Method and apparatus for reverse engineering nano-optical devices based on deep learning

Abstract

A method and apparatus for reverse engineering a nano-optical device based on deep learning are disclosed. The nano-optical device reverse engineering method may include the steps of: setting initial parameters of the nano-optical device; determining a transmission spectrum of the nano-optical device using an analysis model; determining a reward to be input to a neural network based on the transmission spectrum; and performing deep learning with the neural network that receives the reward to reverse engineer the nano-optical device.

Description

Technology Field

[0001] The present invention relates to a method and apparatus for reverse engineering a nano-optical device, and more specifically, to a method and apparatus for searching for optimal parameters corresponding to a nano-optical device of a target value using deep learning. Background Technology

[0002] As nano-optical devices are used in various fields, methods for reverse engineering nano-optical devices are also being studied.

[0003] Conventional nano-optical device inverse engineering methods inversely engineered nano-optical devices by repeating the process of determining the parameters of the nano-optical device through numerical analysis simulations.

[0004] However, the reverse engineering method for nano-optical devices using numerical analysis simulations can be calculated quickly when the difference between the wavelength of the electromagnetic waves incident on the nano-optical device and the size of the nano-optical device is within a certain range. However, when the wavelength of the electromagnetic waves is between 100 μm (3 THz) and 10 mm (30 GHz), the size of the nano-optical device is several nanometers, and the difference in size can be as much as 10,000 to 10,000,000 times the wavelength, so the simulation time becomes very long or impossible, making it impossible to reverse engineer the nano-optical device.

[0005] Therefore, there is a demand for a method to reverse engineer nano-optical devices even when the difference between the wavelength of the electromagnetic wave and the size of the nano-optical device exceeds a certain range. The problem to be solved

[0006] The present invention provides a method and apparatus that can improve processing speed compared to a reverse engineering device using a simulation tool by determining transmittance using an analysis model. means of solving the problem

[0007] A method for reverse engineering a nano-optical device according to one embodiment of the present invention may include the steps of: setting initial parameters of the nano-optical device; determining a transmission spectrum of the nano-optical device using an analysis model; determining a reward to be input to a neural network based on the transmission spectrum; and performing deep learning with the neural network that receives the reward to reverse engineer the nano-optical device.

[0008] The analysis model of the nano-optical device reverse engineering method according to one embodiment of the present invention can determine the electric field of the nano-optical device according to the initial parameters of the nano-optical device, normalize the electric field, and determine the transmittance of the nano-optical device based on the normalized electric field and the ratio of metal to dielectric included in the nano-optical device.

[0009] The step of determining the transmission spectrum of the nano-optical device reverse engineering method according to one embodiment of the present invention can determine a transmission spectrum representing the ratio of the electromagnetic wave transmitted through the nano-optical device based on the electromagnetic wave incident on the nano-optical device and the transmittance of the nano-optical device.

[0010] The reward of the nano-optical device reverse engineering method according to one embodiment of the present invention can be determined based on the peak position, dip position, and electric field amplification at the peak of the process frequency of the nano-optical device.

[0011] The electric field amplification at the peak of the nano-optical device reverse engineering method according to one embodiment of the present invention can be determined based on a coverage ratio representing the ratio of metal to dielectric included in the nano-optical device and the maximum transmittance of the nano-optical device.

[0012] The step of determining the reward in the nano-optical device reverse engineering method according to one embodiment of the present invention may determine the reward by combining the interval between the target value of the peak position and the peak position of the process frequency, the difference between the target value of the interval between the peak position and the diff position and the interval between the peak position and the diff position of the process frequency, and the electric field amplification at the peak.

[0013] The step of determining the reward in the nano-optical device reverse engineering method according to one embodiment of the present invention may determine the reward based on the similarity between the target value of the nano-optical device and the transmission spectrum.

[0014] The nano-optical device of the nano-optical device reverse engineering method according to one embodiment of the present invention is an optical device antenna having a plurality of polygonal holes formed therein, and the neural network can optimize the parameters by receiving the reward and parameters including the horizontal length of each of the polygonal holes, the vertical length of each of the polygonal holes, the spacing between the polygonal holes, and the depth of each of the polygonal holes.

[0015] The nano-optical device of the nano-optical device reverse engineering method according to one embodiment of the present invention is an optical device antenna in which a plurality of polygonal loops are formed, the sides of each of the loops are formed of a dielectric, and the remaining region excluding the sides of each of the loops is formed of a metal, and the neural network can optimize the parameters by receiving the reward and parameters including the horizontal length of each of the polygonal loops, the vertical length of each of the polygonal loops, the spacing between the polygonal loops, the width of the sides of each of the loops, and the depth of the sides of each of the loops.

[0016] A method for reverse engineering a nano-optical device according to one embodiment of the present invention may further include: a step of terminating the operation of a deep learning agent and determining the parameters of the reverse-engineered nano-optical device as optimal parameters corresponding to a target value when the parameters of the reverse-engineered nano-optical device satisfy optimal conditions; a step of correcting the parameters of the nano-optical device when the parameters of the reverse-engineered nano-optical device do not satisfy optimal conditions; a step of recreating the transmission spectrum of the nano-optical device using the corrected parameters; a step of recreating a reward to be input to a neural network based on the recreating transmission spectrum; and a step of performing deep learning with the neural network that receives the recreating reward to reverse engineer the nano-optical device.

[0017] A nano-optical device reverse engineering device according to one embodiment of the present invention may include: a parameter setting unit for setting initial parameters of a nano-optical device; a transmission spectrum determining unit for determining a transmission spectrum of the nano-optical device using an analysis model; a reward determining unit for determining a reward to be input to a neural network based on the transmission spectrum; and a deep learning agent for reverse engineering the nano-optical device by performing deep learning with the neural network that receives the reward.

[0018] The analysis model of the nano-optical device reverse engineering device according to one embodiment of the present invention can determine the electric field of the nano-optical device according to the initial parameters of the nano-optical device, normalize the electric field, and determine the transmittance of the nano-optical device based on the normalized electric field and the ratio of metal to dielectric included in the nano-optical device.

[0019] The reward determination unit of the nano-optical device reverse engineering device according to one embodiment of the present invention can determine the reward by combining the interval between the target value of the peak position and the peak position of the process frequency, the difference between the target value of the interval between the peak position and the diff position and the interval between the peak position and the diff position of the process frequency, and the electric field amplification degree at the peak.

[0020] The nano-optical device of the nano-optical device reverse engineering device according to one embodiment of the present invention is an optical device antenna having a plurality of polygonal holes formed therein, and the neural network can optimize the parameters by receiving the reward and parameters including the horizontal length of each of the polygonal holes, the vertical length of each of the polygonal holes, the spacing between the polygonal holes, and the depth of each of the polygonal holes.

[0021] The nano-optical device of the nano-optical device reverse engineering device according to one embodiment of the present invention is an optical device antenna in which a plurality of polygonal loops are formed, the sides of each of the loops are formed of a dielectric, and the remaining region excluding the sides of each of the loops is formed of metal, and the neural network can optimize the parameters by receiving the reward and parameters including the horizontal length of each of the polygonal loops, the vertical length of each of the polygonal loops, the spacing between the polygonal loops, the width of the sides of each of the loops, and the depth of the sides of each of the loops. Effects of the invention

[0022] According to one embodiment of the present invention, by determining the transmittance using an analysis model, the processing speed can be improved compared to a reverse engineering device using a simulation tool.

[0023] In addition, according to one embodiment of the present invention, by improving the processing speed required for reverse engineering of nano-optical devices, it is possible to reverse engineer terahertz active switching devices or terahertz superconducting active devices that require a large amount of processing. Brief explanation of the drawing

[0024] FIG. 1 is a drawing illustrating a nano-optical device reverse engineering apparatus according to an embodiment of the present invention. Figure 2 is an example of the electric field generated in an optical antenna depending on the gap of the optical antenna. FIG. 3 is an example of a nano-optical device reverse engineering process according to an embodiment of the present invention. Figure 4 is an example of the performance of a nano-optical device determined according to an embodiment of the present invention. FIG. 5 is another example of a nano-optical device reverse engineering process according to one embodiment of the present invention. FIG. 6 is an example of a process for simplifying a nano-optical device to apply an interpretation model according to an embodiment of the present invention. FIG. 7 is an example of information used to generate a reward according to an embodiment of the present invention. FIG. 8 is a flowchart illustrating a nano-optical device reverse engineering method according to an embodiment of the present invention. Specific details for implementing the invention

[0025] Hereinafter, embodiments are described in detail with reference to the attached drawings. However, various modifications may be made to the embodiments, and thus the scope of the patent application is not limited or restricted by these embodiments. It should be understood that all modifications, equivalents, and substitutions to the embodiments are included within the scope of the rights.

[0026] The terms used in the embodiments are for illustrative purposes only and should not be interpreted as intended to be limiting. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprising" or "having" are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0027] In addition, when describing with reference to the attached drawings, identical components are assigned the same reference numeral regardless of drawing symbols, and redundant descriptions thereof are omitted. In describing the embodiments, if it is determined that a detailed description of related prior art could unnecessarily obscure the essence of the embodiments, such detailed description is omitted.

[0028] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

[0029] FIG. 1 is a drawing illustrating a nano-optical device reverse engineering device (100) according to an embodiment of the present invention.

[0030] A nano-optical device reverse engineering device (100) according to one embodiment of the present invention may include a parameter setting unit (110), a transmission spectrum determining unit (120), a reward determining unit (130), a neural network (140), and a deep learning agent (150) as shown in FIG. 1. At this time, the parameter setting unit (110), the transmission spectrum determining unit (120), the reward determining unit (130), the neural network (140), and the deep learning agent (150) may be different processors as shown in FIG. 1, or they may be respective modules included in a program executed on a single processor.

[0031] The parameter setting unit (110) can set the initial parameters of the nano-optical device.

[0032] The transmission spectrum determining unit (120) can determine the transmission spectrum of the nano-optical device using an analysis model. The analysis model may be a model that determines the electric field of the nano-optical device according to the initial parameters of the nano-optical device, normalizes the electric field, and determines the transmittance of the nano-optical device based on the normalized electric field and the ratio of metal to dielectric included in the nano-optical device. In addition, the transmission spectrum determining unit (120) can determine a transmission spectrum representing the ratio of the electromagnetic waves transmitted through the nano-optical device based on the electromagnetic waves incident on the nano-optical device and the transmittance of the nano-optical device determined by the analysis model. For example, the electromagnetic waves incident on the nano-optical device may be terahertz waves, which are electromagnetic waves in the communication frequency band.

[0033] The reward determination unit (130) can determine a reward to be input to the neural network (140) based on the transmission spectrum determined by the transmission spectrum determination unit (120). At this time, the reward determination unit (130) can determine the reward based on the similarity between the target value of the nano-optical device and the transmission spectrum. Additionally, the reward can be determined based on the peak position, dip position, and electric field amplification at the peak of the process frequency of the nano-optical device.

[0034] The electric field amplification at the peak can be determined based on the coverage ratio, which represents the ratio of metal to dielectric included in the nano-optical device, and the maximum transmittance of the nano-optical device. Additionally, the reward determination unit (130) can determine the reward by combining the interval between the target value of the peak position and the peak position of the process frequency, the difference between the target value of the interval between the peak position and the diff position and the interval between the peak position and the diff position of the process frequency, and the electric field amplification at the peak.

[0035] In the case where the nano-optical device is an optical device antenna in which a plurality of polygonal holes are formed, the neural network (140) can optimize the parameters by receiving a reward and parameters including the horizontal length of each polygonal hole, the vertical length of each polygonal hole, the spacing between polygonal holes, and the depth of each polygonal hole.

[0036] In the case where the nano-optical device is an optical device antenna in which a plurality of polygonal loops are formed, each side of the loops is formed of a dielectric material, and the remaining area excluding each side of the loops is formed of metal, the neural network (140) can optimize the parameters by receiving a reward and parameters including the horizontal length of each polygonal loop, the vertical length of each polygonal loop, the spacing between polygonal loops, the width of each side of the loops, and the depth of each side of the loops.

[0037] The deep learning agent (150) can reverse engineer a nano optical device by performing deep learning with a neural network (140) that receives a reward as input.

[0038] When the parameters of the reverse-engineered nano-optical device satisfy optimal conditions, the deep learning agent (150) terminates the operation of the deep learning agent and can determine the parameters of the reverse-engineered nano-optical device as optimal parameters corresponding to the target value.

[0039] If the parameters of the reverse-engineered nano-optical device do not satisfy optimal conditions, the deep learning agent (150) can correct the parameters of the nano-optical device. Then, the transmission spectrum determining unit (120) can recrystallize the transmission spectrum of the nano-optical device using the corrected parameters. Next, the reward determining unit (130) can determine a reward to be input to the neural network (140) based on the recrystallized transmission spectrum. Then, the deep learning agent (150) can reverse-engineer the nano-optical device by performing deep learning with the neural network (140) that received the recrystallized reward.

[0040] The transmission spectrum determining unit (120), the reward determining unit (130), and the deep learning agent (150) can repeat the process described above until the parameters of the reverse-engineered nano-optical device satisfy optimal conditions.

[0041] The nano-optical device reverse engineering device (100) can improve processing speed compared to a reverse engineering device using a simulation tool by determining transmittance using an analysis model.

[0042] In addition, the nano-optical device reverse engineering device (100) can reverse engineer a terahertz active switch device or a terahertz superconducting active device that requires a large amount of processing by improving the processing speed required for reverse engineering of a nano-optical device.

[0043] Figure 2 is an example of the electric field generated in an optical antenna depending on the gap of the optical antenna. In Figure 2, the graph placed on the left side of the image can represent the strength of the magnetic field corresponding to the color of the magnetic field generated at both ends of the optical antenna.

[0044] The images (210, 220, 230, 240) of FIG. 2 sequentially illustrate an example of an optical antenna with the widest distance (gap) (200) between the two ends of the optical antenna, starting from image (210) showing the widest example of an optical antenna with the widest gap (200), up to image (240) showing the widest example of an optical antenna.

[0045] As shown in FIG. 2, the magnetic field generated at both ends of the optical element antenna may be weakest at the optical element antenna with the widest gap (200) and strongest at the optical element antenna with the narrowest gap.

[0046] That is, the narrower the gap (200) of the optical element antenna, the greater the magnitude of the electric field generated in the optical element antenna.

[0047] FIG. 3 is an example of a nano-optical device reverse engineering process according to an embodiment of the present invention.

[0048] The nano-optical device reverse engineering device (100) can reverse engineer a hole nano-optical device (310) including rectangular holes as shown in FIG. 3.

[0049] The parameter (311) of the hole nano-optical device (310) is the horizontal length a of each of the square holes. x , the vertical length of each of the square holes a y , horizontal spacing between square holes l x , vertical spacing between square holes l y It may include the depth h of each of the square holes.

[0050] The transmission spectrum determining unit (120) can determine the transmittance of the hole nano-optical device (310) using an analysis model. The reward determining unit (130) can determine the reward to be input to the neural network (320) based on the transmission spectrum.

[0051] The neural network (320) can receive parameters (311) and rewards as input. And, the deep learning agent (150) can perform reverse engineering to search for optimal parameters corresponding to the target value of the nano optical device by performing deep learning with the neural network (320).

[0052] If the parameters of the reverse-engineered nano-optical device do not satisfy optimal conditions, the deep learning agent (150) can correct the parameters of the nano-optical device.

[0053] The neural network (320) may be a Deep Q Network (DQN) that combines Q learning and deep learning. In this case, the Q output from the neural network (320) can be represented as Equation 1.

[0054]

[0055] At this time, s is a parameter (311), and a may be a correction value that the deep learning agent (150) applies to the parameter (311) to correct the parameter (311). For example, a may be one of increasing, decreasing, or maintaining. If a is increasing, the deep learning agent (150) may increase the value of the initial parameter to determine it as the corrected parameter. If a is maintaining, the deep learning agent (150) may determine the initial parameter as the corrected parameter. If a is decreasing, the deep learning agent (150) may decrease the value of the initial parameter to determine it as the corrected parameter. Also, Θ may be a weight.

[0056] The cost function of the neural network (320) can be defined as Equation 2.

[0057]

[0058] When the parameters of the nano-optical device are determined in the neural network (320), the simulation environment (330) can perform a simulation of the structure of the nano-optical device by applying an analytical solution to the determined parameters of the nano-optical device.

[0059] Figure 4 is an example of the performance of a nano-optical device determined according to an embodiment of the present invention.

[0060] When reverse engineering a nano-optical device using a simulation tool, the peak position (411) and the dip position (412) can be determined as shown in the graph (410).

[0061] When the nano-optical device reverse engineering device (100) reverse engineers the nano-optical device, the peak position (421) and the dip position (422) can be determined as shown in the graph (420).

[0062] The nano-optical device reverse engineering device (100) can determine the transmission spectrum using an analysis model, thereby allowing the peak position (421) and the diff position (422) to converge to an optimized position at a faster speed than the graph (410).

[0063] FIG. 5 is another example of a nano-optical device reverse engineering process according to one embodiment of the present invention.

[0064] The nano-optical device reverse engineering device (100) can reverse engineer a loop nano-optical device (500) that includes rectangular loops as shown in FIG. 5 and has gaps formed on the sides of the loops.

[0065] The transmission spectrum determining unit (120) can determine the transmittance (530) of the loop nano-optical device (500) using an analysis model. Additionally, the transmission spectrum determining unit (120) can determine a transmission spectrum representing the ratio of electromagnetic waves (520) transmitted from the loop nano-optical device (500) to the nano-optical device, based on the electromagnetic waves (510) incident on the loop nano-optical device (500) and the transmittance (530) of the nano-optical device determined by the analysis model. At this time, the parameter (550) of the loop nano-optical device (500) is the horizontal length a of each of the rectangular loops. x , vertical length a of each of the rectangular loops y , horizontal spacing between rectangular loops l x , vertical spacing between rectangular loops l y It may include the width w of each side of the square loops and the depth h of each side of the square loops.

[0066] The reward determination unit (130) can determine a reward (540) to be input to the neural network (560) based on the transmission spectrum determined by the transmission spectrum determination unit (120). At this time, the reward determination unit (130) can determine the reward (540) according to the similarity between the target value of the nano-optical device and the transmission spectrum.

[0067] The neural network (560) can receive parameters (550) and rewards (540) as input. And, the deep learning agent (150) can perform reverse engineering to search for optimal parameters corresponding to the target value of the nano optical device by performing deep learning with the neural network (560).

[0068] At this time, the deep learning agent (150) can correct the parameter (550) according to the output of the neural network (560). Then, the transmission spectrum determining unit (120) can recrystallize the transmission spectrum by calculating the transmittance (530) according to the corrected parameter (550). In addition, the reward determining unit (130) can recrystallize the reward (540) based on the recrystallized transmission spectrum.

[0069] At this time, the neural network (560) can perform deep learning by receiving the re-determined reward (540) and the corrected parameter (550) as input. The deep learning agent (150) can repeat the process described above until the output of the neural network (560) satisfies the optimal condition.

[0070] FIG. 6 is an example of a process for simplifying a nano-optical device to apply an interpretation model according to an embodiment of the present invention.

[0071] The actual nano-optical device may have a structure in which the lower surface of each side of the loop is connected to the dielectric material constituting the sides of the loop, as shown in image (610). However, if the transmittance is determined by considering the structure of the actual nano-optical device, it may be difficult to apply the interpretation model due to the dielectric material passing through the lower surface of the metal. Therefore, the transmission spectrum determining unit (120) can determine the transmittance by setting the nano-optical device to a structure in which dielectric materials are placed only on each side of the loop, as shown in image (620).

[0072] The magnitude of the electric field generated in the gap of a nano-optical device can be determined by boundary conditions. For example, the electric field generated in the gap of a nano-optical device can be expressed as in Equation 3. In this case, E gap This can refer to the electric field generated near the structure when an electric field is incident on the structure (the amplified electric field in the near field).

[0073]

[0074] At this time, , , Each can be defined as in mathematical equations 4, 5, and 6.

[0075]

[0076]

[0077]

[0078] At this time, represents the dielectric function of metals and dielectrics, and the dielectric can be a gap material. Also. Silver can represent the leaking wavevector of light leaking into the metal. It can be expressed as in mathematical formula 7.

[0079]

[0080] In addition, of mathematical formula 6 It can be defined as in mathematical formula 8.

[0081]

[0082] At this time, Is It is an integer multiple, and Is It can be an integer sum of. Also, is the width of the rectangular loop and ε₀ can be the vertical length of the rectangular loop. Also, h₀ is the thickness of the metal thin film, and w₀ is the gap width of the nano gap. Additionally, J in FIG. 6 can be a sinc function as shown in Equation 9.

[0083]

[0084] Finally, the electric field E from the outside out can be expressed as Equation 10 according to the normalization process. In this case, the external electric field E out can refer to the far field electric field that can be observed in transmission experiments.

[0085]

[0086] And, the interpretation model is the electric field E from the outside out The transmittance t of a loop nano-gap structure as shown in Equation 11 can be determined by normalizing the ratio of the portion covered by metal and the portion where the dielectric is placed in the nano-optical device.

[0087]

[0088] FIG. 7 is an example of information used to generate a reward according to an embodiment of the present invention.

[0089] The reward of the inverse design process can be determined based on the peak position, dip position, and electric field amplification at the peak of the process frequency of the nano-optical device.

[0090] At this time, the reward determination unit (130) can determine the field enhancement using the coverage ratio and transmittance t as in Equation 12.

[0091]

[0092] The coverage ratio is the ratio of the portion covered by metal and the portion where the dielectric is placed in a nano-optical device, and can be defined as in Equation 13.

[0093]

[0094] Accordingly, the reward determination unit (130) has an electric field amplification degree at the peak according to mathematical formula 14. can decide.

[0095]

[0096] At this time, the electric field amplification at the peak The value is tens to tens of thousands ( Since it changes, the range of change can be very large. The reward determination unit (130) may apply a logarithmic function as in Equation 15 to include the electric field amplification at the peak in the reward.

[0097]

[0098] The reward determination unit (130) has a target value of the peak position (target position) as in mathematical formula 16, ) and peak position of process frequency The interval between can decide.

[0099]

[0100] The reward determination unit (130) has a target value for the interval between the peak position and the deep position as in mathematical formula 17. and peak position of process frequency Wow, deep location The difference between the intervals can decide.

[0101]

[0102] The reward determination unit (130) can determine the reward by combining the values ​​determined in mathematical formulas 15, 16, and 17. For example, the reward determination unit (130) can input a total reward defined as in mathematical formula 18 into the neural network.

[0103]

[0104] FIG. 8 is a flowchart illustrating a nano-optical device reverse engineering method according to an embodiment of the present invention.

[0105] In step (810), the parameter setting unit (110) can set the initial parameters of the nano-optical device.

[0106] In step (820), the transmission spectrum determining unit (120) can determine the transmission spectrum of the nano-optical device using an analysis model. The transmission spectrum determining unit (120) can determine a transmission spectrum representing the ratio of electromagnetic waves transmitted through the nano-optical device based on the electromagnetic waves incident on the nano-optical device and the transmittance of the nano-optical device determined by the analysis model.

[0107] In step (830), the reward determination unit (130) can determine a reward to be input to the neural network (140) based on the transmission spectrum determined in step (820). At this time, the reward determination unit (130) can determine the reward based on the similarity between the target value of the nano-optical device and the transmission spectrum.

[0108] In step (840), the deep learning agent (150) can reverse engineer the nano optical device by performing deep learning with the neural network (140) that receives the reward as input.

[0109] In step (850), the deep learning agent (150) can check whether the parameters of the nano-optical device reverse-engineered in step (840) satisfy optimal conditions. If the parameters of the nano-optical device reverse-engineered in step (840) satisfy optimal conditions, the deep learning agent (150) can terminate the operation of the deep learning agent and determine the parameters of the reverse-engineered nano-optical device as optimal parameters corresponding to the target value.

[0110] If the parameters of the reverse-engineered nano-optical device do not satisfy the optimal conditions, the deep learning agent (150) can perform step (860).

[0111] In step (860), the deep learning agent (150) can determine the correction direction of the parameters of the nano-optical device. At this time, the deep learning agent (150) can determine one of the correction directions of increasing, decreasing, or maintaining for each of the parameters of the nano-optical device.

[0112] In step (860), the deep learning agent (150) can correct the parameters of the nano-optical device from the initial parameters to parameters according to the correction direction. For example, if the correction direction is increasing, the deep learning agent (150) can increase the value of the initial parameter to determine the corrected parameter. If the correction direction is maintaining, the deep learning agent (150) can determine the initial parameter to the corrected parameter. If the correction direction is decreasing, the deep learning agent (150) can decrease the value of the initial parameter to determine the corrected parameter.

[0113] At this time, the transmission spectrum determining unit (120) can recrystallize the transmission spectrum of the nano-optical device using the corrected parameters. Next, the reward determining unit (130) can determine a reward to be input to the neural network (140) based on the recrystallized transmission spectrum. Then, the deep learning agent (150) can reverse-engineer the nano-optical device by performing deep learning on the neural network (140) that received the recrystallized reward.

[0114] The transmission spectrum determining unit (120), the reward determining unit (130), and the deep learning agent (150) can repeat the process described above until the parameters of the reverse-engineered nano-optical device satisfy optimal conditions.

[0115] The nano-optical device reverse engineering device (100) can improve processing speed compared to a reverse engineering device using a simulation tool by determining transmittance using an analysis model.

[0116] In addition, the nano-optical device reverse engineering device (100) can reverse engineer a terahertz active switch device or a terahertz superconducting active device that requires a large amount of processing by improving the processing speed required for reverse engineering of a nano-optical device.

[0118] Meanwhile, the nano-optical device reverse engineering device based on ignition data or the nano-optical device reverse engineering method based on ignition data according to the present invention is written as a program that can be executed on a computer and can also be implemented on various recording media such as magnetic storage media, optical reading media, and digital storage media.

[0119] Implementations of the various technologies described herein may be implemented as digital electronic circuits, or as computer hardware, firmware, software, or combinations thereof. Implementations may be implemented as computer programs tangibly embodied in a computer program product, for example, a machine-readable storage device (computer-readable medium), for processing by the operation of a data processing device, e.g., a programmable processor, a computer, or a plurality of computers, or for controlling such operation. Computer programs such as the computer program(s) described above may be written in any form of programming language, including compiled or interpreted languages, and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. Computer programs may be deployed to be processed on one computer or a plurality of computers at one site, or distributed across a plurality of sites and interconnected by a communication network.

[0120] Processors suitable for processing computer programs include, for example, both general-purpose and special-purpose microprocessors, and any one or more processors of any type of digital computer. Generally, the processor will receive instructions and data from read-only memory or random access memory, or both. The elements of the computer may include at least one processor that executes instructions and one or more memory devices that store instructions and data. Generally, the computer may include one or more mass storage devices that store data, for example, magnetic, magneto-optical disks, or optical disks, or may be combined to receive data from these, transmit data to these, or both. Information carriers suitable for embodying computer program instructions and data include, for example, semiconductor memory devices, magnetic media such as hard disks, floppy disks, and magnetic tapes; optical media such as CD-ROMs (Compact Disk Read Only Memory) and DVDs (Digital Video Disks); magneto-optical media such as floptical disks; ROMs (Read Only Memory); RAMs (Random Access Memory); flash memory; EPROMs (Erasable Programmable ROM); EEPROMs (Electrically Erasable Programmable ROM); etc. Processors and memory may be supplemented by or included in special-purpose logic circuit organizations.

[0121] Additionally, a computer-readable medium may be any available medium accessible by a computer and may include all computer storage media.

[0122] Although this specification contains details of a number of specific embodiments, they should not be understood as limiting the scope of any invention or claimables, but rather as descriptions of features that may be characteristic of a specific embodiment of a specific invention. Specific features described in this specification in the context of individual embodiments may be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented in multiple embodiments individually or in any appropriate sub-combination. Furthermore, while features may operate in a specific combination and be described as initially claimed, one or more features from the claimed combination may be excluded from the combination in some cases, and the claimed combination may be changed to a sub-combination or a variation of the sub-combination.

[0123] Likewise, although operations are depicted in the drawings in a specific order, this should not be understood as requiring that such operations be performed in that specific or sequential order depicted to obtain a desirable result, or that all depicted operations must be performed. In certain cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of the various device components of the embodiments described above should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and devices can generally be integrated together into a single software product or packaged into multiple software products.

[0124] Meanwhile, the embodiments of the present invention disclosed in this specification and drawings are merely specific examples provided to aid understanding and are not intended to limit the scope of the present invention. It is obvious to those skilled in the art that other variations based on the technical concept of the present invention are possible in addition to the embodiments disclosed herein. Explanation of the symbols

[0125] 110: Parameter setting section 120: Transmission Spectrum Determination Section 130: Reward Decision Department 140: Neural Network 150: Deep learning agent

Claims

Claim 1 A method for reverse engineering a nano-optical device, comprising a parameter setting unit, a transmission spectrum determining unit, a reward determining unit, and a deep learning agent, wherein the parameter setting unit sets initial parameters of the nano-optical device; the transmission spectrum determining unit determines the transmission spectrum of the nano-optical device using an analysis model; the reward determining unit determines a reward to be input to a neural network based on the transmission spectrum; and the deep learning agent performs deep learning on the neural network that receives the reward to reverse engineer the nano-optical device, wherein the step of determining the transmission spectrum is a nano-optical device having a structure in which dielectrics are disposed only on each of the sides of the loops, such that the nano-optical device has a structure in which a plurality of polygonal loops are formed, each of the loops is formed with a variant dielectric, and the lower surface of each of the sides of the loops is connected to the dielectric constituting the sides of the loops, thereby determining the transmittance. Claim 2 A nano-optical device inverse engineering method according to claim 1, wherein the interpretation model determines the electric field of the nano-optical device according to the initial parameters of the nano-optical device, normalizes the electric field, and determines the transmittance of the nano-optical device based on the normalized electric field and the ratio of metal to dielectric included in the nano-optical device. Claim 3 In claim 1, the step of determining the transmission spectrum is a nano-optical device inverse engineering method in which the transmission spectrum determining unit determines a transmission spectrum representing the ratio of the electromagnetic wave transmitted through the nano-optical device based on the electromagnetic wave incident on the nano-optical device and the transmittance of the nano-optical device. Claim 4 In claim 1, the reward is a nano-optical device inverse engineering method determined based on the peak position, dip position, and electric field amplification at the peak of the process frequency of the nano-optical device. Claim 5 A method for reverse engineering a nano-optical device according to claim 4, wherein the electric field amplification at the peak is determined based on a coverage ratio representing the ratio of metal to dielectric included in the nano-optical device and the maximum transmittance of the nano-optical device. Claim 6 In claim 4, the step of determining the reward is a nano-optical device inverse engineering method in which the reward determining unit determines the reward by combining the interval between the target value of the peak position and the peak position of the process frequency, the difference between the target value of the interval between the peak position and the diff position and the interval between the peak position and the diff position of the process frequency, and the electric field amplification degree at the peak. Claim 7 In claim 1, the step of determining the reward is a nano-optical device reverse engineering method in which the reward determining unit determines the reward according to the similarity between the target value of the nano-optical device and the transmission spectrum. Claim 8 delete Claim 9 A method for reverse engineering a nano-optical device according to claim 1, wherein the neural network receives a reward and parameters including the horizontal length of each of the polygonal loops, the vertical length of each of the polygonal loops, the spacing between the polygonal loops, the width of each of the sides of each of the loops, and the depth of each of the sides of each of the loops, and optimizes the parameters. Claim 10 A method for reverse engineering a nano-optical device according to claim 1, further comprising: a step in which, if the parameters of the reverse-engineered nano-optical device satisfy optimal conditions, the deep learning agent terminates the operation of the deep learning agent and determines the parameters of the reverse-engineered nano-optical device as optimal parameters corresponding to a target value; a step in which, if the parameters of the reverse-engineered nano-optical device do not satisfy optimal conditions, the deep learning agent corrects the parameters of the nano-optical device; a step in which the transmission spectrum determining unit recreates the transmission spectrum of the nano-optical device using the corrected parameters; a step in which the reward determining unit recreates a reward to be input to a neural network based on the recreated transmission spectrum; and a step in which the deep learning agent performs deep learning with the neural network that receives the recreated reward to reverse-engineer the nano-optical device. Claim 11 A nano-optical device inverse engineering device comprising: a parameter setting unit for setting initial parameters of a nano-optical device; a transmission spectrum determining unit for determining the transmission spectrum of the nano-optical device using an interpretation model; a reward determining unit for determining a reward to be input to a neural network based on the transmission spectrum; and a deep learning agent for inverse engineering the nano-optical device by performing deep learning with the neural network that receives the reward, wherein the transmission spectrum determining unit determines the transmittance by setting the nano-optical device as a structure in which dielectrics are disposed only on each of the sides of the loops, in the case where the nano-optical device has a structure in which a plurality of polygonal loops are formed, each of the loops is formed with a variant dielectric, and the lower surface of each of the sides of the loops is connected to the dielectric constituting the sides of the loops. Claim 12 In claim 11, the interpretation model is a nano-optical device inverse engineering apparatus that determines the electric field of the nano-optical device according to the initial parameters of the nano-optical device, normalizes the electric field, and determines the transmittance of the nano-optical device based on the normalized electric field and the ratio of metal to dielectric included in the nano-optical device. Claim 13 In claim 11, the reward determination unit determines the reward by combining the interval between the target value of the peak position and the peak position of the process frequency, the target value of the interval between the peak position and the diff position and the difference between the interval between the peak position and the diff position of the process frequency, and the electric field amplification at the peak. This is a nano-optical device inverse engineering apparatus. Claim 14 delete Claim 15 In claim 11, the neural network is a nano-optical device inverse engineering device that receives a reward and parameters including the horizontal length of each of the polygonal loops, the vertical length of each of the polygonal loops, the spacing between the polygonal loops, the width of each of the sides of each of the loops, and the depth of each of the sides of each of the loops, and optimizes the parameters.

Images