Generation and application of sub fs, ultra high brightness, multi GeV electron beams

Transcription

1 Generation and application of sub fs, ultra high brightness, multi GeV electron beams J.B. Rosenzweig UCLA Dept. of Physics and Astronomy X ray Science at the Femtosecond to Attosecond Frontier UCLA, May 18, 2009

2 Ultra short XFEL pulses Investigations at atomic electron spatio temporal scales Angstroms nanometers (~Bohr radius) Femtoseconds (electronic motion, Bohr period) 100 fsec accessible using standard techniques Many methods proposed for the fsec frontier Slotted spoiler; ESASE; two stage chirped pulse Unsatisfactory (noise pedestal, low flux, etc.) Still unproven Use clean ultra short electron beam Myriad of advantages in FEL and beam physics Robust in application: XFEL, coherent optical source, beyond

3 The clean path: ultra low charge electron beam Excellent phase space ( and ) Very low emittance Highly compressible Unprecedented high brightness Ultra short high brightnss beam in FEL Bunch ~ cooperation length; super radiant single spike Short cooperation length; sub femtosecond pulse Pedestal free ultra short pulse Mitigate collective effects dramatically CSR instability Undulator beam pipe wakes

4 Ultra high brightness beams: How? Brightness B = 2I ε n 2 Q σ t ε n 2 Low Q in injector: shorter σ t, smaller ε Velocity bunching at low energy, recover I Chicane bunching (1 or 2 stages) The rules change much in our favor Low charge makes all manipulations easier Higher brightness gives new possibilities, design application with open imagination Illustrate with original example: SPARX (2007)

5 Photoinjector scaling J.B. Rosenzweig and E. Colby, Advanced Accelerator Concepts p. 724 (AIP Conf. Proc. 335, 1995). Beam at lower energy is single component relativistic plasma Preserve optimized dynamics: change Q, keeping plasma frequency (n, aspect ratio) same Dimensions scale σ Q 1/ 3 Shorter beam i Emittances: At low Q, ε x,th dominates (Ferrario WP, SPARC/LCLS) ε n ( mm-mrad) = a 1 Q( nc) 2 / 3 + a 2 Q( nc) 4 / 3 + a 3 Q( nc) 8 / Q( nc) 1/ 3 ( mm-mrad) a 1 = 0.11 a 2 = 0.18 a 3 = 0.23

6 Velocity bunching Enhance current at low energy, avoid bending Inject near zero crossing Apply optimized focusing, manage ε evolution Longitudinal phase space schematic for velocity bunching

7 VB example: SPARX (2007), 1 pc σ x (mm) ε n (mm-mrad) z (cm) σ z (mm) z (cm) - ε growth manageable - 1 order of magnitude compression σ ζ 9 µm z (cm)

8 Thermalized Longitudinal Phase Space Space charge dominated longitudinal focus FLASH case (1 pc, L-band) Plasma type Q 1/3 scaling (original SPARX scenario) σ δp,th 30 Q( pc) 1/ 3 ( ) 1/ 3 [ ] kev/c σ z = 9 [ Q pc ] µm Results focusing dependent Better compression with lower focusing

9 Chicane Compression: the Picture Run off crest in linac, chirp longitudinal PS Remove chirp with chicane compressor Long beams not easy to compress, large longitudinal ε due to RF curvature ε ζ,rms = k RF 2 σ z 3 2

10 Chicane Compression by the Numbers Chicane operation unchanged from standard λ R 56 tan φ RF 0 k RF σ ζ <<1 2π δp th p 0 <<k RF σ ζ cot φ 0 Final bunch length/initial σ ζ * σ ζ = Dominated by thermal type spread from VB RF curvature effects negligible for Q<10 pc In low Q limit, SPARX scenario: ( ) 1 k 2 ( RFσ ζ ) 4 2 +σ δp,th ( ) 4 2 +σ δp,th + ( k RF σ ζ ) 2 cot 2 ( φ 0 ) σ δp,th k RF σ ζ 1 2 k RFσ ζ σ ζ * σ δp,th k RF cot φ 0 ( ) λ RF m ( )cot φ 0 ( )Q pc p 0 ( MeV)cot φ 0 ( ). ( ) 1/ 3 ( )

11 The original goal: single spike XFEL operation 1D dimensionless gain parameter 1D gain length ρ 1D = JJ ( K rms)k rms k p 4k u L g,1d = Cooperation length Single spike operation 4π 2 / 3 λ u 3ρ 1D L c,1d = 4π λ r σ b,ss < 2πL c,1d = 2 3ρ 1D λ r 3ρ 1D

12 Numerical example: INFN LNF Take 2 GeV operation, standard SPARC undulator, λ u =2.8 cm Peak current I=2 ka, Estimate single spike condition: σ b,ss = 0.48 µm (1.6 fsec) Note: with ultra small Q, ρ is enhanced Spike is a bit shorter FEL gain better ρ 1D =

13 Example: SPARX (3/2007) 2 GeV just before undulator Need nm σ * ζ 480 Choose to accelerate 23 forward of crest Deduce upstream beam of σ ζ = 9 µm Must produce from velocity buncher Check consistency with energy spread σ δp cot( φ 0 )( k z σ ζ ) << ρ 1D We have σ δp << ρ 1D ( > ) Final (full) energy compression OK (not for LCLS )

14 Low charge working point Velocity bunching gives σ ζ σ 0 Q nc Need σ 0 10σ ζ 90 µm (0.3 psec) Not that short factor of 10 below present Thus, work with 10 3 smaller Q, or 1 pc Emittance > compensated value Growth in velocity buncher Extremely high brightness beam predicted! ( ) 1/ 3

15 Beam simulations SPARX 1 pc case Summary (3/2007, UCLA PARMELA) Charge 1 pc (6.2E6 electrons) Laser pulse length (full) 1 psec (280 fsec rms) Gun maximum on-axis electric field 110 MV/m Average traveling wave section field 13.5 MV/m Initial laser beam radius (full) 100 microns Thermal emittance mm-mrad Emittance after velocity bunching mm-mrad Final bunch length (rms) 9 m (28 fsec) Energy after velocity bunching section 17.9 MeV Final relative momentum spread 0.31%

16 Compressed beam at SPARX I (A) SPARX example, compress at 2 GeV, σ δp = << ρ FEL Compressor: R cm θ b = 25 mrad Analytical est. of growth in ε, σ δp due to CSR: Δσ δp 10 5 Δε n m-rad Both consistent with simulation With I=260 A, and B = A/m 2 Two orders of magnitude enhanced! t (fsec) ε n = m-rad

17 Genesis simulation: SPARX (3/2007 case) Standard undulator, focusing Start to end from PARMELA/Elegant Do not take advantage of lower ε by changing β (must worry about) diffraction "Z R "= 4πσ x 2 / λ r = 83 cm Single spike operation

18 FEL peak power, profile evolution z Saturation in <30 m ζ Extremely clean pulse Single spike past saturation

19 Power profile Current profile FEL power profile 220 MW peak power (10 below 1 nc design) <1 femtosecond rms pulse. March 2007, I was convinced Shorter than electron beam

20 Spectral properties Nearly Fourier transform limited at onset of saturation σ ω σ t =1.3

21 SPARX design: new regimes Explore velocity bunching optimization Higher charge cases Easier beam measurements More photons for users Examine tolerance issues in realistic design context Many similarities, lessons for PSI XFEL case

22 Push to shorter pulses at SPARX (lowest charge) Q=1 pc Beam transverse size, emittance Beam length during velocity bunching Study 1 and 10 pc case. Use LANL PARMELA (standardization) Relax focusing in velocity bunching, lower ε growth, shorter beam For 1 pc, σ z only 4.7 µm after velocity bunching

23 Push to shorter pulses at SPARX (lowest charge) Q=1 pc case Study 1 and 10 pc case. Relax focusing in velocity bunching, lower ε growth, shorter beam For 1 pc, σ z only 4.7 µm after velocity bunching Use June 2008 version of SPARX lattice compression no longer at end, at 1.2 GeV (Final 2.1 GeV) Much higher final currents, ε nx m - rad some CSR emittance growth, for 1 pc Longitudinal tails, higher peak brightness

24 FEL performance at 1 pc Single spike with some structure > 1 GW peak power at saturation (30 m) 480 attosecond rms pulse at 2 nm z (m) ΔωΔt = 1.67 s (µm) λ (nm)

25 Higher Q SPARX case: 10 pc Longitudinal phase space, 10 pc Current profile; some pedestal visible Put back beam power, X ray photons Velocity bunch to 10.3 µm rms Still space dominated charge scaling ~Q 1/3 1 fs pulse Large emittance growth ε n = m - rad

26 SPARX FEL at 10 pc z (m) s (µm) 10 GW peak FEL power Saturation at 20 m (v. high brightness) Quasi single spike Lower brightness=longer cooperation length 4x10 11 photons, good for many applications

27 Practicality : tolerance studies Concentrate on 1 pc case, most challenging Look at various jitter errors Injection (laser) timing Injection offset Main linac phase, voltage Examine effects on Arrival time Mean energy Emittance Bunch length

28 Injector tolerances Model photoinjector with LANL PARMELA Emphasize laser errors Large injection timing error Pointing jitter Set tolerances compatible with linac, chicane Velocity bunching makes life easier

29 Laser transverse offsets at cathode Focusing in injector set for space charge dominated beam Too strong for centroid (behaves as 1 particle) Offset doesn t demagnify per acceleration (only ~0.5) Must control to fraction of final σ x (70 µm) Effective pointing jitter less than 20 micron on cathode Solution: relay image an aperture Solution: use over illuminated high QE cathode?

30 Injection Timing Studies For standard laser timing jitters of 0.2 psec (~degree at 2856 MHz) rms, all these parameters are acceptable

31 Time of arrival jitter Velocity bunching mitigates timing jitter (w.r.t. RF clock) greatly. Laser timing jitter of 0.2 psec produces 1.5 fsec (!) jitter in arrival time at injector exit Want pump-probe timing? Use e-beam radiation; time stamping

32 Linac and Chicane Jitter Tolerances Δσ z σ z0 = σ z σ zo σ zo <10% ΔE = E E 0 0o < 0.05% E 0 E 0o ΔT a = T a T a 0 <1fs Constraints Change in σ z <10% Experimentally negligible Change in mean beam energy < 0.05% Effect on FEL frequency smaller than BW Change in mean time of flight less than 1 fsec Experimentally negligible Initial condition tolerances set by PARMELA injector studies, used for Elegant input

33 Methodology: quick jitter estimation Apply overall φ shift to analyze performance due to laser gun timing error Apply initial E errors (±0.25MeV) also generated by laser injection error Apply individual RF phase (φ) and voltage (V) errors and analyze performance Establish error tolerance set apply an error distribution simulate many times for statistical performance

34 Injector Timing/Energy Errors Overall φ shift was applied to each linac module VB produces E error at linac entrance Tolerances easily satisfied with VB system \

35 Statistical Runs 2 parameters: RF phase (φ) and voltage (V) Assume uncorrelated jitter between linac modules Choose tightest tolerances, apply statistical error distribution to all elements Error set random, Gaussian distributed (± 3σ) Voltage error of interest ΔV/V=0.4% Phase error to investigate is Δφ=0.25 Repeated (400) ELEGANT simulation set

36 Summary of statistical analysis Δσ z σ z % ( ) All constraints satisfied easily ΔE E ΔT a (fsec)

37 Extension to LCLS case 1 pc does not give quasi single spike operation Beam too long, need to scale with λ/ρ Use 0.25 pc (1.5M e ), obtain ε n =0.033 mm mrad Yet higher brightness; saturation expected in 60 m σ E / E = Over 350A peak

38 LCLS Genesis results Deep saturation achieved Minimum rms pulse: 150 attosec Quasi-single spike, 2 GW σ ω σ t 1.1

39 Not all shots single spike z ζ

40 Alternative scenario: new undulator for very short λ operation Use high brightness beam to push to short wavelength LCLS example Shorter period undulator Shorter still soon available K u 1 λ u (cm) 1.5 β (m) 2-6 U (GeV) 4-14

41 Progress at SLAC on low Q, high brightness 20 pc case measured Diagnostic limit? Excellent emittance after injector 0.15 µm (predicted minimum 0.1 µm) No velocity bunching Chicane compression As low as 2 fs rms, with ε growth to 0.4 µm Measurement resolution inadequate See talks by Z. Huang, others from SLAC

42 Current UCLA R&D for small Q case Blowout regime at PEGASUS Blown out ellipsoid observed Thermal emittance limits Highly linear longitudinal phase space See Musumeci talk Ellipsoidal bunch distribution observed in blowout regime with RF deflector at PEGASUS Blowout regime single shot longitudinal phase space with RF deflector at PEGASUS, with simulation (right)

43 Low Q at SPARC Precursor to SPARX work Velocity bunching measurements w/rf deflector Factor 3 compression observed without ε growth See Ferrario talk

44 Ultra short beam application: coherent, sub cycle radiation Coherent transition radiation Non destructive: coherent edge radiation (CER) Total emitted CER spectrum (BNL ATF, UCLA compressor), measured compared to QUINDI. Coherent THz pulses Coherent optical, sub-cycle pulse (FLASH 1 pc case). Unique source at these wavelengths Would like to try this at LCLS Angular distribution of far field radiation, by polarization: measured in color, QUINDI in contours

45 Alternatives on the horizon: plasma accelerators LWFA gives GeV, fs electron beams: Table-top XFEL? undulator >10 TW LWFA yields GeV beams in cm Beam quality needs to be controlled Naturally gives fs pulses Can this be preserved? Hot topic Projects in EU, USA (Talk by M. Fuchs)

46 Table top XFEL work LMU MPQ (Garching) centered collaboration (BESSY, LBNL, UCLA, etc.) UCLA collaboration on beam transport and advanced undulator Need short λ u high field undulator for GeV important for linac based sources too Use undulator at SPARX? Lase at 6.5 Å (smaller by almost 5) Saturation at 200 MW (2E9 γ s), in <10 m (35 m in standard operation) Hybrid undulator: Pr-based, with SmCo sheath 9 Apply mm with period, LCLS with energy up to Å T

47 Ultra short beam application: IR wavelength PWFA Ultra high brightness, fs beams impact HEP also! Use 20 pc beam (LCLS) in high n plasma (>7E19/cc) Scaling to moderate blowout; more efficient >700 GV/m fields (!) Forming collaboration, looking at possibilities now OOPIC simulation of LCLS case

48 How to proceed? Collaborations (California to the World) UCOP proposal (LBNL UCB UCLA) NSF PIRE International Proposal (UCLA centered) Prof. William Barletta, Massachusetts Institute of Technology (also UCLA, visiting appt.), and Director, US Particle Accelerator School Dr. Vitaly Yakimenko, Director, Accelerator Test Facility, Brookhaven National Laboratory Prof. Joseph Bisognano, Director, University of Wisconsin Synchrotron Radiation Center and University of Wisconsin Dept. of Eng. Physics Prof. Luigi Palumbo, Responsible: SPARC High Brightness Beam Facility and SPARX X-ray FEL, University of Rome La Sapienza and INFN-LNF (Italy) Dr. Stephen Milton, Responsible: FERMI X-ray FEL Sincrotrone Trieste (Italy), also Argonne National Laboratory Dr. Sven Reiche, Head, FEL Physics for PSI-XFEL, Paul Scherrer Institute (Switzerland), also UCLA, Dept. of Physics and Astronomy Prof. Avraham Gover, Ludwig Jokel Chair of Electronics in Tel-Aviv University, Dept. of Physical Electronics, head the Israeli Knowledge Center for Radiation Sources and Applications (Israel) Dr. Luca Giannessi, Staff Scientist, Head, FEL Physics SPARC and SPARX FELs, ENEA-Frascati (Italy) Prof. Joerg Rossbach, leader of accelerator physics group, University of Hamburg Technical head FLASH VUV FEL, DESY (Germany), Dr. Frank Stephan, Head, PITZ Photoinjector Laboratory, DESY-Zeuthen (Germany) Dr. Thorsten Kamps, Staff Scientist, BESSY-Berlin (Germany), Dr. Wim Leemans, Director, l OASIS Laser Accelerator Lab, Lawrence Berkeley National Laboratory Prof. Florian Gruner, project leader MAP Cluster of Excellence for Table-Top Free-Electron-Lasers, project leader in ELI, Ludwig Maxillian Univ./MPQ (Germany) Dr. John Corlett, Deputy Division Director AFRD, Program Head, Center for Beam Physics, Lawrence Berkeley National Laboratory Prof. Tang Chuanxiang, Director, Department of Engineering Physics; Head, Accelerator Laboratory, Tsinghua Univ. (China)

49 Conclusions; future work Very promising options with very low charge Excellent emittance and gain Smaller charge, more photons/electron Can allow shorter undulators, FEL λ What do users want? Do we need single spike or just ultra fast? Important to exploit brightness? Femtoseconds? Attoseconds Excellent beam scenario, but All measurements change Low charge diagnostics needed; experience at UCLA Pegasus Coherent optical signals everywhere. Signal needs to be interpreted Background noise Dark current has much more charge; use dual deflector/collimator? Much more experimental work needed E.g., photoinjector blowout regime Use of FEL in single spike (SPARC, FLASH, LCLS) OPTIMIZE DESIGN from the ground up!