Femtosecond Pump Probe Operation and Plans at the

Femtosecond Pump / Probe Operation and Plans at the LCLS Josef Frisch for the LCLS Commissioning team 1

Ultra-Fast Science Some experiments use multiple images on an already evolving system All feet off the ground Most experiments are pump probe: Stimulate the system (fire bullet) Wait Measure with a probe pulse (flash bulb) H. Edgerton Measurement resolution is set by the length of the pump / probe pulses AND the accuracy of the time delay between the pump and probe. 2

Short Bunch Operation at LCLS Low charge (20 p. C) operating mode for very short pulses No direct pulse length measurement available, but believed to be < 5 fs FWHM Phase = +1 deg Phase =+0. 5 deg Phase = 0 deg ΔT=5. 0 fs Phase = -0. 5 deg ΔT=2. 3 fs Phase = -1 deg ΔT=1. 1 fs Genesis Simulation for over compression: 5 fs FWHM ΔT=1. 9 fs ΔT=4. 2 fs we typically operate here 3

Narrow or Double X-Ray Pulses from a Slotted Foil PRL 92, 074801 (2004). P. Emma, M. Cornacchia, K. Bane, Z. Huang, H. Schlarb (DESY), G. Stupakov, D. Walz 0. 25 mm 0 -6 mm pulses not coherent Power (GW) 10 0 -150 fs 5 2 fs 0 time (fs) 4

Low Charge AND Slotted Foil X-ray spectrum with 20 pc operation – few spikes suggest ~5 fs pulses With 20 pc and slotted foil see single spike spectrum suggests very short pulses No direct measurement but LCLS may be producing ~1 fs X-ray pulses 5

Short Pulse Lasers • Commercial Ti: Sapphire lasers can produce pulses as short as 15 fs. (25 -50 fs more typical). • High harmonic generation can produce ~100 a. S, pulses in the XUV ~100 e. V. • Assume lasers will produce shorter pulses in the future Attosecond XUV generation, Max-Plank_institut fur 6 Quantemoptik / ATLAS

Experiment Requirements • This talk will concentrate on laser pump / X-ray probe experiments – Most common experiment at LCLS • Right now operating with ~10 -100 fs X-ray pulses and ~50 fs laser pulse • In the future we expect few-fs X-rays and few-fs laser pulses • Timing control at the few fs level will be required. – Typical temperature coeficient for either coaxial cables or fiber optics is 2 x 10 -5/C° ->1 meter is 60 femtoseconds / C° – Thickness of a sheet of paper = 100 fs • When describing timing drift or jitter, need to be careful to clarify what reference is used for comparison. 7

Experiment Requirements Pump Laser X-ray beam X-rays to detector System evolves from pump to probe time Timing jitter relative to an external clock isn’t important Ideally would scan time difference Generally OK to let jitter vary the timing and measure shot-to-shot 8

Sources of Timing Jitter Laser Gun RF off crest RF RF Bunch Compressor Laser pulse is compressed typically 2 X in gun, then an additional factor of 100 in the bunch compressors Changes in laser time are compressed, so gun laser jitter is not very important. Beam time is mostly set by the RF in the compression system. (both amplitude and phase contribute) Synchronizing the gun laser to the experiment laser doesn’t fix the jitter 9

Conventional Timing System Stabilized transmitter Stabilized Receiver Femtosecond Laser Amplifier Undulator Beam time pickup X-rays E-beam dump Experiment ~100 M Beam pickup typically responds to electric field of bunch: either RF cavities or electro-optical pickups are used Stabilization system typically feeds back on the length of the cable / fiber. 10

Timing Jitter in LCLS ~ 1 km Master Source 10 ps drift over hours Few fs jitter in a few ms High power RF Phase Shift ~50 fs RMS jitter shot to shot Feedback Stabilized link ~20 fs stability 10 fs jitter, 50 fs stability Phase Detect ~50 fs jitter Laser RF in compressor sets beam time Accelerator Phase cavity FEL Experiment data corrected offline with phase cavity data report 50 -100 fs stability Experiment 11

Beam arrival time cavity (LCLS) Similar to a cavity BPM but use the monopole mode Phase drift from cavity temperature is the most significant problem 1 us time constant, 10 -5 /C° temperature coefficient -> 10 ps/C° (!) Raw Signal Phase slope gives cavity temperature 12 RMS difference between cavities ~12 femtoseconds RMS at 250 p. C, 25 femtoseconds at 20 p. C. Drift is ~100 femtoseconds p-p over 1 day.

RF Phase Detection Limits • Oscillators: unlocked timing noise relative to an “ideal” clock increases with time – Conventional oscillators: 1 fs RMS above 1 KHz – Sapphire oscillators 1 fs RMS above 10 Hz • RF phase measurement (2 X thermal noise) – 1 GHz, 1 ms, 1 m. W power -> 20 a. S (theoretical) – SLAC summer students actually measured a noise level corresponding to 30 a. S in a 1 KHz bandwidth – In a 1 MHz bandwidth, still expect 1 fs. – Phase cavity system noise is about 7 fs RMS. (best conditions) • Electronics noise is not a stringent limit! • Drift: few fs / °C for mixers. • Drift: ~30 fs / °C for 1 M cable. 13

EO Beam Time Measurement (Several versions, simplified concept shown) Short pulse laser Free space or fiber-optic Detector Bunch fields Output intensity depends on relative timing of laser plulse and Ebeam Electric field from bunch Electro-optical intensity modulation F. Loehl et al DESY/FLASH 6 femtosecond timing noise published (Believe ~3 fs achieved) Allows direct conversion from beam timing to optical signal: significant advantage for some types of timing systems 14

Long Distance Timing Transmission Transmitter Adjust Delay mirror Feedback Use fibers: Low loss as high transmitter frequency Good directional couplers Low cost Timing Signal Compare forward and reflected signals Envelope scheme (DESY, MIT Bates): Transmit short (ps) pulses at ~100 MHz rate. Timing of the reflected pulses is used to measure the fiber length. Control fiber length with feedback Pulses detected at the receiver end are used for timing Carrier scheme (LBNL, used at LCLS) Frequency stabilized laser used in an interferometer Interferometer determines fiber length Control fiber length with feedback (feed forward in this case). Both systems work at <20 fs over 100 M fibers Pulses allow direct locking to experiment laser Excellent resolution – based on optical wavelength Difference between phase and group velocity is important an must be compensated 15

Optical to Electronic Conversion • Even with perfect fiber stabilization systems, this can be the performance limit. • Photo-diodes: Tradeoff between noise and linearity – Nonlinearity: Charge extraction -> changes bias voltage -> changes capacitance -> changes phase delay – High frequency diodes have small area, low capacitance. – For S-band (3 GHz) diode -> 150 fs single shot resolution – For X-band (12 GHz) diode -> 60 fs single shot • For high repetition rate systems (oscillators) this isn’t too bad: 68 MHz, 100 us TC -> 1 fs (ideal) • For amplifiers, this is a large problem – single shot measurements are very difficult. • Can in principal use an optical resonant cavity (etalon) to average signals. For Q = 100 -> ~10 fs • Other techniques have been developed for fiber based systems: Rely on electro-optical mixing between laser and RF signal. 16

• Conventional Ti: Sapphire laser oscillators can be locked to ~50 fs to a RF reference. – Several limitations: Laser Stabilization • Phase detection from photodiodes • Acoustic noise changing the cavity length • Pump laser fluctuations change the effective cavity length through nonlinearities – Laser chirp pulse amplifier system can add jitter • Wavelength changes can change the delay through the compressors (if the wavelength response of the amplifier isn’t flat) • Pulse shape changes with laser power from changes in amplifier saturation – Very active area of research both at labs and in industry. – At least at LCLS this is the limit to stability. (A. Winter et al). DESY optical master oscillator • The pulsed DESY / FLASH system allows direct optical cross correlation between the experiment laser and the timing system! 17

Superconducting vs RT Accelerators • The beam timing jitter relative to the accelerator timing reference system is similar for room temperature and superconducting accelerators: 30 -50 fs RMS. Feedback Gun RF Structure Compressor RF Structure Beam time pickup In an superconducing accelerator the beam timing can be measured for each pulse at the ~MHz beam rate, much faster than the typical 100 us energy storage time in the accelerator cavities This allows the use of a fast timing feedback to reduce the timing jitter. 18

Other Limits • Ground Motion – Tidal stretching is 30 um / kilometer. (100 fs/km) – In principal predictable, but in practice tricky – Fast ground motion varies with location. • Measured at SLAC as 10 s of nanometers over 14 M separation. • Needs more study • SASE process – Statistical fluctuations give a minimum timing jitter of [(1/12)r. L]1/2 with r the slippage distance and L the bunch length. – If only part of the bunch lasers, X-ray time will not match electron beam time. • Location of experimental IP (1 um -> 3 fs) • Looks difficult to reach 1 fs even if the individual technical system problems can be resolved. Tides observed in LEP frequency corresponding to ~2 x 10 -8 (L. Araudon et al, CERN SL/94 -07) 19

Optical / X-ray Cross Correlator X-rays Reflected optical beam measured on array sensor Laser Ga. As or similar Tests at SXR (W. Schlotter et al) have demonstrated <60 fs RMS (consistent with 0) single shot X-ray to laser optical timing measurement. Note that electronic timing will still be needed for “crude” 100 fs timing 20

Cross Correlator (very preliminary) (SXR) M. Beye B. Schlotter 21 W. Schlotter et al. (LCLS)

Cross Correlation • Various physics is available, but need to find a way to operate over the full wavelength range and with femtosecond resolution – 250 -12 Ke. V – Operate at few u. J pulse energies (1 fs operation) • Final version should do cross correlation in the experimental chamber – 1 fs is 300 nm, very difficult to control long lengths at this level. • Need to find appropriate physics to use for this – May need an XFEL to study this physics! 22

THz Timing Experiments • THz pump / X-ray probe – The high peak current beams used for XFELs can also serve as sources of very intense THz radiation – This radiation is precisely timed to the electron beam. – Unfortunately since the beams are ultra-relativistic the THz can never “catch” the X-rays THz FEL X-rays THz delayed relative to X-rays. Need to use 2 bunches, one generates THz, second X-rays. For hard X-rays can use crystals to delay to match the THz Timing error limited by mechanical stability 23 Plans to test both schemes at SLAC / LCLS.

Future Timing Systems • “Conventional” systems presently have 50 -100 fs rms timing resolution • Can probably extend to ~10 -30 fs RMS • Conventional lasers now produce <25 fs pulses, with ~100 as available from XUV lasers. • XFELS at <10 fs, with <1 fs likely in the near future. • For single femtosecond timing will need new approaches like direct X-ray / optical cross correlation. 24