Disentangling Intrinsic Ultrafast Excited-State Dynamics of Cytosine

ARTICLE pubs.acs.org/JPCA

Disentangling Intrinsic Ultrafast Excited-State Dynamics of Cytosine Tautomers Jr-Wei Ho, Hung-Chien Yen, Wei-Kuang Chou, Chih-Nan Weng, Li-Hao Cheng, Hui-Qi Shi, Szu-Hsueh Lai, and Po-Yuan Cheng* Department of Chemistry, National Tsing Hua University, Hsinchu, Taiwan, R. O. C. ABSTRACT: Gas-phase ultrafast excited-state dynamics of cytosine, 1-methylcytosine, and 5-fluorocytosine were investigated in molecular beams using femtosecond pump
probe photoionization spectroscopy to identify the intrinsic dynamics of the major cytosine tautomers. The results indicate that, upon photoexcitation in the first absorption band, the cytosine enol tautomer exhibits a significantly longer excited-state lifetime than its keto and imino counterparts. The initially excited states of the cytosine keto and imino tautomers decay with sub-picosecond dynamics for excitation wavelengths shorter than 300 nm, whereas that of the cytosine enol tautomer decays with time constants ranging from 3 to 45 ps for excitation between 260 and 285 nm.

1. INTRODUCTION The remarkable photostability of deoxyribonucleic acid (DNA) that protects life on Earth from the solar ultraviolet (UV) radiation is often attributed to the ability of the four constituent nucleobases to efficiently dissipate dangerous electronic energy through ultrafast excited-state relaxation processes.1,2 Although the excited-state dynamics may be affected by the sugar
phosphate backbone and base paring interaction, knowledge of the photochemistry of isolated nucleobase is essential to our understanding of the underlying mechanism in the photostability of DNA. Femtosecond time-resolved spectroscopic studies of nucleobases in aqueous solutions have indeed revealed the extremely short-lived nature of their excited states.3
8 Gas-phase investigations of the nucleobases have also gained popular attention in recent years with a view to elucidating the intrinsic relaxation dynamics.4,9
16 In the case of cytosine (Cy), several groups have9
12 explored the ultrafast excited-state dynamics of Cy in molecular beams. However, controversial results exist in the literature. Kang et al.9 excited cytosine at 267 nm and probed the excited-state dynamics via multiphoton ionization (MPI) at 800 nm. They interpreted the observed transients with an initial spike arising from nonresonant MPI at the zero delay time and a single exponential decay of 3.2 ps.9 Canuel et al.10 carried out a similar experiment at the same excitation wavelengths (267 nm pump; 400 nm probe) and attributed the observed transients to a combination of two independent exponential decays of 0.16 and 1.86 ps.10 Ullrich et al.11 have measured femtosecond timeresolved photoelectron spectra of Cy excited at 250 nm and resolved three temporal components, an extremely fast decay occurring in less than 50 fs, a second one in 0.82 ps, and a third slower one in 3.2 ps.11 As have been pointed out elsewhere,11,12 one ambiguity in these experiments is the coexistence of two or more tautomeric forms of Cy in the gas phase. There are totally six possible r 2011 American Chemical Society

tautomeric structures for Cy. Among them, the amino-keto, aminoenol, and imino-keto forms (see Figure 1 for the chemical structures) have been identified as the three most stable tautomers by previous experimental17
22 and theoretical studies.20,21,23
26 Hereafter, these tautomers are referred to as the keto, enol, and imino, respectively, for the sake of conciseness. Early experiments have already revealed the simultaneous presence of these tautomers in inert environments. For example, a molecular-beam microwave spectroscopic study estimated a population ratio of 1:1:0.25 for keto:enol:imino at about 570 K;19 and a matrixisolation IR spectroscopic study reported a population ratio of 0.5:1:0.05 at an evaporation temperature of about 490 K.18 Very recently, Bazso et al.21 revisited the problem by fitting a matrixisolation IR spectrum globally with computed IR intensities and determined a mole fraction of 0.22:0.70:0.08 for keto:enol:imino at an evaporation temperature of 450 K.21 A successful UV absorption spectrum simulation was achieved using these “more reliable” mole fractions.21 Several theoretical studies have also predicted that the enol form is the most stable tautomer in the gas phase and that all three forms exist in non-negligible abundances at evaporation temperatures used in typical molecular-beam experiments.11,20,21,23
26 These experimental evidence as well as the theoretical results certainly bring questions to the assignment of the above-mentioned time-resolved experiments. Nir et al.27,28 have employed a “methyl-blocking approach” to identify Cy tautomers by comparing UV spectra of Cy and 1-methylcytosine (1mCy) using nanosecond-laser resonance-enhanced multiphoton ionization (REMPI) spectroscopy. Owing to the methyl substitution at the N1 position (see the labeling and chemical structure in Figure 1), enol-keto tautomerization is infeasible; therefore, 1mCy exists predominantly in the keto form. Nir et al.27,28 observed spectral features in two distinctive Received: June 15, 2011 Published: June 23, 2011 8406

dx.doi.org/10.1021/jp205603w | J. Phys. Chem. A 2011, 115, 8406–8418

The Journal of Physical Chemistry A

Figure 1. Chemical structures of the major tautomers of cytosine (Cy), 1-methyl cytosine (1mCy), and 5-fluorocytosine (5FCy).

regions around 310 and 270 nm. Both Cy and 1mCy were detected in the 310 nm region, while only Cy could be detected in the 270 nm region. Accordingly, they assigned the spectrum observed around 310 nm to the keto form and that observed around 270 nm to the enol form. The band origin for keto-Cy has been determined at 31 826 cm
1 (∼314 nm),27 whereas that for enol-Cy can only be roughly estimated to be around 36 000 cm
1 (∼278 nm).27,28 Although these assignments have been challenged recently,21,29 they are in good agreement with the adiabatic excitation energies of the first bright 1ππ* state predicted by Tomic et al.30 at the DFT/MRCI level of theory: 4.06 eV (305 nm) for keto, 4.50 eV (276 nm) for enol, and 4.46 eV (278 nm) for imino.30 Kosma et al.12 recently measured femtosecond pump
probe ionization transients of Cy excited at five different wavelengths in the range from 260 to 290 nm. On the basis of the spectral carrier identifications proposed by Nir et al.,27 they assigned the dynamics observed with 290 and 280 nm excitation to the keto tautomer.12 The transients were decomposed into three components with lifetimes of τ1 < 0.1 ps, τ2 = 1.1
1.2 ps, and τ3 > 150 ps (290 nm) or τ3 = 55 ps (280 nm).12 The slow decay (τ3) was assigned to a low-lying imino 1nπ* state populated by excited-state keto-to-imino tautomerization, but the possibility due to triplet state formation was also considered.12 The data measured with 270 and 267 nm excitation were interpreted as mixtures of tautomers, and three lifetimes were again resolved: τ1 = 0.21 ps, τ2 = 2.2 ps, and τ3 = 19 ps (270 nm), or τ3 = “long” (267 nm).12 The slower component (τ3) was assigned to the corresponding state in the keto tautomer. For 260 nm excitation, only two components (τ1 = 0.12 ps and τ2 = 3.8 ps) were resolved and assigned to a dominant contribution from the enol, or possibly the imino, tautomer.12 Although these assignments are in line with the spectral carrier identifications based on nanosecond REMPI experiments,27 it is not clear why the keto tautomer does not contribute to the transient at 260 nm excitation, given that the vertical excitation energy of Cy keto tautomer has been calculated to be 4.4
5.4 eV (282
230 nm) by many authors at various levels of theory.21,29
46 For example, Tomic et al.30 have predicted the vertical excitation energies of Cy to be 4.83 eV (256 nm) for keto, 5.14 eV (241 nm) for enol, and 5.26 eV (236 nm) for imino.30 Ab initio spectra simulation at

ARTICLE

various levels of theory have also shown that the Cy keto form absorbs strongly even at wavelengths shorter than 260 nm.21,29,38,39 Recently, Bazso et al.21 reported a spectrum simulation with the linear vibronic coupling model using vertical excitation energies and transition moments obtained from CC3-LR/aug-cc-pVDZ calculations. Combining with the mole fractions of the three Cy tautomers, they successfully reproduced the observed matrixisolation UV absorption spectrum.21 The simulation revealed significant spectral overlap of the three Cy tautomers in the first UV absorption band; in particular, the keto and enol forms seem to contribute equally at the red edge of the band.21 Thus, separation of Cy tautomer excited-state dynamics by spectral selection alone should be taken with caution. In this work, we employ a different approach to discern the excited-state dynamics of Cy tautomers in the gas phase. The approach is similar to that employed by Nir et al.27 but with femtosecond time resolution. Briefly, we study excited-state dynamics of Cy and its two derivatives, 1mCy and 5-fluorocytosine (5FCy), using femtosecond pump
probe photoionization mass spectrometry. 1mCy and 5FCy bear substitutions that greatly change the population ratio of tautomers, making it possible to discern the dynamics of different forms. For 1mCy, the H atom at the N1 position in Cy is replaced by a methyl group, making enol
keto tautomerization infeasible, and therefore, the molecule exists predominantly in the keto form. For 5FCy, the fluoro substitution at the C5 position results in a strong stabilization for the enol form, making it the dominant 5FCy tautomer in the vapor phase. By comparing the transients observed for Cy, 1mCy, and 5FCy in the molecular beam, we discerned the scrambled excited-state dynamics of different Cy tautomers.

2. EXPERIMENTAL SECTION The femtosecond laser system employed in this work consists of a self-mode-locked Ti:sapphire laser (Spectra Physics, Tsunami), a 1 kHz chirped-pulse regenerative amplifier (CPA, Spectra Physics, Spitfire), and two optical parametric amplifiers (OPA, Light Conversion, TOPAS and TOPAS white). Pump pulses between 260 and 310 nm were obtained either by the third-harmonic generation of the CPA output or by suitable nonlinear conversion of the OPA outputs. The probe wavelengths used for Cy experiments were as follows: for 260 nm e λpump < 280 nm, the corresponding fundamental of the third harmonic of the CPA output was used as the probe, i.e., λprobe = 3λpump; for 280 nm e λpump e 290 nm, λprobe = 800 nm; and for λpump = 300 nm, λprobe = 400 nm. For 1mCy and 5FCy experiments, the probe wavelength was kept at 800 nm. The probe beam was directed through a retroreflector mounted on a computer-controlled translation stage to facilitate the optical delay. The experiments were carried out in a two-chamber differentially pumped molecular beam apparatus equipped with a timeof-flight mass spectrometer (TOF-MS). Pure He gas at ∼300 Torr was allowed to flow over sample crystal heated to ∼190
210 °C in an oven and was then expanded through a 100 μm diameter pinhole to produce a continuous jet in the first chamber. The nozzle was heated to ∼230 °C to avoid clogging. No Cy dimer was observed under this condition. The jet was skimmed and then intersected by the femtosecond laser pulses in the extraction region of the TOF-MS. The pump and probe beams were collinearly focused through a f = 500 mm lens. The mutual polarizations of the pump and probe beams were set at 8407

dx.doi.org/10.1021/jp205603w |J. Phys. Chem. A 2011, 115, 8406–8418

The Journal of Physical Chemistry A

ARTICLE

the magic angle (54.7°) to minimize the rotational coherence effect. The laser irradiance was adjusted by placing variable neutral density filters in the pump and probe beam paths. The pulse energies used here were typically 1
0.5 μJ/pulse for the pump and 80
175 μJ/pulse for the probe. The pump-laser irradiance was kept at the lowest level possible so as not to produce appreciable background parent ion signal that may give rise to transient ion depletion.47,48 For this reason, the pump laser was slightly defocused in most cases. Femtosecond massselected transients were obtained by monitoring the ion intensity with a boxcar integrator (Stanford Research SR250) while the pump vs probe delay time was scanned. The fwhm of the effective instrumental response function (IRF) ranged between 170 and 210 fs, depending on the pump and probe wavelengths.

3. RESULTS AND ANALYSES 3.1. Cytosine. Figure 2 shows the Cy transients obtained at several pump wavelengths (λpump) between 260 and 300 nm. These transients were obtained with the lowest pump-pulse energy (e1 μJ/pulse) possible under a relatively soft focusing (f = 500 mm) condition to ensure one-photon excitation. As shown in Figure 3A,B, variations in pulse energies of the pump by a factor of 8 or the probe by a factor of 4 do not change the shape of the transients, except for the spike-like feature near the time zero. Pump-laser irradiance dependence measurements, shown in Figure 3C, indicate that the decay signals in these transients are due to one-photon excitation of Cy from the pump pulse. The transients measured with λpump e 285 nm can be well described by a combination of an “initial spike” located at the zero delay time and the convolution of the instrument response function (IRF) with a biexponential decay function. The initial spike resembles the IRF and exhibits different laser irradiance dependences from signal observed at later delay times, as shown in Figure 3A,B. It can be observed even at pump wavelengths where resonance absorption is absent, e.g., 385 nm. (Figure 3D). We believe that this initial spike is mainly due to the enhancement of nonresonant multiphoton ionization occurring at time zero when pump and probe pulses are temporally overlapped;49
51 therefore, it is mostly irrelevant to the excited-state dynamics discussed below. However, despite the efforts taken to reveal the different nature of the initial spike from the delayed transient signal, we cannot completely rule out the possibility of an unresolved small contribution, hidden in the spike, due to an initial step much faster than our IRF. The excitation-wavelength dependence of the transient is noteworthy. The decay time of the faster component varies marginally, whereas that of the slower component increases dramatically with increasing excitation wavelength, reaching about 45 ps at 285 nm. Figure 4 summarizes the excitation-energy dependence of the decay rates derived from Cy transients measured at 27 excitation wavelengths, including those not shown in Figure 2. The amplitude ratio of the slow to the fast component (A2/A1), also shown in Figure 4, is rather insensitive to the excitation wavelength between 260 and 276 nm. However, when the excitation wavelength is tuned further to the red, the A2/A1 ratio gradually decreases. This behavior is clearly seen in transients measured with λpump = 274 and 280 nm; i.e., somewhere between these two wavelengths the slower component becomes smaller than the faster one. For λpump > 280 nm the slower component continues to diminish and becomes undetectable for λpump g 290 nm where the transients appear to

Figure 2. Femtosecond pump
probe photoionization transients of Cy excited and probed at several different wavelengths. The pump and probe wavelengths are denoted as λpump/λprobe in each figure. The black solid lines are the best fit of the data point (red open circles) to the model described in the text, and the colored lines are the decomposed components (pink = initial spike; green = fast decay; and blue = slow decay). The solid circles in (F) are the magnification (10) of the slow-decay part of the data. Note the difference in time scale in each figure. The insets in each figure show the zoom-in view of the transients in a shorter time scale. The pump and probe-pulse energies used for collecting these data were about 0.5 μJ/pulse and 80 μJ/pulse, respectively.

exhibit a single fast decay (τ1) of about 1 ps in addition to the initial spike, distinctly different from those observed at shorter excitation wavelengths. Because similar experiments on Cy have been reported previously,9
12 it is necessary to compare the present results with those works. Our transient measured with λpump = 266 nm is qualitatively comparable to those reported by other groups using similar excitation wavelengths (266
267 nm).9,10 The transients reported by Kosma et al.12 at five excitation wavelengths share some similarities as well as discrepancies with the present results. The transients observed by both groups exhibit a similar general trend; i.e., the shorter the excitation wavelength, the faster the decay. However, the slower component (τ2) observed in our experiments seems to be always slightly shorter compared to 8408

dx.doi.org/10.1021/jp205603w |J. Phys. Chem. A 2011, 115, 8406–8418

The Journal of Physical Chemistry A

Figure 3. (A) Pump-pulse-energy dependence of the Cy transients measured with the probe-pulse energy kept constant at ∼80 μJ/pulse. (B) Probe-pulse-energy dependence of the Cy transients measured with the pump-pulse energy kept constant at ∼1 μJ/pulse. Note that, for easy comparison, the transients in (A) and (B) are scaled to the same intensity at ∼0.5 ps where the initial spike is negligible. (C) Pumppulse-energy dependence of the Cy transient signal at pump
probe delay time of ∼0.5 ps. The solid line is the best linear fit to the data with a slope close to unity. (D) (Blue trace) Cy transient obtained with λpump = 385 nm where no appreciable resonance absorption is expected. (Red trace) Cy transient measured with λpump = 266 nm for comparison.

Figure 4. (Lower panel) excitation-energy dependence of the decay rate constants (blue solid circles, 1/τ1; red solid circles, 1/τ2) obtained from fitting the Cy transients. The black solid curve is the result of RRKM calculations (for enol-Cy) with a barrier height of 1250 cm
1 (∼0.15 eV) and the band origin located at 290 nm (see text). (Upper panel) excitation-energy dependence of the amplitude ratio of slow to fast components (green solid circles, A2/A1). Note that the excitation energy refers to the excitation photon energy (Ehν), not Eint.

what Kosma et al. observed at the same wavelength. This discrepancy may arise from the difference in the molecular beam conditions; we used a continuous jet with soft expansion, whereas Kosma et al. employed a pulsed nozzle with harder expansion. Because the decay rate is very sensitive to the excitation energy as revealed here, it is plausible that the warmer beam temperature in

ARTICLE

Figure 5. (A) Cy transient measured with a much higher probe (798 nm) pulse energy of 160 μJ/pulse. The pump (266 nm) pulse energy was about 1 μJ/pulse. The inset shows a vertically expanded view of the same transient to reveal the very small long-lived component. The fitting results are τ1 = 0.4 ps and τ2 = 3.9 ps. (B) Transient obtained by collecting all fragment ion signals between m/z = 66 and 69 with λpump = 266 nm. The fitting results are τ1 = 0.6 ps and τ2 = 3.6 ps. (C) Transient obtained by collecting all fragment ion signals between m/z = 40 and 44 with λpump = 266 nm. The fitting results are τ1 = 0.7 ps and τ2 = 3.5 ps. (D) Transient obtained by collecting all fragment ion signals between m/z = 40 and 44 with λpump = 276 nm. The fitting results are τ1 = 1.7 ps and τ2 = 7.5 ps. The insets in (C) and (D) display the corresponding long-scale transients under the same conditions. The pulse energies used in all fragment-ion transients were ∼1 μJ/pulse for the pump and ∼100 μJ/pulse for the probe.

our experiment causes more hot-band excitation, which results in a higher total excitation energy at the same wavelength and, thus, slightly shorter decay times. Another discrepancy is that the transients reported by Kosma et al. contain a long-lived component with lifetimes varied from 19 ps up to g150 ps, depending on the excitation wavelengths.12 This slow component was assigned to a low-lying imino 1nπ* state populated by excited-state keto-to-imino tautomerization.12 Although we did not include it in the fits shown in Figure 2, we have also noted an extremely small long-lived component in our transients at all excitation wavelengths. This long-lived component is noticeable, i.e., greater than the noise level, only when the probe-pulse energy is raised to a much higher level, as shown in Figure 5A, indicating a higher probe-laser irradiance dependence than the preceding faster components. With the probe-pulse energy (∼80 μJ/pulse) used to obtain the transients shown in Figure 2, this long-lived component is negligible. The signal of this long-lived component stays nearly constant within the range of our delay line (∼1 ns), suggesting a decay time of longer than 10 ns. Similar long-lived components were also observed for 1mCy and 5FCy, as will be described below. Because the transients reported by Kosma et al. were obtained by summing the parent and major fragment ion signals altogether,12 we also recorded several fragment ions transients (see Figure 5B
D) under similar conditions with 266 and 276 nm excitation for comparison. A typical mass spectrum recorded 8409

dx.doi.org/10.1021/jp205603w |J. Phys. Chem. A 2011, 115, 8406–8418

The Journal of Physical Chemistry A

Figure 6. Pump
probe time-resolved TOF-MS spectra measured at ∼0.5 ps delay time for Cy with two different probe-pulse energies: (A) 100 μJ/pulse (lower panel); (B) 20 μJ/pulse (upper panel). The pumppulse energy was