Eyes in Space—Sensors for Treaty Verification and Basic Research

Eyes in Space Sensors for treaty verification and basic research

William C. Priedhorsky and contributors

pace-based nuclear threat Laboratories were entrusted with pro- Similar to what is being done today reduction began with the sign- viding the all-important sensors. to carry out the Laboratory’s missions, Sing of the Limited Test Ban Los Alamos was a natural choice to the scientists then applied their Treaty (LTBT) in 1963. The treaty supply these sensitive “eyes” in space. expertise to building a detection sys- prohibited nuclear tests in the atmos- Since the late 1950s, researchers at tem that would behave as planned. phere, outer space, and under water, Los Alamos had used sounding rock- They also initiated new research pro- and was a significant first step toward ets to hoist neutron, gamma-ray, and grams specifically designed to further both slowing the nuclear arms race other detectors into the upper atmos- an understanding of background and curbing the environmental con- phere in order to gather data from sources and create sensors that could tamination associated with above- high-altitude nuclear tests. Those better discriminate nuclear explosions ground tests. But in the tense atmos- same instruments would be adapted from natural signals. Furthermore, phere of the Cold War, neither the for the orbital environment and the they realized from the onset that a United States nor the Soviet Union nuclear detonation (NUDET) detec- system that was sensitive to, say, would trust that the other had com- tion mission. But numerous technical lightning could be used to study light- plied with the treaty without a fool- difficulties surrounded this new mis- ning. Soon, scientists were using proof method of verification. sion, as the sensors would be subject NUDET sensors to conduct world- That method turned out to rely to a host of natural backgrounds and class research in atmospheric science, heavily on earth-orbiting satellites, obfuscating signals. Would something space-plasma science, and even astro- each of which carried a bevy of sen- as common as a lightning flash be physics. sors that would monitor the skies and confused with a nuclear event, or unambiguously detect aboveground would something as exotic as gamma nuclear detonations. The Defense rays from a supernova1 trigger the Advanced Research Projects Agency system? and TRW, Inc., were tasked with designing, building, and fielding these 1This latter question was originally posed satellites called Vela for the Spanish by Stirling Colgate, now a senior fellow at Los Alamos, in a 1959 test ban summit word “velar,” meaning to watch. Los meeting. Colgate perceptively recognized Alamos and Sandia National the connection between mission and basic research.

152 Los Alamos Science Number 28 2003 Eyes in Space

Los Alamos detectors have since nuclear states long before any bomb from the nuclear debris. Both the journeyed over the poles of the sun, was detonated and to address prob- prompt and delayed radiations can be probed two comets, and flown to sev- lems of nuclear materials control and detected by satellite-borne sensors: eral planets to gather data for basic international terrorism. This broader bhangmeters2 for detection of optical research. In turn, these highly visible concept of nuclear threat reduction signals, very high frequency (VHF) space missions aid our nation’s threat required new sensing capabilities: new radio receivers for measurement of the reduction efforts. Consider, for exam- small satellites for space observations, EMP, plus neutron, gamma-ray, and ple, the recent discovery of water on new sensors to monitor effluent x-ray detectors. Mars, discussed in more detail in the streams from factories and power These sensors studded the surface sidebar “Geochemical Studies of the plants, portable sensors for materials and filled the insides of the Vela Moon and Planets” on page 166. trafficking, and sensors that could satellites, which were the first used to Scientists announced this find after operate in cyberspace to detect subtle verify the LTBT. The Velas operated analyzing the neutrons coming from patterns and connections in large in pairs, with satellites occupying the Red Planet. Because they are gen- masses of data. Like the NUDET sys- opposite sides of a nearly circular erated by cosmic rays bombarding the tems, these advanced technologies orbit that lay about one-third of the Martian surface, the neutrons have a double as research tools and have led way between the earth and moon. known energy spectrum, which to more discoveries of our planet, the Their sensitive instruments could see becomes slightly distorted if they col- solar system, and the cosmos. the entire surface of the earth, as well lide with water molecules. Those as a large region of space surrounding spectral distortions were “seen” by the the planet. highly advanced neutron spectrometer Space-Based Nuclear Event on the orbiting Mars Odyssey satel- Detection 2The name “bhangmeter” possibly lite. While the discovery of water on derives from bhang, the Indian name for Mars justly fuels the public’s imagina- Remote detection of nuclear explo- a type of marijuana. Apparently, some tion and promotes basic research, it sions is accomplished with sensors believed that anyone who thought satellite-based optical detection would work also reminds other nations of the that measure the different forms of must have been smoking something. United States’ remarkable capabilities energy coming from the weapon. Equally likely, “b-hang” derives from a in neutron detection, in case any Neutrons, gamma rays, and x-rays are two-syllabic way of pronouncing “bang.” This pronunciation mirrors the detection nation needs reminding. emitted promptly within about 2 mil- of the two distinct optical peaks (one NUDET detection for treaty verifi- liseconds of the detonation. Those short and one long) characteristic of an cation and situational awareness radiations then interact with their sur- atmospheric nuclear explosion. remains a Los Alamos mission. The roundings to produce secondary radia- radiation detection system on the Air tions, including visible light and elec- Force’s Defense Support Program tromagnetic pulses (EMPs), in the part (DSP) satellites and the Global of the radio-frequency (rf) band below Positioning System (GPS) satellites a few hundred megahertz. Delayed —the same satellites that give us gamma rays and neutrons also come hand-held navigation—are being used for NUDET detection. The last DSP satellite will be launched in 2003 or 2004, after which the next generation of GPS satellites, and perhaps another system, will carry on the mission. The end of the Cold War, however, changed the world. We needed to assess the capabilities of aspiring

Number 28 2003 Los Alamos Science 153 Eyes in Space

(a) (b) Optical signal from bhangmeter Log signal intensity

Log time (s) (c) Gamma Gamma burst

Radioactive decay Log energy rate Background level

Figure 1. Nuclear Event Detection Log time (s) (a) No single sensing technology can provide unambiguous detection of a nuclear (d) Neutron event under all circumstances, so a NUDET system employs a host of different Fission types of sensors. For events that take place in the troposphere or the lower strato- Fission sphere (within about 30 km of the earth’s surface), only the optical bhangmeters Thermal and VHF sensors yield information because the gamma rays and neutrons are absorbed in this region. The bhangmeter records an unambiguous double-humped optical signal, shown in (b), which is the result of the atmosphere becoming trans- Log counting rate parent, then opaque, and transparent again as the blast’s shock wave travels Background level through it. Unfortunately, clouds can obscure that signal from the satellite’s view. The electromagnetic pulse (EMP) is produced primarily when gamma rays “collide” Log time (s) with atoms in the atmosphere, freeing electrons. These become accelerated in the (e) X-ray earth’s magnetic field and produce a broad spectrum of radio waves. The VHF portion of the spectrum can penetrate clouds and the earth’s upper atmosphere and then reach satellite-based sensors. This mechanism for producing the EMP becomes ineffective above 30 km because the atmosphere becomes transparent to the gamma rays. But in this intermediate region, the neutron and gamma-ray sensors become useful. Schematic data from these sensors are shown in (c) and (d).

For events in the ionosphere (above 60 or 70 km), NUDET detection is augmented Signal intensity by data from particle detectors and (e) x-ray data. Log time (s)

All told, six pairs of Vela satellites much for system shakedown as for known about that plasma region or were launched between 1963 and treaty verification. Far from being about the effects of that region on sen- 1969. The initial pair (Vela 1 and 2) empty, the space between the sun and sitive instruments. (The Velas were carried only x-ray, neutron, and the earth is filled with charged parti- also subject to hostile cosmic radia- gamma-ray detectors. These would cles that boil from the sun’s surface tion, which comes from outside the see any events that occurred high in and stream through the solar system at solar system. Thus, many skeptics the atmosphere (above about 30 kilo- supersonic speeds (the solar wind). gave the instruments no more than meters) and also in space (see Interactions between the solar wind two weeks to live. But most instru- Figure 1). Even a detonation on the and the earth’s magnetic field create a ments lasted well beyond their design far side of the moon would be detect- tenuous and highly variable plasma, lifetime of six months; some, for as ed because the nuclear blast would known as the magnetosphere, which long as a decade.) expel a gamma-ray-emitting cloud of surrounds the earth. The Velas’ orbit Adopting a bootstrap approach, debris that would quickly be seen. would carry them through the magne- scientists used the data from the first These first detectors were used as tosphere, but in 1963, little was Vela satellites to design new types of

154 Los Alamos Science Number 28 2003 Eyes in Space

sensors that would monitor the plasma background and track particle fluxes The Little Satellite That Could that could cause false signals in the other detectors. These plasma and Diane Roussel-Dupré energetic-particle sensors, plus bhangmeters and VHF sensors that could The two years 1985 and 1986 were bad ones for the U.S. space program. identify explosions that took place in Three major launches failed, and on January 28, 1986, the space shuttle the lower atmosphere (refer to Challenger exploded in full view of the entire world. These calamitous failures stopped all U.S. space launches for Figure 1), were fielded along with the more than a year and left the space other detectors on the Vela 3 through community cautious and conservative. 6 satellites. The last three pairs of satellites (officially known as Quixotically, it was during this guard- Advanced Velas) carried improved ed period that our young experimental NUDET systems, plus sensors that team at Los Alamos chose to field the monitored solar activity, terrestrial Laboratory’s first satellite. The ALEX- lightning, and celestial x-rays and IS satellite was designed to test new gamma rays. soft x-ray and radio-frequency nuclear As a series, the Velas worked detonation (NUDET) detectors. It was funded by the Department of Energy Figure A. ALEXIS superbly and were widely considered and launched by the Air Force Space The ALEXIS satellite demon- to have seen every aboveground Test Program. The rocket was the new strated new technologies for nuclear explosion that was within their Pegasus launch vehicle, which had treaty verification while carrying field of view. They established the mixed success on its first three out- out basic research in astrophysics benchmark for surveillance capability, ings. Its fourth launch on April 25, and atmospheric science. but their legacy was also one of scien- 1993, however, went well, and our tific discoveries. As discussed later in rocket gracefully ferried ALEXIS aloft to an 800-kilometer circular orbit. But this article, much of our early data on the satellite itself ran into complications caused by the launch forces. the solar wind was obtained by the Velas’ particle detectors, whereas their The Pegasus rocket was outfitted with a video camera to monitor the rocket performance and reveal whether the nose cone deployed cleanly. To our hor- gamma-ray detectors were the first to ror, the video footage that was transmitted back to the California tracking sta- observe cosmic gamma-ray bursts, an tion showed that one of the solar panels on ALEXIS had broken loose at the entirely unknown phenomenon that hinge and was dangling freely. We could not tell from the video whether any opened a new doorway into the other damage had occurred or whether the satellite was dead or alive. The first observable universe. attempts to contact the satellite yielded nothing but silence, feeding our team’s Starting in the 1970s, the Air Force worst fears. DSP satellites began carrying NUDET systems, which were continually For six frantic weeks, the team listened for a signal every time the satellite upgraded for sensitivity, dynamic passed over our Los Alamos ground station. We took a second ground station to an Air Force facility, trying to “shout” at the satellite with a bigger dish. We range, and background rejection. But took pictures of our satellite from the Air Force optical tracking station on top the basic instruments remained the of Haleakala in Hawaii to learn about its status, and we optimized our contact same as those on Vela, even though strategy. Our persistence finally paid off with a brief contact from our Los extending system capabilities into the Alamos station, followed by a longer contact and an understanding of the extreme ultraviolet (soft x-ray) and rf satellite’s problems. We formulated a recovery plan, and ALEXIS revived as bands had always been a goal. By the expected. late 1980s, we had concepts for new sensors to operate in those extended ALEXIS was planned as a high-risk, one-year mission. However, as ALEXIS frequency bands. approaches its 10th birthday, it is still fully operational, operated by an auto- Unfortunately, these new devices matic ground station in the Physics Building at Los Alamos. The solar panels are losing the ability to provide charge to the batteries, the commercial nickel- presented us with a problem. While cadmium batteries have some trouble charging, and protons from recent solar we could verify their operation in the storms have damaged parts of its memory, but ALEXIS is still “the little satel- laboratory, in space they would be lite that could.” subject to large and poorly understood backgrounds. We needed to test them

Number 28 2003 Los Alamos Science 155 Eyes in Space

Figure 2. The FORTÉ Satellite (a) the intense electrical discharge pro- and a View of Lightning duces a burst of rf noise that can (a) The sketch shows an artist’s con- mimic the nuclear EMP. The flip side ception of FORTÉ in orbit. The radio is that our VHF/EMP sensors are antenna, which is pointed toward the excellent lightning detectors that can earth, is deployed to 11 m in length be used for basic research. from a storage container the size of an Blackbeard, for example, enhanced office wastebasket. (b) This plot shows our understanding of how the iono- frequency vs time for a particularly sphere modifies lightning-induced rf strong NBE collected by FORTÉ. The two pulses correspond to the direct pulses that pass through it and, in the pulse from the lightning and an echo course of its operation, discovered reflected from the ground. The spacing TIPPs (for transionospheric pulse of the two pulses can be used to infer pairs), or doublets of brief, transient the source height. Free electrons in rf events that form in energetic thun- the earth’s ionosphere cause the derstorms at 8 to 10 kilometers above lower-frequency components of the the earth’s surface. signal to arrive later than the higher- Other successes soon followed. frequency ones. We quantify and When compared with ALEXIS, the remove this effect to deduce when the event would have arrived at the satel- FORTÉ (for fast on-orbit recording of lite if the ionosphere were absent. If (b) transient events) satellite, which was we see the event from four or more launched in 1997 and is still opera- satellites, we can use these timings to tional, was a step-up in size and solve for x, y, z, and t and locate the sophistication. Its primary mission event in three dimensions. was to demonstrate new rf detection technologies that were to be at the core of the V-sensor, a new EMP sensor that will fly on the next genera- Frequency (MHz) tion of GPS satellites. Over the years, FORTÉ mapped optical and rf backgrounds, tested detection algorithms, and provided a wealth of data on the in space, but unproved instruments the spacecraft and on the ground and physics of lightning and the iono- could not be deployed on a commer- to run the satellite almost sphere. cial or military satellite—the cost of autonomously. ALEXIS was the first One of the first and most basic of failure was too high. satellite for which the weight and vol- FORTÉ’s findings was an explanation Unable to fly these sensors on ume of the scientific payload was for the TIPPs observed by someone else’s satellite, we chose to greater than the nonpayload (batteries, Blackbeard. A lightning discharge fly them on our own. We assembled a solar panels, structural components, between clouds in the troposphere small, dedicated team to design and and others) remainder of the space- (the roughly 20-kilometer-thick build Los Alamos’ first satellite and craft, and was one of the first to use atmospheric layer closest to the sur- called in Sandia National Laboratories computer memory instead of a tape face of the earth) produces an rf pulse and AeroAstro, Inc., a start-up small recorder for data storage. After a that reflects from the ground, so that a satellite company, to help. Named somewhat shaky start—recounted in pair of pulses is detected by the sen- ALEXIS (for array of low-energy x- the box “The Little Satellite That sor. TIPPs are closely related to ray imaging sensors), our satellite was Could” on page 155—it performed another unusual lightning phenome- launched in 1993. The first of the beautifully. non, narrow bipolar events (NBEs), “faster, cheaper, better” satellites, its ALEXIS carried a set of soft x-ray which are intense, in-cloud rf events sophisticated design included a novel imaging telescopes and an rf receiver, that occur during thunderstorms and uplink/downlink protocol, similar to called Blackbeard, that was intended last less than about 20 microseconds the file transfer protocol used on the to help us understand lightning (see Figure 2). They are the brightest, Internet, which allowed us to have events. Lightning is a common back- most common form of lighting seen very simple, inexpensive antennae on ground for our VHF sensors because by our orbiting sensors.

156 Los Alamos Science Number 28 2003 Eyes in Space

(a) (b) the spectral calibration and by an inability to fully compensate for atmospheric effects. The multispectral thermal imager (MTI), developed jointly by Sandia and Los Alamos National Laboratories and launched in early 2000, was meant to demonstrate advanced imaging and image-processing techniques that could be used in future systems. A major component of the MTI project was absolute calibration of the instrument, which is excellent and the best in its class. MTI takes data in 15 spectral bands, rang- ing from visible to long-wavelength Figure 3. Imaging with MTI infrared, which, when combined and The MTI is one of the most accurately calibrated thermal imagers ever launched into analyzed, provide information about orbit. It gathers data in 15 frequency bands—from the infrared to the ultraviolet. (a) surface temperatures, materials, water An optical image taken by MTI of a section of the Lake Ontario shoreline at quality, and vegetation health. Rochester, New York, reveals certain types of information, for example, the existence Additional spectral bands provide of offshore sandbars. (b) A thermal image of the same area reveals other features. simultaneous information about the In this false-color image, red represents hot temperatures, whereas blue represents atmosphere, such as the amount of cool ones. We can now see a plume of hot water from a water treatment plant enter- ing the lake. Data from all spectral bands give us valuable information for detecting water vapor and the aerosol content. and characterizing an area or facility. All this information helps us construct the profile of a remotely located facility or area of interest (see Figure 3). As it turns out, the occurrence rate Advanced Systems. Although we Multispectral data are also exceed- and source height of NBEs are excel- are still advancing the science of ingly useful for conducting basic earth- lent statistical indicators of the deep nuclear event detection, the alarming science research. The satellite doubles convective strength of the parent rise of nuclear-capable states in the as a national and international resource storm. Deep convection, or convec- waning years of the twentieth century that provides data to a large number of tion between the lower and upper tro- called for an expanded mission. We researchers. For example, MTI was posphere, is the driving mechanism needed to develop surveillance sys- used to study the volcanic eruption of for several forms of severe weather tems that could be used for detecting Popocatepetl in Mexico in January 2001 on the earth and is a primary means and characterizing facilities that might and the effects of the Cerro Grande fire by which energy—in the form of be producing weapons of mass that swept through Los Alamos in May latent heat—drives the large-scale destruction. But gleaning information 2000. The MTI team at Los Alamos has atmospheric circulation. It is also the about an unknown facility is far more built the Data Processing and Analysis primary means by which the atmos- difficult than gleaning the specifics of Center to distribute data to the national phere injects water into the strato- a nuclear blast. The latter presents a user community. sphere, where it profoundly influ- well-defined signature of gammas, Los Alamos scientists have also ences the radiative and chemical bal- neutrons, and electromagnetic radia- developed ground-based advanced ance of the atmosphere. Once the new tions, whereas the former oftentimes imaging systems. Among them is V-sensor is in orbit, we will be able presents a patchwork of subtle signals RULLI (for remote ultralow light to use its data to map atmospheric that make sense only after detailed imaging), a single-photon detector and deep convective processes in a near- analysis. In general, a modern surveil- imager that can accurately and simul- real-time, global manner, particularly lance system will take images at sev- taneously measure the position and over oceanic regions where weather eral wavelengths, or spectral bands. absolute arrival time of individual radar coverage is limited. Such maps Unfortunately, interpreting and piec- photons coming from a target area. will be used to support commercial ing together the spectral information The result is a data set that contains and military aviation. is often hindered by uncertainties in full three-dimensional (3-D) informa-

Number 28 2003 Los Alamos Science 157 Eyes in Space

(a) (b) 14 12 10 8 6

Height (m) 4 Photon 2 180 200 220 240 260 Photocathode Distance (m) Photoelectrons Figure 4. Imaging with Single Photons (a) The RULLI technology allows us to take 3-D images of subjects that are cloaked MCP in darkness. Individual photons (say from starlight) that reflect off objects are Z-stack sensed by the RULLI detector and converted into well-defined electrical pulses in the crossed delay lines. Coupled with fast analog electronics and a processor, the sensor system measures the position and time of each photon event. If we use a 107 electron charge cloud pulsed laser with known timing characteristics to bounce photons off the subject, we can measure each photon’s roundtrip time of flight. We can deduce the distance Y2 to the subject with an accuracy of a few centimeters and reconstruct a 3-D image. (b) RULLI’s 3-D imaging capabilities allow us to see both the forest and the trees in Y1 a Los Alamos canyon. We observed this local scene from a distance of 235 m by illuminating the trees with a 6-mW pulsed laser (about as bright as a laser X1 CDL readout pointer).The image appears to be a head-on shot, but it is not. The laser and detec- X2 tor were located to the left, looking up the sloping canyon, and the horizontal axis of this picture corresponds to distance. We can reconstruct this view through the trees only because we have full 3-D information.

tion about the area (see Figure 4). allows translating human knowledge hundreds of algorithms, each of which Because it can detect activities con- into an algorithm that can recognize finds (to some degree), the regions of ducted under the darkness of night, objects and patterns in data streams. water in the training set. The program RULLI and its successor technolo- At its core, GENIE is a computer ranks the algorithms according to a set gies can be used for various threat- program that develops other computer of “best-fit” criteria. reduction applications, including air- programs (algorithms). It does so by Although the top-ranked algo- borne, large-area surveillance for using genetic programming techniques, rithms may work very well, typically perimeter protection. which are methods for automatically they do not find all the features of creating a working computer program interest. The analyst then goes by combining, mutating, or rearranging through the training set again, retag- From Outer Space to low-level, nonspecific computer func- ging missed features or flagging Cyberspace tions or programs. As its name implies, incorrect ones, and GENIE reworks genetic programming draws its inspira- the top-ranked algorithms. After a few The body of data returned by tion from biology, where new species such iterations, GENIE “evolves” an advanced systems such as MTI, emerge through the exchange and algorithm that is optimized to find the RULLI, or other signal collection and modification of chromosomes. features of interest (see Figure 5). The imaging systems is huge. Human ana- Training GENIE to find selected analyst can retrieve the optimized lysts face the nearly impossible task of features in a data set is an iterative, algorithm in human-readable code, keeping up with this deluge. evolutionary process. Starting with a automate it, and use it to chew Increasingly, we must turn to computer- small data set, or even a single image, through large, complex data sets. based image-processing tools to auto- an analyst marks features of interest— GENIE is a general-purpose tool mate and assist in the analysis. But a for example, all regions of water. for feature classification. Aside from computer’s ability to analyze image Given this goal and a substantial threat reduction, it has been used suc- data pales in comparison with the library of low-level image-processing cessfully in detecting cancers and remarkable human brain. Hence, we functions (for example, edge detectors pathogens in humans, looking for developed GENIE (for genetic image or spectral filters), GENIE uses genetic topographic features and minerals on exploitation), a new software tool that programming techniques to produce Mars, and mapping ash and debris

158 Los Alamos Science Number 28 2003 Eyes in Space

(a) (b) from the World Trade Center after the New York City terrorist attack. Whereas GENIE enables us to create optimized software, meeting the demands of our expanding threat- reduction mission means optimizing hardware as well. We need to couple a sensor directly to a processor and have the system shoulder much of the real-time data analysis. Unfortunately, in trying to build such a system, we quickly run into size and power restrictions. A general-purpose processing board wastes valuable processing power and real estate because Figure 5. GENIE it provides capabilities that are extra- GENIE is a computer program that develops pattern recognition algorithms from a neous to our purposes. Our data prob- limited body of analyst-supplied training data. (a) For a water-finding task, an ana- lems are so supersized that we need lyst tags pixels of interest (water) in green and undesired pixels (anything but water) in red. GENIE used this initial information to evolve the mask shown in (b), which every hardware gate to be dedicated includes all the water and nonwater in the image. The user is able to influence the to solving our task. Field-programma- evolution of algorithms by providing additional information and by interactively pro- ble gate arrays (FPGAs) deliver this viding additional training data. capability. The FPGA consists of cells that implement logical gate functions, such FPGA board User -defined node as NAND, NOR, or XOR. Each cell Data processed can be configured to perform different in parallel logic functions at different times. A programmable matrix connects the cells to each other, and those connections can be altered by signals sent to the FPGA board. Thus, a user can create different logic circuits (nodes). Similarly, the nodes can be linked together to perform all the steps that are needed for the data analysis (see Figure 6). Furthermore, the nodes process large data sets in parallel, greatly reducing analysis time. Once the task is completed, or the search criteria change, the user can reconfig- ure the FPGA to perform another task. Configurable By adding memory and input/out- Figure 6. Reconfigurable Computing put devices to the FPGAs, we build, The heart of an RCC is the field-programmable gate array (FPGA) circuit board, in fact, a reconfigurable computer whose function can be modified by software. The FPGA consists of millions of sys- (RCC). One system we have built for tem gates, which are the basis for the reconfigurable cells. Each cell can be reconfig- an RCC—we called it POOKA— ured to perform a different low-level logic function, such as AND or OR. Many cells combines genetic programming with are grouped together into a node that performs a complex function, such as edge reconfigurable hardware and allows us detection or spectral filtering. A genetic algorithm reconfigures the collection of nodes to create an optimized analysis algorithm. The advantage of the RCC is that to build a truly optimized analysis many subtasks can be done in parallel. Information is also pipelined so that new data algorithm. How much speed can can start to be analyzed even as old data move through the system. The RCC allows POOKA bring to feature classification us to do complex analysis tasks much faster and with far less hardware than was tools such as GENIE? A lot! With a previously possible.

Number 28 2003 Los Alamos Science 159 Eyes in Space

small data set, new algorithms can be (a) obtained 100 times faster on POOKA than on a conventional computer. Once the system is trained, the optimized algorithm applied to a new data set runs 20 times faster. POOKA is so fast that we are able to search in real time for features in video data streams, for example, from a surveil- (b) lance camera on an unmanned aerial vehicle. Thus, we can train the algorithm to recognize not just spatial or spectral features but also features that vary between video frames. The ability to couple a processor to a sensor and optimize the processor to Figure 7. X-Ray Map of M31, the Milky Way’s Neighbor The X-Ray Multimirror Mission Observatory allows seeing the Andromeda Galaxy perform specific tasks has allowed us (M31) in (a) optical light and (b) x-rays. Although the early Vela x-ray instruments to do multispectral analysis in real could not even detect M31, the observatory is so sensitive that it can resolve time. This achievement has revolu- 600 x-ray sources within that galaxy. Most of these sources are rare double stars tionized our surveillance capabilities that contain a neutron star or black hole. and has also opened up amazing opportunities for basic research. (See the box “Gotcha! You Blinked!” on with a source located within or near Things began to change in 1991, the opposite page 161.) our galaxy. after NASA launched the Compton Theoreticians and experimentalists Gamma-Ray Observatory. The satel- at Los Alamos were extremely active lite viewed the entire sky with an Fundamental Space Science in trying to shed light on the phe- array of relatively large detectors and and Astrophysics nomenon. Collaborating with other recorded hundreds of gamma-ray institutions, Los Alamos researchers bursts. The data clearly showed that In 1973, a Los Alamos team fielded increasingly sensitive gamma- bursts came from all parts of the sky, announced that the gamma-ray detec- ray detectors aboard the Pioneer without any preference for the plane tors aboard the fifth and sixth pairs of Venus Orbiter (launched in 1978), of the Milky Way or for regions Vela satellites had detected 16 very the third International Sun-Earth around the Andromeda Galaxy. The intense “bursts” of celestial gamma Explorer spacecraft (ISEE-3, also likely explanation was that sources rays, each lasting about a minute but launched in 1978), and the Ginga were uniformly distributed throughout consisting of a number of quick, spacecraft, which was launched in the universe. That view was solidified sharp pulses. The astounding feature 1987. But because of their small size, by the Italian-built BeppoSax satellite, of the bursts was their unbelievable those instruments were insensitive to launched in 1996. Data from the satel- brightness—often brighter than the all but the largest bursts. In addition, lite could be analyzed fast enough rest of the gamma-ray universe com- the instruments had limited spatial (within 5 to 8 hours) that ground per- bined! The discovery of bursts imme- resolution, so data had to be com- sonnel could direct onboard x-ray diately raised two scientific ques- bined with that from other spacecraft instruments to observe the source. tions: What astrophysical sources to allow accurately locating the burst BeppoSax was the first to detect an x- could emit such rapid, potent spikes in the sky. Unfortunately, the initial ray “afterglow” following a gamma- of energy, and where were those data analysis often took weeks to ray burst. The x-ray data allowed sources located? Because the intensi- complete, far too long to permit fol- researchers to extract redshifts and ty of light falls off inversely as the low-up studies by higher-resolution hence deduce a distance scale. Most square of the distance, the questions x-ray and optical telescopes. For physicists now agree that the bursts were related. Cosmic sources located many years, those deficiencies come from sources located billions of millions, or even billions, of light- limited the amount of information light-years away. years away would have to emit orders available to the gamma-ray burst Scientists are still searching for a of magnitude more energy compared community. complete picture of how bursts are pro-

160 Los Alamos Science Number 28 2003 Eyes in Space

duced and are relying on data from the latest generation of spacecraft. The Gotcha! You Blinked! HETE-2 (for High-Energy Transient Explorer) satellite, for example, W. Thomas Vestrand launched in 2000 with Los Alamos instrumentation and software, process- We take for granted that the stars in the night sky are stable from night to es burst data within tens of seconds. A night and year to year. But also populating the heavens are short-lived optical fast trigger on the gamma-ray detectors transients such as the bright optical flash of January 23, 1999, that lasted quickly relays to observers worldwide about 90 seconds and reached an apparent magnitude in brightness of 9. Estimated to have originated at a distance of 10 billion light-years, it was the event information, which elicits fast most luminous optical object ever measured by humankind. Unfortunately, responses from ground-based robotic witnessing similar events is frustratingly difficult. The flashes are generally telescopes (such as the RAPTOR sys- not preceded by other events and are often over by the time we can train a tem discussed in the box “Gotcha! You telescope to the right spot. Blinked!”on this page). Spectral information can be gathered during the cru- The solution is to adapt cial first minutes of the event, while technology that is used to the burst is still happening. In fulfill our threat reduction December 2003, NASA will launch the mission and couple optical Swift satellite. With its enormous burst sensors to real-time proces- sors. This procedure has alert telescope and Los Alamos trigger- allowed us to develop the ing and imaging software, Swift will first of a new generation of have an even greater opportunity to “smart” telescopes that can locate and observe hundreds of bursts locate, in real time, celestial per year. optical transients that come Gamma-ray bursts are but one area and go in less than a minute. Figure A. RAPTOR of fundamental space research that This double image of an asteroid approaching was advanced by Los Alamos instru- Our sky-monitoring system, the earth was taken by RAPTOR. Two tele- ments. Another is in the field of x-ray RAPTOR (for rapid telescopes, 38 km apart, took each image (shown astronomy. This work started with a scopes for optical response), in red and blue, respectively). Unlike the dis- is best understood as an tant stars, the asteroid position shifts from simple x-ray telescope that flew on analogue of human vision. one telescope to the other. the Vela satellite. Although modest in The human eye has a size and limited in performance, that wide-field, low-resolution telescope proved to be exceedingly imager (rod cells of the retina), as well as a narrow-field, high-resolution useful because it operated for more imager (cone cells of the fovea). Both eyes send image information to a pow- than 10 years. It allowed us to do erful real-time processor, the brain, running “software” for the detection of long-term studies of x-ray binaries interesting targets. When a target is identified, both eyes are rapidly turned to (peculiar double stars containing a place the target on the central fovea imager for detailed “follow-up” observations with color vision and higher spatial resolution. Because we have two black hole or neutron star) and active eyes viewing the same scene, we can eliminate such image faults as galactic nuclei (supermassive black “floaters” and extract distance information about objects in the scene. holes at the center of galaxies). That Similarly, RAPTOR employs two primary telescope arrays that are separated telescope was the forebear of the opti- by a distance of 38 kilometers to provide stereoscopic imaging (see cal-ultraviolet monitor telescope that Figure A). Each telescope array has a wide-field imager and a central, nar- we helped develop for the giant X-Ray row-field fovea imager. Both arrays are coupled to a real-time data analysis Multimirror Mission Observatory, a system that can identify transients in seconds. Instructions are then relayed to satellite launched by the European point the high-resolution fovea telescopes at the transient. Space Agency in 1999. The observatory has studied the x-ray source popu- The RAPTOR sky-monitoring system, which collected its first data in the summer of 2002, will give astronomers the first unbiased global look at the lation in the Andromeda Galaxy, the variability of the night sky on timescales as short as a fraction of a minute. It Milky Way’s nearest large neighbor has already imaged an asteroid approaching the earth (see the figure), which (see Figure 7). stands out from the stars in its field because of the parallax (position shift) Closer to home, research on the between the images taken by the two telescopes. near-space environment has been

Number 28 2003 Los Alamos Science 161 Eyes in Space

Near Space and ENA Imaging

John T. Gosling and Geoffrey D. Reeves

Far from being empty, the near-space environment is filled heated solar-wind plasma, along with plasma of ionospheric with magnetic fields and solar wind (see Figure A). The latter origin that also resides in the geomagnetic tail, is further ener- is a magnetized plasma consisting primarily of protons and gized and accelerated toward the earth, where it collides with electrons that flee the sun’s surface at supersonic speeds. As and excites particles in the upper atmosphere. The excited par- an ionized gas, the bulk of the solar wind cannot penetrate ticles then emit light that we see as auroras. directly into the earth’s magnetosphere and, therefore, must be diverted around it. Because the solar-wind flow is supersonic, Los Alamos pioneered an effort to image and map the a bow shock stands off upstream of the earth to cause the earth’s entire magnetosphere at one time, a feat that will rev- solar wind to divert around the magnetosphere. olutionize our understanding of this plasma environment. We proposed our innovative imaging technique—known as ener- As a result of its interaction with the solar wind, the day side getic neutral-atom (ENA) imaging—nearly 20 years ago, of the earth’s magnetosphere is compressed. Some of those demonstrated the principle in the 1990s, and have begun to field lines, through the process of magnetic recombination, demonstrate its full potential in the last two years. ENA become interconnected with the magnetic field carried by the imaging relies on the exchange of an electron between an

Magnetosphere

Very Bow shock Deflected solar dense wind particles plasma region Solar Geomagnetic field lines wind Low- density plasma region

Hot, Plasma sheet dense plasma Sun Earth region

Figure A. Magnetic Fields and Solar Wind in the Near-Space Environment The solar wind flows at supersonic speed and is deflected around the earth’s magnetosphere by a bow shock. The solar wind compresses the day side of the magnetosphere. Field lines from the earth’s day side recombine with the magnetic field and form a geomagnetic tail on the earth’s dark side. The tail encloses a plasma sheet of hot, high-density solar-wind plasma. solar wind and are carried far past the earth. The result is a energetic ion and a cold neutral atom. Neutral atoms in space long geomagnetic tail on the earth’s dark side. Far down- are extremely rare, and they seldom collide with ions. But stream, the magnetic interconnection with the solar wind is when they do, the ion gives up its charge and breaks free from broken, and field lines can return to the earth. This enclosed the confines of the planetary or interplanetary magnetic fields. area within the geomagnetic tail is called the plasma sheet, Except for the weak effects of gravity, the neutral atom travels and it holds a relatively high density of captured and heated in a straight line and can be imaged by a detector to “take a solar-wind plasma. During geomagnetic disturbances, this picture” of the distant plasma.

162 Los Alamos Science Number 28 2003 Eyes in Space

(a) Figure B. ENA extremely productive and has led to a 2500 number of fundamental discoveries

-1 Imaging 19.9 keV (a) How would the mag- about the sun’s extended atmosphere, netosphere look if seen 2000 the solar wind, and the interaction of from a satellite 53,000 that atmosphere with the earth’s mag-

s sr keV00) km above the earth? 2 netic field. Measurements by instru- 1500 The earth is seen at the ments on the Vela satellites revealed center, and the north- some of the complexity of the earth’s ern polar cap, where 1000 ENAs are not produced, magnetosphere and led directly to our is seen as the black, discovery of the earth’s plasma sheet, 500 oval shape in the fig- a region of concentrated plasma that ure. An asymmetric ion extends far downstream on the night ° ring, the ring current, Differential ENA flux (cm ENA Differential 125 FOV 0 side of Earth (see the box to the left). produces bright emis- Other measurements by Los Alamos sions around the north instruments on Vela led to the discov- and south polar caps ery that the sun often impulsively as it sweeps around the ejects into interplanetary space large (b) 19.9 keV earth. Ions in the low-

4 amounts of material, which have come -1 altitude “horns” of the magnetic field lines to be known as “coronal mass ejec- 3 interact with the dense tions.” Los Alamos work in the early

s sr keV) portion of the atmos- 1990s revealed that these ejections, 2

2 phere. (b) Viewed from and not solar flares, are the prime (cm

6 one direction, the ENA cause of major solar-wind disturbances

10 1 emissions are an inte-

ion and large geomagnetic storms.

J gral of a 3-D ion distri- 0 The considerable success of the 10 -10 bution convolved with a 5 early Vela measurements prompted

-5 3-D atmospheric distri- ) E )

0 0 E

(R (R NASA to use Los Alamos plasma sen- SM bution. A model of the 5

X -5 Y SM sors on a series of satellites launched 10 inferred equatorial dis- -10 tribution is shown here. in the early 1970s. That was the begin- ning of a long and fruitful collaboration between our two institutions to We produced the first dynamic images of the earth’s magnetosphere in the study the near-space environment, a 1990s, using a satellite-based instrument that was originally designed to measure charged particles. In 2000, NASA’s IMAGE (for imager for magnetopause-to- collaboration that continues to the aurora global exploration) mission was launched with three types of ENA present. Our instruments have sampled imagers (for high-, medium-, and low-energy atoms), each fully optimized to all the different regions of the earth’s image the magnetosphere. Figure B(a) shows what the magnetosphere would magnetosphere and have explored the look like if you were on a satellite 53,000 kilometers above the earth. Models of solar wind in considerable detail. the magnetic field and the distribution of atmospheric neutrals can be used with Figure 8 shows the solar-wind speed these images to determine the distribution of ions as a function of radius and as a function of solar latitude. The data local time. One such distribution is shown in Figure B(b). were obtained by Los Alamos instrumentation on the Ulysses spacecraft, a The upcoming TWINS (for two wide-angle imaging neutral-atom spectrometers) joint endeavor between NASA and the mission, which will be launched in 2003 and 2005, will have two medium-energy ENA instruments on separate satellites. The stereoscopic data will allow us to create European Space Agency. Ulysses was the first 3-D images of the magnetosphere. The data will also help advance our launched toward Jupiter from the understanding of “space weather,” or the variations in the plasma environment that space shuttle Discovery in can adversely affect, among other things, satellite communications and operations, October 1990. The giant planet’s grav- radio and television transmission, the power network, and even the safety of our itational field deflected the craft out of astronauts in space. the ecliptic and into a 5.5-year-long orbit over the poles of the sun. It is the first-ever polar orbit of the sun by a manmade object. Ulysses is now nearing completion

Number 28 2003 Los Alamos Science 163 Eyes in Space

understanding of, among other things, magnetic reconnection, collisionless- shock formation, and charged-particle acceleration and transport. These phe- nomena, in turn, have helped us construct models of basic astrophysical plasma processes. Magnetic reconnection, for example, is a restructuring of a plasma’s magnetic field, in which field lines oriented in opposite direc- tions break and reconnect to each other with a subsequent release of stored magnetic energy. Magnetic recombination, which has been evoked as the likely power source for the acceleration of charged particles into space during solar activity, is also believed to power the relativistic jets of matter that shoot out from quasars. Although magnetic reconnection cannot be studied directly from a quasar located billions of light-years away, our own near-space environment provides us with a remarkable laboratory to study the phenomenon.

Epilogue Figure 8. Solar-Wind Speed as a Function of Solar Latitude A polar plot of the solar-wind speed as a function of solar latitude was measured by As the Laboratory celebrates Los Alamos plasma detectors on Ulysses. The speed trace is color-coded according 60 years of serving society, it also cel- to the observed polarity of the interplanetary magnetic field (IMF) that threads the ebrates 40 years in space. In those heliosphere. Underlying the speed trace is a set of concentric images of the solar 40 years, we have strengthened the corona, the source of the solar wind, obtained from a combination of space and national security with sensor and pro- ground-based telescopes. A striking aspect of the plot is the high and nearly con- cessing systems and used the same stant speed of the solar wind, outward in the northern hemisphere and inward in the southern hemisphere, observed at high heliographic latitudes throughout the orbit. capabilities to explore our world from This high-speed wind originates from relatively dark regions in the solar atmos- the earth outward to the early uni- phere known as coronal holes. Low-speed wind originates in the bright coronal verse. With each generation of nuclear streamers prevalent at low solar latitudes at this phase of the solar cycle. The alter- detection sensors, we do more with nating flows at low latitudes reflect the fact that the solar magnetic dipole had a siz- less, driving our systems into progres- able (20°-to-30°) tilt relative to the solar rotation axis during the interval shown, and sively smaller but more capable pack- as the sun rotated with a periodicity of 25 days, high- and low-speed flows were ages, thanks to advances in onboard directed toward Ulysses at regular intervals. event detection, device miniaturiza- tion, and background processing. The of its second trip over the sun, during second pass. This change was due to curve of performance shows no sign which time the 11-year solar-activity the considerably more complex nature of turning over. As a result, we are cycle rose and peaked. As a pointed of the sun’s magnetic field and corona confident of many more discoveries in reminder of the variability of our local and the increased number of solar- the decades to come. environment, the relatively organized wind disturbances produced by solar nature of the solar wind measured dur- activity at this time. For more information, please visit ing the first orbit (refer to Figure 8) Near-space research has con- http: www.lanl.gov/orgs/nis/ . was noticeably more complex on this tributed substantially to a fundamental

164 Los Alamos Science Number 28 2003 Eyes in Space

William (Bill) Priedhorsky received his Maya Gokhale is deputy leader of the Space Geoffrey Reeves is leader of the Space and bachelor’s in physics from Whitman College Data Systems Group and project leader of Atmospheric Science Group in the and his Ph.D. in physics Deployable Adaptive Nonproliferation and from the California Processing Systems International Security Institute of Technology. (DAPS), which sponsored Division. He has spent After finishing graduate the Pixel-Based much time studying the work, he joined Los Multispectral Image acceleration and transport Alamos National Classification (referred to of energetic particles in Laboratory, where he has as POOKA) Project. the earth’s magnetosphere developed sensors for and their effects on space nonproliferation and systems. treaty verification. Bill’s current research includes muon radiography Jack Gosling and high-energy astrophysics. He is the chief is a Laboratory fellow in the Diane Roussel-Dupré is manager for the proj- scientist for the Nonproliferation and Nonproliferation and International Security ect Cibola Flight Experiment Software Radio International Security Division. Division, where he works Space Demonstration. In on various topics in space the past, she was the plasma physics and has Richard Belian, currently a Laboratory con- flight operations manager authored more than 350 sultant (retired from Los Alamos), was a co- for the ALEXIS and the scientific papers. investigator on the team FORTÉ satellites. that conceived and built the cosmic x-ray detector (XC) flown on the last two pairs of VELA satellites. The XC was the first viable scientific x-ray Cheng Ho is a technical staff member of the David Suszcynsky is a technical staff member instrument ever to be Space and Remote Sensing Sciences Group in and project leader in the Space and flown on a satellite. It the Nonproliferation and Atmospheric Sciences made several important International Security Group of the discoveries in cosmic x-rays, including the first Division and project Nonproliferation and ever report of a cosmic x-ray burst. leader for the Remote International Security Ultra-Low Light Imaging Division. Steven Brumby is co-inventor of the Los (RULLI) Single-Photon Alamos GENIE machine learning/image analy- Sensor Project sis system, and has been working on a range of space missions to Earth, Mars, and the outer plan- Stephen Knox is project leader in the ets. Steven received his Tom Vestrand is the project leader for the Nonproliferation and Ph.D. in theoretical particle RAPTOR (rapid telescopes for optical International Security physics in 1997 from the response) and ROTSE Division’s Research and University of Melbourne, (robotic optical transient Development Program. Australia, and is a techni- search experiment) cal staff member in the Programs at Los Alamos Space and Remote Sensing Sciences Group. National Laboratory.

Edward Fenimore has been a technical staff member at Los Alamos since 1978. At present, he is in the Space and David Lawrence is a technical staff member Remote Sensing Sciences on the Space Physics Jay Schecker came to Los Alamos National Group, where he is work- Team in the Space and Laboratory in 1990 as a postdoctoral fellow in ing on the high-energy Atmospheric Sciences nuclear physics. He has transient experiment Group at Los Alamos been a science writer with (HETE), on the Swift National Laboratory Los Alamos Science since satellites, and on theoreti- 1995. cal studies of gamma-ray bursts.

Number 28 2003 Los Alamos Science 165