Photochemical pathways in nucleobases measured with an X-ray FEL

Abstract

The conversion of light energy into other molecular energetic degrees of freedom is often dominated by ultrafast, non-adiabatic processes. Femtosecond spectroscopy with optical pulses has helped in shaping our understanding of crucial processes in molecular energy-conversion. The advent of new, ultrashort and bright X-ray free electron laser sources opens the possibility to use X-ray-typical element and site sensitivity for ultrafast molecular research. We present two types of spectroscopy, ultrafast Auger and ultrafast X-ray absorption spectroscopy, and discuss their sensitivity to molecular processes. While Auger spectroscopy is able to monitor bond distance changes in the vicinity of an X-ray created core hole, near-edge absorption spectroscopy can deliver high-fidelity information on non-adiabatic transitions involving lone-pair orbitals. We demonstrate these features on the example of the UV-excited nucleobase thymine, investigated at the oxygen K-edge. We find a C–O bond elongation in the Auger data in addition to ππ*/nπ* non-adiabatic transition in X-ray near-edge absorption. We compare the results from both methods and draw a conclusive scenario of non-adiabatic molecular relaxation after UV excitation.

This article is part of the theme issue ‘Measurement of ultrafast electronic and structural dynamics with X-rays’.

1. Introduction

Ultrafast photophysics and photochemistry of organic molecules play a crucial role for processes like light harvesting in plants [1], vision [2], nucleobase photoprotection [3,4] and atmospheric chemistry [5–7]. Photoexcitation triggers a complex interplay of electronic and nuclear degrees of freedom, which cannot be described within the Born–Oppenheimer approximation, which either leads to chemical storage of the absorbed light energy through nuclear rearrangements or the efficient dissipation of the absorbed energy to prevent unwanted structural changes. The ability to prevent unwanted processes originates from outcompeting them on the ultrafast timescale. Ultrafast spectroscopy aims at deciphering the ultrafast pathways of molecules in the complex phase space of electrons and nuclei. Many ultrafast spectroscopy methods have been developed over the last 30 years covering the spectral range from mid IR to VUV and light-matter interactions like photoabsorption, Raman transitions and photoionization. In many instances, attribution of signatures in observables to processes in the molecule is challenging because of the high level of complexity of the molecule and the comparatively low dimensionality of the observables. Novel X-ray spectroscopic methods provide a new and interesting route to perform ultrafast spectroscopy yielding a set of observables, which is complementary to established methods. While visible and UV light interacts with valence electrons of a molecule, X-rays interact with its core electrons. The tightly and locally bound core electrons make X-ray spectroscopy element- as well as site-specific [8–10]. Thus, X-ray spectroscopic observables obtain a qualitatively new aspect: the ability to probe photochemistry locally at different sites of a molecule. In this paper, we will show, using the example of the nucleobase thymine, how this novel capability helps in achieving a deeper understanding of photophysical pathways.

Thymine has been extensively studied by both experimental and theoretical approaches [11–14]. It has been known for a long time to quickly and efficiently transfer photoenergy absorbed in the UV range into vibrational energy via non-adiabatic dynamics through conical intersections. While there is general agreement over a non-adiabatic relaxation mechanism, the exact pathway is still a subject of debate. In this contribution, we show how probing the ultrafast relaxation mechanism of thymine by X-ray induced Auger electron spectroscopy (AES) and near-edge X-ray absorption fine structure (NEXAFS) spectroscopy helps to distinguish the details of the electron–nuclear relaxation and coupling processes. In the following, we will review the literature on thymine dynamics, our experimental set-ups for X-ray experiments and the experimental results from time-resolved (TR) ultrafast AES and NEXAFS experiments. We will discuss how these results compare with each other and what major conclusions can be drawn taking results from both experiments into account.

2. Ultrafast dynamics of nucleobases

Both the purine (adenine, guanine) and pyrimidine (thymine, uracil, cytosine) nucleobases exhibit unsaturated, planar geometries of one or two condensed rings. Their electronic structure is, therefore, typical to unsaturated hydrocarbons. Thus, they exhibit high-lying occupied and low-lying unoccupied π molecular orbitals (MOs) with strong delocalization over the ring(s). The presence of those orbitals leads to exceptionally strong ππ* transitions which dominate the absorption spectrum in the ultraviolet regime. The band associated with a ππ* transition in thymine, for instance, exhibits a cross-section on the order of 30 Mbarn [15], which are comparable to fluorescence dyes [16].

The presence of heteroatoms (nitrogen and oxygen) introduces high-lying occupied lone-pair MOs, which exhibit a high degree of localization at the heteroatom site. Especially in the case of pyrimidine nucleobases, they give rise to a second type of low-lying excited state, nπ* states. The latter are ‘dark’ i.e. not accessible from the ground state via one photon absorption. A third type of excited state, πσ* states, are induced by the presence of unoccupied antibonding σ* orbitals.

All nucleobases exhibit very low fluorescence quantum yields [17], which is a direct proof of excited state lifetimes in the ps or sub-ps regime. It is well known that these ultrashort lifetimes are due to readily accessible conical intersections between the photoexcited ππ* state and other electronic states like the nπ* state and the ground state permitting efficient depopulation of the ππ* state. The conical intersections allow for ultrafast non-adiabatic population dynamics which cannot be described in the framework of the Born–Oppenheimer approximation. The mechanisms of excited state population relaxation have been extensively studied by various experimental and theoretical methods over the past three decades [11–14]. The knowledge about the relaxation mechanisms in the pyrimidine nucleobases can be summarized in the cartoon scheme in figure 1. The pyrimidine bases can be excited to the ππ* excited state by UV light. Calculations predict the nπ* state to be below the ππ* state in the Franck–Condon region. Conical intersections between the ππ* state and the nπ* state have been predicted by several investigations of the excited state potential energy surface [13]. Furthermore, conical intersections between the ππ* state and the ground state have been observed.

Figure 1.

Scheme visualizing possible relaxation paths for UV-excited thymine. Thymine can be photoexcited to a ππ* state. There exists a second, lower and spectroscopically dark nπ* state. There exist relaxation channels from the ππ* state to the nπ* state and to the ground state through conical intersections. The lifetime of the ππ* state is subject to a long-standing debate in the literature. Adapted from [18], Copyright © 2007 American Chemical Society.

It is, however, well known that the mere investigation of excited potential energy surfaces is insufficient for the prediction of relaxation mechanisms, because it cannot be easily predicted if e.g. a conical intersection is visited by an excited state nuclear wavepacket, which is launched in the Franck–Condon region. A number of explicit simulations of the excited state wavepacket dynamics have been published so far [18–30]. Their results are well summarized in ref. [13]. Most of them agree on a rapid relaxation out of the Franck–Condon region and trapping of the ππ* state population for at least several hundreds of fs in a local ππ* minimum. The conical intersection to the nπ* state is accessed over a small barrier, indicated in figure 1.

For the experimental results, we concentrate on gas phase molecules, as we performed the X-ray experiments discussed below for isolated thymine. Weak and strong field photoion studies [31–33], as well as photoelectron experiments [34–37] generally exhibit three timescales in the signals: a short timescale of around 100 fs, an intermediate few picoseconds, and a 100 s of picosecond signal. In photoions, the timescales generally manifest themselves in a decaying signal, while for photoelectrons, a shift in kinetic energy accompanies that time constant. Earlier interpretations of these results assume the electronic relaxation from ππ* to lower PES to be responsible for the 100 fs constant [34–37]. According to the explicit excited state wavepacket simulations, however, the shortest timescale is interpreted as a nuclear relaxation from the FC region to a minimum in the ππ* state, while the intermediate ps time constant is seen as a signature of population lingering in the minimum, having a picosecond timescale to cross it and to immediately decay electronically into lower-lying states [18].

3. Ultrafast X-ray methodology at FELs

Soft X-ray spectroscopy is known to yield element- and site-specific information about the electronic and nuclear structure of molecules [8–10]. The 1 s ionization potentials (K-edges) of the—apart from hydrogen—most abundant elements in organic chemistry, carbon, nitrogen and oxygen, lie in the soft X-ray regime separated by more than 100 eV. Thus, element specificity can be achieved by tuning the soft X-ray wavelength. The site-specificity originates from the strongly localized character of the 1 s orbitals around their respective atomic cores. The most important soft X-ray spectroscopic techniques for investigations in the gas phase are near-edge X-ray absorption fine structure (NEXAFS) spectroscopy (or X-ray absorption near-edge spectroscopy, XANES), X-ray photoelectron spectroscopy (XPS) and Auger electron spectroscopy (AES). We will in the following focus on NEXAFS spectroscopy and AES, because they have been successfully employed for the nucleobase thymine.

The NEXAFS spectrum of thymine at the oxygen edge is reproduced from [38] in figure 2b. The pre-edge region shows a strong transition due to resonances between the core O1 s and two different π orbitals (only one of them shown in figure 2a). Owing to the strong localization of the core orbital at the O atom, the dipole matrix element is localized and thus its amplitude increases with the core-valence orbital overlap. If the molecule is pre-excited in the UV, a valence electron is promoted from an occupied to an unoccupied molecular orbital. This leaves an electron hole in a formerly occupied orbital and, thus, a new pre-edge resonance in time-resolved (TR) NEXAFS. Based on the above explanation, the cross-section of the pre-edge resonance is a measure for the degree of localization of the final state in absorption. The lone-pair n-orbital exhibits a high localization at one of the O atoms, it is essentially an O 2p orbital. Thus, if the nπ* state of the molecule gets populated during thymine's relaxation, a TR-NEXAFS study should show a strong UV-induced pre-edge feature.

Figure 2.

(a) Sketch of π,n, π* valence and one O 1 s core orbital of thymine, shown as an isoamplitude plot (right) and absolute square in two dimensions. The core orbital is deeply bound at 540 eV and tightly confined to one of the oxygen atoms to a few picometers. (b) Absorption spectrum of thymine, showing the O1 s to π* resonance. The splitting results from two slightly different π* orbitals (not shown in (a)). Data digitized from ref. [38].

In TR-AES, we are more sensitive to the nuclear molecular relaxation. To induce a non-resonant Auger process, a core electron is excited to the continuum. In the soft X-ray range, the core hole decays predominantly via Auger emission, i.e. through the simultaneous transition of a valence electron to the core hole and emission of a second valence electron, the Auger electron. The final state of the molecule is a valence-dicationic state, which usually undergoes fragmentation. In all but the simplest molecules, many valence electrons can participate in Auger decay, leading to a large number of decay channels. However, as in the case of NEXAFS spectroscopy, their cross-sections are modulated by the local wave function overlap between the core hole and the valence orbitals resulting in the site-specificity of AES and initially leading to a strong localization of the dicationic charge. Owing to energy conservation, the kinetic energy of the Auger electron is equal to the energy gap between core-ionized intermediate and dicationic final state minus the ionization potential of the Auger electron. In static AES, this provides detailed information about the local valence electronic structure [39].

In TR-AES [40] like in the case of our study of the nucleobase thymine [4], an additional effect on the Auger electron spectra becomes important, the influence of time-dependent structural changes as the nuclei of the molecule relax to an excited state potential. During the Auger decay, two positive charges are locally created at the site probed by AES. Owing to the strong Coulomb repulsion of the two charges, the potential of the dicationic final state is strongly repulsive with respect to chemical bonds of the probed site with neighbouring atoms. The intermediate core hole state, on the other hand, does not exhibit a strongly repulsive character. Thus, a change in the bond length between the probed site and a neighbouring atom during excited state relaxation strongly modulates the gap between core-ionized and dicationic state, which directly transfers into a modulation of the kinetic energy of the Auger electron on the level of several eV. Therefore, TR-AES can act as a highly sensitive probe for local structural dynamics like bond length changes (figure 3).

Figure 3.

Adapted from ref. [3]: schematic representation of the experimental set-up used for the time-resolved near-edge X-ray absorption fine structure spectroscopy experiments. The spectral range is scanned by simultaneously tuning the free electron laser electron pulse energy and adjusting the X-ray monochromator. Timing between optical and X-ray pulses is measured with a cross-correlator timing tool on a shot-by-shot basis. Optical and X-ray pulses are quasi-collinearly overlapped employing an optical mirror with a hole for X-ray transmission. The sample is introduced into the interaction region of the experiment using an effusive oven. Auger electrons are detected by a magnetic bottle time-of-flight spectrometer. The integrated Auger electron yield is a measure for the absorbance of the sample. In TR-AES, the X-ray photon energy is set fixed above the molecular ionization potential and a monochromator is not needed.

(a). Experimental apparatus

Both AES and NEXAFS spectroscopy require similar experimental conditions as provided at the AMO [41,42] and SXR [43] experimental hutches at Linac Coherent Light Source (LCLS) [44]. They need a vacuum chamber equipped with an effusive sample oven [45] and an electron spectrometer. Furthermore, the soft X-ray pulses are spatially and temporally overlapped with UV pulses from an ultrafast optical laser system synchronized to the X-ray free electron laser (XFEL) source. Owing to a considerable jitter in the relative arrival time of optical and soft X-ray pulses, their temporal delay must be measured on a shot-by-shot basis to allow for the improvement of the temporal resolution below 100 fs by resorting in the post processing of the data. This can now be routinely done with the LCLS timing tools [46–48]. In the case of AES, the spectral requirements for the soft X-ray laser pulses are relaxed. The typical bandwidth of several eV and strong shot-by-shot fluctuations of XFEL pulses are transferred into the kinetic energy of the X-ray photoelectrons, whereas the Auger electron kinetic energies are independent of the photon energy of the XFEL pulse. This is different for NEXAFS spectroscopy, where the spectral resolution is determined by the stability of centre and bandwidth of the XFEL pulses. Thus, the XFEL pulses must be monochromatized by either self-seeding schemes [49,50] or the use of a monochromator as available in the SXR hutch [51]. While AES can be performed at a constant but jittering soft X-ray photon energy, the energy must be scanned to obtain NEXAFS spectra. This can be achieved by simultaneously tuning the electron bunch energy of the XFEL and adjusting the monochromator. Both AES and NEXAFS spectra are recorded using an electron spectrometer, in the case of our experiments, a magnetic bottle spectrometer [52]. In the AES experiments, the spectrum can be directly obtained by resolving the kinetic energy of the detected electrons. For NEXAFS spectroscopy, the Auger electron yield is an excellent measure for the number of absorbed photons. Thus, the absorption spectrum can be constructed from an integrated electron yield at different soft X-ray photon energies.

4. Experimental results of X-ray induced auger and X-ray absorption probing

(a). Time-resolved auger probing

Figure 4 shows time-dependent difference Auger electron spectra (photoexcited Auger electron spectra minus static Auger electron spectra) of the nucleobase thymine at the oxygen edge. The spectra are split in regions I–III. Regions I and III show positive features, i.e. an intensity gain induced by the UV photoexcitation, whereas region II shows negative features, i.e. an intensity loss induced by the UV photoexcitation. The earliest time-dependent features appear in regions I and II, while the onset of the feature in region III is clearly delayed with respect to the other features. The positive feature in region I decays on a sub-ps timescale.

Figure 4.

Adapted from ref. [4], (a): False-colour plot of time-dependent difference Auger electron spectra of the nucleobase thymine after photoexcitation at 267 nm. The spectra are divided into regions I–III. The earliest UV-induced changes are a positive signature in region I and a negative signature in region II. The signature in region I decays with a 200 fs time constant. Simultaneously, a positive signature in region III develops. (b) Difference Auger electron spectra at two delays, which are marked in the top spectrum by arrows.

The simultaneous appearance of the negative feature in region II and the short-lived positive feature in region I resemble a shift of the AES of the photoexcited molecules to higher Auger electron kinetic energies with respect to the static AES. Taking into account the above considerations about the sensitivity of Auger electron spectra to structural dynamics, the blue shift suggests lengthening of at least one of the C–O bonds in the molecule. Simulations of the ground state and excited state potential energy surfaces and the excited state dynamics do indeed find evidence for a lengthening of the C–O bond at the oxygen, which is marked in figure 2, by 0.1 Å during relaxation from the ππ* Franck–Condon region to the nπ* state minimum [3,4,18]. Furthermore, simulations of the ππ* state at the Franck–Condon geometry and a geometry with a stretched C–O bond show a shift to higher Auger electron kinetic energies in agreement with the experiment [4]. The ground state and ππ* state spectra at the Franck–Condon geometry, on the other hand, exhibit only minor differences. The positive signal in region I decays with a time constant of 200 fs, which coincides with the signal increase in region III. Thus, the Auger electron spectra of the UV-excited molecules undergo a change from shift to higher kinetic energies to a shift to lower kinetic energies compared to the ground state spectra. This is probably due to a change in the electronic state character.

(b). Time-resolved NEXAFS probing

The ground state NEXAFS spectrum at the oxygen edge of thymine (see the black spectrum in figure 2a) exhibits a double peak structure with maxima at 531.4 and 532.2 eV [3,38]. The lower maximum is assigned to the O1 s → π* transition involving the oxygen which is marked in figure 2, the higher maximum to the O1 s → π* transition involving another π* orbital. The NEXAFS spectrum 2 ps after UV photoexcitation (green spectrum) in figure 2a exhibits a slight weakening of the π* transition of the static spectrum. Moreover, it exhibits a new signature at 526.4 eV, shifted down by approximately the photon energy of the pump UV light (4.65 eV) with respect to the ground state NEXAFS features. The cross-section of this transient feature is large, especially taking into account that it originates only from the UV-excited fraction of the molecules (an estimated 13%) [3]. Inspecting the time dependence of both transient features in difference-NEXAFS spectra (time-dependent UV-excited NEXAFS spectra minus static NEXAFS spectrum) reveals that the onset of the weakening of the ground state resonances precedes the onset of the transient feature at 526.4 eV by (60 ± 30) fs.

The negative signature can only originate from bleaching the ground state spectrum due to UV-photoexciting part of the ground state population. It, therefore, marks the start of the excited state dynamics. The delay of the 526.4 eV signature with respect to the photoexcitation fits perfectly to a relaxation of the photoexcited population from the ππ* excited state to the nπ* excited state through a conical intersection. Photoexcitation to the ππ* excited state opens an O1 s → π transition, which is, however, expected to be weak due to the high degree of delocalization of the π orbital. Upon ππ*/nπ* internal conversion, the electron vacancy switches from the π to the n-orbital opening an O1 s → n transition with a high cross-section due to the high degree of localization of the n-orbital at the oxygen site. The assignment is supported by quantitative simulations of ground state and excited state NEXAFS spectra on coupled-cluster level [3]. The simulations show a remarkable degree of invariance to the molecular geometry, suggesting a selective sensitivity of TR-NEXAFS to rapid changes in the electronic structure as encountered in the present case for the ππ*/nπ* internal conversion.

5. Discussion

(a). Sensitivity of time-resolved Auger versus time-resolved NEXAFS

Excited state Auger electron spectra exhibit a strong sensitivity to bond distances between the probed site and neighbouring atoms, in the case of TR-AES of thymine on the C–O distance [4]. A bond length increase by 0.1 Å results in a blue shift of the excited state Auger electron spectra by several eV. A strong shift to lower Auger kinetic energies was speculatively attributed to a change in the electronic character of the excited state. This effect becomes clearer in a different study, where fragmentation dynamics after multiphoton ionization of thymine were probed with AES [53]. The emergence of sharp peaks from initially only weakly structured Auger electron spectra could be observed within several hundred fs. The peaks could be identified as the Auger electron signatures of HNCO, a known photofragment of thymine. The Auger electron peaks of HNCO were initially shifted to slightly lower kinetic energies than in static Auger electron spectra of HNCO. This shift disappeared on a sub-ps timescale. The effect could be attributed to the separation of the neutral HNCO fragment from the remaining cationic fragment of thymine.

The sensitivity of AES to changes in the electronic structure is, however, substantially more difficult to rationalize than in the case of NEXAFS spectroscopy, because qualitatively accurate pictures are missing and the simulation of TR-AES is still challenging [39]. As opposed to TR-AES, our simulations find only minor sensitivity of the position of the nπ* feature in the TR-NEXAFS spectra (figure 5) to changes in nuclear geometry implying that the potentials of the valence-excited and core-excited states are approximately parallel. This observation agrees with our simulations of excited state Auger spectra, where the core-ionized state potentials are rather featureless and the changes in the Auger electron spectra are mostly introduced by the strong modulations in the dicationic state potentials. Likewise, the intensity of the nπ* feature is not significantly sensitive to the nuclear geometry. Hence, we interpret it as an effect mainly induced by a change in the electronic part of the molecular wave function, the transition of the electron vacancy from the π to the n-orbital. Simulations of TR-NEXAFS spectra of photoexcited ethylene predict rich signatures of the underlying non-adiabatic dynamics [54,55]. The underlying reason is, however, likewise an effect in the electronic structure, a symmetry break induced by pyramidalization around one of the carbon atoms close to the conical intersection.

Figure 5.

Adapted from ref. [3]: (a) near-edge X-ray absorption fine structure (NEXAFS) spectra at the oxygen edge of thymine in the ground state and 2 ps after ultraviolet (UV) photoexcitation. Two UV-induced changes can be identified, the appearance of a strong new feature at 526.4 eV and a weak bleach of the ground state signature at 531.4 eV. (b) False-colour plot of time-dependent difference-NEXAFS spectra, showing the time dependence of the tow UV-induced features. (c) Time-dependence of the integrated features at 526.4 and 531.4 eV.

(b). Consequences of sensitivities for thymine dynamics

The TR-AES and TR-NEXAFS spectroscopy studies draw a consistent picture of the excited state dynamics of thymine. TR-AES provides information about nuclear dynamics by encoding C–O bond elongation of the nuclear relaxation out of Franck–Condon region of the ππ* state. The disappearance of the bond stretching feature in combination with a shift of the TR-AES spectrum to lower kinetic energies is probably due to a change of the electronic state. TR-AES is, however, not very indicative of the character of this state. TR-NEXAFS spectroscopy, on the other hand, pins down the timescale of ππ*/nπ* internal conversion through a conical intersection by monitoring transient changes in the excited state electronic character.

The significance of these experimental probes of structural and electronic dynamics becomes apparent when compared to quantum chemical simulations. Since ππ* excitation transfers an electron to the C–O antibonding π* orbital, a Franck–Condon active C–O stretching mode can be expected. This agrees with multireference wavepacket simulations, which predict the C–O elongation on the way to a local minimum of the ππ* state [18,19,29]. However, they predict trapping of the population in this minimum on the ps timescale, which is in clear contradiction to both our soft X-ray studies. Static investigations of the ππ* and nπ* potential close to the Franck–Condon region on the coupled-cluster theory level also suggest C–O bond elongation early in the relaxation dynamics, but rather on the way from the conical intersection to the nπ* state minimum [3]. The TR-AES feature associated with the bond lengthening disappears on the sub-ps timescale, whereas TR-NEXAFS predict internal conversion to the nπ* state on a similar timescale. Since static investigations of the ππ* and nπ* potential energy surfaces also predict a readily accessible ππ*/nπ* conical intersection and cannot confirm a local ππ* minimum [3,20], the shapes of the calculated excited states in dynamical calculations might be contaminated by artefacts from an unbalanced description of the nπ* and ππ* characters in multireference wave functions. Our experimental investigations, on the other hand, cannot exclude trapping of a minor fraction of the excited state population in the ππ* state.

6. Conclusion and outlook

The combination of TR-AES and TR-NEXAFS spectroscopy provides a compelling tool set for investigations of photophysics and photochemistry, especially for organic molecules like the nucleobase thymine. The wide-spread application of these spectroscopic tools is currently impeded by limited accessibility of ultrafast soft X-ray light sources. This obstacle will be removed in the coming years with the upcoming second generation of XFEL light sources. Furthermore, table-top soft X-ray sources based on high harmonic generation are now becoming intense enough for the investigation of ultrafast photochemistry [56–58]. Owing to their broadband spectra, they are ideal for TR-NEXAFS investigations, whereas TR-AES investigations will probably rely on XFEL sources in the future.

Data accessibility

This article has no additional data.

Authors' contributions

T.W. and M.G. wrote the manuscript together.

Competing interests

We declare we have no competing interests.

Funding

T.W. is supported by the AMOS programme within the Chemical Sciences, Geosciences, and Biosciences Division of the Office of Basic Energy Sciences, Office of Science, US Department of Energy. M.G. acknowledges the Volkswagen Foundation for a Lichtenberg Professorship.