The Challenges For Going Green
Fans of high definition movies and Sony Playstations are grateful to the inventors and producers of the GaN laser. However, what they are probably unaware of is that these scientists and engineers that refined and developed this violet-emitting laser, which can read the ones and zeros off of optical discs, had to deal with a multitude of tough challenges that do not exist in the traditional III-Vs: Limited availability of native substrates; higher growth temperatures and low chemical reactivity at room temperature; polarization effects resulting from a wurtzite crystal structure; heavy effective mass; high acceptor ionization energy; and limited availability of strain-free heterostructures.
More recently, these issues have hampered efforts within the nitride community to extend laser emission to longer wavelengths, a spectral region known as the green gap. Green lasers are wanted for colour projection systems that combine the output of red, green and blue sources. In addition, these single-chip lasers could win sales in a variety of defence, biomedical, industrial and instrumentation applications.
Chip designers trying to select an architecture for a green laser come up against three major impediments: Strain, because it is not possible to make the device with a set of lattice-matched materials; the need to operate at the high current densities to reach optical amplification, a pre-requisite for lasing; and injecting enough carriers uniformly into an active region plagued with an awkward band structure. The remainder of this article will detail each of these three challenges, before offering some insights into ways to overcome them.
Coping with strain
Lattice matching with nitride materials is very tricky. Differences in crystal structure prevent pairing III-Ns with III-Vs – the wurtzite structure of the former is significantly different from the zinc blende structure of the latter (see Figure 1). Substituting another group V element for nitrogen is not easy, either, so there are only three ternary alloys available. Two are AlGaN and InGaN, which cannot be lattice matched, and the third is AlInN, which is rarely used due to the challenges associated with the growth of uniform material.
Figure 1. Differences in crystal structure makes the alloying of wurtzite III/N with traditional III/V compounds challenging
There is also a family of quaternary alloys that allow some degree of bandgap tuning for a given lattice constant: AlInGaN. But this class of material is rarely used, due to strong fluctuations in composition and a tendency for morphology roughening resulting from a large difference in the preferred growth temperatures of AlGaN and InGaN.
Consequently, designers of deep blue and green nitride lasers tend to select AlGaN for the cladding layers and InGaN for the waveguides. To prevent excessive strain, concentrations of aluminium and indium in AlGaN and InGaN are kept within a few percent of each other. However, this leads to a relatively low refractive index contrast between the waveguide core and claddings. Making matters worse, at longer wavelengths, the differences in refractive index between given alloy compositions diminish.
One choice facing every nitride laser designer is that of substrate orientation. Traditionally, nitride lasers have been formed on the c-plane, an orientation that allows for significant strain without defect formation, thanks to a relatively large force needed for glide along available glide planes.
Relaxation in nitrides due to dislocation glide occurs at far lower levels of strain when lasers are formed on semi-polar planes – planes in a wurtzite crystal tilted by an
with respect to the c-plane. There are many semi-polar planes to work with, and it is possible to find orientations that allow sufficient indium incorporation to realise green emission. More importantly, semi-polar planes enable the formation of epitaxial structures with higher optical gain. This is a key advantage for making green lasers, where optical gain is limited (this is discussed in detail later on).
When nitride layers are deposited on semi-polar substrates, strain leads to misfit dislocation arrays at AlGaN/InGaN interfaces between layers over 100 nmthick that have aluminium and indium contents of just a few percent. If the impact of these misfit dislocations is similar to that in other III-V devices, they will jeopardize reliability.
Our team of researchers at Corning has found that misfit dislocations also act as strong non-radiative recombination channels, and in addition they degrade surface morphology, just like they do in other III-V structures. Fortunately, there are special tricks for sidestepping these relaxation effects and ultimately achieving sufficient optical confinement in semi-polar green lasers with good quantum efficiency. Thanks to these tricks, it is possible to realise high optical gain in green lasers.
Keeping misfit dislocations away from the active region is one way to manage strain. Do this, and strain relaxation is actually beneficial. The key is to grow an InGaN waveguide core over the cladding in a way that leads to dislocations forming at this interface. This increases the lattice constant of the InGaN layer, and it enables the subsequent growth of a green InGaN quantum well with much lower stress, thereby preventing relaxation of this layer.
Figure 2. Green laser diodes are composed of latticemismatched heterostructures because most of the conventional III/N alloys are lattice mismatched with GaN substrates
This must be followed up with the growth of a second InGaN waveguiding layer on top of the well that has the same lattice constant as the bottom waveguide. Doing this enables the coherent growth of a multiquantum well region sandwiched between InGaN waveguiding layers.
Once this is done, engineers must then deposit a lattice-mismatched, relaxed top cladding layer, which will lead to the formation of another array of misfit dislocations, this time on the lower interface of the upper cladding. With this approach, dislocations are only located in the interfaces between waveguiding and cladding layers, where they are too far away from the active region to contribute to non-radiative recombination losses. It is worth noting that the possibility of instantly forming a misfit dislocation array at the interface and growing a relaxed and low-defectdensity layer right over it is quite unique for semipolar III-N heterostructures.
There have also been reports of strain relaxation management in III-V zinc-blende structures, such as near-infrared lasers with In(Ga)As quantum wells, quantum dots grown on GaAs substrates, and multijunction solar cells. However, in order to avoid defects in active regions, those realizations require the growing of bulk transition layers or superlattices, instead of using single interface, like in semipolar III-Ns.
Alternatively, misfit dislocations can be completely avoided with strain-compensating layers. The trick is to balance compressive strain in the wells and the InGaN waveguiding layers with tensile strain in Al(In)GaN barriers. Apparently, when the indium concentration in InGaN layers of the waveguide core is sufficiently high, strain compensation is possible without degrading optical confinement.
This is thanks to one beneficial characteristic of InGaN: Its refractive index increases superlinearly with indium concentration. In other words, the refractive index increases much faster when the separation between the lasing photon energy and the InGaN material bandgap is smaller. Intuitively, simple insertion of the Al(In)GaN layers inside the InGaN waveguide core would reduce its average refractive index. However, due to strain compensation it in fact enables the use more indium in the waveguide, which overcompensates the refractive index reduction, so it is still higher than for the design without strain compensation.
Cranking up the current
The second major challenge facing the developers of green lasers is to design and build a device capable of operating at the very high current densities required for light amplification. In InGaN quantum wells, carrier density is relatively high, due to the high effective mass of both types of carrier – masses of holes and electrons in InGaN are more than 1.4me and 0.2me, respectively, compared to values of just 0.51me and 0.063me for electrons and holes in GaAs (me is the mass of a free electron). The higher electron and hole effective masses lead to a higher density of electron and hole states. In turn, the higher density of electron and hole states in the InGaN quantum well leads to a higher transparency carrier density, a pre-requisite for lasing, that is more than twice that required for laser diodes based on InGaAs quantum wells.
This hike in transparency carrier density has unwanted ramifications. Recombination current is a super-linear function of carrier concentration – especially at high current densities where non-radiative Auger recombination dominates – so the current density needed to reach transparency in an InGaN quantum well is several times higher than it is in one made from InGaAs (see Figure 3).
Figure 3. (a) The charge carrier density needed to reach transparency is much higher for InGaN QWs – especially c-plane ones – than III/As QWs. (b) As a result, InGaN QWs need much higher pumping for light amplification. Despite this, the differential gain is much lower in green InGaN QWs. (This figure drew on data from P. Blood et al. J. Appl. Phys. 70 1144 (1991))
Making matters worse, the increases in indium content that push the emission of InGaN quantum wells towards the green also leads to rapid reductions in differential gain. This is not just the result of a hike in non-radiative recombination – there is also significant inhomogeneous line broadening resulting from InGaN alloy fluctuations when the molar concentration of indium in the ternary is around 30 percent, the fraction required for green emission.
Part of the reason behind the high current densities required for lasing in long-wavelength nitride lasers is that the wurtzite III-N crystal has a polar nature. The polarization fields that result from this particular structure reduce electron-hole overlap in the well, a phenomena known as the quantum confined Stark effect (QCSE). To overcome optical losses, current densities in blue and green laser diodes need to hit 2 kA cm-2 and over 4 kA cm-2, respectively. These values are an order of magnitude higher than those for lasers based on InGaAs quantum wells.
The low optical gain found in green lasers can be partially addressed with semi-polar substrates: Differential optical gain doubles when a green InGaN quantum well is grown on a semi-polar orientation, rather than the conventional polar one. This benefit of semi-polar orientation is due to a reduction in the QCSE and anisotropic optical gain by the breaking of the 90-degree rotational symmetry of the wurtzite crystal semi-polar plane – a property not associated with conventional III-V compounds, which have a cubic symmetry.
Working with semi-polar lasers is relatively new, and this type of epitaxial structure is still immature. Limits of this approach are still being explored, including the strain management techniques already outlined above. One downside of working with semi-polar lasers is that optical gain is, in most cases, only high in one direction – and this direction is not always favourable for cleaved facet formation. This state of affairs also has its origins in the breaking of plane rotational symmetry.
Low gain is only partly to blame for high-threshold currents, which also result from high optical loss. In the red and infrared spectral ranges served by III-V lasers, intersubband absorption and free carrier absorption are the primary causes of optical loss, and they can be trimmed to less than 1 cm-1 for 980 nm lasing. With III-N lasers operating in blue-green, however, losses are ten times higher, due to acceptor-bound absorption.
One undesirable trait associated with III-Ns is the high activation energy of acceptors. Very high acceptor concentrations are needed to realise a desired hole conductivity. With GaN and its related alloys, the only practical acceptor is magnesium, which has an activation energy in excess of 160 meV – at least four times that associated with a range of elements for doping more traditional III-Vs. This means that when using magnesium, if the acceptor concentration is about 1x1019 cm-3, less than 2 percent of holes contribute to conductivity. So very high levels of magnesium are needed to form layers that have reasonable pconductivity, a necessity for making devices with a low operating voltage. However, a penalty must be paid – high optical absorption.
Figure 4. Deep charge carrier confinement inside the QW leaves limited opportunity for carrier distribution among several QWs and is the root cause of injection asymmetry. Heavy magnesium doping is needed to obtain p-conductivity. Asymmetry of carrier injection in the active region causes electron leakage to a heavily p-doped region where carriers can recombine non-radiatively if this process is not blocked
Lasers designers can use several tricks to trim the total optical loss of their devices. They can select a relatively high reflectivity for the front mirror, which also leads to a reduced slope efficiency of output light. In addition, they can keep the p-doped region away from the optical mode, while still ensuring that there is sufficient hole injection into the quantum well and the operating voltage of the laser is reasonable. One neat way to do this is to use an asymmetrical waveguide refractive index profile, which shifts the optical mode towards the n-side. Although this slightly reduces optical overlap with the active region, it substantially cuts overlap with the p-layers.
Injecting electrons and holes
The third major challenge facing developers of green III-N lasers is the injection of carriers into quantum wells that are deep and have a band structure that is severely distorted by the QCSE.
With infrared laser diodes, good carrier transport results from the use of a graded-index, separate-confinement heterostructure. In this class of device, the bandgap increases gradually in both directions away from the quantum well region, and carriers can move in and out of each quantum well, leading to similar populations in every well. However, electrons and holes are still confined within the multi-quantum well region, because this is surrounded by a bandgap gradient that creates electric fields preventing carrier out-diffusion. What’s more, this approach also leads to optical confinement, because the material furthest from the wells has the widest bandgap and the lowest refractive index.
Unfortunately, up until now it has proved impossible to transfer this elegant design to green nitride lasers. In our view, this is because of peculiarities associated with the bandstructures of GaN and InGaN used in the barriers and wells. These materials have bandgaps of 3.49 eV and 2.34 eV, so the total band offset exceeds 1 eV. It is hard to determine the precise bandstructure, due to significant variations in the values for the offset ratio of the conduction band to the valence band. However, these differences are not important, because there is no question that the barrier height (given in energy ) for each charge carrier exceeds the value of kBT by a factor of at least ten.
To replicate the graded-index structures found in conventional III-V lasers, engineers must fabricate InGaN barriers with a significantly narrower bandgap than GaN. But this would increase the total amount of indium in the active region, adding compressive strain to wells that are already suffering from this. So the barriers must contain very little or no indium. This means that the rate for carrier thermal escape from a quantum well ground state, which rapidly diminishes with increases in barrier height, is far slower than carrier recombination. Consequently, carriers are trapped in one well until they recombine, and they only travel across the wells by ballistic fly-over.
The upshot of all of this is poor carrier distributions, particularly in c-plane lasers. Here, the piezoelectric effects create potential spikes in the barriers, which lead to a drag on ballistic carrier motion. Holes fare worse than electrons, due to a higher average effective mass and lower mobility, and this results in the accumulation of carriers in the quantum wells near the p-side.
In c-plane green laser diodes and LEDs the well closest to the p-side tends to be the one most populated with carrier. This is part of the reason why the p-n junction is effectively shifted toward the p-side, promoting electron leakage to the p-doped region. Any electrons that get this far undergo parasitic nonradiative recombination. To prevent this from degrading device behaviour, an electron-blocking layer is inserted between the wells and the p-type region. This thin layer stems the flow of electrons to the p-side, but partially blocks hole injection into the quantum wells and fails to address the problem of non-uniform carrier distribution within the active region. When we tried to build c-plane green lasers with several quantum wells, only the one or two closest to the p-side provided optical gain, with the rest absorbing light, because they are under-pumped, unless extremely high current is applied.
Semi-polar lasers don’t suffer the same fate. Hole injection into several quantum wells is much easier, due to a reduction in the strength of the built-in polarization fields. These weaker fields lower barrier spikes, making carrier fly-over more uniform and ultimately leading to a more even distribution of electrons and holes within the wells.
We have found that this superior band structure in semipolar lasers also allows designers to dispense with the electron-blocking layer, which is not used in III-V infrared lasers. Thanks to a deeper penetration of holes in the multiple quantum well region, fewer electrons are attracted to the p-layers, and there is much less shifting of the position of the p-n junction.
There is a silver lining associated with deep carrier confinement in the quantum wells: Reduced sensitivity to changes in temperature. The measure of this sensitivity in laser diodes is the threshold current characteristic temperature T0. In arsenide- or phosphidebased lasers, carrier confinement is typically 0.2 eV or less, and a significant proportion of carriers leak away from the active region at normal operating temperatures. With green nitride lasers, the far stronger carrier confinement prevents any electron escape from quantum wells with temperatures below 400K.
Switching from non-polar to semi-polar green lasers leads to an even higher value of T0. We believe that this stems from a combination of the deep carrier confinement found in all forms of nitride laser, plus a particularly large separation of quantum well quantization states. This leads to a lower thermal population of the higher energy bands.
It is clear that III-N materials hold great promise for covering a wide spectral range. When it comes to green III-N lasers, the challenges are quite different to those facing the designers of III-V red and infrared lasers, so a different approach is needed. With III-N green lasers, strain comes into play, there is low optical gain and high optical loss to deal with, and deep carrier confinement combats uniform carrier distributions in the multiple quantum well – but it has a good side too, in the form of diminished device temperature sensitivity. Understanding all of this, and how to navigate a path through these obstacles, holds the key to improving the performance of green lasers.