Figure 5.10 shows the layer structure and circuit diagram for a typical elec- troluminescent device. The device consists of several epitaxial layers grown on top of a thick crystal substrate. The epitaxial layers consist of a p-n diode
with a thin active region at the junction. The diode is operated in forward
bias with a current flowing from the p-layer through to the n-layer underneath.
The luminescence is generated in the active region by the recombination of electrons that flow in from the n-type layer with holes that flow in from the p-type side.
The microscopic mechanisms that determine the emission spectrum are ex- actly the same as the ones discussed in the context of photoluminescence in Sections 5.3.1—-5.3.3. The only difference is that the carriers are injected electrically rather than optically. At room temperature we therefore expect a single emission line of width ~ kgT at the band gap energy Ey. Hence Ey, determines the emission wavelength.
We pointed out in Section 5.2 that the radiative efficiency of indirect gap materials is low. Commercial electroluminescent devices are therefore made from direct gap compounds. Any direct gap semiconductor can, in principle, be used for the active region, but in practice only a few materials are commonly employed. The main factors that determine the choice of the material are:
(1) the size of the band gap;
(2) constraints relating to lattice matching;
(3) the ease of p-type doping.
5.4 Electroluminescence 103
sample in cryostat
laser
mirror collection PL lenses
entrance slit
/
spectrometer
Fig. 5.9 Experimental arrangement used for the observation of photoluminescence (PL) spectra. The sample is excited with a laser or lamp with photon energy greater than the band gap. The spectrum is obtained by recording the emission as a function of wave- length using a computer-controlled spec- trometer and detector. In photoluminescence excitation spectroscopy (PLE), the detection wavelength is fixed and the excitation wave- length is scanned. In time-resolved photolu- minescence spectroscopy, a pulsed laser is used, and the emission at each wavelength is recorded on a fast detector as a function of time after the pulse has arrived.
We concentrate here on electroluminescence from inorganic semiconductors. In Sec- tion 8.6, we will discuss an alternative ap- proach to making visible LEDs using the bur- geoning technology of light emitting poly- mers.
In the past, some indirect gap materials have been used for lack of practical direct gap alternatives. For example, gallium phosphide based devices have been used for yellow and green LEDs, while silicon carbide has been used for the blue spectral region. The active regions of these devices are often doped to promote recombination via impurities and hence increase the luminescent quantum ef- ficiency. The advent of efficient direct gap nitride LEDs in 1995 has made these indirect gap devices obsolete.
104 Luminescence
Fig. 5.10 (a) Layer structure and (b) circuit diagram for a typical electroluminescent de- vice. The thin active region at the junction of the p- and n-layers is not shown, and the dimensions are not drawn to scale. The thickness of the epitaxial layers will be only
~ lym, whereas the substrate might be
~ 500 um thick. The lateral dimensions of the device might be several millimetres.
Epitaxy is the name given to any crystal growth technique involving the formation of thin high quality layers on top of a thicker substrate crystal. The substrate acts as a support for the epitaxial layers, and also serves as a heat sink where needed. There are a number of techniques commonly used.
Medium quality crystals are grown by liq- uid phase epitaxy (LPE), but the highest quality materials are grown by metal-organic vapour phase epitaxy (MOVPE) — also called metal-organic chemical vapour depo- sition (MOCVD) — and molecular beam epitaxy (MBE). These techniques are crucial to the successful growth of the high quality quantum well structures described in the next chapter.
holes m,~ current
g
p-type —,. + epitaxial
n-type layers
electrons substrate
(a) (b)
The first point is obvious: the band gap determines the emission wavelength.
The second and third points are practical ones relating to the way the devices are made. These are discussed further below.
The term lattice matching relates to the relative size of the lattice parameters of the epitaxial layers and the substrate. The thin epitaxial layers are grown on top of a substrate crystal, as shown in Fig. 5.10(a). This is done for practical reasons. It is hard to grow large crystals with sufficient purity to emit light efficiently. We therefore grow thin ultra-pure layers on top of a substrate of poorer optical quality by various techniques of crystal epitaxy. The crystal growth conditions constrain the epitaxial layers to form with the same unit cell size as the substrate crystal. This means that the epitaxial layers will be highly strained unless they have the same lattice constant as the substrate, that is, that we have ‘lattice matching’ between the epitaxial layers and the substrate. If this condition is not satisfied, crystal dislocations are likely to form in the epitaxial layers, which would severely degrade the optical quality.
Figure 5.11 plots the band gap of a number of IIJ-V materials used in electroluminescent devices against their lattice constant. The lattice constants of the commonly used substrate crystal are indicated at the top of the figure.
The materials separate into two distinct groups. On the right we have the arsenic and phosphorous compounds which crystallize with the cubic zinc blende structure, while on the left we have the nitride compounds which have the hexagonal wurtzite structure. We will discuss the cubic materials first, and then consider the nitrides afterwards.
For many years, the optoelectronics industry has been mainly based on
GaAs. GaAs emits in the infrared at 870 nm, and by mixing it with AlAs to form the alloy Al,Ga;_,As, light emitters for the range 630-870 nm can be produced. (See Example 5.1.) Lattice-matched AlGaAs can easily be grown on GaAs substrates because of the convenient coincidence that the lattice con- stants of GaAs and AlAs are almost identical. AlGaAs emitters are widely used in local area fibre optic networks operating around 850 nm, and also for red LEDs.
AlGaAs is an example of a ‘ternary’ alloy which contains three elements.
‘Quaternary’ alloys such as (AlyGai_y)yIni_„P can also be formed. All of these arsenic and phosphorous alloys suffer from the problem that they become indirect as the band gap gets larger. This limits their usefulness to the red and near-infrared spectral range.
Applications in the fibre optics industry require light emitting devices that operate around 1.3 ~m and 1.55 wm. These are the wavelengths where silica
substrate materials
SiC sapphire GaAs
i em
hexagonal cubic
6F AIN đ direct band gap ơ
O indirect band gap
4 be =-
GaN AlAs
blue
Band gap (eV)
red
<—) fibre
1 <—J optics
3 4 5 6
Lattice constant (A)
fibres have the lowest dispersion and loss respectively. Emitters for these wave- lengths tend to be made from the quaternary alloy Ga, In;_,As,P _,. Lattice matching to InP substrates can be achieved if x ~ 0.47y. This allows a whole range of direct gap compounds to be made with emission wavelengths varying from 0.92 ¿m to 1.65 wm. See Table 5.1.
Until fairly recently, it has been very difficult to make efficient electrolumi- nescent devices for the green and blue spectral regions using III-V compounds.
This is because of the problem that has already been mentioned, namely that the arsenic and phosphorous compounds become indirect as the band gap gets larger. However, in 1995 Shuji Nakamura at Nichia Chemical Industries in Japan made an important breakthrough and reported the successful develop- ment of LEDs based on gallium nitride compounds. GaN has a direct band gap of 3.5 eV at 4 K (see Fig. 5.3) and 3.4 eV at room temperature. By alloying it with InN, which has a direct gap of 1.9 eV at 300 K, the emission wavelength.
can be varied between 360 nm (ultraviolet) and 650 nm (red). This enables
the entire visible spectrum to be covered using nitrides for the blue and green colours, and AlGalInP alloys for the reds.
It is interesting to consider why it took so long to develop the nitride devices.
It was well known that the nitrides would in principle make good blue/green emitters, but no commercial devices were available. The reason for this relates to the third point on our list of factors affecting the choice of electrolumines- cent materials, namely the difficulty of p-type doping. This is a problem that has also dogged other wide band gap materials. For example, the direct gap II-VI compounds like ZnSe and CdSe should also, in principle, make good LEDs for the blue/green/yellow spectral regions, but they have never found widespread commercial application due to the doping problem.
P-type doping is difficult in wide band gap semiconductors because they have very deep acceptor levels. The energies of the acceptors are given by eqn 7.29 with m= replaced by m/. The high value of m; and the relatively small
5.4 Electroluminescence 105
Fig. 5.11 Band gap of selected III-V semi- conductors as a function of the their lattice constant. The materials included in the dia- gram are the ones commonly used for mak- ing LEDs and laser diodes. The lattice con- stants of readily available substrate crystals are indicated along the top axis. The nitride materials on the left grow with the hexagonal wurtzite structure, whereas the phosphides and arsenides on the right have the cubic zinc blende structure. After [2].
Table 5.1 Band gap energy Eg and emission wavelength Ag for several com- positions of the direct band gap qua- ternary III-V alloy GayInj_,AsyPj_y.
The compositions indicated all satisfy the lattice-matching condition for InP substrates, namely x + 0.47y. After [2].
x y Eg (eV) = Ag (um)
0 0 1.35 0.92
0.27 0.58 0.95 1.30
0.40 0.85 0.80 1.55
0.47 1 0.75 1.65
We will see in Section 9.5 that the nitride emitters can also be combined with phos- phors to make efficient white light sources.
There is another point that is surprising about Nakamura’s development of nitride LEDs.
It is apparent from Fig. 5.11 that the best substrate material from the point of view of lattice-matching considerations is silicon carbide. However, SiC is very expensive, and the commercial devices tend to be grown on the cheaper sapphire substrates. This com- bination is far from satisfying the lattice- matching condition, but the devices still work very efficiently. One factor that has made this possible is the growth of a thick ‘buffer’ layer
between the substrate and the active region,
to prevent the crystal dislocations affecting the light-emitting regions too adversely.
106 Luminescence
Fig. 5.12 Band diagram of a light emitting diode at (a) zero bias, and (b) forward bias Vo ~ Eg/e. The diode consists of a p-n diode with heavily doped p and n regions.
The dashed lines indicate the positions of the Fermi levels in the p and n regions. Light is emitted in (b) when the electrons in the n- region recombine with holes in the p-region at the junction.
Note that this is a different type of degen- eracy to that considered in Section 5.3.3. In this case we have full thermal equilibrium at Vo = 0, and there is a unique Fermi energy at each point in the device. Degeneracy here means that the carrier density produced by the doping is so large that the Fermi energies are positive with respect to the band edges.
depletion region 1<—_—>1
P
(b) V, = +E,/e
value of €; increases the acceptor energies, and hence reduces the number of holes which are thermally excited into the valence band at room temperature.
This last point follows from the Boltzmann factor (eqn 5.11) with E equal to the acceptor binding energy, which is significantly larger than kgT. The low hole density gives the layers a high resistivity, which causes ohmic heating when the current flows and hence device failure. Nakamura’s breakthrough came after discovering new techniques to activate the holes in p-type GaN by annealing the layers in nitrogen at 700 °C.
In the next chapter we will describe how the use of quantum well layers has led to further developments in the field of electroluminescent materials. In fact, many commercial devices now routinely use quantum wells in the active region. This is especially true in laser diodes, but it is also increasingly so for
LEDs as well. |