Data treatment in near infrared spectroscopy

Abstract

The near infra red (NIR) takes place between midd l e infrared and visible region of the spectrum. As quartz is tra n s p a rent in near infra red regi o n , t ra n s m i t t a n c e m e a s u rements of liquids can be done using standard s cuvettes and, reflectance measurements of powders can be realised using fiber optics. NIR spectro s c o py is thus a method that re q u i re few or no sample prep a rat i o n . Absorption bands observed in NIR spectra are due to overtones of, mainly, hydrogenic stretching vibrations or combination involving stretching and bending modes of vibration.Those bands are thus broader than in middle infrared and spectra are considerably more complex. Due to this comp l ex i t y, NIR spectro s c o py has soon taken adva n t age of sophisticated calibration techniques and is now using state of the art data treatment. This article will illustrate, first the spectral transformation and then the two current uses of NIR spectroscopy : qualitative discriminant analysis and quantitative applications. ExperimentalQ u a l i t at ive discriminant ap p l i c at i o n : m a n n i t o l , s o r b i t o l P100T, sorbitol P60, sorbitol P20-60, hydroxyethylcellulose, hydroxypropylcellulose (Roquette, Lestrèmes, France).Quantitative application: Paracetamol (Sigma, St. Quentin Fallavier, France) with variable amounts of hydroxypropylcellulose (Roquette) and lactose monohydrate (Sigma).The measurement where done with a BUHLER-NIRVIS s p e c t rometer and a 2 meters optical fi b re. Calculat i o n s wh e re done using NIRCAL 2.0 softwa re (BU H L E R -ANATEC, Uzwil, Switzerland). Spectrum transformationFour different goals can be reached by spectra transformation:Since the last ten years, the use of near-infrared spectroscopy is increasing due to the improvement of instrumentation and software. The calculation capacity of today computers allows a promising future for this technique. According to the complexity of structural attribution of absorption bands, only a chemometric approach is able to solve qualitative and quantitative applications. Depending on spectrophotometers and analytical context, the signal treatment and the calculation strategy may vary. For an identification task, cluster analysis is built on spectral data from only a few wavelengths using disperse spectrophotometer when principal component analysis is applied on overall spectral range with interferometer based apparatus. The main quantitative calculation modes i.e. MLR, PCR and PLS are discussed.

On oxidative degradation of parchment and its non-destructive characterisation and dating

“…For parchment, a direct link between lipid content and collagen degradation has, however, still not been demonstrated. absorptions multiple times across the NIR spectrum, it is necessary to extract the useful information using multivariate chemometric data analysis, such as principal component analysis (PCA) [22][23][24][25] or partial least square calibration (PLS) [21,24,25]. The objective of PCA is to reduce the dimensionality (number of variables) of the dataset, but retain most of the original variability in the data.…”

A few aspects on the importance of particle size enabling proper tabletting

“…It can be concluded that NIR spectroscopy [4,10,11,[20][21][22][23][24][25][26][27][28][29][30][31][32][62][63][64] has now entered a mature stage of development and is currently the most efficient method of performing both qualitative and quantitative analysis; this method can also be used to characterize the particle characteristics, including the sizes of solid samples. Section II.…”

“…As the reflectance of solid samples varies with the concentration, the absorptivity and the scattering coefficient according to the Kubelka Munk theory [22], the NIR spectrum of a solid material depends both on its chemical composition and on its physical properties, such as the particle size and surface characteristics [23][24][25][26].…”

Aplicação da espectroscopia de infravermelho próximo na caracterização de petróleo: simulação de uma unidade de destilação atmosférica.

“…No entanto, as derivadas são também usadas para corrigir deslocamentos de linha base (Chaminade et al, 1998), sendo esta uma das principais causas de sua aplicação, já que ao estar relacionadas à variação da absorbância ao longo do espectro, os espectros que diferem somente no seu nível de linha base apresentarão a mesma derivada. Na figura A.1 ilustra-se um efeito típico da derivação nos espectros NIR.…”

“…As diferentes linhas de regressão são interpretadas como produto da dispersão, e os desvios das linhas como a informação química contida nos espectros (Naes e Isaksson, 1994b et al (1985). (Chaminade et al, 1998) (Meglen, 1991). O NIPALS é um algoritmo eficiente para extrair vetores loadings a partir dos espectros na ordem decrescente de suas contribuições à variância nos espectros de calibração (Haaland e Thomas, 1998).…”