Light emitting diodes (LEDs)
p-electrode p-GaN p-Al6.3Gq.7N Ing.rGao.rN
n-Inn.s2Gas 96N n-Al9.3Gas.7N
333
n-GaN GaN bufTer layer
Fig. 13.4
Typical layer structure of a QW based nirridc LED.
wavelength*. Forming and consolidating the buffer layer is almost an afi form. A black art many say. Howevcr, an elaborate process called epitaxial lateral overgrowth (ELOG) involving chemical vapour deposition (CVD) of a layer of SiOz, subsequently penetrated by windows a few micrometres rvide by conventional masking and etching. This structure is ovcrlaid by tlrther lnGaN MOCVD gading and growth layers. The distortion causes lots of extra strain. but in some regions of the resulting or.'erlaver thc dislocation densitv is below 106cm-2. Unfortunately. there is verv little change in device performance across such a layeq except that for pure GaN" u'hich is not verv- luminescent. there is a marked improvement at lou' defect de nsit-v-. So the practical ansu er is to keep some In and no ELOG. This is bome out b-"" Fig. I 3.5 r.r hich show's how the quantum efficiency varies with In content. fiom u'ork clone a fc,uv years ago. Recently the quantum efficiencv of hiGaN LEDs reached 30o,i, at violet-blue wavelengths at 20 mA current. a po\'" er of 2 I rnW emitted in the blue, and 7 mW in the green, corresponding to about 60-30 lumens W-l . High brightness red LEDs are up to 50% e{hcient. This has led the wav to commercial usage of LEDs. There are large colour displays.*signal and traffic lights. automotive uses including stop and rear lights, also for cycles. The estimated life of InGaN LEDs at present. from accelerated and projected life testing is about l0 years continuous use. For an average room light duty cycle this gives 60 years! The long life. lorv maintenance, is also the main selling point for traffic lights. It makes some local authorities, even in the U.K., look further forward than the next financial year, normally their absolute limit to thinking ahead. So how can we assess the state of InGaN LED technology in early 2009? The nitrides are still difficult to grow as reasonable epitaxial layers, and the very successful LEDs have been made on small crystals. typically 5 nm thick active layer. There is no good theory of why they work. So we are in a situation of technological success leading to a rapidly expanding semiconductor lighting industry without a proper scientific backing. This has happened before, maybe steam engines were somewhat like this before thermodynamics caught up. But following the precise science/technology advance of solid state circuit elechonics, it is a bit of a shock. We rnust hope that all will become clearer
Not only composition defines
the
wavelength. The nitrides are very noncentro symmetric and have a large piezo-
electric constant (Section 10. I I ), so the strains and dislocations cause a big internal field. this gives wavelength shift by the Stark effect, particularly in the quantum rvell structure (Section 13.10.3
and 4).
100
; :80
:60
a
:oo
'2
zo
0'
36'7 369
'
3'7t 3'73 375 311 379
'
'
'
'
'
'
381
Emission peak wavelength (nm)
Fig. 13.5
Relative output power of uv InGaN SQW LEDs as a function of emission peak wavelength. The dots show increasing In mole fractions from practrcally 0% on the left to 4% on the right.
display that can be described
stage
as
large rvithout any exaggeration rvas used
to illuminate the
at the opening
ceremony of thc Beijing Olympics in 2008. lt employed 44000 LEDs in a display having dimensions of36 m x 147
m.
when better material is available. Meanwhile, we can conjeclure that the necessary small quantity of In in GaN leads to localised deep levels that
