Agilent Claims VCSEL Breakthrough

Agilent Technologies Inc. (NYSE: A) claims to have solved the well-known "mirror problem" that has hampered efforts to commercialize long-wavelength Vertical Cavity Surface Emitting Lasers (VCSELs).

Agilent's breakthrough could lead to VCSELs operating at wavelengths of 1310 and 1550 nm reaching the market in under a year and a half, says Mike Tan, lead researcher on the VCSEL project. Such lasers promise to deliver high performance at lower cost than the standard edge-emitting lasers used for the majority of today's telecom applications (see Laser Blazers).

Designing VCSELs to operate at shorter wavelengths around 850 nm is relatively easy. At 1310 and 1550 nm, however, optimising the optical and electrical characteristics at the same time is tricky, the biggest problem being that materials that give out plenty of light at these wavelengths are not much good for making mirrors.

VCSEL mirrors are made from alternating layers of different materials, creating high reflectivity through the constructive inteference of multiple reflections. The bigger the difference in refractive indices between the two materials, the stronger the reflection. But in indium-phosphide compatible materials, which are used for long-wavelength light emission, only a small refractive index difference can be achieved.

Agilent's solution is to replace one of the materials in the mirror with air. Air has a refractive index of 1.0, while indium phosphide has a refractive index of around 3.5, giving the maximum possible index contrast. In fact, these mirrors only require three mirror pairs, says Tan, compared to 65 or more required for typical designs based on indium phosphide materials. That translates to shorter manufacturing times, lower manufacturing costs, and higher performance.

Agilent researchers make the mirror using techniques similar to those used to manufacture MEMS (microelectromechnical systems), according to Tan. First, they deposit layers of indium phosphide and a sacrificial material. After etching to define the shape of the device, they regrow indium phosphide support pillars. The sacrificial material can then be removed with a selective etch.

So far, Agilent has used this technique to make 1310 and 1550 nm devices, but believes it is likely to prove more useful at 1550 nm, the wavelength of most interest to telecoms companies because it enables DWDM applications. "This approach could easily be applied to the fabrication of lasers at 1550nm wavelengths," says Tan. "I believe that's where this technology will really shine."

Agilent isn't the first vendor to work on long-wavelength VCSELs. Until recently, however, only two others -- Bandwidth9 Inc. and the now defunct Coretek -- had managed to bring indium-phosphide-based VCSELs to market (see Coretek Is Closed). But competition is hotting up: two European startups BeamExpress Inc., and VertiLas GmbH -- have popped up on the scene in recent months (see BeamExpress Tunes VCSEL and VertiLas Secures Funding).

Markus Ortsiefer, founder and CTO of Vertilas, wonders if Agilent has addressed all the aspects of the "mirror problem". "Everyone says the problem is because of the low refractive index difference, but that isn't really the point" he contends. "The real problem is how to manage thermal behaviour."

Ortsiefer notes that indium phosphide-based mirrors have a very low thermal conductivity, with the result that the laser heats up, and more current is needed to get light out. Vertilas has done two things to improve matters. One, it uses a "tunnel junction" to reduce the electrical resistance of the laser. Two, it has very good heat sinking, which means removing the substrate so that the heat sink can be placed close to the active region of the device.

Tan says that Agilent also uses a tunnel junction in its device, but he didn't provide details about heatsinking.

Agilent claims to have demonstrated a 1550 nm VCSEL based on the new mirror technology, and to have started reliability tests.

— Pauline Rigby, Senior Editor, Light Reading
redface 12/4/2012 | 9:16:09 PM
re: Agilent Claims VCSEL Breakthrough Agilent seems to have developed a fixation with nothing but air. After their infamous optical switch based on bubble (air-material interface), here is another one that relies on the air-material interface. Let's hope this one will make it to the marketplace.
bw 12/4/2012 | 9:16:07 PM
re: Agilent Claims VCSEL Breakthrough The other fundamental issue with VCSEL is the small active volume that needed for single mode operation. The active area for VCSEL has to be just a few micron wide to ensure single mode operation, this is 10x less active area compare with the edge emitters. As a result, VCSEL may be "efficient", but can't get enough power out. There are tricks to solve this problem, but definitely not easy...

BobbyMax 12/4/2012 | 9:16:01 PM
re: Agilent Claims VCSEL Breakthrough At one there over 30 vendors who were NCSELS in the high wavelength range. Since the DWDM market is not expanding, I see a very limited market opportunity for the VCSEL vendors. There is not much rationale for Agilent to be in this business.
desikar 12/4/2012 | 9:15:44 PM
re: Agilent Claims VCSEL Breakthrough
At the end of the article, it says "Agilent claims to have demonstrated a 1550 nm VCSEL based on the new mirror technology, and to have started reliability tests".

But a mirror does not a device make in the active components world. So it would be interesting to know if the 1550 nm VCSEL demonstration is really a monolithic device, or if the mirrors are separate from or otherwise hybrid-integrated with the gain region. It appears not to be monolithic, based on the statement by Mike Tan of Agilent, as quoted below:

"This approach could easily be applied to the fabrication of lasers at 1550nm wavelengths," says Tan. "I believe that's where this technology will really shine."

The temperature stability of VCSELs using this process may not be as much of an issue as with other approaches to VCSEL mirrors, as Vertilas claims. These (both the InP layers and the air gap) will have the same thermal expansion coefficient as InP. The mode volume issue (i.e. not much power, as 'bw' pointed out) is still relevant, and has slowed down Bandwidth9 and others. This is likely to limit VCSELs to shorter links at first. Unfortunately, shorter links (< 40 km) are predominantly based on 1310, where low cost edge emitters rule and may not be easily displaced by the time the VCSELs are ready. Which may be why Mike Tan says that 1550 is where these are likely to shine, and we run into the power conundrum again... Also wonder what the speed constraints are with VCSELs based on this design.


Half-Inch Stud 12/4/2012 | 9:15:42 PM
re: Agilent Claims VCSEL Breakthrough Air-mirror shows a gripping level of creativity...

Yet, I ask: What it the VCSEL thermal Resistance with this air-mirror?

Seems to met the mirror technique may show to behave as a thermal oven.

Also, such a wonderful mirror appears as a trap for moisture...Moisture exposure once, and then what does the VCSEL operation do?

Half-Inch Stud
The_Escapist 12/4/2012 | 9:15:36 PM
re: Agilent Claims VCSEL Breakthrough Let's see...

Well, with cleaver packaging you can get around that problem. MEMS people do it all the time.

THE problem for long-wavelength VCSELs is heating. The air-semiconductor mirror is not that new of an idea, though I'm not sure anyone has applied it to a electrically pumped VCSEL (assuming that is what Mike is hyping).

One can mount the devices p-side down to extract heat, so you don't have to extract heat through the extremely high thermal resistance air/InP DBR. I'm assuming they are doing this (unless someone else has an idea of how they could get heat out of the device).

The intrinisic speed will helped by reducing the thermal resistance with p-down mounting. Additionally, the air/InP mirror does have the benefit of creating a low penetration depth mirror, reducing the overall mode volume and increasing the intrinsic speed.

Monolithic operation is quite possible. That doesn't mean they have done it, only that it can be done with a bit of work. And that doesn't mean that the device will have the speed and modulation characteristics (rise & fall time, chirp ect.) that make it useful in a real link.

They likely have a long long road ahead.
Sign In