Abstracts

Article

NMR and MRI Obtained with High Transition Temperature DC SQUIDs

R.E. de Souzaa,^‡ ‡ Permanent address: Departamento de Física, Universidade Federal de Pernambuco, Av. Prof. Luiz de B. Freire, s/n, 50670-901 Cidade Universitária, Recife - PE, Brazil , K. Schlengab,^† ‡ Permanent address: Departamento de Física, Universidade Federal de Pernambuco, Av. Prof. Luiz de B. Freire, s/n, 50670-901 Cidade Universitária, Recife - PE, Brazil , A. Wong-Foy^a, R. McDermott^b, A. Pines^a, and John Clarke^b

aDepartment of Chemistry, University of California, Berkeley and Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA

^b Department of Physics, University of California, Berkeley, and Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, 94720, USA

Introduction

In a typical NMR experiment, the signal is proportional to the product of nuclear magnetization¹ and the Larmor frequency. Since both magnetization and precession frequency scale with magnetic field strength, the NMR signal is greatly enhanced as magnetic field is increased. However, for systems with short-range order, such as amorphous and polycrystalline materials, high-field NMR spectra are broad and hard to resolve. In such systems, the random molecular orientation with respect to the applied magnetic field causes different local Larmor frequencies at each molecular site. Therefore, the spectra display a characteristic "powder pattern"². Several techniques have been developed to narrow these spectra and facilitate the acquisition of structural information³. An alternate approach is to detect NMR signals at magnetic fields that are low compared to local fields, so that all equivalent sites have the same resonance frequency; thus, the spectra are considerably simplified. However, for conventional NMR detectors, the signal-to-noise (S/N) ratio is markedly reduced in low field⁴, making data acquisition extremely time consuming. SQUID-detected NMR offers an alternative solution to this problem. While conventional pickup coils measure the time derivative of a magnetic flux⁵, SQUIDs measure the magnetic flux directly⁶. This leads to a much higher S/N ratio at low frequencies.

Over the last decade, low transition temperature (low-T_c) SQUIDs have been employed in numerous experiments to detect NMR and nuclear quadrupole resonance (NQR) signals^7-14. Even for magnetic resonance imaging (MRI), where static and/or time-dependent magnetic field gradients are applied, low-T_c SQUIDs have proven to be a sensitive and stable sensor^15-16. However, low-T_c SQUIDs must be operated at liquid helium temperatures. At this low temperature, the number of samples having sufficiently short spin-lattice relaxation times to allow for signal averaging in pulsed experiments is limited. Low-T_c SQUIDs are inconvenient for NMR investigations of biological samples at room temperature because of the relatively complicated cryogenics involved. Nevertheless, low-T_c SQUID NMR experiments were reported by Seton et al.^12,16 and Kumar et al.^13,14. Later Kumar et al.¹⁷ used a high-T_c SQUID with simpler cryogenics for the detection of NMR signals.

The aim of this article is to present a new high-T_c dc SQUID spectrometer that is able to detect NMR signals from samples at room temperature in magnetic fields as low as 0.059 mT. The feasibility of NMR detection in these low fields is demonstrated with spectral data of ¹H, ¹⁹F, and laser-polarized ¹²⁹Xe nuclei. In addition, we have implemented a system of magnetic field gradient coils to perform MRI under low-field conditions. Two-dimensional proton magnetic resonance images of filled phantoms have been obtained using this apparatus.

Experimental Configuration

Our spectrometer is based on a high-T_c dc SQUID operating in a flux-locked loop. The sensor is a multilayer multiloop integrated magnetometer¹⁸. In the relevant frequency range the system flux noise is 25 mF_o/Hz^1/2, where F_o = h/2e is the flux quantum. Taking into account the effective area of 1.8 mm², we find a magnetic field noise of 30 fT/Hz^1/2. The magnetometer is operated without a tuned input circuit; this allows for broadband detection from DC to 1.1 MHz, which is a desirable feature for multiple resonance spectroscopy. Figure 1 shows a schematic layout of the spectrometer. The voltage across the SQUID is amplified by a low-noise pre-amplifier at room temperature. After further amplification, the signal is integrated and fed back via a resistor into a single-turn coil inductively coupled to the SQUID. The voltage across the feedback resistor is further amplified and filtered (Rockland model 442) and then fed into a spectrum analyzer (Hewlett Packard-model 3561A) and a data acquisition system (Tecmag–ARIES and Macintosh-IIci). The control unit produces audio-frequency drive pulses and performs data digitization and storage. Because high power audio-frequency pulses inevitably couple excessive flux into the SQUID, driving the integrator into saturation, a reset signal is forwarded to the SQUID electronics to short the integrator during excitation pulses. The pulse sequence generated by the Tecmag system is used to gate a function generator (Hewlett Packard 3314A) in order to produce the audio-frequency excitations. These pulses are applied to the transmitter coils after amplification by a power amplifier (Krohn-Hite-model 3322 filter).

Figure 2 shows a simplified drawing of the dewar and coil system. The SQUID magnetometer is operated in vacuum and separated from the sample by a sapphire window; the SQUID-to-sample separation is less than 1 mm. A 0.1 m long sapphire rod thermally anchored to the liquid nitrogen reservoir provides efficient cooling. The dewar is similar to that described in detail in elsewhere¹⁹. A pair of Helmholtz coils, each with a diameter of 0.12 m and 1,300 turns, produces the static magnetic field B₀. The calculated field homogeneity at the center of the coils is better than 0.07% and 0.7% correspond, respectively, to over a cubic volume of 1 cm³ and 8 cm³. A three-layer mu-metal shield enclosing the dewar and coil system attenuates external fields by a factor of 1,430 at zero frequency. All experiments are carried out in an electromagnetically shielded room.

For our pulsed NMR experiments, the excitation field B₁ is produced by a Helmholtz pair oriented perpendicular to the B₀ direction. Each excitation coil has a diameter of 8.6 cm and contains 20 turns; the coils are tuned to the Larmor frequency of the sample. The calculated field homogeneity at the center of these coils is better than 0.12% and 1.2% correspond, respectively, to over a cubic volume of 1 cm³ and 8 cm³. For the purposes of imaging we used a Maxwell pair, oriented parallel to the static field, to produce a magnetic field gradient G. Each gradient coil has a diameter of 0.12 m and contains 50 turns. The linearity of the field gradient is better than 0.3% and 1.0% correspond, respectively, to over a cubic volume of 1 cm³ and 8 cm³ at the center of the coils. All three pairs of coils were wound from insulated copper wire and mounted on a common frame made from G-10 fiberglass and plexiglass. In order to avoid acoustic resonances in the coils, this frame was designed to be extremely rigid.

Precessing magnetization of the sample is induced by magnetic field audio-frequency pulses from the transmitter coils. We have employed Hahn spin echo²⁰ sequences for most of the spectroscopic experiments at various field values. We typically used a p/2 pulse with a field amplitude of 60 mT and a length of 200 ms, followed a few milliseconds later by a p pulse with the same field amplitude. In order to avoid pickup of ringing produced in the coils by the mismatch between the coils and the signal generator, the p pulse is followed by a 100 ms delay before the SQUID feedback loop is switched on. Data acquisition begins 100 ms later. To allow the spins to relax completely the repetition rate is low, approximately 2.5 Hz.

For our NMR imaging experiments, we have employed pulse sequences with Ernst tipping angles²¹ in order to minimize acquisition time. Here we applied pulses with a field amplitude of 60 mT, a length of 120 ms, and a repetition rate of 25 Hz. A static magnetic field gradient of 2 mT/m is applied during the entire acquisition. For two-dimensional imaging experiments, our sample was rotated in angular steps of 15°; for each angle we performed 30,000 averages.

Results and Discussion

Our first experiments were performed on samples with either a high density of nuclear spins or with hyperpolarized ¹²⁹Xe. For our ¹H samples we chose mineral oil because of its relatively short spin-lattice relaxation time T₁ and its relatively long spin-spin relaxation time T₂. This allows us to exploit the advantage of pulsed NMR experiments and to improve the S/N ratio by signal averaging. In addition, we have measured ¹H signals from human tissue. For studies of ¹⁹F nuclei we used deuterated trifluoracetic acid, which has a T₁ value comparable to T₁ in ¹H samples. The samples of ¹²⁹Xe were polarized far above the equilibrium value by means of spin-exchange collisions with optically pumped rubidium, as described elsewhere^11,22.

A. Spectroscopic experiments

Most of our experiments were performed in magnetic fields between 1 mT and 2 mT; depending on the nuclei, the Larmor frequencies were between 10 kHz and 90 kHz. In a magnetic field of 2.03 mT we can detect the proton spin echo signal at 86.67 kHz without signal averaging. The corresponding spectrum after Fourier transformation is shown in Fig. 3(a). Given the extremely weak equilibrium polarization of ¹H nuclear spins (7 x 10^-9) at this field and room temperature, the resulting spectrum demonstrates the high sensitivity of the SQUID detection scheme. We have also performed experiments on hyperpolarized ¹²⁹Xe which had a pumped polarization of about 2 x 10^-2, an enhancement of seven orders of magnitude over the equilibrium polarization at room temperature. We have measured the free induction decay (FID) of ¹²⁹Xe with the multiloop magnetometer after a single p/2 pulse. The sample tube was 1.4 cm long and had an inner diameter of 0.5 cm. Figùre 3(b) shows the Fourier transform of the ¹²⁹Xe FID; the S/N ratio is approximately 100. The extremely high S/N ratio attained in this experiment indicates that laser-polarized ¹²⁹Xe has great potential for use in SQUID MRI²³ as an agent to study voids in materials and in living systems.

In order to improve on the S/N ratio of our first experiments with protons, we averaged 1,000 times under the same conditions. The averaged spectrum of the proton spin echo is shown in Fig. 4(a). In this experiment we obtained a S/N ratio of approximately 44 with an averaging time of 10 min. Signal averaging further allowed us to resolve the proton spin echo produced by 1 mL of mineral oil in a field of 0.059 mT, which is comparable to the Earth’s magnetic field. Even though the equilibrium spin population of protons in mineral oil at room temperature, , is only 2 x 10^-10, we observe a clear peak at 2.5 kHz after 2,000 averages²⁴. We might note that while Packard et al.²⁵ have reported NMR in the Earth’s magnetic field, their experiments were carried out with the aid of prepolarization techniques.

We have demonstrated the system’s ability to measure NMR signals from nuclei other than ¹H in thermal equilibrium. We chose fluorine for its high gyromagnetic ratio and its environmental relevance. Figure 4(b) shows a spectrum obtained from ¹⁹F nuclei in deuterated trifluoracetic acid under the same conditions as in the previous experiment with protons.

In order to test our ability to perform NMR in living systems, we carried out experiments with human tissue. These experiments were performed without the lid of the mu-metal shield. We have measured the ¹H-NMR signal from the index finger of one of the authors under the same conditions as the ¹H experiments in mineral oil. The Fourier transformed spin echo of protons from the tip of an index finger is shown in Fig. 5. This important test has ruled out several expected problems, such as direct noise coupling into the SQUID from the surroundings, noise coupling to the SQUID via the finger (which might behave as an antenna), audio-frequency from the transmitter coils coupling to the SQUID via the finger, and mistuning of the transmitter coils due to the presence of the finger.

B. Imaging experiments

The high S/N ratio obtained with the multiloop magnetometer in fields of 2 mT has enabled us to obtain proton magnetic resonance images. For the purposes of imaging we must take into account the fact that the SQUID magnetometer does not enclose the sample, but rather acts as a surface coil, similar to those used in conventional NMR/MRI²⁶. The intensity of the NMR signal from a given part of the sample depends on its position relative to the magnetometer. Therefore our imaging method involves both frequency encoding and position encoding of the spins in our sample²⁷. Our earliest two-dimensional imaging experiments involved a simple scheme in which the sample is placed on the top of the SQUID magnetometer and sequentially rotated in a magnetic field gradient applied parallel to the plane of the magnetometer. Hence, on resonance and with the gradient G along the direction of the static magnetic field B₀, the signal at a given angular position a is given by:

where f(x, y, z) describes the contribution of an ensemble of dipoles located at the coordinates (x, y, z) to the total magnetic flux through the magnetometer²⁷, r(x, y, z) is the spin density at (x,y,z), and k = gGt/2p, where g is the gyromagnetic ratio; the center of the magnetometer defines the origin in space (0,0,0), and the xz plane is taken to coincide with the plane of the magnetometer. Therefore the function signal (k,a) is intrinsically defined in a polar raster.

The linewidth in a given experiment, along with the desired spatial resolution, defines the minimum value for the magnetic field gradient necessary to obtain magnetic resonance images. From linewidth measurements on the spectrum shown in Fig. 4(a) we calculate a minimum field gradient of 0.25 mT/m in order to achieve a spatial resolution of approximately 0.35 cm. The maximum magnetic field gradient suitable for MRI experiments is mainly set by the S/N ratio attainable. In our MRI experiments we used a static magnetic field gradient of about 2 mT/m.

Using the filtering back-projection reconstruction method^28,29 we calculate the imaging function, given by

This function represents the spin density distribution of the sample projected along the y direction (the vertical direction in our experiments). For objects with axial symmetry, the function f(x, y, z) can be considered constant. For objects with quasi-axial symmetry, f(x, y, z) can be taken to be nearly constant for points in the sample whose distance from the center of magnetometer is roughly comparable to the magnetometer radius. Therefore, for our magnetometer dimensions, the linear sample size should be about 0.4 cm.

We measured signal (k, a) from a sample consisting of 0.6 mL of mineral oil in a cylindrical container of 0.7 cm diameter and 2 cm height. Two glass rods were placed inside the container, aligned with the container axis and positioned roughly along the container diameter. In Fig. 6 we show a gray scale plot of the function Image(x, z) for the phantom described above. From the measured linewidth of 70 Hz without an applied magnetic field gradient, we estimate the spatial resolution to be about 800 mm. We might note that it is possible to acquire the signal function in Cartesian coordinates by using pulsed magnetic field gradients^30,31. With this end in mind, a new system of coils, gradient power amplifiers, and gradient controllers, is currently being developed.

Conclusions

We have developed a novel system for SQUID-detected NMR and MRI studies of samples at room temperature and in magnetic fields ranging from 59 mT to 2000 mT. This system can accommodate samples of 1 cm scale and larger, thereby permitting spectroscopy of biological samples. We have observed NMR signals from protons, fluorine, and hyperpolarized ¹²⁹Xe. In particular, proton NMR signals have been measured from samples in thermal equilibrium at room temperature and in a magnetic field comparable to that of the Earth. We have acquired low-field proton NMR data from the tip of a finger, suggesting that we could perform MRI on living tissue. We have also obtained two-dimensional MRI from phantoms filled with mineral oil. We plan to extend our NMR experiments to other nuclei, as well as to implement NQR to study systems with quadrupole nuclei. Efforts are underway to incorporate pulsed gradients into our two-dimensional imaging scheme. Finally, we note that although our high-T_c dc SQUID-based detector has produced impressive proton NMR signals at 2 mT, one could further improve the S/N ratio, for spectroscopic proposes, by means of prepolarization techniques or dynamic nuclear polarization (DNP), depending on the system in study.

With regard to the ¹²⁹Xe studies, a self-contained gas flow system is currently under construction that will permit continuous optical pumping. The high sensitivity of the SQUID-based detector at low magnetic fields and the high magnetization of hyperpolarized ¹²⁹Xe can be combined to perform high resolution MRI of voids, for instance, in granular and porous media. It should also be possible to perform low-field NMR imaging of hyperpolarized ¹²⁹Xe dissolved in solution, which would be attractive for future in vivo studies.

Acknowledgments

We thank D.M. TonThat for helpful advice during the early phase of this project. The multiloop magnetometer was fabricated in 1994 by Frank Ludwig. This work was supported by the Director, Office of Energy Research, Office of Basic Energy Sciences, Materials Sciences Division, of the U.S. Department of Energy under Contract No. DE-AC03-76SF00098, the Deutsche Forschungsgemeinschaft, and the Brazilian Agency for Research CNPq.

Received: March 29, 1999

^†

Present address: BRUKER Analytische Messtechnik, Wikingerstr. 13, 76189 Karlsruhe, Germany

Publication Dates

History