Unraveling the decarboxylation dynamics of the fluorescein dianion with fragment action spectroscopy

Fluorescein is a dye and fluorescent tracer, and many of its applications exploit its intense solution-phase fluorescence.¹ Several charged and neutral pH dependent forms of fluorescein exist in the solution phase, including a doubly deprotonated dianion (Fl²⁻), which has a fluorescence maximum at 500 nm, corresponding to radiative relaxation from S₁ → S₀.^1,2 While the fluorescence quantum yield in solution is very high, previous studies have indicated that Fl²⁻ does not fluoresce in the gas phase; instead, S₁ decays via electron loss.^2,3 Decarboxylation of the carboxylate group of Fl²⁻ may also be favorable, given the high thermodynamic stability of CO₂, particularly if the Fl⁻ radical anion that is formed following electron loss from the dianion has a carboxyl radical character.^4,5 In fact, multiple photon excitation of Fl²⁻ in the infrared (IR) (via irradiation with a CO₂ laser) and collisionally induced dissociation (CID) of Fl²⁻ were found to exclusively produce decarboxylation with electron loss (i.e., Fl–CO₂⁻ + e⁻ + CO₂), in contrast to excitation in the visible, where electron loss was the dominant channel (i.e., Fl⁻ + e⁻).² Therefore, the mechanism for decarboxylation, as well as the role of excited states in mediating the electron loss and decarboxylation dynamics of Fl²⁻, remains unclear. In this manuscript, we present fragment action spectroscopy of Fl²⁻ in the gas-phase between λ = 260–600 nm in order to gain insight into the competition between electron and CO₂ loss from Fl²⁻.

Dianions, such as Fl²⁻, are characterized by the balance of long range repulsive forces and short range attractive forces, which constitute the repulsive Coulomb barrier (RCB).^3,6–11 There is a multitude of RCBs associated with a dianion corresponding to individual rovibronic states of the parent species and linked to distinct product channels, whether that be electron or anion loss channels.^12,13 Therefore, each RCB forms a barrier to the loss of low kinetic energy electrons or anionic fragments. One key distinction between the study of electron loss and anionic dissociation processes of dianions is the relative prevalence of tunneling through a specific RCB. Resonant and non-resonant electron tunneling through an RCB has been observed for many dianions, on account of the low mass of the electron.^14–17 In contrast, anionic fragmentation of dianions has only been observed as an “over-the-barrier” process.^13,18

Experimentally, the properties of dianions and their electronic RCBs have been effectively studied by photoelectron spectroscopy, where a characteristic low electron kinetic energy (eKE) cutoff is observed in the photoelectron spectra at the minimum height of the RCB.^18–22 Electron loss via tunneling has been observed by the presence of photoelectrons below the cutoff, and these features are often spectrally characterized by a peak at fixed eKE, which is invariant with photon energy.^3,12,23 Photodissociation spectroscopy has also been used to study anion loss from dianions, in particular for IrBr₆²⁻, where the minimum height of the RCB to fragmentation has been reported from the kinetic energy release of the Br⁻ fragment.^13,18,24,25 The groups of Weber and Kappes have used reflectron time-of-flight mass spectrometry to record fragment action spectra,^18,25,26 while other methodologies, such as the collisional activation or laser excitation of dianions in a quadrupole trap or electrostatic ion storage rings, have also been used to study polyanion dissociation dynamics.^2,13,24

Previous work in our group has studied the photodetachment dynamics of Fl²⁻ using frequency-resolved photoelectron imaging.^3,12,27 Unusual and rich photodetachment dynamics were observed, with electron loss occurring via several distinct resonant tunneling pathways that are associated with different mediating dianion states, RCBs and final anion states.¹² While photoelectron spectroscopy can yield a detailed picture of the electron loss dynamics, once the electron has been lost, this methodology is blind to any subsequent dynamical process, such as dissociation. This is the case for Fl²⁻, where CO₂ loss has been observed in conjunction with electron loss under specific experimental circumstances.² So how do such dissociation dynamics take place and can this be reconciled with the electron dynamics? Commonly observed processes involve electron loss from a carboxylate anion, leading to rapid decarboxylation of a transiently formed carboxyl radical via a repulsive potential energy surface.^4,5,28–30 However, it is unclear if the Fl⁻ formed following electron loss from Fl²⁻ would have carboxyl radical character and be expected to undergo a similar repulsive dissociation. Additionally, the photodetachment dynamics indicate that at higher photon energies, new tunneling pathways compete, which leads to the formation of a highly internally excited Fl⁻, both electronic and/or vibrational. The role of vibrational excitation may be important if CO₂ loss occurs via a ground state dissociation, while different electronic states in the anionic product may have different dissociation dynamics. In this manuscript, we build on the previously reported photoelectron imaging studies, which focused on the nature of the RCB, to unravel the decarboxylation dynamics of Fl²⁻ by the application of fragment action spectroscopy.^3,12

The action spectrum of a fragment ion cannot be recorded using linear time-of-flight methods as, to a first approximation, all of the photoproducts have the same velocity as the parent species, making it challenging to separate the time-of-flight of the fragment ions from the parent dianions. In order to overcome this limitation, a secondary mass spectrometer consisting of a reflectron was introduced, which acts as an electrostatic kinetic energy filter, in order to temporally separate the charged photoproducts with different $mz$ ratios.^31,32 We can study both the photodetachment (different z) and photodissociation (different m and sometimes z) of dianions using this methodology.

The ion beam and light source used in this apparatus are the same as for previous photoelectron imaging experiments and have been described in detail elsewhere.^33,34 Briefly, Fl²⁻ was produced using electrospray ionization of a 1 mM solution of sodium fluorescein in methanol. The dianions entered the vacuum via a capillary and were guided and trapped by a series of radio frequency ring electrodes, before acceleration at −1.5 keV using a Wiley–Maclaren time-of-flight mass spectrometer. The dianions of interest were mass selected by time-of-flight at the intersection of the ion beam with the nanosecond tunable output of a Nd:YAG pumped optical parametric oscillator, which induces photochemistry in Fl²⁻. Likely, photo products include photoelectrons, which can be studied by photoelectron imaging,^19,20 and anions produced via electron loss, dissociation, or a combination of both.

The new detector region, including the reflectron secondary mass spectrometer, is shown schematically in Fig. 1. After the interaction of laser light and Fl²⁻, any product anions, including Fl⁻ (blue), continue along the same beam axis as Fl²⁻ (red), past a grounded detector and deflector, and into the reflectron. As the kinetic energy of the dianion (⁠$mz=mFl2=165$ amu) is half that of the anion (⁠$mz=mFl=330amu$⁠), the two ion packets can be temporally separated by the behavior of the reflectron as a kinetic energy filter. The reflectron consists of a grounded front plate, followed by a resistive glass tube (15 cm long) that is biased at either end (front plate, V_f = −700 V and mirror plate, V_m = −3500 V). As the kinetic energy of the anion product is twice as large as the dianion, the mirror voltage of the reflectron must be more than twice the beam energy of the dianion in order to reflect the anions for detection. This is shown in Fig. 1 by the Fl²⁻ (red) and Fl⁻ (blue) ion paths, where the flight distance is substantially longer for the species with larger kinetic energy. Typically, reflectrons consist of a stack of metal plates linked via a chain of resistors,^18,31 but the use of a resistive glass tube here allows the reflectron to act as an infinite chain of electrodes. This minimizes lensing and fringe field effects near individual electrodes, in order to maximize the fragment mass resolution of the reflectron, while maintaining its short length. Similar resistive glass tubes have been used in the velocity map imaging of electrons.³³ Once the ions enter the reflectron, the deflector is pulsed (from 0 V to −3 kV, pulse duration ∼100 µs) to push anions (parent dianions and fragment anions) into the ion detector on the second pass, so that the time-of-flight of each species can be recorded.

If any additional anionic photoproducts are formed following irradiation with laser light, then additional peaks will be observed in the secondary mass spectra. For example, Fl–CO₂⁻ has a time of flight between Fl²⁻ and Fl⁻ (⁠$mz=330−441=286$ amu). Generally, ion peaks are assigned to a specific $mz$ by comparison of the experimental ion arrival times to the results of a simple model, which calculates the acceleration of an ion of a specific $mz$ within the applied fields, extracts an ion turn around the position, and ultimately predicts a time-of-flight.

Following the assignment of the species in the secondary mass spectrum, action spectra can be recorded by monitoring the intensity of the signal from a specific mass-selected species, as a function of laser wavelength. The laser was scanned at a rate of 0.1 nm s⁻¹, and the scans were repeated three times. All of the ion peak intensities in the following action spectra were normalized to the integral of the parent dianion peak in order to account for substantial fluctuations in the parent ion beam. The low intensity of the fragment signal requires additional amplification after the detector, resulting in saturation of the Fl²⁻ ion signal on the final time-of-flight trace (Fig. SI1). However, the effect of this on the reported action spectra is minimal, as the Fl²⁻ ion intensity was very constant, as verified by measurement of the parent ion signal without further amplification and by repeated measurements of the action spectra, which were very similar. We are currently modifying the data acquisition setup to allow us to independently amplify the fragment signal, which should result in a more sensitive normalization procedure in the future. It should also be noted that no attempt has been made to account for the significant changes in laser intensity, which occur over this wavelength range, with particular minima being observed at 400 and 292 nm, which are characteristic of our optical parametric oscillator, while over other regions, the output was found to be relatively constant in terms of pulse energy. Electron yield spectra can also be acquired by simply monitoring the number of electrons on the imaging detector as a function of wavelength.

To aid the interpretation of the results, additional electronic structure calculations were performed on Fl²⁻, Fl⁻, and Fl–CO₂⁻. Ground state density functional theory (DFT) calculations and excited state time-dependent DFT calculations, using the Tamm–Dancoff approximation, were performed using the Gaussian software package at the B3LYP level of theory with the 6-311G ++ 2d,2p basis set.^35–37 

In the absence of laser light, the secondary mass spectrum consisted of a single peak with $mz$ = 165 amu, assignable to Fl²⁻. Once the sample was irradiated, two further peaks were observed over specific wavelength ranges, corresponding to Fl–CO₂⁻ and Fl²⁻. An example time-of-flight spectrum is shown in the supplementary information, Fig. SI1. No evidence was seen for the formation of other products such as Fl–CO₂²⁻, indicating that dissociation only occurred in conjunction with electron loss.

Previous work in our group has studied the photodetachment dynamics of Fl²⁻ using frequency resolved photoelectron imaging.^3,12,27 A brief overview of this work is presented here as its conclusions will guide some of the discussion below. Figure 2 shows the Fl²⁻ photoelectron image recorded at λ = 320 nm, where two clear spectral features are observed. Photoelectron imaging in the visible reported a single fixed eKE feature, which remains even in the UV and corresponds to the inner ring in Fig. 2. This feature arises from resonant tunneling through an RCB associated with the Fl²⁻[S₁] state (RCB₁).³ As this feature’s eKE is fixed with decreasing λ, the excess photon energy following electron loss will be deposited as internal rovibrational energy in the ground-state anion, Fl⁻[D₀].³ For UV excitation, a second higher-lying and competing tunneling pathway (outer ring in Fig. 2) appears following excitation to the S₂ state of Fl²⁻, Fl²⁻[S₂]. A fraction of the population tunnels through a RCB associated with the S₂ state, RCB₂, resulting in a different final anion electronic state, Fl⁻[D₁].¹² The different final states accessed by tunneling arise from different Koopmans’ correlations.¹² At wavelengths in the visible (λ ≤ 540 nm), only the inner feature is observed in the photoelectron images, whereas the outer ring competes at higher photon energies in the UV (λ ≤ 350 nm). In addition, a minor direct detachment channel was observed for λ ≤ 440 nm at electron kinetic energies that overlapped with the resonant tunneling pathway mediated by S₂ (outer ring).

In order to confirm the functionality of the new reflectron, the action spectra of both the electron and Fl⁻ yield arising from photodetachment of Fl²⁻ were recorded at wavelengths between 400 and 600 nm and are shown in black in Fig. 3. The electron action spectrum is smoothed via a 20-point moving average, with the raw data shown in gray. The electron and Fl⁻ action spectra are similar in structure, indicating that the primary process following absorption of a photon is a one-photon photodetachment to form the radical anion i.e., Fl²⁻ → Fl⁻ + e⁻, in agreement with previous work.^2,3 The small differences between the electron and the Fl⁻ yield spectra, shown in black, are likely to be within the noise of the two measurements or attributable to the two spectra being recorded with different detectors.

The black line in Fig. 4 is the action spectrum of Fl²⁻ recorded over a wider wavelength range, λ = 260–600 nm, by monitoring the Fl⁻ fragment. This action spectrum has a similar structure to the previously reported gas-phase action spectrum of Fl²⁻ and the absorption spectrum of Fl²⁻ in water at λ > 400 nm, indicating that the regions of increased signal intensity are due to the presence of an increased photodetachment cross section for Fl²⁻, i.e., the location of an excited electronic state.^1,2 The most intense feature is centered around 500 nm and corresponds to the bright S₀ → S₁ transition in Fl²⁻. This feature shows some vibrational structure, similar to the solution phase absorption spectrum. Irradiation of Fl²⁻ near 400 nm produces very little Fl⁻, indicating that the photodetachment cross section is very low in this region, even though, at this wavelength, direct “over-the-RCB” detachment is an open channel.^3,12 At higher photon energies, an increased yield of Fl⁻ is observed, as higher-lying electronic states of Fl²⁻ become energetically accessible. In particular, S₂ is located near 325 nm, with potential evidence for S₃ at λ < 292 nm. It should be noted that a portion of this action spectrum (λ = 300–600 nm) has been published before.¹² 

The action spectrum of Fl²⁻ recorded by monitoring the yield of the decarboxylated anion Fl–CO₂⁻ was also measured using the reflectron and is shown in red in Fig. 4. The Fl–CO₂⁻ yield spectrum is similar in appearance to Fl⁻ at wavelengths shorter than 400 nm (solid red line) but significantly different at longer wavelengths (dashed red line). The excellent correlation between the Fl–CO₂⁻ and Fl⁻ yield spectra for 300 < λ < 400 nm indicates that a one-photon process can either lead to just electron loss or combined electron and CO₂ loss, with the former being approximately twice as likely around the S₂ excited state.

In the visible range, the discrepancy between the Fl–CO₂⁻ and Fl⁻ yield spectra is due to a multiple photon process, which can be evidenced by the action spectrum taken at lower laser power. As the formation of Fl⁻ is a one-photon process, differences in the relative intensity of the Fl–CO₂⁻ peak would indicate a multiple photon process. Figure SI2 shows the Fl²⁻ action spectra recorded by monitoring the Fl⁻ and Fl–CO₂⁻ yields at wavelengths between 400 and 500 nm and at a third of the laser power used for the measurement in Fig. 4. Based on the large decrease in the relative intensity of Fl–CO₂⁻ compared to Fl⁻ from Fig. 4 to SI2, we conclude that the formation of Fl–CO₂⁻ at λ > 400 nm occurs via a multiple photon process, presumably a two-photon process (dashed red line). The formation of Fl⁻ at all wavelengths and Fl–CO₂⁻ for λ < 400 nm is a one-photon process (solid red line).

Decarboxylation is a common process in carboxylates and carboxyl radicals, driven by the large thermodynamic stability of CO₂.^4,5,38–40 It has been observed to occur via a photodissociation or dissociative photodetachment (DPD) process in carboxylate anions and even via secondary dissociative autodetachment in dicarboxylate dianions.^4,5,40 In the case of combined electron and CO₂ loss, as observed for Fl²⁻, it is likely that electron loss and CO₂ loss occur sequentially rather than via the formation of CO₂⁻, given that CO₂ has a negative electron affinity.⁴¹ Therefore, the mechanism for decarboxylation of Fl²⁻ is likely to be either DPD (Fl²⁻ $→hν$ Fl⁻ + CO₂ + e⁻), which may occur via a repulsive electronic state of the Fl⁻ anion, e.g., Fl⁻[D₁], or via the formation of a vibrationally excited Fl⁻[D₀], which then undergoes unimolecular dissociation or a combination of these processes on different anionic electronic states.

As electron loss is the first step in the formation of both Fl⁻ and Fl–CO₂⁻, the photoelectron imaging of Fl²⁻ (e.g., Fig. 2) will inform on the electron loss processes that lead to specific fragments at a given wavelength (Fig. 4).^3,12 Two resonant electron loss processes have been observed: (1) resonant tunneling through RCB₁ from the Fl²⁻[S₁] to Fl⁻[D₀] at all wavelengths studied (300 nm ≤ λ ≤ 540 nm) and (2) a second resonant tunneling channel through RCB₂ from Fl²⁻[S₂] to Fl⁻[D₁] at λ < 350 nm.^3,12 For λ < 350 nm, both processes compete as both are visible in the photoelectron image (Fig. 2).¹² There is a third electron loss channel due to detachment to Fl⁻[D₀] over RCB₀ (linking Fl²⁻[S₁] and Fl⁻[D₀]) occurring at λ ≤ 440 nm.^3,42 However, this is a very minor channel as evidenced by the low Fl⁻ ion signal observed at non-resonant wavelengths (350 nm ≤ λ ≤ 440 nm).¹² 

At the longest wavelengths studied (440 ≤ λ ≤ 600 nm), absorption of a single photon populates the excited dianion state, Fl²⁻[S₁], and the only available channel is electron tunneling through RCB₁ to form a Fl⁻[D₀],³ with no decarboxylation. The final anionic product Fl⁻[D₀], while internally hot, is stable on a timescale longer than the flight time (∼20 µs) as it is observed in the fragment action spectra (Fig. 4).³ At shorter wavelengths (λ ≤ 350 nm), where the two resonant electron loss channels compete, there is ion intensity in the Fl⁻ and Fl–CO₂⁻ yield spectra (Fig. 4) in a ratio of ∼2:1, indicating a competition between photodetachment and DPD following one-photon excitation of Fl²⁻. One explanation is that Fl⁻[D₁] is a repulsive electronic state, whereas Fl⁻[D₀] is bound, such that CO₂ loss occurs readily following the formation of Fl⁻[D₁], whereas Fl⁻[D₀] is stable on the timescale of the experiment i.e., the flight time from the interaction region to the turnaround point in the reflectron. Evidence for Fl–CO₂⁻ formation via a repulsive dissociation is found in the comparable width of the Fl–CO₂⁻ and Fl⁻ peaks in the time-of-flight spectrum (Fig. SI1). This indicates that the timescale of decarboxylation is shorter than the flight time from the interaction region to the reflectron (∼5 µs) because if dissociation occurred within the reflectron itself, a much broader distribution of ion arrival times would be expected. Hence, the DPD process appears to be prompt rather than statistical in nature.

In order to investigate the dissociation dynamics further, electronic structure calculations were performed on Fl²⁻, Fl⁻, and Fl–CO₂⁻, and the results are shown in Table I. Of most importance to the decarboxylation process is the relative energetics of the Fl⁻[D₀] and Fl⁻[D₁] states compared to the dissociation asymptote Fl–CO₂⁻ + CO₂. The calculations indicate that Fl⁻[D₀] is bound with respect to Fl–CO₂⁻ + CO₂, with a bond dissociation energy of 0.82 eV. The calculated bond dissociation energy of Fl⁻[D₁] is much smaller, at 0.09 eV, which may also indicate that Fl⁻[D₁] is bound with respect to CO₂ loss but is within the error of the calculations and comparable to the expected internal energy of anions at 298 K. Regardless, our relative energies show clearly that decarboxylation is more facile in Fl⁻[D₁] than in Fl⁻[D₀].

The larger calculated bond dissociation energy of the Fl⁻[D₀] compared with the Fl⁻[D₁] state, together with the experimental observations of competition between the formation of Fl⁻ and Fl-CO₂⁻ following excitation to S₂ (λ ≤ 350 nm), and of a relatively rapid dissociation (<5 µs), are consistent with the conclusion that decarboxylation is favored following resonant tunneling through RCB₂. Effectively, Fl⁻[D₁] loses CO₂ rapidly, whereas Fl⁻[D₀] is stable on the timescale of the experiment. The negligible bond dissociation energy for Fl⁻[D₁] suggests that loss of CO₂ is energetically possible whenever Fl²⁻[S₂] can be excited, which matches well with the observed action spectra where one-photon decarboxylation is seen for λ < 350 nm and suggests that the potential energy surface would be, at most, weakly repulsive. A cartoon illustrating the rapid decarboxylation of Fl⁻[D₁], following its formation via tunneling through RCB₂, is shown in Fig. 6(a).

However, if the above picture of dissociation was complete, we would expect to see a fixed ratio of Fl⁻ and Fl–CO₂⁻ once excitation to Fl²⁻[S₂] was energetically accessible. In reality, the relative intensity of the Fl–CO₂⁻ signal compared to Fl⁻ increases with photon energy, as shown in Fig. SI3, suggesting that there is a second dissociation mechanism at play. As we have established that decarboxylation must follow electron loss via tunneling through a RCB, we conclude that one-photon DPD must also be a possible channel from Fl⁻[D₀] when sufficient vibrational energy is present to overcome the bond dissociation energy. The Fl⁻[D₀] state is populated following resonant electron loss from Fl²⁻[S₁] and so decarboxylation becomes energetically accessible for λ < 375 nm, corresponding to the energy of Fl²⁻[S₁] plus the Fl⁻[D₀] dissociation energy [2.48 eV (500 nm) + 0.82 eV = 3.30 eV (375 nm)]. This decarboxylation channel, shown in Fig. 6(b), would be expected to increase in favourability as the photon energy increases because there is a larger likelihood of the vibrational modes relevant to decarboxylation to be populated. This decarboxylation mechanism is likely to be slower than dissociation via Fl⁻[D₁] and may be expected to result in a broader time-of-flight distribution for Fl–CO₂⁻. However, low absolute signal levels make measurement of such broadened features (in a minor channel) not presently feasible, and it should be noted that a prompt (i.e., <5 µs) contribution is also present because it appears in the Fl–CO₂⁻ signal.

In summary, one-photon decarboxylation occurs after excitation to the Fl²⁻[S₂], primarily from Fl⁻[D₁] produced via resonant tunneling through RCB₂, but with a minor contribution from vibrationally excited anions produced via internal conversion from Fl²⁻[S₂] to Fl²⁻[S₁] and subsequent resonant tunneling through RCB₁. The combination of these two mechanisms explains the observed one-photon Fl²⁻ action spectra when combined with our knowledge of the electron dynamics from photoelectron imaging. This picture of the decarboxylation dynamics is summarized in Fig. 6, which provides a comprehensive view of the possible mechanisms at play following photoexcitation. This picture is also consistent with the observations from previous CID and IR excitation experiments, where those excitation methods provided sufficient energy to surpass the Fl⁻[D₀] dissociation barrier after an electron has been lost.² 

The final consideration is the mechanism of the two-photon electron loss and decarboxylation process, which occurs in the visible region. For λ ≥ 400 nm (Fig. 4), electron loss results in the production of Fl⁻[D₀]. The subsequent absorption of a second photon, which results in decarboxylation, is likely to form an electronically excited Fl⁻[D_n]. This is supported by the observation of clear vibrational structure in the action spectrum of Fl⁻ (Fig. 5). The Fl⁻[D_n] electronically excited state is higher lying than the Fl⁻[D₁] state, which lies 0.73 eV above Fl⁻[D₀] (see Table I) and may be dissociative along the CO₂ coordinate. Alternatively, decarboxylation could take place following internal conversion to the weakly bound Fl⁻[D₁] state or from the Fl⁻[D₀], although this would be a statistical and slower dissociation mechanism.

A new reflectron secondary mass spectrometer has been used to record the action spectra of the photodetachment and DPD product channels, following irradiation of Fl²⁻ at wavelengths between 260 and 600 nm. The action spectrum of Fl²⁻, recorded by monitoring Fl⁻, indicates that at least two excited states of Fl²⁻ are observed. In addition, photoelectron imaging demonstrates the presence of electron loss via several distinct resonant tunneling channels through different RCBs. For λ > 400 nm, excitation to the Fl²⁻[S₁] results in photodetachment, with CO₂ loss only occurring via a two-photon process of photodetachment and subsequent decarboxylation of Fl⁻. For λ < 400 nm, CO₂ loss occurs via a one photon process, with increasing relative yields of Fl–CO₂⁻ as the photon energy increases. In addition, electronic structure calculations indicate that Fl⁻[D₀] is bound with respect to decarboxylation, whereas Fl⁻[D₁] is similar in energy to the dissociation asymptote. Taken together, these observations indicate that decarboxylation occurs rapidly following tunneling through RCB₂ and the formation of a Fl⁻[D₁], with a smaller contribution from the formation of highly vibrationally excited Fl⁻[D₀] following tunneling through RCB₁. Finally, the two-photon process of electron loss via direct detachment or tunneling through RCB₁ and subsequent decarboxylation, which is observed in the visible region, is likely mediated by a high-lying repulsive state of Fl⁻.

The supplementary material contains an example time-of-flight trace, a low laser power fragment action spectrum of Fl2−, and a plot showing the proportion of decarboxylation as a function of wavelength.

Jemma A. Gibbard is thankful for the support of a Ramsay Memorial Fellowship. The authors would like to thank Adam Mendoza for developing the data acquisition program used to record the action spectra in this work.

The authors have no conflicts to disclose.

Jemma A. Gibbard: Conceptualization (equal); Formal analysis (lead); Investigation (lead); Writing – original draft (lead); Writing – review & editing (equal). Jan R. R. Verlet: Conceptualization (equal); Writing – review & editing (equal).

The data that support the findings of this study are openly in Zenodo at https://doi.org/10.5281/zenodo.7828814, Ref. 43.