Li et al.: Journal of AOAC International Vol. 98, No. 1, 2015 183 RESIDUES AND TRACE ELEMENTS
Application of Near Infrared Spectroscopy Coupled with Fluidized Bed Enrichment and Chemometrics to Detect Low Concentration of β-Naphthalenesulfonic Acid Wei Li, Xuan Zhang, Kaiyi Zheng, and Yiping Du1
Shanghai Key Laboratory of Functional Materials Chemistry; and Research Center of Analysis and Test, East China University of Science and Technology, Meilong Rd 130, Shanghai 200237, People’s Republic of China Peng Cap, Tao Sui, and Jinpei Geng Yantai Entry-Exit Inspection and Quarantine Bureau, Beima Rd 66, Yantai 264000, People’s Republic of China
A fluidized bed enrichment technique was developed to improve sensitivity of near infrared (NIR) spectroscopy with features of rapidness and large volume solution. D301 resin was used as an adsorption material to preconcentrate β-naphthalenesulfonic acid in solutions in a concentration range of 2.0–100.0 μg/mL, and NIR spectra were measured directly relative to the β-naphthalenesulfonic acid adsorbed on the material. An improved partial least squares (PLS) model was attained with the aid of multiplicative scatter correction pretreatment and stability competitive adaptive reweighted sampling wavenumber selection method. The root mean square error of cross validation was 1.87 μg/mL at PLS factor of 7. An independent test set was used to assess the model, with the relative error (RE) in an acceptable range of 0.46 to 10.03% and mean RE of 3.72%. This study confirmed the viability of the proposed method for the measurement of a low content of β-naphthalenesulfonic acid in water.
β
-naphthalenesulfonic acid is an important dye intermediate with low biodegradability, high polarity, and high toxicity. It is widely used as a textile auxiliary chemical in the dye industry. In the environment it is normally present at a low concentration level, but it is still a potential hazard in particular surface and ground waters that are related to the quality of drinking water. Although there are various techniques that are applicable for detecting low content of β-naphthalenesulfonic acid, most of them are based on HPLC with either UV or fluorescence detection (1–3). Regrettably, these methods usually need complex sample pretreatment procedures that are time-consuming and costly. Also, these methods are not applicable to field detection. Near IR (NIR) spectroscopy is a powerful analytical method that is widely adopted in various fields, such as agriculture (4–6), petrochemicals (7), pharmaceuticals (8–11), and food engineering (4, 12–14). Compared with other analytical techniques, it has the advantages of being nondestructive and eco-friendly, having fast response time, not requiring complex Received December 5, 2012. Accepted by AK March 4, 2013. 1 Corresponding author’s e-mail: [email protected] DOI: 10.5740/jaoacint.12-436
sample pretreatments, and the possibility to simultaneously detect several analytes in a same sample. NIR spectroscopy is easy to miniaturize. It is a method that seems best suited to batch sample analysis and field detection. As a consequence of the low molar absorptivity of NIR signals and the fact that spectra typically contain weak, non-specific, and overlapping peaks, this technique is not suitable for analysis of low levels of analyte or trace analysis. Therefore, establishing new analysis methods that not only utilize its advantages but also make up for its deficiencies and extend its use is important to the development of NIR spectroscopy. In recent years, some sample pre-enrichment techniques were investigated to enhance the sensitivity of NIR spectroscopy determination (15–17), In our previous work, an online enrichment device was developed (18) in which chelating resin 2+ D401 was used as an adsorbent to enrich Cu in water. After the enrichment procedure, the diffuse reflection NIR (DRNIR) 2+ spectra of the adsorbed Cu were captured from the bottom of the device. The enrichment device was designed based on a fluidized bed idea, which is rapid, accurate, and practical as large volume of water samples can be analyzed with this method and no elution step is involved. It is well known that chemometrics plays an indispensable role in NIR analysis, and it has been proved to be an effective process to improve the sensitivity of NIR techniques. Spectrum pretreatment and wavenumber selection are commonly used strategies to ameliorate the models (19–25). Cai et al. (19) developed a variable method based on the principle of Monte Carlo (MC) and uninformative variable elimination (UVE) for selecting important variables. This method built a large number of models with randomly selected calibration samples at first, and then each variable was evaluated with a stability computed by regression coefficients divided by its SD. Variables with poor stability were known as uninformative variables and were eliminated. After being applied to analyze nicotine and sugar contents in tobacco samples, it was proved that the proposed method was able to select important wavelengths from the NIR spectra, which made the prediction more robust and accurate in quantitative analysis. Wei et al. (23) used competitive adaptive reweighted sampling (CARS) and MC-UVE to select the effective wavelengths for the determination of total acid of vinegar. Each program ran 1000 times to get the most typical variables. Then least squares regression models were built on the selected wavelengths. The root mean square errors of prediction (RMSEP) of full spectrum, CARS, and MC-UVE models were 0.21 g/100 mL, 0.13 g/100 mL, and 0.18 g/100 mL, respectively.
184 Li et al.: Journal of AOAC International Vol. 98, No. 1, 2015 It was concluded that NIR spectroscopy coupled with the CARS method seems to be rapid and effective for determining total acid of vinegar. Quite recently, we proposed a modified CARS algorithm called stability competitive adaptive reweighted sampling (SCARS; 25), which considers the stability as important information of each variable. In SCARS, a variable is selected by an index of stability that is defined as the absolute value of the regression coefficient divided by its SD. The performance of the proposed algorithm was evaluated by three NIR datasets: tobacco, corn, and glucose. The results showed that SCARS can select the least variables and supply the least root mean square error of cross validation (RMSECV) and latent variable (LV) number of the partial least squares (PLS) model compared with methods of Moving Window PLS, MC-UVE, and CARS. In this work, a method was developed to investigate the feasibility of determining a low content of β-naphthalenesulfonic acid in water utilizing DRNIR spectroscopy (DRNIRS) associated with sample preconcentration by a fluidized bed enrichment device. Chemometric methods including spectrum pretreatment and wavenumber selection were used to fulfill the task of optimizing the PLS models. Experimental Reagents and Adsorbent All chemical reagents used were of analytical purity grade. β-naphthalenesulfonic acid monohydrate was purchased from Tokyo Chemical Industry Co. Ltd (Tokyo, Japan). All 51 standard solutions of β-naphthalenesulfonic acid in a concentration range of 2.0–100.0 μg/mL were prepared by dissolving β-naphthalenesulfonic acid monohydrate powder and then diluting with ultra-pure water (Sartorius Arium 611D1 Germany, 18.2MΩ*cm) Macroporous weakly basic anion exchanger resin D301 was selected as the sorbent material (26) and was supplied by Shanghai Huazhen Science and Technology Co. Ltd of East China University of Science and Technology (Shanghai, China) with bead size of 0.3–1.2 mm and moisture content of 50–60%. Prior to use, the resin was repeatedly washed with hot water at 50–60°C until the supernatant liquid was not brownish and there were no bubbles. Then the resin was successively soaked with 1 M NaOH and 1 M HCl solution, in turn, and washed with ultrapure water for neutralization after each infusion. Finally, it was infiltrated with 1 M NaOH solution for 2 h and washed with ultrapure water until pH was
Application of Near Infrared Spectroscopy Coupled with Fluidized Bed Enrichment and Chemometrics to Detect Low Concentration of β-Naphthalenesulfonic Acid Wei Li, Xuan Zhang, Kaiyi Zheng, and Yiping Du1
Shanghai Key Laboratory of Functional Materials Chemistry; and Research Center of Analysis and Test, East China University of Science and Technology, Meilong Rd 130, Shanghai 200237, People’s Republic of China Peng Cap, Tao Sui, and Jinpei Geng Yantai Entry-Exit Inspection and Quarantine Bureau, Beima Rd 66, Yantai 264000, People’s Republic of China
A fluidized bed enrichment technique was developed to improve sensitivity of near infrared (NIR) spectroscopy with features of rapidness and large volume solution. D301 resin was used as an adsorption material to preconcentrate β-naphthalenesulfonic acid in solutions in a concentration range of 2.0–100.0 μg/mL, and NIR spectra were measured directly relative to the β-naphthalenesulfonic acid adsorbed on the material. An improved partial least squares (PLS) model was attained with the aid of multiplicative scatter correction pretreatment and stability competitive adaptive reweighted sampling wavenumber selection method. The root mean square error of cross validation was 1.87 μg/mL at PLS factor of 7. An independent test set was used to assess the model, with the relative error (RE) in an acceptable range of 0.46 to 10.03% and mean RE of 3.72%. This study confirmed the viability of the proposed method for the measurement of a low content of β-naphthalenesulfonic acid in water.
β
-naphthalenesulfonic acid is an important dye intermediate with low biodegradability, high polarity, and high toxicity. It is widely used as a textile auxiliary chemical in the dye industry. In the environment it is normally present at a low concentration level, but it is still a potential hazard in particular surface and ground waters that are related to the quality of drinking water. Although there are various techniques that are applicable for detecting low content of β-naphthalenesulfonic acid, most of them are based on HPLC with either UV or fluorescence detection (1–3). Regrettably, these methods usually need complex sample pretreatment procedures that are time-consuming and costly. Also, these methods are not applicable to field detection. Near IR (NIR) spectroscopy is a powerful analytical method that is widely adopted in various fields, such as agriculture (4–6), petrochemicals (7), pharmaceuticals (8–11), and food engineering (4, 12–14). Compared with other analytical techniques, it has the advantages of being nondestructive and eco-friendly, having fast response time, not requiring complex Received December 5, 2012. Accepted by AK March 4, 2013. 1 Corresponding author’s e-mail: [email protected] DOI: 10.5740/jaoacint.12-436
sample pretreatments, and the possibility to simultaneously detect several analytes in a same sample. NIR spectroscopy is easy to miniaturize. It is a method that seems best suited to batch sample analysis and field detection. As a consequence of the low molar absorptivity of NIR signals and the fact that spectra typically contain weak, non-specific, and overlapping peaks, this technique is not suitable for analysis of low levels of analyte or trace analysis. Therefore, establishing new analysis methods that not only utilize its advantages but also make up for its deficiencies and extend its use is important to the development of NIR spectroscopy. In recent years, some sample pre-enrichment techniques were investigated to enhance the sensitivity of NIR spectroscopy determination (15–17), In our previous work, an online enrichment device was developed (18) in which chelating resin 2+ D401 was used as an adsorbent to enrich Cu in water. After the enrichment procedure, the diffuse reflection NIR (DRNIR) 2+ spectra of the adsorbed Cu were captured from the bottom of the device. The enrichment device was designed based on a fluidized bed idea, which is rapid, accurate, and practical as large volume of water samples can be analyzed with this method and no elution step is involved. It is well known that chemometrics plays an indispensable role in NIR analysis, and it has been proved to be an effective process to improve the sensitivity of NIR techniques. Spectrum pretreatment and wavenumber selection are commonly used strategies to ameliorate the models (19–25). Cai et al. (19) developed a variable method based on the principle of Monte Carlo (MC) and uninformative variable elimination (UVE) for selecting important variables. This method built a large number of models with randomly selected calibration samples at first, and then each variable was evaluated with a stability computed by regression coefficients divided by its SD. Variables with poor stability were known as uninformative variables and were eliminated. After being applied to analyze nicotine and sugar contents in tobacco samples, it was proved that the proposed method was able to select important wavelengths from the NIR spectra, which made the prediction more robust and accurate in quantitative analysis. Wei et al. (23) used competitive adaptive reweighted sampling (CARS) and MC-UVE to select the effective wavelengths for the determination of total acid of vinegar. Each program ran 1000 times to get the most typical variables. Then least squares regression models were built on the selected wavelengths. The root mean square errors of prediction (RMSEP) of full spectrum, CARS, and MC-UVE models were 0.21 g/100 mL, 0.13 g/100 mL, and 0.18 g/100 mL, respectively.
184 Li et al.: Journal of AOAC International Vol. 98, No. 1, 2015 It was concluded that NIR spectroscopy coupled with the CARS method seems to be rapid and effective for determining total acid of vinegar. Quite recently, we proposed a modified CARS algorithm called stability competitive adaptive reweighted sampling (SCARS; 25), which considers the stability as important information of each variable. In SCARS, a variable is selected by an index of stability that is defined as the absolute value of the regression coefficient divided by its SD. The performance of the proposed algorithm was evaluated by three NIR datasets: tobacco, corn, and glucose. The results showed that SCARS can select the least variables and supply the least root mean square error of cross validation (RMSECV) and latent variable (LV) number of the partial least squares (PLS) model compared with methods of Moving Window PLS, MC-UVE, and CARS. In this work, a method was developed to investigate the feasibility of determining a low content of β-naphthalenesulfonic acid in water utilizing DRNIR spectroscopy (DRNIRS) associated with sample preconcentration by a fluidized bed enrichment device. Chemometric methods including spectrum pretreatment and wavenumber selection were used to fulfill the task of optimizing the PLS models. Experimental Reagents and Adsorbent All chemical reagents used were of analytical purity grade. β-naphthalenesulfonic acid monohydrate was purchased from Tokyo Chemical Industry Co. Ltd (Tokyo, Japan). All 51 standard solutions of β-naphthalenesulfonic acid in a concentration range of 2.0–100.0 μg/mL were prepared by dissolving β-naphthalenesulfonic acid monohydrate powder and then diluting with ultra-pure water (Sartorius Arium 611D1 Germany, 18.2MΩ*cm) Macroporous weakly basic anion exchanger resin D301 was selected as the sorbent material (26) and was supplied by Shanghai Huazhen Science and Technology Co. Ltd of East China University of Science and Technology (Shanghai, China) with bead size of 0.3–1.2 mm and moisture content of 50–60%. Prior to use, the resin was repeatedly washed with hot water at 50–60°C until the supernatant liquid was not brownish and there were no bubbles. Then the resin was successively soaked with 1 M NaOH and 1 M HCl solution, in turn, and washed with ultrapure water for neutralization after each infusion. Finally, it was infiltrated with 1 M NaOH solution for 2 h and washed with ultrapure water until pH was
