Cyclical process improves film quality

Thin-film deposition and growth processes lie at the heart of compound semiconductor device fabrication. Methods employed include physical vapor deposition (PVD) of metal films for seed layers and interconnects, plasma-enhanced CVD for silicon nitride dielectrics, and MBE and MOCVD growth of epitaxial layers for electronic and photonic structures.

Nanolayer deposition (NLD) can now be added to this list, thanks to the efforts of ourselves and co-workers at Tegal Corporation, which is headquartered in Petaluma, CA. This cyclical deposition technique, which is somewhat akin to atomic layer deposition (ALD) and draws on CVD and MOCVD technologies can produce highly conformal pinhole-free thin films that include metals and metal nitrides. One major plus point is the relatively low growth temperatures. TiN, for example, can be deposited at just 250 °C, which is an advantage when processing GaAs wafers because higher temperatures can lead to dopant diffusion in epiwafers and other undesirable effects.

NLD employs MOCVD precursors to deposit a film that is typically 0.2–1.0 nm thick. This is then exposed to a downstream plasma that converts the film into a very pure, high-density nanolayer. Further film growth to reach its target thickness results from cycling the growth and plasma treatment (figure 1). Since the underlying chemistry is based on MOCVD precursors, many materials can be deposited, including nitrides and oxides, such as TiN, TaN, TiON, ZrO2, and metals such as ruthenium and copper.

There are many applications where NLD offers an alternative to ALD. The latter growth technology is often used for highly conformal metal and metal-nitride film growth and has four major applications, according to IBM s Hyung Giun Kim: the fabrication of diffusion barriers and adhesion-promoting layers; deposition of copper seed or other direct plating layers; growth of low-resistance, chemically and thermally stable liners in contacts and vias; and the creation of forming liners, dielectrics and electrodes in memory-storage cells.

The variety of uses makes ALD a very versatile technique. However, it does have a major, inherent drawback – self-limiting growth. Less than a monolayer is added during each deposition cycle, so many more ALD cycles are required to produce a film of a given thickness than with NLD.

Our NLD process module includes our very own direct vapor draw precursor delivery system that supplies various MOCVD precursors into the reactor. An inductively coupled plasma source is attached to a remote plasma zone, which allows plasma species to be injected into the deposition area. Precise wafer temperature control is ensured by an electrostatic chuck assembly, and turbo pumps provide high vacuum performance. Multicomponent and multilayer films can also be formed through upstream deployment of additional precursor delivery submodules.

The thickness of the monolayer is controlled by the deposition time or the growth temperature, which can be adjusted from less than 100 to more than 500 °C. It s critical to get a good balance between system throughput and film quality: too thick and the quality diminishes through incomplete plasma treatment; too thin and throughput suffers because more cycles are needed.

Plasma treatment is the key to the success of NLD. As-deposited film quality strongly depends on the deposition temperature. High-temperature films are generally superior, but conformality – a measure of the thickness uniformity of a film over complex structures, such as wafers with narrow trenches – tends to suffer. This can be combated with lower growth temperatures that improve conformality through a reduction in the precursor sticking coefficient. However, lower temperatures also increase impurities, predominantly carbon, which in turn impact film quality.

NLD avoids these problems with downstream plasma treatment. This improves as-deposited film quality by removing the impurities through reactions with the plasma species. Volatile compounds are formed instead, which help to "densify" the film. This layer can be treated quickly, which enables high throughput, thanks to the plasma s high density – it is highly dissociated and vigorously reactive. Since the plasma is generated upstream, and because the wafer is not biased, all of the growth surfaces are treated in the same way, including the sidewalls of high-aspect-ratio features.

Film quality also benefits from our direct vapor draw precursor system. Carrier gases are not used to push the precursor into the reactor, and this reduces the total process gas flow and pressure. The upshot should be fewer particles in the process compared with conventional carrier gas designs.

We have verified our NLD tool s capability through the growth of a 20 nm thick conformal TiN film in a very deep trench (figure 3). Grown at a temperature of 250 °C, the film was pinhole free and had a resistivity of 500 µΩ cm.

We believe that our NLD reactor is a very promising tool for high-brightness (HB)-LED production and we have already received an order for our Compact 360 NLD from manufacturers of these devices. Our deposition tool combines low cost of ownership with an ability to produce highly conformal films of excellent quality at low temperatures. This should help to address the grand challenges that are facing HB-LEDs, which include improvements to internal and external quantum efficiency and packaging technology. We are determined to demonstrate how our NLD process can help to overcome these obstacles.

Our tool can also play an important role in 3D packaging, which is a key driver for creating the small, highly integrated components for cell-phone handsets, smart phones and other mass-market consumer electronics products. NLD is capable of fulfilling many of the applications that we highlighted for ALD, and it can also be applied to the creation of "through silicon vias" (TSVs). Three-dimensional TSV ICs have large features and deep structures, and we look forward to the adaptation of NLD processes to the specific requirements of this new and important packaging application.

Further reading
J Crowell 2003 J. Vac. Sci. Technol. A 21 S88.
H Kim 2003 J. Vac. Sci. Technol. B 21 2231.   

View pdf of article