DESY 04-013 January 2004 Terawatt-scale sub-10-fs laser technology – 4 0 key to generation of GW-level attosecond 0 2 pulses in X-ray free electron laser n a J 7 2 E.L. Saldin, E.A. Schneidmiller, and M.V. Yurkov ] h p Deutsches Elektronen-Synchrotron (DESY), Hamburg, Germany - c c a . Abstract s c i s We propose a technique for the production of attosecond X-ray pulses which is y h based on the use of X-ray SASE FEL combined with a femtosecond laser system. A p few-cycle optical pulse from a Ti:sapphire laser interacts with the electron beam in [ a two-period undulator resonant to 800 nm wavelength and produces energy modu- 1 lation within a slice of the electron bunch. Following the energy modulator the elec- v 6 tron beam enters the X-ray undulator and produces SASE radiation. Due to energy 3 modulation the frequency is correlated to the longitudinal position within the few- 1 cycle-driven slice of SASE radiation pulse. The largest frequency offset corresponds 1 0 to a single-spike pulse in the time domain which is confined to one half-oscillation 4 period near the central peak electron energy. The selection of single-spike pulses is 0 achieved by using a crystal monochromator after the X-ray undulator. Our studies / s show that the proposed technique is capable to produce 300 attoseconds long single c i pulses with GW-level output power in the 0.1 nm wavelength range, and is appli- s y cable to the European X-Ray Laser Project XFEL and the Linac Coherent Light h Source at SLAC. p : v i X r a Preprint submitted to Optics Communications 2 February 2008 1 Introduction At the start of this century, we have seen a revolution in synchrotron radiation source intensities. This revolution stemmed from the technique of free electron laser (FEL) based on self-amplified spontaneous emission (SASE), combined with recent progress in accelerator technologies, developed in connection with high-energylinearcolliders.In2001,theVUVFELattheTESLATestFacility at DESY(Hamburg, Germany) has successfully demonstrated saturationfrom 82 nm to 125 nm with GW-level peak power and pulse duration down to 40 fs [1,2]. It is the first result from this device that Wabnitz et al. reported in [3]. They illuminated xenon clusters with high-intensity (1014 W/cm2) VUV FEL pulses and observed an unexpectedly strong absorption of the VUV radiation. Such a highly nonlinear optical interaction between light and matter at VUV wavelength range has never been seen before and these fascinating results show the potential of this new class of light sources for scientific research. While modern third generation synchrotron light sources are reaching their fundamental performance limit, recent successes in the development of the VUV FEL at DESY have paved the way for the construction of the novel type of light source which will combine most of the positive aspects of both a laser and a synchrotron. Starting in 2004, the phase 2 extension of TTF will deliver FEL radiation down to the soft X-ray spectral range with minimum wavelength of about 6 nm in the first harmonic and reaching into ”water window” in the second harmonic [4]. RecentlytheGermangovernment,encouragedbytheseresults,approvedfund- ing a hard X-ray SASE FEL user facility – the European X-Ray Laser Project XFEL. The US Department of Energy (DOE) has given SLAC the goahead for engineering design of the Linac Coherent Light Source (LCLS) device to be constructed at SLAC. These devices should produce 100 fs X-ray pulses with over 10 GW of peak power [5,6]. These new X-ray sources will be able to produce intensities of the order of 1018 W/cm2. The main difference between the two projects is the linear accelerator, an existing room temperature linac for LCLS at SLAC, and a future superconducting linac for European XFEL. The XFEL based on superconducting accelerator technology will make possi- ble not only a jump in a peak brilliance by ten orders of magnitude, but also anincrease in average brilliance by five orders of magnitude compared to mod- ern 3rd generation synchrotron radiation facilities. The LCLS and European XFEL projects are scheduled to start operationin 2008and 2012,respectively. The motivation for the development of XFELs was recently described in de- tail in [5,6]. The discussion in the scientific community over the past decade has produced many ideas for novel applications of the X-ray laser. Brilliance, coherence, and timing down to the femtosecond regime are the three prop- erties which have the highest potential for new science to be explored with 1 an XFEL. In its initial configuration the XFEL pulse duration is about 100 femtoseconds. Even though this is a few hundreds times shorter than in third generation light sources, it can probably be further reduced to about 10 fem- toseconds [7–9]. A novel way to generate sub-10 fs x-ray pulses – the slotted spoiler method (P. Emma, 2003) has been proposed recently. This method is based on spoiling the beam phase density in a part of the electron bunch so that this part will not lase, while preserving lasing in a short length of the bunch. The FEL performance of the spoiled beam approach was computed using the time-dependent GENESIS simulation. It has been shown that it is possible to produce X-ray pulses with duration of 3-4 fs FWHM for nominal LCLS bunch compression parameters [10]. Femtosecond-resolution experiments with X-rays can possibly show directly how matter is formed out of atoms. In fact, X-ray pulse duration even shorter than one femtosecond may be useful for many scientific applications. The reason is that phenomena inside atoms occur on sub-femtosecond timescale. Generating single attosecond ∼ 0.1 nm X-ray pulses is one of the biggest challenges in physics. The use of such a tool will enable to trace process inside the atoms for the first time. If there is any place where we have a chance to test the main principles of quantum mechanics in the pure way, it is there. The interest in the science with attosecond pulses is growing rapidly in the very largelaser community. This community is familiar with attosecond pulses of light at photon energies up to about 0.5 keV (3 nm). This is achieved by focusing a fs laser into a gas target creating radiation of high harmonics of fundamental laser frequency. The key to these developments was the inven- tion of laser systems delivering pulses in the range of 5 fs with pulse energies higher than a fraction of mJ. This approach produced the first measurable XUV pulses in the 200 as regime [11,12]. In principle, table-top ultra-fast X- ray sources have the right duration to provide us with a view of subatomic transformation processes. However, their power and photon energy are by far low. The XFEL is ideally suited the purpose of this emerging field of science. Recently an approach for the generation of attosecond pulses combining fs quantum laser and harmonic cascade (HC) FEL scheme [13,14] was proposed in [15]. The HC FEL scheme has the potential to produce coherent light down to wavelengths of a few nm in an undulator sequence [16]. The analysis pre- sented in [15] shows that this technique has potential to produce 100 as long radiation pulses with MW-level of output power down to 1 nm wavelength. The X-ray SASE FEL holds a great promise as a source of radiation for gener- ating high power, single attosecond pulses. What ultimately limits the XFEL pulse duration? Since the temporal and spectral characteristics of the radia- tion field are related to each other through Fourier transform, the bandwidth of the XFEL and the pulse duration cannot vary independently of each other. There is a minimum duration-bandwidth product (uncertainty principle). The 2 shortest possible X-ray pulse durationgenerated by XFEL islimited by the in- trinsic bandwidth of the SASEprocess. Inthe case ofthe EuropeanXFEL and the LCLS, the FWHM bandwidth near saturation (at 0.1 nm) is about 0.1%, indicating a 300-as coherence time determined by the bandwidth product. Re- cently a scheme to achieve pulse durations down to 400-600 attoseconds at a wavelength of 0.1 nm has been proposed [17]. It uses a statistical properties of SASE FEL high harmonic radiation. The selection of a single 10-GW level attosecond pulses is achieved by using a special trigger in data acquisition system. A promising scheme for attophysics experiments using this approach has been studied and could be implemented in the XFEL design [18]. In this paper we propose a new method allowing to reduce the pulse length of the X-ray SASE FEL to the shortest conceptual limit of about 300 as. It is based on the application of a sub-10-fs laser for slice energy modulation of the electron beam, and application of a crystal monochromator for the selection of single attosecond pulses with GW-level output power. 2 The principle of attosecond techniques based on the use of XFEL combined with fs quantum laser AbasicschemeoftheattosecondX-raysourceisshowninFig.1.Anultrashort laser pulse is used to modulate the energy of electrons within the femtosecond slice of the electron bunch at the seed laser frequency. The seed laser pulse will be timed to overlap with the central area of the electron bunch. It serves as a seed for a modulator which consists of a short (a few periods) undulator. Following the energy modulator the beam enters the X-ray undulator. The process of amplification of radiation in this undulator develops in the same way as in a conventional X-ray SASE FEL: fluctuations of the electron beam current serve as the input signal [19]. The proposed scheme for the genera- tion of attosecond pulses is based on frequency-chirping the SASE radiation Fig. 1. Schematic diagram of attosecond X-ray source 3 Fig.2.SketchofsingleattosecondX-raypulsesynthesationthroughfrequencychirp- ing and spectral filtering Fig. 3. Sketch of two attosecond X-ray pulse sequence synthesation through fre- quency chirping and spectral filtering. Pulse separation is 2T , where T is the 0 0 Ti:sapphire laser oscillation period pulse. When an electron beam traverses an undulator, it emits radiation at the resonance wavelength λ = λ (1+K2/2)/(2γ2). Here λ is the undulator w w period, mc2γ is the electron beam energy, and K is the undulator parameter. The laser-driven sinusoidal energy chirp produces a correlated frequency chirp of the resonant radiation δω/ω ≃ 2δγ/γ. After the undulator, the radiation is passed through a crystal monochromator which reflects a narrow band- width. Since the radiation frequency is correlated to the longitudinal position within the beam, a short temporal radiation pulse is transmitted through the monochromator. Recent technological advances in ultrafast optics have permitted the genera- tion of optical pulses comprising only a few oscillation cycles of the electric 4 and magnetic fields. The pulses are delivered in a diffraction-limited beam [20]. The combination of a X-ray SASE FEL and a few-cycle laser field tech- niques is very promising. Our concept of an attosecond X-ray facility is based on the use of a few-cycle optical pulse from a Ti:sapphire laser system. This optical pulse is used for the modulation of the energy of the electrons within a slice of the electron bunch at a wavelength of 800 nm. Due to the extreme temporal confinement, moderate optical pulse energies of the order of a few mJ can result in an electron energy modulation amplitude larger than 30-40 MeV. In few-cycle laser fields high intensities can be ”switched on” nonadi- abatically within a few optical periods. As a result, a central peak electron energy modulation is larger than other peaks. This relative energy difference is used for the selection of SASE radiation pulses with a single spike in the time domain by means of a crystal monochromator. A schematic, illustrating these processes, is shown inFig. 2.Many different output fields can be realized by using different spectral windowing. For instance, it is possible to generate a sequence of 300-as X-ray pulses, separated by T (or 2T ), where T is the 0 0 0 Ti:sapphire laser oscillation period. Such operation of the attosecond X-ray source is illustrated in Fig. 3. The discussion in this paper is focused on the parameters for the European XFEL operating in the wavelength range around 0.1 nm [5]. Optimization of the attosecond SASE FEL has been performed with the three-dimensional, time dependent code FAST [21] taking into account all physical effects influ- encing the SASE FEL operation (diffraction effects, energy spread, emittance, slippage effect, etc.). In our scheme the separation of the frequency offset from the central frequency by a monochromator is used to distinguish the 300-as pulse from the 100 fs intense SASE pulse. The monochromatization is straightforward:forthe0.1nmwavelength range,Braggdiffractionisthemain tool used for such purposes. In this case, one has to take care that the short pulse duration is preserved. Transmission through the monochromator will produce some intrinsic spreading of the pulse, and the minimum pulse dura- tion which may be selected by this method is limited by the uncertainty prin- ciple. The number of possible reflections which provide the required spectral width is rather limited. We are discussing here only Ge crystals, which have the largest relative bandwidth. This is an important feature which ensures the preservation of the single-spike pulse duration. In its simplest configuration the monochromator consists of Ge crystal diffracting from the (111) lattice planes. We show that it is possible to produce X-ray pulses with FWHM du- ration of 300 as. In some experimental situations this simplest configuration of monochromator is not optimal. In particular, our study has shown that the maximum contrast of the attosecond X-ray pulses does not exceed 80% and is due to the long tail of the intrinsic crystal reflectivity curve. The obvious and technically possible solution of the problem of contrast increase might be to use a premonochromator. One can align the premonochromator so that the main peak of the spectrum is blocked. 5 3 Generation of attosecond pulses from XFEL Inthe following we illustrate the operationofanattosecond SASE FEL forthe parameters close to those of the European XFEL operating at the wavelength 0.1 nm [5]. The parameters of the electron beam are: energy 15 GeV, charge 1 nC, rms pulse length 25 µm, rms normalized emittance 1.4 mm-mrad, rms energy spread 1 MeV. Undulator period is 3.4 cm. 3.1 Slice modulation of the electron beam The parameters of the seed laser are: wavelength 800 nm, energy in the laser pulse 2–4 mJ, and FWHM pulse duration 5 fs (see Fig. 4). The laser beam is focused onto the electron beam in a short undulator resonant at the optical wavelength of 800 nm. Parameters of the undulator are: period length 50 cm, peak field 1.6 T, number of periods 2. Optimal conditions of the focusing correspond to the positioning of the laser beam waist in the center of the undulator. In laser pulses comprising just a few wave cycles, the amplitude envelope and the carrier frequency are not sufficient to characterize and control laser radiation, because the evolution of the light field is also influenced by a shift of the carrier wave with respect to the pulse peak [20]. Recently, the generation of intense, few-cycle laser pulses with a stable carrier envelope phase ϕ was demonstrated [22]. Let us 0 consider the principle question for the design of few-cycle pulse experiments: howdoesthepulsephasebehave duringlinearpropagation?Inordertoanswer ] . u . a [ ϕ = 0 ϕ = π/2 h t g n e r t s 0 d l e i f c i r t c e l E -5 0 5 10 t [fs] Fig. 4. Possible evolutions of the electric field in the 5-fs pulse. carried at a wave- length 800 nm for two different pulse phases (φ = 0,π/2) 6 40 20 ] V e M [ 0 E0 - E -20 -40 5 10 15 20 t [fs] Fig. 5. Energy modulation of the electron beam at the exit of the modulator undu- lator. The laser parameters are λ = 800 nm, W = 800 GW, and FWHM pulse peak duration of τ = 5 fs p this question, we can calculate the evolution of few-cycle pulses in vacuum, which is most conveniently described by a parabolic wave equation and by starting from a Gaussian initial spatial pulse profile [20]. Choosing the initial pulse phase to be ϕ at the beam waist, one reveals that the carrier envelope 0 phase in the far field ϕ(∞) = ϕ − π/2 undergoes a phase shift due to the 0 Guoy phase shift −π/2 [20]. Note that the Guoy phase shift and all the other changes experienced by the pulse during propagation do not depend on the initial phase ϕ . 0 For an attosecond X-ray source it is of great interest to maximize the central peak energy offset, which depends sensitively on theabsolute phase of theseed laser pulse ϕ . We start with an illustration of the few-cycle-driven energy 0 modulation assuming that the peak electric field appears at the peak of the envelope when the laser pulse passes the undulator center (i.e. ϕ = 0 at the 0 Gaussian beam waist). The interaction with the laser light in the undulator thenproducesatimedependentelectronenergymodulationasshowninFig.5. For the laser (FWHM) pulse duration of 5 fs at a laser pulse energy 2-4 mJ, we expect a central peak energy modulation 30-40 MeV. 3.2 Monochromator The width of the spectral distribution of the SASE radiation will be deter- mined by the frequency chirp, provided the frequency chirp is larger than FEL bandwidth. A monochromator may be used to select the pulses of short du- ration, due to correlation between frequency and longitudinal position in the radiation pulse. For 12 keV photons, we consider Bragg diffraction in crystals 7 as a method of bandwidth selection. Special attention is called to the fact that the relative spectral width for the given Bragg reflection is independent of the wavelength or glancing angle of X-rays and is given merely by proper- ties of the crystal and the reflecting atomic planes. In particular, it implies that the choice of a crystal and reflecting atomic planes determines the spec- tral resolution. For example, one can consider Si(111) crystals, which have a FWHM bandwidth of ∆λ/λ = 1.3 × 10−4, or Ge(111)crystals, which have a FWHM bandwidth of ∆λ/λ = 3.4 × 10−4. Monochromators at synchrotron beam lines are most commonly fabricated from silicon. The reason is that the semiconductor industry has created a huge demand for defect-free , perfect single crystals. Silicon is by no means the only choice, and in recent years diamond has become a popular alternative, due to the fact it has the highest thermal conductivity and low absorption. An attosecond X-ray source requires a relatively broadband monochromator. The larger the monochromator bandwidth is, the shorter the minimal pulse duration than can be extracted. We are discussing here only Germanium sin- gle crystals which have the largest relative bandwidth. Although Ge is not as perfect as silicon or diamond, sufficiently large perfect Ge crystals are avail- able today. For 12 keV photons Bragg peaks of Ge crystals have reflectivities of approximately 75%. Figure 6 gives an example of a reflectivity curve for a thick absorbing crystal. The drawing of Fig. 6 shows several interesting fea- tures. The shape is asymmetric and is due to absorption effect. The tails of the reflectivity curve decrease as (∆λ/λ)−2. It should be pointed out that the tail of reflectivity curve plays important role in the operation of the attosec- ond X-ray source, and this characteristic of spectral window and attosecond pulse contrast are ultimately connected. Good crystal quality is required for 0.8 y t 0.6 i v i t c e l f e 0.4 R y t i s n e 0.2 t n I 0.0 -0.10 -0.05 0.00 0.05 0.10 ∆λλ / [%] Fig.6.ReflectivitycurveforathickabsorbingcrystalintheBraggcase.Germanium, 111, 0.1 nm 8