Open Access
3 October 2024 Diffraction casting
Author Affiliations +
Abstract

Optical computing is considered a promising solution for the growing demand for parallel computing in various cutting-edge fields that require high integration and high-speed computational capacity. We propose an optical computation architecture called diffraction casting (DC) for flexible and scalable parallel logic operations. In DC, a diffractive neural network is designed for single instruction, multiple data (SIMD) operations. This approach allows for the alteration of logic operations simply by changing the illumination patterns. Furthermore, it eliminates the need for encoding and decoding of the input and output, respectively, by introducing a buffer around the input area, facilitating end-to-end all-optical computing. We numerically demonstrate DC by performing all 16 logic operations on two arbitrary 256-bit parallel binary inputs. Additionally, we showcase several distinctive attributes inherent in DC, such as the benefit of cohesively designing the diffractive elements for SIMD logic operations that assure high scalability and high integration capability. Our study offers a design architecture for optical computers and paves the way for a next-generation optical computing paradigm.

1.

Introduction

Optical computing is a longstanding and captivating topic in the fields of optics and photonics. It is considered a potential post-Moore computing technology1 that offers distinct advantages, including high bandwidth, rapid processing speed, low power consumption, and parallelism.2,3 Around the 1980s, optical computing was actively explored, with developments in technologies, such as optical vector matrix multipliers47 and optical associative memories.810 Among these, shadow casting (SC) emerged as a prominent optical computing technology of that era.1114 SC facilitated single instruction, multiple data (SIMD) for logical operations through optical and spatially parallel computing. The SC scheme relied on shadowgrams, which optically generated a single output image through massively parallel logic operations from two binary input images. The versatility of SIMD logic operations was attained by altering the illumination pattern of the shadowgrams. Another key aspect involved the computational encoding and decoding of input and output images, respectively, designed to balance light intensities between the zeros and ones in the binary images. This computational process was an obstacle in achieving end-to-end optical computing. Despite the anticipated benefits in speed and energy efficiency, these optical computing technologies in the 1980s stagnated due to limitations in hardware (fabrication) and software (design) for optical components at that time. As a result, they lagged behind the major progress made in electronic computing.

Over the past few decades, significant advancements in microfabrication, mathematical optimization, and computational power have dramatically transformed the field of optical computing from what it was in the 1980s. Several pioneering optical computing techniques have been studied, including waveguide-based photonic circuits and diffractive neural networks (DNNs). Waveguide-based photonic circuits, which integrate waveguide interferometers, have high compatibility with currently existing electronic computers and circuits. They have led to a wide range of applications, such as vector-matrix operations,1517 logical operations,1824 and integrated reconfigurable circuits.2527

DNNs consist of cascaded diffractive optical elements (DOEs), which emulate neural network connections as light waves pass through the DOEs. This configuration utilizes the spatial parallelism of light and realizes fast and energy-efficient computation. A wide range of attractive applications based on DNNs has been proposed, including image classification,2833 image processing,3437 linear transformations,3840 and logic operations.4147

Currently, the demand for computation in SIMD logic operations has intensified, particularly due to advancements in cutting-edge technologies such as image processing, machine learning, and blockchain.4850 Traditional computation with central processing units is often inadequate to meet the computational needs of these fields. Consequently, graphics processing units,51 tensor processing units,52 field-programmable gate arrays (FPGAs),53,54 and application-specific integrated circuits55,56 are employed as SIMD-specific devices. The trend towards high-speed, energy-efficient, and massively parallel computing aligns well with the advantages of optical computation. As a result, there has been a rapid increase in efforts to develop practical optical computing methods for SIMD logic operations. Optical SIMD logic operations have been achieved using waveguide photonic circuits.26,27 However, a drawback of this approach is its limited scalability, which arises from the need for precise yet large-scale fabrication.

The use of DNNs holds potential as a solution to this issue, owing to the parallelism inherent in free-space propagation. For instance, several types of DNN-based logic operations have been proposed to overcome this drawback,4147 but these previous methods typically involve only one or a few types of logic operations on a small number of bits, and the realization of SIMD logic operations using DNNs remains unachieved. Moreover, these methods still require computational encoding and decoding of the input and output, respectively, posing a significant challenge for end-to-end optical computing, similar to the SC scheme.

In this study, considering the background mentioned above, we present a method termed diffraction casting (DC) for conducting all 16 optical SIMD logic operations on more than one hundred bits by incorporating the SC scheme and DNNs. DC revives SC through the use of DNNs. Therefore, DC shares the motivation of SC but exhibits several differences and advantages over SC. Unlike SC, which is based on geometrical optics, DC is grounded in wave optics. As a result, DC incorporates wave phenomena such as diffraction and interference through the use of DOE cascades in DNNs and is anticipated to offer greater integration capability compared to SC. Another advantage of DC is its elimination of the need for computational encoding and decoding of the input and output, respectively, which have been inherent bottlenecks in the SC scheme and previous DNN-based logic operations referred to above. This is enabled by introducing a buffer area around the input pair. In the rest of the paper, we will elaborate on the architectural design of DC, including the forward model, the optimization process, and provide numerical demonstrations.

2.

Materials and Methods

2.1.

Concept of Diffraction Casting

DC is designed for 16 types of SIMD logic operations, processing two input binary images to produce one output binary image. Figure 1 depicts the conceptual architecture of DC. DC consists of a reconfigurable illumination, DOEs, and an input layer. The reconfigurable illumination enables switching between logical operations and casts light on the DOE cascade forming a DNN. In this paper, we focus on the reconfigurable illumination with binary amplitude modulation implemented using a digital micromirror device (DMD) illuminated with coherent light from a laser, and on DOEs with phase modulation, assuming the use of commercially available optical components.

Fig. 1

Schematic diagram of DC. The selection of a logic operation is performed using reconfigurable illumination without any modification to the DOEs.

AP_6_5_056005_f001.png

We place the two input images side by side on the input layer within the DOE cascade to achieve a simple optical setup. The output of the logic operation appears as an intensity distribution at the end of the cascade and is captured with an image sensor. The final result is binarized by assuming a one-bit image sensor or a computational process. The reconfigurable illumination and DOEs are specifically trained to perform the 16 SIMD logic operations on any two binary images, as detailed in the subsequent subsection. Once the training process is completed, DC enables massively parallel optical logic operations on arbitrary binary inputs just by selecting the illumination patterns, without necessitating any modifications to the DOEs.

2.2.

Optical Forward Model

Figure 2 illustrates the forward and backward processes of DC. We consider a total of L types of SIMD logic operations on N parallel bits. These operations are conducted by an optical cascade composed of K layers, including one illumination layer, one input layer, and K2 DOE layers, with the layer index denoted as k{1,2,,K}. The first layer of the optical cascade is the binary reconfigurable illumination rl{0,1}Px×Py, where l{1,2,,L} is the index of the logic operations. An input pair f{0,1}Nx×2Ny, composed of side-by-side binary images, is located on the input layer, denoted as the Kin’th layer in the cascade.

Fig. 2

Forward and backward processes of DC. The reconfigurable illumination, the DOEs, and the scaling factor are optimized through the training process.

AP_6_5_056005_f002.png

Here, Nx and Ny represent the pixel counts of the individual images within the input pair along the x and y directions, respectively, where Nx×Ny=N. The phase distributions for each DOE layer are denoted by ϕkRPx×Py, and Px and Py indicate the pixel counts of the DOEs along the x and y axes, respectively. The result of the logic operations is observed with the image sensor located downstream of the K’th layer in the optical cascade.

We now describe the forward process of DC. The complex amplitude modulation vkCPx×Py, induced by the reconfigurable illumination, input pair, and the phase-only DOEs at the k’th layer in the cascade, is expressed as

Eq. (1)

vk={rlfork=1,I[f]+tfor  k=Kin,exp(jϕk)otherwise,
where j denotes the imaginary unit. Here, I is an operator transforming the input pair into an amplitude image on the input layer, composed of the following two steps. The first step is upsampling along the x and y directions with factors of sxN and syN, respectively. The second step is zero padding to enlarge the upsampled input pair to the DOE size (Px×Py pixels). t{0,1}Px×Py expresses a buffer surrounding the upsampled input pair, as illustrated in Fig. 2 and is defined as follows:

Eq. (2)

t(ux,uy)={0for{ux,uy}(PxsxNx2,Px+sxNx2]×(Py2syNy2,Py+2syNy2],1otherwise,
where uxN and uyN are indices along the x and y directions. This buffer is employed to compensate for light intensities transmitted or blocked on the input pair and enables the removal of computational encoding and decoding of the input and output, processes that are indispensably employed in previous optical logic operation methods, including the SC scheme.

The propagation process passing through the k’th layer in the cascade is written as

Eq. (3)

wk+1=Dk[vkwk],
where wkCPx×Py is a complex amplitude field just before the k’th layer. Dk is a diffraction operator representing the propagation from the k’th layer to the (k+1)’th layer, calculated based on the angular spectrum method.57 The initial field w1 is specified as an all-ones matrix, indicating a uniform field at the start.

The output intensity field of the optical cascade is observed with the image sensor as follows:

Eq. (4)

h=aO[|wK+1|2],

Eq. (5)

g=B[h].
Here, hRPx×Py represents the intermediate field, obtained through an operator O that first crops the central sxNx×syNy pixels from the output intensity field and then downsamples it to the original input image size of Nx×Ny. This process includes scaling the intensity by a factor of aR>0, which corresponds to either amplifying or attenuating the signal. gRNx×Ny is the final result of the logic operation, with the binarization operator B defined as

Eq. (6)

B[b]={0forb<0.5,1forb0.5,
where bR is an arbitrary variable. The binarization process, which converts analog signals to Boolean ones, is implemented using either a one-bit image sensor or through computational means. The variation of the threshold in the binarization is not crucial to the performance of DC. This is because the scaling factor a in Eq. (4) is optimized for the threshold through the gradient descent process, as described in the next section, and thus neutralizes the impact of the threshold.

2.3.

Optimization Process

To realize optical logic operations in parallel, the illumination rl, the DOEs ϕk, and the scaling factor a are optimized based on gradient descent in this study. First, we describe the optimization process by assuming a single logic operation (L=1) and a single input pair for simplicity, where the illumination is defined as rL=1. Then, we extend the optimization process to arbitrary numbers of logic operations and input pairs.

2.3.1.

Derivatives for a single logic operation and a single input pair

We define a cost function e for a single logic operation and a single input pair based on the mean squared error (MSE) as follows:

Eq. (7)

e=1N|e|2,
where represents the summation of all the elements of a tensor on its right side. Here, the error e is defined by

Eq. (8)

e=hg^,
which represents the difference between the intermediate field h and the ground truth of the operation result g^. This is to avoid the intermediate field’s signals around the threshold values in B, ensuring a robust binarization process.

To optimize rL=1 and ϕk based on gradient descent, the partial derivatives of e with respect to these variables are expressed by employing the chain rule as follows:

Eq. (9)

erL=1=v1rL=1·ev1,

Eq. (10)

eϕk=vkϕk·evk.

The right sides of these partial derivatives include the partial derivative of e with respect to vk, calculated as

Eq. (11)

evk=4aNwk*Dk1[vk+1*Dk+11[[vK*DK1[wK+1O1[e]]]]],
where Dk1 and O1 are operators representing the inverse processes of Dk and O, respectively, and the superscript * denotes the complex conjugate.

The partial derivatives with respect to each optimized variable are finally written as follows. The partial derivative with respect to rL=1 is described as

Eq. (12)

erL=1=Re[ev1],
where Re[·] denotes the real part of a complex amplitude. The partial derivative with respect to ϕk is described as

Eq. (13)

eϕk=Re[jvk*evk].
The partial derivative with respect to a is described as

Eq. (14)

ea=2aNhe.

2.3.2.

Derivatives for multiple logic operations and multiple input pairs

Next, we extend the optimization process from a single logic operation and a single input pair as described above, to L logic operations and M input pairs. In this scenario, the cost function E based on the MSE is expressed as

Eq. (15)

E=1LMl,mel,m,
where el,m denotes the cost associated with the l’th logic operation and the m’th input pair, derived from Eq. (7). The partial derivatives of E with respect to rl, ϕk, and a are presented as summations of the partial derivatives of el,m with respect to these variables, derived from Eqs. (12)–(14), respectively,

Eq. (16)

Erl=mel,mrl,

Eq. (17)

Eϕk=l,mel,mϕk,

Eq. (18)

Ea=l,mel,ma.

2.3.3.

Updating procedure

The variables rl, ϕk, and a are updated with the partial derivatives in Eqs. (16)–(18) based on the Adam optimizer.58 The updating processes for the DOEs ϕk and the scaling factor a are described as follows:

Eq. (19)

ϕkϕkAdam[Eϕk],

Eq. (20)

aaAdam[Ea],
where Adam[·] represents an operator to calculate the updating step in the Adam optimizer with the derivatives. To simplify the physical realization of the illumination rl, we assume its binary implementations, such as DMDs, by introducing stochastic perturbations into the gradient descent process.59 The first step in the update process for the variables is as follows:

Eq. (21)

rl˜C[rl˜Adam[Erl]],
where r˜l is an intermediate variable for the backward process in the optimization of rl. Here, C is an operator for clipping the range of values as follows:

Eq. (22)

C[b]={0forb<0,1forb>1,botherwise.
Subsequently, rl in the forward process is updated as follows:

Eq. (23)

rl=B[r˜l+q],
where qRPx×Py is a uniform distribution between ±0.5, introduced to avoid local minima in the binary optimization. After the optimization process, rl is finalized as follows:

Eq. (24)

rl=B[r˜l].

3.

Numerical Demonstration

3.1.

Experimental Conditions

We numerically demonstrated DC with all 16 logic operations (L=16) for the Boolean input pair fleft and fright, as shown in Table 1. In this numerical demonstration, the wavelength of the coherent light for the reconfigurable illumination λ was defined as 0.532  μm. The optical cascade comprised eleven layers (K=11), incorporating nine DOEs, with the input layer positioned as the sixth layer (Kin=6). The intervals between the layers were equally set to 3×104λ(1.60×104  μm). For the illumination pattern rl, the DOEs ϕk, and the input layer, the pixel pitch was 16λ(8.51  μm). This pixel pitch was chosen by considering commercially available spatial light modulators (SLMs), including DMDs, as well as the microfabrication technique used for configuring DOEs,60 and the simplicity of using an integer product.44,45 The pixel count was 160(=Px) along the x axis and 288(=Py) along the y axis, respectively. For the input pair fm, the pixel count of the individual image in the pair was 16(=Nx) along the x axis and 16(=Ny) along the y axis, where the parallel bits N became 256, and the upsampling factors sx and sy were both 8. The width of the region with ones on the buffer t, as defined in Eq. (2), was set to 16 pixels. To prevent the circulant effect on the diffraction calculation, the complex amplitude fields were zero-padded with a width of 64 pixels during the layer-by-layer propagation processes.

Table 1

Logic operations defined on input pair.

Input pair fleftfrightOperation index lLogic operationBoolean 0 0 1 10 1 0 1
Output
g^l
100 0 0 0
2fleftfright (AND)0 0 0 1
3fleftfright¯0 0 1 0
4fleft0 0 1 1
5fleft¯fright0 1 0 0
6fright0 1 0 1
7fleftfright (XOR)0 1 1 0
8fleftfright (OR)0 1 1 1
911 1 1 1
10fleftfright¯ (NAND)1 1 1 0
11fleft¯fright1 1 0 1
12fleft¯1 1 0 0
13fleftfright¯1 0 1 1
14fright¯1 0 1 0
15fleftfright¯ (XNOR)1 0 0 1
16fleftfright¯ (NOR)1 0 0 0

For the optimization process, the input pairs were generated with values initially selected from uniform random distributions between 0 and 1 and were then binarized using randomly selected thresholds, also between 0 and 1. The numbers of input pairs for training and testing were 80,000 and 256, respectively, without any duplication. The batch size M was set to 16 for training. The number of iterations was 5000. The learning rates for the Adam optimizer, as used in Eqs. (19)–(21), for rl, ϕk, and a were set to 3×102, 1×102, and 3×103, respectively. These variables were initially set to uniform random distributions for rl and ϕk, and 10 for a. The final performance of DC with L logic operations was evaluated by the root mean squared errors (RMSEs) between the final result gl,m and the ground truth g^l,m for M test input pairs, as shown in Fig. 2 and described as follows:

Eq. (25)

RMSE=1LMNl,m|gl,mg^l,m|2.

3.2.

Result

The optimization results for the illumination rl and the DOEs ϕk are presented in Figs. 3(a) and 3(b), respectively. The scaling factor a was optimized to 23.8. In the numerical demonstration shown in Fig. 4, DC was performed using two test input pairs with 256 parallel bits. The first pair, shown in Fig. 4(a), was composed of random patterns. The second pair, shown in Fig. 4(d), was composed of characteristic patterns. Ground truths of their 16 SIMD logic operations are displayed in Figs. 4(b) and 4(e), respectively. The operation outputs are shown in Figs. 4(c) and 4(f), respectively, where all operations were successful, and their RMSEs were zero. Furthermore, the RMSEs for 256 test input pairs of random patterns were also found to be zero. These outcomes underscore the promising potential of DC. More detailed discussions are provided in the next section and the appendix.

Fig. 3

Optimization results. (a) Binary amplitude patterns on a DMD for the reconfigurable illumination and (b) phase distributions on the DOEs. Scale bar is 1 mm.

AP_6_5_056005_f003.png

Fig. 4

Examples of the DC process with the optimized illumination and the DOEs shown in Fig. 3. (a) Test input pair of random patterns and its corresponding (b) ground truths of the 16 operations, with the operation index l noted below each, and (c) their operation outputs, with the RMSE noted below each. (d) Test input pair of characteristic patterns and its corresponding (e) ground truths of the 16 operations, with the operation index l noted below each, and (f) their operation outputs, with the RMSE noted below each. Scale bars in (a) and (d) are 1 mm, indicating the physical scale after the upsampling process.

AP_6_5_056005_f004.png

4.

Analysis

We conducted numerical analyses of the performance of DC under various optical conditions. Throughout this analysis, the experimental conditions were consistent with those described in Sec. 3, except where otherwise noted. Further analyses are provided in Appendix A.

4.1.

Number of DOEs

The computational performance of DC, with the number of DOEs set to K2, was evaluated using the RMSEs, as illustrated in Fig. 5. In this evaluation, the number of input parallel bits N was set to 4, 16, 64, 128, and 256. Correspondingly, the upsampling factors along the x and y axes were adjusted to (64, 64), (32, 32), (16, 16), (8, 16), and (8, 8) [=(sx,sy)], respectively, aiming to maintain consistent pixel counts of 128 (=sxNx,syNy) on the input layer after upsampling. The layer index of the input layer Kin was set to (K2)/2+2. When the number of DOEs was zero, only the illumination pattern was optimized.

Fig. 5

Computational errors associated with the varying number of DOEs.

AP_6_5_056005_f005.png

As illustrated in Fig. 5, the calculation error decreased with an increase in the number of DOEs. Additionally, the necessary number of DOEs for achieving error-free calculation increased with the number of input parallel bits N, but at a rate less than proportional to N. This rate of increase was smaller than predicted in previous works,32,38 indicating an advantage of DC in terms of scalability and integration capability through the use of spatially parallelized optical processes for logic operations.

4.2.

Multiplexing Advantage

The DOEs implemented 16 logic operations in a multiplexed manner in DC, as shown in Table 1. We confirmed the computational errors under different numbers of DOEs, denoted as K2, when the optical cascade was designed for single logic operations. The layer index of the input layer, Kin, was set to (K2)/2+2. In Fig. 6, computational errors for AND, OR, NAND, and NOR operations, selected from the 16 operations, are shown. Results for all 16 logic operations are provided in Fig. 8. In most of these results, error-free or nearly error-free calculations for single logic operations were achieved when the number of DOEs was greater than 6. On the other hand, as shown in Fig. 5, the necessary number of DOEs for multiplexing 16 logic operations was 9, which is significantly less than 6×16, for error-free calculations. Furthermore, the computational errors for some logic operations, such as XNOR and NOR, were reduced by multiplexing all the logic operations, which may help prevent falling into local minima in the optimization process. These results verified the advantage of multiplexing logic operations, in terms of both architecture configuration and training process, as well as the integration capability of DC.

Fig. 6

Computational errors associated with the varying number of DOEs for single logic operations.

AP_6_5_056005_f006.png

5.

Conclusion

We revived SC as DC by employing DNNs to achieve scalable and flexible optical SIMD operations. The optical cascade of DC consisted of reconfigurable illumination, DOEs, and an input layer. The illumination patterns and DOEs were designed to perform 16 logic operations on any binary input image pair, and the output intensity of the optical cascade was binarized to produce the final results. In this study, we achieved 16 switchable logic operations on 256 bits, which is an outstanding achievement compared with previous studies.4147 In contrast to these methods, where the optical diffraction processes from each spatial region of logic operations and bits must be separately designed, our method allows both interference between spatial regions of bits and interference between logic operations, enabling these regions to be densely located based on end-to-end designs of the illumination patterns and DOEs. This advantage has enabled high scalability and integration capability with all-optical operation, eliminating the need for computational encoding and decoding, all of which were numerically demonstrated.

An issue with DC for practical applications is its low energy efficiency. This may be addressed by adopting illumination with phase modulation and optimizing physical conditions, including the layer interval, buffer, and image sensor. We are planning the physical implementation of DC based on the setups shown in Sec. 7. Owing to its flexible and reconfigurable architecture offered by a learning-based approach, DC can be extended to a versatile range of inputs and operations beyond SIMD logic operations, such as image processing, including image filtering, and advanced reconfigurable optical computing methods like optical FPGAs. Furthermore, incorporating multiplexing in various optical quantities, such as time,61,62 wavelength,63 polarization,39,64 and orbital angular momentum,42 would enhance the computational capacity in DC. Thus, our study on DC offers a novel design architecture for optical computers and optical accelerators and paves the way for a next-generation optical computing paradigm.

6.

Appendix A: Supplementary Analyses

This appendix provides detailed analyses of DC. Throughout this appendix, the experimental conditions are consistent with those described in Sec. 3, except where otherwise noted.

6.1.

Training Process

Figure 7 illustrates the trends of the cost function based on the MSE without the binarization process in Eq. (15) and the computational error based on the RMSE in Eq. (25) during the training process. At the end of the training process, the cost function converged to nearly zero but not exactly zero. On the other hand, the computational error converged to exactly zero around the final iteration step. This indicates that the binarization process rectified the optical outputs and eliminated their small errors.

Fig. 7

Error trends during the training process.

AP_6_5_056005_f007.png

The code for the numerical demonstrations in this study was implemented with MATLAB 2022a (MathWorks) and was executed on a computer with an AMD EPYC 7763 64-core processor at a clock rate of 2.45 GHz and an NVIDIA A100 SXM4 with 80 GB of memory. The computational time for the entire training process was ∼7 h.

6.2.

Multiplexing Advantage

In Sec. 4.2, the computational errors of DC designed for single-logic operations of AND, OR, NAND, and NOR were selectively presented. The errors for all 16 logic operations are shown in Fig. 8. In most logic operations, error-free or nearly error-free calculations were achieved when the number of DOEs was greater than 6. On the other hand, multiplexing all 16 logic operations achieved error-free calculation when the number of DOEs was 9, as shown in Fig. 5, which was much smaller than 6×16 and supported the advantage of multiplexing the logic operations.

Fig. 8

Computational errors associated with the varying number of DOEs for single logic operations. (a) 1l4, (b) 5l8, (c) 9l12, and (d) 13l16.

AP_6_5_056005_f008.png

6.3.

Physical Volume of DC

The physical volume of the optical cascade in Sec. 3.2 was calculated as 3.89×1012λ3(5.86×1011  μm3), where the pixel pitch on the DOEs was 16λ and the intervals between the layers in the optical cascade were 3×104λ, respectively. We investigated the computational error with respect to the reduction of the physical volume by scaling down both the pixel pitch and the interval with the same magnification ratio, varying the pixel pitch from 1/8λ to 16λ by powers of two. The result is shown in Fig. 9. In this case, the minimal physical volume without computational error was 5.94×107λ3(8.94×106  μm3), where the pixel pitch was λ(=0.532  μm), and the interval was 1.17×102λ(6.23×10  μm). This result indicated that the minimal physical volume of DC is constrained by the diffraction limit.

Fig. 9

Computational errors associated with the varying physical volume of DC.

AP_6_5_056005_f009.png

6.4.

Position of the Input Layer

The computational error was calculated by varying the position of the input layer Kin from 2 to 11 in the optical cascade with 11 layers, as depicted in Fig. 10. This result shows the importance of the DOEs downstream from the input layer in reducing computational error. It suggests that there is an advantage in positioning the input layer at an upper layer, excluding the top one.

Fig. 10

Computational errors under different positions of the input layer.

AP_6_5_056005_f010.png

6.5.

Energy Efficiency

We evaluated the light-energy efficiency of DC using the following definition:

Eq. (26)

Energy efficiency=[O[|wK+1|2]]|l=9,f=1PxPy.
Here, the denominator represents the total input energy to the optical cascade. The numerator is the total energy on the output area of interest, calculated when the logic operation is configured to produce one output (l=9) and all elements of the input pair f are set to 1. The energy efficiency was assessed with respect to the scaling factor a and the width of the buffer t.

6.5.1.

Scaling factor

The light-energy efficiency is associated with the scaling factor a, which amplifies or attenuates the signals captured by the image sensor before the binarization process. A larger a indicates lower energy efficiency and vice versa. In the above demonstrations and analyses, a was included in the optimized parameters, as shown in Eq. (20). Here, a was set to a specific value and was not updated during the optimization process. Once the optimizations of the illumination pattern rl and the DOEs ϕk were completed, the energy efficiency in Eq. (26) and the computational error were calculated. This process was repeated by changing a from 2 to 32. The relationship between the energy efficiency and the computational error is shown in Fig. 11. The RMSEs for the energy efficiencies between 1.81% and 8.71% were less than 2.38×103. Therefore, nearly error-free calculation was achieved within this range of energy efficiencies.

Fig. 11

Relationship between computational errors and energy efficiencies when varying the scaling factor.

AP_6_5_056005_f011.png

The primary sources of energy loss were amplitude modulation on the illumination plane, light leakage from the optical cascade, and the cropping of the limited square area by the image sensor at the end of the optical cascade. The first issue can be solved by employing phase modulation on the illumination plane, although its modulation speed is lower than that of amplitude modulation on currently available SLMs. The second issue may be alleviated by reducing the intervals between layers. The third issue can be addressed by increasing the sensor area, employing anisotropic sampling, or utilizing anamorphic imaging. Another approach to improve energy efficiency is to increase the width of the buffer, as indicated in the next section. The trade-off between energy efficiency and computational error will not be an issue in a proof-of-concept experimental demonstration with an optical setup, such as those shown in Sec. 7, by increasing the illumination intensity. However, this issue must be considered in practical applications, where energy efficiency in computation is a crucial factor.

6.5.2.

Buffer width

The buffer t was introduced into DC to compensate for the light intensities transmitted or blocked by the input pair. It was expected to eliminate the computational encoding and decoding processes employed in previous methods for optical logic operations, including the SC scheme. In the above demonstrations and analyses, the buffer width was set to 16 pixels. The plots in Fig. 12 show the energy efficiencies and computational errors at different buffer widths, including zero width. In this analysis, the scaling factor a was included in the optimization parameters. This result supported the necessity of the buffer for error-free calculation. Furthermore, a larger buffer width increased the energy efficiency.

Fig. 12

Relationship between computational errors and energy efficiencies with the varying buffer width (BW [pixels]).

AP_6_5_056005_f012.png

The RMSE for the logic operations at 1l8 without the buffer was reduced from 4.24×102 to 0 by adding a buffer of one pixel. On the other hand, the RMSE for the logic operations at 9l16 with no buffer was significantly improved from 4.40×101 to 0 by adding a one-pixel buffer. As shown in Table 1, operations at 1l8 do not include the operation with an input of fleft=0,fright=0 and an output of g^l=1. Conversely, the operations at 9l16 include such an operation. This result also verified the role of the buffer—compensating for the balance between the light intensities of the input and output in the optical cascade.

6.6.

Alignment Error

An issue in the physical demonstration of DC will be alignment errors. The system’s performance under alignment errors along the x and z axes is presented in Fig. 13, showing the computational error with one-dimensional alignment on the individual layers, including the illumination and input layers. The alignment error along the y axis was omitted in this analysis because its impact is considered similar to that along the x axis due to their symmetry. As shown in these results, the impact of the alignment error along the x axis was greater than that along the z axis.

Fig. 13

Relationships between computational errors and alignment errors along the x axis on the layers (a) between the illumination and the one before the input (1k5), and (b) between the input and the end (6k11); and those along the z axis on the layers (c) between the illumination and the one before the input (1k5), and (d) between the input and the end (6k11).

AP_6_5_056005_f013.png

Several methods for compensating for alignment errors in DNNs have been proposed, and these can be applied to configure our setup. The first approach is enhancing robustness against alignment errors or model errors by introducing them during the computational training process.65 The second approach is using a closed-loop process to feed back alignment errors or model mismatches to controllable optical elements, such as SLMs.66,67 The third approach is incorporating integrated chip fabrication techniques, which significantly reduce alignment errors or model mismatches.32,43

7.

Appendix B: Optical Setup

Two candidate experimental setups for DC are presented in Fig. 14. Both setups employ DMDs for the reconfigurable illumination and input due to the high speed and high contrast in DMD modulation. The first setup, shown in Fig. 14(a), uses transmissive phase modulation elements, such as DOEs. While it can be bulky, it is easier to align optical components. The polarization state of the light may optionally be controlled to prevent stray light from the DMDs. This type of setup has been demonstrated in DNNs for object classification and classical or quantum logic gates.28,31,41,46 The second setup, shown in Fig. 14(b), uses a mirror and reflective phase modulation elements, such as SLMs. This reflective setup may be more compact and compatible with the closed-loop approach described in Sec. 6.6 to compensate for alignment errors. The diagonal propagation process can be considered when designing the phase modulations using rotational transformation in numerical diffraction.68 This type of reflective setup has been demonstrated in DNNs for a beam mode converter, a quantum gate, and optical reservoir computing.6971

Fig. 14

Candidates for the experimental setups of DC using (a) transmissive phase modulation with DOEs and (b) reflective phase modulation with SLMs.

AP_6_5_056005_f014.png

Disclosures

The authors declare no conflicts of interest.

Code and Data Availability

Data may be obtained from the authors upon reasonable request.

Acknowledgments

We gratefully acknowledge the insightful and thoughtful leadership of the late Dr. Makoto Naruse and extend our deepest condolences. We also thank Mr. Yuya Hidaka at The University of Tokyo for fruitful discussions. This work was supported by Japan Society for the Promotion of Science (Grant Nos. JP20K05361, JP22H05197, and JP23K26567).

References

1. 

G. E. Moore, “Cramming more components onto integrated circuits,” Proc. IEEE, 86 (1), 82 –85 https://doi.org/10.1109/JPROC.1998.658762 (1998). Google Scholar

2. 

H. J. Caulfield and S. Dolev, “Why future supercomputing requires optics,” Nat. Photonics, 4 (5), 261 –263 https://doi.org/10.1038/nphoton.2010.94 (2010). Google Scholar

3. 

K. Kitayama et al., “Novel frontier of photonics for data processing—photonic accelerator,” APL Photonics, 4 (9), 090901 https://doi.org/10.1063/1.5108912 (2019). Google Scholar

4. 

R. P. Bocker, “Matrix multiplication using incoherent optical techniques,” Appl. Opt., 13 (7), 1670 –1676 https://doi.org/10.1364/AO.13.001670 (1974). Google Scholar

5. 

J. W. Goodman, A. Dias and L. Woody, “Fully parallel, high-speed incoherent optical method for performing discrete Fourier transforms,” Opt. Lett., 2 (1), 1 –3 https://doi.org/10.1364/OL.2.000001 (1978). Google Scholar

6. 

R. A. Athale and W. C. Collins, “Optical matrix–matrix multiplier based on outer product decomposition,” Appl. Opt., 21 (12), 2089 –2090 https://doi.org/10.1364/AO.21.002089 ((1982). Google Scholar

7. 

M. Gruber, J. Jahns and S. Sinzinger, “Planar-integrated optical vector-matrix multiplier,” Appl. Opt., 39 (29), 5367 –5373 https://doi.org/10.1364/AO.39.005367 (2000). Google Scholar

8. 

Y. Owechko et al., “Holographic associative memory with nonlinearities in the correlation domain,” Appl. Opt., 26 (10), 1900 –1910 https://doi.org/10.1364/AO.26.001900 (1987). Google Scholar

9. 

E. G. Paek and D. Psaltis, “Optical associative memory using Fourier transform holograms,” Opt. Eng., 26 (5), 428 –433 https://doi.org/10.1117/12.7974093 (1987). Google Scholar

10. 

M. Ishikawa et al., “Optical associatron: a simple model for optical associative memory,” Appl. Opt., 28 (2), 291 –301 https://doi.org/10.1364/AO.28.000291 (1989). Google Scholar

11. 

J. Tanida and Y. Ichioka, “Optical logic array processor using shadowgrams,” J. Opt. Soc. Am., 73 (6), 800 –809 https://doi.org/10.1364/JOSA.73.000800 (1983). Google Scholar

12. 

Y. Ichioka and J. Tanida, “Optical parallel logic gates using a shadow-casting system for optical digital computing,” Proc. IEEE, 72 (7), 787 –801 https://doi.org/10.1109/PROC.1984.12939 IEEPAD 0018-9219 (1984). Google Scholar

13. 

J. Tanida and Y. Ichioka, “OPALS: optical parallel array logic system,” Appl. Opt., 25 (10), 1565 –1570 https://doi.org/10.1364/AO.25.001565 (1986). Google Scholar

14. 

K.-H. Brenner, A. Huang and N. Streibl, “Digital optical computing with symbolic substitution,” Appl. Opt., 25 (18), 3054 –3060 https://doi.org/10.1364/AO.25.003054 (1986). Google Scholar

15. 

D. A. Miller, “Self-configuring universal linear optical component,” Photonics Res., 1 (1), 1 –15 https://doi.org/10.1364/PRJ.1.000001 (2013). Google Scholar

16. 

Y. Shen et al., “Deep learning with coherent nanophotonic circuits,” Nat. Photonics, 11 (7), 441 –446 https://doi.org/10.1038/nphoton.2017.93 (2017). Google Scholar

17. 

N. C. Harris et al., “Linear programmable nanophotonic processors,” Optica, 5 (12), 1623 –1631 https://doi.org/10.1364/OPTICA.5.001623 (2018). Google Scholar

18. 

M. Zhang, L. Wang and P. Ye, “All optical XOR logic gates: technologies and experiment demonstrations,” IEEE Commun. Mag., 43 (5), S19 –S24 https://doi.org/10.1109/MCOM.2005.1453421 (2005). Google Scholar

19. 

Y. ZhangY. Zhang and B. Li, “Optical switches and logic gates based on self-collimated beams in two-dimensional photonic crystals,” Opt. Express, 15 (15), 9287 –9292 https://doi.org/10.1364/OE.15.009287 (2007). Google Scholar

20. 

Y.-D. Wu, T.-T. Shih and M.-H. Chen, “New all-optical logic gates based on the local nonlinear Mach-Zehnder interferometer,” Opt. Express, 16 (1), 248 –257 https://doi.org/10.1364/OE.16.000248 (2008). Google Scholar

21. 

K.-S. Lee and S.-K. Kim, “Conceptual design of spin wave logic gates based on a Mach–Zehnder-type spin wave interferometer for universal logic functions,” J. Appl. Phys., 104 (5), 053909 https://doi.org/10.1063/1.2975235 (2008). Google Scholar

22. 

J. Dong, X. Zhang and D. Huang, “A proposal for two-input arbitrary Boolean logic gates using single semiconductor optical amplifier by picosecond pulse injection,” Opt. Express, 17 (10), 7725 –7730 https://doi.org/10.1364/OE.17.007725 (2009). Google Scholar

23. 

Y. Fu, X. Hu and Q. Gong, “Silicon photonic crystal all-optical logic gates,” Phys. Lett. A, 377 (3–4), 329 –333 https://doi.org/10.1016/j.physleta.2012.11.034 (2013). Google Scholar

24. 

D. G. Sankar Rao, S. Swarnakar and S. Kumar, “Performance analysis of all-optical NAND, NOR, and XNOR logic gates using photonic crystal waveguide for optical computing applications,” Opt. Eng., 59 (5), 057101 https://doi.org/10.1117/1.OE.59.5.057101 (2020). Google Scholar

25. 

Y. Xie et al., “Programmable optical processor chips: toward photonic RF filters with DSP-level flexibility and MHz-band selectivity,” Nanophotonics, 7 (2), 421 –454 https://doi.org/10.1515/nanoph-2017-0077 (2017). Google Scholar

26. 

W. Bogaerts et al., “Programmable photonic circuits,” Nature, 586 (7828), 207 –216 https://doi.org/10.1038/s41586-020-2764-0 (2020). Google Scholar

27. 

Z. Ying et al., “Electronic-photonic arithmetic logic unit for high-speed computing,” Nat. Commun., 11 (1), 2154 https://doi.org/10.1038/s41467-020-16057-3. (2020). Google Scholar

28. 

X. Lin et al., “All-optical machine learning using diffractive deep neural networks,” Science, 361 (6406), 1004 –1008 https://doi.org/10.1126/science.aat8084 SCIEAS 0036-8075 (2018). Google Scholar

29. 

T. Yan et al., “Fourier-space diffractive deep neural network,” Phys. Rev. Lett., 123 (1), 023901 https://doi.org/10.1103/PhysRevLett.123.023901 (2019). Google Scholar

30. 

J. Weng et al., “Meta-neural-network for real-time and passive deep-learning-based object recognition,” Nat. Commun., 11 (1), 6309 https://doi.org/10.1038/s41467-020-19693-x (2020). Google Scholar

31. 

H. Chen et al., “Diffractive deep neural networks at visible wavelengths,” Engineering, 7 (10), 1483 –1491 https://doi.org/10.1016/j.eng.2020.07.032 ENGNA2 0013-7782 (2021). Google Scholar

32. 

H. Zhu et al., “Space-efficient optical computing with an integrated chip diffractive neural network,” Nat. Commun., 13 (1), 1044 https://doi.org/10.1038/s41467-022-28702-0. (2022). Google Scholar

33. 

C. Qian et al., “Dynamic recognition and mirage using neuro-metamaterials,” Nat. Commun., 13 (1), 2694 https://doi.org/10.1038/s41467-022-30377-6 (2022). Google Scholar

34. 

Y. Luo et al., “Computational imaging without a computer: seeing through random diffusers at the speed of light,” elight, 2 (1), 4 https://doi.org/10.1186/s43593-022-00012-4 (2022). Google Scholar

35. 

Ç. Işıl et al., “Super-resolution image display using diffractive decoders,” Sci. Adv., 8 (48), eadd3433 https://doi.org/10.1126/sciadv.add3433 (2022). Google Scholar

36. 

M. Huang et al., “Diffraction neural network for multi-source information of arrival sensing,” Laser Photonics Rev., 17 (10), 2300202 https://doi.org/10.1126/sciadv.add3433 (2023). Google Scholar

37. 

T. Igarashi, M. Naruse and R. Horisaki, “Incoherent diffractive optical elements for extendable field-of-view imaging,” Opt. Express, 31 (19), 31369 –31382 https://doi.org/10.1364/OE.499866 (2023). Google Scholar

38. 

O. Kulce et al., “All-optical information-processing capacity of diffractive surfaces,” Light Sci. Appl., 10 (1), 25 https://doi.org/10.1038/s41377-020-00439-9 (2021). Google Scholar

39. 

J. Li et al., “Polarization multiplexed diffractive computing: all-optical implementation of a group of linear transformations through a polarization-encoded diffractive network,” Light Sci. Appl., 11 (1), 153 https://doi.org/10.1038/s41377-022-00849-x (2022). Google Scholar

40. 

J. Li et al., “Massively parallel universal linear transformations using a wavelength-multiplexed diffractive optical network,” Adv. Photonics, 5 (1), 016003 https://doi.org/10.1117/1.AP.5.1.016003 (2023). Google Scholar

41. 

C. Qian et al., “Performing optical logic operations by a diffractive neural network,” Light Sci. Appl., 9 (1), 59 https://doi.org/10.1038/s41377-020-0303-2. (2020). Google Scholar

42. 

P. Wang et al., “Orbital angular momentum mode logical operation using optical diffractive neural network,” Photonics Res., 9 (10), 2116 –2124 https://doi.org/10.1364/PRJ.432919 (2021). Google Scholar

43. 

S. Zarei and A. Khavasi, “Realization of optical logic gates using on-chip diffractive optical neural networks,” Sci. Rep., 12 (1), 15747 https://doi.org/10.1038/s41598-022-19973-0 (2022). Google Scholar

44. 

Y. Luo, D. Mengu and A. Ozcan, “Cascadable all-optical NAND gates using diffractive networks,” Sci. Rep., 12 (1), 7121 https://doi.org/10.1038/s41598-022-11331-4 (2022). Google Scholar

45. 

X. Liu et al., “Parallelized and cascadable optical logic operations by few-layer diffractive optical neural network,” Photonics, 10 (5), 503 https://doi.org/10.3390/photonics10050503 (2023). Google Scholar

46. 

X. Ding et al., “Metasurface-based optical logic operators driven by diffractive neural networks,” Adv. Mater., 36 (9), 2308993 https://doi.org/10.1002/adma.202308993 (2024). Google Scholar

47. 

X. Lin et al., “Polarization-based all-optical logic gates using diffractive neural networks,” J. Opt., 26 (3), 035701 https://doi.org/10.1088/2040-8986/ad2712 (2024). Google Scholar

48. 

M. Sonka, V. Hlavac and R. Boyle, Image Processing, Analysis and Machine Vision, Springer( (2013). Google Scholar

49. 

J. A. Dev, “Bitcoin mining acceleration and performance quantification,” in IEEE 27th Can. Conf. Electr. and Comput. Eng. (CCECE), 1 –6 (2014). https://doi.org/10.1109/CCECE.2014.6900989 Google Scholar

50. 

A. Reuther et al., “Survey and benchmarking of machine learning accelerators,” in High Perform. Extreme Comput. Conf. (HPEC), 1 –9 (2019). https://doi.org/10.1109/HPEC.2019.8916327 Google Scholar

51. 

J. D. Owens et al., “GPU computing,” Proc. IEEE, 96 (5), 879 –899 https://doi.org/10.1109/JPROC.2008.917757 (2008). Google Scholar

52. 

N. P. Jouppi et al., “In-datacenter performance analysis of a tensor processing unit,” in Proc. 44th Annu. Int. Symp. Comput. Architect., 1 –12 (2017). https://doi.org/10.1145/3079856.3080246 Google Scholar

53. 

J. Rose, A. El Gamal and A. Sangiovanni-Vincentelli, “Architecture of field-programmable gate arrays,” Proc. IEEE, 81 (7), 1013 –1029 https://doi.org/10.1109/5.231340 IEEPAD 0018-9219 (1993). Google Scholar

54. 

X. Zhang et al., “Machine learning on FPGAs to face the IoT revolution,” in IEEE/ACM Int. Conf. Comput.-Aided Design (ICCAD), 894 –901 (2017). https://doi.org/10.1109/ICCAD.2017.8203875 Google Scholar

55. 

K. R. Hari et al., “Cryptocurrency mining–transition to cloud,” Int. J. Adv. Comput. Sci. Appl., 6 (9), 115 –124 https://doi.org/10.14569/IJACSA.2015.060915 (2015). Google Scholar

56. 

E. Nurvitadhi et al.,, “Accelerating recurrent neural networks in analytics servers: comparison of FPGA, CPU, GPU, and ASIC,” in 26th Int. Conf. Field Programm. Logic and Appl. (FPL), 1 –4 (2016). https://doi.org/10.1109/FPL.2016.7577314 Google Scholar

57. 

J. W. Goodman, Introduction to Fourier Optics, Roberts and Company( (2005). Google Scholar

58. 

D. P. Kingma and J. Ba, “Adam: a method for stochastic optimization,” (2014). Google Scholar

59. 

M. Courbariaux, Y. Bengio and J.-P. David, “BinaryConnect: training deep neural networks with binary weights during propagations,” (2015). Google Scholar

60. 

H. Wang et al., “Toward near-perfect diffractive optical elements via nanoscale 3D printing,” ACS Nano, 14 (8), 10452 –10461 https://doi.org/10.1021/acsnano.0c04313 (2020). Google Scholar

61. 

Z. Zhang et al., “Space-time projection enabled ultrafast all-optical diffractive neural network,” Laser and Photonics Rev., 18 2301367 https://doi.org/10.1002/lpor.202301367 (2024). Google Scholar

62. 

J. Zhou, H. Pu and J. Yan, “Spatiotemporal diffractive deep neural networks,” Opt. Express, 32 (2), 1864 –1877 https://doi.org/10.1364/OE.494999 (2024). Google Scholar

63. 

Z. Duan, H. Chen and X. Lin, “Optical multi-task learning using multi-wavelength diffractive deep neural networks,” Nanophotonics, 12 (5), 893 –903 https://doi.org/10.1515/nanoph-2022-0615 (2023). Google Scholar

64. 

X. Luo et al., “Metasurface-enabled on-chip multiplexed diffractive neural networks in the visible,” Light Sci. Appl., 11 (1), 158 https://doi.org/10.1038/s41377-022-00844-2 (2022). Google Scholar

65. 

D. Mengu et al., “Misalignment resilient diffractive optical networks,” Nanophotonics, 9 (13), 4207 –4219 https://doi.org/10.1515/nanoph-2020-0291 (2020). Google Scholar

66. 

M. Nakajima et al., “Physical deep learning with biologically inspired training method: gradient-free approach for physical hardware,” Nat. Commun., 13 (1), 7847 https://doi.org/10.1038/s41467-022-35216-2 (2022). Google Scholar

67. 

A. Momeni et al., “Training of physical neural networks,” (2024). Google Scholar

68. 

K. Matsushima, “Formulation of the rotational transformation of wave fields and their application to digital holography,” Appl. Opt., 47 (19), D110 –D116 https://doi.org/10.1364/AO.47.00D110 (2008). Google Scholar

69. 

N. K. Fontaine et al., “Laguerre-Gaussian mode sorter,” Nat. Commun., 10 (1), 1865 https://doi.org/10.1038/s41467-019-09840-4 (2019). Google Scholar

70. 

Q. Wang et al., “Ultrahigh-fidelity spatial mode quantum gates in high-dimensional space by diffractive deep neural networks,” Light Sci. Appl., 13 (1), 10 https://doi.org/10.1038/s41377-023-01336-7 (2024). Google Scholar

71. 

M. Yildirim et al., “Nonlinear processing with linear optics,” Nat. Photonics, 17 https://doi.org/10.1038/s41566-024-01494-z (2024). Google Scholar

Biography

Ryosuke Mashiko is a graduate student in the Department of Information Physics and Computing at the Graduate School of Information Science and Technology, the University of Tokyo, Tokyo, Japan. He received his BE degree in engineering from the University of Tokyo, in 2023. His work focuses on the development of photonic computing and computational imaging.

Makoto Naruse was a professor in the Department of Information Physics and Computing, Graduate School of Information Science and Technology, the University of Tokyo, Tokyo, Japan. He received his BE, ME, and PhD degrees in engineering from the University of Tokyo, in 1994, 1996, and 1999, respectively.

Ryoichi Horisaki is an associate professor in the Department of Information Physics and Computing at the Graduate School of Information Science and Technology, the University of Tokyo, Tokyo, Japan. He received his BS degree in engineering from Nara National College of Technology in 2005, and his MS and PhD degrees in information science from Osaka University in 2007 and 2010, respectively. His research interests include information photonics and computational imaging.

CC BY: © The Authors. Published by SPIE and CLP under a Creative Commons Attribution 4.0 International License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI.
Ryosuke Mashiko, Makoto Naruse, and Ryoichi Horisaki "Diffraction casting," Advanced Photonics 6(5), 056005 (3 October 2024). https://doi.org/10.1117/1.AP.6.5.056005
Received: 12 May 2024; Accepted: 3 September 2024; Published: 3 October 2024
Lens.org Logo
CITATIONS
Cited by 1 scholarly publication.
Advertisement
Advertisement
KEYWORDS
Logic

Light sources and illumination

Diffraction

Optical computing

Image processing

Energy efficiency

Binary data

RELATED CONTENT


Back to Top