Presentation on theme: "International Linear Collider (ILC) Linac Basics Part I: General Design Constraints Part II: ILC Design Choices Chris Adolphsen SLAC."— Presentation transcript:
1International Linear Collider (ILC) Linac Basics Part I: General Design Constraints Part II: ILC Design Choices Chris Adolphsen SLAC
2Part I: General Design Constraints Linear collider design is complex due to the interrelationships among the various parameters and the soft constraints on their values.I will give one or many possible descriptions of the rationale behind the ILC linac design.Bob Palmer1990
31.3 GHz TESLA CavitiesMade with solid, pure niobium – it has the highest Critical Temperature (Tc = 9.2 K) and Thermodynamic Critical Field (Bc ~ 1800 Gauss) of all metals.Nb sheets are deep-drawn to make cups, which are e-beam welded to form cavities.Cavity limited to ~ 9 cells (~ 1 m Long) to reduce trapped modes, input coupler power and sensitivity to frequency errors.Iris radius (a) of 35 mm chosen in tradeoff for low surface fields, low rf losses (~ a), large mode spacing (~ a3 ), small wakes (~ a-3.5 ).
5The BasicsThe low rf surface losses in superconducting cavities allow essentially 100% RF-to-beam transfer efficiency in steady state:RF Input Power = Cavity Voltage * Beam CurrentFirst look at what limits cavity voltage (gradient * cavity length)Note: highest gradient not always cost optimal !
6Cavity Operating Parameters Operate at 2K (in super-fluid He) to reduce resistivity:1.3 GHz frequency (f) chosen to reduce power loss, sensitivity to thermal instabilities, wakes, and cavity size and to match available sources.At Qo = 1e10, cavity time constant (Qo/w) ~ 1 sec, and at 35 MV/m and 0.1% duty factor, average power loss ~ 1 W/m (but it takes ~ 1 kW/m of AC power to remove heat).188.8.131.52.53.14.6Qo ~ 2e10at T = 2KTemperature (K)Niobium Surface Resistance
7Operating GradientQo varies with gradient due to a number of mechanisms.In recent years, gradients are approaching fundamental limit: Bc * (Grad / B surface) ~ 1800/41.5 ~ 43 MV/mBest 9 Cell Cavity Result to Date: CW Performance of a Cavity Electro-Polished at DESYQoGradient (MV/m)
8The Basics (2)The low rf wall losses in superconducting cavities allow essentially 100% RF-to-beam transfer efficiency in steady state:1.3 GHz Input Power = MV/m * 1m * Beam CurrentNext look at beam structureBeam Current = Bunch Charge / Bunch SpacingBunch Train Length = Number of Bunches * Bunch Spacing
9Bunch Charge and Length Nominal Bunch Charge (N = 2e10) and Length (sz = 300 mm)Mainly determined by damping ring, linac energy spread and IP considerations.Bunch length reduced from 6 mm to 300 microns prior to linac injectionAlso constrained byShort-Range Transverse Wake Kicks (N sz / a 3.5 )Short-Range Loading (N / sz / a 2 )Wakefield in a PETRA cavityBunchaIrisRadius
10Number of Bunches per Pulse, Repetition Rate, Luminosity and AC Power Luminosity ~ Rep Rate ´ Number of Bunches per PulseRepetition rate (5 Hz) constrained by damping ring store time (see next slide).Number of Bunches per Pulse constrained byTrain Length > Cavity Fill Time: for > 50% rf-to-beam efficiency, minimum number of bunches is1870 (at 35 MV/m and 2e10 e/bunch) independent of bunch spacing !Length and cost of damping rings (newer, smaller designs assume smaller bunch separations).Train Length < Max klystron pulse length (limited by pulse heating).Product constrained by practical site power (few hundred MW). Would limit rep rate even damping rings did not.
11Damping Ring Constraints Optimal bunch train length very long (>> linac length), soMinimize bunch spacing in the damping ring – limited by separation (3-20 ns) required to extract individual bunches with a kicker magnet.Still, damping ring is long (17 km circumference in TDR, 6 km in current design), which makes the required store time long (200 ms), even with a few hundred meters of wigglers.RF pulse length (1.4 ms) << store time so ring needs to hold full bunch train.Store time limits machine repetition rate (5 Hz). Could increase wiggler length, but already about 20% of damping ring cost.Dog Bone Damping Ring
12TESLA Design Choices (2001 TDR) Gradient = 23.4 MV/mBunch Charge = 2e10 eRep Rate = 5 Hz# of Bunches = 2820Bunch Spacing = 337 ns Beam Current = 9.5 mAInput Power = 230 kWFill Time = 420 msTrain Length = 950 ms
13Bunch Spacing Non-linac constraints on minimum spacing Peak positron target heatingIP bunch separationWeak linac constraintsBunch coupling from long-range transverse wakefieldsSteady state RF-to-Beam efficiency: w/Qo very high (much different than for a warm machine)Strong linac constraintsPower source costs ~ 1 / Bunch SpacingCryogenic costs ~ Bunch Spacing
14Klystron EconomicsCost of 1.3 GHz Klystron + Modulator (crude approximation)Similar for similar average powerIndependent of peak power.However, rf energy per pulse differs greatly, for example,20 MW peak, 10 msec pulses (300 Hz, 60 kW)5 MW peak, 2 msec pulses (10 Hz, 100 kW)0.1 MW peak, CWNumber of klystrons~ Cavity Input Power / Peak Klystron Power~ 1 / Bunch Spacing (with fixed peak power)With smaller bunch spacing, would not use full average klystron average power capability.
15Bunch Spacing and Fill Time Adjust Qext to match cavity impedance (R/Qo * Qext) to the beam impedance (Gradient / Current). So Fill Time ~ Bunch SpacingFor TDR parameters, Qext = 3e6 so cavity BW = 430 Hz.Need to achieve < 0.1% energy gain uniformity.
16Other Bunch Spacing Considerations Input coupler power limitationsPower ~ 1 / Bunch SpacingBaseline TTF3 design processed to 1 MW and tested up to 600 kW for 35 MV/m operation (1000 hours): long term reliability for required operation at 350 kW not known.InputPowerCoaxial Power Coupler
17TESLA Design Choices (2001 TDR) Gradient = 23.4 MV/mBunch Charge = 2e10 eRep Rate = 5 Hz# of Bunches = 2820Bunch Spacing = 337 ns Beam Current = 9.5 mAInput Power = 230 kWFill Time = 420 msTrain Length = 950 ms
18(NLC/GLC vs TESLA TDR Efficiencies and Average Power) Where Does the Power Go(NLC/GLC vs TESLA TDR Efficiencies and Average Power)
19Cost OptimizationMajor cost components that depend on Gradient (G) and Bunch Train Length (Tb).Cooling Power Length~ A ´ Tb ´ G B ´ Tb C ´ G -1Cost Study: Compute Cost vs G and Tb for fixed Luminosity (L)Assume charge per bunch and number of bunches constantCavity fill time (Tf) scales as G * TbRF pulse length (Trf) = Tb + Tfill
20Relative Total Project Cost* (TPC) -vs-Linac GradientRelative CostGradient ( MV/m)* TPC is for 500 GeV machine in US Options Study but does not include additional unpowered tunnel sections.
26Operational Parameter Plane Rather than a specific machine design, the baseline configuration is for an operating ‘plane’.Four parameters sets proposed to achieve design peak luminosity:Nominal.Low bunch charge (Q).Large vertical IP beam size (Large Y).Low Beam Power (Low P).Not strictly fixed sets -> used to define necessary operational flexibility.
28Main Linac DesignBaseline Configuration Document (BCD) distilled from Snowmass Working Group recommendations in August 2005.Major differences from 2001 Tesla TDR 500 GeV Design.Higher gradient (31.5 MV/m instead of 23.4 MV/m) for cost savings.Two tunnels (service and beam) instead of one for improved availability.The Linac Area Group of the Global Design Effort (GDE) is continuing to evolve design.
291.3 GHz TESLA CavitiesFor ILC, would accept only ‘vertically’ tested cavities (using CW rf without high power couplers) that achieve gradients > 35 MV/m and Q > 8e9 (discard or reprocess rejects).When installed in 8 cavity cryomodules, expect stable operation at an average gradient of 31.5 MV/m and Q = 1e10 (rf system designed for 35 MV/m).Derating due to desire for overhead from quench limit, lower installed performance and limitations from using a common rf source.For a 1 TeV upgrade, expect average gradient = 36 MV/m, Q = 1e10 for new cavities (the TDR 800 GeV design assumed 35 MV/m and Q > 5e9).
30ILC Linac RF Unit (1 of ~ 600) Gradient = 31.5 MV/m Bunch Charge = 2e10 eRep Rate = 5 Hz# of Bunches = 2967Bunch Spacing = 337 ns Beam Current = 9.5 mAInput Power = 311 kWFill Time = 565 msTrain Length = 1000 ms(8 Cavities per Cryomodule)
31Achieved Gradients in Single and 9-Cell Cavities In recent years, single-cell cavity gradients approached fundamental limit: Bc * (Grad / B surface) ~ 1800/41.5 ~ 43 MV/m for Tesla-shape cavities.During past 2.5 years, DESY has produced 6 fully-dressed cavities with Gradients > 35 MV/m and Q > 8e9. Yield for such cavities < 30%.Test Results for Dressed-Cavities that will beused in a ’35 MV/m’ Cryomodule
32Main Production Problem Has Been Poor Reproducibility ILC GoalGradients achieved over time in DESY cavities
33Achieved Gradients in Tesla Test Facility (TTF) 8-Cavity Cryomodules (Cavities not Electro-Polished)Diamonds and Error Bars = Range of Gradients Achieved in Individual CW Cavity Tests. = Average Gradient Achieved in CryomoduleGradient (MV/m)Cryomodule Number
34High Gradient R&D: Low Loss (LL) and Re-Entrant (RE) Cells with a Lower Bpeak/Eacc Ratio
36Studies also underway using single or large grain Nb – could eliminate need for Electro-Polishing (EP)BCP + 120C Baking
37Tuning the CavitiesBoth slow (500 kHz over minutes) and fast (2.5 kHz during the 1.6 ms pulse) tuning required – achieve by compressing the cavity (~ 1 micron per 300 Hz).Want tuners located away from cavity ends to minimize cavity spacing.‘Blade Tuner’ shown below. To date, have not achieved more than ~1kHz range of fast tuning. Final design for BCD not yet chosen.
38RF Fill DynamicsFor ILC, Qext = 4e6 so cavity BW = 325 Hz (DL = 1 micron).Need to achieve < 0.1% energy gain uniformity with Low Level RF (LLRF) systemFeedback to maintain constant ‘sum of fields’ in 24 cavities
40RF Distribution Math (for 35 MV/m Max Operation) 10 MW Klystron35 MV/m * 9.5 mA * m = 345 kW (Cavity Input Power)24 Cavities1/.93 (Distribution Losses)1/.89 (Tuning Overhead)10.0 MW
41Pulse Transformer Style Modulators (115 kV, 135 A, 1.5 ms, 5 Hz)(~ 2 m Long)To generate pulse, an array of capacitors is slowly charged in parallel and then discharged in series using IGBT switches. Will test full prototype in 2006Pulse Transformer Style
43Marx Generator Modulator 12 kV Marx Cell (1 of 16)IGBT switchedNo magnetic coreAir cooled (no oil)
44Modulators Baseline: Pulse Transformer Alternative: Marx Generator 10 units have been built over 10 years, 3 by FNAL and 7 by industry.8 modulators in operation – no major reliability problems (DESY continuing to work with industry on improvements).FNAL working on a more cost efficient and compact design, SLAC building new dual IGBT switch.Alternative: Marx GeneratorSolid state, 1/n redundant modular design for inherent high availability, reliability.Highly repetitive IGBT modules (90,000) cheap to manufacture.Eliminating transformer saves size, weight and cost, improves energy efficiency.
45SNS High Voltage Converter Modulator (Unit installed at SLAC) Other AlternativeModulatorsSNS High Voltage Converter Modulator (Unit installed at SLAC)RECTIFIER TRANSFORMERAND FILTERSBOOST TRANS-FORMERHV RECTIFIERAND FILTER NETWORKSCRREGULATORENERGYSTORAGESWITCHING3Ø(ON/OFF)4mH400A6 EACHRTNAØBØCØ-HV10ohm20mH.03uF.05uFVMONHVOUTPUT13.8KV3ØINPUTLINE CHOKE5thHARMONICTRAP7th50mHAØBØCØRECTIFIER TRANSFORMERAND FILTERSSCRREGULATOREQUIPMENTCONTROL RACKHVCM
46Series Switch Modulator (Diversified Technologies, Inc. ) IGBT Series Switch140kV, 500A switch shown at left in use at CPIAs a Phase II SBIR, DTI is building a 120 kV, 130 A version with a bouncer to be delivered to SLAC at the end of 2006
47KlystronsBaseline: 10 MW Multi-Beam Klystrons (MBKs) with ~ 65% Efficiency: Being Developed by Three Tube Companies in Collaboration with DESYThalesCPIToshiba
48Status of the 10 MW MBKsThales: Four tubes produced, gun arcing problem occured and seemed to be corrected in last two tubes after fixes applied (met spec). However, tubes recently developed other arcing problems above 8 MW. Thales to build two more without changes and two with changes after problem is better diagnosed.CPI: One tube built and factory tested to 10 MW at short pulse. At DESY with full pulse testing, it developed vacuum leak after 8.3 MW achieved – has been repaired and will be tested again.Toshiba: One tube built and achieved operation spec but developed arcing problems above 8 MW – being shipped to DESY for further evaluation.These are vertically mounted tubes – DESY will soon ask for bids on horizontally mounted tubes for XFEL (also needed for ILC).
49Alternative Tube Designs 5 MW Inductive OutputTube (IOT)10 MW Sheet BeamKlystron (SBK)Low Voltage10 MW MBKParameters similar to10 MW MBKVoltage 65 kVCurrent 238AMore beamsPerhaps use a DirectSwitch ModulatorKlystronOutputIOTDriveSLACCPIKEK
50Klystron SummaryThe 10 MW MBK is the baseline choice – continue to support tube companies to make them robust (DESY needs 35 for XFEL although will run at 5 MW).SLAC funding design of a 10 MW sheet-beam klystron (will take several years to develop).Backup 1: Thales 2104C 5 MW tube used at DESY and FNAL for testing – it appear reliable (in service for 30 years) but has lower effiency compared to MBKs (42% vs 65%).Backup 2: With increased DOE funding next year, propose to contract tube companies to develop high efficiency, single-beam, 5 MW klystron.
51RF DistributionBaseline choice is the waveguide system used at TTF, which includes off-the-shelf couplers, circulators and 3-stub tuners (phase control).
52Need more compact design (Each Cavity Fed 350 kW, 1 Need more compact design (Each Cavity Fed 350 kW, 1.5 msec Pulses at 5 Hz)Two of ~ 16,000 Feeds
53And should consider simplifications (circulators are ~ 1/3 of cost) BaselineAlternative Design with No Circulators
54Adjustable Tap-Offs Using Mode Rotation rotatable jointsloadfeed1234mode rotatora2aRotatable section with central elliptical region, matched for both polarizations of circular TE11 mode with differing phase lengths.C. Nantista
55Proposed RF Distribution Layout Adjustable power to pairs of cavitiesNo circulatorsPairs feed by 3 dB hybrids (requires nl +/- 90 degree cavity spacing – only 7 mm longer than TDR/BCD spec)loadscavity couplersflexible waveguidethree-stub tunersdiagnostic directional couplers1.326 mbeam direction3-dB hybridsC. Nantista
56Cryomodules CryoCap Oct 96 50 M1 Mar 97 5 M1 rep. Jan 98 12 M2 Sep 98 TTF ModuleInstallation dateCold time [months]CryoCapOct 9650M1Mar 975M1 rep.Jan 9812M2Sep 9844M3Jun 9935M1*MSSJun 02308M3*M4M5Apr 0319M2*Feb 0416
58Multipacting Simulation of TTF3 Coupler “Cold Side”BellowsPrimaries -Green, Secondaries- RedThese are the first results from the simulation with the eigensolver Omeag3P. They are only possible with new advances in nonlinear eigensolvers which I will touch on in the next slide, Because the couplers are extracting power from the cavity, the egienvalues are complex from which we get what they call the external Q of the modes. We have calculated the external Qs of the HOMs in both the DESY and KEK design and the design challenge now is to optimize the end cells of the cavity to lower the external Qs of the modes to below a threshold so that the beam will be stable. There is a slide on shape optimization later
59Cryomodule Design Relative to the TTF cryomodules Continue with 8 cavities per cryomodule based on experience and minimal cost savings if number increased (12 in TDR).Move quad / corrector / bpm package to center (from end) to improve stability.Increase some of cryogenic pipe sizes (similar to that proposed for the XFEL).Decrease cavity separation from 344 mm to 283 mm as proposed in the TDR.
60Beam-Related Design Issues Optics / Tolerances / Operation similar to that in TDR:One quad per rf unit (three, 8-cavity cryomodules).Few hundred micron installation tolerances for cavity, quad and BPM (demonstrated with TTF cryomodules).Cavity BPM resolution of a few μm (should be achievable).Use quad shunting and DFS tuning algorithms for dispersion control (need to better understanding systematic effects).Assume beamline will follow Earth’s curvature.XFEL will serve as a benchmark although emittance much larger.Alternatives for cost savings.Larger quad spacing at high energy end of the linac where wake and dispersion effects smaller.Halve quad and bpm aperture to allow superferric quad and higher resolution BPMs (increases wakes by 10%).
71ILC Availability Challenge The ILC will be an order of magnitude more complex than any accelerator ever built.If it is built like present HEP accelerators, it will be down an order of magnitude more (essentially always down).For reasonable uptime, component availability must be much better than ever before. Must do R&D and budget for it up-front.
73SC Linac SummaryA superconducting linac allows for efficient acceleration – for the ILC with a 35 MV/m gradient,60% RF-to-Beam Efficiency15% AC-to-Beam EfficiencyA low beam current is required to make this approach cost competitive (matches rf source capabilities).The resulting long bunch trains require large damping rings with untested designs.
74ILC Design Summary Basic linac design complete: converging on details Tradeoffs of operability, availability and cost.Major cost and technical risksProducing cryomodules that meet design gradient at a reasonable cost (cost model still in development, XFEL will provide a reference, and will get new industry-based estimates).Producing a robust 10 MW klystron.Potential Cost SavingsAdopt Marx ModulatorUse simpler rf distribution schemeHave one tunnel although ‘the additional cost is marginal when considering the necessary overhead and equipment improvements to comply with reliability and safety issues.’Reduce cavity aperture to 60 mm for 21% reduction in dynamic cryo-loading and 16% reduction in cavity fill time.