Molecular Dynamics at Electrical- and Optical-Driven Phase Transitions Time-Resolved Infrared Studies Using Fourier-Transform Spectrometers
aa r X i v : . [ c ond - m a t . m t r l - s c i ] A p r Noname manuscript No. (will be inserted by the editor)
Molecular Dynamics at Electrical- andOptical-Driven Phase Transitions
Time-Resolved Infrared Studies Using Fourier-TransformSpectrometers
Tobias Peterseim · Martin Dressel
Received: date / Accepted: date
Abstract
The time-dependent optical properties of molecular systems areinvestigated by step-scan Fourier-transform spectroscopy in order to explorethe dynamics at phase transitions and molecular orientation in the milli- andmicrosecond range. The electrical switching of liquid crystals traced by vibra-tional spectroscopy reveals a rotation of the molecules with a relaxation timeof 2 ms. The photo-induced neutral-ionic transition in TTF-CA takes place bya suppression of the dimerization in the ionic phase and creation of neutral do-mains. The time-dependent infrared spectra depend on temperature and laserpulse intensity; the relaxation of the spectra follows a stretched-exponentialdecay with relaxation times in the microsecond range strongly dependent ontemperature and laser intensity. We present all details of the experimentalsetups and thoroughly discuss the technical challenges.
Keywords
Fourier-transform infrared spectroscopy · step-scan technique · time dependent phenomena · vibrational spectroscopy · liquid crystals · photo-induced phase transition Fourier-transform infrared (FTIR) spectroscopy is a widely utilized methodto investigate the optical response of gasses, liquids and solids [1,2,3,4]. Ingeneral, steady-state properties are measured, however, numerous approacheshave been developed over the years to explore time-dependent phenomena byFourier-transform interferometry [5,6], many of them optimized for a certain
Tobias Peterseim and Martin Dressel1. Physikalisches Institut, Universit¨at Stuttgart,Pfaffenwaldring 57, D-70550 Stuttgart, GermanyTel.: +49-711-685 64946Fax: +49-711-685 64886E-mail: [email protected] Tobias Peterseim, Martin Dressel time regime. Standard rapid-scan techniques are limited by the mirror velocityto a fraction of a second (typically 10 ms), depending on the spectral resolu-tion ∆ν required: ∆t ∝ /∆ν . Since this is often not sufficient, step-scanningis nowadays implemented in several high-end commercial Fourier-transforminstruments because there is no inherent time limit. It covers the largest dy-namical range with a time resolution of the order of nanoseconds determinedby the current detector and electronics technology [6,7,8] and still achievinga high spectral resolution ∆ν > − . The only restriction is the repeata-bility of the process: depending on the resolution, spectral range and desiredsignal-to-noise ratio, it has to be executed hundreds of times.In contrast to the continuously moving interferometer mirror in the rapid-scan configuration, the advantage of the step-scan method is the continuousrecording of the signal at a fixed mirror position. This is repeated after themirror has moved to the next position, eventually composing the completeinterferogram. The step-scan technique is mainly applied in biophysics andpolymer chemistry where, for instance, the photolysis processes of chemicalreactions [9], bacteria systems [10,11,12,13,14] and the time-dependent reori-entation of liquid crystals under the influence of a short electric field [15,16,17,18,19] are studied. But also temperature- and light-induced phase transitionscan be investigated [20,21]. Furthermore, it is used to examine the characteris-tics of lasers and their mode spectra [9] and for photo-reflection measurementsof semiconducting materials as well as quantum wells [22,23].Here we want to present and discuss our experimental setups employed toinvestigate the dynamics of phase transitions. Tracing the vibrational spectraof molecules, we probe the configurational changes after the transition hasbeen triggered either by a short laser pulse (Sec. 4) or by an electrical pulse(Sec. 3). In the left panel of Fig. 1 the data acquisition of a step-scan measurementis schematically depicted. At each mirror position n∆x along the travelingdistance of the mirror the temporal varying reflection signal I ∗ ( n∆x, m∆t )is recorded. The complete interferogram is sampled for various times and re-tardations by successive stepwise moving of the mirror. A subsequent Fouriertransformation for each measured time point m∆t in the intensity “matrix” I ∗ ( nx, m∆t ) derives the time-resolved spectrum S ∗ ( j∆ν, m∆t ). The processcan be repeated several times to improve the signal-to-noise ratio whereas ateach n∆x position the time-dependent signal is averaged (typically 10 to 50times). Additionally, the stability of the mirror influences the signal-to-noiseratio significantly [10,24], therefore, one must take care that the spectrometeris located in a vibration-free and silent environment. For this reason we placeour FTIR-spectrometer on a heavy optical bench mounted on air attenua-tors to decouple the system from the environment. Furthermore, the vacuum olecular Dynamics at Electrical- and Optical-Driven Phase Transitions 3 Fig. 1 (left panel) At each mirror position n∆x the detector signal I ∗ ( n∆x, m∆t ) isrecorded as a function of time. After all mirror positions were passed through, the time-dependent signal is recieved by the execution of a Fourier transformation for each measuredtime point m∆t . (right panel) Illustration of the signal sequence within a step-scan experi-ment. (a) External trigger signal T1 given by the external perturbation source, for examplea pulse generator or laser. (b) The second signal T2 is generated by the spectrometer andsent to the interferometer motor to move the mirror to the next position. (c) After a certainpredefined stabilization time the third signal T3 waits for the next external trigger signalT1 and rises afterwards immediately. It stays high as long as all data points are captured.(d) Signal T4 corresponds to each recorded time point within the signal T3. pumps were placed in a separated room to reduce vibrations and acousticnoise. This way we reached a mirror stability better than 3 nm.Since the data acquisition at each mirror position n∆x has to start alwaysat the same time, the data recording must be synchronized with the externalstimulation source (laser or pulse generator). Thus, the trigger signal sequenceis very important and crucial in a step-scan measurement. The temporal se-quence of the TTL control signals is visualized in the right panel of Fig. 1. Anexternal or an internal trigger signal T1, which is correlated with the beginningof the reaction, controls the data acquisition. Yet, before the recording starts,the spectrometer sends a signal T2 to the interferometer, so that the mirroris moved to the next position. There, it is stabilized for a few milliseconds.After the stabilization procedure ( ∼
20 ms) the next arising trigger signal T1is used as a starting signal for the record window T3. The signal T3 stays highfor the total recording time T = N · ∆t ( N total number of time slices) whichis defined at the beginning of each measurement. As soon as the signal T3 ison, at each TTL-signal T4 the detector signal is captured. Depending on thenumber of averaging spectra this procedure is repeated several times startingagain with the T3 signal. Subsequently, the mirror moves to the next position.The time resolution ∆t depends mainly on the response time of the detec-tor. Standard PC-MCT-detectors, operating in the photo-current mode, havea minimum response time of 1 µ s. Their disadvantage is that the measuredcurrent becomes nonlinear above a certain incident threshold intensity. A pho-tovoltaic (PV) detector has a two order of magnitudes shorter response timedue to the small detector area. Furthermore, the measured signal is alwaysproportional to the incident light intensity. For the measurement a PC-MCTof the model D316 (Bruker Optics, Ettlingen) with a time resolution of about1 µ s and a PV-MCT KMPV11-11-1-J1 from Kolmar Technologies with a the-oretical rise time of 25 ns are utilized. The time resolution also depends onthe amplifier and the A/D-converter. There are two amplifiers, the build-inamplifier of the spectrometer and the KA100-A1 from Kolmar Technologies Tobias Peterseim, Martin Dressel with a bandwidth of 250 MHz. As an A/D-converter the internal converter ofthe spectrometer with its time resolution of 6 µ s, with a dynamical range of24 bit and a maximal input voltage of V pp = 20 V, can be used, or a transientrecorder M3i.4142 from Spectrum Systementwicklung Microelectronic GmbH,Grosshansdorf with a bandwidth of 400 MHz at 16 bit and V pp = 10 V.The standard interferogram of a Fourier-transform spectrometer consistsof an ac- and dc-component: while only the ac-signal contains the impor-tant spectral information the dc-signal is usually removed by an electronichigh-pass filter. The phase correction can only be performed directly for arapid-scan measurement. In a step-scan experiment the spectrum can includepositive as well as negative features. Thus, a phase correction with the rawtime-resolved ac-signal does not work. Two options exist to resolve this com-plication: first, the simultaneously recording of the dc-signal and the ac-signal;the dc-component provides the right phase correction for the ac-component.The second possibility is to use the phase of a previous rapid-scan measure-ment [3,25,26]. Electrically induced phenomena occur mainly in semiconductor leading to non-linear conductivity or negative differential resistance [27,28]. High electricfields can also trigger phase transitions, for instance in charge-density wavesystems [29] or correlated insulators [30,31]. In organic conductors electricalswitching is subject of research for quite some time [32,33,34,35,36,37,21].As an example of electric switching, we here have investigated the polariza-tion-dependent vibrational modes of liquid crystals, which allow us to trace theorientation of the molecule upon applying a voltage by time-resolved infraredmeasurements. In these molecules several vibrational modes exist which canbe assigned to different parts of the molecules. By static and time-resolvedpolarization-dependent measurements we gain on the one hand informationabout the orientation of the individual molecular building blocks in staticpositions and on the other hand about the temporal evolution of the electricswitching process. Hence, we learn about the rotation of the different molecularconstituents with the electric field.3.1 Liquid CrystalsLiquid crystals are widely used for electronic displays due to the possibility tomanipulate the orientation of the complex organic molecules by applying anelectric field and this way to control the transmitted light. We have chosen theFELIX 017/100 mixture from the Clariant as a ferroelectric liquid crystal thatconstitutes a chiral smectic C phase (SmC ∗ ) at ambient conditions creatinga helical structure with θ = 14 ◦ , as depicted in Fig. 2(a). The spontaneouselectric polarization p m = p m · ( h × n ) = p m sin θ is perpendicular to the olecular Dynamics at Electrical- and Optical-Driven Phase Transitions 5 Fig. 2 (a) Chiral smectic phase (SmC ∗ ): going from layer to layer the director n succes-sively helically precedes around the stacking normal vector h in an angle θ . The spontaneouspolarization p m is perpendicular to the director and the normal vector. The total polar-ization p total is zero. (b) Oriented smectic C phase: all directors n are aligned towards thesame direction. By symmetry the polarization p m = p m · ( h × n ) = p m sin θ can adopttwo possible values. (c) Side- and top view of the liquid crystal cell consisting of two CaF plates (light gray, transparent). A thin film of ITO was sputtered onto the plates (brightbrown). Polyimid (green) was deposited by the spin-coating process on top of the ITO layerand serves as an orientation layer for the liquid crystals. UV-glue in connection with smallplastic spheres (shaded area) with a thickness of d = 5 µ m keeps the two windows at afixed distance. (d) Transmission spectrum of CaF (black) measured at room temperaturein comparison to CaF coated with ITO (blue) and CaF in combination with ITO andpolyimid (red). The dashed line indicates a spectral range where water absorption and iceon the detector window corrupt the data. director vector n and the layer normal vector h . While in general the differentorientations average and the total polarization is zero, a uniaxial rubbed sur-face aligns the molecules parallel to each other, leading to a net polarization,sketched in Fig. 2(b). The whole procedure is well-known as “surface stabilizedferroelectric liquid crystal” (SSFLC) [38]. Two energetically equal configura-tions exist with p m arranged in opposite directions, but perpendicular to thesurface with the same tilting angle ± θ . An applied electric field can now inducethe transition between the two polarization states by changing the orientationof the molecule.While standard liquid-crystal cells consist of glass windows transparent inthe visible, we have constructed a cell with two CaF windows suitable forthe infrared spectral range from 1000 cm − to 50 000 cm − [39]. In Fig. 2(d)the transmission spectrum of a CaF window is plotted in the mid-infraredfrequency range. Compared to alternative window materials, such as KBr, itis not hydroscopic, and with n = 1 . n = 4), for instance, reducing the reflection losses considerably. The CaF Tobias Peterseim, Martin Dressel windows (20 ×
20 mm , thickness of 0 . windows are sputtered with a thin film of indium tin oxide(ITO) commonly used as an electrode material for displays because it is lucentover a broad energy range and at the same time conductive. However, opti-cal measurements plotted in Fig. 2(d) reveal that the transmission decreasesrapidly towards low frequencies [40]. In the range of interesting for the C=Cdouble bond vibrations the transmission is between 5 and 20%, which is stillsufficient for our experiment. In order to preorient the liquid-crystal moleculesa very thin film of polyimide was deposited on top of the ITO layer by using thespin-coating technique which was rubbed afterwards unidirectional [Fig. 2(c)].As soon as the molecules were placed on the surface, they align themselvesalong the rubbing direction [41]. As demonstrated in Fig. 2(d) the very thinlayer of polyimide does not affect the overall light transmission of the systemin the relevant frequency range in agreement with previous results [15,42,19].Both plates have been clued together with a UV glue mixed with small macro-scopic plastic spheres with a diameter of d = 5 µ m which serve as spacer tokeep the windows on a constant distance, as sketched in Fig. 2(c). After theglue was cured, the cell was heated up above 80 ◦ C so that the liquid-crystalmixture transformed in the isotropic phase with a reduced viscosity. The liquidcrystals were sucked in the chamber due to the capillary action. Afterwards,the temperature of the melt was slowly lowered so that a single domain wasformed. Due to the large dimension of the cell the distance of the two plateswas not constant and reduces to the center of the cell causing Newton’s rings.It changes again, when placed in a vacumm spectrometer. The contact wireswere made out of copper and fixed on the exposed ITO side areas with indium;they are connected to a Philips-PM 5768B pulse generator to switch the liquidcrystals. The generator sends during the measurement simultaneously to theswitching voltage pulse a second TTL signal to the spectrometer to start thedata acquiring process.The cell is placed in a Bruker Vertex 66v/s Fourier-transform infrared spec-trometer together with a suitable infrared polarizer. The absorption becomesmaximal if the infrared light is polarized parallel to the electronic transitiondipole moment µ e of the molecular vibrational mode. As soon as the externalelectric field E is switched on, a force acts on the liquid crystal. In general, theswitching process takes a few microseconds τ or until most of the molecules arereoriented. The new orientation of the molecules and its director n change thedirection of the transition dipole moment µ e leading to an increase or decreaseof the absorption signal.If the electric field is switched on, the directors shown in Fig. 2(b) spincollectively in one direction and therefore, the azimuthal angle changes. Theswitching velocity and -time can be derived from the equation of motion [43,44,45]: η ˙ χ ( t, z ) = − I ¨ χ ( t, z ) + K ∇ χ ( t, z ) − pE sin χ ( t, z ) , (1) olecular Dynamics at Electrical- and Optical-Driven Phase Transitions 7 with η the damping due the viscosity of the liquid, χ is the angle between theelectric field and the polarization of the molecule, z is the coordinate alongthe direction perpendicular to the window, and K the elasticity constant. Dueto the large moment of inertia I m = 1 × − kg/m the equation can besimplified to η ˙ χ = K ∇ χ − pE sin χ and solved for small angles χχ = χ exp (cid:26) tτ (cid:27) · sin n πzd o (2)with d = 5 µ m as the spacing between the two cell windows. We can extracttwo time constants, τ or for rise with the field and τ reor for fall after the electricfield E is turned off: τ or = ηpE − K π d τ reor = ηd Kπ . (3)The switching time for the here examined material FELIX 017/100 can beestimated using η = 8 × − Ns/cm , p m = 47 nC/cm , K = 5 × − N, E = 10 kV/cm, and d = 5 µ m to yield τ or = 170 µ s and τ reor = 40 ms.3.2 Steady-State Optical PropertiesThe mid-infrared transmission spectrum of the liquid crystal cell filled withFELIX is displayed Fig. 3. The dips in the spectrum between 1000 cm − and1700 cm − as well as between 2800 and 3000 cm − , enlarged in the insets (b) T V Wavenumber (cm -1 ) (a) Wavenumber ( cm -1 ) T V Fig. 3 (a) Transmission spectrum of FELIX 017/100 measured at room temperature. Theoscillations originate from interference effects caused by multi-reflections within the liquidcrystal cell and windows. (b) The detail view of the frequency region which shows mainlyresonance features of the vibrations of the molecular body. (c) The inset shows the featureof the methyl vibrations. Tobias Peterseim, Martin Dressel (b)
Vibrational modes (cm -1 ) 2924 2854 1606 1584 1546 1516 1439 Angle ((cid:176)) (a) FELIX 017/100
Polarization angle ((cid:176)) 0 20 40 60 80 100 120 140 160 S = S V - S V ( a r b . un i t s ) Wavenumber (cm -1 ) Fig. 4 (a) Spectra of the intensity difference ∆S = S V − S V for various polarizationstates of the incident light measured at room temperature. (b) Polarization-dependent mea-surement of the change of the transmitted intensity after creating 5 V on the two windowsfor seven different vibrational modes. Only the strongest modes are displayed. The maximalchange of the transmission appears at 35 ◦ , followed by a minimum at 125 ◦ . The sinusoidalmodification evidences that the liquid crystals react on the applied electric field. and(c), be assigned to the vibrational modes of the liquid-crystal molecules.The features ν − ν are stretching modes of CH - and CH end group. Themodes ν − ν are connected with the vibrations of the C=C double bonds be-longing to the center part of the molecules. The features at 1439 and 1395 cm − can be referred to the wiggling and torsion of the CH group. The low-lyingresonances at 1269 and 1243 cm − belong to an asymmetric stretch oscillationof the C-O-C bond. The position of the detected features perfectly coincideswith the resonance frequencies determined by Huang and Shih [19].Fig. 4 shows a change of the infrared spectra when an electric field of E = 5 V is applied. All measurements were performed at 25 ◦ C. Due to therearrangement of the liquid-crystal molecules a strong modification of the in-tensity could be observed. We can can deduce two different-oriented transitiondipole moments: the first belongs to the vibrational mode of the rigid body ofthe molecule and the other points out from the molecule axis related to the endgroups. The modes containing the CH and CH bonds exhibit a minimum at35 ◦ and a maximum at 125 ◦ , correspondingly. The minimum of the modes be-longing to the molecule body are slightly shifted and can be found at 125 ◦ . Thegreen curve, for instance, belongs to a vibrational mode with an aligned dipolemoment parallel to the central molecular frame. At 35 ◦ it reaches a maximumfor a voltage of 5 V, this means that the dipole moment rotates away fromthe electric field vector of the infrared light. Hence, the transmitted intensityincreases at this specific wavelength. In contrast, the intensity of the dipolemoments related to the end groups of the molecule is reduced because thetransition dipole moments are aligned parallel to the incident radiation. olecular Dynamics at Electrical- and Optical-Driven Phase Transitions 9
500 1000 1500 20000123
Wavenumber (cm -1 ) T i m e ( m s ) -4.5E-03-2.0E-034.6E-042.9E-035.4E-037.9E-03 S (arb. units) (a) 500 1000 1500 2000 -4-202468 S ( a r b . un i t s ) FELIX 017/100T= 25(cid:176)C, U = 6 VTime ( s) - 50 200 1150 1950
Wavenumber (cm -1 ) (b) Fig. 5 (a) Contour plot of time-dependent infrared spectra of FELIX, demonstratingthe switching process of the liquid-crystal cell for an electric field pulse of the strength E = 12 kV/cm with a pulse width of 2 ms. In a spectral region from 1200 to 1800 cm − several features can be recognized corresponding to the orientation of their transition dipolemoment. (b) Time-evolution of the spectra demonstrate for four selected times after thevoltage pulse of 6 V was applied. The spectrum (blue), recorded 50 µ s before the pulse isapplied, exhibits almost no change and is flat. Several features appear at time 200 µ s thatcan be ascribed to certain vibrational modes. The signal saturates after 1150 µ s and showsno further variation of the intensity. ◦ , and voltages between 4 and 6 V appliedbetween the two plates. The time-dependent infrared signal is presented ina contour plot in Fig. 5. The rectangular voltage pulse had a width of 2 mswith a repetition rate of 30 Hz and the signal was recorded for a period of4 ms. To improve the signal-to-noise ratio, five spectra were averaged. Thetime resolution was set to 25 µ s and the spectral resolution was 2 cm − . Wecan conclude that the liquid crystal molecules rotate and reorient like a stiffbody under the influence of an external electric field.The extreme values of ∆S ( ν, t ) are displayed in blue or in red in the contourplot. The transition dipole moments µ e are aligned parallel to E IR lead to areduced and, hence to a negative signal whereas in contrast for a positive µ e isturned away from E IR . This becomes more obvious in Fig. 5(b) where spectraat different times are plotted. As a check, data were taken 50 µ s before thepulse; no variation of the vibrational modes and no shift is observed. Afterapplying the voltage pulse, however, the spectrum dramatically alters andreveals the feature exactly as in Fig. 4(a) which marks the final state. Alsothe absolute values of the change are the same. In accord with Fig. 3(a) and4(a) no variation can be identified below 1000 cm − because the ITO layerabsorbs most of the incident light. The switching process is completed afterabout 1 ms and the signal stays constant over a period of 800 µ s. Fity +Aexp{-t/ } Applied voltage 4 V 5 V 6 V S ( a r b . un i t s ) Time (ms) voltage pulse length T = 25(cid:176)C ) 1584 ( ) 1606 ( ) (b)(a) Resonance frequency (cm -1 ) 1439 ( ) 1516 ( ) Voltage (V)
FELIX 017/100 (c) T i m e ( s ) Fig. 6 (a) The variation of the intensity ∆S ( ν, t ) is displayed for the resonance frequency1606 cm − for three different voltages. The rise of the signal as well as the drop can befitted well by a single exponential function y + A exp( − t/τ ). The signal increases withincreasing voltage. At the same time the switching speed increases. In contrast to that therelaxation process retards with higher voltage. (b) Voltage-dependence of the rise time τ or for five different resonance frequencies. With increasing voltage the switching time decreasescontinuously. (c) Whereas the relaxation time τ reor deceases, but no direct dependence onthe voltage can be discovered. For 6 V τ reor reveals a larger variance. To determine the rise and fall time, the temporal evolution of ∆S ( ν, t )of five vibrational modes were used. The change of the intensity of the ν mode is plotted as a function of time in Fig. 6(a) for three different voltages.With a rise time of less than 10 ns, the voltage pulse basically has a rectangularshape; however, the infrared signal reacts slowly and requires several hundredsof microsecond until it saturates. Due to the viscosity of the liquid crystals themolecules react with a certain delay to the electric field. The switching time τ or diminishes with rising voltage whereas the amplitude increases slightly.However, the signal overshoots at the beginning for 6 V and relapses back tothe value of the 5 V pulse. The rise time can be determined by fitting theexperimental data with a single exponential function y + A · exp {− t/τ } . Theextracted values of τ or are summarized in Fig. 6(b) and (c) as a function ofthe applied voltage. At the end of the pulse the signal relaxes slowly back toits initial values within 2 ms and is at least by a factor of four larger thanthe rise time τ or . The time for reorientation τ reor can be determined from thetemporal evolution of the recovery process by a single exponential function.At a voltage of 4 V almost all vibrational modes reveal a switching time τ or between 210 µ s and 230 µ s; with increasing voltage this range enlarges tomaximum 70 µ s for 6 V whereas the values are between 190 µ s and 120 µ s.The various time constants imply that constituents of the molecules reactdifferently on the electric field. The vibrational modes related to the moleculebody ( ν -mode) rotate slower than the ones ( ν -mode) which are connected tothe end groups. The theoretically calculated value for τ or = 170 µ s (for 5 V)derived from Eq. (3) agrees very well with the experimentally determined τ or olecular Dynamics at Electrical- and Optical-Driven Phase Transitions 11 which is of about 200 µ s. The reason for the small discrepancy may be dueto uncertainties in the distance of the spacer plates or in the material specificparameters η and p m ; they all depend strongly on temperature. From ourlimited data set, we cannot verify the predicted 1 /E characteristic of the risetime; our data indicate a constant time. Probably, τ or exhibits the expectedfield dependence and the divergent behavior for smaller voltages.As it was expected from the theoretical calculations of τ reor the relaxationtimes in Fig. 6(c) are larger than the switching time. A time constant of40 ms was predicted but not overserved. Furthermore, the relaxation rateis a function of the applied voltage but according to Eq. (3) it should beindependent of any external parameter. One possible explanation therefor isthe stronger tilting of the molecules with increasing voltage as well as the totalamount of realigned molecules increases as well. Thereby, the layer of orientedmolecules increases which leads to a kind of self-stabilization effect resulting inthe extended decay. For this reason, the relaxation rate τ reor is not a functionof the intrinsic elastic constant K , but also of the external eclectic field. Over the last decades photo-induced phase transitions were investigated inseveral material classes, for instance, polymers [46], organic charge-transfersalts [47,48,49,50,51,52], transition-metal oxides like vanadium oxides [53],cuprates [54], and manganites [55]. In the following we study the one-dimensionalorganic mixed-stack crystals tetrathiavulvalene-chloranil (TTF-CA) that un-dergoes a ionic-neutral phase transition at T NI = 81 . et al. [46], ultrafast pump-probe experiments have been performed to examinethe photo-induced phase transition PIPT in the femto- and picoseconds timerange [46,56,57,49]. The understanding of these phenomena was boosted bytheories of Nasu and Yonemitsu [58,59,60,61]. In the case of TTF-CA openquestions concern the relaxation of metastable domains, the related time scale,and the modification of the infrared spectrum due to photo-excitation; someof these have recently be addressed by time-resolved FTIR-spectroscopy [20].4.1 Neutral-Ionic Transition in TTF-CAAt ambient conditions the planar molecules of TTF and CA are equidistantlyarranged in alternating stacks along the a -direction (Fig. 7). Since the chargetransfer ρ = 0 . e as determined from optical experiments is rather small[62,63,64], this state is referred to as the neutral phase; upon cooling thecharge transfer increases slightly. At T NI = 81 . ρ increases from 0.3 e to about0.6 e ; Fig. 7(d) displays a sketch of the arrangement. Accordingly the dielectric Fig. 7 (a) Chloranil molecule (CA, C Cl O ) and (b) tetrathiafulvalene (TTF, C S H ).(c) Monoclinic unit cell of TTF-CA at room temperature. The TTF and CA molecules areordered along the crystallographic a -axis. (d) At T = 300 K the space group of the unitcell is P2 /n and the CA and TTF molecules are stacked equally spaced along the a -axis.A further stack is located at z = c = 0 .
5, respectively, at which the TTF-CA pairs aretilted opposite to the a -axis. In the ionic phase the TTF and CA molecules dimerize alongthe stack. By the charge transfer of about ρ = 0 . e electric dipoles are formed along thestacking direction. T NI = 81.5 K T (K) l n ( c m ) -1 ) TTF-CA -1 ) d l n / d T - ( K ) T (K)
Fig. 8
Arrhenius plot of the resistivity ρ dc ( T ) (green) of TTF-CA along the stacking direc-tion. Above the transition it behaves as a classic band insulator with an activation energyof ∆ = 0 .
12 eV. constant diverges at the transition with a pronounced frequency dependentresponse [66]. The application of 11 kbar pressure shifts the transition to roomtemperature [67,68].The resistivity of TTF-CA along the stacking direction increases uponcooling, following an activated behavior as displayed in the Arrhenius plotof Fig. 8. At T NI , ρ dc ( T ) drops by one order of magnitude before it increaseagain at lower temperatures. The strong enhancement of the conductivity isattributed to the rising number of neutral-ionic domain walls in the neutralphase which also explains the dielectric behavior and non-linear transport[69]. Alternatively it was suggested, that the multiphonon Peierls coupling olecular Dynamics at Electrical- and Optical-Driven Phase Transitions 13 Fig. 9 (a) Reflectivity and (b) optical conductivity of TTF-CA for T = 8 K (blue) and 85 K(red) taken for the polarization along the stack. The maximum is mainly caused by emv-coupled modes which gain intensity in the dimerized ionic phase. (c) Moleuclar vibrations ofchloranil (CA) and (b) tetrathiofulvalene (TTF): the upper rows depict the gerade a g -modesand the lower rows the ungerade b u -modes. leads upon approaching the phase transition to a progressive shift of spectralweight and of the coupling strength toward the phonons at lower frequencies,ending in a soft-mode behavior only for the lowest-frequency phonon near thetransition temperature [70,71]. In the proximity of the phase transition, thelowest-frequency phonon becomes overdamped due to anharmonicity inducedby its coupling to electrons.In Fig. 9 the mid-infrared reflectivity and the optical conductivity of TTF-CA are presented for temperatures above and below the neutral ionic tran-sition. In this energy range the intramolecular modes of the TTF and CAmolecules can be identified and assigned [72,63,73,20]; some of them aresketched in panels (c) and (d). Along the a -direction the symmetric a g as wellas the infrared-active b u modes of CA and TTF can be observed whereas thefirst one is only infrared active due to the emv-coupling. Above the phase tran-sition in the neutral phase the TTF and CA molecules are not dimerized andthus the a g modes are only weakly infrared-active and the optical conductivityis low.In the ionic phase the point inversion symmetry is lost and the moleculesbecome dimerized: the intensity of a g modes is enhanced enormously. Time-resolved measurements thus allow us to explore the dynamics at the neutral-ionic phase transition by looking at the change in dimerization of the TTFand CA molecules, the increase of the ionicity. We can investigate the buildingof metastable domains until they eventually annihilate.4.2 Experimental DetailsFor photo-induced experiments a Nd:YAG laser (B.M Industries/Thales, YAG-502DNS-DPS920) was operated in the second harmonics λ = 532 nm (2.35 eV) Fig. 10
Sketch of the optical setup including the laser beam to excite the sample. The setupconsists of three main parts (dashed lines). The centerpiece of the setup is the spectrometercontaining the electronics as the A/D-converter and amplifier, but also the Globar lightsource, the interferometer with the beamsplitter and several mirrors to deflect the lightbeam. The second part is the infrared microscope which is attached to the spectrometer.There, the light is focused on the sample by a Cassegrain reflector. Furthermore, a polarizerand bandpass filter can be mounted in the microscope. The MCT detector is placed at theend of the light beam. of the fundamental wavelength (1064 nm). The pulse length is 8 ns. The repe-tition rate can be selected internally and externally between 1 Hz and 20 Hz.The laser intensity is adjusted continuously by a Brewster plate in addition toneutral density filters between 0.1 and 0.5 optical density. The laser intensitywas checked by a power and energy meter in front of the sample. The longterm stability of the laser power is better than 5%.A telescope arrangement reduces the diameter of the laser beam from 1 cmby a factor of 2, illustrated in Fig. 11. The laser beam is directed from theoptical bench to the infrared microscope via several mirrors (see Fig. 10).There, it is deflected on the sample by a 45 ◦ aluminum coated mirror mountedbelow the Schmidt-Cassegrain objective of the Bruker HYPERION infraredmicroscope. A lens ( f = 400 mm) focuses the beam on the sample, by varyingthe focal length. The light is circular polarized. The laser system and opticalsetup are spatially decoupled from the spectrometer and mounted on an opticaltable to suppress possible external vibrations.The laser pulse sequence is controlled externally to ensure the temporalsynchronization between the laser, the pulse generators and the FTIR-spec-trometer, as depicted in Fig. 11. The charging of the flash lamps is triggeredby an external signal as well as the Pockels cell generating the laser pulse.Therefore, a HP pulse generator (PM 5786 B) sends the trigger signal X1 tothe delay generator (EG&G Princeton Applied Research Model 4144). One ofthe delay generator output signals X2 is forwarded to a second pulse generator olecular Dynamics at Electrical- and Optical-Driven Phase Transitions 15 Fig. 11
The left panel sketches the laser setup. The centerpiece is the pulsed Nd:YAG laserwhich is externally controlled by two synchronized pulse generators. A delay generator isadditionally implemented in the setup to guarantee synchronization between the laser pulseand the voltage pulse generator as well as the oscilloscope. It also triggers the spectrometerto acquire the infrared data. The laser beam (green) is collimated, directed and focused onthe sample via several mirrors and lenses. The pulse sequence to control the laser-inducedexperiment is depicted on the right panel. X1 is the master trigger pulse triggering thedelay generator. It synchronizes the flash lamps (X2 and X3), the voltage pulse generator(Avtech) (X5) and the FTIR-spectrometer (X8) with each other. X4 controls the Pockelscell in the resonator with a minimum length of 6 µ s. The pulse X6 is the voltage pulsewhich is applied to the sample for the photoconductivity measurements, for instance. Theoscilloscope records the variation of the sample current or resistance and is synchronizedwith the other instrument via the signal X7. (HP 214B) which releases a further delayed pulse X3 with a minimum lengthof 150 µ s and minimum height of 5 V. It initializes the charging of the flashlamps of the laser system. After signal X3 drops to zero the lamps are chargedwith a delay of 1.5 ms. It also activates the discharge of the lamp with a delayof 15 µ s. About 30 µ s after the end of the charging and discharge pulse X3 afurther voltage signal X4 from the first HP PM 5768 B pulse generator (length6 µ s, 6 V) is sent to the Pockels cell generating the laser pulse. In the case of thephotoconductivity measurements the delayed trigger signal X5 from the delaygenerator activates the pulse generator (Avtech Electrosystems Ltd., Ottawa)to apply a voltage pulse X6 to the sample. The synchronizing pulse X7 of theAvtech device goes to the Tektronix oscilloscope to start the acquisition of thephotocurrent. The third signal X8 from the delay generator is used to initializethe time-resolved infrared measurement of the FTIR-spectrometer.4.3 Time-Resolved Infrared StudiesFor the PIPT measurements we excite only one of the first transitions of TTF + by the second harmonic of the Nd:YAG laser; all intermolecular transitions ofCA − lie above present photon energy. Also excitations from lower lying bandsinto the valence band are possible. Suzuki et al. compared the dependence of Fig. 12 (a) Illustration of a one-dimensional chain of dimerized TTF + ρ and CA + ρ pairsin the ionic phase in TTF-CA. A TTF + ρ molecules is excited with a laser pulse withthe photon energy hν . (b) Vertical excitation of the HOMOs of TTF + ρ according to theFrank-Condon-principle in the LUMO of TTF + ρ . The excitation is strongly localized to themolecule. (c) Creation of excitons, for instance Frenkel-excitons, which are delocalized acrossthe whole molecules. (d) Via different relaxation processes and channels charge transfer ex-citons are created which triggers to the transition of the neutral phase. (e) The dimerizationis suppressed and the charge between the molecules is redistributed. A neutral domain is cre-ated in the ionic host matrix which is separated by neutral-ionic domain walls. (f) Afterwardsthe neutral domains extend along the one-dimensional chain. By electron-phonon-couplingalso neighboring ionic chains are converted into neutral regions and a three-dimensionaldomain is established. the conversion efficiency on the photon energy; by excitation of intramoleculartransitions no threshold intensity occurs to create neutral domains [57].In Fig. 12 the optical generation of neutral domains is schematically illus-trated. The light pulse will create intramolecular Frenckel-excitons, which areelectron-hole pairs strongly localized on the excited molecules. They decay viavarious channels into several charge-transfer excitons [59] and by that lead toa phase transition. In the case of direct excitation only one exciton is created,which is not enough to establish a macroscopic, metastable domain extendedover several D A pairs. With a sufficient number of photons a multiplicative,non-linear effect can be established, formating metastable domains. By thecreation of the one-dimensional, neutral, non-dimerized region, neutral-ionicdomain walls are formed between the neutral and ionic parts. The excitationenergy of these domain walls is about 0.1 eV [67,69] and corresponds to theactivation energy of 120 meV and 65 meV in the ionic phase [20]. This isin good agreement with predictions of an activation energy between 25 and olecular Dynamics at Electrical- and Optical-Driven Phase Transitions 17 -0.20-0.15-0.10-0.050.000.050.10 I = 0.71 mJ/cm T = 78 K
354 s 156 s 48 s 42 s 12 s 0 s (a)
E a t R Wavenumber (cm -1 ) TTF-CA t R ( t ) no r m a li z ed (c)(b) I (mJ/cm ) 0.71 0.36 0.57 0.30 exp{-(t/ ) } T (K) 68 78 73 exp{-(t/ ) }
Time (ms)
Fig. 13 (a) The temporal sequence of the reflectivity change ∆ t R = R ( t ) − R (0) (solidlines) after photo excitation is depicted for various times. The behavior resembles the staticreflectivity difference ∆ T R = R − R presented in Fig. 9(a). (b) and (c) Normalized ∆ t R ( t ) for various laser intensities recorded at 1390 cm − for T = 78 K and for differenttemperatures for 0.71 mJ/cm . The time profile can be successfully fitted by a stretchedexplonential function exp {− ( t/τ ) β } (dashed linies).
56 meV [74,75,76]. Afterwards, the electronic system couples to the crystal lat-tice and exciting phonons [49] via electron-phonon coupling and hence, createsshock-waves. They can convert neighboring chains into neutral domains.In Fig. 13(a) the time-dependent behavior of the reflectivity change ∆ t R = R ( t ) − R (0) at T = 78 K is plotted for a laser pulse intensity of 0.71 mJ/cm . Itdirectly compares to the static reflectivity change ∆ T R = R − R plottedin Fig. 9(a). By photo excitation ∆ t R becomes negative within a short timewhich is below the experimental time resolution of 6 µ s. The direct comparisonof the ∆ t R shape and the static reflectivity change ∆ T R reveal that the ionicphase was not only dissolved, but also a transition into a neutral state wasinduced. Within several hundreds of microseconds the signal ∆ t R relaxes backto zero which means that the ionic phase is reestablished. Moreover, no changeof the spectral shape with the elapsed time and laser pulse intensity could bedetected. By comparing the shape of ∆ t R with the corresponding difference ofthe reflectivity ∆ T R between 78 K and 85 K, we conclude that the vanishingof this features indicates the dissolving of the dimerization state between theTTF and CA molecules. Moreover, we suggest that metastable, non-dimerizedneutral domains in the ionic matrix are created.To trace the temporal evolution of the PIPT in dependence of the pumpintensity and the sample temperature, we have chosen the very intense ν (a g ) mode of TTF residing at 1390 cm − since we have asserted that thetemporal evolution is the same for the whole spectra. The normalized ∆ t R ( t )is represented in Fig. 12(b) for different pulse intensities. At the beginning thesignal decays very fast and at the end it flattens out. At the vicinity of T NI thefirst component decays faster with decreasing laser intensity. A fit by a simple single or double exponential function [59] does not yield satisfactory results.However, we obtain an excellet agreement when using a stretched-exponentialfunction, which is also called Kohlrausch-William-Watt function ∆ t R ( t ) ∝ exp (cid:8) − ( t/τ ) β (cid:9) , (4)as depicted in Fig. 13(b) and (c). The fitting parameters β and τ are a functionof the laser intensity and decreases from 0.35 to 0.42 and from 3 . × − to2 . × − s with decreasing laser intensities.In Fig. 13(c) ∆ t R ( t ) is displayed for various temperatures T = 68, 73, and78 K. Far below T NI , the temporal dynamics of the reflectivity drops very fastwithin the first 20 µ s and approaches asymptotically a constant value whichis in contrast to the temporal profile at the vicinity ( T =78 K) of T NI whichconstantly diminishes. Similar to the dependence of the fitting parameters onthe laser intensity the effective recombination time τ as well as the stretchingexponent decrease from 2 . × − s to 3 . × − s and from 0.42 to 0.23,respectively, with decreasing sample temperature.The observed time profile can be explained by a random-walk annihilationprocess of the generated neutral-ionic domain walls. Our comprehensive time-resolved infrared study and the random-walk model [20] allow us to concludethat close to the phase transition, large domains are formed due to the valenceinstability. We find that the merger and interaction of the induced domainsplay an important role for the formation of the macroscopic domains anddeduce from the model with decreasing laser intensity, the average domain sizedecreases. At lower temperatures the ionic phase is more robust; the averagedomain size is much smaller and changes less with laser intensity. The randomwalk of the neutral-ionic domain walls is the dominant factor for the relaxationof the metastable domains in the temperature range considered. We have presented two examples where time-resolved infrared investigationsusing step-scan Fourier-transform spectroscopy provide insight in the molecu-lar dynamics at the phase transition and molecular orientation. In the case ofliquid crystals, the application of an electric field switches the orientation of themolecular dipoles changing the transmission of the cell for polarized infraredlight. Watching the time evolution of the pronounced molecular vibrationalmodes allows us to extract the time constant of approximately 2 ms and itsdependence on the applied voltage. The neutral-ionic phase transition of TTF-CA can be photo-excited with a short laser pulse, creating domain walls thatare mobile and eventually annihilate. We follow the time-dependence of thevibrational modes activated by the changed dimerization and ionicity. Therelaxation back to the initial state strongly depends on the temperature andlaser intensity, it extends from a 3 µ s to almost 1 ms. For both examples wegive details of the experimental setups and limitations. We demonstrate the olecular Dynamics at Electrical- and Optical-Driven Phase Transitions 19 applicability of step-scan Fourier-transform spectroscopy in a large dynamicalrange. Acknowledgements
We thank F. Sch¨org and F. Giesselmann for providing the liquidcrystals and G. Untereiner for continuous help; E. Kurz and N. Fr¨uhauf supported us whenbuilding the cell and use of their clean room. Funding by the Deutsche Forschungsgemein-schaft (DFG) is acknowledged. T.P. thanks the Carl-Zeiss-Stiftung for financial support.
References
1. R.J. Bell,
Introduction to Fourier Transform Spectroscopy (Academic Press, New York,1972)2. H.U. Gremlich, H. G¨unzler,
IR-Spektroskopie: Eine Einf¨uhrung , 4th edn. (Wiley-VCH,Weinheim, 2003)3. P.R. Griffiths, J.A. De Haseth,
Fourier transform infrared spectrometry , 2nd edn.(Wiley-Interscience, Hoboken, N.J, 2007)4. B. Smith,
Fundamentals of Fourier Transform Infrared Spectroscopy , 2nd edn. (CRCPress, Taylor & Francis, Boca Raton, FL, 2011)5. J. Sloan, E. Kruus, in
Time Resolved Spectroscopy , vol. 18, ed. by R. Clark, R. Hester(John Wiley & Sons., Chichester, 1989)6. G.D. Smith, R.A. Palmer, in
Handbook of Vibrational Spectroscopy , vol. 1, ed. by J.M.Chalmers, P.R. Griffiths (John Wiley & Sons., Chichester, 2001), pp. 625–6407. A. Laubereau, M. Stockburger (eds.),
Time-resolved Vibrational Spectroscopy , vol. 4(Springer, Berlin, 1985)8. P.R. Griffiths, B.L. Hirsche, C.J. Manning, Vib. Spectrosc. , 165 (1999)9. T.J. Johnson, A. Simon, J.M. Weil, G.W. Harris, Appl. Spectrosc. , 1376 (1993)10. W. Uhmann, A. Becker, C. Taran, F. Siebert, Appl. Spectroscopy , 390 (1991)11. B. Hessling, J. Herbst, R. Rammelsberg, K. Gerwert, Biophys. J. , 2071 (1997)12. R. Rammelsberg, B. Hessling, H. Chorongiewski, K. Gerwert, Appl. Spectroscopy ,558 (1997)13. R. Rammelsberg, S. Boulas, H. Chorongiewski, K. Gerwert, Vib. Spectrosc. , 143(1999)14. M. Schleeger, C. Wagner, M.J. Vellekoop, B. Lendl, J. Heberle, Analytical and Bioan-alytical Chem. , 1869 (2009)15. V.G. Gregoriou, J.L. Chao, H. Toriumi, R.A. Palmer, Chem. Phys. Lett. , 491 (1991)16. N. Katayama, T. Sato, Y. Ozaki, K. Murashiro, M. Kikuchi, S. Saito, D. Demus,T. Yuzawa, H. Hamaguchi, Appl. Spectroscopy , 977 (1995)17. N. Katayama, M.A. Czarnecki, M. Satoh, T. Watanabe, Y. Ozaki, Appl. Spectroscopy , 487 (1997)18. A.L. Verma, B. Zhao, A. Bhattacharjee, Y. Ozaki, Phys. Rev. E , 051704 (2001)19. J.Y. Huang, W.T. Shih, J. Phys.: Condens. Matter , 7593 (2006)20. T. Peterseim, P. Haremski, M. Dressel, EPL (Europhysics Letters) , 67003 (2015)21. T. Peterseim, T. Ivek, D. Schweitzer, M. Dressel, Phys. Rev. B , 12345 (2016)22. J. Shao, F. Yue, X. Lu, W. Lu, W. Huang, Z. Li, S. Guo, J. Chu, Appl. Phys. Lett. ,182121 (2006)23. J. Shao, W. Lu, F. Yue, X. Lu, W. Huang, Z. Li, S. Guo, J. Chu, Rev. Sci. Instrum. , 013111 (2007)24. C. Rodig, F. Siebert, Appl. Spectroscopy , 893 (1999)25. W. Herres, J. Gronholz, Understanding FT-IR data processing (1984). Bruker Ana-lytlsche Messtechnlk GmbH, Karlsruhe, Germany26. T. Peterseim, Organic conductors in equilibrium and nonequilibrium states: The inter-action with electric and light pulses. Ph.D. thesis, Universit¨at Stuttgart (2015)27. S.M. Sze, K.K. Ng, Physics of semiconductor devices , 3rd edn. (Wiley-Interscience,Hoboken, N.J, 2007)28. P.Y. Yu, M. Cardona,
Fundamentals of semiconductors: Physics and materials proper-ties , 4th edn. (Springer, Berlin, 2010)0 Tobias Peterseim, Martin Dressel29. G. Gr¨uner,
Density waves in solids (Addison-Wesley Pub. Co., Advanced Book Pro-gram, Reading, Mass., 1994)30. A. Asamitsu, Y. Tomioka, H. Kuwahara, Y. Tokura, Nature , 50 (1997)31. A. Sawa, Materials Today , 28 (2008)32. R.S. Potember, T.O. Poehler, D.O. Cowan, Appl. Phys. Lett. , 405 (1979)33. Y. Tokura, H. Okamoto, T. Koda, T. Mitani, G. Saito, Phys. Rev. B , 2215 (1988)34. R. Kumai, Y. Okimoto, Y. Tokura, Science , 1645 (1999)35. K. Inagaki, I. Terasaki, H. Mori, T. Mori, J. Phys. Soc. Jpn. , 3364 (2004)36. T. Mori, T. Kawamoto, I. Terasaki, T. Kakiuchi, H. Sawa, Phys. Rev. B , 235103(2007)37. T. Ivek, I. Kovaˇcevi´c, M. Pinteri´c, B. Korin-Hamzi´c, S. Tomi´c, T. Knoblauch,D. Schweitzer, M. Dressel, Phys. Rev. B , 245125 (2012)38. N.A. Clark, S.T. Lagerwall, Appl. Phys. Lett. , 899 (1980)39. E.D. Palik (ed.), Handbook of Optical Constants of Solids, Author and Subject Indicesfor Volumes I, II, and III (Academic Press, San Diego, CA, 2009)40. P.K. Biswas, A. De, N.C. Pramanik, P.K. Chakraborty, K. Ortner, V. Hock, S. Korder,Mat. Lett. , 2326 (2003)41. K. Sakamoto, R. Arafune, N. Ito, S. Ushioda, Y. Suzuki, S. Morokawa, Jpn. J. Appl.Phys. , L1323 (1994)42. A. Verma, B. Zhao, S. Jiang, J. Sheng, Y. Ozaki, Phys. Rev. E , 3053 (1997)43. N.A. Clark, M.A. Handschy, S.T. Lagerwall, Mol. Cryst. Liq. Cryst. , 213 (1983)44. M.A. Handschy, K.M. Johnson, G. Moddel, L.A. Pagano-Stauffer, Ferroelectrics , 279(1988)45. J.W. Goodby (ed.), Handbook of liquid crystals , 2nd edn. (Wiley-VCH, Weinheim, 2014)46. S.Y. Koshihara, Y. Tokura, T. Mitani, G. Saito, T. Koda, Phys. Rev. B , 6853 (1990)47. S.y. Koshihara, Y. Tokura, K. Takeda, T. Koda, A. Kobayashi, J. Chem. Phys. , 7581(1990)48. H. Okamoto, Y. Ishige, S. Tanaka, H. Kishida, S. Iwai, Y. Tokura, Phys. Rev. B ,165202 (2004)49. S. Iwai, Y. Ishige, S. Tanaka, Y. Okimoto, Y. Tokura, H. Okamoto, Phys. Rev. Lett. , 057403 (2006)50. K. Ikegami, K. Ono, J. Togo, T. Wakabayashi, Y. Ishige, H. Matsuzaki, H. Kishida,H. Okamoto, Phys. Rev. B , 085106 (2007)51. H. Okamoto, K. Ikegami, T. Wakabayashi, Y. Ishige, J. Togo, H. Kishida, H. Matsuzaki,Phys. Rev. Lett. , 037405 (2006)52. M. Mitrano, G. Cotugno, Clark, S. R., R. Singla, S. Kaiser, J. St¨ahler, R. Beyer, M. Dres-sel, L. Baldassarre, D. Nicoletti, A. Perucchi, T. Hasegawa, H. Okamoto, D. Jaksch,A. Cavalleri, Phys. Rev. Lett. , 117801 (2014)53. A. Cavalleri, T. Dekorsy, H. Chong, J.C. Kieffer, R.W. Schoenlein, Phys. Rev. B ,161102 (2004)54. J. Demsar, K. Biljakovi´c, D. Mihailovic, Phys. Rev. Lett. , 800 (1999)55. M. Rini, R. Tobey, N. Dean, J. Itatani, Y. Tomioka, Y. Tokura, R.W. Schoenlein,A. Cavalleri, Nature , 72 (2007)56. S.y. Koshihara, Y. Takahashi, H. Sakai, Y. Tokura, T. Luty, J. Phys. Chem. B ,2592 (1999)57. T. Suzuki, T. Sakamaki, K. Tanimura, S.y. Koshihara, Y. Tokura, Phys. Rev. B ,6191 (1999)58. K. Nasu, Relaxations of Excited States and Photo-Induced Structural Phase Transitions (Springer, Berlin, 1997)59. K. Nasu,
Photoinduced phase transitions (World Scientific, Singapore, 2004)60. K. Yonemitsu, K. Nasu, J. Phys. Soc. Jpn. , 011008 (2006)61. K. Yonemitsu, K. Nasu, Phys. Rep. , 1 (2008)62. A. Girlando, R. Bozio, C. Pecile, J.B. Torrance, Phys. Rev. B , 2306 (1982)63. A. Girlando, F. Marzola, C. Pecile, J.B. Torrance, J. Chem. Phys. , 1075 (1983)64. C.S. Jacobsen, J.B. Torrance, J. Chem. Phys. , 112 (1983)65. M. Le Cointe, M.H. Lem´ee-Cailleau, H. Cailleau, B. Toudic, L. Toupet, G. Heger,F. Moussa, P. Schweiss, K.H. Kraft, N. Karl, Phys. Rev. B , 3374 (1995)66. S. Tomic, M. Dressel, Rep. Prog. Phys. , 096501 (2015)olecular Dynamics at Electrical- and Optical-Driven Phase Transitions 2167. T. Mitani, Y. Tokura, Y. Kaneko, K. Takaoka, T. Koda, G. Saito, Synthetic Met. ,515 (1987)68. A. Dengl, R. Beyer, T. Peterseim, T. Ivek, G. Untereiner, M. Dressel, J. Chem. Phys. , 244511 (2014)69. H. Okamoto, T. Mitani, T. Komatsu, Y. Iwasa, T. Koda, G. Saito, Phys. Rev. B ,8224 (1991)70. M. Masino, A. Girlando, A. Brillante, R.G. Della Valle, E. Venuti, N. Drichko, M. Dres-sel, Chem. Phys. , 71 (2006)71. A. Girlando, M. Masino, A. Painelli, N. Drichko, M. Dressel, A. Brillante, R.G.Della Valle, E. Venuti, Phys. Rev. B , 045103 (2008)72. A. Girlando, I. Zanon, R. Bozio, C. Pecile, J. Chem. Phys. , 22 (1978)73. R. Bozio, I. Zanon, A. Girlando, C. Pecile, J. Chem. Phys. , 2282 (1979)74. N. Nagaosa, J. Phys. Soc. Jpn. , 3488 (1986)75. N. Nagaosa, J. Takimoto, J. Phys. Soc. Jpn. , 2745 (1986)76. Z.G. Soos, A. Painelli, Phys. Rev. B75