The Secret in Building 26

by Jim DeBrosse

  Wenger fretted over the future of OP20G. Without the American Bombe, his agency would have no real contribution to the European war and no way to prove the need for his machine-driven approach to cryptanalysis. The M section, with all its technical expertise, was reduced to using hand methods against the easy parts of the Enigma keys already broken by the British. On the rare occasions when the Americans did solve a key, it took the group an embarrassing average of six hundred hours. Moreover, OP20G had to organize itself to deal with a two-front war. Most of its people and resources had to be focused on the Japanese problems: the Navy was engaged in a death struggle with the Japanese, and OP20G had promised Britain that it would carry the codebreaking burden in the Pacific.

  JAPAN WAS A very different cryptanalytic enemy from Germany. Although Japan used some automatic encrypting machines, such as Purple, its vital naval messages made use of rather old-fashioned paper-and-pencil code-additive systems like the ones OP20G had tackled during the 1930s. Message clerks started with code groups of numbers for specific words and terms, then added or subtracted another set of numbers from another book. A machine program for tackling the Japanese additives had begun even before the Bombe project. But until OP20G’s research crew was freed from the Bombe effort, the Japanese project was the child of IBM, not NCR. Attacking a code system via machine called for lots of memory, and stacks of punched cards, such as those used on IBM tabulating machines, were the most practical mass-memory devices of the era.

  One of the central attacks on additives was based on trying various “subtractions” from the code to see if the results matched any known high-frequency code groups. Tabulators and their kin were obvious choices for the enormous task of sorting and listing the thousands of different code groups. With much labor, they could be used to find strings of repeated text in different messages and, hence, point codebreakers to their underlying meanings.

  With IBM already a major supplier of tabulators and other equipment to codebreakers, it was natural for OP20G to turn to Thomas Watson’s company for help in tackling the Japanese codes. IBM agreed to help design and build a range of special “tabs” to supplement the hundreds of standard machines leased by OP20G during the war. In fact, OP20G became one of the two largest users of IBM machines in the Washington area during the war. The other was William Friedman’s Army group, which had moved across the river from OP20G’s new Nebraska Avenue headquarters.

  Engstrom and Meader may have felt some competitive pressure because of IBM’s ability to deliver its products quickly, but they maintained their faith in NCR and Joe Desch, explaining to their superiors how the Bombe was a much more challenging engineering project. *23 They might have also mentioned that IBM was making a healthy profit from codebreakers through its leased tabulators and punch-card sales while NCR was working on a no-profit basis. Navy and Army codebreaking operations were each paying close to one million dollars per year to IBM.

  Still, if the NCR project was to survive, it had to produce working machines and useful intelligence very soon, if not immediately. In late May, Wenger ordered Desch to ready the two temperamental prototype machines for use on Enigma intercepts to be sent from Washington to Dayton. The results were to be forwarded to the British as examples of America’s new abilities. Worried and hopeful, a Navy delegation from Washington—including the head of naval communications, Admiral Joseph Redman—arrived in Dayton near the end of the month, just as a secure communications line was set up between Building 26 and OP20G in Washington.

  But even with the most attentive coaxing from Desch and his team, Adam and Eve were feisty and unpredictable. They could be made to perform now for up to several hours, rather than minutes, before breaking down, but oil leaks and faulty electrical contacts continued to plague them. Problems as basic as the carbon brushes splintering from their contact with the high-speed commutator wheels seemed intractable. As the summer of 1943 drew near, the Bombe again seemed to be too much of a challenge, even for the talented Joe Desch.

  NOT LONG AFTER lunch on May 28, in a secured inner room at NCR’s Building 26, Phil Bochicchio was conducting a test run on Adam when its high-pitched whine suddenly died, and the machine shut off. At first, he thought it was just one more electrical short.

  But then, as it was designed to do, the Bombe came back to life and began slowly to rewind. That stop suggested that a solution might have been found, but more testing was needed to see if a possible Enigma key had been identified.

  Each Bombe solution was the result of a two-step process. To begin a test, the operator first set the starting positions of the wheels to match a codebreaker’s menu—that is, the suspected pairings between the plain and cipher text and their positions within the message. When the Bombe was started, electrical energy raced through the ever-changing circuits of its spinning wheels, trying all the thousands of various settings that might satisfy the letter patterns in the menu. While the Bombe was going through the many unsuccessful settings, all of the electrical wires were “hot.” But if the conditions of the codebreaker’s menu were fulfilled, one circuit would become “cold.”

  At that point, the machine stopped, rewound back to the suspected setting, and in about fifteen seconds made a second or hot test. It polled the Diagonal Board, looking for inconsistencies, such as two letters being steckered to the same letter. If no such contradictions were found, the Bombe would print a “story,” giving wheel positions and possible steckers. With that in hand, a cryptanalyst used an Enigma mock-up to see if the story led to the cipher message becoming sensible German plain text. If so, a Jackpot was declared.

  On cue, Adam began printing out its results during Bochicchio’s test run. Still, he was skeptical. He had a fellow mechanic, K. P. Cook, run the same menu through Eve. Eve shut off in precisely the same rotor positions, went through its hot-point test, and printed out the resulting sequence of settings.

  “We concluded that both machines were working, because they both gave us the same printout,” Bochicchio said. “But we didn’t know what it meant because we hadn’t been cleared that far” for security reasons. Whatever its purpose, he said, “I thought to myself, ‘That’s a hell of a machine.’ ”

  Bochicchio immediately went to the naval offices and gave the printout to Meader, who then cabled the results to Navy intelligence in Washington, using Building 26’s new high-security encryption machine.

  Three or four days later, Bochicchio said, Meader heard some exciting news back from Washington. “They told us that one hit had been worth the entire cost of the project,” Bochicchio recalled.

  That first hit must have seemed like redemption for Desch and the others who had labored so long and so hard on the NCR Bombe, but it didn’t provide information, as Meader had been told, that the Navy could use. The test menu that had been fed into the two machines had been based on an Enigma intercept that was perhaps weeks old and, OP20G would soon learn, had been decrypted first by GCCS.

  Nor had the NCR team overcome all the technical hurdles still in its path. As Dayton entered its steamy summer months, the engineers were still struggling with the machine’s most basic parts. Those large rotors seemed doomed to overheat, lose their shape, and create faulty electrical signals. The carbon sensing brushes also seemed to have been meant only for slow and dependable tabulators. As quick fixes were made to those parts, more oil leaks developed on the prototypes. Those problems raised such fears about the production model’s design that assembly was halted and all message testing was suspended.

  Meanwhile, in the vast expanses of the Atlantic, the battle of men, ships, and wits couldn’t wait. The tide already was beginning to turn—without the help of the American Bombes.

  8

  U-boats on the Run

  June 5, 1943—Azores area, Central Atlantic

  IT WAS A choice morning for submarine hunting—calm seas, a gentle southerly wind, and just a hint of sea haze on the horizon—as Lieutenant “Goose” McAuslun and Lieutenant Richard S. Rogers each flew
their U.S. Navy aircraft over the vast, empty reaches of the Atlantic, midway between Jacksonville and Morocco. McAuslun’s Avenger torpedo-bomber carried four 250-pound depth charges set to detonate at fifteen feet below the surface, just deep enough to catch a submarine in the midst of crash-diving. Rogers’s Wildcat fighter was equipped with six .50-caliber machine guns, an effective armament for clearing the decks of any surfaced U-boat. But in order to defeat their elusive prey, the pilots had something just as important as their weapons: up-to-date intelligence information, in this case from the British, locating a line of seventeen German U-boats, codenamed “Group Trutz,” lying in ambush for a slow, eastbound convoy of thirty-one Allied merchant ships.

  The pilots’ mission: kill the subs before they could sight the Allied vessels. The pair was on the last leg of a five-hour patrol, seventy miles from their escort carrier, USS Bogue, when McAuslun spotted U-217 cruising placidly on the surface about seven miles to their right. McAuslun signaled Rogers, who immediately banked his Wildcat and dove. “He just tore off and left me in his smoke,” McAuslun recalled. After Rogers strafed the sub three times, killing several gunners and starting a fire in the conning tower, the U-boat commander decided to dive. That’s when McAuslun dropped his tubby Avenger down out of the sun.

  Leveling off at an altitude of one hundred feet, he released all four depth bombs on either side of the sub’s submerging hull. The explosions lifted it out of the water and split it nearly in two. The sub sank in just thirty-three seconds, with Rogers strafing the upright stern one last time before it went to the bottom.

  If Rogers dispatched his duties with a bit more fervor than might have seemed necessary, his excitement was understandable. U-217 was the first Nazi submarine destroyed by the U.S. Navy in a purely offensive action. The day before, making use of the same communications intelligence, pilots from the Bogue had damaged and scattered three other U-boats from Group Trutz. Those two days marked a turning point in the Battle of the Atlantic: the Allies were no longer just detouring convoys to evade the submarines, and their escorts were no longer waiting for the wolf packs to attack. They had evolved the organization and the means to pursue the subs before they could strike. The Allies were no longer the hunted, but the hunters.

  The Allies sank forty-one German submarines in May 1943, a nearly threefold increase from the fifteen sunk the month before. By May 24, Dönitz had had enough: he licked his wounds and pulled his subs out of the North Atlantic convoy routes. While Dönitz and the German cryptologists refused to believe that their Enigma codes were being read by the Allies, they were mostly correct in blaming two other key factors for their submarine war’s demise: the superiority of Allied detection equipment and the omnipresent air cover in the Atlantic.

  The U-boat skippers could run, but they had nowhere to hide. In the four months from May through August, the Allies dispatched 106 Atlantic U-boats to the bottom—21 more than in all of 1942. Of the 42 sunken subs credited to the Americans in the critical summer months of June through September, 21—or half—had transmitted or received messages revealing their locations to the Allies. But Shark decrypts at the time were still inconsistent and slow in coming, limiting their usefulness to sub hunters. In fact, that summer the Americans sank only five U-boats within two weeks of their locations being decrypted, and they were still almost wholly dependent on Ultra intelligence from the British, who relied on their three-wheel Bombes, years of cribbing experience, and captured German short-signal codebooks to break into Shark.

  OP20G did not yet play a major role in Enigma codebreaking. Although Desch’s prototype produced their first real Jackpot in late June, the U.S. Navy did not have the machines, the methods, or the trained people it needed to be a true codebreaking partner for several more months.

  As the summer progressed, Desch and his engineers relied on old-fashioned tinkering to get their machines to behave. They found better ways of protecting the Bakelite wheels from heat, perhaps by improving the Bombe’s ventilating system, which on the early production machines consisted of little more than smokestacks. They also improved their methods of polishing, smoothing, and adjusting the spinning rotors so they were less likely to distort. Phil Bochicchio served up a few important tricks, including soaking the machines’ leather seals in oil before installation, so they would swell and prevent leakage, and installing a circuit tester that checked for opens and shorts prior to running the machines.

  Tired of OP20G’s apprentice role to the British, Wenger and Engstrom pressured Meader, and in turn Meader pushed Desch even harder. If OP20G was too late to be the savior of the Atlantic, there was still much to do to vanquish the U-boats. Despite the period of calm after the spring U-boat offensive, no one was certain of the safety of the convoys—all the more troubling given the massive buildup of men and matériel that would be needed for the planned invasion of Europe.

  Even after the heavy losses of U-boats in May 1943, Hitler and Dönitz showed no signs of scaling back their sub offensive. At a Führer Conference on June 3, 1943, Hitler declared: “There can be no talk of letup in submarine warfare. The enemy forces tied up by our submarine warfare are tremendous, even though the actual losses inflicted by us are no longer great.”

  The Kriegsmarine wasn’t the only worry for Allied codebreakers. Germany’s Army, Air Force, rocket-development team, and police agencies might switch over to four-wheel Enigmas. And those German systems that continued to use a three-wheel machine might easily broaden the blackout by changing their procedures or making more sophisticated use of their plugboards.

  Worse, there were fears of dangerous new Nazi technologies, including a new breed of bigger, faster U-boats and sub-launched rockets that could rain terror on U.S. coastal cities. As Samuel Eliot Morison wrote:

  Although the turning point had been passed by the end of May 1943, it was by no means a foregone conclusion that it would not turn again. First one side would obtain an advantage with a new weapon or tactics; then the other would produce a new device or defensive weapon to counteract it. And always there was bright expectations for the Germans, and apprehensions for the Allies, that . . . secret weapons would burst forth like monstrous messengers from Mars, to sweep Allied shipping from the sea.

  In the summer of 1943, the U-boats faced a new frustration in their attempts to locate and destroy Allied shipping. The two-year, nearly unbroken run the Germans had enjoyed in reading British naval Cypher Number 3, the primary code for the convoys, came to an abrupt end in June when the Allies switched to the more secure Cypher Number 5—a move long urged by OP20G’s Laurance Safford.

  The switch in codes was a classic example of the machinations of spy versus spy. Ultra’s early and limited breaks into Shark more than hinted to the British what they had refused to hear from the Americans as early as 1941: that the Germans had information on convoy locations that could be obtained only through codebreaking. The American suspicions had been confirmed in March 1943 when Lieutenant Mahan, an OP20G analyst in Washington, noticed a German intercept in which Dönitz had abruptly canceled an order of a few hours before and ordered a dramatic change of course for his U-boats. Mahan left his post at the OP20G complex near Tenley Circle and rushed downtown to the convoy-routing department in the Main Navy Building, where he insisted on seeing the latest Allied orders to the convoys. After some bureaucratic haggling, he got his way and soon found the detour order for the convoy that had triggered the change of course by Dönitz.

  If any more proof were needed, the switch from Cypher Number 3 to Cypher Number 5 had instant and dramatic results for the fortunes of Allied convoys. Although the U-boats sank forty-one merchant vessels in May 1943, it was the last good month of hunting for Dönitz and his wolf packs in the Atlantic. From that June until the end of the war, the Germans never sank more than fourteen merchant ships in any one month.

  Allied intelligence victories weren’t the only factors that turned the tide in the Atlantic, nor perhaps even the most important. Much credit must be g
iven first of all to the valor, skill, and persistence of the British Navy and Air Force, particularly for waging a devastating air campaign in the Bay of Biscay—the gauntlet that U-boats had to traverse between their bases in western France and their hunting grounds in the Atlantic.

  But beyond the human factor, a host of improvements in antisubmarine warfare converged against the U-boats: improved air cover, weaponry, and detection; better training and coordination among antisubmarine forces; new methods for analyzing U-boat movements; and more aggressive tactics, including those of the new escort class of carriers like the Bogue.

  By 1943, the fusion of science and the military that Vannevar Bush had sought through the National Defense Research Committee began to pay big dividends for the war effort. Allied technical developments came fast and furious in the antisubmarine war. The old “ashcan” type of depth charge, dropped with a wish and a prayer from the sterns of destroyers, was replaced with a rocket-propelled design with a streamlined casing and a contact fuse. The new depth charges—a larger version called the “Hedgehog” for destroyers and the smaller “Mousetrap” for chase vessels—were grouped on the foredeck and launched in deadly patterns ahead of the ship, before the ship’s sonar passed over the U-boat and contact was lost. Unlike the old depth charges, which detonated at a preset depth, the new projectiles did not explode unless they hit something. Their contact fuses eliminated both the danger to the ship from a shallow, premature explosion and the roiling underwater noise that overwhelmed sonar operators trying to maintain contact with the submarine.

  A variety of new weapons were introduced to enhance the air attack as well: powerful five-hundred-pound depth charges, deadly rockets, and the small homing torpedo known as “Fido,” which could chase submarines with devastating accuracy as they began their dives.