Environmental sensitivity of GaN nanofins grown by selective area molecular beam epitaxy

Nanostructures exhibit a large surface-to-volume ratio, which makes them sensitive to their ambient conditions. In particular, GaN nanowires and nanofins react to their environment as adsorbates influence their (opto-) electronic properties. Charge transfer between the semiconductor surface and adsorbed species changes the surface band bending of the nanostructures, and the adsorbates can alter the rate of non-radiative recombination in GaN. Despite the importance of these interactions with the ambient environment, the detailed adsorption mechanisms are still not fully understood. In this article, we present a systematic study concerning the environmental sensitivity of the electrical conductivity of GaN nanofins. We identify oxygen- and water-based adsorbates to be responsible for a quenching of the electrical current through GaN nanofins due to an increased surface band bending. Complementary contact potential difference measurements in controlled atmospheres on bulk m- and c-plane GaN reveal additional complexity with regard to water adsorption, for which surface dipoles might play an important role besides an increased surface depletion width. The sensitive reaction of the electrical parameters to the environment and surface condition underlines the necessity of a reproducible pre-treatment and/or surface passivation. The presented results help to further understand the complex adsorption mechanisms at GaN surfaces. Due to the sensitivity of the nanofin conductivity on the environment, such structures could perform well as sensing devices.

Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. efficiency of related devices, especially due to their high surface-to-volume ratio. Further, meaningful measurements on GaN NSs require repeatable electronic properties like a reproducible surface termination and surface band bending (sbb), which is a requirement that is often challenging to ensure. Molecules from the ambient induce a charge transfer between these adsorbates and the GaN, thereby influencing the charge carrier concentration and surface properties of the NSs [18]. As a prominent example, the ambient environment has a pronounced impact on the overall intensity of the photoluminescence (PL) of GaN NSs [19][20][21] as well as their chemical stability [22] and electrical properties [23,24]. Nevertheless, there is still a lack of understanding the underlying dynamics of the adsorption mechanisms. In addition, results and interpretations in the literature are sometimes contradicting each other, and a generally valid explanation for all experimental observations has not yet been found. A recent study investigated the influence of the main air constituents on the radiative recombination efficiency of GaN NSs and found that oxygen can act either quenching or enhancing on their PL intensity, depending on the temperature [24]. By combining electrical and PL measurements, they conducted that oxygen gives rise to two fundamentally different adsorption states. In this paper, as a complementary approach we report on a detailed study of the influence of oxygen and water on the electrical conductivity of GaN NFs depending on the illumination conditions and the temperature. By combining these results with contact potential difference (CPD) measurements under controlled ambient conditions, we gain a further insight in the complex environmental adsorptionkinetics at GaN surfaces.

Experimental
GaN NFs were grown on either 4 μm thick Ga-polar ironcompensated GaN templates on sapphire or on 2 μm thick Alpolar not intentionally doped AlN templates on sapphire from Saint-Gobain Lumilog and Xiamen Powerway Advanced Material Co., Ltd, respectively. Samples were cut into 10 × 10 mm 2 pieces. The backsides of the sapphire substrates were coated with 200 nm Ti followed by 70 nm Pt in order to enhance the heat absorption from the radiative heater during molecular beam epitaxy (MBE) growth. The effective substrate temperatures noted below refer to thermocouple values of the substrate heater, which overestimate the real temperatures at the substrate surface by about 100°C.
The GaN NSs were grown via MBE in a selective area growth mode. To this end, 10 nm of Ti were evaporated onto the sample surface as a mask material. The Ti mask was patterned via electron beam lithography and wet-chemical etching with arrays of lines, which serve as nucleation spots for the NF growth. By defining the line direction relative to the crystallographic orientation of the substrate, it is possible to grow GaN NFs with either a-or m-plane side facets. The top facets of the NFs are semi-polar ({¯} 1102 and {¯} 1124 facets for m-and a-plane NFs, respectively) [25,26]. Prior to the MBE growth, the AlN substrates were etched with hydrochloric acid (HCl, 36%) in order to remove the native surface oxide and to obtain a reproducible surface termination. The MBE growth started with a nitridation of the samples by exposure to a nitrogen plasma with a power of 425 W at 400°C and 800°C for 10 and 5 min, respectively. This leads to the conversion of the Ti mask into a more stable TiN x layer. The actual GaN NS growth was performed with the same plasma power, a Ga flux of 6 · 10 −7 mbar beam equivalent pressure and a substrate temperature of 960°C. The growth duration was 90 min. A detailed description of the NS fabrication process can be found in the literature [25][26][27][28]. The background pressure of the MBE system was in the 10 −10 mbar range.
After MBE growth, to avoid short circuits the TiN x masks were removed in a mixture of hydrofluoric acid (HF, 50%) and hydrogen peroxide (H 2 O 2 , 31%), followed by a final etch in a pure HF (50%) bath. Subsequently, ohmic Ti/ Al/Ti/Au (20/80/20/120 nm) contacts were evaporated onto arrays of 50 NFs with a period of 1 μm using a tilted substrate rotation in order to also deposit the metals on the NF side walls. The distance between both contacts was 30 μm. If not mentioned otherwise, a constant voltage of 1 V was applied to the NFs during the measurements. Figure 1(a) shows scanning electron microscopy (SEM) images of GaN NFs with a-and m-plane side facets. The NFs are characterized by a homogeneous width and height with smooth side walls. Further, the a-plane NFs are slightly higher compared to their m-plane counterparts (420 nm instead of 350 nm) while being 30 nm thinner (90 nm instead of 120 nm). These differences are a result of an anisotropic Ga diffusion on the NF side facets during MBE growth [26]. An exemplary optical microscopy image of the lateral Ti/Al/Ti/Au contacts on an array of 50 GaN NFs with a period of 1 μm is shown in figure 1(b). As there would be a significant photocurrent of the GaN templates due to ultra-violet (UV) excitation, all experiments under illumination were conducted on GaN NFs grown on an AlN template. However, initial experiments were performed on GaN NFs grown on a GaN template as the growth on this substrate was already established at the time of the start of this study, in contrast to the growth on AlN. Nevertheless, as discussed below there is no qualitative difference of the dark current response to the ambient atmospheres (see figures 2 versus 3).
Heating of the samples was achieved by a resistive heater next to the NF array. A calibration of the heating current was performed by measuring the temperature-sensitive Raman shift of the GaN NF E 2 (high) mode (see figure S3, supporting information).
Oxidation of the GaN NFs was achieved via a microwave oxygen plasma system operating at room temperature (RT) with a power of 200 W. AlO x shells were grown via thermal atomic layer deposition (ALD) using water and trimethylaluminum as precursors at a reactor temperature of 200°C in a hot-wall reactor (Fiji G2, Veeco CNT).
The environmental sensitivity of GaN NFs was investigated in a flow chamber, which can be evacuated by means of a turbo molecular pump to a pressure of approx 5 · 10 −6 mbar. For measurements in gaseous atmospheres, the chamber was flushed with a fixed gas flux of 1.5 l min 1 and pumped via a membrane pump in order to maintain a base pressure of 1 · 10 2 mbar. The influence of different gases was investigated to gain insight into the effects of an inert gas in comparison to the influence of water and oxygen. Nitrogen was used as an inert gas, which was extracted by a phase separator from liquid nitrogen and has oxygen and water impurities below 75 ppm and 30 ppm, respectively. We used nitrogen as it was more cost-efficient compared to other inert gases like argon. Further, as there is no detectable response of the electrical conductivity of the GaN NFs when the ambient atmosphere is changed from dry nitrogen to dry argon (see figure S5, supporting information), we think that dry nitrogen acts sufficiently well as an inert gas. For measurements in an oxygen-containing atmosphere, we used synthetic air (SA), which is a mixture of 20% oxygen and 80% nitrogen (purity 99.999%). For measurements in humid environments, the respective gases were bubbled through deionized water leading to a relative humidity of 100%. The samples can be illuminated through a sapphire window either by a UV LED (Thorlabs M340L5, nominal emission at 340 nm) or a green LED (Thorlabs MINTL5, nominal emission at 554 nm) with an excitation intensity of 4 mW cm −2 and 32 mW cm −2 , respectively. The LED spectra are shown in figure S1 of the supporting information.
All samples (except the passivated ones) have been etched with HCl (36%) and rinsed with deionized water and isopropanol prior to mounting them in the measurement chamber. The aqueous HCl effectively removes a large amount of the native oxide from the GaN surfaces and leaves a chlorinated termination behind, which is expected to delay re-oxidation of the GaN [29][30][31]. Additionally, a lowered sbb is reported after HCl pre-treatment [31,32]. After this HCl treatment, the sample is mounted as quickly as possible into the measurement chamber, which is then evacuated to high vacuum in order to have reproducible starting conditions for the experiment. The measurement chamber was pumped for at least four hours as this was the duration needed to reach stable and good vacuum conditions. Longer pumping times do not influence the environmental sensitivity of the GaN NFs. The necessity of this process becomes apparent in figure 1(c). The plot shows the response of the same GaN NFs to humid nitrogen with and without pumping of the measurement chamber after mounting the sample. While we observe a quenching due to water for the vacuum-exposed NFs, there is a current enhancement in humid nitrogen in the case of the NFs without preceding vacuum storage. Possibly, adsorbed species from the HCl treatment but also from the water and isopropanol rinsing might influence the interaction of the NFs with humidity. These species seem to desorb more efficiently in vacuum compared to solely purging with dry nitrogen.
The measurements in aqueous environments were carried out in a glass beaker, which was closed at the top by a sealing film but allowed for saturation with either SA or nitrogen through in-and outlets. The samples have been passivated by an SU8 photo-resist leaving only the NF arrays exposed to the environment in order to prevent short circuits.
CPD measurements were performed in the same measurement chamber on commercially available not intentionally doped free standing bulk m-plane GaN plates (MSE Supplies LLC) and 4 μm thick Ga-polar Si-doped GaN templates on sapphire (Saint-Gobain Lumilog). The custom-built Kelvin probe setup is equipped with a commercial controller (Kelvin Control 07, Besocke DeltaPhi) and a piezoelectrically driven gold grid with a diameter of 3 mm and a work function (WF) of 4.9 eV as a probe. The CPD voltage is the difference between the WF of the gold probe and the semiconductor: Thus, negative CPD values correspond to a higher WF of the semiconductor.

Sensitivity of a-and m-plane NFs in the dark
The interaction of GaN a-and m-plane facets with the ambient environment differs, especially with regard to their chemical stability [22]. Thus, we investigate whether the ambient conditions have a qualitatively different influence on the electric properties of a-and m-plane NFs. It is known from previous studies that adsorbed molecules are able to change the sbb of GaN NSs [20,24]. An increased sbb reduces the width of the conductive channel in the NFs, which decreases the conductivity of the structure. The a-and m-plane NFs are grown on the same piece of GaN template only a few μm away from each other and their responses to the environment are measured simultaneously in the same measurement run. Figure 2 shows the current through NFs with a-and m-plane side facets with an applied voltage of 2 V and 1 V, respectively. The higher voltage on the a-plane NFs ensures a more comparable current (19.6 μA and 19.0 μA) of both NF arrays. The initial higher resistivity of the a-plane NFs might be due to their smaller width and due to sample fabrication issues like a different quality of the electrical contacts. Figure 2(a) depicts the current response as the chamber gets purged with dry nitrogen, followed by dry SA and back. The data is normalized to the dark current saturation value under vacuum. Note that the left (right) y-axes refer to the a-(m)-plane, respectively, and are quantitatively different. When nitrogen is introduced into the evacuated chamber, the current through both NF-types decreases over a time range of hours. While we observe a current quenching to approx. 99% of the vacuum level for the a-plane NFs, the current through the m-plane NFs is quenched to approx. 94% within 90 min. Note that the quenching is still not saturated after this time. When the dry nitrogen is substituted by dry SA, the quenching accelerates again and the current is reduced below 98% and 88% after additional 180 min, respectively. Changing the atmosphere back to dry nitrogen does slightly attenuate the quenching and leads to a saturation but not to a recovery of the current. Evacuating the chamber back to high vacuum conditions leads to an enhancement of the current. Thus, vacuum is necessary to recover the NFs from the effect of the surrounding atmospheres. This shows the importance of the chamber evacuation after each experiment in order to obtain meaningful results. In we have normalized the current to the saturated dry nitrogen level, which was purged into the freshly evacuated measurement chamber. Further, we fixed the exposure times to the probe gases at 20 min, respectively. Within this time frame, SA leads to a reduction of the current to 99% (95%) of the nitrogen level while humid nitrogen quenches the current even further to <99% (<92%) for the a-(m-)plane NFs, respectively. A slight recovery can be observed when switching from humid nitrogen exposure to dry nitrogen (figure 2(c)). The accompanying immediate overshoot after switching is difficult to explain. Probably, fluctuations in the gas flow through the measurement chamber are responsible for this effect, which we did not observe for the NFs grown on the AlN templates (see figure 3). Figure 2(d) shows the combined effect of oxygen and humidity. After we observed the initial quenching in dry SA, we flushed the chamber with humid SA. This led to the strongest quenching down to approx. 97% (86%) of the nitrogen level. Switching back to dry SA did not interrupt the quenching process. The final dry nitrogen exposure stabilized the currents at their lowest values.
In these experiments, we observed that the a-and mplane NFs behave qualitatively very similarly, but with an overall higher sensitivity of the m-plane NFs towards changes in the ambient conditions. Nevertheless, a repetition of the experiment on a different sample lead to the opposite result of a higher response of the a-plane NFs (figure S2, supporting information). Therefore, the magnitude of the response might be the result of other unknown effects, likely arising from variations of the sample fabrication process. Improvements in the process as well as better statistics are necessary to accurately determine any quantitative differences of side facet sensitivity. We therefore focused on the m-plane NFs for the subsequently performed experiments.

Influence of illumination and temperature
It has been shown in the literature that the current quenching through GaN NFs due to oxygen exposure can be enhanced by increasing temperature and under high-intensity UV excitation [24]. Thus, we want to study the influence of UV and visible light excitation on the environmental sensitivity of GaN NSs with moderate excitation intensities as well as at different temperatures in the dark. To this end, m-plane GaN NFs were grown on an AlN template, which does not generate a measurable photocurrent response at the investigated excitation wavelengths. Figure 3(a) shows that the dark current through these NFs is quenched to 99.5% of the dry nitrogen level within 20 min in dry SA with no recovery when switching back to dry nitrogen. An excitation with sub-band gap light (green LED) leads to a stronger current reduction down to 99.0% with only a very small recovery when switching back to dry nitrogen. When above-band gap light (UV LED) is used to illuminate the NFs, the current reduction in dry SA is most pronounced (98.4%). Further, a noteworthy recovery of this quenching is observable after excitation with UV radiation. The influence of humidity is displayed in figure 3(b). Here, we observe qualitatively the same behavior for humid nitrogen, an overall stronger sensitivity to water and again only a recovery under UV excitation. The presented results indicate a significant role of light-induced charge carriers of the environmental sensitivity to water and oxygen on the NF surfaces.
The ambient temperature significantly influences the adsorption kinetics of molecules on GaN surfaces [24]. Figure 3(c) shows the influence of dry SA on the current through GaN NFs at different temperatures in the dark. Note that the initial absolute current levels show no trend with the temperature. As heating of the GaN NFs changes their carrier density as well as the amount of adsorbed molecules, the absolute current values follow a complicated behavior. At RT, the previously discussed quenching is again observed. When a power of 0.3 W is applied to the heating wires next to the NF array, their temperature is expected to increase. However, the heating was not measurable within the accuracy of our Ramanbased method (see figure S3, supporting information). Nevertheless, at this slightly elevated temperature (RT + ) the current quenching in dry SA is more pronounced. This trend continues for a temperature of 36°C and a slight recovery is observable when switching back to dry nitrogen. At a temperature of 110°C, we observe the strongest and fastest decrease in current. Further, within 20 min in dry nitrogen, the current recovers almost completely to the initial level before the oxygen exposure. The results indicate that the oxygen ad-and desorption kinetics at GaN surfaces depend on the sample temperature. Nevertheless, from these measurements at elevated temperatures it is still hard to estimate whether the observed effects are oxygen-induced as residual water in the system might influence the measurements. Although a lower water adsorption rate on the GaN surfaces is expected at elevated temperatures, ambient-sensitive PL measurements in the literature show that even at temperatures of approx. 120°C water influences the surface properties of GaN NSs [33].

Complementary contact potential difference study
In order to be able to more accurately interpret the response of GaN NFs to humidity and oxygen presented in the previous section, complementary CPD measurements on c-and mplane GaN samples have been performed. The experimental conditions resembled the conditions for the current response investigations as closely as possible. Figures 4(a)-(d) show the CPD response of m-and c-plane GaN to oxygen and humidity, respectively. The gray graphs represent the CPD response in the dark, whereas the violet data points were recorded while illuminating the samples with UV light (340 nm, 6 mW cm −2 ). The surface photovoltage (SPV) is defined as the difference between the CPD in the dark and under illumination and is depicted in red.
In dry nitrogen, m-plane GaN exhibits a CPD level of approx. −0.3 V. When the sample is exposed to dry SA, the CPD increases by 0.1 V. After 20 min, the chamber is purged with dry nitrogen again and the CPD hardly recovers within 10 min. Under UV illumination, the CPD in dry nitrogen is reduced by almost 0.3 V compared to the dark level. When the illuminated m-plane GaN is exposed to dry SA, the CPD increases by more than 0.2 V within 20 min. Further, there is a pronounced recovery in dry nitrogen. The corresponding SPV drops from 0.3 V in dry nitrogen to below 0.2 V in SA. Figure 4(b) depicts the CPD response of m-plane GaN to water. In humid nitrogen, the dark CPD increases by almost 0.02 V, whereas the level of the UV-illuminated GaN drops by 0.06 V within 20 min. Independent of illumination, both CPD curves show a slight recovery when the ambient is switched back to dry nitrogen. The SPV increases in humid nitrogen by more than 0.02 V. Figure 4(c) shows that the CPD levels of the c-plane GaN sample in dry nitrogen is about −0.39 V in the dark and −0.42 eV under UV illumination. When the sample is exposed to dry SA, the CPD in the dark increases by 0.1 V whereas the CPD level under UV light increases only by 0.03 V with a slight overshoot in the first 2 min. In contrast to the m-plane GaN, the c-plane sample shows a significant recovery from the oxygen exposure in the dark. The SPV increases by 0.07 V due to oxygen in the chamber. The influence of humidity on the CPD of c-plane GaN is depicted in figure 4(d). In humid nitrogen, the CPD in the dark decreases by approx. 0.01 V and shows almost no recovery in dry nitrogen. Under UV illumination, the CPD decreases due to the humidity by 0.04 V and shows a slight recovery when switching back to dry nitrogen. The SPV increases in humid nitrogen by 0.03 V.
The discussed CPD values together with their placement relative to the CPD levels under vacuum (corresponding CPD transients shown in figure S4, supporting information) are summarized in figures 4(e)-(f). In the case of m-plane GaN without illumination, dry nitrogen increases the CPD compared to the vacuum level. Oxygen and humidity cause a further enhancement, but with a stronger impact of oxygen. In general, UV illumination lowers the CPD levels in each atmosphere. Further, the CPD decreases when the ambient around the sample under UV illumination is changed from vacuum to dry nitrogen. Humidity causes a further decrease of the CPD under UV illumination, while dry SA lead to an increase in the CPD relative to the dry nitrogen level.
For c-plane GaN, there are only two qualitatively different CPD responses when compared to the m-plane sample: In the dark, the CPD decreases when humidity is induced in nitrogen. Further, in vacuum the CPD in the UV-illuminated case is lower than the dry nitrogen level.

Discussion of the NF response combined with CPD analysis
A comparison of the ambient response of the GaN NFs and the macroscopic CPD measurements allows a deeper insight into the underlying mechanisms of the interaction of GaN with the environment. The environmental sensitivity of the electrical current through GaN NFs is mainly a consequence of the ambient-induced changes in the sbb, which influences the width of the conductive channel within the NF. The response of the CPD measurements to the environment is expected to be influenced by the sbb, but also by electrostatic effects due to the adsorbates from the ambient, which might lead to the formation of surface dipoles [31].

Without illumination
In the dark, the current through GaN NFs decreases when the evacuated chamber is purged with dry nitrogen. Accordingly, the CPD measurements indicate that the sbb of m-and c-plane GaN increases. We assume that residual water and oxygen in the nominally dry nitrogen atmosphere are responsible for this effect since these impurities (see section experimental) can be translated into a partial pressure of approx 10 −2 mbar, which is much higher than the partial pressure of oxygen and water in the high vacuum. Note that there is no change in the current through GaN NFs when the atmosphere is changed from dry nitrogen to either dry CO 2 or dry argon (see figure S5, supporting information). An intentional exposure of the samples to oxygen (dry SA) causes a further upward sbb accompanied by an enhanced quenching of the NF current. These results are in accordance with the literature, which reports an oxygen-induced Fermi level pinning for polar GaN surfaces [34] and GaN NWs [20,35]. The response of GaN NSs to ambient water is more complex: The current through the NFs decreases relative to the dry nitrogen level. The CPD of the m-plane GaN sample increases when exposed to humidity while the c-plane GaN shows the opposite response. Further, water exposure has a larger impact on the electrical current through GaN NFs while oxygen impacts the CPD signal more. Thus, adsorbed water might increase the sbb in the GaN NFs but additionally could lead to a significant accumulation of surfaces dipoles, which noticeably influence the CPD measurements. Figure 2(c) showed that a combination of water and oxygen resulted in the strongest observed current quenching. Wang et al reported that water and oxygen exposure alone have no significant impact on the electronic properties of Ga-polar GaN. Instead, when GaN is exposed to both species, a deep surface depletion region forms as water on the GaN surface acts as a conduit for the oxygen adsorption from the environment [36]. In our case, we already observed a significant current quenching when exposing the GaN NFs to only one of these adsorbates. On the one hand, this might indicate that the nominally dry SA and humid nitrogen are not water-and oxygen-free. Especially water is expected to be omnipresent on the GaN surfaces at RT [37]. On the other hand, a higher sensitivity of our high surface-tovolume ratio NSs exposing also non-polar side facets might even reveal the smaller impact of either water or oxygen alone on the electronic properties of GaN.
At elevated temperatures, a faster and more pronounced current quenching was observed together with a higher degree of reversibility. It is known from the literature that oxygen can bind via dissociative adsorption and bridge adsorption to GaN surfaces [38]. Further, measurements on GaN NFs have shown that at low temperatures oxygen causes mainly an enhancement of the sbb (indicating bridge adsorption). At elevated temperatures, an additional surface termination was observed (dissociative adsorption) [24]. However, the hightemperature dissociative oxygen adsorption state mainly influenced the PL intensity in this study and it did not significantly affect our electrical measurements. In addition, the dissociative oxygen adsorption is more persistent, which is in contradiction to the enhanced reversibility of the electrical current in dry nitrogen at elevated temperatures. More likely, higher temperatures catalyze also the bridge adsorption of oxygen as potential barriers of the oxygen binding process might be overcome more easily, which is supported by the faster response times at elevated temperatures.

With illumination
The UV excitation increases the current through the GaN NFs by up to approx. 8%. However, the exact photocurrent is hard to determine as photo-induced ad-and desorption of molecules additionally impact the NF conductivity. When the GaN NFs are under illumination, their current quenching in SA increases with a stronger impact of above-band gap illumination. Further, a significant degree of recovery when the atmosphere is changed back to dry nitrogen was only observed during UV illumination. Appropriately, the m-plane CPD measurements show an enhanced response in oxygen under UV light exposure. The measurements on Ga-polar cplane GaN deviate from this behavior as the CPD reaction to oxygen and the recovery is stronger without UV illumination. Overall increased oxygen ad-and desorption induced by UV light have previously been reported in the literature, where UV-induced photo-ad-and desorption of oxygen on planar and NS GaN altered the PL intensities of the GaN samples due to Fermi level pinning [20,34,35,39].
In humid nitrogen, the GaN NFs showed a current quenching, which was enhanced by illumination. Interestingly, the CPD of m-and c-plane GaN significantly decreased upon water and UV light exposure. Again, this points to the presence of water-based surface dipoles, which are adsorbed at the GaN surface and might counteract the detection of an increased sbb due to humidity in the CPD measurements. In any case, we observe an enhanced response to humidity and recovery in dry nitrogen when the samples are exposed to UV illumination. This indicates that there might be a UV-induced photo-ad-and desorption of water at the GaN surfaces similarly to the interaction with oxygen. Further, the illumination might modify the strength of the water-based surface dipoles [31].
Note that an enhanced sbb of GaN due to water exposure is in contradiction to previous XPS studies on polar GaN [40,41]. However, as the GaN NFs mainly expose non-polar facets to the ambient, their influence might overweight and counteract the cplane sbb. On the other hand, the electric field on the GaN surface due to the water-based surface dipoles might also influence the binding energies observed in XPS measurements in a similar manner as it affects the presented CPD measurements.
When the GaN NFs are illuminated with green light, electrons from either surface states or from defects in the bulk can be excited into the conduction band. They contribute to the photocurrent of the NSs and can reach the surface via a hopping mechanism [42]. There, they might contribute to an enhanced sensitivity of the GaN to its environment, e.g. due to a transfer of photo-generated charge carriers from the semiconductor to the adsorbed species, thereby influencing the sbb [39]. Nevertheless, the photo-generated charge carriers do not lead to an enhanced desorption of oxygen as we observe no recovery in dry nitrogen under green light. This is reasonable since the UV-induced desorption of oxygen is induced by the transfer of photo-generated holes from the bulk to the surface [20]. However, visible light excites electrons from defects and surface states, leaving no holes behind that could contribute to the oxygen desorption.

Ambient interaction of oxidized GaN NFs
Recently, it has been demonstrated that the surface treatment of GaN NSs influences their environmental interaction and thereby their transient PL properties [24]. In detail, HCl-etched and oxidized NSs behaved very differently. Further, oxidation is expected to passivate the surface states of GaN [43,44]. This motivates the investigation on the influence of the surface oxide also on the electrical response of GaN NFs. Figure 5 shows that there is the expected current quenching of the normally HCletched GaN NFs in dry SA. After oxidation of this sample in a plasma system at room temperature (200 W, 5 min), the total base current in dry nitrogen is reduced by 0.1 mA.
It is difficult to tell whether this reduction can be assigned to the electronic properties of the NFs as the whole sample including the contacts etc have been exposed to the oxygen plasma. However, an increased sbb of oxidized GaN compared to HCl-etched surfaces would be in accordance with the literature [31]. More interestingly, the NFs show almost no response to dry SA after the oxidation process. In contrast, when exposed to humid nitrogen, we observe an increase of the current. The strength of this enhancement is even stronger than the response of the oxygen-induced quenching of the initially HCl-etched sample. Again, this reversal of the response to humidity accentuates out the complexity of the water adsorption processes at the GaN (GaO x ) surfaces. Further, a simple oxidation of GaN is not sufficient to passivate the NFs from the ambient. Even a more sophisticated passivation method like a conformal 20 nm thick AlO x shell grown by ALD is not sufficient for a complete passivation of the sample (figure S6, supporting information).

Sensing of oxygen dissolved in water
So far, all presented measurements were conducted in gaseous environments. Nevertheless, it is of interest whether the sensitivity of the GaN NFs persists in liquid environments. Figure 6 shows the transient current through an array of 5000 NFs. On this sample, only the NFs are exposed to the environment, while the rest of the sample is passivated. After the current was stabilized in ambient air, the sample was immersed in deionized water through which we bubbled ambient air. The current level did not change significantly when the sample was put under water, but the noise level increased drastically. This is likely caused by disturbance from the bubbling. Next, the gas used for bubbling was switched to nitrogen. After a delay of 5 min, the current increases by 40% within 10 min. When the nitrogen is changed to SA, the current rapidly decreases within 2 min and drops even below the initial level in ambient air. In contrast to the measurements in gaseous environments, the oxygen-induced quenching is reversible by solely purging the ambient water with nitrogen. The results demonstrate that the sensitivity of GaN NFs to oxygen / nitrogen is not limited to gaseous environments but is also existent in a liquid ambient, which greatly increases their potential for sensing applications.

Conclusion
In summary, we have presented a detailed study on the environmental sensitivity of the electrical conductivity of GaN nanofins (NFs). Starting from the base level in dry nitrogen, either oxygen in dry synthetic air (SA) or humidity in a nitrogen atmosphere lead to a quenching fn the current through the NFs. Further, a combination of oxygen and water leads to the strongest quenching. Elevated temperatures or optical excitation enhance these effects, with above-band gap illumination being more effective than sub-band gap light. A recovery of the quenched current could be achieved either by increasing the sample temperature or by vacuum or ultra-violet (UV) light exposure. Complementary contact potential difference (CPD) measurements on bulk m-and c-plane GaN reveal that oxygen adsorption mainly increases the surface band bending (sbb). The adsorption of water on the GaN surfaces seems to be more complex. Besides an increase in sbb, the water-based adsorbates might also lead to the formation of a surface dipole dominating the CPD response in humid atmospheres. Interestingly, the sensitivity of GaN NFs to oxygen persists when immersed in liquid water. A reproducible sample conditioning prior to the experiments is essential for achieving meaningful results, as the sample pre-treatment qualitatively changes the electrical response of the GaN NFs to their environment. Our results highlight the complexity and importance of the environmental conditions on the electric properties of GaN NFs. Generally, this article demonstrates the fundamental importance of controlled ambient conditions during the (opto-)electronic characterization of NSs, which is most likely not limited to the GaN-based material system.

Acknowledgments
This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germanys Excellence Strategy -EXC 2089/1 -390776260. Supported by Deutsche Forschungsgemeinschaft (DFG) through TUM International Graduate School of Science and Engineering (IGSSE), GSC 81. The authors thank C Paulus for her essential support in the laboratories and A Henning for the AlO x deposition.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).

ORCID iDs
Florian Pantle https:/ /orcid.org/0000-0001-5592-8891 Max Kraut https://orcid.org/0000-0002-7274-4856 Figure 6. Influence of dissolved oxygen in liquid deionized water on the electrical current through an array of 5000 NFs. Gray represents the current of the NFs in gaseous ambient air, red in untreated liquid deionized water and blue (green) in deionized water purged with a constant gas flow of nitrogen (SA). The right axis shows the current normalized to the level in ambient air.