Study an‎d investigation of copper-based composites, used in the anode of lithium batteries

2/16/2027 12:00:00 AM

Lithium-ion batteries have become the dominant battery in the market over the past decade due to their reasonable cost, high efficiency, an‎d long cycle life. Graphite, commonly used in the anode of these batteries, has a theoretical capacity of 372 mAh/g, which is insufficient for high‑deman‎d applications. Silicon, with a theoretical capacity of 4200 mAh/g, is a suitable alternative; however, its main issue is 300% volumetric expansion during charge–discharge, leading to capacity fading an‎d anode degradation. One solution is to use silicon composites with other materials, such as Si‑Cu‑C, which improve electron transport. Initially, the Si‑Cu‑C composite was prepared using ball milling, followed by coating an‎d battery cell assembly. Characterization an‎d electrochemical tests were performed. In the neural network section, the effects of the number of layers, neurons, an‎d the test data split ratio were investigated. Regression plots, correlation tables, an‎d descriptive statistics were examined. The SiCuC electrode was eva‎luated for use in lithium‑ion batteries. Although it shows capacity loss in the initial cycles, it exhibits stable capacity retention an‎d a coulombic efficiency close to 100% (with an initial efficiency of 64.86%) over long‑term cycling, indicating good electrochemical stability at 0.5 C. Rate capability tests at rates from 0.1 C to 1 C an‎d back to 0.1 C showed capacity recovery to 381–453 mAh/g, suggesting that the capacity loss at higher rates is mainly kinetic in nature an‎d not due to structural degradation. Diffraction analysis an‎d elemental mapping revealed non‑uniform carbon distribution (local agglomeration), a more uniform silicon distribution, limited copper presence, an‎d negligible oxygen. EDX composition was 79.55% carbon, 11.42% silicon, 4.14% copper, an‎d 4.89% oxygen. Statistical analysis of 3340 data points an‎d neural network architectures showed that increasing the number of neurons an‎d layers improves the capacity prediction accuracy. Based on R², MAE, an‎d RMSE, the best performance was achieved with a three‑layer, 30‑neuron architecture (with a test data volume of 0.1), while the worst performance corresponded to a single‑layer, 5‑neuron architecture.

باتري , كامپوزيت , مس , آند باتري‌هاي ‌ليتيومي , شبكه عصبي

Battery , Composite , Copper , Lithium‑ion battery anode , Neural network

Aria Moradi

Mohammadreza Zamani Mimian