Jahn-Teller distortion and dissociation of CCl4+ by transient X-ray spectroscopy simultaneously at the carbon K- and chlorine L-edge

X-ray Transient Absorption Spectroscopy (XTAS) and theoretical calculations are used to study CCl4+ prepared by 800 nm strong-field ionization. XTAS simultaneously probes atoms at the carbon K-edge (280–300 eV) and chlorine L-edge (195–220 eV). Comparison of experiment to X-ray spectra computed by orbital-optimized density functional theory (OO-DFT) indicates that after ionization, CCl4+ undergoes symmetry breaking driven by Jahn–Teller distortion away from the initial tetrahedral structure (Td) in 6 ± 2 fs. The resultant symmetry-broken covalently bonded form subsequently separates to a noncovalently bound complex between CCl3+ and Cl over 90 ± 10 fs, which is again predicted by theory. Finally, after more than 800 fs, L-edge signals for atomic Cl are observed, indicating dissociation to free CCl3+ and Cl. The results for Jahn–Teller distortion to the symmetry-broken form of CCl4+ and formation of the Cl–CCl+3 complex characterize previously unobserved new species along the route to dissociation.


Introduction
The Jahn-Teller (JT) theorem 1 states that degenerate electronic states in non-linear molecules cannot be minima of energy vs. nuclear positions and will therefore undergo nuclear displacements that break the degeneracy. This is oen observed in highly symmetric molecules with partial occupation of symmetry equivalent orbitals. A prototypical example of JT distortion is CH 4 + , which arises from ionization from the triply degenerate highest occupied molecular orbital (HOMO) of CH 4 . The resulting cation rapidly distorts away from the tetrahedral (T d ) geometry of the neutral to a C 2v symmetry structure 2,3 with two long and two short CH bonds. Baker et al. 4 have attempted to measure the time required for the JT process via comparison of high harmonic generation in CH 4 + to CD 4 + , where they report a difference between the relaxation times of geometries of the two molecules up to 1.6 fs with sub-cycle resolution. However, information about longer times was limited by the wavelength of the laser. Recently, Gonçalves et al. 5 calculated the JT timescale in CH 4 + and found that distortion to C 2v occurs as early as 7 fs, while the heavier CD 4 + takes longer to relax to the same structure. Ionization of halogenated methanes is expected to lead to slower (and thus easier to measure) JT distortion, due to the larger substituent masses. However, JT distortion has not been observed for such systems, as the resulting cations are quite short-lived and therefore difficult to study. 6 Indeed, it has been assumed that dissociation of the T d conformation of CCl 4 + to CCl 3 + and Cl occurs without any intermediates, 7,8 as no stable ion is produced by direct photoionization. 9 Theoretical work 8,10 has suggested a stable structure of the form [Cl 3 C-Cl] + , with one Cl 3.5Å away from the C and weakly coordinating to it. However, there are no experiments that validate the existence of these possible CCl 4 + complexes.
The instability of CCl 4 + necessitates the use of a technique that is sensitive to changes in nuclear and electronic structure occurring within the rst few femtoseconds aer formation. Xray Transient Absorption Spectroscopy (XTAS) is such a method, as its time resolution is only limited by the duration of the pump and probe laser pulses and femtosecond resolution is readily achievable. [14][15][16] Following strong-eld ionization by the pump, the X-ray pulse induces transitions from core levels to unoccupied orbitals, which ensures that XTAS is sensitive to the local environment of individual atoms in the probed molecular system. 15,17 It is therefore well suited for probing possible JT distorted transient intermediates that may arise during a dissociation. Indeed, XTAS has previously been used to infer JT distortion in the benzene cation 18 and ring opening in cyclohexadiene. 19 Pertot et al. 20 also employed XTAS at the C Kedge to study the dissociation of CF 4 + , reporting rapid dissociation to CF 3 + and F within $40 fs.
While several aforementioned studies focused on the C Kedge, HHG X-ray pulses span a relatively broad energy range 14,21,22 and can be used to study multiple atomic edges simultaneously. In particular, the Cl L 2,3 -edge (2p levels, $195-220 eV) is a natural complement to the C K-edge (1s level, $280-300 eV) for studying chlorinated methanes. Signal from two elements potentially allows for the identication of species whose spectra might be unresolvable or involve forbidden transitions at any one particular edge. It also permits observation of any Cl dissociation both from the perspective of the departing atomic Cl species and the remaining C-containing molecular fragment.
In this paper, we present the results of a time-resolved experimental and theoretical investigation of strong-eld ionized CCl 4 , by tabletop XTAS simultaneously at both the C K-edge and the Cl L 2,3 -edge. We report evidence for a dissociation pathway of CCl 4 + , illustrated in Fig. 1, that is initiated by an ultrafast ($6 fs) JT distortion driven symmetry breaking, away from the parent tetrahedral (T d ) geometry to an ensemble of symmetry-broken covalently bonded structures where all four Cl atoms remain covalently bonded to C. All these bonded forms can be viewed as vibrationally excited C 2v CCl 4 + , as this geometry is predicted by theory to be the only stationary point on the cation potential energy surface with four C-Cl covalent bonds. Theory further indicates that the ions are only temporarily trapped in the bonded form, based on ab initio adiabatic quasiclassical trajectory (QCT) calculations and computed X-ray spectral features from orbital optimized density functional theory (OO-DFT 23 ) that can be directly compared to experiment. The experiments corroborate the computed spectral features and identify the timescales for the initial symmetry breaking. Similarly, experiment and theory synergistically describe the evolution of the transient covalently bonded cation to a noncovalently bound complex between CCl 3 + and Cl. The latter structure ultimately undergoes irreversible dissociation over relatively long timescales ($800 fs). We also nd evidence for a higher energy channel that results in CCl 3 and atomic Cl + , which is expanded upon in the ESI. †

Methods
The tabletop experimental apparatus was previously described in ref. 14. We provide a brief summary that also touches on relevant upgrades. The pump pulse that induces strong-eld ionization is centered at 800 nm, spectrally broadened, and compressed to about 6 fs in duration (as described in the ESI †). The maximum energy of the compressed pulse at the sample is 150 mJ at 1 kHz repetition rate. It is focused to 65 mm full width at half maximum (FWHM) to achieve an electric eld with peak intensity up to 3 AE 1 Â 10 14 W cm À2 , as estimated by numerical calculations and by observed relative ionization rates of Ar, as explained in the ESI. † The X-ray probe is generated by high harmonic generation (HHG), using a 10 fs pulse, centered at 1300 nm, focused into a semi-innite gas cell lled with 2.3 bar He. This tabletop setup provides X-ray photons with energies up to 370 eV. The X-ray monochromator has 0.2 eV spectral resolution at the C K-edge Fig. 1 The sequence of computationally characterized intermediates for the dissociation of CCl 4 + identified in this experiment. Geometries were optimized with uB97M-V 11 /aug-pcseg-3. 12 Zero-point energies were computed at the same level of theory. Electronic energies at these geometries were subsequently found using CCSD(T) 13 extrapolated to the complete basis set (CBS) limit. The structures of the closest local minima are used for the symmetry-broken covalently bonded form and non-covalently bonded complex; although, the experiment samples a large range of nuclear configurations, due to vibrational energy. The energy is set to 0 for the dissociated ion. Time constants shown are gathered from the DOD data in Fig. 2b and c and their lineouts (as described in the ESI †). We refer to the symmetry-broken covalently bonded form as SBCB and non-covalently bonded complex as NBC, in subsequent figures and tables. and the experimental data are measured as a change in absorbance, or DOD, which is separated into spectral components by multivariate tting, described in detail in the ESI. † The temporal cross-correlation of the experiment is measured to be 8 AE 2 fs by the autoionization in Ar L 2,3 lines 24 (as discussed in the ESI †). Experiments are performed over different time ranges to cover a wider range of dynamics; the shortest have a step size of 1 fs and extend to 80 fs, and the longest have variable step sizes and extend to 10 ps. Additionally, experiments are run with varying pump power, from $1À3 Â 10 14 W cm À2 , to assess how the power affects the temporal dynamics and nal states. These scans are taken as closely in time as possible to make the comparisons between scans more consistent. Timescales are extracted by tting lineouts in time to unimolecular kinetics (as described in the ESI †). Each t includes convolution with a Gaussian of 8 AE 2 fs FWHM to account for the crosscorrelation of the experiment, using the 95% condence intervals as error bars. CCl 4 was obtained from Sigma-Aldrich at 99.5% purity and was vaporized by exposing the liquid to vacuum at room temperature. The CCl 4 was probed in a nite gas cell with a 4 mm pathlength with a foreline pressure of 12 mbar.
Quantum chemical calculations were performed with the Q-Chem 5 soware. 25 Structures were optimized with the uB97M-V 11 density functional and the aug-pcseg-3 (ref. 12) basis set. Zero-point energies were found at the same level of theory. Relative ground state electronic energies at the optimized geometries were computed with CCSD(T) 13 extrapolated to the complete basis set (CBS) limit, as detailed in the ESI. † Ab initio adiabatic trajectory calculations on CCl 4 + were performed with uB97M-V/aug-pcseg-1, starting from the equilibrium CCl 4 T d structure and with quasiclassical velocities 26 for nuclei (this ensures each normal mode of the neutral species has the associated zero point energy). A total of 256 trajectories (out of the 512 possible ones for the CCl 4 molecule with 9 normal modes) were run. The trajectory calculations did not incorporate an external electric eld and thus can only provide a rst estimate of timescales that can be compared to experiment.
X-ray absorption spectra were simulated with OO-DFT, 23 utilizing the SCAN 27 functional, aug-pcX-2 basis 28 on the site of core-excitation, and aug-pcseg-2 basis 12 on all other atoms. This approach has been shown to be accurate to $0.3 eV root-meansquared error for the core-level spectra of electronic ground states of both closed-shell 29 and open-shell 30 species, without any need for empirical energy translation of the spectra. Ref. 29 and 30 provide detailed protocols for running OO-DFT calculations. Excited state orbital optimization was done with the square gradient minimization (SGM 31 ) and initial maximum overlap method (IMOM 32 ) algorithms, for restricted open-shell and unrestricted calculations, respectively.

General features of experimental spectrum
The ground state static spectrum of CCl 4 is presented in Fig. 2a. The Cl L 2,3 -edge comprises a number of peaks, several of which overlap, largely due to the C l2p spin-orbit splitting of 1.6 eV.
The dominant features correspond to excitations to the 7a * 1 and 8t * 2 symmetry adapted linear combinations (SALCs) of s * CCl antibonding orbitals. Transitions from the L 3 (2p 3/2 ) level to 7a * 1 and 8t * 2 occur at 200.4 eV and 203.2 eV respectively. 33 The corresponding L 2 (2p 1/2 ) transitions occur at 202.0 eV and 204.8 eV. All of these excitations are dipole allowed. Transitions to Rydberg levels lead to additional peaks (such as the one at 205.5 eV (ref. 33)), and a rising ionization edge begins and persists around 207.5 eV. We note that the L 3 ionization energy of CCl 4 from X-ray photoelectron spectroscopy is 207 eV. 34 In contrast, the C K-edge spectrum of neutral CCl 4 consists of a single intense peak at 290.9 eV, 35 followed by a rising edge at $295 eV. The 290.9 eV peak arises from the C 1s/8t * 2 excitation, as the transition to the lower energy 7a * 1 orbital is dipole forbidden. 35 OO-DFT predicts the C 1s/8t * 2 excitation to be at 290.8 eV, in excellent agreement with experiment, and the forbidden C 1s/7a * 1 transition is predicted at 289.0 eV. The HHG process leads to lower photon ux at the higher C K-edge energy range than the Cl L-edge, contributing to less signal-tonoise in the former. There is also signicant absorption from the tail of the chlorine edge in the experimental spectrum, which further reduces the signal-to-noise at the C K-edge.
Strong-eld ionization by the pump pulse induces dynamics that leads to differences in absorption (DOD), as shown in Fig. 2b and c. The many spectral overlaps at the Cl edge ( Fig. 2b) make it difficult to identify individual signals in the DOD data. Negative signal from ground state bleach is predominant from 200 to 210 eV with positive signal appearing in the 211-216 eV range, possibly due to a positive local charge on the chlorines leading to a net blue shi. The C edge is easier to resolve, with only a ground state bleach at 290.8 eV, and positive absorption at both higher ($292 eV) and lower (287-290 eV) energies.
Within these broad experimental signals, several notable features can be observed. The rst is a small positive feature at the Cl edge, labeled feature 1 in Fig. 2b. This feature appears transiently at early times and rapidly decays within a few femtoseconds. The second is at 209.5 eV, feature 2, which initially drops to a negative DOD but decays back towards zero values on the timescale of 100 fs. A similar timescale is observed in the decay of feature 3, at 289.2 eV, which suggests that the two decays measure the same process. Additionally, the positive signal of feature 3 initially rises on the same time scale as the decay observed in feature 1, suggesting that the early time behavior of features 1 and 3 reect the same process. A fraction of feature 3 undergoes a continuous evolution (289.2 to 287.1 eV) to feature 4 at 287.1 eV aer $20 fs. Feature 4 continues to grow aer that time, concurrent with the decay of feature 3. Finally, at delays of several hundred femtoseconds to a few picoseconds, much longer times than those shown in Fig. 2, a series of sharp spectral features between 204 and 207 eV become resolvable from the broader features, shown more clearly in Fig. 6 and described in Section III. † Based on the three timescales involved in the evolution of the noted features, few-femtoseconds, tens of femtoseconds, and several hundred femtoseconds, we conclude that three separate processes are responsible for the dynamics of the dissociation of CCl 4 + . In order to better understand the nature and quantitative times of these processes, these features are averaged and tted along the time axis, and assignments are made by comparison to theory, as discussed below. In particular, we note that our probe would not be able to directly track a hole in the Cl lone pair levels of pure 3p character. The Cl 2p / 3p transition is dipole forbidden, and the C 1s / Cl 3p process involves local sites on distinct atoms with essentially no overlap. Changes in electronic and nuclear structure arising from such Cl 3p holes would however lead to other transient features, as discussed later. Transitions from Cl 2s levels (L 1 edge) to a 3p hole are dipole allowed. However Cl L 1 edge transition peaks are very broad and not very intense (as shown through an example in the ESI †). We therefore did not pursue analysis in that regime ($270 eV).

General considerations for C K-edge
Any distortion away from T d geometries would alter the symmetry of the s * CCl levels, which would no longer form the 7a * 1 and 8t * 2 SALCs. The most general symmetry-broken case would be of C 1 symmetry, with four unequal bond lengths. This leads to four nondegenerate s* levels that permit dipole allowed transitions from the C 1s level, although intensities will be lower for the MOs with greater C 2s character. The shorter bonds would feature stronger C-Cl interactions, leading to higher energy s* MOs, while the longer bonds would conversely lead to lower energy s* levels. In particular, the lowest energy s * CCl level is expected to be dominated by the longest C-Cl bond, and therefore the lowest energy C K-edge feature in the transient absorption spectrum should correspond to this bond.
Conversely, the shortest bonds should lead to the highest energy absorption feature. More symmetric congurations can permit multiple bonds to make comparable contributions to any given s* MO, preventing assignment of a transition to one particular bond. In general, however, the longest bonds should lead to the lower energy features in the C K-edge spectrum, while the higher energy features should arise from shorter bonds. However, there is no such simple rule of thumb available for the Cl L-edges, due to the complexity of the spectrum.

Jahn-Teller distortion of CCl 4 +
Lineouts of the few-femtosecond process are shown in Fig. 3a.
These show the decay of feature 1 in 6 AE 2 fs and the rise of feature 3 in 7 AE 3 fs. These lifetimes, along with those extracted from the other ts, are compiled in Table 1 (with further details about the ts being provided in the ESI †). The ground state bleach at 200.4 eV from depletion of the neutral CCl 4 represents the instrument response function (i.e., the temporal broadening introduced by the experiment). Comparison of features 1 and 2 to the ground state bleach in Fig. 3a show these lifetimes are signicant beyond the experimental cross-correlation. We assume the strong-eld ionization process abruptly populates a CCl 4 + cation state or states in a Franck-Condon-like From theory, the closest local minimum to the T d starting structure is a C 2v symmetry distorted tetrahedron with two long (1.82Å) and two short (1.70Å) C-Cl bonds. The calculations indicate that a C 2v form of CCl 4 + is lower in energy by 0.4 eV, and there is no energy barrier between this minimum and the initial T d geometry. This energy stabilization is smaller than the 1.5 eV stabilization observed for the analogous CH 4 structure. 3 This result is unsurprising as the CCl 4 ionization is from Cl lonepairs while the CH 4 electron loss is from bonding orbitals. The energy stabilization from JT distortion will be available to the vibrational modes of the ion, so many different nuclear congurations will be accessible around the C 2v local minimum conguration. This range of congurations will be referred to as the symmetry-broken covalently bonded form of CCl 4 + (or covalently bonded, for simplicity). The OO-DFT C K-edge spectrum of the C 2v stationary point was computed in order to determine if the covalently bonded forms were contributing to feature 3. At the C 2v stationary point, all four s* SALCs are nondegenerate and the C 1s transitions to these SALCs are all formally dipole allowed. The two lower energy SALCs correspond to the long C-Cl bonds, which is computed to lead to absorption at 289.8 eV, and the shorter bonds lead to higher energy s* SALCs that are computed to absorb at $292 eV. A comparison of these energies in Fig. 4a with the absorption of feature 3, corrected for ground state bleach, shows that the energies match well. This validates the assignment that feature 3 arises from distorted CCl 4 + with four covalent bonds. In order to get an idea of the timescales of this distortion, Fig. 3b shows the longest C-Cl bond distance for a random subset of 100 calculated trajectories. The longest C-Cl bond is used here as a measure of distortion away from T d , for best comparison to the experimental X-ray absorption. It shows that at least one C-Cl bond rapidly elongates to the C 2v value of 1.82 A at $5 fs on average.
The $5 fs time from the trajectories is comparable to the 6 AE 2 fs decay of feature 1 and the 7 AE 3 fs rise of feature 3. Both time constants should measure the same process, the lifetime of the CCl 4 + T d geometry, with feature 1 more directly measuring the decay of the initial T d state and feature 3 measuring formation of the distorted state. They conrm the rapidity of the JT process and the barrierless energy surface between the two. We also note that on changing the pump intensity, the lifetimes exhibited no power dependence beyond what is necessary for ionization within the error bounds.

Covalent bond breakage of CCl 4 +
The evolution of some fraction of feature 3 to the much lower energy feature 4 in the experimental spectrum suggests further The longest C-Cl bond length for a set of 100 trajectory calculations, because that bond will correspond to the lowest energy absorption at the C K-edge. It shows that the bond elongates into the symmetry-broken covalently bonded form (labeled SBCB) rapidly, assisting the assignment that feature 1 corresponds to the T d . C-Cl bond stretching. OO-DFT calculations conrm this, with Fig. 4b showing the absorption of energy in CCl 4 + as a function of the longest C-Cl bond distance (with all other nuclear coordinates being optimized). Fig. 4b further explains the lower energy tail of the experimental covalently bonded form spectrum in Fig. 4a, as the experiment will sample many molecules spanning a wide range of covalently bonded geometries. Feature 4 at 287.1 eV absorbs at the same energies as an extremely elongated C-Cl distance ($3Å) in Fig. 4a and b, suggesting that feature 4 corresponds to a CCl 3 + moiety. Indeed, this feature corresponds to a transition to the unoccupied 2p orbital in CCl 3 + . This CCl 3 + orbital originates from the lowest energy s * CCl level of the covalently bonded form, which has decreasing contribution from the dissociating Cl atom as the C-Cl distance increases. The observed spectrum is thus consistent with the theoretical prediction that the lowest energy forms of An energy decomposition analysis (EDA 41 ) calculation reveals that the À0.15 eV interaction energy between atomic Cl and CCl 3 + in the minimum energy complex geometry is mostly (56%) from the polarization of the atom by the cation, with a much smaller amount (19%) arising from charge-transfer (the remaining 25% being permanent electrostatics, Pauli repulsion and dispersion). The system effectively acts as a charge-induced dipole complex. It is also worth noting that these noncovalent complex structures do not have any covalent Cl-Cl interactions, contrary to some early assignments. [42][43][44] The timescale of the symmetry-broken covalently bonded CCl 4 + species to Cl-CCl 3 + complex transition is shown through lineouts of features 2, 3, and 4 in Fig. 5a. This shows decay times of features 2 and 3 in 90 AE 10 fs and 80 AE 30 fs, respectively. Feature 4 appears only aer a delay of 23 AE 8 fs, which implies that the nuclear motions take at least 23 fs to get to the point that the ion starts to resemble its dissociated form. Aer this delay, it has an additional lifetime of 50 AE 20 fs towards reaching its asymptotic value. Because of the delay from nuclear movement, the experimental time for the covalently bonded to noncovalent complex process is taken from the more precise feature 2, 90 AE 10 fs, and no pump power dependence is observed. This time is expected to be signicantly longer than the JT distortion due to the computed barrier of 0.17 eV for this process. In varying the pump power, we do not observe a difference in the covalently bonded population decay times, possibly due to the large intensity already required to ionize the CCl 4 . The trajectory calculations in Fig. 5b show that a portion of the ions continue their initial distortion and dissociate immediately, which is observed experimentally by the direct evolution of the feature 3 / feature 4 signal. Indeed, this regime of the experimental signal reports direct dynamics from the covalently Fig. 4 (a) The experimental features 3 and 4 as well as the neutral static are disambiguated and corrected for ground state bleach to obtain an absorption spectrum, using a method described in the ESI. † Notably, this shows that feature 3 DOD signal at 289.2 eV corresponds to an actual maximum at 289.8 eV. These spectra are compared to OO-DFT calculations for energies and oscillator strengths of the excitations (given by the peak position and heights, respectively). This allows for assignment of feature 3 to the symmetry-broken covalently bonded form (labled SBCB) and feature 4 to the noncovalent Cl-CCl 3 + complex (labeled NBC) or CCl 3 + moeity. (b) OO-DFT absorption spectra for various CCl 4 + geometries. The third plot from the top shows absorption as a function of a single increasing bond distance (and other nuclear coordinates being optimized with this constraint). This shows splitting of absorption energies with the lower energies corresponding to the longest bond and higher energies to the shortest. The region of maximum bond extension without cleavage is roughly shown in the symmetry-broken covalently bonded region. These longer bonds account for the low energy tail in the covalently bonded experimental spectrum.
bonded form to the complex for a fraction of the molecules. Another portion remains trapped in the covalently bonded form for longer than a vibrational period, showing longest C-Cl distances between 1.8 and 2.05Å, which serves to dene the symmetry-broken covalently bonded region. We however note that the trajectory calculations show a signicant population persisting in the covalently bonded form for a longer time than the experimentally measured lifetime. This likely arises from the trajectory calculations not including the excess vibrational energy added by the pump pulse. The calculations only include the zero-point energy of neutral CCl 4 , leading to the covalently bonded form persisting longer than what is experimentally observed. A further discussion about the long time behavior of the trajectories is given in Section IX of the ESI. †

Atomic Cl dissociation
Although the C K-edge cannot distinguish between noncovalent complex and free Cl, the Cl L 2,3 edge can make this distinction, as atomically sharp Rydberg lines corresponding to atomic Cl become evident upon completion of the dissociation, which is shown in Fig. 6a and b. Very little change is observed in the DOD spectra between 250 fs and 3.5 ps at either the C K-edge or chlorine L 2,3 -edge, other than the appearance of these sharp lines, which is consistent with the previous assignment of the noncovalent complex. While weakly bound in the complex, the atomic lines are broadened by a combination of two factors. Firstly, the transitions are already of low intensity due to their Rydberg character, the lowest energy lines being of 2p / 4s character. Secondly, the diffuse nature of the Rydberg levels showing the longest C-Cl bond distance. It shows a population that temporarily remains in the covalently bonded form with longest bonds going between 1.8 and 2.05Å. It also shows a portion of ions have sufficient velocity in the appropriate chlorine to break the C-Cl covalent bond immediately, while others require additional time to redistribute the energy such that the bond may be broken. The trajectories show more population remaining in the covalently bonded form for a longer time, but the difference is likely due to excess vibrational energy in the experiment or another particular of the strong-field ionization process, which are not included in the trajectories. See Section IX of ESI † for a discussion about the long time behavior of trajectories. Fig. 6 (a) Lineouts in energy for 220 fs and 2.5 ps at the Cl L 2,3 -edges. They show that although the overall shape and intensity of the DOD signal remains constant, sharp spectral features appear at longer delays. The difference between 3.5 ps and 220 fs is taken and added to an error function to obtain the spectrum in b. (b) A comparison of the sharp spectral features is made to atomic Cl 39 and to molecular Cl 2 , 40 which shows that the features match well with atomic Cl. (c) A high-pass spectral filter is applied to the DOD data and an average of the two most prominent lines at 204.2 and 206.6 eV is shown as a lineout against time. A fit is taken starting from t ¼ 0, due to noise preventing a better determination of a start point. This fit with $3 Â 10 14 W cm À2 pump intensity has a lifetime of 800 AE 200 fs. means that the excitation energy would be sensitive to the precise position of the CCl 3 + entity, and the atness of the complex ground state potential energy surface permits a large range of accessible nuclear congurations. However, over time, the complex irreversibly dissociates, leading to atomic Cl lines at 204.2 eV and upward from 206 eV, which agrees with previous experimental absorption data. 39 Thus, we can use the Cl atomic lines to track the time it takes for the complex species to completely dissociate to CCl 3 + and Cl. We note that molecular Cl 2 has absorption in this energy range as well; 40 however, comparison to experimental atomic absorbance in Fig. 6b shows much better agreement with atomic Cl. A high-pass lter is applied to the transient spectrum to separate the sharp atomic Cl lines from broader molecular signals. A simple exponential t of these lines in Fig. 6c gives a lifetime longer than 800 fs. The exponential starts at time zero, despite the Cl signal not clearly departing from 0 until $300 fs. This may correspond to the minimum time required for atomic Cl to move far enough to show the Rydberg lines, although the signal-to-noise of this experiment is not sufficient to make that distinction.
The 800 fs time scale is much longer than the 90 fs observed for noncovalent complex formation, despite the excess of energy available to the system. However, the interactions and thus rate of energy transfer between the CCl 3 + moeity and the neutral Cl are much weaker than a covalent bond. This situation is similar to van der Waals complexes, which sometimes show lifetimes exceeding milliseconds with vibrational energy in the molecular moeity. 45,46 The 800 fs value is the fastest Cl formation time and was measured from the most intense pump pulse of 3 Â 10 14 W cm À2 . Datasets collected at lower pump power show slower dissociation times, and some do not exhibit signicant atomic Cl formation up to 3 ps. The noncovalent complex signal from feature 4 at 287.1 eV is still present, suggesting that the strong-eld ionization may be capable of forming long-lived noncovalently bound CCl 4 + complexes; although times longer than a few ps are outside the scope of this paper. It has been observed that otherwise unstable parent ions can be generated by fewcycle strong eld ionizing laser pulses, such as tetramethyl silane and CS 2 , 47,48 and an analogous complex may be a potential explanation of these signals. A quantitative comparison of the power dependence on this timescale was not carried out. We also note that the lifetime of a molecular species against collision with an electron (released by strong-eld ionization) is on the order of $5 ps with our setup (as discussed in the ESI †), which is an additional factor to consider, but only at very long times.
4 The high intensity case and the formation of Cl + While the observed dissociation to CCl 3 + and Cl can be well explained by the intermediate forms and pathways discussed above, another channel appears to be present in the data. This channel is clearest at the chlorine L 2,3 -edge where very sharp lines with similar widths to the atomic Cl lines are observed at 214 eV with less obvious components at 212.6 eV, shown in the ESI. † These energies are in the range that the calculations predict for the Rydberg states (2p / 3d, 4s) of Cl + . These atomically sharp peaks begin to appear with a delay of 37 AE 6 fs relative to the onset of the main cationic CCl 4 signal, and from that point, they show a time to grow in of 85 AE 10 fs for the best signal-to-noise dataset. Low energy satellite features in the C 1s spectrum $285.5 eV are also observed, which can correspond to transitions to the singly occupied level of the CCl 3 radical, computed to be at 285.5 eV. These simultaneous features appear to suggest the existence of a channel that results in CCl 3 and Cl + . The formation of Cl + is a much higher energy channel, with nal energies about 4.8 eV higher than the normal dissociation channel. Cl + formation has been observed in previous CCl 4 ionization experiments in low quantities from electron impact, 49 singlephoton ionization 9,50 and in higher quantities from strong-eld ionization. 51 Further discussion about Cl + formation is provided in the ESI. †

Conclusions
In this paper, we have shown experimental and theoretical evidence for both a transient Jahn-Teller distorted symmetrybroken covalently bonded CCl 4 + , similar to the stable ion of CH 4 + , and a noncovalently bound complex between CCl 3 + and Cl. Neither of these intermediates have been observed previously in experiments. A summary of each of the time constants experimentally extracted is shown in Table 1. The transition from the initial tetrahedral ion to the C 2v state occurs in 6 AE 2 fs, showing that the Jahn-Teller distortion is very fast, on the order of the duration of the pump laser pulse. It provides an order of magnitude estimate for other symmetric molecules undergoing JT distortion, which can aid chemical simulations as well as help predict the vibrational energies and nuclear dynamics of those other systems, especially given that this symmetry-broken covalently bonded intermediate was not assumed to exist prior to this work. 7,8 The noncovalent Cl-CCl 3 + complex is bound largely by the polarization of the neutral species, forming an ion-induced dipole complex with a lifetime much longer than the covalent bond cleavage time. The nal dissociation time for free Cl or Cl + seems to be slightly dependent on the strong-eld power, but for the highest power used in this experiment, the lifetime is more than 800 fs (vs. 90 fs for noncovalent complex formation). This observed long lifetime is a potential explanation for otherwise unstable parent ions appearing in mass-spectra of ultra-short strong-eld ionization.