![]()
Advances in MBE Selective Area Growth of III-Nitride Nanostructures:
From NanoLEDs to Pseudo Substrates
Steven Albert*, Ana Maria Bengoechea-Encabo*, Francesca Barbagini, David Lopez-Rormero,
Miguel Angel Sanchez-Garcia and Enrique Calleja
ISOM-Dept. Ing. Electronica, ETSIT, Univ. Politecnica, E-28040 Madrid, Spain
[email protected] Pierre Lefebvre
ISOM-Dept. Ing. Electronica, ETSIT, Univ. Politecnica, E-28040 Madrid, Spain
and CNRSâLaboratoire Charles Coulomb (L2C), UMR5221, F-34095 Montpellier, France
Xiang Kong, Uwe Jahn and Achim Trampert
Paul-Drude Institut, Hausvogteiplatz 5-7, 10117 Berlin, Germany
Marcus MĂŒller, Frank Bertram, Gordon Schmidt, Peter Veit, Silke Petzold and JĂŒrgen Christen
Otto-von-Guericke University, UniversitÀtsplatz 2, 39106 Magdeburg, Germany
Philippe De Mierry and Jesus Zuñiga-Perez
CRHEA-CNRS, 06560 Valbonne, France
Received 30 June 2014
Accepted 12 July 2014
The aim of this work is to provide an overview on the recent advances in the selective area growth (SAG) of (In)GaN nanostructures by plasma assisted molecular beam epitaxy, focusing on their potential as building blocks for next generation LEDs.
The first three sections deal with the basic growth mechanisms of GaN SAG and the emission control in the entire ultraviolet to infrared range, including approaches for white light emission, using InGaN disks and thick segments on axial nanocolumns. SAG of axial nanostructures is developed on both GaN/sapphire templates and GaN-buffered Si(111).
As an alternative to axial nanocolumns, section 4 reports on the growth and characterization of InGaN/GaN core-shell structures on an ordered array of top-down patterned GaN microrods. Finally, section 5 reports on the SAG of GaN, with and without InGaN insertion, on semi-polar (11-22) and non-polar (11-20) templates. Upon SAG the high defect density present in the templates is strongly reduced as indicated by a dramatic improvement of the optical properties. In the case of SAG on non-polar (11-22) templates, the formation of nanostructures with a low aspect ratio took place allowing for the fabrication of high-quality, non-polar GaN pseudo-templates by coalescence of these nanostructures.
Keywords: Selective area growth; nanostructures; InGaN; GaN; LED; photoluminescence; white-light emission; single color emission; core-shell; non-polar; semi-polar; pseudo substrates; coalescence.
1. Introduction
Light emitting diodes (LEDs) are anticipated to have great potential as a replacement for traditional lamp-based lighting systems due to a higher efficiency, with the greatest expected impact for high-quality white-light sources. The main white light source up to now is the incandescent light bulb which converts only about 5% of electricity into visible light. Due to that lighting is the second largest user of energy in in-house lighting1 causing greenhouse gas emissions of 1900 Megatons of CO2 per year2. That corresponds to over three times the emissions caused by aircraft traffic3. All these drawbacks led to a gradual ban of incandescent light bulbs by many governments.
The strongest two competitors as a replacement for incandescent lamps are compact fluorescent lamps (CFLs) and non-organic LEDs. Due to the relatively high price of non-organic LEDs, CFLs are the dominant replacement for general household lighting at the moment. They have a four times higher efficiency than incandescent bulbs and a lifetime of up to 10000h (compared to 1000h for incandescent light bulbs). A drawback of CFLs is the use of mercury which is highly toxic. This can apparently be considered as a severe issue considering the fact that most states already banned mercury thermometers. In addition the efficacy of CFLs is not expected to exceed 100 lm/W. Due to that there is a high need to develop highly efficient white light sources that do not contain any toxic materials, making non-organic GaN-based LEDs the perfect lamp choice for the future. Unlike every conventional light source, LEDs directly transfer electrical energy into light. The most common approach for white light generation using LEDs is a blue LED that pumps phosphors that emit at longer wavelengths, i.e. phosphor converted LEDs (pc-LEDs). The simplest pc-LEDs combine a blue LED (λ = 440-460 nm) with a YAG:Ce3+ phosphor (λ = 560). The problem with that approach is a rather low color rendering index (CRI) of 70-80 and a correlated color temperature (CCT) of 4000 to 8000K. With these characteristics these LEDs are only suitable for less demanding applications such as outdoor lighting. For indoor applications CRIs between 80 and 90 are needed4. The figures of merit of a pc-LED can be improved by adding a second red-emitting phosphor such as nitrodosilicate. With this approach commercial white LEDs with CRIs of 90, efficacies of 55 lm/W and CCTs of 3000-6000K have already entered the market. An alternative to pc-LEDs is the RG(Y)B approach in which solely LEDs are used for white light generation. With this approach CRIs higher than 95 as well as increased efficiencies can be achieved5,6. With the RGB approach a dynamic color control is possible that would allow for a real-time spectral tuning. As a main advantage over pc-LEDs, an efficiency improvement due to the absence of color filters or phosphors is expected. Assuming an efficiency of around 25% for each part, efficacies of around 160 lm/W are theoretically possible. The disadvantages of this approach up to now are different efficiencies as well as temperature dependencies of the efficiencies when using different material systems such as GaAlInP materials for the red and InGaN materials for the blue and green, leading to unacceptable variations of the color unless compensated by current drivers and feedback loops for each color which would add extra cost.
The two main obstacles to overcome, before large scale commercialization of RGB-LEDs, are; i) the green gap problem and ii) the efficiency droop of InGaN LEDs which refers to a decrease of efficiency with increasing injection current. The green gap problem refers to the fact that efficient LED emission could not be demonstrated yet across the entire visible spectrum, particularly from green to yellow. A good performance can only be achieved in the blue range using InGaN based LEDs and red range using AlGaInP based LEDs. The efficiency drop of AlGaInP based materials at shorter wavelengths is caused by an indirect band gap crossover in the green-yellow range which can hardly be solved. A solution may be the use of InGaN for bridging the green-yellow gap since it has a direct bandgap across the complete visible range with no intrinsic roadblock to high efficiency optical emission. Up to now most research effort has been focused on InGaN/GaN based c-plane oriented, two-dimensional quantum well (QW) structures. In these planar structures the green gap problem is assumed to be caused by: (i) the reduction of the radiative recombination rate induced by the quantum confined Stark effect due to spontaneous and piezoelectric polarizations, and (ii) the high density of non-radiative defects due to the increasing lattice mismatch (strain) between GaN and InGaN alloys with increasing In content. The strong polarization effects have their origin in the wurtzite crystal structure leading to piezoelectric and spontaneous polarization along the polar [0001] crystal direction causing large (>1MV/cm) electrostatic fields in the QW. At this point it has to be emphasized that the piezoelectric polarization depends on strain, leading to a reduction in radiative recombination rate for higher In contents assuming that InGaN is grown coherently strained on GaN.
In addition to its impact on the polarization, strain also causes a number of additional effects. In general the growth of InGaN requires a rather low growth temperature due to the reduced thermal stability of InN7. At a given growth temperature the strain in an InGaN/GaN heterostructure reduces the incorporation of indium due to compositional lattice pulling effects8,9. Due to that even lower growth temperatures are needed in order to achieve higher In content. The combination of a low growth temperature and an increasing strain energy leads to the formation of a number of structural defects, e.g. point defects10, incorporation of impurities11 and V-defects12,13 causing a decrease of the optical quality of the active region.
A potential solution for all problems discussed above may be the use of ordered nanocolumnar structures as building blocks for next generation LEDs. It has been shown several years ago that dislocation- and strain- free group-III nitrides can be grown on Si(100) and (111), as well as on amorphous SiO2 substrates in the form of one-dimensional structures in a self-assembled fashion14-18. When growing nanocolumnar InGaN/GaN heterostructures a higher strain can be accommodated (due to the strain relief by lateral relaxation) before dislocations generate19. However, LEDs based on InGaN/GaN self-assembled nanocolumns (NCs) always show polychromatic emission that derives from an inhomogeneous axial and radial strain and In% distribution, geometry dispersion, and an inherent tendency of InGaN alloys to develop composition fluctuations20,21. In addition, efficient and reliable LEDs based on self-assembled NCs are hindered by limitations due to strong electrical characteristics dispersion22.
In order to overcome these limitations, selective area growth (SAG) of GaN23-26 has been developed. It allows the growth of NCs with well-controlled position and diameter, resulting in periodic arrays of NCs with very little morphology dispersion.
This review summarizes rec...
