The most accurate thermometer in the known universe sits in a rather nondescript white building in Teddington, England, on the campus of the U.K.’s National Physical Laboratory (NPL). It looks nothing like a slender tube filled with mercury or colored alcohol. Instead, it’s a copper vessel about the size of a large cantaloupe, filled with dilute ultrapure argon gas and studded with microphones and microwave antennas, precisely shaped by a diamond-tipped lathe so that its radius varies with an uncertainty of only about 12 atomic layers of copper. The purpose of this thermometer is not really to measure temperature, however. Rather, new results from this and other similar devices could soon allow scientists to redefine temperature completely and bring it in line with the meter and other standard international units of measurement.
What the device actually measures is the relation between energy, as measured in joules, and temperature, as measured in the international standard unit, the kelvin. This relation is expressed as the Boltzmann constant and, in a perfect world, would be the kelvin’s ideal physical basis. That it’s not is purely a historical accident born of the fact that most of our planet’s surface is covered with liquid water, a substance which conveniently changes to ice or vapor at well-known thresholds of temperature.
Because our standard conceptions of temperature are intimately linked to these phase changes of water, in principle we currently only know two temperatures exactly: absolute zero, the temperature at which atomic motion ceases, and 273.16 kelvins. That latter temperature is the so-called triple point of water, which can exist stably there as ice, liquid and vapor. In 1954 an international agreement defined the kelvin as 1/273.16 the difference between absolute zero and water’s triple point.
All thermometers around the world are ultimately calibrated against this triple-point temperature, which itself is calibrated using specially constructed triple-point cells filled with a carefully concocted mixture of water that has a certain, specific ratio of hydrogen and oxygen isotopes. (The international standards community refers to it as the Vienna standard-mean ocean water.) This system generally works quite well—a recent comparison of triple-point cells from across Europe revealed temperature variations between them of only some 20 millionths of a kelvin—but it begins to break down for temperatures far from water’s triple point, such as those found in cryogenic systems or within stars. Extremely low or high temperatures must be measured using standards such as the freezing point of copper or the electrical noise rippling through a resistor, all of which must be circuitously tied back to water’s triple point with varying amounts of uncertainty and approximation.
“It’s bonkers that the kelvin doesn’t directly relate temperature to energy,” says Michael de Podesta, an NPLmetrologist who leads the team responsible for the world-record thermometer. “It only happened this way because people started measuring temperature long before they knew what it actually was, before temperature was known to just be atoms and molecules buzzing around. Now that we know better and have the opportunity to correct it, we should.”
In pursuit of that goal, de Podesta and researchers at Cranfield University in the U.K. and the Scottish Universities Environmental Research Center spent the past five years building and refining their thermometer. It’s technically an “acoustic resonator”—when certain frequencies of sound are piped in through its microphones, it will ring like a bell. Comparing that sonic resonance with the vessel’s radius, which is measured via microwaves, de Podesta and his team can pin down the speed of sound within the gas-filled cavity, and thus the average speed at which the argon molecules are moving—that is, how much kinetic energy they contain. By doing all this while also holding the resonator’s temperature steady at 273.16 kelvins, at water’s triple-point, de Podesta and his team have made the most accurate measurement yet of the Boltzmann constant, pinning it at 1.38065156 (98) X 10-23 joules per kelvin. That “(98)” is a statistical measure of doubt about the two preceding digits and corresponds to an uncertainty of 0.7 part per million (ppm). The team’s findings are published July 11 in the journal Metrologia.