- Количество слайдов: 1
A tool for automated background correction in on-line gradient LC-FTIR Guillermo Quintás, Bernhard Lendl* Inst. of Chemical Technologies and Analytics, Vienna University of Technology, Vienna, Austria. Getreidemarkt 9 -164, A-1060 Vienna, Austria. *: Phone: +43 1 5880115140. Fax: +43 1 5880115199. e-mail: blendl@mail. zserv. tuwien. ac. at In this contribution we present a new automatic procedure for background correction that, for the first time, can be successfully applied in both isocratic and gradient on-line reversed phase LC-FTIR. The proposed approach uses an eluent spectral reference data set and it is based on the assumption that using characteristic absorptions of the eluent. An eluent spectrum can be calculated from the reference matrix. This eluent (background) spectra is then subtracted from the spectra containing the eluting analyte. Based on this approach highly accurate subtraction of the eluent absorption is achieved at minimal computational cost. The method solves two problems simultaneously: the background correction in gradient HPLC and also correct possible changes in the intensity of the absorption spectra of the eluent during analyte elution. ABSTRACT BACKGROUND CORRECTION METHOD BACKGROUND SPECTRAL CHANGES DURING A GRADIENT ELUTION 1. REFERENCE SPECTRAL MATRIX (RSM) MEASUREMENT Water band shifting at different acetonitrile CN stretching (ν 2) relative concentratrions. The developed background spectral subtraction approach it is based on the use of a Reference Spectra Matrix (RSM) of different mobile phase solutions covering a range of compositions. Then, according to the composition of the eluent in every sample spectra, the spectra of the eluent is calculated as a linear combination of two spectra included in the RSM and subtracted from the sample spectrum. The RSM measurement is carried out in practice by using a linear LC gradient. The interval of concentrations of the mobile phase components in the RSM will define the interval of application of the background subtraction procedure in further sample injections. On the other hand, the slope of the linear gradient defines the size of the RSM and thus the accuracy of the whole procedure. 2. CALCULATION OF A REGRESSION MODEL CALIBRATION METHOD: Partial Least Squares SPECTRAL REGION: 2345 – 2145 cm-1 NUMBER OF SPECTRA INCLUDED IN THE CALIBRATION: 5500 (RSM) REFERENCE VALUE: Absorbance at 1639 cm-1 CONCENTRATION RANGE: Water: Acetonitrile (40: 60) up to (1: 99) NUMBER OF LVs: 6 RMSECV: 0. 0009 combination band of both CH 3 bending (ν 3) and C-C stretching (ν 4) 1. The absorption bands from water and acetonitrile change their relative intensity and shape with their relative concentrations, and so, the mobile phase spectra can not be directly calculated from the spectra of the pure compounds. 2. Acetonitrile presents three intense absorption bands at 1442, 1416 and 1377 cm-1 with relative intensities depending on the water relative concentration. On the other hand, it can also be seen two sharp bands at 2252 and 2291 cm -1. 3. The CN stretching band at 2252 cm-1 can be deconvoluted into two Lorentzians separated by 5 cm-1 (2254 and 2251 cm-1). The 2254 cm-1 band can be assigned to free acetonitrile molecules, and 2251 cm-1 band to hydrogen-bonded acetonitrile molecules. During the LC gradient the absolute and relative intensities of the two Lorentzian bands change, but their wavenumbers do not change significantly. This change in the CN stretching band shape is used here to calculate the contribution of the eluent to the final absorption spectra throughout the chromatographic runs. PLS calibration spectral region PLS reference value EXPERIMENTAL RESULTS FROM LC-FTIR INJECTIONS On-line isocratic LC-FTIR The proposed approach performs the background subtraction dynamically, according to Correlation coefficient = 0. 99 Effect of the number of PLS variables used in the calibration model on the noise obtained after background correction in a LC-FTIR gradient elution. defined spectral features of every spectrum from the sample data set so, increasing the robustness of the technique as long as possible changes in the composition of the mobile phase are automatically corrected. To estimate whether the proposed subtraction procedure provide accurate results the noise at three different solvent characteristic wavenumbers in the extracted chromatograms from a blank injection were calculated. Previously measured vanillic acid spectra Wavenumber 3. LC-FTIR MATRIX ANALYSIS USING THE PLS MODEL The PLS analysis of the LC-FTIR matrix from a sample injection provides, for every spectra included in the sample matrix, a reference value, CSAMPLE, i , the calculated absorbance at 1639 cm-1, directly related to the water concentration and so, to the mobile phase composition. Spectra extracted using the proposed method of eluent subtraction during the elution of a 2 mg ml-1 Caffeine standard. LC conditions: 60% acetonitrile: 40% water (1% HAc). (cm-1) 171 2 164 3 144 2 1280 NOISE 3. 1 Noise values * 104 7. 6 4. 7 2. 3 The measured noise levels are statistically comparable to those obtained using a constant reference background. 4. BACKGROUND SPECTRA CALCULATION Spectra extracted using the proposed method of eluent subtraction during the elution of a 2 mg ml-1 Vanillic acid standard. LC conditions: 70% acetonitrile: 30% water (1% HAc). Using the reference value previously calculated, for every spectra included in the sample matrix the position of the minimum of the absolute value of the difference between CSAMPLE, i and the absorbance at 1639 cm-1 (CREFERENCE) of every spectra from the RSM is calculated. p-2 |(Creference-Csample)| p+2 Correlation coefficient = 0. 93 Two pesticides, Diuron and Atrazine, were injected using the gradient indicated in the figure. Correlation coefficient = 0. 94 The accuracy of the subtraction process was evaluated by comparing the extracted spectra of both pesticides at the peak apex to pesticide reference spectra previously measured, obtaining correlation factors of 0. 94 and 0. 93 for Atrazine and Diuron, respectively. On the other hand, in the following figure it is shown the extracted chromatogram at 1547 cm-1. Minimum = p p+1 p-1 On-line gradient LC-FTIR p time (min) % CH 3 CN 0 60 40 10 Number of the reference spectra included in the RSM % Water 1 99 Results obtained from the background correction of a blank gradient elution in the same range of concentrations. Finally, the spectra to be subtracted is calculated as a linear combination of two spectra form the RSM: the corresponding to the minimum (spectra number p) and the second closest spectra which can be the spectra number p+1 or p-1. Csample Cp-1 Csample Cp β α Raw spectral matrix Cp Cp+1 α MATLAB GUI FILE β p-1 for Csample < Cp p+1 for Csample > Cp 5. SPECTRAL SUBTRACTION The last step of the procedure involves the direct subtraction of the calculated spectra from the original sample spectra. Spectral matrix after background correction Top: Extracted FTIR spectra of both pesticides at the peak apex obtained after removing the eluent contribution. Bottom: Extracted chromatogram at 1547 cm-1. Snapshot of the Matlab GUI file developed to facilitate the use of the proposed procedure. It includes three additional univariated procedures to perform background correction in online LC-FTIR, also based on the use of a RSM. FTIR measurement conditions: Bruker IFS 55 spectrometer equipped with a MCT detector, a Globar source and a KBr beamsplitter was employed for spectral measurements, also using a flow cell with Ca. F 2 windows, an optical pathlength of 25 m, 3 mm broad and a length of 32 mm. The spectra were recorded by co addition of 25 scans with a spectral resolution of 8 cm-1 at a mirror velocity of 180 k. Hz He. Ne frequency also using an aperture of 8 mm for the IR beam. Prior to data acquisition a background spectrum of the sample compartment after removing the flow cell was acquired. LC measurement conditions: The LC system consisted in a Waters (Waters, Milford, USA) 7100 quaternary pump with a six-port injection valve (Rehodyne, Bensheim, Germany) equipped with a 12. 5 l injection loop. Chromatographic separation was achieved by a C 18 reversed-phase column (Nucleosil 100 RP C-18 column, 150 mm × 2. 1 mm, 3 m, VDS Optilab, Berlin, Germany) using of acetonitrile and H 2 O (1% v/v acetic acid) as mobile phase components at 25ºC also using a 150 L min-1 flow-rate. CONCLUSIONS Results from this study show the potential use of the proposed approach to perform background correction in continuous liquid flow systems. Results obtained from the correction of the raw data from isocratic and gradient LC injections were used to verify of its capability to correct background signals in the presence of changing eluent compositions, as well as during the analyte elution. Furthermore, it is not needed complex data pre-treatment, prior knowledge, standards or reference spectra of the analytes and there is no user interaction thus increasing the repeatability of the results. Moreover, the accuracy of the whole subtraction procedure can be easily improved without any additional modifications of the method by using larger reference spectral data matrix. The use of PLS extends the applicability of the procedure to more complex systems in which the calculation of the composition can not be carried out using univariated approaches.