Keywords

12.1 Introduction

Metals are always considered important materials in manufacturing, science, engineering and technology, and commercial aspects (Santos 2017). These bulk metals, when fragmented to nanosize having dimensions 1–100 nm, are termed metal nanoparticles or metallic nanoparticles (MNP) (Venkatesh 2018). The advent of MNP marked a revolutionary change in sensing technology and biology because of distinct electronic, optical, biochemical, and physicochemical properties. Facile synthesis and surface modification of MNPs with diverse functional moieties like antibodies, enzymes, ligands, proteins, and drugs of interest facilitate target-oriented binding with analytes rendering selectivity and an efficient sensing platform. The large surface-area-to-volume ratio and spatial confinement of free electrons offer massive numbers of binding sites on the surface of metal nanoparticles (MNP). It brings an excellent scaffold to immobilize with large quantities of ligands and biomolecules, making it highly interactive with the analytes (Venkatesh 2018; Siontorou 2019). Metal nanoparticles (like gold, silver, copper, and platinum nanoparticles) have the exceptional feature of absorption and scattering of light that originates from the collective oscillation of surface electrons to unique optical properties of MNPs (Vasquez et al. 2018; Maghsoudi et al. 2021). When exposed to appropriate frequencies of electromagnetic waves (light), it induces the excitation of electrons on the surface of MNPs known as surface plasmon resonance (SPR). This property of MNPs is also responsible for their vibrant color in an aqueous solution and can be easily tuned by changing the shape and size of the metal nanoparticle. For example, gold nanospheres of ~20 nm diameters appear wine red in color. However, their color becomes purple and blue as the size increases to ~100 nm. Likewise, the silver nanoparticle of 20 nm is yellow colored in an aqueous solution, and with the increasing size of the silver nanoparticle, color changes to red (Shrivas et al. 2015; Venkatesh 2018; Willner and Vikesland 2018). This size-dependent change in color of metal nanoparticles can be exploited to develop visual colorimetric sensors (chemical sensors, biosensors, and electro-optical sensors) where small nanoparticles aggregates in the presence of analytes change in color of the solution (Chen et al. 2014b). The metal nanoparticles are synthesized by two approaches, top-down and bottom-up (Wang and Xia 2004). Top-down synthesis involves bulk material as a precursor, broken down to nano-range particles using different physical lithography techniques such as soft lithography and electron-beam lithography. Top-down plays a vital role in the large-scale fabrication of nanostructure; it has limitations such as imperfections in resulting material, expensive, and time-consuming (Khandel et al. 2018). Bottom-up synthesis relies on the assembly of molecules or atoms to build complex nano-constructs. Some common processes used in bottom-up methods include sol-gel (Epifani et al. 2000), chemical vapor deposition (Murty et al. 2013), laser ablation (Kumari et al. 2014), and solvo-thermal method (Choi et al. 2013), but the most popular is the chemical reduction that provides the advantage of fine control over shape and size of the nanoparticle. The chemical reduction method of nano-metal synthesis involves reducing precursor metal salt in the presence of a suitable stabilizer. The quasi-spherical-shaped nanoparticles are thermodynamically most stable; therefore, the synthesis of spherical nanoparticles with different size ranges can be easily achieved by altering the concentration of precursor salt, concentration and rate of addition of reductant, and temperature. Since the optical properties of nanostructures vary with shape and size, a process to synthesize non-spherical nanoparticles has been developed to utilize these anisotropic nanostructures in different applications. Anisotropic metal nanostructures are synthesized by step-wise growth in the presence of nanoparticle seed and structure-directing agents, like cetyl-trimethyl ammonium bromide (CTAB) (Murphy et al. 2011; Chang and Murphy 2018). High surface energy and short inter-particle distances of nanoparticles make them unstable and coalesce, forming thermodynamically favored stable bulk particles. In the absence of interfering repulsive forces, metal nanoparticles attract each other resulting in larger aggregates. To maintain spatial confinement in the nano range, stabilization of metal nanoparticles is essential, which can be accomplished by steric or electrostatic stabilization using stabilizing agents like polymer, ligands, and surfactants having suitable functional groups (Olenin 2019; Sperling and Parak 2010). Stabilization of metal nanoparticles leads to the formation of an electric double layer that furnishes repulsive force to remain without aggregated in dispersed form (Venkatesh 2018; Polte 2015). Functionalization of metal nanostructures with proper capping agents not only provides stability to nanoparticles but also renders specificity toward target analytes and, thus, finds several applications in sensing (Yu and Li 2019), drug delivery (Ghosh et al. 2008), cell imaging, and photothermal therapeutics (McQuaid et al. 2016).

In the present rapid industrialization scenario, heavy metal contamination in the environment is a significant problem globally. The presence of excessive levels of heavy metal pollutants in soil and water affects the quality of surface and groundwater, resulting in a severe threat to human health and the deterioration of environmental resources. The heavy metal pollutant enters water bodies by various natural and anthropogenic sources, including mining, manufacturing, industrial, and municipal waste discharge (Poornima et al. 2016). Arsenic, chromium, lead, mercury, and cadmium are some of the most concerning heavy metals due to their high toxicity, even at low concentrations. Therefore, regulatory organizations like Environmental Protection Agency (EPA) and World Health Organization (WHO) set a standard on permissible limits of heavy metal consumption (Varun and Kiruba Daniel 2018; Zhang et al. 2019a; Li et al. 2013; Tchounwou et al. 2012). Consumption of heavy metals above permissible limit triggers bio-toxic effects by altering cellular activities and developing severe disorders, including cancer. Since they are not quickly metabolized or excreted, they tend to accumulate in soft tissues for years, which slowly causes mental and central nervous dis-functioning, damage to the liver, kidneys, lungs, and other vital organs (Li et al. 2013). The traditional techniques for detecting these toxic metals are based on either spectroscopy or chromatography, which includes inductively coupled plasma mass spectrometry (ICP-MS), atomic adsorption spectroscopy (AAS), inductively coupled plasma optical emission-spectrometry (ICP-OES), X-ray fluorescence spectrometry (XRF), and high-performance liquid chromatography (HPLC). These techniques are highly selective, sensitive, and efficient in quantification; their large-scale implementation is still a challenge due to their complexity and sophisticated installation procedure. Moreover, they require technical expertise to operate, involve toxic chemicals, multiple sample preparation steps, time-consuming, and dedicated laboratory setup, and time-to-time maintenance of instruments makes analysis highly expensive (Zhang et al. 2019a; Lu et al. 2018; Buledi et al. 2020). Thus, there is a need to develop an inexpensive heavy metal detection technology that is rapid, easy-to-handle, user-friendly, portable, and operated as a point-of-use device. In this context, nanoparticle-enabled colorimetric sensing technology has huge potential to detect metal toxins on-site with improved performance as a device in terms of selectivity, sensitivity, and reproducibility that can be readily developed into commercial products. A general schematic of MNP synthesis and its advantages in sensing application has been shown in Fig. 12.1. This chapter reviews the colorimetric sensing strategies for heavy metals based on aggregation and dispersion of metal nanostructure, specially focusing on gold nanoparticle-based sensors. It also highlights recent advances in developing a miniaturized, point-of-use colorimetric sensor for metal toxins on paper substrate. Moreover, the integration of smartphone camera readouts and machine learning approach with colorimetric sensors introduces a new lab-on-mobile concept, which has been discussed in the later section of this chapter.

Fig. 12.1
figure 1

General schematic of MNP synthesis and their use in sensing applications of toxic metal ions

Full size image

12.2 Metal Nanoparticle-Based Sensor

Sensors are devices that convert the chemical or physical properties of a specific analyte into a measurable signal proportional to the analyte concentration (Jayabal et al. 2015a). Metal nanoparticle-based sensing devices are characterized in three units, i.e. (a) metal nanoparticle, (b) a recognition component that furnishes selectivity, and (c) a signal transduction system, which supplies information about the presence and absence of analyte (metal toxin) in a sample (Willner and Vikesland 2018; Mahato et al. 2018). The resultant signals originating from the MNP sensor can be of different types, and based on these signals, sensors can be categorized as optical, electrochemical, and piezoelectric sensors (Willner and Vikesland 2018). Optical sensor depends on the interaction of toxic metals with electromagnetic radiation (like ultraviolet, visible, or infrared light) in the form of emission or absorption, which spectroscopic techniques can monitor. Colorimetric and fluorescence are two methods commonly used as reporting signals in optical sensors. Colorimetric sensing is based on surface plasmon resonance of metal nanoparticles. SPR peak of the metal nanoparticle is highly sensitive to the inter-particle distance between nanoparticles. Plasmon coupling causes aggregation of metal nanoparticles with pronounced color change and concomitant red-shift of SPR peak. Most of the gold/silver nanoparticle-based colorimetric sensors explore the property of color change coupled with aggregation and dispersion of nanoparticles in the presence of target heavy metal (Doria et al. 2012). The fluorescence sensor consists of a fluorophore as a signal-transducing element, which exhibits the property of photoluminescence. When the fluorophore is irradiated with electromagnetic radiation, it absorbs the photon energy, and its orbital electrons are excited to a higher energy level (singlet state). Fluorescence occurs when the excited electron relaxes to a lower energy state (ground state) by emitting a photon. Upon interaction with the heavy metal toxin, the fluorescent signal changes as either “turn-off” or “turn-on.” The metal nanoparticles of size 3 nm or smaller (like nanodots and nanoclusters) can be directly used as a fluorescent marker as they exhibit inherent fluorescence properties. Moreover, MNPs which lack their fluorescence can be functionalized with a fluorophore to obtain a fluorescent sensor. Quenching and restoration of fluorescence property indicate interaction of toxic metal and nanoparticle-based changes in the sensor (Xiong et al. 2019; Willner and Vikesland 2018). An electrochemical sensor is a device that transforms chemical reactions into electrical signals (Alam et al. 2020). When an electrical circuit is introduced to heavy metal, the molecular binding of toxic metal near electrode surface initiates oxidation/reduction process, which generates or modulates electrical current in the form of charge (electron) transfer between the electrode and toxic metal ion. This charge transfer may lead to the completion of an incomplete circuit or alteration in current, potential, or resistance measured by instruments like potentiostat or galvanostat (Willner and Vikesland 2018; Doria et al. 2012). Electrochemical sensors may exist in a two-electrode system or three-electrode system (Andrea Scozzari 2008). In the two-electrode arrangement, a working electrode (WE) is coupled with a counter electrode (Auxiliary, CE), and the difference of electric potential is measured between WE and the potential of CE. Examples of the two-electrode system are amperometric sensor (measures electric current between CE and WE in the presence of a constant electric potential) (Sahin and Kaya 2019) and potentiometric sensor (measures the potential difference between two electrodes, i.e., WE and CE in the absence of current flow) (Isildak and Özbek 2020). Three-electrode consists of a reference electrode (RE) along with WE and CE. The best example for a three-electrode system is a voltammetric sensor, which measures current response as a function of applied potential. Current is linearly dependent on the concentration of electro-active species (toxic metal ion) (Power and Morrin 2013). Among these sensors, colorimetric sensors have several advantages: simplicity, unmatched sensitivity, and inexpensive and fast detection time. Moreover, it operates in a visible range of the electromagnetic spectrum. The resultant signals can be detected by naked eyes, making it possible for wide-scale use by the common people. Therefore, it gained huge attention for quick detection of metal toxins in solutions (Kim et al. 2012; Jayabal et al. 2015b).

12.3 MNP-Based Colorimetric Detection Strategies

Metal nanoparticles are highly flexible because of precise control on size, shape, composition, assembly, and optical properties during the synthetic process. Thus, it has been extensively investigated for colorimetric sensing of toxic metal ions. WHO has standardized the consumption limit for these toxic metal ions; for instance, the permissible limit for arsenic is 10 ppb, chromium 50 ppb, lead 15 ppb, mercury 2 ppb, copper 15 ppb, and cadmium 5 ppb. Thus, their trace level monitoring is essential, and for which researchers are coming up with new techniques based on metal nanoparticles. This section deals with MNPs-based colorimetric sensors for the determination of toxic metals. Zhou et al. reported 4-mercaptobenzoic acid (4-MBA) modified silver nanoparticles (AgNP) for colorimetric sensing of CuII ion in water. 4-MBA consists of –SH group that binds with AgNP and –COOH groups exposed on the surface that chelates CuII forming carboxylate–CuII–carboxylate bridges. Chelation of CuII causes aggregation of AgNP and color changes from yellow to purple (Fig. 12.2a) (Zhou et al. 2011). Chen and coworkers demonstrated a paper-based colorimetric device (PCD) for HgII detection using citrate stabilized PtNP. 3,3,5,5-tetramethylbenzidine (TMB) and H2O2 produce blue color in the presence of PtNP that mimics the peroxidase activity by catalyzing the reaction. However, the introduction of HgII in the reaction system inhibits the catalytic activity of PtNP resulting in a color change from blue to colorless. (Fig. 12.2b) (Chen et al. 2016). Shrivas et al. reported colorimetric sensing of PbII using polyvinyl alcohol (PVA) functionalized AgNP and paper-based analytical devices. After interaction of PbII with AgNP-PVA, the color changes from yellow to red and color intensity were recorded on a smartphone followed by processing in ImageJ software (Fig. 12.2c) (Shrivas et al. 2019). Similar color-based sensing methods have been reported for toxic metals like AsIII, CrIII/VI, CdII, PbII, CuII, and HgII using modified metal nanoparticles including Au, Ag, Pt, and Pd nanoparticles, which have been listed in Table 12.1. Silver and gold nanoparticles exhibit prominent SPR-based properties associated with a color and have been widely explored for the same. Though Pt-like nanoparticles exhibit SPR features, they are examined mainly for their catalytic activity and enzyme mimetic behavior for inducing an indirect color change in sensor application. Among all these metallic nanoparticles, gold nanoparticles (GNP) received much attention in colorimetric sensing applications (Singla et al. 2016). GNP can be prepared by simple methods with high stability and provides a suitable platform for multi-functionalization with various biological and organic ligands for selective binding of target toxins. They have a high surface-to-volume ratio and exhibit unique optoelectrical and catalytic properties imparting useful surface plasmon behavior that generates a detectable response in the presence of metal ions. These distinctive properties make them “star” among the other nanoparticles providing researchers a broad spectrum for sensor application (Jans and Huo 2012).

Fig. 12.2
figure 2

Metal nanoparticle-based colorimetric sensor for heavy metal. (a) 4-MBA-AgNP-based colorimetric detection of CuII with color change from yellow to purple. (b) Paper-based analytical device for detection of HgII using PtNP in the presence of TMB and H2O2. (c) Colorimetric sensing of PbII using PVA-AgNP coated paper substrate showing color change from yellow to red. Reproduced from Ref. Zhou et al. (2011) (a). Open access with proper citation (b). Reproduced from Ref. Shrivas et al. (2019) (c)

Full size image
Table 12.1 MNP-based colorimetric sensing of toxic metal in solution and paper-based analytical devices
Full size table

12.4 Synthesis, Functionalization, Properties, and Sensing Strategy of GNP

12.4.1 Synthesis of GNP

The sensing application of gold nanostructures is dependent mainly on their shape, size, and surface functionality, and thus selection of suitable synthetic procedures is crucial in designing a sensor (Yeh et al. 2012; Yu and Li 2019). In the last few decades, numerous chemical and physical procedures are adopted to precisely control the shape, size, and mono-dispersity of nanoscale gold. However, green synthesis technologies involving biological entities are gaining much attention nowadays due to the environmental-friendly and biologically safe approach to synthesizing gold nanoparticles (Zhang et al. 2020). The details of some commonly used synthesis procedures for nanoscale gold are summarized here.

12.4.1.1 Turkevich–Frens Method

Turkevich first reported the synthesis of gold nanoparticles in 1951 (Turkevich et al. 1951). In this method, gold chloride salt (HAuCl4) is heated (~90 °C) in the presence of a reducing agent like sodium citrate resulting in monodispersed spherical gold nanoparticle suspended in the water of around 20–40 nM in diameter (Fig. 12.3a). Here, citrate plays the role of both stabilizing and reducing agents (Saha et al. 2012). Later, in 1973, Frens improved this protocol and prepared gold nanoparticles of different sizes ranging between 16 and 147 nm. This method provides precise control over the size of the gold nanoparticle by adjusting the proportion of chloroauric acid salt to sodium citrate solution (Razzaque et al. 2016). The citrate reduction provides a negative surface charge to gold nanoparticles and prevents aggregation by imposing repulsion induced by Coulombic force (Chen et al. 2014a). Citrate forms a weak coordination layer with gold nanoparticles that adds easy replacement of citrate with functionalizing agents like thiols (Zhu et al. 2003), polymers, and biomolecules (Nghiem et al. 2010).

Fig. 12.3
figure 3

Various synthesis procedures of gold nanoparticles. (a) Turkevich–Frens method. (b) Brust–Schiffrin method. (c) Seed-mediated growth method for anisotropic gold (GNR). (d) Green synthesis of GNP using plant extract

Full size image

12.4.1.2 Brust–Schiffrin Method

In 1994, Brust and Schriffin introduced another protocol for synthesizing gold nanoparticles in an organic medium that results in water-immiscible gold nanoparticles (Brust et al. 1994). This synthetic strategy involves bi-phasic reduction of gold salt to produce thiol-protected gold nanoparticles (Li et al. 2011; Perala and Kumar 2013). The gold chloride salt is transferred from the aqueous phase to the organic phase (toluene) using the surfactant tetraoctylammonium bromide (TOAB) followed by the addition of dodecanethiol as a stabilizing agent. Next, sodium borohydride (a strong reducing agent) is added that reduces gold salt to thiol-stabilized gold nanoparticles producing deep brown color in toluene (Fig. 12.3b). This method results in 1.5–5 nm gold nanoparticles having low dispersity. Alkanethiol forms a monolayer on GNP surface enabling easy modification with various functional groups (Saha et al. 2012).

12.4.1.3 Seed-Mediated Growth

The seed-mediated growth is the most reputed procedure used in the preparation of anisotropic gold nanoparticles like nanorod (GNR), introduced by Murphy and coworkers (Jana et al. 2001) and El-Sayed group (Nikoobakht and El-Sayed 2003). Cetyltrimethyl ammonium bromide (CTAB), a surface-active molecule, is used as the template for directed growth of anisotropic nanostructures (Meng et al. 2019). The synthesis of gold nanorod is achieved in two steps (Fig. 12.3c). Step 1 involves the seed solution synthesis where a golden brown–colored gold seed is prepared by reducing gold chloride salt (HAuCl4) with freshly prepared ice-cold borohydride (NaBH4) sodium in the presence of CTAB. Step 2 involves the growth of nanorods in the presence of CTAB and silver nitrate (AgNO3). Gold salt is reduced with a mild reductant ascorbic acid followed by seed solution. The growth solution is left undisturbed for 12 hrs to grow the seed crystals into gold nanorod under the action of surfactant (CTAB). The nanorods of desired aspect ratio (length/width) are achieved by altering the concentration of gold chloride salt and seed (Murphy et al. 2011; Li et al. 2018a).

12.4.1.4 Green Synthesis of Gold Nanoparticles

Green synthesis of nanoparticles emerged as an attractive substitute for conventional chemical synthesis procedures. This method involves the use of unicellular and multicellular biological entities like plant extracts, bacteria, actinomycetes, fungus, yeast, and viruses. The biological entities act as a factory for nanoparticle synthesis that offers non-toxic, inexpensive, and environmental-friendly approaches both extra and intra-cellularly without using toxic chemicals during synthesis, thus also termed as “Green chemistry” (Fig. 12.3d) (Baranwal et al. 2016; Zhang et al. 2020; Salem and Fouda 2021; Sengani et al. 2017) Das and coworkers synthesized 20 nm spherical gold nanoparticles by using flower extracts from Nyctanthes arbor-tristis (night jasmine) (Das et al. 2011). Narayanan and Sakthivel used leaf extracts of Coriandrum sativum to synthesize gold nanoparticles of size 7–58 nm (Narayanan and Sakthivel 2008). Husseiny group reported the extracellular preparation of GNP using a bacterial strain Pseudomonas aeruginosa. The mechanism involves the transfer of an electron from nicotinamide adenine dinucleotide hydrogen (NADH)-dependent reductase enzyme resulting in the reduction of Au3+ to Au0 and itself oxidized to NAD+ (Husseiny et al. 2007). Synthesis of nanoparticles using plant extracts is comparatively easier, faster, and cost-effective than bacterial synthesis as it does not require complex and multiple-step processes like isolation, culturing, and maintenance of bacterial strain (Iravani 2011).

12.4.2 Colloidal Stability of Gold Nanoparticles and Functionalization for Sensor Application

Colloidal stability of gold nanostructure depends largely on the surface energy, surface composition, and the inter-particle behavior arising from the surface and intermolecular forces. Nanoparticles exhibit high surface energy and short inter-particle distance, which make them unstable and result in aggregation. The chemistry behind the aggregation of the nanoparticle is complicated due to the involvement of different kinds of forces like electrostatic repulsion, van der Waals, and magnetic forces (Rance et al. 2010). However, the van der Waals force of attraction dominates due to short inter-particle distance, which compels them to form aggregates. Therefore, to avoid the agglomeration of nanoparticles, repulsive force is introduced by adding a capping agent during the synthesis of nanoparticles. The capping agents bind on the surface of the nanoparticle, providing two types of stabilization, electrostatic and steric. In an aqueous environment, most of the nanoparticles carry some surface charge due to ionization of the surface group or adsorption of charged molecules or ions. To balance the surface charge, a cloud of opposite charges is created. This charged cloud consists of the inner stern layer and outer diffuse layer forming an electric double layer (EDL), creating an electrostatic repulsive force between particles. In the case of steric stabilization, a physical barrier is created by the adsorption of ligands on the particle surface that prevents particles from aggregation (Amina and Guo 2020; Moore et al. 2015). The stabilization of gold particles was described by DLVO (Derjaguin–Landau–Verwey–Overbeck) theory of balance between the repulsive (electrostatic interactions) and the attractive force (van der Waal). According to DLVO theory, the sum of electrostatic and van der Waals force between two nanoparticles represents the total force acting on the colloidal solution (Zhou et al. 2009; Aldewachi et al. 2018).

Due to the nanometer size, gold nanoparticles have a high surface to volume ratio making them extremely active, and therefore, surface capping is required to lower the surface energy and increase stability. Surface functionalization of gold nanoparticles also renders specificity for the target analyte during the sensing procedure. Therefore, surface modification of GNP accomplishes two objectives: (1) chemical stability and (2) target-specificity (Zhang 2013). Gold nanoparticles can be functionalized by thiol-containing ligands, biomolecules, and polymers using different strategies like covalent coupling (Au-S bonding), specific recognition (e.g., antibody-antigen, DNA, aptamer), and electrostatic interaction. Sulfur-containing molecules are highly effective functionalizing agents, since, Au-thiol bonds are strong, resulting in highly stable gold nanoparticles. Many thiol-containing compounds like thioctic acid, glutathione, thioguanine, cystamine, thiolated–diethylenetriaminepentaacetic acid (SH-DTPA), and thiolated–PEG (SH-PEG) are used to modify gold nanoparticles and implicate in sensing of toxic metal ions (Mahato et al. 2019). Chai and coworkers reported glutathione functionalized gold nanoparticle (GSH-GNP) for PbII detection based on aggregation and red to blue color change of GSH-GNP (Chai et al. 2010). Xue and group demonstrated 6-mercaptonicotinic acid and L-Cysteine co-functionalized GNP as CdII sensor (Xue et al. 2011). Wang and coworkers reported 4-amino-3-hydrazino-5-mercapto-1,2,4-triazole (AHMT) functionalized gold nanoparticles for colorimetric CdII detection. AHMT consists of thio, amino, and triazole groups which can form bonding with GNP surface. However, among these groups, thiol preferentially binds with GNP forming (Au–S) bond while other groups are involved in the chelation of CdII ion (Wang et al. 2013). Compared to other molecules, thiol-ligands bind easily with the gold nanoparticle surface, which can be attributed to the mechanism of ligand exchange that replaces the already bound ligand with the thiol compound without altering the structural integrity of GNP. Liu and Lu fabricated a lead biosensor using DNAzyme-directed aggregation of DNA modified gold nanoparticles. DNAzyme consists of an enzyme specific to PbII ions. DNA functionalized GNP forms a blue-colored assembly in the presence of DNAzyme. The introduction of PbII in reaction activates DNAzyme to cleave DNA stand and change the color of GNP from blue to red (Liu and Lu 2003). Lee and Mirkin developed a highly selective HgII detection assay based on thymine-Hg-thymine base paring. The GNP surface was functionalized with a thiol modified DNA probe (probe 1 and probe 2). Though the thiol modified-GNP is stable, it loses stability in the presence of HgII, forming a bridge of thymine-Hg-thymine, leading to mismatch in T–T base pairs. This causes red-colored GNP to turn blue leading to aggregation (Lee and Mirkin 2008).

12.4.3 Optical Properties

The refinement of bulk gold to nanoscale dimension allows them to interact with light, causing strong absorption at specific wavelengths. When gold nanoparticles are exposed to light, the electromagnetic field of light causes polarization of free electrons present on GNP’s surface, resulting in their collective oscillation. When the frequency of incident light coincides with the frequency of collective oscillation of surface electrons, it absorbs the radiation of that particular wavelength giving an absorption band known as surface plasmon resonance (SPR). The SPR of gold nanostructures ranges from the visible to the near-infrared region of the electromagnetic spectrum depending on the size and shape of the gold nanoparticles (Fig. 12.4a) (Wang and Yu 2013). A spherical gold nanoparticle of 20 nm possesses SPR peak at 520 nm in the visible region responsible for the red color of the colloidal solution. As the size of the GNP increases, the SPR band gradually shifts to a higher wavelength with a concomitant color change of colloidal solution. Thus, the colored appearance (red, orange, brown, purple, and blue) of the colloidal gold solution is dependent on the size of gold nanoparticles (Amina and Guo 2020). Moreover, as the symmetry changes from spherical to nanorod, the SPR splits into two Plasmon bands known as transverse bands arising due to electron oscillation along the short axis (width) and longitudinal band due to electron oscillation along the long axis (length) (Fig. 12.4a) (Yasun et al. 2013). Richard Gans, in 1912, explained that the change in the shape of nanoparticles leads to alteration in position and number of SPR band and thus depends on the aspect ratio (length/width) of nanoparticles, not absolute dimension. Thus, a change in SPR band can be used to track the interaction of analytes with gold nanoparticles (Nath et al. 2018b).

Fig. 12.4
figure 4

Schematic illustration of surface plasmon excitation of GNP and GNR (a), metal ion-mediated aggregation of GNP (b); metal ion-induced side by side (c), end to end interaction (d), both side by side and end to end interaction forming total aggregate of GNR (e), and metal ion-mediated etching of GNR to spherical-shaped nanoparticle (f) responsible for the change in optical properties of gold nanostructures

Full size image

The detection mechanism of colorimetric sensors is dependent on the change in color and absorption band associated with the aggregation and disaggregation of nanoparticles. Interaction of heavy metals with spherical gold nanoparticles forms aggregates with a shift in SPR and red to blue color change (Fig. 12.4b). However, gold nanorod aggregates in different ways, namely end to end, side by side, aggregation, and etching of nanorod (Vigderman et al. 2012). End to end interaction occurs when metal toxins bind at the edge of nanorod forming chain/wire-like structure resulting in bathochromic shift of longitudinal band. Metals toxins when bound on the longer edge of the nanorod result in side by side interaction with a blue shift in longitudinal SPR. Aggregation-based assembly first initiates with end to end and ends side by side, forming a total-aggregate nanorod with decreased absorption band intensity. Some metal toxins etch nanorod at longitudinal edge forming spherical shape coupled with red shift and change in color from purple to pink. The change in optical properties with toxic ion-mediated aggregation has been shown in Fig. 12.4c, d, e, f.

12.5 GNP as a Colorimetric Sensor for Heavy Metals

As discussed in Sect. 12.4.3, gold nanoparticles are extremely responsive to the local dielectric, which results in aggregation with change in SPR and the color of colloidal solution indicating the presence of test analyte. Researchers have extensively investigated this feature to develop a color-based sensor for metal ions, which has been discussed in this section. Chen et al. reported DMSA functionalized GNP for trace level detection of CrIII/VI. Chromium exists as an aqueous complex with six water molecules coordinated with the CrIII ion. Cr(III) has empty orbitals that accept one pair of electrons from the oxygen of DMSA, forming a metal–O coordinate covalent bond by replacing water molecules from aqueous chromium. However, in the case of Cr2O72−, the CrVI ion lacks coordination sites as they are already occupied by oxygen atoms. Therefore, chromium binding occurs through hydrogen bonding OH----O involving carboxyl –OH of DMSA and –O of Cr2O72− (Fig. 12.5a). (Chen et al. 2015) Nath et al. reported a red to blue color change sensor for arsenic (III and V) using 2-mercapto-4-methyl-5-thiazoleacetic acid (MMT) and europium chloride (EuCl3) functionalized gold nanoparticle (GNP-MMT@Eu). The SPR peak intensity of GNP-MMT@Eu at 525 nm decreased while a new peak at 650 nm appeared due to arsenic-mediated aggregation of nanosensor. GNP-MMT@Eu exhibits Eu-OH group exposed on GNP surface which are the sole binding sites for arsenic ion. The As-OH/As-O− groups of arsenic and –OH groups of Eu(III) bind forming an inner-sphere arsenic complex between GNP with the release of H2O and OH− moieties. The response of GNP-MMT@Eu for As(V) was quick compared to that of As(III). The nanosensor surface attains a partial positive charge at pH ~6-7 as the Eu-OH converts to Eu-OH2+ which initiates the binding between H2AsO4−/HAsO42− and GNP-MMT@Eu through electrostatic interaction. Thus, both covalent and electrostatic modes of binding prevail between arsenic and the nanosensor, which are accountable for rapid response to AsV (Fig. 12.5b) (Nath et al. 2018a). Priyadarshni et al. demonstrated aggregation-based detection of AsIII/V using gold nanorod (GNR) modified with mPEG-SH and DMSA. After interaction with AsIII/V, the bluish-purple color of the GNR sensor turns colorless with a small shift (778 to ~802-820 nm) and decrease in SPR peak suggesting side to side and end to end binding forming total aggregate. DMSA contains two thiols (–SH) groups; one binds with GNR and other complexes with AsIII/V. At pH ~7, AsIII and AsV exist as H3AsO3 and H3AsO4/H2AsO4− while thiol remains protonated. As−OH groups participate in As−S bond formation and release H2O due to removal of hydrogen from –SH and displacement of –OH− to form a complex between AsIII/V and GNR, inducing arsenic-mediated aggregation of nanorods (Fig. 12.5c) (Priyadarshni et al. 2018).

Fig. 12.5
figure 5

(a) Interactions between DMSA-GNP and CrIII (top) and CrVI (down). (b) AsIII/V mediated aggregation of GNP-MMT@Eu. (c) AsIII/V induced aggregation of DMSA functionalized gold nanorod. Reproduced from Ref. Chen et al. (2015)) (a). Reproduced from Ref. Nath et al. (2018a)) (b). Reproduced from Ref. Priyadarshni et al. (2018) (c)

Full size image

12.5.1 Detection on Paper Substrate

Conventional detection methods for metal toxins have constantly improvised with new-age technological developments to miniaturize the setup and provide a decentralized approach. The emergence of microfluidic techniques resulted in the notion of the “lab-on-chip (LOC)” concept back in the 1990s (Guan and Sun 2020; Sackmann et al. 2014). In recent years, paper-based microfluidics have emerged as promising LOC sensing devices (Li et al. 2012; Yetisen et al. 2013; Kumar et al. 2015). Microfluidics that couple paper-based devices with colorimetric analysis are particularly attractive, attributed to easy fabrication, portability, and inexpensive, i.e., provide a cheaper alternative for point-of-use testing. Moreover, the paper has a porous matrix that offers self-pumping and capillary flow to the solution (Mahato et al. 2020; Xiong et al. 2020; Mahato et al. 2017). This section discusses some paper-based colorimetric methods for sensing toxic metal ions in water. Nath et al. reported trace-level determination of AsIII on Y-shaped microfluidic paper device using thioctic acid and thioguanine conjugated gold nanoparticle (Au-TA-TG). The two arms of the device were used as the inlet, each for Au-TA-TG and AsIII, and the reaction occurs on the paper surface resulting in red to blue color change, suggesting the existence of AsIII (Fig. 12.6a) (Nath et al. 2014). Zhang and coworkers demonstrated the detection of CuII by etching of nanorod in the presence of hexadecyltrimethylammonium bromide (HBr) on a paper substrate. When CuII in combination with HBr is added to nanorod, HBr induces transformation of Au(0)-to-Au(I) and CuII catalytically etches the longitudinal edge of GNR accompanied with color change from blue to red (Fig. 12.6b) (Zhang et al. 2014). Paper-based colorimetric metal ion detection using gold nanoparticles has been shown in Fig. 12.6 and listed in Table 12.2.

Fig. 12.6
figure 6

(a) Au-TA-TG on Y-shaped paper strip for colorimetric detection of AsIII. (b) Paper-based colorimetric detection of CuII by etching of nanorod in the presence of HBr turns blue-to-red. (c) Paper-based colorimetric detection of CrVI by aggregation of BSA-AuNP/STCP. Reproduced from Ref. Nath et al. (2014)) (a). Reproduced from Ref. Zhang et al. (2014) (b). Reproduced from Ref. Guo et al. (2016) (c)

Full size image
Table 12.2 GNP- and GNR-mediated colorimetric detection of metal toxins in solution, paper substrate, and smartphone
Full size table

12.6 Smartphone and Machine Learning (Color Readout)-Based Quantification of Heavy Metals

The enhanced technical capabilities of smartphones, especially wireless connectivity and high definition cameras, enable various innovative ideas for detecting environmental toxins like heavy metals (Mutlu et al. 2017). The addition of a simple colorimetric sensing apparatus to a smartphone makes it lab-on-phone, cost-effective, portable, and accurate (Sajed et al. 2020; Wang et al. 2019). Most of the integrated smartphone detection systems introduced so far rely on RGB (red, green, blue) intensities of the colorimetric sensor. Chen et al. developed a smartphone integrated colorimetric sensor using meso-2,3-dimercaptosuccinic acid (DMSA)-capped GNP to detect chromium ions (III and VI). The detection system was dependent on RGB extraction using an application software “color scan” and the ratio of green to red was calculated to obtain a calibration curve and determine the amount of chromium in the real water sample (Fig. 12.7a) (Chen et al. 2015). Faham and coworkers showed a paper-based colorimetric sensor integrated with smartphone readout for chromium (III) detection in solution. 2,2′-thiodiacetic acid-modified gold nanoparticle (TDA-GNP) was spotted on a paper disc and treated with various concentrations of CrIII followed by image capturing using a Samsung E5 smartphone. Color intensities of the paper disc were obtained using adobe photoshop CS5 and applied to obtain CrIII concentration in real samples (Faham et al. 2018). Smartphone-based detection provides real-time on-site application of colorimetric sensors with the help of simple software available by third-party service providers. However, to obtain analytical values from color, the images are highly processed and compressed, leading to alterations in final analytical data and cannot be trusted completely. Moreover, simple analytical models fail to detect the number of independent variables like in the multi-analyte sensor. The drawbacks of illumination and smartphone detection systems can be overcome by applying artificial intelligent systems like machine learning (Mutlu et al. 2017).

Fig. 12.7
figure 7

(a) RGB extraction and smartphone readout-based detection of CrIII/VI using DMSA-GNP. (b) General work flow of machine learning and its advantages in colorimetric sensing. (c) Smartphone readout and regression-based model to process RGB features extracted from color change on interaction of PbII with oligonucleotide-GNP. Reproduced from Ref. Chen et al. (2015) (a). Reproduced from Ref. Sajed et al. (2020) (c)

Full size image

Machine learning (ML) is a subset of artificial intelligence (AI) that acquires information from raw data by extracting its features and uses it to tackle problems without human intervention. It is a computer program or algorithm that makes machines more intelligent in behavior and decision by enabling them to learn from past experiences and develop their own program (Cui et al. 2020; Lussier et al. 2020). In the field of sensing, ML is employed as a tool for data processing to extract features like color intensity and utilize these features to predict the concentration of analyte and toxic ions directly (Cui et al. 2020). A general workflow of machine learning and its advantages in colorimetric sensing is presented in Fig. 12.7b. ML is grouped into two classes: supervised and unsupervised learning (Ayodele 2010). In unsupervised learning, the algorithm is not trained with the input data (training data); instead, it learns from the pattern of untagged data, builds a concept, and predicts the output. k-Means clustering is the most used unsupervised learning algorithm.

On the contrary, supervised learning involves training of ML algorithms with input data called training data. Based on training data, the algorithm predicts the output for the unknown input data called testing data. Convolutional neural network (CNN), artificial neural network (ANN), support vector machine (SVM), and multiple linear regression (MLR) are some supervised ML algorithms, gaining attention in chemical and biological colorimetric sensors development (Cui et al. 2020). Sajed group reported a novel detection method for HgII ions in water utilizing smartphone-based machine learning regression. Aptamer modified gold nanoparticle changes color from red to purple in HgII, and the corresponding color change was captured on a smartphone camera. The obtained color cards trained the machine learning regression model to interpret mercury ion concentration based on RGB values. The smartphone was fabricated with an optoelectronic component using three-dimensional printing technology, an attachment to any smartphone. The optoelectronic device comprises three compartments: LCD board and camera’s depth focus chamber, and cuvette holder. This setup provides an advantage of blocking ambient light interference and results in reproducible LCD illumination (Sajed et al. 2019). In another work, Sajed et al. reported similar detection apparatus for PbII ion using oligonucleotide modified gold nanoparticle and machine learning regression on the mobile platform (Fig. 12.7c) (Sajed et al. 2020). Machine learning combined with a smartphone provides high sensitivity, accuracy, and easy operation with ubiquitous detection using Lab-on-phone apparatus.

12.7 Conclusion

Metal nanoparticle-based colorimetric sensing has gained special attention in the detection of environmental toxins because of distinct chemical and optical characteristics, including catalytic behavior, easy synthesis, and a broad array of surface functionalization molecules. SPR-based color change associated with the aggregation of nanoparticles is the primary mechanism to develop a colorimetric sensor. Gold and silver nanoparticles exhibit distinct SPR features. On the other hand, the platinum-type nanoparticle is explored mainly for its catalytic activity and enzyme mimetic behavior to induce an indirect color change in sensing application. Among all these metallic nanoparticles, gold nanoparticles (GNP) have emerged as a versatile platform for detecting toxic metal ions because of their stability, ease of synthesis, and functionalization with biomolecules and unique optical properties. Compared to traditional methods, GNP-based colorimetric sensor provides rapid and inexpensive detection techniques specifically for the metal toxins with a detection limit of micro-to-pico molar level. Recently, a paper-based analytical device coupled with colorimetric assay emerged as a cheaper alternative to conventional methods to develop the point-of-use testing system. This technique involves immobilization of modified gold nanoparticles on paper, and metal toxins are allowed to flow through it, exploring the self-wicking property of paper substrate toward metal ion detection. The paper-based colorimetric devices provide a field-deployable miniaturized sensing platform. The developments in sensing strategies have given a new lab-on-phone concept, which includes the addition of a simple colorimetric sensing apparatus to a smartphone. This smartphone-based sensing apparatus relies on RGB (red, green, blue) features extracted from the color intensities of samples. However, the images become highly processed and compressed during feature extraction, and thus results from the smartphone-based devices cannot be trusted completely. The drawbacks of smartphone detection systems can be overcome by applying artificial intelligence systems like machine learning algorithms. ML is employed as a tool for data processing to extract features like color intensity and utilize these features to predict the concentration of toxic metal ions directly in the field of sensing.