Electronic and Structural Elements That Regulate the Excited-State Dynamics in Purine Nucleobase Derivatives

The excited-state dynamics of the purine free base and 9-methylpurine are investigated using experimental and theoretical methods. Femtosecond broadband transient absorption experiments reveal that excitation of these purine derivatives in aqueous solution at 266 nm results primarily in ultrafast conversion of the S2(ππ*) state to the vibrationally excited 1nπ* state. Following vibrational and conformational relaxation, the 1nπ* state acts as a doorway state in the efficient population of the triplet manifold with an intersystem crossing lifetime of hundreds of picoseconds. Experiments show an almost 2-fold increase in the intersystem crossing rate on going from polar aprotic to nonpolar solvents, suggesting that a solvent-dependent energy barrier must be surmounted to access the singlet-to-triplet crossing region. Ab initio static and surface-hopping dynamics simulations lend strong support to the proposed relaxation mechanism. Collectively, the experimental and computational results demonstrate that the accessibility of the nπ* states and the topology of the potential energy surfaces in the vicinity of conical intersections are key elements in controlling the excited-state dynamics of the purine derivatives. From a structural perspective, it is shown that the purine chromophore is not responsible for the ultrafast internal conversion in the adenine and guanine monomers. Instead, C6 functionalization plays an important role in regulating the rates of radiative and nonradiative relaxation. C6 functionalization inhibits access to the 1nπ* state while simultaneously facilitating access to the 1ππ*(La)/S0 conical intersection, such that population of the 1nπ* state cannot compete with the relaxation pathways to the ground state involving ring puckering at the C2 position.

A high-intensity absorption band is also observed at energies above 5.6 eV (< 220 nm in Figure 1). This absorption band red-shifts and decreases in intensity in going from acetonitrile to aqueous solution in both the purine free base and 9-methylpurine, and a further red-shift and decrease in intensity is observed for 9-methylpurine in cyclohexane. This complex solvatochromic behavior suggests that multiple electronic transitions are contributing to this absorption band in different solvents, and from both the N7H and N9H tautomers in the case of the purine free base. This is in agreement with the present and previous calculations, 7 which suggest a superposition of electronic transitions with * and n* character above 5.6 eV for both of the N7H and N9H purine tautomers in vacuum.
The results presented in this work show that 9-methylpurine and the purine free base exhibit very small fluorescence yields of ca. 10 -3 . For 9-methylpurine, the emission maximum blue-shifts and the fluorescence yield decreases approximately fourfold in going from water to cyclohexane. In contrast, the emission maximum in the purine free base does not change appreciably, while its fluorescence yield exhibits approximately a twofold decrease upon going from water to acetonitrile. The comparison of the emission results for 9-methylpurine and the purine free base indicates that both the N7H and N9H purine tautomers contribute to the observed fluorescence. Moreover, based on the theoretical calculations, it is proposed that the observed fluorescence emission originates from the 1 n* state of both the N7H and N9H purine tautomers and also from the 1 n* state of 9-methylpurine. This is consistent with the ultrafast decay of the 1 * state and the hundreds-of-picoseconds intersystem crossing of the 1 n* state to the triplet manifold. In fact, the experimental fluorescence maximum for 9-methylpurine of ~370 nm in cyclohexane, as well as the experimental E 0,0 energy reported in Table S1 of the SI, are in good agreement with the computed vertical (375 nm) and adiabatic (326 nm) emission energies. The S 0 (3.45 D) to S 1 (2.54 D) change in the dipole Figure 1 are in good agreement with those reported by Drobnik and Augenstein in 1966. 4 They are also supported by single photon counting experiments, 23 where a long-lived fluorescence lifetime of 5 ns was reported for the purine free base in liquid N 2 at 77K.

Broadband Transient Absorption Spectrometer.
The experimental setup and data analysis procedure have been described previously. [24][25][26] Briefly, a Quantronix Integra-i/e 3.5 Laser generating laser pulses centered at 800 nm with a repetition rate of 1 kHz was used to seed an optical parametric amplifier (OPA, TOPAS, Quantronix / Light Conversion). Laser pulses with a full-width-half-maximum of 100 fs were measured using a GRENOUILLE (Model 8-50, Swamp Optics), which is based on the frequency-resolved optical gating (FROG) technique developed by Trebino and coworkers. 27 The OPA was tuned to the excitation wavelength of 266 nm. Contributions to the excitation beam by other wavelengths or polarizations were removed using a reflective wavelength filter (-filter) and a Glan-Taylor prism. The excitation pulses were attenuated to the desired intensity of 1 mJ using a neutral density optical filter in order to minimize cross phase modulation effects and sample degradation. 28,29 Long-lived transient absorption signals originating from hydrated electrons formed by two-photon ionization of the water solvent 30 were effectively minimized under the experimental conditions used in this work. Rotational relaxation effects were removed from the transient absorption signals by randomizing the polarization of the excitation pulses by using a depolarizing plate in the spectrometer.
A broadband transient absorption spectrometer (Helios, Ultrafast Systems, LLC) was used for data acquisition. A continuously moving 2 mm CaF 2 crystal was used for white light continuum generation (WLC) giving access to the spectral range from 350 to 675 nm. The probe pulses were corrected for group velocity dispersion (GVD) using a home-made LabView (National Instruments, Inc.) program as described previously. 24 A pump/probe beam diameter ratio of three was used.
Data analysis was performed using Igor Pro 6.12A software (Wavemetrics, Inc.), as described previously. 31 Nineteen kinetic traces were selected from the multidimensional data set for each compound at selected probe wavelengths and analyzed using a global fitting subroutine set up in the Igor Pro software. The global and target analysis method 32,33 based on a sequential kinetic model 34 was used to obtain the excited-state lifetimes and decay associated spectra. The sequential model rate law was composed of three exponential components. 34 This function was convoluted with a Gaussian-shaped instrument response function with a FWHM of 200 fs. The time constants were linked for all the traces while the amplitudes were left wavelength-dependent. The reported uncertainties for the lifetimes shown in Table 1 are twice the standard deviation (2σ) obtained from the global analysis of at least three independent sets of experiments for each compound.
The absorbance of 9-methylpurine (9MP) and purine (P) solutions were in the range of 0.5 to 1 at the excitation wavelength of 266 nm, which corresponds to millimolar concentrations of the bases. A 2 mm optical path length cell was used in the pump-probe experiments. The samples were continuously stirred using a Teflon-coated stir bar and a magnetic stirrer to avoid re-excitation of the excited volume by successive laser pulses.
Solutions were carefully monitored for photodegradation using a spectrophotometer (Cary 100, Varian, Inc.) and were replaced with fresh solutions before a 10 % decrease in steady-state absorbance at the excitation wavelength was observed.

Methodology used for Estimating E 0,0 and Fluorescence Quantum Yields. Steady-state
absorption and emission spectra were measured at room temperature using Cary 100 Bio and Cary Eclipse spectrophotometers (Varian, Inc.), respectively. Background corrections were performed by subtracting a solvent-only scan under the same experimental conditions. Fluorescence spectra were taken at high PMT voltage with slit-widths of 5 nm and averaging times of 2 s. All solutions were made with an optical density of ~0.2 at the excitation wavelength of 267 nm.
Calculations of fluorescence quantum yield were performed using Equation 1: where Φ f represents the fluorescence quantum yield, OD is the optical density (i.e. steadystate absorption at the excitation wavelength), I is the integrated area under the emission spectrum, and n is the refractive index of the respective solvent. 35 The subscripts s and un stand for the fluorescence standard and unknown, respectively. When calculating quantum yields, it is common to use the same solvent for the standard and unknown; thus the n terms are made equal by assuming the low concentrations used do not change the solvent refractive index. While this was done for the solutions in PBS, using the reported Φ f s for Tryptophan (Trp) of 0.13 ± 0.1, 35 the necessary standard Φ f of Trp in acetonitrile and cyclohexane could not be found; therefore, the refractive indices of these solvents were used to calculate the respective Φ Fl of purine free base and 9-methylpurine.
The transition energy between the lowest vibrational levels of the ground state and 1 ππ* state were determined for the purine free base and 9-methylpurine from the crossing point of their absorption and emission spectra (Table S1 and Figures S1 and S2). Absorption spectra were used rather than excitation spectra for these E 0,0 determinations due to the DOI: 10.1021/ja512536c J. Am. Chem. Soc. 2015, S8 difficulty of obtaining accurate excitation spectra at wavelengths shorter than 300 nm, especially working with compounds having low fluorescence quantum yields. 36 Emission spectra were obtained on a wavelength (nm) scale using a constant band-pass setting; therefore, following their conversion to wavenumbers (cm -1 ), the emission intensities needed to be corrected by multiplying each intensity by its collection wavelength squared (λ 2 ). 35 Table S1. Fluorescence quantum yields and zero-zero energies for purine free base (P) and 9-methylpurine (9MP) in aqueous buffer solution (PBS, pH 7), acetonitrile (ACN), and cyclohexane (CHX)

Base
Solvent Φ Fl  10 -3 a E 0,0 (cm -1 ) E 0,0 (eV)  Figure S1. Emission spectra of purine free base (P) and 9-methylpurine (9MP) in aqueous buffer solution, pH 7 (PBS), acetonitrile (ACN), and cyclohexane (CHX) recorded in backto-back with identical optical densities at the excitation wavelength (OD = 0.20 at 267 nm).  Figure S2. Absorption and emission spectra used to calculate the fluorescence quantum yields and zero-zero energies of the purine free base (P) and 9-methylpurine (9MP).   Table S2 and Figure S5 were performed using the Gaussian 03 suite of programs. 37 Groundstate optimizations were performed using the parameter-free PBE0 functional. 38 Optimized ground-state geometries were calculated without any geometrical restriction. The gas-phase optimized geometries for 9MP and 9MP(H 2 O) 3 complex were confirmed to be local minima on the ground-state potential energy surface through vibrational frequency analysis at the PBE0/6-31G level of theory. The ground-state geometry of 9MP was further optimized in vacuum at the PBE0/6-31+G(d,p) level of theory. Excited-state calculations were performed using the PBE0 functional. The PBE0 functional provides accurate excited-state energies, 39,40 singlet-triplet energy gaps, 41 excited-state ordering, 22,42 and ground-state absorption spectra, 39 particularly when solvent effects are taken into account. 39,42,43 Solvent effects on the excitedstate vertical energies were modeled by performing self-consistent reaction field (SCRF) calculations using the polarizable continuum model (PCM) 44 with the integral equation formalism (SCRF=IEFPCM). 45 Additional calculations were performed that include explicit solute-solvent interactions when modeling the water solvent. In this case, vertical excitation energies for the 9MP(H 2 O) 3 complex also include bulk solvent effects. Basis set superposition errors were not considered as they are not expected to significantly influence the semi-quantitative results obtained in this "super-molecule" approach. Excited-state energies are presented at the TD-PBE0/IEFPCM/6-311++G(d,p) level of theory throughout the calculations presented in Table S2 and Figure S5.  Table S2. TD-DFT/PBE0/IEF-PCM/6-311++G(d,p)//B3LYP/6-31+G(d,p) vertical singlet and triplet excitation energies for 9-methylpurine (9PM) and the complex between 9methylpurine and three water molecules (9MP(H 2 O) 3 )) a ground state optimized at PBE0/6-31+G(d,p); b S 4 in the gas phase. S 3 has mostly n* character (5.18 eV, 0.0014) in the gas phase.

CASSCF and CASPT2
Calculations. The ground-state absorption spectra of N9H purine and 9-methylpurine were calculated at the equilibrium geometries optimized with the complete active space self-consistent field (CASSCF) method 46  Framed in blue the (12,9) active space used during the dynamics simulations and framed in red the orbitals involved in the S 1 (n*) and S 2 (*) excitations discussed throughout the manuscript. Minimum energy paths (MEPs) in Figures S11 to S16 were computed following the intrinsic reaction coordinate (IRC) algorithm to locate energy-accessible stationary points (Table S4 and Figure  In general, the CASSCF and CASPT2 methods find different qualitatively and/or quantitatively descriptions of the potential energy surfaces in the vicinity of the 1 (*) minimum and at the degeneracy regions between different electronic states of the same or different multiplicity. The inclusion of dynamical correlation typically increases the energy gap between the states involved in the surface crossings. In these cases, singlet/singlet and singlet/triplet crossings were re-optimized with the help of the gradient difference vector, DOI: 10.1021/ja512536c J. Am. Chem. Soc. 2015, S17 until an energy difference smaller than 0.2 eV between the two intersecting states was reached at MS-CASPT2 level of theory. For the correct description of the early stages of the deactivation mechanism, we optimized the 1 (*) minimum using CASPT2 gradients and the smaller active space (6,5).
The probability of intersystem crossing along the MEP was estimated by calculating spin-orbit coupling terms at critical points of the potential energy surfaces, namely at the Franck-Condon geometry, singlet minima and singlet/triplet minimum energy crossing points.
Vertical spectra, minimum energy paths, geometry optimizations, final energies, and spin-orbit couplings were obtained with the MOLCAS-76 program. 52 The optimization of singlet-singlet and singlet-triplet minimum energy crossing points was done with MOLPRO 2009. 53

Comparison of Simulated and Experimental Transient Absorption Spectra.
Experimental spectra shown in Figure 8 were extracted from the transient data for 9methylpurine in acetonitrile. The spectra were chosen at time delays where the signals reached a maximum intensity before beginning to decay. As mentioned above, the absorption spectrum of each individual excited-state was simulated by Gaussian broadening (50 nm width-at-half maximum) of the calculated transition oscillator strengths ( Figure S9). At each of the four chosen time delays, the spectrum was simulated by linear combination of multiple computed excited-state absorption spectra. The excited-state spectra included in each linear combination were selected based on the time delay and the predicted major relaxation pathway from the static and dynamics simulations. The contribution of each excited-state absorption to the linear combination was weighted until the simulated spectra best matched the experimental spectra as quantitatively as possible. The coefficient required for each excited-state absorption in the linear combination, which is reported in the following paragraph, was then used to support the decay mechanism of the purine free base and 9methylpurine.  Figure 8. The absorption spectra of the individual excited states were computed for 9-methylpurine in vacuum using MS-CASPT2. The spectrum of each excited state was scaled so that their linear combination best matched the experimental spectrum at the given time delay. The oscillator strengths shown here are those that have been scaled to produce the linear combinations (unscaled oscillator strengths given in Table S7). The coefficients used for each linear combination are given in the text below and in the caption of nm excitation, 7% to S 3 , 80% to S 2 , and 13% to S 1 ( Figure 6); if we assume that some decay from the S 2 (ππ*) FC to the (S 2 (ππ*)/S 1 (nπ*)) CI occurs faster than our time resolution (supported by the steep slope of the PES between these two regions, see Figure 7).
At a time delay of about 1.4 ps ( Figure S9b), the major decay is internal conversion to populate three major areas of the potential energy surfaces: the S 2 (ππ*) MIN , the vibrationallyhot (unrelaxed) S 1 state, S 1 (nπ*) UR , and the relaxed S 1 , S 1 (nπ*) R with an ultrafast lifetime of 0.15  0.05 ps. The populations estimated from the simulated spectrum at 1.4 ps is as follows.
The band at ~520 nm is assigned to the S 2 (ππ*) MIN absorption because of the prominent and simultaneous decay of the band at ~350 nm where the S 2 (ππ*) MIN has a very high relative oscillator strength (see Figures 2 and 8 and Table S7 for clarification of this fact).
The major decay mechanism with a tens of picoseconds lifetime (maximum signal at 42 ps, Figure S9c) is vibrational cooling. The population that was left at the (S 2 (ππ*)/S 1 (nπ*)) CI internally converts to populate S 1 (nπ*) UR (0.36). The population that was already at S 1 (nπ*) UR vibrationally cools to populate S 1 (nπ*) R (0.52). A small fraction of the population that was already at S 1 (nπ*) R intersystem crosses to populate the T 2 (ππ*) state Finally, the excited-state population fully reaches the T 1 (ππ*) state minimum within 2600 ps ( Figure S9d) through intersystem crossing from the S 1 (nπ*) R and subsequent conformational relaxation from T 2 (ππ*) area of the 3 ππ* potential energy surface (see Figure   7). Apparently, conformational relaxation in the 3 ππ* potential energy surface (i.e., T 2 (ππ*)  T 1 (ππ*)) occurs on an ultrafast time scale such that there is no further buildup of the T 2 (ππ*) population than the population already observed at 42 ps. This is consistent with the dynamics simulations where no appreciable population in the T 2 (ππ*) is observed ( Figure   S18, top panel).

Dynamical Calculations and Initial Conditions.
The absorption spectrum of Figure 6 has been calculated using a Wigner distribution encompassing 1000 geometries, as described in Refs. 54 and 55 with frequencies obtained using MP2/def2-svp. 56 The spectrum is composed as the sum of Gaussian functions assigned to all transitions of each of the 1000 geometries, computed at the SA-CASSCF (12,9)/def2-svp level of theory, averaged over 5 singlet states.
A sample of 94 initial conditions is used to run ab initio molecular dynamics during 1 ps. The distribution into the initial electronic states is selected according to the computed excitation energies and oscillator strengths. 57,58 In the same way as the experimental excitation window is centered at the maximum of the spectrum, the simulated excitation window was centered at the theoretical absorption maximum of 5.7 eV. The width of the theoretical excitation window was chosen as 0.5 eV, which lead to 12 trajectories starting in the S 1 , 75 starting in the S 2 , and 7 starting in the S 3 . Trajectories were terminated earlier than 1 ps if relaxation to the ground state or to the T 1 state occurred. The employed time step in the nuclear dynamics was 0.5 fs while the time step for integration of the time-dependent Schrödinger equation was 0.02 fs. The energies, gradients, non-adiabatic couplings, and spinorbit couplings were calculated on-the-fly using SA-CASSCF(12,9)/def2-svp, state-averaged over 4 singlet and 6 triplet states. The dynamical simulations were done with the program SHARC. 59,60 Singlet-singlet, triplet-triplet and singlet-triplet hopping geometries were used as starting points for optimizations of minimum energy crossings at SA-CASSCF(12,9)/6-31G* level of theory. These optimizations, as well as the on-the-fly quantum calculations have been carried out using the MOLPRO 2012 program package. 61 The populations plotted in Figure S18 are adiabatic energies (S 0, S 1 , S 2 and so on).
Since in the dynamics the planar symmetry of the chromophore is generally lost due to the general motion of the molecule, nπ* and ππ* states mix, e.g., at some geometries S 1 might be primarily nπ* and at other geometries ππ*. Hence, it is difficult for the dynamics to disentangle the populations in terms of these state characters. However, in the case of the purine, the S 1 state mostly correlates to nπ* and the S 2 state to the ππ* character. The triplet        Figure S16. CASSCF minimum energy paths from S 2 (ππ*)/T 2 (nπ*) and S 1 (nπ*)/T 1 (ππ*) intersystem crossings and T 2 (nπ*)/T 1 (ππ*) conical intersection following the indicated gradients.

Figure S1
initio molecular dynamics tautomer of the purine free base obtained from ab initio molecular dynamics. CI and ISC stand for conical intersection and intersystem crossing, respectively.

18.
initio molecular dynamics tautomer of the purine free base obtained from ab initio molecular dynamics. CI and ISC stand for conical intersection and intersystem crossing, respectively.
. Top panel: initio molecular dynamics tautomer of the purine free base obtained from ab initio molecular dynamics. CI and ISC stand for conical intersection and intersystem crossing, respectively.
Top panel: initio molecular dynamics tautomer of the purine free base obtained from ab initio molecular dynamics. CI and ISC stand for conical intersection and intersystem crossing, respectively.
Top panel: initio molecular dynamics tautomer of the purine free base obtained from ab initio molecular dynamics. CI and ISC stand for conical intersection and intersystem crossing, respectively.
Top panel: initio molecular dynamics tautomer of the purine free base obtained from ab initio molecular dynamics. CI and ISC stand for conical intersection and intersystem crossing, respectively.
Top panel: Semi initio molecular dynamics tautomer of the purine free base obtained from ab initio molecular dynamics. CI and ISC stand for conical intersection and intersystem crossing, respectively. Semi initio molecular dynamics simulations. tautomer of the purine free base obtained from ab initio molecular dynamics. CI and ISC stand for conical intersection and intersystem crossing, respectively.  Figure S1 dynamical simulations. degrees.  Figure S1 dynamical simulations. degrees.       Table S7. MS-CASPT2 excitation energies and oscillator strengths calculated at particular points along the potential energy surfaces given in Figure 7. These were used to simulate the experimental transient spectra as shown in Figures 8 and S9.