All you need are a kiln, some hydrofluoric acid, and a little bit of luck

By Aalyia Shaukat,
contributing writer

Do-it-yourself transistors have
been generated from a few hobbyists online including NeilSteiner and some guy named Jim back in 2009. The most notable
homebrew transistors, however, come from Jeri Ellsworth, a tinkerer with a knack for
device physics. She built a home-based fabrication facility (also known as “fab”)
in 2010 and managed to generate some simple circuits with her homemade chips.
Interestingly enough, no other well-documented DIY transistors have since been
published. This is likely due to the fact that this type of manufacture
necessarily leads to the handling of dangerous chemicals and can take years to perfect
(it took Jeri two years). Moreover, because transistors are the building block
for electronics, they can be purchased cheaply to create much more complex ICs
that perform something that can be more directly functional. With that being
said, it is a pretty substantial achievement for a hobbyist to build a
transistor from scratch with a home laboratory, so one would imagine there is a
personal sense of accomplishment that comes with making one.

Nowadays, most fabless firms
outsource the actual production overseas because the overhead cost of employing
the required number of engineers and technicians to operate the plant at a
profit is too high following the $10 billion invest that it typically costs to
construct a state-of-the-art fabless plant. Most companies cannot sustain the
level of sales needed to justify that kind of financial investment.

Being that there’s no other option
to generate transistors cheaply, it makes sense why these foundries are making
a killing with a new record of $400 billion in sales. In essence, it might
actually make sense to explore more accessible modular transistor fabrication,
particularly for prototyping, despite the major cost and technological hurdles because
the gigantic facilities do have some drawbacks.

Jeri Ellsworth and Neil Steiner
have both created transistors from prefabricated germanium diodes and cadmium
sulfide photocells, respectively. This is not the same as building a transistor
from just bare doped wafer because the diode and photocell are prefabricated.

Building a transistor from scratch
would require some tools that are not commonly found in a home-based lab. This
video from San Zeloof is a pretty thorough tour of his
home chip fab and is reminiscent of Jeri’s
lab. Some more uncommon tools include
the following:

 

Some of these materials are more
harmful than others. Anyone handling HF must do so very carefully because this
chemical can penetrate tissue, causing some pretty gnarly burns. A nitrogen
tank is not completely necessary, but it helps in controlling the atmosphere of
the kiln to more predictably grow the oxide layer on silicon, and when you’re
growing layers that are only several hundred angstroms (Å) in thickness, it will probably
save a lot of time. Also worth noting, doping pure silicon is another project
in and of itself, so purchasing predoped prime-grade silicon online might be
more feasible. Some more mundane hobbyist equipment includes a power supply,
oscilloscope, tweezers, CPU fan, conductive epoxy, and solder.

How are transistors
fabricated?
A transistor is fabricated through
the process of photolithography, which, in essence, patterns the
surface of a substrate to form various transistor topologies. One such topology
is that of a metal-oxide field-effect transistor (MOSFET), as shown below.

Image source: Shutterstock.

Once again, the bulk p-type
substrate can be readily bought online.

Growing the initial oxide layer is
accomplished through time in the kiln. Jeri, for instance, states that it takes
her six hours to grow a 500- to 600-Å-thick layer of oxide with the
addition of steam pumped into the furnace. There is no need to pull out a super
tiny caliper and microscope to measure this thickness as it is apparent by the
color. A 600-Å-thick layer corresponds to a green color.

Etching, or removing the oxide to accomplish a particular pattern, is
accomplished through the use of HF. For many etch steps, part of the wafer is
protected from the etchant by a “masking” material, which resists
etching. The vinyl sticker mask is precut to resist the etchant in certain
areas.

The source and drain region is created by spinning the phosphosilicate film onto the
wafer so that there is a thin layer of the liquid on the wafer piece; this is
normally accomplished with a CPU fan.

Once again, the device is placed
in the kiln at 1,000oC for a certain amount of time to deposit high
concentrations of phosphorus onto the surface of the substrate and oxide layers
— creating two doped n-type regions and, ultimately, a channel for electrons to
flow.

Finally, vinyl masks are used to
etch away the gate region and place it back into the kiln to grow a gate oxide
layer. Jeri specifies a thickness of 800 to 1,000 Å, which is a pink to dark red
color.

Contacts to the gate, source, drain, and substrate can be made with
conductive epoxy. This is accomplished with another vinyl mask and additional
etching to make access points down to the source and drain to place bits of
epoxy that can be soldered to with leads.

More details are listed on a number of different websites
with just a little bit of research. It is apparent that this process gets
exponentially more complicated when designing masks and planning on the steps
for a topology with multiple transistors. The size and scope of the task can
get unwieldy; thus, the design of huge fabrication facilities in which
transistor gate widths are down to nanometers and CAD programs for the layout
of millions of transistors.