Issue 34 Testing Summer 2009

To Catch a Bug

Michael D. Gordin

They call it the “bug-catcher,” even though bugs don’t fly at 18,500 feet. Humans probably shouldn’t, either—not in these conditions. Air Weather Service B-29s, only a few years after the Second World War, are doing double-duty missions from Alaskan air bases. Their primary function is to record weather patterns in the Arctic, over Japan, and along the Pacific border of the Soviet Union. No mystery why you might eventually want to know that. It’s some compensation for the icing fog, frozen carburetors, inadequate polar charts, and magnetic anomalies that make flying torture.

But they also have a secondary, more bizarre, mission: manipulating a Rube Goldberg contraption held together by spit, sealing wax, and immense effort and ingenuity. In an unpressurized section of the plane, two separate devices are routinely shoved out through the fuselage into the icy air—changed mid-flight, too, exposing the crew to -60˚F conditions every few hours. Each device (two per plane, and several planes a day) consists of a slab of 9 x 22-inch activated charcoal squeezed between two layers of open-mesh wire screen, designed to filter the particulate dust of the upper atmosphere. This is the so-called bug-catcher. It is not a term of affection.

What would you call a test that you can’t really fail? I don’t mean that you are guaranteed to pass it, or that it is easy, but a test where the term “failure” is like dividing by zero: it is strictly undefined. Is this really an examination, or an assay, or a sampling? What’s the point of taking a test that you can either pass or, if you don’t, will constitute a tremendous waste of time and resources but still won’t count as a failure? What if passing isn’t entirely up to you, but depends on your ability to determine whether someone somewhere far away has passed an entirely different kind of test, one where failure and success are both well defined? Now imagine that you really, really don’t want to know the answer.

It is 1949.

* * *

The real test wasn’t taking place in America; in fact, the Americans did not know whether it was taking place at all. The point of the bug-catcher was to—maybe, possibly, conceivably—detect a Soviet nuclear blast from its telltale signatures. Those were essentially two: either you focused on the fact that a nuclear blast was a Really Big Explosion and built something to measure things like shock waves; or you focused on the fact that a nuclear blast was Really Radioactive. In an atomic explosion, the extremely hot fireball would scoop up dust and earth into the column of the mushroom cloud, where molten fission products would condense around the dust particles and then rise on the hot winds into the upper atmosphere, where they would be carried on prevailing winds. Assuming the particles were carried far enough and were present in a density large enough to be collected, they could be filtered out of the air and then presented for chemical analysis.

The bug-catcher was the central instrument in what the recently minted United States Air Force (made autonomous from the Army in a unified Department of Defense in 1947) called Operation Whitesmith, and later renamed Bequeath. Such name changes were standard procedure for ensuring the continued secrecy of military projects, but nobody was especially concerned in this specific instance. Most American estimates, including the official top-secret assessments of the fledgling CIA (also a 1947 creation), predicted that the soonest possible Soviet test blast would occur in 1951, and more likely in 1953. The Americans possessed an absolute monopoly on the atomic weapon, and the various desks scattered throughout the American bureaucracy interested in Soviet progress toward their own bomb—the Atomic Energy Commission, State Department, Defense Department, National Security Council, CIA, Joint Chiefs of Staff, and so on—had no direct sources of information (that is, no spies). If they wanted to know what was happening behind what Winston Churchill had recently dubbed the “Iron Curtain,” they had to go the indirect route.

Reading through popular accounts of atomic-test detection both at the time and since, one might come away with the impression that finding out whether there had been a nuclear test in the Soviet Union was much like checking one’s answering machine—a light beeps, you press a button, and you learn the news. In reality, it was much more like dusting the entirety of the Empire State Building for fingerprints every day on the off chance that the fingerprints of a specific individual might show up, someone you do not expect to arrive for roughly a decade. Except it was even more complicated than that; for “Empire State Building,” substitute “atmosphere of the northern hemisphere.”

This program had been rendered operational, as an interim stopgap measure, in April 1949, mostly to assuage Lewis L. Strauss (pronounced “Straws”), a self-described “Hoover Republican” who represented President Harry Truman’s effort for bipartisanship on the infant Atomic Energy Commission (1947 again). Since assuming office, he had been pushing for some form of airborne radiological detection—specifically airborne—even though most of the cognoscenti pointed out such a method would be provisional, unreliable, and easy to fool (by detonating an explosive device at high altitude to minimize fallout, for example). This did not deter Strauss, a veteran at manipulating the slow levers of military bureaucracy. By spring 1947, bug-catching flights ran daily.

Strauss pushed the Air Force into the air to look for bits of earth that had passed through atomic fire. The United States Navy, the long-standing rival to the Air Force’s monopoly on matters nuclear, completed the compass of Aristotelian elements. They would look for Soviet nukes in water.

Parallel to Bequeath, and running at just over ten percent of the cost ($150,000, the equivalent of $1.34 million in 2009), was project Rainbarrel, initiated by the Naval Research Laboratory (NRL). Shortly after the Trinity nuclear test in the deserts of Alamogordo, New Mexico, on 16 July 1945, scientists at Eastman Kodak discovered that batches of film had become contaminated by stray lines and spots. A little investigating yielded the source of the spotting: radioactive strawboard used to package the film, rendered hot by the water in Midwestern rivers that had been used in their composition. The NRL followed the logic: rainwater brought dust particles down from the troposphere without having to fly up to fetch them. (The disadvantage, obviously, was that it was dependent upon good—or poor, depending on your perspective—weather conditions involving lots of precipitation.)

The brainchild of Roger Revelle in spring 1947, Rainbarrel was another episode in the Navy’s struggle not to be shut out of nuclear matters by the Air Force. Seventy-five measuring devices, each consisting of a nest of seven large Geiger counters hooked up in anti-coincidence (to eliminate the natural background from cosmic rays), were operational by the time of the second wave of postwar nuclear tests—the Sandstone shots of 1948. In June of that year, NRL scientists obtained water from the Virgin Islands (whose entire water supply was provided by rainwater trapped in tremendous cisterns) and, using five-hundred-gallon decontamination trucks from the Chemical Weapons Service, treated the water to settle the flocculent and siphon off the clear water. After treating 2,500 gallons, they obtained about five gallons of “floc,” which was found to contain the fission products yttrium-91, cerium-141, and cerium-144 in ratios indicating origins in the Sandstone tests. The Navy passed their information on to AFOAT-1, the division of the Air Force assigned nuclear detection by then Chairman of the Joint Chiefs of Staff Dwight D. Eisenhower. They received little information in return, but that did not stop them from setting up their own storage tanks to gather rainwater for testing. It was an iffy solution to a problem that not everyone was certain needed to be resolved, but at least it was in place.

Meanwhile, several times a week, on multiple routine flights between Alaska and the Soviet Pacific coast, scientists and soldiers dusted the world for fingerprints. By the beginning of September 1949, alert levels of 50 counts per minute (cpm) had been exceeded 111 times. Each time the Geiger counters in Alaska registered an alert level, the graphite filters were whisked to Berkeley, California, where a private subcontractor named Tracerlab analyzed the particulate matter for trace radioactive isotopes. So far, nothing that couldn’t be explained by natural causes, such as radioactive dust hurled into the atmosphere from volcanic eruptions. One hundred and eleven false alarms.

* * *

On Saturday, 3 September 1949, Lieutenant Robert Johnson, on the Loon Charlie route between Japan and Alaska, flew his B-29 weather plane at eighteen thousand feet while his crew exposed and changed the bug-catchers as usual. Of the filters used on this flight, two were somewhat out of the ordinary. The first, exposed to the winds wafting across the fuselage for three hours, had a Geiger count rate of 85 cpm—enough to trigger Alert 112, but only because the count rate had been lowered from 100 to 50 cpm on 1 August. The second filter registered 153 cpm (over three hundred percent of the alert rate). As per standard operating procedure, more planes were sent up over the next two days to gather samples, and the filters were sent to Tracerlab’s facilities for isotopic analysis.

Doyle Northrup and his team assembled at AFOAT-1’s Data Analysis center in Washington, DC, to await Tracerlab’s report. Meanwhile, data flowed in from other bug-catching flights across the Pacific. On Monday, a plane originating from Guam registered 1,000 cpm at 10,000 feet over the North Pacific. By the time all the data was gathered—a period that lasted less than two weeks—ninety-two previously unscheduled flights had been flown by the Americans, not counting the standard routes, which continued as before. Alert 112 was now given a codename: Vermont.

3:30 AM, Wednesday, 7 September: Tracerlab phoned. Barium and cerium were found in the samples. Before 9:00 AM, the phone rang again: molybdenum. All three of these elements were byproducts of plutonium fission and were present in levels far too high to be explained by a natural event. Between 3 and 16 September, more than five hundred radioactive samples had been gathered, 167 of them above 1,000 cpm. By 9 September, even ground-based units started registering elevated gamma rays. This did not look like another false alarm.

On Thursday, 8 September, William Webster, Deputy Secretary of Defense for Atomic Energy under the new and imperious secretary Louis Johnson, decided that this might in fact have been what they had been searching for: a Soviet nuclear detonation. The Americans were passing their test. Webster called Carroll Wilson, general manager of the Atomic Energy Commission (AEC), and informed him of AFOAT-1’s tentative conclusions. Wilson in turn summoned radiochemist Spofford G. English and the AEC’s new director of intelligence, Walter F. Colby, to examine the data Webster had sent along.

By this time, the radioactive air mass had traveled over the United States and was somewhere over the mid-Atlantic. To continue this American test of the Soviet test, the Yankees needed to peek at someone else’s exam. They had to let the British know. On 10 September, Alex Longair, assistant scientific attaché at the British Embassy in Washington, was rushed into the telex room and informed about the ongoing investigation. He contacted London, where British atomic-energy officials were beginning to intensify their investigation of allegations of espionage against Klaus Fuchs, the head of the theory division at the Harwell establishment, Britain’s principal atomic research center.

(Fuchs, the most notorious atomic spy of the era, had been present at Los Alamos during the war years, working on the development of the first nuclear devices. He had been present at the Trinity test at Alamogordo. Although information he transmitted to Soviet handlers proved in some ways instrumental for the Soviet detonation, he was understandably not invited to that test.)

As the British report on Vermont recalled: “A conference was held by teletype in the American Embassy in London and the British were informed that a mass of air containing activity was about to pass north of Scotland. It was estimated that the activity would be approximately ¼ of a disintegration per minute per cubic foot.” Temporarily putting the Fuchs investigation on hold, they scrambled more planes and intensified their sampling efforts along two routes, Bismuth and Nocturnal, which had their own bug-catching flights. Their conclusion was no less dreary than the Americans’, if perhaps colored by a touch of classic British understatement: “It is perhaps worth noting that the maximum activity collected by the British filters, for flights of equal duration at similar altitudes, was roughly 20 times that obtained for any of the Sandstone tests.”

* * *

Sandstone had been the calibration, the practice examination. Of course, in a practice test you can look up the answers. With Sandstone, the Americans had known when and where the blasts were going to happen, because they had set them off themselves. The Soviets would not be so kind. Even with the heads-up, the practice exam did not go especially well.

For all the importance of the Sandstone atomic tests in making long-range detection even remotely feasible, the tests were primarily tests of bomb designs, not of detection apparatus. AFOAT-1 had to piggyback on the first major postwar innovations on nuclear core design. AEC Commissioner Lewis Strauss, who claimed he was almost forced to front his own money for the instrumentation, understood that these tests would be crucial for solving the calibration difficulty of long-range detection.

The project for testing the proposed detection methods (Operation Fitzwilliam) was directed by Dr. Ellis A. Johnson, an MIT graduate who had taken a post at the Carnegie Institution in 1935 and worked at the Naval Ordnance Laboratory during the war. (By chance, he had been present at Pearl Harbor during the attack of 7 December 1941.) He organized 466 air sampling sorties for a total of 4,944 hours in the air, as well as seismic teams from the Coast and Geodetic Survey for short-range work on Runit, Parry, and Aniyaaii atolls near Eniwetok. Naval Ordnance Laboratory seismographs were distributed at eight different sites in the Pacific, including Kwajalein and Eniwetok. These methods had all been cursorily tested in the 1946 Crossroads shots at the Bikini Atoll, though not with instruments of such sensitivity.

They tested new methods as well. The Signal Corps and the Naval Ordnance Laboratory established sonic sensors of various sorts. The Army set up an array of over twenty acoustical sensors, while the Air Force sent theirs up attached to a series of Project Mogul balloons, the latter launched from Kwajalein (450 miles from ground zero), Guam (1,200 miles), Hawaii (2,750 miles), and even from as far away as New Mexico and Alabama. Two Army Signal Corps teams made efforts to detect any light from the test that might be reflected off the surface of the moon—one of the more far-fetched detection attempts—through telescopes attached to photoelectric devices at Guam and Eniwetok. The Naval Ordnance Laboratory erected magnetometers to measure post-atomic electromagnetic disturbances at Eniwetok and Kwajalein. A hypothesized “ionospheric dimple” above the point of detonation (theoretically detectable through radio-wave interference) was searched for from an installation on Kwajalein. At Los Alamos (5,000 miles), scientists searched for minute changes in the sky’s illumination.

Most of the results were terrible. The Yoke shot (equivalent to forty-nine kilotons of TNT), part of the Sandstone tests that took place two years later, was the largest nuclear explosion the world had yet seen and still could not be detected by seismometers at distances of over five hundred miles away. Ellis Johnson despaired: “In connection with the seismic program … in the present state of the art, there is only about a one to one chance of success in distinguishing between an earthquake and a major explosion.” Sonic measurements were slightly (but only slightly) better, detecting Yoke at 1,700 miles, but Zebra, an eighteen-kiloton Sandstone shot, at only a thousand miles. This was woefully inadequate if one wanted to detect a supposed test in the Soviet heartland.

Among the three major detection methods then, this left the radiological method, collected by airborne filters. The results were better: Flight C of the 373rd Reconnaissance Squadron, Very Long Range Weather, based in Lagens in the Azores (halfway around the world), had definitely registered the US tests. Partisans of long-range detection interpreted the data as a qualified success.

* * *

September 1949 again. Back in the US, Rainbarrel also had something to report. The air monitor at the Washington NRL site received elevated readings on 10 September, and the rain of Tuesday, 13 September, was collected and filtered. The full data from 15 August to 23 September 1949 was plotted and a clear peak emerged. In the final report of 22 September, they noted that the “birthday” of the activity ratios in the sample was “probably not earlier than 24 August.” And the isotopic mix indicated that the explosion was not only a bomb, but probably a plutonium one, a conclusion derived from radium and thorium compounds.

This particular hot potato did not belong to the Navy, however, but to AFOAT-1, and the latter based its conclusions not on rainwater analysis or the British assessments (both rather cavalierly appended to the conclusions), but instead on bug-catching and Tracerlab. For these analysts, the main question to be resolved in mid-September was not whether an explosion had happened—they had few doubts on that score—but when it had happened. They had to find a birthday. To determine this, they had to factor in two different variables: the isotopic mix (which fixed the times of origin and quantities of the fission products), and the wind speed of the cloud of radioactive debris (for the hypothesized location of origin culled from the vast geographic distribution of samples). Finding a birthday produced a judgment about birthplace.

The isotopic mix was comparatively unproblematic. Tracerlab had found barium-140, molybdenum-99, zirconium-95, and protactinium-144, and in quantities sufficient to produce a regression curve. Unfortunately, there was no reliable data for the wind speed and direction over the first few days of the cloud’s trajectory for the obvious reason that the Soviets did not make a habit of releasing this kind of information. In the absence of exact values, AFOAT-1 had to guess, and the only approximation they could make was a bad one—and they knew it. In the absence of better options, the analysts decided to substitute the geostrophic wind for actual wind for large portions of the data. This is the theoretical wind that would result from the exact balance of the pressure gradient (the natural flow of air from regions of high pressure to regions of low pressure) and the Coriolis force (attributed to the rotation of the earth). It is also liable to twenty-five percent errors in either direction.

The results came in the form of several maps of the Soviet Union, colored with crayon to indicate the different probabilities of location across the enormous span of 35˚ and 170˚ east longitude. Factoring in all the uncertainties, they concluded that the probable date for the Soviet explosion was on 27 August 1949. They had a result. They had passed the test.

* * *

Sort of.

Passing the test was a relative matter. The Americans were convinced the Soviet Union had detonated a nuclear device. Convinced enough, anyway, for President Truman to issue a press release on 23 September 1949, announcing the fact to the world. From that point on, the Cold War assumed a familiar course, a course that was certainly not obvious while the test was still undefined. And there were still lots of open questions. The Americans did not know whether it was the Soviets’ first. (It was.) They were convinced, until the mid-1950s, that the date of the test was 27 August. (It wasn’t. The correct date was the 29th, at 7 AM in the deserts of Kazakhstan at an elaborately designed test site called Semipalatinsk-21.)

The real point here is not so much epistemological or geopolitical but emotional. The worst possible news, the fact that the Soviet Union had indeed detonated an atomic bomb, ironically came trailing wisps of relief. For now the anxiety, if not assuaged—certainly not that—at least had a concrete shape. It was a fear you could put words to, something you could shadowbox against. For even worse than failing the test—this test of the existence of a prior Soviet test—was not knowing what the absence of evidence meant. The Americans had imputed meaning to a probe that had left failure un-defined. They had indeed passed.

Michael D. Gordin is a professor of history and history of science at Princeton University. His books include A Well-Ordered Thing: Dmitrii Mendeleev and the Shadow of the Periodic Table (Basic Books, 2004), Five Days in August: How World War II Became a Nuclear War (Princeton University Press, 2007), and the forthcoming Red Cloud at Dawn: Truman, Stalin, and the End of the Atomic Monopoly (Farrar, Straus and Giroux, 2009).

Cabinet is a non-profit organization supported by the Lambent Foundation, the Orphiflamme Foundation, the New York Council on the Arts, the NYC Department of Cultural Affairs, the National Endowment for the Arts, the Andy Warhol Foundation for the Visual Arts, the Katchadourian Family Foundation, Goldman Sachs Gives, the Danielson Foundation, and many generous individuals. Please consider making a tax-deductible donation by visiting here.