A physics group at Peking University and the Helmholtz Institute Mainz has published a room-temperature magnetometer sensitive enough, on paper, to register the brain’s electrical activity — without the liquid-helium cryogenics or magnetically shielded rooms that the two current gold-standard approaches both require. The result, reported in Science on August 6, 2026, is an instrumentation paper, not a neuroscience or clinical study: it demonstrates the sensor’s raw sensitivity and physical footprint, not a working brain scan. For labs weighing biomagnetic instrumentation, the relevant question is what infrastructure this could eventually remove, and what it has not yet been shown to do.
What was published
Wei Ji, Changhao Xu, Guofeng Qu and Dmitry Budker describe a “levitated sensor for magnetometry in ambient environment” in Science vol. 393, issue 6811, pp. 607-610 (DOI: 10.1126/science.adx1707). The device is a diamagnetically stabilized, magnetically levitated magnet magnetometer: a sensor magnet smaller than one millimetre is suspended in a vacuum chamber — roughly the size of a food-storage container, per Nature‘s coverage of the paper — between a “lifting” magnet above it and a diamagnetic material below it that pushes back against the lifting field to hold the sensor stable. A laser tracks the sensor magnet’s motion optically as it wobbles in response to a nearby magnetic field.
The authors report a sensitivity of 32 femtoteslas per square root of hertz, which they describe as adequate for a range of applications in biology, chemistry and fundamental physics and as matching the performance of superconducting quantum interference devices (SQUIDs) and atomic magnetometers — while operating at room temperature and under Earth’s ambient magnetic field, rather than inside a cryostat or a magnetically shielded enclosure. In an editor’s summary accompanying the paper, Science‘s Jelena Stajic notes that “because of its operability at ambient conditions and the potential to further increase its sensitivity, the sensor may find wide applicability.”
The standoff-distance advantage
Both incumbent high-sensitivity magnetometer types impose a physical cost that has nothing to do with raw sensitivity. SQUID arrays need liquid-helium cryoplants, and spin-exchange relaxation-free (SERF) atomic magnetometers — the type used in optically pumped magnetometer (OPM) MEG systems — need near-perfect shielding from Earth’s and the lab’s own stray magnetic fields. Per Nature‘s news coverage of the paper, both requirements also push the sensing element physically further from whatever is being measured. The levitated-magnet design reduces that sample-to-sensor standoff to a few hundred micrometres, and is reported to be orders of magnitude more sensitive than diamond-based (NV-center) magnetometers, an alternative room-temperature approach.
Ji told Nature that reaching femtotesla-scale sensitivity required an “intricate combination of noise reduction techniques” — magnetic coils and damping systems to suppress stray fields and mechanical vibration — and offered a sense of the tolerances involved: “When you’re probing the femtotesla scale, even a thin film of aluminium foil can introduce 100 femtoteslas of noise.” A femtotesla is roughly ten billion times weaker than Earth’s own magnetic field.
Why this is an infrastructure story for biomagnetism, not just a sensitivity number
Magnetoencephalography (MEG) is gated by infrastructure at least as much as by physics. Conventional SQUID-based MEG systems require a liquid-helium cryoplant and a magnetically shielded room, putting a full system’s cost and footprint out of reach for most labs outside dedicated neuroimaging centers. OPM-MEG, the newer wearable alternative, traded cryogenics for compact SERF sensors — but those still need a magnetically shielded environment to function, and typically sit at a coarser standoff from the scalp than the levitated design reported here. A sensor that operates at room temperature, tolerates Earth’s ambient field, and closes the sample-to-sensor gap to a few hundred micrometres would, if it scales, change which labs and settings can run biomagnetic measurements at all — not just make an existing measurement marginally better. The same standoff and room-temperature advantages are why Nature‘s coverage also flags interest from the fundamental-physics side, including dark-matter and axion searches that similarly benefit from getting a sensitive magnetometer physically closer to a sample.
What the paper does not yet show
It is worth being precise about what has and hasn’t been demonstrated. The Science paper is a single-sensor physics and instrumentation demonstration: it establishes the sensitivity and stabilization mechanism of one levitated-magnet magnetometer under laboratory conditions. It is not a multichannel array, has not been validated against a real physiological signal such as brain or heart activity in a living subject, and has not been compared head-to-head with an existing SQUID or OPM-MEG system on the same task. Turning a single high-sensitivity sensor into a usable MEG instrument requires scaling to a multichannel array, engineering it for use near a human head, and validating it against established systems — none of which this paper addresses. Instrumentation groups and neuroimaging labs evaluating this work should treat it as a sensitivity and design proof-of-concept, not as a characterized alternative to shared-instrumentation-funded SQUID or OPM-MEG systems already in use.
Frequently asked questions
Is this a new type of brain scanner?
Not yet, and the paper does not claim to be one. It is a demonstration of a room-temperature magnetometer whose sensitivity is, on paper, adequate to register the brain’s electrical activity. Building an actual MEG instrument from this sensor design — a multichannel array, engineered for use near a human head, validated against existing systems — is future work the authors identify as a direction, not a result already achieved.
How does 32 femtoteslas per √Hz compare to existing MEG sensors?
The authors describe it as matching the performance of SQUIDs and atomic (SERF-type) magnetometers, the two technologies used in essentially all current MEG systems. The claimed advantage is not raw sensitivity beyond either — it’s reaching comparable sensitivity without the cryogenics SQUIDs require or the magnetic shielding SERF/OPM sensors require, and at a much shorter standoff distance from the sample.
What is a diamagnetically stabilized levitated magnet magnetometer?
It’s a sub-millimetre magnet held in a fixed position by opposing magnetic forces — a “lifting” magnet pulling it up and a diamagnetic material pushing back against that pull — rather than by mechanical contact, which would introduce friction and vibration noise. A laser optically tracks tiny changes in the suspended magnet’s orientation as it responds to an external magnetic field, and that motion is the measured signal.
Does this affect research still relying on SQUID- or OPM-based MEG?
Not directly, and not yet. Existing SQUID and OPM-MEG systems remain the only clinically and experimentally validated options. This result is worth tracking for instrumentation and neuroimaging groups making multi-year equipment decisions, but it has not been built into, or benchmarked against, a working MEG system.
As of August 7, 2026, the sensitivity figure, device design and standoff-distance claims above are drawn directly from the peer-reviewed Science paper and Nature’s news coverage of it, cited below. No MEG-specific validation, multichannel array, or in-vivo comparison has been reported for this device as of this writing.
