Recently, my wife and I attended a dinner meeting of the
American Association for Crystal Growth at my old stomping
grounds, Bell Labs, now the world headquarters of Lucent
Technologies. Needless to say, there was much gloomy talk
about Lucent”s current precarious financial situation. Fortunately,
the speaker of the evening talked about a lighter subject, literally –
light emitting diodes, a subject dear to my heart. Last week I
mentioned that, prior to a performance of “Carousel” at the Paper
Mill Playhouse, Eddie Bracken warned the audience to turn off
any cellular phones. Our speaker could have used Eddie. Part
way through his talk, a cellphone erupted at our table. The
speaker asked our embarrassed tablemate to show the audience
the color of the display on his phone. It was green.
Coincidentally, the speaker had just finished saying that that the
eye is most sensitive to the color green. That annoying phone call
was turned to the speaker”s advantage.
But, later in the talk, another phone rang persistently. If it didn”t
disturb the speaker, it certainly disturbed me. The cellular phone
situation has truly gotten out of hand. Or do I think so just
because I seem to be the only one who doesn”t have one? And it”s
not going to get any better. The June 25th New York Times had
an article stating that IBM is announcing a new transistor that will
play a key role in future generation cellphones. In fact, IBM
claims this transistor is the world”s fastest “silicon-based”
transistor. I must have missed it, but the newspaper article says
that a few weeks before, Intel also announced the world”s fastest
silicon-based transistor! The article manages to soften any
skepticism about the conflicting statements by saying that the
transistors are two different types with different functions.
Apparently, IBM and Intel can both feel comfortable in their
claims.
What makes a transistor fast? Essentially, the speed depends on
how long it takes for electrons to get from one part of the
transistor to another or, more simply, how long does it take an
electron to get from point A to point B? The shorter the time, the
faster the transistor. One way to make the transistor faster is
simply to make it smaller, that is, make the distance from A to B
shorter. The electron doesn”t have as far to travel and gets there
quicker. Transistors have indeed been getting smaller and
smaller, and faster and faster, following the often-quoted Moore”s
Law. But they”re getting so small now that this approach is not
going to work forever. Excuse me while I find out what Intel
does to make its ”fastest transistor”.
I”m back, having just returned from the Intel Web site. Intel, co-
founded by Gordon Moore, naturally wants to keep Moore”s Law
going forever and, by golly, they seem to be doing it! Workers at
Intel have shrunk the size down to as small as 20 nanometers with
certain layers only 3 atoms thick. You can hardly go much
smaller than that. Moore”s Law targets 30 percent smaller
transistors, twice as many transistors per wafer and half the cost
per transistor, all taking place every two years. This has been
going on since before 1970 and they expect it to continue for at
least the rest of this decade. Today”s fastest Pentium processors
have about 40 million transistors. If the new speedster goes into
production in 2007 as planned, the Moore”s Law projection is 1
billion transistors on the chip!
What does Intel envision this will mean for you? Your computer
will be smarter, understand your spoken commands or read your
handwriting, monitor how you work and meet your needs to get
the job done faster. An example of the latter suggests that you
will be able to do your Christmas shopping in five minutes. You
tell the computer what you want, maybe even show it pictures,
and it”ll go online and order everything. I would need more help
than that. I would want my computer to tell me what my wife
really would appreciate so I knew what to order. I”m sure that will
be possible with a billion transistors on the chip. The Intel
transistor is rated at over a trillion cycles per second and the
projected frequency of the processor containing these transistors
in 2007 is 20 billion cycles per second (20 gigahertz).
What about the IBM fastest transistor? The Times article says
that the IBM transistor is based on an “exotic” type of silicon
manufacturing process, silicon germanium. Darn! I must have
missed the boat again. Back at Bell Labs back in 1955, I co-
authored an article on silicon-germanium. Nine years later, I co-
authored another one. At that time, we certainly didn”t consider
silicon-germanium to be ”exotic”. In fact, the germanium-silicon
system is a rather dull system in some respects.
Silicon and germanium form what we in the trade call a
“continuous series of solid solutions”. This is a slightly fancy
way of saying that you can make alloys of germanium and silicon
in any proportion. Put another way, if you start with a crystal of
silicon and replace the silicon with germanium atom by atom until
you have pure germanium the crystal structure doesn”t change.
(It”s like mixing Scotch and water – they mix in all proportions.)
The only thing that happens is that as you add the germanium the
distance between the atoms in the crystal gets larger, by four
percent when you get to pure germanium. Otherwise, nothing
interesting happens. Of course, that”s not really true. Scientists
can always find something interesting in just about anything and
then write a paper about it. I mentioned our own two papers!
While the chemist or metallurgist may think the germanium-
silicon system is pretty tame, the electrical engineer sees
something interesting. Remember we said that we can make our
transistors faster by making them smaller, by shrinking the
distance from A to B. There is another way. Suppose you make
your transistor out of a material in which the electrons are
inherently speedier. There is such a material – gallium arsenide, a
compound of gallium and arsenic. In a gallium arsenide transistor
the electrons move a lot faster than in silicon. So, to get a fast
transistor you don”t have to make the transistor as small. In fact,
gallium arsenide is used to make high frequency transistors for
certain applications.
But gallium arsenide is not cheap. A Japanese source quoted on
the Cavendish Laboratory Web site compares the cost of a square
millimeter of gallium arsenide with the cost of silicon in the form
found in computer memory chips. The site quotes the cost of
Tokyo real estate and the silicon both as only a penny a square
millimeter. I haven”t done the math but I”m sure the cost per acre
will be appreciable! But the site quotes the cost of gallium
arsenide at $2.00 a square millimeter, 200 times more expensive!
Suppose we can find a less costly material that still has speedy
electrons. Germanium is such a material. The electrons are a
good bit speedier than in silicon but not as fast as in gallium
arsenide. However, we only have to pay 60 cents for a square
millimeter versus the $2 for the gallium arsenide. So, let”s add
some germanium to the silicon, the approach used by IBM. Hey,
the electrons are speeding up, not as much as in gallium arsenide
but still speedier than in silicon itself. IBM”s transistor is
rated at 210 gigahertz, 200 billion cycles per second. My faithful
old Dell computer has a 266-megahertz processor, 750 times slower.
What about that ”exotic” technology? Actually, it”s not that easy to
grow silicon-germanium material on silicon. Recall I said that the
distance between the atoms is 4 percent larger in germanium than
in silicon. This difference doesn”t sound like much but if there”s
enough germanium in the silicon-germanium material, the
mismatch can cause a lot of strain between the silicon and the
silicon-germanium transistor. This strain can lead to cracking, not
a good thing. The Times article said that the IBM transistor is
only 100 atoms thick. By sticking to very thin layers, it”s possible
to grow materials that are strained but not enough to cause
cracking. The strain can even result in beneficial electrical
properties. Without going into detail, silicon-germanium happens
to be in this category. So, today”s silicon-germanium is indeed
much more exotic than we ever imagined over four decades ago!
Oh, back to our speaker at the Bell Labs. During the question and
answer period, the same muffled ringing popped up again and
nobody answered it. By this time, the speaker was justifiably
perturbed and asked in an irritated manner, “Whose phone is
that?” You guessed it. Someone said, “I think it”s yours!” Sure
enough the muffled phone emerged from the speaker”s briefcase!
With these superfast chips on the way, the cellphones of the future
will no doubt be loaded with even more features. I propose that a
very desirable feature would be that a cellphone would
automatically sense when the owner is (a) in a theater or a
meeting, (b) in a restaurant or anywhere else where there are
people who are not thirsting to hear conversations other than their
own or (c) in a moving vehicle in a location where the driver can
pick it up. In any of the above situations, the phone would
automatically shut itself off and become inoperable! One can
only dream.
Allen F. Bortrum
