Sn-Seeded GaAs Nanowires as Self-Assembled Radial p-n Junctions.

The widespread use of Au as a seed particle in the fabrication of semiconductor nanowires presents a fundamental limitation to the potential incorporation of such nanostructures into electronic devices. Although several other growth techniques have been demonstrated, the use of alternative seed particle metals remains an underexplored but potentially very promising way to influence the properties of the resulting nanowires while simultaneously avoiding gold. In this Letter, we demonstrate the use of Sn as a seed particle metal for GaAs nanowires grown by metal-organic vapor phase epitaxy. We show that vertically aligned and stacking defect-free GaAs nanowires can be grown with very high yield. The resulting nanowires exhibit Esaki diode behavior, attributed to very high n-doping of the nanowire core with Sn, and simultaneous C-doping of the radial overgrowth. These results demonstrate that the use of alternative seed particle metals is a potentially important area to explore for developing nanowire materials with controlled material properties.

Au is widely used as a seed particle in the growth of semiconductor nanowires. However, the incompatibility of Au with for example conventional silicon semiconductor processing has led to numerous efforts to find alternative techniques for nanowire growth.[1] Most common among these are self-seeded[2,3] and selective-area[4,5] nanowire growth, both of which avoid the introduction of any foreign catalyst material but have relatively narrow parameter space for tuning the resulting nanowire properties. An interesting alternative is to use a different foreign metal as a seed particle, specifically selected to yield properties of the nanowire different from those achieved with gold or alternative techniques. New properties may mean for example crystal structures or directions not observed with conventional growth techniques but may also mean designed material properties directly caused by incorporation of trace amounts of the foreign metal. This has been demonstrated for example in Al-seeded Si nanowires, where very high doping levels could be achieved by incorporation of Al atoms during Si nanowire growth.[6] For III–V materials, there have to date been very few examples of seeded epitaxial nanowire growth using alternative seed metals,[7−10] unlike Si and Ge nanowires, for which alternative seed particles have been extensively reported.[1] The potential for designing exotic materials is as yet virtually unexplored for III–V materials, with the notable exception of works by Martelli et al.[11] and Jabeen et al.,[12] who used Mn seed particles to induce magnetic properties in III–V nanowires.

In this Letter, we report the use of Sn as a seed material for GaAs nanowires grown by metal–organic vapor phase epitaxy (MOVPE). Sn is a potentially interesting material due to its low melting point (232 °C) and high boiling point (2602 °C), which ensure that it can form liquid droplets over a wide potential growth temperature window without relying on the formation of specific low-temperature alloys (allowing for so-called vapor liquid solid, VLS, growth). In this way, it differs from nearly all other demonstrated alternative seed particle metals, which typically remain solid during nanowire growth (so-called vapor–solid–solid, VSS, growth).[1] Although VSS introduces certain possible advantages (such as potentially sharper heterojunction interfaces[13]), vertical alignment of the resulting nanowires is often more difficult due to the lack of a flat interface between the particle and the substrate.[1] Sn is therefore an interesting test case for exploring and developing the potential of alternative VLS seed metals.

Sn is a group-IV element and does not cause midgap electronic states in Si (making it potentially compatible with semiconductor processing). Several reports exist that demonstrate the use of Sn droplets for growth of Si nanowires using a variety of techniques.[14−20] In III–V materials, it acts as a dopant; while in principle amphoteric, it is reported to act only as an n-dopant in GaAs.[21,22] This makes it an advantageous choice compared to other group-IV dopants, as higher doping concentrations can in principle be achieved. However, doping of seeded nanowires remains challenging, and novel techniques for obtaining selected doping profiles are highly desirable.[23] In particular, n-doping of GaAs nanowires has proven difficult,[24] with most n-dopants incorporating preferentially in radial overgrowth rather than into the nanowire itself, potentially limiting the flexibility of the obtained structures. Successful Sn-doping of GaAs nanowires has been reported using tetraethyltin (TESn) as an in situ dopant element during Au-seeded nanowire growth.[25] The authors report moderate doping levels with a narrow growth window, which they attribute to the difficulty for Sn atoms to pass through the Au seed particle. Here, we show that the use of Sn seed particles for GaAs nanowire growth results in high levels of self-doping; more remarkably, the nanowires show Esaki diode behavior with very high current levels. It is deduced that the nanowires exhibit self-assembled radial p–n junctions. The doping profile and nanowire dimensions are consistent with those predicted to be suitable for photovoltaics,[26] which make these structures potentially interesting for applications of this type. The Sn particles are formed in situ using TESn, which is a widely available doping source in conventional MOVPE. The resulting nanowires grown from Sn particles show epitaxial alignment with the substrate and pure zinc blende (ZB) crystal structure (free of stacking defects) under growth conditions that give high densities of stacking defects in Au-seeded GaAs nanowires.

Nanowire growth is performed using GaAs (111)B substrates placed in a horizontal-flow MOVPE reactor (Aixtron 200/4) for growth at 10 kPa in a hydrogen carrier gas flow of 13 L/min. Substrates are first heated to an annealing temperature of 630 °C in arsine (AsH3) background with molar fraction 1.54 × 10–3. Following 10 min annealing, samples are cooled to 550 °C in AsH3. Once the temperature has stabilized, AsH3 is changed to 7.69 × 10–5, and a TESn flow with molar fraction 1.17 × 10–5 is introduced. Precision in mass flow is attained using Epison concentration measurements together with mass flow controllers. Note that two different mass flow controllers are used to cover the large flow range required for AsH3. Following a 15 min deposition unless specified otherwise, TESn is turned off, and the samples are cooled to the desired growth temperature (in the range 475–535 °C) in AsH3 ambient. Note that reactor temperature is calibrated using a LayTec setup for optical reflectance measurements of the Si–Al eutectic. Once growth temperature is attained, trimethylgallium (TMGa) is turned on with a molar fraction of 4.29 × 10–5. Nanowires are grown for 10 min unless otherwise stated. The nanowires here are grown with a V/III ratio of 1.9, which is rather low compared to Au-seeded GaAs nanowires grown in MOVPE. However, we observed no growth of nanowires for V/III ratio of more than 4. It is at present unclear why such a low V/III ratio is necessary and how this is related to other parameters such as temperature. It should be noted that the use of a very low V/III ratio means the process used here resembles self-seeded growth.[2,3]

Morphology is characterized by scanning electron microscopy (SEM; Hitachi SU8010 at 15 kV); structure is characterized with transmission electron microscopy (TEM; JEOL 3000F at 300 kV); and composition is characterized by X-ray energy dispersive spectroscopy (EDX) in high-angle annular dark-field scanning TEM (HAADF-STEM) mode. Samples are prepared for TEM analysis by mechanically breaking the nanowires near the base onto carbon film-coated Cu grids.

To confirm that the introduction of a high flow of TESn prior to nanowire growth results in Sn droplets on the surface, the sample shown in Figure 1, panel a was cooled down directly after a 15 min TESn deposition step without introducing TMGa. As shown, particles are formed on the surface with an average diameter of 59 ± 7 nm and surface density of 11 ± 2/ μm2. This size distribution is surprisingly narrow; it is for example much narrower than nanoparticles achieved using annealing of evaporated metal thin films and only moderately wider than colloidal or aerosol nanoparticles.[27] A narrow size distribution among metal droplets can be attributed to Ostwald ripening and is an indication that the surface diffusivity of Sn on these substrates is high at the temperatures used. Figure 1, panels b and c show a top view and 30° tilted view of a single nanowire grown from particles formed for 15 min (as in Figure 1a). The tips of the nanowires exhibit truncated spherical caps of average diameter 54 ± 7 nm, which may provisionally be identified as Sn seed particles. It is clear that the nanowires are relatively tapered toward the base, with smooth side facets and a triangular cross-section. Tapering may be an indication of radial growth in parallel with the Sn-seeded axial growth but may also be an indication that the Sn droplet is shrinking in diameter as the nanowire grows. Both of these possibilities will be discussed further. To determine the effect of TESn deposition time on nanowire growth, a series was grown with varied TESn deposition of 5 min, 10 min, 15 min, and 30 min (Figure 1d–g; nanowires were grown for 10 min in all cases). The nanowires grow homogeneously across the substrate in all cases. It is clear that, first, shorter deposition times lead to smaller Sn droplets and narrower nanowire diameter, and second, that nanowires grown following shorter deposition times have a higher growth rate. In particular, nanowires grown from 5 min particle deposition have a much higher aspect ratio than those grown following 15 min of Sn deposition.

Figure 1

SEM images of (a) Sn particles formed on the substrate by exposure to TESn at 550 °C. (b) 30° tilted magnified view of a single nanowire grown from 15 min Sn particles and (c) top view of the same nanowire showing a triangular cross-section. 30° tilted view of GaAs nanowires grown from Sn particles deposited for (d) 5 min, (e) 10 min, (f) 15 min, and (g) 30 min. Scale bars are 500 nm unless otherwise specified.

To investigate the crystal structure of the nanowires and confirm the Sn-seeding mechanism, nanowires were next analyzed by TEM and EDX (Figure 2). Figure 2, panels b and c show TEM images of nanowires grown at the same conditions as in Figure 1, panel b, imaged in the ⟨1̅01⟩ zone axes. The nanowires exhibit ZB crystal structure without any stacking defects along the entire length of the nanowire. This is somewhat surprising since gold-seeded nanowires grown under identical conditions exhibit high densities of stacking defects. Defect-free ZB structure in GaAs nanowires is typically associated with either very low temperature,[28] high V/III ratio,[29,30] or high total precursor flows.[31] Clearly, the selection of seed particle material strongly influences the resulting crystal structure. The nanowires side facet orientation is identified as {112}A by comparison of atomically resolved images and diffraction patterns with the HAADF intensity profiles (Figure 2a,c,d). The composition of the nanowires and seed particle were also investigated by EDX. The seed particle was found to consist primarily of Sn, with an average As content of up to 18 atomic % (and negligible postgrowth Ga content). Since the nanowires are cooled in AsH3 and the melting point of Sn lies below the growth temperature, the phase observed ex situ clearly is not indicative of the phase during growth. However, the presence of a Sn-rich particle confirms the Sn-seeded mechanism. A weak Sn peak is also observed in the EDX spectrum of the nanowire itself (see Figure 2e), which suggests Sn incorporation into the nanowire. This specific EDX spectrum was taken from the middle part the nanowire; EDX spectra from different parts of the nanowire are shown in Supporting Information S1. Since the sensitivity of the detection technique is on the order of 1 atomic %, it is difficult to accurately quantify the observed Sn content. However, to be detected at all levels, it must be very high. If this amount of Sn is incorporated into the nanowire, the doping level would be very high; however, such high Sn-doping of GaAs has been reported.[32] On the other hand, it may also be that there is Sn present on the surface of the nanowire or in clusters that are not electrically active. Although we have not observed these effects, we cannot definitively rule them out.

Figure 2

TEM analysis of a Sn-seeded GaAs nanowire, imaged in the ⟨1̅01⟩ zone axes. The diffraction pattern in panel a shows a pure ZB structure, which is also seen in the high-resolution TEM image in panel c. From the HAADF-STEM image in panel b, an (d) intensity line profile from the dashed area is used, in combination with panels a and c, to deduce the facet indices of the wire. The intensity line profile shows a triangular cross-section with one facet parallel to the beam. In panel e, two EDS spectra are overlaid: blue spectrum from the seed particle, clearly showing Sn- and As-related peaks, and red spectrum from the same area as the dashed box in panel b, showing a weak but detectable Sn signal.

The effect of temperature on the nanowire morphology is shown in Figure 3, panels a–f. At the lowest temperature investigated (475 °C), the nanowires exhibit a high probability of kinking, while nanowires grown at or above 487 °C exhibit perfect vertical alignment with a significant reduction in length with increasing temperature. Additionally, at higher temperatures, the nanowires exhibit rough side facets (Figure 3e), which TEM analyses show are associated with twinning (Figure 3g–j). It is interesting to note that, with the exception of occasional twin planes at high temperature, the nanowires are pure ZB over the entire temperature window, which indicates that the parameter window for obtaining defect-free crystal structure is much wider than for nanowires grown with gold.[28−31] It is difficult to determine the reason for the very different crystal structure observed for Sn- and Au-seeded nanowires in the absence of comparative data from other seed nanoparticle materials, although we can speculate that differences in local supersaturation (due to different precursor solubility) or different interfacial energetics are likely to play a role. When the growth temperature is set to values higher than 535 °C, no nanowire growth is observed (Figure 3f). The temperature sensitivity of the growth system is on the order of 5 °C.

Figure 3

(a–f) 30° tilted SEM images of Sn-seeded GaAs nanowires grown at different temperatures with magnified single nanowire as insets in which scale bars are 100 nm. (g) Bright field overview TEM image of a single nanowire grown at 525 °C viewed along the ⟨1̅01⟩ zone axes, with a higher magnification of the same nanowire shown in panel h. A twinned ZB crystal structure, 180° rotational twinning around a ⟨1̅01⟩-type direction, of the nanowire can be recognized from the (i) characteristic diffraction pattern as well as from the (j) high resolution image. The latter displays an enlarged part of the nanowire shown in panels g and h.

To combine the higher axial growth rate at lower temperature with the vertical alignment exhibited at higher temperature, we used a two-temperature approach similar to that previously demonstrated for growing vertically aligned gold-seeded nanowires at low temperatures.[28] The nanowires shown in Figure 4, panel a were nucleated at 500 °C for 3 min, followed by lowering the temperature to 475 °C and continuing the growth for an additional 7 min. Interestingly, these nanowires exhibit regularly spaced stacking defects, which are shown in Figure 4, panels b–e. The success of this two-temperature technique demonstrates that the nanowire growth itself is not limited to the temperature range shown in Figure 3, panels a–f; rather, it is the nucleation process that is temperature-limited (similar to Au-seeded nanowires).

Figure 4

(a) 30° tilted view of nanowires grown with a two-temperature process with an inset showing a higher magnification of a single nanowire. (b) Bright-field overview TEM image of nanowire grown with a two-temperature approach viewed along the ⟨1̅01⟩ zone axes. (c) Dark-field image of the same nanowire exhibits regularly spaced twin planes; twin 1 and twin 2 can be recognized from (d) characteristic diffraction pattern and also the (e) high-resolution image.

Electrical characterization next was performed to investigate the properties and to confirm the hypothesis of Sn-doping. For electrical characterization, a high aspect ratio was required; nanowire growth was performed at 500 °C for 20 min following TESn deposition of 5 min to yield nanowires longer than 1 μm. Nanowires were deposited onto Si(n++)/SiO2(100 nm) substrates and viewed in SEM using a low magnification. Contacts were fabricated using electron beam lithography, followed by 30 s oxygen plasma ashing, 15 s HCl/H2O (1:10) etching, metal evaporation of Ni/Au (25/100 nm), and lift-off.

Two-probe electrical characterization, with contacts placed at the tip (source, S) and base (drain, D) of the nanowire (Figure 5a), revealed strongly nonlinear current–voltage (IDS–VDS) characteristics. A pronounced region of negative differential resistance was observed for positive VDS applied to the base of the nanowires, see Figure 5, panel b. Such characteristics are typical for highly doped p–n junctions (Esaki diodes), where band-to-band tunneling is the dominant transport mechanism for reverse bias, and small forward bias. Without intentional doping, the devices exhibited surprisingly high peak-to-valley current ratios, up to 12 at T = 295 K and 18 at T = 4.2 K, and peak current densities of 2.5 kA/cm2.[33] To shed more light on where the doping is located along the length of the nanowire, devices with three contacts were also fabricated, see Figure 5, panel c. Electrical measurements of the top segment (contacts II–III) revealed close-to-linear IDS/VDS characteristics with a low resistivity (ρ = 2.9 × 10–2 Ωcm). The positive slope for the back-gate voltage (VGS) dependence of IDS is a clear indicator that the dominant carrier type is electrons, a result of strong n-doping. The base of the nanowire (contacts I–II) shows the same electrical properties as the nanowire in Figure 5, panel a. From these measurements, we thus conclude that the p-doping is primarily located close to contact I in Figure 5, panel c, near the base of the nanowire. Electrical measurements performed on nanowires grown at a lower temperature do not show this Esaki diode behavior (see Supporting Information S2) and have n-type conductivity up to five-times lower than the sample shown in Figure 5, which suggests that Sn incorporation can be controlled with temperature.

Figure 5

(a) SEM image of a contacted Sn-seeded GaAs nanowire grown at 500 °C. (b) Absolute drain current versus drain voltage, VDS, applied to the base of the nanowire in panel a, at T = 295 K and 4.2 K. (c) SEM image of a nanowire from the same growth, with three contacts. (d) Electrical measurements of the bottom segment (I–II) and top segment (II–III) of the nanowire in panel c at 295 K. Top inset shows the back-gate dependence of the current in the upper segment, and the lower inset is a magnification of the peak associated with band-to-band tunneling in the lower segment.

The formation of p–n junctions is somewhat surprising since the nanowires form in a single growth step. Sn is potentially an amphoteric dopant and may in principle act as either donor or acceptor in III–V materials. However, in GaAs, it is almost universally reported to act as an n-dopant. A simple explanation for the origin of the p–n junction is that the n-doping is a result of incorporation of Sn from the seed particle into the core of the nanowire during growth. This hypothesis is in agreement with the finding that the tip of the nanowire, where radial overgrowth is negligible, shows strong n-doping. It is interesting that the n-doping is so efficient despite the low V/III ratio, which might favor incorporation into As sites. However, it is consistent with previous studies of Sn in GaAs, which do not show amphoteric behavior regardless of growth conditions. Conversely, the p-doping we attribute to incorporation of background C from the metal–organic precursors into the radial overgrowth. It is well-documented that incorporation of C is low during axial III–V nanowire growth[34−36] but that it can be used to p-dope radial overgrowth on GaAs nanowires.[23,36] Although the high C incorporation necessary to produce the currents observed here is not typical of GaAs nanowires grown in MOVPE, Sn-seeded nanowires in this study are grown at unusually low V/III ratio compared to typical gold-seeded nanowires. Very low V/III ratio has been associated with formation of As vacancies and enhanced incorporation of parasitic C in GaAs layer growth.[37,38] For the nanowires studied here, the overgrowth is primarily located around the base of the nanowires where the diameters are the largest and where the surfaces have been exposed to the C-containing vapor for the longest time. Very thin p-layers further up on the nanowire may also have been removed in the HCl etch during contact fabrication, which explains the very clear Esaki diode behavior observed. To confirm that the amount of Sn lost from the seed particle is indeed sufficient to provide the high doping levels observed, we have performed a simple order-of-magnitude calculation of the number of atoms (see Supporting Information S3). Also worth noting is that the small decrease in relative Sn content detected by EDX toward the base would be consistent with a shell that is not Sn-doped (see Supporting Information S1); however, since the Sn signal is barely detectable in EDX, this should be interpreted with caution.

Two control samples were grown and processed to confirm that the growth conditions used indeed would lead to strong p-type doping. Both samples were grown under conditions resulting in radial overgrowth. The first sample studied was based on Au-seeded GaAs nanowires grown under “normal” V/III conditions for MOVPE (V/III = 236), and the second on Au-seeded GaAs nanowires grown under conditions similar to the Sn-seeded nanowires, with a V/III ratio of 1.4. Electrical measurements of the first sample, Figure 6, panels a and b, showed very high two-probe resistivity (ρ = 9 × 104 Ωcm), partly explained by the formation of Schottky contacts due to a low carrier concentration. The back-gate dependence, with increasing conductivity for decreasing VGS, is a signature of p-type doping. The second sample, however (Figure 6c,d), prepared under low V/III, showed several orders of magnitude lower resistivity values (ρ = 2 × 10–2 Ωcm) and increasing conductivity with decreasing VGS (inset of Figure 6d), indicative of strong p-type doping. Also worth noting is that the sample grown at higher V/III ratio exhibited defect-free ZB structure (similar to Sn-seeded nanowires), while the sample grown at low V/III ratio exhibited frequent stacking defects in a predominantly ZB structure. These results thus show that possible C incorporation in radial overgrowth is likely related to V/III ratio rather than to nanowire crystal structure.

Figure 6

(a) SEM image of a contacted GaAs Au-seeded reference nanowire grown under V/III = 236. (b) Corresponding I–VDS measurement, for two different VGS, showing high resistive, p-type transport. (c) SEM image of a second Au-seeded GaAs reference nanowire grown under a lower V/III = 1.4. (d) Corresponding I–VDS measurement showing a very low resistance, and the IDS versus VGS in the inset is indicative of p-type transport.

In summary, we have demonstrated the use of Sn as a seed particle for GaAs nanowire growth, with Sn particles formed in situ using the standard metal–organic precursor tetraethyltin. The resulting nanowires exhibit uniform vertical alignment and defect-free ZB crystal structure for optimized growth temperature. Nanowires grown at high temperature exhibit low growth rate, but low growth temperature is associated with kinking. We demonstrate that straight nanowires with higher growth rate can however be achieved by a two-temperature approach, nucleating at higher temperature and growing at lower temperature. The nanowires in this work also exhibit Esaki diode behavior with high current levels despite the absence of intentional doping during nanowire growth. Electrical evaluation indicates that Sn primarily incorporates in the core of the nanowire, likely via the seed particle leading to strong n-doping, and that the p-doping is located in the radial overgrowth, where the high doping level is an effect of the low V/III ratio used during growth. Incorporation of both of these dopants is expected to depend on growth conditions: specifically, C-doping is anticipated to be very sensitive to V/III ratio, while Sn-dopant incorporation is known to be strongly temperature-dependent. An important next step for future studies will therefore be to map the nanowire properties as a function of accessible growth parameters to determine the extent to which dopant levels can be controlled in Sn-seeded nanowires. In conclusion, this work demonstrates that the use of alternative seed particle materials is an interesting and potentially simple approach to introduce alternative properties and functions into semiconductor nanowires while at the same time avoiding the use of industrially incompatible gold.