AlN: Opening doors with low-temperature epitaxy
By delivering record electron and hole concentrations in AlN,
low-temperature epitaxy promises to unleash a new generation of extreme
BY CHRISTOPHER MATTHEWS, HABIB AHMAD, KEISUKE MOTOKI, SANGHO LEE, AHELI GHOSH, EMILY MARSHALL, AMANDA TANG AND W. ALAN DOOLITTLE FROM GEORGIA INSTITUTE OF TECHNOLOGY
Unfortunately today’s materials are not going to meet tomorrow’s needs. That’s not to say that GaN and SiC have not had substantial success – they are transforming power electronics and making solid-state lighting a reality. However, these wide bandgap materials have limitations when employed for high-power diodes and transistors for grid-scale electronics and electric vehicles, and they cannot be used to make ultraviolet emitters for sterilisation and lithography.
Offering much promise on all these fronts are a number of materials with even wider band gaps that have been attracting increasing attention in recent years. They include Ga2O3, diamond, and very recently, AlN.
For all of these materials, progress at the device level is far from easy. Ga2O3 has a number of great attributes, including a bandgap of 4.9 eV, a high critical electric field and compatibility with inexpensive bulk growth methods. But there are weaknesses: devices are limited to those that are unipolar, due to only n-type doping; and recent results have highlighted the low thermal conductivity of this oxide, hindering its applicability to power electronics.
Figure 1. Minimum required device thickness versus grid voltage for various points along the power transmission grid, assuming the theoretical critical electric field limits (taken from J. Y. Tsao et al. Adv. Electron. Mater. 4 1600501 (2018)). Thicknesses below 200 µm are considered reasonable for semiconductor devices. For higher voltage transmission applications, every material requires an unreasonably high device thickness. However, AlN is the only material compatible with substation-level applications at both 230 kV and 138 kV. While multiple materials can hold off lower voltages, AlN devices require the least material, which makes them a comparatively economical choice.
Offering an even higher bandgap of 5.5 eV is diamond, which benefits from a high thermal conductivity and other favourable material properties. However, despite impressive recent substrate advances, diamond is difficult to dope and expensive to produce. This material is also hampered by its indirect band gap, making it unsuitable for optoelectronic applications, which have been a major leveraging factor for driving early investment and research in other semiconductors, such as GaN.
AlN has also garnered interest, thanks to its direct bandgap of 6.1 eV, its high critical electric field of 15.4 V/cm, its high thermal conductivity of 319 W/mK, and its compatibility with the established III-nitride infrastructure. However, doping AlN is challenging. Prior to 2020, reported carrier concentrations were limited to around 1010 cm-3 for holes and 1015 cm-3 for electrons.
Our team at Georgia Institute of Technology has smashed through this ceiling with metal-modulated epitaxy (MME), a low-temperature variant of MBE. Using MME, we have realised p- and n-type doping in AlN with beryllium and silicon dopants, respectively, obtaining carrier concentrations greatly exceeding previously reported values – they are well above 1018 cm-3. Read on to discover how low-temperature MME enabled this success.
In III-nitride semiconductors, the doping efficacy is largely influenced by the concentration of unintentional point defects. They include: contaminant atoms, typically oxygen, carbon, and hydrogen; cation vacancies; N-vacancies (VAl and VN for AlN); and reconfigurable defects, such as the DX-centre. Thus, it is of utmost importance to limit the incorporation of undesired impurities and the formation of vacancies, especially when trying to dope notoriously insulating materials, such as AlN.
These impurities, arising from various sources, are dependent on the growth technique. For oxygen and carbon contamination, two competing thermal mechanisms are at play. Higher temperatures increase the extent of the degassing of these elements from structural components within a growth chamber, but this is somewhat counterbalanced by a simultaneous decrease in their sticking coefficient on a semiconductor surface. To limit the incorporation of oxygen and carbon in thin films, growth is typically conducted at diametric temperatures – that is, as low or as high as possible.
For III-nitride epitaxy, the formation of vacancies is often thought to be related solely to the III/V (or V/III) ratio, with VAl occurring for N-rich growth and VN occurring for metal-rich growth, with the majority flux used to replace atoms lost to desorption. More generally, the concentration of vacancies is related exponentially to the temperature, so one useful strategy to reduce the vacancy concentration is to dramatically reduce the temperature.
One example of the diametric approach is MOCVD. This epitaxial growth technique employs high temperatures and thermodynamically produces high concentrations of VN. However, kinetically, the massive V/III ratio that’s used in MOCVD replaces these VN as they are produced.
The opposite approach, involving low temperature methods, never creates the enormous surface concentrations of vacancies by removing the thermodynamic driving force. Due to this, even excessive aluminium-rich III/V ratios lead to a low concentration of VN, regardless of the flux that’s used.
Figure 2. Representative calculations of the hoping rates and surface diffusion lengths as a function of various atomic diffusion pathway barriers. Nitrogen in a gallium bilayer has a diffusion length on the order of one micron, even at low growth temperatures.
In most cases, the epitaxy of compound semiconductor layers is conducted near the highest temperature possible for that material, typically helping to promote the growth of smooth films. For the III-nitrides, this approach has been pursued to optimise films that are grown under nitrogen-rich growth conditions. However, there is a dramatic shift in physics for metal-rich growth that’s been overlooked. Essentially, high temperatures are needed in the nitrogen-rich regime to overcome large surface binding energy barriers and promote long-range adatom hopping, critical for the formation of smooth films. These barriers are much lower when adatoms exist on a metal-rich surface, ensuring that a similar degree of diffusion is possible at much lower temperatures.
It is easy to see these factors at play when comparing the diffusion lengths of the limiting reactants in GaN growth on different GaN surface terminations. For N-terminated (or dry) GaN surfaces, the diffusion barrier for gallium is between 1.4 eV and 1.8 eV. In comparison, the barrier is only 0.7 eV for nitrogen diffusion on a gallium terminated surface, and it is as low as just 0.12 eV for surfaces covered in a bilayer of gallium. Due to these variations, adatom surface diffusion lengths are significantly higher for metal-rich growth, exceeding 3 µm, compared with lengths for nitrogen-rich growth, even at much lower temperatures.
Figure 3. Typical flux profiles as a function of time in metal-modulated epitaxy. Instantaneous metal flux (ΦIII) is higher than the nitrogen flux (ΦN), ensuring that growth occurs in metal rich conditions. The time-averaged metal flux (ΦIII,avg) is lower than (ΦN), eliminating droplet build-up.
Based on this observation, reconsidering the use of high growth temperatures is worthwhile. After all, high quality III-nitride films can be realised at low growth temperatures using metal-rich conditions, while high growth temperatures have several downsides: there is an increase in impurity generation due to outgassing of structural components; greater thermal expansion takes place, introducing unwanted strain; and doping can be impaired.
Unfortunately, it’s not as easy as simply lowering the growth temperature and using exclusively metal-rich conditions. If that approach is adopted, it leads to an excess droplet build-up of metal, and ultimately has adverse effects on III-nitride film quality. What’s needed is a modulated growth technique, alternating between metal- and nitrogen-rich growth conditions. Typically implemented as a variation of MBE, this approach takes advantage of the benefits of metal-rich growth and limits the accumulation of excess metal.
Chief among these modified growth methods is MME. With this form of epitaxy, the MBE metal flux is in excess of the nitrogen flux by between 30 percent and 200 percent. A key part of the process is physically shuttering the metal sources, undertaken in a manner that ensures that the time-averaged growth conditions are nitrogen-rich through periodic removal of all metal accumulation (see Figure 3). When the metal shutters are opened, the surface is quickly terminated by a metal adlayer. Accumulation continues until the metal shutters are closed, at which point the metal adlayer is consumed by the continuous supply of active nitrogen. Once the metal adlayer is fully depleted, the film briefly stops growing until the metal shutters are re-opened and the cycle repeats. However, since the growth reaction only depends on the amount of adsorbed metal on the surface, rather than the metal flux, growth (almost) always occurs in metal-rich conditions.