Reading these accident reports persuaded Stevens that the safety of America’s nuclear weapons couldn’t be assumed. The available data was insufficient for making accurate predictions about the future; a thousand weapon accidents were not enough for any reliable calculation of the odds. Twenty-three weapons had been directly exposed to fires during an accident, without detonating. Did that prove a fire couldn’t detonate a nuclear weapon? Or would the twenty-fourth exposure produce a blinding white flash and a mushroom cloud? The one-in-a-million assurances that Sandia had made for years now seemed questionable. They’d been made without much empirical evidence.
Instead of basing weapon safety on probabilistic estimates, Stevens wanted to ground it in a thorough understanding of abnormal environments — and how the components of a nuclear weapon would behave in them. During a single accident a weapon might be crushed, burned, and struck by debris, at a wide range of temperatures and velocities. The interplay among those factors was almost impossible to quantify or predict, and no two accidents would ever be exactly the same. But he thought that good engineering could invent safety devices that would always respond predictably.
Bill Stevens hired half a dozen staff members to explore how to make nuclear weapons safer. Stan Spray was one of the first Sandia engineers to be recruited, and he soon led the research on abnormal environments. Spray had been concerned about weapon safety for years. While visiting the Naval Ordnance Test Station near Cape Canaveral, Florida, he’d watched a bent pin nearly detonate an atomic bomb during a routine test. The accident could have obliterated a large stretch of the Florida coast. In the early 1960s Spray investigated a series of electrical faults in nuclear weapons, analyzing more than a dozen anomalous events prompted by crashes, handling mistakes, and design errors. He had a rare ability to focus intently on a problem for hours, to the exclusion of almost everything around him, until it was solved.
Spray and his team began to gather components from existing weapons and subject them to every kind of abuse that might be encountered in an abnormal environment. It helped that Sandia had the world’s largest lightning simulator. Ever since Donald Hornig babysat the first nuclear device during a lightning storm, the night before the Trinity test, various forms of electromagnetic radiation had been considered a potential trigger of accidental detonations. The Navy tested many of its weapons by placing them, unarmed, on the deck of an aircraft carrier, turning on all the ship’s radars and communications equipment, and waiting to see if anything happened. The electroexplosive squibs of a Navy missile detonated during one of those shipboard tests — and similar squibs were used in some nuclear weapons. By 1968 at least seventy missiles with nuclear warheads had already been involved in lightning accidents. Lightning had struck a fence at a Mace medium-range missile complex, traveled more than a hundred yards along the fence, damaged three of the eight missiles, and knocked out the power to the site. Each missile carried a Mark 28 thermonuclear warhead.
Four Jupiter missiles in Italy had also been hit by lightning. Some of their thermal batteries fired, and in two of the warheads, tritium gas was released into their cores, ready to boost a nuclear detonation. The weapons weren’t designed to sit atop missiles, exposed to the elements, for days at a time. They lacked safety mechanisms to protect against lightning strikes. Instead of removing the warheads or putting safety devices inside them, the Air Force surrounded its Jupiter sites with tall metal towers to draw lightning away from the missiles.
Stan Spray’s group ruthlessly burned, scorched, baked, crushed, and tortured weapon components to find their potential flaws. And in the process Spray helped to overturn the traditional thinking about electrical circuits at Sandia. It had always been taken for granted that if two circuits were kept physically apart, if they weren’t mated or connected in any way — like separate power lines running beside a highway — current couldn’t travel from one to the other. In a normal environment, that might be true. But strange things began to happen when extreme heat and stress were applied.
When circuit boards were bent or crushed, circuits that were supposed to be kept far apart might suddenly meet. The charring of a circuit board could transform its fiberglass from an insulator into a conductor of electricity. The solder of a heat-sensitive fuse was supposed to melt when it reached a certain temperature, blocking the passage of current during a fire. But Spray discovered that solder behaved oddly once it melted. As a liquid it could prevent an electrical connection — or flow back into its original place, reconnect wires, and allow current to travel between them.
