Autoionization from the plasmon resonance in isolated 1-cyanonaphthalene

For more than three decades, polycyclic aromatic hydrocarbons (PAHs) have been conjectured to be abundant in space;^1–3 they are surmised to dominate mid-infrared emissions,^4–7 including the so-called aromatic infrared bands (AIBs) in UV-irradiated astrochemical environments.^6,8,9 These IR emission bands are ubiquitous in the universe, observed across many astrochemical environments, including the galactic interstellar medium, planetary nebulae, star forming regions, and other galaxies.^10,11 Only in the past few years have specific PAHs been identified in the cold, dark molecular cloud TMC-1 (Taurus Molecular Cloud-1) through comparing radioastronomy observations with spectra recorded in the laboratory.^12–14 Significantly, the observed abundances of two cyanonaphthalene isomers were six orders of magnitude higher than astrochemical modeling predicted;¹² there are similar abundance discrepancies between observations and modeling for the other identified PAHs.^13–15 Prior to these identifications, it was thought that PAHs with less than ≈50 atoms would not radiatively stabilize following ionizing interactions.^16–18 

In a step toward characterizing the dynamic processes leading to the preponderance of 1-cyanonaphthalene (1-CNN, Fig. 1) in TMC-1, some of the current authors used the cryogenic Double ElectroStatic Ion storage Ring ExpEriment (DESIREE) facility at Stockholm University to demonstrate efficient radiative cooling of the radical cation 1-CNN⁺ through recurrent fluorescence (RF).^23,24 RF is a relaxation mechanism associated with radiative emission from thermally populated electronic states, accessed through a process known as inverse internal conversion.^25,26 The occurrence of RF appears common in isolated PAH cations^27–32 and maybe more active than previously thought in many PAHs because Herzberg–Teller coupling increases the intensity of the low-lying electronic transitions.^23,33 The DESIREE study demonstrated that 1-CNN⁺ can radiatively stabilize efficiently through RF when formed with up to 5 eV worth of vibrational energy,²³ which is ≈2 eV beyond the bond dissociation threshold (≈3 eV).³⁴ Because the ionization potential of 1-CNN is ≈8.59 eV,³⁵ 1-CNN⁺ generated through interaction with H⁺ + e⁻ recombination radiation (13.6 eV) or Lyman-α (10.2 eV) should be resilient to decomposition of the organic backbone because RF cooling outcompetes dissociation. However, despite the astrochemical importance of neutral 1-CNN, little is known about its detailed photoionization dynamics.

In this work, we used the CiPo beamline³⁶ at the Elettra-Sincrotrone Trieste laboratory (Trieste, Italy) to record photoelectron velocity-map images for 1-CNN following photoionization with hν = 9.0–19.5 eV radiation (in 0.5 eV increments). Details of the velocity-map imaging end station and experimental arrangement are available elsewhere.³⁷ Briefly, ≈3 g of 1-CNN (Sigma-Aldrich, $>$99% purity) was loaded into a sample vessel connected to a heated (temperature-controlled) nozzle (2 mm diameter) with a skimmer assembly. The sample was kept at T = 323 K and the nozzle at T = 358 K. No sample decomposition was evident. These conditions gave a working pressure in the range from high-10⁻⁸ to low-10⁻⁷ mbar on a base vacuum chamber pressure of 1 × 10⁻⁸ mbar. Velocity-map images of the photoelectrons were accumulated in coincidence with the parent ion in order to avoid photoelectron signals originating from fragments or contaminants, such as background H₂O. The images were processed using antialiasing and polar onion peeling algorithms to obtain photoelectron spectra and photoelectron angular distributions (PADs).³⁸ The velocity-to-energy conversion was calibrated using the atomic photoelectron spectra of neon and xenon.³⁹ The ΔE/E resolution of the imaging assembly is ≈6%, and extracted PADs (as β₂ values) have uncertainties of typically ±0.1. The photoionization spectrum was normalized with respect to photon flux at each photon energy. The photoionization spectrum in this work is often referred to as the total co-incidence photoelectron spectrum (or total electrons in co-incidence with the parent ion for photon energies after fragmentation channels become accessible). Higher-order effects in the synchrotron radiation were negligible, as evidenced from the velocity-map images.

Photoelectron spectra for 1-CNN are shown in Fig. 2(a), with the hν = 9.0 and 9.5 eV spectra shown expanded in Fig. 2(b). The corresponding PADs, quantified as β₂ values, are shown in Fig. 2(c). The photoionization spectrum associated with the formation of 1-CNN⁺ (measured in coincidence) is given in Fig. 2(d), along with a He I (21.2 eV) photoelectron spectrum (horizontal abscissa corresponds to electron kinetic energy) for 1-CNN from Ref. 20, and also our computed vertical ionization potentials. Computed vertical electronic transition energies of 1-CNN are shown in Fig. 2(e).

The photoelectron spectra reveal several bands, most prominently appearing at hν < 11.5 eV, where the photoionization yield is maximized [Fig. 2(d)]. The photoelectron spectrum at hν = 9.0 eV shows a vibronic structure, which has been assigned to the D₁[1A″] ← S₀ ionization with the structure associated with fundamental and combination bands involving 1430 and 1700 cm⁻¹ in-plane stretching modes. This vibronic structure is consistent with earlier He I (21.2 eV) and He II (40.8 eV) photoelectron spectra^20,35 and provides the ionization potential at 8.60 ± 0.03 eV, in agreement with the literature values of 8.59 and 8.61 eV (no uncertainties given). Note that He I and He II photoelectron spectra should be (predominantly) non-resonant and, thus, reflect only direct photoionization processes. Two further photoelectron peaks corresponding to binding energies of ≈9.35 eV (9.35 eV in Ref. 20) and ≈10.28 eV (10.31 eV in Ref. 20) are associated with the second (D₁[2A″] ← S₀) and third (D₂[3A″] ← S₀) lowest ionization transitions, respectively, and are similarly consistent with the He I photoelectron spectrum. As described below, β₂ values associated with the first two lowest ionization transitions are reproduced using a model combining Dyson orbitals and Coulomb wave electron ejection.

For 11 ≤ hν ≤ 18 eV, our excited state calculations on 1-CNN indicate that there are ≈20 ionization transitions, resulting in complex photoelectron spectra and β₂ values. Furthermore, this photon energy spans the so-called plasmon resonance. The term plasmon resonance describes a fascinating property of isolated PAHs that manifests as a broad autoionizing resonance in the hν = 14–18 eV range (closer to hν = 11.5–17 eV in this work) in electronic absorption spectra, photo-ion yield curves, and electron energy loss spectra.^42,43 This feature has been interpreted as arising from a high density or “pile-up” of indistinguishable excitations to π* and σ* states whose excitation cross section is augmented by electron correlation effects.^43,44 Because these states, which are technically shape and Feshbach resonances, are all strongly overlapping in excitation energy and are coupled to some degree, they are treated as a collective. By analogy to similar electronic structures in fullerenes, graphene, and three-dimensional materials, this feature is commonly referred to as the plasmon resonance in isolated PAHs.⁴⁵ Although the plasmon resonance is situated in the detachment continuum and is open to spontaneous electron ejection through autoionization,⁴⁶ some combination of centrifugal, polarization, and exchange forces leads to a temporary binding interaction with a valence-localized character.⁴⁷ The plasmon resonance for 1-CNN was modeled [Fig. 2(e)] by integrating the calculated oscillator strengths (≈600 states up to 20 eV) and assuming absorption profiles for each transition with a 0.5 eV broadening (FWHM) determined by Franck-Condon factors rather than lifetime effects. We note that the feature is dominated by π–π* transitions since they typically have oscillator strengths an order of magnitude or more larger than transitions to σ* states. Due to the number of excited states, our simple integration model of the plasmon resonance does not consider the vibronic structure of the individual transitions. However, we note that the model does satisfactorily account for the peak in the photoionization spectrum at ≈13 eV [Fig. 2(e)].

Computed direct photoionization cross sections to the accessible A″ (π-like ejection) and A′ (σ-like ejection) states of the neutral are shown in Fig. 3(a). The corresponding Dyson orbitals for the first four ionized states of each symmetry class are shown in Fig. 3(b). The sum of the state-specific direct photoionization (DPI) cross sections is shown in Fig. 3(c) and is unable to account for either the shape of photoionization spectrum or the broad peak at ≈12.5 eV. On the other hand, our plasmon model [Fig. 3(c), orange], combining the DPI model and the absorption cross section curve [Fig. 2(e)], reproduces the shape and the peak feature at ≈12.5 eV in the photoionization spectrum. Our model is simplistic—it assumes that individual electronic transitions in the plasmon resonance all have a 0.5 eV width absorption profile and all excitations lead to autoionization.⁴⁶ The latter of these assumptions is consistent with the fact that photodissociation was observed only for hν > 16 eV [Fig. 2(c)], although the dissociation signal level in the experiment was only a few percent compared with the parent photoion signal. The main photofragments corresponded to loss of HCN, H, and C₂H₂, consistent with observations from a collision-induced dissociation study.³⁴ Ultimately, we conclude that excitation to the plasmon resonance followed by autoionization is more probable than direct photoionization and that the probability for dissociation, directly or indirectly (i.e., after internal conversion) following excitation of the plasmon resonance, is very low.

Comparing modeled β₂ values with experiments is straightforward for the first few ionization potentials leading to A″ states [Fig. 3(d) and supplementary material] because the spectral bands in the photoelectron spectra are well separated, with minimal contamination from autoionizing states. The consistency between theory and experiment gives confidence that the PADs for direct photoionization can be modeled adequately. On the other hand, the large number of ionization thresholds and excited electronic states of the neutral over the hν = 11–18 eV range mean that the measured β₂ values result from the sum (incoherent superposition) of many direct photoionization and prompt autoionization processes. Broadly speaking, simulated β₂ values for ionization to A′ states are positive, while those for ionization to A″ states are negative (see the supplementary material). Experimental β₂ values, averaged over the P1 and P2 features [as indicated in Fig. 2(a)], are shown in Fig. 3(e) and are predominately negative. This trend correlates with the main fraction of photoelectrons arising from prompt autoionization of π–π* states, where prompt implies that electron ejection occurs more rapidly than molecular rotation; such a process resembles direct photodetachment to an A″ state. We note that the P2 feature shows quite negative β₂ values for hν = 15–18 eV [Fig. 3(e)], consistent with several strong π–π* transitions followed by prompt autoionization. In summary, trends in simulated and experimental β₂ values are consistent with the interpretation that autoionization from the plasmon resonance is the dominant electron ejection process for 1-CNN over the hν = 11.5–17 eV range. In our interpretation, the division of the plasmon resonance between P1 and P2 regions is somewhat arbitrary but was guided by the double-peaked feature in the photoelectron spectra [Fig. 2(a)]. For larger PAHs with more excitations contributing to the plasmon resonance and further direct photoionization thresholds, this structure will likely disappear.

This work has considered the photoionization and autoionization dynamics of 1-CNN, a molecule that has been identified through radioastronomy in the cold, dark molecular cloud TMC-1. The present study indicates that, when subjected to VUV and soft XUV radiation, such as 13.6 eV photons generated through H⁺ + e⁻ recombination or the Lyman-α line at 10.2 eV,⁵² the molecule is most likely to be photoexcited to the plasmon resonance, which promptly autoionizes. These hydrogen emission lines correspond to the most abundant photon energies in many astrochemical environments.⁵² The probability for photodissociation of 1-CNN, either in an excited state or after internal conversion to the ground electronic state, is low over the microsecond timescale. This conclusion is consistent with the fact that the photodissociation signal in this work, albeit weak, is present for hν > 16 eV, although we note that the observation of dissociation could be restricted by the short time between ionization and ion detection (≈1 µs), i.e., the experimental window. While slower, statistical dissociation (and potentially isomerization) may occur, this contribution is likely small based on our 1-CNN⁺ DESIREE experiments, where we monitored quenching of neutral production (through dissociation) by RF.^23,24 Combining the results from this work with those from our recent DESIREE studies on 1-CNN⁺,^23,24 which showed that cations formed with an internal vibrational energy of up to 5 eV (or a minimum photon energy of 13.6 eV when ionizing neutral 1-CNN) are able to radiatively stabilize through RF without dissociating,²³ leads to the sequence of dynamics shown in Fig. 4. The astrophysical significance of these dynamics is that 1-CNN is unlikely to exist as neutral molecules in UV-dominated regions of space, such as photodissociation regions (PDRs);⁵³ rather, the organic framework will be present as 1-CNN⁺. In dark molecular clouds, such as TMC-1, the 1-CNN vs 1-CNN⁺ charge balance will be determined by the competition between photoionization (by VUV photons and cosmic rays) and electron–ion/ion–ion recombination.^52,54

We propose that similar autoionization and radiative cooling dynamics are key ingredients for the VUV and soft XUV photoresilience of other PAHs that can survive in space. The photoionization and radiative cooling dynamics characterized for 1-CNN challenge the conventional wisdom^16–18 that energized interstellar PAHs radiatively cool only through infrared emission, which occurs substantially more slowly and less efficiently than RF cooling,^30,59–62 and, consequently, that only PAHs with more than ≈50 atoms can radiatively stabilize.² It is clear that further experiments on the radiative cooling and photoionization dynamics of PAHs that are thought to exist in space, including the known PAHs indene and 2-cyanoindene,^13,14 are needed to establish “rules of thumb” for determining interstellar PAH propensity.

Finally, we note that recent XUV-IR experiments have sought to investigate ultrafast relaxation dynamics of PAH cations in their so-called correlation band (CB,⁶³ an amalgamation of bound, ionized states situated below the lowest double ionization threshold).^64,65 In contrast, the plasmon resonance discussed in this work is due to the amalgamation of excited states situated above the lowest ionization threshold of the neutral molecule. For naphthalene, the XUV-IR study used ≈23 eV photons to generate the cation; this photon energy is situated well above the expected plasmon resonance⁴⁵ for (neutral) naphthalene and at an energy where the total direct photoionization cross section should exceed the excitation cross section. In future experiments, it would be interesting to apply time-resolved strategies to 1-CNN, or similar molecules, with the pump pulse tuned to excite the plasmon resonance.

The supplementary material contains details on expanded plots of all photoelectron spectra and β2 values, illustrations of Dyson orbitals, all modeled state-specific photoionization cross sections for the A′- and A″-symmetry states, β2 values for the 1A″ and 2A″ ionized states modeled with different basis sets, modeled state-specific direct photoionization cross sections, and β2 values for each ionized state.

Funding was provided by the Swedish Foundation for International Cooperation in Research and Higher Education (STINT) Grant for Internationalization program (Grant No. PT2017-7328 to M.H.S. and J.N.B.) and an EPSRC New Investigator Award (Grant No. EP/W018691 to JNB). E.K.A. acknowledges the University of East Anglia for a doctoral studentship. The work was supported by the MAECI Italy-Sweden project “Novel molecular tools for the exploration of the nanoworld” and the PRIN Grant No. 20173B72NB project “Predicting and controlling the fate of bio-molecules driven by extreme-ultraviolet radiation.” H.Z. acknowledges the Swedish Research Council for an individual project grant (Contract No. 2020-03437). This article is based upon work from COST Action CA18212—Molecular Dynamics in the GAS phase (MD-GAS), supported by COST (European Cooperation in Science and Technology). Electronic structure calculations were carried out on the High Performance Computing Cluster supported by the Research and Specialist Computing Support service at the University of East Anglia. Part of the theoretical work used resources from the iOpenShell Center for Computational Studies of Electronic Structure and Spectroscopy of Open-Shell and Electronically Excited Species (http://iopenshell.usc.edu).

The authors have no conflicts to disclose.

This work was based on a beamtime application prepared by M.H.S. and J.N.B. Analysis of experimental data was performed by E.K.A., J.E.N.N., B.Z., and J.N.B. P.B., J.E.N.N., B.Z., R.R., N.P., J.C., and L.A. participated in experimental data acquisition. Electronic structure calculations were performed by C.S.A., E.K.A., and J.N.B. The manuscript was prepared by J.N.B. with assistance from E.K.A., and was discussed by all authors.

James N. Bull: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Supervision (equal); Writing – original draft (lead); Writing – review & editing (equal). Paola Bolognesi: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal). Cate S. Anstöter: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – review & editing (equal). Eleanor K. Ashworth: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). José E. Navarro Navarrete: Data curation (equal); Investigation (equal); Writing – review & editing (equal). Boxing Zhu: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – review & editing (equal). Robert Richter: Data curation (equal); Investigation (equal); Methodology (equal); Resources (equal); Writing – review & editing (equal). Nitish Pal: Data curation (equal); Investigation (equal). Jacopo Chiarinelli: Data curation (equal); Investigation (equal). Lorenzo Avaldi: Data curation (equal); Investigation (equal); Methodology (equal); Writing – review & editing (equal). Henning Zettergren: Project administration (equal); Supervision (equal); Writing – review & editing (equal). Mark H. Stockett: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the authors upon reasonable request.