Heteroatom oxidation controls singlet–triplet energy splitting in singlet fission building blocks

Abstract

Singlet fission (SF) is a promising multiexciton-generating process. Its demanding energy splitting criterion – that the S₁ energy must be at least twice that of T₁ – has limited the range of materials capable of SF. We propose heteroatom oxidation as a robust strategy to achieve sufficient S₁/T₁ splitting, and demonstrate the potential of this approach for intramolecular SF.

---

Singlet fission (SF) has shown potential to improve the power conversion efficiency in photovoltaic devices beyond the Shockly–Queisser limit by promoting the splitting of a photon-absorbing singlet exciton into two triplet excitons.¹ SF involves the excitation of a ground state (S₀) chromophore to an excited singlet state (S₁) upon absorption of light, followed by energy transfer to a second chromophore. The initial S₁ state is coupled to a triplet pair (¹TT) state, a process which may be mediated by low-lying charge transfer (CT) states or may proceed directly, via a resonance mechanism.^2,3 The triplet pair then evolves into physically separate and energetically independent triplets (T₁), one on each chromophore.

Among the requirements necessary for a system to be capable of SF, the most stringent is that the process be thermodynamically possible, meaning that the energy of the S₁ state must be no less than twice that of the T₁ state: ΔE_ST = E(S₁) − 2E(T₁) ≥ 0. This energy splitting term ΔE_ST is therefore the most relevant target property in the discovery of new SF materials.^1,4,5 It has been shown that (i) extending the conjugation, and increasing the (ii) biradicaloid⁴ or (iii) quinoidal⁶ character of a chromophore can improve ΔE_ST, as summarized in Scheme 1. These strategies have drawbacks. For instance, compounds with high diradicaloid and quinoidal character tend to suffer from chemical instability.⁷

Scheme 1 Proposed strategies to increase singlet–triplet splitting (ΔE_ST) in organic chromophores.

A particular challenge arises in designing materials which fall into the ΔE_ST ≥ 0 regime: the S₁ and T₁ energies tend to move in parallel. When the excitation energies are stabilized to the point that ΔE_ST is fulfilled, T₁ is often too low to be of value for device applications. A historically relevant example of this is the acene family, in which the excited state energies decrease with an increasing number of fused rings. In early reports of SF, in anthracene (3 rings), SF was not favored due to a negative ΔE_ST and, therefore, the endothermicity of SF.⁸ This was also the case for tetracene (4 rings),⁹ while pentacene (5 rings) became the poster child for SF due to it being the first acene in which SF is exergonic (ΔE_ST > 0), although its T₁ energy is already somewhat lower (0.9 eV) than desirable.¹⁰ The next acene, hexacene,¹¹ exhibits much more favorable ΔE_ST for SF, but has a far too low T₁ energy (0.4 eV).

A moiety which stabilizes T₁ incrementally to the point that it can be tuned to remain above 1 eV (for exciton injection into silicon for instance, whose band gap is 1.1 eV) without also lowering S₁ substantially would be greatly beneficial (Scheme 1, right panel). Here, we identify a chemical functionality, heteroatom oxidation, which modulates the ΔE_ST in potential SF chromophores in a foreseeable way (Scheme 1 (iv)). This approach is motivated by the experimental observation that the oxidized form of thiophene (thiophene-S,S-dioxide) is an effective acceptor in donor–acceptor copolymers capable of intramolecular SF (iSF),^12–14 and that nitrone/N-oxide groups were found in large numbers in a recent screening of thousands of crystal structures for compounds with high ΔE_ST.⁵ The present results show that, indeed, heteroatom oxidation governs the S₁ and T₁ energies and thus can be used to improve the singlet–triplet splitting of potential SF chromophores.

To establish if a systematic improvement in ΔE_ST can be achieved through heteroatom oxidation, we constructed a dataset consisting of 11 heteroatom-containing building blocks found widely in the organic electronics literature (see Fig. S1, ESI† for full dataset). These are classified by the number of heteroatoms in the conjugated system, as shown in Scheme 2a. All oxidized derivatives of these compounds were generated by placing one oxygen atom at the electron pair of all sp²-hybridized nitrogen atoms, thereby forming N-oxide (nitrone) moieties, and one or two oxygen atoms at the electron pairs of all sulfur atoms, forming S-oxide or S,S-dioxide moieties, respectively (as shown for bithiophene in Scheme 2b). This produced a total of 67 oxidized compounds (all structures shown in ESI†). The oxidation of sp³ nitrogens was not considered, as this would lead to charged or radical species. Compound geometries were relaxed using density functional theory (ωB97X-D/6-31G*), and the S₁ and T₁ exited-state energies were computed both vertically and at their minima using time-dependent DFT within the Tamm–Dancoff approximation at the same level of theory (see ESI† for details).

Scheme 2 (a) Examples of building blocks studied in this work. (b) Five oxidized derivatives of 2,2′-bithiophene.

Representative results for the effect of oxidation on the vertical and adiabatic S₁, T₁ and ΔE_ST energies of bithiophene are shown in Fig. 1a and results for all other compounds are given in the ESI.† We observe a constant difference between the vertical and adiabatic excitation energies. This allows us to extend our previous observation in dimers¹³ – that the adiabatic energy splitting cutoff (ΔE^adia_ST ≥ 0 eV) can be expressed as ΔE^vert_ST ≥ −1 eV in the Franck–Condon regime – to smaller (monomer) building blocks, due to the linear relationship between the two ΔE_ST values (Fig. S2, ESI†). Although it is immediately clear that increasing the number of oxygens (regardless of their position) stabilizes T₁ in an additive fashion across all compounds, the effect on S₁ is less evident. While mono-oxidation of sulfur leads to a sharp reduction in both S₁ and T₁ energies, which has little positive effect on ΔE_ST, a second oxidation of the same sulfur increases the S₁ energy while further stabilizing T₁, leading to a strong improvement in ΔE_ST. For example, the dioxide derivatives of bithiophene have ΔE^adia_ST ≥ 0 eV (above the grey line in Fig. 1a) while bare bithiophene and its mono-oxidized derivatives do not. The same conclusions can be drawn with all other sulfur-containing units (Fig. S3–S6, ESI†): as outlined in Fig. 1b, single sulfur oxidations have little effect on ΔE_ST as they stabilize S₁ more than T₁, while double oxidations of sulfur are invariably beneficial to ΔE_ST due to their similar stabilization of T₁ but smaller impact on S₁. N-Oxidation systematically improves ΔE_ST through a robust stabilization of T₁ (Fig. S7 and S8, ESI†). These trends are observed regardless of the degree of oxidation of all others heteroatoms, resulting in a remarkably simple cumulative effect across all compounds: the most highly oxidized structures have the highest ΔE_ST of all combinations of oxidation products (Fig. 1 and Section S2 of ESI†).

Fig. 1 (a) Vertical and adiabatic S₁, T₁, and ΔE_ST energies of bithiophene and its oxidized derivatives. Grey line indicates the ΔE_ST cut-off. (b) Summary of the change of adiabatic S₁, T₁, and ΔE_ST upon S-, S,S- and N-oxidation for all compounds, showing averages (white points), 1st–3rd quartiles (black bars), and maximal/minimal values (whiskers). See ESI† for details.

To understand the effect of oxidation on S₁ and T₁ energies, we turned to the nature of the excitations in the simplest sub-unit, thiophene (Table 1). The T₁ state of thiophene is dominated by a local HOMO → LUMO transition in the carbon backbone, while the S₁ state involves predominantly charge transfer (CT) from sulfur (HOMO−1) into the backbone. In thiophene-S-oxide, the T₁ excitation is a localized HOMO−1 → LUMO transition on the backbone, as the HOMO is located on the oxygen. It is instead the S₁ excitation which corresponds to a HOMO → LUMO CT state from oxygen into the backbone π* orbital, which explains the significant stabilization of S₁ upon mono-oxidation. Finally, in thiophene-S,S-dioxide both S₁ and T₁ are characterized by backbone HOMO → LUMO (π → π*) transitions, as the O and S orbitals are much lower in energy.

Table 1 Excitation energies, character and molecular orbitals most involved in excitations for thiophene (Th), thiophene-S-oxide (Th-1O) and thiophene-S,S-dioxide (Th-2O). Atoms involved in orbitals contributing to CT character are highlighted in green. BB = carbon backbone. Orbitals shown in Fig. S11 (ESI)

Compound		State	State character	Orbitals	E (eV)	ΔE^STvert (eV)
	Th	T1	Local BB	H → L	3.90	−1.47
		S1	CT S-to-BB	H−1 → L	6.33
	Th-1O	T1	Local BB	H−1 → L	3.01	−2.24
		S1	CT O-to-BB	H → L	3.78
	Th-2O	T1	Local BB	H → L	2.80	−0.94
		S1	Local BB	H → L	4.66

---

Similarly, in benzodithiophenedione (BDO) and thienopyrroledione (TPD, see Fig. S12, ESI†), S₁ is stabilized by CT states from the oxygen n orbitals into the π* orbital of the backbone. The difference compared to thiophene is that BDO and TPD already contain oxygens in their conjugated systems by virtue of their carbonyls, such that S₁ in non-oxidized TPD and BDO is described by CT from the carbonyls into the heterocycle. A first S-oxidation stabilizes S₁ through CT from the S=O moiety, as with thiophene-S-oxide, while a second oxidation shifts the source of CT back to the carbonyls. Therefore, the nature of the CT (i.e. the n orbitals involved) changes, depending on the structure of the unit, but the stabilizing effect of a single S-oxidation on S₁ remains constant across all compounds. And yet, the oscillator strength of S₁ tends to drop significantly for S-mono-oxidized compounds (see Fig. S14, ESI†), which may have consequences on the SF decay pathway and overall mechanism.¹⁵ Less impact is expected on the photophysical properties of the S₁ state of S,S- and N-oxidized compounds.

Recent work has rationalized ΔE_ST based on ground- and excited-state aromaticity.^16,17 To explain the effect of these substitutions on aromaticity, we computed the nucleus independent chemical shifts (NICS) of thiophene, TPD, and thiazole (Fig. S15, ESI†). While thiophene is aromatic in the ground state and anti-aromatic in the first triplet state, thiophene-S-oxide is much less aromatic in the ground state, but still significantly anti-aromatic in the triplet. The absence of lone pairs on the sulfur atom of thiophene-S,S-dioxide leads to non-aromatic character of both the ground state singlet and first triplet. This is reflected in the bond order of the backbone which, like butadiene, is reversed in the triplet (CH–CH=CH–CH) compared to the singlet (CH=CH–CH=CH). This is not the case for non-oxidized or mono-oxidized thiophene (Fig. S16, ESI†). The TPD ring aromaticity is similarly suppressed upon the double oxidation of sulfur. Put together, these results suggest that the destabilization of S₁ upon S,S-dioxidation, and its consequently beneficial effect on ΔE_ST, originate from the SO₂ moiety eliminating the aromatic character of the heterocycle and instead inducing a polarized butadiene-like behavior to the backbone.¹⁸ In this way, T₁ is sufficiently stabilized, while the stabilizing effect of CT from S (in thiophene rings) or O (in thiophene-S-oxide rings) into the π-system observed in S₁ is eliminated. This is similar to the ‘breaking’ of conjugation in polycyclic hydrocarbons through boron-doping, an approach proposed to build molecules that fulfill ΔE_ST.¹⁹

CT, primarily from oxygen into sulfur (HOMO → LUMO+1), also accounts for the stabilization of S₁ in N-oxidized benzothiadiazole (BT, see Fig. S13, ESI†), compared to non-oxidized BT, which has a local π → π* (HOMO → LUMO) character. The T₁ states are also described by a π → π* transition in both non-oxidized and N,N′-dioxidized BT. The inclusion of the N–O moiety in the conjugated system reduces the π/π* energy gap, leading to an extreme lowering of both the S₁ and T₁ energies by approximately 2 eV. N-Oxidation has the effect of strongly reducing the antiaromatic character of the triplet (Fig. S15, ESI†), while retaining ground state aromaticity, explaining the significant T₁ stabilization and consequent increase in ΔE_ST.

iSF has been demonstrated experimentally in donor–acceptor (D–A) polymers,^12,20,21 in which the triplet pair formation is mediated through low-lying donor-to-acceptor CT states, while the spatial separation of the acceptors by the absorbing donor leads to a weakly bound ¹TT state. We have recently proposed a protocol with which to identify potential polymer candidates for iSF based on the ΔE_ST of the constituent monomers and their relative frontier molecular orbital (FMO) energies.^13,14 To assess the performance of these new oxidized units to form iSF-capable D–A pairs, we treat all those in the dataset that fulfill ΔE^adia_ST ≥ 0 (and ΔE^vert_ST ≥ −1 eV, vide supra) as acceptor monomers (34 compounds). The FMOs of each acceptor were compared to all other buildings blocks (2244 monomer pairs), and only those whose FMO arrangement is conducive to CT (i.e. donor HOMO higher than acceptor HOMO and donor LUMO higher than acceptor LUMO; see earlier work for details¹⁴) were retained. For these 631 D–A combinations, the dimers were generated, their ground state geometries optimized, and their vertical excited states were evaluated at the same level of theory as the monomers. All dimers exhibit energy splitting above the vertical threshold ΔE^vert_ST ≥ −1 eV (Fig. S17, ESI†), which is consistent with our observation¹⁴ that the dimer ΔE_ST originates from the monomer with the higher (i.e. more positive) ΔE_ST.

We have previously outlined two other requirements beyond ΔE_ST for iSF to be possible in D–A systems: S₁ must have significant donor-to-acceptor CT character to drive triplet-pair formation (S₁ → ¹TT), and T₁ must be located on the acceptor to promote dissociation of the triplet states (¹TT → T₁ + T₁).^13,14 Deactivation of S₁ towards higher energy triplet states is neglected but the S₁ → ¹TT transition required for iSF is expected to be the most efficient decay pathway (see Section S6). Quantum chemical descriptors were introduced to quantify these criteria using the character of excited states (see ESI†). Fig. 2 shows the fraction of CT from donor to acceptor in S₁ (Ω^S1_D→A) and the fraction of local T₁ character on the acceptor (Ω^T1_A→A). The upper righthand corner corresponds to the ‘ideal’ region in which these two criteria are fulfilled simultaneously. The majority of the present dimers display a highly localized T₁ (Ω^T1_A→A = 0.5–1.0) and non-negligible S₁ CT character (Ω^S1_D→A = 0.1–0.4), but are nonetheless not in the ideal region.

Fig. 2 Donor-to-acceptor charge-transfer character of S₁ (Ω^S1_D→A, x-axis) and local acceptor character of T₁ (Ω^T1_A→A, y-axis) in dimers, colored by the dihedral between the donor and acceptor (φ_D–A).

A striking exception are dimers containing N,N-dioxidized benzothiadiazole along the top of Fig. 2, indicating pure localization of T₁ on the acceptor (Fig. S17, ESI†). In addition to stabilizing T₁ to achieve positive ΔE_ST, this acceptor induces dihedral torsion to the D–A linkage (φ_D–A ≈ 50°), thereby contributing to very high CT in S₁ (up to 0.8). The best of these dimers is shown in Fig. 2 (compound A), and is revealed to have a donor partner which differs from the acceptor only with regard to the sulfur oxidation. However, the N-oxidation of compound A stabilizes T₁ too much for it to be of practical use if extracted (0.41 eV). All other dimers constructed with this acceptor suffer from this problem (T₁ = 0.3–0.9 eV). Two other dimers (B and C) have a similar acceptor, albeit without N-oxidation, which leads to promising excited state behavior in the dimer and appropriate ΔE_ST (as with A), but importantly, they retain attractive T₁ energies (1.47 eV and 1.24 eV, respectively; see Table S2, ESI†). The absence of nitroxides leads to smaller dimer dihedrals (30° and 19°) and therefore slightly lower CT compared to A, but are still near the ideal region. This analysis demonstrates that through judicious chromophore oxidation, both ΔE_ST and T₁ can be fine-tuned without losing the CT character which mediates the SF process in D–A copolymers.

We have disclosed heteroatom oxidation as a convenient handle through which to modulate singlet–triplet splitting in SF building blocks. Beneficial ΔE_ST through double oxidation of sulfur is obtained by suppressing aromaticity while maintaining overall conjugation, thereby stabilizing T₁ compared to non-oxidized analogs, while having a smaller impact on S₁. A higher number of heteroatom oxidations stabilizes T₁ additively, making it possible to drive the T₁ energy down as far as necessary to achieve exergonic splitting. The utility of this approach is demonstrated using new S- and N-oxidized compounds to construct D–A materials for iSF, although this method is equally valid in intermolecular SF materials design. D–A systems based on a new benzothiadiazole-S,S-dioxide acceptor may be excellent candidates, as sulfur oxidation modulates the excited state energies for SF to be thermodynamically possible while ensuring that the resulting T₁ is appropriate for injection into silicon (1.1–1.7 eV). N-Oxidations, on the other hand, also systematically improve ΔE_ST, but at the expense of an attractive T₁ energy. While these specific units have not been described in the literature, previously reported preparation of S-oxidized²² and S,S-(di)oxidized^22,23 analogs of benzothiadiazole, as well as N-oxidized thiazoles,²⁴ bithiazoles,²⁵ and thiadiazoles²⁶ suggest that they are synthesizable.

The authors are grateful to the EPFL for financial support. M. F. acknowledges funding from European Union's H2020 research and innovation, under MSCA-IF-2018 (G.A. #836849).

Conflicts of interest

There are no conflicts to declare.

References

1. M. B. Smith and J. Michl, Chem. Rev., 2010, 110, 6891–6936.
2. E. G. Fuemmeler, S. N. Sanders, A. B. Pun, E. Kumarasamy, T. Zeng, K. Miyata, M. L. Steigerwald, X.-Y. Zhu, M. Y. Sfeir and L. M. Campos, ACS Cent. Sci., 2016, 2, 316–324.
3. D. Casanova, Chem. Rev., 2018, 118, 7164–7207.
4. I. Paci, J. C. Johnson, X. Chen, G. Rana, D. Popović, D. E. David, A. J. Nozik, M. A. Ratner and J. Michl, J. Am. Chem. Soc., 2006, 128, 16546–16553.
5. D. Padula, Ö. H. Omar, T. Nematiaram and A. Troisi, Energy Environ. Sci., 2019, 12, 2412–2416.
6. B. C. Streifel, J. L. Zafra, G. L. Espejo, C. J. Gómez-García, J. Casado and J. D. Tovar, Angew. Chem., Int. Ed., 2015, 54, 5888–5893.
7. Z. Zeng, X. Shi, C. Chi, J. T. López Navarrete, J. Casado and J. Wu, Chem. Soc. Rev., 2015, 44, 6578–6596.
8. S. Singh and B. Stoicheff, J. Chem. Phys., 1963, 38, 2032–2033.
9. C. E. Swenberg and W. T. Stacy, Chem. Phys. Lett., 1968, 2, 327–328.
10. S. Yoo, B. Domercq and B. Kippelen, Appl. Phys. Lett., 2004, 85, 5427–5429.
11. N. R. Monahan, D. Sun, H. Tamura, K. W. Williams, B. Xu, Y. Zhong, B. Kumar, C. Nuckolls, A. R. Harutyunyan, G. Chen, H.-L. Dai, D. Beljonne, Y. Rao and X. Y. Zhu, Nat. Chem., 2017, 9, 341–346.
12. E. Busby, J. Xia, Q. Wu, J. Z. Low, R. Song, J. R. Miller, X. Zhu, L. M. Campos and M. Y. Sfeir, Nat. Mater., 2015, 14, 426.
13. J. T. Blaskovits, M. Fumanal, S. Vela and C. Corminboeuf, Chem. Mater., 2020, 32, 6515–6524.
14. J. T. Blaskovits, M. Fumanal, S. Vela, R. Fabregat and C. Corminboeuf, Chem. Mater., 2021, 33, 2567–2575.
15. M. Fumanal and C. Corminboeuf, J. Phys. Chem. Lett., 2021, 12, 7270–7277.
16. O. El Bakouri, J. R. Smith and H. Ottosson, J. Am. Chem. Soc., 2020, 142, 5602–5617.
17. K. J. Fallon, P. Budden, E. Salvadori, A. M. Ganose, C. N. Savory, L. Eyre, S. Dowland, Q. Ai, S. Goodlett and C. Risko, J. Am. Chem. Soc., 2019, 141, 13867–13876.
18. M. M. Oliva, J. Casado, J. T. L. Navarrete, S. Patchkovskii, T. Goodson, M. R. Harpham, J. S. Seixas de Melo, E. Amir and S. Rozen, J. Am. Chem. Soc., 2010, 132, 6231–6242.
19. J. Stoycheva, A. Tadjer, M. Garavelli, M. Spassova, A. Nenov and J. Romanova, J. Phys. Chem. Lett., 2020, 11, 1390–1396.
20. G. Grancini, M. Maiuri, D. Fazzi, A. Petrozza, H. Egelhaaf, D. Brida, G. Cerullo and G. Lanzani, Nat. Mater., 2013, 12, 29.
21. J. Hu, K. Xu, L. Shen, Q. Wu, G. He, J.-Y. Wang, J. Pei, J. Xia and M. Y. Sfeir, Nat. Commun., 2018, 9, 2999.
22. T. Linder, E. Badiola, T. Baumgartner and T. C. Sutherland, Org. Lett., 2010, 12, 4520–4523.
23. D. Pinkowicz, Z. Li, P. Pietrzyk and M. Rams, Cryst. Growth Des., 2014, 14, 4878–4881.
24. E. Amir and S. Rozen, Chem. Commun., 2006, 2262–2264.
25. R. A. Mirabal, L. Vanderzwet, S. Abuadas, M. R. Emmett and D. Schipper, Chem. – Eur. J., 2018, 24, 12231–12235.
26. L. S. Konstantinova, E. A. Knyazeva, N. V. Obruchnikova, Y. V. Gatilov, A. V. Zibarev and O. A. Rakitin, Tetrahedron Lett., 2013, 54, 3075–3078.

---

Footnote

† Electronic supplementary information (ESI) available: Computational details, dataset, excited state energies for individual compounds, projected density of states, nucleus-independent chemical shifts. See DOI: 10.1039/d1cc06755a

---