The Secret in Building 26

by Jim DeBrosse

  Two miles from U-177, Pinnell ran into heavy fire from the sub’s flak cannon and twin twenty-millimeter antiaircraft guns. The exploding flak “was so close it sounded like a sledgehammer pounding the cabin,” Pinnell recalled. “There was dust falling down from the ceiling.”

  He took evasive action, wobbling the big plane up and down as he dove, but finally was forced to pull up. “It broke my heart to do that. I thought, ‘Boy, this is a good way to get hit, and we haven’t even dropped our load.’ ”

  Pinnell could have called off the attack and waited for his wing partner but he feared the U-boat would slip away. Instead, he attacked again about three quarters of a mile from the sub—an act of heroism that earned him the Distinguished Flying Cross. This time, the B-24’s bow gunner quieted the sub’s guns just long enough that, at one hundred feet, Pinnell was able to straddle the sub’s hull with three depth charges on each side.

  As Pinnell completed his pass, machine-gun fire flashed at the B-24’s exposed belly—but not for long. By the time he had banked and started his second pass, the force of the depth charges had lifted the sub’s hull out of the water and ripped it apart. As U-177 sank, the water above it was a roiling stew of debris, with only its conning tower still visible a few feet below the surface.

  Pinnell dropped his remaining three depth charges, flew off, and returned again to find fourteen survivors struggling in the water. The B-24 crew dumped a life raft for them, then radioed their position for ships in the area. Pinnell roared back to Ascension Island.

  Despite their late arrival in the war, the American Bombes were making a significant contribution to the destruction of the Nazi U-boat fleet by the fall of 1943. U-177 was the third sighting of a sub in as many months by Pinnell’s squadron alone, all without the aid of radar. Clearly, the decrypts were giving pilots accurate information on their targets.

  November 1943 was when the U.S. Bombes at last began breaking the keys to the Shark code on a routine basis. By December, the average time for gaining entry into Shark was thirty-six hours—down dramatically from the embarrassing early months of 1943 when the Americans had needed an average of twenty-five days to break Shark, mostly by hand. The new average was also better than the British Bombes had been able to do during their best times in 1943.

  By March 1944, the British trusted the Americans enough to transfer the major responsibility for deciphering the Shark keys to OP20G. With new shipments of Bombes coming from Dayton every week, the Americans now had nearly one hundred in operation at the Naval Communications Annex, where the workforce was at last up to speed. It was not only a matter of trust: the Americans had demonstrated that the NCR four-wheel Bombe was more reliable and more adept at getting the job done than the British models.

  The timing was fortuitous, because the job of cracking Shark only got tougher in the months that followed. Admiral Dönitz had decided to end the wolf-pack attacks in the North Atlantic and had dispersed his fleet to hunt vulnerable Allied ships on their own, thus greatly reducing the number and quality of U-boat messages to and from headquarters. Even so, the Americans continued to conquer Shark as well as make contributions to the attacks on several other German naval systems. By the summer of 1944, hundreds of submarine messages were being read the same day, some within minutes of their transmission, giving Allied antisubmarine forces a fresh bead on the subs’ whereabouts.

  With the faster decrypt times generated by the U.S. Bombes, the percentage of U-boats sunk with the aid of Ultra intelligence climbed steeply. In 1943, Ultra decryptions had a role in just 10 percent of sub kills, but they were a direct factor in almost 30 percent of the U-boats dispatched by American forces in 1944, and 25 percent in 1945. Although British sub sinkings are more difficult to categorize, the Ultra-assisted percentage of sub kills rose from 20 percent in 1943 to 26 percent in 1944. *26

  It was payback time for the dark days of 1942 and 1943. In May 1944, the Allies sank more than half the operating U-boats—destroying them at a rate faster than the Germans could replace them. Over the next three months that summer, the percentage of operating U-boats sent to the bottom reached a high of 76 percent. More important, the loss of Allied merchant shipping to U-boats remained low throughout the year—with almost no casualties in the Atlantic—in large part because of the more secure British Cypher No. 5.

  According to a jubilant and perhaps overstated July 14, 1944, OP20G memo, written when the American attack against Shark was at its high point,

  the effect of the U.S. “bombes” on solving the Atlantic U-boat cipher has exceeded all expectations. . . . For the last half of each day, we can read messages to and from Atlantic and Indian Ocean U-boats simultaneously with the enemy. In fact, during these hours the translation of every message sent by a U-boat is at hand about 20 minutes after it was originally transmitted. At present, approximately 15 percent of these keys are solved by the British and the remainder by OP20G.

  The American contribution to the victory over Shark and other Enigma naval systems was hard-won but still far from secure. Wenger, Engstrom, Desch, and their crews battled a stream of near-constant upgrades to the Enigma and its use. While the Germans did not make frequent and significant changes to their cipher systems—their math-beguiled cryptologists were too smug about the theoretical challenges of the Enigma machine to force such radical changes—they did see the need to improve their systems on a regular basis. At times, too, the German military had enough nagging doubts about the security of the Enigma to make larger adjustments—such as introducing a double turnover ring between the fast rotor and the second as it did in 1944. But given the cost and logistics of modifying hundreds and even thousands of Enigma machines scattered around the globe, the Germans took the conservative course, in part because many in the military believed that the Allies’ superior radar was the prime culprit for their U-boat troubles.

  New German systems and, later, Japanese codes, along with Germany’s latest Enigma innovations, kept OP20G off balance and in a reactive mode until the end of the war. Although there were attempts to move far ahead of their adversaries, the bright young theoreticians in Engstrom’s M section and the engineers in charge of machine design did not have the time to turn their most advanced ideas into hardware. Desch and the other engineers had to continue to build the types of machines that could be produced fast enough to meet crises, with the risk, of course, that they soon would be outmoded.

  The Navy engineers did not have the time to design a more automatic and general-purpose machine—a luxury available during the war only to the Army’s codebreakers. With the British decoding a nearly unbroken stream of German Army and Air Force messages, the U.S. Army was able to step back and design its more adventurous Madame X and Superscritcher machines, albeit with unimpressive results. Even though Madame X used thousands of telephone relays to attack the three-wheel Enigma—more relays than the telephone exchange for a town of eighteen thousand people—it nonetheless proved very slow after its completion in 1943 and was used mostly as a research machine. Plans for a four-wheel version of the machine were scrapped because it would have needed another five hundred thousand relays, then in short supply. The Superscritcher, a more advanced design that used vacuum tubes and electronics rather than relays, wasn’t ready until after the war.

  Unlike the Army machines, Desch’s Bombe needed skilled humans to change the wheels, switches, and cable connections for each run. The setups took as long or, at times, longer, than it took a Bombe to find a solution. But Desch’s choice was wise given the alternative: a much more complex, unpredictable, and hard-to-manufacture device.

  The machine’s balance of speed, power, and reliability paid off. The American Bombes were far faster than the British three-wheel machines and faster and more reliable than GCCS’s new four-wheel Bombes. On four-wheel runs, the American devices could complete their task in one twentieth of the time needed by the older British models. When setup times for runs were included, however, the American advantage in speed d
ropped to about a threefold edge over the British, depending on the number of wheels to be changed and switches to be reset. The American Bombes also were faster than the British four-wheel version, by about 50 percent, and much less prone to break down. *27

  The power of the 120 or so American Bombes was equal to or greater than that of the 220 British Bombes at Bletchley Park. The British, who were charged with attacking many more German systems than were the Americans—including the three-wheel Army and Air Force Enigma traffic known as Bovril—asked for time on the American Bombes.

  Fortunately for the Allies, Desch had had the foresight to design the Bombe with the flexibility to attack both three-wheel and four-wheel Enigma problems—all it took was a flip of a switch on the machine. That simple feature became a major blessing. In part because the Desch Bombes were so successful against the four-wheel Shark traffic, they soon developed excess capacity that could be diverted to help the British with their three-wheel problems. During the last two years of the war, more than 60 percent of the U.S. Bombes’ running time was devoted to Bovril traffic. *28

  Bovril was still predominantly a British operation, however. GCCS sent over the intercepts, cribs, and even menus and lists of priorities for runs on the American machines, sometimes to the frustration of Engstrom and Wenger, who wanted to focus OP20G’s efforts in the hunt for U-boats. The German Army and Air Force problems were divided into many different systems and produced a deluge of messages for cryptanalysis. Hoping to solve as many keys as quickly as they could, the British kept the easier problems to themselves and delegated the stickier ones to Washington—a move that displeased the Americans.

  As the Americans began to help with other Enigma systems, they made further adjustments to their Bombes. Two cipher systems that linked the German and Japanese navies as well as their attachés and blockade-runners—Seahorse and Sunfish—required a Bombe that could handle a long string of menu letters. The Desch machine, whose sixteen banks of simulated Enigmas could handle up to sixteen menu letters, was inadequate for the job. Too embarrassed to have to ask for time on the larger British Bombes, Howard Engstrom requested in the summer of 1943 a new double-unit Bombe with thirty-two banks, dubbed Granddad. Driving twice as many Enigma banks at the speeds required for the fast wheels and also supplying those banks with enough electrical current to keep track of their movements was a major task for NCR’s engineers. Joe Desch and his staff soon found solutions: they substituted a larger motor and a belt drive for the geared shaft in the regular Bombes and silver brushes for copper to carry more current through the banks. *29

  ALTHOUGH THE BRITISH continued to work on Japanese systems at GCCS and smaller centers in Ceylon and Australia, the United States bore the brunt of codebreaking duties in the Pacific and was the senior partner in the alliance, just the reverse of the European Ultra arrangement. The Navy’s Pacific codebreakers had relied on traditional approaches and hundreds of IBM tabulators, including the specially designed NC (Navy Change) IBM machines. Then in November 1943, just as the Navy was about to launch its toughest series of attacks yet against the Japanese-held islands in the Pacific, codebreakers asked for help from OP20G’s M section and the team at Building 26.

  The Japanese naval systems presented challenges different from, and in some ways more demanding than, those of the Enigma. With few exceptions, they used codes, not ciphers, and many were additive systems, such as the main Japanese naval code, JN-25. *30

  At times, the Japanese problems were more frustrating and demanded more manpower than Enigma. For Japanese systems like JN-25, each message presented a separate challenge until Allied codebreakers could reconstruct the current book of additives and identify the keys being used. As Admiral Nimitz prepared for his late 1943 surge into the Pacific islands, the Navy had to ensure that Japanese radio intelligence kept flowing. The pressure was on OP20G.

  In October 1943, OP20G’s research section was asked to design and build machines that would be many times faster than the best IBM tabulators to help speed up the attack against the Japanese codes. An ex-MIT student, Lawrence Steinhardt, was put in charge of the effort, and it soon became one of OP20G’s and Dayton’s highest priorities. Navy engineers conferred with the cryptanalysts working on the Japanese problems and began outlining a series of five revolutionary machines, called Copperheads. By November, Desch’s team started work on the first Copperhead model—the simplest of the five machines and the only one that seems to have been completed. All five designs were based on the principles of the early Vannevar Bush RAM devices: finely punched and ultrahigh-speed tapes, light-sensing circuits, and electronics.

  The first Copperhead was designed to search through the thousands of Japanese messages for “double-hits” of the same enciphered code groups, which indicated that two messages might have been enciphered with the same additives. But the project soon ran into technical troubles: its expensive and complex tape punches proved too slow and unreliable. At best, they could punch only forty messages of average length per hour when hundreds of such messages were needed for a single tape. It took close to one year to work out all the bugs in Copperhead, and it wasn’t in operation until November 1944. The other Copperhead plans were shelved.

  In addition to using difficult codes, the Japanese Navy showed signs that it might shift more of its most important messages to a machine encryption device. OP20G became aware that the new Jade (JN-157) machine was being used on at least some high-level naval messages in summer 1942, but it had been unable to put a team on the problem until late in the year. Concerned that the use of Jade might become widespread, one of OP20G’s brightest was given responsibility. Frank Raven, a 1934 Yale graduate in mathematics and a Phi Beta Kappa scholar, had both cryptanalytic and people skills.

  Raven suspected that the Jade device might be a more complex version of the Japanese Purple machine, which had been cracked already and was providing significant information to the Allies. Japan’s diplomats continued to use Purple throughout the war, unaware that the Allies were reading their messages. Purple was based on the use of telephone stepping switches, and Raven believed it was a good bet that Jade used them as well.

  Raven’s group untangled Jade after almost one year’s work, then quickly turned to Building 26 to produce a set of analog machines that would both speed decryption and help find settings. Raven also asked for more: a distinct and unique machine to solve Jade keys.

  By the end of October 1943, Desch’s men were busy constructing the analogs to Jade, called Vipers, while others in Building 26 were designing a new Bombe-like machine for attacking Jade, called Rattler. Delivered to Washington in May 1944, Rattler at last contained an innovation that OP20G had sought in early 1942 for the Enigma problem: electronic rotors. The Rattler and the Vipers were technical triumphs but disappointing in their use—the Japanese sent few messages over Jade, and the encryption device was retired less than three months after Rattler was delivered.

  But the experience with Rattler and the Vipers found other vital applications. Raven’s team suspected that Jade might be an extension of the latest version of a machine for the Japanese naval attachés’ communications, dubbed Coral, that had defied Navy codebreakers since its introduction in late 1939. Agnes Driscoll, who had been successful against earlier Japanese attaché machines, had labored against it for almost a year, without result, until late 1940, when she was ordered to concentrate on the German problems. Driscoll and her small group of experienced civilians were unable to return to the Coral challenge until the last months of 1943.

  In early 1944, one member of Raven’s group decided to see if he could help with the Coral and its code system, known as JNA-20. By then, Driscoll and her assistants were near giving up, after months of compiling mountains of statistics trying to understand the machine. They did not object to an outsider’s help. With encouragement and help from Bletchley Park, Raven’s team soon conquered Coral. By March 1944, OP20G was ordering analogs of the JNA-20 machines, dubbed Pythons in the tradi
tion of naming the Pacific codebreaking devices after deadly snakes. Building 26 produced many of those new Pythons and helped convert the Rattlers to run Coral problems.

  By April 1944, the work on JNA-20 was paying off. With Japanese attachés stationed in the major Axis capitals and many neutral nations, Coral provided valuable insights, including information on German technical capabilities and intentions. The messages of the very diligent Japanese attaché in Berlin provided many heads-ups on new German weaponry and technologies and helped the Allies determine whether countertechnologies were needed.

  Awed by the Nazis’ scientific wizardry, the attaché in Berlin regularly radioed his superiors in Tokyo the entire scope of the German efforts to counter the Allies’ antisubmarine strategies. He was especially impressed by the Germans’ new radar-search detectors, by their pattern-running torpedoes (Federapparat, or FAT), which could loop through a convoy with greater odds of hitting a ship, and by “Aphrodite,” a balloon launched by U-boats as a radar decoy. He described in detail the German Navy’s emergency training and tactics for counterattacking destroyers, for evading attack from surface ships, and for breaking through convoy-escort screens. Most important, though, he was the first to tip the Allies to the newer and deadlier classes of U-boats envisioned by Dönitz.

  The list included the development of the dreaded Walter submarine, which could stay submerged for days at a time, attack at underwater sprint speeds of eighteen knots, and beat a hasty retreat. The submarine was to be powered by a special engine, designed by German scientist Helmut Walter, that used hydrogen peroxide as a source of oxygen for combustion, eliminating the need for a snorkel or to surface. The Allies feared the Walter U-boats might be developed in time to blockade British ports and to launch a mass attack against the D Day invasion fleet. But by that time, the Germans were short on facilities for making either submarines or hydrogen peroxide, thanks to the merciless Allied bombing.