Five Birds with One Stone: Photoelectron Photoion Coincidence Unveils Rich Phthalide Pyrolysis Chemistry

Phthalide pyrolysis has been assumed to be a clean fulvenallene source. We show that this is only true at low temperatures, and the C7H6 isomers 1-, 2-, and 5-ethynylcyclopentadiene are also formed at high pyrolysis temperatures. Photoion mass-selected threshold photoelectron spectra are analyzed with the help of (time-dependent) density functional theory, (TD-)DFT, and equation-of-motion ionization potential coupled cluster, EOM-IP-CCSD, calculations, as well as Franck−Condon simulations of partly overlapping bands, to determine ionization energies. The fulvenallene ionization energy is confirmed at 8.23 ± 0.01 eV, and the ionization energies of 1-, 2 and 5-ethynylcyclopentadiene are newly determined at 8.27 ± 0.01, 8.49 ± 0.01 and 8.76 ± 0.02 eV, respectively. Excited state features in the photoelectron spectrum, in particular the Ã A′ band of 1-ethynylcyclopentadiene, are shown to be practical to isomer-selectively detect species when the ground-state band is congested. At high pyrolysis temperatures, the C7H6 isomers may lose a hydrogen atom and yield the fulvenallenyl radical. Its ionization energy is confirmed at 8.20 ± 0.01 eV. The vibrational fingerprint of the first triplet fulvenallenyl cation state is also revealed and yields an ionization energy of 8.33 ± 0.02 eV. Further triplet cation states are identified and modeled in the 10−11 eV range. A reaction mechanism is proposed based on potential energy surface calculations. Based on a simplified reactor model, we show that the C7H6 isomer distribution is far from thermal equilibrium in the reactor, presumably because irreversible H loss competes efficiently with isomerization.


■ INTRODUCTION
Polycyclic aromatic hydrocarbons (PAHs) are emitted into the atmosphere on a large scale as byproducts of natural and anthropogenic combustion, 1 and have adverse health effects. 2 Looking beyond the bio-and anthroposphere, PAHs are also abundant in the interstellar medium (ISM) as evidenced by their ubiquitous IR emission bands. 3−5 In the ISM, PAHs are believed to be formed in outflows of carbon-rich stars by processes reminiscent of those in combustion. 6,7 PAHs containing pentagonal structures are also widespread and may play an important role in the formation of curved PAHs and perhaps ultimately fullerenes. 8−12 It is thus critical to understand the reactivity and spectroscopy of such pentagonbearing hydrocarbons.
Hydrocarbons of C 7 H 6 composition are commonly formed in combustion experiments. 13 There are four cyclopentadienebased C 7 H 6 isomersfulvenallene (FA), 1-ethynylcyclopentadiene (1ECP), 2-ethynylcyclopentadiene (2ECP), and 5ethynylcyclopentadiene (5ECP) (see Figure 1)and of these structures, FA is the most stable one. 14 Fulvenallene is formed in the thermal decomposition of benzyl radicals. 15,16 All four cyclopentadiene-based C 7 H 6 isomers yield the resonance-stabilized fulvenallenyl radical (FArad)a PAH and soot precursordirectly via H loss. 17 This motivated experimental and computational studies of these species in the past.
In a He I photoelectron spectroscopic study, Muller et al. 18 reported the ionization energy of FA as 8.29 eV to the X̃+ 2 A 2 ground state of the cation. They pyrolytically induced the isomerization of 1,2-diethynylcyclopropane, which had previously been shown to yield bicyclo[3.2.0]hepta-l,4,6-triene, FA, and ECP isomers together with heptafulvalene, depending on the pyrolysis conditions. 19 Early experiments on the thermolysis of the lactone phthalide (2-benzofuran-1(3H)one, C 8 H 6 O 2 , see Figure 1 for the structure) revealed that FA and ECP isomers are formed, among other minor breakdown products. 20, 21 Botter et al. 22 pyrolyzed phthalide, isolated and stabilized the product using tetracyanoethylene, and recorded a photoelectron spectrum yielding 8.22 eV as the FA ionization energy. The photoelectron spectrum also showed two unresolved bands at 9.14 and 10.15 eV, which were attributed to electronically excited cation states of 2 B 1 and 2 B 2 symmetry, respectively. The notation assumes yz to be the principal plane in C 2v symmetry, which we will use herein, too. Three decades later, Steinbauer et al. 23 studied the photoelectron spectroscopy of FA and FArad. In their work, phthalide was pyrolyzed in a SiC microreactor and the threshold photoelectron spectra (TPES) of the products were recorded using synchrotronbased imaging photoelectron photoion coincidence (iPEPI-CO) spectroscopy. A strong resonance was found at 8.22 eV in photoion mass-selected (ms-)TPES of m/z 90 and ascribed to ionization of FA to the 2 A 2 ground state of the ion. The comparably clean C 7 H 6 spectrum is intriguing in light of the potential energy surface (PES) study by Kvaskoff et al.,24 which showed that the various isomers are connected by relatively low-lying open-shell singlet (OSS) biradical species, making it difficult to prepare one isomer selectively in a hot environment. Steinbauer et al. 23 also reported an ionization resonance at 8.19 eV in the m/z 89 ms-TPES and attributed this to ionization of FArad to its X̃+ 1 A 1 ground state.
Li et al. 25 attempted to determine the isomeric composition of C 7 H 6 species detected in combustion and flame studies. Hansen et al. 13 found that C 7 H 6 was formed in a fuel-rich cyclopentene flame by synchrotron-based photoionization mass spectrometry. With a lack of experimental ionization energies (IEs) for most C 7 H 6 isomers, they computed adiabatic IEs to assign the photoionization signal. They concluded that the measured ionization threshold at 8.23 eV can be attributed to either FA or 1ECP, and because of their somewhat higher computed ionization energy, it was concluded that 2ECP and 5ECP do not contribute to the m/z 90 signal. Evidently, the differences in the published experimental IEs for FA and the proximity of the computed ionization thresholds of the isomers make it hard to base the isomer-specific assignment of C 7 H 6 on the ionization energy alone.
Because of spectral congestion, the detailed and wellresolved characterization of the TPES of the C 7 H 6 isomers and the resonance stabilized C 7 H 5 • radical in general and the identification of excited-state spectral signatures in particular are crucial to identify these species in the gas phase. In the past, we have shown that this approach works well for complex chemical systems where many isomers may be formed. 26 In this work, we revisit the photoelectron spectroscopy of the C 7 H 6 isomers and of FArad, formed upon pyrolysis of phthalide, by combining double imaging photoelectron photoion coincidence (i 2 PEPICO) experiments with quantum chemical computations and resolve contradicting assignments in the literature. Pyrolysis time-of-flight (TOF) mass spectra and photoion mass-selected threshold photoelectron (ms-TPE) spectra are recorded for various pyrolysis temperatures. Isomeric products are assigned by comparing the measured TPE spectra to Franck−Condon simulations. The C 7 H 6 potential energy surface is explored, and a simplified reactor model is presented to explain the observed products and point out the remaining open questions.
■ METHODS Experimental Section. The experiments have been performed using a pyrolysis microreactor connected to the double imaging photoelectron photoion coincidence endstation (CRF-PEPICO) at the vacuum ultraviolet (VUV) beamline of the Swiss Light Source. Details about the beamline, 27,28 the endstation, 29 and the Chen-type pyrolysis reactor 30 can be found elsewhere, and only a brief summary of the experiment is provided here.
The phthalide sample (98%, Sigma−Aldrich) was put in a 1 / 4 in. tube with a 100 μm pinhole (Lennox Laser) between a small aluminum pellet and glass wool. The tube was mounted in a temperature-controlled copper block in the source chamber and heated to 343 K to ensure sufficient vapor pressure of the sample. The pyrolysis microreactora ca. 2 cm long and a 1 mm internal diameter SiC tube resistively heated using a DC power supplywas mounted in front of the pinhole. The phthalide vapor was picked up by a 20 sccm flow of argon. The gas passed through the tube and expanded into the hot pyrolysis microreactor through the pinhole at the end. The temperature is estimated from the heating power based on prior calibration measurements. There is a large uncertainty (±100 K) in the absolute temperature inside the reactor that is derived from the heating power, but the relative temperatures between the various measurements reported here are accurate.
The surviving phthalide precursor and the pyrolysis products exit the microreactor with the argon flow into the source chamber, which is kept at a pressure of 6 × 10 −5 mbar by one cryogenic and two turbomolecular pumps. The resulting molecular beam is skimmed using a 2 mm diameter skimmer from Beam Dynamics, before it enters the detection chamber, which is kept at a background pressure of 2 × 10 −6 mbar by a turbomolecular pump. The synchrotron radiation is generated using a bending magnet, dispersed with a 150 grooves/mm blazed grating, and passed through a differentially pumped gas filter containing a mix of krypton, argon, and neon to remove light resulting from higher grating orders. The light is focused onto a 200 μm exit slit, located before the detection chamber, resulting in a resolution of 8 meV at a flux of ca. 10 12 photons/ s in the studied energy range.
The molecular and the VUV beams intersect in the detection chamber. Ions and electrons formed by photoionization of species entrained in the molecular beam are extracted in opposite directions by a constant, 218 V cm −1 electric field and are detected in delayed coincidence by Roentdek delay line detectors. 29 The electrons are velocity map imaged to measure their kinetic energy. Thanks to their negligible flight time, they also provide the start time for the time-of-flight (TOF) measurement of the coincident ions. The photoionization mass spectrum is constructed by plotting the ion counts as a function of TOF for all electrons, i.e., regardless of the electron energy. To plot the ms-TPES, only electrons imaged onto the center of the detector are selected for the TOF start signal. The center signal is still contaminated by nonthreshold electrons that happen to have no off-axis momentum component. This hot electron contamination was approximated based on a ring area around the center spot and subtracted from the center signal to obtain the threshold photoionization signal. 31 Computational Methods. The experimental data are supported by computational studies using Gaussian 16 and Q-Chem 4.3. 32,33 To identify the isomers that may ionize at a given photon energy, adiabatic ionization energies were computed using the CBS-QB3 composite method first. 34 Vertical ionization energies were also calculated at the EOM-IP-CCSD/cc-pVTZ//B3LYP/6-311++G(d,p) level of theory. 35 CBS-QB3 adiabatic ionization energies are reported with and EOM-IP results are reported without zero-point corrections. Franck−Condon simulations can be used to establish isomer-specific vibronic fingerprints unambiguously, even when they arise in the midst of broader progressions in a mix of isomers. We carried out geometry optimizations and frequency calculations using density function theory (DFT) for the ground neutral and cation states and time-dependent (TD-)DFT calculations for selected excited states to assign the ms-TPES features in an isomer-selective manner. Franck−Condon factors were calculated in the double harmonic approximation including the Duschinsky rotations using ezSpectrum. 36 The simulated stick spectrum was convoluted with a Gaussian function with a full width at half-maximum (fwhm) of 160 cm −1 to account for the rotational broadening and the resolution of the experiment and then compared with the recorded ms-TPES.
To rationalize the pyrolysis and isomerization reactions, we explored the potential energy surface (PES) at the B3LYP/6-311++G(d,p) level of theory using constrained optimizations and internal coordinate scans. On the basis of CBS-QB3 energetics, we also discuss a simplified reactor model to account for the observed product distributions.

■ RESULTS AND DISCUSSION
Mass Spectrometry. Temperature-dependent pyrolysis mass spectra show that phthalide (m/z 134) pyrolysis sets in at 1030 K and leads to C 7 H 6 formation at m/z 90 ( Figure 2): As shown by Steinbauer et al. 23 and evidenced by the strong m/z 105 peak in the electron ionization mass spectrum of phthalide, 37 dissociative photoionization (DPI) of phthalide readily yields m/z 105 via: Dissociative photoionization is enhanced in pyrolysis by the incomplete vibrational cooling of the precursor in the molecular beam. 38 At a reactor temperature of 1130 K, the parent molecule gets depleted nearly completely, as is evidenced by the simultaneously weakened m/z 105 and 134 peaks. As the temperature is increased further to 1166 K, a signal grows in at m/z 89, which is ascribed to hydrogen loss from the primary pyrolysis product at m/z 90: Identification of the m/z 90 Products. Mass-selected threshold photoelectron spectra were recorded at various reactor temperatures to reveal if there is a change in the isomeric composition of the pyrolysis products. Overview TPE spectra are available in the Supporting Information ( Figure S1) and we focus on the energy ranges with resolved vibronic structure here. First, the ms-TPE spectra recorded at two temperatures, 1030 and 1166 K, are shown in Figure 3 in the 8.1−9.0 eV photon energy range. There is a clear dependence of the photoelectron spectrum on the pyrolysis temperature. In the low-temperature spectrum, a single sharp resonance is seen at 8.22 eV followed by a series of weaker vibronic resonances at higher energy. This sharp resonance was previously detected at 8.22 eV and ascribed to FA. 23 When the temperature is increased to 1166 K, an additional, strong resonance appears at 8.26 eV, and further subtle changes are apparent in the shape of the TPES, particularly at 8. 44 CBS-QB3 calculations have been performed to establish the adiabatic ionization energies (IEs) of the C 7 H 6 isomers (Table  1). Based on these calculations, FA or 1ECP could both be responsible for the strong origin transition observed in the spectrum at 8.22 eV. It is only with the help of Franck− Condon simulations that we can establish ionization to the ground 2 A 2 state of FA as the main carrier of the m/z 90 spectrum (purple simulation in Figure 3), because of the strong resonance at ca. 8.4 eV. This assignment confirms that   23 CBS-QB3 predicts an ionization energy of 8.25 eV for 1ECP, which makes this isomer a prime candidate to explain the changing vibrational structure as a function of pyrolysis temperature. The simulated TPE spectrum of 1ECP is shown in Figure 3b, and it is evident that the strongest resonance in the 1166 K spectrum at 8.28 eV can be attributed to this isomer. The agreement between the experimental and the Franck−Condon simulated spectrum is further improved by the inclusion of a third isomer, 2ECP, which has a calculated IE of 8.49 eV. This is mainly due to the reproduction of the weak, but clearly visible peaks at 8.48 and 8.57 eV. The fourth C 7 H 6 isomer, 5ECP, has a calculated IP of 8.71 eV and is found to account for the weak resonances above 8.7 eV. Due to the mobility of the hydrogens in cyclopentadienyl ring moiety, sigmatropic rearrangements are likely to occur and are responsible for the observation of all three ECP isomers, as was shown previously. 40,41 A spectrum composed of the sum of the four contributing isomers is shown in red in Figure 3 and agrees well with all main resonances. The experimental ionization energies are obtained by shifting the onset of the Franck−Condon simulations to best fit the recorded ms-TPE and are listed in Table 1. To our knowledge, this spectral analysis yields the first experimentally determined ionization energies for 1ECP, 2ECP, and 5ECP. Note that the values listed in Table 1 have been corrected for the 0.01 eV Stark shift in the extraction field, while the spectra shown in this manuscript have not been corrected.
The high energy part of the m/z 90 ms-TPE spectrum recorded at 1166 K and stretching from 9.8 to 10.9 eV is shown in Figure 4. This part of the spectrum reveals vibronic structure that results from ionization to an excited state of one of the identified products (vide supra). Contributions by excited states of FA can be ruled out, as no resolved substructure is observed in the 1030 K TPE spectrum, fully attributed to FA ( Figure S1). Moreover, the ionization resonance to the ground state of 1ECP has a much stronger ground state transition than those of 2ECP and 5ECP, which makes it more likely that 1ECP features dominate the highenergy part of the spectrum, as well. EOM-IP-CCSD/cc-pVTZ calculations predict the vertical ionization energies to the first excited states of 1ECP and 2ECP at 10.47 and 10.34 eV, respectively. EOM-IP-CCSD geometry optimizations have failed to converge but suggest that the adiabatic ionization energies are at least ca. 0.1 eV lower. This was confirmed using the TD-DFT optimized geometry of the Ã+ 2 A′ state of 1ECP. Furthermore, vibrational analysis of the TD-DFT optimized geometries of the first excited states of 1ECP and 2ECP allowed for Franck−Condon simulation of this band. As shown in Figure 4, the simulated Ã+ 2 A′ 1ECP band matches the measured vibronic resonances very well. The Franck−Condon simulation of the Ã+ 2 A″ 2ECP band appears to account for the red wing of the main transition, but this assignment is less certain. Aside from these states, vertical excitations are predicted by EOM-IP-CCSD at 9.28 and 10.23 eV to the Ã+ 2 B 1 and B̃+ 2 B 2 states of FA, respectively, as well as at 10.86 and 10.63 eV to the B̃+ 2 A″ state of 1ECP and B̃+ 2 A′ state of 2ECP, respectively. There are two pronounced bands in the ms-TPES at ca. 9.2 and 10.1 eV in the low-T ms-TPES (see   Figure S1), corresponding to the energy range of the Ã+ and B̃+ states of the FA cation, respectively. These bands lack fine structure, which suggests that they undergo significant geometry relaxation and/or are short-lived. Similar behavior was also found for fulvenone, which exhibits the same cyclopentadienyl moiety and exhibits a structureless Ã+ state band. 42 B̃+ states of the 1ECP and 2ECP cations appear to be beyond the energy range studied herein. It is interesting to compare our m/z 90 ms-TPES (IE = 8.23 eV) with the He I photoelectron spectra of Botter et al. 22 (IE = 8.22 eV) and of Muller et al. 18 (IE = 8.29 eV). We agree with the overall assignment of Muller et al. 18 that the X̃+ 2 A 2 , Ã+ 2 B 1 , and B̃+ 2 B 2 states are found in the 8−11 eV energy range. However, our FA threshold photoelectron spectrum exhibits a strong ground-state band, followed by weak and broad peaks belonging to the excited state. In contrast, the He I spectrum shows three similarly intense bands. The m/z 90 ms-TPES of Steinbauer et al. 23 (IE = 8.22 eV) also displays a drop of signal intensity, although, based on their m/z 89 results (vide infra), it appears that the signal intensity dropped at higher photon energies because of an experimental artifact. More importantly, both He I photoelectron spectra exhibit a three-membered vibronic progression in the ground-state band, where the peak heights are by and large comparable. Our FA TPES, on the other hand, exhibits a ground-state band structure with only two strong resonances. Furthermore, we could assign the relatively sharp peak at ca. 10.1 eV unambiguously to an isomer of FA, because it is completely absent in the T = 1030 K spectrum. Its presence in the He I photoelectron spectra suggests that Botter et al. 22 and Muller et al. 18 studied a mix of C 7 H 6 isomers instead of a sample of pure FA. This is not entirely surprising, as they synthesized and separated bulk FA and introduced it in the spectrometer, assuming that it maintained its isomeric purity, which may not have been entirely true.
The Fulvenallenyl Radical: m/z 89 ms-TPES. An overview spectrum showing the ms-TPE spectra of the m/z 89 mass channel is available from the Supporting Information ( Figure S2). In Figure 5, only the (low-energy) region between 8.1 and 9.0 eV (top panel) and the (high-energy) region between 10.1 and 11 eV (bottom panel) are highlighted, as this is where vibrationally resolved transitions are observed. Two strong and resolved resonances are seen in the low-energy part of the spectrum at 8.20 and 8.32 eV. The former was previously attributed to ionization to the ionic ground state ( 1 A 1 ) of the resonance-stabilized FArad (C 7 H 5 • ). 23 Indeed, the simulated spectrum of the singlet ground state of the fulvenallenyl cation, shown in red in Figure 5, matches the first two resonances seen at 8.20 and 8.26 eV very nicely. Curiously, the second strong resonance at 8.32 eV and the strong and broad feature centered around 8.5 eV were not reported before. We believe that dropping overall sample levels or VUV flux may be a reason for the missing signal above 8.3 eV in the spectrum of Steinbauer et al. 23 This structure can only be explained if we consider the lowest-lying ã+ 3 B 2 triplet state of the FArad cation, the Franck−Condon simulation of which is plotted in blue in Figure 5. We note that EOM-IP-CCSD calculations predict the first singlet excited state Ã+ 1 B 2 to be near-degenerate with the first ã+ triplet state. However, TD-DFT calculations have not converged for the Ã+ 1 B 2 state, which implies strong coupling with the X̃+ state. Thus, the actual low-energy band (top panel Figure 5) may contain three contributing states.
There is vibronic structure in the 10.1−11.0 eV range of the m/z 89 ms-TPES of C 7 H 5 • , as shown in Figure 5b. Our  The Journal of Physical Chemistry A pubs.acs.org/JPCA Article tentative assignment, based on triplet TD-DFT frequency analysis, Franck−Condon simulations, and EOM-IP-CCSD calculations, suggests that transitions to b̃+ 3 A″ ( 3 A 2 in C 2v ) and c̃+ 3 A′ ( 3 A 1 in C 2v ) are seen in this energy range. Vertical ionization energies are 10.61 and 10.78 eV to these states, and adiabatic ionization energies, calculated at the EOM-IP-CCSD/cc-pVTZ level at the TD-DFT optimized C s geometries, were found to be 10.50 and 10.67 eV, respectively. The TD-DFT energies of these states are within 30 meV of each other, but the different symmetries and geometries, as well as the EOM-IP results, confirm that two different states are at play. This part of the spectrum is, thus, best reproduced by a combined b̃+ and c̃+ triplet simulation of FArad, which also yields ionization energies to these states (see Table 2). The vertical ionization energy to the C̃+ 1 A 2 singlet state lies also in this energy range at 10.74 eV. TD-DFT geometry optimizations have failed for this state, which may indicate strong coupling with the lower-lying singlet states. This would imply a short lifetime and a broad and structureless contribution of this state to the TPES. Thus, it is likely that this state contributes only to the baseline of the ms-TPES in the 10.1−11.0 eV energy range. Potential Energy Surface Calculations. Internal coordinate scans and constrained optimizations were carried out at the B3LYP/6-311++G(d,p) level of theory to explore the potential energy surface (PES) and rationalize the detected products. The formation of the C 7 H 6 species INT1 by CO 2loss from phthalide, as well as the reaction steps leading to products are described in detail in the Supporting Information.
Here, we only focus on the rate-limiting transition states that connect the initially formed INT1 to the detected pyrolysis products. A summary of the PES is shown in Figure 6.
INT1 is formed by CO 2 -loss from the hot phthalide precursor. Unrestricted density function theory calculations with a symmetry-broken initial guess confirm it to be an openshell singlet (OSS) species with ⟨S 2 ⟩ = 1.02. The energy of INT1 shown in Figure 6 is determined by applying the Ziegler−Cramer correction, 43 similarly to a recent study by Kvaskoff et al. 24 INT1 is the only OSS singlet species in Figure  6. The lowest barrier from this intermediate is TS1 at 231 kJ mol −1 leading to fulvenallene (FA). Alternatively, INT1 can cross TS13 located at 276 kJ mol −1 and form 5ECP. This transition state is lower than the one found to connect INT1 and 5ECP in the earlier study by Kvaskoff et al. 24 Hydrogen migration over barriers TS11 and TS12, located at 143 and 131 kJ mol −1 , respectively, result in 1ECP and 2ECP. The latter two products are the two most energetically stable isomers on this branch of the reaction. The fulvenallenyl radical (C 7 H 5 • at an energy of 333 kJ mol −1 (shown with the dotted line in Figure 6) can form from the reaction products shown in blue on the PES via hydrogen abstraction without a reverse barrier. The PES provides valuable qualitative insights into the formation of products as a function of energy and, thus, temperature. The channel to FA opens up first, and the formation path to ECP and eventually FArad only opens up when the energy of the system is increased further. Simplified Reactor Model. The residence time in the reactor is on the order of 100 μs, which is sufficiently long for species to rethermalize multiple times. 44 In order to establish what the thermal equilibrium in the microreactor would look like, we refined the DFT-calculated energetics in Figure 6 using the CBS-QB3 method and arrived at 0, 40, 10, 14, and 334 kJ mol −1 for FA, 5ECP, 1ECP, 2ECP, and FArad + H, respectively. On the basis of these energies and the harmonic state functions, we calculated the Boltzmann distribution and the isomer distribution of the C 7 H 6 species as a function of temperature. However, once C 7 H 6 is energetic enough to lose a hydrogen atom, it may form C 7 H 5 irreversibly. To account for unimolecular fragmentation, we introduce a rethermalization cycle count; in each cycle, the population of C 7 H 6 species above the H-loss threshold dissociates and contributes to the C 7 H 5 abundance. Assuming equal photoionization cross sections, the population of FArad is ca. one-third at 1160 K as observed in Figure 2. To reproduce this value, we arrived at a rethermalization cycle count of 390. If the remaining C 7 H 6 population is in thermal equilibrium, the product ratios should correspond to Figure 7 as a function of reactor temperature. The model reproduces the quickly rising m/z 89 signal with rising temperature in Figure 1. However, in contrast with the large 1ECP signal seen in Figure 3, the ECP species are always minor products with a plateauing abundance at ca. 1000 K, and the more stable FA should always dominate the spectrum. Therefore, FA and the ECP products are not in thermal equilibrium but are already disconnected below the H-loss barrier ( Figure 6). Branching from INT1 is determined by kinetic effects, and the large stabilization of the C 7 H 6 products formed results in slow isomerization between FA, INT1, and the ECP species in the narrow energy window below the Hloss threshold. Thus, we can show that although various C 7 H 6

■ CONCLUSIONS
We pyrolyzed phthalide and recorded mass spectra as a function of temperature to identify pyrolysis products at m/z 90 and 89. The experiments revealed a richer phthalide pyrolysis chemistry than previously reported. It furthermore led us to rediscover C 7 H 6 isomerism, as previously unveiled by D'Amore et al. 19 Photoion mass-selected threshold photoelectron spectra of the C 7 H 6 channel were used to identify the temperature-dependent contributions of the FA and three ECP isomers. The latter are newly assigned pyrolysis products of phthalide. The C 7 H 5 • radical can be formed by H loss from any of the C 7 H 6 isomers, and its photoelectron spectrum has also been recorded. Accurate ionization energies were determined for FA, 1ECP, 2ECP, and 5ECP and were compared to computed adiabatic ionization energies. A vibrationally resolved excited-state band was detected around 10.1 eV and was attributed to the sum of the first electronically excited state of 1ECP and 2ECP. The ionic ground state (singlet) of the FArad radical was found at 8.20 eV and the triplet ground state was found to be 0.13 eV higher in energy. Furthermore, a vibrationally resolved excited-state band in the photoelectron spectrum of C 7 H 5 • was assigned to the first and second excited triplet states. The temperature-dependent formation of the various products was rationalized by complementary quantum chemical computations. A simplified reactor model shows that irreversible hydrogen atom loss from C 7 H 6 may stop the evolution of thermal equilibrium, and the branching ratio between the FA and ECP products is determined by kinetic effects instead.
Temperature-dependent ms-TPES, ground-and excitedstate calculations as well as Franck−Condon models helped assign the congested ground-state band of C 7 H 6 and the excited-state peaks in an isomer-selective manner. They also helped us determine the singlet and triplet state contributions to the FArad photoelectron spectrum. This illustrates how careful analysis of PEPICO data can contribute to an isomerspecific understanding of reaction processes as well as providing reference data for analytical purposes. Different isomers may exhibit different reactivity with respect to PAH growth, as illustrated by the dependence of ring condensation of C 7 H 6 to form heptafulvalene on the reaction conditions, originally reported by D'Amore et al. 19 Thus, isomer-selective detection of elusive PAH precursor intermediates may help us understand PAH growth and soot formation in more detail.
Temperature evolution of the overview spectra of the m/ z 90 product and an overview TPE spectra of the m/z 89 radical product, the potential energy surface of phthalide dissociation, and the full C 7 H 6 potential energy surface (PDF) Figure 7. Predicted phthalide pyrolysis product distribution as a function of temperature assuming that C 7 H 6 species above the H-loss barrier lose hydrogen irreversibly in 390 rethermalization cycles, while the remaining C 7 H 6 product distribution is in thermal equilibrium.