Reactive Sputtering of Gallium Nitride for Electroluminescent Applications 

      I began working on GaN in 1987 When I was a Senior in college. My first Notebook entry from the research is dated July 8, 1990. One of my research advisors, E. E. Crisman, was familiar with the work of Pankove and others and thought the technique of Reactive Sputtering might solve some old problems in the material system.

     At the time, I was working on eptaxial growth of GaAs in a Laminar Flow LPE system and had just been introduced to the magical world of Semiconductor Electro-Optics. The LPE project I was working on was for Solar Cells, and I had studied enough to understand heterostructure lasers and LEDs as well. I was really taken by the idea of active-element flat panel displays, and worked for the next 6 years off and on on GaN and related compounds. This lecture will give a brief overview of some of the more interesting aspects of that research, with special emphasis on the parts I played.

     The work of Pankove, Maruska and others in the 60's and 70's had shown that GaN could be grown from Ga melts and the vapor phase, and that epitaxial GaN on Al₂O₃ showed luminescence at a variety of wavelengths, depending on impurities and preparation.

Some Properties of GaN and Related Compounds

Compound	Decomposition Temperature, °C	Lattice Constant, a, Å	Lattice Constant, c, Å	Energy Gap, EV	Dielectric Constant
BN (cubic)	3000	3.615	3.615	6.4	7.1
AlN (hex.)	2800	3.111	4.978	6.28	9.1
GaN (hex.)	1000	3.186	5.178	3.45	5.2
GaN (cubic)	?	4.54	4.54	3.26	?
InN	620	3.541	5.705	1.95	?

     The two main ways to produce photons from electricity in a semiconductor are the Light Emitting Diode (LED), which is essentially a current device, and the Metal Insulator Semiconductor (MIS) diode, which is essentially a field device. The LED has several advantages over the MIS, but requires both positive majority carrier (p-type) and negative majority carrier (n-type) material to be grown in a homojunction. All the early results were from MIS structures, though, because it was difficult to produce p-type GaN.

      A very convincing argument is made by Wager that the reason p-type GaN is hard to make is that it is both thermodynamically and kinetically unstable. We concluded that a growth process at a low, "non-equilibrium" temperature might get around this problem. The basic problem is that as the Enthalpy of formation of an ionized donor vacancy approaches the band gap energy of the material, it is energetically more favorable to create a new ionized vacancy than to allow an uncompensated acceptor impurity.

      Based on this information, and because of what equipment we had available, we decided to proceed with an investigation of reactive sputtering of GaN and related "Group III Nitrides" (GaN, AlN, InN and alloys thereof) Sputtering is a deposition or etching technique (depending on bias) whereby atoms or ions from a plasma impact a surface, knocking off atoms, molecules or chunks of the material being sputtered. Sputtering is usually accomplished with Argon gas, since even Ar ions are not chemically reactive, and Ar is easilly obtained in a pure state from liquid air, and is therefore cheap. In our experiments, we used mixtures of Ar and N₂ and NH₃ to mix a chemcally reactive species into the sputtering plasma.

      The basic procedure for all the nitride materials we made was to make a target and sputter it under different conditions. Normally, metals are machined into disks and clamped into the sputtering head, but there is a special property of Ga that leads to what is referred to by physicists as a gravitational symmetry breaking. The low melting point of Ga (30° C) means that it must be treated as a liquid under normal laboratory conditions. I remember I got a real kick out of the fact that we had solid gallium metal in plastic squeeze bottles. Therefore, all targets were prepared by alloying Ga with various dopants and pouring the molten alloy into a Cu disk with a cup machined into it. We initially had trouble getting the Ga to "wet" the Cu, but when we swabbed the hot Cu disk with H₂SO₄ as a flux. No sulphur was found in subsequent chemical analysis by SIMS, AES and XPS. Targets were prepared with impurity concentrations (in mg/g, or parts per thousand by weight) of 0.1 and 20 mg/g. One target was prepared with 7 atomic percent In.

      Since the whole point of this project was nominally to produce flat-panel displays, it was important to demonstrate working EL devices and measure their spectra. We were able to produce material with cathodoluminescence and devices with electroluminescence at blue, yellow and red wavelengths. Unfortunately, the devices were not very efficient, and quantitative EL spectra were elusive. Then one day, a salesman for a company that made optical multi-channel analyzers came by. He offered to demonstrate his machine on one of our specimens. Reproduced below is the one and only trace we got, which has now been in at least three publications before this.

      The main parameters varied in the materials preparation were:

• Composition of Target
• Composition of Sputtering Plasma
• Plasma Density (Usually adjusted to optimize sputtering rate)
• Substrate Material
• Substrate Temperature

      The films turned out to be amorphous or polycrystalline, as measured by x-ray diffraction. Amorphous films resulted from sputtering in NH₃ while columnar polycrystalline material was produced in N₂. P-type material was produced from targets with High Zn doping, and high resistivity N-type material was produced from undoped targets. We concluded from the crystallinity results that, with NH₃ as a sputtering gas, N ions reacted at the target (Ga) surface, forming GaN which was then sputtered onto the substrate. We believe that with N₂, some less reactive species (Free N ?) reacts with sputtered Ga near or at the substrate, growing columnar polycrystalline material. The columnar nature of the films is related to attachment kinetics, and Akasaki and Amano have shown that this can be overcome by growing GaN films on AlN buffer layers.

      Some of the main physical properties of the process are listed below:

      I was involved in doing resistivity measurements in a Hall effect system I set up for standard thin-film analysis. I was able to consistently measure hole mobilities on the order of ~7 cm²/V-s. I was to see a lot more of the Hall Effect when I measured the Quantum Hall Effect in GeSi for my thesis.

      Much of the device processing for our first run of EL devices was done in a single sputtering system. We were able to produce the Al and GaN layers from the DC sputtering head, and the SiO₂ from an AC sputtering head. Both the SiO₂ and the Indium Tin Oxide (ITO) transparent electrode layer were deposited in atmospheres with intentional O₂ partial pressures. To avoid problems associated with etching and wet chemistry, all the depositions were done either all over the surface or through small stencil "shadow-masks" Alignment was rough, but sample purity was maintained, and the Process was potentially automatic, with a spindle changing masks inside the deposition chamber.
As it was, we think that oxygen contamination of materials may have been a problem. Oxygen was found in analysis of all films, and the system had no load lock, so targets would oxidize with each opening.