Answers are grounded in this paper and cite their sources.
An interlaboratory trial was performed to assess the transferability of a near-infrared spectral library for the identification of 15 common solvents. Ten laboratories, recruited from the pharmaceutical industry and academia, participated. Instruments were of both diffraction grating and Fourier transform types using a wide range of sampling accessories which included fibre optic probes fitted with transflectance caps and direct transmission modules. One hundred and seventy three solvent spectra were returned to the organising centre, of which 139 were internal and 34 external to the spectral library. Second-derivative spectra of the solvents over the wavelength range 1136–2000 nm were compared against a master reference library using correlation in wavelength space with the critical value of r set to >0.97 for a positive identification. To compare spectra to the master library, purpose-written software was needed to circumvent problems of incompatibility between the software of different instrument manufacturers. Spectra originally recorded in wavenumber were first converted to equally spaced data points in wavelength using cubic spline interpolation. In the majority of cases high correlations (r > 0.99) were obtained for solvents included in the library and identified correctly. When r ⩽ 0.97 the causes were traced to incorrect labelling, sample contamination or the use of an uncalibrated instrument. Apart from industrial methylated spirits 74 OP, all solvents external to the library were correctly identified as unknowns.
No takes yet. Share an insight, caveat, or question.
Yoon et al. (2000) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: