Superluminescent Diodes- A Novel Approach For Droop-free Solid-state Lighting?
Superluminescent diodes marry the virtues of LEDs and laser diodes by offering droop-free emission from a high-quality beam that can deliver speckle-free projection
BY ANNA KAFAR FROM UNIPRESS
One tremendous talking point of recent times has been the cause of droop, the decline in the efficiency of an LED as the current passing through it is cranked up. Getting to the bottom of this has been a goal for many researchers in industry and academia, who have hoped that by uncovering the cause of this mysterious malady, they will unlock the door to a generation of droop-free LEDs that will accelerate the adoption of solid-state lighting.
It did not take long for these investigators to realise that droop depends on the cube of carrier density. But it has taken far longer for them to come up with convincing evidence for its cause, which is non-radiative Auger recombination. Direct evidence only arrived in 2013, through a partnership between scientists at the University of California, Santa Barbara, and à‰cole Polytechnique, France (see Compound Semiconductor p.46 8 (2013)).
Auger recombination is a material property, so supressing droop is very difficult. However, it is possible to sidestep it, rather than address it head on, by turning to a droop-free light source: a laser diode. When this devices is driven above its lasing threshold, its carrier density is constant, clamping droop-related recombination. Thanks to this, it can then operate with a high differential efficiency in a fashion described as droop free. For this reason, it has been recently proposed by scientists at Sandia National Laboratory that laser diodes, in partnership with phosphors, can be efficient sources of white light (the possibility of using such devices has been broadly discussed in the article "Solid-state lighting: Are laser diodes the logical successors to LEDs?" in Compound Semiconductor p.40 8 (2013)).
The laser, however, does not have a monopoly on droop-free light emission "“ this accolade is shared by the superluminescent diode (SLD), a device that we have been developing and investigating at Unipress in Warsaw, Poland. This lesser-known light emitter combines many characteristics of LEDs and laser diodes, and can be viewed as an edge-emitter that is fabricated in a way very similar to that of the laser diode.
While it may look like a laser, the SLD is distinct, because it does not lase, due to far higher losses at the mirrors or at the end of the waveguide. This means that light is generated in a process of spontaneous emission, just like it is in an LED; although it is also amplified by stimulated emission in the waveguide, which is what happens in a laser.
The interplay of spontaneous and stimulated emission gives rise to a light-versus-current curve with an exponential region (see Figure 1 (a)). High mirror losses that are engineered by a specific device design prevent light from reflecting at both or one of the facets. This leads to a shape of emission spectrum resembling that of a laser diode below threshold (see Figure 1(b)). But the device can reach optical power comparable to laser diodes above threshold and, like a laser, it emits a high quality light beam.
Figure 1: The output power profile of the superluminescent diode is similar to that of a laser, and it has a spectral profile resembling that of an LED.
Thanks to these attributes, arsenide and phosphide SLDs are already common light sources in applications such as optical coherence tomography and fibre-optic gyroscopes. And we believe that nitride SLDs, which are now less mature than their phosphide and arsenide cousins, can be used in the future in the same fields, as well as alternatives to laser diodes in droop-free light generation.
Turning to SLDs for lighting might raise a few eyebrows, because as these devices do not reach lasing threshold, one might expect that they are prone to droop, just like an LED. But our experience suggests otherwise. Our studies show that SLDs can easily emit high optical power at current densities three orders of magnitude beyond those associated with the peak quantum efficiency for an LED, which often occurs around
10 A cm-2. At current densities of 1000 A cm-2 or more, droop would be expected to govern the light-current characteristics, but measurements show that's not the case, with the output power steadily increasing with current. It is possible, therefore, that SLDs are even immune to droop.
The supreluminescent diode is not overwhelmed by droop, thanks to stimulated emission. Credit: S. StaÅ„czyk, Unipress.
Our development of SLDs has drawn on the efforts of others, including the pioneers of this device, who built the first emitters in the early 1970s using the arsenide material system. Nearly 40 years later, the SLD family expanded, with the first nitride SLDs emitting at around 420 nm emanating from the à‰cole Polytechnique Fédérale de Lausanne. Five years of intense study of SLDs followed, with many groups focusing their research on approaches for cavity suppression that could unlock the door to high-power devices spanning a wide range of emission wavelengths.
We started to develop SLDs in 2010, and the output of our latest violet devices is state-of-the-art at more than 200 mW. In comparison, the world record for blue-emitting variants, held by Osram, is above 100 mW. A reduction in output power at longer wavelengths is to be expected, and mirrors the behaviour of laser diodes. For both classes of device, when indium is added to push emission from the violet to the green, the strength of the built-in electric field increases, leading to reduced electron-hole overlap. A fall in gain results, lowering output power. But it is not all bad news for these SLDs that are following in the footsteps of the laser diodes, as many improvements to the latter can be applied to the former. Thanks to this, we believe that it should be possible to realise really fast developments in SLD performance.
Mimicking laser manufacture
The fabrication of a nitride SLD is similar to the making of a laser, and begins by growing a separate confinement heterostructure on a GaN substrate. This is followed by the addition of waveguides and contacts. Performance of the device is governed by the cavity suppression mechanism, which can be realised by many means, including: the insertion of a waveguide that is titled with respect to the facet, the addition of a high-quality anti-reflection coating, and the deployment of an absorbing region.
Most of today's SLDs employ a tilted waveguide, which has an axis inclined to the cleavage planes (the future chip facets). With this design, light generated within the active area is guided towards the window while undergoing amplification, just as it would in a laser diode. However, light does not impinge on the facet perpendicularly. This means that a proportion of the light reflected back into the chip is directed outside the waveguide (see Figure 2(a)). Consequently, light does not oscillate between the facets, and no lasing takes place. Note that the light that is emitted out of the chip exits at an angle determined by the refractive index of the material in the laser, and the relative angles of the facet and waveguide.
Figure 2: Superluminescent diodes feature a bend waveguide geometry. At the front facet a) light is reflected outside of the waveguide to prevent the device from lasing, whereas at the rear facet b) light is coupled back to the waveguide to lengthen the amplification path. The team at Unipress employ a "j"-shape design c), where half of the waveguide is straight, like it is in laser diodes.
One great strength of this design is that without substantial changes in standard laser diode fabrication, it is possible to reach an extremely low value of front facet reflectivity (below 10-4) "“ this is highly desirable, as it quashes light oscillations. This low-reflectivity front facet can be combined with a rear facet with high reflectivity to form an SLD with double-pass amplification. When the SLD operates in this manner, the light initially travelling towards the rear facet is coupled back into the waveguide, creating an amplification path that is twice that of the length of the chip (see Figure 2 (b)).
Doubling the amplification path has a tremendous impact on output power, because intensity rises exponentially with path length. Creating a bend is the key to forming such devices, and in our case we use a "˜j-shaped' waveguide. Further improvements in facet properties are possible with additional antireflection and high-reflection coatings, but care is required in order to preserve proper cavity suppression, and any increase in rear-facet reflectivity must go hand-in-hand with a trimming of front-facet reflectivity.
Although a front-facet reflectivity below 10-4 can reduce light oscillations substantially, it cannot eliminate them. Any measurements on superluminescent diodes operating at a reasonable current will uncover modulations or ripples in the spectral output. These arise from the facets reflecting some light into the waveguide and are a sign of a high-loss Fabry-Pérot resonator. The depth of modulation increases with current until it reaches the full height of the spectrum, and then the device starts to lase (this is the limit of superluminescence).
During device optimisation, one of the biggest goals is to reduce the ripple depth as much as possible. In our case, because we employ a tilted waveguide design, this occurs through optimising the bend angle. Varying this angle produces a series of deep minima for front-facet reflectivity, and an optimised device geometry is realised from a minima associated with a very low reflectivity of the front facet.
As a function of current, the efficiency profile of the SLD is vastly different from that of an LED. Instead of peaking at around 10 A cm-2, it hits its peak performance at far higher current densities, where, surprisingly, droop does not appear to play a role. For a typical device, light output first increases exponentially with current, in agreement with theory, before entering a linear regime that is free from lasing (this is evident in measurements of emission spectra). What is particularly interesting is that a laser diode and SLD fabricated side-by-side produce the same slope efficiencies (see Figure 4), so it should be possible for an SLD to deliver the same output power as the laser, if it operates at a higher current.
Figure 3: Emission spectra of a superluminescent diode measured close to the onset of amplified spontaneous emission (200 mA) and at high current (500 mA).
Figure 4: The waveguide geometry impacts the behaviour of the superluminescent diode, which has a spectra shown in Figure 3.
The linear regime of the SLD is associated with the saturation of optical gain at high currents. To measure this gain, we have developed an approach that can compare amplified and spontaneous emission, and thus explicitly uncover the presence of gain saturation. This saturation is not that surprising, because our SLD operates at extremely high carrier densities. Unlike a laser, there is no clamping of the carrier density at threshold, and instead it climbs throughout the entire operating range. We have constructed a simple model for our SLD, drawing on the well-known ABC model used to explain LED behaviour. We determined values for A, B and C by fitting the spontaneous emission of the SLD, and the results of this suggest that for current densities in the linear regime, the gain value is so high that we approach its material limits (defined by Fermi-Dirac statistics), which we observe as saturation. On top of this, a high photon density can cause carriers to be "˜burned out', if they are recombining quicker than the rate of new carrier supply. This situation also promotes gain saturation.
Our explanation thus far does not offer a clear answer to this question: Why can superluminescent diodes be immune to droop? To gain an insight into why this is the case, we calculated how Shockley-Read-Hall recombination, bimolecular recombination and Auger recombination depend on current density, using an ABC model.
This study shows that Auger recombination overshadows other mechanisms at rather low current densities, such as 100 A cm-2. However, the key point is that there is another mechanism that we have ignored in our calculations "“ stimulated emission. This is a very fast process that is influenced by the photon density, and at high current densities this consumes carriers at a far higher rate than the Auger process. Consequently, droop is quashed at high current densities.
Figure 5: Comparing the rates of different recombination mechanisms using the ABC model and a fit of measured data suggests that Auger recombination strongly dominates in the operating regime of superluminescent diodes. However, this model ignores stimulated emission, which occurs at a rate that is far faster than Auger, and accounts for the majority of carriers.
Virtues of the SLD over the LED are not restricted to immunity to droop, and include lower fabrication costs. This stems from the far smaller chip size, which is essential for cost-competitive nitride devices that are grown on native substrates. In addition, there is no need to devote time and effort to improving the extraction efficiency of the SLD (like in all edge emitters), while the high beam quality enables efficient fibre coupling and focusing.
What's more, the SLDs have advantages over laser diodes for white-light generation. A lack of time coherence makes this source safer for general applications, and it should be possible to employ higher light intensity before damaging the phosphor.
The superluminescent diode is much smaller than the LED, so material costs for this potential lighting source are lower. Credit: S. StaÅ„czyk, Unipress.
Note, also, that GaN SLDs are a perfect light source for pico-projectors. This form of display requires a high-quality light beam, and when a laser is employed, the image is compromised by speckle, an interference effect related to the high degree of time coherence of the source. SLDs have low time-coherence, and red, green and blue forms of this device promise to combine to form an excellent projector.
This is just one of many opportunities for the SLD. With a slope efficiency that matches that of the laser diode, it promises to increase in power, and serve multiple markets.
The work was supported by Wroclaw Research Centre EIT+ within the project "The Application of Nanotechnology in Advanced Materials" "“ NanoMat (POIG.01.01.02-02-002/08) co-financed by the European Regional Development Fund (operational Programme Innovative Economy, 1.1.2).