PERLE accelerator complex is arranged in a racetrack configuration; hosting two cryomodules (containing four, 5-cell, cavities operating at 802 MHz), each located in one of two parallel straights, completed with a vertical stack of three recirculating arcs on each side. The straights are about 10 meter long and the 180° arcs are 5.5 meter across. Additional space is taken by 4 meter long spreaders/recombiners, including matching sections. As illustrated in Figure 1, the total ‘footprint’ of PERLE is: 24 m × 5.5 m × 0.8 m; the last dimension reflecting 40 cm vertical separation between the arcs. Each of the two cryomodules provides 65.5 MeV energy boost. Therefore, in three turns, a 393 MeV energy increase is achieved. Adding initial injection energy of 5 MeV yields the total energy of 398 MeV - call it ‘400 MeV’.

Multi-pass energy recovery in a racetrack topology explicitly requires that both the accelerating and the decelerating beams share the individual return arcs. This in turn, imposes specific requirements for the TWISS function at the linacs ends: the TWISS functions have to be identical for both the accelerating and decelerating linac passes converging to the same energy and therefore entering the same arc.

The spreaders are placed directly after each linac to separate beams of different energies and to route them to the corresponding arcs. The recombiners facilitate just the opposite: merging the beams of different energies into the same trajectory before entering the next linac. As illustrated in Figure 2 , each spreader starts with a vertical bending magnet, common for all three beams, that initiates the separation. The highest energy, at the bottom, is brought back to the initial linac level with a chicane. The lower energies are captured with a two-step vertical beamline. The vertical dispersion introduced by the first step bends is suppressed by the three quadrupoles located appropriately between the two steps. The lowest energy spreader is configured with three curved bends following the common magnet, because of a large bending angle (45°) the spreader is configured with. This minimizes adverse effects of strong edge focusing on dispersion suppression for a lower energy spreader. Following the spreader, there are four matching quads to ‘bridge’ the TWISS function between the spreader and the following 180° arc (two betas and two alphas).

All six, 180° horizontal arcs are configured with the FMC optics to ease individual adjustment of M56 in each arc (needed for the longitudinal phase-space re-shaping, essential for operation with energy recovery). The lower energy arcs (1, 2, 3) are composed of four 45.6 cm long curved 45° bends and of a series of quadrupoles (two triplets and one singlet), while the higher arcs (4, 5, 6) use ‘double length’, 91.2 cm long, curved bends. The usage of curved bends is dictated by a large bending angle (45°). If rectangular bends were used, their edge focusing would have caused significant imbalance of focusing, which in turn, would have had adverse effect on the overall arc optics. Another reason for using curved bends is to eliminate the problem of magnet sagitta, which would be especially significant for longer, 91.2 cm, bends. Each arc is followed by a matching section and a recombiner (mirror symmetric to previously described spreader and matching section). As required in case of mirror symmetric linacs, matching conditions described in the previous section, impose a mirror symmetric arc optics (identical betas and sign reversed alphas at the arc ends). A complete lattice for arc 1 at 70.5 MeV, including a spreader, 180° horizontal arcs and a recombiner, is illustrated in Figure 3. Presented arc optics features high degree of modular functionality to facilitate momentum compaction management, as well as orthogonal tunability for both the beta functions and dispersion.