Report

Femtosecond XANES Study of the Light-Induced Spin Crossover Dynamics in an Iron(II) Complex

See allHide authors and affiliations

Science  23 Jan 2009:
Vol. 323, Issue 5913, pp. 489-492
DOI: 10.1126/science.1165733

Abstract

X-ray absorption spectroscopy is a powerful probe of molecular structure, but it has previously been too slow to track the earliest dynamics after photoexcitation. We investigated the ultrafast formation of the lowest quintet state of aqueous iron(II) tris(bipyridine) upon excitation of the singlet metal-to-ligand-charge-transfer (1MLCT) state by femtosecond optical pump/x-ray probe techniques based on x-ray absorption near-edge structure (XANES). By recording the intensity of a characteristic XANES feature as a function of laser pump/x-ray probe time delay, we find that the quintet state is populated in about 150 femtoseconds. The quintet state is further evidenced by its full XANES spectrum recorded at a 300-femtosecond time delay. These results resolve a long-standing issue about the population mechanism of quintet states in iron(II)-based complexes, which we identify as a simple 1MLCT→3MLCT→5T cascade from the initially excited state. The time scale of the 3MLCT→5T relaxation corresponds to the period of the iron-nitrogen stretch vibration.

There is a large class of iron(II)-based molecular complexes that exhibit two electronic states closely spaced in energy: a low-spin (LS) singlet and a high-spin (HS) quintet state. They therefore manifest spin crossover (SCO) behavior, wherein conversion from a LS ground state to a HS excited state (or the reverse) can be induced by small temperature or pressure changes or by light absorption (1, 2). The SCO phenomenon has been much studied using steady-state (2) and ultrafast (36) optical spectroscopies, with the goal of identifying the elementary steps leading to formation of the HS state. A representative energy level diagram of all Fe(II)-based complexes is shown in Fig. 1 (7). The main difference between them concerns the absolute energies of states, not their energetic ordering (2). All crystallographic studies point to an Fe-N bond elongation by ∼0.2 Å in the HS compared to the LS state (1, 2). Theoretical studies show that the Fe-N bond length of the singlet and triplet metal-centered (MC) 1,3T states lies halfway between those of the LS and HS states (7). Obviously, accessing the HS excited state by absorption of light from the LS ground state is forbidden by the spin selection rules. Therefore, the doorway to the HS state is ideally via the singlet metal-to-ligand-charge-transfer (1MLCT) that exhibits strong absorption bands in the visible spectrum, or via the weakly absorbing and lower-lying 1,3T states (1, 2). The time scale and the route going from the initially excited 1MLCT state to the lowest-lying quintet state are still the subject of debate. Steady-state spectroscopic studies at cryogenic temperatures showed that excitation into the MC 1,3T states leads to population of the 5T2 state with a quantum efficiency of ∼80% (2). Researchers therefore concluded that the relaxation cascade from the 1MLCT state to the HS 5T2 state proceeds via the intermediate 1,3T states. However, for excitation of the 1MLCT state, the relaxation process was reported to occur with 100% efficiency at both 10 K (2) and at room temperatures (8, 9), thus questioning the involvement of the intermediate 1,3T states. Ultrafast laser studies established that the relaxation cascade from the initially excited 1MLCT state to the lowest excited quintet state 5T2 is complete in <1 ps (35), but this result was indirectly inferred, as neither the intermediate MC states nor the final 5T2 state have spectroscopic transitions in the region of the probe (>350 nm). McCusker and co-workers (6) proposed that the 1MLCT state relaxes to a manifold of strongly mixed singlet and triplet MC states down to the quintet state, the latter being considered to be the only one clearly defined by its spin quantum number S. However, the steady-state spectroscopic studies of Hauser and coworkers (2) point to a clear classification of all MC states according to their spin character, thus excluding strong state mixings.

Fig. 1.

(A) Representative potential energy curves of Fe(II)-based SCO complexes as a function of the Fe-N bond distance (7). The manifold of MLCT states is shown as a shaded area. [FeII(bpy)3]2+ has predominantly Oh symmetry with a trigonal (D3) distortion. The MC states are represented by their symmetry character (A, T, and E) in the D3 group: the LS 1A1 ground state has a completely filled e4a12 configuration (deriving from the t2g6 subshell in Oh symmetry), whereas the anti-bonding e (eg in Oh symmetry) orbital is empty. Per electron that is promoted from the t2g subshell to the eg subshell (for easier reading we will use the Oh nomenclature hereafter), the metal-ligand bond length increases by as much as 0.1 Å (1, 7). For the series of 1,3T(t2g5eg) states, the Fe-N bond length is expected to lie between the values observed for the ground and the high-spin 5T2 (t2g4eg2) states. (B) Relaxation cascade as determined by ultrafast laser spectroscopy upon excitation of aqueous [Fe(bpy)3]2+ at 400 nm (5). The intermediate MC states are not shown because they are optically silent in the region >350 nm and were therefore not observed in (5). (C) For the [Fe(bpy)3]2+ complex, the Fe-N bond length is 1.97 Å in the low-spin 1A1(t2g6 ground state (32) but increases by 0.2 Å in the high-spin 5T2 (t2g4eg2) state, as determined by picosecond XAS experiments (11).

Iron(II)-tris(bipyridine) ([FeII(bpy)3]2+), which is the molecule studied here, serves as a model system for the family of Fe(II)-based SCO complexes. Early events of the relaxation cascade in aqueous [FeII(bpy)3]2+ were recently investigated using femtosecond resolved fluorescence and transient absorption by Gawelda et al. (5) upon 400-nm excitation of the 1MLCT state. They observed a prompt (∼30 fs) intersystem crossing (ISC) to the 3MLCT state, followed by a departure from this state within ∼120 fs (Fig. 1B). The subsequent steps and the arrival into the HS state were not observed directly, and the final step of the photo-cycle, the radiationless HS→LS transition, was identified via the recovery of the ground-state bleach with its 665-ps lifetime. For 400-nm excitation, the relaxation cascade from the initially excited 1MLCT state to the HS state implies dissipation of 2.6 eV of energy in <1 ps and, were it to proceed via the intermediate MC states, it would entail a back electron transfer, followed by at least three ISC events, as well as an Fe-N bond elongation by 0.2 Å. This elongation was recently measured by x-ray absorption spectroscopy (XAS) studies with 50- to 100-ps resolution on [FeII(tren(py)3)]2+ (10) and [FeII(bpy)3]2+ (11, 12) in solution. The structural change manifests itself through substantial modifications of the x-ray absorption near-edge structure (XANES) at the Fe K-edge, which we exploit in the present study of the ultrafast light-induced SCO of aqueous [FeII(bpy)3]2+.

So far, most x-ray studies with subpicosecond time resolution have used diffraction to investigate strain, coherent phonon dynamics, or melting phenomena in solid materials (1315). Scattering does not require wavelength tunability, and sources of (monochromatic) femtosecond x-ray pulses (obtained by plasma emission from metal targets struck by intense ultrashort laser pulses) have readily been available for some time now. Diffraction is also a collective phenomenon in crystals, delivering rather strong signals. For chemical and biological systems that may be disordered and diluted in solution, x-ray absorption spectroscopy is a more suitable probe (12, 16, 17). However, it requires rather stable sources of tunable ultrashort x-rays. Subpicosecond x-ray plasma sources have been implemented for time-resolved XAS studies on the few picoseconds (18) to tens of picoseconds time scale, but their use is challenging because of their poor shot-to-shot stability and low fluxes (19, 20). Synchrotron sources (12, 17) deliver very stable radiation with reasonable fluxes, although the pulse durations lie in the 50- to 150-ps range. The recently developed slicing scheme (21) has allowed the extraction of tunable femtosecond x-ray pulses from a synchrotron and was first implemented for soft x-ray absorption studies of the electronic changes resulting from the photo-induced ultrafast insulator-metal phase transition in VO2 bulk crystals (22) and the ultrafast demagnetization dynamics in solid nickel (23). For structural determination, hard x-rays (>2 keV) are better suited, and the recent implementation of the slicing scheme for 5 to 20 keV radiation at the Swiss Light Source (SLS, PSI-Villigen) (24) opens the possibility of carrying out ultrafast XANES studies on dilute molecular systems in liquids. By applying this technique, we have succeeded in following the structural changes in real time upon visible light excitation of aqueous [FeII(bpy)3]2+, and moreover we have unraveled the mechanism of the ultrafast spin crossover in this class of molecules.

Briefly (25), a 100-μm-thick free-flowing liquid jet of an aqueous solution of 50 mM [FeII(bpy)3]2+ was excited by an intense 400-nm laser pulse (115-fs pulse width, repetition rate 1 kHz), and a tunable femtosecond hard x-ray pulse from the slicing source was used to probe the system in transmission mode at 2 kHz. The flux of the femtosecond x-ray source was about 10 photons per pulse at 7 keV. We recorded the transient difference absorption spectra by alternating detection of signals from the laser-excited and the unexcited sample, thus achieving a precise intrinsic energy calibration that compensates for drifts of the laser or synchrotron energies and fluxes. The time resolution was <250 fs (25).

Figure 2A shows the Fe K-edge XANES spectra of aqueous [FeII(bpy)3]2+ in its ground (LS) and excited (HS) state. The spectrum of the latter was retrieved from the LS spectrum and the difference spectrum recorded 50 ps after laser excitation (red dots in Fig. 2B) (11). The strongest increase in absorption upon LS to HS conversion occurs at the so-called B-feature (arrow in Fig. 2), which was previously identified as a structure-sensitive above-ionization multiple-scattering resonance (26). The increase in intensity of the B-feature is concomitant with the increase in Fe-N bond distance upon LS to HS conversion, reflecting a well-established correlation between edge absorption intensity and bond distance [see, e.g., (27)]. For the present system, this correlation was confirmed (fig. S1) by a simulation of the XANES spectrum using the Minuit XANES (MXAN) code (11, 28), which additionally shows a nearly linear relationship between the Fe-N bond elongation and the intensity of the B-feature. The B-feature intensity is therefore a signature of the Fe-N bond elongation, and it allows us to distinguish the various states that can be grouped by their similar Fe-N bond distances: (i) the LS ground and the 1,3MLCT states; (ii) the 1,3T states, which exhibit an elongation of 0.1 Å relative to the ground state (7); and (iii) the 5T state, which exhibits a 0.2 Å elongation (11). Based on this correlation, we analyze the observed light-induced changes at the B-feature as a function of the time delay between the optical pump pulse and the x-ray probe pulse in the femtosecond to picosecond time domain.

Fig. 2.

(A) Fe K-edge XANES spectrum of the LS state of aqueous [FeII(bpy)3]2+ (black trace) and of the HS quintet state (red dots). The latter is determined from the LS spectrum and the transient spectrum (B) measured at a time delay of 50 ps after laser excitation at 400 nm (11). (B) Transient XANES spectrum (difference in x-ray absorption between the laser-excited sample and the unexcited sample) recorded 50 ps after laser excitation at 400 nm (red dots) (11). Note the increase in absorption at the so-called B-feature. The blue stars represent the transient spectrum recorded at a time delay of 300 fs in the present work. Error bars, ±1 SD (25).

Figure 3 shows the transient signal at the B-feature as a function of time delay (the inset shows an expanded region out to 10 ps). It is characterized by a steep rise followed by a plateau beyond 250 to 300 fs, which suggests that the system has reached the HS state within this time frame. This suggestion is confirmed by the energy scan recorded at a time delay of 300 fs (blue stars in Fig. 2B), which agrees with the transient absorption spectrum recorded at 50-ps time delay. Considering a simple four-level kinetic model 1A11MLCT→3MLCT→5T, we simulated the signal with no adjustable parameters, assuming (i) an optical/x-ray cross-correlation of 250 fs; (ii) the 1MLCT and 3MLCT decay times measured in (5) (Fig. 1B); (iii) the cross sections at the B-feature for the LS and HS states (Fig. 2A), as well as for the intermediate 1,3T states, derived from the relationship between the Fe-N bond elongation and the B-feature intensity (25); and (iv) the absorption cross section of the MLCT state(s). For the latter, based on our previous study of the analogous [RuII(bpy)3]2+ molecule (29), the MLCT and LS XANES are expected to be similar, except for a shift to higher energies of the MLCT XANES spectrum, due to the oxidation of the central metal atom (over the time the system remains in the MLCT manifold). At the Fe K-edge, this oxidation state shift amounts to at most +2 eV based on a study of Fe(II)- and Fe(III)-hexacyanide (30).

Fig. 3.

(A) Time scan of the signal (blue points) at the B-feature (at 7126 eV) (Fig. 2B) as a function of laser pump/x-ray probe time delay after excitation of aqueous [FeII(bpy)3]2+ at 400 nm. The inset shows a long time scan up to a 10-ps time delay. The red trace is the simulated signal assuming a simple four-step kinetic model 1A11MLCT→3MLCT→5T to describe the spin conversion process [see (B)]. The vertical arrow displays the expected effect of the elongation of 0.2 Å for the Fe-N bond elongation ΔR between the LS and HS states. Error bars, ±1 SD (25). (B) Simulated transient absorption traces of the individual states (black, green, and blue) and total (red) trace based on a four-step kinetic model 1A11MLCT→3MLCT→5T, with the 1MLCT→3MLCT intersystem crossing taking place in 20 to 30 fs and the depopulation of the 3MLCT state taking place in 120 fs, as determined by ultrafast fluorescence and transient absorption studies (5). Neglect of the absorption decrease due to the MLCT states (dashed traces) does not affect the final simulated signal. The kinetics were convoluted with a cross-correlation of 250 fs between laser and x-ray pulse.

Figure 3B shows the simulated time evolution of the signal due to the various states, including (solid curves) and excluding (dashed curves) the MLCT states, as well as the resulting total signal (red traces). The blue shift of the MLCTspectrum should lead to an initial decrease in absorption of the signal, which we do not observe. The simulations (green trace in Fig. 3B) confirm that our pump-probe correlation time brings this initial signal decrease below our experimental sensitivity, thus accounting for its absence in the data. Also, we note that time zero is hardly affected by inclusion of the signal due to the MLCT state, and its value changes by at most 40 fs when comparing the simulations with and without the (temporary) oxidation shift. Finally, introducing the 1,3T state(s) can reproduce the data only for a fitted relaxation time of <60 fs (25). Such a short lifetime, however, is unrealistic because it corresponds to the period of high frequency modes of the system, which in addition would need to be shared among the 1,3T states that all have potential curves with identical equilibrium distances and curvatures along the Fe-N coordinate (Fig. 1A). Consequently, the agreement between the experimental and the simulated time trace (Fig. 3A) implies that the rise time (∼150 fs) of the x-ray absorption of the HS state corresponds to the decay of the 3MLCT state (5). Thus, the population of the 3MLCT state proceeds to the quintet state directly and bypasses the intermediate 1,3T states. Furthermore, the derived relaxation time scale corresponds to the period of the Fe-N stretch mode, which lies in the 130- to 160-fs range for all Fe(II)-based complexes, according to Raman studies (31). Therefore, here the observation of the structural dynamics allows us to unambiguously unravel the population relaxation pathway.

Because the 3MLCT state lies about 1.5 eV above the quintet state (Fig. 1), the latter is populated in high vibrational levels. However, we do not observe vibrational cooling in the quintet state, because XANES spectroscopy is in general not very sensitive to Debye-Waller factors, which reflect the uncertainty in atomic coordinates due to thermal motion.

The general picture of the light-induced SCO process that emerges from this study becomes very simple and is summarized in Fig. 1B. The full cascade reduces a two-step ISC process, 1MLCT→3MLCT→5T. The bypassing of the intermediate 1,3T states resolves the issue of multiple ultrafast ISC steps among states that are quasiparallel with respect to the Fe-N coordinate (Fig. 1A). Dissipation of the energy difference in the ultrafast cascade is accounted for by storage of vibrational energy in the quintet state. Finally, the unit quantum efficiency of the SCO process from the 1MLCT state into the quintet state makes sense in the context of excluding any leakage back to the ground state (9) through the bypassing of the 1,3T states.

Considering that [FeII(bpy)3]2+ is a model system for all Fe(II)-based SCO complexes, we believe that our results are of general validity to this family. These results also underscore the power of ultrafast x-ray absorption spectroscopy for the study of molecular structural dynamics of dilute systems. In the present case, resolving the structural dynamics unravels the pathways of spin and electronic relaxation.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1165733/DC1

Materials and Methods

Fig. S1

References

References and Notes

View Abstract

Navigate This Article