Functional neuroimaging has led to important advances in our understanding of the neural events that support behavior and has emerged as an important technique in the study of psychiatric and neurological disorders such as schizophrenia, dementia, depression, and epilepsy. Magnetoencephalography (MEG) is a particularly valuable technique that detects the extremely weak (10 fT – 1 pT) magnetic fields produced by neural currents with an array of > 200 highly sensitive magnetometers arrayed around the scalp. Despite its unique ability to localize neural currents with sub-centimeter spatial and millisecond temporal resolution, MEG’s potential as a research and clinical tool has yet to be realized due in large part to the high costs associated with the whole-head array of cryogenic superconducting quantum interference device (SQUID) magnetometers used in conventional MEG.
In an effort to greatly reduce the size and expense of MEG, we are developing small, fiber-coupled atomic magnetometers (AMs) that could be scaled up to form an array of magnetometers that covers the entire head. The acquisition and maintenance costs of an AM system will be significantly reduced in comparison to conventional SQUID-based systems for several reasons. AMs do not require cryogenic cooling. By removing the liquid helium cryogenic infrastructure from an MEG system, a significant portion of the acquisition cost and the bulk of the operating cost is eliminated. Furthermore, the overall size of the system can be significantly reduced. This further lowers acquisition cost because the magnetically shielded environment can be much smaller and less expensive. Along with the reduced size comes the possibility of having a portable system, further increasing the availability of MEG. As an added bonus, an AM developed for MEG could be straightforwardly adapted to benefit other SQUID-based techniques such as ultra low-field MRI.
Many highly sensitive magnetometers utilize free-space pump and probe beams that propagate orthogonally to each other and intersect in the vapor cell. This is primarily due to the ease of implementing this configuration in a laboratory setting. However, for the MEG application where compactness is essential to make an array surrounding the head, we orient the pump and probe beam such that they propagate along a single optical axis and deliver the light to vapor cell using a polarization maintaining single-mode optical fiber. The pump and probe beam are tuned to different strongly absorbing optical transitions out of the rubidium ground state and into the first excited electronic state. The "D1 line" at 795 nm must be used to optically pump the atoms. The "D2 line" at 780 nm cannot efficiently optically pump the cloud but is well suited as a linearly polarized probe beam. By pumping and probing at different wavelengths it is possible to maintain separate control of the pump and probe beam powers and detunings.
In this single axis configuration the Faraday rotation response is zero for fields oriented along the optical axis of the magnetometer and, in principle, independent of the orientation of the magnetic fields applied perpendicular to the optical axis. To define a sensitive axis, we apply a ~ 1 kHz oscillatory magnetic field perpendicular to the optical axis. By demodulating the signal using lock-in detection, we achieve a linear response to a magnetic field along the modulation direction which is centered around zero field with a dynamic range of a few nanotesla.
The single axis magnetometer is shown schematically in the figure. Light exits the optical fiber and passes through and polarizer and a dichroic waveplate. The dichroic waveplate converts the linearly polarized 795 nm beam into circular polarization for optical pumping while maintaining the linear polarization of the 780 nm probe light. The beams are collimated and enter an electrically heated rubidium vapor cell. The beams pass through the vapor cell, retroreflect, pass back through the vapor cell, and are then focused onto a quadrant photodiode polarization analyzer. Pump light is removed prior to polarization analysis with an interference filter that passes light at 780 nm but not at 795 nm.
Below is a picture of the sensor. The units of the tape measure are in inches. It is constructed entirely of nonmagnetic materials with supporting structure made of G-10 fiberglass composite. The vapor cell has a 2.5 cm diameter and the microporous ceramic oven has a footprint on the head of 5 cm X 5 cm. The sensor is a 4-channel device. By replacing the single element photodiode (as shown in the picture) with a segmented quadrant photodiode, signals from different parts of the cloud are obtained. These signals are subtracted from each other to remove common-mode noise, resulting in a gradiometric measurement. High quality common-mode rejection only occurs with precise imaging of the cloud onto the segmented photodiode. The inferred single-channel sensitivity as determined from gradiometric measurements is 5 ft/Hz1/2 at 3-11 Hz. The 3 dB bandwidth of the magnetometer is 11 Hz. We have used this device to measure evoked responses in both the somatosensory and auditory cortex in a single human subject. Comparison of data measured with the atomic magnetometer to that of a conventional SQUID MEG system shows that both devices measure a similar response with comparable signal-to-noise ratios.