Sebtal on Big Science Fusion

This is a rather good look seeinto the world of big fusion research that has sucked up a maximal budget fordecades with not much sense that we are much closer to a practicalsolution.  Yet his criticism ofalternative protocols is well taken. This is science that really hates to scale.

This well worth the read and yes,throwing billions of dollars at it is an excellent use of governmentspending.  As I pointed out decades ago,not one dollar ever landed on the moon and not one dollar spent here is ever goingto Iraqand the appropriate Swiss bank account.

While we are at it, also throwbillions at all the alternative approaches. We have to discover something important.

Details of Current Fusion Energy Work from Commenter Sebtal

JANUARY 12, 2011

In response to a posting on UK nuclear fusion research therewere several detailed comments from someone knowledgable in nuclear fusion.

1. Sure, they modelled it, and this is part of the ongoing work. Remember,Culham (I did my PhD there) used to be known as UKAEA and has been in magneticconfinement fusion from the beginning, being home of the team from the westsent to verify the then amazing temperatures being claimed by the SovietTokamak people, when the west were messing around with Torsatrons and ReversedField Pinches.

This is not some Johnny-come-lately "new scheme for fusion"bull-snot, nor (as you are correct in saying) is it some new silver bullet:these people ARE the Europeans who have been trying dozens of topologies etc.and more of that endless theorizing: it never stopped!

Plasmas are hideously non-linear and difficult tomodel. We still don't know exactly how burning plasmas will behave in fusionreactors. Such models, being horribly non-linear and massively computerintensive, and can always be improved. Sure, the potential benefits of alphasin generating "free" current and transitioning to steady stateoperation without the need for enormous indirect current drive have been theorized,and even modeled before.

The significance of this paper is that it represents an advance in the accuracyof the modeling and thus improves the credibility of the idea. One must realizethat the mainstream fusion programme is still very much in the physics stage.ITER is designed as a physics experiment, not an engineering test bed. Themodels of all the stuff that goes on in a fusion plasma are not exact, and inthe past such approaches have turned out to diverge from the reality. This,incidentally, is one of the reasons to have a healthy dose of skepticism aboutall these small private Fusion researchers... MCF fusion in the public sectorstarted out with a plethora of ideas of how to do fusion and converged onTokamaks and Stellarators as they seemed to work best experimentally.Nevertheless, with each new generation of machines performed less well thantheory suggested as new physics kicks in (one of the main reasons that fusionis always 20 years away), and likely those exploring new concepts will find thesame.

Anyway, these analytical and numerical results need to be compared against realdata before they can be truly believed, and this need is one of the mainreasons for building ITER: to better confirm our present knowledge of MCFplasmas and verify what we think we know about the behavior of a burning MCFplasmas. ITER does not represent the ideal commercial reactor, it representsthe ideal Physics experiment. After that, we have DEMO (and probably, DEMOs) toact as engineering testbeds and to try out optimized machines that use all thetricks we have discovered to be smaller, simpler and cheaper. 

2. Tom Craver – 

While it certainly makes sense to study hot plasmas, I often wonder if anengineering approach isn't exactly what is missing from fusion research.

No, not the engineering approach of "optimize an existing system based onknown physical principles". More of the "hackaround with wild ideas to try coming up with a new approach, and THEN applyknown physics to beat on the idea until nothing is left - but maybe sometimesSOMETHING is left that's worth trying

This is more of the "Science fiction as inspiration" approach toengineering - an engineer reads about something that is currently impossible ina SF story, thinks it is just too cool to continue NOT existing, and startsthinking about odd-ball approaches and trying to see what physics might applyto make them work.

"Dang I wish I had a Tri-corder/communicator/space-drive/ray-gun/holodeck!But how would it work? How can I MAKE it work?!" Where a good scientistwould look at the thing, think of 3 or 4 good reasons why it is impossible anddismiss it, a good engineer will be thinking "Well, yeah, butwhaddabout...." - looking for loop-holes in the reasons why it can't work.

That's what I see happening now in the various alt-fusion approaches. Granted,most or perhaps all of them will fail - but then, so have the traditionalscience-based "study plasma" approaches, so far. It's just that thescience-base approach cares at least as much about learning about plasmas andfusion, as they do about getting a working fusion reactor.

"Collapsing Fractal Magnetic fields!" "Magnetically Compressed shellsof plasma!" "X-Ray beams!" "Steam-punk fusor!""Kinked magnetic fields!" "Immaterial electron grid!" 

3. Well, really it needs both. In an ideal world, I guess a fusionprogram would consist of a team of senior engineers and physicists working outwhat the next machine in a path towards something that met thespecification for a "commercially competitive reactor" would looklike, iterating between sound engineering principles and the requirements ofthe physics, and modeling the kind of performance you expect of the plasma.Where there is a question mark, it would the role of physicists and engineers to come up with experimentson the available machines, new diagnostics, modeling etc. to fill in thatquestion mark, verify the model etc. As soon as the new machine has beenconceptually designed, the work of securing funding to build this conceptualdesign would go forward.

In practice, the program is definitely too physics led in my view. Thescientists are not employed like an engineer in the Apollo program with a clear product in mind, rather,they run the normal academic treadmill of publish or perish. This tends to meanthey are perfectly happy and able to make an interesting an valuable scientific career out of running experimentalcampaigns on a given machine using every possible diagnostic they can think of(verifying an old result on a new machine is still publishable, andscientifically valuable too) until literally everything that can be done has beendone. This is useful, because governments fund MCF experiments like Telescopes, they tend to be not so sympathetic to buildingnew ones until you can say you have run out of things to do with the old one.Furthermore, a lot of the programs are run by state institutions with lots ofpermanent staff, so you have to worry about things like redundancies, budget tohire new staff... a lot of work in bidding for new machines is structuredaround the idea of maintaining jobs etc. as well as doing new and interestingstuff. On top of that, the whole thing is taking place at arms length to therest of science and engineering. New ideas can take a long time to permeatethrough (I remember telling someone about the idea of using diamond as a plasmafacing material about six years ago, and being laughed at "you are goingto embed rocks in the divertor?"... the guy knew nothing of chemicalvapour deposition!).

This tends to lead to the engineering being relegated to "here is a budgetto make this kind of device", which works well for scientific experiments,which can be large, costly, but one off ventures, but not so well for ensuringthe concept that is eventually evolved is truly viable for commercial use. Somuch is dependent on things like machine geometry, I wonder if what we learnfrom ITER will be generally applicable, or contain hidden requirements that themachine look substantially similar to ITER, and not be easily translatable to,say, a compact spherical tokamak-stelarator hybrid that might actuallyrepresent the peak of the design space. I worry that in the total parameterspace of viable fusion reactors, we could be missing a giant mountain as wetwiddle around to find the local peak.

That said, the plasma physics really is nightmarishly complex on it's own,irrespective of the device or scheme. I am highly skeptical that there is aparticular device that overcomes that complexity through a clever trick.Working with these small ideas, with the potential for looming complexity whenyou scale up, is tantamount to banging around in the dark and hoping for thebest, though by all means, try them out. With funding restricted though, itmakes more sense to plug on with what we are most advanced in, which isstellarators and tokamaks rather than throwing money at lots of small machinesand hoping for something new. Japanstill maintains a lot of university sized machines though. Besides that, fusionresearch did go through the process of random creative ideas: inertial,electrostatic, reversed field... small experiments of this type were abundant,with big labs running several experiments in parallel. We are left with ICF,stellarators and tokamaks because they worked best and were most worthy offurther investigation. Not because Hubristic people said "not inventedhere". In fact, the reverse was the case: investigation on reversed fieldpinches and stellarators dropped radically in the west, where most interestwas, because the soviets Tokamak concept massively outperformed them.

Those still pushing fusors and RF configurations are in part the same people,who just never gave up on them (which is not necessarily a good thing!) and/orpeople who do not have access to lots of the original research which can bedifficult to get hold of as it is often not on-line, may never have been publishedas it was a null result or because the people working on it were not soconcerned with publication as they are now. A lot of it may still be locked upin the minds of senior or now-retired researchers, most of whom are working inthe MCF programmes, who's response to such ideas is to chuckle to themselvesand say "everyone knows that won't work! We tried it back in the 50's andit was a disaster." but not really bother with debunking it as they don'ttake programmes outside the main fusion programme seriously. And if thatinformation is conveyed to the researchers, it can look a lot like "notinvented here" and "we don't take you at all seriously".

Even within the MCF mainstream, the publication and "new machine"problem means you find things you think are new and then discover resultssimilar to your own that are older than you are, but which have been forgottendue to technical limitations. In the 70's they lacked the computational powerto do so, so assumed it is micro scale turbulence model it as an effectivediffusion, even though the experimental evidence shows transport at the edgewas and still is intermittent, coherent bursts of plasma being shot out of themachines. People just smoothed the data to remove the spikes, fitted aneffective diffusivity, called it anomalous and moved on. This is fine for somepurposes, as long as you remember that the transport is NOT laminar anddiffusive, and the "effective" diffusivity is just a crudeparametrisation suitable for some tasks but not for others. Over twenty tothirty years though, people tend to forget.

4. The stuff you actually have to worry about in big machines like ASDEXetc. and larger is not best understood from a strictly particle view.It's stuff like turbulence. MCF plasmas tend to be quasineutral, with electricfields effectively screened after a few micrometers. Even when at temperaturessufficient for fusion, their gyro-orbits (10cm) are way smaller than the plasmadimensions (meters). These problems were overcome decades ago, and part of thereason why people thought that Fusion would be a lot easier than it has turnedout to be.

The reality is most transport of heat and particles in modern large scaledevices is "anomalous" which is the name we give to stuff that wedon't understand. A mix of instabilities; drifts we hadn't taken into accountarising from, well, all sorts of things from small electric field perturbationsto ripples, wells, and islands in the magnetic field; and properly nasty fluidstuff like turbulence. These problems kick in at different sale lengths, whichis why Fusion has been "20 years away" for 60 years. Cautious groundsfor optimism then that we might not find too many nasty surprises (and perhapssome nice ones!) in ITER as the plasmas are now big and hot enough not toexpect us to be looking at plasmas through the wrong kind of model (particlerather than fluid, for example).

But that is still a lot of nasty nonlinear bits of physics interacting witheach other, and a lot of self organisation is going on in the plasma, so it isan analytical nightmare, and computationally horrible. Only recently has theresources to model more than a bundle of flux tubes in 3-D. I am hoping thatGPGPU computing is really going to help here. Further, they are very difficultto properly measure (when it is happening in the centre of a plasma), hard todo fully repeatable experiments. Your control knobs are rather indirect, andthe precise way the plasma in a given scenario evolves can be highly dependenton the precise condition of the machine, including impurity levels andcleanliness of the wall). Experimental MCF physics is pretty horrible!

A lot has been done with empirical scaling laws (beware!) but if we really wantto design good reactors, my feeling is we need to understand the physics of theplasmas and exploit all the tricks we can. Though it is possible to design amachine that will ignite in ohmically heated machines without such tricks, itwould have to be huge. And probably far far too expensive to ever be acompetitive reactor design I would have thought. On the other hand, this mighthave been a smarter way to go for ITER. The reduced costs in magnetic volumehave been replaced with the complexity of a design that has higher requirementsin other areas, like materials, required control diagnostics and complicatedplasma scenarios, restricting the explorable parameter space.

Now, I'm an experimentalist and worked primarily in the edge (though I've juststarted a bit of work on neoclassical transport in stellarators), so I'm notthe best person to ask about the significance of this work. I don't follow therhyme or reason of NBF seizing on this particular paper (there are plenty ofothers of a similar bent). I would guess incremental as another bolt towardsmaking *better* plasma scenarios that achieve higher confinement times,densities and temperatures, rather than something that suddenly "lets usdo fusion". Actually, we already probably know enough to "do Q=10fusion" in a very large L mode tokamak. Nobody has done it (though it wasthe original ITER design), largely because people keep cutting the budgetsevery ten years. Nevertheless, it would not make for a good power plant andwould be completely unoptimised.

The key areas in progressing from what we currently know how to do, which ismake big physics experiments, to moving onto something more compact, simple andsuitable for commercialisation is:

1. Transport barriers.

Transport barriers (which I think I am correct in saying we still don'tfully understand) in simple terms: the dominant transport mechanism seems to beturbulent, and if you get a velocity shear layer in the plasma, you can createa local barrier in the plasma that blocks the transport. This leads to thingslike H-mode (High confinement mode), that blocks transport at the edge andpushes up the core pressure quite dramatically, and "advanced mode",involving an internal transport barrier that does the same again. These arethings that allow us to get to higher densities, temperatures and confinementtimes in the core, which in turn means smaller devices with less auxiliaryheating, and possibly less stored energy, which is a good thing as a disruptingplasma can dump a fantastically high power loading onto a very small area ofthe wall, which can pollute the machine with lots of nasty heavy metal atomsthat can be very difficult to remove, but ensure all your future plasmasradiate their energy away in line emission and bremsstrahlung.

What would be the holy grail in this area is understanding the exact causalrelationship between turbulent transport and the shear layers, how the shearlayers form, and if we can design plasma scenarios where they formspontaneously or can be induced by outside methods.

2. Indirect current drive. Tokamaks use the poloidal magnetic field to confinethe plasma, the toroidal component just adds stability. The poloidal magneticfield is generated through a toroidal current in the plasma, which is driveninitially by induction. Your plasma is essentially a single turn secondarywinding. This means that your machine is intrinsically pulsed. Just about finefor very large physics experiments, not so fine for a reactor, as high pulsecurrents mean your machine is being continually put under stress and strain. Arequirement for incredibly high vacuum and low levels of elements like oxygenand nitrogen contaminating the plasma, combined with a vacuum vessel that isbeing whacked every few hours or so (or less) with pretty hefty impulses fromhigh power coils is not a great combination. Indirect current drive throughradio waves, or self-organising currents in the plasma, offer the opportunityfor the shot to continue after the solenoid swing has been exhausted, eitherallowing an extended duty cycle or possibly continuous operation.

The particular reason for Culhams interest is that the UK fusionprogramme has been concentrating on a variation on Tokamaks known as theSpherical Tokamak. By opting for a shape that is less donut and more coredapple, you can get some benefits to confinement and radically shrink themagnetic volume (the biggest single cost of a reactor) required to obtain agiven core temperature and pressure. However, this leaves less space for thecentral solenoid, or rather, less space for the armour required to protect yourcentral solenoid from being bombarded with neutrons and (if superconducting)quenching or (if copper) being turned gradually to cheese. So, again, indirectcurrent drive here would be a big advantage as you could have your solenoidactivate to begin the plasma scenario, and then after sufficient self organisedor indirect current exists in the plasma, drop the solenoid out and into a pitsafe from pesky neutron damage.

This is where this work fits in... understanding the theory better now means wecan start to design scenarios and experiments to run on ITER and, ultimately,design machines that are more credible power plants in the future.

3. Materials.

100 displacements per atom over the life time of the machine from neutrondamage. Tritium bred in situ. Power loadings on the divertor of severalmegawatts per square meter (which must be both conducting and non-porous), withmuch higher peak loadings for transient instabilities like Edege Limited Modes(a depressing side effect of the H-mode). Enough said really.

On top of that, some people are starting to suggest that some of the advancedplasma control methods that are required to operate in these advanced scenarios(analogy: high performance fighter jets are no longer aerodynamically stableand rely on active feedback controls) rely on diagnostic techniques that may nolonger work in the kinds of plasmas ITER is supposed to run. Fun and games allaround.

Make no mistake, these are huge challenges... my overall my feeling for fusionis not overly optimistic (though not necessarily totally pessimistic) and I doworry that the political and public understanding of ITER is radicallydifferent from what the physicists think, and also that not enough work hasbeen put into thinking through the feasibility of these things in a marketplace where politicians no longer sign a piece of paper and institute nationalinfrastructure. I recently met a researcher for the EU Parliament who had beenworking on ITER funding. There is a dangerous mix here of Physicists pushing anagenda that fits what they imagine fusion power should look like, and EUpoliticians looking for "an equivalent to the apollo programme" whoactively yearn for the days when big state infrastructure programmes existed.This seems to me to be a recipe for bad strategy for the direction of theprogramme as a whole. I do think we might be better off opting for a stagingpoint with fission-fusion hybrid machines, but generally the community is nowlocked into supporting ITER. Anyone hoping to realistically jump from ITER to aproof of principle commercial reactor producing electricity that is alsosuitable to compete with fission, gas or coal plants is in for a nasty shock Ireckon. Naturally, Physicists tend to act as though a thermal Q=10 means thework is done. In reality, it means the work is just starting. But this isslightly off topic, you can read more about my views on that on a comment here:

As for the start ups with their RF configurations and polywells, well, I thinkthey are treading the well worn path of massive optimism followed by the harshrealities of nasty physics kicking in at larger powers and scale lengths.


Two minutes after posting this: "I am highly sceptical that there is aparticular device that overcomes that complexity through a clever trick."I read that MIT have built a machine (Large Dipole Experiment) that somehowmakes turbulence, the bane of all controlled fusion research, work to confinethe plasma... :)

So, perhaps there are new thoughts to be had, but a lot of the stuff reportedhere on polywells, reversed field configuration and beam-beam fusion is stuffthat was tried before and was discarded in favour of Stellarators, Tokamaks andICF... so I would take with a big pinch of salt that smaller teams with lessfunding (though it buys better technology now) are going to crack the problem. Iam particularly sceptical when such claims are backed by nice straight linegraphs that don't appear to be subject to any regime changes as the devicesscale with power density or spatial scale lengths.

5. Goatguy

Sebtal, a magnificent reply. I feel humbled, and pleased that you took the timeto enlighten, obviously using the watered-down, yet still almost tangible argotof the physics you practice. Gyroradius, line emission, Bremsstrahlung, ohmicheating, turbulent/shear modes, ... I had no idea that the ion mean excursionoutside individual flux tubes was on the order of microns. Stuff there to chewon.

But I also want to point out to the few brave readers that have made it thisfar - note that the discussion, necessarily, edged from science toward designgoals, then from there to the realities (and unrealities) of the financing ofthe research, the politics of the financing, and the impetus of the politicalparties to politic the issues. It is starkly clear that as one wag said some 25years ago, "the only thing standing between us now and jaw-droppinglysuccessful fusion... is the political will to fund the science to resolve theissues that stand in the way."

On this I agree.

Secondly, Sebtal notes that the (relative) microfunding of alternate plasmaresearch is likely, when scaled up, to hit the walls of the unpredicted effectsof scaling itself. We shall see. I too remain not very optimistic about most ofthe alt.fusion proposals. Farnsworth fusor is magnificent in its simplicity andpresent-tense ability to generate copious neutrons from relatively pedestrian("high school geek home-made") apparatus, but "copiousneutrons" is an enormous distance from practical power-generator levels offused nucleons. As in 10 to 15 orders of magnitude more. 10,000,000,000× to1,000,000,000,000,000× That's a lot of scaling.

6. Sebtal

Er, sorry, it did get a bit long and wonkish.

No comments:

Post a Comment