Exploring correlation in infrared spectroscopy

Abstract

Using theoretical and experimental transmission, transflection, and attenuated total reflection (ATR) spectra, we investigated how well corresponding absorbance spectra correlate with true absorbance, defined as the absorption index function multiplied by the wavenumber, using poly(methyl methacrylate) layers on CaF₂, Si, and gold substrates. To improve correlation, the substrate spectrum is often subtracted from the sample spectrum. A typical example is layers on CaF₂, where this approach is sufficient to establish a strong linear correlation. However, in many cases, the substrate spectrum is not a suitable reference for removing unwanted physical contributions, such as substrate-related effects. One such example is layers on Si substrates, where high reflectance causes the spectrum to be dominated by interference fringes. Instead of using the spectrum of an uncoated substrate, one must use the spectrum of a substrate with a non-absorbing layer that has the same refractive index in the transparency region between the MIR and visible spectral regions. For ATR spectra, a simple multiplicative correction based on the wavelength dependence of the penetration depth significantly increases the Pearson coefficient, though not to levels high enough for spectral recognition. To achieve higher accuracy, the Poor Man’s ATR Correction can be employed. For transflection spectra, all relatively simple methods generally fail, and only methods that ultimately determine the optical constant function show promise for success.