Electron delocalization in the S₁ and T₁ metal-to-ligand charge transfer states of trans-substituted metal quadruply bonded complexes

Abstract

The singlet S₁ and triplet T₁ photoexcited states of the compounds containing MM quadruple bonds trans-M₂(TⁱPB)₂(O₂CC₆H₄-4-CN)₂, where TⁱPB = 2,4,6-triisopropylbenzoate and M = Mo (I) or M = W (I^′), and trans-M₂(O₂CMe)₂((N[ⁱ Pr ])₂CC ≡ CC₆H₅)₂, where M = Mo (II) and M = W (II^′), have been investigated by a variety of spectroscopic techniques including femtosecond time-resolved infrared spectroscopy. The singlet states are shown to be delocalized metal-to-ligand charge transfer (MLCT) states for I and I^′ but localized for II and II^′ involving the cyanobenzoate or amidinate ligands, respectively. The triplet states are MoMoδδ* for both I and II but delocalized ³MLCT for I^′ and localized ³MLCT for II^′. These differences arise from consideration of the relative orbital energies of the M₂δ or M₂δ* and the ligand π^∗ as well as the magnitudes of orbital overlap.

Conjugated organic polymers have been intensely studied over the past two decades because of their fascinating optoelectronic properties. Aside from the sheer scientific curiosity that they aroused, it soon was recognized that a plastic electronics industry was possible and a good deal of this expectation has already been realized. Conjugated organic polymers find commercial applications as field-effect transistors (1), light emitting diodes (2), and photovoltaic devices (3, 4). As an extension of this field there is considerable interest in incorporating metal ions into conjugated organic systems and numerous reports are to be found in the literature concerning the role of metal ions in tailoring the optoelectronic properties of the organic conjugated systems. For example, the attachment of π-conjugated ligands to Ir(III) has allowed the luminescence to cover the entire region of the visible spectrum (5, 6) and the incorporation of Pt(II) into conjugated ethynylthiophenes has led to a significant enhancement of the efficiency of a bulk heterojunction photocell (7).

Knowledge of electronic structure is key to the understanding of these properties and the ability to manipulate electronic structure by selection of metal-organic components holds the promise of custom design. The MM quadruply bonded unit (MM = Mo₂, MoW, or W₂) has many attractive features due to the tunability of the energy of the M₂δ orbital by selection of the metal and its attendant ligands and due to M₂δ to organic π-conjugation. In this report, we show how this can influence the charge dynamics and delocalization of singlet and triplet photo-excited states that may be delocalized or localized (valence trapped) metal-to-ligand charge transfer (MLCT) states or MMδδ* states. To achieve this we have prepared the trans-substituted compounds M₂(TⁱPB)₂(O₂CC₆H₄-4-CN)₂, where TⁱPB = 2,4,6-triisopropylbenzoate and M = Mo (I) or W (I^′), and M₂(O₂CMe)₂((N[ⁱ Pr ])₂CC ≡ CC₆H₅)₂, where M = Mo (II) or W (II^′), whose structures are shown in Fig. 1. Here, the bulky TⁱPB and amidinate ligands favor the trans-substitution and allow the extended conjugation of the two trans-ligands via Lπ-M₂δ-Lπ conjugation. Evidence of this is seen in the molecular structures found in the solid-state that reveal the near coplanarity of the aryl groups of the p-cyanobenzoate and amidinate ligands in I and II^′, respectively.

Fig. 1.

ORTEP diagram of I•2THF (Left) and II^′ (Right) drawn with 50% probability displacement ellipsoids with hydrogen atoms omitted. Structure I•2THF contains a crystallographic inversion center. Structure II^′ was reproduced from ref. 26 (reproduced by permission of the Royal Society of Chemistry).

The ground state geometry and symmetry of these discrete compounds presents a situation where added charge could reside on either of two interchangeable redox active ligand sites. This situation is, by definition, described by the term mixed valency. In the case of compounds I, II, and their tungsten analogs I^′ and II^′, an electron can be added to the ligand sites by photoexcitation of the MLCT transition and they therefore can be further classified as excited state mixed valence compounds (8). Depending on how strongly the ligand sites interact with one another, the excited electron can be completely localized (valence trapped) on one ligand, completely delocalized over both ligands, or only partially delocalized.

These various classifications of the mixed-valence states can be visualized by the potential energy surfaces (PESs) shown in Fig. 2. When the ligands are noninteracting, there are two isoenergetic PESs in the MLCT, S₁ state. This is known as Class I under the Robin and Day scheme (9). As the electronic coupling, H_ab, begins to increase, the ligand states interact to give a lower, in-phase (S₁), and upper, out-of-phase (S₂), combination of the surfaces. When the coupling is small, two distinct ligand states remain and the lower PES is a double well. An electron in this state can be said to be localized on a single ligand and is known as Class II, shown in Fig. 2. An excited state transition can occur within this mixed-valence state that overcomes the barrier between the two surfaces and transfers the electron between the ligand sites, analogous to the intervalence charge transfer (IVCT) band of ground-state mixed-valence compounds (10). As the strength of the coupling increases further, the two states have greater mixing and the electron can begin to become delocalized over both ligands. As a result, the PESs move apart and the potential energy barrier between the two sites decreases. Eventually, a point is reached where there is no longer a barrier and the electron is nearly equally shared. This occurs at the class II/III border and is shown in the right panel of Fig. 2. When the electron finally becomes completely delocalized, class III, the PESs represent completely bonding and antibonding molecular orbitals and the transition between S₁ and S₂ becomes like that for a molecular species. In this case the IVCT transition becomes known as a charge resonance band (11).

Fig. 2.

Potential energy surface representation of the mixed-valence states for a symmetrical system. S₁ represents the state where an excited electron occupies the in-phase L-π^∗ and S₂ the state where an excited electron occupies the out-of-phase L-π^∗ combination. (Left) Class I; weakly coupled localized. (Right) Class III strongly coupled, delocalized. (Center) Class II-II/III; 0 < a < 1. Modified from ref. 11.

The electronic coupling of the two organic π-systems arises from the interaction of the M₂δ orbital with the carboxylate, CO₂, or amidinate, NCN, π orbitals shown in A and B of Scheme 1, below. The orbital interactions shown in A involve the in- and out-of-phase combinations of the π^∗of the CO₂ or NCN moiety. The out-of-phase combination has a symmetry match with the M₂δ orbital and is a metal-to-ligand backbonding interaction. On the other hand, the interactions in B involve the filled orbitals of the carboxylate or amidinate ligands and based on orbital energetics it is the former interaction, A, that is more important. Given that O is more electronegative than N, the π^∗ of the carboxylates is lower in energy than the amidinate and thus the mixing with the M₂δ orbital and the coupling between the trans-ligands is greater in the p-cyanocarboxylates. Also, because the M₂δ orbital in closely related compounds is roughly 0.5 eV higher in energy for M₂ = W₂ relative to Mo₂, the interactions in the tungsten complexes are greater.

Scheme 1.

These qualitative bonding descriptions are supported by density functional theory (DFT) calculations that are described in SI Text and Fig. S1. The calculations reveal the energy splitting of the in- and out-of-phase ligand π* combinations due to mixing with the M₂δ orbital. Also consistent with the calculations, electrochemical studies reveal metal based oxidation and, for I and I^′, ligand based reduction waves by cyclic voltametry. Reduction waves for II and II^′ were not seen within the solvent window due to the higher energy of their π* orbitals.

Results and Discussion

Photophysical Studies: Electronic Spectroscopy.

All of the compounds are strongly colored due to MLCT and their electronic absorption spectra are shown in Fig. 3. Though allowed by symmetry, the ¹MM δδ* transition is weak due to the relatively poor overlap of the metal d orbitals. In each case, the lowest energy transition is characterized as MLCT (12, 13). Although the planar D_2h structure of these molecules is the ground state structure, thermal energy leads to a Boltzmann distribution of conformers/rotamers in which the aryl groups of the ligand deviate from coplanarity with the carboxylate CO₂ or amidinate NCN moieties. The room temperature solution spectra represent the combined absorptions of an ensemble of molecules, which together with vibronic features arising from the displacement of the ground and excited state potential energy surfaces leads to the broadening of these MLCT bands. Upon cooling to 77 K in a 2-methyltetrahydrofuran glass, these bands sharpen with the peak shifting to lower frequency and the vibronic features becoming better resolved, as shown in Figs. S2 and S3. At higher energy to these MLCT bands associated with the conjugated ligands, we also observe weaker MLCT bands involving the CO₂ moieties of the TⁱPB (in I^′) and acetate ligands (in II^′) and ligand based ππ* transitions.

Fig. 3.

Electronic absorption spectra showing the molar absorptivity (ϵ) of I (red) I^′ (purple), II (green), II^′ (blue) at room temperature in THF.

All of the compounds are emissive and compounds I (Figs. S3 and S4), II (14), and II^′ (14) each emit from both the S₁ and T₁ states. In compound I^′, the T₁–S₀ energy gap is estimated to be small and this favors nonradiative decay so only emission from the S₁ state is observed (Figs. S3 and S4). Interestingly, the S₁ states of both II and II^′ show a solvent dependent emission whereas the MLCT absorptions do not. Furthermore, the phosphorescence of the tungsten complex (II^′) shows a solvent dependence but that of the molybdenum compound (II) does not. The more limited range of solubilities of compounds I and I^′ made the solvent dependent study less clear. However for II, our interpretation of the solvent dependence is that the initial absorption occurs without a significant change in dipole moment whereas that associated with the emission from the S₁ states is significant (14). Because the ground state molecules are symmetrical (D_2h) and have a center of inversion, the Franck–Condon absorption conserves the point group symmetry to place the excited electron delocalized over both ligands. Further relaxation of the Franck–Condon S₁ states in polar solvents leads to the localization of the excited electron mostly on an individual ligand, increasing the dipole moment of the S₁ state and causing the emission solvent polarity dependence.

Similarly, T₁ emission for II^′ (W) is solvent dependent but emission from T₁ for II (Mo) is not. At low temperature, the emissive T₁ state for II shows vibronic features assignable to the MoMo stretching frequency with ν(MoMo) ∼ 400 cm^-1 (14). The vibronic spacing together with the lack of solvent dependence in the T₁ state emission for II is consistent with the assignment of this state as the ³MoMoδδ^∗ state. Furthermore we should note that the lowest energy emission for II^′ also shows solvent dependence and suggests an assignment of ³MLCT for the T₁ state of this complex. Again, our evaluation of the electron delocalization within the MLCT excited states is further substantiated by the femtosecond time-resolved infrared (fsTRIR) spectroscopy experiments, below, which also confirm our assignments of the S₁ and T₁ states as MLCT or δδ*.

Time-Resolved Studies.

We have examined the compounds by femtosecond and nanosecond transient absorption spectroscopy (fsTA and nsTA), detecting both the S₁ and T₁ states in all cases. The lifetimes for the S₁ MLCT states measured for these types of quadruply bonded compounds are generally on the order of 10 ps (15–17) and the compounds under investigation here range from approximately 3–20 ps. On the other hand, the lifetimes of the T₁ states generally range from 0.010–100 μs depending largely upon the metal and give an indication of their nature as ³MLCT or ³δδ∗ (12, 13, 16, 18). For carboxylate complexes, if the T₁ state is ³MLCT, the lifetime is on the order of hundreds of nanoseconds or less. In the case of I^′, the T₁ state is detected in the fsTA as a long-lived transient but the lifetime is shorter (τ < 10 ns) than the time resolution of the nsTA experiment. On the other hand, if the T₁ state is ³δδ^∗, the lifetimes are longer, lasting for approximately 100 μs. In amidinate complexes, the lifetimes are observed to increase as a result of the ligand based orbitals originating at higher energy compared to the carboxylate ligands. In II^′, the T₁ state has τ = 4.6 μs and in II it is approximately 105 μs. In general we find that T₁ for Mo complexes is ³δδ^∗ and the lifetimes are 3–4 orders of magnitude longer than T₁ for W complexes that are ³MLCT. The transient absorption spectra along with representative kinetic traces for I and I^′ are shown in Fig. S5 and those for II and II^′ are shown in ref. 14. Based upon the absorption, emission, and transient absorption results we can formulate a Jablonski diagram describing the general photophysics of the four compounds, and this is shown in Fig. 4.

Fig. 4.

Jablonski diagram showing excited states in compounds I (red), I^′ (purple), II (green), and II^′ (blue) observed after MLCT photoexcitation.

Photophysical Studies: Vibrational Spectroscopy (fsTRIR).

Particularly convenient for studying the nature of these types of photoexcited states is the presence of the IR active C ≡ N and C ≡ C functional groups within the ligands. As a consequence, the compounds have also been subjected to fs time-resolved, visible pump, IR probe (fsTRIR) spectroscopy.

The p-cyanobenzoates, I and I^′, show ν(C ≡ N) ∼ 2,230 cm^-1 in the ground state and upon excitation a single new band occurs that is shifted to lower energy by approximately 60–70 cm^-1, as shown in the time-resolved infrared (TRIR) spectra above, Fig. 5. In the case of I (Mo), this band decays completely with a lifetime of approximately 3.9 ps, similar to that of the ¹MLCT state determined by fsTA, and no new transient band is observed thereafter. In the case of I^′, however, this band decays with a lifetime of 7.2 ps (again similar to that of the ¹MLCT state lifetime by fsTA) to give rise to a long-lived transient that is present after 1,000 ps at nearly the same wavenumber. This result is a clear indication that for I^′ (W) both the S₁ and T₁ states are MLCT and further suggests that the charge distribution in both the singlet and triplet states is similar. The spectroscopic features of the ligands in MLCT excited states commonly resemble those of the reduced ligand. For this reason, we have completed DFT calculations on the model anions to predict the change in ν(C ≡ N) upon reduction of the ligand. Our expectation is that, because the anion calculations were done at the optimized ground state geometry with D_2h symmetry, the calculated shift predicts the experimental shift in the situation that the photoexcited electron is delocalized over both ligands (Class III). For both compounds I and I^′, the calculated shift is approximately 65 cm^-1. Because the experimental values are similar, it implies that the ¹MLCT excited states for both complexes as well as the ³MLCT excited state for I^′ are delocalized or Class III mixed-valence (MV) ions in the Robin and Day scheme. The results are also consistent with the expectation that in a fully delocalized excited state both of the trans-ligands are equivalent and both should have identical stretching frequencies. A summary of the TRIR experimental and DFT calculated C ≡ N frequencies are shown in Table 1 and representative kinetic traces from the fsTRIR experiments are shown in Fig. S6.

Fig. 5.

TRIR spectra for I (Upper) and I^′ (Lower) in THF at room temperature.

Table 1.

Summary of ν(C ≡ N) values (cm^-1)

	Experimental (TRIR)			Calculated (DFT)
	Ground state	Excited state	Shift	Neutral	Anion	Shift
Compound	ν(C ≡ N)	ν(C ≡ N)∗	Δν	ν(C ≡ N)	ν(C ≡ N)-	Δν
I (1MLCT)	2,230	2,157	73	2,345	2,278	67
I′ (1MLCT)	2,225	2,164	61	2,343	2,279	64
I′ (3MLCT)	2,225	2,162	63	—	—	—

Relatively few examples occur in the literature of excited state infrared spectra for molecules involving cyano groups. One example that does exist is 4-dimethylaminobenzonitrile (DMABN). It is generally agreed that, in polar solvents, the lowest excited state of DMABN is described by the transfer of an electron from the amino group to benzonitrile and there is a 90° dihedral angle between these groups, which has been named the twisted-intramolecular charge transfer state (TICT) state (19). There are two reports of the TRIR in the C ≡ N stretching frequency region of the TICT state for this molecule (20, 21) In THF, the shift downward is 102 cm^-1, significantly larger than that observed for I and I^′. This further supports the view that the excited electron is delocalized across two cyanobenzoate ligands in I and I^′ whereas in DMABN the electron can occupy only one benzonitrile unit.

The vibrational spectroscopic features associated with the compounds II and II^′ are even more interesting. In the ground state these compounds show weak IR bands close to 2,200 cm^-1 associated with ν(C ≡ C). Upon photoexcitation, the S₁ states of these compounds have two distinct transient bands that have both been shifted to lower wavenumber than that in the ground state, see Fig. 6. One band has weaker intensity at approximately 2,150 cm^-1 and the other, with much stronger intensity, occurs at much lower wavenumber between 1,950 and 2,000 cm^-1. Similar to the p-cyanobenzoates, these initial transient bands decay with lifetimes 18.8 and 6.8 ps for II and II^′ (Fig. S7), respectively, which is in agreement with the ¹MLCT lifetimes determined by fsTA. For II (Mo), the transient decays completely and no new transient band is observed thereafter. On the other hand, the decay of the initial transients at 2,150 and 1,990 cm^-1 for II^′ (W) is accompanied by the formation of a new set of transient bands at 2,115 cm^-1 and 1,940 cm^-1 that are long-lived and persistent. The presence of these persistent transient bands in II^′ clearly indicates that the T₁ state is ³MLCT and the absence of these bands in II implies that the T₁ state can be assigned to ³MoMoδδ^∗.

Fig. 6.

TRIR spectra for II (Upper) and II^′ (Lower) in THF at room temperature.

Another interesting feature of the TRIR spectra for II^′ is that the transient bands associated with ν(C ≡ C) in the ³MLCT state are shifted to even lower wavenumber than those for the corresponding ¹MLCT state. This suggests that in the T₁ state the negative charge is more localized on the CC triple bond relative to the S₁ state. It is also consistent with the general view that singlet states are more diffuse than triplet states, which are lower in energy and more tightly bound.

To evaluate the electronic coupling within the Robin and Day scheme for the compounds of type II, a set of calculations on the model anions were again completed. These calculations predict a shift for ν(C ≡ C) of approximately 130 cm^-1 to lower wavenumber in the reduced anion. This shift is notably larger than what was calculated for the p-cyanobenzoates. In Fig. S6, the molecular orbital (MO) picture for the lowest unoccupied molecular orbital (LUMO), as determined from DFT, shows that the electron density has a much greater contribution from the C ≡ C antibonding orbital in II^′ than that contribution from the C ≡ N antibonding orbital in I^′. Therefore, the observed shifts are larger for the ligands with C ≡ C moieties compared to the C ≡ N moiety. As stated above, the MO calculations pertain to the delocalized model or MV ion of Class III. The observed TRIR shift of the major peak in the compounds II and II^′ is notably greater than the calculated value, which together with the solvent dependent emission implicates a polarized MLCT state where one ligand has been mostly reduced by the photoexcited electron with only a small “spillover” to the other ligand. In other words, the amidinate complexes, II and II^′, exist as only partially delocalized mixed-valence Class II ions in both their relaxed ¹MLCT and ³MLCT states. The observation of two distinct transient peaks in the TRIR spectra is an experimental signature of this situation. Also, recall the earlier statement that the Franck–Condon absorption is to a delocalized state. Because the initial spectra after excitation contain two peaks, it appears that the charge localization process that places the excited electron mostly on one ligand occurs more rapidly than can be detected and within the lifetime of the ¹MLCT states.

The specific C ≡ C stretching wavenumbers determined for compounds II and II^′ are collected in Table 2 along with the calculated wavenumbers for the model compounds and their corresponding anions. There it can be seen that Δν₂ for the ¹MLCT state of II is notably larger than that for II^′, which may be an indication of the more C ≡ C triple bond localized nature of the charge in the ¹MLCT state for the molybdenum compound.

Table 2.

Summary of ν(C ≡ C) values (cm^-1)

	Experimental (TRIR)					Calculated (DFT)
	Ground state	Excited state		Shift		Neutral	Anion	Shift
Compound	ν(C ≡ C)	ν1(C ≡ C)∗	ν2(C ≡ C)∗	Δν1	Δν2	ν(C ≡ C)	ν(C ≡ C)-	Δν
II (1MLCT)	2,200 *	2,155	1,959	45	241	2,293	2,152	141
II′ (1MLCT)	2,180	2,152	1,989	28	191	2,281	2,153	128
II′ (3MLCT)	2,180	2,116	1,942	64	238	—	—	—

^*Value obtained from Raman spectrum due to weak IR intensity.

There have been several reports of excited state infrared spectra for square-planar transition metal complexes of platinum involving phenylacetylide ligands that monitor the C ≡ C triple bond stretching wavenumber (22–25). Of these, the most notable comparison to compounds II and II^′ is trans-Pt[(CCPh)₂](PBu₃)₂ in its ³MLCT state(22). This molecule has a highest occupied molecular orbital (HOMO) that is composed of ligand π orbitals with some admixture of Pt 5d_xy and the LUMO is totally ligand π*. Photoexcitation is thus a mixture of ππ* and MLCT and this yields a triplet state that has a transient IR band assignable to ν(C ≡ C) that is shifted approximately 360 cm^-1 to lower wavenumber and another that is shifted to higher wavenumber approximately 10 cm^-1 relative to the ground state ν(C ≡ C) value. This is consistent with a localized triplet state arising from ππ* + MLCT where one phenylacetylide is in the photoexcited state and the other feels the positive charge on the platinum that lessens the Pt 5d to ligand π* backbonding. It is thus evident that the higher energy M₂δ orbitals and the smaller HOMO-LUMO gap for MM quadruply bonded complexes of Mo or W allows for greater coupling of the two ligand π* systems.

Concluding Remarks

The nature of the photoexcited states in compounds I, II, and their tungsten analogs can be correlated with the electronic structures of the molecules in their ground states. First, the lower energy of the Mo₂ δ and δ* orbitals leads to the T₁ states being ³MMδδ^∗ and this can be expected to be generally the case unless the ligand π* orbital is very low in energy and the ¹MLCT absorption falls in the near infrared. Conversely, the higher energy of the W₂ δ and δ* will lead to T₁ being ³MLCT unless the energy of the ligand π* orbital is notably high in energy and the ¹MLCT absorption is in the high energy region of the UV/visible spectrum. Second, the coupling of the two trans ligands is dependent primarily on the energy separations between the M₂δ and the ligand based LUMO π* orbital. The smaller the energy gap the greater the coupling and this can lead to fully delocalized S₁ and T₁ MLCT states. As the coupling of the two ligands via the M₂δ decreases, the photoexcited state can take on mixed valence Class II characteristics. The initial photoexcitation may be to a delocalized state that rapidly relaxes to the charge localized Class II MV state. This is seen for the amidinate compounds and their ground state and photoexcited state potential energy surfaces can be represented by Fig. 2, above. Third, the extent of charge delocalization in the singlet and triplet MLCT states is in general expected to be different with the higher energy S₁ state being more diffuse than the lower energy T₁ state. This was very nicely revealed in the TRIR spectra of II^′.

Collected in Table 3 is a summary for compounds I, I^′, II, and II^′ of the S₁ and T₁ state assignments and their classifications under the Robin and Day scheme. It is particularly appealing that these four molecules have revealed these limiting cases. Furthermore, based on considerations of orbital energies involving the M₂δ and the ligand π* we anticipate these findings will be applicable to the photoexcited states of conjugated polymers incorporating MM quadruple bonds (15).

Table 3.

Summary of photophysical assignments and Robin and Day classification

Compound (M)	State	Assignment	Classification
I (Mo)	S1	1MLCT	Class III
	T1	3δδ∗	N.A.
I′ (W)	S1	1MLCT	Class III
	T1	3MLCT	Class III
II (Mo)	S1	1MLCT	Class II
	T1	3δδ∗	N.A.
II′ (W)	S1	1MLCT	Class II
	T1	3MLCT	Class II

Materials and Methods

All reactions were carried out in an inert atmosphere using Schlenk line techniques whereas samples for measurements were prepared inside a glove box. All solvents were dried, distilled, and degassed prior to use. Steady state absorption, emission, and infrared spectra, nanosecond and femtosecond transient absorption spectra, electrochemistry, electronic structure calculations, fsTRIR, and crystal structure details are presented in the SI section materials and methods. Details for the synthesis of compounds I and I^′ can also be found in the SI Text. Compounds II and II^′ were synthesized according to published procedures (26).