Introduction

The growing demand for food with the increase in the world’s population has sparked the need of enlarged cultivated areas and an increase in productivity in agriculture, thus generating an increasing use of pesticides. Unfortunately, some of these pesticides are not biodegradable and have limited solubility in aqueous solutions, which can then lead to contamination of the environment, especially for groundwater and drinking water. One of such pesticides is atrazine, a widely used herbicide [1], which acts as an inhibitor of photosynthesis [2, 3] and may present endocrine disruptor or carcinogenic effects depending on its concentration [4–6]. It is not easily biodegraded due to its low solubility. Thus, it is necessary to develop more sensitive and selective analytical methods to monitor residual amounts of this pesticide.

In a search in the literature, we found distinct reports describing a few analytical methods to detect atrazine. The chromatographic methods [7, 8] show great selectivity and sensitivity; however, they demand sophisticated sample preparation. Electrochemical sensors using mercury [9, 10] and solid electrodes [11–14] are the most explored for atrazine detection. However, besides the mercury, their detection limit is normally around 10−6 to 10−9 M, which means that in many cases such sensors are not useful to ensure quality control of drinking water. The low concentrations allowed in drinking water according to regulations in different countries vary from 3 μg/L (1.39 × 10−8 M or 3 ppb) in the USA, determined by the US Environmental Protection Agency (USEPA) [15], to 0.1 μg/L (4.64 × 10−10 M or 0.1 ppb) in the European Union [16].

Since atrazine possesses a relatively strong Raman signal, we were motivated to try and detect it using surface-enhanced Raman scattering (SERS) [17], in which a signal enhancement between 103 and 106 can be achieved [18, 19], making it possible to be applied to diluted systems [20, 21]. The enhancing phenomenon is achieved by placing the target molecules close enough (touching) to metallic nanoparticles. The enhancement factors depend on the size, shape, and aggregation of nanoparticles; dielectric functions of the surrounding medium and of the metal (nanoparticle); and the frequency of the exciting radiation. Nanoparticles of noble metals are normally applied with exciting radiations (lasers) in the visible range [18, 19], as these nanoparticles can sustain localized surface plasmon resonances (LSPRs) that are able to amplify the electric field surrounding them. This, in turn, increases the induced dipole of the target molecule placed near the metallic nanoparticle, which is in the origin of the SERS phenomenon [18, 19]. For molecules placed in the interstices of nanoparticles (hot spots), the enhancement factor can reach up to 1010, with which single molecules can be detected [22–24]. SERS has been applied in sensing experiments under various experimental conditions, with isolated nanoparticles, colloids, and films, and for a variety of target molecules, including pesticides [20, 25–28].

In this study, we report on atrazine detection with SERS using Ag colloids for concentrations between 10−7 and 10−12 M. For comparison purposes, we also used impedance spectroscopy with a sensor array as in typical work of electronic tongues [29, 30] and cyclic voltammetry [31], in a methodology that has been applied to detect environmental pollutants. In such measurements, the sensing units were made of a bare electrode (Pt interdigitated for impedance and indium tin oxide (ITO) for cyclic voltammetry), and electrodes were coated with layer-by-layer (LbL) films containing gold or silver nanoparticles (AuNPs or AgNPs). The cartoon in Fig. 1a, b illustrates the approach applied to SERS, impedance, and voltammetric measurements. The data were analyzed using multidimensional projection techniques in order to demonstrate the viability of distinguishing trace amounts of atrazine, whose molecular structure is given in Fig. 1c.

Fig. 1
figure 1

a Atrazine in Ag colloid for SERS measurements. b Immersion of LbL film of PAH/AgNPs and PAH/AuNPs into atrazine aqueous solution for impedance spectroscopy (electrode: Pt interdigitated coated by LbL film) and cyclic voltammetry (electrode: ITO coated by LbL film) measurements. PAH. c Atrazine molecular structure

Full size image

Materials and methods

Reagents

The following reagents were acquired from Sigma-Aldrich: silver nitrate (AgNO3, molar mass (MM) = 169.88 g/mol), gold(III) chloride trihydrate (HAuCl4 · 3H2O, MM = 393.83 g/mol), sodium citrate (C6H5Na3O7 · 2H2O, MM = 294.10 g/mol), hydroxylamine hydrochloride (NH2OH · HCl, MM = 69.49 g/mol), sodium hydroxide (NaOH, MM = 40.00 g/mol), and poly(allylamine hydrochloride) (PAH, MM = 15,000 g/mol) polyelectrolyte. The atrazine pesticide (C8H14ClN5, MM = 215.68 g/mol), purity = 98.8 %, was purchased from Fluka Analytical. All the chemicals were used without further purification. Ultrapure water with resistivity of 18.2 MΩ cm and pH 5.6, acquired from a Milli-Q system, model Simplicity, was used to prepare the solutions and LbL films.

Au colloid synthesis by citrate reduction

The AuNPs were obtained by citrate reduction of HAuCl4, following the method proposed by Lee and Meisel [32], and used as synthesized. They were prepared by dissolving 240 mg of HAuCl4 in 500 mL of ultrapure water and heated up to boiling point. Then, 50 mL of an aqueous solution of sodium citrate 1 % water/volume (w/v) was added into the first solution under stirring, which was kept boiling and stirring for 1 h [32]. The final concentration of the AuNPs dispersion is ca 1.0 mM.

Ag colloid synthesis by hydroxylamine reduction

The AgNPs obtained by hydroxylamine reduction were synthesized according to the methodology described by Leopold and Lendl [33]. The synthesis consisted in adding 4.5 mL aqueous solution of NaOH 0.1 M, at room temperature, into 5 mL of NH2OH · HCl 43.3 mM solution. This solution was added to 90 mL of AgNO3 at 1.2 mM under stirring. The final concentration of AgNPs dispersion is ca 1.0 mM [33].

Impedance spectroscopy

The sensing units for impedance spectroscopy characterization were fabricated by the modification of Pt interdigitated electrodes through deposition of LbL films, which were grown following the experimental procedure reported in Aoki et al. [34]. Briefly, the electrode was immersed into distinct solutions according to the following sequence: PAH solution (3 min) → ultrapure water gently stirred to remove nonadsorbed PAH (1 min) → AgNPs or AuNPs colloidal dispersion (3 min) → ultrapure water to remove nonadsorbed nanoparticles (1 min). This four-step sequence led to the first bilayer, and multilayered LbL films were grown by repeating this sequence. The growth of PAH/AgNPs and PAH/AuNPs is depicted in Fig. SI1 in the Supporting Information.

The impedance spectroscopy measurements were carried out with a Solartron 1260A impedance analyzer. The sensor array was composed of three sensing units: a bare Pt interdigitated electrode and Pt interdigitated electrodes coated with five-bilayer LbL films of PAH/AuNPs and PAH/AgNPs. The bare Pt electrode is used to monitor any change in the electrical response caused by the thin films. This sensing array was applied to discriminate ultrapure water and atrazine aqueous solutions at 1.1 × 10−10, 1.1 × 10−9, 1.7 × 10−8, 5.6 × 10−8, and 1.1 × 10−7 M prepared by adding aliquots of atrazine aqueous solutions from a stock solution into ultrapure water. Impedance was measured as a function of frequency from 1 up to 106 Hz, with AC voltage of 50 mV. The Pt interdigitated electrodes contained 50 pairs of digits, each digit being 10 μm wide, 0.5 mm long, and 100 nm high, spaced by 10 μm from each other. The theoretical fitting of the experimental curves was carried out using the ZView software.

Raman scattering and UV-Vis absorption spectroscopy

Micro-Raman analysis and optical microscopy were obtained using a micro-Raman Renishaw spectrograph, model inVia, equipped with a Leica microscope, whose × 50 objective lens allows for collecting the spectra with ca 1 μm2 spatial resolution. The spectrograph contains a charge-coupled device (CCD) detector, laser at 633 nm, 1,800 grooves per millimeter grating with additional edge filters. UV-Vis absorption spectra were collected using a spectrophotometer Varian, model Cary 50.

Solutions for SERS measurements

An atrazine stock solution at 1.0 × 10−4 M was diluted in Ag colloid to achieve the SERS effect. The final solution concentrations of atrazine were 5.0 × 10−7, 5.0 × 10−8, 5.0 × 10−9, 5.0 × 10−10, 5.0 × 10−11, and 5.0 × 10−12 M. The stock solution was prepared by dissolving 5.40 mg of atrazine in 250 mL of ultrapure water under sonication. In order to acquire SERS spectra, a small droplet of atrazine solution was placed in a holder under the microscope, and the laser focus was adjusted onto the air/water interface.

Data analysis

Data from sensing and biosensing experiments have long been treated with statistical methods, such as principal component analysis (PCA), which is a statistical tool employed to reduce the dimensionality of the data [35]. For biosensing, in particular, other multidimensional projection techniques have been proven superior to PCA [36], especially in the cases where an optimization procedure is exploited. In these methods that are said to belong to the realms of information visualization, one attempts to project data from a multidimensional space onto a 2D or 3D plot with maximum preservation of similarity relationships. Formally, the data in the original space are represented by X = {x1,x2,..,xn}, and δ(xi, xj) is defined as the distance between two data instances i and j. They are projected onto a 2D plot with graphical markers represented by Y = {y1,y2,..,yn}, which are determined in an optimization procedure using an injective function f: X → Y that minimizes |δ(xi, xj) − d(f(xi),f(yj))| ≈ 0, ∀xi, xj ∈ X [37], where d(yi, yj) is the distance function on the projected plane. The flexibility of this optimization approach arises from the availability of several cost (or error) functions used for placing the graphical markers on the 2D plot. Here, we used the so-called Sammon’s mapping [38] and Interactive Document Map (IDMAP) [39], whose error functions are defined, respectively, as follows:

$$ {S}_{S am}=\frac{1}{{\displaystyle \sum_{i< j}\delta \left({x}_i,{x}_j\right)}}{\displaystyle \sum \frac{{\left( d\left({y}_i,{y}_j\right)-\delta \left({x}_i,{x}_j\right)\right)}^2}{\delta \left({x}_i,{x}_j\right)}} $$
(1)

where δ and d are the distance functions defined above

and

$$ {S}_{IDMAP}=\frac{\delta \left({x}_i,{x}_j\right)-{\delta}_{\min }}{\delta_{\max }-{\delta}_{\min }}-\left({y}_i,{y}_j\right) $$
(2)

where δ min and δ max are the minimum and maximum distances between the samples.

Cyclic voltammetry

The voltammetric measurements were carried out in a voltammetric cell of three electrodes: Ag/AgCl as the reference electrode, Pt wire as the counter electrode, and ITO electrode modified with a ten bilayer LbL film of PAH/AgNPs or ten bilayer LbL film of PAH/AuNPs as working electrodes. The study of the interaction of atrazine/modified electrode was carried out by adding aliquots from 4.95 × 10−7 M of atrazine standard solution containing 0.1-M KCl solution as supporting electrolyte (pH 1.88). The analyses were conducted in continuous flux of high-purity nitrogen gas. All solutions were prepared using the ultrapure water (18.2 MΩ cm).

Results and discussion

Sensing atrazine with SERS

SERS was applied to detect atrazine solutions down to 5 × 10−12 M concentration, using Ag colloid. Figure 2 shows the SERS spectra collected for all atrazine concentrations diluted in the Ag colloid. A typical spectrum for the Ag colloid is given as reference for the identification of the atrazine vibrational bands. The several Ag colloid spectra recorded are given in the Supporting Information (Fig. SI2). There is an overlap of bands between 1,339 to 1,619 cm−1 for atrazine and Ag colloid spectra. According to Costa et al. [40], atrazine bands in this spectral region are assigned to CH3 + CH + NH bending at 1,339 cm−1, NH + CH bending + CH2 wagging at 1,365 cm−1, CH3 bending at 1,453 cm−1, and NH bending + C–N stretching at 1,619 cm−1.

Fig. 2
figure 2

SERS spectra of atrazine solutions diluted in Ag colloid at 5.0 × 10−7, 5.0 × 10−8, 5.0 × 10−9, 5.0 × 10−10, 5.0 × 10−11 and 5.0 × 10−12 M concentrations. The Ag colloid spectrum is given as a reference at the bottom in order to distinguish the vibrational bands of atrazine

Full size image

Vibrational bands arising only from SERS of atrazine solution are highlighted by dotted lines, with the following assignments: CH2 rocking + CH3 wagging at 751 cm−1, C–C stretching + CH3 wagging at 800 cm−1, CH3 twisting at 917 cm−1, C–C stretching at 964 cm−1, NH bending at 1,183 cm−1, CH2 twisting at 1,275 cm−1, and NH bending at 1,505 cm−1 [40]. The SERS spectrum of atrazine solution at 5.0 × 0−7 M is compared in Fig. SI3 with the SERS spectrum of the cast film made with the same solution concentration and with the Raman spectrum of atrazine powder. The main vibrational bands in the SERS spectra for the solution and cast film and in the conventional Raman for the powder are highlighted by dotted lines. Most of SERS bands are shifted in relation to the Raman bands of the atrazine powder, which is explained by the different environments for the atrazine molecules, as discussed further later on in this paper. The average intensity of SERS was higher for the spectra collected in solution (5.0 × 10−7 M) than that in the cast film, which means that the conformation of atrazine molecules adsorbed on AgNPs plays a central role in this ultrasensitive analysis [40, 41].

According to Costa et al. [40], the isopropyl group of atrazine molecule causes a steric effect, avoiding the approximation of AgNPs in this region of the molecule. They proposed that the atrazine molecule adsorbs onto AgNPs via the N between the ethyl radical and the Cl atom. This hypothesis was confirmed by Bonora et al. [41], which observed a band at 1,610 cm−1 assigned to the deformation of the ring due to the vibration of the C–N–H atoms positioned between the ethyl radical and Cl atom. Here, we can assume the same adsorption mechanism of the atrazine molecule onto AgNPs once the observed bands in our SERS spectra are all related to the molecular vibrations at this position.

We have also performed a quantitative computational analysis of the SERS spectra using a multidimensional projection technique, whose details are given in Paulovich et al. [42] and Oliveira et al. [36]. The suitability of SERS to distinguish atrazine solutions demonstrated in Fig. 2 is revealed by treating the SERS spectra using the Sammon’s mapping [38] technique to project the data in a 2D plot, as shown in Fig. 3. We also obtained the visualization with IDMAP, but the distinction appeared slightly superior with the Sammon’s mapping. The results are clustered according to the similarity of analyzed solutions. The closer the circles, the more similar the SERS spectra are. The spectra corresponding to different atrazine concentrations are positioned apart from each other, indicating that these samples can be clearly distinguished down to 5 × 10−12 M. The latter highlights the usefulness of projection techniques not only to handle the data but also to optimize the sensing performance.

Fig. 3
figure 3

Sammon’s mapping layout of the SERS spectra obtained from distinct atrazine solutions. No labels are assigned to the axes because in this type of plot what matters is the relative distance between data points, where similar samples should be placed close to each other

Full size image

The large dispersion in the data for the lowest concentration (5 × 10−12 M) could be expected on the basis of fluctuations on the SERS spectra affecting bandwidth, band shape, Raman shift, and absolute and relative intensities when approaching highly diluted solutions. Indeed, we observed such SERS spectral variation in experiments of single-molecule detection, which is associated with local changes of the molecular environment [43]. For 5 × 10−12 M, the number of atrazine molecules per picoliter (10−12 L) is estimated to be three, where the picoliter scale was chosen because this is the order of magnitude of the volume probed by the laser in single-molecule experiments [44, 45]. To our knowledge, this is the lowest concentration ever detected for atrazine solutions, approaching single-molecule levels. In fact, only a few reports regarding atrazine detection by SERS are found [40, 46]. In references Costa et al. [40] and Bonora et al. [41], the main goal was to determine how the atrazine is adsorbed onto the nanoparticle surfaces under distinct experimental conditions, and the atrazine concentrations used were 10−7 M and in the ppm range, respectively. Carrillo-Carrion et al. [46] coupled chromatography and SERS for sensing experiments, reaching a detection limit of ca 0.9 μM.

Sensing atrazine with impedance spectroscopy

The combination of interdigitated electrodes, ultrathin films, and impedance spectroscopy is a well-established approach to detect trace levels of different analytes, including water pollutants [30, 47]. As a complementary approach, the performance of PAH/AgNPs LbL films as transducers in sensing units was checked by immersing it into aqueous solutions containing atrazine at different concentrations. Besides the five bilayers of the PAH/AgNPS LbL film, the sensor array was composed of a bare Pt interdigitated electrode and a Pt electrode coated with five bilayers of PAH/AuNPs. Figure 4a, b shows the Bode plots with capacitance versus frequency (C vs. f) curves and the Nyquist plots (Z″ vs. Z′), respectively, for all sensing units immersed into ultrapure water. As expected, even nanostructured films affect the electrical response of the Pt electrode, whose impedance curves were modeled with electrical equivalent circuits (EEC)s inspired in the work by Taylor and Macdonald [48]. The circuit in Fig. 4c corresponds to an electrode coated by an insulating film dipped in an electrolyte solution, with the film being represented by the elements CPEb and Rb in series with the electrolyte impedance. The latter contains three components: the geometric capacitance of the electrode immersed in an electrolyte (CPEg), the electrical double layer capacitance (CPEd), and the total resistance of the electrolyte represented by (1/Rd + 1/Rt), where Rd is the electrical double layer resistance and Rt is the charge transfer though the interface electrolyte/film. The electrical response is governed by different physicochemical phenomena depending on the frequency range, as follows [49]: At the low-frequency region (100–102 Hz), impedance is mainly related to the formation of an electrical double layer at the film/solution interface [49]; in the mid-frequency region (102–105 Hz), it is related to the properties of the film coating the electrode [49], and at the high-frequency region (>105 Hz), the impedance response is dominated by the geometric capacitance of the interdigitated electrodes [49] and by the dielectric constant of the solution [50].

Fig. 4
figure 4

a Capacitance versus frequency (C vs. f) curves, b Nyquist plot (Z″ vs. Z′) for all sensing units immersed into ultrapure water, and c electrical equivalent circuit proposed by Taylor and Macdonald. Dotted curves are the experimental results, and solid lines represent the theoretical fittings in a and b. In c CPEg is the geometric capacitance of the electrode, CPEd is the electrical double-layer capacitance, Rd is the electrical double-layer resistance, Rt is the charge transfer resistance through the electrolyte/film interface, CPEb and Rb are the capacitance and resistance of the film coating the electrode, respectively

Full size image

The major changes induced by the LbL films occur at low and intermediate frequencies, as predicted by the model of Taylor and Macdonald. The parameters used for fitting are shown in Table 1, where optimized fitting was obtained upon removing the Rb element (film resistance) in the EEC, suggesting that the sensing units used here present a dominant capacitive behavior. Upon film deposition, the charge transfer resistance Rt decreased by one order of magnitude, probably owing to the presence of metallic nanoparticles, while the double-layer resistance Rd increased owing to an increased surface area for the coated electrodes, which also caused the film capacitance CPEb to decrease.

Table 1 Parameters used for fitting the impedance curves of bare Pt electrode, PAH/AgNPs, and PAH/AuNPs sensing units immersed in ultrapure water
Full size table

The capacitance versus frequency curves for the sensing units made with LbL films PAH/AgNPs and PAH/AuNPs are shown in Fig. 5, and the data for the bare electrode are not shown lest the plot would be overcrowded. A visual inspection does not allow one to infer whether it is possible to distinguish the samples with different atrazine concentrations, which is the reason why we resorted to statistical and computational methods for analyzing the data.

Fig. 5
figure 5

Capacitance versus frequency curves for a PAH/AgNPs and b PAH/AuNPs sensing units immersed into atrazine aqueous solution with concentrations from 1.1 × 10−10 up to 1.1 × 10−7 M

Full size image

When the data for the three sensing units are combined, the projections onto a 2D plot with PCA [35] and IDMAP [39] show clear separation according to the atrazine concentration. This is illustrated in Fig. SI4 in the Supporting Information for PCA and in Fig. 6 for IDMAP, in both cases with a shift of the clusters to the right with increasing atrazine concentrations. We also used Sammon’s mapping for comparison, but the results with IDMAP were superior. An interesting feature in the plot is that a measurement with ultrapure water at the end of the experiments indicated that the sensing units were affected by the measurements since the data points for “final water” did not coincide with those of the “initial water.” Therefore, though it is clear that the addition of small aliquots of atrazine solution to increase the concentration leads to changes in the electrical response, thus corresponding to a high sensitivity, such changes are irreversible. As in similar work with other analytes [45, 30], irreversible adsorption on the LbL films may occur, with adsorbed molecules not being removed in the washing procedures. In Aoki et al. [45], such adsorption was confirmed with SERS experiments, but this could not be done here because the SERS signal for the LbL films was negligible. Caution should then be taken in analytical procedures, for the results may depend on the order with which the experiments are performed. This is clearly a disadvantage of the impedance spectroscopy, compared with SERS which does not suffer from this limitation. Owing to this limitation, we have not attempted to detect even lower concentrations of atrazine to try and match the experiments for SERS.

Fig. 6
figure 6

IDMAP projection of the capacitance data obtained a from all sensing units combined in a single projection to discriminate the solution concentrations. b Data of all sensing units separately projected on a single graph

Full size image

The IDMAP technique can also be used to identify the contribution from each sensing unit for the distinction ability. The color of the circles denotes the sensing unit as indicated in the label, while the concentration is shown next to each circle in the plot. The closer the circles, the more similar the electrical responses are. The latter allows a direct visualization of the performance of each sensing unit in discriminating atrazine solutions. The bare Pt electrode is positioned apart from the other sensing units, confirming that the modification introduced by coating the Pt electrodes is significant. The results of the sensing unit composed of the PAH/AgNPs LbL film are placed separately from those of the PAH/AuNP sensing unit. There are no overlapping areas among those data. Therefore, the replacement of AgNPs for AuNPs in the LbL films gives rise to distinct data points in the plot, showing that both sensing units play a role, i. e., there is clear dependence of the electrical response on the metallic nanoparticles.

Sensing atrazine with cyclic voltammetry

Complementary, Fig. 7 shows voltammograms obtained to check the sensing performance of the cyclic voltammetry technique in terms of atrazine detection using the same kind of thin films applied to impedance spectroscopy, i. e., PAH/AuNPs and PAH/AgNPs LbL films, however, now to modify the ITO electrodes (instead of Pt interdigitated electrodes). The cyclic voltammograms obtained for both modified electrodes show one anodic peak at −571 mV versus Ag/AgCl, which may be associated with InO3 at low pH values (InO3 present on ITO electrode surface). The PAH/AgNPs-modified electrode presented a redox process at E pa = 148 mV and E pc = −42 mV versus Ag/AgCl, which is ascribed to the oxidation of Ag and the reduction of Ag oxides on the ITO electrode surface [51], while the PAH/AuNPs-modified electrode presented a redox process at E pa = 895 mV and E pc = 654 mV versus Ag/AgCl, which is ascribed to reduction of Au oxides at the nanoparticles [52]. In the presence of atrazine in solution, the PAH/AgNPs-modified electrode showed a nonlinear variation of the current values with an increase of atrazine concentration. Thus, it was not possible to obtain the analytical curve. Anyway, the lowest concentration that we were able to measure was limited at 10−9 M. On the other hand, a great decrease of current values in the presence of atrazine was observed for PAH/AuNPs-modified electrodes, which may be related to loss of charge transfer. Indeed, the AgNPs showed to be more sensitive to the presence of atrazine in solution, however, without a linear variation with atrazine concentration, while AuNP response was not favorable to atrazine determination. The latter results reinforced the potential of SERS as a suitable tool to detect atrazine above and below its concentration limit allowed for drinking water.

Fig. 7
figure 7

a Cyclic voltammograms of ITO bare electrode, ten bilayers of PAH/AgNPs and ten bilayers of PAH/AuNPs-modified electrodes in the absence of atrazine in solution. Cyclic voltammograms for b PAH/AgNPs and c PAH/AuNPs in the presence of different atrazine concentrations. All measurements were carried out at 0.1-M KCl solution as supporting electrolyte at pH 1.88 and with 50 mV s−1. d Variation of absolute current values of PAH/AgNP-modified electrode versus atrazine concentration

Full size image

Conclusions

We have shown that SERS can be used to detect atrazine in AgNPs colloid with the lowest concentration ever reported to the best of our knowledge, where distinction was made absolutely clear using a Sammon’s mapping plot. Most significantly, distinguishing an atrazine concentration down to 5 × 10−12 M, as was done here, is sufficient to monitor the quality of drinking water even according to the most stringent international regulations and is an important step toward single-molecule detection. The importance of detection via SERS was further illustrated in a comparative study where atrazine was detected in aqueous solutions using the e-tongue concept with impedance spectroscopy and cyclic voltammetry applied to LbL films made with AuNPs and AgNPs. Though distinction of the different concentrations of atrazine could be made using impedance spectroscopy, we noted that the electrodes were irreversibly affected by the measurements, a limitation that does not apply to SERS. Finally, the analysis of SERS data with information visualization methods, which only recently has been introduced [43] and was exploited here, paves the way for detecting atrazine and other pesticides in real samples of drinking water.