Imaging the influence of oxides on the electrostatic potential of photovoltaic InP nanowires

Nanowires require surface passivation due to their inherent large surface to volume ratio. We investigate the effect of embedding InP nanowires in different oxides with respect to surface passivation by use of electron beam induced current measurements enabled by a nanoprobe based system inside a scanning electron microscope. The measurements reveal remote doping due to fixed charge carriers in the passivating POx/Al2O3 shell in contrast to results using SiOx. We used time-resolved photoluminescence to characterize the lifetime of charge carriers to evaluate the success of surface passivation. In addition, spatially resolved internal quantum efficiency simulations support and correlate the two applied techniques. We find that atomic-layer deposited POx/Al2O3 has the potential to passivate the surface of InP nanowires, but at the cost of inducing a field-effect on the nanowires, altering their electrostatic potential profile. The results show the importance of using complementary techniques to correctly evaluate and interpret processing related effects for optimization of nanowire-based optoelectronic devices.


Introduction
Nanowire (NW) solar cells have the potential to reach similar or even higher efficiencies as thin-film solar cells while at the same time only using fractions of the expensive III-V semiconductor materials [1]. Furthermore, these low dimensional material building blocks enable a facile combination of materials with different lattice parameters [2,3] and the possible integration in flexible polymers after peel-off [4,5]. Due to the high research interest in NW solar cells, several reviews on this topic have been published recently [6][7][8][9].
However, the reduced dimensions of the NWs pose new challenges. One of those is the high surface-to-volume ratio that can lead to unfavorable surface recombination of charge carriers. It is thus of great importance to passivate the surface of most III-V NWs, and different passivation strategies have been presented based on wet chemical treatment [10], epitaxial shell growth [11,12] and dielectric layers [13][14][15]. Even though InP has a relatively low surface recombination velocity VSR [16], we have recently observed that using different insulation layers strongly affects device performance [17]. It is challenging to passivate InP because no obvious high bandgap, latticematched passivating material is available to be grown in-situ. Alternatively, atomic layer deposition (ALD) of amorphous oxides such as a stacked layer of POx/Al2O3 has been shown to increase the photoluminescence decay lifetime of nominally intrinsic, selective area grown wurtzite InP NWs from 1.8 to 5.4 ns [14]. On the other hand, it has been shown, that passivation is necessary but may not be sufficient to optimize NW devices, as fixed charges in a passivation layer can influence the electrostatic potential of the p-n junctions in the NW solar cells due to a field-effect [18]. A fixed positive (negative) charge at the interface between the semiconductor NW and the dielectric insulation layer increases the electron (hole) density and induces n-type (p-type) behavior. Up to now, on a single NW level, this has only been simulated [18].
Electron beam induced current (EBIC) measurements have been proven to be a successful tool to investigate p-n junctions of both single NWs [17,[19][20][21][22], and processed NW solar cells [23][24][25]. Using this technique, single NWs are contacted in a scanning electron microscope (SEM) equipped with a piezocontrolled tungsten probe, enabling to measure EBIC and IV-characteristics. Time-resolved photoluminescence (TRPL) is often used to study photovoltaic materials [26,27] and especially the surface passivation thereof [14,15]. Both EBIC and TRPL enable to obtain information about the photovoltaic quality of NWs rapidly without the need of any time-consuming processing.
In this work, we investigated photovoltaic InP NWs with p-i-n junctions by use of EBIC to determine the influence of insulating oxides on the electrostatic potential. The nanowires were measured as-grown after synthesis, and again after depositing SiOx [17] or a POx/Al2O3 stack [14]. We detected a field-effect on the electrostatic potential when the NWs were capped with POx/Al2O3. We then prepared uniformly doped nanowires using the same dopant molar fractions during synthesis as in the respective segments of p-i-n junctions to characterize and compare the mobile charge carrier lifetimes of nominally intrinsic, p-and n-type InP NWs by use of TRPL. Additionally to as-grown NWs, surface passivated NWs using the same oxides were measured. The dependence of both doping and VSR on EBIC measurements was calculated with the help of spatially resolved internal quantum efficiency (SIQE) [28] modelling. The simulations agree with the experimental data and demonstrate the influence of surface passivation and dopant concentrations on the electrostatic potential of NWs with p-i-n junctions thus helping to understand the correlation of TRPL and EBIC measurements. An EBIC signal along the full NW is desired for optimal short circuit density [17,28], but a possible change of the EBIC signal after oxide deposition has been ignored. We show that with EBIC the displacement of a p-n junction due to a field-effect can be measured. Understanding the effect of the use of different oxides on the electrostatic potential of semiconductor NWs is crucial, as many nanowire-based devices, especially NW solar cells with axially defined junctions [17,29,30], employ insulation layers to prevent electric short circuits.
Two different oxides were deposited by ALD. SiOx was deposited in a Veeco/CNT Savannah ALD chamber at 255 °C and a base pressure of 100 mbar. Tris(tert-butoxy)silanol (TTBS) and trimethylaluminium (TMAl) were used as precursors of which TTBS was heated to 90 °C. 18 cycles were deposited, and each cycle consisted of one 0.06 s TMAl pulse and three 3 s TTBS pulses. The resulting thickness of SiOx on the NWs was 30 nm and contained a low Al content [38]. The POx/Al2O3 was deposited on other samples in a Veeco/CNT Fiji ALD chamber at 50 °C at a base pressure of 80 mbar. Trimethyl phosphate (TMP) and TMAl were used as precursors. First POx was deposited by 100 cycles of 2 s TMP pulses each terminated by a 5 s O2 plasma oxidation step. Then, Al2O3 was deposited using 133 cycles of 0.08 s TMAl pulses and 5 s oxidation by use of O2 plasma. The resulting thickness of POx/Al2O3 on the NWs was 12 nm.
All of the TRPL experiments were performed in a dry nitrogen atmosphere and at room temperature. A Ti:Sapphire 770 nm 100 fs pulsed laser (Spectra-Physics, Tsunami) with a repetition rate of 81 MHz was used for sample excitation. The laser beam was focused on the sample with an angle of 70°, which is close to the Brewster angle of InP. The PL emission from samples was collected via two lenses with a focal length of 50 mm, and focused on a spectrograph (Chromex) having a 50 lines·mm −1 grating. Then, the output of the spectrograph was sent into the streak camera (Hamamatsu C6860) set at a slit width of 20 μm. In the data processing, background subtraction, shading and spectral sensitivity corrections of the PL decay images were utilized with a calibrated standard lamp (Ocean Optics, LS-1-CAL). Efforts were taken to keep the alignment of the experimental set-up the same for all the samples studied.
EBIC was measured in a Hitachi SU8010 SEM by a Point Electronic EBIC amplifier. Samples were cleaved and glued on a 90° stub such that the NW axis was perpendicular to the electron beam. A Prober Shuttle from Kleindiek Nanotechnik was used to contact single NWs with a tungsten probe. At least four NWs were measured per sample. In order to contact the gold alloy particle of the coated NWs, the oxide was scratched away at the tip of the NWs using the tungsten probe while taking care to avoid charging of the oxide layer under the electron beam of the SEM. All measurements were conducted using 5 kV acceleration voltage and a beam current in the order of tens of pA.
A three-dimensional drift-diffusion model was employed to simulate the carrier distribution and splitting process in InP nanowire. The technical equation and simulation parameters were stated in Ref. [28]. In the SIQE simulation, a cubic box with a side length of 10 nm was placed on the surface of the nanowire. In the box, an initial carrier density of 1 × 10 21 cm −3 was fixed and the current was calculated with this initial condition at a short circuit situation. The ratio between the number of carriers in the current and the number of carriers in the box is defined as SIQE which shows the ability to split electron-hole pairs at this local position. In the band diagram simulation, an oxide shell was added to the nanowire and the contact was extended from the tip of the nanowire to the sidewall. Thus the surface of the oxide shell has the same electrical potential as the top contact. The oxide was simulated as a very high bandgap semiconductor (9 eV [39]) with type I band alignment between InP and oxide shell. and 1(c) show schematics of as-grown and coated NW arrays respectively. The electrostatic potential of nanowires is strongly influenced by doping and VSR. Therefore, to deepen the understanding of EBIC measurements of capped NWs, we firstly investigated the passivation effect of SiOx and POx/Al2O3 on nominally intrinsic and intentionally doped as-grown InP NWs with TRPL. A constant dopant flow was used during synthesis, but it should be noted that the doping efficiency can vary with the length of NWs [40][41][42]. Subsequently, TRPL was measured on the same NWs coated by the two aforementioned oxides. As the passivation by POx/Al2O3 is expected to improve over time or after a thermal annealing step [14], these NW samples were thermally annealed for 1 min at 250 °C in N2 atmosphere after characterization and then re-measured.

Time-resolved photoluminescence
In order to extract the charge carrier lifetime of nominally intrinsic NWs, a bi-exponential decay was fitted to the TRPL measurements in Fig. 1(d). We observed that coating the nominally intrinsic InP NWs with any oxide reduced the charge carrier lifetime. In particular, at an excitation flux of 3 × 10 12 photons·cm −2 ·pulse −1 the charge carrier lifetime decreases from 650 ps for as-grown NWs to 507 ps for SiOx covered NWs and 474 ps for POx/Al2O3 covered NWs (see Table S1 in the Electronic Supplementary Material (ESM)). These results show significantly lower lifetimes than those reported by Black et al. [14], possibly because a different characterization tool using a different time range was studied. It should be pointed out that in this study a streak camera was used that has a much shorter (ca. two orders of magnitude) instrument response function than time-correlated single photon counting (TCSPC), allowing to resolve the dynamics of PL at short time scales, which could be completely missed by TCSPC. Indeed, when measured with TCSPC, our samples exhibit lifetimes ranging from 4 to 7 ns, as shown in Fig. S1 in the ESM.
We observed that nominally intrinsic NWs exhibit the longest lifetimes. Some strategies have been presented to produce doping-free NW solar cells based on carrier-selective contacts [43], but generally and especially for two-terminal multijunction solar cells, doping is used. Therefore, NWs with the same dopant molar fractions used in the respective segments of p-i-n junctions were grown and characterized by TRPL. The measurements of uniformly doped NWs are shown in Fig. 2. The carrier lifetime decreases when doping is introduced. Interestingly, using small amounts of DEZn for compensation doping decreases the lifetime more than heavy p-doping. Though the exact origin is unclear, one explanation could be that outdiffusing Zn forms a thin ZnO layer after exposure to air which has been reported to passivate InP [44]. The same figure includes the measurements performed of the samples with the different passivation schemes. While the lifetime of the doped NWs decreases after deposition of SiOx, both n-and p-doped nanowires covered with POx/Al2O3 have longer lifetimes than as-grown NWs (see Table S1 in the ESM). The passivation is further improved after annealing (for 1 min at 250 °C under N2 atmosphere), indicated by increased amplitudes of the longer decays.

Electron beam induced current measurements
It has been shown that fixed charges of an insulation layer can induce a radial change of the carrier density and thus affect the performance of NW solar cells [18]. In order to validate if the used oxides show a field-effect on InP NW samples, a series of InP NWs with p-i-n junctions using the same dopant molar fractions as for the uniformly doped nanowires (Fig. 2) were grown. The doping of the middle segment was The scatter plot shows the measured data points, while the solid line is a bi-exponential fit to the data. After deposition of SiOx (blue) the optical properties of as-grown InP NWs (black) degrade. After an annealing step, the POx/Al2O3 (green) improves the photoluminescence. varied by increasing the DEZn molar fraction in the middle segment from χZn = 0, to χZn = 0.3 × 10 −7 , to χZn = 2.6 × 10 −7 . This corresponds to a nominally intrinsic, compensation doped, and overcompensation doped middle segment respectively, as depicted in Figs. 3(a)-3(c). Figure 3(d) shows a representative composition image of the greyscale SEM image and EBIC measurement for an as-grown NW from Fig. 3(c).
We observe that the metallurgical p-n junction of as-grown NWs is located at the p-i interface for NWs with nominally intrinsic middle segments (Fig. 3(a)), but at the i-n interface for NWs with compensation doped middle segments (Figs. 3(b) and 3(c)) [17,21]. The latter is advantageous because due to strong light absorption in the NWs most charge carriers are generated at the tip of the NWs [28]. Generally, a flat EBIC profile is desired to maximize charge carrier generation and short-circuit current [21].
The measurements after oxide deposition reveal that the oxides have a strong effect on the peak position and width of the EBIC signal. We observe that if the EBIC profile of as-grown NWs is flat due to a compensation doped middle segment, the EBIC profile before and after SiOx deposition looks similar. For a n-as well as for a p-type middle segment, the EBIC signal in the middle segment decreases and the peak gets much sharper. The question remains whether this SiOx induces a field-effect or not since the position of the EBIC signal is not significantly moved in any direction.
The use of POx/Al2O3, on the other hand, leads to an n-type field-effect which moves the position of the junction closer towards the bottom p-segment of the NWs. The nominally intrinsic middle segment ( Fig. 3(a)) already shows n-type behavior of but the position of the EBIC signal peak is still moved even closer to the substrate because of the n-type fieldeffect on the p-segment. In the case of an overcompensated p-type middle segment (Fig. 3(c)), the field-effect of the oxide counteracts the intended doping and the EBIC signal flattens out (i.e., moving towards a the ideal p-i-n case with an uniform EBIC profile along the NW). If the middle segment is compensation doped (Fig. 3(b)), the use of POx/Al2O3 leads to a peak at the p-i interface in addition to the signal at the i-n interface. This double-peak can be explained by a field-effect caused by fixed charges in the shell. In Fig. S2 in the ESM, it can be seen how the potential of fixed charges in the oxide provokes band bending at both the p-i and i-n interface. As the EBIC signal peaks at the p-i interface we deduce that the field-effect leads to an inversion of the middle segment from slightly p-type to slightly n-type.
In order to get a better understanding of the different EBIC profiles, simulations for the SIQE of InP NW p-i-n junctions with different doping in the middle segment and varying VSR have been calculated and are shown in Fig. 4. The carrier lifetime is inversely proportional to VSR in the model used for calculations and therefore short lifetimes are represented by fast VSR.
The SIQE of Fig. 4(a) (n-type middle segment) and 4(b) (p-type middle segment) with a VSR of 10 4 cm·s −1 resembles the measured EBIC profiles of as-grown NWs from Figs. 3(a) and 3(b) respectively. Further simulations (see Fig. S3 in the ESM) show, that the measured EBIC profile of as-grown NWs from Fig. 3(c) with increased χZn cannot be explained by only increasing the dopant concentration. The measured EBIC profile is better resembled by the simulation using an increased VSR in the order of 10 5 cm·s −1 .
For nominally intrinsic NWs we did not observe any improvement on the passivation by either oxide (Fig. 1(d)) with TRPL. Accordingly, for p-i-n junctions with nominally intrinsic middle segments, both oxides decrease the EBIC signal and the profiles match with simulations for increased VSR. For compensation doped NWs, SiOx showed the shortest lifetimes in TRPL (Fig. 2). That is why the EBIC measurements after SiOx deposition are sensitive to the doping level of the middle segment as seen in Fig. 3.
After deposition of POx/Al2O3 we did not observe significantly improved lifetimes with TRPL for compensation doped NWs. Therefore, we attribute the flattened EBIC profile of NWs with a p-type middle segment in Fig. 3(c) to the field-effect of POx/Al2O3 and not to a decrease in VSR. Instead, we believe that due to fixed positive charges in the oxide the electron concentration is increased towards the interface between NW and oxide which compensates the overcompensation of Zn. Another indication for this hypothesis is that for the double peak EBIC profile with a lower χZn the highest signal is detected at the p-i interface suggesting an inversion of the middle segment from p-type to n-type.

Conclusions
By measuring TRPL on nominally intrinsic and doped InP NWs, we showed that the influence of insulation layers is doping dependent. Then, we measured EBIC on a similar set of InP NWs capped with insulation layers to characterize the effect of the dielectric layer on the electrostatic profile of the nanowires and show the importance of such layers in the performance optimization of NW solar cells. Indeed, a field-effect can be induced by embedding the nanowires in the oxides, that can change the electrostatic potential of the NWs and even invert the polarity due to fixed charges in the oxide layer. This is especially the case for POx/Al2O3 due to the n-type field effect as indicated by the EBIC measurements. All EBIC measurements Figure 3 Representative EBIC profiles along the NW axes including the Au particle. The DEZn molar fraction in the middle segment is increased from (a) χZn = 0, to (b) χZn = 0.3 × 10 −7 to (c) χZn = 2.6 × 10 −7 . POx/Al2O3 induces an n-type field-effect that moves the peak position of the EBIC signal for all samples. SiOx decreases the EBIC signal in the middle segments, but increases the EBIC signal in the p-segment of the NWs. (d) Composition image of SEM (greyscale) and EBIC (colorbar) signal of the as-grown NWs from (c). could be reproduced and explained by simulated SIQE for varied dopant concentrations and VSR. By comparing the simulations to the measured EBIC profiles and considering the passivation properties obtained from TRPL, we demonstrate that neither a change in surface passivation nor a change in dopant concentration alone can explain the measurements. Instead, both effects have to be recognized and kept in mind for device optimization.
Electronic Supplementary Material: Supplementary material (fitting parameters, band diagram after field-effect, and additional SIQE profiles) is available in the online version of this article at https://doi.org/10.1007/s12274-021-3344-9.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.