Photocatalytic CO₂ conversions on copper nanoparticles investigated by Raman spectral changes using convolutional neural networks

A convolutional neural network (CNN) deep learning process is employed to analyze in situ Raman scattering data for CO₂ capture and its photocatalytic conversions onto copper sulfide hollow nanospheres (CuSHNSs) and copper nanocubes (CuNCs) in microalgae solution of Spirulina maxima. Raman spectra under visible light at 633 nm in a microfluidic solution provided representative vibrational marker bands of C[dbnd]O features at ∼2100 cm⁻¹ and CH₂/CH₃ bending vibrations at ∼1400 cm⁻¹ that are correlated with CO₂ reduction products of carbon monoxide (C₁) and multi‑carbon species such as propanol (C₃), butanol (C₄), respectively. Accumulated Raman spectra were trained and analyzed to estimate photocatalytic pathways using CNN algorithm. The presence of Spirulina maxima microalgae on the alteration of photocatalytic processes is studied by analyzing collective Raman spectral changes. The main observation is that strong CO peaks in Raman spectra of CO₂ adsorbed by CuNCs almost disappeared after treatment with microalgae, whereas their intensities were slightly increased in case of CuSHNS. The CNN deep learning process for Raman spectra was effective to differentiate photocatalytic mechanisms of CO₂ conversion onto nanoparticle surfaces.