Printed Disposable Colorimetric Array for Metal Ion Discrimination M. Ariza-Avidad,† A. Salinas-Castillo,† M. P. Cuéllar,‡ M. Agudo-Acemel,† M. C. Pegalajar,‡ and L. F. Capitán-Vallvey*,† †
ECsens. Department of Analytical Chemistry, Faculty of Sciences, University of Granada, Campus Fuentenueva, Granada 18071 Granada Spain ‡ Department of Computer Science and Artiﬁcial Intelligence, E.T.S. Ingenierías Informática y de Telecomunicación, University of Granada, C/Periodista Daniel Saucedo Aranda s/n, Granada 18071, Spain S Supporting Information *
ABSTRACT: One of the main limiting factors in optical sensing arrays is the reproducibility in the preparation, typically by spin coating and drop casting techniques, which produce membranes that are not fully homogeneous. In this paper, we increase the discriminatory power of colorimetric arrays by increasing the reproducibility in the preparation by inkjet printing and measuring the color from the image of the array acquired by a digital camera, using the H coordinate of the HSV color space as the analytical parameter, which produces robust and precise measurements. A disposable 31 mm × 19 mm nylon membrane with 35 sensing areas with 7 commercial chromogenic reagents makes it possible to identify 13 metal ions and to determine mixtures with up to 5 ions using a two-stage neural network approach with higher accuracy than with previous approaches.
microprocessors can be programmed to carry out the image processing and mathematical data treatment. Diﬀerent approaches have been described for optical electronic tongues, 5−7 but with regard to metal ion discrimination alone, three diﬀerent formats have been devised: membranes,8,9microtiter plates,8,10−15 and paper-based sensing devices.16−18 The microtiter plate format arranges the reagents in the wells either immobilized as a membrane at the bottom10,11,15 or as a solution placed in vials that are arranged in an array before acquiring the analytical signal.12 That format needs only a small volume of sample, even at submicroliter level,14 although in the usual application with water analysis it is not the main problem. Membrane and paper-based devices oﬀer the additional advantage of preconcentration, lowering the detection limit. For instance, the arrangement of metal ion reagents in ormosils coated on a cellulose acetate/nitrate membrane has been proposed for the discrimination of eight metal ions with detection limits of 50 nM.9 Diﬀerent types of paper-based devices have been described: Hossain and Brennan16 used a ﬂower design with wax-printed hydrophobic channels and inkjet-printed detection areas in which the stem is the sampling area. Feng et al.17 described an eight detection area wheel design paper device with a sampling area in the middle that improves the preconcentration, including a water absorbent cell on each detection area and a water absorbent resin placed at the end of each hydrophilic channel. In this way it is possible to
he identiﬁcation and determination of metallic ions, mainly multivalent transition metals with potential toxicity or ecotoxicity,1 in the lab is a well-established subject that uses an ample panoply of analytical techniques, including inductively coupled plasma/atomic emission spectrometry (ICP-AES), inductively coupled plasma/mass spectrometry (ICP-MS), atomic absorption spectroscopy (AAS), and atomic ﬂuorescence spectroscopy (AFS). However, the persistence of these metals in both natural and contaminated environments and their involvement in the development of human cancer and neurodegenerative diseases, encourages the development of fast, cheap procedures that can be performed in situ by untrained personnel, able to manage a large number of samples that provide qualitative and quantitative information on metal ions, particularly in drinking and natural waters and wastewaters.2 One of the most interesting approaches for rapid multianalyte analysis is a sensor arrays designnot those based on speciﬁc receptors because of selectivity problemsbut rather those based on nonspeciﬁc sensors producing a set of high dimensional analytical signals. The processing of signals through advanced mathematical procedures of pattern recognition and/or multivariate analysis3,4 makes it possible to obtain qualitative and quantitative information. From the existing sensing array schemes, optical arrays (also known as optical electronic tongues) oﬀer a reasonable combination of discriminatory power, sensitivity and simplicity to convey the requirements of these analytical devices. These are mainly combined with imaging devices for simultaneous signal acquisition, such as scanners and digital cameras and more recently, with digital cameras integrated into portable devices such as tablets and smartphones, because the built-in © 2014 American Chemical Society
Received: April 21, 2014 Accepted: August 4, 2014 Published: August 4, 2014 8634
dx.doi.org/10.1021/ac501670f | Anal. Chem. 2014, 86, 8634−8641
ﬁlter through a sample volume as high as 800 μL. A diﬀerent design presented by Feng et al.18 uses a 3 × 3 membrane array prepared as above with a wax-printing technique for the pattern and by drop coating the reagents. The location of a piece of ﬁlter paper under the array makes it possible for analytes (600 μL) to pass through the spots enriched with the metals. Diﬀerent chemistries, both colorimetric and luminescent, have been employed for the described devices. Conventional colorimetric reagents both alone9 or in combination with an enzymatic test like the β-galactosidase-based test16 have been used for the discrimination of 7−8 metal ions. Additionally, some luminescent reagents have been employed on the basis of the combination of diﬀerent coordination chemistries and signal transduction schemes, such as turn-on, turn-oﬀ, and ratiometric schemes to generate discriminatory data for metal classiﬁcation.11,13,14 In other cases, the array is composed of a family of reagents that combine receptor and ﬂuorophore in the molecule to introduce subtle diﬀerences to increase the discriminatory power. Examples include reagents based on di2-picolyamine derivatives as receptors and BODIPY as the ﬂuorophore12 or 8-hydroxyquinoline as the receptor and diﬀerent oligoﬂuorenes as the ﬂuorophore.15 In general, diﬀerent imaging techniques have been used to obtain analytical signals ranging from a conventional scanner16 to a ﬂuorescence scanner,13−15 a CCD camera for lifetime measurements,10,11 and a digital camera.12,18 One of the weak points of these devices is the reproducibility of the preparation because the most common way to deposit the reagents in the detection area is by drop casting, which increases the inhomogeneity of the recognition area, reducing the reliability of the device and the precision in quantitative determinations. In this paper, we study the preparation of colorimetric arrays for metal ions based on conventional metal-ion reagents emphasizing the reproducibility of the preparation by printing techniques and the use of the H coordinate from HSV color space as the analytical signal19 in order to increase the discriminatory power of the array and the precision of the determinations. The membrane array was prepared with a drop-on-demand printing method based on piezoelectric inkjet technology, a reliable system for membrane production taking into account the low volume of chemical ink required, the disposable piezo inkjet used, and the low cost of the array. In this case, the colorimetric response of any membrane in the sensor array is a sigmoid-shape nonlinear function of the color coordinate with respect to the metal concentration.20 Thus, we need to use a multivariate nonlinear mathematical technique able to provide a calibration model for the whole array. In this work, we use artiﬁcial neural networks (ANN)21 because of their multivariate and nonlinear nature and their features such as noise tolerance and generalization properties for the calibration data. These types of techniques have been widely and successfully used in similar tasks. 22 More speciﬁcally, ANNs are used in our approach to identify up to 13 metal ion mixtures in solution and to estimate the corresponding metal concentration from a sensing array composed of seven sensing membranes that use the hue color feature as an analytical parameter.
[(3-carboxy-5-methyl-4-oxo-2,5-cyclohexadien-1-ylidene)(2,6dichloro-3-sulfophenyl)methyl]-2-hydroxy-3-methylbenzoic acid trisodium salt (Chromazurol S, CS); 3-(2-pyridyl)-5,6diphenyl-1,2,4-triazine-4′,4″-disulfonic acid sodium salt (Ferrozine, FER); 1-(2-hydroxy-1-naphthylazo)-2-naphthol-4-sulfonic acid sodium salt (Calcon, CAL); 1-(2-pyridylazo)-2-naphthol (PAN); 2-carboxy-2′-hydroxy-5′-sulfoformazyl-benzene monosodium salt (Zincon, ZIN); pyrocatecholsulfonphthalein (Pyrocatechol Violet, VP); 1-(4-nitrophenyl)-3-(4phenylazophenyl)triazene (Cadion, CAD); diphenylcarbazide (DPC), dimethylglyoxime (DMG), and diphenylthiocarbazone (Dithizone, DTZ). The chemical inks were prepared from solutions in hydroalcoholic media of the above cited chromogenic reagents with viscosity, surface tension, and pH adjusted to printer requirements (10−12 cPs viscosity; 28−33 dyn/cm surface tension; 1 g/cm3 density;