A Working Map of Quantum Sensing

In February 2022, a group at the University of Birmingham published a paper in Nature describing a gravity measurement taken outdoors, on a road, using two clouds of cold rubidium atoms held about a meter apart. The instrument located a buried utility tunnel roughly two meters across by comparing how hard gravity pulled on the upper cloud against the lower one. No excavation, no prior map.

The same two words, quantum sensing, also describe the cesium clock disciplining timing across a mobile network, the superconducting loop reading magnetic fields off a patient’s scalp, and a laboratory proposal to build radar out of entangled microwave photons. Three of those are things an organization can purchase. One is a paper.

That spread is why the term confuses people arriving from an engineering background. Quantum sensing names a method rather than a product category, and vendors, funding agencies, and trade press each stretch it in a different direction. What follows is the map we use when we teach the subject: a way to sort several dozen device types into categories that fit in your head, and a way to judge, for any one of them, how far it is from something you could write into a specification.

Three questions that pin down any quantum sensor

Every device in this field can be located with three questions.

What does it measure? The measurand: time and frequency, acceleration, rotation, gravity, magnetic field, electric field, photons, temperature, force, pressure.

What does the measuring? The platform: free-falling neutral atoms, trapped ions, a defect in a diamond lattice, a superconducting circuit, a beam of light in a non-classical state.

What keeps it indoors? The constraint: a vacuum chamber, cryogenic cooling, a magnetically shielded room, a rack of stabilized lasers, or simply size, weight, and power draw.

Most published taxonomies sort by the second question, including the one on our sister site PostQuantum. That’s the physicist’s cut, and it organizes the literature well, because the papers cluster by platform. We sort by the first, because an engineer arrives with a measurement problem rather than a preference for rubidium.

Some readers will object that sorting by measurand hides the physics, and it does. We accept the trade because the same platform then shows up in four different categories below, and that repetition is the lesson. An atom interferometer is a gravimeter, an accelerometer, and a gyroscope, depending mostly on which way you point it and what you hold fixed.

What the quantum part actually buys

A conventional sensor averages the behavior of an enormous number of particles and reads out a bulk property. A quantum sensor prepares and reads individual quantum systems, or ensembles held in a controlled quantum state, and extracts the measurement from how that state changes. Four resources recur across the whole field, and almost every device below uses at least one of them.

A reference that nature supplies. Energy levels in an atom are set by physical constants, so every cesium-133 atom has the same transition frequency. A strain gauge drifts and needs periodic recalibration against a standard. A cesium atom is the standard. This is where the absolute measurement claim comes from, and it is the single most commercially proven advantage in the field.

Phase accumulated in superposition. A quantum system can be prepared so that it occupies two states at once, a condition called superposition. Split a system into two paths, let the paths experience slightly different conditions, then recombine them, and the accumulated phase difference shows up as an interference pattern. That is interferometry. Do it with light and you measure length. Do it with atoms and you measure acceleration and rotation, because a matter wave accumulates phase in proportion to the acceleration it experiences and to the square of the time it spends in flight.

Spin as a probe. An electron or nuclear spin precesses at a rate set by the local magnetic field. Prepare the spin, let it precess for a known interval, read the angle, and you have a field measurement traceable to a frequency. Spins in a heated vapour cell and spins in a diamond lattice both work this way.

Non-classical light. Send N independent photons at a measurement and the uncertainty in the result falls only as the square root of N. That floor is called shot noise, or the standard quantum limit, and it comes from photons arriving independently rather than from any defect in the detector. Squeezing redistributes quantum uncertainty between two conjugate properties of a light field so that the one you care about gets quieter. Entanglement, meaning correlations between particles that cannot be reproduced by any independent preparation, can in principle improve the scaling from one over root N to one over N.

The fourth resource is the one that dominates press coverage and the one with the fewest fielded examples. The clearest exception is gravitational-wave detection, where squeezed light has been injected into the LIGO interferometers since 2019 and buys a few decibels of noise reduction at high frequency, which translates directly into more detected events per observing run. Outside that, most working quantum sensors run on the first three.

The map by measurand

Time and frequency

An atomic clock counts oscillations of a specific transition between two energy levels, and the SI second is defined as 9,192,631,770 of them in cesium-133. Because the reference is the atom, the clock does not age the way a quartz oscillator ages. Cesium fountain standards such as NIST-F2 would neither gain nor lose a second over roughly 300 million years.

Optical clocks move the reference from microwave frequencies near 9 GHz to optical frequencies in the hundreds of terahertz, using strontium or ytterbium atoms in an optical lattice or a single trapped aluminum ion. Finer ticks mean finer division of the second, and the best reported systematic uncertainties now sit below one part in 10 to the 18th.

Here is where a clock stops being only a clock. Gravitational redshift shifts a clock’s frequency by roughly 1.1 parts in 10 to the 16th for every meter of elevation. A clock resolving one part in 10 to the 18th therefore resolves height differences of about a centimeter, and geodesists have started treating clock networks as gravity instruments. The categories in this map are permeable, and this is the first place you see it.

Deployment is well ahead of the rest of the field. Chip-scale atomic clocks fit in under twenty cubic centimeters and draw around a tenth of a watt, which puts them in vehicles, radios, and network equipment. The commercial drivers are unglamorous and real: holdover timing when satellite navigation signals are jammed or blocked, phase synchronization across mobile base stations, and regulatory timestamping. Under MiFID II, high-frequency trading systems in Europe must keep business clocks within 100 microseconds of UTC, and something in the rack has to guarantee that.

Acceleration, rotation, and gravity

Cool a cloud of atoms to near absolute zero, drop it, and hit it with a sequence of laser pulses. The first pulse puts each atom into a superposition of two momentum states, so the cloud follows two slightly different trajectories at once. A later pulse brings them back together. The interference pattern in the recombined cloud encodes the acceleration the atoms felt while separated.

Point that apparatus down and you have an absolute gravimeter reading the local value of g. Arrange the pulse sequence so the two paths enclose an area and you have a gyroscope reading rotation. Nothing about the atoms changes.

Gradiometry is the trick that got the technique outdoors, and it explains the Birmingham result. Near the Earth’s surface, gravity weakens with height by about three parts in 10 to the 7th per meter. That is the signal. Ground vibration from traffic and machinery shakes an instrument platform by amounts thousands of times larger. That is the noise, and it defeats a single gravimeter on a roadside.

A gradiometer runs two interferometers simultaneously from the same laser pulses, one above the other. Platform vibration accelerates both atom clouds identically, so it cancels when you subtract one phase from the other. The gravity gradient does not cancel, because it is by definition the difference between the two locations. The subtraction is the instrument. Everything else is engineering around it.

Applications follow from what gravity can see that other methods cannot. Voids, tunnels, and buried infrastructure show up as mass deficits regardless of soil conductivity, which is where ground-penetrating radar struggles. Groundwater storage, magma movement, and ice mass show up as slow changes in g. And an atom accelerometer paired with an atom gyroscope forms an inertial unit whose drift is set by atomic physics rather than by mechanical wear, which is the case for satellite-denied navigation on submarines, ships, and aircraft.

The constraints are severe. The atoms need ultra-high vacuum, the lasers need stabilizing, the measurement is cyclic rather than continuous, so sample rates are typically a few hertz to a few tens of hertz, and tilt has to be controlled to sub-milliradian levels. Field-usable systems exist and are sold. They are not small.

Magnetic fields

Set the scale first, because the numbers do most of the explaining. The Earth’s field runs around 50 microtesla. The magnetic field of a beating human heart, measured at the chest, is around 50 picotesla. The field from neural currents in the brain, measured at the scalp, runs from roughly 100 femtotesla to a few picotesla. That is nine orders of magnitude between the ambient field and the signal, and it explains why magnetically shielded rooms exist.

Three quantum technologies compete in this space.

Superconducting quantum interference devices (SQUIDs) have been commercial since the 1970s. A superconducting loop interrupted by one or two Josephson junctions turns magnetic flux threading the loop into an oscillating voltage, and the period of that oscillation is one flux quantum, a fixed constant. Noise floors of a few femtotesla per root hertz are routine. The price is liquid helium, a cryostat, and the vacuum gap the cryostat imposes between the sensor and whatever you are measuring.

Optically pumped magnetometers (OPMs) use a heated cell of alkali vapour, usually rubidium or cesium. One laser aligns the atomic spins, the field makes them precess, and a second laser reads the precession as a rotation of polarization. Operated in the spin-exchange relaxation-free (SERF) regime, in near-zero residual field, these reached 0.54 femtotesla per root hertz in a 2003 Nature result and have been improving since. No cryogens are required, and a sensor head can be the size of a sugar cube.

That size difference changed a clinical technology. Magnetoencephalography (MEG) maps brain activity from its magnetic signature, and for thirty years it meant a fixed helmet of SQUIDs with a cryostat wall holding the sensors roughly two centimeters off the scalp. A magnetic dipole field falls off steeply with distance, so that gap costs signal. In a 2018 Nature paper, a Nottingham group built a wearable MEG helmet from OPMs sitting a few millimeters from the head, reported four to five times more signal, and recorded usable data while the subject moved, drank tea, and bounced a ball. The physics didn’t improve; the standoff distance did.

Nitrogen-vacancy (NV) centers are point defects in diamond, where a nitrogen atom sits beside a vacancy in the carbon lattice. The defect holds an electron spin that can be prepared with green light, manipulated with microwaves, and read out from its fluorescence, all at room temperature. Sensitivity for a bulk ensemble lands in the picotesla range, well short of a SQUID, but a single NV center is a probe a few nanometers across. That buys spatial resolution nothing else offers: mapping current flow inside a packaged integrated circuit, imaging magnetic domains in a material, or reading a field profile with a scanning tip.

The selection rule between the three is not sensitivity alone. It is sensitivity at the standoff distance and spatial scale your problem actually has.

Electric and radio-frequency fields

Excite an atom’s outer electron to a very high energy level, a Rydberg state, and the electron orbits far from the nucleus. The atom becomes physically large and exquisitely responsive to electric fields.

A vapour cell full of Rydberg atoms works as a radio receiver. An incoming radio-frequency field shifts and splits the atomic energy levels, and a probe laser passing through the cell reads that splitting as a change in transmission. The receiver is a glass cell and two lasers, with no antenna, no mixer, and no amplifier chain in front of the measurement.

Self-calibration is the property that makes this more than a curiosity. The splitting of the atomic line is proportional to the field strength multiplied by the atomic transition dipole moment, and that dipole moment is calculable from atomic structure. So the instrument converts a measured optical frequency into volts per meter using a computed constant, and the result is traceable to the SI second rather than to a calibration house. National metrology institutes care about this a great deal, because everything else measuring field strength has to be calibrated against something.

By choosing which Rydberg states to address, one apparatus covers frequencies from the low megahertz to hundreds of gigahertz, which no single antenna does. Demonstrations have received ordinary broadcast and Wi-Fi signals directly. The open problems are laser stability, sensitivity relative to a well-designed conventional receiver at any single frequency, and the awkward fact that a glass vapour cell is not a rugged part.

Superconducting circuits work the other end of the same territory. Josephson parametric amplifiers add close to the minimum noise quantum mechanics permits, which is why they read out superconducting qubits and why axion dark matter searches use them to listen for microwave signals that would otherwise be buried. Kinetic inductance detectors absorb single submillimeter photons and change their resonant frequency, and large arrays of them sit in submillimeter astronomy instruments today. Both need dilution refrigerators, which fixes them to installations rather than platforms.

Photons, images, and range

This category carries the most hype relative to what has shipped.

What is genuinely in service is single-photon detection and time-correlated counting. Modern lidar systems and depth cameras count individual photon arrivals and time them to picoseconds, which extends range at a fixed eye-safe laser power and helps discriminate returns from fog and foliage. Whether that counts as quantum sensing is a definitional argument. It works, and it ships.

What is real in the laboratory and moving slowly outward is quantum-enhanced imaging. Ghost imaging, demonstrated with entangled photon pairs by Pittman, Shih and colleagues in 1995, forms an image using a detector that never sees spatial detail, by correlating its clicks with a camera watching the twin beam that never touched the object. Squeezed light has improved signal-to-noise in optical microscopy and particle tracking in living samples, where you cannot simply turn up the illumination without cooking the specimen. That constraint, a hard ceiling on how many photons the sample tolerates, is exactly where non-classical light earns its cost.

What does not exist is quantum radar as described in defense trade press. The theory is sound and specific: Lloyd’s 2008 quantum illumination proposal, and the Gaussian-state analysis by Tan and colleagues shortly after, showed an advantage in error probability over the best classical strategy at equal transmitted energy, in a regime of high background noise and low signal. The advantage survives even though the entanglement itself does not survive the round trip, which is the counter-intuitive part. But the bound is asymptotic, it applies at very low signal power, it requires storing the idler photon coherently until the return arrives, and translating any of it into a microwave system with useful range remains unsolved. Treat any procurement conversation about quantum radar accordingly.

Temperature, force, and pressure

A short category, held together by the fact that each of these devices reads a quantum system whose resonance shifts with the quantity of interest.

NV centers in diamond shift their spin resonance with temperature as well as with magnetic field, which turns a nanodiamond into a thermometer small enough to sit inside a living cell. A 2013 result demonstrated millikelvin-level resolution at that scale, which no thermocouple approaches, because a thermocouple large enough to read is large enough to perturb.

Optomechanical sensors couple light to a mechanical element, usually a membrane or a suspended mirror, and read its displacement through a cavity. Cooled toward its quantum ground state, such an element measures forces at the attonewton scale. LIGO’s 40-kilogram mirrors are the extreme case of the same principle, and nanogram oscillators used to weigh single particles are the other end of it.

Pressure is the quiet success. A cold-atom vacuum standard infers pressure from the rate at which background gas knocks trapped atoms out of a trap, with the loss rate calculable from atomic collision physics. That gives a primary vacuum standard, meaning one that does not need calibration against another gauge, in a pressure range where existing gauges drift.

How close each category is to the field

Category	Representative device	Readiness	Main constraint
Time and frequency	Chip-scale and rack atomic clocks	In service, decades of production	Cost per unit at optical performance
Magnetic fields	SQUID arrays, OPM sensor heads	In service, clinical and geophysical	Magnetic shielding, ambient field rejection
Photon counting	Single-photon lidar, depth sensors	In service	Definitional argument over what counts
Gravity and inertial	Absolute gravimeters, gradiometers	Field trials, early commercial	Vacuum, vibration, size and power
NV imaging	Scanning and wide-field NV microscopes	Early commercial, instrument market	Diamond material quality, averaging time
Radio-frequency	Rydberg vapour-cell receivers	Late laboratory, metrology pilots	Laser stability, ruggedization
Quantum illumination	Quantum radar concepts	Theory and bench demonstrations	No path yet to a fielded microwave system

Read the readiness column as the honest answer to a common question. Quantum sensing is not five years away. Parts of it shipped before most of the audience was born, other parts are being trialed on trucks and in hospitals right now, and a small, loud fraction is a research program.

Five questions to ask a vendor

These are the questions we teach people to ask, and they separate a specification from a brochure.

1. Is the measurement absolute or relative? An absolute instrument reads the quantity against a physical constant and needs no calibration against a reference device. A relative one drifts and needs a tie-in. This single distinction explains most of the price difference between quantum and conventional instruments in the same category.
2. What is the noise floor per root hertz, and measured under what conditions? A femtotesla figure inside a three-layer shielded room and a femtotesla figure in an active hospital are different products. Ask what shielding, what temperature, and what averaging time produced the number.
3. What is the bandwidth and the duty cycle? Atom interferometers are cyclic. Between measurement shots they are blind. If the platform is moving, that dead time is a specification, not a footnote.
4. What does the environment cost to provide? Cryogen supply, vibration isolation, a shielded enclosure, laser maintenance intervals, and power draw are all part of the installed price and none of them appear in the sensitivity figure.
5. What is the conventional alternative doing this year? MEMS gyroscopes, fluxgate magnetometers, and ground-penetrating radar are all improving. A quantum instrument has to beat what the incumbent will be in the year it deploys, not what it was when the research program started. Where the quantum device wins outright, it usually wins on something the incumbent cannot do at all, such as absolute accuracy or nanoscale resolution, rather than on a factor of two.

One further item belongs on a procurement checklist rather than a physics one. Several jurisdictions tightened export controls on quantum hardware during 2024, and gravimeters and high-sensitivity magnetometers are the sensing categories most often named. Check classification before you plan a cross-border deployment.

Where to start if you are learning this

The field looks large because the applications are scattered across geophysics, medicine, defense, and metrology. The underlying material is much narrower than the application list suggests, and we sequence it in four steps in our own courses.

Start with measurement and noise, not with quantum mechanics. Noise spectral density expressed per root hertz, Allan deviation, the difference between accuracy and stability, and where the standard quantum limit comes from. Almost every specification argument in this field is an argument about one of those five things, and none of them require a Hamiltonian.

Then learn the two-level system properly: Rabi oscillation, Ramsey interferometry, coherence time, and why a longer coherence time buys sensitivity. Every atomic clock, every OPM, and every NV sensor is a variation on that one piece of physics.

Third, learn interferometric phase, first for light and then for matter waves. Once the phase accumulation is clear, gravimeters, gyroscopes, and gradiometers stop being three topics.

Fourth, and only then, take on the platform specifics: laser cooling and trapping, superconductivity and the Josephson effect, and solid-state defect spins. For a technical reference to work alongside this, the review titled “Quantum sensing” by Degen, Reinhard and Cappellaro in Reviews of Modern Physics 89, 035002, published in 2017, remains the standard entry point, and it assumes exactly the four steps above.

For a deeper technical treatment of the individual device families and their use cases, our sister site PostQuantum.com covers the same territory organized by platform, which is a useful second pass once the measurand map is in place. If you are weighing quantum sensing as a direction rather than a topic, QuantumCareers.com maps the roles that actually hire for it, and they skew toward metrology, laser engineering, and systems integration rather than toward theory.

Quantum Academy’s sensing curriculum follows the sequence above, with worked measurement problems in place of derivations, and it is built for people who will specify, evaluate, or integrate these instruments rather than design them. You can see the current programs and enrollment options at quantumacademy.com/.