SERF-OPMs: Spin-Exchange Relaxation-Free Sensors

• SERF-OPMs are atomic-vapor quantum sensors that operate in a near-zero magnetic field regime, where rapid spin-exchange collisions preserve coherence for magnetic detection.
• They employ optical pumping and various readout methods, including probe-based optical rotation and single-beam detection, to accurately extract weak magnetic signals.
• Recent advances extend SERF-OPMs to multimodal applications by introducing dual-axis sensing, noise engineering, and adaptations for operation beyond standard low-field conditions.

Searching arXiv for recent and foundational SERF-OPM papers to ground the article. Spin-Exchange Relaxation-Free optically pumped magnetometers are atomic-vapor quantum sensors in which optical pumping, spin-exchange collisions, and other relaxation channels drive a dense alkali ensemble into a non-equilibrium steady state suitable for magnetic-field detection. Their defining operational regime is the SERF regime, where the spin-exchange rate is much larger than the characteristic magnetic precession rate, so spin-exchange collisions redistribute spin population without destroying net polarization, thereby preserving coherence useful for sensing (Sousa et al., 10 Apr 2026). In practice, SERF-OPMs are implemented with optically pumped alkali vapors such as $^{87}$Rb, K, or Cs, typically at high atomic density and low magnetic field, and are read out through absorption or optical rotation. Recent work has broadened the topic from low-field sensitivity alone to thermodynamic state preparation, noise engineering, multichannel timing behavior, dual-axis and gradiometric readout, structured-light control of sub-ensembles, and extensions toward high-field or geomagnetic operation via related suppression mechanisms (Mouloudakis et al., 2024).

1. Regime definition and dynamical basis

SERF-OPMs operate by using a pump laser to drive an atomic vapor into a Non-Equilibrium Steady State. In dense atomic vapors, the combined action of spin-exchange and spin-destruction collisions, together with the pump rate and polarization, determines the properties of this state (Sousa et al., 10 Apr 2026). The operative distinction is between spin-exchange collisions, denoted $\Gamma_{\mathrm{SE}}$, and spin-destruction collisions, denoted $\Gamma_{\mathrm{SD}}$. In the SERF regime, $\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$, and spin-exchange collisions redistribute spin population without destroying net polarization, whereas spin-destruction processes, including wall and buffer-gas-related channels, lead to true relaxation (Sousa et al., 10 Apr 2026).

A representative density-matrix model for alkali atoms in a vapor cell is written as

$$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$

where $R_{\mathrm{op}}$ is the optical pumping rate and $\mathbf{s}$ is the photon spin vector (Sousa et al., 10 Apr 2026). This expression formalizes the interplay of unitary spin evolution, optical pumping, spin exchange, and spin destruction, and it is the basis for recent thermodynamic and metrological treatments of SERF state preparation.

The same core SERF logic appears across distinct experimental realizations. A potassium SERF magnetometer for zero- and ultralow-field NMR uses a high-buffer-gas, high-density vapor cell so that the spin-exchange collision rate far exceeds the Larmor precession frequency, suppressing first-order spin-exchange relaxation at zero field (Hong et al., 2024). A microfabricated $^{87}$Rb magnetometer based on optically modulated zero-field dispersive resonance likewise operates near zero field with high density so that spin-exchange broadening is strongly suppressed (Jimenez-Martinez et al., 2014). Hybrid optical pumping realizations using $^{39}$K-$^{85}$Rb-$\Gamma_{\mathrm{SE}}$0He or $\Gamma_{\mathrm{SE}}$1Cs-$\Gamma_{\mathrm{SE}}$2Rb-$\Gamma_{\mathrm{SE}}$3He are also explicitly described as SERF atomic magnetometers requiring very low magnetic field and sufficiently high alkali density (Liu et al., 2017).

This shared framework has an important corollary: conventional SERF-OPMs are intrinsically low-field devices. One 2024 comparison states that traditional SERF OPMs suppress spin-exchange relaxation by operating at near-zero magnetic field and high alkali density, and therefore require magnetic shielding or active compensation because finite field revives spin-exchange relaxation (Schönau et al., 2024). This low-field condition underlies both their exceptional sensitivity and many of their engineering constraints.

2. Optical pumping, steady-state preparation, and thermodynamic interpretation

Recent analysis has recast SERF-OPM initialization as a non-equilibrium thermodynamic process. The system evolves from a thermal equilibrium state of maximal entropy to a less entropic, polarized Non-Equilibrium Steady State, with irreversibility quantified by the relative entropy

$\Gamma_{\mathrm{SE}}$4

and with the entropy production rate peaking during the transient and vanishing at steady state (Sousa et al., 10 Apr 2026). Within this description, useful energy is quantified by ergotropy,

$\Gamma_{\mathrm{SE}}$5

and the spin-polarization efficiency is defined by

$\Gamma_{\mathrm{SE}}$6

High $\Gamma_{\mathrm{SE}}$7 and high $\Gamma_{\mathrm{SE}}$8 lead to high $\Gamma_{\mathrm{SE}}$9, while small cell radius or high wall relaxation lower $\Gamma_{\mathrm{SD}}$0; the SERF regime is identified as the regime of optimally achievable $\Gamma_{\mathrm{SD}}$1 (Sousa et al., 10 Apr 2026).

The same work relates state preparation directly to magnetic metrology. For estimation of a magnetic field component $\Gamma_{\mathrm{SD}}$2, with coupling $\Gamma_{\mathrm{SD}}$3, the Quantum Fisher Information is

$\Gamma_{\mathrm{SD}}$4

and the quantum Cramér-Rao bound gives

$\Gamma_{\mathrm{SD}}$5

Higher preparation efficiency $\Gamma_{\mathrm{SD}}$6 yields higher QFI for transverse directions, while $\Gamma_{\mathrm{SD}}$7 when the state is diagonal in the pump basis (Sousa et al., 10 Apr 2026). The paper further reports strict linear proportionality between transverse-direction QFI and relative entropy above thermal, $\Gamma_{\mathrm{SD}}$8, linking thermodynamic distance from equilibrium to the fundamental sensitivity bound (Sousa et al., 10 Apr 2026).

This thermodynamic formulation is consistent with more traditional SERF design guidance. Hybrid optical pumping theory states that sensitivity varies with buffer gas, quench gas, pumping rate, external magnetic field, cell effective radius, measurement volume, cell temperature, and measurement time, and provides the shot-noise formula

$\Gamma_{\mathrm{SD}}$9

with

$\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$0

together with a practical refinement for hybrid systems (Liu et al., 2017). The thermodynamic treatment does not replace these relations; a plausible implication is that it reorganizes them into a resource-theoretic language centered on entropy production, ergotropy, and state-preparation efficiency.

3. Readout architectures and signal extraction

SERF-OPM readout is commonly performed with a probe laser that measures optical rotation after passage through the polarized vapor, but several papers in the data set emphasize alternative architectures intended to reduce perturbation or technical noise. In pump-probe systems, a second probe beam introduces extra perturbation and additional optical complexity. A 2015 study demonstrated that the electron-spin polarization in a Cs SERF magnetometer can instead be extracted from the transmitted intensity of the circularly polarized pump beam itself (Fang et al., 2015). The key relations are

$\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$1

so that

$\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$2

The method was experimentally validated in a SERF magnetometer, operates in a silent mode, and provides real-time observation (Fang et al., 2015). The same study added a diffusion term to the magnetic-field-response model to account for discrepancies between theory and experiment, underscoring the importance of atomic transport in compact cells (Fang et al., 2015).

Single-beam readout has also been pursued at the device level. A compact Cs-Ne SERF magnetometer uses a single beam in a 3 mm diameter by 3 mm length cylindrical vapor cell, compatible with silicon-glass bonding micro-machining, and introduces a differential laser power noise suppression method in which a reference beam bypasses the cell and is subtracted electronically from the measurement channel (Chen et al., 2021). The paper states that the large background detection offset and laser amplitude noise are the main noise source of a single-beam absorption SERF magnetometer, and reports a factor of 2 improvement in power noise suppression together with a sensitivity of $\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$3 at 30 Hz (Chen et al., 2021).

Probe-based optical rotation remains central in many high-performance systems. The potassium SERF magnetometer developed for ZULF NMR uses a linearly polarized probe on the K D2 transition and a photo-elastic modulator to shift detection to 50 kHz, with top and bottom photodiodes separated by 1 mm to permit gradiometric readout (Hong et al., 2024). After amplitude and phase calibration, the gradiometric channel reached a magnetic noise floor of $\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$4 at 20–30 Hz, a 7-fold enhancement relative to the single channel (Hong et al., 2024). This configuration illustrates a standard SERF pattern: strong optical pumping, dispersive optical readout, and common-mode noise rejection by multi-channel subtraction.

More unconventional detection has been introduced via inverse weak measurement. In that approach, the spatial profile of the probe beam serves as the pointer, the polarization state acts as the quantum system, and a Sagnac interferometer introduces a small spatial shift between $\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$5 and $\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$6 (Cao et al., 17 Aug 2025). With post-selection,

$\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$7

and the pointer displacement is

$\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$8

which in the inverse weak measurement regime becomes

$\Gamma_{\mathrm{SE}} \gg \Gamma_{\mathrm{SD}}$9

The paper reports a magnetic-field sensitivity of $$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$0 and states that the detected signals depend only on the internal degrees of freedom of the probe laser, making the system robust against fluctuations in laser power; Allan deviation analysis showed one to two orders of magnitude improvement in stability relative to conventional detection (Cao et al., 17 Aug 2025). This suggests a distinction between sensitivity optimization and stability optimization: the IWM scheme is not presented as the lowest-noise SERF readout in the data set, but as a detection method with specific immunity to probe-power fluctuations.

4. Noise, sensitivity limits, and bandwidth

SERF-OPM performance is shaped by quantum noise, technical noise, and geometry-dependent relaxation. A 2024 treatment of spin projection noise develops a mean-field density-matrix model that includes the degree of spin polarization, intra- and interhyperfine correlations, decoherence, atom-light coupling, and spin dynamics in the spin-noise spectra (Mouloudakis et al., 2024). Especially in the SERF regime, that work reports a new SERF feature: reduction of spin-projection noise at the spin precession frequency as a consequence of strongly correlated hyperfine spins that attenuate and redistribute SPN when properly probed (Mouloudakis et al., 2024). The measured noise spectrum is written as

$$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$1

and the sensitivity as

$$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$2

The formal significance is that the quantum noise floor itself becomes a function of probe detuning and hyperfine correlations rather than a fixed stretched-state estimate.

Shot-noise and relaxation-based sensitivity modeling also appears in hybrid SERF theory. For $$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$3K-$$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$4Rb-$$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$5He and $$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$6Cs-$$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$7Rb-$$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$8He systems, the fundamental sensitivity is analyzed as a function of buffer gas, quench gas, pump rate, field, cell radius, volume, temperature, and measurement time (Liu et al., 2017). Under one stated set of conditions, the paper reports $$\frac{d\rho}{dt} = \frac{1}{i\hbar}[H_0,\rho] + R_{\mathrm{op}}[\phi(1+2\mathbf{s}\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SE}}[\phi(1+4\langle\mathbf{S}\rangle\cdot\mathbf{S})-\rho] + \Gamma_{\mathrm{SD}}[\phi-\rho],$$9 for the $R_{\mathrm{op}}$0K-$R_{\mathrm{op}}$1Rb-$R_{\mathrm{op}}$2He magnetometer and $R_{\mathrm{op}}$3 for the $R_{\mathrm{op}}$4Cs-$R_{\mathrm{op}}$5Rb-$R_{\mathrm{op}}$6He magnetometer, with an estimated optimized lower bound superior to $R_{\mathrm{op}}$7 (Liu et al., 2017). These are theoretical projections, not reported device-level measurements, and they illustrate how SERF performance analysis often separates projected quantum limits from achieved technical noise floors.

Experimentally, the data block spans sensitivities from the low-picotesla scale to femtotesla and attotesla projections. The optically modulated zero-field $R_{\mathrm{op}}$8Rb microfabricated SERF magnetometer demonstrated $R_{\mathrm{op}}$9, limited by photon shot noise and weak pumping in that implementation (Jimenez-Martinez et al., 2014). The compact Cs-Ne single-beam device reached $\mathbf{s}$0 at 30 Hz (Chen et al., 2021). The potassium ZULF-NMR sensor achieved $\mathbf{s}$1 in single-channel mode and $\mathbf{s}$2 in calibrated gradiometric mode (Hong et al., 2024). A SERF magnetometer discussed for CASPEr demonstrated sensitivity $\mathbf{s}$3 with an effective sensing volume of $\mathbf{s}$4, while citing literature values of $\mathbf{s}$5 in gradiometer configuration and a quantum noise limit of $\mathbf{s}$6 (Wang et al., 2017).

Bandwidth is a recurrent limitation. The CASPEr study states that the natural bandwidth of a SERF-OPM is a few Hz and can be increased up to 200 Hz or higher with a bias field along the pump direction, though sensitivity decreases above that (Wang et al., 2017). By contrast, an Earth-field vector OPM based on light narrowing rather than standard SERF operation reported a bandwidth $\mathbf{s}$7 and a white noise floor below $\mathbf{s}$8 between 100 Hz and 600 Hz (Schönau et al., 2024). This contrast is central: SERF-OPMs achieve their highest sensitivity under conditions that tend to narrow their operational field range and bandwidth, whereas alternative suppression mechanisms trade part of that operating paradigm for robustness in large fields.

5. Axiality, vector sensing, arrays, and spatial multiplexing

A common misconception is that SERF-OPMs are intrinsically scalar. In fact, much of the recent literature is concerned with vector reconstruction, dual-axis readout, and spatial multiplexing. A dual-axis $\mathbf{s}$9-pulse SERF magnetometer applies a small DC bias field and a comb of magnetic DC $^{87}$0 pulses along the pump direction so that $^{87}$1 and $^{87}$2 are upconverted into synchronous AC signals at the pulse repetition frequency with orthogonal phases (Zhivun et al., 2018). In the spherical-basis description,

$^{87}$3

and under resonance $^{87}$4,

$^{87}$5

The paper reports technical noise floors of $^{87}$6 (x) and $^{87}$7 (y) at 0.01 Hz, compared with $^{87}$8 for a single-axis DC SERF on the same apparatus, and minima of $^{87}$9 (x) and $^{39}$0 (y) at 10 Hz (Zhivun et al., 2018). The purpose is explicit suppression of $^{39}$1 probe noise at low frequencies, particularly for biomagnetism.

Related dual-axis concepts appear in alkali-metal–noble-gas comagnetometers operated with pulsed optical pumping. In that case the measurement occurs in the dark after a 6 ms pump pulse and during a 20 ms dark period, and the transient probe signal is fitted as

$^{39}$2

with $^{39}$3 and $^{39}$4 encoding two transverse axes (Wang et al., 2024). The paper describes simultaneous two-axis sensitivity, elimination of pump-induced light shifts during measurement, suppression of pump-power fluctuations, strong mitigation of $^{39}$5 noise, and magnetic-field suppression factors as high as $^{39}$6 (Wang et al., 2024). This is not a pure SERF-OPM in the narrow sense, but it draws directly on SERF magnetometer sensitivity while showing how noble-gas coupling changes the noise and vector-response structure.

Spatial multiplexing within a single SERF ensemble has been developed with micrometer-scale polarization control. In “Sub-Ensemble Isolation in SERF Magnetometry Enabled by Micrometer-Scale Polarization Control” (Liang et al., 29 Jun 2025), a space-variant polarization metasurface creates position-dependent pumping polarization in a miniaturized $^{39}$7Rb cell, producing distinct sub-ensembles with antiparallel or zero net polarization. The polarization and light-angular-momentum fields are expanded in Laplacian eigenmodes,

$^{39}$8

with dynamics

$^{39}$9

and

$^{85}$0

The paper reports an average scale factor of $^{85}$1, crosstalk between adjacent channels up to 32 dB using a fictitious magnetic field of 3.5 nT at 30 Hz, compared with 20 dB without sub-ensemble isolation under the same experimental condition (Liang et al., 29 Jun 2025). This directly challenges the conventional understanding that rapid exchange forces uniform time evolution of the whole SERF ensemble.

Optical modulation is another array-relevant architectural choice. A microfabricated $^{85}$2Rb SERF magnetometer with zero-field dispersive resonance uses a light-shift beam to generate an optically modulated fictitious magnetic field rather than an RF magnetic field, precisely because RF modulation can induce cross-talk among adjacent sensors or perturb the source under measurement (Jimenez-Martinez et al., 2014). This all-optical modulation strategy is specifically presented as useful for array-based magnetometers.

6. Materials, cell engineering, timing, and extensions beyond the standard low-field paradigm

SERF-OPM performance is tightly coupled to cell construction, buffer gas selection, laser systems, and timing behavior in multichannel platforms. A compact Cs-Ne SERF magnetometer identifies 3 Amagats of neon as optimal for a 3 mm cylindrical cell and states that this is the first demonstration of a Cs-Ne SERF magnetometer (Chen et al., 2021). Neon is argued to reduce wall relaxation without the leakage disadvantages of helium or the large spin-destruction cross-section of nitrogen in microfabricated cells (Chen et al., 2021). The wall-relaxation expression used for the cylindrical geometry is

$^{85}$3

The result illustrates how SERF engineering increasingly depends on microfabrication-compatible gas choices rather than only on bulk vapor-cell heuristics.

Laser infrastructure is another enabling layer. A pulsed auto-locking external-cavity diode laser system with interference-filter stabilization, tapered waveguide amplification to 2 W CW, and AOM pulse generation down to 20 ns was developed specifically for high-precision magnetometry and described as relevant to portable rubidium SERF magnetometers with sensitivities of $^{85}$4 (Pouliot et al., 2018). Numerical simulations in that work state that, in high buffer/quencher gas conditions, the Rb D1 transition at 795 nm is preferred because it pumps atoms into a dark stretched state with very high polarization, whereas D2 is less optimal (Pouliot et al., 2018). This aligns with multiple SERF implementations in the data set that use D1 pumping under high-pressure conditions.

Commercial multichannel systems introduce a different engineering issue: timing. A 2025 cross-platform study of four commercial SERF-OPM systems reports frequency-dependent delays of 1–10 ms, group delays of 1–15 ms, intra-channel spreads up to $^{85}$5 ms, and settling times of 2–55 ms (Elzenheimer et al., 26 Sep 2025). In the 20–140 Hz band, the time delay deviation between channels is in the sub-millisecond range in all systems, which the paper states is sufficient for magnetoencephalography source localization (Elzenheimer et al., 26 Sep 2025). At the same time, longer settling times in some platforms limit performance for rapid stimulation protocols. This moves SERF-OPM evaluation beyond sensitivity alone toward waveform fidelity, synchronization, and application-specific latency budgets.

Several papers in the data set address extensions or alternatives when the standard low-field SERF condition is impractical. One route is pulsed parametric resonance at near-Earth-scale fields. A sequence of AC-coupled $^{85}$6 pulses repeated at the Larmor frequency suppresses spin-exchange relaxation by a factor of the duty cycle, with

$^{85}$7

and experimentally enables resonant transverse pumping in magnetic fields as high as 0.1 G (Korver et al., 2013). Another route is light narrowing in a strong bias field. A microfabricated Cs vector magnetometer immersed in a homogeneous bias field of about $^{85}$8 uses the light narrowing effect to suppress spin-exchange relaxation even at large field amplitude, demonstrating a white noise floor below $^{85}$9 between 100 Hz and 600 Hz and bandwidth $\Gamma_{\mathrm{SE}}$00 kHz (Schönau et al., 2024). A third route is the proposal that atoms with nuclear spin $\Gamma_{\mathrm{SE}}$01 can operate in the SERF regime even at high magnetic field, leading to a projected fundamental sensitivity of about $\Gamma_{\mathrm{SE}}$02 at geomagnetic fields in a dual-species potassium–atomic-hydrogen magnetometer (Dikopoltsev et al., 2022).

These approaches do not abolish the conceptual centrality of SERF; rather, they show that the suppression of spin-exchange relaxation can be generalized beyond the original near-zero-field alkali-only architecture. A plausible implication is that “SERF-OPM” now denotes both a specific low-field operating regime and a broader design objective: preserving coherence in dense spin-exchanging media while retaining practical readout.

7. Applications, limitations, and conceptual boundaries

SERF-OPMs are used where extreme magnetic sensitivity is required under non-cryogenic conditions. Biomagnetism is a major theme. The dual-axis $\Gamma_{\mathrm{SE}}$03-pulse SERF magnetometer is explicitly motivated by fetal magnetocardiography and magnetoencephalography, where low-frequency biomagnetic signals make $\Gamma_{\mathrm{SE}}$04 noise suppression essential (Zhivun et al., 2018). Commercial multichannel systems are evaluated with respect to stimulation-evoked responses, brain-computer interfaces, and closed-loop neuromodulation, emphasizing the importance of timing characterization for multichannel biomagnetic sensing (Elzenheimer et al., 26 Sep 2025). High-spatial-resolution biomagnetic mapping is also a stated application of sub-ensemble isolation via metasurface-induced polarization patterning (Liang et al., 29 Jun 2025).

NMR and fundamental-physics detection are another important application class. Potassium SERF magnetometers have been integrated with zero- and ultralow-field NMR, with custom vacuum-chamber design used to place the sample close to the hot cell while protecting it from thermal load (Hong et al., 2024). In CASPEr, a SERF magnetometer with $\Gamma_{\mathrm{SE}}$05 sensitivity and $\Gamma_{\mathrm{SE}}$06 effective sensing volume is proposed for NMR detection, with superconducting flux transformers considered to suppress the large leading magnetic field and expand dynamic range (Wang et al., 2017). SERF magnetometers have also been proposed for searches for exotic spin-dependent interactions by moving polarized or unpolarized test masses next to the vapor cell; the electron spins inside the cell respond to effective fields in the same way they respond to a magnetic field, allowing projected improvement in coupling constraints below 0.01 m interaction range (Chu et al., 2016).

At the same time, the data set makes the limitations of standalone SERF sensors explicit in nonmagnetic-coupling searches. Numerical comparison with noble-gas–alkali-metal comagnetometers shows that a self-compensating comagnetometer can reduce sensitivity to low-frequency magnetic fields without loss of sensitivity to nonmagnetic couplings, and that its response to neutron pseudo-magnetic coupling is about five orders of magnitude stronger, and to proton pseudo-magnetic coupling about three orders of magnitude stronger, than that of the SERF magnetometer (Padniuk et al., 2021). This is not a refutation of SERF-OPM sensitivity; it is a statement about selectivity. SERF-OPMs are highly sensitive to real magnetic fields, and that very sensitivity can be a disadvantage when the experimental target is a pseudo-magnetic coupling obscured by magnetic noise.

The conceptual boundary of SERF itself is also under active discussion. A 2025 cesium study reports anomalous suppression of spin-exchange relaxation in alignment signals under linearly polarized pumping in ultra-weak magnetic fields, with measured linewidths down to about 6.4 Hz, comparable in manifestation to the SERF effect known for orientation signals (Petrenko et al., 20 Jan 2025). The paper notes that conventional angular-momentum-conservation arguments for SERF orientation do not obviously guarantee alignment preservation, yet certain alignment components show no measurable contribution from spin-exchange broadening in zero or perpendicular fields (Petrenko et al., 20 Jan 2025). This suggests that the usual identification of SERF exclusively with orientation dynamics may be incomplete.

Taken together, the literature portrays SERF-OPMs as a mature but still rapidly diversifying sensor class. Their classical definition remains the low-field, high-density alkali vapor regime in which spin-exchange collisions cease to dominate transverse decoherence. Around that core, current work extends the framework toward thermodynamic optimization, hyperfine-noise engineering, dual-axis and gradiometric operation, silent or single-beam readout, micrometer-scale spatial channelization, multichannel timing control, and alternative relaxation-suppression strategies for operation outside the standard near-zero-field envelope (Sousa et al., 10 Apr 2026).