NUCLEAR REACTOR
The Opening
Hold your hand in front of your face. The atoms in your skin are mostly empty space. If an atom were a football stadium, the nucleus would be a marble on the 50-yard line. The electrons would be gnats orbiting in the parking lot.
All the mass is in that marble. All the energy holding the marble together is staggering. And for 13.8 billion years, nobody figured out how to crack it open.
Then, in December 1942, under the bleachers of a squash court in Chicago, a team led by Enrico Fermi stacked 45,000 blocks of graphite into a pile, threaded uranium through it, and slowly withdrew a cadmium rod. A neutron counter began clicking faster. And faster.
They had built a machine that:
├── Splits atoms on command
├── Releases 50 million times more energy per reaction than fire
├── Sustains itself — each split triggering the next
├── Can be throttled from a trickle to a torrent
├── Runs for 18 months on a single fuel load
├── Produces zero carbon
└── Can melt through its own floor if you lose control
A nuclear reactor is the most energy-dense machine humanity has ever built. And the most consequential. Get it right, and you power a city for decades on a truckload of fuel. Get it wrong, and you render a city uninhabitable for centuries.
Let's build one.
───
PHASE 1: Split an Atom
You want energy. The question is: where does energy live inside matter?
Pick up a lump of coal. Burn it. You get heat — about 4 electron volts (eV) per chemical reaction. That's the energy stored in the bonds between atoms. Electrons rearrange, carbon meets oxygen, CO₂ forms, and you get warmth.
Now look deeper. Past the electrons. Past the electron cloud. Down into the nucleus itself — protons and neutrons packed together in a space 100,000 times smaller than the atom.
The force holding those protons and neutrons together is the strong nuclear force. It's the strongest force in the universe — 100 times stronger than electromagnetism, 10³⁸ times stronger than gravity. But it only works at nuclear distances: about 10⁻¹⁵ meters. Step one femtometer too far and it vanishes.
The energy locked in those nuclear bonds dwarfs chemical bonds the way the Sun dwarfs a match.
The binding energy curve — where the energy actually is
Every nucleus pays a price to stay together. Protons repel each other (positive charges), but the strong force glues them anyway. The difference between "what the parts weigh separately" and "what the assembled nucleus weighs" is the binding energy.
More binding energy per nucleon = more tightly bound = more stable.
Binding energy
per nucleon (MeV)
│
8.8 ─┤ ● Fe-56 (PEAK: 8.79 MeV)
│ ● Cr ● Ni
8.5 ─┤ ● Si ● Zn
│ ● O ● Kr
8.0 ─┤ ● C ● Ba-141
│
7.5 ─┤ ● Li ● Xe
│ ● U-235 (7.59 MeV)
7.0 ─┤● He ● U-238
│
│
1.1 ─┤ ● H-2
└─────┬─────┬─────┬─────┬─────┬─────┬─────→
20 60 100 140 180 220 240
Atomic mass number (A)Iron-56 sits at the peak. Nuclei lighter than iron release energy by FUSING (moving right, uphill). Nuclei heavier than iron release energy by SPLITTING (moving left, uphill). This is why stars fuse hydrogen into iron and stop — and why we split uranium.
Iron is the ash of the universe. You can't extract nuclear energy from iron. It's the bottom of the energy hill in both directions. Stars spend millions of years fusing hydrogen toward iron. We split uranium back toward iron. Both paths release energy because both paths climb toward that peak.
This is the same curve that governs stellar nucleosynthesis — the reason heavy elements exist at all.
Split uranium-235: the mass defect calculation
Here's the reaction. A slow neutron hits a uranium-235 nucleus. The uranium absorbs it, becomes uranium-236 for a fleeting instant, then tears itself apart:
n + ²³⁵U → ²³⁶U* → ¹⁴¹Ba + ⁹²Kr + 3n
Now derive the energy. We need the masses:
BEFORE:
├── 1 neutron: 1.008665 amu
├── U-235 nucleus: 235.043930 amu
└── Total: 236.052595 amu
AFTER:
├── Ba-141: 140.914411 amu
├── Kr-92: 91.926156 amu
├── 3 neutrons: 3.025995 amu (3 × 1.008665)
└── Total: 235.866562 amu
MASS DEFECT:
Δm = 236.052595 − 235.866562 = 0.186033 amu
Where did the mass go? It became energy.
E = Δm × c²
E = 0.186033 × 931.5 MeV/amu
E = 173.3 MeV
Add gamma rays and kinetic energy of neutrons: ~200 MeV total0.186 atomic mass units vanish. They don't go anywhere — they convert directly into kinetic energy of the fragments, neutron kinetic energy, and gamma radiation. Einstein's E = mc² is not a theoretical curiosity here. It is the operating principle of the machine.
200 MeV. That's per single fission event. One atom splitting.
How much is that? Let's build a comparison ladder.
The comparison ladder: nuclear vs chemical
Burning coal — one carbon atom meets one oxygen molecule:
C + O₂ → CO₂ + 4.1 eV
Splitting one uranium atom:
n + U-235 → Ba + Kr + 3n + 200,000,000 eV
Reaction Energy released
─────────────────────────────────────────────
Burning coal (C + O₂) 4.1 eV
Burning hydrogen (H₂ + O₂) 2.96 eV ← best chemical fuel
TNT explosion 4.6 eV
Fission of U-235 200 MeV ← 50 MILLION × coal
Fusion of D + T 17.6 MeV ← per reaction (lighter fuel)
Per kilogram of fuel:
─────────────────────────────────────────────
Coal: 24 MJ/kg ← one hot bath
Gasoline: 46 MJ/kg ← drives a car 15 km
TNT: 4.6 MJ/kg ← less than butter
Uranium-235: 82,000,000 MJ/kg ← powers a city for a day
Ratio: 1 kg of uranium = 3,400 TONS of coalThis is not an incremental improvement. It's not "10% better" or "twice as good." Nuclear fission releases 50 million times more energy per reaction than burning carbon. One fuel pellet the size of your fingertip — about 7 grams — contains as much energy as 1 ton of coal or 480 liters of oil.
One reactor fuel pellet. The size of a pencil eraser. Holds the energy of a truckload of coal.
This is why nuclear power exists. Not because it's elegant. Not because physicists wanted to play with atoms. Because the energy density is so absurdly, incomprehensibly high that ignoring it would be insane.
DESIGN SPEC UPDATED:
├── Energy source: strong nuclear force (binding energy)
├── Reaction: n + U-235 → Ba-141 + Kr-92 + 3n + 200 MeV
├── Mass defect: 0.186 amu converts to energy via E = mc²
├── Energy per fission: 200 MeV (50 million × chemical burning)
├── Energy density: 82 TJ/kg (vs 24 MJ/kg for coal)
└── Iron-56 is the peak — fission moves heavy nuclei toward it
───
PHASE 2: Start a Chain Reaction
You split one atom and got 200 MeV. Impressive — but useless. One atom is nothing. You need to split trillions per second.
Here's the gift that uranium gives you: each fission releases 2 to 3 free neutrons. Those neutrons can hit OTHER uranium atoms. Which split. Which release MORE neutrons. Which hit MORE uranium.
This is a chain reaction. And it changes everything.
Without a chain reaction, you'd need to fire neutrons into uranium from an external source — one by one, like feeding coins into a vending machine. Pathetically slow. With a chain reaction, the process feeds itself. One fission triggers two. Two trigger four. Four trigger eight.
The question isn't whether it works. The question is whether you can STOP it.
The multiplication factor k — the single most important number
Define k as the effective multiplication factor: the average number of neutrons from one fission that go on to cause another fission.
k < 1 SUBCRITICAL Each generation has fewer fissions.
│ The reaction dies out.
│ Pull out fuel rods → reactor shuts down.
│
k = 1 CRITICAL Each generation has exactly as many fissions.
│ Steady state. Constant power output.
│ THIS IS WHERE A REACTOR OPERATES.
│
k > 1 SUPERCRITICAL Each generation has MORE fissions.
│ Power increases exponentially.
│ Controlled: used during startup to raise power.
│ Uncontrolled: a bomb.
Generation: 1 2 3 4 5 6 7 8
──────────────────────────────────────────────────
k = 0.9: 1 0.9 0.81 0.73 0.66 0.59 0.53 0.48 → dying
k = 1.0: 1 1 1 1 1 1 1 1 → steady
k = 1.001: 1 1.00 1.00 1.00 1.01 1.01 1.01 1.01 → slow rise
k = 2.0: 1 2 4 8 16 32 64 128 → explosionA reactor operates at k = 1.0000. Not 1.01. Not 0.99. Exactly 1. The control systems adjust k by fractions of a percent. A bomb operates at k ≈ 2 with no control system at all.
The difference between a nuclear reactor and a nuclear bomb is not the fuel, not the physics, not the reaction. It's the value of k. A reactor holds k at exactly 1. A bomb lets k run to 2 or higher.
How fast does a bomb reach full yield?
In a bomb, the neutron generation time is about 10 nanoseconds — the time for a fast neutron to cross a few centimeters of dense metal and hit another uranium nucleus.
Start with 1 fission. Each generation doubles (k ≈ 2). How many generations to release the energy of a 20-kiloton bomb — the yield of the Hiroshima weapon?
Hiroshima yield: 20 kilotons TNT = 8.4 × 10¹³ J
Energy per fission: 200 MeV = 3.2 × 10⁻¹¹ J
Fissions needed: 8.4 × 10¹³ / 3.2 × 10⁻¹¹ = 2.6 × 10²⁴ fissions
With k = 2, after n generations: N = 2ⁿ fissions total
2ⁿ = 2.6 × 10²⁴
n × ln(2) = ln(2.6 × 10²⁴)
n × 0.693 = 56.2
n = 81 generations
Time: 81 × 10 ns = 810 nanoseconds
Less than ONE MICROSECOND from first fission to full yield.81 doublings. That's all it takes to go from a single neutron to the destruction of a city. The chain reaction completes before the bomb has time to blow itself apart. This is why nuclear weapons work — the exponential is faster than the expansion of the material.
81 generations. 810 nanoseconds. In less time than it takes sound to travel one foot, 2.6 × 10²⁴ atoms split and release as much energy as 20,000 tons of TNT.
This is the exponential at its most terrifying. And it's the same exponential a reactor must tame.
Why a reactor can't explode like a bomb
A reactor's fuel is enriched to 3-5% U-235. The rest is U-238, which doesn't sustain a chain reaction with slow neutrons. A bomb needs >80% enrichment.
At 3% enrichment, k can barely reach 1.0 with perfect moderation. There's no physical mechanism to achieve k = 2. Even in a total meltdown — worst case — the geometry changes, moderation degrades, and k drops.
A reactor can melt. It can release radiation. It can contaminate a region. But it cannot produce a nuclear explosion. The physics forbids it. The enrichment is too low. The geometry is wrong.
Chernobyl was a steam explosion and graphite fire — not a nuclear detonation. Hiroshima required 64 kg of 80%-enriched uranium compressed by shaped explosive charges. A reactor has neither the enrichment nor the compression.
DESIGN SPEC UPDATED:
├── Chain reaction: each fission releases 2-3 neutrons → self-sustaining
├── k = 1.0000: critical (steady power) — the operating point
├── k > 1: supercritical (power rising) — startup or bomb
├── k 80%)
└── A reactor cannot produce a nuclear explosion — physics forbids it
───
PHASE 3: Slow the Neutrons Down
You have a chain reaction. But there's a problem — your neutrons are moving too fast to be useful.
When U-235 splits, the neutrons come out fast. Really fast.
Fast neutron energy: ~2 MeV
Fast neutron speed: ~20,000 km/s (6.7% the speed of light)
At these speeds, a neutron screams past a U-235 nucleus without interacting. The fission cross-section — a measure of how likely a neutron is to cause fission — is tiny for fast neutrons.
Cross-section
(barns)
│
600 ─┤ ● ← thermal (0.025 eV): 585 barns
│ ╲
│ ╲
100 ─┤ ╲
│ ╲ resonance peaks
10 ─┤ ╲ ╱╲╱╲╱╲
│ ╲ ╲
1 ─┤ ──── ● fast (2 MeV): ~1.2 barns
│
└──────┬──────┬──────┬──────┬──→
0.025 eV 1 eV 1 keV 2 MeV
Neutron energyAt thermal energy (0.025 eV), a neutron's fission cross-section for U-235 is 585 barns. At fast energy (2 MeV), it's about 1.2 barns. That's a factor of ~500. Slow neutrons are 500 times more likely to split uranium. This single fact drives the entire design of a thermal reactor.
585 barns vs 1.2 barns. Slow a neutron down and it becomes 500 times more likely to cause fission.
You need to slow 2 MeV neutrons to 0.025 eV — thermal energy, the kinetic energy of molecules at room temperature. From 20,000 km/s to 2.2 km/s. A speed reduction of 10,000×.
How to slow a neutron — the billiard ball physics
A neutron has no charge. You can't use electric or magnetic fields. You can't grab it with chemistry. The only way to slow it down is to bounce it off other nuclei, like a billiard ball hitting other billiard balls.
Neutron (mass 1) hits a nucleus (mass A):
Maximum energy transferred per collision:
4A
f = ──────────
(1 + A)²
If A = 1 (hydrogen): f = 4(1)/(1+1)² = 4/4 = 1.00 (100%)
If A = 2 (deuterium): f = 4(2)/(1+2)² = 8/9 = 0.89 (89%)
If A = 12 (carbon): f = 4(12)/(1+12)² = 48/169 = 0.28 (28%)
If A = 238 (uranium): f = 4(238)/(1+238)² = 952/57121 = 0.017 (1.7%)This is the same physics as pool. A cue ball (neutron) transfers the most energy when it hits a ball of equal mass. Hit a bowling ball (uranium) and it barely slows down. Hit a ping-pong ball — impossible, nuclei are heavier. Hydrogen is the closest match to a neutron's mass.
A head-on collision with hydrogen transfers ALL the kinron's energy in one shot. But collisions aren't always head-on. On average, a neutron loses about 63% of its energy per collision with hydrogen.
How many collisions to go from 2 MeV to 0.025 eV?
The collision count — derive it
Each collision with hydrogen reduces energy by a factor of about e⁻¹ on average (the logarithmic energy decrement ξ = 1 for hydrogen).
Number of collisions = ln(E_initial / E_final) / ξ
n = ln(E₀ / E_f) / ξ
E₀ = 2 MeV = 2 × 10⁶ eV
E_f = 0.025 eV
ln(2 × 10⁶ / 0.025) = ln(8 × 10⁷) = 18.2
Moderator A ξ Collisions (n) Absorbs neutrons?
──────────────────────────────────────────────────────────────────
Hydrogen (H₂O) 1 1.000 18 yes (some)
Deuterium (D₂O) 2 0.725 25 almost none
Carbon (graphite) 12 0.158 115 very little
Oxygen 16 0.120 152 very little
Iron 56 0.035 520 absorbs a lot
Uranium 238 0.008 2,172 absorbs everythingHydrogen wins by a landslide: 18 collisions. But ordinary water absorbs some neutrons (hydrogen captures them to form deuterium). Heavy water (D₂O) takes 25 collisions but absorbs almost nothing — so you can use natural uranium. Graphite takes 115 collisions but is solid and doesn't boil away. Each choice defines a fundamentally different reactor design.
18 bounces off hydrogen nuclei. That's all it takes to slow a neutron from 6.7% of light speed to room temperature.
Those 18 collisions happen in about 1 millisecond — the neutron pinballs through water molecules, losing energy with each hit, until it's drifting at thermal speed.
This is why water is the most common moderator. It's cheap, it's everywhere, and it slows neutrons in the fewest collisions. But it comes with a cost — water absorbs some neutrons, which means you need enriched uranium (3-5% U-235) instead of natural uranium (0.7% U-235).
Canada's CANDU reactors use heavy water (D₂O) instead. Deuterium absorbs almost no neutrons, so CANDU reactors run on natural uranium — no enrichment needed. The trade-off: heavy water costs $300-600 per kilogram to produce.
Why graphite works — and why it's dangerous
The Soviet RBMK reactor (the Chernobyl design) used graphite as its moderator and water only as coolant. Carbon-12 takes 115 collisions instead of 18, but it doesn't absorb neutrons and it doesn't boil.
This matters because of a critical design implication: in a water-moderated reactor, if the water boils away, you lose your moderator. No moderator → neutrons stay fast → fewer fissions → reactor shuts itself down. This is inherently safe.
In a graphite-moderated, water-cooled reactor, if the water boils away, you lose your coolant but KEEP your moderator. The graphite keeps slowing neutrons. Fissions continue. But now there's nothing to carry heat away.
This is the design flaw that killed Chernobyl. We'll return to it in Phase 7.
DESIGN SPEC UPDATED:
├── Fast neutrons (2 MeV) miss U-235 — cross-section too small
├── Thermal neutrons (0.025 eV) are 500× more likely to cause fission
├── Moderator slows neutrons via elastic collisions (billiard ball physics)
├── Hydrogen (water): 18 collisions, but absorbs some neutrons → need enriched fuel
├── Deuterium (heavy water): 25 collisions, minimal absorption → natural uranium OK
├── Graphite: 115 collisions, good moderator, but decoupled from coolant
└── Water loss = moderator loss = shutdown (inherently safe) — except in RBMK
───
PHASE 4: Control the Beast
You have a chain reaction sustained by moderated neutrons. But the chain reaction is exponential. How do you control something that can double its power in milliseconds?
If every neutron in a reactor were prompt — released at the instant of fission — the reactor's response time would be the neutron generation time: about 0.1 milliseconds.
At k = 1.001 (just barely supercritical), power would double every 0.1 seconds. k = 1.01 would double power every 10 milliseconds. No mechanical system can move control rods fast enough to keep up. No human could react. The reactor would be as uncontrollable as a bomb, just slower.
This is the nightmare scenario. And it's why nuclear reactors would be impossible to operate — except for one lucky accident of nuclear physics.
Delayed neutrons — the gift that makes reactors possible
Not all neutrons appear at the instant of fission. About 0.65% of neutrons from U-235 fission are delayed.
Here's what happens: some fission fragments are neutron-rich isotopes that undergo beta decay. After decay, the daughter nucleus is so excited it ejects a neutron. But the beta decay takes TIME.
Group Half-life Fraction of delayed neutrons
─────────────────────────────────────────────────────
1 55.7 s 0.033%
2 22.7 s 0.219%
3 6.2 s 0.196%
4 2.3 s 0.166%
5 0.61 s 0.026%
6 0.23 s 0.010%
─────────────────────────────────────────────────────
Total delayed fraction (β): 0.65%
99.35% of neutrons: appear in 10⁻¹⁴ seconds (prompt)
0.65% of neutrons: appear in 0.2 to 56 seconds (delayed)0.65% sounds trivial. It's not. Those delayed neutrons stretch the effective generation time from 0.1 milliseconds to about 0.1 SECONDS — a factor of 1,000. This makes the reactor respond on a human timescale instead of a machine-gun timescale.
Without delayed neutrons, a reactor's generation time is ~0.0001 seconds.
With delayed neutrons folded in, the EFFECTIVE generation time becomes ~0.1 seconds.
That factor of 1,000 is the difference between "uncontrollable" and "a human can operate it with a dial."
Reactor operators keep k in the range where the reactor is supercritical on prompt + delayed neutrons combined, but subcritical on prompt neutrons alone. This regime is called delayed critical. The reactor needs the delayed neutrons to sustain itself. Pull them out (by absorbing more neutrons) and the reaction dies on a timescale of seconds — plenty of time for control rods to move.
Control rods — the throttle and the brake
Control rods are made of materials with enormous neutron absorption cross-sections:
Boron-10: absorption cross-section = 3,840 barns
Cadmium-113: absorption cross-section = 20,600 barns
Hafnium: absorption cross-section = 104 barns (but doesn't swell or become brittle)
For comparison, U-235's fission cross-section is 585 barns. Cadmium absorbs neutrons 35 times more readily than uranium fissions them.
Push a control rod into the reactor core → it absorbs neutrons → fewer neutrons cause fission → k drops below 1 → power falls.
Pull a control rod out → fewer neutrons absorbed → more neutrons cause fission → k rises above 1 → power rises.
┌─ control rod (boron/cadmium)
│
▼
┌───────────────────────┐
│ ░░░░░░░█████░░░░░░░ │ ← rod inserted: absorbing neutrons
│ ░░fuel░░█rod█░░fuel░░ │ k < 1, power decreasing
│ ░░░░░░░█████░░░░░░░ │
└───────────────────────┘
┌───────────────────────┐
│ ░░░░░░░░░░░░░░░░░░░ │ ← rod withdrawn: neutrons free
│ ░░fuel░░░░░░░░░fuel░░ │ k = 1, steady power
│ ░░░░░░░░░░░░░░░░░░░ │
└───────────────────────┘
▲
│
└─ rod pulled up out of coreA typical PWR has 40-60 control rod assemblies, each containing 20-24 individual rods. In an emergency, all rods drop into the core by gravity in under 2 seconds. This is called a SCRAM — and it shuts the reactor down from full power to subcritical in seconds.
The negative temperature coefficient — the reactor that fixes itself
Water-moderated reactors have a built-in safety feature that requires no human intervention: the negative temperature coefficient of reactivity.
Here's how it works:
├── Reactor gets too hot
├── Water expands (becomes less dense)
├── Less dense water = fewer hydrogen atoms per cm³
├── Fewer hydrogen atoms = less moderation
├── Less moderation = fewer thermal neutrons
├── Fewer thermal neutrons = fewer fissions
├── Fewer fissions = less heat
└── Reactor cools itself down
The feedback loop is NEGATIVE. Hotter → less reactive → cooler. This is inherently stable. You could walk away from a PWR and it would not melt down from a power excursion — it would throttle itself.
The RBMK had the opposite: a positive void coefficient. As water boiled, the graphite moderator kept working, but the water's neutron absorption disappeared. Fewer absorbed neutrons → MORE fissions → MORE heat → MORE boiling → MORE fissions. A runaway positive feedback loop. We'll see exactly how this killed Chernobyl in Phase 7.
DESIGN SPEC UPDATED:
├── Delayed neutrons (0.65%): stretch response time from 0.1 ms to 0.1 s
├── Without delayed neutrons, reactors would be uncontrollable
├── Control rods (boron, cadmium): absorb neutrons, adjust k
├── SCRAM: all rods drop by gravity in <2 seconds
├── Negative temperature coefficient: hotter → less moderation → self-limiting
└── Positive void coefficient (RBMK): hotter → MORE reactive → runaway
───
PHASE 5: Carry the Heat Away
You're splitting atoms and controlling the chain reaction. But all that energy is heat — 200 MeV per fission, delivered as kinetic energy of fragments slamming into surrounding material. You need to move that heat OUT of the core and into something useful.
A typical large reactor produces 3,000 MW of thermal power. That's 3 billion watts of heat in a volume the size of a large room.
For comparison:
├── Your body at rest: 80 watts
├── A car engine: 100,000 watts
├── A Boeing 777 at takeoff: 200,000,000 watts
├── A nuclear reactor core: 3,000,000,000 watts
└── The Sun per square meter of surface: 63,000,000 watts
Your reactor core produces 48× more heat per square meter than the surface of the Sun. You MUST carry this heat away continuously or the fuel melts. There is no "pause" button. Even after shutdown, decay heat from radioactive fission products keeps pumping out megawatts.
Three loops — keep the radiation in, get the heat out
You can't run steam from the reactor core directly to a turbine. The water touching the fuel is intensely radioactive — activated by neutron bombardment, contaminated with fission products. Send it to a turbine and you'd need to put the entire power plant inside a radiation containment building.
Solution: heat exchangers. Transfer heat without transferring fluid.
PRIMARY LOOP (radioactive) SECONDARY LOOP (clean) TERTIARY LOOP
┌──────────────────┐ ┌──────────────────┐ ┌────────────┐
│ │ │ │ │ │
│ REACTOR CORE │ ← 320°C → │ STEAM GENERATOR │ → steam │ TURBINE │
│ ░░░░░░░░░░░░ │ hot leg │ ┌────────────┐ │ → 270°C │ ↓ │
│ ░░ fuel ░░░░ │ │ │ primary │ │ │ GENERATOR │
│ ░░ rods ░░░░ │ │ │ tubes │ │ │ ↓ │
│ ░░░░░░░░░░░░ │ │ │ ≋≋≋≋≋≋≋≋≋ │ │ ← water │ CONDENSER │
│ │ ← 290°C → │ └────────────┘ │ ← 50°C │ ↓ │
│ PRESSURIZER │ cold leg │ │ │ COOLING │
│ 155 atm │ │ │ │ TOWER │
└──────────────────┘ └──────────────────┘ └────────────┘
Water at 155 atm Steam at ~70 atm River/ocean
Never boils (327°C Drives the turbine or tower
boiling point at ~270°C dumps waste
at this pressure) heat
Primary water never leaves containment.
Secondary steam is clean — no radiation.
Tertiary water goes to the environment.The primary loop runs at 155 atmospheres to keep water liquid at 320°C (water boils at 100°C at 1 atm, but at 327°C at 155 atm). The steam generator transfers heat from the primary to the secondary loop through thousands of thin metal tubes — the fluids never mix. This is the same principle as the radiator in your car, except the stakes are somewhat higher.
Why 155 atmospheres? Because you want the hottest water possible — hotter water carries more energy per kilogram — but you can't let it boil. Boiling creates steam voids that change moderation. So you pressurize it far above the boiling point.
The pressurizer is a separate vessel connected to the primary loop, half-filled with water and half with steam. Electric heaters and cold-water sprays maintain exact pressure. It's the pressure regulator for the entire reactor.
Why only 33% efficiency? The Carnot limit.
Your reactor produces 3,000 MW of heat. But the electrical output is only about 1,000 MW. Where did 2,000 MW go?
The second law of thermodynamics. No heat engine — not even a perfect one — can convert all heat into work. The maximum possible efficiency is set by the Carnot limit:
T_cold
η_max = 1 − ──────────
T_hot
T_hot = 593 K (320°C primary coolant)
T_cold = 308 K (35°C condenser/river water)
η_max = 1 − 308/593 = 1 − 0.519 = 0.481 = 48.1%
That's the THEORETICAL maximum. Real turbines, pumps,
and heat exchangers have losses:
Actual efficiency: ~33%
So: 3,000 MW thermal × 0.33 = ~1,000 MW electrical
Remaining 2,000 MW → waste heat → cooling towers/river
Comparison:
────────────────────────────────────────────────
Coal plant: ~36% efficiency (hotter steam: ~540°C)
Nuclear (PWR): ~33% efficiency (limited to ~320°C)
Gas turbine: ~40% efficiency (combustion at ~1,400°C)
Combined cycle: ~60% efficiency (gas + steam turbine)Nuclear runs cooler than coal or gas because the fuel cladding — Zircaloy alloy tubes surrounding the fuel pellets — starts degrading above ~400°C and melts at 1,850°C. You can't just "make it hotter" without melting your own fuel rods. This material limit caps the efficiency.
Could you make a nuclear plant more efficient by running it hotter? In principle, yes. The Carnot efficiency climbs as T_hot rises. At 800°C (some advanced reactor designs), η_max = 1 − 308/1073 = 71%.
The problem: Zircaloy cladding. The thin tubes that contain the fuel pellets are made of zirconium alloy because zirconium absorbs almost no neutrons (critical for reactor economy) and is strong at moderate temperatures. But it oxidizes violently in steam above ~1,200°C, producing hydrogen gas. And it melts at 1,850°C.
This is not a theoretical concern. At Fukushima, overheated Zircaloy cladding reacted with steam, generating hydrogen that accumulated in the reactor buildings and exploded. The material that makes the reactor efficient is also the material that fails catastrophically.
DESIGN SPEC UPDATED:
├── Thermal power: 3,000 MW → electrical output: ~1,000 MW (33% efficiency)
├── Three loops: primary (radioactive, 155 atm) → secondary (steam) → tertiary (cooling)
├── Carnot limit: η = 1 − T_cold/T_hot = 48% theoretical, ~33% actual
├── Limited by Zircaloy cladding: can't exceed ~400°C safely
├── Zircaloy + steam above 1,200°C → hydrogen gas (explosion risk)
└── 2,000 MW of waste heat must be continuously dumped to the environment
───
PHASE 6: Contain the Unthinkable
What if everything goes wrong at once? The chain reaction runs away, the coolant drains, the fuel melts. What stands between a molten reactor core and the outside world?
Nuclear safety is built on a principle called defense in depth. Not one barrier. Not two. Multiple independent barriers, each sufficient alone, stacked so that the failure of any single barrier doesn't release radiation.
It's the same philosophy as designing an airplane: multiple hydraulic systems, triple-redundant flight computers, backup power, manual reversion. Except the stakes are higher and the timescales are longer.
Five barriers between fission products and the public
BARRIER 1: THE FUEL ITSELF
┌─────────────────────────────────────┐
│ UO₂ ceramic fuel pellet │ Melting point: 2,865°C
│ ● Fission products are trapped │ Retains most gases up to 1,200°C
│ inside the crystal lattice │ First line of defense: the fuel
│ │ matrix itself locks in radioactivity
└──────────────┬──────────────────────┘
│
BARRIER 2: FUEL CLADDING
┌──────────────┴──────────────────────┐
│ Zircaloy tube (0.6 mm thick) │ Melting point: 1,850°C
│ ● Sealed at both ends │ Contains fission gases (Xe, Kr)
│ ● Coolant flows OUTSIDE │ Fails above ~1,200°C (oxidation)
└──────────────┬──────────────────────┘
│
BARRIER 3: REACTOR PRESSURE VESSEL
┌──────────────┴──────────────────────┐
│ Carbon steel + stainless liner │ Thickness: 20-25 cm
│ ● 155 atm operating pressure │ Design pressure: ~175 atm
│ ● Weighs 300-500 tons │ Inspected for cracks every refueling
└──────────────┬──────────────────────┘
│
BARRIER 4: CONTAINMENT BUILDING
┌──────────────┴──────────────────────┐
│ Reinforced concrete + steel liner │ Concrete: 1.2-1.8 m thick
│ ● Steel liner: 6-10 mm │ Steel liner: gas-tight
│ ● Design pressure: ~5 atm │ Withstands: aircraft impact,
│ ● Volume: ~50,000-80,000 m³ │ tornado, earthquake, internal
│ │ steam explosion
└──────────────┬──────────────────────┘
│
BARRIER 5: EXCLUSION ZONE
┌──────────────┴──────────────────────┐
│ Distance + emergency planning │ 1-3 km exclusion zone
│ ● Emergency evacuation plans │ 16 km emergency planning zone
│ ● Radiation monitoring network │ 80 km ingestion pathway zone
└─────────────────────────────────────┘Every barrier is independent. The fuel pellet retains fission products even if the cladding fails. The cladding contains gases even if the vessel leaks. The vessel holds pressure even if the containment is breached. Five layers, each designed assuming the previous one has failed completely.
The containment building — calculate the pressure
In a Loss of Coolant Accident (LOCA), the 155-atmosphere primary loop ruptures. Superheated water at 320°C flashes to steam. How much pressure must the containment hold?
Primary loop water inventory: ~300 m³ at 320°C, 155 atm
If all this water flashes to steam inside containment:
Containment free volume: ~50,000 m³
Steam from 300 m³ of water at 320°C:
At atmospheric pressure, water expands ~1,600× as steam
Volume of steam = 300 × 1,600 = 480,000 m³
But containment is only 50,000 m³, so steam is compressed:
P ≈ 480,000 / 50,000 × 1 atm ≈ ~4.5 atm absolute
(actual calculation accounts for temperature, condensation
on cold structures, and spray systems)
Design pressure: ~5 atm (with margin)
The 1.2 m concrete wall must resist:
Hoop stress in a cylinder: σ = P × r / t
P = 5 atm = 507 kPa
r = ~20 m (typical containment radius)
t = 1.2 m
σ = 507,000 × 20 / 1.2 = 8.45 MPa
Reinforced concrete tensile strength: ~3-5 MPa (concrete alone)
With steel rebar: effective strength >30 MPa
Safety margin: 3.5× or higherThe containment doesn't need to hold hundreds of atmospheres — it only needs to hold the steam from the entire primary loop flashing at once. That's about 5 atm. The real engineering challenge is making it leak-tight (the steel liner) and impact-resistant (the reinforced concrete).
The China Syndrome — myth vs reality
The 1979 movie proposed that a melted reactor core would burn through the containment floor, through the Earth's crust, all the way to China. Physics says otherwise.
A fully molten core (~100 tons of UO₂ and steel, called corium) reaches about 2,800°C. It will melt through steel (melting point: 1,510°C) but what happens when it hits concrete?
Concrete ablates slowly. The calcium carbonate decomposes endothermically (absorbs heat). Water bound in the concrete flashes to steam, cooling the corium. The melt spreads laterally as it penetrates, increasing surface area and heat loss.
Experiments at Sandia National Laboratory (MACE tests) showed that corium penetrates about 1-2 meters into concrete before cooling enough to solidify.
The real concern isn't burning to China. It's that corium-concrete interaction generates hydrogen and carbon monoxide (flammable gases) and aerosolizes radioactive material. Containment spray systems and hydrogen recombiners are designed to handle exactly this scenario.
At Three Mile Island, the core partially melted. Corium reached the bottom of the reactor vessel — and stopped. The vessel held. The containment held. No radiation reached the public at levels above background.
DESIGN SPEC UPDATED:
├── Five barriers: fuel pellet → cladding → vessel → containment → exclusion zone
├── Reactor vessel: 20-25 cm carbon steel, 300-500 tons
├── Containment: 1.2-1.8 m reinforced concrete, ~5 atm design pressure
├── Safety margin on containment: >3.5× design loads
├── Corium penetrates 1-2 m into concrete before solidifying (not to China)
└── TMI core melted — vessel held, containment held, public dose: negligible
───
PHASE 7: Learn from Disaster
Three reactors have failed catastrophically in 70 years of commercial nuclear power. Each failure taught the industry something that no simulation, no analysis, and no safety review could have revealed. Each was a specific engineering mistake, not a random act of God.
Nuclear disaster is not about bad luck. It's about designs that look safe until one particular combination of events exposes a hidden vulnerability. Understanding exactly HOW each one failed is more instructive than knowing that they failed.
Three Mile Island (1979) — the accident that worked
Pennsylvania, USA. March 28, 1979.
What happened:
├── A valve stuck open after a transient event
├── Coolant drained from the primary loop through the stuck valve
├── Operators saw RISING water level on a misleading instrument
│ (the instrument measured pressurizer level, not core level)
├── They REDUCED emergency cooling water — making it worse
├── The core partially uncovered for ~2 hours
├── About 45% of the core melted
├── 20 tons of corium pooled in the bottom of the reactor vessel
└── The vessel held. The containment held.
Radiation released: negligible. The maximum dose to any member of the public was less than a chest X-ray. No injuries. No deaths. No measurable health effects in the 40+ years since.
Lesson: The instruments were right. The operators' mental model was wrong. They didn't understand what the instruments were actually measuring. TMI led to a complete redesign of control room interfaces, operator training programs, and the creation of the Institute of Nuclear Power Operations (INPO).
The containment building — that 1.2 meters of reinforced concrete — did exactly what it was designed to do. It contained a partial meltdown with zero harm to the public.
Chernobyl (1986) — the design that killed
Pripyat, Ukraine (Soviet Union). April 26, 1986.
The RBMK reactor had three fatal design features:
1. Positive void coefficient. As water boiled, reactivity INCREASED. (Phase 3 explained why — graphite moderator stays, water absorber leaves.)
2. No containment building. The RBMK was housed in a standard industrial building. No reinforced concrete dome. No steel liner. No pressure rating beyond normal atmospheric.
3. Control rod design flaw. The control rods had graphite tips. When first inserted, the graphite tips DISPLACED water at the bottom of the core. Water absorbs more neutrons than graphite. So the first few seconds of inserting the rods actually INCREASED reactivity before the absorbing boron section entered the core.
What happened:
├── Operators ran a safety test (ironically) with inadequate preparation
├── They disabled multiple safety systems to run the test
├── Reactor dropped to very low power — xenon poisoning built up
├── They pulled too many control rods to compensate (210 of 211 withdrawn)
├── At 1:23 AM, they triggered the test
├── Power surged — operators hit the emergency shutdown button (AZ-5)
├── The graphite-tipped rods briefly INCREASED reactivity
├── Power spiked to 30,000 MW (10× rated power) in 4 seconds
├── Fuel shattered, steam pressure blew the 1,000-ton reactor lid off
├── Graphite caught fire — burned for 10 days
└── Fallout spread across Europe
Power
(MW)
│
30000─┤ ● EXPLOSION
│ ╱
│ ╱
│ ╱ AZ-5 pressed — but graphite
│ ╱ tips INCREASE reactivity first
3200─┤─── rated power ──●╱
│ │
200─┤── ── ── ── ── ──● ← low power, test begins
│ │
└───────────────────┼────→ time
1:23:40 1:23:44
4 secondsFrom 200 MW to 30,000 MW in 4 seconds. The thermal shock shattered the fuel. Steam pressure blew the 1,000-ton upper biological shield into the air. Without a containment building, there was nothing between the burning core and the atmosphere.
Deaths: 2 from the explosion, 29 from acute radiation syndrome in the following weeks, and an estimated 4,000-16,000 excess cancer deaths over subsequent decades (WHO/UNSCEAR estimates vary).
The 30-km exclusion zone around Chernobyl remains largely uninhabited to this day.
Lesson: Design trumps operation. No amount of operator training can save a reactor with a positive void coefficient and no containment building. Every reactor built since has a negative void coefficient and a containment building. The RBMK design was known to have these flaws. Engineers raised concerns. The Soviet system suppressed them.
Fukushima Daiichi (2011) — the disaster nobody imagined
Miyagi Prefecture, Japan. March 11, 2011.
The plant survived the earthquake. The reactors scrammed automatically. Control rods inserted. Chain reactions stopped.
Then the tsunami arrived.
├── 15-meter wave overtopped the 5.7-meter seawall
├── Seawater flooded the basement — where the emergency diesel generators sat
├── All backup power lost (station blackout)
├── Without power: no pumps → no coolant flow → no heat removal
├── But the chain reaction was stopped — why was there still heat?
Decay heat. Even after shutdown, radioactive fission products in the fuel continue to decay, releasing energy:
Time after shutdown Decay heat (% of full power)
──────────────────────────────────────────────────────
0 seconds ~6.5% = 200 MW for a 3,000 MW reactor
1 hour ~1.5% = 45 MW
1 day ~0.4% = 12 MW
1 week ~0.2% = 6 MW
1 month ~0.1% = 3 MW
1 year ~0.02% = 0.6 MW
Even at 1%: 30 MW = enough to melt the core in hours without coolingDecay heat cannot be turned off. It's not the chain reaction — that stopped instantly with the SCRAM. It's the radioactive inventory that accumulated during years of operation. The fission products will decay on their own schedule regardless of what you do. You MUST keep pumping coolant or the fuel melts. This is the fundamental vulnerability of nuclear power: you can turn off the chain reaction, but you cannot turn off the heat.
At Fukushima:
├── 3 reactors lost cooling
├── Core temperatures climbed past 1,200°C
├── Zircaloy cladding oxidized: Zr + 2H₂O → ZrO₂ + 2H₂
├── Hydrogen accumulated in reactor buildings
├── Hydrogen exploded in Units 1, 3, and 4 — destroying the buildings
├── Fuel melted in Units 1, 2, and 3
├── Contaminated water leaked into the Pacific
Deaths from radiation: zero (as of 2025). One worker death later attributed to lung cancer from radiation exposure. The evacuation itself caused ~2,300 deaths from stress, disruption, and inadequate care of elderly residents.
Lesson: Plan for the disaster you can't imagine. The seawall was designed for a historical tsunami. The 2011 tsunami exceeded all historical records. Emergency generators were in the basement — the worst possible location for flood protection. Every nuclear plant worldwide has since been re-evaluated for "beyond design basis" events.
DESIGN SPEC UPDATED:
├── TMI (1979): containment worked — partial meltdown, zero public harm
├── Chernobyl (1986): positive void coefficient + no containment = catastrophe
├── Fukushima (2011): station blackout + decay heat → hydrogen explosions
├── Decay heat: ~6.5% at shutdown, can melt core in hours without cooling
├── Every disaster was a specific engineering failure, not random bad luck
└── Post-Chernobyl: all new reactors require containment + negative void coefficient
───
PHASE 8: Outlast Civilization
The reactor ran for 18 months. The fuel is spent. But the waste it produced will be dangerous for longer than human civilization has existed.
Spent nuclear fuel is the most concentrated form of hazardous material humans have ever created. A single spent fuel assembly — about the size of a telephone pole, weighing 300 kg — contains:
├── Unreacted U-235 and U-238 (~95%)
├── Plutonium and other transuranics (~1%)
├── Fission products: Cs-137, Sr-90, I-131, Tc-99, and hundreds more (~4%)
└── Enough radioactivity to deliver a lethal dose in minutes at close range
The spent fuel is removed from the reactor and placed in a spent fuel pool — a deep pool of water that shields radiation and carries away decay heat. It will sit there for 5-10 years, cooling down, before it can be moved to dry storage.
Half-life math — when does it become safe?
Radioactive decay follows a precise exponential law:
N(t) = N₀ × (1/2)^(t / t_half)
Each isotope decays at its own rate. The dangerous ones span an enormous range:
Isotope Half-life Danger
──────────────────────────────────────────────────────
I-131 8.02 days thyroid cancer (Chernobyl)
Sr-90 28.8 years bone seeker (mimics calcium)
Cs-137 30.2 years gamma emitter (Chernobyl, Fukushima)
Am-241 432 years alpha emitter (smoke detectors)
Pu-239 24,100 years alpha emitter, weapons-usable
Tc-99 211,000 years mobile in groundwater
I-129 15.7 million years very long-lived, low activity
U-235 704 million years the fuel itself
U-238 4.47 billion years natural background
After 10 half-lives, activity drops to (1/2)¹⁰ = 1/1024 ≈ 0.1%Short-lived isotopes are intensely radioactive but burn out quickly. Long-lived isotopes are weakly radioactive but persist forever. The dangerous middle ground — Cs-137 and Sr-90 with ~30-year half-lives — dominates the hazard for the first few centuries.
The short-lived isotopes (I-131, 8 days) are ferociously radioactive but burn out in weeks. The truly long-lived ones (U-238, 4.47 billion years) are so weakly radioactive they're essentially harmless — natural uranium ore exists in the ground all around you.
The danger window is the middle: Cs-137 and Sr-90 dominate for the first 300 years. Then plutonium isotopes dominate for tens of thousands of years.
When is spent fuel as safe as natural uranium ore?
This is the question that defines the waste problem. Natural uranium ore exists in deposits worldwide. It's radioactive, but not dangerous at normal distances. When does spent fuel decay to that level?
Radioactivity
(relative to
natural ore = 1)
│
10⁶─┤ ● ← at discharge: 1 million × natural ore
│ ╲
10⁵─┤ ╲
│ ╲
10⁴─┤ ╲ ← 10 years: still intensely hot
│ ╲
10³─┤ ╲
│ ╲ ← 100 years: Cs-137/Sr-90 declining
10²─┤ ╲
│ ╲
10─┤ ╲ ← 1,000 years: most fission products gone
│ ╲
1─┤─ ─ ─ ─ ─ ─ ─ ─ ─ ─╲─ ─ ─ natural ore level ─ ─ ─
│ ╲___________
└───┬──────┬──────┬──────┬──────┬──────┬──→
10 100 1k 10k 100k 1M years
Crossover point: ~300,000 yearsAfter about 1,000 years, the fission products have largely decayed. What remains are the transuranic elements — plutonium, americium, neptunium — which are alpha emitters. Alpha radiation is stopped by a sheet of paper but deadly if ingested. The waste must be isolated from groundwater for ~300,000 years to reach natural ore radioactivity levels.
300,000 years. For context:
├── Modern humans have existed: ~300,000 years
├── Agriculture was invented: 12,000 years ago
├── Writing was invented: 5,500 years ago
├── The oldest surviving structure: ~6,500 years (Knap of Howar)
└── You need to store waste safely for longer than civilization has existed
This is the hardest engineering problem nuclear power has ever posed. Not the reactor. Not the containment. The waste.
Deep geological storage — bury it and forget it?
The leading solution: dig a repository 300-500 meters deep in geologically stable rock. Place the waste in copper or steel canisters. Surround them with bentonite clay (which swells when wet, sealing cracks). Backfill the tunnels. Seal the shafts.
Finland's Onkalo repository — the world's first permanent deep geological repository — is being built in 1.8-billion-year-old gneiss bedrock. It's expected to begin accepting waste around 2025-2026.
Sweden is building a similar facility in granite.
The U.S. spent $15 billion studying Yucca Mountain in Nevada before political opposition killed the project in 2010. The geology was suitable. The engineering was sound. The politics were impossible.
And there's one final challenge: how do you warn civilizations that don't exist yet?
The waste will be dangerous for 300,000 years. No human language has survived more than 5,000 years. No symbol system we know has lasted 10,000 years. The U.S. Department of Energy commissioned studies on creating warning systems for Yucca Mountain. Proposals included:
├── Giant concrete thorns covering the site (hostile architecture)
├── Engineered "atomic priesthood" — a self-perpetuating knowledge cult
├── Genetically modified cats that glow near radiation ("ray cats")
├── Embedding information in the landscape itself
└── Or: simply bury it deep enough that no civilization could accidentally reach it
None of these proposals have been proven to work over the required timescale. The waste problem remains unsolved — not technically, but socially. We know HOW to store it safely. We don't know how to COMMUNICATE the danger across geological time.
DESIGN SPEC COMPLETE:
├── Spent fuel: 1 million × more radioactive than natural ore at discharge
├── Cs-137 and Sr-90 (~30 yr half-life): dominate hazard for first 300 years
├── Pu-239 (24,100 yr): dominates from 1,000 to 100,000 years
├── Crossover to natural ore level: ~300,000 years
├── Deep geological storage: 300-500 m in stable rock (Finland's Onkalo, first in world)
├── Yucca Mountain: technically sound, politically dead
└── 300,000-year warning problem: no human communication system has lasted that long
───
FULL MAP
Nuclear Reactor
├── Phase 1: Split an Atom
│ ├── Energy source: strong nuclear force (binding energy)}
│ ├── Reaction: n + U-235 → Ba-141 + Kr-92 + 3n + 200 MeV}
│ ├── Mass defect: 0.186 amu converts to energy via E = mc²}
│ ├── Energy per fission: 200 MeV (50 million × chemical burning)}
│ ├── Energy density: 82 TJ/kg (vs 24 MJ/kg for coal)}
│ └── Iron-56 is the peak — fission moves heavy nuclei toward it}
│
├── Phase 2: Start a Chain Reaction
│ ├── Chain reaction: each fission releases 2-3 neutrons → self-sustaining}
│ ├── k = 1.0000: critical (steady power) — the operating point}
│ ├── k > 1: supercritical (power rising) — startup or bomb}
│ ├── k 80%)}
│ └── A reactor cannot produce a nuclear explosion — physics forbids it}
│
├── Phase 3: Slow the Neutrons Down
│ ├── Fast neutrons (2 MeV) miss U-235 — cross-section too small}
│ ├── Thermal neutrons (0.025 eV) are 500× more likely to cause fission}
│ ├── Moderator slows neutrons via elastic collisions (billiard ball physics)}
│ ├── Hydrogen (water): 18 collisions, but absorbs some neutrons → need enriched fuel}
│ ├── Deuterium (heavy water): 25 collisions, minimal absorption → natural uranium OK}
│ ├── Graphite: 115 collisions, good moderator, but decoupled from coolant}
│ └── Water loss = moderator loss = shutdown (inherently safe) — except in RBMK}
│
├── Phase 4: Control the Beast
│ ├── Delayed neutrons (0.65%): stretch response time from 0.1 ms to 0.1 s}
│ ├── Without delayed neutrons, reactors would be uncontrollable}
│ ├── Control rods (boron, cadmium): absorb neutrons, adjust k}
│ ├── SCRAM: all rods drop by gravity in <2 seconds}
│ ├── Negative temperature coefficient: hotter → less moderation → self-limiting}
│ └── Positive void coefficient (RBMK): hotter → MORE reactive → runaway}
│
├── Phase 5: Carry the Heat Away
│ ├── Thermal power: 3,000 MW → electrical output: ~1,000 MW (33% efficiency)}
│ ├── Three loops: primary (radioactive, 155 atm) → secondary (steam) → tertiary (cooling)}
│ ├── Carnot limit: η = 1 − T_cold/T_hot = 48% theoretical, ~33% actual}
│ ├── Limited by Zircaloy cladding: can't exceed ~400°C safely}
│ ├── Zircaloy + steam above 1,200°C → hydrogen gas (explosion risk)}
│ └── 2,000 MW of waste heat must be continuously dumped to the environment}
│
├── Phase 6: Contain the Unthinkable
│ ├── Five barriers: fuel pellet → cladding → vessel → containment → exclusion zone}
│ ├── Reactor vessel: 20-25 cm carbon steel, 300-500 tons}
│ ├── Containment: 1.2-1.8 m reinforced concrete, ~5 atm design pressure}
│ ├── Safety margin on containment: >3.5× design loads}
│ ├── Corium penetrates 1-2 m into concrete before solidifying (not to China)}
│ └── TMI core melted — vessel held, containment held, public dose: negligible}
│
├── Phase 7: Learn from Disaster
│ ├── TMI (1979): containment worked — partial meltdown, zero public harm}
│ ├── Chernobyl (1986): positive void coefficient + no containment = catastrophe}
│ ├── Fukushima (2011): station blackout + decay heat → hydrogen explosions}
│ ├── Decay heat: ~6.5% at shutdown, can melt core in hours without cooling}
│ ├── Every disaster was a specific engineering failure, not random bad luck}
│ └── Post-Chernobyl: all new reactors require containment + negative void coefficient}
│
└── Phase 8: Outlast Civilization
├── Spent fuel: 1 million × more radioactive than natural ore at discharge}
├── Cs-137 and Sr-90 (~30 yr half-life): dominate hazard for first 300 years}
├── Pu-239 (24,100 yr): dominates from 1,000 to 100,000 years}
├── Crossover to natural ore level: ~300,000 years}
├── Deep geological storage: 300-500 m in stable rock (Finland's Onkalo, first in world)}
├── Yucca Mountain: technically sound, politically dead}
└── 300,000-year warning problem: no human communication system has lasted that long}
───