Effect of Ionic Liquid Structure on the Oxygen Reduction Reaction under Humidified Conditions

The oxygen reduction reaction (ORR) is widely studied in room temperature ionic liquids (RTILs), but typically in dry environments. Since water is known to affect diffusion coefficients and reaction outcomes, the influence of water on the ORR is expected to be significant. We have therefore studied the effect of RTIL structure on the ORR at different relative humidity (RH) levels using cyclic voltammetry. A broad range of cations including imidazolium-, ammonium-, pyrrolidinium-, pyridinium, sulfoniumand phosphonium-, and anions such as [BF4], [PF6], [NTf2] and [FAP]were employed. The cation was found to have a large effect on the reduction current of oxygen even at low humidity levels (< 40 RH %), whereas the anion mainly influenced the current at higher humidity levels (> 65 RH%). Consequently, the choice of cation needs to be carefully considered when selecting a suitable RTIL solvent for oxygen reduction in humidified environments. The size, structure and hydrophobicity of the ions was found to dictate the degree at which the RTIL is susceptible to changes in humidity. The physical characteristics of the RTIL electric double layer on platinum electrode surfaces were further investigated by atomic force microscopy force-curve studies in three selected RTILs. The results suggest that there is a significant amount of water incorporated at the electrode-RTIL interface in [C2mim][NTf2] and [N4,1,1,1][NTf2], but not in the more hydrophobic [P14,6,6,6][NTf2]. The presence of moisture has a significant impact on ORR currents in [C2mim][NTf2] even at extremely low humidity levels, which was verified by the higher level of water incorporation in [C2mim][NTf2] compared to [N4,1,1,1][NTf2] and [P14,6,6,6][NTf2]. Hydrophobic and large RTIL cations and anions (e.g. [P14,6,6,6] and [FAP]-) are recommended for applications where a stable ORR current response is required under humidified conditions.


Introduction
The reliable monitoring of gases under extreme conditions is an important criteria for sensors in industry, and is especially important for health and safety requirements. 1 Commercially available sensors usually consist of a threeelectrode setup, with an aqueous based electrolyte (usually water/H2SO4), and a gas permeable membrane that prevents solvent evaporation at high temperatures. 2 However, this membrane also reduces the diffusion rate of gases towards the electrode. 3 An alternative, proposed by Buzzeo et al., is the use of room-temperature ionic liquids (RTILs) as a non-volatile electrolyte, which removes the need for a membrane layer. 4 RTILs are composed of cations and anions that are liquid at room temperature due to asymmetry in at least one of the ions 5 and weaker ion-ion interaction forces. 6 RTILs are increasingly used in synthesis, 7 as well as in electrochemistry applications, [8][9][10][11] due to their promising characteristics, such as negligible volatility, high stability up to certain temperatures without decomposition, and high conductivity. 12 Due to their intrinsic conductivity, they have been extensively explored as alternative electrolyte materials. 13 The wide electrochemical windows (ca. 4-6 V) 14 of these solvents make them suitable candidates for electrochemical sensing applications; with the detection of gaseous analytes like hydrogen, oxygen or ammonia already reported in the literature. 15 However, the hygroscopic nature of RTILs is an issue for their use in practical environments that currently hinders their implementation in commercial membrane-free sensors. The tendency for RTILs to absorb moisture can lead to instability and unreliability in the measurements when the gas-sensor is subjected to real-atmospheres with changing humidity levels. [16][17][18] It is hence clear that water is a significant impurity in RTIL-based gas-sensors that need to be investigated; this can be indirectly monitored by studying the oxygen reduction reaction (ORR).
The ORR has been widely reported in aprotic RTILs for the purposes of oxygen gas detection. 4,8,[19][20] Katayama et al. published an almost reversible redox couple in ammonium-based ionic liquids with a current density ratio of 0.97. 21 It has also been shown that the reduction of oxygen can undergo different mechanistic pathways, depending on the presence and concentration of protic species such as water, which could cause a shift from a 1e -(equation 1) to a 2e -(equation 2) or even a 4e -(equation 3) reduction reaction. [22][23][24] This significantly complicates the reliable calibration of a sensor when exposed to different humidity levels, since water can act as an effective proton source, 25 opening up electrochemical reaction pathways that are dependent upon the concentration of water.
Therefore, several groups (including ours), have used hydrophobic RTILs, [26][27][28] or mixed the electrolyte with a hydrophobic polymer to reduce the water uptake. [29][30] However, this did not eliminate the water absorption problem entirely.
When examining the impact of absorbed moisture on the electrochemical processes at the electrode-RTIL interface, the structure of the RTIL at the interface must be taken into account. 17 Research performed by other groups has shown that as with conventional solvents, ionic liquids form electrical double layers (EDLs) at the electrode/solution interface. [31][32] The EDL structure at the electrode interface could have a severe impact on the transport of analytes such as gases or moisture towards the surface, which is required to induce an electrochemical reaction. Several methods including atomic force microscopy (AFM), [33][34] surface force apparatus (SFA), 35 neutron reflectometry, 36 sum frequency generation vibrational spectroscopy (SFG-VS) 37 and surface-enhanced infrared absorption spectroscopy (SEIRAS) 38 have been used to try to elucidate this behaviour. For RTILs, the thickness and number of layers at the electrode is strongly dependent on the charge density at the surface and the nature of the cation and anion. Depending on the cation and anion structure, the EDL that is formed when a potential is applied can be either more or less pronounced. 34,[39][40][41] Single as well as ion-pair layers have been detected at the interface using AFM. 33,42 With regards to layer thickness, theoretical simulations of these EDL structures are in good accordance with experimental data. 43 It is already well known that when the electrode is highly charged, more than one cation or anion layer can be formed near the electrode 44  there is a strong interaction between water and ionic liquid anion in the bulk phase. 47 This means that for different RTILs that have the same anion, the mutual solubility of water only varies slightly, while for RTILs with same cation and different anions, the effect is more pronounced. 47 It should also be noted that the bulk-phase behaviour of water in RTILs can be very different to the behaviour near a charged electrode surface. For example, Bi et al. 48 showed that the electrosorption of water on electrodes in humid, hydrophobic ILs was higher than in hydrophilic RTILs even though the bulk phase solubility of water showed the opposite trend. Therefore, the effect of water on electrode processes may be different to the trends expected based on the hydrophobicity/hydrophilicity of the RTILs.
In this work, we demonstrate the influence of the RTIL structure at the EDL -varying both the cation and the anion -and show its effect on the ORR in humidified environments using cyclic voltammetry (CV). Six different cation types (imidazolium-, pyrrolidinium-, pyridinium-, ammonium-, sulfonium-and phosphonium-) with different alkyl chain lengths and several different anions are analysed. Atomic force microscopy (AFM) is then performed under unbiased and biased conditions to provide insights into the impact of water on the EDL structure of RTILs on a charged surface, and to give an explanation for the electrochemistry results.
[N x,n,n,n ] + P C n H 2n+1  Figure S1 for the NMR data). The chemical structures of the ionic liquid cations and anions used in this study are summarized in Figure 1. attempted. However, the interaction of superoxide with the LFP material led to a change in ORR peak shape and current over consecutive scans, making measurements unreliable. Therefore, to investigate the influence of humidity on CV peak potentials for selected RTILs, ferrocene was added in-situ, and potentials were shifted such that the midpoint of the ferrocene/ferrocenium (Fc/Fc + ) redox couple was at 0 V.
Seven µL of the ionic liquid was drop-cast to cover all three electrodes on the TFE and purged for at least 45 min in a high purity nitrogen stream at a flow rate of 500 mL min -1 to remove dissolved gases and impurities such as oxygen and carbon dioxide. The relatively high volume of electrolyte on the small chip (compared to our previous experiments) ensures that the electrogenerated products from each scan can be effectively diluted, so that their accumulation is negligible. For all oxygen measurements, the integrated CE and WE were used. After purging the electrolyte with nitrogen gas for at least 45 min, oxygen was introduced into one arm of a modified glass T-cell. 51 The oxygen line was additionally connected to a humidity generator (Owlstone Humidity Generator OHG-4, Owlstone, Cambridge, UK) that separates the gas flow into 'dry' and 'wet' streams by bubbling the gas through a container of 400 mL of ultrapure water at a flow rate of 500 mL min -1 . Adjusting the dry to wet flow ratio results in different humidity levels and the measured dew point (Td) was converted to obtain the relative humidity percentage (RH%), using equation (4) 52 : with the constants m = 7.59 and Tn = 240.73 °C, the ambient temperature Tamb and the dew point Td in °C (Humidity Conversion Formulas, Vaisala). 53 The dew point was measured at a location close to the electrochemical cell to ensure that the actual humidity introduced into the T-cell is not affected by condensation or adsorption along the tubing (see Figure 2). Figure 2. Schematic of the gas humidifying system used for voltammetric measurements on thin-film electrodes (TFEs).
Since the humidity generator is not able to detect humidity levels lower than 1 RH%, measurements were conducted at a starting humidity level of ~1 RH%. It is also known in the literature that even with prolonged vacuum purging, a significant amount of water is retained (in the ~10-100 ppm range) 54 in the RTIL. An equilibration time of 45 min was employed before performing CV scans at each humidity level to ensure that the gas was fully saturated, particularly into the most viscous ionic liquid. Repeated CV scans for O2 reduction were carried out under constant humidity conditions to identify the duration for a stable peak current to be achieved (see Figures S3-S6 in the supporting information). 45 mins was found to be more than sufficient for equilibrium to be attained across all RTILs and humidities. Hence, before the commencement of experiments, the RTILs were also first purged under N2 gas for more than 45 mins. It is noted that water uptake times of up to 180 min were observed for RTILs at different humidity environments, 55 but that study employed a larger volume of ionic liquid and used a static set-up, in contrast to the large surface-to-volume (seven microliter RTIL droplet spread out over a thin-film electrode) and the constant flow system used in our experiments.
All experiments were carried out at least in triplicate and all data points were used to establish the trendline that was fitted by a logistic regression analysis. In order to remove the contributions caused by different solubilities and diffusion coefficients in the RTILs, the currents were normalised, where the measured currents at different humidities were divided by those measured at 1 RH%.

AFM measurements
Atomic force microscopy (AFM) experiments were carried out with a Bruker Dimension Icon interfaced to a computer with NanoScope 9.4 software in contact mode using activated Pt TFEs. Force curves were collected continuously with a scan rate of 0.16 Hz while the ramp size was between 100 and 160 nm. A silicone barrier (Selleys Silicone Sealant, Padstow, NSW, Australia) was constructed around the electrode cell to confine the ionic liquid during the AFM measurements. All experiments were performed with a silicon nitride AFM tip (SNL-10, spring constant 0.35 Nm -1   The PM3 calculated SEs of selected single isolated RTIL molecules are also listed in Table S1 to provide the local affinity of each ion towards surrounding water molecules; the greater the SE value, the more hydrophobic the molecule is likely to be. It is noted that these values are relevant on the inter-molecular scale but may not apply for the bulk RTIL as they do not account for RTIL-RTIL and bulk interactions, and hence may not coincide with bulk properties of hygroscopicity and hydrophobicity. The SEs (listed in  Figure 3. A reversible one-electron reduction to superoxide (see equation 1) was observed in a dry environment (< 1 RH%, black line). The CV shape is typical for an electrochemically quasi-reversible process, which is well reported for the oxygen/superoxide redox couple in RTILs. 16 At increased humidity levels, the reductive peak current becomes larger and the reduction process became more irreversible, resulting in a decreased superoxide oxidation peak.

Computational estimation of ion dimensions and water-affinity
Ferrocene was added to the solution as an internal reference due to significant potential shifting occurring at the Pt quasi-RE in the presence of the humidified gas. The voltammograms were shifted so that the midpoint of the ferrocene/ferrocenium redox process was at 0 V, and a substantial shift (~600 mV) of the ORR potential, from ca.

Effect of Anion
The type of anion is known to have a significant impact on the hygroscopic behaviour of RTILs. 54,57 The bulk solubility of water in ionic liquids containing the [FAP]anion is much lower compared to [NTf2] -, [PF6] -, or [BF4] --see trend in equation 6, and also in reference [57]. It was previously shown that electro-absorption of water in the EDL is more pronounced for hydrophilic anions at negative potentials. 58 In an ionic liquid, absorbed water molecules are known to preferentially interact with the anion. 6,47,[58][59] Hence, it is important to investigate how different anionic species may affect the ORR reaction at different humidity levels. Figure 4 shows the normalised ORR current as a function of increasing humidity levels in four different [C4mim] + -based RTILs. The current is mostly unaffected at low humidity levels (< 15 RH%) suggesting that the cation is the determining factor at low humidity levels. This is consistent with AFM 32-33, 39, 60 and SEIRAS 38 studies, where it is well known that the innermost layer at a negatively biased surface is mainly composed by cations. The normalised ORR current then begins to increase above 15 RH%. The current at high humidity levels (> 95 RH%) follows the trend: which is consistent with the computationally predicted molecular-hydrophobicity and volume trends for individual anionic molecules (see trends 5 and 6), and by COSMO-RS calculations 47 from a separate study.
For the [BF4] -RTIL, there is a lower than expected current trend between 20 and 80 RH%. It was suspected that this may be due to insufficient equilibration time. However, waiting 130 min instead of 45 min before recording the first scan did not affect the current response (see ESI Figure S3). Instead, repeated CV scanning caused an increase in the current, suggesting a possible side-reaction and build-up of electrogenerated products contributing to the current. This is not surprising considering that [BF4] -, as well as [PF6] -, can undergo hydrolysation. 61 At 90 RH%, the electrogenerated products are likely saturated at the EDL from the initial scan; consecutive scans at this humidity level can cause severe changes in the ORR current due chemical reactions with previously electrogenerated species, such as H2O2. Overall, the results in Figure 4 suggest that bulky, hydrolytically stable and hydrophobic anions such as [FAP]are recommended for applications where a stable ORR response in humidified environments is required, e.g. for oxygen sensing.

Effect of cation with a hydrophobic anion, [NTf2] -
A negatively polarised electrode during oxygen reduction voltammetry implies that the EDL structure at the electrode-RTIL interface will be dominated by the RTIL cation. In contrast with anions, cations generally only have a slight influence on the solubility of water in RTILs. 45,47 Properties such as the molecular affinity to water (predicted by SE values), coupled with cation size (based on CPK volumes) -which may sterically hinder the adsorption of other species at the electrode surface (see Table S1 in the ESI) -need to be considered. In this study, only RTILs displaying quasi-reversible ORRs were included, i.e. the pyridinium ionic liquid was omitted due to an irreversible ORR process (see ESI, Figure S7). The presence of an obvious superoxide oxidation peak for the remaining RTILs implies that any impurities in the RTILs (e.g. left over from the synthesis procedure) are minimal, and do not affect the electrochemical processes occurring during the oxygen reduction reaction. appears to have a more significant influence on the ORR current at lower humidity levels (i.e. < 40 RH%). There is a clear difference between the humidity level at which the ORR current starts to sharply increase -referred to as the "onset-humidity" -typically occurring at below 30 RH%, and follows the trend:  (8) This is in contrast with the different anions discussed in Section 3.1.1, where the onset-humidity appears to be almost unaffected by the type of cation present. Significant variability in the data points towards higher humidity levels (see discussions in section 3.2.3) prevents a confident evaluation of the 95 RH% current trends, although it is clear that the ORR current in the [P14,6,6,6] + RTIL is much less affected by water compared to the other cations. The variability in current could be due several factors including increased follow-up chemical reactions and even bubble formation at the electrode-RTIL interface due to water-splitting reactions (see Figure S11 in the ESI).
Zhong et al. 60 Figure 5. A layer of ammonium cations will form at a negatively charged surface. At sufficiently high negative surface polarisation, the positively charged nitrogen group will be pulled towards the negative surface charge, inducing a reorientation of the short alkyl chains. Ultimately, this results in a dense structure of cations at the electrode interface, where longer alkyl groups are expected to point away from the charged surface, orientating in the most stable energetic state to maximise interaction with the anion layer above. However, the three shorter alkyl chains of the ammonium RTILs studied are too short and less flexible to effectively cover the electrode surface area between repelling cations, and do not form a highly dense hydrophobic layer at the electrode interface. This can leave cavities and pathways for water transportation to the electrode surface.  Fedorov and Kornyshev's 62 study of the ionic liquid (IL) EDL at the electrified solid-IL interfaces reveals a "generally more complicated than expected" structure, due to the tendency of ILs to self-assemble. Previous work suggests an EDL structure that is comparable to aqueous electrolytes, consisting of an adsorbed ion layer succeeded by an electrostatically bound diffuse layer. [63][64][65] However, it was found from AFM studies that a multilayer of both single and cation-anion pair layers can form at the interface, and the number of layers increased and were more tightly organised at higher potentials. 34 conditions has been previously reported, but these conclusions may not be valid for a significantly rougher surface.

General trends in AFM force-curves
Overall, the force-curves in Figure 7 for both charged and uncharged surfaces show step-like features, superimposed on a broad curve with the force gradually increasing towards 0 nm separation. The broad curve is related to the compression of stable EDL structures by the tip as well as the interaction of the tip with the rough electrode surface. The steps correspond to tip-interaction with structured layers of molecules (e.g. cations, anions, water, etc.) orientated and assembled over the electrode surface. Subtle step-features are also present in between the more pronounced ones, which are attributed to either different conformations of ions, the presence of interstitial water or may be caused by roughness of the electrode surface. In the following discussions, the more pronounced layers will be referred to as layers 1, 2, 3, etc., with layer 1 being the surface-adsorbed innermost ion layer.
The thickness of the EDL can be observed by the occurrence of a "snap-on" event. This is defined as the distance at which the descending tip experiences the first significant attractive force towards the diffuse layer, and is observed as an abrupt drop in the measured force coinciding with the start of the compression behaviour. Within the EDL, the pronounced layer-thicknesses can be estimated by measuring the width between each step of the force-curves, and are expected to be a function of the ion or ion-pair sizes. The force required for the AFM tip to puncture through a layer is referred to as the "push-through" force. Zhong et al. 60  silica-terminated silicon AFM tip (of 5-10 nm radius). 55 The steps appeared more pronounced under humidified conditions (43 RH%), and a negligible compression behaviour was observed. This is probably because of the atomically smooth and uniform surface of Mica as well as the fact that water is required to charge the Mica surface. The uniformly charged and smooth surface of Mica allows highly ordered establishment of the first layer, which can then allow more ordered subsequent layers to form, whereas more disordered and less defined structures are expected on the rougher Pt-TFE surface used in our study (see discussions in ESI).
Also, an interface-selective vibrational spectroscopy study on the behaviour of ammonium ILs at Au interfaces showed that alkyl chains form a loose structure that reaches into the anion layer, resulting in a more interdigitated structure rather than a clearly separated alternating cation/anion layer. 38 This behaviour could further add to the contribution of the superimposed compression-like feature, especially for the longer chain RTILs (e.g. [P14,6,6,6][NTf2]).
For this work, we have chosen to examine the EDL on the non-ideal Pt-TFE (also used for the CV experiments) since this is a more typical representation of an electrode surface found in real sensors.

Comparison of the EDLs of individual RTILs
To aid with the discussion, estimations of the layer thicknesses were obtained from computationally optimised molecular structures of each ion, as shown in Figure S13 in the ESI. Orientations of the ions were based on suggestions from other research. 37-38 . Since only the first few layers are expected to have a strong influence on the electrochemistry at the Pt-RTIL interface, they will be focussed on in the following sections. [C2mim] + is known to rest relatively flat and well-ordered on the surface. 37 The stability of this orientation will determine the effectiveness of the cation to block moisture adsorption at the electrode surface.
Comparing the layer thicknesses with Figure 7, it can be assumed that the innermost layer almost equals the size of a cation (≈0.25 nm), followed by an anion layer (≈0.6 nm), as expected. Further layers are significantly thicker (around 0.5nm) that would match a water + cation layer, or an anion layer. So the layer thickness suggests water incorporation between the ion layers, which was also concluded by Zhong  However, for [P14, 6,6,6][NTf2], the long hydrophobic alkyl chains are able to significantly shield the influence of the surface electric field, thus reducing the stabilising effect of a negative potential bias and therefore no obvious changes in the structuring can be seen. Our experimental results agree well with a molecular dynamics simulation study, 48  the layers remain relatively unaffected by negative biasing, which further substantiates this interpretation. In conclusion, potential biasing results in an increased force required to push through the innermost layers, and for the small RTIL cations, obvious structural layers begin to vanish due to the competition between water and IL ions.

Conclusions
The ORR mechanism was studied by CV on Pt-TFEs in several different RTILs under humidified conditions. Peak potentials and currents were found to vary with humidity levels, showing a gradual shift in the reduction mechanism with increasing humidity (from < 1 to 95 RH%). The degree by which a given RTIL is susceptible to humidity changes was found to be a function of ion size, structure, and water affinity. Generally speaking, the cation influences the ORR current at humidity levels below 40 RH% due to the more pronounced structuring of the ions near the negatively charged electrode, while the anion has a more significant impact on the ORR reaction above 65

Acknowledgements
This work was supported by an Australian Research Council (ARC) Future Fellowship Award (FT170100315) for DSS.