Depth charges exploding after being dropped by the destroyer HMS VANOC over the spot indicated by the submarine detecting apparatus, which reported a contact during an Atlantic Convoy, May 1943. Some crew members can be seen at the stern watching the explosion.

A Type IXD2 under attack from US aircraft. The two flakvierling appears to be pointed in different directions, indicating it is under attack by more than one aircraft.

Slow Convoy, SC 127 eluded U-boats at one of the most difficult times in the Battle of the Atlantic. The ocean was so full of U-boats that the first sea lord feared that “We can no longer rely on evading the U-boat packs and, hence, we shall have to fight the convoys through them.” In addition, the B-Dienst was at the height of its powers, solving 5 to 10 percent of its intercepts in time for Grand-Admiral Karl Dönitz C-in-C U-boat command, to use them in tactical decisions. Early information sometimes enabled him to move his U-boats so that a convoy would encounter the middle of the pack, enabling more boats to attack than if the convoy met only one wing of the patrol line.

But the first signs of German weakness had begun to appear. Stronger Allied defenses—more escorts, more airplanes—kept the U-boats from attacking with the vigor and daring of the previous years. Dönitz’s exhortations grew shriller, complaining that anyone who failed to engage the enemy closely was “no true U-boat man.” The rate of success declined. The great convoy battle of March 1943, during which U-boats sank Allied ships at twice the rate at which they were being built, was followed in April by a fight that brought poorer results: the Germans sank twelve merchant vessels, but at a cost of seven U-boats. The situation worsened the following month.

“In the Atlantic in May,” wrote Dönitz in his war diary, “the sinking of 10,000 tons was paid for with the loss of one U-boat, while not very long before that time one boat was lost for the sinking of about 100,000 tons.” He called such losses “unbearable,” and on May 24 he pulled the seventeen submarines on the North Atlantic convoy routes out and sent them to what he thought was a “less air-endangered area” to the south. From there they could operate against the convoys between the United States and the Strait of Gibraltar, through which supplies for the American forces in North Africa had to pass. But this was not the vital traffic whose loss would defeat Britain and keep the Allies from mounting an assault against Festung Europa. The move marked a major defeat for the Germans in the vital Battle of the Atlantic.

The success of Allied convoy diversions in January and February 1943 had again raised Dönitz’s suspicions about the security of his ciphers. For two and a half weeks in January, U-boat sweeps had discovered no convoys along the North Atlantic routes to Britain; for the first time since the United States entered the war, merchant ship losses in all Atlantic areas fell below one a day. In February, the few convoys that were not sighted by chance were spotted only by single boats at the ends of patrol lines, suggesting that the convoys were going around the wolfpacks. Dönitz’s concern was intensified when Allied destroyers came upon the U-459 as it was refueling an Italian U-boat some 300 miles east of St. Paul’s Rock, the desolate traditional division between the North and the South Atlantic, far from any destroyer bases and far from the normal convoy lanes. And the B-Dienst’s solutions of Allied U-boat situation reports raised suspicions. On April 18, for example, an intercept of an Allied submarine situation report showed that the Americans suspected the presence of twenty submarines in the rectangle running from 48° to 54° north latitude and from 38° to 45° west longitude. And the report was correct: TITMOUSE was in the area with eighteen boats.

Dönitz asked Vice-Admiral Erhard Maertens, Chief of Office of Naval Intelligence, Naval War Command, to investigate, as he had done in 1941. Again Maertens exculpated Enigma. The British U-boat situation reports themselves stated that the Allies’ information on submarine locations was coming from direction-finding, he said. Documents found in a French Resistance agent’s radio station showed that the Allies were obtaining information from the Resistance on departure times for U-boats and on whether they were headed for the North or the South Atlantic, enabling the foe, Maertens said, to estimate submarine movements with some accuracy. The British information about the wolfpacks DOLPHIN and FALCON was vague; if the information had come from cryptanalysis, it would have been exact. At worst, capture, perhaps of a cue word, which—contrary to all regulations—would have to have been written down, might have given the Allies insight into some messages. The chief of the Naval War Staff conceded that a capture was possible, and he approved Maertens’s plan to establish separate regional key nets.

Maertens was supported in his position by the coincidental discovery on February 2, in a British bomber downed at Rotterdam, of a new type of radar. It was based on the cavity magnetron, a block of copper with eight cylindrical holes bored in it parallel to and around a central axis. These hollows enabled the radar to operate on a wavelength of 9.7 centimeters, much shorter than the earlier 1.5 meters. Because its wavelength was measured in centimeters, the device was called “centimetric radar.” It gave the British two advantages: it depicted objects—coastlines, buildings—on the radar screen, which the older radar could not do, and the U-boats’ radar warning receivers, which were tuned to the longer wavelength, could not detect it. With centimetric radar, British airplanes could thus locate surfaced U-boats from a distance without alerting the submarines and could attack them by surprise. The Royal Air Force Coastal Command had begun doing just this with some success against U-boats traversing the Bay of Biscay. Though Dönitz had as yet no evidence that centimetric radar was being used in the Battle of the Atlantic, the use of this powerful new weapon could not be excluded.

So Dönitz accepted Maertens’s view that Kriegsmarine ciphers were secure and that the leaks were elsewhere. “With the exception of two or three doubtful cases,” he confided to his war diary, “enemy information about the position of our U-boats appears to have been obtained mainly from extensive use of airborne radar, and the resultant plotting of these positions has enabled him [the enemy] to organize effective diversion of convoy traffic.” And when SC 127 circumvented a wolfpack, he gave as the most probable reason that “the enemy has an extraordinary location device, usable from airplanes, whose effect cannot be observed by our boats.”

Nevertheless, suspicion that the Allies were solving naval Enigma messages would not die. Dönitz tried to reconcile his concern with Maertens’s reassurances, but he was not always able to. On April 27, as SC 127 was slogging across the ocean, the Allies, in a U-boat situation report that the B-Dienst solved, reported five U-boats within a 150-mile radius of 50° north, 34° west. “For some time resupplying has been carried out here,” Dönitz noted. “It remains disquieting that they were suspected precisely in the area in which no radioing had been done for several days.”

A few days later, Dönitz, for reasons that went beyond his fears about cryptosecurity, fired Maertens, sending him to Kiel to run a shipyard. He replaced him with the glass-eyed Captain Ludwig Stummel, Maertens’s chief of staff, promoting him to rear admiral. Stummel maintained, as always, that Enigma “had, on the basis of repeated and thorough investigations, proved itself up to the present as unbreakable and militarily resistant.” Dönitz apparently believed him, for in June he was telling the Japanese ambassador that U-boat losses were due to a new Allied direction-finding system.

Despite his claims, Stummel began in 1944 to prepare a measure that would carry the Kriegsmarine’s basic cryptosecurity principle to its logical conclusion. By subdividing the navy’s cryptosystem into as many key nets as necessary, Stummel sought to reduce the number of messages in a common key. As the volume of traffic grew, Enigma key nets had expanded from one in the early 1930s to separate home and foreign key nets and to the addition of a U-boat net and many others by 1943, when traffic averaged 2,563 radio messages a day. Now Stummel proposed to give each U-boat its own key.

Individual keys were issued to some submarines shortly after D-Day, June 6, 1944; they began to be widely used in November, and by February 1945 they were carrying practically all the operational traffic of the U-Boat Command. In that month, Dönitz told Hitler that Allied knowledge of wolfpacks came from radar and betrayal. By then Stummel had also been ousted, but his program of individual keys justified his faith in Enigma: G.C.&C.S. solved only three keys for brief periods. Perhaps not coincidentally, sinkings rose steadily from November 1944 to April 1945 in the North Atlantic and North Sea, although the absolute number remained small. Solution of these individual keys would have required a great increase in personnel and in bombes, but G.C.&C.S. felt confident that it would have been able to do it. Germany’s surrender saved it from this test.

Long before that happened, Dönitz mourned the loss of the source of information that he said gave him half of his intelligence: the B-Dienst. He had feasted on it for so long in part because the Germans had no monopoly on cryptographic failure. In this respect the British were just as illogical as the Germans. The surprise of the North African invasion confirmed the Admiralty’s belief that its cryptosystems were secure, just as Fricke had argued that the operations of British ships gave no indication that the British were reading German messages. And G.C.&C.S. retained confidence in its superencipherment (even though it had solved similar systems before the war) because it was encountering increasing difficulty in solving high-grade Italian codes after the summer of 1940 and fewer problems with nonnaval Enigma; this logic resembles the Kriegsmarine’s argument that Enigma must be secure because it was unable to break the American naval cipher machine.

The cherished beliefs of the British were wrong. In December 1942, they learned from their Enigma solutions that the Germans were reading Naval Cypher No. 3, the main cryptosystem for convoy arrangements in the North Atlantic. And in Washington, in March 1943, Lieutenant McMahan of OP-20-G saw a German intercept that canceled an order by Dönitz of a few hours earlier and directed a radical change of course. McMahan thought that only a German solution of a message diverting an Allied convoy could have caused Dönitz to react like that. He went downtown to Convoy and Routing in Main Navy and, after some difficulty, persuaded them to let him see the messages to Allied convoys. His discovery of the very message that had ordered the detour brought together compartmentalized elements and confirmed the Allies’ recognition that the Germans were reading their traffic.

In June, when Naval Cypher No. 5 replaced Nos. 3 and 4, the B-Dienst made no real progress against it. Concerns about the security in heavy traffic of the superencipherment, called the long sub-tractor system, had been raised as early as 1940; G.C.&C.S. devised a replacement—the stencil subtractor—by 1941, but the services did not decide to adopt it until after extensive trials that ended in March of 1942. Design and production of the devices and printing of the tables took the rest of the year, distribution for the Royal Navy until the middle of 1943, and distribution within the U.S. Navy until January 1, 1944—a record of cryptographic negligence that compares favorably with Germany’s. Still, from the middle of June 1943, the B-Dienst was effectively shut out from its vital Anglo-American intelligence. In May 1944, Hitler asked his naval codebreakers which English systems could be broken. They had to confess that although they were solving a number of secondary systems and a convoy system for stragglers, “The two main English systems cannot be read, the one [the main warship cryptosystem] since the start of 1944 and the other [the convoy system] since the start of June 1943.”


This admission unwittingly confirmed the Allied victory in cryptology. In August 1943, the British and the Americans had begun reading Enigma messages nearly always currently. The capture of the U-505 by an American task force on June 4, 1944, provided a copy of the Adressbuch that provided the keys for disguising grid positions; from then on the Allies read them as easily as the Germans did.

But solving German messages did not always mean the successful diversion of convoys. It is true that in January and February 1943, when solutions were almost uninterrupted, the Allies suffered far fewer losses than in March, when for days no solutions were achieved. On the other hand, two convoys out of three escaped detection in August and September 1942, during the ULTRA blackout, while less than half avoided being spotted in the first five months of 1943, when solutions were frequent. The totality of other factors eclipsed ULTRA: the number of U-boats on patrol, the quantity of very long range aircraft the Allies had, centimetric radar, shipboard direction-finding, operational research, the arrival of escort aircraft carriers, the increase in escort vessels. But when ULTRA worked with these new Allied strengths, particularly after Dönitz withdrew his U-boats from the North Atlantic on May 24, the results could be spectacular. On September 21, 1943, Churchill announced to the Commons that, in the third of a year just ending, not one merchant ship had been lost to enemy action in the North Atlantic. The House erupted in cheers.