Probing the Charge Transfer Reaction Coordinate of 4

Jun 22, 2010 - ... and narrowing of the 1581 cm−1 quinoidal C═C stretch (ν8a) on the ∼7 ps time scale ... modes that are strongly anharmonicall...
0 downloads 0 Views 3MB Size
14646

J. Phys. Chem. B 2010, 114, 14646–14656

Probing the Charge Transfer Reaction Coordinate of 4-(Dimethylamino)benzonitrile with Femtosecond Stimulated Raman Spectroscopy† Justin M. Rhinehart, Randy D. Mehlenbacher, and David McCamant* Department of Chemistry, UniVersity of Rochester, Rochester, New York 14627-0216 ReceiVed: March 16, 2010; ReVised Manuscript ReceiVed: June 2, 2010

Femtosecond stimulated Raman spectroscopy (FSRS) and femtosecond transient absorption have been used to probe the photoinduced charge transfer (CT) dynamics of 4-(dimethylamino)benzonitrile in methanol and n-hexane. Through a combined analysis of temporal changes in the Raman modes and transient absorption kinetics, a more complete picture of the reaction coordinate of the intramolecular charge transfer process has been established. FSRS spectra of the phenyl CdC stretching mode (Wilson mode 8a) at 1607 cm-1, which shifts to 1581 cm-1 in the CT state, and transient absorption measurements ranging from 360 to 700 nm support internal conversion from the locally excited to the charge transfer state in 4-5 ps and then a subsequent vibrational relaxation within the CT state manifold on a 6-8 ps time scale. Dramatic shifting and narrowing of the 1581 cm-1 quinoidal CdC stretch (ν8a) on the ∼7 ps time scale indicates that the quinoidal distortion is an important probe of the CT reaction dynamics. The cause of the spectral shifts is determined by comparing the observed shifts in the vibrational spectrum to anharmonic couplings computed for the benzonitrile radical anion by density functional theory (DFT) and with quantitative theoretical models of the solvent induced vibrational peak shifts. The DFT calculations indicate that the 10 cm-1 downshift of the CdC stretch is most likely attributable to significant vibrational excitation in nontotally symmetric modes that are strongly anharmonically coupled to the CdC stretch. Introduction The dynamics of photoinduced charge transfer determine the efficiency of a number of critical molecular systems from photosynthesis to solar energy conversion to DNA repair.1-3 Because of its structural simplicity, 4-(dimethylamino)benzonitrile (DMABN, Figure 1) has been one of the most investigated charge transfer (CT) compounds historically. However, a complete description of its electronic and structural dynamics has eluded researchers for the last four decades.4-6 Its popularity is in some part due to its unique dual fluorescence in polar solvents. The dual fluorescence is attributed to the presence of two emissive electronic states, a higher-energy ππ* state localized on the benzene orbitals, termed the locally excited LE state, and a second state with an extremely red-shifted emission, termed the CT state. As the polarity of the solvent is decreased, the CT emission blue-shifts and decreases in intensity until, in nonpolar solvents, only the LE emission is observed.6 Much of the previous work has focused on the twist of the amine in the CT state, which may lead to a twisted intramolecular CT (TICT) state, in which the p-orbital of the amine is completely decoupled from the benzene π-system. The most widely debated alternative model suggests a planar intramolecular CT (PICT) state, in which a formal double bond exists between the amine nitrogen and the benzene carbon, inducing a quinoidal benzene ring. Most of this work has employed steady state and transient absorption and fluorescence to infer the structural dynamics in the excited state, despite the inherent insensitivity of electronic spectroscopy to the subtle details of molecular structure.6-10 In the last five years, breakthroughs in excited state quantum chemistry have allowed detailed examinations of DMABN’s excited state potential energy surface.11-15 A prevailing model †

Part of the “Michael R. Wasielewski Festschrift”. * Corresponding author. E-mail: [email protected].

Figure 1. Representative energy level diagram for the photoexcitation of DMABN in a polar solvent. Given are the excitation, λpump, and absorption maximum, λmax, wavelengths as well as the Raman pump and Raman probe wavelengths. The lower energy level of the CT state relative to the LE state is indicative of the use of a polar solvent; a higher CT state relative to the LE state is observed in nonpolar solvents.

has developed that incorporates aspects of both the TICT and PICT pictures. Time dependent density functional theory (TDDFT) calculations find that the stable minimum on the CT surface contains a 90° twist of the amine, as in the TICT model, and a quinoidal benzene, as in the PICT model.11 In addition, a decrease in the nitrile CN frequency was predicted, consistent with the PICT valence bond structure, and the C4-N stretch

10.1021/jp1023982  2010 American Chemical Society Published on Web 06/22/2010

Charge Transfer Reaction Coordinate of DMABN was also predicted to downshift, consistent with the lack of amine-benzene coupling in the TICT model. Interestingly, this work also predicted a relatively flat potential along the amine twist coordinate, a finding that was contradicted by later coupled cluster (CC2) calculations.12 The CC2 calculations also supported a twisted, quinoidal CT state structure but found that a global minimum in vacuum also requires pyramidalization of C4, adjacent to the amine. However, as stated in that work, it is likely that in polar solvents, the planar C4 (C2V) structure is still the most stable structure because the pyramidalization actually decreases the CT state dipole moment.12 In addition, each of these studies found that the optically active S2 (La) state had an extremely shallow or no minimum near the planar FranckCondon structure and that the LE (S1, Lb) minimum occurred at a planar, benzenoid structure, which was roughly isoenergetic with the CT (La) state in vacuo. Gomez et al. performed complete active space self-consistent field (CASSCF) calculations of DMABN in vacuo to determine the excited state minima on S2 and S1 and the adiabatic and conical intersection pathways that connect them.13 They observed a shallow PICT minimum on S2, a minimum on S1 corresponding to a planar benzenoid LE state, and a TICT minimum on S1, all at energies that roughly agree with the CC2 calculations.12 In addition, the CASSCF calculations revealed a rehybridized ICT (RICT) potential energy minimum containing a bent C-C-N cyano group, but the authors stated it was at too high an energy to participate in the dynamics. Importantly, this work also determined that the S2/S1 conical intersection occurs along a seam that is parallel to the amine torsional coordinate. The conical intersection degeneracy was lifted by distortion along both the quinoidal stretch and an asymmetric skeletal distortion of the benzene, suggesting that relaxation from the FC state through the conical intersection may excite normal modes contributing to those degeneracy lifting coordinates. The importance of the quinoidalization coordinate was further emphasized by Cogan et al. in their own CASSCF calculations.14 This model emphasizes the molecular orbital picture in which the benzonitrile group accepts the donated electron into one of the standard π*-orbitals of benzene, that is, the 3a or 3b orbitals. Occupation of these can lead to either an antiquinoidal or quinoidal benzene ring. Their calculations determined four minima on the S2/S1 surface: a benzenoid LE (Lb) state, which was nearly isoenergetic with an antiquinoidal twisted CT state (Lb); a quinoidal planar La/CT state; and a quinoidal twisted CT state, which was the global minimum on the excited state surface.14 On this surface, the LE (Lb) minimum lies quite close to the ground-state minimum, that is, the FC geometry, but an adiabatic path connects LE to the TICT structure via distortion along the quinoidal and twisting coordinates. As in the work of Gomez et al., a nonadiabatic relaxation path also connects the FC state in S2 to the TICT minimum.13,14 This work was also consistent with previous TDDFT and CASSCF calculations by Xu et al.16 An excellent review of the pros and cons of these various theoretical methods for DMABN was presented by Parusel et al.17 Relatively few theoretical calculations have attempted to calculate the excited state structures of DMABN in the presence of solvent. Minezawa used the reference interaction site model to conclude that the twist angle is the dominant relaxation mechanism between the LE and TICT states, since the solvent reorganization only slightly modifies the LE-TICT energy difference.18 Scalmani et al. performed TDDFT computations in the presence of a polarizable continuum.15 This work predicted that the La and Lb states were nearly degenerate at

J. Phys. Chem. B, Vol. 114, No. 45, 2010 14647 the Franck-Condon state, with the La state 0.04 eV above Lb. However, upon solvent reorganization, the La state drops below the Lb state by 0.09 eV and intramolecular geometry relaxation, excluding the twist, then brings the two states back into near degeneracy, with La 0.03 eV below Lb. Finally, the twisting of the amine dramatically drops the energy of the TICT state so that it lies 0.85 eV below the LE (Lb) minimum. Chiba and co-workers attempted to correct TDDFT’s typical problems with CT structures by incorporating long-range corrections.19 They also reached the conclusion that the La and Lb states were nearly degenerate at the FC state. However, this work also suggested that the dual fluorescence bands in polar solvents are both attributable to the CT (La) state, which contradicts the seminal fluorescence anisotropy experiments.20 Generally, these results support the conclusion that the La and Lb states are nearly degenerate and that the primary relaxation coordinates leading to the CT state are a quinoidal contraction of the benzene ring and a 90° twist of the amine. Over the course of the past decade, there have been many time-resolved vibrational spectroscopy experiments that have probed the excited state structures of DMABN.21-28 Early work by Chudoba et al. measured the 4-ps CT state formation time apparent in the time-dependent growth of the 2112 cm-1 CN stretching frequency.21 The frequency of this mode ruled out the bent nitrile structure of the RICT state, but no other features could be determined that would report on the structure of the CT state. Later, Okamoto et al. were able to observe the downshift of the ph-N stretch from 1372 cm-1 in S0 to 1272 cm-1 in a picosecond infrared (psIR) study of the CT state, supporting the TICT structure in which the π-conjugation across this bond is broken.22 The same downshift was observed by Kwok, Ma, and co-workers in the picosecond time-resolved Raman (psTR2) spectrum.25 In addition, Okamoto observed that the psIR spectrum of the pyrrole-benzonitrile CT state appears identical to the spectrum of DMABN’s CT state, supporting the conclusion that the donors are vibrationally and electronically decoupled from the common acceptor, the benzonitrile radical anion.29 This fact was later confirmed in the psTR2 spectra of a variety of methyl substituted DMABN analogs.27 However, each of these studies examined only the relaxed CT structure of DMABN with a single spectrum taken at a time point between 6 and 50 ps. Only a few infrared studies have observed the time-dependent changes in the vibrational spectrum as the CT state is formed.26-29 Okamoto, in probing the pyrrole-benzonitrile CT state, observed that the 1219 cm-1 ph-CN upshifts and narrows on the 10-20 ps time scale, whereas the 964 cm-1 phenyl distortion peak (Wilson mode 12) is constant over this same time range.29 This indicated that vibrational energy is deposited and remains isolated in specific vibrational modes during CT state formation. Kwok and Ma et al. have observed the LE f ICT reaction by fsIR of the nitrile stretch region and have observed the subsequent formation of a hydrogen bonded ICT (HICT) species in protic solvents.26 In methanol, no spectral shifts were observed as the ICT CN stretch at 2109 cm-1 grew in with a 5.5 ps time constant. The downshifted HICT CN stretch appeared at 2091 cm-1 and appeared with a 13 ps time constant. Their later work measured the ICT and HICT formation times in butanol, in which the CN stretch was observed to upshift 1 cm-1 in the first 3 ps and then downshift 15 cm-1 on the 30-500 ps time scale in response to the slow solvent reorganization.28 However, to our knowledge, no time-resolved Raman spectra have been reported that observe the structural evolution of DMABN as it relaxes from the FC state to the LE and ICT states.

14648

J. Phys. Chem. B, Vol. 114, No. 45, 2010

Rhinehart et al.

Recently, femtosecond stimulated Raman spectroscopy (FSRS) has emerged as a promising vibrational technique to probe ultrafast CT dynamics and the accompanying intramolecular structural reorganization the CT induces.30 This work was undertaken to observe the time-resolved evolution of DMABN’s vibrational spectrum as the CT state is formed. These spectra report on the reaction coordinate and its projection onto the various normal modes of the system and can thereby support or contradict theoretical descriptions of the CT mechanism. Experimental Section Transient absorption and femtosecond stimulated Raman spectra were carried out using the output of a regeneratively amplified Ti:sapphire laser (Spectra-Physics Spitfire) with an 800-nm, 100-fs fundamental and a repetition rate of 1 kHz. For the transient absorption, third harmonic generation of the Ti: sapphire output produced the 266-nm, 0.400-mW pump pulse. The white-light continuum probe was generated by focusing a small portion of the 800-nm Ti:sapphire fundamental into a CaF2 crystal, giving a usable spectral region ranging from 360 to 630 nm. After the sample, the probe was dispersed onto a photodiode array (Spectronic Devices) and then relayed to a PC controlled by LabView (National Instruments). Rotational diffusion effects were eliminated by orienting the pump and probe polarizations at the magic angle (54.7°). Transient absorption spectra are the result of averaging 40 scans with 200-ms exposure times with solvent subtraction. Time resolution was determined by optical Kerr effect (OKE) measurements performed between the 266-nm pump and continuum probe pulses within a 150-µm-thick quartz slide. Resulting signals were fit to wavelength-dependent Gaussian instrument response functions (IRF) that established the fullwidth at half-maximum to be 220 fs throughout the spectrum. The IRF was later convoluted with multiexponential molecular response functions in the kinetic fitting procedures. The probe chirp, also measured by OKE, produced a 0.050 ps delay between probe wavelengths of 406 and 550 nm. This wavelength dependent delay was incorporated into the fitting algorithm; all kinetic traces are displayed with the chirp subtracted for clarity. Possible multiphoton absorption processes were ruled out on the basis of a linear power dependence of the absorption signal monitored at 400 nm. The sample was illuminated in a wireguided open air liquid jet in which the solution flowed in a vertical “sheet” with a net thickness of ∼150-200 µm.31 4-(Dimethylamino) benzonitrile (DMABN, Aldrich) was used as received. Methanol (MeOH, Mallinckrodt Chemicals) and hexanes (95% n-hexane, Mallinckrodt Baker, Inc.) were spectroscopic grade and used as received. For the FSRS spectra, a 266-nm 100-fs excitation pulse was used, along with a Raman pump pulse created by spectral filtering a portion of the Ti:sapphire fundamental output.32 To create the 400-nm, 2-ps Raman pump, the spectrally filtered fundamental was frequency-doubled in a 3-mm BBO crystal. To increase the probe intensity in the Raman Stokes region relative to that of the transient absorption probe, the FSRS probe was generated by focusing a 400-nm second-harmonic of the Ti:Sapph fundamental into CaF2. The 266-nm pulse was used to initiate photochemistry (actinic pump) while subsequent to this excitation, the Raman pump and probe together obtained a Raman spectrum of the sample. The stimulated Raman spectrum is observed in the amplification of the probe beam at specific Stokes Raman frequencies relative to the Raman pump wavelength. The Raman pump pulse energy was 1.0 µJ at the sample and the 266-nm pulse energy was 1.1 µJ. The ground state and

Figure 2. (a) Transient absorption spectra of DMABN in methanol for pump-probe delay times ranging from -4 to 30 ps with excitation pump wavelength of 266 nm. (b) Kinetic traces (points) and fits (solid lines) observed at 410, 435, and 550 nm. The initial absorption spectrum is formed by ∼1 ps, after which the CT state signal at 410 nm grows in with an 8-ps time constant, and the LE signal at 550 decays with a 4-ps time constant.

positive time delay spectra were obtained by averaging 1200 scans with exposure times of 200 ms. Difference spectra shown are the result of subtracting the ground state and solvent from a scaled positive pump-probe time delay spectra. The scaling factor is chosen to normalize the solvent peaks so they are eliminated in the difference spectra. Hence, the difference spectra display the contributions from the excited state peaks and ground state bleach only. Optical density of the sample at the excitation wavelength (266 nm) was monitored throughout the experiment and varied between 0.88 and 1.08. The liquid jet sample apparatus was also used for the FSRS data, with the same net thickness as the transient absorption data. Time resolution for the FSRS experiments is the same as for the transient absorption experiments, 220 fs. The empirical spectral resolution was 19 cm-1, determined primarily by the spectrograph slit and Raman pump pulse bandwidth. Computations were performed in the Gausian03 software package, using density functional theory incorporating the B3LYP methodology and the 6-31G(d) basis set.33 Anharmonic vibrational analysis was performed using Barone’s automated methodology within Gaussian03.34,35 Results and Discussion The time-resolved transient absorption spectra of DMABN in methanol at room temperature (Figure 2a) consist of two main components: a broad absorption band ranging from 440 to 700 nm that initially appears over the first picosecond and an intense absorption band peaked at 400 nm that grows in over the first 10-30 ps. The broad absorption band is assigned to the LE state and decays on a 4-ps time scale. Its decay coincides with the growth of the intense band at 400 nm, which is assigned to

Charge Transfer Reaction Coordinate of DMABN the CT state absorption and grows in on an ∼8-ps time scale. Select kinetic traces and fits of the LE and CT state spectral evolution are shown in Figure 2b. As mentioned previously, OKE cross-correlation signals were recorded for each wavelength and the transient absorption kinetics were fit to a convolution of the cross-correlation and an exponential molecular response. Kinetic fits for the LE state decay were performed at 490 and 550 nm, giving 3.71 ( 0.14 (1σ) and 4.21 ( 0.17 ps time constants, respectively (Figure 2b, Table S1 of the Supporting Information). Investigations of the growth of the absorption band peaked at 400 nm resulted in time constants ranging from 8.1 to 8.7 ( 0.5 ps taken at wavelengths between 406 and 420 nm (Figure 2b, Table S1 of the Supporting Information). These absorption band assignments are in agreement with the results of DMABN transient absorption by Druzhinin and Gustavsson in acetonitrile.8,36 Transient absorption measurements of DMABN in n-hexane were also carried out for comparison (Figure S1 of the Supporting Information). It is known that nonpolar n-hexane prevents dual-fluorescence from occurring, since stabilization of the CT does not occur.4,5,7,20 Upon absorption of the 266-nm excitation pulse, the transient absorption spectra of DMABN in n-hexane show the same initial broad absorption feature assigned to the LE state previously mentioned. In comparison, no intense absorption band peaked around 400-nm corresponding to the CT state is observed. The lack of any significant evolution of the transient absorption spectrum of DMABN in n-hexane is in agreement with previous transient absorption studies and the lack of CT emission in this nonpolar solvent.6,24,37 In methanol, the difference in the rise time of the CT state absorption and the decay time of the LE state band of DMABN requires the introduction of an intermediate state between these two species, presumably an unequilibrated CT state species. The unequilibrated CT state can relax via either solvation dynamics (i.e., outer-sphere solvent reorganization) or vibrational relaxation within the newly formed CT state (i.e., inner-sphere intramolecular vibrational relaxation). Although transient absorption elucidates the presence of either one of these dynamic processes, distinguishing between the two is difficult, if not impossible, using just this method. The initial LE absorption features grow in over the course of the first picosecond due to a wavelength-dependent time resolution. We do not believe that the short time growth is indicative of a rapid electronic relaxation on this time scale. Other experiments (shown in Figure S2 of the Supporting Information) probing this time region with enhanced time resolution do not exhibit any exponential growth. Kwok et al. discussed 267-nm wavelength excitation of DMABN as initially populating the La state of the LE manifold where this population then experiences rapid relaxation, expected to be on the order of