Appearance of a Photoinduced Hidden State in the Electron Donor–Acceptor Type Metal–Organic Framework (NPr4)2[Fe2(Cl2An)3]

Electron donor–acceptor (DA)‐type metal–organic frameworks (MOFs) with valence instability are promising molecular materials for the design of photo‐responsive and electronic/magnetic functional materials. Here, ultrafast photoinduced dynamics in a DA‐type layered MOF that exhibits a charge‐transfer (CT)‐type phase transition are reported: (NPr4)2[Fe2(Cl2An)3], where NPr4+ = tetra‐n‐propylammonium and Cl2An2− = 2,5‐dichloro‐3,6‐dihydroxo‐1,4‐benzoquinonate. At room temperature (300 K: RT), ultrafast photoinduced CT between the Fe and Cl2An ions induces a sensitive change in the state associated with the valence instability from a single‐chain, electron‐correlated state to a new photoinduced, structurally modulated state. In the photoinduced state, two absorption bands are observed, one on the higher‐energy side of the CT band and the other in the mid‐IR range. This strongly implies that the local inversion center on the Cl2An ion that exists in the initial state disappears instantly upon photoexcitation, causing an ultrafast change in the lattice structure due to the softening of rigid bonds. This has never been realized in thermal excitation. These findings demonstrate that a new electronic state with a unique lattice structure—i.e., a photoinduced hidden state—appears in this MOF system at ultra‐high speed (within 110 fs) upon photoexcitation at RT.


Introduction
The photo-response of electron donoracceptor (DA)-type thermally driven electron-transfer (TDET) systems [1][2][3][4][5][6][7][8][9] is an attractive target for achieving fast, nonthermal, and contactless control of crystal functionality.The reason is that a chargetransfer (CT)-type transition-which has the lowest energy of photoexcitation in such TDET systems-can be an efficient mechanism for controlling electron/spin functionality.Such a DA-type TDET system is usually a molecular material with a low-dimensional, electron-correlated lattice, as defined by anisotropically directed electronic correlations or chemical bonds between the constituent molecules/ions.The low dimensionality of such an electroncorrelated lattice may induce several kinds of instability associated with the electrons, lattice, and spin.These are key factors for achieving an efficient photo-response because of the inherently cooperative electron-electron, electron-lattice, and magnetic interactions that are strongly expected.
Most neutral-ionic (N-I) phase-transition systems [10][11][12] are 1D DA-type systems, although some interchain Coulombic interactions have been observed. [12]In particular, TTF-CA crystals (where TTF = tetrathiafulvalene and CA = p-chloranil) exhibit highly efficient photoinduced N-I phase conversions [13][14][15][16][17][18][19] -so-called "photoinduced phase transitions". [20,21]In this case, the I phase is accompanied by a dimerization-type lattice distortion due to a spin-Peierlslike instability, which is essential for realizing electronic ferroelectricity. [22]We note that the N-I phase transition in TTF-CA can be regarded as a non-magnetic-to-non-magnetic phase transition, since the I phase is a spin-singlet due to dimerization.From the viewpoint of discovering magnetic photo-responsive systems, it is therefore quite interesting to explore TDET systems that involve multiple spin changes of CT-related constituent units/ions with limited lattice distortions.
The functionalization of molecular-framework systemswidely known as metal-organic frameworks (MOFs) [9,[23][24][25] -has been developed significantly during the past two decades and has opened a new avenue in molecular materials science, especially because of their porous nature.The potential  IM o ). [40]The Fe ions are depicted as colored circles, and the Cl 2 An ions are represented by thick colored lines.The arrows inside the circles and lines represent the spins.
[31] In addition, most MOFs-especially three-dimensionally infinite MOFshave robust and electronically inert frameworks, with rigid electronic structures.They also have high crystallinity and porosity, which makes it difficult for them to respond very strongly to external stimuli.However, recent studies have focused intensively on flexible MOFs [32][33][34] and electron-conjugating low-dimensional MOFs. [35]Flexible MOFs have thus become attractive targets from both a structural and an electronic point of view for designing stimulus-responsive materials tuned, for example, by photoexcitation. [36]Indeed, photoexcitation can lead to a variety of fascinating situations, such as simultaneously photoinduced local energy and charge transfer between D and A ions, photoconductive phenomena, and the photo-thermal effect on relatively slow timescales. [34,36,37]Regrettably, ultrafast photocontrol of physical properties based on the cooperative interactions in MOF systems has rarely been reported.Nevertheless, since DA-type MOFs that exhibit CT-type phase transitions have become much more popular recently, [8,12,[38][39][40] it is possible that some MOFs may be ideal systems for observing fast photo-responses of the electronic states.
In the present study, we focused on the DA-type layered MOF (NPr 4 ) 2 [Fe 2 (Cl 2 An) 3 ] (where NPr 4 + = tetra-npropylammonium and Cl 2 An 2− = 2,5-dichloro-3,6-dihydroxo-1,4-benzoquinonate). [40]This compound has a stable layered structure that can be regarded as a D 2 A 3 -type 2D honeycomb layer crystallized in the space group P2 1 /n, where D and A correspond to the Fe and Cl 2 An ions/ionic units, respectively. [40]igure 1a shows a schematic illustration of the structure of the 2D honeycomb layer of [Fe 2 (Cl 2 An) 3 ] 2-.The NPr 4 + cations are placed between the 2D honeycomb layers, but they are omitted from this figure for clarity.Interestingly, this compound shows a two-step CT-type phase transition, with the steps occurring at 317 K [T 1/2(1) ] and 354 K [T 1/2 (2) ].The phases distinguished by T 1/2(1) and T 1/2(2) are named a low-temperature phase (LT phase), an intermediate phase (IM phase), and a high-temperature phase (HT phase) (Figure 1b). [40]The charge and spin patterns in the honeycomb layer of each phase have been confirmed by magnetic, structural, and Mössbauer spectroscopic measurements.Despite its honeycomb layer structure, the charge alignment of the LT phase consists of quasi-1D chains of Fe 3+ and Cl 2 An 3- along the a+c direction, which are connected with neighboring chains by the diamagnetic Cl 2 An 2− (S = 0) ions.Since the Fe 3+ (S = 5/2) and Cl 2 An 3− (S = 1/2) ions are antiferromagnetically coupled along the chain, these ferrimagnetic chains are magnetically isolated; this magnetic structure in the LT phase is confirmed to be a single-chain magnet (SCM). [40,41]The IM phase appears upon increasing the temperature above T 1/2(1) , and the HT phase appears above T 1/2(2) (Figure 1b).The IM phase may take the form of either a charge-disproportionate ordered state (IM o ) or a delocalized state (IM d ), with a 1:1 ratio of Fe 2+ to Fe 3+ and a 1:1 ratio of Cl 2 An 2-to Cl 2 An 3-in the chain.In contrast, the HT phase consists solely of a set of Fe 2+ and Cl 2 An 2− ions; i.e., it is a paramagnetic phase.The CT-type phase transition thus corresponds to a phase transition from an SCM-related charge alignment to a paramagnetic charge alignment.The driving mechanism for these phase transitions has been discussed from the viewpoint of multi-stability, reflecting the balance between the valence change and the Coulomb interaction, as in an N-I phase-transition system. [12,42]However, a robust framework structure without appreciable lattice deformation-like the dimerization in TTF-CA-even at each T 1/2 is quite fascinating for investigating ultrafast photoinduced CT-type phase changes.
Herein, we report photoinduced dynamics in the DA-type layered MOF system, (NPr 4 ) 2 [Fe 2 (Cl 2 An) 3 ], which we examined using fs-time-resolved spectroscopy at room temperature (300 K = RT).The characteristics of the transient state imply an ultrafast and efficient appearance of a photoinduced CT state caused by the cooperative valence instability in the LT phase.The change from the SCM-related state to the paramagnetic state occurred within 110 fs.This photo-induced CT phenomenon was not limited to a  and 0.9 eV.e) Temperature Dependence of the integrated () between 2.5 and 2.7 eV.The dashed lines represent the transition temperatures T 1/2(1) (317 K) and T 1/2(2) (354 K) of the charge-transfer (CT)-type phase transitions based on previous magnetic and structural studies. [40]ngle DA pair but extended to several DA pairs by virtue of cooperative interactions.Moreover, the ultrafast formation of new spectral structures in the CT transition and the mid-IR energy ranges suggests the disappearance of the local inversion centers on the Cl 2 An ions due to a change in the lattice structure.Specifically, this fact means that even in a robust MOF system, both the charge and the lattice structure may become soft immediately following photoexcitation.This demonstrates for the first time the appearance of a new, transient electronic state in the present MOF system, with a unique lattice structure that is different from the one observed in thermal equilibrium, i.e., it is a photoinduced hidden state. [21,43]

Results and Discussion
We obtained anisotropic reflectivity spectra of a single crystal of (NPr 4 ) 2 [Fe 2 (Cl 2 An) 3 ] at RT using linearly polarized light, which are plotted in Figure 2a.The spectra reflect the anisotropic electronic structure of this crystal.We observed a large difference between the spectrum with the polarization parallel to the 1D chain (the a+c axis, blue line) and perpendicular to the chain (the b axis, orange line).Figure 2b shows the optical-conductivity [()] spectra obtained from a Kramers-Kronig (KK) transformation of the reflectivity.We observed two absorption peaks, at ≈0.8 and 2.6 eV.Quantum-chemical calculations based on density-functional theory (DFT) revealed that the peak at ≈0.8 eV has the character of a CT transition between the Fe and Cl 2 An ions, while the peak at ≈2.6 eV is attributed to an intramolecular transition in the Cl 2 An ions.Indeed, DFT calculations for an isolated Cl 2 An ion showed a similar valence change, as shown in Figure S1e (Supporting Information), which strongly supports our assignment of the intramolecular transition at ≈2.6 eV.The details of the DFT calculations are described in the Supporting Information (see Supporting Information, Note S1).
We measured the () spectra with E || chain at several temperatures in the range 296-380 K, and the spectra at 296, 340, and 380 K depicted in Figure 2c are typical of the LT, IM, and HT phases, respectively.We integrated the () spectra with E || chain at each temperature in the photon-energy ranges 0.7-0.9eV and 2.5-2.7 eV to determine the spectral weights of the CT and intramolecular transitions, as shown in Figure 2d,e, respectively.The spectral weight of each peak shows a steep drop at around each T 1/2 due to the LT-IM and IM-HT phase transitions [the vertical dashed lines in panels (d) and (e)].The variations of the spectral weights are consistent with the expected valence changes of the Fe and Cl 2 An ions that accompany the two-step phase transition. [40]These integrated intensities can therefore be utilized as suitable probes for detecting the degree of CT between the Fe and Cl 2 An ions.
We performed time-resolved reflectivity measurements using an optical pump-probe technique to confirm the occurrence of a photoinduced CT-type phase transition in this MOF system.We used a pump light with a photon energy of either 0.73 or 0.8 eV (E || chain) and a 90 fs pulse width (see Figure S3, Supporting Information Note S3 for details).We monitored the relative reflectivity change (∆R/R) induced by photoexcitation of the CT band over a wide range of probe-photon energies (0.1-3.0 eV) at RT, i.e., in the LT phase.Typical temporal profiles of ∆R/R show a sudden change of ∆R/R just after the photoexcitation (Figure proves).The value of ∆R/R relaxes to zero within 1 ms, and then the observed transient states are meta-stable states with a lifetime much shorter than 1 ms. Figure 3a shows the isothermal static reflectivity spectra measured at the temperatures 296, 340, and 380 K-corresponding to the LT, IM, and HT phases, respectively-and with E || chain.The broad photon-energy range covers the range of intramolecular transitions (2.0-3.0 eV), the CT transition (0.5-1.5 eV), and molecular vibrations (mid-IR, 0.15-0.25 eV).These three photonenergy ranges are labeled as green-, brown-, and yellow-shaded regions, respectively.(The spectrum at 296 K is almost the same as that with E || chain shown in Figure 2a).The differences (∆R/R) in the isothermal reflectivity spectra between 296 and 340 K and between 296 and 380 K are shown as orange and green lines, respectively, in Figure 3b.They correspond to the ∆R/R spectra expected in thermally induced phase changes from the LT phase to the IM and HT phases, respectively.The ∆R/R spectra observed just after photoexcitation (0 ps, open black circles) and at 10 ps (open pink circles) with a pump fluence of 1.0 mJ cm −2 are also plotted in Figure 3b.We observed considerable differences between the spectral shapes of the photoinduced ∆R/R (open black and red circles) and the thermally induced values (the orange and green lines) both in the region of the CT transition (the brown-shaded area) and in the mid-IR region (the yellowshaded area).For simplicity, we next consider the transient reflectivity spectra in these three energy ranges separately and hereafter discuss in detail what we observed in each energy range in order to obtain further insight into the effect of photoexcitation in the present MOF.
Figure 3c depicts the ∆R/R spectra in the range 2.0-3.0 eV, in which intramolecular transitions associated with the valence of the Cl 2 An ion are active.The open black circles show the ∆R/R spectrum with a 0 ps delay at RT.To evaluate the photoinduced ∆R/R spectra quantitatively, it is necessary to consider the spatial distribution of the density of photoinduced states because the intensity of the pump light weakens as it penetrates deeper into the crystal from the light-irradiated sample surface.We therefore adopted the method of analysis described in refs.[17] and [44].We assumed a homogeneous distribution of the density of photoinduced states at the depth z below the sample surface, with an exponential decay type z-dependence (shown schematically in Figure S5, Note S5, Supporting Information).Then we constructed the dielectric function [()] at depth z as a uniform mixture of the dielectric functions of the initial [ ini ()] and photoinduced [ PI ()] states.Given these assumptions, () at depth z is given by: where  is the frequency of the light, and  ini () is () at 296 K.For discussing the intramolecular transition range 2.0-3.0 eV, we assumed  PI () to be () for the HT phase at 380 K [ HT ()], since it is reasonable to assume that the photoinduced CT excitation modulates the valence of the Cl 2 An 3-ions (in the initial LT phase) into that of the Cl 2 An 2− ions (in the HT phase).The quantity r 0 (0 < r 0 < 1) represents the yield of the observed photoinduced state at the surface.The value of d is the specific length of the exponential-decay type distribution of photoinduced states, which we assume to be equal to the penetration depth of the pump light, ≈162 nm.We determined this value from the KK analysis using the static reflectivity at 296 K.The reflectivity spectrum of the transient state can be calculated by using the transfer-matrix method [45] for the accumulation of thin films with uniform (, z).As indicated by the thick purple line in Figure 3c, the calculated ∆R/R spectrum reproduces the experimentally observed spectrum well.We determined the value of r 0 to be 0.48 ± 0.002 from a least-squares fit to the 0 ps spectra.This value implies that the photoconversion efficiency corresponds to about two DA pairs for each photon (see Supporting Information Note S7).Hence, the initial process is not restricted to a single DA pair because the cooperative valence instability involves multiple DA pairs.Since r 0 can be evaluated at any delay time using the same method of analysis (see Supporting Information Note S6), we obtained the time dependence of r 0 shown in Figure 3d, where we performed the fitting analysis using Equation (2): + a 2 exp This result revealed a combination of two types of relaxation, a very fast relaxation on a timescale of 140 ± 32 fs ( 1 ) and a slow relaxation on a timescale of 3.40 ± 0.24 ps ( 2 ).In the following, we focus our discussion solely on the photoinduced state observed just after photoexcitation because the spectral shapes of ∆R/R for the photoinduced state in the observed energy range did not show any apparent change during ≈15 ps, and only the density change can explain the delay-time dependence (as discussed in Supporting Information, Notes S6, S8, and S9) As discussed in the previous paragraph, the observed spectral change at ultrafast speed in the intramolecular-transition range can be explained by an efficient photoinduced valence change from the LT phase (Cl 2 An 3-) to the HT phase (Cl 2 An 2− ).This raises the naïve question of whether the photoinduced state resembles a pure HT phase.To obtain an adequate answer to this question, it is essential to evaluate quantitatively the photoinduced ∆R/R spectra in the CT transition region as well as in the mid-IR region, as we discuss in the following sections.
We first emphasize that, as shown in Figure 3b, the shape of the photoinduced ∆R/R transient spectrum in the CT-transition energy range at RT is different from that of the thermally induced spectrum.Figure 4a depicts the variation of R over the photonenergy range 0.5-1.5 eV, which corresponds to the CT transition energy at 0 ps delay (black open circles).Note that a simulation of the R variation using Equation ( 1) and assuming that the photoinduced state is the HT phase (the purple line in Figure 4a) fails to reproduce the experimental data, even though utilizing the same parameter values for r 0 and d that obtained from the analysis for the intramolecular transition energy range (0 ps).This spectral discrepancy is due to the appearance of a shoulder located on the higher-energy side of the reflectivity spectrum, between 0.8 and 1.2 eV (indicated by the solid pink fill in Figure 4a).We observed this unexpected spectral modification not only immediately after photoexcitation but even after a 10 ps delay (see Figure S8, Supporting Information Note S8).To reproduce this unexpected spectral feature, we added one Lorentzian oscillator  l () to  PI () in Equation ( 1) to represent another absorption feature, i.e.,  PI () =  HT () +  l (), with where  0 , S 0 , and  0 are the resonance frequency, oscillator strength, and damping constant, respectively, of the Lorentzian oscillator.We determined the optimal values of these parameters from a least-squares fit to the data at 0 ps.The resulting calculated photoinduced reflectivity spectrum (the thick red line in Figure 4a) matches the experimental data (open black circles) well.The simulated () spectrum of the photoinduced state based on this analysis is shown as the thick red curve in Figure 4b.This feature looks like a superposition of the spectrum of the HT phase (the dashed green line) and the spectrum of the new absorption feature (the dashed orange line).The peak position of the new absorption is ≈1.02 ± 0.003 eV-with  0 = 0.22 ± 0.001 eV and S 0 = 0.20 ± 0.001 eV 2 -that is significantly higher than the energy of the CT transition for either the LT or the HT phases.It is therefore reasonable to assume that the new absorption band can be assigned to a CT transition between D and A, i.e., to a double-peaked structure produced by the splitting of the CT transition.We discuss the possible origin of this splitting later in connection with the photoinduced spectral change in the mid-IR region, i.e., in the molecular-vibration region.The transient spectra in the mid-IR energy range cannot be explained simply by photoconversion from the LT phase to the HT phase, as shown in Figure 3b.The amplitude of the photoinduced ∆R/R is smaller than that of the thermally induced ∆R/R in this energy range.Since there are many spiky structures due to intramolecular vibrational modes, the energy resolution of the data in Figure 3b, where the wavelength resolu-tion is ≈150 cm −1 , is not sufficient for this analysis.Therefore, we measured the ∆R/R spectra much more precisely to obtain better insight into the spectral nature of the photoinduced state in this range.In the setup used for this experiment, we sent the light reflected from the sample directly into a monochromator to obtain high-energy-resolution spectra with a wavelength resolution of ≈6 cm −1 , so that the intramolecular vibration modes can be distinguished, even at the expense of a temporal resolution degraded to the order of 1 ps (see Supporting Information, Note S10). [46]The static reflectivity spectra in the intramolecularvibration range 1200-2000 cm −1 at 296, 340, and 380 K with E || chain, are shown in Figure 5a.The transient ∆R/R spectra with E || chain (black line) and E ⟘ chain (brown line) at 0 ps and RT are shown in Figure 5b.These spectra imply that the photoinduced spectral change is strongly anisotropic.The spectrum obtained with E || chain produced a broad dispersion-type spectral shape overlapped with spiky structures due to the intramolecular vibrational modes, while the spectrum obtained with E ⟘ chain showed essentially no change.
As a first step in analyzing these results, we calculated the photoinduced ∆R/R spectrum with E || chain using the r 0 and d values obtained in the previous analyses and assuming the photoexcited phase to be the HT phase.This result is represented by the thin-dashed purple curve in Figure 5b.There is an obvious discrepancy between this simulated curve and the experimental spectrum (the black line in Figure 5b).This result supports the hypothesis that the photoinduced state differs from the thermally induced state.To test this hypothesis, we next included the additional Lorentzian oscillator absorption not observed in the HT phase as a second step in the simulation.We performed this spectral simulation successfully using Equations ( 1) and (3), and the calculated ∆R/R spectrum that includes the additional oscillator (the thick red line in Figure 5b) reproduces the curve of the experimental data considerably better, except for the spiky structures due to intramolecular vibrations.Using the least-squares fit, we determined the optimal values of the parameters for the added new absorption in the transient state to be  0 = 1570 ± 9.06 cm −1 , S 0 = 72.60 ± 4.34 cm −2 , and  0 = 403.28± 5.06 cm −1 .The difference in the spiky structure due to the intramolecular vibration modes-especially those found below 1500 cm −1 -may reflect a difference in the molecular structure between the HT phase and the photoinduced state.Figure 5c shows the simulated () spectrum for the photoinduced state (the thick red line) including the additional new absorption (the dashed orange curve).The peak position of the new absorption is higher than those of any of the peaks assigned to intramolecular vibrational modes, which are observed as spiky peaked structures in the spectra at 296 and 380 K (the thin blue and green dashed lines).
To explore the origin of the new absorption in the mid-IR energy range, we compared the transient () spectrum with Raman spectra (Figure 5d,e), in which "xx" and "xy" mean that the polarization directions of the analyzer are respectively parallel and perpendicular to the polarization of the input light (E || chain).Figure 5d shows these spectra for the HT phase (380 K), and Figure 5e shows the "xx" alignment spectra measured at the temperatures 293, 340, and 380 K.The Raman spectra in the HT phase exhibit a characteristic mode at 1596 cm −1 , which is assigned to a C-O stretching mode [47] with A g symmetry, as judged from the intensity ratio between "xx" and "xy"; [48] the inset of Figure 5d depicts a schematic illustration of the atomic motions in this mode (the red arrows). [42]Interestingly, we did not observe this C─O stretching mode at 1596 cm −1 in the LT phase (Figure 5e), indicating that this mode is related to the Cl 2 An 2− ions.The peak position of this mode is very close to that of the new photoinduced absorption.This suggests that the Raman mode became IR-active in the photoinduced state.Considering the strongly anisotropic nature of the broad structure of ∆R/R observed and the A g symmetry of the Raman mode at 1596 cm −1 , we conclude that the newly observed absorption band at ≈1600 cm −1 in the photoinduced state can be attributed to an electron-molecular-vibration (e-MV) coupled mode.The e-MV coupled modes in DA complexes have been intensively studied in TTF-CA, [49] in which the CT energy between the D and A molecules becomes asymmetric because of the disappearance of the inversion centers in these molecules due to the dimerizationtype lattice distortion.In other words, a totally symmetric A g mode can become an IR-active mode in this situation.A characteristic of the e-MV coupled mode is its strong oscillator strength borrowed from the CT transition. [50]The A g mode inherent to a single Cl 2 An 2− ion can thus become the origin of the observed broad photoinduced IR band based on the same mechanism.That is, local inversion-symmetry breaking due to lattice modulation in the photoinduced state may be able to induce the emergence of a new IR absorption band in the present system.Two possible types of structural modulations in the photoinduced state are shown schematically in Figure 6c,d.Model 1 is a DA- type dimer modulation similar to that in the ferroelectric I phase of TTF-CA, and model 2 represents an ADA-type trimer modulation.In model 1, the Fe-related e-MV coupled mode may be active.Therefore, observations of the Fe-related e-MV coupled modes associated with the symmetry around Fe ions are key for determining that mechanism is the actual origin of this state.Regrettably, we have not yet been able to carry out such experiments because the Fe-related e-MV coupled modes are characterized by wavenumbers-e.g., ≈600 cm −1 -that are lower than those in our measurable range.
However, the lattice-modulation scenario also can be explained by another piece of evidence that is due to the appearance of the new absorption band in the CT transition region described above: the double-peaked CT transitions tentatively associated with the disappearance of the inversion center from the Cl 2 An ion in the smallest DAD model (see Supporting Information, Note S11).Indeed, similar observations of double-peaked structures have been reported in segregated [51] and mixed [52] stacked DA complexes with dimerization.In refs.[50] and [52], an asymmetric CT energy explained a double-peaked structure due to the disappearance of the inversion centers on the constituent molecules.
In addition to CT, the presence of lattice modulation in the 1D chain self-consistently explains the observed photoinduced spectral change in this system over the wide range of photon energies extending from the region related to intramolecular transitions to the mid-IR region.The remaining issue in this discussion is a concern about the speed of the lattice deformation.From the temporal dependence of ∆R/R observed in the CTtransition energy range, the structural change occurred immediately after photoexcitation, i.e., within the ≈110 fs time resolution of our study (see Supporting Information, Note S4).This timescale is much shorter than that for a typical change in the lattice structure, which, for example, was found to be ≈600 fs in TTF-CA. [18]The TTF-CA system is constructed from - stacking between D and A, i.e., it can be described approximately as a "soft bonding" between D and A. In contrast, the present MOF system is composed of relatively "rigid bonds" between D and A, so the structural changes may be related primarily to the C─O stretching mode (≈1600 cm −1 ) and the Fe-O stretching mode (≈600 cm −1 ), [53] which have vibrational periods of ≈20 or 60 fs, respectively.Thus, an ultrafast structural change within our time resolution (110 fs) may be reasonable.
Finally, the effect of the NPr 4 + cations should be discussed because we neglected them during the above discussions.The initial photoexcitation occurs inside the 2D honeycomb layer, and then the successive dynamics are induced by energy transfer from the locally excited state.The time scale of the energy transfer should be evaluated by the inverse of the energy scale of the interaction.Then we believe the dynamics related to the NPr 4 + cation should be slower than the dynamics within the layer and can be neglected in the time scale we discussed here.

Conclusions
We have performed ultrafast time-resolved photoexcitation experiments using a well-shaped crystal sample of the DA-type layered MOF (NPr 4 ) 2 [Fe 2 (Cl 2 An) 3 ].We found that it exhibits an efficient photoinduced CT conversion between Fe and Cl 2 An ions at RT by virtue of the cooperative valence instability.Two absorption bands-at the higher-energy side (≈1.02 eV) of the main CT band and at ≈1570 cm −1 in the mid-IR energy range-emerged immediately after photoexcitation.The emergence of these two new peaks, which can only be observed in the photoinduced transient state, implies the generation of a photoinduced, thermally hidden state.That is, this phenomenon is clearly distinct from what we see in the thermally induced CT states.The hidden state may be produced by unexpected lattice modulations along the chain direction, with accompanying symmetry breaking of the local inversion centers on the Cl 2 An ions.Two types of lattice modulations are expected, which may demonstrate a change from the SCM state to a paramagnetic and ferroelectric state similar to that of TTF-CA (model 1) or to a paramagnetic state (model 2).It remains for the future to confirm whether or not this structural change involves long-range ordering as well as longrange charge-ordering.This will require lattice-structure measurements and magneto-optical measurements resolved on fs timescales.Ultrafast photocontrol of the charge, spin, and lattice structures in MOF systems may be quite challenging, but they will provide opportunities to develop ultrafast phase-switching systems.We also mention that the present study can be considered as a first step toward the development of high-speed photoresponsive MOFs with magnetic natures.
Static-Reflectivity Measurements: Static optical-reflectivity spectra were measured using a Fourier-type IR spectrometer (Thermo 6700) with a Cassegrain microscope for the mid-IR range (0.08-1.0 eV).The reflectivity spectra in the near-IR to the visible range (0.9-3.0 eV) were measured using a homemade microscope-spectroscopy system with a diffraction-type monochromator (Bunkokeiki M25) and adequate photodiode detectors.The temperature-dependent static optical spectra were measured using a commercial cryostat with BaF 2 or quartz windows, Microstat He (Oxford Instrument).
Raman-Spectra Measurements: The Raman spectra were measured using a JASCO NRS-4500.The single-crystal sample was placed on a copper plate, and a film heater under it controlled its temperature.The sample was irradiated with a 532 nm Raman excitation laser.We measured the Raman spectra with the polarization directions of the analyzer oriented in the x-and y-directions of the incident and scattered light.
Femtosecond Pump-Probe Reflection Spectroscopy: We measured the transient reflectivity change (∆R/R) using a pump-probe technique.A Ti:sapphire regenerative amplifier system was used (center wavelength, 792 nm; pulse width, 90 fs; repetition rate, 1 kHz; Spectra-Physics Solstice Ace) as the light source.The optical-pulse output from the amplifier system was divided into pump and probe pulses using a beam-splitter plate.The pump pulse delivered from the optical parametric amplifier (OPA: Light conversion, TOPAS) was either 0.73 or 0.8 eV, and we tuned the repetition rate to half of the fundamental frequency (500 Hz) using an optical chopper (New focus 3501).We converted the probe pulse into the 0.12-3.0eV energy range using frequency-mixing processes.Here, the OPA was used either in the range 0.5-0.7 eV (idler light) or the range 0.8-1.0eV (signal light).We employed differential frequency generation between the signal and idler light in the range 0.12-0.4eV, second-harmonic generation of the idler light in the range 1.1-1.5 eV, signal light in the range 1.6-2.1 eV, and fourth-harmonic generation of the idler light in the range 2.2-3.0 eV.The widths of the pump and probe pulses depend on their energies; a typical probe-pulse width was 150 fs.∆R/R was measured using MCT (HgCdTe) (0.1-0.5 eV), InGaAs (0.51-1.1 eV), Si (1.1-2.5 eV), and GaP (2.5-3.0 eV) detectors in the photon-energy ranges of the probe light.The signals from the detectors were collected using a gated integrator (Stanford research, SR250) and an analog-to-digital converter (NI, NI-6143), which were synchronized with the frequencies of the optical chopper and the Ti:sapphire regenerative amplifier.The samples were maintained at RT, with both pump, and probe pulses entering the sample at near-normal incidence.To measure time-resolved mid-IR ∆R/R (1200-2000 cm −1 ), the probe light reflected from the sample was detected using the MCT detectors after passing it through a monochromator (Bunkokeiki M10).The energy resolution of this system was ≈4 cm −1 .
Theoretical Calculations: The DFT calculations were performed using Gaussian16 [54] with the B3LYP functional and employing the LANL2DZ basis set for the Fe atoms and the 6-31G(d) basis set for the C, Cl, and O atoms.Details of the method of calculation were provided in Supporting Information (Supporting Note S1)

Figure 1 .
Figure 1.a) Structure of the D 2 A 3 -type honeycomb layer in (NPr 4 ) 2 [Fe 2 (Cl 2 An) 3 ] (where D = donor and A = acceptor).The portion surrounded by the thick dashed line is schematically illustrated in panel b.b) Schematic charge and spin patterns in the low-temperature (LT), intermediate-temperature (IM), and high-temperature (HT) phases.Two possibilities are proposed for the IM phase, a delocalized state (IM d ) and an ordered state (IM o).[40]The Fe ions are depicted as colored circles, and the Cl 2 An ions are represented by thick colored lines.The arrows inside the circles and lines represent the spins.

Figure 2 .
Figure 2. a) Steady-state reflectivity (R) spectra of a single crystal of (NPr 4 ) 2 [Fe 2 (Cl 2 An) 3 ] polarized with E || chain (blue line) and E ⟘ chain (orange line) at room temperature (300 K: RT).b) Optical conductivity [()] spectra of the single crystal with E || chain (blue line) and E ⟘ chain (orange line) at RT.The insets show schematic representations of the assignments of the absorption peaks.c) Temperature dependence of the () spectra with E || chain.d) Temperature dependence of the integrated () between 0.7 and 0.9 eV.e) Temperature Dependence of the integrated () between 2.5 and 2.7 eV.The dashed lines represent the transition temperatures T 1/2(1) (317 K) and T 1/2(2) (354 K) of the charge-transfer (CT)-type phase transitions based on previous magnetic and structural studies.[40]

7
Figure 2. a) Steady-state reflectivity (R) spectra of a single crystal of (NPr 4 ) 2 [Fe 2 (Cl 2 An) 3 ] polarized with E || chain (blue line) and E ⟘ chain (orange line) at room temperature (300 K: RT).b) Optical conductivity [()] spectra of the single crystal with E || chain (blue line) and E ⟘ chain (orange line) at RT.The insets show schematic representations of the assignments of the absorption peaks.c) Temperature dependence of the () spectra with E || chain.d) Temperature dependence of the integrated () between 0.7 and 0.9 eV.e) Temperature Dependence of the integrated () between 2.5 and 2.7 eV.The dashed lines represent the transition temperatures T 1/2(1) (317 K) and T 1/2(2) (354 K) of the charge-transfer (CT)-type phase transitions based on previous magnetic and structural studies.[40]

Figure 3 .
Figure 3. a) Temperature-dependent R spectra of a single crystal of (NPr 4 ) 2 [Fe 2 (Cl 2 An) 3 ] with E || chain.b) Transient reflectivity (∆R/R) spectra with E || chain for pump and probe pulses at ∆t = 0 (black circles) and 10 ps (pink circles) at RT (LT phase).The fluence of the pump pulse is I ex = 1.0 mJ cm −2 .The ∆R/R spectrum with E || chain between 296 K (the LT phase) and 380 K (the HT phase) is shown as a green line and that between 296 and 340 K (the IM phase) is shown as an orange line.The energy ranges corresponding to intramolecular transitions (2.0-3.0 eV), the CT transition (0.5-1.5 eV), and molecular vibrations (mid-IR, 0.15-0.25 eV) are designated by green-, brown-, and yellow-shaded regions, respectively.c) ∆R/R in the intramolecular transition region (≈2.6 eV).The black circles represent the photoinduced values of ∆R/R at 0 ps and RT.The purple line shows the values of ∆R/R at 0 ps calculated by considering the distribution of the photoinduced state (see text).In this calculation, the photoinduced state is assumed to be the HT phase (380 K).The parameters are d = 162 nm and r 0 = 0.48.d) The black circles show the values of r 0 at each delay time, along with a fit to the data (the solid red line).The temporal dependences of the three terms in Equation (2) are shown as dashed lines.

Figure 4 .
Figure 4. a) Reflectance (R) spectrum of the photoinduced state of (NPr 4 ) 2 [Fe 2 (Cl 2 An) 3 ], obtained using a pump with E || chain and I ex = 1.0 mJ cm −2 , at ∆t = 0 and RT around the CT-transition range (0.5-1.5 eV).The observed spectrum is plotted as black circles.Calculated R spectra at 0 ps are also shown; they consider the photoinduced state either to be the 380 K (HT phase; the purple line) or to be the sum of the 380 K (HT phase) and a Lorentzian oscillator (the thick red line).Details of the calculation are presented in the text.The solid pink fill shows the difference between the purple and red lines.The dashed lines represent the R spectra of the LT (296 K) and HT (380 K) phases.b) Calculated () spectra of the photoinduced state (thick red line), 296 K (dashed blue line), and 380 K (dashed green line).The dashed orange line shows the new photoinduced-absorption peak.The resonance frequency of the Lorentzian oscillator is indicated by purple arrows.

Figure 5 .
Figure 5. a) High-resolution, temperature-dependent R spectra of (NPr 4 ) 2 [Fe 2 (Cl 2 An) 3 ] in the mid-IR range (1200-2000 cm −1 ) with E || chain.b) ∆R/R with E || chain (black line) and E ⟘ chain (brown line) probes for the pump with E || chain and I ex = 1.0 mJ cm −2 at ∆t = 0 ps and RT.The R spectra at 0 ps, calculated using d = 162 nm and r 0 = 0.48, consider the photoinduced state to be either the 380 K (HT) phase (the purple dashed line) or the sum of the 380 K (HT) phase and a Lorentzian oscillator (the thick red line).c) The calculated () spectrum just after photoexcitation is shown as a thick red line.The photoinduced absorption in the model is also shown, as an orange dashed line.The blue and green dashed lines are the observed () spectra of the LT (296 K) and HT (380 K) phases, respectively.d) Polarized Raman spectra with "xx" and "xy" polarization geometries (see main text) at 380 K.The vertical dashed lines represent the center wavenumber of the oscillator and of the Raman spectral peak.e) Raman spectra at 293 K (blue line), 340 K (orange line), and 380 K (green line) with "xx" polarization geometry.

Figure 6 .
Figure 6.a) Schematic representation of the charge and spin patterns in the LT phase of the 2D-layered system [Fe 2 (Cl 2 An) 3 ] 2− .The donor units, Fe 3+ and Fe 2+ , are represented as pink, and brown circles, respectively, whereas the acceptor units, Cl 2 An 3− and Cl 2 An 2− , are represented as blue, and green cylinders, respectively.b) Schematic 1D chain with DA units before photoexcitation.(c) and (d) Candidates for the photoinduced state at 0 ps.c) DA-type dimer modulation between DA units, similar to that in the ferroelectric TTF-CA.d) ADA-type trimer modulation.