Number 727, April 14, 2005
by Phil Schewe and Ben Stein
A New Kind of Equilibrium
Normally heat will flow from a hot place to a neighboring cold place.
In a new form of thermoelectric refrigerator, proposed by Tammy Humphrey
(University of Wollongong, Australia) and Heiner Linke (University of
Oregon), temperature imbalances can be held at bay by electrochemical
imbalances. The implications? Possibly much more efficient forms of
no-moving-parts electric refrigerators.
Heat and electricity are two
forms of energy, and in a special circuit, made from thermoelectric
materials, a temperature difference can generate electricity and, conversely,
a voltage difference can bring about a temperature difference. A thermoelectric
circuit usually consists of two semiconductors joined at two junctions.
One of the semiconductors is of the p type with a surplus of holes,
the other of the n type with a surplus of electrons.
Here’s how you
can generate heat or electricity in contrary phenomena. In the Peltier
effect, a voltage imbalance will pull electrons and holes out of one
of the junctions, thus cooling that junction and warming the other junction.
In the Seebeck effect, things work in reverse: a temperature imbalance
between the junctions will set electrons and holes in motion, thus constituting
an electric current. The Peltier effect is at work, for example, in
on-chip cooling of critical microcircuitry. The Seebeck effect is used
in powering spacecraft (too far from the sun for photocells to be of
use), where the heat from a radioactive source is used to make electricity.
What keeps thermoelectric devices from greater applicability is the
poor efficiency, typically 10%.
One of the main problems is that some
of the heat (applied at one junction) used to drive a current through
the circuit is carried by electrons to the other junction, reducing
the thermal gradient and therefore sapping the process of generating
electricity. What one needs is a circuit good for electric conduction
but poor for thermal conduction by electrons. And this is what Humphrey
(firstname.lastname@example.org ) and Linke's proposed circuit would do
(see figure at www.aip.org/png
The p-leg and n-leg parts of the circuits would consist not of bulk
matter but of quantum dots, nanoscopic pieces of matter in which only
select electron energies are allowed. Engineer the dots to discourage
the higher-energy electrons carrying thermal energy, heat leakage will
drop, and the overall efficiency will go up. The best thermoelectric
efficiencies are about 10%. If efficiencies could be pushed to 50%,
the thermoelectric approach (silent, less bulky, no refrigerant, long
lived) would compete to take over even bulk household refrigeration,
Humphrey says. (Physical
Review Letters, 11 March 2005; lab website www.humphrey.id.au,
Cooling of Bulk Material
Cooling of bulk material has been achieved with a solid-state refrigerator.
At the heart of the NIST-Boulder device is a tiny sandwich-shaped diode
whose layers are successively a normal metal, an insulator, and a superconductor.
The stack has the effect of pulling the hottest electrons out of the
normal-metal layer. This no-moving-parts refrigerator is not the first
to achieve 100 mK temperatures but it is the first to do so with technologically
useful cooling powers.
The NIST micro-fridge chilled a cube of germanium
about 250 microns on a side and with a mass of 80 micrograms. This sounds
like a small speck of matter, but it was enormous compared to the size
of the refrigerating junctions (see figure at www.aip.org/png
). Indeed, the ratio of the volume of the cube to the volume of the
junctions is 11,000. This is equivalent to a refrigerator the size of
a person chilling something the size of the Statue of Liberty. In preliminary
tests, the cube was cooled from 320 mK down to 240 mK.
should lower the base temperature to near 100 mK. According to NIST
physicist Joel Ullom (email@example.com), their refrigerator works
best at temperatures below 1 K, so it won't be used to cool foods. But
it will be very useful for chilling circuitry on chips and maybe samples
as large as the centimeter size. (Clark et al., Applied Physics
Letters, upcoming article)