How a semiconductor gets manufactured is an interesting process,
well-documented on several websites. Sematech, an industry
consortium, describes the process simply in 9 steps, from growing the
silicon ingot to dicing the individual ICs:
http://www.sematech.org/public/news/mfgproc/mfgproc.htm
Infrastructure, an electronics industry publication, describes the
growing of the silicon ingot over a 24-hour period on this web page:
http://www.infras.com/Tutorial/sld002.htm
Once the ingot is grown, it cools and is ground down to the size
required by the wafer slicers, which are actually diamond saws. Once
sliced into wafers, they are market with an ID number by laser. The
wafers are too rough for production yet, so they also must be
polished. The entire process of growing the ingot, marking, slicing
and polishing takes less than one week. This is a Department of
Energy technical article showing additional information, such as yield
losses at each level of processing:
http://www.nrel.gov/silicon/Nmulticrs.htm
According to Intersil, a semiconductor manufacturer, the balance of
the process of getting from wafer to die is another 10-30 days,
depending on the process:
http://rel.intersil.com/docs/lexicon/manufacture.html
So the total crystal-to-die process is done in a 17-37 day period!
For an interesting and detailed look at producing ICs, the Industry
Initiatives for Science and Math Education (IISME) produced this
step-by-step introduction to the process. Its well supported by
photographs of each step:
http://iisme.org/etp/Silicon_Wafer_Processing.pdf
Now: how many are produced each year? The Semiconductor Industry
Association (www.semichips.org) produces quarterly reports, so you
detailed reports are publicly available. The numbers are for merchant
IC manufacturers, which the SIA notes are the great majority of the
worlds IC production. Not covered are captive manufacturers, such
as IBM Microelectronics, which produces some integrated circuits for
use in IBM computers.
The report through Q2, 2002 has lots of detail and shows that wafer
starts are down this year over 2001. In 2001, there were about 5.2
million wafers produced in 8 equivalents. The industry is using 8
equivalents because over half of the production is now in 8 wafers;
7% in newer 300mm (or 12); and the balance in 5 wafers. Details are
on the SIA website here:
http://www.semichips.org/downloads/SICASQ22002.pdf
Please note that 8 wafers have a surface area of almost 50 square
inches, while 5 wafers have an area of 19.5 so it takes about 2 ½
5 wafers to produce as many die as the larger wafers.
Though market share for semiconductors is available, with the SIA
giving the U.S. 51% of the worlds $139 billion market, numbers for
U.S. production of wafers dont appear to be available. Major
suppliers, such as Shin-Etsu Chemical, which claims 25% of the world
wafer market, often produce silicon ingots wherever customers are
located. Shin-Etsu has two plants in Louisiana, in addition to
factories in Japan.
The computer research organization Gartner Group tracks semiconductor
and equipment sales, but apparently its figures by country dont seem
to be available on the Net:
http://www3.gartner.com/1_researchanalysis/focus_areas/semiconductors/semimkt_fa.html
Gartner Group and VLSI Research, another market research firm tracking
the semiconductor industry, introduce another way of looking at wafer
production using millions of square inches. In 2002, VLSI and
Gartner both foresee growth to between 4 million square inches (VLSI)
and 4.7 million square inches. The business will continue to grow by
11-12% per year for the next five years, both companies forecast.
Semiconductor sales by region are detailed by SIA, showing market
shares over a 10-year period here:
http://www.semichips.org/downloads/ACF14E.PDF
I hope that you find these web links as interesting as I did! |