Design of 1 TWe Fusion Power Plant

Sitting in a nuclear engineering lecture hall around 2009 I was about to hear a seminar fusion for power plants. The general public was/is pessimistic, but this guy who works in the field is sure to bring us optimism. I was wrong. The entire time was a long stream of “if you thought it was a long-shot, it’s worse”. What do you mean tokamaks can’t operate at steady-state? They create current in the plasma with what?! Hold on, how many things does the DEMO commercial plant hand-wave?

I walked out with a new take — why are you in this field? As a society, should we have completely abandoned fusion? Why are you doing this?

I’ve gone many years without sufficiently answering that question to myself. At last, I will do that here. But for this to work, I need to work backwards from the glorious O’Neill vision of the future. That is the future where we live in rotating space habitats, possibly at L4 and L5 orbital points. Importantly, automation and AI allows for a massive industrial base, bootstrapped by the reusable Starship. At this time, many of these confinment and logistical troubles can be solved by greater scale. Surely, our future selves will have use for 1 TW of power, and much more. Can scaling be the hero that fusion need?

Is fusion power cooked?

The story of the future, as we know it, has Starship fulfilling its HLS obligations and landing US astronauts on the moon, exactly 1 month before China does it. However, this time we do it the “right” way and the program has a sustained presence. All the while SpaceX and other “commercial partners” pivot the archecture they’ve built to also service private money-making endeavors. They build the mass-driver which launches AI sattelites in to sun-synchronous orbits, allowing that industry to continue to scale, and so our future as a space-faring civilization begins.

It is well-published that the moon has many shortcuts to making solar panels (see anthrofuturism for reference, not getting specific here). The costs on solar panels on Earth have been falling exponentially. The AI sattelites will use solar panels. So will the moon base. A giant cislunar economy will be powered by solar panels with no shortage of collection area. There is no reason to think this pro-solar trend will not continue. Combined with battery technology (on Earth) it displaces fossil fuels. The batteries are not necessary for large space infrastructure. If that’s not good enough, we’ll beam solar power from space to Earth.

To turn this into an insult, the Kardashev scale literally measures how advanced a civilization is in units of solar panel area. With dramatic space development potentially on the horizon, this terminology has been adopted by an increasingly wide circle of technologists. This asserts the dominance of solar power as an assumption.

It is painfully clear that solar can do the job, and that fusion is not needed in the cislunar space. For the outer solar system, people tend to recognize that nuclear power is needed, but by that they usually mean fission power!

It takes a really tough contrarian to stand by fusion power.

But why? Is fusion really that hard? Well yes, it is. But, why is it hard-hard? And when, if ever, will the script flip in favor of the fusion optimists?

The Scale Dividend

We know very specifically terms what we want to build. Almost every timeline assumes a particular physical form. First we build Iter, then DEMO. Or SPARC and then ARC. It’s what every power-point slide says. Go look, look at this slide.

There is historical template for this expectation. Fission plants started small and then got bigger as we built more, and wanted to economize on the economies-of-scale better. The first set of power-producing reactors are relatively bad economically (both fusion and fission) but then that gets the data you need to for scale-up later on.

But physics has a voice, and I worry that it hasn’t been given its say. There are other unrelated reasons to question the conventional story. When the US plants were scaling up, we were the workshop of the world. We simultaneously hit the maximum practical plant size while many other factors stalled traditional industrial growth (to later be replaced by computers), economically.

R and B

I’ll introduce a reference here: On the size of tokamak fusion power plants

This very clearly recognizes trade-off between magnetic field (externally generated, but in plasma volume) B and size R. A greater size can compensate for a weaker field. Every single word, even if the abstract, is painfully explicit in the both direct treatment of this tradeoff, and assumptions that this tradeoff exists.

Two routes to FPPs are discussed: the more conventional one increasing R, based on the assumption that B is limited by present technology; and an alternative approach assuming the availability of new technology for superconducting coils, allowing higher B.

(R is the major radius of the tokamak, “a” often denotes the minor radius)

This frames the entire idea of what the fusion challenge is. Returning to my original question — why do we not have practical tokamak fusion power? The answer is that R and B are insufficient. Oh, and BTW, don’t let the plasma destroy itself through instability modes.

If you put yourself in the shoes of a 1970s researcher, it is clear that dumbly increasing R to get break-even is a fool’s errand. Why? Because you go broke before you break-even. Break-even being Q>1. You can get net energy with a large enough R, the problem is that you can’t build big enough!

Basic Proportionality Equations

This is all to give the “why” of why we don’t have fusion power now. I’ll re-state these equations in my own semi-equation form here.

From this, you would think that making the magnetic field (B) stronger is the right approach. And the entire field would agree with you! For the last 50 years that’s exactly what they’ve been trying to do, and thankfully we have fancy superconductors to help with that. But clearly, we’re doing all that we can there, and there are some limits. If you expand out the magnetic field term, there are new limits.

But with an electromagnetic so powerful, you will need support structures to brace it.

I looked into the SPARC details, and it seems that the ratio of conductor material to structural material is surprisingly close to 1-to-1. This suggests that we are already near an optimum point. If you apply a stronger field (and I’m sure we will) then your reactor chamber is going to hit the squared term for more structural mass.

So if you meditate on this, it works best to assume that we’re just already getting the best B that we can. If that is mostly set, then you’ve got R, the tokamak major radius, to work with, or (volume) in the equation above.

Now, mapping the plasma volume to the vacuum chamber volume (coil size) is a little less trivial. But there’s this thing called gyroradius, which is how large of a circle ions travel in the magnetic field. So again, with constant B, there’s this very important ratio of gyroradius to the distance to the wall (we can use minor radius “a” here). At first glance, it’s super obvious that the gyroradius needs to be less than R, but in practice is needs to be much less due to turbulence, resonance time, and other stuff I still don’t understand well enough. But it seems to me that (gyroradius) / (physical scale) is a good proxy for a form of containment quality.

This is a good case where you can use an “exception proves the rule”. There is a fusion reactor called “Alcator C-Mod” and its tokamak major radius is R≈0.67 m, and has a toroidal field up to about 8 Tesla. Woah! That’s a high field, on par with the big boys like SPARC. So this is a small reactor, high-field. No one is even considering this a fusion reactor, it’s just a plasma demonstration. Because R is too small to get good confinement. Because of the gyroradius.

Digression: ok ok, I’ve learned enough to walk through more steps if needed. First, you have relative particle loss rate, then it’s the energy loss rate (Tau_E), how long it holds onto the energy input, which is then used in the Lawson criteria. So you can do more work and get the same answer.

I hope it’s smacking you in the face at this point where I’m going. Obvious question — “what if we go in the other direction?”

Reason for 1 TWe

Due to many technical reasons, when we get a commercial fusion plant, it will still suck. But this is a power plant that looks like a power plant as we know it. I’m interested in a different question — what does the fusion power plant of a Kardashev Type I civilization look like?

To articulate the specific divergence, we (as in Iter and the like) are in the path of scaling up R and B to get good enough confinement to flirt with basic Q>1 operation over seconds or hopefully 500 seconds sometime in the next decade. What I’m asking here is, do we do ourselves any favors to just… jump way ahead? Assume a giant reactor. This will help confinement, but just fundamentally, I’m jumping way ahead to a high-output reactor.

The size of fission power plants stopped at around 1.2 GWe. Don’t get me wrong, there’s a good reason we did. They became megaprojects, and if anything, people are now trying to make them smaller. Iter costs $20 billion-ish. I can’t bring this up in polite company. ChatGPT takes detailed hand-holding to entertain the concept. But say we continue the initial scaling up trajectory and don’t stop.

We could absolutely not build such a plant on Earth, because we don’t build structures that large. But we need to listen to what the physics tells us. And to do that, we need to entertain the idea of not just 4–5x scaling up to commercial, but 4,000x — 5,000x scaling up, and just find out what that would look like.

Because remember the competition. Solar is highly effective in space, and will be able to scale up to larger power scales than what the grid on Earth is. What new, additional usefulness, does fusion bring to the table? One way to answer that is to offer a bridge to a more compact power source when solar collection area becomes absurdly large, like a significant fraction of the size of the moon. At that point, fusion might provide something useful, which is not otherwise obtainable by its competitors.

AI Answers for Design of 1 TWe Plant

What would a 1 TWe electric plant look like? It’s 2026, so of course we start by asking the AIs. I was so invested in this problem that I asked several with the same prompt, and collected the answers in this github repository. Then I refined from there.

github.com/AlanCoding/giant-fusion-reactor

First things first — every single AI rejected implosion fusion. I already had the same gut feel myself. This was not due to the difficulty of confinement, but of the engineering economics to operate consistently. Interestingly, I don’t believe any of these designs are impossible, and believe the core argument is that magnetic confinement is simply better. Very interesting to note, though, that this choice is not obvious on the small scale, but I will argue it is obvious at the large scale.

After digesting all of these answers, I have realized the main design elements are the following. This is mainly going from the most-firm observations to the less-firm observations.

1. Ideal confinement type is a long tube (maybe with stellarator-like helical twists) with some active control to maintain positioning
2. You benefit dramatically from a “minor radius” significantly larger than current designs, so the ratio of this dimension to the overall circumference or length of the active area would be low, like, less than 20
3. It is speculatively beneficial to have a straight pipe as opposed to a torus, because the ends offer opportunity for direct energy conversion and geometry (trapping all ion paths) gets a lot easier

This is still all too open-ended to be concrete. Consider these 3 existing types to choose from.

I’ll make a representative example of

• left: tokamak, SPARC
• middle: stellarator, Type One Energy Infinity One
• right: magnetic mirror, Novatron

These established, first thing we do is to ax the tokamak. Not good enough for continuous operation like we need. Like implosion fusion, this is rejected fairly early in the process. Even if the continuous operation problem is solved (a big “if”) the problem of maintaining current is still very major and a lot of the AI bots thought it just wasn’t ever going to be solved.

First thing that will be obvious to everyone is that power is proportional to wall area. This isn’t from fusion physics, it’s basic heat removal. The heat flux hits the wall, and we have to take that away to the thermal cycle or the radiator.

High Aspect Ratio

The ideal aspect ratio (ratio of major radius to minor radius) is high for super large fusion plants, but not crazy high. Most of the AI bots got this wrong. Interestingly, claude gave a stand-out answer, although it was the minority voice. To break down the qualitative answers:

- Long tube: meta, grok, ChatGPT
— Stellarator: claude
— weird: vibe, gemini

The “long tube” camp tended to give a length of 10 or 12 km. This was interesting academically for some first-thoughts on large fusion plants, but actually mostly wrong for reasons I’ll go into later. But claude gave a minor radius of 50 meters, which seems to have been for the right reasons.

Generally, for any design that’s “good”, we’re looking at about 50 meter minor radius and 1 km major radius or circumference. Looks about like this:

These about about the general vibe of the 1 TW reactor. I have concluded that both of them are good designs. The stellarator is more conservative, because those already exist, and you would just be scaling them up (aside from the greater aspect ratio, which you’d arrive at in the detailed design phase). But the math works the same for both, which is nice for this analysis.

You would use a higher aspect ratio because that’s part of how you spend the “size dividend”. The larger size makes a ton of things easier, but one of the very direct and qualitative ways is that it can reduce the size of the magnetic mirrors compared to the overall machine — less cusp losses. For the stellarators, it allows for smoother twisting geometries. Because particles circling through the stronger inner magnetic field and weaker outer magnetic field creates problems, and this is generally less if you have a skinnier torus.

But you still want to pay some of the size dividend in the minor radius, and I’ll explain why.

Scaling the Minor Radius — Bringing the Hammer

I said that the AI bots were generally wrong on a 12 km long tube. The reason is because you want to increase the radius of the tube itself (resulting in a shorter length, same wall area). Note: you can increase the plasma mass loading per length of the tube with the same overall B, that’s fine.

The reason is the same reason that Alcator C-Mod can’t do fusion. The gyroradius, and scale in general. The main way that pressure is maintained is the toroidal, or length-wise, field. That will have a certain B value due to the coils being a certain fixed mass/volume fraction of the machine. This is true for big or small. Because of this, the gyroradius is small, proportionally, for larger machines. This is saying “bigger is better” very directly.

Having the walls be 1 gyroradius away is not enough to declare confinment. There is radial drift even if the primary field for confinement works. There is turbulence and a bunch of other nasty things that come up, which are absolutely what the plasma physicists study today.

The only question we should be trying to answer is whether these problems get easier with larger scale. And yes. Yes, this is easy to answer, they do. The helical twisting fields of the stellarator are one of the answers to some of these things. I’ve convinced myself to think of it this way — for each gyroradius, you get a shot at your other systems stopping the particle from escaping, when it something (like an extra collision, or turbulence) causes it to drift further. But trying to confine with 2 or 3 gyroradius values to work with is like trying to play Galaga with just a few vertical rows. You just can’t beat it.

The more space you give yourself, the more likely the twist is able to do its job. Give yourself enough space, and even somewhat badly engineered solutions can manage to work well enough for confinement to work.

Back to the design process, this is where we throw out the super long tubes. We would still want a long-ish tube to minimize the end losses, particularly in the straight tube design. So, this is how it went roughly in my own head.

I can’t emphasize enough how increasing the minor radius of a fusion plant is a gargantuan green flag. It makes fusion, as a power source, spectacularly better. The main problem with fusion is confinement, and this makes confinement easier. So to further emphasize this effect I want to give some scaling numbers. An DEMO-like number for the minor radius would be 2.5 meters. So I will start from a tube that size (would be 12 km long) and then scale up the minor radius. I’m only interested in the cross-sectional analysis. Here, I will look at 3 scenarios.

1. Constant B as it is scaled up, the most conservative, no material efficiency benefit
2. Middle ground, relaxed B somewhere between 1 and 2
3. Lowest viable B, keep the gyroradius to the radius ratio to the same, no confinement benefit

So the basically idea is that we illustrate an engineering trade-off of either (1) improved confinement to (e) more efficient material usage. Why not both?

Of course we would actually seek a middle-ground which is what scenario (2) is and is the most realistic scenario. This gets math-heavy enough that you might want to look at my workbook. But I’ll hit a few select highlights here. Firstly, let me formalize what B is doing in the scenarios:

The scenarios 2 and 3 are explicitly and intentionally lessening B. This reduces coil mass, but also, we can’t reduce it by much because of the B² term, and overall the metrics being very highly sensitive to B value, but we can reduce it some while getting the same confinement FOM (3) or still somewhat-better (2).

But how can this possibly work? I direct you to plasma density. We need to formalize that cross-sections are basically going to set the T (temperature) that we work with, and changing that was never under consideration. So here is a density proxy, given as n. The reaction rate trends with n² because it is a binary reaction.

But this still misses something, which is the plasma inventory. Even if we have the state equation set via T and P, we can still have more meaning that it occupies more space in the center of the tube. The distinctions between scenarios start to really matter here. Relaxing B in scenario (2) or (3) requires compensating with higher plasma inventory, which should be intuitive.

Now as a final sanity check, I’ll give our confinement quality proxy here. The gyroradius of an alpha particle from fusion is given as rho. We care about the gyroradius relative to the size of the tube which can be given by rho/a. Lower values mean better confinement. You can see that scenario (3) does not get any better confinement, because that’s the whole definition of that scenario.

Let me recap this with a table that states everything again. High-level, we’re looking at scale up from 2.5 meters to 50 meters, and how that affects confinement quality and material needs. Here, eta is a factor showing how we split the size dividend between lower field and better confinement.

That confinement time looks really good. This is spectacular. It starts to look like something actually useful as a power plant. This also corresponds to much higher inventory — these go hand in hand, because lower density means lower power intensity per mass, you need more fuel mass in the reactor at any given moment.

Lowering field strength, probably something like scenario 2, is also really really useful for building a power plant. The lower value of beta sounds bad, but it’s actually really good. This means the external magnetic field is proportionally stronger relative to the plasma and this will help the confinement topics. And then the lower magnetic field obviously means you need less material. Let me just put those in numbers to show the scale of the benefit.

Really Basic Render

This is, by no means, detailed engineering of a fusion reactor. But it certainly accomplishes the basic idea. I want to clarify that I have not done any emperical calculations for the tradeoff between the tube radius and length. So the exact 50 meter selection (again, I credit claude for the original choice for the right reasons) is striking a qualitative balance between the 2 competing factors of the material & confinement benefits described here, and the end losses in the case of a long tube. In the case of a stellarator, we don’t have end losses, but twists are harder to get right in a more “compact” design, so a less compact design is expected to be better.

Originally, I thought that the helical coils from a stellarator should be borrowed for the linear design. But with further study of the instability modes… no, actually probably not. Oh well, I made this Blender render before I realized that.

This is trying to show a main helical coil, like helical fusion, as the main windings. Then a winding for control. I did just kind of think it looked cool. So now imagine this where the diameter is the size of a football field. It certainly feels Sci-Fi.

If you’re asking me what my design for a 1 TWe reactor is, this is basically it. A mostly-straight tube with a mostly-uniform magnetic field seems to be better than a stellarator. I would be sure to borrow from a few other current fusion designs. I think you’ll need a lot of magnetic control like what Thea energy gives as an alternative to the manufactured stellarator coils. But in our case these will be mostly for the stability issues.

Confinement Topics

While I now believe that the main field would be a straightforward B field in a straight tube with magnetic mirrors on the end (or FRC), there are also instabilities you have to deal with. It’s hard for me to wrap my head around these. So here’s an image of potential instability modes for the specific giant-long-tube design.

And for some more detail, here’s a table and whether scale hurts or helps.

There are green flags all over the place here. It looks scary, but fusion is really hard in general. Increased size makes the vast majority of these way better, but not all. Active control will probably be a simple fix for the center-line stability problem.

The flute instabilities seems like the worst remaining problem. It’s unclear if the increased size makes these worse, but it could. Some current papers, like Novatron specifically, talk about ways to address this in their own designs, but it has tie-ins to the overall design of the device which is not a great sign… or just a sign that the fully straight tube is still a bit naive.

You Can’t Skip Steps

I want to be clear that I’m not advocating that we skip a 1 GWe plant. I’m just speculating that maybe we build a 1 GWe plant. Then build a 4 GWe plant… then a 10 GWe plant. And keep doing a multiplier every time. It’s not unreasonable to imagine that these are all deeply uneconomical. But they could lead us to a future where the 1 TWe plant becomes economical. Even if disappointing, this gives a grounded and satisfying answer to the question of when fusion power will contribute significantly to our energy use.

Even if you didn’t need the step-ladder of smaller plants for the industrial proof-of-concept, you still need them to

1. prove the various stability modes are addressed.
2. make enough Tritium for the next plant

As this is going on, in a future of technological abundance, I’m sure there will be other uses of fusion. Long before fusion competes for power, it will be a contender for extra-solar probe propulsion. We will get plenty of neutrons. I don’t know what good the neutrons are, but we’ll probably find something.

Screaming Hot Heat Rejection

Because we are positioning this device to be competitive with space-based solar power, we have to be extremely skeptical of anything that would call for a very large radiator. Why? Because why would you even bother building this if it’ll demand a radiator area similar to the area you would need for a solar array? Solar panels in space are basically self-radiating. In either case (solar or fusion) you need to have a separate cooling means for the load demand so I do not want to take any interest in that part of the problem because there is no comparative advantage either way.

I looked into thermal cycles that might be possible at higher radiator temperatures. The final conclusion I got looks like:

FLiBe primary loop→intermediate loop→closed recuperated He or He–Xe Brayton cycle

You can get a radiator temperature of perhaps 340 C, but only if you assume close to theoretical performance and don’t do it with real working fluids. Once you include any real physical cycle, you easily demand around 200 to 250 C. Due to the T⁴ factor, that gets you about a 10x reduction in area on the high side relative to solar power. Again, solar power is a really good option in the very beginning.

I have to rate the thermal cycle options as not just bad, but outright barftastic. They are god-awful. Anything you might reasonably do runs into the thermal death cycle, where you need to use really bad efficiencies or really large areas. But really large radiator areas have to be completely rejected, with prejudice. To be clear, I’m fine with running at low efficiency, because we have plenty of Duterium. Throw the fuel in the trash, I do not care, we can burn through however many nueli we want. But the problem is that a fusion reactor needs to self-power with electricity. The efficiency floor comes about from the need to break-even on energy production. I was throwing around 340 C and then 250 C because those assumed a 10% power re-investment. Maybe this is too high, but maybe it’s too low. Obviously our superior confinment helps a lot for that number, but it’s still putting us in a very unconformable territory where we risk coming full circle back to fusion being economically non-viable. Just use solar!

This is where we come to my real conclusion. For fusion to be competitive, not just work, but be superior to solar, direct conversion isn’t just a good idea, it’s completely mandatory. If you look at fusion reactions, we have a maximum fraction of the energy that we can get from direct conversion.

• DT 20%
• DD 66.4% (but we probably can’t fuse these)
• catalyized DD 61.8%

Well I wanted to use catalyized DD anyway. But even if I couldn’t, after looking at how awful the thermal cycle options were, I’d take the 20% over a thermal cycle. Let me be perfectly clear what I’m saying, don’t bother with a thermal cycle. Take the heat produced by the reactor, and throw it in the trash. Straight to the radiator.

With our FLiBe blanket, that can get us 700 C temperatures. That can get us a 120x size reduction compared to solar arrays, and the direct conversion efficiency is almost certainly sufficient to allow this to happen.

Is this safe? Lord no. Any place remotely near to the active fusion rector will be red-hot in terms of temperature and radiation. I don’t mean within the structures, I mean we will pump reactor-grade fluids straight through the radiator, and the nearest 1,000s of km are probably going to be a no-go zone for spacecraft. If you go there, your propellant tanks will heat and pop. Even from Earth, this radiator will be an ominous red dot. This is a fusion-reactor-of-death, designed to pack a Type I civilization into a tight, neat, little package. This is the major-leagues of being a space faring civilization. You have torch ships? We have torch power plants, you shouldn’t have expected any less.

For Type 1, producing as much power as sunlight is incident on Earth, you would need 1 million square km. That’s the size of Bolivia. Would you rather use an entire planet for your heat production? Or Bolivia?

Maybe someone will get mad at this for being inefficient, meaning, it uses too much Deuterium inefficiently. Well firstly, it’s not that inefficient. We already have lots of ~30% efficient thermal plants. Direct conversion can go higher, I don’t see the problem. And also, it’s Deuterium, what do you care? We’re not going to run out of it, and by the time it matters, we’ll have even bigger reactors and will fuse more isotopes. We’ll eventually work our way up the neucleosysnthesis chain.

Does this have a claim to logically fit within the historical arc of energy development and use by people? I recently saw a book on the history of energy. It is called “More and More and More”. And yes, that is basically what this is. The ideas I am giving here are grotesque. But our use of the environment today would probably be called that by someone living in the 1700s.

But let me re-echo the relative terms involved here.

• 29.9 km² — one 1 TWe FLiBe radiator, about half of San Marino (61.2 km²); if you flatten it into a disc, the disc diameter is about 6.17 km, which is about 0.174x the 35.37 km tube length, while the disc area is about 54x the tube’s side area.
• 9.45 km² — one 1 TWe tin radiator, about one-sixth of San Marino; flattened disc diameter is about 3.47 km, or 0.098x the 35.37 km tube length, and the disc area is about 17x the tube’s side area.
• 2,449 km² — one 1 TWe space-solar array at 30% collection efficiency, about Luxembourg (2,586 km²).
• 5.18 million km² — combined FLiBe radiator area for the Type I benchmark of 173,549 plants, about 1.6 Indias (3,287,263 km² each).
• 1.64 million km² — combined tin radiator area for the same Type I benchmark, about Iran (1,648,195 km²).
• 1 Earth face-on for Type 1 energy from a solar array, by definition.

The reactor plus associated systems are very very clearly on the ~ 1 km scale. This is big, but it’s also a lot smaller than the entire nation of Bolivia. So physically, the radiator still dominates. Some other designs suggested a first-wall fluid of Tin, which has a boiling point around 2600 C. This could get us another size reduction of 30x or more. It would also put us in the “Warm Candlelight” category if you classified the radiator as a light bulb. Except we’re not illuminating a bedroom, but a large fraction of the around around the Earth-moon system. Maybe this is just a branding problem, where we need attitude corrections to sometimes direct the thermal output towards Earth and call it a night light.

Comparisons to other Energy Sources

Do I think this would win over an implosion type plant? On the one hand, you don’t need the giant coils for the implosion plant… but you do need the giant lasers. Also, this has potential for direct conversion, which is a very big deal. I’m going to say this wins over implosion fusion. I don’t think there’s any contest. On Earth, we have some legitimate doubt about the viability of a 1 GWe plant. Those doubts probably don’t exist, not in the same way, for a 1 TWe plant.

What about solar panels? Oof, it does not look good. Let’s bring those mass intensity numbers back up.

Advanced solar panels are probably in the 100–500 W/kg range. Ouch! The coils and structural mass alone will cost as much mass as solar panels. If that wasn’t bad enough, Neodymium is used in the current superconductors we use. Either we have to develop superconductors free of rare Earths, or 1 TWe reactor will use several-years worth of current production. And asteroids and the moon don’t look good for alternative sources.

So direct head-to-head… even for a Type 1 human civilization, I’m going to have to call it, I am disappointed to say that solar beats this fusion reactor.

I wanted to come up with circumstances where this fusion plant wins. But boy is it hard. The obvious argument is that you need a concentrated power source. This seems valid, but remember we’re looking at maybe ~50% efficiency, and a red-hot radiator for the plant. You can’t have a red-hot radiator for your habitats because you would have to run a thermal cycle, and all of a sudden your power needs have exploded. Maybe you still do that?! Assuming you don’t, your radiator is roughly the mass of your solar array (ok, within a factor of… 3). If you’ve got the area for that radiator, it’s silly to say you don’t have the area for a solar array.

The one exception I can think of comes from the designs I’ve seen for orbital data centers. Classic computronium. Those (even now) propose a 90° C radiator because the computer chips can still operate at these temperatures. Because of this, they have large solar arrays and comparatively small radiators. Given this, fusion can win on the basis of area-efficiency. However, even if you compare 90° C to 20° C, that’s still only an area multiple of about 2.4 for radiator size. But, there’s also the solar power collection efficiency on the front-end, so maybe that adds another factor of 3. I think we’re looking at an area ratio of maybe 5-ish in favor of fusion. This is still not great, but nonetheless it is an arguable conditional win for fusion over solar for heavy industry where any human habitations would need a heat pump to stay alive. This requires dramatic assumptions, and I simply do not believe those conditions be possible on the TW scale. I think Type I energy levels would have to be involved, putting it in the PW scale. That doesn’t necessarily require PW scale individual fusion reactors, but as we’ve seen, larger scale will help the plasma physics.

What about batteries? I can see fusion being superior to batteries. Because those might give several 100s of Wh/kg, and if fusion can get >1,000 W/kg, then it seems pointless to bother with storage.

Of course, the outer solar system makes this viable. Easily.

Final Lukewarm Verdict

Should you change career to fusion power in an anticipatory move? I’m going to say “no”. It’s really hard for me to see this as the mainstream power source, even assuming radical technological progress, space-faring transition, etc. But I’ll be more specific here.

The superconducting coil mass required is the main thing that waters down my enthusiasm. The larger plant size helps this factor, but not enough, even at the 1 TWe scale. Because it still fails to answer the “why not just use solar?” question. What might change this is breakthroughs in superconductors. If it gets vastly easier to manufacture superconducting coils (with more ordinary materials), then I will change my position.

To be fair, this is only because solar power is outrageously good in this context.

For the other question — does radical scaleup mostly fix the teething problems we can predict today for fusion power? I’m going with “yes”. The ideal unit output for a fusion power plant is vastly larger than anything that we might reasonably fit into our current industrial civilization. If you assume we can build at that scale, it looks like a pretty solid, reliable, continuous, power source. It’s just not better than solar at 1 AU and closer, because nothing is.

Fusion is still valid as a minor power source in an advanced civilization. To whatever extent you need power diversity (which is debatable), steady-state plasma fusion will be able to play second fiddle. It’s not there now, but most of the reasons for this can be remedied by greater scale. Additionally there are cases where fusion can contextually shift to the main power source — those being high temp & high density industry, and far-from-sun large-scale civilizations. I am fairly well convinced these would use giant fusion reactors.

What does this mean for the present? I started on the thesis that fusion engineers intuitively understood that the pains the face now will be eliminated with greater scale (and possibly shortcut with better confinement technology), one day. I have a mixed verdict on that. My key points are that:

• Per-unit-output needs for coil & structural mass do not offer any near-term scaling benefit. It’s marginal for < 10 MWe, if it’s anything. Magnetic field must remain at the bleeding edge of what’s possible, and will be fighting with heat removal and breeding systems for the foreseeable future.
• Confinement challenges are absolutely lessened with increasing scale, both near and long term.

It will be exciting to see confinement for Q>1 come into fruition in our lifetimes. I believe it will, but this comes from scaling R up, scaling B up, and the brutally hard work of plasma physicists. After that, a new phase will happen where there’s less technological movement, and it’s more apparent that the rest of the road is boring scaling of what is already demonstrated. We’re not there yet.

I wanted to conclude with some idea of how the long-view of the future reactors should influence the avenues of approach for current reactors. But I think the answer is that normal fusion plasma physics is the ticket. I would be the least original person ever to tell you that General Fusion isn’t a good idea. No, it doesn’t scale well, but it also wasn’t a good idea before that. Even First Light Fusion looks like a dead-end, even if it works. Because unless your fusion can eventually evolve into direction conversion, it’s only a transitional solution.