Could an Iron Man Arc Reactor Work? The Fusion Numbers • The Science 360

In a cave in Afghanistan, Tony Stark builds a glowing disc about the size of a hockey puck and tells Yinsen that, if his math is right, it puts out three gigajoules per second. That is 3 gigawatts. Three large nuclear reactors' worth of electricity, sitting in a hole in a man's sternum, built from scrap.

Could anything like it work? The useful way to answer is not "fusion is hard" (it is) but to ask what the number itself demands. Put 3 GW in a volume of about 126 cm³ and you can work out, with real fusion data, how dense and hot the fuel must be, how strong the magnets must be, how many neutrons come out and how much heat has to go somewhere. Each answer is wrong by a factor between ten and a quarter of a million. The interesting part is which problem is the real killer, because it is not the one most people guess.

Quick answerNo. Three gigawatts from a palm-sized core is about 24 TW per cubic metre, 240,000 times a fission reactor core and millions of times any fusion plasma design. Confining it would need fields of roughly 300 tesla, and even "clean" aneutronic fusion would give the wearer a lethal neutron dose in milliseconds.

Infographic: Could an Iron Man arc reactor work? A glowing 8 cm reactor disc is labelled 3 GW. Headline: its power density would be 240,000 times a fission reactor core, 24 TW per cubic metre. Confining it needs about 300 tesla; neutrons kill in 15 microseconds. A log-scale bar chart of power density runs from the Sun's core at 276 W per cubic metre, the human body at 1.5 kW, ITER at 0.6 MW, SPARC at 7 MW and a fission core at about 100 MW, up to the arc reactor at 24 TW. Tiles: 0.77 kg of fuel a day, 580 kg of water per second, 12 ms to a lethal dose with proton-boron. — Tap the image to view it full size. View the accessible, editable SVG.

What does 3 gigawatts in your chest actually mean?

Start with the most basic number in reactor engineering: power density, the watts produced per cubic metre. I will treat the movie reactor as a disc 8 cm across and 2.5 cm thick, which matches the prop's proportions reasonably well. It is an assumption, but changing it by a factor of two changes nothing that follows.

Power density of the arc reactor

\begin{aligned}V &= \pi \times (0.04\ \mathrm{m})^2 \times 0.025\ \mathrm{m} \\ &= 1.26\times10^{-4}\ \mathrm{m^3}\ (126\ \mathrm{cm^3}) \\ \frac{P}{V} &= \frac{3\times10^{9}\ \mathrm{W}}{1.26\times10^{-4}\ \mathrm{m^3}} \\ &= \mathbf{2.4\times10^{13}\ W/m^3}\ (24\ \mathrm{TW/m^3})\end{aligned}

For comparison: a large pressurised water reactor core runs at about 100 MW/m³ (100 kW per litre), and the SPARC compact tokamak design at about 7 MW/m³.

\begin{aligned}\text{ITER:}\ \frac{500\ \mathrm{MW}}{840\ \mathrm{m^3}} &= 0.6\ \mathrm{MW/m^3} \\ \text{vs fission core:}\ \frac{2.4\times10^{13}}{10^{8}} &= \mathbf{240{,}000}\end{aligned}

Horizontal bar chart of power density on a logarithmic scale from 100 watts to 100 terawatts per cubic metre. The Sun's core is 276 watts per cubic metre, a resting human body about 1.5 kilowatts, the ITER plasma 0.6 megawatts, the SPARC plasma 7 megawatts, a large fission reactor core 100 megawatts, and the arc reactor 24 terawatts per cubic metre, 240,000 times the fission core. — Figure 1. Power per cubic metre on a log scale. Every real fusion device sits below a fission core, because a fusion plasma is a near-vacuum. The arc reactor's 24 TW/m³ is my calculation from the film's 3 GW and an assumed 126 cm³ core. The human body figure is roughly 100 W over 0.066 m³.

Figure 1 holds the first surprise. Fusion is not a high power density process in the places where it happens. The centre of the Sun produces about 276 watts per cubic metre, less than your body does sitting still. The Sun is bright because it is enormous. Magnetic fusion devices on Earth do far better than the Sun, but even the most aggressive compact design, SPARC, aims for about 7 MW/m³, roughly fourteen times less than an ordinary fission core.

So the common framing, "the arc reactor is a miniaturised fusion reactor", misses the scale of the ask. Miniaturising a fusion reactor is not like miniaturising a computer. The arc reactor needs a power density three to four million times SPARC's, from a technology whose plasma is, by design, thinner than air.

How dense and hot would the plasma have to be?

Fusion power density in a 50:50 deuterium-tritium (D-T) plasma has a tidy formula. If n is the total ion density, each cubic metre has n/2 deuterons and n/2 tritons, and the reaction rate depends on how often they collide at high speed, captured in the reactivity ⟨σv⟩:

\frac{P}{V} = \frac{n}{2}\cdot\frac{n}{2}\,\langle\sigma v\rangle E_{\mathrm{fus}} = \frac{n^2\langle\sigma v\rangle E_{\mathrm{fus}}}{4}

E_{\mathrm{fus}}\ (\text{D-T}) = 17.6\ \mathrm{MeV} = 2.82\times10^{-12}\ \mathrm{J}

The reactivity depends steeply on temperature. Figure 2 plots it from the tabulated values in the NRL Plasma Formulary for three fuels.

Log-log plot of fusion reactivity against ion temperature from 1 to 1000 keV for D-T, D-D and D-He3 fuels. D-T rises fastest, peaking near 9 times ten to the minus 22 cubic metres per second around 50 to 100 keV. D-He3 overtakes D-D above about 20 keV and peaks near 2.4 times ten to the minus 22 at 200 keV. A dashed line marks the D-T optimum near 14 keV. — Figure 2. Maxwell-averaged fusion reactivity ⟨σv⟩ from the NRL Plasma Formulary table (dots), joined by log-log interpolation. D-T is easiest by a wide margin; D-He³ needs several times higher temperature for the same rate. Proton-boron (p-¹¹B) is not in the table and is harder still.

Now solve for the density the arc reactor needs, at about the temperature where D-T is easiest to confine (around 14 keV, or 160 million K):

Step 1. Density needed for 24 TW/m³

$\langle\sigma v\rangle$ at 14 keV ≈ 2.1 × 10⁻²² m³/s (interpolated from the table).

\begin{aligned}n &= \sqrt{\frac{4\,P/V}{\langle\sigma v\rangle E_{\mathrm{fus}}}} \\ &= \sqrt{\frac{4 \times 2.4\times10^{13}}{2.1\times10^{-22} \times 2.82\times10^{-12}}} \\ &= \mathbf{4.0\times10^{23}\ ions/m^3}\end{aligned}

SPARC's design density is about 3 × 10²⁰ m⁻³, so this is about 1,300 times denser.

Step 2. Plasma pressure (ions plus an equal number of electrons)

\begin{aligned}p &= 2nkT \\ &= 2 \times 4.0\times10^{23} \\ &\quad \times 14\times10^{3}\ \mathrm{eV} \\ &\quad \times 1.602\times10^{-19}\ \tfrac{\mathrm{J}}{\mathrm{eV}} \\ &= \mathbf{1.8\times10^{9}\ Pa}\end{aligned}

That is about 18,000 atmospheres.

Eighteen thousand atmospheres of gas at 160 million kelvin. No material wall can touch that, which is why tokamaks hold plasma with magnetic fields. The field's ability to push is its magnetic pressure, $B^2/2\mu_0$ , and the ratio of plasma pressure to magnetic pressure is called beta ( $\beta$ ). Tokamaks run at $\beta$ of a few percent. So:

Step 3. Field needed to hold it

B = \sqrt{\frac{2\mu_0 p}{\beta}}

\begin{aligned}B_{\beta=1} &= \sqrt{2 \times 1.257\times10^{-6}} \\ &\quad \times \sqrt{1.8\times10^{9}} \\ &= \mathbf{67\ T} \\ B_{\beta=0.05} &= \frac{67}{\sqrt{0.05}} = \mathbf{300\ T}\end{aligned}

The first line is the ideal limit; the second is a realistic tokamak. Strongest continuous lab magnet: about 45 T. ITER's peak coil field: 11.8 T.

Log-log plot of required magnetic field against ion temperature from 5 to 300 keV. For D-T at beta 1 the minimum is about 67 tesla near 10 to 20 keV; at beta 0.05 it is about 300 tesla. For D-He3 the requirements are about 200 tesla at beta 1 and 900 tesla at beta 0.05. Horizontal lines mark the strongest continuous magnet at 45 tesla and ITER's 11.8 tesla. — Figure 3. Field needed to confine a plasma producing 24 TW/m³, computed from the reactivity data in Figure 2. Even the impossible β = 1 case is above the world's strongest continuous magnet. Switching to D-He³ to cut neutrons roughly triples the field.

There is a shortcut that gives the same answer and shows why bigger magnets are the whole game in fusion. At fixed β and temperature, fusion power density scales as B⁴. That is why SPARC and MIT's ARC power plant design use high-temperature superconducting magnets: doubling the field buys sixteen times the power density. Run it backwards from SPARC:

\begin{aligned}\text{density gain} &= \frac{2.4\times10^{13}}{7\times10^{6}} \\ &= 3.4\times10^{6} \\ \text{field gain} &= (3.4\times10^{6})^{1/4} = 43 \\ B &= 12.2\ \mathrm{T} \times 43 \approx \mathbf{520\ T}\end{aligned}

That is at SPARC's $\beta$ .

Two different routes, a few hundred tesla either way. Magnets of that strength would tear themselves apart: magnetic pressure at 300 T is $B^2/2\mu_0 = 3.6\times10^{10}$ Pa, about 350,000 atmospheres, beyond what any known conductor and support structure can hold for long. Fields of hundreds of tesla do exist briefly in labs, in explosively driven or single-turn coils that destroy themselves in microseconds.

The real killer: neutrons

Suppose Stark solved confinement. In D-T fusion, each reaction makes a 3.5 MeV helium nucleus and a 14.1 MeV neutron. The neutron carries 80% of the energy and, having no charge, ignores every magnetic field you could build.

Neutron output at 3 GW

\begin{aligned}\text{reactions/s} &= \frac{3\times10^{9}\ \mathrm{W}}{2.82\times10^{-12}\ \mathrm{J}} \\ &= 1.06\times10^{21}\ \mathrm{s^{-1}} \\ P_{n} &= 0.8 \times 3\ \mathrm{GW} = \mathbf{2.4\ GW}\end{aligned}

That is 10²¹ neutrons per second, about 150 GW/m² through the reactor's surface (about 0.016 m²).

\begin{aligned}\text{fuel} &= 1.06\times10^{21} \times 5.03\ \mathrm{u} \\ &\quad \times 1.66\times10^{-27}\ \mathrm{kg/u} \\ &= 8.9\ \mathrm{\mu g/s} \approx 0.77\ \mathrm{kg/day}\end{aligned}

Power plant designs worry about neutron wall loads of a few megawatts per square metre. This is tens of thousands of times more, and the wall in question is Tony Stark.

Here is a deliberately generous estimate. Assume only 1% of the neutron energy is absorbed by his body (most of the rest flies off in other directions or passes through), take a body mass of 70 kg, and use 5 gray (joules per kilogram) as a lethal whole-body dose.

Time to a lethal dose, D-T

\begin{aligned}P_{\mathrm{abs}} &= 0.01 \times 2.4\times10^{9}\ \mathrm{W} = 2.4\times10^{7}\ \mathrm{W} \\ \dot D &= \frac{2.4\times10^{7}\ \mathrm{W}}{70\ \mathrm{kg}} = 3.4\times10^{5}\ \mathrm{Gy/s} \\ t_{5\,\mathrm{Gy}} &= \frac{5}{3.4\times10^{5}} = \mathbf{15\ \mu s}\end{aligned}

This is the misconception most "is the arc reactor real?" articles skip. They focus on whether fusion can be miniaturised, as though a tiny reactor that worked would be fine to wear. The opposite is true: the better it worked, the faster it would kill you. Shielding does not save it either. Fast neutrons are typically attenuated by about a factor of e for every 10 cm or so of water or concrete (a rough figure that depends on material and energy), so cutting 10²¹ neutrons per second to a tolerable level takes metres of shielding, which is exactly what surrounds every real reactor.

Would aneutronic fusion fix it?

This is the best real-physics defence of the arc reactor, and it deserves a fair hearing. Some fusion reactions release their energy mostly as charged particles, which a magnetic field can steer and which could in principle be converted directly to electricity without a steam turbine. The two usual candidates:

Deuterium-helium-3: D + ³He → ⁴He (3.6 MeV) + p (14.7 MeV). The main reaction makes no neutrons, but D-D side reactions in the same plasma do, leaving roughly 5% of the energy in neutrons. Helium-3 is also extremely scarce on Earth.
Proton-boron-11: p + ¹¹B → 3 ⁴He + 8.7 MeV. Side reactions put roughly 0.1% of the energy into neutrons. The catch is that its Lawson requirement (the density, temperature and confinement time product needed for net energy) is about 500 times harder than D-T's.

Log-log plot of time to a lethal 5 gray dose against the fraction of 3 gigawatts emitted as neutrons. D-T at 80 percent neutrons gives 15 microseconds, D-He3 at 5 percent gives 0.23 milliseconds, and proton-boron at 0.1 percent gives 12 milliseconds, all far shorter than a 0.3 second blink. — Figure 4. Time to a 5 Gy whole-body dose versus neutron fraction, assuming 1% of the neutron power is absorbed by a 70 kg wearer of a 3 GW reactor. My estimate; real absorption depends on geometry and neutron energy, but even a factor of 100 error leaves every fuel lethal within seconds.

Figure 4 is the honest outcome. Aneutronic fuel moves the time to a lethal dose from 15 microseconds to about 12 milliseconds. That is a factor of 800 improvement and still less than the time it takes to blink. There is also a second radiation problem the neutron fraction hides: a plasma hot enough to burn boron (hundreds of keV) radiates intense X-rays as electrons swerve past nuclei, and at those temperatures that X-ray loss is comparable to the fusion power itself. That is the main reason p-¹¹B is so hard, and every one of those X-rays would also be heading into Stark's chest.

Where would 3 gigawatts of heat go?

Even a perfect reactor turns its output into heat somewhere. Assume a heroic 50% conversion to electricity inside the reactor, leaving 1.5 GW to remove from the device itself. Two ways to get rid of it:

Option A: radiate it away (Stefan-Boltzmann law, emissivity 1)

\begin{aligned}T &= \left(\frac{P}{\sigma A}\right)^{1/4} \\ \sigma &= 5.67\times10^{-8}\ \mathrm{W/(m^2\,K^4)}\end{aligned}

For the reactor face alone, $A = 0.005\ \mathrm{m^2}$ :

\begin{aligned}\sigma A &= 5.67\times10^{-8} \times 0.005 \\ &= 2.84\times10^{-10} \\ T &= \left(\frac{1.5\times10^{9}}{2.84\times10^{-10}}\right)^{1/4} \\ &= \mathbf{48{,}000\ K}\end{aligned}

Spread over the whole suit, $A \approx 2\ \mathrm{m^2}$ , it is still 10,700 K, nearly twice the surface of the Sun.

Option B: boil water

\begin{aligned}\text{per kg} &= 4{,}186 \times 75 + 2.26\times10^{6} \\ &\approx 2.6\times10^{6}\ \mathrm{J/kg} \\ \text{water} &= \frac{1.5\times10^{9}}{2.6\times10^{6}} = \mathbf{580\ kg/s}\end{aligned}

That covers heating from 25 °C plus boiling.

Log-log plot of surface temperature against radiating area for 1.5 gigawatts of waste heat. The reactor face of 0.005 square metres would need 48,000 kelvin, a whole suit of 2 square metres 10,700 kelvin, and a hectare of radiator 1,100 kelvin. Reference lines mark the Sun's surface at 5,772 kelvin and tungsten's melting point at 3,695 kelvin. — Figure 5. Radiating 1.5 GW: temperature needed for a given area. To stay below tungsten's melting point the suit would need a radiator of over 100 m²; to be merely red hot, about a hectare. Computed from the Stefan-Boltzmann law.

Your body, for scale, sheds about 100 W at rest and perhaps a kilowatt or so in hard exercise. Real power plants dump gigawatts into rivers, seas or cooling towers the size of cathedrals for this reason.

Is it a power problem or an energy problem?

Yinsen also says the reactor could run Stark's heart "for fifty lifetimes". That claim can only be true if the reactor almost never runs at 3 GW. Fifty 80-year lifetimes at full power is 3.8 × 10²⁰ J. D-T fusion converts about 0.38% of the fuel's mass into energy, so that output would burn about 1.1 million kilograms of fuel. The device must hold a tiny load (the chest electromagnet, watts at most) almost all the time and only surge to gigawatts for flight and fighting. That is a battery's job description, not a power plant's, and it is a fair reading of how the films use it.

So what would a physicist's arc reactor look like? Fusion fuel is wonderfully compact: at 3 GW you burn under a kilogram a day. The bulk lives in everything else: magnets to confine the plasma, metres of shielding against neutrons, and radiators or cooling water to carry away the heat. That is why the most compact real fusion power plant concept, MIT's ARC, has a major radius of 3.3 m and a 9.2 T superconducting field to produce 200 to 250 MW of electricity. Compact for a power station is the size of a building.

The scorecard: power density 240,000 times a fission core; confining field about 300 T against a 45 T world record; lethal neutron dose in 15 µs for D-T and about 12 ms even for proton-boron; 1.5 GW of waste heat needing a hectare of radiator or 580 kg of water per second. The fuel is the only part that fits in a chest.

FAQ

Is the arc reactor possible in real life?

Not as shown. Producing 3 gigawatts in a palm-sized volume means a power density about 240,000 times that of a commercial fission core and millions of times that of any fusion plasma design. Confining such a plasma would need a magnetic field of several hundred tesla, and the neutrons and waste heat would kill the wearer almost instantly.

How much power does Iron Man’s arc reactor produce?

In Iron Man (2008) Tony Stark says it produces three gigajoules per second, which is 3 gigawatts. That is about the electrical output of three large nuclear power reactors.

Is the arc reactor based on real fusion?

Loosely. Its ring shape echoes a tokamak, the doughnut-shaped magnetic fusion device used by ITER and the compact SPARC design. Real tokamaks are building-sized because confining a fusion plasma takes a large volume and strong magnets, and their power density is a few megawatts per cubic metre.

Would aneutronic fusion make an arc reactor safe?

Safer, not safe. Proton-boron fusion releases only about 0.1 percent of its energy as neutrons through side reactions, but 0.1 percent of 3 gigawatts is still 3 megawatts of neutrons. On a rough estimate that delivers a lethal dose to a person wearing it in about a hundredth of a second, and the fuel needs conditions roughly 500 times harder to reach than deuterium-tritium.

What is MIT’s ARC reactor?

ARC is a 2015 design study for a compact, high-field tokamak power plant: a major radius of 3.3 metres, a 9.2 tesla field from high-temperature superconducting magnets, and about 200 to 250 megawatts of net electricity. Compact for a power station, it is still the size of a building.

Sources

Iron Man Wiki, Arc Reactor (3 GJ/s, "fifty lifetimes" line attributed to Yinsen): ironman.fandom.com/wiki/Arc_Reactor
Inverse, "The MCU's first smash hit gets one thing right about nuclear fusion" (tokamak inspiration): inverse.com
Wikipedia, "Nuclear fusion" (NRL Plasma Formulary reactivity table, reaction energies, neutronicity, relative Lawson criteria): en.wikipedia.org/wiki/Nuclear_fusion
Wikipedia, "ITER" (500 MW fusion power, 840 m³ plasma volume, 11.8 T peak coil field): en.wikipedia.org/wiki/ITER
A. J. Creely et al., "Overview of the SPARC tokamak", Journal of Plasma Physics 86 (2020) (12.2 T, R = 1.85 m, P_fusion ≈ 140 MW, ⟨n_e⟩ ≈ 3 × 10²⁰ m⁻³, P/V ≈ 7 MW/m³): impact.ornl.gov
B. N. Sorbom et al., "ARC: A compact, high-field, fusion nuclear science facility and demonstration power plant with demountable magnets" (2015): arxiv.org/abs/1409.3540
Pressurised water reactor core power density (about 100 kW per litre), Taylor and Francis topic summary: taylorandfrancis.com
Wikipedia, "Sun" (core power density 276.5 W/m³): en.wikipedia.org/wiki/Sun
National MagLab, hybrid magnet and 45 T continuous field record: nationalmaglab.org and nationalmaglab.org (45.5 T test coil)

Method note: the reactor volume (8 cm by 2.5 cm), the 50% conversion efficiency, the 1% body absorption and the 10 cm neutron attenuation length are my assumptions, chosen to be generous to the reactor. Reactivities are interpolated from the NRL Plasma Formulary table.

Header image: NASA/JPL-Caltech (image PIA23879), regraded for this article.

What does 3 gigawatts in your chest actually mean?

How dense and hot would the plasma have to be?

The real killer: neutrons

Would aneutronic fusion fix it?

Where would 3 gigawatts of heat go?

Is it a power problem or an energy problem?

FAQ

Is the arc reactor possible in real life?

How much power does Iron Man’s arc reactor produce?

Is the arc reactor based on real fusion?

Would aneutronic fusion make an arc reactor safe?

What is MIT’s ARC reactor?

Related on The Science 360

Sources