Fusion needs tens of millions of degrees just to strike a spark. Fission asks for something almost mundane by comparison: enough of the right atoms, packed close enough together. But "enough" has a hard floor — and small modular reactors are, at bottom, a bet on how close to that floor you can actually build.
Every fission reactor ever built, from a 1940s pile under a Chicago football stadium to the four-loop plants running today, is solving the same problem: get enough fissile atoms into a small enough space, arranged the right way, that each fission event reliably triggers another. Get the arrangement wrong — too little material, too dilute, too much surface area relative to volume — and the chain reaction sputters out no matter how good your fuel is. Get it right, and you have a reactor. The size of "right" is not a design choice; it's a property of the atom you picked, and it sets a floor under how small you can build.
Small modular reactors (SMRs) — and their even smaller cousins, microreactors — are, at the time of writing, the most talked-about corner of nuclear energy: dozens of vendors, government demonstration programs, and a wave of commercial interest tied to data-center power demand. Nearly all of that conversation is about economics, licensing, and deployment timelines. Underneath it sits a physics question that gets asked far less often: how small can a fission reactor actually go, and what has to be true for it to still work?
In the introduction we saw that a fission chain reaction is an accounting exercise for neutrons: it sustains itself when the multiplication factor k equals one, whether the neutrons doing the work are slow (thermal) or fast. Here we ask what sets the size of a core that can reach that point.
Two things push k down from its best-case value. The first is absorption: some neutrons get captured by non-fuel material, or by fuel nuclei without causing fission, instead of going on to cause a new one. The second, and the one that actually sets a size floor, is leakage: neutrons that simply escape the surface of the assembly before they hit anything at all. A bigger, more compact core has proportionally less surface for its volume, so it leaks a smaller fraction of its neutrons — which is exactly why there's a minimum size below which even perfect, undiluted fuel can't sustain a chain reaction.
That minimum size is the critical mass — and, for a given shape, the equivalent critical radius. For a bare sphere of fissile material, elementary neutron diffusion theory gives an estimate of the form:
The corresponding critical mass follows from the sphere's volume: Mc = ρ · (4/3)πRc3. Wrap a reflector around the sphere — a shell of material that scatters escaping neutrons back inward instead of losing them — and both numbers drop further; the same trick that shrinks a nuclear device also shrinks a reactor core, and it matters more, not less, at small scale, since a small core already has a large surface-to-volume ratio working against it. Dilute the fuel into a moderated lattice, as ordinary power reactors do, and the numbers move the other way, trading compactness for a more forgiving, thermal-neutron-driven chain reaction.
That formula is an order-of-magnitude estimate; real critical masses and sizes differ nuclide by nuclide, and not always in the direction intuition suggests. The two panels below show representative bare-sphere critical diameters and critical masses across a set of fissile and near-fissile actinides — the classic reactor and weapons-relevant isotopes (U-233, U-235, Pu-239) alongside less common neighbors (Np, Cm, Cf, Es).
Two things are worth pulling out of that pair of charts. First, moving away from the familiar U-235/Pu-239 pair generally costs you: critical size and mass climb toward the heavier, more neutron-rich actinides, as fission cross-sections fall and spontaneous-fission background rises. Second — and this is the more useful, less obvious point — diameter and mass are not two views of the same trend. A sphere's mass scales with the cube of its radius, but also directly with density; two nuclides can land at nearly the same diameter and still need noticeably different masses, because one happens to pack more atoms into the same volume. Critical size, in other words, is a property of the specific material, not a single number you can read off a chart of atomic mass alone — which is exactly why choosing a fuel cycle is a physics decision, not just a supply-chain one.
Everything so far assumed the most compact arrangement possible: pure metal, no moderator, neutrons that stay at their velocity depending on the model assumed. For a real reactor we choose splitting the SMR landscape in two.
In a thermal reactor a moderator, usually water or graphite, slows the neutrons down to thermal energies, where uranium-235 captures them far more readily, so low-enriched fuel is enough (3–5 %, or up to 19.75 % for the smallest designs).
In a fast reactor there is no moderator, and the chain is sustained by fast neutrons on fuel enriched to about 15–20 %, or on plutonium-bearing fuel.
Does that choice change how big the core must be, or how much fuel it holds?
Fig. 2 puts the two routes on the same axes for enrichments from 3 to 20 %, using coarse diffusion estimates that have been fitted to measured critical systems (see the box below for the references). For the thermal route the curves describe a uniform mixture of uranium and water; for the fast route, a sphere of uranium metal. "Reflected" means surrounded by a thick water reflector (thermal) or by 20 cm of depleted uranium (fast); "bare" means no reflector.
Size first. At 20 % enrichment the two critical diameters are almost the same: about 27–35 cm for a water-reflected thermal sphere and about 33 cm for a fast sphere with a depleted-uranium reflector. The gap opens at lower enrichment, where a fast core must grow quickly (about 50 cm at 10 %, against 30–37 cm for the thermal one). The moderator adds volume around the fuel, but a thermal system uses each neutron so much more effectively that the overall geometry ends up comparable, and a reflector shrinks both. Over the range that matters for compact designs, the critical diameter is not what separates the two routes.
Mass is where they differ. A thermal system goes critical with a few kilograms of 235U (about 1–3.5 kg at 10–20 % when reflected, depending on how much water is mixed in), because the moderator makes every atom of fuel far more effective. A fast system needs one to two orders of magnitude more: about 70 kg of 235U at 20 % enrichment with a reflector, and nearly twice that at 10 %, rising steeply as enrichment falls. The measured FCA-I assembly, a 20 % enriched fast core with a natural-uranium blanket, needed 95 kg. That is why fast cores are loaded with highly enriched or plutonium-bearing fuel, and why their fissile inventory is a central design and cost question.
So the choice between thermal and fast barely changes the minimum size at high enrichment, but it changes the fuel inventory by a factor of twenty or more, and it decides which enrichments are workable at all: a thermal core is comfortable at 3–5 %, a fast core is not.
Fig. 3 then asks what surrounds that critical core. The critical core is only the innermost shell: reflector, pressure vessel and shielding add far more weight and size than the fuel. In the illustrative case shown, the core is about 1 % (thermal) to 8 % (fast) of the mass inside the vessel, and 1 % or less of the total once the shield is included. Real cores are also much larger than critical (the 4-loop PWR core is 3.4 m across and the NuScale core about 1.5 m, Fig. 2, top), because heat removal, burnup and operating margin, not criticality, set the real size. That is the subject of the next episodes.
Two questions come up whenever "small reactor" turns into "should we actually build this": what do you do with the heat, and what do you do with the waste?
On heat: shrinking a reactor core doesn't relax the thermodynamics. Almost every design, SMR or not, still turns fission heat into electricity through a conventional steam (Rankine) or gas (Brayton) cycle, and that cycle's efficiency is capped by the same Carnot limit regardless of how small the core is — going small is an engineering story, not a thermodynamic loophole. What does change at small scale is the opportunity: some Gen-IV concepts, particularly molten-salt and liquid-metal-cooled fast designs, can deliver primary-loop temperatures upward of 1500 °C — several times hotter than a light-water reactor — which opens the door to selling high-value process heat (hydrogen production, steelmaking, chemical synthesis) alongside, or instead of, electricity.
None of this is specific to small reactors — but fuel throughput and waste inventory both scale down with core size, which means an SMR or microreactor fleet built on a closed, reprocessed fuel cycle carries a proportionally smaller version of the same waste story as a large plant, not a new one.
That's the physics floor under "small modular": a hard critical-size limit that varies fuel by fuel, and a heat-and-waste story that changes in degree, not in kind, as the core shrinks. The next step is numerical: composition-dependent criticality, reflector sizing and fuel comparisons for cores in the SMR-to-microreactor range. That work is the subject of a technical paper now in preparation, and the choice of fuel and breeding chain will have its own episode in this series.
Dataphysics Limited works on energy-systems modelling and physics-informed AI. → dataphysics.ae