Lawrence Berkeley National Laboratory Lawrence Berkeley National Laboratory

Transcription

1 Lawrence Berkeley National Laboratory Lawrence Berkeley National Laboratory Title Performance Comparison of Nb3Sn Magnets at LBNL Permalink Athor Chiesa, L. Pblication Date escholarship.org Powered by the California Digital Library University of California

2 SC-MAG#722 LBNL Performance Comparison of Nb 3 Sn Magnets at LBNL L. Chiesa, S. Caspi, M. Coccoli, D.R. Dietderich, P. Ferracin, S.A. Gorlay, R.R. Hafalia, A.F. Lietzke, A.D. Mcintrff, G. Sabbi and R.M. Scanlan Abstract-The Spercondcting Magnet grop at Lawrence Berkeley National Laboratory has been sccessflly developing Nb,Sn high-lield dipole magnet technology for the last ten years. Noteworthy magnet tests inclde D20 (SOnn bore, 4-layer coss, 12.8 T, accelerator qality dipole), and recent racetrack dipoles: I) RTI (2-layer, 12 T, no bore, no training), 2) RD3b (3-layer, 14.7 T, 10mm bore), 3) RD3c (3-layer, 10 T,low-harmonics 3Smm bore), and 4) some small NbJSn magnets that tilized new technology. The performance of these magnets is smmarized, comparing 1) cable and magnet geometry parameters, 2) training behavior, 3) ramp rate sensitivity, 4) RRR measrements,s) peak temperatres and voltages, and 6) fast flx adjstments that occr dring ramping. Index Terms-Spercondcting Magnets, NbjSn, CosS, Racetrack, Comparison. THE I. INTRODUCfION spercondcting magnet grop at the Lawrence Berkeley National Laboratory has been developing highfield, spercondcting Nb1Sn magnet technology for ftre accelerators. Over the last ten years. improvements in condctor properties reslted in a significant increase of the maximm dipole field. At the same time new design concepts and fabrication techniqes allowed redcing the magnet cost. D20 demonstrated that Nb1Sn condctor cold be sed in an accelerator qality cos8 dipole [I]. A desire for more cost effecti ve magnets motivated the exploration of flat racetrack common coil magnets. Recent efforts have focsed pon assessing their ability to achieve higher fields, at less cost, while retaining the reliability and field qality of cos8 magnets. RTI showed exceptionally fast training in a 12 T test of resable racetrack mod les with improved condctor [2l, [3]. RD3b [4]-[6] verified that a high field insert-coil cold be made with this condctor and sed at 14.7 T withot degradation. RD3c tilized a larger bore insert modle with a flat racetrack coil for field qality improvement [7]. Meanwhile, a Sbscale Magnet Test Facility has been developed to investigate innovative approaches to mechanical. condctor and fabrication isses in the T field range Manscri pt received Agst 6, This work was spported by Office of Science, High Energy & Nclear Physics. Div. High Energy Physics. U.S. DOE. Contract DE-AC03-76SFO0098. All the athors are wi th Lawrence Berkeley Nat ional Laboratory, CA USA ( ; fax: ; Ichi esa@lhl.oov, AFLi etzke@ lbl. oov). before applying them to a fll-scale magnet [8], [9J. In this paper, several important featres of these Nb,Sn magnets, and the reslts that have been achieved ths far are compared. Recently, interest has developed in Nb1Sn cos8 qadrpoles for the LHC pgrade [10]. II. MAGNET DEStGNS A. General Feat res The magnets were all assembled from vacm-epoxyimpregnated coil modles. To avoid an internal splice, all modles were wond as doble-layers. Table I shows the parameters for all magnets. TABLE I GEOMETRIC FEATURES AND GENERAL PARAMETERS MAGNET 020 RTi RD3b RD3c SM-Ol Geometry CosO Race- Racetrack track track track Race Race NeoUs Leoil.!!!!! Dborc-hole (nmt! Dsc-frce (mm) Imax (ka) Dbore (T, 4.3K) llpcnk (T. inner) t llpcak (T, oter) Ntrns linncr) Ntrns foter) Indctance (mil) Energy (0)!.I < 20k! * Calclated vales considering no degradation are 11.9 ka and 10.9 T. Training was aborted to assess frther training improvements with a later thermal cycle. D20 is a cos8 dipole, assembled from two doble-layer coil modles. Wedge-shaped spacers were lilized to provide accelerator field qality and compensate for the partially keysloned cable. In contrast, in the RD (Racetrack Dipole) and SM (Sbscale Magnet) magnets, the coil modles were all constrcled with (doble-layer) racetrack windings. No spacers were sed in these coi ls, nlil the attempl 10 improve the field qality with an insert in RD3c. The magnets also differed in their bore diameters (Dborehole), and condctor-free distances near the bore (Dsc-free), parameters that inflence the bore's field magnitde and field qality. D20 and RD3c had bores that were large enogh to insert a warm rotating coil field probe. While the Roman arch

3 2LCOl 2 geometry of D20's coils was self-spporting, the racetrack coils reqired bore-plates that were rgged enogh to keep the bore holes from yielding nder the coil pre-stress. There were also large variations in stored energy and indctance. Stored energy for the D20 and RD magnets is of the same order, despite a factor three difference in the indctance, de to the difference in short sample crrent. For the sbscale magnets (SM series), the stored energy is mch lower, allowing fast qench recovery dring the test with low helim consmption. B. Cable Parameters Table II smmarizes the strand and cable parameters of these magnets. In all cases, the cable packing factor was arond 85%. The cable inslation (abot 0.125mmltrn) was a woven fiberglass sheath arond the bare cable that was epoxy impregnated after winding and reaction. TABLE II STRAND AND CABLE PARAMETERS MAGNET 020 RTI RD3b RD3e SMOI Jc (A/on 2 ) 960, (l2t,4.2k) , Jc (Alon l ) 2240, (lzt, 4.2K) , No. strands No. trns CiSC 0.43, , , Strand diam (mm) Thickness (mm) Widtb(mm) Pitch length (mm) Cable vales are an average of the known strand vales... When two rows exist. the pper row is associated with the inner mod le (if it existed). The boltom row indicates oter modle vales.... The second of two vales separated by comma refers to an idenlica l coil with a different condctor. Note that RD3b and RD3c sed the same oter coils, one of which was also sed in RTl, demonstrating the reliability of Nb 3 Sn coils. There was a significant improvement in J, between D20 and RD3c's strand. The ltimate J, target for the Condctor Development Program is 3000 Almm' (12 T, 4.2 K) [II]. Another goal is to enhance the overall crrent density by redcing the copper to spercondctor ratio and decreasing inslation thickness. C. Magn et Protection The magnet prolection system is triggered when the halfmagnet vollage imbalance exceeds a preset threshold (abot V). Two qench protection healer power spplies were fired shortly after (25-40 ms). The heater geometry was determined by the desire to keep the peak coil temperatre below 200 K, and the peak internal voltage difference less than 400 V. The heater efficiency depends on how fast the vollage threshold is reached by the qench front, and on heat diffsion from the heater into the cable throgh the kapton inslation. After a delay, sfficient to observe the heater effect, the crrent extraction power spply is fired to accelerate energy removal. This delay varied according to the experimental need (qench propagation stdies, or protection). In fll scale magnets, each coil layer had an actively heated, Kapton-encapslated qench-heater foil that was palled, and in intimate contact, with the coil. This foi l had a geometry that cold directly heat more than 70% (average) of the trns in the straight section. Being a distribted heater, the active part was only 40% for 020 and 35% for the RD series (average). With sfficient applied energy, this cold initiate a global qenching within IOms. The set of heaters were wired in two series circits 10 minimize the internal magnet voltage differences in case of failre of one PS circit. The energy dmp resistor was adjsted to keep the magnet's terminal voltage below 500 V. In order to redce cost and assembly time the sbscale magnets did not have qench heaters and the short UR decay times made their elimination reasonable. A. Training History III. TEST RESULTS Fig. 1 shows the 4.4 K training histories of all the magnets. J -" "-!j 8.OO OOO.' <>......,g " 0 IJ J -5 0 c.' O! o o..... TRAINING lllstory,-. : x:x,,"xxx 00 '<>o... eo. r "'... L.. T H ee o.eoooo. -o " 20 " Qench Nmber RTI.. ROll> Ie SM.(ll. : SM-O l b - R03,',,', 30 "..., Fig. 1. Training history: nonnalized qench cultent vs. qench nmber. All bt two of the magnets (RT! and SM-OI b) exhibited slow training, reqiring between qenches to reach the expected condctor limit. D20 reqired several qenches at 1.8 K to achieve its 4.4 K condctor limit [I]. RD3b reached its short sample after 40 qenches, bt not consistently. Below 13.7 T, RD3b experienced most of its training qenches in the virgin inner modle. Above this vale, most of the qenches started in the previosly tested oter coil modles and Ihe highest field of 14.7 T was recorded. Mosl of RD3c's qenches were also in these same oter coils, while its virgin insert coil experienced only two qenches. This modle was only welded and no vertical and axial pre-tress was applied. On the inner modle of RD3b the pre-stress was 56 MPa vertically and 25 MPa axially. RD3c reached a platea at 90%

4 3 of its short sample dring its thermal cycle. SM-O l a experienced troble exceeding 95% and its thermal cycle was aborted to change the pre-stress. SM-Olb was assemble sing the same coils and lower pre-stress [8]. Only RTi and SM-Olb performed well and reached their short sample limit within a few qenches. Both magnets had low horizontal pre-stress so that coil-coil separation occrred at 50% of maximm field. B. Ramp Rate Sensitivity The 4.4 K ramp rate sensitivity of these same magnets (Fig. 2) shows a relatively weak dependence at low ramp-rates [I], [3], [6], followed by a "cliff-like" drop to a flat platea as the ramp-rate increases. The fll-scale coils experienced their drop-off between 3-5 Tlmin in the bore-field. In the small scale coils, the drop-off was less clearly defined. c U V) r'..c 0.2 g QI RAMP RATE SENSITIVITY <r-(> RTl RD3b _ SM-Ola... RD3c 0 1 I I! I!! I o ]0 15 W n " ID Ramp Rate (TeslalMinte) Fig. 2. Ramp-rate sensitivity of the maximm magnet crrent. Normalized qench crrent is plotted as a fnction of the ramp rate of the bore field. C. RRR measrements Table III lists the 20K Residal Resistivity Ratios (RRR = R(300K)IR(20K)) that were measred for these magnets. D20 had higher RRR's than most of the RD-series, with one that was extremely high whose condctor was prodced from a different manfactrer (IGC) than the other three (TWC). The high J, condctor in the RD-series exhibited lower RRR's. TABLE III 20K RRR MEASUREMENT MAGNET Coil RRR pper inner D20 pper oter 44.9 lower inner 47.5 lower oter 51.6 oter RTI oter oler I 13.4 RD3b oler inner 3a 3 inner 3b oter RD3c oter inner SM-O! SC-OI 39.7 SC This was de to the dration of the high-temperatre reac tion cycle, which was case of poisoning of copper with tin that penetrated the diffsion barrier. The relatively low RRR's appeared to have no serios effect on performance bt some changes were observed. In particlar, larger qench voltage signals, shorter trn-trn delay times, decreased LIR crrent decay times, and no significant effect on Ihe peak coil temperatre. The larger RRR's in recent racetrack coils have correlated with shorter high-temperatre reaction times. D. Peak Temperatres and Voltages Peak lemperatres have been determined for D20, RTI and RD3b, si ng the coil segment resistances, and the T(R) calibration measred dring cool-down. The highest temperatre observed in D20 was 205 K. For RTi and RD3b the highest measred temperatre was 215 K. The measred peak voltages did not reach vales higher than 250 V. This voltage is well below the critical threshold of 400 V since in all Ihe qenches both heaters and dmp resistors were sed. E. Fast Flx Challges(FFC's) The time-derivative of the voltage developed by some portion of a magnel circit provides a freqency-weighted view of the flx changes experienced by that circit. Dring short sdden movements of a crrent-carrying condctor, it provides an easy way to visalize and qantify the relative acceleration experienced by the condctor and the size of its movement. Assming the Lorentz force always moves a crrent-carrying condctor toward a larger net enclosed flx, the polarity of a signal prodced by electrically nlling one part of a magnet against another allows the search for movement locations with a signal that is relatively insensitive to external noise sorces. A detection threshold allows collecting only those events above a preset level of motion. An examination of signals collected while ramping a magnet often reveals two distinctly different classes of signals. The most freqent signal has a sdden plse onset, followed by a reverse swing and a rather slow decay of the excited speclrm (Fig. 3). Voltage Derivative Imbalance qench nmber 15 - RD3c Time (ms) Fig. 3. One of several thosand half-magnet-imbalance, Fast-Flx-Changes (FFC's) that were recorded dring RD3c's training. In bold, a low freqency imbalance observed at 10% of calclated short sample dring ramp Q15 of RD3c's trai ning.

5 4 This imbalance has been of intense experimental interest at LBNL since before D20 [I], and interpreted as evidence of stick-slip condctor motion. This signal is freqently observed immediately before the start of a spontaneos training qench (Fig. 4). It was observed that the freqency of these events sally increases dramatically while ramping a magnet into a "virgin" Lorentz-stress range. The same freqency in "nonvirgin" regions sally redces monotonically as the magnet "trains". In addition, the magnet's training rate correlates with the rate of redction of the "non-virgin" event freqency. At the same time in any "training fa ll-back" not only the crrent redces bt also the "non-virgin" event freqency increases ' > ' Voltage Derivative Imbalance casing qench nmber 15 RD3c horefield 9.97 T 2.>,., _ dv/dt (right. lofl,ido=.,ro.e,a,""1 ",. ptct La \he bon) Time (ms) Fig. 4. The mass ive FFC lms before the onset of resistance that tenninated ramp QI 5 (the highest field observed dring R03c's training). Low freqency flx changes (Fig. 3 bold crve) have been observed when the magnet crrent is low (10-50% of the cable's crrent-limit). The nmber of these "slow" events is nearly independent from the direction of crrent ramping, the nmber of complete ramps since the last qench, and the level of the magnet training. At higher ramp-rates, the nmber of events decreases, and the magnet imbalance voltage increases. When a half-magnet imbalance asymmetry exists, the asymmetry inverts when the ramp-rate inverts [7]. IV. DISCUSSION Virgin magnet performance contines to be npredictable for several reasons, inclding the large nmber of parameters that can inflence training, and the difficlty in qantifying (and controlling) these parameters. With the new approach for magnet assembly and disassembly achieved with the RD3-style magnet containment of flat racetrack coils, one can imagine inexpensive tests of changes that cold affect training. For example, the pre-stress redction to allow RTi-style coi l separation in SM-OI [8] took 2 days (removal to reinsertion in its cryostat). Even if coil separation is intolerable for field qality inside accelerator magnets, low pre-stress crrently appears to be a simple means to accelerate the assessment of magnet technology innovations. The relatively similar ramp rate sensitivity of the large coils is not too srprising. given their relatively similar preparations. Stdies of qench propagation on different cables throgh stimlated qenches cold be helpfl in stdying inter/intra- 4> strand crrent sharing behavior in the cable, especially for the new cable design sch as mixed strand and cored cable [II]. The RRR vales appear rather controllable. The present target is RRR > I 0, with a condctor heat treatment target of less than 200 hors, withot losing crrent density [I I]. A small-magnet test is being planned to investigate the relationship between peak condctor temperatre and condctor margin degradation, sing an integrated ribbon spot heater and local thermometer near the maximm field. The freqent presence of high-freqency FFC's immediately before spontaneos qenching, and the strong correlation between magnet training rates and FFC training rates, spports the notion that controlling the freqency and amplitdes of FFC's cold accelerate magnet training and improve reliability. Locating these condctor movements cold help magnet designers design more reliable magnets and "retro-fix" slowtraining magnets more systematically. The slow fast-flx changes are believed to reslt from flx jmping in excessively large Nb1Sn filaments, and will likely disappear as the condctor improves. V. CONCLUSION For several years the Spercondcting Magnet Grop at LBNL has been developing technology for high-field Nb1Sn accelerator magnets. Dipole magnets with different geometries and characteristics have been sccessflly tested. Common coil magnets have been explored the past few years. While the cost redctions and record dipole fields have been very encoraging, many isses reqire frther improvements, inclding: training, condctor movements, ramp rate dependence, and flx jmps. The recently initiated Sbscale Magnet Program is expected to address many isses in a more cost-effective manner. REFERENCES [I] A. Mcintrff et al., ''Test Reslts for a High Field (I3T) Nb)Sn Dipole", Particle Accelerator Conference, Vancover, Canada, p [2] S.A. Gorlay et ai., Fabrication and Test Reslts of a prototype, Nb3Sn Spercond4cting Racetrack Dipole Magnet", 1999 PAC New York, NY, March 1999, Vol. I, p.174. [3] S.A. Gorlay et ai., "Fabrication and Test of Nb)Sn Racetrack Coils at High Field", IEEE TrailS. Appl. SperCOIldllct., Yol. II, No. I. March p [4] K. Chow et al., "Mechanical Design of a High Field Common Coil Magnet", 1999 PAC. New York, NY, March 1999, Yol. 5, p ] S.A. Gorlay et al., "Design and Fabrication of a 14T, Nb)Sn Racetrack Dipole Magnet", IEEE TraIlS. Appl. SlIpercolldllcr., Yol. 10, No. I, March 2000, p.294. [6] A.F. Lietzke et ai., "Fabrication and Test Res lts of a High Field, Nb)Sn Spercondcting Racetrack Dipole Magnet", 2001 PAC, Chicago, IL, Jne 2001, p.20s. [7] A.F. Lietzke et ai., ''Test Reslts for Rd3c, a Nb)Sn Racetrack Dipole Magnet", this Conference. [S] R.R. Hafalia et al., "An Approach For Faster High Field Magnet Technology Development", this Conference. [9] M. Coccoli, L. Chiesa, "SM-Ola and SM-Olb Test Res lts", Internal Note LBNL SC-MAG 775, Janary (10] G.L. Sabbi et al., "Nb)Sn Qadrpole Magnets for the LHC IR", thi s Conference. {J 11 R. Scanlan, O.R. Dietderich, "Progress and Plans for the High Energy Physics Condctor Development Program", This Conference.