Growth of branched nanowires via solution-based Au seed particle deposition

Nanowires offer unprecedented flexibility as nanoscale building blocks for future optoelectronic devices, especially with respect to nanowire solar cells and light-emitting diodes. A relatively new concept is that of charge carrier diffusion-induced light-emitting diodes, for which nanowires offer an interesting architecture by use of particle-assisted core-branch growth. The branches should be homogenously distributed along the cores. However, most deposition techniques, such as aerosol particle deposition, mainly yield particles at the nanowire tips for dense nanowire arrays. In this study, we demonstrate a liquid-based approach for homogeneously distributed formation of catalytic Au particles on the core nanowire sidewalls which is cost and time-efficient. Subsequently, we demonstrate the synthesis of dispersed nanowire branches. We show that by changing the deposition parameters, we can tune the number of branches, their dimensions, and their growth direction.


Introduction
Energy is one of humanity's biggest challenges in the 21 st century, where the use of fossil fuels has led to global warming which has become of immediate concern. To meet the goal of limiting global warming to an increase of 1.5°C compared to pre-industrial temperature level, two main approaches are being followed: (i) utilizing renewable energy harvesting such as wind power and solar power, and (ii) finding solutions to reduce the energy cost of technology currently in use [1,2]. Lighting consumes about 20 % of the yearly electricity production worldwide and an increase of 40 % in energy use for lighting is expected by 2030 [3]. Recent EU energy label standards have shifted theA++ efficiency class marking to C, in an effort to push for further research and development in light bulbs and other devices in the electro-optical components sector [4]. Nanowires (NWs) are promising candidates for future optoelectronic and photonic devices such as solar cells [5][6][7][8][9], photodetectors [10,11], and light-emitting diodes (LEDs) [12]. Due to the efficient strain relaxation in NWs [13,14], heterostructures with large lattice mismatch can be grown, suitable for LEDs. Standard LEDs which dominate the market consist of planar heterojunction structures in which the active region (usually a multi-quantum well (MQW) stack) is embedded in between a p-n junction. A novel concept of charge carrier diffusion-induced LEDs was demonstrated in the nitride materials system [15]. This architecture allows to design LEDs without the need to sandwich the opto-electrically active layers in-between electrical contacts. In this design, electrons and holes are injected by applying a bias over the structure, reach the active region with a smaller bandgap by bipolar diffusion, and recombine there. This concept opens the possibility for flexibility in device design and processing. The nanowire geometry is especially interesting with respect to this kind of LED structure. Theoretical simulations have shown that in NWs with a diameter below a threshold value, the generated photons are emitted efficiently because the electric field leaks out of the NW [16]. It opens up to completely avoid total internal reflection; the main challenge of extracting light from semiconductors. The goal is to synthesize heterostructure core-branch NWs, where core p-i-n doped NWs with a diameter of~180 nm and a high bandgap are grown via 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.
the vapor-liquid-solid (VLS) method [17], followed by a subsequent growth of NW branches with lower bandgap and thinner diameter (∼25-50 nm), as shown in figure 1(a)). The difference in bandgap will induce carrier diffusion towards the branches, where the carriers can recombine, and light is efficiently emitted. In figure 1(b)) a cross-section scanning electron microscope (SEM) image of core-branch InP NWs is shown.
Highly pure Au aerosol particles have previously been used for branched NW synthesis on NW cores randomly distributed on a substrate and in regular NW arrays [18][19][20][21]. In our pattern of 1 μm pitch NWs in a hexagonal array, it turns out that this technique mainly deposits particles at the tip of the NWs [21,22] and not along the NWs which is necessary for the synthesis of branched device structures. We address this challenge by depositing Au particles on the sidewalls of NWs by immersion of the substrate in a liquid solution [23][24][25][26], which is a cost and time-efficient process. The seed particles enable the growth of randomly distributed branch structures over the main part of the NWs.
We use an aqueous solution of HAuCl 4 for the galvanic deposition [24] and perform a systematic study of the Au deposition parameters by varying the temperature of the sample-liquid interaction, concentration, and time in order to investigate how these parameters affect the particle formation and which level of control can be achieved. We demonstrate a dense formation of particles distributed over the entire NW core. In order to study the synthesis of branched NW structures without the additional complexity of heterostructure growth, we focus on the growth of InP branches on InP cores in this study. We find that we can grow approximately 70 branches per NW.

Experimental methods
The periodic NW array which acts as the core for the branches was prepared by patterning the substrate using Displacement Talbot Lithography [5,[27][28][29][30][31] using a mask with 1 μm pitch of circular holes with a diameter of 120 nm. After the lithography step, the photoresist was developed for 60 s in MF24A and an undercut was created. A Temescal electron beam evaporator was used to deposit 65 nm of Au on the wafer. The excess resist with Au on top of it was removed in a lift-off process using Remover 1165 after which the wafer was cleaned using deionized water (DI H 2 O). The wafer was then diced using a Disco Dicer into 9×11 mm 2 sized samples.
The p-i-n InP NWs were grown using a laminar flow Aixtron 200/4 MOVPE reactor with a reactor pressure of 100 mbar and a total flow of 13 l min −1 . To preserve the gold pattern on the substrate a low temperature nucleation step at 280°C was used (as described in [32]) by introducing trimethylindium (TMIn) and phosphine (PH 3 ) with molar fractions of χ TMIn = 4.46 × 10 −5 and χ PH3 = 6.92 × 10 −3 , respectively. The temperature was then increased to 550°C for a high temperature annealing step under χ PH3 = 3.46 × 10 −2 to remove any surface oxides. To start the nanowire growth the reactor temperature was decreased to 440°C and χ PH3 reduced to 6.92 × 10 −3 . TMIn with a molar fraction χ TMIn = 4.46 × 10 −5 was introduced at the growth temperature to start the growth. Diethylzinc (DEZn) with a molar fraction χ DEZn = 4.95 × 10 −6 was used for the p-doped segment and then decreased to χ DEZn = 3.02 × 10 − 8 for growth of the compensation doped, intrinsic region [33][34][35]. For the n-segment, tetraethyltin (TESn) was added to the flow in a molar fraction of χ TESn = 2.12 × 10 −5 [34,36]. During the entire growth of the NWs, a flow of HCl (χ HCl = 1.23 × 10 −2 ) was used for in-situ etching to impede radial growth [37]. After growth, the NWs were decorated with Au particles. To deposit Au particles on the side walls of the NWs, a 10 mM HAuCl 4 aqueous solution was prepared. HAuCl 4 · 3H 2 O (Sigma Aldrich) was mixed with Millipore DI H 2 O to form an aqueous solution. The stock solution was subsequently diluted into 0.25, 0.5, 1, 2, and 4 mM solutions. In addition to the variation of the concentration, the effects of the deposition temperature and time on the Au particle formation were studied. To deposit Au particles on the NWs, the samples were submerged in the solution, and the metal particles deposit on the semiconductor surface via galvanic displacement [24]. Since the growth of cores and branches is done in two separate growth runs, we found that it is important to remove the seed particle from the cores, to avoid axial growth of the cores during branch growth. In order to selectively remove the Au seed particle from the NW tips without affecting the Au particles formed on the sidewalls, S1818 photoresist was spin-coated on the samples at 5000 rpm for 60 s, followed by a soft bake at 115°C for 90 s [38]. To access the seed particles from the cores the samples were flood exposed under UV light for 1 s, developed with MF319 for 2 min, and flushed with DI H 2 O for 2 min. The seed particle was etched by wet chemical etching: 25 s in buffered etch oxide (BOE) 1:10, 25 s in potassium iodide (KI) and then 3 min DI H 2 O. The S1818 was removed from the samples by use of Remover 1165, after which the samples were flushed in DI H 2 O for 3 min. A Zeiss Leo 1560 SEM was used to inspect the samples and ImageJ was used to count the number of particles, branches, and their length and diameter. Figure 2 shows 30°tilt SEM images of the process.
The samples were then inserted again into the reactor to grow the branches. The precursors used for growth were TMIn with a molar fraction χ TMIn = 7.43 × 10 −5 and PH 3 with a molar fraction χ PH3 = 6.92 × 10 −3 . The growth time was kept constant at 1 min.

Results and discussion
Au deposition Deposition temperature variation The first studied parameter was the deposition temperature. Figure 3 shows SEM images after the Au deposition for 1 min at 20 − 80°C in steps of 20°C in a 1 mM solution. At room temperature, the particles are distributed along the entire core NW length. By increasing the temperature, we observe that larger size particles are formed on the NWs in fewer amounts compared to room temperature. A higher temperature accelerates the galvanic displacement reaction [39]. In addition to the accelerated reaction rate, the effective concentration in the solution changes during the deposition at higher temperatures. Considering the water-based nature of the solution, when increasing the temperature close to the boiling point of water, water desorption occurs at an increased rate. The decrease in water content changes the effective concentration of the solution. In the supplementary information (si), figures S1-S4 show the experimental results of the temperature variation for 30 s, 2, 3, and 5 min, respectively.
We have shown that the Au deposition can be achieved on denser NW arrays as well, for example, an array with a 500 nm pitch. The temperature variation has been tested on NW cores grown on a 500 nm pitch geometry and compared with the previous results. In figure 4, a comparison between a 3 min deposition at 60 C,  is shown for the two different arrays. We can observe that the results are very similar between the two arrays, with few particles formed along the NW core. Although it is not possible to discern a substantial difference, there is an indication that the process is mass flow limited as the core NWs on a 1 μm pitch have particles of a larger size.

Concentration variation
The concentrations used for the study were 0.25, 0.5, 1, 2, and 4 mM.

Deposition time variation
In order to further evaluate the parameters for the formation of Au particles from the solution, we carried out time-dependent deposition studies. Figure 6 shows SEM images of Au formation from a 1 mM solution on the NWs for several different times. The complete set of experiments can be found in figure S6 which shows SEM images of the Au formation for immersion times ranging from 1 s to 10 min in irregular steps. From 1 s to 30 s, the deposition time studied is 1, 5, 10, 15, 25, and 30 s, and from 1 min to 10 min it is in steps of 1 min. Increasing the deposition time, leads to an increase in the particle density, followed by agglomeration of particles at longer time scales. We have also observed that the Au particles preferentially form along the intersection of NW facets. Increasing the deposition time, it would be expected that the size and density of the agglomerations increases as well, but this is not what we observe. After a deposition time of 2 min, a saturation of the agglomerates is observed. We do not have an indication of Ostwald ripening occurring as the smaller particles are not consumed to form bigger agglomerations. The other possibility is that the NW sidewalls are covered with particles, but not all the Au nuclei are observable in the SEM due to their small size. Complete coverage would halt the redox reaction and no further deposition of Au occurs [40] as there would not be any free semiconductor surface left for the reaction to occur. A final observation that we have made is related to the age of the NWs and their relation to the Au formation. In short, we observed that the deposition is very sensitive to the age of the NWs and deposition should be performed within a few hours of the NW growth for the best results. Compared to immediate deposition after synthesis of the core NWs, deposition on the aged NWs leads to fewer particles with increased size deposited. The main hypothesis is that it is related to changes in the surface oxide of the NWs, which changes the surface energies and interaction of the Au containing solution that leads to particle deposition. In the SI, a comparison between the deposition on fresh and aged NWs is shown in figure S7. A detailed description of the observation is provided along with SEM images of the depositions for different times (figure S8), and InP branch growth for the different Au deposition times (figure S9). Plots of the particle/branch number and particle/branch diameter are shown in figures S10 and S11, respectively.

InP branch growth
To study the branch growth from the deposited Au particles, we have grown InP branches. Figure 7 shows top view SEM images of the intrinsic InP branches grown on the core NWs. The effect of the concentration on the size and number of particles formed strongly influences the growth results. The cores have a 3-fold symmetry [20], and the branches are expected to grow in the three symmetry related 〈111〉B directions. However, the core NWs have many twin planes which cause a rotation of the crystal structure by 60°, resulting in 6 downward pointing 〈111〉B growth directions [18,20,22]. In the case of the 0.25 mM solution ( figure 7(a))), the branches grow aligned to one another with little spreading along the growth axis, making the apparent 6-fold symmetry visible. These branches have a narrow distribution of their length and diameter. The results from using the solutions with higher concentrations leads to branches that have a broader distribution in diameter and which are not perfectly aligned along the facets of the cores. When the concentration of the solution used for deposition is increased, the parasitic growth of nanowires on the substrate is observed to become dominant over branch growth. The particles on the substrate have a larger size as compared to the ones on the sidewalls of the cores and reduce the material available that can contribute to branch growth by parasitic group III adsorption. Figure S12 shows SEM images of the branches grown using different concentrations of the solution tilted at 30°.
In figure 8, we have plotted the average number of particles formed and branches grown as a function of the solution's concentration. For each sample, 9 NW cores have been investigated to count the particles and the branches formed. The plot shows that the number of particles measured before growth is about a factor of three higher than the number of branches counted after growth. This difference indicates that the Au particles move along the nanowire and merge to form bigger Au particles before branch nucleation occurs. This results in fewer branches with a larger diameter as compared to the diameter expected would the branches be grown from the originally deposited particles. The estimations were performed by manually counting the number of particles and branches as contrast differences were too small to distinguish between core and particle/branch using imageJ. We have used the tilt images for the estimations and since only half of the NW is visible, the counted particles/branches were multiplied by 2 to account for the other half. Figures S13 and S14 show a plot of branch diameter and length for the different concentrations of the solution during deposition. In figures S15 and S16, histograms of the diameter and length distribution of the branches for the different concentrations are shown.
To further inspect and characterize the branches, we have taken cross-sectional SEM images of the samples, shown in figure 9. We have analyzed the branches grown on the samples where the Au particles were deposited at different times. The images reveal a surprising result. The expected growth direction of the branches would be the 〈111〉B direction, but instead, we observe that the branches grow almost purely in the 〈110〉 direction for the shortest particle deposition time of 1 s. Branches grown in the 〈111〉B direction become more frequent when the deposition time is longer than 10 s and become the dominant growth direction for depositions longer than 2 min. In figure 9(g)) a plot is shown which relates the ratio of branches grown in the 〈110〉 and 〈111〉B to the deposition time. For each sample, 7 NW cores have been investigated to count the branches growing in the 〈110〉 and 〈111〉B direction. The number of 〈111〉B directed branches increases for depositions up to 2 min, and then a constant ratio close to 0.1 is reached. The full set of cross-sectional SEM images can be found in figure S17.
It has previously been shown that the diameter of the NW can be used to tune the growth direction of NWs [41], where the growth direction changes from 〈111〉 to 〈110〉 for a NW diameter below 20 nm. However, in our  case, we have not observed a similar dependence. For example, in figure 9(e)) several thick branches grow in the 〈110〉 direction even though neighboring branches with smaller diameters grow in the 〈111〉B direction. We raise the hypothesis that this change in the growth direction is related to the content of the Au containing solution. Inside the solution, the HAuCl 4 is dissociated into H + and AuCl 4 − ions. When the redox reaction occurs, the AuCl 4 − ions are further dissociated into Au 3+ ions which reduce to Au 0 and Cl − ions that can act as an etchant to the native oxide surrounding the cores after growth and exposure to air [39]. Furthermore, the H + ions could interact with the Cl − ions and form HCl, which is known to etch InP [37]. We speculate that during the formation process of the Au particles, the native oxide is partly being etched. The discussion can be divided between short and long deposition times. For shorter times (< 2min), the Cl − ions and the HCl partially etch the native oxide. When the samples are inserted into the MOVPE reactor for the branch growth and heated to the growth temperature, the Au particles can migrate on the core which leads to the formation of bigger Au particles by agglomeration. During this migration, the Au particles can desorb the oxide locally underneath them [42], but they are still surrounded by the native oxide which remains after the deposition is terminated. The oxide that surrounds the Au particles becomes a barrier in the growth process, and it forces the branches to grow in another direction than (111)B [43][44][45]. For longer deposition times (>2min) the etching process is longer, and it is possible that the entire native oxide is etched. Upon re-exposure to the ambient after the Au deposition process, an oxide will regrow around the cores, but it may have a different stability or thickness. Thus, depending on the deposition time, upon re-exposure to air both the initial oxide and a newly formed oxide will be present. The properties of the newly formed oxide could lead to differences in its partial or complete removal during heating up for growth as compared to the initial oxide which desorbs only underneath the Au particles. The complete removal of the new oxide causes the particles to not be trapped by an oxide anymore and allows the growth to proceed mostly in the expected 〈111〉 B direction [44,45]. Further research is necessary to provide a definite answer on the growth direction change. The characterization and identification of an oxide at the nm level around sub 20 nm Au particles is challenging. Figure 10 shows a bright field transmission electron microscope (TEM) image performed on one NW with branches grown using the 1 mM solution. In the TEM image, we observe branches that grow both in the 〈110〉 and 〈111〉B direction. In both types of branches, the lattice planes propagate from the cores to the branches, meaning that the branches grow epitaxially and contain twin planes because of the cores. In the branches grown in the 〈111〉 B direction, the (111) twinning planes that originate from the core propagate into the epitaxially grown branches. These twins become terminated at the surface of the branch nanowire when it has a certain length set by the angle of growth. Between a branch growing in the 〈111〉B and the core NW, the observed angle is 109.5°. Then the normally occurring twins perpendicular to the growth direction are observed. For 〈110〉 NWs that grow perpendicular to the 〈111〉B core NWs, the twins propagate through the entire length of the branched NWs [22].

Conclusions
In conclusion, in this study, we have used a solution method to deposit Au particles on the side walls of NWs to enable the growth of nanotree structures. To study the parameter space of the particle deposition we varied (i) the concentration of the solution, (ii) the time which we submerge the substrates, and (iii) the temperature of the solution during deposition and found 1 mM concentration, 30 s deposition, and 20°C as the optimal deposition parameters. By changing the concentration of the solution, we have observed that the growth direction of the branches can be the 〈111〉B or the 〈110〉 direction. We have successfully grown InP branches on p-i-n doped InP cores and aim to develop heterostructures such as InP p-i-n doped cores with InAsP branches emitting in the infrared (IR) region, interesting for optical communication at 1.3 μm and 1.55 μm. Other material combinations such as GaInP/GaInP, with the branches having a lower Ga content than the cores are interesting for use in LEDs emitting in the red visible region, for highly efficient lighting solutions.