SUB-PICOSECOND OPTICAL PULSES AT THE SLS STORAGE RING

G. Ingold, A. Streun, B. Singh, R. Abela, P. Beaud, G. Knopp, L. Rivkin, V. Schlott, Th. Schmidt,H. Sigg, J.F. van der Veen, A. Wrulich, Paul Scherrer Institut, Villigen, Switzerland

S. Khan, BESSY, Berlin, Germany

Abstract

We report on the feasibility and expected performance ofa sub-picosecond X-ray source at the Swiss Light Source(SLS), based on the electron-beam slicing method. Thismethod has recently been demonstrated at the AdvancedLight Source (ALS), using a bending magnet to produceultrafast optical pulses [1]. To improve the flux of such asource for user experiments, we have studied its realizationboth in the hard and soft X-ray regime, employing smallgap insertion devices (IDs) installed either in one or in twosuccessive straight sections. The most favourable geome-try is one in which the modulator and the radiator are posi-tioned in the same straight section of the storage ring.

1 INTRODUCTION

The ID based photon sources [2] at the SLS [3] cover theenergy range 10 eV to 40 keV. The energy range 10 eV to18 keV is covered by undulators, and for photon energiesup to 40 keV, a hybrid wiggler is employed. To achievehigh flux and brightness at a storage ring of 2.4 GeV, smallgap undulators and wigglers have to be installed. In addi-tion, for a versatile photon source the options of a flexiblepolarisation and a short-pulsed time structure of the photonbeam is of central importance to a variety of experimentsin magnetism and structural research. The notable featuresof the ID-based photon sources at the SLS are:� For soft X-rays: the use of elliptical twin-undulators

with variable linear and circular polarisation, allowing op-positely polarized, rapidly switched (0.01-1 kHz) beamsunder otherwise identical conditions.� For hard X-rays: the use of small gap, short period

undulators to extend the high brightness synchrotron radi-ation (SR) to '18 keV (i.e. into the domain of high energySR facilities), operating on higher (11th/13th) harmonics.

The method to generate sub-picosecond synchrotron ra-diation as recently demonstrated at the ALS, Berkeley, em-ploys a femtosecond laser to modulate the energy of anelectron bunch over the distance of the laser pulse length.The optical laser pulse modulates the energy of a storedelectron beam bunch as they co-propagate through themodulator wiggler. In a dispersive section, the modulatedelectrons become spatially separated from the unperturbedpart of the electron bunch (’electron-beam slicing’). Thespatially separated electron slice can be used to generatesub-picosecond X-rays in the so-called ’radiator’. The radi-ator can be a dipole magnet in the arc-sector or an insertiondevice in a straight section.

To enhance the overall efficiency, a laser system witha high power and a high repetition rate is needed. Using

state-of-the-art lasers, an upper limit is set on the energy ofthe storage ring by the requirement that the energy modu-lation must be several time larger than the energy spread ofthe electron beam (�E � 5�E ' 13 MeV at the SLS). Toefficiently employ the electron-beam slicing method, thebeam energy cannot be much higher than 2.5 GeV. There-fore, for intermediate energy storage rings such as the SLS(2.4- 2.7 GeV), this technique opens a unique avenue fortime-resolved experiments:� For the production of short photon pulses, the com-

bined effect of the electron beam slicing method and theuse of higher undulator harmonics means that high bright-ness sub-picosecond X-rays up to 18 keV are accessible atundulator beamlines specially developed at user facilitieswith a beam energy of about 2.5 GeV.� For the conceptual design of an ID-based sub-

picosecond photon source at the SLS, one can profit fromseveral unique attributes of the facility:� Low emittance optics� Large momentum acceptance� Long straight sections� Small gap, short period wiggler as a modulator� Linear/circular polarized twin-undulator as a radiator

for soft X-rays� Small gap, short period in-vacuum undulators as a ra-

diator for hard X-rays

2 LATTICE MODIFICATIONS

At the PX-beamline 6S we recently have installed theshort period (24 mm) in-vacuum undulator U24. As the(radiator) source for sub-picosecond X-rays, we studied toinstall small gap (4-6 mm) undulators of this type also inthe long straight section 5L (10.8 m). For this purpose, tolower the vertical �-function from nominally 3 m to � 1 mat the center of the radiator, the optics in the straight section5L needs to be modified.

An additional quadrupole triplet mounted asymmetri-cally splits the long straight 5L into two short straights ofdifferent length with �y=1.6 m and �y=0.6 m respectivelyat the center (Figure 2). The change of the ring optics isconfined between sextupole pairs in the arcs adjacent to thestraight 5L. Therefore no optical function at any sextupolewill be affected. The betatron phase advance introduced bythe modification is ��x=0 horizontally and ��y = � ver-tically (”�-trick”). Since in a regular (non-skew) lattice allnormal multipoles contain only even powers of the verticalcoordinate y, the non-linear beam dynamics remains basi-cally unchanged [4]. Marginal deteriorations are due to aslight increase of the integrated sextupole strength and due

to a deviation of ��y from � for off-momentum particles(j� p/pj � 6%).

The locally modified optics allows the installation of twosmall gap insertion devices, either two undulators (’Two-Straight Scenario’, Figure 1 and 2) or one wiggler and oneundulator (’One-straight Scenario’, Figure 3 and 4).

Figure 1: Schematic view: lattice modification and 2-straight scenario.

A sizeable dispersion function (�) is necessary toachieve a transverse spatial separation (5 �) of the energy-modulated electron slice at the radiator. Because of thesmall vertical beam size, a vertical �-function will be im-plemented.

2.1 Two-Straight Scenario

Two undulators (U24, U17) in section 5L cover the pho-ton range 2-18 keV. Laser modulation of the electron beamoccurs in the modulator (revolver-type wiggler W61/W90)installed in section 4S, immediately upstream from the ra-diator straight 5L (Figure 1 and 2). A revolver-type modu-lator is proposed to ensure the highest horizontal flux den-sity for the Material Science (MS) beamline at 4S usingthe small gap wiggler W61 (under construction) when thesub-picosecond X-ray facility is not operational. Operat-ing the laser on the 2nd harmonic in resonance with themodulator wiggler W90 has the advantage that the regularMS-beamline operation is not much affected by the sub-picosecond operation in section 5L.

Four skew quadrupoles in the arcs adjacent to 4S areused to generate a vertical asymmetric �-bump by cou-pling the horizontal � in the arc to the vertical plane atthe location of the modulator [4]. An effective dispersion�eff=8.5 mm and � 0eff ' 0 at the radiator is generatedby a vertical betatron oscillation of the energy modulatedelectrons induced by the skew quadrupoles. The princi-ple is explained in Figure 2. For electrons with an energymodulation �0.5%, a vertical spatial separation of 43 �mis achieved. The vertical beam size is 8 �m. To keep theemittance coupling low (' 0.5%), various options for cre-ating the asymmetric �-bump are under study.

2.2 One-Straight Scenario

The first radiator (undulator U24) in Figure 1 is replacedby a small gap wiggler as the modulator. The modula-tor and the radiator are now installed in the same straightsection. This scheme, shown in Figure 3 and 4, provides

Figure 2: The modulator (in 4S, with asymmetric �-bump)and the radiator(s) (in 5L, with mini-beta lattice) are in-stalled in two successive straight sections. Solid curve:closed vertical dispersion bump of the stored beam. Dashedcurve: vertical dispersion of the sliced beam.

Figure 3: Schematic view: 1-straight scenario with verticalchicane at the wiggler (modulator).

the smallest temporal stretching of the electron bunch slice(due to the non-isochronicity of the SLS storage ring) andthus the shortest X-ray pulses. We estimate that insteadof 0.3-0.4 ps the photon pulses can now be as short as 0.1ps using a 50 fs laser. In addition, the modulator wiggler(W110) is now resonant to the laser fundamental (800 nm),since a simultaneous operation with a second beamline isnot required. The closed vertical �-bump around the modu-lator is generated by a vertical 4-dipole chicane, providing�=8.8 mm and � 0=0 for the 5�-separation (48�m) of theshort photon pulses. The vertical displacement ('8 mm)

Figure 4: Mini-beta lattice with chicane for pure spatialbeam separation. The modulator W110 and the radiatorU17 are installed in the long straight section 5L. Solidcurves: vertical beta function and vertical dispersion of thestored beam. Dotted curves: horizontal beta function andvertical dispersion of the sliced beam.

of the stored beam in the modulator W110 with respect tothe radiator U17 is used to separate and block the unwantedmodulator radiation. Minimizing the photon background iscritical for creating a clean image of the source (see section3.3).

3 SUB-PICOSECOND X-RAY SOURCE

3.1 Modulator and Radiator

The following parameters, summarized in Figure 5, 6and 7, are used to predict the characteristics of the sub-picosecond X-ray source:� One-straight scenario (5L):

Ti:sapphire laser: 800 nm (fundamental), 50 fs [FWHM],1 mJ/pulse, 5 W, 5 kHz (10 kHz); modulator W110: pe-riod 110 mm, gap 7.5 mm, peak field 2.5 T, K-parameter25.6, length 2m; radiator U17: gap 4 mm, period 17 mm,peak field 1.0 T, energy range 5-18 keV (3.-11. harmonic);slicing efficieny (400 bunches): 5.5 E-9 (5 kHz) (the slicingefficiency is defined as the number of electrons with�E �0.5% / second/ beam current (400 mA)).� Two-straight scenario (4S / 5L):

Simultaneous operation of the material science (MS) beam-line 4S requires that the laser is in resonance with the sec-ond harmonic of the modulator wiggler. Ti:sapphire laser:400 nm (2nd harmonic, conversion efficiency 30-50%), 50fs [FWHM], 1 mJ/pulse, 2 W, 2 kHz (5 kHz); modulatorW90 (revolver type W61/W90): period 90 mm, gap 7.5mm, peak field 2.35 T, K-parameter 19.7, length 2m; (wig-gler W61, under construction, for regular operation at theMS-beamline (4S), period 61 mm, gap 7.5 mm, peak field1.95 T, K-parameter 11.0, length 2 m); radiator U24: gap5 mm, period 24 mm, peak field 1.10 T, energy range 2-8keV (1.-7. harmonic); radiator U17: gap 4mm, peak field1.0 T, energy range 5-18 keV (3.-11. harmonic); slicingefficiency (single bunch): 7.5 E-7, slicing efficiency (400bunches): 4.8 E-9 (5 kHz).

3.2 Laser

Initial experimental experience with the sub-picosecondX-ray source can be gained with a reliable commercial sys-tem. Commercial systems offer average output powers oftypically 1 W independent of the repetition rate (100 kHz /10 �J - 10 Hz / 100 mJ). We expect that a 50 fs Ti:sapphirelaser system based on the chirped pulse amplification withan average power of 3 W (1 mJ/pulse, 3 kHz) will commer-cially be available at the end of year 2001.

Simultaneously a system delivering pulses of 50 fs andan energy of 1 mJ scalable in repetition rate (10 kHz)will be developed at the Combustion Research Labora-tory at PSI. Such a system could be based on the Coher-ent RegA9050 system that provides 50 fs pulses with ener-gies in the microjoule range and a repetition rate variableup to 250 kHz. Here the switching is performed acousto-optically (cavity dumper). No further electro-optic switch-ing unit is then needed for further amplification. The am-

Modulator Wiggler

Wiggler W110 One Straight

length LW 2 mperiod length �W 110 mmno. periods MW 18max. �eld B0 2.5 T (Laser, 1 st Harm.)

mag. gap 7.5 mmK-value 25.6

Wiggler W90 Two Straights

length LW 2 mperiod length �W 90 mmno. periods MW 22max. �eld B0 2.35 T (Laser, 2nd Harm.)

mag. gap 7.5 mmK-value 19.7

Figure 5: Modulator parameters.

Radiator Undulator- hard X-rays -

U17/U24, 'multi-undulator' type

U17 undulator U24 undulator(in-vacuum) (in-vacuum)

type hybrid hybrid

magn. material Sm2Co17 NdFeB

magn. length LU 2 m 2 m

no. periods MU 117 81

magn. gap gU 4 mm 5 mm

max. �eld B0;U 0.92 T 1.10 T

K-value KU 1.46 2.50

energy range 5-18 keV 2-10 keV

no. harmonics 3-11 1-7

polarization linear linear

Figure 6: Radiator parameters.

plification of the �J pulses to the mJ range will be basedon the chirped pulse amplification technique [5], usingan asymmetric grating pair pulse stretcher, prior to theRegA9050 regenerative amplifier, followed by a double-pass Ti:sapphire amplifier and a grating-pair pulse com-pressor (Figure 8).

A schematic view of the amplifier is shown Figure 9.For the case that a double-pass is not sufficient to saturatethe amplifier, a multi-pass amplifier in a butterfly configu-ration will be used. Assuming saturated amplification, thestability of the amplified femtosecond pulse will be that ofthe pump laser. The design is scalable to higher repetitionrates by incorporating future developments in pump lasertechnology and additional amplifier stages. The problemof thermal lensing by the high average powers absorbed inthe Ti:sapphire crystals can be solved by cryogenic coolingof the amplifier rods [6].

Laser: Ti:Sapphire

1 st Harmonic

wavelength �L 800 nmenergy/pulse AL 1.0 mJpulse length �L 50 fs

no. optical cycles ML 18repetition rate fL 1-5 kHz (phase 1)

5-10 kHz (phase 2)

energy modulation �E:modulator W110 � 12 MeV (= 5 �E)

2nd Harmonic

wavelength �L 400 nm conv. e�.:energy/pulse AL 1.0 mJ 30-50%pulse length �L 50 fs

no. optical cycles ML 37repetition rate fL 1-2 kHz (phase 1)

2-5 kHz (phase 2)

energy modulation �E:modulator W90 � 12 MeV

Figure 7: Laser parameters.

Figure 8: Femtosecond Ti:sapphire laser system.

The laser slicing efficiency is determined by the numberof electrons per bunch with an energy modulation �E �0.5% after the laser interaction in the modulator. By nu-merical integration we calculate�E = �e R ~v? � ~E(~x; t)dtwhere ~v? is the electron transverse velocity and ~E(~x; t) theelectrical field of the laser. The waist of the laser beam atthe center of the modulator is typically 200 �m (H) x 200�m (V).

Figure 9: Schematic view of the double two-passTi:sapphire 10 kHz amplifier. P: polarizer; FR: Faradayrotator.

Figure 10: Calculated momentum and spatial distributionof the electron bunch following interaction with the laser inthe modulator. An energy modulation�0.5% is required.

3.3 Sub-Picosecond Synchrotron Pulses

The simulated slicing efficiency 5 E-9 for a 5 kHz lasersystem is in good agreement with the analytical estimategiven in [1]. Using this value, the intensity of the sub-picosescond X-ray pulses listed in Figure 12 is calculatedfrom the values given in the 3rd column, which are validfor the regular multibunch operation (400 mA, 500 Mhz)of the SLS storage ring. To obtain the values in the 3rd col-umn of Figure 12, we earlier have studied the reduction ofbrightness due to undulator phase errors, the emittance andthe energy spread of the electron beam for harmonic num-ber 1-13 [2] (Figure 11). The on axis flux density F 00:=d3N/dt(d!=!) d j�=0 [photons/sec/mrad2/0.1% bw] hasbeen calculated numerically including (i) phase errors in arealistic undulator field, (ii) electron beam emittance (�=4.8nm-rad, �x=2.6 m, �y=1.2 m, coupling 1%) and (iii) en-ergy spread (�E/E=0.09%). The natural brightness is de-fined as Bn=F00/2��2 and � =

p2�L=2� (�0 =

p�=2L;

� is the wavelength of the X-rays and L the length of theundulator.).

Imaging will be used to separate the short photon pulses.We follow the concept adopted at the SIM-beamline 11Mfor time-resolved photoelectron-microscopy on magneticmaterials [7] at the SLS (section 4). A focussing mir-ror having demagnification 1/M, located behind the doublecrystal monochromator (Si(111)), images the two parallellydisplaced source points (see Figure 2 and 4) for the slicedand unperturbed electrons in the radiator onto the interme-diate image plane. At the intermediate focus the imagesare separated by �0 = �=M ' 22�m. A pair of slits,displaced 5�y above the optical axis, will select the sub-picosecond X-ray pulses and block the long-pulse back-ground. A mechanical chopper (1-5 kHz) in front of theslits is used to match the time structure of the X-ray beamto the laser system (Figure 13).

N h� F00 Bn � �E/E

[keV] 1) 2)

1 1.5 2.1E+18 4.2E+21 2.7E-1 8.7E-13 4.6 2.1E+18 1.3E+22 1.5E-1 5.0E-15 7.7 1.4E+18 1.4E+22 1.1E-1 3.3E-17 10.8 8.1E+17 1.1E+22 8.7E-2 2.5E-19 13.9 4.3E+17 7.7E+21 7.4E-2 2.0E-111 17.0 2.2E+17 4.8E+21 6.3E-2 1.6E-113 20.0 1.1E+17 2.8E+21 5.4E-2 1.4E-1

N Phase Phase Beam ALLError Error Size Factors

1)+2) 3),1.50 4),40 5) 1)-3), 5)

1 2.4E-1 1.00 0.99 7.3E-2 1.7E-23 1.0E-1 0.99 0.95 2.8E-2 2.8E-35 5.8E-2 0.97 0.87 1.8E-2 1.0E-37 3.7E-2 0.95 0.76 1.3E-2 4.6E-49 2.6E-2 0.93 0.64 1.0E-2 2.4E-411 1.9E-2 0.90 0.51 8.3E-3 1.4E-413 1.4E-2 0.88 0.31 7.1E-3 8.7E-5

Figure 11: SLS U17 undulator: brightness reduction fac-tors.

Flux: F [photons/sec/0.1% bw]

N h� F F F

[keV] [20ps, 500MHz] [20ps, 5kHz] [100fs, 5kHz]

1 1.5 3.5E+15 3.5E+10 1.8E+73 4.6 1.3E+15 1.3E+10 6.5E+65 7.7 4.6E+14 4.6E+9 2.3E+67 10.8 1.6E+14 1.6E+9 8.0E+59 13.9 6.6E+13 6.5E+8 3.3E+511 16.9 2.6E+13 2.6E+8 1.3E+513 20.0 9.3E+12 9.5E+7 4.8E+4

Brilliance: B [photons/sec/mm2/mrad2/0.1% bw]

N h� B B B

[keV] [20ps, 500MHz] [20ps, 5kHz] [100fs, 5kHz]

1 1.5 7.1E+19 7.0E+14 3.5E+113 4.6 3.6E+19 3.6E+14 1.8E+115 7.7 1.4E+19 1.4E+14 7.0E+107 10.8 5.1E+18 5.0E+13 2.5E+109 13.9 1.8E+18 1.8E+13 9.0E+911 16.9 6.7E+17 6.5E+12 3.3E+913 20.0 2.4E+17 2.4E+12 1.2E+9

Figure 12: Average flux and brightness of the sub-picosecond X-ray pulses for the U17 undulator (integratedover 0.25 mrad (H) x 0.05 mrad (V)).

4 PICOSECOND SYNCHROTRONPULSES

We also studied the two-straight scenario for the straightsections 10S and 11M, where the elliptically polarizingtwin-undulators UE56 for beamline 11M would be used asradiators. A similar scheme as shown in Figure 2 wouldagain be applicable. Using a 5 kHz laser, for the energyrange 0.09-2 keV a flux and brightness of (0.5-1.0)E+6 and(0.6-1.3)E+10 for sub-picosecond, linear polarized pho-tons is expected. For circularly polarized photons the sub-picosecond flux and brightness would be (1.6-3.3)E+5 and(0.2-0.4)E+10 respectively.

However, time scales of interest for magnetic studies arerather in the pico- and nanosecond range, where instead of

Figure 13: Short pulse separation: imaging of the source.

slicing the intrinsic time structure of the storage ring (20 ps(rms) pulses spaced by 2-800 ns) will be used.

Figure 14: Photon beam separation: parallel vs. tiltedbeams.

The novel feature of beamline 11M is the concept usedfor helicity switching. A chicane with 7 dipoles createsa small (� 0.5mm) parallel displacement of the electronbeam in the horizontal direction which is opposite for bothundulators. The optical paths for the two photon beamswith opposite circular or linear polarization overlap almostcompletely. A single set of relatively small mirrors is usedto accept and guide both photon beams. By imaging thetwo source points with a demagnifying (4:1) optics, bothphoton beams are again separated at the intermediate focus(Figure 14). At this location a mechanical chopper alterna-tively blocks one of the two photon beams. The two photonbeams can be overlapped on the sample using a single re-focussing mirror.

5 SUMMARY

We propose to develop at the SLS a facility for sub-picosecond X-ray pulses based on electron-beam slicingmethod with the characteristics of U17 as the radiator. Theundulator will likely be the one planned for the micro-XAFS/diffraction beamline. The facility will enable time-dependent studies of the structural dynamics of condensedmatter and photochemistry and will complement time re-solved studies in the pico- and nanosecond range plannedat the photoelectron-microscopy beamline for soft X-rays.

6 REFERENCES

[1] R.W. Schoenlein, S. Chattopadhyay, H.H. Chong, T.E.Glover, P.A. Heimann, C.V. Shank, A.A. Zholents, M.S.Zolotorev, ”Generation of Femtosecond Pulses of Syn-chrotron Radiation”, SCIENCE 287 (2000) 2237 and Appl.Phys. B71 (2000) 1.

[2] G. Ingold, ”Insertion Devices: New Concepts and Perfor-mance”, Proceedings of EPAC 2000, Vienna, Austria (2000),220; G. Ingold, T. Schmidt, ”Insertion Devices for SLS”, PSIScientific Report 2000, Volume VII.

[3] A. Streun for the SLS team, ”Commissioning of the SwissLight Source”, Proceedings of PAC 2001, Chicago, UnitedStates (2001).

[4] Y. Wu, H. Nishimura, D.S. Robin, A.A. Zholents, E. For-est, ”Mini-Beta Lattice for the Femto-Second X-Ray Sourceat the ALS”, Proceedings of EPAC 2000, Vienna, Austria(2000), 1098.

[5] D. Strickland, G. Mourou, Appl. Phys. B 58 (1985) 211;M. Pessot, P. Maine, G. Mourou, Opt. Commun. 62 (1987)419; P. Maine, D. Strickland, P. Bado, M. Pessot, G. Mourou,IEEE J. Quant. Electron. 24 (1988) 398.

[6] P.A. Schulz, S.R. Henion, IEEE J. Quant. Electron. 27 (1991)1039.

[7] C. Quitmann, U. Flechsig, L. Patthey, Th. Schmidt, G. Ingold,M. Howells, M. Janousch, R. Abela, ”A Beamline for TimeResolved Photoelectron-Microscopy on Magnetic Materialsat the Swiss Light Source (SLS)”, Proc. of 2nd LEEM/PEEMWorkshop, Paris, Sept. 2000, ”Surface Science”.