Conference Contribution Details
Mandatory Fields
T C Sadler, H N Li, V Zubialevich, M Conroy, Z Quan, P J Parbrook
4th International Symposium on Growth of III-Nitrides
Routes to Aluminium Gallium Nitride Buffer Layers for UV-LED Growth
St Petersburg, Russia
Poster Presentation
Optional Fields

Ultra violet light emitting diodes (UV-LEDs) have many potential applications in

biomedical engineering, and sterilisation, with the biggest potential market in the

treatment of drinking water. However the efficiencies of UV-LEDS remains very low at

1-2 %, compared to over 70% for blue emitting GaN/InGaN LEDS. Reasons for this

include high threading dislocation densities (TDDs) in the AlGaN buffer layers, low

quantum efficiencies in AlxGa1-xN/AlyGa1-yN quantum wells and the use of absorbing

GaN p-type caps. In this study we compare routes towards achieving relaxed AlGaN

buffers with low TDDs, grown by MOCVD on c-plane sapphire substrates, avoiding

expensive bulk AlN substrates. AlGaN cannot be grown directly on sapphire, which

necessitates that growth is based on an AlN or GaN template. If a GaN layer, this must

be removed along with the sapphire substrate in a flip-chip processing geometry, but

this is advantageous for UVLEDs anyway. This initial AlN or GaN layer must be

followed by layers that relax the strain without creating threading dislocations,

introducing cracks or that lead to surface roughening. We compare three different routes

to growing high quality relaxed Al0.5Ga0.5N buffers: using GaN/AlN superlattices on

both AlN and on GaN substrates, and using low temperature AlN interlayers on a GaN

substrate. Our GaN templates have a lower TDD compared to AlN: 002 FWHM are

250 for GaN and 350 for AlN, and 101 FWHM are 500 for GaN and 1000 for AlN.

AlGaN grown directly on GaN without any strain relief relaxes by creating screw

type dislocations, with a 002 FWHM of 450 and 101 FWHM of 1000. Using a 100

period 0.25 nm GaN / 0.25 nm AlN superlattice prevents cracking in a 1 μm

Al0.5Ga0.5N, but at the consequence of roughening (with values of 2nm for AlN, and 15

nm for AlGaN).

AlN/GaN superlattices grown on GaN will also similarly prevent AlGaN buffers

from cracking, but relax by increasing surface roughness. Superlattices with a small

period of 1 nm AlN /1 nm GaN did not relax the AlGaN layer, where as thicker 5 nm

AlN /5 nm GaN superlattices with a period of 5nm did relax, but by creating a large

number of screw dislocations, the 002 FWHM was 800 and 101 FWHM 900.

We have found the best way to grow relaxed Al0.5Ga0.5N buffers is to use multiple

10 nm thick AlN interlayers grown at the low temperature of 650C, interspaced with

10 nm GaN layers. Using two interlayers gives a smooth AlGaN surface, with only

moderate cracks around the wafer edge. 002 FWHM of 250 and 101 FWHM of 1000,

show that only edge dislocations are created to enable relaxation. When more

interlayers are used the crack density reduces further but more edge dislocations are

created, whereas only using one resulted in the cracks penetrating further towards the

wafer centre.

These buffer layers have been used to grow prototype polarisation-matched

InAlN/AlGaN multiple quantum wells.