Unless otherwise specified, all chemicals were sourced from Fisher Scientific (Loughborough, UK). Polystyrene pellets (441,147) and graphite flakes (28,286–3) were supplied by Sigma-Aldrich (Poole, UK). Acrylonitrile butadiene styrene (ABS) granules, MG94 resin were supplied by Saudi Basic Industries Corporation (SABIC; sourced from OS3DP.com). High Impact Polystyrene (HIPS) filament (HIPS175W1) was obtained from CPC Farnell (Preston, UK) and Pyrograf III carbon nanofibers from Pyrograf Products, Inc. (pyrografproducts.com, PR-24-XT HHT). Deionised water was produced using a Purite RO200-Stillplus HP System (Purite Water Purification Solutions, Oxon, UK). Stock solutions of 0.2 M orthophosphoric acid and 0.2 M trisodium phosphate were prepared by dissolving the appropriate mass in deionised water. The 0.2 M trisodium phosphate solution was titrated with the 0.2 M orthophosphoric acid solution to produce phosphate buffer solutions at pH 2, 4, 6, and 8. Stock solutions of 2,4-DNP were prepared by dissolving the required amount in acetonitrile to achieve a concentration of 10 mM. Working standards for voltammetric and chronoamperometric studies were prepared by diluting the primary stock solution with phosphate buffer and acetonitrile to achieve a final solution of 0.1 M phosphate buffer, 10% acetonitrile.

Cyclic voltammetry and chronoamperometry were undertaken using an EmStat3 potentiostat (Ivium Technologies, Eindhoven, the Netherlands) interfaced to a PC for instrument control and data acquisition. The voltammetric cell contained a carbon rod auxiliary electrode, a Ag/AgCl/KCl(3M) reference electrode, and a 3D printed carbon nanofiber–graphite–polystyrene electrode as the working electrode. The electrodes were connected to the potentiostat via suitable crocodile clips attached to coaxial cable leads inserted into the appropriate sockets on the instrument. The 3D printed carbon electrode was rinsed with deionised water, polished with tissue paper, rinsed a second time in deionised water, and dried under a stream of nitrogen gas before each scan.

Cyclic voltammograms were recorded with solutions of 0.1 M phosphate buffer containing 10% acetonitrile, and then in the same solution containing 1.0 mM 2,4-DNP. Samples were purged with nitrogen gas (BOC, Guildford, UK) for 5 min to remove dissolved oxygen. Cyclic voltammetric investigations were undertaken using either (i) a starting and end potential of 0.0 V, an initial switching potential of −1.5 V, and a second switching potential of +1.0 V, or (ii) a starting and end potential of 0.0 V, an initial switching potential of +1.0 V, and a second switching potential of −1.5 V. The effect of the scan rate was investigated over 20 mV/s to 200 mV/s.

Double potential step chronoamperometric studies (Mortimer, 2017) were undertaken in solutions of 0.1 M pH 2.0 phosphate buffer, containing 10% acetonitrile using the potential waveform shown in Scheme 1. Step 1: applied potential at −1.4 V for 60 s. Step 2: applied potential of +0.8 V for 60 s. All measurements were recorded in triplicate.

The cyclic voltammetric behaviour of 2,4-DNP was studied over the range pH 2 to pH 8 (representative voltammograms obtained at pH 2 shown in Figure 2). Using an initial negative-going scan, two reduction peaks were recorded and designated R1 and R2. These are thought to result from the reduction of the two nitro groups to their corresponding hydroxylamines (Equation 1). On the subsequent positive-going scan, two oxidation peaks are observable (designated O1 and O2) corresponding to the oxidation of the electrochemically-generated hydroxylamine to the corresponding nitrosamine (Equation 2). R − NO 2 + 4 e − + 4 H +   →   R − N H − OH + H 2 O (1) R − N H − OH  →   R − N = O + 2 e − + 2 H + (2)

Further studies were undertaken to investigate this proposed mechanism. If the voltammetric scan was first implemented in a positive direction, no oxidation peaks were observable. Thus, the oxidative responses did not occur unless the molecule had undergone some previous reduction process. Hence, it was concluded that the two oxidation peaks result from the oxidation of the hydroxylamine (Equation 2) formed via the process described in Equation 1.

The peak potential (Ep) for peaks R1, R2, O1, and O2 was obtained over the range pH 2–8 (Figure 3). All of the peaks observed were found to be pH dependent. Plots of the Ep values of peaks R1, R2, and O2 against pH were found to give slope close the Nernst theoretical values of 59 mV/pH indicating an equal number of protons and electrons were involved in their reductions or oxidations. However, peak O1 was found to have a break point at pH 4.1. It is possible that this is a result of acid dissociation (pKa) associated with the phenol group. The pKa of 2,4-dinitrophenol is 4.13 (Pearce and Simkins, 1968) which is close to the result obtained. Consequently, we believe that oxidation process seen at O1 is related to the oxidation of the phenolic group. However, this oxidation is not observable without prior reduction of the molecule. This was concluded to be due to the stabilisation of the phenolate anion by the electron withdrawing properties of the hydroxylamine groups formed from at higher pH values. At pH values below the pKa, this stabilising effect is removed allowing for the oxidation of phenol. This effect can be seen in the change in the slope of the plot of Ep verses pH for O1 shown in Figure 3. At pH values above pH 4 a slope of 55 mV/pH unit is results, close to the theoretical Nernst value for an oxidation process involving an equal number of electron and protons. However, a slope of 120 mV/pH unit is recorded below pH 4; indicative of an oxidation process involving a two to one ratio of protons to electrons. We believe that this is the result of the loss of the stabilising effect of the electron withdrawing nitro groups in the prior reduction step. At this lower pH and without stabilisation the nitro groups are reduced to their corresponding amine groups. As amines act as electron donating groups, their presence allows for the oxidation of the phenol to a radical, and then via the loss of a further proton from one of the amine groups to a residence stabilised phenolate. The possibility of direct electrochemical oxidation of 2,4-DNP via one-electron oxidation of the phenolic group has been demonstrated previously (Yin et al., 2012). However, we have not seen evidence for this in the cyclic voltammetric investigation conducted here. Both of the reduction processes (R1 and R2) showed slopes of 55 mV/pH unit; this is close to the theoretical Nernst value for reduction processed involving an equal number of electron and protons.

Chronoamperometry offers the possibility of simple operation and utility outside of the laboratory with small sample volumes. In comparison to other voltammetric techniques, chronoamperometry can give an improved signal-to-noise ratio as the current can be integrated over longer time intervals. As we were interested in determining low concentrations of 2,4-DNP, we chose to investigate the possibility of utilising a chronoamperometric measure step for the determination of 2,4-DNP. We focused our investigations on the anodic oxidation of the electrochemically generated hydroxylamine species (Equation 2). This offers analytical advantages, as it avoids interference from the reduction of sample components, such as oxygen. The double step amperometric measurement (Mortimer, 2017) step is potentially open to interferences from the redox behaviour of metal ions, such as Pb²⁺ and Cu²⁺. However, as previously shown (Honeychurch, 2019), metal ions exhibit poor electrochemical activity in phosphate-based electrolytes like those employed here. Primarily Na⁺ and K⁺ salts are the only ones readily soluble in water and are not electrochemically active under the conditions employed here. The anodic region, particularly at low positive potentials, offers several analytical advantages for electrochemical detection. Background currents in this region are typically lower than those observed in the cathodic range, resulting in an improved signal-to-noise ratio and enhanced sensitivity. Additionally, fewer electroactive species are present at these potentials which reduces the likelihood of signal overlap and interference thereby improving the clarity and reliability of the analytical response. Oxidation processes, especially those involving pre-generated intermediates such as hydroxylamines, also tend to produce more stable and reproducible signals compared to direct reductions in complex environmental matrices.

We first investigated the chronoamperometric behaviour of a quiescent 10% acetonitrile 0.1 M pH two phosphate buffer in the presence and absence of 1 mM 2,4-DNP. A potential of −1.4 V was applied for 30 s to reduce 2,4-DNP to its corresponding hydroxylamine (Equation 1). The potential was step to +1.0 V and held for 60 s to oxidise the hydroxylamine formed to the nitrosamine, giving the analytical signal (Figure 4a). Plots of current (i) vs the inverse of the square root of time (t^−½) showed the response to follow Cottrell type behaviour (Figure 4b) (Pletcher et al., 2001).

The chronoamperometric current response was found to be pH dependent over the range studied (Figure 5) and was found to decrease in a near linear decrease (R² = 0.994, −1.395 μC/pH) with increasing pH. This was likely due to the stabilisation of the nitro groups in 2,4-DNP through charge delocalisation facilitated by the phenolate anion. The maximum peak current of the oxidation peak O1 was obtained at pH 2.0. Therefore, subsequent studies were undertaken using a supporting electrolyte of 0.1 M pH 2.0 phosphate buffer.

We first optimised the applied reduction potential used in the first step of the chronoamperometric measurement. The effect of this reduction potential was investigated over the range −0.4 V to −1.5 V by monitoring the resulting oxidation current recorded at +1.0 V (Figure 6). The oxidative chronoamperometric response was found to increase with increasing negative potential over the range −0.4 V to −0.9 V. At more negative potentials the oxidative response was found to reach a constant. However, as the potential beyond −1.0 V was made more negative the precision of the measurement improved with the lowest coefficient of variation (3.2%) being obtained at −1.4 V. Consequently, an applied potential of −1.4 V was used for step 1 (reduction) of the chronoamperometric measurement in further studies.

Subsequently, the oxidative measurement step (step 2) was examined. The effect of applied potential was studied over the range +0.2 V to +1.0 V. In this investigation, the resulting chronoamperometric signal was found to increase with increasing positive potential over the range +0.2 V to +0.8 V. The signal was found to become constant at more positive potentials and so further studies were undertaken using an applied step 2 potential of +0.8 V.

The effect of step 1 reduction time in an unstirred, quiescent solution on the resulting oxidative chronoamperometric response was investigated over the range 10 s–120 s. The oxidation signal increased with time and exhibited a logarithmic relationship (Figure 7). In this proof-of-concept investigation based on a balance of time against sensitivity, a step 1 reduction time of 60 s was selected and used in further investigations.

Initial studies were undertaken to study the effect of 2,4-DNP concentrations on the magnitude of the chronoamperometric response. The calibration plot was found to be linear from 50 μM to 1.0 mM (0.4399 μA/mM; R² = 0.9980), with a detection limit of 7.8 µM (based on a signal-to-noise ratio of three). A coefficient of variation of 4.7% was obtained for a 0.05 mM 2,4-DNP solution. The LOD was calculated based on measurement of the blank signal; the signal measured when no analyte is present. This was determined using the following equation (Equation 3): L O D = 3   σ b l a n k   S (3)

The optimised procedure was used to determine the concentration of 2,4-DNP in fortified and unfortified environmental pond water samples (typical amperograms shown in Figure 8). A mean recovery of 106% was obtained for a 0.05 mM 2,4-DNP fortified environmental pond water sample with an associated coefficient of variation of 3.6%. These results show that the optimised chronoamperometric method is capable of determining 2,4-DNP in environmental water samples.

The findings demonstrate that these 3D-printed working electrodes serve as effective alternatives to those made from conventional materials, providing reliable data for the determination of 2,4-DNP. The performance characteristics are similar to those previously reported at traditional electrode materials such as Hg modified Ag electrodes with detection limits ranging between 2.7 µM (Fischer et al., 2007) to 10 µM (Danhel et al., 2009) being reported. Notably, the detection limit is an improvement on that at Pt and Au electrodes of 85 µM (Khachatryan et al., 2005). A number of reports have shown that lower limits of detection can be obtained at carbon-based electrodes fabricated from more traditional materials such as glassy carbon. However, these have utilised secondary modifications to the electrode surface; including surfactants (Wang et al., 2006), graphene and molecularly printed polymers (Liu et al., 2012), metal nanoparticles (Yang et al., 2011), and sample preparation steps such as solid phase extraction (Lezi et al., 2014) and liquid chromatography (Karaová et al., 2016; Ruana et al., 1993). A dual-function 3D printed electrochemical device based on recycled PLA combined with carbon black, and graphite that acts both as a collector and detector has recently been developed (Souza et al., 2024). Using square-wave voltammetry for the determination of TNT, limit of detections of between 0.88 and 3.42 µM were reported. They demonstrated the possibility of determining TNT residues on a metal surface following an explosion by either rubbing the electrode over the surface or by tapping the electrode on the contaminated surface. Similarly, a 3D printed electrochemical sensor made from a graphite/polylactic acid (Gpt-PLA) composite was evaluated using square-wave voltammetry for the detection of TNT in environmental water samples (Siqueira et al., 2023). Three reduction peaks with two being selected for quantification giving limits of detection of 0.52 μM and 0.66 μM. Using reduction processes as the analytical signal can be problematic as commonly occurring sample components, such as oxygen and metal ions, can interfere. Our reported approach could offer some advantages over this.