Reversible Formation of Gold Halides in Single‐Crystal Hybrid‐Perovskite/Au Interface upon Biasing and Effect on Electronic Carrier Injection

Solar cells, light emitting diodes, and X‐ray detectors based on perovskite materials often incorporate gold electrodes, either in direct or indirect contact with the perovskite compound. Chemical interactions between active layers and contacts deteriorate the operation and induce degradation, being the identification of the chemical nature of such interfacial structures an open question. Chemical reactivity of gold in contact with the perovskite semiconductor leads to reversible formation of oxidized gold halide species and explains the generation of halide vacancies in the vicinity of the interface. Electrical biasing induces contact reaction and produces modifications of the current level by favoring the ability of perovskite/Au interfaces to inject electronic carriers. The current injection increment does not depend on the halogen source used, either extrinsically by iodine vapor sublimation of Au electrodes, or intrinsically by bias‐driven migration of bromide ions. In addition, the formation of a dipole‐like structure at the reacted electrode that lowers the potential barrier for electronic carriers is confirmed. These findings highlight adequate selection of the external contacts and suggest the need for a deeper understanding of contact reactivity as it dominates the operation characteristics, rather than being governed by the bulk transport properties of the charge carriers, either electronic or ionic.


Solar cells, light emitting diodes, and X-ray detectors based on perovskite materials often incorporate gold electrodes, either in direct or indirect contact with the perovskite compound. Chemical interactions between active layers and contacts deteriorate the operation and induce degradation, being the identification of the chemical nature of such interfacial structures an open question. Chemical reactivity of gold in contact with the perovskite semiconductor leads to reversible formation of oxidized gold halide species and explains the generation of halide vacancies in the vicinity of the interface. Electrical biasing induces contact reaction and produces modifications of the current level by favoring the ability of perovskite/Au interfaces to inject electronic carriers. The current injection increment does not depend on the halogen source used, either extrinsically by iodine vapor sublimation of Au electrodes, or intrinsically by bias-driven migration of bromide ions.
In addition, the formation of a dipole-like structure at the reacted electrode that lowers the potential barrier for electronic carriers is confirmed. These findings highlight adequate selection of the external contacts and suggest the need for a deeper understanding of contact reactivity as it dominates the operation characteristics, rather than being governed by the bulk transport properties of the charge carriers, either electronic or ionic.
conversion in the near future. In addition to photovoltaic applications, perovskitebased materials are being explored as sensitive layers for high-energy radiation detectors and imaging devices for medical diagnostics. [2] These detection technologies rely upon the ability of perovskite compounds of absorbing high-energy photons and convert their energy into electronic carriers that are finally collected at the outer contacts. For soft X-ray photons (<10 keV), absorbing layer thickness of tens of micrometers suffices to stop the radiation. But hard X-ray radiation (10-50 keV) possesses much longer penetration length (≈100 µm) in compounds as methylammonium lead iodide or bromide (MAPbI 3 and MAPbBr 3 ). [3] Hence, relatively thick layers (0.1-1 mm) need to be employed to achieve sufficient electrical signal. Despite the large electronic carrier mobility and mobility-lifetime product, [4,5] such thick absorbing layers should be externally biased in order to increase the detector responsivity. Electrical fields as high as ≈0.1-1 V µm −1 are commonly needed for poly-crystalline perovskite deposits incorporated into X-ray detectors. [3,5] Large field strength requirements are in some amount lowered when devising single-crystal approaches for X-ray detection using perovskite materials. [4,6] In any case, it is widely recognized that the application of an external bias promotes the displacement and interfacial built-up of intrinsic mobile ions. [7,8] Ion accumulation finally shields the electrical field within the absorbing layer bulk, reducing as a consequence the detector sensitivity. The applicability of perovskite compounds for gamma photon (50 keV to 1 MeV) detection faces similar issues caused by electrical biasing. [9][10][11] Apart from electric field shielding, mobile ions might chemically interact with the materials at the outer electrodes. Migrating halides approaching the contacts, in the presence of high concentrations of electrons and holes, can produce electrochemical reactions. For example, generation of molecular I 2 has been reported which in turn induces electrochemical reactions like the oxidation of the external metal electrode to generate the corresponding halides, [12,13] i.e., Ag to AgI, where the formal oxidation state of the metal is modified from 0 to +1 as detected several times. Electrochemical reactions have also been observed

Introduction
Perovskite solar cells (PSCs) have shown rapid improvements in efficiencies, above 20% in the last few years, [1] indicating the promise of this technology for a prominent role in energy with the commonly used spiro-OMeTAD hole extraction layer, which can be reduced producing highly resistive neutral species. [14] Alternatively, anion exchange reactions with the anions of the external contacts have also been reported like in the transformation of ZnO into ZnI 2 . [2] Moreover, reversible reactions with TiO 2 have also been detected by Raman spectroscopy which are related to hysteresis in the j-V measurements. [15] The implications of this chemical reactivity is very important as it totally modifies the energy levels and injection barriers at the contacts and may induce performance degradation. [14] Apart from degradation caused by chemical interactions between intrinsic mobile ions and device constituents, extrinsic molecules (moisture, oxygen) are known to severely affect performance. Some strategies (encapsulation and interface engineering) to address that issue have been reported. [16] Interestingly, postdevice ligand treatments have been suggested consisting on direct reaction with ligand vapors to form stable low-dimensional perovskite complexes at the contact surface. [17] As far as high-energy photon detectors is concerned, usual contacting materials include metals of reduced reactivity. For instance, interdigitate Au/Ti electrodes were employed to analyze conductive properties of spin-or spray-coated MAPbI 3 X-ray sensitive layers. [3] To enhance cathode electron extraction in single-crystal MAPbBr 3 detectors, electron-transporting layers comprising 20 nm C 60 and 8 nm BCP interlayers were deposited on the bottom surface before Ag metal contact, while an Au (25 nm) anode was evaporated on the top surface. [4] More sophisticated contacting materials are formed by poly(3,4ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) and phenyl-C 61 -butyric acid methyl ester (PCBM) and ZnO as hole-selective and electron-selective contact and buffer layers, respectively. [5] In any case, gold is still extensively used either in direct contact with the perovskite absorber, [4,18] or covering proper interlayers. [6] However, it is already known that Au may interact with the sensitive perovskite layers causing degradation. For instance, gold migrates through the hole-transporting layer spiro-OMeTAD, and reacts with the perovskite material affecting the device performance metrics under working photovoltaic conditions. [19,20] Therefore, direct interactions of perovskite compounds and, specifically, their mobile ionic constituents with gold is an issue that deserves further analysis because of its important technological implications. This will be then the main subject of this work. Here single-crystal MAPbBr 3 is investigated in order to focus on contact reactivity effects and avoid grain-boundary interferences. Bromide perovskite compound is also chosen because of its greater ambient stability in comparison with iodide counterparts.
It is reported here that electrical biasing produces modifications of the current level by favoring the ability of perovskite/ Au interfaces to inject electronic carriers. Current level is determined by formation of new species comprising AuX (Au in oxidation state I or III) bonds (X = I and Br) that appear at the surface of the electrodes. Moreover, the effect of favoring current injection is observed to be highly similar irrespective of the halogen source, either extrinsic by iodine vapor sublimation of Au electrodes, or intrinsic by bias-driven migration of bromide ions. In addition, it is confirmed the formation of a dipole-like structures at the reacted electrode that lower the carrier potential barrier. Our findings confirm the important role that mobile intrinsic defects have on the device operation and performance deterioration through contact reactivity.

Results and Discussion
In this work monocrystalline MAPbBr 3 perovskite samples were prepared by using the inverse temperature crystallization method as described by Bakr and co-workers, [21] and conditions are described in the methods section. Symmetric devices are fabricated by evaporation of gold contacts in opposite sides of the single crystals, see the inset of Figure 1. Measurements of j-V curves were carried out in the dark at room temperature for different fresh crystals that have not been previously electrically biased. Current flowing through the device was registered initially from 0 V toward either positive or negative applied bias as in Figure 1a,b, respectively. Symmetrical contacts would entail symmetric j-V curves. However, preferential direction in one quadrant is observed when sweeping the voltage at low scan rates of 50 mV s −1 , indicating that biasing distorts the symmetric response. Interestingly, the initial bias direction determines the quadrant at which the current level is preferentially favored. Indeed, if the device is initially polarized toward positive voltages, high current is observed at positive voltages ( Figure 1a) and the opposite is observed when the device is initially polarized toward negative values ( Figure 1b). Next, in order to understand if the injection of carriers can be further promoted by poling the device at a fixed voltage, the two devices were polarized either at +5 V ( Figure 1a) or −5 V (Figure 1b) during 120 s. As can be observed in Figure 1, this polarization approximately doubles the injection current. Alternatively, the current can be partially or totally suppressed if polarization is carried out at the opposite quadrant, where current injection is not favored. Here, it is important to note that magnitudes of current and degree in which polarization is suppressed varies from sample to sample, but we have recurrently observed that the overall effect is largely reproducible. It is also noticeable that in the direction of preferential injection, current hysteresis is further promoted.
The previously reported experiment leads us to consider outer interfacial effects on the device current driven by the applied bias. In particular, here we propose that reversible reactions at the contact layer modify the interface electronic properties after halogen interaction with the Au electrode. As later demonstrated, the observed current has preferentially an electronic origin, in such a way that the increment/suppression of injected currents is connected to the interface reactivity assisted by the intrinsic halogen ion migration inside the perovskite bulk.
One related observation is the often reported generation of a remnant open-circuit potential (V oc ) in the device after polarization that persists for long time. [22] In particular, a device that has not been previously polarized shows a V oc = 0 V as expected for a symmetric device. However, after poling during 5 min it typically shows a V oc = 0.5-0.6 V, pointing to the presence of asymmetric contacts with some electrostatic charges remaining in one of them. This suggests us the formation of a dipole-like structure at the reacted electrode [8] with surface charge concentration of order 10 14 cm −2 resulting from simple electrostatic calculations. The net electrical effect is then lowering the potential barrier for electronic carriers at interfaces. The open question is now identifying the chemical nature of such interfacial structures.
We next investigate the effect of the scan rate on the electrical response of symmetrical device structure. However, it is adopted here a sandwich configuration (Figure 2a) since it provides superior experimental versatility as later explained. Interestingly, if a fresh and previously nonpolarized device is first measured at high scan rates of 2000 mV s −1 (Figure 2e), rather Ohmic response is observed in the form of an approximate straight line, as expected for a symmetric structure. If the device is then measured at 50 mV s −1 (low scan rates), as in Figure 1, preferential injection in the quadrant initially polarized is observed again. The effect of scan rate seems to be crucial as the increment in current is induced at low enough values (a full set of scan rate values is shown as supporting information). Importantly, the order in which the different scan rates curves are measured matters in the final recorded response. Indeed, a j-V curve measured at 2000 mV s −1 after biasing the device at slow scan of 50 mV s −1 does not suppress the large injection current that equals that observed in the slow measurement, as below demonstrated. In other words, this fact moves us to consider that modification of the injection properties at outer interfaces needs of long-time poling (very slow kinetics) to be effective, as the device shows some memory on the previous history/measurements.   One candidate process accounting for the interface reactivity is the generation of species containing Au-halogen bonds fostered by intrinsic ion migration upon biasing. In order to assess this mechanism, Au-halogen contact electrodes are presynthesized in the configuration glass/FTO/Au-I, and their electrical response is compared with a glass/FTO/Au reference electrode. Here, it is important to note that the chemistry of Au with the different halogens (Cl, Br, and I) is very similar, as reported elsewhere. [23,24] The Au-halogen contact electrode is synthesized by sublimation of molecular iodine onto a thin layer of evaporated gold. As can be observed in the scanning electron microscopy (SEM) images when compared to the reference glass/FTO/Au (Figure 2b), formation of Au-I (Figure 2d) does not take place homogeneously leaving unreacted Au islands (bright spots). Energy dispersive X-ray spectroscopy (EDX) confirms the presence of iodine atoms in the films. Alternatively, Raman spectroscopy (Figure 3a) reveals three bands in the region of 100-195 cm −1 , with wavenumbers typical for AuI(III) bonds of plasmon resonance signal for gold particles beyond 250 cm −1 . [25] The following reaction is then proposed for modified electrodes 2Au 0 3I vapor 2AuI III 2 3 ( ) where gold is oxidized and iodine is reduced. However, the reaction can be much more complex as Au chemistry is very rich and up to three different oxidation states have been reported to coexist in some gold halides. [23] Then, modified FTO/Au-I electrodes were used as external contacts in asymmetric configuration by sandwiching the perovskite monocrystal with a reference electrode of glass/FTO/ Au (Figure 2c). j-V curves were measured for the asymmetric configuration as shown in Figure 2f where high injection of carriers occurs in the prereacted electrode, regardless of the scan rate used during the measurement. Moreover, although the prereacted electrode exhibits larger current levels, the effect of favoring current injection is observed to be highly similar (Figure 2e,f) irrespective of the halogen source, either extrinsic by iodine vapor sublimation of Au electrodes, or intrinsic by bias-driven migration of bromide ions as next shown.
Considering the importance of the change in the chemical environment of the external contacts due to the migrating halide ions, here we propose that gold contact reacts with mobile bromide defects to form AuBr bonds (Figure 4). Very importantly, this chemical reactivity takes place in the same time scale (seconds-minutes) as the voltage sweep during the j-V curve measurement. Here we polarize devices at +5 V (see Figure 5), and observe that if devices are prepoled, high injection currents are observed at positive voltages even when the scan rate is 2000 mV s −1 . This electrical response readily entails that the prepoled process modifies the contact properties. Interestingly, a close inspection by Raman spectroscopy   ( Figure 3a) of an external electrode of glass/FTO/Au that has been previously incorporated in a symmetric device and poled shows bands in the region 80-200 cm −1 that are not present in the pristine reference FTO/Au electrode. This is a strong evidence that there are new species comprising oxidized Au such as AuX (I) or AuX (III) bonds (X = I and Br) being formed at the surface of the electrodes, either by effect of sample biasing or vapor sublimation. Further characterization was carried out by X-ray photoelectron spectroscopy (XPS) as shown in Figure 3b by registering a nonpolarized Au sample (glass/FTO/Au) and a polarized sample (glass/FTO/AuBr). The binding energy (BE) at the maximum of the Au 4f 7/2 and 4f 5/2 signals was found around 84.0 and 87.7 eV, respectively, which means a separation for the spin-orbit components about 3.7 eV. Further details on the analysis is provided as supporting information. These values are typical for pure metallic Au. [26,27] But also, a slight difference is observed in the signal obtained of Au polarized, an asymmetric widening of both contributions of Au 4f signal was observed toward higher BE, as is shown in Figure 3b. This asymmetric widening can be assigned to the presence of oxidized species of Au, which can be generated due to Au-halogen bonds. Typically, these Au species show a signal slightly shifted toward higher BE. [27,28] The reversible nature of the chemical reaction is probed by repeating Raman and XPS measurements after 6 months, see supporting information. After this period of time, only metallic gold is detected using both techniques. Our findings indicate then that electrode reactivity and enhancement of injection currents are closely related phenomena.
We note here that the basic coordination and redox chemistry using different halides should be similar. See, for example, early work by Braunstein et al. where synthesis of Au(I) and Au(III) is carried using similar chemistry for the three halides. [29] Of course, the reaction of the halide with the metal will take place with the halide according to the standard electrode potential of each of the halogens following the series Cl > Br > I being the most favored the reaction with Cl 2 . [30] However, the nature of the species generated by the different halides would depend on the halide used, i.e., Au(I) versus Au(III). Similarly, careful consideration on the thermodynamics and kinetics of this redox chemistry will be important when selecting the electrode metal as it will determine if this type of reaction will take place. We expect metals with negative standard electrode potentials (i.e., Ca, Na, or Al) to readily react with the perovskite leading to irreversible reactions. Alternatively, metal with positive standard electrode potentials such as Ag, Pt, or Au should better stand the highly oxidative conditions, in some cases leading to reversible reactions. In any case, each combination metal electrode/ perovskite should be carefully chosen.
Formation of oxidized Au(I) or Au(III) electrode is explained relying on the diagram shown in Figure 4. When an external bias is applied, the positive contact is charged with a large concentration of holes, which intervene in the metal contact oxidation (Figure 4a). The fact that halide ions are spatially close to Au atoms, and that PbX bonds are relatively weak, implies that halide exchange between metallic centers can be stabilized by forming Au(I) or Au(III) (Figure 4b). On the perovskite side of the interface, this halogen exchange leads to formation of a halide vacancy (V X + ), as those recently reported during observation of gold interdigitated electrodes. [31] In relation to this fact, we observed the signature of intrinsic ionic migration by using impedance spectroscopy for monocrystals of MAPbBr 3 . [32] In the long term upon continuous biasing, halide migration through a vacancy mechanism occur toward the negative electrode, that would lead to increased concentration of iodine ions in the positive electrode as observed by XPS measurements in a different study. [33] As noted above, the halogen exchange leads to the formation of halide vacancies (V X + ) in the perovskite side of the interface. This might lead one to consider the current increment observed upon biasing as exclusively originated by the transport of vacancies from the positively polarized electrode to the interior of the perovskite bulk (Figure 4c). However, the current has dominant electronic component as next explained. In a hypothetical ionic mechanism as a sort of electrochemical current generation, a continuous oxidation of the gold sustains the charge flux until the total consumption of one of the reactants, either Au + or X − , leads to a current drop. By examining Figure 5a using a fresh sample, one can infer that long-time biasing (+5 V) produces an almost constant current after 8 h and, therefore one can safely discard this possibility. By integration of the  chronoamperometric profile, we calculate a total charge approximately equal to Q ≈ 0.2 C cm −2 intervening in the experiment. In order to keep the current flowing by Au electrode oxidation, the thickness of the metallic contact d = Qv/e should be as large as 850 nm, with e being the elementary charge. This last calculation assumes a gold unit cell volume of v = 0.0679 nm 3 . However, the actual gold contact is only 60 nm thick. It should also be stressed the incompatibility between the current response of Figure 5a, exhibiting a steady increase over time, and models that predict current decays caused by ion-blocking, nonreacting electrodes. [34] One can estimate the hypothetical density of formed halide vacancies after 15 h of continuous polarization considering the thickness of the perovskite layer ≈1 mm. The V X + density results as high as 10 19 cm −3 , certainly a physically unrealistic value that would imply a huge defect density alteration, as the sample is a monocrystal with typical reported densities of about 10 12 cm −3 . [35] The experiment in Figure 5b corroborates the electronic origin for the current mechanism. In the case of being the current exclusively of ionic origin, a significant current reduction is expected for high-rate biasing because ionic transport-limited mechanisms. On the contrary, large current levels are reported, even for high scan rates (2000 mV s −1 ), as noted also previously. Therefore, our findings indicate that the reactivity of the contact, by oxidation of the gold electrode and the formation of metallic halide at the interface, is able to modify the injection properties for electronic carriers that are the responsible for the measured current.

Conclusions
In this work, we have investigated the importance of the chemical reactivity of Au electrode in contact with single-crystal perovskite semiconductor layers. This contact has been the preferred choice in several electronic applications. It is proposed that specific interface reactivity occurs through the generation of species containing Au-halogen bonds fostered by intrinsic ion migration upon biasing. The consequence of having metal halide compounds at the electrode interface is the net increase of the electrical current. The effect of favoring current injection is observed to be highly similar irrespective of the halogen source, either extrinsic by iodine vapor sublimation of Au electrodes, or intrinsic by bias-driven migration of bromide ions. The overall current mechanism can be rationalized as a modification of the injection properties (lowering potential barriers) of electronic carriers at the outer interface by intrinsic mobile ion electrode reactivity originating dipole-like interlayers. Contact reactivity then dominates the operation characteristics, rather than being governed by bulk transport properties of the charge carriers, either electronic or ionic. In addition, it is observed that the Au electrode oxidation is a highly reversible process that follows rather slow kinetics.

Experimental Section
The MAPbBr 3 single crystal with variable dimensions, typically in around 3 × 3 × 1 mm 3 , was prepared by an inverse temperature crystallization technique. [21] Lead bromide (PbBr 2 , TCI) and methylammonium bromide (CH 3 NH 2 ⋅ HBr, MABr, and Dyesol) were dissolved in N,Ndimethylformamide (DMF, Aldrich) to obtain a 1 m solution of MAPbBr 3 . This solution was then filtered with a polyvinylidene fluoride (PVDF) filter (0.45 µm) into a clean vial. The synthesis of MAPbBr 3 crystals took place in an oil bath. The initial temperature was set at 60 °C. The solution was kept at this temperature for 30 min, and then the temperature was increased every 15 min by 5 °C. Finally, to grow the MAPbBr 3 crystal of required size, the solution was kept for 30 min at 80 °C. The crystal was then extracted from the solution, rinsed in diethylether and dried with a nitrogen stream.
Devices were prepared by evaporation of gold (60 nm) onto two opposite sides of the crystals with an active area of 4 mm 2 . j-V and polarization curves of the MAPbBr 3 crystal was carried out using an Autolab PGSTAT-30. First, j-V curves were registered from previously non-measured crystals from zero voltage without any polarization. Then we polarized the crystal with positive (+5 V) or negative voltage (−5 V) and immediately measured j-V curves, again, from zero voltage. We also investigated the effect of measuring j-V curves from 0 V toward negative values and poling during 120 s at either +5 or −5 V.
Electrodes containing Glass/Au/Au-I were prepared by evaporation of 60 nm of gold on glass/FTO substrate and then promoting the chemical reaction by placing upside down the electrode in the lid of a petri dish and sublimation of iodine crystals from the bottom of the Petri dish during 12 h. In this way we obtained AuI/AuI 3 with islands of unreacted gold. Devices were prepared by sandwiching a MAPbBr 3 crystal with a soft separator between glass/FTO/Au electrode and a second electrode that can be either glass/FTO/Au or glass/FTO/Au-I. j-V curves of two different structures were measured Au-MAPbBr 3 -Au and Au-MAPbBr 3 -Au-I.
Raman spectroscopy measurements were carried out with a dispersive spectrometer NRS-3100 (Jasco) with energies of 80 mW and excitation times of 1 s used with a holographic filter with a cutoff edge at 70 cm −1 . X-ray photoelectron spectroscopy was used to study the Au chemical bonding states in the samples. The spectra were registered by using a Kratos Axis UltraDLD spectrometer, with monochromated Al Kα radiation (1486.6 eV) and a 20 eV pass energy. Electrostatic charging effects could be stabilized with the help of a specific device developed by Kratos.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.