Drop Tests in Liquid Oxygen Reveal Which Ship Materials Could Ignite on Hydrogen Vessels

As the shipping industry edges toward hydrogen as a marine fuel, an unexpected hazard is coming into focus: liquid oxygen. Liquid hydrogen is stored at roughly minus 253 degrees Celsius, far below the boiling point of oxygen at about minus 183 degrees. Any surface thermally coupled to a hydrogen tank—an under-insulated pipe, a valve stem acting as a thermal bridge, or a breached vacuum jacket—can chill below the condensation point of atmospheric oxygen. When that happens, liquid oxygen can condense and pool unintentionally in drip trays or on decks, creating a powerful oxidizer in contact with ordinary ship materials. A new experimental campaign has now tested how representative marine materials respond to mechanical impact while submerged in liquid oxygen, and the results show that the standard aerospace pass/fail criterion may miss important hazards.

The study, motivated by the scarcity of liquid-oxygen compatibility data for marine applications, used an Army Ballistic Missile Agency (ABMA)-type impact tester following the ASTM D2512 standard, the reference method developed for aerospace propulsion systems where liquid oxygen is a deliberate bulk oxidizer. Six specimen types were selected after considering whether each could plausibly contact liquid oxygen aboard a liquefied-hydrogen carrier or a hydrogen-fueled ship: uncoated SM400C rolled steel, the same steel carrying a thick-film epoxy anticorrosion coating, A5083-O aluminium alloy, the engineering thermoplastic PEEK, multilayer insulation (MLI) blankets, and a two-part cryogenic epoxy adhesive. Unlike aerospace materials chosen for oxygen compatibility, these marine materials are selected primarily for seawater corrosion resistance, weldability, and fabrication economy, and their behavior under mechanical impact in liquid oxygen was largely undocumented.

A key departure from aerospace practice was the impact energy itself. The conventional criterion treats a material as compatible if it shows no ignition under a 98-joule mechanical impact, but aboard ships, liquid oxygen formation is an incidental consequence of fault sequences rather than an intended operating condition, and the mechanical energy realistically available—say, from a dropped tool or the closing of mechanical parts—may be considerably lower. The researchers therefore varied the drop height to deliver 98, 49, and 25 joules, separating the effect of impact energy from that of contact area. Each impact was recorded with synchronized video and still imaging, and all post-test specimens were examined by optical microscopy. Ten drops were performed per condition rather than the twenty prescribed by the standard, following the practice at the Kakuda facility, with a material classified as non-reacting only if none of the ten specimens ignited.

The binary results at the standard energy were striking. Uncoated steel and the aluminium alloy showed no reaction in thirteen impacts each, whereas reactions occurred for the marine-coated steel (three of three specimens), PEEK (four of thirteen), the multilayer insulation (four of four), and the cryogenic adhesive (eight of twelve, pooled over surface states). Because the coated and uncoated steels share an identical substrate, the reactivity of the coated specimens is attributed to the organic anticorrosion coating, which introduces an oxidizable fuel at the surface. The authors note, however, that with fewer than twenty impacts these observations do not constitute formal ASTM D2512 compatibility determinations, and the uncoated steel was tested only at 98 joules, so the attribution rests on a comparison at a single energy.

Reducing the impact energy changed the picture in a material-dependent way. The marine-coated steel showed no ignition at 25 joules and only a single marginal event among ten specimens at 49 joules, later classified as charring detectable only by microscopy. PEEK, which reacted in four of thirteen specimens at 98 joules, showed no ignition at 49 joules. In sharp contrast, the multilayer insulation reacted in essentially every specimen at all three energies—four of four, seven of seven, and four of five at 98, 49, and 25 joules respectively. Post-test microscopy of an ignited insulation specimen revealed melted film and fibrous reaction products consistent with vigorous local combustion supported by oxidizer held between the laminated sheets. The cryogenic adhesive occupied an intermediate position and appeared sensitive to the surface condition of the cured material.

To go beyond pass/fail outcomes, the team applied the Impact Sensitivity index, a quantitative measure from the aerospace epoxy literature that assigns each impact to a category—burning, explosion, flash, or charring—weights each category by severity (1.0, 0.9, 0.6, and 0.4 respectively), and averages the weighted scores over the test series. The index ranges from 0 percent, meaning no reaction in any impact, to 100 percent, meaning the most severe reaction in every impact. For the marine-coated steel, the index fell from 70 percent at 98 joules, based on one explosion and two flashes among just three specimens, to 4 percent at 49 joules and 0 percent at 25 joules. Fisher’s exact test found significant differences between 98 and 49 joules (p = 0.014) and between 98 and 25 joules (p = 0.0035) for this material, though the 98-joule result rests on only three specimens, so the trend is preliminary and does not define a threshold energy.

PEEK’s Impact Sensitivity dropped from 23 percent at 98 joules—two explosions and two flashes among thirteen specimens—to 0 percent at 49 joules, but the difference was not statistically significant (p = 0.104), so an energy dependence was not confirmed. The multilayer insulation behaved markedly differently, maintaining a high index of 93, 79, and 76 percent at 98, 49, and 25 joules respectively, with no clear energy dependence. The authors suggest this persistent sensitivity may stem from the specimen architecture rather than film chemistry alone: achieving the standard 1.27-millimeter thickness required stacking fifteen to thirty thin sheets, creating interlaminar spaces, surface undulations, and irregular contact points that may retain liquid oxygen near the impact interface while introducing multiple sites of local energy concentration. This architectural effect was not experimentally separated from the intrinsic reactivity of the constituent film and spacer materials.

The cryogenic adhesive introduced a third variable: surface condition. Specimens cured in aluminium cups were prepared in four surface states—as-cured scratchy, six scratches, four scratches, and without deliberate scratches—with the scratches intended to create narrow crevices where liquid oxygen could be retained at the instant of impact. The results were suggestive but inconclusive. For the unscratched surface, the index was 36 percent at 98 joules and 43 percent at 49 joules, a non-monotonic response that hints at how specimen-to-specimen heterogeneity, such as curing-induced voids, can obscure the underlying energy dependence when sample sizes are small. Because most surface-condition and energy combinations were represented by only one to a few specimens, and because visual inspection revealed considerable variability in how the adhesive filled the cup, no causal relationship between surface scratching and reactivity could be established. A significant difference was found between 98 and 25 joules for the pooled adhesive data (p = 0.0115), but the response could not be resolved further.

The findings carry two practical messages. First, the surface presented to the oxidizer—not just the bulk material—can dominate the response: the coated steel reacted while the identical uncoated substrate did not, and the insulation and adhesive showed substantial variability among nominally identical specimens. Meaningful testing should therefore reproduce in-service surface conditions, including coatings, finish, and fabrication-induced porosity or contamination, as faithfully as practicable. Second, applying the aerospace criterion—the absence of reaction in twenty impacts at 98 joules—may not fully represent marine conditions, because materials chosen for seawater resistance, weldability, and cost react at that energy. The authors stress that this does not justify adopting a lower acceptance energy; instead, energy-resolved Impact Sensitivity data may serve as descriptive input to a risk assessment that also weighs the likelihood of liquid oxygen forming at a given location and the credible impact energy there, consistent with established oxygen-system practice and the risk-based alternative-design provisions of the IGF Code.

Important limitations bound these conclusions. Specimen numbers per condition ranged from one to thirteen, below the twenty impacts the standard prescribes, and some series were terminated early once ignition was confirmed. The severity weights were adopted from the aerospace literature and remain provisional for marine use, and the index is neither an ignition probability nor a validated risk metric. Camera sensitivity settings were adjusted case by case, potentially affecting the detectability of marginal events, and inter-observer agreement was not evaluated. The tests were conducted in pure liquid oxygen at ambient pressure, whereas oxygen-enriched condensate and pressurized conditions—where impact sensitivity is known to increase—remain untested, and the impact energies credible aboard ships are themselves unknown. The authors propose a future workflow: identify where liquid oxygen may form, characterize the exposure mode, define credible onboard impact events and energies, and use the index at the relevant energy as descriptive input when assessing material suitability. Firming up the quantitative picture, they conclude, will require at least twenty impacts per condition, controlled specimen preparation, more closely spaced energy levels, and tests in oxygen-enriched liquid air under pressure.

Subject of Research: Liquid oxygen impact compatibility testing of marine materials for liquefied-hydrogen carriers and hydrogen-fueled ships

Article Title: Liquid-oxygen impact reactivity of marine materials for liquefied-hydrogen carriers/fueled vessels: effects of impact energy and surface condition

Article References: Yanagimoto, F., Suga, D., Sunami, K., Tazawa, A., Kado, Y., & Takada, S. (2026). Liquid-oxygen impact reactivity of marine materials for liquefied-hydrogen carriers/fueled vessels: effects of impact energy and surface condition. Results in Engineering, 32, Article 113741. https://doi.org/10.1016/j.rineng.2026.113741

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113741

Keywords: liquid oxygen, liquid hydrogen, hydrogen-fueled ships, liquefied hydrogen carriers, mechanical impact testing, ASTM D2512, Impact Sensitivity, multilayer insulation, cryogenic adhesive, marine coatings, PEEK, decarbonization