Ultrafast Potential Energy Surface Softening of One ... - ACS Publications

Jul 2, 2010 - Alke Meents and Christian Sager. HASYLAB at DESY, Notkestrasse, D-22607 Hamburg, Germany. Jav Davaasambuu, Gerhard Busse and ...
1 downloads 0 Views 691KB Size
J. Phys. Chem. A 2010, 114, 7677–7681

7677

Ultrafast Potential Energy Surface Softening of One-Dimensional Organic Conductors Revealed by Picosecond Time-Resolved Laue Crystallography Marc Messerschmidt* SLAC, Menlo Park, California, 94025, and HASYLAB at DESY, Notkestrasse, D-22607 Hamburg, Germany

Thomas Tschentscher European XFEL Project GmbH, Albert Einstein Ring 19, 22761 Hamburg, Germany

Marco Cammarata European Synchrotron Radiation Facility, 38043 Grenoble, France

Alke Meents and Christian Sager HASYLAB at DESY, Notkestrasse, D-22607 Hamburg, Germany

Jav Davaasambuu, Gerhard Busse, and Simone Techert Max Planck Institute for Biophysical Chemistry, IFG Structural Dynamics of (Bio)chemical Systems, ¨ ottingen, Germany 37070 G ReceiVed: May 4, 2010; ReVised Manuscript ReceiVed: June 22, 2010

Time-resolved Laue crystallography has been employed to study the structural dynamics of a one-dimensional organic conductor (tetrathiafulvalene-p-chloranil) during photoexcitation in the regime of the neutral to ionic phase transition. Exciting this crystalline system with 800 nm 100 fs long optical pulses leads to ultrafast population of a structural intermediate as early as 50 ps after excitation with a lifetime of at least 10 ns. Starting from the neutral phase, this intermediate has been assigned as a precursor state toward the photoinduced population of the ionic phase. The observed intensity changes are significantly different from comparable equilibrium structures. The interpretation of this structural data is that the potential of this intermediate is being softened during its population in a dynamical process. The depopulation proceeds through thermal processes. The nonmonotonic control of properties of matter by absorption of optical photons using photoinduced phase transitions (PIPT) is one of the most promising concepts to be applied in photonics, data storage, or optoelectronics.1 PIPTs have therefore been investigated in a wide range of materials and technologically important compounds. Several systems promise further access to ultrafast time scales, and initial investigations of the light-induced dynamics of phase transitions in spin-crossover compounds,2 one-dimensional organic conductor systems,3,4 semiconductors,5 and self-organizing membrane model systems6 have previously been carried out. The underlying physicochemical mechanism of the PIPT can differ largely between different systems, and complex phenomena such as strong electron correlation or highly cooperative effects caused by strong electron-phonon interaction play an important role. With timeresolved X-ray diffraction, the structural properties of the molecular configuration and of the crystal lattice can be resolved during a PIPT with atomic resolution.7,8 The biggest modifications affect, e.g., the unit cell dimension or symmetry. Organic, in particular metallo-organic, compounds exhibit in many cases photoswitching between their para- and ferroelectric phases.3 One of the compounds, for which the ultrafast structural and spectroscopic response functions of the PIPT have been * Corresponding author. Electronic address: [email protected].

studied in great detail, is tetrathiafulvalene-p-chloranil (TTFCA), whose structure was originally reported in 1979.9 At Tc ) 82 K TTF-CA exhibits a temperature driven phase transition from an ionic to a neutral phase.10 In spectroscopic studies of TTF-CA it is found that near Tc it is possible to photoinduce in this system a macroscopic phase transition that nonmonotonically modifies its ferroelectric properties.3 With time-resolved X-ray diffraction under similar conditions, it is found that a symmetry change of the molecular crystal, attributed to the macroscopic phase transition from the neutral to the ionic phase, is delayed by nearly a nanosecond relative to the absorption of 800 nm optical photons. The origin of the delay had been interpreted as the time required to switch a large amount of crystal domains by highly cooperative effects.4 This measurement was limited in time resolution to about 100 ps. On a much faster time scale, ultrafast optical reflectometry reveals that, with excitation in the charge transfer band at 1900 nm, coherent oscillations in TTF-CA with damping times of about 2 ps are created.11,12 In this work we add another important aspect about the dynamics of structural properties in TTF-CA close to the PIPT. In view of the ultrafast coherent oscillations and the delayed development of the symmetry change of the crystal lattice structure, it is of interest if a much faster structural intermediate is formed. Such an intermediate structure would serve as a

10.1021/jp104081b  2010 American Chemical Society Published on Web 07/02/2010

7678

J. Phys. Chem. A, Vol. 114, No. 29, 2010

Messerschmidt et al.

precursor state for the structural transition indicated by the symmetry change on the 1 ns time scale. Using picosecond timeresolved Laue diffraction, we observe a structural response in TTF-CA populated within 80 ps. This newly found intermediate state is generated with full conservation of the unit cell as no spot position shifts are observed within the first nanosecond despite immediate intensities changes. I. Experimental Section The time-resolved X-ray experiments have been performed at ID-09B (ESRF, Grenoble). Carried out in 16 bunch mode with a pulse separation of 196 ns, a high speed chopper has been used to isolate single bunches. The stroboscopic pumpprobe experiments achieved a time resolution of 80 ps determined by the duration of X-ray pulses, while the synchronization to the femtosecond optical laser pulse has a resolution of a few picoseconds. Given the high flux in the polychromatic experiments (109 photons/pulse for a bandwidth of 3%) very small crystals with 10-20 µm thickness can be measured by accumulating data for just 30 single pulses per exposure at 20 Hz. Since the extinction length of 800 nm is around 1 µm for TTF-CA, up to 10% of the crystal can be directly transformed. The extinction coefficient has been determined by optical transmission spectroscopy based on the Lambert-Beer law of polycrystalline TTF-CA in a KBr matrix. The penetration depth µ-1 is defined as µ-1 ) ε · c/oD, where oD is the optical density (equals one), ε is the molar decadic extinction coefficient, and c is the concentration of TTF-CA polycrystals in the matrix. In fact, the observed electron density differences agree with smaller populations of excited molecules even for photoexcitation intensities just below the damage threshold of TTF-CA single crystals. The time delay between the exciting laser pulse and probing X-ray is set by varying the arrival times of the femtosecond laser pulses with respect to the synchrotron X-ray pulses. Delay times are varied from 50 ps to 1 µs. Due to the time resolution of the apparatus, the absolute time zero setting is within (80 ps. Therefore, a formal time zero might be any time point within the resolution of the experiment. Additionally, reference measurements are performed at negative time delay. Only temperatures just above the phase transition temperature (Tc ) 82 K) are set using an open flow liquid nitrogen cooling system. Temperature increases up to 90 K do not result in pronounced intensity changes in the X-ray diffraction of TTFCA, while even at only 2 K temperature difference, small changes in Bragg positions have been observed. Furthermore, the optical laser fluence has been varied during the course of the experiment for 800 nm excitation. Below laser fluences of 10 mJ · cm-2 no light induced changes in the quality of the diffraction pattern are observable. Crystal damage is found above 40 mJ · cm-2, which has been recorded as crystal cracking. Measurements of significant responses in the diffraction pattern have been recorded for optical laser fluences between 15 and 40 mJ · cm-2. We record several full data sets for different time delays each covering a 90° rotation in 1° steps. The Laue method is highly sensitive for monitoring the positions of a diffraction spot in reciprocal space, because at any given orientation the reflection condition is met for some of the reflections. In monochromatic experiments a small change in the lattice parameter corresponds only to small shifts in the Bragg reflection positions in reciprocal space and therefore might spatially not be resolved. For the Laue case the same change can result in a shift of the reflecting wavelength for several reflections. This results in no change of reflection angle for some reflections, while others show a more

Figure 1. Time evolution of the relative spot size (the amount of integrated pixels per spot) for two data sets at 40 mJ · cm-2 and one at 15 mJ · cm-2 optical laser excitation fluences. Negative delay times (NEG) and no laser exposure (OFF) are displayed for reference.

pronounced shift, which becomes observable in the diffraction pattern even for slight lattice changes. Hence, we are able to measure very small shifts in the Bragg diffraction positions, which relate to relative lattice parameter changes below 0.1%. In comparison, the thermally activated lattice changes are more than 5 times bigger for temperature variations of a few Kelvin. For a series of diffraction patterns at varying crystal angles, we evaluate sets of reflections differing only in the reflecting wavelength and resulting in several reflections with very high redundancy. This feature is especially useful for time-resolved experiments on exposure of sensitive samples (either with optical methods or with respect to X-rays) where minimization of the experimental exposure time is important and high significance in observed data are needed to observe small structural effects. Using Laue diffraction, we thus obtain the sensitivity needed to monitor small lattice changes, and therefore to investigate new and yet unobserved structural changes, which are not observable with monochromatic diffraction techniques. A characteristic of the transient X-ray diffraction pattern of TTF-CA is that changes in the Laue reflection spot shape immediately result in integral intensity changes on at least a 10% level. Since these diffraction spot shape changes can potentially bias the data analysis, we give some introducing remarks to this effect and use it for selecting useful data sets. The results of the measurements of the spot sizes as a function of optical laser fluence are illustrated in Figure 1 and can be used to monitor the development of quality of the photoexcited crystalline structure during the time course of the structural response. Starting from very small changes at early delay times, a continuous growth of the spot size is observed. At late times (10 ns and beyond), streaking of reflections occurs and indicates strong lattice deformations in particular for high laser fluences. For further data evaluation, we therefore mainly focus on transient diffraction data sets with optical excitation conditions that show at long time delays moderate shifts of the diffraction spots in reciprocal space and very moderate changes in the spot shapes. The data evaluation, including their normalization, is based on the program packages LaueGUI13 and Precognition/Epinorm.14 In the final scaling, λ-curves are refined and they closely match the measured undulator curves of the ID-09 beamline. ShelX15 is used for structure refinements and merging of all symmetry equivalent reflections. The resulting sets of independent reflections scaled to an overall structure show significant

PES Softening of 1D Organic Conductors

J. Phys. Chem. A, Vol. 114, No. 29, 2010 7679

Figure 2. Time evolution of the integral intensities of various Bragg peaks averaged over all reflected wavelengths. Data sets at 40 (a, b) and 15 (c) mJ · cm-2 laser excitation. A single green marker was added to label zreo delay time.

changes for several reflections, as can be observed from Figure 2. The statistics of the reflections can be improved by using their high redundancy and merging the relative changes of these reflections (similar to Coppens16). Measurements for negative time delays and without laser should agree to within a small error. Therefore, data sets with differences in these reference measurements exceeding 10% are rejected to limit the influence of hot pixels and shadowing by the laser beam stop.

Figure 3. (A) Structure of neutral CA in the ab plane. The elements are assigned according to cgs notation. (B) Projection of the neutral structure of TTF-CA with view along the a-axis emphasizing the stacking character of this one-dimensional electron donor/acceptor system.

II. Results and Discussion Figure 2 summarizes the time evolution of the integral intensities of selected Bragg reflections as a function of time. The reflections have been assigned according to their Miller indices. As can clearly be seen, the intensities of the Bragg reflections already modulate at very early time points, within the time resolution of the experiment (80 ps). For the selected reflections, the changes are positive, constant, or negative. We find for several reflections a strong variation of the integral intensity corresponding to the formation of a new structure, while the overall intensities are found to agree within 2%. Note that changes in the Debye-Waller factor due to a temperature increase by optical laser excitation would lead to an intensity decrease of all diffraction peaks. The overall effect on the TTF-CA structure after photoexcitation can be visualized using difference Fourier synthesis. This way, difference electron densities can be directly generated between positive time points and the reference measurement. We observe a time resolution limited population of an intermediate structural state at short times that stays significantly populated for at least 10 ns. Figure 3 shows the ground state structure corresponding to the neutral TTF-CA structure at 100 K. The color of the elements is according to cgs notation. Figure 3A shows the unit cell of neutral TTF-CA in the ab plane. In Figure 3B a projection within the planes is shown emphasizing the stacking character of this one-dimensional electron donor (TTF)/acceptor (CA) system. Figure 4 visualizes the difference electron density map of CA and TTF 50 ps after photoexcitation (Figure 4A,C) and near the thermal phase transition (Figure 4B,D). For the structure determination the same conditions have been used as for the structure presented in Figure 1 (red curve). For the structure determination the conditions as Figure 4A shows the photoinduced electron density difference at the CA molecule 50 ps after photoexcitation, which corrresponds to the electron density of the photoexcited molecule minus the electron density of the molecule in the neutral structure at a negative time point (-1 ns). By convention, electron density decreases

Figure 4. Difference electron density maps of TTF and CA (explanations in text). (A) Photoinduced electron density change of CA 50 ps after photoexcitation minus the electron density of the structure at negative time point (-1 ns). (B) Electron density change as the difference between the electron density of the ionic phase structure minus the electron density of the neutral phase structure. (C) and (D) present respective difference maps for TTF. The assignment of the difference map to 50 ps resembles the time resolution of the apparatus rather than a precise time description. The contours were set to (0.3 e · Å-3 (A, C) and (1.0 e · Å-3 (B, D), respectively.

are assigned as negative electron difference densities, indicated by blue contours. Electron density increases correspond to positive difference density and are represented by red contours. For the photoinduced electron density map the contours are set to (0.3 e · Å-3 and 1375 unique reflections have been used

7680

J. Phys. Chem. A, Vol. 114, No. 29, 2010

Messerschmidt et al.

TABLE 1: Time Evolution of Difference Electron Density ∆ED/e · Å-3 a delayb

min

max

rms

50 ps 200 ps 1 ns 1 µs

-0.355 -0.397 -0.352 -0.728

0.516 0.508 0.496 1.326

0.083 0.113 0.103 0.171

a Data taken at 90 K with 40 µJ laser power. b The negative delay time (-2 ns) is taken as reference.

for the structural refinement. Figure 4B visualizes the corresponding electron density change as the difference between the electron density of the ionic phase structure (70 K structure) minus the electron density of the neutral phase structure (78 K structure) near the thermal phase transition temperature of Tc ) 82 K (70 K structure minus 78 K structure). Here, the contours are set to (1 e · Å-3 with a comparable number of unique reflections for the structural refinement as in the timeresolved experiment. Parts C and D of Figure 4 present the corresponding differences for the TTF structure: the photoinduced electron density changes versus the thermally induced electron density change by going through the thermal phase transition. First we describe the found time-resolved experiments (Figure 4A,C): In particular, for early time points (between 50 ps and 1 ns), an electron density difference of about ∆E = 0.5 e · Å-3 has been observed (see table 1). These difference electron densities can be refined for all early time points up to 1 ns. As reference, t ) -2 ns has been used. For late time points (t ) 1 µs), the density difference increases substantially to ∆E ) 1.33 e · Å-3, which coincides with the increase of streakiness in the diffraction peaks. As explained in the Experimental Section, this increase can be assigned to the strong response of the TTFCA lattice late after photoexcitation. The observed photoinduced transformation seems to be homogeneous over the range investigated here, since we cannot find splitting of reflections even in cases of clear shifts of the Bragg reflections. Differences similar to the observed ones of Table 1 can be modeled by assuming an overall increase of 5% in the vibrational amplitude in TTF-CA. The increase of >10% for some Bragg peaks indicates additional localized reorganization, while the overall structure change is small, as indicated by good agreement of most reflections independent of the time delay. Concerning the thermally induced electron density changes (Figure 4B,D), by going from the neutral (high temperature) to the ionic phase (low temperature phase), our measurements confirm the results obtained by the very detailed analysis of Lecomte et al.17 The anisotropic electron density change around the sulfur atoms of TTF and the chlorine and oxygen atoms of CA shows very clearly the tilt of the TTF and CA planes on one side of these molecules toward each other: one side loses electron density where on the other side of the molecule electron density increases. Note that, for comparison to the time-resolved experiment, the difference map ionic phase minus neutral phase has been chosen. The shown difference map depicts changes around the phase transition, using data differing by just 8 K. Therefore, when going from the neutral phase to the ionic phase, thermal effects and structural effects due to the change of the electronic state in TTF and CA can be separated. Now we come back to the time-resolved crystallographic studies, which reveal another type of electron density modulation (Figure 4A,C): In contrast to the thermal phase transition, the electron density map around chlorine (CA) and sulfur (TTF) is

of isotropic character. In addition, the unit cell dimensions do not change for time points before 10 ns. Note that Laue crystallography is very sensitive to unit cell changes. The blue electron density contour plots therefore emphasize a decrease in the electron population at the heavy atom sites that is not compensated by an electron density increase in specific locations as revealed in the thermally induced phase transition. This key difference in the photoexcited structure change versus thermally driven phase transition can be explained as follows: in the case of photoexcitation, the initial structural response resides inside the cell while the lattice response requires much more time to react. Therefore, the lattice is in a nonequilibrated state. The photoinduced neutral to ionic phase transition of TTF-CA belongs to the group of electron transfer reactions, which can be explained within the picture of the Marcus theory for diabatic electron transfer. This nonequilibrated lattice can be pictured as a modification of the gradient of the potential energy hypersurface whose coordinate is the distance between one TTF and one CA molecule inside the cell with respect to the equilibrium state. Regarding the picture of electron transfer theory, the isotropically distributed electron density map indicates an additional inner reorganization energy around the heavy atoms that is equivalent to a dynamical potential softening in the modes and coordinates in which the heavy atoms are involved. The decrease of electron density means a dislocation of electronic population into disorder, as a higher vibrational excitation in TTF and CA. Note that this higher vibrational excitation corresponds to the increase of inner reorganization energy, as explained above. The change of inner reorganization energy and change of state population occurs within 80 ps, as can be seen in the experimental data of Figure 2. From examination of the position of the Bragg reflection, it can be concluded that the lattice transformation of TTF-CA and the thermal processes concerning the lattice transformation happen at long delay times (10 ns and later), as also indicated by significant position shifts. These changes in the crystalline properties can be assigned to changes in the outer (crystalline) reorganization energy, according to the diabatic electron transfer picture. The current structural findings are therefore in good agreement with time-resolved reflectometry investigations in ref 3. As a consequence of stronger vibrations, simulations for different Debye-Waller factors of the sulfur atom as well as the experimental observed changes in the photocrystallographic experiments suggest a displacement increase of the sulfur atom on the order of 0.1 Å (in maximum, harmonic approximation). In the case of the sulfur-including vibrations, we suggest that the relevant vibrations are dominantly low frequency lattice vibrational modes, as they are out-of-plane and in-plane deformation, asymmetric and symmetric stretching vibrations, and wagging and torsion motions. Note that temperature increases due to laser heating are dominantly seen in electron difference maps using electron densities taken without laser excitation as reference. Using electron densities taken at negative time points as reference, as we have done, eliminates these pure thermal effects.20 III. Conclusion In conclusion, within the time resolution of the experiment we have investigated a new structural intermediate forming instantly after photoexcitation by time-resolved Laue diffraction. The electron density difference map reveals an isotropic distribution around the heavy atom sulfur in the electron donor TTF, and chlorine and oxygen in the electron acceptor CA.

PES Softening of 1D Organic Conductors These findings are in contrast to the structural changes observed in the thermally induced phase transition. For the thermally induced phase transition the electron difference is of clear anisotropic character, which can be assigned to a structural tilt of the CA and TTF apart from each other. In the photoinduced structural change, the difference map coincides with an electron density decrease around the TTF sulfur atoms and the CA chlorine atoms and a very small increase/decrease around the CA oxygen atoms. The dominant isotropic distribution is explained by a dynamical potential softening in the modes and coordinates in which the heavy atoms are involved during the PIPT. These findings confirm that the mechanism of the PIPT of TTF-CA, already at very early time points, substantially differs from a thermally induced phase transition. Acknowledgment. We acknowledge the European Synchrotron Radiation Facility for provision of synchrotron radiation facilities at beamline ID-09b. This work was supported by EUFLASH MC-8041. S.T. is grateful to the DFG, SFB 602, Aventis Foundation and Fonts of the Chemical Industry. J.D. was funded out of EU-Grant NEST MI-3567. The Advanced Study Group of the Max Planck Society is thanked for continuous support. Supporting Information Available: Table of distances between TTF and CA molecules and experimental and calculated undulator spectrum of ID09B. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Nasu, K. Relaxation of Excited States and Photo-induced Structural Phase Transitions; Springer-Verlag: Berlin, Heidelberg, 1997. (2) Guetlich, P.; Hauser, A.; Spiering, H. Angew. Chem., Int. Ed. 1994, 33, 2024.

J. Phys. Chem. A, Vol. 114, No. 29, 2010 7681 (3) Koshihara, S.; Takahashi, Y.; Sakai, H.; Tokura, Y.; Luty, T. J. Phys. Chem. B 1999, 103, 2592. (4) Collet, E.; Lemee-Cailleau, M. H.; Cointe, M. B.-L.; Cailleau, H.; Wulff, M.; Luty, T.; Koshihara, S. Y.; Meyer, M.; Toupet, L.; Rabiller, P.; Techert, S. Science 2003, 300, 612. (5) Lindenberg, A. M. Science 2005, 308, 392. (6) Quevedo, W.; Petri, M.; Busse, G.; Techert, S. J. Chem. Phys. 2008, 129, 024502. (7) Cole, J. Chem. Soc. ReV. 2004, 33, 501–513. (8) Bourgeois, D.; Schotte, F.; Brunori, M.; Vallone, B. Photochem. Photobiol. Sci. 2007, 6, 1047–1056. (9) Mayerle, J. J.; Torrance, J. B.; Crowley, J. I. Acta Crystallogr. 1979, B35, 2988–2995. (10) Le Cointe, M.; Leme´e-Cailleau, M. H.; Cailleau, H.; Toudic, B.; Toupet, L.; Heger, G.; Moussa, F.; Schweiss, P.; Kraft, K. H.; Karl, N. Phys. ReV. B 1995, 51, 3374–3386. (11) Iwai, S.; Ishige, Y.; Tanaka, S.; Okimoto, Y.; Tokura, Y.; Okamoto, H. Phys. ReV. Lett. 2006, 96, 057403. (12) Iwai, S.; Tanaka, S.; Fujinuma, K.; Kishida, H.; Okamoto, H.; Tokura, Y. Phys. ReV. Lett. 2002, 88, 057402. (13) Messerschmidt, M.; Tschentscher, T. Acta Crystallogr. 2008, A64, C611. (14) Ren, Z. J. Synchrotron Rad. 1999, 6, 891–917. (15) Sheldrick, G. M. SHELXL97; University of Goettingen: Germany, 1997. (16) Coppens, P. J. Synchrotron Rad. 2009, 16, 226–230. (17) Garca, P.; Dahaoui, S.; Katan, C.; Souhasso, M.; Lecomte, C. Faraday Discuss. 2007, 135, 217–235. (18) Frisch, M. J.; et al. Gaussian 98; Gaussian Inc.: Pittsburgh, PA, 1998. (19) Painelli, A.; Girlando, A. J. Chem. Phys. 1986, 84, 5655–5671. (20) The model is based on a quantum chemical simulations of the potential and vibrational modes of neutral and ionic TTF-CA dimer and tetramer units. For the DFT simulations the Gaussian 98 program package (B3LYP, 6-31 G basis set)18 has been used. It should be noted that stationary infrared measurements reveal that the active modes supporting the equilibrium neutral-to-ionic phase transition are strongly affected by vibronic and electron-phonon couplings.19 From the softening of the potential energy hypersurface in the coordinates describing oscillations with strong vibronic couplings, we estimate a possible increase of the displacement of the sulfur atom by maximally 0.1 Å for TTF-CA in the excited state.

JP104081B