Introduction
The project aims to design the next generation of sustainable AI hardware
The PhD student will explore how atomically thin semiconductors can be engineered into devices that combine sensing, memory, and computing in one platform. You’ll investigate cutting-edge approaches to in-memory computing and retinomorphic vision sensors, inspired by the human brain and eye, with the goal of reducing the energy footprint of AI.
Project Overview
Conventional computing architectures face severe energy bottlenecks due to the separation of memory and processing. This so-called von Neumann bottleneck limits the efficiency of artificial intelligence (AI) systems, particularly in data-intensive tasks such as image recognition and real-time sensing. In this project, you will explore how atomically thin (2D) materials – such as layered ferroelectrics and semiconductors – can be engineered into novel devices that combine sensing, memory, and processing within the same physical platform.(1)
Research Challenges:
In-memory computation: How can ferroelectric resistive switching in ultrathin materials be harnessed for energy-efficient logic and memory that can mimic the human brain?
Retinomorphic sensing: Can we design photonic/electronic devices that directly process visual information (edge detection, contrast enhancement) at the sensor level, mimicking the human retina?
Scalability: What device engineering strategies will improve device reliability (2) and enable integration into larger arrays while maintaining low-power, high-speed operation?
Sustainability: How can these emerging device technologies contribute to Sustainable AI by reducing the carbon footprint of training and deploying AI hardware?
Methods and Approach
As part of the MINDS lab and Sustainable Electronic Technologies group, the student will build on recently pioneered van der Waals contact technology to fabricate and characterise reliable nano devices based on 2D ferroelectrics, semiconductors, and heterostructures (3,4). Electrical and optical testing will probe their memory, switching, and sensory capabilities. The project combines nanofabrication at Southampton's world-class cleanrooms, with opportunities for collaboration in system-level modelling and AI applications.
Collaborations with the University of Cambridge and National Physical Laboratory on materials science and characterisation will allow benchmarking against industry standards.
References:
(1) Feng Guangdi, et al. "Retinomorphic hardware for in‐sensor computing." InfoMat 5.9 (2023): e12473.
(2) Wang Yan, Sarkar Soumya et al. "Critical challenges in the development of electronics based on two-dimensional transition metal dichalcogenides." Nature Electronics 7.8 (2024): 638-645.
(3) Sarkar, Soumya, et al. "Spin injection in graphene using ferromagnetic van der Waals contacts of indium and cobalt." Nature Electronics 8.3 (2025): 215-221.
(4) Sarkar, Soumya, et al. "Multistate ferroelectric diodes with high electroresistance based on van der Waals heterostructures." Nano letters 24.42 (2024): 13232-13237.