Exploration and Engineering

A little later in 2004, the inheritance reviews held with Lockheed engineers out in Denver uncovered a more systemic problem. Lockheed, like JPL itself, required every piece of flight hardware that was delivered to the vehicle assembly floor to be accompanied by a package of documentation, analysis, and test data validating the hardware’s conformance to its requirements. The 2001 lander project had already reviewed and accepted all this documentation and its associated hardware; Goldstein, Parks, and Lockheed project manager Ed Sedivy decided not to accept the 2001 project’s certifications. Instead, they reran the hardware certification reviews. “What was amazing,” Parks recalled, “was that it wasn’t a few flaws that we found in the delivery packages; it was that the shortcomings were the rule, not the exception.” Documentation was often missing, and testing was “wholly inadequate.”7 The review panel Parks had assembled for the telecommunications subsystem had made this point most strongly in their review outbriefing, stating, “There is [sic] voluminous missing background data on inherited components,” and so the panel could not certify that the review objectives had been met.8

Goldstein and Parks’s finding that much more work would have to be done to requalify the supposedly qualified hardware produced a major cost increase, estimated at about $20 million in early 2005. At the same time, Wayne Lee, who had become the manager of JPL’s EDL group after completing his service on the MER project, was advocating increasing the Phoenix EDL task funding. Lee thought Goldstein’s team had not planned for enough testing of the EDL system and its various components, and he had taken his argument up JPL’s chain of command. In the risk-averse environment of the period, his arguments were accepted, and Goldstein made the pitch to NASA headquarters officials for more money. This gained Phoenix a further increase of $18.1 million. By May, the project’s cost cap had been raised to $385 million.10

More troubling news reached JPL in the summer of 2005, as Loren Zumwalt’s analysis of the vehicle’s thermal state came together. The thermal model showed a large variety of violations, with some components, such as the star trackers, getting too warm and others, including the lander legs and their deployment hardware, far too cold.11 The heat shield separation nuts were also too cold, and the thermal team had to add heaters to them. But by far the worst thermal exceedance, discovered sometime in the summer of 2005, was by the separation connectors that were supposed to release the aeroshell from the cruise stage prior to entry into the martian atmosphere. At the temperatures Zumwalt’s model estimated for Phoenix’s arrival at Mars, these were some 60°C below their limits. When Lockheed’s mechanical test engineers tried to operate them at those predicted temperatures, they were, in essence, frozen solid.12

Those thrusters, though, gave the project team additional headaches. In simulating the aeroshell’s aerodynamics, Langley Research Center engineers discovered that during a short portion of the vehicle’s hypersonic flight, the thrusters could cause the vehicle to tumble. This happened because at certain velocities, the effect of the thrusters reversed. When a thruster fired to tip the aeroshell in one direction, the aeroshell would tilt the opposite way. (This had to do with the way the thruster flow fields interacted with the airflow around the backshell at different velocities.) The flight software would not understand that and would again fire that same thruster to compensate, eventually causing the vehicle to tumble out of control. This was actually a “known problem” within the aerodynamics community, first discovered during the X-15 flight program in the 1960s and long ago fixed within NASA’s human flight programs. But the knowledge had not transferred over to the Mars program’s robotic efforts in the 1990s. Discovered very late in the project, December 2006, only eight months before launch, this “hypersonic thruster efficacy” challenge was fortunately easy to fix. The team reprogrammed the spacecraft to turn off the attitude control thrusters during that short piece of the entry sequence and reactivate them after it had slowed below the dangerous velocity range.16

The initial radar test took place in February 2006, at the manufacturer’s plant in Minnesota. Dara Sabahi, the chief engineer, reflected later that they found five separate, unique problems with the radar that day.18 He drafted several specialists from JPL’s radar section to help understand and fix them; in March they found ways to either eliminate or mitigate four. But in retests done at Dryden Flight Research Center in April, the team found still more problems. The radar took far longer to lock on to the surface than it should, and worse, the beam that usually locked up first was not the vertical beam intended to sense altitude but one of the side-looking beams intended to measure velocity. This produced an incorrect altitude measure. The side-looking Doppler beams also had a positive velocity bias, sensing a meter/second or so of speed that was not actually present—an acceptable amount of error for the jet fighter the radar was built for but not for a Mars lander that could not tolerate more than 2 m/sec of velocity at impact. And there were other troubles, too.

Sabahi established a tiger team to begin digging deeply into the radar. Understanding why it had these problems was made somewhat difficult by the contractual arrangements involved. The radars had been bought “off the shelf” from Honeywell, which sold thousands of them to many customers; JPL was not a large client that carried much weight with the company. And the radar was proprietary, so JPL was in a weak position to be asking for access to the company’s intellectual property. The legal challenges were made more difficult by the presence of Lockheed as a middleman, which, as the prime contractor, had actually purchased the radars, and, of course, the flight software that received and interpreted the radar signals was proprietary to Lockheed. So getting the JPL tiger team access to two sets of proprietary software was its own challenge. Sabahi credited his counterpart at Lockheed, Tim Gasparrini, for pushing the negotiations within his company forward so that the problems could be investigated and resolved.19

With only a little more than a year before launch, the tiger team’s investigation pursued several tracks simultaneously. More flight and drop tests took place in April and May out at Dryden, and technical interchange meetings with Honeywell engineers began that month, too. From his pot of EDL robustness money, Wayne Lee had had the foresight to finance development of a simulator that would allow “dropping” the flight radar thousands of times in simulation relatively quickly, and assembly of that system also began in May.

The May tests at Dryden uncovered still more anomalies. Lockheed had built a test rig that enabled a tethered radar to be dropped from a helicopter at the right altitude and velocity for a Mars lander, permitting more realistic tests than those flown by Honeywell back in February. After 33 flight and drop tests, they had accumulated 13 different problems.20 Due to the high risk posed by the radar’s increasing troubles, JPL senior management formed a radar advisory group to oversee the tiger team’s work.

Technical interchange meetings with Honeywell engineers determined that the company already knew about some of the faults and had fixed them in its current production units, but of course the units Phoenix had were built in the late 1990s for Surveyor 2001 and obsolete. And the project’s two new radars were being built to replicate the 2001 model, not the current production model, because it was important to the project’s engineers that all their radars behaved the same way. So fixes were already known for some of the flaws, narrowing what the team had to address.21

During the summer, the tiger team’s engineers were able to fix most of the remaining problems by revising the radar’s “firmware,” software that was embedded in the radar’s electronics, or by altering the lander’s software to deactivate some of the radar’s potentially problematic modes or to reject obviously wrong measurements (“data editing,” they called this). They also chose to simply accept some of the radar’s less-critical problems after analyzing them.22 In September, they gained another new problem, however, when the side-looking antennae that enabled measurement of horizontal velocity failed in the spacecraft’s thermal vacuum tests. They had been made of two different metals with different thermal expansion rates; in the extreme cold of space, the antennae became distorted and created unacceptable beam patterns. The project had to design, procure, and install new antennae by March 2007.23

But the radar team was not quite out of trouble. Later that month, the radar simulation effort turned up a new problem: the radar could lock on to the heat shield and, thinking it had detected the surface close by, raise its pulse repetition frequency and trigger the final landing sequence. But the lander would be too far above the actual surface to land before running out of fuel. This was a problem that could not have been discovered by the drop test program, which had no way of simulating a heat shield dropping away from the radar. Sabahi, who had been keeping a “fever chart” summarizing the tiger team’s activities and the associated risk, had initially painted December’s risk level green. In January, he switched December to yellow, and he painted January red.26

At the same time, the project uncovered another, potentially very expensive, problem. Late in 2006, Lockheed’s engineers had pulled the 2001 lander’s backshell out of storage and found that its thermal protection coating did not meet the requirements. It was not the correct density. It had been put into storage in that condition; George Pace’s engineers had documented that and, in the risk-taking posture of the era, accepted the resulting risk. Phoenix’s engineers could not accept the risk without verification that it would not produce a failed mission, so they took samples of the material for arc jet testing at Ames Research Center. The team also attempted to replicate the material to see if newly made coating would pass the tests, but they found the material could not be replicated. Something was wrong with the manufacturing process. And when they put the vehicle into its first thermal vacuum test in December, the protective coating cracked—they could not know how deeply until they pulled it out again in January, disassembled it, and performed some tests. But by then, the project was out of money.

By this time, NASA officials had become very uncomfortable about the Phoenix project. Dara Sabahi remembers that they increasingly did not want to fly the vehicle—too many severe problems had been found. But they had already spent a lot of money, and critics of NASA would doubtless use cancellation as a justification for attacking the agency.29 Forcing Goldstein to finish the project within the cap was an even worse option. If the project did not finish the radar rework effort, or the backshell thermal protection replacement, or fix their latest problem (during the backshell separation test, the backshell hung up on the lander’s thrusters and bent some of them) due to insufficient funds but launched anyway, failure of the mission seemed very likely. The critics would pounce. Science Directorate officials had no good options other than approving the increase, and they did.

Goldstein had gone into the termination review assuming he would be the review’s casualty. Managers of JPL cost-capped missions who overran their budgets were generally removed; by 2007, this had happened to a few project managers in JPL’s projects funded by the Discovery program. Because of the difficulty inherent in actually cancelling a project that was far along in its development phase, this was really the only way any kind of accountability could be imposed by either NASA or JPL management. “I survived, though,” he recalled.30

During the remaining few months before launch, the project was able to clear up its known problems, and a few more. Radar tests in April 2007 indicated that the fix the tiger team had thought up for the heat shield problem, changing the pulse repetition frequency, would resolve the issue without introducing new ones. But soon after, they also found a new radar problem. The intended flight radar, newly built to meet the traceability requirements, took several tens of seconds to lock on to the surface again if it was suddenly shut off and then turned back on again (“power cycled”) when it was at 250 feet altitude. While this was not supposed to happen, the radar’s own fault-detection procedures could cause the radar to power cycle automatically. Investigation by Honeywell’s engineers determined that the fault occurred in only a single temperature range, due to an error in the radar’s firmware (somewhat ironically, the original pair of radars built for the 2001 lander did not have this error).31 The team fixed this problem too, even though it was unlikely that the radar would reboot itself, at exactly that temperature and altitude, during the real landing. And they scheduled still more flight tests, this time at very low altitudes, for the following February to make certain there were no more low altitude anomalies.

Goldstein’s engineers got Phoenix off the launch pad and on its way to Mars on August 4, 2007.34 The launch itself was flawless, and so was the post-launch checkout, with one exception: TEGA had water contamination that made it unlikely to meet one of its requirements, to be able to distinguish between normal and “heavy” water. The project’s mitigation plan was a series of long bake-outs during cruise, heating the instrument ovens to drive the undesired water out of the instrument.35

Peter Smith, the Phoenix principal investigator, was responsible for the site selection, and he relied on the help of Ray Arvidson of Washington University in St. Louis to manage the process. Principal-investigator-led missions did not require the open community process that MER and Mars Pathfinder had used; instead, site selection was done through a pair of workshops that were largely internal to the project science team, with some assistance from the Mars program office.

The science team’s site survey had generated four areas around the region that seemed to have appropriate characteristics. Data from the Mars orbiter camera and thermal emission imaging system (THEMIS) suggested that individual sites existed within those areas that met the standards, and Smith and Arvidson decided to focus on what they called “Region B,” a wedge between 120° and 140° east longitude in the interior of the Vastitas Borealis. Region B had the lowest elevation, providing the longest entry sequence (and valuable “timeline margin”). It was not, based on the MOC images, the region likely to have the lowest rock abundance, but the science team convinced themselves that the region would still be within the standard established by the Viking 2 landing site—preferred by the engineering team because it was a known, flat plain with many very small rocks.37 Because of the deep cold of the northern polar plains, the THEMIS data was not definitive about the rock abundance, though, and Mike Malin suspected that the team’s analysis was wrong. He thought dark areas in his images were likely to have lots of large rocks.

The landing site selection subteam chose a landing ellipse within Region B in the summer of 2006 and asked Alfred McEwen to image within it using his HiRISE camera while Mars Reconnaissance Orbiter was still in its “transition” orbit phase. This had to be done because by the time MRO reached its science orbit, there would no longer be enough sunlight for HiRISE to produce images at its highest resolution. McEwen’s team got eight images within Region B during September and October. The images revealed that Region B, and especially the part of it they’d been focusing their attention on, was full of boulder fields. Rocks exceeding 8 meters in size seemed to be frequent; the mission design and science teams simply couldn’t fit Phoenix’s landing ellipse anywhere within Region B that didn’t include at least one of these boulder fields. This produced what Matt Golombek termed a “freak-out period,” for they had very little time to pick new spots in the martian high latitudes for HiRISE to target for new imaging before the Sun set and the north polar region began its long, cold, dark winter. So by the tenth science team meeting in mid-October, the team had abandoned Region B and had reopened investigation of Regions A and D, far away from their original site and dominated by different geology.38 These regions were higher altitude, requiring some rework of the EDL timeline by the engineers, but MOC and THEMIS data suggested were likely to be much lower in rockiness.

Arvidson put a graduate student of his to work automating the process of rock identification and counting, with the aid of programmers at JPL.39 They relied on new HiRISE images of the Viking 2 landing site to calibrate the model, which was designed to automatically identify and count rocks within HiRISE imagery. This automation was necessary because HiRISE full-resolution images were enormous—whole walls of the MRO operations area at JPL were covered by printouts of single images—and the team lacked time and funds to do manual rock counting and then compare the results to the much larger supply of MOC and THEMIS images. While that model was being prepared and tested, the landing site effort turned to the THEMIS thermal imagery again, looking for areas of Regions A and D that appeared very cold in early morning, an indication of low rockiness.40

The area they finally chose was a broad valley in Region D, on the other side of Mars from their original site. One end of the landing ellipse just missed a large crater.41 Tentatively chosen in February 2007, the landing area could not be imaged again by HiRISE until late in 2007—well after launch—because of the onset of winter. What confidence in the safety of the landing site they could develop came from the efforts to cross-compare THEMIS, MOC, and the handful of HiRISE images already available to validate the automatic rock-counting model that was in the works. New HiRISE images obtained in the martian spring of 2008 would confirm the site safety; the project’s planned trajectory correction maneuvers would have to be used to retarget, if necessary.

Fortunately for the project, when the Sun finally rose again over the martian south pole, new HiRISE images confirmed that their valley (they named it “Green Valley”) had low slopes and few rocks.42 At the site certification review in March 2008, Lockheed’s Tim Gasparrini presented a Monte Carlo analysis indicating a touchdown failure probability of under 5 percent for the site.43 The rockiness of the Green Valley site was acceptable.

Jim Erickson, who had become project manager for MRO, had been asked in February 2006 if there was a way to photograph Phoenix with HiRISE during its descent. The inquiry had provoked a good deal of examination within the MRO project. MRO’s first priority was recording Phoenix’s telemetry stream, and that function could not be put at risk. But getting an image would be very useful for failure reconstruction—if there were a failure—and it would be a great public relations coup if Phoenix landed safely. The telemetry requirement already put MRO in almost the right position to capture an image; the challenge was to get the spacecraft’s slew rate right and HiRISE set up correctly. The camera’s pushbroom nature made that task easier. It could scan a large part of the landing ellipse, so unless Phoenix was far off its trajectory, the HiRISE team was likely to get an image. Lockheed’s engineers thought it could be done; Rob Manning, then in the Mars program office, really wanted to try to get one or two images; at the same time, in the more risk-averse climate in the 2000s, Erickson’s team had to convince JPL reviewers that they had put sufficient rigor into their analyses and understood all the risks. He gained JPL’s institutional approval after a meeting on May 1.

Out in Arizona, attention turned to getting the lander into commission. Data back from Mars Odyssey on the first overpass after Phoenix’s landing showed the solar arrays and instrument masts had all deployed correctly, and the vehicle was generating power. Krajewski’s timeline had the first sol on the surface taken up by imaging of the spacecraft and its landing site, the second sol for unstowing the robot arm and the beginning of instrument checkouts, and sols 3–6 for continuing instrument and landing site checkout. By sol 7, he hoped to get the first surface sample into TEGA. The surface itself was flat, with polygonal structures like those typically found in Arctic tundra throughout the visible landscape. It was devoid of large, and even medium-sized, rocks. Instead, like Meridiani, it had scatterings of pebbles and small rocks, mostly smaller than the robot arm’s scoop.

Krajewski’s timeline did not evolve quite as expected after the nearly perfect landing day. The biobarrier did not completely retract but could be worked around. Commands sent up on sol 2 to release the robotic arm did not reach the spacecraft due to a glitch on MRO, so Phoenix kept running a backup command sequence that generated more images.49 This put the operations team a day behind the timeline; meanwhile, another problem cropped up in the checkout of TEGA. The telemetry stream seemed to indicate that the instrument had an intermittent short circuit.50 JPL’s design requirements ensured that the short could not kill the spacecraft itself, but it could damage, or even disable, TEGA. As understanding of the nature of the problem percolated up through the project and then to NASA management, it became the single biggest threat to achieving a fully successful mission.

Ice beneath. The Phoenix lander’s robot arm camera took this image under the lander shortly after its descent. The silver objects at the top of the image are the lander’s descent engine nozzles. The project scientists interpreted the white patches as ground ice blown clear of loose surface material by the descent thrusters.

JPL image PIA10741, courtesy NASA/JPL-Caltech / University of Arizona / Max Planck Institute.

Meanwhile, on sol 16, Mars threw Boynton’s TEGA team a bit of a curve ball. The day after the sprinkle technique was tested, the team ordered another shaking of the TEGA screen, and this time it worked. The soil sample that had been sitting on top of the fourth oven’s screen suddenly fell into the oven.53 Nobody knew why. But the team ordered Phoenix to start testing that sample on sol 18.

In digging its first trench, the lander had turned up some bright chunks of material that some science team members thought had been ice, and these chunks—rather, their disappearance over the next four sols—became the subject of the project’s first big post-landing press conference. On June 20, the science team announced that the chunks must have been water ice. Carbon dioxide ice would not have still been present at the landing site’s temperature, and the other alternative, some kind of salt, would not have simply sublimated away. Only water ice would have gradually vanished over a few days. “It must be ice,” said Peter Smith at the news conference. “These little clumps completely disappearing over the course of a few days, that is perfect evidence that it’s ice. There had been some question whether the bright material was salt. Salt can’t do that.”54

Their next step was using the new sprinkle technique to feed Michael Hecht’s Microscopy, Electrochemistry, and Conductivity Analyzer (MECA) instrument, which was designed to examine soil chemistry and electrical conductivity.55 This worked well. But TEGA’s short-circuit problem reared its head again on sol 26. It appeared that the shaking they had ordered on oven #4, hoping to get the sample to drop in, had exacerbated the short circuit. The science team delayed delivery of the next sample to oven #5 while the engineers tried to develop a better diagnosis of the problem’s severity. Instead, they began testing the rasp’s ability to dig into what appeared to be an ice layer a few centimeters below the surface. But the engineers’ conclusion was that the short circuit was bad enough that the next use of an oven might be the instrument’s last test.

A few days later, Phoenix became subject of a strange eruption of media coverage, due to an Aviation Week and Space Technology article that claimed the project had briefed the White House on major new findings “concerning the ‘potential for life’ on Mars.”59 The reporter, veteran Craig Covault, had apparently heard and misunderstood some discussions among the MECA team about a find that they were still arguing about internally. Some of the MECA team members thought the instrument had found a salt called perchlorate. But TEGA, which should also have found perchlorate’s constituent atoms, had not. Perchlorate is toxic to humans (and thus relevant to human habitability), but bacteria can coexist with it. By itself, perchlorate’s existence has little bearing on the presence or absence of life. The argument over whether perchlorate was really there or not was taking place within the Phoenix team at the same time that the confirmation of water ice was happening; when Covault did not hear the controversial “find” announced at the press conference on sol 65, he apparently thought the news was being suppressed for political reasons.

All of this occurred while the science and operations teams were making two transitions. The perchlorate press conference was on sol 70; sol 76 was to be the day the project shifted back from Mars to Earth time, and sol 90, the end of the primary mission phase, was the day the project shifted to distributed operations. The engineering support team moved from Arizona back to JPL and Lockheed’s Denver facility, and most of the scientists returned to their home institutions. The overall pace of operations slowed, too. In part, this was because of the return to Earth time, making the planning cycle more challenging for the team, but primarily it was due to the rapidly waning martian Sun. Phoenix’s power state was already declining and would fall more quickly after sol 90, reflecting the return of martian fall. The lander might be able to continue operating the meteorology experiments until perhaps sol 150, but the arm, MECA, and especially TEGA would be useable less and less frequently.

The Phoenix team ultimately got five surface samples into TEGA, four during the primary mission and only one during the extended mission, on sol 136. And the sol 136 sample was not fully analyzed before the lander failed on sol 152.62 Similarly, the wet chemistry lab component of MECA was only able to analyze samples in three of its four cells; apparently, the sample delivered by the scoop to one of the cells got stuck in the cell’s inlet, blocking it.

Phoenix’s end came as a result of the predicted decline in power available from the waning Sun, combined with a dust storm and the buildup of water-ice clouds above the landing site in late October. The lander put itself into low-power safe mode on October 27, and for a few more days Goldstein’s engineers were able to get engineering data, and some science data from the meteorology instruments, back via Mars Odyssey. They hoped that shutting down some of the heaters for the robot arm, TEGA, and other functions would enable them to gain a couple more weeks of meteorology data. But November 2 was the last time Phoenix communicated with any of the orbiters; NASA officials decided to stop listening for it December 1, which would have been Phoenix’s sol 183.63

After the martian Sun rose again and the carbon dioxide ice had retreated from the site in 2010, JPL tried to reach Phoenix without success. The Mars Reconnaissance Orbiter was able to get a clear image of Phoenix in May, though, showing that one, and perhaps both, of the solar panels had been torn off.64 They had never been designed to withstand the weight of the dry ice that had certainly built up on them during the long winter and had probably simply broken away.

Some, but not all, of the cost difference is attributable to Pathfinder’s relative paucity of scientific instrumentation and the associated procurement, integration, and science team costs. Some is in testing. For example, Tom Rivellini did 15 full-scale airbag test drops for Mars Pathfinder, while MER did 52.66 Phoenix did not have airbags, but its radar team ultimately did 73 test drops.67 Testing rigor, and cost, expanded after Polar Lander’s demise. The remainder of the cost differential is likely in documentation. The MER project was not able to recover Pathfinder’s blueprints for its “build to print” goal; Dara Sabahi could not recover Pathfinder’s parachute data either. Similarly, Phoenix could not recover some of the lander test data from Polar Lander and Surveyor 2001. All this had to be recreated, and that cost money. One of the tenets of the faster-better-cheaper argument was reduced documentation; the consequence was that future missions could not reuse designs without also re-creating the original test data, raising their own costs. Reduced documentation was a short-term savings with a substantial longer-term cost.

Phoenix’s scientific results are, as of this writing, unclear. Hecht’s team confirmed the perchlorate find, and in addition to the confirmation of water ice at the site from TEGA, Bill Boynton’s team found that the site contained calcium carbonate, which they attributed to formation in thin films of liquid water.68 The perchlorate finding threw into question the old Viking landers’ results, which had not found organic material despite a constant influx of organic material on meteors. Like TEGA, the Viking landers’ mass spectrometers had heated their samples, and perchlorate would have combusted with any organics, destroying them. Another result Tamppari found memorable was the first snowfall seen on Mars. It was photographed on the meteorology experiment’s mirror, which reflected images of a wind telltale provided by Denmark.

Phoenix’s short, exciting life paralleled the onset of another planning cycle for the Mars program, though unlike that undertaken in 2000, this “cycle” would drag on for years without resolution. The Mars science community, via its Mars Exploration Program Analysis Group, remained dedicated to the dream of sample return. During 2009 and 2010, this dedication was validated by the National Academy of Sciences’s Committee on Planetary and Lunar Exploration’s decadal survey, chaired by Steve Squyres. Published in 2011, Squyres’s survey declared Mars sample return to be the highest-priority flagship-class planetary mission for the 2013–2022 period.71

Yet these were poor years in which to restart sample return planning. The worst banking crisis since the Great Depression resulted in a focus on budget cutting, and NASA was not spared. The Vision for Space Exploration program also ran far over budget, and the Obama administration, faced with shrinking tax revenues and growing costs, reopened the question of whether to develop a new launch vehicle. Portions of the aerospace industry had never been happy with the program’s focus on development of a shuttle-derived, government owned and operated launch vehicle. The United States already had commercial launch vehicles with sufficient lift capacity to serve the International Space Station with cargo, and which could be improved to carry astronauts as well; the builders of those rockets, understandably, wanted the government’s business. They found inroads into the new administration, and the administration empaneled (yet another) blue ribbon review of NASA’s exploration policy. That panel concluded the Vision for Space Exploration’s launch vehicle development needed about $59 billion over the FY 2010 budget guidelines it had been given by the White House, or $159 billion over the next decade. Completion on schedule would need, in essence, an increase of the NASA budget of about $6 billion per year.72 That was not in the cards.

The administration’s FY 2011 and 2012 budget submissions had projected NASA being flat-funded at $18.5 billion overall, with its Science Directorate also being flat funded for five years at just over $5 billion for the next five years. Much of the cut in 2012 came from funds that had once financed shuttle operations, so the agency’s bet that it could retain those funds for its commercial crew development and heavy lifter development did not pan out. In order to expand the funds available to the commercial crew effort back to what it had intended in 2011, it again had to raid other accounts. The Science Directorate’s contribution to the commercial crew development was about $162 million.

That $162 million contribution, though, hid the true damage to planetary science, which was cut $309 million.75 The remaining funds largely went to fund an enormous cost overrun on the James Webb Space Telescope, which was approved for development in 2004 for $2.5 billion but will exceed $8 billion.76 The $309 million cut to the planetary science budget amounted to a slice of about 20 percent. Mars exploration received the largest cut in absolute dollars, of $226.2 million or 40 percent of its previous budget; the smaller lunar science program begun in the Bush administration was cut by half, and will be zeroed out in 2015 unless Congress intervenes; and outer planets science was cut by about a third.77

These external pressures on the Mars program were mirrored by internal problems. Mars Phoenix’s cost overrun paled compared to its successor, the Mars Science Laboratory, which, like Mars Observer two decades before, saw its launch delayed from 2009 to 2011. The proximate cause of the delay was the inability of the vendor responsible for the “actuators,” the electric motors and gears that drove the vehicle, to deliver on schedule. The consequence was a cost increase from $1.6 billion to $2.3 billion.78 Most of this came from inside the Mars program itself: the sample return technology development program, known as “Safe on Mars,” was eliminated and the funds available for a planned mission in 2016 were reduced, leading NASA to seek a partnership with the European Space Agency for a joint mission in that year.79 The second Mars Scout mission was also delayed two years, in part due to conflicts of interest within the selection committee, and its funds became available. But funds for the next outer planet mission were cut, too. All of these transfers were supposed to be made up for in future years, but the external cuts imposed in 2011 and 2012 made some of those paybacks impossible.

NASA Mars Program budget

The consequence of all this was programmatic turmoil. After the selection in September 2008 of the second Mars Scout mission, Mars Atmosphere and Volatile Evolution (MAVEN), and the first Mars mission to be managed by Goddard Space Flight Center in Maryland, the Scout program was suspended indefinitely. During 2009 and 2010, NASA and European Space Agency officials negotiated and funded a pair of joint “ExoMars” missions to be carried out in 2016 and 2018; NASA abruptly withdrew from them in 2012, and ESA turned to Russia instead.80

Reflecting the dismay among Mars scientists at this surprising turn of events, the chairman of the Mars Exploration Program Analysis Group wrote that his colleagues were

shocked by the severe reduction in the President’s proposed budget profile for NASA’s Planetary Science Division, even though its programs have dramatically advanced our understanding of the Solar System and been so compelling to the public … The cuts threaten the very existence of the Mars Exploration Program, which has been one of the crown jewels of the Agency’s planetary exploration. This occurs at a time when the significant discoveries of the last 15 years were about to inform and enable missions to Mars that will advance dramatically our quest of discovering life beyond the Earth, pave the way for human exploration, energize future scientists, inspire the public, and enhance U.S. prestige in space exploration.81

But the damage was already done. Neither protests from the scientific community nor from the equally dismayed Planetary Society had any significant effect on the decision.82 Thus, after MAVEN’s November 2013 launch, NASA’s Mars Exploration Program, originally built around a demand for two launches every other year, will likely not have a launch until at least 2020. The next approved mission to Mars, InSIGHT, will instead be funded by the Discovery program (as Mars Pathfinder had been). And it will use a variant of Lockheed’s lander, not JPL’s Mars Science Laboratory–derived Skycrane.

The Phoenix project’s success at reengineering Polar Lander was not mirrored by success in reengineering a sustainable Mars exploration program.

Conclusion

If the Mars Phoenix mission represents the final end of NASA’s faster-better-cheaper saga, it also represents the completion of the Mars program’s “follow the water” strategy. Mars’s water has been definitively found, and via the radars on Mars Express and Mars Reconnaissance Orbiter, Mars’s shallow subsurface ice has also been fairly well mapped globally. While questions about the planet’s hydrologic past will continue to be asked—the heatflow measurements to be made by InSIGHT will hopefully address questions about whether Mars sustains deep subsurface reservoirs of liquid water, for example—scientific interest appears to be migrating back toward the Viking project’s old focus on a search for life. Or, at least, a search for the remnants of ancient life. Water is only one of the needs of Earthlike life; living things on Earth also need nitrogen compounds (or produce them as waste products), and they’ve not yet been found on Mars.

It’s important to understand, finally, that the Mars exploration program’s sinking fortunes did not reflect a similar shrinkage of public enthusiasm for Mars. Quite the opposite was true. The very same years that the Mars program was being cut witnessed continued expansion in public interest—and in the offices of the Obama administration’s NASA administrator.