Ultrafast and compact photonic-electronic leaky integrate-and-fire circuits based upon resonant tunnelling diodes Journal: opticajournal Publication type: Research Article
Joshua Robertson, 1,* Dylan Black, 1 Giovanni Donati, 1 Qusay Raghib Ali Al-Taai, 2 Ekaterina Malysheva, 3 Bruno Romeira, 4 José Figueiredo, 5 Victor Dolores Calzadilla, 3 Edward Wasige, 2 and Antonio Hurtado 1 · 2025
Read Ultrafast and compact photonic-electronic leaky integrate-and-fire circuits based upon resonant tunnelling diodes Journal: opticajournal Publication type: Research Article by Joshua Robertson, 1,* Dylan Black, 1 Giovanni Donati, 1 Qusay Raghib Ali Al-Taai, 2 Ekaterina Malysheva, 3 Bruno Romeira, 4 José Figueiredo, 5 Victor Dolores Calzadilla, 3 Edward Wasige, 2 and Antonio Hurtado 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-17133.txt.
Introduction
Joshua Robertson and ten co-authors present a 2025 paper on tiny circuits that mimic nerve cells using light and electronics together. The circuits are built around a device whose current jumps sharply past a threshold. Fed with light pulses, they build up charge and fire once it crosses a set line, like a neuron.
The team built two forms: one pairs the device with a separate light sensor, and one folds sensing into the device itself. Both fired in response to several fast light pulses landing within a window under a nanosecond. Single pulses ran about one to two hundred picoseconds. The circuits point to small, swift hardware for brain like computing.
Catalog record
| CREATOR | Joshua Robertson, 1,* Dylan Black, 1 Giovanni Donati, 1 Qusay Raghib Ali Al-Taai, 2 Ekaterina Malysheva, 3 Bruno Romeira, 4 José Figueiredo, 5 Victor Dolores Calzadilla, 3 Edward Wasige, 2 and Antonio Hurtado 1 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 23 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-17133
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Ultrafast and compact photonic-electronic leaky integrate-and-fire circuits based upon resonant tunnelling diodes Journal: opticajournal Publication type: Research Article Joshua Robertson, 1,* Dylan Black, 1 Giovanni Donati, 1 Qusay Raghib Ali Al-Taai, 2 Ekaterina Malysheva, 3 Bruno Romeira, 4 José Figueiredo, 5 Victor Dolores Calzadilla, 3 Edward Wasige, 2 and Antonio Hurtado 1 Abstract This work provides a first report of ultrafast and compact photonic-electronic neuromorphic temporal leaky integrate-and-fire neuronal circuits built with Resonant Tunnelling Diodes (RTDs). We demonstrate experimentally that multiple fast ( ∼ \sim 100-200 ps) optical input pulses, arriving within a short (sub-ns long) temporal window, control the triggering of excitable responses in two different photonic-electronic RTD circuit architectures. These architectures are an electronic RTD coupled externally to a photodetector (referred to as a PD-RTD), and an integrated opto-electronic RTD device with inherent photodetection capability at infrared telecom wavelengths (referred to as an Optical RTD-PD). For both RTD systems, we reveal that the high-speed optically-triggered integrate-and-fire spiking operation can be precisely controlled by acting on the voltage bias applied to the RTD devices, or via the intensity of incoming optical pulses. Finally, we demonstrate the application of the leaky integrate-and-fire behaviour to a pattern recognition task at high-speed, with the systems triggering fast ns-long electrical spikes in response to optical inputs of weighted 4-bit digital headers. 1 Introduction Neuromorphic computing, the emulation of biological neural functionality and neural circuits, has in recent years emerged as one of the key solutions for the next generation of computing architectures [ 1 ] . Driven by the desire to improve substantially energy efficiency and achieve processing at the edge, neuromorphic systems enable the implementation of artificial neural networks, and in many cases Spiking Neural Networks (SNNs), using numerous electronic and photonic technologies [ 2 ] . Devices based on memristors [ 3 , 4 ] , phase-change materials [ 5 , 6 ] , modulators [ 7 , 8 ] and semiconductors lasers [ 9 , 10 ] (amongst others), are all seeing success by exploiting the nonlinear processes (all-optical, all-electrical or opto-electronic) in their approach, generating a plethora of neuromorphic contenders with a variety of benefits, timescales, and use-cases (e.g. image classification or real-time audio recognition) [ 11 ] . One type of promising neuromorphic device is Resonant Tunnelling Diodes (RTDs). RTDs are chip-scale semiconductor devices that are rapidly developing as candidates for high-speed and efficient artificial opto-electronic spiking neurons [ 12 , 13 , 14 ] . RTDs contain embedded double barrier quantum wells (DBQW) that give rise to highly non-linear I-V characteristics, producing regions of negative differential resistance (NDR) and different complex nonlinear dynamics, including crucially excitability and bursting [ 12 , 13 , 14 ] , enabling spike firing responses analogous biological neurons. RTDs have also demonstrated high bandwidth operation for applications such as THz oscillation generation and detection [ 15 , 16 ] , thus offering promise for the realisation of ultrafast artificial spiking neuron operation. Further, opto-electronic RTDs, fabricated with embedded photodetecting layers, have shown the ability to deterministically trigger spiking regimes when subject to optical stimulation [ 17 , 18 ] . The versatility of RTDs, whether they are operated purely electronically or opto-electronically [ 19 ] , makes them suitable candidates for scalable, multimodal, non-linear spiking systems for present and future neuromorphic computing hardware. However, in SNN architectures several neuronal models can be implemented, each with different toolkits of functionality. One of these neuronal models, most commonly used for spike-based processing, is the leaky integrate-and-fire neuronal model [ 20 ] . These neurons are characterised by their ability to collect and combine input signals over time, with a counteracting ’leaky’ decay constant, until the total contribution crosses a firing threshold and produces an excitable response. Leaky integrate-and-fire neurons, like those shown in Figure 1 , therefore allow multiple inputs (temporal and spatial) from different dendritic pathways to contribute to the same spike activation, given the inputs occur within a temporal window shorter than the decay time of the system. Leaky integrate-and-fire neurons directly allow for the processing of input information, and for this reason several neuromorphic devices have seen investigations into this behaviour [ 4 , 10 , 21 , 22 ] . In this work, we demonstrate for the first time photonic-electronic RTD neuronal circuits exhibiting a temporal leaky integrate-and-fire behaviour when subject to time-multiplexed optical pulses (see Figure 1 ). Two RTD architectures are investigated, which include an electronic RTD externally-coupled to a fast photodetector, and an opto-electronic RTD with an embedded infrared light detection semiconductor layer. In this work, both RTD circuits are applied to a practical 4-bit pattern recognition task to showcase their potential as scalable opto-electronic devices for neuromorphic systems. Figure 1: Schematic of a leaky integrate-and-fire neuron model and an optical RTD-PD leaky integrate-and-fire neuron. In the optical RTD-PD neuron multiple time-multiplexed optical inputs arrive on the neurons where temporal integration occurs within the device and an electrical spike is fired. 2 Methods Two types of neuromorphic spiking RTD structures were investigated in this work. The first structure is a 3 \mu m radius mesa, fully electronic, RTD featuring an AlAs/InGaAs/AlAs double barrier quantum well (DBQW) [ 14 ] . It is fed both a DC and an RF electrical input (via RF probes and a bias tee) that arrive from a power supply and a fast amplified photodetector (PD), respectively. In this configuration (referred to here as a PD-RTD), optical inputs incident on the PD are converted to electrical signals for injection into the RTD device. The optical signals are generated by modulating the output of a 1546 nm tuneable laser source with a Mach-Zehnder (MZ) intensity modulator and an Arbitrary Waveform Generator (AWG). A 50:50 splitter is used to analyse the RF output of the PD-RTD following the injection of input signals. A schematic of the PD-RTD setup is shown in Figure 2 (a). The second structure studied in this work is an opto-electronic RTD with a similar DBQW stack, but with the inclusion of an embedded spacer layer (that enables light absorption) and top layer nanopillar structure (for improved carrier confinement) [ 18 ] . In this configuration (referred to here as an Optical RTD-PD), the modulated optical inputs from a 1546 nm tuneable laser are directly introduced to the RTD device via a lensed fiber. The lensed fibre delivered an average power of 0.885 mW to the light absorbing spacer layer of the RTD. The RTD device was driven with a DC bias from a power supply ( V_{bias} ), and again the RF output of the RTD was monitored for analysis. A schematic of the optical RTD-PD setup is shown in Figure 2 (c). In both configurations a 12 GSa/s (5 GHz) arbitrary waveform generator was used to encode fast optical pulses in the system, and a fast 40 GSa/s (16 GHz) oscilloscope was used to analyse and capture the corresponding electrical responses. Full details of the structures, dimensions and epilayer structures of the RTD devices used in this work can be found in [ 14 , 18 ] . The current-voltage (I-V) characteristics of both the PD-RTD and optical RTD-PD devices are shown in Figure 2 (b) and Figure 2 (d),