prediction result aggregating device PC905, and copied from the portable storage to the prediction result aggregating device PC905, thereby storing the reference chamber feature map data D107 in the prediction result aggregating device PC905.

[0062] The semiconductor device manufacturing equipment system 915 including the plasma processing equipment (PPE) 200 and the plasma processing equipment diagnostic equipment PC910 may be configured such that the plasma processing equipment diagnostic equipment PC910 is connected to the network NW904 as shown by the dashed line LL91 in Fig. 9. Alternatively, the plasma processing equipment diagnostic equipment PC910 may not be connected to the network NW904, but may be connected to the prediction result aggregating equipment PC905 by another network as shown by the dashed line LL92 in Fig. 9. When the reference chamber feature map data D107 is stored in the prediction result aggregating equipment PC905, the semiconductor device manufacturing equipment system 915 may not be connected to the prediction result aggregating equipment PC905. When an update operation of the reference chamber feature map data D107 occurs, it is preferable to electrically connect the semiconductor device manufacturing equipment system 915 and the prediction result aggregating equipment PC905 by a network.

[0063] 9 can be regarded as a semiconductor device manufacturing system having a platform that is connected to the semiconductor manufacturing equipment (201a, 201b, 201c) via a network NW904 and has an application for identifying the performance difference between the processing chambers of the semiconductor manufacturing equipment (201a, 201b, 201c). The application executes the steps of obtaining features for data collected from the semiconductor manufacturing equipment (201a, 201b, 201c), mapping the features to a two-dimensional or higher graph with the parameters of the manufacturing conditions of the semiconductor manufacturing equipment (201a, 201b, 201c) as axes, and creating feature map data, and identifying the performance difference based on the created feature map data. The application is stored in the memory (MEM) 207 or storage (STG) 208 in FIG. 2 and executed by the processor (PROC) 206. EXAMPLES

[0064] The second embodiment relates to a feature optimization method for selecting feature values ​​with high identification accuracy in identifying a performance difference between chambers for feature values ​​stored in the reference chamber feature value map data D107 and the correction target chamber feature value map data D112 in the first embodiment.

[0065] FIG. 10 is a flowchart of the second embodiment. The flowchart in FIG. 10 is a feature optimization algorithm for correcting the difference between chambers, which is executed by the diagnostic device 302. In FIG. 10, components with the same reference numerals as in FIG. 1 represent the same steps or data of the same format as in FIG. 1. That is, S103, D104, S105, S106, S108, D109, S110, S111, and S113-S116 in FIG. 10 are the same as in FIG. 1, and D121, D122, D123, and S124 in FIG. 10 are data and steps that have been changed from FIG. 1. Only steps that have been changed from FIG. 1 are shown below.

[0066] The reference chamber C119 and the correction target chamber C120 are the same chamber and are in a state where there is little change over time. The correction target chamber C120 is in a state where a pseudo offset occurs in the recipe parameters for which the correction amount is identified by editing the internal control parameters of the plasma processing apparatus 201. On the other hand, the reference chamber C119 is in a state where no pseudo offset occurs in the recipe parameters for which the correction amount is identified, without editing the internal control parameters of the plasma processing apparatus 201.

[0067] The reference chamber feature map data D121 and the correction target chamber feature map data D122 are data that hold all candidates for features used in template matching as elements of a feature array.

[0068] Step S124: Using the correction amount data D123 as the result data when template matching is performed for all the candidate feature amounts, the correction amount calculated from the pseudo offset is compared with the correction amount identified by template matching to perform feature amount optimization. That is, the performance difference calculated from the pseudo offset value is compared with the identified performance difference, and the feature amount that brings them closer is selected as the optimized feature amount. By using such a feature amount optimization method, the optimized feature amount can be selected and used. By using the selected optimal feature amount in the algorithm shown in FIG. 1, it is possible to identify highly accurate correction amount data. The highly accurate correction amount data is made the content of the message to be transmitted to the plasma processing apparatus 201 by step S118 of the algorithm in FIG. 1. The content of the message is transmitted from the external IF 205 (see FIG. 3) of the diagnostic apparatus 302 to the plasma processing apparatus 201.

[0069] That is, the equipment diagnostic device (302) according to the second embodiment optimizes the feature amount in order to identify with high accuracy the performance difference between different chambers, the performance difference caused by the change over time of the same chamber, or the performance difference caused by the replacement or cleaning of parts of the same chamber. By using the feature amount selected by the optimization of the second embodiment in the equipment diagnostic device (302) according to the first embodiment, it becomes possible to identify with high accuracy the performance difference between different chambers, the performance difference caused by the change over time of the same chamber, or the performance difference caused by the replacement or cleaning of parts of the same chamber, and correct the performance of the chamber to be corrected to be equivalent to the performance of the reference chamber by the identified highly accurate correction amount. The equipment diagnostic device (302) transmits the identified highly accurate correction amount as an electronic message to the second semiconductor device manufacturing equipment (201).

[0070] This makes it possible to generate plasma in the chamber of plasma processing apparatus 201, which compensates for performance differences between different chambers, performance differences due to aging in the same chamber, or performance differences due to part replacement or part cleaning in the same chamber. This has the unique effect of suppressing fluctuations in the performance of semiconductor devices manufactured by plasma processing apparatus 201, thereby enabling the performance of semiconductor devices to be made uniform.

[0071] FIG. 11 is a graph in which the correction amount data D123 is plotted with the correction amount (CCA) calculated from the offset given pseudo to the recipe parameter on the horizontal axis and the correction amount (ICA) identified by template matching on the vertical axis. The result R1001 displayed with a white circle is the correction amount data identified when an ineffective feature is used, and the result R1002 displayed with a black circle is the correction amount data identified when an effective feature is used. In addition, in the graph of FIG. 11, the slope (TI) of the dotted line L11 is 1 (TI=1). In step S124 of FIG. 10, optimization is performed by using such results to select a feature that reduces the difference between the correction amount calculated from the pseudo offset and the correction amount identified by template matching as an effective feature. As an optimization method, a feature with a coefficient of determination close to 1 may be selected using the identified correction amount as a predicted value, or a machine learning method may be used to search for an optimal feature.

[0072] The features of the device diagnostic device according to the present invention can be summarized as follows.

[0073] 1) In an equipment diagnostic device that identifies performance differences between processing chambers of semiconductor manufacturing equipment, A feature amount for the data collected from the semiconductor manufacturing equipment is obtained; Feature quantity map data is created by mapping the feature quantities onto a two- or more-dimensional graph with the parameters of the manufacturing conditions of the semiconductor manufacturing device as axes; The performance difference is identified based on the created feature amount map data.

[0074] 2) In the device diagnostic device described in 1) above, The performance differences are identified by applying template matching to the feature map data.

[0075] 3) In the device diagnostic device described in 1) above, The performance difference includes a performance difference between different processing chambers, a performance difference caused by aging of the same processing chamber, and a performance difference caused by replacement or cleaning of parts of the same processing chamber.

[0076] 4) In the device diagnostic device described in 1) above, Corrections to the parameters that correspond to the identified performance differences are determined.

[0077] 5) In the device diagnostic device described in 1) above, The semiconductor manufacturing equipment is a plasma etching equipment.

[0078] 6) A semiconductor manufacturing equipment system includes the equipment diagnostic device described above in 1) to which semiconductor manufacturing equipment is connected via a network.

[0079] 7) A semiconductor device manufacturing system including a platform connected to a semiconductor manufacturing device via a network and having an application implemented thereon for identifying a performance difference between processing chambers of the semiconductor manufacturing device, determining a feature amount for data collected from the semiconductor manufacturing equipment; creating feature quantity map data by mapping the feature quantities on a two- or more-dimensional graph with the parameters of the manufacturing conditions of the semiconductor manufacturing apparatus as axes; and identifying the performance difference based on the created feature map data, the step being executed by the application.

[0080] 8) A method for diagnosing a semiconductor manufacturing device for identifying a performance difference between processing chambers, comprising: determining a feature amount for data collected from the semiconductor manufacturing equipment; creating feature quantity map data by mapping the feature quantities on a two- or more-dimensional graph with the parameters of the manufacturing conditions of the semiconductor manufacturing apparatus as axes; and identifying the performance difference based on the created feature amount map data.

[0081] 9) In 8) above, The performance difference includes a performance difference between different processing chambers, a performance difference caused by aging of the same processing chamber, and a performance difference caused by replacement or cleaning of parts of the same processing chamber. [Explanation of symbols]

[0082] 20, 20a, 20b, 20c: semiconductor device manufacturing equipment system (semiconductor manufacturing equipment system) 200: Plasma processing device (first semiconductor device manufacturing device) 201, 201a, 201b, 201c: Plasma processing apparatus (second semiconductor device manufacturing apparatus) 302, PC901, PC902, PC903: Plasma processing equipment diagnostic equipment (equipment diagnostic equipment) D107, D121: Reference chamber feature map data D112, D122: Chamber feature map data to be corrected PC905: Prediction result aggregation device C101, C119: Reference chamber C102, C120: Chambers to be corrected 900: Semiconductor device manufacturing system

Claims

1. 1. A semiconductor device manufacturing system comprising: a platform connected to a plasma etching apparatus via a network, and having an application implemented thereon for identifying a performance difference between processing chambers of the plasma etching apparatus, determining a feature value for data collected from the plasma etching apparatus; creating feature quantity map data by mapping the feature quantities on a two- or more-dimensional graph having parameters of manufacturing conditions of the plasma etching apparatus as axes; and identifying the performance difference based on the created feature map data, the step being executed by the application; The graph is divided into cells corresponding to the manufacturing conditions for each set value of the parameter, The semiconductor device manufacturing system according to claim 1, wherein the mapped feature amount is stored in the cell corresponding to the manufacturing condition related to the feature amount.

2. A method for diagnosing a performance difference between processing chambers of a plasma etching apparatus, comprising: determining a feature value for data collected from the plasma etching apparatus; creating feature quantity map data by mapping the feature quantities on a two- or more-dimensional graph having parameters of manufacturing conditions of the plasma etching apparatus as axes; and identifying the performance difference based on the created feature amount map data, The graph is divided into cells corresponding to the manufacturing conditions for each set value of the parameter, The method for diagnosing an apparatus, comprising the steps of: storing the mapped feature quantity in the cell corresponding to the manufacturing condition related to the feature quantity.

3. 3. The apparatus diagnostic method according to claim 2, The performance difference includes a performance difference between different processing chambers, a performance difference caused by aging of the same processing chamber, and a performance difference caused by replacement or cleaning of parts of the same processing chamber.

Images