Short answer: Atomic magnetometers measure magnetic fields using the spins of alkali atoms in a vapour cell, pumped and read out with light. Published experiments report sensitivities in the femtotesla range, and one 2024 chip-scale design reports 20 fT per root hertz. They are laboratory demonstrations aimed at areas such as biomagnetic measurement.
How the measurement works
An overview on arXiv (physics/0611246, 2006) describes some of the most sensitive magnetic field measurements as using optical pumping and the interaction of resonant light with atoms. It says atomic magnetometers have reached sensitivities rivalling, and in some cases surpassing, most SQUID-based devices, and that the most sensitive atomic optical magnetometer is the spin-exchange-relaxation-free (SERF) type, with demonstrated sensitivity better than 10^-15 tesla per root hertz and projected fundamental limits below 10^-17. Quantum mechanics sets limits, including projection noise from the finite number of polarised atoms. The paper also says SERF devices offer the possibility of spatially resolved measurement with millimetre resolution.
Radio-frequency sensing
A 2005 Physical Review Letters paper describes an alkali-metal magnetometer for weak radio-frequency fields. High sensitivity comes from tuning the Zeeman resonance of the atoms to the radio frequency and partly suppressing spin-exchange collisions in the vapour. The authors demonstrate a sensitivity of 2 fT per root hertz at 99 kHz with a resonance width of 400 Hz. They also derive a simple expression for the fundamental sensitivity limit and say about 0.01 fT per root hertz could be reached in a practical system with a 200 cubic centimetre measurement volume. That is a projection, not a measurement. The experimental cell was a 3.8 cm glass cell held at 190 degrees Celsius.
Towards chip-scale devices
A 2024 Nature paper (Microsystems and Nanoengineering) reports a miniaturised single-beam hybrid optical pumping SERF magnetometer built on a microfabricated vapour cell. In it, optically thin caesium atoms are spin-polarised while the dense rubidium atoms determine the experimental results. The authors report enhanced signal strength and narrower resonance linewidth, and about a five-fold suppression of optical noise using differential detection. They find the cell temperature markedly affects performance, and report an optimal magnetic sensitivity of 20 fT per root hertz. They present it as a basis for chip-scale integration of sensitive magnetometers, for possible future magnetocardiography and magnetoencephalography applications.
What the numbers mean
The three papers use different designs, field frequencies and conditions, so their sensitivities cannot be ranked against one another. Hot vapour cells, and in SERF devices, magnetic shielding, are part of the setups described. None of the papers reports a product.
Why the sensitivity figures are hard to compare
Magnetometer sensitivity is quoted in tesla per root hertz, and it depends on the frequency band, the cell, the temperature and how the noise is measured. The 2005 result is at a high radio frequency, the 2024 result is from a miniaturised cell, and the overview quotes demonstrated and projected limits for SERF devices in general. A lower number in one paper does not mean a better device in another paper's setting. The 2024 authors also report that cell temperature markedly affects performance, which is one reason such devices are characterised carefully before any application.
FAQ
What is a SERF magnetometer?
A spin-exchange-relaxation-free magnetometer, which the 2006 overview calls the most sensitive atomic optical magnetometer.
How sensitive are they?
Demonstrated sensitivity better than 10^-15 tesla per root hertz in the overview, 2 fT per root hertz at 99 kHz in the 2005 radio-frequency device, and 20 fT per root hertz for the 2024 chip-scale design.
Do they rival SQUIDs?
The 2006 overview says atomic magnetometers have rivalled and in some cases surpassed most SQUID-based devices.
What is a chip-scale magnetometer for?
The 2024 authors point to magnetocardiography and magnetoencephalography as possible future uses.
Is the 0.01 fT figure measured?
No. It is a projection in the 2005 paper for a practical system.
All field notes