Postselected amplification applied to atomic magnetometers

Abstract

We propose to embed

the atomic magnetometer (AM) into an optical Mach-Zehnder interferometer (MZI). We analyze the effect of amplification of the Faraday rotation angle of the probe laser light, by properly postselecting the path-information state of the laser photons when passing through the MZI. The proposed scheme provides a postselected metrological protocol of probing weak magnetic fields, having a potential to further enhance the sensitivity of the nowadays state-of-the-art optical AM.

Introduction

.— Detection of weak magnetic fields plays important role in diverse fields of fundamental science PXH21 ; PXH23 ; Rom22 and practical technologies such as in geology Ger20 , aeromagnetic investigation Fot20 , medical and military studies Bow22 ; Kaz20 , etc. Accordingly, various techniques and instruments have been developed, including such as fluxgate sensors Rip92 ; Rip22 , superconducting quantum interference devices (SQUIDs) Fag06 , Hall effect magnetic sensors Kar20 , and atomic magnetometers (AMs) Bud07 ; PXH21-b . Among these, SQUID may be the most popular and advanced one. However, it works under extremely low temperatures, requiring thus cryogenic maintenance which causes, inevitably, expensive apparatus and operational costs. In contrast, the full optical AM Bud07 , has the advantage of operating at or above room temperature and getting smaller in volume with the potential for miniaturization.

Actually, the AMs have been developed for longer than half a century Kas50 ; Deh57 ; Bloom57 ; Bloom62 . In particular, along with the advent of tunable diode lasers and progress of optical pum** techniques and producing dense atomic vapours with long ground-state spin relaxation times (in some cases $\sim$ 1 sec), the atomic magnetometers have achieved sensitivities comparable to or even surpassing that of most SQUID-based magnetometers, having thus become a leading magnetometry of ultrasensitive magnetic field measurements Bud17 . By constantly improving, the most advanced atomic magnetometers have stepped at the precision level of subfemtotesla Bud00 ; Rom02 ; Rom10 .

In recent experiments PXH21 ; PXH23 ; Rom22 , the performance of the ⁸⁷Rb AM was improved by mixing ¹²⁹Xe gas into the ⁸⁷Rb gas in the same vapor cell, with the ¹²⁹Xe nuclear spins playing a role of a pre-amplifier of weak magnetic fields. The amplified effective magnetic field acting onto the electron spins of the ⁸⁷Rb atoms can be two orders of magnitude larger than the original field. This new technique has also been discussed to probe signals from such as Goldstone bosons of new high-energy symmetries, CP-violating long-range forces, and axionic dark matter, etc.

Figure 1: (a) Schematics of principle of the atomic magnetometer (AM). Atoms (e.g. ⁸⁷Rb) in the vapor cell are polarized under the illumination of a pump laser light (circularly polarized), and a linearly polarized probe laser light is propagated along the $x$-direction. The probe light suffers an optical Faraday rotation (FR) when passing through the vapor cell. Measuring the FR angle allows to infer an unknown magnetic field. (b) Proposal to insert the AM into one of the arms of an optical Mach-Zehnder interferometer (MZI). Proper postselection for the “path-state” of the probe light can realize the effect of postselected amplification (PSA) for the FR angle. Abbreviations of device elements in the setup: LP (linear polarizer), BS (beam splitter), PBS (polarizing beam splitter), PS (phase shifter), PD (photodiode), M (mirror), and DAQ(data acquisition).

In this work, we propose a postselected metrological scheme to further enhance the probe sensitivity of the ⁸⁷Rb AM (the principle is also applicable for other AMs). The idea of postselection was recently reanalyzed in a broader context Llo20 ; Shu22 ; Yang23 , as a generalization of the weak-value amplification (WVA) AAV1 ; AAV2 . The technique of WVA has been successfully demonstrated in precision measurements Kwi08 ; How09a ; How10a ; How10b ; How13 ; Sto12 ; Lun17 ; ZLJ20 , showing profound advantages in the presence of detector’s saturation, technical noises, and some environmental disturbances. Our proposal is suggesting to insert the AM into an optical Mach-Zehnder interferometer (MZI) (see Fig. 1), which renders the photons of the probe laser light having a path-information (“which path”) degree of freedom in addition to the optical polarization degree of freedom. Then, by properly postselecting the path information from the path-and-polarization entangled state (owing to path-dependent Faraday rotation), we will show the effect of postselected amplification (PSA) of the Faraday rotation (FR) angle. This technique is anticipated to benefit the nowadays AM in probing tiny magnetic fields.

Scheme of Postselected Amplification

.— The basic principle of an AM (e.g., the ⁸⁷Rb AM) consists of two operations, as schematically shown in Fig. 1(a). First, the ⁸⁷Rb atoms in the vapor cell are polarized under the illumination of a pump laser light (circularly polarized). The pump laser couples the outmost valence electron of the ⁸⁷Rb atom to resonant transition between $5^{2}S_{1/2}$ and $5^{2}P_{1/2}$ (i.e., the D1 line at 794.98 nm). After experiencing a few dynamic processes, most or all of the atoms are transferred to the $m_{J}=+1/2$ state along the $\hat{z}$-axis, with the ⁸⁷Rb atoms in the vapor cell having a polarization vector $\textbf{P}=P_{z,0}\textbf{e}_{z}$. Second, let the polarization vector P precess around a magnetic field B, and propagate a linearly polarized probe laser light along the $\hat{x}$ direction, which is detuned from the D2 transition line at 780.2 nm between $5^{2}S_{1/2}$ and $5^{2}P_{3/2}$. The probe light will suffer an optical FR after passing through the atom vapor cell. The FR angle is proportional to the $x$-component of P. Then, measuring the FR angle allows us to infer the unknown magnetic field, which is contained in B.

Now, consider to insert the AM into one of the arms of an optical MZI, as schematically shown in Fig. 1(b). The polarized probe light is an ensemble of photons with the same polarization and the same frequency. We thus adopt the quantum mechanical description of single photon for the probe light. For the setup under consideration, the transverse spatial wavefunction of the light beam is irrelevant to the problem, therefore each single photon has two degrees of freedom: one is the polarization, which can be described using the basis states $|H\rangle$ and $|V\rangle$; another is its path information when passing through the MZI, using $|1\rangle$ and $|2\rangle$ to denote it. We assume that the probe light is prepared initially with the polarization of $|H\rangle$. After entering the MZI through the first beam splitter (BS1), the path state of the photon is $|i\rangle=(|1\rangle+|2\rangle)/\sqrt{2}$. Only propagating along path “1”, the photon’s polarization $|H\rangle$ will suffer a Faraday rotation of angle $\theta$, owing to magnetic polarization of the electron spins of the atomic medium in the vapor cell. In contrast, if the photon propagates along path “2”, its polarization remains unchanged. This path-dependent Faraday rotation can be elegantly described as

$$|\Psi(\theta)\rangle=e^{-i\theta\hat{A}\hat{\sigma}_{x}}|i\rangle|H\rangle\,.$$ (1)

Here we introduced $\hat{A}=|1\rangle\langle 1|$, and $\hat{\sigma}_{x}$ is the Pauli operator defined in the Hilbert space expanded by $\{|H\rangle,|V\rangle\}$.

To realize the effect of PSA, let us assume to measure only the light outgoing from the exit port along $x$-direction through the second beam-splitter of the MZI (BS2, see Fig. 1(b)), and to extract the FR angle by a difference-signal data analyzer (i.e., the DAQ shown in Fig. 1(b)). This one-exit-port measurement (discarding the light in the other exit port) corresponds to a postselection of the path state. More specifically, the state of postselection is $|f\rangle=(|1\rangle+e^{i\beta}|2\rangle)/\sqrt{2}$. Here the phase factor $e^{i\beta}$ is introduced by a phase shifter which is inserted in path “2”, as shown in Fig. 1(b). Via modulating $\beta$, one can realize an effect of PSA for the FR angle. Below we outline the key treatment for achieving this desired effect.

After postselection with $|f\rangle$, the polarization state of the postselected photon becomes

$$|\Phi_{f}(\theta)\rangle=\frac{1}{2\sqrt{p_{f}}}\left[(e^{-i\beta}+\cos\theta)|H\rangle-i\sin\theta|V\rangle\right]\,.$$ (2)

Mathematically, the postselection is described by $|\widetilde{\Phi}_{f}(\theta)\rangle=\langle f|\Psi(\theta)\rangle$. $|\Phi_{f}(\theta)\rangle$ is the normalized version of $|\widetilde{\Phi}_{f}(\theta)\rangle$, where the normalization factor is associated with the postselection probability $p_{f}$, which is given by $p_{f}=\langle\widetilde{\Phi}_{f}(\theta)|\widetilde{\Phi}_{f}(\theta)\rangle=(1+\cos\theta\cos\beta)/2$.

It would be instructive to get a preliminary insight for the effect of amplification through postselection, by computing first the quantum Fisher information (QFI) about $\theta$, contained in $|\Psi(\theta)\rangle$ and $|\Phi_{f}(\theta)\rangle$, respectively. For $|\Psi(\theta)\rangle$, simple calculation gives the QFI about $\theta$ Cra16 ; Rao92 , as ${\cal I}(\theta)=2$. However, remarkably, for $|\Phi_{f}(\theta)\rangle$, the QFI is obtained as

$${\cal I}^{\rm ps}(\theta)=\frac{4p_{f}-\sin^{2}\theta}{4p^{2}_{f}}\,.$$ (3)

One can check that, ${\cal I}^{\rm ps}(\theta)$ can be anomalously larger than $2$. This means that each postselected photon carries more QFI than the photon before postselection, in the entangled state $|\Psi(\theta)\rangle$. This result reflects the key feature of the recently advocated postselection filtering technique, which has been termed as postselected metrology Llo20 . The origin of the anomalous QFI has been profoundly connected with the unique quantum nature of negativeness of the Kirkwood-Dirac quasiprobability, and it was pointed out that the postselected metrology can outperform the optimal postselection-free experiment Llo20 . Indeed, in a subsequent experiment Shu22 , the metrological advantage of this technique was demonstrated, showing that it can amplify the information by two orders of magnitude, in per postselected photon.

Figure 2: (a) The QFI ${\cal I}^{\rm ps}(\theta)$, carried by per postelected photon in state $|\Phi_{f}(\theta)\rangle$, under different choices of the postselection parameter $\beta$, compared with the QFI encoded in the entangled state $|\Psi(\theta)\rangle$ given by Eq. (1) (plotted here by the black line). (b) The PSA-FR angle $\widetilde{\theta}$ versus the true FR angle $\theta$, under different choices of $\beta$. The slope of each curve in the small $\theta$ regime characterizes the amplification factor, say, $\widetilde{\theta}/\theta$. Compared with the black line (result without postselection), the PSA effect is evident.

In Fig. 2(a), for our setup, we display the QFI ${\cal I}^{\rm ps}(\theta)$ carried by per postelected photon under different choice of the postselection parameter (say, the phase shift parameter $\beta$). We see that, drastically, ${\cal I}^{\rm ps}(\theta)$ can exceed the QFI encoded in the entangled state $|\Psi(\theta)\rangle$, i.e., ${\cal I}(\theta)=2$. ${\cal I}^{\rm ps}(\theta)$ can exceed also the QFI associated with the conventional measurement (CM), shown in Fig. 1(a), which encodes the FR angle $\theta$ in the polarization state of a probing photon through the transformation $|\Phi^{\rm cm}(\theta)\rangle=e^{-i\theta\hat{\sigma}_{x}}|H\rangle$, which carries the QFI ${\cal I}^{\rm cm}(\theta)=4$. Notice that, ${\cal I}(\theta)$ is only a half of ${\cal I}^{\rm cm}(\theta)$. The reason is that, when a photon transmits through the MZI, it is affected by the parameter $\theta$ with only a half probability.

Measurement and Signal Analysis

.— Below we continue our analysis for practical measurement of the amplified signal of the FR angle. In conventional measurement, the FR angle $\theta$ is extracted from the probability $P_{V}=\sin^{2}\theta$, of the $|V\rangle$ component in the state $|\Phi^{\rm cm}(\theta)\rangle$, which is generated by the FR transformation. In the postselected measurement, based on Eq. (2), the probability of the $|V\rangle$ component in the postselected state $|\Phi_{f}(\theta)\rangle$ is given by

$$\widetilde{P}_{V}(\theta)=\sin^{2}\theta/4p_{f}\equiv\sin^{2}\widetilde{\theta}\,.$$ (4)

Here, by analogy with conventional measurement, we define an effective angle $\widetilde{\theta}$, which is termed as PSA-FR angle hereafter in this work. This result indicates that, with the decrease of $p_{f}$, $\widetilde{\theta}$ can be much larger than the true FR angle $\theta$, realizing thus a remarkable amplification effect. In Fig. 2(b), we illustrate how the FR angle $\theta$ is amplified to $\widetilde{\theta}$, by modulating the postselection parameter $\beta$. For smaller $\theta$, the amplification effect can be more prominent, for instance, it can be amplified by several orders of magnitude.

In practice, $\widetilde{P}_{V}$ can be obtained by measurement as follows. That is, the postselected photons are guided to a PBS (polarizing beam splitter), as shown in Fig. 1(b), which is set to allow the $|H\rangle$-component to transmit through it, and the $|V\rangle$-component to be reflected. Then, measure the photon flux intensities of the $|H\rangle$ and $|V\rangle$ polarizations, by two photo-detectors, which yield photo-currents $I_{H}=I_{0}\widetilde{P}_{H}$ and $I_{V}=I_{0}\widetilde{P}_{V}$, respectively. The difference-signal data analyzer, i.e., the DAQ shown in Fig. 1(b), gives the ratio

$$R=(I_{V}-I_{H})/(I_{V}+I_{H})=2\widetilde{P}_{V}-1\,,$$ (5)

from which we straightforwardly obtain the key quantity $\widetilde{P}_{V}$.

Below we further specify our consideration to connect with the real ⁸⁷Rb atomic magnetometer, which was improved in a recent experiment PXH21 , by mixing ¹²⁹Xe gas into the ⁸⁷Rb gas in the same vapor cell, and exploiting the ¹²⁹Xe nuclear spins playing role of a pre-amplifier to measure weak magnetic field. It was demonstrated in Ref. PXH21 that the amplified effective magnetic field acting onto the electron spins of the ⁸⁷Rb atoms can be two orders of magnitude larger than the original field. Thus, roughly speaking, the FR of polarization of the probe light is amplified by two orders of magnitude. This amplification technique has been discussed to be useful in probing extremely weak magnetic effects, including such as searching dark matters.

Below, using the real parameters of the ⁸⁷Rb AM, let us carry out an estimation for the time dependent FR angle $\theta(t)$. We assume that the electron spins of the ⁸⁷Rb atoms are polarized in the $z$-direction, see Fig. 1(a), by the pump laser light, and that a bias magnetic field (relatively large) is applied along the $z$-direction. Then, in the absence of any other magnetic fields, the polarization vector P does not precess around the bias magnetic field. However, if a weak magnetic field (to be estimated) is present, e.g., in the $y$-direction, the polarization vector P will precess around the total magnetic field. As a result, the component $P_{x}(t)$ of $\textbf{P}(t)$ in the $x$-direction will cause an optical rotation of the linearly polarized probe laser light after transmitting through the vapor cell (along the $x$-direction), with the FR angle proportional to $P_{x}(t)$ as PXH21

$$\theta(t)=\frac{1}{4}lr_{e}cfnD(\nu)P_{x}(t)\,.$$ (6)

In this result, $l$ is the optical path length of the probe light transmitting through the vapor cell; $r_{e}=e^{2}(4\pi\varepsilon_{0}m_{e}c^{2})^{-1}=2.8$ fm is the classical radius of electron; $c$ is the speed of light; and $f$ is the oscillator strength of electron (about 1/3 for the D1 transition and 2/3 for the D2 transition). The Lorentzian lineshape function, $D(\nu)=(\nu-\nu_{0})/{[(\nu-\nu_{0})^{2}+(\Delta\nu/2)^{2}]}$, characterizes the effect of detuning between $\nu$ (frequency of the probe light) and $\nu_{0}$ (frequency of the D2 transition). In the real experiment PXH21 , the probe light is blue-detuned by 110 GHz from the D2 transition. $\Delta\nu$ is the full-width at half-maximum (FWHM), owing to level broadening, or the dam** rate of oscillating dipole in classical treatment. The off-resonance D2 transition, or the respective classical electric polarization of the atoms, causes a change of the refractive index $n(\nu)$ of the atomic medium to the light interacting with it. If $P_{x}\neq 0$, which indicates the electron spin probabilities $p_{+1/2}\neq p_{-1/2}$ (the spin component is projected along the $x$ direction), then the left ($L$) and right ($R$) circularly polarized lights will cause different electric polarizations of the atoms (owing to the selection rule of angular momentum conservation), resulting thus in different refractive indices, $n_{+}(\nu)\neq n_{-}(\nu)$. From $\theta=\frac{\pi\nu l}{c}[n_{+}(\nu)-n_{-}(\nu)]$, one obtains the result of Eq. (6).

To carry out specific result of the FR angle, we need to obtain $P_{x}(t)$, from the Bloch equation

$$\frac{d}{dt}\textbf{P}=\gamma_{e}\textbf{B}\times\textbf{P}+R_{\rm op}(\textbf{e}_{z}-\textbf{P})-R_{\rm rel}\textbf{P}\,.$$ (7)

In this equation, P is the Bloch vector of the electron spin; $\gamma_{e}=g_{s}\mu_{B}$, with $g_{s}\approx 2$ the electron’s Land$\acute{e}$ $g$ factor and $\mu_{B}$ the Bohr magneton; $R_{\rm op}$ is the optical pum** rate and $R_{\rm rel}$ is the spin relaxation rate. Moreover, $\textbf{e}_{z}$ is the unit vector in the direction of the pum** light, and B is the total magnetic field. Following the experiment PXH21 , we assume a relatively large constant bias field $B_{z}(t)=B_{z}$ applied in the $z$-direction, and a weak alternating field, $B_{y}(t)=B_{y}\cos(2\pi\nu t)$, to be measured, in the $y$-direction. Under the condition $R_{\rm op}\gg\gamma_{e}B_{y}$, one can first obtain the approximate solution $P_{z}(t)\simeq P_{z,0}=R_{\rm op}/(R_{\rm op}+R_{\rm rel})$, then obtain $P_{x}(t)={\cal M}P_{z,0}B_{y}\cos(2\pi\nu t+\theta_{y})$. Both the amplitude modulation factor ${\cal M}$ and the phase delay factor $\theta_{y}$ are $B_{z}$ dependent, but have lengthy expressions (thus not shown here).

Figure 3: Time dependent FR signals, by comparing the true FR angle $\theta$ (dark-green curve, left coordinate) with the PSA-FR angle $\widetilde{\theta}$ (purple curves, right coordinate). The results are simulated using parameters referred to Ref. PXH21 , which are summarized in the main text of present work.

In Fig. 3, we display the numerical result of PSA of the FR angle, for a few choices of the postselection parameter $\beta$, using parameters referred to the experiment PXH21 . Some parameters not specified below Eq. (6) are: the optical path length $l=1$ cm, the density of atom numbers $n=14~{}{\rm cm}^{-3}$, and the FWHM $\Delta\nu=59$ Hz. The values of the quantities in the Bloch equation are: the gyromagnetic ratio $\gamma_{\rm e}=2\pi\times 28~{}{\rm Hz/nT}$, the pum** rate $R_{\rm op}=2800~{}{\rm sec}^{-1}$ and the relaxation rate $R_{\rm rel}=1000~{}{\rm sec}^{-1}$, the magnetci fields $B_{z}=759~{}{\rm nT}$ and $B_{y}=30\times\cos(2\pi\times 8.96t)~{}{\rm pT}$. Using these parameters, our simulation gives the result shown in Fig. 3. The effect of amplification is prominent. For example, for the postselection using $\beta=0.997\pi$, the amplification factor $\eta=\widetilde{\theta}/\theta$ is larger than 100.

Discussion and Summary

.— The precision of an AM is largely affected by shot noise of the discrete photon numbers of the probe light beam. Let us consider the quantity $R$, measured in real experiment, given by Eq. (5). We may denote the collected photon numbers in the two photo-detectors as $N_{1}=p_{f}N\cos^{2}\widetilde{\theta}$ and $N_{2}=p_{f}N\sin^{2}\widetilde{\theta}$. Then, we have $R=(N_{1}-N_{2})/(N_{1}+N_{2})$. In general, the shot noise (uncertainty of the photon numbers) is $\delta N_{1}=\sqrt{N_{1}}$ and $\delta N_{2}=\sqrt{N_{2}}$. The error of $R$, originated from $\delta N_{1}$ and $\delta N_{2}$, is determined by $\delta R=[(\frac{\partial R}{\partial N_{1}}\delta N_{1})^{2}+(\frac{\partial R}{\partial N_{2}}\delta N_{2})^{2}]^{1/2}$. Simple algebra gives $\delta R=\sin 2\widetilde{\theta}/\sqrt{p_{f}N}$. Based on Eqs. (5) and (4), we further obtain $\delta\widetilde{\theta}=1/(2\sqrt{p_{f}N})$.

We see that the signal-to-noise ratio of the PSA scheme, $\widetilde{\cal R}_{S/N}=\widetilde{\theta}/\delta\widetilde{\theta}$, remains the same as the conventional measurement, ${\cal R}_{S/N}=\theta/\delta\theta$, by noting that $\widetilde{\theta}\simeq\theta/(2\sqrt{p_{f}})$ (in the case of small $\theta$ and $\widetilde{\theta}$) and $\delta\theta\sim 1/\sqrt{N}$. This implies that the small number of postselected photons contain almost all the metrological information of all the photons before postselection. As a result, the PSA provides an approach to ensure that the detector operates under the saturation threshold even for a large number of input photons, leading thus to remarkably outperforming the conventional measurement, as analyzed in theory Lun17 and demonstrated by experiment ZLJ20 .

For the case of AM, in the presence of saturation of the two photo-detectors (PD1 and PD2 in Fig. 1), one can expect that increasing the intensity of the probe light can make the PSA scheme remarkably outperform the conventional measurement without postselection. Moreover, the PSA metrology has been proposed and demonstrated to amplify miniscule physical effects, holding the potential for enhancing measurement sensitivity and overcoming technical noises and some environmental disturbances. For the setup of AM considered in this work, we may remain the full analysis of its performance in the presence of practical imperfections and technical noises for future investigation.

To summarize, we analyzed in this work the effect of amplification of the FR angle via properly postselecting the path state, by considering to embed the optical AM into an MZI. Physically speaking, this is because the postselected photon, compared with the photon in the conventional measurement without postselection, contains much more information about the FR angle, while the FR angle is related with (proportional to) the weak magnetic field to be estimated. Therefore, the scheme proposed in this work provides a postselected amplification metrology of probing tiny magnetic fields, holding the advantage of further enhancing the sensitivity of the nowadays optical AM.

Acknowledgements.

— This work was supported by the NNSF of China (Nos. 11675016, 11974011 & 61905174).