Molecular scale structure and dynamics at an ionic liquid/electrode interface

After a century of research, the potential-dependent ion distribution at electrode/ electrolyte interfaces is still under debate. In particular for solvent-free electrolytes such as room-temperature ionic liquids, classical theories for the electrical double layer are not applicable. Using a combination of in situ high-energy X-ray re ﬂ ectivity and impedance spectroscopy measurements, we determined this distribution with sub-molecular resolution. We ﬁ nd oscillatory charge density pro ﬁ les consisting of alternating anion- and cation-enriched layers at both cathodic and anodic potentials. This structure is shown to arise from the same ion – ion correlations dominating the liquid bulk structure. The relaxation dynamics of the interfacial structure upon charging/ discharging were studied by impedance spectroscopy and time resolved X-ray re ﬂ ectivity experiments with sub-millisecond resolution. The analysis revealed three relaxation processes of vastly di ﬀ erent characteristic time scales: a 2 ms scale interface-normal ion transport, a 100 ms scale molecular reorientation, and a minute scale lateral ordering within the ﬁ rst layer.


Introduction
In the early 20th century Gouy solutions near a charged electrode. [1][2][3] However, room-temperature ionic liquids (ILs) consist solely of ions. They are intensively studied as future environmentallyfriendly working uids in applications ranging from catalysis to solar cells and supercapacitors. 4 Therefore, the diluted solution approximation is clearly invalid. [5][6][7] Detailed understanding of the function and performance optimization of such devices requires a molecular-resolution knowledge of the electrode's interfacial structure and its dynamics during the charging/discharging processes. 8 Therefore, a variety of experimental, theoretical, and computational techniques have been employed to shed light on the structure and dynamics of ILs near interfaces and in connement. 6 For ILs composed of cations with short alkyl side chains, molecular dynamics (MD) simulations suggest an interfacial prole comprising alternating cationand anion-enriched layers. 9 This leads to an oscillatory interfacial concentration prole decaying gradually into the uniform bulk composition. Such proles deviate signicantly from the exponentially decaying concentration prole of a diffuse electric double layer predicted by the classical Gouy-Chapman theory. Furthermore, they cannot be described by the approaches developed for highly concentrated electrolyte solutions and molten salts, taking into account the nite size of ions. [10][11][12] It has been suggested that the observed proles are a consequence of asymmetric ions 13,14 and strong ion-ion correlations in the absence of solvent molecules. 15,16 Different experimental techniques have been used to study the molecular-scale structure of ILs at solid/liquid interfaces. Interfacial layering was observed in atomic force microscopy (AFM) [17][18][19][20] and X-ray reectivity (XRR) measurements [21][22][23][24] for several ILs on various substrates. In thin IL lms, long range ordered structures have been found by helium atom scattering. 25 Starting in 2010, initial attempts have been made to investigate the response of the interfacial structure to electrode potentials by XRR 26 and neutron reectivity. 27 However, in these early studies substrate reconstruction on gold surfaces 28 and a limited q-range in neutron reectivity rendered the extraction of the molecular-scale ion structure near the interface highly ambiguous. More recently, different groups have reported synchrotron XRR studies at IL/electrode interfaces under controlled electric potentials. 29,30 In all of these studies, distinct changes in the interfacial ion distribution were found upon potential variation. Time resolved experiments, covering the relaxation dynamics on the seconds to minute scale, indicated the presence of ultraslow interfacial processes. 31,32 Scanning tunneling microscopy and AFM studies indicate that the substrate-adsorbed cation layer is affected by an applied potential. 20,28,[33][34][35] Sum frequency generation (SFG) spectroscopy 36 detected molecular reorientations upon variation of the applied potential. Impedance spectroscopy (IS) studies showed that the interfacial dynamics are governed by at least 3 relaxation processes on time scales ranging from milliseconds to minutes. 18,20,28 However, based only on these electrochemical studies, it is not possible to unambiguously assign the observed processes to specic spatial rearrangement of ions near the solid/liquid interface.
Here, we present an in situ study of the structure and dynamics of the IL 1butyl-1-methylpyrrolidinium tris(pentauoroethyl)triuorophosphate 37 [bmpy] + [FAP] À at an inert electrode during the charging/discharging process. Using high-energy XRR, we determined interface-normal ion proles with molecular-scale resolution. Comparison with bulk X-ray scattering revealed the origin of the observed spatial ion distribution. Its temporal response to applied potentials was determined by time-resolved XRR and electrochemical IS experiments. This combined approach enabled us for the rst time to directly study the structural response of an electrolyte at an electrode on the millisecond to minute time scale.

Materials
The IL 1-butyl-1-methylpyrrolidinium tris(pentauoroethyl)triuorophosphate [bmpy] + [FAP] À was obtained from Merck in high-purity grade. Pure [bmpy] + [FAP] À can be supercooled well below its melting point of 4 C, allowing experiments at À12 C. At this temperature, a liquid mass density of r m ¼ 1.62 g cm À3 was determined by pycnometry. 38 To remove moisture and volatile residues, the IL was kept in a vacuum oven (1 mbar, 90 C) overnight prior to the experiments.

In situ setup
For in situ investigation of the IL/electrode interfaces, a new experimental cell for simultaneous XRR and electrochemical measurements was developed (Fig. 1). The setup is inspired by concepts employed in our previous high energy X-ray reectivity studies on deeply buried interfaces, 21,23,39 and recent developments in the eld of electrochemical in situ X-ray scattering techniques. 40,41 For visual sample inspection during alignment and measurements in a vacuum or inert Fig. 1 Sketch of the setup for in situ XRR experiments (total height 165 mm). The sample (red circle and inset) is contained in a gas tight cell (light blue). The central glass tube allows visual inspection during alignment and measurements in a vacuum or inert atmosphere. Kapton windows (orange) for the incident and reflected X-ray beam; sample post (Cu, brown) with bore hole for the cooling fluid (isopropanol, dark blue) of a closed cycle thermostat. The insets shows the BDD working electrode (top) and the IL reservoir (PTFE, bottom). Connection is made via a free standing meniscus by moving the overfull IL reservoir against the working electrode (arrows).
atmosphere, the center part of the cell is made of a glass tube (height 140 mm, outer diameter 160 mm, wall thickness 5.5 mm; Schott Duran). For connection with the surrounding stainless steel elements, the glass part is equipped with two DN150 at anges (DIN 12214) at the top and bottom. At the side, two KF50 anges (ISO 2861/1) for the X-ray windows (thickness 50 mm; Kapton) point in the opposite direction. The electrical potential across the solid/liquid interface ( Fig. 2) is controlled by a potentiostat in the three electrode conguration (PGSTAT302, Autolab). Using an inert boron doped diamond (BDD) working electrode (WE), we avoid surface reconstruction, which plagues metal electrode measurements. 28,42 The working electrode is a monocrystalline BDD plate (size 4 mm Â 4 mm Â 0.3 mm, boron concentration 10 19 cm À3 ; Element Six) in (100) orientation. A polycrystalline, highly boron doped diamond (electrochemistry grade, boron concentration > 10 20 cm À3 ; Element Six) and a ame annealed 2 mm diameter platinum wire served as the counter electrode and quasi reference electrode, respectively. The solid/liquid interface is formed by touching the BDD electrode to the meniscus of the IL, kept in a temperature controlled PTFE reservoir (Fig. S1 †). A 350 mm thick single crystalline corundum plate and a polyether ether ketone spacer electrically isolate the electrochemical cell from the supporting structure. The sample temperature was controlled by a closed cycle thermostat and monitored by two PT-100 sensors underneath the RE and above the WE, respectively. To reduce dris in the WE/IL interface position, the WE is mechanically decoupled from the lower copper sample post by an aluminum frame and a stainless steel rod.

Impedance spectroscopy (IS)
Impedance spectra for electrode potentials À2.5 V # +1.5 V were recorded in 0.25 V steps under a dry nitrogen atmosphere. For each potential, IS was measured for frequencies between 10 kHz $ u/2p $ 0.01 Hz at an excitation voltage of 10 mV. Fig. 2 The potential U between the working electrode (WE) and the quasi reference electrode (RE) is controlled by a potentiostat in a standard 3-electrode configuration with a counter electrode (CE). For XRR measurements with sub-millisecond time resolution, a periodic square wave potential U(t) was applied by an external function generator. A Picoharp device assigned a time stamp to each photon counting event recorded by the detector and the rising edge of the potential steps.

Faraday Discussions Paper
Faraday Discuss. This journal is © The Royal Society of Chemistry 2017

X-ray reectivity (XRR)
XRR experiments were performed at the high energy micro diffraction (HEMD) setup at ID15, ESRF (wavelength l ¼ 0.0178 nm). Details are described elsewhere. 38,39,43 Aer transfer to the measurement chamber, the sample was degassed and kept under helium. This protocol prevents water adsorption, reduces the background signal in XRR measurements, and improves the temperature homogeneity across the sample. X-rays enter and leave the IL through the free standing meniscus formed between the top edge of the PTFE reservoir and the WE. Specular XRR curves R(q) at xed potentials of +1.5 V, 0 V and À2.5 V were recorded by a single photon counting NaI scintillation detector (Cyberstar CBY48NA05B, Oxford Danfysik). Before each measurement, the interface was equilibrated by applying a 50 Hz oscillatory potential between À2.5 V and +1.5 V.
Aer several minutes the oscillation amplitude was decreased slowly. At the same time, the potential offset was gradually shied from À0.5 V to its nal value. The background, stemming mainly from bulk IL scattering, was collected at a AE0.1 offset in the incidence angle relative to the specular condition. Repeated scans demonstrate that reproducible XRR data up to a scattering angle 2q ¼ 1.8 can be attained while keeping the radiation damage of the sample at an acceptable level. The specular and background signals from multiple scans were averaged and interpolated to a regular grid in q ¼ 4p l sinðqÞ, where q is the X-rays' grazing angle of incidence with respect to the interface and 2q is the total scattering angle. Aer background subtraction, footprint corrections were applied to account for the nite beam and sample sizes.

Time resolved experiments
Time resolved XRR signals were recorded at a xed incidence angle q ¼ 0.35 , corresponding to a momentum transfer q ¼ 4.3 nm À1 . At this angle, excellent counting statistics were obtained. To study slow processes on the minute time scale, cyclic voltammetry (CV) between À2.5 V and +1.6 V was performed at a scanning speed of 10 mV s À1 . Fast processes were investigated by monitoring the relaxation of the XRR signal during periodic square wave potential cycles 44,45 ( Fig. 2). Alternating steps between À2.5 V and +1.6 V with 0.02 ms rise time were applied to the WE at a frequency of 50 Hz. Each single photon counting event, recorded by the X-ray detector, was logged by a Picoharp (PicoQuant). Submillisecond time resolution was achieved by calculating the histogram of counting events recorded over several minutes vs. delay with respect to the rising edge of the potential steps.

Bulk X-ray scattering
Bulk X-ray scattering was measured in the transmission geometry on a self constructed instrument using Cu K a radiation (l Ka ¼ 0.154 nm, Rigaku MicroMax 007 microfocus rotating anode X-ray generator, Osmic Max-Flux confocal multilayer optics). 46

X-ray reectivity
Interfacial proles. For quantitative interpretation of the XRR curves, the experimental data was analyzed by a modied distorted crystal model. [48][49][50] It is based on a parametrization for the total electron density prole r e (z) that was successfully used in previous XRR studies to analyze the interfacial structure of ILs. 21,23,38,51 Originally, the prole normal to the solid/liquid interface was composed of ionic contributions from cations (c), anions (a), and the working electrode r WE e . This model was extended to account for the excess charge of the surfaceadsorbed (ad) ion layer c ad , controlled by the applied potential. The partial electron densities r a,c are calculated from the composition of the respective ion species (eqn (2)). Z a,c is the number of electrons per anion/cation, r m the IL bulk mass density, M the molecular mass of the IL, and N A the Avogadro constant. Following Névot and Croce, the cumulative normal distribution function accounts for the surface roughness of the solid working electrode s WE . 52,53 First adsorbed layer. To account for adsorption and desorption of counter ions, the rst ion layer is modeled by a single slab.
The distance z 0 controls the separation of the adsorbed layer from the electrode. A full monolayer of adsorbed ions corresponds to an area density r ad d I . As motivated by the work of Fedorov et al., 5 the dimensionless parameter g limits the maximum local ion concentration within the adsorbed layer to c ad (z) < 1/g. The slab thickness Dz was set to obey charge neutrality. It is determined by the surface charge s(U) and calculated numerically by the condition where s ML ¼ AEed I r a,c /Z a,c is the charge equivalent of a monolayer of monovalent anions or cations with elementary charge e. The effective width s 0 0 is obtained from s 0 by comparison of eqn (4) with the Gaussian density distributions (eqn (7)), modeling the subsequent alternating anion and cation enriched layers that are equidistantly spaced by 1/2d I .

Faraday Discussions Paper
Faraday Discuss. This journal is © The Royal Society of Chemistry 2017 Eqn (6) is calculated from the constraint that at the surface charge s(U) ¼ 1/ 2s ML the area and maximum value of c ad (z) agree with the corresponding Gaussian density distribution for n ¼ 0.
Subsequent anion and cation layers. The remaining oscillatory ionic part of the interfacial prole composed of anion and cation enriched layers was parametrized via a binary DC model composed by a series of Gaussians representing the alternating cation and anion enriched layers. Subsequent Gaussians, centered at a distance z n from the electrode, are equidistantly spaced by 1/2d I (eqn (7c)). Their width s n is composed of the intrinsic sizes of anions and cations s a,c plus incremental broadening s b , and potential dependent broadening s p (U) (eqn (7d)). Mass conservation is taken into account by eqn (7b). XRR curves. Reectivity curves were numerically calculated using the Parratt formalism aer dividing the prole into 0.02 nm slabs of constant density. 53,54 Dispersion effects were included using X-ray form factors from the NIST database. 55 Density proles were extracted by simultaneous tting of all three XRR data sets using a simulated annealing algorithm. 56 For large separations z from the solid/liquid interface, the interfacial density prole in the IL r e (z) can be approximated 57-59 by the generic form Here, the oscillatory prole adjacent to the solid/liquid interface is solely characterized by its amplitude A, phase 4, periodicity d I , and decay length x I .
For sufficiently large z/d I , the proles calculated by eqn (7) exhibit the asymptotic behavior predicted by eqn (8). 38,60 Therefore, the effective parameters for the periodicity d I and correlation length x I of the oscillatory interfacial structures were numerically determined from the anion and cation contributions to the total electron density proles.

Impedance spectroscopy
For IS, a small sinusoidal potential variation with frequency u/(2p) is applied to a capacitor in equilibrium at a constant potential U. The measured complex capacitanceĈ is modeled by a sum of Cole-Cole expressions, each representing a relaxation process of strength DC j (U) and time constant s j . 61 Expression (9) includes diffusive processes such as electrode polarization as well as interfacial relaxations. 18,62 The exponent 0 # a j # 1 describes the deviation from an ideal Debye process (a j ¼ 1) and is related to the width of the relaxation time distribution around its mean value s j .

Surface charge
The impedance spectra (Fig. 3) reveal two distinct capacitive processes on different time scales s j and with different relaxation strengths DC j , each represented by a half circle in the complex capacitance plane (Fig. 4). At low frequencies, the onset of a third slow process is observed. With the total differential capacitance DC(U), the surface charge difference on the electrode Ds(U) can be obtained by numerical integration.
The surface charge difference relative to U 0 ¼ 0 V was calculated as Ds(+1.5 V) ¼ 1.8 mC cm À2 and Ds(À2.5 V) ¼ À1.9 mC cm À2 . This amounts to approx. AE10% Fig. 3 Imaginary part (red triangles) and real part (blue circles) of the complex differential capacitance at the potentials +1.5 V (top), 0 V (middle) and À2.5 V (bottom). Lines are fits to the Cole-Cole expression (eqn (9)). Curves are vertically shifted by 1 unit for clarity.

Ion prole
Ion distributions normal to the interface were studied by XRR at xed potentials. To highlight changes induced by rearrangement of cations and anions, the experimental XRR patterns R(q) were normalized by the Fresnel reectivity R F (q) of an ideally at and abrupt IL/BDD interface (Fig. 5b). The À2.5 V XRR curve shows a pronounced dip at q z 8 nm À1 , corresponding to a distance of 2p/q z 0.8 nm. Its position and width are close to those of the rst scattering peak of the bulk liquid (Fig. 5a). Quantitative analysis by Fourier transformation of the total structure function yield a bulk periodicity d B ¼ 0.80 nm and a bulk correlation Arrows point to frequencies corresponding to the relaxation times s 1 z 2 ms and s 2 z 120 ms. The inset shows the potential dependence of the differential capacitance C (red circles) and the relaxation time s 1 (blue squares) of the fast process. length x B ¼ 1.46 nm. 38 Bulk X-ray and neutron scattering measurements as well as MD simulations on similar ILs suggest that this periodicity corresponds to the average bulk separation between same-charge ions. 16,[63][64][65][66] Increasing the potential from À2.5 V to +1.5 V results in an almost structureless XRR pattern. Compared to XRR curves measured at IL/gold interfaces, 26 the signal modulations in this study are signicantly enhanced. This results from a better scattering contrast between diamond and the IL than that between Au and the IL, demonstrating the high sensitivity of our method for detecting potential-induced ion rearrangement near the electrode. For a quantitative interpretation, the measured XRR curves were tted by the modied distorted crystal model described in Section 3.1. All three measured XRR curves recorded at different potentials were tted simultaneously with the same values for all bulk-related parameters, the same surface roughness of the BDD working electrode, and the xed surface charge difference Ds(U), determined by IS. The XRR ts (lines in Fig. 5b) yield layered density proles for both the anion and cation (Fig. 6). The resultant effective interfacial layer periodicity d I ¼ 0.73 nm and the decay length x I ¼ 1.44 nm are in accordance with the XRR measurements on negatively charged sapphire substrates, 21,38 and AFM measurements that revealed a layer periodicity of 0.9 nm. 28 The good correspondence between d I and d B as well as x I and x B indicates that the interfacial structure is governed by the same ion-ion correlations dominating the bulk structure. 23,51,58,67 The ion concentrations in each layer (Fig. 7) were determined by integration of the interfacial model proles, derived from the XRR ts.
The interfacial proles agree qualitatively with the results from MD simulations and continuum theory at comparable surface charges. 9,15 In comparison with the parameters used in the continuum theory model, in our system the normalized bulk correlation length x B /d B is four times larger. This leads to stronger oscillations in the relative cation/anion concentrations. A cation excess was found in the substrate adsorbed layer at all three potentials. At +1.5 V, the surface charge, i.e. the sum over all layers, amounts to an equivalent of approx. 60% of the charge in a cation monolayer. Thus, the potential of zero charge must  occur at high anodic potentials, where a maximum in C(U) is predicted for symmetric ions. 12,15 This is in agreement with the observation of a monotonically increasing C(U) between a potential of À2.5 V and +1.5 V (Fig. 4 inset).
Our choice of a chemically inert but semiconducting BDD electrode material leads to an exponential charge carrier density prole within the working electrode. 68 This limits the surface charge difference between À2.5 V and +1.5 V to approx. 20% of a cation monolayer. Therefore, in our experiments neither the potential of zero charge followed by the exchange of cations with anions in the adsorbed layer, nor lattice saturation and the crowding regime, is reached. [69][70][71] However, the observed potential dependent XRR patterns are fully reversible and reproducible. In contrast to a recent study by the group of P. Dutta using silicon working electrodes, 32 no formation of extended, several nanometer thick interfacial adsorbed layers were observed in this study.
Two main features of ion density proles were found to vary signicantly with applied potential. The rst is the center of mass of the substrate-adsorbed layers. At +1.5 V the center is observed at 0.20 nm. For more negative potentials, the center is shiing away from the electrode surface. At 0 V we nd a separation of 0.23 nm and at À2.5 V a value of 0.28 nm. Similar shis in the substrate-adsorbed layer position were observed in AFM force-distance curves. 28,34,35 Moving away from the electrode, details related to the specic molecular organization of the cations in the adsorbed layer become indistinguishable. At these larger distances z, the oscillatory density proles are solely characterized by their effective amplitudes, phases 4, periodicities d I , and decay lengths x I (eqn (8)). The latter two are linked to their corresponding bulk values d B and x B , respectively. 38,59,72 Therefore, these two parameters are expected to remain unchanged upon potential variation. However, changes in the adsorbed layer result in phase shis of the IL's oscillatory prole relative to the solid electrode (eqn (8) and Fig. 6). The XRR signal R(q)/R F (q) originates from the interference of waves reected at gradients in the interfacial electron density proles. Therefore, XRR is highly sensitive to such phase shis. Demonstratively, phase shis in the oscillatory prole lead to different interference patterns with the waves reected from the BDD electrode having an electron density about twice the IL bulk value ( Fig. 6 and S2 †). These interference effects explain the strong variation of the XRR signal at different potentials despite the rather small changes in the charge concentrations in all layers, shown in Fig. 7.

Interfacial dynamics
The dynamics of ion rearrangement at the IL/electrode interface during the charging/discharging process were studied by combining the IS measurements ( Fig. 4) with two complementary time-resolved XRR measurements. The rst, yielding sub-millisecond time resolution, recorded the evolution of the XRR signal at a xed q following an abrupt positive/negative switching of the potential U(t) (Fig. 8b). The second, providing information on a longer time scale, recorded the evolution of the same xed-q XRR signal during a slow cyclic voltammetry (CV) scan U(t) (Fig. 8a). Thus, the structural rearrangement of cations and anions adjacent to the electrode during the charging/discharging process was investigated on time scales spanning several orders of magnitudes.
In the abrupt switching measurements, a periodic square wave potential with 4 V steps and 0.02 ms rise time were applied to the IL/electrode interface while recording XRR. The resultant XRR signal shows a small but signicant modulation with a relative amplitude of approx. 2%. Relaxation times were determined by tting the measured XRR with a sum of two decaying exponentials and a constant baseline.
To extract the long relaxation time T 2 of the slow component from 50 Hz cycles, R 0 ¼ R(Dt / N) was xed to the values determined from the static XRR curves at q ¼ 4.3 nm À1 (Fig. 5b).
The t (Fig. 8b) yields T 1 ¼ 2 ms in good agreement with the relaxation time s 1 z 2 ms of the rst, fast, process observed above by IS. The ts by eqn (9) to the IS

Faraday Discussions Paper
Faraday Discuss. This journal is © The Royal Society of Chemistry 2017 data yield a value for a 1 of 0.93. This is close to unity as expected for diffusion driven electrode polarization. 62 Furthermore, on Au electrodes, a Vogel-Fulcherlike temperature dependence was found for this relaxation process. 18 Such dependence is characteristic of the bulk ion conductivity, suggesting that the fast relaxation process is connected to ion transport from/to the interface, limited by the ion conductivity. Indeed, its capacitive strength, DC 1 z 0.6 mF cm À2 , dominates the total capacitance found by IS (Fig. 4) and supports this interpretation. The magnitude R 2 of the intensity modulations associated with the second exponential (dashed lines in Fig. 8b) is more than an order of magnitude larger than R 1 . Our model ts of the XRR curves, recorded at different static potentials, show that these large modulations primarily originate from shis of the rst cation layer normal to the electrode surface. SFG spectroscopy suggests such shis to result from potential-dependent reorientations of the asymmetric [bmpy] + substrate-adsorbed cations. 36 The corresponding relaxation time T 2 ¼ 50 ms is less than half of s 2 z 120 ms of the second relaxation process observed by IS. However, the time scales are of the same order of magnitude. The difference may arise from the specic experimental conditions. In IS an equilibrated system is probed by small perturbations. In contrast, for the fast XRR measurements potential steps of 4 V are applied to the IL/electrode interface. This leads to a highly non-equilibrium ion conguration and a relaxation pathway with a faster time constant T 2 . Apparently, this reorientation is governed by specic ionelectrode interactions and happens on much longer time scales T 2 than the ion transport. The broad relaxation time distribution obtained from IS with a 2 z 0.6 may reect electrode surface inhomogeneities. The Arrhenius-like temperature dependence of s 2 found on gold electrodes 18 supports our assignment of this process to molecular reorientation within the rst adsorbed cation layer. Finally, note that compared to the fast ion transport process, the slower reorientation process has only a small capacitive strength DC 2 of approx. 0.2 mF cm À2 . This may arise from the relaxation of the rst cation layer's distance from the electrode surface, as well as the adsorption of additional cations on vacancies formed aer reorientation.
In the low frequency regime, i.e. on the time scale above 10 s, the IS data (Fig. 4) indicates the onset of a third, very slow, process. This agrees with the existence of a hysteresis loop in the XRR signal recorded during CV (Fig. 8a). The presence of such a loop conrms the occurrence of structural rearrangements on a time scale over which a signicant potential variation is affected in a CV scan, i.e. 10-100 s. Such slow dynamics could be caused by a lateral reorganization and eventually 2D ordering of the rst layer of interface-adsorbed cations, as observed in scanning tunneling microscopy 33 that also shows very slow dynamics upon potential variation. 73 Likewise, in XRR studies on electried IL/graphene interfaces, relaxation times on the 10 s scale have been observed. 29,31 Here, the slow dynamics was attributed to the presence of a bistable system with a large energy barrier of approx. 9k B T. The stable states are represented by the structures having either an anion or cation layer adsorbed on the electrode. However, our combined IS and XRR study shows that the ion transport to and from the interface occurs on a much faster millisecond time scale.

Summary and conclusions
In conclusion, combining electrochemical and time-resolved XRR and IS measurements, we presented a comprehensive picture of the molecular-scale structure of an IL/electrode interface. Its response to applied potential comprises multiple time scales ranging from a few milliseconds to hundreds of seconds. At all investigated potentials the interface-normal ion concentration proles exhibit a distinct layering structure. The measured XRR curves are reproduced by a single layering periodicity and a single decay length which are independent of the applied potential. They are close to those of the bulk correlation, implying that bulk correlations dominate also the interfacial structure. The time-resolved measurements suggest a three-step structure-variation scenario for the charging/discharging process at an IL/electrode interface (Fig. 9). Specically, switching the voltage from À2.5 V to +1.5 V reduces the surface charge by approx. 20% of a monolayer-equivalent. The diffusion-limited ion transport from and to the interface happens on a millisecond time scale. In addition, a shi occurs in the rst cation layer's position relative to the electrode surface. This process exhibits a small capacitive strength and slow relaxation time on the order of 100 ms. We tentatively assign this process to a reorientation of substrate-adsorbed cations. Due to ion-electrode interactions, this reorientation process is strongly hindered and sensitive to electrode inhomogeneities. These inhomogeneities lead to a broad relaxation time distribution. Finally, based on observations by scanning probe techniques, 33,73 we suggest that the even slower third process, observed in CV on the 10-100 s time scale, is related to a lateral reorganization of substrate-adsorbed cations. However, at deeply-buried solid/ liquid interfaces such processes can not be probed directly by grazingincidence X-ray diffraction due to the intrinsically strong background originating from the IL bulk.