Superconducting magnets at the heart of modern accelerators are complex state-of-the art systems. They operate at extreme conditions of deep cryogenic cooling and heavy mechanical stress. Accelerator magnet are inductors, and therefore store magnetic energy when in operation: a single 14-m long dipole of LHC stores ~7 MJ at nominal operating current. This significant energy can be suddenly released in a process called “quenching” when a portion of the superconducting winding transitions into a normal state. The threshold for this to occur is nearly 1010 times lower than the stored energy: a staple dropped from a 3 cm height would be sufficient to quench a magnet. This demonstrates the enormous challenge of designing magnets that minimize localized material failures and motion under stress—potential causes of quenches. But quenches do occur, and once a quench begins, its detection should be as quick and reliable as possible to allow for a timely activation of protection systems preventing thermal damage to the magnet and infrastructure.

Quench in a superconducting magnet

Left: A Berkeley Lab magnet called CCT3, immersed in a liquid helium bath, quenches spontaneously at 11 kA. CCT3 is one of the Canted Cosine Theta magnets we have been developing; the plan is that this winding style will intercept stresses that could cause conductors to move under high magnetic fields. Right: The cryogenic camera at our Magnet Test Facility.

Traditionally, to detect and localize quenches, superconducting magnet voltages are monitored across taps distributed along the coil windings. While this approach normally works well for smaller magnets, new design concepts such as the Canted Cosine Theta that is under development by the MDP, or the 4 m-long QXF quadrupoles developed by LARP/AUP for the LHC luminosity upgrade, call for more informative and less invasive diagnostic techniques. Also, the race towards operational fields in excess of 15 T makes it inevitable that he presently-used NbTi and Nb3Sn low-temperature superconductors will be augmented or supplanted by high-temperature superconductors like YBCO or Bi-2212. The latter class of superconductors is characterized by a much slower initial rate of quench development, difficult to detect using voltage-based techniques.

To address these challenges, ATAP's Superconducting Magnet Program, which is a partner in the multi-lab US Magnet Development Program, has been leading the way in developing novel diagnostics and bringing the existing ones to a new level. Below are some highlights of the diagnostic techniques that are established or under development in the group.

Magnetic Quench Antennas

One way of learning about the quench is by sensing its magnetic signature. This approach was proposed in the early 90s by CERN and KEK researchers, who introduced magnetic "quench antennas" capable of providing real-time information of quench location and quench propagation velocity. These antennas are less invasive than traditional voltage taps, but need to be in close proximity to superconducting coils to measure detectable signals. We have taken this concept further by introducing “axial field” quench antenna designs that are sensitive to field disturbances caused by quenching, yet insensitive to variations of the principal field component of the magnet. This makes them less affected by noise and more efficient in picking up quench signatures. For smaller scale conductor studies, cryogenic linear Hall arrays were recently developed, allowing for real-time measurements of quench propagation and detection of current sharing irregularities along the quenching conductor. Quench antennas enable use of new conductor geometries for high-field magnets or new applications like fusion energy.

Various axial-field magnetic quench antennas developed at LBNL for accelerator magnets.

Quench propagation in a Conductor on Round Core (CORC®) cable made from high-temperature superconductor, shown as a 2D plot of field variation vs. time along a linear array of Hall sensors (shown in the top image).

Acoustic Emission Diagnostics

A superconducting magnet is a complex and very heterogeneous system, so under stress it would typically experience numerous mechanical transient effects such as epoxy cracking, delamination, cable or structural part motion, un-sintering of superconducting cable strands, or dislocation motion within the conductor material. These effects are often the cause of premature quenching, or of another puzzling phenomenon called “training." A superconducting magnet reaches its operational field strength only after a series of quenches, usually at progressively increasing currents (although occasional de-training during this process is also common). This break-in process constitutes a costly part of magnet development and testing, and it is a long-standing yet unsolved problem in the field. Once trained, magnets are expected to “remember” the maximal current (and stress state) they have reached; such memory is crucial for stable long-term operation of the accelerator.

To learn about magnet transient mechanics, we employ an Acoustic Emission (AE) diagnostics technique that allows us to localize and potentially “fingerprint” various mechanical disturbances. The AE technique was pioneered in the early 80s by MIT, and has recently gained popularity due to advances in hardware and data processing. The original cryogenic acoustic sensors were developed at LBNL in 2013 and have been adopted by a number of magnet groups within the USMDP collaboration and internationally.

Cryogenic acoustic emission sensors developed at LBNL. They operate from room temperature to 1.9 K, and provide a 1 MHz usable bandwidth for the acoustic signals.

Spectrogram of a typical acoustic emission event

When energized, magnets emit sounds in the broad (up to 300 kHz) ultrasonic range; it can be heard when slowed down ~10-20 times in time. The high-frequency portion of it potentially contains information about the nature of the emitting event, while the ringdown "tail" is representative of the internal mechanical resonances of magnet structure.

Cross-sectional view of high-field Nb3Sn dipole HD3 instrumented with four acoustic sensors S1-S4 locations shown in red

Trace of the acoustic signals recorded as HD3 magnet approaches quenching . Sensor S1 (blue) -> Left sound channel; sensor S4 (red) -> Right sound channel.

Original sound was slowed down 10 times.


By timing acoustic wave arrival at the sensors, the emission source can be localized within the magnet. When quench is driven by a crack propagation or local conductor motion, such techniques allows us to determine the origin of the quench with a precision of a few cm. In the example shown at the left, quench locations are triangulated using 8-sensor array and plotted on top of the CCT magnet 3D model (top). With every new current ramp the quench current is progressively increasing, which manifests the magnet "training" process (middle plot). Training is costly and its origins are not yet fully understood. We analyze acoustic "fingerprints" of magnets to determine the origins of training and develop ways of mitigating it. As training progresses, one can hear increasing "ratcheting" sounds from the magnet beyond quenching, pointing to a slip-and-stick type of internal motion. Once trained, magnets usually "memorize" the highest current they have reached, and maintain the post-training level of performance in regular accelerator operations.

Magnet memory also clearly manifests itself on a microscopic scale through the so-called "Kaiser effect": newly built magnets emit sounds only when energized beyond the current / field level they have already "seen" before, while staying mostly quiet below that current level. This is though of being a result of progressive crack propagation and percolation in epoxy-impregnated magnet windings. However, as training progresses, it behavior changes, and sounds are now emitted in every ramp, with peak amplitude progressively increasing with the current. This is presumably happening due to a transition to the slip-stick mechanism of internal motion. We found, that the cracking and slip-stick regimes exhibit different acoustic "signatures," , and also separated by a "kink" in the training curve. We are working on investigating these phenomena to understand and reduce training in future magnets built by MDP collaboration, such as CCT5 at LBNL and the 15-T dipole at Fermilab.

An example of "Kaiser effect" in CCT4 magnet: acoustic emission and current ramping traces are shown as function of time for different quench numbers. The training plot of this magnet is shown in the inset on the right.

Active Quench Detection for the High-Temperature Superconductors (HTS)

Quench propagation velocity in HTS materials is slower than in low-temperature superconductors (< 50 mm/s at fastest), which translates into a very localized hot spot that does not generate much resistive voltage. Therefore an HTS coil could burn out before a quench was detected by the traditional means of voltage taps. We recently proposed to use the conductor itself as distributed thermometer, and monitor local temperature variations using corresponding changes in the sound velocity v=√(E/ρ) , where E is the Young's modulus varying ~1-10 ppm per degree Kelvin. While this change is quite small, it is detectable using the active acoustic technique. This technique allows to detect quenches without using voltage tap wires and provides an additional level of magnet-protection redundancy.

Operational principle of the technique :

  1. An object is pulsed by a sender transducer
  2. A “ring-down” transient waveform reverberates across it multiple times
  3. Transient oscillation is acquired by a receiver transducer; and stored as “reference” URef (t). Its shape is uniquely defined by the object geometry, density and elastic modulus E(T)
  4. Pulsing and transient acquisitions are repeated periodically; every new transient Ux(t) is compared to URef(t) using cross-correlation: A(dt) = Ux(t+dt)*URef(t), and the mutual time shift dt that yields maximal cross-correlation is found.
  5. When a hot spot develops, E(T) decreases locally, thus and adding delay to the wave propagation. This proportionally increases dt, which is being monitored as function of time.

In a tape HTS conductor, in-plane shear waves appear to be most useful for detection purposes, as they can propagate long distances without being damped by the liquid cryogen or structural elements surrounding the conductor. We excite these waves using a plate-shaped piezo-transducer attached to the tape surface (shown at the right).

ANSYS simulation of in-plane wave propagation in HTS tape conductor

In a demonstration experiment shown below, we were able to detect a developing quench in a 1.2 m long superconducting HTS tape immersed in liquid nitrogen, when a temperature of a localized (~1 cm long) hot spot rose 1.6 K above the temperature of the nitrogen bath. A differential variant of the technique was used, comparing time shift for the wave arriving at the tape ends from a single pulsing transducer installed in the middle of the conductor. This allowed for a drastic improvement of detection sensitivity due to cancellation of signals due to temperature fluctuations of the nitrogen bath. As excitation and detection involve a frequency range above 100 kHz, the technique is mostly insensitive to mechanical noise and ambient vibrations.

Transient waveforms acquired by the receiver transducers at the tape ends (offset vertically for clarity). Sub-waveforms shown below were selected for quench monitoring.

Sketch of the experimental setup

Results of quench detection experiment with permanent magnet placed on the tape in location “B” (in order to create a "weak spot" where quenching would start)

We are further developing active detection towards applications in sub-scale magnets and cables. In magnets where the conductor is epoxy-impregnated, its cooling is less efficient, and a higher hot spot temperature is expected which would be easier to detect. Also, acoustic waves can travel across windings turn-to-turn rather than along the conductor, thus reducing the required sender-receiver distance and further improving the sensitivity. Pilot experiments were conducted on HTS magnet coils developed by the USMDP, including a successful test at 4.2 K on Bi-2212 sub-scale cable. Being non-invasive, the technique may especially benefit complex coil configurations with limited access to the conductor, such as those designed for fusion applications.

Apart from quench detection, our technique is applicable for monitoring mechanical contacts within the magnet. If those change during operation or training, the wave emitted by the pulser would reach the receiver over a different path, thus acquiring an additional time shift.

To learn more...

  1. “Axial-Field Magnetic Quench Antenna for the Superconducting Accelerator Magnets”, M. Marchevsky, A. R. Hafalia, D. Cheng, S. Prestemon, G. Sabbi, H. Bajas, G. Chlachidze, IEEE Trans. Appl. Supercond. 25, 9500605 (2015)
  2. q“Magnetic Quench Antenna for MQXF quadrupoles”, M. Marchevsky, G. Sabbi, S. Prestemon, T. Strauss, S. Stoynev and G. Chlachidze, IEEE Trans. Appl. Supercond. 27, v. 4, 9000505 (2017)
  3. “Acoustic emission during quench training of superconducting accelerator magnets”, M. Marchevsky, G. Sabbi, H. Bajas, S. Gourlay, Cryogenics 69, 50 (2015)
  4. "Localization of Quenches and Mechanical Disturbances in the Mu2e Transport Solenoid Prototype Using Acoustic Emission Technique", M. Marchevsky, G. Ambrosio, M. Lamm, M. A. Tartaglia, M. L. Lopes, IEEE Trans. Appl. Supercond. 26, 4102105 (2016)
  5. “Acoustic thermometry for detecting quenches in superconducting coils and conductor stacks,“ M. Marchevsky and S. A. Gourlay, Appl. Phys. Lett. 110 (2017)
  6. Quench Detection for High-Temperature Superconductor Conductors Using Acoustic Thermometry”, M Marchevsky , E. Hershkovitz, X. Wang, S. A. Gourlay, S. Prestemon, IEEE Trans. Appl. Supercond., v 28, Issue 4 (2018)
  7. "Eigenfrequency Thermometry", patent application filed (2017)