Identification of organic compounds using artificial neural networks and refractive index

Abstract

Identification of chemical compounds has many applications in science and technology. However, this process still relies significantly on the knowledge and experience of chemists. Thus, the development of techniques for faster and more accurate chemical compound identification is essential. In this work, we demonstrate the feasibility of using artificial neural networks to accurately identify organic compounds through the measurement of refractive index. The models were developed based on the refractive index measurements in different wavelengths of light, from UV to the far-infrared region. The models were trained with about 250,000 records of experimental optical constants for 60 organic compounds and polymers from published literature. The models performed with accuracies of up to 98 %, with better performance observed for the refractive index measurements across the visible and IR regions. The proposed models could be coupled with other devices for autonomous identification of chemical compounds using a single-wavelength dispersive measurement.

Keywords

• machine learning
• anns
• classification
• deep learning
• materials identification