Sn-seeded GaAs nanowires grown by MOVPE

It has previously been reported that in situ formed Sn nanoparticles can successfully initiate GaAs nanowire growth with a self-assembled radial p–n junction composed of a Sn-doped n-type core and a C-doped p-type shell. In this paper, we investigate the effect of fundamental growth parameters on the morphology and crystal structure of Sn-seeded GaAs nanowires. We show that growth can be achieved in a broad temperature window by changing the TMGa precursor flow simultaneously with decreasing temperature to prevent nanowire kinking at low temperatures. We find that changes in the supply of both AsH3 and TMGa can lead to nanowire kinking and that the formation of twin planes is closely related to a low V/III ratio. From PL results, we observe an increase of the average luminescence energy induced by heavy doping which shifts the Fermi level into the conduction band. Furthermore, the doping level of Sn and C is dependent on both the temperature and the V/III ratio. These results indicate that using Sn as the seed particle for nanowire growth is quite different from traditionally used Au in for example growth conditions and resulting nanowire properties. Thus, it is very interesting to explore alternative metal seed particles with controllable introduction of other impurities.


Introduction
Epitaxial synthesis of III-V nanowires has almost exclusively used Au as a catalyst particle since III-V nanowires were demonstrated by Hiruma et al [1]. To implement this promising one-dimensional nanostructure with its outstanding properties like direct band gap and high carrier mobility into electronic and optical devices, it needs to be compatible with the existing semiconductor technology. However, gold severely limits the possibility to integrate these building blocks into, for example, conventional Si industry since gold can easily diffuse into silicon creating mid-gap electronic states, which will degrade device performance drastically [2].
There has been much research of gold-free nanowire growth methods including selective area, self-seeded, and non-gold foreign metal seeded nanowire growth [3][4][5][6][7]. In the case of selective area nanowire growth, a pre-defined opening pattern is used to initiate the nanowire growth and no seed particle is required, but one-dimensional epitaxy strongly relies on the selected growth conditions to enable anisotropic growth rates. Thus, it usually has a rather narrow growth window [8]. For self-seeded nanowire growth, one of the elements that composes the nanowire acts as the seed material; usually Ga and In droplets are formed and used due to their low melting points. However, it is challenging to apply this method simultaneously in various nanowire material systems to achieve for example axial heterostructures where Original content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. the group III species is switched to induce the change in the material system.
So far reports on III-V nanowires seeded from metal particles different from gold are limited. Thus, exploring alternative foreign seed particle materials other than gold is of great interest. It may also be interesting for device applications to incorporate impurities from suitable seeding material to achieve for example self-doping without the use of extra dopants. By investigating the nanowire growth behavior from different seed particles, it may also give us some insights into why gold can be used to initiate the nanowire growth of any III-V material system in a wide growth parameter window and allows crystallographic manipulation of the nanowires. It is well documented that Sn seed particles have been used for group IV, Si and Ge nanowire synthesis by different growth techniques [9,10]. We have previously demonstrated the use of in situ formed Sn particles for GaAs nanowire growth, resulting in nanowires with a self-assembled p-n junction having a Sn-doped n-type core and a C-doped p-type shell [11]. The n-doping of the core is due to the consumption and incorporation of the Sn seed particle into the nanowire during growth, while the p-doping of the shell is attributed to the incorporation of C into the radial overgrowth from decomposition of metal organic precursors. Our first investigation suggested that Sn-seeded GaAs nanowire growth was successful only in a quite narrow temperature window; straight nanowires were observed between 487°C and 525°C. At temperatures lower than 487°C, nanowires were kinked and at temperatures higher than 525°C we did not observe nanowire growth. Transmission electron microscopy (TEM) and x-ray energy dispersive spectroscopy analysis revealed a zinc blende crystal structure of the nanowires and primarily Sn in the seed particle of all investigated nanowires with a weak but detectable Sn signal also along the nanowire.
In order to further develop the growth, it is important to understand how morphology and crystal structure of the nanowires are related to both temperature and V/III precursor ratio. In this paper, we present a more detailed and systematic study on the aspects of fundamental growth parameters for Sn-seeded GaAs nanowires. We demonstrate how the V/III ratio, obtained by changing either group III or group V precursor flows, allows to control nanowire crystal structure and nanowire kinking. We find that a lower V/III ratio can lead to twin plane formation. In addition, it is interesting that we observe a high ratio of nanowire kinking at both high group V flow and high group III flow at identical growth temperature. The effect of temperature on nanowire growth is related to the actual amount of available group III material, which is not surprising as the latter is suggested to be temperature-dependent. Furthermore, photoluminescence measurements were carried out on selected nanowires to investigate the simultaneous 'seed-doping' from the Sn particles where we find that the doping level is very high. We also find that carbon is incorporated in a temperaturedependent way.

Experimental methods
GaAs nanowires were grown on (111)B GaAs substrates in a low-pressure horizontal metal organic vapor phase epitaxy reactor (Aixtron 200/4) at a pressure of 100 mbar, a total flow of 13 l min −1 with H 2 as carrier gas as well as in an Aixtron 3×2″ close coupled showerhead (CCS) reactor at a pressure of 100 mbar and a total flow of 8 l min −1 . Most of the experiments were carried out in the low-pressure horizontal reactor unless otherwise specified. Trimethylgallium (TMGa) was used as group III precursor and AsH 3 as group V precursor. The GaAs substrates were first heated to an annealing temperature of 630°C in AsH 3 ambient with a molar fraction of 1.54×10 −3 . After 10 min annealing the temperature was ramped down to a set value of 550°C. Once the set temperature was reached, tetraethyltin (TESn) with a molar fraction of 1.17×10 −5 was introduced into the reactor for a 15 min Sn particle formation step with an AsH 3 background of 7.68×10 −5 . Then, TESn was turned off and the reactor temperature was set to the desired growth temperature. Once the temperature stabilized, TMGa was turned on and AsH 3 was set to the desired molar fraction for the nanowire growth. The particle deposition step was kept the same with an average diameter of 59±7 nm and a density of 11±2 μm −2 which is identical to the conditions as reported in [11]. The growth temperature and the flows of group V and group III precursors were varied to determine the effect of these parameters on the resulting nanowire growth. The ranges of molar fractions that group V and group III precursors were varied are from 7.68×10 −5 to 2.30×10 −4 and from 2.01×10 −5 to 1.04×10 −4 , respectively. GaAs nanowires were grown for 10 min unless otherwise stated and cooled down in AsH 3 ambient. Ex situ characterization of nanowire morphology and crystal structure was performed by scanning electron microscopy (SEM: Hitachi SU8010 operated at 15 kV) and transmission electron microscopy (TEM: JEOL 3000F operated at 300 kV). Low temperature (7 K) photoluminescence spectroscopy (PL) measurements were performed on single nanowires which were broken off from the native growth substrate and transferred onto a gold-covered Si substrate which had markers for easy location of the nanowires. The gold cover also slightly enhanced the emission intensity. A frequency doubled yttrium-aluminum-garnet-laser was employed for excitation of the nanowires and an optical microscope to collect the luminescence, which was dispersed onto a thermoelectrically cooled charge-coupled device. Typically five single nanowires of each sample were characterized.

Results and discussion
3.1. Dependence of nanowire growth on V/III ratio changes: TMGa precursor flow variation We have studied the effect of TMGa precursor flow on the GaAs nanowire morphology and crystal structure. We will start with discussing the impact of TMGa molar fraction variation from 2.01×10 −5 to 1.04×10 −4 while keeping the AsH 3 flow constant at a molar fraction of 7.68×10 −5 resulting in V/III ratios ranging from 3.81 to 0.74. These experiments were carried out at 500°C, which was found in our previous study to be the optimum growth temperature giving vertically aligned zinc blende GaAs nanowires with the most uniform morphology. A variation of the V/III ratio by changing the TMGa precursor flow has a strong effect on the nanowire morphology, especially the length and crystal structure are affected. Figure 1 shows SEM images of the nanowires grown with increasing TMGa flows, thus decreasing V/III ratio. As illustrated in figures 1(a)-(e), nanowires all grow vertically with increasing TMGa precursor flow, in other words decreasing the V/III ratio from 3.81 to 0.95. The length of the nanowires is strongly dependent on the TMGa precursor flow (see also supporting information S1) and higher V/III ratios result in shorter nanowires. This indicates that the length of the nanowire mainly depends on the flow of group III precursor. This is in agreement with what has been reported for gold-seeded GaAs nanowires in which the length of the nanowire is linearly or sublinearly dependent on the flow of TMGa precursor [12].
With a further increase of TMGa flow leading to a V/III ratio of 0.74, the nanowires start to kink instead of growing vertically from the substrate as illustrated in figure 1(f). This will be discussed more in detail later. It is interesting to note that the density of nucleated vertical nanowires is identical to the density of pre-deposited Sn droplets (for all samples exhibiting predominantly vertical nanowire growth). However, it is quite difficult to tell the density for the kinked nanowires since they have very irregular and random morphology which makes it hard to determine whether they are successfully nucleated or not. To a first approximation, the density of identifiable crystallites seems consistent with the density of Sn droplets.
Another interesting phenomenon observed is that when a high flow of TMGa is supplied, the nanowire side facets appear rough which is noted from the alternating contrast at the side facet of the nanowire, as shown in the inset of figure 1(e), instead of rather smooth side facets as has been observed at lower TMGa flows (figures 1(a)-(d)). These microfacets are correlated with multiple twin planes along the nanowires growth axis [13,14]. Note that in figure 1(d), one can also see that the side facets are not perfectly smooth in the nanowires; instead there are one or two facet rotations, associated with twin planes, in some of the nanowires. This clearly indicates that the twin formation in Sn-seeded GaAs nanowires is progressively interconnected with lowered V/III ratio [15]. Previously we reported that Sn-seeded nanowires mostly have a pure zinc blende crystal structure with {112} A side facets at optimized growth conditions [11]. Similarly to what has been demonstrated for Au-seeded GaAs nanowires, the {112}-type sidewalls usually have anisotropic growth rates as a result of different polarities [16]. Here, the finally observed {112} A side facets imply that {¯¯¯} 112 B planes grow faster instead of the {112} A ones as for the Au-seeded case reported in [16]. {112}-type planes are partially polar (being constructed of microfacets of the {111}and {100}-type respectively), indicating that {112} A planes have more dangling bonds available on Ga atoms, while {¯¯¯} 112 B planes have more As atoms with non-saturated bonds. Considering our very low V/III ratio growth condition outlined here, relatively more Ga species should be available in this environment which will quickly bind to the accessible As spots. Consequently the As dominant {¯¯¯} 112 B planes would grow much faster causing them to outgrow and finally leaving nanowires with only {112} A side facets. Or in other words, the surface energy of {112} A facets is significantly lower compared to the {¯¯¯} 112 B counterparts under the applied low V/III ratio growth conditions which makes a fast overrespective outgrowth of the latter highly probable resulting in {112} A -terminated nanowires. A similar V/III-ratio dependence of surface energy of polar III-V surfaces in GaAs has been calculated and reported as a function of the chemical potential in [17].
However, in the case of frequently occurring rotational twinning along the nanowire growth axis the geometrical/ morphological result of overgrowth processes is slightly different which will be discussed in the following. From TEM images in figure 2 (which represents the analysis of a single nanowire from the sample shown in figure 1(e)), we find a zinc blende crystal structure with a high number of semiperiodic rotational twins (figures 2(a)-(c)). The spacing of the twin planes, as shown by the conventional dark field images in figure 2(c), correlates to a good degree with the features observed as rough facets in figure 1(e), meaning that the rough side facets in the SEM image are associated with the multiple twin planes here. Unlike Au-seeded GaAs nanowires which usually have short and dense twin segments [18][19][20], here we can see that the length of each twin segment can be up to 60 nm. There is also distinct radial overgrowth on the alternating, twinned zinc blende segments as highlighted by the conventional dark field images in figures 2(e)-(g). Here we see brighter contrast occurring at similar positions on every other zinc blende segment and alternated (60°rotated) on consecutive segments when using (111)-related twincharacteristic diffractions spots which is especially visible in figure 2(g). A possible explanation for the overgrowth is that in the early stage of a twin formation, the nanowires primarily possess {111}-type facets. Whenever a twin plane appears, the segments are rotated by 60°with respect to each other around the¯¯á ñ 111 B growth axis. Thus, the {111} A facets are meeting {¯¯¯} 111 B facets at alternating twin planes in the axial direction with the global tendency to finally form{112}-type facets as discussed above. However, as we also discussed above {¯¯¯} 112 B facets preferably grow faster in order to attain {112} A side facets on both segments separated by a twin plane. From a purely geometrical perspective this overgrowth would result in triangular cross sectional segments stacked along the¯¯á ñ 111 B growth axis, terminated by 3 {112} A -type facets, and with every other segment rotated by 60°with respect to each other. Although this situation is not realistic, we do still observe the alternating overgrowth on the zinc blende segments but not with the ultimate geometrical consequences as described in the previous sentence.

Dependence of nanowire growth on V/III ratio changes: AsH 3 precursor flow variation
We also investigated the effect of AsH 3 precursor flow on GaAs nanowire growth, in which the flow of TMGa precursor was set to a constant molar fraction of 4.03×10 −5 at 500°C, while varying the AsH 3 molar fraction from 7.68×10 −5 to 2.30×10 −4 , corresponding to V/III ratios between 1.91 and 5.71. As illustrated in figure 3(a) and the inset, at a low V/III ratio the nanowires have rough side facets, which are associated with twin planes perpendicular to the growth direction. This correlates to what we have observed for the nanowires grown at high TMGa supply ( figure 1(e)), also resulting in a low V/III ratio. It seems that the formation of twin planes can be associated with a low V/III ratio but not with absolute precursor flows within the investigated flow regime. There is no obvious relation between the lengths of the nanowires with AsH 3 flow (see also supporting information S1). Moreover, a distinct trend in the rate of kinking nanowires with increasing AsH 3 flow is observed from figure 3. At a critical V/III ratio of 3.34, around 6% of the nanowires start to kink into a nonvertical growth direction (see supporting information S2). An even larger amount of kinked nanowires occurs with increasing AsH 3 flow and thus at higher V/III ratio until eventually at a V/III ratio of 5.71 no straight, vertically aligned nanowires are observed. And we expect with V/III ratios higher than 6, there will not be any successful growth of straight and vertically aligned nanowires at this temperature and instead all nanowires will be kinked. This kinking effect is consistent with what has been reported for Au-seeded GaAs nanowires grown at high V/III ratios [15]. One possible reason behind this is that high AsH 3 flows could lead to a surface reconstruction on GaAs substrate [21] which might assist another growth direction other than¯¯á ñ 111 B causing nanowire kinking. On the other hand, it is surprising that kinking also occurs at high TMGa flows as shown in figure 1(f). It was demonstrated previously that in III-V/IV heterostructure system, the particle volume, the contact angle of the droplet particle, and the match between particle and nanowire growth interface are fundamental factors in determining the nanowire morphology, more specifically nanowire kinking [22]. We have previously reported that from EDX analysis on our Sn-seeded GaAs nanowires, there is a substantial fraction of As in the seed particle (see also figure 2(i) here) in contrast to the case of gold [11]. The composition of the Sn particle is very likely to be a dynamic state during nanowire growth. Thus, excessive group V or group III precursor availability could lead to a volume change in the Sn seed particle and also changes of the contact angle. These two extreme cases could both lead to nanowire kinking, however, from different mechanisms. Indeed, in figure 1(f) (high TMGa), the shape and diameter of the nanowires change substantially, the nanowires lost its shape and become like a cluster, despite the fact that they are also kinking. However, in figure 3(f) (high AsH 3 ), it seems that the nanowires still grow straight after kinking happens in an early stage (see also supporting information S3). Thus, it is possible that the kinking of these Sn-seeded GaAs nanowires originates from two different mechanisms for the two regimes (high TMGa versus high AsH 3 ).
As discussed above, Sn-seeded GaAs has a quite low suitable V/III ratio for the growth compared to Au-seeded reference growth, and the nanowires have either pure zinc blende crystal structure or zinc blende with a few rotationally twinned segments. No wurtzite crystal structure nanowires, not even short segments of wurtzite have been observed so far in the growth range investigated. In the case of Au-seeded nanowire growth, the V/III ratio can be up to 240 and denser twin planes or even twin plane superlattices can be observed under certain growth conditions [23,24]. Wurtzite structure can also be achieved at selected growth conditions. However, nanowires seeded from both Sn and Au nanoparticles follow the same trend with different V/III ratio, which is the density of twin defects is closely related to lower V/III ratio and the formation of twinning can be prevented by either increasing AsH 3 precursor flow or decreasing the TMGa precursor flow [15].

Broadening the growth temperature window
In our previous investigation of Sn-seeded GaAs nanowires, a temperature range from 475°C to 535°C was used for the GaAs nanowire growth at a V/III ratio of 1.9 where straight vertically aligned nanowires were only observed in a limited range, from 487°C to 525°C. This is narrower compared to Au-seeded GaAs nanowire growth, which usually occurs within the range from 380°C to 540°C [23,25].
For Sn-seeded GaAs nanowires we previously found that there was barely any nanowire nucleation and the vapor-solid surface growth was dominant at a temperature of 535°C and we expected the same for temperatures above that. Conversely most of the nanowires were still growing but just kinked when the temperature was lower than 487°C. In addition it is well documented that for Au-seeded GaAs nanowire growth the radial overgrowth or tapering is substantially minimized at lower temperature [24,25]. This makes it interesting to explore further the nanowire growth in the low temperature range to improve the yield of straight versus kinked nanowires. At low temperatures, the resulting nanowire kinking and growth along other directions different from¯¯á ñ 111 B appeared similar to the case with a high AsH 3 supply as shown in figure 3(f) [11].
It was previously shown that Sn-seeded GaAs nanowire kinking could be prevented by a two-step approach consisting of a nucleation step at high temperature and then cooling down to a lower temperature for the growth [11]. This implies that there could be a similar nucleation problem for Sn-seeded GaAs nanowires at low temperatures following the same trend as reported for Au-seeded nanowires [26]. Considering the fact that changes in temperature will affect the decomposition rate of precursors it is likely that the actual V/III ratio at the growth front will change with temperature even for constant precursor flows. To investigate whether the kinking of nanowires observed at low temperatures is associated to a relatively higher effective V/III ratio, another series of experiments was carried out at a temperature of 455°C with increasing TMGa flow while holding the AsH 3 flow constant at 7.31×10 −5 . As can be seen from SEM images in figure 4, there is less tapering of the resulting nanowires due to the lower growth temperature and the proportion of vertical nanowires increases from around 13% to almost 100% with increasing TMGa supply. From this we can conclude that AsH 3 is closely related to nanowire kinking and increasing the TMGa supply at low temperature can be used to counterbalance this effect of AsH 3 resulting in straight vertically aligned nanowires. One possible reason for this could be that As trimers form on the substrate surface at low temperature with a higher effective V/III ratio leading to an As-rich environment interfering with nanowire nucleation. We have also discussed previously that two extreme regimes could cause nanowire kinking with either excessive Ga or As in the Sn particle during growth. Hence, another possible reason could be that abundant As would accumulate in Sn seed particles at low temperature which would result in kinked nanowires, but adding more TMGa could again balance the composition in the Sn particle and prevent nanowire kinking.
Based on this observation and to further optimize the growth result, the experiments were transferred to an Aixtron CCS 3×2″ reactor in which higher TMGa flows could be attained technically. We find that the temperature window can be broadened to as low as 400°C by stepwise increasing the TMGa with decreasing temperature. As shown in figure 5, at 'optimized' conditions of 420°C with a V/III ratio of 0.86, the radial shell overgrowth can be significantly reduced due to the lower growth temperature used. Also worth noting is that the axial nanowire growth rate decreases remarkably with decreasing temperature and here the nanowire growth time is 30 min. From SEM and TEM studies presented in figure 5, we find that the nanowires exhibit zinc blende crystal structure with few rotational twins. From EDX spectra we further find detectable signals from both As and Ga in the Sn seed particle. A weak but detectable Sn signal can also be measured at the GaAs nanowire body which is consistent with what we previously reported for nanowires grown at higher temperatures [11]. Since the shell growth was previously found to be highly p-doped by C incorporation, the possibility to tune the growth to either include or eliminate radial overgrowth makes it potentially very interesting for device applications.

Optical properties
We have investigated selected nanowire samples to represent characteristic points in a growth series by photoluminescence (PL) spectroscopy to study the effect of for example temperature. Figure 6 shows the low temperature PL spectra of the nanowires grown at different V/III ratios at a constant growth temperature (figure 6(a)) and at different temperatures, but constant V/III ratio ( figure 6(b)). The spectra are normalized and plotted with an offset for clarity. We observed a strong Burstein-Moss shift, which we attribute to the highly n-type  doped core. The heavy doping shifts the Fermi level into the conduction band, which increases the average recombination energy and broadens the spectra. It should be noted that the PL did not show any noticeable changes with excitation power density apart from the intensity. In order to acquire a rough estimate of the free carrier concentration we used the Fermi-tail fitting method [27], taking into consideration changes in the effective carrier mass and the bandgap renormalization. The carrier concentration is on the order of 10 19 cm −3 for all the samples (figure 6(a)), which is consistent with our estimation in the previous study that according to the consumption of Sn nanoparticle during growth, the doping density can be up to 10 20 cm −3 [11]. Calculations of carrier concentration for nanowires grown at different V/III ratios show a trend of Sn incorporation which is more favorable on the As sites with decreasing V/III ratio.
Our previous studies of GaAs nanowires grown at 500°C with a V/III ratio of 1.91 showed strong Esaki diode behavior, with the C-doped p-type overgrowth (the shell) and the heavily Sn-doped n-type core [11]. This complicates the analysis further making an interpretation of the spectra not trivial. The low energy shoulder at 1.48 eV can be caused by crystal defects, such as rotational twins, where spatially indirect recombination takes place [20]. However, PL measurements show decreased intensity of the low energy shoulder with increasing temperature ( figure 6(b)), despite the higher density of rotational twins. This makes the crystal defects very unlikely to be the origin of the low energy emission in our nanowires. As an alternative explanation we speculate that it is related to carbon impurities in the p-doped shell. This suggests that carbon incorporation decreases with increasing V/III ratio as well as with increasing growth temperature, both of which are consistent with layer growth studies [28,29]. It can also be related to other point defects incorporated during the growth. The high energy peak cannot be assigned to any optical transitions involving band-to-band transitions. We speculate, that it may be related to states created by Sn below the conduction band at the L point. The L point has energy of 1.815 eV where the valence band maximum is at 0 eV.

Summary and conclusion
In summary, we have investigated the growth of Sn-seeded GaAs nanowires in detail by varying the temperature as well as group III and group V precursor flows to further understand the nanowire growth mechanism. The growth rate of nanowires is strongly dependent on group III precursor supply (TMGa). High amounts of either TMGa or AsH 3 flows result in nanowire kinking, suggested to be caused by different mechanisms due to the composition change in the Sn seed droplet during nanowire growth. Moreover the twin plane formation in nanowires is found to be closely related to V/III ratio and happens at low V/III ratio regardless of whether group V or group III precursor is changed. We demonstrate that vertical GaAs nanowires with reduced shell growth can be obtained in a broadened temperature range by adjusting the TMGa precursor flow together with decreasing temperature, which could be attributed to a different effective V/III ratio involved in the nanowire growth at lower temperature. Furthermore, PL results reveal the temperature and V/III dependence of carbon incorporation related doping and confirm the high level of Sn-doping in the nanowires. Our results could be used in order to grow highly n-type GaAs nanowires without the use of Au seed particles or addition of extra dopants for future device applications. (a) Spectra taken from nanowires grown at 500°C but different V/III ratio. (b) PL spectra of nanowires grown at constant V/III ratio at 1.91 but different temperatures. Broad spectra as well as blue-shifting with respect to the bandgap energy of GaAs indicates a high doping level as well as a high concentration of defects. Black spectra is smoothed and grey is raw data.