Plasma-assisted atomic layer deposition (PA-ALD) is a cyclic, low-temperature thin film deposition method, in which the substrate surface is exposed to sequential pulses of precursor molecules and plasma species separated by evacuation and/or purging periods. When compared to other techniques, ALD stands out with its self-limiting growth mechanism, which enables the deposition of highly uniform and conformal thin films with sub-angstrom thickness control. These features make PA-ALD a promising and alternative technique for the low-temperature deposition of III-nitrides and their alloys in post-CMOS processing and 3D integration technology.

In our previous reports on the PA-ALD of polycrystalline wurtzite AlN thin films at temperatures ranging from 100-500 °C using trimethlyaluminum as the Al precursor, films deposited at temperatures within the ALD window (100–200 °C for both NH₃ and N₂/H₂ processes) were C-free and had relatively low O concentrations (<3 at.%). Our initial efforts for depositing GaN thin films, however, resulted in amorphous thin films with high O concentrations (~20 at.%). Following experiments revealed the source of this O contamination as the quartz tube of the inductively coupled RF-plasma source itself. In view of these circumstances, the choice of N-containing plasma gas (N₂, N₂/H₂ or NH₃) determined the severity of O incorporation into the deposited AlN and GaN thin films.As an effort to completely avoid this contamination problem, we integrated a stainless steel hollow-cathode plasma (HCP) source to the ALD system, and thereby reported on hollow cathode PA-ALD (HCPA-ALD) of nanocrystalline AlN and GaN thin films with low impurity concentrations at 200 °C using trimethylmetal precursors.Within the scope of the same study, Al_xGa_1-xN thin films were also deposited via digital alloying, where alloy composition was determined by the relative number of AlN and GaN subcycles in the main HCPA-ALD cycle.

In this presentation, we will review our recent efforts on the development of low-temperature HCPA-ALD processes for III-nitride alloys including GaN, InN, In_xGa_1-xN, and In_xAl_1-xN thin films. In-detail materials characterization results including structural, optical and electrical properties as well as potential device architectures for post-CMOS processing and 3D integration will be presented and discussed.

Metalorganic Chemical Vapor Deposition (MOCVD) is a technology of choice for large scale production of GaN based LED and Power Electronic devices. For the last 20 years MOCVD equipment evolved from small R&D oriented deposition systems (three 50 mm wafers per run) to large industrial cluster type systems (two hundred sixteen 50 mm wafers per run) with very sophisticated in-situ devices and process control. Evolution of the production scale MOCVD equipment is driven by one major goal – Cost of Ownership (CoO) reduction. Industry is achieving this goal utilizing several trends:

Migration from R&D to production requirements. GaN MOCVD systems started as an R&D tool. Further development of these systems is the path from universality and flexibility typical for R&D tools to stability and simplicity required for production environment.

Increasing batch size. This is the most obvious way to improve CoO as the cost to manufacture system with two times more wafers per run is less than the factor of two. All major MOCVD equipment companies follow this trend and release new larger batch systems every 3-5 years. One of the most important questions for scaling up is the limit of this trend.

Move from the single reactor system to the cluster and increase level of automation. Majority of modern MOCVD systems migrated from the single reactor to the cluster type multi-reactor design with central loading module and wafer carrier transfer robot.

Increasing role of the in-situ devices for wafer parameters measurement and control. Evolution of in-situ devices for production system includes the following sequence: thermocouple - conventional pyrometer – reflectometer - emissivity compensated pyrometer - deflectometer (wafer bow measurements). There is also a trend for more sophisticated control methods that move from PID to predictive and model based algorithms.

Increased wafer carrier complexity. The wafer carrier is a unique component of MOCVD system that to a large degree defines system yield. Complexity of the wafer carriers is constantly increasing with the goal to improve deposition uniformity. Wafer carriers are a subject of majority MOCVD equipment patents.

Increased role of process modeling. Troubleshooting and process optimization in production environment exclude “trial and error” approach and require good computational models for flow dynamic and process chemistry that are fine-tuned based on experimental data.

In this presentation we will describe above trends in detail and make an attempt to predict next steps in the development of the equipment for large scale production of GaN based materials.

This presentation will describe results recently obtained with pulsed electron beam deposition (PED) of GaN on sapphire, silicon (111), and 2 nm germanium coated silicon (111) substrates. The PED technique is potentially useful for growth of III-nitrides at lower substrate temperatures, a capability that can allow use of new buffer layer materials, introduction of chemically dissimilar lattice-matched materials, and help solve wafer bowing and cracking problems during growth. The introduction of this technique could lead to improvements in device quality and fabrication of vertical LED structures. In this study, GaN was deposited on sapphire at a substrate temperature of 750°C, and on silicon (111) and Ge/Si (111) at 600°C in a UHP N₂ (15 mTorr) environment (without any surface pre-treatment such as pre-nitridation). A high power electron gun pulse (Neocera, Inc) was used to ablate the GaN target (1” dia. x 0.250” thick, 99.99% pure) stationed at 5 cm vertical distance from the substrate. The electron pulses were generated at 15KV, 0.3 J/pulse at 1 Hz for initial few nm of growth, and then increase to a 3 Hz pulse rate. Scanning Electron Microcopy (SEM), X-ray Diffraction (XRD), Rutherford backscattering, and optical absorption characterization were performed. SEM imaging confirms a rough surface morphology with the presence of 30 nm to 300 nm scaled GaN crystallites (for the GaN/Sapphire sample) while smaller but more coalesced crystallites of 30-50 nm size is observed for GaN/Si (111) and GaN/Ge/Si (111) samples. The average film thickness is 350 nm for the samples, yielding a growth rate of 0.16 angstrom/pulse. From SEM, it appeared that high aspect ratio filament structures have grown over the crystallites. XRD θ–2θ scans from 2θ = 0° to 2θ = 70° on the GaN on sapphire showed only two other peaks, besides the peaks from the sapphire, near 2θ = 34.6°. The peaks near 2θ = 34.6° consist of a stronger peak at 34.668° and a much weaker peak at 36.903°. These peaks correspond to the (0002) and (10-11) orientations for GaN, respectively. XRD θ–2θ scans from 2θ = 0° to 2θ = 70° on the GaN on Si (111) and GaN on Ge/Si (111) samples show presence of only polar GaN (0002) peak at 34.7° besides the Si (111) peak at 2θ = 28.5°. The XRD results clearly show that the deposited GaN material is not polycrystalline. Optical absorption spectroscopy over a 1.2 eV to 6.2 eV spectral range, for the GaN/Sapphire sample, showed an abrupt absorption edge at 3.4 eV, a clear indication of interband transitions in binary GaN. These results confirm that our PED-grown GaN is highly c-axis oriented and suitable for the initial growth of GaN on various substrate materials.

The unique optical and electrical properties of InN and related ternary InGaN alloys make the material system attractive for various optoelectronic device applications, including but not limited to high-speed electronics, photovoltaic solar cells, or light emitting devices. Even though progress has been made in establishing the base properties of the binaries InN and GaN, the growth of high-quality InN and indium-rich ternary InGaN epilayers and heterotructures is an open challenge. In previous work, we demonstrated the stabilization of InN and InGaN epilayers utilizing superatmospheric MOCVD (also denoted as HPCVD) to suppress the decomposition at higher growth temperatures.