A radio receiver turned edge-on to the transmitter goes quiet. Atomic magnetometers suffer from a similar problem: at certain angles between the sensor and the magnetic field, the signal nearly vanishes, and the instrument goes blind. Engineers have spent years trying to make the receiver 'listen' equally well from all directions. In the new work, the light beam inside the cell is split into two crossed beams, like two conversationalists who interrupt each other but together never fall silent.

When the pumping is turned off, the atoms keep spinning for a while like tops, and this precession decays. The rotation frequency depends on the magnetic field, and that is what the instrument measures. Previously, it was thought that the two regions worked independently, but the experiment showed that atoms excited in one region partially migrate to the other. This cross-talk, with a fraction of 0.28, explained why the signal does not disappear at any orientation.

The orientation error, that is, the difference between the measured and true field when the sensor is tilted, does not exceed 0.7 nT in a 50 μT field. That is like measuring the length of a football field to within the thickness of a hair, even if you hold the tape measure at an angle. A closed loop allows tracking fast field changes at rates up to 10,000 nT/s.

Such accuracy is already needed by geologists searching for ores via magnetic anomalies and by navigators working without GPS. In the future, the sensor could become the basis of a vector magnetometer that determines not only the strength but also the direction of the field. To get there, one remaining challenge is how to switch the bias field and correct the error in real time. But the main barrier, the dead zone, has already been removed, and the radio no longer goes silent no matter how you turn it.