Near-infrared spectroscopy of LiNH 3 : First observation of the electronic spectrum

Electronic spectra of LiNH3 and its partially and fully deuterated analogues are reported for the first time. The spectra have been recorded in the near-infrared and are consistent with two electronic transitions in close proximity, the Ã2E−X̃2A1 and B̃2A1−X̃2A1 systems. Vibrational structure is seen in both systems, with the Li–N–H bending vibration (ν6) dominant in the ÃE−X̃A1 system and the Li–N stretch (ν3) in the B̃A1−X̃A1 system. The prominence of the 60 band in the ÃE−X̃A1 spectrum is attributed to Herzberg–Teller coupling. The proximity of the B̃A1 state, which lies a little more than 200 cm−1 above the Ã2E state, is likely to be the primary contributor to this strong vibronic coupling. © 2011 American Institute of Physics. [doi:10.1063/1.3570824]


I. INTRODUCTION
All of the alkali metals dissolve in liquid ammonia. 1 The resulting solutions have properties that depend strongly on the alkali/ammonia mole ratio.At high alkali concentrations, the solutions have a shiny metallic appearance and show high electrical conductivities consistent with metalliclike behavior.In contrast, dilute solutions of the alkali in liquid ammonia are electrically insulating and show a deep blue color.The electrical and optical properties of these solutions have long been attributed to the valence electron of the alkali metal, which can detach from the atom and enter the liquid ammonia to form a solvated electron when the solution is sufficiently dilute.It is transitions of these solvated electrons that are responsible for the color of the solution and the absorption wavelength is determined by the precise environment in which the electron is located, which in turn depends upon the alkali/ammonia ratio.
The study of clusters of alkali metal atoms with ammonia molecules in the gas phase provides a means of probing the impact of the solute/solvent ratio on the electronic properties in a system of finite size.The latter is particularly valuable because small complexes are, in principle, amenable to high quality ab initio calculations, which can then be compared directly with experimental findings. 2The smallest, and ostensibly the simplest, alkali-ammonia cluster is LiNH 3 .Any attempt to understand small alkali-ammonia clusters, and to extrapolate their properties as a function of cluster size, should include this most fundamental complex.However, the only previous experimental studies of LiNH 3 have been carried out in solid argon matrices. 3,4 sing infrared spectroscopy, the fundamental frequencies of several vibrational modes of LiNH 3 and its deuterated analogues were determined.a) Author to whom correspondence should be addressed.Electronic mail: andrew.ellis@le.ac.uk.Telephone +44 (0)116 252 2138; Fax +44 (0)116 252 3789.b) Electronic mail: tim.wright@nottingham.ac.uk.Telephone +44 (0)115 846 7076; Fax +44 (0)115 951 3562.
Here we report the first observation of the electronic spectrum of LiNH 3 .This is part of a series of investigations of alkali/ammonia complexes and complements our earlier spectroscopic work on the ground electronic states of these complexes. 5,6 he only previous study of the electronic spectroscopy of Li(NH 3 ) n complexes was a recent investigation of Li(NH 3 ) 4 originating from our laboratory, which employed photodissociation spectroscopy. 7Very recently, this work has been complemented by a velocity map imaging study to explore the photodissociation dynamics in the first excited electronic state of Li(NH 3 ) 4 . 8More extensive spectroscopic work has been carried out on small Na(NH 3 ) n complexes and the first such study was reported by Nitsch and co-workers, 9 who successfully obtained spectra for NaNH 3 .This and subsequent work by Schulz and co-workers, 10 and also by Rodham and Blake, 11 on both NaNH 3 and NaND 3 , has yielded resolved vibrational structure.Extension to larger complexes, with up to n = 6, was achieved by Brockhaus and co-workers, 12 but the spectra were broad and unresolved, unlike that of NaNH 3 .The transitions observed for the n = 1−6 complexes all occur in the near-infrared and correlate asymptotically with the 3p ← 3s transition of atomic sodium.One of the defining features of the NaNH 3 spectrum is strong vibronic coupling, as evinced by intense bands originating from excitation of a low frequency degenerate bending vibration, whose observation should be symmetry forbidden in the Franck-Condon limit for a molecule of C 3v symmetry.The origin of this vibronic coupling, whether a Jahn-Teller effect or Herzberg-Teller coupling, or indeed a combination of the two, has not been determined.Predictions of the vertical transition energies to low-lying excited electronic states of both LiNH 3 and NaNH 3 have recently been published, 13,14 but neither geometries nor vibrational frequencies were reported for these complexes in their excited electronic states.
Electronic spectra of both LiNH 3 and its partiallyand fully-deuterated analogues are presented in the current study.These spectra were successfully obtained using both two-color resonance-enhanced multiphoton ionization (REMPI) and a photodepletion technique.The spectra show a number of features, which appear to be vibrational or vibronic in origin, and their assignment is discussed in the light of results obtained for partially-and fully-deuterated molecules.

II. EXPERIMENTAL
6][7] Briefly, LiNH 3 complexes were produced by laser ablation of a solid lithium target in the presence of gaseous NH 3 and the resulting mixture was expanded into vacuum to form a supersonic jet.The central portion of the jet was extracted by a skimmer and the resulting molecular beam passed into the source region of a time-of-flight mass spectrometer.To gain some information on the species formed during the expansion, single-photon ionization in the near-UV was employed using the output from a pulsed dye laser.The dominant ions present were LiNH 3 + and Li(NH 3 ) 4 + , although significant quantities of other Li(NH 3 ) n + ions were also observed.
To record optical spectra, the output from a LaserVision optical parametric oscillator/amplifier (OPO/A) was employed.This system was pumped by the output from an injection-seeded Nd:YAG laser (Surelite II-10), although the injection seeder was not essential for the experiments described here, since it did not affect the observable resolution.Since all of the spectra were recorded in the near-infrared, only the OPO part of the OPO/A system was used.This gave a wavelength-tuneable output capable of pulse energies of up to 15 mJ, although the pulse energy incident on the molecular beam was typically 0.5 mJ.The beam from the OPO was gently focused into the source region of the mass spectrometer and aligned such that it spatially overlapped the UV laser beam.The delay between the firing of the two laser pulses was controlled using a delay generator.
Both two-color (1+1 ) REMPI and photodepletion spectroscopy were used to record spectra, with the latter yielding much higher quality spectra, as detailed later.
Ammonia was obtained from a standard liquid ammonia cylinder (BOC Gases, 99% purity) and was used without purification.To assist with spectral assignments some experiments were also carried out using ND 3 (Sigma Aldrich, 99% D).

III. COMPUTATIONAL DETAILS
Ab initio calculations on LiNH 3 have been carried out in support of the experimental work.We have previously reported ab initio calculations on the ground electronic state of LiNH 3 at levels up to and including UCCSD(T) with an aug-cc-pVQZ basis set. 15These high level calculations were used to predict both the vibrational frequencies and the Li-N bond dissociation energy.The aims of the new calculations were twofold: (1) to predict vibrational isotopic shifts for the ground state and (2) to explore the low-lying excited electronic states.For the ground state calculations, UMP2/6-311++G** calculations were initially performed using the GAUSSIAN 03 program. 16Spin contamination was found to be negligible, as shown by the calculated value of <S 2 > = 0.7501.The geometry was optimized and harmonic vibrational frequencies calculated at the potential energy minimum.In addition, RCCSD(T) calculations were also performed, with again both geometry optimization and harmonic frequency calculations being undertaken: these calculations employed MOLPRO 2008. 17For N and H, standard Dunning aug-cc-pVTZ basis sets were employed.For Li, the cc-pVTZ basis set was extended by adding a single set of diffuse s, p, and d basis functions, extrapolated in an even-tempered fashion from the existing basis functions with the smallest exponents.
Hashimoto and Daigoku have recently computed the vertical transition energies to low-lying electronic states of LiNH 3 using multireference calculations (MRCI-SD). 14In the current study, we undertook a series of CASSCF and CASSCF+MRCI calculations.For the CASSCF calculations, the basis set was as described above for the RCCSD(T) calculations, while for the CASSCF+MRCI calculations, we employed aug-cc-pVDZ basis sets for N and H and the cc-pVDZ basis set for Li.MOLPRO 2008 (Ref.17) was used for all of the CASSCF and CASSCF+MRCI calculations.The aim was to obtain excitation energies, as well as geometries and vibrational frequencies for the ground and excited states.These calculations proved to be far from straightforward, especially with regard to calculation of the vibrational frequencies.
Initially, the strategy adopted was to use state-averaged CASSCF, with the triple-ζ basis set mentioned above.The four lowest energy roots arising from the interaction of Li with NH 3 were considered, where Li had the unpaired electron in either the 2s or 2p orbitals.As will be discussed in more detail later, these four roots correspond to the X 2 A 1 , Ã2 E, and B2 A 1 states, where C 3v symmetry is assumed in these labels.Both the Li and N 1s orbitals were kept doubly occupied during this procedure, although their coefficients were allowed to vary.Analytic gradients were employed for optimizing the geometry, by using the RS2 procedure within MOLPRO,17 but not allowing any excitations; however, the second derivatives were calculated numerically and this required convergence at all displaced geometries.Unfortunately, even with small step sizes in the numerical gradient calculation, the vibrational frequencies for the 2 E state came out with imaginary frequencies and severe nondegeneracy of the vibrational modes with nominal e symmetry (although the geometries and the energies of the two components of this state were essentially identical).Indeed, the two components of the 2 E state, calculated separately, did not give vibrational frequencies in even approximate agreement with each other.Even for the nondegenerate X and B states, nondegeneracy in the e vibrations was observed, even when C 3v symmetry was imposed.Similar results were obtained when each root was considered separately in single-reference CISD calculations.
Further calculations were then undertaken where multireference CI was performed on the CASSCF wavefunctions obtained with the double-ζ basis set.It was found that the values obtained for the X state were now in good agreement with   those from the RCCSD(T)/aug-cc-pVTZ calculations.There were, however, still nondegenerate and imaginary vibrational frequencies for the Ã2 E state.As a consequence, we have been unable to obtain vibrational frequencies for the 2 E state.
The same set of calculations did, however, yield a full set of real vibrational frequencies for the X 2 A 1 state and the B2 A 1 states, and which also showed the expected degeneracies, and the results of this latter set of calculations are discussed below.

A. Ab initio calculations
We begin with a discussion of the results from the ab initio calculations, since these provide information that has proved to be important in achieving a spectral assignment.The minimum energy structures in all three electronic states of LiNH 3 , derived from CASSCF+MRCI calculations, are summarized in Table I.For comparison purposes, we also include structures for the ground states of both LiNH 3 and LiNH 3 + obtained from CCSD(T) calculations.Caution is needed in interpreting the theoretical findings for the Ã2 E state since, as mentioned earlier, some imaginary vibrational frequencies were calculated for the two components of this state.Nevertheless, in all of the neutral and ionic states the minimum energy structure was found to correspond to C 3v symmetry.According to the calculations the minima of the Ã2 E and B2 A 1 states are separated by <350 cm −1 , a point of significance which will be encountered again later.It should also be noted that the calculated adiabatic ionization energy obtained from these calculations, 4.32 eV (inclusive of zero-point vibrational energy), is in excellent agreement with the experimentally determined value of 4.339 ± 0.003 eV. 18n important take-home message from these calculations is that the change in structure in moving from the X 2 A 1 state to the Ã2 E state is small, whereas excitation to the B2 A 1 state leads to a much more significant change in structure, specifically in the Li-N bond length.Indeed the Li-N bond is found to be even shorter in the B2 A 1 state than in the cation, which would not be expected if the unpaired electron was simply a nonbonding electron.In fact in the limit of a nonbonding electron the Ã2 E state should possess the shortest bond, since the orientation of the unpaired electron, which would now lie in orbitals that correlate with the 2p x,y orbital pair on the Li atom, would expose the N atom to an increase in effective positive charge on the Li atom.By way of contrast, in the B2 A 1 state the Li-N bond length would be expected to be longer than that in the X 2 A 1 state, since increased electron-electron repulsion would arise from the unpaired electron residing in what is approximately a 2p z orbital pointing directly at the N atom.This simple electrostatic picture is clearly deficient for LiNH 3 and suggests that the unpaired electron has a significant covalent bonding role, particularly in the B2 A 1 state.
Under C 3v symmetry LiNH 3 will have six distinct vibrational frequencies and Table II shows the ab initio predictions alongside experimental values for the ground electronic state of LiNH 3 derived from IR matrix isolation spectroscopy. 3,4 though the argon matrix will perturb the vibrations, the effect on the vibrational frequencies is normally small and the shifts relative to the gas phase tend to be a few cm −1 at most.Confirmation that the matrix IR assignments are reasonable is provided by the relatively good agreement with our ab initio vibrational frequencies.The only exception to this statement is for mode ν 3 , the Li-N stretch, where a significant discrepancy between theory and experiment is seen.The same conclusion has been reached in a recent DFT study of MNH 3 complexes, 19 where M = Li-Fr, and therefore the matrix assignment of the Li-N stretching frequency in LiNH 3 looks to be erroneous.Also, the corresponding data for LiND 3 are shown in Table II.The effect of deuteration in the matrix experiments is faithfully reproduced by the calculated vibrational frequencies, a fact that will become important later in the spectral assignment process.
Table III compares the vibrational frequencies calculated for the X 2 A 1 state with those obtained for the B2 A 1 state of LiNH 3 and the X 1 A 1 state of LiNH 3 + .There is good agreement between the CASSCF/MRCI and RCCSD(T)

B. Survey scan for LiNH 3
Initial attempts to observe the near-infrared electronic spectrum of LiNH 3 were made using photodepletion spectroscopy, as used for other Li(NH 3 ) n complexes in our laboratory. 5,7 n the case of LiNH 3 this technique involves monitoring the LiNH 3 + ion signal produced by laser photoionization at a wavelength of 283 nm, while scanning the wavelength of the OPO.In order to register IR absorption, this process must induce fragmentation of the neutral, which then leads to a reduction in the monitored LiNH 3 + signal; consequently, IR absorption is registered by dips in this ion signal.UCCSD(T) calculations of the Li-N bond dissociation energy of LiNH 3 (4430 cm −1 ) 15 suggest that this energy is easily exceeded by the photon energies used in the present work (see below).Successful observation of a depletion spectrum is also determined by intramolecular decay dynamics, since some of the energy from the initial IR absorption must move into the Li-N stretching vibration, such that Li-N fission can occur on the time scale of the experiment.Since we have electronic excitation of LiNH 3 , and the dissociation process must take place on the ground state potential energy surface, the rate of internal conversion is critical.
The spectrum in Fig. 1 is a survey scan of LiNH 3 obtained in photodepletion mode.This spectrum covers the region from 11300 to 13600 cm −1 and consists of several wellresolved bands accompanied by a number of weaker and more congested features on the high frequency side of the spectrum.All of the bands are relatively broad, with full widths at half maximum of approximately 40 cm −1 .The linewidth of the OPO is approximately 3 cm −1 , so the widths of these bands must be dominated by other factors.We can rule out power broadening, since experiments using reduced OPO pulse energies showed no significant decrease in observed bandwidths.Consequently, the most likely source is lifetime broadening, most probably caused by rapid internal conversion following laser excitation.As a result, no rotational information is obtained from the current work.
One problem with depletion spectroscopy is that laserinduced dissociation of higher mass complexes can sometimes lead to contributions to the ion signal in a particular lower mass channel.For example, depletion of Li(NH 3 ) 2 to produce LiNH 3 will ultimately lead to an increase in LiNH 3 + signal on photoionization.This leads to ion signal enhancements which would be seen as structure in the spectrum pointing in the opposite direction to the photodepletion bands seen in Fig. 1.From the low signal levels of the corresponding cations in the mass spectrum, we conclude that Li(NH 3 ) 2 and Li(NH 3 ) 3 are more than an order of magnitude less abundant than LiNH 3 in the gas mixture.Furthermore, the absence of any absorption features due to excess ion production, which would produce bands extending below the baseline level in Fig. 1, suggests that cascade effects from larger clusters are negligible.Nevertheless, these observations do not definitively rule out contributions from one or more larger complexes, since there is some possibility, albeit small, of coincident positive-and negative-going signals canceling out each other.
Firm proof that the spectrum arises only from LiNH 3 was obtained by recording the two-color REMPI spectrum in FIG. 1. Electronic spectrum of LiNH 3 in the near-infrared obtained using photodepletion spectroscopy.This spectrum was constructed from two separate scans, which are joined at 12250 cm −1 .See text for details of the assignment.As detailed in the text, the signal/noise ratio for REMPI spectra was found to be inferior to photodepletion spectra.
the LiNH 3 + mass channel.The wavelength of the frequency doubled dye laser was set at 288 nm introducing a photon with an energy just below the ionization threshold of LiNH 3 (4.334eV or 285.7 nm). 18When the IR output was added and the wavelength scanned, a REMPI spectrum was observed in which all of the main bands seen for LiNH 3 in the photodepletion mode were also obtained in the REMPI spectrum.However, the signal/noise ratio for this spectrum was not as good as that recorded by photodepletion, as can be seen in Fig. 2, which shows part of the REMPI spectrum.The lower quality of the REMPI spectrum presumably arises because the decay of the excited electronic state of LiNH 3 is so rapid, which is also consistent with the broad but relatively intense photodepletion bands.

C. Spectral assignments
The positions of the bands observed in the electronic spectra of LiNH 3 and LiND 3 are summarized in Table IV.To assist with the assignment, we can gain useful clues from previous work, including matrix isolation IR studies of LiNH 3 ,  b The numbers in parentheses show the positions of the bands relative to the respective assigned electronic origin transition.
previous work on the electronic spectroscopy of NaNH 3 , and ab initio calculations, including our own.We will begin by discussing the low frequency part of the spectrum in Fig. 1.We will start with LiNH 3 and will then subsequently draw on a comparison with its partially-and fully-deuterated analogues to check the consistency of the assignments made.The band observed at the lowest wavenumber in the spectrum of LiNH 3 is at 11423 cm −1 .Scans further to the red, extending down to 10500 cm −1 , found no other identifiable bands suggesting that this band is likely to be an electronic origin (0 0 0 ) transition.The Ã2 E− X 2 A 1 transition is expected to correlate with the strongly allowed 2p ← 2s transition of atomic lithium.In the free Li atom this transition takes place at 14903 cm −1 and so in LiNH 3 it is shifted substantially to the red by the presence of the NH 3 group.0][11][12] Assuming C 3v point group symmetry for LiNH 3 , the Ã2 E state will correlate with the 2p x,y orbitals and the B2 A 1 state has the Li 2p z orbital as its principal contributor.It is plausible that transitions to both excited electronic states could be observed in our spectra if the states are close enough in energy.
The two lowest frequency vibrations of LiNH 3 are, by a large margin, the Li-N stretch (ν 3 ) and the Li-N-H bend (ν 6 ).If we assume that the lowest frequency band, at 11 423 cm −1 , originates from the Ã2 E− X 2 A 1 0 0 0 transition, then we cannot achieve a sensible band assignment for the three bands immediately above this nominal origin band based on a single electronic transition, even invoking the nontotally symmetric (e symmetry) ν 6 bending mode (also known as the rocking mode).This suggests a contribution from two distinct electronic transitions, the Ã2 E− X 2 A 1 and B2 A 1 − X 2 A 1 transitions.To gain further information, we also recorded spectra for LiNH 2 D, LiNHD 2 , and LiND 3 .Fig. 3 shows a comparison of the spectra of these isotopologues with that of LiNH 3 .
We begin our assignment with the aforementioned assumption that the lowest frequency band, at 11423 cm −1 , corresponds to the Ã2 E− X 2 A 1 0 0 0 transition.We can then use the isotopic shifts, along with other information, to assign vibrational features in the spectrum.The ratio of frequencies in a given vibrational mode for the nondeuterated versus fully deuterated molecule should be approximately retained from one electronic state to another.Our calculations for the X 2 A 1 and B2 A 1 states confirm this, with the vibrational frequency ratios for the two states agreeing to within 1% for all of the vibrational modes.Consequently, we expect to be able to predict a reliable estimate of this ratio in the Ã state, even though ab initio vibrational frequency calculations on the Ã state were unsuccessful, by making use of the ab initio data for the X state, which has a very similar equilibrium geometry, as noted above.The assignment of the third lowest energy band of LiNH 3 , at 12035 cm −1 , to the 6 1 0 transition, leads to a response to deuteration in line with expectations (experimental ν 6 D/H ratio is 0.79 versus 0.76 from theory).Furthermore, we have carried out additional ab initio calculations on the various isotopologues of LiNH 3 in its X state, which support this assignment.As can be seen from Table V, the calculations show that the degenerate 6 1 0 band should split into two distinct bands in LiNH 2 D (an Li-N-H bend and an Li-N-D bend).Accordingly, we see in the spectra a clear splitting of the This results in continuously changing isotopologue ratios as the spectral scan progresses and so these spectra are intended only to show correlations in band positions on isotope substitution: the intensity ratios will not be reliable.
tentatively assigned 6 1 0 feature into two bands, with an approximate separation of 60 cm −1 , in going from LiNH 3 to LiNH 2 D (see Fig. 2).The ab initio calculations predict a smaller splitting for LiNHD 2 and a significant shift to the red for the two resulting features.In agreement with this prediction we see a clearly broadened, but unresolved, band in Fig. 3 which shows a notable redshift relative to the 6 1 0 band of LiNH 3 .The bending mode assignment is further confirmed by the additional redshift seen on full deuteration to give LiND 3 .
Immediately beyond the 6 1 0 band is a second band which is also redshifted on deuteration.The only other low frequency vibration which could possibly be responsible for this band is the ν 3 vibration, where ν 3 is the Li-N stretch.Unlike the 6 1 0 transition, which is nominally forbidden on the basis of simple symmetry arguments, the 3 1 0 transition is symmetrically allowed in the Franck-Condon limit.The absence of any splitting of this band on deuteration is consistent with this vibrational assignment.This band could be assigned to the 3 1 0 transition in the Ã2 E − X 2 A 1 system, but this would require a huge increase in the frequency of ν 3 , from roughly 440 to 739 cm −1 .Not only do the molecular structure changes predicted for the Ã2 E − X 2 A 1 transition (see Table I) not support this large change in ν 3 but also a comparison with the corresponding band in LiND 3 gives a drastically different ratio of vibrational frequencies (0.87) when compared with the ab initio prediction (0.98).
To achieve a sensible assignment, we attribute the bands at 11 656 and 12 035 cm −1 to the B2 A 1 − X 2 A 1 0 0 0 and 3 1 0 transitions, respectively.Comparison with the corresponding transitions seen for LiND 3 now gives a fully deuterated/nondeuterated ν 3 frequency ratio of 0.98, in perfect agreement with the value predicted from ab initio calculations.In addition, the frequency extracted for ν 3 in the B2 A 1 state of LiNH 3 , 506 cm −1 , is quite close to the MRCI harmonic estimate of 560 cm −1 (see Table III).
The conclusions drawn so far, based on the first four bands in the LiNH 3 spectrum, are that there are two electronic transitions, the Ã2 E− X 2 A 1 and B2 A 1 − X 2 A 1 transitions, in close proximity.Specifically, we obtain a separation of only 233 cm −1 between the two electronic origins.Our CASSCF/MRCI calculations predict a T e separation of 343 cm −1 , which is certainly consistent with the experimental separation given the margins of error that are typical for such excited state transition energy calculations.Previous calculations by Hashimoto and Daigoku using MRCI methodology predicted a separation of vertical transitions of around 1200 cm −1 , 14 which again is reasonably close to the observed separation in view of the likely margin of error in those calculations.Furthermore, Hashimoto and Daigoku predicted almost identical oscillator strengths for the two electronic transitions, which is in accord with the similar band intensities seen in the spectrum in Fig. 1.It is also interesting to note the effect of deuteration on the electronic origin transition energies.For a transition where the binding is stronger in the excited electronic state, deuteration would normally shift the origin of transition to the red.This is indeed the case for the B − X system, where both the ν 3 vibrational assignment and the ab initio predictions point to a marked strengthening of the Li-N bond on electronic excitation.Interestingly, the Ã − X 0 0 0 transition shows a modest blueshift on deuteration.A small blueshift has also been reported previously for the Ã2 E− X 2 A 1 0 0 0 transition of NaNH 3 and this was tentatively explained by the suggestion that the binding in the NH 3 entity was weakened by electronic excitation. 10The ab initio calculations carried out in the present work do not provide any specific support for this suggestion in the case of LiNH 3 , but these calculations were hampered by an inability to calculate vibrational frequencies for the Ã2 E state.
Having found ν 6 structure in the Ã2 E− X 2 A 1 system, we might reasonably look for evidence of excitation of the other low frequency vibration, ν 3 .However, we see no additional low frequency features, which would allow such an assignment.The conclusion we draw is that either the ν 3 structure is hidden beneath existing bands, e.g., the 3 1 0 and 6 1 0 bands coincide, or the 3 1 0 band is simply very weak and is therefore not observed.The latter suggestion is plausible given the negligible change in the Li-N bond length predicted by the ab initio calculations (see Table I), which would confer a small Franck-Condon factor to the 3 1 0 transition.Likewise, in the B2 A 1 − X 2 A 1 manifold we have successfully identified structure due to ν 3 but not due to ν 6 .However, single quantum excitation of ν 6 is forbidden in the absence of vibronic coupling (see Sec. IV D) and we would not expect any first-order vibronic coupling in the B2 A 1 state.
To close this section on band assignments, we turn to the collection of weaker bands in the higher frequency section of the spectra.There is a trio of bands lying between 12 500 and 13 000 cm −1 in Fig. 1.An unambiguous assignment for these bands is not possible because more than one plausible transition can contribute.Furthermore, the data from partly and fully deuterated molecules are less definitive for these and higher bands because of their low intensities and the increased congestion.Nevertheless, we have viable suggested assignments which are summarized in Table IV.A totally symmetric vibration that could be active in this region is the NH 3 umbrella mode, ν 2 .Bands have been assigned to 2 1 0 in both the Ã − X and B − X systems for both LiNH 3 and LiND 3 .On the basis of the Ã − X 2 1 0 assignment, the frequency of ν 2 in the Ã state is 1173 cm −1 , which is close to the calculated ground state value of 1194 cm −1 (Table II; RCCSD(T) value).The D/H ratio of vibrational frequencies is 0.75, which is very similar to that expected for the electronic ground state (0.76) based on ab initio calculation lending further credence to this assignment.The corresponding ν 2 frequency in the B state is 1152 cm −1 , which is significantly below the 1294 cm −1 prediction from the CASSCF/MRCI calculations (Table III).Furthermore, the experimental D/H ratio is 0.83 suggesting that the B − X 2 1 0 assignment in LiNH 3 or LiND 3 , or perhaps both, may not be sound.
There are additional, weak, bands above 13 000 cm −1 , but there is major congestion in this region and, therefore, specific assignments would be highly speculative.However, we note that several transitions could plausibly lie within this region, including the B − X 3 3 0 band and combination bands such as the Ã − X 2 1 0 6 1 0 band.

D. Vibronic coupling
The observation of activity in ν 2 and ν 3 in the spectrum of LiNH 3 is not unexpected, since these vibrations are totally symmetric.More surprising is the prominence of the 6 1 0 band in the Ã2 E − X 2 A 1 system, since mode ν 6 is nontotally symmetric in the C 3v point group and its excitation is therefore symmetry-forbidden at the single quantum level ( v = ±1) in the Franck-Condon limit, if this point group symmetry is maintained in both electronic states.We note that a prominent 6 1 0 band has also been reported in the corresponding spectrum of NaNH 3 and there has been some (unresolved) debate about whether this arises from a Jahn-Teller effect or Herzberg-Teller coupling. 11Evidence is presented here that shows that Herzberg-Teller coupling dominates in LiNH 3 .
The Jahn-Teller effect arises from the coupling of vibrational angular momentum with electronic orbital angular momentum within a degenerate electronic state of a nonlinear molecule.When this coupling is significant, the vibrational quantum numbers for Jahn-Teller active vibrations are no longer good quantum numbers, but they can still be used in an approximate sense as a guide to the expected vibrational behavior.In particular, if Jahn-Teller coupling is prominent, we would expect vibrational selection rules for the active vibration(s) of v = ± 0, ±1, ±2, etc. 20 In effect, Jahn-Teller distortion converts a nominally nontotally symmetric vibration into a totally symmetric one, in the point group of the distorted structure, subject to the normal Franck-Condon selection rules, and thus a progression in this mode is possible.On the other hand, in Herzberg-Teller coupling, the corresponding vibrational selection rule for the active nontotally symmetric vibration is v = ±1, ±3, etc. 21 We cannot use these selection rules to distinguish between Herzberg-Teller and Jahn-Teller coupling in LiNH 3 since the 6 2 0 and 6 3 0 bands would lie in congested regions of the spectrum where other bands are also clearly present.
However, we have other evidence from the present work which points to a dominant role for Herzberg-Teller coupling.In particular, Herzberg-Teller coupling can persist as the vibrational symmetry is lowered, whereas Jahn-Teller coupling cannot, since all vibrational angular momentum is quenched for LiNH 2 D and LiND 2 H. Consequently, no significant ν 6 structure would be expected for LiNH 2 D and LiND 2 H if Jahn-Teller coupling was responsible for the 6 1 0 bands seen in the spectra of LiNH 3 and LiND 3 .Furthermore, although the vibrational symmetry is lowered for LiNH 2 D and LiND 2 H, making all vibrational transitions fully allowed in the Franck-Condon limit even without Jahn-Teller or Herzberg-Teller coupling, observable progressions in what were previously nontotally symmetric vibrations would not be expected because there is no change in the equilibrium structure of the molecule on isotopic substitution.Since we observe the 6 1 0 band just as readily for LiNH 2 D and LiND 2 H as we do for LiNH 3 and LiND 3 , we therefore conclude that Herzberg-Teller coupling is responsible for the appearance of the nominally forbidden 6 1 0 band.The prominence of this band must derive from substantial intensity borrowing from one or more electronically allowed transitions.The vibrational wavefunction for the v 6 = 1 level in LiNH 3 has e symmetry and can potentially mix with the wavefunction of the Ã2 E state to give vibronic states with A 1 and E symmetries.An A 1 vibronic state would now have the correct symmetry to interact with other electronic states of A 1 symmetry and in particular the nearby B2 A 1 state.According to the ab initio predictions by Hashimoto and Daigoku the B2 A 1 − X 2 A 1 electronic transition has virtually the same oscillator strength as the Ã2 E− X 2 A 1 transition. 14onsequently, the "borrowing" of substantial intensity from the B2 A 1 − X 2 A 1 transition is a plausible source of the strong vibronically-allowed 6 1 0 band seen in the Ã2 E− X 2 A 1 system of LiNH 3 .

V. CONCLUSIONS
The electronic spectrum of the prototypical alkaliammonia complex, LiNH 3 , has been recorded for the first time.This spectrum, occurring in the near-infrared, shows bands assigned to both the Ã2 E -X 2 A 1 and B2 A 1 − X 2 A 1 electronic transitions.In particular the origin bands for these two transitions are found to be in very close proximity (∼200 cm −1 ).Vibrational structure in these two band systems has been identified and assigned with the aid of isotope substitution studies and a series of supporting ab initio calculations.
The Ã2 E − X 2 A 1 system is characterized by prominent activity of the Li-N-H bending vibration (ν 6 ) implying strong vibronic coupling in the excited electronic state.Jahn-Teller activity is possible in the Ã2 E state, which could give rise to observation of Franck-Condon-forbidden features, such as the intense 6 1 0 band.However, evidence from isotope substitution experiments suggests that the vibronic activity is a consequence of Herzberg-Teller coupling.In particular, the proximity of the B2 A 1 state to the Ã2 E state provides the opportunity for strong vibronic coupling.

124304- 5 FIG. 2 .
FIG.2.Part of the two-colour (1+1 ) REMPI spectrum of LiNH 3 showing the region from 11400 to 12400 cm −1 .As detailed in the text, the signal/noise ratio for REMPI spectra was found to be inferior to photodepletion spectra.

a
Only bands that have significant signal/noise ratios and are clearly resolved are listed in this table.

FIG. 3 .
FIG. 3. Comparison of the low frequency regions in the depletion spectra of LiNH 3 , LiNH 2 D, LiNHD 2 , and LiND 3 .The spectra for the partly and fully deuterated isotopologues were recorded simultaneously by gating the signals for the respective cations in the mass spectrum.A small quantity of ND 3 was injected into the NH 3 flow allowing the H and D atoms to exchange.This results in continuously changing isotopologue ratios as the spectral scan progresses and so these spectra are intended only to show correlations in band positions on isotope substitution: the intensity ratios will not be reliable.

TABLE I .
Calculated equilibrium parameters for LiNH 3 and LiNH 3 + .
This is the first ionization energy of LiNH 3 excluding zero point energy effects.Inclusion of zero point energy corrections in the neutral molecule and cation gives an adiabatic ionization energy of 34845 cm −1 (4.32 eV).
a Note that these parameters for the cation agree closely with those described in Ref.15, which were also calculated using UCCSD(T) methodology.b

TABLE II .
Vibrational frequencies a (in cm −1 ) of the X 2 A 1 states of LiNH 3 and LiND 3.
a The experimental data are vibrational fundamental frequencies while the calculated values are unscaled harmonic vibrational frequencies.bThesedescriptions describe the approximate character of the vibrational modes.cFrominfrared spectra of LiNH 3 and LiND 3 isolated in an argon matrix (Refs.3 and 4).The values for LiND 3 in parentheses were estimated from a force field analysis (Ref.4).Downloaded 25 Jun 2013 to 143.210.120.177.This article is copyrighted as indicated in the abstract.Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

TABLE III .
Comparison of calculated vibrational frequencies a (in cm −1 ) of the X 2 A 1 and B2 A 1 states of LiNH 3 and the X 1 A 1 state of LiNH 3 + .A 1 state of LiNH 3 .The CASSCF/MRCI calculations also yielded vibrational frequencies for the B2 A 1 state.Of interest, is the very large increase in the frequency of the Li-N stretching vibration, ν 3 , which shifts from 434 to 560 cm −1 on excitation from the X 2 A 1 state of LiNH 3 to the B2 A 1 state.Indeed the calculated frequency for ν 3 in the B2 A 1 state exceeds that in the ground state of the cation, 554 cm −1 , a finding that fits with the shorter Li-N bond for the former species discussed earlier.

TABLE IV .
Band positions a and assignments for LiNH 3 .

TABLE V .
Comparison of the calculated vibrational frequencies of the X 2 A 1 state of LiNH 3 with its partially and fully deuterated analogues a .
a The values not in parentheses are from MP2/6-311++G** calculations, while those in parentheses are from RCCSD(T) calculations using a triple-ζ basis set (see text for details).The harmonic frequencies shown are unscaled.Only those vibrations with frequencies below 2000 cm −1 , the region of interest in the current work, are listed here.b These descriptions give the approximate character of the vibrational modes.