PRL 100, 063001 (2008)
PHYSICAL REVIEW LETTERS
week ending
15 FEBRUARY 2008
Heteronuclear and Homonuclear High-Spin Alkali Trimers on Helium Nanodroplets
Johann Nagl, Gerald Auböck, Andreas W. Hauser, Olivier Allard, Carlo Callegari, and Wolfgang E. Ernst*
Institute of Experimental Physics, TU Graz, Petersgasse 16, A-8010 Graz, Austria, EU
(Received 27 August 2007; published 13 February 2008)
The electronic excitation spectra of all possible homo- and heteronuclear high-spin (quartet) trimers of
K and Rb (Kx Rb3x , x 0 . . . 3) assembled on the surface of superfluid helium droplets, are measured in
the spectral range from 10 600 to 17 400 cm1 . A regular series of corresponding bands is observed,
reflecting the similar electronic structure of all these trimers. For the assignment and separation of
overlapping bands, we determine x directly, with mass-selected beam depletion, and indirectly with a
V-type double-resonance scheme. The assignment is confirmed by high-level ab initio calculations of the
electronic structure of the bare trimers. The level structure is rationalized in terms of harmonic-oscillator
states of the three valence electrons in a quantum-dot-like confining potential. We predict that three should
be a magic number for high-spin alkali clusters.
DOI: 10.1103/PhysRevLett.100.063001
PACS numbers: 33.20.Kf, 31.15.A, 42.62.Fi, 67.25.dw
Helium nanodroplet isolation spectroscopy has become
a powerful tool for the production and probing of molecules, clusters, and chemical reactions at low temperatures
[1]. Dopant species are cooled down to the internal temperature of the droplet (0.37 K [2]) and readily form
complexes, weakly bound ones being favored [3].
Among all dopants, alkali atoms are unique in that they
remain on the surface [4,5] rather than becoming solvated
inside the droplet; because of the different formation energy, high-spin molecules are almost exclusively observed,
instead of the expected statistical mixture [3,6].
High-spin (quartet) alkali trimers were immediately recognized as a model system to investigate van der Waals
forces between constituents with highly deformable outer
electron shells [6]. Soon, the importance of three-body
forces was shown [7,8]; in this description one has chosen
to view valence electrons as bound each to its own atom. In
large metal clusters, one describes them as delocalized
within a confining potential, thus emphasizing the electronic structure over the geometric structure. This leads to a
shell model, which is of great interest in the study of the
electronic structure and excitation of metal clusters and
quantum dots, particularly as perturbed by a magnetic field
[9].
The explosive growth of Bose-Einstein condensation
research [10,11], largely based on spin-polarized diluted
gases of alkali atoms, has made such topics as two- and
three-body collisions [12,13], molecule formation, especially by magnetic tuning of Feshbach resonances [14,15],
and Efimov states [16,17], of utmost interest. To date,
several electronic transitions of high-spin dimers and
trimers have exclusively been measured in He droplets
[3,6,7,18], where spin has also been a handle to initiate
simple unimolecular reactions [6,18].
Recently, we measured the spin relaxation of alkali
atoms and dimers on He droplets within a magnetic field,
and demonstrated the possibility to create a net spin orientation and to optically address Zeeman substates [19]. In
triplet dimers, a rich physics emerges from the combina0031-9007=08=100(6)=063001(4)
tion of spin-orbit coupling and dopant-droplet interaction
[20], further enriched in quartet trimers by the Jahn-Teller
effect [21].
To date, no one has attempted a systematic investigation
of quartet alkali trimers, perhaps deterred by the spectral
congestion expected for the heavier trimers, and by the
difficulty of a reliable assignment. In this Letter, we approach these two aspects with a combination of massselected beam depletion (MSBD) spectroscopy and
double-resonance (DR) spectroscopy, backed by complete
active space self-consistent field (CASSCF) ab initio calculations of the electronic structure of Kx Rb3x , x
0 . . . 3, quartet trimers. We find a common level pattern,
which will be interpreted in terms of a shell structure. The
equilibrium structure is rigorously an equilateral triangle
(A02 state, D3h symmetry) for K3 and Rb3 , slightly distorted
to an isosceles triangle (B2 state, C2v symmetry) for K2 Rb
and KRb2 . To emphasize the common level pattern, in the
text we identify all states with their D3h labels, which
correlate with C2v as follows: A01 ! A1 , A02 ! B2 , E0 !
A1 B2 , A001 ! A2 , A002 ! B1 , E00 ! A2 B1 .
The experimental setup has been described in detail
elsewhere [22,23]. In brief, He droplets are produced via
supersonic expansion in vacuum of He gas through a cold
nozzle (diameter 5 m, T 14 K, stagnation pressure
60 bar, for an average size of N 104 atoms). The droplet
beam is doped in two sequential pickup cells loaded with K
and Rb metal, respectively, and separately heated. The cell
temperature determines the statistical distribution of dopant atoms per droplet [24,25], and thus the spectra that are
observed and their intensity. In the experiment described
here, the doped droplet beam crosses two laser interaction
zones, which we call the depletion zone and probe zone,
respectively; the latter is fitted with a laser-inducedfluorescence (LIF) detector. Further downstream, the
beam crosses a surface ionization detector (LangmuirTaylor detector, LT) [26], hitting the ionizing rhenium
ribbon at grazing incidence; most of the beam proceeds
unhindered into a quadrupole mass spectrometer (QMS)
063001-1
2008 The American Physical Society
week ending
15 FEBRUARY 2008
PHYSICAL REVIEW LETTERS
1.0
LIF signal [cts/(mW s)]
operating in counting mode. The lasers used are listed in
Ref. [27].
LIF spectra were measured at a set of different pickup
cell temperatures, as suitable for the target trimer. Of
course, not all droplets are doped with the desired number
of atoms, and the spectra of K [4], Rb [28], or their dimers
[22] may appear. Up to now, the lower-lying transitions of
dimers and homonuclear trimers have been addressed, and
the occurrence of overlapping spectra has been rare. This is
not so in the present investigation where a multitude of
complexes is formed. When sufficient, these are identified
via the unfolding of LIF signal versus vapor density inside
the pickup cell (growth analysis) [23]. Beyond this, particularly for mixed trimers, we use two spectroscopic
schemes relying on desorption of the dopants from their
host droplet upon excitation [29].
The first scheme is a two-color V-type double-resonance
measurement: The tagging laser crosses the beam in the
depletion zone, where it excites a known transition, thus
desorbing the target species. The second laser crosses in
the probe zone, where the intensity of the LIF signal from
the unknown transition is affected by the tagging laser only
if the two transitions share the same ground state. The
second scheme is based on mass-selective beam depletion:
the ion signal from the QMS, set at the mass of the
candidate species, is monitored as a laser is scanned across
the transition. Again, a depletion signal is observed if the
assignment is correct. The first scheme is more sensitive
but relies on an accessible, known transition of the candidate species. The second is more flexible, but has a poorer
signal.
With Rb only, we observe, in the spectral range of the
Ti:Al2 O3 laser, two new bands extending from 11 500 to
12 000 cm1 (Figs. 1 and 2) and from 14 000 to
14 250 cm1 (Fig. 2), which, based on signal growth
analysis, we assign to the 24 E0
14 A02
14 A02 and 14 A001
trimer transitions. With K only, we observe the known
24 E0
14 A02 trimer transition [18].
With K and Rb simultaneously, we observe a complex
LIF spectrum (Fig. 1, top panel). We decompose it by
MSBD into four contributions: two from the above 24 E0
14 A02 bands of K3 , Rb3 , and two new ones from the corresponding bands of K2 Rb, KRb2 . Traces in the bottom panel
show MSBD signals for the most abundant isotopologues:
at ion mass 163 (K2 Rb ) and 209 (KRb
2 ). A scan at mass
)
proves
the
method
to
work
reliably,
as LIF and
255 (Rb
3
MSBD spectra have exactly the same shape. Unexpectedly,
the signal at mass 117 (K
3 ) was unstable and no useful
MSBD spectrum could be recorded. As a further check, we
verify that a weighted sum of the four trimers’ spectra fully
accounts for the profile of a typical LIF spectrum (Fig. 1,
top panel. Also see Ref. [30]).
Four unstructured peaks, each from a different molecule,
appear in the spectral range between 14 000 and
15 000 cm1 (Fig. 2). The first, in order of increasing
photon energy, belongs to Rb3 (see above). The fourth
8
0.8
ion signal [arb. units]
PRL 100, 063001 (2008)
6
0.6
0.4
4
0.2
2
0.0
0.5
0.4
0.3
0.2
0.1
0.0
11400
0
11600
11800
12000
-1
wave number [cm ]
12200
FIG. 1 (color online). Bottom panel: MSBD electronicexcitation spectra of Kx Rb3x trimers, weighted to best fit the
LIF spectrum above. Thick solid (red) line, KRb2 ; thick dashed
(blue) line, K2 Rb; thin solid (black) line, Rb3 ; thin dashed
(orange) line, K3 , LT beam-depletion spectrum. Top panel:
Comparison between a typical LIF spectrum (gray line) and
the sum of the individual depletion spectra (black line).
appears upon the presence of K; DR with the tagging laser
set at 11 920 cm1 (K3 ) results in a 70% decrease of LIF,
whereas tagging at 13 900 cm1 (K2 ) produces no effect.
The second and third peaks come from heteronuclear
molecules. No dimer transitions are expected, according to
potential energy curves [31]. Because of poor signal-tonoise ratio, no growth analysis was performed; we had
previously observed, however, that an increase in Rb density in the pickup cell shifts the relative strength towards
the second band, which is thus assigned to KRb2 . The third
1A1"
2E'
3E'
Rb3 (D3h)
3A1 4B2
3A2
KRb2 (C2v)
4B2 3A1
3A2
K2 Rb (C2v)
2E'
1A1"
3E'
K3 (D3h)
11500 12000
14000 14500 15000 15500 16000
-1
wave number [cm ]
FIG. 2 (color online). Survey of the electronic-excitation
bands of homo- and heteronuclear Kx Rb3x trimers. Vertical
lines mark the position of the ab initio transition energies (all
quartet states).
063001-2
PRL 100, 063001 (2008)
week ending
15 FEBRUARY 2008
PHYSICAL REVIEW LETTERS
band is assigned to K2 Rb: DR with the tagging laser tuned
at 11 944 cm1 , where the 24 E0
14 A02 band of K2 Rb is
strong, and that of KRb2 weak, indeed shows 30%
depletion. The pattern of these four peaks suggests a
similar electronic configuration of the participating states;
14 A02 band.
we assign them to the 14 A001
Based on double resonance with 14 A001
14 A02 , a new
1
band of Rb3 at 15 000 cm is assigned to the 34 E0
14 A02 transition. This band had already been observed in
LIF excitation spectroscopy of Rb-doped helium droplets
[32]. Dispersed-fluorescence spectra had shown emission
from the 11 u ! 11
g transition of free Rb2 , consistent
with the excitation energy, but inconsistent with the concurrent emission from free Rb atoms: dissociation of Rb2
from the 11 u state into any excited-atom channel is
energetically forbidden. DR reveals that the excitation
spectrum consists, in fact, of two transitions and allows
their separation. The tagging laser is set to excite the Rb3
band at 14 100 cm1 , the probe laser is scanned twice, with
and without the tagging laser, and LIF spectra are recorded.
The difference signal (shown in Fig. 2) represents then a
quartet Rb3 spectrum, which accounts for the free-atom
emission, while the unaffected portion (not shown)
matches the Franck-Condon envelope of the Rb2 11 u
transition (calculated using potentials from
11
g
14 A02 transition of
Ref. [33]). Likewise, with the 24 E0
K3 as a tag, we assign a similar structure at 16 200 cm1
to a mixture of the 11 u
11
g band of K2 (potentials
4 0
from Ref. [34]) and the 3 E
14 A02 band of K3 (Fig. 2).
The range between 15 000 and 16 000 cm1 is very congested (three dimer transitions also fall here), so we are
unable to recognize and separate mixed trimers spectra. We
also do not see (either in the experiment or in the calculations) the band previously observed at 12 800 cm1
14 A02 transition of
and tentatively assigned to the 14 A001
K3 [18]. Because it does not fit the regular pattern we
uncovered, we believe it must originate from some other
oligomer.
Assignment of all the trimer transitions is based on
ab initio calculations, performed with MOLPRO [35].
Details will be published separately; in brief, we use
relativistic small-core effective-core potential basis sets
of the Stuttgart/Cologne group [36]. We first optimize the
geometry of the electronic ground state for all trimers at the
RCCSD(T) level of theory (restricted coupled cluster calculations with single, double, and noniterative triple excitation). In agreement with Ref. [8], K3 and Rb3 are
equilateral triangles with bond length and binding energy
5.06 Å, 1260 cm1 , and 5.52 Å, 929 cm1 , respectively.
K2 Rb and KRb2 are isosceles triangles with bond lengths
5.26 and 5.29 Å, angles 57.61 and 62.19 , and binding
energies 1159 and 1049 cm1 , respectively.
We calculate the vertical excitation energy at the global
minimum geometry for the eight lowest optically active
electronic transitions using the state-averaged CASSCF
method of Werner and Knowles [37,38], with three elec-
trons in the active space; because of restrictions in the
computer code, all calculations are done in C2v symmetry.
At large internuclear separations, our calculations deviate
from the experimental asymptotic (atomic) excitation
energies [39]. We enforce consistency by ad hoc shifts of
the calculated states: 1500 and 1730 cm1 for those
correlating, respectively, to excited K and excited Rb; this
gives the level pattern shown in Fig. 2. The splitting of the
2E0 manifold for K2 Rb and KRb2 , due to the lower molecular symmetry, is clearly visible in the experimental
spectra and is well reproduced by the calculations. The
complicated structure of the same band in Rb3 is due to the
Jahn-Teller distortion with strong spin-orbit coupling, and
will be analyzed separately [21].
The level structure underlying the observed spectra can
be clarified in terms of delocalized orbitals, which reveal
the existence of a shell structure. As often observed for
single-particle states in quantum dots [9], the canonical
orbitals from the ab initio calculations resemble harmonicoscillator (HO) eigenfunctions. Because of the different
strength of the potential in-plane and off-plane, we label
orbitals, according to their nodal structure, with quantum
numbers n; ‘; nz corresponding to a basis of twodimensional HO eigenfunctions (Fock-Darwin orbitals
[9]) in the molecular plane r; times one-dimensional
HO eigenfunctions in the z direction; HO eigenvalues are
hc~
2n j‘j 1 hc~
n 1 and hc~
z nz 1=2.
The energies E of the lowest 13 orbitals, measured from
0; 0; 0, fit well to a 5-parameter formula with anharmonic
and cross terms: E=hc n ~ nz ~z x n2 ~
p
xz n2z ~z xz n nz ~ ~z . Because we only have nz 0
or 1, xz is redundant and we set it to 0. The best fit
parameters and rms of the residuals are reported in
Table I. In a high-spin configuration, each orbital can
accommodate only one of the three electrons (Pauli principle); in the ground state, the 0; 0; 0 orbital and the
0; 1; 0 pair are occupied. The observed bands correspond to excitations of one of the 0; 1; 0 electrons into
0; 2; 0 for 2E0 , and 0; 1; 1 for 1A001 . The 3E0 band is a
mix of shells arising from the degeneracy of 1; 1; 0,
0; 3; 0, and of the doubly excited state with both
0; 1; 0 electrons promoted to 0; 2; 0.
As the ground state corresponds to a complete j‘j 1
shell, we infer that the trimer should be more stable than
the planar tetramer. This shell structure is the half-filledorbitals analog of the celebrated observation in low-spin
TABLE I. Parameters of fit of orbital energies.
K3
K2 Rb
KRb2
Rb3
063001-3
~
(cm1 )
~z
(cm1 )
x
xz
xz
Res. rms
(cm1 )
14 963
14 269
13 975
13 312
19 589
18 747
18 104
17 329
0.1237
0.1201
0.1161
0.1136
0
0
0
0
0.3571
0.3353
0.2963
0.2754
930
1100
1210
1070
PRL 100, 063001 (2008)
PHYSICAL REVIEW LETTERS
Na clusters [40]; it is remarkable that it occurs for high
spins too, already at such a small number of atoms.
In summary, we measured the excitation spectra of all
possible quartet-state homo- and heteronuclear trimers of
K and Rb (Kx Rb3x , x 0 . . . 3) in the spectral range from
10 600 to 17 400 cm1 ; we have separated and assigned
them with a combination of pickup statistics, massselected beam depletion, and double-resonance spectroscopy. A regular pattern of bands common to all these
trimers is observed, shifting to progressively lower photon
energies across the series K3 , K2 Rb, KRb2 , Rb3 . Our highlevel ab initio calculations of the corresponding excitation
energies agree well with experimental data; the obtained
electronic structure is interpreted in terms of harmonicoscillator orbitals and indicates that a shell model is applicable to these high-spin adducts. The experimental method
can be extended to any desired alkali-metal trimer and
almost certainly will remain applicable to larger clusters,
where the shell model will be of great help to assign and
interpret the measured spectra.
Beyond the scope of this Letter, Rb3 turns out to be a
model system for the study of Jahn-Teller distortion combined with strong spin-orbit coupling. Also, our calculations yield the first ground-state potentials of mixed
trimers, which are of importance for the investigation of
cold atom-molecule collisions, and, more generally, of the
interaction between spin-polarized alkali atoms.
We thank Pavel Soldán for advice on ab initio calculations. This research is supported by the Austrian Science
Fund (FWF, Grant No. P18053-N02), and the European
Research and Training Network (Contract No. HPRN-CT2002-00290).
*Corresponding author.
wolfgang.ernst@tugraz.at
[1] Special issue on Helium Nanodroplets: A Novel Medium for Chemistry and Physics [J. Chem. Phys. 115
(2001)].
[2] M. Hartmann, R. E. Miller, J. P. Toennies, and A. Vilesov,
Phys. Rev. Lett. 75, 1566 (1995).
[3] J. Higgins et al., J. Phys. Chem. A 102, 4952 (1998).
[4] F. Stienkemeier et al., Z. Phys. D 38, 253 (1996).
[5] F. Ancilotto et al., Z. Phys. B 98, 323 (1995).
[6] J. Higgins et al., Science 273, 629 (1996).
[7] J. Higgins, W. E. Ernst, C. Callegari, J. Reho, K. K.
Lehmann, G. Scoles, and M. Gutowski, Phys. Rev. Lett.
77, 4532 (1996).
[8] P. Soldán, M. T. Cvitas, and J. M. Hutson, Phys. Rev. A 67,
054702 (2003).
[9] S. M. Reimann and M. Manninen, Rev. Mod. Phys. 74,
1283 (2002), and references therein.
week ending
15 FEBRUARY 2008
[10] J. Herbig et al., Science 301, 1510 (2003).
[11] M. W. Zwierlein, C. A. Stan, C. H. Schunck, S. M. F.
Raupach, S. Gupta, Z. Hadzibabic, and W. Ketterle,
Phys. Rev. Lett. 91, 250401 (2003).
[12] A. Simoni, F. Ferlaino, G. Roati, G. Modugno, and
M. Inguscio, Phys. Rev. Lett. 90, 163202 (2003).
[13] J. M. Hutson and P. Soldán, Int. Rev. Phys. Chem. 26, 1
(2007).
[14] T. Köhler, K. Góral, and P. S. Julienne, Rev. Mod. Phys.
78, 1311 (2006).
[15] J. M. Hutson and P. Soldán, Int. Rev. Phys. Chem. 25, 497
(2006).
[16] T. Kraemer et al., Nature (London) 440, 315 (2006).
[17] E. Braaten and H. W. Hammer, Ann. Phys. (N.Y.) 322, 120
(2007).
[18] J. H. Reho et al., J. Chem. Phys. 115, 10265 (2001).
[19] J. Nagl, G. Auböck, C. Callegari, and W. E. Ernst, Phys.
Rev. Lett. 98, 075301 (2007).
[20] G. Auböck et al., J. Phys. Chem. A 111, 7404 (2007).
[21] G. Auböck et al. (unpublished).
[22] O. Allard et al., J. Phys. B 39, S1169 (2006).
[23] J. Nagl et al., J. Phys. Chem. A 111, 12386 (2007).
[24] M. Hartmann et al., J. Chem. Phys. 110, 5109 (1999).
[25] S. Vongehr and V. V. Kresin, J. Chem. Phys. 119, 11124
(2003).
[26] F. Stienkemeier, M. Wewer, F. Meier, and H. Lutz, Rev.
Sci. Instrum. 71, 3480 (2000).
[27] All lasers from Coherent, Inc. The pump laser is an
Innova 200, power stabilized at 21 W, multiline visible.
The tunable lasers, all multimode, are an 899-01 Ti:Al2 O3
ring laser with mid- and shortwave optics, a 599 linear dye
laser operating on DCM special, and a 699 ring dye laser
operating on Rhodamine 6G.
[28] F. R. Brühl et al., J. Chem. Phys. 115, 10220 (2001).
[29] C. Callegari et al., J. Phys. Chem. A 102, 95 (1998).
[30] The LIF spectrum was taken in an earlier experimental run
at slightly different conditions. Lacking an MSBD spectrum for the K3 contribution, we use instead an older LT
beam depletion spectrum.
[31] S. Rousseau, A. R. Allouche, and M. Aubert-Frécon,
J. Mol. Spectrosc. 203, 235 (2000).
[32] F. R. Brühl et al., J. Chem. Phys. 115, 10275 (2001).
[33] S. Park et al., J. Mol. Spectrosc. 207, 129 (2001).
[34] S. Magnier et al., J. Chem. Phys. 121, 1771 (2004).
[35] H.-J. Werner et al., MOLPRO, version 2006.1, a package of
ab initio programs (2006); see http://www.molpro.net.
[36] I. S. Lim et al., J. Chem. Phys. 122, 104103 (2005); http://
www.theochem.uni-stuttgart.de/pseudopotentials/.
[37] H.-J. Werner and P. J. Knowles, J. Chem. Phys. 82, 5053
(1985).
[38] P. J. Knowles and H.-J. Werner, Chem. Phys. Lett. 115,
259 (1985).
[39] Y. Ralchenko et al., NIST Atomic Spectra Database version 3.1.2 (2007); http://physics.nist.gov/asd3.
[40] W. D. Knight et al., Phys. Rev. Lett. 52, 2141 (1984).
063001-4