Probe-assisted Depopulation Pumping in Low-pressure Alkali-metal Vapor Cells for Magnetometry

For precision atomic magnetometry, inert buffer gas is included in alkali-metal vapor cells to significantly broaden hyperfine transitions, which facilitates optical pumping and reduces diffusive relaxation, while also providing non-radiative excited state quenching. We show low-buffer gas pressure (below 50 Torr) alkali vapor cells with resolved hyperfine manifolds can also yield high-performance magnetometers. For high polarization in $^{87}$Rb, we optically pump $F=2$ states with narrow linewidth $σ_+$ light, while tuning a probe beam to depopulate $F=1$ states ($Δν= 6.8$ GHz from $F=2$). The probe tuning then also provides $F=2$ detection with high optical rotation and low probe broadening; we demonstrate top-bottom gradiometry, within a single 25 Torr, 0.5 cc cell, that yields an Earth’s field free-precession magnetometer sensitivity of 18 fT/$\sqrt{\text{Hz}}$ with a 1 kHz bandwidth, as well as RF magnetometer sensitivity of 12 fT/$\sqrt{\text{Hz}}$ in a small band about 110 kHz.

💡 Research Summary

The paper presents a novel “probe‑assisted depopulation pumping” technique that enables high‑performance atomic magnetometry using low‑pressure alkali‑metal vapor cells (≤ 50 torr buffer gas). Conventional high‑pressure cells broaden the D₁ hyperfine transitions, allowing broadband lasers to pump both F = 1 and F = 2 ground manifolds simultaneously, but the broadening reduces the maximum optical rotation and introduces excess probe‑induced relaxation. In contrast, the authors work with a 0.47 cc ⁸⁷Rb cell containing only 25 torr of an Ar:N₂ (3:2) mixture, where the hyperfine manifolds remain spectrally resolved. A narrow‑linewidth σ⁺ pump laser is tuned to the F = 2 transitions and drives the atoms into the m_F = +2 edge state, achieving near‑unity polarization of the F = 2 manifold. However, without additional action the remaining F = 1 population would cause spin‑exchange relaxation and limit the coherence time.

To solve this, a linearly polarized (σ⁰) probe beam is tuned exactly to the F = 1 → excited‑state transitions, which are separated from the F = 2 transitions by the 6.8 GHz hyperfine splitting. The probe therefore continuously depopulates the F = 1 ground state, optically pumping atoms into the already‑polarized F = 2 manifold. Because the probe is far detuned from the F = 2 transitions, it provides a large optical rotation signal with minimal additional broadening (probe‑induced Γ_P). The combined action yields three key benefits: (1) maximal edge‑state polarization suppresses spin‑exchange relaxation; (2) the probe’s detuning ensures high‑SNR optical rotation while keeping probe‑induced decoherence negligible; (3) the overall relaxation rate is limited only by wall collisions (Γ_W) and the modest probe broadening, giving a transverse coherence time T₂ ≈ 0.33 ms.

The authors develop a quantitative model using Voigt profiles for absorption and dispersion, incorporating Doppler (Γ_G ≈ 0.57 GHz) and pressure broadening (Γ_L ≈ 18 MHz·p/760 torr). At the optimal pressure (~20 torr) the total decoherence rate is ≈ 2π·30 kHz, matching the experimentally observed T₂. Using the Cramér‑Rao lower bound (δB ∝ T^{-3/2}) they identify the optimal single‑shot duration as ≈ 2 T₂, which in practice translates to a 0.4 ms detection window and a 2 kHz repetition rate, yielding a 1 kHz effective bandwidth.

Experimentally, the cell is heated to 90 °C, and both pump and probe lasers (795 nm D₁ line, 0.5 MHz linewidth) deliver 5–10 mW each. The pump is σ⁺‑polarized via an AOM and quarter‑wave plate; the probe is linearly polarized, passes through the cell, and its rotation is measured with a balanced polarimeter and split photodiodes for top‑bottom gradiometry. By scanning pump and probe frequencies, three regimes are identified: “assisted” (⋆) where probe depopulation of F = 1 enhances F = 2 polarization, “mid” (○) with moderate rotation, and “suppressed” (×) where probe counteracts pumping. In the assisted regime the optical rotation signal is maximized while the probe remains detuned by 6.8 GHz from the F = 2 line, ensuring low probe‑induced broadening.

For scalar magnetometry, after a pump pulse a π/2 tipping pulse places the spins transverse to a 44 µT (≈ 3.1 kHz) bias field. The free‑precession signal is recorded for 0.4 ms, fitted to a decaying sinusoid, and the Larmor frequency is converted to magnetic field. Despite a 200 fT·Hz⁻¹ᐟ² current‑supply noise, the top‑bottom differential measurement achieves a sensitivity of 17.8 ± 0.3 fT·Hz⁻¹ᐟ² with a 1 kHz bandwidth. This represents a ten‑fold increase in bandwidth compared with prior high‑pressure cells (≈ 90 Hz) while maintaining comparable sensitivity, owing to the reduced probe‑induced decoherence and the long T₂ afforded by low pressure.

The same cell is then operated as an RF magnetometer. Raising the temperature to 130 °C increases vapor density, and continuous probe‑assisted pumping is maintained. The bias field is set so that the Larmor frequency lies near 110 kHz; a weak RF coil drives Rabi oscillations in the rotating‑wave approximation. The detected RF response yields a magnetic‑field sensitivity of 12.1 ± 0.4 fT·Hz⁻¹ᐟ² and a resonance linewidth of ≈ 3 kHz, approaching the spin‑projection noise limit (δB < 10 fT·Hz⁻¹ᐟ²) for the given atom number and probe volume.

In summary, the work demonstrates that low‑pressure alkali vapor cells, when combined with a carefully tuned depopulating probe, can achieve high polarization, long coherence, and large optical rotation without the drawbacks of high‑pressure broadening. The technique is applicable to other alkali species (K, Cs) and opens a pathway to compact, low‑power, high‑bandwidth atomic magnetometers suitable for applications such as portable MEG, magnetic navigation, and neutron‑EDM searches, where gradient tolerance and operation in Earth‑scale fields are essential.