Photodissociation dynamics of methyl iodide across the A-band probed by femtosecond extreme ultraviolet photoelectron spectroscopy

In order to aid comparison with previously reported MPI measurements performed at pump wavelength of 269 nm and 255 nm, we initially consider the signal obtained at binding energies related to ionization of the FC geometry into the $\tilde {X}$ states of the cation as measured by both probes. The intensity profiles over these features for each pump wavelength are plotted in figure 4 and cover a binding energy range of 4.91–6.14 eV, 4.96–5.95 eV, 4.70–5.70 eV and 4.55–5.38 eV for the 279 nm (a), 269 nm (b), 254 nm (c) and 243 nm (d) data respectively. The error bars are derived from a bootstrapping analysis of the data and mark the 95% confidence intervals. The time-dependence of the integrated photoelectron intensity in these regions were fitted to a sum of exponentially decaying functions convolved with a Gaussian instrument response function appropriate for the pump wavelength used. The equation used had the form:

$$\begin{equation}I(t)=A\enspace {\mathrm{e}}^{-\lambda (t-{t}_{\text{offset}})}\left[1+\mathrm{e}\mathrm{r}\mathrm{f}\left(\frac{\left(t-{t}_{\text{offset}}\right.-{\sigma }^{2}\lambda }{\sqrt{2}\sigma }\right)\right],\end{equation} \tag{ 1 }$$

where I(t) is the time-dependent intensity, λ is the decay rate constant and σ is the pump–probe cross-correlation width equivalent to FWHM$/2\sqrt{2\enspace \mathrm{l}\mathrm{n}\enspace 2}$. t_offset is a temporal offset that defines the delay in reaching a maximum intensity. This is measured relative to the experimental time-zero which is defined as the time at which the pump and probe are temporally overlapped. The results of the fits are given in table 1 with the decay represented by the lifetime (τ = 1/λ) rather than the associated decay constant (λ). For the 279 nm, 269 nm and 243 nm pump wavelengths this only required a single exponential term. However, the 254 nm data required two exponentially decaying terms to obtain a good fit with the second term having a variable time offset, with Δt being the separation between the two terms. This behaviour was previously seen at a similar pump wavelength (255 nm) when measured using an MPI probe. The fits to equation (1) are presented as solid lines in figure 4, with the two components of the fit to the 254 nm data presented along with the total.

Zoom In Zoom Out Reset image size

The intensity profiles for the 279 nm, 269 nm and 243 nm data sets all show a rapid decay in intensity. Fitting the data from each pump wavelength to equation (1) then provides a lifetime equivalent to the time taken for the excited state wavepacket to leave the observation window associated with the FC region of the excited state potential energy surface. From an initial inspection of both figures 3 and 4 the 279 nm data appears longer lived. This appearance is however misleading, the fits to the data show that the apparent extended observation time is due to the much longer pulse duration of the 279 nm pump relative to that of the 269 nm and 243 nm pump as shown in table 1. The relatively long instrument response at 279 nm means we cannot reliably extract a time-constant associated with population leaving the FC region. We obtain our most satisfactory fit with a simple Gaussian envelope which places an upper limit on the lifetime of approximately 25 fs, suggesting very rapid motion out of the FC region. This picture is consistent with that observed at a pump wavelength of 269 nm. The shorter pulse duration at 269 nm means we can extract a lifetime of 23 ± 6 fs utilising the XUV probe which is in good agreement with that previously obtained with a MPI probe (shown as an inset) of 32 fs.

The equivalent 243 nm pump data is plotted in figure 4(d). Fitting to equation (1) provides a characteristic observation lifetime of 46 ± 14 fs which, while slightly longer than those obtained for the 279 nm and 269 nm pump wavelengths, suggests the excited state population rapidly dissociates at this higher pump wavelength as well. While each of these three traces show slight differences in their observation lifetime, they are all consistent with a rapid reduction in measured population that can be accurately described by a simple exponential decay of population out of the FC region. This suggests that the dynamics are broadly similar at each pump wavelength.

The 254 nm data shown in figure 3(c) has a strikingly different appearance. As with the other pump wavelengths, we plot the integrated intensity over the initial excited state features associated with ionization from the FC geometry covering the binding energy range 4.7–5.7 eV, in figure 4(c). This data shows a significant increase in the observation time of the initial excited state features, with significant intensity observed out to over 150 fs. In this case we cannot obtain a satisfactory fit to the data using a single exponentially decaying function. In the previous MPI work at 255 nm, a satisfactory fit was obtained using an empirical fitting function consisting of a sum of two exponentially decaying functions, where one of the two components is delayed relative to time zero. It should be stressed that this provides a purely empirical measure of the intensity profile and care should be taken in trying to extract mechanistic information from the numbers alone. The numbers simply serve to highlight the extended period over which significant signal is maintained and to highlight the difference between this energy region and the others. We use a similar empirical approach here and use the same function to fit the measured data as in the MPI experiments to aid comparison. The results from the MPI experiment are also plotted as an inset in figure 4(c), with the fits in both plotted as solid lines with the associated parameters given in table 1. Comparing the two data sets we see that both show a similar temporal structure but with a much higher intensity delayed component in the MPI measurements than observed in the XUV experiments. The fits to the two data sets provide very similar results for the delay between the two components and their associated lifetimes but with quite different relative amplitudes.

To further compare the MPI and XUV ionization photoelectron spectroscopy measurements around 255 nm, we plot the time-dependent data over the FC energy region that both probes measure in figures 5(a) and (b) respectively. In both data sets the time step used is the same but the MPI data has a significantly higher ultimate energy resolution as defined by the detector pixel spacing for the MPI data while the XUV data resolution is defined by the TOF binning size. The MPI resolution is 0.005 eV as opposed to 0.1 eV for the XUV measurements predominately due to the much lower electron kinetic energies measured, ∼1 eV and ∼17 eV in the MPI and XUV measurements respectively. Despite the difference in energy resolution there are a number of key similarities. In both cases, significant intensity is observed out to almost 200 fs, significantly later than any of the other pump wavelengths. The MPI data shows an initial energy shift to a slightly lower binding energy. The shift of approximately 0.04 eV is smaller than the energy resolution of the XUV data; however, as the pulse has a finite energy width we do see a shift in the centre of the photoelectron peaks in the XUV data to lower binding energies that matches the observations in the MPI data. In the MPI measurement we also observe a long lived feature at 5.5 eV which is highlighted by the white dashed boxes in figure 5. This feature was assigned to ${\tilde {X}}^{2}{E}_{3/2}\enspace {\nu }_{2}^{2}$ with ionization through a resonant Rydberg state. Although the XUV data does not have the energy resolution to separate this feature from the ${\tilde {X}}^{2}{E}_{1/2}$ peak we do not observe any unexpected enhancement of the peak, supporting our original assignment of the feature at 5.5 eV. The resonant enhancement of this ionization process is most likely why the delayed feature appears much stronger in the MPI data than in the XUV data. The similarity in appearance of the photoelectron spectrum and obtained lifetimes from both the MPI and XUV measurements suggest that the extended excited state lifetime observed here is not due to any accidental resonance as previously suggested [19], but is a consequence of the excited states dynamics of methyl iodide in the A-band. The accidental resonance appears to have enhanced the relative intensity of the long lived component in the MPI measurement but its origin is due to some extended excited state dynamics of the system.

Zoom In Zoom Out Reset image size

To analyse the longer term dynamics of the 279 nm, 254 nm and 243 nm data we follow a similar method to that outlined in Warne et al [30] and plot the integrated photoelectron intensity across a number of other energy regions as a function of pump probe delay, figure 6. The energy regions are selected to provide a series of discrete energy bands below the onset of the ground state spectrum (approx. 9.54 eV) where we can observe the changing intensity profile with time. From the bottom to top of figure 6 the binding energy of the corresponding spectral region plotted increases. The bottom trace in each plot, labelled with an $\tilde {X}$, is equivalent to that plotted in figure 4 and corresponds to ionization into the $\tilde {X}$ state of the methyl iodide cation from the FC region. The second row corresponds to the binding energy region between that covered in $\tilde {X}$ and the initial energy associated with ionization into the quartet state. The third trace, labelled with a Q, covers the energies associated with ionization into the quartet state from FC geometries, while the fourth plot in the 254 nm and 243 nm data covers the energy region between that associated with the quartet state and the onset of the ground state features. This is absent from the 279 nm data due to a lack of separation between the initial quartet and ground state peaks. The solid lines plotted for the $\tilde {X}$ state region in the 243 nm, 254 nm and 279 nm are the same as those described above for figure 4. For all of the other energy regions the solid lines are fits of the experimental data to equation (1). The scatter of the data at these higher binding energies means that the fits are used more as a qualitative guide to the eye rather than for the extraction of quantitative information from the absolute values obtained. This means that we cannot extract accurate time-dependent shifts for the new data, as we have previously demonstrated on the 269 nm pump data, [29] in order to extract approximate C–I bond lengths. The data at 279 nm and 243 nm however show the same time-dependent shift towards higher binding energy as a function of pump probe delay as seen at 269 nm. The intensity at binding energies between those associated with ionization into the $\tilde {X}$ and quartet states peak at a later time which is indicative of the increasing C–I bond length as the molecule dissociates. The intensity of the quartet state signal masks the continued shift to some extent but in the 243 nm data the continued shift can be seen at the highest binding energy region plotted. The data at 243 nm and 279 nm therefore provides a consistent picture of the rapid dissociation along the C–I with what appears to be a slightly longer lifetime observed at 243 nm.

Zoom In Zoom Out Reset image size

The data measured at a pump wavelength of 254 nm shows a much longer excited state observation lifetime across each energy region considered. This is most clearly shown in the comparison of the highest energy region covered in the 243 nm and 254 nm data. The 243 nm data at 8–8.6 eV binding energy shows a very sharp peak that reaches a maximum around 50 fs after time zero and decays back to the baseline level within a further 50–60 fs. The equivalent region in the 254 nm data shows a similar delayed increase reaching a maximum intensity at a similar delay to that seen at 243 nm, but in this case it does not return to the baseline level for approximately another 150 fs. Similar increases in the observation time at 254 nm, relative to the other wavelengths measured, are observed in the other binding energy regions plotted suggesting more complex dynamics in the excited state before dissociation. We stress again that the fits here are empirical and, while there are hints of signal related to peaks from multiple components in the time-dependent traces these cannot be asserted with any confidence within the current data set. Current signal levels are limited by a combination of the high-harmonic flux and the gas density at the interaction region.

Numerous measurements at wavelengths close to 266 nm have shown that the dominant dissociation product channel forms I* $\left({}^{2}P_{1/2}\right)$ in conjunction with the CH₃ co-fragment in its vibrational ground state. Previous product formation experiments measured the appearance time (as defined by the time taken for the C–I bond to reach a length of 13 Bohr) of this dominant fragment was recently measured to be around 114 fs. A shorter appearance time, of approximately 75 fs, was also observed for the next most abundant peak which corresponds to formation of the ground state $\text{I}\left({}^{2}P_{3/2}\right)$ in conjunction with the CH₃ co-fragment in its vibrational ground state. Interestingly, the measurements showed that the product channels associated with the vibrationally excited umbrella mode in CH₃ fragments have shorter appearance time than their vibrationless equivalent when produced in conjunction with the excited I* $\left({}^{2}P_{1/2}\right)$ co-fragment, and longer appearance times when associated with the ground $\text{I}\left({}^{2}P_{3/2}\right)$ co-fragment. The shorter appearance times (84 fs for the ν₂ = 1 channel) observed for the vibrationally excited fragments in the I* $\left({}^{2}P_{1/2}\right)$ channel is difficult to explain and is at odds with the modelling performed with both 4D and 9D potential surfaces which consistently show very similar or slightly longer lifetimes for the various vibrational states covered [19]. Our experiments show it takes ∼25 fs for the wavepacket to leave the initial FC region. This gives the wavepacket 50–90 fs to travel about 4.4 Å to be consistent with the product formation experiments. Although timescales measured are not directly comparable, they are, however, consistent, with both showing rapid dissociation. Due to the fact we cannot quantum state-resolve the individual product channels or the vibrational distributions within them, our measurements will be a weighted average of the individual contributions to the lifetime convolved with the laser cross-correlation duration.

Other experiments have shown that at the shorter wavelength of 243 nm the I* $\left({}^{2}P_{1/2}\right)$ in conjunction with the CH₃ co-fragment in its vibrational ground state continues to be the dominant product channel. The appearance time was recently measured to be shorter than that observed at 266 nm standing at approximately 83 fs and there is a commensurate reduction in the appearance time (53 fs) of the equivalent $\text{I}\left({}^{2}P_{3/2}\right)$ channel [19]. This is perhaps unsurprising given the extra kinetic energy available upon dissociation at this higher energy but is contrary to the extended excited state lifetime observed in our photoelectron spectroscopy experiments. The product state measurements also show that vibrationally excited fragments are produced with higher yield at 243 nm and the appearance times increase with the level of vibrational excitation observed. This is more in line with the supporting simulations, however, the magnitude of the observed increase is much larger than those calculated. The vibrationally excited fragments typically appear within 10 fs of the vibrationless product in the calculation while experimental measurements suggest the delay can be as much as 50–60 fs for some fragments [19]. We suggest that the higher prevalence of the vibrationally excited products and their increased lifetimes would result in an effective increase in the excited state lifetime observed in our photoelectron spectroscopy measurements explaining why the measured lifetime was ∼20 fs longer than lifetime measured for 269/279 nm. The increased lifetime therefore reflects the different vibrational dynamics associated with the various dissociation channels. This discussion highlights clear issues around comparing time frames obtained from various measurement techniques. The selectivity provided by REMPI ionization and product fragment recoil detection, allows measurement of product fragment appearance times with full vibrational and spin–orbit state-resolution but is limited in only measuring the dynamics of the system at the end of the dynamic process. In contrast the photoelectron spectroscopy probes provide a more general probe of the system over all geometries but this is necessarily averaged over all quantum states. A combination of all techniques will clearly be required to disentangle the dynamics of systems where multiple product states are populated.

As far as we are aware, there are no time-resolved product state measurements at wavelengths close to 254 nm which makes comparison difficult. In the work of Murillo-Sánchez et al [19], it was suggested that the extended lifetime observed in the MPI photoelectron spectroscopy measurements could be due to accidental resonances in the ionization step. Our one-photon measurements presented here rule this suggestion out. The accidental resonances appear to have enhanced the amplitude of the longer lived features when compared with the one-photon measurements but the extended lifetime measured, and the unusual profile of the excited state decay transient, persists when this effect is removed. The authors also suggested that the enhanced lifetime may be due to transfer of kinetic energy into product vibrational modes, as seen at 243 nm, as opposed to an enhanced contribution of the ¹ Q₁ state as originally suggested in reference [29]. Both of these options are possible. We observe the extension of the lifetime due to vibrational dynamics in the 243 nm measurements but the appearance of the time-delay transient at 243 nm is strikingly different to that measured at 254 nm. If a new region of the ³ Q₀ potential energy surface became accessible at 254 nm where FC overlap with the methyl cation $\tilde {X}$ state were reduced, this may explain the time-dependent modulation of the ionization cross section observed in this measurement. However, we would also expect to continue to see the effect of this on the shape of the delay transients at shorter wavelengths, yet there is no evidence of this in the 243 nm measurement. We also expected some vibrational excitation of the room temperature ensemble of methyl iodide, with about 7% in the ν₃ vibration (C–I stretch) and 1% in the ν₆ vibration (CH₃ rock). This may have an effect on the observed photoelectron spectrum but why this would only be observed at 254 nm is not clear. Therefore, while we cannot rule out the possibility that the extended lifetime is due to vibrational dynamics, the stark difference in the appearance of the delay transient at 254 nm makes us lean towards an alternative theory, such as enhanced population of the ¹ Q₁ state at this wavelength, followed by geometric relaxation on the ¹ Q₁ surface, for the changes in the photoelectron spectrum. What is clear is that more measurements are required and we hope that our measurements encourage directly comparable measurements around 254 nm to allow a direct comparison of the quantum state distributions obtained with the excited state dynamics measured.