Energy Efficient Memristor Based on Green-Synthesized 2D Carbonyl-Decorated Organic Polymer and Application in Image Denoising and Edge Detection: Toward Sustainable AI

Abstract

According to the United Nations, around 53 million metric tons of electronic waste is produced every year, worldwide, the big majority of which goes unprocessed. With the rapid advances in AI technologies and adoption of smart gadgets, the demand for powerful logic and memory chips is expected to boom. Therefore, the development of green electronics is crucial to minimizing the impact of the alarmingly increasing e-waste. Here, it is shown the application of a green synthesized, chemically stable, carbonyl-decorated 2D organic, and biocompatible polymer as an active layer in a memristor device, sandwiched between potentially fully recyclable electrodes. The 2D polymer's ultramicro channels, decorated with C═O and O─H groups, efficiently promote the formation of copper nanofilaments. As a result, the device shows excellent bipolar resistive switching behavior with the potential to mimic synaptic plasticity. A large resistive switching window (10³), low SET/RESET voltage of ≈0.5/-1.5 V), excellent device-to-device stability and synaptic features are demonstrated. Leveraging the device's synaptic characteristics, its applications in image denoising and edge detection is examined. The results show a reduction in power consumption by a factor of 10³ compared to a traditional Tesla P40 graphics processing unit, indicating great promise for future sustainable AI-based applications.

Keywords: 2D polymers; green synthesis; memristor; neuromorphic computing; sustainable electronics.

Figure 1

a) The schematic representation of the synthesis of TpDb. b) The SEM image of TpDb top view. c) The schematic of preparation of TpDb solution. d) Cytotoxicity test of TpDb; which illustrate the cell viability with different concentrations of TpDb. e–h) Cell staining results where the green and red one indicate the living and dead cells, respectively, e) shows control, i.e., without TpDb f) with 0.5 µg mL⁻¹ concentration of TpDb, g) with 5 µg mL⁻¹ concentration of TpDb h) with 25 µg mL⁻¹ concentration of TpDb.

Figure 2

a) Schematic illustration of the biological synapse and the equivalent artificial/electronic synapse Cu/TpDb/Pt memristor device. b) The cross‐section TEM images of the devices with the scale bar of 20 nm. c) EDS elemental line profile to validate the various elements in the device. d) EDS mapping of various elements Pt, C, O, and Cu presented in the Cu/TpDb/Pt memristive device. e) Wide scan spectra X‐ray photo spectroscopy of TpDb/Pt. f) C 1s spectra of X‐ray photo spectroscopy of TpDb/Pt. g) O 1s spectra of X‐ray photo spectroscopy of TpDb/Pt.

Figure 3

Design and electrical characterization. a) I–V characteristics of the TpDb‐based device for 1st, 100th, 500th, and 1000th cycle. b) DC endurance or (LRS/HRS current) of the TpDb device. c) Device to device uniformity of D1, D4, D7, and D10 devices. d) Synaptic behavior of long‐term potentiation (LTP) and depression (LTD) of the TpDb device by 50 SET pulses and 50 RESET pulses. e) The PPF index of the device and Voltage pulse scheme for Paired pulse Facilitation (PPF) of the synaptic device (inset). f) Repeatability of LTP and LTD characteristics of the device with total number of 100 epochs.

Figure 4

Illustration of the electrochemical processes during resistive switching. a) Schematic of the fresh device. b) When positive (+) bias is applied on the top electrode, Cu starts oxidizing and Cu+ ions start moving toward TpDb layer. c) Migration of Cu+ ions in the TpDb layer indicating oxidation and reduction reactions happening before Cu+ reaches to the bottom electrode. d) The conductive filament (CF) grows continuously and at a certain point, the device attains the low resistance state (LRS) which implies the SET process. e) By applying the reverse bias, the conductive filament ruptures due to the joule heating effect (HRS/RESET).

Figure 5

a) Network structure of CNN autoencoder b) The kernels (left) are obtained from the ideal software test. The offline mapping kernels (right) are mapping from the corresponding ideal kernel values with device conductance. c) The figure visualization for different steps. d) The training loss and PSNR result with the training epochs. e) The comparison result between GPU and our memristor device.

Figure 6

a) The procedure of hardware‐based convolutional image processing performed on the KAUST campus image. b) Different convolutional operators can help extract features of the image for different detection applications.

Figure 7

Comparison of our TpDb memristor with previously reported organic and polymeric‐based memristor devices.