Stabilization of p-GaAs electrode surfaces in organic solvent by bi-phenyl rings for spin dependent electron transfer studies

Biphenyl-4-thiol (B4t) functionalised p-GaAs photocathodes were compared to non-functionalised pGaAs ones in the presence of methyl viologen cation radical (MV) as a redox mediator. Cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) studies revealed that at bare pGaAs photocathodes the radical interacts strongly with the electrode surface, displaying unwanted surface states (SS) mediated electron transfer (ET) even in the dark. Functionalising the electrode with B4t passivated the interface against this unwanted process, while ensuring that the conduction band (CB) electrons still facilitated faradaic reduction of MV to MV with unhindered efficiency.


Introduction
GaAs semiconductors attract considerable interest in many fields, particularly in photovoltaics and in conversion of solar to chemical energy. [1][2][3][4][5][6][7] Previous research has often highlighted the chemical instability of GaAs to be one of the main problems in its adoption as an electrode material in electrochemistry. [8][9][10][11][12][13] This issue is especially prevalent in the case of p-GaAs, where the majority charge carriers are holes. p-doped GaAs (p-GaAs) is also reported as an interesting material for spin dependent electrochemical studies, 14,15 our field of interest.
In our approach, a spin-oriented electron transfer (ET) is induced from a p-GaAs semiconductor to a radical redox mediator polarized in solution by a magnetic field. The spin orientation of the electrons of the semiconductor occurs during the promotion of valence band (VB) electrons to the empty conducting band (CB) through optical pumping using a polarized light under weak illumination, 16 i.e. conditions for which a spin-dependent ET kinetics have been already observed. 14,15 In such a process, the spin polarization of the CB electrons must be conserved during the ET, so any process involving spin relaxation prior to the ET must be avoided.
The question of the ET pathways at semiconductor/solution interface has been discussed extensively in the literature. [17][18][19][20] We have recently characterised the photocathodic behaviour of p-GaAs towards 1,4-benzoquinone (BQ) and Tris(2,2'-bipyridine)ruthenium(II) hexafluorophosphate (Ru(bpy)3 2+ ) in acetonitrile under weak illumination conditions. 21 Our study has highlighted that in the presence of BQ the ET originates from the CB, whereas in the presence of Ru(bpy)3 2+ , the surface states (SS) of the semiconductor play an important role in the ET process.
We also recently reported on the spin dependent ET at chiral and non-chiral electrodes using methyl viologen cation radical (MV +• ) as redox mediator to explore chirality-induced spin selectivity (CISS). 16 In this work, the chiral electrodes were made of an α-helical polypeptide layer terminated with Ag nanoparticles deposited on an Au-covered GaN substrate, while the non-chiral ones were constituted by bare p-GaAs. Although the spin-dependent ET to MV +• was visible in both cases, the lower than expected effect on bare p-GaAs could be attributed to the presence of SS that caused partial spin relaxation, and also to the low semiconductor surface stability. Actually, both SS removal and electrode stability enhancement must be achieved to facilitate spin dependent studies in solution when using MV +• as redox mediator, which has so far been reported as the best probe for spindependent studies. 15 Previously GaAs/electrolyte interfaces have been stabilised by functionalising the semiconductor surface with self-assembled monolayers. [22][23][24][25][26][27][28] The surface grafting was also shown to modify the energetics and the electron transport properties of the interface. [29][30][31][32] Among the best characterised functionalisation methods is the use of aromatic rings in combination with thiol chemistry. [33][34][35][36][37][38][39][40][41] Here we report the effect of biphenyl-4-thiol (B4t) functionalisation on the surface stability of p-GaAs in acetonitrile under weak illumination when using MV +• as the redox mediator. We show that the functionalisation can remove the unwanted non-spin polarised ET processes present even in the dark, while maintaining the photo-induced ET from the CB. Also, we show that the B4t layer does not hinder the ET efficiency related to MV +• reduction.
The B4t modified p-GaAs electrode surfaces were characterised by X-Ray Photoelectron Spectroscopy (XPS) and Atomic Force Microscopy (AFM), while their electrochemical behaviour was qualified by cyclic voltammetry (CV) and Electrochemical Impedance Spectroscopy (EIS).
Single crystal p-GaAs with Arsenic rich [111]B orientation (2.9×10 17 cm -3 Zn-doping) was used as working electrode (WE). An ohmic contact was established to the back side by evaporating 200 nm gold after oxygen plasma etching the wafer from both sides. This contacting method has previously been found to be adequate for electrochemical experiments. 9,42,43 A copper wire was contacted to the gold layer with silver paste, and the contact was protected from the sample solution with silicone.

Methods
The general procedure for p-GaAs functionalisation with a thiol containing monolayers has been well documented. 33,38 Briefly, wafers were sonicated in acetone for 3 min and then rinsed with Milli-Q water (18.2 MOhm/cm). Then the wafers were etched in 6 M HCl for 60 s, rinsed with Milli-Q water and ethanol before dipping into a functionalisation solution (2 mM B4t in dry EtOH). The wafers were incubated at 50⁰C for 20 h inside a glove box after which they were sonicated for 30 s in absolute ethanol, rinsed with absolute ethanol and finally blown dry with nitrogen.

J o u r n a l P r e -p r o o f
The non-functionalised electrodes were prepared as described above concerning the functionalised electrodes up to the point of incubation in the B4t solution. All electrodes were stored in a glove box and only taken out in an assembled and sealed test tube.
The experiments were performed in a 3 mm by 3 mm ID square glass tube. Due to the limited size of the tube, a Pt wire tip and a Pt coil were used as pseudo-reference electrode (RE) and counter electrode (CE), respectively. The potential of the Pt tip RE was compared to that of an organic Ag/Ag + RE, where an Ag wire was immersed into an ACN solution containing 0.2 M of TBAP and 0.01 M of AgNO3. The Ag/Ag + RE was separated from the sample solution by a frit. All of the electrochemical data has been adjusted to the potential of the Ag/Ag + couple, as described in S1 (supplementary information).
A 830 nm laser diode was used as a light source, where the non-collimated light was guided to the electrode surface through a channel while protecting the electrochemical cell from the ambient light. A Gamry potentiostat (Interface 1000E) was used to record the CV and EIS data. The EIS was performed from 0.5 Hz to 100 kHz with 10 mV modulation of the electrode potential, recording 6 points per decade. Thanks to the low laser intensities used, all the EIS data were recorded under a steady state current, limited by the availability of the CB electrons.
A two compartment U-cell was used to generate MV +• by electrolysis from MV 2+ at a Pt electrode, where a frit separating the two sides of the cell prevented the diffusion of the counter electrode species to the sample solution. The final concentration of MV +• was either 1 or 2 mM in 0.2 M TBAP used as supporting electrolyte. Before starting the electrolysis, MV 2+ solutions were heated gently to promote the solubility of the mediator. The MV 2+ was converted to MV +• quantitatively. To ensure the complete electrolysis after depleting [MV 2+ ] close to zero, small amounts of MV 0 were generated in the U-cell to convert the residual amount of MV 2+ thanks to the comproportionation reaction described by equation 2 below. The final composition of the sample solution was checked with a 10 µm diameter Pt ultramicroelectrode (UME) to confirm that the electrolysis was complete (see S2, supplementary information).

Surface characterisation
Atomic Force Microscopy (AFM) measurements were recorded with a Park Systems model XE-100 atomic force microscope scanning a 1 µm by 1 µm surface area. Three samples were studied: 1) a native surface sonicated 3 min in acetone and rinsed with Milli-Q water and EtOH, 2) sonicated 3 min in acetone, rinsed with Milli-Q water, etched 60 sec in 6M HCl followed by Milli-Q and EtOH rinse, 3) as 2) but subsequently functionalised in a B4t solution followed by sonicating 30 s in EtOH and EtOH rinse. The AFM tip used was an OTESPA-R3 from Bruker with a spring constant and a resonance frequency of 26 N/m and 300 kHz, respectively. X-Ray Photoelectron Spectroscopy (XPS) measurements were carried out using a PHI VersaProbe II scanning XPS microprobe (Physical Instruments AG, Germany). Analyses were performed using a monochromatic Al Kα X-ray source at 24.8 W power with a beam size of 100 µm. The pass energy was 20 eV yielding a full width at half maximum of 0.70 eV for the Ag 3d 5/2 peak. The spectra were corrected for C-1s (C-C) peak at 285.0 eV. The peaks were fitted with a Voigt profile. 38,44 Either a Shirley or a Tougaard type background was subtracted from the raw data, depending on the type of the background and the chemical specie under investigation. Three samples identical to those listed for AFM were studied. Focusing on As 3d spectra, Figure 1 (i) reveals the disappearance of a broad band situated around 44 and 45 eV for the HCl etched and functionalised surfaces. These two signals losses, attributed to As2O3 and As2O5, 44 are in agreement with the removal of the oxide layer from the GaAs surface. A characteristic feature of the HCl etched surface is a doublet for high elemental Arsenic (As 0 ) contents centred around 42.1 eV, whereas the functionalised surface displays a doublet for stoichiometric lattice Arsenic (As-Ga) centred around 41.4 eV. 33 In Figure a (ii) the functionalised surface shows the doublet attributable to the As-Ga as the main feature, and a residual As-O signal centred at 44 eV. A tentative assignment of Arsenic covalently bound to Sulphur (As-S) is visible at 42.8 eV, although strongly overlapped by the As-Ga. The As 2p3(2 core-level spectra were also recorded due to their higher surface sensitivity compared to As 3d, and the results are shown in Figure 1b (i). The main feature for the non-treated surface is the strong As-O signal at 1326.2 eV, which is greatly attenuated for the HCl etched and functionalised surfaces. Likewise to the 3d spectra, the HCl etching results to the strong presence of As 0 at the surface at 1323.7 eV. 45 The deconvolution of the XPS spectrum for the functionalised surface in Figure 1b (ii) reveals a residual As-O signal at 1326.2 eV and a lattice As-Ga signal at 1323.2 eV. Most importantly, and in agreement with previous study on thiol functionalised p-GaAs 111B surface, a clear signal for As-S can be observed at 1324.8 eV, confirming the successful attachment of the B4t to the electrode surface. 38 The XPS studies also showed that the B4t modified GaAs surface was resistant to any reoxidation within a time scale of days, even when the electrodes were left to an ambient atmosphere.

AFM characterisation
The surface topography was characterised by AFM on a 1 µm 2 electrode surface area. The surface morphology is shown in Figures 2a(i) to 2c(i), and the detailed surface topography in Figures 2a(ii) to 2c(ii) for native, HCl etched and B4t functionalised samples, respectively. All of the surfaces show general amorphous features, the grain size seemingly growing from native to B4t functionalised.
The untreated p-GaAs surface had a typical Root Means Square roughness (Rq) of 3.3 (±0.4) Å, in accordance with previous observations for As rich 111B surface ( Fig. 2 a(i)). 37 The HCl etching prior to the functionalisation did not change the surface appreciably, providing a surface with Rq of 3.5 (±0.5) Å. The most significant change was observed for B4t functionalised surface with Rq of 8.8 (±3.3) Å.
Regarding the surface topography, HCl etching had only limited effects towards [111]B surface (2a(ii) vs 2b(ii)), indicating that the greater surface roughness observed for the B4t functionalized sample (2c(ii)) is the result of the biphenyl layer and not due to etching induced surface roughening. Together with XPS measurements, this strongly suggests that the p-GaAs surface is covered with a B4t layer, covalently bonded to the As species via As-S bonds.
J o u r n a l P r e -p r o o f

General features
In acetonitrile methyl viologen di-cation MV 2+ can go through two consecutive reduction processes, each involving one electron. The first reduction leads to a cation radical MV +• and the second one to a neutral specie MV 0 according to equation 1: Both processes are close to the reversibility with a peak to peak potential separation (ΔEp) close to 70 mV at a 3 mm diameter Au electrode (Supplementary S1). The electrochemical process described by equation 1 is affected by a comproportionation / disproportionation reaction depicted by equation 2: The equilibrium constant for this reaction in acetonitrile has been reported to be around 1 × 10 7 , meaning that the disproportionation of MV +• does not occur appreciably in our conditions. 46 In addition, the dimerization of MV +• leading to a radical spin-paired cation and/or the association of J o u r n a l P r e -p r o o f MV +• with MV 0 has been observed in aqueous solutions, but are absent in organic solvents such as propylene carbonate and acetonitrile. 47,48 Finally, MV +• is also sensitive to O2(g) and UV light, but working in an electrochemical cell prepared in a glove box and subsequently sealed has provided MV +• solutions stable over time.
It should be highlighted that the electrochemical studies described below were conducted with a radical MV +• as the bulk redox mediator obtained by the electrolysis of a solution of MV 2+ at 1 or 2 mM (see experimental part). This is contrary to typical electrochemical studies concerning methyl viologen, where the starting species is the MV 2+ di-cation, and highlights our need for a radical species in the bulk solution for the spin dependent charge transfer studies. 16

Cyclic voltammetry
The qualitative behaviour of a p-GaAs electrode in a 1 mM MV +• solution is shown in Figure 3a. The first redox process related to the conversion of MV 2+ into MV +• peaking at around -0.9 V occurs in the dark and is almost completely independent of the light intensity ( Fig. 3a black CV). This suggests a valence band (VB) and/or SS mediated charge transfer of holes, as the conduction band for a p-type semiconductor is empty in the dark. [49][50][51] To confirm that the observed dark process originates from VB and involves SS, electrochemical studies were performed also on a MV 2+ solution (S3). The absence of the dark process with MV 2+ confirmed that the source of this phenomenon is the interaction of MV +• with the p-GaAs surface, likely generating SS energetically at or close to the formal potential of the MV +•/2+ .
A strong laser illumination (260 µW) introduces the second redox process, related to the conversion of MV +• into MV 0 , peaking at around -1.2 V and arising from the transfer of the CB electrons of the semi-conductor.
Another interesting feature in Figure 3a is the shape of the oxidation peak corresponding to the MV +• /MV 0 redox couple during the backward scan which is presumably composed of two peaks. One can notice the inversion of the peak potentials for the MV +• /MV 0 redox couple, the oxidation peak being situated at a more negative potential than the reduction peak during the forward scan. This could be attributed to a VB mediated oxidation, where the energetic overlap between the VB edge or SS and the density of states of MV 0 species at the interface determines the oxidation process, not the energy of the Fermi-level which is continuously changed during the potential sweep.  Figure 3b shows the quantitative current response of a bare p-GaAs electrode towards low laser intensities. The dark process corresponding to the MV +• /MV 2+ redox couple at around -0.8 V was again present and largely independent of the laser intensity. The current that belongs to the second redox couple MV 0 /MV +• exhibited a steady state profile, and the current of this plateau increased linearly with the laser power when measured at the potential of -1.32 V (Fig. 3b inset). At low laser intensity, there were no significant changes in the plateau current while changing the CV's scan rate (S4), thus confirming that the current was limited by the availability of the CB electrons and not by the diffusion of MV +• to the electrode surface. 13 On the return scan the anodic current peak around -0.85 V can be assigned to the oxidation of MV 0 , the intensity of which was proportional to the amount of MV 0 generated during the forward scan and thus to the laser intensity. Figure 3c shows the behaviour of a p-GaAs surface functionalised with B4t (black and green CVs), and its comparison with a non-functionalised one (dashed black line). The dark process that was observed for the non-functionalised p-GaAs has vanished completely, suggesting that the SS mediated hole transfer has been suppressed and that the surface is now stable towards MV +• . The effect of the laser J o u r n a l P r e -p r o o f irradiation on the functionalised electrode is described by the green CV, which shows that the photogenerated CB electrons are still transferred to the MV +• radicals, thus attesting that the photo induced ET was not hindered by the surface modification. Figure 3d shows the dependency of the photo current on the laser intensity for B4t functionalised p-GaAs. As for the bare surface, the limiting photocurrent at -1.32 V still evolved linearly with the laser power (Fig. 3d inset). When compared to a non-functionalised electrode (Fig. 3b), one can observe the suppression of the dark process related to MV +• /MV 2+ but also the disappearance of the oxidation of MV 0 during the return scan at around -0.85 V whatever the applied laser power. The removal of this latest feature is most likely due to the suppression of the SS at the interface by the B4t layer, thus preventing any SS mediated charge transfer between VB and MV 0 .

J o u r n a l P r e -p r o o f
When normalising the plateau currents in Figures 3b and 3d also by the laser power incident to the electrode surface (data not shown, see [21] for the measurement method), the quantum yields were found to be 41 and 37 (±3) % for the functionalised and non-functionalised electrodes, respectively. Also, the slopes of the plateau currents vs the laser power plots (Figs. 3b and d insets) were found to be -1.67 and -2.77 for the non-functionalised and the functionalised electrodes, respectively, suggesting that the functionalised surface was slightly more sensitive to the laser irradiation.
Most importantly, the bi-phenyl monolayer does not seem to passivate the p-GaAs towards ET, which has been the case in aqueous solvents. 27,33,34,52 Thus, while clearly attenuating the unwanted VB/SS charge transfer, the B4t monolayer maintains the facility of the charge transfer between the p-GaAs and MV +• , as could be expected from π-electron containing biphenyl rings. 40 Figure 3a. The oxidation of MV +• can readily occur through an electron transfer to the VB, but the reduction of MV 2+ without the presence of photoelectrons remains problematic. Tentatively the SS of p-GaAs could equilibrate with the E1 0 ' of MV +•/2+ couple and thus facilitate an electron transfer to the MV 2+ species. 56 Therefore we have added the SS to in the vicinity of the E1 0 ' to the proposed diagram in 3a, although we stress that the dark process is specific to the use of MV +• as a starting species, and was not observed when using MV 2+ as a bulk redox mediator (S3).
As for the observed photoreduction of the MV 0/+• couple, the CB edge is 0.35 V above the E2 0 '. This indicates that a wide range of reorganisation energies could bring the density of states of the oxidized species of the MV 0/+• couple to overlap with the CB edge, thus facilitating the second reduction process. When considering the oxidation processes for both couples during the return scan (Fig. 3a), we assume a VB process, as the densities of states of the MV 0 do not overlap with the CB edge. This notion is supported by previous work conducted with n-GaAs in water/EtOH (1:1) mixture in the dark, where during a forward scan two reduction peaks were observed for MV 2+ , whereas in the return scan no oxidative processes were observed, as the holes were absent from the VB. 57 This is contrary to the observations made in aqueous solutions at pH 2, where the oxidation of MV 0 has been observed to occur via an electron injection to the CB of n-GaAs. 58 For the functionalised surface in Figure 4b the interfacial processes are simplified. The VB edge is now situated at -0.27 V vs Ag/Ag + , placing the CB edge to -1.69 V. Exactly the same Fermi-level pinning was observed as for the bare surface however (-0.82 V), suggesting a band bending of 0.46 V at the equilibrium. As shown in Figure 3, the functionalisation of the p-GaAs surface with a B4t monolayer removes the dark process occurring from the VB completely, and only the reduction initiated from the CB by photoelectrons takes place. This is clearly attributable to the effect of the B4t monolayer blocking the interaction between the MV +• and the electrode surface. Thus, we have not drawn any SS contribution to the interface, as the EIS data presented below effectively suggested their absence from the interface.

Electrochemical Impedance Spectroscopy
EIS was used to characterise the B4t-modified GaAs electrodes. The equivalent circuit used to fit the impedance data is presented in Figure 5a. Similar circuit has been suggested before for 4,4'-thio-bisbenzene-thiolate and biphenyl 4,4′ -dithiol functionalised p-GaAs in H2SO4 solutions, 37,38 but in our case the SS and adsorption related components were removed, as their incorporation did not result to further increase in the fit quality. In addition to the solution resistance (Rsoln), the equivalent circuit comprises two sub-circuits. The first sub-circuit is made of a parallel combination of a space charge layer capacitance (CSC) with a charge transfer resistance (RCT) and Warburg impedance (W). The second sub-circuit, in series with the first sub-circuit, is constituted by a monolayer capacitance (CB4t) and resistance (RB4t).
To improve the fit, constant phase elements (CPE) were employed instead of ideal capacitors to model any inhomogeneities of the interface. The impedance of a CPE can be defined as ZCPE = Y0(jω) -α where α is the frequency component, 0 indicating a pure resistance and 1 pure capacitance. 35,59 It was also necessary to introduce a Warburg element (W) to fit the spectra at low frequencies, most likely due to the diffusion of ions within the B4t 41,60 which could indicate the existence of pinholes within the layer, or some π-π interactions between MV +• and the biphenyl layer. The inclusion of the Warburg element clearly acts as an approximation, as this element is generally used to represent a linear diffusion across a semi-infinite space. 60 Figure 5b shows the EIS spectra of the functionalised electrode stepped to a potential of -1.32 V (i.e. at the limiting current, see Fig. 3d) in the dark and when illuminated with a laser at 9 and 22 µW powers for a 2 mM MV +• solution. The solid lines are the corresponding EIS fits utilising the circuit depicted in Figure 5a. Most importantly, the possibility to fit the EIS data without incorporating the SS effects confirms that their influence is negligible for the B4t modified electrodes. This has already been observed for functionalised GaAs, namely that the SS capacitance (CSS) is not detectable in EIS spectra if CSS / CSC < 0.1, 35,52 and the same assumption was used also here to simplify the fitting and analysis of the EIS data. 61,62 The missing influence of the SS in the EIS data correlates with the absence of the oxidative peaks in the return scans for the CVs displayed in Figure 3d. This finding further confirms that the oxidation of MV 0 on non-functionalised electrodes involves the SS, a process which is clearly hindered for the functionalised surfaces.
As the frequency ranges for the CB4t and the CSC overlapped during an EIS measurement, it was difficult to distinguish between the different capacitance contributions. Indeed, it was possible to fit the EIS data recorded at -1.32 V with a simple Randles circuit without losing any quality of the fit while obtaining results within a couple of % points from those with Eq. 3. We have attempted to deconvolute the EIS data to individual CB4t, CSC, RB4t and RCT values in S6.
The CTOT value for the B4t coated p-GaAs has been measured before to be 0.45 µF/cm 2 . 34 Although performed in aqueous solvent, this result is similar order of magnitude than our results in Fig. 6a, which shows the decreasing CTOT with increasing bias. The α value of the B4t CPE was above 0.95 for all the measurements, indicating that the B4t SAM behaved almost as a pure capacitance. Figure 6b shows the total resistance RTOT of the interface as a function of the applied potential and the light intensity. As observed, RTOT is strongly influenced by the applied potential and the laser intensity, the dominant component being the RCT rather than RB4t (S6). The increasing laser intensity diminishes the overall resistance, as the current increases and the rate of Faradaic reduction increases. The fact that the RB4t contributes only slightly towards the overall charge transfer resistance is promising for spin dependent experiments, as the monolayer is not expected to introduce significant hindrance towards the rate of ET and thus towards unwanted relaxation of the spin polarisation.
As a comparison, systematic EIS studies performed on non-functionalised p-GaAs in MV +• solutions turned out to be unsuccessful, partly due to the instability of the p-GaAs in presence of the radical species, and partly due to the complicated dark process that could not be satisfactorily fitted to any feasible equivalent circuit.

Conclusions
Functionalising p-GaAs with B4t monolayer was shown to stabilise the semiconductor surface against MV +• radical, removing the SS mediated unwanted MV +• /MV 2+ process that was observed even in the dark on bare p-GaAs. The EIS analyses of the functionalised surface revealed that a simple space charge-monolayer model fitted the data adequately, and the addition of more fitting parameters was not needed. This suggests that the influence of the SS at the interface was attenuated to a lower level, i.e. not detectable by electrochemical methods.
This situation is ideal for future studies on spin dependent processes at p-GaAs under optical excitation, as radicals required in the sample solution will not destabilise the electrode, and the absence of SS at the interface should lead to lesser degree of SS mediated relaxation of electrons from the CB to VB.
The passivation of the electrode surface against unwanted processes did not lead to a lower Faradaic yield, displaying the ability of the biphenyl monolayer to facilitate the electron transfer from the CB of the p-GaAs to the radical in the solution. This is contrary to previous studies in aqueous solution, where up to 50 fold increase in the interfacial resistance after functionalisation has been observed. 33,34 However, the stability of the interface was far from that observed in aqueous solutions (up to 22h) most likely due to some π-π interactions between MV +• and the biphenyl layer, providing a stable photo-response up to a couple of hours only, depending on the magnitude of the photocurrents generated.

Author Statement
The authors declare no competing financial or personal interests.

Declaration of interest
The authors declare no competing financial or personal interests.