Semiconductor photonic device

By using a three-dimensional photonic integrated circuit to perform sub-operations of convolution operations with optical lenses and spatial light modulators, the problem of high power consumption in complex calculations of electronic integrated circuits is solved, and efficient and low-power artificial intelligence computing is realized.

CN223808612UActive Publication Date: 2026-01-16TAIWAN SEMICONDUCTOR MANUFACTURING CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202520227379.3
Authority / Receiving Office
CN · China
Patent Type
Utility models(China)
Current Assignee / Owner
Priority Date
2024-03-20
Filing Date
2025-02-13
Publication Date
2026-01-16
Estimated Expiration
2035-02-13

AI Technical Summary

Technical Problem

Existing electronic integrated circuits consume high power and require large resources when performing complex artificial neural network calculations, making it difficult to meet the rapidly growing demand for computing performance.

Method used

A three-dimensional photonic integrated circuit is used to perform sub-operations of convolution, including Fourier transform, multiplication and inverse Fourier transform, by using optical lenses and spatial light modulator structures. The calculation is achieved by passively propagating optical signals in three-dimensional space.

Benefits of technology

It significantly reduces computing power consumption, decreases reliance on electronic integrated circuits, and enables efficient processing of complex artificial intelligence tasks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN223808612U_ABST
    Figure CN223808612U_ABST
Patent Text Reader

Abstract

The embodiment of the utility model relates to a semiconductor photonic device, which comprises a three-dimensional photonic integrated circuit, and the three-dimensional photonic integrated circuit comprises an optical assembly which is configured to implement an artificial neural network such as a convolutional neural network (CNN) or a part of the CNN. For example, semiconductor photonic devices described herein may include a three-dimensional photonic integrated circuit including an optical lens structure and a spatial light modulator (SLM) structure configured to perform sub-operations of a convolution operation, including Fourier transform operations, multiplication operations, and inverse Fourier transform operations.
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] Embodiments of the present utility model relate to artificial neural network photonic integrated circuits and methods of forming the same. BACKGROUND

[0002] A convolutional neural network (CNN) is a type of artificial neural network designed primarily for processing and analyzing visual data. Visual data can include electronic visual data, such as electronic images and videos. CNNs are very effective in tasks such as image recognition, object detection, and even image generation. SUMMARY

[0003] Embodiments of the present utility model relate to a semiconductor photonic device comprising: a first spatial light modulator (SLM) structure; a first lens structure; a second SLM structure, wherein the first lens structure is between the first SLM structure and the second SLM structure; and a second lens structure, wherein the second SLM structure is between the first lens structure and the second lens structure.

[0004] Embodiments of the present utility model relate to a semiconductor photonic device comprising: a first spatial light modulator (SLM) layer comprising one or more first SLM structures; a first lens layer comprising one or more first lens structures; a second SLM layer comprising one or more second SLM structures, wherein the first lens layer is between the first SLM layer and the second SLM layer; and a second lens layer comprising one or more second lens structures, wherein the second SLM layer is between the first lens layer and the second lens layer.

[0005] Embodiments of the present utility model relate to a method comprising: providing a first spatial light modulator (SLM) structure of an artificial neural network photonic integrated circuit of a semiconductor photonic device; providing a first lens structure of the artificial neural network photonic integrated circuit over the first SLM structure, wherein the first lens structure is adjacent to the first SLM structure; providing a second SLM structure of the artificial neural network photonic integrated circuit over the first lens structure, wherein the first lens structure is between the first SLM structure and the second SLM structure; and providing a second lens structure of the artificial neural network photonic integrated circuit over the second SLM structure. BRIEF DESCRIPTION OF DRAWINGS

[0006] Aspects of the disclosure are best understood from the following detailed description when read in conjunction with the accompanying drawings. It is to be noted, however, that the various

[0007] Figure 1is a diagram of an example of an optical signal propagation by a photonic integrated circuit included in a semiconductor photonic device described herein.

[0008] Figure 2A and Figure 2B is a diagram of an example of an optical signal propagation by a photonic integrated circuit included in a semiconductor photonic device described herein.

[0009] Figures 3A to 3F is a diagram of an example of a semiconductor photonic device (or portion thereof) described herein.

[0010] Figure 4 is a diagram of an example of a semiconductor photonic device (or portion thereof) described herein.

[0011] Figure 5 is a diagram of an example of a semiconductor photonic device (or portion thereof) described herein.

[0012] Figure 6 is a diagram of an example of a semiconductor photonic device (or portion thereof) described herein.

[0013] Figure 7A and Figure 7B is a diagram of an example of a semiconductor photonic device (or portion thereof) described herein.

[0014] Figure 8 is a diagram of an example of a semiconductor photonic device (or portion thereof) described herein.

[0015] Figure 9 is a flowchart of an example process associated with a photonic integrated circuit forming a semiconductor photonic device described herein.

[0016] Figure 10 is a flowchart of an example process associated with a semiconductor photonic device described herein. DETAILED DESCRIPTION

[0017] The following disclosure provides many different embodiments, or examples, for implementing different aspects of the provided subject matter. Specific examples of components and arrangements are described below to simplify the present disclosure. These are, of course, merely examples and are not intended to be limiting. For example, in the following description, a first component forming a second component above or on the second component can include embodiments in which the first component and the second component are formed directly one on the other, and also embodiments in which additional components can be formed between the first component and the second component such that the first component and the second component can not be directly in contact. Additionally, the present disclosure can repeat use of reference numerals in the various examples and / or drawings in order to simplify the drawings and the present disclosure. This re-use of reference numerals is not to be construed as a limitation on the scope or functionality of the various embodiments and / or configurations described herein but is merely an aide to the reader in

[0018] Moreover, spatially relative terms, such as "beneath", "below", "lower", "above", "upper", and the like, can be used herein for ease of description to describe one element or structure's relationship to another element(s) or structure(s) as illustrated in the figures. The spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientations depicted in the figures. The devices can be otherwise oriented (rotated 90 degrees or at other orientations) and the spatially relative descriptors used herein interpreted accordingly.

[0019] A convolutional neural network (CNN) is composed of various layers, including convolution layers, pooling layers, and / or fully connected layers, among other examples. For image recognition, an input image can be passed through a series of convolution layers to detect features of the input image, such as edges, contours, patterns, and / or textures, among other examples. Pooling layers can be used to downsample feature maps of the detected features. Typically, a max-pooling operation is performed to downsample the feature maps. After down-sampling, the feature maps can be processed through fully connected layers to combine the features and identify a classification property of the input image using an activation function, such as a softmax and / or a rectified linear unit (ReLU) function, among other examples.

[0020] Each convolution layer can be implemented by a plurality of electronic integrated circuits of a semiconductor device, such as a graphics processing unit (GPU) or an artificial intelligence (AI) accelerator. The electronic integrated circuits are configured to perform a convolution (or correlation) operation of the convolution layer. The convolution operation involves applying a filter across an input image and extracting features of the input image using the filter. The convolution operation can include performing a Fourier transform on a plurality of input electrical signals (e.g., signals carrying data associated with the input image) to generate a plurality of transformed electrical signals. The Fourier transform transforms the input image from a spatial domain to a frequency domain to identify frequency components of the input image. Then, a multiplication operation of the convolution operation is performed, in which each frequency component of the transformed electrical signals is multiplied by the filter to modulate the frequency components of the input image based on a frequency response of the filter. Then, an inverse Fourier transform is performed to transform the output of the multiplication operation back into the spatial domain.

[0021] Convolution operations are computationally complex operations that typically require a large amount of processing resources (e.g., a large number of transistors and other types of integrated circuit devices) for electronic integrated circuits of semiconductor devices. CNNs can include many convolution layers, each of which performs one or more convolution operations on an input. In recent years, as the use and adoption of computer-implemented intelligence has significantly proliferated, the complexity of artificial neural networks used to implement machine learning, deep learning, AI, and other computer-implemented intelligence has increased exponentially. In some cases, the computational performance requirements for artificial neural networks have doubled approximately every 2 to 3 months in recent years, and are expected to continue to grow at this rate (or even faster). Implementing artificial neural networks with sufficient computational performance for complex computer-implemented intelligence tasks can often require the use of supercomputers or computing systems with hundreds or thousands of GPUs and / or AI accelerators. Advances in the manufacture of electronic integrated circuits can not keep pace with advances in the complexity of computer-implemented intelligence tasks. Even so, the power consumption of these electronic integrated circuits at exascale (or even zettascale) computational performance can be on the order of hundreds to thousands of megawatts.

[0022] In some implementations described herein, a semiconductor photonic device includes a three-dimensional photonic integrated circuit that includes optical components configured to implement an artificial neural network, such as a CNN or a portion thereof. For example, a semiconductor photonic device described herein can include a three-dimensional photonic integrated circuit that includes optical lens structures and spatial light modulator (SLM) structures configured to perform sub-operations of a convolution operation of an optical input signal, including a Fourier transform operation, a multiplication operation, and an inverse Fourier transform operation.

[0023] Performing these sub-operations of a convolution operation of an optical input signal using optical lenses and SLMs enables these sub-operations to be performed passively by enabling the optical input signal (and associated optical transform signals) to propagate through the photonic integrated circuit in three dimensions. For example, a first SLM structure can be placed at a front focal plane of a first optical lens structure, such that the first SLM structure can provide a modulated optical input signal (e.g., associated with an input image or another type of input) to the first optical lens structure, which passively performs a Fourier transform operation on the modulated optical input signal. A second SLM structure can be placed at a back focal plane of the first optical lens structure to receive a transformed optical signal and perform a multiplication operation (e.g., based on an optical filter signal) on the transformed optical signal to produce an output optical signal. The output optical signal from the multiplication operation is provided to a second optical lens structure, which passively performs an inverse Fourier transform operation on the output optical signal to produce a transformed output optical signal and provide the transformed output optical signal to one or more optical detectors.

[0024] As such, the three-dimensional photonic integrated circuits described herein are able to use photons (which have both amplitude and phase) to implement encoded optical signals that can be passively transformed to perform complex artificial intelligence tasks. Thus, the photonic integrated circuits described herein consume significantly less power and can be less complex than electronic integrated circuits (e.g., a three-dimensional photonic integrated circuit described herein can be able to process N optical input signals, whereas a semiconductor device can have 2N 2 electronic integrated circuits to process N electrical input signals), enabling further expansion of artificial neural networks.

[0025] Figure 1 FIG. is an example 100 of a semiconductor photonic device 102 including a photonic integrated circuit 104 described herein. Figure 1 A perspective view of the example 100 of the semiconductor photonic device 102 and related exploded view of the semiconductor photonic device 102 are illustrated, where details of layers of the semiconductor photonic device 102 are illustrated. The photonic integrated circuit 104 includes a portion of an artificial neural network, such as a CNN. The photonic integrated circuit 104 can be configured to perform convolution operations on the CNN. The convolution operations can include Fourier transform operations, multiplication operations, and inverse Fourier operations.

[0026] As shown in Figure 1 The semiconductor photonic device 102 can include a layer stack. In some implementations, the layer stack is formed on a substrate. In some implementations, the substrate can include a silicon (Si) substrate, a germanium (Ge) substrate, a binary semiconductor substrate (such as a III-V semiconductor substrate or a II-IV semiconductor substrate (e.g., a gallium nitride (GaN) substrate, a gallium arsenide (GaAs) substrate)), a silicon carbide (SiC) substrate, and / or another type of semiconductor substrate. In some implementations, the substrate is omitted from the semiconductor photonic device 102.

[0027] The layer stack can include a light source layer 106, an SLM layer 108 above the light source layer 106, a lens layer 110 above the SLM layer 108, an SLM layer 112 above the lens layer 110, a lens layer 114 above the SLM layer 112, and a photodetector layer 116 above the lens layer 114. The layers 106-116 can be stacked and arranged in the semiconductor photonic device 102 in a z-direction (e.g., a vertical direction). In some implementations, one or more additional layers are included in the layer stack. For example, one or more buffer layers can be included between two or more layers in the layer stack. As another example, one or more polarizer layers can be included between two or more layers in the layer stack. As another example, one or more filter layers can be included between two or more layers in the layer stack. As another example, one or more phase mask layers can be included between two or more layers in the layer stack.

[0028] The light source layer 106 can include one or more light source structures 118. The SLM layer 108 can include one or more SLM structures 120 formed or placed on an optically transparent substrate. The optically transparent substrate can include silicon oxide (SiO x ), glass, undoped silica glass (USG), and / or another type of optically transparent material. The lens layer 110 can include one or more lens structures 122 formed or placed on an optically transparent substrate. The SLM layer 112 can include one or more SLM structures 124 formed or placed on an optically transparent substrate. The lens layer 114 can include one or more lens structures 126 formed or placed on an optically transparent substrate. The photodetector layer 116 can include one or more photodetector structures 128.

[0029] The photonic integrated circuit 104 can include one or more light source structures 118, SLM structures 120, lens structures 122, SLM structures 124, lens structures 126, and one or more photodetector structures 128 stacked or arranged in the z-direction in the semiconductor photonic device 102. The light source structures 118, SLM structures 120, lens structures 122, SLM structures 124, lens structures 126, and photodetector structures 128 can correspond to a CNN or a portion thereof, such as a convolutional layer of the CNN. The light source structures 118, SLM structures 120, lens structures 122, SLM structures 124, lens structures 126, and photodetector structures 128 are arranged in the z-direction to enable optical signals to propagate in the z-direction (e.g., vertically) in the semiconductor photonic device 102 from the light source structures 118 to the photodetector structures 128 through the SLM structures 120, lens structures 122, SLM structures 124, and lens structures 126. Alternatively, the light source structures 118, SLM structures 120, lens structures 122, SLM structures 124, lens structures 126, and photodetector structures 128 are arranged in the x-direction or y-direction to enable optical signals to propagate in the x-direction or y-direction in the semiconductor photonic device 102 from the light source structures 118 to the photodetector structures 128 through the SLM structures 120, lens structures 122, SLM structures 124, and lens structures 126.

[0030] The light source structures 118 can include coherent light source structures (e.g., laser structures) and / or another type of light source structure capable of producing coherent optical signals (e.g., laser signals). Examples of such light source structures include semiconductor lasers (e.g., semiconductor laser diodes), such as edge-emitting diode lasers, quantum well lasers, and / or vertical cavity surface emitting laser (VCSEL) structures, among other examples. In some implementations, the one or more light source structures 118 are configured to produce one or more optical input signals associated with an input (e.g., an input image) of a convolution layer of the photonic integrated circuit 104. In some implementations, the one or more light source structures 118 are configured to produce one or more optical filter signals associated with a filter to be applied by a convolution layer of the photonic integrated circuit 104. In some implementations, the light source structures 118 are configured to produce an extended beam from a signal laser source. In some implementations, multiple light source structures 118 are configured to produce an array of laser signals phase-locked by a single pilot laser.

[0031] In some implementations, the light source structures 118 each include one or more semiconductor materials, such as silicon (Si), silicon germanium (SiGe), gallium arsenide (GaAs), indium phosphide (InP), gallium nitride (GaN), aluminum gallium arsenide (AlGaAs), indium gallium nitride (InGaN), and / or quantum dot semiconductor materials, among other examples. In some implementations, the light source structures 118 include a P-N junction between a p-type semiconductor layer and an n-type semiconductor layer. The p-type semiconductor layer can include a semiconductor material (e.g., silicon (Si)) doped with one or more p-type dopants (e.g., boron (B), aluminum (Al), and / or gallium (Ga), among other examples). The n-type semiconductor layer can include a semiconductor material (e.g., silicon (Si)) doped with one or more n-type dopants (e.g., phosphorus (P), arsenic (As), and / or antimony (Sb), among other examples).

[0032] The SLM structure 120 is positioned at the front focal plane of the lens structure 122 in the z-direction. The SLM structure 120 can extend in the x-y plane such that a first surface of the SLM structure 120 faces the light source structures 118 in the light source layer 106, and a second (opposite) surface of the SLM structure 120 faces the lens structure 122 in the lens layer 110.

[0033] The SLM structure 120 is configured to receive one or more optical input signals from the light source structure 118 and generate complex-valued optical input signals for input to the CNN. The complex-valued optical input signals can include a plurality of modulated optical input signals used to generate feature maps associated with input to the CNN. The SLM structure 120 includes a transmissive SLM structure that allows optical signals to pass (or be transmitted) through the SLM structure 120. Since the photons of the optical input signals received from the light source structure 118 have both an amplitude and a phase, the SLM structure 120 is able to generate modulated optical input signals based on the amplitude and phase of the optical input signals. In particular, the SLM structure 120 is configured to modulate both the amplitude and the phase of the optical input signals to generate the modulated optical input signals.

[0034] In some implementations, the SLM structure 120 includes a plurality of layers and / or structures including semiconductor materials, metallic materials, dielectric materials, and / or liquid crystal materials among other examples. For example, the SLM structure 120 can include a semiconductor layer (e.g., a silicon-based layer) that includes control electronics of the SLM structure 120. As another example, the SLM structure 120 can include one or more electrodes, a liquid crystal layer, and / or a dielectric passivation layer among other examples. Control circuitry in the semiconductor layer can be configured to control the refractive index in the liquid crystal layer, enabling modulation of the optical input signals. The electrodes can include one or more metals such as tungsten (W), cobalt (Co), ruthenium (Ru), titanium (Ti), aluminum (Al), copper (Cu), or gold (Au), another metal, and / or alloys thereof. The dielectric passivation layer can include one or more dielectric materials such as silicon oxide (SiO x ), silicon nitride (Si x N y ), silicon oxynitride (SiON), tetraethyl orthosilicate oxide, phosphosilicate glass (PSG), borophosphosilicate glass (BPSG), fluorinated silicon dioxide glass (FSG), carbon-doped silicon oxide, and / or another dielectric material.

[0035] In the semiconductor photonic device 102, the modulated optical input signals can propagate from the SLM structure 120 toward the lens structure 122 in the z-direction. Furthermore, in the semiconductor photonic device 102, the optical input signals can be spatially distributed in the x- and y-directions. Thus, the modulated optical input signals propagate through the photonic integrated circuit 104 in three dimensions, enabling the lens structure 122 to passively perform (e.g., without electrical input) a Fourier transform of the modulated optical input signals.

[0036] The lens structure 122 is positioned in the z-direction between the SLM structure 120 and the SLM structure 124. As indicated above, the lens structure 122 is configured to passively perform a Fourier transform of the modulated optical input signal to produce a transformed optical signal. The lens structure 122 can extend in the x-y plane such that a first surface of the lens structure 122 faces the SLM structure 120 in the SLM layer 108 and a second (opposite) surface of the lens structure 122 faces the SLM structure 124 in the SLM layer 112. The lens structure 122 can include a semiconductor lens (e.g., silicon (Si)), a dielectric lens (e.g., silicon oxide (SiO x ) lens, and / or a lens formed of another optically transparent dielectric material) and / or another type of lens structure. The lens can be convex, plano-convex, bi-convex, bi-concave, concave, and / or another lens shape. In some implementations, multiple lens structures 122 are included between the SLM structure 120 and the SLM structure 124 for performing the Fourier transform.

[0037] The SLM structure 124 is positioned at a back focal plane of the lens structure 122 in the z-direction and at a front focal plane of the lens structure 126 in the z-direction. The SLM structure 124 can extend in the x-y plane such that a first surface of the SLM structure 124 faces the lens structure 122 in the lens layer 110 and a second (opposite) surface of the SLM structure 124 faces the lens structure 126 in the lens layer 114.

[0038] The SLM structure 124 is configured to receive the transformed optical signal from the lens structure 122 and modulate the transformed optical signal to produce an output optical signal. In particular, the SLM structure 124 is configured to perform a multiplication operation of the CNN, which can include modulating the transformed optical signal based on a transformed filter optical signal (e.g., a spatial inverse filter).

[0039] The SLM structure 124 includes a transmissive SLM structure that allows optical signals to pass (or be transmitted) through the SLM structure 124. In some implementations, the SLM structure 124 includes multiple layers and / or structures that include semiconductor materials, metallic materials, dielectric materials, and / or liquid crystal materials, among other examples. For example, the SLM structure 124 can include a semiconductor layer (e.g., a silicon-based layer) that includes control electronics for the SLM structure 124. As another example, the SLM structure 124 can include one or more electrodes, a liquid crystal layer, and / or a dielectric passivation layer, among other examples. Control circuitry in the semiconductor layer can be configured to control the refractive index in the liquid crystal layer, enabling the modulated optical signal to be transformed. The electrodes can include one or more metals, such as tungsten (W), cobalt (Co), ruthenium (Ru), titanium (Ti), aluminum (Al), copper (Cu), or gold (Au), another metal, and / or alloys thereof. The dielectric passivation layer can include one or more dielectric materials, such as silicon oxide (SiO x ), silicon nitride (Si x N y ), silicon oxynitride (SiON), tetraethyl orthosilicate oxide, phosphosilicate glass (PSG), borophosphosilicate glass (BPSG), fluorinated silicon dioxide glass (FSG), carbon-doped silicon oxide, and / or another dielectric material.

[0040] In the semiconductor photonic device 102, the output optical signal can propagate in the z-direction from the SLM structure 124 toward the lens structure 126. Further, in the semiconductor photonic device 102, the optical input signal can be spatially distributed in the y-direction and the x-direction. Thus, the output optical signal propagates through the photonic integrated circuit 104 in three dimensions, enabling the lens structure 126 to passively perform (e.g., without electrical input) an inverse Fourier transform of the complex-valued optical input signal.

[0041] The lens structure 126 is positioned between the SLM structure 126 and the photodetector structure 128. As indicated above, the lens structure 126 is configured to passively perform an inverse Fourier transform of the optical output signal to produce a transformed optical output signal. The lens structure 126 can extend in the x-y plane such that a first surface of the lens structure 126 faces the SLM structure 124 in the SLM layer 112, and a second (opposite) surface of the lens structure 126 faces the photodetector structure 128 in the photodetector layer 116. The lens structure 126 can include a semiconductor lens (e.g., silicon (Si)), a dielectric lens (e.g., silicon oxide (SiO x) and / or another type of lens structure. The lens structure 126 can be convex, plano-convex, lenticular, biconcave, concave, and / or another lens shape. In some implementations, a plurality of lens structures 126 are included between the SLM structure 124 and the light detector structure 128 for performing an inverse Fourier transform.

[0042] The light detector structure 128 is positioned at a back focal plane of the lens structure 126 in the z-direction, and can be configured to receive the transformed optical output signal from the lens structure 126. The light detector structure 128 can include a semiconductor photodetector structure capable of generating an electrical signal based on the transformed optical output signal. In some implementations, the photonic integrated circuit 104 includes a plurality of light detector structures 128 arranged in an array in the y-direction and / or the x-direction in the semiconductor photonic device 102. In some implementations, each light detector structure 128 is configured to receive an associated transformed optical output signal, and generate an electrical signal based on the transformed optical output signal (e.g., based on an intensity or amplitude of the transformed optical output signal). In some implementations, the input feature map can be encoded on the electrical signal.

[0043] In some implementations, the light detector structures 128 each include a P-N junction diode, a P-I-N junction diode (e.g., a diode including a p-type semiconductor material / intrinsic semiconductor material / n-type semiconductor material junction), and / or another type of semiconductor structure capable of generating an electrical signal (referred to as a photocurrent) based on photons received in an optical signal. Photons generate electron / hole pairs in an absorption region (intrinsic semiconductor material) of the light detector structure 128, and the electrons and holes are separated and collected at opposite doped collection regions (p-type and n-type semiconductor materials).

[0044] Accordingly, the photonic integrated circuit 104 includes the SLM structure 120, the lens structure 122 adjacent to the SLM structure 124 in the z-direction, the SLM structure 124 adjacent to the lens structure 122 in the z-direction (the lens structure 122 positioned between the SLM structure 120 and the SLM structure 124 in the z-direction), and the lens structure 126 adjacent to the SLM structure 124 in the z-direction. The light source structure 118 can be positioned adjacent to the SLM structure 120 in the z-direction, such that the SLM structure 120 is between the light source structure 118 and the lens structure 122 in the z-direction. The light detector structure 128 can be positioned adjacent to the lens structure 126 in the z-direction, such that the lens structure 126 is between the light detector structure 128 and the SLM structure 124 in the z-direction.

[0045] SLM structures 120 and / or 126 can each include a transmissive SLM structure through which optical signals can pass. Lens structure 122 and lens structure 126 form a 4f optical system (e.g., a 4- focal point optical system), where SLM structure 120 is positioned at a front focal plane of lens structure 122, SLM structure 124 is positioned at a back focal plane of lens structure 122 and a front focal plane of lens structure 126, and light detector structure 128 is positioned at a back focal plane of lens structure 126.

[0046] As indicated above, providing Figure 1 as examples. Other examples can differ from what is described. Figure 1 described.

[0047] Figure 2A and Figure 2B is a diagram of an example 200 of optical signal propagation through a photonic integrated circuit 104 included in a semiconductor photonic device 102 described herein. Figure 2A illustrates a side view of a photonic integrated circuit 104 included in a semiconductor photonic device 102. As Figure 2A shown in FIG. 1, SLM structures 120 from SLM layer 108, lens structures 122 from lens layer 110, SLM structures 124 from SLM layer 112, and lens structures 126 from lens layer 114 are arranged (e.g., vertically arranged) in the z-direction in semiconductor photonic device 102, which enables optical signals to propagate in the z-direction in semiconductor photonic device 102 through and between SLM structures 120, lens structures 122, SLM structures 124, and lens structures 126. Photonic integrated circuit 104 can further include a plurality of light source structures 118a and 120b from light source layer 106, and a plurality of light detector structures 128a and 130b from light detector layer 116. Figure 2A The number of light source structures 118a and 120b and the number of light detector structures 128a and 130b illustrated in FIG. 1 are examples, and other numbers of light source structures and other numbers of light detector structures are within the scope of the present disclosure.

[0048] As indicated above, the SLM structure 120, the lens structure 122, the SLM structure 124, and the lens structure 126 are arranged in a 4f system that includes four focal planes 202, 204, 206, and 208. The focal plane 202 corresponds to the location of the SLM structure 120 and is the front focal plane of the lens structure 122 in the z-direction. The focal plane 204 corresponds to the location of the SLM structure 124 and is the back focal plane of the lens structure 122 in the z-direction. The focal plane 206 also corresponds to the location of the SLM structure 124 and is the front focal plane of the lens structure 126 in the z-direction. The focal plane 208 corresponds to the location of the light detector structure 128 and is the back focal plane of the lens structure 126 in the z-direction. The focal plane 202 can correspond to the object plane of the 4f system, the focal planes 204 and 206 can correspond to the pupil planes of the 4f system, and the focal plane 208 can correspond to the image plane of the 4f system.

[0049] The SLM structure 120 and the lens structure 122 can be separated in the z-direction by a focal length fl. The SLM structure 124 and the lens structure 122 can be separated in the z-direction by a focal length f2. The SLM structure 124 and the lens structure 126 can be separated in the z-direction by a focal length f3. The lens structure 126 and the light detector structures 128a and 130b can be separated in the z-direction by a focal length f4. In some implementations, the focal length fl and the focal length f2 are approximately equal. In some implementations, the focal length fl and the focal length f2 are different focal lengths. In some implementations, the focal length fl is greater than each of the focal lengths f3 and f4. In some implementations, the focal length f2 is greater than each of the focal lengths f3 and f4. In some implementations, the combined focal length of the focal lengths fl and f2 is greater than the combined focal length of the focal lengths f3 and f4. In some implementations, the focal length f3 and the focal length f4 are approximately equal. In some implementations, the focal length f3 and the focal length f4 are different focal lengths.

[0050] As Figure 2A As further shown in the example 200, the light source structures 118a and 120b can each generate respective optical input signals 210a and 210b. The optical input signals 210a and 210b can propagate in the z-direction toward the SLM structure 120. The optical input signals 210a and 210b can also propagate in the x-direction and / or the y-direction in the semiconductor photonic device 102. Figure 2A The number of optical input signals 210a and 210b illustrated in the example 200 is an example, and other numbers of optical input signals are within the scope of the present disclosure.

[0051] The optical input signals 210a and 210b can pass through the SLM structure 120, and the SLM structure 120 can modulate the optical input signals 210a and 210b to generate a plurality of modulated optical input signals 212a-c and a plurality of optical filter signals 214a-c. The modulated optical input signals 212a-c and the optical filter signals 214a-c can propagate in the z-direction toward the lens structure 122. The modulated optical input signals 212a-c and the optical filter signals 214a-c can also propagate in the x- and y-directions in the semiconductor photonic device 102. Figure 2A The number of modulated optical input signals 212a-c and the number of optical filter signals 214a-c illustrated in FIG. 2 are examples, and other numbers of modulated optical input signals and other numbers of optical filter signals are within the scope of the present disclosure.

[0052] The modulated optical input signals 212a-c and the optical filter signals 214a-c can be received at the lens structure 122, and the lens structure 122 can perform a Fourier transform of the modulated optical input signals 212a-c and the optical filter signals 214a-c. This results in the lens structure 122 generating transformed optical signals 216a-c from the modulated optical input signals 212a-c and transformed optical filter signals 218a-c from the modulated optical filter signals 214a-c. Figure 2A The number of transformed optical signals 216a-c and the number of transformed optical filter signals 218a-c illustrated in FIG. 2 are examples, and other numbers of transformed optical signals and other numbers of transformed optical filter signals are within the scope of the present disclosure.

[0053] The transformed optical signals 216a-c and the transformed optical filter signals 218a-c propagate in the z-direction from the lens structure 122 to the SLM structure 124 in the semiconductor photonic device 102. The transformed optical signals 216a-c and the transformed optical filter signals 218a-c also propagate in the x- and / or y-directions in the semiconductor photonic device 102.

[0054] Transformed optical signals 216a-216c and transformed optical filter signals 218a-218c can pass through SLM structure 124, and SLM structure 124 can perform a multiplication operation based on transformed optical filter signals 218a-218c to modulate transformed optical signals 216a-216c. For example, transformed optical signal 216a can be multiplied by transformed optical filter signal 218a, transformed optical signal 216b can be multiplied by transformed optical filter signal 218b, transformed optical signal 216c can be multiplied by transformed optical filter signal 218c, etc. This results in SLM structure 124 producing a plurality of modulated optical output signals 220a-220c and a plurality of modulated optical output signals 222a-222c. Modulated optical output signals 220a-220c and modulated optical output signals 222a-222c can propagate in the z-direction toward lens structure 126. Modulated optical output signals 220a-220c and modulated optical output signals 222a-222c can also propagate in the x-direction and / or y-direction in semiconductor photonic device 102. Figure 2A The number of modulated optical output signals 220a-220c and the number of modulated optical output signals 222a-222c described in the summary are examples, and other numbers of modulated optical output signals are within the scope of the present disclosure.

[0055] Modulated optical output signals 220a-220c and 222a-222c can be received at lens structure 126, and lens structure 126 can perform an inverse Fourier transform of modulated optical output signals 220a-220c and 222a-222c. This results in lens structure 126 producing transformed optical output signals 224a-224c from modulated optical output signals 220a-220c, and transformed optical output signals 226a-226c from modulated optical output signals 222a-222c. Figure 2A The number of transformed optical output signals 224a-224c and the number of transformed optical output signals 226a-226c described in the summary are examples, and other numbers of transformed optical output signals are within the scope of the present disclosure.

[0056] In the semiconductor photonic device 102, the transformed optical output signals 224a-c can propagate in the z-direction to the photodetector structure 128a, and the transformed optical output signals 226a-c can propagate in the z-direction to the photodetector structure 128b. The transformed optical output signals 224a-c and 226a-c can also propagate in the x- and / or y-directions in the semiconductor photonic device 102. The photodetector structure 128a can generate electrical output signals based on the transformed optical output signals 224a-c, and the photodetector structure 128b can generate electrical output signals based on the transformed optical output signals 226a-c.

[0057] As shown in Figure 2B the various types of waves. For example, the lens structure 122 can transform the spherical waves of the modulated optical input signals 212a-c and the spherical waves of the optical filter signals 214a-c into planar waves of the transformed optical signals 216a-c and the transformed optical filter signals 218a-c, respectively. As another example, the lens structure 126 can transform the planar waves of the modulated optical output signals 220a-c and 222a-c into the spherical waves of the transformed optical output signals 224a-c and 226a-c, respectively.

[0058] As indicated above, the provision of Figure 2A and Figure 2B are provided as examples. Other examples can differ from what is described with respect to Figure 2A and Figure 2B described.

[0059] Figures 3A to 3F is a diagram of an example 300 of forming a semiconductor photonic device (or a portion thereof) described herein. In particular, the example 300 includes an example of forming the photonic integrated circuit 104 in the semiconductor photonic device 102. In some implementations, one or more semiconductor processing operations described in connection with Figures 3A to 3F are performed using one or more semiconductor processing tools, such as a deposition tool, an exposure tool, a developer tool, an etch tool, a plating tool, a planarization tool, an ion implantation tool, a wafer / die transport tool, and / or another type of semiconductor processing tool.

[0060] In some implementations, the photonic integrated circuit 104 of the semiconductor photonic device 102 can be formed on and / or over a substrate. In some implementations, one or more components of the photonic integrated circuit 104 can be provided or placed on the substrate. The substrate can be provided in the form of a wafer, a die, and / or another type of workpiece. Alternatively, the substrate can be provided as a carrier wafer, a handle wafer, a semiconductor die frame, and / or another type of substrate. For example, the substrate can include a silicon-on-insulator (SOI) wafer, a dielectric wafer, a semiconductor wafer (e.g., a silicon wafer), and / or another type of wafer. As another example, the substrate can include an SOI die, a dielectric die, a semiconductor die (e.g., a silicon die), and / or another type of die. As another example, the substrate can include a removable substrate, such as a carrier wafer, a handle wafer, a die frame, a die holder, a die carrier, and / or another type of removable or temporary substrate.

[0061] As shown in Figure 3A In some implementations, one or more light source structures 118 of the photonic integrated circuit 104 can be fabricated in the light source layer 106, and the light source layer 106 (together with the light source structures 118 formed thereon) can be provided. For example, the light source layer 106 can be disposed over a substrate.

[0062] As shown in Figure 3B In some implementations, SLM structures 120 of the photonic integrated circuit 104 can be fabricated in the SLM layer 108, and the SLM layer 108 (together with the SLM structures 120 formed thereon) can be placed on or over the light source layer 106. In some implementations, the SLM structures 120 are fabricated in or placed on a transparent substrate of the SLM layer 108.

[0063] As shown in Figure 3C In some implementations, lens structures 122 of the photonic integrated circuit 104 can be fabricated in the lens layer 110, and the lens layer 110 (together with the lens structures 122 formed thereon) can be placed on or over the SLM layer 108. In some implementations, the lens structures 122 can be fabricated in or placed on a transparent substrate of the lens layer 110.

[0064] As shown in Figure 3D In some implementations, SLM structures 124 of the photonic integrated circuit 104 can be fabricated in the SLM layer 112, and the SLM layer 112 (together with the SLM structures 124 formed thereon) can be placed on or over the lens layer 110. In some implementations, the SLM structures 124 are fabricated in or placed on a transparent substrate of the SLM layer 112.

[0065] As shown in Figure 3EAs shown in FIG. 1, the photonic integrated circuit 104 can include a lens layer 114, an SLM layer 112, a light detector layer 116, and a light source layer 108. The lens layer 114 can include one or more lens structures 126 that can be formed in the lens layer 114 and / or placed on or above the lens layer 114. The SLM layer 112 can include one or more SLM structures 120 that can be formed in the SLM layer 112 and / or placed on or above the SLM layer 112. The light detector layer 116 can include one or more light detector structures 128 that can be formed in the light detector layer 116 and / or placed on or above the light detector layer 116. The light source layer 108 can include one or more light source structures 118 that can be formed in the light source layer 108 and / or placed on or above the light source layer 108.

[0066] As shown in FIG. 1, the photonic integrated circuit 104 can include a lens layer 114, an SLM layer 112, a light detector layer 116, and a light source layer 108. The lens layer 114 can include one or more lens structures 126 that can be formed in the lens layer 114 and / or placed on or above the lens layer 114. The SLM layer 112 can include one or more SLM structures 120 that can be formed in the SLM layer 112 and / or placed on or above the SLM layer 112. The light detector layer 116 can include one or more light detector structures 128 that can be formed in the light detector layer 116 and / or placed on or above the light detector layer 116. The light source layer 108 can include one or more light source structures 118 that can be formed in the light source layer 108 and / or placed on or above the light source layer 108. Figure 3F As shown in FIG. 1, the photonic integrated circuit 104 can include a lens layer 114, an SLM layer 112, a light detector layer 116, and a light source layer 108. The lens layer 114 can include one or more lens structures 126 that can be formed in the lens layer 114 and / or placed on or above the lens layer 114. The SLM layer 112 can include one or more SLM structures 120 that can be formed in the SLM layer 112 and / or placed on or above the SLM layer 112. The light detector layer 116 can include one or more light detector structures 128 that can be formed in the light detector layer 116 and / or placed on or above the light detector layer 116. The light source layer 108 can include one or more light source structures 118 that can be formed in the light source layer 108 and / or placed on or above the light source layer 108.

[0067] In some implementations, one or more of the light source structures 118, the SLM structures 120, the lens structures 122, the SLM structures 124, the lens structures 126, and / or the light detector structures 128 can be provided after partial fabrication, and the fabrication of these components can be completed after being provided.

[0068] In some implementations, one or more components of the photonic integrated circuit 104 can be formed from a semiconductor layer, such as the light source structures 118, the SLM structures 120, the lens structures 122, the SLM structures 124, the lens structures 126, and / or the light detector structures 128. For example, a hard mask layer can be formed on and / or above a semiconductor layer, and a pattern in the hard mask layer can be used to etch the semiconductor layer to remove portions of the semiconductor layer based on the pattern to form the light source structures 118, the SLM structures 120, the lens structures 122, the SLM structures 124, the lens structures 126, and / or the light detector structures 128.

[0069] A deposition tool can be used to deposit a hard mask layer on the semiconductor layer (e.g., using a chemical vapor deposition (CVD) technique, a physical vapor deposition (PVD) technique, and / or another type of deposition technique) and to deposit a photoresist layer on the hard mask layer (e.g., using a spin-on technique and / or another type of deposition technique). The hard mask layer can include silicon nitride (Si x N ymaterial or another hard mask material. The photoresist layer can include a photosensitive material that can be patterned using an exposure tool, such as a deep ultraviolet (DUV) lithography tool and / or an extreme ultraviolet (EUV) lithography tool, among other examples. An exposure tool can be used to expose the photoresist layer to a source of radiation to form a pattern in the photoresist layer. A developer tool can be used to develop and remove portions of the photoresist layer to expose the pattern. An etch tool can be used to etch the hard mask layer to transfer the pattern from the photoresist layer to the hard mask layer. An etch tool can then be used to etch the semiconductor layer based on the pattern in the hard mask layer to remove material from the semiconductor layer to form a pattern to form the light source structures 118, the SLM structures 120, the lens structures 122, the SLM structures 124, the lens structures 126, and / or the photodetector structures 128. In some implementations, the etching operation includes a plasma etching operation, a wet chemical etching operation, and / or another type of etching operation. In some implementations, a photoresist removal tool removes remaining portions of the photoresist layer (e.g., using a chemical stripper, a plasma ashing, and / or another technique).

[0070] In some implementations, one or more portions of the light source structures 118, one or more portions of the SLM structures 120, one or more portions of the lens structures 122, one or more portions of the SLM structures 124, one or more portions of the lens structures 126, and / or one or more portions of the photodetector structures 128 are formed from the semiconductor layer. Additional portions of the light source structures 118, additional portions of the SLM structures 120, additional portions of the lens structures 122, additional portions of the SLM structures 124, additional portions of the lens structures 126, and / or additional portions of the photodetector structures 128 can be formed from other layers and / or structures in the semiconductor photonic device 102.

[0071] As shown in Figure 3F , the light source layer 106, the SLM layer 108, the lens layer 110, the SLM layer 112, the lens layer 114, and the photodetector layer 116 of the semiconductor photonic device 102 can be formed such that the light source structures 118, the SLM structures 120, the lens structures 122, the SLM structures 124, the lens structures 126, and the photodetector structures 128 of the photonic integrated circuit 104 are arranged in the z-direction in the semiconductor photonic device 102.

[0072] As indicated above, the provision Figures 3A to 3F is by way of example. Other examples can differ from what is described with respect to Figures 3A to 3F what is described.

[0073] Figure 4 is an example 400 of a semiconductor photonic device 102 including the photonic integrated circuit 104 described herein. Figure 4A perspective view of an example 400 of the semiconductor photonic device 102 and related exploded view of the semiconductor photonic device 102, in which details of layers of the semiconductor photonic device 102 are illustrated. The photonic integrated circuit 104 includes portions of an artificial neural network, such as a CNN. The photonic integrated circuit 104 can be configured to perform convolution operations on the CNN. The convolution operations can include Fourier transform operations, multiplication operations, and inverse Fourier operations.

[0074] As shown in Figure 4 In the example 400, the semiconductor photonic device 102 includes a layer stack similar to the example 100 in Figure 1 with the exception that the semiconductor photonic device 102 includes multiple SLM layers 108a and 108b between the light source layer 106 and the lens layer 110 in the z-direction. Thus, in the example 400, the photonic integrated circuit 104 includes multiple SLM structure components 120a and 120b between the light source structure 118 and the lens structure 122 in the z-direction. The SLM structure component 120a can be adjacent (e.g., vertically adjacent) to the light source structure 118 in the z-direction, and the SLM structure component 120b can be adjacent (e.g., vertically adjacent) to the lens structure 122 in the z-direction.

[0075] The SLM structure components 120a and 120b can be configured to perform respective functions for modulating an optical input signal received from the light source structure 118. For example, the SLM structure component 120a can include a transmissive phase-only SLM structure configured to perform phase modulation of the optical input signal, and the SLM structure component 120b can include a transmissive amplitude SLM (e.g., amplitude modulation SLM structure component) configured to perform amplitude modulation of the optical input signal. The combined phase and amplitude modulation performed by the SLM structure components 120a and 120b enables a modulated optical input signal to be generated. In some implementations, the SLM structure component 120b (e.g., amplitude SLM) includes a transmissive phase-only SLM in combination with one or more polarizers.

[0076] As indicated above, a transmissive phase-only SLM is provided Figure 4 as an example. Other examples can differ from what is described with respect to Figure 4 .

[0077] Figure 5 is a diagram of an example 500 of a semiconductor photonic device 102 including a photonic integrated circuit 104 described herein. Figure 5A perspective view of an example 500 of a semiconductor photonic device 102 and a related exploded view of the semiconductor photonic device 102 are illustrated, where details of layers of the semiconductor photonic device 102 are illustrated. The photonic integrated circuit 104 includes portions of an artificial neural network, such as a CNN. The photonic integrated circuit 104 can be configured to perform convolution operations on the CNN. The convolution operations can include Fourier transform operations, multiplication operations, and inverse Fourier operations.

[0078] As shown in Figure 5 In the example 500, the semiconductor photonic device 102 includes a layer stack similar to the example 100 in Figure 1 with the exception that the semiconductor photonic device 102 includes a plurality of SLM layers 112a and 112b between the lens layer 110 and the lens layer 114 in the z-direction. Thus, in the example 500, the photonic integrated circuit 104 includes a plurality of SLM structure components 124a and 124b between the lens structure 122 and the lens structure 126 in the z-direction. The SLM structure component 124a can be adjacent to the lens structure 122 in the z-direction, and the SLM structure component 124b can be adjacent to the lens structure 126 in the z-direction.

[0079] The SLM structure components 124a and 124b can be configured to perform respective functions for modulating transformed optical signals received from the lens structure 122. For example, the SLM structure component 124a can include a transmissive phase-only SLM structure configured to perform phase modulation of the transformed optical signals, and the SLM structure component 124b can include a transmissive amplitude SLM (e.g., amplitude modulation SLM structure component) configured to perform amplitude modulation of the transformed optical signals. The combined phase and amplitude modulation performed by the SLM structure components 124a and 124b enable a modulated optical output signal to be generated. In some implementations, the SLM structure component 124b (e.g., amplitude SLM) includes a transmissive phase-only SLM in combination with one or more polarizers.

[0080] As indicated above, the provision Figure 5 is provided as an example. Other examples can differ from what is described with respect to the Figure 5 described.

[0081] Figure 6 is a diagram of an example 600 of a semiconductor photonic device 102 including a photonic integrated circuit 104 described herein. Figure 6 A perspective view of the example 600 of the semiconductor photonic device 102 and a related exploded view of the semiconductor photonic device 102 are illustrated, where details of layers of the semiconductor photonic device 102 are illustrated. The photonic integrated circuit 104 includes portions of an artificial neural network, such as a CNN. The photonic integrated circuit 104 can be configured to perform convolution operations on the CNN. The convolution operations can include Fourier transform operations, multiplication operations, and inverse Fourier operations.

[0082] As Figure 6 illustrated in example 600, the semiconductor photonic device 102 includes a layer stack similar to that of example 100 in Figure 1 , except that the semiconductor photonic device 102 includes multiple SLM layers 108a and 108b between the light source layer 106 and the lens layer 110 in the z-direction, and multiple SLM layers 112a and 112b between the lens layer 110 and the lens layer 114 in the z-direction. Thus, the photonic integrated circuit 104 includes multiple SLM structure components 120a and 120b between the light source structure 118 and the lens structure 122 in the z-direction, and multiple SLM structure components 124a and 124b between the lens structure 122 and the lens structure 126 in the z-direction. The SLM structure component 120a can be adjacent to the light source structure 118 in the z-direction, and the SLM structure component 120b can be adjacent to the lens structure 122 in the z-direction. The SLM structure component 124a can be adjacent to the lens structure 122 in the z-direction, and the SLM structure component 124b can be adjacent to the lens structure 126 in the z-direction.

[0083] The SLM structure components 120a and 120b can be configured to perform respective functions for modulating an optical input signal received from the light source structure 118. For example, the SLM structure component 120a can include a transmissive-only phase SLM structure configured to perform phase modulation of the optical input signal, and the SLM structure component 120b can include a transmissive amplitude SLM (e.g., an amplitude modulation SLM structure component) configured to perform amplitude modulation of the optical input signal. The combined phase and amplitude modulation performed by the SLM structure components 120a and 120b enables a modulated optical input signal to be generated.

[0084] The SLM structure components 124a and 124b can be configured to perform respective functions for modulating a transformed optical signal received from the lens structure 122. For example, the SLM structure component 124a can include a transmissive-only phase SLM structure configured to perform phase modulation of the transformed optical signal, and the SLM structure component 124b can include a transmissive amplitude SLM (e.g., an amplitude modulation SLM structure component) configured to perform amplitude modulation of the transformed optical signal. The combined phase and amplitude modulation performed by the SLM structure components 124a and 124b enables a modulated optical output signal to be generated.

[0085] As indicated above, providing Figure 6 as an example. Other examples can differ from what is described with respect to Figure 6 .

[0086] Figure 7A and Figure 7Bis a diagram of an example of a semiconductor photonic device 102 including a photonic integrated circuit 104 described herein. Figure 7A A perspective view of an example 700 of a semiconductor photonic device 102 is illustrated, along with a related exploded view of the semiconductor photonic device 102, in which details of layers of the semiconductor photonic device 102 are illustrated. The photonic integrated circuit 104 includes portions of an artificial neural network, such as a CNN. The photonic integrated circuit 104 can be configured to perform convolution operations on the CNN. The convolution operations can include Fourier transform operations, multiplication operations, and inverse Fourier operations.

[0087] As shown in Figure 7A , in the example 700, the semiconductor photonic device 102 includes a layer stack similar to the example 100 in Figure 1 , except that the semiconductor photonic device 102 includes one or more spacer layers 702, 704, 706, 708, and / or 710. Each spacer layer 702-710 includes a layer of optically transparent material, such as silicon oxide (SiO x ), glass, USG, and / or another type of optically transparent material. Each spacer layer 702-710 can be included between two layers of the semiconductor photonic device 102 in the z-direction. For example, the spacer layer 702 can be included vertically (e.g., in the z-direction) between the light source layer 106 and the SLM layer 108, the spacer layer 704 can be included vertically (e.g., in the z-direction) between the SLM layer 108 and the lens layer 110, the spacer layer 706 can be included vertically (e.g., in the z-direction) between the lens layer 110 and the SLM layer 112, the spacer layer 708 can be included vertically (e.g., in the z-direction) between the SLM layer 112 and the lens layer 114, and / or the spacer layer 710 can be included vertically (e.g., in the z-direction) between the lens layer 114 and the photodetector layer 116.

[0088] In some implementations, the semiconductor photonic device 102 includes a greater number of spacer layers than the number illustrated in Figure 7A . In some implementations, the semiconductor photonic device 102 includes a lesser number of spacer layers than the number illustrated in Figure 7A . In some implementations, the semiconductor photonic device 102 includes a different arrangement of spacer layers than the number illustrated in Figure 7A .

[0089] The spacer layers 702-710 can be included to enable bonding between components of the photonic integrated circuit 104 Figure 2A and Figure 2BThe described focal distances fl to f4 can be achieved. For example, a spacer layer 702 between the SLM layer 108 and the light source layer 106 and / or a spacer layer 704 between the SLM layer 108 and the lens layer 110 can enable the SLM structure 120 and the lens structure 122 of the photonic integrated circuit 104 to be separated in the z-direction by a focal distance fl. The thickness of the spacer layer 702 and / or the thickness of the spacer layer 704 can be selected to achieve the focal distance fl. As another example, a spacer layer 706 between the lens layer 110 and the SLM layer 112 can enable the SLM structure 124 and the lens structure 122 of the photonic integrated circuit 104 to be separated in the z-direction by a focal distance f2. The thickness of the spacer layer 706 between the lens layer 110 and the SLM layer 112 can be selected to achieve the focal distance f2.

[0090] As another example, a spacer layer 708 between the SLM layer 112 and the lens layer 114 can enable the SLM structure 124 and the lens structure 126 of the photonic integrated circuit 104 to be separated in the z-direction by a focal distance f3. The thickness of the spacer layer 708 between the SLM layer 112 and the lens layer 114 can be selected to achieve the focal distance f3. As another example, a spacer layer 710 between the lens layer 114 and the photodetector layer 116 can enable the lens structure 126 and the photodetector structure 128 of the photonic integrated circuit 104 to be separated in the z-direction by a focal distance f4. The thickness of the spacer layer 710 between the lens layer 114 and the photodetector layer 116 can be selected to achieve the focal distance f4.

[0091] In some embodiments, the spacer layers 702, 704, 706, 708, and 710 each have substantially the same thickness. In some embodiments, two or more of the spacer layers 702, 704, 706, 708, and / or 710 have different thicknesses.

[0092] Figure 7BA perspective view illustrating an example 712 of a semiconductor photonic device 102. The example 712 of a semiconductor photonic device 102 is similar to the example 700 of a semiconductor photonic device 102, except that one or more spacer inserts 714, 716, 718, 720, and / or 722 are included between vertically adjacent layers of the semiconductor photonic device 102 in place of one or more spacer layers 702, 704, 706, 708. For example, a spacer insert 714 can be included vertically (e.g., in the z-direction) between the light source layer 106 and the SLM layer 108, a spacer insert 716 can be included vertically (e.g., in the z-direction) between the SLM layer 108 and the lens layer 110, a spacer insert 718 can be included vertically (e.g., in the z-direction) between the lens layer 110 and the SLM layer 112, a spacer insert 720 can be included vertically (e.g., in the z-direction) between the SLM layer 112 and the lens layer 114, and / or a spacer insert 722 can be included vertically (e.g., in the z-direction) between the lens layer 114 and the photodetector layer 116. The spacer inserts 714, 716, 718, 720, and / or 722 can be positioned at comers of layers of the layer stack, such that the spacer inserts 714, 716, 718, 720, and / or 722 do not obstruct the light transmission path of the photonic integrated circuit 104.

[0093] In some implementations, the semiconductor photonic device 102 includes a greater number of spacer inserts than the number illustrated in Figure 7B In some implementations, the semiconductor photonic device 102 includes a greater number of spacer inserts than the number illustrated in Figure 7B In some implementations, the semiconductor photonic device 102 includes a greater number of spacer inserts than the number illustrated in Figure 7B In some implementations, the semiconductor photonic device 102 includes a different configuration of spacer inserts than the number illustrated in In some implementations, two or more of the inserts 714, 716, 718, 720, and / or 722 have different thicknesses.

[0094] As indicated above, the examples of Figure 7A and Figure 7B are provided as examples. Other examples can differ from what is described with respect to Figure 7A and Figure 7B .

[0095] Figure 8 is a diagram of an example 800 of a semiconductor photonic device 102 including a photonic integrated circuit 104 described herein. Figure 8A perspective view illustrating an example 800 of the semiconductor photonic device 102 and related exploded view of the semiconductor photonic device 102 in which details of layers of the semiconductor photonic device 102 are illustrated. The photonic integrated circuit 104 includes portions of an artificial neural network, such as a CNN. The photonic integrated circuit 104 can be configured to perform convolution operations on the CNN. The convolution operations can include Fourier transform operations, multiplication operations, and inverse Fourier operations.

[0096] As shown in Figure 8 In the example 800, the semiconductor photonic device 102 includes a layer stack similar to the example 100 in Figure 1 except that one or more layers of the semiconductor photonic device 102 are coupled with a movable gantry 802. The movable gantry 802 can be used to move one or more of the light source layer 106, the SLM layer 108, the lens layer 110, the SLM layer 112, the lens layer 114, and / or the light detector layer 128 in the z-direction in the semiconductor photonic device 102. Thus, the movable gantry 802 can enable selection and / or modification of the coupling between components of the photonic integrated circuit 104 Figure 2A and Figure 2B described focal lengths fl to f4. For example, the movable gantry 802 can be used to enable the SLM structure 120 and the lens structure 122 of the photonic integrated circuit 104 to be spaced apart in the z-direction by the focal length fl, and / or to tune or modify the focal length fl. The movable gantry 802 can enable the SLM layer 108 to be moved relative to the lens layer 110 and / or the lens layer 110 to be moved relative to the SLM layer 108 to select a particular focal length fl. As another example, the movable gantry 802 can be used to enable the SLM structure 124 and the lens structure 122 of the photonic integrated circuit 104 to be spaced apart in the z-direction by the focal length f2, and / or to tune or modify the focal length f2. The movable gantry 802 can enable the SLM layer 112 to be moved relative to the lens layer 110 and / or the lens layer 110 to be moved relative to the SLM layer 112 to select a particular focal length f2.

[0097] As another example, the movable gantry 802 can be used to enable the SLM structure 120 and the lens structure 126 of the photonic integrated circuit 104 to be spaced apart in the z-direction by a focal length f3, and / or to tune or modify the focal length f3. The movable gantry 802 can enable the SLM layer 112 to be moved relative to the lens layer 114 and / or the lens layer 114 to be moved relative to the SLM layer 112 to select a particular focal length f3. As another example, the movable gantry 802 can be used to enable the photodetector structure 128 and the lens structure 124 of the photonic integrated circuit 104 to be spaced apart in the z-direction by a focal length f4, and / or to tune or modify the focal length f4. The movable gantry 802 can enable the photodetector layer 116 to be moved relative to the lens layer 114 and / or the lens layer 114 to be moved relative to the photodetector layer 116 to select a particular focal length f4.

[0098] In some implementations, the light source layer 106, the SLM layer 108, the lens layer 110, the SLM layer 112, the lens layer 114, and / or the photodetector layer 128 are supported at opposite ends by the movable gantry 802. In some implementations, the light source layer 106, the SLM layer 108, the lens layer 110, the SLM layer 112, the lens layer 114, and / or the photodetector layer 128 can be supported at the bottom of the light source layer 106, the SLM layer 108, the lens layer 110, the SLM layer 112, the lens layer 114, and / or the photodetector layer 128 by the movable gantry 802.

[0099] As indicated above, providing Figure 8 As an example. Other examples can differ from what is described Figure 8 as described with respect to

[0100] Figure 9 is a flow diagram of an example process 900 associated with forming a photonic integrated circuit of a semiconductor photonic device described herein. In some implementations, Figure 9 One or more process blocks of

[0101] As shown in Figure 9 process 900 can include providing a first SLM structure of an artificial neural network photonic integrated circuit of a semiconductor photonic device (block 910). For example, one or more semiconductor processing tools can be used to provide an SLM structure 120 of an artificial neural network photonic integrated circuit (e.g., photonic integrated circuit 104) of a semiconductor photonic device 102 as described herein. In some implementations, the SLM structure 120 can be placed over a substrate.

[0102] AsFigure 9 Further to the foregoing, process 900 can include providing a first lens structure of the artificial neural network photonic integrated circuit over the first SLM structure (block 920). For example, one or more semiconductor processing tools can be used to provide the lens structure 122 of the artificial neural network photonic integrated circuit over the SLM structure 120, as described herein.

[0103] As Figure 9 Further to the foregoing, process 900 can include providing a second SLM structure of the artificial neural network photonic integrated circuit over the first lens structure (block 930). For example, one or more semiconductor processing tools can be used to provide the SLM structure 124 of the artificial neural network photonic integrated circuit over the lens structure 122, as described herein. In some implementations, the lens structure 122 is interposed between the SLM structure 120 and the SLM structure 124.

[0104] As Figure 9 Further to the foregoing, process 900 can include providing a second lens structure of the artificial neural network photonic integrated circuit over the substrate (block 940). For example, one or more semiconductor processing tools can be used to provide the lens structure 126 of the artificial neural network photonic integrated circuit over the SLM structure 124, as described herein.

[0105] Process 900 can include additional implementations, such as those described below and / or any individual implementation or any combination of implementations described in connection with one or more other processes described elsewhere herein.

[0106] In a first implementation, the first SLM structure 120, the first lens structure 122, the second SLM structure 124, and the second lens structure 126 are vertically arranged in the semiconductor photonic device 102.

[0107] In a second implementation, alone or in combination with the first implementation, providing the first lens structure 122 includes providing the first lens structure 122 on an optically transparent substrate of the lens layer 110.

[0108] In a third implementation, alone or in combination with one or more of the first or second implementations, providing the first SLM structure 120 includes providing a first plurality of SLM structures 120, where the first plurality of SLM structures includes on separate optically transparent substrates.

[0109] In a fourth implementation, alone or in combination with one or more of the first through third implementations, providing the first plurality of SLM structures 120 includes providing only phase SLM structure components 120a and providing amplitude SLM structure components 120b.

[0110] In a fifth implementation, alone or in combination with one or more of the first through fourth implementations, process 900 includes providing one or more coherent light source structures (e.g., light source structures 118, 120a, 120b) such that the one or more coherent light source structures are adjacent to the first SLM structure 120, and providing one or more light detector structures (e.g., light detector structures 128, 130a, 130b) such that the one or more light detector structures are adjacent to the second lens structure 126.

[0111] Although Figure 9 Example blocks of process 900 are shown, but in some implementations, process 900 includes additional blocks, fewer blocks, different blocks, or differently arranged blocks than those depicted in FIG. 10. Additionally, or alternatively, two or more of the blocks of process 900 can be performed in parallel. Figure 10 Example blocks of process 900 are shown, but in some implementations, process 900 includes additional blocks, fewer blocks, different blocks, or differently arranged blocks than those depicted in FIG. 10. Additionally, or alternatively, two or more of the blocks of process 900 can be performed in parallel.

[0112] Figure 10 is a flow diagram of an example process 1000 associated with forming a semiconductor photonic device described herein. In some implementations, Figure 10 One or more process blocks of the processes described in FIG. 10 are performed using one or more semiconductor processing tools, such as a deposition tool, an exposure tool, a developer tool, an etch tool, an electroplating tool, a planarization tool, an ion implantation tool, a pick-and-place tool, a wafer / die transport tool, and / or another type of semiconductor processing tool.

[0113] As further shown in Figure 10 In some implementations, the light source layer 106 can be disposed over a substrate. For example, one or more semiconductor processing tools can be used to provide a light source layer 106 of a semiconductor photonic device 102 as described herein. In some implementations, the light source layer 106 can be disposed over a substrate.

[0114] As further shown in Figure 10 In some implementations, a spacer layer 702 is disposed on the light source layer 106, and the SLM layer 108 is disposed on the spacer layer 702. For example, one or more semiconductor processing tools can be used to provide a SLM layer 108 of a semiconductor photonic device 102 over a light source layer 106 as described herein. In some implementations, a spacer layer 702 is disposed on the light source layer 106, and the SLM layer 108 is disposed on the spacer layer 702.

[0115] As further shown in Figure 10In further implementations, process 1000 can include providing a first lens layer of the semiconductor photonic device over the first SLM layer (block 1030). For example, one or more semiconductor processing tools can be used to provide a lens layer 110 of semiconductor photonic device 102 over SLM layer 108, as described herein. In some implementations, a spacer layer 704 is disposed on SLM layer 108, and lens layer 110 is disposed on spacer layer 704.

[0116] As Figure 10 In further implementations, process 1000 can include providing a second SLM layer of the semiconductor photonic device over the first lens layer (block 1040). For example, one or more semiconductor processing tools can be used to provide a SLM layer 112 of semiconductor photonic device 102 over lens layer 110, as described herein. In some implementations, a spacer layer 706 is disposed on lens layer 110, and SLM layer 112 is disposed on spacer layer 706.

[0117] As Figure 10 In further implementations, process 1000 can include providing a second lens layer of the semiconductor photonic device over the second SLM layer (block 1050). For example, one or more semiconductor processing tools can be used to provide a lens layer 114 of semiconductor photonic device 102 over SLM layer 112, as described herein. In some implementations, a spacer layer 708 is disposed on SLM layer 112, and lens layer 114 is disposed on spacer layer 708.

[0118] As Figure 10 In further implementations, process 1000 can include providing a photodetector layer of the semiconductor photonic device over the second lens layer (block 1060). For example, one or more semiconductor processing tools can be used to provide a photodetector layer 116 of semiconductor photonic device 102 over lens layer 114, as described herein. In some implementations, a spacer layer 710 is disposed on lens layer 114, and photodetector layer 116 is disposed on spacer layer 710.

[0119] Process 1000 can include additional implementations, such as those described below and / or any individual implementation or any combination of implementations described in conjunction with one or more other processes described elsewhere herein.

[0120] In a first implementation, SLM layer 108, lens layer 110, SLM layer 112, and lens layer 114 are arranged vertically in semiconductor photonic device 102.

[0121] In a second implementation, alone or in combination with the first implementation, providing SLM layer 108 includes providing a first SLM layer 108a and providing a second SLM layer 108b over the first SLM layer 108a.

[0122] In a third implementation, alone or in combination with one or more of the first or second implementations, providing the SLM layer 112 includes providing a first SLM layer 112a and providing a second SLM layer 112b on the first SLM layer 112a.

[0123] Although Figure 10 Examples of blocks of the process 1000 are shown, but in some implementations, the process 1000 includes additional blocks, fewer blocks, different blocks, or differently arranged blocks than those depicted in FIG. 10. Additionally, or alternatively, two or more of the blocks of the process 1000 can be performed in parallel. Legend Examples of blocks of the process 1000 are shown, but in some implementations, the process 1000 includes additional blocks, fewer blocks, different blocks, or differently arranged blocks than those depicted in FIG. 10. Additionally, or alternatively, two or more of the blocks of the process 1000 can be performed in parallel.

[0124] As such, the semiconductor photonic devices described herein include three-dimensional photonic integrated circuits that include optical components configured to implement artificial neural networks, such as CNNs or portions thereof. For example, the semiconductor photonic devices described herein can include three-dimensional photonic integrated circuits that include optical lens structures and SLM structures configured to perform sub-operations of convolution operations, including Fourier transform operations, multiplication operations, and inverse Fourier transform operations. The three-dimensional photonic integrated circuits described herein are capable of implementing encoded optical signals using photons that can be passively transformed for performing complex artificial intelligence tasks. As such, the photonic integrated circuits described herein consume significantly less power and can be less complex electronic integrated circuits, enabling further scaling of artificial neural networks.

[0125] As described in greater detail above, some implementations described herein provide a semiconductor photonic device. The semiconductor photonic device includes a photonic integrated circuit including a first SLM structure, a first lens structure, a second SLM structure, and a second lens structure. The first lens structure is between the first SLM structure and the second SLM structure, and the second SLM structure is between the first lens structure and the second lens structure.

[0126] As described in greater detail above, some implementations described herein provide a semiconductor photonic device. The semiconductor photonic device includes a photonic integrated circuit including a first SLM structure, a first lens structure, a second SLM structure, and a second lens structure. The first lens structure is between the first SLM structure and the second SLM structure, and the second SLM structure is between the first lens structure and the second lens structure.

[0127] As described in greater detail above, some embodiments described herein provide a method. The method includes providing a first SLM structure of an artificial neural network photonic integrated circuit of a semiconductor photonic device. The method includes providing a first lens structure of the artificial neural network photonic integrated circuit over the first SLM structure. The method includes providing a second SLM structure of the artificial neural network photonic integrated circuit over the first lens structure, wherein the first lens structure is between the first SLM structure and the second SLM structure. The method includes providing a second lens structure of the artificial neural network photonic integrated circuit over the second SLM structure, wherein the second SLM structure is between the first lens structure and the second lens structure.

[0128] The terms “approximately” and “substantially” can indicate that a value of a given quantity varies by 5% of the value (e.g., ±1%, ±2%, ±3%, ±4%, ±5% of the value). These values are merely examples and are not intended to be limiting. It should be understood that, in accordance with the present disclosure, the terms “approximately” and “substantially” can refer to a percentage of a value of a given quantity.

[0129] The foregoing summary of features of several embodiments enables a person skilled in the art to better understand aspects of the present disclosure. The person skilled in the art will appreciate that they can readily use the present disclosure as a basis for designing or modifying other processes and structures for performing the same purposes and / or achieving the same advantages introduced herein. The person skilled in the art will also recognize that these equivalent constructions do not depart from the spirit and scope of the present disclosure and that they can be varied in various ways, replaced, and modified herein without departing from the spirit and scope of the present disclosure.

[0130]

[0131] 100: example

[0132] 102: semiconductor photonic device

[0133] 104: photonic integrated circuit

[0134] 106: light source layer

[0135] 108: spatial light modulator (SLM) layer

[0136] 108a: first SLM layer

[0137] 108b: second SLM layer

[0138] 110: lens layer

[0139] 112: SLM layer

[0140] 112a: first SLM layer

[0141] 112b: second SLM layer

[0142] 114: lens layer

[0143] 116: photodetector layer

[0144] 118: light source structure

[0145] 118a: light source structure

[0146] 120: SLM structure

[0147] 120a: SLM structure component

[0148] 120b: amplitude SLM structure component

[0149] 122: lens structure

[0150] 124: SLM structure

[0151] 124a: SLM structure component

[0152] 124b: SLM structure component

[0153] 126: lens structure

[0154] 128: photodetector structure

[0155] 128a: photodetector structure

[0156] 128b: photodetector structure

[0157] 200: example

[0158] 202: focal plane

[0159] 204: focal plane

[0160] 206: focal plane

[0161] 208: focal plane

[0162] 210a: optical input signal

[0163] 210b: optical input signal

[0164] 212a to 212c: modulated optical input signal

[0165] 214a to 214c: optically filtered signal

[0166] 216a to 216c: transformed optical signal

[0167] 218a to 218c: transformed optically filtered signal

[0168] 220a-c: modulated optical output signals

[0169] 222a-c: modulated optical output signals

[0170] 224a-c: transformed optical output signals

[0171] 226a-c: transformed optical output signals

[0172] 300: example

[0173] 400: example

[0174] 500: example

[0175] 600: example

[0176] 700: example

[0177] 702: spacer layer

[0178] 704: spacer layer

[0179] 706: spacer layer

[0180] 708: spacer layer

[0181] 710: spacer layer

[0182] 712: example

[0183] 714: spacer insert

[0184] 716: spacer insert

[0185] 718: spacer insert

[0186] 720: spacer insert

[0187] 722: spacer insert

[0188] 800: example

[0189] 802: movable gantry

[0190] 900: process

[0191] 910: block

[0192] 920: block

[0193] 930: block

[0194] 940: block

[0195] 1000: process

[0196] 1010: block

[0197] 1020: block

[0198] 1030: block

[0199] 1040: block

[0200] 1050: block

[0201] 1060: block

[0202] f1 to f4: focal length.

Claims

1. A semiconductor photonic device, characterized by It comprises: a first spatial light modulator (SLM) structure; a first lens structure; a second SLM structure, wherein the first lens structure is between the first SLM structure and the second SLM structure; and a second lens structure, wherein the second SLM structure is between the first lens structure and the second lens structure. The first SLM structure comprises a transmissive SLM structure.

2. The semiconductor photonic device of claim 1, wherein The second SLM structure comprises another transmissive SLM structure.

3. The semiconductor photonic device of claim 2, wherein The first SLM structure comprises a plurality of SLM structure components; and 4. The semiconductor photonic device of claim 1, wherein wherein the plurality of SLM structure components comprises: a phase-only SLM structure component; and an amplitude-modulating SLM structure component. The second SLM structure comprises another plurality of SLM structure components; and 5. The semiconductor photonic device of claim 4, wherein wherein the another plurality of SLM structure components comprises: another phase-only SLM structure component; and another amplitude-modulating SLM structure component. The first SLM structure is positioned at a front focal plane of the first lens structure; and 6. The semiconductor photonic device of claim 1, wherein wherein the second SLM structure is positioned at a back focal plane of the first lens structure. The second SLM structure is positioned at a front focal plane of the second lens structure.

7. The semiconductor photonic device of claim 6, wherein It comprises:

8. A semiconductor photonic device, characterized by a first spatial light modulator (SLM) layer comprising one or more first SLM structures; a first lens layer comprising one or more first lens structures; a second SLM layer comprising one or more second SLM structures, wherein the first lens layer is between the first SLM layer and the second SLM layer; and a second lens layer comprising one or more second lens structures, wherein the second SLM layer is between the first lens layer and the second lens layer. It further comprises:

9. The semiconductor photonic device of claim 8, wherein a light source layer comprising a plurality of light source structures. The first SLM layer is between the light source layer and the first lens layer.

10. The semiconductor photonic device of claim 9, wherein ​