The world's leading source of technology news and analysis
Search Spectrum IEEEXplore Digital Library Submit
Font Size: A A A
IEEE
Home [Alt + 1] Magazine [Alt + 2] Bioengineering [Alt + 3] Computing [Alt + 4] Consumer [Alt + 5] Power/Energy [Alt + 6] Semiconductors [Alt + 7] Communications [Alt + 8] Transportation [Alt + 9]

Chip Making's Singular Future Continued By Rajendra Singh and Randhir Thakur

emailEmail PrintPrint CommentsComments ()  ReprintsReprints NewslettersNewsletters

The solution, we believe, lies in a fundamental change in the machines that process the wafers: a switch from batch to single-wafer manufacturing.

The single-wafer approach is a completely serial one, in which just one wafer is processed at a time, all the way through the factory from start to finish. There is never a time when the machines work on a large batch of wafers at the same time, as they do today. The single-wafer technique will solve the oversupply problem by shortening the time it takes to make a finished, packaged chip to less than one month, rather than the three months or more that is typical today. Basically, with single-wafer manufacturing, semiconductor companies will be able to produce chips quickly when the orders come in, in the exact quantities specified by those orders. There will be no need to build up huge inventories that may just sit on shelves until they become obsolete.

And it isn't just boutique chips, which are made in small quantities, that would benefit from the single-wafer approach. Even commodities like static random-access memory (SRAM) and microcontroller chips, which suffer from periodic oversupply and the resulting price plunges and reduced profits, would benefit from a more agile response to changing market demands.

So what will it take to shift to single-wafer manufacturing? First, consider today's typical semiconductor plant. It combines single- and batch-processing steps; some of the machines process wafers in groups, while others already process them singly. True single-wafer manufacturing eliminates all the batch processes and uses only machines that process wafers one at a time. Today, only a few semiconductor plants have switched over completely to single-wafer manufacturing.

In 2001, Trecenti Technologies Inc. of Hitachinaka, Japan (now part of Renesas Technology Corp.), adopted 100 percent single-wafer processing for the fabrication of advanced semiconductor ICs on 300-mm wafers. The company's experience with this technique has been remarkable. It has found that it can reduce manufacturing time from 90 to 30 days, and the number of days needed for each chip layer has dropped from 2.25 to 0.25. Even more remarkable is the improvement in the fabrication time for a wafer of SRAM chips made up of 130-nanometer structures. That time has dropped from about 60 days to fewer than six days.

Several other IC manufacturers are also currently considering 100 percent single-wafer processing. Freescale, Philips, and STMicroelectronics have formed the Crolles2 Alliance. Its 300-mm wafer facility, in Crolles, France, uses single-wafer processing for most steps. Tokyo-based Toshiba Corp.'s minifab, in Oita City, Japan, is another example of IC manufacturing dominated by single-wafer processing.

The fabrication of a finished, packaged IC has three key groups of processing steps, regardless of whether it happens in a conventional or a single-wafer setup. First, the transistors are made; next, they're wired together into circuits on the chip; and finally, the small slivers of silicon, known as die, are packaged into finished chips [see illustration, Once and Future Process].

Making transistors, described simply, is a sequential process that builds the four major transistor components: the source, drain, channel, and gate. Basically, the first step in the formation of the transistor is to coat a silicon wafer with a photosensitive material, called a photoresist. The photoresist is exposed to light in the pattern of the areas in which transistors are to be built. Each of these regions will eventually contain the transistor's four components.


« Previous Page 2 of 6 Next »
emailEmail PrintPrint CommentsComments ()  ReprintsReprints NewslettersNewsletters