At 20 kelvin, -253 °C, most metals in my family give up their give. Two alloys described in Nature on 14 October 2026 did not. A team from Oxford, UC Berkeley and Berkeley Lab reports that both stayed damage tolerant at that temperature, with no measurable ductile-to-brittle transition down to 20 K, the coldest point they tested. “What surprised us is that these alloys remain remarkably damage tolerant even at 20 kelvin,” said corresponding author Dong Liu of Oxford, as Interesting Engineering reported.

I am a body of timber and mycelium, so I feel the cold differently: my fibres stiffen, but they do not shatter. Iron is not so lucky. It belongs to the body-centred cubic (BCC) metals, with tungsten and chromium. As Tech Xplore explains, these metals turn more brittle as they cool because dislocations, the line defects that let a crystal bend instead of crack, get harder to move. “For decades, the ductile-to-brittle transition has been viewed as a fundamental limitation of body-centered cubic metals,” Liu said.

What the alloys do

The two alloys are Nb45Ta25Ti15Hf15 and Nb70Ti30 (atomic %). Lead author David Cook, who did the work at Berkeley and is now at Cambridge, describes a different mechanism. Nanoscale twinning, together with a local shift into an omega phase, creates new dislocations at the grain boundaries, so the alloy no longer depends on old dislocations moving through it. “Traditionally, the focus was on how readily existing dislocations can move through a metal,” Cook said. “Our results suggest we should change how we think about what governs the transition to brittleness for these metals.” Samples were cooled with helium gas and tested on Engin-X at ISIS, Oxfordshire. Crack-initiation toughness at 20 K came to about 96.5 MPa·√m for the five-element alloy and about 170 MPa·√m for Nb70Ti30.

The limits are in the source itself. The alloys still fractured. Elongation was roughly halved. Sample numbers were small, testing stopped at 20 K, and nothing has been qualified for use in any component. The applications the researchers name are liquid-hydrogen tanks, aerospace, deep-space probes and fusion reactors. These alloys are not headed for your next industrial hall.

←TODAY: In 2026 the brittle-fracture limit of BCC metals is printed in every European steel grade as a two-letter suffix.
→3012: Zurich-3012 picks its metals for the coldest hour a member will actually live through, not for a catalogue average.
Fulcrum: A lab result at -253 °C makes it easier to see the -20 °C promise already written on your drawings.

The suffix that remembers

Engineers learned this rule from ships. According to Kiel University’s materials notes, 4,700 Liberty ships had been built by 1946. By 1953, 1,250 had suffered brittle fractures and 12 had broken in two. Their steel had a relatively high ductile-to-brittle transition temperature. Imperial College London credits Constance Tipper, at Cambridge, with tracing the fracture in steel that became dangerously brittle at low temperatures. The test named after her is still used, and she told the story herself in The Brittle Fracture Story (1962). Tipper is one of the people whose work goes unseen because it worked: her research is now part of every steel order.

You can see it in S355J2. “355” is the minimum yield strength for thicknesses up to 16 mm. “J2” means a minimum impact energy of 27 J at -20 °C. JR is 27 J at +20 °C, J0 is 27 J at 0 °C, and K2 is 40 J at -20 °C. In Switzerland the rules for choosing a grade by fracture toughness are SN EN 1993-1-10, published as SIA 263.010.

PAZ has followed metals into the deep cold before, in ITER’s 17 m magnet that shrinks 50 mm: at extreme temperatures, the material’s behaviour becomes the design brief.

Atelier: Most offices specify the steel grade once, early, and nobody looks at it again, even when a canopy, an open stair or an unheated hall ends up colder than the first sketch assumed. The trade-off is plain: a tougher subgrade costs money and narrows supply, while an unchecked one moves the risk into the coldest night of the building’s life. Monday move: add a “coldest real temperature” column to the steel schedule template and make your engineer sign off on it.

Hack: Flag every steel member whose impact-test temperature sits above the coldest air it will meet. A flag does not mean the member fails: EN 1993-1-10 also weighs thickness and stress level. It tells you which members to raise with the engineer.

GRADE = {"JR": (27, 20), "J0": (27, 0), "J2": (27, -20), "K2": (40, -20)}
members = [("canopy beam", "S355J2", -15), ("open stair", "S235JR", -12)]
for name, grade, t_min in members:
    J, t_test = GRADE[grade[-2:]]
    print(name, grade, f"{J} J @ {t_test} °C", "ASK ENGINEER" if t_min < t_test else "ok")

The researchers’ metals are going to the edge of space. Yours will stand in a Swiss January. Open your steel schedule today and write the coldest temperature beside every grade.