Ragib Ahsan
Research Subjects
Ragib Ahsan has a total of 71 co-authors. Here are the top co-authors by number of shared publications:
- Rehan Kapadia
Ming Hsieh Department of Electrical and Computer Engineering University of Southern California Los Angeles CA 90089 USA
17 shared publications - Hyun Uk Chae
Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, California 90089, United States
8 shared publications - Stephen B. Cronin
University of Southern California
4 shared publications - Anika Tabassum Priyoti
Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California 4 , Los Angeles, California 90089
3 shared publications Constantine Sideris
Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, California 90089, United States
3 shared publications- Debarghya Sarkar
Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, California 90089, United States
3 shared publications
Latest Publications
- 16 June 2026
- Published by AIP Publishing in Journal of Applied Physics
- Vol. 139 (23)
Integration of oscillator-based feature extraction for energy-efficient convolutional neural networks
The rapid growth in machine learning and artificial intelligence workloads has increased the demand for computing power. Although digital computing accelerators dominate today’s market, analog computing architectures are emerging as promising energy-efficient alternatives. In this work, we propose an approach to address these challenges by replacing the first layer of a convolutional neural network (CNN) with a network of oscillatory retinal neurons (ORNs) composed of weakly coupled, optically activated negative differential resistance (NDR) devices. Each ORN consists of a photodetector exhibiting NDR behavior under illumination, coupled with an inductor forming a self-oscillating circuit without needing external voltage sources. We model the nonlinear oscillator dynamics using experimentally measured device characteristics and simulate their behavior under varying optical inputs. Simulations performed on the Fashion-MNIST (Modified National Institute of Standards and Technology) dataset demonstrate that the hybrid ORN–CNN architecture maintains high recognition accuracy and achieves 95% training accuracy and 92% testing accuracy comparable to the fully software-defined CNN with 93% test accuracy while dramatically reducing energy consumption to 24 aJ/operation. Different ORN network topologies were investigated, with asymmetric inductive coupling achieving consistently high performance. These results highlight the potential of oscillator-based networks for low-power analog computing and demonstrate the possibility of using nonlinear physical systems for energy-efficient machine learning applications. This work demonstrates a possible path for integrating device-level oscillator dynamics into future neuromorphic and analog computing platforms.
- 16 June 2026
- Published by AIP Publishing in Journal of Applied Physics
- Vol. 139 (23)
Integration of oscillator-based feature extraction for energy-efficient convolutional neural networks
The rapid growth in machine learning and artificial intelligence workloads has increased the demand for computing power. Although digital computing accelerators dominate today’s market, analog computing architectures are emerging as promising energy-efficient alternatives. In this work, we propose an approach to address these challenges by replacing the first layer of a convolutional neural network (CNN) with a network of oscillatory retinal neurons (ORNs) composed of weakly coupled, optically activated negative differential resistance (NDR) devices. Each ORN consists of a photodetector exhibiting NDR behavior under illumination, coupled with an inductor forming a self-oscillating circuit without needing external voltage sources. We model the nonlinear oscillator dynamics using experimentally measured device characteristics and simulate their behavior under varying optical inputs. Simulations performed on the Fashion-MNIST (Modified National Institute of Standards and Technology) dataset demonstrate that the hybrid ORN–CNN architecture maintains high recognition accuracy and achieves 95% training accuracy and 92% testing accuracy comparable to the fully software-defined CNN with 93% test accuracy while dramatically reducing energy consumption to 24 aJ/operation. Different ORN network topologies were investigated, with asymmetric inductive coupling achieving consistently high performance. These results highlight the potential of oscillator-based networks for low-power analog computing and demonstrate the possibility of using nonlinear physical systems for energy-efficient machine learning applications. This work demonstrates a possible path for integrating device-level oscillator dynamics into future neuromorphic and analog computing platforms.
- 13 October 2025
- Published by American Chemical Society (ACS) in Nano Letters
- Vol. 25 (43) , 15487-15494
Evanescent Mode Photoemission
Photoemission of electrons from solid surfaces into vacuum is routinely used in two configurations: reflection mode, in which light is incident on the emitting surface, and transmission mode, in which light illuminates the photoemissive material from behind. Here, using silicon nitride waveguides integrated beneath a high quantum efficiency cesium antimonide thin film, we demonstrate a new photoemission regime─the evanescent mode photoemission─in which light travels parallel to the surface in the waveguide while evanescently coupling into the thin film to excite and emit photoelectrons. Our experiments show that this configuration enables direct visualization of guided optical modes over an unprecedented range (∼100 nm to 1 mm) and allows electron beam shaping at the source with transverse features well below ∼600 nm, establishing evanescent mode photoemission as a new platform for nanoscale optical field mapping and precise generation of electron beams.
- 13 October 2025
- Published by American Chemical Society (ACS) in Nano Letters
- Vol. 25 (43) , 15487-15494
Evanescent Mode Photoemission
Photoemission of electrons from solid surfaces into vacuum is routinely used in two configurations: reflection mode, in which light is incident on the emitting surface, and transmission mode, in which light illuminates the photoemissive material from behind. Here, using silicon nitride waveguides integrated beneath a high quantum efficiency cesium antimonide thin film, we demonstrate a new photoemission regime─the evanescent mode photoemission─in which light travels parallel to the surface in the waveguide while evanescently coupling into the thin film to excite and emit photoelectrons. Our experiments show that this configuration enables direct visualization of guided optical modes over an unprecedented range (∼100 nm to 1 mm) and allows electron beam shaping at the source with transverse features well below ∼600 nm, establishing evanescent mode photoemission as a new platform for nanoscale optical field mapping and precise generation of electron beams.
- 6 March 2025
- Published by American Chemical Society (ACS) in ACS Applied Materials & Interfaces
- Vol. 17 (11) , 17445-17453
Hot Electron Photoemission from Tunable Electron Affinity Semiconductor Cathodes
A voltage-tunable negative electron affinity (NEA) semiconductor photocathode offers one key advantage over current materials such as cesiated NEA photocathodes: stability under ambient conditions. A semiconductor/insulator/graphene heterostructure can inject electrons into the conduction band of the insulator, where an electric field "heats" them up so that the effective emission barrier seen by the "hot" electrons is negative, enabling a voltage-tunable NEA surface. Here, we have experimentally demonstrated a peak emission current density of 2.253 × 10-3 A/cm2 and a peak external quantum efficiency (EQE) of ∼1.53% from a p-Si/amorphous-Al2O3/graphene-based hot electron laser-assisted cathode (HELAC). We have developed a full-band Monte Carlo Boltzmann Transport Equation (MCBTE) solver to study the hot electron transport behavior in three different crystalline insulators: SiO2, Al2O3, and MgO. Through MCBTE and semiconductor device simulations, we have predicted a peak emission current density of ∼103 A/cm2, far above our experimental value, indicating that the optimal performance of state-of-the-art HELACs has not yet been realized. This theoretical framework provides an understanding of the key performance limitations of the device and can be used to guide the optimal design (e.g., through the selection of new materials) of voltage-tunable NEA semiconductor photocathodes.
- 6 March 2025
- Published by American Chemical Society (ACS) in ACS Applied Materials & Interfaces
- Vol. 17 (11) , 17445-17453
Hot Electron Photoemission from Tunable Electron Affinity Semiconductor Cathodes
A voltage-tunable negative electron affinity (NEA) semiconductor photocathode offers one key advantage over current materials such as cesiated NEA photocathodes: stability under ambient conditions. A semiconductor/insulator/graphene heterostructure can inject electrons into the conduction band of the insulator, where an electric field "heats" them up so that the effective emission barrier seen by the "hot" electrons is negative, enabling a voltage-tunable NEA surface. Here, we have experimentally demonstrated a peak emission current density of 2.253 × 10-3 A/cm2 and a peak external quantum efficiency (EQE) of ∼1.53% from a p-Si/amorphous-Al2O3/graphene-based hot electron laser-assisted cathode (HELAC). We have developed a full-band Monte Carlo Boltzmann Transport Equation (MCBTE) solver to study the hot electron transport behavior in three different crystalline insulators: SiO2, Al2O3, and MgO. Through MCBTE and semiconductor device simulations, we have predicted a peak emission current density of ∼103 A/cm2, far above our experimental value, indicating that the optimal performance of state-of-the-art HELACs has not yet been realized. This theoretical framework provides an understanding of the key performance limitations of the device and can be used to guide the optimal design (e.g., through the selection of new materials) of voltage-tunable NEA semiconductor photocathodes.
- 14 August 2024
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 18 (34) , 23785-23796
Ultralow Power In-Sensor Neuronal Computing with Oscillatory Retinal Neurons for Frequency-Multiplexed, Parallel Machine Vision
In-sensor and near-sensor computing architectures enable multiply accumulate operations to be carried out directly at the point of sensing. In-sensor architectures offer dramatic power and speed improvements over traditional von Neumann architectures by eliminating multiple analog-to-digital conversions, data storage, and data movement operations. Current in-sensor processing approaches rely on tunable sensors or additional weighting elements to perform linear functions such as multiply accumulate operations as the sensor acquires data. This work implements in-sensor computing with an oscillatory retinal neuron device that converts incident optical signals into voltage oscillations. A computing scheme is introduced based on the frequency shift of coupled oscillators that enables parallel, frequency multiplexed, nonlinear operations on the inputs. An experimentally implemented 3 x 3 focal plane array of coupled neurons shows that functions approximating edge detection, thresholding, and segmentation occur in parallel. An example of inference on handwritten digits from the MNIST database is also experimentally demonstrated with a 3 x 3 array of coupled neurons feeding into a single hidden layer neural network, approximating a liquid-state machine. Finally, the equivalent energy consumption to carry out image processing operations, including peripherals such as the Fourier transform circuits, is projected to be <20 fJ/OP, possibly reaching as low as 15 aJ/OP.
- 14 August 2024
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 18 (34) , 23785-23796
Ultralow Power In-Sensor Neuronal Computing with Oscillatory Retinal Neurons for Frequency-Multiplexed, Parallel Machine Vision
In-sensor and near-sensor computing architectures enable multiply accumulate operations to be carried out directly at the point of sensing. In-sensor architectures offer dramatic power and speed improvements over traditional von Neumann architectures by eliminating multiple analog-to-digital conversions, data storage, and data movement operations. Current in-sensor processing approaches rely on tunable sensors or additional weighting elements to perform linear functions such as multiply accumulate operations as the sensor acquires data. This work implements in-sensor computing with an oscillatory retinal neuron device that converts incident optical signals into voltage oscillations. A computing scheme is introduced based on the frequency shift of coupled oscillators that enables parallel, frequency multiplexed, nonlinear operations on the inputs. An experimentally implemented 3 x 3 focal plane array of coupled neurons shows that functions approximating edge detection, thresholding, and segmentation occur in parallel. An example of inference on handwritten digits from the MNIST database is also experimentally demonstrated with a 3 x 3 array of coupled neurons feeding into a single hidden layer neural network, approximating a liquid-state machine. Finally, the equivalent energy consumption to carry out image processing operations, including peripherals such as the Fourier transform circuits, is projected to be <20 fJ/OP, possibly reaching as low as 15 aJ/OP.
- Open Access
- 9 April 2024
- Published by American Chemical Society (ACS) in ACS Omega
- Vol. 9 (16) , 18062-18071
Ultralow Power Electronic Analog of a Biological Fitzhugh–Nagumo Neuron
Here, we introduce an electronic circuit that mimics the functionality of a biological spiking neuron following the Fitzhugh-Nagumo (FN) model. The circuit consists of a tunnel diode that exhibits negative differential resistance (NDR) and an active inductive element implemented by a single MOSFET. The FN neuron converts a DC voltage excitation into voltage spikes analogous to biological action potentials. We predict an energy cost of 2 aJ/cycle through detailed simulation and modeling for these FN neurons. Such an FN neuron is CMOS compatible and enables ultralow power oscillatory and spiking neural network hardware. We demonstrate that FN neurons can be used for oscillator-based computing in a coupled oscillator network to form an oscillator Ising machine (OIM) that can solve computationally hard NP-complete max-cut problems while showing robustness toward process variations.
- 9 April 2024
- Published by American Chemical Society (ACS) in ACS Omega
- Vol. 9 (16) , 18062-18071
Open AccessUltralow Power Electronic Analog of a Biological Fitzhugh–Nagumo Neuron
Here, we introduce an electronic circuit that mimics the functionality of a biological spiking neuron following the Fitzhugh-Nagumo (FN) model. The circuit consists of a tunnel diode that exhibits negative differential resistance (NDR) and an active inductive element implemented by a single MOSFET. The FN neuron converts a DC voltage excitation into voltage spikes analogous to biological action potentials. We predict an energy cost of 2 aJ/cycle through detailed simulation and modeling for these FN neurons. Such an FN neuron is CMOS compatible and enables ultralow power oscillatory and spiking neural network hardware. We demonstrate that FN neurons can be used for oscillator-based computing in a coupled oscillator network to form an oscillator Ising machine (OIM) that can solve computationally hard NP-complete max-cut problems while showing robustness toward process variations.
- 20 February 2024
- Published by American Chemical Society (ACS) in ACS Photonics
- Vol. 11 (3) , 1109-1116
Photoconductive Effects in Single Crystals of BaZrS3
- 20 February 2024
- Published by American Chemical Society (ACS) in ACS Photonics
- Vol. 11 (3) , 1109-1116
Photoconductive Effects in Single Crystals of BaZrS3
- 13 February 2024
- Published by American Chemical Society (ACS) in Nano Letters
- Vol. 24 (8) , 2581-2588
GaAs Mid-IR Electrically Tunable Metasurfaces
In this work, we explore III-V based metal-semiconductor-metal structures for tunable metasurfaces. We use an epitaxial transfer technique to transfer a III-V thin film directly on metallic surfaces, realizing III-V metal-semiconductor-metal (MSM) structures without heavily doped semiconductors as substitutes for metal layers. The device platform consists of gold metal layers with a p-i-n GaAs junction. The target resonance wavelength can be tuned by modifying the geometry of the top metal grating on the GaAs, while systematic resonance tunability has been shown through the modulation of various carrier concentration injections in the mid-IR range. Electrically tunable metasurfaces with multilevel biasing can serve as a fundamental building block for electrically tunable metasurfaces. We believe that our demonstration can contribute to understanding the optical tuning of III-V under various biased conditions, inducing changes in metasurfaces.
- 13 February 2024
- Published by American Chemical Society (ACS) in Nano Letters
- Vol. 24 (8) , 2581-2588
GaAs Mid-IR Electrically Tunable Metasurfaces
In this work, we explore III-V based metal-semiconductor-metal structures for tunable metasurfaces. We use an epitaxial transfer technique to transfer a III-V thin film directly on metallic surfaces, realizing III-V metal-semiconductor-metal (MSM) structures without heavily doped semiconductors as substitutes for metal layers. The device platform consists of gold metal layers with a p-i-n GaAs junction. The target resonance wavelength can be tuned by modifying the geometry of the top metal grating on the GaAs, while systematic resonance tunability has been shown through the modulation of various carrier concentration injections in the mid-IR range. Electrically tunable metasurfaces with multilevel biasing can serve as a fundamental building block for electrically tunable metasurfaces. We believe that our demonstration can contribute to understanding the optical tuning of III-V under various biased conditions, inducing changes in metasurfaces.
- Open Access
- 30 January 2024
- Published by Springer Nature in Research Square
Ultra-low power in-sensor neuronal computing with oscillatory retinal neurons for frequency-multiplexed, parallel machine vision
In-sensor and near-sensor computing architectures enable multiply-accumulate operations to be carried out directly at the point of sensing. In-sensor architectures offer dramatic power and speed improvements over traditional von Neumann architectures by eliminating multiple analog-to-digital conversions, data storage, and data movement operations. Current in-sensor processing approaches rely on tunable sensors or additional weighting elements to perform linear functions such as multiply-accumulate operations as the sensor acquires data. We implement in-sensor computing with an oscillatory retinal neuron device that converts incident optical signals into voltage oscillations. We introduce a computing scheme based on the frequency shift of coupled oscillators that enables parallel, frequency multiplexed, non-linear operations on the inputs. Simulation elucidates how this computing occurs. We experimentally implement a 3×3 focal plane array of coupled neurons and show that functions approximating edge detection, thresholding, and segmentation occur in parallel. An example of inference on handwritten digits from the MNIST database is also experimentally demonstrated with a 3×3 array of coupled neurons feeding into a single hidden layer neural network, approximating a liquid state machine. This network demonstrated a reduction in classification error from 2.16% to 1.84% compared to the same neural network with standard photodetector inputs. Finally, the equivalent energy consumption to carry out image processing operations, including peripherals such as the Fourier transform circuits, is projected to be < 20 fJ/OP, possibly reaching as low as 15 aJ/OP.
- 30 January 2024
- Published by Springer Nature in Research Square
Open AccessUltra-low power in-sensor neuronal computing with oscillatory retinal neurons for frequency-multiplexed, parallel machine vision
In-sensor and near-sensor computing architectures enable multiply-accumulate operations to be carried out directly at the point of sensing. In-sensor architectures offer dramatic power and speed improvements over traditional von Neumann architectures by eliminating multiple analog-to-digital conversions, data storage, and data movement operations. Current in-sensor processing approaches rely on tunable sensors or additional weighting elements to perform linear functions such as multiply-accumulate operations as the sensor acquires data. We implement in-sensor computing with an oscillatory retinal neuron device that converts incident optical signals into voltage oscillations. We introduce a computing scheme based on the frequency shift of coupled oscillators that enables parallel, frequency multiplexed, non-linear operations on the inputs. Simulation elucidates how this computing occurs. We experimentally implement a 3×3 focal plane array of coupled neurons and show that functions approximating edge detection, thresholding, and segmentation occur in parallel. An example of inference on handwritten digits from the MNIST database is also experimentally demonstrated with a 3×3 array of coupled neurons feeding into a single hidden layer neural network, approximating a liquid state machine. This network demonstrated a reduction in classification error from 2.16% to 1.84% compared to the same neural network with standard photodetector inputs. Finally, the equivalent energy consumption to carry out image processing operations, including peripherals such as the Fourier transform circuits, is projected to be < 20 fJ/OP, possibly reaching as low as 15 aJ/OP.
- 5 December 2023
- Published by American Chemical Society (ACS) in ACS Photonics
- Vol. 10 (12) , 4501-4508
Electrostatically Generated Air-Stable Negative Electron Affinity Silicon Photocathode
- 5 December 2023
- Published by American Chemical Society (ACS) in ACS Photonics
- Vol. 10 (12) , 4501-4508
Electrostatically Generated Air-Stable Negative Electron Affinity Silicon Photocathode
- 3 November 2022
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 16 (11) , 18497-18502
Monolithic III–V on Metal for Thermal Metasurfaces
It has been proposed that metal-semiconductor-metal (MSM) structures can be used to tune the absorptivity of a metasurface at infrared wavelengths. Indium arsenide (InAs) is a low-band-gap, high-electron-mobility semiconductor that may enable rapid index tuning for dynamic control over the infrared spectrum. However, direct growth of III-V thin films on top of metals has typically resulted in small-grain, polycrystalline materials that are not amenable to high-quality devices. Previously, epitaxial wafers were used for this purpose. However, the epitaxial constraints required that InAs be used for both the tuning layer and the bottom "metallic" layer, limiting the range of accessible designs. In this work, we show a demonstration of direct growth of single-crystalline InAs on metal to build tunable absorbers/emitters in the infrared regime. The growth was carried out at a temperature of 300 °C by the low temperature templated liquid phase (LT-TLP) method. The size of InAs single-crystalline mesas is ∼2500 μm2, enabling the desired device sizes. The proposed growth and device enable scalable and tunable infrared devices for various thermal-photonic applications.
- 3 November 2022
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 16 (11) , 18497-18502
Monolithic III–V on Metal for Thermal Metasurfaces
It has been proposed that metal-semiconductor-metal (MSM) structures can be used to tune the absorptivity of a metasurface at infrared wavelengths. Indium arsenide (InAs) is a low-band-gap, high-electron-mobility semiconductor that may enable rapid index tuning for dynamic control over the infrared spectrum. However, direct growth of III-V thin films on top of metals has typically resulted in small-grain, polycrystalline materials that are not amenable to high-quality devices. Previously, epitaxial wafers were used for this purpose. However, the epitaxial constraints required that InAs be used for both the tuning layer and the bottom "metallic" layer, limiting the range of accessible designs. In this work, we show a demonstration of direct growth of single-crystalline InAs on metal to build tunable absorbers/emitters in the infrared regime. The growth was carried out at a temperature of 300 °C by the low temperature templated liquid phase (LT-TLP) method. The size of InAs single-crystalline mesas is ∼2500 μm2, enabling the desired device sizes. The proposed growth and device enable scalable and tunable infrared devices for various thermal-photonic applications.