Introduction

Graphene-based nanomaterials will be indispensable parts of next-generation sensors, microelectronics, and energy storage and conversion devices. Their rising importance is due to the outstanding properties of graphene, a low dimensional material with a high specific surface area, high conductivity, low density, and ultra-high mechanical stability1,2. However, a considerable portion of graphene’s popularity originates from the nature of its surface chemistry and suitable surface modifications3,4. The broad range of applications of graphene includes those in which graphene directly interacts with an aqueous solution (e.g., seawater desalination5,6, sensors7,8, and supercapacitors9,10). Nevertheless, our understanding of graphene-aqueous electrolyte interactions is severely lacking and has not been sufficiently explored. Even though the arrangement of ions and molecules of the electrolyte in the vicinity of the graphene is determined by the graphene’s inherent hydrophobicity, the nature of the interfacial interactions on actual samples is affected by the presence of local surface charges and other defects11. Water contact angle (WCA) measurements can be used to quantify the effect of wettability on graphene/electrolyte interactions. An ideal graphene lattice is formed by networks of sp2-hybridized carbon atoms, generating a highly symmetric π-electron cloud. Hence, introducing common external dopants (e.g., molecular oxygen) or other functionalities (e.g., –OH, –COOH) disrupts the orbital symmetry and generates local polarities (defects). Therefore, the dipole orientations of water molecules and ions vary as the graphene chemistry changes12. However, in confined spaces (such as nanochannels13), an opposite correlation between wettability and capacitance has been reported. In such environments with low WCA, solvent molecules instead of ions have been found to be adsorbed onto the surface at the interface, leading to a reduced capacitance while higher wettability results in the strong electrosorption of solvent molecules that can impede the dielectric screening ability, ultimately leading to a decrease in capacitance14.

Multiple solution parameters, including pH, temperature, dissolved oxygen (DO), the solutions ability to oxidize chemical species (i.e., its ability to gain electrons from these species, known as its reduction potential, commonly referred to as oxidation-reduction potential, ORP), and more importantly, ionic strength, influence the graphene properties simultaneously15,16 through the interfacial electrochemical double layer (EDL). Therefore, the study of graphene-aqueous electrolyte interfaces requires well-controlled conditions where the co-influence of parameters is minimized. The EDL describes the arrangement of molecules and ions at the surface and indicates the distance from which the solid surface potential fades into the solution potential. The extent of the EDL and its impact on graphene properties will be affected by the ionic strength of the electrolyte17. Therefore, the ionic strength of the electrolyte and its influence on the thickness of the EDL are instrumental to understanding the impact of the EDL on graphene, particularly at high concentrations where the EDL characteristics may be altered due to interfacial species, topographic features, and electric field distributions14.

Electrochemical measurements involving graphene devices require a sufficient ionic strength of the electrolyte to provide cell conductance18, posing challenges for exploring the structure of the EDL and its impact on the graphene at low ionic strength. Designs based on field effect transistors or chemiresistive geometries probe the interface with a current passing parallel to the interface and are thus suitable for studying a broader range of electrolytes19. Moreover, the graphene electrode polarization is shown to be both frequency and voltage-dependent20, meaning a high applied potential or oscillating frequency may result in a change of the carrier dispersion within graphene through modulation of the Fermi energy, which affects the electrochemical potential of the electrolyte relative to the Fermi level of the graphene21. Accordingly, the kinetics and thermodynamics of EDL formation will be affected. Hence, understanding the structure of the graphene interface with an aqueous electrolyte (i.e., EDL) requires an alternative (or parallel) measurement method that operates at a relatively low applied potential and continuously monitors the graphene response to variations in the structure of the EDL.

One approach to examining the EDL of graphene is the measurement of the decay length of the electrostatic force of the ions. The Debye–Hückel (D-H) theory is commonly used to predict the relationship between the electrostatic screening length (\({\lambda }_{\text{D}}\)) and the ionic strength of the solution according to Eq. (1)22:

$${{\rm{\lambda }}}_{D}=\sqrt{\frac{{\varepsilon }_{{\rm{o}}}{\varepsilon }_{{\rm{r}}}{KT}}{2I{{\rm{e}}}^{2}}}$$
(1)

where \({{\rm{\varepsilon }}}_{{\rm{o}}}\) is the vacuum permittivity, \({{\rm{\varepsilon }}}_{{\rm{r}}}\) is the relative permittivity of the solution, \(k\) is the Boltzmann constant, \({{T}}\) is the temperature, and \(I\) represents the ionic strength22. The ionic strength of a solution is measured by \(I=\mathop{\sum }\nolimits_{1}^{n}{c}_{i}{z}_{i}^{2}\) (Ci and Zi are the concentration and charge of ions, respectively), and is used to directly determine the activity coefficient of an ion in a solution. The Debye length is also considered the characteristic decay length of the EDL forces proposed by the Poisson–Boltzmann (PB) theory. The PB theory assumes that the ions are non-polarizable point charges immersed in a continuous solution at a large distance from each other, meaning the ion-ion interactions can be neglected. Accordingly, the \({\lambda }_{\text{D}}\) calculation will only be valid in dilute solutions23. Although Debye theory is limited in the consideration of radii of ions, smaller ions have a greater effect on the ionic strength of a solution due to their higher charge density and greater interaction with the solvent molecules. Thus far, there are multiple theories to predict the change in screening length for electrolytes at high ionic strength; however, they either ignore the role of the substrates (conductive or insulator)22, are not dedicated to aqueous electrolytes24, or are not consistent with deviation from ideality25. Finding the concentration at which the solution starts deviating from D-H theory requires the direct measurement of experimental decay length. Most experimental measurements of the screening length are done using a surface force apparatus (SFA) or atomic force microscope (AFM), requiring complex equipment that may not be easily accessible. The former issue originates from the complexity of confining an electrolyte between two statically charged surfaces and changing the ionic concentrations while measuring the forces within nanometer distances26. The AFM tip is sensitive (both materials and mode) to graphene topography and the results are very substrate dependent. Investigating the EDL on free-standing graphene would substantially complicate the experimental design even further, but a discussion of the role of the substrate in the deviation from ideality is missing in the literature20. The majority of the techniques exploring the screening length of an electrolyte (i.e., SFA or SFG) are based on the substrates with static charges such as SiO222. Therefore, conductive graphene derivatives have been less explored. Additionally, the impacts of graphene thickness, defect density and edge termination have been ignored in determining EDL composition in graphene devices24,25.

In this research, we report the use of a simple graphene-based chemiresistive platform as an indirect tool for monitoring changes in the experimental screening lengths of aqueous solutions as a function of increasing ionic strength. With the help of electrochemical measurements, we propose a mechanism for the deviation of the screening length from the D-H theory at certain concentrations of salts, and elaborate on the roles of graphene defect density, cation size, and DO in the formation of the EDL. This research aims to develop a more complete picture of the mechanism by which the Debye–Hückel theory is violated in the proximity of graphene surfaces at much lower concentrations (10–50 mM) than previously reported values for other systems (1 M).

Results and discussion

Graphene and its interactions with ions

The structure of the EDL at the electrolyte/graphene interface depends on the defect density, surface charge, geometry, concentration, charge, and size of the ions in the electrolyte. The total capacitance resulting from the formation of the EDL on graphene originates from the adsorption capacitance upon ion adsorption, Helmholtz Layer capacitance, and bulk capacitance. This means that the ions adsorbed onto the surface should be considered a separate capacitator parallel to where the Stern Layer containing both the inner and outer Helmholtz layer is located19,20,22. Therefore, prior to any experiments, understanding whether the ion adsorption could dope graphene chemically is vital. The Raman spectrum of the ultrasonically exfoliated few layer graphene (FLG) materials used in this research (Fig. 1a, black spectrum) exhibits the expected D, G and 2D bands at 1385, 1587, and 2684 cm−1, respectively. The D band results from structural asymmetries caused by sp3 hybridization in the graphene crystal, while the G band is a feature of the long-range ordered sp2 hybridized network of carbon atoms27. The ratio of D to G intensities (or areas) is a measure of the defectivity of graphene28. Here, the \(\frac{{I}_{{\rm{D}}}}{{I}_{{\rm{G}}}}\) ratio of the FLG is calculated to be 0.38, indicating the existence of various defects, such as oxygen atoms in the form of different functional groups. These functional groups mainly originate from the sonication process through the direct interaction of graphene and solvents (IPA and H2O)29. Therefore, electron-withdrawing oxygen atoms generate more holes as majority carriers and p-dope the graphene. Since oxygen is more electronegative than carbon, oxygen dopants cause a negative zeta potential on the graphene surface exposed to aqueous electrolytes30. Even though atomically thin graphene offers advantages over FLG in terms of uniformity, thin graphene layers can be transparent to the substrate properties, and the arrangement of ions near the graphene will be affected by the substrate’s hydrophobic or hydrophilic properties11. Therefore, the configuration of the EDL and/or the decay length deviation from ideality will depend not only on the variations in graphene but also on the mobile or static surface charges of the substrate, interfering with the graphene/aqueous electrolyte interactions. In addition, the characteristics of a thin graphene produced by CVD are highly influenced by the number of layers and edge configuration, and offers a low saturation concentration that makes it unsuitable for high-range concentration measurements16. CVD-grown single layer graphene will therefore have very different properties than networks of FLG flakes, making it unsuitable as a model system to understand the behavior of electrodes used in sensing or EDL capacitor applications which require porosity and large available electrochemical surface areas. Hence the choice of FLG networks over monolayer graphene in this study.

Fig. 1: Ionic strength-dependent behavior of graphene.
figure 1

Raman spectra of FLG before (black) and after (red) (a) exposure to NaCl solution during chemiresistive measurements and (b) drop casting 0.1 mM NaCl solution. Chemiresistive response of FLG films to the addition of NaCl in (c) 0 ppm DO and (d) 7 ppm DO solutions (red numbers: NaCl concentrations in mM). e Calibration curves of the chemiresistive response of the FLG films to the addition of NaCl to the electrolyte in the absence (black circles) and presence of DO (red circles). The inset shows the response inversion for 0 ppm DO solutions between 0–10 mM NaCl.

Full size image

The impact of the ionic strength on the extent of the EDL (screening length) at the graphene/electrolyte interface can be investigated by measuring the response of the graphene chemiresistive device to the addition of salts to the aqueous electrolyte. However, since graphene/electrolyte interactions occur through two main pathways of electrostatic gating (known as capacitive charging) and charge transfer through redox reactions (known as pseudocapacitive charging), it is necessary to exclude charge transfer (chemical doping) effects during the experiments. Therefore, alkali chlorides (i.e., LiCl, NaCl, and KCl) were used to adjust the ionic strength since they dissociate into non-redox-active hydrated ions31,32 (see Supplementary Fig. 1 for the measured ionic strengths of LiCl, NaCl, and KCl solutions as a function of concentration)33. Therefore, we do not expect to observe chemical doping by the cations, which is confirmed by the Raman spectrum of graphene exposed to 0.1 M NaCl aqueous solution for 2 days, shown in Fig. 1a, red, compared to the blank graphene (Fig. 1a, black). The comparison of the band positions and intensities reveals that (1) the G band is slightly blue-shifted, indicating charge doping of the surface of graphene (n or p doping cannot be distinguished); (2) the 2D band undergoes a blueshift, demonstrating hole-doping of graphene, and (3) a slightly higher \(\frac{{I}_{{\rm{D}}}}{{I}_{{\rm{G}}}}\) ratio for graphene exposed to NaCl31,32. This shows that the graphene has been slightly p-doped by extended exposure to an aqueous electrolyte, but no trace of Na+ doping (ion trapping or charge transfer) is observed. A direct Faradaic charge transfer and n-doping of graphene would have manifested itself in the Raman spectra as a 2D peak shifts toward lower wavenumbers, irrespective of the type of metal chloride dopant31.

In contrast, when 0.1 M NaCl is drop-cast on FLG and dried at 100 °C, it forces the ions to trap on and between graphene sheets. In this case, shown in Fig. 1b red, a small redshift in the G band (from 1578 to 1572 cm−1) is observed. This observation is in agreement with literature reports that exposure to ions from metals with low ionization energy (e.g., Na+, K+, Ca2+) can result in n-doping of the graphene surface at a high concentration when directly applied31. The presence of the n-doping effect is additionally verified through the observation of a slight redshift in the 2D peak following drop casting. These findings can be substantiated by the electrochemically driven adsorption of cations onto the surface of graphene. This adsorption taking place through dehydration or adsorption mechanisms will depend on the zeta potential of the graphene surface. Accordingly, cation doping often occurs when a cation encounters a graphene surface with a high negative zeta potential. These interactions can be regulated by the chemical composition of the graphene34.

Non-ideal behavior of solutions at the graphene surface

Figure 1c represents the chemiresistive response of the device to the addition of NaCl. To minimize the impact of dissolved redox-active gases during the experiments, the solution was purged continuously with dry N2(g), yielding nearly 0 ppm DO in the solution (Supplementary Fig. 2). As seen in Fig. 1c, the addition of up to 10–20 mM NaCl reduces the current across the graphene chemiresistor. However, adding even more NaCl causes an unexpected increase in current, so the graphene response to salts is inverted. For a detailed explanation of chemiresistive measurements, calibration curves, and a comparison to electrochemical measurements, please refer to Supplementary Discussion 1, incl. Supplementary Figs. 1 and 2. The behavior in low concentrations is expected following the conventional EDL formation mechanism in graphene devices. NaCl fully dissociates into hydrated Na+ and Cl− ions. Due to the presence of negatively charged oxygen atoms in the functional groups at the graphene surface, the Stern Layer is dominated by the Na+ ions. Since underpotential deposition of alkali metals is not possible from aqueous electrolytes under the given conditions, the Na+ ions will retain their solvation spheres and cannot directly adsorb onto the surface35 (For the voltage-dependent EDL response of graphene, see Supplementary Fig. 5 in Supplementary Discussion 2). Hence, these two layers of charges (Na+ being positive, graphene being negative, and water as the dielectric) can be considered as a standard capacitor. This means the formation of the Stern layer by Na+ ions electron-dopes the surface, and the chemiresistive current decreases. The level of n-doping increases as more NaCl is added to the solution, mainly due to the considerable reduction in theoretical Debye length (Supplementary Fig. 6). Recent theoretical investigations have unveiled the potential impact of excess charge on substrates and surface chemistry concerning the partial stripping of solvation shells36. While this concept aligns with certain well-established theories such as the Gouy-Chapman-Stern-Grahame model, which regard the inner Helmholtz layer as a region containing unsolvated ions in close proximity to the surface37, it deserves special consideration due to the consequential loss of solvation spheres. This loss, particularly at low applied potential or at OCP, may lead to chemical doping or the formation of surface complexes, which could potentially impact the accuracy of capacitive measurements in graphene36.

The latter part of Fig. 1c (above 10–20 mM) is the result of deviation from ideal behavior, meaning that from a particular concentration, \({\lambda }_{D}\) increases as a function of salt concentration. In 2019, Gaddam and Ducker found this concentration to be around 1 M for NaCl, nearly independent of substrate22. These findings may be explained by the balance between the electrostatic double-layer forces of the ions with the van der Waals38,39,40 and hydration interactions41,42. At low concentrations, ions can be considered separated, i.e., the local electric field at each ion is caused only by that ion. Therefore, the long-range forces in the double layer forces, attracting ions to the charged surface of graphene. At higher concentrations, the deviations from the assumptions underlying D-H theory alter the balance of forces. One theory proposes that the dominant Coulombic repulsive forces between similar ions (Na+-Na+) cause the EDL to expand and \({\lambda }_{\text{D}}\) to increase43,44. Accordingly, the magnitude of EDL capacitive charge screening at the graphene surface decreases, leading to lower resistances (larger current). Some theoretical studies predict more complex surface structures, including overscreening of surface charge at the surface for low potentials and ion accumulation at interfaces where the surface potential is high. This substrate surface potential undergoes a “charge reversal” due to the non-electrostatic contributions of cations interacting with water. Early studies suggested that only divalent cations could cause the formation of a strongly correlated liquid at the surface while monovalent ions do not induce a charge reversal45. Nevertheless, this phenomenon was later observed for monovalent ions at concentrations around 1 M22. In contrast to the literature stating that no charge reversal occurs up to 0.15 M, we have seen this inversion at around 10 mM for Na+ in the absence of DO. Hence, the role of the substrate’s surface chemistry needs to be considered.

In the presence of DO, the structure of the EDL formed by monovalent or divalent cations becomes more complex. The graphene chemiresistive response to the EDL in an electrolyte in equilibrium with air (7 ppm DO) (Fig. 1d) also exhibits current inversion upon addition of NaCl, however at much higher concentrations of between 50–75 mM (see Fig. 1e for the calibration curves). This could be due to two simultaneous phenomena: (1) direct absorption of DO and surface oxidation, and (2) charge neutralization that diminishes the electrostatic forces caused by Na+. The former is due to the redox activity of the DO toward withdrawing electrons from graphene. The latter results from the formation of the Stern layer by positively charged ions. These results are confirmed by electrochemical impedance spectroscopy (EIS) of graphene in DO concentrations of 0 and 7 ppm at open circuit potentials (OCP). Compared to the case where DO is present, the N2 purged solution exhibits a lower charge transfer resistance (Rct), a smaller diffusion resistance, and slightly more ideal capacitive EDL characteristics (Supplementary Fig. 7). These indicate that the presence of DO may disrupt the long-range arrangement of Na+ ions and diminishes their capacitive charges to the graphene. Also, the possible presence of adsorbed DO at the graphene surface should be considered, which would alter the graphene surface charge. However, since the DO adsorption on the surface is strongly ionic strength dependent, it can be concluded that the change in FLG response to DO is due to the presence of other ions in the Stern layer (here Na+ and Cl−).

The origin of solution’s non-ideality on graphene

A reversal in chemiresistive response may also occur when the Fermi level crosses the Dirac (neutrality) point in the band structure of graphene. Graphene is inherently p-doped due to exposure to ambient oxygen and oxygen-containing defects. Charge screening by the Stern layer induces electrons (n-doping), which counteracts the p-doping and moves the graphene’s Fermi energy closer to the neutrality point, until that point is passed, and electrons become the majority carriers. A series of experiments was conducted to rule out this mechanism. The graphene devices were initially exposed to pH variations from 3 to 5 and back down to 3, and a stepwise increase of the ionic strength was conducted shortly after in the same solution. After 1 day of exposure to a highly concentrated salt solution (around 600 mM NaCl), the pH experiment (3 to 5 to 3) was repeated at the higher ionic strength. As seen in Supplementary Fig. 8, the pH response of graphene is somewhat lower but going in the same direction. The pH response of graphene with the Raman \(\frac{{I}_{{\rm{D}}}}{{I}_{{\rm{G}}}}\) ratio of 0.38 is believed to be due to the presence of pH-responsive functional groups of the surface namely carboxyl (–COOH), hydroxyl (–OH), amine (–NH2), and indirectly ketone (–CO–) and aldehydes (–CHO). Considering the p-doped nature of graphene, the current increases at lower pH due to protonation of –COO− to COOH at a pH around its pKa (3.74). As seen, after exposure to salt for a day, the pH response remains the same, meaning the nature of the charge carriers has not changed. It is worth noting that the pseudocapacitive charging mechanisms of functional groups may dominate the surface capacitive interactions during pH changes, provided that the \(\frac{{I}_{{\rm{D}}}}{{I}_{{\rm{G}}}}\) ratio of the graphene exceeds 0.35 ± 0.146. Therefore, it is crucial to carefully select the type of ions to minimize the electrochemical interactions with graphene via losing their hydration spheres, control the pH and ORP of the solution to ensure no abrupt change occurs upon the addition of salts (we purged the salt solutions with air and N2 before adding them to the reaction vessel to prevent any variations in ORP or pH due to dissolved O2). Finally, all measurements were carried out at OCP ± 10 mV to ensure that the applied potential does not result in external currents, and faradic contributions of redox species are mitigated.

A recent electrochemical study of ITO working electrodes reported that the contribution of ion diffusion to the overall cell conductance becomes substantial at 10 mM NaCl25. Below 10 mM the formation of the Stern layer causes charge screening while above 10 mM the decay length of ions extends to somewhere in the diffuse layer. The study concluded that the experimentally measured \({\lambda }_{D}\) match with theory above 10 mM. \({\lambda }_{D}\) initially increases and starts to drop above 10 mM25. To compare our results with literature reports, our samples were also analyzed by EIS when exposed to varying concentrations of NaCl (Fig. 2a). Increasing the NaCl concentration leads to smaller internal resistances (Supplementary Fig. 9) obtained from the sum of the electrode (Fig. 2b, RA) electrolyte (Fig. 2b, RAB), and diffuse layer resistances (Fig. 2b, RBC). Although equivalent circuits of Nyquist plots could reveal valuable information, in our case the circuit components change with salt concentrations and the obtained CEDL values will not be meaningfully comparable. Nevertheless, our results show that the absolute value for the diffuse layer resistance increases with increasing NaCl concentration (Fig. 2d) while the measured slope of the capacitive region in the EIS spectra drops at concentrations around 10–20 mM (Fig. 2c). Therefore, increasing the NaCl concentration initially increases the slope, indicating the formation of a more “ideal” capacitor (slope → ∞ or vertical line represents an ideal capacitor, i.e., a circuit component independent of Zreal). At a NaCl concentration of 10–20 mM, this trend is broken and further increases in the NaCl concentration no longer change the slope by much. The results mirror the chemiresistive response, with the minimum in chemiresistive current matching the minimum in the capacitive slopes. Even though these results contradict the reported trend in the literature, the electrochemical measurements are complicated by low ionic conductivity and the results may not be consistent. This inconsistency is also apparent in cyclic voltammograms of FLG exposed to NaCl (Fig. 2e). The obtained specific EDL capacitance (CEDL) experiences a drop at 10–20 mM NaCl and then starts to rise again. Two conclusions can be drawn from this observation: (1) since the chemiresistive current depends only on the graphene conductivity, the system can function at dilute solutions; (2) assuming the formation of a Stern layer below 10–20 mM, chemiresistive currents must drop within this range (Fig. 1c) due to the adsorption of Na+, while above 10–20 mM counter ions (Cl−) dominate the capacitive charging of graphene, resulting in a current increase depending on the DO concentration. The total capacitance (CT) of the system depends on the contributions of all capacitances. Here, the graphene quantum capacitance (CQ) and CEDL act as two capacitors in series (i.e., \(\frac{1}{{\text{C}}_{\text{T}}}=\frac{1}{{\text{C}}_{\text{EDL}}}+\frac{1}{{\text{C}}_{\text{Q}}}\))11. By considering an average value of 2 µF cm−2 for FLG on SiO2 (at zero gate voltage47), CT (obtained from CV) follows a similar trend as CT (obtained from EIS) at concentrations below 50 mM (see Fig. 2e). However, for concentrations above 50 mM, the bulk solution capacitance (CB) starts to contribute as another capacitor in series (see Supplementary Fig. 10a, b, for the calculation of the capacitance from CV in Supplementary Discussion 3). The very small variation of CT as a function of NaCl concentration is also indicative of a small contribution of non-faradic interactions compared to pseudocapacitive charging mechanisms of FLG, perhaps due to existing defects and edges. This is seen in the very small variation of CV graphs as a function of NaCl concentration, shown in Supplementary Fig. 11a–c. Furthermore, upon increasing the ionic strength of the electrolyte, there is a corresponding increase in the potential difference across the EDL of graphene. As a result, the OCP of graphene will change. It can be generally observed that as the ionic strength of the electrolyte is increased, the OCP at the graphene surface decreases (Supplementary Fig. 12). Additionally, the measurement of the OCP is subject to significant uncertainty, as demonstrated in Supplementary Fig. 12, due to a variety of factors. These include the low conductivity of the solution, slow diffusion of ions, and the buildup of charge at the electrode surface48,49. Another contributing factor is the capacitive charging of other adsorbates including dissolved gasses or instantaneously formed ions such as carbonate. As seen Supplementary Fig. 12, the instability of graphene OCP is considerably reduced when air-purged solutions are used. Therefore, it is essential to ensure that an equilibrium between graphene and the aqueous electrolyte is established at low ionic strength prior to any experiment, especially in the absence of DO.

Fig. 2: Evaluation of ion-induced changes in the impedance of graphene.
figure 2

a Nyquist plots of FLG exposed to varying concentrations of NaCl, b qualitative description of a Nyquist plot associated with the samples, c change in the slope of low frequency EDL region as a function of NaCl concentrations (slope of the graph after dashed line C), d change in the diffuse layer resistance as a function of NaCl concentration (represented by the resistance between B and C in the Nyquist plot), e The CT and CEDL obtained from cyclic voltammograms and EIS spectra of FLG exposed to varying concentration of NaCl, respectively. Error bars represent the standard deviation from measurements of three samples per condition.

Full size image

To understand the graphene response to DO concentration at various ionic strengths, graphene samples were equilibrated in NaCl electrolytes overnight, and the device responses to changes in DO concentration between 0 ppm (N2 purged) and 11 ppm (Air purged) were recorded. Figure 3a, b shows that the graphene device does not respond to changes in DO concentration in 1 mM and 10 mM Na+. At 100 mM DO, the introduction of DO causes a stepwise variation in chemiresistive current (Fig. 3c). Notably, upon purging with N2 and decreasing the DO concentration in a 100 mM NaCl solution, the graphene is significantly influenced by the densification of the Stern layer and the current decreases, demonstrating that the impact of DO on the EDL is stronger at short Debye screening lengths. A schematic of the role of DO in the response of graphene to ionic strength is shown in Fig. 3d.

Fig. 3: Graphene response to variations in dissolved oxygen.
figure 3

Chemiresistive response of graphene to variation of DO between 0 (N2 purged) and 11 ppm (air purged) in (a) 1 mM NaCl, (b) 10 mM NaCl, (c) 100 mM NaCl solutions, (d) schematic illustration of EDL at graphene/electrolyte interface focusing on the role of DO.

Full size image

Defining an experimental decay length

As discussed above, the chemiresistive response of graphene shows a current inversion with increasing salt concentration, similar to what has already been reported for the screening length of the solution. The plot of \({\lambda }_{{\rm{D}}}\) as a function of graphene sensor response (Fig. 4a, red dots) reveals a linear dependence in both regions. The slope of the fitted lines represents the dependence of sensor response on the screening length of the EDL at low (green) and high (blue) concentrations. Two regions with slopes of 0.0286 and −0.454 can be obtained for NaCl (Supplementary Fig. 13a, b). Since the deviation from ideality occurs at around 50–75 mM for NaCl, two different responses can be considered for the concentration range above 50–75 mM: (1) experimentally measured sensor response (\({ \% }_{{\rm{s}}}\)) obtained from the blue dots, (2) the theoretical response at high concentration in which \({\lambda }_{{\rm{D}}}\) is reduced continuously as a function of NaCl concentration (\({ \% }_{{\rm{D}}}\)) and is obtained from the relationship between \({\lambda }_{D}\) and the sensor response at low concentrations. The difference between %s and %D is representative of the discrepancy between the experimental screening length (\({\lambda }_{{\rm{s}}}\)) and its theoretical value (\({\lambda }_{{\rm{D}}}\)).

Fig. 4: Deviation of the experimental decay length from Debye–Hückel theory.
figure 4

a Graph of change in sensor response as a function of theoretical Debye screening length (blue) for two sections of ideal (black) and non-ideal (red) solutions. b The deviation of experimental screening length from ideal behavior measured by chemiresistive response (%S and %D represent the chemiresistive response based on D-H theory and experimental condition, respectively). Chemiresistive sensing responses of FLG to (c) LiCl, (d) NaCl, and (e) KCl.

Full size image

The ratio between ionic radius a and screening length \({\lambda }_{{\rm{D}}}\), (\(a/{\lambda }_{{\rm{D}}}\)), is related to the concentration of the solution. When \(a/{\lambda }_{{\rm{D}}}\ll 1\), ionic radii are significantly smaller than the screening length, meaning the solution is dilute, and the surface potential of the ion decays into the solution before reaching other ions. When \(a/{\lambda }_{{\rm{D}}}\ge 1\), the surface potential of the ion’s decays within a distance equal to or less than the ionic radius. Based on the literature, the deviation from ideality occurs at a ratio of \(a/{\lambda }_{{\rm{D}}}\approx 1\) when the forces are measured using a surface force apparatus (SFA) at surfaces without static charges38. By plotting \({ \% }_{s}\)–\({ \% }_{{\rm{D}}}\) as a function of \(a/{\lambda }_{{\rm{D}}}\) (Fig. 4b), it can be seen that up to 50 mM Na+, \({ \% }_{{\rm{s}}}\)–\({ \% }_{{\rm{D}}}\) is nearly zero, meaning that the theoretical and experimental screening lengths are in good agreement. Nevertheless, at \(a/{\lambda }_{{\rm{D}}}\approx\) 0.06, the value for \({ \% }_{{\rm{s}}}\)–\({ \% }_{{\rm{D}}}\) starts to increase since the screening length of the EDL no longer follows the D-H theory.

The analysis of \(a/{\lambda }_{{\rm{D}}}\) also highlights that the response of the graphene chemiresistive platform depends on the size of the cation. When KCl is used as an alternative salt to NaCl with a similar anion, the \({\lambda }_{{\rm{D}}}\) will be comparable for both salts at a particular concentration (the spectroscopically measured relative dielectric constants are almost the same for both Na+ and K+)50,51. However, it has been reported that the dielectric constant at the interface (EDL) is nearly 10% less than that in the bulk solution. This value can reach the permittivity saturation of water6 when a strong electric field (107 V m−1) is present22. Therefore, as an increase in concentration results in a decrease of \({\lambda }_{D}\), the larger cation is expected to reach non-ideality earlier due to a larger \(a/{\lambda }_{D}\) ratio. This hypothesis is confirmed by the chemiresistive response of graphene to LiCl and KCl compared to NaCl. As seen in Fig. 4c, adding LiCl does not lead to any current inversion up to 500 mM. Two simultaneous phenomena contribute to this effect: (1) the smaller size of Li+ compared to \({\lambda }_{{\rm{D}}}\) requires considerably higher concentrations to reach the current inversion zone (low \(a/{\lambda }_{{\rm{D}}}\)), higher than what is used here; (2) the possibility of Li+-intercalation or injection processes into graphene flakes. This is a commonly observed phenomenon in applications where graphene-based materials are in contact with electrolytes containing Li+. Accordingly, upon Li+ intercalation, the graphene is n-doped and the chemiresistive current decreases. Figure 4d, e shows chemiresistive data for NaCl and KCl, respectively. A comparison of the curves shows that the current inversion for KCl happens at lower concentrations (10–20 mM) compared to NaCl (50–75 mM), demonstrating that the EDL response of graphene is cation size dependent. This is consistent with the faster increase of \({ \% }_{{\rm{s}}}\)–\({ \% }_{{\rm{D}}}\) from 0 in KCl compared to NaCl in Fig. 4b.

The level of graphene defectivity may also impact the arrangement of ions in the EDL. The presence of defects alters the graphene solution interactions. For example, according to Supplementary Fig. 14a, the addition of NaCl causes a stepwise reduction in the chemiresistive current due to capacitive n-doping of graphene (\(\frac{{I}_{{\rm{D}}}}{{I}_{{\rm{G}}}}\) = 0.38) by the Na+ ions. However, this scenario is not seen in graphene oxide that has been reduced at 350 °C in a H2/N2 gas flow for 12 h (Supplementary Fig. 14b), 9 h (Supplementary Fig. 14c) or 6 h (Supplementary Fig. 14d), resulting in \(\frac{{I}_{{\rm{D}}}}{{I}_{{\rm{G}}}}\) ratios of 1.11, 1.25 and 1.43, respectively. This observation is in agreement with our recent study on the competing roles of the EDL and defect density in graphene during the interaction with the aqueous environment12. Accordingly, by increasing the defect density, three mechanisms may influence the formation and effect of EDL: (1) formation of alkali metal complexes with oxygen atoms in surface functionalities, (2) dominant defect-induced interactions caused by surface functionalities, and (3) discontinuity of the Stern layer42. The last phenomenon occurs primarily due to greater surface charges of graphene caused by oxygen. Hence, an interpretation of graphene surface defectivity is necessary to understand the impact of the EDL, and the conclusions of this study pertain to specifically graphene with an \(\frac{{I}_{{\rm{D}}}}{{I}_{{\rm{G}}}}\) ration in the range of 0.3–0.5, containing pH-responsive oxygen based functional groups.

We have presented a systematic study of the performance of graphene devices exposed to concentrated aqueous electrolytes. The change in the current passing through the graphene thin film upon interaction with an aqueous electrolyte was shown to be a means to indirectly determine the experimental screening length of the solutions. A more detailed picture of the EDL including the Stern layer, the diffuse layer, and the bulk solution, was elucidated, focusing on the role of DO at the graphene surface. It was demonstrated that the presence of cations in the Stern layer n-dopes the graphene through electrostatic charge screening, and these charges are neutralized by surface adsorption of DO. We also demonstrate that the inversion of the trend in chemiresistive current response of graphene devices upon adding NaCl is due to deviation from D-H prediction, and the experimental screening length increases as a function of salt concentration. This concentration was shown to be dependent on DO concentration, cation size, and defect density at the graphene surface. Accordingly, a lower DO concentration, larger cation size, and lower defectivity cause the current inversion to occur at lower concentrations. This study is a starting point for further exploration of the substrate-dependent experimental screening length of concentrated aqueous electrolyte solutions.

Methods

Materials

Graphite powder (99.99%) was purchased from Sigma-Aldrich. Glass slides were purchased from VWR. Isopropanol (99.99%) was purchased from VWR. Ultrapure water with a resistance of 18.2 MΩ cm was obtained from a Millipore Simplicity UV water purification system. Sodium chloride (99.99%) and lithium chloride were purchased from VWR. Potassium chloride was purchased from Caledon Laboratories Ltd). Sodium hydroxide (99%) and hydrochloric acid (37.2%) were used to adjust the pH in the experiments if necessary. Graphene oxide was purchased from ZEN Graphene Solutions Ltd, containing 33% oxygen atoms in the graphene structure.

Device fabrication

The fabrication procedure of a chemiresistive sensor from ultrasonically exfoliated few layer graphene (FLG) has been reported previously42. Briefly, 40 mg of graphite powder mixed with H2O: IPA (10.5:4.5 ml) are sonicated for 6 h in a bath sonicator (Elmasonic P60H ultrasonic cleaner) at 37 kHz and 30 °C. Then, the suspension is centrifuged twice using an Eppendorf MiniSpin Plus microcentrifuge, once at 14,000 rpm (13,140 × g) for 5 min, the supernatant is separated, then centrifuged at 14,000 rpm (13,140 × g) for 15 min, and the sedimented flakes are kept for the sensor fabrication.

To fabricate the chemiresistive sensors, the glass slides were rinsed with methanol and dried with N2 (g) to eliminate surface contaminations. Then, the slides are pre-patterned with two rectangles of 9B pencil trace to reduce the contact resistance between the active layer and the contacts. The glass slides are heated to 100 °C, and the FLG is airbrushed onto the surface uniformly until the average film resistance reaches ∼5 kΩ. Subsequently, copper tape is attached on both sides of the FLG film, covering the pencil-drawn rectangles. Lastly, parafilm (M) is used to mask the contacts to avoid direct contact with the solution. To ensure the sealing quality of the parafilm, the substrate should be heated to 70 °C.

For the measurement of DO with a commercial DO electrode, two calibration solutions were prepared. One aqueous solution was in equilibrium with air, and one purged with N2 (g) overnight. The electrode was placed in each solution for 10 min before calibration, and the DO concentrations for air equilibrated N2 (g) purged were set at 7 and 0 ppm, respectively.

Characterization

Raman spectroscopy was done using a Renishaw inVia instrument in the range of 500–3500 cm−1 with a spectral resolution of 2 cm−1. The 633 nm laser was focused using a 50× objective lens on the sample. The laser power was set to 1% to avoid surface damage during the analysis. The spectra were taken at three spots on each sample to ensure data reproducibility.

The UV-visible spectra of FLG in the presence of salts were obtained by placing 1 ml of solution (0.3 ml of salt solution with the desired concentration in 0.7 ml FLG-ethanol-DI water suspension) into a quartz cuvette in an Orion Aquamate 8000 spectrophotometer over a range from 250 to 600 nm.

For chemiresistive measurements, an eDAQ EPU452 Quad Multifunction isoPod was used to measure the sensing behavior of the sensor devices. The channel type for chemiresistive sensing was set to “biosensor”, with an applied voltage of 50 mV unless noted otherwise. For solution conductivity measurements, a conductivity probe (cell constant K = 1.057) was calibrated by using 0.1 mM KCl solution (12.64 mS cm−1) for 30 min. The pH electrode used in the experiments was calibrated with pH 4 and 7 calibration solutions before the experiments.

The electrochemical measurements were performed using a three-electrode configuration on an EC301 electrochemical workstation (PalmSens). The electrolyte was purged with dry N2 (g) for 40 min to eliminate the dissolved O2 prior to electrochemical measurements. Similarly, when desired, the dissolved O2 content was stabilized by purging with air for 40 min. All electrochemical measurements were done by using FLG, graphite, and Ag/AgCl as working, counter, and reference electrodes, respectively. For EIS, the AC and DC potentials were set 10 and 5 mV, respectively. The frequency range was chosen between 1 MHz to 10 kHz. Cyclic voltammetry measurements were conducted within a potential window from −0.2 to 0.8 V with varying scan rates of 20 and 50 mV s−1.