High-Field Optical Cesium Magnetometer for Magnetic Resonance Imaging

We present a novel high-field optical quantum magnetometer based on saturated absorption spectroscopy on the extreme angular-momentum states of the cesium ${\mathrm{D}}_2$ line. With key features including continuous readout, high sampling rate, and sensitivity and accuracy in the ppm range, it represents a competitive alternative to conventional techniques for measuring magnetic fields of several teslas. The prototype has four small separate field probes, and all support electronics and optics are fitted into a single 19-inch rack to make it compact, mobile, and robust. The field probes are fiber coupled and made from nonmetallic components, allowing them to be easily and safely positioned inside a 7 T MRI scanner. We demonstrate the capabilities of this magnetometer by measuring two different MRI sequences, and we show how it can be used to reveal imperfections in the gradient coil system, to highlight the potential applications in medical MRI. We propose the term EXAAQ (EXtreme Angular-momentum Absorption-spectroscopy Quantum) magnetometry, for this novel method.

Popular Summary

High magnetic fields are today a crucial element in many branches of science and technology. When measuring magnetic fields of several teslas, four conventional techniques can be employed, each with their characteristic pros and cons. The choice of the sensor depends on the application. Some applications, such as fusion reactors, particle accelerators, and MRI scanners, are pushing the limits of the conventional high-field magnetometry techniques, and in some cases the ideal sensor simply does not exist.

We have developed a new kind of optical quantum magnetometer, challenging the decades long status quo in high-field magnetometry. The technology is based on tracking a magnetic-field-dependent optical resonance in cesium atoms. That is, a particular frequency of infrared light, absorbed by an atomic cesium gas, which changes in response to different magnetic-field strengths. Even though this is the first demonstration of the technology, we demonstrate an accuracy rivaling the established paradigms and powerful features including high bandwidth and low electromagnetic interference.

We show that the sensor can be used to clearly detect imperfections in a 7-T MRI scanner coil system, highlighting the potential use in MRI image improvement. We note that the sensor also seems like an attractive tool for use in fusion reactors and particle accelerators. Future work will improve on the current prototype and investigate applications in MRI.

Figure 1

Optical setup. Probe laser light is modulated by ${\mathrm{EOM}}_{\mathrm{FM}}$, split by a free-space beam splitter, and then further split by a $1\times 4$ fiber splitter, for a total of five different paths. Each path is again modulated by an EOM, before the probe light is sent to the reference probe at 0 T and probes 1–4 in the MRI scanner. Colored cables represent optical fibers, and black cables represent the analog feedbacks realizing the accurate measurement scheme.

Figure 2

Exploded view of the optical-probe assembly. The probe light comes from the blue fiber, a polarizing beam splitter and a quarter-wave plate prepares ${\sigma }_{+}$ polarization before the light enters the cesium vapor cell, it is attenuated by an optical filter and reflected back through the vapor cell, is reflected off the polarizing beam splitter, and exits through the orange fiber. High-power laser light from the yellow fiber heats the vapor cell, to increase the cesium atomic density. The assembled probe dimensions are $90\times 33\times 10{\mathrm{mm}}^3$.

Figure 3

The sideband-spectroscopy frequency-shift measurement scheme is illustrated. The probe-laser frequency ${\nu }_{\mathrm{L}}$ is stabilized 97.5 GHz above the relevant transition, as measured at 0 T, by spectroscopy with the lower fifth sideband of ${\nu }_0=19.5\mathrm{GHz}$ using the reference probe. For each of the probes $i=\{1,2,3,4\}$, a first upper sideband of about 0.5 GHz follows the transition as measured in the 7 T MRI scanner. The total frequency shift, $\mathrm{\Delta }{\nu }_{+}$, is then found as $5\times {\nu }_0+{\nu }_i$.

Figure 4

The analog feedback used to control the frequencies of ${\mathrm{EOM}}_{1\text{--}4}$. The signal from the frequency-selective photodetector is passed through a lock-in amplifier, to generate the error signal $e_i$. This is integrated by an analog op-amp integrator, to produce a voltage $U_i$, controlling a VCO, producing a frequency ${\nu }_i$, which is sent to ${\mathrm{EOM}}_i$.

Figure 5

EXAAQ magnetometer prototype. (1) On-off switch. (2) Toptica DLC Pro laser controller. (3) Cable reels with 19 m fibers and probes. (4) Optical breadboard with free-space beam splitting, synthesizer and amplifier for ${\nu }_0$, ${\mathrm{EOM}}_0$, and magnetic shield with reference. (5) $1\times 4$ fiber splitter, amplifiers for ${\nu }_{1\text{--}4}$, and ${\mathrm{EOM}}_{1\text{--}4}$. (6) VCOs. (7) Lock-in amplifiers and integrators for probes 1–4. (8) Subrack with photodetectors and heating lasers. (9) Probe laser, ${\mathrm{EOM}}_{\mathrm{FM}}$, and lock-in amplifier and integrator for reference. (10) Oscillators for ${\nu }_{\mathrm{FM}}$ and VCO voltage scanning.

Figure 6

Error signals $e_i$, recorded as a function of ${\nu }_i$. This range of frequencies is produced by sweeping $U_i$ about 2 V. A fit to the linear region around resonance, $e_i=0$, returns the slopes $s_i$.

Figure 7

Example measurements, in the magnetic field of the MRI scanner. The sampling rate is 40 kHz. The field is assumed to be a perfect $7000.066$ mT. Colors blue, red, green, and yellow correspond to probes 1–4, respectively. Probe 4 is the one that has the least noise, with a standard deviation of $3.9\text{μ}\mathrm{T}$, over a 1 s measurement. Noise common among the probes is attributed to imperfect frequency stabilization of the probe laser.

Figure 8

The PSD of a measurement with probe 4. The shaded lower area shows the noise of the oscilloscope, used for measuring $U_i$, with disconnected input channels.

Figure 9

Magnetic-field measurements during an EPI sequence with probes 1–4, in colors blue, red, green, and yellow, respectively. In the upper plot are shown the differences between the first upper sidebands and the atomic resonances. In the middle plot are shown the VCO frequencies. In the lower plot are shown the magnetic fields calculated using Eqs. (2) and (4). The reader familiar with MRI sequence development will recognize slice selection, prewinder, $k$-space coverage, and spoiler [52].

Figure 10

Magnetic-field measurements during a spiral imaging sequence with probes 1–4, in colors blue, red, green, and yellow, respectively.

Figure 11

A small section of the EPI sequence shown in Fig. 9, recorded with the probes distributed along the $y$-axis. Blue, red, green, and yellow traces correspond to positions $-15$, $-7.5$, $+7.5$, and $+15\mathrm{cm}$ respectively.

Figure 12

Nonlinearity in the magnetic field of the $y$-gradient coil. In the upper plot, the magnetic-field values of the first gradient plateau in Fig. 11 are shown, along with an ideal 39.87 mT/m gradient. In the lower plot, the residuals between the measured field and the ideal gradient are shown. At $\pm 15$ cm we see deviations of about 4%.

Figure 13

Measurement of the first 16 ms after the spiral sequence shown in Fig. 10, recorded with the probes distributed along the $y$-axis. Blue, red, green, and yellow traces correspond to positions $-15$, $-7.5$, $+7.5$, and $+15\mathrm{cm}$ respectively. Oscillating eddy currents are clearly seen. Notice how the measurements have drifted about $20\text{--}30\text{μ}\mathrm{T}$, since the calibration done 15 min earlier.

Figure 14

Magnetic-field gradient along the $y$-axis calculated using the measurements by probes 1 and 4, shown partly in Fig. 13. In the upper plot, the time-domain signal is shown starting directly after the spiral sequence has finished and ending 45 ms later when the decaying oscillations are no longer visible. In the lower plot the corresponding spectrum is shown. The dashed lines show the known oscillations that the ECC system tries to compensate.