This article was originally featured in the magazine:
Volume 23 Issue 7

Increasing Market Penetration Of The Quantum Dot


Can quantum dots flourish in telecom lasers, LIDAR and quantum communication? By Dieter Bimberg from Institute For Solid State Physics, Technical University Berlin

The most promising gain material for photonic devices is the quantum dot. Armed with an array of these nanostructures, devices have the potential to combine great temperature stability with high efficiency. This allows quantum dot lasers to target a wide variety of applications, ranging from the automotive sector to access networks.

But where did this revolution start? And where do we stand today? Read on to hear a story of errors and recent breakthroughs.

The advent of semiconductor nanostructure photonics can be traced back to 1976, when researchers Ray Dingle and Chuck Henry, working at Bell Labs, were awarded a US patent entitled Quantum effects in heterostructure lasers. In this patent they argued that threshold current can plummet by an order of magnitude with every ‘reduction of dimensionality’ – that is, with the move from three-dimensional material to double heterostructure lasers; and again with progression to two dimensional materials, as found in quantum well lasers; and eventually to one dimensional materials, which would lie at the heart of quantum wire lasers.

Since this proclamation by the Bells Labs’ duo in the mid 70s, much progress has been made. Today, the production of semiconductor lasers is dominated by heterostructure designs featuring quantum wells. Manufacture is relatively easy, using well-established steps that include the epitaxial growth of a structure that is an extension of the double heterostructure. Three layers are replaced with five or more, allowing the gain layer to have a thickness of the order of nanometres. Separate confinement for the optical modes results from fine tuning the combined thickness of the well and the layers surrounding it.

Note, however, that a move from wells to wires has never taken off. Fabricating the latter is complex, with this class of device proving unreliable and failing to fulfil its promise.

Interestingly, the ultimate level of confinement, provided by zero-dimensional structures that are referred to as dots, did not get a mention in the Bell Labs patent. But it did ten years on, in a seminal paper by Yasuharu Suematsu and co-workers from the Tokyo Institute of Technology. This group evaluated the potential of GaAs and InP lasers with reduced dimensionality, and predicted tremendous improvements in gain and threshold current density with the switch from three-dimensional lasers to ‘quantum box’ lasers, how they called them.

Experimental work from that era also highlighted the promise of lower-dimensional structures. As far back as 1982, Yasuhiko Arakawa and Hiroyuki Sakaki witnessed a lowering of threshold current with increased confinement. They observed this when studying quantum well lasers in increasingly high magnetic fields that reduced the temperature dependence of the threshold current. They correctly attributed this behaviour to increasing carrier localization.

At that time, work on lower dimensional structures was restricted to lattice-matched structures, because this condition was considered imperative for fabricating dislocation-free heterostructures. Due to this limitation, efforts focused on the combination of lithography and epitaxy to form lattice-matched, quantum box lasers, which are now known as quantum dot lasers. Even after seven years of development, the best devices were not that impressive, having to overcome huge threshold current densities to deliver a pulsed mode output when cooled to 77 K.

Self-organised success

A new dawn arrived in the early 1990s, with the re-writing of the rulebook for the production of quantum dot structures. At that time, researchers found that the deposition of a few monolayers of non-lattice matched material led to the formation of similar sized pyramids that were coherent – that is, defect-free (see Figure 1). These pyramids, with dimensions of nanometres, present self-organised quantum dots.

Figure 1. High-resolution, top-view, transmission electron microscopy image of a layer of InAs/GaAs quantum dots, showing four pyramidal dots. The basis of the pyramids are oriented parallel to [100] directions of the semiconductor. Taken from a chapter by D. Bimberg in Optical Fiber Telecommunications ed. by I.P. Kaminov, T. Li and A.E.Willner, Elsevier (2008) and D.Bimberg and U.W. Pohl, Materials Today 14 388 (2011)

This class of dots may be grown by just modifying conventional semiconductor epitaxial processes. With both MBE and MOCVD, engineers have turned to previously undiscovered growth regimes to produce defect-free strained layers. By avoiding electron-beam lithography, production is simpler and cheaper. The dots that result have size- and shape-dependent electronic and optical properties, including their emission wavelength.

With these self-organised dots, efficient recombination is possible at 1.3 μm by inserting InAs quantum dots in GaAs. This combination provides an alternative to the InP material system for the near-infrared. Encouragingly, the material gain for these InAs quantum dots is orders of magnitude higher than that for the bulk form of this material. As the dot layers can easily be stacked on one another many times, modal gain can be comparable, or even exceed, that of stacked quantum wells.

Note that when quantum dots are said to be self-similar, that does not implying that they are identical. Adjacent dots do not have exactly the same number of atoms, or identical interfaces. However, they do have similar shapes, and thus symmetry properties. Entropy accounts for the differences, which are actually a benefit for almost all device applications.

For a single quantum dot, its density of states is described by a delta-function. This is replicated in its emission line shape (see Figure 2), which shows temperature-dependent Lorentzian broadening.

Figure 2. Electroluminescence, delta-function-like spectrum of a single quantum dot embedded in a p-i-n diode (see Figure 6). Emission consists of a superposition of two polarized emission lines a few hundred microelectronvolts apart, which are not resolved in this picture. It is remarkable, that across a range of 400 nm only the emission of this single QD is observed. Taken from W. Unrau and D. Bimberg, Laser and Photonics Review 8 276 (2014)

With a collection of dots, slight differences in the emission wavelength of individual dots impact the overall emission profile. Consider, for example, the active area of a 100 μm-wide, 1 mm long edge-emitting laser. If the density of dots is 1011 per square centimetre, then the active area will have around 108 non-interacting dots, each emitting at a very slightly different wavelength. Emission produced by the laser will be the superposition of that produced by each of the dots, and well described by a Gaussian shaped envelope with a typical a half width of between 30 meV and 100 meV, depending on growth conditions. Thanks to this Gaussian broadening, mode-locked quantum dot lasers can produce femtosecond pulses, and semiconductor optical amplifiers can provide ultra-fast, multiple-wavelength amplification without cross-talk.

Measurements on the first generation of lasers with stacked quantum dot layers revealed device properties that were far superior to those of quantum well lasers. In addition, these results were superior to those expected – a rare occurrence in device development.

A key figure of merit for the semiconductor laser is its threshold current density. This has decreased over many decades. By 2001, quantum dot layers were leading the way, setting a new benchmark of less than 10 A cm-2 /dot layer (see Figure 3).

Figure 3. The threshold current density of diode lasers falls by four orders of magnitude when moving from homojunction to heterojunction, quantum well and finally to quantum dot based structures. Taken from a chapter by D. Bimberg in Optical Fiber Telecommunications ed. by I.P. Kaminov, T. Li and A.E. Willner, Elsevier (2008) and D.Bimberg and U.W. Pohl, Materials Today 14 388 (2011)

Placing dots within quantum wells can yield a particularly strong device performance. When combined with an appropriate scheme for p-type doping, lasers can exhibit a temperature-independence threshold current density at temperatures of up to about 70°C. In such devices, there is complete decoupling between the gain of the quantum dots and the index of refraction of the light guiding layer. That’s possible because the total number of carriers residing in the dots is many orders of magnitude smaller than the number of carriers residing in the carrier reservoir.

One of the strengths of this type of laser is that it is capable of ultra-fast gain recovery following gain depletion, thanks to the very fast down scattering of carriers from the reservoir to dots via an Auger effect. Another merit of the device is that gain and phase modulation can be carried out independently, enabling higher-order modulation schemes. Bit rates of 100 Gbit/s or more are possible.

Additional attributes of the quantum dot laser are that it: increases coupling efficiency into a fibre by suppressing filamentation of the transverse ground mode of the edge-emitting laser; increases radiation hardness; lowers facet overheating, raising the level for catastrophic optical mirror damage; and enhances the stability for external optical feedback by more than 20 dB, leading to the elimination of optical isolators in fibre-based systems.

It may raise an eyebrow to hear that billions of quantum dot photonic devices have already been sold and used. However, these devices are not made with the arsenide or phosphide material systems – they are based on GaN. In GaN-based lasers and LEDs, the InGaN wells are thin and highly strained, giving rise to formation of localisation centres. While these dots have been largely been formed inadvertently, they provide the fundamental advantages associated with quantum dot based photonic devices, enabling the commercial breakthroughs of these device classes.

Aiding the networks

The challenge is to now replicate this success with GaAs- and InP-based nanodevices. Up until very recently, this seemed unlikely, due to a lack of ‘pull’ from a market that had no interest in changing a winning formula: the InP-based quantum well laser. But the situation is now changing, due to demand for sources increasing, and a pressure to cut prices. This scenario is playing out particularly strongly in the market for fibre-to-the-home systems, which are being installed in huge quantities in all industrialized countries, and are the final point in metropolitan and access networks.

Within fibre-optic networks, installation of metropolitan area network traffic hardware is tipped to grow far faster than it is in long-haul networks. This means that there will be a rapid increase in the number, diversity and complexity of photonic devices serving these networks that operate at around 1.3 μm. Controlling the power consumption and cost of the networks will require a sea-change in how the photonic systems are put together. One option for driving success is to turn to InAs/GaAs quantum dot-based photonic modules, because these outperform all the alternatives. They are more energy efficient, more temperature stable, far cheaper, and have merits not found in higher-dimensional devices.

A backbone for these high bit-rate optical communication networks is the mode-locked laser, which can emit electrical and optical pulse trains. This device can be used to make an optical clock, and when paired with a modulator, it can form a transmitter for optical time-division multiplexing systems. Further uses for the mode-locked laser exist in other fields of communications, such as multi-carrier generation and all-optical sampling. Meanwhile, in terahertz and photonics radio-over-fibre systems, mode-locked lasers can simplify source architectures.

The lowest cost mode-locked lasers are passive. Savings result from easily adding a saturable absorber into the cavity. This absorber, like the gain section, is simply direct-current biased. One of the merits of the addition of the absorber is that it ensures a self-start of mode-locking, without the need for an external radio-frequency reference source. When saturable absorbers are paired with 1.3 μm lasers, recovery is typically just 700 fs under large bias. This enables the generation of sub-picosecond pulses with low timing jitter at frequencies beyond 80 GHz. Repetition rates of 160 GHz were presented by multiplexing (see Figure 4).