Ultimate Alchemy


Research into artificial atoms could lead to one startling endpoint:
programmable matter that changes its makeup at the flip of a switch.

By Wil McCarthy

The hardest thing you can ask them is how old they are. The question seems
to rock them back, to give them pause. "I guess I'm 38," one of them tells
me uncertainly. "I must be 54," another answers, after even longer
deliberation. It's not that these men are slow, it's that they're
physicists. And they're involved in a new research area as promising as it
is strange, so if they seem a little distracted, well, c'est la vie. Despite
modesty and caution so deeply ingrained that it might well be genetic, they
also project an air of barely contained excitement. They're building the
future, and they know it.

To understand how and why, it helps to be at least passingly familiar with
the areas of nanotechnology and micro-electromechanical systems. Both offer
great promise in creating forms of matter that are substantially more
capable than anything nature has to offer. But nanotechnology faces inherent
limitations in communications, power, and materials distribution, while MEMS
have short operating lifetimes and will always have trouble creating precise
micro- and nanostructures other than simple chemicals.

There will of course be myriad applications, some of them quite amazing, but
by the time they find their way into the real world, both nanotech and MEMS
may wind up looking less magical than even a humble television screen, which
after all can change its appearance instantly and completely. But there may
be a truly programmable substance in our future that is capable of changing
its apparent physical and chemical properties as easily as a TV screen
changes color. Lead to gold, baby, on demand. Call it programmable matter.

The keys to this phenomenon are found in a mysterious realm called the
mesoscale. At the nanoscale (10-9 meters) we find very tiny, very simple
objects, like the water molecule, that possess a high degree of symmetry and
can be modeled precisely using the closed-form equations of quantum field
theory. The behavior of a simple molecule can always be predicted. At the
microscale (10-6 meters), where we find larger objects like micron-wide
blood cells or silicon gears, these quantum effects blur and average across
billions of particles into the familiar smear we call classical physics. You
know: mechanics, electricity, stuff like that. Again, easily modeled and
predicted, because the individual particles can be ignored, and their
aggregate behavior is completely predictable.

But these scales are separated by three orders of magnitude, and in between
them lies the mesoscale (from the Greek mesos, or middle), an area largely
uncharted in both theory and experiment. The study of mesoscale effects is a
key aspect of condensed-matter physics, which Britannica defines as "the
study of the thermal, elastic, electrical, magnetic, and optical properties
of solid and liquid substances."

It is here, in this scientific hinterland, that we find Marc Kastner, head
of the physics department at MIT; Charles Marcus, physics professor with
Harvard's Center for Imaging in Mesoscale Structures; and Howard Davidson,
distinguished engineer and "quantum mechanic" at Sun Microsystems.

Their specialty: mesoscopic semiconductor structures with bizarre new
properties.




Tuning Up the Semiconductor


Most materials are either conductors, which permit the free flow of
electrons, or insulators, which resist it. Semiconductors are insulators
capable of conducting electrons within a certain narrow energy band - a
useful trick that makes integrated circuits and other electronics possible.

The electrical properties of a semiconductor like silicon are fixed by the
laws of physics, but in a process known as doping, very small and very
precise amounts of another material can be scattered through the crystal
lattice. Often, this doping is controlled almost to the level of individual
atoms, and typically about one dopant atom is added per million atoms of
substrate. This tiny impurity can wreak large changes in a semiconductor's
behavior, so that, for example, room-temperature electrons have a good
chance of jumping up into the conduction band when voltage is applied.

Silicon doped with electron donor atoms such as phosphorus becomes an N, or
negative-type, semiconductor, which contains one excess electron for every
atom of dopant. These excess electrons have nowhere to bind, so they flow
easily through the material, just as the more numerous excess electrons do
inside a conductive metal.

Doping with electron borrowers like aluminum produces a P, or positive
material, which conducts "holes," or spaces where an electron isn't. It
seems counterintuitive, but electron holes can be manipulated and moved
around as though they were positively charged particles. The analogy is that
little puzzle where you slide the squares around to unscramble a picture or
a sequence of numbers - you rearrange the puzzle by moving the hole where
you want it. Anyway, with P-type silicon you get one extra hole per atom of
dopant, meaning that a small, precise number of excess electrons can be
absorbed by the material, sharply inhibiting their free flow.

This may sound rather abstract, but it's a trillion-dollar factoid: A P
layer adjacent to an N layer creates a structure known as a P-N junction,
which is a kind of electrical valve or gate that permits electrons to flow
easily in one direction but not the other. This effect is critical in
electronic components such as diodes, LEDs, rectifiers, and transistors. In
fact, the latter half of the 20th century was built almost entirely on P-N
junctions; without them, we would not have the compact computers and
communications devices that have made numerous advances possible.

And since the late '80s, another application for P-N junctions has been
discovered that may, in the end, prove even more revolutionary. When an N
layer is sandwiched between two Ps, a kind of trap is created that attracts
electrons into the middle layer and doesn't let them out. This is a useful
trait all by itself and leads to a couple of exotic variants on the standard
transistor. But if the N layer is really thin - about 10 nanometers (equal
to 0.00001 millimeters or 50 atoms) - then something weird starts to happen.
The size of the trap approaches a quantum-mechanical transition point: the
de Broglie wavelength of a room-temperature electron.

The result? Along the vertical axis of the trap, the excess electrons can no
longer move and propagate in the Newtonian way. Their positions and
velocities take on an uncertain, probabilistic nature. They become waves
rather than particles.

These P-N-P devices, known as quantum wells, are easy and cheap to produce
and have the interesting property of generating photons of very precise
wavelength, which means they can be used to make laser beams. Quantum wells
find practical use in computers, fiber-optic networks, and those cute little
$7 laser pointers you can buy for your keychain.

Like the meat inside a sandwich, a quantum well's electrons are confined in
a two-dimensional layer. But if the meat and the top bread layer are cut
away on two sides, leaving a narrow stripe of P-N sandwich sitting on a
slice of P bread, the electrons take on wavelike behavior along an
additional axis. This structure is called a quantum wire and is used to
produce intense laser beams that can be switched on and off much more
rapidly than quantum well lasers can - up to 10 gigahertz, or 10 billion
times per second. Quantum wires can also be used as precision waveguides,
and of course as actual wires.

  The electrons trapped in a quantum dot arrange themselves as if they were
part of an atom. With a big difference: This particle has no nucleus.

But quantum wires are only a stepping-stone here. They lead us to a final
configuration: etching away the ends of the stripe to leave a tiny square of
meat and bread atop the lower layer, producing a "quantum dot" that confines
the electrons in all three dimensions. Unable to flow, unable to move as
particles or even hold a well-defined position, the trapped electrons must
instead behave as de Broglie standing waves, or probability-density
functions, or strangely shaped clouds of diffuse electric charge. Strangely
shaped because, even as waves, the negatively charged electrons will repel
one another and attempt to get as far apart as their energies and geometries
permit.

If this sounds familiar, it's because there's another, more familiar place
where electrons behave this way: in atoms. Electrons that are part of an
atom will arrange themselves into orbitals, which constrain and define their
positions around the positively charged nucleus. These orbitals, and the
electrons that partially or completely fill them, are what determine the
chemical properties of an atom - such as what other sorts of atoms it can
react with, and how strongly.

This point bears repeating: The electrons trapped in a quantum dot will
arrange themselves as though they were part of an atom, even though there's
no atomic nucleus for them to surround. Which atom they resemble depends on
the number of excess electrons trapped inside. What's more, the electrons in
two adjacent quantum dots will interact just as they would in two real atoms
placed at the equivalent distance, meaning the two dots can share electrons
between them - they can form connections equivalent to chemical bonds. Not
virtual or simulated bonds, but real ones.

Amazing, right? If you're not amazed, go back and read the last three
paragraphs again.

Now we'll take it a step further: Quantum dots needn't be formed by etching
blocks out of a quantum well. Instead, the electrons can be confined
electrostatically by electrodes whose voltage can be varied on demand, like
a miniature electric fence around a corral. In fact, this is the preferred
method, since it permits the dots' characteristics to be adjusted without
any physical modification of the underlying material. We can pump electrons
in and out simply by varying the voltage on the fence.

This type of nanostructure is called an artificial or designer atom, because
it can be manipulated to resemble any atom on the periodic table. It's not a
science-fictional device, but a routine piece of experimental hardware used
in laboratories throughout the world.




Building Atoms


Where do artificial atoms come from? The poetic answer is that they arise
from the sweat and dreams of human beings such as Marc Kastner and Charlie
Marcus, and the dozens of eager grads and undergrads and postdocs and
visiting fellows who work for and with them. There are a few dozen labs
worldwide engaged in this research, and the atmosphere among them is
laid-back and clubby. This is basic research, geared toward methodical
discovery rather than near-term commercial payoff. Collaborations and data
sharing are the norm. Marcus, one of the youngest and most energetic of the
lot, is always in motion, always on an errand, always engaged in jovial
conversation with someone, with everyone. Laughing about it, he tells me,
"Why bother hogging glory when there are so many hard problems still to be
solved? The greatest threat out here is dying of loneliness."

The less poetic answer is that artificial atoms usually come from the same
sort of semiconductor laboratories that produce exotic computer chips. This
isn't surprising, since these labs are already expert at creating precision
nanostructures, and since shrinking electronics are pushing their way into
the mesoscale anyway. The marriage of the two fields is a natural: easy to
explain, to fund, to justify.

For the price of coffee, any condensed-matter physicist will happily explain
the future of computing. Quantum dots can serve as really, really small
transistors, whose logic is many times more efficient than that of today's
transistors and operates on single electrons. They can also serve as
cellular automata - spreadsheet-like systems in which each "cell" contains a
formula that defines its state as a function of the states of neighboring
cells (a great way to simulate weather or fluid mechanics). In addition,
scientists have shown that an array of quantum dots passing around excess
electrons is, mathematically speaking, a kind of neural network. Quantum-dot
"neurons" may one day display some of the same traits as biological ones,
despite being orders of magnitude smaller.

Scientists are most interested, though, in quantum computing applications,
where the potential performance gains are nearly unimaginable. (See "Liquid
Logic," Wired 9.09, page 152.) Future quantum dot processors may well
combine all the best features of digital, analog, and quantum processing. In
principle, the same hardware could be used for any and all of these
operations, and might even switch modes dynamically, depending on the
problem being solved. In the face of such technology, words like
"supercomputer" simply fail. But computer chips - even blindingly superior
ones - have all but lost their power to amaze us. Fortunately, the promise
of artificial atoms goes way beyond mere computing.




The Chemistry Set


It turns out that you can have a quantum dot without a P-N-P junction, as
long as you have some other means to confine the electrons. Researchers have
achieved excellent results with simple geometries, such as tiny beads of
semiconductor with a thin, insulative coating. Amazingly, this means that
quantum dots can also be grown in a chemistry lab. It takes about 15 minutes
to heat and mix the appropriate chemicals, and then minutes or hours for the
resulting nanocrystals to grow to the desired size, after which they can be
deposited onto a chip, where they will line up as neatly as oranges in a
box. MIT chemist Moungi Bawendi uses this surprisingly precise technique to
make artificial solids with interesting electrical and optical properties.

These chemically produced dots unfortunately don't have tiny electrodes
running up to them, so they can't be controlled individually as artificial
atoms. But by running a voltage through the semiconductor substrate, we can
drive electrons into and out of all the dots simultaneously. Because of
slight irregularities in size, shape, and number of dopant atoms, there's no
guarantee that all these artificial atoms are behaving in precisely the same
way, but the average number of excess electrons can be controlled, and the
resulting properties measured. Bawendi has grown very large solids: tens of
centimeters across and tens of microns thick. Held together only by surface
tension forces, these solids are waxy in texture and easily scratched or
broken.

  Programmable Matter (n): any physical substance whose properties can be
adjusted precisely and repeatably through electrical or optical stimulation.

Unfortunately, at the lower energy states, with few electrons aboard the
dots, these materials are also highly insulative, which makes it nearly
impossible to drive additional current through them unless the electrodes
providing the voltage are very close together. Typically, interesting
effects are observed only in samples that are at most a few microns (or
thousands of nanometers) across, and with just a few layers of dots on top
of the semiconductor. We're talking about specks, really. Nevertheless, they
give us our first look at what programmable substances are like, and what
they may be capable of.

One of the properties of Bawendi's artificial matter is fluorescence: very
precise fluorescence at narrow frequency ranges - something that does not
occur in nature. Just turn the voltage source on, and away it glows. These
solids can also be excited optically rather than electrically - they have
the fascinating ability to drink in light of almost any wavelength and spit
it back out in a nearly monochromatic stream. The light source can be of
almost any type: white, colored, laser, ultraviolet - what comes out is a
single, bright color determined by the exact characteristics of the quantum
dots. These crystals also reflect, refract, and absorb light in intriguing -
and electrically variable - ways. Other properties that have been
investigated include the photoelectric effect (as in solar panels),
photodarkening (as in light-sensitive sunglasses), photoconductivity, and
electrodarkening. In related work, researchers in London have created an
artificial solid that is normally opaque but turns transparent when excited
by a laser.

These substances are not, to say the least, ordinary semiconductors. As the
fatherly Kastner explains in patient yet unmistakably enthusiastic tones,
the properties of an artificial solid can be completely different from those
of the semiconductor it's made from. "Gold" has been created from a grid of
cadmium selenide, although the substrate still has an effect on the final
properties. Not real gold, then, but some closely related and decidedly
improbable material: gold-cadmium-selenide, straight from the monkish,
13th-century dreams of alchemist Roger Bacon.




Not Natural


Artificial atoms have been a boon to quantum physicists because they provide
a new window into natural atoms. Because their structure does not rely on
electrons' attraction to a positively charged nucleus, but on electrostatic
repulsion and the geometry of P-N-P junctions, artificial atoms are
generally at least 50 times larger than natural ones and physically overlap
with at least 50 by 50 by 50 of the atoms in the semiconductor substrate.
That's an important difference.

Also, notably, the orbital shapes of natural atoms - spherical, dumbbell,
pinwheel, tetrahedral, et cetera - arise from the purely spherical symmetry
of the nucleus' electric field. Artificial atoms do not share this
characteristic, and in fact many of them are "pancaked," foreshortened along
one dimension. And, depending on the shape and voltage of the repulsive
fence, artificial atoms can have square or triangular or any other sort of
symmetry. Multipart fences with different voltages on each section can even
lead to asymmetrical atoms.

Another difference of artificial atoms is that the energy of the electrons
is a direct function of the physical size of the quantum dot they reside in.
Bigger atoms mean lower energies and blurrier distinctions between energy
levels. This is bad, because the randomizing effects of "thermal noise" can
destroy any atomlike behavior in the system. For this reason, today's large
quantum dots - some of them a thousand nanometers or more in diameter - show
interesting properties only at liquid helium temperatures: 3 degrees Kelvin
and below. Smaller quantum dots are highly desirable, because they have
higher energy and sharper definition. The closer you get to the size of a
natural atom, the better you can mimic its behavior. The ultimate goal, our
scientists agree, is to construct artificial atoms that function at room
temperature and above.

To put it another way, the chemical/optical/electrical behavior of real
atoms is defined by the outermost layer of electrons. In the words of Sun's
Howard Davidson, "Everyday life happens in the outer shell, between a half
and about 2 electron-volts." An electron-volt is a measure of energy
equivalent to 1.6 x 1019 joules, and according to Marcus, this implies an
atomic size in the sub-nanometer range. This doesn't mean our electric fence
has to be that small, though: The electrons, repelled from the fence, tend
to crowd toward the center of their enclosure. The artificial atom they form
is significantly smaller than the fence itself. Anyway, small is good, small
is versatile, and in fact, by erecting concentric fences atop a quantum
well, we should be able to create dots with ever more tunable sizes and
energies, which would be more versatile still.

This arrangement - an artificial atom of variable energy and shape - could
be further enhanced by placing a second, identical set of electrodes on the
quantum well's lower face, probably with an insulating layer below that. The
atom's characteristics would then be fully adjustable in three dimensions.
This is important because "pancake" elements, lacking one whole dimension to
cram electrons into, have a much simpler periodic table than the one
Mendeleyev derived for natural atoms in the late 19th century. If you want
the full richness of nature - and more! - you really want that third
dimension.

Even more complex arrangements might be possible, but remember these
structures are so small - a few thousand atoms at most - that we don't have
the broad flexibility we do at larger scales. It's like building with
Legos - the smaller you get, the more restricted your design choices.

A final and critical difference between natural and artificial atoms is that
nuclear forces limit the number of protons in a reliable atomic nucleus to
92, the atomic number of Uranium. Unfortunately, since protons and electrons
are paired, this means that atoms containing more than 92 electrons have
short half-lives - that is, they're radioactive and therefore annoyingly
difficult to use for anything other than bomb and power-plant fuel. This is
where the real killer app of artificial atoms comes in: Since they're not
burdened with a nucleus, they can remain stable with hundreds or even
thousands of electrons crammed inside, forming gigantic new orbitals
classical chemists could never have imagined.

The remarkable properties of these new orbitals have been studied by Paul L.
McEuen of Cornell, and also by Marcus, Kastner, and another MIT scientist
named Raymond C. Ashoori. Their finding: much weirdness. In natural atoms,
electrons move in fixed orbits based on their energy and spin, and while
this is also true of artificial atoms, the orbits can be quite different
even for an atom with very few electrons. Some orbitals exhibit electron
bunching, a phenomenon that permits one or more electrons to slip in and out
of the atom with zero energy cost. Ashoori has found that this occurs when
the artificial atom's electron cloud splits into two pieces: an inner circle
and an outer ring - something that would never happen in a natural atom.

  Stable natural elements are limited to 92 electron states. Artificial
atoms can have hundreds, even thousands, making today's periodic table look
puny.

Even where this bunching doesn't occur, recent progress in single-electron
imaging has revealed the structure of artificial orbitals, some of which
turn out to be very thready and discrete, more like billiard ball
trajectories than clouds. The images are spectacular, and the physicists,
being hackers of a sort, are nearly as proud of their detection methods as
they are of the actual results. In discussion, the two are so deeply
intertwined that they might as well be one thing. Perhaps, for the folks
down here in the trenches of discovery, they are one thing.

Another finding these guys are mulling over is that larger orbitals are
influenced by magnetic fields much more than small ones in a natural atom
are. Effects can be easily produced in quantum dots that would appear in
atoms only at field strengths of a million tesla or more - about 10,000
times stronger than we can presently create. And in larger quantum dots -
particularly asymmetric ones with large numbers of electrons - the shell
structure breaks down altogether, yielding electron-gas structures with
bizarre magnetic properties. Indeed, switchable exotic magnets may be an
important application for artificial atoms.

Clearly, artificial atoms - while remarkably similar to natural ones - are
capable of exhibiting a wide range of structures, characteristics, and
behaviors that do not occur in nature. In fact, the 92 natural structures
are tiny and by no means preferred islands in the sea of this technology's
overall capability. This we now know: Doping a semiconductor with artificial
atoms can modify its physical, optical, and electrical properties in
decidedly unnatural ways, with decidedly unnatural results.

And keep in mind that the artificial atoms don't just sit there; like
cellular automata, they interact with their neighbors. We can not only form
chemical bonds between them, but also turn the bonds on and off as electrons
are pumped in and out. This is virtual chemistry, or pseudochemistry, or
artificial chemistry, but chemistry nonetheless. Nano- and microtechnology
promise to rearrange the shape and texture of materials, which is great but
not really so different from what we can already achieve manually, with a
machine shop or even a simple potter's wheel. Nanotechnology may even be
able to rearrange atoms, albeit slowly, like a sort of mechanized plant or
fungus. But here is something entirely new: a material capable of changing
its very substance, instantaneously.




This Way to the Future


Up to this point, I've been coloring mainly inside the lines of existing
research, not straying too far into future applications. This reflects the
prejudices of the researchers themselves: Despite the enormous potential of
this field, no respectable scientist has stepped into the media limelight to
discuss its future goals and benefits, or even its state of the art.
Cornell's Paul McEuen, one of the more outspoken figures in the field, went
on record in Science magazine, saying, "The next step is to assemble these
atoms into artificial molecules ... and new kinds of solids - ones that
could not be realized with real atoms." The comment was subsequently picked
up by Science News, with its 1.2 million readers, making the quote perhaps
the most public prediction any scientist has yet made about programmable
matter.

That tantalizing turn of phrase, though, is one that crops up constantly in
the field's own literature. It's almost a direct quote from Marc Kastner, in
a 1993 Physics Today paper, in which he incidentally coined the term
"artificial atom." What's interesting is that this prediction is generally
found in the summations and dismissals at the end of technical papers, and
has not visibly evolved in the nine years since it first appeared. When
pressed on this point, university scientists grow taciturn, or talk airily
about the differences between basic and applied research, or science and
engineering. "Scientists," Charlie Marcus observes dryly, "don't necessarily
make better futurists than the people down at the doughnut shop. So much
advancement depends on serendipity - the laser came out of microwave
research, not a desire by Schawlow and Townes to improve surgery or record
players."

Kastner is more direct: "The nation is not willing any longer to invest in
basic research in the physical sciences, and I would argue that
condensed-matter physics has been treated worst of all. This makes us
nervous about making claims that could be attacked, even if they are right."

When they've warmed up a bit, though, Kastner and Marcus do eventually start
dropping veiled hints about switchable ferromagnets and high-temperature
superconductors, about quantum pumps and filters and repeaters. The focus is
microscopically tight, on devices rather than systems. I want to hear what
the walls will look like in a world of programmable matter, but belaboring
this point is something of a disservice: There's a bright future ahead of
us, but making promises about it is not in their job description. "It's easy
to raise expectations," Ashoori warns.

For better or worse, things are a bit looser in the world of industry; when
I sat down with Joel Martin, founder and chair of the Quantum Dot
Corporation, the future was our main topic of conversation. A pre-IPO
startup in the heart of Silicon Valley, QDC takes Bawendi's chemically
synthesized quantum dots into the realm of bioscience. Where all the money
is, remember?

Interestingly, these Qdot particles are not mounted on any substrate at all,
but simply float around as ultra-tiny particles in a biological solution.
They glow furiously under any light source and are bright enough to show up
as winking, monochromatic pinpoints under an ordinary microscope. And when
chemical receptors are attached to them, they seek out and illuminate
individual molecules, allowing biologists to observe the chemical processes
going on inside a cell, in vivo, while they're actually happening. With
different types of molecules, each tagged with its own color, the cell
becomes a kind of road map, alive with twinkling traffic.

  Funding for condensed-matter physics has tanked. "This makes us nervous
about making claims that could be attacked, even if they are right."

"It takes the hubris of entrepreneurs to start a company," Martin opines.
"You don't have the shoulders of giants to stand on here; you have to make
all your own mistakes. Knowing a little about the science doesn't tell you
anything about the problems of manufacturing - you have to get in there in a
very Edisonian way and just try all the combinations. But in doing that, you
create a huge pool of intellectual property and know-how, which can be used
in the future."

About the materials-science applications for quantum dots, the 44-year-old
Martin - a PhD chemist as well as an MBA venture capitalist - waxes
enthusiastic. "I'm actually working on one right now. The real applications
will be in things that you're familiar with today: detectors, fiber-optic
repeaters, paints and coatings, security inks - making things better, not
making these nanorobots that bumble around inside you. This nanoscale
technology gives us another degree of control over the properties of matter,
of surfaces, that we don't have by other means."

Alas, too much money can shut people up as surely as too little; Martin is a
Valley insider who has participated in nine startups, two of them his own.
His board of directors is crowded with other venture capitalists, who eat
nondisclosure agreements for breakfast and expect a return on their $40
million investment. I don't bother asking what he's working on - if it pans
out, we'll know soon enough. But when I ask his opinion on whether prospects
are good for programmable matter in the abstract, he has a one-word answer:
"Absolutely."

As a physicist on Sun's staff, Howard Davidson is free to take things a
couple of steps further. He points out that the frequency of a laser depends
on the gain frequency of the material generating it - an electrical
property - and the resonant frequency of the mirrored chamber in which it
builds up - an optical property related to the index of refraction. So with
quantum dot materials you could achieve one of the holy grails of the
photonics industry: a tunable laser.

Davidson also points out that our nanoscale fabrication techniques have
already produced materials that are roughly a million times more sensitive
to magnetic fields than are any naturally occurring substance. IBM, for
example, uses these in the read heads of disk drives, he says, and "the
ability to switch that property on at will is of great technological
interest." And on the subject of superconductors: "There is no good theory
here. We still don't know how the high-temperature ones work." But with much
better computers and more versatile materials, he reasons, we should
eventually be able to plumb these mysteries, and begin designing
high-temperature superconductors rather than discovering them randomly.

So all in all, it's looking damned good for programmable matter. But when it
comes to describing its future concretely, even Davidson seems to draw
something of a blank. That sort of gonzo extrapolation is, properly, the job
of science fiction writers. Since I'm one of these, I'll cautiously suggest
the next couple of steps myself.




Follow Me


To date, I'm not aware of any experiment that has placed more than a handful
of individually programmable quantum dots together. This is primarily
because our manufacturing techniques aren't up to the task, but it isn't
difficult to imagine, say, a 1-millimeter-square microchip covered with a
grid of artificial atoms. The properties of Moungi Bawendi's "2-D artificial
solids" can be adjusted only en masse, but this designer chip would be the
real McCoy: a slice of matter that is fully programmable at the atomic
level.

Now, the spacing of the quantum dots on our chip is problematic, since they
need to be close enough to interact, but not so close that the electrodes of
one will have a major disruptive effect on the contents of its neighbors.
For argument's sake, let's say the whole thing - electrodes, confinement
space, and safety margins around the outside - is a square 20 nanometers
(about 100 silicon atoms) on a side. This size should permit the sort of
room-temperature and visible-light interactions we're interested in. This
also means the 1-mm2 chip will hold 50,000 rows of 50,000 dots each, or 2.5
billion artificial atoms.

Let's further suggest that, for maximum flexibility, each quantum dot is
controlled by 16 electrodes with independent voltage sources. This means 16
separate conductor traces feeding into the chip for each of our several
billion dots. That's a lot of wires, and a lot of independent voltage
sources. Impractical? An obvious simplification is to break the grid up into
smaller "tiles," say groups of 8 by 8, or 64, quantum dots. Each dot on a
tile would be controlled independently of the others, but each tile of 64
would behave the same as every other tile. The 16 voltages controlling any
given dot are also passed along to the same location in the neighboring
tiles. Thus, only 1,024 different voltages (16 by 64) are needed to control
a tile-floor chip of arbitrary size.

This may sound like a limitation, but if each electrode can be set, for
example, to 256 discrete voltages, each designer atom will have 25616, or
3.4 x 1038 possible states. Compared with the 92 states of the periodic
table, this is a staggering number, and if we place three designer atoms
together, the number climbs to 10 quadrillion googol, or 1.02 x 10115 -
higher than most calculators can count. So an 8 by 8 grid - more than 21
times as large - represents an absurd and downright spooky wealth in
undreamed-of materials. Finding needles in that cosmic-scale haystack will
be the work of lifetimes.

Controlling the chip itself, however, is relatively easy: With only 1,024
signals to worry about, our only problem is splitting and routing these to
the individual quantum dots.

Laying a few hundred of these chips side by side will result in exactly what
I promised earlier: a TV screen that changes substance as easily as it
changes color. With minuscule power consumption, it could easily switch from
lead to gold and back again, many times a second. And since it isn't limited
to the 92 natural elements, it would be capable of taking on characteristics
that natural substances can't. It's a reasonable bet that there'll be better
superconductors than today's yttrium barium copper oxides, better reflectors
than the mercury and silver we use today, better photoelectric converters
than silicon. In fiction, I've even posited the existence of programmable
matter "superreflectors" and "superabsorbers," which process light in a
given frequency band with 100 percent efficiency.

  They won't change mass or shape. But with the flick of a bit, artificial
atoms will change from one miraculous pseudosubstance to another.

Really, such chips would be capable of doing and being so many things that
it's easier to start from the other end and list their limitations. They
can't change their mass. They can't change their shape, although they can be
mounted on the surface of something that can. They also can't
self-replicate, although they can presumably be mass-produced by a
sufficiently advanced nanotechnology. Also, while their chemical properties
are real, they're not straightforward - the atoms of the semiconductor
substrate don't simply go away. At best, you'll have an atomically thin
programmable layer sitting on a bed of silicon or gallium arsenide. At
worst, you'll have discrete programmable islands jutting up from the
substrate like stones in a Japanese garden.

Too, since their electron orbitals are about 50 times larger than those of
natural atoms, they won't interact with natural atoms in a natural way.
Clever choice of quantum dot settings could allow bonding between artificial
and natural atoms, but even so, the spacing of the dots is a major
limitation. For example, we could tile the chip's surface with ersatz
glucose molecules, but these would be so big our taste buds wouldn't
recognize them. Still, if we really want the chip to taste sweet, or sour,
or like filet mignon, future engineers may find some dot settings to
approximate it.




Wellstone: A Logical Endpoint


A final, important shortcoming of this technology is its lack of 3-D
structure. The programmable layer is a nanoscopically thin veneer on the
surface of the chip, capable of mimicking only two-dimensional molecules.
This rules out the vast majority of organic substances, inorganic crystals,
and nanomachine components. You can't command a diamond coating to appear on
the chip, or even a quartz one. Fortunately, this limitation also has a
solution: We roll the chip into a long, thin fiber. With the P-N-P layers of
the quantum well, the conducting traces on top of them, and the memory and
insulation layers beneath, this fiber would have a minimum diameter of about
60 to 80 nanometers (300 to 400 atoms), meaning we could fit 10 to 13
artificial atoms around its circumference and a potentially infinite number
along its length.

Once we have these fibers, we can string them up in a 3-D lattice not unlike
the skeleton of a building, or else weave them together tight as basket
wicker. This is a tough nano-assembly job either way, but once it's complete
we have artificial atoms bumping right up against one another, able to bond
with neighbors on the same fiber and/or adjacent fibers. Now we can create
not only a thin film of goldlike pseudomatter but a three-dimensional solid
with the mass of wickered silicon but the physical, chemical, and electrical
properties of an otherwise-impossible gold/silicon alloy. Or mixtures of
other metallic or nonmetallic substances, including the "unnatural" and
"impossible" ones discussed above. And with the flick of a bit, the voltages
on the quantum dots can be altered, to change the solid from one miraculous
pseudosubstance to another.

In a discussion on this subject in the summer of 1998, Gary E. Snyder of
Pioneer Astronautics and I coined the name "quantum wellstone" (or simply
"wellstone") to describe this hypothetical but plausible form of
programmable matter. It's a term that has served me well in my fiction.

As for the material's mechanical strength, the bonding between quantum dots
on the same fiber is limited by their spacing, which is a function of how
small we can reasonably make the fences. These bonds will be weak: at best,
about .0025 percent as strong as a 5-ev carbon bond. If you programmed in a
bunch of carbonlike atoms, you'd get back a sort of watered-down diamondoid
crystal. The bulk properties of such a material are difficult to estimate,
but they're probably quite different from both silicon and carbon, in the
same way fiberglass is different from bulk SiO2 and polymer resin. And
remember, we have an infinity of electron patterns to play with, so if we
don't like the properties we end up with, we can simply adjust them.

Anyway, imagine a building constructed of such programmable matter. Or a
suit of clothing. Or an implantable prosthetic skeleton. Wellstone provides
more than simple access to "impossible" physical states; it provides the
ability to change states simply by shuffling electrons around. Electronic
devices built of wellstone could use the quantum dot arrays themselves as
computing elements, bringing a whole new meaning to the term "smart
materials." They could also lay out - and instantly rearrange - conventional
circuit traces with the amazing power of changing from conductor to
semiconductor to superconductor to insulator. To tunable laser, sure, or
tunable magnet. To window or mirror or self-optimizing solar collector.

I imagine a gingerbread cottage in the woods, its walls and roof a patchwork
of shifting colors. Sunlight filters through the forest canopy, its rays
falling on moving squares of gold or chrome, or vanishing into superabsorber
blackness. Parts of the cottage glow a malevolent red, or slice between the
trees with occasional blasts of laser. I notice a window, framed in silver,
inching its way along the walls. It oozes around a corner, and suddenly I
can see the inside of the house: a couch, a coffee table, a pair of fuzzy
slippers. And wearing the slippers, on the couch, a withered old lady in
sweatpants. Not a witch, no - this is my mother. She lives here. It's a
perfectly ordinary home, with a car in the driveway and a pink flamingo on
the lawn.

This kind of dramatic and instantaneous effect is serious Clarke's law
mojo - virtually indistinguishable from magic. It will probably be 10 to 20
years before the first serious programmable-matter chip is built, and
another 20 before anything like wellstone becomes anything like possible.
From there, it will take still more time before the technology diffuses into
mainstream society. But when it does - even if it's only a fraction as
capable as I've hinted at here - its transformative effects will be
staggering, a tumble through the rabbit hole for all of us. Some pundits
might fret about the corrupting effects of such power and instant
gratification, but as Alice herself remarked, "When I used to read fairy
tales, I fancied that kind of thing never happened, and now here I am in the
middle of one!"


--------------------------------------------------------------------------------

Wil McCarthy (wmccarth@sprynet.com) , an occasional rocket scientist, is the
author of numerous works of science and science fiction, including Bloom and
The Collapsium. Some portions of this piece were adapted from an article
that will appear in Analog magazine.


-- 

Ken

"Buddhism elucidates why we are sentient."
"Buddhism follows thought throughout the Universe."
"Karma means that you don't get away with anything."