Test results for a Nb/sub 3/Sn dipole magnet

A cosine theta type dipole magnet using Nb/sub 3/Sn conductor have been designed, built and tested. D19H is a two-layer dipole magnet with a Nb/sub 3/Sn inner layer and a recycled NbTi outer layer. Coil-pairs are connected with two of the four Nb/sub 3/Sn splices in a high field region, and compressed by a ring and collet system. The ramp-rate sensitivity and the splice resistances were pleasingly low; and the 4.4 K training was rapid. At 1.8 K, however, the unusually high frequency of outer-coil fast-motion events increased with current, effectively creating a training-ceiling at 90% of the expected outer-layer limit (10.2 T). A low end-load applied to a relatively fluffy outer layer is believed to have caused this training limit. The end-load was increased; but a retest was aborted after the magnet failed a precautionary hipot test.


INTRODUCTION
A major goal of the Superconducting Magnet Program at the Lawrence Berkeley National L,aboratoly is the advancement of magnet technology for accelerator magnets. As it is presently necessary to work with brittle materials for B > 10 T, Nb3Sn is currently being investigated as a prototypical brittle superconductor. Two different Nb&, 50 mm (bore), 1 meter long dipole designs are being evaluated.
A four-layer Nb3Sn design (D20), is under construction, in an attempt to explore the 12-14 Tesla range. Its details have been published previously (2,3). The magnet has the following principal goals: (1) reach a significantly higher Bfield, (2) validate the technology required to reliably achieve these field levels, (3) provide a facility to test insertion coils made with other high-field superconductors.
A second magnet @19H) was constructed to provide a faster test-bed for the new Nb3Sn magnet technology. It utilized the two-layer "D 1 ! Y design that had been previously developed at the Lawrence Berkeley National Laboratory to reach 10T using NbTi at 1.8 K. This design was modified to replace the inner layer conductor with a Nb,Sn cable. Its details are published elsewhere in these proceedings[ 11. This magnet has the following goals: (1) demonstrate efficient production of Nb3Sn magnet coils (once the tooling and magnet structure have been developed), (2) evaluate the feasibility of ring and collet precompression of Nb3Sn coils,

A. Instrumentation
Each Nb3Sn coil had 15 voltage-taps, positioned in a manner to monitor four anticipated trouble regions: the pole turn (highest B-field), each of the three wedges (highest discontinuity-related strain), the mid-plane turn (highest stress), and each splice (joint resistance). The outer coil had only full-coil taps. All voltage-tap signals were processed by analogderivative amplifiers (1V output for lOOV/s input). "Fast-Imbalance" signals were produced by nulling a coil section against its closest neighbor. These signals assisted the determination of quench onset.
Each Nb3Sn/NbTi splice was soldered under pressure with a eutectic solder for a distance of 100 imm. A twisted pair of wires was attached across each joint and monitored with an HP3458A voltmeter. A thermometer was also embedded within each splice-block to monitor any temperature rise.
Four strain gages (onekoiVlayer) monitored the pole-turn pressures; and load-cells (sidend) monitored the end-force applied by each of twelve end-load loading-bolts.

B. Test procedures
The temperature and resistance of each coil was monitored during cool-down and warm-up. The strain gages were also monitored to determine the timing of the cool-down loading losses. When 4.4K was achieved, the magnet was ramped to quench at a slow rate (5A/s), until we later discovered that 2ONs was equivalent for training purposes.
After establishing the existence of a thermally-limited plateau, the ramp-rate sensitivity was measured. After cooldown to 1.8K, the above sequence was repeated.
After warming the magnet, both end-loads were increased substantially, in preparation for another round of tests. Testing was aborted when the magnet failed a standard 700V hi-pot and quickly degraded to 2 1OV (considered inadequate to safely extract energy from the magnet during quenching).

A. Coil Resistance and Strain-Gage Histories
Two different superconducting transitions were easily visible at the expected temperatures pig. 1). Both outer (NbTi) coil resistances decreased (RRR = 75) from their initial values (each 136 mohms). The inner coils differed by 4% at 300K (100 mohms, 96 mohms). This relative difference increased during cooldown to 0.33 mohms and 0.24 mohms, prior to transition. This yielded significantly different RFtR-values (300 and 400, respeztively). The strain-gages also changed during cool-down (Figs. 2,3), showing an anticipated relaxation everywhere. The leadend almost completely unloaded. During training, the outerlayer pole pressures decreased (Fig. 4), while the inner-layer pole-pressures and the end-loads (Fig. 5) increased. After warm-up, all strain-gages indicated higher stress-states than before cooldown. Two end-load increases were made in preparation for further testing. Testing was aborted when the magnet failed a standard hi-pot test.

B. Magnet Ramping
During ramping, the magnet transfer function slope was measured to be 1.13T/KA initially, reducing to 0.81 T/KA above 6T. The outer poles (Fig. 4) unloaded symmetrically and nearly linearly with the square of the magnet current (excepting the first ramp where some jumps were observed). The inner poles were also symmetrical, decreasing nonlinearly above 8 Tesla. The end-loads increased non-linearly (Fig. 5 ) at approximately 15% of the total Lorentz end-load. The I2 dependencies were more linear for the outer layer than the inner layer, and all pole-pressures decreased more rapidly above 8 KA.
Although the inner layer did unload the fastest, it did not unload as fast or as much as expected.
1 .1 (KAV) Fig. 5. End Forces: These four end-load-cell signals illustrate the axial asymmetry of the pre-load (RE=Retum-end, LE=Lead-end) and the relative axial symmetry of the nonlinear I' dependence. The rate of total end-load variation accounted for only 10-20% of the calculated Lorentz end-load.
The first 4.4K quench (Fig. 6) occurred at 7.4T (90% of plateau). Plateau was achieved after six outer-layer (NbTi) quenches, one of which fell-back. The 1.8K training began at current-ramp #26, with inner-layer (Nb3Sn) quenching at 80% of the expected maximum (E910.2T). The rate of improvement was significantly slower, and was eventually dominated by outer-layer (NbTi) quenching after ramp # 46. The frequency of "fall-backs'' was also high. Training was aborted in order to measure the 1.8K ramp-rate sensitivity. Testing was aborted to warm the magnet and increase the end-load.
Quench onsets were most easily identified with the "Fast-Imbalance" signals ( Fig. 7, 8). The early, 1.8K, inner-layer training quenches (# 26-46) seldom showed the fast-motion signatures normally observed at quench onset during training. They also showed significantly (lox) slower quench propagation speeds than the training-limiting outerlayer quenches (Fig. 8). The number of Fast-Imbalance events increased rapidly with current during ramping (Fig. 9). At 4.4 K, most of them occurred shortly before the quench, the number reducing with training. At 1.8K, the rate of increase was more linear, starting at 5 KA. No decrease with training was observed above 7 KA. The Fast-Imbalance history (Fig. 10) also illustrates the opposing trends of the two trainings. At 4.4K, the total number generally decreased, even though the current increased, while at 1.9K the total number increased with the current (and fell back when the current fell-back). The pole-turn (high-field) splice resistances were equal (1 .O n-ohm), nearly independent of current (I<9KA), and three times the average: mid-plane (low-field) splice resistance. Embedded thermometers in their copper splice blocks showed only small splice temperature rises (0-15 mK, during the 0-9 KA ramp shown in Fig. 11). One mid-plane splice showed a small ramp-rate dependence, while the other mid-plane splice showed no heating at all. One pole-splice showed a steeply rising current dependence prior to magnet quenching, while the other pole-splice, using a germanium resistor, gave unreliable temperatures in the magnetic field. The voltage break-down discovered after warm-up was localized near the lower coil's inner-to-outer-layer bgh-field splice. Testing was terminated until this condition could be corrected.

IV. DISCUSSION
The early 1.8K inner-layer mid-plane training quenches, with several showing simultaneous upper and lower coil quench-starts, suggests the possibility of some mid-plane slippage and/or mismatched mid-plane surfaces.
The fast-motion-initiated outer-coil training-ceiling and the persistently high frequency of outer layer fast-imbalances is strong evidence for inadequately restrained outer layer coils. This is believed to have been aggravated by the failure to stretch the outer layer during collaring (as was done for D19A), in addition to the conservatively low total end-load that had been rigidly applied to two coil-layers having markedly Werent stiffness [l]. The un-potted outer coil appears to have been so loose as to preclude fast-motion training, a prerequisite for increasing the quench-current.
The radial loading from the outside shell apparently provided enough friction that only a fraction of the Lorentz load is delivered directly to the end-structure. The nonlinear end-gage response suggests that some of this friction was increasingly overcome as the Lorentz load increased. The larger relative cooldown loss of the lead-end is believed to have resulted from its relatively larger plastic fraction, caused by epoxy-filling twice as many end-padcable gaps.
The low ramp-rate dependence suggests fairly good interstrand isolation, until a 200-250 A/s punch-through occurred.
The extremely slow Nb3Sn quench propagation speed supports the existence of a large relative margin.

V. CONCLUSION
A Nb,Sn/M>Ti hybrid magnet was tested to validate several portions of our growing body of Nb3Sn magnet technology. The pleasingly low ramp-rate sensitivities and low splice resistances validated the coil-reaction and splice technologies used for the upcoming magnet D20. The low Nb3Sn quench velocities support the high, anticipated margin of the inner conductor. A NbTi outer-coil "trainingceiling" unfortunately limited our ability to extend our observations to the 10.2 Tesla outer-coil limit. An unusually high frequency of outer-layer fast-motion events did not decrease w i t h training (as usual). This leads us to suspect that looseness in the weakly loaded outer coil-ends limited the stress-range of this test. This could be corrected by substantially improving the packing and loading of the outer coils. However, the cost/benefit ratio of this change alone does not look very attractive with our present commitments,