Research On The Analysis Method Of Characteristic Bands Of Near Infrared Spectroscopy

Near infrared?NIR?spectroscopy has been generally used in all fields as a fast and green analysis method.Bands in NIR are overlapped and the absorbance intensity are week,which make the interpretation of NIR complicated.Chemometrics based variable selection needs to repeatedly do in different object.The selected variable between different chemometrics method have low reproducibility and repeatability.we could not give a proper explanation.NIR characteristic variable selection based on molecule structure not only provide theoretical knowledge to the use of NIR,but also settle substruction to establish the causality between molecule structure property and NIR spectra.And then to improve the interpretive of chemical compound NIR spectra and improve the quantitative and qualitative prediction accuracy of NIR model.Based on the above issues,this paper is to explore the methods of NIR characteristic variable selection method and the scientific nature and reliability of the method.Mainly in virtue of derivates spectrum,tow dimensional correlation spectroscopy,differential spectrum.The mainly research contents are as fellows:In aspect of NIR absorption theory,we using deuterated solvents as object,compared the NIR absorption frequency difference of deuterated solvents and corresponding non-deuterated solvents by anharmonicity theory.The results indicated that substances have NIR absorption when its IR absorption is above 2000cm-1.This conclusion has enriched the basic theory of NIR.In aspect of NIR spectra interpretation:?1?Established the NIR collecting method to spectra interpretation.The spectra collected using deuterated chloroform?CDCl₃?and carbon tetrachloride?CCl₄?as solvent with transmittance spectroscopy was chosed.?2?Established the concentration dependent two dimensional correlation analysis method.The influence of number and concentration region of the sample on tow dimensional synchronous spectra was discussed using CCl₄ as solvent.Finally select 9 samples between 0.05?0.3mol/L for n-hexane simple system and 9 samples between 0.015?0.153mol/L for TCM chemical composition.?3?Established the difference spectra based NIR spectra interpretation method.The results indicated that the energy of sp2 hybrid mode in benzene was greater than the sp3 hybrid mode in cyclonexane,which make the C-H bands intensity of benzene was greater than cyclohexane;The band absorption intensity in 2310 nm has a positive correlation with the number of substituted methyl,methylbenzene:dimethylbenzene:mesitylene?1:2:3?,the more substituted methyl the greater the absorption intensity is;the PLS mode multivariate detection limit of benzene,methylbenzene and dimethylbenzene in 2330 nm were gradually decrease,which illuminated the relationship between absorption intensity,emphasized the importance of variable selection;The framework C-H absorption intensity of methy substituted benzene has a negative correlation with the molecule symmetry,the better the symmetry,the strong the absorption intensity;The framework C-H absorption intensity of methy substituted benzene in 1138 nm and 1670 nm are moved to high wavelength with a order of methylbenzeneethybenzene<o-xylene<m-xylene<p-xylene<mesitylene;the difference between o-xylene,m-xylene and p-xylene are in 2460 nm;O-H substituted benzene have a strong single peak in 1410 nm,which has great difference with the second combination frequency of alkyl.In aspect of method validation:?1?NIR spectra of phenylpropenoids,alkaloids,terpenoids in CDCl₃ were collected.Discriminate model was established by partial least square discriminate?PLS-DA?.The identification rate of external validation after OSC preprocessing was 100%.It was suggested that the spectra collected include efficiency information for discrimination,can be used for characteristic bands assignment.?2?Characteristic bands of phenylpropenoids,alkaloids,terpenoids in CDCl₃ were select by derivates spectra,which can be a guidance for NIR spectra interpretation.?3?The feature bands of n-hexane,cyclohexane,benzene,methylbenzene were investigated by concentration dependent two dimensional correlation analysis.The feature band position from the synchronous spectra were the same to the literature reported.It suggests that the feature bands from concentration dependent synchronous spectra is reliable.The linear relationship between absorption and concentration of menthol in 1410nm was discussed.The correlation coefficient was 0.998,which indicate that NIR spectra has a linear relationship with absorption intensity and provided a data support to NIR quantitative analysis.?4?Feature bands of tanshinone ?A and cryptotanshinone were investigated under different conservation in CDCl₃ respectively.Then using the feature bands selected to establish the PLS quantitative model of tanshinone ?A and cryptotanshinone in salvia extract.Compare to the full region PLS model,the RPD of cryptotanshinone and tanshinone ?A were increased from 6.5 and 6.1 to 7.3 and 6.2 repsectivly.It indicate that the feature band selected under different concentration is applicability in complex system and more interpretive and robustness.