The Magnetometer That Switches A Laser Off

A laser running at more than 50 milliwatts goes completely dark when a microwave field comes onto resonance at 2.87 gigahertz. Not dimmer — dark. That is the measurement at the centre of a paper published in Nature Communications on 22 October by Florian Schall, Jan Jeske and colleagues at the Fraunhofer Institute for Applied Solid State Physics in Freiburg, with partners in New York, Japan and Australia, and it is the first experimental demonstration that a magnetic field can shift the threshold of a laser containing nitrogen-vacancy diamond.

The number that follows from it is 100 per cent optical contrast. Conventional magnetometry with nitrogen-vacancy centres gets a few per cent. That gap is not an incremental improvement; it changes which term dominates the sensitivity. Fraunhofer IAF puts the gain at up to a factor of 590 in the ratio of sensitivity limit to dynamic range — the figure of merit that has constrained this whole class of sensor — measured against conventional fluorescence-based nitrogen-vacancy sensors and vapour-cell magnetometers.

A nitrogen-vacancy centre is a defect in diamond: a nitrogen atom substituted for a carbon, sitting next to a missing carbon. In its negatively charged state it has a spin-triplet ground state whose sublevels are split by 2.87 gigahertz at zero field, and the splitting moves with an applied magnetic field at 28 gigahertz per tesla. Illuminate the defect with green light and it fluoresces red; drive the spin transition with resonant microwaves and the fluorescence drops. That drop, read as a function of microwave frequency, is optically detected magnetic resonance, and the position of the dip tells you the field.

It is a beautiful measurement and it has a structural problem. The fluorescence drop is small — typically a few per cent of a signal that is itself weak, because the fluorescence is emitted into 4π and most of it never reaches a detector. The sensitivity of the measurement scales with that contrast, so the whole field has spent twenty years fighting for it: better collection optics, higher nitrogen-vacancy densities, pulsed protocols, lock-in detection, flux concentrators.

Most of those tricks cost dynamic range. A pulsed protocol that narrows the resonance improves the slope you measure on, but a narrower resonance means a smaller field excursion before the signal leaves the linear region. A flux concentrator multiplies the field at the diamond and divides the range you can measure. The practical result is the trade-off Schall describes: the most sensitive nitrogen-vacancy magnetometers work over a few microtesla, which is a problem when the Earth's own field is around 50 microtesla and you would like to take the instrument outside the shielded room.

The idea behind laser threshold magnetometry is to stop collecting photons that the defect happens to emit and instead make the defect decide whether a laser oscillates at all.

The nitrogen-vacancy centre has a second optical transition that gets much less attention than the red one. Its singlet system — the intersystem-crossing pathway that makes optical spin polarisation work in the first place — has a transition at 1042 nanometres. The population in those singlet states depends on the spin state of the triplet ground state, and therefore on the magnetic field and on whether the microwaves are resonant. So at 1042 nanometres, a nitrogen-vacancy ensemble is an absorber whose absorption is a function of the magnetic field.

Put that absorber inside a laser cavity that oscillates at 1042 nanometres and you have made the field into an intracavity loss. Near threshold, output power is a steep and highly nonlinear function of net round-trip gain: a small change in loss moves the threshold, and if the pump sits just above the old threshold and just below the new one, the laser turns off entirely. The sensor is no longer counting the few per cent of photons the spin state changed. It is using the spin state as a trip switch on a 50-milliwatt beam.

This has been the plan at Freiburg for a long time. In June 2022 the same group reported the first measurement of magnetic-field-dependent stimulated emission from nitrogen-vacancy centres, with a contrast of about 33 per cent — a record at the time, and still a long way from a laser that switches. Jeske described the goal then in the same terms he uses now.

The hard part of this scheme is that diamond is a poor host for a laser cavity and 1042 nanometres is an awkward wavelength to force a gain medium onto. The Freiburg solution, described in the preprint version of the work posted in September 2025, is a two-media cavity: a semiconductor provides the gain, the diamond provides the field-dependent loss, and neither has to do the other's job.

The gain element is a membrane external-cavity surface-emitting laser — a thin semiconductor membrane on a GaAs substrate containing around eleven InGaAs quantum wells, optically pumped at 808 nanometres by a fibre-coupled diode focused to roughly a 150-micrometre spot. Left alone, its gain peaks near 1075 nanometres. A three-millimetre quartz Lyot filter set at Brewster's angle pulls the oscillation down to 1042 nanometres, onto the nitrogen-vacancy singlet line. That Lyot plate is doing the whole experiment's alignment of wavelengths, and it is the reason a membrane laser was chosen: the gain is broad enough to be dragged 33 nanometres off its peak and still oscillate.

The cavity is about 98.5 millimetres long, closed at one end by the membrane laser's own back mirror and at the other by a concave output coupler of 95.1 per cent reflectivity and 100-millimetre radius of curvature. The mode is about 140 micrometres across on the gain chip. The diamond sits 28 millimetres from the chip, where the mode has expanded to about 280 micrometres, and is tilted to Brewster's angle so that p-polarised light passes with low loss — a conventional trick, and a necessary one when the intracavity element is a 0.3-millimetre plate of material with a refractive index of 2.4.

The diamond itself is a (100)-oriented type 1b high-pressure, high-temperature plate about 0.306 millimetres thick, annealed at 1800 °C in hydrogen, irradiated with 2 MeV electrons to create vacancies, then annealed again for two hours at 1000 °C so the vacancies migrate to the nitrogen. The resulting negatively charged nitrogen-vacancy concentration is 1.91 parts per million. A separate 532-nanometre laser pumps the defects, mode-matched into the cavity through a 125-millimetre lens, and a microwave antenna drives the spin transition near 2.87 gigahertz with a permanent magnet supplying a bias field along the (100) direction.

With the green pump at 5 watts, the lasing threshold sits at 1.53 watts of 808-nanometre pump when the microwaves are off resonance, at 2 gigahertz, and moves to 1.82 watts when they are brought onto the nitrogen-vacancy resonance. The slope efficiency falls from 22.0 per cent to 18.7 per cent across the same change. Without any green pumping at all, the threshold is 1.23 watts and the slope efficiency 27.4 per cent — which is the honest measure of how much the diamond costs the laser even before the spins are addressed.

Park the pump at 1.82 watts, exactly at the resonant threshold, and the picture is the headline: off resonance the laser emits more than 50 milliwatts, on resonance it emits nothing. That is the 100 per cent contrast. A second configuration, at 1.44 watts of pump in the vector arrangement where all four nitrogen-vacancy orientations are resolved, gives about 100 per cent contrast on all eight resonances — but with a baseline of 2 milliwatts rather than 50. A third, with a 16-milliwatt baseline, gives about 97 per cent. Contrast and output power trade against each other, and the published figures sit at different points on that trade.

The sensitivity claims follow the same discipline. The paper reports a photon-shot-noise-limited sensitivity of 670 femtotesla per root hertz at the optimal point of the resonance, over a dynamic range of ±280 microtesla set by the width of the optically detected magnetic resonance line divided by the electron gyromagnetic ratio. Fraunhofer's own announcement states a figure below 400 femtotesla per root hertz for each of the three vector components, again assuming photon shot noise dominates. Both numbers are projections of what the measured signal would support if shot noise were the limiting term, obtained from plain continuous-wave resonance measurements, without a lock-in amplifier and without flux concentrators. They are not a measured noise floor, and the paper does not present one.

That distinction matters more here than usual. The whole argument is that contrast approaching unity improves sensitivity faster than linearly — a scaling that had been predicted and that the Freiburg data now demonstrate experimentally — and a superlinear gain in the shot-noise-limited figure is only worth having if the real instrument can be made shot-noise-limited. A 98.5-millimetre free-space cavity with 5 watts of green light dumped into a diamond plate has thermal lensing, pointing drift and intensity noise to contend with before shot noise is anywhere near the floor.

The dynamic range of ±280 microtesla is set by the resonance linewidth, which means it is inherited from ordinary nitrogen-vacancy physics rather than created by the laser. What the cavity does is let the team keep that range while taking the contrast — and therefore the sensitivity — to a place that previously required narrowing the line. The factor of 590 is a ratio, and reading it as a 590-fold sensitivity improvement would be wrong.

Then there is the apparatus. This is a bench instrument with a Lyot filter, a Brewster-angled diamond, a separate green pump laser and a microwave antenna, occupying the better part of a metre of optical table. The applications Fraunhofer names — magnetocardiography, magnetomyography, magnetoencephalography, magnetic navigation, anomaly detection for resource exploration and security — all want something that fits on a person or in a vehicle. The work sits inside NeuroQ, a flagship project funded by Germany's Federal Ministry of Research, Technology and Space, whose stated aim is a quantum sensor that reads brain activity well enough to drive an exoskeleton through a brain-computer interface. The gap between a 98.5-millimetre external cavity on a table and a sensor in a headband is the next decade of the work, not a detail.

The co-authors give some sense of where the pieces would come from: Twenty-One Semiconductors on the semiconductor gain side, the College of Staten Island at CUNY and the National Institutes for Quantum Science and Technology in Japan on diamond materials, and RMIT University in Australia, where Andrew Greentree has worked on the theory of laser threshold magnetometry since the idea was first proposed.

The quantum-sensing framing undersells what has been built. Strip the application away and this is an intracavity absorption spectrometer in which the absorber's cross-section is controlled by a radio-frequency field, operated at the point of maximum nonlinearity in the laser's transfer function. Intracavity laser absorption spectroscopy has been doing exactly this trick — using threshold to amplify a tiny loss — since the 1970s, and the enhancement factors it reaches are enormous. What is new is a loss channel that can be switched coherently and that reports a vector quantity.

The engineering problems are laser problems, not quantum problems. The diamond is an intracavity plate of index 2.4 held at Brewster's angle, and inserting it costs the cavity 0.3 watts of threshold before a single spin is addressed; reducing that insertion loss, at the surface quality a laser cavity demands and without introducing stress birefringence, is a polishing and coating brief. The Lyot filter is a 3-millimetre quartz plate; a volume Bragg grating or a coated etalon would hold 1042 nanometres more tightly against thermal drift. The 95.1 per cent output coupler is a compromise between extracting power and keeping the cavity sensitive to a small loss change, and that compromise is a coating specification. The step from this experiment to an instrument runs through optics and coatings, and the people who can make that step are not, for the most part, quantum physicists.

Until 22 October, laser threshold magnetometry was a proposal with partial evidence: stimulated emission had been shown to depend on the magnetic field, and contrast had been pushed to a third. What was missing was the threshold itself — the demonstration that the field can decide whether the device lases. That is now on the record, with a laser going from 50 milliwatts to zero.

What it costs is honesty about the rest. The sensitivity figures are shot-noise projections from continuous-wave data, the instrument is a bench cavity, and the dynamic range comes from the diamond rather than from the laser. None of that diminishes the result; it locates it. A measurement principle that had been argued about for more than a decade has been shown to work, and the remaining work has moved from physics into optical engineering — which is where it can be bought, specified and improved.