In Japan, Europe and the US, money is being poured into the development of synchrotron technology, specifically the development of a compact electron ring small enough to be incorporated into a semiconductor wafer fabrication facility. This goal, when achieved, will lead to cmaller, denser, cheaper silicon chips – and the rewards will be rich for the first to come up with the goods. The current state of play has the UK and Japan at the forefront of research and development, with the US slightly behind and buying in some compact technology.
Helios
IBM has recently announced the shipment of UK-based Oxford Instrument’s Helios ring to its integrated circuit facility in New York State, an announcement that immediately added 42 pence to the value of Oxford’s shares. But although IBM deemed the ring sufficient to be shipped, it’s giving no clues as to when it will actually be used in chip fabrication. The Helios can store an electron beam of over 50mA with a lifetime of one hour at the full electron beam energy of 700 MeV. Oxford says the ring could be used to make 64M-bit and 256M-bit memory chips and that the machine will probably be ready for test exposures on both in the second half of next year. The ring is considerably smaller than most others, measuring five metres in length, two metres in width and three metres in height. Oxford Instruments expects to sell the machines at around $25m apiece. Development of the technology is phenomenally expensive, which explains why there have been so many collaborations. In October last year, IBM teamed up with Motorola in an attempt to head off increasing Japanese competi-tion. Motorola is supplying an undisclosed sum to help IBM’s research and development effort in return for a crack at trying out the X-ray generating machine on some of its own parts. Other Western tie-ups include the agreement between IBM and Siemens in January, in which the two decided jointly to develop a 64M-bit dynamic ram for commercial introduction in the mid-1990s. Siemens’ president, Karlheinz Kaske said the move was part of a long term effort to supply locally fabricated integrated circuits to European industry. Other DRAM research and development collaborations include the Megaproject joint venture between Philips NV and Siemens AG – if Philips can afford to keep up the subscriptions needed to keep it in the game, and the US Sematech operation. But there are also joint ventures in DRAM manufacturing between US and Japanese companies, Intel and Minebea Co’s NMB Semiconductor affiliate of Japan announced one at the beginning of the year.
By Sonya McGilchrist
Japanese corporate and government funding for research and development is perhaps the most dramatic. It includes several large-scale projects targetting 0.2 to 0.25-micron geometries which, as with Oxford’s machine, are aimed at producing 256M-bit memory chips. Nippon Telegraph & Telephone Corp is moving to upgrade its existing equipment built around a compact superconducting synchrotron orbital radiation ring. Within three years it believes it will have 0.04-micron mechanical alignment accuracy – needed for the 256M-bit dynamic RAM devices. Sortec is a group of 13 Japanese electronics companies. Its $90m research project is 70% funded by the Japanese Ministry of International Trade & Industry, the balance being provided by the 13 member companies. These are Toshiba, NEC, Hitachi, Fujitsu, Matsushita Electric, Mitsubishi Electric, Oki Electric, Sanyo, Sharp, Sumitomo Electric Industries, Sony Corp, Canon and Nikon. Sortec has completed a twin-ring, high-power synchrotron orbital radiation ring capable of supplying the working-level 200mA constant current, necessary for lithography purposes. The ring consists of a 40MeV injector and a pair of rings 15 yards in diameter including a 1,000MeV accelerator ring and a 200mA storage ring. In March it completed installation of an aligner. So far, the development has cost the Sortec consortium $19m to build. The group is confident of meeting a self-imposed target of developing basic X-ray lithography tech
nologies for 256M-bit parts by 1995.
Obsolete
But whoever finishes the development and gets there first, when electron rings do become small enough and cheap enough to be viable for commercial chip production, traditional optical lithography will eventually become obsolete and the new chips should revolutionise the industry – not least in price terms. Optical etching technologies cannot be used to create circuits with geometries finer than 0.5 microns or so, because beyond that point, diffrac-tion makes accurate focussing impossible. With X-rays, diffraction only starts to cause trouble at the nanometre level, enabling much finer circuits to be etched and many more devices to be put onto one chip. One chip will therefore be able to replace many smaller capacity chips and provided cooling problems can be solved, complete top-end mainframe processors will be avble to be shrunk to a handful of integrated circuits from the 130 or so that make up the heart of IBM’s 3090 processor. They will also be faster, because everything will be closer together, on one chip, instead of over several. Once the technology is fully developed, making one chip should be a lot cheaper than making three or four, and this should affect box prices, leading to another tenfold advance in price-performance – and another round of headaches for the companies that make a living selling hardware.
