Data on the directional changes of a full magnetization vector during cycling to cryogenic temperatures can provide important insights into the low-temperature magnetic properties of natural and synthetic materials. These data also provide an empirical basis for the application of low-temperature treatments in paleomagnetism, for example, the removal of viscous magnetization in magnetite-bearing rocks. However, existing instruments only allow continuous measurement of magnetization along a single axis, hampering experimental and theoretical advances in rock magnetism and the implementation of low-temperature techniques into regular paleomagnetic practices. Here we describe development of a novel low-temperature insert designed in collaboration with William S. Goree Inc., which allows measurement of directional behavior of a full magnetization vector during zero-field low-temperature cycling. Pilot experiments on well-controlled polycrystalline samples of pseudo-single-domain (PSD) and multidomain magnetite as well as on a natural sample containing PSD magnetite indicate that the orientation of a saturation isothermal remanent magnetization (SIRM) imparted at room temperature remains constant during low-temperature cycling to 20 K. This observation lends additional support to low-temperature cycling as a cleaning technique in paleomagnetism. The SIRM imparted in an individual crystal of magnetite showed systematic, albeit small changes upon both cooling and warming through the Verwey temperature, which may reflect switching between the easy magnetization directions. However, the switching effect may be significantly attenuated by crystallographic twinning in magnetite below the transition. Overall, our results demonstrate the potential of the directional low-temperature magnetometry for the advancement of our understanding of the properties of natural and synthetic materials.
Although the study of low-temperature properties of magnetic minerals has a long history [e.g., Li, 1932; Verwey, 1939; Bickford, 1953; Nagata et al., 1964], subsequent advances have been somewhat hampered by the technical complexity and high cost of low-temperature magnetic measurements. The commercial development during the last two decades of devices for cryogenic magnetometry (for example, the Magnetic Property Measurement System by Quantum Designs and the Vibrating Sample Magnetometer equipped with a cryostat by Princeton Measurement Corporation) has promoted a considerable expansion of the experimental database on low-temperature magnetic behavior of natural magnetic minerals [e.g., Dunlop and Özdemir, 1997; Kosterov, 2007, and references therein]. Nevertheless, in spite of these successes, two factors remain that impede experimental and theoretical advances in rock magnetism, and the implementation of low-temperature techniques into regular paleomagnetic practices. First, the specialized equipment necessary for such studies is still not easily affordable for a typical university-based rock/paleomagnetic laboratory. Second, existing instruments do not allow measurement of the directional behavior of a full magnetization vector during low-temperature cycling. Therefore, our understanding of the behavior of magnetic minerals at low temperatures is based almost entirely on magnetic moment measurements along a single axis.
However, directional measurements could provide valuable constraints on the physical processes governing the magnetic properties of minerals at low temperatures. Some of these processes, such as those associated with the Verwey transition in magnetite, still remain obscure [e.g., Walz, 2002; Subias et al., 2004; Rozenberg et al., 2006]. In addition, three-component measurements can test the assumption that low-temperature cycling does not change the direction of a primary magnetization vector. While this assumption is crucial for using low-temperature treatment for paleomagnetic cleaning, a few authors have suggested recently that some low-temperature data may be better explained assuming that the magnetic moment orientation changes during low-temperature treatment [e.g., Liu et al., 2003]. Finally, understanding the potential changes of magnetization vector upon low-temperature cycling may allow one to separate not only a stable component but also secondary NRM components based on continuous measurement of directional changes of natural remanent magnetization (NRM) during low-temperature demagnetization. Such an approach could be especially useful for rocks containing metastable magnetic carriers prone to chemical alteration upon heating (e.g., titanomaghemite).
In this paper, we report our efforts to develop a device to be used with a (de facto) standard superconducting rock magnetometer for continuous three-component magnetization measurement during low-temperature cycling. A prototype of this device was built by William S. Goree Inc. and underwent several modifications based on testing performed at the Paleomagnetic Laboratory of the University of Rochester between 2002 and 2006 [Smirnov et al., 2004]. Here we describe the final modification of the device, and report the results of our test experiments using well-controlled polycrystalline samples of nearly stoichiometric pseudo-single-domain (PSD) and multidomain magnetite, a natural sample containing PSD magnetite, and a large individual crystal of magnetite.
2. Overview of Low-Temperature Properties of Magnetite
Magnetite (Fe3O4), the most important and common magnetic mineral on Earth, is characterized by unique magnetic properties at low temperatures. When cooled below room temperature, magnetite undergoes two magnetic transitions. The first, known as the isotropic point (Ti), occurs at ∼130 K when the first constant of magnetocrystalline anisotropy (K1) passes through zero and changes sign [e.g., Syono, 1965] and the easy magnetization axes change their orientation from  (above Ti) to  (below Ti [e.g., Stacey and Banerjee, 1974]). Magnetite on further cooling undergoes a phase transition from cubic to lower (probably monoclinic [Zuo et al., 1990]) crystallographic symmetry at TV ≈ 120 K, called the Verwey transition [Verwey, 1939]. The transition is accompanied by a change in magnetocrystalline anisotropy from cubic to uniaxial, when one of the cubic  directions becomes the monoclinic c axis [Kakol, 1990], which also becomes a new easy magnetization axis. As a result of this change, a complete reorganization of the domain structure is expected [e.g., Halgedahl and Jarrard, 1995]. The Verwey transition is also accompanied by abrupt changes of magnetic properties controlled by magnetocrystalline anisotropy, such as remanent magnetization [e.g., Aragón et al., 1985; Aragón, 1992].
On warming from TV to room temperature, the remanence shows irreversible behavior with some fraction of the initial magnetization recovered after low-temperature cycling. The recovery has been linked with the stiffening of domain walls (due to the increasing magnetocrystalline anisotropy) on heating [e.g., Muxworthy and McClelland, 2000]. The recovered fraction (the low-temperature memory) is largest for single-domain (SD) grains and decreases as grain size increases, but even millimeter size magnetite crystals may exhibit some memory [e.g., Heider et al., 1992; Özdemir and Dunlop, 1999]. A parameter commonly used to characterize the low-temperature memory is the ratio (RLT) of the remanence recovered after low-temperature cycling to the initial remanence [e.g., Halgedahl and Jarrard, 1995; King and Williams, 2000]. In addition to the grain size, the ratio depends on the initial type of remanence [e.g., Muxworthy et al., 2003] and internal stress [e.g., Hodych et al., 1998; King and Williams, 2000].
The effect of low-temperature treatment on remanent magnetization is the basis for selective low-temperature demagnetization of remanence carried by multidomain (MD), pseudo-single-domain (PSD) or equidimensional SD grains when they are cycled through TV [Ozima et al., 1964]. However, in spite of the experimental and theoretical efforts, the exact processes governing low-temperature cycling behavior of magnetite are still unresolved [e.g., Hodych et al., 1998; Özdemir and Dunlop, 1999; Muxworthy et al., 2003].
3. Low-Temperature Insert for Directional Measurement of Remanent Magnetization
A prototype of the low-temperature insert for directional measurement of remanent magnetization was designed and built by William S. Goree Inc. We tested the prototype with a 2G DC superconducting quantum interference device (SQUID) horizontal magnetometer (with a 4 cm access and high-resolution coil geometry) at the University of Rochester. The probe consisted of three parts (Figure 1). The first part is a 1.5 m long double-walled fiberglass tube (3.8 cm outer diameter, 2.6 cm inner diameter) with a vacuum valve, which was inserted into the magnetometer access opening, so that its ends protruded from both sides of the magnetometer (Figure 1). During experiments, air was pumped out of the space between the walls (using a Welch DirecTorr 8905 vacuum pump) to isolate the cooled interior of the insert from the magnetometer. The vacuum level was better than 5 mtorr. In order to accelerate the homogenization of temperature, several 10 inch strips of single crystal quartz were glued inside the tube (along the tube axis) in the measurement region.
The second part of the insert, a helium supply assembly consists of a fiberglass tube (2.5 cm outer diameter, 1.0 cm inner diameter) with a port for a helium transfer line on one side and a bronze nonmagnetic diffuser on the other side (Figure 1). The assembly was inserted into the main tube from the cold head side of the magnetometer so that the diffuser was located about two inches from the silicon strips (i.e., seven inches from the center of magnetometer's sensor region). Liquid helium was supplied from a dewar through a transfer line equipped with a flow regulator (adjustable between 0 and 2 L/min) custom built by Kadel Engineering Corporation. The transfer line was inserted into the assembly until it was pressed against the diffuser.
The gas temperature was measured with a Lakeshore DT-470 silicon diode temperature sensor (hereinafter, Diode A) mounted on the diffuser. Diode thermometry is based on the temperature dependence of the forward voltage drop in a p-n junction biased at a constant current (15 μA in our experiments). Advantages of using a diode temperature sensor (in comparison with a thermocouple) include its relatively high voltage output (between 0.5 and 1.7 V in our installation) and an inverse rather than direct proportionality between the voltage (V) and temperature (T).
In order to regulate the gas temperature during experiments, a noninductive 50 W heater was wound around the diffuser (Figure 1). The heater was controlled by a Lake Shore Model 331 temperature controller able to provide up to 50 W of continuous power to a resistive heater load. The control output was calculated using a proportional integral derivative (PID) algorithm based on temperature set point and feedback from the control sensor (diode A). A slower PID algorithm was utilized to avoid overshooting of the set point temperature. The temperature controller had a built-in calibration curve for the DT-470 diode and allowed immediate reading (in °K) the temperatures of diodes A and B (with display updates twice each second). No effect of the heater current on the SQUID readings was observed.
The third part of the insert, a sample holder assembly is a fiberglass tube (2.5 cm outer diameter, 1.0 cm inner diameter) with a 30 cm long Mylar tube (1 cm in diameter) extension (Figure 1). A multiple O-ring seal system allowed moving the sample in and out of the magnetometer sensor region. As a sample holder we used a nonmagnetic plastic cylinder (1 cm in diameter and 0.8 cm high) with a central opening for a sample and several additional holes to allow helium gas flow to the region behind the sample toward the helium release line (Figure 1, inset). A sample, sealed in a gelatin capsule, was tightly fixed into the central opening of the holder. The holder with a sample was then inserted into the Mylar tube extension and fixed using nonmagnetic Kapton tape.
A second Lakeshore DT-470 temperature sensor (Diode B) was located close to the base of the Mylar tube extension, to avoid interference with the magnetometer pickup coils. The outer end of the sample holder assembly had a port for the helium release line (a stainless steel double-walled tube with a vacuum valve). Before an experiment, air was pumped out of the space between the walls to enhance the thermal isolation between the gas release line and the magnetometer. For measurements, the assembly was inserted into the low-temperature probe, so that the sample was in the center of the magnetometer sensing region (Figure 1).
Because of rapid changes of magnetization and temperature during our experiments, using the standard 2G software was impractical. In addition, the standard software does not allow continuous recording of magnetometer readings. Therefore, in order to record the magnetic remanence and temperature changes, we used a fast data acquisition unit (Agilent Technologies 34970A) controlled by a specialized software. While the unit allows up to 600 DC voltage readings per second, in our low-temperature experiments we mostly used a recording rate of 0.5–1 readings per second for each channel.
The analog voltage output from the three SQUID control display consoles (X, Y and Z axes) of the magnetometer together with the voltage output from both temperature sensors (diodes) were directed to the Agilent unit through the first five channels of a 16 channel reed multiplexer (HP 34902A) (Figure 1). The voltage output versus time (V(t)) was monitored in real time to identify potential problems (such as flux jumps) during the experiment. After an experiment, the records from the five channels were saved with time marks in an ASCII file for further processing and visualization.
The full voltage output range for SQUID consoles was ±10V which corresponds to ±1 magnetic flux quantum (Φ0) at the 1X measurement range, ±10 Φ0 at the 10X range, ±100 Φ0 at the 100X range, and ±1000 Φ0 at the 1000X range. Accordingly, the magnetic moment components (Mi) were calculated using the expression
where Vi is the voltage output from 2G SQUID sensors, Ci is the magnetometer's calibration constant (Cx = −3.77 · 10−5 emu/Φ0, Cy = −3.87 · 10−5 emu/Φ0, and Cz = −1.805 · 10−5 emu/Φ0), and R is a number corresponding to the measurement range used (1, 10, 100 or 1000 Φ0/V). Because the 2G SQUID output is reset to zero whenever ±10V is reached, for our experiments we selected the measurement range so that the output from SQUIDs stayed within the ±10V range to avoid flux jumps (identified by a saw tooth pattern on the V(t) curves). The total magnetic moment, declination and inclination were calculated using the following formulas:
Temperature was calculated from the voltage output from diodes A and B using the DT470 calibration curve.
4. Experimental Results
4.1. Experimental Procedure
The generic experimental sequence is described below.
1. The insert (the main tube) was inserted into the magnetometer and fixed to prevent any accidental motion of the insert with respect to the magnetometer during the experiment.
2. The helium supply assembly was inserted into the main tube and the transfer line was inserted into the assembly.
3. Recording of the output from SQUID sensors and temperature diodes was started and continued for about one minute before the next step to obtain a “zero (base)” line.
4. The sample holder assembly with a sample was inserted into the main tube and the magnetization components were recorded for 30–60 s at ambient temperature before the next step.
5. The other end of the transfer line was put in a liquid helium dewar. In a short time (usually one or two minutes), cold gaseous and, then, liquid helium started to proceed into the insert and cool the sample. In order to prevent fluctuations of liquid He flow, a small (∼0.25 psi) pressure gas helium was applied to the dewar.
6. For cooling, the set point was set on the temperature controller to a desired temperature. After the readings from the both temperature diodes equalized, the set point was set to a lower temperature, the temperature equilibration achieved, and so forth until the lowest desired temperature was reached. The temperature step varied between 5 and 15 K. The voltage output from SQUID sensors and temperature diodes was constantly recorded during cooling. The duration of cooling to 20 K typically did not exceed 1 h.
7. Warming was done in the same fashion. However, the equalization of diode temperatures took much longer time (typically several hours).
8. After the sample temperature reached room temperature, the sample was taken out of the insert, and the zero level and temperature were recorded for about one minute to check for a possible drift of the zero line.
9. The recording of the voltage output from the SQUID sensors and the diodes was stopped.
Before the measurement of our test samples, we measured the directional behavior of magnetization of the empty insert (with the helium supply and sample holder assemblies) upon cycling to 20 K. The intrinsic magnetic moment of the insert at room temperature was 1.71 · 10−9 Am2 (1.71 · 10−6 emu). We observed no significant variation of the intensity and direction of the insert magnetic moment upon low-temperature cycling.
We tested the device using well-controlled polycrystalline samples of nearly stoichiometric pseudo-single-domain and multidomain magnetite, a natural magnetite-bearing sample, and an individual magnetite crystal. The low-temperature properties of magnetite are well studied for a variety of grain sizes and oxidation states [e.g., Dunlop and Özdemir, 1997; Kosterov, 2007], which makes the mineral an ideal test material. In addition, the easily identifiable Verwey transition at ∼120 K in nearly stoichiometric magnetite may serve an independent temperature indicator.
For two synthetic polycrystalline samples, M1.5 and M30 (Table 1) the temperature dependence of a room temperature (RT) SIRM was previously measured using a Quantum Design Magnetic Property Measurement System (MPMS) (Figure 2). The grain size and shape distributions of the samples were determined by visual inspection of images obtained using scanning electron microscopy (SEM) [Smirnov, 2006]. In both samples, magnetite grains were characterized by irregular, nearly equidimensional shape with grain size distributions well approximated by a lognormal distribution. The magnetic hysteresis data indicated pseudo-single-domain and multidomain magnetic behavior for samples M1.5 and M30, respectively, consistent with their grain size distributions (Table 1). Thermal demagnetization of low-temperature (LT) SIRM imparted at 20 K showed a sharp Verwey transition [Smirnov, 2006] (Table 1) for both samples indicating their nearly stoichiometric composition [e.g., Özdemir et al., 1993].
Table 1. Grain Size Distribution and Magnetic Properties of the Test Samplesa
Parameters dlower and dupper represent the lower and upper limits of grain size in a sample, so that subjectively, approximately 95% of all the grains in a sample fall between dlower and dupper. The parameter dmode approximately corresponds to the mode of grain size distribution [Smirnov, 2006]. Hc, coercivity; Hcr, coercivity of remanence; Mrs, saturation remanence; Ms, saturation magnetization. The Verwey transition temperature (TV) is defined as the temperature of corresponding maximum of the first derivative of a LT SIRM demagnetization curve. RLT is the ratio of the RT SIRM recovered after low-temperature cycling to the initial RT SIRM.
For our experiments, the synthetic magnetite powders were mixed with Omega cement (∼1–2% of magnetite in the mixture). We strove to disperse the magnetite powder in the matrix as homogeneously as possible. The samples were made of spherical shape (about 3 mm in diameter) to avoid the effects of sample shape anisotropy on the magnetization direction. Before experiments, a room temperature SIRM was imparted in our samples by applying a 1 T magnetic field for 15 s.
In addition to the synthetic samples, we measured a natural sample (MT, Table 1) of a ∼2.45 Ga Matachewan mafic dike [Smirnov and Tarduno, 2004]. Over 95% of the oxide minerals in this sample were relatively large (several hundreds of microns to >1 mm) grains, containing one or several subordinate sets of trellis type lamellae [e.g., Haggerty, 1991]. Between the lamellae, fine (from submicron to 2–3 μm in size) intergrowths of two phases were observed. SEM and low-temperature magnetic analyses indicated that these phases are nearly stoichiometric magnetite and ilmenite produced by oxyexsolution [Smirnov and Tarduno, 2005]. The sample manifests PSD hysteresis behavior (Table 1) consistent with SEM data.
Finally, we measured a ∼3 mm octahedral single crystal of magnetite (CR), which manifests multidomain magnetic hysteresis behavior (Table 1). The crystal was given an SIRM in a 1 T field along the  crystallographic axis [e.g., Özdemir et al., 1995; Dunlop and Özdemir, 1997]. For the low-temperature experiment, the crystal was oriented so that the  axis was aligned with the magnetometers Z axis, and the  and  axes with the X and Y axes, respectively. However, we cannot rule out a potential orienting uncertainty up to ±5°. The misalignment could have occurred when the crystal was inserted into a gelatin capsule, the capsule was placed into the holder, and/or when the sample holder assembly was inserted into the LT insert.
For all the samples, we measured the behavior of an RT SIRM upon zero-field cycling to 20 K. The synthetic polycrystalline samples were placed into the magnetometer so that the SIRM direction was close to the magnetometer X-Y plane.
The magnetization of the PSD sample (M1.5) monotonically decreased upon cooling from room temperature to TV (Figures 3a and 3b). The demagnetization curve showed no humping, consistent with nearly stoichiometric magnetite [Özdemir and Dunlop, 2010]. At TV, approximately 60% of the initial magnetization was demagnetized and no further changes in magnetization occurred during further cooling the sample to 20 K and rewarming to TV. During warming above TV, the magnetization monotonically increased and about 53% of the initial magnetization was recovered at room temperature (Figure 3a and Table 1). The temperature dependence of a RT SIRM measured with the LT insert is practically identical to that measured for the same sample using an MPMS (Figure 3a). The magnetization direction manifested no significant change during the low-temperature cycling (Figure 3c) with both declination and inclination varying within half a degree (with inclination about twice as variable as declination) (Figures 3c–3e).
In general, the temperature dependence of an RT SIRM measured from the MD sample (M30) was similar to that observed from the sample M1.5 (Figure 4). However, a larger portion of the initial magnetization (87%) was demagnetized at TV and a smaller fraction (33%) was restored after low-temperature cycling (Figures 4a and 4b). The temperature dependence of RT SIRM for M30 was nearly identical to its counterpart measured with an MPMS (Figure 4a). No significant directional change was observed for M30 (Figures 4c–4e) with both declination and inclination varying within half a degree.
The natural rock sample (MT) was given an RT SIRM in a 1 T field after the sample was demagnetized in an alternating field with the initial magnitude of 0.12 T. The initial SIRM direction of the sample was close to the Z axis of the magnetometer (Figure 5c). The low-temperature behavior of the RT SIRM closely resembled the behavior of the synthetic samples (Figures 5a and 5b). The fractions of the initial magnetization preserved at TV and restored at room temperature are 28% and 48%, respectively (Figure 5a). Similarly to the synthetic samples, no systematic variation of magnetization direction during the low-temperature cycling was observed (Figures 5c–5e).
The individual natural magnetite crystal showed the most interesting behavior during the low-temperature cycling. A RT SIRM was imparted in the crystal in a 1 T field along one of its  axes. However, the initial SIRM deviated from the  axis by approximately 20° (Figure 6c). Upon cooling, the total magnetic moment gradually decreased to about 5% of the initial SIRM at TV. The magnetization sharply increased to ∼64% just below the Verwey temperature and stabilized at the ∼60% level upon further cooling (Figure 6a). On warming from 20 K to TV, the magnetization behavior was nearly reversible including the peak just below TV. During warming above TV, the magnetization slightly increased to ∼14% of the initial SIRM at room temperature (Figure 6a).
Interestingly, the magnetite crystal manifested systematic changes in the magnetization direction during the temperature cycling. On cooling, a well expressed steepening of inclination by ∼5–6° (toward the  axis) occurred in the vicinity of the Verwey transition (Figures 6c and 6e). However, this change was not accompanied by any noticeable change in declination (Figure 6d). During cycling between TV and 20K, the direction remained nearly constant. However, on warming through the Verwey transition, the inclination shifted by 2–3° toward the initial direction (Figures 6c and 6e) and this shift was accompanied by a ∼3–4° change in declination (Figure 6d). The magnetization direction remained nearly constant during further warming to room temperature (Figures 6c–6e).
5. Discussion and Conclusion
The low-temperature insert developed and tested at the University of Rochester allowed us to obtain high-quality data on the behavior of a full magnetization vector during low-temperature cycling from a suite of synthetic and natural samples. To the best of our knowledge this is the first data set of this kind ever published in peer-reviewed literature. Despite the limited size of our data set, our experiments convincingly demonstrate that the low-temperature insert at the University of Rochester allows reliable measurement of the full vector of remanent magnetization during low-temperature cycling to temperature at least as low as 20 K.
The low-temperature behavior of the SIRM intensity measured from the polycrystalline synthetic samples is in a good agreement with the observations reported by other authors from similar samples [e.g., Muxworthy et al., 2003]. The observed magnitudes of low-temperature memory (Table 1, RLT) are consistent with the magnetic domain state of the samples [e.g., Heider et al., 1992]. More importantly, the almost identical shape of the temperature dependences of RT SIRM measured using the insert and using an MPMS from the same samples (Figures 3a and 4a) indicates that the insert provides an adequate control of the sample temperature within 1–2°C. Although we do not have the data on the RT SIRM behavior upon low-temperature cycling measured with an MPMS from our natural sample, the magnetization behavior at cryogenic temperatures is consistent with the observations reported from Matachewan and other diabase dikes by Hodych et al. . The value of low-temperature memory (RLT = 0.49) is also consistent with PSD state of the sample.
The absence of significant directional changes of magnetization during low-temperature cycling observed from our synthetic and natural polycrystalline samples is not surprising, especially taking into account the relatively small directional changes observed from our individual magnetite crystal. More importantly, these results unequivocally demonstrate that low-temperature demagnetization does not change the direction of primary magnetization, hence lending strong support to the fidelity of low-temperature demagnetization as a paleomagnetic cleaning technique.
The systematic (albeit small) change of the SIRM direction observed in a large individual magnetite crystal may reflect switching of the easy magnetization direction between the crystallographic axes in the vicinity of the Verwey transition. Although, theoretically, such switching should occur at the isotropic point temperature, our data (Figures 6a and 6b) do not show any clear separation between the Verwey transition and isotropic point similar to that reported by Özdemir and Dunlop  (in their Figure 4). The observed directional behavior could be explained by a relatively significant role of other sources of magnetic anisotropy such as magnetostriction (possibly also accounting for the initial deviation of the imparted SRM from the  axis). Additional experiments are needed to clarify this issue.
The small magnitude of the observed directional changes may be explained by the crystallographic twinning in the monoclinic magnetite. To reduce the spontaneous strain caused by lowering of crystal symmetry below TV, different areas of a magnetite crystal (twin domains) have their monoclinic c axes aligned in different directions [e.g., Chikazumi et al., 1971; Medrano et al., 1999; Kasama et al., 2010]. In the absence of an external field, all three  directions have an equal chance to become the monoclinic c axis. In the twin domains with c axes approximately orthogonal to the field direction for the initial SIRM acquisition, the remanence would be randomized and would not contribute to any systematic change in direction. While further experiments are needed to confirm and refine our observations, the systematic directional changes we report provide a perfect example of the magnetic behavior that would be impossible to observe using the conventional (single-axis) instruments for low-temperature magnetometry such as an MPMS.
The results reported here clearly demonstrate the great potential of directional low-temperature magnetometry. We note that other groups, most notably the Institute for Rock Magnetism (University of Minnesota) have taken up the challenge of further developing and implementing low-temperature inserts for SQUID magnetometers. This new approach will undoubtedly result in breakthroughs toward better understanding of the low-temperature magnetic properties of natural and synthetic materials.
The development of the device here relied heavily on the tremendous support of the late William (Bill) Goree. We also thank Pavel Doubrovine for help with experiments. Support for this research was provided by the National Science Foundation.