In the early hours of February 2, with more than 700,000 gallons of cryogenic propellant flowing into the Space Launch System at Kennedy Space Center, a sensor flagged what mission controllers were hoping they would not see: hydrogen concentrations climbing past allowable limits at the tail service mast umbilical, the 35-foot structure at the base of the rocket where a ground fuel line meets the vehicle. Controllers paused the flow, let the seal warm, cycled pressure, and tried again, a procedure borrowed directly from the Artemis I campaign three years earlier. It didn’t hold. A spike during terminal countdown pressurization tripped the automated launch sequencer, killing the simulated count at T-minus 5 minutes and 15 seconds.
Within hours, NASA confirmed what the data already made clear: the February window was gone. Artemis II, the first crewed mission beyond low Earth orbit since Apollo 17 in 1972, would not fly before March. The crew (commander Reid Wiseman, pilot Victor Glover, mission specialists Christina Koch and Jeremy Hansen) ended a pre-flight medical quarantine and were sent home to await a new date.
The delay is frustrating, but it is not surprising. Liquid hydrogen has been doing this to American rockets for decades. Understanding why, and understanding what has and hasn’t changed between Artemis I and Artemis II, is the difference between seeing this as a temporary setback and recognizing it as a structural feature of flying the most unforgiving propellant in aerospace.
The smallest molecule, the hardest seal
Hydrogen is the lightest element in the universe, and in its diatomic molecular form it is preposterously small, small enough to migrate through crystal lattice gaps in metals that would stop any other cryogenic fluid cold. At the temperatures required to keep it liquid (minus 423 degrees Fahrenheit), the metals and elastomers surrounding it contract and stiffen, opening microscopic leak paths that did not exist at ambient temperature. Every joint, every seal, every quick-disconnect fitting is a potential escape route.
This isn’t an SLS-specific weakness. It is a hydrogen-specific one. The Space Shuttle used liquid hydrogen to feed its three RS-25 main engines, the same engine design now bolted to the bottom of the SLS core stage, and hydrogen leaks scrubbed Shuttle launches with what Ars Technica once calculated was a frequency of roughly one scrub per launch attempt across the program’s 30-year history. The worst stretch came in the summer of 1990, when hydrogen leaks grounded the entire Shuttle fleet for more than six months. NASA dispatched engineer Bob Schwinghamer from Marshall Space Flight Center to Kennedy with what his supervisors described as a one-way ticket: he was not to return to Alabama until the leak was solved. His team spent three months building an exhaustive fault tree before tracing the problem to glass bead contamination in Columbia’s main engine disconnect hardware left over from post-flight maintenance.
ULA’s Delta IV Heavy, which also relied on liquid hydrogen through its RS-68 engines, faced its own version of the problem. A 2018 launch of the classified NROL-71 payload was scrubbed after elevated hydrogen concentrations were detected around the rocket, pushing the flight from December into January 2019.
The pattern is consistent across programs and decades. It is not a question of whether a given liquid-hydrogen vehicle will leak during ground processing, but when, where, and whether the leak rate can be managed within flight constraints.
A timeline of SLS hydrogen leaks
The SLS has now been through two campaigns, and both have produced hydrogen leaks at the same interface. The history is worth laying out in sequence, because it shows how consistent the failure mode has been.
| Date | Mission | Event | Outcome |
|---|---|---|---|
| Apr 2022 | Artemis I | First wet dress rehearsal; hydrogen leak at TSMU detected during tanking | Test scrubbed; rollback to VAB ordered |
| Jun 2022 | Artemis I | Second and third WDR attempts; leak managed to reduced criteria | WDR declared complete with caveats; additional rollback for repairs |
| Aug 29, 2022 | Artemis I | First launch attempt; hydrogen leak at TSMU quick-disconnect, plus engine sensor issue | Launch scrubbed |
| Sep 3, 2022 | Artemis I | Second launch attempt; larger hydrogen leak at different QD location, 2 to 3x flammability limit | Launch scrubbed; rollback to VAB for seal replacement |
| Nov 16, 2022 | Artemis I | Third launch attempt; hydrogen managed within limits | Successful launch after a controlled manual leak fix during tanking |
| Feb 2, 2026 | Artemis II | Wet dress rehearsal; TSMU leak exceeded limits during fast-fill; spike during terminal count pressurization | Countdown terminated at T-5:15; February window forfeited |
Six tanking events across two vehicles, and five of them produced hydrogen leaks at the tail service mast umbilical that required operational intervention. That is not a fluke. It is a design characteristic.
What’s actually happening at the tail service mast
The tail service mast umbilical is a ground-support structure roughly 35 feet tall that carries cryogenic propellant lines and electrical connections from the mobile launcher to the base of the SLS core stage. Where the ground-side plate meets the flight-side plate, a pressure-assisted seal must maintain integrity across a temperature swing of nearly 500 degrees Fahrenheit while hydrogen flows at varying rates and pressures. The seal also has to release cleanly at liftoff. It is, by design, a quick-disconnect fitting, meant to break free as the rocket lifts off the pad.
That dual requirement, sealing tight under cryogenic stress and then separating cleanly under launch loads, is the core engineering tension. During the Artemis II rehearsal, the leak pattern was telling: the seal held during the low-pressure slow-fill phase, began leaking as controllers ramped into fast-fill, and then spiked hard when the core stage was pressurized for terminal count.
To put the numbers in context: NASA measures hydrogen leakage as a concentration percentage, representing how much hydrogen is present in the ambient air around the seal interface. Anything above 4 percent is considered flammable. After Artemis I, NASA set an operational cutoff of 16 percent, a threshold based on external testing that determined hydrogen at that concentration could not be made to ignite. During the Artemis II rehearsal, Artemis launch director Charlie Blackwell-Thompson described the leak rate at 12 to 14 percent during filling, below the cutoff, but the reading reached 16 percent during terminal count pressurization, triggering the automated termination of the count.
NASA officials have speculated, without yet confirming, that the 12-hour rollout from the Vehicle Assembly Building to Pad 39B could be a factor. The rocket travels the four-mile crawlerway at less than one mile per hour, but the vibration and stress environment during that journey is difficult to fully model. Associate Administrator Amit Kshatriya described each SLS as its own unique vehicle and noted that the rollout environment is “very complicated.” The TSMU is mated to the stack during rollout, and engineers cannot fully characterize what happens to the seal interface during transit.
NASA’s Artemis Mission Management team chair John Honeycutt said the Artemis II leak “caught us off guard,” noting the initial data suggested either misalignment, deformation, or debris on the seal. That candor matters. Three years elapsed between Artemis I’s launch and the Artemis II rehearsal. During that time, the mobile launcher’s TSMU hardware was refurbished, the launch tower was repaired from Artemis I exhaust damage, and new pad-access platforms were installed. Despite all of that work, the same interface produced the same failure mode. This suggests, though does not yet prove, that the problem is inherent to how the SLS/mobile launcher mating geometry responds to cryogenic loading, not to a specific hardware defect that a swap can permanently resolve. NASA’s own stated hypotheses (misalignment, deformation, debris) are consistent with that interpretation: each points to a systemic vulnerability in the interface rather than a one-time manufacturing error.
What NASA has fixed, and what it can’t
By February 8, NASA reported that technicians had replaced two seals inside the tail service mast umbilical and were analyzing the removed components. The agency plans a second wet dress rehearsal before committing to a launch date, though no date for that rehearsal has been announced.
There are meaningful improvements from the Artemis I campaign. The most consequential is that the SLS’s flight termination system, the explosive charges that destroy the rocket if it goes off course, governed by Space Force range safety rules, can now be accessed and re-certified at the pad. During Artemis I, any issue with the termination system batteries required a rollback to the Vehicle Assembly Building, a weekslong process. New work platforms on the mobile launcher eliminate that constraint, giving NASA the ability to stay at the pad for two consecutive launch periods without rolling back.
That operational flexibility is significant because of the calendar pressure. The SLS’s solid rocket boosters have propellant certified through 2028, and Orion’s CO2 scrubbing system is checked out through mid-2027, so long-term expiration is not an immediate concern. But NASA officials have said that if the mission does not launch in March, they likely will have to roll back to service batteries in the interim cryogenic propulsion stage, a move that would push any launch attempt further out.
The March windows and what comes after
NASA’s updated mission availability document, confirmed accurate on February 4, identifies five primary launch dates in March: the 6th through the 9th and the 11th. Backup opportunities exist on April 1, April 3 through 6, and April 30. Each window is dictated by the alignment of the Earth, Moon, and Sun needed for Orion’s free-return trajectory and its constrained skip-reentry profile.
To hit any March date, NASA must complete a successful second wet dress rehearsal, conduct a flight-readiness review, bring the crew back to Kennedy Space Center, re-enter quarantine protocols, and execute final countdown preparations, all within approximately four weeks. That is achievable but not comfortable, particularly if the second rehearsal produces another leak.
If March slips, the April windows are available, but a rollback for battery servicing would consume the better part of two weeks in transit and processing alone, compressing the timeline for any April attempt. The final listed date, April 30, would be a last-resort window before the mission faces a longer delay into summer or fall, with cascading effects on the Artemis III landing mission currently targeted for 2028.
The signal to monitor is not whether the second wet dress produces a leak (it very likely will, based on six prior tanking events) but whether the leak rate stays within the 16 percent threshold through terminal count pressurization. Artemis I eventually launched on its third attempt after controllers managed a known leak in real time during tanking. A similar outcome for Artemis II, where the leak exists but stays within margins, would be sufficient for a “go.” A repeat of the February 2 pressurization spike would be a more serious problem, potentially requiring hardware changes beyond seal replacement.
The bigger picture: living with hydrogen
There is a reason newer launch vehicles are moving away from liquid hydrogen. SpaceX’s Falcon 9 and Falcon Heavy use RP-1 (refined kerosene), which is far denser and far less prone to leakage. SpaceX’s Starship uses liquid methane, which boils at a higher temperature than hydrogen and, while still cryogenic, is vastly easier to contain. Blue Origin’s New Glenn also runs on methane. ULA’s Vulcan Centaur, which replaced the hydrogen-fueled Delta IV, uses two methane-fueled BE-4 engines on its first stage while retaining hydrogen only on the Centaur upper stage, where volumes are smaller and the exposure to ground-system interfaces is more limited.
NASA does not have the option of switching fuels. The SLS core stage was designed around four RS-25 engines, themselves derived from the Space Shuttle Main Engine, which is specifically a hydrogen-oxygen engine. Sixteen of these engines were harvested from retired Shuttle orbiters to power the first four SLS flights. The architecture (hydrogen fuel, Shuttle-heritage engines, the mobile launcher interfaces at Pad 39B) was locked in by congressional mandate and contractor relationships more than a decade ago, with NASA itself noting SLS was developed to fulfill the NASA Authorization Act of 2010 mandate.
That inheritance comes with a recurring operational cost. Every SLS launch will require the same cryogenic ballet with the same unforgiving molecule at the same vulnerable interfaces. The question for Artemis II, and for every SLS mission after it, is whether NASA can manage that reality fast enough to keep a launch campaign from stretching into the kind of months-long saga that defined Artemis I.
Four astronauts, a 10-day trip around the far side of the Moon, and the first crewed spacecraft to travel beyond low Earth orbit in 53 years are all waiting on the answer.
