Sensors & Actuators: B.

Chemical 359 (2022) 131551

Contents lists available at ScienceDirect

Sensors and Actuators: B. Chemical

journal homepage: www.elsevier.com/locate/snb

Identification and determination of multiple heavy metal ions using a

miniaturized paper-based optical device
Hoda Sharifi a, Javad Tashkhourian a, *, Bahram Hemmateenejad a, b, **
a
Chemistry Department, Shiraz University, Shiraz 71454, Iran
b
Medicinal and Natural Products Chemistry Research Center, Shiraz University of Medical Sciences, Shiraz, Iran

A R T I C L E I N F O A B S T R A C T

Keywords: A simple colorimetric paper-based sensor array (optical tongue), which can quickly and accurately identify and
Colorimetric discriminate twenty-five kinds of heavy metal ions, is developed. It functions by immobilizing six metal in­
Paper-based optical tongue dicators on the paper substrate using polyvinyl chloride. Exposing the device to the aqueous solution of metal
Immobilized indicator
ions produces a unique colorimetric signature for each ion, enabling discrimination of the ions by visual in­
Heavy metal ions
Discriminant analysis
spection. Further analysis of the colorimetric patterns by chemometrics data analysis methods leads to a clus­
Water pollution tering map for accurate classifying of the metal ions at different concentrations (10.0, 5.0, 2.5, and 1.0 mg L− 1).
In addition, the device was applied to analyze binary and ternary mixtures of metal ions as well. An interesting
feature of this tasting device is speciation analysis of an element of different oxidation states and their mixtures.
Beside to qualitative analysis, it represented quantitative analysis ability, e.g., for Zn2+ and Co2+, in the con­
centration ranges of 0.10–10.00 mg L− 1 (R2 = 0.99) and 0.25–12.50 mg L− 1 (R2 = 0.99) and detection limits of
0.03 and 0.08 mg L− 1, respectively. This optoelectronic tongue shows promising results to be used potentially of
on-site analysis of water for its quality.

1. Introduction ions have been analyzed by electrochemical [12] and chromatographic

methods [13,14] too. However, these mentioned techniques are com­
Heavy metal ions as not biodegradable chemicals can be harmful for plex, use expensive instruments, usually are time-consuming and may
aquatic ecosystems if their levels are higher than the maximum need a trained operator.
permissible level [1–3]. These ions, which usually exist mainly in Analyzing the hazardous metal concentration in water requires col­
cationic forms, mostly emit from the waste of industries [4,5]. Although lecting samples on-site, transferring samples to the laboratory and then
some metal ions are essential for human and animal health and the performing analysis in the laboratory by using expensive and timely
function of micro-organisms, many of them (e.g., heavy metals) can pose procedures [15]. So, as an important goal of the research programs, the
serious health hazards to the human body and the environment [1,4,6, development of inexpensive analytical methods that can rapidly identify
7]. So, the development of reliable, fast, and simple analytical methods and detect environmental pollutions in the sampling site has attracted
for the analysis of metal ions in environmental and food samples is an considerable attention [16]. Recently, assay procedures based on paper
important topic in chemical analysis. substrates have been used widely in different subjects including but not
Analytical methods for the determination of metal ions are mainly limited to early detection of diseases, food quality, and environmental
based on spectroscopic methods: (i) direct analysis by atomic spectro­ analysis. Interest on paper is because some papers feature like its flexi­
scopic method like atomic absorption spectrometry [8], inductively bility, portability, ease to produce and low cost, accessibility, and
coupled plasma atomic emission spectrometry (ICP-AES) [9], induc­ biocompatibility [7,17,18].
tively coupled plasma mass spectrometry (ICP-MS) [10], inductively The paper-based analytical device (PAD) was first offered by the
coupled plasma optical emission spectroscopy(ICP-OES) [11], or (ii) pioneering work of Whitesides and co-workers [19]. These devices are
indirect analysis by molecular spectroscopic methods including UV/Vis fabricated in a way that is more suitable for use by non-technical
molecular absorption and fluorescence spectroscopies. Moreover, metal personnel [15,20–23]. Fabrication of PADs is performed by patterning

* Corresponding author.
** Corresponding author at: Chemistry Department, Shiraz University, Shiraz 71454, Iran.
E-mail addresses: [email protected] (J. Tashkhourian), [email protected] (B. Hemmateenejad).

https://doi.org/10.1016/j.snb.2022.131551
Received 17 December 2021; Received in revised form 3 February 2022; Accepted 6 February 2022
Available online 9 February 2022
0925-4005/© 2022 Elsevier B.V. All rights reserved.
H. Sharifi et al. Sensors and Actuators: B. Chemical 359 (2022) 131551

Fig. 1. (A) Schematic illustration of the analytical procedure for metal ions assay and (B) actual dimensions of used PAD.

the hydrophilic/hydrophobic microchannel networks on the surface of improved sensitivity and discrimination capability toward twenty-five
the paper by technologies such as ultraviolet photolithography [24], different heavy-metal ions.
wax- printing [25], inkjet-printing [26], plasma treatment [27], and
other processing techniques. The hydrophilic part of the paper used in 2. Experimental
these devices prepared a high surface area/high optical contrast sub­
strate for various detection types [15], including colorimetric assay 2.1. Chemicals and equipment
[28], electrochemical sensing [29], and fluorescence [4] measurement.
On the other hand, rapid, simple, and portable analytical methods Pyrocatechol violet (PV), pyrogallol red (PR), chrome azurol S (CAS),
are critically important for on-site analysis of these chemical hazardous 1-(2-pyridylazo)− 2-naphthol (PAN, ≥ 97.0% purity), 4-(2-pyridylazo)
pollutants. From this view, one can consider PAD technology as a po­ resorcinol (PAR, ≥ 98.0% purity), 1,5-diphenylcarbazone (DPC), tetra­
tential candidate for on-site analysis of metal ions. For example, in 2013, hydrofuran (THF, ≥ 99.9% purity), high molecular weight poly(vinyl
Feng et al. [7] used eight pyridylazo compounds in an array format on a chloride) (PVC), Cadmium nitrate tetrahydrate (Cd(NO3)2.4 H2O, 98.0%
paper substrate and developed a highly selective PAD for the sensitive purity), and tricaprylmethylammonium chloride (Aliquat 336, ≥ 97.0%
discrimination of some heavy metal ions (Hg2+, Cd2+, Pb2+, Ag+, Ni2+, purity) were purchased from Sigma-Aldrich.
Cu2+, Zn2+, and Co2+). This group [4] also developed a fluorometric Zinc (II) nitrate hexahydrate (Zn(NO3)2.6 H2O, ≥ 98.0% purity), iron
PAD for the determination of some heavy metal ions. Very recently, a (III) chloride hexahydrate (FeCl3.6 H2O, ≥ 98.0% purity), iron (II)
portable PAD for on-site analysis of 5 metal ions has been suggested by chloride tetrahydrate (FeCl2.4 H2O, ≥ 99.0% purity), manganese (II)
Muhammad-aree and Teepoo [30]. In the above-mentioned PADs, pre­ nitrate tetrahydrate (Mn(NO3)2.4 H2O, 99.0% purity), lanthanum (III)
cise addition of a certain amount of sample solution is required. In nitrate hexahydrate (La(NO3)3.6 H2O), cobalt (II) nitrate hexahydrate
addition, because of the weak interaction of indicators with the paper (Co(NO3)2.6 H2O, 98.0% purity), copper (I) chloride (CuCl, 97.0% pu­
surface, they may leak in the presence of sample solution, which can rity), tin (II) chloride dihydrate (SnCl2.2 H2O, 98.0% purity), vanadium
affect the precision and sensitivity of the methods. (IV) oxide sulfate (VOSO4, 97.0% purity), tin (IV) chloride solution
Very recently, we showed that immobilization of indicators on paper (SnCl4, 98.0% purity), uranium (I) nitrate (U2(NO3)2), uranyl nitrate
by polyvinyl chloride (PVC) increased the sensor stability and fabrica­ (UO2(NO3)2), cerium (III) nitrate hexahydrate (Ce(NO3)3.6 H2O, ≥
tion reproducibility [31]. The objective of this work was to develop a 99.9% purity), 4-(2-hydroxyethyl)− 1-piperazineethanesulfonic acid
PVC-based PAD in an array format to develop an optoelectronic tongue (HEPES, ≥ 99.5% purity), arsenic (V) standard solution (H3AsO4),
for the detection and discrimination of many metal ions. An array of standard antimony (III) solution (Sb2O3), and dimethyl sulfoxide
metallochromic indicators, which were immobilized into a paper texture (DMSO, ≥ 99.9% purity) were acquired from Merck chemical company.
using PVC, could prevent leaking of indicators and hence made the PAD Copper (II) nitrate trihydrate (Cu(NO3)2.3 H2O, ≥ 98.0% purity),
highly stable such that it could be used by dip immersion. In this aluminum (III) nitrate nonahydrate (Al(NO3)3.9 H2O, ≥ 98.0% purity),
manner, no sampling device was required, and the device could be used nickel (II) nitrate hexahydrate (Ni(NO3)2.6 H2O, ≥ 98.5% purity), bis­
by direct contact with the liquid samples. The PAD was fabricated in a muth (III) nitrate (Bi(NO3)3, ≥ 98.0% purity), dibutyl phthalate (DBP,
very small size and therefore it needed a very small volume of the sample 98.0% purity) were purchased from Fluka. Lead (II) nitrate (Pb(NO3)2,
solution. Here, the heavy metal ions could be reacted with PVC-indicator ≥ 99.0% purity), mercury (II) nitrate dihydrate (Hg(NO3)2.2 H2O, ≥
and formed a uniform color response. The experimental results analyzed 97.0% purity), and silver nitrate (AgNO3, ≥ 99.8% purity) were ob­
with chemometrics methods showed that our method indicated largely tained from Riedel-de Haën. Vanadium (V) standard solution (NH4VO3)

2
H. Sharifi et al. Sensors and Actuators: B. Chemical 359 (2022) 131551

and titanium (IV) chloride (TiCl4) bought from BDH Laboratory. All calculated are used.
chemicals and materials were used as received without further The differences in color intensities of each PAD are collected in a
purification. vector of length 18 (6 sensing elements, each is described by 3 color
Whatman grade 2 filter paper was supplied from Whatman GE elements RGB). A response data matrix is provided by staking the vector
Healthcare (U.S.). responses of all samples. This response data matrix, whose numerical
AutoCAD 2019 software was used to design the PADs, which were data were extracted by ImageJ, is analyzed by principal component
printed using an HP LaserJet 1320 Printer. The devices were heated in analysis (PCA), and hierarchical cluster analysis (HCA) as unsupervised
an Oven (Memert, Germany). Images were captured using Can­ exploratory data analysis, and linear discriminant analysis (LDA) as a
oScanLiDE 220 color scanner. ImageJ software (National Institutes of supervised classification method. All data analyses are performed in
Health; http://imagej.nih.gov) was used for analyzing the color images. MATLAB R2018a (version 9.4).
Elemental analyses were performed by inductively coupled plasma-
optical emission spectrophotometer (ICP-OES) (Varian 720-ES, Varian 3. Results and discussion
Inc., Palo Alto, CA, USA).
3.1. Design principle of the sensor array
2.2. Fabrication of the PAD and assay procedure
In our previous article [31], we addressed some problems with the
The PADs were designed in AutoCAD software and then printed on microfluidic paper system such as instability of reagent spot such that it
one side of filter paper (0.18 mm thick) with a resolution of 600 dpi. The can be dissolved in the sample solution and therefore leach from the
printed paper was heated in the oven at 200 ◦ C for 45 min. This allows paper surface. We suggested immobilization of reagent on paper using
the polymeric binder of the printer toner to melt and penetrate into the PVC. This immobilization is provided by the interaction between PVC
paper’s width [26]. This provides hydrophobic barriers in the printed and oxygen atoms of reagents [32]. Another issue is the interaction of
area. As shown in Fig. 1, the designed PAD contains six hydrophilic metal ions of high charge density with carboxylic and hydroxyl groups of
detection zone with diameters of 3.6 mm located in a hydrophobic paper. Thus, when sample solution tends to move through hydrophilic
pentagonal area. channels of paper microfluidics, these metal ions may entrap in these
According to our previous work [31], the colorimetric reagents were channels and do not arrive in the detection zone [21]. Adding some
immobilized onto paper using PVC solution, which was prepared by portion of water after sample introduction can help to diminish this
dissolving 35 mg PVC and 60 mg dibutyl phthalate as a plasticizer in effect since added water pushes the metal ions toward the detection
3 ml THF. To prepare PVC-indicator solutions with concentration of zone. However, to do that, special microfluidic designs are required and
0.005 M, each indicator was first dissolved in 300.00 µL of DMSO and at the same time, precise measurement of the volume of the sample
then mixed with PVC solution. To achieve a homogeneous solution, the solution and the eluent solution is required.
phase catalysis reagent aliquot 336 was added to the indicators’ solu­ Here, fixing the reagents on the paper substrate by PVC allowed us to
tions. As shown in Fig. 1, the PAD is composed of 6 hydrophilic zones, use the device as dip immersion to minimize movement of sample so­
which are used as detection zone. A 0.35 µL aliquot of the mixture of lution through hydrophilic channels and to ignore using precise volume
PVC and metallochromic reagents was added to these hydrophilic zones. measurements. Dip immersion usually needs a large volume of sample
Then, 0.50 µL of HEPES buffer (0.2 M, pH 7) was dropped onto the same solution compared to microfluidic design. To reduce the required vol­
zones. ume of sample solution, the size of the device was reduced as small as
The assay procedure is very simple. Before interacting the device possible. Different shapes were examined (Fig. S2), and we found that
with the sample, the image of the device is recorded (before image). the maximum number of PADs in a A4 paper was accessible (396 PADs)
After that, some amount of standard solution of cations or sample so­ when 6 indicators were spotted on a piece of paper of pentagonal shape
lution (not less than 150.0 µL) is added in a Petri dish and the PAD is (Fig. 1). Also, the spaces between the detection zones were made hy­
immersed into the sample solution. After a specified time, the PAD pulls drophobic using laser printing. This hinders the diffusion of solution into
out from the sample solution and lets to dry at room temperature. Then, the void spaces on paper and thus reduces the volume of sample solution
the device image is recorded again (after image). required for interaction with the PAD.
The steps for fabrication of PAD and assay procedure are shown in It should be noted that compared with our PAD reported in the
Fig. S1 of the supplementary section. Moreover, some video clips are previous article [31], the size of the PAD used in this study is smaller, so
given as Supplementary materials to explain how the device was pre­ the consumption of paper and printer ink is lesser. The detection zones
pared and then was worked for analysis of water samples. Since the are directly exposed to analyte solution, and unlike the microfluidic
indicators are fixed on the PAD, the PAD can be in contact with the PADs, here there are no channel and reservoir zones, which reduces the
samples by dip immersing in the sample or by putting it on the surface of risk of analytes’ entrapment. Also, the dip-based design used in this
the sample or by direct contact with the dropping tap water. This means essay does not need precise measurement of sample volume. Moreover,
that precise measurement of it sample volume for introduction to the in this essay, we use an array of indicators that can identify metal ions of
PAD is not required. Nevertheless, we investigated the minimum different elements with variable oxidation states.
amount of the sample solution that can be analyzed by the PAD. It was The response of the sensor array depends on some experimental
found that PAD could analyze samples of having volumes as low as parameters such as pH, concentrations of indicators and buffer, and
150 µL. reaction time. Therefore, the effect of these parameters on the device’s
performance was investigated and the optimum experimental condition
2.3. Data processing was determined. The details of optimization are given in the supple­
mentary material section. As an optimized condition, HEPES buffer so­
All scanned images of the PADs are analyzed by ImageJ color anal­ lution of pH 7 was used. To control the pH of indicators’ solutions on the
ysis software. For each detection zone, the average color intensity over PAD, 0.5 µL portions of this buffer were added to every sensing zones of
all pixels is calculated. The colorimetric difference map (CDM) is the PAD. Also, it was found that the PAD represented the highest
calculated by calculating the differences in the average color values RGB discrimination ability when 0.35 µL portions of the indicators’ solutions
(extracted by ImageJ) of the ‘‘after image’’ and ‘‘before image’’. The of 0.005 M were added to the sensing zones. Concerning response time,
differences in color intensities are then changed to CDM images. Since it was observed that the discrimination ability (or selectivity) of the PAD
color value cannot be negative, the absolute differences are used when was almost the same in the time range of 2–20 min after exposing the
calculating CDMs. However, for data analysis, the difference values as PAD to the sample solutions. However, the color change intensity, which

3
H. Sharifi et al. Sensors and Actuators: B. Chemical 359 (2022) 131551

Fig. 2. (A) Images of PAD before and after exposure to twenty-five heavy metal ions and (B) corresponding colorimetric difference maps for 2 min.

may affect on sensitivity, was increased by time and was highest at device of a very small size. Therefore, highly selective reagents were not
20 min. So, in the feature analyses, the reaction time of 2 min, as the appropriate for our purpose. Instead, general metallochromic regents
shortest reaction time, which offered enough accuracy, and the reaction that interact with many metal ions but with different tendencies were
time 20 min, which offered both selectivity and sensitivity, were used more suitable. On the other hand, since environmental monitoring of
and the results were compared. The device was stable for 2 months when heavy and transition metal ions are more important than e.g., alkali and
keeping in a desiccator. alkali earth metal ions, the indicators that prefer heavy and transition
We intended to fabricate an optoelectronic tongue capable to detect metal ions were considered in this study. Moreover, obvious visual color
and discriminate a large set of metal ions (25 metal ions) on a single changes when interacting with metal ions were another important

4
H. Sharifi et al. Sensors and Actuators: B. Chemical 359 (2022) 131551

1
Fig. 3. (A) PCA and (B) HCA analysis for twenty-five heavy metal ions at concentration of 10.0 mg L− and control for 2 min.

feature that was considered for choosing the suitable indicators. Among differential interaction strength between the metal ions and indicators
available different commercial metallochromic indicators, we selected and the differences in color changes of indicators in the presence of
six following general indicators that possess the above-mentioned metal ions. It is rare to have two metal ions that interact with a list of
characteristics: Pyrocatechol violet (PV), 1-(2-pyridylazo)− 2-naphthol indicators with the same strength and at the same time produce the same
(PAN), chrome azurol S (CAS), 4-(2-pyridylazo) resorcinol (PAR), py­ color changes. These two important features caused all the studied metal
rogallol red (PR), and 1,5-diphenylcarbazone (DPC). The molecular ions to produce a distinctive color change schemes, expect for the pair of
structures of these six indicators are presented in Fig. S3. As shown in Zn2+ and Ni2+, for which very similar color differences are observed.
Fig. 2, these indicators represented clear color changes after interacting However, they could be discriminated based on their small color dif­
with different metal ions. Metals can have different hard/soft properties ferences with indicators DPC and PV.
depending on their oxidation states. Hence, according to Hard-Soft Acid-
Base (HSAB) theory, they can interact with ligands differently. Also,
3.2. Discriminant and cluster analyses
metal ions, depending on their nature and oxidation state, can exhibit
different coordination numbers when bound to a ligand. In this case, the
The potential of the fabricated analytical device to discriminate a
shared d orbitals of the same metal ions differ with different indicators
large number of metal ions was assessed by recording colorimetric re­
[33,34]. Differences in hard-soft properties and coordination number
sponses of the PAD toward 25 metal ions (Ce3+, Ag+, Cu2+, Co2+, Ni2+,
affect the complex formation constant of a series of metal ions with an
Sn4+, Cd2+, Fe2+, Fe3+, Al3+, Zn2+, Hg2+, Pb2+, Mn2+, Bi3+, La3+, U22+,
indicator (see Table S1 of supplementary material for the complex for­
Ti4+, VO3-, AsO43-, Sb3+, UO22+, Cu+, Sn2+, and VO2+) at concentrations
mation constants of some metal ions and indicators, which compiled
of 10.0, 5.0, 2.5, and 1.0 mg L− 1. As shown in Fig. 2 A, the PAD repre­
from [35,36]. When designing a colorimetric sensor array for metal ions
sents almost unique colorimetric signatures for the studied ions. In the
based on metallochromic indicators, in addition to differences in for­
shown images, clear color changes of some indicators toward some
mation constant of metal ions with the indicators, differences in color
metal ions are evident. For example, blue color of CAS in the presence of
change of an indicator when interacting with a set of metal ions [33,35]
Cu2+, Fe2+ and Al3+, pinkish color of PAN in the presence of Cu+, Cu2+,
should be considered. Therefore, the selectivity gained by the simple
Mn2+, Ni2+, Cd2+ and Zn2+ and color change of PR in the presence of
array of metallochromic indicators used here can be explained by both
almost all studied ions, except arsenate (V). Interestingly, the indicator

5
H. Sharifi et al. Sensors and Actuators: B. Chemical 359 (2022) 131551

Fig. 4. The accuracy obtained for cross-validation by the PCA-LDA model for four different concentrations (10.0, 5.0, 2.5, 1.0 mg L− 1) of metal ions at first ten PCs
for (A) 2 and (B) 20 min.

PAR, with a chemical structure like PAN, behaved differently, e.g., the results, PAN showed a higher response for transition metals (like Cu2+,
color of PAN and PAR indicators in presence of Co2+ were changed from Cu+, Fe2+, Mn2+, Co2+, Cd2+, Zn2+, Ni2+, and VO3-) than other ions.
yellow to dark gray and red, respectively. In the presence of two metal Metal ions, including Cu2+, Fe2+, and Al3+ had a strong tendency to
ions AsO43- and Pb2+, the color of CAS turned to light yellow, whereas its react with CAS. In addition, CAS also showed a relatively high response
color changed to bright yellow for Sn4+. The metal ions VO3-, Sb3+, Ti4+, to other metal ions such as Sn4+, U22+, VO3-, Fe3+, and Cu2+. The next
and Bi3+ could change the color of this metal indictor to purple. sensor, PV, played an important role in discriminating of Sn4+ and Ti4+.
The CMDs, represent visual discrimination of metal ions better than All but a few of the cations, including Ag+, AsO43-, Sb3+, UO22+, Sn2+,
original images. As one can see from Fig. 2, transition metals (like Cu2+, had shown a tendency to react with PR, which led to discrimination
Cu+, Fe2+, Mn2+, Ni2+, and VO3-), post-transition metals (like Cd2+ and between these two groups of cations. Like PAN, PAR performed the same
Zn2+), and U22+ could change the color of PAN and PAR indicators. function against cations. Of course, the signal intensity of these two
VO2+ and post-transition metals like Pb2+, Bi3+ and Hg2+ could change sensors to these cations was different (the higher responses belonged to
the color of PAR indicator but not that of PAN. PAN). The results also showed that DCP produced larger response for
In addition, PV acted an important role in discriminating elements of nine cations, such as Ce3+, Ag+, Sn4+, Al3+, Mn2+, Bi3+, La3+, Ti4+, and
different oxidation states. Al3+, Sn2+, Cu2+, and VO3- were able to Sb3+, thereby giving extra selectivity to the sensor array.
change the color of this indicator while the intensity of this color change The response signal outputs of the PAD toward all metal ions,
was lesser for Cu+. PV sensor could also help to discriminate Hg2+, Ti4+, generated a training data matrix, which was firstly subjected to PCA and
and Sn4+, because in the presence of these metals, it turned yellow. HCA as unsupervised clustering methods. The results are given in Fig. 3
Results in Fig. 2 reveal the important role of DPC for the discrimi­ for metal ions, all with concentration of 10.0 mg L− 1 and response time
nation of transition metal from others. Also, it was also observed (see of 2 min. It is observed that almost all twenty-five metal ions could be
Fig. 2 and S16) that Fe3+ could change the color of these indicators if it is contributed as separated clusters in the space of the first three principal
exposed to PAD for 20 min. This is because at the paper surface, there is components (capturing 59.9%, 14.5% and 10.7%, respectively). In the
a competition between indicator and hydroxide ion to react with Fe3+ PC score plot shown in Fig. 3, one can observe that the metal ions were
and at a longer time, the indicator won. arranged based on Hard-Soft Acid-Base (HSAB) theory in the order from
To have an overall overview of the cross-selectivity and cross- borderline to soft metals and then from soft to hard metals. For complete
sensitivity of the designed PAD to the studied metal ions, Euclidean discrimination of the analytes, PCA of larger dimension (5 PCs) is
distances of the response of all sensors against all metal ions (at con­ required. However, for visualization we limit to just 3 PCs. On the other
centration of 10.0 mg L− 1) was plotted (Fig. S17). This plot clearly hand, HCA as another unsupervised clustering uses total dimensional
shows the selectivity of the set of indicators’ responses. There are not a space of the data. As shown in the lower panel of Fig. 3, the dendrogram
pair of metal ions with the same response pattern. According to the of HCA plot exhibits complete clustering of the analytes in their

6
H. Sharifi et al. Sensors and Actuators: B. Chemical 359 (2022) 131551

Fig. 5. PCA score plot for the binary and ternary mixtures of Hg2+ with four different metal ions (Sn4+, Co2+, Cu2+ and Fe3+, the total metal ion concentration was
5.0 mg L− 1) for 2 min (B) corresponding color change patterns of the sensor array for these mixtures.

respective groups. this diluted solution. A close look at the HCA plot shown for this con­
As noted previously, the sensor responses were collected at both centration reveals that the low accuracy is because of inaccurate
reaction times of 2 min and 20 min. For comparison, cluster analysis of discrimination of some limited metal ions including Ag+, UO22+, Ti4+
the sensor response recorded at 20 min was done and the results are and La3+.
given in Fig. S18. Obviously, clustering of the metal ions at longer re­ The classification accuracy can be partly attributed to the repro­
action time is very similar with those of shorter reaction time. This ex­ ducibility of the device fabrication and response measurement.
plains that the device selectivity was the same for both short and long Repeated measurements of each sample solution for 3 times allowed us
reaction times, although large signals were obtained at longer reaction to report relative standard deviation (RSD)values for all metal ions at
times. four different concentrations (10.0, 5.0, 2.5, and 1.0 mg L− 1) and re­
In the next, cluster analysis of the diluted metal ion solutions (5.0, action times of 2 and 20 min. The shown results in Table S2 explain that
2.5, and 1.0 mg L− 1) by the PAD was investigated. The results are given the PAD had RSDs lower than 10% for most of metal ions. However, for
in Figs. S19-S24. Results confirmed that our sensor array could perform some metal ions larger RSDs, in rare cases larger than 20% are observed.
well even for discrimination of cations at concentration as low as Interestingly, this large within group variances did not hinder accurate
2.5 mg L− 1 and it failed to accurately cluster the metal ions at concen­ classification of the analyses. In cluster analysis, the between group
tration of 1.0 mg L− 1. PCA and HCA as unsupervised methods are not variances, which can be translated as sensor selectivity, plays more
able to quantify the clustering performance. To explain classification important role. Actually, the ratio of between group variance (sensor
ability of the PAD for different concentrations, PC-LDA was utilized. The selectivity) to within group variance (sensor reproducibility) determines
extracted principal components (PCs) from the response data matrices the overall classification accuracy.
were used as input of separate LDA models. In Fig. 4 are shown the In addition to the discrimination of heavy metal at the same mass
changes in PC-LDA accuracy as a function of the number of PCs entered concentration (mg L− 1), the performance of PAD was also investigated
in the model for different concentrations of metal ions and at two for discrimination of the metal at the same molar concentration of
recording times. The plots for concentrations of 10.0, 5.0 and 2.5 mg L− 1 100 µM. As shown in Fig. S25, for both times of 2 and 20 min, PAD
are very similar, which means that the PAD behaved similar for these showed accurate discrimination for all twenty-five metals at target
concentrations. According to the shown plots in Fig. 4, PC-LDA model of molar concentration.
6 PCs can be considered as optimum considering higher accuracy and Speciation analysis is an important issue when analyzing metal ions.
lesser model complexity. Accordingly, accuracy is 100% for the con­ An interesting feature of our PAD is its potential for speciation analysis
centration of 10 mg L− 1 whereas for concentrations of 5.0 and of metal elements of different oxidation states. As shown in Supporting
2.5 mg L− 1, the accuracy is 96%. information Figs. S26 and S27, the PAD performed excellent for
As obvious from Fig. 4, the PADs failed to accurately discriminate discrimination of the couples Cu2+-Cu+, Fe3+-Fe2+, Sn4+-Sn2+, U22+-
metal ions at a concentration of 1.0 mg L− 1. For recording time of UO22+ and VO2+-VO3-. These metal ions of different oxidation states
20 min and by using 8 PCs, the accuracy reaches to more than 85% for have been discriminated well at different studied concentrations. The

7
H. Sharifi et al. Sensors and Actuators: B. Chemical 359 (2022) 131551

Fig. 6. PCA score plots for 4 groups of (A) post-transition metals, (B) transitions metals, (C) metalloids, and (D) actinides/lanthanides at three different concen­
trations (2.5, 5.0 and 10.0 mg L− 1) for 2 min colorimetric color change patterns of the sensor array for (E) Zn2+ and (F) Co2+ at different concentrations and
Euclidean distances of the sensor array plotted against the different concentrations of these metal ions for 2 min. Insect: linear relationship of Euclidean distance of
the sensor array and logarithms of these metal ions’ concentration. The error bars show the standard deviation of three independent measurements.

represented data explain that speciation has been achieved better at a been clearly discriminated into distinct clusters without any overlap in
longer recording time (20 min) for the couples U22+-UO22+ and VO2+- the factor space of the first 3 PCS. Among the studied metal ions in this
VO3-. This means that kinetics of reactions with indicators are different study, Hg2+ and Cu2+ are two metal ions that usually interfere with the
for these oxo-metal pairs. determination of each other. So, discrimination of their individual and
One important future of a sensor array is its ability in discrimination mixture solutions is an important development in the analysis of these
of mixtures of analytes. Firstly, binary mixtures of Sn2+ and Sn4+ of metal ions, which was achieved here.
different mass ratios (Sn2+:Sn4+= 0:10, 2:8, 4:6, 6:4, 8:2 and 10:0) with To reproducibility of the sensor for replicate measurements is
total metal ion concentration of 5.0 mg L− 1 were analyzed by the PAD explained in the supplementary section. As shown, the relative standard
using response data recorded at reaction times of 2 and 20 min. The deviations for 8 replicate measurements are always less than 5% and
color change patterns of sensor array for these mixtures are shown in mostly are less than 3%.
Fig. S28. Clustering of these mixtures in the 3D PCA score plot is also
given in this Figure. Obviously, the mixtures were discriminated accu­
rately by proper arrangement in the shown plot with the order of mass 3.3. Quantitative analysis
ratios. This result demonstrated the array’s potential ability for the
identification and semi-quantitative detection of elements of different Besides the qualitative determination of the metal ions, which was
oxidation states. achieved by cluster and discriminant analysis, the PADs were examined
In the next, the ability of the sensor array for the discrimination of for semi-quantitative and quantitative analysis of the metal ions too. For
closely related metal elements in their mixtures was assessed. To do so, semi-quantitative analyses, responses of the PAD to metal ion solutions
binary, and ternary mixtures of Hg2+ with Co2+, Cu2+ and Fe3+ were of variables concentrations (2.5, 5.0 and 10.0 mg L− 1) were analyzed by
analyzed by the PAD. Hg2+ was selected for its strong toxicity. To pre­ cluster analysis methods. Since showing the data points of all studied
pare mixtures, equimolar solutions of the metal ions with total con­ metal ions at 3 concentrations is hard to be understood, the metal ions
centrations of 5.0 mg L− 1 were prepared. This neglects the effect of total were classified into 4 groups of post-transition metals, transitions
concentration on discrimination. As shown in Fig. 5 and S29, the solu­ metals, metalloids, and actinides/lanthanides and the result for each
tions of individual metal ions and their binary and ternary mixtures have group is shown separately. The results for reaction time of 2 min are
given in Fig. 6A-D. In all cases, the same metal ions of different

8
H. Sharifi et al. Sensors and Actuators: B. Chemical 359 (2022) 131551

Fig. 7. (A) PCA and (B) LDA score plots for real samples (tap water and industrial wastewater samples) for 2 min (C) Images of sensor array after exposure to (top)
tap water and (down) industrial wastewater sample.

concentrations are grouped close to each other whiles the same metal Finally, the metalloid group elements AsO43- and Sb3+ at different
ions of different concentrations are obviously discriminated from each concentrations are clearly clustered in two classes (Fig. 6D).
other. For quantitative analysis, the response vector of each PAD was
In the post-transition metals group (see Fig. 6A), it was observed that converted to a scalar response by calculating the Euclidean norm of that
metal ions were distributed in the factor space of PCs based on Hard-Soft vector. Then, this response was plotted against the analyte’s concen­
Acid-Base (HSAB) theory. The hard ions (like Sn4+ and Al3+), soft ions tration. Here, the results are given for just Co2+ and Zn2+ as potential
(like Bi3+, Pb2+ and Cd2+) and ions with borderline properties (like poison heavy metal ions. As shown in Fig. 6E and F, there is a loga­
Zn2+) are in the left, middle and right side of the PCA plot, respectively. rithmic relationship between the Euclidean distance of the PAD response
The same trend was observed for the transition metal ions. As shown in and the analyte’s concentration. Hence, well-defined linear relation­
Fig. 6B, soft metal ions (like Ag+), hard Lewis acids (like Ti4+, VO2+ and ships are observed between the PAD response and logarithm of the
Mn2+) and borderline (Co2+, Cu2+ and Fe2+) are distinctly separated. analytes’ concentration in the range of 0.10–10.00 mg L− 1 (R2 = 0.99)
Discrimination of lanthanides from actinides is evident from Fig. 6C. for Zn2+ and 0.25–12.50 (R2 = 0.99) mg L− 1 for Co2+. Accordingly, the

9
H. Sharifi et al. Sensors and Actuators: B. Chemical 359 (2022) 131551

calculated detection limits for these analytes were 0.03 and 0.08 mg L− 1 Table1
for Zn2+ and Co2+, respectively. These obtained DLs were below the Comparison of the current work with different sensor arrays for discrimination
maximum permission limits reported by the World Health Organization of heavy metal ions.
(WHO) for zinc and cobalt ions in drinking water (5.00 and 0.10 mg L− 1 Method Type of sensor Number Number Accuracy [Ref]
for Zn2+ and Co2+, respectively) [37,38]. elements of sensors of metal (%)
ions

3.4. Identification of discriminatory indicators Solution Coumarin-pyridine 4 11 100 with [39]

fluorescence LDA and
receptors HCA
In order to understand the importance of each sensing element used
Solution Gold-nanoparticle 3 6 100 with [40]
in this study, the factor loadings for the first 3 PCs of LDA model were (AuNP) LDA
investigated. As shown in Fig. S30, all used indicators represented sig­ Paper BODIPY (4,4- 9 7 100 with [4]
nificant contributions in discrimination. However, PV, PR and PAR device difluoro-4-bora- HCA
showed more impact than the others. The relative significant contribu­ 3a,4a-diaza-s-
indacene) and di-2-
tions of the indicators vary for the different PCs used. For example, PAR picolyamine (DPA)
contributes significantly to the first PCS whereas PV and PR contribute derivatives
more significantly to the second and the third PCs, respectively. Paper Pyridylazo 8 8 100 with [7]
device compounds HCA
Paper Metal indicators 6 25 100 with This
3.5. Application for real samples
device PCA, HCA work
and LDA
Many potential applications for this sensor can be considered. Here,
we used the PAD for untargeted metallomics investigation. Since the
PAD showed differential sensitivities to individual metal ions, in a sys­ sample solution on the PAD). According to the comparison made be­
tem of a mixture of metal ions, any change in composition may be tween our method and other reported sensor arrays for metal ions
revealed by the sensor. Therefore, we used the sensor to discriminate discrimination (Table 1), one can found that our PAD, in addition to its
drinking and wastewater as system of mixtures of metal ions. Six do­ advantages explained above, had the ability to discriminate a much
mestic tap water samples (drinking water) collected from different re­ larger number of metal ions.
gions in Iran and seven water industrial wastewater samples of Moreover, the previously published paper-based devices for detec­
sugarcane, petroleum and petrochemical industry were selected and tion of metal ions (e.g., those reported in [21–23]), could detect/identify
evaluated by the PAD. The images of the PADs and distribution of the limited number of metal ions (2–4 species) whereas since we used
samples in the spaces of PCA scores and LDA canonical variates are given general metallochromic indicators not highly selective ones, we could
in Fig. 7. Obviously, the PADs’ images of drinking and industrial waters identify a larger number of metal ions. Also, our device in this worked
have different colorimetric signatures, which is the results of different in based on immobilized indicators, which means it can be used as dip
metal ion compositions of two types of water samples. Also, they immersion without need to precise measurement of sample volume.
represent clear clusters, confirming the ability of the design PAD to Therefore, using immobilized indicators in PADs, simplifies the
discriminated polluted water from the drinking water. It was found that measuring experiment and gives extra stability to the sensor arrays.
by combining the PAD with chemometrics data analysis, drinking water This method like all sensor arrays reported so far, has limitations in
can be differentiated from the wastewater with 100% sensitivity and quantitative analysis of complex mixtures. However, by post-processing
100% specificity (see Table S3). of the sensor array data by multivariate calibration methods, one can
To explain the basis for discrimination of drinking and wastewater specialize the PAD for analysis of a given mixture of metal ions. Also, the
samples by the PADs, the composition of 3 drinking and 3 wastewater PAD may not have enough sensitivity for ultra-trace analyses. However,
samples (those samples represented closer canonical variates in Fig. 7, in one may combine it with a pre-concentration method to increase its
other words, the most similar samples from two groups) were analyzed sensitivity. Moreover, discrimination of more complex mixtures would
by ICP-OES method for content of their metallic specie. The results are be possible by our PAD. These extra researches to overcome these lim­
given in Table S4. It is observed that in some instances, the wastewater itations of the PAD are under investigation in our group.
samples contain larger amounts of heavy metals than drinking water.
CRediT authorship contribution statement
4. Conclusions
Hoda Sharifi: Methodology, Investigation, Writing – original draft.
In summary, we have presented a miniaturized paper-based colori­ Javad Tashkhourian: Conceptualization, Writing – review & editing.
metric sensor array, which is able to accurately identify and determi­ Bahram Hemmateenejad: Conceptualization, Supervision, Writing –
nation of twenty-five metal ions. In this array, six commercial metal review & editing.
indicators immobilized into PVC membrane were spotted on a very
small piece of filter paper. The discrimination performance of the array Declaration of Competing Interest
was assessed using pattern recognition methods (PCA, LDA and HCA).
This array could effectively identify the binary mixtures of Sn2+/Sn4+, The authors declare that they have no known competing financial
metal-ion mixtures and different metal ions at different concentrations. interests or personal relationships that could have appeared to influence
The metal ions were successfully identified at low concentration of the work reported in this paper.
1.0 mg L− 1 with allowable classification accuracy. Furthermore, excel­
lent linearity and reproducibility were revealed. At last, this six-sensor Acknowledgements
array demonstrated excellent potential for discrimination between tap
water and industrial wastewater samples. We thankfully acknowledge financial support from Shiraz University
Our PAD is highly stable (because of PVC immobilization), fast Research Council.
(response time of 2 min), inexpensive (because of miniaturization, it
needs just 1.8 ×10− 9 mole of reagent per PAD), high throughout put Appendix A. Supporting information
(396 no. of sensors can be prepared on one filter paper of A4 size) and
simple to use (just dip it in sample solution or drop small amount of Supplementary data associated with this article can be found in the

10
H. Sharifi et al. Sensors and Actuators: B. Chemical 359 (2022) 131551

online version at doi:10.1016/j.snb.2022.131551. [24] B. Li, W. Zhang, L. Chen, B. Lin, A fast and low-cost spray method for prototyping
and depositing surface-enhanced Raman scattering arrays on microfluidic paper
based device, Electrophoresis 34 (2013) 2162–2168.
References [25] R. Pratiwi, M.P. Nguyen, S. Ibrahim, N. Yoshioka, C.S. Henry, D.H. Tjahjono,
A selective distance-based paper analytical device for copper (II) determination
[1] M. Shirani, F. Salari, S. Habibollahi, A. Akbari, Needle hub in-syringe solid phase using a porphyrin derivative, Talanta 174 (2017) 493–499.
extraction based a novel functionalized biopolyamide for simultaneous green [26] M. Taghizadeh-Behbahani, B. Hemmateenejad, M. Shamsipur, A. Tavassoli,
separation/preconcentration and determination of cobalt, nickel, and chromium A paper-based length of stain analytical device for naked eye (readout-free)
(III) in food and environmental samples with micro sampling flame atomic detection of cystic fibrosis, Anal. Chim. Acta 1080 (2019), 138-45.
absorption spectrometry, Microchem. J. 152 (2020), 104340. [27] X. Li, J. Tian, T. Nguyen, W. Shen, based microfluidic devices by plasma treatment,
[2] Y. Ding, S. Wang, J. Li, L. Chen, Nanomaterial-based optical sensors for mercury Anal. Chem. 80 (2008) 9131–9134.
ions, TrAC Trend Anal. Chem. 82 (2016) 175–190. [28] G.O. da Silva, W.R. de Araujo, T.R. Paixão, Portable and low-cost colorimetric
[3] J. Li, J. Liu, W. Lu, F. Gao, L. Wang, J. Ma, et al., Speciation analysis of mercury by office paper-based device for phenacetin detection in seized cocaine samples,
dispersive solid-phase extraction coupled with capillary electrophoresis, Talanta 176 (2018) 674–678.
Electrophoresis 39 (2018) 1763–1770. [29] E. Rafatmah, B. Hemmateenejad, Dendrite gold nanostructures electrodeposited on
[4] L. Feng, H. Li, L.-Y. Niu, Y.-S. Guan, C.-F. Duan, Y.-F. Guan, et al., A fluorometric paper fibers: Application to electrochemical non-enzymatic determination of
paper-based sensor array for the discrimination of heavy-metal ions, Talanta 108 glucose, Sens. Actuators B Chem. 304 (2020), 127335.
(2013) 103–108. [30] S. Muhammad-Aree, S. Teepoo, On-site detection of heavy metals in wastewater
[5] A.R. Timerbaev, Element speciation analysis using capillary electrophoresis: using a single paper strip integrated with a smartphone, Anal. Bioanal. Chem. 412
twenty years of development and applications, Chem. Rev. 113 (2013) 778–812. (2020), 1395-405.
[6] L. Liu, M.-R. Xie, F. Fang, Z.-Y. Wu, Sensitive colorimetric detection of Cu2+ by [31] H. Sharifi, J. Tashkhourian, B. Hemmateenejad, A 3D origami paper-based
simultaneous reaction and electrokinetic stacking on a paper-based analytical analytical device combined with PVC membrane for colorimetric assay of heavy
device, Microchem. J. 139 (2018) 357–362. metal ions: Application to determination of Cu (II) in water samples, Anal. Chim.
[7] L. Feng, X. Li, H. Li, W. Yang, L. Chen, Y. Guan, Enhancement of sensitivity of Acta 1126 (2020), 114-23.
paper-based sensor array for the identification of heavy-metal ions, Anal. Chim. [32] N. Alizadeh, A. Moemeni, M. Shamsipur, Poly(vinyl chloride)-membrane ion-
Acta 780 (2013) 74–80. selective bulk optode based on 1,10-dibenzyl-1,10-diaza-18-crown-6 and 1-(2-
[8] M. Tüzen, Determination of heavy metals in soil, mushroom and plant samples by pyridylazo)-2-naphthol for Cu2+ and Pb2+ ions, Anal. Chim. Acta 464 (2002)
atomic absorption spectrometry, Microchem. J. 74 (2003) 289–297. 187–196.
[9] O.T. Yayintas, S. Yılmaz, M. Turkoglu, Y. Dilgin, Determination of heavy metal [33] P. Atkins, T. Overton, Shriver and Atkins’ Inorganic Chemistry, Oxford University
pollution with environmental physicochemical parameters in waste water of Press, USA, 2010.
Kocabas Stream (Biga, Canakkale, Turkey) by ICP-AES, Environ. Monit. Assess. 127 [34] R. Janes, E.A. Moore, Metal-ligand Bonding, Royal Society of Chemistry, 2004.
(2007) 389–397. [35] K.L. Cheng, K. Ueno, T. Imamura, CRC handbook of organic analytical reagents,
[10] B. Dai, M. Cao, G. Fang, B. Liu, X. Dong, M. Pan, et al., Schiff base-chitosan grafted CRC Press, 1992.
multiwalled carbon nanotubes as a novel solid-phase extraction adsorbent for [36] L. SOMMER, G. ACKERMANN, Commission on analytical reactions and reagents.
determination of heavy metal by ICP-MS, J. Hazard. Mater. 219 (2012) 103–110. Quantitative characterization of procedures using ultra violet and visible molecular
[11] A.E. Fathabad, N. Shariatifar, M. Moazzen, S. Nazmara, Y. Fakhri, absorption spectrophotometry, Pure Appl. Chem. 58 (1986) 1015–1022.
M. Alimohammadi, et al., Determination of heavy metal content of processed fruit [37] M. Kumar, A. Puri, A review of permissible limits of drinking water, Indian J.
products from Tehran’s market using ICP-OES: a risk assessment study, Food Occup. Environ. 16 (2012) 40.
Chem. Toxicol. 115 (2018) 436–446. [38] Z. Wang, X. Xing, Y. Yang, R. Zhao, T. Zou, Z. Wang, et al., One-step hydrothermal
[12] C. Fan, L. Chen, R. Jiang, J. Ye, H. Li, Y. Shi, et al., ZnFe2O4 Nanoparticles for synthesis of thioglycolic acid capped CdS quantum dots as fluorescence
Electrochemical Determination of Trace Hg (II), Pb (II), Cu (II), and Glucose, ACS determination of cobalt ion, Sci. Rep. 8 (2018) 1–12.
Appl Nano Mater. 4 (2021) 4026–4036. [39] L. Chen, L. Li, D. Wu, X. Tian, D. Xia, L. Lu, et al., Construction of multi-channel
[13] U. Araujo-Barbosa, E. Peña-Vazquez, M.C. Barciela-Alonso, S.L.C. Ferreira, A.M. fluorescence sensor array and its application for accurate identification and
P. Dos Santos, P. Bermejo-Barrera, Simultaneous determination and speciation sensitive quantification of multiple metal ions, Sens. Actuators B Chem. 303
analysis of arsenic and chromium in iron supplements used for iron-deficiency (2020), 127277.
anemia treatment by HPLC-ICP-MS, Talanta 170 (2017) 523–529. [40] X. Li, S. Li, Q. Liu, Z. Chen, Electronic-tongue colorimetric-sensor array for
[14] Q. Hu, G. Yang, J. Yin, Y. Yao, Determination of trace lead, cadmium and mercury discrimination and quantitation of metal ions based on gold-nanoparticle
by on-line column enrichment followed by RP-HPLC as metal-tetra-(4- aggregation, Anal. Chem. 91 (2019), 6315-20.
bromophenyl)-porphyrin chelates, Talanta 57 (2002) 751–756.
[15] N. Ratnarathorn, O. Chailapakul, C.S. Henry, W. Dungchai, Simple silver
Hoda Sharifi received her B.S. and M.S. degrees in pure chemistry and analytical chem­
nanoparticle colorimetric sensing for copper by paper-based devices, Talanta 99
istry from Shahid Chamran University in 2012 and 2015. She started her Ph.D. course at
(2012) 552–557.
the Department of Chemistry, Shiraz University in 2016. Her research interests focus on
[16] B. Su, Z. Zhang, Z. Sun, Z. Tang, X. Xie, Q. Chen, et al., Fluonanobody-based
the development of paper-based analytical device for monitoring environmental hazard­
nanosensor via fluorescence resonance energy transfer for ultrasensitive detection
ous materials. She is currently working as a visiting researcher at Heidelberg university on
of ochratoxin A, J. Hazard. Mater. 422 (2022), 126838.
paper-based analytical sensors.
[17] S. Kim, D. Kim, S. Kim, Simultaneous quantification of multiple biomarkers on a
self-calibrating microfluidic paper-based analytic device, Anal. Chim. Acta 1097
(2020) 120–126. Javad Tashkhourian received his M.Sc. and Ph.D. degrees in analytical chemistry from
[18] X. Qin, J. Liu, Z. Zhang, J. Li, L. Yuan, Z. Zhang, et al., Microfluidic paper-based Shiraz University in 1999 and 2004, respectively. He was a member of chemistry
chips in rapid detection: current status, challenges, and perspectives, TrAC Trends department of Persian Gulf University (2004–2010). He joined the chemistry department
Anal. Chem. (2021), 116371. of Shiraz University in 2010 where he is now an Associate Professor. His research is
[19] A.W. Martinez, S.T. Phillips, M.J. Butte, G.M. Whitesides, Patterned paper as a focused on design and construction of chemical sensors (optical and electrochemical) and
platform for inexpensive, low-volume, portable bioassays, Angew. Chem. 119 synthesis and applications of nanomaterials in electrochemical analysis.
(2007) 1340–1342.
[20] T.G. Henares, K. Yamada, S. Takaki, K. Suzuki, D. Citterio, “Drop-slip” bulk sample Bahram Hemmateenejad is a professor of Chemistry at Chemistry Department of Shiraz
flow on fully inkjet-printed microfluidic paper-based analytical device, Sens. University. He received his B.S. degree in pure chemistry from Shiraz University in 1996,
Actuators B Chem. 244 (2017) 1129–1137. M.S. and Ph.D. in Analytical Chemistry from Isfahan University of Technology and Shiraz
[21] H. Wang, Y.-j Li, J.-f Wei, J.-r Xu, Y.-h Wang, G.-x Zheng, based three-dimensional University in 1998 and 2002, respectively. As fellow of Alexander von Humboldt Foun­
microfluidic device for monitoring of heavy metals with a camera cell phone, Anal. dation, he has visited different research institutes and universities in Germany. He won
Bioanal. Chem. 406 (2014) 2799–2807. many national and international awards such as Razi Festival award, Analytical Chemistry
[22] M. Li, R. Cao, A. Nilghaz, L. Guan, X. Zhang, W. Shen, “Periodic-table-style” paper Award of the Iranian Chemical Society, Elsevier Chemolab Award and TWAS young
device for monitoring heavy metals in water, Anal. Chem. 87 (2015) 2555–2559. affiliate membership. He runs a Lab on the development and applications of different
[23] N. Idros, D. Chu, Triple-indicator-based multidimensional colorimetric sensing chemometrics methods in the chemical disciplines of bioanalytical chemistry, chemo-
platform for heavy metal ion detections, ACS Sens. 3 (2018) 1756–1764. biological interactions, computer-aided drug design and colorimetric sensors and
currently is active on development of paper-based analytical sensors.

11