Fast and non-invasive Raman imaging of material distributions at interfaces

Monitoring the shift in the energy of photons that are inelastically scattered by molecules forms the basis of Raman spectroscopy - an analytical technique widely used for studying the chemical properties of materials. However, the phenomenon of Raman scattering is very weak - bombard 100 million photons onto a material and only 1 of them would spontaneously Raman scatter. Modern day Raman microscopes can acquire a Raman spectrum with appreciable signal-to-noise ratio (SNR) after 1-10 s of accumulation and using a laser dose of 10 mW/µm2. Naturally, this is not a very long time if only a spectral identification of the probed material is required. In the case where a spatial map of the chemical properties of a material is necessary, especially in heterogeneous materials, Raman imaging has to be performed. A typical Raman image made at a resolution of 100 x 100 pixels and at an exposure time of 1 s/pixel would take ≈ 3 hours. To lower this time and facilitate quick non-invasive (minimal laser dose) characterization of materials, one could reduce the laser power and/or the exposure time/pixel, but this would compromise on the SNR of the spectrum. To improve the SNR of the collected Raman spectra, the mathematical technique of Principal Component Analysis (PCA) is applied for denoising, which we term as “Algorithm-improved Confocal Raman Microscopy (ai-CRM). ai-CRM is used to study a variety of systems wherein a limitation on the maximum probed laser dosage results in low SNR of the acquired Raman data. This includes the requirements of fast and non-invasive spatial mapping of 2D materials (low SNR due to low integration time and laser power used), low limit of detection and imaging of analyte molecules on 2D materials (low SNR due to sub-nM analyte concentrations) as well as fast 3D mapping (low SNR due to fast acquisition of a large-sized dataset)