Getting to grips with nitride laser failures
Shipments of nitride lasers are going up and up. These devices are already being used for optical storage systems, such as Blu-ray players, and as time goes on they will also start to be deployed in laser projection systems based on red, green and blue lasers, plus a variety of biomedical systems.
Today, Blu-ray players and recorders represent by far the biggest market for the InGaN-based laser. In this application 405 nm sources are focused to a spot size of 580 nm and used to extract data from discs storing up to 25 GB per layer. Reading the data from the discs requires just 5-10 mW, but writing demands far higher powers. 100-300 mW sources are not uncommon, and even higher powers are needed to realise really fast writing speeds, because then the laser has even less time to transfer energy to the media.
Reaching high optical power levels requires operating at current densities of up to 10 kA/cm2 and electrical input powers in excess of 1W. The effective penetration of these devices into commercial applications is governed by their reliability; while some studies have demonstrated lifetimes of several thousand hours, device reliability during operation at high output power levels is still a critical issue. Many research groups around the world are trying to understand the cause of laser failure and putting forward conflicting views regarding its primary cause.
The first generation of laser diodes, which were made in the 1990s, had very short lifetimes. In many cases high dislocation densities were to blame, which led to catastrophic failure. To address this and other problems related to dislocation densities, improved growth techniques such as Epitaxial Lateral Over Growth (ELOG) were introduced to improve crystal quality. Dislocation densities then fell to 106 cm-2 or less.
Despite these important improvements, InGaN-based laser diodes still suffer from a gradual decrease in device performance during operation. This weakness, which is known as parametric degradation, is something of a hot topic within the nitride laser community and is being avidly discussed at conferences and in the academic literature.
Is it this, that or something else?
One view is that InGaN-based laser failure stems from degradation at the facets, which leads to increasing mirror losses, higher threshold currents and reduced slope efficiency. Others argue that over time there is an increase in point defect density in the active layer, which increases the non-radiative recombination rate and ultimately the threshold current density, while there is also a school of thought that as the laser is driven, current confinement under the ridge worsens, inducing an increase in the threshold current and a shift in emission wavelength. Meanwhile, some are saying that operating the laser for many hours leads to the formation of dislocations in the active region, and subsequently the generation of dark regions, as recently demonstrated for MBE-grown samples.
The lack of consensus over the cause of parametric degradation is partly because the physical mechanisms responsible for the degradation of InGaN-based laser diodes have not been univocally identified. This is generating important discussion in the GaN-laser community.
Our group at the University of Padova, Italy, in close co-operation with Panasonic Corporation, has been able to demonstrate that the increase in non-radiative recombination – due to the generation of point defects – plays a major role in determining the degradation of InGaN-based lasers. We have also shown that long term degradation of lasers is not strongly related to the degradation of the facets, nor to the worsening of the confinement or injection efficiency. Our results strongly support the hypothesis that a necessary step towards the development of highly reliable laser diodes is the optimisation of the active layer, including its radiative efficiency, over the full lifetime of the device.
We arrived at these conclusions after extensively comparing the electro-optical degradation of laser diodes and LED-like samples – devices with the same epitaxial structure as the lasers, but devoid of ridges and facets. Results indicate that the same degradation mechanism, namely the decrease in internal efficiency, is present both in laser diodes and LED-like samples. In other words, devices with strong geometrical constraints, mirrors and cavities degrade in the same way as simple structures with no ridges or facets.
Analysis was carried out on triple-quantum-well devices grown on a GaN substrate and emitting at 405 nm (Blu- Ray technology). A split-wafer experiment was designed as follows (see Figure 1): half of the wafer was processed in order to obtain laser diodes with 600 μm-long cavities; the other half was used to yield LEDs with an area of 75 x 200 μm2. Devices were submitted to constant current stress, with current densities in the range 2-4 kA/cm2, and a case temperature equal to 75 °C.
Fig. 1: The structure of the lasers analysed within this work, and of the split-wafer experiment that was carried out to study laser degradation
The most noticeable consequence of this stress is the increase in laser diode threshold current (Figure 2). No significant variation in slope efficiency was detected after stress, demonstrating that carrier injection efficiency is unaffected by ageing. Furthermore, during ageing, threshold current increases with the square-root of stress time. This result has been described by several research groups, and indicates that a diffusion process is involved in degradation.
Fig. 2: Degradation of the optical characteristics of one of the analysed lasers. Inset: threshold current increase measured during stress time on one of the analysed samples. Threshold current increases according to the square-root of stress time, indicating that a diffusion process is involved in degradation
Many groups have tried to identify the impurity or defect involved in this diffusion process: both an experimental effort based on Secondary Ion Mass Spectroscopy (SIMS) and a theoretical approach based on the analysis of the degradation kinetics were recently adopted. SIMS investigation by Piotr Perlin and colleagues from Unipress, Poland, which was reported at the EMRS Fall meeting in 2010, indicated that degradation is not correlated to the instability of magnesium and hydrogen, two species that were previously indicated to be involved in the degradation process.
Meanwhile, the theoretical approach taken by Kenji Orita and co-workers from Panasonic demonstrated that under nominal operating conditions the diffusion coefficient of the impurity involved in the degradation process is around 1.9 x 1019 cm2/s. This is very high compared to the diffusion coefficient of the most common impurities in GaN.
Getting to the point
We also found that our LED-like samples exhibited a significant degradation during constant current stress: Degradation was found to be more prominent at low measuring current levels, indicating that stress induces an increase in non-radiative recombination. Furthermore, laser diodes and LED-like samples submitted to constant current stress showed similar degradation kinetics, strongly suggesting that the two different kinds of devices degrade due to the same physical process, which is the generation of point defects.
Fig. 3: Recombination dynamics in laser diodes
It is possible to gain further insight into the degradation process by performing extensive characterisation of the electrical characteristics of the devices during stress time. We have found that stress induces a significant increase in defect-related current components, such as generation, recombination and reverse currents. This suggests that the defectiveness of the active layer is increasing during stress time. Interestingly, defect-related current was found to increase according to the square-root of stress time, exactly as we have found for the optical power degradation.
Our set of results allows us to draw several important diodes conclusions regarding the origin of degradation. First, it is clear that the slope efficiency of the lasers is unaffected by stress, which indicates that long-term degradation of the device is not related to a decrease in injection efficiency (the proportion of electrons that are injected into the quantum wells). We can also say that a diffusion process can be responsible for the measured optical degradation, because degradation kinetics follow the square root of stress time.
Another point worth noting is that the optical degradation of LED-like samples is more prominent at low measuring currents, indicating an increase in non-radiative recombination components under these conditions within the active layer. Since degradation occurs both in laser diodes and in LED-like samples, we can argue that the laser’s ageing is not due to the high optical field present in the cavity. And finally, we note that optical degradation is strongly correlated to the increase in defect-related current components: Stress, therefore, is considered to induce the generation/propagation of defects within the active layer of the samples.
Is current confinement to blame?
Recently, Jens Müller and co-workers from Osram suggested that the optical degradation of lasers could be related to the worsening of current confinement under the ridge, which could lead to an increase in threshold current.
However, our study demonstrates that degradation occurs in both laser diodes and in LED-like samples. This strongly supports the hypothesis that degradation is not correlated to the geometry of the devices, nor to a degradation of the mirrors.
Despite the important results presented recently in the literature, the challenge is still open: Several issues must be solved to identify the origin of the degradation of InGaN-based laser diodes. First of all, although we know that degradation is related to a diffusion process, the impurity or defect involved in diffusion has not been univocally identified. In addition, although we indicate that non-radiative recombination plays a major role in the degradation process, the defect or deep-level responsible for recombination is still to be exposed. How to identify this villain remains a big question. Finally, most of the studies on laser reliability have been carried out on Blu- Ray devices, which have a relatively good lattice quality. What about longer-wavelength samples? Will green lasers show similar degradation characteristics? The road to the optimisation of InGaN-laser diodes is still long.
© 2011 Angel Business Communications. Permission required.
Further reading
T. Schoedl et. al. J. Appl. Phys. 97 123102 (2005)
M. Meneghini et. al. IEEE Electron Device Letters 9 578 (2008)
S. Tomiya et. al. Proceedings of the IEEE 98 1208 (2010)
J. Müller et. al. Appl. Phys. Lett. 95 051104 (2009)
M. Rossetti et. al. Appl. Phys. Lett. 92 151110 (2008)
K. Orita et al., Proc. IRPS 2009
P. Perlin et al, MRS Proceedings 2010
M. Meneghini et. al. Appl. Phys. Lett. 97 263501 (2010)
