Uncertain future looms for dilute-nitride lasers

These dilute-nitride lasers can either take the form of edge-emitters, or exploit AlGaAs distributed Bragg reflector technology as VCSELs. This class of emitter also has an intrinsic advantage over InP-based lasers - a higher conduction band offset that is claimed to improve high-temperature performance. Despite these attractive attributes and a significant research effort from academia and industry, dilute nitrides have failed to fulfill their potential, and the material system is still regarded as a difficult one to grow. Infineon leads the way One maker of VCSELs containing dilute-nitride active layers is Germany-based Infineon Technologies. The company produces 1310 nm VCSELs at its Munich facility on 4 inch GaAs substrates using MBE growth.

Infineon, considered by many to be the leading exponent of dilute-nitride laser technology, held on to this product range following the sale of its optical-transceiver business to US company Finisar earlier this year (Compound Semiconductor March p6). However, the firm is looking for a buyer for its fiber-optic-components business as it does not view this as part of its core activities. Infineon plans to end its involvement in fiber optics by the end of the year.

Ayad Abul-Ella, general manager and vice-president of Infineon s remaining fiber-optics business unit, says that the company started developing dilute-nitride lasers in the mid 1990s. "In 2001, when the bubble burst, cost reduction became a dominant driver in the industry. We saw our 1310 nm VCSEL technology as one enabler for cost reduction, and invested heavily from 2001 onwards," said Abul-Ella. This year the firm completed lifetime testing and chip qualification, and produced engineering and qualification samples, but it has not moved into volume production.

Abul-Ella says that Infineon s 1310 nm dilute-nitride VCSELs are cheaper to manufacture than equivalent emitters containing a distributed feedback (DFB) structure. He thinks that the technology could also replace Fabry-Pérot lasers for longer-distance transmission, but cost savings would be modest.

Growing dilute-nitride VCSELs involves a series of trades-offs between achieving a wavelength close to 1310 nm, obtaining a suitable bandwidth, producing sufficient output power, and fabricating a design that can couple sufficient light into an optical fiber. Abul-Ella says that the greatest of these challenges is getting the right wavelength while coupling sufficient light into the fiber. Many researchers fall short of this target.

Infineon s dilute nitrides have to operate at temperatures of up to 85 °C and comply with internationally recognized standards for optical carrier systems operating at data rates such as 155 and 622 Mbit/s, and 2.5 Gbit/s. Abul-Ella explains that the firm achieves the best all-round performance for its devices by targeting an emission wavelength for its lasers of 1300 nm, which falls within the specifications for the 1310 nm standards. Infineon has qualified its process for lasers operating at data rates of up to 2.5 Gbit/s, and has also produced prototypes that can operate at 10 Gbit/s, although these do not yet conform to the necessary specifications. Abul-Ella says that the company has encouraging data on its 2.5 Gbit/s laser chip, including 5000 hours of laser qualification. The same performance for less Abul-Ella believes that the first application genuinely demanding dilute-nitride VCSEL technology could be 4 Gbit/s FibreChannel. He claims that this standard can be used in applications requiring transmission through up to 10 km of fiber, which rules out the use of Fabry-Pérot lasers. "The technology generally used is DFB lasers from the likes of Sumitomo," said Abul-Ella. He claims that Infineon s VCSELs are one-third to one-half the cost of the 4 Gbit/s DFB lasers that sell for at least $20-$30, while meeting the same specifications.

Abul-Ella explains that efforts must now be directed at chip packaging, which has a big effect on the coupling efficiency of the light into the fiber. However, since Infineon is exiting the fiber-optics business, the company has ceased to invest in the 1310 nm VCSELs that it has spent 10 years developing.

One drawback of the technology is that the yields associated with dilute-nitride device production are not particularly high. However, this has never been a major focus for Infineon, because high-volume production is yet to happen. Abul-Ella does point out, though, that yields must remain reasonable (typically more than 30%) to ensure a stable process, and if volumes increase then yields will improve. Taiwanese competition Another company that has been developing 1.3 μm dilute-nitride lasers is Taiwan-based Chungwa Telecom. At the Indium Phosphide and Related Materials conference in Glasgow, UK (see box), Nien-Tze Yeh, a researcher at Chungwa s advanced technical research laboratory, outlined Chungwa s development of lasers emitting at 1.26 μm. He says that when additional nitrogen was added to the quantum well to push emission to 1.3 μm, the crystal quality of the InGaAsN active layer deteriorated and no laser emission was seen. Yeh believes that laser emission could be possible, however, by inserting either GaAsN or InGaAs layers into the structure.

According to Yeh, several difficulties exist for dilute-nitride growth: nitrogen-incorporation efficiency is low, crystal quality deteriorates with nitrogen incorporation, and it is challenging to grow good-quality InGaAsN layers on top of the aluminum-containing layers that are usually used for cladding regions.

As a result, Chungwa is to stop its dilute-nitride laser program and focus on InP-based 1.3 μm emitters. The company is now developing 1.3 μm, 10 Gbit/s DFB lasers, which it intends to release by the end of this year.

10 Gbit/s modulation from edge-emitting dilute-nitride lasers operating at 1.3 μm has been demonstrated by researchers at France s Centre National de la Recherche Scientifique, working in conjunction with the III-V lab owned jointly by Alcatel and Thales. The team used MBE growth to produce structures with three Ga0.63In0.37N0.01As0.99 quantum wells, 7 nm thick, surrounded by 100 nm-thick GaAs barriers. The cladding regions contained n- and p-doped 1.5 μm-thick Al0.8Ga0.2As, as well as thin layers next to the active region that graded the aluminum content to zero.

The processed lasers contained a 2 μm-wide ridge, typically emitted at 1335 nm, and showed a threshold current density of 1.1 kA cm-2 when operated at 20 °C in pulsed mode. One of these MBE-grown lasers, which emitted at 1346 nm was able to produce 2.5 Gbit/s transmission rates along 2 km of fiber at up to 85 °C, and data transfer reached 10 Gbit/s at 25 °C. The researchers have also produced single-quantum well devices with 1.5 μm-thick InGaP cladding regions by MOCVD that deliver 2.5 Gbit/s transmission rates at up to 85 °C.

The French team s results, along with those of Infineon, show that dilute nitrides can compete with their InP rivals. However, with Infineon no longer owning a transceiver unit, and Chungwa ditching the technology altogether, the near-term future of dilute-nitride lasers looks decidedly unpromising. Further reading Yeh et al. IPRM Con. Proc. 2005.
Dagens et al. 2005 IEEE Photon. Technol. Lett. 17 p971.