A Multi Purpose X-band structure for FELs M. Dehler, A. Citterio, A. Falone, J.-Y. Raguin Paul Scherrer Institut A. Grudiev, G. Riddone, R. Zennaro, D. Gudkov, A. Samoshkin, W. Wuensch CERN G. D‘Auria Sincrotrone Trieste

Why - 250 MeV Injector at PSI Introduction of X band technology at PSI - the shopping list: Klystron at European X Band (development by SLAC for CERN/PSI/ELETTRA) Accelerating structure described in the following X Band component – loads, splitters etc (developments by I. Syratchev) FERMI@ELETTRA joined in (similar needs for their project).

A CERN/PSI/ST collaboration Motivation for CLIC: Another data point in high gradient test program Validation of design and fabrication procedures A true long term test in another accelerator facility Motivation for the FEL projects: An X band structure to compensate long. phase space nonlinearities High gradient/power requirements of CLIC = a design for safe operation at the more relaxed parameters of the PSI X-FEL RF design (mostly) by PSI, fabrication & LL RF test (mostly) at CERN, mechanical support & other parts by FERMI

Special considerations for FEL Operating structure at relatively low beam energies (PSI injector: 250 MeV) High sensitivity to transverse wakefields! Strategy: Passive: Try to have open structure while maintaining good efficiency and breakdown resilience Active: Wake field monitors See offsets before they show up as emittance dilution Possibly measure higher order/internal misalignments (tilts, bends ….)

A priori specifications Beam voltage 30 MeV at a max. power of 45 MW Mechanical length <1017 mm Iris diameter > 9 mm Wake field monitors Operating temperature 40 deg. C Constant gradient design, no HOM damping Fill time < 1 usec Cooling assuming 1 usec/100 Hz RF pulse

Do a castrated NLC type H75 without damping manifolds! The strategy Use 5π/6 phase advance: Longer cells: smaller transverse wake Intrinsically lower group velocity: Good gradient even for open design with large iris Needs better mechanical precision Long constant gradient design (efficiency!) No HOM damping Wake field monitors to insure optimum structure alignment Do a castrated NLC type H75 without damping manifolds!

NLC type H75 Well optimized design (iris aperture, thickness and ellipticity varying along structure) Original design gives 65 MV/m for 80 MW input power Sucessfully tested up to 100 MV/m with SLAC mode launcher (below) |E| r z |B| r z

Constant gradient design 72 cells, active length 750 mm Relatively open structure: mean aperture 9.1 mm Average gradient 40 MV/m (30 MeV voltage) with 29 MW input power Group velocity variation: 1.6-3.7% Fill time: 100 nsec Average Q: 7150

Lower dipole band versus cell number Measuring the alignment via dipole wakes Lower dipole band versus cell number May be damped by output coupler, strong coupling to beam Trapped modes Strong coupling to beam Modes may couple to input coupler Weak kick (nowhere coupling to fsync ) From distribution, we see distinct frequency bands

Cell 36 as upstream monitor See contributions from the first half of the structure in the band 15.3-15.8 GHz

Cell 63 as downstream monitor Restricted by bandwidth of dipole band: Contributions from cells 40-63 Signal bandwidth 15.8-16.1 GHz

HOM coupler a la NLC DDS Electric short on one side TE type coupling minimizes spurious signals from fundamental mode and longitudinal wakes Need only small coupling (Qext<1000) for sufficient signal Minor loss in fundamental per- formance: 10% in Q, <2% in R/Q Output wave guides with coaxial transition connecting to measurement electronics Axial signal output wave guides

Modelling transverse wakes via Bane-Gluckstern equivalent circuit model Quasi physical model of hybrid two mutually coupled chain of resonators describing TE and TM components. Parameters derived from numerically calculated dispersion curves and kick factors Advantages: Modelling lowest two dipole bands Very good fit of dispersion relations Physically reasonable variation of kick factor with phase. K.L.F. Bane, R.L. Gluckstern, The transverse wakefield of detuned X-band accelerator structure, SLAC-PUB-5783, 1992 Extension to include damping manifold: R.M. Jones et al., PRST-AB 9:102001

Transverse wake spectrum

Output signal spectra

Alternative time domain measurement? An approximate idea about time domain behavior (looking only at cell 45) As the beam passes cell 45, it will excite a backward wave at the synchronous frequency The wave will travel back to where this frequency has a phase advance of pi, where it is reflected From there it advances forward until it reaches the point with phase advance zero (cell 63..) Wake signals from different cells (with their respective synchronous frequency) will arrive with a different delay at the coupler … Alternative time domain measurement?

Signal envelopes of wake monitors Signal at time t is correlated with frequency – is correlated with cell number….. Can we learn something about internal misalignments?

Beam axis Structure tilt Tilted Ref. - offset

What limits the resolution? Dilution by Noise in RF front – only an issue for low bunch charges Spurious signal from fundamental, longitudinal wakes - negligible due to TE coupling, waveguide length plus (if still necessary) additional filtering Random misalignment of individual cells:

Comparing random misalignment systematic offset Comparing both signals gives a estimate for the resolution being double the random cell to cell misalignment Syst. offset

Validation results with full structure simulation using the ACE3P codes from SLAC Computing full structure using ACE3P codes on NERSC super computer Franklin (Cray XT4) Goals: Validate electrical/mechanical design Get refined information about longitudinal/transverse wakes See influence of mechanical tolerances on performance

The accelerating mode |E| of 5 π/6 mode Below: monitor at cell 36 66 cell substructure: Omit power couplers, matching cells 500’000 elements, 10’000’000 unknowns (3rd order approach required) Computed resonance frequency: F = 11.99235 GHz (w/o losses) ~ F=11.9912 GHz (including losses) Design: F=11.991648 GHz |E| of 5 π/6 mode Below: monitor at cell 36

Ez on axis Some reflections due to monitors Monitor locations |E| [a.u.] Some reflections due to monitors z [mm]

Phase error of accelerating gradient Monitor locations Δφ [deg] z [mm] Only 25 deg phase error! Excellent!!

Driven problem in s3p Include matching cells and power couplers Include conduction losses Using 2nd order elements (somewhat higher numerical error)

Ez on axis

Phase error

Acceleration at nominal power (29 MW) V [MeV] F [GHz]

Reflection at input port S11 [dB] F [GHz]

Transverse wakes from eigenmodes 72 cell model (omitting power couplers) 2nd order elements Comparison to equivalent circuit distorted by Matching cells Waveguide stubs

Transverse spectrum Im(Zt) [kOhm/mm/m] F [GHz]

Comparing both to TD wake calculation WT [V/pC/mm/m] TD wake uses full structure (including power couplers) and include damping effects of all waveguide ports!! s [m]

Mechanical model Each two structures for structures for PSI (SwissFEL) and ST (Sincrotone Trieste) with wakefield monitors under fabrication Wakefield monitor details 33

Prototype disks from VDL ETG – flatness profile Done by Didier GLAUDE 34

Short test stack done with diffusion bonding Bonding at 1040°C for 90 minutes under H2 Metallurgical polishing + etching 75 s in Ammonium peroxodisulfate (NH4)2S2O8 35

Site of Interest 1: Outer side of disc stack Done by Markus AICHLER 21/07/2010 Site of Interest 1: Outer side of disc stack Grains grew down across the joining plane Joining plane 36

Summary Design for joint CERN/PSI/FERMI X band structure Special challenge transverse wake fields 5π/6 design: simultaneously open and efficient Wake field monitors Separate decoupled signals for up and downstream half Additional information from time domain envelopes of output signals Resolution of 10 μm determined by internal structure alignment Validating of circuit simulations/full structure simulations using SLAC codes S3P/Tau3P Successful metrology and bonding tests with prototype disks at CERN Delivery of disks for first two structures expected until the end of the year. ☺ Big thanks to ACE3P group at SLAC for access to codes, lots of support, specially A. Candel, R. Lee, Z. Li ☺