Storage rings and free-electron lasers (FELs) make use of undulators to deliver highly brilliant photon beams to a vast user community. Undulators are special magnets consisting of an array of dipoles that produce a sinusoidal magnetic field. Undulators emit radiation at specific wavelengths λ _n : λ n = λ U 2 n γ 2 1 + K 2 2 ; K = e B λ U 2 π m e c , (1) where λ _U is the undulator period, n = 1,2,3,... is an integer describing the harmonic number, and γ is the Lorentz factor. K is the “so-called” undulator parameter, e is the electron beam charge, B is the peak magnetic field on the axis, m _e is the electron mass, and c is the speed of light. Most of the undulators used nowadays are based on permanent magnet technology. As shown in Equation (1), in order to reduce the photon wavelength/increase the photon energy without increasing the electron beam energy at a fixed peak magnetic field on axis B, the undulator period λ _U needs to be reduced. On the other hand, for the same technology and geometry, by reducing λ _U , the peak magnetic field on axis B and, therefore, the tunability range (the available photon energy range) are also decreased.

Different efforts have been made in the past decades to reduce the period of the undulators while keeping a high magnetic field on axis B. Figure 1 shows the evolution of these efforts. On the top left is sketched a permanent magnet undulator (PMU) with the magnet arrays out of the vacuum chamber. The first step has been the inclusion of the magnetic arrays inside the electron beam vacuum chamber. In this way, the magnets are closer to the electron beam, which, therefore, is subject to a higher magnetic field. These are the “so-called” in-vacuum undulators (IVUs). The second step consisted in cooling the magnetic arrays in the IVUs. By cooling the magnets to approximately 100 K, their remanence increases and so does the peak field on the magnetic axis. These are the “so-called” cryogenic permanent magnet undulators (CPMUs). In order to further increase the peak field on the axis, superconducting magnet technology has been applied. Superconducting undulators (SCUs) have dipole arrays made of superconducting coils cooled down to approximately 4 K.

A comparison in terms of peak field on the magnetic axis between the aforementioned technologies is shown in Figure 2. A beam stay clear/vacuum gap of 5 mm has been considered. The curves reported in Figure 2 are obtained by the following parametrization curves. For the magnetic field on axis B (T) of the PMUs and IVUs with, respectively, a magnetic gap g of 7 mm and 5.2 mm, the following function [1] is considered: B = 3.381 ⋅ exp − 4.73 ⋅ g λ U + 1.198 ⋅ g λ U 2 . (2)

The magnetic field of the CPMUs with g = 5.2 mm is described by [2]. B = 3.502 ⋅ exp − 3.604 ⋅ g λ U + 0.359 ⋅ g λ U 2 . (3) The one for the SCUs is described by [3]. B = ( 0.28052 + 0.05798 ⋅ λ U − 0.0009 ⋅ λ U 2 + 5.10 − 6 ⋅ λ U 3 ) ⋅ exp − π ⋅ g λ U − 0.5 , (4) where λ _U and g are in mm and B is in T. The blue curve shown in Figure 2 for SCUs is valid by operating at a temperature of 4.2 K with a temperature margin of 1 K for both of the following cases: a) a magnetic gap of 6 mm applying ≈80% of the critical current and b) a magnetic gap of 6.5 mm applying ≈85% of the critical current. The last one, which is the case for S-PRESSO, the first SCU to be installed at the European XFEL, is described in Section 4. SCUs show potential to further increase their performance, as shown in the blue curve in Figure 2. This is possible by applying superconductors with higher critical temperatures and current densities, such as Nb₃Sn or high-temperature superconducting tapes or bulks. Both technologies are not mature yet. Another possibility is to work with the well-proven technology based on the NbTi wire using a cooling scheme, allowing it to reach lower temperatures (the cyan curve in Figure 2). Applying superfluid helium cooling schemes as performed for the Nb superconducting radio frequency (RF) cavities improves the peak field on the axis by 35%. The cyan curve in Figure 2 is obtained by multiplying the right term of Equation (4) by 1.35. Simulations made with FEMM (http://www.femm.info/wiki/HomePage) confirmed this parametrization for 10, 15, 18, 20, and 25-mm period length SCUs with a magnetic gap of 6 mm and a temperature margin of 1 K.

An additional advantage of NbTi SCUs with respect to permanent magnet-based undulators is radiation hardness, which has been demonstrated in NbTi magnets used in different colliders such as the Tevatron [4], HERA [5], and LHC [6] for several decades.

The choice of the undulator period is driven by the selected photon energy range and can be made once the electron energy is fixed. To cover an even larger photon energy range, it is possible to 1) vary the electron beam energy, which is performed, for example, at the European XFEL and 2) use undulators with period doubling. These possibilities complicate the parameter space, and for the sake of simplicity, they should be kept as options and not as guiding lines. It should also be remembered that the large tunability range of the photon energy presents challenges for both the optics and the experimental techniques.

SCUs offer elegant solutions, using the same winding body, to duplicate or triplicate the period length, which can further increase the photon energy tunability range considerably [7].

SCUs based on NbTi coils have been serving users since more than 5 years at the Advanced Photon Source (APS) [8] and the Karlsruhe Institute of Technology (KIT) light source [9]. SCUs are now commercially available from the company Bilfinger Noell GmbH (BNG). A long-term development carried out at KIT in collaboration with the industrial partner BNG has led to the successful commercialization of SCUs [10]. Recently, an SCU produced by BNG and measured by KIT has been installed in the Australian Light Source [11].

It is now time to implement SCUs in FELs. The potential of SCUs for X-ray FELs is addressed in the following section. Afterward, the plans for implementing SCUs at the European XFEL and the ongoing activities necessary to realize them are described.

In terms of cost reduction, the main advantage of using short-period and strong magnetic field undulators, such as SCUs, for an FEL with respect to PMUs, is that, to reach a defined photon energy, it is possible to reduce the electron beam energy.

In the case of an undulator with a shorter period but keeping the same K value, the saturation length is shorter, and consequently, the length of the undulator line is shorter. For a green field facility, this would result in considerable savings in civil construction costs.

Advantages of SCU over PMUs include the following:

• SCUs allow for building more compact and, therefore, less expensive FELs.

Currently, X-ray FELs use PMUs and IVUs. However, there are no plans, to the best of authors’ knowledge, to install CPMUs, SHINE, LCLS, and European XFEL working on the implementation of SCUs [12]. The Chinese CW X-ray FEL SHINE plans a complete undulator line with SCUs [13]. The LCLS facility is planning to install a cryostat with two 1.5-m-long SCU coils with a cold intersection in between consisting of a superconducting phase shifter, a copper quadrupole, and an RF cavity BPM [14]

SCUs are part of the European XFEL facility development program. The development of SCUs will enable us to build the know-how inside the facility for the state-of-the-art technology and further improve the know-how technology to be later transferred to industries. The benefits of SCUs to the European XFEL strategic plan are manifold:

• The European XFEL has the highest electron beam energy among all the similar facilities worldwide [15]. SCUs with a period between 15 and 20 mm, and a vacuum gap (magnetic gap) of 6 mm (7 mm) offer the possibility to establishing an FEL source above 30 keV toward 100 keV, which is unprecedented to date [16,17].

The European XFEL has three undulator lines serving seven instruments. There are two hard X-ray lines and a soft X-ray line named SASE1, SASE2, and SASE3, respectively. By changing the electron energy range between 8.5 GeV and 17.5 GeV and/or the undulator gap, the undulator lines can be tuned to cover the following photon beam energy ranges: 3.6–25 keV (SASE1/2) and 0.25–3 keV (SASE3).

A total of six SCU modules are planned to be installed downstream of the SASE2 PMUs at the European XFEL. The Superconducting undulator PRE-SerieS mOdule (S-PRESSO) is the pre-series module for the Free-Electron laser SuperconducTing undulator Afterburner (FESTA). More details on the design of S-PRESSO can be found in the following section. The small-series production FESTA consists of five modules.

A sketch of the SASE2 undulator line at the European XFEL showing the position of the planned six SCU modules is presented in Figure 4. The length of each SCU module cryostat is 5 m, identical to the length of the installed PMUs. This allows us to use the same intersections (violet rectangle).

Each intersection contains a quadrupole magnet to focus the electron beam transversely below ≈30 μm, which is necessary for the FEL process to take place; a beam position monitor and a beam loss monitor for electron beam diagnostics; air coils to trim the vertical and horizontal first and second field integrals of the undulators; and a phase shifter to compensate the phase advance of the emitted photons with respect to the electron beam. In the PMU lines, the intersections are 1.1 m long. An RF valve will be placed at the intersection separating the last PMU and the first SCU. The intersections interfacing with the SCUs also need to host RF bellows. The RF bellows permit vertical and horizontal alignment of the SCUs with beam and length compensation for the thermal shrinkage of the SCU modules after cooldown. Each SCU module will sit on four cam movers that are remotely controllable.

All SCU modules will be cooled with cryocoolers, following a cryogen-free conduction-cooled scheme similar to the one developed by the KIT/Bilfinger Noell collaboration. We are planning to use six cryocoolers per SCU module. Each SCU module contains two 2-m-long SCU coils with superconducting corrector coils co-wound at the last grooves of the iron yoke and two pairs of vertical and horizontal iron-free coils located between the main coils with a superconducting phase shifter between them (see the upper plot in Figure 4). The layout of each module is planned to be the same as the S-PRESSO module, and it is described in the following section.

Each intersection also contains an absorber to screen the downstream undulator from the radiation produced by the upstream undulators. The PMU vacuum chambers are elliptical with horizontal and vertical axes of 15 mm and 8.6 mm, respectively. The absorbers in the PMU line have a diameter of 8 mm. The SCU vacuum chamber has a racetrack shape with horizontal and vertical axes of 10 mm and 5 mm, respectively. Therefore, absorbers with a diameter of 4 mm are planned.

Simulations were run to estimate the heat load on the six SCU modules. The simulations were performed using BDSIM [18], a tracking code that also models the interactions between the particles and the beamline components. The simulation tracked the incoherent synchrotron radiation produced in SASE2 until after the SCU modules, simulated with SPECTRA. Some of this radiation is absorbed by copper absorbers and other beamline components during the simulation. The absorbers throughout SASE2 have an aperture of 8 mm. However, the absorber above the S-PRESSO module and all absorbers downstream of the S-PRESSO module have an aperture of 4 mm. The results of the simulation are used to estimate the rate at which energy is deposited into the SCU modules when the machine is running at full power. The results, reported in Figure 5, show that the maximum heat load is ≈ 0.1 mW, which does not affect the SCU module’s operation.

A test 4-mm-diameter absorber has been installed downstream of the last PMUs of SASE2 in the winter shutdown of 2022/2023. The electron beam successfully passed through this test absorber. Radiation-sensitive detectors are placed downstream, and the effects of the radiation on the devices installed nearby are under investigation. The first results show low radiation levels.

The SCU coils are foreseen to have a period of 18 mm and a peak field on the axis of 1.82 T. We are mainly focusing on applications above 30 keV for electron beam energies > 16 GeV. S-PRESSO and FESTA can be operated by amplifying the first harmonic of the PMUs up to ≈40–50 keV. To reach higher photon energies, we planned to use the second harmonic bunching generated by the FEL process in the PMUs and amplify it with the fundamental of the SCUs. From first estimations, considering an electron beam energy of 16.5 GeV, a normalized emittance of 0.4 mm rad, and an initial energy spread of 3 MeV, the number of photons per pulse 30 fs long is ≳ 10¹⁰ up to ≈50 keV and > 10⁹ at ≈ 60 keV [17]. Such values are up to two orders of magnitude larger than what can be reached in diffraction-limited storage rings such as the ESRF-EBS and APS-U [16]. Further studies, including wakefields, tapering, and optimized electron beam parameters, are foreseen. A promising way to generate brilliant photon beams at high energies is harmonic lasing [19]. A scheme using phase shifters and filters [20] for suppression of the fundamental is currently under numerical study. Experimental tests at low energies are also planned in the near future.

S-PRESSO is the pre-series module for FESTA. The technical requirements are reported in the study by [16] and summarized as follows. The main goals of S-PRESSO are to demonstrate the following:

• it is possible to contribute to the FEL process with such an SCU by reaching the needed accuracy in the alignment of the two 2-m-long SCU coils, and the mechanical manufacturing and assembly of the SCU coils;

The contract of S-PRESSO has been assigned to Bilfinger Noell GmbH. The main parameters of the SCU coils of S-PRESSO are shown in Table 1. The peak fields on the axis and period have been chosen such that the fundamentals of the SCUs and SASE2 PMUs overlap at the high-photon-energy end of the SASE2 PMU tuning range. For example, for an electron beam energy of 16.5 GeV, the lowest photon energy of the fundamental of the SCUs is 25.3 keV and will overlap with the fundamental of the PMUs up to K ≈ 0.765, corresponding to 50 keV. As mentioned in Section 2, for the CW upgrade, it will be necessary to lower the electron beam energy to 7–8 GeV. For an electron beam energy of 8 GeV, the lowest photon energy of the fundamental of the SCUs is 5.9 keV and will overlap with the fundamental of the PMUs up to ≈ 11.8 keV.

The main coils will be wound with a rectangular NbTi wire of 0.6 mm × 1 mm (bare dimensions). The wire has polyvinyl Formvar insulation with a thickness of 0.04 mm and 2,100 filaments with a diameter of 13 µm. All grooves are filled with 55 turns except the first and last three ones, which have a different winding scheme for end field compensation similar to the one used for the SCUs installed in the KIT and Australian light sources [10,11]. The geometry of a one-and-a-half period is shown in Figure 6.

The upstream and downstream SCU coils will be powered by two power supplies up to a maximum current of 900 A. With 900 A of operating current, 2D and 3D magnetic field simulations produce a peak magnetic field on the axis that slightly exceeds the required 1.82 T. At a nominal current of 900 A, the conductor has a temperature margin of 1 K by operating at 4.2 K. To reduce the power consumption during operation, the power supplies will be placed in the tunnel close to S-PRESSO.

For the correctors and phase shifters, nine power supplies with 30 A and 40 V are foreseen. These are standard power supplies developed by the Deutsches Elektonen-Synchrotron (DESY) and used for the room temperature correctors of EuXFEL and PETRA III. The correctors and phase shifter will operate at a current of 10 A with a large temperature margin > 3 K.

The needed mechanical accuracies are reported in Table 2. They have been set considering that for the half periods, the following limits on K/2 hold: |∆K/K| _rms < 0.0015 and |∆K/K| < 0.006. With the aforementioned restrictions on K/2, the pulse energy emitted by the six SCU modules is not significantly reduced [21]. FEL simulations show that the same requirements for mechanical accuracies are well-compatible with a complete SCU line with a 15-mm period length. Such mechanical accuracies reduce the mean FEL power at saturation length by less than 5% [22] and are, therefore, not notably harming the performance of the FEL process. In case that the magnetic field measurements show that the required mechanical accuracies/range of deviation of the undulator parameter are not reached in the manufacturing and assembly processes or are not maintained after cooldown, shim coils are foreseen.

The shim coils can be wound as racetracks (see the orange coil in Figure 7) with a thin NbTi wire with a diameter of 0.152 mm (insulated). Each shim coil wound around one pole can correct approximately 1% of the field with a current of 5 A. Using two neighboring shim coils with a pole in between and powered with 5 A, it is possible to correct up to 2% of the magnetic field. These values are sufficiently larger with respect to the maximum specified variations on the K-value. More configurations are studied and reported in [23]. A maximum of 11 power supplies of the same type as those for the corrector coils are foreseen for the shim coils.

A misalignment between the two coils causes a correlated error in K-values along the magnetic axis. Calculations made by [22] show that for an SCU line, the two SCU coils in one module, as reported in Figure 8, need to be aligned with respect to each other within the values listed in Table 3. To fulfill the requirements on alignment, a rigid support structure, sustained by eight Ti rods, is designed to hold the SCU coils straight within 30 µm over a length of approximately 4.5 m. The magnetic gap is accessible at some places for coordinate measurement machine (CMM) measurements in assembled conditions. The alignment of the coils will be performed at room temperature and atmospheric pressure. The fiducialization procedure is foreseen to be performed as described in the study by [9].

A copper vacuum chamber separates the cold mass, placed in an insulation vacuum, from the electron beam vacuum. The racetrack chamber, with horizontal and vertical axes of 10 mm and 5 mm, respectively, will be 0.3 mm thick and will be made of copper with a residual-resistance ratio ≈100. The internal surface of the copper electron beam vacuum chamber must have a roughness R _a < 0.8 µm and an oxide layer thinner than 20 nm. In such conditions, the beam heat load due to resistive wall heating is always below 10 W for all relevant operation modes [16]. Geometric impedance will contribute to less than 0.1 W by keeping steps in the vacuum chamber cross-section < 0.1 mm. Synchrotron radiation will be screened by the copper absorbers placed in the upstream intersection (see Section 3).

A passive quench protection system in the form of cold rectifier diodes is foreseen. Each SCU single coil (top or bottom coil of one of the two SCU coils in the module) is protected by a set of six cold diodes, amounting to a total of 24 diodes for the two SCU coils. In case of a quench, the output of the power supply is deactivated via its interlock function, and the stored energy is dissipated within the coils and the diode racks. The quench detection is based on the balanced voltage principle. A quench is detected as soon as the difference between the voltages across the upper and lower coils increases above a certain threshold for longer than a given amount of time.

The cooling system is divided into two circuits: one for cooling the magnets and one for the vacuum chamber. S-PRESSO is cooled by six Gifford McMahon (GM) cryocoolers from Sumitomo: three are of type SRDE-412D4 with a cooling power of 1.25 W at 4 K and three are of type RDK-408S with a cooling power of 5.4 W at 10 K. The second stages of the three SRDE-412D4 cryocoolers are connected to the 4.2 K components, including the high-temperature superconducting current leads, while the three RDK-408S cryocoolers are connected to the electron beam chamber at approximately 10 K. The first stages of all six cryocoolers are used to intercept radiation on the thermal shield and the heat due to thermal conduction on the support rods, the electron beam chamber, and the resistive part of the current leads. The upstream and downstream SCU coils are powered by two power supplies through three current leads designed to operate at 900 A. This is of considerable advantage in terms of heat load and, therefore, needed cooling power compared to the normal design with four current leads carrying the operating current. The foreseen cooldown time is 7 days, and the warm-up time is 4 days.

Two test facilities, SUNDAE1 [24] and SUNDAE2 [25] (Superconducting UNDulAtor Experiment), are under development to qualify the magnetic field properties of SCU coils. SUNDAE1 and SUNDAE2 are planned, keeping in mind the present needs for S-PRESSO and FESTA, and possible future upgrades of the European XFEL. SUNDAE1 is also foreseen to measure SCU test coils with novel winding schemes and superconducting wires and tapes carrying higher currents than NbTi, such as high-temperature superconducting (HTS) tapes (see Section 6). SUNDAE1 and SUNDAE2 are located at DESY in a hall with a cryoplant. Tests at 2 K are foreseen in SUNDAE1 and will be possible in SUNDAE2 with minor refurbishments of the helium distribution box. Measurements at 2 K might be of interest for possible future upgrades of the European XFEL, where complete SCU lines might be used. In such a case, cooling would be performed with a cryoplant instead of cryocoolers. To further improve the performance of the NbTi SCUs, similar solutions, as adopted for the superconducting RF cavity cryo-modules, could be used to cool the developed SCUs to 2 K (see Figure 2).

Both test stands will use their own in-house developed control systems, specifically based on Beckhoff Automation components and PLC software frameworks, as well as .NET applications. Data from the experiments will be stored in the InfluxDB database. Real-time visualization of time series will be performed via the Grafana web-based framework (see https://grafana.com/) connected to a database. Data analysis, such as calculating field integrals, will be performed via in-house developed software.

To improve the undulator’s performance, Nb₃Sn or HTS materials like rare earth barium copper oxide (ReBCO) instead of NbTi superconductors can be utilized. These materials have higher irreversibility fields, critical temperatures, and critical current densities. The field quality of an SCU is strongly determined by the mechanical accuracies reached. Although Nb₃Sn requires heat treatment, which reduces the reachable mechanical accuracies, this is not the case for ReBCO tapes, making them a good candidate for undulator applications.

Companies producing ReBCO tapes have made important improvements in the past few years to increase the engineering current density by doping ReBCO tapes with Zr and reducing the thickness of the substrate from 50 μm to 30 µm. This increases the reachable peak field on the axis even further. Once uniformity, repeatability, mechanical properties, available length, and production rate of high-quality ReBCO tapes are improved, ReBCO tapes will become the working horse to produce full-scale length magnets, including undulators, to be reliably operated in storage rings and FELs.

The best performance in terms of peak field on the axis can be reached by filling the SCU grooves with the HTS tape, taking advantage of the high engineering current density. The peak field on the axis can be increased by a factor of ≈ 1.7 with respect to NbTi by using the same geometry and the 11-K temperature margin operating at 4 K. This is reached with a current density of 5000 A/mm² in the HTS tape. The values of the critical current have been obtained from the high-temperature superconducting wire critical current database maintained by the Victoria University of Wellington, Australia [27]. However, HTS tapes are not yet available in long lengths (the maximum length is approximately 100 m), and they are roughly 10 times more expensive than the NbTi wire. To use the high engineering current density of the ReBCO tape in the most effective location to increase the peak field on the magnetic axis, HybriSCU, a graded SCU combining NbTi and the ReBCO tape, has been proposed and studied. The concept foresees winding the first layers in the SCU grooves with the NbTi wire and the last layers closer to the electron beam with the ReBCO tape. An additional advantage of placing the ReBCO tape closer to the electron beam is its high temperature margin.

The geometry taken into consideration foresees the use of the 4-mm HTS tape and an undulator period of 16.8 mm. The improvement of B _max at 16.8 mm for a magnetic gap of 6 mm and a temperature margin of 1 K for NbTi and 11 K for the ReBCO tape is approximately 15% by adding seven layers of the ReBCO tape to the 15 layers of NbTi and approximately 37% by considering 13 layers of NbTi and 30 layers of the ReBCO tape [28]. Further studies on the realization process of HybriSCU and ReBCO coils are planned.

Superconducting undulators are successfully operated in synchrotron light sources. The European XFEL plans to demonstrate the FEL process with S-PRESSO, a pre-series superconducting undulator, and provide photons for users with S-PRESSO and FESTA in the unexplored FEL regime with photon energies > 30 keV. This project is pivotal to implementing complete SCU lines for X-ray FELs, including those operating in the CW mode.