Communication: XUV transient absorption spectroscopy of iodomethane and iodobenzene photodissociation Free

Studies of chemical dynamics can benefit from the vibrant field of attosecond science¹ in many intriguing ways. The short duration of attosecond pulses implies broad and continuous frequency spectra with photon energies extending into the extreme ultraviolet (XUV), well suited for inner-shell absorption measurements in molecules.² Core-to-valence transitions are element-specific probes of the intramolecular local chemical environment, since an electron from a highly localized and energetically distinct core orbital is lifted to a hole in the valence shell, affected by chemical bonding. Core-to-valence transitions appear just below the ionization edges in the molecular absorption spectrum and are targeted by NEXAFS (Near Edge X-Ray Absorption Fine Structure) spectroscopy.³ With the development of ultrashort XUV and X-ray pulses from high harmonic generation (HHG), slicing of synchrotron radiation, and free-electron lasers, this powerful tool of chemical structure analysis is currently being translated to the ultrafast time domain.^4–9 Table-top transient absorption studies with attosecond pulses and attosecond pulse trains (APTs) from HHG benefit from their appealing temporal and spectral properties.^10–14 

Here we probe the pre-N_4,5 edge region of iodine around 50 eV photon energy by transient XUV absorption of an APT. These transitions originate in the outermost 4d core shell in the iodine atom. We study photodissociation of iodomethane (CH₃I) and iodobenzene (C₆H₅I) in the A-band, pumped by a UV femtosecond laser pulse centered around 265 nm wavelength. Comparing transient XUV absorption of the well-explored photodissociation of these two molecules^15,16 promises to be particularly interesting, due to the intramolecular location of the valence shell hole created by the UV pump pulse. In CH₃I the dominant transition originates from promotion of an electron from a non-bonding (n) iodine orbital to an anti-bonding (σ^*) orbital localized on the C–I bond. In C₆H₅I, the dominant π → σ^* excitation involves the phenyl π orbital, which is distant from the iodine reporter atom.

Photodissociation of CH₃I was recently examined using a very similar experimental scheme. Through subtraction of the dominant transitions, Attar et al.¹⁷ have found transient core-to-valence excitations at distinct energies. In contrast, we interpret our CH₃I results as the immediate emergence of a persisting new target state for XUV absorption upon molecular n → σ^* UV excitation. During dissociation, the molecular orbitals separate into atomic orbitals of iodine and molecular orbitals of the methyl fragment. This causes the UV-created hole in the valence shell to adjust, leading to a time-dependent shift in the XUV transition energy.

A-band excitation in CH₃I is split by the strong spin-orbit interaction into three overlapping states ³Q₁ (E), ³Q_0⁺ (A₁), and ¹Q₁ (E), dipole-allowed from the A₁ ground state.¹⁸ While the $X ̃ A 1 → 3 Q 0 +$ transition carries the majority of the oscillator strength at 265 nm,^15,19 a non-adiabatic crossing with the higher-lying ¹Q₁ state is encountered upon elongation of the C–I bond (see Fig. 1(a)), leading to population redistribution.²⁰ Asymptotically, the ¹Q₁ state connects to ground-state I (²P_3/2), while the ³Q_0⁺ state leads to spin-orbit excited I^* (²P_1/2).

An apparatus usually used for photoion and -electron spectroscopy with short APTs and intense IR fields²¹ was adapted for the present work. XUV pulses were generated via HHG in argon by a 6 fs duration laser pulse, obtained by spectral broadening in a neon-filled hollow-core fiber and chirped-mirror compression. UV pulses were produced by third harmonic generation of the 35 fs duration pulses from the 1 kHz repetition rate Ti:Sa laser system. The recombined beams were focussed into the absorption cell by a toroidal mirror. The transmitted XUV pulses were analyzed with a flat-field spectrometer coupled to a multichannel plate/phosphor screen stack and a digital camera. The spectrometer was calibrated using an aluminium filter’s L_2,3-edge and known absorption lines of argon. To minimize the influence of XUV spectral fluctuations, static and pump-probe absorption spectra were recorded in quick alternation using a chopper wheel. The linear absorption difference signal was transformed to a change of absorbance (ΔA) by taking its (base-10) logarithm. A small contribution of scattered light was subtracted. The observed linewidths in CH₃I are consistent with a spectral resolution of 0.3-0.4 eV. To determine the UV-XUV cross correlation and zero time-delay, the control of autoionization line shapes of argon by semi-strong UV pulses was recorded accompanying each measurement.^11,22 In Figs. 3 and 5 the renormalized absolute peak area of the UV on/off difference of the 3s²3p⁶  →  3s3p⁶np transition at 28 eV is plotted in grey. Taking into account the 23.3 fs lifetime,¹¹ cross correlations of (103 ± 6) fs and (110 ± 6) fs were extracted for the CH₃I and C₆H₅I measurements, dominated by the duration of the chirped 265 nm pulses.

Fig. 2(a) shows the XUV absorption spectrum of ground-state CH₃I, obtained as the difference of a spectrum recorded with and without CH₃I in the target cell. It is very similar to the reference spectrum from the literature,²³ obtained with (e,e) spectroscopy. The three fully resolved peaks are fitted with Voigt profiles and a static background. Their mean energies agree well with the literature values²³ (see Table I). The dominant peaks are the iodine spin-orbit doublet at 50.6 and 52.4 eV, assigned to the molecular 4d_5/2 → σ^* and 4d_3/2 → σ^* transitions (see Fig. 1(c)). Additional partially resolved features belong to transitions to higher valence orbitals (markers in Fig. 2 and Table I). In Fig. 2(b), we show the spectral change in absorbance of the dissociated CH₃I with respect to the unpumped molecule (the photodissociation action spectrum). The dominant peaks from the ground-state spectrum (A, B) are depleted (4, 5), while a strong new doublet appears at 45.9 and 46.7 eV (1, 2). These lines correspond to 4d_5/2 (²P_3/2)  →  5p_3/2 (²D_5/2) and 4d_3/2 (²P_1/2)  →  5p_1/2 (²D_3/2), transitions in atomic I and I^* (Fig. 1(d), Table I). A very weak peak at 47.6 eV is assigned to the I: 4d_3/2 (²P_3/2)  →  5p_3/2 (²D_3/2) transition, which is known to be nine times weaker than the I transition at 45.9 eV.^24,25 The peak areas scale linearly with pump intensity, confirming a 1-photon transition.

The excited atom fraction of the iodine photodissociation products $Φ I * = I * / I + I *$ can be extracted from the fitted peak areas in the action spectrum, after scaling with the respective oscillator strengths (see Table I). Using the atomic spin-orbit doublet assigned to I and I^* (peaks 1,2), $Φ I * ,=0.73 ± 0 . 03 0 . 01$ is obtained. The top value (±0.01) is the uncertainty due to the fit errors, while the bottom value (±0.03) arises from the uncertainty of the relative oscillator strength from the literature.²⁵ Alternatively, the depleted molecular transitions measure the relative CH₃I photodissociation yield, giving $Φ I * ,=1− Φ I ,=0.76 ± 0 . 05 0 . 02$ (using peaks 1 and 4) and $0.74 ± 0 . 05 0 . 03$ (peaks 1 and 5). All values are consistent within the error bars, with a mean of Φ_I^* = 0.74 ± 0.06, centered in the range 0.70 - 0.81 reported in the literature for 265 nm photodissociation.²⁰ 

In Fig. 3 integrated peak areas from the action absorption spectra are shown as a function of pump-probe delay. The atomic I and I^* lines (1 and 2 in Fig. 2(b)) rise and the iodine transitions in CH₃I (sum of lines 4 and 5) are depleted immediately upon UV photon absorption within the experimental time-resolution. This important result is explained by inspecting the valence shell molecular orbitals (Fig. 1(b)). The A-band excitation drives a transition from the Highest Occupied Molecular Orbital to the Lowest Unoccupied Molecular Orbital (HOMO to LUMO). The resulting (2e)⁻¹ hole immediately provides a new target for XUV absorption from the iodine 4d core. Since the (2e)⁻¹ hole is centered on the iodine atom (see inset of Fig. 1(b)), the oscillator strength is increased as compared to the 4a₁ target (see inset) and the positive ΔA associated with the re-filling of this hole exceeds the negative ΔA associated with the 4d → σ^* transition (Fig. 2(b)). Since UV excitation populates 4a₁, the depletion of the iodine transitions in CH₃I is also immediate. The valence shell of the atomic iodine product consists of a fully occupied 5s² shell and a 5p⁵ shell with a single vacancy. The atomic 5s² shell originates predominantly from the 2a₁ orbital of CH₃I.¹⁸ The 5p⁵ shell is formed from the depleted 2e(n) shell (3 electrons and 1 hole) and the 3a₁ (σ) shell (2 electrons) (see Fig. 1). Importantly, since the UV-depleted 2e shell of the molecule corresponds to iodine lone pairs, which are not involved in chemical bonding, the energy shift between the 2e shell and the atomic 5p shell is expected to be small. The hole created by the pump pulse is hence expected to persist, with minor adjustment during dissociation.

An independent particle model is convenient for rationalizing the direction of the energetic shifts of the emerging XUV transitions upon dissociation.¹⁷ At CH₃I equilibrium bond length the 4d → n(I) transition should appear at energies corresponding to the difference between the 4d(I) → σ^* and n(I) → σ^* transitions. Upon dissociation, the XUV transitions converge to the atomic transitions (see Fig. 1). The 4d core is split into a 4d_3/2 spin-orbit manifold and a 4d_5/2 spin-orbit manifold. The I(5p) atomic state is split into 5p_3/2 and 5p_1/2, and the 2e(I) molecular orbitals are split into n_3/2 and n_1/2. The ³Q_0⁺ A-band branch represents a n_1/2 → σ^* transition, with a transition energy of 4.7 eV²⁰ and leads asymptotically to the I^* product, i.e., (n_1/2)⁻¹ in the molecule correlates to (5p_1/2)⁻¹ in the atom. Hence, at equilibrium bond length an XUV transition should appear at (52.3 -4.7) eV = 47.6 eV and converge to 46.7 eV in the asymptotic region (see Figs. 1(c) and 1(d)). In an atom, the 4d_5/2 → 5p_1/2 transition is forbidden. The analysis given for the triplet branch ³Q_0⁺ can be repeated for the singlet branch ¹Q₁, which leads to the I ground state product. Since the C–I coordinate-dependent potential energy surfaces of the core-excited states are not known, we take the energetics at the equilibrium bond length for this estimate. ¹Q₁ corresponds to n_3/2  → σ^*, with a transition energy of 5.2 eV and correlates to (5p_3/2)⁻¹ in the atom. Hence, at the equilibrium bond length XUV transitions should appear at (50.6-5.2) eV = 45.4 eV and (52.3-5.2) eV = 47.1 eV and converge to 45.9 eV and 47.6 eV, respectively, in the asymptotic region. Note that the 45.9 eV transition is nine times more intensive (Table I). In summary, the independent particle model predicts that the dominant asymptotic XUV lines I: 4d (²P_3/2) → 5p (²D_5/2) and I^*: 4d (²P_1/2) → 5p (²D_3/2) at 45.9 eV and 46.7 eV reach their final position via a sub-eV shift to higher (45.4 to 45.9 eV) and lower energies (47.6 to 46.7 eV), respectively.

For each time step in the pump-probe delay scan, we fit Voigt profiles to the dominant emerging XUV transitions and plot the center positions as a function of time-delay (Fig. 4). In line with the expectation from the independent particle model, we find a clear, small shift to lower energies for the dominant I^* transition and to somewhat higher energies for the dominant I transition. The shifts are about −0.15 eV (I^*) and +0.10 eV (I), i.e. less than expected, but showing the predicted relative trend. Smaller measured than expected shifts may be partly due to the long UV pulse duration on the time scale of the dynamics. In fact, a fit of a convolution of an exponential growth/decay function and the fixed Gaussian IRF shows no significant time-constant of the shifts of the XUV lines within our experimental resolution, as is the case for the emergence of these transitions (see above).

The very rapid emergence and shift of the new XUV transitions is in marked contrast to the ≈90 fs rise time of the atomic mass signal in femtosecond UV pump - multi-photon ionization probe experiments.²⁶ While the latter probe the asymptotic atomic products, the shifts of the emerging XUV transitions represent the relaxation of the valence shell vacancy, which we find to take place during the very first moments of dissociation.

We now turn to iodobenzene, which features a delocalized π-system, responsible for a large fraction of the UV oscillator strength. Low-lying excitations of C₆H₅I include dissociative π  →  σ^* and bound π  →  π^* transitions, as well as the n →  σ^* transitions also found in iodomethane. The mixed A-band is located in the 240-320 nm range²⁹ and leads to production of I^* (τ  ≈  350 fs), I (τ  ≈  350 fs, dominant channel) and I (τ  ≈  700 fs, weak channel) fragments.^30–33 The main doorway state at 265 nm is 3B₁ (π  →  σ^* transition).³⁴ The 3B₁ potential energy surface is repulsive along the C–I coordinate, asymptotically forming ground state I. It crosses with the repulsive 5B₁ and 5A₂ states, which also converge to I, and with the repulsive 4A₁ state, which produces I^*. The slow channel is believed to originate from the initial excitation of the predissociative 4B₂ (π  →  π^*) state, which undergoes an intersystem crossing.³⁴ At 265 nm, the 1A₁ → 4B₂ oscillator strength is an order of magnitude below that of 1A₁ → 3B₁, consistent with the direct dissociation accounting for 84% of the total yield.³³ 

In Fig. 5(a) we show the XUV action spectrum of iodobenzene photodissociation at 265 nm (cf. Fig. 2(b)). A depleted molecular transition at 50.6 eV (4d_5/2)⁻¹ is observed. We hence expect the (4d_3/2)⁻¹ counterpart at 52.3 eV, where indeed evidence of a peak is visible. The atomic I: 4d_5/2 → 5p_3/2 transition at 45.9 eV is much stronger than the I^*: 4d_3/2 → 5p_1/2 transition at 46.7 eV. Furthermore the I: 4d_3/2 → 5p_3/2 transition at 47.6 eV is now clearly visible and has the expected relative intensity. Fitting peak areas and scaling with the relative oscillator strength (Table I), we obtain $Φ I * =0.33 ± 0 . 03 0 . 03$⁠, in good agreement with the previously reported experimental value of Φ_I^* = 0.32 ± 0.05 from Sage et al.¹⁶ 

Fig. 5(b) shows the evolution of the integrated peak areas as a function of pump-probe delay. The iodine transitions in the bound molecule are depleted on a time scale indistinguishable from the cross correlation (panel PhI), as was the case in CH₃I. Interestingly, the atomic features appear with noticeable delay, namely, with rise times of (40 ± 9) fs and (53 ± 31) fs for I and I^*, respectively. The position of the emerging I peak at 45.9 eV is independent of time-delay in contrast to the CH₃I case; for the weaker transition in I^* the signal-to-noise level and proximity of the intense I peak precludes an analysis of the shift.

While a detailed explanation has to be deferred to a later publication and involve theory, we can speculate on the marked difference observed between the two molecules. In contrast to CH₃I, the dominant π  →  σ^* UV excitation in iodobenzene creates a vacancy, which is not localized on the reporter atom but rather the distant phenyl ring. Therefore, it is initially not accessible for a vertical XUV transition from the iodine core. An evolution of the valence shell is necessary for the hole to migrate and the new XUV transitions, which converge to the asymptotic atomic lines, to emerge. On the other hand, the unoccupied orbital in the ground-state C₆H₅I is populated immediately with UV absorption, indicating dominant σ^* character of the UV-excited state. The absence of a time-dependent shift for the transition converging to I: 4d_5/2 → 5p_3/2 may also be understood within the independent particle model as follows: The transition energy for the dominant UV excitation 1A₁ → 3B₁ is 4.5 eV.³⁴ Therefore at the equilibrium bond length the XUV transition should appear at (50.6-4.5) eV = 46.1 eV and converge to 45.9 eV in the asymptotic region. Given the time-delay of the transition and smaller than expected shifts for CH₃I, the shift is probably too small to be observed in our current experiment. Consistent with the case of CH₃I the emergence of the XUV transitions is still much faster than the rise time of the atomic mass signal in femtosecond photoionization experiments (350 fs for the dominant direct channel³²).

In their recent comparable work on CH₃I, Attar et al. find weak shoulders upon the emerging atomic iodine transitions, which subsequently disappear during the dissociation.¹⁷ After subtracting the scaled atomic transitions, the shoulders are interpreted as new, short-lived intermediate core-to-valence electronic states at distinct XUV energies which are assigned to the transition-state region. These transient resonances are found to reach a maximum at ≈40 fs and decay within ≈90 fs. Our experimental raw data for CH₃I appear to be consistent with the data of Attar et al. While we cannot rule out such distinct transient intermediate resonances in CH₃I, we believe that our interpretation of the data as a weak energy shift of the valence shell hole is a straightforward alternative that should be considered. The much shorter time scales as compared to complete dissociation²⁶ suggest that in CH₃I very little nuclear motion is involved in the relaxation of the valence shell vacancy, attributed to its nature originating from the iodine lone pair. In view of the marked difference observed for iodobenzene, XUV transient absorption is anticipated to be sensitive to the coupled electronic and structural dynamics in molecules. In contrast to time-resolved photoelectron spectroscopy,³⁵ the dynamics is projected on sparse core-excited states rather than dense cation states, which offers a different coordinate dependence. The recent development of theoretical methods dedicated to femtosecond NEXAFS of molecular dynamics³⁶ is hence very timely; it will aid to reach a detailed understanding of the results obtained in the current groundbreaking experiments.

We thank Oleg Kornilov (MBI) for help with the laser system, fruitful discussions, and comments on the manuscript. We also thank Gerry O’Sullivan (UCD, Ireland) for helpful discussion on XUV absorption cross sections of iodine, Rune Lausten (NRC, Canada) for technological advise and Roman Peslin (MBI) for excellent technical assistance. G.R. thanks the Netherlands Organization for Scientific Research (NWO) for financial support (Rubicon 68-50-1410).