Optical arithmetic device and method for controlling optical arithmetic device

Abstract

The present invention realizes an optical arithmetic device capable of performing more various optical arithmetic operations than conventional optical arithmetic devices. An optical arithmetic device (1) comprises: an image sensor (12ai) that includes a plurality of light receiving cells; an optical modulation element (11ai) that includes a plurality of cells in which phase modulation amounts can be set independently of each other; an optical branch element (13ai) that causes signal light to branch into monitor signal light which enters the image sensor (12ai) and arithmetic signal light which enters the optical modulation element (11ai); and a control unit (14) that sets the phase modulation amount of the cells in the optical modulation element (11ai) in accordance with the intensity of the monitor signal light detected at the light receiving cells in the image sensor (12ai) corresponding to said cells.

Description

[Technical field]

[0001] The present invention relates to an optical arithmetic device including an optical modulation element, and a method for controlling such an optical arithmetic device. [Background technology]

[0002] There is known an optical modulation element that has multiple cells and is designed to optically execute a predetermined operation by causing the signal light transmitted through each cell to interfere with each other. Optical operations using such optical modulation elements have the advantage of being faster and consuming less power than electrical operations using a processor. In addition, multiple-stage optical operations can be realized by sequentially applying the signal light to two or more optical modulation elements arranged side by side.

[0003] Patent Document 1 discloses an optical neural network having an input layer, an intermediate layer, and an output layer. The above-mentioned optical modulation element can be used, for example, as the intermediate layer of such an optical neural network. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] U.S. Patent No. 7,847,225 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the phase modulation performed in each cell of the light modulation element is a linear phase modulation (φ→φ+Δφ) in which a predetermined phase Δφ is added to the phase φ of the signal light before modulation. Therefore, the optical calculations that can be realized by the light modulation element are only limited. For example, each neuron that constitutes a neural network performs a threshold calculation using an activation function, but it is difficult to realize such a threshold calculation by the light modulation element.

[0006] One aspect of the present invention has been made in consideration of the above-mentioned problems, and its purpose is to realize an optical arithmetic device capable of performing more diverse optical arithmetic operations than conventional optical arithmetic devices that perform linear phase modulation in each cell of an optical modulation element. [Means for solving the problem]

[0007] An optical computing device according to one embodiment of the present invention includes an image sensor including at least one light receiving cell, an optical modulation element including a plurality of cells whose phase modulation amounts can be set independently of each other, an optical branching element that branches signal light into a monitor signal light that is incident on the image sensor and an operation signal light that is incident on the optical modulation element, and a control unit that sets the phase modulation amount of each cell of the optical modulation element in accordance with the intensity of the monitor signal light detected by a light receiving cell of the image sensor corresponding to that cell.

[0008] A control method according to another embodiment of the present invention is a control method for an optical calculation device comprising an image sensor including at least one light receiving cell, an optical modulation element including a plurality of cells whose phase modulation amounts can be set independently of each other, and an optical branching element that branches signal light into a monitor signal light that is incident on the image sensor and a calculation signal light that is incident on the optical modulation element, and includes a step of setting the phase modulation amount of each cell of the optical modulation element in accordance with the intensity of the monitor signal light detected by a light receiving cell of the image sensor corresponding to that cell. Effect of the Invention

[0009] According to one aspect of the present invention, it is possible to realize an optical arithmetic device capable of performing more diverse optical arithmetic operations than conventional optical arithmetic devices that perform linear phase modulation in each cell of an optical modulation element. [Brief description of the drawings]

[0010] [Figure 1] 1 is a block diagram showing a configuration of an optical arithmetic device according to an embodiment of the present invention; [Diagram 2] FIG. 2 is a block diagram showing a modified example of the optical arithmetic device shown in FIG. [Diagram 3]FIG. 2(a) is a plan view of a reflective light modulation element included in the optical computing device shown in FIG. 1, and FIG. 2(b) is a cross-sectional view of a cell constituting the light modulation element. [Figure 4] FIG. 3(a) is a plan view of a transmissive light modulation element included in the optical computing device shown in FIG. 2, and FIG. 3(b) is a cross-sectional view of a cell constituting the light modulation element. [Diagram 5] 1. FIG. 4 is a cross-sectional view of another reflective light modulation element included in the optical computing device shown in FIG. [Figure 6] 3 is a cross-sectional view of another transmissive optical modulation element included in the optical computing device shown in FIG. 2. [Figure 7] (a) and (b) are block diagrams of a configuration in which an integrated circuit is used instead of a microcomputer as the control unit shown in Fig. 1. In (a), each cell of the light modulation element and each light receiving cell of the image sensor correspond one-to-one, and in (b), each cell of the light modulation element and each light receiving cell of the image sensor correspond one-to-four. [Figure 8] FIG. 8 is a circuit diagram showing the adder circuit shown in FIG. 7(b). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] (Configuration of Optical Computing Unit) The configuration of an optical arithmetic unit 1 according to an embodiment of the present invention will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the optical arithmetic unit 1.

[0012] As shown in FIG. 1, the optical calculation device 1 includes a light modulation element group 11, an image sensor group 12, a light branching element group 13, and a control unit .

[0013] The light modulation element group 11 is a set of at least one reflective light modulation element 11a1-11an (n is any natural number equal to or greater than 1). In this embodiment, in order to realize multi-stage optical calculation, a set of three light modulation elements 11a1-11a3 is used as the light modulation element group 11. The reflective light modulation element 11ai (i is a natural number equal to or greater than 1 and equal to or less than n) is a plate-like element composed of a plurality of cells whose phase modulation amount can be set independently of each other, and performs optical calculation by interfering with signal light reflected by each cell while being phase-modulated. Here, performing optical calculation refers to converting the two-dimensional intensity distribution of the signal light from a two-dimensional intensity distribution representing information before calculation to a two-dimensional intensity distribution representing information after calculation. An example of the configuration of the reflective light modulation element 11ai will be described later with reference to a different drawing.

[0014] In this embodiment, the light modulation elements 11a1-11an are arranged side by side such that one of the main surfaces (signal light incident / exiting surface) of each light modulation element 11ai is located on the same plane P. The light modulation elements 11a1-11an may be integrated. For example, the light modulation elements 11a1-11an may be embedded in a single substrate, or n regions of a single light modulation element may be used as the light modulation elements 11a1-11an. Alternatively, n light diffraction layers formed in a structure that transmits signal light, such as a dry gel, may be used as the light modulation elements 11a1-11an.

[0015] The image sensor group 12 is a set of image sensors 12a1-12an, the number of which is the same as that of the light modulation elements 11a1-11an. In FIG. 1, a set of three image sensors 12a1-12a3 is illustrated as an example of the image sensor group 12. Each image sensor 12ai is a plate-shaped element including at least one light receiving cell, and detects a two-dimensional intensity distribution of the signal light. In this embodiment, each cell of the light modulation element 11ai and each light receiving cell of the image sensor 12ai correspond one-to-one. For example, when the light modulation element 11ai is composed of 200×200 cells arranged in a matrix, the image sensor 12ai is also composed of 200×200 light receiving cells arranged in a matrix. Therefore, the intensity of the signal light incident on each cell of the light modulation element 11ai is detected by the light receiving cell of the image sensor 12ai corresponding to that cell. Specific examples of the image sensors 12a1 to 12an include a charge coupled device (CCD) camera and a complementary metal oxide semiconductor (CMOS) camera.

[0016] In this embodiment, the image sensors 12a1-12an are arranged side by side such that one of the main surfaces (the incident surface of the signal light) of each image sensor 12ai is located on the same plane Q. Here, the plane Q is a plane parallel to the above-mentioned plane P. The image sensors 12a1-12an may be integrated. For example, the image sensors 12a1-12an may be embedded in a single substrate, or n regions of a single image sensor may be used as the image sensors 12a1-12an.

[0017] The optical branching element group 13 is a collection of optical branching elements 13a1-13an, the number of which is the same as the number of optical modulation elements 11a1-11an. In Fig. 1, a collection of three optical branching elements 13a1-13a3 is illustrated as an example of the optical branching element group 13. Each optical branching element 13ai branches the signal light into a monitor signal light and an operation signal light. Each optical branching element 13ai may be a half mirror or a beam splitter.

[0018] In this embodiment, the light branching elements 13a1 to 13an are arranged side by side such that the reflecting surfaces of the light branching elements 13ai are located on the same plane R. Here, the plane R is a plane that is located between the above-mentioned planes P and Q and is parallel to the above-mentioned planes P and Q.

[0019] The control unit 14 is configured to control the phase modulation amount of each cell of each light modulation element 11ai according to the intensity of the signal light detected by each light receiving cell of the image sensor 12ai corresponding to that cell. In this embodiment, a microcomputer, which is one aspect of a computer, is used as the control unit 14. However, in one aspect of the present invention, a personal computer, which is another aspect of a computer, or an integrated circuit can be used as the control unit 14. In addition, when an integrated circuit is used as the control unit 14, the integrated circuit may be one whose logic circuit is rewritable (for example, an FPGA (Field-Programmable Gate Array)) or one whose logic circuit is non-rewritable. A non-rewritable integrated circuit is a collection of electronic circuits composed of resistors, amplifiers, etc., that is, an integrated circuit. The control unit 14 composed of such an integrated circuit will be described later with reference to FIG. 7 and FIG. 8.

[0020] The signal light L0 input to the optical arithmetic device 1 is split by the optical branching element 13a1 into a monitor signal light L0 m and the calculation signal light L0 c Here, the monitor signal light L0 m is the signal light transmitted through the optical branching element 13a1, and the calculation signal light L0 c is the signal light reflected by the optical branching element 13a1. m On the other hand, the calculation signal light L0 c At this time, the control unit 14 controls the calculation signal light L0 cThe amount of phase modulation of each cell of the light modulation element 11a1 acting on the monitor signal light L0 detected by the light receiving cell of the image sensor 12a1 corresponding to that cell is m Set according to the strength of the

[0021] The signal light L1 reflected by the optical modulation element 11a1 is split by the optical branching element 13a2 into monitor signal light L1 m and the calculation signal light L1 c Here, the monitor signal light L1 m is the signal light transmitted through the optical branching element 13a2, and the calculation signal light L1 c is the signal light reflected by the optical branching element 13a2. m On the other hand, the calculation signal light L1 c At this time, the control unit 14 controls the calculation signal light L1 c The amount of phase modulation of each cell of the light modulation element 11a2 acting on the monitor signal light L1 detected by the light receiving cell of the image sensor 12a2 corresponding to that cell is m Set according to the strength of the

[0022] The signal light L2 reflected by the optical modulation element 11a2 is split by the optical branching element 13a3 into monitor signal light L2 m and the calculation signal light L2 c Here, the monitor signal light L2 m is the signal light transmitted through the optical branching element 13a3, and the calculation signal light L2 c is the signal light reflected by the optical branching element 13a3. m On the other hand, the calculation signal light L2 c At this time, the control unit 14 controls the calculation signal light L2 c The amount of phase modulation of each cell of the light modulation element 11a3 acting on the monitor signal light L2 detected by the light receiving cell of the image sensor 12a3 corresponding to that cell is m Set according to the strength of the

[0023] The signal light L3 reflected by the light modulation element 11a3 is outputted to the outside as a signal light representing the result of the calculation by the optical calculation device 1.

[0024] In this embodiment, a configuration is adopted in which the signal light L0 representing the input information is input to the optical arithmetic device 1, but the present invention is not limited to this. For example, a configuration can be adopted in which an electrical signal representing the input information and a carrier light are input to the optical arithmetic device 1. In this case, for example, an optical modulation element 11a0 (shown by a dotted line in FIG. 1) that generates the signal light L0 representing the input signal by optically modulating the carrier light with the electrical signal representing the input information may be added to the front stage of the optical modulation element 11a1.

[0025] The amount of phase modulation of each cell of the optical modulation element included in the conventional optical computing device is constant, regardless of the intensity (light intensity) P of the signal light incident on that cell. Therefore, if the phase of the signal light incident on that cell is φ and the amount of phase modulation in that cell is Δφ, the phase φ' of the signal light emitted from that cell is given by φ'=φ+Δφ (constant). In other words, if the signal light incident on the cell is expressed as a complex number P×e i×φ If the signal light emitted from the cell is expressed as a complex number P×e i×φ ×e i×Δθ It is expressed as:

[0026] On the other hand, the phase modulation amount of each cell of the optical modulation element 11ai included in the optical arithmetic device 1 according to this embodiment is i-1 c In other words, the intensity P of the monitor signal light L detected by the light receiving cell of the image sensor 12ai corresponding to that cell is i-1 m Therefore, the signal light for calculation L incident on the cell is i-1 c The phase of the signal light L emitted from the cell is expressed as φ and the phase modulation amount in the cell is expressed as Δφ(P). i The phase φ' of the signal light for calculation L incident on each cell is given by φ'=φ+Δφ(P). i-1 c to the complex number Pe iφWhen expressed by, the signal light L emitted from the cell i is a complex number P×e i×φ ×e i×Δθ(P) is expressed by. Therefore, according to the optical arithmetic unit 1 according to the present embodiment, it is possible to perform a wider variety of optical arithmetic operations than a conventional optical arithmetic unit that performs linear phase modulation in each cell of the optical modulation element.

[0027] Note that the intensity P i-1 c of the arithmetic signal light L c and the intensity P i-1 m of the monitor signal light L m may or may not match. However, even in the latter case, there is a proportional relationship P c =ρP m between these intensities, so the control of the phase modulation amount corresponding to the intensity P i-1 c of the arithmetic signal light L c can be performed. Here, ρ is a constant determined according to the branching ratio (transmittance and reflectance) of the optical branching element 13ai, etc.

[0028] Note that in the present embodiment, the control unit 14 sets the phase modulation amount Δφ(P) of each cell of the optical modulation element 11ai according to Δφ(P)=P×α+β. Here, α and β are constants determined for each cell of the optical modulation element 11ai. However, the present invention is not limited to this. That is, the relationship between the phase modulation amount Δφ(P) of each cell and the intensity P of the monitor signal light L i-1 m detected by the light receiving cell of the image sensor 12ai corresponding to that cell is arbitrary. For example, the control unit 14 may set the phase modulation amount Δφ(P) of each cell of the optical modulation element 11ai such that Δφ(P)=0 when P<P0 and Δφ(P)=γ when P≧P0. Here, γ is a constant determined for each cell of the optical modulation element 11ai.

[0029] The constant part of the phase modulation amount of each cell of the light modulation element 11ai can be set by using, for example, machine learning. As a model used in this machine learning, for example, the two-dimensional intensity distribution of the signal light L0 input to the light modulation element 11a1 is input, and the signal light L0 output from the light modulation element 11an is n A model that outputs a two-dimensional intensity distribution of the signal light L0 input to the optical modulation element 11a1 may be used, which includes constant parts of the phase modulation amount in each cell (α, β, and γ in the above example) as parameters. Here, the two-dimensional intensity distribution of the signal light L0 input to the optical modulation element 11a1 refers to, for example, a set of intensities of the signal light input to each cell constituting the optical modulation element 11a1. Also, the signal light L0 output from the optical modulation element 11an may be used n The two-dimensional intensity distribution refers to, for example, a collection of intensities of signal light input to each cell constituting a two-dimensional image sensor arranged downstream of the light modulation element 11an.

[0030] In addition, the monitor signal light L by each image sensor 12ai i-1 m , the control unit 14 controls each of the light modulation elements 11ai, and each of the light modulation elements 11ai detects the signal light L i-1 c Therefore, the phase modulation amount of all the cells of each light modulation element 11ai converges to a constant value, and the signal light L expected from the optical calculation device 1 n Considering this, the control unit 14 controls the control circuit 14 to output the signal light L, which represents the correct calculation result, after the phase modulation amount of all the cells of each light modulation element 11ai converges to a constant value. n It is preferable to output a signal (which may be an optical signal or an electrical signal) indicating that the phase modulation amount of all the cells of each light modulation element 11ai is being output. In this way, at the output destination of this signal, the signal light L representing an incorrect calculation result that may be output from the optical calculation device 1 before the phase modulation amount of all the cells of each light modulation element 11ai converges to a certain value is output. n The risk of being referred to can be reduced. In addition, the output destination of this signal can be, for example, the signal light L nExamples of the control unit of a device that uses the information represented by the information include a control unit of the device (which may be a computer connected to the device, or a microcomputer or integrated circuit built into the device).

[0031] In addition, in this embodiment, a configuration is adopted in which each cell of the light modulation element 11ai and each light receiving cell of the image sensor 12ai correspond one-to-one, but the present invention is not limited to this. For example, a configuration may be adopted in which each cell of the light modulation element 11ai and each light receiving cell of the image sensor 12ai correspond one-to-K (K is a natural number of 2 or more). In this case, the control unit 14 calculates the phase modulation amount Δφ(P) of one cell of the light modulation element 11ai based on the monitoring signal light L detected by the K light receiving cells of the image sensor 12ai corresponding to that cell. i-1 m The value of K is set according to the intensity P of the monitor signal light L detected by one light receiving cell of the image sensor 12ai. The case where K=4 will be described later with reference to FIG. 7(b). Alternatively, a configuration may be adopted in which each cell of the light modulation element 11ai corresponds to each light receiving cell of the image sensor 12ai in an L-to-1 correspondence (L is a natural number of 2 or more). In this case, the control unit 14 calculates the phase modulation amount Δφ(P) of the L cells of the light modulation element 11ai based on the intensity P of the monitor signal light L detected by one light receiving cell of the image sensor 12ai corresponding to those cells. i-1 m L may be the total number of cells constituting the light modulation element 11ai. In this case, the phase modulation amount Δφ(P) of all the cells of the light modulation element 11ai is set according to the intensity P of the monitor signal light L detected by one light receiving cell of the image sensor 12ai corresponding to those cells. i-1 m The value is set according to the strength P of the

[0032] In addition, when the n light diffraction layers formed in the dry gel are used as the light modulation elements 11a1-11an, it is preferable to use a gel that shrinks while maintaining a similar shape by dehydration shrinkage, for example, a gel used in the Implosion Fabrication method. This makes it possible to easily manufacture the light modulation element group 11 in which the light modulation elements 11a1-11an are arranged with high precision by drying the swollen gel in which the n light diffraction layers are formed.

[0033] (Example of the configuration of a reflective light modulation element) A configuration example of the reflective light modulation element 11ai included in the optical calculation device 1 will be described with reference to Fig. 3. Fig. 3(a) is a plan view of the light modulation element 11ai according to this example. Fig. 3(b) is a cross-sectional view of a microcell C constituting the light modulation element 11ai according to this example.

[0034] As shown in FIG. 3A, the optical modulation element 11ai is composed of a plurality of microcells C whose phase modulation amounts are set independently of each other. i-1 c When the light is incident on the microcell C, the signal light L for calculation is reflected by each microcell C while being phase modulated. i-1 c The signal light L representing the calculation result is generated by mutual interference. i is formed.

[0035] In this specification, the term "microcell" refers to a cell having a cell size of less than 10 μm, for example. Also, the term "cell size" refers to the square root of the area of ​​the cell. For example, when the shape of the microcell C in a plan view is a square, the cell size of the microcell C is the length of one side of the microcell C. The lower limit of the cell size of the microcell C is, for example, 1 nm.

[0036] 3A is composed of 200×200 microcells C arranged in a matrix. The shape of each microcell C in plan view is a square of 500 nm×500 nm, and the shape of the light modulation element 11ai in plan view is a square of 100 μm×100 μm.

[0037] Each microcell C constituting the light modulation element 11ai can be composed of a polarizing plate C11, a reflecting plate C12, a first electrode C13, a magnetization free layer C14, an insulating layer C15, a magnetization fixed layer C16, and a second electrode C17, for example, as shown in (b) of Figure 3.

[0038] The polarizing plate C11 and the reflecting plate C12 are arranged to face each other. The first electrode C13, the magnetization free layer C14, the insulating layer C15, the magnetization fixed layer C16, and the second electrode C17 are stacked in this order and sandwiched between the polarizing plate C11 and the reflecting plate C12. Here, the stacking direction of the first electrode C13, the magnetization free layer C14, the insulating layer C15, the magnetization fixed layer C16, and the second electrode C17 is perpendicular to the stacking direction of the polarizing plate C11 and the reflecting plate C12. Therefore, the first side surface of the magnetization free layer C14 is in surface contact with one of the main surfaces of the polarizing plate C11, and the second side surface of the magnetization free layer C14 facing the first side surface is in surface contact with one of the main surfaces of the reflecting plate C12. The calculation signal light L i-1 c (1) enters the inside of the magnetization free layer C14 via the polarizing plate C11, (2) is reflected by the reflecting plate C12, and (3) exits to the outside of the magnetization free layer C14 via the polarizing plate C11.

[0039] The magnetization free layer C14 is made of, for example, a conductive and light-transmitting soft magnetic material (for example, CoFeB). The magnetization fixed layer C16 is made of, for example, a conductive hard magnetic material (for example, Permalloy). The polarizer C11 is selected to selectively transmit a polarized light component whose polarization direction P is parallel to the magnetization direction M of the magnetization fixed layer C16. FIG. 3B illustrates a case where the magnetization direction M and the polarization direction P are parallel to both the main surface of the polarizer C11 and the main surface of the magnetization fixed layer C16.

[0040] When a potential difference is applied between the first electrode C13 and the second electrode C17, a spin current (a flow of spin-polarized electrons) is injected from the magnetization fixed layer C16 to the magnetization free layer C14 through the insulating layer C15 by the tunnel effect, and magnetization occurs in the magnetization free layer C14. Here, the magnetization generated in the magnetization free layer C14 is magnetization parallel to the magnetization direction M of the magnetization fixed layer C16, that is, the magnetization direction of the calculation signal light L incident on the magnetization free layer C14 through the polarizing plate C11. i-1 c The magnetization is parallel to the polarization direction P of the signal light for calculation L. i-1 c The phase of the magnetic field is delayed by the transverse Kerr effect while propagating through the magnetic free layer C14.

[0041] Here, the signal light L for calculation in cell C i-1 c The amount of phase change of the magnetization free layer C14 is determined according to the magnitude of magnetization generated in the magnetization free layer C14. The magnitude of magnetization generated in the magnetization free layer C14 is determined according to the magnitude of the spin current injected into the magnetization free layer C14. The magnitude of the spin current injected into the magnetization free layer C14 is determined according to the potential difference applied between the first electrode C13 and the second electrode C17. Therefore, by controlling the potential difference applied between the first electrode C13 and the second electrode C17, the phase modulation of the cell C can be set to a desired value.

[0042] In this configuration example, the microcell C having the same configuration as the STT (Spin Transfer Torque) type MRAM (Magnetoresistive Random Access Memory) has been described, but the present invention is not limited thereto. For example, the microcell C having the same configuration as the SOT (Spin Orbit Torque) type MRAM may be used. Such a microcell C can be realized by removing the insulating layer C15, the magnetization fixed layer C16, and the second electrode C17 from the structure shown in FIG. 3(b). In this case, for example, the first electrode C13 contains a heavy metal, and a pulse voltage or pulse current is applied to the first electrode C13, thereby efficiently injecting a spin current into the magnetization free layer C14.

[0043] The cell C may be configured similarly to a LCOS (Liquid Crystal On Silicon) cell. In this case, the cell C is configured, for example, as shown in FIG. 5, by a glass substrate C21, a transparent electrode C22, a liquid crystal layer C23, and a reflective electrode C24. The glass substrate C21, the transparent electrode C22, the liquid crystal layer C23, and the reflective electrode C24 are configured to reflect the signal light L for calculation. i-1 c The layers are stacked in this order from the incident side.

[0044] Signal light L for calculation i-1 c (1) passes through the glass substrate C21 and the transparent electrode C22, (2) passes through the liquid crystal layer C23, (3) is reflected by the reflective electrode C24, (4) passes through the liquid crystal layer C23, and (5) passes through the transparent electrode C22 and the glass substrate C21. The liquid crystal layer C23 is composed of liquid crystal molecules oriented in a direction parallel to the main surface of the glass substrate C21, and has a refractive index according to the potential difference between the transparent electrode C22 and the reflective electrode C24. For this reason, the calculation signal light L i-1 c is subjected to phase modulation when passing through the liquid crystal layer C23. The amount of phase modulation of the cell C can be set to a desired value by controlling the potential difference applied between the transparent electrode C22 and the reflective electrode C24.

[0045] (Modification of the optical computing device) A modified example of the optical arithmetic device 1 (hereinafter, referred to as optical arithmetic device 1A) will be described with reference to Fig. 2. Fig. 2 is a block diagram showing the configuration of the optical arithmetic device 1A.

[0046] The optical arithmetic device 1A, like the optical arithmetic device 1, includes a light modulation element group 11, an image sensor group 12, a light branching element group 13, and a control unit 14. The optical arithmetic device 1 differs from the optical arithmetic device 1A in (1) the light modulation element group 11 is composed of transmissive light modulation elements 11b1-11bn, and (2) the arrangement of the light modulation elements 11b1-11bn included in the light modulation element group 11 and the light branching elements 13a1-13an included in the light branching element group 13.

[0047] The transmissive light modulation element 11bi is composed of a plurality of cells whose phase modulation amounts can be set independently of each other, and performs optical calculations by causing the signal light transmitted through each cell to interfere with each other while being phase modulated. An example of the configuration of the transmissive light modulation element 11bi will be described later with reference to a different drawing.

[0048] In this modification, the light modulation elements 11b1 to 11bn are arranged side by side on a straight line S such that the principal surface (the incident surface and the exit surface of the signal light) of each light modulation element 11bi is perpendicular to a plane Q. Here, the plane Q is a plane on which one principal surface (the incident surface of the signal light) of each image sensor 12ai is arranged, and the straight line S is a straight line parallel to the plane Q.

[0049] The optical branching element 13a1 is disposed on the straight line S before the optical modulation element 11b1. The optical branching element 13aj (j is a natural number between 2 and n) is disposed on the straight line S after the optical modulation element 11bj-1 and before the optical modulation element 11bj. The optical branching element 13ai is oriented such that the normal to the reflecting surface of the optical branching element 13ai is included in the plane T, the reflecting surface of the optical branching element 13ai and the straight line S form an angle of 45°, and the monitoring signal light L i-1 m is determined to face the image sensor 12ai. Here, the plane T is a plane that includes the straight line S and is perpendicular to the plane Q.

[0050] The signal light L0 input to the optical arithmetic device 1A is split by the optical branching element 13a1 into monitor signal light L0 m and the calculation signal light L0 c Here, the monitor signal light L0 m is the signal light reflected by the optical branching element 13a1, and the calculation signal light L0 c is the signal light transmitted through the optical branching element 13a1. m On the other hand, the calculation signal light L0 c At this time, the control unit 14 controls the calculation signal light L0 cThe amount of phase modulation of each cell of the light modulation element 11b1 acting on the monitor signal light L0 detected by the light receiving cell of the image sensor 12a1 corresponding to that cell is m Set according to the strength of the

[0051] The signal light L1 transmitted through the optical modulation element 11b1 is split by the optical branching element 13a2 into monitor signal light L1 m and the calculation signal light L1 c Here, the monitor signal light L1 m is the signal light reflected by the optical branching element 13a2, and the calculation signal light L1 c is the signal light transmitted through the optical branching element 13a2. m On the other hand, the calculation signal light L1 c At this time, the control unit 14 controls the calculation signal light L1 c The amount of phase modulation of each cell of the light modulation element 11b2 acting on the monitor signal light L1 detected by the light receiving cell of the image sensor 12a2 corresponding to that cell is m Set according to the strength of the

[0052] The signal light L2 transmitted through the optical modulation element 11b2 is split by the optical branching element 13a3 into monitor signal light L2 m and the calculation signal light L2 c Here, the monitor signal light L2 m is the signal light reflected by the optical branching element 13a3, and the calculation signal light L2 c is the signal light transmitted through the optical branching element 13a3. m On the other hand, the calculation signal light L2 c At this time, the control unit 14 controls the calculation signal light L2 c The amount of phase modulation of each cell of the light modulation element 11b3 acting on the monitor signal light L2 detected by the light receiving cell of the image sensor 12a3 corresponding to that cell is m Set according to the strength of the

[0053] The signal light L3 transmitted through the light modulation element 11b3 is outputted to the outside as a signal light representing the result of the calculation by the optical calculation device 1A.

[0054] In this embodiment, a configuration is adopted in which the signal light L0 representing the input information is input to the optical arithmetic device 1A, but the present invention is not limited to this. For example, a configuration can be adopted in which an electrical signal representing the input information and a carrier light are input to the optical arithmetic device 1A. In this case, for example, an optical modulation element 11b0 (shown by a dotted line in FIG. 2) that generates the signal light L0 representing the input signal by optically modulating the carrier light with the electrical signal representing the input information may be added to the front stage of the optical modulation element 11b1.

[0055] The phase modulation amount of each cell of the optical modulation element 11bi included in the optical calculation device 1A according to this modification is the calculation signal light L i-1 c In other words, the intensity P of the monitor signal light L detected by the light receiving cell of the image sensor 12ai corresponding to that cell is i-1 m Therefore, the signal light for calculation L incident on the cell is i-1 c The phase of the signal light L emitted from the cell is expressed as φ and the phase modulation amount in the cell is expressed as Δφ(P). i The phase φ′ of the signal light for calculation L incident on each cell is given by φ′=φ+Δφ(P). i-1 c to the complex number Pe iφ When expressed by, the signal light L emitted from the cell i is the complex number Pe iφ ×e iΔθ(P) For this reason, the optical arithmetic device 1A according to this modification can perform more diverse optical arithmetic operations than the conventional optical arithmetic device that performs linear phase modulation in each cell of the optical modulation element.

[0056] In addition, the calculation signal light L i-1 c The strength of P c and monitor signal light L i-1 m The strength of Pm Sometimes they match, and sometimes they don't. However, even in the latter case, there is a proportional relationship P c =ρP m between these intensities, so that the phase modulation amount can be controlled according to the intensity P i-1 c of the operation signal light L c . Here, ρ is a constant determined according to the branching ratio (transmittance and reflectance) of the optical branching element 13ai, etc.

[0057] In this modification, the control unit 14 sets the phase modulation amount Δφ(P) of each cell of the optical modulation element 11bi according to Δφ(P)=P×α + β. Here, α and β are constants determined for each cell of the optical modulation element 11bi. However, the present invention is not limited to this. That is, the relationship between the phase modulation amount Δφ(P) of each cell and the intensity P i-1 m of the monitor signal light L detected by the light receiving cell of the image sensor 12ai corresponding to that cell is arbitrary. For example, the control unit 14 may set the phase modulation amount Δφ(P) of each cell of the optical modulation element 11bi such that Δφ(P)=0 when P<P0 and Δφ(P)=γ when P≧P0. Here, γ is a constant determined for each cell of the optical modulation element 11bi.

[0058] The setting of the constant part of the phase modulation amount of each cell of the optical modulation element 11bi can be realized, for example, using machine learning. As a model used in this machine learning, for example, a model that takes the two-dimensional intensity distribution of the signal light L0 input to the optical modulation element 11b1 as an input and outputs the two-dimensional intensity distribution of the signal light L n output from the optical modulation element 11bn can be used, and a model including the constant part (α, β, and γ in the above example) of the phase modulation amount in each cell as a parameter. Here, the two-dimensional intensity distribution of the signal light L0 input to the optical modulation element 11b1 refers to, for example, the set of intensities of the signal light input to each cell constituting the optical modulation element 11b1. Also, the signal light L nThe two-dimensional intensity distribution refers to, for example, a collection of intensities of signal light input to each cell constituting a two-dimensional image sensor arranged downstream of the light modulation element 11bn.

[0059] (Example of the configuration of a transmissive light modulation element) A configuration example of the transmissive light modulation element 11bi included in the optical calculation device 1A will be described with reference to Fig. 4. Fig. 4(a) is a plan view of the light modulation element 11bi according to this example. Fig. 4(b) is a cross-sectional view of a microcell C constituting the light modulation element 11bi according to this example.

[0060] As shown in FIG. 4A, the optical modulation element 11bi is composed of a plurality of microcells C whose phase modulation amounts are set independently of each other. i-1 c When the light is incident on the microcell C, the signal light L for calculation is transmitted through each microcell C while being phase modulated. i-1 c The signal light L representing the calculation result is generated by mutual interference. i is formed.

[0061] 4A is composed of 200×200 microcells C arranged in a matrix. The shape of each microcell C in plan view is a square of 500 nm×500 nm, and the shape of the light modulation element 11bi in plan view is a square of 100 μm×100 μm.

[0062] Each microcell C constituting the light modulation element 11bi can be composed of, for example, a polarizing plate C11, a polarizing plate C18, a first electrode C13, a magnetization free layer C14, an insulating layer C15, a magnetization fixed layer C16, and a second electrode C17, as shown in (b) of Figure 4.

[0063] The polarizing plate C11 and the polarizing plate C18 are arranged to face each other. The first electrode C13, the magnetization free layer C14, the insulating layer C15, the magnetization fixed layer C16, and the second electrode C17 are stacked in this order and sandwiched between the polarizing plate C11 and the polarizing plate C18. Here, the stacking direction of the first electrode C13, the magnetization free layer C14, the insulating layer C15, the magnetization fixed layer C16, and the second electrode C17 is perpendicular to the stacking direction of the polarizing plate C11 and the polarizing plate C18. Therefore, the first side surface of the magnetization free layer C14 is in surface contact with one of the main surfaces of the polarizing plate C11, and the second side surface of the magnetization free layer C14 facing the first side surface is in surface contact with one of the main surfaces of the polarizing plate C18. The calculation signal light L i-1 c (1) enters the inside of the magnetization free layer C14 via the polarizing plate C11, (2) passes through the magnetization free layer C14, and (3) exits from the magnetization free layer C14 via the polarizing plate C18.

[0064] The magnetization free layer C14 is made of, for example, a soft magnetic material (e.g., CoFeB) having electrical conductivity and light transmission. The magnetization fixed layer C16 is made of, for example, a hard magnetic material (e.g., Permalloy) having electrical conductivity. The polarizers C11 and C18 are selected as polarizers that selectively transmit the polarized light component whose polarization direction P is parallel to the magnetization direction M of the magnetization fixed layer C16. FIG. 4B illustrates a case where the magnetization direction M and the polarization direction P are parallel to both the main surface of the polarizer C11 and the main surface of the magnetization fixed layer C16.

[0065] When a potential difference is applied between the first electrode C13 and the second electrode C17, a spin current (a flow of spin-polarized electrons) is injected from the magnetization fixed layer C16 to the magnetization free layer C14 through the insulating layer C15 by the tunnel effect, and magnetization occurs in the magnetization free layer C14. Here, the magnetization generated in the magnetization free layer C14 is magnetization parallel to the magnetization direction M of the magnetization fixed layer C16, that is, the magnetization direction of the calculation signal light L incident on the magnetization free layer C14 through the polarizing plate C11. i-1 c The magnetization is parallel to the polarization direction P of the signal light for calculation L. i-1 c The phase of the magnetic field is delayed by the transverse Kerr effect while propagating through the magnetic free layer C14.

[0066] Here, the signal light L for calculation in cell C i-1 c The amount of phase change of the magnetization free layer C14 is determined according to the magnitude of magnetization generated in the magnetization free layer C14. The magnitude of magnetization generated in the magnetization free layer C14 is determined according to the magnitude of the spin current injected into the magnetization free layer C14. The magnitude of the spin current injected into the magnetization free layer C14 is determined according to the potential difference applied between the first electrode C13 and the second electrode C17. Therefore, by controlling the potential difference applied between the first electrode C13 and the second electrode C17, the phase modulation of the cell C can be set to a desired value.

[0067] In this configuration example, the microcell C having the same configuration as the STT (Spin Transfer Torque) type MRAM (Magnetoresistive Random Access Memory) has been described, but the present invention is not limited thereto. For example, the microcell C having the same configuration as the SOT (Spin Orbit Torque) type MRAM may be used. Such a microcell C can be realized, for example, by removing the insulating layer C15, the magnetization fixed layer C16, and the second electrode C17 from the structure shown in FIG. 4(b). In this case, for example, the first electrode C13 contains a heavy metal, and a pulse voltage or pulse current is applied to the first electrode C13, thereby efficiently injecting a spin current into the magnetization free layer C14.

[0068] The cell C may be configured similarly to a cell of an LCD (Liquid Crystal Display). In this case, the cell C is configured by a first glass substrate C21a, a first transparent electrode C22a, a liquid crystal layer C23, a second transparent electrode C22b, and a second glass substrate C21b, as shown in Fig. 6. The first glass substrate C21a, the first transparent electrode C22a, the liquid crystal layer C23, the second transparent electrode C22b, and the second glass substrate C21b transmit the signal light L for calculation. i-1 c The layers are stacked in this order from the incident side.

[0069] Signal light L for calculation i-1 c (1) passes through the first glass substrate C21a and the first transparent electrode C22a, (2) passes through the liquid crystal layer C23, and (3) passes through the second transparent electrode C22b and the second glass substrate C21b. The liquid crystal layer C23 is composed of liquid crystal molecules oriented in a direction parallel to the main surface of the glass substrate C21, and has a refractive index according to the potential difference between the first transparent electrode C22a and the second transparent electrode C22b. For this reason, the calculation signal light L i-1 c undergoes phase modulation when passing through the liquid crystal layer C23. The amount of phase modulation of the cell C can be set to a desired value by controlling the potential difference applied between the first transparent electrode C22a and the second transparent electrode C22b.

[0070] (Modification of the control unit) A modified example of the control unit 14 shown in Fig. 1 will be described with reference to Fig. 7 and Fig. 8. Fig. 7(a) and (b) are block diagrams of a configuration in which an integrated circuit is used as the control unit 14 instead of a microcomputer. Fig. 7(a) and (b) show a light modulation element 11a1 as an example of the light modulation element 11ai, and an image sensor 12a1 as an example of the image sensor 12ai. Fig. 8 is a circuit diagram of the addition circuit SA shown in Fig. 7(b).

[0071] In FIG. 7(a), each cell C of the light modulation element 11a1 and each light receiving cell PDC of the image sensor 12a1 correspond one-to-one. On the other hand, in FIG. 7(b), each cell C of the light modulation element 11a1 and each light receiving cell PDC of the image sensor 12a1 correspond one-to-four. That is, when comparing the configuration in FIG. 7(b) with the configuration in FIG. 7(a), the resolution of the light receiving cell PDC in the image sensor 12a1 is twice as high vertically and horizontally, totaling four times. In addition, in FIG. 7(a) and (b), the arrows between the cell C and the light receiving cell PDC facing each other indicate that the cell C and the light receiving cell PDC facing each other face each other.

[0072] In this modification, two resistors R a ,R b By using the voltage divider circuit VD, the voltage corresponding to the intensity of the monitor signal light detected by the light receiving cell PDC is a ,R b The voltage is divided according to the ratio of the resistance values ​​of R b / (R a +R b ) is applied to the cell C. The voltage applied to the cell C from the voltage dividing circuit VD corresponds to α in the phase modulation amount Δφ(P)=P×α+β of each cell C of the light modulation element 11ai set by the control unit 14.

[0073] For example, if the light modulation element 11a1 is composed of 200 x 200 cells C and the image sensor 12a1 is composed of 200 x 200 light receiving cells PDC, the control unit 14 corresponding to the light modulation element 11a1 and the image sensor 12a1 is composed of an integrated circuit including 200 x 200 voltage division circuits VD.

[0074] In addition, when each cell C of the light modulation element 11a1 and each light receiving cell PDC of the image sensor 12a1 have an L-to-1 correspondence (i.e., cell C is composed of L subcells), the single voltage divider circuit VD shown in Figures 7(a) and 7(b) can be replaced with L voltage divider circuits arranged in parallel. Each voltage divider circuit corresponds to a subcell of cell C, and the voltage divided by each dividing circuit is applied to each subcell of cell C. In this way, by providing a voltage divider circuit corresponding to each subcell, the voltage to be divided and applied can be adjusted for each subcell.

[0075] Also, as shown in FIG. 7B, when each cell C of the light modulation element 11a1 and each light receiving cell PDC of the image sensor 12a1 correspond to one to four, each light receiving cell PDC is composed of sub light receiving cells PDC1 and PDC2 and two sub light receiving cells not shown in FIG. 7B. In this case, the intensity of the monitor signal light detected by the light receiving cell PDC corresponding to the cell C is obtained by averaging the intensity of the monitor signal light detected by the above-mentioned four sub light receiving cells. In the configuration shown in FIG. 7B, an adder circuit SA is used to obtain the average intensity of the monitor signal light detected by the four sub light receiving cells.

[0076] As shown in Fig. 8, the adder circuit SA includes two amplifiers AP1 and AP2 and seven resistors R1 to R7. In this modification, the adder circuit AP is configured so that the resistance values ​​of the resistors R1 to R4 are equal, the resistance value of R5 is 1 / 4 of the resistance values ​​of R1 to R4, and the resistance values ​​of the resistors R6 and R7 are equal. When voltages Vin1 to Vin4 corresponding to the intensities of the monitor signal lights detected by the four sub-light receiving cells are input to the four input ports of the adder circuit AP configured in this way, Vm = -(Vin1 + Vin2 + Vin3 + Vin4) / 4 is obtained at the intermediate port, and Vout = (Vin1 + Vin2 + Vin3 + Vin4) / 4 is obtained at the output port.

[0077] In this modification, an image sensor including a light receiving cell PDC that outputs a voltage corresponding to the intensity of the detected monitor signal light is used as the image sensor 12ai. However, in one aspect of the present invention, an image sensor including a light receiving cell PDC that outputs a current (charge) corresponding to the intensity of the detected monitor signal light can also be used as the image sensor 12ai. An example of an image sensor including a light receiving cell PDC that outputs a voltage is a CMOS camera, and an example of an image sensor including a light receiving cell PDC that outputs a current is a CCD camera. When an image sensor including a light receiving cell PDC that outputs a current is used, a transimpedance amplifier is provided in the rear stage of the light receiving cell PDC to convert the output current into a voltage. According to this configuration, the voltage divider circuit VD and the adder circuit SA described above can be applied in the same way as when a light receiving cell that outputs a voltage is used.

[0078] (summary) The optical computation device according to aspect 1 of the present invention comprises an image sensor including at least one light receiving cell, an optical modulation element including a plurality of cells whose phase modulation amounts can be set independently of each other, an optical branching element that branches signal light into a monitor signal light that is incident on the image sensor and an computation signal light that is incident on the optical modulation element, and a control unit that sets the phase modulation amount of each cell of the optical modulation element in accordance with the intensity of the monitor signal light detected by a light receiving cell of the image sensor corresponding to that cell.

[0079] According to the above configuration, it is possible to perform more diverse optical operations than in a conventional optical operation device that performs linear phase modulation in each cell of the optical modulation element.

[0080] In the optical computing device according to aspect 2 of the present invention, in addition to the configuration of aspect 1, the control unit sets the phase modulation amount Δφ(P) of each cell of the optical modulation element according to Δφ(P) = P × α + β (where P is the intensity of the monitor signal light detected in the cell of the image sensor corresponding to that cell, and α and β are constants).

[0081] According to the above configuration, it is possible to perform more diverse optical operations than in a conventional optical operation device that performs linear phase modulation in each cell of the optical modulation element.

[0082] In an optical arithmetic device according to aspect 3 of the present invention, in addition to the configuration of aspect 1 or 2, the optical arithmetic device is provided with a plurality of reflective optical modulation elements as the optical modulation elements, the monitor signal light is signal light transmitted through the optical branching element, and the arithmetic signal light is signal light reflected by the optical branching element.

[0083] According to the above configuration, multi-stage optical calculations can be performed using a reflective light modulation element.

[0084] An optical computing device according to a fourth aspect of the present invention employs, in addition to the configuration of the third aspect, a configuration in which the plurality of reflective optical modulation elements are integrated.

[0085] According to the above configuration, the number of parts of the optical calculation device can be reduced. Also, alignment adjustment between the reflective optical modulation elements, which is required when a plurality of reflective optical modulation elements are separate, can be omitted.

[0086] In an optical arithmetic device according to aspect 5 of the present invention, in addition to the configuration of aspect 1 or 2, the optical arithmetic device is configured to include a plurality of transmissive optical modulation elements as the optical modulation elements, the monitor signal light is signal light reflected by the optical branching element, and the arithmetic signal light is signal light transmitted through the optical branching element.

[0087] According to the above configuration, multi-stage optical calculations can be performed using a transmissive light modulation element.

[0088] In an optical computing device according to aspect 6 of the present invention, in addition to the configuration of any one of aspects 1 to 5, the optical computing device has a configuration in which a combination of the optical modulation element, the image sensor, and the optical branching element is provided in multiple stages, and the optical branching element in each stage from the second stage onwards receives as input signal light generated by the optical modulation element in the stage preceding that stage.

[0089] According to the above configuration, multi-stage optical calculations can be performed.

[0090] In an optical computing device according to aspect 7 of the present invention, in addition to the configuration of any one of aspects 1 to 6, a configuration is adopted in which the control unit outputs a signal indicating that signal light representing a correct computation result is being output after the phase modulation amounts of all cells of the optical modulation element converge to a constant value.

[0091] According to the above configuration, at the output destination of the signal, it is possible to reduce the risk that a signal light representing an incorrect calculation result, which may be output from the optical calculation device before the phase modulation amount of all cells of the optical modulation element converges to a certain value, will be referenced.

[0092] A control method according to an eighth aspect of the present invention is a control method for an optical calculation device comprising an image sensor including at least one light receiving cell, an optical modulation element including a plurality of cells whose phase modulation amounts can be set independently of each other, and an optical branching element that branches signal light into a monitor signal light that is incident on the image sensor and a calculation signal light that is incident on the optical modulation element, and includes a step of setting the phase modulation amount of each cell of the optical modulation element in accordance with the intensity of the monitor signal light detected by a light receiving cell of the image sensor corresponding to that cell.

[0093] According to the above method, it is possible to perform more diverse optical operations than in the conventional optical operation device which performs linear phase modulation in each cell of the optical modulation element.

[0094] (Additional Notes) The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means included in the above-described embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0095] 1 Optical calculation device 11 Light modulation elements 11a1~11an Light modulation element (reflective type) 11b1~11bn Light modulation element (transmissive type) 12 Image sensors 12a1~12an image sensor 13 Optical branching element group 13a1~13an Optical branching element

Claims

1. an image sensor including at least one light receiving cell; an optical modulation element including a plurality of cells whose phase modulation amounts can be set independently of each other; an optical branching element that branches a signal light into a monitor signal light that is incident on the image sensor and an arithmetic signal light that is incident on the optical modulation element; a control unit that sets the amount of phase modulation of each cell of the light modulation element in accordance with the intensity of the monitor signal light detected by a light receiving cell of the image sensor corresponding to the cell. An optical computing device comprising:

2. The control unit sets the phase modulation amount Δφ(P) of each cell of the light modulation element according to Δφ(P)=P×α+β (where P is the intensity of the monitor signal light detected by the light receiving cell of the image sensor corresponding to the cell, and α and β are constants); 2. The optical computing device according to claim 1 .

3. The optical arithmetic device includes a plurality of reflective optical modulation elements as the optical modulation elements, the monitor signal light is a signal light transmitted through the optical branching element, the calculation signal light is signal light reflected by the optical branching element, 3. The optical computing device according to claim 1, wherein the optical computing device is a first optical computing device.

4. The plurality of reflective light modulation elements are integrated together.

4. The optical computing device according to claim 3.

5. The optical arithmetic device includes a plurality of transmissive optical modulation elements as the optical modulation elements, the monitor signal light is a signal light reflected by the optical branching element, the calculation signal light is a signal light transmitted through the optical branching element, 3. The optical computing device according to claim 1, wherein the optical computing device is a first optical computing device.

6. The optical calculation device includes a plurality of stages each including a combination of the optical modulation element, the image sensor, and the optical branching element, The optical branching element of each stage from the second stage onward receives the signal light generated by the optical modulation element of the previous stage.

3. The optical computing device according to claim 1, wherein the optical computing device is a first optical computing device.

7. the control unit outputs a signal indicating that a signal light representing a correct calculation result is being output after the phase modulation amounts of all the cells of the optical modulation element have converged to a constant value.

3. The optical computing device according to claim 1, wherein the optical computing device is a first optical computing device.

8. A method for controlling an optical arithmetic device including an image sensor including at least one light receiving cell, an optical modulation element including a plurality of cells whose phase modulation amounts can be set independently of each other, and an optical branching element that branches a signal light into a monitor signal light incident on the image sensor and an arithmetic signal light incident on the optical modulation element, comprising: setting an amount of phase modulation of each cell of the light modulation element in accordance with an intensity of the monitor signal light detected by a light receiving cell of the image sensor corresponding to the cell; 23. A method for controlling an optical computing device comprising:

Images