Dynamics of ultraviolet-induced DNA lesions: Dewar formation guided by pre-tension induced by the backbone

For the study a T(6-4)T dinucleotide containing a bio-isosteric formacetal linker instead of the natural phosphodiester was synthesized using a published procedure [32]. Briefly, the formacetal-linked thymine dinucleotides were prepared and purified as described previously [34]. The compounds were irradiated with UV light at 254 nm under anaerobic conditions to give the T(6-4)T dinucleotides [35]. Subsequently, the T(6-4)T dinucleotides were analysed and purified by reverse-phase HPLC.

Steady-state spectra were recorded with an UV/Vis absorption spectrometer (Perkin-Elmer, Lambda 750), a fluorescence spectrometer (Horiba, Fluorolog 3) and a Fourier transform IR spectrometer (FTIR) (Bruker, IFS 66). To avoid an inner filter effect when recording fluorescence spectra, the optical density of the sample solution was kept below 0.05 (for a path length of 1 cm). The fluorescence spectra were corrected for the spectral sensitivity of the instrument. 2-(1-Naphthyl)-5-phenyloxazole (1-NPO; supplied by Lambda Physics) in cyclohexane was used as a reference dye for the determination of the fluorescence quantum yields.

Reaction quantum yields were determined using the FTIR spectrometer as a detector. The fourth harmonic at 266 nm of a Q-switched Nd:YAG laser (model: NY60-B, Continuum) served as a light source. The average power of the actinic light amounted to 7.4 mW.

2.3.1. Fluorescence

2.3.2. Transient absorption spectroscopy

2.4.1. Quantum chemical calculations of electronic excited states

2.4.2. IR- and difference IR-absorption spectra

Figure 2 shows ground-state absorption spectra of the dinucleotides depicted in figure 1. The dinucleotide TpT (grey) exhibits a strong electronic absorption band centred around 260 nm as expected for the intact DNA base thymine [17]. UV excitation of this ππ* transition can lead to the formation of the T(6-4)T lesion (black). Compared to TpT, the lowest ππ* absorption band of the T(6-4)T lesion is lower in oscillator strength and red-shifted by about 75 nm. It peaks in the UV-A range at about 325 nm. With the formation of the T(Dewar)T by illumination at e.g. 355 nm the characteristic ππ* absorption of the T(6-4)T lesion at 325 nm disappears. The T(Dewar)T does not show a characteristic absorption band in the near UV. The absorption is shifted below 250 nm, which hinders the investigation of the formation of the Dewar isomer in the UV range.

Zoom In Zoom Out Reset image size

The fluorescence spectrum of the T(6-4)T (blue) (after excitation at 325 nm) shows an emission peak at 400 nm. A similar value of 393 nm has been reported for the (6-4) lesion of TpT by Blais et al [53]. The small shoulder of the emission spectrum at ∼350 nm is due to the Raman scattering signal from deuterated water. The fluorescence quantum yields of intact DNA bases are extremely small, of the order of 10^–4 (due to ultrashort excited state lifetimes) [17, 54–57]. The fluorescence quantum yield of T(6-4)T was estimated via an analysis according to Strickler and Berg [58], resulting in a value of the order of 1% in general agreement with the value of 3% reported in [53].

To gain information on the lifetime of the excited state of T(6-4)T after photo-excitation (at 320 nm), the fluorescence emission has been monitored on the picosecond and nanosecond time scale with a streak camera setup. The results of these measurements are given in figure 3. The upper diagram (figure 3(a)) is a contour representation of the fluorescence emission observed between 350 and 600 nm at 301 K. The plot covers a time window of up to 800 ps and shows a fluorescence emission with a spectral shape similar to the steady-state emission spectrum (see figure 2). A global fit of the emission spectrum reveals more details. The fit routine yields two decay components (given as dotted and solid lines on the right axis) peaking at 420 and 405 nm. They show small differences in spectral shape. The slower component exhibits a much weaker amplitude, indicating that the corresponding emitting state was populated to a smaller extent. Assuming the same emission cross-section for both components, one can deduce a relative population of 13 and 87%, respectively.

Zoom In Zoom Out Reset image size

The kinetics of the fluorescence emission at 301 K is shown in figure 3(b) by plotting the emission intensity at 432 nm (with logarithmic scaling) versus the delay time in nanoseconds. The data are well reproduced by a sum of two exponentials (red line) representing the two components mentioned above: (i) a strong component 1 decays with a time constant of ∼106 ps and (ii) a weaker component 2 (by a factor 0.14) with a time constant of ∼910 ps. At late delay times (5 ns) a small deviation between fit and data is visible. This can be attributed to the instrumental offset at very small emission intensities. Performing experiments at different temperatures (between 291 and 334 K) reveals a strong temperature dependence of the decay times. An overview of the temperature-dependent results is given in table 1. Rising the temperature from 291 to 334 K leads to a decrease of the time constants from 137 to 64 ps (for decay component 1) and from 1.24 to 0.5 ns (for decay component 2); the relative amplitudes remain essentially constant. An Arrhenius plot of the data showing the observed temperature dependence of the rate constants for decay component 1 (red squares) and 2 (black dots) is given in the supplementary data (figure S1), available from stacks.iop.org/NJP/14/065006/mmedia. Both decay times decrease with increasing temperature with slightly different slopes. Analysis of the data with the Arrhenius equation k = k⁰ exp(−E_A/k_BT) yields calculated activation barriers of E_A1 = 1090 cm⁻¹ and E_A2 = 1467 cm⁻¹ for the fast and the slow component, respectively. The derived values are within an error margin of the order of 10% (see the supplementary data).

The absorption changes induced after excitation at 323 nm of the T(6-4)T lesion have been recorded in the UV and Vis spectral ranges. In figure 4(a), a contour plot shows the observed transient absorption between 350 and 700 nm. Positive absorption changes (induced absorption) are given in yellow/red, while negative absorption changes are indicated in cyan/blue. The delay time is given in a linear scale around time zero until 1 ps and with logarithmic scaling thereafter. Immediately after excitation a broad induced absorption is observed throughout most of the Vis spectral range. An absorption decrease is found in the range of the fluorescence emission at about 400 nm. Figure 4(b) depicts transients at the indicated probing wavelengths covering the areas of induced absorption in the Vis (550 nm) and UV (360 nm) part of the spectrum as well as in the range of about 400 nm with negative absorption changes (416 nm).

Zoom In Zoom Out Reset image size

The data were analysed using a global fit based on the sum of exponentials (solid blue lines in figure 4(b)). The fit delivers three decay times of 2.9 ps, 120 ps and 1.1 ns and an offset describing the absorption changes at very late delay times (>3 ns). The spectral characteristics of the observed absorption changes can be seen in figures 5(a) and (b). The upper panel (a) shows the transient spectra between 1 ps (black line) and 500 ps (blue line). One can see the overall decrease of induced absorption (between 450 and 700 nm) and negative absorption changes (between 400 and 450 nm). After 500 ps a residual positive and broad absorption change over the whole spectral region is visible. The residual absorption change at late delay times is given in more detail in figure 5(b), where absorption changes between 0.5 ns (blue line) and 3.5 ns (black line) are depicted. There is a further decrease of the induced absorption between 450 and 700 nm, while an increase in absorption occurs between 390 and 450 nm. For comparison, we plot also the absorption difference spectrum of the minimal model molecule 1-methyl-2(1H)-pyrimidinone (1MP; grey broken) 3 ns after photoexcitation (redrawn from [33]). These traces suggest that the long-time offset of both molecules may originate from the same transient species.

Zoom In Zoom Out Reset image size

The basic features found in emission and UV/Vis absorption experiments are as follows: the early absorption changes are the result of the population of the initially excited state and represent excited state absorption (ESA) and stimulated emission of the excited T(6-4)T lesion. On the time scale of a few picoseconds a first transient occurs (∼2.5 ps) with significant changes of the amplitude and spectral position of the stimulated emission. This indicates that the processes monitored are related to vibrational relaxation in the electronically excited state combined with solvation [59]. Subsequently, most of the ESA and the fluorescence emission decay with a time constant in the 100 ps range. Here we observe the dominant part of the decay of the excited electronic state. On the 1 ns time scale one finds the decay of ESA and fluorescence emission with much weaker amplitudes. After these processes the UV/Vis absorption changes indicate that most of the excited molecules have decayed to the electronic ground state, while small absorption changes remain. The latter are compared with the results obtained for 1MP. The obtained difference spectrum after 3 ns (dashed grey line) is characteristic of the triplet state of 1MP. The close resemblance between the 3 ns spectrum of 1MP and the 3.5 ns spectrum of the T(6-4)T lesion is an indication that a considerable amount of triplet states are formed after UV excitation of the T(6-4)T lesion. Since the Dewar valence isomer does not show absorption in the Vis or near UV the experiments cannot give indications on its formation. More detailed information on the photophysical and photochemical reaction pathways will be obtained below from steady-state and time-resolved IR spectroscopy.

The steady-state IR absorption spectra of the T(6-4)T and the T(Dewar)T lesion are given in figure 6. In the plotted spectral range (1625–1830 cm⁻¹) the absorption is dominated by C = O groups. In figure 6(a), the ground state absorption spectrum of the T(6-4)T lesion (red line) is compared with the ground state absorption spectrum of the T(Dewar)T lesion (black line) obtained after extended illumination by light at 355 nm. The resulting difference spectrum is given in figure 6(b) (black line). In the T(6-4)T lesion one can identify three absorption bands with peaks at 1650 cm⁻¹ (shoulder on the left), 1665 cm⁻¹ and 1712 cm⁻¹. In the T(Dewar)T lesion the absorption shows well-separated peaks at 1673 and 1714 cm⁻¹ and one at 1781 cm⁻¹. This latter region is typical of C = O oscillators where the C atom is incorporated into an aliphatic ring (e.g. as in a Dewar isomer). Therefore this band can be used as a marker for the formation of the T(Dewar)T lesion. Recently, the absorption band at 1781 cm⁻¹ was used to determine the formation yield of T(Dewar)T lesions. A quantum yield of about 8% has been determined in irradiation experiments with 355 nm excitation [32].

Zoom In Zoom Out Reset image size

For further interpretation of the IR spectra, quantum chemical calculations were performed (see the next section for details). Panel (c) in figure 6 shows the calculated ground state absorption spectra of T(6-4)T (red) and T(Dewar)T (black) as well as the absorption of the T(6-4)T triplet state (green). Difference spectra for different ratios of ground and triplet state contributions are given in figure 6(d). The calculated absorption spectrum shows three absorption bands for the singlet ground states of T(6-4)T. In the T(Dewar)T lesion a characteristic absorption band occurs at 1780 cm⁻¹. As expected this band represents the C = O vibration at the Dewar moiety. The calculated triplet state of T(6-4)T (green) shows two absorption bands with peaks at 1706 and 1674 cm⁻¹ which overlap with the two lower absorption bands of the T(Dewar)T. In the calculated difference spectra, the participation of a triplet state intermediate results in characteristic spectral signatures in the range between 1650 and 1725 cm⁻¹. This suggests that the presence of a triplet state can be monitored with time-resolved IR spectroscopy.

The results of time-resolved IR spectroscopy are given in figure 7. In the contour plot of figure 7(a), blue colour indicates negative absorption changes (bleach), while red colour indicates positive absorption changes. From 1 to 600 ps, one finds a strong bleach signal between 1625 and 1690 cm⁻¹. This bleach signal decays on the 100 ps time scale. On the same time scale there is the decay of an increased absorption centred at 1710 cm⁻¹. In the range of the T(Dewar)T marker band (1780 cm⁻¹, signal amplitudes scaled by a factor of 10), a positive absorption band grows concomitant with the decay in the lower frequency range.

Zoom In Zoom Out Reset image size

Looking at the transient spectra given in figure 7(b) one can see the spectral characteristics of the absorption changes on the picosecond time scale in more detail. The broad bleach between 1625 and 1690 cm⁻¹ occurs in the range of the ground state absorption of the T(6-4)T lesion with a peak at about 1665 cm⁻¹ and a shoulder at 1640 cm⁻¹. The absorption increase at 1710 cm⁻¹ can be tentatively assigned to an excited state C = O vibration. A global analysis of the data reveals several dynamic processes. One finds a fast 3.4 ps transient in accordance with the results of time-resolved experiments in the UV/Vis. Recovery of the ground state absorption and the formation of the marker band at 1780 cm⁻¹ occur with the time constant of τ = 128 ps. The appearance of this band gives unequivocal evidence for the formation of the T(Dewar)T on this time scale.

In an additional transient IR experiment, we investigated the processes on the nanosecond to the microsecond time scale. The results are given in figures 7(c) and (d) together with the ground state absorption of the T(6-4)T lesion (broken curve). The transient spectra recorded between 0.2 and 2 ns in figure 7(c) show a strong change of the bleach in the region of the ground state absorption band at 1663 cm⁻¹. The dynamics can be separated into a fast (about 1 ns) and a slower component. The latter leads to a further recovery of the ground state absorption on the 500 ns time scale. The situation is different in the low-frequency part below 1650 cm⁻¹ where absorption recovers partially on the 100 ps range but stays constant afterwards. Within the experimental accuracy the T(Dewar)T marker band at 1780 cm⁻¹ also does not change after t > 500 ps. At the end of our observation period at 10 μs we observe a very good match between the transient spectrum (blue line) and the steady-state difference spectrum (light grey area) in figure 7(d). With this information we can conclude that the T(Dewar)T is formed predominantly on a time scale of 130 ps.

The changes in the range of the ground state absorption of the T(6-4)T lesion (1600–1700 cm⁻¹) on the ns time scale need further discussion. The results from time-resolved UV/Vis experiments suggest the rapid formation of a triplet state (see the comparison with 1MP given in figure 5(b)). This is in line with the fact that intersystem crossing (ISC) to a triplet state is promoted by carbonyl functions of the T(6-4)T moiety [60]. For ISC one would expect to find IR absorption changes as derived for the calculated difference spectra in figure 6(d). Indeed, one finds prominent changes around 1660 cm⁻¹ on the 500 ns time scale as expected for a triplet decay.

The experimental results obtained can be summarized in a reaction model as given in figure 8. UV excitation from the ground state (blue) populates T(6-4)T in the Franck–Condon region (FC T(6-4)T*, red). From the Franck–Condon region the molecules evolve on the excited state surface. We find a time constant of about 2.5 ps for the relaxation to local minima on the excited state surface. The experimental observation of two time constants in the decay of the excited electronic state is indicative of the population of two local minima in the excited state (light and dark purple boxes). The amplitudes in fluorescence emission point to relative populations of 13 and 87% for the components with life times of 1 ns and 130 ps, respectively. The experiments clearly show that the T(Dewar)T (orange) is formed via the 130 ps process with an overall quantum yield of 8% and that Dewar formation goes in parallel with the decay of the excited singlet state. At later times (>1 ns) there are signatures for a triplet state (green). The triplet is populated with a significant quantum yield in the range of 2–10% and decays on the 500 ns time scale (for details of the estimate, see supporting data (available from stacks.iop.org/NJP/14/065006/mmedia)). There are no indications that the triplet state is involved in the formation of the Dewar isomer. Further information addressing reaction paths, possible energy barriers and conical intersections will be discussed below using quantum chemical calculations.

Zoom In Zoom Out Reset image size

4.1.1. Singlet excited state relaxation

The observed excited state lifetime on the 100 ps time scale suggests the existence of an excited state minimum. In figure 9, different areas of the potential energy surfaces involved are shown. The left part displays the four lowest singlet states along a linear interpolated coordinate (LIC) from the Franck–Condon (FC) structure to the S₁ minimum. Optical excitation occurs into the lowest ππ* state (S₂, green). The excitation energy S₀ → S₂ of 4.48 eV on the highest-level ONIOM(MS-CASPT2-CAS(12/9):HF) calculation is in good agreement with the experimental value of 3.84 eV. On this level of theory the ππ* state lies between the dark states with n_Nπ* (red) and n_Oπ* (brown) character. Relaxation (dashed black arrow) from S₂ towards the S₁ minimum involves a curve crossing from S₂ to S₁. Thereby the electronic character of S₁ changes from n_Nπ* in the Franck–Condon region to n_Oπ* with some amount of ππ* at the S₁ minimum. The close succession of electronic states in T(6-4)T* is intriguingly similar to the isolated chromophore 5M2P [61, 62]. The S₁ minimum in T(6-4)T has nearly planar structure with butterfly distortions. The right part of figure 9 connects the S₁ minimum to the reactive part of the potential energy surface including the conical intersection (CoIn) region along the intrinsic reactive coordinate (IRC). It is calculated on the ONIOM(CAS(12/9):HF level of theory. The S₁ minimum is below the CoIn (by 0.4 eV) and the potential energy surface in the excited state is shallow. Nevertheless, the S₀/S₁ CoIn_Dewar is not barrierless accessible, as indicated in the grey region where the connection between the two parts of the potential energy surface is sketched. The small changes in potential energy along the reaction coordinate support the experimental observation of an excited state decay on the 100 ps time scale.

Zoom In Zoom Out Reset image size

4.1.2. Reactive conical intersection in T(6-4)T

Conical intersections are known to play a dominant role in many ultrafast photochemical reactions. Their accessibility and topography determine reaction velocity and product yield. In the direct vicinity of a CoIn the potential energy can be expressed as the first order of a Taylor expansion in the two-dimensional-branching space [63] defined by the two vectors x₁ (gradient difference) and x₂ (derivative coupling) and the coordinates x₁ and x₂. The topography of the CoIn can be described by [50]

$$\begin{equation} E_{1/2} = d_{{\bf{x}}_{\bf{1}} {\bf{x}}_{\bf{2}} } \left[ \sigma _{{\bf{x}}_{\bf{1}} } x_1 + \sigma _{{\bf{x}}_{\bf{2}} } x_2 \pm \left( {\frac{1}{2}\left( {x_1 ^2 + x_2 ^2 } \right) + \frac{{\Delta _{{\bf{x}}_{\bf{1}} {\bf{x}}_{\bf{2}} } }}{2}\left( {x_1 ^2 - x_2 ^2 } \right)} \right)^{1/2} \right] \end{equation} \tag{ 1 }$$

with the parameters

$$\begin{eqnarray} &&\begin{array}{ll} \displaystyle \sigma _{\bf x_1} = \frac{{\bf S}\cdot{\bf{x}}_{\bf{1}}}{d_{{\bf{x}}_{\bf{1}} {\bf{x}}_{\bf{2}} } },\qquad \sigma _{{\bf{x}}_{\bf{2}} } = \frac{{\bf S}\cdot{\bf{x}}_{\bf{2}}}{d_{{\bf{x}}_{\bf{1}} {\bf{x}}_{\bf{2}} } }, \\ \displaystyle \Delta _{{\bf{x}}_{\bf{1}} {\bf{x}}_{\bf{2}} } = \frac{{\bf{x}}_{\bf{1}}^2 - {\bf{x}}_{\bf{2}} ^2 }{d_{{\bf{x}}_{\bf{1}} {\bf{x}}_{\bf{2}}^{2} } },\qquad d_{{\bf{x}}_{\bf{1}} {\bf{x}}_{\bf{2}} } = \left( {{\bf{x}}_{\bf{1}}^2 + {\bf{x}}_{\bf{2}}^2 } \right)^{1/2} \end{array} \end{eqnarray} \tag{ 2 }$$

and the gradient sum vector S.

The result of the Taylor expansion leads to the famous double-cone structure of the CoIn. The tilt of the double cone from the vertical axis is given by σ_xi. Note that Δ_x1x2 describes the deviation from cylindrical symmetry and d_x1x2 characterizes the slope of the cone. A CoIn with perfect cylindrical symmetry (Δ_x1x2 = 0, σ_xi = 0) is designated as a peaked CoIn. All other situations lead to a sloped CoIn. The shape of the cone has a strong influence on the branching ratio of the reaction since it controls the initial gradient towards the product channels. A photoexcited system stays only for an ultrashort period in the vicinity of the CoIn. Once the system has reached the CoIn, the transfer to the low-lying state occurs immediately.

We analyse the reactive CoIn_Dewar from which Dewar formation is possible (IRC = 0.57 in figure 9, right). In this context, we compare the topography of CoIn_Dewar in the dinucleotide with that in the isolated 5M2P (CoIn_5M2P). The cone parameters of both CoIn are given in table 2. The resulting cones are visualized in figure 10. CoIn_5M2P (left) is almost peaked, while CoIn_Dewar (right) is sloped with a stronger tilt from the vertical axis (σ_xi) and a larger deviation from cylindrical symmetry (Δ_x1x2). The peaked CoIn topography in 5M2P efficiently mediates a photophysical excited state deactivation as indicated by the yellow arrows. No Dewar valence isomerization occurs. The motions of hydrogen atoms of the 5M2P ring (see the structure in figure 10, left) are free and the lowest energy CoIn_5M2P is characterized by a sofa structure [64, 65]. To reach CoIn_5M2P a crucial sp² → sp³ rehybridization at C4 is required and facilitated by the mobility of the hydrogen atoms. The situation changes for the dinucleotide where a strong clamping effect hinders the sp² → sp³ rehybridization. The sp² character of the C4 atom is maintained and CoIn_Dewar becomes the lowest accessible nonradiative relaxation funnel. Displacement along the direction of the derivative coupling vector x₂ further reduces the N3–C6 distance, which is a prerequisite to form the new Dewar bond. In figure 10 (right), the reactive channel is indicated by the red arrow and the photophysical deactivation by the yellow arrow. In the case of the dinucleotide, the excited state potential surface is shallow with small barriers between the Franck–Condon range and the CoIn_Dewar. Therefore Dewar formation is mainly determined by the gradients at the CoIn. The calculations and the experiments clearly show that the Dewar formation only occurs when the dinucleotides are connected by a backbone. In the studied molecule the backbone was a formacetal linker. Other isosteric linkers as in the DNA backbone are expected to act in the same way.

Zoom In Zoom Out Reset image size

Table 2. Cone parameters (in a.u.) characterizing the potential energy surfaces of S₀ and S₁ in the vicinity of CoIn_Dewar and CoIn_5M2P.

Structure	σx1	σx2	Δx1x2	dx1x2
CoInDewar	−0.574	−0.915	−0.240	0.104
CoIn5M2P	−0.121	−0.574	0.000	0.141

4.1.3. The role of triplet states

Further support for the participation of the triplet channel in the photophysical deactivation of T(6-4)T* can be inferred from the comparison of experimental and simulated IR spectra shown in figure 6. The IR absorption of T(6-4)T ground state (red lines in figures 6(a) and (c)) shows three strong C = O bands in the 1650–1730 cm⁻¹ frequency region. The three C = O stretch vibrations are strongly coupled. Accordingly, the displacement of the three carbonyl groups is correlated. The lowest energy normal mode at ν = 1668 cm⁻¹ arises from an anti-symmetric linear combination of two carbonyl groups, the pyrimidinone C = O, as well as one C = O of the adjacent thymidilyl ring. The third C = O stretch vibration (located on the thymidilyl ring) has negligible displacement contributions. The mode at ν = 1682 cm⁻¹ is composed of a symmetric linear combination of all three carbonyl groups and carries the highest intensity. The mode at ν = 1713 cm⁻¹ shows the weakest intensity and arises mainly from the two thymidilyl carbonyl groups, the C = O group of the pyrimidinone chromophore contributes only weakly. Compared to the experimental steady-state IR spectrum (red line in figure 6(a)) we find excellent agreement in the intensity pattern of all three normal modes. Note the blue shift of the low-frequency bands (<1750 cm⁻¹) by 20 cm⁻¹ in the calculated spectrum.

In T(Dewar)T the modified chemical environment of the Dewar C = O group leads to a pronounced shift (87 cm⁻¹) towards higher frequencies (ν = 1775 cm⁻¹, black line in figure 6(c)). The now localized Dewar C = O mode serves as a unique probe for the kinetic and quantum yield of the valence isomerization reaction [32] (section 3.1.5). The two remaining carbonyl groups, located on the thymidilyl ring, show up as symmetric (ν = 1714 cm⁻¹) and anti-symmetric (ν = 1683 cm⁻¹) C = O stretch modes. As the Dewar C = O mode decouples from the thymidilyl C = O modes the IR absorption in the low-energy region is reduced.

Additionally, we depict the calculated IR spectrum of the triplet T₁ of the T(6-4)T lesion (green line in figure 6(c)). The spin density is predominantly localized on the pyrimidinone entity, with an unpaired electron occupying the lowest anti-bonding orbital with C = O* character. As a consequence, the pyrimidinone carbonyl bond is elongated from 1.23 to 1.71 Å. The weakend C = O bond shifts to 1537 cm⁻¹ (data not shown). Similar shifts upon triplet formation have been found by Hare and coworkers for thymine and thymidine in acetonitrile [68]. The two C = O modes of the thymidilyl entity are slightly red-shifted (Δν = 8 cm⁻¹; symmetric C = O stretch: ν = 1706 cm⁻¹; asymmetric: ν = 1674 cm⁻¹).

Simulated IR difference spectra for the T(6-4)T to T(Dewar)T and triplet conversion are depicted in figure 6(d). We consider three limiting cases. The black line represents the exclusive Dewar formation. The green line represents the exclusive triplet formation. The blue line represents equal amounts of Dewar and triplet formation. Although the IR spectra of triplet and Dewar are similar, the C = O region between 1650 and 1700 cm⁻¹ of the difference spectra can be used to identify a triplet state contribution. The major influence of a triplet state is seen at about 1685 cm⁻¹. At about 1665 cm⁻¹ smaller changes are expected. Comparison with the time-resolved IR data on the ns time scale (figure 7(c)) reveals a similar pattern as expected with the shift by −20 cm⁻¹ (discussed above). While Dewar concentration is not affected between 2 ns and 10 μs (see figure 7(c), the marker band at ν = 1780 cm⁻¹ stays practically constant) the observed pattern points to the decrease of triplet concentration in this time range.

From the calculations we could identify two rotational conformers of the T(6-4)T lesion which can be connected to four different T(Dewar)T isomers. They differ by the relative position of the newly formed Dewar bond (see the supplementary data (figure S3), available from stacks.iop.org/NJP/14/065006/mmedia) and the respective rotation of the former pyrimidone unit around the 6-4 σ-bond. Related structures have been proposed in the literature based on x-ray and NMR measurements [12, 69]. The Dewar isomers differ in their free reaction energies and exhibit small changes in their calculated IR spectra (see the supplementary data (figure S4)). The small variations in the reactant free energy differences (ΔG = 0.05 eV) suggest that different local minima are excited by light absorption. They may be responsible for the occurrence of side reactions. It is expected that not all side reactions have the same probability for Dewar formation or internal conversion. This can explain the experimental observation of two fluorescing states with different decay times (see figure 8).