Three days before Christmas, an Alpha rocket built by Firefly Aerospace lifted off from Vandenberg Space Force Base in California. The initial phases of the flight appeared to go as expected, placing the upper stage and its satellite payload, a Lockheed Martin technology demonstration satellite called Tantrum, into a parking orbit. All that was left, Firefly said as it wrapped up its launch webcast, was a second burn of the upper stage about 40 minutes later to circularize the orbit.
Hours passed, though, without an update from Firefly about that second burn and satellite deployment. In the meantime, initial tracking data from the U.S. Space Force showed two objects from the launch — assumed to be the upper stage and Tantrum — in elliptical orbits with perigees of only 215 kilometers, less than half of the intended circular orbit. At that altitude, the satellite and upper stage will likely remain in orbit for several weeks before atmospheric drag causes them to reenter.
About 12 hours after launch, Firefly finally confirmed that the second stage malfunctioned. “Alpha’s scheduled stage 2 engine relight did not deliver the payload to its precise target orbit,” the company said. It did not elaborate on the malfunction but said it would work with Lockheed and the government to investigate the problem.
That failure capped a rough year for upper stages. Among Western launch vehicles alone there were five partial or complete failures on orbital launches in 2023 (six when counting the second, suborbital test flight of SpaceX’s Starship in November.) While there is no common technical cause for the failures, they illustrate the often-overlooked complexity and challenges of upper stages that can, in some respects, be greater than those of lower stages.
From the mundane to the complicated
Sometimes, upper stage failures are relatively basic. The upper stage on Virgin Orbit’s LauncherOne had performed as designed on four launches before it shut down prematurely on the rocket’s sixth launch in January 2023 from England’s Spaceport Cornwall. (The upper stage was not used on the rocket’s first launch because the first stage engine shut down seconds after ignition.) The design of the rocket had not changed, so some kind of manufacturing defect seemed likely.
A month after the failure, Dan Hart, chief executive of the company, revealed that the failure was caused when a filter in a fuel line “was dislodged and caused mischief downstream,” he explained at a SmallSat Symposium conference panel. The company later said the filter affected the performance of a fuel pump for the upper stage’s engine, causing it to overheat and shut down.
“This is like a $100 part that took us out,” Hart said at the conference. It would have been relatively simple to fix, he said, and Virgin Orbit had already decided to use a different filter on the next LauncherOne mission. However, the company, already in financial distress, filed for bankruptcy before it could launch again, its assets later liquidated.
The Rocket Lab Electron on the right suffered an unspecified stage separation anomaly that prevented the rocket’s Capella Space payload from reaching orbit during its Sept. 19 launch. Rocket Lab returned to flight Dec. 14 using the Electron on the left to successfully launch a radar imaging satellite for the Japanese company iQPS. Credit: Rocket Lab webcast
By contrast, the failure experienced by Rocket Lab’s Electron in September was far less obvious. The Rutherford engine in the upper stage ignited, only to shut down almost immediately, leading to the loss of a radar imaging satellite for Capella Space.
“This was always going to be a highly complex issue to figure out,” said Peter Beck, chief executive of Rocket Lab, in an earnings call Nov. 8, a month and a half after the failure. He said the challenge was exacerbated by the fact the company only had 1.6 seconds of data from the first sign of a problem with the rocket to the total loss of telemetry.
Beck turned a portion of that earnings call into an electrical engineering lecture as he discussed how the upper stage failed. “We’re all going to take a little lesson in Paschen’s Law and Paschen curves,” he told an audience of financial analysts.
The failure revolved around a phenomenon known as Paschen’s Law, which describes how electrical arcs can form as a function of atmospheric conditions. In some partial vacuum conditions, like the upper atmosphere, it becomes much easier for electrical arcs to form than at both higher and lower pressures. That coincides, he said, with the region of the atmosphere where stage separation takes place.
That was the case on the failed September launch: the conditions described by Paschen’s Law allowed an electrical arc to form in the upper stage’s power system, shorting it out. That arc also required what he called a “tiny, undetectable” flaw in the wiring insulation for the arc to form. With the power system shorted out, the pumps feeding propellant into the Rutherford engine shut down.
“Look, I don’t generally believe in luck as an engineer, but in this instance, I would say that so many things had to line up that most people would say that the current probability of this occurring would be largely improbable,” Beck concluded. “If any of these things were not present, then the failure would not have occurred.”
While that failure was “largely improbable,” as he put it, that broader issue of engine failure is more common. Two first flights of new rockets weeks apart in March 2023 — Japan’s H3 and Relativity Space’s Terran 1 — failed when their upper stage engines failed to ignite.
Relativity, in a statement released a month after the launch, described the failure as a cascading series of events. The main valves for that Aeon Vac engine opened slower than expected, which affected the timing of propellant reaching the engine, including the gas generator that drives the engine’s turbopumps. The precise sequence of events needed to ignite the engine was thrown off. .
“Due to off-nominal propellant pressure and timing, the [gas generator] did not light, and the engine did not reach full power,” the company concluded.
Relativity Space’s Terran 1 rocket blasts off from Launch Complex 16 at Cape Canaveral Space Force Station March 22, 2023. Credit: Relativity Space
On the inaugural H3 launch, the upper stage engine also failed to ignite. Neither the Japanese space agency JAXA nor Mitsubishi Heavy Industries (MHI), the prime contractor for the rocket, have publicly released many details about the failure.
A Japanese-language JAXA presentation from July found that the rocket received the signal to ignite the upper stage engine, but a failure in an electrical system kept onboard computers from issuing commands to open valves and start the ignition process. The report identified several scenarios that could cause the electrical failure, but engineers did not have enough information at the time to narrow those down to a single root cause.
Because the H3 upper stage engine is similar to the one used on the H-2A, JAXA grounded the H-2A while the investigation was ongoing. The H-2A resumed launch in September, successfully launching an X-ray astronomy satellite and a lunar lander.
“We defined the corrective actions and some of them applied to the H-2A launch vehicle,” said Iwao Igarashi, vice president and general manager of MHI, during a panel discussion at World Satellite Business Week just after the H-2A’s return to flight. He did not go into details about those corrective actions, but said MHI had turned its attention to getting the H3 back on the pad. It is slated to make its next flight in mid-February.
Complex and unforgiving
The failures linked to upper stage engines not igniting point to one of the reasons why upper stages seem prone to failures. They operate in unforgiving environments with often far less margin for error than for first stages. .
“If your booster engine fails to ignite, you scrub and try again another day,” said one industry source not directly involved with the companies that suffered failures in the last year. “If your upper stage engine doesn’t ignite, it’s game over.”
Upper stages can also be more complex than boosters. A first stage’s mission typically lasts only a few minutes, from ignition on the launch pad to engine burnout and stage separation. An upper stage, though, may need to sometimes operate for hours, performing multiple burns to place the payload into the right orbit. Each of those maneuvers needs to take place precisely or else the payloads may go into the wrong orbit — as happened on the Firefly Alpha — or be lost.
The unique environments that upper stages operate in can be difficult to test on the ground, as Rocket Lab discovered with the electrical arc linked to a quirk of Paschen’s Law. All three Electron failures were caused by issues with the rocket itself (the first Electron launch failed because of a range safety problem) involving the upper stage. A 2020 launch failed because of what Peter Beck called a “really unusual thermal problem” that affected an electrical connection, while a 2021 launch failed because of a “previously undetectable failure mode” in the engine’s ignition system.
The company has previously vowed to focus more on inspections, testing and quality control for the vehicle. “We took a big step back and a had a look across the whole vehicle, and as a result, we’ve made a bunch of changes to work instructions and quality signoffs,” Beck said after the 2020 failure, suggesting those efforts go only so far.
More testing can help find those problems, but tests can be time-consuming and expensive. ArianeGroup, the prime contractor for Europe’s Ariane 6, used a special test stand operated by the German aerospace agency DLR to perform tests of the rocket’s upper stage. That included a successful full-duration test of the stage in September.
However, a Dec. 7 test of the upper stage at the same facility, to see how it performs in “degraded” conditions, was aborted two minutes in for reasons that, a month later, were still under investigation by ArianeGroup and the European Space Agency.
“We are confident that these investigations will not impact the schedule to the Ariane 6 inaugural flight,” scheduled for mid-2024, ESA said in a statement. However, given the reputation of upper stages, it wouldn’t hurt to test again, just to be sure.
This article first appeared in the January 2024 issue of SpaceNews magazine.