Diamond-defect magnetometry gets ready for the field

When diamond is bombarded with nitrogen ions accelerated to a few thousand electron volts, its formerly pristine crystal lattice is imbued with point-like defects. Most prized and intriguing of those defects is the negatively charged nitrogen–vacancy (NV) center, shown in figure 1, which consists of a nitrogen atom adjacent to a vacant lattice site.

Figure 1.

The unpaired electrons in the dangling bonds surrounding the vacancy together behave like a spin-1 atom, with a trio of electron-spin quantum states that can be externally manipulated. Shielded from their surroundings by the diamond lattice, NV centers have a long spin coherence time that makes them appealing as building blocks for a quantum computer. And because the spin states shift in energy in response to an external magnetic field, the defects also serve as tiny magnetometers that can pick up magnetic signals in single living cells or picoliter-sized samples. (See Physics Today, May 2018, page 21 .)

Of course, the need for magnetic field measurements isn’t confined to the microscopic world. Archaeologists, for example, use magnetic mapping as a nondestructive tool for surveying sites; it can reveal the presence of not just magnetic metals but also brick, burned soil, and other materials (see Physics Today, March 2014, page 24 ). And all-magnetic navigation, based on maps of anomalies in Earth’s magnetic field, is valuable in military settings where GPS is vulnerable to disruption. ¹ Plenty of technologies exist already for detecting macroscopic fields. But sensors based on ensembles of NV centers promise some distinct advantages, including small size, low power consumption, and the ability to detect all three components of the field vector with a single device.

Now Danielle Braje and her colleagues at MIT Lincoln Laboratory—including lead authors Hannah Clevenson, Linh Pham, and Carson Teale—have taken an important step toward creating an NV magnetometer that’s suitable for use in real-world conditions. ² Their detection scheme is robust to changes in temperature and other parameters that would otherwise necessitate frequent recalibration. And it provides fast and continuous measurements of the field vector for magnitudes up to at least 11 mT. For comparison, the magnitude of Earth’s magnetic field at ground level is orders of magnitude smaller, ranging from 25 µT to 65 µT, and previous real-time NV magnetometry schemes are limited to dynamic ranges of just a few microteslas.

Microwave resonances

An NV center’s ground-state sublevels are characterized by spin quantum numbers m_s of −1, 0, and +1. At room temperature and under zero magnetic field, the −1 and +1 levels are degenerate with each other and separated from the 0 level by the energy of a 2.87 GHz microwave photon. Raising the temperature lowers that energy gap, and applying a magnetic field introduces an energy splitting between the formerly degenerate levels.

A major advantage of NV centers over other atom-like crystal defects is that although the m_s splitting energies lie in the microwave regime, they can be measured optically. When a visible-wavelength laser promotes the defects from the m_s = 0 level of the ground state into an electronic excited state, relaxation back to the ground state is accompanied by detectable fluorescence. When the +1 and −1 levels are optically excited, on the other hand, the NV centers are prone to nonradiative relaxation. As a result, if a microwave field applied in tandem with the laser causes the measured fluorescence to dip, it must be of just the right frequency to drive the NV centers out of the m_s = 0 level and into the m_s = +1 or −1 level. (For more on the physics of NV-center measurements, see the article by Lilian Childress, Ronald Walsworth, and Mikhail Lukin, Physics Today, October 2014, page 38 .)

The engineering challenge, then, is how best to find and keep track of those resonant microwave frequencies. For an ensemble of NV centers, there can be as many as eight resonances to measure. In the tetrahedral crystal lattice of a single diamond crystal, NV centers can have four possible spatial orientations. Each responds differently to an external magnetic field, and each has separate resonances corresponding to promotion to the +1 and −1 levels.

One way to measure them all is to sweep a microwave field over the full range of possible frequencies and look for the dips in the fluorescence spectrum. ³ But that takes at least several seconds for a single measurement—a long time from the perspective of navigating a fast-moving vehicle. Furthermore, the microwave resonant frequencies depend on both temperature and magnetic field. If either of those parameters changes over the course of a measurement, interpretation of the fluorescence spectrum is greatly complicated.

Another approach, once the approximate location of a resonance is known, is to keep track of it over time by rapidly oscillating the frequency of an applied microwave field over a narrow range and watching for changes in the fluorescence. ⁴ That has the benefit of speed, and one can even track multiple resonances at the same time by using a different oscillation period for each of them. ⁵ However, the technique is limited to measuring magnetic field changes of just a few microteslas, and furthermore, it requires the resonance line shapes to be constant over time. Fluctuations in the optical laser power can cause changes in the line shape and are thus a significant source of error.

Frequency following

Braje and colleagues’ design improves on the second technique by incorporating closed-loop feedback circuits to allow the oscillating microwave frequencies to follow the resonances as they move. Naturally, that enhancement allows operation over a much larger range of magnetic field. Moreover, because of the difference in the way the resonant frequencies are derived from the measured signal, it frees the measurement from the assumption of constant laser power and constant line shape. Attenuating the laser power by even a factor of 20 doesn’t change the ultimate field measurement; a normal 5% drift in laser intensity has virtually no effect at all.

For their demonstration, the researchers used a piece of diamond 2 mm on each side and containing an estimated trillion NV centers of all four orientations. Using the trick of different frequency-oscillation periods, they probed the resonances two at a time. They always measured the +1 and −1 resonances of the same orientation simultaneously, which is important for removing the effect of temperature drift. And they cycled through the four orientations with a dwell time of 0.1 s on each.

To tell the orientations apart, they applied a strong bias field of 7 mT. As shown in figure 2a, that caused the eight resonance frequencies to separate over a range of several hundred megahertz. A typical field to be measured, up to tens or hundreds of microteslas, imposes relatively small additional frequency shifts; the resonances remain distinct, and in principle a stronger bias field would allow an even larger dynamic range.

Figure 2.

Figure 2b shows how the NV magnetometer can measure both the magnitude and the direction of an external field. In response to 10 µT fields applied sequentially in the z, y, and x directions, the resonances exhibit distinct patterns of frequency shifts. The response is fast and stable over time, with little noise. The measurement is overdetermined—four NV orientations are used to reconstruct a three-dimensional field vector—and that redundancy can be helpful in correcting for lingering sources of instrumental error.