Fei Ma and
Arkady Yartsev*
Chemical Physics, Department of Chemistry, Lund University, Box 124, SE-21000, Lund, Sweden. E-mail: Arkady.Yartsev@chemphys.lu.se
First published on 20th April 2016
Photoisomerization of 1,1′-diethyl-2,2′-carbocyanine iodide (pinacyanol) in alcohols was investigated by means of femtosecond time-resolved absorption spectroscopy. Only one out of two possible types of photoisomer was found. Viscosity-dependent kinetics indicate that the excited state dynamics consist of two parts, associated with a barrier-crossing observed in the blue side of stimulated emission (SE) and the barrierless isomerization observed in the red side of SE. These two processes occur sequentially. The blue barrier SE part is characterized by a decay lifetime (τ1, 5.2 ps in methanol) that is constant over the observed spectral region. Dependence of τ1 on solvent viscosity (η) is linear and follows Kramer’s law for barrier-crossing reactions. The red barrierless part is faster and is characterized by a rise in time constants, which increases with the wavelength (τ2λ, from 0.2 to 0.6 ps for 720–1000 nm range in methanol). τ2λ depends on η as a fractional power function and follows BFO theory for barrierless photoisomerization. The overall photoisomerization rate is limited by the barrier-crossing rate, τ1−1. A two-state two-mode model with a saddle-type of potential energy surface (PES) was used to explain the observed spectral dynamics. Along the stretching mode it is barrierless while along the torsion mode there is a very small barrier. As a result, the PES becomes ridge-like after displacement along the stretching coordinate. Comparing pinacyanol and 1,1′-diethyl-2,2′-cyanine iodide (1122C) with similar excited state lifetimes (5.2 vs. 4.9 ps) but quite different photoisomer quantum yields (QY, 3.6% vs. 26%, respectively), we conclude that excited state PES properties play the most important role in determining when the torsion mode couples with the stretching mode, and as a result the photoisomer QY.
Torsion motion is the key step for photoisomerization. Depending on the intrinsic energy barrier along the torsion coordinate, photoisomerization is divided into two categories: barrierless and barrier types. For the barrierless type (Chart 1A), the topology of the excited state PES is often ridge-like.7–12 The torsion and stretching modes are usually both activated upon excitation. Thus a molecule starts to elongate and twist simultaneously, corresponding to a downhill motion of the excited state population. The barrierless photoisomerization can be described by the long-standing theory proposed by Bagchi, Fleming and Oxtoby (BFO).9,13 For the barrier type, the topology of the excited state PES is valley-like (Chart 1C).14,15 Photoexcitation induces stretching motion from the Franck–Condon (FC) point to a local PES minimum where a planar configuration remains, defined as Minplanar. The starting point of the torsion mode corresponds to Minplanar. It is usually the rate-limiting step of the overall photoisomerization and its rate is determined by the value of the intrinsic activation energy. After crossing the barrier, torsion motion drives the molecule towards a global PES minimum in the vicinity of the CI sink with a ∼90° twisted configuration (Mintwist). This step is usually fast and is thus difficult to resolve. An intermediate barrier situation has also been predicted (Chart 1B). When the barrier is small enough, the topology of the excited state PES is valley–ridge-like.6 The torsion mode may start before Minplanar is reached. Thus this type of photoisomerization has both barrier and barrierless properties. Furthermore, the photoisomerization process and thus its outcomes can be drastically affected by PES topology changes under an influence of the environment. For example, all-trans-retinal in the protein matrix in bacteriorhodopsin undergoes one predominant type of isomerization18,19 whereas in solution it can form several photoisomers with similar QYs.16,17
Chart 1 Two-mode excited state PES with ridge (A), ridge valley (B), and valley (C), structures (top, reprinted (adapted) with permission from ref. 6, Copyright (2000) American Chemical Society), and the corresponding SE shapes (bottom) obtained by mathematical simulations. |
Cyanines have been widely studied in relation to photoisomerization. The excited state energy barrier for isomerization increases with the number of conjugated CC bonds. If the number is larger than 1 then more than one form of photoisomer may be formed. A variety of theoretical6,20,21 and experimental7–12,14,22–27 studies have been carried out to understand isomerization dynamics and even to control it. 1,1′-Diethyl-2,2′-cyanine iodide (1122C, molecular formula in Fig. 1) has only one CC bond and undergoes fast, barrierless photoisomerization.10,28–32 The rate is determined by the downhill evolution of the excited state population and its Brownian motion around the CI sink.13 Adaptive feedback control was successfully used to control the yield of the photoisomer by changing the distribution of momentum components in the initial wavepacket.12,28
In this work, 1,1′-diethyl-2,2′-carbocyanine iodide (pinacyanol, molecular formula in Fig. 1) containing two CC bonds between quinoline rings was investigated. Pinacyanol has been used as a sensitizer in photography,33 as a vital tool in staining leukocytes34 and as an indicator for solvent polarity.35 Besides, in aqueous solutions this molecule undergoes hypsochromic self-aggregation.36 NMR studies and quantum chemistry calculations established that at room temperature, the most stable configuration of pinacyanol is the trans configuration with a 23.4° distorted angle between the pairs of quinoline rings due to steric hindrance,37 which is smaller than the distorted angle of 1122C, of 46°.38 The difference in the excited state FC configuration between pinacyanol and 1122C is associated with the different PES shapes. Early studies have addressed the dynamics of pinacyanol, and it was suggested that there is nearly no barrier present during isomerization.22,39–42 However, this conclusion could be influenced by the relatively low time resolution used. Therefore, deeper investigation is needed to identify the nature of photoisomerization in this molecule. As two isomer configurations may be formed, an additional question is raised concerning which photoisomer (or both) is formed. To resolve these issues, fluorescence emission and time-resolved absorption spectroscopy were employed. We have measured the fluorescence emission spectral shape, wavelength-dependent stimulated emission (SE) dynamics and photoisomer QY of pinacyanol and compared these findings with those reported for barrierless 1122C10,31 and 1,1′-diethyl-2,2′-dicarbocyanine iodide (DDI, containing three CC bonds and is of the barrier type,43 molecular formula in Fig. 1). Since the barrierless and barrier reactions are supposed to have different viscosity dependencies, we have studied the solvent variation effects on isomerization dynamics. As a result, a clear picture of the excited state PES properties and photoisomerization dynamics of pinacyanol was obtained.
In the progression 1122C–pinacyanol–DDI, the fluorescence emission spectra exhibit dramatic changes (Fig. 1A). Fluorescence of DDI (pink line) has a narrow bandwidth and a small Stokes shift of 441 cm−1, representing an approximate mirror image of the absorption spectrum. The small bandwidth and structured emission spectrum points to a bound S1 PES and the mirror image emission indicate that only minor geometrical changes occur between the absorptive and emissive states. This is consistent with the observation that a barrier is present in DDI photoisomerization.42 On the contrary, 1122C (green line) emission has a much larger Stokes shift of 1421 cm−1, and is much broader and featureless. All these features indicate a drastic change between the absorptive and emissive state geometries, in very nice agreement with the nature of the ultrafast isomerization process which was proven to be barrierless.10 The shape of the pinacyanol (black line) emission spectrum falls in between the 1122C and DDI shapes. Its Stokes shift is 556 cm−1 and it possesses to some extent a symmetry with the absorption spectrum, yet the emission spectrum is clearly broader than the absorption. Moreover, a pronounced tail appears in the fluorescence in the low frequency region. These features suggest that more than one form of deactivation (and isomerization) pathway exists in pinacyanol. Therefore, an intuitive method was employed to decompose the emission spectrum into two parts (Fig. 1B). The frequency value at the intersection point of the absorption and emission spectra is set as a symmetry point and then the true mirror image of absorption was produced, which is then scaled and used as part 1 (barrier path, orange line) of the emission spectrum decomposition. Subtraction of part 1 from the emission spectrum results in part 2 (blue line). In Fig. 1B the scaling value is set to 0.5. The obtained part 2 is broad and structureless, which therefore can be interpreted as an indication of a barrierless process. It should be noted that this method is just an attempt to separate two prominent characteristics and may not represent the real emission spectra, associated with different processes. There are two possible ways to account for the two-part emission of pinacyanol. First, the molecule undergoes two different isomerization pathways (one pathway with a barrier and the other barrierless), as a dicarbocyanine molecule potentially can form two isomer configurations. Second, the topology of the S1 PES changes gradually from a bound shape to an unbound shape during excited state population evolution toward the CI sink. Consequently the emission spectrum changes from a mirror shape to a structureless shape. The second possibility is consistent with the CASSCF calculation results of C5H6NH2+, a simplified model of pinacyanol, showing that the surface topology in the FC region is initially valley-like (corresponding to barrier evolution) and then becomes ridge-like (corresponding to barrierless evolution) after partial displacement along the stretching coordinate.6 To clarify the mechanism, investigations of the isomerization dynamics are needed.
Transient absorption spectra of pinacyanol in methanol after 610 nm excitation are shown in Fig. 2. Initially, excited state absorption (ESA), ground state bleaching (GSB) and stimulated emission (SE) dominate the spectral regions at 470–530 nm, 530–640 nm and 640–1000 nm, respectively. After a few picoseconds, absorption of a photoisomer appears at 610–650 nm. In detail, upon excitation three sharp negative (560, 608 and 665 nm) and one positive (507 nm) signals appear simultaneously. Then these peaks become broader and smooth out at 0.2 ps. Meanwhile, the SE peak red-shifts from 663 to 668 nm. This early time evolution of the differential spectra resembles the differential dynamics spectra reported in ref. 10 and is assigned to the hole burning-like phenomena. Yet it should be noted that some contamination of the measured spectra due to the cross-phase modulation artifact during 0–0.2 ps may be present. After 0.8 ps, the ESA, SE and GSB decay with a ∼5.2 ps time constant, remaining nearly unchanged in shape.
An apparent ESA signal appears only below 570 nm, which is in contrast to the case of 1122C where transient ESA was observed from 400 to 1000 nm.10 That is because SE dominates the long wavelength region in pinacyanol at all times before the excited state population decays to zero. ESA is not very sensitive to the shape of the excited state PES and the population distribution on PES and thus directly reflects the dynamics of the overall S1 population. Although GSB contributes a little in the recorded wavelength region (470–530 nm), its contribution does not exceed 3% at 470 nm, estimated by comparing the relative amplitudes of the steady-state and transient spectra at 470 and 610 nm. The kinetic trace at 470 nm rises within the instrumental response time and decays to zero with a time constant of 5.0 ps (Fig. 3A). This excited state lifetime is in agreement with the fluorescence lifetime reported previously.39 GSB has an additional slow recovery component (≥1 ns), which is attributed to the back-conversion of the photoisomer. Evolution associated decay spectra are shown in Fig. S1.†
The contribution of the rise component Rrise (Arise/(Adecay + Arise)) also increases with wavelength (Fig. 3B). It implies a red-shift of SE during the first ps delay, which is a characteristic feature of a barrierless process. For comparison, in the case of the clear barrierless photoisomerization of 1122C,10 both the rise and decay times of SE increase with increasing wavelength. Thus, the S1 PES of pinacyanol seems to be composed of two sequential parts, instead of a single barrier or barrierless process.
A two-state two-mode model (Chart 1B) inspired by the calculated PES for C5H6NH2+6 and 1,1′-diethyl-2,2′-pyridocyanine (1122P)32 can explain these SE features of pinacyanol: (1) upon excitation, an excited state wavepacket on S1 PES is created. As dissipation processes are fast and efficient enough, this wavepacket quickly loses its phase memory and turns into an unphased excited state population. In the vicinity of the FC point, there is a pronounced barrier along the torsion coordinate and the topological structure of the S1 state PES is valley-like. In this region the excited state population evolves only along the stretching coordinate, with the CC bond of pinacyanol being elongated while the trans configuration remains. Thus the dihedral angle in the emissive state configuration does not change much compared to the FC configuration. Radiative deactivation occurs mainly from the bottom of this bound state and gives close to mirror image emission. Torsion motion also occurs from the same configuration and involves barrier-crossing. This barrier-crossing process is the rate-limiting step of the excited state isomerization and its rate is determined by the barrier height. As the SE lifetime of pinacyanol (5.2 ps) is close to that of the barrierless 1122C (4.9 ps),10 we assume that the barrier energy is very small; (2) after the initial displacement along the stretching coordinate, the pinacyanol molecule undergoes torsion motion. The PES becomes ridge-like after crossing this valley–ridge inflection point and the subsequently excited state population evolves downhill toward Mintwist. Behind the barrier, in the barrierless region, an increase of the torsion angle corresponds to a longer SE wavelength. Furthermore, a larger torsion angle needs a longer time to be achieved. This process is responsible for the broad red tail of the SE and the increased τrise with observation wavelength; and (3) when the excited state population arrives at Mintwist a quasi-equilibrated excited state is formed at the dihedral angle of ∼90°. In Fig. 3B one can see that in the red-most SE of our experiment (1000 nm) the rise time constant has not reached yet a wavelength-independent range. This implies that 1000 nm does not correspond yet to the quasi-equilibrated excited state in the vicinity of Mintwist but could be not very far from it, as τrise appears to approach saturation at this wavelength (Fig. 3B). Apparently, the excited state population undergoes random Brownian-type motion along all coordinates when the quasi-equilibrated excited state is reached. If the CI sink is encountered, non-radiative transition to either the trans or cis ground state takes place.
The dynamic process discussed above can be described in a simplified form as the following sequence of steps, each with its own rate: FC point valley–ridge inflection point CI sink, with k1 ≤ k2λ and k2λ1 > k2λ2 > k2λ3⋯> k2λn (λ1 < λ2⋯< λn). According to this description, we fit our data and obtain k1 as being equal to (5.2 ps)−1, while k2720 nm and k21000 nm are equal to (0.2 ps)−1 and (0.6 ps)−1, respectively. It is important to note that although barrierless torsion motion occurs after stretching in the SE kinetics, the increased τrise component appears before slower decay, which corresponds to the excited state population before the barrier. This is a conventional appearance of the inversion of the decay times in kinetics when the first step of a reaction is slower than its second step.
We have measured differential spectra of pinacyanol in ethanol and propanol to support the above conclusion. In the spectral range from 720 to 980 nm, the SE rise time increases from the instrument response limited in the blue side to 0.91 ps for ethanol and to 1.38 ps for propanol in the red (Fig. S2†).
The proposed sequential barrier-to-barrierless photoisomerization can only be observed when the barrier energy is small enough so that the excited state PES can undergo transformation from the valley to ridge topology. When the barrier for torsion motion is rather high, only an instrument response-limited rise of a mirror image SE spectrum should appear in the differential spectra.
In contrast to 1122C, pinacyanol may hypothetically form two types of cis isomer via rotating around one of the two CC bonds in the polymethine chain: (2Z,10E) and (2E,10Z) (Fig. 1). Here we denote them as P1 and P2, respectively. Although a combination of barrier and barrierless parts due to the specific property of the excited state PES agrees with just one isomerization pathway, there is still a need to consider the existence of the second isomerization pathway. Previously, the S1 state PES of pinacyanol in a vacuum was calculated by semi-empirical methods, showing that formation of P2 is barrierless while formation of P1 has a small activation energy barrier of 1.5 kcal mol−1 or 528 cm−1.44 The 0 → 0 absorptions were calculated to be at 590, 530 and 610 nm, for trans, P1 and P2, respectively. Assuming that these calculation results are still valid in solution, both photoisomerization pathways are expected to occur and two isomer absorption bands, blue-shifted P1 and red-shifted P2, should be observed. Nevertheless, only one photoisomer absorption band appears on a long time scale, being ∼20 nm red-shifted compared to the trans form, which is consistent with the calculation results for P2. The results of our measurements do not present any evidence for the blue-shifted P1, neither spectrally nor kinetically. The barrierless part in the long wavelength SE cannot be due to the second barrierless pathway. Indeed, if the two pathways would occur in parallel, SE at different wavelengths should exhibit different decay dynamics, contrary to our observations. Additionally, the time scale, ∼0.6 ps, seems to be too fast even for the complete barrierless photoisomerization of cyanines. For example, during photoisomerization of NK88 which undergoes both pathways, the time constant of the barrierless pathway was reported as 2–3 ps.26 Due to the appearance of only a red-shifted isomer absorption band, we suggest that the only formed photoisomer is P2 and speculate that the results of calculation are valid only for a vacuum whereas in solution the activation energy for P2 should substantially increase.
Since only a single type of photoisomer is formed, we can estimate the isomerization QY adopting the method employed for 1122C.10 We assume that the ground state absorption spectrum of the photoisomer is similar, but spectrally shifted and scaled, to that of the trans ground state. Then fitting of the spectral shape of ΔA(λ) at 100 ps with a superposition of trans state bleaching and the photoisomer photoinduced absorption, with the spectral shift and the scaling factor used as fit parameters, gives an estimate for the extinction coefficient of the photoisomer, εcis(λ). As shown in Fig. 4, the spectral shift and εcis627 nm are 527 cm−1 and 8.83 × 104 M−1 cm−1, respectively.
Now, the photoisomer QY in methanol can be calculated according to the expression below:
(1) |
It is interesting to note that although pinacyanol and 1122C have similar excited state lifetimes (5.2 and 4.9 ps, respectively), the QYs of the photoisomer are very different (3.6% vs. 26%, respectively). We compare these two cases to examine what properties of the excited state PES control the outcome of a photoreaction. Chart 1 shows three PES structures and their corresponding SE shapes calculated according to each model. The two-mode excited state PES of 1122C possesses a ridge-like feature (Chart 1A) and there is no barrier along the torsion coordinate.10 The excited state population then evolves entirely downhill to the PES minimum, implying that the 1122C molecule is elongated and twisted at the same time. As a result, a quasi-equilibrated state is formed at about a 90° dihedral angle. At this point, Brownian motion drives the excited state population to move randomly along the torsion coordinate among others, and non-radiative transition to the ground state PES takes place if the CI sink is encountered. The excited state PES of long-chain polyene molecules such as DDI is valley-like with a barrier along the torsion coordinate (Chart 1C). From the FC point, the CC bond is elongated until Minplanar is reached. Then barrier-crossing towards the CI sink along the torsion coordinate takes place. For pinacyanol, as discussed above, there is a small barrier along the torsion coordinate and the PES is of a valley–ridge topology. After excitation, the molecule is elongated first, and after a while starts to twist before Minplanar is reached. It then evolves downhill to the CI sink, just as in 1122C. The above processes are summarized in a simple form in Chart 2. Comparing these three cases, the main difference is the consequence of the different excited state PES topologies related to the FC configuration and to the location of the point where the effective torsion mode couples to the stretching mode. The later this coupling occurs, the greater the amount of excited molecules deactivated back to the trans ground state via non-radiative IC is, and therefore fewer photoisomers are formed. Thus, although pinacyanol exhibits similar excited state lifetimes with 1122C, the effective photoisomerization is less competitive with the other deactivation pathways.
Concerning the excitation photon energy-dependence of the photoisomerization of pinacyanol, it has been reported previously that more photoisomers are formed when the molecule is excited by a shorter wavelength.45 To test this statement, we have chosen 0 → 0 and 0 → 1 absorption transitions for excitation. In both cases, the characteristic time constants and QYs (3.6% vs. 3.5%) are very similar. We rationalize this observation considering that the vibronic bands in the steady-state absorption spectrum are attributed to the stretching mode of the N–C bond, excitation of which does not contribute to the CC bond torsion.
Following ref. 31, we assume that at η = 0 the SE rise time should be limited by the experimental temporal resolution (100 fs). From Fig. 5C, one can see that the lifetime (k1)−1 obtained from either ESA or SE shows a linear dependence on η. When fitted with a power-law function, an index of 1.0 is obtained. The linear dependence further proves a barrier character of the process as according to Kramer’s expression, non-radiative relaxation rates of thermally activated barrier-crossing are inversely proportional to the viscosity value in low viscous solvents.14,46
The dependencies of the rise time constants on η (Fig. 5D) are not linear and can be fitted with a power-law function. The power indexes of 0.60, 0.64 and 0.65 are obtained for 820, 900 and 990 nm, respectively. These values are very close to those reported for the barrierless case of 1122C.31 BFO theory has predicted that barrierless relaxation rates are fractional power-dependent on η, and increasing the index value corresponds to a smaller PES slope. Thus, these results further support our assignment that the red part of SE corresponds to a barrierless process.
The photoisomer QY is also solvent viscosity-dependent: it is 3.6% in methanol, 2.9% in ethanol and 2.1% in isopropanol. As we have seen, the excited state population dynamics on both the barrier and barrierless parts of PES slow down with η. Therefore more time is needed for the population to arrive to the CI sink, i.e., for the molecule to twist up to 90° of the dihedral angle. As a result, due to competition with the IC process more excited state molecules are transferred back to the trans ground state and the photoisomerization becomes less efficient, leading to a decrease of the photoisomer QY with η. The back-isomerization rate of the photoisomer is also viscosity-dependent; however, the time constants cannot be obtained accurately within our experimental time scale.
In this work, sequential barrier and barrierless photoisomerization was observed. This indicates that the excited state PES of pinacyanol corresponds to a transition of topology between a valley-like and ridge-like PES. To the best of our knowledge, this work constitutes the first dynamic study of a valley–ridge excited state PES.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra03299k |
This journal is © The Royal Society of Chemistry 2016 |