Generation of Femtosecond Electron Pulses

Transcription

1 Generation of Femtosecond Electron Pulses W. D. Kimura STI Optronics, Inc., 755 Northup Way, Bellevue, WA , USA Two techniques for generation of femtosecond electron pulses will be presented. The first is from the Staged Electron Laser Acceleration (STELLA) experiment located at the BNL ATF. The second is a laserslicing technique demonstrated by LBL on the ALS ring. Differences between these techniques and the problems of preserving the femtosecond electron pulses will be discussed. Introduction A new era in electron beam (e-beam) technology has arisen with the advent of different means for generation of femtosecond e-beam pulses. These pulses have many potential applications including generation of femtosecond x-ray pulses for studying ultrafast phenomena, and as a means to pump and/or probe media with the electrons directly. While there have been different efforts on generating femtosecond e-beam pulses, 1 two techniques that utilize external laser beam interaction with the e-beam are reviewed here. The first is the Staged Electron Laser Acceleration (STELLA) experiment being conducted at the Brookhaven National Laboratory (BNL) Accelerator Test Facility (ATF). The second is a laser-slicing technique 3 developed by Lawrence Berkeley National Laboratory (LBL). These two techniques are similar to each other in that both rely on the free electron laser (FEL) effect to exchange energy between the laser beam and electrons. However, the electron pulses created by the techniques are fundamentally different because in STELLA the laser pulse length is much longer than the e-beam pulse; while in the LBL technique, the e-beam pulse is much longer than the laser pulse. This difference is summarized in Fig. 1. As will be shown, the former leads to the creation of a train of femtosecond microbunches; whereas, the latter generates a pair of femtosecond electron pulses and was not designed to create a train of microbunches. Fig. 1. The two techniques discussed in this paper differ in that for STELLA (top figure) the e-beam pulse has much shorter time duration than the laser pulse. For the LBL laserslicing (bottom figure), the laser pulse is much shorter than the e-beam pulse.

2 Staged Electron Laser Acceleration (STELLA) The goal of the STELLA experiment was to demonstrate staging between two laser accelerators. This work was motivated by the great progress being made in laser acceleration where gradients of >1 GeV/m have been demonstrated. 4 However, thus far these experiments have involved a single pass of the laser beam with the e-beam and have been limited to interaction lengths of a few millimeters. In order to eventually realize a practical accelerator it will be necessary to stage the acceleration process; whereby, the laser beam repeatedly interacts with the e-beam resulting in high net energy gain. STELLA used a pair of inverse free electron lasers (IFEL) 5 for the laser accelerators. The first IFEL (IFEL1) acted as an energy modulator of the e-beam (i.e., prebuncher), which resulted in the creation of femtosecond microbunches. The second IFEL (IFEL) then accelerated these microbunches. A key issue is achieving proper phase synchronization between the microbunches formed by the prebuncher and the laser field in the accelerator (IFEL). In an IFEL, the e-beam copropagates with a laser beam inside a magnetic array called a wiggler or undulator. The undulator causes the electron trajectory to oscillate in the plane of the laser beam electric field as illustrated in Fig., thereby projecting a component of this field in the direction of the electron motion. Depending on the sign of the electric field (i.e., its phase relative to the electrons), the field can accelerate or decelerate the electrons. ELECTRON BEAM S N N S S N N S S N N S UNDULATOR MAGNET ARRAY LASER BEAM Fig.. Illustration for an IFEL based upon a planar undulator. In order to achieve net energy exchange along the length of the undulator, the following resonance condition must be satisfied 6 γ λ w K = 1+, (1) λ L where λ L = laser wavelength, λ w = wiggler wavelength, γ = Lorentz factor, K = eb o λ w /πmc, B o = peak magnetic field, m = mass of electron, and c = speed light. Higher energy exchange is possible if the undulator is also tapered, 7 in which either the magnetic gap or magnet period varies along the undulator length.

3 A schematic of the STELLA experiment is depicted in Fig. 3. The ATF CO laser beam is split into two beams with approximately 4 MW sent to the prebuncher (IFEL1) and up to 3 MW sent to the accelerator (IFEL). Axicon lenses convert the Gaussianprofile laser beam into an annular one. Focusing telescopes focus the beams at the center of each undulator. An adjustable optical delay stage permits changing the phase of the laser beam entering IFEL relative to the beam entering IFEL1. Each laser beam enters the beam line vacuum pipe through windows and is directed to the undulators using invacuum mirrors with central holes for transmission of the e-beam. The separation distance between the exit of IFEL1 and the entrance to IFEL is m. At the end of the beam line is a spectrometer 8 featuring a wide energy acceptance (±%) capable of measuring the entire electron spectrum in a single shot. CO LASER BEAM ADJUSTABLE OPTICAL DELAY STAGE FOCUSING LENSES DIPOLE MAGNET Accelerator (IFEL) Prebuncher (IFEL1) E-BEAM FOCUSING LENSES VACUUM PIPE SPECTROMETER VIDEO CAMERA UNDULATOR MAGNET ARRAY E-BEAM FOCUSING LENSES MIRROR WITH CENTRAL HOLE UNDULATOR MAGNET ARRAY MIRROR WITH CENTRAL HOLE = QUADRUPOLE MAGNET E-BEAM Fig. 3. Schematic of the STELLA experiment. In STELLA, the e-beam pulse length is 3 ps; whereas, the laser pulse length is 18 ps. This means the electrons enter the STELLA prebuncher distributed uniformly over all phases of the laser field inside the undulator. This is depicted in Fig 4(a), which is a model simulation of the STELLA prebuncher. Hence, the laser imparts a sinusoidal energy modulation on the e-beam [see Fig. 4(b)] with an amplitude of ±.5% for 4 MW into IFEL1. This amount of modulation is chosen so that after drifting m to IFEL, the fast electrons catch up with the slow ones resulting in longitudinal density bunching of the electrons into microbunches [see Fig. 4(c)]. Because this modulation is induced by the laser field, these microbunches have bunch lengths a fraction of the laser wavelength. And, since the laser wavelength (1.6 µm) is much shorter than the e-beam pulse length, a train of 3-fs microbunches is formed with each microbunch spaced apart by the laser wavelength (~3 fs). It is thus by this means that STELLA generates femtosecond microbunches. These microbunches are then sent into IFEL where they are accelerated. Figure 5 shows false-color raw video images from the spectrometer camera. Figure 5(a) shows the energy spectrum for the e-beam only. After modulation by the prebuncher, the spectrum changes into a symmetric double-peaked one shown in Fig. 5(b). With the prebuncher and accelerator both operating and the phase delay between

4 (a) (b) (c) Fig. 4. Model simulations for the prebuncher showing the electron energy-phase distribution and longitudinal density distribution. (a) At entrance to prebuncher. (b) At exit to prebuncher. (c) After drifting m to the accelerator. Fig. 5. Spectrometer output showing false-color images of the e-beam energy spectrum with white representing saturation. (a) With e-beam only. (b) With prebuncher only. (c) With prebuncher and accelerator, and phase delay set for near-maximum acceleration. (d) With prebuncher and acceleration, and phase delay set 18 from (c).

5 the laser beams entering the prebuncher and accelerator adjusted for maximum acceleration, a clear peak can be seen in the spectrum [see Fig. 5(c)] representing the accelerated microbunches. By changing the phase delay by 18, these microbunches can be decelerated [see Fig. 5(d)]. The microbunches can be characterized by examining how the e-beam energy spectrum is altered when the microbunches interact with the laser beam inside the accelerator (IFEL). The STELLA model predicts distinctive changes in the energy spectrum depending on the microbunch characteristics and the phase at which they enter the accelerator. The model is based upon a classical FEL simulation 9 and includes 3D effects, such as emittance and beam misalignments, and 1D space-charge effects. It ray-traces the electron trajectories through the drift region between IFEL1 and IFEL. The parameters for the STELLA experiment that were used in the model are listed in Table I. E-beam energy E-beam energy spread E-beam macropulse length Wiggler wavelength Table I. Parameters for STELLA experiment. Parameter Value 45.6 MeV*.4% (1σ) 3 ps 3.3 cm Number of wiggler periods 1 Wiggler K parameter.9 Laser wavelength 1.6 µm Laser pulse length Laser peak power to prebuncher Laser peak power to accelerator Measured energy modulation by prebuncher (IFEL1) *E-beam energy selected to satisfy resonance condition given by Eq ps 4 MW -3 MW ±.5% Figure 6 compares the model predictions with the experimental data for the phase delay set at near-maximum acceleration and MW delivered to the accelerator. The raw video signal of the energy spectrum is shown in Fig. 6(d); its line profile is plotted in Fig. 6(c). We see there is good agreement between the model spectrum and the line profile. Figure 6(a) shows the electron phase distribution that gave rise to the model energy spectrum in Fig. 6(c). A concentration of electrons representing the microbunch can be clearly seen. These electrons projected onto the phase axis [see Fig. 6(b)] indicate that the microbunch length is ~.8 µm long (FWHM) corresponding to ~.7 fs in duration.

6 Changing the phase delay between the laser beams driving the IFELs causes the microbunch to move within the energy spectrum as demonstrated in Fig. 7. We see once again that the model and data agree well at all phase positions. Acceleration of well-formed microbunches can be disrupted if the conditions are not correct. For example, sending too little or too much laser power to the prebuncher can cause the microbunches to have their maximum density distribution (i.e., smallest bunch length) either downstream or upstream of the accelerator, respectively. Figure 8 shows the data and model predictions for the case when too much laser power is delivered, i.e., there is overmodulation occurring in the prebuncher. (a) Energy Shift (%) (b) Bunch Length (µm) (c) (d) Electron Distribution 1 Model Data Spectrometer output Fig. 6. Comparison of data with model for the case of near-maximum acceleration of the microbunch. The laser powers driving the two IFELs in Fig. 8 are comparable to each other. This results in a breaking apart of the concentration of electrons seen previously in Fig. 6(a) into separate bands of electrons [see Fig. 8(a)]. The effect of these bands is the appearance of four peaks within the energy spectrum as confirmed by the data [see Fig. 8(c)]. The net result is loss of the microbunch [see Fig. 8(b)].

7 Electron Distribution Energy Shift (%) Data 1 Model Phase Delay Electron Distribution Model Energy Shift (%) Data Phase Delay Electron Distribution Model Energy Shift (%) Data Phase Delay 31 Phase Delay Fig. 7. Model and data energy spectra as a function phase delay between the laser beams entering the two IFELs. Zero phase has been arbitrarily chosen to correspond to maximum acceleration. Electron Distribution Data 3 1 Model Energy Shift (%) The staging process demonstrated during STELLA was remarkably stable. Even though there was no active phase stabilization used and the various mirrors directing the laser beams were separated by many meters, phase synchronization could be maintained over periods of many minutes. In order to help improve this phase stability, the current program, called STELLA-II, will modify the experiment so that a single laser beam drives both the prebuncher and accelerator. This entirely eliminates any phase jitter related to using separate laser beams to drive the IFELs. This new system is depicted in Fig. 9. STELLA-II will be using the upgraded ATF laser capable of delivering several hundred gigawatts of peak power. Because this laser power is much larger than needed to drive the prebuncher, the undulator for the prebuncher will be replaced with a 3-period electromagnet that is intentionally detuned in order to still provide a modulation of ±.5% even with 1 s of GW of laser power passing through it. Using a single laser also requires a minimum drift space between the IFELs. Thus, a short-length magnetic chicane will be utilized between the new prebuncher and accelerator. The laser beam will also be passing through the chicane. Thus, to minimize the possibility of additional laser-induced modulation occurring, the chicane will be oriented with its magnetic field orthogonal to the prebuncher.

8 (a) (c) (d) Energy Shift (%) Data Model (b) Electron Bunch Length (µm) Spectrometer output Fig. 8. Comparison of model and data for the case of 9 MW sent to the prebuncher and 115 MW to the accelerator. CO LASER BEAM FOCUSING TELESCOPE DIPOLE MAGNET PREBUNCHER (IFEL1) ACCELERATOR (IFEL) E-BEAM FOCUSING LENSES VACUUM PIPE SPECTROMETER VIDEO CAMERA E-BEAM FOCUSING LENSES TAPERED UNDULATOR ARRAY CHICANE MIRROR WITH CENTRAL HOLE = QUADRUPOLE MAGNET E-BEAM Fig. 9. Schematic of planned STELLA-II experiment.

9 A final important modification during STELLA-II will be to use a tapered undulator for the accelerator. This will permit better trapping of the microbunches and greater energy gain using the higher laser power that will be available. A model simulation for STELLA-II is shown in Fig (a) Energy Shift (%) Electron Bunch Length (µm) (b) (c) Fig. 1. Model predictions for STELLA-II for 1 GW laser power and 5% taper in the accelerator undulator. With 1 GW of laser power and 5% taper in the accelerator undulator, the model predicts an energy gain of 13 MeV for a 45.6 MeV e-beam [see Fig. 1(c)]. As can be seen in Fig. 1(a), the microbunch has been cleanly separated from the unaccelerated electrons and rotates in phase-space approximately 3/4 of a synchrotron period. In addition, the accelerated microbunch has an energy spread of 1.% FWHM. Hence, STELLA-II will demonstrate monoenergetic acceleration of the microbunches. This process is important for eventually building practical laser accelerators and for enabling femtosecond electron microbunches created in this manner to be used for other applications. LBL Laser-Slicing The laser-slicing technique demonstrated by LBL had as its motivation the desire to generate femtosecond x-rays in synchrotron storage rings. As explained in Ref. 3, femtosecond x-rays have many applications such as studying ultrafast structural dynamics associated with phase transitions in materials and various processes.

10 The basic scheme for the LBL laser-slicing technique is depicted in Fig. 11. Similar to STELLA, LBL uses a wiggler and the FEL interaction to exchange energy between the laser beam and e-beam. However, as mentioned earlier, in the LBL case the laser pulse length is.1 ps and the e-beam pulse length is 3 ps. This means, as shown in Fig. 11(a), that the laser beam only modulates a short section of the e-beam pulse. Fig. 11. Schematic for LBL laser-slicing technique used to generate femtosecond electron pulses. The e-beam is then sent through an energy-dispersion bend magnet in order to spatially separate the modulated electrons from the main electron pulse. These different groups of electron travel through different arc trajectories as depicted in Fig. 11(b). Consequently, the synchrotron x-ray emission from these groups of electrons is emitted at slightly different angles. An x-ray mirror collects this radiation and a slit can be used to allow only the femtosecond x-ray radiation from, say, the accelerated electrons to pass through. Various diagnostics are used to monitor and optimize the FEL interaction as depicted in Fig. 1. A camera and spectrometer are used to detect spatial and spectral matching between the laser beam and the spontaneous emission from the wiggler. This approach is different from STELLA, where the spontaneous emission from the undulators was not monitored. In addition, LBL also optimizes the differential gain by using a pair of optical filters centered at the maximum and minimum gain points in the FEL spectrum. Again, STELLA does not do this because the e-beam is tuned to be resonant with the fixed-gap undulators. In the LBL situation, the wiggler gap is tuned for the e-beam energy of the ALS beam (1.5 GeV) and wavelength of the laser driver (8 nm).

11 Fig 1. Schematic of FEL diagnostics for LBL laser-slicing experiment. The experimental parameters for the LBL femtosecond pulse demonstration are given in Table II. Table II. Parameters for LBL femtosecond pulse demonstration experiment. E-beam energy E-beam energy spread E-beam macropulse length Wiggler wavelength Parameter Value 1.5 GeV (ALS beam).8% (rms) 3 ps 16 cm Number of wiggler periods 19 Wiggler K parameter 13 Laser wavelength.8 µm Laser pulse length Laser peak power to prebuncher.1 ps 4 MW Measured energy modulation ±.4% Although the laser pulse length was 1 fs, the electron pulse sliced from the main pulse spread out to 3 fs due to the relatively long distance between the wiggler and the bend magnet. This is because the energy spread imparted on the modulated electrons by the femtosecond laser pulse causes the electrons to mix and spread in longitudinal space. Unlike STELLA, LBL is able to directly measure the pulse length of the femtosecond electron pulse by performing correlation measurements between the laser pulse and the synchrotron radiation (SR) passing through the slit produced by the electrons. This is shown in Fig. 13. Figure 13(a) shows the measured gain in the

12 amplified laser pulse energy as a function of time delay between the laser pulse and the electron main bunch. This indicates a width for the electron main bunch of σ = 16.6 ps. Separating the modulated electrons from the main electron bunch in the dispersion magnet means there will be a femtosecond hole left in the main bunch. When this bunch emits SR, a femtosecond dip in the x-ray emission occurs as shown in Fig. 13(b). This data was taken by setting the slit (see Fig. 11) to pass the central ±3σ of the SR emitted from the bend magnet. If the slit is adjusted to pass the +3σ to +8σ of the SR, then primarily the radiation from the accelerated femtosecond electron pulse is detected [see Fig. 13(c)]. This correlation data indicates the electron pulse has a length of σ = 161 fs. (a) (b) (c) Fig. 13. Correlation measurements showing various pulse lengths during LBL femtosecond electron pulse demonstration. (a) Measured gain on laser pulse obtained by sweeping laser pulse through e-beam. (b) Loss of x-ray emission caused by hole left in the main electron bunch due to separation of femtosecond electron pulse after traveling through dispersion magnet. (c) X-ray emission from modulated electron pulse. Another possible method for separating the femtosecond electrons has been suggested by Zolotorev and Zholents. 1 A septum magnet can be positioned at the point of maximum dispersion with the septum plates used to separate the high and low energy modulated electrons. This scheme is depicted in Fig. 14. This permits the high and low energy modulated electrons to be used for other applications beside generation of x-rays. For example, one set of electrons could be sent to a target and used as a pump for some process. The other set of electrons could be used as the probe to analyze the effects of the pump electrons on the target. However, this method does require isochronous relaying of the e-beam from the output of the wiggler to the target, which can increase the complexity of the imaging optics. Because the 1-fs laser pulse only slices out ~1/3th of the electrons in the main electron bunch, and because these modulated electrons are accelerated and decelerated into two separate groups of electrons, the amount of charge within the femtosecond electron pulse is limited. This is different from STELLA where up to 5%

13 bunching is possible with the accelerated and decelerated electrons combining to make a train of microbunches. Fig. 14. Alternative scheme for separating femtosecond electron pulses. Nonetheless, the LBL laser-slicing technique has several attractive features. It uses commercially available femtosecond lasers and conventional wigglers. These lasers are capable of high repetition rates (e.g., 1 khz). The method is compatible with storage ring operations. The laser slicing process does not disrupt the main circulating electron pulse. In fact, the hole caused by the laser slicing within the main bunch tends to fill in again as the electron bunch recirculates around the ring, thereby allowing the laser slicing to be repeated again on the main electron bunch. The x-rays generated are automatically synchronized with the drive laser, which makes it easier to perform experiments utilizing both beams. Preserving Femtosecond Electron Pulses The e-beam transport system must be carefully designed to minimize pulse spreading. As already mentioned, a factor of three spreading occurred in the LBL experiment simply because the bend magnet was too far away from the wiggler. (Future LBL laser-slicing experiments are planned with the bend magnet much closer to the wiggler.) In STELLA, modeling analysis has shown that micron-level errors in longitudinal electron trajectories within the microbunches can lead to significant spreading for 3-fs microbunches. Different situations can lead to pulse spreading. The energy spread of the electron pulse coupled with any dispersive beam line optics, such as dipoles, can cause the electrons to disperse spatially. Path length differences between electrons within the femtosecond pulse can arise due to beam divergence and/or angular misalignment of the beam through focusing optics (e.g., quadrupoles). This latter effect was an important issue during STELLA because of the triplet located between the prebuncher and accelerator.

14 Plotted in Fig. 15 are the model predictions for the microbunch length as a function of the e-beam centroid trajectory angle entering the triplet downstream of the prebuncher. The affect is different in the vertical and horizontal planes because the polarities of the quads within the triplet cause it to act like a position-dependent chicane in the vertical plane. In the vertical plane, we see an angle error of.5 mrad can double the length of a 3-fs microbunch. There are different ways to help minimize these bunch smearing effects. As mentioned, for the LBL laser-slicing technique, positioning the bend magnet closer to the wiggler will help. For STELLA-II, the triplet will be eliminated between the prebuncher and accelerator. Trapping the femtosecond microbunches within an acceleration potential well can also help by using the laser fields to contain the electron motion in phase space. Some laser acceleration processes also have significant radial fields, which in principle could be used to help restrict radial movement of the electrons within the microbunch Microbunch Length (µm) Vertical Plane Horizontal Plane Trajectory Angle at Wiggler Exit (mrad) Fig. 15. Model prediction for bunch length smearing due to angular misalignment of the e-beam through the triplet between the prebuncher and accelerator in STELLA. Conclusions This paper discusses two approaches that have demonstrated generation of femtosecond electron pulses. Other related methods, in particular those associated with other laser acceleration mechanisms, would also be expected to produce femtosecond pulses. In all cases the issue of preserving the femtosecond electron pulses must be considered. This makes transporting ultrashort pulses over long distances a challenge and may require new thinking in the design of transport systems able to control electron trajectory errors to micron levels. Finally, electron pulses lengths less than 1 fs may be produced in the near future. While exciting in its own right, this brings up the interesting question of how to measure the lengths of these super-ultrashort pulses.

15 Acknowledgements The author wishes to acknowledge the other members of the STELLA collaboration: M. Babzien, I. Ben-Zvi, L. P. Campbell, D. Cline, C. E. Dilley, J. Gallardo, S. C. Gottschalk, P. He, K. P. Kusche, R. H. Pantell, I. V. Pogorelsky, D. C. Quimby, J. Skaritka, L. C. Steinhauer, A. van Steenbergen, V. Yakimenko, and F. Zhou. He also wishes to thank R. Schoenlein and M. Zolotorev for providing figures and reviewing material on the LBL laser-slicing experiment. ÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃ ÃSee for example K. N. Ricci and T. I. Smith, Phys. Rev. ST Accel. Beams 3, p. 381 ().Ã ÃW. D. Kimura, et al., Phys. Rev. Lett. 86, p. 441 (1).Ã ÃR. W. Schoenlein, et al., Science 87, p. 37 ().Ã ÃW. P. Leemans and E. Esarey, in Proceedings of Advanced Accelerator Concepts 8th Workshop, Baltimore, MD, Jul. 5-11, 1998, edited by W. Lawson, C. Bellamy and D. Brosius (American Institute of Physics, New York, 1999), p. 174.Ã ÃR. B. Palmer, J. Appl. Phys. 43, p. 314 (197).Ã ÃE. D. Courant, C. Pellegrini and W. Zakowicz, Phys. Rev. A 3, p. 813 (1985).Ã ÃN. M. Kroll, P. L. Morton and M. N. Rosenbluth, IEEE J. Quant. Electron. QE-17, p (1981).Ã ÃV. Yakimenko, [Online] Available: See Sec. 8 on World Wide Web page.ã ÃD. C. Quimby, J. M. Slater and J. P. Wilcoxon, IEEE J. Quan. Elect. QE-1, p. 979 (1985).Ã ÃM. S. Zolotorev and A. A. Zholents, Phys. Rev. Lett. 76, p. 91 (1996).Ã ÃR. D. Romea and W. D. Kimura, Phys. Rev. D 4, p. 187 (199).Ã