Вы находитесь на странице: 1из 6

LETTERS

PUBLISHED ONLINE: 25 JANUARY 2016 | DOI: 10.1038/NPHOTON.2015.264

Femtosecond and nanometre visualization of


structural dynamics in superheated nanoparticles
Tais Gorkhover1,2*, Sebastian Schorb1,2, Ryan Coffee1,3, Marcus Adolph2, Lutz Foucar4,5, Daniela Rupp2,
Andrew Aquila1,6,7, John D. Bozek1,8, Sascha W. Epp4,9, Benjamin Erk4,9,10, Lars Gumprecht6,
Lotte Holmegaard6,11, Andreas Hartmann12, Robert Hartmann12, Gnter Hauser13, Peter Holl12,
Andre Hmke4,9, Per Johnsson14, Nils Kimmel13, Kai-Uwe Khnel9, Marc Messerschmidt1,15,
Christian Reich12, Arnaud Rouze16,17, Benedikt Rudek4,9,18, Carlo Schmidt4,9, Joachim Schulz6,7,
Heike Soltau12, Stephan Stern6,19, Georg Weidenspointner13,20, Bill White1, Jochen Kpper6,19,
Lothar Strder12,21, Ilme Schlichting4,5, Joachim Ullrich4,9,18, Daniel Rolles4,5,22, Artem Rudenko4,8,22,
Thomas Mller2 and Christoph Bostedt1,3,23,24*
The ability to observe ultrafast structural changes in nanoscopic samples is essential for understanding non-equilibrium
phenomena such as chemical reactions1, matter under extreme
conditions2, ultrafast phase transitions3 and intense lightmatter
interactions4. Established imaging techniques are limited either
in time or spatial resolution and typically require samples to be
deposited on a substrate, which interferes with the dynamics.
Here, we show that coherent X-ray diffraction images from
isolated single samples can be used to visualize femtosecond
electron density dynamics. We recorded X-ray snapshot images
from a nanoplasma expansion, a prototypical non-equilibrium
phenomenon4,5. Single Xe clusters are superheated using an
intense optical laser pulse and the structural evolution of the
sample is imaged with a single X-ray pulse. We resolved ultrafast
surface softening on the nanometre scale at the plasma/vacuum
interface within 100 fs of the heating pulse. Our study is the rst
time-resolved visualization of irreversible femtosecond processes
in free, individual nanometre-sized samples.
Direct imaging of ultrafast processes in single isolated nanometre-sized samples is virtually impossible with conventional
methods. Electron microscopy requires the samples to be deposited
on a substrate, interfering with their dynamics, and is limited in
time resolution. Conversely, optical light scattering within the femtosecond domain lacks spatial resolution. When aiming for a combination of high temporal and spatial resolution, most techniques
fail or require integration over many images. X-ray free-electron
lasers (FELs) are a promising tool for investigating phenomena
with a single exposure on a femtosecond timescale and with nanometre spatial resolution. Initial studies have yielded insights into the
dynamics during laser ablation6, lattice dynamics in nanocrystals7,
shock waves in materials8 and protein quakes9. To date, direct
single-exposure imaging of ultrafast dynamics has been restricted
to macroscopic specimens6 or spatially xed nanosample ensembles10, as well as to picosecond time resolutions. For many nonequilibrium phenomena, particle ensembles can hide the relevant
effects due to averaging over the laser focus intensity distribution,
or because of sample heterogeneity11,12. Also, sample supports
may lead to energy and charge transfer between the substrate and
the sample, substantially modifying the ultrafast dynamics4,1315.
In this study, we have imaged sub-picosecond structural changes

in individual and laser-heated nanosamples not interacting with


the surroundings. Our study paves the way to following diverse
dynamical processes via direct imaging of a single nanospecimen
under well-dened conditions.
Individual clusters with radii ranging from 15 to 30 nm were
exposed to a single near-infrared (NIR) laser pump and a subsequent
X-ray probe pulse at the Linac Coherent Light Source (LCLS)16. The
delay between pulses was varied from the femtosecond up to the picosecond regime. A schematic of the experiment is presented in Fig. 1.
The focused X-ray pulses exhibited a 0.8 nm wavelength, pulse energies up to 2 mJ and pulse durations of 50 fs, reaching intensities up to
1 1017 W cm2, which is sufcient to record single-shot images of
single non-crystalline nanoparticles. X-ray diffraction images from
single samples were recorded with a pn junction charge coupled
device (pnCCD) located downstream of the interaction region, with
scattering angles from 1.4 to 8.6 (ref. 17; see Methods). For each
scattering pattern the coincident ion spectra were recorded (see
Methods). The heating NIR laser pulse had a wavelength of 800
nm and pulse duration of 70 fs, focused to an intensity up to 2
1015 W cm2. The temporal overlap between the X-ray and NIR
lasers was determined by two-colour ionization of nitrogen molecule
dication N22+ (ref. 18; see Methods).
For the data analysis we only considered events from clusters in
the centre of both laser foci under well-dened conditions (see
Methods and Supplementary Information). Pump and probe
delays on the picosecond scale lead to a dramatic decrease in the
total diffraction signal intensity in the brightest images (see
Methods and Supplementary Information) as shown in Fig. 2.
High-quality diffraction patterns from pristine particles (top left
panels in Fig. 2a,b) exhibit the ring structure characteristics of
spherical particles. However, the recorded diffraction data differ
dramatically for the NIR-laser-heated clusters. Even after 1 ps, all
higher-order information has completely faded. After 2 ps, the
detected parts of the zero-order signal can no longer be distinguished from the isotropic uorescence background (see Methods).
Images from similarly sized clusters, but recorded with delays
below 1 ps (Fig. 2b), show a less pronounced decay of the coherent
diffraction signal, but exhibit signicant modications in the intensity distribution of the diffraction pattern compared with unheated
particles. The time-dependent changes in the scattering pattern can

A full list of author afliations appears at the end of the paper.


NATURE PHOTONICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephotonics

2016 Macmillan Publishers Limited. All rights reserved

LETTERS

NATURE PHOTONICS

DOI: 10.1038/NPHOTON.2015.264

NIR

Xe cluster source
X-ray FEL

z = 220 mm

Variable delay

Figure 1 | Experimental set-up. Single NIR and X-ray laser pulses were collinearly focused on single Xe nanoclusters. The NIR pulse superheats the cluster
into a rapidly expanding nanoplasma. The X-ray pulse diffracts from the nanoplasma and projects the transient structural information on the pnCCD detector.
The temporal delays between the pumping NIR and probing X-ray laser pulses were varied from 100 fs to several picoseconds.

b
0 fs

0.50
0.25

1 ps

1.5 ps

0.0

0.5

3 ps

250 fs

5 ps

500 fs

1.0

0.0

0.5

1.0

Intensity (a.u.)

1.00
0.75

100

Intensity (a.u.)

2 ps

25

100

Intensity (a.u.)

FEL only

q (nm1)

100

Resolution, d (nm)
15
10

5
0 fs

10
1
0.1
250 fs

10
1
0.1

500 fs

10
1
0.1
0.25

0.50

0.75

1.00

Spatial frequency, q (nm1)

Figure 2 | Time-dependent diffraction patterns. a, Single-shot X-ray diffraction patterns from single Xe clusters recorded with picosecond delays between
the NIR laser heat and imaging FEL pulse. Only shots recorded with the highest FEL intensities were selected (see Supplementary Information). The coherent
diffraction rings from a pristine cluster (upper left panel) indicate the hard-sphere shape of the sample. The number of rings is reduced for delayed X-ray
pulses. After 2 ps, the coherent diffraction intensity deceases dramatically. The strong isotropic uorescence indicates that the signal was produced near the
highest X-ray laser intensities (see Methods and Supplementary Information). b, Diffraction patterns produced during the rst 500 fs after sample heating.
The brightest diffraction patterns of similarly sized nanosamples (r = 18 2 nm) and their radial averages are shown corrected for uorescence and parasitic
beamline stray light. The images were recorded coincident with and 250 and 500 fs after the exciting infrared pulse. The higher-order diffraction rings
disappear into the noise level with increasing time delay between the NIR and FEL laser pulses.

be distinguished best in the radial average plots of the single-shot


images. The radial plots were corrected for the uorescence background (see Methods) and exhibit a high signal-to-noise ratio of
over more than three orders of magnitude. The data show that a
clear loss of higher-order information has already occurred within
the rst hundred femtoseconds after laser heating of the clusters,
as demonstrated in Fig. 2b. The radial average of the diffraction
pattern from the pristine cluster shows maxima up to the fth
order and the signal reaching the edge of the detector at 1.15 nm1
spatial frequency. Both the diffraction image and the radial average
show that, after 250 fs, the signal can be identied only up to the
third order and after 500 fs only up to the second order.
2

Based on the established view of the interaction between the NIR


pulse and the cluster4, one would expect very different diffraction
patterns. During the ionizing NIR laser pulse excitation, electrons
are rapidly injected into the surrounding vacuum and charge the
nanosample early in the pulse. The growing positive space charge
on the nanometre-sized sample connes the electrons released
during the laser pulse and forms a non-equilibrium nanoplasma.
Under the current experimental conditions, charge states up to Xe20+
are created, but fewer than 1% of the delocalized electrons can
escape from the cluster and so most electrons are bound to the deep
cluster Coulomb potential. Subsequently, the conned hot electrons
initiate the expansion as they transfer their kinetic energy to the ions.

NATURE PHOTONICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephotonics

2016 Macmillan Publishers Limited. All rights reserved

NATURE PHOTONICS

Particles density prole

Measured and calculated X-ray diffraction

Visualization

1.0

Calculated diffraction
rsolid = rini = 17 nm

100

0.8
0.6

Counts

0 fs

Normalized 3D density

Temporal delay

LETTERS

DOI: 10.1038/NPHOTON.2015.264

Solid density
100 vol%

0.4
0.2

10

Measurement

0.0
0

10

15

20

25

30

35

0.2

0.4

1.0

100

0.8

Pristine cluster
shape

0.6

Counts

100 fs

Normalized 3D density

68 vol%

0.4

Excited cluster

0.2

Simulated shape
with a solid core

10
1

10

15

20

25

30

0.2

35

0.4

100
Counts

Normalized 3D density

1.0
0.6
54 vol%

0.4

1.0

10
1

0.2
0.0
5

10

15

0.2

20 25 30 35 40

0.4

0.6

1.0

100

0.8
0.6

40 vol%

0.4

0.8

1.0

Spatial frequency, q (nm1)

Counts

Normalized 3D density

0.8

rsolid = 16.4 nm
rini = 20.2 nm

Distance (nm)

500 fs

0.6
Spatial frequency, q (nm1)

0.8

rsolid = 15.5 nm
rini = 17.8 nm
Recovered initial cluster

Distance (nm)

250 fs

1.0

0.0
0

0.8

0.6
Spatial frequency, q (nm1)

Distance (nm)

0.2

rsolid = 12.5 nm
rini = 17 nm

10
1

0.0
0

10

15

20

25

30

35

0.2

0.4

0.6

0.8

1.0

Spatial frequency, q (nm1)

Distance (nm)

Figure 3 | Simulation of diffraction patterns. ad, Evolution of the superheated nanoclusters at simultaneous NIRFEL excitation (a), and 100 fs (b), 250 fs (c)
and 500 fs (d) after the NIR pulse. Right panels: Measured diffraction integrated radially (histogram bars), tted diffraction patterns (solid blue lines) and
recovered diffraction from the pristine cluster (dashed lines). The experimental data indicate an inhomogeneous density distribution of the sample in bd.
The decrease in higher-order diffraction signal intensity suggests that the outer parts of the sample dilute due to expansion of the sample. Using the Guinier
approximation, we tted the observed diffraction patterns (solid line) with density proles (shown on the left). Left panels: Comparison of tted electron
density proles of the excited clusters (solid blue lines) and the recovered pristine shapes (dashed lines). The simple model of diffraction from a hard sphere
already fails at 100 fs after the heating infrared laser pulse, as shown in b. Here, the sample is already better described as a solid intact core with radius rsolid
(solid blue) and an expanding shell (solid yellow) with an exponentially decaying density prole. With increasing time delay from b to d, the results strongly
support the scenario of a rapidly growing expansion front. For reference, we also calculated the pristine cluster shape (dashed line, density plots) with the
initial radius rini and the corresponding diffraction pattern (dashed lines, right panels). The comparison with the expected diffraction from the intact clusters
outlines the rapidly increasing damage due to the expansion dynamics. Note that the electron density distribution scale is normalized to the pristine
cluster size.

The underlying concept of X-ray imaging is that X-rays scatter


off the electrons bound to atoms inside the sample and form a diffraction pattern. These diffraction patterns can then be used to
reconstruct the shape of the object. Intuitively, one would rst
expect that the explosion of the superheated cluster leads to a narrower ring spacing in the diffraction pattern, akin to an increasing
aperture size leading to ner line spacing in optical diffraction.
Second, one might also expect that the high fraction of electrons
delocalized from the Xe atoms during the NIR laser heating
process would result in a strongly decreased coherent scattering
response from the clusters during or immediately after the NIR pulse.
Neither aspect is observed in the data in this form. The position
of the rst minimum, reecting the particle size, does not vary

signicantly for sub-picosecond delays. The overall coherent scattering photon signal, mainly concentrated in the zero-order diffraction
peak, does not change signicantly within the rst 500 fs. This
nding suggests that hot electrons bound to the cluster Coulomb
potential continue to participate in the elastic scattering process.
On longer timescales of 1 ps and beyond, a statement about the
hot electrons contribution becomes more difcult because all
structure in the scattering pattern has faded into the uorescence
background. However, the overall time-dependent data indicate
that the diffraction response shifts from electrons bound to atoms
to electrons conned inside the cluster Coulomb potential well.
The changes in the diffraction pattern must therefore be correlated to the ultrafast structural changes in the overall nanoparticle

NATURE PHOTONICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephotonics

2016 Macmillan Publishers Limited. All rights reserved

LETTERS

NATURE PHOTONICS

electron density prole. The symmetry of the detected diffraction


pattern mirrors the spherical shape of the Xe clusters. The X-ray diffraction pattern for spherical objects can be calculated analytically in
the limit of small scattering angles with the Guinier approximation19, resulting in the wellknown Legendre polynomials. Using
the Guinier approximation (see Methods), the density prole of
the particles can also be deduced from the diffraction patterns in
a converging tting routine (left column in Fig. 3). The Guinier
t of the diffraction radial plot conrms that the pristine particle
can be approximated as a solid, uniform sphere within the achieved
resolution (Fig. 3a). For the superheated nanoparticles, a coreshell
structure emerges from the tted data. With increasing time delays
after the pump pulse, the solid-density core becomes smaller, and a
larger surface shell expands rapidly. We note that the obtained size
of the particle core is very sensitive to the tting parameters,
describing the data down to the nanometre regime below the
spatial resolution of the measurement of 8 nm (see Methods).
The data show that the particle expansion starts with the outermost surface and evolves towards the inner particle core. After 100 fs,
more than 30% of the atoms overall are in the expanding surface
shell, and this number increases to 60% after 500 fs. The observed
surface shell expansion compares well to theoretical predictions for
a nite plasma expansion into vacuum2023, so imaging of nanoplasma expansion dynamics allows the direct determination of critical
plasma parameters. The observed plasma expansion suggests a plasma
speed of sound of 1 104 m s1, in good agreement with theoretical
predictions20,21. Assuming an average charge state of Xe20+ (observed
in ion time-of-ight measurements), one can estimate an electron
temperature of 200 eV inside the super-heated nanoplasma, consistent
with full plasma simulations11,24. On the picosecond timescale, the
solid core dilutes, as evidenced by the strong decay in the diffraction
pattern at scattering angles greater than 1.4.
A detailed knowledge of the sample density prole also allows the
size of the undamaged particle to be recovered under the assumption
that the number of scattering electrons remains constant on the
subpicosecond timescale. In the right column of Fig. 3. the calculated
scattering proles of the undamaged, recovered particle (dashed line)
and the tted coreshell structure (solid line) are compared to the data
(histogram). In a direct comparison of the simulated diffraction patterns from pristine clusters and measured data, the distinct changes
due to particle expansion become very clear. The higher orders disappear rapidly, but the peaks also broaden towards higher scattering
angles, as best observed for the 500 fs data point. Moreover, the
minima of the pumped particle are located at larger scattering
angles than the minima of the simulated undamaged particles, as
the small-angle information is dominated by the solid density core.
This leads to the counterintuitive situation that the larger expanding
particles appear smaller than the unheated particles. This ngerprint
of nanoplasma formation can be used to identify potential sample
damage in single-particle imaging with X-ray laser pulses25.
In summary, using time-resolved single-shot imaging of individual gas-phase samples with femtosecond X-ray pulses, we have
imaged the expansion of non-equilibrium nanoplasma into vacuum
with femtosecond temporal resolution and nanometre spatial resolution. With improved optical and X-ray pulse synchronization, the
time resolution of our experimental approach can be pushed
towards sub-10 fs (refs 2628). Our method could be applied in the
hard X-ray regime to follow structural changes with ngstrm resolution on the timescale of electron motion, for example, conformational changes in aerosols29, shock fronts in dense plasmas12 and
plasmon oscillations13. It overcomes the complications and limitations from laser focal volume and ensemble averaging, as well as
energy dissipation to supports. The results pave the way for unprecedented observations of ultrafast dynamics in matter under extreme
conditions, including inertial connement fusion, warm-dense
matter, nanophotonics and non-equilibrium phase transitions.
4

DOI: 10.1038/NPHOTON.2015.264

Methods
Methods and any associated references are available in the online
version of the paper.
Received 19 October 2015; accepted 30 November 2015;
published online 25 January 2016

References
1. Miller, R. D. Femtosecond crystallography with ultrabright electrons and X-rays:
capturing chemistry in action. Science 343, 11081116 (2014).
2. Fletcher, L. et al. Ultrabright X-ray laser scattering for dynamic warm dense
matter physics. Nature Photon. 9, 274279 (2015).
3. Rousse, A. et al. Non-thermal melting in semiconductors measured at
femtosecond resolution. Nature 410, 6568 (2001).
4. Ditmire, T. et al. Nuclear fusion from explosions of femtosecond laser-heated
deuterium clusters. Nature 398, 489492 (1999).
5. Bostedt, C. et al. Ultrafast X-ray scattering from single free nanoparticles as a
probe for transient states of matter. Phys. Rev. Lett. 108, 093401 (2012).
6. Barty, A. et al. Ultrafast single-shot diffraction imaging of nanoscale dynamics.
Nature Photon. 2, 415419 (2008).
7. Clark, J. et al. Ultrafast three-dimensional imaging of lattice dynamics in
individual gold nanocrystals. Science 341, 5659 (2013).
8. Milathianaki, D. et al. Femtosecond visualization of lattice dynamics in shockcompressed matter. Science 342, 220223 (2013).
9. Arnlund, D. et al. Visualizing a protein quake with time-resolved X-ray
scattering at a free-electron laser. Nature Methods 11, 923926 (2014).
10. Chapman, H. N. et al. Femtosecond time-delay X-ray holography. Nature 448,
676679 (2007).
11. Gorkhover, T. et al. Nanoplasma dynamics of single large xenon clusters
irradiated with superintense X-ray pulses from the Linac Coherent Light Source
free-electron laser. Phys. Rev. Lett. 108, 245005 (2012).
12. Hickstein, D. D. et al. Observation and control of shock waves in individual
nanoplasmas. Phys. Rev. Lett. 112, 115004 (2014).
13. Zherebtsov, S. et al. Controlled near-eld enhanced electron acceleration from
dielectric nanospheres with intense few-cycle laser elds. Nature Phys. 7,
656662 (2011).
14. Gomez, L. F. et al. Shapes and vorticities of superuid helium nanodroplets.
Science 345, 906909 (2014).
15. Sellberg, J. A. et al. Ultrafast X-ray probing of water structure below the
homogeneous ice nucleation temperature. Nature 510, 381384 (2014).
16. Emma, P. et al. First lasing and operation of an ngstrom-wavelength freeelectron laser. Nature Photon. 4, 641647 (2010).
17. Strder, L. et al. Large-format, high-speed, X-ray pnCCDs combined with
electron and ion imaging spectrometers in a multipurpose chamber for
experiments at 4th generation light sources. Nucleic Instrum. Methods A 614,
483 (2010).
18. Schorb, S. et al. X-rayoptical cross-correlator for gas-phase experiments at the
Linac Coherent Light Source free-electron laser. Appl. Phys. Lett. 100,
121107 (2012).
19. Guinier, A. & Fournet, G. Small-Angle Scattering of X-Rays (Structure of Matter
series, Wiley, 1955).
20. Mora, P. Collisionless expansion of a Gaussian plasma into a vacuum. Phys.
Plasmas 12, 112102 (2005).
21. Medvedev, Y. V. Expansion of a nite plasma into a vacuum. Plasma Phys.
Control. Fusion 47, 1031 (2005).
22. Hau-Riege, S. P. & Chapman, H. N. Modeling of the damage dynamics of
nanospheres exposed to X-ray free-electron-laser radiation. Phys. Rev. E 77,
041902 (2008).
23. Peltz, C., Varin, C., Brabec, T. & Fennel, T. Time-resolved X-ray imaging of
anisotropic nanoplasma expansion. Phys. Rev. Lett. 113, 133401 (2014).
24. Chung, H.-K., Chen, M., Morgan, W., Ralchenko, Y. & Lee, R. FLYCHK
generalized population kinetics and spectral model for rapid spectroscopic
analysis for all elements. High Ener. Dens. Phys. 1, 312 (2005).
25. Aquila, A. et al. The LINAC coherent light source single particle imaging road
map. Struct. Dynam. 2, 041701 (2015).
26. Harmand, M. et al. Achieving few-femtosecond time-sorting at hard X-ray freeelectron lasers. Nature Photon. 7, 215218 (2013).
27. Hartmann, N. et al. Sub-femtosecond precision measurement of relative X-ray
arrival time for free-electron lasers. Nature Photon. 8, 706709 (2014).
28. Schulz, S. et al. Femtosecond all-optical synchronization of an X-ray freeelectron laser. Nature Commun. 6, 5938 (2015).
29. Loh, N. D. et al. Fractal morphology, imaging and mass spectrometry of single
aerosol particles in ight. Nature 486, 513517 (2012).

Acknowledgements
T.G. acknowledges a Peter Ewald fellowship from the Volkswagen Foundation. Parts of this
research were carried out at the Linac Coherent Light Source (LCLS) at the SLAC National
Accelerator Laboratory. LCLS is an Ofce of Science User Facility operated for the US

NATURE PHOTONICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephotonics

2016 Macmillan Publishers Limited. All rights reserved

NATURE PHOTONICS

LETTERS

DOI: 10.1038/NPHOTON.2015.264

Department of Energy Ofce of Science by Stanford University. This work is supported by


the US Department of Energy, Ofce of Science, Ofce of Basic Energy Sciences, Division of
Chemical, Geological and Biological Sciences (contract nos. DE-AC02-06CH11357 (C.B.),
DE-AC02-76SF00515 (C.B. and R.C.) and DE-FG02-86ER13491 (D.Ro. and A.R.)). T.M.
acknowledges nancial support from BMBF projects 05K10KT2 and 05K13KT2 as well as
DFG BO3169/2-2. P.J. acknowledges support from the Swedish Research Council and the
Swedish Foundation for Strategic Research. M.M. acknowledges support from the National
Science Foundation (award no. 1231306). The authors acknowledge the Max Planck
Society for funding the development and operation of the CAMP instrument within the
ASG at CFEL. The authors thank T. Fennel for discussions, and M. Swiggers, J.-C. Castagna
and all LCLS staff for their help in setting up and performing the experiments.

Author contributions
C.B. conceived the idea and coordinated the project together with T.G., S.Sc., R.C., D.Ro.,
A.Ru. and T.M. The experimental setup was designed by all authors. The laser system was

prepared by R.C., T.G., L.H., P.J., J.K., A.Ro. and B.W. The pnCCD detectors were
developed and operated by A.H., R.H., G.H., P.H., N.K., C.R., G.W., H.S. and L.S. The
experiment was performed by T.G., S.Ss., R.C., M.A., L.F., A.A., J.D.B., B.E., R.H., P.H.,
N.K., K.-U.K., C.R., B.R., J.S., G.W., J.U., D.Ro., A.Ru., T.M. and C.B. The CASS online
and ofine data analysis software was developed by L.F. The data were analysed by T.G. The
results were interpreted by T.G., T.M. and C.B. The manuscript was written by T.G. and
C.B. with contributions from T.M. and I.S. as well as input from all authors.

Additional information
Supplementary information is available in the online version of the paper. Reprints and
permissions information is available online at www.nature.com/reprints. Correspondence and
requests for materials should be addressed to T.G. and C.B.

Competing nancial interests

The authors declare no competing nancial interests.

Linac Coherent Light Source, SLAC National Accelerator Laboratory, Stanford, California 94309, USA. 2 Institut fr Optik und Atomare Physik, Technische
Universitt Berlin, Hardenbergstr. 36, Berlin 10623, Germany. 3 PULSE Institute and SLAC National Accelerator Laboratory, 2575 Sand Hill Road, Menlo Park,
California 94025, USA. 4 Max Planck Advanced Study Group, Center for Free-Electron Laser Science, Notkestrasse 85, 22607 Hamburg, Germany.
5
Max-Planck-Institut fr medizinische Forschung, Jahnstr. 29, Heidelberg 69120, Germany. 6 Center for Free-Electron-Laser Science (CFEL), DESY,
Notkestrasse 85, Hamburg 22607, Germany. 7 European XFEL GmbH, Albert-Einstein-Ring 19, Hamburg 22761, Germany. 8 Synchrotron SOLEIL, LOrme des
Merisiers Saint-Aubin, BP 48 91192, GIF-sur-YVETTE CEDEX, France. 9 Max-Planck-Institut fr Kernphysik, Saupfercheckweg 1, Heidelberg 69117, Germany.
10
Photon Science DESY, Notkestrasse 85, 22607 Hamburg, Germany. 11 Department of Chemistry, Aarhus University, Aarhus C 8000, Denmark.
12
PNSensor GmbH, Otto-Hahn-Ring 6, Mnchen 81739, Germany. 13 Max-Planck-Institut fr extraterrestrische Physik, Giessenbachstrasse, Garching 85741,
Germany. 14 Department of Physics, Lund University, PO Box 118, Lund 22100, Sweden. 15 National Science Foundation BioXFEL Science and Technology
Center, 700 Ellicott Street Buffalo, New York 14203, USA. 16 Max-Born-Institut, Max-Born-Strae, Berlin 12489, Germany. 17 FOM-Institute AMOLF, 1098 XG,
Amsterdam, The Netherlands. 18 Physikalisch-Technische Bundesanstalt (PTB), Bundesallee 100, 38116 Braunschweig, Germany. 19 Department of Physics and
Center for Ultrafast Imaging, University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany. 20 Max-Planck-Institut Halbleiterlabor, Otto-HahnRing 6, Mnchen 81739, Germany. 21 Universitt Siegen, Emmy-Noether Campus, Walter Flex Str. 3, 57072 Siegen, Germany. 22 J.R. Macdonald Laboratory,
Kansas State University, Manhattan, Kansas 66506, USA. 23 Argonne National Laboratory, 9700 South Cass Avenue, Lemont, Illinois 60439, USA.
24
Department of Physics and Astronomy, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208, USA. * e-mail: taisgork@slac.stanford.edu;
cbostedt@anl.gov

NATURE PHOTONICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephotonics

2016 Macmillan Publishers Limited. All rights reserved

LETTERS

NATURE PHOTONICS

Methods

Experimental setup. The experiment was performed at the AMO end station at
LCLS using the CFEL-ASG Multi-Purpose (CAMP) instrument17. FEL pulses with
up to 2 mJ pulse energy were tuned to a wavelength of 0.8 nm and pulse length of
50 fs (calculated as two-thirds of the electron bunch length). The FEL beam was
focused by Kirkpatrick Baez mirrors to a spot size of 9 m2 (full-width at
half-maximum, FWHM) with intensities up to 1 1017 W cm2 (ref. 18).
NIR pulses were generated by a chirped-pulse regenerative amplied Ti:sapphire
system producing 800 nm wavelength, 70 fs pulse length and up to 1.5 mJ pulse
energy. The laser pulses were focused by a f = 1,000 mm lens to a spot size of
1,600 m2 (FWHM). The maximum average NIR intensity yielded up to
2 1015 W cm2 (estimated from ionization yield measurements performed
on Ar atoms30).
The infrared laser beam was aligned collinearly to the FEL beam, and both foci
were spatially overlapped with the aid of a microscope and focusing of attenuated
beams on a silicon nitride screen. The Xe cluster jet was intersected perpendicular to
the two laser beams inside the foci. The cluster jet was produced by supersonic
expansion of Xe gas into vacuum through a conical nozzle with an orice of 200 m
and an opening angle of 4 half-angle. The nozzle was mounted on a pulsed solenoid
valve synchronized to the FEL repetition rate to reduce the gas load into the vacuum
system. During the supersonic expansion, the thermal energy of the gas is converted
into a directed ow, leading to supersaturation and subsequent condensation of the
Xe gas. The valve was operated at room temperature with a backing pressure of
10 bar Xe. The resulting mean cluster radius was 21 6 nm, agreeing well with the
established scaling laws and a previous detailed characterization of the cluster
source31. Subsequent to the expansion, the cluster jet was shaped through two
3 mm skimmers. An adjustable piezo-driven skimmer immediately before the
interaction regions allowed the number of particles inside the FEL focus to
be controlled5,11.
The NIR and the FEL pulses were synchronized using conventional feedback
techniques correlating the arrival time relative to the radiofrequency driving the FEL
accelerator. The arrival time of both laser pulses at the interaction region was rst
determined using a photodiode up to 50 ps relative time difference. Finer temporal
correlation was achieved using an ion time-of-ight detector placed perpendicular to
the FEL beam and the cluster jet. The relative ion yields from N2 molecule Coulomb
explosions were used to determine the pulse arrival times. The shot-to-shot
temporal resolution of the experiment was limited by the systematic error
introduced by the relative jitter between the FEL and infrared pulses, estimated to be
100 fs r.m.s. (ref. 18).
The diffraction images were recorded by a single-photon sensitive pnCCD
detector17 with an active area of 76.8 mm 37 mm placed 220 mm downstream
from the interaction region and around 5 mm higher than the FEL beam in the
vertical direction. An infrared-absorbing, 2 m thin C22H10N2O5 foil coated with Al
was placed in front of the pnCCD to protect the detector from stray light from the
NIR laser. The detector position was optimized for maximal scattering angle
coverage, including the lowest angles with the highest X-ray scattering yield.
Although the detector had a centre passage hole for the FEL laser beam, the high
divergence of the collinear infrared laser beam was the major factor limiting the
geometry of the setup.
Selection of diffraction images. The diffraction patterns displayed in Figs 2 and 3
represent shots from clusters illuminated in the centre of the FEL where the highest
FEL intensities are present. The diffraction images were extracted from around
3 105 shots using selection criteria that were independent of the coherent
diffraction signal (see Supplementary Information). Thus, clusters exposed to high
FEL intensities in the centre of the FEL focus could be identied, even when the
diffraction patterns had faded, as for example in the images recorded with arrival
times greater than 1 ps. The selection criteria and statistical analysis are described in
detail in the Supplementary Information.
Pre-heated cluster melting is observed throughout the entire data set,
independent of cluster size (21 6 nm), as described in the Supplementary
Information. To illustrate the cluster expansion process, we focused on similarly
sized clusters and similar coherent diffraction intensities in the zero-order
diffraction ring (containing more than 90% of the overall diffracted photons). We
selected smaller clusters with radii of 18 2 nm for the discussion, because their
zero-order diffraction peak extended to greater scattering angles, which were better
covered by the detector.

DOI: 10.1038/NPHOTON.2015.264

The nanoparticle size was estimated from the ring pattern spacing of the
rst fully detected ring. Note that we could assume a uniform NIR intensity of
2 1015 W cm2 for all detected cluster images as the focus size ratio between the
NIR and the imaging FEL was greater than 160.
Correction of diffraction images. The coherent diffraction images were corrected
for the uorescence photon background on a shot-to-shot basis and for the average
beamline stray light.
In bright images from the best hit clusters, there was signicant signal from
uorescence photons. Using the energy-dispersive single-photon-sensitive pnCCDs,
the uorescence photons could be identied and subtracted from each image during
post-processing. This correction could be applied in domains with low incoming
photon ux (less than one photon per nine pixels). Additionally, the diffraction
patterns were corrected for parasitic stray light from the beamline. The stable
beamline stray light was recorded in the absence of clusters, averaged over 1,000
shots and subtracted from each cluster image. Both corrections signicantly
improved the overall signal-to-noise ratio of the diffraction images.
Simulation. The shape and electron density distribution of superheated clusters
were analytically calculated from the diffraction images, which are projections of the
electron density e in k (momentum) space. The spherical symmetry of the
nanosamples reduced the complexity of the problem and allowed for integration
over density uctuations in dependence on r (distance to the cluster centre). Due to
the low imaginary refractive index of solid Xe at 1,500 eV, the effects from
absorption could be neglected.
The diffraction images were simulated using the Guinier approximation for
small-angle X-ray scattering. The expected intensity variations Isc(q) inside the
diffraction patterns depend on the scattering angle (or the spatial frequency q in k
space) and were calculated according to19

Isc (q) = Ie (q)

e (r)

sin(qr)
4r2 dr
qr

2
(1)

and
q() =

4 sin(/2)

(2)

where Ie is the intensity scattered by a single electron and is the photon wavelength.
We tted a solid coreGaussian expanding shell for the density distribution e(r) of
pre-heated clusters21:

e (r) =

d 2 r2

0 e2(a(R)d)
0

if x > |d|
if x |d|

(3)

where d is the radius of the solid core and a is a normalization factor that depends on
the initial cluster radius R.
Assuming that the number of scattering electrons N remains nearly constant
during cluster expansion, the value of a can be used to reconstruct the pristine cluster
radius a posteriori (as outlined in Fig. 3):

4
4r2 e (r, a)dr
N = R3 =
3
0
The major parameter in the tting routine is the dimension of the intact core as
the sharp edge produces characteristic ring patterns. Note that the shape and the
electron density of the expanding shell might deviate from the pure Gaussian
density distribution.

References
30. Bryan, W. et al. Atomic excitation during recollision-free ultrafast multi-electron
tunnel ionization. Nature Phys. 2, 379383 (2006).
31. Rupp, D. et al. Generation and structure of extremely large clusters in pulsed jets.
J. Chem. Phys. 141, 044306 (2014).

NATURE PHOTONICS | www.nature.com/naturephotonics

2016 Macmillan Publishers Limited. All rights reserved

Вам также может понравиться