WATCH SCHOOL: Why Cold Makes a Mechanical Watch Run Wrong

Late in a COSC test, a tray of uncased movements goes into a cabinet that's cooled down to 8°C and stays there for 24 hours. A couple of days later the same movements spend a day at 38°C. The lab subtracts the cold rate from the hot rate and divides by 30. To be certified as a chronometer, a movement has to come out at no more than 0.6 seconds of deviation a day per degree.

Put a watch with a plain steel hairspring and a brass balance through the same test and the figure would be around 11 seconds a day per degree. Across the 30-degree span, that adds up to 330 seconds, or five and a half minutes a day. For most of the history of the mechanical watch, temperature was the biggest source of error, and almost all of it came from one part, the balance spring.

The balance and its spring form an oscillator. The spring pushes the balance back toward its resting position, the balance overshoots, and the spring pushes it back again. How quickly this happens depends on two things. One is the stiffness of the spring. The other is how the balance’s mass is distributed around its axis, which watchmakers call its moment of inertia. A stiffer spring speeds the balance up. A rim that carries more of its mass farther from the center slows it down.

Steel gets softer as it warms and stiffer as it cools. Engineers describe this through the elastic modulus, a material’s resistance to bending, and in ordinary steel it falls steadily as the temperature rises. A hairspring is a very thin strip of metal flexing back and forth several times a second, so a small change in modulus reflects directly in the rate. In heat the spring weakens and the watch loses time. In the cold, it gains, because the stiffer spring drives the balance through each swing faster. The balance adds to the effect, since a brass rim shrinks slightly in the cold and brings its mass closer to the center. Both changes push the rate in the same direction, but the spring accounts for most of it.

The first attempt at fixing this problem focused on the balance. In the late 18th century, the English chronometer makers John Arnold and Thomas Earnshaw built balances whose rims were made of two metals, brass on the outside and steel on the inside. Brass expands and contracts more than steel. If you cut the rim near each arm so the ends are free, a change in temperature makes those ends curl. In heat they bend inward, moving mass toward the center and speeding the balance up to make up for the weaker spring. In cold they bend outward and slow it down. Earnshaw fused the brass directly onto the steel instead of riveting the two together, and his version became the standard marine chronometer balance from around 1800.

If you have ever looked at an old pocket watch movement and seen a split balance studded with screws, this is what you were looking at. The screws let a watchmaker adjust where the mass sits, and getting the compensation right took time and skill. It also had a limit. The balance’s correction changes roughly in a straight line with temperature, and the steel spring’s error does not. You could make the two cancel at two temperatures, but between those points the watch gained, and outside them it lost time. This became known as the middle temperature error, and chronometer makers spent much of the 19th century adding auxiliary devices to reduce it.

The lasting solution came from a physicist. Charles-Édouard Guillaume was born in 1861 in Fleurier, in the Val-de-Travers, the son of a watchmaker. In 1883 he joined the International Bureau of Weights and Measures at Sèvres, outside Paris, where he spent his whole career and served as director from 1915 to 1936. The bureau’s standard meters were made of platinum-iridium, which was stable and very expensive. Guillaume went looking for a cheaper alloy that would keep its length.

He found it in 1896. An iron alloy with about 36 percent nickel expanded far less with heat than any metal known at the time, and he named it Invar. It went into clock pendulums and surveyors’ tapes, and it also interested the watchmaker Paul Perret, who asked Guillaume for a sample and made a hairspring from it. A steel spring made a watch lose about 11 seconds a day for every degree of warming. Perret’s Invar spring made it gain about 18. Since the two materials erred in opposite directions, some nickel steel between them should barely err at all.

Guillaume and Perret spent the summer of 1897 testing alloys this way, building them into hairsprings and timing the watches at different temperatures. In 1899 Guillaume went after the middle temperature error directly. He replaced the steel inner layer of the compensation balance with a nickel steel of his own, later called Anibal, whose expansion behaved in a way that matched the spring’s curve. The integral balance, now usually called the Guillaume balance, nearly eliminated the error and was used in the best observatory and chronometer movements. In 1919 he added chromium to the iron and nickel and produced Elinvar, from élasticité invariable. Its elasticity hardly changed with temperature, so a spring made from it no longer needed a cut bimetallic balance to correct it. A watch could run a plain, solid balance instead.

In 1920 Guillaume received the Nobel Prize in Physics for his work on the anomalies of nickel steels and their value to precision measurement. It was a physics prize for materials research, and watchmaking was one of the fields that got the most out of it. Within a few years, Elinvar springs were standard in mass-produced watches.

Elinvar was not the end point, however. In the early 1930s Reinhard Straumann, an engineer at Thommen in Waldenburg, developed a spring alloy of iron and nickel with chromium, titanium and beryllium, working with the German metals firm Heraeus to melt it in a vacuum. He called it Nivarox, from the German nicht variabel oxydfest, meaning non-variable and non-oxidizing. Straumann understood that the spring’s temperature coefficient could be tuned by adjusting the mix. A balance still expands slightly when warm, so the right spring has a small thermal effect of its own, set to cancel the balance’s. Paired with uncut balances in Glucydur, a beryllium bronze that also arrived in the 1930s, Nivarox became the standard spring of the Swiss industry. Nivarox-FAR, now part of the Swatch Group, still supplies most of them.

Silicon is the newest answer, and its first attempt went badly. In 2002 Ludwig Oechslin, then working for Ulysse Nardin, in collaboration with the Swiss research center CSEM, fitted experimental silicon hairsprings to Unitas movements. Pure silicon softens a lot as it warms, and those prototypes drifted by 106 seconds a day over a 31-degree range. The fix was a layer of silicon dioxide grown on the surface of the spring. Silicon dioxide gets stiffer as it warms, the opposite of silicon, so an oxide layer of the right thickness cancels the core’s drift. Patek Philippe, Rolex and the Swatch Group funded the work through CSEM, and in 2006 Patek Philippe introduced its Spiromax spring in a material called Silinvar, named with a nod to Invar. Breguet used silicon springs the same year, Omega followed in 2008, and Rolex now uses its own oxidized silicon spring, Syloxi.

So what happens to your watch on a cold morning? On your wrist, very little, because the movement stays close to body temperature. Left overnight on a cold windowsill, a modern watch should drift by a fraction of a second per degree. If yours runs noticeably fast or slow after a day outside in winter, the hairspring is probably not the cause. Thickened oil is the more likely suspect, and that is the subject of this week’s long read.