Quantum-dot lasers on silicon

Integrating quantum-dot lasers on silicon photonic chips promises to create high-speed devices for datacom and other applications.

BY ARTEM PROKOSHIN AND YATING WAN FROM KING ABDULLAH UNIVERSITY OF SCIENCE AND TECHNOLOGY

Since its invention in 1962, the semiconductor laser has played a phenomenal role in changing our world. This source of monochromatic emission is now an essential component in CD and DVD players and recorders, laser printers, barcode readers, and most importantly, fibre-optic communication systems that connect our world through the internet.

Internet traffic continues to climb at an eye-watering pace, with analysts calculating a compound annual growth rate of 25 percent that, in 2022, propelled global traffic to 4.8 zettabytes – that’s 4.8 trillion gigabytes. Given this tremendous rise in internet traffic, which shows no sign of abating, it is more important than ever to trim the power consumption of data transmitters and receivers. Efforts that are underway in that regard are not restricted to long-haul optical communications, and are also considering short-distance interconnects in datacentres, which now account for up to 2 percent of global electricity. The target is to drop below 1 picojoule per bit.

Fulfilling this goal will have far-reaching benefits. As well as helping to curb the carbon footprint in the datacom and telecom sectors, energy savings will help where there is a booming interest in the use of photonic chips in neural networks, known as photonic neural networks. With growing popularity of large language models, the cost of training and operating them is increasing. Photonic neural networks, driven by highly efficient semiconductor lasers, offer a promising solution to meeting this growing energy demand.

Helping to turn such dreams into reality by integrating quantum-dot lasers on silicon photonic chips is our team at the King Abdullah University of Science and Technology. By marrying these two technologies, we are uniting incredibly efficient sources with mature, high-volume semiconductor processing techniques.

Figure 1. The fabrication process for quantum-dot lasers on silicon.

Why silicon?
Silicon, by far the most widely used semiconductor material since the latter half of the last century, is the backbone of the microelectronics industry. When used to produce integrated photonics, silicon provides the most advanced material platform, developed over the last 20 years and relying heavily on CMOS technology. Merits of the silicon photonics platform include: a high refractive index contrast, leading to low-loss; high-confinement waveguides; and efficient grating couplers.

Drawing on silicon’s doping technology, chipmakers can produce high-speed modulators based on p-n junctions, while this material’s compatibility with germanium ensures fast and efficient silicon-germanium photodetectors. What’s more, silicon photonics provides the highest manufacturing volume and the lowest cost, thanks to the opportunity to manufacture silicon chips on 300 mm wafers.

Unfortunately, despite all these strengths, silicon is not the perfect choice. Its biggest disadvantage for photonic applications is its indirect bandgap, preventing it from providing an efficient light source. While several silicon-germanium laser diodes have been demonstrated, they are unsuitable for real-world applications, due to their feeble output power and broad linewidth.

The lack of a silicon-based laser has led to the pursuit of two options for producing PICs with this material system. One involves combing silicon photonic chips, as purely passive devices, with an external light source. In this case, the downsides are high coupling losses, typically exceeding 3 dB, and increased packaging complexity. The second option is to integrate efficient lasers based on III-V semiconductors, such as GaAs or InP, directly onto silicon photonic chips. This approach reduces coupling loss to typically below 0.5 dB and simplifies packaging, but leads to constrained production volumes and increased costs.

Why quantum dots?
Since the introduction of the first semiconductor lasers, there have been a number of improvements to their design, along with the development of various different architectures. The first lasers were homojunction diodes, using the same material for the waveguide core material and its surroundings. To improve optical confinement, a breakthrough that ensured the first continuous-wave operation, engineers introduced a double heterostructure, sandwiching the active region between a pair of cladding layers with a wider bandgap and a lower refractive index. For example, this has been realised by surrounding a GaAs active region with AlGaAs. Another important advance followed, with a move to an active region employing a multiple quantum well structure that improves the efficiency of the radiative recombination process.

The next logical step, pursued by many in recent decades, is a move from a multi-quantum-well active region to one based on quantum dots. With this refinement, carrier confinement switches from one lateral direction to all three dimensions. Quantum dots, also referred to as ‘zero-dimensional’ structures, have dimensions on the order of tens of nanometres. They can be formed with a self-assembly process, involving epitaxial growth of InAs on a lattice-mismatched GaAs substrate.

The father of the quantum dot laser is Yasuhiko Arakawa from the University of Tokyo. In 1982 he proposed this device, claiming it had the potential to provide a lower threshold current and better temperature stability than its quantum-well-based cousins. Thanks to an atom-like structure and a discrete density of states, the quantum-dot laser provides an almost symmetric gain spectrum and a low linewidth enhancement factor, enabling narrow-linewidth and isolator-free operation. Additional opportunities for quantum dots are found in low-dark-current photodetectors and efficient Stark-effect amplitude modulators.

In the context of integration with silicon, the main advantage of quantum dots is a reduced sensitivity to crystal defects. These nanostructures offer improved in-plane carrier confinement, with diffusion lengths on the order of several microns – that compares with tens of microns in quantum wells, and ensures that dots are highly tolerant to dislocation defects.

The great potential of quantum dots is realised in devices. Measurements of quantum-dot lasers demonstrate that they can operate with a long extrapolated-lifetime at high temperatures, thus relaxing the stringent requirements on temperature control of photonic chips. This class of laser also offers a threshold current of less than 1 mA, making it an ideal candidate for addressing the ever-increasing energy consumption in optical communication networks.