WO2020157967A1

WO2020157967A1 - Substrate processing device and method for manufacturing semiconductor device

Info

Publication number: WO2020157967A1
Application number: PCT/JP2019/003676
Authority: WO
Inventors: 護大石
Original assignee: 株式会社ＫｏｋｕｓａｉＥｌｅｃｔｒｉｃ
Priority date: 2019-02-01
Filing date: 2019-02-01
Publication date: 2020-08-06
Also published as: CN215933545U; JPWO2020157967A1; JP7055226B2

Abstract

The present invention makes it possible to: detect, with high accuracy, contact between a substrate transfer machine or a substrate on a substrate transfer machine and a boat or a substrate on a boat; and detect very small abnormalities that are present before a state of malfunction is reached.　The present invention comprises: a substrate transfer machine equipped with a substrate holder for holding a substrate, an end effector for directly operating the substrate, and a drive mechanism for causing the end effector to move, the substrate transfer machine transferring the substrate to the substrate holder or transferring the substrate from the substrate holder; a vibration sensor for detecting vibration in the end effector; a control unit for controlling the substrate transfer machine; and an analysis device for using a normal space created by analyzing, through the MT method, time region data for the vibration in the end effector as detected by the vibration sensor when a substrate transfer operation by the substrate transfer machine is carried out normally, calculating the Mahalanobis distance, in the normal space, of time region data detected by the vibration sensor while the substrate transfer machine places a substrate and performs a prescribed transfer operation, and assessing that contact has been made between the end effector or the substrate on the end effector and the substrate holder or the substrate on the substrate holder when the calculated Mahalanobis distance exceeds a prescribed threshold value. The control unit outputs, to the analysis device, signals indicating the start and the end of the prescribed transfer operation.

Description

Substrate processing apparatus and semiconductor device manufacturing method

The present disclosure relates to a method for manufacturing a substrate processing apparatus and a semiconductor device.

A substrate processing apparatus transfers a wafer from a storage container (hereinafter also referred to as a pod) in which a substrate (hereinafter also referred to as a wafer) is stored by a substrate transfer machine to a boat that is a substrate holder, and the boat holding the wafer reacts. It is configured to be loaded into a furnace and to perform a predetermined process on the wafer in the reaction furnace.

The substrate transfer machine transfers a wafer between a pod and a boat, and has a function of advancing/retracting, moving up/down, and rotating with a wafer placed on a tweezer of the substrate transfer machine. .. Teaching work is performed so that the wafer can be placed at a proper position when the wafer is placed from the pod to the tweezers or when the wafer placed on the tweezers is loaded on the boat by the substrate transfer machine. ..

However, the wafer may be loaded at a position different from the teaching work due to the deformation of the wafer, the vibration of the pump provided in the substrate processing apparatus, the secular change of the tweezers, etc., and the wafer may be displaced. When the displacement of the wafer becomes large, the tweezers or the wafers on the tweezers may come into contact with the boat, the wafers on the boat, or the like, and the wafers may be damaged or the boat may tip over.

Conventionally, in order to prevent this fall accident, the vibration at the time of collision is detected by a collision sensor provided in the substrate transfer machine (for example, refer to Patent Document 1). Further, the vibration generated in the lifting pins on the support table that supports the wafer is detected, and the mounting state is determined based on the total value of the periods when the detected vibration intensity is larger than the threshold value. (See, for example, Patent Document 2).

Japanese Patent Laid-Open No. 2003-109999 Japanese Patent No. 6244317

However, according to the conventional technology, there were many erroneous determinations, which caused interruption of the transfer operation. Moreover, since the change in the analog amount is detected by the threshold value, it is difficult to detect a slight abnormality such as contact.

An object of the present disclosure is to detect with high accuracy that a substrate transfer device or a substrate on the substrate transfer device and a substrate holder or a substrate on the substrate holder are in contact with each other, and to detect a slight error before a failure state. It is to provide a configuration capable of detecting an abnormality.

According to one aspect of the present disclosure,
A substrate holder for holding the substrate,
A substrate transfer machine that includes an end effector that directly operates the substrate and a drive mechanism that moves the end effector, transfers the substrate to the substrate holder, or transfers the substrate from the substrate holder. ,
A vibration sensor for detecting the vibration of the end effector,
A control unit for controlling the substrate transfer machine,
A normal space created by analyzing the time domain data of the vibration of the end effector detected by the vibration sensor when the substrate transfer operation by the substrate transfer machine is normally performed by the MT method is used. , The Mahalanobis distance in the normal space of the time domain data detected by the vibration sensor while the substrate transfer machine places the substrate and performs a predetermined transfer operation, and the calculated Mahalanobis distance A substrate on the end effector or the end effector, and an analysis device for determining that the substrate holder or the substrate on the substrate holder are in contact with each other when the threshold value exceeds a predetermined threshold value,
A technique is provided in which the control unit outputs a signal indicating the start and end of the predetermined transfer operation to the analysis device.

According to the present disclosure, it is possible to detect with high accuracy that the substrate transfer device or the substrate on the substrate transfer device and the substrate holder or the substrate on the substrate holder are in contact with each other, and to detect a slight error before the failure state. Anomalies can be detected.

FIG. 3 is a perspective view of a storage chamber preferably used in the first embodiment of the present disclosure. It is a cross-sectional view of a substrate processing apparatus preferably used in the first embodiment of the present disclosure. It is a perspective view of a substrate transfer machine suitably used in one embodiment of the present disclosure. FIG. 3 is a schematic configuration diagram of a controller of a substrate processing apparatus that is preferably used in an embodiment of the present disclosure, and is a block diagram of a control system of the controller. (A) is a figure which shows the time domain data for explaining MT method, (B) is a figure which shows the variation|change_quantity and existence amount for every sample line extracted from the time domain data of (A). Is. It is a figure for demonstrating the multivariate data used when creating a normal space. FIG. 7 is a diagram for explaining the operation of the substrate transfer machine that is preferably used in the embodiment of the present disclosure. (A) is a flow chart for explaining the operation of an analysis apparatus preferably used in an embodiment of the present disclosure, and (B) is a real-time calculation result of the Mahalanobis distance (MD value) by the analysis apparatus. FIG. (A) is a figure which shows an example of time domain data of the output value by the vibration sensor in the transfer operation at the time of normal, (B) is the feature-value extracted using the time domain data of (A). FIG. (A) is a figure which shows an example of the time domain data of the output value by the vibration sensor in the transfer operation at the time of abnormality, (B) is a feature-value extracted using the time domain data of (A). FIG.

Hereinafter, an embodiment of the present disclosure will be described.

(1) Configuration of Substrate Processing Apparatus As shown in FIG. 1, in the present embodiment, the substrate processing apparatus 4 is configured as a vertical heat treatment apparatus (batch type vertical heat treatment apparatus) for performing a heat treatment step in an IC manufacturing method. Has been done. In this vertical heat treatment apparatus, a FOUP (Front Opening Unified Pod) 20 is used as a carrier for transporting a wafer W as a substrate. The substrate processing apparatus 4 includes a processing furnace 8, a housing chamber 12, and a transfer chamber 16 described later.

(Accommodation room)
An accommodation chamber 12 for carrying the pod 20 into the apparatus and accommodating the pod 20 is disposed on the front side of the housing of the substrate processing apparatus 4. On the front side of the housing of the housing chamber 12, a loading/unloading port 22A that is an opening for loading/unloading the pod 20 into/from the housing chamber 12 is provided so as to communicate the inside and outside of the housing of the housing chamber 12. The loading/unloading port 22A may be configured to be opened and closed by a front shutter. An AGV port (I/O stage) 22 as a load port (pod mounting device) is provided inside the housing of the carry-in/out port 22A. A transfer port 42 is installed on the wall surface between the storage chamber 12 and the transfer chamber 16. The pod 20 is loaded into the substrate processing apparatus 4 on the AGV port 22 by an in-process transfer apparatus (inter-step transfer apparatus) outside the substrate processing apparatus 4, and is also unloaded from the AGV port 22.

A storage shelf (pod shelf) 30A for storing the pod 20 is installed in two steps above and below the AGV port 22 in the front of the housing of the storage chamber 12. In addition, storage shelves (pod shelves) 30B that store the pods 20 are installed in a matrix at the rear of the housing of the storage chamber 12.

∙ The OHT ports 32 as load ports are installed side by side on the same straight line in the horizontal direction as the upper storage rack 30A in front of the housing. The pod 20 is carried into the OHT port 32 from above the substrate processing apparatus 4 by the intra-process carrying apparatus (inter-process carrying apparatus) outside the substrate processing apparatus 4, and is carried out from the OHT port 32. The AGV port 22, the storage shelf 30A, and the OHT port 32 are configured so that the pod 20 can be slid in the front-rear direction between the mounting position and the delivery position by the horizontal drive mechanism 26. Hereinafter, the AGV port 22 may be referred to as a first load port and the OHT port 32 may be referred to as a second load port.

As shown in FIG. 2, the space between the front storage rack 30</b>A and the rear storage rack 30</b>B in the housing of the storage chamber 12 forms a pod transfer area 14. In the pod transfer area 14, the pod 20 is connected. Are delivered and transported. A rail mechanism 40A is formed on the ceiling of the pod transport area 14 (the ceiling of the accommodation chamber 12) as a traveling path of the pod transport mechanism 40 as a pod transport device. Here, the delivery position is located in the pod transport area 14, and is, for example, a position directly below the pod transport mechanism 40.

The pod transport mechanism 40 that transports the pod 20 includes a traveling unit 40B that travels along a traveling path, a holding unit 40C that holds the pod 24, and an elevating unit 40D that is coupled to the traveling unit 40B and that vertically elevates the holding unit 40C. Prepare By detecting the encoder of the motor that drives the traveling unit 40B, the position in the traveling path 40B can be detected, and the traveling unit 40B can be moved to an arbitrary position.

(Transfer room)
A transfer chamber 16 is formed adjacent to the rear of the storage chamber 12. On the transfer chamber 16 side of the accommodation chamber 12, a plurality of wafer loading/unloading ports for loading/unloading the wafer W into/from the transporting chamber 16 are arranged in a line in the horizontal direction, and are transferred to each wafer loading/unloading port. Each port 42 is installed. At the transfer port 42, the mounting table 42B on which the pod 20 is mounted is horizontally moved and pressed against the wafer loading/unloading port, and the pod is opened by a FIMS (Front-Opening Interface Standard) opener (not shown) as a lid opening/closing mechanism (lid opening/closing device). Twenty lids are unfolded. When the lid of the pod 20 is unfolded, the wafer W is transferred into and out of the pod 20 by the substrate transfer device 86 as a substrate transfer device. The substrate transfer machine 86 is fixedly installed in the transfer chamber 16 and transfers the wafer W between the boat 58 and the pod 20 described later.

(Processing furnace)
The processing furnace 8 is provided above the transfer chamber 16. The processing furnace 8 is provided with a reaction tube 50 that constitutes a reaction container (processing container). The reaction tube 50 is formed in a cylindrical shape having an upper end closed and a lower end opened. A processing chamber 54 is formed in the hollow portion of the reaction tube 50. The processing chamber 54 is configured to be accommodated by a boat 58 as a substrate holder that holds a wafer W as a substrate in a horizontal posture and in a vertically aligned state in multiple stages.

A plurality of nozzles are provided in the processing chamber 54 so as to penetrate the lower portion of the reaction tube 50, and a plurality of types of processing gases are supplied to the processing chamber 54.

A seal cap 78 is installed below the reaction tube 50 as a furnace port lid that can hermetically close the lower end opening of the reaction tube 50. An O-ring is provided on the upper surface of the seal cap 78 as a seal member that comes into contact with the lower end of the reaction tube 50. The seal cap 78 is configured to come into contact with the lower end of the reaction tube 50 from the vertically lower side.

The boat 58 arranges a plurality of wafers W, for example, 25 to 200 wafers W in a horizontal posture and vertically aligned with the centers thereof aligned with each other in a multi-stage manner, that is, at an interval. Is configured to let.

A rotation mechanism 80 as a boat rotation device that rotates the boat 58 is installed on the opposite side of the seal cap 78 from the processing chamber 54. The rotating shaft 80A of the rotating mechanism 80 penetrates the seal cap 78 and is connected to the boat 58. The rotation mechanism 80 is configured to rotate the wafer W by rotating the boat 58. An auxiliary sensor 301 as an auxiliary vibration sensor that detects the vibration of the boat 58 is attached to the rotating shaft 80A.

(2) Configuration of Substrate Transfer Machine Next, the configuration of the substrate transfer machine 86 will be described in detail with reference to FIG.

The substrate transfer machine 86 is mainly acoustically coupled to the tweezers 86a as an end effector for directly operating the wafer W, a drive mechanism 308 for moving the tweezers 86a, and the tweezers 86a, thereby vibrating the tweezers 86a. The vibration sensor 300 for detecting and the rotation mechanism 310 for rotating the drive mechanism 308 around the vertical axis are provided.

The tweezers 86a have, for example, a U-shaped thin plate shape, a plurality of (five in the present embodiment), and are horizontally provided at equal intervals in the vertical direction. The tweezers 86a are of a scooping type and have a guide or a step provided on the upper surface at a position slightly outside the end of the wafer W to restrict the position of the wafer W, and the wafer W is simply placed in the guide. Alternatively, the tweezers 86a can be selected from an edge grip type, a drop edge grip type, a suction type, and the like. The attraction type is a vacuum attraction, Bernoulli chuck, or Johnsen-Rahbek type electrostatic chuck, and the wafer W can be handled in a non-contact manner.

The drive mechanism 308 is configured as a linear (linear) type, and mainly has a fixed part 304 and a guide part 302. The fixing portion 304 holds the roots of the tweezers 86a and fixes them at equal intervals. A drive mechanism for the edge grip, a variable distance mechanism, a horizontal/vertical fine movement mechanism, and the like may be provided in the fixed portion 304.

The guiding section 302 holds the fixing section 304 from below and moves it back and forth on one horizontal axis. In the guide portion 302, for example, two guide rails 302a (not shown) that guide the tweezers 86a in the uniaxial direction are formed substantially parallel to each other. Then, the fixed portion 304 slides along the guide rail 302a, whereby the tweezers 86a move forward and backward with respect to the guide portion 302 and are moved in the front-rear direction.
The rotating mechanism 310 rotates the guide part 302 in the left-right direction while holding the guide part 302 near the center of the lower surface.

Further, the substrate transfer machine 86 is mounted on the substrate transfer machine elevator 306 described later. The substrate transfer machine elevator 306 holds the rotation mechanism 310 of the substrate transfer machine 86 and moves the substrate transfer machine 86 up and down.

That is, the tweezers 86 a of the substrate transfer machine 86 are moved forward and backward along the guide rails 302 a by the drive mechanism 308, and are rotated in the left and right directions by the rotation of the drive mechanism 308 by the rotation mechanism 310, and the substrate transfer machine elevator 306. It is moved up and down by moving up and down.

The vibration sensor 300 is a triaxial acceleration sensor. The vibration sensor 300 is provided on the fixed portion 304 at the base of the stacked tweezers 86a. Vibration sensor 300 may require an amplifier (not shown). The amplifier is preferably provided near the vibration sensor 300 and may be provided in the drive mechanism 308, for example.

(3) Configuration of Controller As shown in FIG. 4, a controller 210, which is a control unit (control means), includes a CPU (Central Processing Unit) 212, a RAM (Random Access Memory) 214, a storage device 216, and an I/O port 218. Is configured as a computer equipped with. The RAM 214, the storage device 216, and the I/O port 218 are configured to exchange data with the CPU 212 via the internal bus 220. An input/output device 222 configured as, for example, a touch panel or the like is connected to the controller 210. In addition, the controller 210 is connected to an analysis device 400 that detects an abnormality during the transfer operation of the wafer W.

The storage device 216 is composed of, for example, a flash memory, an HDD (Hard Disk Drive), or the like. In the storage device 216, a control program for controlling the operation of the substrate processing apparatus, a process recipe in which a procedure and conditions of the substrate processing described later, and the like are stored in a readable manner. The process recipe is a combination that causes the controller 210 to execute each procedure in the substrate processing process described below and obtains a predetermined result, and functions as a program. Hereinafter, the process recipe, the control program, and the like are collectively referred to simply as a program. When the term program is used in the present specification, it may include only the process recipe, only the control program, or both. The RAM 214 is configured as a memory area (work area) in which programs and data read by the CPU 212 are temporarily stored.

The I/O port 218 is connected to the pod transport mechanism 40, the horizontal drive mechanism 26, the

sensors

25B and 28A, the drive mechanism 308, the substrate transfer machine elevator 306, the rotation mechanism 310, the rotation mechanism 80, the boat elevator 82, etc. described above. ing.

The CPU 212 is configured to read the control program from the storage device 216 and execute the control program, and read the process recipe from the storage device 216 in response to input of an operation command from the input/output device 222. The CPU 212 charges the wafer W of the substrate transfer machine 86 by the transfer operation of the pod 20 by the pod transfer mechanism 40, the drive mechanism 308, the rotation mechanism 310, and the substrate transfer machine elevator 306 so as to follow the contents of the read process recipe. And a discharging operation, and an up-and-down operation of the boat 58 by the boat elevator 82 are controlled.

The controller 210 can be configured by installing the above program stored in an external storage device (for example, a magnetic disk such as a hard disk, an optical disk such as a CD, a semiconductor memory such as a USB memory) 224 into a computer. it can. The storage device 216 and the external storage device 224 are configured as a computer-readable recording medium. Hereinafter, these are collectively referred to simply as a recording medium. When the term “recording medium” is used in this specification, it may include only the storage device 216 alone, may include only the external storage device 224 alone, or may include both of them. Note that the program may be provided to the computer by using communication means such as the Internet or a dedicated line without using the external storage device 224.

(4) Operation of Substrate Processing Apparatus Next, a series of operations using the above-described substrate processing apparatus 4 will be described.

(Carrier loading process: S10)
When the pod 20 is supplied to the AGV port 22 or the OHT port 32, the pod 20 above the AGV port 22 or the OHT port 32 is carried into the substrate processing apparatus 4. The loaded pod 20 is automatically transported by the pod transport mechanism 40 to the designated stage 25 of the storage shelf 30 and transferred to and temporarily stored, and then is transferred from the storage shelf 30 to one transfer port 42. It is transported and delivered, or directly transported to the transfer port 42.

(Lid development step: S11)
When the pod 20 is mounted on the mounting portion of the transfer port 42, the lid of the pod 20 is opened by the FIMS opener.

(Wafer charging process: S12)
When the lid of the pod 20 is opened, the plurality of wafers W in the pod 20 are loaded into the boat 58 (wafer charge) by the substrate transfer machine 86. Prior to this, mapping for confirming the storage state of the wafer W in the pod 20 may be performed.

(Boat loading process: S13)
When the plurality of wafers W are loaded into the boat 58, the boat 58 is loaded into the processing chamber 54 by the boat elevator 82 (boat loading). At this time, the seal cap 78 is in a state of hermetically closing the lower end of the reaction tube 50 via the O-ring.

(Film forming process: S14)
Then, by supplying a processing gas to the wafer W in the processing chamber 54, a film is formed on the wafer W.

After the film forming process is completed, the processing chamber 54 is purged, and the gas and reaction byproducts remaining in the processing chamber 54 are removed from the processing chamber 54. After that, the atmosphere in the processing chamber 54 is replaced with an inert gas, and the pressure in the processing chamber 54 is returned to normal pressure.

(Boat unloading process: S15)
After the atmospheric pressure is restored, the boat elevator 82 lowers the seal cap 78 to open the lower end of the reaction tube 50. Then, the processed wafer W is unloaded from the lower end of the reaction tube 50 to the outside of the reaction tube 50 while being supported by the boat 58 (boat unloading).

(Wafer discharging process: S16)
The processed wafer W is taken out from the boat 58 by the substrate transfer machine 86 (wafer discharge).

(Carrier unload process: S17)
The processed wafer W is stored in the pod 20 by the substrate transfer machine 86, and the pod 20 in which the processed wafer W is stored is subjected to the operation opposite to the carrier load, and the load port (AGV port 22, OHT port It is returned to 32) and collected by an external transfer device.

(5) Analysis Device Next, the analysis device 400 in this embodiment will be described.

The analysis apparatus 400 according to the present embodiment detects an abnormality during the transfer operation of the wafer W in the wafer charging step (S12) and the wafer discharging step (S16) using the MT (Mahalanobis Taguchi) method. is there.

The MT method is a method of multivariate analysis in quality engineering, and is based on the idea that a normal state is judged by calculating a distance (Mahalanobis distance, MD value) from the normal state as a reference. is there.

The analysis apparatus 400 according to the present embodiment generates multivariate data by extracting the characteristic amount from the time domain data of the output value detected by the vibration sensor 300, and the multi-variate data is generated during the transfer operation of the wafer W using the MT method. It is configured to detect anomalies. Further, as shown in FIG. 4, the analysis device 400 is connected to the vibration sensor 300 and the auxiliary sensor 301, and is configured to be able to acquire the output values thereof.

First, a specific method of extracting the amount of change and the amount of existence as feature amounts from the time domain data of the output value of the vibration sensor 300 will be described with reference to FIGS. 5(A) and 5(B). Note that FIG. 5A is data obtained by acquiring the acceleration of a certain axis at a sample rate of 100 sps during 19.5 seconds during the transfer operation. FIG. 5B is a diagram showing the amount of change and the amount of existence for each sample line extracted from the time domain data shown in FIG. 5A.

Here, the amount of change is the differential characteristic of the time domain data, and is the number of times the sample lines set in the time domain data straddle each sample line. The abundance is the integral characteristic of the time domain data, and is the number of data having a value larger than the sample line for each sample line set in the time domain data.

Further, as shown in FIG. 5A, a plurality of (7 in FIG. 5A) sample lines are preliminarily arranged in the time axis direction with respect to the time domain data of the output value (analog value) of the vibration sensor 300. (Y=-40, -20, 10, 20, 25, 50, 75) is set. Each sample line is set as a plurality of thresholds, and a value to be monitored or a value characteristically showing a change in environment is set.

Then, for each sample line, the number of times the waveform crosses the sample line is counted as the amount of change. Further, for each sample line, the number of values larger than the sample line is calculated as the existing amount. That is, the feature amount of two items of the variation amount and the existing amount is extracted for one sample line. In addition to the amount of change and the amount of existence, the maximum value, the minimum value, the average value, and the like of the waveform may be used as the feature amount.

Then, as shown in FIG. 5(B), sample data that is 14-dimensional multivariate data extracted by using the amount of change and the amount of existence in the transfer operation under normal conditions as the feature amount is calculated.

Then, as shown in FIG. 6, N pieces of sample data calculated using a common sample line are calculated from the time domain data in a plurality of times (N pieces in FIG. 6) of transfer operation in a normal state, A normal space is created using the N sample data. Specifically, the average vector and the covariance matrix are calculated using N sample data in the normal state. The normal space is a space spanned by the eigenvectors of the covariance matrix.

Then, the Mahalanobis distance MD in the normal space is calculated as a data vector (column vector) which is multivariate data created based on the time domain data of the output value detected by the vibration sensor 300 during the transfer operation. Using the mean vector of multivariate data in the normal space and the inverse matrix of the covariance matrix, it is shown as follows.

If the Mahalanobis distance (MD value) exceeds a preset threshold value, it is determined that an abnormal situation has occurred during the transfer operation. The time domain data on which the multivariate data is based preferably has the same number of samples (1950 samples), but the multivariate data can also be divided by the number of samples for normalization. Further, the Mahalanobis distance may be threshold-determined with the squared value thereof being kept.

Next, the operation of the analysis device 400 according to the present embodiment will be described with reference to FIGS. 7, 8A and 8B.

FIG. 7 is a view showing how the wafer W is transferred by the substrate transfer machine 86 between the pod 20 of the transfer port 42 and the boat 58 in the transfer chamber 16. A camera 226 is installed in the transfer chamber 16, and images of how the wafer W is transferred and how the boat 58 is moved up and down are photographed and recorded in response to a command from the controller 210.

The analysis apparatus 400 is configured to transfer the wafer W to the boat 58 during the wafer charging process (S12) and the wafer discharging process (S16), which are transfer processes, and the tweezers 86a, the wafer W on the tweezers 86a, and the boat 58. It is detected that the wafer W and the wafer 58 on the boat 58 are in contact with each other.

For example, in the wafer charging step (S12) described above, the controller 210 controls the substrate transfer machine 86, the rotation mechanism 310, and the substrate transfer machine elevator 306 to transfer the wafer W to be transferred. Abnormality is detected in all or part of the transfer operation. Specifically, with the tweezers 86a facing the wafer W to be transferred in the pod 20, the transfer machine 86 advances the tweezers 86a to the lower side of the wafer W and raises the tweezers 86a to lift the wafer W. After mounting, the tweezers 86a are retracted. Then, the substrate transfer machine elevator 306 moves up and down to the mounting position of the wafer W on the boat 58. Further, the rotation mechanism 310 rotates the transfer device 86 in the left-right direction so that the transfer device 86 faces the boat 58. Then, the tweezers 86a are moved forward to the wafer mounting position of the boat 58 by the transfer machine 86 and then moved in the vertical direction to transfer the wafer W to the boat 58 and retract in the backward direction. After that, the transfer machine 86 is moved up and down to the height of the next transfer target wafer W, and is rotated so as to directly face the transfer target wafer W. The controller 210 repeats this series of operations until all the wafers W have been transferred. At this time, an abnormality can be detected in the operation from the forward movement of the tweezers 86a toward the boat 58 to the backward movement thereof, which is a section in which there is a high possibility of contact in the series of operations. In this example, the operation of this abnormality detection section is the same every time, and the required time is constant.

Similarly, in the wafer discharging process described above, the processed wafer W is delivered to the pod 20 by the operation opposite to the wafer charging process.

The analyzer 400 previously creates a normal space for the transfer operation of the wafer W to the boat 58 at the time of adjustment after installation of the device, and records it inside. Specifically, the time domain data of the output value of the vibration sensor 300 when the transfer operation of the wafer W to the boat 58 by the substrate transfer machine 86 is normally performed in the presence of the worker is performed a plurality of times. A normal space is created by setting acquired sample lines and analyzing a plurality of acquired time domain data by the MT method.

FIG. 8A is a flowchart for explaining the operation of the analysis device 400. During operation, the controller 210 transmits signals indicating the start and end of the abnormality detection section in transfer to the analysis device 400. The analysis device 400 receives the start signal, starts the operation of FIG. 8, and continues until it receives the end signal. In this example, in the abnormality detection section, time domain data is acquired at 100 sps to 100 ksps (samples/second), and feature quantity extraction (multivariate data conversion) and Mahalanobis distance calculation are performed with time domain data of less than 2048 samples as a unit. Calculate. At this time, it is desirable to calculate the Mahalanobis distance within 3 seconds after the occurrence of the contact event. More preferably, the Mahalanobis distance calculation cycle is 0.5 seconds or less, or the time from the start of acquisition of the time domain data to the Mahalanobis distance calculation is 1 second or less.

First, the analysis apparatus 400 acquires M samples of the time domain data of the output value of the vibration sensor 300 while the substrate transfer machine 86 performs the transfer operation (step S100). Here, M is the same number as the number of samples of the time domain data used when creating the normal space, which is 1950 in this example. M may be set such that the time of M samples is an integer multiple of the period of the particular noise (eg, power hum). The acquired M samples may partially overlap the already acquired data. For example, time domain data can be acquired using a sliding window that is set while shifting by M/d samples. Where d is the number of divisions and is selected so that 2≦d≦M and M/d is an integer.

Next, multivariate data is generated by extracting the amount of change and the amount of existence for each sample line as a feature amount based on the time domain data acquired in step S100 (step S101). When a sliding window is used, this processing may be performed and held in units of M/d samples, and the latest d pieces may be added together.

Next, the Mahalanobis distance in the normal space is calculated for each transfer operation from the multivariate data created in step S101, the average value of the multivariate data in the normal space, and the covariance matrix (step S102). When the sliding window is used, the squared value of the Mahalanobis distance may be calculated and held in units of M/d samples in the time domain data, and the latest d pieces may be summed.

Next, it is determined whether the calculated Mahalanobis distance exceeds a preset threshold value (step S103). When the Mahalanobis distance does not exceed the preset threshold value (No in step S103), the analysis operation is continued during the transfer of the wafer W to the boat 58 (No in step S104), and the wafer W is transferred to the boat 58. When the conveyance (abnormality detection section) of is completed (Yes in step S104), the analysis operation is stopped. Note that the threshold value determination is not limited to one time determination, and may be performed according to the number of times the threshold value is continuously exceeded, for example.

When the Mahalanobis distance exceeds the preset threshold value (Yes in step S103), it is determined that an abnormal situation has occurred (step S105), and the substrate transfer machine 86 is stopped urgently. That is, it is determined that an abnormal situation has occurred in which the tweezers 86a and the wafer W on the tweezers 86a and the boat 58 and the wafer W on the boat 58 are in contact with each other, and the emergency stop signal is transmitted from the analysis device 400 to the controller 210. The substrate transfer machine 86 is stopped urgently.

FIG. 8B is a diagram showing a time transition of the Mahalanobis distance (MD value) calculated by the analysis device 400. The threshold value is set to about 3.8, and the MD value finally exceeds the threshold value, indicating that an emergency stop is performed. Even if the sample rate of the time domain data is 100 sps, by setting d=50, the Mahalanobis distance is calculated at intervals of about 0.4 seconds, and the abnormality can be detected substantially in real time. The abnormalities that can be detected from the MD value include collisional contact that means that the wafer cannot be used as a product, scratching of the wafer by the tweezers 86a, and the end surface of the wafer W held by the tweezers 86a. Friction with the groove of the boat 58 may also be included. Such friction can be regarded as a sign of a more serious contact, but in a situation where frequent occurrence of emergency stop due to slight friction is not desired, a normal space is created by using data including friction or the MD value An emergency stop can be suppressed by adjusting the threshold value.

FIG. 9A is a diagram showing an example of the time domain data of the output value by the vibration sensor 300 in the transfer operation at the normal time, and FIG. 9B shows the time domain data of FIG. 9A. It is a figure which shows the feature-value extracted using. 10A is a diagram showing an example of the time domain data of the output value by the vibration sensor in the transfer operation at the time of abnormality, and FIG. 10B is the time domain data of FIG. 10A. It is a figure which shows the feature-value extracted using.

The Mahalanobis distance (MD value) in the normal space based on the feature amount extracted from the time domain data shown in FIG. 9(B) was 0.9. In this transfer operation, the MD value is below the threshold value 3.8, so it is determined to be in a normal state, and the transport operation is continued.

On the other hand, the Mahalanobis distance (MD value) in the normal space based on the feature amount extracted from the time domain data shown in FIG. 10(B) was 33.1. In this transfer operation, the MD value greatly exceeds the threshold value, so that it is determined to be in an abnormal state.

The transfer chamber 16 is provided with a camera 226 that captures the transfer operation of the substrate transfer machine 86, and the image captured by the camera 226 is associated with time and stored in the storage device 216 as a recording device or external storage. Recorded in device 224. In other words, the transfer operation by the substrate transfer machine 86 is recorded as an operation log in the storage device 216 or the external storage device 224 in association with the time, and when an abnormal situation is detected by the analysis device 400, a remote connection is made. The input/output device 222 displaying the operation screen is configured so that the user can grasp the time of the transportation error, the location of the transportation error, the content of the transportation error, and the like. Further, when a transport error occurs, the replay video and live video before and after the transport error by the camera 226 can be displayed on the input/output device 222.

Further, the analysis device 400 is generated by extracting the amount of change and the amount of existence based on the time domain data of each of the front, rear, left, right, and top and bottom axes of the vibration sensor 300 and the time domain data of the auxiliary sensor 301 as the feature quantity. By creating a normal space using multiple multivariate data, selecting the corresponding normal space from the multiple normal spaces, and calculating the Mahalanobis distance in the selected normal space, The accuracy is improved. For example, from the three-axis data of the vibration sensor 300, one normal space is created by combining those feature amounts, and from the data of the auxiliary sensor 301, a separate normal space is created. Then, the abnormality due to contact can be determined in consideration of the co-occurrence of the Mahalanobis distance in each normal space. That is, when the two Mahalanobis distances simultaneously exceed the threshold value, the contact is more strongly estimated, otherwise another anomaly is estimated.

Further, the analysis device 400 uses the number of wafers W being transferred, the moving speed and acceleration of the tweezers 86a, the insertion amount of the tweezers 86a into the boat 58, the amount of wind flowing around the boat 58, the temperature of the boat 58, or the boat. A plurality of normal spaces are created for each temperature of the processing chamber accommodating the 58, and a normal state that matches the state when the transfer operation by the substrate transfer machine 86 is performed from the plurality of normal spaces. Anomalies can be detected by selecting a space and calculating the Mahalanobis distance in the selected normal space. This improves the detection accuracy of the abnormality. Alternatively, in the time domain data during transfer operation, if there is a section in which the value tends to be always large (for example, the start or end point of acceleration or deceleration, the vicinity of the switching point of operation, etc.) In addition to the first normal space created from the data including the section, the second normal space is created from the data not including the large value section, and the abnormality is detected in the second normal space in the section. Good. At this time, the time domain data becomes temporally discontinuous before and after the exclusion of the large value section, but the analysis device 400 can treat them as continuous without any problem.

(6) Effects of this Embodiment According to this embodiment, at least one or more of the following effects (a) to (d) are achieved.

(a) According to this embodiment, it is possible to detect with high accuracy that the substrate transfer device 86, the wafer W on the substrate transfer device 86, and the boat 58 and the wafer W on the boat 58 contact each other. Can detect even a slight abnormality before the contact. In other words, there is a possibility that abnormal vibrations before a failure can be detected from the vibrations generated by the substrate transfer machine 86 or the like itself due to aging or the like.

(B) According to this embodiment, since the auxiliary sensor 301 for detecting the vibration of the boat 58 is provided, the contact detection accuracy is improved.

(C) According to this embodiment, an abnormality can be detected and an emergency stop can be performed within 3 seconds after the occurrence of contact. Since the time for the tweezers 86a to advance and retreat once is about 3 seconds, it is expected that an emergency stop will be performed in the middle of the advance and retreat to prevent the spread of wafer damage. Furthermore, since the analysis device 400 acquires the time domain data by the sliding window, the abnormality detection cycle can be shortened with a relatively small calculation load. Therefore, even if the analyzer 400 is configured with a commercially available industrial sequencer, the emergency stop can be performed at a sufficient speed. On the other hand, in spectrum analysis using FFT, a resolution of a predetermined value (for example, 512 points) or more is required, but even if a sliding window is used, past calculation results cannot be used and the calculation load is large. In other words, the calculation cycle is limited by the amount of calculation, and even if the sample rate is increased, the calculation cycle cannot be shortened.

(d) According to the present embodiment, the transfer operation by the substrate transfer machine 86 is recorded in the storage device 216 or the external storage device 224 as an operation log in association with the time, and the analysis device 400 detects an abnormal situation. Then, the input/output device 222 that is remotely connected to display the operation screen is configured so that the user can grasp the time of the transport error, the location of the transport error, the content of the transport error, and the like. ing. Further, when a transport error occurs, a recorded image or live image screen immediately before the transport error by the camera 226 installed can be displayed on the input/output device 222.

(7) Other Embodiments It is needless to say that the present disclosure is not limited to the above embodiments, and various modifications can be made without departing from the scope of the invention.

In the above-described embodiment, a normal space is created in the operation of the substrate transfer machine 86 between the boats 58 from the time the tweezers 86a are faced to the boat 58 until the wafer W is transferred to the boat 58 and retracted. Although the Mahalanobis distance in the normal space is calculated, the present invention is not limited to this, and a normal space may be created for the operation of the substrate transfer machine 86 between the pods 20 and the Mahalanobis distance in the normal space may be calculated. As a result, the contact between the substrate transfer device 86, the wafer W on the substrate transfer device 86, and the pod 20 or the wafer W in the pod 20 is detected with high accuracy, and a slight abnormality before the failure state is detected. Can be detected.

Further, in the above-described embodiment, the example in which the auxiliary sensor 301 is attached to the rotating shaft 80A as the auxiliary vibration sensor to detect the vibration of the boat 58 has been described, but the present invention is not limited to this, and the auxiliary sensor 301 is not limited to the rotating mechanism 80 or the boat. It may be attached to a heat insulation unit provided between 58 and the rotating shaft 80A, or may be attached to a table on which the boat 58 is placed. As the auxiliary vibration sensor, a sound collecting microphone that has directivity toward the tip of the tweezers 86a and collects vibrations in the air may be used. At this time, the sound collecting microphone is mounted on the substrate transfer machine 86 or in the transfer chamber 16. A laser Doppler vibrometer that irradiates the boat 58 with laser light to detect the vibration may be used as the auxiliary vibration sensor. At this time, the laser Doppler vibrometer is installed in the transfer chamber 16. Note that a plurality of auxiliary vibration sensors may be used. The accuracy of the abnormality detection of the analysis device 400 is improved by using the plurality of auxiliary vibration sensors.

Further, in the above-described embodiment, the example in which the vibration sensor 300 is provided in the fixing portion 304 that stacks and fixes the roots of the plurality of tweezers 86a has been described. A sensor may be provided. In this case, a plurality of normal spaces are created for each time domain data of the output value of the vibration sensor for each tweezer 86a, and the Mahalanobis distance in the corresponding normal space is calculated from the plurality of normal spaces. Thereby, the tweezers 86a in which the abnormality is detected can be specified.

Although the MT method is applied to the signal of the vibration sensor 300 or the like without any special preprocessing in the above-described embodiment, the present invention is not limited to this, and preprocessing such as a filter may be performed. .. The pre-processing can be configured as a filter that suppresses noise of a specific frequency or passes a frequency that is peculiar to contact, or an equalizer that compensates the frequency characteristic of a detection system including a tweezer having a natural frequency. When the frequency generated from scratch depends on the moving speed of the substrate transfer machine 86, or when the variable frequency vacuum pump is used and the noise frequency fluctuates, the center frequency and cutoff frequency of the filter are changed so as to tune to the frequency. Can be controlled.

In the above embodiments, the case of processing a wafer has been described, but the present disclosure can be applied to all substrate processing apparatuses that process substrates such as glass substrates of liquid crystal panels and magnetic disks and optical disks. Further, in the above-described embodiment, an example in which a batch-type substrate processing apparatus that processes a plurality of substrates at one time is used has been described. The present disclosure is not limited to the above-described embodiment, and can be suitably applied to, for example, the case of using a single wafer processing apparatus that processes one or several substrates at a time.

(8) Preferred Embodiments of the Present Disclosure Hereinafter, preferred embodiments of the present disclosure will be additionally described.

(Appendix 1)
According to one aspect of the present disclosure,
A substrate holder for holding the substrate,
A substrate transfer machine that includes an end effector that directly operates the substrate and a drive mechanism that moves the end effector, transfers the substrate to the substrate holder, or transfers the substrate from the substrate holder. ,
A vibration sensor for detecting the vibration of the end effector,
A control unit for controlling the substrate transfer machine,
A normal space created by analyzing the time domain data of the vibration of the end effector detected by the vibration sensor when the substrate transfer operation by the substrate transfer machine is normally performed by the MT method is used. , The Mahalanobis distance in the normal space of the time domain data detected by the vibration sensor while the substrate transfer machine places the substrate and performs a predetermined transfer operation, and the calculated Mahalanobis distance A substrate on the end effector or the end effector, and an analysis device for determining that the substrate holder or the substrate on the substrate holder are in contact with each other when the threshold value exceeds a predetermined threshold value,
The substrate processing apparatus is provided in which the control unit outputs a signal indicating the start and end of the predetermined transfer operation to the analysis apparatus.

(Appendix 2)
The substrate processing apparatus according to Appendix 1, preferably:
The analysis device extracts the amount of change in the time domain data, which is the number of times the sample line is crossed for each of a plurality of predetermined sample lines, and the existing amount, which is the number of data having a value larger than the sample line, as feature amounts. To generate multivariate data,
An average value of multivariate data generated based on a plurality of time domain data of vibration of the end effector detected by the vibration sensor when the substrate transfer operation by the substrate transfer machine is normally performed, and Calculating a covariance matrix and defining the normal space using the covariance matrix,
The Mahalanobis distance in the normal space is calculated during the transfer operation from the multivariate data created based on the time domain data detected by the vibration sensor during the transfer operation by the substrate transfer machine. ..

(Appendix 3)
The substrate processing apparatus according to attachment 2, preferably:
The analysis device includes the number of substrates being transferred, the moving speed of the end effector, the insertion amount of the end effector into the substrate holder, the air volume of air flowing around the substrate holder, and the substrate holder. A plurality of normal spaces are stored for each temperature or each temperature of the processing chamber that accommodates the substrate holder,
A normal space that matches the state when the transfer operation by the substrate transfer machine is performed is selected from the plurality of normal spaces, and the Mahalanobis distance in the selected normal space is calculated.

(Appendix 4)
The substrate processing apparatus according to Supplementary Note 2 or Supplementary Note 3, preferably:
The substrate transfer machine is fixedly installed in the substrate processing apparatus, and transfers the substrate to and from the substrate holder carried out from the processing chamber,
The vibration sensor is a triaxial acceleration sensor,
An auxiliary sensor attached to a member holding the substrate holder, a sound collecting microphone that collects vibrations in the air having directivity toward the tip of the end effector, or laser light on the substrate holder. Further comprising an auxiliary vibration sensor including at least one of a laser Doppler vibrometer for irradiating the
The analysis device, a plurality of multivariate data generated by extracting the amount of change and the amount of presence based on each of the time domain data of each axis of the vibration sensor and the time domain data of the auxiliary vibration sensor as a feature amount. Is used to create the normal space.

(Appendix 5)
The substrate processing apparatus according to Supplementary Note 2 or Supplementary Note 3, preferably:
The number of time domain data when creating the normal space is 24576 samples or less, and the multivariate data is multivariate at intervals of less than 1024 samples.

(Appendix 6)
The substrate processing apparatus according to Appendix 1, preferably:
A camera for capturing the transfer operation, and a recording device for recording the captured image in association with time,
The control unit controls the transfer operation to be recorded in the recording device as an operation log in association with time.

(Appendix 7)
According to another aspect of the disclosure,
A substrate transfer machine provided with an end effector that directly operates the substrate and a drive mechanism that moves the end effector transfers the substrate to a substrate holder that holds the substrate, or transfers the substrate from the substrate holder. And a transfer process for transferring
Time domain of vibration of the end effector detected by a vibration sensor that detects vibration of the end effector when the substrate transfer operation by the substrate transfer machine in the transfer step is normally performed by the analysis device Using the normal space created by analyzing the data by the MT method, the substrate transfer device detects the vibration while the substrate is placed and a predetermined transfer operation is performed in the transfer step. The Mahalanobis distance in the normal space of the time domain data is calculated, and when the calculated Mahalanobis distance exceeds a predetermined threshold value, the end effector or the substrate on the end effector, the substrate holder or the substrate holding An analysis step of determining that an abnormal situation has occurred in which the board on the tool is in contact with each other;
An output step of outputting a signal indicating the start and end of a predetermined transfer operation by the substrate transfer machine to the analysis device;
A method for manufacturing a semiconductor device is provided.

4 Processing equipment (substrate processing equipment)
20 pods (container)
58 Boat (substrate holder)
86 Substrate transfer machine 210 Controller 300 Vibration sensor 301 Auxiliary sensor 400 Analysis device

Claims

A substrate holder for holding the substrate,
A substrate transfer machine that includes an end effector that directly operates the substrate and a drive mechanism that moves the end effector, transfers the substrate to the substrate holder, or transfers the substrate from the substrate holder. ,
A vibration sensor for detecting the vibration of the end effector,
A control unit for controlling the substrate transfer machine,
A normal space created by analyzing the time domain data of the vibration of the end effector detected by the vibration sensor when the substrate transfer operation by the substrate transfer machine is normally performed by the MT method is used. , The Mahalanobis distance in the normal space of the time domain data detected by the vibration sensor while the substrate transfer machine places the substrate and performs a predetermined transfer operation, and the calculated Mahalanobis distance A substrate on the end effector or the end effector, and an analysis device for determining that the substrate holder or the substrate on the substrate holder are in contact with each other when the threshold value exceeds a predetermined threshold value,
The substrate processing apparatus wherein the control unit outputs a signal indicating the start and end of the predetermined transfer operation to the analysis apparatus.
The analysis device extracts the amount of change in the time domain data, which is the number of times the sample line is crossed for each of a plurality of predetermined sample lines, and the existing amount, which is the number of data having a value larger than the sample line, as feature amounts. To generate multivariate data,
An average value of multivariate data generated based on a plurality of time domain data of vibration of the end effector detected by the vibration sensor when the substrate transfer operation by the substrate transfer machine is normally performed, and Calculating a covariance matrix and defining the normal space using the covariance matrix,
The Mahalanobis distance in the normal space is calculated during the transfer operation from the multivariate data created based on the time domain data detected by the vibration sensor during the transfer operation by the substrate transfer machine. The substrate processing apparatus according to claim 1.
The analysis device includes the number of substrates being transferred, the moving speed of the end effector, the insertion amount of the end effector into the substrate holder, the air volume of air flowing around the substrate holder, and the substrate holder. A plurality of normal spaces are stored for each temperature or each temperature of the processing chamber that accommodates the substrate holder,
The Mahalanobis distance in the selected normal space is calculated by selecting a normal space that matches the state when the transfer operation is performed by the substrate transfer machine from the plurality of normal spaces. Substrate processing equipment.
The substrate transfer machine is fixedly installed in the substrate processing apparatus, and transfers the substrate to and from the substrate holder carried out from the processing chamber,
The vibration sensor is a triaxial acceleration sensor,
An auxiliary sensor attached to a member holding the substrate holder, a sound collecting microphone that collects vibrations in the air having directivity toward the tip of the end effector, or laser light on the substrate holder. Further comprising an auxiliary vibration sensor including at least one of a laser Doppler vibrometer for irradiating the
The analysis device, a plurality of multivariate data generated by extracting the amount of change and the amount of presence based on each of the time domain data of each axis of the vibration sensor and the time domain data of the auxiliary vibration sensor as a feature amount. The substrate processing apparatus according to claim 2, wherein the normal space is created by using the normal space.
A substrate transfer machine provided with an end effector that directly operates the substrate and a drive mechanism that moves the end effector transfers the substrate to a substrate holder that holds the substrate, or transfers the substrate from the substrate holder. And a transfer process for transferring
A time region of vibration of the end effector detected by a vibration sensor that detects vibration of the end effector when the substrate transfer operation by the substrate transfer machine is normally performed in the transfer step by the analysis device. The normal space created by analyzing the data by the MT method is used, and in the transfer step, the substrate transfer device detects the vibration while the substrate is mounted and a predetermined transfer operation is performed. The Mahalanobis distance in the normal space of the time domain data is calculated, and when the calculated Mahalanobis distance exceeds a predetermined threshold, the end effector or the substrate on the end effector, the substrate holder or the substrate holding An analysis step of determining that an abnormal situation has occurred in which the board on the tool is in contact with each other;
An output step of outputting a signal indicating the start and end of a predetermined transfer operation by the substrate transfer machine to the analysis device;
A method for manufacturing a semiconductor device, comprising: