The application of multivariate correction in FT-IR quantitative analysis

Due to the birth of Fourier infrared spectrometer, infrared quantitative analysis has been greatly promoted A digital infrared spectrum with high signal-to-noise ratio and wide linear range can be obtained from the Fourier infrared spectrometer Therefore, modern statistical methods can be used to analyze the infrared spectrum The introduction of multivariable correction method has greatly improved the ability of infrared quantitative analysis and qualitative analysis Problems that were previously thought to be unsolvable can now be solved by infrared spectroscopy These statistical methods belong to the category of chemical statistics Because FT-IR contains a large number of spectral data, chemical statistics, which is used to transform spectral data into chemical information, has become a rapidly developing field in recent years Therefore, it is very important for chemists to have a clear understanding of various chemical statistical methods of quantitative FT-IR analysis The multivariate correction methods include classical Zui small two multiplication (CLS), inverse Zui small two multiplication (ILS), Q-matrix, partial Zui small two multiplication (PLS) and basic component regression (PCR) The classical Zui small two multiplication and inverse Zui small two multiplication are often called k-matrix and P-matrix by infrared workers CLS means that the correction mode conforms to beer's law, that is, the spectral absorbance is expressed as a linear function of component concentration Other methods are that concentration is expressed as a linear function of spectral intensity or a linear combination of intensities These correction methods are not limited by the fact that the components in the analyzed spectrum should be consistent with those in the corrected sample, and can be corrected with less or more linear components Therefore, these methods have great flexibility However, there may be experimental deviations in beer's law, which are usually called nonlinearity When the general equation is used to solve the problem, ILS is limited by the number of spectral absorption degrees used, and CLS has the ability of full spectrum analysis ILS uses the reverse form of beer's law, that is, concentration is expressed as a linear function of spectral absorbance Multi point linear regression (MCR) usually refers to ILS method in near infrared literature, because CLS method is rarely used in near infrared analysis PCR and pls are two factor analysis methods Factor analysis is a set of data analysis, it can mathematically complete the separation of multi-component spectrum PCR, also known as target factor analysis, has the ability of full spectrum analysis and flexibility PCR and pls have been widely used in quantitative analysis of infrared spectrum Although multivariate statistical methods have the ability to extract useful information from noisy data, it must be emphasized that they need high quality calibration samples At the beginning of quantitative analysis, how to design experiments is very important 2 The experimental application of quantitative infrared analysis for solution and gas, transmission method is common in Zui However, potassium bromide tablet pressing method, paste method, membrane method and direct determination of solid raw materials are also available Attenuated total reflection (ATR), especially ATR with circulation cell, is generally suitable for the determination of solution, but the selective absorption of solute on ATR element must be considered ATR method can also be used for the analysis of powder samples, textiles, rubber products and plastic products, but it must be ensured that the contact between the samples and ATR elements is good and can be reproduced It is possible to measure the quantitative specular reflection from a thin film on a reflecting surface However, if the thickness of the thin film is the same as the wavelength of the infrared radiation light, the nonlinear effect between the absorption band and the interference fringe becomes very important Diffuse reflectance method starts from FT-IR quantitative analysis of solid samples, which is widely used in near infrared spectroscopy Photoacoustic spectroscopy can be used for quantitative analysis of solids Infrared emission spectrometry has been used in thermal gas analysis and solidification phase quantitative analysis With the development of infrared microscopy, quantitative FT-IR infrared microscopy has become possible, but the influence of stray light must be considered, because stray light can cause obvious nonlinear phenomenon Maris et al Determined the concentration of gaseous light alkanes by ILS method according to the C-H stretching vibration zone Kathryn et al Quantitatively determined the ratio of cellulose to lignin on the surface of cellulose filler by CLS method Charles et al Determined the morphological characteristics of pet by pls Mann et al Analyzed anisaldehyde in o-and m-xylene by orthogonal design Cahn et al Analyzed xylene isomers by PCR and pls Painter et al Used CLS method to make quantitative determination of minerals in coal Haaland et al Determined the content of OH and B2O3 in glass samples by CLS, pls and PCR Hans Rene et al Successfully quantified the crystal form of silicon at low concentration Linn et al Used PLS method to measure the amount of oxygen in the middle of silicon wafer Fuller et al Successfully analyzed six components of decontamination agent by pls Kisner et al Analyzed cholesterol and liquid plasma components by ILs Koenig et al Determined the components in the mixture of polyphenylene oxide and polystyrene by PCR according to the infrared spectrum of the film Lee et al Determined the hydroxyl number in polyester by PLS, and compared the near infrared technology with the mid infrared technology Garcia et al Determined methanol and MTBE in gas oil by pls Lin et al Determined the concentration of NaCl in aqueous solution by PCR In recent years, there are many articles about the analysis and determination of sugar, such as: Dupuy et al Used PLS method to determine glucose, fructose and sucrose in sugar powder; Dupuy et al Compared the advantages and disadvantages of PCR method and PLS method to determine sugar and organic acid in fruit juice; ward et al Used PLS method to determine the concentration of blood glucose; Schindler et al Used PLS method and sequential injection analysis technology to successfully determine glucose in sugar solution Glucose, fructose and sucrose were determined by pls Cadet et al Established a rapid and accurate method for the determination of sucrose content by PCR 3 Preparation and design of samples when preparing analytical samples, it is very important to mix the samples as well as to make the samples representative In order to eliminate the nonlinearity of absorption band height and shape caused by scattering, the size of solid particles should be uniform or smaller than the wavelength of infrared radiation light used Interference fringes produced by parallel reflecting surfaces must be avoided or corrected The sample or sample cell used may also cause quantitative errors When any part of the light leaks through the sample, there will be errors, such as pinholes, cracks, bubbles, uneven particle size or infrared light passing through the edge of the sample The compactness, particle size, temperature, humidity and height of samples are important for diffuse reflectance quantitative analysis These factors may also have great influence on photoacoustic spectrum The temperature should be constant This is especially true for gas samples and samples with interaction between component molecules Because infrared light can cause a significant temperature rise, the temperature balance of the sample in the optical path of the spectrometer may also be important The pressure of the gas sample must be controlled For very dilute gas and solution, more attention should be paid to the adsorption and reaction between the sample and the infrared cell wall The linearity of beer's law is influenced by the interaction between molecules and also by the factors mentioned above The influence of these factors can usually be reduced by diluting the sample Multiple reflections can lead to nonlinearity due to the existence of an unsteady reflectivity of the sample that causes distortion of the absorption band The dispersion of refractive index in the strong absorption band also leads to the nonlinearity of the result This is particularly important in the analysis of materials with high refractive index, such as Ge or Si The scattering phenomenon related to particle size can also lead to absorption band distortion It is also possible to introduce large errors into the impurity components in the sample which have the characteristics of masking spectrum Therefore, these components should also be included in the calibration sample or at least tested The selection of calibration sample is very important for the success of quantitative analysis, especially when the component fraction in the sample is very large Generally, the calibration samples can be prepared independently, and effective statistical experimental design is used to design calibration samples, such as factor design and mixed design In general, in order to get more information of Zui from a limited number of samples, the component concentration is designed to be orthogonal Such a design allows for interaction and nonlinearity in the sample When the calibration sample can not be prepared independently, it is necessary to take random samples from a large number of samples In this case, the general accuracy is relatively low It is difficult to ensure that all the deviation sources and the whole deviation range existing in the unknown sample are also included in the calibration sample If all the independent deviation sources existing in the unknown sample are not included in the correction sample, the quantitative error will be brought to the unknown sample with more independent deviation sources In the design of calibration samples, the target component should be the center of the design, and the concentration range should be larger than that of independent analysis These properties of experimental design must be considered to improve the accuracy of FT-IR analysis and prevent recalibration when the target concentration changes When the concentration range increases, the nonlinearity and the complexity of the model increase For example, with the use of pure components, they usually can not represent the real mixed samples, and in the sample mixture, there are many non-linear sources Therefore, careful design of calibration samples is an important part of FT-IR quantitative analysis of Zui Once the calibration sample is prepared, the problem is focused on the spectrometer In order to ensure that the deviation related to resolution in beer's law does not reduce the accuracy of analysis results, the resolution of spectrometer should be high enough Ramsy believes that the errors caused by the low resolution of the spectrometer will increase with the increase of the absorption intensity In order to obtain high precision of Zui, the resolution value of spectrometer should be much smaller than that of absorption band width of analyte Ramsy believes that when the band-pass of the spectrometer is 1 / 5 of the absorption band, the measurement error is 3% When the band-pass of the spectrometer is only 1 / 2 of the absorption bandwidth, there is 24% measurement error The non-linearity of the response of the spectrometer often exists, and it becomes more important with the increase of absorptivity Although SNR is important, the deviation from other aspects is far greater than that from detector noise The deviation of the same sample in the spectrometer is larger than that of the detector noise alone It is mainly caused by the heterogeneity, thickness, surface, temperature difference and reflection difference of the sample These factors can be avoided as much as possible by reproducibly introducing samples to the same location Liquid cell or liquid ATR is used for liquid sample and gas cell is used for gas sample When the sample is replaced, the cell does not need to be taken out from the spectrometer The influence of these factors can also be reduced by taking the average spectrum after repeatedly measuring the spectrum 5 conclusion at present, more and more quantitative software is produced by FT-IR spectrometer manufacturers Such as specfit quant, PCR, quant, etc However, no correction method is good for Zui, they have their own advantages and disadvantages The correction method is selected according to the type of information obtained Through the detailed understanding of the concept of these methods, we can determine the appropriate correction methods for specific analysis Only by understanding various factors that may affect the accuracy and accuracy of quantitative analysis can we get satisfactory experimental results Author: Xu Xinyuan (Shanghai Institute of drug control 200)