Journal of Electron Spectroscopy and Related Phenomena 183 (2011) 114117

Contents lists available at ScienceDirect

Journal of Electron Spectroscopy and

Related Phenomena
journal homepage: www.elsevier.com/locate/elspec

Angle-resolved photoemission: From reciprocal space to real space

F.J. Himpsel
Physics Department, University of Wisconsin Madison, 5108 Chamberlin Hall, 1150 University Ave., Madison, WI 53706, United States

a r t i c l e i n f o a b s t r a c t

Article history: The prospects for converting angle-resolved photoemission data from reciprocal space to real space are
Available online 12 March 2010 evaluated. Examples are given for one- and two-dimensional systems, where wave functions have been
obtained for quantum well states conned between steps and for planar molecules adsorbed at a surface.
Keywords: The extension to three-dimensional systems via variable photon energies is discussed. The uncertainty
Photoemission relation between position and momentum makes such a capability complementary to imaging in real
Angle-resolved
space by scanning tunneling microscopy. Angle-resolved photoemission provides access to the internal
Wave function
structure of an electronic orbital and penetrates below the surface.
Fourier transform
Imaging 2010 Elsevier B.V. All rights reserved.

It was recognized as early as 1970 by Siegbahn and coworkers
[1,2] that photoelectrons have a distinct momentum distribution.
Early results are shown in Fig. 1 (from [1]), where the angular
dependence of the photoemission intensity from a NaCl single crys-
tal is plotted versus the polar angle . The Na 1s and Na 2s core
levels, as well as the Cl 3p valence band, exhibit distinct modula-
tions of up to 30% which seem to be tied to the crystal lattice in
real space. It was quickly found [3] that this phenomenon was not
restricted to ionic insulators but rather wide spread. Such obser-
vations eventually led to the concepts of photoelectron diffraction
[46] and photoelectron holography [7]. The angular distribution of
elastic photoelectrons from a core level is explained by scattering
of the outgoing photoelectrons at neighbor atoms. In the forward
direction this leads to a focusing effect along rows of atoms which
explains the connection with the crystal lattice in real space rather
than the Brillouin zone in reciprocal space. By appropriate model-
ing [46] or by direct holographic inversion [7] it is now possible
to determine the positions of atoms at surfaces. This is particularly
useful for nding the location of adsorbed molecules relative to the
substrate lattice.
When going from X-rays to the vacuum ultraviolet and from
core levels to valence electrons one nds again a strong angu-
lar dependence of the photoemission intensity, as shown in Fig. 2
for NaCl(1 0 0) (from [810]). The angular dependence is shaped
by the outgoing photoelectrons for the data in Figs. 1 and 2,
i.e., one observes predominantly a nal state effect. This can be
veried by varying the photon energy with synchrotron radia-
tion and exciting the same nal state from different initial states.
With valence electrons, however, there is also a strong initial
Tel.: +1 608 263 5590; fax: +1 608 265 2334.
E-mail address: fhimpsel@wisc.edu.
state effect when the electron escape depth becomes small com-
pared to the electronphonon scattering length. That eliminates
electronphonon scattering, which scrambles the momentum
information without changing the energy signicantly.
For truly elastic photoelectrons it becomes possible to deter-
mine the momentum distribution of the valence electrons in the
initial state and measure their E(k) band dispersion and Fermi sur-
face (for early work see [11] and references therein, for reviews
see [1214]). Initial and nal state effects can be separated using
tunable synchrotron radiation, which allows constant initial state
spectra (CIS) and constant nal state spectra (CFS) to be taken [15].
Synchrotron radiation also makes it possible to tune the momen-
tum perpendicular to the surface, k , which can be varied with
the photon energy. An example is shown in Fig. 3 for LiF(1 0 0),
now at energies high enough to avoid electronphonon scattering
(from [16]). The angular distribution is shown for several initial
state energies E across the F 2p valence band. E is measured rel-
ative to the valence band maximum which lies at k = 0 ( -point).
The valence band minimum lies at the X-point of the Brillouin zone
boundary at 3.5 eV. The upper row of images is taken at a photon
energy where k corresponds to the (3 0 0) X-point in normal emis-
sion (with k in units of 2/a). Consequently, one sees a bright dot at
the center when E is set to 3.4 eV, close to the X-point (left end of
the upper row). In the lower row of images the photon energy is set
for the (4 0 0)  -point in normal emission. Consequently, one sees
a bright dot at the center when E is set to +0.2 eV, near the  -point
(right end of the lower row).
While the photoelectrons in the nal state behave like a nearly
free electron with an inner potential at higher energies, the elec-
trons in the initial state are strongly affected by the crystal potential
and exhibit correlations with the other electrons in the solid. There-
fore, a more sophisticated theory is required to describe initial state
effects. In Fig. 3 a many-body, quasiparticle approach was used

0368-2048/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.elspec.2010.03.007
 F.J. Himpsel / Journal of Electron Spectroscopy and Related Phenomena 183 (2011) 114117
Fig. 1. Observation of angular (momentum) dependence of the photoemission signal from a NaCl single crystal using MgK radiation with a photon energy  = 1.25 keV.
From Siegbahn et al. [1].
to describe the initial states and a nearly-free-electron model for
the nal states [16]. Sophisticated many body theory is required
to get the correct band width for correlated electrons, such as the
localized F 2p electrons in the valence band of LiF. In Fig. 3 the
calculated angular pattern is shown in the bottom half of each
image and compared to the experiment in the top half. All the
experimental features are reproduced by this rst-principles the-
ory, and even the intensities can be compared in semi-quantitative
fashion.
With both initial and nal states under control, one can ask
whether it is possible to extend the concept of photoelectron
diffraction from core levels (where the initial state is a simple -
function) to valence states (where one has to deal with extended
orbitals). Experimentally, the analogy is rather straightforward:
just measure the angular photoemission pattern of valence elec-
trons instead of core levels. Due to the nite energy width of the
valence band one obtains a set of data at several initial state ener-
gies across the band width, as shown in Fig. 3. These correspond
to different initial state wave functions. Can a straightforward
inversion procedure, such as the reconstruction of a hologram [7],
produce a reasonable initial state wave function? In contrast to pho-
toelectron holography, one cannot use the spherical wave emitted
from a core level as reference wave.
115
The same problem is currently receiving a lot of attention for
the reconstruction of a coherent X-ray diffraction pattern from an
isolated nano-object. One faces the usual problem of X-ray crys-
tallography: only the intensity distribution in reciprocal space is
known, not the phases. For periodic crystals there are several stan-
dard ways to overcome the phase problem, such as substitution
by heavy atoms and anomalous scattering at an absorption edge.
For an aperiodic object, such as a single protein molecule or a sin-
gle nanocrystal, the phase problem can be overcome as long as the
object is located in a well-dened spatial region (for example inside
a small aperture) and isolated from neighboring objects [1720]. In
such a situation one can employ an iterative method to retrieve the
phases. To begin, the measured intensities in reciprocal space are
combined with arbitrary phases to enable an inverse Fourier trans-
form into real space. The spurious intensity outside the aperture is
then set to zero and the rest of the data is Fourier transformed into
reciprocal space. The resulting diffraction intensities are corrected
to the experimental values while keeping the phases. A second
inverse Fourier transform leads back into real space. This procedure
can be iterated many times, and it eventually converges to the cor-
rect result. Quite a few renements have been devised, but there is
a common condition for this method to work: The diffraction pat-
tern needs to be oversampled in reciprocal space, i.e., the k-points
116 F.J. Himpsel / Journal of Electron Spectroscopy and Related Phenomena 183 (2011) 114117

Fig. 2. Angle-dependent photoemission from the Cl 3p valence band of NaCl(1 0 0),
varying both the nal energy Ef an the initial energy Ei = Ef  and of the photo-
electrons using synchrotron radiation. From [8] (see also [9,10]).
need to be dense enough that diffraction features from the aper-
ture are still well-sampled. Using electrons instead of X-rays for
coherent imaging is analogous, except for the multiple scattering
problem at low energies. Various techniques have been developed
by electron microscopists ever since Gabors invention of hologra-
phy. A recent example is the coherent imaging of carbon nanotubes
and the electric potential around the tip of a nanotube eld emitter
[2123].
While X-rays and high energy electrons image the core electrons
and thereby provide structural information, one has to use lower
energy photons with angle-resolved photoemission to image the
valence electrons and their orbitals. This has been demonstrated
for quantum well states residing on a terrace of a stepped Au(1 1 1)
surface [24], as shown in Fig. 4. A coherent diffraction experiment is
ideally performed with a single nano-object, but an array of identi-
cal terraces is sufcient to reconstruct meaningful wave functions,
Fig. 4. Momentum-resolved photoemission data are transformed from reciprocal
space (left) to real space (right), thereby revealing the wave functions of quantum
well states conned to a terrace on the stepped Au(23 23 21) surface. This provides
proof of principle that the phase problem can be overcome by utilizing the knowl-
edge that the wave function is conned to nite region in space (here a terrace).
From [24].
as long as the electrons do not scatter coherently between different
terraces. The short mean free path of low energy electrons provides
a rather efcient cutoff that serves as affective aperture. The tran-
sition from coherent to incoherent electrons scattering between
different terraces has been investigated systematically in [25]. The
reconstruction of electron wave functions from a regular array of
terraces can be performed in the same way as for coherent X-ray
diffraction by over-sampling and iterative phase retrieval. By set-
ting the energy of the spectrometer to the three lowest quantum
well states and mapping their intensity distributions in k-space,
three different wave functions are obtained. Even when the wrong
phase is chosen as starting point (for example the phase of the
n = 2 state with the intensity data for the n = 1 state) the iteration

Fig. 3. Comparison between experiment (top half of the images) and theory (bottom half) for angular photoemission patterns from LiF(1 0 0). The momentum perpendicular
to the surface k is varied by changing the photon energy. The upper row of images has the (3 0 0) X-point at the center (Ei = 3.5 eV), and the lower row the (4 0 0)  -point
(Ei = 0 eV). From [16].
 F.J. Himpsel / Journal of Electron Spectroscopy and Related Phenomena 183 (2011) 114117 117

are some subtle differences. Photoemission measures the imagi-

nary part of the Greens function G, multiplied by a matrix element.
Standing waves in STM involve G and T, the reection amplitude at
the structure that produces a standing wave (for example at a pair
of steps). STM probes surface wave functions, while photoemission
penetrates a few layers deep.
STM and photoemission complement each other nicely for
imaging wave functions: working in real space with STM becomes
easier for larger features, while working in reciprocal space with
photoemission becomes easier for small distances. Specically,
STM captures wave functions larger than the orbital at the end of the
STM tip, while angle-resolved photoemission is limited to features
smaller than the Fourier transform of the momentum resolution.
There is an analogy to time-resolved experiments, where one can
use either the time domain or the energy domain. STM and photoe-
mission probe the spatial and momentum domains, respectively.
Ideally, one could combine both techniques and use the overlap
region at intermediate length scales as cross-check.

Acknowledgment

This work was supported by NSF under contracts DMR-0705145

and DMR-0537588 (SRC).

References

[1] K. Siegbahn, U. Gelius, H. Siegbahn, E. Olson, Phys. Lett. A 32 (1970) 221.

[2]
[3]
[4]
[5]
Fig. 5. Two-dimensional transformation of momentum-resolved photoemission [6]
data from reciprocal space (top) to real space (bottom, A) for an array of planar
molecules. The result is compared to theory (bottom, B). From [26]. [7]
[8]
procedure eventually yields the correct wave function. For a one- [9]
dimensional system it is possible to calculate the effective potential [10]
directly from the wave function via the Schrdinger equation. The [11]
[12]
resulting potential V is shown as black curve in Fig. 4. [13]
Recently, this type of wave function imaging has been extended
to a two-dimensional array of planar molecules adsorbed on a sur-
[14]
face [26], as shown in Fig. 5. The resulting planar charge distribution
in panel A is in good agreement with the charge distribution calcu- [15]
lated from rst-principles in panel B. An extension of wave function [16]
imaging to three dimensions has been achieved as well [27], where
[17]
the charge density of bulk copper was obtained from a photoe- [18]
mission intensity map in the irreducible wedge of the Brillouin
zone. In the three-dimensional case one needs to utilize tunable [19]
photon energy to vary the momentum perpendicular to the sur- [20]
face k [6]. This is already clear from Fig. 3, where the two rows
of data represent two different k values with completely differ- [21]
ent angular distributions. In the future, it would be desirable to [22]
push wave function imaging towards a single nano-object. Vari- [23]
ous efforts on spatially resolved photoemission are under way with
[24]
the eventual goal to reach the current focusing limit of soft X-rays
(about 25 nm). That should be sufcient to focus in on a wide variety
of isolated nanostructures, such a nanocrystals, catalytic particles, [25]
and nanowires. [26]
It is useful to consider the alternative way of obtaining wave
functions by scanning tunneling microscopy (STM) combined with [27]
spectroscopic imaging. It is possible to see standing waves created
[28]
by back-reection of surface electron waves at articial structures, [29]
such as a quantum corral or a step. In fact, the quantized terrace
wave functions in Fig. 3 have long been seen by STM [28,29]. There
K. Siegbahn, U. Gelius, H. Siegbahn, E. Olson, Phys. Scripta 1 (1970) 272.
C.S. Fadley, S..L. Bergstrm, Phys. Lett. A 35 (1971) 375.
C.S. Fadley, Prog. Surf. Sci. 16 (1984) 275.
D.P. Woodruff, A.M. Bradshaw, Rep. Prog. Phys. 57 (1994) 1029.
D.H. Rosenblatt, J.G. Tobin, M.G. Mason, R.F. Davis, S.D. Kevan, D.A. Shirley, C.H.
Li, S.Y. Tong, Phys. Rev. B 23 (1981) 3828.
J.J. Barton, Phys. Rev. Lett. 61 (1978) 1356.
F.J. Himpsel, Ph. D. Thesis, Ludwig Maximilians University Munich, 1976
(unpublished).
F.J. Himpsel, W. Steinmann, Phys. Rev. Lett. 35 (1975) 1025.
F.J. Himpsel, W. Steinmann, Phys. Rev. B 17 (1978) 2537.
N.V. Smith, M.M. Traum, Phys. Rev. Lett. 31 (1973) 1247.
F.J. Himpsel, Adv. Phys. 32 (1983) 1.
S.D. Kevan (Ed.), Angle resolved photoemission theory and current applica-
tions, Studies in Surface Science and Catalysis, vol. 74, Elsevier, Amsterdam,
1992.
S. Hfner (Ed.), Very High Resolution Photoelectron Spectroscopy, Lect. Notes
Phys. 715, Springer-Verlag, Berlin Heidelberg, 2007.
G.J. Lapeyre, J. Anderson, P.L. Gobby, J.A. Knapp, Phys. Rev. Lett. 33 (1974) 1290.
E.L. Shirley, L.J. Terminello, J.E. Klepeis, F.J. Himpsel, Phys. Rev. B. 53 (1996)
10296.
J. Miao, P. Charalambous, J. Kirz, D. Sayre, Nature 400 (1999) 342.
D. Shapiro, P. Thibault, T. Beetz, V. Elser, M. Howells, C. Jacobsen, J. Kirz, E. Lima,
H. Miao, A.M. Neiman, D. Sayre, PNAS 102 (2005) 15343.
S. Eisebitt, J. Lning, W.F. Schlotter, M. Lrgen, O. Hellwig, W. Eberhardt, J. Sthr,
Nature 432 (2004) 885.
M.A. Pfeifer, G.J. Williams, I.A. Vartanyants, R. Harder, I.K. Robinson, Nature 442
(2006) 63.
J. Cummings, A. Zettl, M.R. McCartny, J.C.H. Spence, Phys. Rev. Lett. 88 (2002)
056804.
X. Lin, V.P. Dravida, Appl. Phys. Lett. 69 (1996) 1014.
J.M. Zuo, I. Vartanyants, M. Gao, R. Zhang, L.A. Nagahara, Science 300 (2003)
1419.
A. Mugarza, J.E. Ortega, F.J. Himpsel, F.J. Garca de Abajo, Phys. Rev. B 67 (2003)
081404, See also: J.E. Ortega, F.J. Himpsel, Atomic chains at surfaces, (Chapter
6) in [14].
J.E. Ortega, S. Speller, A.R. Bachmann, A. Mascaraque, E.G. Michel, A. Nrmann,
A. Mugarza, A. Rubio, F.J. Himpsel, Phys. Rev. Lett. 84 (2000) 6110.
P. Puschnig, S. Berkebile, A.J. Fleming, G. Koller, K. Emtsev, T. Seyller, J.D. Riley,
C. Ambrosch-Draxl, F.P. Netzer, M.G. Ramsey, Science 326 (2009) 702.
M. Mnsson, T. Claesson, M. Finazzi, C. Dallera, N.B. Brookes, O. Tjernberg, Phys.
Rev. Lett. 101 (2008) 226404.
Ph. Avouris, I.-W. Lyo, Science 264 (1994) 942.
L. Brgi, O. Jeandupeux, A. Hirstein, H. Brune, K. Kern, Phys. Rev. Lett. 81 (1998)
5370.