Nuclear Instruments and Methods in Physics Research A 618 (2010) 43 47 Contents lists available at ScienceDirect Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima Development of a superconducting transverse holding magnet for the HIgS frozen spin target P.-N. Seo a,b,, D.G. Crabb c, R. Miskimen b, M. Seely d, H.R. Weller a a Triangle Universities Nuclear Laboratory, Durham, NC 27708, USA b Department of Physics, University of Massachusetts, Amherst, MA 01003, USA c Department of Physics, University of Virginia, Charlottesville, VA 22904, USA d Meyer Tool & Manufacturing, Inc., Oak Lawn, IL 60453, USA article info Article history: Received 17 February 2010 Accepted 20 February 2010 Available online 2 March 2010 Keywords: DNP Superconducting magnet Saddle coil Frozen spin target abstract We describe the design, construction, and performance of a set of saddle coils used to maintain the spin of a polarized proton target in the transverse direction with respect to the incident gamma-ray beam direction. The transverse coil assembly consists of two racetrack shaped coils formed from a single strand of thin NbTi superconducting wire. Four layers of NbTi per saddle coil were wet wound in a racetrack shape and then installed on a cylindrical support tube and epoxyed to prevent them from moving when the coils were energized. As expected from our model calculation, a set of two saddle coils produced 0.36 T in the middle of the coils with a 25 A current at 4 K, which is 63% of the critical current of the wire. The measured homogeneity of the field over the target volume has a maximum variation from the average value of 0.6%. & 2010 Elsevier B.V. All rights reserved. 1. Introduction Corresponding author at: Triangle Universities Nuclear Laboratory, Durham, NC 27708, USA. Tel.: +1 919 660 2631; fax: +1 919 660 2634. E-mail address: pilneyo@tunl.duke.edu (P.-N. Seo). High nucleon polarization can be achieved using the dynamic nuclear polarization (DNP) technique in a solid state target [1]. In DNP, the high electron polarization from paramagnetic impurities in the target material is transferred via microwave irradiation to the nucleon. For the DNP process the target has to be in a high and homogeneous magnetic field of 2.5 7 T and at temperatures of about 0.3 1 K. The superconducting wire of the polarizing magnets forms a thick material layer, which scattered particles can only pass through with large energy loss. Moreover, low energy charged particles are bent so that they may never emerge from the magnet. If a solenoid, used to polarize a DNP target, is kept as part of the experiment spectrometer then only particles scattered at small angles within the magnet bore can reach the detectors. For a transverse field the magnet is more complicated and with more stringent requirements on the particle trajectories. Using a frozen spin target, these limitations can be avoided or at least mitigated. In our frozen spin target, protons are first polarized using DNP. After saturation of the proton polarization, the target material is cooled down to sub-kelvin temperatures (50 mk): the polarization decay time is a function of the temperature and magnetic field and is very long at such temperatures. Therefore, the magnetic field can be lowered until an acceptable rate of polarization decay is reached, such that re-polarization is necessary at infrequent intervals. The low (holding) field is obtained from a separate set of coils, typically operating at around 0.5 T; these are thin and allow a much better acceptance for the scattered particles to be detected. In operation, the magnetic field of the polarizing magnet is lowered and at the same time, the field of the holding magnet is raised. When the field of the solenoid is zero and the field of the holding magnet is at the experimental operating point, the solenoid magnet can be removed from around the target. In the HIgS Frozen Spin Target (HIFROST) experiments [2,3] a transversely polarized proton target will be used with circularly polarized incident photons to measure the double spin asymmetry for Compton scattering. Measurements of the double asymmetries allow us to access the presently unknown spin polarizabilities of the proton. We plan to polarize protons via DNP at 2.5 T with a solenoid magnet, a 0.5 T holding coil, and a horizontal dilution refrigerator. The holding coil is maintained inside the dilution refrigerator at helium temperatures. In the HIFROST experiments, a transverse holding field of about 0.5 T is planned. In the experiment the beam is incident on the proton target through the cryostat axis. The polarizing magnet is a 2.5 T superconducting solenoid in its own dewar separate from the target cryostat. The low mass transverse holding magnet was 0168-9002/$ - see front matter & 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2010.02.129
44 P.-N. Seo et al. / Nuclear Instruments and Methods in Physics Research A 618 (2010) 43 47 constructed with two superconducting saddle coils mounted on the isolation vacuum tube of the dilution refrigerator with high cooling power [4]. In this article, we report the design and construction of a superconducting polarization holding magnet and present the results of preliminary tests of the magnet. We discuss possible future improvements to the magnet. 2. Holding magnet for HIFROST 2.1. Magnet design Even using the holding magnet, the material thickness introduced by the holding coil for scattered particles has to be small compared to the material thickness introduced by the several concentric walls of the refrigerator. We therefore chose a thin superconducting wire for the coils. Fig. 1 shows a single wire model of the holding magnet formed by two saddle coils. The radius (a) of the semi-circles at the ends of the saddle coils is defined by the diameter of the coil support tube, the length of the coils (2b) is defined by the length of the cylindrical target, a number of layers in each coil are mainly defined by the maximum field strength and the critical current density of the wire. The homogeneity of the field over the target volume should be no worse than about 10 3. With geometric inputs and parameters of the coils, the magnetic field was calculated by integrating numerically the field using the Biot Savart law for each segment and then the total field was obtained by summing the field of each segment. Each saddle coil was in a four-layer configuration. Optimized numbers of turns for each layer of the coil are 136, 136, 122, and 48 starting from the top layer which becomes the most inner layer when the coils are mounted on the support tube. Using a current of 25 A, 17.5 cm length of the coils, and other constraints, the field strength at the center of the target was calculated to be 0.362 T. The field profiles are shown in Fig. 2. According to the design, a total of 442 turns of the wire produces a force between the coils of about 707 N. Assuming that this force is 8 4 -b y 2 5 x Fig. 1. A single-wire model was used to calculate the field profiles of the holding magnet formed by two saddle coils. Here the beam direction and the field direction are along z and x, respectively. 1 6 3 b a 7 z Fig. 2. Using the model the field profiles of the holding magnet formed by two saddle coils are calculated and plotted along the x-, y-, and z-axis (top). The outer diameter of the coil support tube is 4.05 cm and a 25 A current was used. The 2-dimensional plot (bottom) shows the field in the x y plane in the center of the target.
P.-N. Seo et al. / Nuclear Instruments and Methods in Physics Research A 618 (2010) 43 47 45 spread uniformly over the dimensions of the inner coil, the pressure on the coil support tube is 16.3 psi. A more detailed calculation using a wire model gives a total pressure difference of 13 psi (radial component) between the bottom and the top of the coil. This force could cause a small distortion on the 0.5 mm thick stainless steel (SS) tube on which the coils are mounted. The motion of the coils or wire could lead the magnet to quench. To reduce the possibility of motion of the coil, the support tube must be made thicker (1mm wall thickness). The calculated inductance of the magnet is 1.58 mh and the stored energy at 25 A is 0.5 J. 2.2. Magnet construction A 177-mm diameter niobium titanium (NbTi) superconducting wire stabilized by a copper matrix [5] was chosen for the coils. The wire is bonded in an epoxy matrix to prevent movement under the Lorentz force when the magnet is energized; even an imperceptible movement of a wire can cause sufficient frictional heating that quenches the magnet. The characteristics of a good superconducting magnet are high wire-packing density and a minimum of epoxy-rich or void regions. The racetrack shape coils are difficult to wind because wire tension during winding does not naturally translate into a compressional force on previous layers to increase wire packing and reduce epoxy rich regions. Therefore, wires in the straight sections of the coils have to be compressed after winding, which often leads to epoxy rich regions at the ends of these sections. Since the epoxy is relatively soft, this can lead to motion of a wire and consequently to a quench of the magnet. We established the coil winding system at TUNL and attempted a new technique of winding a pair of saddle coils. All coils were wound on a rotating welding table. The basic idea to make the saddle coil a racetrack shape in multiple layers was to wind a coil around stacked plastic shims which were precisioncut and the straight sections of the coil were pushed into place as they were wound. The thickness of the plastic shims was equal to the diameter of the wire. The ends of the coils were turned on a radius to match the spacing of the straight sections resulting in a racetrack-shape coil. Fig. 3 shows a schematic showing the basic construction of a single saddle coil. Fig. 3 (left) is a top view and Fig. 3 (right) is a cross-sectional view of a wound coil in four layers around the shims. Each layer has inner lengths of [L1, L2, L3, L4] ¼[13.4,13.4,14.0, and 17.15] in cm, distances between turnaround centers of [l1, l2, l3, l4]¼[12.1,12.1,12.3,13.1] in cm, and distances between straight sections of [a1, a2, a3, a4]¼[1.4,1.4,1.9,4.4] in cm. In the middle of each shim, there are three 1.27-cm diameter holes to allow the shim to be held on the winding fixture. Winding was performed on a fixture, which consists of three rigid Teflon-coated aluminum plates that were mounted on the rotating table. First we applied vacuum grease on the surface of the center plate to allow the removal of a completed coil without breaking the wire. We then stacked four plastic shims together on the plate. The center plate rested on screws which were threaded through each corner of the lower plate. These screws allowed the gap between the upper and center plates to be adjusted to be a total thickness of four shims. The gap between the two plates needed to be very carefully adjusted so that the wire did not get in between the two layers. Then, the coils were wound around the shims in four layers using the Formvar insulated NbTi wire which has a copper to superconductor ratio of 1.13:1 [5]. As modeled, the number of turns from top to bottom layer in Fig. 3 (right) was 136, 136, 122, and 48, respectively. We had to be careful to add enough epoxy to hold the coils together but not so much that there was a lot of l4 L4 a4 a3 a1, a2 Fig. 3. Schematic design of plastic shims for a single racetrack coil (top) and a cross-sectional view of a wound coil in four layers around the stacked shims (bottom). The number of turns per layer shown on the right is chosen only as an example. Fig. 4. During the cure of the epoxy the straight sections of the coils were compressed by plastic bars in white to help minimize epoxy rich regions (left). The completed set of the racetrack-shaped saddle coils in a single strand of wire was backed on the micro-polyester mesh with the epoxy (right). excess. We started wet winding, which means that we applied the epoxy while we were winding. The epoxy was an epoxy-resin with lower hardness when cured and low viscosity. We accurately weighed the resin and hardener in the recommended ratio [6]. After blending the components by hand for 2 3 min, the mixture was put into a small vacuum vessel and pumped for about 10 min to remove any trapped air introduced during the mixing operation. After completing the first coil of the pair, we continued winding the second one. The center slot in the top plate runs across both sides of the plate to allow the completed one section of the pair to be passed through the plate and thus avoid the need for a joint between the upper and lower halves of the coil. As the epoxy was cured over 2two days after winding, the two straight sections of the coil were compressed by plastic bars to help minimize epoxy rich regions as in Fig. 4 (left). Fig. 4 (right) shows a completed set of two racetrack coils after removing the plastic shims. The major design difference from previous trials lies in the different pre-stressing technique for the coils. The reason for
46 P.-N. Seo et al. / Nuclear Instruments and Methods in Physics Research A 618 (2010) 43 47 holding the racetrack coil assembly under tension is to minimize any movement of these coils due to current-induced forces when the magnet is energized. This set of coils backed on the micropolyester mesh were glued together with the epoxy. The coils were then wrapped around the stainless steel support tube using a rough nylon thread to hold them to the carrier. 2.3. Magnet test The transverse holding magnet for HIFROST was tested at TUNL. A cryogenic insert to suspend the coils mounted on the tube in a liquid helium (LHe) dewar for the test was designed and built at TUNL [7]. The insert consisted of two stainless steel (SS) flanges connected by three thin and smaller diameter SS supporting tubes. The tube on which the coils were mounted was attached to the lower flange, while the upper flange formed a seal with the top flange of the dewar. Two brass tubes with 0.624 cm diameter and with wall thickness of 0.082 cm were used to carry a current from room temperature to liquid helium temperature through a Teflon bushing on the top flange. During the first cool down the two current wires of the magnet were attached with screws on the cold ends of the brass tubes. The brass tubes were electrically insulated from the insert using Kapton tape wrapped around the tubes. At room temperature the measured resistance of the magnet circuit was 483 O across the brass leads on the top flange of the insert. Outside the insert a 12-guage solid copper wire was used to carry currents from a 40-V/50-A power supply to the brass current leads. A 91.4-cm long helium level sensor was used to monitor the LHe level in the dewar. The surface of the LHe was kept several inches above the superconducting normal conducting joint. The resistance of the coil system was monitored during the cool down to see when the magnet reached the superconducting state. A transverse hall probe with a transverse field sensor was positioned along the axis of the coil in the center of the top flange of the insert. The hall probe was initially positioned with the active end roughly at the midpoint of the magnet. The distance from the midpoint of the magnet to the top of the hall probe was measured, and the hall probe marked at that distance from its lower end for reference. During the initial test of the magnet, a quench occurred at 20 A, and resulted in damage to the magnet. The quench apparently started in a part of the superconducting leads connecting the two saddle coils, since a section of superconducting lead was observed to be vaporized, as were several layers of wire in one of the coils where the lead entered the coil. Since the lead entrance point into the saddle coils was identified as a weak point, the connections between the brass tubes and the superconducting wires were improved. As shown in Fig. 5, a one-inch long end of the superconducting wire, after the isolating coating was removed, was soldered onto the surfaces of a 0.5-mm thick OFHC Cu foil using 60/40 Sn/Pb solder alloy. Heat conductivity from the brass tube to the NbTi wire through an OFHC Cu connector and the Cu foil is gradually decreasing because the cross-sectional areas gradually decrease. 1.2 8 mm Cu connector A bottom piece of the Cu connector was soldered onto the inner side of the brass tube after cutting out a section that is one half of the 0.8-cm long tube. A very thin indium layer was inserted between the top of the Cu connector and the Cu foil, and also between the bottom of the Cu connector and the Cu foil, and then all the parts were held together with a screw as shown in Fig. 5. The wire on the surface of the foil allows the current from the copper foil to flow to the copper matrix of the superconducting filaments and then to filaments of the NbTi wire through a large cross-section. Since most of power (I 2 R) loss was caused by the resistance of the brass tube, we carefully monitored the LHe level in the dewar with respect to the position of the joint in order to minimize the amount of the brass tube submerged in the LHe. Even with the electrically improved joint, we had two quenches and in both events one of the magnet lead contacts was broken when the LHe level went below the joint. However, with this improved superconducting normal conducting joint, the current was successfully ramped up incrementally to the design current of 25 A and produced 0.363 T at the center of the magnet. Fig. 6 shows the measured field vs. supplied currents and demonstrates the linearity of the current field relationship. B (T) 0.4 0.3 0.2 0.1 0 0 10 20 30 I (A) Fig. 6. Measured field in the center of the magnet as a function of the applied current after the improvement of the magnet lead joint. B (T) 0.25 0.20 0.15 0.10 0.05 Measurement Calculation Cu foil NbTi wire 8 mm Brass tube Fig. 5. Connection between the NbTi wire and the brass tubes. -6.0 0.00-4.0-2.0 0.0 2.0 4.0 6.0 Distance of the probe from the center of the coil (inches) Fig. 7. Measured field profile at 14 A along the coil axis (z-axis) are compared with the results of our calculation.
P.-N. Seo et al. / Nuclear Instruments and Methods in Physics Research A 618 (2010) 43 47 47 We also mapped the region over the target volume at about one half of our maximum testing current, 14 A. The measured field profile along the coil axis is shown in Fig. 7. As expected, the field is flat around the center of the magnet and then drops as the probe is positioned farther from the center. The maximum variation in the homogeneity over the expected target length, about 10 cm along the axis, is 0.6% from the average. The variation from the center out to 5 cm, which is the size of many targets for this type of experiment, is 0.11%, which is close to our calculated value of 0.09%. Stability of the field over a longer time was also observed. During 30 min with the current of 18 A, the change in the field was less than 0.01%. 2.4. Discussion The transverse holding magnet for HIFROST produced 0.363 T with 25 A at LHe temperature. The measured field strength agrees with our calculated value to 0.5%. We used the measured critical currents of the wire at different magnetic fields provided by the manufacturer [5] and extrapolated these values to our value of the magnetic field strength using the equation of Ref. [8]. With our design current of 25 A, the critical current density in the wire was found to be 3931 A/mm 2 corresponding to 40 A. Based upon the linearity of the measured fields vs. currents in Fig. 6, our magnet will produce 0.5 T at 35 A. At the present test the performance of the magnet at LHe temperature was limited by the power supply and the absence of the quench protection circuit. However, an increase in current density would make the magnet more prone to quenches. Without an effective quench protection, an increase of the stored energy also increases the probability of permanent damage to the magnet in a quench. This concern will likely be the limiting factor of the magnet performance at lower temperatures. The magnet can potentially perform even better if operated at higher current, which requires the quench protection system. In anticipation of the construction of the cryogenic system [9] for the HIFROST experiments, reconstruction of the system is currently underway at University of Virginia [10]. One of the most promising techniques for achieving the highest field is a hybrid design, where the inner coil is made from high temperature superconducting (HTS) material where the field is highest and the outer coils are made from LTS material Nb 3 Sn where the field is somewhat lower [11]. With the rapid development of second generation HTS materials and magnet manufacturing technology, a magnet made from a pair of saddle coils with higher than 0.5 T may become possible. With the higher current required to produce the higher field, cooling of the leads becomes more problematic, as does the connection between the normal and superconducting states. The most serious problem is the forces between the coils. During the HIFROST experiments where no liquid helium is present in the leads joint, the leads will be cooled by a convection of lowpressure 4 He vapor pumped out of the 1 K pot because the 1 K pot has inadequate thermal conductance to keep the leads cold with the power loss. The temperature in the leads region is expected to be o4 K, which is low enough for thermal conductance. This cooling scheme has previously been operated successfully in the frozen spin mode [4,9], where a 20-A current did not increase the pot boil-off or the pressure in the still even at the lowest boil-off rates. Because of the dependence of the magnetic field and the temperature in the thermal equilibrium spin polarization mode, a small holding field of 0.36 T requires lower temperature operation of the dilution refrigerator. In the holding mode a temperature of about 55 mk in a proton target is needed with enough cooling power to compensate the various heat leaks. The relaxation time for a 0.5 T transverse holding field at 55 mk target temperature was measured to be 30 days [9,12], which allows us to polarize the target every 2 3 days. This result is based on a measurement using a butanol target and it may be different with a polarized scintillating target [3], doped with TEMPO for dynamic polarization at 2.5 T. 3. Conclusions We have developed the transverse superconducting holding magnet for HIFROST with low inductance and increased field strength in a new design. The magnet was tested using the temporary cryostat designed for this purpose. After the first quench of the magnet, we improved the magnet lead joint to make the thermal gradient less steep between the superconducting wire and the brass tube which carries the current from room temperature to LHe temperature. We observed that it was critical to keep the magnet lead joint cool to prevent the magnet from quenching. At LHe temperature, the measured field strength in the central region of the magnet is 0.363 T at the design current of 25 A, which agrees with our calculation within 0.5%. The TUNL transverse holding magnet will be further developed using a quench protection circuit and is expected to reach 0.5 T at 35 A, which will meet the requirement for the HIFROST experiments in order to get reasonable relaxation times at 55 mk when a scintillating target material doped with TEMPO is used. Acknowledgements The authors would like to thank C. Keith and J. Brock at Thomas Jefferson Laboratory National Accelerator Facility for equipments and helpful discussion, P. Martel for his help and M.W. Ahmed for his support and C. Westfield at TUNL and B. Jelinek at the physics department machine shop for their technical help. This work was supported in part by the Department of Energy under Grant no. DE-FG02-88ER40415 A025. References [1] D.G. Crabb, W. Meyer, Annu. Rev. Nucl. Part. Sci. 47 (1997) 67. [2] H.R. Weller, et al., Prog. Part. Nucl. Phys. 62 (2009) 257. [3] R. Miskimen, et al., the High Intensity Gamma-ray Source PAC-09. [4] T.O. Niiniikoski, Nucl. Instr. and Meth. 97 (1971) 95. [5] Supercon Inc., Shrewsbury, MA 01545, USA. [6] Emerson & Cuming, Billerica, MA 01821, USA. [7] P. Kingsberry, et al., TUNL Progress Report XLV 2005-06, p. 163. [8] P.A. Hudson, F.C. Yin, H. Jones, IEEE Trans. Magn. 17 (1981) 1649; Y. Iwasa, M.J. Leupold, Cryogenics 22 (1982) 477. [9] H. Dutz, et al., Nucl. Instr. Meth. A 356 (1995) 111. [10] D. Crabb, et al., TUNL Progress Report XLV 2005-06, p. 159. [11] J. Cozzolino, et al., Presented at the Applied Superconductivity Conference, Houston, TX, 2002, unpublished. [12] H. Dutz, et al., Nucl. Instr. and Meth. A 340 (1994) 272.