.@physicstoday 'It sounds like science fiction, but the fusion (between quantum physics and machine learning) is happening right now.' @carrasqu * @rgmelko used Monte Carlo-sampled equilibrium spin configurations of an Ising Model to train a feedforward neural network https://t.co/qctnlEvF7Z with the Adaptive Moments (Adam) algorithm (using a cross-entropy loss function w L2 regularization) to identify paramagnetic/ ferromagnetic states; after training, network correctly classified samples not seen before. +, it located Tc & found the critical scaling exponents! pic.twitter.com/dWOPgo0NTh
Figure above shows critical point Tc for the triangular lattice Ising model (4m C&M @arxiv 1605.01735v1). PT article also discusses work of @gppcarleo& @matthiastroyer on Restricted #BoltzmannMachines & quantum states in detail; + has nice intro to #TensorNetwork Representations.
Karlsruhe, the Southwest German town, home to the Forschungszentrum Karlsruhe [The Karsruhe Research Center - a major German center for nuclear research) and the Karlsruhe Institut fur Technologie, will host two separate Conferences on Materials for Energy Applications this year - in July and October respectively.
The July Conference (EnMat 2010) will mainly deal with materials for non-nuclear energy applications - Hydrogen Storage, Fuel Cells, Thermoelectrics, and related topics (though there will also be a plenary talk on Fusion Materials - this is especially interesting since Fusion does represent, well, a fusion of hydrogen and nuclear technologies). Extremely interestingly, a Fusion plant can be conceived as a complete hydrogen economy - it uses two isotopes of hydrogen - deuterium and tritium as fuel, generates (or breeds) tritium as a byproduct, and the resulting fusion heat can be used to split water either thermo-chemically or electrochemically to yield molecular hydrogen - which can be used in fuel cells to generate electricity, or burnt in internal combustion engines directly. [I discussed this fascinating possibility in my presentation Nuclear Hydrogen Production: Re-examining the Fusion Option and the accompanying paper at the Canadian Hydrogen Association Meeting in 2007.] Fusion does indeed look even more interesting when viewed from the Hydrogen Economy prism.
The October Conference (NuMat 2010) will deal mainly with Materials for Nuclear Applications - fuel materials as well as structural materials for nuclear plants. NuMat 2010 will be a combined venue for several conferences on related topics which have previously been occurring separately, and there will be 6 major themes at NuMat 2010:
* Thermodynamics and Thermophysics of Nuclear Fuels
* Materials Models and Simulations for Nuclear Fuels
* Radiation Stability of Complex Microstructures
* Molten Salts for Nuclear Applications
* Structural and Functional Materials for Fission Reactors
* Structural Materials Modelling and Simulation
Fusion has long suffered from the unfortunate impression of being a technology that is always 'thirty years into the future'. Reality has always been more complicated, even if sometimes the impression seems to ring true. Since nuclear fusion has been understood for longer than nuclear fission, the seeming lack of progress in commercializing fusion has appeared especially frustrating. While technical issues certainly remain to be addressed, the lack of a compelling future scenario within which nuclear fusion could be seen as a 'natural' energy solution has also been a barrier in the techno-scientific as well as policy discourse surrounding fusion.
If seen merely as an attempt to extract energy by fusing deuterium and tritium atoms (in the simplest conception), fusion appears less compelling than when seen as a natural part of a future carbon-free hydrogen economy. Deuterium (D) and tritium (T), after all, are isotopes of hydrogen, and the energy they yield on fusion is usefully seen as nuclear hydrogen energy. But what if the heat yielded by the neutrons in D-T fusion were further used in thermochemical schemes to create molecular hydrogen, from which chemical or electrochemical hydrogen energy could be extracted? If this is successfully done, the transportation sector of the future could well come to be powered indirectly by fusion.
I sketch out and elaborate this vision for a fusion-driven hydrogen economy of the future in my paper Nuclear Hydrogen Production: Re-examining the Fusion Option. I discuss more generally a vision for a Fusion Island (first sketched out by Nuttall & Glowacki), in which a complete hydrogen economy is envisaged - a scheme which uses all the isotopes of hydrogen (protium, deuterium, tritium) in all forms of matter (solid, liquid, gas, plasma). I discuss the new perspective in which fusion appears when seen through such a hydrogen economy prism, the policy implications thereof, and the likely present-day economic actors who might find such a vision of the future hydrogen economy sufficiently compelling to begin more actively participating in and funding fusion R&D today. Such a new perspective on fusion also sees both fusion and fission as complements instead of substitutes, and offers novel possibilities such as fusion breeders of fission fuels, as well as, for example, fusion-fission hybrids, and fission breeders of fusion fuels.
Update Presidential Science Adviser John Holdren, giving the Rose Lecture at MIT on 25 October 2010, discussed the role of fusion and fission in providing future energy options that would mitigate climate change. He mentioned that both fission and fusion represent energy sources with 'nearly inexhaustible' fuel supplies, and though the fusion fuel supply was 'much more inexhaustible' (paraphrasing), that was not much of an advantage over fission since fission was 'quite inexhaustible already'! However, he also stressed that he personally was in favor of funding fusion R&D, since the number of such 'nearly inexhaustible' fuel options was so small. However, this funding could only be sustained if the overall funding pie for energy R&D of all kinds was increased. Here's the video of part of his talk where he discusses this issue:
Dr. Holdren also presents a number of quantitative projections for the future of nuclear power that are worth summarizing in brief. The world currently has about 440 nuclear reactors which produce a total of 375 GWe of electrical energy, constituting about 13% of world total electricity supply, a percentage that is declining even as new plants are being built - since other sources of supply are growing faster in the aggregate. He feels that in the next 90 years, that is, out to the year 2100, the world total supply of nuclear power would fall considerably short of the 3500 GWe total that some analysts have hoped for [and which would have been an order of magnitude larger than current capacity].
He feels, however, that by the year 2050, a rough quintupling of current supply, to about 1700 GWe could happen. However, I found the most remarkable figure in his talk to be the estimate of Remaining Ultimately Recoverable Uranium (RURU) as 100 Million tons, based on a recent MIT study. What this means is that a once-through fuel cycle option using natural or lightly enriched uranium will remain competitive, and that reprocessing and breeding options may not need to be commercialized for several decades yet. Of course, the issue of how this recoverable uranium is actually distributed throughout the world, as well as how widely the technology of extraction will become available, remains. Different countries who feel uranium-constrained may still very well choose to pursue fuel cycle options that include reprocessing and breeding technologies.
