1. Klystrons C Lingwood, Lancaster University/Cockcroft Institute on behalf of HEIKA (High Efficiency International Klystron Activity)

2. Introduction Increase in RF generation efficiency is high priority for the future large accelerators (CLIC, ILC, FCC, ESS) “Recent” klystron developments targeted high power neglecting high efficiency Few high power klystrons offer 65%+ efficiency Fewer with long pulse length Proton machines can piggy back on developments for large projects

3. Future Machines…large numbers! FCC FCC ee : CW, 0.8 GHz, P RF total= 110 MW circular linear ILC e+e- : Pulsed, 1.3 GHz, P RF total= 88 MW 0.5 TeV 3.0 TeV CLIC e+e- : Pulsed, 1.0 GHz, P RF total= 180 MW Achieved efficiency at 10 MW peak RF power level The existing MBK klystrons provide efficiency very close to 70%

4. Operating Principal (Velocity modulation) DC beam passes through input Electrons accelerated or decelerated according by the gap voltage – Beam is velocity modulated Bunches of electrons are formed – Beam is spatially modulated Output cavity is excited by the bunches Power is coupled out to load RF output RF input Collector Cathode (DC Beam) Thanks M Jensen More cavities give better bunching. – (Very) Roughly ¼ plasma wavelength apart

5. Traditional Approach For high efficiency traditionally we chase low perveance: – High voltages or – Low currents (many beams) For high power both become “unpleasant”. Performance limited by the slowest electrons (must avoid reflecting electrons) Traditional theoretical efficiency limited to 80% @ 0.1- 0.2ish microperv

6. Klystron Technology Limitations (Frequency) Low frequency klystrons are BIG Lower bound is something like 300 MHz Depends how many are needed How much space you have If something else works… ESS have 352 MHz 2.8 MW klystrons Tetrode might be nice… Upper bound need not concern us

7. Klystron Technology Limitations (Power) Pulsed power can go very high ~3-4 ms is more or less CW (ish) Limited by output window to around 1 MW average (Toshiba sell 1.2 MW CW at ~1.2GHz) More windows = more power = more complexity = more cost Convert from short pulse tube to long pulse Bigger collector Windows

8. Why not 100% Efficient The simple answer is – Imperfect bunching – Residual Velocity – energy still in the beam Significant charge outside bunch. Velocities aligned… Many electrons miss bunch. Significant energy left in bunch! Bunching monotonic – electrons move to center of bunch

9. Collector depression Decelerate the electrons into the collector to recover energy Efficiency increases with number of stages Hard to cool electrodes Adds to the complexity and cost of the tube Easier to just to design the previous parts better

10. The (massive) problem with protons Unlike electron machines, need a ramp in power – heavy so not relativistic at lower energies Range of powers of sources End up underrating tubes Also control overhead is Needed so tubes cannot be run at saturation ESS Power per Cavity, thanks ESS, M Jensen

11. Underrating Constant voltage and current Manipulate input power Constant impedance Constant voltage Manipulate current with mod anode Constant perveance Drop modulator voltage (easy)

12. Mod annode Allows you to run at a lower output power Adds ceramic to copper brazes Risk of arcing + reduced reliability “Better” to just reduce operating voltage. 12

13. Anything else we can do? If you drop the voltage you change the reduced plasma wavelength Cavities are no longer well spaced so efficiency reduced Change the output Q to extract more power and restore some efficiency Unstable Power Efficiency Thanks to C Marrelli, ESS

14. State of the art? ESS most recently acquiring long pulse tubes – 352 MHz – 704 MHz Can argue this is more or less the proton linac klystron state of the art

15. ESS 352 MHz Klystron Frequency: 352MHz Output power: 2.8 MW Voltage: 108 kV Current: 46.5 A Pluse width (catalogue 1.5ms): 3.5 ms Efficiency: 53%

16. ESS 704 MHz Klystron Nominal output power1.5 MW Frequency704.42 MHz BW≥ +/- 1 MHz Pulse width3.5 ms Repetition rate14 Hz Perveance0.6*10 -6 Efficiency>60% VSWRUp to 1.2 Power Gain≥ 40 dB Group Delay≤ 250 ns Harmonic Spectral content≤ -30 dBc Spurious Spectral content≤ -60 dBc Three prototypes are being procured, from three different manufacturers (Thales, Toshiba and CPI) Delivery expected in March (Thales), May (Toshiba) and July (CPI) 2016 Thanks Chiara Marrelli

17. Other power requirements Klystron focusing magnet – < 20kW often <10kW – At 1 MW -> 0.1% (very hard to care) Cooling? Control and HV a modulator issue.

18. Potential Performance Deeper understanding of the klystron physics, new ideas and modern computational power will help us towards 90% efficiency HEIKA collaboration of many experts working towards this For protons most tubes under active development are either – Wrong frequency – Short pulse All is not lost technology should transfer – Frequency scaling is ok (so long as we don’t want Hz or THz) – Pulse length is down to heat dissipation

19. Method to get high efficiency Core Oscillation (space charge debunching) Bunching split into two distinct regimes: – non-monotonic: core of the bunch periodically contract and expand (in time) around center of the bunch – outsiders monotonically go to the center of the bunch Core experiences higher space charge forces which naturally debunch Outsiders have larger phase shift as space charge forces are small Very long, very efficient tubes result. Phase Traditional bunching Core oscillations Cavity Space Core oscillations Traditional approach

20. 90% Efficient Klystron Efficiency increases with number of core oscillations and reaches 88-90% for 4- 5 oscillations

21. Methods to get high efficiency BAC Method (I. Guzilov) Again based on core oscillations Interaction space is wasted “waiting” for space charge forces to debunch. A cavity can achieve the same thing in a shorter space by aligning electron velocities Structure half the length while maintaining efficiency. Thanks to SLAC

22. Electron velocity/density The fully saturated (FS) bunch Final compression and bunch rotation prepare congregating FS bunch. After deceleration all the electrons have identical velocities. Mission accomplished Process in the high efficiency klystron (bunch rotation)

23. HEKCW Tube HEIKA/HEKCW working team: I.I. Syratchev (CERN) II.C. Lingwood (Lancaster) III.G. Burt (Lancaster) IV.D. Constable (Lancaster) V.V. Hill (Lancaster) VI.R. Marchesin (Thales) VII.Q. Vuillemin (Thales/CERN) VIII.A. Baikov (MUFA) IX.I. Guzilov (VDBT) X.C. Marrelli (ESS) XI.R. Kowalczyk (L-3com) 16 beams MBK cavity R/Q = 22 Ohm/beam Tube parameters: 1.5MW Voltage: 46 kV Total current: 36A N beams: 16 µK/beamx10 6 : 0.213 N cavities: 8 Bunching method #1: COM HEKCW 3.54m

24. Magic HEKCW #08-03 Cavity 1 voltage, 0.83 kV: Nice Stable output No reflected electrons Slightly odd modulation current Stable in cavity 7 signal. Efficiency….. Output power Cavity 8 Volts Cavity 7 Volts 83%

25. Magic HEKCW #08-03 Electron Animations Bunch ”bounces” in output gap Superficially nightmarish phase space Good chunk decelerated well though Quite a good slab bunch but not great in at r=0 Cavity 7 Cavity 8

26. Onwards!

27. HEIKA Time Scales 0.5-1 years further R&D to get to optimal MBK 1.5 years technical design to build “Igor’s wish”: Test 2018 If all goes well and it is funded.

28. Other prospects

29. 5045 Retrofit SLAC, A. Jensen Retrofitting a 5045 S-Band) with BAC 60 -> 80MW 45% -> 55% 4 more cavities Plug compatible (needs new solenoid) X Band Short pulse Results will be reported at IVEC in April, 2016

30. S-Band Hardware Development VDBT I. GUZILOV The first commercial S‐band MB tube employs the new bunching technology: 40 beams Permanent Magnets focusing system Low voltage: 52 kV Peak power: 7.5 MW Efficiency: 77% (in simulations) Pulse length: 5 microsecond Repetition rate: 300 Hz Average power: 30 kW

31. Kladistron (Adiabatic Bunching) CEA, A. Mollard Bunch slowly to keep velocity spread small Many cavities Retrofit a TH2166 (5GHz, 56kW, 26kV) as proof of principal 50% -> 55% in preliminary redesign Soon ordering some components

32. RF source typeGain [db] Maximum output power pulsed [kW] Rise time [us] Pulse length range [ms] Rep rate range [Hz] Max. output power CW [kW] Efficiency at working point [%] High voltage needs [kV] Frequency range [MHz] Single Beam40-501000-3000<<1 (~300 ns) -4 <1200 kW 55 (65 max) ~90- 120kV.3GHz-1.5GHz MBK40-5010,000-15,000 (up to 1.5ms at least) <<1 (~300 ns) -4 <1200k W (no point) 60 (70 max) ~90- 120kV.3GHz-1.5GHz Future Single Beam --- -- -70+ 40-60 kV.3GHz-1.5GHz Future MBK------80+40-60 kV.3GHz-1.5GHz Performance Figures (well it depends) Health warning: some numbers mutually exclusive Cost: 300k€ - 1M€ depending on complexity, number, novelty

33. Conclusion and Outlook Using new bunching theory 80%+ looks possible for FCC/CLIC/ESS/ILC klystrons – No new materials or manufacturing techniques needed – Little additional complexity – Simply existing technology reconfigured – Prototypes for proof of concept in progress – Lower voltages combined with high efficiency appears achievable 83% Achieved in PIC 90% and stable in PIC so far elusive, but not ruled out Prototypes and further validation required Lessons learned directly applicable to proton driver tubes International collaboration at work