The Production Process of Silicon Chips
The silicon chip is the most important component that builds up the
microprocessors which runs the PCs we use everyday. It is a piece of almost
pure silicon, usually less than one centimeter square and about half a
millimeter thick. On the chip are hundreds of thousands of micro miniature
electronic circuit components, including diodes, capacitors, and resistors
but mainly transistors. All these components are packed and interconnected
in many layers, referred to as an IC (integrated circuit) chip. Most chips
have 4~6 layers, but some have more than 15. On the surface of the chip,
there is a grid of thin metallic strips, which are the electrical connections
via wires to the outside world.
Since silicon is the second most abundant element on earth,
crude silicon is abundant for chip manufacturers. The silicon used to produce
chips is usually refined from quartz rocks. The first step in the manufacturing
process involves melting the silicon crystals and molding them into long
strips. In the silicon chip manufacturing process, using pure silicon is
absolutely necessary. Silicon that is ready for fabricating into chips
need to reach a purity level of 99.9999999%. If even more than one part
in a thousand million is impure, the final yield of working chips is likely
to be too low to be economical. The process that could produce this kind
of purity must be something special; yet the inventor of this purification
method, William Pfann at Bell laboratories didn’t publish his discoveries
until 10 years after his discovery. He thought that it was so simple that
everyone has already knew about it. He made his discovery by noticing that
when silicon ingots were withdrawn slowly from the furnace (so that they
solidified gradually from one end) the impurities tended to accumulate
at the end of the silicon bar that solidified last. This is because the
impurities, faced with an advancing wall of solidifying silicon, tended
in most cases to maintain a higher concentration in the liquid part of
the bar than in the solid/liquid boundary. This process made the initially
uniform distribution of impurities come out of the furnace with most of
them swept to one end. The end containing the higher concentration of impurities
was sawn off and the process repeated. A more improved method known as
zone refining was later devised as it was discovered that it was more efficient
to use a heating coil to move along the length of the bar, rather than
to move and melt the complete silicon bar. By slowly moving the coil along
the length of the bar, a molten zone traversed its length, carrying a useful
proportion of the impurities with it. As soon as the first coil was far
enough away from the starting point, a second coil would follow it, creating
another molten zone. In this way, several sweeps could be carried out simultaneously
and the whole process repeated until there are no further reductions in
the impurity content.
At the end of the zone refining process, the silicon bar is sufficiently
pure. However, at this stage, the atoms in this solid silicon will not
be arranged in the simple periodic structure that is characteristic of
a single crystal. To transform this material into a pure crystal, a process
called crystal growth must be carried out. One option to achieve this is
by converting the solid into a liquid and then to allow this liquid to
solidify by moving it decisively into the correct position in the solid.
This is done by allowing the new solid to form on the surface of a “seed
crystal.” The seed is a small nucleus of crystal that internally is highly
perfect and, thus can nucleate the structure of the new crystal. The seed
crystal is dipped into the melted silicon liquid and then slowly withdrawn.
Using the perfect lattice of the seed crystal as a template, the silicon
in the crucible can be pulled out as a single crystal, which can be 10
cm in diameter and 2 meters long. It is also at this crystal-pulling process
that the silicon is doped by adding either minute quantities of boron or
phosphorus.
The virtually pure silicon crystal bar is then cut into slices, called
wafers, using a diamond-edged circular saw. The wafers are usually 0.5mm
thick and 4~8 inches in diameter. One side of the wafer is then highly
polished. There cannot be any scratches or defects on it. The surface must
be uniformly flat to within one wavelength of visible light. Hundreds of
wafers come from a single crystal ingot, and hundreds of silicon chips
will eventually be produced from each wafer. Add to the fact that a large
chip production plant can produce tens of thousands of wafers every week;
we can understand how it is possible to produce large quantities of silicon
chips at such a low price.
Now then, we are finally ready to make a microelectronic circuit on
the refined silicon chip but first we would have to design the circuit.
The basic conception of the design is supplied to a computer together with
approximate positions and sizes of all the circuit elements involved. Then
the behavior of the circuit is calculated in detail by a circuit-analysis
program, which also tells the designer which parts are going to have the
greatest effect if they are changed. Any part of the layout can be displayed,
magnified several times on the screen. The circuit isn’t a simple flat
pattern on the silicon surface; it will be made out of several layers,
all interconnected. The designed circuit is then printed onto a photographic
plate. The print is produced by scanning a spot of light controlled by
the computer across the plate. From this original copy, an image that is
the correct final size of the circuit is reproduced hundreds of times in
rows and columns to fill an area the size of the silicon wafer. The final
print is the photo mask or stencil for the particular layer of the microcircuit.
Before implementing the circuit onto the wafer, the wafer needs to go through
one more process. The silicon wafer is exposed to extreme heat and gas
to let it grow a thin layer of silicon dioxide, which is one of the best
insulators known, on its surface. It is then coated with a substance called
photo resist. Photo resist becomes soluble when exposed to ultraviolet
light. Finally, in a process called photolithography, ultraviolet light
is then passed through the mask, which defines the patterns of openings
to be opened in the oxide layer on the silicon wafer. The mask protects
parts of the wafer from the light, while areas exposed to the light will
turn into a gooey layer of photo resist. Each layer on the microprocessor
uses a mask with a different pattern. The gooey photo resist is then completely
dissolved by an acid solvent. Further washing in another chemical will
then dissolve the oxide but leave the exposed silicon surface and the hardened
photo resist unaffected. If we were using a boron doped silicon in this
process, then the pattern of the electronic device must involve the introduction
of phosphorus to a controlled depth on selected areas of the wafer. Thus,
the aim of all this is to introduce phosphorus into certain precisely defined
areas on the wafer below the silicon surface, leaving the rest of the wafer
untouched. These areas of bare silicon are positioned using the photo mask.
All that remains is to put the wafer in a furnace, this time with a phosphorus
atmosphere, and let the phosphorus work its way in.
We now have an array of buried microelectronic circuit components in
the silicon wafer, protected by an insulating layer on top and absolutely
untouched by human hands. All that remains is to provide them with metallic
top contacts that can then be used for electrical connections to the outside
world. The preferred metal for top contacts is aluminum, and this is deposited
as a thin film by evaporation. The wafer is placed above a small crucible
of aluminum in a vacuum chamber, and a film of aluminum about one micrometer
thick is deposited. The aluminum can be selectively removed from the wafer,
leaving only a set of contacts on the diodes, using another photo mask
and the same procedure as with the oxide layer. A computer-controlled circuit
tester then electrically probes each chip on the wafer. No attempts are
made to repair circuits that don’t work, since that would be impracticable
and uneconomical. After the initial test, the one that pass are sectioned
into individual chips by marking a grid of scratch lines and then breaking
it. The working chips are mounted in protective packages with thin aluminum
wires joining their contact pads to the pins on the package. Finally, the
complete device is tested again and then finally they are sent to companies
that will use them to produce everyday items.
references:
References:
http://intel.com/education/chips/introduction.htm
http://wiscinfo.wisc.edu/warf.boi/p90019us.html
http://www.smithsonianmag.com/smithsonian/issues00/jan00/intel.html
Silicon chips and you, Renmore, C.D, BEAUFORT BOOKS, INC, New York,
1980
An introduction to microelectronic technology, Morgan, DV, John Wiley
& Sons, New York, 1985