10 Jul 2008
Combining a saturable absorber and a gain section in a single semiconductor structure could lead to a new generation of compact ultrafast lasers. Jacqueline Hewett talks to ultrafast pioneer Ursula Keller of ETH Zurich, Switzerland, to find out about her latest idea.
Today's vertical external-cavity surface-emitting lasers (VECSELs) are an elegant ultrafast technology with many compelling advantages. Now, researchers at ETH Zurich in Switzerland have unveiled an enhanced design that they believe holds promise for semiconductor-based high-volume wafer-scale fabrication of compact ultrafast lasers.
"One of my goals is to move compact ultrafast technology into everyday life," Ursula Keller from ETH's Ultrafast-Laserphysics Group told OLE. "We have successfully integrated the pulse-forming saturable absorber and the VECSEL gain into a single semiconductor structure. We call this new class of laser the modelocked integrated external-cavity surface-emitting laser (MIXSEL)."
The interest in VECSELs stems from a list of advantages that includes their power scaling potential and their ability to reach wavelengths that are not easily accessible with established solid-state gain materials.
Ultrafast VECSELs have three main cavity elements: the gain structure, an output coupler and a semiconductor saturable absorber mirror. The essence of Keller's idea is to integrate the gain and the saturable absorber into a single structure. With the only other cavity element being an output coupler, Keller thinks that MIXSELs could become a true wafer-scale technology.
"The cavity is formed by the semiconductor structure and the output coupler," she explained. "You can consider this as a semiconductor wafer and a glass wafer, where you etch the output coupler into the glass. You then glue the two wafers together and then at the end you just dice it up."
The team's MIXSEL structure (figure 1) is based on a gallium-arsenide substrate and contains two highly reflective distributed Bragg reflectors (DBRs), quantum-well gain layers, a saturable absorber based on quantum dots and a final anti-reflection (AR) coating. The quantum wells are 8 nm-thick InGaAs layers, while the saturable absorber is made from self-assembled InAs quantum dots.
"When you passively modelock a semiconductor, the saturable absorber must saturate faster than the gain. These are all high-repetition-rate lasers with minimal self-phase modulation in the cavity, so the saturable absorber needs to do the pulse formation," explained Keller. "This is why we used quantum dots but we also moved to a stronger field enhancement – so the field is a bit stronger in the absorber compared with the gain. You can then optimize the saturation fluence and the modulation depth independently, and this was the key to successful vertical integration."
Keller's team grows the structure using molecular beam epitaxy, but crucially no regrowth or processing is required – it is one single growth that takes several hours. "This is what we would call a 'right-side-up' structure but ultimately it limits the output power," commented Keller. "Doing the growth upside-down, starting with the AR coating, would be easier in the long run. But as this was our first feasibility study we did it right-side-up because of the faster turnaround. We will try upside-down growth in the future."
While Keller admits that the quantum dots were the key to the MIXSEL, she adds that a significant amount of time was spent perfecting all aspects of the growth. For example, ensuring that the gain was high quality and the DBRs were highly reflective was equally important. The bottom (laser) DBR consists of 30 pairs of AlAs/GaAs, while the intermediate (pump) DBR is made up of AlAs and AlGaAs.
"The laser DBR must be highly reflecting, which is why we have a 30 double-pair DBR," said Keller. "You really care about fractions of a percent of loss as these are not acceptable. The pump DBR does not need to have such a high reflectivity and in fact it helps to have this DBR as it minimizes the defects in the gain structure."
With the growth aspect complete, the final MIXSEL uses a laser diode emitting at 808 nm to pump the integrated semiconductor structure, and an output coupler as the second end mirror in the cavity. The MIXSEL emits around 950 nm (depending on the exact composition of the quantum-well gain layers), delivers an average power of 185 mW and picosecond pulses at a repetition rate dictated by the length of the cavity.
"I would like to see this as a laser technology that goes from a repetition rate of a few gigahertz right through to the 100 GHz regime. A repetition rate of 50 GHz requires a 3 mm cavity," said Keller. To put this into perspective, the integrated semiconductor stack is only 8 µm in height and the rest of the cavity is essentially free space. This means that the light's interaction length with the gain is very short compared with a typical edge-emitting laser.
"Edge-emitters always have a long interaction time with the gain material, which in turns leads to strong nonlinearities," said Keller. "Here, it [the interaction length with the gain] is always the same and is independent of the pulse repetition rate for the same peak power. In principle, you do not need to re-engineer the whole modelocking process for the same peak power. It should be truly scalable in terms of the pulse repetition rate."
To date, the Keller group has concentrated on MIXSELs producing picosecond pulses because it is specifically targeting its devices at optical clocking of multicore processors – an application that does not need or want femtosecond pulses.
"If we ever run hundreds of multicore processors and truly use them in parallel computing then they must talk to each other at the clock rate," explained Keller. "Between the cores there is easy access to optics and I think that there is huge potential for MIXSELs in this application. I can visualize a multicore processor clocked by a laser on my desktop."
Another application that could benefit from the MIXSEL technology is supercontinuum generation, although the team would have to push the pulse duration into the femtosecond regime. "The vertical structure could play a role in developing cheaper and more compact supercontinuum sources," said Keller. "The idea here would be to couple the emission from the structure into a fibre, which would go on to create the continuum."
Keller and her colleagues are now pursuing several avenues of research. One of these is designing an electrical pumping architecture that is compatible with passive modelocking, which Keller believes should be achievable, and work is currently ongoing to reach this crucial goal (Applied Physics B 91 257). Another area under exploration is using quantum dots in the gain layers. "We are looking at this for the inhomogenous broadening and potentially shorter pulses that can be achieved," said Keller.
• This article originally appeared in the July/August 2008 issue of Optics & Laser Europe magazine.
View pdf of article