Determination method, determination program, substrate processing device, and information processing device

Abstract

To detect an occurrence of an embedding failure.SOLUTION: A determination method includes a step a) and a step b). In the step a), a pattern including a recessed part is formed, spectral measurement of a substrate in which an embedding material is embedded in the recessed part is performed, and either spectrum of an absorbance spectrum or a reflectance spectrum of the substrate is measured. In the step b), an embedded state of the recessed part is determined from an integration value of a spectrum value in each wavelength range of one spectrum of the substrate measured by the step a) by using a model generated by mechanically learning data obtained by associating the integration value of the spectrum value in each wavelength range about a plurality of wavelength ranges of the one spectrum of a plurality of substrates in which the embedding material is embedded with the embedded state of the recessed part.SELECTED DRAWING: Figure 14

Description

[Technical field]

[0001] The present disclosure relates to a determination method, a determination program, a substrate processing apparatus, and an information processing apparatus. [Background technology]

[0002] Patent Document 1 listed below discloses a substrate processing method including: "a first step of supplying a processing gas onto a substrate for manufacturing a semiconductor device, the substrate having an upper surface on which a surface of a first layer is exposed and a recess in which a second layer is exposed on at least the side walls and bottom surface of the substrate, to form a third layer and fill the recess with the third layer; a second step of etching the third layer and stopping the etching when the upper surface of the substrate is exposed and the third layer remains in the recess; and a latter step of supplying a processing gas onto the substrate to form a third layer and fill the recess with the third layer, the second step being characterized in that when the processing gas is supplied, an incubation time on the surface of the first layer is longer than an incubation time on the surface of the second layer." [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2017-174902 A Summary of the Invention [Problem to be solved by the invention]

[0004] The present disclosure provides a technique for detecting the occurrence of filling defects. [Means for solving the problem]

[0005] The determination method according to one embodiment of the present disclosure includes steps a) and b). In step a), a spectroscopic measurement is performed on a substrate on which a pattern including recesses is formed and a filling material is filled into the recesses, and one of the absorbance spectrum or reflectance spectrum of the substrate is measured. In step b), the filling state of the recesses is determined from the integral value of the spectral value for each wavelength range of the spectrum of one of the substrates measured in step a) using a model generated by machine learning data that associates the integral value of the spectral value for each wavelength range of the spectrum of one of the substrates on which the filling material is filled with the recesses with the filling state of the recesses. Effect of the Invention

[0006] According to the present disclosure, the occurrence of filling defects can be detected. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is a schematic cross-sectional view showing an example of a schematic configuration of a film forming apparatus according to an embodiment. [Diagram 2] FIG. 2 is a diagram showing a state in which the substrate is raised from the mounting table in the film forming apparatus according to the embodiment. [Diagram 3] FIG. 3 is a schematic configuration diagram showing another example of the film forming apparatus according to the embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of a schematic configuration of a control device according to the embodiment. [Diagram 5] FIG. 5 is a diagram illustrating an example of a substrate before film formation according to an embodiment. [Figure 6] FIG. 6 is a diagram illustrating an example of a substrate after a film is formed according to an embodiment. [Figure 7] FIG. 7 is a diagram illustrating an example of an index for the filling state of the recessed portion according to the embodiment. [Figure 8A] FIG. 8A is a diagram illustrating an example of an index of the state of the base of the substrate according to the embodiment. [Figure 8B] FIG. 8B is a diagram illustrating an example of an index of the state of the base of the substrate according to the embodiment. [Figure 9] FIG. 9 is a diagram illustrating an example of data stored in the learning data according to the embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of data stored in the learning data according to the embodiment. [Figure 11A] FIG. 11A is a diagram illustrating an example of an overall flow of a determination method according to an embodiment. [Figure 11B] FIG. 11B is a diagram illustrating another example of the overall flow of the determination method according to the embodiment. [Figure 11C] FIG. 11C is a diagram illustrating another example of the overall flow of the determination method according to the embodiment. [Figure 11D] FIG. 11D is a diagram illustrating another example of the overall flow of the determination method according to the embodiment. [Figure 11E] FIG. 11E is a diagram illustrating another example of the overall flow of the determination method according to the embodiment. [Figure 12] FIG. 12 is a diagram showing an example of the S / N ratio of the reflectance spectrum. [Figure 13] FIG. 13 is a flowchart showing an example of the flow of a generation process including a generation step of the determination method according to the embodiment. [Figure 14] FIG. 14 is a flowchart showing an example of the flow of a film forming process including a determination step of the determination method according to the embodiment. [Figure 15] FIG. 15 is a diagram showing an example of a cross-sectional configuration of a substrate W used for the inspection. [Figure 16A] FIG. 16A is a diagram for explaining the verification results of Case 1. In FIG. [Figure 16B] FIG. 16B is a diagram for explaining the verification results of Case 2. In FIG. [Figure 16C] FIG. 16C is a diagram for explaining the verification results of Case 3. In FIG. [Figure 17] FIG. 17 is a diagram for explaining the verification results when the width of the wavelength range is changed. [Figure 18] FIG. 18 is a schematic configuration diagram showing another example of the film forming apparatus according to the embodiment. [Figure 19] FIG. 19 is a diagram illustrating an example of a schematic configuration of a measurement unit according to an embodiment. [Figure 20] FIG. 20 is a diagram illustrating another example of a schematic configuration of the measurement unit according to the embodiment. [Figure 21] FIG. 21 is a diagram illustrating an example of a schematic configuration of an information processing device according to an embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] Hereinafter, embodiments of the determination method, the determination program, the substrate processing apparatus, and the information processing apparatus disclosed in the present application will be described in detail with reference to the drawings. Note that the disclosed determination method, the determination program, the substrate processing apparatus, and the information processing apparatus are not limited to the embodiments.

[0009] In the manufacture of semiconductor devices, a film forming apparatus forms a conductive film or an insulating film on a substrate such as a semiconductor wafer on which a pattern including recesses is formed. The film forming apparatus places the substrate in a chamber that is kept at a predetermined vacuum level, supplies a film forming source gas into the chamber, and forms a film on the substrate using reaction-assisting energy such as heat or plasma. Known film forming techniques include, for example, thermal CVD (Chemical Vapor Deposition), thermal ALD (Atomic Layer Deposition), PE-CVD (Plasma Enhancement CVD), and PE-ALD (Plasma Enhancement ALD).

[0010] However, with the advancement of miniaturization of semiconductor devices, the recesses of the patterns formed on the substrate are becoming finer and have higher aspect ratios. When a film formation process is performed to fill the recesses of such a substrate with a filling material, there is a risk of defective filling occurring, in which the recesses are filled with voids remaining in the recesses.

[0011] Therefore, a technology for detecting the occurrence of filling defects is desired.

[0012] [Embodiment] [Device configuration] Next, an embodiment will be described. First, an example of the substrate processing apparatus of the present disclosure will be described. In the following, the substrate processing apparatus of the present disclosure is a film forming apparatus 100, and a case where a film is formed as a substrate processing by the film forming apparatus 100 will be mainly described as an example. FIG. 1 is a schematic cross-sectional view showing an example of a schematic configuration of the film forming apparatus 100 according to the embodiment. In this embodiment, the film forming apparatus 100 corresponds to the substrate processing apparatus of the present disclosure. In one embodiment, the film forming apparatus 100 is an apparatus that forms a film on a substrate W. The film forming apparatus 100 shown in FIG. 1 has a chamber 1 that is airtight and electrically grounded. This chamber 1 is cylindrical and is made of, for example, aluminum or aluminum oxide having an anodized coating formed on the surface. A mounting table 2 is provided in the chamber 1.

[0013] The mounting table 2 is formed of a metal such as aluminum, nickel, aluminum oxide, or aluminum nitride, or a ceramic. A substrate W such as a semiconductor wafer is mounted on the upper surface of the mounting table 2. The substrate W has a pattern including a recess. The mounting table 2 horizontally supports the mounted substrate W. The lower surface of the mounting table 2 is electrically connected to a support member 4 formed of a conductive material. The mounting table 2 is supported by the support member 4. The support member 4 is supported by the bottom surface of the chamber 1. The lower end of the support member 4 is electrically connected to the bottom surface of the chamber 1 and is grounded via the chamber 1. The lower end of the support member 4 may be electrically connected to the bottom surface of the chamber 1 via a circuit adjusted to reduce the impedance between the mounting table 2 and the ground potential.

[0014] The mounting table 2 has a built-in heater 5, and the heater 5 can heat the substrate W mounted on the mounting table 2 to a predetermined temperature. The mounting table 2 may have a flow path (not shown) formed therein for circulating a coolant, and a coolant whose temperature is controlled by a chiller unit provided outside the chamber 1 may be circulated and supplied within the flow path. The mounting table 2 may control the substrate W to a predetermined temperature by heating with the heater 5 and cooling with the coolant supplied from the chiller unit. Note that the mounting table 2 may not be equipped with the heater 5, and may control the temperature of the substrate W only with the coolant supplied from the chiller unit.

[0015] An electrode may be embedded in the mounting table 2. A DC voltage is supplied to this electrode, which generates an electrostatic force that allows the mounting table 2 to adsorb the substrate W placed on its upper surface.

[0016] The mounting table 2 is provided with lifter pins 6 for raising and lowering the substrate W. FIG. 2 is a diagram showing a state in which the substrate W is raised from the mounting table 2 in the film forming apparatus 100 according to the embodiment. In the film forming apparatus 100, when the substrate W is transported or when spectroscopic measurement is performed on the substrate W, the lifter pins 6 are protruded from the mounting table 2, and the back surface of the substrate W is supported by the lifter pins 6 to raise the substrate W from the mounting table 2. For example, a side wall of the chamber 1 is provided with a loading / unloading port (not shown) for loading and unloading the substrate W. A gate valve for opening and closing the loading / unloading port is provided at the loading / unloading port. When loading and unloading the substrate W, the gate valve is opened. The substrate W is loaded into the chamber 1 from the loading / unloading port by a transport mechanism (not shown) in the transport chamber. The film forming apparatus 100 controls a lifting mechanism (not shown) provided outside the chamber 1 to lift the lifter pins 6 and receive the substrate W from the transport mechanism. After the transport mechanism has left the film forming apparatus 100, the film forming apparatus 100 controls the lifting mechanism to lower the lifter pins 6 and place the substrate W on the mounting table 2.

[0017] A showerhead 16 formed in a substantially disk shape is provided above the mounting table 2 on the inner surface of the chamber 1. The showerhead 16 is supported on the upper part of the mounting table 2 via an insulating member 45 such as ceramics. This electrically insulates the chamber 1 from the showerhead 16. The showerhead 16 is made of a conductive metal such as nickel.

[0018] The shower head 16 has a top plate member 16a and a shower plate 16b. The top plate member 16a is provided so as to close the inside of the chamber 1 from above. The shower plate 16b is provided below the top plate member 16a so as to face the mounting table 2. A gas diffusion space 16c is formed in the top plate member 16a. The top plate member 16a and the shower plate 16b have a large number of gas discharge holes 16d formed in a dispersed manner and opening toward the gas diffusion space 16c.

[0019] The top plate member 16a is formed with a gas inlet 16e for introducing various gases into the gas diffusion space 16c. The gas inlet 16e is connected to a gas supply path 15a. The gas supply path 15a is connected to a gas supply unit 15.

[0020] The gas supply unit 15 has gas supply lines connected to the gas supply sources of various gases used in film formation. Each gas supply line is appropriately branched according to the film formation process, and is provided with control devices for controlling the flow rate of the gas, such as valves such as an opening / closing valve and a flow rate controller such as a mass flow controller. The gas supply unit 15 can control the flow rate of various gases by controlling the control devices, such as the opening / closing valve and the flow rate controller, provided on each gas supply line.

[0021] The gas supply unit 15 supplies various gases used in film formation to the gas supply path 15a. For example, the gas supply unit 15 supplies a raw material gas for film formation to the gas supply path 15a. The gas supply unit 15 also supplies a purge gas and a reactive gas that reacts with the raw material gas to the gas supply path 15a. The gas supplied to the gas supply path 15a is diffused in the gas diffusion space 16c and discharged from each gas discharge hole 16d.

[0022] The space surrounded by the lower surface of the shower head 16 and the upper surface of the mounting table 2 forms a processing space where a film formation process is performed. The shower head 16 is paired with the mounting table 2 and functions as an electrode plate for forming a capacitively coupled plasma (CCP) in the processing space. A high-frequency power source 10 is connected to the shower head 16 via a matching device 11. When plasma is generated in the processing space, the high-frequency power source 10 supplies high-frequency power (RF power) in accordance with the supply of gas from the gas supply unit 15. As a result, the high-frequency power is applied to the gas supplied from the shower head 16 to the processing space 40, and plasma is generated in the processing space. The high-frequency power source 10 may be connected to the mounting table 2 instead of being connected to the shower head 16, and the shower head 16 may be grounded.

[0023] An exhaust port 71 is formed at the bottom of the chamber 1. An exhaust device 73 is connected to the exhaust port 71 via an exhaust pipe 72. The exhaust device 73 has a vacuum pump and a pressure adjustment valve. The exhaust device 73 can reduce and adjust the pressure inside the chamber 1 to a predetermined vacuum level by operating the vacuum pump and the pressure adjustment valve.

[0024] The film forming apparatus 100 performs spectroscopic measurement on the substrate W in the chamber 1, and is capable of detecting the state of the film formed on the substrate W. Spectroscopic measurement includes a method (transmission method) in which light is irradiated onto the substrate W and the light (transmitted light) transmitted through the substrate W is measured, and a method (reflection method) in which the light (reflected light) reflected from the substrate W is measured. In spectroscopic measurement using the transmission method, an absorbance spectrum is measured. In spectroscopic measurement using the reflection method, a reflectance spectrum is measured. The absorbance spectrum represents the amount of light for each wavelength absorbed by the film when the light passes through the film. The reflectance spectrum represents the phenomenon in which the reflectance changes when the light is reflected from the substrate W at the surface or the lower part of the film, due to the light being absorbed or attenuated in the film, as the amount of light reflected for each measured wavelength.

[0025] The film forming apparatus 100 shown in FIG. 1 shows an example of a configuration capable of performing spectroscopic measurement by a transmission method. The chamber 1 has windows 80a and 80b on side walls facing each other across the mounting table 2. The window 80a is provided at a high position on the side wall. The window 80b is provided at a low position on the side wall. The windows 80a and 80b are fitted with a member having optical transparency, such as quartz, and are sealed. An irradiation unit 81 that irradiates light is provided outside the window 80a. A detection unit 82 that can detect light is provided outside the window 80b. In this embodiment, the irradiation unit 81 and the detection unit 82 correspond to the measurement unit of the present disclosure.

[0026] 2, in the film forming apparatus 100, the lifter pins 6 are protruded from the mounting table 2, and the substrate W is raised from the mounting table 2. The positions of the window 80a and the irradiation unit 81 are adjusted so that the light irradiated from the irradiation unit 81 is irradiated onto the upper surface of the raised substrate W through the window 80a. The positions of the window 80b and the detection unit 82 are adjusted so that the transmitted light of the light that has passed through the raised substrate W is incident on the detection unit 82 through the window 80b.

[0027] In spectroscopic measurement, the measurement light used in the measurement is preferably light that is transparent to the substrate W. For example, when the substrate W is a silicon substrate, the measurement light is preferably infrared light that is transparent to the silicon substrate. Also, for example, when a SiO film or a SiN film is formed as a filling material by CVD, the measurement light may be light with a short wavelength equal to or shorter than infrared light, for example, light in the visible light region. In particular, when the depth of the recess formed in the substrate W is relatively shallow, the measurement light may be light with a short wavelength of about 0.1 um to 0.22 um, or visible light.

[0028] When detecting the state of a film formed on the substrate W, the irradiation unit 81 irradiates measurement light. The irradiation unit 81 is disposed so that the irradiated measurement light strikes a predetermined area near the center of the raised substrate W through a window 80a. The detection unit 82 is disposed so that transmitted light that has passed through a predetermined area of ​​the substrate W enters through a window 80b. The detection unit 82 detects the transmitted light that has passed through the substrate W.

[0029] The irradiation unit 81 incorporates a light source that emits measurement light, and optical elements such as mirrors and lenses, and is capable of irradiating interference measurement light. For example, the irradiation unit 81 splits the intermediate portion of the optical path of the measurement light generated by the light source until it is emitted to the outside into two optical paths using a half mirror or the like, and varies the optical path length of one path relative to the optical path length of the other path to change the optical path difference and cause interference, thereby irradiating measurement light of various interference waves with different optical path differences. Note that the irradiation unit 81 may be provided with multiple light sources, and the measurement light of each light source may be controlled by an optical element to be capable of irradiating measurement light of various interference waves with different optical path differences.

[0030] The film forming apparatus 100 detects filling defects of the film formed on the substrate W from the transmitted light detected by the detection unit 82. Specifically, the film forming apparatus 100 obtains an absorbance spectrum of the transmitted light detected by the detection unit 82. Then, the film forming apparatus 100 detects filling defects of the film formed on the substrate W from an integral value of absorbance for each wavelength range for a plurality of wavelength ranges of the absorbance spectrum.

[0031] 1 and 2, an example has been described in which the film formation apparatus 100 is configured to enable spectroscopic measurement by a transmission method. However, the film formation apparatus 100 may be configured to enable spectroscopic measurement by a reflection method. FIG. 3 is a schematic diagram showing another example of the film formation apparatus 100 according to the embodiment. The film formation apparatus 100 shown in FIG. 3 shows an example in which the film formation apparatus 100 is configured to measure reflected light reflected by a substrate W.

[0032] In the film forming apparatus 100 shown in FIG. 3, a window 80a and a window 80b are provided at positions facing each other on the side wall of the chamber 1 with the mounting table 2 interposed therebetween. An irradiation unit 81 that irradiates light is provided outside the window 80a. A detection unit 82 that can detect light is provided outside the window 80b. The positions of the window 80a and the irradiation unit 81 are adjusted so that the light irradiated from the irradiation unit 81 is irradiated onto the substrate W through the window 80a. The positions of the window 80b and the detection unit 82 are adjusted so that the light reflected by the substrate W is incident on the detection unit 82 through the window 80b. A loading / unloading port (not shown) for loading / unloading the substrate W is provided on a side wall of the chamber 1 different from the windows 80a and 80b. A gate valve for opening and closing the loading / unloading port is provided at the loading / unloading port.

[0033] When detecting filling defects, the irradiation unit 81 irradiates measurement light. The irradiation unit 81 is disposed so that the irradiated measurement light strikes a predetermined area near the center of the substrate W through a window 80a. The detection unit 82 is disposed so that the reflected light reflected from a predetermined area of ​​the substrate W enters through a window 80b. The detection unit 82 detects the reflected light reflected from the substrate W. In this manner, the film forming apparatus 100 shown in FIG. 3 is capable of spectroscopic measurement by the reflection method. In spectroscopic measurement by the reflection method, a reflectance spectrum indicating the reflectance for each wavelength is obtained from the reflected light.

[0034] The film forming apparatus 100 may be configured to change the incident angle and irradiation position of the measurement light incident on the substrate W from the irradiation unit 81. For example, in Fig. 1 and Fig. 3, the irradiation unit 81 is configured to be vertically movable and rotatable by a driving mechanism (not shown), so that the incident angle and irradiation position of the measurement light incident on the substrate W from the irradiation unit 81 can be changed.

[0035] The operation of the film forming apparatus 100 configured as above is generally controlled by a control device 110. The control device 110 controls each part of the film forming apparatus 100.

[0036] [Device configuration of control device 110] Next, an example of the configuration of the control device 110 according to the embodiment will be described. Fig. 4 is a diagram showing an example of a schematic configuration of the control device 110 according to the embodiment. The control device 110 is, for example, an information processing device such as a computer.

[0037] The control device 110 has an external I / F (interface) unit 111, a display unit 112, an input unit 113, a storage unit 114, and a control unit 115. Note that the control device 110 may have various functional units that a known computer has in addition to the functional units shown in FIG.

[0038] The external I / F unit 111 is an interface for inputting and outputting information to and from other devices. For example, the external I / F unit 111 is an interface for controlling communication with other devices. One aspect of the external I / F unit 111 is a network interface card such as a LAN card. For example, the external I / F unit 111 transmits and receives various data to and from the film forming apparatus 100 and other devices via a network. The external I / F unit 111 may be an interface such as a Universal Serial Bus (USB) port.

[0039] The display unit 112 is a display device that displays various types of information. Examples of the display unit 112 include a display device such as an LCD (Liquid Crystal Display) or a CRT (Cathode Ray Tube). The display unit 112 displays various types of information.

[0040] The input unit 113 is an input device for inputting various information. For example, the input unit 113 may be an input device such as a mouse or a keyboard. The input unit 113 receives an operation input from an administrator or the like, and inputs operation information indicating the received operation content to the control unit 115.

[0041] The storage unit 114 is a storage device that stores various data. For example, the storage unit 114 is a storage device such as a hard disk, a solid state drive (SSD), an optical disk, etc. Note that the storage unit 114 may be a semiconductor memory in which data can be rewritten, such as a random access memory (RAM), a flash memory, or a non-volatile static random access memory (NVSRAM).

[0042] The storage unit 114 stores an OS (Operating System) and various programs executed by the control unit 115. For example, the storage unit 114 stores a program for executing a generation process and a film forming process, which will be described later. Furthermore, the storage unit 114 stores various data used in the programs executed by the control unit 115. For example, the storage unit 114 stores processing condition data 120, learning data 121, and model data 122. The storage unit 114 can also store other data in addition to the data exemplified above. Furthermore, the various programs and data may be stored in a computer-readable computer recording medium (for example, a hard disk, an optical disk such as a DVD, a flexible disk, a semiconductor memory, etc.). Furthermore, the various programs and data can be transmitted from other devices at any time and used online.

[0043] The processing condition data 120 is data that stores processing conditions for a substrate processing performed on the substrate W. In this embodiment, the processing condition data 120 stores processing conditions for a film formation processing performed on the substrate W. The processing conditions include the type and flow rate of a processing gas used during film formation, and the power of high frequency power supplied from the high frequency power supply 10. The learning data 121 is learning data used to generate a prediction model. Details of the learning data 121 will be described later. The model data 122 is data for a prediction model that predicts the embedding state.

[0044] The control unit 115 is a device that controls the control device 110. As the control unit 115, electronic circuits such as a central processing unit (CPU) and a micro processing unit (MPU), and integrated circuits such as an application specific integrated circuit (ASIC) and a field programmable gate array (FPGA) can be adopted. The control unit 115 has an internal memory for storing programs and data. The control unit 115 reads out various programs stored in the storage unit 114 and executes the processing of the read programs. The control unit 115 functions as various processing units by the operation of the programs. For example, the control unit 115 has a film formation control unit 130, an acquisition unit 131, a generation unit 132, a measurement control unit 133, and a determination unit 134. In this embodiment, a case where the control unit 115 has the film formation control unit 130, the acquisition unit 131, the generation unit 132, the measurement control unit 133, and the determination unit 134 will be described as an example. However, the functions of the film formation control unit 130, the acquisition unit 131, the generation unit 132, the measurement control unit 133, and the determination unit 134 may be realized by distributing them among a plurality of control units.

[0045] The film formation control unit 130 controls each part of the film formation apparatus 100 and controls the film formation process.

[0046] The acquisition unit 131 acquires various types of information. For example, the acquisition unit 131 acquires various types of information by displaying an input screen on the display unit 112 and receiving input of various types of information from the input unit 113. The acquisition unit 131 also acquires various types of information from other devices via a network (not shown). For example, the acquisition unit 131 acquires the state of the base of the substrate W, the filling state of the recesses 91 of each substrate W, and the film thickness of the film formed on the substrate W.

[0047] Here, a flow of the film forming apparatus 100 according to the embodiment performing a film forming process as a substrate process on the substrate W under the control of the film forming control unit 130 will be briefly described. When performing a film forming process in the film forming apparatus 100, the substrate W is placed on the mounting table 2 by a transport mechanism such as a transport arm (not shown). A pattern including a recess is formed on the substrate W. When performing a film forming process on the substrate W, the film forming control unit 130 controls the exhaust device 73 to reduce the pressure inside the chamber 1 by the exhaust device 73. The film forming control unit 130 controls the gas supply unit 15 to supply various gases used in film formation from the gas supply unit 15 and introduce a process gas into the chamber 1 from the shower head 16. Then, the film forming control unit 130 controls the high frequency power source 10 to supply high frequency power from the high frequency power source 10 to generate plasma in the processing space, and performs film forming on the substrate W.

[0048] Incidentally, as semiconductor devices become finer, the recesses of the patterns formed on the substrate W become finer and have higher aspect ratios. For example, in the manufacturing process of semiconductors for VLSI (very large scale integration), miniaturization has already progressed to the nm (nanometer) range, and market demands for even higher integration are driving not only miniaturization but also 3D. When a film formation process is performed to fill recesses in such a substrate W with a filling material, there is a risk of filling defects occurring, in which the recesses are filled with voids formed in the recesses. Such voids are called voids or seams. In the following, the voids formed in the recesses are also called voids.

[0049] FIG. 5 is a diagram for explaining an example of a substrate W before a film is formed according to the embodiment. FIG. 5 shows a schematic cross section of the substrate W. The substrate W is made of, for example, silicon (Si). A pattern 90 is formed on the substrate W in correspondence with an area that will become a chip of a semiconductor device. The pattern 90 includes recesses 91 of various shapes and depths. FIG. 6 is a diagram showing an example of the substrate W after a film is formed according to the embodiment. FIG. 6 shows a schematic state in which a film 92 is formed on a pattern 90 having a recess 91. As a result of the film formation, some recesses 91a of the substrate W are not filled with the film 92, and voids 93, which are voids, are formed in the recesses 91a. In FIG. 6, the recesses 91 in which the voids 93 are generated are shown as "NG potion".

[0050] Therefore, the film formation apparatus 100 according to the embodiment performs spectroscopic measurement on the substrate W on which a film has been formed, and measures one of the absorbance spectrum or reflectance spectrum of the substrate W. For example, when the film formation apparatus 100 is configured to be capable of spectroscopic measurement by the transmission method as shown in Figs. 1 and 2, the film formation apparatus 100 measures the absorbance spectrum of the substrate W. When the film formation apparatus 100 is configured to be capable of spectroscopic measurement by the reflection method as shown in Fig. 3, the film formation apparatus 100 measures the reflectance spectrum of the substrate W. The film formation apparatus 100 determines the filling state of the recesses 91 of the substrate W from the results of the spectroscopic measurement, using a prediction model that predicts the filling state of the recesses 91 of the substrate W.

[0051] The film forming apparatus 100 according to the embodiment generates a prediction model by machine learning. For example, the storage unit 114 stores learning data 121 used to generate the prediction model. The learning data 121 stores data that associates the integral value of the spectrum value for each wavelength range of one of the absorbance spectrum or the reflectance spectrum of the substrates W in which the filling material is embedded with the filling state of the recesses 91. For example, when the film forming apparatus 100 performs spectroscopic measurement using a transmission method, the learning data 121 stores data that associates the integral value of the absorbance for each wavelength range of the absorbance spectrum of the substrates W in which the filling material is embedded with the filling state of the recesses 91. Also, for example, when the film forming apparatus 100 performs spectroscopic measurement using a reflection method, the learning data 121 stores data that associates the integral value of the reflectance for each wavelength range of the reflectance spectrum of the substrates W in which the filling material is embedded with the filling state of the recesses 91. The learning data 121 may be generated by the film forming apparatus 100, or may be received from another apparatus. For example, the film forming apparatus 100 performs a film forming process on a plurality of substrates W in advance, and embeds the embedding material in the plurality of substrates W. The film forming apparatus 100 also performs spectroscopic measurement on the plurality of substrates W embedded with the embedding material, and calculates one of an absorbance spectrum showing the absorbance for each wavelength of each of the plurality of substrates W, or a reflectance spectrum showing the reflectance for each wavelength. For example, when the film forming apparatus 100 performs spectroscopic measurement by a transmission method, the film apparatus 100 performs spectroscopic measurement on the plurality of substrates W embedded with the embedding material, and calculates an absorbance spectrum. For example, when the film forming apparatus 100 performs spectroscopic measurement by a reflection method, the film apparatus 100 performs spectroscopic measurement on the plurality of substrates W embedded with the embedding material, and calculates a reflectance spectrum.

[0052] Furthermore, the film forming apparatus 100 acquires the filling state of the recesses 91 of each of the plurality of substrates W. For example, an inspector such as a user in charge of inspecting the filling state inspects the plurality of substrates W filled with a filling material by the film forming apparatus 100, and determines the filling state of the recesses 91. For example, the inspector checks an SEM image of a cross section of the substrate W, and indexes the filling state of the recesses 91.

[0053] FIG. 7 is a diagram for explaining an example of an index of the filling state of the recess 91 according to the embodiment. FIG. 7 shows a state in which the recess 91 of the substrate W is filled by forming a film 92 of a filling material. In FIG. 7, the filling state is classified into six levels from level 0 to 5, with a good filling state in which the recess 91 is filled without any gaps being level 0 and a bad filling state in which a large void 93 is generated in the recess 91 being level 5. For example, an inspector checks an SEM image of a cross section of the substrate W and indexes the filling state of the recess 91 of the substrate W into six filling index values ​​from 0 to 5 with reference to FIG. 7. The inspector may index the filling state of the recess 91 into fewer filling index values ​​(2 to 5 levels). The inspector may also index the filling state of the recess 91 into more filling index values ​​than six levels. The indexing of the filling state of the recess 91 may be realized by software. For example, image recognition may be performed on an SEM image of the cross section of the substrate W to identify the area of ​​the void 93 contained in the recess 91, and the filling state of the recess 91 may be indexed according to the shape and size of the identified area of ​​the void 93.

[0054] The acquisition unit 131 acquires the embedding state of the recesses 91 of each of the multiple substrates W. For example, the acquisition unit 131 displays an input screen on the display unit 112 and receives an input of an embedding index value for each of the multiple substrates W from the input unit 113, thereby acquiring the embedding state of the recesses 91 of each of the multiple substrates W. The acquisition unit 131 associates the integral value of the spectrum value for each wavelength range of one of the absorbance spectrum or the reflectance spectrum of the multiple substrates W in which the embedding material has been embedded with the embedding state of the recesses 91, and stores the associated data in the storage unit 114 as learning data 121. For example, when the film forming apparatus 100 performs spectroscopic measurement by the transmission method, the acquisition unit 131 associates the integral value of the absorbance for each wavelength range of the absorbance spectrum with the embedding index value for each of the multiple substrates W, and stores the associated data in the storage unit 114 as learning data 121. Furthermore, for example, when the film forming apparatus 100 performs spectroscopic measurement using a reflectance method, the acquisition unit 131 associates, for each of a plurality of substrates W, the integral value of the reflectance for each wavelength range of the reflectance spectrum with the embedded index value, and stores the associated data in the memory unit 114 as learning data 121.

[0055] The generation unit 132 generates a prediction model by performing machine learning on the learning data 121. The generation unit 132 stores data of the generated prediction model in the storage unit 114 as model data 122.

[0056] The film forming apparatus 100 according to the embodiment performs spectroscopic measurement on the substrate W on which a film has been formed, and determines the filling state of the recesses 91 of the substrate W from the results of the spectroscopic measurement using a prediction model.

[0057] For example, the measurement control unit 133 controls spectroscopic measurement of the substrate W. For example, the measurement control unit 133 controls the irradiation unit 81 and the detection unit 82, irradiates measurement light from the irradiation unit 81, and detects light transmitted through or reflected by the substrate W with the detection unit 82. The detection unit 82 outputs data on the signal intensity of the detected light to the control device 110. The measurement control unit 133 performs spectroscopic analysis such as Fourier transform on the signal intensity data input from the detection unit 82 to calculate one of the absorbance spectrum or the reflectance spectrum. For example, when the film forming apparatus 100 performs spectroscopic measurement by a transmission method, the measurement control unit 133 calculates the absorbance spectrum from the signal intensity data input from the detection unit 82. Also, for example, when the film forming apparatus 100 performs spectroscopic measurement by a reflection method, the measurement control unit 133 calculates the reflectance spectrum from the signal intensity data input from the detection unit 82.

[0058] The determination unit 134 predicts the state of the film formed on the substrate W from the calculated one spectrum using the prediction model. The determination unit 134 obtains a spectral value for each wavelength range of the one spectrum. For example, when the film forming apparatus 100 performs spectroscopic measurement using a transmission method, the determination unit 134 obtains an integral value of absorbance for each wavelength range of the absorbance spectrum of the substrate W. Also, for example, when the film forming apparatus 100 performs spectroscopic measurement using a reflection method, the determination unit 134 obtains an integral value of reflectance for each wavelength range of the reflectance spectrum of the substrate W. The determination unit 134 predicts the filling state of the recess 91 of the substrate W from the integral value of the spectral value for each wavelength range using the prediction model. For example, when the film forming apparatus 100 performs spectroscopic measurement using a transmission method, the determination unit 134 predicts the filling state of the recess 91 of the substrate W from the integral value of the absorbance for each wavelength range using the prediction model. Furthermore, for example, when the film forming apparatus 100 performs spectroscopic measurement by a reflectance method, the determination unit 134 predicts the filling state of the recesses 91 of the substrate W from the integral value of the reflectance for each wavelength range using a prediction model. Then, the determination unit 134 determines the filling state of the recesses 91 of the substrate W from the prediction result.

[0059] The learning data 121 may further be associated with the film thickness of the embedding material and the state of the base of the substrate W on which the embedding material is formed. For example, the inspector inspects a plurality of substrates W to determine the state of the base on which the embedding material is formed. The state of the base may be determined on the substrate W before the embedding material is embedded, or on the substrate W after the embedding material is embedded. For example, the inspector checks an SEM image of a cross section of the substrate W to index the state of the base on which the embedding material is formed. In addition, the inspector inspects a plurality of substrates W on which the embedding material is embedded by the film forming apparatus 100 to measure the film thickness of the embedding material. The acquiring unit 131 acquires the film thickness of the embedding material and the state of the base for each of the plurality of substrates W by receiving the input of the film thickness of the embedding material and the state of the base for each of the plurality of substrates W through the input unit 113. The acquiring unit 131 associates the integral value of the spectrum value for each wavelength range of the plurality of substrates W on which the embedding material is embedded, the embedding state of the recess 91, the film thickness of the embedding material, and the state of the base, and stores them in the storage unit 114 as the learning data 121. The generation unit 132 may generate a prediction model by machine learning the learning data 121 including the film thickness and the state of the base of the substrate W. The determination unit 134 may use the prediction model to predict the state and film thickness of the film formed on the substrate W from one of the spectra.

[0060] Here, a specific example will be described.

[0061] First, a specific example of indexing the state of the base of the substrate W will be described. FIGS. 8A and 8B are diagrams for explaining an example of an index of the state of the base of the substrate W according to the embodiment. FIGS. 8A and 8B show a recess 91 of the substrate W in which a filling material is filled under the same film forming conditions. A plurality of protruding core parts 95 are formed as a pattern 90 on the substrate W. The recess 91 is formed between the core parts 95. In FIGS. 8A and 8B, the cross-sectional shape of one core part 95 is shown by a broken line 95a. In FIG. 8A, the core part 95 is formed with the same width in the vertical direction of the substrate W. The side wall 96 of the core part 95 is perpendicular to the vertical direction of the substrate W. In FIG. 8B, the core part 95 is formed in a tapered shape with a narrower width toward the upper side in the vertical direction of the substrate W. The side wall 96 of the core part 95 is inclined with respect to the vertical direction of the substrate W. The filling state of the recess 91 varies depending on the state of the sidewall 96 constituting the recess 91 of the substrate W. For example, as shown in FIG. 8A, when the cross-sectional shape of the core part 95 is rectangular, the filling material does not easily reach a deep position of the recess 91 during film formation, and therefore voids 93 tend to occur easily. On the other hand, as shown in FIG. 8B, when the cross-sectional shape of the core part 95 is tapered, the filling material easily reaches a deep position of the recess 91 during film formation, and therefore voids 93 are less likely to occur than in FIG. 8A, and even if voids 93 occur, they tend to be small. Therefore, an inspection is performed on the substrate W to determine the state of the substrate W. For example, an inspector checks an SEM image of the cross section of the substrate W on which a film is to be formed, and indexes the state of the substrate W. For example, the inspector indexes the state of the substrate W into two stages, by setting the state of the substrate to 0 when the sidewall 96 is perpendicular to the vertical direction of the substrate W, and setting the state of the substrate to 1 when the sidewall 96 is considered to be oblique to the vertical direction of the substrate W. The inspector may index the condition of the base of the substrate W into a multi-stage base index value according to the angle of the sidewall 96 with respect to the vertical direction of the substrate W. The indexing of the base condition may also be realized by software. For example, an image recognition may be performed on an SEM image of a cross section of the substrate W to identify the base region, and the base condition may be indexed according to the shape of the identified base region.

[0062] 9 is a diagram illustrating an example of data stored in the learning data 121 according to the embodiment. In FIG. 9, as an example of a plurality of substrates W, embedment index values, film thicknesses, and base index values ​​for substrates 1 to 5 are shown.

[0063] FIG. 10 is a diagram for explaining an example of data stored in the learning data 121 according to the embodiment. In FIG. 10, a case where the film forming apparatus 100 is configured to be capable of spectroscopic measurement by the reflection method and the reflectance spectrum is measured by spectroscopic measurement will be explained as an example. In FIG. 10, waveforms of the reflectance spectra SP1 to SP3 of three substrates W are shown in outline. In the determination method according to the embodiment, for each substrate W, an integral value of the reflectance for each wavelength range of the reflectance spectrum of the substrate W is obtained. For example, in the determination method according to the embodiment, a predetermined wavelength is set as a reference wavelength, and an integral value of the reflectance for each wavelength range before and after the reference wavelength is obtained. For example, the reference wavelength is set to a wavelength at which the change in absorbance is large when a filling defect such as a void 93 occurs. For example, the reference wavelength is set to a wavelength at which the change in absorbance is the largest. The width of the multiple wavelength ranges is preferably set to any of 75 to 200 nm. The multiple wavelength ranges are set to 1 to 5 so as to be continuous before and after the reference wavelength. In Fig. 10, the reference wavelength is set to 480 nm, the width of the wavelength range is set to 80 nm, three wavelength ranges WE1 to WE6 are defined before and after the reference wavelength, and the integral value of the reflectance is calculated for each of the wavelength ranges WE1 to WE6. For example, in Fig. 10, the integral value of the reflectance for each of the wavelength ranges WE1 to WE6 for the reflectance spectra SP1 to SP3 is calculated. Note that the multiple wavelength ranges may be defined as ranges obtained by dividing the wavelength range where the change in absorbance due to filling defects such as voids 93 is large into ranges of a certain wavelength width.

[0064] Returning to FIG. 4, the learning data 121 stores the integral value of the spectrum value for each wavelength range of one of the absorbance spectrum or the reflectance spectrum of the substrates W in association with the filling state of the recesses 91. For example, when the film forming apparatus 100 performs spectroscopic measurement using a transmission method, the learning data 121 stores the integral value of the absorbance for each wavelength range of the absorbance spectrum of the substrates W in association with the filling state of the recesses 91. Also, for example, when the film forming apparatus 100 performs spectroscopic measurement using a reflection method, the learning data 121 stores the integral value of the reflectance for each wavelength range of the reflectance spectrum of the substrates W in association with the filling state of the recesses 91. For example, the learning data 121 stores the integral value of the reflectance for each wavelength range WE1 to WE6 of the reflectance spectrum as shown in FIG. 10 in association with the filling index value for each substrate W.

[0065] The generating unit 132 performs machine learning on the learning data 121 to generate a prediction model. For example, the generating unit 132 performs machine learning on the learning data 121 using a linear regression algorithm to generate a prediction model that has learned the relationship between the data of the learning data 121. For example, the generating unit 132 generates a prediction model that has learned the relationship between the integral value of the spectrum value for each wavelength range and the embedded index value as linear regression. For example, when the learning data 121 is data in which the integral value of the absorbance for each wavelength range is associated with the embedded index value, the generating unit 132 generates a prediction model that has learned the relationship between the integral value of the absorbance and the embedded index value as linear regression. Also, for example, when the learning data 121 is data in which the integral value of the reflectance for each wavelength range is associated with the embedded index value, the generating unit 132 generates a prediction model that has learned the relationship between the integral value of the reflectance and the embedded index value as linear regression.

[0066] In this embodiment, a linear regression algorithm is used as machine learning to generate a prediction model so as to prevent overfitting in the learning data 121. However, it is also possible to use a more advanced machine learning algorithm such as a neural network or a more advanced regression analysis than linear regression.

[0067] The determination unit 134 uses the prediction model thus generated to determine the embedding state of the recess 91 of the substrate W. The determination unit 134 derives the embedding index value from the prediction model by using the integral value of the spectrum value for each wavelength range as an explanatory variable and the embedding index value as an objective variable. For example, when the prediction model is generated by machine learning the relationship between the integral value of the absorbance and the embedding index value, the determination unit 134 derives the embedding index value from the prediction model by using the integral value of the absorbance for each wavelength range as an explanatory variable and the embedding index value as an objective variable. Also, for example, when the prediction model is generated by machine learning the relationship between the integral value of the reflectance and the embedding index value, the determination unit 134 derives the embedding index value from the prediction model by using the integral value of the reflectance for each wavelength range as an explanatory variable and the embedding index value as an objective variable. The determination unit 134 determines the embedding state of the recess 91 based on the derived embedding index value. For example, when the derived embedding index value is greater than a predetermined threshold value, the determination unit 134 determines that the embedding is defective.

[0068] Next, an example of the overall flow of the determination method according to the embodiment will be described. In the following, a case where the film forming apparatus 100 is configured to be capable of spectroscopic measurement by the reflection method and a reflectance spectrum is measured by spectroscopic measurement will be described as an example. FIG. 11A is a diagram that shows an outline of an example of the overall flow of the determination method according to the embodiment. FIG. 11A shows a case where a filling index value is predicted from an integral value of reflectance using a prediction model. The determination method according to the present embodiment is divided into a generation process of generating a prediction model and a determination process of determining a filling state of the recess 91 of the substrate W on which a film is actually formed using the prediction model.

[0069] The generating step is performed at least once before the determining step in order to generate a prediction model. In the generating step, learning data 121 is prepared. The learning data 121 is stored for each of a plurality of substrates W, in which the integral value of the reflectance for each wavelength range of the reflectance spectrum is associated with the embedded index value. In the generating step, the generating unit 132 performs machine learning on such learning data 121 to generate a prediction model.

[0070] In the determination step, a spectroscopic measurement is performed on the substrate W on which a film is formed in the film forming apparatus 100 to calculate the reflectance spectrum of the substrate W, and an integral value of the reflectance for each wavelength range of the reflectance spectrum of the substrate W is obtained. Then, in the determination step, a prediction model is used to predict the embedding state of the substrate W from the integral value of the reflectance for each wavelength range of the reflectance spectrum. For example, in the determination step, the integral value of the reflectance for each wavelength range is used as an explanatory variable, and the embedding index value is used as an objective variable to derive an embedding index value from the prediction model. The prediction model is generated by machine learning the learning data 121 in which the integral value of the reflectance and the embedding index value are associated with each other. Therefore, in the determination step, the embedding index value can be predicted from the integral value of the reflectance using the prediction model. In the determination step, the embedding state of the recess 91 is determined based on the derived embedding index value. For example, when the derived embedding index value is greater than a predetermined threshold value, the determination unit 134 determines that the embedding is defective.

[0071] The learning data 121 may further be associated with the film thickness of the filling material and the state of the underlying surface of the substrate W. The prediction model may further be a model that learns the relationship between the film thickness and the state of the underlying surface of the substrate W.

[0072] Fig. 11B is a diagram showing another example of the overall flow of the determination method according to the embodiment. Fig. 11B shows a case where a prediction model is used to predict a filling index value from an integral value of reflectance and a film thickness of a filling material. Fig. 11B is partially similar to Fig. 11A, so the following mainly describes the differences.

[0073] 11B, the learning data 121 stores, for each substrate W, an integral value of the reflectance for each wavelength range of the reflectance spectrum, an embedding index value, and a film thickness of the embedding material in association with each other, for a plurality of substrates W. In the generation step, the generation unit 132 performs machine learning on such learning data 121 to generate a prediction model.

[0074] In the determination step, the substrate W on which the film is formed in the film forming apparatus 100 is subjected to spectroscopic measurement to calculate the reflectance spectrum of the substrate W, and the integral value of the reflectance for each wavelength range of the reflectance spectrum of the substrate W is obtained. The film forming apparatus 100 also obtains the film thickness of the filling material. The film thickness may be obtained by measuring in the film forming apparatus 100, or may be obtained as a result of measurement in another apparatus. In the determination step, the embedding state of the substrate W is predicted from the integral value of the reflectance for each wavelength range of the reflectance spectrum and the film thickness of the filling material using a prediction model. For example, in the determination step, the integral value of the reflectance for each wavelength range and the film thickness of the filling material are used as explanatory variables, and the embedding index value is used as a target variable, and an embedding index value is derived from the prediction model. The prediction model is generated by machine learning learning data 121 in which the integral value of the reflectance, the embedding index value, and the film thickness of the filling material are associated with each other. Therefore, in the determination step, the embedding index value can be predicted from the integral value of the reflectance and the film thickness of the filling material using the prediction model. In the determination step, the embedding state of the recess 91 is determined based on the derived embedding index value.

[0075] Fig. 11C is a diagram showing another example of the overall flow of the determination method according to the embodiment. Fig. 11C shows a case where the embedding index value and the thickness of the embedding material are predicted from the integral value of the reflectance using a prediction model. Fig. 11C is partially similar to Fig. 11A and Fig. 11B, so the following mainly describes the differences.

[0076] 11C, as in the case of Fig. 11B, the learning data 121 stores, for each of a plurality of substrates W, an integral value of the reflectance for each wavelength range of the reflectance spectrum, an embedding index value, and a film thickness of the embedding material in association with each other for each substrate W. In the generation step, the generation unit 132 performs machine learning on such learning data 121 to generate a prediction model.

[0077] In the determination step, the substrate W on which the film is formed in the film forming apparatus 100 is spectroscopically measured to calculate the reflectance spectrum of the substrate W, and the integral value of the reflectance for each wavelength range of the reflectance spectrum of the substrate W is obtained. In the determination step, the embedding state of the substrate W and the film thickness of the embedding material are predicted from the integral value of the reflectance for each wavelength range of the reflectance spectrum using a prediction model. For example, in the determination step, the embedding index value and the film thickness of the embedding material are derived from the prediction model using the integral value of the reflectance for each wavelength range as an explanatory variable and the embedding index value and the film thickness of the embedding material as objective variables. The prediction model is generated by machine learning the learning data 121 in which the integral value of the reflectance, the embedding index value, and the film thickness of the embedding material are associated with each other. Therefore, in the determination step, the embedding index value and the film thickness of the embedding material can be predicted from the integral value of the reflectance using the prediction model. In the determination step, the embedding state of the recess 91 is determined based on the derived embedding index value. In the determination step, it may be further determined whether the film thickness of the embedding material is within an appropriate range. For example, when the thickness of the derived filling material is outside a predetermined range, the determining unit 134 determines that the thickness is defective.

[0078] Fig. 11D is a diagram showing another example of the overall flow of the determination method according to the embodiment. Fig. 11D shows a case where a prediction model is used to predict an embedding index value from an integral value of reflectance and a base index value. Fig. 11D is partially similar to Fig. 11A, so the following mainly describes the differences.

[0079] 11D, the learning data 121 stores, for each substrate W, an integral value of the reflectance for each wavelength range of the reflectance spectrum, an embedded index value, and a base index value in association with each other, for a plurality of substrates W. In the generation step, the generation unit 132 performs machine learning on such learning data 121 to generate a prediction model.

[0080] In the determination step, a spectroscopic measurement is performed on the substrate W on which a film is formed in the film forming apparatus 100 to calculate the reflectance spectrum of the substrate W, and an integral value of the reflectance for each wavelength range of the reflectance spectrum of the substrate W is obtained. In the determination step, a prediction model is used to predict the embedding state of the substrate W from the integral value of the reflectance for each wavelength range of the reflectance spectrum and the base index value. For example, in the determination step, the embedding index value is derived from the prediction model using the integral value of the reflectance for each wavelength range and the base index value as explanatory variables and the embedding index value as a target variable. The prediction model is generated by machine learning the learning data 121 in which the integral value of the reflectance, the embedding index value, and the base index value are associated with each other. Therefore, in the determination step, the embedding index value can be predicted from the integral value of the reflectance and the base index value using the prediction model. In the determination step, the embedding state of the recess 91 is determined based on the derived embedding index value.

[0081] Fig. 11E is a diagram showing another example of the overall flow of the determination method according to the embodiment. Fig. 11E shows a case where a prediction model is used to predict a filling index value and a film thickness of a filling material from an integral value of reflectance and a base index value. Fig. 11E is partially similar to Figs. 11A to 11D, and therefore the following mainly describes the differences.

[0082] 11E, the learning data 121 stores, for each substrate W, an integral value of the reflectance for each wavelength range of the reflectance spectrum, an embedding index value, a film thickness of the embedding material, and a base index value in association with each other, for a plurality of substrates W. In the generation step, the generation unit 132 performs machine learning on such learning data 121 to generate a prediction model.

[0083] In the determination step, the substrate W on which a film is formed in the film forming apparatus 100 is spectroscopically measured to calculate the reflectance spectrum of the substrate W, and the integral value of the reflectance for each wavelength range of the reflectance spectrum of the substrate W is obtained. In the determination step, the embedding state of the substrate W and the film thickness of the embedding material are predicted from the integral value of the reflectance for each wavelength range of the reflectance spectrum and the base index value using a prediction model. For example, in the determination step, the embedding index value and the film thickness of the embedding material are derived from the prediction model using the integral value of the reflectance for each wavelength range and the base index value as explanatory variables, and the embedding index value and the film thickness of the embedding material as objective variables. The prediction model is generated by machine learning the learning data 121 in which the integral value of the reflectance, the embedding index value, the film thickness of the embedding material, and the base index value are associated with each other. Therefore, in the determination step, the embedding index value and the film thickness of the embedding material can be predicted from the integral value of the reflectance and the base index value using the prediction model. In the determination step, the embedding state of the recess 91 is determined based on the derived embedding index value. In the determination step, it may be further determined whether the film thickness of the embedding material is within an appropriate range.

[0084] 11A to 11E are an example of a case where the film forming apparatus 100 is configured to be capable of spectroscopic measurement by the reflection method and the reflectance spectrum is measured by spectroscopic measurement. Even when the film forming apparatus 100 is configured to be capable of spectroscopic measurement by the transmission method and the absorbance spectrum is measured by spectroscopic measurement, the determination method according to the embodiment can be performed in the same manner as in FIGS. 11A to 11E.

[0085] Next, the reason why the determination method according to the embodiment uses the integral value of the spectral value for each wavelength range, rather than the spectral value of the spectrum, will be described. The following will take the reflectance spectrum as an example. FIG. 12 is a diagram showing an example of the S / N ratio of the reflectance spectrum. FIG. 12(B) shows the S / N ratio of the signal component (S) and the noise component (N) included in the reflectance of each wavelength of the reflectance spectrum. FIG. 12(A) shows an example of the S / N ratio of the signal component (S) and the noise component (N) included in the integral value of the reflectance for each wavelength range of the reflectance spectrum.

[0086] The reflectance spectrum measured by spectroscopic measurement contains a noise component such as white noise, along with a valid signal component corresponding to the film state. Therefore, as shown in Figure 12(B), the reflectance spectrum fluctuates due to the noise component with the signal component as the reference at each wavelength. The noise component fluctuates randomly within a certain numerical range, and the average and variance of the entire data are zero.

[0087] FIG. 12(A) shows the S / N ratio of the integral value of the reflectance for each wavelength range of 100 nm for wavelengths of 200 to 900 nm of the reflectance spectrum. As described above, the noise component randomly varies within a certain numerical range. Therefore, when the reflectance of each wavelength of the reflectance spectrum is integrated, the S / N ratio of the integral value of the reflectance is small. This improves the detection sensitivity. For example, in FIG. 12(A), the detection sensitivity is improved by 10 db compared to FIG. 12(B). In FIG. 12, the reflectance spectrum is described as an example, but the S / N ratio of the integral value of the absorbance is also small for the absorbance spectrum. Therefore, in the determination method according to the embodiment, the filling state of the recess 91 is determined using the integral value of the spectrum value of one of the absorbance spectrum or the reflectance spectrum.

[0088] Next, a process flow of the determination method according to the embodiment will be described. As described above, the determination method according to the embodiment is divided into a generation process and a determination process. In the following, a case where the learning data 121 is generated in the film forming apparatus 100 and a prediction model is generated will be described.

[0089] FIG. 13 is a flowchart showing an example of the flow of a generation process including a generation step of the determination method according to the embodiment.

[0090] The substrate W on which a pattern including recesses is formed is carried into the film forming apparatus 100 by a transport mechanism such as a transport arm (not shown) and placed on the mounting table 2 (step S10). The film forming control unit 130 controls each part of the film forming apparatus 100 to perform a film forming process (step S11).

[0091] The measurement control unit 133 controls each unit of the film forming apparatus 100 to perform spectroscopic measurement on the substrate W (step S12). For example, the measurement control unit 133 controls the irradiation unit 81 and the detection unit 82 to irradiate measurement light from the irradiation unit 81 and detect light transmitted through or reflected by the substrate W with the detection unit 82. The detection unit 82 outputs data on the signal intensity of the detected light to the control device 110. The measurement control unit 133 performs spectroscopic analysis such as Fourier transform on the signal intensity data input from the detection unit 82 to calculate one of the absorbance spectrum or the reflectance spectrum. The acquisition unit 131 obtains an integral value of the spectral value for each wavelength range of the one spectrum, and stores the integral value of the spectral value for each wavelength range of the one spectrum in the learning data 121 for each substrate W. In this embodiment, step S12 corresponds to the process c) of the present disclosure. When the film forming apparatus 100 performs spectroscopic measurement by the transmission method, the learning data 121 stores the integral value of the absorbance as the spectral value. Furthermore, when the film forming apparatus 100 performs spectroscopic measurement using a reflectance method, the learning data 121 stores an integral value of the reflectance as a spectral value.

[0092] The substrate W placed on the mounting table 2 is carried out from the film forming apparatus 100 by a transport mechanism (not shown) (step S13).

[0093] The generation unit 132 determines whether or not film formation has been performed on a predetermined number of substrates W or more (step S14). The predetermined number is determined in advance according to the number of substrates W required for the learning data 121. If film formation has not been performed on the predetermined number of substrates W or more (step S14: No), the process proceeds to step S10 described above.

[0094] On the other hand, when film formation has been performed on a predetermined number of substrates W or more (step S14: Yes), the acquisition unit 131 acquires the embedding state of the recesses 91 of each of the substrates W on which film formation has been performed (step S15). For example, the acquisition unit 131 displays an input screen on the display unit 112 and receives an input of an embedding index value for each of the substrates W from the input unit 113. The acquisition unit 131 stores the input embedding index value in the learning data 121 by associating it with an integral value of the spectral value for each wavelength range for each substrate W. The acquisition unit 131 may further acquire a film thickness and a base index value. For example, the acquisition unit 131 may display an input screen on the display unit 112 and receive an input of a film thickness and a base index value for each of the substrates W from the input unit 113. The acquisition unit 131 may store the integrated value of the spectral value for each wavelength range, the embedding index value, the film thickness, and the base index value in the learning data 121 by associating them with each other for each substrate W. The acquiring unit 131 may also acquire the embedding index value, the film thickness, and the base index value via a network (not shown) from another device that stores the embedding index value, the film thickness, and the base index value for each of the multiple substrates W. In this embodiment, step S15 corresponds to steps d), f), and g) of the present disclosure.

[0095] The generation unit 132 performs machine learning on the training data 121 to generate a prediction model (step S16). For example, the generation unit 132 performs machine learning on the training data 121 using a linear regression algorithm to generate a prediction model that has learned the relationships between data in the training data 121. In this embodiment, step S16 corresponds to step e) of the present disclosure.

[0096] The generation unit 132 stores the generated prediction model data in the storage unit 114 as model data 122 (step S17), and ends the generation process.

[0097] FIG. 14 is a flowchart showing an example of the flow of a film forming process including a determination step of the determination method according to the embodiment.

[0098] In the film forming apparatus 100, a substrate W on which a pattern including recesses is formed is carried in by a transport mechanism such as a transport arm (not shown) and placed on the mounting table 2 (step S20). The film forming control unit 130 controls each part of the film forming apparatus 100 to perform a film forming process (step S21).

[0099] The measurement control unit 133 controls each unit of the film forming apparatus 100 to perform spectroscopic measurement on the substrate W (step S22). For example, the measurement control unit 133 controls the irradiation unit 81 and the detection unit 82 to irradiate the substrate W with measurement light, and the detection unit 82 detects the light transmitted through or reflected by the substrate W. The detection unit 82 outputs data on the signal intensity of the detected light to the control device 110. The measurement control unit 133 performs spectroscopic analysis such as Fourier transform on the signal intensity data input from the detection unit 82 to calculate one of the absorbance spectrum or the reflectance spectrum. For example, when the film forming apparatus 100 performs spectroscopic measurement by a transmission method, the measurement control unit 133 calculates the absorbance spectrum from the signal intensity data input from the detection unit 82. Also, for example, when the film forming apparatus 100 performs spectroscopic measurement by a reflection method, the measurement control unit 133 calculates the reflectance spectrum from the signal intensity data input from the detection unit 82. In this embodiment, step S22 corresponds to step a) of the present disclosure.

[0100] The determination unit 134 reads out the model data 122, and derives an embedding index value from one spectrum using a prediction model of the read out model data 122 (step S23). For example, when the prediction model has learned the relationship between the integral value of the spectral value and the embedding index value, the determination unit 134 derives the embedding index value from the prediction model using the integral value of the spectral value for each wavelength range of one spectrum as an explanatory variable and the embedding index value as an objective variable. For example, when the film forming apparatus 100 performs spectroscopic measurement using a transmission method, the determination unit 134 derives the embedding index value from the integral value of the absorbance for each wavelength range using the prediction model. Also, for example, when the film forming apparatus 100 performs spectroscopic measurement using a reflection method, the determination unit 134 derives the embedding index value from the integral value of the reflectance for each wavelength range using the prediction model.

[0101] The determination unit 134 determines the embedding state of the recess 91 based on the derived embedding index value (step S24). For example, when the derived embedding index value is greater than a predetermined threshold, the determination unit 134 determines that the embedding is defective. In this embodiment, steps S23 and S24 correspond to step b) of the present disclosure.

[0102] The determination unit 134 outputs the determination result (step S25). For example, when the determination unit 134 determines that there is an embedding defect, the determination unit 134 displays the fact that an embedding defect has occurred on the display unit 112. In addition, the determination unit 134 notifies an external device such as a management terminal used by the administrator that an embedding defect has occurred.

[0103] The substrate W placed on the mounting table 2 is carried out from the film formation apparatus 100 by a transport mechanism (not shown) (step S26), and the film formation process is completed.

[0104] Next, a specific example of verifying the prediction result by the prediction model will be described. The verification was performed by measuring the reflectance spectrum using a film forming apparatus 100 configured to perform spectroscopic measurement by the reflection method. The verification was performed using a plurality of substrates W on which patterns 90 of approximately the same shape, including recesses 91, were formed. The plurality of substrates W were filled with a filling material under the same film forming conditions by the film forming apparatus 100. FIG. 15 is a diagram showing an example of a cross-sectional configuration of the substrate W used for the verification. The substrate W is a silicon substrate or an amorphous silicon substrate. The substrate W has a plurality of core material parts 95 made of SiO formed thereon, and the entire surface is covered with an amorphous silicon film 97. The substrate W has a recess 91 between the core material parts 95 covered with the amorphous silicon film 97. The substrate W has SiN formed thereon as a filling material, and the amorphous silicon film 97 is covered with a SiN film 98. The plurality of substrates W used for the verification include those having slightly different conditions of the base. For example, the multiple substrates W used in the verification include those in which the cross-sectional shape of the core portion 95 is tapered.

[0105] In the verification, spectroscopic measurement was performed on each of the substrates W in which the filling material was embedded, the reflectance spectrum was calculated, and the integral value of the reflectance for each wavelength range of the reflectance spectrum of each substrate W was obtained. In addition, in the verification, SEM images of the cross sections of the substrates W in which the filling material was embedded were checked, and the embedding index value, the film thickness of the embedding material, and the base index value were identified. For example, in the verification, the base index value was identified from the cross-sectional shape of the core material part 95 including the amorphous silicon film 97. In the verification, the embedding index value was identified from the embedding state of the recess 91 with the SiN film 98. In the verification, the film thickness of the SiN film 98 on the upper part of the core material part 95 was measured as the film thickness of the embedding material.

[0106] First, the verification of case 1 will be described. In case 1, a learning substrate W was randomly selected from a plurality of substrates W, and the remaining substrates W were used as test substrates W. In case 1, for each learning substrate W, the integral value of the reflectance for each wavelength range of the reflectance spectrum, the embedding index value, the film thickness of the embedding material, and the base index value were associated with each other and stored in the learning data 121. In case 1, such learning data 121 was subjected to machine learning to generate a prediction model. Then, in case 1, the embedding index value of each learning substrate W and the test substrate W was derived using the generated prediction model. For example, in case 1, the integral value of the reflectance for each wavelength range and the base index value were used as explanatory variables, and the embedding index value and the film thickness of the embedding material were used as objective variables, and the embedding index value and the film thickness of the embedding material were derived from the prediction model.

[0107] FIG. 16A is a diagram for explaining the verification results of Case 1. FIG. 16A shows a graph with the embedding index value (actual measurement value) specified in advance on the horizontal axis and the embedding index value (predicted value) derived by the prediction model on the vertical axis. The verification results of the learning substrate W and the test substrate W are plotted on the graph. The coefficient of determination R between the actual measurement value and the predicted value of the learning substrate W is 2 The coefficient of determination R between the actual measured value and the predicted value of the test board W is 0.76. 2is 0.71. From the verification results of Case 1, even if multiple substrates W include substrates with slightly different base conditions, the embedding index value can be derived with high accuracy by using the integral value of the reflectance and the base index value as explanatory variables.

[0108] Next, verification of case 2 will be described. In case 2, substrates W having the same substrate state (for example, substrate index value=0) were selected from a plurality of substrates W, and some of the selected substrates W were used as learning substrates W, and the remaining selected substrates W were used as test substrates W. In case 2, for each learning substrate W, the integral value of the reflectance for each wavelength range of the reflectance spectrum, the embedding index value, and the film thickness of the embedding material were associated with each other and stored in the learning data 121. In case 2, such learning data 121 was subjected to machine learning to generate a prediction model. Then, in case 2, the embedding index values ​​of each learning substrate W and the test substrate W were derived using the generated prediction model. For example, in case 2, the integral value of the reflectance for each wavelength range and the film thickness of the embedding material were used as explanatory variables, and the embedding index value was used as a response variable, and the embedding index value was derived from the prediction model.

[0109] FIG. 16B is a diagram for explaining the verification results of Case 2. FIG. 16B shows a graph with the embedding index value (actual measurement value) specified in advance on the horizontal axis and the embedding index value (predicted value) derived by the prediction model on the vertical axis. The verification results of the learning substrate W and the test substrate W are plotted on the graph. The coefficient of determination R between the actual measurement value and the predicted value of the learning substrate W is 2 The coefficient of determination R between the actual measured value and the predicted value of the test board W is 0.98. 2 is 0.97. Case 2 shows a case where substrates W having the same base material index value are selected from a plurality of substrates W, and the substrates W having the same base material state are verified. Generally, in semiconductor manufacturing, substrates W having the same base material state are transported to the film forming apparatus 100, and a film of a filling material or the like is formed in the film forming apparatus 100. From the verification result of Case 2, in the case of substrates W having the same base material state, the filling index value can be predicted with high accuracy even without using the base material index value as an explanatory variable.

[0110] Next, the verification of Case 3 will be described. In Case 3, the embedding index value and the film thickness of the embedding material of each of the learning substrates W and the test substrates W selected in Case 2 were derived using the prediction model generated in Case 2. For example, in Case 3, the embedding index value and the film thickness of the embedding material were derived from the prediction model generated in Case 2 using the integral value of the reflectance for each wavelength range as the explanatory variable and the embedding index value and the film thickness of the embedding material as the objective variables.

[0111] FIG. 16C is a diagram for explaining the verification results of Case 3. FIG. 16C shows a graph with the embedding index value (actual measurement value) specified in advance on the horizontal axis and the embedding index value (predicted value) derived by the prediction model on the vertical axis. The verification results of the learning substrate W and the test substrate W are plotted on the graph. The coefficient of determination R of the actual measurement value and the predicted value for the embedding index value of the learning substrate W is 2 The coefficient of determination R between the actual and predicted values ​​for the embedding index of the test substrate W is 0.96. 2 is 0.98. In addition, for each learning substrate W and test substrate W, the thickness of the embedded material specified in advance (actual measurement value) and the thickness of the embedded material derived by the prediction model (prediction value) were verified. The coefficient of determination R 2 The coefficient of determination R between the actual and predicted values ​​for the film thickness of the test substrate W is 0.97. 2 is taken as 0.99. The optical transmittance and reflectance of the substrate W change depending on the film thickness of the embedding material. For this reason, the waveform of the reflectance spectrum obtained by spectroscopic measurement of the substrate W shifts in response to the film thickness of the embedding material. For this reason, there is a correlation between the integral value of the reflectance for each wavelength range of the reflectance spectrum and the film thickness of the embedding material. From the verification results of Case 3, it is possible to predict the embedding index value and the film thickness of the embedding material with high accuracy from the integral value of the reflectance for each wavelength range.

[0112] Next, a specific example will be described in which the width of the wavelength range over which the reflectance is integrated is changed. In the verification, the width of the wavelength range is changed within a range of 25 to 300 nm, and verification of Case 3 is carried out for each wavelength range width. FIG. 17 is a diagram for explaining the verification results when the width of the wavelength range is changed. In FIG. 17, the horizontal axis represents the width of the wavelength range, and the coefficient of determination R 2 The graph has lines L1 to L4. The line L1 represents the coefficient of determination R 2 The line L2 shows the change in the coefficient of determination R of the embedding index value of the test substrate W with respect to the change in the width of the wavelength range. 2 The line L3 shows the change in the coefficient of determination R of the film thickness of the learning substrate W with respect to the change in the width of the wavelength range. 2 The line L4 shows the change in the coefficient of determination R of the film thickness of the test substrate W with the change in the width of the wavelength range. 2 As shown in the line L2, the coefficient of determination R 2 For example, the coefficient of determination R of the embedding index value of the test substrate W is 2 The optical transmittance is significantly decreased when the wavelength range width is around 50 nm, and tends to decrease as the wavelength range width becomes longer. Therefore, the wavelength range width is preferably set to 75 to 200 nm, and more preferably to 100 nm.

[0113] Thus, the determination method according to the embodiment includes steps a) and b). In step a), a spectroscopic measurement is performed on the substrate W on which a pattern 90 including recesses 91 is formed and the recesses 91 are filled with a filling material, and one of the absorbance spectrum or reflectance spectrum of the substrate W is measured. In step b), the filling state of the recesses 91 is determined from the integral value of the spectral value for each wavelength range of the spectrum of one of the substrates W measured in step a) using a model generated by machine learning data that associates the integral value of the spectral value for each wavelength range of the spectrum of one of the substrates W filled with the filling material with the filling state of the recesses 91. In this way, the determination method according to the embodiment can detect the occurrence of a filling defect.

[0114] Moreover, the determination method according to the embodiment further includes steps c), d), and e) before steps a) and b). In step c), spectroscopic measurement is performed on the multiple substrates W in which the embedding material is embedded, and one of the spectra of the multiple substrates W is measured. In step d), a first index value (embedding index value) that indexes the embedding state of the recesses 91 of the multiple substrates W is obtained. In step e), for each of the multiple substrates W, data in which the integral value of the spectrum value for each wavelength range of the one of the spectra measured in step c) and the first index value acquired in step d) are associated with each other is machine-learned to generate a model (prediction model). In step b), the integral value of the spectrum value for each wavelength range of the one of the spectra of the substrate W measured in step a) is used as an explanatory variable, the first index value is used as a response variable, and the first index value is derived from the model generated in step e), and the embedding state of the recesses 91 is determined based on the derived first index value. In this way, the determination method according to the embodiment can detect the occurrence of embedding defects.

[0115] In step e), the determination method according to the embodiment performs machine learning using a linear regression algorithm to generate a model that learns the relationship between the integral value of the spectral values ​​for each wavelength range and the first index value. As a result, the determination method according to the embodiment can accurately detect the first index value corresponding to the filling state of the recesses 91 using the model, and therefore can accurately detect the occurrence of filling defects.

[0116] Moreover, the determination method according to the embodiment further includes a step f) before the step e). In the step f), the film thickness of the embedding material of the multiple substrates W in which the embedding material has been embedded is acquired. In the step e), for each of the multiple substrates W, the integral value of the spectral value for each wavelength range of one of the spectra measured in the step c), the first index value acquired in the step d), and the film thickness acquired in the step f) are associated with each other to generate a model by machine learning. In the step b), the integral value of the spectral value for each wavelength range of one of the spectra of the substrate W measured in the step a) is used as an explanatory variable, and the first index value and the film thickness are used as objective variables to derive the first index value and the film thickness from the model generated in the step e). As a result, the determination method according to the embodiment can derive the film thickness in addition to the first index value using the model.

[0117] Moreover, the determination method according to the embodiment further includes step g) before step e). In step g), a second index value (base index value) that indexes the state of the base of the plurality of substrates W in which the embedding material is embedded is obtained. In step e), for each of the plurality of substrates W, a model is generated by machine learning data that associates the integral value of the spectrum value for each wavelength range of one spectrum measured in step c), the first index value acquired in step d), and the second index value acquired in step g). In step b), the integral value of the spectrum value for each wavelength range of one spectrum of the substrate W measured in step a) and the second index value are used as explanatory variables, and the first index value is used as a response variable to derive the first index value from the model generated in step e). As a result, the determination method according to the embodiment can accurately detect the first index value corresponding to the embedding state of the recess 91 using the model even when the state of the base of the substrate W changes, and therefore can accurately detect the occurrence of embedding defects.

[0118] Although the embodiments have been described above, the disclosed embodiments should be considered to be illustrative and not restrictive in all respects. Indeed, the above-described embodiments may be embodied in various forms. Furthermore, the above-described embodiments may be omitted, substituted, or modified in various forms without departing from the scope and spirit of the claims.

[0119] For example, in the above embodiment, the irradiating unit 81 is configured to be vertically movable and rotatable, and the incident angle and irradiation position of the light incident on the substrate W are changeable. However, this is not limited to this. For example, optical elements such as mirrors and lenses may be provided in the optical path of the light irradiated from the irradiating unit 81 or the optical path of the light incident on the detecting unit 82, and the incident angle and irradiation position of the light incident on the substrate W may be changed by the optical elements.

[0120] In the above embodiment, the measurement light is transmitted through the center of the substrate W to determine the embedding state of the substrate W. However, the present invention is not limited to this. For example, an optical element that reflects light may be provided in the chamber 1, and the measurement light may be irradiated to a plurality of locations, such as the center and periphery of the substrate W, by the optical element, and the transmitted light or reflected light may be detected at each location to determine the embedding state of each of the plurality of locations of the substrate W.

[0121] In the above embodiment, the substrate processing apparatus of the present disclosure has been described as a single-chamber type film forming apparatus 100 having one chamber. However, the present disclosure is not limited to this. The substrate processing apparatus of the present disclosure may be a multi-chamber type film forming apparatus having multiple chambers.

[0122] Fig. 18 is a schematic configuration diagram showing another example of the film formation apparatus 200 according to the embodiment. As shown in Fig. 18, the film formation apparatus 200 is a multi-chamber type film formation apparatus having four chambers 201 to 204. In the film formation apparatus 200, film formation processes are performed in each of the four chambers 201 to 204.

[0123] The chambers 201 to 204 are connected to the four walls of a vacuum transfer chamber 301, which has a heptagonal planar shape, via gate valves G. The inside of the vacuum transfer chamber 301 is evacuated by a vacuum pump and maintained at a predetermined vacuum level. Three load lock chambers 302 are connected to the other three walls of the vacuum transfer chamber 301 via gate valves G1. An atmospheric transfer chamber 303 is provided on the opposite side of the load lock chamber 302 from the vacuum transfer chamber 301. The three load lock chambers 302 are connected to the atmospheric transfer chamber 303 via gate valves G2. The load lock chambers 302 control the pressure between atmospheric pressure and vacuum when transferring the substrate W between the atmospheric transfer chamber 303 and the vacuum transfer chamber 301.

[0124] Three carrier attachment ports 305 for attaching carriers (FOUPs or the like) C that accommodate substrates W are provided on the wall of the atmospheric transfer chamber 303 opposite to the wall on which the load lock chamber 302 is attached. In addition, an alignment chamber 304 for aligning the substrates W is provided on the side wall of the atmospheric transfer chamber 303. A downflow of clean air is formed within the atmospheric transfer chamber 303.

[0125] A transfer mechanism 306 is provided in the vacuum transfer chamber 301. The transfer mechanism 306 transfers a substrate W to and from the chambers 201 to 204 and the load lock chamber 302. The transfer mechanism 306 has two transfer arms 307a and 307b that are independently movable.

[0126] A transfer mechanism 308 is provided in the atmospheric transfer chamber 303. The transfer mechanism 308 transfers the substrate W to the carrier C, the load lock chamber 302, and the alignment chamber 304.

[0127] The film forming apparatus 200 includes a control device 310. The operation of the film forming apparatus 200 is generally controlled by the control device 310. The control device 310 has a similar configuration to the control device 110.

[0128] In the film forming apparatus 200 configured as above, the measuring unit 85 for performing spectroscopic measurement on the substrate W may be provided in any one of the chambers 201 to 204. For example, in the film forming apparatus 200, the measuring unit 85 for performing spectroscopic measurement on the substrate W is provided in any one of the vacuum transfer chamber 301, the load lock chamber 302, the atmospheric transfer chamber 303, and the alignment chamber 304. FIG. 19 is a diagram showing an example of a schematic configuration of the measuring unit 85 according to the embodiment. The measuring unit 85 has an irradiation unit 81 for irradiating light and a detection unit 82 for detecting light. The irradiation unit 81 and the detection unit 82 are disposed outside the housing 86 of the vacuum transfer chamber 301, the load lock chamber 302, the atmospheric transfer chamber 303, and the alignment chamber 304. The irradiation unit 81 and the detection unit 82 are connected to light guide members 87a and 87b such as optical fibers. Ends of the light guide members 87a and 87b are disposed inside the housing 86. The light output from the irradiation unit 81 is output from an end of the light guide member 87a. The end of the light guide member 87a is arranged so that the light is incident on the substrate W at a predetermined angle of incidence (for example, 45°). The end of the light guide member 87b is arranged so that the light reflected from the substrate W is incident on the end of the light guide member 87b. The light incident on the end of the light guide member 87b is detected by the detection unit 82 via the light guide member 87b. The measurement unit 85 performs spectroscopic measurement of the substrate W to measure the reflectance spectrum of the substrate W. The control device 310 uses a prediction model to determine the filling state of the recesses 91 based on the integral value of the reflectance for each wavelength range of the reflectance spectrum of the substrate W measured by the measurement unit 85. This makes it possible to detect the occurrence of filling defects in the recesses 91 of the substrate W in-line even in the film forming apparatus 200. The measurement unit 85 may perform spectroscopic measurement by making light incident perpendicularly to the substrate W. FIG. 20 is a diagram showing another example of a schematic configuration of the measurement unit 85 according to the embodiment. 20, the ends of light-guiding members 87a and 87b arranged in a housing 86 are coaxial double optical fibers, and are arranged above the substrate W so as to be perpendicular to the substrate W. Light output from the irradiation unit 81 is output from the end of the light-guiding member 87a and perpendicularly enters the substrate W. The light that enters the substrate W is reflected and enters the end of the light-guiding member 87b. The light that enters the end of the light-guiding member 87b is detected by the detection unit 82 via the light-guiding member 87b.In this manner, the measuring unit 85 may measure the reflectance spectrum by performing spectroscopic measurement with light being incident perpendicularly onto the substrate W. The control device 310 may use a prediction model to determine the filling state of the recesses 91 based on the integral value of the reflectance for each wavelength range of the reflectance spectrum of the substrate W measured by the measuring unit 85.

[0129] Furthermore, when the depth of the recesses 91 of the pattern 90 formed on the substrate W is relatively shallow (for example, within 0.5 um), the wavelength region of the light used for the spectroscopic measurement may be infrared light on the short wavelength side or light in the visible light region with a wavelength of about 200 to 1000 um. In that case, it is not necessary to angle the irradiating unit 81 and the detecting unit 82, and for example, spectroscopic measurement may be performed by irradiating light perpendicularly to the substrate W as shown in FIG.

[0130] As described above, in the above embodiment, the substrate processing apparatus of the present disclosure is disclosed as a single-chamber type having one chamber for processing substrates W one by one in the chamber, or as a multi-chamber type single-wafer substrate processing apparatus having multiple chambers. However, the present disclosure is not limited to this. For example, the substrate processing apparatus of the present disclosure may be a batch type substrate processing apparatus capable of processing multiple substrates at once, or may be a carousel type semi-batch type substrate processing apparatus.

[0131] In the above embodiment, the case where the occurrence of filling defects is detected in the film forming apparatuses 100 and 200 has been described as an example. However, the present invention is not limited to this. The spectroscopic measurement may be performed by a measurement unit such as a measurement device that performs spectroscopic measurement other than the film forming apparatuses 100 and 200. The filling state of the recess 91 may be determined by an information processing device such as a computer other than the film forming apparatuses 100 and 200. For example, the substrate W on which the filling material is formed in the film forming apparatuses 100 and 200 may be transported to a measurement device, and the measurement device may perform spectroscopic measurement to measure one of the absorbance spectrum or reflectance spectrum of the substrate W. The information processing device 400 may acquire data of one spectrum of the substrate W from the measurement device, and determine the filling state of the recess 91 of the substrate W from the integral value of the spectrum value for each wavelength range of the one spectrum of the substrate W indicated by the acquired data. FIG. 21 is a diagram showing an example of a schematic configuration of the information processing device 400 according to the embodiment. Since the information processing device 400 is partially the same as the control device 110, the same parts are denoted by the same reference numerals and description thereof will be omitted below, and mainly different parts will be described. The acquisition unit 131 acquires data of one of the absorbance spectrum or reflectance spectrum of the substrate W from a measurement unit (measurement device). For example, the acquisition unit 131 acquires data of one of the spectra of the substrate W from the measurement device via a network (not shown). The judgment unit 134 judges the filling state of the recess 91 of the substrate W from the integral value of the spectrum value for each wavelength range of the one of the spectra of the substrate W indicated by the data acquired by the acquisition unit 131 using a prediction model. In this way, the information processing device 400 according to the embodiment can detect the occurrence of filling defects.

[0132] It should be noted that the embodiments disclosed herein are illustrative in all respects and should not be construed as limiting. Indeed, the above-described embodiments may be embodied in various forms. Furthermore, the above-described embodiments may be omitted, substituted, or modified in various forms without departing from the scope and spirit of the appended claims.

[0133] Furthermore, the following supplementary notes are disclosed regarding the above embodiment.

[0134] (Appendix 1) a) performing spectroscopic measurement of a substrate on which a pattern including recesses is formed and a filling material is filled into the recesses, thereby measuring one of an absorbance spectrum or a reflectance spectrum of the substrate; b) determining a filling state of the recesses from an integral value of a spectral value for each wavelength range of the spectrum of one of the substrates measured in step a) using a model generated by machine learning data correlating an integral value of a spectral value for each wavelength range of the spectrum of one of the substrates of the plurality of substrates in which the filling material is embedded with the filling state of the recesses; A determination method comprising:

[0135] (Appendix 2) Before the step a) and the step b), c) performing spectroscopic measurement of the plurality of substrates in which the embedding material has been embedded, and measuring a spectrum of the one of the plurality of substrates; d) obtaining a first index value that indexes the filling state of the recesses of the plurality of substrates; e) performing machine learning on data associating, for each of the plurality of substrates, an integral value of the spectral value for each of the wavelength ranges of the one spectrum measured in step c) with the first index value acquired in step d) to generate the model, In the step b), an integral value of the spectral value for each wavelength range of the one spectrum of the substrate measured in the step a) is used as an explanatory variable, and the first index value is used as a response variable, to derive the first index value from the model generated in the step e), and a filling state of the recess is determined based on the derived first index value. The determination method described in Appendix 1.

[0136] (Appendix 3) In the step e), machine learning is performed using a linear regression algorithm to generate a model that has learned the relationship between the integral value of the spectral value for each wavelength range and the first index value. The determination method described in Appendix 2.

[0137] (Appendix 4) Before step e), f) acquiring a thickness of the filling material for the plurality of substrates into which the filling material has been filled; In the step e), for each of the plurality of substrates, data correlating an integral value of the spectral value for each of the wavelength ranges of the one spectrum measured in the step c), the first index value acquired in the step d), and the film thickness acquired in the step f) with each other is machine-learned to generate the model; In the step b), an integral value of the spectrum value for each wavelength range of the one spectrum of the substrate measured in the step a) is used as an explanatory variable, and the first index value and the film thickness are used as objective variables to derive the first index value and the film thickness from the model generated in the step e). The method for determining whether or not a substance is soluble in water is described in Appendix 2 or 3.

[0138] (Appendix 5) Before step e), g) acquiring a second index value that indexes the state of the base of the plurality of substrates into which the embedding material has been embedded; In the step e), for each of the plurality of substrates, data correlating an integral value of the spectral value for each wavelength range of the one spectrum measured in the step c), the first index value acquired in the step d), and the second index value acquired in the step g) with each other is machine-learned to generate the model; In the step b), the integrated value of the spectral value for each wavelength range of the one spectrum of the substrate measured in the step a) and the second index value are used as explanatory variables, and the first index value is used as a response variable, to derive the first index value from the model generated in the step e). A determination method according to any one of appendices 2 to 4.

[0139] (Appendix 6) The width of each of the wavelength ranges is between 75 and 200 nm. A determination method according to any one of appendices 1 to 5.

[0140] (Appendix 7) The wavelength ranges are set so that the wavelengths are continuous. A determination method according to any one of appendices 1 to 6.

[0141] (Appendix 8) The wavelength ranges are set to a range of 200 to 700 nm. A determination method according to any one of appendices 1 to 7.

[0142] (Appendix 9) a) performing spectroscopic measurement of a substrate on which a pattern including recesses is formed and a filling material is filled into the recesses, thereby measuring one of an absorbance spectrum or a reflectance spectrum of the substrate; b) determining a filling state of the recesses from an integral value of a spectral value for each wavelength range of the spectrum of one of the substrates measured in step a) using a model generated by machine learning data correlating an integral value of a spectral value for each wavelength range of the spectrum of one of the substrates of the plurality of substrates in which the filling material is embedded with the filling state of the recesses; A judgment program that causes a computer to execute the above.

[0143] (Appendix 10) a measurement unit configured to perform spectroscopic measurement of a substrate on which a pattern including recesses is formed and a filling material is filled into the recesses, and to measure one of an absorbance spectrum and a reflectance spectrum of the substrate; a determination unit configured to determine a filling state of the recesses from an integral value of a spectral value for each wavelength range of the one spectrum of the substrates measured by the measurement unit, using a model generated by machine learning data that associates an integral value of a spectral value for each wavelength range of the one spectrum of the substrates of the substrates into which the filling material is embedded with the filling state of the recesses, the integral value being determined for each wavelength range of the one spectrum of the substrates; A substrate processing apparatus comprising:

[0144] (Appendix 11) an acquisition unit configured to perform spectroscopic measurement of a substrate on which a pattern including recesses is formed and a filling material is filled into the recesses, and to acquire data of one of an absorbance spectrum and a reflectance spectrum of the substrate from a measurement unit that measures one of the absorbance spectrum and the reflectance spectrum of the substrate; a determination unit configured to determine a filling state of the recesses from an integral value of a spectral value for each wavelength range of the one spectrum of the substrate indicated by the data acquired by the acquisition unit, using a model generated by machine learning data that associates an integral value of a spectral value for each wavelength range of the one spectrum of the substrates of the plurality of substrates in which the filling material is embedded with the filling state of the recesses, the integral value of the spectral value for each wavelength range of the one spectrum of the substrate; An information processing device having the above configuration. [Explanation of symbols]

[0145] 1 Chamber 2. Placement table 81 Irradiation unit 82 Detection unit 85 Measurement section 86 Case 87a Light guide member 87b Light guide member 90 Patterns 91, 91a recess 92 membrane 93 Void 95 Core 96 Side wall 97 Amorphous silicon film 100 Film deposition equipment 110 Control device 112 Display section 113 Input section 114 Storage section 115 Control Unit 120 Processing condition data 121 Training Data 122 Model Data 130 Film formation control unit 131 Acquisition Department 132 Generation part 133 Measurement control section 134 Judgment section 200 Film deposition equipment 201~204 Chamber 301 Vacuum Transfer Chamber 302 Load Lock Chamber 303 Atmospheric Transfer Room 304 Alignment Chamber 305 Carrier Mounting Port 306 Conveyor Mechanism 307a Transport arm 307b Transport arm 308 Conveyor Mechanism 310 Control device 400 Information processing device W substrate

Claims

1. a) performing spectroscopic measurement of a substrate on which a pattern including recesses is formed and a filling material is filled into the recesses, thereby measuring one of an absorbance spectrum or a reflectance spectrum of the substrate; b) determining a filling state of the recesses from an integral value of a spectral value for each wavelength range of the spectrum of one of the substrates measured in step a) using a model generated by machine learning data correlating an integral value of a spectral value for each wavelength range of the spectrum of one of the substrates of the substrates filled with the filling material with the filling state of the recesses; A determination method comprising:

2. Prior to step a) and step b), c) performing spectroscopic measurement of the plurality of substrates into which the embedding material has been embedded, and measuring a spectrum of the one of the plurality of substrates; d) obtaining a first index value that indexes the filling state of the recesses of the plurality of substrates; e) performing machine learning on data associating, for each of the plurality of substrates, an integral value of the spectral value for each wavelength range of the one spectrum measured in step c) with the first index value acquired in step d) to generate the model, In the step b), an integrated value of the spectral value for each wavelength range of the one spectrum of the substrate measured in the step a) is used as an explanatory variable, and the first index value is used as a response variable, to derive the first index value from the model generated in the step e), and a filling state of the recess is determined based on the derived first index value. The method according to claim 1 .

3. In the step e), machine learning is performed using a linear regression algorithm to generate a model that has learned the relationship between the integral value of the spectral value for each wavelength range and the first index value. The method according to claim 2 .

4. Prior to step e), f) acquiring a thickness of the filling material for the plurality of substrates in which the filling material has been filled; In the step e), for each of the plurality of substrates, data correlating an integral value of the spectral value for each wavelength range of the one spectrum measured in the step c), the first index value acquired in the step d), and the film thickness acquired in the step f) with each other is machine-learned to generate the model; In the step b), an integral value of the spectrum value for each wavelength range of the one spectrum of the substrate measured in the step a) is used as an explanatory variable, and the first index value and the film thickness are used as objective variables to derive the first index value and the film thickness from the model generated in the step e). The method according to claim 2 .

5. Prior to step e), g) acquiring a second index value that indexes a state of the base of the plurality of substrates into which the embedding material is embedded; In the step e), for each of the plurality of substrates, data correlating an integral value of the spectral value for each wavelength range of the one spectrum measured in the step c), the first index value acquired in the step d), and the second index value acquired in the step g) with each other is machine-learned to generate the model; In the step b), the integrated value of the spectral value for each wavelength range of the one spectrum of the substrate measured in the step a) and the second index value are used as explanatory variables, and the first index value is used as a response variable, to derive the first index value from the model generated in the step e). The method according to claim 2 .

6. The width of each of the wavelength ranges is between 75 and 200 nm. The method according to claim 1 .

7. The wavelength ranges are set so that the wavelengths are continuous. The method according to claim 1 .

8. The wavelength ranges are set to a range of 200 to 700 nm. The method according to claim 1 .

9. a) performing spectroscopic measurement of a substrate on which a pattern including recesses is formed and a filling material is filled into the recesses, thereby measuring one of an absorbance spectrum or a reflectance spectrum of the substrate; b) determining a filling state of the recesses from an integral value of a spectral value for each wavelength range of the spectrum of one of the substrates measured in step a) using a model generated by machine learning data correlating an integral value of a spectral value for each wavelength range of the spectrum of one of the substrates of the substrates filled with the filling material with the filling state of the recesses; A judgment program that causes a computer to execute the above.

10. a measurement unit configured to perform spectroscopic measurement of a substrate on which a pattern including recesses is formed and a filling material is filled into the recesses, and to measure one of an absorbance spectrum and a reflectance spectrum of the substrate; a determination unit configured to determine a filling state of the recesses from an integral value of a spectral value for each wavelength range of the one spectrum of the substrates measured by the measurement unit, using a model generated by machine learning data that associates an integral value of a spectral value for each wavelength range of the one spectrum of the substrates of the substrates into which the filling material is embedded with the filling state of the recesses, the integral value being determined for each wavelength range of the one spectrum of the substrates; A substrate processing apparatus comprising:

11. an acquisition unit configured to perform spectroscopic measurement of a substrate on which a pattern including recesses is formed and a filling material is filled into the recesses, and to acquire data on one of an absorbance spectrum and a reflectance spectrum of the substrate from a measurement unit that measures the one of the absorbance spectrum and the reflectance spectrum of the substrate; a determination unit configured to determine a filling state of the recesses from an integral value of a spectral value for each wavelength range of the one spectrum of the substrate indicated by the data acquired by the acquisition unit, using a model generated by machine learning data that associates an integral value of a spectral value for each wavelength range of the one spectrum of the substrates of the plurality of substrates in which the filling material is embedded with the filling state of the recesses, the integral value of the spectral value for each wavelength range of the one spectrum of the substrate; An information processing device having the above configuration.

Images