Why some metals turn brittle in the cold
The transition temperature, and the failures it has caused.
Some metals turn brittle in the cold because the same slow, absorbing kind of deformation that lets them bend and yield before breaking at room temperature depends on thermal energy the metal's own atoms need to slip past one another, and once a metal is cold enough that its atoms no longer have that energy readily available, it stops yielding gracefully and starts fracturing suddenly instead, with barely any warning bend beforehand.
The physics of a metal's ductile-brittle transition
A metal that behaves in a reassuringly forgiving way at room temperature, bending visibly and absorbing energy before it finally gives way, is relying on internal planes of atoms sliding past one another under load, a process called slip, which needs a certain amount of thermal energy to happen easily. At lower temperatures that energy becomes scarcer, and slip becomes correspondingly harder to trigger, so the metal increasingly resists deforming gracefully and instead accumulates stress internally until it releases all at once through a sudden crack racing across the material rather than through the slow, absorbing stretch that characterised the same metal only a little warmer. This shift is called the ductile-brittle transition, and it is not a gradual fade from one behaviour to the other so much as a genuine crossover, often concentrated within a fairly narrow band of temperature, on one side of which the same metal, same shape, same load, behaves completely differently to how it behaves on the other. A finer grain structure inside the metal generally pushes that crossover colder, since a boundary between smaller, more numerous grains interrupts a growing crack more often on its way through, which is why two batches of nominally the same steel, processed slightly differently, can show meaningfully different transition temperatures despite matching every other property on the datasheet.
The twist-tie comparison
Twisting the thin wire inside an ordinary plastic-coated bread-bag tie is a small, familiar pleasure precisely because it bends smoothly around a finger and stays exactly wherever it is bent, giving no hint of resistance before settling into its new shape. The same tie, left overnight in a car's glovebox through a genuinely cold night and twisted again the next cold morning, can be felt to snap cleanly at the very first bend, the identical piece of wire behaving as though it had become an entirely different material purely because of the temperature it happened to be at when asked to bend. Warming that same snapped tie back up in a closed fist for a minute or two and trying a fresh piece from the same reel confirms nothing structural was actually lost overnight, since the wire itself is unchanged, only the ease with which its atoms could rearrange under load, which is precisely why the transition is a property of temperature and material together rather than a one-way piece of damage.
Why welded steel structures are especially exposed to this risk
Ordinary structural steels are among the metals most affected by the ductile-brittle transition, and a welded joint concentrates several of the conditions that make the risk worse in one place at once, a rapid weld cool-down that can leave the surrounding steel in a slightly different, more transition-prone internal structure than the parent plate, a stress concentration at the weld's own toe where a crack finds it easiest to start, and in a large welded structure, a single continuous path of steel a crack can run along for a considerable distance once it begins. Eberhart's Why Things Break documents in detail how a class of welded cargo ships built to a single wartime design suffered a number of sudden, catastrophic hull fractures, several of them occurring while the ship sat still in cold harbour water under no unusual load at all, traced afterward to steel that had simply crossed below its own brittle transition temperature at exactly the wrong moment. Investigations into that whole series of failures did more than explain one wartime shipbuilding programme, since they pushed fracture behaviour itself into becoming a property engineers specify and test deliberately, rather than an assumption quietly carried over from how a metal happened to behave on whatever day it was last tested.
One figure worth keeping in mind
The transition for some ordinary structural steels sits uncomfortably close to an ordinary winter's coldest days, while other steels, alloyed and processed specifically to push that transition down, remain safely ductile at temperatures cold enough to freeze mercury solid, a genuinely large spread in behaviour hidden behind two metals that can otherwise look, and cost, almost identically.
What this changes in practice
Choosing steel for anything that will see genuinely cold service, a ship's hull, a pipeline running through a cold climate, structural steelwork on an exposed site through winter, means checking where that specific grade's transition temperature actually sits rather than assuming any steel behaves the same way any other steel does, since two steels of similar strength at room temperature can differ enormously in how cold they can be taken before losing their forgiving, ductile behaviour altogether. Impact testing at the coldest temperature a structure will realistically see is the direct way this gets checked, rather than trusting a room-temperature strength figure to say anything reliable about performance in the cold. This is usually done with a notched impact test, commonly called a Charpy test, measuring how much energy a small standard sample absorbs when struck sharply at a chosen temperature, a figure that falls away sharply once the test temperature crosses below a given steel's own transition.
Where this stops being true
Not every metal shows a sharp ductile-brittle transition at all. Metals with a face-centred cubic crystal structure, aluminium and copper among the most common structural examples, generally remain ductile all the way down to genuinely cryogenic temperatures, with no equivalent sudden crossover into brittle behaviour the way body-centred cubic metals like ordinary steel show. This is one of the reasons aluminium alloys are chosen for structures that must operate reliably in extreme cold, since the specific failure mode that makes cold such a serious concern for steel simply does not apply to them in anything like the same way.