All-Solid-State Potentiometric Sensor Based on Graphene Oxide as Ion-to-Electron Transducer for Nitrate Detection in Water Samples

Abstract

Graphene oxide (GO) was used as an ion-to-electron transducer for all-solid-state nitrate electrodes based on an alkyl ammonium salt as the sensing element. Commercially available carbon screen-printed electrodes modified with GO were used as conductive substrates, whose morphology and distribution along the surface were evaluated by scanning electron microscopy and Raman spectroscopy. The potentiometric performance of the GO-based electrodes revealed a Nernstian slope of −53.5 ± 2.0 mV decade⁻¹ (R² = 0.9976 ± 0.0015) in the range from 3.0 × 10⁻⁶ to 10⁻² M and a lower limit of detection of 1.9 × 10⁻⁶ M. An impressive reproducibility between equally prepared electrodes (n = 15) was demonstrated by a variation of <6% for the calibration parameters. Constant current chronopotentiometry and water layer tests were used to evaluate the potential signal stability, providing similar performance to previously published works with graphene-based ion-selective electrodes. Notably, the GO-based sensors showed the absence of a water layer, a long-term drift of 0.3 mV h⁻¹, and a stable performance (LOD and sensitivity) over 3 months. The applicability of the proposed sensors was demonstrated in determining nitrate levels in water samples with great accuracy, yielding recovery values from 87.8 to 107.9%, and comparable (p > 0.05) results to a commercial nitrate probe. These findings demonstrate the use of GO as an alternative ion-to-electron transducer for the fabrication of all-solid-state potentiometric electrodes.

1. Introduction

Nitrate ions (NO₃⁻) are primary compounds involved in the biogeochemical nitrogen (N) cycle that occurs in natural ecosystems like soils and aquatic environments [1]. However, the development of modern agriculture has expanded the use of inorganic fertilizers, increasing the levels of NO₃⁻ in surface and groundwater resources [2]. This contamination causes disturbances in the ecological N cycle, and excessive nutrient concentration in aquatic systems leads to uncontrolled algae growth, resulting in oxygen depletion and poor water quality [3]. This phenomenon, known as eutrophication, has negative consequences for biodiversity, fisheries, and recreational activities. On the other hand, the presence of NO₃⁻ in natural waters at levels higher than 50 mg L⁻¹, which is the guideline value defined by the European Union (EU) [4], makes them inappropriate for consumption. Their ingestion may cause severe adverse effects on human health, including reproductive problems, methemoglobinemia, colorectal cancer, and thyroid disease [5,6]. NO₃⁻ content is thus a critical water quality indicator, and its monitoring is crucial for water ecosystem preservation, aquaculture system efficiency, and human safety.

Current standard analytical methodologies for determining NO₃⁻ are based on liquid chromatography, capillary electrophoresis, or spectrophotometry [7,8]. However, these methods require expensive instrumentation and laborious sample pre-treatment, increasing the time and cost per analysis. Electrochemical sensors are an exciting approach to overcoming these issues, particularly potentiometric ion-selective electrodes (ISEs), due to their simplicity of use, cost-effectivity, wide linear analytical range, short response time, and applicability in turbid samples [9,10,11,12]. Despite their successful use since the 1970s [13,14,15], the presence of an internal filling solution limited their application in decentralized measurements due to high maintenance and complications with miniaturization. These early ISEs evolved into all-solid-state ISEs by replacing the inner liquid solution with a solid substrate between the ion-selective membrane (ISM) and the conductive wire, paving the way for miniaturization and incorporation into portable platforms [11,16]. However, the lack of a thermodynamically well-defined electrochemical interface and/or the formation of a thin water layer between the ISM and the inner electron conductive substrate results in potential drift and poor reproducibility, restricting their long-term application [17,18,19]. These issues have led to significant research into ion-to-electron transducers that have been incorporated directly into the ISM or introduced as intermediate layers between the ISM and the conductive substrate [20,21].

Among the available materials, conducting polymers (CPs) have widespread popularity as they are inexpensive, possess good electrical properties, and their conductivity can be readily tuned via structural modifications or doping with ions [22]. Poly(3-octylthiophene), polypyrrole, and polyaniline are the most representative examples [23,24,25]. However, they are usually vulnerable to light, redox species, CO_2, and O₂, and their hydrophilicity is prone to generating a water layer, which may result in a potential drift [16,26]. On the other hand, poly(3,4-ethylenedioxythiophene) (PEDOT) is insensitive to light, oxidation, and pH [27], but it has been formed onto the electrode’s surface by electropolymerization, which is a tedious fabrication process compared to the easy and scalable process of drop-casting [28]. Despite the unquestionable benefits of CPs’ application as transducers, the field is open for other materials to ensure stable potential readings of all-solid-state potentiometric sensors.

A new generation of solid-contact transducers has recently emerged based on the use of nanostructured materials [29], such as carbon nanotubes [26], macroporous carbon [30], gold nanoparticles [31], and graphene [32]. Their inherent advantages, such as insensitivity to redox species, light, and hydrophobicity, are crucial characteristics for in situ measurements [33]. While carbon nanotubes [26,34] and reduced graphene oxide [35,36,37] have been investigated as solid-contact transducer materials in NO₃⁻ potentiometric sensors, graphene oxide (GO) in its native form has not been reported yet. The insulating properties and relative hydrophilicity [38] open up the question of a possible effect on the potential stability and electrochemical parameters of all-solid-state sensors. Nevertheless, GO can be an attractive material for application as a transducer in potentiometric sensors due to an additional pseudo-capacitance effect of attached oxygen-containing functional groups compared with graphene [39]. Moreover, the potentialities of GO increase if the functional groups can exchange ions in the ISM, as already demonstrated for CPs [23], carbon nanotubes [40], and carboxy-functionalized graphene [41].

Herein, this work presents an all-solid-state NO₃⁻–ISE based on GO as an ion-to-electron transducer and a plasticized membrane based on a quaternary ammonium salt as a sensing material. Commercially available carbon screen-printed electrodes (CSPE) modified with GO were used as conductive substrates. The analytical performance and electrochemical properties were compared with those obtained from the CSPE modified only with the sensing membrane. The proposed sensors were used to determine NO₃⁻ in drinking and groundwater samples.

2. Materials and Methods

2.1. Chemicals and Solutions

All chemicals were of analytical reagent grade and were purchased from Sigma-Aldrich (Portugal): sodium nitrate (NaNO₃, ref. S5506), sodium nitrite (NaNO₂, ref. S2252), sodium chloride (NaCl, ref. S9625), sodium carbonate (Na₂CO₃, ref. 223530), sodium sulfate (Na₂SO₄, ref. 239313), disodium hydrogen phosphate (Na₂HPO₄, ref. S9763), sodium dihydrogen phosphate (NaH₂PO₄, ref. 71505), sodium bromide (NaBr, ref. 220345), sodium iodide (NaI, ref. 217638), sodium perchlorate (NaClO₄, ref. 310514), sodium hydroxide (NaOH, ref. S5881 71690), and phosphoric acid (H₃PO₄, ref. W290017). The materials to prepare the ISMs were of Selectophore grade, namely tetradodecylammonium nitrate (TDDA-NO₃, ref. 87252), bis(2-ethylhexyl)sebacate (DOS, ref. 84818), poly (vinyl chloride) (PVC, ref. 81395), and tetrahydrofuran (THF, ref. 87369).

All solutions were prepared by dissolving the appropriate salts in 18.2 MΩ cm⁻¹ double-deionized water (Milli-Q water systems, Merck Millipore, Algés, Portugal). The stock standard solution of NO₃⁻ was prepared at a concentration level of 1 M and stored at room temperature. The calibration working solutions were prepared by dilution to the final concentration of 10⁻³ and 10⁻¹ M. The influence of pH on sensor response was investigated by adjusting the pH value with concentrated H₃PO₄ or NaOH (measured with a SevenCompact™ S210 pH Meter, Metter Toledo, Cornellà de Llobregat, Barcelona, Spain).

2.2. Preparation of All-Solid-State NO₃⁻–ISEs

The all-solid-state ISEs were based on commercially available carbon screen-printed electrodes (CSPE, ref. DRP-110) and CSPE modified with graphene oxide (CSPE/GO, ref. 110GPHOX), which were purchased from Metrohm Dropsens (Gomensoro Potencial Zero, Lisbon, Portugal).

The ISM for the NO₃⁻ was prepared by dissolving 4.0 wt% TDDA-NO₃, 31.0 wt% PVC, and 65.0 wt% DOS in 3 mL THF (total mass of 300 mg). The ISM cocktail was drop-casted (4 × 5 µL) onto the conductive surface relative to the working electrode (CSPE/GO/ISM). Each layer was allowed to dry for 10 min at room temperature for THF evaporation. For comparison, all-solid-state ISEs without GO were also prepared by coating the bare CSPE with the ISM directly on the carbon working area (CSPE/ISM). Finally, all electrodes were conditioned overnight in 10⁻³ M NaNO₃ solution before the measurements (~12 h). When not in use, they were stored dry at room temperature and protected from the light. Five identical electrodes were prepared and examined for each type of sensor configuration. Ten additional electrodes of the CSPE/GO/ISM configuration were prepared for a critical assessment of the reproducibility between electrodes.

2.3. Characterization of GO Solid-Contact

Raman spectroscopy was performed using an Alpha 300R Access Witec miniconfocal Raman microscope (WITec, Ulm, Germany). The Raman spectra were acquired in six locations on the bare CSPE and CSPE/GO for 0.5 s and 20 accumulations using a 633 nm excitation laser line, 300 line mm⁻¹ grating, and a 20x objective. Raman spectra were processed by correcting the baseline and cosmic ray removal as well as obtaining the average spectrum from 3 to 6 single spectra using WITec Project 5.2 software and Spectragryph v1.2.16.1 software.

Scanning electron microscopy (SEM) analysis was performed on the bare CSPE and CSPE/GO using a FEI Quanta 650 FEG field emission microscope and an accelerating voltage of 10 kV. Energy-dispersive X-ray spectroscopy (EDX) was also executed to generate information about the chemical composition of the electrode’s surface.

Electrochemical characterization was carried out by cyclic voltammetry (CV) using an EmStat4S potentiostat connected to a computer equipped with PSTrace software 5.9 version (PalmSens BV, Houten, The Netherlands). The measurements were performed in a solution of 5 × 10⁻³ M ferro/ferricyanide prepared in 10⁻² M PBS using a three-electrode electrochemical cell composed of a single junction Ag/AgCl/KCl 3 M (MF-2056, BASI Inc., West Lafayette, IN, USA) as a reference electrode, a platinum wire (MW-1032, BASI Inc., USA) as an auxiliary electrode, and bare CSPE or CSPE/GO as a working electrode. CV cycles were recorded between −0.5 and +0.5 V at a scan rate of 100 mV s⁻¹. All experiments were performed at a controlled room temperature (20 ± 1 °C).

2.4. Evaluation of All-Solid-State NO₃⁻–ISEs

The potentiometric measurements were performed using the EmStat4S potentiostat (PalmSens BV, Houten, The Netherlands) at a controlled room temperature (20 ± 1 °C). PS Trace software 5.9 version was also used for data acquisition and visualization. The electromotive force (EMF) was measured in stirred solutions (every 1 s) by the potential difference between the all-solid-state NO₃⁻–ISEs and the Ag/AgCl/KCl 3 M reference electrode. The former was connected to the potentiostat using a flexible cable (ref. CAC) acquired from Metrohom Dropsens (Gomensoro Potencial Zero, Algés, Portugal). Dynamic calibration curves were constructed for NO₃⁻ in the ion concentration range from 10⁻⁷ to 10⁻² M. Activity coefficients were calculated using the Debye–Hückel approximation to transform experimental concentrations into activities [42]. The measured EMF at the steady state was plotted against each logarithmic activity to determine the sensitivity (slope), intercept (standard potential, E⁰), and linear range of response (LRR). The lower limit of detection (LOD) was calculated as the activity related to the cross point between extrapolation of the lines defining the non-responsive range and LLR of the electrode [43]. All the calibrations were carried out in triplicate (i.e., three subsequent calibrations using the same electrode).

The response time was determined as the time that elapses between the instant when the activity of the target ions is changed and the first instant when the ∆$E$/∆t becomes equal to 1.0 mV min⁻¹ [43]. The potentiometric selectivity coefficients were evaluated using the fixed interference method (FIM) at two interfering concentration levels (10⁻⁴ and 10⁻² M) [44]. The reversibility was evaluated by switching the NO₃⁻ concentration between 10⁻⁴ and 10⁻³ M and performing calibration curves from higher to lower concentrations and vice versa. The influence of pH on the potentiometric response was investigated at 10⁻³ M NaNO₃ solution by measuring the EMF after changing the pH value from 2 to 11 with the addition of concentrated H₃PO₄ or NaOH.

The formation of a water layer between the ISM and the underneath conductive substrate was carried out on both CSPE/ISM and CSPE/GO/ISM configurations [19], which were previously conditioned overnight (~12 h) in 10⁻¹ M NaNO₃ solution. The EMF was firstly measured in 10⁻¹ M NaNO₃ solution for 2 h, then in 10⁻¹ M NaCl solution for 2 h, and finally in 10⁻¹ M NaNO₃ solution for 16 h (i.e., long-term drift). The potential stability was also evaluated by changes in the general analytical parameters after the repetitive calibration for 12 weeks.

The capacitance of the all-solid-state-ISEs was evaluated along with the short-term potential stability by reversed-current chronopotentiometry in a 10⁻¹ M NaNO₃ solution using phosphate buffer as the electrolyte (10⁻¹ M and pH 5.0) [45]. A constant current of +1 nA was applied on the working electrode for 60 s, followed by −1 nA current for another 60 s, while the EMF was continuously recorded. The resistance in charge transfer was studied in the same NaNO₃ solution by EIS in the frequency range from 0.1 Hz to 100 kHz with a potential amplitude of 10 mV at the open-circuit potential. All impedance spectra were fitted to equivalent electrical circuits using PSTrace software 5.9. The same EmStat4S potentiostat, together with the Ag/AgCl/KCl 3 M reference electrode and the platinum wire as an auxiliary electrode, were used for these measurements.

2.5. Water Samples Analysis

The determination of NO₃⁻ content was carried out in different water samples: three groundwater samples were collected from different wells located in the North of Portugal, one tap water sample was collected directly from our research institution, and commercial water was purchased in a local supermarket. The NO₃⁻ concentrations were measured as soon as collected (on the same day). The well water samples were filtered through 0.45 μm pore size filters coupled to syringes before the analysis, while the others were directly tested. The CSPE/GO/ISM sensors were calibrated in a 10⁻¹ M phosphate buffer to ensure an adequate performance. After calibrating, the EMF was measured in each sample three times and interpolated in the previous calibration plots.

A commercial NO₃⁻–ISE inner filling solution (perfectION^TM, ref. 51344727) connected to a pH/Ion meter S220 (SevenCompact^TM, ref. 30019028) from Metter-Toledo was used as a comparative instrument. The measurements followed the manufacturer’s recommendations for proper functioning and ion determination.

3. Results and Discussion

3.1. Characterization of GO Layer

The surface morphology of the CSPE and CSPE/GO bare electrodes was characterized by SEM observation. As shown in Figure 1a, the surface of the bare CSPE is porous and displays a typical sheet layer structure. In comparison, small particles and fewer sheet-like structures are observed on the CSPE/GO (Figure 1b). Notably, a clear change was observed at the 50,000x magnification, in which the granular appearance on the CSPE (Figure 1c) highly contrasts with the smoother build-up of thin sheets on the CSPE/GO electrode (Figure 1d). Additionally, the EDX analysis revealed the presence of oxygen and carbon atoms in the CSPE/GO electrode, typical of the GO material, while only carbon atoms were found in the bare CSPE (Figure S1).

Raman analysis was performed to evaluate the presence and distribution of GO on the surface in comparison with the bare CSPE. The Raman spectra of the CSPE/GO displayed prominent D and G bands, which are typical of graphitic materials [46], in contrast with the CSPE based only on amorphous carbon (Figure 1e). These bands arise from graphene lattice defect sites (D band) and in-plane vibrations of sp2-carbon atoms (G band) [47]. The D band at 1332 cm⁻¹, with comparable intensity to the G band at 1595 cm⁻¹, indicates significant structural disorders in the electrode surface due to oxygen functional groups typical of GO. The intensity ratio of the D and G bands (I_D/I_G) was calculated as 1.00 ± 0.03, close to those reported by other authors [46,48]. Notably, the Raman results were similar over different sampling locations (Table S1), proving the distribution of GO along the surface.

The electrochemical characterization of the GO layer was performed through cyclic voltammetry (CV) in a 5 × 10⁻³ M ferro/ferricyanide solution. The cyclic voltammograms presented in Figure 1f show a decrease in oxidation–reduction current peaks in CSPE/GO, which is related to the insulating properties of GO, already reported by other authors [38]. This translates into a lower electroactive surface area, calculated from the Randles–Sevcik equation [49], for the CSPE/GO (0.08 ± 0.01 cm²) in comparison with the bare CSPE (0.12 ± 0.01 cm²). Nevertheless, the key aim of this work is to assess the applicability of GO, in its native form, as a new ion-to-electron transducer and an alternative material for the preparation of stable and robust all-solid-state potentiometric ISEs.

3.2. Potentiometric Response of GO-Based All-Solid-State NO₃–ISEs

The analytical performance of the CSPE/GO/ISM sensors was evaluated under steady-state conditions against a commercial Ag/AgCl reference electrode. The dynamic potentiometric response was recorded by the progressive addition of different amounts of NO₃⁻ within the concentration range from 10⁻⁷ to 10⁻² (Figure 2a). The EMF at the steady state was plotted against the corresponding logarithmic activity, revealing Nernstian behavior (Figure 2b). The slope calculated from the calibration plots (n = 15 from three calibrations on five identical electrodes) was −53.5 ± 2.0 mV decade⁻¹ (R² = 0.9976 ± 0.0015) over the linear range of response (LRR) from 3.0 × 10⁻⁶ to 10⁻² M, with coefficients of variation (CV) < 3.8% for the calibration parameters. The lower limit of detection (LOD) was 1.9 × 10⁻⁶ M [43]. Notably, the great repeatability of the potentiometric response, assessed by subsequent calibrations of the same electrode, was demonstrated by a variation <5.8% for the slope and the intercept (Table S2). Additionally, the calibration parameters were found to be impressively reproducible for equally prepared electrodes (slope of −52.3 ± 1.5 mV decade⁻¹ and intercept of 223.9 ± 12.1 mV, n = 15 from one calibration on fifteen sensor units), with CV < 5.4% (Table S3). Analogous analytical parameters (

Table 1. Potentiometric response characteristics of the proposed all-solid-state electrodes for detection of NO₃⁻ (n = 15 calibrations from five identical electrodes).

Parameter	CSPE/GO/ISM	CSPE/ISM
Slope (mV dec−1)	−53.5 ± 2.4	−55.4 ± 5.0
Intercept (mV)	215.2 ± 7.3	174.2 ± 33.0
Coefficient of determination (R2)	0.9976 ± 0.0015	0.9950 ± 0.0058
Linear range of response (M)	3.0 × 10−6–10−2	3.0 × 10−6–10−2
Limit of detection (M)	1.9 × 10−6	2.2 × 10−6
Working range (pH)	3–6	3–6
Response time (s)	10–20	10–20
Reproducibility slope (SD, mV)
Intra-electrode 1	<3.1	<4.1
Inter-electrode 2	2.0	5.0
Reproducibility intercept (SD, mV)
Intra-electrode 1	<3.8	<16.6
Inter-electrode 2	7.3	33.0

¹ Three consecutive calibrations performed on the same sensor; ² Fifteen calibrations performed from five equally prepared sensors. SD—Standard deviation.

), but with a slight shift in the E⁰, were achieved with the CSPE/ISM sensors: the slope of −55.4 ± 5.0 mV dec⁻¹, LRR from 3.0 × 10⁻⁶ to 10⁻² M, and LOD of 2.2 × 10⁻⁶ M (n = 15 from three calibrations on five identical electrodes).

The response time ranged between 10 and 20 s for changes at high (10⁻³ M) and low (10^−5.5 M) NO₃⁻ concentrations, respectively (inset of Figure 2a). The proposed CSPE/GO/ISM sensors exhibited enhanced potential reversibility (Figure 3a) when conducting the repetitive cycles between 10⁻⁴ and 10⁻³ M NaNO₃ solutions (standard deviation of 2.5 and 2.8 mV for the lower and high concentrations, respectively). Additionally, the reversibility was evaluated by performing calibration curves from low to high NO₃⁻ concentrations and vice versa (Figure S2). A very reproducible response was evidenced by the standard deviation of 0.3 mV and 1.0 mV for slope and intercept, respectively.

The influence of pH on the potentiometric response was evaluated at two concentration levels of NO₃⁻ (Figure 3b). A stable response (i.e., EMF variation lower than ±5 mV) was obtained over the wide pH range from 3 to 11 at 10⁻³ M NaNO₃ solution, which is equivalent to the guideline value for groundwater and drinking water in Europe (50 mg L⁻¹ [4]). Nevertheless, at the lower concentration (10⁻⁵ M NaNO₃), a slight decrease in the EMF was observed when the pH values increased over six units, which may worsen the LOD of the proposed sensors. Therefore, the working pH range was defined as optimal between 3 and 6 units to reach the detection of low levels of NO₃⁻.

All potentiometric characteristics for the investigated all-solid-state-ISEs are summarized in Table 1. Since the slope and lower detection limit are mainly defined by ISM composition, comparable analytical performance was observed. Notably, an improved reproducibility for the intercept value was found for the CSPE/GO/ISM sensors, proving the great impact of the GO layer as an ion-to-electron transducer.

The potentiometric selectivity coefficients ($K_{{\mathrm{N}\mathrm{O}}_3^{-},{\mathrm{X}}^{-}}^{\mathrm{P}\mathrm{o}\mathrm{t}}$) of the all-solid-state NO₃⁻–ISEs were determined by the fixed interference method (FIM), following the equation (Equation (1)):

$$K_{{NO}_3^{-},X^{-}}^{Pot}=\frac{a_{{NO}_3^{-}}}{a_{X^{-}}^{\raisebox{1ex}{$z_{{NO}_3^{-}}$}\!\left/ \!\raisebox{-1ex}{$z_{X^{-}}$}\right.}},$$

(1)

where $a_{{NO}_3^{-}}$ and $a_{X^{-}}$ are the activities of the primary (NO₃⁻) and interfering ion ($X^{-}$), respectively, and $z$ are the respective charge numbers [44]. The ion concentration of $X^{-}$ remained constant (10⁻⁴ and 10⁻² M), while the NO₃⁻ concentration varied from 10⁻⁷ to 10⁻² M (Figure S3). The $K_{{NO}_3^{-},X^{-}}^{Pot}$ values summarized in

Table 2. Potentiometric selectivity coefficients ($K_{{NO}_3^{-},X^{-}}^{Pot}$) of the proposed all-solid-state NO₃⁻ electrodes.

$\mathbf{Interfering}\mathbf{Ion}({\mathit{X}}^{-}$)	${\mathit{L}\mathit{o}\mathit{g}\mathit{K}}_{{\mathit{N}\mathit{O}}_3^{-},{\mathit{X}}^{-}}^{\mathit{P}\mathit{o}\mathit{t}}$
	CSPE/GO/ISM		CSPE/ISM
	10−4 M	10−2 M	10−4 M	10−2 M
CO32−	−3.2	−4.0	ND.	−4.5
SO42−	−3.2	−4.2	ND.	−4.5
HPO42−	−3.2	−4.3	ND.	−4.5
Cl−	−1.3	−2.7	ND.	−2.7
NO2−	−1.2	−1.8	ND.	−1.9
Br−	−0.8	−1.1	ND.	−1.2
NO3−	-	-	ND.	-
I−	1.1	0.5 1	ND.	0.5 1
ClO4−	1.2 1	0.5 1	ND.	0.5 1

¹ No Nernstian response was observed for these interfering ions at the examined concentrations and, thus, the selectivity coefficients presented should be considered minimum values. ND.—not determined.

demonstrate a selectivity behavior governed by the ion lipophilicity, which is typical for potentiometric sensors based exclusively on an ion exchanger immobilized in a plasticized PVC membrane [50]. Therefore, the selectivity followed the Hofmeister series [51], exhibiting minimal response to CO₃²⁻, SO₄²⁻, and HPO₄²⁻ and enhanced response to I⁻ and ClO₄⁻. Even when the concentration of the interfering ions decreased to 10⁻⁴ M, these last two anions showed substantial interference (Figure S4 and Table 2). Nevertheless, the selectivity coefficients were similar to the ones reported for a similar ISM composition, either in electrodes with an inner liquid configuration [52,53] or with different solid-contact transducers [37,54]. This indicates that the selectivity of the studied sensors is a function of the ISM composition rather than the nature of the inner transducing element.

3.3. Electrochemical Studies

EIS experiments evaluated the charge transfer resistance and the transduction mechanism of the proposed all-solid-state-ISEs in 10⁻¹ M NaNO₃ solution. The impedance spectrum (i.e., Nyquist plot) of the CSPE/GO bare electrode was first examined, showing a capacitive line down to low frequencies (0.1 Hz) and a deviation from the capacitive element at high frequencies, still indicating a fast transduction across the GO film/electrolyte solution (Figure S5). When compared with the impedance spectra of CSPE/ISM and CSPE/GO/ISM sensors (Figure 4a), the signal at high frequencies becomes mainly dominated by the bulk ISM resistance coupled with the contact resistance between the electrode surface and ISM. The diameter of the semicircle was similar in both sensor configurations: 3.1 ± 0.3 MΩ and 2.7 ± 0.9 MΩ for the CSPE/ISM and CSPE/GO/ISM configuration, respectively (average of three sensor units). On the other hand, the absence of a low-frequency semicircle shows that the ion-to-electron transduction occurs properly in both configurations. Notably, a clear difference is observed at low frequencies, whereas the impedance spectrum of the CSPE/GO/ISM comes closer to the real impedance axis than the CSPE/ISM configuration, which results from the increase in the low-frequency capacitance given by the GO. These findings are corroborated by analyzing the Bode plots (Figure S6). The impedance was almost independent of frequency in the range from 0.1 to 10³ Hz, while the phase angle was close to zero. This demonstrates that recorded impedance was practically equal to resistance, particularly the ohmic resistance of the PVC-based ISM [55]. For frequencies higher than 10³ Hz, the impedance starts to decrease, and the negative value of the phase angle significantly increases. This indicates the growing role of a capacitive element—in this case, the geometric capacitance ($C_g$) of the PVC-based ISM, which is connected to the ohmic resistance [45]. The $C_g$, estimated from the impedance ($Z$) recorded for a phase angle close to −80°, can be calculated by Equation (2):

$$C_g=\frac{1}{2\pi fZ},$$

(2)

where $f$ is frequency. The estimated $C_g$ is around 10⁻¹¹ F, consistent with other data reported for PVC-based membranes [45].

On the other hand, the capacitance of the solid contact, which characterizes the ion-to-electron transduction ability, is related to the low-frequency semicircle of the impedance spectrum and was evaluated along with the short-term potential stability using reversed-current chronopotentiometry [45]. The potential jump ($E$) in the chronopotentiograms (Figure 4b) was used to calculate the total resistance ($R_{Tot}$) of the ISEs from Equation (3):

$$R_{Tot}=\frac{E}{I},$$

(3)

where $I$ is the applied current (1 nA). The calculated $R_{Tot}$ was 7.1 ± 1.0 and 6.5 ± 1.9 MΩ for the CSPE/ISM and CSPE/GO/ISM, respectively. From the slope (∆$E$/∆t), the potential drift was 41.6 ± 18.5 and 30.7 ± 2.3 µV s⁻¹, respectively. The low-frequency capacitance ($C_{LF}$) was calculated from Equation (4) [45]:

$$C_{LF}=\frac{I}{\raisebox{1ex}{$∆E$}\!\left/ \!\raisebox{-1ex}{$∆t$}\right.},$$

(4)

as 32.6 ± 2.4 µF for the CSPE/GO/ISM, which is slightly higher than the obtained value for the CSPE/ISM sensors (27.2 ± 11.0 µF). The capacity values reported herein are overall better than those obtained with reduced graphene oxide [36,56] and comparable (or at least in the same order of magnitude) with those obtained for carbon nanotubes- [34,57] or graphene-based [41,58] sensors. These findings prove the successful use of GO as a reliable solid-contact transducer material for all-solid-state electrode design.

3.4. Water Layer Test and Potential Stability

The formation of a water layer between the ISM and the conductive surface leads to long equilibration times, sensor reading drift, and sensitivity to CO₂ partial pressure [18]. Over time, this water layer can continue to spread over the interface, affecting the ISM’s adhesion and ultimately leading to delamination. The introduction of GO could avoid the formation of this inner aqueous layer, which was assessed using the water layer test proposed by Fibbioli et al. [19]. Briefly, the sensors were initially conditioned overnight in the primary ion solution (10⁻¹ M NaNO₃) and then exposed to an interfering ion solution (10⁻¹ M NaCl) for 2 h. The fast EMF shift when changing solutions is caused by the change in phase boundary potential at the ISM/solution interface and is a function of the selectivity coefficient and the concentrations of the primary and interfering ions (Figure 5). After 2 h, the initial primary ion solution replaced the interfering ion solution, and the EMF was restored to the initial value. The continuous recording for nearly 16 h revealed excellent long-term stability for the CSPE/GO/ISM sensor, whose potential drift was 0.32 mV h⁻¹. Under the same conditions, the potential drift observed in the CSPE/ISM (1.87 mV h⁻¹) could be related to the formation of the water layer, as similarly reported earlier for this type of electrode [18].

The potential stability and sensor longevity of GO-based all-solid-state NO₃⁻ electrodes were studied in detail by performing repetitive calibration curves for 12 weeks and assessing the changes in the analytical parameters. The electrodes were stored dry and conditioned overnight in 10⁻³ M NaNO₃ solution before use. Both sensor configurations exhibit stable ion sensitivity within the linear range from 10⁻⁵ to 10⁻² M, with variations in the slope of −55.9 ± 3.0 and −54.5 ± 2.8 mV decade⁻¹ for the CSPE/GO/ISM and CSPE/ISM sensors, respectively. However, the drift of the recorded EMF values from calibration to calibration was much higher for the CSPE/ISM sensors, with coefficients of variation (CV) of 13.7 and 14.7% for the lower and higher limits of the linear range, respectively. Notably, the corresponding values obtained with the CSPE/GO/ISM sensors ranged from 6.8 and 7.9%. These findings corroborated the formation of the aforementioned water layer in the CSPE/ISM sensors and revealed higher potential stability with the CSPE/GO/ISM sensors. This is ultimately demonstrated by the lower variation of the E⁰ (213.4 ± 20.1 mV, CV of 10%) compared to the observed in the CSPE/ISM sensors (167.0 ± 43.0 mV, CV of 20%). Additionally, the lower limit of detection remained stable for the CSPE/GO/ISM sensors (2.3 ± 0.8 µM) over the 12 weeks, while it became worse for the CSPE/ISM sensors (4.9 ± 3.2 µM). The significantly higher stability in the sensors prepared with GO highlights its use as a transducer in potentiometric all-solid-state electrodes.

3.5. Analytical Application of GO-Based All-Solid-State NO₃⁻ Electrodes

To assess the applicability of the proposed GO-based all-solid-state NO₃⁻ electrodes, NO₃⁻ calibrations in phosphate buffer (10⁻¹ M) were first performed to adjust the ionic strength to 10⁻¹ M and mimic high electrolyte backgrounds (Figure S7 and Table S5). No significant difference was observed compared with the dynamic response and the respective calibration curve obtained with those found in Milli-Q water (Figure 2 and Table 1). Additionally, the NO₃⁻ calibration curves in the real sample matrix (Figure S8), namely domestic (DW) and agriculture (AW) well water, similar to those obtained in phosphate buffer, proved the nonexistence of matrix interference. Recovery studies were accomplished by the analysis of spiked real samples with 2.0 × 10⁻³ (DW.1 and AW.1) and 6.9 × 10⁻³ M (DW.2 and AW.2) NO₃⁻ concentrations, which are equivalent to 123.8 and 430.9 mg L⁻¹, respectively. Each sample was analyzed in triplicate using the proposed CSPE/GO/ISM sensors and a commercial electrode of the inner filling solution. The recoveries were calculated according to Equation (5):

$$Recovery\left(\%\right)=\frac{{NO}_{3Found}^{-}-{NO}_{3Initial}^{-}}{{NO}_{3Added}^{-}}\times 100\%,$$

(5)

where ${NO}_{3Found}^{-}$ is the concentration of ${NO}_3^{-}$ measured in the spiked sample, ${NO}_{3Initial}^{-}$ is the concentration of ${NO}_3^{-}$ found in the water sample, and ${NO}_{3Added}^{-}$ is the amount of ${NO}_3^{-}$ added to the sample. The recovery percentages were acceptable for both ISEs, showing the appropriate accuracy of the CSPE/GO/ISM sensor (

Table 3. Recovery values of the proposed CSPE/GO/ISM sensors and a commercial electrode in spiked water samples.

Sample	Added (mg L−1)	CSPE/GO/ISM			Commercial ISE
		Found 1 (mg L−1)	CV 2 (%)	Recovery (%)	Found 1 (mg L−1)	CV 2 (%)	Recovery (%)
DW	0	34.9 ± 1.0	2.8	-	37.8 ± 2.3	6.0	-
DW.1	123.8	160.8 ± 11.4	7.1	101.7	160.5 ± 5.8	3.6	99.1
DW.2	431.1	500.1 ± 30.0	6.0	107.9	488.7 ± 5.4	1.1	104.6
AW	0	78.5 ± 3.8	4.9	-	81.0 ± 1.9	2.3	-
AW.1	123.8	187.2 ± 11.7	6.3	87.8	200.1 ± 8.9	4.5	96.2
AW.2	431.1	543.5 ± 26.6	4.9	107.9	551.7 ± 30.6	5.5	109.2

¹ Average ± standard deviation (n = 3); ² Coefficient of variation.

).

Finally, the feasibility of the proposed all-solid-state electrodes for determining NO₃⁻ concentration in different water samples was investigated. Well, tap, and commercial water samples were analyzed by direct immersion of the ISEs into the sample, including the commercial one, for comparison. Before the analysis, each sensor was calibrated from 10⁻⁵ to 10⁻² M NO₃⁻ in a phosphate buffer background to ensure proper functioning. In all cases, the difference in NO₃⁻ concentrations provided by the two ISEs was lower than 8.5% (

Table 4. Quantification of nitrate levels in different water samples using the proposed CSPE/GO/ISM sensor and the commercial ISE.

Sample	NO3− (mg L−1) 1		% Difference	p-Value 2
	CSPE/GO/ISM	Commercial ISE
Commercial	4.7 ± 0.3	4.7 ± 0.2	0.4	0.370
Tap	4.3 ± 0.5	4.5 ± 0.1	6.0	0.443
Domestic well	34.9 ± 1.0	37.8 ± 2.3	8.5	0.129
Agriculture well 1	50.5 ± 6.3	51.4 ± 0.2	1.9	0.152
Agriculture well 2	78.5 ± 3.8	81.0 ± 1.0	3.3	0.932

¹ Average ± standard deviation (n = 3); ² p-value calculated from the bilateral t-test with a 95% confidence level.

). This agreement was evidenced by the obtained p-values (>0.05) from a t-test analysis at the 95% confidence level.

Regarding water safety, the NO₃⁻ levels found in the commercial, tap, and domestic well samples were below the guideline value defined by the EU (50 mg L⁻¹ [4]) for drinking water, making them acceptable for consumption without risk to human health. However, higher levels were found in waters from the agriculture well, which could be explained by the runoff of inorganic fertilizers from the soil to the groundwater [59]. Nevertheless, these agriculture wells are included in a geographical area already considered a Nitrate Vulnerable Zone by the Portuguese Government [60], defined by the EU as an area of land that drains into polluted waters or waters at risk of pollution from agricultural activities [4]. Overall, the highly accurate and precise results demonstrate the reliability of the CSPE/GO/ISM sensors for NO₃⁻ measurements without any pre-treatment.

3.6. Comparison of GO-Based All-Solid-State NO₃⁻ Electrode with Previously Reported Electrodes

The electrochemical and analytical performance of the herein proposed GO-based electrode were compared with relevant NO₃⁻ all-solid-state potentiometric sensors reported in the literature (Table S4). The linear response range, the lower limit of detection, and the sensitivity are comparable with earlier reported sensors. Notably, the fast response time recorded as 10 s, the long-term potential stability (0.3 mV h⁻¹), and the impressive lifetime of 3 months outperform the most recently reported NO₃⁻ electrodes. Regarding the electrochemical performance, the proposed sensor exhibits a capacitance value better than reduced graphene oxide [36,56] and comparable (or at least in the same order of magnitude) with those obtained for carbon nanotubes- [34,57] or graphene-based [41,61] sensors. Overall, these findings prove the successful use of GO as a reliable and alternative carbon solid-contact transducer for all-solid-state electrode design.

4. Conclusions

This work used graphene oxide (GO) as an ion-to-electron transducer to develop all-solid-state nitrate electrodes. The analytical performance of the proposed electrodes (LOD, sensitivity, selectivity coefficients, and long-term drift) is comparable with those of other solid-contact arrangements. The proposed sensors exhibited improved potential stability and the absence of a water layer compared to the carbon bare electrodes (using the same sensing membrane). The applicability was demonstrated by accurately determining nitrate levels in water samples. These results expand the scope of graphene-based sensors, particularly by implementing GO as an alternative transducer for constructing stable and durable all-solid-state ISEs, which hold great promise for routine sensing applications.