When an atom fissions, big chunks of the nucleus break off as new, smaller-atomic-number atoms, but extra individual neutrons
also fly out of the nucleus. If you put enough easily-fissionable atoms
close enough together and start them off by shooting some neutrons in,
you can get a self-sustaining reaction:
the first generation of atoms are fissioned, emitting enough neutrons
to fission another generation of atoms, and so on. This is what happens
in nuclear reactors.
Nuclear engineers have a choice about neutrons: do they want fast ones or slow ones? When a neutron hits a nucleus, it can either fission it or be absorbed by it. Slow neutrons
are more likely to fission a given nucleus for any given collision, but
they can only fission very high quality fuel– specifically fissile
nuclides like the uranium 235 isotope. Fast neutrons
are less likely to fission a nucleus for any given collision, but they
can split less-fissionable fuel, they produce more extra neutrons per
fission, and they don’t get absorbed by the smaller fission products as
readily. The extra neutrons produced allow the reactor to breed fissile
Plutonium-239 from typically-uninteresting Uranium-238, allowing fast
reactors to get substantially more energy out of the amount of Uranium
we have on Earth than slow (“thermal”) reactors could.
Loading mercury into the system.
We can keep the neutrons at fast speeds by having only relatively
heavy nuclei in a reactor. Water, the traditional coolant, is made up
of two hydrogen atoms and one oxygen atom. The oxygen atom is more massive than a neutron by 16 times (because oxygen has 16 nucleons),
but the hydrogen has about the same mass as a neutron, and there are
twice as many of them. Neutrons traveling through a water coolant hit
oxygens, which slow them down a little, and hydrogens, which (on
average) slow them down a lot (by conservation of momentum). The use of
water as a coolant is the reason most reactors today are slow (thermal) reactors.
Thermal reactors can be much smaller than fast reactors because the
neutrons are more readily absorbed by the fissile nuclei. (Imagine a
fissile nucleus as a bar magnet and a neutron as a steel ball….if the
steel ball goes very slowly past the magnet, it’s much more likely to be
attracted to the magnet than if it was zinging by at, say, Mach 3!)
But wait! The neat thing is that there’s no rule that water has to be your coolant. Metals are much more massive than neutrons (how much more massive depends on atomic number),
so when neutrons collide with metal atoms, they retain much more of
their kinetic energy, like a ping-pong ball bouncing off of a bowling
ball. Using metals as coolants can let us build fast-neutron reactors
and use our vast “waste” reserves of depleted U238 uranium in the
reactor to be converted to Plutonium-239.
TerraPower has looked at using liquid sodium metal as a coolant. A big problem with sodium is that as an alkali metal with only one valence electron, sodium is very reactive, and the high temperatures involved would make it even more reactive.
In the pump system, any bits that could let in air or impurities
would give the sodium something to react with. This means, essentially,
that any moving parts would be hazards.
The magnetohydrodynamic (MHD) pump uses conduction to force the
liquid to circulate, so it has no moving parts at all. As a side effect,
this pump has a totally steady flow.
**Many thanks to Nick Touran from TerraPower for helping explain the
physics with the coolant to me, and to Jon McWhirter for conceiving of
the pump concept and for editorial comments.
How it works:
Moving a charged particle in a magnetic field creates a force on that charge in this way:
where 
is the force on the particle,
q is the charge of the particle,
is the particle’s velocity, and
is the magnetic field.
In Zihong’s pump, the big red anode and black cathode wires in the picture carry current through the system. They aren’t directly connected; the two electrodes are separated by a tube full of mercury (the prototype uses mercury, not liquid sodium, because mercury is liquid under standard conditions
while sodium is not). Conventional current flows from the anode through
the mercury, which is electrically conductive, to the cathode.
On top of and beneath this segment of tube are two magnets. This
means that the mercury, while it is carrying a moving charge, is in a
strong magnetic field. Like all moving charges in magnetic fields, it
experiences a force, and that force pushes it down the tube and through
the system. As long as current flows, this force (and thus the flow of
metal coolant) are maintained. See diagram.
The prototype runs on 101 A at 0.3 V.