Report on Design and Growth of Ingan Nanowires for Emission Wavelength >550 Nm

Abstract

Green solid state emitters with good internal quantum efficiency are a major concern in order to get white light by colour mixing. InGaN being a potential candidate for green emitter has been studied here. Shorter wavelength emitting LEDs, in visible range, made of III-nitrides are already there in the market. So, bandgap tuning of GaN is done by incorporating InN in order to get bandgap corresponding to λ~550 nm (mid-green). This report addresses the growth mechanism of InGaN nanowires, problems in tuning the bandgap i.e. difficulties in incorporating high concentrations of Indium, which is necessary to reduce bandgap of GaN and how these problems are resolved by optimizing the growth conditions.

Introduction

InGaN finds application in many optoelectronic devices such as solar cells, LEDs, Laser diodes, Photodetectors etc. The wide range of applications of InGaN is because of its tuneable bandgap. The bandgap can be varied from 0.69eV (λ=1.7µm for 𝐼𝐼𝐼𝐼𝑁𝑁) to 3.4eV (λ=365nm for 𝐺𝐺𝐺𝐺𝑁𝑁) which covers nearly the entire solar spectrum and visible wavelength range. Therefore, it is very useful in heterojunction solar cells, and visible light LEDs. Alloying is a common tool for tuning physical properties, such as lattice constant, as well as optical properties like band gap. Specifically for LEDs, III-nitrides are a promising material for high efficiency solid state lighting. To obtain white light, phosphors are used to partially convert blue LED light to yellow/green range. But this conversion comes with high energy loss (~25%) known as Stokes’ loss limiting the highest attainable efficiency to much below 100%. Whereas, InGaN/GAN LEDs are offering efficiencies more than 80%. So, the phosphor conversion for getting white light can be eliminated by using LEDs emitting different wavelengths such as red, green/yellow and blue. Not only higher efficiency but solid state lighting offers longer lifetimes as compared to phosphor based lighting.

The issue in transition from phosphor based lighting to LED based is that Green LEDs have lower efficiency compared to red (III-nitride based) and blue ones (III-phosphide based ). Phosphor free white LEDs need at least a green emitter (λ~550nm). So, the efficiency of white LED is limited by maximum efficiency of green emitter. Even phosphor based green emitters have higher efficiency than that of semiconductor based. So, for White LEDs green emitter is very important. The red and blue LEDs are made of III- phosphides respectively. While looking for a green emitter, the phosphides and arsenides have also been explored but the only material (other than nitrides) with wide bandgap(required for green region) is AlInP (aluminium indium phosphide) for which convenient substrate have not been found and the material-growth system characteristics are uncertain such as concentration of Al for green emission etc. Therefore, the option left is III-nitrides for green emitter.

Efficiency obtained with InGaN/GaN layered LEDs is limited to 40%. The reasons behind this low efficiency could be larger defect concentration due to unavailability of native substrates (defects contribute non-radiative recombination sites to the film), lower concentration of In in InGaN due to compositional pulling effect, etc. But to get Green 2 emission from InGaN, high concentration of In in the crystal is required to reduce the bandgap to green wavelength. In planar structure the thickness is limited to a critical value after which the layer will break by plastic deformation but for high In content in the layer we need thick layers. So, planar layers do not serve the purpose. All these issues can be resolved by switching to 3D growth of InGaN e.g. nanowires. The nanowires are free from defects as the strain due to lattice mismatch is relaxed within very few monolayers, due to high surface to volume ratio, and rest of the wire grows without strain so compositional pulling is absent in nanowires.

This review report focuses on growth mechanisms of InGaN by understanding GaN nanomaterial (3D) growth. The growth methods that are in use to grow nitrides are CVD, VLS, MBE, etc. Vapor-Liquid- Solid (VLS) method is the most common technique used to grow Semiconductor Nanowires using a metal catalyst that act as nucleation sites. For spontaneous growth without using a catalyst, MBE can be used by optimizing the various parameters of growth such as temperature, Ga and N fluxes, and the various intermediate layers. In this paper, we will study the problems with planar InGaN layers and how they are prevented using InGaN nanowires. InN when alloyed with GaN gives poor crystal quality and inhomogeneity due a large covalent radii difference between Ga (1.26 A°) and In (1.44 A°) which also causes large internal strains.

Another issue that needs attention is in phase segregation in defect spaces which reduces In content in the InGaN alloy crystals and disturbs the stoichiometry affecting bandgap, lattice constant and some other properties of interest. This problem arises because of high growth temperatures, high vapor pressure of InN as compared to that of GaN, difference in formation enthalpies of InN and GaN etc. Therefore, it can be minimized by optimizing the growth parameters, such as the use of relatively low growth temperatures, high V/III flux ratio (i.e. N/Ga flux), low growth rate and low growth pressure. Large V/III ratios were found to be able to suppress the indium segregation during growth of InGaN. All the issues are addressed in detail in the later parts of this report.

Literature review

The literature review is intended to study the Growth mechanism of InGaN nanowires and how it is different from planar layers, and also the compositional pulling effect in InGaN. Most of the work studied is on spontaneous (without catalyst) MBE growth of InGaN over Si(111). Some groups have used buffer layers (AlN) to bridge the lattice mismatch between InGaN and Si. Ristic et al have studied effect of pre-deposited Ga droplets on nucleation of GaN Nanowires. Bare Si(111) wafers have been used for patterning with Ga droplets to avoid effects of buffer layer morphology, strain differences, crystalline quality, or thickness on nucleation process. The Ga droplet patterning was done at temperatures below Ga evaporation temperatures to prevent its desorption. The Ga flux used for patterning and GaN Nanocolumns growth was same. The metallic Ga forms droplets on Si surface by coalescence 3 to reduce surface energy.

After patterning Si substrates with Ga Droplets, the substrate temperature was increased to growth temperature. Ga desorption can occur at growth temperature but partial nitridation reduces the desorption rate. The Nanocolumn grown substrates were studied by SEM (scanning electron microscopy). What they have observed was that the nanocolumns were formed everywhere except the Ga droplet sites. They concluded that Ga droplets hindered the nanocolumn growth on them. As the Ga droplet diameter decreased (substrates with variable droplet diameters were prepared for nanocolumn growth) nanocolumns were observed on droplet sites also but were slightly tilted with respect to the substrate as if they emerged from the facets of partially nitride Ga droplets. Nitridation was observed more in case of smaller droplets. They have reported that Ga droplets only serve as reservoirs of Ga, supplying Ga to the Nanowires when growth rates are faster and V/III ratio is high. This is confirmed by two observations (1) high densities of Nanowires around Ga droplets and (2) the density of nanocolumns was more than that of Ga droplets. Both these observations confirm that Ga droplets do not act as nucleation sites but reservoirs of Ga.

The two concerns addressed by Ristic et al. are the mechanisms responsible for constant diameter of nanocolumns throughout the length and nucleation process that determines nanocolumn size and density. It was anticipated that nanocolumn growth started from random distribution of GaN islands. First a 2D layer forms and then it transforms to 3D to minimize energy and strain build-up. When the strain reaches threshold, nanocolumns are formed simultaneously on the entire substrate. Nanocolumn nucleation takes long times to form over the entire substrate surface as it occurs when GaN islands reach a saturation density. These research groups have not reported 2D to 3D growth mode change instead a direct nucleation of islands on reaching saturation density i.e. following Volmer- weber (VW) growth mode have been reported. It is supported by the SEM and HRTEM analysis which does not indicate any wetting layer formation in GaN nanocolumns. SixNy layer over Si have been reported which can be due to reaction of Si with active nitrogen plasma.

Nucleation can be explained by capillarity theory given by Volmer and Weber. According to the theory there is a critical nuclei size, which increases with increasing substrate temperature, above which a nuclei becomes stable and below which it decays by desorption or by diffusing into other larger nuclei. So, all the clusters having radii smaller than critical nuclei size disappear and all those which have radii greater than critical radii survive and keep growing into larger ones by diffusing atoms. This determines the minimum nucleation sites size at a given growth temperature. So, large diameter nanocolumn growth can be predicted at higher temperatures. For this to happen, sufficient Ga flux is needed to compensate for desorbing Ga. The Ga ad-atom mean diffusion length depends on substrate temperature and III/V flux ratio. The distance between two nuclei should be equal to twice the mean diffusion length of Ga for saturation to occur. The diffusion length decreases with decrease in III/V ratio. This determines the diameter of nanocolumn and density.

When III/V ratio is very small (N-rich condition), stable nucleation sites do not coalesce as the incoming Ga atoms will get incorporated on their top. This behaviour is explained by tendency of hexagonal III-nitrides to grow into columnar grains. When III/V ratio is large i.e. high Ga flux, vertical growth saturates and lateral growth proceeds leading to diameter increase and eventually coalescence to form 2D layer. Since, Ga has smaller diffusivity, smaller flux rate (for vertical growth higher N-flux is needed), therefore the time required to reach the nanocolumn density saturation is more. Daruka and Barabasi studied that the growth mode depends upon the total amount of material deposited and lattice mismatch. They showed that growth mode and morphology can change from Frank-van der Merwe (FM) to SK and then to VW as the lattice mismatch increases. They also concluded that GaN nanocolumn nucleation on Si(111) occurs via VW growth mode due to high lattice mismatch. The substrate surface gets covered by different size GaN islands during nucleation and results in varying diameter Nanocolumns. Since, the nucleation does not start at the same point of time at all the points on substrate, therefore, the height of nanocolumns also varies. Height variation can be due to Ga diffusion along the sidewalls of nanocolumn to its top. So, larger diameter nanocolumns grow slower. When nucleation is achieved, the further nanocolumn growth depends upon two Ga diffusion processes (1) Ga atoms impinging directly on the top of islands, and (2) Ga diffusion through sidewalls to top when the atoms impinge on the substrate surface. This theory can be supported by the observations that Ga droplets act as reservoirs promoting neighbouring island growth, so the density in the neighbourhood of Ga droplet increases, and also when the Ga flux is increased nanocolumn diameter increases only at the top because more Ga atoms diffuse to the top of nanocolumn. The nanocolumn growth is explained by considering high Ga diffusion length along nanocolumn side walls and high sticking coefficient of tip of nanocolumn i.e. the bottom plane (c-plane). For a given surface, ad-atom, and temperature Qdes is fixed. Such surface would have high diffusion length giving very short time for ad-atom adsorption into crystal. So, for vertical growth i.e. ad-atom adsorption on c-plane, Ga ad-atoms on this plane must have lower diffusion length due to high concentration of ad-atom adsorption sites on this surface and higher Qd as c-plane is the polar surface with more dangling bonds. This very well explains the vertical growth. The diameter increase at the top can be due to accumulation of Ga ad-atoms on the top surface resulting in residence time increase which eventually leads to incorporation into the top sidewalls. When buffer layer such as AlN is used, lattice mismatch reduces enabling better wetting properties and hence promotes SK growth mode which makes islands to coalesce and larger islands are formed between which nanocolumns can grow due to excess N. If the thickness of these islands increases, it results in strain relaxation and increase in surface roughness. This rough compact material co-exists along with nanocolumns. The compact layer can still be 6 avoided by increasing the N-flux further even if AlN buffer layer is used. It was concluded that AlN use increases Ga diffusion length which can be reduced back by increasing the N- flux. This is observed that N-excess has two effects (1) it decreases Ga diffusion length, and (2) it promotes vertical (c-axis) growth.

Bertness et al. have studied the spontaneous GaN nanowire growth using MBE on Si(111) substrates using AlN as buffer layer. The sequence of growth initiates with AL pre-layer growth, then AlN buffer layer followed by rough GaN matrix layer for nanowire nucleation. The growth was carried out in presence of high N- flux as compared to Ga or Al flux and substrate temperature was kept in between 810-830 °C. FESEM and AFM confirmed the nanowire nucleation in the hexagonal pits formed in GaN matrix layer. They have observed that nanowire nucleation depends on N-flux, substrate temperature, and AlN buffer layer thickness.

If Al and AlN layers are skipped, no nucleation was observed and amorphous GaN was deposited even if high N-flux was maintained. At higher N-flux rates nanowire as well as matrix layer growth rates were found to be increased, which is in contrast to the observation reported by J. Ristic´. The nucleation rate increased as the thickness of AlN layer was increased. They have stated that the cause of these discrepancies could be substrate outgas procedure and growth system hardware which affects SiNx formation. Both atomic N and excited molecular N2 have been used to study growth of nanowires and found that excited N2 has lower energy than atomic N and hence it lowers diffusivity of Ga. So, atomic N is preferred for GaN growth.

They have also used NH3 as nitrogen source and found out that no nanowire growth occurred, but good high hexagonal pit density was observed. For vertical growth of nanowires in hexagonal pits they have accounted higher sticking coefficient of c-plane as compared to sidewalls. The role of atomic N has been described as a stabilizer for the formation of c-plane on AlN surface. More control of MBE growth parameters can give better nanowire growth.

Goodman et al. reported growth of InGaN nanowires by Plasma assisted MBE. Using technique reported by Kikuchi et al., first GaN nanowires were grown for short period of time and then over these InGaN nanowires were grown by altering the growth conditions such as lower substrate temperature and lower Ga flux to incorporate In in the crystal. The growth conditions such as low III/V ratio etc. were maintained as in case of GaN nanowires. The condition that needs to be changed is substrate temperature. Indium has low vaporization temperature so at high growth temperatures it cannot be incorporated into the nanowires. Since, temperature plays an important role in InGaN growth, temperature variation effects have been studied by Goodman et al. At lower growth temperatures no nanowire growth was observed. So, the temperature was increased but kept below the temperature required for GaN nanowire growth. This increase resulted in incorporation of In into crystal as well as nanowire formation.

They have also grown InGaN nanowires without using GaN nanowire nucleation layer and found that they were same as obtained when nucleation layer was used however the densities were different in the two cases. Higher densities were obtained in case of nucleation layer which was grown on a Ga droplet patterned substrate (Ga droplet patterning gives higher densities of GaN nanowires). They have observed nanowires of different diameter and height. The nanowires which have larger height were thinner and the ones that have smaller height were thicker. The possible cause of this phenomenon could be that at lower growth temperatures the metal atoms do not get sufficient energy to distribute among the wires. But after studying GaN diameter variation they concluded that the variations in case of InGaN are due to incorporated In content. The main aim of the study by Goodman et al. was to grow InGaN nanowires with constant In content throughout the wire length. But they observed that the content varies from 0-35% along nanowire length. Also variation of In content was observed from wire to wire in same growth sample.

Kong et al. have studied the above mentioned issue i.e. In concentration variation through the length of the wire. They found out that as InGaN epil-ayer grows over GaN nanowires the strain increases due to the large lattice mismatch between InN and GaN (~11%). This elastic strain can result in variation in content of In in nanowires. It is also called as compositional pulling or crystal pulling effect in InGaN/GaN nanowires. In this effect, the In incorporation into nanowires is hindered by the lattice mismatch, hence it pulls the composition of growing InGaN nanowires towards that of substrate GaN i.e. In concentration increases with increasing height and is minimum at the GaN/InGaN interface in order to match the lattice. Pereira et al. has also reported this effect. They have studied this effect by Depth-resolved cathodoluminescence (CL) and Rutherford backscattering spectrometry (RBS) and reported that In mole fraction increases for thicker InGaN samples (more relaxed). To reduce the compositional pulling effect, the lattice relaxes elastically by increasing Nanowire diameter as well as plastically by introducing misfit dislocations at the interface. When the lattice is relaxed In content in InGaN increases reducing the compositional pulling effect and lattice mismatch. Indium incorporation alters the bandgap and hence the emission wavelength.

Future Work

The aim of this study is to find an efficient growth method for InGaN nanowire. From the above studied growth methods it can be concluded that MBE is a technique that can give spontaneously grown nanowires free from impurities (incorporated from catalyst). The conditions for spontaneous growth are low III/V flux ratio, optimum substrate temperature (a range decided by factors such as In desorption and quality of crystal) and growth rate, GaN nanowire buffer layer to get higher density of InGaN nanowires, etc. Also high In and Ga flux rates can improve growth rate but III/V ratio must be maintained. Nanowires offer better structural and optical properties such as less or no defects and impurities (can be grown without catalyst). Absence of compositional pulling allows good amount of In incorporation permitting desired tuning of bandgap. Hence InGaN nanowires can be used as active layer in LEDs. Nanowires restrict carrier movement in one direction improving directionality and hence external quantum efficiency of LED. Localisation of carriers using quantum well can also help in improving radiative emission by reducing non- radiative recombination. So, a quantum well based LED structure can be fabricated by sandwiching InGaN nanowires in between p-GaN and n-GaN.

References

1. J. Ristic´, E. Calleja , S. Ferna´ ndez-Garrido, L. Cerutti , A. Trampert , U. Jahn , K. H. Ploog, J. Crystal growth, 310(2008) 4035.
2. K.A. Bertness, A. Roshko, N.A. Sanford, J.M. Barker, A.V. Davydov, J. Crystal Growth 287 (2006) 522.
3. K.A. Bertness, A. Roshko, L.M. Mansfield, T.E. Harvey, N.A. Sanford, J. Crystal Growth 300 (2007) 94.
4. E. Calleja, J. Ristic´ , S. Ferna´ ndez-Garrido, L. Cerutti, M.A. Sa´nchez-Garcı´a, J. Grandal, A. Trampert, U. Jahn, G. Sa´ nchez, A. Griol, B. Sa´ nchez, Phys. Stat. Sol. (b) 244 (2007) 2816.
5. H. Luth, Surfaces and Interfaces of Solid Materials, Springer, Berlin, 1993.
6. K.A. Bertness, A. Roshko, L.M. Mansfield, T.E. Harvey, N.A. Sanford, J. Crystal Growth 300 (2007) 94.
7. R.K. Debnath, R. Meijers, T. Richter, T. Stoica, R. Calarco, H. Lu¨ th, Appl. Phys Lett 90 (2007) 123117.
8. R. Calarco, R.J. Meijers, R.K. Debnath, T. Stoica, E. Sutter, H. Lu¨ th, Nanoletters 7 (2007) 2248.
9. T. Zywietz, J. Neugebauer, M. Sheffler, Appl. Phys. Lett. 73 (1998) 487; T. Zywietz, J. Neugebauer, M. Scheffler, J. Northrup, Chris G. Van de Walle, MRS Internet J. Nitride Semicond. Res. 3 (1998) 26.
10. I. Daruka, A.-L. Barabasi, Phys. Rev. Lett 79 (1997) 3708.
11. A. Kikuchi, M. Kawai, M. Tada, K. Kishino, Jap. J. Appl. Phys. 43 (2004) L1524.
12. K. D. Goodman, V. V. Protasenko, J. Verma, T. H. Kosel, H.G. Xing, and D. Jena,J.Applied Physics 109, 084336 (2011)
13. X. Kong, S. Albert, A. B. Encabo, M. A. Sanchez-Garcia, E. Calleja, and A. Trampert, Phys. Status Solidi A 212, No. 4, 736–739 (2015)
14. S. Pereira, M. R. Correia , E. Pereira, K. P. O’Donnell, C. Trager-Cowan, F. Sweeney, and E. Alves, Physical Review B, Volume 64, 205311(2001).
15. T. Detchprohm, M. Zhu, W. Zhao, Y. Wang, Y. Li, Y. Xia, and C. Wetzel, Compound Semicond. 15, 21 (2009).
16. Mike Cooke, semiconductorTODAY Compounds;Advanced Silicon • Vol. 7 • Issue 8 • October 2013
17. T. Nagatomo, T. Kuboyama, H. Minamino, O. Omoto, Japan J. Appl. Phys. 28 (1989) L1334. 10
18. N. Yoshimoto, T. Matsuoka, T. Sasaki, A. Katsu, Appl. Phys. Lett. 59 (1991) 2251.
19. W. Van der Stricht, I. Moerman, P. Demeester, L. Considine, E.J. Thrush, J.A. Crawley, MRS Internet J. Nitride Semicond. Res. 2 (1997) 16.
20. P. Waltereit, O. Brandt, K.H. Ploog, M.A. Tagliente, L. Tapfer, Phys. Status Solidi b 228 (2001) 49.
21. R. A. Oliver (2016) Critical Assessment 23: Gallium nitride-based visible light-emitting diodes, Materials Science and Technology, 32:8, 737-745