Michael KreponNuclear Weapon Accidents

Accidents happen. The least accident-prone nuclear weapons are the ones that are not in motion — but not always: see Eric Schlosser’s account of the Damascus incident in Command and Control.

Nuclear weapons in transit are more accident prone. The dangers associated with transit multiply with the number of vehicles carrying weapons in transit, which can spike during a crisis. The most accident-prone nuclear weapons are those in motion when hostilities commence, when standard operating procedures are subject to change. In this event, the most accident-prone nuclear weapons are those moved close to the forward edge of battle.

Nobel Prize-winning economist and strategic thinker Thomas C. Schelling (right) wrote about accidents in the September 1960 issue of the Bulletin of Atomic Scientists. Here are some excerpts from “Meteors, Mischief and War:”

“The point is that accidents do not cause war. Decisions cause war. Accidents can trigger decisions; and this may be all that anybody meant. But the distinction needs to be made, because the remedy is not just preventing accidents but constraining decisions.

“If we think of the decisions as well as the accidents we can see that accidental war, like premeditated war, is subject to “deterrence.” Deterrence, it is usually said, is aimed at the rational calculator in full control of his faculties and his forces; accidents may trigger war in spite of deterrence. But it is really better to consider accidental war as the deterrence problem, not a separate one…

“Thus the accident-prone character of strategic forces—more correctly, the sensitivity of strategic decisions to possible accidents—is closely related to the security of the forces themselves. If a country’s retaliatory weapons are reasonably secure against surprise attack, preemptive or premeditated, it need not respond so quickly. Not only can one wait and see, but one can assume that the enemy himself, knowing that one can wait and see, is less afraid of a precipitate decision, less preoccupied with his own need to preempt.

“And it is apparent that there can be quite a difference between an accident-prone system, and an accidental-war-prone system…

“What matters is whether this affects the way we wish to conduct the war. If the concept of ‘accidental war’—or whatever we choose to call a war that is not initiated altogether deliberately—has any meaning, it is probably a war in which our urge for revenge and retaliation is less than our urge to curtail the consequences of the error, regardless of whose error it was. If our object, in the event war should come, is to save as much of the country as possible and to provide for its further security, we should think not only about how to deter war, and how to enter it most effectively if it comes, but how to terminate it to best advantage.”

Accidents involving nuclear weapons are most likely to occur in states that have a rising learning curve and a high tempo of nuclear operations, whether due to a crisis or to paranoia. The probability of accidents also grows when safety devices are insufficient to prevent detonation or radiological dispersion in the event of an accident. Inadequate safety and security mechanisms also increase the probability of unauthorized use.

If a nuclear weapon-related accident occurs in the absence of a crisis or hostilities, accidental war might possibly be averted depending, at a minimum, on where the accident occurs and whether accurate, credible information can be released quickly about the circumstances surrounding this event. If, however, the nuclear accident occurs during a crisis or during the onset of military operations, escalation control could be more difficult, regardless of where the accident occurs. The initiation of military operations might reflect a conscious choice by national leaders, but a nuclear accident could mock the best laid plans.


  1. j_kies (History)

    Color me confused; accidents are one thing – detonation with nuclear yield is another entirely. Does this posting state that some risk exists that a ‘broken arrow’ style event occurring at the time of a crisis could presage an accidental war leading to nuclear escalation?

    • krepon (History)

      You are rarely confused.
      One scenario for an “accidental detonation” would be if a weapon is struck by incoming — perhaps delivered by aircraft — at the onset of hostilities. Perhaps a short-range system, perhaps deployed to signal an adversary, perhaps a failure of deterrence, perhaps thought to be carrying conventional warheads, perhaps not.

    • j_kies (History)

      Unless nations are transporting armed warheads (as in all the safing systems are in ‘unsafe’ states) nuclear yield in accident is extremely improbable. As Cthippo points out, weapons require simultaneously achieving geometry and effects that are difficult to accomplish on purpose.

      If the weapons are armed; the ‘accident’ term only applies to the intent and command/control.

    • Jonah Speaks (History)

      Krepon raises the issue of whether a nuclear weapon could be accidentally detonated by a conventional attack. If yes, this is an additional issue of concern if two nuclear-armed nations get into a conventional spat. So far as I am aware, there are no historical precedents for conventional attacks on nuclear weapons. This is a technical issue–any thoughts on whether accidental nuclear detonation by conventional attack is possible or likely?

      The notion of accidental nuclear war is far broader than that of accidental nuclear detonations. Usually the notion is based on false alarms or unanticipated escalations. Alan F. Philips, “20 Mishaps That Might Have Started Accidental Nuclear War,” provides only one example in twenty where a close-call, hypothetical nuclear detonation might have lead to nuclear war. http://nuclearfiles.org/menu/key-issues/nuclear-weapons/issues/accidents/20-mishaps-maybe-caused-nuclear-war.htm

      In the January 21, 1968 B-52 Crash near Thule, a “pilotless plane flew over the Thule base before crashing on the ice 7 miles miles offshore. Its fuel and high explosive component of its nuclear weapons exploded, but there was no nuclear detonation. At that time, the ‘one point safe’ condition of the nuclear weapons could not be guaranteed, and it is believed that a nuclear explosion could have resulted from accidental detonation of the high explosive trigger.”

    • krepon (History)

      maybe we need a new category: purposeful conventional/accidental nuclear warfare

    • John Schilling (History)


      There are well-understood mechanisms by which nuclear weapons can be rendered absolutely safe against any threat that does not involve a person with effective control of that weapon making a deliberate decision to ready it for imminent use. So-called “One-point safety” can be a part this, but is not necessary as there are other means to the same end. Some of these techniques are relatively simple to implement, amenable to any plausible weapon design, and involve only modest operational limitations.

      It is, of course, impossible to rule out the possibility that weapons designers in e.g. Pakistan or North Korea, have not taken these precautions. And depending on how severe a “conventional” conflict you are envisioning, it is also possible that a weapon’s owners will have fully armed the weapon in anticipation of imminent need.

      But the major threats are still the ones that lead to a person in control of the weapon deciding to use the weapon. Misinformation propagating through the command and control chain, or a weapon falling into the hands of an unauthorized party capable of putting it to use. There are technical countermeasures that are applicable here as well, but by nature not infallible ones.

    • Jonah Speaks (History)

      John, Thank you (and j_kies and Cthippo) for explaining the technical side on this. I will take it as unlikely or impossible for a conventional attack on a nuclear weapon to cause a nuclear detonation, unless the nuke is already armed. So, for example, if India discovered and attacked 50 not-yet-armed Pakistani nukes, it would be highly unlikely that any of the 50 conventional attacks would result in a nuclear detonation.

      To clear up a loose end, in the event “a weapon’s owners will have fully armed the weapon in anticipation of imminent need,” is there any reason why the weapon must be armed while it is still on home turf? That is, (human error or perversity aside) should we assume that all nuclear bombs and warheads are never armed while on home turf, because adequate electrical or mechanical means are always built in to ensure that the weapon is armed only after it has been launched and is at or near its intended target?

    • Cthippo (History)

      It’s hard to say how a particular weapon might be fused, but let’s take the Goldsboro accident as an example. In that case there were six separate parameters that had to be satisfied before the bomb could detonate. The one parameter that was not met, and the reason the bombs did not detonate, is that the arm / safe switch in the cockpit had not been set to arm when the bomb was dropped. Other sensors to prevent detonation include a barometric switch to ensure that the bomb, reached the correct high altitude before being dropped, wires that were attached to the aircraft that detected whther a bomb was actually dropped (as opposed to being inside an aircraft plummeting towards the earth), a delay timer to give the aircraft time to get away. These plugs were the only safety feature that failed in this accident since they are not supposed to activate in a breakup scenario.

      This might be easier to understand from the perspective of the bomb. When the B-52 broke up in mid flight over North Carolina the plugs pulled out and the bomb thought it had been dropped and “woke up”, activating the fuzing system.

      The bomb first checked “am I at the correct altitude”? Yes

      Have I been properly dropped? Yes. (actually no, it was confused)

      Have I fallen the programmed amount of time? Yes

      Was the Arm / safe switch set to arm when i was dropped? No.

      As a result of the one parameter not being met, the bomb did not arm itself and instead hit the ground and sat there until the truck came to get it the next day.

      This is admittedly a sort of cartoon example of how fuzing systems work, but hopefully you get the idea. Modern systems use a more advanced Permissive Action Link (Or Prohibited Action Link depending on who you ask) system in which a code has to be typed into a panel on the aircraft or silo before the weapon is activated. Until this code is entered and the weapon “unlocked” it cannot detonate. Typically in addition to the PAL there will also be other safety sensors including barometers and accelerometers to determine if it has been correctly launched and traveled a far enough distance before activating.

      To answer your implied question of when could a weapon be accidentally detonated, the only thing I can think of is if an aircraft had taken off with the intention of actually dropping the weapon and the PAL had been activated, but the aircraft was shot down before it reached the target then conceivably the weapon could detonate on impact. Theoretically a missile that was launched with a live warhead at a target could also detonate if hit by an interceptor at the terminal stage, but my guess is that the warhead would be destroyed by the impact rather than detonate.

      Based on what we know I think India’s weapons are one-point safe with effective, if not top of the line PALs, Pakistan’s weapons are almost certainly one point safe with PALs and North Korea probably has at least a primitive PAL system, even if the weapons (if they exist) may not be one point safe. I say “if they exist” because I’m not sure they have been weaponized into something that can be dropped or launched, and so may more properly be devices.

      Despite the cold war, persistent rumors suggest the US was willing to help new nuclear weapons states implement at least primitive PAL systems. I think the idea was that while we didn’t like them having the bomb in the first place, now that they did we REALLY didn’t want them to have one accidentally go off. Kind of like if you’re going to keep a gun in the house, at lease store it with the safety on. I’ve seen documents posted on the National Security Archive that report on the discussion about giving this technology to Pakistan when they developed their bomb, but in the end the US decided not to share. The fact that they were even having the discussion indicates to me that the technology was probably shared, or at least talked about being shared in the past.

  2. David Scott (History)

    There are, of course, different categories of nuclear weapons ‘accidents.’ The Damascus incident represents a technical failure in which the operator loses full control of the weapon. Transport increases this risk. Assuming the weapon does not detonate, clearly a far bigger disaster, adverse outcomes include contamination, misinterpretation of the event by an opponent or permanent loss of control, ie theft.

    As nuclear proliferation creeps forward the risk of a motor vehicle or aircraft accident putting a nuclear weapon into play slowly increases. How secure are the weapons of Pakistan and India, for example, while in transit and should we assume their PALs are up to the task of keeping them secure? Even if they are what countries would bet a major city on the invulnerability of these nation’s PALs?

    • krepon (History)

      Those who know about the state of PALs in Pakistan and India aren’t talking. Those who are talking acknowledge how little they know.

  3. Fred Miller (History)

    “Accidents can trigger decisions”, but many other factors can influence decisions: JFK’s exposure to popular media curbed his willingness to risk nuclear war over Cuba. Protests in Germany in the 1950s constrained Eisenhower’s decisions around nuclear weapons strategy and deployment in Europe.

    The total scientific work that has been published addressing the effect of nuclear war on local and global climate is still very modest. Would leaders of nuclear states be as willing to fund large nuclear arsenals if they knew that 90% of scenarios in which nuclear weapons are used would lead to massive, negative climatic consequences?

    Reducing the number of nuclear weapons accidents can happen entirely within the nuclear weapons bureaucracy. Increasing decision makers’ awareness of the full cost of nuclear explosions on public health, environmental stability and the global and national economy happens in many spaces,from news and entertainment media to political campaigns. It may lead to specific goals being met, such as a particular treaty being signed or a weapon being cut from the budget. It may also lead to consequences not specifically foreseen, such as hawkish leaders pursuing dovish policies and regimes losing vital public support.

  4. Cthippo (History)

    The fact that there has never been an accidental detonation despite all the opportunities throughout the years, I think speaks to safety being an inherent design parameter in everybody’s bomb. If there is an accident, it’s going to happen in the owner’s territory (unless you’re flying airborne alerts) and so safety elements have been built into every device since the first one.

    The other relevant fact is that it’s just not that easy to get a nuke to go off. Lots of things have to happen in a very specific order and if any of them fail you get a small chemical explosion and a mess to clean up.

    An accidental detonation of a nuclear weapon remains a possibility so long as the weapons exist in an assembled form, but I think the threat is far less than most people realize.

    • Kevin (History)

      “…safety elements have been built into every device since the first one.”

      — a little sanguine for my taste.

    • Cthippo (History)

      A little sanguine, perhaps, but also literally true.

      Fully half of the weight of the Hiroshima and Nagasaki was their armored casings, built to withstand a direct flak hit on the B-29 carrying them. Starting with the Mark 3 design, in-flight insertion (IFI) of the core was utilized so that the weapon would remain in a nuclear safe configuration until it was ready to be dropped. The use of IFI designs only stopped when the newer half crit and fractional crit designs came out which were experimentally determined to be one-point safe.

      More empirically, the fact that there has never been an accidental detonation of a weapon despite them having been burned, blown up, dropped from altitude, blown out of silos, etc etc, says something about the inherent safety of the designs.

      That’s not to say that nuclear weapons are 100% accident proof, nothing made by humans is. They are, however, probably just about the most safety engineered objects on the planet. Everyone recognizes that they are every bit as dangerous to the nation wielding them as to the nation being targeted, and for the most part they are treated appropriately.

    • Captain Ned (History)

      Do we really know what happened over Goldsboro, NC in 1961?

    • krepon (History)
    • Jonah Speaks (History)

      Given that designs for nuclear weapons are classified, each new nation must come up with its own design for each new weapon type. Obviously, not blowing yourself up is a strong motivation for designing in safety features, but mistakes do happen. Also, some nations may be more risk-tolerant or mistake-prone than others, particularly if the nation is in a hurry to acquire nuclear weapons or lacks adequate engineering talent.

      The absence of accidental nuclear detonations, so far, rules out high-frequency probability (e.g. once per year or once per decade), but does not rule out low-frequency probability (e.g., once per century or once per millennium).

    • Kevin (History)

      Cthippo –

      I rather imagine it was pucker time when the Enola Gay left Tinian, at least until it cleared the runway without crashing.

      On a contemporary note, I the CHE-IHE “debate” would be nonexistent absent concerns (real? imagined?) over accidental detontation under abnormal circumstances.

    • Cthippo (History)


      The debate over using insensitive High Explosives (IHE) versus Conventional High Explosives (CHE) is an interesting one, but nuclear detonation in an accident isn’t part of it. Modern designs are one-point safe regardless of the type of explosives used, and so in an accident scenario the worst that’s likely to happen is that the chemical explosives will detonate and scatter the plutonium around.

      IHE has been used in conventional aircraft bombs and missiles since sometime before the Vietnam war and this has led to crash rescue crews being trained that they have 10 minutes to fight the fire and perform rescue operations even around bombs that are fully engulfed in fire. One of the reasons that the fire on the USS Forrestal was so devastating is that the deck crews had been trained this way, but didn’t know that there were bombs dating from before WW2 mixed in the magazine, and these weapons detonated during the time period when the crash crew thought they were safe to fight the fire.

      The difference between a weapon using IHE versus CHE is that the IHE are harder to get to explode, and release less energy when they do so. In an aircraft crash, fire, silo explosion, etc a weapon made with IHE is less likely to have the explosives detonate than one made with CHE. Once again though, in neither case is the detonation of the chemical explosives going to lead to a nuclear detonation.

      The tradeoff is that the IHE’s slower burning rate reduces the efficiency of the compression in the primary, reducing yield. In order to get the whole weapon into a size suitable for use as a MIRV or on a tactical aircraft the primary has to be very very efficient in order to ignite the secondary, and some stockpiled weapons used CHE in place of IHE in order to reach this threshold. Part of some of the life extension programs for stockpiled weapons involved replacing the CHE with IHE, which would reduce the likelihood of a chemical explosion in an accident scenario, but also increases the possibility of the primary not producing enough energy to ignite the secondary resulting in a fizzle if used. Since we can’t test (at full scale anyway) whether the IHE would have sufficient energy to reliably meet the compression threshold necessary for weapons function it becomes an added risk.

      The issue of IHE vs CHE is one of risk of accidental chemical detonation versus weapons reliability. the likelihood of a nuclear detonation in an accident scenario is equally small with either type.

  5. tobias piechowiak (History)

    So which country is most prone to an accident like that?
    Pakistan, India?
    The article made me remember the story of lieutenant
    Stanislaw Petrow who probably avoided nuclear war…

  6. Jonah Speaks (History)

    “But it is really better to consider accidental war as the deterrence problem, not a separate one…”

    This may be an example of Schelling vs. Schelling. I believe it was Schelling himself who talked about the risk of things getting out of control, and the alleged usefulness manipulating that risk, if it would cause one’s adversary to back down. In other words, “accidental” nuclear war could be the by-product of deliberate risk taking, not easily deterred…

    I believe this was partly the justification for NATO to threaten or risk nuclear war if the Soviets invaded Western Europe. It was a threat that left too much to chance. I hope no nuclear-armed adversary starts to think that way today.

  7. krepon (History)

    Cthippo and others:
    Many thanks for your comments, which are very helpful.
    Is it fair to presume that the controls and safeguards you describe are less rigorous for a first generation, relatively small (in size and punch) tactical nuclear weapon? How about a first generation tactical nuclear weapon that has been armed?

  8. Cthippo (History)

    My feeling is that any deliverable nuclear weapon would have at least some sort of PAL because 1) it’s important and 2) a PAL is simpler to build than, say, the firing circuit.

    As for things like accelerometers and barometric pressure switches, on the one hand they’re much cheaper and more redially available today, but on the other there has been a move away from high altitude bombing as a delivery system, so… Not sure.

    For a missile warhead I would expect at a minimum a PAL and an accelerometer switch due to the relatively high likelihood of launch failure. If I were designing it I would use a PAL to unlock the warhead, an acceleromter to verify that it had actually launched, and a basic timer to guarantee a certain distance from the launch point before detonation is possible. All these parts are easially available on eBay and High School electronics student should be able to design the system.

    On the other hand, a first generation weapon is far less likely to be one point safe and so there is a greater risk of a small nuclear yield in a crash scenario. I really don’t know of an emerging nuclear power would put the effort into things like in-flight insertion, or if they would just not put the weapon on the aircraft until they were really really ready to use it.

    One of these days I would like to have a collaborative discussion with the other hardware folks on here about what the likely capabilities and specifications of a modern first generation weapon would look like.

  9. Bradley Laing (History)


    The prevailing Western view of Russia’s Cold War strategic nuclear weapons policy is that it resulted from a two-part interplay between the leaders of the Communist Party and the military. Steven J. Zaloga has found that a third contributor—the Russian defense industry—also played a vital role.

    Drawing from elusive Russian source material and interviews with many proud Russian and Ukrainian engineers, Zaloga presents a definitive account of Russia’s strategic forces, who built them, and why. The book is the first in English to refer to the weapons by their actual Soviet names, providing the bedrock for future works. Helpful appendices list U.S., NATO, and other designations, and the illustrations provide clear visual references

  10. Bradley Laing (History)


    This 6000 tons of nuclear submarine can only carry four k4 class ballistic missile or 12 k-15 ballastic missile’s with range of 750km with 1 tons warhead. This proves the existence of a soft policy on the nuclear matters by India , though a new government front runner have put a thought of abolishing this policy, but they were quick to retract from there statement

  11. Bradley Laing (History)


    Russia is reportedly planning to create a unified naval base network on its Arctic territories to host warships and submarines, in a bid to bolster the protection of its interests and borders.

    Speaking at Russia’s Security Council meeting, Russia President Vladimir Putin said that the country needs to bolster its military infrastructure.

    “In particular, to create in our part of the Arctic a unified network of naval facilities for new-generation ships and submarines,” Putin said.