A laser is like a tuning fork that rings at a single note. But even in complete silence, its sound trembles slightly. Previously, it was thought that this tremor, random kicks from spontaneous emission of light by atoms inside the laser, was the main limit to precision. New work shows there is a second source of noise, quieter but omnipresent.

The authors derived that the minimum frequency shift (how weak a signal can still be detected) for an ideal single-mode laser is the geometric mean between the measurement bandwidth and the Schawlow-Townes linewidth, multiplied by a coefficient. Under certain conditions, this coefficient can be much larger than unity. In other words, record sensitivity is not just a narrow line but a balance between measurement speed and fundamental phase noise.

Three schemes were tested: an unbalanced Mach-Zehnder interferometer, a passive Fabry-Perot cavity, and heterodyning with a reference laser. For each, conditions were found where vacuum noise ceases to be an unnoticed background and begins to dictate the limit. This matters for sensors that detect gravitational waves, monitor seismicity, or search for dark matter.