Assess the ability of detecting low concentration analyte with near-infrared spectroscopy based on dynamic enrichment Xuan Zhang a , Yiping Du a, ⁎, Peijin Tong a , Wei Li a , Jibran Iqbal b , Ting Wu a , Huilian Hu a , Weibing Zhang a a Shanghai Key Laboratory of Functional Materials Chemistry, Research Center of Analysis and Test, East China University of Science and Technology, Meilong Rd 130, Shanghai 200237, China b Interdisciplinary Research Centre in Biomedical Materials, COMSATS Institute of Information Technology, Lahore, Pakistan abstract article info Article history: Received 26 November 2013 Received in revised form 12 March 2014 Accepted 15 March 2014 Available online 21 March 2014 Keywords: Kinetic modeling Fluidized bed Liquid film diffusion Near-infrared diffuse reflectance spectroscopy (NIRDRS) Detection limit Dynamic adsorption The ability to detect low concentration analyte with near-infrared diffuse reflectance spectroscopy (NIRDRS) based on dynamic enrichment method has been assessed. A special design of fluidized bed enrichment device was used to enrich a large volume of analyte's dilute solution before spectrum detection, in order to improve the detection sensitivity of NIRDRS. A kinetic model, which considers the mass transfer with liquid film diffusion, has been used to characterize the adsorption process in this device. The developed model agreed with the exper- imental results very well in a wide range of the influent flow rate (F) and solution concentration (C 0 ). Based on this model, the lowest detectable concentration was estimated, at the same time the effects of liquid fluid rate and operation time on this value were also investigated. Meanwhile, a comparison between this model and the static adsorption model was made. Furthermore, a series of carbaryl aqueous solutions at different concentra- tions were treated with the enrichment device to verify the estimated lowest concentration. This study reveals that the specially designed fluidized bed device is able to enrich enough amount of analyte in a quite short time, and based on this dynamic adsorption model, it is possible to detect analyte in solution quantitatively at ppm-level by NIRDRS. © 2014 Elsevier B.V. All rights reserved. 1. Introduction Near-infrared spectroscopy (NIRS) has proven to be a powerful analytical tool and has been widely applied in many fields, such as agriculture [1,2], food [3,4], pharmaceutical [5,6], routine chemical analysis [7–9] and the petroleum industry [10,11]. It has attracted grow- ing attention due to its advantages of rapidness, non-destruction, simple-pretreatment, few-reagent and low cost. However, the low molar absorptivity of near infrared signals which considerably reduces its sensitivity, makes common NIR technique impossible for micro- and trace level analysis. Therefore, it is very meaningful to improve the sensitivity of NIRS for low concentration analysis. In recent years, aiming to improve the sensitivity of NIR technique, some researchers have reported the ppm (mg/L) level detection [12–24]. Among these work, adsorption technique as an effective pre- concentration process was often introduced into NIR technique to im- prove the sensitivity. In this combination method, solid adsorbents are employed to enrich analytes selectively from dilute solutions, and then the solid materials with adsorbed analytes are directly measured by NIR without any elution process. The advantages of this method are obvious. It simplifies the enrichment process, reduces or even elim- inates the interference of water, avoids the loss of analytes and aban- dons the use of toxic reagents that are commonly encountered during elution process. Meanwhile, the work using this combination method usually show good results with very small values of root mean square error of cross validation (RMSECV) and/or root mean square error of prediction (RMSEP), which seem to be satisfactory for quantitative analysis. However, if we take a closer look at the RMSECV and RMSEP values, we will find that they are not small enough comparing with the concentrations of the analyte to build a satisfactory model. If we evaluate these work based on the value of relative error (RE), which is calculated by the RMSECVs or RMSEPs divided by the mean concentra- tion of sample solutions, some work do not show a good performance with large relative errors (normally larger than 15%). More details about this conclusion has been discussed and reported in our previous work [25], it is unnecessary to go into a lot of details here. As mentioned in our previous research [25], the reason of the large errors is the low content of the analyte adsorbed on the adsorbent, i.e. mass of adsorbed analyte per unit weight of adsorbent (q t ), which are directly related to the NIR spectra. In some work, several or tens of milliliter solutions were used, so it was hard to obtain a high q t , e.g. 20 mL of 2 mg/L analyte solution only contains 0.04 mg analyte, and if it adsorbed on the 2 g adsorbent completely, q t is only 0.02 mg/g. Thus, increasing the concentration of the adsorbed phase q t (mg/g) is particularly important to improve the sensitivity of the method based on enrichment. There are two important factors influencing the value of q t , which are mass of adsorbent and volume of sample solution denoted by m and V. Too small m will affect the spectral accuracy, there- fore, larger V should be used to ensure that more amounts of analytes could be adsorbed on the adsorbent. Chemometrics and Intelligent Laboratory Systems 134 (2014) 58–66 http://dx.doi.org/10.1016/j.chemolab.2014.03.008 0169-7439/© 2014 Elsevier B.V. All rights reserved. Contents lists available at ScienceDirect Chemometrics and Intelligent Laboratory Systems journal homepage: www.elsevier.com/locate/chemolab