Towards Iron(II) Complexes with Octahedral Geometry: Synthesis, Structure and Photophysical Properties

Abstract

The control of ligand-field splitting in iron (II) complexes is critical to slow down the metal-to-ligand charge transfer (MLCT)-excited states deactivation pathways. The gap between the metal-centered states is maximal when the coordination sphere of the complex approaches an ideal octahedral geometry. Two new iron(II) complexes (C1 and C2), prepared from pyridylNHC and pyridylquinoline type ligands, respectively, have a near-perfect octahedral coordination of the metal. The photophysics of the complexes have been further investigated by means of ultrafast spectroscopy and TD-DFT modeling. For C1, it is shown that—despite the geometrical improvement—the excited state deactivation is faster than for the parent pseudo-octahedral C0 complex. This unexpected result is due to the increased ligand flexibility in C1 that lowers the energetic barrier for the relaxation of ³MLCT into the ³MC state. For C2, the effect of the increased ligand field is not strong enough to close the prominent deactivation channel into the metal-centered quintet state, as for other Fe-polypyridine complexes.

1. Introduction

Ligand-field splitting is an important parameter with a deep impact on the photophysics of transition metal-based complexes. Indeed, its fine tuning is crucial to populate long-lived metal-to-ligand charge transfer (MLCT) states that are required for triggering light induced functions such as photochemically induced electron transfer [1,2,3,4,5]. The control of the ligand-field splitting is particularly challenging in iron (II) polyimine complexes where the MLCT states, first populated upon photo excitation, undergo an ultrafast deactivation into the low-lying metal-centered (MC) states, thus leading to the loss of the suitable photophysical properties [6,7].

N-heterocyclic carbenes (NHC) ligands, ideally combining a strong σ-donating and a weak-to-moderate π-accepting character, have been reported to destabilize the MC states over the MLCT manifold, thus impressively slowing down the deactivation pathways. In this context, tridentate pyridyldicarbene ligands with appropriate modifications of both the NHC side and the central azine have been reported by us and other groups as leading to remarkable ³MLCT lifetimes in the 9–32 ps range for the corresponding iron (II) complexes [4,8,9,10,11,12,13,14]. Other ligands giving rise to more rigid complexes [15] or different coordination architectures [16] have achieved impressive MLCT lifetimes beyond hundreds of picoseconds. A comprehensive review of the rapid advances in this area was recently published [17].

The ligand-field splitting can also be enhanced by reducing the angular strain around the metal center [18,19,20,21,22], i.e., when the coordination geometry is as close as possible to a perfect octahedron, as seen for example in some heavy metal complexes. For example, the angularly strained [Ru(tpy)₂]²⁺ complex (tpy = 2,2′:6′,2′’-terpyridine) experiences an ultrafast population of MC states [23], while the weakly distorted bidentate homologue [Ru(bpy)₃]²⁺ (bpy = 2,2′-bipyridine) [24] is strongly luminescent at room temperature with excited-state lifetimes up to 1 ms since the access to deactivating MC states is hampered [25].

In the case of Ru-based compounds, the angular strain has been minimized by switching from tridentate to bidentate ligands. We have recently shown that the use of a bidentate pyridylNHC ligands also improved notably the excited state lifetimes of iron complexes compared to their tridentate analogues [11].

However, in contrast to tridentate ligands, the dissymmetrical bidentate ligands generally form mixtures of fac and mer isomers, which can be problematic when a vectorial electron transfer is required [26]. In addition, it has also been shown that the fac and mer isomerism may lead to unexpected photophysical consequences [11,13,14]. Thus, the possibility of achieving tridentate iron complexes with reduced angular strain still appears extremely attractive for the design of photoelectrochemical active agents. While some elegant approaches to increase the ligand field using tridentate ligands have been developed [18,27,28,29] there is a lack of such advances for iron (II) complexes, and the benefit of an improved octahedral coordination on the excited state relaxation dynamics has only be explored for an Fe(II) complex with 2,6-bis(2-carboxypyridyl)pyridine ligands [30]. Here, it was shown that the higher ligand field splitting destabilizes the ⁵MC, leading to a faster deactivation of this quintet state.

Herein we report the synthesis, structural characterization and photophysics of the novel homoleptic complexes C1 and C2 (Figure 1) based on relevant tridentate ligands selected for their ability to create an almost perfect octahedral geometry of the coordination sphere. In particular, we focus on the dqp (2,6-diquinolylpyridine) ligand, which has been reported to give a close to perfect octahedral complexes having a great impact on the MLCT lifetimes that reached the microsecond scale for the homoleptic Ru(dqp)₂²⁺ complex (C2) [18]. A second ligand structure achieved by introducing methylene spacers on the 2,6-Bis(1-methylimidazolylidene) pyridine ligand has been used to make C1. It shall be noted that the 2,6-Bis(1-methylimidazolylidene) pyridine has been previously reported to give C0, a complex with a distorted octahedral Fe²⁺ coordination exhibiting MLCT lifetimes of 9 ps [8].

By such modifications, we investigated the effects induced by the optimal coordination geometry on the photophysical properties of these new iron complexes aiming to extend the excited-state lifetimes analogously to the reported case of ruthenium.

2. Results

2.1. Synthesis and Characterization

2.1.1. Synthesis of Ligands and Complexes

The access to C1 required the synthesis of the pyridylimidazolium salt precursor L1. It was obtained in 80% yield from 2,6-bis(bromomethyl)pyridine and 1-methylimidazole according to the literature [31]. L1 was subsequently involved in the coordination process by a first reaction with FeBr₂ followed by the addition of tBuOK to generate the carbene (Scheme 1). When performed at room temperature, the reaction led to degradation of L1. However, the coordination was successfully made by decreasing the temperature to 0 °C affording C1 in 19% yield. The low coordination yield obtained is in agreement with those obtained previously with Fe(II)NHC ligands [4,9,13].

The diquinolinylpyridine (dqp) ligand required for the synthesis of C2 has been obtained in 65% yield employing a Suzuki cross-coupling between 8-quinoline boronic acid and 2,6-dibromopyridine according to a reported procedure [18]. The coordination of this ligand with iron was first attempted by mixing FeBr₂ and dqp, where regardless of the temperature and solvent used, no complex formation was detected. Switching to a microwave irradiation (60 W) in refluxing DMF for 4 h gave the expected complex C2 in 30% yield (Scheme 2). Another iron precursor has been tested by using ammonium iron(II) sulfate hexahydrate [Fe(NH₄)₂(SO₄)₂]₆ H₂O. Overnight reaction at room temperature in acetonitrile gave C2 also in 30% yield.

2.1.2. X-Ray Structures

Single crystals of the two complexes suitable for X-ray diffraction analysis were obtained from slow evaporation of acetonitrile solution. X-ray crystal structures of the complexes C1 and C2 are shown in Figure 1 and Figure 2, and selected bond lengths and bond angles are given in

Table 1. Bond lengths and coordination angles of C0 and C1.

Bond Length (Å)				Angles (°)
C0		C1		C0		C1
Fe1-N3	1.919(3)	Fe1-N1	2.035(11)	C1-Fe1-N3	79.5(1)	C1-Fe1-N1	87.7(5)
Fe–N8	1.930(3)	Fe1-N1′	2.035(11)	C11-Fe1-N3	79.2(1)	C12-Fe1-N1	88.5(4)
Fe1-C11	1.965(3)	Fe1-C12	1.981(11)	C1-Fe1-C11	158.0(2)	C12-Fe1-C1	176.1(5)
Fe1-C14	1.965(3)	Fe1-C1′	1.981(10)	C14-Fe1-N8	79.0(1)	C1′-Fe1-N1′	87.7(5)
Fe1-C1	1.966(3)	Fe1-C1	1.981(10)	C22-Fe1-N8	79.0(1)	C12′-Fe1-N1′	88.5(5)
Fe-C22	1.970(3)	Fe1-C12′	1.981(11)	C14-Fe1-C22	158.0(2)	C1′-Fe1-C12′	176.1(5)

and

Table 2. Bond lengths and coordination angles of C2.

Bond Length (Å)		Angles (°)
Fe1-N2	1.961(2)	N2-Fe1-N3	90.36(9)
Fe1-N2′	1.957(2)	N3′-Fe1-N1	90.88(8)
Fe1-N1	1.997(2)	N3-Fe1-N1	177.58(8)
Fe1-N1′	1.997(2)	N2′-Fe1-N1′	90.15(9)
Fe1-N3	1.986(2)	N2′-Fe1-N3′	90.48(9)
Fe1-N3′	1.995(2)	N3′-Fe1-N1′	179.21(9)

.

In contrast to its parent complex C0, which has a bite angle of 158° [8], the geometry around the central iron atom in C1 is very close to fully octahedral as shown by the increased bite angles (176.1°) afforded by the flexibility provided by the methylene extension. The Fe–Pyridine bonds are longer than those of C0 (2.035 vs. 1.930 Å), which agrees with the widening of the coordination sphere in C1.

Like its ruthenium homologue [18], the crystallographic angles confirm the optimization of the octahedral symmetric coordination sphere of C2 (Figure 3). Indeed, ideal coordination angles close to 90° on average have been obtained (Table 2), while maintaining bite angles close to 180°. The Fe–N (quinoline) bonds are slightly longer than the two Fe–N (pyridine) bonds due to the better charge delocalization in quinoline, increasing the electronic density and the amount of antibonding interactions with the iron d orbitals.

2.1.3. Electronic and Electrochemical Properties

The electronic and optical properties of both complexes were investigated by means of steady-state UV-vis spectroscopy (Figure 4) and cyclic voltammetry (

Table 3. Photophysical and electrochemical properties of C0, C1 and C2.

Complex	λabs-Max a (nm) ε(M−1.cm−1)	Eox (FeIII/FeII) b (V/SCE)	Ered1 (V/SCE)	ΔE c (eV)
C0	393 (9000) 460 (15,900)	0.71 (rev)	−2.00 (irrev)	2.75
C1	525 (4800)	0.31 (rev)	−2.05 (irrev)	2.36
C2	575 (14,200)	0.96 (rev)	−1.35 (rev)	2.31

^a Measured in CH₃CN at 25 °C. ^b First oxidation potential. Potentials are quoted vs (SCE). Under these conditions, E_1/2 (Fc+/Fc) = 0.39 V/SCE. Recorded in CH₃CN using Bu₄N⁺,PF₆⁻ (0.1 M) as supporting electrolyte at 100 mV.s⁻¹. ^c Electrochemical band gap (ΔE = E_ox − E_red1).

).

The UV-Vis spectra of both complexes present two types of absorption bands. Very intense bands are in the UV range (<350 nm) and are ascribed to bright π−π* transitions localized on the ligands. For each complex, one distinct, broader, and less intense band is found at longer wavelengths in the visible range. The absorption spectrum of C2 presents a better spectral coverage, its absorption band at 575 nm (14,200 M⁻¹ cm⁻¹) is notably red-shifted by 50 nm as compared to the maximum absorption band of C1, which peaks at 525 nm (4800 M⁻¹ cm⁻¹). This bathochromic shift can be attributed to the combination of the geometric stabilization and the electron-withdrawing effects of quinolines stabilizing the MLCT states. One can note that for C1, the characteristic absorption band of the Fe-carbene transitions is no longer visible or resolved as it is the case in its parent complex C0, there with a maximum at 393 nm. This transition is most probably hidden in the broad MLCT absorption band in the 420–500 nm range. The decrease in the absorption capacity observed in C1 could be related to the small barrier between singlet and triplet MLCT states (see Figure 8) producing an increased triplet character.

The redox properties of both complexes were measured by cyclic voltammetry with SCE as standard electrode and compared in Table 3.

For C2, as compared to C0, electrochemical measurements show an increase of the Fe^II/Fe^III oxidation potential (0.96 V vs. 0.71 V) due to the π-acceptor effect of quinolines. The reduction potential is also increased (−1.35 V vs. −2.00 V), which could be ascribed in a first approximation to a lower LUMO energy level due to a greater stabilization through extended π-conjugation in the bqp ligand compared to the NHC ligand in C0. The lower electrochemical gap is in good agreement with the bathochromic effect observed in the UV-Vis spectra.

For C1, a strong decrease of the Fe^II/Fe^III oxidation potential is observed compared to its parent complex C0 (0.31 V vs. 0.71 V) which is indicative of the marked increase of the σ-donating character of the carbene moieties which are no more conjugated with the pyridine as in C0. However, C1 exhibits an irreversible monoelectronic transfer at −2.05 V occurring at about the same potential as C0 indicating for these two complexes higher energy levels of the π* orbitals than for C2.

2.2. Excited State Relaxation Dynamics

The effects of the ligand geometry of the coordination sphere on the excited state lifetimes was investigated for compounds C1 and C2 by femtosecond transient absorption spectroscopy (fs-TAS) and compared to C0, as reference [8].

2.2.1. Fs-TAS of C1

C1 dissolved in acetonitrile (ACN) was excited with ≈50 fs pump pulses at 495 nm, and the pump-induced absorbance changes ΔA, with no more than 10% ground state depletion, were probed by a white-light continuum generated in a 2-mm thin CaF₂ plate in the range of 320–640 nm. Figure 5 displays the time-dependent ΔA, and the sign-inverted ground state absorption spectrum (GSA) for comparison. The TAS data show at all times a negative ground state bleach (GSB) in the range of 410–540 nm at early delays, with a peak at 507 nm. Positive ΔA due to excited state absorption (ESA) is identified in the blue part of the GSB, below 400 nm, and in the red, above 520–540 nm. Note that both ESA bands have significant overlap with the GSB, since the latter is visible only in a limited range of the GSA spectrum, due to the larger absorption cross section of ESA in the ranges where ΔA > 0. Clear spectral evolution occurs on a sub-300 fs time scale in the blue ESA band, shifting the ESA-GSB crossover around 400 nm by 20 nm, and in the red ESA during a longer time period, i.e., the first 10 ps (25 nm red-shift). This indicates that different species contribute to the ESA, populated along a relaxation scheme we will identify in what follows.

Another interesting observation is related to periodic modulations in the ESA amplitude. Figure 5C highlights these for the blue and red ESA (366, 576, 630 nm), and for the longer wavelength part of the GSB (510 and 535 nm), but not for 450 nm. The oscillation period is difficult to determine precisely as damping is strong, but it can be determined to be of the order of 0.5 ps. Maxima of the oscillatory features in ESA correspond to minima at the wavelengths of negative ΔA. Keeping in mind that the latter is a result of overlapping ESA and GSB, one can conclude, in particular for λ > 450 nm, that the modulations are in the ESA signal only. In other words, the excited state transition dipole moment is modulated by a strongly damped but coherent vibrational motion with a frequency of ≈60 cm⁻¹. See SI for a more quantitative analysis, and higher frequency ground state oscillations.

Regarding the excited state lifetimes and relaxation scenario, Figure 5B plots kinetic traces on a semi-log scale, so as to highlight the components of exponential decay. Dashed black lines with a slope of 1.5 ps overlay with the initial portion of most of the kinetic traces. When the signal-to-noise ratio is high enough, an additional slower component, 10–15 ps, is observed (450, 510, 576 and 600 nm). All kinetic traces can be well fitted by a sum of three exponentials with wavelength-dependent amplitudes and decay constants, ${{\displaystyle \sum }}_{i=1}^3{A}_i\left(\lambda \right)e^{-t/{\tau }_i\left(\lambda \right)}$, convoluted with a Gaussian instrument response function (width 50 fs). The results of single wavelength fits are displayed as solid black lines in Figure 5C. They reveal a rise time, in the ≈0.3 ps range, for the red ESA part (576 and 630 nm). Table S16 in the Supplementary Materials gives the fit parameters for individual wavelengths. The fastest ESA decay and GSB recovery time is τ₂ = 0.8–1.5 ps, and the second decay is found by τ₃ = 7–17 ps, with an overall 8–10 times smaller amplitude than τ₂.

Global analysis, i.e., fits of the full data set, but with wavelength-independent lifetimes, ${{\displaystyle \sum }}_{i=1}^3{A}_i\left(\lambda \right)e^{-t/{\tau }_i}$, were carried out, using OPTIMUS [32] and the results displayed as decay-associated difference spectra (DADS,$A_i\left(\lambda \right)$) in Figure 6A. The oscillatory components are neglected in these fits, but they represent the qualitative trends of Figure 5. The validity of such global fitting is assessed in detail in the Supplementary Materials (cf. Figure S7). First, the negative amplitude of the 0.3-ps DADS in the blue and red ESA regions reflect the observed rise. Likewise, its positive amplitude in the 380–450 nm range is consistent with the slightly rising GSB signal in this time and wavelength ranges (Figure 5C). Second, the double exponential decay character of both ESA and GSB is reliably reproduced in the DADS of τ₂ = 1.3 ps and τ₂ = 7.3 ps. We find that the amplitude of τ₂ is approx. 10 times larger than the one of τ₃ in the blue ESA band and above 580 nm. However, since the ESA amplitudes depend on the respective cross sections and overlap with the GSA, it is not straightforward to conclude about the relative populations of these channels.

What is the excited state relaxation scenario emerging from these spectroscopic data? The sub-picosecond τ₁ is usually assigned to the transition from the optically excited ¹MLCT to ³MLCT and vibrational relaxation of the latter, reflected in the ultrafast rise of ESA below 380 and above 500 nm. The positive DADS of τ₁ in the 380–450 nm range can be related to a reshaping of ESA, i.e., vibrational relaxation of ³MLCT. On the other hand, the DADS of τ₂ and τ₃ are very similar to the situation we recently reported for complexes with bidentate ligands [14], where two excited states seem to co-exist, and is rather at odds with the single excited state scenario we reported for tridentate Fe–NHC complexes [12]. Indeed, for the latter τ₃ was dominant in intensity and τ₂ with minor amplitude rather related to structural relaxation in the single excited state. Hence, the TAS data are consistent with the existence of two excited triplet states, T₁ and T₂, populated via relaxation from ¹MLCT and subsequent vibrational cooling during τ₁. Their precise nature in terms of diabatic states (MLCT, MC or mixtures of both) cannot be determined from the TAS difference spectra, and will be addressed by TD-DFT calculations (Section 2.3). However, the experimental data suggest a parallel decay back to the ground state S₀, according to the following Equation (1):

$${}^1{MLCT}^{+}\stackrel{<IRF}{\to }\{\begin{array}{c}{}^3{{MLCT}_1}^{+}\\ {}^3{{MLCT}_2}^{+}\end{array}\stackrel{{\mathsf{\tau }}_1}{\to }\{\begin{array}{c}{T}_1\stackrel{{\mathsf{\tau }}_3}{\to }{S}_0\\ T_2\stackrel{{\mathsf{\tau }}_2}{\to }{S}_0\end{array}$$

(1)

Here, the “+” index denotes vibrationally excited electronic states, inevitably present in the first hundreds of femtoseconds after laser excitation. This scenario was tested with a 50/50 population of T1 and T2 and the species-associated difference spectra (SADS) obtained using GLOTARAN [33] are given in Figure 6C. They provide reasonable difference spectra for these triplet states, with clearly separated bands and transition energies. In the absence of any further knowledge, e.g., from simulations about the ESA spectra of these states, Equation (1) appears to be a viable interpretation of the data. Note that similar excited state branching scenarios, though with different lifetimes and state assignments, were recently published for Fe–NHC complexes with four carbene ligands [34,35].

On the other hand, an alternative purely sequential relaxation scheme can be postulated according to Equation (2):

$${}^1{MLCT}^{+}\stackrel{<IRF}{\to }{}^3{MLCT}^{+}\stackrel{{\mathsf{\tau }}_1 }{\to }{T}_1\stackrel{{\mathsf{\tau }}_2 }{\to }\mathrm{X}\stackrel{ {\mathsf{\tau }}_{3 }}{\to }{S}_0$$

(2)

The corresponding evolution-associated difference spectra (EADS) are given in Figure 6B. The excited state relaxation populates a triplet state T₁ within 0.3 ps, which decays on a 1.3 ps timescale into a state X, which has a 7.3 ps lifetime. While it cannot be excluded that X is an excited state of unknown nature, the EADS of this species suggests its absorption spectrum to be very similar to the GSA of C1, the only differences being a slight 20–30 nm red-shift of the main transitions (see Supplementary Materials). It is therefore very likely that, in this sequential scenario, X represents a vibrationally excited, “hot” form of S₀. Further information comes from TD-DFT calculations, which will be discussed in Section 2.3.

2.2.2. Fs-TAS of C2

C2 dissolved in ACN was excited with ≈50 fs pump pulses at 400 nm, and the pump-induced absorbance changes ΔA, in the range of 350–650 nm are displayed in Figure 7A,B together with the sign-inverted ground state absorption spectrum (GSA) in panel A. The TAS data show a negative ground state bleach (GSB) at all delays probed and over the full wavelength range. A small positive ΔA on the 0.2 ps spectrum may be due to excited state absorption (ESA) from an early decaying state, but since these data are not corrected for cross-phase modulation around time zero, the assignment is unclear. As shown by the kinetic traces (Figure 7C), the bleach signal increases slightly between time zero and ≈1.0 ps, and decays thereafter on a time scale longer than the observation window (1.1 ns). Double exponential fits of traces at individual wavelengths, displayed in Figure 7C, yield a GSB rise time τ₁ of 0.2–0.5 ps, and a decay time τ₂ = 3.2 ± 0.5 ns.

Cast in a global lifetime analysis, values of τ₁ = 0.45 ± 0.1 ps, and τ₂ = 3.0 ± 0.1 ns are found. The DADS and the SADS using a sequential relaxation scheme are displayed in Figure S9. The lifetime and ΔA spectrum of the long-lived species is consistent with the one observed for the ⁵MC quintet state of [Fe(tpy)₂]²⁺ in H₂O [36], in particular the very weak ESA, if any, in the red part of the spectrum distinguishes it from an ³MLCT state (cf. C1, and other Fe–NHC complexes). Note that the negative ΔA signal level <360 nm is much smaller than the expected GSB (inverted GSA), pointing to absorption of the MC quintet, in good agreement with reports probing ⁵MC of [Fe(tpy)₂]²⁺down to 290 nm [37]. In analogy to other polypyridyl complexes of Fe(II), the data are consistent with an ultrafast spin crossover transition from the ^1/3MLCT manifold, identified by the SADS of τ₁ (Figure S3). Hence, the relaxation scheme for C2 is:

$${}^1{MLCT}^{+}\stackrel{<IRF}{\to }{}^3{MLCT}^{+}\stackrel{{\mathsf{\tau }}_1=0.45 ps }{\to }{}^{5}MC\stackrel{ {\mathsf{\tau }}_2 = 3.0ns}{\to }{S}_0$$

(3)

It must be noted, however, that the ultrafast relaxation into ⁵MC leads necessarily to a vibrationally excited quintet. The vibrational relaxation time on a similar spin cross-over complex was found to be in the 10 ps range, as identified by fs mid-IR spectroscopy [38]. Apparently, these cooling processes go unnoticed or do not show up with significant spectral changes in the near UV/VIS for C2.

2.3. Excited-State Decay Mechanism Based on TD-DFT Calculations

The most relevant electronic states involved in the photoresponses of C1 and C2 are displayed in Figure 8. We have adopted the computational methodology used in our previous works on Fe–NHC complexes with bidentate [11,13,14] and tridentate ligands [12] in order to obtain fully comparable results.

As in other transition metal complexes, light absorption populates bright high-lying ¹MLCT states of C1. As commented in the previous sections, ultrafast intersystem crossing leads to the population of the triplet ³MLCT manifold in a sub-picosecond regime, followed by vibrational relaxation to the lowest-lying ³MLCT state (T₁). From a potential energy surface (PES) perspective, all these phenomena take place in regions close to Franck-Condon and S₁ min areas, in which the ¹MLCT and ³MLCT states are also close in energy (see Figure 8). The term S₁ min areas refer to the PES regions where the energy of the S₁ state is minimum. On longer time scales, the triplet state must evolve to reach non-adiabatic crossing points with the ground state that mediate non-radiative decay channels. These mechanisms are intrinsically different between Fe–NHC complexes with bidentate [11,13,14] and tridentate [12] ligands and have different implications for the longer ESA components τ₂ and τ₃. In the former type of complexes, relaxation of the lowest-lying triplet state T₁ leads to a ³MC region through a barrierless path encompassing an extended area of spin-crossover character, i.e., an area where S₀ lies above T₁, in which the system can be trapped for several picoseconds [13]. On the contrary, in complexes with tridentate ligands, the T₁ relaxation leads to a well-defined ³MLCT minimum close to the Franck-Condon region [12]. The stretching of Fe–N bonds, always involved in the stabilization of the dissociative ³MC states, is clearly the dominant degree of freedom in driving the excited state decay. Due to the different topology of the PESs, Fe–N bonds are spontaneously stretched in the triplet manifold of complexes with bidentate ligands, while this molecular motion has an energy cost in Fe–NHC systems with tridentate ligands, explaining the triplet lifetimes of tens of picoseconds in the latter [12].

Despite the tridentate coordination, the ES decay mechanism of C1 lies in between these two extreme cases and somehow closer to the bidentate type mechanism. The T₁ min has a clear ³MC nature (see Figure 9), with Fe–N distance stretched up to 2.60 Å (see

Table 4. Fe–N distances (in Å) corresponding to the structures displayed in Figure 8.

Fe–N	S0 min	S1 min	T1 min	Q1 min
C1
Fe–N1	2.120	2.151	2.604	2.491
Fe–N1′	2.122	2.042	2.417	2.490
C2
Fe–N1	2.013	2.003	2.281	2.190
Fe–N2	1.975	1.960	1.955	2.179
Fe–N3	2.013	2.019	2.283	2.189
Fe–N1′	2.013	2.010	2.121	2.188
Fe–N2′	1.976	1.973	1.937	2.179
Fe–N3′	2.013	2.024	2.120	2.191

). As a matter of fact, the T₁ min geometry is close to but does not perfectly represent a singlet-triplet crossing (STC) with S₀, since the singlet-triplet energy gap is still 0.28 eV, although this relatively low energy splitting suggests that the STC must be in the surroundings. As a matter of fact, the T₁ potential energy surface shown in Figure S10 confirms the access to the singlet-triplet crossing region with the ground state with an energy barrier < 0.05 eV. On the other hand, the energy difference between the T₁–T₃ states at the T₁ min region is smaller as compared to that of previous bidentate Fe–NHC complexes, supporting a possible sequential decay mechanism (T₁/T₂/T₃→S₀) in which τ₂ is ascribed to the population of the ground state and τ₃ belongs to vibrational relaxation of S₀. A parallel mechanism like the one proposed for pure bidentate complexes (see Scheme 1 and refs [11,13,14,39], in which the T₁ and T₂ states are associated to different ES paths operating in parallel, cannot be excluded, although it seems less plausible in this particular case given the T₁–T₃ energy degeneracy at S₁ min and T₁ min regions.

The analysis of the photodynamics of C1 is also of particular value since it reinforces important general conclusions about the photophysics of this type of Fe(II) complexes. In particular, the flexibility added by the presence of alkyl spacers between the pyridine and imidazole rings opens ES decay paths that are forbidden for the more rigid tridentate complex, that is, longer Fe–N elongations leading to accessible ³MC minima. This is reflected in the TAS signals (two long components in bidentates vs. one component in tridentates) and in the photochemical landscape shown in Figure 8. Hence, a subtle equilibrium between the ideality of the octahedral coordination, destabilizing the ³MC states in the Franck-Condon region, and limited flexibility, hampering an efficient population of ³MC states at regions far from the Franck-Condon area via Fe–N elongations, should be carefully considered when designing photoactive iron complexes.

On the contrary, the ES decay mechanism of C2, as also highlighted by TAS results, follows a completely different scenario. A major difference as compared to C1 is the pivotal role played by the much more stable Q₁ state, which is degenerated with T₁ already at the Franck-Condon geometry (see Figure 8). Whereas the optimization of the S₁ (¹MLCT) state destabilizes Q₁, this state is on the contrary greatly stabilized with the Fe–N stretching associated to the T₁ relaxation. Indeed, the Q₁ state becomes degenerated with the ground state at the T₁ min region, which is of ³MC nature (see Figure 9). Full optimization of the Q₁ state gives the Q₁ min, in which the state is at a relative energy close to that of the ground state at its corresponding equilibrium geometry, hence fully justifying the observed light-induced spin crossover capabilities of C2.

Therefore, as stated above, the data presented in Figure 8 indicate that the Q₁ state is the main actor of C2′s photoresponse, in coherence with the TAS analysis depicted above. This state is populated in the Franck-Condon region or in the surroundings in ~0.3 ps and is expected to rapidly evolve towards the Q₁ min region. At this region, the S₀ is not accessible (note that its relative energy is of 1.15 eV, more than 1 eV above Q₁), and therefore one can expect a significant barrier (at least 0.5 eV, according to the solid black arrow shown Figure 8) to repopulate the ground state. This description is in good agreement with the long 3.0 ns lifetime measured experimentally for the quintuplet state.

3. Materials and Methods

3.1. General Information

Solvents and commercially available reagents were used as received. Thin layer chromatography (TLC) was performed by using silica gel 60 F-254 (Merck) plates and visualized under UV light. Chromatographic purification was performed by using silica gel 60 (0.063–0.2 mm/70–230 mesh). ¹H (400 MHz) and ¹³C NMR (100 MHz) spectra were taken on a DRX400 Bruker spectrometer at ambient temperature. The chemical shifts (δ) were calibrated by using either tetramethylsilane (TMS) or signals from the residual protons of the deuterated solvents and are reported in parts per million (ppm) from low to high field. High-resolution mass spectrometry (HRMS) data was obtained by using Bruker micrOTOF-Q spectrometer. UV vis spectra were recorded in a 1 cm path length quartz cell on a LAMBDA 1050 (PerkinElmer), spectrophotometer. Cyclic voltammetry was performed on a Radiometer PST006 potentiostat using a conventional three-electrode cell. The saturated calomel electrode (SCE) was separated from the test compartment using a bridge tube. The solutions of studied complexes (0.2 mM) were purged with argon before each measurement. The test solution was acetonitrile containing 0.1 M Bu₄NPF₆ as supporting electrolyte. The working electrode was a vitreous carbon rod (1 cm²) wire, and the counter-electrode was a 1 cm² Pt disc. After the measurement, ferrocene (Fc) was added as the internal reference for calibration. All potentials were quoted versus SCE. In these conditions the redox potential of the couple Fc+/Fc was found at 0.39 V. In all the experiments the scan rate was 100 mV.s⁻¹.

3.2. Xray Diffraction

Two single crystals of complexes C1 and C2 were selected and measured with a Rigaku Oxford Diffraction SuperNova 4-circles diffractometer. This diffractometer was equipped with a microfocus X-ray source with a Molybdenum anode (MoKα, λ = 0.71073 Å) and an Atlas CCD detector. An Oxford Cryosystems Cryostream nitrogen blower was used to fix the sample temperature at 100 K for C1 and 110 K for C2. Using Olex2 software [40], structures were solved with XT structure solution program [41] and structures were refined with the XL refinement package [41] by using Least Square minimization. CCDC 1992132 (C1) and CCDC 1982173 (C2) contain the supplementary crystallographic data for this paper. These data are provided free of charge by the Cambridge Crystallographic Data Centre.

3.3. Ultrafast Transient Absorption Spectroscopy

An amplified 5kHz Ti: sapphire laser generates 30 fs 0.5 mJ pulses to pump a commercial optical parametric amplifier (TOPAS: Light Conversion), from which 50–60 fs pulses are derived to excite the samples: 495 nm for C1, and 400 nm for C2. A white light continuum is generated in a 2 mm thick CaF₂ crystal, mounted on an oscillating loudspeaker to reduce photo-damage. The white light is optimized for minimized fluctuations (rms < 0.4% at 220 Hz). The beam is split in two: the probe that is sent through the sample, and a reference beam is used for measuring and compensating the white-light intensity fluctuations. The polarization of the probe beam is set at magic angle (54.7°) with respect to the pump. A 1 mm path length cuvette in fused silica contains the complexes dissolved in CH₃CN. Time-resolved spectra I (λ) are acquired in a range ≈300 nm wide window with an adjustable central wavelength as a function of pump-probe delay, by a combination of 25 cm focal length spectrometer (resolution 2 nm) and a Peltier-cooled CCD with 220 Hz acquisition rate. A second chopper blocks the pump beam at 110 Hz and a home-made software computes and displays the differential spectra $\Delta A=-lo{g}_{10}\left(\frac{I_{pump-on}}{I_{pump-off}}\right)$. A solvent-only sample is measured, and the data are processed to (i) remove a residual signal background at negative delay times, (ii) the solvent Raman signal and the coherent interactions of the pump and probe in the cell (except for C2), (iii) to correct for the group velocity dispersion of the probe beam, characterized in the solvent-only data set. The temporal resolution is determined by the 50±5 fs FWHM of the solvent Raman response. Data are fitted at individual wavelengths according to $\Delta A\left(\lambda ,t\right)=IRF\left(t_0,\sigma \right)\otimes \left(A_0+{{\displaystyle \sum }}_{i=1}^n{A}_i\left(\lambda \right)e^{-t/{\tau }_i\left(\lambda \right)}\right)$, (1) a sum of n-exp functions and an un-resolved contribution (A0), convoluted with a Gaussian instrument response function (IRF, σ = 35–50 fs, FWHM). Although data are chirp-corrected and t0, is adjustable for each wavelength. Global and target analysis was performed with the software packages OPTIMUS [32] and GLOTARAN [33].

3.4. Computational Details

The computational protocol has been widely validated in previous works and it will be only summarized here [11,12,13,14,39,42]. Unless otherwise stated, all calculations have been performed using the Gaussian 16 program [43]. The ground state geometries have been optimized using the DFT/B3LYP method, whereas the S₁ state has been optimized using TD-DFT in combination with the pure HCTH functional [44]. The T₁ and Q₁ states have been optimized using the unrestricted uDFT/HCTH method. All optimizations have been performed employing the 6-31+G(d,p) basis set. Later, single-point calculations with the larger 6-311G(d,p) basis set have been performed on top of the converged geometries to obtain the final energies displayed in Figure 8. The energies of S₁ and T₁ have been obtained with the TD-DFT/HCTH method, while the energies of Q₁ have been determined making use of the uDFT/HCTH ansatz. Solvent effects (acetonitrile) have been included by means of the polarizable continuum model. The Tamm-Dancoff approximation has been used in all TD-DFT calculations [45]. For C2, due to the inherent π-stacking interaction of the ligands, empirical dispersion has been included in all calculations through Grimme’s original D3 damping function [46]. Analysis of the nature of the excitations (Figure 9) have been performed by computing the natural transition orbitals (NTOs) [47] with the NANCY_EX code [48,49].

3.5. Synthesis of Ligands and Complexes

3.5.1. Synthesis of Ligands

Synthesis of L1 [31]. A solution of 2,6-bis(bromomethyl)pyridine (1.337 g, 5.0 mmol) and 1-methylimidazole (0.831 g, 10.0 mmol) was stirred in dioxane (80 mL) at 100 °C for 12 h. After cooling, the formed solid was collected and purified by repetitive precipitation from MeOH/Et2O mixtures and finally by recrystallization from CH₂Cl₂/Et₂O. The product was then collected and solubilized in a small amount of water, then precipitated using a saturated aqueous solution of KPF₆ to give L1 (2.27 g, 80%). ¹H-NMR (CD₃CN, 400 MHz): δ 8.47 (s, 2H), 7.90 (t, J = 7.8 Hz, 1H), 7.44 (d, J = 7.8 Hz, 2H), 7.35 (s, 4H), 5.38 (s, 4H), 3.85 (s, 12H) ppm.

Synthesis of dqp [18]. An oven-dried flask was charged with 8- quinoline boronic acid (413 mg, 2.39 mmol), 2,6-dibromo-pyridine (265mg, 1.12mmol), Pd(dba)2 (13mg, 0.023mmol), 2-dicyclohexylphosphino-2’,6´-dimethoxybiphenyl (0.019 g, 0.046 mmol) and grinded K₃PO₄ (2.27 g, 10.7 mmol). The flask was evacuated and charged with N₂. Dry toluene (10 mL) was added via a syringe and the resulting suspension was stirred at 100˚C for 15 h. The mixture could cool to room temperature, diluted with CH₂Cl₂ (30 mL) and filtered. The solvent was removed and the remaining solid purified by column chromatography using silica gel and 2.5% MeOH/CH₂Cl₂ as eluent to give dqp as an off-white solid (244mg, 65%). ¹H-NMR (CDCl₃, 400 MHz,): δ 9.00 (dd, J = 4.2, 1.9 Hz, 2H), 8.28 (dd, J = 7.3, 1.5 Hz, 2H), 8.24 (dd, J = 8.3, 1.9 Hz, 2H), 8.13 (d, J = 7.7 Hz, 2H), 7.96 (t, J = 7.7 Hz, 1H), 7.89 (dd, J = 8.2, 1.5 Hz, 2H), 7.67 (dd, J = 8.2, 7.3 Hz, 2H), 7.46 (dd, J = 8.3, 4.2 Hz, 2H) ppm.

3.5.2. Synthesis of Complexes

Synthesis of C1. To a solution of L1 (0.129 g, 0.231 mmol) in 2 mL of anhydrous DMF was added FeBr₂ (15 mg, 0.115 mmol) and degassed the mixture with N₂ for 10 min. Then t-BuOK was added (104 mg, 0.927 mmol) to the above mixture and stirred at 0 °C and temperature left slowly rising to room temperature. A saturated solution of KPF₆ was added (10 mL), and the precipitate was collected by filtration. Then the crude was further purified on silica gel column chromatography using acetone/H₂O/KNO₃ (sat) = 10:3:3 mixture. The pink fraction was collected and after the evaporation of acetone, the left solution was treated with a saturated solution of KPF₆. C1 (20 mg, 19%) was finally obtained after filtration. ¹H-NMR (CD₃CN, 400 MHz): δ 7.58 (t, 2H, J = 7.7 Hz), 7.30 (d, 4H, J = 7.7 Hz), 7.07 (d, 4H, J = 1.9 Hz), 6.67 (d, 4H, J = 1.9 Hz), 5.41 (s, 4H), 3.40 (s, 4H), 2.05 (s, 12H) ppm. ¹³C-NMR (CD₃CN, 100 MHz): δ 191.4, 157.3, 137.3, 126.3, 124.2, 120.1, 53.5, 34.5. HR-MS (ESI) calcd for C₃₀H₃₄FeN₁₀P₂F₁₂ m/z = 295.1153 [M – 2 PF₆]²⁺. Found: 295.1145.

Synthesis of C2. A 5 mL microwave tube was charged with dqp (65 mg, 0.195 mmol) and FeBr₂ (20mg, 0.093 mmol). After addition of DMF (3 mL) the tube was sealed and heated for 4h using microwave heating (60W). A saturated solution of KPF₆ was added (10 mL), and the precipitate was collected by filtration. Then the crude was further purified on silica gel column chromatography using Acetone/H₂O/KNO₃ (sat) = 10:3:3 mixture. The dark pink fraction was collected and after the evaporation of acetone, the left solution was treated with a saturated solution of KPF₆. C2 (28 mg, 30%) was finally obtained after filtration. ¹H-NMR (400 MHz, CD3CN): δ 8.17 (dd, J = 8.2, 7.8 Hz, 2H), 8.07 (m, 8H), 7.89 (m, 4H), 7.73 (dd, J = 7.5, 1.3 Hz, 4H), 7.68 (dd, J = 8.3, 1.3 Hz, 4H), 7.45 (dd, J = 8.3, 7.5 Hz, 4H), 7.05 (dd, J = 7.9, 5.3 Hz, 4H) ppm. ¹³C-NMR (DMSO-d6, 100 MHz): δ 164.50, 164.25, 161.34, 149.95, 138.89, 138.68, 132.36, 130.84, 128.10, 127.88, 127.32, 126.25. HR-MS (ESI) calcd for C₄₆H₃₀FeN₆P₂F₁₂ m/z = 361.0935 [M – 2 PF₆]²⁺. Found: 361.0944.

4. Discussion and Conclusions

The present study of Fe(II) homoleptic tridentate complexes C1 and C2 with octahedral coordination symmetry aimed at providing evidence for the effect of a larger ligand field splitting on the excited state (ES) dynamics, in particular aiming at obtaining longer-lived ³MLCT states. The change in geometry had a positive effect on the absorption properties inducing notable red shifts of the Fe-pyridine MLCT absorption bands compared with the non-octahedral C0, in agreement with the decrease of the HOMO-LUMO gap also confirmed by cyclic voltammetry. However, as pointed out above, the ligand modifications also bring about a larger conformational flexibility, the effect of which appears to dominate the ES quantum dynamics.

When comparing C1 with the proto-typical C0 ([Fe(bmip)₂]²⁺) two important differences are observed. First, the ES lifetime of C1 is roughly six times shorter than that of C0. The TD-DFT calculations indicate that, due to the higher ligand flexibility, the barrier hindering the relaxation of ³MLCT into the metal-centered ³MC in [Fe(bmip)₂]²⁺ is lowered in C1, leading to a faster ES deactivation (Figure 10 center). The TD-DFT calculations indicate that, most likely, one single triplet state minimum of ³MC character is transiently populated (Figure 8) hinting to the “sequential decay” scenario being more plausible. Secondly, a strongly damped excited state vibrational coherence is observed with a frequency of 55–65 cm⁻¹, in addition to ground state vibrations in the range of ≈350 cm⁻¹ (cf. Figure S8). The higher frequency is in the usual range of Fe–N or Fe–C stretch vibrations, and the 60 cm⁻¹ is similar to the low frequency breathing mode identified for C0 at 110 cm⁻¹ [50]. It was recently identified to be associated with vibrationally coherent population of ³MC through ultrafast branching from the ¹MLCT manifold in C0 [36]. We rather attribute it to the initially populated ³MLCT. In any case, its existence and lower frequency point to a higher flexibility of the ligands of C1 as compared to C0. For C2, the relaxation from ³MLCT into the metal-centered ³MC and ⁵MC states (Figure 10, left) seems to be almost unaffected since the 0.45 ps lifetime is only slightly longer than the one observed for [Fe(tpy)₂]²⁺ [8].

In summary, the comparison of the photophysics of C1 and C2 evidences important aspects about the Fe(II) coordination sphere directly impacting ES decay mechanisms and ES lifetimes of Fe–NHC complexes. The main conclusion is that the destabilization of ³MC states at the Franck-Condon region does not guarantee longer triplet MLCT lifetimes since these states quickly relax out of this region if the system is flexible enough, as demonstrated by the current C1 complex. A similar conclusion was drawn from our previous comparison of bidentate vs. tridentate coordination where the associated rigidity dictates the participation of the triplet states and their ability to stretch Fe–N bonds [14]. Longer ES lifetimes may be achieved by blocking the elongations of the Fe–N bonds [51] or through their substitution by more photoresistant bonds such as Fe–C ones [15,52,53,54].

In the case of C2, while the increase of the rigidity of the molecular scaffold could allow the improvement of the photophysical properties, the presence of six Fe–N bonds strongly diminishes the ligand-field splitting and hence strongly shortens the ES lifetime due to the stabilization of the metal centered quintet state that is actively participating in the relaxation. All in all, these contrasting observations clearly points to the subtle equilibrium between the different electronic and geometrical factors susceptible of altering the photophysical outcome of iron(II)-based complexes. The competition between the different factors needs to be fully considered when proceeding to rational molecular design strategy for improving the optical response.