With a convulsive rumble, followed by billowing flames and exhaust, a sleek 60-foot rocket emerged from its silo at California's Vandenberg Air Force Base.
It was a test of the backbone of the nation's missile defense system. If North Korea or Iran ever launched nuclear weapons against the United States, the interceptors at Vandenberg and remote Ft. Greely, Alaska, would be called on to destroy the incoming warheads.
Scientists conducting the test at Vandenberg on Sunday, Jan. 31, 2010, had left little to chance. They knew exactly when the target missile would be launched from an atoll in the Marshall Islands 4,900 miles away. They knew its precise dimensions, expected trajectory and speed.
Based on this and other data, they had estimated the route the interceptor's heat-seeking "kill vehicle" would have to follow to destroy the target.
Within minutes, the interceptor's three boosters had burned out and fallen away, and the kill vehicle was hurtling through space at 4 miles per second. It was supposed to crash into the mock enemy warhead and obliterate it.
At a cost of about $200 million, the mission had failed.
Eleven months later, when the U.S. Missile Defense Agency staged a repeat of the test, it failed, too.
The next attempted intercept, launched from Vandenberg on July 5, 2013, also ended in failure.
The Ground-based Midcourse Defense system, or GMD, was supposed to protect Americans against a chilling new threat from "rogue states" such as North Korea and Iran. But a decade after it was declared operational, and after $40 billion in spending, the missile shield cannot be relied on, even in carefully scripted tests that are much less challenging than an actual attack would be, a Los Angeles Times investigation has found.
The Missile Defense Agency has conducted 16 tests of the system's ability to intercept a mock enemy warhead. It has failed in eight of them, government records show.
Despite years of tinkering and vows to fix technical shortcomings, the system's performance has gotten worse, not better, since testing began in 1999. Of the eight tests held since GMD became operational in 2004, five have been failures. The last successful intercept was on Dec. 5, 2008. Another test is planned at Vandenberg, on the Santa Barbara County coast, later this month.
The GMD system was rushed into the field after President George W. Bush, in 2002, ordered a crash effort to deploy "an initial set of missile defense capabilities." The hurried deployment has compromised its effectiveness in myriad ways.
"The system is not reliable," said a recently retired senior military official who served under Presidents Obama and Bush. "We took a system that was still in development — it was a prototype — and it was declared to be 'operational' for political reasons.
"At that point, you couldn't argue anymore that you still needed to develop and change things. You just needed to build them."
Dean A. Wilkening, a physicist at Lawrence Livermore National Laboratory in Livermore, Calif., offered a similar assessment. Wilkening served on a National Academy of Sciences panel that issued a 2011 report on missile defense.
GMD remains a "prototype system" that "has performed less well than people had hoped," he said at a May 28 policy conference in Washington, D.C. "If you're going to rely on that as an operational system, one shouldn't be too surprised that it does tend to fail more than you'd like."
At a separate conference this month, Wilkening called the system's test record "abysmal."
The Times interviewed missile defense scientists and current and former Defense Department officials, and reviewed thousands of pages of congressional testimony and reports by the Government Accountability Office, the Pentagon's independent testing office, the National Academy of Sciences and the Defense Science Board.
Official pronouncements about the GMD system, The Times found, have overstated its reliability.
Early on, the Pentagon's chief weapons buyer told Congress he had high confidence an attack could be foiled by firing one to three missiles at each enemy warhead.
Under that scenario, "the effectiveness would be in the 90% range," Defense Undersecretary Edward C. "Pete" Aldridge Jr. told the House Armed Services Committee in 2003.
Navy Adm. Timothy J. Keating, then head of the U.S. Northern Command, was even more emphatic when he told the Senate Armed Services Committee in 2007: "I appear before you today as confident as I know how to be in the employability and efficacy of that system."
But given GMD's record in flight tests, four or five interceptors probably would have to be launched to take out a single enemy warhead, according to current and former government officials familiar with the Missile Defense Agency's projections.
The system's 30 interceptors — four at Vandenberg and 26 at Ft. Greely — could be overwhelmed by an attack with multiple missiles.
The threat would be even greater if enemy missiles were outfitted with decoys or shed metal debris, which could confuse GMD's radar and sensors.
Despite GMD's problems, influential members of Congress have protected its funding and are pushing to add silos and interceptors in the Eastern U.S. at a potential cost of billions of dollars.
Boeing Co. manages the system for the Pentagon. Raytheon Co. manufactures the kill vehicles. Thousands of jobs in five states, mostly in Alabama and Arizona, depend directly or indirectly on the program.
The Obama administration, after signaling that it would keep the number of interceptors at the current 30, now supports expanding the system. Defense Secretary Chuck Hagel has called for deploying 14 new interceptors at Ft. Greely by late 2017.
Missile Defense Agency officials declined to be interviewed for this article. A spokesman, Richard Lehner, said in a statement that the agency was working "to conduct component testing and refurbishment of the interceptors currently deployed to ... improve their reliability."
The agency's director, Vice Adm. James D. Syring, told a Senate subcommittee Wednesday that officials had identified the causes of the two most recent flight-test failures, and that the underlying problems had been fixed, or would be by the end of this year.
Asked about the system's ability to defend against an attack with multiple missiles, Syring said his goal was to "greatly improve where we are today in terms of the number of interceptors that we fly at each threat."
Raytheon referred questions about GMD to Boeing, the prime contractor for the system.
A Boeing spokesman, Dexter Q. Henson, said the company "remains confident in the system's ability to defeat potential adversaries."
Missiles launched from North Korea or Iran probably would fly over the Arctic Circle on their way to the U.S., the most direct route. The GMD system is designed to destroy incoming warheads at roughly the midpoint of their arcing journey, as they begin their descent toward Earth — hence the term "midcourse."
Intercepting a ballistic missile is a supreme technical challenge. Scientists liken it to hitting one speeding bullet with another.
The GMD system's bullet is the 5-foot-long, 150-pound kill vehicle. During flight, it is subjected to extreme stresses: blazing heat and violent vibrations, followed by frigid temperatures outside Earth's atmosphere. Each kill vehicle has more than 1,000 components. The slightest glitch can foil an attempted intercept.
"Fly, then buy" is a maxim in the defense and aerospace fields, meaning that customers should wait until a complicated new system has been rigorously tested before purchasing.
With GMD, the government's approach was the opposite: "Buy, then fly."
Then-Secretary of Defense Donald H. Rumsfeld exempted the Missile Defense Agency from standard procurement rules and testing standards, freeing it to use research and development money to buy and deploy a system quickly.
The rocket interceptors were essentially prototypes rather than finished products when put in the field. The first model of kill vehicle was not flight-tested against a mock warhead until September 2006 — two years after the vehicles had been placed in the silos.
Because each of the kill vehicles is handmade, no two are identical. A fix that works with one interceptor might not solve problems with others. The piecemeal approach has left the system short of spare parts for critical components.
Pressure to produce and deploy the interceptors at a breakneck pace made it difficult to revise engineering drawings to correct shortcomings exposed in flight tests or keep up with technological advances.
One senior official involved in the system described his frustration at learning that some computers aboard the kill vehicles lacked the processing power of common cellphones.
About a third of the kill vehicles now in use — the exact number is classified — are the same model that failed in the 2010 tests, according to people familiar with the system who spoke on condition of anonymity. That model has yet to intercept a target.
Because interceptors used in test flights burn up when they reenter the atmosphere or are lost in the ocean, scientists have been hard-pressed to pinpoint the causes of the failures.
But some of the system's problems can be traced to the kill vehicles' internal guidance center — the electronic brain that dictates final speed and trajectory.
This crucial component, called the "inertial measurement unit," malfunctioned in preliminary factory testing and during seven subsequent flight tests, according to interviews with missile defense scientists and federal auditors and reports by the GAO and the Pentagon's testing and evaluation office.
Scientists suspect that intense vibration during the interceptors' ascent is the cause of some of the test failures. A GAO report in April described vibration as a "systemic problem."
It could take years of additional engineering work to solve this and other technical problems in the kill vehicles, scientists said.
Lehner, the Missile Defense Agency spokesman, said vibrations were successfully dampened in a January 2013 flight test. The test did not involve an attempt to intercept a target.
Philip E. Coyle III, who oversaw several early test flights as the Pentagon's director of operational testing and evaluation from 1994 to 2001, said that even the system's eight successful interceptions should be viewed skeptically because of the staged conditions.
"The tests are scripted for success," said Coyle, who has also served as a science advisor in the Obama White House. "What's amazing to me is that they still fail."
Engineers who have worked with the system acknowledge that because each kill vehicle is unique, even a successful test might not predict the performance of interceptors launched in combat.
For decades, America's defense against the nuclear threat was based on the Cold War doctrine of "mutually assured destruction," which held that the prospect of massive retaliation would deter both the U.S. and the Soviet Union from striking first.
The 1972 Anti-Ballistic Missile Treaty entrenched this doctrine by prohibiting either side from fielding systems to intercept intercontinental missiles. The idea was that the prospect of mutual annihilation had proved effective and missile defense would undermine it; if either side thought it could block a retaliatory response, the temptation to attack would be greater.
After taking office in 2001, Bush withdrew from the treaty, which he said was outmoded and prevented the U.S. from protecting itself against new threats from the world's "least responsible states," notably North Korea.
Bush was not the first president to see promise in missile defense technology. In 1983, President Reagan launched the Strategic Defense Initiative, which envisioned ground- and space-based systems that would render nuclear weapons "impotent and obsolete." An estimated $30billion was spent on the effort, but no system was ever deployed.
The nation's defense against a massive attack by Russia or China continues to be based on mutually assured destruction. The GMD system was conceived as a safeguard against a "limited nuclear attack" by a less imposing adversary.
In 1995, Congress attached to the Defense Department budget a provision requiring deployment of such a system by 2003.
President Clinton vetoed the bill, saying it "would waste billions of dollars" on a negligible threat. But the administration said it would test elements of a missile defense system for three years before deciding whether the technology was reliable enough to deploy.
Missile defense proponents in Congress responded by appointing an advisory commission chaired by Rumsfeld. In a 1998 report, the commission said U.S. intelligence agencies had failed to recognize the threat posed by North Korea, which it said had the "capability to deploy chemical or biological warheads on missiles."
The panel suggested that Iran, Iraq and North Korea were developing nuclear weapons. Perhaps most ominous, it said a country committed to developing long-range missiles would need only about five years to do so.
In July 1999, Clinton signed the National Missile Defense Act, which called for a system to defend against limited ballistic missile attack. The law said an "effective" system should be deployed "as soon as is technologically possible."
Two of the first three flight tests of rocket interceptors, conducted in 1999 and 2000, failed. The Clinton administration supported continued research but declined to deploy a system.
That policy changed when Bush took office in 2001 and appointed Rumsfeld secretary of Defense. On Dec. 16, 2002, Bush signed a presidential directive requiring "deployment of a set of missile defense capabilities in 2004."
The week before, two flight tests of the fledgling system had failed. At that point, its performance record stood at five successful intercepts in nine attempts. Bush's directive called the results "impressive."
Instead of delaying deployment until the system had been rigorously tested, the Missile Defense Agency, beginning in 2004, placed interceptors in silos at Vandenberg and Ft. Greely and declared the system operational.
Rumsfeld and his aides said it made more sense to improve the system over time rather than to try to field a finished version at the start.
"The way this program is evolving, every one of the new missiles will be better than the one that previously was produced," Michael W. Wynne, acting undersecretary for acquisition, technology and logistics, told the Senate Armed Services Committee on March 11, 2004. "Each one of the missiles will add to the reliability of the system."
Appearing before a Senate subcommittee a month later, the Missile Defense Agency's director, Air Force Lt. Gen. Ronald T. Kadish, acknowledged some "technical challenges" but said: "Things are going all in the right direction."
Signs of trouble soon emerged.
Launches planned for Dec. 15, 2004, and Feb. 14, 2005, were scrubbed when the interceptors remained stuck in their silos. Officials attributed the first failure to a "problematic software configuration" and the second to a silo support arm that "did not retract, triggering an automatic abort."
By then, a dozen interceptors had been placed in silos. The system's proponents maintained that this skeletal shield provided unprecedented protection.
"If we didn't have this … we would be absolutely naked, given an unexpected attack," said Sen. John Cornyn (R-Texas) at an April 7, 2005, Armed Services Committee hearing.
A new Missile Defense Agency director, Lt. Gen. Henry A. Obering III, told the senators that day: "We maintain our confidence in the system's basic design, its hit-to-kill effectiveness and its inherent operational capability."
Obering conceded, however, that 38% of the software and 33% of other components in the kill vehicles had not been validated through flight testing.
"I am confident that the kill vehicle will work," he said under questioning. "But we have yet to prove that."
Project engineers were concerned that some components of the kill vehicles, known as CE-1 models, were already outdated. Spare parts, including replacements for the vehicles' guidance system, were out of stock. This prompted development of a partly redesigned kill vehicle, designated CE-2.
"The CE-2 was going to be the salvation of the program," recalled a Defense Department engineer.
Yet neither the original kill vehicle nor the CE-2 featured modular designs that would enable technicians to swiftly remove and replace suspect components. It typically takes a year or more to disassemble and restore a kill vehicle.
Defense officials bought the CE-2 models from 2008 to 2010 knowing that preliminary factory testing had found problems with the inertial measurement units, according to federal auditors. Though the design of those units was tweaked and they were ultimately cleared as acceptable at the factory, malfunctions arose during flight testing.
Some of those working on the GMD system believed it was time to stop buying more interceptors and embark on a comprehensive redesign.
"But the pressures did not allow that to happen," said a former high-ranking Pentagon official, referring to lobbying by contractors and demands from members of Congress. "There was a drive — we had to keep going to more and more" interceptors.
Boeing and Raytheon are among the top four defense contractors worldwide in revenue. From 1999 through March of this year, Boeing spent $261.6 million on general lobbying of the federal government and Raytheon spent $144.4 million, public records show.
Raytheon referred questions to Boeing, which offered no comment on its lobbying.
One of the staunchest advocates for speedily expanding the system has been Sen. Jeff Sessions, a Republican from Alabama, where missile-defense jobs are heavily concentrated.
Sessions, the senior Republican on the Senate subcommittee responsible for missile defense, has fought moves to slow the production of the rockets and has warned repeatedly about what North Korea or Iran might do.
Alabama's other senator, Richard C. Shelby — the ranking Republican on the Appropriations Committee — has also sought to deflect concerns about the test
failures and the program's cost.
"We're interested in cost," Shelby said at an appropriations subcommittee hearing on July 17, 2013. "We're also interested in defending this country."
Though both North Korea and Iran have launched crude unarmed missiles, U.S. intelligence assessments provided to Congress indicate that neither country has the capability to deliver a long-range, nuclear-tipped missile to the United States.
In late 2008, Bush appointed an experienced physicist to lead the Missile Defense Agency.
Army Lt. Gen. Patrick J. O'Reilly discovered that the agency was spending more to prepare the harsh terrain at Ft. Greely for a planned third field of interceptors than to improve the problem-plagued kill vehicles.
O'Reilly persuaded Defense Department officials to support postponing the new field so the money could be used for design and engineering fixes to the interceptors, according to people familiar with the matter.
Defense Secretary Robert M. Gates agreed to delay the work.
A group of senators, including Sessions, privately complained to Gates about the decision. Others in Congress, including Rep. Michael R. Turner (R-Ohio), suggested that O'Reilly was looking past the North Korean threat.
At a House Armed Services subcommittee hearing on May 21, 2009, Turner told O'Reilly that not deploying more interceptors would undermine "our industrial base," adding: "There's concerns in second- and third-tier suppliers who, reports indicate, could be without work at the end of this year."
O'Reilly replied that his priority was to "ensure that we are addressing the obsolescence" of the electronics in the interceptors.
He soon learned that Gates was planning to visit Ft. Greely. O'Reilly voiced interest in joining Gates but said he was blocked by the Defense secretary's aides and never got the chance to explain personally his rationale for postponing the expansion.
Gates later ordered the third field of silos built. In his book "Duty: Memoirs of a Secretary at War," Gates said he reversed his earlier decision after "seeing how close they were to completion." Gates did not respond to messages seeking comment for this article.
As O'Reilly and his staff focused on the next test flight, new concerns arose about the CE-2 kill vehicle.
In March 2009, the GAO reported that the test was being delayed, partly because of problems with the CE-2's guidance center, the inertial measurement unit. The office also noted that 10 of the kill vehicles "will have been manufactured and delivered before that first flight test demonstrates the CE-2 capability."
Ten months later, the CE-2 was at last scheduled for its first flight test.
Preparations were elaborate. Commercial air and shipping traffic was warned to stay out of a wide swath of the Pacific on the day of the test.
At 3:46 p.m. Sunday, Jan. 31, 2010 — six minutes after a target missile was launched from Kwajalein Atoll in the Marshall Islands — the rocket rose from its silo at Vandenberg. The attempted intercept failed.
After a months-long review, the Missile Defense Agency attributed the failure in part to "kill vehicle and system sensor" malfunctions.
The review also faulted a contractor for failing to install a lock wire — described by engineers as a "twist-tie" device — when assembling the kill vehicle.
Federal auditors reported that the sea-based radar used to track the target missile had become confused by pieces of metal that sloughed off the projectile, underscoring the system's potential vulnerability to decoys.
Sessions continued to press the Pentagon to buy and deploy more interceptors.
"This administration, to me, is penny-wise, pound-foolish," Sessions told O'Reilly at a Senate Armed Services Committee hearing on April 20, 2010. "I'm concerned about whether or not we're building enough, we're deploying enough, of these missiles."
On Dec. 15, 2010, engineers once again launched a CE-2 kill vehicle from Vandenberg toward a target missile.
This time the sea-based radar "performed as planned," the missile agency said, but the kill vehicle missed the target.
O'Reilly told a Senate subcommittee on April 13, 2011, that he wanted to halt the planned purchase of seven interceptors and use the savings "to do more refurbishments" and conduct additional testing.
"That looks to me like you're robbing Peter to pay Paul," Sessions replied, adding: "We don't need to stop short of the number of interceptors we need in the ground and prepared."
Asked for comment, Sessions' office said the senator's advocacy for the GMD system was motivated by concern for national security.
Shelby said in a statement that the work done on GMD in Alabama helped "protect the U.S. and our allies from rogue and unpredictable countries.... Given the increasingly dangerous world we live in, I will continue to advocate for missile defense to deter and defend against our enemies."
In his first round of congressional appearances as O'Reilly's successor, Syring said he had "great confidence we have addressed the causes" of the failed December 2010 test.
Appearing before a House Armed Services subcommittee on May 8, 2013, Syring said the next flight test would "demonstrate the improvements made" to the fleet of interceptors. For this test, the missile agency would use one of the original CE-1 kill vehicles.
The test was held July 5, 2013 — 31 months after the last attempted intercept, which failed.
After burning its boosters to reach space, the interceptor failed to separate from the rocket, preventing it from striking the target.
Syring is asking Congress for $99.5 million to begin what he described Wednesday as a "redesign improvement" of the kill vehicle. The work would stop short of a complete redesign, according to people familiar with the matter.
Frank Kendall III, undersecretary of Defense for acquisition, technology and logistics, broke ranks recently with those who have given upbeat assessments of GMD.
"We recognize the problems we have had with all the currently fielded interceptors," Kendall told a defense industry conference in Washington in February. "The root cause was a desire to field these things very quickly and very cheaply.... We are seeing a lot of bad engineering, frankly, and it was because there was a rush."