WO2025177324A1 - Optical computation device and method for adjusting mach-zehnder interferometer in optical computation device - Google Patents

Abstract

This optical computation device is provided with: a multi-stage MZI array (1) in which Mach-Zehnder interferometers (MZI) arranged in odd-numbered columns are arranged in even-numbered rows, and Mach-Zehnder interferometers arranged in even-numbered columns are arranged in odd-numbered rows in a matrix of a plurality of rows and a plurality of columns, and the Mach-Zehnder interferometers (MZI) are arranged in a rectangular shape; and monitor photodetectors (MPD) connected to first output ports Por3 of the Mach-Zehnder interferometers (MZI) arranged in the first row.

Description

Optical computing device and method for adjusting a Mach-Zehnder interferometer in the optical computing device

This disclosure relates to an optical computing device having a multi-stage Mach-Zehnder interferometer (MZI) array in which multiple MZIs are arranged in a rectangular configuration, and a method for adjusting the MZIs in the optical computing device.

In an optical computing circuit that uses analog computing technology to perform linear calculations using the interference of light traveling through an optical waveguide, a Clements-type mesh, which is a rectangular arrangement that is compact and has low optical loss compared to a Reck-type mesh that has a triangular arrangement, is shown in Non-Patent Document 1 as a multi-port interferometer mesh.
The Clements-type mesh shown in Non-Patent Document 1 has N input and N output detectors and M rows of MZIs.
The number of MZIs arranged in odd-numbered columns is N/2, and the number of MZIs arranged in even-numbered columns is (N/2-1).

Furthermore, calibration to compensate for phase errors due to manufacturing variations for each MZI in the Clements-type mesh shown in Non-Patent Document 1 is first performed along the longest diagonal line in the mesh, starting from the bottom MZI towards the top MZI, then along the line connecting the MZI in the row above the bottom MZI, excluding the bottom MZI, towards the top MZI, and similarly, along the line connecting the MZI in the last column, the top MZI is performed, and similarly, for each odd-numbered row below.

Christopher Alexiev et al. “Calibrating rectangular interferometer meshes with external photodetectors” OSA Continuum vol.4,N0.11/15 Nov.2021.P.P.2892-2904

The Clements-type mesh shown in Non-Patent Document 1 is configured as described above, and therefore the calibration time for all MZIs becomes long.
Furthermore, the Clements-type mesh shown in Non-Patent Document 1 is constructed from the tail end of a plurality of connected MZIs, and therefore, depending on the initial conditions, the detected light intensity may become very weak, making construction difficult.
Moreover, in the Clements-type mesh shown in Non-Patent Document 1, an uncalibrated MZI is always present in the optical waveguide path from the input node where light is input to the output node where light is output, and therefore the light detection accuracy at the output node decreases.

The present disclosure aims to solve the above-mentioned problems by providing an optical computing device having a multi-stage MZI array in which multiple MZIs are arranged in a rectangular shape, in which adjustment of the MZIs in the multi-stage MZI array can be performed in a short time.

The optical computing device disclosed herein comprises a multi-row, multi-column matrix in which Mach-Zehnder interferometers arranged in odd-numbered columns are arranged in even-numbered rows, and Mach-Zehnder interferometers arranged in even-numbered columns are arranged in odd-numbered rows, with the Mach-Zehnder interferometers arranged in a rectangular configuration; and a monitor photodetector connected to the first output port of the Mach-Zehnder interferometer arranged in the first row.

According to this disclosure, adjustment of MZIs in a rectangularly arranged multi-stage MZI array in an optical computing device can be performed in a short time.

1 is a schematic diagram conceptually showing a configuration of an optical arithmetic device according to a first embodiment; FIG. 2 is a diagram showing a configuration of an MZI in the optical arithmetic device according to the first embodiment; FIG. 2 is a diagram showing a configuration of an MZI in the optical arithmetic device according to the first embodiment; FIG. 2 is a diagram showing a configuration of an MZI in the optical arithmetic device according to the first embodiment; 4 is a flowchart showing a procedure in a method for adjusting an MZI in the optical computing device according to the first embodiment. 1 is a schematic diagram showing an optical waveguide path for adjusting an MZI ₃₃ that is an adjustment target in the optical computing device according to the first embodiment. FIG. FIG. 10 is a schematic diagram conceptually showing the configuration of an optical arithmetic device according to a second embodiment. 10 is a flowchart showing a procedure in a method for adjusting an MZI in an optical computing device according to a second embodiment. FIG. 10 is a schematic diagram showing the optical waveguide paths for adjusting MZI ₉ and MZI ₂₅ , which are adjustment targets, in the optical computing device according to the second embodiment. FIG. 11 is a schematic diagram conceptually showing the configuration of an optical arithmetic device according to a third embodiment. FIG. 10 is a schematic diagram conceptually showing the configuration of an optical arithmetic device according to a fourth embodiment.

Embodiment 1.
An optical arithmetic device according to a first embodiment will be described with reference to FIGS.
The optical arithmetic device according to the first embodiment is an optical arithmetic device including an optical arithmetic circuit that uses analog computing technology to perform linear arithmetic using the interference of light traveling through an optical waveguide.
The optical arithmetic circuit is composed of a Clements-type multi-stage MZI array, which is a rectangular arrangement formed on the surface of a substrate.

The optical computing device according to the first embodiment includes a multi-stage MZI array 1 and a plurality of monitor photodetectors MPD.
The multistage MZI array 1 is an integrated optical unitary converter that unitarily converts an optical signal input to an input node and outputs the converted signal to an output node.

The multi-stage MZI array 1 is a matrix of multiple rows and multiple columns, in which Mach-Zehnder interferometers (hereinafter referred to as MZIs) arranged in odd-numbered columns are arranged in even-numbered rows, and MZIs arranged in even-numbered columns are arranged in odd-numbered rows, with the MZIs arranged in a rectangular shape.
Each MZI has a first input port Port1, a second input port Port2, a first output port Port3, and a second output port Port4.
The plurality of monitor photodetectors MPD are photodiodes connected to the first output ports of the corresponding MZIs arranged in the first row.

In the multistage MZI array 1 of the first embodiment, if the number of MZIs arranged in the first column, that is, the number of MZIs in the input stage, is N (N is a natural number of 2 or more), the number of rows is 2N+1 and the number of columns, that is, the number of stages is 2N.
The number of MZIs arranged in each row is N.
The number of stages may be less than 2N.

The indexes for the MZI are given as follows:
That is, the index of the MZI in the first row, second column is set to 0, the index of the MZI arranged in the first row is increased by 1 in the order of the columns, the MZI in the second row, first column is set after the MZI in the first row, Nth column, the index of the MZI arranged in the second row is increased by 1 in the order of the columns, and similarly, the indexes are assigned in order up to [N × (2N + 1) - 1] up to the MZI in the 2N + 1 row, Nth column.
In each row, the index of the MZI to be arranged is a smaller number when the number of columns is smaller.
In the following description, when each MZI is specifically referred to, it will be described as MZI _i , where i is an index number ranging from 0 to [N×(2N+1)−1].

In the following description, as shown in FIG. 1, a 4×4 Clements-type multistage MZI array will be taken as an example, in which four MZIs are arranged in the input stage, the number of input nodes is eight, and the number of output nodes is eight.
In FIG. 1, the numbers 0 to 35 enclosed in square frames indicate MZIs and also indicate the index numbers of the MZIs.
In the multistage MZI array 1 shown in FIG. 1, N is 4, the number of rows is 9 (2N+1), the number of columns is 8 (2N), the input stage is the first column, and the output stage is the eighth column, which is the final column.

MZIs with indexes 0 to 4 are placed on the first row, MZIs with indexes 5 to 7 on the second row, MZIs with indexes 8 to 11 on the third row, MZIs with indexes 12 to 15 on the fourth row, MZIs with indexes 16 to 19 on the fifth row, MZIs with indexes 20 to 23 on the sixth row, MZIs with indexes 24 to 27 on the seventh row, MZIs with indexes 28 to 31 on the eighth row, and MZIs with indexes 32 to 35 on the ninth row.

The number of MZIs arranged in each row is 4(N).
Therefore, the optical signal propagates through the same number of MZIs on all paths from the input node to the output node.
As a result, in a multi-stage MZI array, the waveguide loss of each path is constant, resulting in good path symmetry.

The first input port Port1 and the second input port Port2 of the MZI arranged in the input stage are connected to input nodes x0 _{to x7} in order from MZI ₄ arranged in the second row to MZI ₁₂ arranged in the fourth row, MZI ₂₀ arranged in the sixth row, and MZI ₂₆ arranged in the _eighth row.
Light-emitting elements (not shown), which are light-emitting diodes that output optical signals, are connected to the input nodes x ₀ to x ₇ .

Of the MZIs arranged in the eighth column, which is the final stage, the second output port Port4 of MZI ₃ arranged in the first row is connected to output node _y0 .
Of the MZIs arranged in the final stage, the first output port Port3 of the MZI ₃₅ arranged in the ninth row, which is the bottom row, is connected to the output node _y7 .
Of the MZIs arranged in the final stage, the first output port Port3 and the second output port Port4 of MZI ₁₁ arranged in the third row between the first and ninth rows, MZI ₁₉ arranged in the fifth row, and MZI ₂₇ arranged in the seventh row are connected to output nodes _y1 to _y6 in that order.
Optical detectors (not shown) that are photodetectors to which optical signals are input are connected to the output nodes y ₀ to y ₇ .

In each of MZI ₀ , MZI ₁ , MZI ₂ , and MZI ₃ arranged in the first row, the first input port Port1 is connected to the corresponding monitor input node p ₀ to p ₃ , the second input port Port2 is connected by an optical waveguide to the first output port Port3 of MZI ₄ to MZI ₇ arranged in the adjacent front column in the second row, and the first output port Port3 is connected to the corresponding monitor photodetector MPD ₀ to MPD ₃ .

In each of the MZIs _i except for the input stage and output stage and the MZIs in the first and bottom rows, the first input port Port1 is connected by an optical waveguide to the second output port Port4 of the MZI arranged in the adjacent previous column in the adjacent previous row, and the second input port Port2 is connected by an optical waveguide to the first output port Port3 of the MZI arranged in the adjacent previous column in the adjacent subsequent row.

In each of MZI ₃₂ , MZI ₃₃ , MZI ₃₄ , and MZI ₃₅ arranged in the bottom ninth row, the first input port Port1 is connected by an optical waveguide to the second output port Port4 of MZI ₂₈ to MZI ₃₁ arranged in the adjacent previous column in the eighth row, and the second input port Port2 is connected to the corresponding monitor input nodes _p4 to _p7 .

Each of MZI ₀ to MZI ₃₅ is one of the MZIs shown in FIGS. 2 to 4, which has a configuration that allows compensation for manufacturing variations in branching ratio (branching ratio variations) and the like through adjustment.
The MZI shown in FIG. 2 includes a first optical waveguide connecting a first input port Port1 and a first output port Port3, a second optical waveguide connecting a second input port Port2 and a second output port Port4, and has a θ phase shifter.

The MZI shown in FIG. 3 includes a first optical waveguide connecting a first input port Port1 and a first output port Port3, a second optical waveguide connecting a second input port Port2 and a second output port Port4, and has a φ phase shifter and a θ phase shifter.
The MZI shown in FIG. 4 includes a first optical waveguide connecting a first input port Port1 and a first output port Port3, and a second optical waveguide connecting a second input port Port2 and a second output port Port4, and has a φ phase shifter, a θ phase shifter, and a ψ phase shifter.

The MZI shown in Figures 2 to 4 can be in two states: a through state in which input light to the first input port Port1 is output to the first output port Port3 and input light to the second input port Port2 is output to the second output port Port4, and a cross state in which input light to the first input port Port1 is output to the second output port Port4 and input light to the second input port Port2 is output to the first output port Port3.

The MZIs shown in FIGS. 2 to 4 are fabricated using a semiconductor process, and the first and second optical waveguides are formed by optical waveguides called silicon photonics, which use single crystal silicon as the core material.
Instead of single crystal silicon as the core material for forming the optical waveguide, silicon nitride (SiN), alumina, quartz, or the like may be used.

Each of the monitor photodetectors MPD ₀ to MPD ₃ is a photodiode having a pin structure, and is made of germanium, which is compatible with the manufacturing process of silicon photonics.
Instead of a photodiode using germanium, a photodiode formed by integrating compound semiconductors such as InGaAs by heterogeneous material junction may be used.

Next, a method for adjusting the MZI in the optical computing device according to the first embodiment will be described.
That is, the procedure of the MZI adjustment method for optimizing the phase adjustment region for each MZI _i in order to compensate for the branching ratio variation of the MZI _i will be described.

In the method for adjusting the MZI in the optical arithmetic device according to the first embodiment, for the MZI _i to be adjusted to compensate for the phase error, the MZI arranged in the front row of the adjustment target MZI _i in the optical waveguide path passing through the adjustment target MZI _i is set to a through state, and the MZI arranged in the rear row in the optical waveguide path is set to a cross state, and adjustment is performed to compensate for the branching ratio variation in the adjustment target MZI _i.

In the method for adjusting the MZI in the optical computing device according to the first embodiment, an optical signal is input to the first input port Port1 of the MZI _i arranged in the first row to be adjusted, and an optical signal is output to the monitor photodetector MPD _k connected to the first output port Port3 of the MZI i, and adjustment is performed, and the MZI _i arranged in the second row or later to be adjusted is performed by using only the investigated MZI _i as the optical waveguide path.
The MZI adjustment method in the optical computing device according to the first embodiment is performed in the order of indexes from the MZI with index 0 to the MZI with index [N×(2N+1)−1].

Hereinafter, a specific procedure for adjusting the MZI in the optical computing device according to the first embodiment will be described with reference to FIGS. 5 and 6 for an example in which N is set to 4. FIG.
In step ST1, MZI ₀ to MZI ₃ arranged in the first row are adjusted as adjustment targets in the order of their indexes.
That is, for each of MZI ₀ to MZI ₃ , light is input to the corresponding monitor input nodes p ₀ to p ₃ in the order of the indexes, and the light is monitored by the corresponding monitor photodetectors MPD ₀ to MPD ₃ , and adjustments are made to compensate for the branching ratio variations of each of MZI ₀ to MZI _3. After the adjustments, each of MZI ₀ to MZI ₃ is set to a through state.

The optical waveguide path for adjustment is shown below.
The adjustment is performed by using the monitor input node _p0 for the adjustment target _MZI0 and the optical detector _MPD0 along the optical waveguide path _p0 → _MZI0 → _MPD0 .
In the following description, the optical waveguide paths will be indicated by symbols to avoid the description becoming complicated.
For the MZI ₁ to be adjusted, p ₁ →MZI ₁ →MPD ₁ .
For the MZI ₂ to be adjusted, p ₂ →MZI ₂ →MPD ₂ .
For the MZI ₃ to be adjusted, p ₃ →MZI ₃ →MPD ₃ .
In step ST1, adjustment of MZI ₀ to MZI ₃ arranged in the first row is completed and the characteristics become known.

In step ST2, MZI ₄ to MZI ₇ arranged in the second row are adjusted as adjustment targets in the order of their indexes.
At this time, light is input from the input node _x0 to the first input port Port1 of the MZI ₄ in the input stage.
Moreover, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

In the optical waveguide path shown below, the state of MZI _i is represented as a through state [T] and a cross state [C] after MZI _i .
For the MZI ₄ to be adjusted, x ₀ →MZI ₄ →MZI ₀ [C] →MPD ₀ . After adjustment, MZI ₄ is set to the through state, and MZI ₀ is also set to the through state.
For the MZI ₅ to be adjusted, x ₀ →MZI ₄ [T] →MZI ₀ [T] →MZI ₅ →MZI ₁ [C] →MPD ₁ . After adjustment, MZI ₅ is set to the through state, and MZI ₁ is also set to the through state.

For the MZI ₆ to be adjusted, x ₀ →MZI ₄ [T] →MZI ₀ [T] →MZI ₅ [T] →MZI ₁ [T] →MZI ₆ →MZI ₂ [C] →MPD ₂ . MZI ₆ is set to the through state after adjustment, and MZI ₂ is also set to the through state after adjustment.
For the MZI ₇ to be adjusted, x ₀ → MZI ₄ [T] → MZI ₀ [T] → MZI ₅ [T] → MZI ₁ [T] → MZI ₆ [T] → MZI ₂ [T] → MZI ₇ → MZI ₃ [C] → MPD _3. After adjustment, MZI ₇ is set to the through state, and MZI ₃ is also set to the through state.

In short, light is input from the input node _x0 to each of MZI ₄ to MZI ₇ arranged in the second row, and is monitored by the monitor photodetectors MPD ₀ to MPD ₃ , respectively.
In adjusting MZI ₄ to MZI ₇ arranged in the second row, the input light passes through MZIs whose characteristics are known and have been adjusted until it is input from input node x ₀ , and the light output from the object to be adjusted passes through MZIs whose characteristics are known and have been adjusted until it reaches each of monitor photodetectors MPD ₀ to MPD ₃ , so that only the characteristics of the MZI to be adjusted can be accurately extracted by each of monitor photodetectors MPD ₀ to MPD ₃ .

Furthermore, since the light does not pass through an MZI whose state, whether through or cross, is unknown, all of the light guided through the adjustment waveguide path (excluding power lost due to waveguide loss) can be detected by each of the monitor photodetectors MPD ₀ to MPD ₃ , maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.

In step ST3, MZI ₈ to MZI ₁₁ arranged in the third row are adjusted in the order of their indexes.
At this time, light is input from the input node _x1 to the second input port Port2 of the MZI ₄ in the input stage.
Furthermore, every time an adjustment target is adjusted, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

For the MZI ₈ to be adjusted, x ₁ →MZI ₄ [T] →MZI ₈ →MZI ₅ [C] →MZI ₁ [C] →MPD ₁ .
For the MZI ₉ to be adjusted, x ₁ →MZI ₄ [T] →MZI ₈ [T] →MZI ₅ [T] →MZI ₉ →MZI ₆ [C] →MZI ₂ [C] →MPD ₂ .

For the MZI ₁₀ to be adjusted, x ₁ → MZI ₄ [T] → MZI ₈ [T] → MZI ₅ [T] → MZI ₉ [T] → MZI ₆ [T] → MZI ₁₀ → MZI ₇ [C] → MZI ₃ [C] → MPD ₃ .
For the MZI ₁₁ to be adjusted, x ₁ → MZI ₄ [T] → MZI ₈ [T] → MZI ₅ [T] → MZI ₉ [T] → MZI ₆ [T] → MZI ₁₀ [T] → MZI ₇ [T] → MZI ₁₁ → y ₁ .
The MZI ₁₁ to be adjusted is monitored by a photodetector connected to the output node _y1 .
In the following description, the output node shown in the optical waveguide path for adjustment includes the photodetector connected to the output node.

In short, light is input from input node _x1 to each of MZI ₈ to MZI ₁₁ arranged in the third row, and is monitored by monitor photodetectors MPD ₁ to MPD ₃ and a photodetector connected to output node _y1 .
In adjusting MZI ₈ to MZI ₁₁ arranged in the third row, the signal passes through MZIs whose characteristics are known and have already been adjusted, so that only the characteristics of the MZI to be adjusted can be extracted with high precision.
In addition, since the light does not pass through the MZI whose state (through or cross) is unknown, all the light guided through the adjustment waveguide path can be detected, maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.

In step ST4, MZI ₁₂ to MZI ₁₅ arranged in the fourth row are adjusted in the order of their indexes.
At this time, light is input from the input node _x2 to the first input port Port1 of the MZI ₁₂ in the input stage.
Furthermore, every time an adjustment target is adjusted, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

For the MZI ₁₂ to be adjusted, x ₂ →MZI ₁₂ →MZI ₈ [C] →MZI ₅ [C] →MZI ₁ [C] →MPD ₁ .
For the MZI ₁₃ to be adjusted, x ₂ →MZI ₁₂ [T] →MZI ₈ [T] →MZI ₁₃ →MZI ₉ [C] →MZI ₆ [C] →MZI ₂ [C] →MPD ₂ .

For the MZI ₁₄ to be adjusted, x ₂ →MZI ₁₂ [T] →MZI ₈ [T] →MZI ₁₃ [T] →MZI ₉ [T] →MZI ₁₄ →MZI ₁₀ [C] →MZI ₇ [C] →MZI ₃ [C] →MPD ₃ .
For the MZI ₁₅ to be adjusted, x ₂ →MZI ₁₂ [T] →MZI ₈ [T] →MZI ₁₃ [T] →MZI ₉ [T] →MZI ₁₄ [T] →MZI ₁₀ [T] →MZI ₁₅ →MZI ₁₁ [C] →y ₁ .

In short, light is input from input node _x2 to each of MZI ₁₂ to MZI ₁₅ arranged in the fourth row, and is monitored by monitor photodetectors MPD ₁ to MPD ₃ and a photodetector connected to output node _y1 .
In adjusting MZI ₁₂ to MZI ₁₅ arranged in the fourth row, the signal passes through MZIs whose characteristics are known and have already been adjusted, so that only the characteristics of the MZI to be adjusted can be extracted with high precision.
In addition, since the light does not pass through the MZI whose state (through or cross) is unknown, all the light guided through the adjustment waveguide path can be detected, maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.

In step ST5, MZI ₁₆ to MZI ₁₉ arranged in the fifth row are adjusted in the order of their indexes.
At this time, light is input from the input node _x3 to the second input port Port2 of the MZI ₁₂ in the input stage.
Furthermore, every time an adjustment target is adjusted, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

For the MZI ₁₆ to be adjusted, x ₃ → MZI ₁₂ [T] → MZI ₁₆ → MZI ₁₃ [C] → MZI ₉ [C] → MZI ₆ [C] → MZI ₂ [C] → MPD ₂
For the MZI ₁₇ to be adjusted, x ₃ → MZI ₁₂ [T] → MZI ₁₆ [T] → MZI ₁₃ [T] → MZI ₁₇ → MZI ₁₄ [C] → MZI ₁₀ [C] → MZI ₇ [C] → MZI ₃ [C] → MPD ₃ .

For the MZI ₁₈ to be adjusted, x ₃ →MZI ₁₂ [T] →MZI ₁₆ [T] →MZI ₁₃ [T] →MZI ₁₇ [T] →MZI ₁₄ [T] →MZI ₁₈ →MZI ₁₅ [C] →MZI ₁₁ [C] →y ₁ .
For the MZI ₁₉ to be adjusted, x ₃ →MZI ₁₂ [T] →MZI ₁₆ [T] →MZI ₁₃ [T] →MZI ₁₇ [T] →MZI ₁₄ [T] →MZI ₁₈ [T] →MZI ₁₅ [T] →MZI ₁₉ →y ₃ .

In short, light is input from input node _x3 to each of MZI ₁₆ to MZI ₁₉ arranged in the fifth row, and is monitored by monitor photodetectors MPD ₂ and MPD ₃ and photodetectors connected to output nodes y ₁ and y _3, respectively.
In adjusting MZI ₁₆ to MZI ₁₉ arranged in the fifth row, the characteristics of the MZIs that are already known and adjusted are passed through, so that only the characteristics of the MZIs to be adjusted can be extracted with high precision.
In addition, since the light does not pass through the MZI whose state (through or cross) is unknown, all the light guided through the adjustment waveguide path can be detected, maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.

In step ST6, MZI ₂₀ to MZI ₂₃ arranged in the sixth row are adjusted in the order of their indexes.
At this time, light is input from the input node _x4 to the first input port Port1 of the MZI ₂₀ in the input stage.
Furthermore, every time an adjustment target is adjusted, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

For the MZI ₂₀ to be adjusted, x ₄ → MZI ₂₀ → MZI ₁₆ [C] → MZI ₁₃ [C] → MZI ₉ [C] → MZI ₆ [C] → MZI ₂ [C] → MPD ₂ .
For the MZI ₂₁ to be adjusted, x ₄ → MZI ₂₀ [T] → MZI ₁₆ [T] → MZI ₂₁ → MZI ₁₇ [C] → MZI ₁₄ [C] → MZI ₁₀ [C] → MZI ₇ [C] → MZI ₃ [C] → MPD ₃ .

For the MZI ₂₂ to be adjusted, x ₄ → MZI ₂₀ [T] → MZI ₁₆ [T] → MZI ₂₁ [T] → MZI ₁₇ [T] → MZI ₂₂ → MZI ₁₈ [C] → MZI ₁₅ [C] → MZI ₁₁ [C] → y ₁ .
For the MZI ₂₃ to be adjusted, x ₄ → MZI ₂₀ [T] → MZI ₁₆ [T] → MZI ₂₁ [T] → MZI ₁₇ [T] → MZI ₂₂ [T] → MZI ₁₈ [T] → MZI ₂₃ → MZI ₁₉ → y ₃ .

In short, light is input from input node _x4 to each of MZI ₂₀ to MZI ₂₃ arranged in the sixth row, and is monitored by monitor photodetectors MPD ₂ and MPD ₃ and photodetectors connected to output nodes y ₁ and y _3, respectively.
In adjusting MZI ₂₀ to MZI ₂₃ arranged in the sixth row, the characteristics of the MZIs that are already known and adjusted are passed through, so that only the characteristics of the MZIs to be adjusted can be extracted with high precision.
In addition, since the light does not pass through the MZI whose state (through or cross) is unknown, all the light guided through the adjustment waveguide path can be detected, maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.

In step ST7, MZI ₂₄ to MZI ₂₇ arranged in the seventh row are adjusted in the order of their indexes.
At this time, light is input from the input node _x5 to the second input port Port2 of the MZI ₂₀ in the input stage.
Furthermore, every time an adjustment target is adjusted, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

For the MZI ₂₄ to be adjusted, x ₅ → MZI ₂₀ [T] → MZI ₂₄ → MZI ₂₁ [C] → MZI ₁₇ [C] → MZI ₁₄ [C] → MZI ₁₀ [C] → MZI ₇ [C] → MZI ₃ [C] → MPD ₃ .
For the MZI ₂₅ to be adjusted, x ₅ → MZI ₂₀ [T] → MZI ₂₄ [T] → MZI ₂₁ [T] → MZI ₂₅ → MZI ₂₂ [C] → MZI ₁₈ [C] → MZI ₁₅ [C] → MZI ₁₁ [C] → y ₁ .

For the MZI ₂₆ to be adjusted, x ₅ → MZI ₂₀ [T] → MZI ₂₄ [T] → MZI ₂₁ [T] → MZI ₂₅ [T] → MZI ₂₂ [T] → MZI ₂₆ → MZI ₂₃ [C] → MZI ₁₉ [C] → y ₃ .
For the MZI ₂₇ to be adjusted, x ₅ → MZI ₂₀ [T] → MZI ₂₄ [T] → MZI ₂₁ [T] → MZI ₂₅ [T] → MZI ₂₂ [T] → MZI ₂₆ [T] → MZI ₂₃ [T] → MZI ₂₇ → y ₅ .

In short, light is input from input node _x5 to each of MZI ₂₄ to MZI ₂₇ arranged in the seventh row, and is monitored by monitor photodetector MPD ₃ and photodetectors connected to output nodes _y1 , _y3 , and y5 _, respectively.
In adjusting MZI ₂₄ to MZI ₂₇ arranged in the seventh row, the signal passes through MZIs whose characteristics are known and have already been adjusted, so that only the characteristics of the MZI to be adjusted can be extracted with high precision.
In addition, since the light does not pass through the MZI whose state (through or cross) is unknown, all the light guided through the adjustment waveguide path can be detected, maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.

In step ST8, MZI ₂₈ to MZI ₃₁ arranged in the eighth row are adjusted in the order of their indexes.
At this time, light is input from the input node _x6 to the first input port Port1 of the MZI ₂₈ in the input stage.
Furthermore, every time an adjustment target is adjusted, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

For the MZI ₂₈ to be adjusted, x ₆ → MZI ₂₈ → MZI ₂₄ [C] → MZI ₂₁ [C] → MZI ₁₇ [C] → MZI ₁₄ [C] → MZI ₁₀ [C] → MZI ₇ [C] → MZI ₃ [C] → MPD ₃ .
For the MZI ₂₉ to be adjusted, x ₆ → MZI ₂₈ [T] → MZI ₂₄ [T] → MZI ₂₉ → MZI ₂₅ [C] → MZI ₂₂ [C] → MZI ₁₈ [C] → MZI ₁₅ [C] → MZI ₁₁ [C] → y ₁ .

For the MZI ₃₀ to be adjusted, x ₆ → MZI ₂₈ [T] → MZI ₂₄ [T] → MZI ₂₉ [T] → MZI ₂₅ [T] → MZI ₃₀ → MZI ₂₆ [C] → MZI ₂₃ [C] → MZI ₁₉ [C] → y ₃ .
For the MZI ₃₁ to be adjusted, x ₆ →MZI ₂₈ [T] →MZI ₂₄ [T] →MZI ₂₉ [T] →MZI ₂₅ [T] →MZI ₃₀ [T] →MZI ₂₆ [T] →MZI ₃₁ →MZI ₂₇ [C] →y ₅ .

In short, light is input to each of MZI ₂₈ to MZI ₃₁ arranged in the eighth row from input node _x6 , and is monitored by monitor photodetector MPD ₃ and photodetectors connected to output nodes y ₁ , y ₃ , and y ₅ , respectively.
In adjusting MZI ₂₈ to MZI ₃₁ arranged in the eighth row, the signal passes through MZIs whose characteristics are known and have already been adjusted, so that only the characteristics of the MZI to be adjusted can be extracted with high precision.
In addition, since the light does not pass through the MZI whose state (through or cross) is unknown, all the light guided through the adjustment waveguide path can be detected, maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.

In step ST9, MZI ₃₂ to MZI ₃₅ arranged in the ninth row are adjusted in the order of their indexes.
At this time, light is input from the input node _x7 to the second input port Port2 of the MZI ₂₈ in the input stage.
Furthermore, every time an adjustment target is adjusted, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

For the MZI ₃₂ to be adjusted, x ₇ → MZI ₂₈ [T] → MZI ₃₂ → MZI ₂₉ [C] → MZI ₂₅ [C] → MZI ₂₂ [C] → MZI ₁₈ [C] → MZI ₁₅ [C] → MZI ₁₁ [C] → y ₁ .
For the adjustment target MZI ₃₃ , x ₇ → MZI ₂₈ [T] → MZI ₃₂ [T] → MZI ₂₉ [T] → MZI ₃₃ → MZI ₃₀ [C] → MZI ₂₆ [C] → MZI ₂₃ [C] → MZI ₁₉ [C] → y _3. In Figure 6, the optical waveguide path for adjustment for the adjustment target MZI ₃₃ is shown by a thick line.

For the MZI ₃₄ to be adjusted, x ₇ → MZI ₂₈ [T] → MZI ₃₂ [T] → MZI ₂₉ [T] → MZI ₃₃ [T] → MZI ₃₀ [T] → MZI ₃₄ → MZI ₃₁ [C] → MZI ₂₇ [C] → y ₅ .
For the MZI ₃₅ to be adjusted, x ₇ → MZI ₂₈ [T] → MZI ₃₂ [T] → MZI ₂₉ [T] → MZI ₃₃ [T] → MZI ₃₀ [T] → MZI ₃₄ [T] → MZI ₃₁ [T] → MZI ₃₅ → y ₇ .

In short, light is input from input node _x5 to each of MZI ₃₂ to MZI ₃₅ arranged in the ninth row, and is monitored by photodetectors connected to output nodes _y1 , _y3 , _y5 , and y7 _, respectively.
In adjusting MZI ₃₂ to MZI ₃₅ arranged in the ninth row, the signal passes through MZIs whose characteristics are known and have already been adjusted, so that only the characteristics of the MZI to be adjusted can be extracted with high precision.
In addition, since the light does not pass through the MZI whose state (through or cross) is unknown, all the light guided through the adjustment waveguide path can be detected, maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.

Although an example of an optical waveguide path to be adjusted for the MZI _i to be adjusted has been shown, in short, the MZI _i arranged in the first row is directly monitored and adjusted by the monitor photodetector MPD _k , and when the MZI _i arranged in the second row or later is to be adjusted, the MZI _i to be adjusted can be adjusted by placing only the adjusted MZI whose state, whether it is a through state or a cross state, in the optical waveguide path through which the light for adjustment passes.
Furthermore, when adjusting MZI _i arranged in the second row or later, adjustment is performed in the order of the index of MZI _i for each row starting from the second row, and the MZI _i to be adjusted can be adjusted to only MZIs that have already been adjusted on the optical waveguide path through which the light for adjustment passes.

The optical computing device according to the first embodiment is provided with a multi-stage MZI array 1 in which MZIs arranged in odd-numbered columns are arranged in even-numbered rows, and MZIs arranged in even-numbered columns are arranged in odd-numbered rows in a matrix of multiple rows and columns, and the MZIs are arranged in a rectangular shape, and a monitor photodetector MPD _k connected to the first output port of the MZI arranged in the first row, so that all of the MZIs constituting MPD _k can be adjusted in a short time.

Furthermore, for the MZI _i to be adjusted that is monitored by the monitor photodetector MPD _k , the number of adjusted MZIs that exist in the optical waveguide path through which light passes to the MZI _i to be adjusted can be reduced, thereby reducing loss in the optical waveguide path.
As a result, the monitored optical intensity can be increased, for example, maximized, the signal-to-noise ratio can be maximized, and the MZI can be adjusted with high precision.

Moreover, in the optical arithmetic device according to the first embodiment, the MZI _i arranged in the first row can be directly monitored and adjusted by the monitor photodetector MPD _k , and for adjustment of the MZI _i arranged in the second row and thereafter, there is an _MZI whose characteristics are known and which has already been adjusted in the optical waveguide path through which the adjustment light passes, and adjustment can be performed without there being any unadjusted MZI, so it is possible to precisely extract only the characteristics of the MZI i to be adjusted, and highly accurate adjustment of the MZI can be performed.

Furthermore, the optical arithmetic device according to the first embodiment can adjust the MZI _i to be adjusted without the presence of an MZI whose state, whether it is a through state or a cross state, being unknown in the optical waveguide path through which the adjustment light passes, and therefore the light propagating through the optical waveguide path through which the adjustment light passes through the MZI _i to be adjusted can be efficiently detected by the monitor photodetector MPD _k or a photodetector connected to the output node, so the intensity of the light to be monitored can be increased, for example maximized, the signal-to-noise ratio can be maximized, and the MZI can be adjusted with high precision.

Embodiment 2.
An optical arithmetic device according to a second embodiment will be described with reference to FIGS.
The optical arithmetic device according to the second embodiment is different from the optical arithmetic device according to the first embodiment in that it further includes a second monitor photodetector MPD connected to the second output port Port4 of the MZI _i arranged in the bottom row, but is otherwise the same.
Therefore, the following description will focus on the differences.
In the following description, the monitor photodetector MPD connected to the first output port of the MZI arranged in the first row will be referred to as the first monitor photodetector.
7 to 9, the same reference numerals as those in FIGS. 1 to 6 designate the same or corresponding parts.

In each of MZI ₃₂ , MZI ₃₃ , MZI ₃₄ , and MZI ₃₅ arranged in the bottom 9th row, the first input port Port1 is connected by an optical waveguide to the second output port Port4 of MZI ₂₈ to MZI ₃₁ arranged in the adjacent previous column in the 8th row, the second input port Port2 is connected to the corresponding monitor input nodes _p4 to _p7 , and the second output port Port4 is connected to the corresponding second monitor photodetectors _MPD4 to _MPD7 .

The second monitor photodetectors MPD ₄ to MPD ₇ each employ a photodiode having a pin structure, and are photodiodes made of germanium, which is compatible with the manufacturing process of silicon photonics.
Instead of a photodiode using germanium, a photodiode formed by integrating compound semiconductors such as InGaAs by heterogeneous material junction may be used.

Next, a method for adjusting the MZI in the optical computing device according to the second embodiment will be described.
The multiple rows (2N+1) in the multistage MZI array 1 are divided into two, a first half (N+1) and a second half (N), and adjustments are made to simultaneously compensate for branching ratio variations in the MZIs arranged in the first half rows and the MZIs arranged in the second half rows, from the first row to the bottom row (N+1) of the first half, and from the bottom row (2N+1) to the top row (N+2) of the second half, and adjustments are made to compensate for branching ratio variations in the MZIs arranged in the first half rows and the MZIs arranged in the second half rows, and adjustments are made with only the MZIs that have been investigated as optical waveguide paths for the MZIs arranged in each row.

The MZIs in the first half of the multi-stage MZI array 1 are adjusted in index order from the MZI with index 0 to the MZI with index [N×(N+1)-1], and the MZIs in the second half are adjusted in index order from the MZI with index [N×(2N+1)-1] to the MZI with index [N×(N+1]].

Hereinafter, a specific procedure for adjusting the MZI in the optical computing device according to the second embodiment will be described with reference to FIGS. 8 and 9 for an example in which N is set to 4. FIG.
The adjustment of MZI ₀ to MZI ₃ arranged in the first row in the first half of the multistage MZI array 1 and the adjustment of MZI ₃₂ to MZI ₃₅ arranged in the ninth row in the second half are performed simultaneously in parallel.
The adjustment of MZI ₄ to MZI ₇ arranged in the second row in the first half and the adjustment of MZI ₂₈ to MZI ₃₁ arranged in the eighth row in the second half are performed simultaneously in parallel.

The adjustment of MZI ₈ to MZI ₁₁ arranged in the third row in the first half and the adjustment of MZI ₂₄ to MZI ₂₇ arranged in the seventh row in the second half are performed simultaneously in parallel.
The adjustment of MZI ₁₂ to MZI ₁₅ arranged in the fourth row in the first half and the adjustment of MZI ₂₀ to MZI ₂₃ arranged in the sixth row in the second half are performed simultaneously in parallel.

MZI ₁₆ to MZI ₁₉ arranged in the fifth row are adjusted.
The adjustment of the first half of MZI ₀ to MZI ₁₉ in the multistage MZI array 1 is performed in steps ST1 to ST5, similar to the MZI ₀ to MZI ₁₉ shown in the MZI adjustment method in the optical arithmetic device according to the first embodiment, so the explanation will be omitted.

In step ST1, MZI ₃₂ to MZI ₃₅ arranged in the ninth row are adjusted in the order of indexes as adjustment targets, corresponding to MZI ₀ to MZI ₃ arranged in the first row.
For the MZI ₃₂ to be adjusted, p ₄ →MZI ₃₂ →MPD ₄ .
For the MZI ₃₃ to be adjusted, p ₅ →MZI ₃₃ →MPD ₅ .
For the MZI ₃₄ to be adjusted, p ₆ →MZI ₃₄ →MPD ₆ .
For the MZI ₃₅ to be adjusted, p ₇ →MZI ₃₅ →MPD ₇ .
In step ST1, adjustment of MZI ₃₂ to MZI ₃₅ arranged in the ninth row is completed, and the characteristics become known.

In step ST2, MZI ₂₈ to MZI ₃₁ arranged in the eighth row are adjusted in the order of their indexes as adjustment targets, simultaneously with MZI ₄ to MZI ₇ arranged in the second row.
At this time, light is input from the input node _x7 to the second input port Port2 of the MZI ₂₈ in the input stage.
Furthermore, every time an adjustment target is adjusted, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

For the MZI ₂₈ to be adjusted, x ₇ →MZI ₂₈ →MZI ₃₂ [C] →MPD ₄ . After adjustment, the MZI ₂₈ is set to the through state, and the MZI ₃₂ is also set to the through state.
For the MZI ₂₉ to be adjusted, x ₇ →MZI ₂₈ [T] →MZI ₂₉ →MZI ₃₂ [T] →MZI ₃₃ [C] →MPD ₅ . After adjustment, MZI ₂₉ is set to the through state, and MZI ₃₃ is also set to the through state.

For the MZI ₃₀ to be adjusted, x ₇ →MZI ₂₈ [T] →MZI ₃₂ [T] →MZI ₂₉ [T] →MZI ₃₃ [T] →MZI ₃₀ →MZI ₃₄ [C] →MPD ₆ . MZI ₃₀ is set to the through state after adjustment, and MZI ₃₄ is also set to the through state after adjustment.
For the MZI ₃₁ to be adjusted, x ₇ → MZI ₂₈ [T] → MZI ₃₂ [T] → MZI ₂₉ [T] → MZI ₃₃ [T] → MZI ₃₀ [T] → MZI ₃₄ [T] → MZI ₃₁ → MZI ₃₅ [C] → MPD ₇ .

In short, light is input from input node _x7 to each of MZI ₂₈ to MZI ₃₁ arranged in the eighth row, and is monitored by the photodetectors connected to second monitor photodetectors MPD ₄ to _MPD7 , respectively.
In adjusting MZI ₂₈ to MZI ₃₁ arranged in the eighth row, the signal passes through MZIs whose characteristics are known and have already been adjusted, so that only the characteristics of the MZI to be adjusted can be extracted with high precision.
In addition, since the light does not pass through the MZI whose state (through or cross) is unknown, all the light guided through the adjustment waveguide path can be detected, maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.

In step ST3, MZI ₂₄ to MZI ₂₇ arranged in the seventh row are adjusted in the order of indexes as adjustment targets, simultaneously with MZI ₈ to MZI ₁₁ arranged in the third row.
At this time, light is input from the input node _x6 to the first input port Port1 of the MZI ₂₈ in the input stage.
Furthermore, every time an adjustment target is adjusted, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

For the MZI ₂₄ to be adjusted, x ₆ →MZI ₂₈ [T] →MZI ₂₄ →MZI ₂₉ [C] →MZI ₃₃ [C] →MPD ₅ .
For the adjustment target MZI ₂₅ , x ₆ → MZI ₂₈ [T] → MZI ₂₄ [T] → MZI ₂₉ [T] → MZI ₂₅ → MZI ₃₀ [C] → MZI ₃₄ [C] → MPD _6. In Figure 9, the optical waveguide paths for adjustment for the adjustment target MZI ₉ and adjustment target MZI ₃₃ in step ST3 are shown by bold lines.

For the MZI ₂₆ to be adjusted, x ₆ → MZI ₂₈ [T] → MZI ₂₄ [T] → MZI ₂₉ [T] → MZI ₂₅ [T] → MZI ₃₀ [T] → MZI ₂₆ → MZI ₃₁ [C] → MZI ₃₅ [C] → MPD ₇ .
For the MZI ₂₇ to be adjusted, x ₆ → MZI ₂₈ [T] → MZI ₂₄ [T] → MZI ₂₉ [T] → MZI ₂₅ [T] → MZI ₃₀ [T] → MZI ₂₆ [T] → MZI ₃₁ [T] → MZI ₂₇ → y ₆ .

In short, MZI ₂₄ to MZI ₂₇ arranged in the seventh row each receive light input from input node _x6 , and the light is monitored by second monitor photodetectors MPD ₅ to MPD ₇ and the photodetector connected to output node _y6 .
In adjusting MZI ₂₄ to MZI ₂₇ arranged in the seventh row, the signal passes through MZIs whose characteristics are known and have already been adjusted, so that only the characteristics of the MZI to be adjusted can be extracted with high precision.
In addition, since the light does not pass through the MZI whose state (through or cross) is unknown, all the light guided through the adjustment waveguide path can be detected, maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.

In step ST4, MZI ₂₀ to MZI ₂₃ arranged in the sixth row are adjusted in the order of indexes as adjustment targets, simultaneously with MZI ₁₂ to MZI ₁₅ arranged in the fourth row.
At this time, light is input from the input node _x5 to the second input port Port2 of the MZI ₂₀ in the input stage.
Furthermore, every time an adjustment target is adjusted, the MZI arranged in the front row in the optical waveguide path passing through the adjustment target is set to the through state, and the MZI arranged in the rear row is set to the cross state.

For the MZI ₂₀ to be adjusted, x ₅ →MZI ₂₀ →MZI ₂₄ [C] →MZI ₂₉ [C] →MZI ₃₃ [C] →MPD ₅ .
For the MZI ₂₁ to be adjusted, x ₅ → MZI ₂₀ [T] → MZI ₂₄ [T] → MZI ₂₁ → MZI ₂₅ [C] → MZI ₃₀ [C] → MZI ₃₄ [C] → MPD ₆ .

For the MZI ₂₂ to be adjusted, x ₅ → MZI ₂₀ [T] → MZI ₂₄ [T] → MZI ₂₁ [T] → MZI ₂₅ [T] → MZI ₂₂ → MZI ₂₆ [C] → MZI ₃₁ [C] → MZI ₃₅ [C] → MPD ₇ .
For the MZI ₂₃ to be adjusted, x ₅ → MZI ₂₀ [T] → MZI ₂₄ [T] → MZI ₂₁ [T] → MZI ₂₅ [T] → MZI ₂₂ [T] → MZI ₂₆ [T] → MZI ₂₃ → MZI ₂₇ [C] → y ₆ .

In short, light is input from input node _x4 to each of MZI ₂₀ to MZI ₂₃ arranged in the sixth row, and is monitored by monitor photodetectors MPD ₅ to MPD ₇ and a photodetector connected to output node _y6 .
In adjusting MZI ₂₀ to MZI ₂₃ arranged in the sixth row, the characteristics of the MZIs that are already known and adjusted are passed through, so that only the characteristics of the MZIs to be adjusted can be extracted with high precision.

In addition, since the light does not pass through the MZI whose state (through or cross) is unknown, all the light guided through the adjustment waveguide path can be detected, maximizing the light intensity during adjustment. This maximizes the signal-to-noise ratio and enables highly accurate adjustment.
For convenience, MZI ₁₆ to MZI ₁₉ arranged in the fifth row are considered to be the first half, but they may also be considered to be the second half, or they may be considered to be the intermediate part between the first half and the second half.

The optical arithmetic device according to the second embodiment has the same effects as the optical arithmetic device according to the first embodiment, but in addition, since the first half MZIs and the second half MZIs in the multi-stage MZI array 1 can be adjusted in parallel for each row, adjustment can be completed in a shorter time.

Embodiment 3.
An optical arithmetic device according to a third embodiment will be described with reference to FIG.
The optical arithmetic device of embodiment 3 differs from the optical arithmetic device of embodiment 1 in that, while monitor photodetectors MPD ₀ to MPD ₃ are arranged corresponding to MZI ₀ to MZI ₃ arranged in the first row, the optical arithmetic device of embodiment 3 has a single aggregated MPD ₁ that is common to all MZI ₀ to MZI ₃ arranged in the first row, but is otherwise the same.
Therefore, the following description will focus on the differences.
In FIG. 10, the same reference numerals as those in FIGS. 1 to 6 denote the same or corresponding parts.

The monitor photodetector MPD _I is an aggregated photodiode having four input terminals, from the 0th input terminal to the 3rd input terminal, to which the first output ports Port3 of each of MZI ₀ , MZI ₁ , MZI ₂ , and MZI ₃ arranged in the first row are respectively connected.
The aggregated photodiode is a pin structure photodiode having a light absorption layer connected to one end of a waveguide, the other end of which is connected to the first output port Port3 of each of MZI ₀ , MZI ₁ , MZI ₂ , and MZI ₃ .

The specific procedure of the method for adjusting the MZI in the optical arithmetic device according to the third embodiment is substantially the same as the specific procedure of the method for adjusting the MZI in the optical arithmetic device according to the first embodiment, and the only difference is that in steps ST1 to ST9, monitoring is performed by the monitor photodetector MPD _I instead of by the monitor photodetectors MPD0 ₁ to MPD ₃ , respectively.

However, in step ST1, the adjustment order for MZI ₀ to MZI ₃ arranged in the first row is performed in the order of indexes in the first embodiment, but it is performed in the reverse order of indexes.
That is, in step ST1, for each of MZI ₀ to MZI ₃ , light is input to the monitor input nodes p ₀ to p ₃ corresponding to the indices in reverse order, and the light is monitored by the monitor photodetector MPD _I to perform adjustments to compensate for the branching ratio variations of each of MZI ₀ to MZI _3. After the adjustments, each of MZI ₀ to MZI ₃ is set to a through state.

In short, adjustment is performed by using the input node _p3 for the adjustment target MZI ₃ and the optical waveguide path _p3 →MZI ₃ →MPD _I , with the photodetector being MPD _I. After adjustment, MZI ₃ is set to the through state.
The optical waveguide paths are formed and adjusted in the following order.
For the MZI ₂ to be adjusted, p ₂ →MZI ₂ →MPD _I .
For the MZI ₁ to be adjusted, p ₁ →MZI ₁ →MPD ₁ .
For the MZI ₀ to be adjusted, p ₀ →MZI ₀ →MPD _I .
In step ST1, adjustment of MZI ₀ to MZI ₃ arranged in the first row is completed and the characteristics become known.

In a multi-stage MZI array 1 having 2N+1 rows and 2N columns, the adjustment order of the MZIs arranged in the first row is the reverse order of the indexes, N-1, N-2, ..., 1, 0, that is, MZI _N-1 , MZI _N-2 , ..., MZI ₁ , MZI ₀ .
Steps ST2 to ST9 are the same as those in the first embodiment.

The optical arithmetic device according to the third embodiment has the same effects as the optical arithmetic device according to the first embodiment, and in addition, since it is integrated as a monitor photodetector MPD _I , the number of circuit elements can be reduced, and the yield and manufacturing costs can be improved.

Embodiment 4.
An optical arithmetic device according to a fourth embodiment will be described with reference to FIG.
In the optical computing device according to the second embodiment, first monitor photodetectors MPD ₀ to MPD 3 are arranged corresponding to MZI ₀ to MZI ₃ arranged in the first row, respectively, and second monitor photodetectors MPD ₄ to MPD ₇ are arranged corresponding to MZI ₃₂ to MZI ₃₅ arranged in the bottom row, the _ninth row, respectively.

In contrast to this, in the optical computing device according to the fourth embodiment, a single aggregated first monitor photodetector MPD _I is arranged in common for all of MZI ₀ to MZI ₃ arranged in the first row, and a single aggregated second monitor photodetector MPD _II is arranged in common for all of MZI ₃₂ to MZI ₃₅ arranged in the ninth row.

The optical arithmetic device according to the fourth embodiment differs from the optical arithmetic device according to the second embodiment in that a first monitor photodetector MPD _I and a second monitor photodetector MPD _II are provided, but is otherwise the same.
Therefore, the following description will focus on the differences.
In FIG. 10, the same reference numerals as those in FIGS. 1 to 6 denote the same or corresponding parts.

The first monitor photodetector MPD _I is an aggregated photodiode having four input terminals, from the 0th input terminal to the 3rd input terminal, to which the first output ports Port3 of each of MZI ₀ , MZI ₁ , MZI ₂ , and MZI ₃ arranged in the first row are respectively connected.
The aggregated photodiode is a pin structure photodiode having a light absorption layer connected to one end of a waveguide, the other end of which is connected to the first output port Port3 of each of MZI ₀ , MZI ₁ , MZI ₂ , and MZI ₃ .

The second monitor photodetector MPD _II is an integrated photodiode having four input ends, from the 0th input end to the 3rd input end, to which the first output ports Port3 of MZI ₃₂ , MZI ₃₃ , MZI ₃₄ , and MZI ₃₅ arranged in the 9th row are respectively connected.
The aggregated photodiode is a pin structure photodiode having a light absorption layer, one end of which is connected to the first output port Port3 of each of MZI ₃₂ , MZI ₃₃ , MZI ₃₄ , and MZI ₃₅ , and the other end of which is connected to a waveguide.

The specific steps of the method for adjusting the MZI in the optical arithmetic device according to the third embodiment are substantially the same as the specific steps of the method for adjusting the MZI in the optical arithmetic device according to the second embodiment, with the only difference being that in steps ST1 to ST5, monitoring by the first group of monitor photodetectors MPD ₀ to MPD ₃ is performed by the first monitor photodetector MPD _I , and monitoring by the second group of monitor photodetectors MPD ₀ to MPD ₇ is performed by the second monitor photodetector MPD _II .

However, in step ST1, in the first embodiment, the adjustment order for MZI ₀ to MZI ₃ arranged in the first row is in the order of indexes, and the adjustment order for MZI ₃₂ to MZI ₃₅ arranged in the ninth row is in the order of indexes, but in this embodiment, the adjustment order is reversed for the first row and the ninth row.

That is, in step ST1, for each of MZI ₀ to MZI ₃ , light is input to monitor input nodes p ₀ to p ₃ corresponding to the reverse order of the indexes, and adjustments are made to compensate for the branching ratio variations of each of MZI ₀ to MZI ₃ by monitoring with a monitor photodetector MPD _I , and at the same time, light is input to monitor input nodes p ₄ to p ₇ corresponding to the reverse order of the indexes for each of MZI ₃₂ to MZI _35, and adjustments are made to compensate for the branching ratio variations of each of MZI ₃₂ to MZI ₃₅ by monitoring with a monitor photodetector MPD _II .
After the adjustment, MZI ₀ to MZI ₃ and MZI ₃₂ to MZI ₃₅ are each set to the through state.

In short, the adjustment target MZI ₃ is set as the monitor input node _p3 , and the adjustment is performed along the optical waveguide path of _p3 →MZI ₃ →MPD _I , with the photodetector being MPD _I.
At the same time, the adjustment target MZI ₃₅ is set as a monitor input node _p7 , and adjustment is performed along the optical waveguide path _p7 →MZI ₃₅ →MPD _II , with the photodetector being MPD _II .
After adjustment, MZI ₃ and MZI ₃₅ are set to the through state.

The optical waveguide paths are formed and adjusted in the following order.
For the MZI ₂ to be adjusted, p ₂ →MZI ₂ →MPD _I , and for the MZI ₃₄ to be adjusted, p ₆ →MZI ₃₄ →MPD _II .
For the MZI ₁ to be adjusted, p ₁ →MZI ₁ →MPD _I , and for the MZI ₃₃ to be adjusted, p ₅ →MZI ₃₃ →MPD _II .

For the MZI ₀ to be adjusted, p ₀ →MZI ₀ →MPD _I , and for the MZI ₃₂ to be adjusted, p ₄ →MZI ₃₂ →MPD _II .
In step ST1, adjustment is completed for MZI ₀ to MZI ₃ arranged in the first row and MZI ₃₂ to MZI ₃₅ arranged in the ninth row, and the characteristics become known.

In a multi-stage MZI array 1 having 2N+1 rows and 2N columns, the adjustment order of the MZIs arranged in the first row is the reverse order of the indexes, N-1, N-2, ..., 1, 0, that is, MZI _N-1 , MZI _N-2 , ..., MZI ₁ , MZI ₀ .
In addition, the adjustment order of the MZIs arranged in the bottom row (2N+1) is the reverse order of the index, that is, in descending order of the column number, that is, (2N+1)N-1, (2N+1)N-2, ..., (2N+1)N-N, that is, MZI _(2N+1)N-1 , MZI _(2N+1)N-2 , ..., MZI _(2N+1)N-N .
Steps ST2 to ST5 are the same as those in the second embodiment.

The optical arithmetic device according to the fourth embodiment has the same effects as the optical arithmetic device according to the second embodiment, and in addition, since it is integrated into a first monitor photodetector MPD _I and a second monitor photodetector MPD _II , the number of circuit elements can be reduced, and the yield and manufacturing costs can be improved.

In addition, in the fourth embodiment, instead of the aggregated first monitor photodetector MPD 1 that is disposed in common for all of MZI ₀ to MZI ₃ arranged in the first row, first monitor photodetectors MPD ₀ to MPD ₃ may be disposed corresponding to MZI ₀ to MZI ₃ arranged in the first row, respectively, as shown in the optical arithmetic device according to _the second embodiment.
That is, the first monitor photodetectors MPD ₀ to MPD ₃ may be arranged corresponding to MZI ₀ to MZI ₃ arranged in the first row, respectively, and a single aggregated second monitor photodetector MPD _II may be arranged corresponding to all MZI ₃₂ to MZI ₃₅ arranged in the ninth row.

Furthermore, in the fourth embodiment, instead of the aggregated second monitor photodetector MPD _II that is disposed in common corresponding to all of the MZI ₃₂ to MZI ₃₅ arranged in the ninth row, second monitor photodetectors MPD ₄ to MPD ₇ may be disposed corresponding to the MZI ₃₂ to MZI ₃₅ arranged in the ninth row, respectively, as shown in the optical arithmetic device according to the second embodiment.

That is, a single aggregated first monitor photodetector MPD 1 may be arranged in common for all of MZI ₀ to MZI ₃ arranged in the first row, and second monitor photodetectors MPD ₄ to MPD ₇ may be arranged corresponding to _{MZI 32} to MZI ₃₅ arranged in the ninth row, respectively.

Furthermore, it is possible to freely combine the embodiments, modify any of the components in each embodiment, or omit any of the components in each embodiment.

The optical computing device disclosed herein is applicable to applications using analog computing technologies such as machine learning and quantum computing.

1 multi-stage MZI array, MZI ₀ to MZI ₃₅ Mach-Zehnder interferometers, MPD ₀ to MPD ₃ , MPD _I (first) monitor photodetector, MPD ₄ to MPD ₇ , MPD _II second monitor photodetector, x ₀ to x ₇ input nodes, y ₀ to y ₇ output nodes.

Claims (15)

a multi-stage MZI array in which Mach-Zehnder interferometers arranged in odd-numbered columns are arranged in even-numbered rows, and Mach-Zehnder interferometers arranged in even-numbered columns are arranged in odd-numbered rows in a matrix of multiple rows and multiple columns, and the Mach-Zehnder interferometers are arranged in a rectangular shape;
a monitor photodetector connected to a first output port of the Mach-Zehnder interferometer arranged in the first row;
An optical computing device comprising: In the multistage MZI array, where the number of Mach-Zehnder interferometers arranged in the first column is N (N is a natural number greater than or equal to 2), the number of rows is 2N+1 and the number of columns is 2N; the index of the Mach-Zehnder interferometer in the first row and second column is set to 0, and the indexes of the Mach-Zehnder interferometers arranged in the first row are increased by one in order of column; the Mach-Zehnder interferometer in the second row and first column is placed after the Mach-Zehnder interferometer in the first row and Nth column; the indexes of the Mach-Zehnder interferometers arranged in the second row are increased by one in order of column; and similarly, the indexes are assigned in order up to [N×(2N+1)-1] up to the Mach-Zehnder interferometer in the 2N+1 row and Nth column. An optical computing device according to claim 1 or claim 2, wherein the monitor photodetectors are arranged corresponding to each of the Mach-Zehnder interferometers arranged in the first row. An optical computing device as described in claim 1 or claim 2, wherein the monitor photodetector is commonly arranged for all Mach-Zehnder interferometers arranged in the first row. An optical computing device according to any one of claims 1 to 4, further comprising a second monitor photodetector connected to a second output port of the Mach-Zehnder interferometer arranged in the bottom row. The optical computing device described in claim 5, wherein the second monitor photodetectors are arranged corresponding to the Mach-Zehnder interferometers arranged in the bottom row. The optical computing device described in claim 5, wherein the second monitor photodetector is arranged in common for all Mach-Zehnder interferometers arranged in the bottom row. A method for adjusting a Mach-Zehnder interferometer in an optical computing device including a multi-stage MZI array in which Mach-Zehnder interferometers arranged in odd-numbered columns are arranged in even-numbered rows and Mach-Zehnder interferometers arranged in even-numbered columns are arranged in odd-numbered rows in a matrix of multiple rows and multiple columns, and the Mach-Zehnder interferometers arranged in the even-numbered columns are arranged in odd-numbered rows, and a monitor photodetector connected to a first output port of the Mach-Zehnder interferometer arranged in the first row,
A method for adjusting a Mach-Zehnder interferometer in an optical computing device, in which an adjustment is made to compensate for branching ratio variations in a Mach-Zehnder interferometer to be adjusted, with a Mach-Zehnder interferometer arranged in a front row in an optical waveguide path through which light passes through the Mach-Zehnder interferometer to be adjusted being set to a through state, and a Mach-Zehnder interferometer arranged in a rear row in the optical waveguide path being set to a cross state. The Mach-Zehnder interferometer arranged in the first row to be adjusted receives an optical signal at its first input port, and outputs the optical signal to the monitor photodetector connected to its first output port, thereby performing adjustment;
The Mach-Zehnder interferometers arranged in the second row and thereafter to be adjusted are adjusted with only the investigated Mach-Zehnder interferometers as the optical waveguide path.
A method for adjusting a Mach-Zehnder interferometer in an optical computing device according to claim 8. The monitor photodetectors are arranged corresponding to the Mach-Zehnder interferometers arranged in the first row,
The Mach-Zehnder interferometers arranged in the first row are adjusted in ascending order of the number of columns in which they are arranged, by inputting an optical signal to their first input port and outputting an optical signal to their corresponding monitor photodetectors connected to their first output port.
10. A method for adjusting a Mach-Zehnder interferometer in an optical computing device according to claim 8 or 9. the monitor photodetector is disposed in common for all Mach-Zehnder interferometers disposed in the first row,
The Mach-Zehnder interferometers arranged in the first row are adjusted in descending order of the number of columns in which they are arranged, by inputting an optical signal to their first input port and outputting an optical signal to a common monitor photodetector connected to their first output port.
10. A method for adjusting a Mach-Zehnder interferometer in an optical computing device according to claim 8 or 9. In a matrix with multiple rows and columns, Mach-Zehnder interferometers arranged in odd-numbered columns are arranged in even-numbered rows, and Mach-Zehnder interferometers arranged in even-numbered columns are arranged in odd-numbered rows. The Mach-Zehnder interferometers are arranged in a rectangular configuration. The multi-stage MZI array includes a monitor photodetector connected to the first output port of the Mach-Zehnder interferometer arranged in the first row. In the multi-stage MZI array, if the number of Mach-Zehnder interferometers arranged in the first column is N (N is a natural number greater than or equal to 2), the number of rows is 2N+1, the number of columns is 2N, and the index of the Mach-Zehnder interferometer in the first row and second column is 0. The index of the Mach-Zehnder interferometer arranged in the first row is increased by one in the order of the columns, with the Mach-Zehnder interferometer in the second row, column 1 being placed after the Mach-Zehnder interferometer in the first row, column N; the index of the Mach-Zehnder interferometer arranged in the second row is increased by one in the order of the columns, and similarly, the index [N×(2N+1)-1] is assigned up to the Mach-Zehnder interferometer in the (2N+1)th row, column N; and an adjustment to compensate for branching ratio variations in the Mach-Zehnder interferometer is performed in the order of the indexes from 0 to [N×(2N+1)-1]. A method for adjusting a Mach-Zehnder interferometer in an optical computing device including: a multi-stage MZI array in which Mach-Zehnder interferometers arranged in odd-numbered columns are arranged in even-numbered rows and Mach-Zehnder interferometers arranged in even-numbered columns are arranged in odd-numbered rows in a matrix of multiple rows and multiple columns, the Mach-Zehnder interferometers being arranged in a rectangular shape; a first monitor photodetector connected to a first output port of the Mach-Zehnder interferometer arranged in the first row; and a second monitor photodetector connected to a second output port of the Mach-Zehnder interferometer arranged in the bottom row,
An optical signal is input to a first input port of a Mach-Zehnder interferometer arranged in the first row to be adjusted to compensate for branching ratio variations, and the optical signal is output to the first monitor photodetector connected to its first output port, thereby performing adjustment; and at the same time, an optical signal is input to a second input port of a Mach-Zehnder interferometer arranged in the bottom row to be adjusted to compensate for branching ratio variations, and the optical signal is output to the second monitor photodetector connected to its second output port, thereby performing adjustment;
The multiple rows are divided into a first half and a second half, and adjustments are made to simultaneously compensate for branching ratio variations for the Mach-Zehnder interferometers arranged in the first half rows and the Mach-Zehnder interferometers arranged in the second half rows, starting from the second row in the first half to the bottom row in the first half, and from the bottom row to the row immediately before that to the top row in the second half, and the Mach-Zehnder interferometers arranged in each row are adjusted with only the investigated Mach-Zehnder interferometer as the optical waveguide path.
A method for adjusting a Mach-Zehnder interferometer in an optical computing device. the first monitor photodetectors are arranged corresponding to the Mach-Zehnder interferometers arranged in the first row,
the second monitor photodetectors are arranged corresponding to the Mach-Zehnder interferometers arranged in the bottom row,
The Mach-Zehnder interferometers arranged in the first row are adjusted in ascending order of the number of columns in which they are arranged, by inputting an optical signal to their own first input port and outputting an optical signal to their corresponding first monitor photodetectors which are connected to their own first output port;
The Mach-Zehnder interferometers arranged in the bottom row are adjusted in ascending order of the number of columns in which they are arranged, by inputting an optical signal to their first input port and outputting an optical signal to their corresponding second monitor photodetectors connected to their first output port.
A method for adjusting a Mach-Zehnder interferometer in an optical computing device according to claim 13. the first monitor photodetector is disposed in common to all the Mach-Zehnder interferometers disposed in the first row;
the second monitor photodetector is disposed in common to all of the Mach-Zehnder interferometers disposed in the bottom row;
The Mach-Zehnder interferometers arranged in the first row are adjusted in descending order of the number of columns in which they are arranged, by inputting an optical signal to their first input port and outputting an optical signal to a common first monitor photodetector connected to their first output port;
The Mach-Zehnder interferometers arranged in the bottom row are adjusted in descending order of the number of columns in which they are arranged, by inputting an optical signal to their first input port and outputting an optical signal to a common second monitor photodetector connected to their first output port.
A method for adjusting a Mach-Zehnder interferometer in an optical computing device according to claim 13.