The trouble with fast breeders is that you need a huge fissile inventory to get one started.

A study of the PRISM design, a modern fast breeder that uses integral reprocessing to reduce the inventory of plutonium that's sitting around, not fissioning, indicates that if we reprocessed all spent fuel in the U.S. today, we'd have enough Pu to power about 33 MWe worth of breeders, about 1/3 of the current fleet.

It comes down to energy vs power. "Fast" breeders can tap a huge energy resource, but they can only produce a low level of power because they tap that energy slowly. That's the whole reason they use highly reactive sodium as a coolant; to reduce fissile inventory per unit of power, you need a coolant with incredibly high heat conductivity. Lead is a much easier material to handle, but a lead-cooled fast reactor has 1/3 the power density of a sodium reactor, which means 3x the fissile inventory.

Fast breeders, therefore, can sustain energy production over a long period of time, but they can't drive a rapid expansion of nuclear energy, like the 5-fold increase we'd need to replace fossil fuels and stop global warming -- it becomes hard to justify the economics of a plutonium economy in the same way... You have to be thinking more than "seven generations" ahead to see the economic boon.

Thermal reactors effectively trade a moderator material (water, graphite, etc.) for (relatively scarce) fissile material. Thermal reactors (even today's LWR) extract 10-20x as much power from fuel than do fast reactors with less aggressive design (although the LWR extracts only 2% as much energy, in the long term, as a breeder could.)

A thermal breeder, based on thorium, could provide the best of worlds. With small inventory, it's possible to meet high power requirements, but by using abundant thorium instead of rare U-235, be able to sustain that power for thousands of years.

> if we reprocessed all spent fuel in the U.S. today, we'd have enough Pu to power about 33 MWe worth of breeders, about 1/3 of the current fleet.

That should be "GWe" not "MWe".

Also, I dispute your choice of statistic. Why not show what happens if you start fast breeders on mined uranium, just like LWRs?

Sorry I missed a factor of 1000. Fast breeders on mined uranium have the same problem as plutonium breeders... Except you can breed plutonium faster (but less efficiently) in LWRs!

I mean, it's an odd statistic. You can use U-235 directly, or you can burn it in an LWR, getting back just 1/5th its equivalent in Pu-239, and use that. If you're trying to maximize fissile input for fast breeders, LWR spent fuel isn't an efficient way.

And moreover, with mined uranium there's no limit to your rate of expansion other than mining. You don't have to sit and wait for your atoms to reproduce (doubling time 20-40 years), when you can just dig up new ones.

Some back of the envelope numbers:

Assuming 3% enriched fuel burned to 40 GWd/ton, and 33% thermodynamic efficiency, the US nuclear fleet (~100 GWe) consumes around 100 tons/year of U-235 in fuel. At about 10 kg/MWe fissile inventory, just 10 years' of present US uranium consumption would be enough (1,000 tons) to start up a 100 GWe fleet of fast breeders -- same as the current LWR fleet.

I found a table of fissile inventories here:

http://nuclear.inl.gov/deliverables/docs/msr_deliverable_doe...

Over the next 40 years there is little economic reason to move to breeder reactors. However, they are a much better long term solution to waste than simply storing the stuff in the ground. Also, because the LWR to Breeder cycle spends so much time on the Breeder side of the equation a fixed number of LWR can feed a growing number of Breeder reactors over time. Thus you need to map out what happens over long time periods to get the balance correct.

Edit: If a LWR extracts 2% of total energy in 0.2% of the time, a breeder would need 10x the fuel to produce that much power and it would take 500 times as long to use up that fuel. Thus the stable ratio of LWR to breeders would be 1:50 relative to power output, but it would take ~750 LWR fuel cycles to reach that balance assuming you did not start wait for a huge stockpile to begin. (Breeders are not 100% efficient but the math is not all that far from reality.)

And that gets us to the central problem of our energy policy.

Utilities are moving towards a short-term thinking model where they install natural gas turbine with a low capital cost and use electricity deregulation to offload the risk of volatile natural gas prices to the consumer. "Renewables" are also driven by short-term motives: making a quick buck by exploiting government subsidies.

Nuclear energy can give us a sustainable energy source the thousand-year timescale (energy-wise, breeders can certainly consume seawater U238 profitably,) but we need to be thinking on the timescale of decades, not the next quarter, to get there.

Your 100 and 1000 year assumptions are just silly. You assume energy demands will remain constant, while they are exponential.

Just think what will happen if electric cars become popular. We will need a lot of electricity.

Resource usage tends to follow a http://en.wikipedia.org/wiki/Logistic_function not pure exponential growth. Picking a good limit is hard but assuming energy use per person is stays around 5x our current value and US population is limited to 1 billion people 1000 years is an extremely conservative estimate.

Conceder: Between 1980 and 2006, the worldwide annual growth rate was 2%. (http://en.wikipedia.org/wiki/World_energy_resources_and_cons...) If you assume world energy usage will grow 2% per year for the next thousand years you end up with 400,000 times current energy usage however, the sun only supplies the earth with ~10,000 times our current energy use so we would fry the planet with waste heat far before that level. (Think global warming to 800 degrees Fahrenheit.)

PS: In 2009, world energy consumption decreased for the first time in 30 years (-1.1%) or 130Mtoe, as a result of the financial and economic crisis (GDP drop by 0.6% in 2009) so growth is not necessarily even guaranteed.

Your assumption that we will stay exclusively on this planet for the next 1000 years is even more silly.

At no point did I assume that. If, I extrapolate based on something that is not true (aka we stop mining uranium) and look at it in a given area (aka the land area currently occupied by the USA or the Earth) then clearly I am not trying to predict human behavior in some other location.

If you want a prediction of human behavior, fine. I suspect that we will move to a mixture of fusion, wind, and solar power long before our supply’s of fissile materials become an issue. I also expect that our initial expansion into space will primarily involve exporting technology into space and collecting raw materials from space. However, initial space exploration is going to have minimal impact on life on earth with few material goods being sent back and forth. I also suspect that it’s going to take far longer than 1000 years before 1 billion people live outside the earths atmosphere.

Edit: The Americas where "discovered" 518 years ago and dispite being there for the taking and far more hospitable than space 200 million more people live in India than all of the Americas put together.