Machine learning method, laser annealing system, and laser annealing method

Abstract

This machine learning method comprises: acquiring image data generated from reflected light of light illuminating a first region that is of a semiconductor film on a substrate and has been laser-annealed with pulsed laser light; acquiring measurement data of semiconductor characteristics in the first region; generating training data in which the image data as an input and the measurement data as an output are associated with each other; and performing machine learning using a neural network on the basis of the training data to generate a learned model.

Description

[Technical field]

[0001] The present disclosure relates to machine learning methods, laser annealing systems, and laser annealing methods. [Background technology]

[0002] TFTs (thin film transistors) are used as the driving elements for flat panel displays that use glass substrates. To realize high-definition displays, it is necessary to create TFTs with high driving power. Polycrystalline silicon and IGZO (indium gallium zinc oxide) are used for the semiconductor thin film that serves as the channel material for TFTs. Polycrystalline silicon and IGZO have higher carrier mobility than amorphous silicon, and have excellent on / off characteristics for transistors.

[0003] Thin-film semiconductors are also expected to be applied to 3D-ICs, which will realize more advanced devices. 3D-ICs can be realized by forming active elements such as sensors, amplifier circuits, and CMOS circuits on the top layer of integrated circuit devices. For this reason, there is a demand for technology to manufacture higher quality semiconductor thin films.

[0004] Furthermore, with the diversification of information terminal devices, there is an increasing demand for flexible displays and computers that are small, lightweight, consume little power, and can be freely bent, which has led to a demand for the establishment of technology to form high-quality semiconductor thin films on plastic substrates such as PET (polyethylene terephthalate).

[0005] In order to form high-quality semiconductor thin films on glass substrates, integrated circuits, or plastic substrates, it is necessary to crystallize the semiconductor thin films without thermally damaging these substrates. For glass substrates used in displays, 400°C is required, for integrated circuits, 400°C is required, and for PET, a plastic substrate, a process temperature of 200°C or less is required.

[0006] Laser annealing is used as a technique for crystallizing a semiconductor thin film without thermally damaging the underlying substrate. This method uses pulsed ultraviolet laser light that is absorbed by the upper semiconductor thin film to suppress damage to the substrate due to thermal diffusion.

[0007] When the semiconductor thin film is silicon, a XeF excimer laser with a wavelength of 351 nm, a XeCl excimer laser with a wavelength of 308 nm, a KrF excimer laser with a wavelength of 248 nm, etc. are used. These gas lasers in the ultraviolet region have the advantage that the laser light has lower coherence than solid-state lasers, has excellent energy uniformity on the laser light irradiated surface, and can anneal a wide area uniformly with high pulse energy. [Prior art documents] [Patent documents]

[0008] [Patent Document 1] International Publication No. 2006 / 055003 [Patent Document 2] US Patent Application Publication No. 2017 / 345140 [Non-patent literature]

[0009] [Non-Patent Document 1] Paul C. van der Wilt, "Excimer-Laser Annealing: Microstructure Evolution and a Novel Characterization Technique", SID 2014 DIGEST, 149-152 (2014) Overview

[0010] A machine learning method according to one aspect of the present disclosure includes acquiring image data generated from reflected light of illumination light irradiated onto a first region of a semiconductor film on a substrate, the first region being laser annealed with pulsed laser light, acquiring measurement data of semiconductor characteristics in the first region, generating training data in which the image data is input and the measurement data is output, and performing machine learning using a neural network based on the training data to generate a trained model.

[0011] A laser annealing system according to one aspect of the present disclosure includes a laser device that outputs pulsed laser light for laser annealing a first region of a semiconductor film on a substrate, an illumination light source that irradiates illumination light onto the first region, an imaging unit that generates image data from reflected light of the illumination light, and a processor that is configured to be able to access a trained model that receives the image data as input and outputs an estimated value of a semiconductor characteristic, calculates the estimated value using the image data and the trained model, and corrects the irradiation conditions of the pulsed laser light based on the estimated value.

[0012] A laser annealing method according to one aspect of the present disclosure includes annealing a first region of a semiconductor film on a substrate with pulsed laser light, irradiating the first region with illumination light and generating image data from reflected light of the illumination light, calculating an estimate using a trained model that takes the image data as input and outputs an estimate of a semiconductor characteristic, and correcting the irradiation conditions of the pulsed laser light based on the estimate. [Brief description of the drawings]

[0013] Some embodiments of the present disclosure will now be described, by way of example only, with reference to the accompanying drawings. [Figure 1] FIG. 1 shows a schematic configuration of a laser annealing system according to a comparative example. [Diagram 2] FIG. 2 is a plan view showing an example of a mask. [Diagram 3] FIG. 3 is a plan view showing an example of a method for laser annealing a substrate by the step-and-repeat method. [Figure 4] FIG. 4 shows an example of setting data of irradiation conditions for the irradiation regions Sr(1) to Sr(Kmax). [Diagram 5] FIG. 5 is a flowchart showing the operation of the laser annealing processor in the comparative example. [Figure 6] FIG. 6 is a flowchart showing details of the initial adjustment process shown in FIG. [Figure 7] FIG. 7 is a flowchart showing details of a process for controlling the laser annealing apparatus and the laser apparatus shown in FIG. [Figure 8] FIG. 8 illustrates a schematic configuration of a machine learning system according to the first embodiment. [Figure 9] FIG. 9 is a flowchart showing the operation of the machine learning processor in the first embodiment. [Figure 10] FIG. 10 is a flow chart showing the details of a process for producing an annealed sample using the laser annealing system. [Figure 11] FIG. 11 is a flowchart showing details of a process for acquiring image data by the image data generating unit. [Figure 12] FIG. 12 is a flowchart showing details of a process for acquiring measurement data of semiconductor characteristics by the semiconductor characteristic measuring unit. [Figure 13] FIG. 13 is a flowchart showing the details of the processing for generating the training data. [Figure 14] FIG. 14 shows an example of the teacher data generated by the process of FIG. [Figure 15] FIG. 15 is a flowchart showing details of the process of performing machine learning to generate a trained model. [Figure 16] FIG. 16 shows a schematic diagram of an example of a deep structure. [Figure 17] FIG. 17 is a graph showing the evaluation results of the trained model using the evaluation data in S65 of FIG. [Figure 18] FIG. 18 shows a schematic configuration of a laser annealing system according to the second embodiment. [Figure 19] FIG. 19 shows a schematic configuration of the imaging section. [Figure 20] FIG. 20 is a flowchart showing the operation of the laser annealing processor in the second embodiment. [Figure 21] FIG. 21 is a flowchart showing details of the initial adjustment process shown in FIG. [Figure 22] FIG. 22 is a flowchart showing the details of the process for calculating the estimated value Eme(k) of the electron mobility shown in FIG. [Diagram 23] FIG. 23 is a flowchart showing details of the process of updating the irradiation conditions shown in FIG. [Figure 24] FIG. 24 is a flowchart showing a first example of a process for correcting the irradiation condition shown in FIG. [Diagram 25] FIG. 25 is a graph showing a typical relationship between the target value of fluence Ft(k) and the estimated value of electron mobility Eme(k). [Figure 26] FIG. 26 is a table for explaining the criteria for determining the control direction of the target value Ft(k) of the fluence in the second embodiment. [Figure 27] FIG. 27 is a flowchart showing a second example of the process of correcting the irradiation conditions shown in FIG. [Figure 28] FIG. 28 is a flowchart showing a third example of the process of correcting the irradiation condition shown in FIG. [Figure 29] FIG. 29 is a graph showing a typical relationship between the number of irradiation pulses Np(k) and the estimated value Eme(k) of electron mobility. [Diagram 30] FIG. 30 is a flowchart showing the operation of the laser annealing processor in the third embodiment. [Diagram 31] FIG. 31 is a flowchart showing details of the initial adjustment process shown in FIG. [Diagram 32] FIG. 32 is a flowchart showing details of the process of calculating the estimated value Vthe(k) of the gate threshold voltage shown in FIG. [Diagram 33]FIG. 33 is a flowchart showing details of the process of updating the irradiation conditions shown in FIG. [Diagram 34] FIG. 34 is a flowchart showing a first example of the process of correcting the irradiation condition shown in FIG. [Diagram 35] FIG. 35 is a graph showing a typical relationship between the target value of the fluence Ft(k) and the estimated value of the gate threshold voltage Vthe(k). [Diagram 36] FIG. 36 is a table for explaining the criteria for determining the control direction of the target value Ft(k) of the fluence in the third embodiment. [Figure 37] FIG. 37 is a plan view showing an example of a method for laser annealing a substrate by the step-and-scan method. [Figure 38] FIG. 38 is a plan view showing the illumination range of the substrate by the illumination light source. [Figure 39] FIG. 39 is a plan view showing the illumination range of the substrate by the illumination light source. [Diagram 40] FIG. 40 is a flowchart showing the operation of the laser annealing processor in the fourth embodiment. [Diagram 41] FIG. 41 is a flowchart showing details of the initial adjustment process shown in FIG. [Diagram 42] FIG. 42 is a flowchart showing details of the annealing setting and annealing start processing for the scan area Sc(m) shown in FIG. [Diagram 43] FIG. 43 is a flowchart showing details of the process of counting Np+n pulses shown in FIG. [Diagram 44] FIG. 44 is a flowchart showing details of the process of acquiring image data shown in FIG. Embodiment

[0014] <Contents> 1. Comparative Example 1.1 Configuration of the laser annealing system 10 1.1.1 Laser device 12 1.1.2 Laser annealing device 14 1.2 Operation 1.2.1 Laser device 12 1.2.2 Laser annealing device 14 1.2.3 Pulsed laser light irradiation sequence 1.2.4 Irradiation conditions 1.2.5 Operation of the Laser Annealing Processor 100 1.2.5.1 Main flow 1.2.5.2 Initial adjustment 1.2.5.3 Control of the laser annealing device 14 and the laser device 12 1.3 Issues with the Comparative Example 2. Machine Learning Systems 2.1 Configuration 2.1.1 Laser annealing system 10 2.1.2 Image data generating unit 20 2.1.3 Semiconductor characteristic measuring unit 40 2.1.4 Machine Learning Processor 500 2.2 How Machine Learning Works 2.2.1 Main flow 2.2.2 Preparation of annealed samples 2.2.3 Acquisition of image data 2.2.4 Acquisition of measurement data of semiconductor characteristics 2.2.5 Generating training data 2.2.6 Machine Learning 2.3 Effect 3. Laser annealing system 10b that estimates electron mobility and corrects irradiation conditions 3.1 Configuration 3.2 Operation 3.2.1 Main flow 3.2.2 Initial adjustment 3.2.3 Calculation of the estimated electron mobility Eme(k) 3.2.4 Update of irradiation conditions 3.2.4.1 First example of processing to correct illumination conditions 3.2.4.2 Second example of processing to correct illumination conditions 3.2.4.3 Third example of processing to correct illumination conditions 3.3 Effect 3.4 Other configuration examples 4. Laser annealing system 10b that estimates gate threshold voltage and corrects irradiation conditions 4.1 Configuration 4.2 Operation 4.3 Effect 5. Step-and-scan laser annealing system 10b 5.1 Configuration 5.2 Operation 5.3 Operation of the Laser Annealing Processor 100 5.3.1 Main flow 5.3.2 Initial adjustment 5.3.3 Annealing setting for scan area Sc(m) and annealing start 5.3.4 Counting Np+n Pulses 5.3.5 Acquiring image data 5.4 Effect 5.5 Other configuration examples 6.Other

[0015] Hereinafter, the embodiments of the present disclosure will be described in detail with reference to the drawings. The embodiments described below show some examples of the present disclosure, and do not limit the contents of the present disclosure. Furthermore, not all of the configurations and operations described in each embodiment are necessarily essential as the configurations and operations of the present disclosure. Note that the same components are given the same reference symbols, and duplicated explanations will be omitted.

[0016] 1. Comparative Example 1.1 Configuration of the laser annealing system 10 1 is a schematic diagram showing the configuration of a laser annealing system 10 according to a comparative example. The comparative example of the present disclosure is a form that the applicant recognizes as being known only by the applicant, and is not a publicly known example that the applicant acknowledges. The laser annealing system 10 includes a laser device 12, an optical path tube 13, and a laser annealing device 14.

[0017] 1.1.1 Laser device 12 The laser device 12 is a discharge excitation type laser device that outputs ultraviolet pulsed laser light. The laser device 12 is, for example, an excimer laser device that uses a laser gas containing argon gas or krypton gas as a rare gas, fluorine gas as a halogen gas, and neon gas as a buffer gas. The pulse time width of the pulsed laser light is, for example, 10 ns or more and 100 ns or less.

[0018] 1.1.2 Laser annealing device 14 The laser annealing apparatus 14 includes an illumination optical system 170 , a frame 172 , an XYZ stage 174 , and a laser annealing processor 100 .

[0019] An irradiation optical system 170 and an XYZ stage 174 are fixed to a frame 172. A substrate SUB is supported on a table 176 of the XYZ stage 174. The XYZ stage 174 is configured to be able to move the substrate SUB in directions intersecting the incident direction of the pulsed laser light to the substrate SUB, for example, the X-axis direction and the Y-axis direction, and in a direction having the same directional component as the incident direction, for example, the Z-axis direction. The XYZ stage 174 corresponds to the stage in this disclosure.

[0020] The substrate SUB is, for example, a glass substrate on which a film of amorphous silicon, IGZO, or IZO (indium zinc oxide) is formed, an integrated circuit, or a plastic substrate.

[0021] The irradiation optical system 170 includes high-reflection mirrors 111, 112, 113, 114, and 115, an attenuator 120, a shaping optical system 130, a mask 140, a projection optical system 145, a window 146, and a housing 150. The high-reflection mirrors 111, 112, 113, 114, and 115 are arranged in this order on the optical path of the pulsed laser light that has passed through the optical path tube 13.

[0022] The attenuator 120 is disposed in the optical path between the high reflection mirror 111 and the high reflection mirror 112. The attenuator 120 includes two partial reflection mirrors 121 and 122 and rotation stages 123 and 124. The rotation stages 123 and 124 are configured to be able to change the angles of incidence of the pulsed laser light with respect to the partial reflection mirrors 121 and 122, respectively. The transmittance of the attenuator 120 is adjusted by the angles of incidence with respect to the partial reflection mirrors 121 and 122.

[0023] The shaping optical system 130 and the mask 140 are disposed in the optical path between the high-reflection mirror 113 and the high-reflection mirror 114. The shaping optical system 130 includes a fly-eye lens 134 and a condenser lens 136. The fly-eye lens 134 is disposed so that the focal plane of the fly-eye lens 134 and the front focal plane of the condenser lens 136 coincide with each other. The condenser lens 136 is disposed so that the rear focal plane of the condenser lens 136 and the position of the mask 140 coincide with each other. The mask 140 is Koehler illuminated by the shaping optical system 130.

[0024] 2 is a plan view showing an example of the mask 140. The mask 140 is, for example, a dot mask in which a large number of dot-shaped metal thin film pieces 141 are arranged in a lattice pattern on the surface of a synthetic quartz substrate. "Arranged in a lattice pattern" refers to a state in which dot-shaped metal thin film pieces 141 are arranged at the intersections of a number of imaginary vertical lines and a number of imaginary horizontal lines. The metal thin film pieces 141 and the surrounding shaded areas of the mask 140 block the pulsed laser light.

[0025] 1, the projection optical system 145 is disposed in the optical path between the high reflection mirror 115 and the substrate SUB. The projection optical system 145 may be a reduction projection optical system configured by a combination of a plurality of lenses 143 and 144.

[0026] The window 146 is disposed in the optical path between the projection optical system 145 and the substrate SUB. The housing 150 is sealed with an O-ring or the like to prevent outside air from entering the housing 150. 2It may be purged with gas.

[0027] The laser annealing processor 100 is a processing device including a memory 101 in which a control program is stored, and a CPU (central processing unit) 102 that executes the control program. The laser annealing processor 100 is specially configured or programmed to execute various processes included in the present disclosure.

[0028] The laser annealing processor 100 outputs data on the target pulse energy Et and a trigger signal to the laser device 12. The laser annealing processor 100 also controls the operations of the attenuator 120 and the XYZ stage 174. The laser annealing processor 100 controls a workpiece transport robot (not shown) for setting the substrate SUB on a table 176 of the XYZ stage 174.

[0029] 1.2 Operation 1.2.1 Laser device 12 The laser device 12 outputs a pulsed laser beam for laser annealing an amorphous semiconductor film on a substrate SUB. The laser device 12 outputs a pulsed laser beam based on data of a target pulse energy Et and a trigger signal received from the laser annealing processor 100. The laser device 12 may include an energy monitor (not shown) and feedback control the pulse energy based on the output result of the energy monitor.

[0030] 1.2.2 Laser annealing device 14 In the laser annealing apparatus 14, the XYZ stage 174 positions the table 176 so that an inverted image of the mask 140 is formed by the projection optical system 145 in the irradiation region Sr(k) of the substrate SUB.

[0031] The pulsed laser light output from the laser device 12 enters the laser annealing device 14 via an optical path pipe 13 .

[0032] The pulsed laser light incident on the laser annealing device 14 passes through the attenuator 120 at a set transmittance. The pulsed laser light transmitted through the attenuator 120 has its light intensity distribution spatially uniformed by the shaping optical system 130 , and then enters the mask 140 .

[0033] According to the pattern formed on the mask 140, a part of the pulsed laser light is blocked and the other part is transmitted through the mask 140. The pulsed laser light that has transmitted through the mask 140 is reduced and projected onto the surface of the substrate SUB by the projection optical system 145. When the amorphous silicon film on the surface of the substrate SUB is irradiated with the laser light, the amorphous silicon film is heated and melted, and then crystallized.

[0034] 2 is formed on the surface of the substrate SUB, the light-shielded portions shielded by the metal thin film pieces 141 are prevented from increasing in temperature, and therefore crystallization begins earlier than in other portions after melting due to the energy of the pulsed laser light. Crystal growth advances in all four directions from each light-shielded portion, and ends when the outer edge of the portion reaches the outer edge of the adjacent crystal. This makes it possible to form a polycrystalline silicon film with small variations in crystal size and characteristics.

[0035] 1.2.3 Pulsed laser light irradiation sequence Fig. 3 is a plan view showing an example of a step-and-repeat laser annealing method for a substrate SUB. In Fig. 3, the surface of the substrate SUB includes Kmax irradiation regions Sr(1) to Sr(Kmax). One of the irradiation regions Sr(1) to Sr(Kmax) is designated as Sr(k). k can take an integer value from 1 to Kmax. The shape of the irradiation region Sr(k) corresponds to the shape of the beam cross section of the pulsed laser light on the substrate SUB. The irradiation region Sr(k) is an example of a first region in this disclosure.

[0036] 3 indicate the order of laser annealing. For example, when the substrate SUB is moved in the -X direction by the XYZ stage 174, the position of the beam cross section of the pulsed laser light with respect to the substrate SUB moves relatively in the X direction. By controlling the positions of the XYZ stage 174 in the X-axis direction and the Y-axis direction, the irradiation regions Sr(1) to Sr(Kmax) are laser annealed in this order.

[0037] The procedure for laser annealing using the step-and-repeat method is as follows. First, the XYZ stage 174 is operated to move and stop the substrate SUB so that the initial irradiation region Sr(1) coincides with the region where the inverted image of the mask 140 is formed. Next, the attenuator 120 is controlled so that the fluence of the pulsed laser light on the substrate SUB becomes a target value Ft(1). Then, the pulsed laser light having the set number of irradiation pulses Np(1) is irradiated onto the irradiation region Sr(1) via the mask 140. Thereafter, the output of the pulsed laser light is stopped.

[0038] Next, the XYZ stage 174 is operated to move and stop the substrate SUB so that the next irradiation region Sr(2) coincides with the region where the inverted image of the mask 140 is formed. Then, the attenuator 120 is controlled so that the fluence of the pulsed laser light becomes a target value Ft(2), and the pulsed laser light having the number of irradiation pulses Np(2) is irradiated onto the irradiation region Sr(2) via the mask 140. Thereafter, the output of the pulsed laser light is stopped. Thereafter, the irradiated regions Sr(3) to Sr(Kmax) are similarly laser annealed.

[0039] Each of the irradiation regions Sr(1) to Sr(Kmax) has a rectangular shape, and the lengths of its two sides are Bx and By, respectively. Bx and By are the beam widths of the pulsed laser light on the surface of the substrate SUB, and correspond to the beam widths in the X-axis and Y-axis directions, respectively. The transmittance of the attenuator 120 is Ta, and the total transmittance of the optical system from the output position of the laser device 12 to the substrate SUB when the transmittance Ta of the attenuator 120 is maximum is Tt. At this time, the fluence in each of the irradiation regions Sr(1) to Sr(Kmax) is calculated by Tt·Ta·Et / (Bx·By). Therefore, in order to set the fluence in the irradiation region Sr(k) to a value near the target value Ft(k), the transmittance Ta of the attenuator 120 is controlled to a value calculated by the following formula. Ta = Ft(k) (Bx By) / (Tt Et)

[0040] The method of controlling the transmittance Ta of the attenuator 120 is not limited to the above. An energy monitor (not shown) may be disposed in the optical path of the pulsed laser light transmitted through the attenuator 120, and the transmittance Ta may be feedback-controlled based on the output result of the energy monitor.

[0041] 1.2.4 Irradiation conditions 4 shows an example of setting data of irradiation conditions for irradiation regions Sr(1) to Sr(Kmax). The setting data of irradiation conditions includes target values ​​of fluence Ft(1) to Ft(Kmax) and numbers of irradiation pulses Np(1) to Np(Kmax). The numbers of irradiation pulses Np(1) to Np(Kmax) refer to the number of pulses of the pulsed laser light irradiated to one location on the substrate SUB. The setting data of irradiation conditions is stored, for example, in memory 101 (see FIG. 1).

[0042] In order to anneal different irradiation regions Sr(1) to Sr(Kmax), different fluence target values ​​Ft(1) to Ft(Kmax) may be set, or different numbers of irradiation pulses Np(1) to Np(Kmax) may be set. Alternatively, the fluence target values ​​Ft(1) to Ft(Kmax) may be a common value, or the numbers of irradiation pulses Np(1) to Np(Kmax) may be a common value.

[0043] 1.2.5 Operation of the Laser Annealing Processor 100 1.2.5.1 Main flow 5 is a flowchart showing the operation of the laser annealing processor 100 in the comparative example. In the comparative example, laser annealing is performed by the step-and-repeat method in the following manner.

[0044] In S100, the laser annealing processor 100 performs an initial adjustment. Details of S100 will be described later with reference to FIG. In S130, the laser annealing processor 100 sets the value of k to an initial value of 1. The value of k corresponds to that described with reference to FIGS.

[0045] In S140, the laser annealing processor 100 sets irradiation conditions for the irradiation region Sr(k). The irradiation conditions include a target value Ft(k) of the fluence and a number of irradiation pulses Np(k). In S150, the laser annealing processor 100 controls the laser annealing device 14 and the laser device 12 so as to perform laser annealing on the irradiation region Sr(k). Details of S150 will be described later with reference to FIG.

[0046] In S200, the laser annealing processor 100 determines whether the value of k has reached Kmax. If the value of k has not reached Kmax (S200: NO), the laser annealing processor 100 proceeds to S210. In S210, the laser annealing processor 100 adds 1 to the value of k to update the value of k. After that, the laser annealing processor 100 returns the process to S140. If the value of k reaches Kmax (S200: YES), the laser annealing processor 100 ends the process of this flowchart.

[0047] In this manner, laser annealing is performed from the irradiation regions Sr(1) to Sr(Kmax).

[0048] 1.2.5.2 Initial adjustment Fig. 6 is a flowchart showing details of the initial adjustment process shown in Fig. 5. The flowchart shown in Fig. 6 corresponds to the subroutine of S100 in Fig. 5.

[0049] In S101 , the laser annealing processor 100 controls a workpiece transport robot (not shown) so that the substrate SUB is set on the table 176 of the XYZ stage 174 .

[0050] In S102, the laser annealing processor 100 reads setting data of the irradiation conditions of the irradiation regions Sr(1) to Sr(Kmax). The irradiation conditions are, for example, those described with reference to FIG. 4. The irradiation conditions set in the above-mentioned S140 are the irradiation conditions corresponding to the irradiation region Sr(k) among the irradiation conditions read in S102. After S102, the laser annealing processor 100 ends the process of this flowchart, and returns to the process shown in FIG.

[0051] 1.2.5.3 Control of the laser annealing device 14 and the laser device 12 Fig. 7 is a flowchart showing details of the process for controlling the laser annealing apparatus 14 and the laser apparatus 12 shown in Fig. 5. The flowchart shown in Fig. 7 corresponds to the subroutine of S150 in Fig. 5.

[0052] In S151, the laser annealing processor 100 controls the attenuator 120 based on the target value Ft(k) of the fluence and the pulse energy of the pulsed laser light output from the laser device 12. As the pulse energy of the pulsed laser light, the target pulse energy Et set in the laser device 12 is used. The laser annealing processor 100 controls the attenuator 120 so that the transmittance Ta of the attenuator 120 becomes the value calculated as described above.

[0053] In S152, the laser annealing processor 100 controls the position of the XYZ stage 174 in the X-axis direction and the Y-axis direction so that the pulsed laser light is irradiated onto the irradiation region Sr(k), and stops the XYZ stage 174 at that position.

[0054] In S153, the laser annealing processor 100 controls the position of the XYZ stage 174 in the Z axis so that an image of the mask 140 is formed on the surface of the substrate SUB, and stops the XYZ stage 174 at that position.

[0055] In S154, the laser annealing processor 100 transmits a trigger signal of Np(k) pulses at a repetition frequency f to the laser device 12. As a result, while the XYZ stage 174 is stopped, a pulsed laser beam of Np(k) pulses is output from the laser device 12 and irradiated onto the irradiation region Sr(k), thereby performing laser annealing. After S154, the laser annealing processor 100 ends the process of this flowchart, and returns to the process shown in FIG.

[0056] 1.3 Issues with the Comparative Example In the comparative example, the amorphous silicon film is irradiated with pulsed laser light to perform laser annealing, and the amorphous silicon film is turned into a polycrystalline silicon film, thereby improving the semiconductor characteristics. However, various factors such as the film thickness and crystal grain size of the polycrystalline silicon film are related to the semiconductor characteristics, and the irradiation conditions of the pulsed laser light required to obtain the target semiconductor characteristics may differ depending on the case. In addition, in order to measure the semiconductor characteristics, it is necessary to move the substrate SUB to a measuring instrument such as a device analyzer. Furthermore, in some cases, the semiconductor characteristics cannot be measured until the polycrystalline silicon film is processed to fabricate a semiconductor device. For this reason, it is difficult to measure the semiconductor characteristics during the laser annealing, and it is difficult to adjust the irradiation conditions of the pulsed laser light.

[0057] Laser annealing of IGZO and IZO differs from laser annealing of amorphous silicon in that it does not crystallize them, but laser annealing of IGZO and IZO has the same problems as laser annealing of amorphous silicon.

[0058] In some embodiments described below, image data of the irradiated region Sr(k) after laser annealing is used as input, and measurement data of the semiconductor characteristics is used as output to perform machine learning using teacher data associated with the image data, generating a trained model. The trained model is used to calculate an output for new image data, thereby calculating an estimated value of the semiconductor characteristics.

[0059] 2. Machine Learning Systems 2.1 Configuration 8 illustrates a schematic configuration of a machine learning system according to the first embodiment. The machine learning system includes a laser annealing system 10, an image data generating unit 20, a semiconductor characteristic measuring unit 40, and a machine learning processor 500.

[0060] 2.1.1 Laser annealing system 10 The laser annealing processor 100 included in the laser annealing system 10 receives setting data of irradiation conditions and an instruction to execute laser annealing from the machine learning processor 500. The laser annealing processor 100 also notifies the machine learning processor 500 of the completion of the operation of a work transport robot (not shown) and the completion of laser annealing. The laser annealing system 10 executes laser annealing on a substrate SUB for an annealing sample.

[0061] In other respects, the configuration of the laser annealing system 10 may be similar to that described with reference to Fig. 1. Note that in Fig. 8, some components of the laser annealing system 10 are omitted from the illustration.

[0062] 2.1.2 Image data generating unit 20 The image data generating section 20 includes an image data generating processor 200 , a transfer optical system 245 , an XYZ stage 274 , and an image generating unit 280 .

[0063] The XYZ stage 274 is fixed to a frame (not shown). A table 276 of the XYZ stage 274 supports a substrate SUB for use as an annealing sample that is laser annealed.

[0064] The transfer optical system 245 is disposed in an optical path between the image generating unit 280 and the substrate SUB. The transfer optical system 245 may be a reduction projection optical system configured by a combination of a plurality of lenses 243 and 244.

[0065] The image generating unit 280 includes an illumination light source 281 , an imaging section 282 , and a half mirror 283 . The illumination light source 281 is configured to output illumination light toward the half mirror 283. The illumination light includes a plurality of wavelength components of visible light. For example, the illumination light is white light. A part of the illumination light is reflected by the half mirror 283, passes through the transfer optical system 245, is irradiated onto the substrate SUB, and is reflected by the substrate SUB. The imaging unit 282 includes an image sensor. The imaging unit 282 receives a portion of the reflected light of the illumination light via the transfer optical system 245 and the half mirror 283. An image of the surface of the substrate SUB is formed on the light receiving surface of the imaging unit 282 by the transfer optical system 245. The imaging unit 282 generates and outputs image data from the image formed on the light receiving surface.

[0066] The image data generation processor 200 is configured to control the XYZ stage 274, the illumination light source 281, and the imaging unit 282. The image data generation processor 200 receives image data from the imaging unit 282. The image data generation processor 200 controls a work transport robot (not shown) for setting the substrate SUB on the table 276 of the XYZ stage 274.

[0067] The image data generation processor 200 receives an instruction to generate image data from the machine learning processor 500. In addition, the image data generation processor 200 notifies the machine learning processor 500 of the completion of the operation of the work transport robot and the completion of the generation of image data, and transmits image data.

[0068] 2.1.3 Semiconductor characteristic measuring unit 40 The semiconductor characteristic measurement section 40 includes a semiconductor characteristic measurement processor 400 , a device analyzer 430 , and an XYZ stage 474 .

[0069] The XYZ stage 474 is fixed to a frame (not shown). A substrate SUB on which a semiconductor device is fabricated as an annealing sample is supported on a table 476 of the XYZ stage 474. The semiconductor device is fabricated by the semiconductor manufacturing apparatus 30.

[0070] The device analyzer 430 brings a probe 431 into contact with a terminal of a semiconductor element formed on the surface of the substrate SUB, measures the semiconductor characteristics of the semiconductor element, and outputs the results. The semiconductor characteristics include, for example, electron mobility. Alternatively, the semiconductor characteristics include a gate threshold voltage.

[0071] The semiconductor characteristic measurement processor 400 is configured to control the XYZ stage 474 and the device analyzer 430. The semiconductor characteristic measurement processor 400 receives measurement data of semiconductor characteristics from the device analyzer 430. The semiconductor characteristic measurement processor 400 controls a work transport robot (not shown) for setting the substrate SUB on a table 476 of the XYZ stage 474.

[0072] The semiconductor characteristic measurement processor 400 receives an instruction to measure semiconductor characteristics from the machine learning processor 500. In addition, the semiconductor characteristic measurement processor 400 notifies the machine learning processor 500 of the completion of the operation of the work transport robot and the completion of the measurement of the semiconductor characteristics, and transmits measurement data.

[0073] 2.1.4 Machine Learning Processor 500 The machine learning processor 500 is a processing device including a memory 501 in which a control program is stored, and a CPU 502 that executes the control program. The machine learning processor 500 is specially configured or programmed to execute various processes included in the present disclosure.

[0074] 2.2 How Machine Learning Works 2.2.1 Main flow 9 is a flowchart showing the operation of the machine learning processor 500 in the first embodiment. In the first embodiment, machine learning is performed in the following manner to generate a trained model.

[0075] In S10, the machine learning processor 500 causes the laser annealing system 10 to perform laser annealing and prepare an annealed sample. The laser annealing is performed by irradiating a pulsed laser light onto a semiconductor film on a substrate SUB for the annealed sample. Details of S10 will be described later with reference to FIG.

[0076] In S20, the machine learning processor 500 causes the image data generating unit 20 to generate image data, and acquires image data of each of the irradiation regions Sr(1) to Sr(Kmax) of the annealed sample from the image data generating unit 20. The image data is generated from the reflected light of the illumination light irradiated onto the annealed sample. Details of S20 will be described later with reference to FIG.

[0077] In S30, the machine learning processor 500 determines whether the semiconductor process by the semiconductor manufacturing equipment 30 has been completed and a semiconductor device has been fabricated. The semiconductor process includes processes for forming resist films and other films, etching processes, ion implantation processes, and the like. If the semiconductor process has not been completed (S30: NO), the machine learning processor 500 waits until the semiconductor process is completed. If the semiconductor process is completed (S30: YES), the machine learning processor 500 proceeds to S40.

[0078] In S40, the machine learning processor 500 causes the semiconductor characteristic measuring unit 40 to measure the semiconductor characteristics, and acquires measurement data of the semiconductor characteristics of each of the irradiation regions Sr(1) to Sr(Kmax) of the annealed sample from the semiconductor characteristic measuring unit 40. Details of S40 will be described later with reference to FIG.

[0079] In S50, the machine learning processor 500 generates teacher data in which image data is input and measurement data of semiconductor characteristics is output in association with each other. Details of S50 will be described later with reference to FIGS. 13 and 14.

[0080] In S60, the machine learning processor 500 performs machine learning using a neural network based on the training data to generate a trained model. Details of S60 will be described later with reference to FIGS. 15 and 16. After S60, the machine learning processor 500 ends the processing of this flowchart.

[0081] 2.2.2 Preparation of annealed samples 10 is a flowchart showing details of a process for producing an annealed sample by the laser annealing system 10. The flowchart shown in FIG. 10 corresponds to a subroutine of S10 in FIG.

[0082] In S11, the machine learning processor 500 transmits setting data of the irradiation conditions for preparing the annealing sample to the laser annealing processor 100 of the laser annealing system 10. The irradiation conditions for preparing the annealing sample may be the same as those described with reference to FIG. 4. However, the target values ​​Ft(1) to Ft(Kmax) of the fluence are, for example, 400 mJ / cm. 2 More than 700mJ / cm 2 It is desirable to set the values ​​within the following range so as not to be biased toward a specific value. Also, it is desirable to set the number of irradiation pulses Np(1) to Np(Kmax) within the range of, for example, 1 to 15 so as not to be biased toward a specific value.

[0083] In S12, the machine learning processor 500 determines whether or not the substrate SUB for the annealing sample has been placed on the table 176 of the laser annealing system 10, i.e., whether or not the operation of a workpiece transport robot (not shown) of the laser annealing system 10 has been completed. If the substrate SUB has not been placed on the table 176 (S12: NO), the machine learning processor 500 waits until the substrate SUB is placed on the table 176. If the substrate SUB has been placed on the table 176 (S12: YES), the machine learning processor 500 advances the process to S13.

[0084] In S13, the machine learning processor 500 transmits an instruction to execute laser annealing to the laser annealing processor 100, thereby causing the laser annealing system 10 to execute laser annealing.

[0085] In S14, the machine learning processor 500 judges whether or not the laser irradiation to all the irradiation regions Sr(1) to Sr(Kmax) of the substrate SUB for the annealing sample is completed, that is, whether or not the laser annealing is completed. If the laser irradiation to all the irradiation regions Sr(1) to Sr(Kmax) is not completed (S14: NO), the machine learning processor 500 waits until the laser irradiation to all the irradiation regions Sr(1) to Sr(Kmax) is completed. If the laser irradiation to all the irradiation regions Sr(1) to Sr(Kmax) is completed (S14: YES), the machine learning processor 500 ends the processing of this flowchart and returns to the processing shown in FIG. 9.

[0086] 2.2.3 Acquisition of image data 11 is a flowchart showing details of the process of acquiring image data by the image data generating unit 20. The flowchart shown in FIG. 11 corresponds to a subroutine of S20 in FIG.

[0087] In S22, the machine learning processor 500 determines whether or not the substrate SUB for the annealing sample has been placed on the table 276 of the image data generating unit 20, i.e., whether or not the operation of the work transport robot (not shown) of the image data generating unit 20 has been completed. If the substrate SUB has not been placed on the table 276 (S22: NO), the machine learning processor 500 waits until the substrate SUB is placed on the table 276. If the substrate SUB has been placed on the table 276 (S22: YES), the machine learning processor 500 advances the process to S23.

[0088] In S23, the machine learning processor 500 transmits an instruction to generate image data to the image data generation processor 200, thereby causing the image data generation unit 20 to generate image data.

[0089] In S24, the machine learning processor 500 judges whether or not the generation of image data of all the irradiation regions Sr(1) to Sr(Kmax) of the substrate SUB for the annealing sample has been completed. If the generation of image data of all the irradiation regions Sr(1) to Sr(Kmax) has not been completed (S24: NO), the machine learning processor 500 waits until the generation of image data of all the irradiation regions Sr(1) to Sr(Kmax) is completed. If the generation of image data of all the irradiation regions Sr(1) to Sr(Kmax) has been completed (S24: YES), the machine learning processor 500 advances the process to S25.

[0090] In S25, the machine learning processor 500 receives image data from the image data generating unit 20. After S25, the machine learning processor 500 ends the processing of this flowchart, and returns to the processing shown in FIG.

[0091] 2.2.4 Acquisition of measurement data of semiconductor characteristics 12 is a flowchart showing details of the process of acquiring measurement data of semiconductor characteristics by the semiconductor characteristic measuring unit 40. The flowchart shown in FIG. 12 corresponds to a subroutine of S40 in FIG.

[0092] In S42, the machine learning processor 500 determines whether or not the substrate SUB for the anneal sample has been placed on the table 476 of the semiconductor characteristic measuring unit 40, i.e., whether or not the operation of the work transport robot (not shown) of the semiconductor characteristic measuring unit 40 has been completed. If the substrate SUB has not been placed on the table 476 (S42: NO), the machine learning processor 500 waits until the substrate SUB is placed on the table 476. If the substrate SUB has been placed on the table 476 (S42: YES), the machine learning processor 500 advances the process to S43.

[0093] In S43, the machine learning processor 500 transmits an instruction to measure the semiconductor characteristics to the semiconductor characteristics measurement processor 400, thereby causing the semiconductor characteristics measurement unit 40 to measure the semiconductor characteristics.

[0094] In S44, the machine learning processor 500 determines whether or not the measurement of the semiconductor characteristics of all the irradiation regions Sr(1) to Sr(Kmax) of the substrate SUB has been completed. If the measurement of the semiconductor characteristics of all the irradiation regions Sr(1) to Sr(Kmax) has not been completed (S44: NO), the machine learning processor 500 waits until the measurement of the semiconductor characteristics of all the irradiation regions Sr(1) to Sr(Kmax) has been completed. If the measurement of the semiconductor characteristics of all the irradiation regions Sr(1) to Sr(Kmax) has been completed (S44: YES), the machine learning processor 500 advances the process to S45.

[0095] In S45, the machine learning processor 500 receives measurement data of the semiconductor characteristics from the semiconductor characteristic measuring section 40. After S45, the machine learning processor 500 ends the processing of this flowchart and returns to the processing shown in FIG.

[0096] 2.2.5 Generating training data 13 is a flowchart showing details of the process for generating teacher data. The flowchart shown in FIG. 13 corresponds to the subroutine of S50 in FIG.

[0097] In S51, the machine learning processor 500 stores the image data received from the image data generating unit 20 in the memory 501 in association with the irradiation regions Sr(1) to Sr(Kmax).

[0098] In S52, the machine learning processor 500 stores the measurement data of the semiconductor characteristics received from the semiconductor characteristic measuring unit 40 in the memory 501 in association with the irradiation regions Sr(1) to Sr(Kmax). After S52, the machine learning processor 500 ends the processing of this flowchart and returns to the processing shown in FIG.

[0099] Fig. 14 shows an example of training data generated by the process of Fig. 13. The training data is obtained by associating image data and measurement data of semiconductor characteristics with each of the irradiation regions Sr(1) to Sr(Kmax). Fig. 14 shows an example in which electron mobilities Em(1) to Em(Kmax) are measured as the semiconductor characteristics. Alternatively, the semiconductor characteristic may be a gate threshold voltage.

[0100] The training data may be created separately for each semiconductor film forming condition and for each mask 140 used for laser annealing. When the pulsed laser light whose light intensity distribution is spatially uniformized by the shaping optical system 130 is irradiated onto the irradiation region Sr(k) without using the mask 140, the training data may be created without using the mask 140.

[0101] 2.2.6 Machine Learning 15 is a flowchart showing details of the process of generating a trained model by performing machine learning. The flowchart shown in FIG. 15 corresponds to a subroutine of S60 in FIG.

[0102] In S61, the machine learning processor 500 reads the training data from the memory 501. The training data is the one described with reference to FIGS.

[0103] In S62, the machine learning processor 500 distributes the teacher data so that a part of the teacher data is used as learning data and the other part is used as evaluation data. For example, of the Kmax data sets associated with the irradiation areas Sr(1) to Sr(Kmax), 90% is used as learning data and 10% is used as evaluation data. This distribution is performed randomly except for the distribution ratio. In this disclosure, "random" includes the use of pseudo-random numbers created by a computer according to a certain algorithm. Also, if the irradiation conditions for the irradiation areas Sr(1) to Sr(Kmax) are set randomly, the distribution operation does not have to be random. For example, the teacher data associated with the irradiation areas Sr(1) to Sr(Kmax / 10) may be used as evaluation data.

[0104] In S63, the machine learning processor 500 sets parameters for a deep structure. A deep structure is a type of neural network, and is a structure in which functions that mimic the actions of nerve cells are arranged in multiple layers.

[0105] Figure 16 shows an example of a deep structure. When image data is input to the deep structure, the functions of each layer receive data from the previous layer and output data generated according to each function to the subsequent layer. These functions perform processes such as convolution, pooling, smoothing, and full connection.

[0106] Referring again to FIG. 15, in S64, the machine learning processor 500 trains the deep structure using the learning data. Specifically, image data included in the learning data is input to the deep structure. Then, the values ​​of the coefficients included in each function are determined so that the sum of errors between the data output to the final stage of the deep structure and the measurement data of the semiconductor characteristics included in the learning data is minimized. The deep structure trained in this manner is referred to as a trained model in this disclosure.

[0107] In S65, the machine learning processor 500 evaluates the trained model using the evaluation data. Specifically, image data included in the evaluation data is input to the trained model. Then, an error between data output to the final stage of the trained model and measurement data of semiconductor characteristics included in the evaluation data is calculated.

[0108] In S66, the machine learning processor 500 determines whether the accuracy of the trained model is sufficient. For example, the accuracy of the trained model is determined based on the average or variance of the error calculated in S65. If the accuracy of the trained model is not sufficient (S66: NO), the machine learning processor 500 returns to the process of S63, resets the parameters, and restarts the training of the deep structure. If the accuracy of the trained model is sufficient (S66: YES), the machine learning processor 500 ends the process of this flowchart and returns to the process shown in FIG. 9.

[0109] If it is determined that the accuracy of the trained model is sufficient, image data whose semiconductor characteristics are unknown can be input into the trained model, and the output of the final stage of the trained model can be used as an estimate of the semiconductor characteristics.

[0110] FIG. 17 is a graph showing the evaluation results of the trained model using the evaluation data in S65 of FIG. 15. Electron mobility is adopted as the semiconductor characteristic. The horizontal axis of FIG. 17 shows the estimated value of electron mobility output from the trained model by inputting image data included in the evaluation data to the trained model. The vertical axis shows the measured electron mobility included as measurement data in the evaluation data. As shown in FIG. 17, although there was some error between the estimated and measured electron mobilities, they generally agreed well, and it was found that the method can be used sufficiently to estimate the electron mobility.

[0111] 2.3 Effect (1) In the machine learning method according to the first embodiment, image data is acquired from reflected light of illumination light irradiated onto an irradiation region Sr(k) of a semiconductor film on a substrate SUB that has been laser annealed with pulsed laser light. Measurement data of the semiconductor characteristics in the irradiation region Sr(k) is also acquired. Teacher data is generated in which the image data is used as input and the measurement data is used as output. Machine learning is also performed using a neural network based on the teacher data to generate a trained model. This makes it possible to predict the semiconductor characteristics based on the image data of the irradiation region Sr(k) whose semiconductor characteristics are unknown, and makes it possible to adjust the irradiation conditions of the pulsed laser light even while laser annealing is being performed.

[0112] (2) In the machine learning method according to the first embodiment, after image data is acquired, a semiconductor process is performed to fabricate a semiconductor device, and then measurement data is acquired. According to this, even if a semiconductor characteristic cannot be measured until a semiconductor device is fabricated, it is possible to predict it based on the image data of the irradiation region Sr(k). Examples of the semiconductor characteristic that cannot be measured until a semiconductor device is fabricated include electron mobility and gate threshold voltage. Other examples of the semiconductor characteristic include contact resistance.

[0113] (3) The machine learning method of the first embodiment further includes evaluating the trained model, generating the trained model using a portion of the teacher data as training data, and evaluating the trained model using another portion of the teacher data as evaluation data. According to this, since evaluation data that has not been used as training data is used, it is possible to appropriately evaluate the trained model. In other respects, the first embodiment is similar to the comparative example.

[0114] 3. Laser annealing system 10b that estimates electron mobility and corrects irradiation conditions 3.1 Configuration 18 is a schematic diagram showing the configuration of a laser annealing system 10b according to the second embodiment. The laser annealing system 10b includes an image generating unit 180 in addition to the configuration of the laser annealing system 10 according to the comparative example, and includes a beam splitter 115b instead of the high-reflection mirror 115. The memory 101 stores a trained model in which image data is input and an estimated value Eme(k) of electron mobility is output, and the laser annealing processor 100 is able to access the trained model.

[0115] The image generating unit 180 includes an illumination light source 181, an imaging section 182, and a half mirror 183. These components are similar to the components of the image generating unit 280 in the first embodiment.

[0116] The beam splitter 115b is a dichroic mirror that transmits one of the visible illumination light output from the illumination light source 181 and the ultraviolet pulsed laser light output from the laser device 12 and reflects the other. In the example shown in Fig. 18, the beam splitter 115b is configured to transmit the illumination light and reflect the pulsed laser light. Alternatively, a high-reflection mirror 115 similar to that in the comparative example may be disposed in place of the beam splitter 115b, and the illumination light output from the illumination light source 181 may be configured to be incident on the substrate SUB at an angle of incidence different from that of the pulsed laser light.

[0117] A portion of the illumination light output from the illumination light source 181 is reflected by the half mirror 183, passes through the beam splitter 115b and the projection optical system 145, is irradiated onto the irradiation region Sr(k) of the substrate SUB, and is reflected by the substrate SUB. The imaging unit 182 receives a portion of the reflected light of the illumination light via the projection optical system 145, the beam splitter 115b, and the half mirror 183, and generates image data. The projection optical system 145 may include a combined lens of calcium fluoride and synthetic quartz in order to reduce chromatic aberration between the ultraviolet pulsed laser light and the visible illumination light.

[0118] 19 illustrates a schematic configuration of the imaging unit 182. The imaging unit 182 includes a shutter 184, an image sensor 185, a drive circuit 186, and an imaging controller 187.

[0119] The shutter 184 opens and closes in response to a shutter trigger signal received from the imaging controller 187, and in the closed state, blocks the illumination light reflected by the substrate SUB. This adjusts the timing and duration of the reflected light incident on the image sensor 185. The image sensor 185 includes charge-coupled devices (CCDs) arranged in a matrix, for example, and captures an image of the surface of the substrate SUB. The drive circuit 186 reads out image data by detecting the charge of each charge-coupled device. The imaging controller 187 controls the detection of the charge by the drive circuit 186 and processes the image data read out by the drive circuit 186. The imaging controller 187 receives a control signal from the laser annealing processor 100 and transmits image data to the laser annealing processor 100. Alternatively, the imaging controller 187 may be included in the laser annealing processor 100. The image sensor 185 may use a complementary metal oxide semiconductor (CMOS) transistor instead of a charge-coupled device.

[0120] 3.2 Operation 3.2.1 Main flow 20 is a flowchart showing the operation of the laser annealing processor 100 in the second embodiment. In the second embodiment, laser annealing is performed by the step-and-repeat method in the following manner. In S100b, the laser annealing processor 100 performs an initial adjustment. Details of S100b will be described later with reference to FIG. The processes from S130 to S150 are similar to those described in the comparative example.

[0121] In S170b, the laser annealing processor 100 acquires image data of the laser annealed irradiation region Sr(k). In S150, the XYZ stage 174 is stopped, and the pulsed laser light of Np(k) pulses is irradiated onto the irradiation region Sr(k), and then, by the time the process of S170b is performed, the XYZ stage 174 is not yet moved. The laser annealing processor 100 controls the illumination light source 181 and the imaging unit 182 to acquire image data of the irradiation region Sr(k).

[0122] In S180b, the laser annealing processor 100 calculates an estimated value Eme(k) of the electron mobility using the image data and the trained model. Details of S180b will be described later with reference to FIG.

[0123] In S190b, the laser annealing processor 100 updates the pulse laser light irradiation conditions based on the estimated value Eme(k) of the electron mobility. The update of the irradiation conditions includes updating the irradiation conditions without changing them and updating the irradiation conditions after correcting them. Details of S190b will be described later with reference to Figs. 23 to 29.

[0124] The processes of S200 and S210 are the same as those described in the comparative example. The next irradiation area Sr(k+1) is irradiated with the pulsed laser light after 1 is added to the value of k. Therefore, the irradiation conditions are updated in S190b before the next irradiation area Sr(k+1) is irradiated with the pulsed laser light. The next irradiation area Sr(k+1) corresponds to the second area in this disclosure.

[0125] 3.2.2 Initial adjustment Fig. 21 is a flowchart showing details of the initial adjustment process shown in Fig. 20. The flowchart shown in Fig. 21 corresponds to the subroutine of S100b in Fig. 20. The process of S101 is the same as in the comparative example.

[0126] In S102b, the laser annealing processor 100 reads the setting data of the irradiation conditions of the irradiation region Sr(1) and the target value Emt of the electron mobility. The irradiation conditions of the irradiation region Sr(1) include the target value Ft(1) of the fluence and the number of irradiation pulses Np(1). The irradiation conditions of the irradiation region Sr(2) and after are set in S190b. After S102b, the laser annealing processor 100 ends the processing of this flowchart, and returns to the processing shown in FIG.

[0127] 3.2.3 Calculation of the estimated electron mobility Eme(k) Fig. 22 is a flowchart showing details of the process for calculating the estimated value Eme(k) of electron mobility shown in Fig. 20. The flowchart shown in Fig. 22 corresponds to the subroutine of S180b in Fig. 20.

[0128] In S181, the laser annealing processor 100 reads the trained model from the memory 101. In S182, the laser annealing processor 100 inputs the image data acquired in S170b into the trained model. In S183b, the laser annealing processor 100 calculates an estimate of the electron mobility Eme(k). After S183b, the laser annealing processor 100 ends the processing of this flowchart, and returns to the processing shown in FIG.

[0129] 3.2.4 Update of irradiation conditions Fig. 23 is a flowchart showing details of the process of updating the irradiation conditions shown in Fig. 20. The flowchart shown in Fig. 23 corresponds to the subroutine of S190b in Fig. 20.

[0130] In S191b, the laser annealing processor 100 calculates the difference ΔEm(k) between the estimated value Eme(k) of the electron mobility calculated in FIG. 22 and the target value Emt of the electron mobility read in FIG. 21 by the following formula. ΔEm(k)=Eme(k)-Emt

[0131] In S192b, the laser annealing processor 100 determines whether or not the absolute value of the difference ΔEm(k) calculated in S191b is equal to or smaller than a first threshold value Emtr. If the absolute value of the difference ΔEm(k) is equal to or less than the first threshold value Emtr (S192b: YES), the laser annealing processor 100 advances the process to S193. In S193, the laser annealing processor 100 sets the current irradiation conditions, fluence target value Ft(k) and irradiation pulse number Np(k), as the next irradiation conditions Ft(k+1) and Np(k+1), respectively. In other words, the irradiation conditions are not changed. When the absolute value of the difference ΔEm(k) is greater than the first threshold value Emtr (S192b: NO), the laser annealing processor 100 advances the process to S194b. In S194b, the laser annealing processor 100 corrects the irradiation conditions of the pulsed laser light based on the estimated value Eme(k) of the electron mobility. Details of the process of S194b will be described later with reference to FIGS. 24 to 29.

[0132] After S193 or S194b, the laser annealing processor 100 ends the processing of this flowchart, and returns to the processing shown in FIG.

[0133] 3.2.4.1 First example of processing to correct illumination conditions Fig. 24 is a flowchart showing a first example of the process of correcting the irradiation conditions shown in Fig. 23. The flowchart shown in Fig. 24 corresponds to the subroutine of S194b in Fig. 23.

[0134] In S1946b, the laser annealing processor 100 judges whether the product of the change amount of the estimated value Eme(k) of the electron mobility and the change amount of the target value Ft(k) of the fluence is equal to or greater than 0. That is, it judges whether the following formula is satisfied. (Eme(k)-Eme(k-1))×(Ft(k)-Ft(k-1))≧0 If this formula is satisfied (S1946b: YES), the laser annealing processor 100 advances the process to S1947b. If this formula is not satisfied (S1946b: NO), the laser annealing processor 100 proceeds to S1948b.

[0135] FIG. 25 is a graph showing a typical relationship between the target value Ft(k) of the fluence and the estimated value Eme(k) of the electron mobility. The curve showing the relationship between the target value Ft(k) of the fluence and the estimated value Eme(k) of the electron mobility has an upwardly convex shape, and the target value Emt of the electron mobility is set near the maximum value of the estimated value Eme(k). If the estimated value Eme(k) of the electron mobility is equal to or greater than the value Emt-Emtr obtained by subtracting the first threshold value Emtr from the target value Emt and is equal to or less than the value Emt+Emtr obtained by adding the first threshold value Emtr to the target value Emt, the determination is YES in S192b, and the process of FIG. 24 is not performed. On the other hand, if the estimated value Eme(k) of the electron mobility is outside the above-mentioned range, the target value Ft(k) of the fluence is either in the low region L or in the high region H. When the target fluence value Ft(k) is in low region L, the relationship of the estimated electron mobility value Eme(k) to the target fluence value Ft(k) rises to the right. When the target fluence value Ft(k) is in high region H, the relationship of the estimated electron mobility value Eme(k) to the target fluence value Ft(k) falls to the right.

[0136] FIG. 26 is a table for explaining the criteria for determining the control direction of the target value Ft(k) of the fluence in the second embodiment. 26, the direction of change in the target fluence value Ft(k) indicates whether the amount of change Ft(k)-Ft(k-1) in the target fluence value Ft(k) is positive or negative. That is, if the target fluence value Ft(k) was increased in the previous update, the amount of change Ft(k)-Ft(k-1) will be a positive number, and if the target fluence value Ft(k) was decreased in the previous update, the amount of change Ft(k)-Ft(k-1) will be a negative number. The change direction of the electron mobility estimate Eme(k) indicates whether the change amount Eme(k)-Eme(k-1) of the electron mobility estimate Eme(k) is positive or negative. That is, if the electron mobility estimate Eme(k) is higher than before the previous update, the change amount Eme(k)-Eme(k-1) is a positive number, and if the electron mobility estimate Eme(k) is lower than before the previous update, the change amount Eme(k)-Eme(k-1) is a negative number.

[0137] If both of these variations are positive or both are negative, then the product of the variations is positive. If one of the variations is positive and the other is negative, then the product of the variations is negative. If the product of the changes is a positive number (S1946b: YES), the current target fluence value Ft(k) is in the low region L, so by setting the control direction of the target fluence value Ft(k) to positive, the estimated electron mobility value Eme(k) can be brought closer to the target value Emt. If the product of the changes is negative (S1946b: NO), the current target fluence value Ft(k) is in the high region H, so by making the control direction of the target fluence value Ft(k) negative, the estimated electron mobility value Eme(k) can be brought closer to the target value Emt.

[0138] Referring again to FIG. 24, in S1947b, the laser annealing processor 100 calculates the next fluence target value Ft(k+1) by adding a positive number ΔFt to the current fluence target value Ft(k). In S1948b, the laser annealing processor 100 calculates the next fluence target value Ft(k+1) by subtracting a positive number ΔFt from the current fluence target value Ft(k).

[0139] After S1947b or S1948b, the laser annealing processor 100 ends the processing of this flowchart, and returns to the processing shown in FIG. The first example of the process for correcting the irradiation conditions is carried out as described above.

[0140] 3.2.4.2 Second example of processing to correct illumination conditions Fig. 27 is a flowchart showing a second example of the process of correcting the irradiation conditions shown in Fig. 23. The flowchart shown in Fig. 27 corresponds to the subroutine of S194b in Fig. 23.

[0141] In S1941, the laser annealing processor 100 judges whether or not the absolute value of the difference ΔEm(k) calculated in S191b is equal to or smaller than a second threshold value Emtrmax. The second threshold value Emtrmax is a value larger than the first threshold value Emtr in S192b. When the absolute value of the difference ΔEm(k) is equal to or less than the second threshold Emtrmax (S1941: YES), the laser annealing processor 100 advances the process to S1946b. The process from S1946b onward is similar to that described with reference to FIG. When the absolute value of the difference ΔEm(k) is greater than the second threshold value Emtrmax (S1941: NO), the laser annealing processor 100 advances the process to S1942.

[0142] In S1942, the laser annealing processor 100 controls the XYZ stage 174 to move the substrate SUB so that the pulsed laser light is irradiated onto a trial irradiation area of ​​the substrate SUB. Alternatively, a substrate for trial irradiation (not shown) may be set on the table 176 instead of the substrate SUB.

[0143] In S1943, the laser annealing processor 100 obtains the relationship between the target value Ft(k) of the fluence and the estimated value Eme(k) of the electron mobility. Specifically, while changing the target value Ft(k) of the fluence according to each value of k, the corresponding irradiation region Sr(k) is trial-irradiated to perform laser annealing, image data is obtained each time, and the estimated value Eme(k) of the electron mobility is calculated.

[0144] In S1944, the laser annealing processor 100 determines the fluence Ftadj that minimizes the difference ΔEm(k) between the estimated electron mobility Eme(k) and the target value Emt, based on the relationship between the fluence target value Ft(k) and the electron mobility estimated value Eme(k).

[0145] In S1945, the laser annealing processor 100 sets the fluence Ftadj as the target value Ft(k+1) of the next fluence. After S1945, S1947b, or S1948b, the laser annealing processor 100 ends the processing of this flowchart, and returns to the processing shown in FIG. The second example of the process for correcting the irradiation conditions is carried out as described above.

[0146] According to the second example, even if the difference between the target fluence value Ft(k) and the actual fluence becomes large due to deterioration of the optical system included in the laser annealing apparatus 14, the relationship between the target fluence value Ft(k) and the estimated electron mobility value Eme(k) can be acquired and the target fluence value Ft(k) can be corrected. This correction can be performed based on the relationship between the target fluence value Ft(k) and the estimated electron mobility value Eme(k) without measuring the electron mobility with the device analyzer 430.

[0147] In S1941, the deterioration of the optical system may be determined by another method. For example, an energy monitor (not shown) may be disposed in the optical path of the pulsed laser light transmitted through the attenuator 120, and the processes of S1942 to S1945 may be performed when the output of the energy monitor decreases. Also, the processes of S1942 to S1945 may be performed when the cumulative number of pulses of the laser annealing by the laser annealing device 14 becomes equal to or greater than a predetermined value.

[0148] 3.2.4.3 Third example of processing to correct illumination conditions Fig. 28 is a flowchart showing a third example of the process of correcting the irradiation conditions shown in Fig. 23. The flowchart shown in Fig. 28 corresponds to the subroutine of S194b in Fig. 23.

[0149] In S1949, the laser annealing processor 100 determines whether the number of irradiation pulses Np(k) is less than a threshold value Nmax. If the number of irradiation pulses Np(k) is less than the threshold value Nmax (S1949: YES), the laser annealing processor 100 advances the process to S1950. If the number of irradiation pulses Np(k) is equal to or greater than the threshold value Nmax (S1949: NO), the laser annealing processor 100 advances the process to S1951.

[0150] FIG. 29 is a graph showing a typical relationship between the number of irradiation pulses Np(k) and the estimated value Eme(k) of electron mobility. If the estimated value Eme(k) of electron mobility is equal to or greater than the value Emt-Emtr obtained by subtracting the first threshold value Emtr from the target value Emt and is equal to or less than the value Emt+Emtr obtained by adding the first threshold value Emtr to the target value Emt, then YES is determined in S192b, and the process of FIG. 28 is not performed. On the other hand, if the estimated value Eme(k) of electron mobility is outside the above-mentioned range, the number of irradiation pulses Np(k) is either in the small region L or in the large region H. If the number of irradiation pulses Np(k) is in the small region L, the relationship of the estimated value Eme(k) of electron mobility to the number of irradiation pulses Np(k) rises to the right. When the number of irradiation pulses Np(k) is in a large region H, the relationship of the estimated value Eme(k) of the electron mobility to the number of irradiation pulses Np(k) drops sharply to the right.

[0151] In the region H, it may be difficult to determine the control direction of the number of irradiation pulses Np(k), and the time and cost required for laser annealing may increase because the number of irradiation pulses Np(k) is large. Therefore, a threshold value Nmax of the number of irradiation pulses Np(k) is set so that the number of irradiation pulses Np(k) does not reach the region H. If the number of irradiation pulses Np(k) is less than the threshold value Nmax (S1949: YES), the number of irradiation pulses Np(k) is in the small region L, so by setting the control direction of the number of irradiation pulses Np(k) to positive, the estimated electron mobility Eme(k) can be brought closer to the target value Emt.

[0152] Referring again to FIG. 28, in S1950, the laser annealing processor 100 calculates the next number of irradiation pulses Np(k+1) by adding 1 to the current number of irradiation pulses Np(k). In S1951, the laser annealing processor 100 sets the current number of irradiation pulses Np(k) as the next number of irradiation pulses Np(k+1). In other words, the irradiation conditions are not changed.

[0153] After S1950 or S1951, the laser annealing processor 100 ends the processing of this flowchart, and returns to the processing shown in FIG. The third example of the process for correcting the irradiation conditions is carried out as described above.

[0154] As a process for correcting the irradiation conditions, one of the first to third examples may be performed. Alternatively, a process for correcting the target value Ft(k) of the fluence and the number of irradiation pulses Np(k) may be performed by combining the first and third examples, or by combining the second and third examples.

[0155] 3.3 Effect (4) A laser annealing system 10b according to the second embodiment includes a laser device 12 that outputs a pulsed laser light for laser annealing an irradiation region Sr(k) of a semiconductor film on a substrate SUB, an illumination light source 181 that irradiates the irradiation region Sr(k) with illumination light, an imaging unit 182 that generates image data from reflected light of the illumination light, and a laser annealing processor 100. The laser annealing processor 100 is configured to be able to access a trained model that receives image data as input and outputs an estimated value Eme(k) of electron mobility, calculates the estimated value Eme(k) using the image data and the trained model, and corrects the irradiation conditions of the pulsed laser light based on the estimated value Eme(k). In the laser annealing method according to the second embodiment, an irradiation region Sr(k) of a semiconductor film on a substrate SUB is annealed with a pulsed laser beam. An illumination light is irradiated onto the irradiation region Sr(k), and image data is generated from the reflected light of the illumination light. An estimated value Eme(k) is calculated using a trained model in which the image data is input and an estimated value Eme(k) of electron mobility is output. An irradiation condition of the pulsed laser beam is corrected based on the estimated value Eme(k). This makes it possible to predict the electron mobility based on image data of the irradiation region Sr(k) whose semiconductor characteristics are unknown, and makes it possible to adjust the irradiation conditions of the pulsed laser light even during laser annealing.

[0156] (5) The laser annealing system 10b according to the second embodiment further includes a beam splitter 115b disposed in the optical path of the pulsed laser beam. The illumination light source 181 irradiates the illumination region Sr(k) with illumination light via the beam splitter 115b, and the imaging unit 182 receives reflected light of the illumination light via the beam splitter 115b. This allows irradiation of the pulsed laser light, irradiation of the illumination light by the illumination light source 181, and acquisition of image data by the imaging section 182 to be performed from approximately the same direction via the beam splitter 115b.

[0157] (6) In the laser annealing system 10b according to the second embodiment, the pulsed laser beam includes a first wavelength component of ultraviolet light, the illumination beam includes a plurality of second wavelength components of visible light, and the beam splitter 115b is a dichroic mirror that transmits one of the first and second wavelength components and reflects the other. This can reduce the loss of the pulsed laser light in the beam splitter 115b. Also, the pulsed laser light reflected by the substrate SUB can be prevented from entering the imaging section 182.

[0158] (7) The laser annealing system 10b according to the second embodiment further includes an XYZ stage 174 that moves the substrate SUB in the X-axis direction and the Y-axis direction intersecting the incidence direction of the pulsed laser light to the substrate SUB. The irradiation region Sr(k) is irradiated with the pulsed laser light of the set irradiation pulse number Np(k) while the XYZ stage 174 is stopped. Thereafter, the XYZ stage 174 moves the substrate SUB so that the next irradiation region Sr(k+1) different from the irradiation region Sr(k) is irradiated with the pulsed laser light. The laser annealing processor 100 corrects the irradiation conditions before the next irradiation region Sr(k+1) is irradiated with the pulsed laser light. According to this, the irradiation conditions can be corrected every time one irradiation region Sr(k) is laser annealed in the step-and-repeat method.

[0159] (8) In the laser annealing system 10b according to the second embodiment, the irradiation conditions include the fluence of the pulsed laser light, and the semiconductor characteristics include the electron mobility. The laser annealing processor 100 corrects the irradiation conditions to make the fluence higher when the estimated value Eme(k) of the electron mobility increases as a result of increasing the fluence, and corrects the irradiation conditions to make the fluence lower when the estimated value Eme(k) decreases as a result of increasing the fluence. Also, corrects the irradiation conditions to make the fluence lower when the estimated value Eme(k) increases as a result of decreasing the fluence, and corrects the irradiation conditions to make the fluence higher when the estimated value Eme(k) decreases as a result of decreasing the fluence. According to this, the control direction of the fluence (positive or negative) can be accurately determined from the relationship between the change direction (positive or negative) of the fluence and the change direction (positive or negative) of the estimated value Eme(k) of the electron mobility.

[0160] (9) In the laser annealing system 10b according to the second embodiment, when the absolute value of the difference ΔEm(k) between the estimated value Eme(k) of the electron mobility and the target value Emt is greater than the second threshold Emtrmax, the laser annealing processor 100 performs trial irradiation while changing the irradiation conditions, acquires image data, and calculates the estimated value Eme(k). As a result, the relationship between the irradiation conditions and the estimated value Eme(k) is acquired, and the irradiation conditions are corrected based on this relationship. According to this, even if the optical system of the laser annealing device 14 is deteriorated, the irradiation conditions can be appropriately corrected based on the relationship between the irradiation conditions and the estimated value Eme(k).

[0161] (10) In the laser annealing system 10b according to the second embodiment, the irradiation conditions include the number of irradiation pulses Np(k) of the pulsed laser light irradiated to one location on the substrate SUB. When the absolute value of the difference ΔEm(k) between the estimated value Eme(k) of the electron mobility and the target value Emt is greater than a first threshold value Emtr, the laser annealing processor 100 corrects the irradiation conditions to increase the number of irradiation pulses Np(k). According to this, by increasing the number of irradiation pulses Np(k), it is possible to bring the electron mobility closer to its target value Emt. In other respects, the second embodiment is similar to the comparative example.

[0162] 3.4 Other configuration examples The laser annealing system 10b in the second embodiment may be used instead of the laser annealing system 10 and the image data generating unit 20 included in the machine learning system in the first embodiment. That is, the laser annealing system 10b may be used to perform laser annealing of the substrate SUB for an annealing sample and acquisition of image data of the annealing sample by a step-and-repeat method.

[0163] 4. Laser annealing system 10b that estimates gate threshold voltage and corrects irradiation conditions 4.1 Configuration The third embodiment differs from the second embodiment in that the laser annealing processor 100 estimates the gate threshold voltage instead of estimating the electron mobility. The memory 101 included in the laser annealing processor 100 stores a learned model that receives image data as input and outputs an estimated value Vthe(k) of the gate threshold voltage. In other respects, the configuration of the third embodiment is similar to that of the second embodiment.

[0164] 4.2 Operation FIG. 30 is a flowchart showing the operation of the laser annealing processor 100 in the third embodiment. FIG. 31 is a flowchart showing details of the initial adjustment process shown in FIG. FIG. 32 is a flowchart showing details of the process of calculating the estimated value Vthe(k) of the gate threshold voltage shown in FIG. FIG. 33 is a flowchart showing details of the process of updating the irradiation conditions shown in FIG.

[0165] 30 to 33 are similar to the processes shown in Figures 20 to 23, except that the electron mobility in Figures 20 to 23 is replaced with the gate threshold voltage. Note that the target value of the gate threshold voltage is Vtht, the estimated value of the gate threshold voltage is Vthe(k), the difference between the estimated value Vthe(k) of the gate threshold voltage and the target value Vtht is ΔVth(k), and the first threshold is Vthtr.

[0166] Fig. 34 is a flowchart showing a first example of the process of correcting the irradiation condition shown in Fig. 33. The flowchart shown in Fig. 34 corresponds to the subroutine of S194c in Fig. 33. FIG. 35 is a graph showing a typical relationship between the target value of the fluence Ft(k) and the estimated value of the gate threshold voltage Vthe(k). FIG. 36 is a table for explaining the criteria for determining the control direction of the target value Ft(k) of the fluence in the third embodiment.

[0167] 35, the curve showing the relationship between the target value Ft(k) of the fluence and the estimated value Vthe(k) of the gate threshold voltage has a downwardly convex shape, and the target value Vtht of the gate threshold voltage is set near the minimum value of the estimated value Vthe(k). Therefore, the control direction of the target value Ft(k) of the fluence according to the positive or negative of the product of the amount of change is opposite to that of the second embodiment.

[0168] In FIG. 34, if the product of the change in the estimated value Vthe(k) of the gate threshold voltage and the change in the target value Ft(k) of the fluence is equal to or greater than 0 (S1946c: YES), the laser annealing processor 100 advances the process to S1947c. If the product of the amounts of change is less than 0 (S1946c: NO), the laser annealing processor 100 advances the process to S1948c.

[0169] In S1947c, the laser annealing processor 100 calculates the next fluence target value Ft(k+1) by subtracting a positive number ΔFt from the current fluence target value Ft(k). In S1948c, the laser annealing processor 100 calculates the next fluence target value Ft(k+1) by adding a positive number ΔFt to the current fluence target value Ft(k). The first example of the process for correcting the irradiation conditions is carried out as described above.

[0170] The second example of the process for correcting the irradiation conditions is similar to that of the second embodiment except for the following (1) and (2). (1) In order to determine the fluence Ftadj by the test irradiation from S1942 to S1945 in FIG. 27, the estimated value Vthe(k) and the target value Vtht of the gate threshold voltage are used instead of the estimated value Eme(k) and the target value Emt of the electron mobility. (2) Instead of the processes from S1946b to S1948b in FIG. 27, the processes from S1946c to S1948c in FIG. 34 are performed.

[0171] The third example of the process for correcting the irradiation conditions is the same as that of the second embodiment. That is, as in Fig. 28, every time the irradiation conditions are corrected, the next irradiation pulse number Np(k+1) is calculated by adding 1 to the current irradiation pulse number Np(k) (S1950). However, if the irradiation pulse number Np(k) reaches the threshold value Nmax, the current irradiation pulse number Np(k) is set as the next irradiation pulse number Np(k+1) (S1951).

[0172] 4.3 Effect (11) In the laser annealing system 10b according to the third embodiment, the irradiation conditions include the fluence of the pulsed laser light, and the semiconductor characteristics include the gate threshold voltage. The laser annealing processor 100 corrects the irradiation conditions to make the fluence higher when the estimated value Vthe(k) of the gate threshold voltage is lowered as a result of increasing the fluence, and corrects the irradiation conditions to make the fluence lower when the estimated value Vthe(k) is higher as a result of increasing the fluence. Also, corrects the irradiation conditions to make the fluence lower when the estimated value Vthe(k) is lowered as a result of decreasing the fluence, and corrects the irradiation conditions to make the fluence higher when the estimated value Vthe(k) is higher as a result of decreasing the fluence. According to this, the control direction of the fluence (positive or negative) can be accurately determined from the relationship between the change direction (positive or negative) of the fluence and the change direction (positive or negative) of the estimated value Vthe(k) of the gate threshold voltage. In other respects, the third embodiment is similar to the second embodiment.

[0173] 5. Step-and-scan laser annealing system 10b 5.1 Configuration The fourth embodiment differs from the second and third embodiments in that laser annealing is performed by a step-and-scan method instead of a step-and-repeat method. In a laser annealing system 10b according to the fourth embodiment, a mask 140 may have a rectangular opening, unlike the dot mask shown in Fig. 2. Alternatively, the mask 140 may not be used, and a pulsed laser beam having a rectangular beam cross section may be irradiated onto the substrate SUB.

[0174] 5.2 Operation Fig. 37 is a plan view showing an example of a laser annealing method for a substrate SUB by the step-and-scan method. In Fig. 37, the surface of the substrate SUB includes Mmax scan regions Sc(1) to Sc(Mmax). One of the scan regions Sc(1) to Sc(Mmax) is designated as Sc(m). m can take an integer value from 1 to Mmax.

[0175] 37 indicates the movement direction of the beam cross section B of the pulsed laser light relative to the substrate SUB. The position of the beam cross section B corresponds to the incidence position of the pulsed laser light on the substrate SUB. The beam cross section B moves by controlling the positions of the XYZ stage 174 in the X-axis direction and the Y-axis direction. For example, when the substrate SUB is moved in the -X direction by the XYZ stage 174, the beam cross section B moves relatively in the X direction.

[0176] The procedure for laser annealing using the step-and-scan method is as follows. First, the substrate SUB is moved so that the position of the beam cross section B coincides with the initial position In(1) away from the scan area Sc(1) in the -X direction. Next, the substrate SUB starts moving in the -X direction, and is accelerated to a speed -Vx until the beam cross section B reaches the scan area Sc(1). Then, transmission of a trigger signal with a repetition frequency f to the laser device 12 is started, and the scan area Sc(1) is laser annealed. When the laser annealing of the scan area Sc(1) is completed, transmission of the trigger signal is stopped, and the speed of the substrate SUB in the -X direction is decelerated.

[0177] Next, the substrate SUB is moved so that the position of the beam cross section B coincides with the initial position In(2) away from the scan area Sc(2) in the X direction. Next, the substrate SUB starts moving in the X direction, and is accelerated to a speed Vx until the beam cross section B reaches the scan area Sc(2). Then, transmission of a trigger signal with a repetition frequency f to the laser device 12 is started, and the scan area Sc(2) is laser annealed. When the laser annealing of the scan area Sc(2) is completed, transmission of the trigger signal is stopped, and the speed of the substrate SUB in the X direction is decelerated. Thereafter, the scan regions Sc(3) to Sc(Mmax) are laser annealed in the same manner.

[0178] The beam cross section B has a rectangular shape, and the lengths of its two sides are Bx and By. The number of irradiation pulses of the pulsed laser light irradiated to one location on the substrate SUB is Np. The number of irradiation pulses Np may be a constant value regardless of the irradiation position on the substrate SUB.

[0179] While the laser device 12 outputs a pulsed laser beam with a repetition frequency f and an irradiation pulse number Np, the substrate SUB moves at a speed Vx by a distance Bx. That is, the time required to irradiate the pulsed laser beam with a repetition frequency f and an irradiation pulse number Np is Np / f, and the time required for the beam cross section B to move by Bx at the speed Vx is Bx / Vx, and these required times are equal. Therefore, the speed Vx is set as follows. Vx=f Bx / Np

[0180] 38 and 39 are plan views showing an illumination range IL of the substrate SUB by the illumination light source 181. The illumination range IL of the illumination light by the illumination light source 181 includes the beam cross section B of the pulsed laser light on the substrate SUB and its periphery. The positional relationship between the illumination range IL and the beam cross section B is fixed. The dashed arrow indicates the movement direction of the illumination range IL and the beam cross section B relative to the substrate SUB. That is, when the substrate SUB moves in the -X direction, the illumination range IL and the beam cross section B move in the X direction relative to the substrate SUB as shown in FIG. 38, and when the substrate SUB moves in the X direction, the illumination range IL and the beam cross section B move in the -X direction relative to the substrate SUB as shown in FIG.

[0181] Within the illumination range IL, the areas that have already been laser annealed are indicated by diagonal lines. As the beam cross section B moves, areas that have not yet been laser annealed are sequentially laser annealed.

[0182] In the fourth embodiment, image data of the laser annealed region is sequentially acquired by the imaging unit 182 while the substrate SUB is moved by the XYZ stage 174. The region on the substrate SUB that corresponds to the position and size of the beam cross section B is an example of the first region in the present disclosure. When the substrate SUB is moved in the X direction, the first region also moves in the X direction, and image data is acquired when the first region overlaps with the image acquisition range IM.

[0183] The image acquisition range IM is located on the opposite side to the movement direction of the beam cross section B. That is, when the substrate SUB is moved in the X direction by the XYZ stage 174 and when it is moved in the -X direction, the imaging unit 182 acquires image data from a different position within the illumination range IL as the image acquisition range IM.

[0184] It is desirable that the image acquisition range IM is adjacent to the beam cross section B. Although Fig. 38 and Fig. 39 show a case where the dimensions of the image acquisition range IM are the same as the dimensions of the beam cross section B, the present disclosure is not limited thereto. The dimensions of the image acquisition range IM may be different from the dimensions of the beam cross section B. However, it is desirable that the dimensions of the image acquisition range IM are the same as the dimensions of the imaging target of the image data used to create the trained model.

[0185] In order to prevent the pulsed laser light from being captured when acquiring image data of the image acquisition range IM, it is desirable that the imaging unit 182 captures an image after the end of emission of one pulse of the pulsed laser light and before the start of emission of the next pulse. Therefore, it is desirable that the open time Ts of the shutter 184 satisfy the following condition. Ts<1 / f-τ Here, τ is the pulse time width of the pulsed laser light. Therefore, the open time Ts of the shutter 184 is shorter than the repetition period 1 / f of the pulsed laser light. The timing of the shutter trigger signal transmitted from the imaging controller 187 to the shutter 184 is controlled so that the opening and closing of the shutter 184 is completed after the emission of one pulse of the pulsed laser light ends and before the emission of the next pulse begins.

[0186] Furthermore, in order to suppress image blurring due to the movement of the substrate SUB, it is desirable that the open time Ts of the shutter 184 satisfy the following condition. Ts<Δx / Vx Here, Δx is the allowable limit of image blur.

[0187] 5.3 Operation of the Laser Annealing Processor 100 5.3.1 Main flow 40 is a flowchart showing the operation of the laser annealing processor 100 in the fourth embodiment. In the fourth embodiment, laser annealing by the step-and-scan method is performed in the following manner.

[0188] In S100d, the laser annealing processor 100 performs an initial adjustment. Details of S100d will be described later with reference to FIG.

[0189] In S110d, the laser annealing processor 100 sets the value of m to an initial value of 1. The value of m corresponds to that described with reference to FIG.

[0190] In S120d, the laser annealing processor 100 performs annealing setting for the scan area Sc(m) and starts laser annealing for the scan area Sc(m). Details of S120d will be described later with reference to FIG.

[0191] In S130, the laser annealing processor 100 sets the value of k to an initial value of 1. As described below, in the fourth embodiment, while moving the beam cross section B in the scan area Sc(m), image data acquisition (S170d), calculation of the estimated value Eme(k) of electron mobility (S180b), and update of the irradiation conditions (S190d) are performed once each, and the value of k is updated each time these are performed. During laser annealing of one scan area Sc(m), the value of k is updated multiple times.

[0192] In S160d, the laser annealing processor 100 counts the trigger signal of Np+n pulses. Assuming that the processes from S170d to S200d can be completed in a short time, the time for Np+n pulses approximately corresponds to the update period of the irradiation conditions. n is an integer equal to or greater than 0. Details of S160d will be described later with reference to FIG.

[0193] In S170d, the laser annealing processor 100 acquires image data of the laser annealed image acquisition area IM. Details of S170d will be described later with reference to FIG.

[0194] In S180b, the laser annealing processor 100 calculates an estimated value Eme(k) of the electron mobility using the image data and the trained model. The process of S180b is similar to that of the second embodiment.

[0195] In S190d, the laser annealing processor 100 updates the pulse laser light irradiation conditions based on the estimated value Eme(k) of the electron mobility. In S190d, the process of updating the target value Ft(k) of the fluence as the irradiation condition is the same as in the second embodiment, and for example, either the first example or the second example described above can be applied. However, in the fourth embodiment, the number of irradiation pulses Np does not need to be updated.

[0196] In S200d, the laser annealing processor 100 determines whether or not the laser annealing of the scan area Sc(m) has been completed. If the laser annealing of the scan area Sc(m) is not completed (S200d: NO), the laser annealing processor 100 advances the process to S210. In S210, the laser annealing processor 100 adds 1 to the value of k to update the value of k. After that, the laser annealing processor 100 returns the process to S160d, and after outputting the Np+n pulses, acquires image data (S170d), calculates the estimated value Eme(k) of the electron mobility (S180b), and updates the irradiation conditions (S190d). If the laser annealing of the scan area Sc(m) is completed (S200d: YES), the laser annealing processor 100 advances the process to S220d.

[0197] In S220d, the laser annealing processor 100 judges whether or not the laser annealing of the scan area Sc(Mmax) is completed, that is, whether or not the value of m is equal to or greater than Mmax. If the laser annealing of the scan area Sc(Mmax) is not completed (S220d: NO), the laser annealing processor 100 proceeds to S230d. In S230d, the laser annealing processor 100 adds 1 to the value of m to update the value of m. After that, the laser annealing processor 100 returns the process to S120d. When the laser annealing of the scan area Sc(Mmax) is completed (S220d: YES), the laser annealing processor 100 ends the processing of this flowchart.

[0198] 5.3.2 Initial adjustment Fig. 41 is a flowchart showing details of the initial adjustment process shown in Fig. 40. The flowchart shown in Fig. 41 corresponds to the subroutine of S100d in Fig. 40. The process of S101 is similar to that in the comparative example and the second embodiment.

[0199] In S102d, the laser annealing processor 100 reads the target value Ft(1) of the fluence, the number of irradiation pulses Np, and the target value Emt of the electron mobility. In S103d, the laser annealing processor 100 controls the position of the XYZ stage 174 in the Z axis so that an image of the mask 140 is formed on the surface of the substrate SUB, and stops the XYZ stage 174 at that position. After S103d, the laser annealing processor 100 ends the processing of this flowchart, and returns to the processing shown in FIG.

[0200] 5.3.3 Annealing setting for scan area Sc(m) and annealing start Fig. 42 is a flowchart showing details of the annealing setting and annealing start processing of the scan area Sc(m) shown in Fig. 40. The flowchart shown in Fig. 42 corresponds to the subroutine of S120d in Fig. 40.

[0201] In S121, the laser annealing processor 100 sets a target value Ft(m) of the fluence and the number of irradiation pulses Np. When the value of m is 1, the target value Ft(m) of the fluence is the target value Ft(1) of the fluence read in S102d. When the value of m is 2 or more, the target value Ft(m) of the fluence is the final target value Ft(k) of the fluence in the previous scan region Sc(m-1).

[0202] In S122, the laser annealing processor 100 controls the attenuator 120 based on the target value Ft(m) of the fluence and the pulse energy of the pulsed laser light output from the laser device 12.

[0203] In S123, the laser annealing processor 100 controls the position of the XYZ stage 174 in the X-axis direction and the Y-axis direction so that the beam cross section B of the pulsed laser light is positioned at the initial position In(m) of the scan region Sc(m).

[0204] In S124, the laser annealing processor 100 sets the scanning direction of the beam cross section B to the positive direction if m is an odd number, and to the negative direction if m is an even number. The scanning direction is the X direction or −X direction indicated by the dashed arrows in FIGS. 37 to 39.

[0205] In S125, the laser annealing processor 100 starts moving the substrate SUB by the XYZ stage 174, and accelerates it to a velocity Vx or −Vx.

[0206] In S126, the laser annealing processor 100 starts transmitting a trigger signal to the laser device 12 at a repetition frequency f. As a result, while moving the substrate SUB, the pulsed laser light output from the laser device 12 is irradiated onto the scan area Sc(m) to perform laser annealing. Note that when the laser annealing of the scan area Sc(m) is completed (S200d: YES), the movement of the substrate SUB in the X direction or −X direction by the XYZ stage 174 may be stopped, and the transmission of the trigger signal may be stopped until the laser annealing of the next scan area Sc(m+1) is started.

[0207] After S126, the laser annealing processor 100 ends the processing of this flowchart, and returns to the processing shown in FIG.

[0208] 5.3.4 Counting Np+n Pulses Fig. 43 is a flowchart showing the details of the process of counting Np+n pulses shown in Fig. 40. The flowchart shown in Fig. 43 corresponds to the subroutine of S160d in Fig. 40.

[0209] In S161, the laser annealing processor 100 sets the value of a counter Nc for counting the number of pulses to an initial value of 0.

[0210] In S162, the laser annealing processor 100 determines whether or not a trigger signal has been sent to the laser device 12. If a trigger signal has not been sent (S162: NO), the laser annealing processor 100 waits until a trigger signal is sent. If a trigger signal has been sent (S162: YES), the laser annealing processor 100 advances the process to S163.

[0211] In S163, the laser annealing processor 100 adds 1 to the value of the counter Nc to update the value of Nc.

[0212] In S164, the laser annealing processor 100 judges whether the value of the counter Nc is equal to or greater than Np+n. If the value of Nc is less than Np+n (S164: NO), the laser annealing processor 100 returns the process to S162. If the value of Nc is equal to or greater than Np+n (S164: YES), the laser annealing processor 100 ends the process of this flowchart and returns to the process shown in FIG.

[0213] Since n is an integer equal to or greater than 0, each time the laser device 12 outputs Np or more pulses of pulsed laser light, a determination is made as to whether or not to correct the irradiation conditions of the pulsed laser light. For example, when the value of n is set to 0, the update period of the irradiation conditions corresponds approximately to the time for Np pulses, and the irradiation conditions such as the target value Ft(k) of the fluence can be updated frequently. However, for example, when the target fluence value Ft(k) is corrected and a time equivalent to Np pulses has elapsed, the image acquisition range IM includes a portion that has been laser annealed in accordance with the target fluence value Ft(k-1) before correction.

[0214] Therefore, it is desirable to determine whether or not to correct the irradiation conditions of the pulsed laser light every time the laser device 12 outputs a pulsed laser light of 2×Np pulses or more. For example, when the value of n is Np, the update cycle of the irradiation conditions is approximately equivalent to the time for 2×Np pulses. After the time for 2×Np pulses has elapsed after correcting the target value Ft(k) of the fluence, the image acquisition range IM is laser annealed according to the corrected target value Ft(k). By using image data obtained by capturing this image acquisition range IM, the next target value Ft(k+1) of the fluence can be corrected more appropriately.

[0215] 5.3.5 Acquiring image data Fig. 44 is a flowchart showing details of the process of obtaining image data shown in Fig. 40. The flowchart shown in Fig. 44 corresponds to the subroutine of S170d in Fig. 40.

[0216] In S171, the laser annealing processor 100 sets an image acquisition range IM to a position shifted in the opposite direction to the scanning direction from the beam cross section B of the pulsed laser light. An example of the image acquisition range IM is as described with reference to FIGS. 38 and 39.

[0217] In S172, the laser annealing processor 100 captures an image of the image acquisition area IM. After S172, the laser annealing processor 100 ends the processing of this flowchart, and returns to the processing shown in FIG.

[0218] In Fig. 44, the image acquisition range IM is set and then the image acquisition range IM is imaged so that the amount of data obtained by imaging can be reduced. In order to perform such processing, the drive circuit 186 (see Fig. 19) of the imaging unit 182 may be configured to detect charges in a specified range of all the pixels of the image sensor 185, rather than detecting charges in all the pixels of the image sensor 185. In this case, the imaging controller 187 specifies the range of all the pixels of the image sensor 185 in which the drive circuit 186 detects charges, and acquires image data from the drive circuit 186.

[0219] The method of acquiring the image data is not limited to the above-mentioned method. For example, the driving circuit 186 may acquire an image by detecting electric charges from all pixels of the image sensor 185, and the imaging controller 187 may acquire image data in the image acquisition range IM from the image.

[0220] 5.4 Effect (12) The laser annealing system 10b according to the fourth embodiment includes an XYZ stage 174 that moves the substrate SUB in the X-axis direction and the Y-axis direction intersecting the incidence direction of the pulsed laser light to the substrate SUB. The XYZ stage 174 moves the substrate SUB by a distance corresponding to the beam width Bx of the pulsed laser light while the semiconductor film is irradiated with the pulsed laser light having the set irradiation pulse number Np. The laser annealing processor 100 determines whether or not to correct the irradiation conditions every time the semiconductor film is irradiated with the pulsed laser light having a pulse number equal to or greater than the irradiation pulse number Np. According to this, the irradiation conditions can be corrected while moving the substrate SUB in the step-and-scan manner. Also, the stop time of the substrate SUB can be made shorter than in the step-and-repeat manner.

[0221] (13) In the laser annealing system 10b according to the fourth embodiment, the laser annealing processor 100 determines whether or not to correct the irradiation conditions each time the semiconductor film is irradiated with pulsed laser light having a pulse number equal to or greater than twice the irradiation pulse number Np. According to this, image data of the semiconductor film laser annealed according to the irradiation conditions after the previous correction can be obtained, so that the correction of the irradiation conditions can be performed more appropriately.

[0222] (14) In the laser annealing system 10b according to the fourth embodiment, the imaging section 182 acquires image data by imaging the first region while the XYZ stage 174 is moving the substrate SUB. This can prevent the substrate SUB from being stopped for the purpose of acquiring image data.

[0223] (15) In the laser annealing system 10b according to the fourth embodiment, the imaging unit 182 includes a shutter 184 that blocks reflected light of the illumination light, and the open time Ts of the shutter 184 is shorter than the repetition period 1 / f of the pulsed laser light. According to this, image data can be acquired after the end of emission of one pulse of the pulsed laser light and before the start of emission of the next pulse, so that the pulsed laser light is not captured when acquiring the image data.

[0224] (16) The laser annealing system 10b according to the fourth embodiment includes an XYZ stage 174 that moves the substrate SUB in the X-axis direction and the Y-axis direction intersecting the incidence direction of the pulsed laser light on the substrate SUB. The illumination light source 181 is configured to irradiate an illumination range IL including the incidence position of the pulsed laser light on the substrate SUB and its periphery with illumination light. The imaging unit 182 includes an image sensor 185 capable of imaging the illumination range IL, a drive circuit 186 that reads out image data of a specified range from the image sensor 185, and an imaging controller 187 that controls the drive circuit 186. When the first region laser annealed by the pulsed laser light is moved to another position within the illumination range IL by the XYZ stage 174, the imaging controller 187 specifies the moved first region and acquires image data of the first region. According to this, since the first area in the illumination range IL that can be imaged by the image sensor 185 is specified to acquire image data, the amount of data obtained by imaging can be small.

[0225] (17) In the laser annealing system 10b according to the fourth embodiment, the XYZ stage 174 is configured to be movable in the X direction and the −X direction. The imaging unit 182 acquires image data of different positions within the illumination range IL when the XYZ stage 174 moves in the X direction and when the XYZ stage 174 moves in the −X direction. According to this, the acquisition range of image data can be changed when the XYZ stage 174 is moving in the X direction and when it is moving in the -X direction, so that the positional relationship between the incidence position of the pulsed laser light on the substrate SUB and the illumination range IL by the illumination light source 181 can be fixed.

[0226] (18) The laser annealing system 10b according to the fourth embodiment includes an XYZ stage 174 that moves the substrate SUB in the X-axis direction and the Y-axis direction intersecting the incidence direction of the pulsed laser light on the substrate SUB. The illumination light source 181 is configured to irradiate an illumination range IL including the incidence position of the pulsed laser light on the substrate SUB and its periphery with illumination light. The imaging unit 182 includes an image sensor 185 that can image the illumination range IL, and an imaging controller 187 that cuts out a part of the image of the illumination range IL to acquire image data. When the first region laser annealed by the pulsed laser light is moved by the XYZ stage 174 to another position within the illumination range IL, the imaging controller 187 cuts out an image of the moved first region from the image of the illumination range IL to acquire image data. According to this, even without a complex driving circuit 186, image data of a necessary range can be extracted from the image of the illumination range IL.

[0227] In other respects, the fourth embodiment is similar to the second embodiment. Alternatively, similar to the third embodiment, the gate threshold voltage may be estimated instead of estimating the electron mobility in the fourth embodiment.

[0228] 5.5 Other configuration examples The laser annealing system 10b in the fourth embodiment may be used instead of the laser annealing system 10 and the image data generating unit 20 included in the machine learning system in the first embodiment. That is, the laser annealing system 10b may be used to perform laser annealing of the substrate SUB for an annealing sample and acquisition of image data of the annealing sample by a step-and-scan method.

[0229] 6.Other The above description is intended to be illustrative rather than restrictive. Thus, it will be apparent to one skilled in the art that modifications may be made to the disclosed embodiments without departing from the scope of the claims. It will also be apparent to one skilled in the art that the disclosed embodiments may be used in combination.

[0230] Terms used throughout the present specification and claims should be construed as "open ended" terms unless otherwise specified. For example, terms such as "include," "have," "comprise," and "include" should be construed as "not excluding the presence of elements other than those listed." In addition, the modifier "a" should be construed as meaning "at least one" or "one or more." In addition, the term "at least one of A, B, and C" should be construed as "A," "B," "C," "A+B," "A+C," "B+C," or "A+B+C." Furthermore, it should be construed to include combinations thereof with elements other than "A," "B," and "C."

Claims

1. acquiring image data generated from reflected light of illumination light irradiated onto a first region of a semiconductor film on a substrate, the first region being laser annealed with pulsed laser light; acquiring measurement data of semiconductor characteristics in the first region; generating teacher data in which the image data is used as an input and the measurement data is used as an output; Performing machine learning using a neural network based on the teacher data to generate a trained model; machine learning methods, including

2. 2. The machine learning method of claim 1, After acquiring the image data, a semiconductor process is performed to manufacture a semiconductor device, and then the measurement data is acquired. Machine learning methods.

3. 2. The machine learning method of claim 1, Further comprising evaluating the trained model; generating the trained model using a portion of the teacher data as training data; The trained model is evaluated using another part of the teacher data as evaluation data. Machine learning methods.

4. a laser device that outputs a pulsed laser beam for laser annealing a first region of a semiconductor film on a substrate; an illumination light source that irradiates the first region with illumination light; an imaging unit that generates image data from the reflected light of the illumination light; 1. A processor comprising: The image data is input, and a trained model having an estimated value of a semiconductor characteristic as an output is accessible; Calculating the estimate using the image data and the trained model; correcting the irradiation condition of the pulsed laser beam based on the estimated value; the processor; 1. A laser annealing system comprising:

5. 5. The laser annealing system according to claim 4, Further comprising a beam splitter disposed in an optical path of the pulsed laser beam; the illumination light source irradiates the first region with the illumination light via the beam splitter; The imaging unit receives the reflected light via the beam splitter. Laser annealing system.

6. 6. The laser annealing system according to claim 5, the pulsed laser light includes a first wavelength component of ultraviolet light, The illumination light includes a plurality of second wavelength components of visible light. The beam splitter is a dichroic mirror that transmits one of the first and second wavelength components and reflects the other. Laser annealing system.

7. 5. The laser annealing system according to claim 4, a stage that moves the substrate in a direction intersecting an incidence direction of the pulsed laser light on the substrate; the first region is irradiated with the pulsed laser light having a set number of irradiation pulses while the stage is stopped; Then, the stage moves the substrate so that a second region different from the first region is irradiated with the pulsed laser light; The processor corrects the irradiation conditions before the second region is irradiated with the pulsed laser light. Laser annealing system.

8. 5. The laser annealing system according to claim 4, the irradiation conditions include a fluence of the pulsed laser light, the semiconductor properties include electron mobility; The processor, If the estimated value becomes higher as a result of increasing the fluence, the irradiation conditions are corrected to increase the fluence; correcting the irradiation conditions to lower the fluence if the estimated value becomes lower as a result of increasing the fluence; correcting the irradiation conditions to lower the fluence if the estimated value becomes high as a result of lowering the fluence; If the estimated value becomes low as a result of lowering the fluence, the irradiation conditions are corrected so as to increase the fluence. Laser annealing system.

9. 5. The laser annealing system according to claim 4, The processor, when an absolute value of a difference between the estimated value and a target value of the semiconductor characteristic is greater than a predetermined value, performing trial irradiation while changing the irradiation condition, acquiring the image data, and calculating the estimated value, thereby acquiring a relationship between the irradiation condition and the estimated value, and correcting the irradiation condition based on the relationship. Laser annealing system.

10. 5. The laser annealing system according to claim 4, the irradiation conditions include the number of irradiation pulses of the pulsed laser light irradiated to one location on the substrate, The processor, correcting the irradiation conditions so as to increase the number of irradiation pulses when an absolute value of a difference between the estimated value and the target value of the semiconductor characteristic is greater than a predetermined value; Laser annealing system.

11. 5. The laser annealing system according to claim 4, the irradiation conditions include a fluence of the pulsed laser light, the semiconductor characteristic includes a gate threshold voltage; The processor, If the estimated value becomes low as a result of increasing the fluence, the irradiation conditions are corrected to increase the fluence; correcting the irradiation conditions to lower the fluence if the estimated value becomes high as a result of increasing the fluence; correcting the irradiation conditions to lower the fluence if the estimated value is lowered as a result of lowering the fluence; If the estimated value becomes high as a result of lowering the fluence, the irradiation conditions are corrected so as to increase the fluence. Laser annealing system.

12. 5. The laser annealing system according to claim 4, The method further includes a stage that moves the substrate in a direction intersecting an incidence direction of the pulsed laser light to the substrate, the stage moving the substrate by a distance corresponding to a beam width of the pulsed laser light while the semiconductor film is irradiated with the pulsed laser light of a set number of irradiation pulses, the processor determines whether or not to correct the irradiation conditions each time the semiconductor film is irradiated with the pulsed laser light having a number of pulses equal to or greater than the number of irradiation pulses. Laser annealing system.

13. 13. The laser annealing system of claim 12, the processor determines whether or not to correct the irradiation conditions each time the semiconductor film is irradiated with the pulsed laser light having a pulse number equal to or greater than twice the irradiation pulse number. Laser annealing system.

14. 13. The laser annealing system of claim 12, the imaging unit acquires the image data by imaging the first region while the stage is moving the substrate. Laser annealing system.

15. 13. The laser annealing system of claim 12, the imaging unit includes a shutter that blocks the reflected light, and an opening time of the shutter is shorter than a repetition period of the pulsed laser light. Laser annealing system.

16. 5. The laser annealing system according to claim 4, a stage that moves the substrate in a direction intersecting an incidence direction of the pulsed laser light on the substrate; the illumination light source is configured to irradiate an illumination range including an incidence position of the pulsed laser light on the substrate and a periphery thereof with the illumination light; the imaging unit includes an image sensor capable of imaging the illumination range, a drive circuit that reads out image data of a designated range from the image sensor, and an imaging controller that controls the drive circuit, and when the first region laser annealed by the pulsed laser light is moved by the stage to another position within the illumination range, the imaging controller specifies the moved first region and acquires the image data of the first region. Laser annealing system.

17. 17. The laser annealing system of claim 16, The stage is configured to be movable in a first direction and a second direction, the imaging unit acquires the image data at different positions within the illumination range when the stage is moving in the first direction and when the stage is moving in the second direction. Laser annealing system.

18. 5. The laser annealing system according to claim 4, a stage that moves the substrate in a direction intersecting an incidence direction of the pulsed laser light on the substrate; the illumination light source is configured to irradiate an illumination range including an incidence position of the pulsed laser light on the substrate and a periphery thereof with the illumination light; the imaging unit includes an image sensor capable of imaging the illumination range, and an imaging controller configured to acquire the image data by cutting out a portion from the image of the illumination range, and when the first region laser annealed by the pulsed laser light is moved by the stage to another position within the illumination range, the imaging controller cuts out an image of the moved first region from the image of the illumination range to acquire the image data. Laser annealing system.

19. 20. The laser annealing system of claim 18, The stage is configured to be movable in a first direction and a second direction, the imaging unit acquires the image data at different positions within the illumination range when the stage is moving in the first direction and when the stage is moving in the second direction. Laser annealing system.

20. annealing a first region of a semiconductor film on a substrate with pulsed laser light; Illuminating the first region with illumination light and generating image data from reflected light of the illumination light; calculating the estimated value using a trained model that has the image data as input and the estimated value of the semiconductor characteristic as output; correcting the irradiation condition of the pulsed laser beam based on the estimated value; A laser annealing method comprising:

Images