Deconvoluting Kinetics and Transport Effects of Ionic Liquid Layers on FeN4-based Oxygen Reduction Catalysts

9 ABSTRACT: The use of ionic liquid layers has been reported to improve both the activity and durability 10 of several oxygen reduction catalysts. However, the development of this technology has been hindered 11 by the lack of understanding of the mechanism behind this performance enhancement. In this work, we 12 use a library of ionic liquids to modify a model FeN 4 catalyst (iron phthalocyanine), to decouple the 13 effects of ionic liquid layers on oxygen reduction kinetics and oxygen transport. Our results show that 14 oxygen reduction activity at low overpotentials it determined by the ionic liquid s’ influence on the *OH 15 binding energy on the active sites, while oxygen solubility controls transport at high overpotentials. 16 Finally, using porosimetry, we have demonstrated that the distribution of the ionic liquids on the catalyst 17 is inhomogeneous, and depends on the nature of the ionic liquid used.


Introduction
To mitigate global warming, green H2 and its utilization in fuel cells have been identified as key technologies.The biggest drawback of low temperature hydrogen fuel cells is the high content of platinum-group-metals (PGM), required to drive the sluggish oxygen reduction reaction (ORR).Despite recent efforts on decreasing platinum, at the present state 0.1 to 0.2 gPt kW -1 are still required, 1 accounting for 40-50% of the overall cost of fuel cells for automotive applications. 2he same time, PGM-free ORR catalysts have been developed, with impressive recent advances in both stability and activity, with a particular attention to iron and nitrogen-doped carbons (Fe-N-C). 3,4rtheless, their performance is still far from the activity and durability targets set by the US department of energy and the European commission's fuel cell and hydrogen joint undertaking. 5,60][11][12][13][14][15][16][17][18][19][20][21][22][23][24] This technique, which was first reported for platinum-based catalysts, 11 is also effective for some noble-metal-free catalysts. 13,21,23,25As we highlighted in a recent review, this technique holds a remarkable potential to improve activity, durability and even selectivity of oxygen reduction catalysts in a facile and inexpensive manner. 26However, progress in this direction is limited by the lack of understanding of the mechanism behind performance enhancement in presence of ionic liquid layers.
18,23 However, this explanation alone fails to rationalise all observed phenomena, in particular, the lack of a linear dependence of ORR kinetics on oxygen concentration. 18e complexities are compounded by the challenges in measuring O2 transport in ionic liquids, as a result of the low solubility of oxygen. 27Oxygen solubility results reported in different manuscripts [27][28][29][30][31][32][33][34] display large variations, with a percentage error of up to 60% (Figure S6a). 26Moreover, to the best of our knowledge, results obtained with different experimental techniques have never been compared in a single study. 27,31,32,35Most studies neglect the difference between O2 concentration and thermodynamic activity, which is inappropriate for oxygenophilic ionic liquids.Since the field -thus farhas lacked a quantitative parametrization of O2 transport through ionic liquids, it has been unable to deconvolute the effects of O2 diffusivity and solubility that leads to the maximum ORR activity as a function of thickness of the ionic liquid layer. 20,23,25,36,37as been explained by the combined effect of the high oxygen solubility in ionic liquids, responsible for the initial increase in activity, and the low oxygen diffusivity, which introduces transport limitations in IL layers.
On top of transport-related phenomena, the ionic liquid effect on ORR performance has also been attributed to kinetics causes.For example, the reduced water presence at the interface could cause a dehydration and consequent destabilization of the HO* intermediate, which should cause an activity increase for those catalyst on the weak HO* binding side of the Sabatier volcano. 38,39Hunag et al showed that on Pt/C, the peak commonly attributed to the adsorption of oxygenated species moves to more positive potential in presence of the ionic liquid MTBD BETI 37 indicating a destabilization of the HO* intermediate. 38,39However, there is no systematic study of the effect of ionic liquids on the binding energy of oxygenated intermediates and on the effect of this phenomenon on the activity of the catalyst.
Finally, an aspect that has been largely overlooked is the distribution of the ionic liquid on the catalyst support.All the observations reported so far assume that ionic liquids form a thin homogeneous layer on the surface of the catalysts and that all the active sites are in contact with the ionic liquid.
In this work, we deconvolute and quantify the above-mentioned phenomena controlling O2 reduction on non-precious metal catalysts modified by ionic liquid thin films.We report for the first time, an experimental method to distinguish between oxygen solubility and thermodynamics activity.
Additionally, we compare experimental oxygen transport results obtained using different techniques, obtaining a reliable estimate of oxygen transport parameters in ionic liquids.By deriving equations from first principles, we quantify the effect of oxygen solubility, diffusivity and thickness of the ionic liquid layer on the overall rate of reaction.Our analysis offers a facile and novel way to predict the effect of ionic liquids on the oxygen transport and to forecast the ideal properties of ionic liquid layers.To account for the role of kinetics, we use voltametric analysis to monitor the effect of a library of ionic liquids on the *OH binding energy of Fe(II) phthalocyanine (FePC).Finally, using thermal and porosimetry analysis, we explore how the distribution of the ionic liquids on the surface depends on the nature of the ionic liquid itself.

Results and Discussion
1][42][43][44] Therefore, in this manuscript we decided to use iron(II) phthalocyanine (FePC) as a model FeN4 catalyst with well-defined active site.FePC was deposited on conductive graphene and modified with the addition of 5 selected ionic liquids (20%w), shown in Figure 1a (characterization of the pristine catalyst can be found in the supplementary information, section S11).The effect of the ionic liquids on the overall ORR performance of the catalyst was studied using a three-electrode cell, with a rotating disc working electrode (RDE).Linear sweep voltammograms were collected in oxygen saturated 0.1M KOH and are shown in Figure 1b.Clearly, the addition of the ionic liquid has a significant impact on both ORR kinetics at low current densities and oxygen transport at high current densities, effects that vary significantly with the chemical structure of the ionic liquid itself.Data were collected in oxygen saturated 0.1M KOH, with a scan rate of 10mVs-1 and rotational speed of 1200rpm, at a temperature of 25 O C. The pristine catalyst loading was maintained constant at 0.14 mg/cm 2 with an addition of 20%w ionic liquid.Before recording the oxygen reduction activity, the catalyst was pre-conditioned with 5 cyclic voltammograms in nitrogen-saturated electrolyte, in a potential range of 0.1V vs RHE to 1.0V vs RHE.The ORR activity of the pristine catalyst is comparable to that reported in literature, as shown in Figure S1e  c) Background-subtracted square wave voltammogram of FePC/G modified by the addition of ionic liquids, zoom-in on the high potential peak.Obtained in N2-saturated 0.1M KOH, using a potential step of 4mv, modulation amplitude of 20mV and a frequency of 2Hz, resulting in a scan-rate of 8mVs -1 , at a temperature of 25 O C. The pristine catalyst loading was maintained constant at 0.14 mg/cm 2 with an addition of 20%w ionic liquid.Before recording the square-wave voltammogram, the catalyst was pre-conditioned with 5 cyclic voltammograms in nitrogen-saturated electrolyte, in a potential range of 0.1V vs RHE to 1.0V vs RHE.After the CVs, SWV measurements were repeated 3 times in the same potential range and no change was observed between the scans.The second scan is shown here d) Potential for 0.1mAcm-2 (in O2-saturated 0.1M KOH), as a function of the position of the square-voltammetry peak in N2 saturated 0.1M KOH.Data presented are the average of three independent measurements and the error represents the highest and lowest data measured e) Ex-situ XANES of FePC/G modified by the addition of 20%w of the ionic liquids EIM NTF2 and BMIM PF6, at the First of all, we focused our study on the effect of the selected ionic liquids in the kinetics region of the LSV.The mechanism of oxygen reduction on iron-based catalysts is still debated, even for the case of model molecules with defined structure, like iron phthalocyanine.Nevertheless, it is still possible to gain some insight by simply monitoring the position of the cyclic voltammetry (CV) peaks.In fact, the high potential CV peak of FePC is attributed to *OH desorption, [45][46][47] and its position is therefore an experimental measurement of the *OH binding energy on the catalyst site.Rather than cyclic voltammetry, we used square-wave voltammetry (SWV) to obtain a more accurate identification of the peak position, thanks to high rejection of capacitative current and better sensitivity to Faradaic processes (as shown in Figure S1b and S1c). 48As show in in Figure 1c, the presence of ionic liquids has a visible effect on the position of the high-potential peak, suggesting an alteration of the *OH binding energy.In particular, all the ionic liquids except EIM NTF2 move the peak to less positive potentials, suggesting a strengthening of the *OH bond.
Figure 1d shows ORR activity as a function of the position of the SWV peak.There is a good correlation between the catalyst performance and the peak position, suggesting that the effect on *OH binding energy dominates the change in ORR kinetics caused by the ionic liquids.In line with reports that the pristine catalyst (FePC) should lie on the strong-binding side of the Sabatier activity volcano, 49 the ionic liquids that strengthen the *OH bond are found to decrease the ORR activity. 45This appears in contrast to reports that the same ionic liquid layers that are hereby detrimental for ORR kinetics, have a positive impact on platinum-based catalysts. 15,37,50This observation shows how the effect of ionic liquids depends not only the nature of the liquid itself, but also on that of the catalyst.
A possible explanation of the effect of ionic liquids on the *OH bond originates from their hydrophobic nature.The decrease in water concentration around the active sites could cause a dehydration of the *OH intermediate and its consequent destabilization. 51,52This could explain the weakening of the *OH bond reported after the addition of ionic liquid on Pt/C catalysts, 15 but not strengthening observed on FePC.In this case, the change in *OH binding energy could be attributed to co-adsorption of the IL cation, similarly to what has been previously reported for electrolyte cations. 53,54This hypothesis would be in line with the observation that the cation of the ionic liquid has a more significant effect on the *OH binding energy than the anion.
Although we take the view that the effect on the *OH bond has a strong influence on ORR kinetics, we cannot exclude the possibility that other factors might play an important role.Even in the absence of liquid water and potential control, ex-situ XANES data obtained at the Fe K-edge of FePC/G (Figure 1e) show that the ionic liquids influence the propensity of the Fe towards oxidation, causing a slight shift in the whiteline and a modification of the pre-peak intensity.In particular, EIM NTF2 appears to decrease the oxidation state of Fe, while BMIM PF6 has the opposite effect.This could cause a change in affinity to *OH and *O and explain the observed weakening of *OH bond for EIM NTF2 and strengthening for BMIM PF6.
Finally, it cannot be excluded that the ionic liquids might block access to some active sites, thus decreasing catalyst utilization.However, this phenomenon alone would not explain the observed modification of the *OH bond.

Reliably determining oxygen transport data in ionic liquids
In the previous section we isolated the effect of ionic liquid layers on the kinetic of oxygen reduction and identified a descriptor (the *OH binding energy), to predict the effect of ionic liquids on the ORR kinetics.Now, we will move on to discuss the effect of ILs on oxygen transport, starting from the reliable determination of oxygen solubility in ionic liquids.
Given the high variability in the oxygen transport data reported in the literature and the unavoidable sources of error associated with experimentally determining transport parameters, we decided to compare oxygen permeability results obtained using three of the most common techniques: gravimetric, volumetric and electrochemical.The first two are based on the incremental addition of oxygen, while monitoring the weight or pressure, respectively, while the electrochemical method relies on the 1electron reaction of dissolved oxygen on a microelectrode.Details of the experimental procedure can be found in the method section, while an in-depth discussion of the advantages, sources of error of each technique and choice of fitting times can be found in the supplementary information, section S2, and S3.
Figure 2 shows the solubility (a), diffusivity (b) and permeability (c) results obtained with the different techniques.Solubility results obtained with all the methods show the same trend, with the anion having the biggest influence on oxygen solubility (with NTF2>PF6).However, while the volumetric and gravimetric methods are in good agreement, the electrochemical method appears to consistently predict lower values.
Despite all the possible sources of inaccuracy, the significant difference observed between the electrochemical and other methods cannot be solely attributed to experimental errors.We instead propose that the electrochemical method calculates the activity of oxygen (aO2), rather than its solubility.
Activity is a thermodynamics concept which Permeability is the product of concentration and diffusion coefficient and controls oxygen transport to the active site.According to the results obtained via the electrochemical method, oxygen permeability in BMIm NTf2 and EIm NTf2 is higher than in the electrolyte, indicating that oxygen transport should be faster in these ionic liquids than in aqueous electrolyte.

The effect of ionic liquid layers on oxygen transport
Once we obtained reliable experimental measurements for oxygen solubility, thermodynamic activty and diffusivity in ionic liquids, these results were used to predict the effect of ionic liquid layers on the transport of oxygen to the surface of the rotation-disc electrode.In the RDE configuration and transport-limited conditions, oxygen transport is controlled by diffusion through the so-called diffusion layer, as shown by the oxygen concentration profiles in Figure 3a and Figure 3b.In absence of the ionic liquid, oxygen concentration is zero at the catalyst surface and equal to oxygen solubility in the electrolyte at a distance equal to the diffusion thickness.When a thin film on ionic liquid is added on the surface of the catalyst, discontinuity in oxygen concentration is introduced at the ionic liquid/electrolyte interface.Assuming that the resistance to oxygen transfer across the interface is negligible, the catalyst surface is flat and the ionic liquid is a homogeneous film, oxygen transport can be calculated as shown in Equation 3 (see supplementary information section S5 and S6).
Where   ,   are the thickness of the diffusion and ionic liquid layer respectively,  and   are the diffusion coefficients of oxygen in the electrolyte and ionic liquid.SR is the the ratio between oxygen solubility in the ionic liquid ( 2,IL ) and the electrolyte ( 2, ).Finally,  , and  , are the concentrations of oxygen at the interface, on the electrolyte and ionic liquid sides respectively.
The results of this model, compared to the experimental diffusion-limited current at 1600rpm, are shown in Figure 3e.As expected, the transport-limited current increases with the permeability of oxygen in the ionic liquid but the dependence is not linear as oxygen transport is controlled by the layer with lowest oxygen permeability.This results in the current decreasing much faster for ionic liquid with permeability lower than the electrolyte, than it increases in the opposite case.The main conclusion is that oxygenophilic ionic liquid can be used to improve oxygen transport, but not as much as a linear correlation with oxygen permeability would predict.
Using Equation 3, we can also observe the effect of various parameters, such as the thickness of the ionic liquids or of the diffusion layer, on the current density.Figure 3f shows the ratio of the diffusionlimited current in presence and absence of the ionic liquids, as a function of the rotational speed.The effect of the ionic liquid layer becomes more pronounced with higher rotational speeds.The model fits the experimental data reasonably well and the discrepancies might result from an inhomogeneous distribution of the ionic liquid on the catalyst, which is discussed in the following section.
So far, we have discussed the effect of oxygen transport in diffusion-limited conditions.When both the oxygen transport and reaction kinetics play a role, the effect of the ionic liquid is more complex.On one side, the ionic liquid affects the transport of oxygen as described above, while on the other side it changes the concentration of oxygen at the catalyst surface, influencing the rate of reaction.In this case, the total oxygen flux or oxygen reaction rate can be calculated according to Equation 4.
Where  2 is the rate of consumption of oxygen,   is the oxygen reaction constant (which is potential dependent) and  0 is the concentration of oxygen at the catalyst surface.The reaction constant can be further related to the overpotential (η) via the Butler-Volmer equation for irreversible reactions (Equation 5), where F is the Faraday constant, z is the number of electrons, R is the universal gas constant, α is the charge transfer coefficient, T is the temperature and ko is an empirical quantity.
This equation highlights how the ionic liquids affect both the rate of reaction (through oxygen solubility) and the transport of oxygen (through oxygen permeability), with the relative effect depending on the applied potential.Given a library of ionic liquids with known oxygen transport properties, equation 4 offers an analytical expression to determine which ionic liquid would provide the highest current at any overpotential.Clearly, the ideal ionic liquid should feature both high oxygen solubility and diffusivity.
However, these two parameters cannot usually be independently tuned. 29Unless this limit is overcome, the optimal ionic liquid layer should maximize the oxygen solubility, while minimizing the thickness.

The distribution of ionic liquids on the catalyst support
From the above analysis, it can be deduced that, for practical applications, the ideal ionic liquid layer for oxygen reduction should have high oxygen solubility, minimum thickness and homogeneous distribution.This observation leads to the necessity of understanding how ionic liquids distribute on the catalyst surface, which we will hereby achieve via thermal and porosimetry analysis.The red data point refer to the experimental results obtained at 1600 rpm, while the grey line is the result of the transport model described above, with a 100nm-thick ionic liquid layer f) Ratio between the diffusion-limited current density in presence and absence of the ionic liquid, as a function of the rotational speed.The full line shows the theoretical data obtained with the equation described above and using the solubility and diffusivity data obtained using the electrochemical method.The thickness of the ionic liquid layer was adjusted to fit the experimental data (squared data points) and was found to be 100 nm.
Differential scanning calorimetry (DSC) and thermogravimetric analysis couple with mass spectroscopy (TG-MS) were used to confirm the interaction between the ionic liquid and the catalyst, exploiting the fact that the properties of an liquid confined into porous materials are different from those of liquid in the bulk. 55DSC measurements on bulk EIM NTF2 and on the EIM NTF2-modified catalyst show that the typical glass transition, crystallization and melting peaks present in the pure ionic liquid disappear upon impregnation on the porous catalysts (Supplementary information, section S7).This confirms that EIM NTF2 has been successfully incorporated on the catalyst and that the ionic liquid remains in an unaffected liquid state up to -100°C.
Figure 4 shows the results of TG-MS.While the bulk ionic liquid starts to decompose at 222°C, the onset decomposition temperature of the ionic liquid incorporated in the catalyst decreases to 200°C (supplementary information section S7). 56,57When comparing different ionic liquid loadings (Figure S10), it can also be noticed that at higher loading the onset decomposition temperature of the confined ionic liquid approaches that of the bulk, confirming that the thermal stability drop originates from interaction of the ionic liquid with the support.Porosimetry analysis further allowed us to study IL distribution in the catalyst pores, with an incremental loading of 2%w to 80%w. Figure 5a and Figure 5b show the N2 sorption isotherm and pore diameter distribution of the catalyst modified with EIM NTF2.As expected, the impregnation with the ionic liquid leads to a decrease in both surface area and pore volume.The micropore volume decreases rapidly and is almost nil already at 10%w ionic liquid loading (Figure 5d), while the average pore diameter remains quite stable throughout the ionic liquid addition (Figure 5e), suggesting that EIM NTF2 is completely filling up some pores (starting from the micropores), leaving others dry.In fact, if the ionic liquid were to distribute uniformly on the surface of the catalyst, one would expect a gradual broadening From grey to dark purple, the curves represent the FePC/G catalyst, modified with an incrementally increasing amount of the ionic liquid EIm NTf2 c) Surface area calculated by BET (normalized to the mass of the pristine catalyst), as a function of the loading of the ionic liquid EIM NTf2 d) Micropore volume calculated using t-plot analysis (normalized to the mass of the pristine catalyst), as a function of the loading of the ionic liquid EIM NTf2 e) Average pore diameter, for pores < 4 nm (green) and <18 nm (orange), as a function of the loading of the ionic liquid EIM NTf2 f) Experimental pore filling degree determined from the pore volume distribution, as a function of the theoretical pore filling degree, calculated from the mass of the ionic liquid added.g), h), i) l) Schematic of the ionic liquid distribution in pores: g) shows the pristine support, h) an ionic liquid ideally distributed to form a monolayer and the ionic liquid distribution that we reconstruct from the porosity analysis, for the case of i) EIM NTF2 and l) BMIM PF6.
Ideal ionic liquid distribution Experimentally observed EIM NTF2 distribution Experimentally observed BMIM PF6 distribution Pristine Support  of the pore diameter distribution and a linear decrease in the average diameter.This indicates that the support-ion interactions are weaker than ion-ion interactions.The fact that the BET surface area initially decreases, but remains constant above 40%w IL loading (Figure 5c), also supports this hypothesis, as above this value the ionic liquid keeps filling up increasingly larger pores and depositing in the form of droplets, while leaving a significant proportion of the surface uncovered.Since it has been reported that different anions can cause a different distribution of the ionic liquid on a support, 58 we studied the effect of the other anion considered in this study (PF6), by looking at BMIM PF6.This ionic liquid showed a similar behavior to EIM NTF2, with a rapid decrease in micropore volume and constant average pore diameter (Figure S12).However, as it can be observed in Figure 5f, for the case of BMIM PF6, the pore filling degree experimentally determined by the porosimetry analysis is much higher than the volume of ionic liquid that was added to the sample.This suggests that this BMIM PF6 can sit at the entrance of some pores, blocking access to them without fully filling their volume.
The IL distribution deduced from this study are summarized in Figure 5 g-l.Rather than forming an ideal monolayer (Figure 5h), our experiments suggest that EIM NTF2 distribute inhomogeneously, gradually fill up the some pores starting from the smallest (Figure 5i) while BMIM PF6 offers a similar behavior while also blocking access to some pores (Figure 5l).

Conclusion
In this work, the effects of ionic liquids on the transport of oxygen and kinetics of oxygen reduction were studied and deconvoluted.
• To account for the effect of ionic liquids on kinetics, we correlated the voltametric peak for *OH adsorption to the kinetic activity.There is a strong correlation between the peak position and oxygen reduction activity.Ionic liquids were found to generally strengthen the bond, having a detrimental effect on the ORR kinetics of FePC.We hypothesize that the co-adsorption of the IL cations might cause the observed change in *OH bond.
• To address the lack of reproducibility in the reported values for oxygen transport in ionic liquids, we compared the results obtained with three common methods.We developed a facile method to experimentally distinguish oxygen solubility (via gravimetric and volumetric techniques) and the thermodynamic activity of oxygen in electrochemical conditions (via electrochemical methods).
• Having characterized oxygen permeability in the ionic liquids, we quantified the effect of ILs on the transport of oxygen.We derived an equation to allow the calculation of oxygen reduction current at a given potential, knowing the kinetics of oxygen reduction on the pristine catalyst and the transport properties of oxygen in the ionic liquids.This equation additionally allowed us to conclude that the highest ORR activity can be obtained by maximizing oxygen solubility in ILs, at the expenses of diffusivity, while using a homogeneous layer of ionic liquid with minimum thickness.
• Finally, using porosimetry analysis, we were able to characterize the distribution of ionic liquids on the catalyst surface.EIM NTF2 was observed to deposit in clusters, gradually filling the pores, starting from the smallest.BMIM PF6 showed a similar behavior but was additionally found to sit at the entrance of some pores, blocking access to them.This highlights the potential of further improving solid catalysts with ionic liquid layers, by optimizing the synthesis method, to allow the formation of uniform layers.
We envisage that our findings are also applicable to other electrochemical reactions.In particular, the hydrophobic nature of these ionic liquids could be further beneficial for the reactions that compete with hydrogen evolution, such as CO2 reduction 59,60 and N2 reduction. 61Additionally, oxygenophilic ionic liquids can also be used to significantly increase the concentration of carbon dioxide at the active site.
In fact, the solubility of CO2 in ionic liquids is three orders of magnitude higher than in water. 62,63The method hereby proposed to experimentally distinguish between oxygen solubility and thermodynamic activity in ionic liquids could also be applied to carbon dioxide (and in principle to any other gas), to obtain a more reliable prediction of their performance, which could find application in CO2 capture and CO2 reduction.
In the synthesis of a typical catalyst, FePc and graphene were grinded in a mortar in a 1:1 mass ratio.
The resulting powder was dispersed in a mixture of 30% IPA in water at a concentration of 4 mg/ml.
The of ionic liquid was added to the mixture and ultrasonicated for 10 minutes using a tip sonicator.
Unless otherwise stated, 20%w of the ionic liquid was added with respect to the pristine catalyst.The resulting solution was finally mixed for 48 hours to allow full contact between the catalyst and ionic liquid.The pristine catalyst was subject to the same treatment, to account for any modifications that might result from sonication and stirring.
The resulting solution was freeze dried, to obtain the catalyst powder, which is referred to as FePC/G + IL.

Electrochemical testing
Catalyst inks at a concentration of 2mgFePC/GmL -1 were prepared by dispersing 2mg of pristine catalyst in 1 ml of 30%IPA in water with 18μL of Nafion, followed by 20 minutes of bath sonication and 10 minutes of probe ultrasonication.
5μL of the catalyst ink was spin-coated on a rotating disc electrode (3mm radius, RDE, Metrohm, previously polished with a micropolish cloth and 0.05μm alumina suspension), resulting in a loading of 0.14 mg/cm 2 (based on the mass of the FePC/G catalyst).The electrode was left to dry for 2 hours before testing.
Electrochemical measurements were performed in a 3-electrode electrochemical cell, featuring a rotating disk working electrode, a graphite rod counter electrode and a Hg/HgO reference electrode.
The reference electrode was calibrated regularly and all the potentials in this paper are reported vs RHE.The ohmic drop was determined after every LSV experiment, by fitting the high frequency semicircle obtained from impedance spectroscopy.The ohmic drop was found to be 13Ω ca. for all the samples tested.All the reported data have been iR corrected.
In a typical experiment, the electrolyte (0.1M KOH, Suprapur®, Merch) was saturated with oxygen (³99.9998%Ultrapure Plus, Air Products) and linear sweep voltammograms (LSV) were collected at a scan rate of 10 mVs -1 and rotational speed ranging from 400 to 2400rpm.Cyclic voltammograms (CV) were collected in static, N2-saturated (³99.99998%BIP® Plus, Air Products) 0.1M KOH at a scan rate of 50mVs -1 .Finally, square-wave voltammograms (SWV) were collected under the same conditions, using a potential step of 4mv, modulation amplitude of 20mV and a frequency of 2Hz, resulting in a scan-rate of 8mVs -1 .

Oxygen transport -gravimetric
Oxygen solubility with gravimetric measurements was determined using a gravimetric gas sorption analyser (Isochema IGA-002).This instruments features an ultrasensitive microbalance, with a capacity of 1g, resolution of 0.1μg and ±1μg and allows the precise control of pressure, from submillibar to 20bar.
Experiments were typically performed at 1, 2 and 3 bar and all the absorption measurements in this work were performed in dynamic mode, in order to ensure an exhaustive control of the set-point pressure.The temperature was maintained at 25°C and regulated by an external water bath.
In a typical experiment, around 150mg of the ionic liquid was loaded in the machine and degassed at 50°C for 24h under vacuum (10 -7 bar).Temperature was therefore reduced to 25°C and the weight was recorded.Oxygen was introduced, up to a set pressure, under constant temperature and the sample weight was monitored.A minimum equilibration time of 2 hours was utilized, and the samples were typically equilibrating in 3 to 4 hours.Desorption experiments were performed in a similar manner.
All the collected data were subsequently buoyancy-corrected.Firstly, an experiment was performed with the empty sample holder (SH), in order to determine the mass and volume of the sample holder (  ,   ).In absence of a sample, the mass read ( , ) is equal to the real mass of the sample holder (which does not change with pressure) and the buoyancy effect (which is proportional to the density of oxygen and the volume of the sample holder).Knowing how the gas density () changes with pressure (data obtained from NIST 64 ), the volume of the sample holder is calculated.Similarly, from the solubility experiments, the mass of the sample holder and sample ( + ) can be calculated from the balance reading (  ) by subtracting the effect of buoyancy, knowing the volume of the sample holder and that of the sample (obtained from the calculated mass and the known density).
, =   −     =  + −  (  +   ) For both the gravimetric and volumetric methods, the diffusion coefficient is calculated by fitting the time-dependent data to the solution of Fick's second law, as shown in Equation 6.
Where CS and CO are the saturation and initial concentration respectively, t is time and L is the thickness of the ionic liquid tested.The diffusion coefficient, D, is determined by fitting the experimental data to this equation by non-linear regression.Although this equation contains an infinite summation, it has been shown that only a few terms are sufficient for an accurate calculation of D. 65 The first 10 terms have been used in the fitting.For both the case of the volumetric and gravimetric analysis, the sample thickness was calculated using the mass of the sample, the density of the ionic liquid and the diameter of the sample holder Oxygen transportvolumetric A micromeritics 3Flex proposity analyser was used to calculate oxygen transport parameters volumetrically.Using micromeritics VacPrep, 2mL of ionic liquids were degassed at 10 -6 bar, 50°C for 24 hours.The samples were consequently transferred to the porosity analyser, using a check seal, to avoid contact with atmosphere.
In a typical experiment, 2 mL of sample are used and the mass is accurately determined by weighting the sample before and after degassing.Oxygen is supplied at ambient temperature at a set-point pressure through a manifold, of known volume.Then the decrease in pressure in monitored, and from the steady-state value (  ), the amount of gas absorbed is determined ( , ), by the difference between the expected pressure and the observed one, according to Equation 7. To calculate the expected pressure, from the volume of amount of gas introduced in the same, it is necessary to know the free space, i.e. the space in the sample holder that is not occupied by the sample, accurately.The free space (  ) is calculated repeating the same experiment with argon as a absorptive gas, which is assumed not to be absorbed by ionic liquid, or only in negligible amounts.The free volume is therefore calculated starting from the known volume of gas introduced ( , ), at standard temperature (  ) and pressure (  ) the temperature of the sample (  ), according to Equation 8.

Figure 1 :
Figure 1: a) Structure and abbreviation of the ionic liquids used in this work.B) Cathodic linear sweep voltammograms of the FePC/G catalyst, modified with the addition of 20%w ionic liquids.Data were collected in oxygen saturated 0.1M KOH, with a scan rate of 10mVs-1 and rotational speed of 1200rpm, at a temperature of 25 O C. The pristine catalyst loading was maintained constant at 0.14 mg/cm 2 with an addition of 20%w ionic liquid.Before recording the oxygen reduction activity, the catalyst was pre-conditioned with 5 cyclic voltammograms in nitrogen-saturated electrolyte, in a potential range of 0.1V vs RHE to 1.0V vs RHE.The ORR activity of the pristine catalyst is comparable to that reported in literature, as shown in FigureS1e c) Background-subtracted square wave voltammogram of FePC/G modified by the addition of ionic liquids, zoom-in on the high potential peak.Obtained in N2-saturated 0.1M KOH, using a potential step of 4mv, modulation amplitude of 20mV and a frequency of 2Hz, resulting in a scan-rate of 8mVs -1 , at a temperature of 25 O C. The pristine catalyst loading was maintained constant at 0.14 mg/cm 2 with an addition of 20%w ionic liquid.Before recording the square-wave voltammogram, the catalyst was pre-conditioned with 5 cyclic voltammograms in nitrogen-saturated electrolyte, in a potential range of 0.1V vs RHE to 1.0V vs RHE.After the CVs, SWV measurements were repeated 3 times in the same potential range and no change was observed between the scans.The second scan is shown here d) Potential for 0.1mAcm-2 (in O2-saturated 0.1M KOH), as a function of the position of the square-voltammetry peak in N2 saturated 0.1M KOH.Data presented are the average of three independent measurements and the error represents the highest and lowest data measured e) Ex-situ XANES of FePC/G modified by the addition of 20%w of the ionic liquids EIM NTF2 and BMIM PF6, at the Fe K edge Figure 1: a) Structure and abbreviation of the ionic liquids used in this work.B) Cathodic linear sweep voltammograms of the FePC/G catalyst, modified with the addition of 20%w ionic liquids.Data were collected in oxygen saturated 0.1M KOH, with a scan rate of 10mVs-1 and rotational speed of 1200rpm, at a temperature of 25 O C. The pristine catalyst loading was maintained constant at 0.14 mg/cm 2 with an addition of 20%w ionic liquid.Before recording the oxygen reduction activity, the catalyst was pre-conditioned with 5 cyclic voltammograms in nitrogen-saturated electrolyte, in a potential range of 0.1V vs RHE to 1.0V vs RHE.The ORR activity of the pristine catalyst is comparable to that reported in literature, as shown in FigureS1e c) Background-subtracted square wave voltammogram of FePC/G modified by the addition of ionic liquids, zoom-in on the high potential peak.Obtained in N2-saturated 0.1M KOH, using a potential step of 4mv, modulation amplitude of 20mV and a frequency of 2Hz, resulting in a scan-rate of 8mVs -1 , at a temperature of 25 O C. The pristine catalyst loading was maintained constant at 0.14 mg/cm 2 with an addition of 20%w ionic liquid.Before recording the square-wave voltammogram, the catalyst was pre-conditioned with 5 cyclic voltammograms in nitrogen-saturated electrolyte, in a potential range of 0.1V vs RHE to 1.0V vs RHE.After the CVs, SWV measurements were repeated 3 times in the same potential range and no change was observed between the scans.The second scan is shown here d) Potential for 0.1mAcm-2 (in O2-saturated 0.1M KOH), as a function of the position of the square-voltammetry peak in N2 saturated 0.1M KOH.Data presented are the average of three independent measurements and the error represents the highest and lowest data measured e) Ex-situ XANES of FePC/G modified by the addition of 20%w of the ionic liquids EIM NTF2 and BMIM PF6, at the Fe K edge Figure 2c also shows a comparison between the permeability values calculated with these methods.

Figure 2 :
Figure 2: a) solubility, b) diffusion coefficientand c) the product of the two, of oxygen in the selected ionic liquids.In yellow are the results obtained with electrochemical, in blue volumetric and in light blue gravimetric techniques.In all the graphs, the value for 0.1M KOH, the most common electrolyte in alkaline conditions, is shown as a grey dashed line for comparison.

Figure 3 :
Figure 3: a), b) Schematic representation of the oxygen concentration profile (line in red) in a classic rotating disc electrode a) before and b) after the addition of an ionic liquid layer.c), d) Schematic representation of the ionic liquid thickness, estimated starting from the BET surface area for low IL loadings (c) and from the geometric area for high loadings (d) e) Diffusion-limited current density as a function of oxygen permeability in the ionic liquid/ aqueous electrolyte.The red data point refer to the experimental results obtained at 1600 rpm, while the grey line is the result of the transport model described above, with a 100nm-thick ionic liquid layer f) Ratio between the diffusion-limited current density in presence and absence of the ionic liquid, as a function of the rotational speed.The full line shows the theoretical data obtained with the equation described above and using the solubility and diffusivity data obtained using the electrochemical method.The thickness of the ionic liquid layer was adjusted to fit the experimental data (squared data points) and was found to be 100 nm.

Figure 4 :
Figure 4: Mass spectrometry (top) and thermal gravimetric measurements of, from left to right, the pristine catalyst, the bulk ionic liquid and the catalyst with a layer of EIM NTF2.To make the mass spectrometry data ore readable, only selected masses are shown

Figure 5 :
Figure5: a) N2-sorption isotherm and b) pore diameter distribution, normalized to the mass of the pristine catalyst.From grey to dark purple, the curves represent the FePC/G catalyst, modified with an incrementally increasing amount of the ionic liquid EIm NTf2 c) Surface area calculated by BET (normalized to the mass of the pristine catalyst), as a function of the loading of the ionic liquid EIM NTf2 d) Micropore volume calculated using t-plot analysis (normalized to the mass of the pristine catalyst), as a function of the loading of the ionic liquid EIM NTf2 e) Average pore diameter, for pores < 4 nm (green) and <18 nm (orange), as a function of the loading of the ionic liquid EIM NTf2 f) Experimental pore filling degree determined from the pore volume distribution, as a function of the theoretical pore filling degree, calculated from the mass of the ionic liquid added.g), h), i) l) Schematic of the ionic liquid distribution in pores: g) shows the pristine support, h) an ionic liquid ideally distributed to form a monolayer and the ionic liquid distribution that we reconstruct from the porosity analysis, for the case of i) EIM NTF2 and l) BMIM PF6.