Photoelectron spectroscopy of anthracene and fluoranthene radical anions Free

The delocalized π systems present in polycyclic aromatic hydrocarbon (PAH) molecules give rise to some interesting electronic structure properties. For example, in crystals, some of these PAHs, such as tetracene and pentacene, exhibit efficient singlet fission phenomenon,^1,2 where one molecule with an excited singlet state can share its excitation energy with a neighboring ground state molecule and both are converted to triplet excited states. These π systems can also directly influence molecular reactivity and exhibit resonance states that allow for efficient capture of a free electron.^3,4 Therefore, studies of PAHs to understand their electronic and related structural properties have gained momentum in recent times due to their possible applications in photovoltaics and relevance in astrophysics.^5–12 To provide a more detailed picture of the electronic states within these PAH molecules, we present here a photoelectron (PE) spectroscopy study of the anthracene and fluoranthene radical anions that directly access the neutral molecule singlet and triplet states. Both molecules studied here are composed of three six-membered rings, as shown in Fig. 1. In anthracene, the three rings are fused linearly, whereas fluoranthene can be considered as having a naphthyl covalently bound to a benzyl group. The π systems in both molecules are fully delocalized, and the difference in connectivity provides an interesting comparison.

The vibrational levels of the neutral anthracene $X̃$¹A_g ground state have been well-studied in the condensed and gas phase. The IR and Raman vibrational frequencies in crystals and solutions have been collected in Refs. 13 and 14. In the gas phase, a fluorescence study of the S₁-S₀ transition (origin at 3.43 eV) observed a_g and b_3g symmetry vibrations, the intensities of the latter vibrations are attributed to vibronic coupling involving the S₁ state.¹⁵ Previous PE spectroscopy studies yielded an electron affinity (EA) of 0.53 eV (4275 cm⁻¹)^16,17 and a S₀-T₁ splitting of 1.869 eV (15 074 cm⁻¹),¹⁷ in good agreement with a phosphorescence study of the T₁ state that gave a similar value of 1.851 eV (14 927 cm⁻¹).¹⁸ There are comparably fewer studies on fluoranthene and due to the lower symmetry of this molecule, the assignments of the observed ground state vibrations are less conclusive.^19–24 Its EA has been measured to be 0.63 eV by electron capture processes.^25,26

Anion PE spectroscopy has often been applied in studies of neutral radical species starting from the corresponding closed-shell singlet anions, thus providing a unique access to reactive open-shell systems or parts of the neutral potential energy surface that are otherwise difficult to access. However, the relaxed selection rules of anion PE spectroscopy are also advantageous for studying the complex electronic structures of PAH molecules, as they allow for direct comparisons of singlet and triplet states that are accessed from the same anion ground state. Moreover, breakdowns of Born-Oppenheimer approximation, i.e., vibronic coupling, are directly evidenced by observations of non-totally symmetric vibrations.²⁷ The anion PE spectroscopy method we employ here is the higher-resolution slow electron velocity map imaging (SEVI) variant,^27,28 which provides the resolution necessary to examine the vibronic structures of these two PAH molecules in detail. However, formation of the radical anion precursors necessarily imparts significant internal energies to the anions, which are not always sufficiently removed by the limited cooling effects of supersonic expansions past the ionization region. This often yields spectral broadening and increased spectral congestion,²⁹ limiting the achievable experimental resolution and further complicating the spectral analysis of these already complex systems. Here, we couple a pulsed valve supersonic source to a cryogenic ion trap to further cool these radical anions after their formation and prior to spectroscopic interrogation, yielding vibrationally resolved SEVI spectra of the PAH molecules in the lowest energy singlet and triplet states. The results show evidence of vibronic coupling involving out-of-plane motions in all the electronic states studied.

SEVI spectra were obtained using our homebuilt cryogenic SEVI instrument described in detail previously.³⁰ Briefly, anthracene or fluoranthene was seeded into Ar carrier gas by heating the solid samples to ∼80 °C. The gas mixture, at a stagnation pressure of 200 psi, was expanded into the vacuum chamber through an Even-Lavie pulsed valve operating at 10 Hz. The radical anions were formed via electron impact in the supersonic expansion using a circular filament ionizer floated at −100 V. The anions passed through a skimmer and were transferred via a hexapole ion guide into a 3D quadrupole ion trap held at 10 K by a closed-cycle helium cryocooler. A short pulse of buffer gas consisting of 10% D₂ in helium was introduced into the ion trap to thermalize the anions. After a delay of ∼95 ms to allow evacuation of the buffer gas from the trap volume, the cooled anions were ejected into a time-of-flight mass spectrometer for mass separation. The desired anions were mass-selected via pulsed re-referencing prior to entering the multiplate velocity map imaging (VMI) region,³⁰ where they were intersected with the output of a 10 Hz tunable Nd:YAG pumped optical parametric oscillator (OPO) laser with 5–7 cm⁻¹ linewidth. The photoelectrons were imaged using a pair of 40 mm active area micro-channel plates (MCPs) coupled to a phosphor screen located 32 cm from the laser interaction region and captured by a 2048 × 2048 pixels CMOS camera. The SEVI images presented here were all acquired with the VMI repeller at −500 V. To obtain the SEVI spectrum, the images were circularized,³¹ quadrant symmetrized, inverse Abel transformed, and radially integrated. Anisotropy information on individual features were obtained by determining the β value³² via fitting the most intense part of a feature, typically 5–6 pixels in width, to the equation $Iθ=σ4π1+βP2cos⁡θ$⁠, with P₂ being the second-order Lagrange polynomial. Photoelectron kinetic energy calibrations were carried out using known photoelectron features of S⁻ for images acquired with photon energies <25 000 cm⁻¹ and Br⁻ for >25 000 cm⁻¹. The SEVI spectra are presented with the x-axis corresponding to electron binding energy (eBE), which is the difference between the photon energy and the measured electron kinetic energy. The reported peak position and uncertainty correspond to the fitted Gaussian center and half width at half maximum of the spectral feature in the nearest threshold SEVI spectra. All SEVI spectra used in the analysis are included in the supplementary material.

Geometry optimization and vibrational frequency calculations were carried out at the cam-B3LYP and B2PLYP levels with the def2-TZVP basis set using Gaussian 09.³³ The harmonic frequencies were scaled by 0.986 for B2PLYP and 0.969 for cam-B3LYP, obtained from comparing the calculated benzene vibrational frequencies with experimental values.³⁴ The Franck-Condon (FC) spectra were calculated using ezSpectrum³⁵ using the optimized geometries, vibrational displacements, and scaled frequencies of the Gaussian calculations and include Duschinsky rotations. The FC spectra are aligned to the experimental SEVI spectra at the 0-0 transition. In the case of anthracene, the molecule has a D_2h symmetry, in which the B₁/B₂/B₃ designation depends on which molecular axis is defined as x/y/z. Here we will use the definition utilized by Gaussian calculations, as noted in Fig. 1. Note that this designation differs from earlier notations^36,37 as well as other calculation software. Finally, vibrations in each electronic state are sorted according to their symmetry and numbered in order of decreasing frequencies. This allows for easier comparison of vibrations across the different electronic states. All the calculated geometries and vibrational frequencies are listed in the supplementary material.

The anionic anthracene radical has a D_2h $X̃$²B_3u ground state, with a calculated electron configuration of …σ(a_g)²π(b_3u)²π(b_2g)²π(a_u)²π(b_1g)²π(b_2g)²π(b_3u)¹. The unpaired electron resides in a b_3u orbital as shown in Fig. 1, and the removal of this electron from the singly occupied molecular orbital (SOMO) results in the $X̃$¹A_g singlet ground state (S₀) of the neutral anthracene. The removal of an electron from the HOMO-1 b_2g orbital, also shown in Fig. 1, results in the ³B_1u lowest energy triplet state (T₁) of the neutral molecule.

The SEVI spectra associated with the neutral S₀ state acquired using 5986 cm⁻¹ and 7994 cm⁻¹ energy photons are shown in Fig. 2(a). Peak a is the most intense feature in the spectra, and we do not observe any transition with lower eBE. Hence, peak a is assigned as the 0-0 origin band, leading to an observed EA of 4290(24) cm⁻¹ [0.532(3) eV], in good agreement with previous PE spectroscopy results^16,17 and the calculated EA at both levels, as listed in Table I.

The SEVI spectra also contain six other major features at higher eBEs, labeled b-g. The cam-B3LYP calculated FC spectrum, compared in Fig. 2(a), has a similar appearance to the SEVI spectra, and the FC active modes are in good agreement with the observed experimental frequencies. This indicates that most of the observed vibrational activities are due to the differences in geometry between the anion and neutral ground states (the B2PLYP FC spectrum has a similar appearance and is included in the supplementary material together with geometry details). The major FC active modes are vibrations with a_g symmetry, and the numbering in Fig. 2 corresponds to only the 12 a_g vibrations in anthracene. The overall agreement between calculation and experiment allows for assignments of most of the observed SEVI features, listed in Table I. The most active vibration is (a_g)v₁₂, a symmetric in-plane skeletal distortion mode, giving rise to peak b at eBE = 4681(21) cm⁻¹ (+391 cm⁻¹ from origin) and peak d at eBE = 5070(30) cm⁻¹ (+780 cm⁻¹ from origin). Peak f at +1267 cm⁻¹ is assigned to the (a_g)v₇ vibration and peak g at +1416 cm⁻¹ is assigned to the (a_g)v₆ vibration; both vibrations are ring stretching modes with (a_g)v₇ also having CH bending motions. The experimental frequencies observed here are all within 5 cm⁻¹ of those determined previously.¹⁵ 

There are two additional features which are clearly present in the experimental spectrum but absent in the FC simulation. These are peak c at +728 cm⁻¹ and peak e at +877 cm⁻¹. The nominally allowed modes in PE spectroscopy are the totally symmetric modes, i.e., either an a_g vibration or two quanta of non-a_g vibrations, such as (a_g)ν₁₀ at 753 cm⁻¹ or two quanta in (b_3u)ν₅ at 380 cm⁻¹, (b_3u)ν₄ at 463 cm⁻¹ or (b_1g)ν₅ at 477 cm⁻¹.^15,38–40 None of these possible assignments offer a particularly great fit with peak c or peak e. Moreover, their presence would mean either the anion or the neutral geometry is not accurately described by theory, resulting in an otherwise FC active mode to have no intensity. Should this be the case, we would expect to observe greater differences between simulation and experiment in the other a_g vibrations as well. Another possibility is that peaks c and e correspond to fundamentals of non-a_g vibrations which gain intensity via vibronic coupling with another state, such as intensity borrowing via Herzberg-Teller coupling. Based on their frequencies, the best tentative assignment of these features would the (b_3u)ν₃ at 726 cm⁻¹ and (b_3u)ν₂ at 874 cm⁻¹.³⁹ Both of these vibrations are C–H out-of-plane bending modes. Note that there are other possible assignments to (b_2g)ν₃ at 896 cm⁻¹, (a_u)ν₂ at 858 cm⁻¹, or (a_u)ν₃ at 743 cm⁻¹,¹³ all of which are out-of-plane modes, but the b_3u modes offer the best fit.

The SEVI spectra associated with the neutral triplet T₁ state acquired using 21 500 cm⁻¹ and 22 503 cm⁻¹ energy photons are shown in Fig. 2(b). The most intense feature here, peak h at eBE = 19 387(25) cm⁻¹, is again assigned to the 0-0 origin band. This provides an experimentally measured S₀-T₁ adiabatic splitting of 15 097(25) cm⁻¹ [1.872(3) eV], in good agreement with previous experiments.^17,18 The B2PLYP calculated value for this transition is at 19 447 cm⁻¹, in excellent agreement with the experiment, and the cam-B3LYP value shows only a little more deviation. Compared to the S₀ state, the T₁ SEVI spectra display lower FC activities, indicating a smaller change in geometry between the anion ground state and the lowest energy triplet state. This is consistent with the calculated FC spectrum predicting an origin band at least three times more intense than any other feature, as shown in Fig. 2(b).

We will focus on the assignment of the five main features labeled i-m. Peaks k, l, and m are in good agreement with the predicted FC active (a_g)ν₇ CH bend, (a_g)ν₆ ring stretch, and (a_g)ν₅ ring stretch and CH bend vibrations. Table I shows that the frequencies of these SEVI features are in good agreement with previously observed a_g vibrations in the T₁ state.⁴¹ However, calculated frequencies show larger deviations from the experimental values than those for the S₀ state, hinting that either the calculated geometry or the harmonic approximation has more inaccuracies for the T₁ state.

Similar to the S₀ state, there are two features in the T₁ SEVI spectra that are not captured by the FC simulations. These are peak i at +516 cm⁻¹ and peak j at +711 cm⁻¹. The best assignment for peak j is the (b_3u)ν₃ mode, calculated to be at 712 cm⁻¹, and previously observed at 719 cm⁻¹.⁴² This is the same out-of-plane CH bend (b_3u)ν₃ mode that is active in the S₀ state SEVI spectra. For peak i, calculations point to three possible assignments: the (a_u)ν₄ out-of-plane ring deformation at 521 cm⁻¹, the (b_2g)ν₅ out-of-plane ring deformation at 542 cm⁻¹, and the (b_3g)ν₁₀ in-plane ring deformation at 518 cm⁻¹. Given the uncertainties in the calculated frequencies in the triplet state, it is not possible to make a more definitive assignment here.

The analysis of anthracene SEVI spectra reveals that FC simulations work well in predicting the presence of totally symmetric modes, but in addition to those vibrations, there is a clear presence of vibronic coupling involving out-of-plane modes in both the S₀ and T₁ states. The assignments of the anthracene SEVI spectra are aided by the wealth of previous vibrational studies in condensed and gas phase. What we learned here will be used to help the analysis of the fluoranthene results.

The anionic fluoranthene radical has a C_2v $X̃$²A₂ ground state with a calculated electron configuration of …σ(a₂)²π(b₁)²π(a₂)²π(b₁)²π(b₁)²π(a₂)²π(a₂)¹. The unpaired electron resides in an a₂ orbital, as shown in Fig. 1, and the removal of this electron results in the $X̃$ ¹A₁ singlet ground state (S₀) of the neutral fluoranthene. The SEVI spectra associated with the S₀ state acquired using 7994 cm⁻¹ and 8888 cm⁻¹ energy photons are shown in Fig. 3(a). Similar to anthracene, the lowest energy eBE feature, peak a, is also the most intense feature in the spectra. It is assigned to the 0-0 origin band, leading to an observed EA of 6108(15) cm⁻¹ [0.757(2) eV], higher than previous estimation from electron capture processes.^25,26 The excess electron in the radical anion is well delocalized throughout the entire molecule, as the MO in Fig. 1 shows, and this larger π system leads to the higher EA observed for fluoranthene compared to anthracene. The experimentally determined EA is in good agreement with the values of 5611 cm⁻¹ and 6377 cm⁻¹ calculated at the B2PLYP and cam-B3LYP levels, respectively.

The SEVI spectra also contain six other major features at higher eBE, labeled b-g. The cam-B3LYP calculated FC spectrum, compared in Fig. 3(a), has a similar appearance to the SEVI spectra, and the FC active modes with a₁ symmetry are in good agreement with the observed experimental frequencies. This generally good agreement allows us to assign most of the main features except peak b, and the assignments are listed in Table II. Peaks e (+1425 cm⁻¹), f (+1461 cm⁻¹), and g (+1610 cm⁻¹) are assigned to (a₁)ν₁₀, (a₁)ν₉, and (a₁)ν₆ ring stretch and CH bend vibrations, with (a₁)ν₁₀ and (a₁)ν₆ mostly localized to the naphthyl half of the molecule. These frequencies are in good agreement with previous observations.²² Peak d at +1382 cm⁻¹ is assigned to the (a₁)ν₁₂ ring stretch and CH bend vibration based on the good agreement with the calculated frequency and intensity. Peak c at +565 cm⁻¹ is assigned to the (a₁)ν₂₃ skeletal deformation, the frequency is again in good agreement with previous observations.²⁰ 

The relatively intense peak b at +426 cm⁻¹ exhibits a distinctively different anisotropy compared to the origin band and the FC allowed transitions; also see the larger image in the supplementary material. This observation allows us to definitively attribute this peak to a transition gaining intensity through Herzberg-Teller coupling. There are two possible assignments, the (b₁)ν₉ and (a₂)ν₉ vibrations, which have been previously observed in solid state at 427 cm⁻¹ and 429 cm⁻¹.²¹ Both of these vibrations are out-of-plane ring deformation modes, with the former mostly localized to the benzyl half of the molecule. Although these two vibrations are distinguished by their different symmetries, it is not clear which upper state is involved in the vibronic coupling, and hence we cannot make a more definitive assignment here.

The removal of an electron from the HOMO-1 a₂ orbital, shown in Fig. 1, results in the lowest energy triplet T₁ state of the neutral fluoranthene. The SEVI spectra associated with this transition acquired using 26 996 cm⁻¹ and 28 495 cm⁻¹ energy photons are shown in Fig. 3(b). The most intense feature here, peak h at eBE = 24 829(15) cm⁻¹, is assigned to the 0-0 origin band. This provides an experimentally measured S₀-T₁ splitting of 18 721(15) cm⁻¹ [2.321(2) eV]. The calculations for this state are more problematic. The B2PLYP calculations yielded a C_2v ³A₁ state, but with an unphysically large 20 442 cm⁻¹ in-plane ring stretching b₂ vibration that has the largest displacements on the naphthyl half. On the other hand, the cam-B3LYP calculations yielded a C_s ³A^′ state which is distorted from the C_2v geometry via displacements similar to the unphysical b₂ vibration in the B2PLYP calculation. This C_s minimum lies only 357 cm⁻¹ below the C_2v saddle point. Nevertheless, the cam-B3LYP calculated S₀-T₁ splitting of 18 341 cm⁻¹ is in very good agreement with the experimental result. The complications in the fluoranthene T₁ state is most likely related to the low-lying T₂ state, nominally ³B₂, with a calculated vertical energy difference of only 0.4 eV in the C_2v symmetry. The energetic proximity of these two states may be leading to the minimum distorting away from C_2v via pseudo-Jahn-Teller distortion.

Despite the lower symmetry of the cam-B3LYP optimized geometry for T₁, the calculated FC spectrum, shown in Fig. 3(b), shows fairly low intensity in all the active modes. The B2PLYP calculated FC spectrum is included in the supplementary material and shows fewer FC active modes and worse agreement with the SEVI spectra. Most notably, both FC simulations fail to reproduce the two most intense vibrational features in the SEVI spectra, labeled peak i at eBE = 25 168(12) cm⁻¹ (+339 cm⁻¹) and peak j at eBE = 25 414(10) cm⁻¹ (+585 cm⁻¹). We note that unlike peak b in the S₀ state, these two features are not associated with an obvious anisotropy difference compared to the 0-0 transition; also see the larger image in the supplementary material. Peak i corresponds well in frequency to the (a′)ν₄₈ in-plane skeletal distortion vibration, but the intensity is significantly different between calculation and experiment. The other FC active a′ modes in the eBE = 25 000-25 500 cm⁻¹ region do not exhibit such large deviations in intensity, i.e., they all correspond to weak features in the SEVI spectrum. The other possible assignment for peak i is the (a″)ν₁₈ out-of-plane ring deformation vibration, calculated at 356 cm⁻¹, and corresponds to the same (a₂)ν₉ vibration possibly observed in the S₀ state. Peak j at +585 cm⁻¹ does not match well in frequency with any of the FC predicted in-plane modes. From the calculated vibrational frequencies, the two possible assignments are the (a″)ν₁₄ out-of-plane ring deformation vibration calculated at 568 cm⁻¹ or the (a′)ν₄₃ in-plane skeletal distortion vibration calculated at 602 cm⁻¹. The former assignment would indicate additional vibronic coupling, while the latter indicates large inaccuracies in the T₁ geometry.

In the analyses of the spectra presented above, we focused on the assignments of the more intense features in the SEVI spectra of anthracene and fluoranthene. It is apparent in the experimental spectra above and those included in the supplementary material that there are other minor features present as well. A few of these appear only in spectra acquired at a specific photon energy, hinting that they are related to autodetachment processes similar to those found in resonance or dipole-bound states.^10,43 Most of the other features are too weak to make conclusive assignments, especially considering the vibronic coupling that is active in all the electronic states. Hence, their analyses are omitted, but all the calculated vibrational frequencies are included in the supplementary material for further considerations.

Finally, comparisons of the anthracene and fluoranthene results show some clear similarities between these two systems. The origin bands for all the electronic states are the dominant features in their respective spectra, indicating that the geometry changes are well distributed throughout the molecule, resulting in the dominance of the 0-0 transition. Additionally, the singlet ground states in both molecules are better reproduced by the harmonic FC simulations than the triplet states, with interesting similarities even in the vibronically coupled modes. Specifically, anthracene shows clear presence of the same nominally forbidden out-of-plane (b_3u)ν₃ vibration in both S₀ and T₁ states, and the same may be true for the (b_3u)ν₂ vibration (it is present in S₀ and may be present in the T₁ spectra as a higher eBE shoulder to peak j). The T₁ state further shows additional evidence of vibronic coupling, i.e., peak i, and although the exact assignment is uncertain, all the possibilities are non-a_g vibrations that can only acquire intensity via vibronic coupling. The situation is almost replicated in fluoranthene, where the out-of-plane (a₂)ν₉ vibration can account for the non-FC feature in the S₀ spectra as well as one of the non-FC features in the T₁ spectra, and the triplet spectra show an additional relatively intense feature, peak j, that is not reproduced in the FC simulation. A possible explanation for increased vibronic coupling in the T₁ state compared to the S₀ state is the energetic proximities of the higher lying states in the triplet manifold. A rough estimate can be acquired by looking at time-dependent density functional theory (TDDFT) vertical energies, which are listed in the supplementary material. For example, while the S₁ state in both molecules are ∼3.5 eV above S₀, the T₂ state is closer to T₁ with an energy gap of 2 eV in anthracene and only 0.9 eV in fluoranthene in C_s symmetry. Moreover, the fluoranthene T₁ state appears to distort away from C_2v due to interactions with the T₂ state, although the validity of this geometry requires more robust multireference calculations to confirm. Unfortunately, these PAH systems require relatively large basis sets to sufficiently describe the delocalized π-system, and the numerous states required in multireference considerations make such calculations quite expensive.

We present here a SEVI study that allows for an examination of the vibronic characters in the lowest energy singlet and triplet states in two similarly sized PAH molecules, anthracene and fluoranthene. PE spectroscopy provides an opportunity to explore the S₀ and T₁ states on equal footing starting from the radical anion ground state. Moreover, SEVI, together with cryogenic ion trap cooling, provides the necessary resolution to obtain vibrationally resolved PE spectra. The results presented here show some interesting similarities and differences in these two three-ringed PAH molecules. Fluoranthene, having a larger π system than anthracene, has a larger EA, with the excess electron residing in a MO that is well-delocalized throughout the molecule. This more precisely determined EA value is higher than the previously estimated value. Fluoranthene also generally exhibits less FC activity in both S₀ and T₁ states compared to anthracene, and its T₁ state exhibits possible pseudo-Jahn-Teller distortion. Beyond these differences, both molecules exhibit similar behaviors relating to some aspects of vibronic coupling. For example, out-of-plane vibrations in the SEVI spectra of the S₀ state are clearly present, even though the S₀ state is energetically far from excited states in the singlet manifold in both molecules. Moreover, the same out-of-plane vibration surprisingly shows up with higher intensity in the SEVI spectrum of the T₁ state, in addition to other vibronically coupled modes. The apparent difficulties in properly describing the T₁ states also point to possible calculation errors in describing the radical anion, which can certainly impact the overall analysis. A more in-depth exploration of the electronic structures in these molecules, including a proper multireference analysis, should yield a deeper insight into these PAH systems.

See supplementary material for calculated geometry details of the anion ground state and neutral S₀ and T₁ states of anthracene and fluoranthene, together with all the calculated vibrational frequencies and vertical excited energies, as well as β values and additional SEVI spectra and images relating to Figs. 2 and 3 with FC spectra calculated at B2PLYP level.

This work was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Award No. DE-SC0010326. The computational resources used in this work are supported by National Science Foundation Grant No. CHE-0840494.