Missile Wars
homethe threatthe technologythe strategydiscussion
Why missile defense won't work
by Theodore A. Postol

In this April 2002 article from Technology Review, MIT professor Theodore Postol analyzes U.S. missile defense tests and reveals the basic flaws that went unreported -- and how a primitive adversary can defeat the system with the simplest of technologies.


Theodore A. Postol is professor of science, technology, and national security policy at MIT.

On June 23, 1997, a prototype of a U.S. military "kill vehicle" designed to intercept nuclear missiles lifted off from a launch pad on the South Pacific atoll of Kwajalein. Its purpose was not to seek out and destroy. Instead, it was to fly by and observe a group of objects that had been launched into space more than 20 minutes earlier from Vandenberg Air Force Base near Santa Barbara, GA, almost 8,000 kilometers away-and determine whether it was possible to distinguish a cloud of decoys from the mock warhead they protected.

It was a big day for nuclear missile defense. Since the decoys used in this experiment were of very simple design, if the experiment showed that the warhead could not be reliably identified, it could mean the whole Star Wars defense plan would for all practical purposes be unworkable, since the most primitive of adversaries could defeat it with the simplest of decoys. Of even greater importance, it would also be a clear demonstration of the fundamental physical reasons why any missile defense that relied on kill vehicles of this type could never be successful.

It worked -- at least that's what we were told. But shortly after the experiment flew, three courageous people -- a former employee of defense contractor TRW turned whistle-blower, a TRW retiree and a U.S. Department of Defense investigator -- brought new evidence to light. Their information, coupled with my own investigation and repeated calls for a full accounting from U.S. representatives Howard Berman and Edward Markey, pointed to a different story -- one of failure, a finding seemingly confirmed this February by a draft of a Government Accounting Office follow on study, as reported by the journal Science. I believe that the top management of the Pentagon's Missile Defense Agency (previously known as the Ballistic Missile Defense Organization) and its contractors have misrepresented or distorted the results derived from the experiment and rigged the follow-on test program that continues to this day. These deliberate actions have hidden the system's critical vulnerabilities from the White House, Congress and the American citizens whom the missile defense program was supposed to protect.

How the Defense System is Supposed to Work

As envisioned since 1996, the U.S. National Missile Defense effort consists of three main elements: infrared early warning satellites, ground-based radars to precisely track warheads and decoys from thousands of kilometers away, and multistage, rocket-powered homing interceptor missiles launched from underground silos. The most critical element of this defense is the roughly 1.5-meter-long "exoatmospheric kill vehicle" that the homing interceptor deploys after being launched to high speed by its rocket stages. After deployment, the kill vehicle has about a minute to identify the warheads in a cloud of decoys as it closes on the targets at high speeds. To that end, it carries its own infrared telescope and has small rocket motors that enable it to home in on its prey. The kill vehicle does not carry a warhead. Rather, it is designed to destroy its quarry by force of impact.

When an enemy missile is launched, it typically takes 30 to 60 seconds to reach altitudes where the infrared early-warning satellites can detect the hot exhaust from its engines. These satellites orbit at an altitude of 40,000 kilometers and can be kept over the same point on the earth's surface. Once two or more detect the rocket, they can crudely track it in three dimensions by stereo-viewing. However, the satellites can only see the hot exhaust from the rocket's engines, so their tracking ends abruptly when the engines shut down -- an event that typically happens in space at between 200 and 300 kilometers in altitude.

Roughly three minutes after engine shutdown, the rocket's upper stage and the just released warhead and decoys rise above the horizon, where they can be tracked by radar. The radar systems originally planned for this task operate on a very short wavelength (three centimeters at a frequency of 10 gigahertz), which allows them to identify objects to an accuracy of 10 to 15 centimeters from many thousands of kilometers away. This makes it possible to observe distinct reflections from different surfaces -- even the seams on an object as it tumbles through space. The spacing and intensity of these signals, and the way their echoes vary as the orientation of a target object changes, can in some circumstances be used to determine which object is a warhead and which a decoy. If all goes well, this information will be used to deploy one or more interceptors within about 10 minutes of an attack's being launched. The interceptors will fly to the defense, destroying their targets about 18 minutes after launch.

That, at any rate, is how the system was initially supposed to work. President Bush's latest proposal does not include this high-resolution radar, making tracking and identification of enemy missiles harder and delaying the interception time. But even with the more advanced original system, big problems surround the scenario. For starters, an adversary could alter the reflections from decoys and warheads by covering surfaces and seams with wires, metal foil or radar-absorbing materials. These simple strategies would render the radar unable to reliably sort out warheads from their armadas of decoys.

Compounding this problem is a simple fact: in the near vacuum of space, a feather and a rock move at the same speed, since there is no air drag to cause the lighter object to slow up relative to its heavier companion. This basic vulnerability makes it even easier for an adversary to devise decoys that will look like warheads to radar or an infrared telescope observing them from long range.

What's more, an adversary would likely deploy decoys and warheads close together and in multiple clusters. Under these conditions, even if the radar could initially identify a warhead among all the decoys, it couldn't track it accurately enough to predict the relative locations of the different objects when the kill vehicle encountered them some eight minutes later. Consequently, the kill vehicle must be able to identify warheads and decoys without help from satellites, ground radars or other sensors. If it cannot perform this task, the defense cannot work. This is where the infrared telescope comes in -- and it was really this critical part of the system that the June 1997 test was all about.

How the Kill Vehicle Identifies Warheads

During a typical intercept attempt, the closing speed between the kill vehicle and targets is around 10 kilometers per second. If targets can be detected from a distance of 600 kilometers, that doesn't leave much time -- a minute or less -- to distinguish between warheads and decoys and maneuver to ram into the right target. The resolving power of the kill vehicle's telescope is quite limited, so all objects look like points of light. Still, the distinction can be made -- by measuring the brightness of each object, and to some extent its wavelength or "color," which in turn can give clues to its infrared temperature.

If, for instance, one object is a tumbling, featureless sphere, no orientation will look different from any other, and its signal will be steady. However, if another object is of a different shape, the different faces it presents to the kill vehicle will show varying degrees of brightness as it tumbles end over end through space; a rod, for example, will be brighter when its more luminous side area is exposed to the telescope than when viewed end-on and will appear to the kill vehicle as a distant point of light that increases and decreases in brightness twice during each complete rotation. So if there is prior knowledge that one target is a tumbling rod and the other is a featureless sphere, it will be clear which is which.

That's the theory. The truth is more complicated. For one thing, measuring temperature with this infrared equipment is not possible when objects in space are observed close to the earth, because their signals are routinely contaminated by reflected infrared radiation from the planet's surface; they are further confused by such factors as the amount of cloud cover, time of year and which part of the earth the target is over.

Even without such earthly interference, the limited strategies available to the defense for distinguishing warhead from decoys put it at a disadvantage. For example, one simple way for an adversary to make discrimination impossible is to put the warhead inside a balloon and deploy it with many additional balloons of different sizes and surface coatings. The temperature of a balloon exposed to the sun can be drastically altered, as can the amount of infrared heat it radiates and reflects from the earth and sun, depending on its size and surface coating. Balloons of different dimensions and with different coatings would each look slightly different. Since there would be no way to know why this was so, there would be no way to know which balloons were empty and which contained warheads -- and discrimination by the kill vehicle's infrared telescope would be impossible.

This is the central point that backers of missile defense have not been able to circumvent.

So far, there have been seven tests, the most recent last December. In each case, a payload of targets has been launched by a modified Minuteman II intercontinental ballistic missile (ICBM) from Vandenberg Air Force Base toward Kwajalein. The target-carrying Minuteman completes powered flight in about three minutes and deploys a rocket-powered vehicle called the Multi-Service Launch System. This vehicle takes another four and a half minutes to deploy its payload, but only after rotating nearly 90 degrees so it can release the targets along a single downward direction in space. Since the kill vehicle telescope has a field of view roughly equal to that of a person looking through a soda straw (about one to two degrees), the payload deployment along a single direction assures that the targets will all be in its limited sights when they arrive at Kwajalein some 20 minutes later. Since simultaneous observation of targets is critical to quickly distinguishing decoys from warheads, this specialized deployment geome try gives the kill vehicle a significant advantage -- one it is hardly likely to have in a real-world attack. (If, instead, the targets were deployed in many directions, the kill vehicle would have to slew between many clusters of targets, viewing each for tens of seconds to get the same readings. Even if it could identify the right target, there would likely not be time to maneuver and intercept.)

When the first flight test was performed, 10 targets were to be observed by the kill vehicle. These included a roughly two-meter-long, spin-stabilized mock warhead; two cone-shaped rigid decoys that were of roughly the same shape and size as the mock warhead; four spherical balloons (two with a diameter roughly equal to that of the base diameter of the mock warhead, and two about half that size); a small cone-shaped balloon; a large spherical balloon; and the upper rocket stage that deployed the decoys and warhead.

At first glance, it might seem that this ragtag collection of decoys is just what an enemy would throw at us. But since the makeup of these objects and the space infrared environment in which they operated were fully known -- all the tests have been carried out around the same time in the early evening, assuring that the geometry of the sun and earth are essentially the same in every experiment -- it was possible, at least in principle, to predict how each would look to the kill vehicle. The predictions indicated, for instance, that the two medium balloons would have scintillating signals as bright as that of the spin-stabilized mock warhead, which had roughly the same diameter. Each of the rigid cone-shaped decoys was expected to look like a tumbling warhead. The large balloon and upper rocket stage were expected to look much brighter than all the other objects, while the small spherical balloons and the cone-shaped balloon would stand out for their dimness. Under these simplified conditions, and with detailed prior knowledge of the characteristics of each object, it must have seemed quite likely that the kill vehicle could pick out the "warhead" from among the decoy companions.

The results of the actual trial were quite unexpected, however, and must have been extremely disconcerting to then director of the Ballistic Missile Defense Office Lt. Gen. Lester L. Lyles and his engineering team. Lyles reported that the trial had proven that discrimination of warheads in a cloud of decoys was possible. However, we now know there was a serious basic problem with the first integrated flight test that would likely, even with the targets' expected characteristics known prior to launch, make any of the data gathered by the kill vehicle essentially useless.

To begin with, one of the medium-sized balloons failed to fully inflate, resulting in its looking half as bright as expected. The fluctuation characteristics of the mock warhead's signal, meanwhile, changed over time, making the probability of its being the warhead appear at different moments more than five times higher or lower than expected. Indeed, the fluctuation characteristics of all the objects were either substantially different from the predictions or changed in time so drastically that if they could be matched to the template of expected values at one time, they could not be matched to the same template even seconds later.

That was bad enough. But the real problem was that the kill vehicle's main infrared sensor failed to cool to its 12-degree-Kelvin design temperature, achieving instead a temperature no lower than 13.5 Kelvin. This difference is the same as if a space suit had been designed to keep an astronaut in a temperature environment of 20°C but instead put her in an environment of 66°C.

Since the sensor was very hot relative to its design operating temperature, the measured target signals were contaminated with heat-generated electronic noise. The unexpectedly high temperature also caused unpredictable changes in the efficiency with which each of the tens of thousands of tiny, independent infrared sensors in the kill vehicle converted infrared signals to electronic. These sensors are arrayed at the telescope's "eye-piece," so that an electronic image of the instrument's field of view can be formed, much the way images are formed by solid-state TV cameras.

Since each infrared sensor's performance was different in detail from the performance of the other sensors, and since the details of how the performance of each sensor changed with the unexpectedly high temperature were unknown, it was not possible to accurately measure the brightness of distant targets, or even the brightness of these targets relative to each other. Because knowing the brightness of a target is critical to identifying it, the singular fact that the main sensor was not at its operating temperature, and that the performance of each of its tens of thousands of elements was unknown, means that the kill vehicle's capacity to discriminate its target was severely compromised.

To understand this basic point, imagine that an object is being observed by two infrared sensors with different conversion efficiencies. When light from the object strikes one of the sensors, a certain brightness is recorded. When it strikes the other sensor, a different brightness is recorded. Unless the conversion efficiencies of both sensors are known, the actual brightness of the object cannot be determined.

This well-known problem and others associated with it would typically have been dealt with by calibrating the performance of each sensor over the range of expected operating temperatures prior to the experiment. However, the experimental team had not performed calibration measurements on the array at 13.5 Kelvin and higher because it had not expected such a massive failure in the sensor's cooling system. Additional sensor calibration data was supposed to have been obtained by observing an infrared star of known brightness, Alpha Bootes, but the noise in the many sensor elements, and changing sensor array temperature during the test, rendered this measurement useless.

So, I believe, when the carefully contrived test failed, the true results of the experiment were hidden through careful selection of the data used in the analysis -- and the way in which those data were interpreted. The kill vehicle collected about 63 seconds of data, starting at a range of roughly 460 kilometers and continuing until it flew by the targets at a speed of 7.3 kilometers per second. The first 30 seconds of data were so severely contaminated by heat-generated electronic noise that none of them could be used in the postflight analysis. For various other reasons -- some scientifically legitimate, but also including the fact that one of the medium-sized balloon decoys suddenly began to look more and more like the warhead -- the last 16 seconds of the flyby were also removed.

That leaves the data collected during the 17-second period between 33 and 16 seconds prior to flyby as the only data officially reported by the contractor. The first five seconds of this period were eventually excluded as well, because changes in both the measured brightness and the fluctuation in brightness of each target caused three different targets to look like the warhead during this short interval. The remaining 12 seconds, then, were the only time when the signals were sufficiently stable that the observed data could be matched to a template of expected warhead and decoy characteristics. But because the sensor measurements involved unknown conversion efficiencies, and it was therefore impossible to use the original template, a new template was created after the test to fit the uncalibrated sensor data. It was this after-the-fact template, matched to almost certainly inaccurate measurements, that formed the basis of the claims about the experiment's success.

Such claims, it almost goes without saying, are meaningless.

Insisting on Success

As I have noted, in spite of the numerous and fundamental experimental failures in the first trial, TRW and the Defense Department reported that the experiment was an unqualified success.

A second, similar test was launched on January 16, 1998 -- and once again, fundamental signs of the system's inadequacy continued to be overlooked. Chief system architect Keith Englander claimed that in both tests "we were able to pick the reentry vehicle out of the target complex." Lieutenant General Lyles and his successor, Lt. Gen. Ronald Kadish, also praised the experimental results before Congress. Kadish went so far as to assert that the first two experiments had "demonstrated a robustness in discrimination capability that went beyond the baseline threat." The Lincoln Laboratory scientists who helped review the experimental claims for the Department of Defense after Nira Schwartz, the TRW whistle-blower, had raised the alert made no mention of the sensor array problems in their public report, issued in late 1998.

Between mid-1998 and December of 2001, five other trials were conducted. The decoys that were the most difficult to discriminate from warheads in the first two tests were removed from these and all subsequent missile defense development tests. These included the coneshaped decoys that had the same size and appearance as the mock warhead, the striped balloons with the same base diameter as the warhead and the small cone-shaped balloons that could easily be made to look like warheads if their surface coating and/or dimensions were slightly altered.

The only "decoy" flown in the three tests immediately following the first two trials was a very large balloon, which was easily identifiable because it was known prior to the test to be seven to 10 times brighter than the mock warhead. When the seventh test was ultimately flown, last December 3, the diameter of the large balloon was reduced somewhat-from 2.2 meters to 1.6 meters -- but it was still three to five times brighter than the warhead. And for future trials, according to accounts in the New York Times, a completely new set of infrared decoys is to be unveiled. These are to be made up only of spherical balloons composed of uniformly unvaried materials and without stripes, virtually guaranteeing that they will have perfectly steady and unvarying signals. By contrast, the dummy warheads will intentionally be deployed so as to tumble end over end. This simulates the most primitive ICBM technology, where the warhead is not spin-stabilized-so as to maintain its orientation in space and make its entry into t he atmosphere and subsequent flight path more predictable-and causes its signal brightness to scintillate wildly.

The implication of this carefully contrived choice of new decoys is chillingly clear. All the problematic shortfalls in the defense system discovered in the first two experiments have been removed through the painstaking designing of a set of decoys that would never be used by any adversary, but would make it possible to distinguish warheads from decoys in flight tests.

This should be of profound concern to every U.S. citizen. The officers and program managers involved in developing the antimissile system have taken oaths to defend the nation. Yet they have concealed from the American people and Congress the fact that a weapon system paid for by hard-earned tax dollars to defend our country cannot work.

How a Successful Missile Defense System Might Work

Whether or not one believes there is any threat serious enough to require deployment of a national missile defense, it makes no sense to advocate a concept that will not work. There is a way, though, to provide a defense that would likely be highly effective, a strategy that avoids the serious and as yet unsolvable problems posed by space-deployed decoys that I have discussed.

A "boost-phase" missile defense would target intercontinental ballistic missiles in their first few minutes of flight, while they are still being accelerated up to speed by their rocket engines. Because such a system would consist of very fast, short-range (perhaps a thousand kilometers) interceptors positioned only a few hundred kilometers from the "rogue" nations likely to attack the United States, it would be effective only over a relatively small region of the earth. While the system would be devastating when used against geographically small emerging missile states, it would be largely useless against missiles launched from vast countries such as Russia or China; it would simply not be feasible to position enough interceptors close enough to their launch sites. This is good news too, however, since it would allow the U.S. to target the Third-World states it claims to be most concerned about without provoking negative reactions from Russia and China.

In the case of North Korea, ships or converted Trident submarines could serve as launch platforms for these interceptors. Silos deployed in eastern Turkey would be effective for covering launches from inside Iraq. If a defense were required against Iran, its larger size and location would require defense sites in Turkey, Azerbaijan, Turkmenistan or the Caspian Sea.

When an ICBM was launched, it would be detected and tracked by sensors on the ground, in unmanned aircraft, aboard ships or on satellites. The interceptors would accelerate to 8 to 8.5 kilometers per second in a little over a minute. At these speeds, even if their launch were delayed for a minute or more in order to establish the enemy missile's trajectory, the interceptors could still destroy the ICBM while it was in powered flight, causing its warhead to fall far short of its target.

Unlike the proposed space-based system, this defense would be difficult to counter. Countries seeking to defeat it might try to reduce the boost-phase flight time, thereby narrowing the window of opportunity for a successful intercept. But that would require the development of highly advanced solid propellant ballistic-missile technology -- an innovation that is in a completely different league than the liquid-fuel, Scud missile technology that is currently the foundation for the missile programs of North Korea, Iran and Iraq. In addition, the technology needed to implement this defense is far less demanding than that needed for midflight intercepts in space. Because boost-phase interceptors would only need to detect the very hot plume of the booster and not the cooler warhead or decoys, such interceptors could use higher-resolution short-wavelength sensors that are easier to build and much less costly than the long-wavelength sensors used by the exoatmospheric kill vehicles of the planned nuclear-missile defense system. Finally, the ICBM booster target is large and would be destroyed by a hit almost anywhere, so the probability of a successful intercept would be very high.

Some boost-phase defense systems would certainly face significant geopolitical obstacles. Getting countries such as Azerbaijan or Turkey, for instance, to allow basing of interceptors in their territory could be a challenge. If a deployment against Iran were needed, it would also require close cooperation between Russia and the United States, which would likely increase existing Chinese concerns about a U.S. Russia alliance.

However, these and other problems are all far more manageable than those raised by the currently planned space based nuclear-missile defense system. Even the first phase of this fragile and easily defeated defense is threatening to create serious problems with both Russia and China -- while providing the U.S. with essentially no meaningful protection against them or any other potential enemy state.

A Plea for Scientific and Political Leadership

In the wake of the terrifying attacks on the World Trade Center and Pentagon, the entire civilized world will need to work to defeat the forces of ignorance, intolerance and destruction. In my view, the current attitude of the Bush administration that "we can go it alone" is one of the most dangerous and ill-considered security policies to be adopted and pursued by the United States in recent memory.

The current U.S. approach to missile defense is a direct outgrowth of the irrational idea that "we" can deal with the world without working with others. It is not only an irrational position when examined in terms of social realities, it is also irrational in terms of basic principles of physical science. It is sad and disturbing that the most technologically advanced and wealthy society in human history has displayed so little scientific and political leadership on matters that will almost certainly affect every aspect of global development in the 21st century.

Reprinted with permission from the April 2002 issue of Technology Review. Copyright 2002 by Technology Review.

home · the threat · the technology · the strategy · introduction · map
timeline · producer's chat · interviews · quiz: missile test · discussion
tapes & transcripts · press reaction · credits · privacy policy
FRONTLINE home · wgbh · pbsi

photograph © 1996 CORBIS; original image courtesy of NASA/CORBIS
web site copyright WGBH educational foundation

SUPPORT PROVIDED BY