Exploring the light-induced dynamics in solvated metallogrid complexes with femtosecond pulses across the electromagnetic spectrum

Photoinduced spin-switching (PSS) in solvated d⁴–d⁷ transition metal (TM) complexes is a fundamental molecular process, which is driving a growing number of advanced technologies, including ion sensing, photocatalysis, and drug delivery (Kumar and Ruben, 2017; Halcrow, 2013). The vast majority of solvated TM compounds that operate through efficient PSS are mononuclear Fe^II complexes that switch from the low-spin (LS) (t_2g)^6 1A₁ state to the high-spin (HS) (t_2g)⁴(e_g)^2 5T₁ state upon light irradiation. Despite definite success in laboratory settings, the efficiencies reported so far cannot yet meet the demands of chemical applications in real-life because of the limitations associated with rather restricted intramolecular distortions and photodecomposition. The prospect of overcoming these hurdles is currently spurring a significant synthetic effort toward oligonuclear complexes, since they often interconvert between stable and adaptable conformers that respond to the changes in the surroundings (Bodenthin et al., 2009; Halder, 2002; and Wang et al., 2013). For example, homometallic, heterometallic, and mixed valence [2 × 2] grid complexes, which are actively pursued for high-density data storage and miniaturization in the field of molecular electronics, have recently emerged as self-assembling nanosized platforms that can sustain multiple spin-switching (Schäfer et al., 2016; Matsumoto et al., 2014; Schneider et al., 2013; Schneider et al., 2010; Wu et al., 2009; and Nihei et al., 2005), charge transfer (Ohkoshi et al., 2011; Liu et al., 2010; and Zhang et al., 2010), and redox modulation (Shen et al., 2015; Wang et al., 2013; Schneider et al., 2010; Ruben et al., 2004; and Uppadine et al., 2004). The optimization of these properties could ultimately produce photosensitizers, reactive centers, or catalysts achieving superior operando performances. However, unlike for the single-crystal and film morphologies, functionalities based on manipulating the spin degrees of freedom with visible photons have not been demonstrated to date for solvated metallogrid complexes (Kumar and Ruben, 2017). This status can be ascribed to the fact that their basic photophysical properties have rarely been investigated in the solution phase. A fundamental difficulty lies in tracing the detailed map of the competing deactivation pathways that are mediated by intersystem crossing (ISC), internal conversion (IC), internal vibrational relaxation (IVR), and vibrational cooling (VC) to the solvent. Consequently, establishing clear correlations between the nuclearity of the large molecular assembly, the connectivity of the multi-site ligand scaffold, and the intrinsic rate of the PSS process requires combining the quantitative results of several ultrafast spectroscopic measurements. From the accumulated experience about the PSS in solvated mononuclear Fe^II complexes, transient optical absorption (TOA) spectroscopy in the UV–visible range can monitor the population of the bright states reached through dipole-allowed transitions, such as the metal-to-ligand-charge-transfer (MLCT) state, which subsequently leads to the formation of the lowest-lying quintet HS state. The technique also delivers the lifetime of the metastable state through the kinetics of the ground-state bleach recovery (i.e., HS → LS transition). In few cases, transient infrared absorption (TIA) spectroscopy (Wolf et al., 2008; Zerdane et al., 2018a; Collet et al., 2019) and transient Raman spectroscopy (Smeigh et al., 2008 and Collet et al., 2019) have been used to follow the vibrational relaxation toward the dark thermalized HS state in related mononuclear Fe^II complexes. Utilizing the high photon energies delivered by the X-ray Free Electron Laser (XFEL) facilities, transient x-ray emission (TXE) spectroscopies of the Fe^II Kα lines (2p_3/2,1/2 → 1s) and Fe^II Kβ lines (3p_3/2,1/2 → 1s) have been employed to uncover the spin multiplicity of the optically dark metal-centered (MC) states involved in the formation of the HS state (Zhang et al., 2014; Zhang and Gaffney, 2015) and, more generally, to map out the photoinduced spin-dynamics in homoleptic and heteroleptic mononuclear Fe^II complexes (Zhang et al., 2017; Kjær et al., 2018; 2019; and Kunnus et al., 2020). The present work reports the first steps toward obtaining exhaustive characterizations of the photoinduced dynamics in oligonuclear TM complexes using mid-IR, UV–vis, and x-ray femtosecond pulses. The reversible photocycle of a solvated pyrazolate-bridged [2 × 2] Fe^II metallogrid complex is probed with femtosecond TOA and TIA spectroscopies along with TXE spectroscopy at the FXE instrument (Galler et al., 2019; Khakhulin et al., 2020) of the European XFEL (EuXFEL) facility (Altarelli, 2015; Tschentscher et al., 2017). Correlating the relevant time scales associated with the coupled electronic, vibrational, and spin dynamics allows building a detailed picture of the PSS process in this important family of oligonuclear TM complexes that will guide the development of their light-driven applications based on spin-switching in the solution phase.

The [2 × 2] metallogrid complex [Fe₄(L)₄](BF₄)₄ (noted Fe4) is formed around the pyrazole-based compartmental ligand bis(2,2′-bipyridine)-6,6′-(3,5-pyrazol) [noted HL, see Fig. 1(a)] (van der Vlugt et al., 2008). The coordination sphere of each Fe^II center consists of two quasi-orthogonal terpyridine-like binding units, each belonging to two compartmental deprotonated ligands L. The molecular structure of Fe4 with the $BF4−$ counter-anion determined through x-ray crystallography depends on the temperature (Schneider et al., 2010). At 133 K, a single Fe^II center is in the (t_2g)⁶ LS configuration, while three Fe^II centers are in the (t_2g)⁴(e_g)² HS configurations [Fig. 1(b)]. The average metal–ligand bond lengths, R_av, are 1.99 Å for the LS center and 2.17 Å, 2.18 Å, and 2.19 Å for the 3 HS centers. At 233 K, the Fe^II center in the LS state undergoes a thermal transition to the HS state [Fig. 1(c)]. The R_av are then 2.16 Å for the switching center and 2.18 Å for the other 3 HS centers. The LS → HS transition at room temperature is also accompanied by a greater degree of angular distortion (Schneider et al., 2010). Figures 1(d) and 1(e) provide the simplified sketches of the molecular structures that will be employed to illustrate the changes in spin speciation upon solvation and photoexcitation.

Figure 2(a) shows the Mössbauer measurements on Fe4 in frozen in MeCN at 80 K, which indicates a spin-state ratio LS:HS of 52:48. For the magnetic characterization, dimethylformamide (DMF) was chosen as a solvent since the solubility of Fe4 in MeCN is not sufficiently high to allow meaningful superconducting quantum interference device (SQUID) magnetometry. Figure 2(b) displays the magnetic measurements for the Fe4 complex dissolved in DMF as a function of temperature from 2 K to 350 K. Over the range 40–200 K, the [2LS–2HS] species is exclusively present in solution. As the temperature rises, the population of HS centers increases. Using χ_MT ([1LS–3HS]) = 9.3 cm³ mol⁻¹ K (i.e., 100% of [1LS–3HS]) and χ_MT ([2LS–2HS]) = 6.2 cm³ mol⁻¹ K (i.e., 0% of [1LS–3HS]), the experimental value χ_MT of 8.09 cm³ mol⁻¹ K obtained at 295 K gives 39:61 for the ratio [2LS–2HS]:[1LS–3HS], which translates into an overall LS:HS ratio of 35:65% ± 5%.

X-ray emission spectroscopy (XES) measurements of the Kα_1,2 and Kβ lines were then conducted in order to establish the spin speciation in MeCN at room temperature. When the orbital angular momentum of the valence shell is equal to zero and when relativistic interactions can be neglected, the multiplet theory predicts a linear relationship between the (2p, 3d) exchange energy splitting and the number of unpaired electrons (Glatzel). As a result, the width of the (2p_3/2 → 1s) Kα₁ line reflects the spin at the 3d metal center. Figure 3(a) shows the Kα_1,2 (2p_3/2,1/2 → 1s) lines for Fe4 in MeCN after normalization by the total area. The asymmetry in the Kα₁ line shape is related to the significant fraction of the HS state in the solvated sample. Figure 3(b) shows the Kβ (3p_3/2,1/2 → 1s) lines for the Fe4 metallogrid complex in MeCN after normalization by the area. The accurate determination of the spin-state from these lines is achieved through monitoring the satellite feature at 7045 eV. Spectral decomposition as a linear combination fit (LCF) based on the solvated mononuclear LS and HS references [Fe^II(bpy)₃](PF₆)₂ (where bpy = 2,2′-bipyridine) in MeCN and [Fe^II(H₂O)₆](BF₄)₂ in H₂O yields 37% ± 3% of LS and 63% ± 3% of HS for Fe4 in MeCN. The spin state composition is illustrated with the sketch in Fig. 3(c). The Kβ profiles for the mononuclear reference complexes are reported in Sec. S.I.1. The comparison between the spin distributions in the solution, powder, and single-crystalline phases summarized in Table S.I.2 shows that the Fe4 metallogrid complex is strongly affected by the environment.

The UV–visible absorption spectrum of Fe4 in MeCN at room temperature is displayed in Fig. 4. The line shape is consistent with the one reported in the literature (Schneider et al., 2010). The oscillator strength below 350 nm is attributed to overlapping ligand-based transitions, while the broad band spanning the 450–700 nm range is assigned to a metal-to-ligand-charge-transfer (MLCT) transition of mixed singlet–triplet character. The values of the extinction coefficient are given in Sec. S.I.3 for the selected wavelengths. Assuming that the individual Fe^II centers are only weakly electronically coupled, they should retain their isolated properties once they are linked. Therefore, the absorbance of the [2LS–2HS] species should be around twice that of the [1LS–3HS] species in the visible range. Temperature-dependent measurements show that the oscillator strength is due to the LS units and that a thermalized HS state does not contribute to the absorbance in this spectral region (see Sec. S.I.4).

Figure 5(a) presents the photoinduced dynamics following the femtosecond laser excitation of Fe4 in MeCN at 385 nm at room temperature for a fluence of about 0.3–1.1 mJ/cm². The setup is described in Sec. S.I.5. The singular value decomposition (SVD) analysis based on a simple sequential model identifies three clear components with distinct time constants (Sec. S.I.6). The decay associated spectra (DAS) are displayed in Fig. 5(b). The first component (red line) is formed instantaneously with a time constant τ₁ of about 200 fs, which is instrument-limited. It comprises a bleach signal (B1) and a broad excited state absorption (ESA) signal that cover the blue (A₁) and red (⁠$A1′$⁠) regions of DAS₁. The second component (blue line) that corresponds to the time constant τ₂ of 15 ps reflects the growth of the bleach (B₂) and a weaker ESA in the blue (A₂) and red (⁠$A2′$⁠) regions of DAS₂. The third component matches the negative of the absorption spectrum (B₃), which is shown as the gray trace in Fig. 5(b). It decays over a time scale exceeding the temporal window covered by the transient absorption setup employed for this measurement. Figure 5(c) shows the optical kinetics that track the temporal evolution of the ground state bleach at 460 nm and the excited manifold at 360 nm. The high-frequency oscillations in the first few picoseconds are presented in the inset. It should be noted that the oscillatory behavior is not observed in the 460 nm kinetics so that there is no impulsive excitation on the PES of the LS (Cannizzo et al., 2010). Fitting a double-exponential to the kinetics trace at 360 nm and subtracting it in order to extract the residual oscillations deliver a period of 600 ± 20 fs (Sec. S.I.7). The schematic in Fig. 5(d) illustrates the global PSS in Fe4 solvated in MeCN. The nanosecond kinetics following excitation at 400 nm are shown in Fig. 5(e). The single-exponential lifetime is fitted to 210 ± 5 ns. As determined with SQUID and steady-state XES measurements, the Fe4 metallogrid complex solvated in MeCN is a mixture of two molecular species, namely, [2LS-2HS] and [1LS-3HS] species at room temperature. The analysis of the transient optical absorption measurements in the UV–visible indicates that a single photoexcited intermediate state leads to the formation of a long-living metastable state. This transient species is characterized by its well-defined single-exponential behavior with a time constant of 15 ps. However, its exact nature cannot be established based on the TOA measurements alone because its spectral line shape is broad and featureless.

TXE measurements with femtosecond resolution were then performed in order to assign unambiguously the spin multiplicity of the intermediate excited states. The experimental setup and the data extraction procedures are described in Sec. S.I.8. Figures 6(a) and 6(b) show the transient Kα_1,2 and Kβ signals (black dot traces), respectively, at a time delay Δt = 2 ps following 400 nm excitation with a fluence of 55 mJ/cm² (after 3-point rebinning). Comparing with signals built from the mononuclear reference traces shows that the intermediate and metastable excited-states belong to the quintet manifold. Direct scaling to these reference signals (green line traces) delivers comparable excited state fractions of 17% ± 5% and 20% ± 8% from the Kα₁ and the Kβ measurements, respectively. The power dependency of the Kα₁ upon the incident fluence is given in Sec. S.I.9.

Figure 7 shows the Kα₁ kinetics acquired within 6400.2–6409.4 eV (see Sec. S.I.8). Fitting the rise by a Gaussian-broadened Heaviside function suggests a σ = 140 ± 70 fs, hence a full-width at half maximum (FWHM) of 330 ± 165 fs for the instrument response function (IRF), in agreement with published reports at the time of these XFEL experiments (Kirkwood et al., 2019) (see Sec. S.I.10). In other words, the quintet character is acquired on the sub-picosecond time scale.

Further information about the relaxation dynamics from the initial Franck–Condon state is obtained with TIA spectroscopy. The first study on the sub-picosecond time scale was performed on the two related mononuclear Fe^II complexes [Fe(btpa)]²⁺ and [Fe(b(bdpa))]²⁺ (where btpa = N,N,N′,N′-tetrakis(2-pyridylmethyl)-6,6′-bis(aminomethyl)-2,2′-bipyridine and b(bdpa) = N,N′-bis(benzyl)-N,N′-bis(2-pyridylmethyl-6,6′-bis(aminoethyl))-2,2′-bipyridine) possessing similar multidentate coordination spheres (Wolf et al., 2008). For these systems, the most pronounced differences were observed in the spectral range of 1000–1065 cm⁻¹. Comparison with density functional theory (DFT) calculations showed that the modes responsible for the transient signals in these systems have significant metal–ligand stretching character. Figure 8(a) shows the steady-state spectrum of Fe4 in MeCN-d3 and its evolution following photoexcitation at 385 nm in the 1400–1650 cm⁻¹ spectral range for a power of 0.4 mW. The steady-state spectrum resembles that of the closely related mononuclear complex [Fe(terpy)₂]²⁺ (see Sec. S.I.11). The transient signal is characterized by the quasi-instantaneous appearance of differential bands within the response function of the instrument. The spectral evolution is on the tens of ps time scale before reaching a line shape that remains constant over the covered temporal window. As for the UV visible data, SVD analysis can be performed with a sequential model and two single-exponential decays [Fig. 8(b)]. The first component appears within the response function of the instrument of 200 fs and displays a time constant of 14 ps. This matches the values obtained for the intermediate species monitored in the UV–vis region.

For a molecular assembly as large as the Fe4 metallogrid complex, calculating the IR frequencies was beyond the computational resources that could be accessed at the time of this study. However, useful insight can be obtained from simple considerations. The first component (green trace) matches very well the derivative (gray trace) of the GS spectrum (black trace), which is the signature of a global shift for all the frequencies to lower wavenumbers (i.e., to lower energies). This can be ascribed to a weakening of the elongated metal–ligand bonds, resulting in lower force constants. For the long-lived nanosecond component (purple trace), some of the initial frequencies gain intensities (features A, B, C, and D in the purple trace) while narrowing (features A and B in the purple trace). This finding follows the general trend observed in the photoexcited Fe^II mononuclear complexes, where the absorption cross section in the infrared (IR) is enhanced upon LS → HS transition (Wolf et al., 2008; Lawson Daku and Hauser, 2010; Lawson Daku, 2018; and Lawson Daku, 2019), associated with increased entropy.

The analysis of kinetics obtained from TOA, TXE, and TIA delivers several time scales that should be correlated in order to yield a detailed description of the PSS process within the Fe4 metallogrid complexes solvated in MeCN. The ultrafast optical kinetics in the UV–visible and IR range indicate that a well-defined photoexcited species, characterized by its specific single-exponential behavior, evolves with a time constant of tens of ps toward a metastable state that lives for hundreds of ns. The ultrafast Kα₁ x-ray kinetics show that this intermediate state possesses a quintet character. Photoexcitation of the solvated Fe4 does not appreciably change the vibrational spectrum in the IR region but induces a sub-ps weakening of all the force constants attributed to the population of the antibonding orbitals of e_g character in the HS manifold. The vibrational coherence observed during the single-exponential decay is particularly notable considering the fact that the initial sample is a mixture of two large molecular species. These findings suggest that the Fe^II LS centers in the two types of metallogrids can be independently photoexcited to HS states, as anticipated from an assembly that is weakly electronically coupled. DFT calculations using BLYP-d3 in the gas phase and with a COSMO model for solvation were performed in order to establish the structures and energetics of the [4LS–0HS], [3LS–1HS], [2LS–2HS], [1LS–3HS], and [0LS–4HS] species. The average metal–ligand bond lengths, R_av, for the four centers in each species are summarized in Sec. S.I.12. The gas phase structures are in good agreement with the ones obtained in previous computational studies conducted with a similar methodology (Zueva et al., 2011; Borshch and Zueva, 2012). The calculated R_av of the [1LS–3HS] and [0LS–4HS] species reproduce the tendency reported upon thermal switching in the crystallographic structures measured at 133 K and 233 K, considering the expected solid-state effects, e.g., packing forces. The energy diagram for the solvated species is displayed in Fig. 9, with [Fe(terpy)₂]²⁺ in MeCN for comparison (see Sec. S.I.12). The values of the energy gaps are given in Sec. S.I.12.

It should be noted here that state-of-the-art DFT methods are still generally failing to provide accurate predictions of the HS–LS energy difference, even for mononuclear transition metal complexes. In this work, the choice of BLYP-d3 was motivated by the fact that it belongs to the family of the so-called GGA functionals. This made it possible to carry out the efficient computational study of the large Fe4 grid targeted in this work. It is well-known that GGA functionals tend to overestimate the stability of the LS state with respect to the HS state in mononuclear Fe(II) complexes (Lawson et al., 2012). In addition, a purely electronic calculation does not account for the crucial vibrational and entropic contributions. For both reasons, the absolute energy differences obtained through the present DFT optimizations are not expected to be strictly comparable to the ones of relevance in the experiment, including kT at room temperature. However, intensive benchmarking has established that the relative energy spacings in closely related complexes are accurately captured by DFT calculations (e.g., Lawson Daku, 2019; Lawson Daku et al., 2005; Vargas et al., 2013; Phan et al., 2017; Bowman and Jakubikova, 2012; and Petzold et al., 2017). This argument can be used to explain certain trends in physicochemical observables, e.g., here the lifetime of the photoexcited [1LS-3HS] and [0LS–4HS] in Fe4 and of the HS in [Fe(terpy)₂]²⁺. The rates of HS → LS transitions vary in accordance with the inverse energy gap law, which is holding in the solution phase (Hauser et al., 1991; Hauser, 1995; 2004; and Liu et al., 2017). Since ΔE₀([1LS–3HS] − [2LS–2HS]) = 0.9812 eV and ΔE₀([0LS–4HS] − [1LS–3HS]) = 0.9807 eV, the decay kinetics of the metastable HS states back to their respective ground state [1LS–3HS] → [2LS–2HS] and [0LS–4HS] → [1LS–3HS] are expected to be very similar.

Overall, the photoinduced dynamics do not exhibit any detectable dependency upon the spin composition of a given metallogrid. Therefore, they can be described in terms of intra-site transitions. Upon photoexcitation, a vibrational wavepacket is created in the quintet manifold on the few hundreds of fs by the coherent excitation of modes that are similar to the ones in the LS state, but downshifted in energy. The vibrationally hot HS state thermalizes within tens of ps to a long-living HS state. This time scale is typical for the equilibration of the photoinduced HS state in mononuclear Fe^II complexes with a flexible coordination sphere (Wolf et al., 2008). The reversible photocycle is summarized in Fig. 10.

The photoinduced dynamics in Fe4 can be compared to the ones reported for the closely related mononuclear complex [Fe(terpy)₂]²⁺ that also imposes a tridentate coordination sphere for the Fe^II center. In this complex, photoabsorption in the visible range creates a coherent superposition of excited vibrational modes (i.e., a nuclear wavepacket) in the quintet manifold. The vibrationally hot HS thermalizes and decays back to the LS ground state within a few nanoseconds in MeCN and water (Liu et al., 2017; Canton et al., 2014; Zhang et al., 2015; and Vankó et al., 2015). The prolonged lifetimes are in line with the larger energy gap ΔE₀([HS]–[LS]) calculated for the LS and HS state of [Fe(terpy)₂]²⁺ in D_2d and D₂ symmetries (see Fig. 9).

Besides energetic factors, further modulation of the HS → LS transition rate in [Fe(terpy)₂]²⁺ arises due to the deviation of the reaction coordinate from an isotropic single configuration coordinate. For this complex, the reaction coordinate cannot be approximated by a simple average Fe–N bond elongation, e.g., it is anisotropic (Nance et al., 2015; Canton et al., 2014; 2015; and Zhang et al., 2015). As revealed by powerful studies combining ligand design, crystallography, and modeling tools, such anisotropy is the key to stabilizing the HS state (Steinert et al., 2016; Stock et al., 2016; 2012; 2017a; 2017b; Phan et al., 2017; and Kroll et al., 2019). A significant increase in the HS lifetime of mononuclear complexes occurs when radial and angular motions are coupled during the rearrangement of the ligand scaffold (Marchivie et al., 2005; Buhks et al., 1980; Hauser et al., 1991; Hauser, 1995; 2004; and Gütlich et al., 1994), e.g., through a trigonal prismatic deformation known as the Bailar twist (Bailar et al., 1958), which was already proposed in the 1980s (Rodger and Johnson, 1988; Vanquickenborne and Pierloot, 1981; McCusker et al., 1992; and 1993). The present work illustrates how linking several metal centers into oligonuclear complexes can be developed as an efficient synthetic strategy to increase the lifetime of the photoinduced metastable spin-state, hence preventing thermal spin-scrambling. Comparing the lifetime observed for photoexcited Fe4 (∼210 ns) with the one reported for photoexcited Fe3 (∼123 ns) (Naumova et al., 2020) pinpoints the global nuclearity of the architecture as a tailoring parameter. Eventually, the balance between the connectivity and the rigidity of the molecular assembly can be further tuned toward promoting vibrational coherence while influencing the overall kinetics.

Finally, Fig. 10 suggests that the metallogrid architecture will be well-suited for creating non-Boltzmann distributions of spin multiplicities. In the particular case of Fe4, the decay [1LS–3HS] → [2LS–2HS] and [0LS–4HS] → [1LS–3HS] takes place on very similar time scales. However, for example, if [0LS–4HS] → [1LS–3HS] could be accelerated (e.g., in a heteroleptic oligonuclear assembly), the photoinduced population of the [1LS–3HS] species would be strongly out of equilibrium on the ns to μs time scale, thereby imparting unusual transient properties to the mixture of species.

The present study highlights the fact that tapping into the reactivity of the hot spin-states produced by PSS in metallogrid complexes is well within reach. However, several open questions have to be answered before mechanistic models can be built to guide their systematic integration into new photoconversion schemes. Synergetic methodologies based on synthetic chemistry, ultrafast spectroscopies, and theoretical modeling should be devised in order to refine the current understanding of the PSS process as a function of fluence (Bertoni et al., 2015; Zerdane et al., 2018), excitation wavelength (Zerdane et al., 2017; 2019), and solvent (Zerdane et al., 2018b). The impact of the local coordination geometry on the lifetime of the HS state has been reported in mononuclear complexes, e.g., bidentate in [Fe(bpy)₃]²⁺ (Gawelda et al., 2007; Cannizzo et al., 2010; Auböck and Chergui, 2015; Chergui and Collet, 2017; and Lawson Daku, 2018) vs tridentate in [Fe(terpy)₂]²⁺(Canton et al., 2014; Zhang et al., 2015; Bowman and Jakubikova, 2012; Petzold et al., 2017; and Lawson Daku, 2019). Improving the monodispersity of the oligonuclear samples in their ground state would clarify all the spectroscopic measurements since the mixtures of species complicate their analysis. Architectures where the number of connected units can be systematically incremented from the mononuclear building block to the complete metallogrid complex would be particularly valuable for studies aiming at determining accurate absorption cross sections and unraveling the impact of multiple excitation events. Capabilities for large-scale calculations of vibrational manifolds should be implemented for assigning the IR-active marker bands to vibrational modes of the complex undergoing PSS. Transient XES measurements with higher temporal resolution and higher signal to noise ratios will clarify the exact nature of the intermediate manifold and the role of metal centers, as for mononuclear Fe^II complexes. Finally, transient x-ray absorption spectroscopy (Bressler et al., 2009; Cammarat et al., 2014; Lemke et al., 2013; 2017; and Britz et al., 2020) and transient wide-angle x-ray scattering displaying increased structural sensitivity with hard x rays (Khakhulin et al., 2019) could also be applied to capture the local and global structural dynamics on the relevant time scales and length-scales associated with the photoreactivity triggered by the PSS process in solvated metallogrid complexes.

In conclusion, the present work initiates multi-spectroscopic studies of the PSS in solvated [2 × 2] iron(II) metallogrid complexes. The ultrafast dynamics are characterized with femtosecond pulses in the mid-infrared, UV–visible, and x-ray spectral ranges. The reversible photocycle can be described in terms of independent intra-sites transitions. Following photoexcitation, the Franck–Condon state of the mixed singlet–triplet MLCT character rapidly evolves into a vibrationally hot quintet state while exhibiting coherent behavior over 1–2 ps. The thermalization toward the equilibrated state takes around 15 ps. The metastable HS state lives for several hundreds of nanoseconds, in stark contrast with the few nanosecond lifetimes reported for the related mononuclear complex [Fe(terpy)₂]²⁺. This finding evidences the crucial role of the ligand connectivity and global nuclearity in stabilizing the photoproducts of the PSS. Independently addressable, energy-rich, and long-lived metastable states are favorable for achieving selective manipulation of the transient spin-state in oligonuclear complexes at the atomic level. With the rapid progress expected from cross-disciplinary investigations, the control of PSS in oligonuclear complexes will contribute to the elaboration of photoconversion schemes that can be efficiently triggered by ultrafast spin-switching.

The ligand HL and the complex [Fe₄(L)₄](BF₄)₄ (noted Fe4) were synthesized according to the previously reported procedures (Vlugt et al., 2008; Schneider et al., 2010). The crystallographic structures in the CCDC are CCDC-768536 [1(BF4)4 · 4 DMF (133 K)] and CCDC-768537 [1(BF4)4 · 4 DMF (233 K)].

The complex bis(2,2′:6′,2″-terpyridine iron(II)) dihexafluorophosphate ([Fe^II(terpy)₂](PF₆)₂) was synthesized according to a previously published method (Machan et al., 2012) and was additionally recrystallized from the water/acetone mixture (1:2.5 v/v) and dried under vacuum before use. The synthesis of [Fe^II(bpy)₃](PF₆)₂ was performed according to literature procedures (Kumar et al., 2016).

Mössbauer spectra were recorded with a ⁵⁷Co source in a Rh matrix using an alternating constant acceleration Wissel Mössbauer spectrometer operated in the transmission mode and equipped with a Janis closed-cycle helium cryostat. Isomer shifts are given relative to iron metal at ambient temperature. Simulation of the experimental data was performed with the Mfit program using Lorentzian line doublets: E. Bill, Max-Planck Institute for Chemical Energy Conversion, Mülheim/Ruhr, Germany.

Temperature-dependent magnetic susceptibility data were measured using a Quantum-Design MPMS-XL-5 SQUID magnetometer at a magnetic field of 5000 Oe. The sample was prepared by dissolving crystalline Fe4 in MeCN and sealing the solution in an nuclear magnetic resonance (NMR)-tube. Each raw data file for the measured magnetic moment was corrected for the diamagnetic contribution of MeCN. The molar susceptibility data were corrected for the diamagnetic contribution according to χ_Mdia(sample) = −0.5M × 10⁻⁶ cm³ mol⁻¹.

The ultrafast optical characterization was performed on two setups achieving femtosecond and nanosecond temporal resolution. They are described in Sec. S.I.5. The details for the data treatment and subsequent analysis are given in Secs. S.I.5–S.I.7.

The TXE measurements with femtosecond resolution were performed at the FXE instrument of the European XFEL facility in the standard optical pump–x-ray probe configuration. The incoming x-ray energy was set to 9.3 keV. The filling pattern was the standard 40 pulses at 374 kHz in a 110 µs train with energy ∼1 mJ/pulse. The x-ray beam was focused down to a 20 µm spot. The Kα_1,2 and Kβ emission lines were collected by 5 Ge(440) and 8 Si(531) crystals, respectively, in the von Hamos geometry and focused, in turn, on a 2D CCD GreatEyes detector. The optical excitation wavelength of the PPLS I laser system was 400 nm. The pump pattern was fixed to 40 pulses/train at 374 kHz. The pulse duration was 15 fs fwhm, and the spot size was 30 µm. The pump and the probe were overlapped in time and space on a flowing liquid round jet of 100 µm diameter. The jet speed was adjusted to insure a refreshed spot for each pump–probe event. The concentration of Fe4 was ∼3 mmol in MeCN. No sample damage was observed either in the laser_off XES spectra or in the UV–vis absorption spectra taken before/after the x-ray measurements. The details of the XES data processing and analysis are given in Sec. S.I.8.

The geometries of the metallogrid complex Fe4 in the different spin states have been optimized in the gas phase and in acetonitrile (MeCN) with the ADF program package [1], using the COSMO (conductor like screening model) implicit model of solvation for MeCN [2] and the dispersion-corrected BLYP-d3 functional [3], combined both with the Slater-type TZP basis set of triple-zeta polarized quality from the ADF basis set database [4]. The calculations were carried out within the frozen core approximation with the cores frozen up to the 1s level for the C and N and up to the 2p level for the Fe atoms. They were run restricted for the [4LS–0HS] state and unrestricted for the [3LS–1HS], [2LS–2HS], [1LS–3HS], and [0LS–4HS] states for which the z-component (quantum number MS) of the total electronic spin (quantum number S) was constrained to M_S = +2 (S = 2), M_S = +4 (S = 4), M_S = +6 (S = 6), and M_S = +8 (S = 8), respectively. The Fe–N bond lengths and the electronic energy gaps, ΔE₀, are given in the tables of S.I.12. The complete reference list for this section is given in Sec. S.I.13.

See the supplementary material for (Sec. S.I.1) a linear combination fit for the Kβ line of Fe4 in MeCN, (Sec. S.I.2) comparison of the spin-state speciation in Fe4 solvated in MeCN with magnetic and spectroscopic techniques, (Sec. S.I.3) extinction coefficient at the selected wavelengths, (Sec. S.I.4) temperature dependent measurements of the UV–visible spectrum, (Sec. S.I.5) ultrafast optical characterization of photoexcited Fe4 in MeCN, (Sec. S.I.6) SVD analysis of the transient optical and infrared spectra, (Sec. S.I.7) fitting of the oscillations in the transient optical absorption signal, (Sec. S.I.8) acquisition and extraction of steady-state and time-resolved Kα and Kβ XES at the FXE instrument of the European XFEL, (Sec. S.I.9) intensity dependency of the transient XES signal at 2 ps for 400 nm excitation, (Sec. S.I.10) fit of the XES kinetics, (Sec. S.I.11) infrared spectrum of [Fe(terpy)₂]²⁺, (Sec. S.I.12) selected structural and energetic parameters for Fe4 from DFT optimizations, and (Sec. S.I.13) complete reference list for the DFT optimizations.

The authors gratefully acknowledge the European XFEL in Schenefeld, Germany, for provision of x-ray free-electron laser beamtime at FXE and would like to thank the instrument group and facility staff for their expert assistance. M.A.N. would like to thank DESY for financial support. J.W.L.W., S.D., and F.M. gratefully acknowledge financial support from the Deutsche Forschungsgemeinschaft (Grant No. SFB 1073, project B06). J.M. acknowledges financial support from the China Scholarship Council. A.K. and M.B. acknowledge financial support from the BMBF TReXHigh and PPFocus64 (Grant Nos. 05K18PPA and 05K19PP1) as well as the DFG SPP 2102 (Grant No. BA 4467/7-1). V.M. acknowledges financial support from the BMBF MatDynamics Project (No. 05K16PX1). A.G., P.Z., F.L., K.K., M.B., A.B., W.G., D.K., and C.B. acknowledge financial support from the European XFEL. W.G. acknowledges partial financial support from the National Science Centre (NCN) in Poland under SONATA BIS 6 (Grant No. 2016/22/E/ST4/00543). C.B. gratefully acknowledges financial support from the Deutsche Forschungsgemeinschaft (Grant No. SFB 925, project A4) and from the Centre for Ultrafast Imaging (CUI). S.E.C. gratefully acknowledges funding from the Helmholtz Recognition Award. The ELI-ALPS project (No. GINOP-2.3.6-15-2015-00001) was supported by the European Union and co-financed by the European Regional Development Fund. This work was performed, in part, at the Center for Nanoscale Materials, a U.S. Department of Energy Office of Science User Facility, and supported by the U.S. Department of Energy, Office of Science, under Contract No. DE-AC02-06CH11357.