Nanowires for UVC emission

AlN nanowires with short-period superlattices based on AlN and GaN are trailblazing a path to a new class of UVC-emitting devices.

BY RÉMY VERMEERSCH AND CORENTIN GUÉRIN, BOTH AFFILATED TO CNRS-GRENOBLE AND CEA-GRENOBLE; GWÉNOLÉ JACOPIN AND JULIEN PERNOT FROM CNRS-GRENOBLE; AND BRUNO DAUDIN FROM CEA-GRENOBLE

IT IS WELL KNOWN that exposure to mercury is not good for our health. It can cause a number of symptoms, including tremors, insomnia, memory loss, neuromuscular effects, headaches and cognitive and motor dysfunction. Due to this, the current implementation of the Minamata convention, signed in 2017 under the auspices of the United Nations, is progressively banning the use of mercury-containing devices for environmental and sanitary reasons.

The phasing out of mercury has major implications for UV lighting. For emission in this spectral range, mercury lamps are the incumbent source, widely used for disinfection. As they will have to be replaced in the short term, the research and development of new UV-emitting devices is on the rise. Hopefully this will address the increasing need for versatile, efficient sources in the UVC range that spans 200 nm to 280 nm.

Optical sources that emit in this spectral domain can serve a variety of applications. That includes disinfection, as UVC light damages DNA (or RNA) chains, thereby preventing the reproduction of viruses and bacteria. However, while highly beneficial for disinfection, the sensitivity of DNA to UVC sources within this wavelength range is also problematic – it may severely hurt human skin, have long-term effects and cause melanoma. These downsides complicate the practical use of 265 nm UVC sources for the growing market of domestic disinfection applications, such as putting LEDs in a refrigerator, using them for tap water sanitisation and providing air purification.

Fortunately, humanity can enjoy the benefits of UVC sources while avoiding their pitfalls. What’s needed is to decrease the wavelength within the UVC to the 220-230 nm range. At such short wavelengths the penetration range in skin is limited to stratum corneum, preventing DNA damage to retina as well as deeper epidermis and dermis layers.

Despite the detrimental effect of UV light with wavelengths above 230 nm to the integrity of dermis, most current research is directed at developing sources emitting in the 260-280 nm range to target disinfection applications.

Incidentally, sources emitting at longer wavelengths than this also have health benefits. Controlled exposure to emission in the UVA (315-400 nm) and UVB (280-315 nm) has been shown to cure various diseases, including psoriasis, lupus vulgaris and rickets, and may even have mood-enhancing effects through stimulated vitamin D production.

Figure 1. Emission in the UVC (200-280nm) exhibits bactericidal properties, with maximum DNA damage observed at 265 nm. At 220 nm, the reduced depth penetration of UV light prevents damage to eye cornea and skin, opening the path to surface sterilisation of objects. Other medical applications, such as psoriasis treatment, are documented around 312 nm. The stimulated production of vitamin D by exposure to UVA is recognised to have beneficial effects to cure a variety of diseases, including rickets and mood troubles.

Nitrides: Pros and cons
Depicted in Figure 1, III-nitride semiconductors – namely AlN, GaN, InN and their alloys – are well-suited to the realisation of heterostructures spanning the entire UV range. In particular, the bandgaps of AlGaN ternary alloys are capable of fully covering the UVC and UVB bands. This capability has ultimately stimulated wide interest in the scientific community.

Unfortunately, the practical realisation of UV LEDs is far from trivial: III-nitride semiconductors suffer from a significant lattice mismatch with the most common substrates used to grow them, namely sapphire, silicon and SiC. Due to this mismatch, strain accumulation increases with layer thickness until relaxation occurs, which tends to result in the formation of a high density of misfit dislocations.

The presence of these dislocations within the epilayers and also of growth-related threading dislocations impairs efficient radiative recombination.

Figure 2. Following the growth of n-doped GaN nanowires (NWs) spontaneously nucleated on silicon (111), an n-doped AlN NW section is deposited, followed by the growth of an active region consisting of either thin GaN quantum wells separated by AlN barriers or gallium-rich nanoclusters in an undoped AlN section. The top part consists of a p-doped AlN NW section covered by a 10 nm thick p⁺⁺ GaN section. This structure is processed using standard lithography techniques and a semi-transparent electrically conducting layer is deposited on top.

Another challenge with light emitters based on III-nitrides is the realisation of sufficient doping. Electrical doping of a deep-UV LED is not easy, as for n-type and p-type doping, using silicon and magnesium, respectively, dopant ionisation energy increases with the AlN molar fraction in AlGaN. That means that doping is increasingly challenging at shorter wavelengths. Due to this, current injection is a key issue for UV devices.

As well as issues related to strain and carrier creation and transport, light extraction is far harder in UV devices than their visible cousins. Current-injection layers absorb UV light, and there is a large optical index for these nitrides with respect to air. Due to this, specific packaging strategies are required to prevent light from being trapped within the chip.

One consequence of all these difficulties is that the external quantum efficiency of today’s UV LEDs – that is, the ratio of extracted UV photons to injected electron pairs – is very low. This efficiency is just a few percent at 265 nm, and even lower at 230 nm.

Figure 3. Deposition of monolayer (ML)-thick GaN quantum wells (QWs) is easily controlled in MBE. For increasing deposition of GaN, a series of peaks corresponding to QW thickness of one to four MLs is observed at wavelength emission values of 239 nm, 267 nm, 289 nm and 304 nm, respectively. The reduced cathodoluminescence (CL) intensity quenching between 5K and 300K suggests a reduced influence of thermally activated non-radiative defects. The high-resolution, transmission electron microscopy image of one-ML-thick GaN QWs illustrates the interface abruptness and the absence of interdiffusion.

The nanowire revolution
A promising pathway to addressing all performance limitations is the nanowire, a monocrystalline semiconductor structure with at least one nanometric dimension. Nanowires are typically one or more microns in length, and 50 nm to 200 nm in the lateral dimensions. Often exhibiting facets corresponding to low-energy crystallographic planes, these miniature marvels can be grown by various techniques, either using or not using a metallic catalyst (for the former, metal droplets seeded on the substrate act as nanowire nucleation centres).

For catalyst-free formation, specific to III-nitride materials, nanowires may be grown by MBE on virtually any kind of substrate – that includes sapphire, graphene, oxides and metals. Unlike the classical vapour-liquid-solid growth mode, which is a thermodynamic process that is triggered by the precipitation of atomic constituents at the interface between a substrate and a gold catalyst droplet featuring dissolved elementary species, catalyst-free growth of nanowires is governed by kinetics. What this means is that growth is influenced by diffusion of adsorbed atomic species (adatoms) on both the substrate and the nanowire surface.

With catalyst-free growth by MBE, following the spontaneous nucleation process, elongation proceeds through upward diffusion of metallic species, which impinge on the side walls until they reach the top surface. However, depending on the detailed growth parameters and their influence on metal adatom diffusion, it is possible that the nucleation probability on the side walls is not strictly zero. Thanks to a ratio for the longitudinal growth rate to the lateral one of typically 30 for GaN, independent, long and non-coalesced nanowires are formed.

Figure 4. The optical signature of monolayer-thick incomplete GaN quantum wells with a limited lateral extension in a single nanowire (NW) consists of a set of sharp lines at wavelengths shorter than 239 nm, corresponding to carrier localisation at the scale of small, flat GaN islands in AlN shown in the high-resolution, transmission electron microscopy image. The scale bar in the inset is 200 nm.

The remarkable properties of these nanowires make them promising candidates to overcome the difficulties that plague the efficiency of conventional UV LEDs, which are based on layer heterostructures.

First, with nanowire heterostructures, the large free surface is very favourable to elastic strain relaxation, which can proceed without producing dislocations detrimental to radiative recombination efficiency.

The second major merit of these miniature structures comes from the higher limit for solubility of electrical dopants in nanowires than layers. With LEDs with a conventional, planar architecture, introducing more foreign atomic species increases the stress in the heterostructure until it reaches a maximum amount, corresponding to the limit of solubility. Beyond this, additional dopants lead to precipitation of this species, as well as mechanical damage, such as crack formation. It’s a markedly different state of affairs for nanowires. Thanks to easy elastic relaxation associated with a large surface-to-volume ratio, it is possible to incorporate more electrical doping species –and this favours current injection.

Last but by no means least is the third key advantage of the nanowires – due to their discrete nature, an ensemble of separated nanowires behaves as a medium with an optical index that is intermediate between the index value of bulk III-nitride material and air. Due to this, nanowires offer enhanced light extraction over their planar cousins.