Hybrid Perovskites & Ferroelectric Polymers for Designing Artificial Rod Photoreceptors

Article: Perovskites used to make efficient artificial retina
Source: King Abdullah University of Science and Technology (Saudi Arabia)
Published: February 23, 2022
Article: New photoreceptors could replicate our eye's sensitivity to light
Published: February 12, 2022

The capacitive photoreceptors' sensitivities to different LED illumination,
indicating tunable properties, with peak sensitivity for green-yellow light

Neuromorphic engineering is the design of electronic circuits to mimic the neuro-biological architecture found in nervous systems. Neuromorphic vision sensors, which reflect the architecture of the eye, such as the retina's photoreceptors, could inspire the design of robotic vision systems and perceptive intelligence (security) applications. Researchers in Saudi Arabia are exploring the use of a hybrid nanocomposite material with capacitive and dielectric properties to create biomimetic photoreceptors that require less energy to operate. In this case, they engineered a hybrid composite of methylammonium lead bromide perovskite (MAPbBr₃) and the terpolymer polyvinylidene fluoride trifluoroethylene-chlorofluoroethylene (PVDF-TrFE-CFE). The perovskite is a strong absorber of visible light, currently being explored in applications like solar cell research, while the ferroelectric polymer has a high dielectric constant, which is useful for efficient charging and discharging. Termed a light intensity capacitive photoreceptor (CPR), the 4 x 4 metal-insulator-metal capacitor array responded to various wavelengths of visible light when illuminated with different colored LEDs. Moreover, this CPR mimics the spectral sensitivity curve of human photopic vision, that is, with peak sensitivity for green-yellow light. Additionally, the authors report that the hybrid material is highly stable, with no degradation or change in response after storage for 129 weeks (~2.5 years) in ambient conditions. Finally, they observed that the CPR with 100 output neurons could, after unsupervised training, recognize handwritten digits with > 70% accuracy; they hypothesize that accuracy could reach 95% with a network of 1,500 neurons or if supervised deep learning algorithms were used. The team next plan to expand the size of the array, optimize interface circuitry, and improve accuracy with a multilayered neural network.

Spiking neural network processing the spike train, generated after seeing handwritten digits

My rating of this study: ⭐⭐

Vijjapu MT, Fouda ME, Agambayev A, et al. "A flexible capacitive photoreceptor for the biomimetic retina." Light: Science & Applications  11:3. 1 January 2022. https://doi.org/10.1038/s41377-021-00686-4