RBMK Reactor: The Emergency Shutdown That Triggered Chernobyl

How an RBMK-1000 Works: Boiling Water in a Graphite Stack

The RBMK-1000 (Reaktor Bolshoy Moshchnosti Kanalnyy, high-power channel reactor) was designed by NIKIET under chief designer Nikolai Dollezhal, with the Kurchatov Institute as scientific lead. The first unit, Leningrad-1, reached the grid in 1973. It is rated at 3,200 MWt of heat and 1,000 MWe of electricity, a gross efficiency of 1,000 ÷ 3,200 ≈ 31%.

• The core. A cylinder of graphite blocks about 11.8 m across and 7 m tall, pierced by vertical channels on a 250 mm grid. Graphite is the main moderator, slowing fission neutrons to thermal speeds where U-235 fissions far more readily.
• The pressure tubes. 1,661 fuel channels of zirconium–2.5% niobium with an 80 mm bore, each holding two stacked 18-pin bundles of uranium dioxide enriched to 2% U-235.
• The coolant. Water is pumped up every channel and boils as it rises, leaving at ~7 MPa and ~284 °C as a mix only ~14% steam by mass. Drum separators send the steam straight to two 500 MWe turbines.
• The controls. 211 boron-carbide control rods in their own water-cooled channels, most driven down from above and 24 shortened rods entering from below.

Soviet planners liked the layout: no giant forged pressure vessel, easy scaling by adding channels, and refuelling while running. The price was a huge, loosely coupled core that was hard to keep in balance, with no Western-style containment building.

The Positive Void Coefficient: Steam That Adds Power

A reactor holds steady when each fission causes, on average, exactly one more: the multiplication factor k = 1. The departure from that is the reactivity, ρ = (k − 1)/k. A small fraction of neutrons, β ≈ 0.005 in partly burned uranium fuel, are delayed, emitted seconds after fission, and they make a reactor controllable. Below ρ = β, power changes on their slow timescale. At ρ ≥ β the core is prompt-critical: prompt neutrons alone sustain growth, and power can multiply many times over in a fraction of a second.

The key feedback is the void coefficient, the change in ρ when coolant turns to steam. In Western PWRs and BWRs the water is also the moderator, so steam means fewer slowed neutrons and reactivity falls. In the RBMK the graphite does most of the moderating, and the water mainly absorbs neutrons, chiefly on hydrogen. Replace water with steam and absorption drops, so reactivity rises: more power, more steam, more power.

Worked example. Post-accident analyses put the void coefficient of Unit 4's core at roughly +2×10⁻⁴ per percentage point of steam volume. Going from just critical to prompt critical takes ρ = β = 0.005, so 0.005 ÷ 0.0002 = 25 percentage points of extra steam would do it; voiding the whole core would be worth ~0.02, or ~4β. At full power the negative fuel-temperature (Doppler) coefficient outweighs this, leaving the power coefficient slightly negative. At low power the balance flips. The test ran at ~200 MWt, just 200 ÷ 3,200 ≈ 6% of rated.

Two things made it worse. Almost all the fixed additional absorbers had been replaced with fuel as the core burned in; Unit 4 had one left, pushing the void coefficient further positive. And xenon-135, a fission product with an absorption cross-section of ~2.6 million barns, had built up after the long hold at half power and the later plunge to low power, forcing the crew to pull nearly every rod out.

The Graphite Displacer and the Positive Scram Effect

Most of the top-entry control rods carried a 4.5 m graphite displacer hanging below the boron-carbide absorber. Without it, a withdrawn rod would leave its channel full of absorbing water; graphite there made each rod worth more. The flaw was the length. A 4.5 m displacer centred in a 7 m core leaves (7 − 4.5) ÷ 2 = 1.25 m of water above it and 1.25 m below it. Here is what a scram does from that fully withdrawn position:

• 1. Release. AZ-5 sends the rods down at ~0.4 m/s.
• 2. The bottom gains. The displacer slides into the lowest 1.25 m of the channel, pushing out absorbing water and replacing it with graphite that barely absorbs. Reactivity in the lower core goes up.
• 3. The top barely loses yet. At the same moment the boron carbide is only beginning to enter the top of the core.
• 4. The window. Clearing the bottom water takes 1.25 m ÷ 0.4 m/s ≈ 3.1 s. Meanwhile the net effect depends on the power shape; where the lower core carries high weight, the bottom gain can beat the top loss.
• 5. Real shutdown, much later. The rods deliver their full negative worth only as the absorber sweeps the full height: 7 m ÷ 0.4 m/s ≈ 17.5 s, ~18–20 s for the full stroke. A typical Western PWR rod drops by gravity in roughly 2–3 s.

This positive scram effect was seen at Ignalina-1 in 1983. The designers were informed, but the rods were never modified before April 1986, and operators were not warned that the shutdown button could briefly push power up.

26 April 1986, Second by Second

The test was meant to show that a coasting turbine could keep the main circulation pumps turning for the roughly one minute emergency diesel generators needed to take the load.

• 25 April. Power reduction begins after midnight. By early afternoon Unit 4 is at 1,600 MWt and the emergency core cooling system is isolated, per the test program. The Kyiv grid dispatcher asks the unit to stay online, and it holds at half power for about nine hours while xenon builds.
• 26 April, 00:28. On the night shift under foreman Aleksandr Akimov, with deputy chief engineer Anatoly Dyatlov directing the test, power collapses to ~30 MWt during a switch between automatic control modes. Fighting xenon, the crew withdraws rods and recovers only ~200 MWt instead of the planned 700–1,000 MWt.
• 01:03 and 01:07. Two standby pumps start, making eight. The heavy flow leaves the channel water only just below boiling.
• 01:22:30. A computer printout shows a reactivity margin of ~6–8 rods, against a 15-rod minimum.
• 01:23:04. Steam to turbine 8 is shut off and the test begins. As the four pumps it powers slow down, flow falls and boiling increases.
• 01:23:40. AZ-5 is pressed; whether as the planned end-of-test shutdown or in reaction to something on the panel is not known with certainty.
• The next few seconds. Displacers add reactivity at the bottom, lower-core water flashes to steam, the steam adds more, and the core passes prompt critical. Power reaches an estimated ~30,000 MWt: 30,000 ÷ 3,200 ≈ 9.4, about 10× rated and ~150× the test's 200 MWt. The rods stall roughly a third of the way in as channels deform.
• Moments later. Fuel shatters, pressure tubes burst, and a steam explosion heaves the ~17 m upper biological shield, weighing well over 1,000 t, off the reactor. A second explosion scatters graphite and fuel.

Two workers died that night and 28 more died of acute radiation syndrome within months, of 134 diagnosed. Pripyat's ~49,000 residents were evacuated on 27 April.

Measuring Reactivity: Margins, Rod Worth and Design Rules

Reactor physicists often express reactivity in units of β, called dollars: a reactivity of one dollar is exactly prompt critical. A control rod's worth is measured in start-up physics tests by moving it and timing how fast power then changes (the reactor period). Void and temperature coefficients are found the same way, by perturbing flow, temperature or power.

RBMK crews tracked the operational reactivity margin (ORM): the negative reactivity held by the inserted parts of all rods, expressed as an equivalent number of fully inserted manual rods. It was not displayed live; the SKALA computer calculated it with the PRIZMA program for periodic printouts. A margin of ~6–8 rods, roughly half the 15-rod floor, meant most rods were parked fully withdrawn, the exact position from which the displacer effect was worst.

Western licensing demands what the RBMK lacked. US rule 10 CFR 50 Appendix A, General Design Criterion 11, requires that in the power operating range the prompt inherent nuclear feedback tends to compensate for a rapid increase in reactivity; Criterion 26 requires two independent reactivity control systems of different design principles. PWRs and BWRs meet Criterion 11 chiefly through a negative fuel-temperature (Doppler) coefficient, backed by moderator and void coefficients that are negative at power, and scram within seconds, by gravity in a PWR and hydraulic accumulators in a BWR.

Warnings, Verdicts and the Retrofits That Followed

Warnings came early: a 1975 fuel-channel failure at Leningrad-1 during low-power operation, and Ignalina-1's 1983 scram observation. Neither led to a fix in time.

The IAEA's INSAG-1 (1986) relied largely on the Soviet account and blamed operator violations. After a Soviet commission led by Nikolai Shteinberg re-examined the evidence in 1991, INSAG-7 (1992) revised that verdict: some supposed violations broke no rule in force, and the design flaws, above all the positive scram effect and positive void coefficient, carried major weight alongside a weak safety culture. Dyatlov, plant director Viktor Bryukhanov and chief engineer Nikolai Fomin had each been sentenced to 10 years in 1987.

Every remaining RBMK was modified:

• Fuel enrichment raised from 2.0% to 2.4%, and later to ~2.6–2.8% with an erbium burnable absorber, so the fuel takes a larger share of neutron absorption and losing the water matters less.
• Around 80 extra fixed absorbers added, and the minimum reactivity margin raised to several dozen rod-equivalents.
• Rod design and withdrawal limits changed so that no water column is left below the displacer, and full insertion speeded up to ~12 s.
• A fast-acting emergency system of 24 rods that insert in ~2.5 s.

Together these cut the void effect to below about +1β. Chernobyl's other units closed by 2000 and Ignalina's two RBMK-1500s by 2009; a dwindling number of retrofitted RBMK-1000s still run in Russia.

Misconceptions: Not a Bomb, Not Fukushima

• It was not a nuclear bomb. A weapon uses highly enriched uranium or weapons-grade plutonium, driven strongly supercritical by chemical explosives so fast neutrons multiply in well under a millisecond and release kilotons of energy (Hiroshima: ~15 kt). Fuel at 2% U-235 spread through graphite cannot do that. The first explosion was a steam explosion, and the cause of the second is still debated. The energy was orders of magnitude below a weapon's; the harm came from scattering the core's radioactivity.
• It was not Fukushima. At Fukushima Daiichi in 2011 the BWRs shut down correctly when the earthquake struck. The tsunami destroyed backup power, and decay heat, ~1–1.5% of rated power an hour after shutdown or ~25–35 MW in a 2,381 MWt unit, melted three cores; hot zirconium reacting with steam made the hydrogen that wrecked the buildings. Chernobyl was a reactivity accident in a running reactor; Fukushima was a cooling failure after shutdown. Both rate INES level 7, but Chernobyl released far more caesium-137, ~85 PBq against ~6–20 PBq to the air.
• It did not melt first and explode later. Unit 4 blew apart within seconds of the surge; the lava-like corium in the basement, including the Elephant's Foot, formed afterwards.
• The graphite tips were not the whole story. In a core with a healthy margin at higher power, the scram effect would have been a small blip. It became the trigger because low power, xenon, near-boiling water, rods parked fully out and a strongly positive void coefficient lined up at once.

The RBMK accident compared with the nuclear events it is most often confused with

Case	Moderator and coolant	What went wrong	Energy behind the damage
Chernobyl Unit 4, RBMK-1000 (1986)	Graphite moderator; light water boiling in 1,661 pressure tubes	Positive void coefficient plus positive scram effect: the running core went prompt-critical after AZ-5	Power surge to an estimated ~30,000 MWt (~10× rated) in seconds, then a steam explosion
Fukushima Daiichi Units 1–3, BWRs (2011)	Light water is both moderator and coolant	Rods inserted correctly on the earthquake; the tsunami then cut power and cooling to the shut-down reactors	Decay heat, ~1–1.5% of rated power an hour after shutdown; cores melted over hours to days, then hydrogen explosions
Three Mile Island Unit 2, PWR (1979)	Light water is both moderator and coolant	Reactor tripped normally; a stuck-open relief valve and misread instruments let coolant escape	Decay heat; about half the core melted, with little radioactivity released
Nuclear weapon (Hiroshima, 1945)	No moderator; highly enriched uranium or weapons-grade plutonium	Built to be driven far past prompt critical by chemical explosives	Fast-neutron chain reaction in well under a millisecond; ~15 kt of TNT