The Secret in Building 26

by Jim DeBrosse

  In spite of considerable effort, no other messages have been found to break on these keys or any component part of the keys. From the keys recovered and from the unsuccessful effort applied, it seems probable that each U-boat is equipped with a set of keys having the following properties: a) they are unrelated to daily keys; b) they are unrelated to keys for any other U-boat; c) they involve a change of wheels or steckers, or both, at least every 10 days. . . . Under the most favorable conditions, the breaking of each Sonder message is equivalent in difficulty to breaking a new month [of keys in another system], and the situation is further complicated by an almost complete absence of cribs.

  A few Sonder keys were broken before the end of the war, but only with heroic effort. A March 1945 attack alone took 5,300 Bombe hours. Again, German procedures saved the Allies. Sonder traffic carried important messages, but fortunately it remained a small percentage of all U-boat transmissions. The logistics of introducing the new system to its entire U-boat fleet must have seemed too daunting to the Germans.

  OP20G and GCCS also found that a potentially devastating fast-flash radio-transmission system called “Kurier” was not being exploited by the Germans. If they had used it more often and more wisely, its short transmission bursts—usually lasting less than half a second to send a complete message—could have thwarted Allied radio interceptors and even direction-finding equipment.

  WHILE ALLIED CODEBREAKERS wrestled with Sonder, the U-boat threat again surfaced in the Atlantic in the late fall of 1944. Sinkings of U-boats dropped from more than 70 percent of operating vessels in August 1944 to 31 percent in September and to less than 25 percent in November: the Germans were again building more boats than they were losing.

  After the disruption in operations caused by the D Day invasion, the U-boats were again on the prowl, many of them equipped with the new Schnorchel apparatus that allowed them to stay submerged longer. In the winter of 1944–1945, an average of nearly forty U-boats per day menaced the ocean lanes, and the loss of Allied merchant vessels began to rise. December brought a 75 percent increase in tonnage lost, followed by another 20 percent increase in January. The losses remained at more than sixty thousand tons a month until the end of the war. The Kriegsmarine was sinking more merchant ships in the Atlantic and Arctic per month—an average of thirteen each month in 1945 from January through April—than in any month since July 1943. Even so, the shipping losses were far short of the disastrous toll in March 1943.

  The Allies were concerned about the Atlantic turnaround, but there are no documents to indicate the kind of panic the British and Americans had felt in 1942–1943. At any rate, the Allies regained the upper hand in March 1945, when U.S. bombers began concentrated attacks against the U-boat pens in Germany and the Baltic. The Kriegsmarine paid dearly for its renewed attacks on Allied shipping. By April 1945, the Germans were sacrificing a U-boat for every four merchant ships sunk—a doubling of their loss ratio from the year before.

  But for Joe Desch and his team at NCR, the Atlantic had ceased to be the primary battlefront. They had turned their attention to the even more punishing developments in the Pacific.

  November 1944—Dayton, Ohio

  SITTING ALONE IN the communications sanctum of Building 26, Joe Desch knew as he hung up the secure phone line to Washington that the information he had just given his superiors at OP20G would mean the deaths of thousands of men. What Desch had told Washington was the solution to a technical problem involving one of the Japanese codebreaking machines, a task that had consumed him nonstop for the past twenty-four hours. For his pains, Desch had been told the purpose and the urgency of his work: it was needed to locate a Japanese convoy laden with troops. Although there are no records to confirm it, the advice OP20G sought may have been related to the decryption of a November 14 intercept in the Philippines, a request for an air escort for two convoys transporting the Japanese 23rd Infantry Division from Manchuria to Luzon. The decrypt gave the precise course and location of the two convoys. Now, U.S. Navy submarines, seeking revenge for the savage kamikaze attacks in the Battle of Leyte Gulf just weeks before, were lying in ambush, ready to send the Japanese troopships and their escorts to the bottom of the Philippine Sea.

  Surprise and vengeance in the Pacific had been the name of the game since Pearl Harbor. It seemed to Desch the madness and cruelty—and the constant pressure from the Navy—would never end, “that damned, dirty business of the war,” he told his daughter years later.

  He snapped. After nearly two years of fourteen-hour days and impossible production deadlines, he walked quietly out of the communications room and out of Building 26, determined never to return. His defiance was the culmination of years of stress, both at home and at work. As he told his daughter not long before his death: “They pushed me and pushed me and pushed me, and told me I had to get all this stuff done because our guys were dying.”

  But in November 1944, the war was far from over and there was still more work to be done—work that came looking for Desch, no matter how hard he tried to escape.

  Tailed by his security shadows, Desch went home that evening and collapsed into bed. Early the next morning, he called up his lifelong friend Mike Moran and asked him a favor. Then he loaded a sledgehammer, spikes, and an ax in the trunk of his car and drove twenty miles east of Dayton to Moran’s family farm, not far from Xenia. In the surrounding thickets, he set himself to a simple, thought-numbing series of tasks: clearing out dead trees and splitting wood. When the trunk of his car was filled, he drove home again to the cottage on Greenmount Avenue and burned his day’s labor that evening in the fireplace, sitting and talking for hours with Dorothy. It was an intimate scene that even his houseguest and dedicated watchdog, Commander Ralph Meader, dared not intrude upon.

  Nearly every day for the next six weeks, Desch went out to the Moran farm and repeated this ritual of oblation—until one day Joe Wenger, the head of OP20G, came from Washington and approached him there. Wenger, who had had his own breakdown, must have understood what Desch was feeling.

  Desch and Wenger weren’t the only OP20G managers who became victims of their top-secret duties. Howard Engstrom, head of OP20G’s research division, also suffered under the weight of the U.S. Bombe project and from the intense guilt he felt over his role in the aerial assassination of Japanese admiral Isoroku Yamamoto, the naval genius behind the surprise attack on Pearl Harbor. On April 14, 1943, a Japanese message was intercepted and broken revealing that Yamamoto, flying under escort of six Japanese fighter planes, was due to arrive on April 18 for an inspection tour on Ballale, a small island in the Solomons. Sixteen P-38 fighters were dispatched from Henderson Field on Guadalcanal, hastily fitted with long-range fuel tanks for the thousand-mile round-trip journey. Over open ocean and in daylight, with no stars to help them navigate, the pilots relied on dead reckoning and miraculous luck in spotting Yamamoto’s airborne entourage. They shot down three Zeros and both Japanese bombers, including the one in which Yamamoto had been flying.

  During a self-interview in 1992, Engstrom’s daughter, Kristina, broke down several times while recounting her father’s wartime experience:

  You know, there were people who were victims of the war that weren’t just shot. There were people like [my father]. . . . His secretary [after the war] said, the last time I talked to her, that Daddy used to talk about Admiral Yamamoto and how terrible he felt about his death and that he felt somehow responsible for it because he had been involved in breaking the Japanese code and that’s how they knew where Yamamoto was.

  Historians appear not to have studied the high frequency of breakdown among codebreakers, and national security officials decline to discuss the issue. But Robert Hogan, a pioneer in personality psychology and a hiring consultant based in Tulsa, Oklahoma, said the long hours of challenging mental work, the emotional strain of war production, and the isolation from families and friends demanded by high-security jobs would have pushed even the most stable personalities to the brink. Added to t
hese external pressures, he said, engineers and research scientists often have a collection of personality traits that make them even more susceptible to stress.

  Members of both professions typically score very high on standard personality tests for “prudence conscientiousness,” Hogan said. As a result,

  they experience a lot more stress than they should because they’re so concerned with meeting deadlines and doing a good job. They really need to learn how to relax, lighten up, back off, but often they can’t. This was what Freud was all about: the essence of psychoanalysis was reducing the negative effects of the superego. The superego was all about very strong rules, about getting things done. If you put a person in a high-stress job who also is extremely conscientious, that’s just a recipe for a nervous breakdown.

  Even before the start of the war, Laurance Safford, then head of OP20G, noted with concern how the strain of codebreaking work was affecting his small staff. In a September 5, 1940, letter of recommendation for a former staffer, Safford said of the man that

  his industry and attention to duty were carried to the extreme that he suffered a nervous breakdown. This may be understood better when I state that my force experienced two other breakdowns and one suicide during the past two years, due to the strain of overwork. [His] physical condition is immeasurably better than when he left the Navy Department and he should be able to handle routine office work without difficulty.

  Nor were the university researchers and corporate engineers recruited by the Navy during the war always suited for its military style of management, said Hogan, who served on a destroyer during the Gulf of Tonkin crisis in Vietnam.

  I’m a retired naval officer, and I promise you I know the kind of pressures the Navy can bring to bear. These people are really quite ruthless. It’s not about patriotism and a professional military, it’s about advancing your career. If you can get the problem solved, your career looks good. They’ll use anything to motivate you. They’ll threaten you with prison. They’ll threaten you with execution. They’ll use guilt, shame. . . . I promise you it’s part of the [military] culture. They just don’t care about people. People are expendable.

  But Wenger cared enough about Desch and the course of the war effort—and, yes, perhaps his own career—that he traveled to Dayton in December 1944 to talk with him man to man. Wenger told Desch that his country desperately needed him: only Desch had the expertise to tackle an engineering problem that had arisen in cracking the Japanese codes. He had to come back. Desch finally said yes, but on the condition that his hours be limited and his workload confined to the Japanese problem. And he insisted that Meader no longer live in his home. But on that point, the Navy brass wouldn’t budge.

  While Desch had been gone, his assistant, Bob Mumma, was put in charge of the German problems and NCR’s managerial responsibilities for Building 26. His task was eased by having on his staff talented Navy engineers such as Ralph Palmer—IBM’s leading electronics engineer prior to the war—and by the Germans’ continuing inability to implement widespread code changes.

  Even so, the unceasing battle of wits continued to the last days of the war. In April 1945, the last full month of hostilities with Germany, the Kriegsmarine began using a variable starting position on its Enigma machines that made codebreaking far more difficult and time-consuming. The first few days of the new Atlantic keys were tough to break, but they were broken in time to read the German order setting up the final wolf pack of the war. Allied forces sank at least three of the pack’s six U-boats.

  An American intercept of one of the war’s last U-boat messages also helped calm fears about Japan gaining the Nazis’ most advanced weaponry. Although Germany was near to surrendering, some of its leaders had planned as a final act of vengeance to provide Japan with the details of its devastating V-2 rockets and jet-powered fighters, far faster than anything the Allies could put in the air. On April 15, a month before the surrender, U-234 was dispatched from Germany to Japan. On board were a high-ranking German general, nine other German naval and Air Force officers, two civil engineers, two Japanese, and a “valuable cargo of plans,” according to a message report from U.S. naval command to Britain’s Admiralty. While the U-boat was en route to the Far East, the Germans officially surrendered and ordered their submarine skippers to give up the fight. U-234 commander Johann Heinrich Fehler decided to obey the order, but the two Japanese passengers felt they had no other choice: they committed suicide rather than surrender to the Americans. The scientific papers never reached Japan.

  After the war, when teams of Allied investigators seized documents and interrogated German codemen, it was discovered that the Nazis, if they had shown better foresight and less confidence in their cipher systems, could have rendered the Bombes useless. They had designs and prototypes for machines such as the “39” version of the Enigma that would have been as secure as America’s ECM encrypting machine, which used fifteen rotors as well as irregular rotor movements to baffle its enemies.

  The German mistakes allowed Building 26 to concentrate on the pressing Japanese problems.

  January 1945—Dayton, Ohio

  WHEN DESCH RETURNED to work at NCR, it was clear that Japan would eventually be defeated. The questions were When? and At what cost? The invasion of the Philippines was turning out to be more difficult than expected, and casualty rates soared as war-weary U.S. Marines were forced to root the Japanese from caves and tunnels on island strongholds. All intelligence pointed to Japan’s military leaders committing that nation to a suicidal last stand, with hundreds of thousands of Allied casualties the price of invading Japan itself.

  OP20G and its centers in the Pacific faced a formidable foe, with an uncanny habit of changing its code systems just before a major strategic move by either side. A significant intelligence blackout hampered the Allies’ Marianas campaign in mid-1944 because of such a well-timed Japanese change. OP20G feared a repeat of that experience. It realized as well that, as U.S. forces pressed closer to Japan, its processing capabilities might be overwhelmed by too many intercepts.

  Wenger and Engstrom felt a duty to provide advanced technologies for attacking the Japanese codes, especially JN-25, which was the key to Japan’s fleet operations, and the daunting weather code, JN-37, which could provide vital information for the U.S. Navy and Army Air Forces groups operating in the region. Britain and the United States had developed efficient methods for attacking JN-25 and at least the fundamentals of an attack on JN-37, but these were tedious and time-consuming. What was needed now was machinery faster and much more powerful than even the special NC tabulators.

  The first to get attention was the additive problem of the Japanese codes. In November, OP20G launched its program to build the Selector, an advanced codebusting machine that used electronics and a massive memory. The Selector added numbers to the code groups in a message—the reverse of the subtraction process performed by the Japanese code clerks. Then the Selector instantly searched a huge memory bank holding already known code groups for matches.

  Desch and his team again had to admit that a fully electronic device was too costly and unreliable, and they soon turned to a combination of huge banks of telephone relays, IBM card readers, and electronics. The group, led by IBM electronics expert Ralph Palmer, worked quickly and well. By February 1945, the concept began to turn into hardware. In May, the first testing was performed, and a month later the Selector—big enough to fill a grade-school classroom—was making its first contributions. By the end of the war, the device had proved valuable though slow, and a new contract was awarded to NCR to try to build an expanded and more electronic model.

  Desch may also have played a role in creating plans for a faster device to break the JN-37 weather code. Despite the use of a high-speed film machine called Hypo, cracking the system took too much time for its information to be of any operational use. OP20G decided in June 1945 to follow a plan suggested by John Howard and build a set of high-speed, film-optic scanning machines, each with the ca
pacity to compare three to five billion code lineups in a single day. Before the million-dollar contract was let to Eastman Kodak, however, the newly conquered Japanese islands, where the Allies were now entrenched, began providing enough weather information in the Pacific that the machine was deferred. The code was no longer worth breaking.

  August 14, 1945—Dayton, Ohio

  EVELYN VOGEL WILL never forget that evening, how it started raining soon after the news of victory over Japan reached the WAVES at Sugar Camp and how it wouldn’t stop raining—and yet the summer torrent was no match for the joy of the occasion. Thousands of revelers headed downtown to drink, dance, and run barefoot through the puddles in the streets. “Commander Meader said we [WAVES] were all free to leave camp and join the celebration, but to remember our manners,” Vogel said. “Anyway, we all went downtown, and I had never seen such a scene. Rain all over, but people were shouting and laughing, and, of course, open bottles were being passed around.”

  Vogel ran into some other WAVES who had made their way downtown, and they decided to head back to the closest thing they knew to their family home: the Desches’ cottage on Greenmount Avenue.

  It’s hard to know how much Joe Desch savored the victory that night and whether he had properly gauged the significance of his own contribution to ending the war. Certainly, he must have been relieved that the violence was at last over. Desch did not discover until years later that he had been part of one of the most brutal and decisive acts of the war. The atom bombs that destroyed Hiroshima and Nagasaki late that summer were developed with the help of a high-speed electronic counter he had designed earlier in the war.