Ae

La

b

c

a

ARRAA

KUFUUS

1

tmbsi2ttitncTwiw

mno

0h

Micron 63 (2014) 40–46

Contents lists available at ScienceDirect

Micron

j our na l ho me page: www.elsev ier .com/ locate /micron

proposal for fs-electron microscopy experiments on high-energyxcitations in solids�

. Piazzaa, P. Musumecib, O.J. Luitenc, Fabrizio Carbonea,∗

Laboratory for Ultrafast Microscopy and Electron Scattering, LUMES, ICMP, Ecole polytechnique fédérale de Lausanne, CH-1015 Lausanne, SwitzerlandDepartment of Physics and Astronomy, UCLA, Los Angeles, CA 90095, USAEindhoven University of Technology, Department of Applied Physics, P.O. Box 513, 5600 MB Eindhoven, The Netherlands

r t i c l e i n f o

rticle history:eceived 14 October 2013eceived in revised form 21 January 2014ccepted 22 January 2014vailable online 8 February 2014

eywords:

a b s t r a c t

Recent advances in ultrafast technology enable both the study and the control of materials propertiesthanks to the ability to record high temporal resolution movies of their transformations, or the ability togenerate new states of matter by selecting ad hoc an excitation to drive the system out of equilibrium.The holy grail of this type of experiments is to combine a high tuneability of the excitation with a wideobservation window. For example, this is achieved in multidimensional optical spectroscopy where theresponse to several excitation energies is monitored in a broad energy range by a large bandwidth optical

ltrafastemtosecondltrafast electron microscopyltrafast electron diffractionolid state physics

pulse. In this article, the possibility to combine the chemical sensitivity of intense tuneable X-rays pulsesfrom a free electron laser, with the wide range of observables available in an ultrafast transmissionelectron microscope is discussed. The requirements for such experiments are quantified via estimatesbased on state of the art experiments and simulations, and it is proposed that ultrafast electron imaging,diffraction and spectroscopy experiments can be performed in combination with a chemically selective

ials.

X-ray excitation of mater

. Introduction

The development of fs lasers opened the possibility to inves-igate the temporal evolution of photoexcitation in solids and

olecules. The research field of transient states in molecules haseen referred to as femtochemistry (Zewail, 1994), and ultrafastpectroscopy has been successfully applied to several domainsncluding solid state and biological physics (Chergui and Zewail,009). Despite the huge potential and the evident success of theseechniques, some limitations to their applicability persisted overhe years, confining them to specific situations. In a nutshell, whilet is often possible to measure the value of a given physical quan-ity by a large choice of static techniques, the out-of-equilibriumature of time-resolved techniques makes their output more spe-ific and its theoretical modeling more challenging and less direct.his is due to the need of referencing in time the observable,

hich is typically done via photoexcitation, and therefore is an

ntimate characteristic of such experiments. There has to be aell-defined excitation in time to be able to probe the temporal

� This is an open-access article distributed under the terms of the Creative Com-ons Attribution-NonCommercial-No Derivative Works License, which permits

on-commercial use, distribution, and reproduction in any medium, provided theriginal author and source are credited.∗ Corresponding author. Tel.: +41 216930562.

E-mail address: fabrizio.carbone@epfl.ch (F. Carbone).

968-4328/$ – see front matter © 2014 The Authors. Published by Elsevier Ltd. All rights

ttp://dx.doi.org/10.1016/j.micron.2014.01.006

© 2014 The Authors. Published by Elsevier Ltd. All rights reserved.

evolution of a system. Naturally, when the ground state proper-ties are already elusive, the description of its excitations and itsout-of-equilibrium state is a further complication. In molecules,this issue has proven less severe than in solids because electronicexcitations are more univocally described and have smaller cou-pling with the other degrees of freedom, allowing for a betterdescription of the pump excitation process. On the contrary, if oneconsiders strongly correlated systems, many-body effects induce abroad spectrum of complex electronic states extending all the wayto the visible-light region (Maldague, 1977; Stephan and Horsch,1990), often making the description of the photoexcitation chal-lenging.

Recent advances in ultrafast technology provided a large vari-ety of observables. Depending on the phenomenon investigated,nowadays time resolution between ms and as can be achievedwith X-rays and electrons, giving direct information on both thestructure and the chemistry of materials via either diffraction orspectroscopy (Carbone et al., 2012). Free-electron laser facilitiesguarantee a unique X-ray flux in a highly transversally coherentbeam, capable of few fs time resolution (Chapman et al., 2006;Rosenzweig et al., 2008; Ding et al., 2009), while state of the artelectron-beamlines and transmission electron microscopes offer

higher sensitivity and versatility at the cost of coherence and toa less extent time resolution (Fulvia Mancini et al., 2012; Siwicket al., 2003; Raman et al., 2008; Aidelsburger et al., 2010; VanOudheusden et al., 2010; Musumeci et al., 2010; Hastings et al.,

reserved.

icron

2c

vstswosse2otsmacf

•

•

tltttvfat

L. Piazza et al. / M

006; Eichberger et al., 2013). Despite this, transverse coherencean be increased using larger beamsizes.

In short, current ultrafast techniques allow the direct obser-ation of the out-of-equilibrium properties of materials via bothtructural probes and broad-band spectroscopy. The need for pho-oexcitation and consequently the observation of a perturbedystem is a key aspect, especially for the study of complex materials,here a great challenge lies in the understanding and description

f the lowest energy excitations of the ground state, which wouldeem logic to observe by the least perturbing possible probes. Thisaid, a great potential hides in time-resolved experiments: if thexcitation is carefully prepared (Fausti et al., 2011; Mansart et al.,013), it can decay on a well-defined final state, and the real-timebservation of the latter can have major advantages: (i) not onlyhe amplitude, but also the phase of the wave-function describinguch a state can be obtained, (ii) the temporal evolution of differentany-body objects, i.e. their consequentiality, can also be obtained

nd therefore the evolution of a system through a phase transitionan be described in much more detail (Mansart et al., 2012). Theollowing two examples can clarify this point:

Graphite: the phonon spectrum of graphite has been known foralmost a century. Despite this, only recently it was discovered thatsome of the vibrational modes couple to the electronic structurein a particular sequence, giving rise to photoinduced structuralchanges that can even turn graphite into diamond during a fewfs, a process that takes ages under extreme conditions in nature(Raman et al., 2008; Carbone et al., 2008; Harb et al., 2011;Carbone et al., 2009; Kanasaki et al., 2009; Acocella et al., 2010).These results shed new light on the dynamics of the structureof graphite which could not be inferred by the knowledge of itsequilibrium properties.High temperature superconductors: In recent ultrafast experi-ments on high-temperature superconductors (Mansart et al.,2013), 1.5 eV photons were used to induce a direct transitionbetween the ground state of the material and higher energy elec-tronic states, ascribed to the excitation of charge stripes andthe Cu-O charge transfer bands (Lorenzana and Seibold, 2003;Uchida et al., 1991), see schematics in Fig. 1B. In a conventionalstatic Raman experiment, as depicted in Fig. 1A, one can excitethe system with a continuous wave (CW) photon beam at 1.5 eVand analyze the Raman-shifted outgoing photons, obtaining theenergy spectrum of the low-energy excitations, structural andelectronic. In a pump-probe experiment, the final state afterimpulsive excitation at 1.5 eV is the same as in a static Ramanexperiment. Only this time, instead of analyzing the outgoingphoton, one probes the system with weak non-perturbing fspulses while the excited state still survives (few ps after exci-tation). As a result, an equivalent of the Raman spectrum isobtained in the time domain, Fig. 1C and D, via the ImpulsiveStimulated Raman Scattering mechanism (ISRS). This scenario isconfirmed by the agreement between the data (FFT) and con-ventional Raman experiments (Sugai et al., 2003), in Fig. 1E andF. These results provided the first observation of the coherentoscillation of the Cooper pair condensate in the time domain.

The aim of this article is to propose a new class of experimentshat can exploit the ISRS mechanism to provide a selective popu-ation of the different excited states and at the same time combinehe advantages of pulsed X-rays and electron beams for the inves-igation of solids. Particular emphasis is put in the discussion onhe advantages of a chemically selective excitation, as achievable

ia X-rays. In what follows, the technical requirements and theeasibility of experiments based on an X-ray photoexcitation andn EM probe will be discussed. A quantitative discussion of whatype of excitations and physics can be probed in solids by such a

63 (2014) 40–46 41

technique is proposed. These ideas borrow concepts from the com-munity of ultrafast technology, electron microscopy and condensedmatter physics, and provide an interdisciplinary challenge aimed atexploiting the most recent technologies made available worldwide.

2. The role of the excitation in time-resolved experiments

As was briefly introduced above, a key ingredient of this dis-cussion is the preparation of the excited state via the photoexcitingpulse. In fact, not only are X-rays a unique tool for probing materialsand molecules, they also provide the possibility to create chemi-cally selective excitations. Typically, ultrashort laser pulses in thevisible light region are used as a pump because of their easy avail-ability. In this energy range, temperature jumps, charge transfers,and phase transitions can be triggered. However, at these energies,the strong orbital hybridization makes the chemical selectivity ofthe pump absent in many systems, and in some cases like stronglycorrelated solids, it is often very hard to decipher the effect oflight excitation in the sample, since a thorough description of thestatic optical absorption can be lacking or be model dependent.Very novel and interesting experiments have been put forwardwhere a more precise selectivity of the excitation has been obtainedvia the use of THz pump pulses (Fausti et al., 2011). In the lowenergy region of the infrared spectrum, light can excite selectivelyatomic motions, providing a way to distort “ad hoc” the structureof a system for probing its consequent dynamics. This approachtakes advantage of the interaction between light and low-energymany-body bosonic excitations and allows a direct manipulationof collective modes in materials. However, because of the way THzpulses are generated and their intrinsic characteristics, sub-ps tem-poral resolution is often challenging. Another possibility to achieveselectivity, although local, in the excitation would be to use X-raypulses tuned to specific core levels while monitoring the conse-quent dynamics with another photon or an electron pulse.

In the case of pump-probe experiments in high-temperaturesuperconductors, it is clear that in certain circumstances one canregard a pump-probe optical experiment as a resonant Ramanexperiment. The main difference is that instead of detecting theoutgoing photon and its polarization, to obtain the low-energyexcitations spectrum, information is obtained directly in the timedomain via an appropriate ultrafast probe. In the end, by Fouriertransforming the time-domain probe signal, one ends up with aRaman spectrum. In addition, the time domain observation yieldsthe phase of the probed state and its temporal evolution throughphase transitions. In this context, as much as Resonant Inelastic X-ray Scattering (RIXS) provides additional information with respectto low-energy Raman experiments (Ament et al., 2011) thanks tochemical and orbital selectivity, an (X-ray pump)-(time resolvedprobe) experiment would take advantage of the same benefits. InRIXS, resonance with certain core-levels can greatly enhance theinelastic cross-section, i.e. the population of excited states; X-rayphotons can also probe the dispersion of low-energy excitationsand can even differentiate between different sites occupied by thesame element, if they have distinguishable absorption features. Inthe energy range of the L-edge of transition metals, X-rays have apenetration depth of few hundreds of nm, matching both the pene-tration depth of high energy electrons and infrared light. One of themain challenges of a RIXS experiment is the presence of a relativelybroad elastic peak whose tail hides the lowest energy excitations(in the THz, meV regime). On the contrary, the technique is veryeffective for energy losses in the region of a fraction of an eV to

a few eV. This behavior is complementary to the performance oftime-resolved experiments, because in this case, the lowest energy(meV, THz) excitations translate into slower oscillations requiringfew tens to few hundreds of fs time resolution. The high energy

42 L. Piazza et al. / Micron 63 (2014) 40–46

Fig. 1. (A) Schematics of a Raman experiment (left), and static Raman spectrum in a La2−xSrxCuO4 (LSCO) high temperature superconductor. (B) Scheme of the Raman processi al susF pectr

A

frcoaccttaprswT

nvolved in an ultrafast ISRS experiment. (C and D) Temporal evolution of the optic) Fourier transform of the temporal traces and comparison with the static Raman s

dapted from Mansart et al. (2013) and Sugai et al. (2003).

eatures like charge transfers instead are found in the eV region,equiring a temporal resolution within the single fs regime. Thisoncept is depicted in Fig. 2, where we show the typical excitationsbserved via RIXS; the graph is adapted from Ament et al. (2011),nd a temporal scale was added to the top of the panel, making theorrespondence between the energy of the electronic states and itsonsequent time-period. In one example, the low-lying orbital exci-ations (orbitons) which are speculated to hide from sight withinhe elastic peak of a conventional RIXS experiment may be visible in

time-resolved measurement as a hundreds of fs oscillation of therobed quantity. Moreover, looking at the example of the ultrafast

esponse of LSCO, the THz Raman spectrum of a high-temperatureuperconductor was obtained through a time domain experiment,ith no information on the dispersion of the observed features.

uning the photoexcitation to an appropriate core-level and taking

ceptibility of LSCO in different Raman geometries, B1g and B2g respectively. (E anda of the superconducting gap in LSCO.

advantage of matrix elements and selection rules, one couldperform an experiment that is sensitive also to the dispersion ofthe electronic states and, more importantly, will have a betterchemical and orbital sensitivity. Chemical selectivity of X-rays hasbeen also proposed to induce photodoping in materials throughcascade decay processes of the electron hole pairs, or by directlypromoting itinerant carriers in valence band orbitals (Campi et al.,2003). This said, in any pump-probe experiment, a certain amountof unwanted excitations will be present. For example, ultrafastexcitation of perovskites often results in strong coherent phononsbeing excited together with the electronic excitations (see the La

vibration mode observed in Mansart et al. (2013) for example).However, the ability of an EM to combine diffraction, EELS, (Piazzaet al., 2013a) and imaging would give further ability to disentanglethe different excitations during the probing.

L. Piazza et al. / Micron 63 (2014) 40–46 43

Fig. 2. Top: X-ray absorption process and generation of a final excited state. Bottom: energy and temporal scale of final excited states.A

3e

3

t(pbomoar(T1p

dapted from Ament et al. (2011).

. Quantitative analysis of an X-ray pump-TEM probexperiment

.1. The fs-TEM probe

Currently, fs-resolved TEMs operate in the regime of 1–10 elec-rons per pulse at repetition rates between 200 kHz and 2 MHzPiazza et al., 2013b; Zewail, 2010). Overall, the number of particleser second ranges from 105 to 107, of which only 10 % are detectedy the most common cameras. Since the temporal resolution isbtained by simple modification of commercial TEM columns, theseachines preserve the versatility characteristics of TEMs and can

perate in a variety of modes including imaging, diffraction andnalytical microscopy. XFEL sources can on the other hand provideepetition rates between few Hz (Elettra, 2014) and tens of kHz

xfel, 2014). Therefore, to retain the same signal per second in theEM probe, the electron pulses should contain between 103 and04 particles each. To have such an amount of charge in a sub-psulse and yet retain the energy spread within few eV and the spatial

coherence required for imaging and diffraction, a careful design ofthe TEM source and column is needed.

The spatial or transverse coherence length is given by L⊥ ≡�/��, with � being the De Broglie wavelength and �� the uncorre-lated angular spread. In order not to waste any electrons, the beamdiameter D should be matched to the sample size. Then the properfigure of merit for spatial coherence (transverse beam quality) isthe relative coherence length L⊥/D, a conserved quantity whichis inversely proportional to the normalized emittance εn : L⊥/D ≈�C/(4εn), with �C = h/mc = 2.43 pm the Compton wavelength. Foramplitude contrast imaging at 100 keV with near-atomic resolu-tion a relative coherence length around 10−3 is ideal, i.e. εn ≈0.5 nm · rad. Similar numbers are required for the most demandingdiffraction measurements, i.e. diffraction of micron-sized proteincrystals or nano-diffraction of solid state samples. However, by

properly correcting aberrations, imaging with sub-nm resolutionis possible also with a completely incoherent source (Pennycookand Jesson, 1990). A normalized emittance of 0.5 nm·rad can inprinciple be achieved by photoemission with a micron diameter

4 icron 63 (2014) 40–46

lbtswgsaacsepbtiogqbtvarrali

bOTwctmtnrtMIrespop(e

apc1

stccewmtpo

Fig. 3. Electron pulses duration and energy spread as a function of the beam charge.

4 L. Piazza et al. / M

aser spot on the photocathode, as was recently demonstratedy Kirchner et al. (2013) using pulses containing one only elec-ron. To extract 103 electrons in a single pulse from a micronized spot a field strength of a few tens of MV/m is required,hich can be done quite straightforwardly, using an RF photo-

un. Alternatively, one could use the recently developed ultracoldource, based on photoionization of a laser-cooled and trappedtomic gas (Engelen et al., 2013; McCulloch et al., 2013), which haslready been demonstrated to deliver the required combination ofharge and emittance in a picosecond pulse. Since the ultracoldource allows a much larger source size for the same normalizedmittance, smaller extraction fields are required and space chargeroblems are much less of an issue. Due to Coulomb forces theunch will expand to many picosecond lengths during its passagehough the TEM. Compressing the bunch back to sub-ps lengths willnevitably lead to an increase of the energy spread, as the productf uncorrelated energy spread �U and bunch length �t – the lon-itudinal normalized emittance ε‖ ≈ �U�t/mc – is a conserveduantity in absence of nonlinear forces. The required longitudinaleam quality is �U�t ≈ 1 ps · eV, corresponding by changing unitso ε‖ � 0.5 nm · rad, and, interestingly, very similar to the trans-erse beam quality requirement: the relative coherence should bepproximately equal in all directions. Producing bunches with theequired initial longitudinal emittance is not a problem using eitheregular photocathodes or the ultracold source. The challenge is topply a compression technique which conserves transverse andongitudinal emittance while the originally dilute, many-ps bunchs converted into a dense, space-charge-dominated, sub-ps bunch.

In the past few years several bunch compression methods haveeen devised (Van Oudheusden et al., 2010; Fill et al., 2006; Vanudheusden et al., 2007; Tokita et al., 2010). Grzelakowski andromp (2013) recently proposed a particularly elegant method,hich employs the mirror symmetry of a spherical electrostatic

apacitor. As it is a passive, static device, no additional synchroniza-ion is required, which is an important advantage. However, this

ethod still needs to be proven experimentally. Here we considerhe possibility of importing in time-resolved TEMs the same tech-ique which has allowed a significant improvement in temporalesolution in ultrafast electron diffraction systems, that is longi-udinal (temporal) focusing using a radiofrequency cavity (Fulvia

ancini et al., 2012; Gao et al., 2013a; Chatelain et al., 2012).t has been shown that the resonant electromagnetic fields in aadiofrequency cavity can be used to impart a velocity-chirp to thelectron beam, thus reversing the longitudinal space charge expan-ion which otherwise limits the pulse length for larger number ofarticles. After passage through the cavity, the particles in the backf the pulse have higher velocity and can overtake the front of theulse so that at the sample the beam reaches a longitudinal focusVan Oudheusden et al., 2010; Fill et al., 2006; Van Oudheusdent al., 2007).

In the TEM, designing and tuning the radiofrequency cavity tochieve maximum compression is less straightforward than in sim-ler UED machines due to multiple transverse crossovers along theolumn. Preliminary estimates indicate that it is possible to obtain03 particles in <200 fs at the sample.

To estimate the achievable beam parameters upon RF compres-ion in a TEM, we simulated using the General Particle Tracer codehe condenser stage of a TEM with a modified gun to allow photo-athode illumination and dynamic time-resolved mode. The TEMolumn and the position of the lenses for the simulation werextracted from Reed et al. (2010). The results of the simulationithout any radiofrequency cavity are in good agreement with the

easurements of pulse length and energy spread carried out using

he PINEM effect (Piazza et al., 2013b). In Fig. 3, the experimentalulses duration (blue symbols) and energy spread (black symbols)f an uncompressed beam are shown as a function of the charge in

Inset: bunches duration and energy spread for an RF compressed beam. (For inter-pretation of the references to color in the text, the reader is referred to the webversion of the article.)

the bunches together with the results of GPT simulations (blue andblack lines).

Using this model, we can then investigate the effect of insert-ing a radiofrequency cavity for beam compression in the column. Aradio-frequency TM010 mode cavity resonant at 3 GHz was insertedat z = 60 cm from the virtual cathode source point to compress thebunch. Preliminary estimates indicate that it is possible to obtain103 particles in <200 fs at the sample with comparable transversebeam quality with current designs. Since the mechanism for com-pression is to accelerate the particles in the back of the beam anddecelerate the ones in the front, a larger beam energy spread isinduced by the cavity. This does not present a major problem fordiffraction and imaging, although the latter can be affected by chro-matic abberations, but it certainly hinders the possibility of highresolution spectroscopy. With a combination of multiple (not justone) longitudinal lenses then one could design a “condenser” stagefor the time-energy phase space and control both the beam bunchlength and energy spread at the sample. We leave this possibilityas a topic for further more detailed study. In the inset of Fig. 3, thepulses duration and energy spread are displayed for an RF com-pressed beam.

It is important to notice at this point that the imaging per-formances of such an instrument will be different from those of aconventional TEM. However, when combining space and time res-olution, no technique exists which is capable of reaching fs and nmresolution in time and direct space simultaneously. There is a greatdeal of phenomena that could be investigated successfully with aspatial resolution of tens to hundreds of nm and fs temporal resolu-tion. To name a few, direct imaging of magnetic domains and theirmotion could be performed. Also, the dynamics of the vortex latticein superconductors, or the temporal evolution of charge domainsand phase separation are topics for which the ability to combinetime and direct-space imaging will be a breakthrough. In a TEM,bright field and dark field images already give a contrast to mor-phological features of a specimen, while in out-of-focus Lorentzmicroscopy enhanced contrast to magnetic domains and chargepatterns can be obtained (Cottet et al., 2013).

3.2. The XFEL pump

The impulsive stimulated Raman scattering effect has beenexploited for the study of coherent structural and electronic exci-tations via the observation of changes in the optical constants

icron

(auape1orh

eawvetapt�cieIctocisurrs

epeo

3

swetstsebsaue(sIoRittt

Ding, Y., Brachmann, A., Decker, F.-J., Dowell, D., Emma, P., Frisch, J., Gilevich, S.,Hays, G., Hering, P., Huang, Z., et al., 2009. Physical Review Letters 102, 254801.

L. Piazza et al. / M

Mansart et al., 2013; Stevens et al., 2002). For visible-light pumps,bsorbed fluences in the range of 100 �J/cm2 per pulse have beensed for photoexcitation. In typical pump-probe experiments, suchn excitation is found to produce a background of linearly excitedarticle-hole pairs on top of the desired Raman excitations. In thexample from Mansart et al. (2013), it was possible to obtain about0−3 to 10−2 ISRS excited states per site, yielding a signal in therder of 10−3 in variation of the dielectric function. In recent fs-esolved electron scattering experiments signals of similar strengthave been detected successfully.

To estimate the yield of the Raman excited particles per site forxcitation with X-rays one needs to estimate how many photonsre absorbed by the material at a given energy. For the examplee used, LSCO, it is well known that the Cu L-edge (930 eV) pro-

ides a very good yield for resonant Raman experiments (Amentt al., 2011). At this energy, the absorption of a 200 nm sample, idealhickness for being transparent to 300 keV electrons, accounts forbout the 70 % of the impinging beam (Henke, 2014). For an XFELump, each pulse can be expected to contain around 1012 pho-ons. If one considers these photons to be absorbed in a volume of

× 50 × 50 × 0.1 �m3, the number of photons absorbed per sitean be as much as 10−1. The fluorescence yield at the Cu L-edges in the order of 10−2, of which up to 20 % can be due to inelasticvents (Ament et al., 2011; Krause, 1979), leaving a net number ofSRS excitation per site in the order of 10−3, in the range of the opti-al ISRS experiments. Such an excitation yield per pulse is expectedo provoke changes in the spectroscopic and diffraction propertiesf the material in the order of the fraction of a percent, while weannot make an estimate of the effects of such changes observablen imaging. The visibility of such a change depends in turn on theignal to noise ratio and integration time of the experimental set-p. In the optical experiments for example, these phenomena wereesolved by summing as many as 1.6 million pulses at a repetitionate of 1 KHz for a total acquisition time of 20 min, having a shot tohot noise in the experimental probe at the percent level.

Based on these estimates, we argue that for a large and gen-ral class of pump-probe experiments aimed at investigating theroperties of coherent modes, the combination XFEL-pump andlectron-probe offers a competitive alternative to more standardptical pump-probe schemes.

.3. The synchronization

Practically, X-ray pump/fs-TEM experiments would require theynchronization of a laser source to the X-ray source. The fs laserill serve as a drive for the TEM photocathode, allowing to generate

lectron bunches with different duration and charge via the pho-oemission process. Synchronization between a laser and an X-FELource has been recently discussed, proposing different approacheshat can lead to a stabilization of the relative delay between theources in the order of the pulse duration itself (Winter, 2005; Kimt al., 2008). An alternative way to solve the problem is offeredy time-stamping each image, enabling data sorting in postproces-ing the images. This has been successfully demonstrated at XFELchieving sub-10 fs temporal resoultion (Riedel et al., 2013). Forltrafast electron pulses time-stamping was pioneered for MeVlectron diffraction setup using electro-optic sampling methodsScoby et al., 2010). New techniques have recently been demon-trated to work also for non-relativistic electrons (Gao et al., 2013b).n addition, a RF-fs laser synchronization technique has been devel-ped which allows the measurement of the phase of the amplifiedF field in the cavity with respect to the fs laser pulse for each

ndividual shot with 30 fs resolution (Brussaard et al., 2013). Sincehe main cause of arrival time jitter is due to RF phase jitter ofhe compression cavity, the latter technique should allow accurateime-sorting of the individual images.

63 (2014) 40–46 45

In the case discussed here, an extra complication arises from theneed to use an RF electron gun in the TEM. Such a device requiresthe synchronization between an RF cavity and the laser source,which can be done also with high accuracy using the techniquedescribed in Brussaard et al. (2013). In total, the jitter of such atime-resolved experiment will be mainly affected by two compo-nents: the laser-XFEL jittering, and the laser-TEM jittering. Each ofthese two components has different sources: (1) the XFEL-laser jit-ter originates from the laser rep-rates instabilities, mainly due totemperature drifts in the cavity, and the intrinsic noise of the XFELsource. (2) The laser-TEM jitter is mainly caused by the synchro-nization between the RF cavity, temperature drifts plus the noiseof the RF source, and the laser itself, temprature drift of the opticalcavity. These sources of non-idealtiy can be considered to a firstapproximation as independent, and therefore, summed quadrati-cally. Based on current technology available, the overall jitteringof such a set-up will be expected to be about 50 % higher then thelaser pulse duration.

4. Conclusions

In this article, we discussed the feasibility of combining ultra-short X-ray sources, currently under development, with an ultrafastTEM. The goal of these experiments is to provide a uniquelybroad-band view of materials and nanostructures and their trans-formations happening in the fs time-scale. Such an approach wouldbenefit from both most recent achievements in ultrafast science, i.e.time-resolved TEM and XFEL technology, and promises to delivercombined structural, spectroscopic and morphological informationon an extraordinarily wide class of systems, ranging from bio-molecules to strongly correlated solids.

Acknowledgements

The authors acknowledge useful discussions with prof. M. Cher-gui. FC acknowledges support from an ERC starting grant N◦ 258697USED.

References

Acocella, A., Carbone, F., Zerbetto, F., 2010. Journal of the American Chemical Society132, 12166.

Aidelsburger, M., Kirchner, F., Krausz, F., Baum, P., 2010. Proceedings of the NationalAcademy of Sciences 107, 19714.

Ament, L.J., van Veenendaal, M., Devereaux, T.P., Hill, J.P., van den Brink, J., 2011.Reviews of Modern Physics 83, 705.

Brussaard, G., Lassise, A., Pasmans, P., Mutsaers, P., van der Wiel, M., Luiten, O., 2013.Applied Physics Letters 103, 141105.

Campi, G., Di Castro, D., Dell’omo, C., Agrestini, S., Saini, N., Bianconi, A., Bianconi,G., Barba, L., Colapietro, M., 2003. International Journal of Modern Physics B 17,836.

Carbone, F., Baum, P., Rudolf, P., Zewail, A.H., 2008. Physical Review Letters 100,035501 (see also Park and Zuo (2010) and Carbone et al. (2011)).

Carbone, F., Kwon, O.-H., Zewail, A.H., 2009. Science 325, 181.Carbone, F., Baum, P., Rudolf, P., Zewail, A.H., 2011. Physical Review Letters 106,

139901.Carbone, F., Musumeci, P., Luiten, O., Hebert, C., 2012. Chemical Physics 392, 1.Chapman, H.N., Barty, A., Bogan, M.J., Boutet, S., Frank, M., Hau-Riege, S.P., March-

esini, S., Woods, B.W., Bajt, S., Benner, W.H., et al., 2006. Nature Physics 2,839.

Chatelain, R.P., Morrison, V.R., Godbout, C., Siwick, B.J., 2012. Applied Physics Letters101, 081901.

Chergui, M., Zewail, A.H., 2009. ChemPhysChem 10, 28.Cottet, M., Cantoni, M., Mansart, B., Alexander, D., H ebert, C., Zhigadlo, N., Karpinski,

J., Carbone, F., 2013. Physical Review B 88, 014505.

Eichberger, M., Erasmus, N., Haupt, K., Kassier, G., von Flotow, A., Demsar, J., Schwo-erer, H., 2013. Applied Physics Letters 102, 121106.

http://www.elettra.trieste.it/FERMI/index.php?n=Main.MachineEngelen, W., van der Heijden, M., Bakker, D., Vredenbregt, E., Luiten, O., 2013. Nature

Communications 4, 1693.

4 icron

F

FF

G

GGH

H

hK

KKKLMM

M

M

M

PPP

6 L. Piazza et al. / M

austi, D., Tobey, R., Dean, N., Kaiser, S., Dienst, A., Ho mann, M., Pyon, S., Takayama,T., Takagi, H., Cavalleri, A., 2011. Science 331, 189.

ill, E., Veisz, L., Apolonski, A., Krausz, F., 2006. New Journal of Physics 8, 272.ulvia Mancini, G., Mansart, B., Pagano, S., Van Der Geer, B., De Loos, M., Carbone, F.,

2012. Nuclear Instruments and Methods in Physics Research Section A: Accel-erators, Spectrometers, Detectors and Associated Equipment 691, 113–122.

ao, M., Lu, C., Jean-Ruel, H., Liu, L.C., Marx, A., Onda, K., Koshihara, S.-y., Nakano, Y.,Shao, X., Hiramatsu, T., et al., 2013a. Nature 496, 343.

ao, M., Jiang, Y., Kassier, G., Miller, R.D., 2013b. Applied Physics Letters 103, 033503.rzelakowski, K.P., Tromp, R.M., 2013. Ultramicroscopy 130, 36.arb, M., Jurgilaitis, A., Enquist, H., Nüuske, R., Schmising, C.v.K., Gaudin, J., Johnson,

S., Milne, C., Beaud, P., Vorobeva, E., et al., 2011. Physical Review B 84, 045435.astings, J., Rudakov, F., Dowell, D., Schmerge, J., Cardoza, J., Castro, J., Gierman, S.,

Loos, H., Weber, P., 2006. Applied Physics Letters 89, 184109.ttp://henke.lbl.gov/optical constants/filter2.htmlanasaki, J., Inami, E., Tanimura, K., Ohnishi, H., Nasu, K., 2009. Physical Review

Letters 102, 087402.im, J., Cox, J.A., Chen, J., Kärtner, F.X., 2008. Nature Photonics 2, 733.irchner, F., Lahme, S., Krausz, F., Baum, P., 2013. New Journal of Physics 15, 063021.rause, M.O., 1979. Journal of Physical and Chemical Reference Data 8, 307.orenzana, J., Seibold, G., 2003. Physical Review Letters 90, 066404.aldague, P.F., 1977. Physical Review B 16, 2437.ansart, B., Cottet, M.J., Penfold, T.J., Dugdale, S.B., Tediosi, R., Chergui, M., Carbone,

F., 2012. Proceedings of the National Academy of Sciences 109, 5603.ansart, B., Lorenzana, J., Mann, A., Odeh, A., Scarongella, M., Chergui, M., Carbone,

F., 2013. Proceedings of the National Academy of Sciences 110, 4539.cCulloch, A., Sheludko, D., Junker, M., Scholten, R., 2013. Nature Communications

4, 1692.usumeci, P., Moody, J., Scoby, C., Gutierrez, M., Westfall, M., Li, R., 2010. Journal of

Applied Physics 108, 114513.ark, H., Zuo, J.M., 2010. Physical Review Letters 105, 059603.ennycook, S., Jesson, D., 1990. Physical Review Letters 64, 938.iazza, L., Ma, C., Yang, H., Mann, A., Zhu, Y., Li, J., Carbone, F., 2013a. Structural

Dynamics 1, 014501.

63 (2014) 40–46

Piazza, L., Masiel, D., LaGrange, T., Reed, B., Barwick, B., Carbone, F., 2013b. ChemicalPhysics 423, 79.

Raman, R.K., Murooka, Y., Ruan, C.-Y., Yang, T., Berber, S., Tománek, D., 2008. PhysicalReview Letters 101, 077401.

Reed, B., LaGrange, T., Shuttlesworth, R., Gibson, D., Campbell, G., Browning, N., 2010.Review of Scientific Instruments 81, 053706.

Riedel, R., Al-Shemmary, A., Gensch, M., Golz, T., Harmand, M., Medvedev, N., Pran-dolini, M., Sokolowski-Tinten, K., Toleikis, S., Wegner, U., et al., 2013. NatureCommunications 4, 1731.

Rosenzweig, J., Alesini, D., Andonian, G., Boscolo, M., Dunning, M., Faillace, L.,Ferrario, M., Fukusawa, A., Giannessi, L., Hemsing, E., et al., 2008. Nuclear Instru-ments and Methods in Physics Research Section A: Accelerators, Spectrometers,Detectors and Associated Equipment 593, 39.

Scoby, C., Musumeci, P., Moody, J., Gutierrez, M., 2010. Physical Review SpecialTopics-Accelerators and Beams 13, 022801.

Siwick, B.J., Dwyer, J.R., Jordan, R.E., Miller, R.D., 2003. Science 302, 1382.Stephan, W., Horsch, P., 1990. Physical Review B 42, 8736.Stevens, T., Kuhl, J., Merlin, R., 2002. Physical Review B 65, 144304.Sugai, S., Suzuki, H., Takayanagi, Y., Hosokawa, T., Hayamizu, N., 2003. Physical

Review B 68, 184504.Tokita, S., Hashida, M., Inoue, S., Nishoji, T., Otani, K., Sakabe, S., 2010. Physical Review

B 105, 215004.Uchida, S., Ido, T., Takagi, H., Arima, T., Tokura, Y., Tajima, S., 1991. Physical Review

B 43, 7942.Van Oudheusden, T., De Jong, E., Van der Geer, S., t Root, W., Luiten, O., Siwick, B.,

2007. Journal of Applied Physics 102, 093501.Van Oudheusden, T., Pasmans, P., Van Der Geer, S., De Loos, M., Van Der Wiel, M.,

Luiten, O., 2010. Physical Review Letters 105, 264801.Winter, A., 2005. Proceedings of the 27th International Free Electron Laser Confer-

ence , Stanford, California.https://www.xfel.eu/overview/facts and figures/Zewail, A.H., 1994. Femtochemistry: Ultrafast Dynamics of the Chemical Bond, vol.

1. World Scientific.Zewail, A.H., 2010. Science 328, 187.