Technical Insight
An elemental change to laser design
Today's telecom lasers are plagued with Auger-related losses, which drive down efficiency and make device cooling mandatory. The solution: Switch to an active region with alloys featuring a bismuth content of more than 10 percent, claim members of the European team BIANCHO.
The days when we could only access the internet with a PC are long gone: Most of us now spend more time browsing the web than making a call on our mobile; and if we don't like what's on television, many of us will surf through listings on internet TV.
These activities are placing increasing strain on optical networks and data centres, which have to handle a 60 percent rise in traffic every year. This is good news for telecom component makers, because it creates a market for the sale of lasers operating at higher data rates. But it is not good news for the environment, because it drives up the energy consumption associated with the internet. In Europe, telecommunication and data networks already account for as much as 3 percent of the continent's electricity, and this figure is only going to go up over the next few years.
To try and prevent electrical consumption in these networks from reaching an exorbitant level, our team is pursuing the development of far more efficient telecom components through a four-year research and development initiative called BIANCHO - BIsmide And Nitride Components for High temperature Operation. This effort, which is supported by the European Union Framework 7 programme and backed by over €2 million of funding, involves five leading European research groups from industry and academia. Our groups have complementary expertise in: epitaxy; fabrication; device physics and modelling; characterization of materials and devices; and commercialisation of semiconductor technologies.
The primary goal of our project is very simple "“ to provide a substantial improvement in the efficiency of semiconductor lasers and semiconductor optical amplifiers that are deployed in telecommunication and data networks. To achieve this, we are developing a new class of bismuth alloys for use in optoelectronic device. This novel semiconductor suppresses non-radiative Auger recombination, which dominates conventional telecom lasers and SOAs, wasting about 80 percent of the input electrical power.� Thanks to this energy saving, our devices will generate far less heat, and will not require power-hungry thermoelectric coolers for temperature control.
Kerstin Volz (left) and Peter Ludewig (right) standing next to the MOCVD tool at Philipps University Marburg, Germany.
Addressing Auger
The energy sapping, Auger loss mechanism that severely degrades the efficiency in today's InP-based devices stems from the recombination of an electron in the conduction band and a heavy-hole in the valence band (see Figure 1). Instead of interacting to emit a photon, this pair of oppositely charged carriers excites a hole from near the valence band maximum into the spin-split-off band. The hole then relaxes, releasing energy in the form of heat.
Figure 1: (Left) Auger recombination is the dominant loss mechanism in conventional InP-based lasers. Here a conduction electron (1) recombines with a valence hole (2), with the released energy exciting a valence hole (3) to the spin-split-off band (4) instead of creating a photon.� (Right) This Auger loss mechanism can be suppressed when the spin-orbit-splitting energy exceeds the band gap. When this occurs, the Auger recombination process is forbidden, due to conservation of energy, because the energy required to excite a hole to the spin-split-off band exceeds the energy produced by electron-hole recombination
Over the years, incremental approaches have been pursued to reduce these Auger-related inefficiencies, but they have failed to address its fundamental cause: It originates from the electronic band structure associated with the constituent materials in the device's active region. Manipulating the band structure is the only way to tackle this issue head-on "“ and that is what we are doing by turning to the unique properties of bismuth-containing alloys, which enable the design of Auger-free lasers.
One attractive attribute of bismide alloys is the behaviour of their energy gap: It decreases very rapidly with bismuth composition, allowing growth of telecom lasers on a GaAs substrate. But even more important than that "“ and the key idea behind the suppression of the dominant Auger loss process "“ is that GaBi, in contrast to conventional III-V materials emitting in the near-infrared, is predicted to have a very large spin-orbit-splitting energy (see Figures 2 and 3). Its value, which is the difference between the valence band maximum and the lower lying spin-split-off valence band, is of the order of 2.2 eV� "“ large and controllable spin-orbit-splitting energies are also possible with bismide alloys, such as GaBiAs, InGaBiAs and GaBiNAs. If this spin-orbit-splitting energy exceeds the bandgap energy of the telecom laser, which is typically below 1 eV, the law of conservation of energy dictates that there will be a significant reduction in Auger recombination in this device.
Figure 2: Spin-orbit splitting energy for various III-V materials. The very large value for the spin-orbit-splitting energy ï„SOfor GaBi holds the key to suppressing Auger-related losses in GaBiAs/GaAs-based lasers
Figure 3: Comparison of the experimental and theoretical values of the energy gap (Eg) and spin-orbit-splitting energy (ï„SO) for epitaxially grown GaBixAs1-x samples on GaAs substrates. Auger losses are suppressed when the bismuth composition is greater than about 10 percent, because at concentrations at this level and higher, ï„SO�exceeds Eg.
We are aiming to design and fabricate a device that behaves just like this. Our efforts kick-started with a study of the bandstructure of epitaxially grown GaBixAs1-x samples on a GaAs substrate. Photo-modulated reflectance spectroscopy and atomistic theoretical calculations undertaken by us have revealed that the introduction of bismuth into GaAs has the desired effect on the band structure.
This combined theoretical and experimental effort by our team has garnered three key insights: the band gap energy of the alloy GaBixAs1-x decreases dramatically with bismuth composition, thereby offering a possibility to achieve 1550 nm emission on a GaAs substrate; spin-orbit-splitting energy increases rapidly with bismuth richness and exceeds the bandgap energy at a content of 9-10 percent; and GaBixAs1-x has a type-I band offset relative to the GaAs substrate, a condition favourable for realising large optical gain and ultimately an efficient laser.
Building bismide lasers
Our next step has been to form bismide quantum wells with high optical quality on a GaAs substrate. We have adopted a two-pronged approach, using both MOCVD and MBE to try and obtain high-quality heterostructures. Producing these structures is challenging because the epitaxial growth of metastable GaBiAs requires very low growth temperatures compared with conventional III-Vs. What's more, the photoluminescence signal is always very sensitive to the bismuth content and the growth conditions.
Both growth technologies have produced some noteworthy success. MBE is, to date, capable of alloys with higher bismuth content, while MOCVD has produced epiwafers that have been processed to yield the world's first electrically pumped, dilute-bismide laser.
Using MBE, we have grown ternaries with a bismuth content greater than 10 percent. Higher values are possible "“ Tom Tiedje's group from University of Victoria, Canada, have recently reported values in excess of 20 percent.
Meanwhile, with MOCVD, we have grown a laser structure in a commercially available AIX 200-GFR reactor system, using palladium-purified hydrogen as the carrier gas at a reduced reactor pressure of 50 mbar. For the quantum well growth, triethyl gallium is used as a group III precursor, while tertiarybutyl arsine and trimethyl bismuth are used as the group V precursors, since low growth temperatures (around 400 �C) are required.� By carefully designing these growth conditions, we have been able to form high-quality GaBiAs single-quantum-well lasers with a bismuth incorporation of 2.2 percent. These devices, measured in "˜as-cleaved' form under pulsed operation to minimize heating effects, produce 950 nm emission at room temperature and have a threshold current density of 1560 A cm-2 (see Figure 4).Our next goal is to systematically increase the bismuth composition in the active region of these lasers, with the aim of achieving 1550 nm lasing while the spin-orbit splitting energy exceeds the bandgap.
Commercialisation of our technology is being driven by CIP Technologies of Martlesham Heath, UK. This company, which is now a part of Huawei, has been developing designs and fabrication processes for making lasers and modulators with these new materials.
Figure 4: (Left) The BIANCHO team have fabricated the world's first bismuth-based laser "“ an electrically pumped GaBi0.022As/AlGaAs single quantum well laser diode. (Right) Light-current relationship for a 50 �m� x 1000 �m GaBiAs/AlGaAs single quantum well laser at room temperature. The lasing spectrum is shown in the inset.�
Efficient terahertz generation using dilute bismide alloys
Fibre-coupled photoconductive terahertz detector featuring a GaBiAs
Dilute bismide layers are not just promising for the fabrication of telecom lasers and modulators "“ they also offer significant opportunities for the development of low-cost, efficient terahertz technologies.� At the Center for Physical Sciences and Technology (FTMC) in Vilnius, Lithuania, Arunas Krotkus' group are exploiting very short photoexcited electron trapping times, which were typical in the first GaBiAs epitaxial layers grown with large bismuth content.
The sub-picosecond trapping times in GaBiAs, together with the narrow bandgap and relatively high electron mobility, make this material very attractive for manufacturing photoconductive antennas for terahertz emission and detection. Such devices are in demand for spectroscopic, imaging, and security applications. Until recently, terahertz emitters and detectors were mainly manufactured from epitaxial layers of GaAs grown by MBE at low substrate temperatures.
This approach is not ideal, because GaAs is transparent to wavelengths beyond 850 nm, so bulky and expensive femtosecond Ti:sapphire lasers are required for carrier excitation in this class of optoelectronic terahertz radiation system. In contrast, compact fibre or diode lasers emitting in the 1.0-1.5 �m range can activate optoelectronic terahertz emitters and detectors developed at Vilnius. Bismide-based systems have already been commercialized. They are available from TERAVIL,�a spin-off company of the FTMC.
Future targets
Funding for our project continues to July 2014, and over the coming months we will spend our time focusing on increasing the bismuth content in single- and multi-quantum-well GaBixAs1-x lasers, until the spin-orbit splitting energy in these devices exceeds the bandgap energy. This should unlock the door to an Auger-free 1550 nm laser. This is our highest priority, because it will enable devices with greatly reduced cooling (and hence energy) requirements and consequently a simpler circuit design. Our epitaxy experts are trying to grow heterostructures by MOCVD and MBE that feature bismide-based wells with a bismuth content in excess of 10 percent, and our team is also pursuing the practical realization of temperature-insensitive, electro-absorption modulators based on the quaternary alloy GaBixNyAs1-x-y and grown on a GaAs substrate.
In addition, we are investigating the growth and characteristics of other bismides. This includes the alloy GaBiNAs, which offers significant scope for tailoring the optical and electronic properties of new devices. Nitrogen and bismuth have opposite effects on material strain, making it possible to grow lattice-matched GaBiNAs layers that combine narrow band gaps with almost independently controllable band offsets and an enhanced spin-orbit splitting. In other words, GaBiNAs has the potential to produce high-efficiency, mid-infrared emitters on GaAs substrates.
One class of device that falls into this category is the mid-infrared VCSEL. This could combine reflective mirrors built from GaAs and AlGaAs, a pair of materials with a significant difference in refractive indices, with an efficient mid-infrared active region. Such a device could be used for low-cost sensors for environmental monitoring and portable medical diagnostic equipment. Other types of device could also benefit from the introduction of alloys containing nitrogen and bismuth. Lattice-matched GaBiNAs layers have the potential to form the 1 eV (and other) junctions for multi-junction solar cells, while lattice-matched GaBiAs/GaNAs superlattices could find application in mid-infrared detectors.
There is also the opportunity for InP substrates to provide the foundation for bismuth-based materials, such as GaBiNAs. With this particular quaternary, a bismuth content of just over�4 percent is required to realise an optimised band structure and allow the construction of high-efficiency lasers beyond 2 �m. Using this material system provides an opportunity for InP foundries to diversify into mid-infrared devices with little or no change in infrastructure, whilst opening up new markets in sensing and defence. The potential for bismuth-containing alloys is enormous, and today we are just skimming the surface.
�The BIANCHO team
The BIANCHO team, led by Tyndall, is pursuing the suppression of Auger recombination via the incorporation of bismuth into GaAs to form the alloy GaBiAs. This approach was originally presented and patented in 2010 by academic Stephen Sweeney from the University of Surrey.
Five institutions are involved in BIANCHO, with each performing different tasks:
