Fifty years of vacuum technology marked by evolutionary advances
10/01/2007
EXECUTIVE OVERVIEW
Vacuum technology is not often considered as leading semiconductor manufacturing. However, a look back at the last 50 years reveals that it has played an essential supporting role. Complex interactions between vacuum and manufacturing technologies have inspired the transition to dry pumps and the evolution of turbomolecular pumps. Meanwhile, vacuum processes have increased in number as fabs moved to producing ICs with tighter device geometries.
In the beginning of the semiconductor era, most high-vacuum systems consisted of oil-sealed, positive-displacement, mechanical pumps, and diffusion pumps for those processes that needed lower pressure. Improvements in residual gas analysis revealed that the primary source of contamination was the hydrocarbon oil in the mechanical rough pump. New oils were developed with lower vapor pressures, which led to faster pumping and lower pressures. The earliest integrated circuits needed vacuum only for metal interconnect deposition-lithography and etch were wet processes. The vacuum system needed to deposit the single layer of aluminum was typically no more than a bell jar connected to an oil diffusion pump [1].
In the 1970s, the appearance of low pressure chemical vapor deposition (LPCVD), plasma enhanced chemical vapor deposition (PECVD), and plasma etch and resist stripping processes brought a dramatic change in vacuum requirements. These processes did not need the high vacuum levels of sputter and evaporative deposition processes, but they did need pumps that could move a large volume of gas at a low vacuum. Oil sealed mechanical pumps fit the bill.
Unfortunately, the reactive gases and process byproducts were hard on the pumps, and vacuum manufacturers expended tremendous effort hardening them. The oil, used to seal and lubricate the pumps, was particularly vulnerable to breakdown and posed significant safety risks because of its flammability. New perfluoropolyether oils, eliminated the flammability issue, but were expensive. Soon chip manufacturers were spending heavily on oil recycling programs and vacuum pump manufacturers invested in improved filter development. Equipment downtime for pump maintenance added significantly to the cost of ownership.
When dry pumps were introduced in the mid 1980s (Fig. 1), they were not a new idea, but had always been dismissed as too expensive because of the higher precision machining they required. In comparison to the true cost of running oil based pumps, however, dry pumps became the clear winner.
Dry pumps brought their own set of challenges. Because of their tight mechanical clearances, they were intolerant of contamination, deposition, and corrosion processes that had been less troublesome for wet pumps. The solution was, and continues to be, customization for the requirements of specific processes: high temperature, caustic or corrosive materials, and high particulate content. Success in dry pump design for semiconductor applications requires extensive expertise in handling the byproducts of a wide variety of manufacturing processes.
Turbopumps provide another example of interactive development in vacuum and semiconductor technologies. Early turbopumps used extensively for aluminum etch, had oil lubricated bearings and bare aluminum rotors. The acid exhaust of the etch process quickly degraded the oil, and pumps could fail in as little as a few weeks. Initially, manufacturers attempted to protect the oil with nitrogen purging. However, deposition of AlCl3 from the process exhaust can only be prevented by operating the pump at high temperature, which destroys the lubricating oil in the bearing.
The ultimate solution was the introduction of magnetic bearings. These eliminated the need for oil and thermally isolated the rotor, permitting operation at temperatures high enough to prevent AlCl3 deposition, and dramatically increasing the time between chamber cleans. Today nearly all turbopumps in the industry use magnetic levitation, especially as the pump size has scaled up from 300 liters/sec class pumps in the 1980s to 3000 liters/sec today (Fig. 2). This scale-up has been almost linear with the shrink in device features.
Looking ahead
Progress in vacuum technology has come down to a game of particulars, with continuous evolutionary advances improving existing technologies by adapting them to specific applications. A current example of this kind of adaptation is atomic layer deposition (ALD).
In ALD, a volatile precursor is adsorbed as a monolayer on a surface. The monolayer is then exposed to a second volatile chemical that combines with it to form a permanent monolayer. If both precursors are pumped away through the same vacuum system, they can interact to form deposits that quickly seize the pump and are very difficult to remove. The deposits can be prevented by providing a system that switches automatically between separate vacuum circuits for the precursor materials.
The escalation of energy prices has focused increased attention on energy efficiency. As much as 20% of fab energy consumption can be attributed to process pumps. The latest generation pumps offer considerable energy savings relative to their predecessors, and lower power pumps are available for specific so-called “light duty” applications such as load-locks. The industry should also be poised to introduce further savings by active control of the pump utilities, standby modes etc.
A related trend to energy efficiency is the growing sensitivity to environmental issues, particularly the production of greenhouse gases and other noxious substances. Integrated systems that safely combine vacuum and abatement technologies in carefully engineered, process-specific solutions, are also more effective and more efficient than ad hoc solutions or a one-size-fits-all approach.
Conclusion
For much of the final quarter of the last century, semiconductor manufacturing processes tended to converge as the industry focused on shrinking planar CMOS designs. That trend has reversed in the first decade of this century, as manufacturers try to surmount some fundamental barriers with diverse new designs, processes, and materials. New materials will introduce new and more difficult challenges for the vacuum pumping systems. We are likely to see the continued refinement of existing technology for specific but divergent applications.
It is likely that the need for vacuum will increase in fabs. For example, if lithography moves to extreme ultraviolet (EUV), e-beam, or x-ray, it will need vacuum. Even nanoimprint will probably need some vacuum processing-if only to make the tooling. It is difficult to imagine building really small things without the clean, controlled environment provided by vacuum.
Reference
1. R.K. Waites, “Semiconductor Manufacturing and Vacuum Technology: A Memoir,” Solid State Technology, May 1997.
Stephen Ormrod received his BSc and PhD in electrical and electronic engineering from the U. of Nottingham. He is the CTO of Edwards, Manor Royal, Crawley, West Sussex, RH10 9LW, UK; ph 44/1293-528-844; e-mail [email protected].
Vacuum’s vital role in electronics manufacturing
We all learned early on about Thomas Edison’s invention of the light bulb and his heroic search for just the right material to make a long lasting filament. The incandescent light bulb was actually patented in Great Britain a year before Edison’s patent in 1879.
Edison’s key observation was that the length of time the filament lasted was directly related to the vacuum level in the bulb. His investigations focused on finding a viable method to produce a vacuum in a volume manufacturing operation. His name appears on 11 vacuum-related patents in the seven years following his “invention” of the light bulb [1]. It wasn’t the filament; it was the vacuum. Looking back now, one must wonder if Edison’s highly publicized search for a filament might not have been cleverly disguised misinformation intended to distance competitors.
Edison’s subsequent investigations led to his discovery of the Edison Effect, in which he observed one-way electrical current flowing between the filament and a nearby metal strip in the vacuum environment of the light bulb. That work was expanded by Fleming and De Forest in the next decade to create the vacuum tube, which was the foundation on which the electronics industry grew through the first half of the 20th century.
With the discovery of the transistor and the shift to solid state (that is non-vacuum state) electronics, pundits predicted an end to vacuum tube manufacturing and to the electronics industry’s all important drive for vacuum technology. From the 1970s, with the introduction of the IC, the opposite was to prove to be the case.
Reference
1. R.K. Waites, “Edison’s Vacuum Technology Patents,” J. Vac. Sci. Technol., A 21(4), July/Aug. 2003.