Femtosecond spin-state switching dynamics of spin-crossover molecules condensed in thin films

Abstract

The photoinduced switching of Fe(II)-based spin-crossover complexes from singlet to quintet takes place at ultrafast time scales. This a priori spin-forbidden transition triggered numerous time-resolved experiments of solvated samples to elucidate the mechanism at play. The involved intermediate states remain uncertain. We apply ultrafast x-ray spectroscopy in molecular films as a method sensitive to spin, electronic, and nuclear degrees of freedom. Combining the progress in molecule synthesis and film growth with the opportunities at x-ray free-electron lasers, we analyze the transient evolution of the Fe L3 fine structure at room temperature. Our measurements and calculations indicate the involvement of an Fe triplet intermediate state. The high-spin state saturates at half of the available molecules, limited by molecule-molecule interaction within the film.

Figures (5)

• Figure 1: Low- and high-spin configurations in spin-crossover molecules. The calculated geometries of an individual [Fe(pypypyr)$_2$] molecule in the low-spin ($S = 0$, left) and high-spin ($S = 2$, right) state. Concomitantly with the low- to high-spin transition, the Fe-N bond length changes by approximately 10 % as indicated by the corresponding average bond lengths inferred from density functional theory calculations. The corresponding ligand twisting in the high-spin state is indicated on the right by the dashed line and the bended arrow. The excitation of the molecule by an external stimulus, such as a pump laser pulse at photon energy $h \nu_\text{pump}$, can trigger this transition. The splitting of $3d$ states of the Fe$^{2+}$ ion into $t_{2g}$ and $e_g$ in an octahedral ligand field is represented by the 10Dq value as indicated at the left and right side for $S = 0$ and $S = 2$, respectively. In the condensed form, the molecules are surrounded by peers, as illustrated with molecules in light colors in the background, which can further modify the switching properties.
• Figure 2: Time-resolved x-ray absorption spectroscopy at the Fe $L_3$ edge. (a) Schematic of the experimental setup of the Spectroscopy and Coherent Scattering Instrument at the European XFEL. (b) Schematic x-ray absorption process for the measured fine structure at the Fe $L_3$ absorption edge. (c) An x-ray absorption spectrum of the molecular film measured at room temperature at the unpumped window (open circles, blue) is depicted together with a pumped spectrum (filled circles, red) at time delay $\Delta t = 3$ ps, which was recorded at a laser pumped membrane window. Both spectra are corrected with the simultaneously measured $I_0$ on a reference Si$_3$N$_4$ window (Supporting Information). The difference of the pumped and the unpumped spectrum represents the pump-induced change (bottom panel). The right axis is given relative to the while line maximum of the unpumped spectrum. The two features in the spectrum at photon energies of 707.1 eV and 708.9 eV, see vertical lines, show changes in intensities that are highlighted by arrows. This spectral modification is represented by the pump-induced change and corresponds to the spin-state switching from $S = 0$ to $S = 2$, induced by the pump laser at an incident fluence of $10$ mJ/cm$^2$.
• Figure 3: Intermediate state analysis of the spin-crossover process. (a) Top panel: Normalized pump-induced change at 707.1 eV (black symbols) and 708.9 eV (pink symbols) photon energy as a function of pump-probe delay $\Delta t$. The lines guide the eye. The gray dashed line is the absolute value of the guide to the eye of the negative transient for 708.9 eV. The right ordinate indicates the ratio to the maximum in the unpumped spectrum, see Fig. \ref{['fig:scs']}c. Inset: The data of the main panel are presented for a reduced time delay range. Bottom panel: The difference of the two transient absorption traces shown in the top panel are depicted. The data are reconstructed out of 26 time-delay sweeps, for which the time zero reference is adjusted as discussed in the SI. The timing of the two transients on the time delay has been defined to a separate reference measurement and does not take drifts into account. (b) X-ray absorption spectra at three different time delays $\Delta t$: before pumping (blue), at 256 fs (black), and at 3 ps (red). In the region of the gray bar the spectrum at 256 fs exhibits larger absorption than the one at 3 ps. The spectrum $\Delta t = 256\pm80$ fs is an average of three transient ($\Delta t = 191$, 238, and 338 fs). The concomitantly measured spectra on unpumped windows (in the LS state) were normalized to their peak intensity at a photon energy of 708.9 eV. The same normalization factors were used to scale the corresponding pumped spectra. The lower panel shows the difference between the $\Delta t = 256$ fs and the $\Delta t = 3$ ps spectra (filtered with a 5-point moving average).
• Figure 4: Calculated potential energy surfaces. Total energies of the metal-centered $S = 0$, $S = 1$, and $S = 2$ (thick lines) states, as well as $^1$MLCT singlet (turquoise area), and $^3$MLCT (gray area) as a function of the average change in Fe-N bond length as inferred from time-dependent DFT. As a starting point, we consider a photo-excitation from the $S=0$ to the manifold of $^1$MLCT states. The arrows illustrate hypothetical relaxation pathways. The Fe-N distance change varies along two coordinates to represent the low-, intermediate-, and high-spin states.
• Figure 5: Pump fluence dependence of the pump-induced change. The normalized pump-induced change at 3 ps time delay is plotted as a function of incident pump fluence in the x-ray absorption at the two features in the fine structure as indicated by the given photon energy. For clarity, the absolute value of the negative pump-induced change at 708.9 eV photon energy is depicted. The considerable fluctuations of the pump-induced changes are assigned to modifications in spatial overlap of pump and probe pulse upon varying the fluence. Within the available data quality, we identify an increase in the pump-induced change to about 5 mJ/cm$^2$ and a saturation of these signals for larger pump fluence.