Efficient RF sources for Linear Accelerators

Dr Chris Lingwood

Motivation

• LARGE numbers of RF sources are required for future linear colliders.

• According to CDR 2012 - CLIC requires 1638 @ 15 MW

• Supply large amount of power at affordable cost (high efficiency)

• Current state of the art– 15 MW klystrons can achieve

65% efficiency

From CLIC CDR (2012)

CLIC MBK Study• Collaboration with CERN and Thales (Erk Jensen, Igor Syratchev,

Phillipe Thouvenin, Rodohple Marchesin).• Efficiency as main target• Evaluated configuration options, multiple beam klystron• Targeted a conservative (plausible) design• Targeted TESLA/ILC specification• Theoretical efficiency: 80% (beyond state of the art)

• Low perveance leads to higher efficiency.

• Low current -> lower space charge forces -> better bunching -> higher efficiency

• 20 beams – trade off between beam voltage and complexity due to beams

Why many beams?

Ib = 8.2AVb = 115V

𝑝𝑒𝑟𝑣𝑒𝑎𝑛𝑐𝑒𝐾=𝐼

𝑉 3 /2

Cavity choices

5. Whispering GalleryTM 10 1TM 0 1

3. & 4. Coaxial Cavity

1. Re-entrant 2. Recessed Re-entrant

TM 0 1• Comparison of multiple cavity types.

• Re-entrant & HOM cavities -> Low R/Q

• Recessed re-entrant and coax cavity -> high R/Q

Interaction structure• Optimised 6 cavity• (single 2nd harmonic)• Low R/Q structure 70%

• High R/Q structure 20 beam structure up to 80%

Optimisation

• Developed and published a new way to design klystron amplifiers:– ~14 Decisions (frequencies, drifts, Qe’s)– 3-4 objectives (efficiency, length,

bandwidth, slowest electron– 5000-10,000 evaluations

• Novel publishable optimisation concepts (recombination operator).

• Impractical without high throughput computing (CI HTCondor Pool)• Use spare clock cycles of desktop pcs

So it’s all done then?

• Conservative approach lead to complex tube• Many, many, many beams

• Push the voltage (always the plan)• Don’t rule out newer techniques• Be braver on layout

• Fundamentally: do you want 50MW in an MBK?

Cavity HOMsCPI(estimated)

Ours

Quadrupole

Cavity radius

Nor

mal

ised

freq

uenc

yR

Can model coaxial cavity as a piece of ridged waveguideVery good agreement even for HOMs

• Large diameter (35cm) at 15MW• More power -> more beams ->

larger still– Dipole mode gets closer for

larger cavities

Current Collaboration with CERN• Working with I. Syratchev, C. Marelli• Attempting to formalise empirical relationship

between efficiency and perveance

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.160

65

70

75

80

85

xNo harmx2nd Harm

Proposed task

• Many questions still surround the RF sources• Requirements still push state of the art

– Power, efficiency, cost, lifetime

• Evaluate options (SBK, MBK, MBIOT…)• Solution is probably klystrons

– Configuration (MBK/SBK)

• Work towards helping CERN become the intelligent customer

Maximising Efficiency

• Tight bunching isn’t enough• The key to higher efficiency is:

– slow all your electrons down as much as possible in the output gap without stopping or reflecting them

• This isn’t trivial• Potential improvements

– Low frequency penultimate cavities– Travelling wave output structures

Higher harmonic cavities

• To get the tightest bunch from a single cavity– sawtooth waveform (includes high harmonics)

• 2nd harmonic cavities well understood• Will 3rd or higher harmonic cavities help more?

– Effect on bandwidth?– Effect on velocity spread?

Reduce velocity spreadIVEC 2013

• Detune penultimate cavity to achieve π phase change– When phase change is good, coupling to the

beam is bad– Two gap cavity can help with control

• Tested in 63 W C-band tube, increase efficiency by 8%, 25% reduced voltage.

• Does this approach scale to MW?

Klystron Configuration• Some interesting configuration

options under consideration.• Some beyond state of the art• Some just beyond ….

New software• A number of klystron specific codes exist. Only one

generally available– AJDisk

• For instance cavity voltages can be unreliable disagree as much as 50% between GdfidL and AJDisk and 1500% (!) with klys2D (Thales).

• Closed source so difficult to identify the issue• Also difficult to integrate with other codes• Propose to develop new “open” disk model code

for klystron research from existing code at Lancaster.

PIC Simulations

• Simplified models can only get us so far• Detailed verification of designs to demonstrate

improvements• 1 week simulation time for klystron 1GHz up to 10 μs

– 8 cores• Scope for improvement with HPC (available at

Hartree Centre, Sci-Tech Daresbury)• Careful benchmarking of code (V-SIM) against

MAGIC

MB-IOTs for Linear Accelerators• Reduce power lost to collector by switch beam• Short “excursions” above rated power

allowable.• Very useful when high efficiency and variable

output power are needed.• Even better if “headroom” is needed

• IOTs exist up to about 100kW• Not a great deal in the context of proposed

large LINACs

• Just like klystron MBK -> MB-IOT• 10 beams -> 1MW, more plausible.• ESS seriously considering.

• Road block - guns

• Worth doing the sums.

Klystron

IOT

Figures from IOT based High Power Amplifiers, Morten Jensen, TIARA Workshop on RF Power Generation for Accelerators – Uppsala June 2013

Milestones• 2014 Investigate suitability of single and multi-beam

klystrons, MBIOTs, magnetrons. • 2014 Produce a bunched beam vacuum tube model.

Transcode and improve existing code to interface with PIC codes for output cavity. Benchmark against existing codes.

• 2015 Evaluate new and existing techniques to improve efficiency.

• 2015 Benchmark V-SIM and CST against MAGIC for bunched vacuum tubes.

• 2016 Design most appropriate tube and validate using PIC.

Deliverables

• 2014: Tube model complete and open sourced• 2015: Design of tube interaction structure for

drive beam - report• 2016: PIC simulations and verification of

proposed interaction structure - report

FinancialStaff (Lingwood) 1 1 1 3RA3 6+6 6+6 6+6 18+18Student (TBD) 12 12 12 36Materials 5+7 4+5 0+0 9+12Travel 1+2 1+2 1+2 3+6

1 RA and 1 Phd student for three years

Summary• Re-evaluate options for RF sources• Push efficiency and plausibility• Scope for improvement in simulation times• Develop more flexible and open klystron

simulation tools.

• Produce candidate structure using lessons learned.