week ending

PRL 109, 123001 (2012) PHYSICAL REVIEW LETTERS 21 SEPTEMBER 2012

Molecular-Frame 3D Photoelectron Momentum Distributions by Tomographic Reconstruction

Jochen Maurer,1 Darko Dimitrovski,2 Lauge Christensen,3 Lars Bojer Madsen,2,* and Henrik Stapelfeldt1,4,†
1
Department of Chemistry, Aarhus University, 8000 Aarhus C, Denmark
2
Department of Physics and Astronomy, Lundbeck Foundation Theoretical Center for Quantum Systems Research, Aarhus University,
8000 Aarhus C, Denmark
3
Department of Physics and Astronomy, Aarhus University, 8000 Aarhus C, Denmark
4
Interdisciplinary Nanoscience Center (iNANO), Aarhus University, 8000 Aarhus C, Denmark
(Received 11 June 2012; published 18 September 2012)
Naphthalene molecules are fixed in space by a laser field and rotated, in 2
steps, over 180
. For each
orientation, they are ionized by an intense, circularly polarized femtosecond laser pulse, and the 2D
projection of the photoelectron momentum distribution is recorded. The molecular-frame 3D momentum
distribution is obtained by tomographic reconstruction from all 90 projections. It reveals an anisotropic
electron distribution, angularly shifted in the polarization plane, that is not accessible by the 2D
momentum images. Our theoretical analysis shows that the magnitude of the angular shift is very
sensitive to the exact form of the laser-modified molecular potential.

DOI: 10.1103/PhysRevLett.109.123001 PACS numbers: 33.60.+q, 33.80.Eh, 33.80.Rv, 42.50.Hz

An important tool in medical imaging is x-ray computed by measuring the spatial orientation of each molecule from
tomography, enabling three-dimensional (3D) imaging of which electrons are being recorded. This is possible if the
the inside of the human body. The basic idea of computed ionization event leads to fragmentation in a way that the
tomography is to record two-dimensional (2D) x-ray im- emission direction of recoiling ions, detected in coinci-
ages for a large number of irradiation angles of an object dence with the photoelectrons, identifies the molecular
and retrieve its 3D spatial distribution by tomographic frame at the instant of ionization. Photoelectron-ion coin-
reconstruction from all 2D images. Recently, the principle cidence techniques have enabled a range of seminal studies
of tomography made its entry into photoelectron imaging on small, typically linear or symmetric top molecules
with the aim of determining the 3D photoelectron momen- where only a single molecular axis needs to be determined
tum distribution (PMD) stemming from ionization of [7–11]. For more complex molecules, the unlikely exis-
atoms or molecules by laser pulses. For atoms, it was tence of a fragmentation pattern that uniquely identifies the
demonstrated that the 3D PMD could be retrieved from a molecular orientation makes coincidence methods a non-
sufficiently large set of projections on a 2D imaging viable route to retrieve MF PMDs [12]. Furthermore, often
detector [1,2]. The different projections were obtained by ionization does not lead to fragmentation which, obviously,
rotating the polarization state of the ionizing laser pulse excludes the application of coincidence methods.
with respect to the detector because the emission direction An alternative and more general approach to access the
of the photoelectrons from spherically symmetric atoms molecular frame is to fix the molecular orientation prior to
is uniquely determined by the polarization state. the ionization event. Here, we show that the ability, based
Tomographic reconstruction was extended to randomly on moderately intense, nonresonant laser pulses [15], to
oriented molecules [3], but this only gives the 3D PMD confine the spatial orientation of molecules and controlla-
in the laboratory frame. To extract maximum information bly rotate them combined with tomographic reconstruction
from PMDs, it is necessary that they are measured with opens a new approach to measurements of MF 3D PMDs.
respect to the molecular frame (MF). Exploiting tomogra- We use this new ability to determine the MF 3D PMDs
phy to retrieve a MF 3D PMD requires, therefore, that the created by strong field ionization with intense circularly
molecular-frame rather than the laboratory-frame PMD is polarized laser pulses and show that it is a sensitive probe
rotated with respect to the detector. As demonstrated in the of the asymptotic laser-modified molecular potential.
present Letter, this can be obtained by sharply aligning The principle of the experiment is illustrated in Fig. 1
molecules and rotating them over 180
. (see Ref. [16]). A cold molecular beam of naphthalene
The motivation for our studies is that the MF 3D PMD is molecules (rotational temperature 1 K) enters a velocity
the most differential quantity from which more commonly map imaging (VMI) spectrometer where it is crossed at
used observables, like the kinetic energy distribution of the 90
by two pulsed laser beams. The first pulse (YAG pulse:
photoelectrons [4,5], can be derived. One established ex-
¼ 1064 nm, FWHM ¼ 10 ns, I0 ¼ 8  1011 W=cm2 ) is
perimental approach is photoelectron-ion coincidence elliptically polarized, shown by the red ellipse in Fig. 1,
techniques. The detector type, typically employed, pro- with an intensity ratio of 3:1 between the major and the
vides 3D PMDs [6], and the molecular frame is accessed minor polarization axes. It adiabatically 3D aligns the long

0031-9007=12=109(12)=123001(5) 123001-1 Ó 2012 American Physical Society

 week ending
PRL 109, 123001 (2012) PHYSICAL REVIEW LETTERS 21 SEPTEMBER 2012

interpretation of the images was given recently [18,19],

notably that the suppression results from the nodal plane
that coincides with the molecular plane, in the orbitals
from where electrons are detached. This can be understood
qualitatively by noting that ionization with a field in the
nodal plane is forbidden because there is no electronic
population in the nodal plane. For naphthalene, the highest
occupied orbital has a structure similar to the atomic fxyz
orbital, possessing three nodal planes.
The tomographic reconstruction was performed for each
2D slice along the laser beam propagation direction by the
filtered backprojection method for parallel projections, a
fast, efficient, and widely used method for tomographic
FIG. 1 (color online). (a) Schematic of the experimental setup reconstruction [20,21]. During the reconstruction, the fil-
showing a naphthalene molecule aligned by the elliptically
tering step was performed with a Hann-type filter in the
polarized YAG pulse (red ellipse) and ionized by the circularly
polarized probe pulse (red spiral). The VMI spectrometer frequency space with a cutoff frequency of half of the full
projects the photoelectron momentum distribution onto the de- frequency range. Furthermore, the reconstructed data were
tector, and the 2D electron images are recorded for 90 align- smoothened with a spatial Gaussian filter after the recon-
ments of the molecule over a 180
interval obtained by rotating struction. Both methods are commonly used in the case of
the YAG polarization ellipse in steps of 2
around the x axis. reconstruction of low statistics data [22–24]. The close
Examples of electron images for the major axis of the ellipse similarity of the unprocessed reconstructed data with the
parallel to (vertical) and perpendicular to (horizontal) the detec- symmetrized and filtered input data confirmed that no
tor are shown in (b) and (c), respectively. information was lost and no artifacts were introduced
during the reconstruction procedure.
molecular axis (y0 axis) along the major polarization axis Figure 2 represents the outcome of the reconstruction for
and the short molecular axis (z0 axis) along the minor a RCP probe pulse by showing an isosurface of the mo-
polarization axis (Fig. 1). The degree of alignment is not mentum distribution, i.e., points in momentum space
quantitatively characterized, but due to the high polariz- where the MF 3D PMD, Pðpx0 ; py0 ; pz0 Þ, attains a constant
ability (anisotropy) of naphthalene, the alignment is ex- value. The isosurface resembles a modulated torus with a
pected to be high and similar to that of 2,6 groove centered at px0 ¼ 0. This is the structure originating
difluoroiodobenzene [17]. The second pulse (probe pulse: from suppression of electron emission in orbital nodal
800 nm, 30 fs, 9  1013 W=cm2 ) is circularly polarized. Its planes, see also [Figs. 1(b) and 1(c)]. Second, a hollow
electric field vector is given by FðtÞ ¼ pFffiffi20 cos2 ðt=Þ  out in the center of the surface along the x0 axis is observed.
½cosð!tÞez  sinð!tÞey , with the plus sign referring to a
left-circularly polarized (LCP) and the minus sign to a
right-circularly polarized (RCP) pulse, ! corresponding
to the wavelength of 800 nm, and  the pulse duration. The
probe pulse is synchronized to the peak of the YAG pulse,
where the molecular alignment is strongest, and it singly
ionizes the naphthalene molecules. The photoelectrons are
projected onto a 2D detector, and the images are captured
by a CCD camera.
To provide the complete set of projections of the PMD,
needed for tomographic reconstruction, electron images
were recorded for 90 equally spaced alignments of the
molecules in a 180
interval obtained by rotating the
polarization ellipse over 180
in steps of 2
. This step
size is sufficiently fine to avoid any significant artifacts
FIG. 2 (color online). Tomographic reconstruction of the mo-
in the reconstruction. The circular polarization state of the
lecular frame 3D photoelectron momentum distribution, repre-
probe pulse was kept fixed. Examples of two electron sented by an isosurface, when naphthalene molecules are ionized
images are depicted in Fig. 1, one with the long molecular by a RCP probe pulse. The momenta at the isosurface value of
axis parallel (b) and one with the axis perpendicular (c) to 0.035 are given in atomic units (a.u.). The result obtained with a
the detector. The images show that the electrons are ejected LCP probe pulse is almost identical except that the angular shift
in a stripe parallel to the polarization plane of the probe of the isosurface in the field polarization plane y0 and z0 is in the
pulse with a pronounced suppression for px0 ¼ 0. The opposite direction.

123001-2
 week ending
PRL 109, 123001 (2012) PHYSICAL REVIEW LETTERS 21 SEPTEMBER 2012

Although this structure also manifests itself in the 2D exhibited by the circular polarization state of the ionizing
images as a formation of a waist along the horizontal center laser in this plane. It is seen that the angular shift, defined
line, its true magnitude is only revealed in the 3D distri- as the counterclockwise (viewed from the side of the
butions. Third, the peak emission direction does not coin- incoming pulse in the -pz0 -py0 plane) offset of the PMDs
cide with either the long (y0 ) or the short (z0 ) molecular from the short molecular axis, is largest for the px0 interval
axis. This characteristic cannot be seen in the 2D electron containing the smallest px0 values and that it gradually
images. In the following, we focus on this angular shift of decreases as the values in the px0 interval increase [31]. It
the MF 3D PMD and analyze its origin and magnitude. is also seen that the maximum in the radial distribution
The angular shift of PMDs, associated with ionization shifts to larger values as the px0 interval increases, reflect-
by intense laser pulses, has been addressed in several ing the fact that high momenta electrons are produced from
previous works [25–29]. The main emphasis has been on focal volume regions where the probe pulse intensity is
ionization of atoms with an elliptically polarized laser high. We identify the maximum in the radial momentum
pulse [25–27]. Here, the electron is released when the laser distribution for each px0 interval—either directly from the
field is strongest, i.e., pointing along the major polarization data points or from an asymmetric Gaussian model fit [32]
axis, and the shift results from the subsequent distortion of and assign an effective probe pulse intensity using the
the laser driven electron trajectory by the field from the calculated dependence of the maximum radial momentum
cation. In the current work, the anisotropy in the electron as a function of probe pulse intensity [34]. Hereby we are
emission is not due to any anisotropy in the polarization able to plot the angular shift versus the effective probe
state of the ionizing field (since it is circularly polarized), intensity. The results (the numerical value of the angular
but rather to the dependence of the ionization yield on the shifts) for both helicities of the probe pulse are shown in
orientation of the field vector with respect to the mole- Fig. 4.
cule—a phenomenon that is general for most anisotropic To qualitatively understand the origin of the angular
molecules [30]. shift, we note that during an optical cycle the electric field
The analysis is based on determining the PMD in the vector of the probe pulse, FðtÞ, completes a full rotation in
polarization plane, i.e., the y; z plane which, for (perfectly) the plane of the naphthalene molecules. For naphthalene,
aligned molecules coincides with the molecular y0 , in the tunneling limit, ionization occurs most efficiently for
z0 -plane. In particular, we obtain the PMD in and parallel the direction of the field where the ionization potential,
to the molecular plane for the five px0 intervals, defined in Stark shifted by the polarizability interaction with FðtÞ, is
the caption to Fig. 3, by projecting the 3D PMD pertaining lowest [18,35–37]. This occurs when FðtÞ points along the
to these intervals onto the polarization plane. This is an long molecular axis where the polarizability difference
example of observables that can only be accessed due to between the neutral molecule and its cation is largest.
the availability of the 3D PMD. The results for a RCP Upon release, the electron is influenced by the remainder
probe pulse, displayed in Fig. 3, reveal anisotropic PMDs, of FðtÞ which would cause the electron to be emitted 90

each with a peak emission direction in between the long ( 90
) away from the long molecular axis, i.e., along the
and the short molecular axis despite the circular symmetry
50
LCP (Asymmetric Gauss)
RCP (Asymmetric Gauss)
40 LCP (Direct Maximum)
RCP (Direct Maximum)
|angle| [degrees]

30

20

10

0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
14 2
Intensity [10 W/cm ]

FIG. 3 (color online). Projections of the 3D momentum FIG. 4. Numerical values of the experimental (symbols) and
distribution onto the polarization plane from 3D aligned naph- theoretical (curves) angular shifts of the electron momentum
thalene molecules ionized with a RCP probe pulse. The five distribution in the polarization plane as a function of probe pulse
different projections correspond to five different px0 intervals: intensity for naphthalene. Full curve: Perfectly aligned mole-
(a) 0.117–0.194 a.u., (b) 0.194–0.272 a.u., (c) 0.272–0.350 a.u., cules, full ionic potential; Dashed curve: Perfectly aligned, only
(d) 0.350–0.427 a.u., (e) 0.427–0.583 a.u. (f) The geometry of the Coulomb potential. Dotted curve: Nonperfectly aligned mole-
projections with respect to the molecule and the probe pulse cules (see text), full ionic potential. The experimental points are
polarization. The probe pulse propagates into the paper. plotted for both LCP and RCP probe pulses. See text for details.

123001-3
 week ending
PRL 109, 123001 (2012) PHYSICAL REVIEW LETTERS 21 SEPTEMBER 2012

short molecular axis for a LCP (RCP) probe pulse. The probe pulse may provide a method for time-resolved prob-
angular shift results from the fact that the electron is, in ing of changes in the molecular potential induced by a
addition to FðtÞ, also influenced by the Coulomb field of pump pulse.
the molecular cation and the electric field from the dipole Our method to retrieve the MF 3D PMD will be gen-
moment induced in the cation by the probe pulse. erally applicable to a broad range of nonpolar molecules
Theoretically, we analyze the angular shift of the PMD independent of the ionization process leading to fragmen-
using a semiclassical two-step model, where the first step is tation or not. The resolution in the reconstructed PMDs can
tunneling and the second step involves propagation of be increased by employing methods to further sharpening
classical trajectories from the tunnel exit point [38,39]. the 3D molecular alignment [19]. An interesting challenge
The probability of tunneling depends on the Stark-shifted is the extension to polar molecules where a uniform ori-
ionization potential Ip ðFÞ ¼ Ip ð0Þ þ 12 FT ðN   I ÞF, entation (head-versus-tail asymmetry) is required for the
where F is the instantaneous electric field vector and N 3D tomographic reconstruction for all alignment positions
and I are the static polarizability tensors of the neutral of the molecules. This may be achieved by adjusting the
and the cation, respectively, with components taken from static electric field in the VMI spectrometer such that the
Ref. [40]. The naphthalene potential is modeled as component along the permanent dipole moment axis re-
mains constant as the molecules are rotated.
Z ðI FÞ  r The work was supported by the Danish Council for
VðrÞ ¼   ; (1)
r r3 Independent Research (Natural Sciences), The Lundbeck
Foundation, the Carlsberg Foundation, the European
for not too small distances [36]. Marie Curie Initial Training Network Grant No. CA-
The angular shift of the PMD (Fig. 4) is calculated for ITN-214962-FASTQUAST, and ERC-StG (Project
both perfectly and imperfectly aligned molecules and by No. 277767-TDMET). We acknowledge helpful discus-
taking into account only the most probable classical tra- sions with Matthias Wollenhaupt and Thomas Baumert.
jectory—the trajectory launched at the peak of the pulse
when F points along the long molecular axis. The angular
shift calculated as a function of the probe pulse intensity is
shown in Fig. 4 together with the experimental results. The
*bojer@phys.au.dk
experimental data are not compatible with the calculation †
henriks@chem.au.dk
that only includes the Coulomb field from the cation [1] M. Wollenhaupt, M. Krug, J. Köhler, T. Bayer, C. Sarpe-
(dashed curve in Fig. 4). Instead they scatter around the Tudoran, and T. Baumert, Appl. Phys. B 95, 647
curve calculated with the full ionic potential included (full (2009).
curve in Fig. 4). The reason for the larger angular shift [2] C. Smeenk, L. Arissian, A. Staudte, D. M. Villeneuve, and
when only the Coulomb potential is included is the smaller P. B. Corkum, J. Phys. B 42, 185402 (2009).
tunnel exit point as compared to the exit point when the full [3] P. Hockett, M. Staniforth, and K. L. Reid, Mol. Phys. 108,
potential of Eq. (1) is present. Moreover, the induced 1045 (2010).
dipole term ‘‘shields’’ the electron from the influence of [4] A. Stolow and J. G. Underwood, Adv. Chem. Phys. 139,
the Coulomb potential (attractive force) in the initial stages 497 (2008).
of propagation in the continuum (immediately after tun- [5] K. L. Reid, Mol. Phys. 110, 131 (2012).
[6] R. Dörner, V. Mergel, O. Jagutzki, L. Spielberger, J.
neling). The induced dipole moment, i.e., the laser modi-
Ullrich, R. Moshammer, and H. Schmidt-Böcking, Phys.
fication to the molecular potential, also explains why the Rep. 330, 95 (2000).
shift decreases when the probe intensity increases. As the [7] J. A. Davies, R. E. Continetti, D. W. Chandler, and C. C.
intensity is increased the induced dipole moment becomes Hayden, Phys. Rev. Lett. 84, 5983 (2000).
larger and so does the shielding of the Coulomb field. For [8] A. M. Rijs, M. H. M. Janssen, E. H. Chrysostom, and C. C.
completeness, we also calculated the angular shift for Hayden, Phys. Rev. Lett. 92, 123002 (2004).
molecules with an alignment distribution that simulates [9] O. Gessner et al., Science 311, 219 (2006).
the one expected in the experiment [41]. The results (dot- [10] M. S. Schöffler et al., Science 320, 920 (2008).
ted curve in Fig. 4) show angular shifts only slightly [11] J. Wu, L. P. H. Schmidt, M. Kunitski, M. Meckel, S. Voss,
smaller than those for perfectly aligned molecules and H. Sann, H. Kim, T. Jahnke, A. Czasch, and R. Dörner,
are, thus, still in good agreement with the experimental Phys. Rev. Lett. 108, 183001 (2012).
[12] We note that for the simplest asymmetric tops, i.e., bent
findings.
triatomics, two recent experiments have demonstrated MF
The results presented here show that the magnitude of PMDs using multiple-coincidence techniques—see
the angular shift of the PMD is determined by the cationic Refs. [13,14].
molecular potential that acts on the departing electron. As [13] M. Yamazaki, J. Adachi, T. Teramoto, A. Yagishita,
such, measurements of the angular shift of photoelectrons M. Stener, and P. Decleva, J. Phys. B 42, 051001 (2009).
created by ionization of molecules with an intense circu- [14] M. Yamazaki, J. Adachi, Y. Kimura, M. Stener, P. Decleva,
larly polarized (or elliptically polarized) femtosecond and A. Yagishita, J. Chem. Phys. 133, 164301 (2010).

123001-4
 week ending
PRL 109, 123001 (2012) PHYSICAL REVIEW LETTERS 21 SEPTEMBER 2012

[15] H. Stapelfeldt and T. Seideman, Rev. Mod. Phys. 75, 543 [30] T. K. Kjeldsen, C. Z. Bisgaard, L. B. Madsen, and H.
(2003). Stapelfeldt, Phys. Rev. A 71, 013418 (2005).
[16] J. L. Hansen et al., Phys. Rev. A 83, 023406 (2011). [31] Similar results are obtained with a LCP probe pulse except
[17] I. Nevo, L. Holmegaard, J. H. Nielsen, J. L. Hansen, H. that the angular shifts are now clockwise.
Stapelfeldt, F. Filsinger, G. Meijer, and J. Küpper, Phys. [32] Use of an asymmetric Gaussian model to fit the radial
Chem. Chem. Phys. 11, 9912 (2009). momentum distribution was motivated by tunneling theory
[18] L. Holmegaard et al., Nature Phys. 6, 428 (2010). for strong field ionization—see, e.g., Ref. [33].
[19] J. L. Hansen, H. Stapelfeldt, D. Dimitrovski, M. Abu- [33] N. B. Delone and V. P. Krainov, J. Opt. Soc. Am. B 8, 1207
samha, C. P. J. Martiny, and L. B. Madsen, Phys. Rev. (1991).
Lett. 106, 073001 (2011). [34] J. Maurer, Ph.D. thesis, Aarhus University, 2012.
[20] A. C. Kak and M. Slaney, Principles of Computerized [35] D. Dimitrovski, M. Abu-samha, L. B. Madsen, F.
Tomographic Imaging (IEEE Press, New York, 1988). Filsinger, G. Meijer, J. Küpper, L. Holmegaard, L.
[21] T. M. Buzug, Computed Tomography from Photon Statistics Kalhøj, J. H. Nielsen, and H. Stapelfeldt, Phys. Rev. A
to Modern Cone-Beam CT (Springer-Verlag, Berlin, 2008). 83, 023405 (2011).
[22] D. G. Politte, G. R. Hoffman, D. E. Beecher, D. C. Ficke, [36] D. Dimitrovski, C. P. J. Martiny, and L. B. Madsen, Phys.
T. J. Holmes, and M. M. Ter-Pogossian, IEEE Trans. Nucl. Rev. A 82, 053404 (2010).
Sci. 33, 428 (1986). [37] J. L. Hansen, L. Holmegaard, J. H. Nielsen, H. Stapelfeldt,
[23] F. O’Sullivan, Y. Pawitan, and D. Haynor, IEEE Trans. D. Dimitrovski, and L. B. Madsen, J. Phys. B 45, 015101
Med. Imaging 12, 653 (1993). (2012).
[24] C. X. Wang, W. E. Snyder, G. Bilbro, and P. Santago, [38] A. N. Pfeiffer, C. Cirelli, M. Smolarski, D. Dimitrovski,
Comput. Biol. Med. 28, 13 (1998). M. Abu-samha, L. B. Madsen, and U. Keller, Nature Phys.
[25] M. Bashkansky, P. H. Bucksbaum, and D. W. Schumacher, 8, 76 (2012).
Phys. Rev. Lett. 60, 2458 (1988). [39] N. I. Shvetsov-Shilovski, D. Dimitrovski, and L. B.
[26] S. P. Goreslavski, G. G. Paulus, S. V. Popruzhenko, and N. I. Madsen, Phys. Rev. A 85, 023428 (2012).
Shvetsov-Shilovski, Phys. Rev. Lett. 93, 233002 (2004). [40] S. M. Smith, A. N. Markevitch, D. A. Romanov, X. Li,
[27] P. Eckle, A. N. Pfeiffer, C. Cirelli, A. Staudte, R. Dörner, R. J. Levis, and H. B. Schlegel, J. Phys. Chem. A 108,
H. G. Muller, M. Büttiker, and U. Keller, Science 322, 11063 (2004).
1525 (2008). [41] This was done using an analytical probability distribution
[28] M. Odenweller, N. Takemoto, A. Vredenborg, K. Cole, for aligned molecules [42], in which the only parameter is
K. Pahl, J. Titze, L. P. H. Schmidt, T. Jahnke, R. Dörner, the expectation value of the cosine square of the alignment
and A. Becker, Phys. Rev. Lett. 107, 143004 (2011). angle—here, a conservative value of 0.85 is employed.
[29] J. Henkel, M. Lein, V. Engel, and I. Dreissigacker, Phys. [42] B. Friedrich and D. Herschbach, Phys. Rev. Lett. 74, 4623
Rev. A 85, 021402 (2012). (1995).

123001-5