week ending

PRL 107, 193003 (2011) PHYSICAL REVIEW LETTERS 4 NOVEMBER 2011

Structure and Stability of the Negative Hydrogen Molecular Ion

B. Jordon-Thaden,1 H. Kreckel,1,* R. Golser,2 D. Schwalm,1,3 M. H. Berg,1 H. Buhr,3,1 H. Gnaser,4 M. Grieser,1
O. Heber,3 M. Lange,1 O. Novotný,1,5 S. Novotny,1 H. B. Pedersen,1,† A. Petrignani,1,‡ R. Repnow,1
H. Rubinstein,3 D. Shafir,3 A. Wolf,1 and D. Zajfman3
1
Max-Planck-Institut für Kernphysik, 69117 Heidelberg, Germany
2
Universität Wien, Fakultät für Physik–Isotopenforschung, 1090 Wien, Austria
3
Department of Particle Physics, Weizmann Institute of Science, Rehovot 76100, Israel
4
Department of Physics and Research Center OPTIMAS, University of Kaiserslautern, D-67663 Kaiserslautern, Germany
5
Columbia Astrophysics Laboratory, 550 West 120th Street, New York, New York 10027, USA
(Received 17 August 2011; published 3 November 2011)
We present the results of a Coulomb explosion experiment that allows for the imaging of the rovibra-
tional wave function of the metastable H
2 ion. Our measurements confirm the predicted large internuclear
separation of 6 a.u., and they show that the ion decays by autodetachment rather than by spontaneous
dissociation. Imaging of the resulting H2 products reveals a large angular momentum of J ¼ 25  2,
quantifying the rotation that leads to the metastability of this most fundamental molecular anion.

DOI: 10.1103/PhysRevLett.107.193003 PACS numbers: 33.15.Dj, 79.77.+g

Almost 50 years ago Taylor and Harris [1] showed that ion trap measurements. It predicts a shallow potential
the 2
þ

u H2 complex is instable relative to autodetachment minimum formed by a centrifugal barrier at large internu-
into the H2 þ e
continuum for internuclear distances of clear distances that supports vibrational states with v ¼ 0
less than 3 a.u. More recent theoretical considerations [2,3] for rotational angular momenta up to J ¼ 27, which are
indicated extremely short lifetimes ( 10–15 s), which metastable against autodetachment into H2 þ e
and frag-
would render a direct observation of H
2 impossible. mentation into H
þ H [8,12].
Experimentally, H
2 ions are hard to identify as they have To study the structure of metastable H
2 , we have per-
almost the same charge to mass ratio as the stable D
ion. formed foil-induced Coulomb explosion imaging (CEI)
Nevertheless, mass spectroscopic detection of H
2 was first measurements [13,14]. By triggering a Coulomb explosion
described in 1957 by Khvostenko and Dukel’Skii [4], on a time scale of 100 as [13], these measurements allow us
followed by reports in the 1970s of H
2 creation in a to determine the square of the nuclear wave function of the
duoplasmatron source [5] and observations of HD
and molecule as a function of the internuclear distance R [15].
D
2 in ion beams at accelerator facilities [6]. However, Negative ions with 2 atomic mass units were extracted
when a careful—albeit unsuccessful—search for meta- from a cesium sputter source utilizing a solid HfH2 target.
stable negative hydrogen ions was carried out by Bae The ion beam with a current of 500 pA was accelerated to
et al. in 1984, the authors concluded that ‘‘H


2 (and H3 ), 0.97 MeV by a linear radiofrequency quadrupole accelera-
do not, in fact, exist metastably’’ [7]. Since then, reported tor and magnetically guided into the CEI beam line
detections were traditionally met with skepticism based on (Fig. 1). The beam consisted mainly of D
ions
the experimental difficulties and lack of reproducibility. (> 99:9%) with only a small fraction of H
2 . Moreover,
This notion was corrected only recently by the definitive of the metastable H
2 states produced in the source, only
detection of H



2 , D2 , and H3 at the VERA tandem accel- the longest-lived state, with a lifetime of 8:2  1:5
s
erator [8], by ensuing lifetime measurements [9], and by a [9,12], is expected to survive the flight time of 11
s
photofragment imaging experiment of D
2 [10]. from the source to the CEI setup.
In reaction dynamics, the H
2 complex has long been Behind the last bending magnet, the ion beam was
assumed to play a key role as a transient state in many collimated by two apertures of 1 and 1.5 mm diameter,
fundamental reactions like dissociative attachment (DA: respectively, before traversing an ultrathin diamondlike
e
þ H2 ! H


2 ! H þ H ) and associative detachment carbon (DLC) foil of 0:8
g=cm2 . Upon entrance in
(AD: H þ H ! H2 ! H2 þ e
), since the magnitude


the foil all electrons are stripped off instantaneously while
of the cross section for these processes suggests a resonant the bare protons and deuterons traverse and exit the foil
mechanism (see [3] and references therein). The latter almost unperturbed. The Dþ ions were discriminated from
process is the dominant formation mechanism of H2 mole- the protons by a magnetic field separator behind the strip-
cules in the epoch of first star formation in the early ping region; thus only the two protons resulting from the
Universe [11]. The nonlocal theory that was employed to Coulomb explosion of diatomic hydrogen were detected.
model the H
2 collision complex in [3,11] has also been The time scale of electron stripping in the foil amounts
applied to explain long-lived H
2 states [12] observed in the to 10
16 s, which is much shorter than the time scales of

0031-9007=11=107(19)=193003(4) 193003-1 Ó 2011 American Physical Society

 week ending
PRL 107, 193003 (2011) PHYSICAL REVIEW LETTERS 4 NOVEMBER 2011
5 kV
chopper collimator 1 3D detector
H++H+

Kinetic Energy Release (eV)

dipole
steerer collimator 2 DLC foil +
H Coulomb
0.97 MeV +
20 20
− − D Coulomb + Rotation
D and H2 −
beam
auxilliary magnetic
magnetic field separation field
10 10
134 cm 300 cm 192 cm 297 cm

FIG. 1. Schematic of the CEI beam line (distances not to

0 0
scale). After collimation the ions pass through a diamondlike
carbon (DLC) foil that strips off all electrons. The stripping foil H+H
covers an aperture of 2 mm diameter in a stainless steel holder. A 0.8 H2 (X1Σg, J=26)

Potential Energy (eV)

magnetic field guides the two protons from the Coulomb ex-
plosion of H
þ
2 towards a 3D imaging detector, while D ions— |ψ(R)|2(H2-,v=0,J=27)

0.4
resulting from electron stripping of D , the unwanted main
constituent of the beam—are discriminated. An auxiliary mag-
H2- (2Σ+u, J=27) H + H-
netic field can be introduced before the DLC foil to deflect all 0.0
charged particles prior to entering the stripping region. |ψ(R)|2(H2,v=0,J=26)
-0.4
molecular rotation (10
12 s) or vibration (10
14 s) and
0 5.0 10.0 15.0
therefore the internuclear distance R can be considered R (atomic units)
frozen during the stripping process. Without the binding
electrons, the two bare protons are driven away from one FIG. 2 (color online). Illustration of the Coulomb explosion
another by their mutual Coulomb repulsion and the poten- principle for molecular hydrogen. Because of the fast stripping
tial energy EC  e2 =R is rapidly converted into kinetic process the initial wave functions are projected vertically onto
energy [13]. The total energy release is determined by a the Coulomb potential between the two bare protons. The
three-dimensional (3D) imaging detector [14] located 3 m measured kinetic energy release distributions are thus mirror
downstream of the stripping foil, where the protons have images of the square of the initial wave functions, which are
gained distances of a few centimeter. For 3D imaging, the slightly shifted to higher energies due to the high angular
detector records the spatial distance of the protons upon momentum of the H
2 ions. Shown in the lower panel are
exemplary molecular potential curves for H
2 and for the auto-
impact on a microchannel plate as well as their relative
detachment product H2 as well as the square jðRÞj2 of the
impact times for one molecule at a time. A fast high- nuclear wave functions corresponding to the v ¼ 0, J ¼ 27 state
voltage chopper before the last bending magnet prevents of H
2 [20] and the v ¼ 0, J ¼ 26 state of H2 [21]. The kinetic
fragments from more than one molecule to reach the energy release as a function of the internuclear distance R and
detector per event. the resulting kinetic energy release distributions are displayed in
In a semiclassical description of the Coulomb explosion the upper panel.
process (see Fig. 2) the kinetic energy distribution PðEkin Þ
is related to the square of the nuclear wave function
;J ðRÞ by [16] influence of the magnetic separation field and the detector







efficiency. The momenta of the bound protons due to rota-



dR


PðEkin Þ  j;J ðRÞj2








: (1) tion are included in the simulation as initial velocities

dEkin



@ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
The measured kinetic energy Ekin for an individual mo- vJ ðRÞ ¼ JðJ þ 1Þ (3)
2
R
lecular event is connected to the internuclear distance R
upon entrance into the foil by perpendicular to the molecular axis. (In principle the zero-
@ 2 point motion of the molecular vibration also adds to the
Ekin ðRÞ ¼ EC ðRÞ þ JðJ þ 1Þ; (2) initial velocity, but this contribution is negligible in the
2
R2
present case [16].) So-called ‘‘wake effects’’ arising from
with
denoting the reduced mass of H
2 . We performed the interaction of the fast particles with the foil atoms [19]
model calculations that showed that the semiclassical treat- can be effectively eliminated by choosing only events that
ment results in a very good approximation of the fully are near parallel to the detector plane. Therefore, all events
quantum mechanical results. with cos < 0:4 (with  being the angle between the mo-
The detailed Coulomb interaction, as well as small per- lecular axis and the detector plane) are discarded. The foil
turbations inside the foil, are incorporated in a Monte Carlo interactions are well understood and contribute around 20%
code of the Coulomb explosion process [17]. It computes to the width of the measured distribution.
classical trajectories, taking into account multiple scatter- Figure 3 displays the outcome of the CEI measurement
ing and charge exchange inside the foil [18], and the (black dots) before cuts. The distribution shows a peak at a

193003-2
 week ending
PRL 107, 193003 (2011) PHYSICAL REVIEW LETTERS 4 NOVEMBER 2011

6000 By varying the pressure in the beam line, we found an

5000
H2- upper limit of 10% for the fraction of the H2 signal caused
H2 (x5) by stripping on residual gas. Hence the majority of the H2
Counts / 0.2 eV 4000 events stem from autodetachment. This finding is corrobo-
rated by tests with metastable D
2 ions for which the
3000
sensitivity to residual gas induced D2 events is much larger,
2000 as the relevant D
2 states have considerably longer life-
times against autodetachment [9,10,12]. Note that the ex-
1000
periment is not sensitive to the spontaneous dissociation of
0 H


2 into H and H . The two protons resulting from such an
0 5 10 15 20 25 30
Kinetic Energy Release (eV)
event behind the foil would hit the detector too closely
together to be detected as two-body events.
FIG. 3 (color online). Measured kinetic energy release after From the relative intensities of the two peaks and the
Coulomb explosion. The black dots show the distribution re- length of the straight section between steerer magnet and
corded with the H
2 beam before cuts. The green squares show stripping foil, we can estimate the partial decay rate for
the KER distribution for the autodetachment product H2 , mea- autodetachment to be 2þ1:3 5
1

0:6  10 s . The largest contri-
sured with the auxiliary magnetic field in front of the DLC target bution to the uncertainty is the influence of Earth’s magnetic
being switched on. field, which leads to a slight difference in transmission
between the H2 and H
2 ions through the beam collimators.
kinetic energy release of 5 eV, attributed to H
2 , and a The inverse of this autodetachment rate corresponds to a
second smaller peak around 16 eV, which coincides with lifetime of 5  2
s, which is shorter, but consistent within
the kinetic energy release expected for neutral H2 mole- the error margins, with the previous lifetime measurement
cules. To verify whether the second peak is caused by that yielded 8:2  1:5
s [9]. The present result thus re-
neutral molecules, we introduced an auxiliary magnetic veals that autodetachment is indeed the dominant decay
field between the second collimator and the stripping foil channel for the most long-lived H
2 resonance.
(see Fig. 1), thereby deflecting all charged particles before The above-mentioned deflection of the H
2 beam by the
they enter the stripping region. In the distribution measured Earth’s magnetic field leads to a small separation between
with this arrangement (green squares) the peak at 5 eV was the center-of-mass (c.m.s.) of the neutral H2 events and the
absent while the 16 eV structure remained. Consequently, H
2 events in the detector (xy) plane. By applying 1 c.m.s.
we assign the smaller peak to neutral H2 molecules, created cuts in the respective x and y coordinates we can thus
between the last bending magnet and the target foil. readily separate the H2 and H
2 events. The measured H2

18
(a) H2- 7 (b) H2
16 (v=0,J=25)
(v=0,J=27) 6
14 (v=0,J=26) sum of
Intensity (arb. units)

84% (v=0,J=25)
Intensity (arb.units)

12 5 16% (v=1,J=25)

10 4

8
3 (v=0,J=0)
6
2
4

2 1

0 0
0 5 10 15 5 10 15 20 25 30
Kinetic Energy Release (eV) Kinetic Energy Release (eV)

FIG. 4 (color online). Kinetic energy release (KER) distributions measured for (a) H
2 and (b) H2 after all cuts. The error bars reflect
the statistical uncertainties on a 1 level. The accuracy of the reconstructed kinetic energies is on the order of the symbol size. Also
shown are fits with simulated distributions based on calculated wave functions. For H
2 (a), the theoretical wave functions for the two
longest-lived states are used [12,20]. Best agreement is found for the state with the maximum predicted lifetime of 6:3
s (assigned
with v ¼ 0, J ¼ 27) [12], while the distribution for the (v ¼ 0, J ¼ 26) state with the next-longest predicted lifetime of 0:38
s is
already clearly falling short of explaining the data. Note that the calculated H

2 potential only supports states with v ¼ 0 and J  27
[12]. For the neutral autodetachment product H2 (b), the KER distribution is best described by (v ¼ 0, J ¼ 25) when assuming only a
single H2 state to be populated (dashed blue curve). The agreement between simulation and data can be further improved by allowing
also for a small v ¼ 1 contribution (solid red curve).

193003-3
 week ending
PRL 107, 193003 (2011) PHYSICAL REVIEW LETTERS 4 NOVEMBER 2011

KER distribution is plotted in Fig. 4(a). The distribution V. Kh. Lichtenstein for providing the DLC foils. This
peaks at 5 eV, directly revealing—independent of its research was supported by the German-Israeli Foundation
detailed theoretical interpretation—that the average dis- for Scientific Research and Development (G.I.F.) under
tance of the two protons in H
2 , measured 11
s after Grant No. I-900-231.7/2005. D. Sch. acknowledges sup-
its production, amounts to hRi  6 a:u: Moreover, the bell- port by the Weizman Institute through the Joseph
shaped KER distribution readily identifies v ¼ 0 as the Meyerhoff program. H. B. was supported by the
dominant vibrational state. When comparing the measured European Project ITS LEIF (HRPI-CT-2005-026015).
KER distribution to simulated distributions based on theo-
retical H
2 wave functions calculated by Čı́žek et al.
[12,20], best agreement is found for the (v ¼ 0, J ¼ 27)
state. This state is also predicted to have the longest life- *[email protected]
time of 6:3
s [12], close to the measured lifetime of Present address: Chemistry Department, University of
8:2  1:5
s [9]. †
Illinois, 61801 Urbana, IL, USA.
Direct experimental information on the rotational quan- Present address: Department of Physics and Astronomy,
tum number of the H
University of Aarhus, DK-8000 Aarhus, Denmark.
2 state can be inferred from the ‡
analysis of the KER distribution of the H2 decay product Present address: Leiden Observatory, Leiden University,
PO Box 9513, NL-2300 RA Leiden, The Netherlands.
shown in Fig. 4(b). Using simulated distributions based on
[1] H. S. Taylor and F. E. Harris, J. Chem. Phys. 39, 1012
the well-known rovibrational H2 wave functions [21], a (1963).
2 fit of the measured distribution, assuming only a single [2] P. L. Gertitschke and W. Domcke, Phys. Rev. A 47, 1031
(v ¼ 0) state to contribute, results in J ¼ 25. Remaining (1993).
differences between the measured and simulated distribu- [3] M. Čı́žek, J. Horácek, and W. Domcke, J. Phys. B 31, 2571
tions may be either due to a small contribution of a (v ¼ 1) (1998).
state, which are still energetically accessible in the decay [4] V. I. Khvostenko and V. M. Dukel’Skii, Sov. Phys. JETP 7,
of H
2 for J < 26, or due to an additional broadening effect
709 (1958).
caused by the divergence of the H2 beam, which is difficult [5] R. Hurley, Nucl. Instrum. Methods 118, 307 (1974).
to account for. Given the precision of the data and the [6] W. Aberth, R. Schnitzer, and M. Anbar, Phys. Rev. Lett.
34, 1600 (1975).
quality of the fits, we conclude that our measurement
[7] Y. K. Bae, M. J. Coggiola, and J. R. Peterson, Phys. Rev. A
confines the rotational quantum number of the autode- 29, 2888 (1984).
tached H2 molecules to J ¼ 25  2. [8] R. Golser et al., Phys. Rev. Lett. 94, 223003 (2005).
As theoretical considerations suggest that autodetach- [9] O. Heber et al., Phys. Rev. A 73, 060501 (2006).
ment will be dominated by an outgoing p-wave (J ¼ [10] L. Lammich et al., Phys. Rev. A 80, 023413 (2009).
1) electron [22], with a strong preference for J ¼
1 [11] H. Kreckel et al., Science 329, 69 (2010).
for energetic reasons, this result supports the J ¼ 27 as- [12] M. Čı́žek, J. Horáček, and W. Domcke, Phys. Rev. A 75,
signment for the most long-lived H
2 state.
012507 (2007).
In summary, we present experimental results for the [13] Z. Vager, R. Naaman, and E. P. Kanter, Science 244, 426
structure and lifetime of the most fundamental molecular (1989).
anion H
[14] R. Wester et al., Nucl. Instrum. Methods Phys. Res., Sect.
2 . After decades of speculation, our measurements
A 413, 379 (1998).
confirm the predicted mechanism lending metastability to
[15] Z. Amitay et al., Science 281, 75 (1998).
the H
2 complex not only on a qualitative level, but also [16] Z. Amitay et al., Phys. Rev. A 60, 3769 (1999).
quantitatively, and they provide a comprehensive verifica- [17] D. Zajfman et al., Phys. Rev. A 46, 194 (1992).
tion of the theoretical description. [18] D. Zajfman et al., Phys. Rev. A 41, 2482 (1990).
CEI measurements of the more long-lived D
2 ion are [19] R. Garcia-Molina et al., Nucl. Instrum. Methods Phys.
presently being analzyed. Further motivation to study Res., Sect. B 230, 41 (2005).
small negative ions comes through recent theoretical cal- [20] M. Čı́žek (private communication). The wave functions
culations that predict stable negative triatomic hydrogen were calculated using an adapted molecular potential,
ions [23] and the suggestion that H
3 may be detectable in
which was introduced in Ref. [12] to better account for
interstellar space [24]. Therefore it is foreseen to carry out the lifetime measured for the most long-lived H
2 state.
[21] R. J. Le Roy, LEVEL 8.0, University of Waterloo
Coulomb explosion measurements with H


3 and D3 in the Chemical Physics Research Report No. CP-663, 2007;
near future. see http://leroy.uwaterloo.ca/programs/.
The authors thank M. Čı́žek for providing his calculated [22] M. Berman, C. Mündel, and W. Domcke, Phys. Rev. A 31,
H
2 wave functions. The authors also thank S. Altevogt, D. 641 (1985).
Bing, J. Hoffmann, and M. B. Mendes for technical [23] M. Ayouz et al., J. Chem. Phys. 132, 194309 (2010).
assistance, L. Lammich for helpful discussions, and [24] M. Ayouz et al., Phys. Rev. A 83, 052712 (2011).

193003-4