The uninformed reader might be misled into believing that optical magnetometers (see Physics Today, Physics Today 0031-9228 56 7 2003 21 https://doi.org/10.1063/1.1603066 July 2003, page 21 ) will soon displace SQUIDs for biomagnetic measurements. I have spent 30 years using superconducting quantum interference device (SQUID) magnetometers and miniature induction-coil magnetometers to measure biomagnetic fields with ever-increasing spatial resolution. Several significant obstacles lie between optical magnetometers and high-resolution measurements of the brain’s magnetic field. 1
Several points are worthy of careful analysis. First, the typically reported sensitivity of SQUIDs, 1 fT/Hz1/2, is not a fundamental limit set by the Josephson effect, but a practical one. One must consider thermal (Johnson and thermoelectric) noise within the Dewar flask and the object being studied; the object’s intrinsic noise—for example, brain noise (thermal currents and background neural activity within the brain); fluctuations of magnetic shields; environmental magnetic noise; the desired spatial resolution; and the dynamic range of the electronics. SQUID sensitivities of nearly 10−17 T/Hz1/2 have been reported. 2 With the exception of SQUID microscopy, more sensitive SQUIDs are not generally needed for biomagnetism, because the existing ones have reached the noise floor set by the Dewar, the object being studied, or the environment.
Because superconducting pickup coils can either trap or exclude magnetic flux, SQUIDs have the tremendous advantage of being able to measure femtotesla magnetic fields in the presence of tesla-strength steady fields and microtesla time-varying ones. A well-balanced SQUID gradiometer operating in the geomagnetic field can achieve a common mode rejection of up to 107. Electronic gradiometry allows high-quality SQUID recordings of the adult and fetal magnetoencephalogram (MEG) in unshielded laboratories. 3 The magnetic shields and cancellation coils needed for the Kominis optical magnetometer reduce the external magnetic field by a factor of 106, four orders of magnitude greater than the shielding factor of approximately 200 for the standard magnetically shielded rooms often used for biomagnetic measurements. That both the optical magnetometer and the object producing the magnetic signals have to operate in such a low magnetic field is a severe practical constraint, such that Kominis and colleagues 1 suggest using superconducting magnetic shields to reduce “the overall noise of the magnetometer.”
The most important point missed in the discussion of the recent optical advance is that the ability of any magnetometer to discriminate between two adjacent sources, such as a pair of cortical columns (modules), is determined not only by the size of the magnetometer but also by the cortex–magnetometer separation. A good rule of thumb is that the magnetometer size and the source-to-sample distance should be approximately equal, lest the spatial resolution of a small magnetometer be blurred by a large separation. To record the MEG from the intact brain, the cortex-to-magnetometer separation would be at least 1 cm—the skull thickness. Hence, millimeter-sized magnetometers would provide little additional benefit.
To be competitive with current commercial SQUID MEG instruments, the magnetic field would have to be measured at about 300 separate locations on the surface of the skull, and ideally more than one field component would be measured. At a minimum, the field component perpendicular to the skull would be needed, which means that each normal-component magnetometer would be measuring the field in a different direction, complicating the control and readouts for optical magnetometers. Although multichannel SQUID MEG systems can already localize a single cortical dipole current source to within a millimeter, many clinically significant MEG sources are distributed over 10 cm2 or more. An optical magnetometer with submillimeter spatial resolution could be used effectively only if the cortical column being measured were a comparable distance from the magnetometer. That would require removal of the skull—readily achieved in whole-animal or brain-slice preparations, but not achieved noninvasively. Finally, the optical magnetometer operates in a 180°C oven, and noiselessly protecting a 37°C brain from a hot magnetometer may be harder than from a −270°C SQUID, since a warm Dewar may be a lot noisier than a cold one!
SQUIDs are regularly recording the MEG, and the magnetometer cost is only a small fraction of the cost of a clinical MEG facility. SQUIDs are already within striking distance of the cortical column: Microscopes with a 4-K, 100-µm SQUID now operate 100 µm from biological tissue. 4 The sensitivity continues to improve: A new 40 × 40-μm SQUID sensor has less than 1 pT/Hz1/2 noise and is well matched to the 60 pT field predicted to exist 50 µm from a single cortical column, according to Franz Baudenbacher of Vanderbilt University.
I applaud the continued development of optical magnetometers. These devices offer interesting prospects for two-dimensional imaging and fundamental measurements, but it is premature to consider their replacing SQUIDs for most applications.
