APP

LIED

PHYS

ICA

LSC

IEN

CES

Site-selective measurement of coupled spin pairsin an organic semiconductorS. L. Baylissa,b,c,1, L. R. Weissb,1, A. Mitioglud, K. Galkowskie, Z. Yange, K. Yunusovaa, A. Surrentee, K. J. Thorleyf,J. Behrendsc, R. Bittlc, J. E. Anthonyf, A. Raob, R. H. Friendb, P. Plochockae, P. C. M. Christianend, N. C. Greenhamb,2,and A. D. Chepelianskiia,2

aLaboratoire de Physique des Solides, Universite Paris-Sud, CNRS, UMR 8502, F-91405 Orsay, France; bCavendish Laboratory, University of Cambridge,Cambridge CB3 0HE, United Kingdom; cBerlin Joint Electron Paramagnetic Resonance Laboratory, Fachbereich Physik, Freie Universitat Berlin, D-14195Berlin, Germany; dHigh Field Magnet Laboratory–European Magnetic Field Laboratory (HFML-EMFL), Radboud University, 6525 ED Nijmegen, TheNetherlands; eLaboratoire National des Champs Magnetiques Intenses, Universite Joseph Fourier, Universite Paul Sabatier, Institut National des SciencesAppliquees (CNRS-UJF-UPS-INSA), 31400 Toulouse, France; and fDepartment of Chemistry, University of Kentucky, Lexington, KY 40506-0055

Edited by Zachary Fisk, University of California, Irvine, CA, and approved March 30, 2018 (received for review October 29, 2017)

From organic electronics to biological systems, understandingthe role of intermolecular interactions between spin pairs is akey challenge. Here we show how such pairs can be selectivelyaddressed with combined spin and optical sensitivity. We demon-strate this for bound pairs of spin-triplet excitations formed bysinglet fission, with direct applicability across a wide range of syn-thetic and biological systems. We show that the site sensitivity ofexchange coupling allows distinct triplet pairs to be resonantlyaddressed at different magnetic fields, tuning them betweenoptically bright singlet (S = 0) and dark triplet quintet (S = 1, 2)configurations: This induces narrow holes in a broad optical emis-sion spectrum, uncovering exchange-specific luminescence. Usingfields up to 60 T, we identify three distinct triplet-pair sites, withexchange couplings varying over an order of magnitude (0.3–5 meV), each with its own luminescence spectrum, coexisting ina single material. Our results reveal how site selectivity can beachieved for organic spin pairs in a broad range of systems.

organic semiconductors | exchange coupling | triplet excitons |singlet fission | organic spintronics

Spin pairs control the behavior of systems ranging from quan-tum circuits to photosynthetic reaction centers (1, 2). In

molecular materials, such pairs mediate a diverse range of pro-cesses such as light emission, charge separation, and energyharvesting (3–5). The relevant spin pair may consist of two spin-1/2 particles, in the form of either a bound exciton or a weaklycoupled electron–hole pair, or spin-1 pairs, which have recentlyemerged as alternatives for efficient light emission and harvest-ing (6–9). A key challenge in understanding and using such pairsis accessing the local molecular environments which supporttheir generation and evolution within more complex structures,information which could ultimately lead to active control of theirproperties.

Here we demonstrate that the joint dependence of spin andelectronic interactions on pair conformation provides a handleto separate such states and extract their discrete environmentsfrom a broader energetic landscape. We apply this technique tomeasure distinct triplet pairs formed by singlet fission (Fig. 1A),a process which generates two spin S = 1 excitons from a photo-generated S = 0 singlet exciton and is of great current interest forsolar energy conversion (10–12). We simultaneously extract theexchange energies and optical spectra of three different triplet-pair sites within the same material. Using a magnetic field, wetune different triplet pairs into excited-state avoided crossings,which we detect as spectral holes in an inhomogeneously broad-ened photoluminescence (PL) spectrum. This enables combinedspin and optical characterization of these states: The fieldsrequired to induce level crossings directly measure the set ofpair exchange-coupling strengths, while the spectral holes pro-vide narrow, spin-specific optical profiles of the states. We extractmultiple triplet-pair states with exchange couplings varying by

an order of magnitude and decouple their distinct luminescencespectra from an otherwise inhomogeneously broadened back-ground, reaching subnanometer spectral linewidths. Our resultsreveal more means of determining structure–function relationsof coupled spins and identify unambiguous pair signatures. Thisapproach is directly applicable to a range of organic systems:from electron–hole pairs in next-generation light-emitting diodesto coupled excitons in artificial and naturally occurring lightharvesters.

Results and DiscussionMethod for Selectively Addressing Exchange-Coupled Triplets.Despite their key role in light emitters and harvesters, tripletpairs have only recently been discovered to form exchange-coupled states (13–17)—we start by outlining how such statescan be selectively addressed to provide a site-specific mea-surement of their exchange interactions and associated opticalspectra. Here we describe the specific case of singlet fission, but

Significance

Pairs of spins in molecular materials have attracted signif-icant interest as intermediates in photovoltaic devices andlight-emitting diodes. However, isolating the local spin andelectronic environments of such intermediates has provedchallenging due to the complex structures in which theyreside. Here we show how exchange coupling can be usedto select and characterize multiple coexisting pairs, enablingjoint measurement of their exchange interactions and opti-cal profiles. We apply this to spin-1 pairs formed by photonabsorption whose coupling gives rise to total-spin S = 0, 1 and2-pair configurations with drastically different properties. Thispresents a way of identifying the molecular conformationsinvolved in spin-pair processes and generating design rules formore effective use of interacting spins.

Author contributions: S.L.B., L.R.W., and A.D.C. designed research; S.L.B., L.R.W., A.M.,K.G., Z.Y., K.Y., A.S., P.P., P.C.M.C., and A.D.C. performed research; K.J.T. and J.E.A. con-tributed new reagents/analytic tools; S.L.B., L.R.W., and A.D.C. analyzed data; S.L.B.,L.R.W., J.B., R.B., A.R., R.H.F., and N.C.G. wrote the paper; and J.B., R.B., A.R., R.H.F., andN.C.G. supervised research.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.

Data deposition: The numerical data underlying each figure in this paper have beendeposited with the University of Cambridge Apollo depository, www.repository.cam.ac.uk, and may be accessed at https://doi.org/10.17863/CAM.22102.1 S.L.B. and L.R.W. contributed equally to this work.2 To whom correspondence may be addressed. Email: alexei.chepelianskii@u-psud.fr orncg11@cam.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1718868115/-/DCSupplemental.

Published online May 2, 2018.

www.pnas.org/cgi/doi/10.1073/pnas.1718868115 PNAS | May 15, 2018 | vol. 115 | no. 20 | 5077–5082

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

0

A B

C

D

Fig. 1. Selective addressing of exchange-coupled triplet–exciton pairs. (A) Schematic of spin-pair generation by singlet fission for an ensemble of pairsites with different exchange interactions. Photon absorption generates a spin-singlet exciton (S1), which can radiatively decay, producing PL, or undergofission into a pair of triplet excitons (TT). Fusion of this triplet pair reforms the singlet exciton, while dissociation destroys it. (B and C) Triplet-pair levelanticrossings for a single exchange energy. A magnetic field tunes optically dark triplet or quintet spin sublevels into near degeneracy with the bright singletstate, resulting in selective reductions in the PL at fields proportional to the exchange interaction J. ∆PL/PL = [PL(B)− PL(0)]/PL(0). (D) The magnetic-field–induced anticrossings create spectral holes linked to a specific triplet pair. This enables the narrow associated emission profiles of triplet pairs with differentexchange interactions to be extracted.

emphasize that the approach can be directly translated to manyother molecular systems since spin-conserving transitions are ageneral feature of such materials.

Fig. 1A outlines the process of triplet-pair generation by sin-glet fission, where both fission and the subsequent fusion processare spin conserving. This makes the spectral regions associatedwith triplet pairs sensitive to their spin states, which can be res-onantly tuned with an external magnetic field (Fig. 1B). Forstrongly exchange-coupled triplets, the eigenstates at zero mag-netic field consist of the pure singlet (S = 0), triplet (S = 1),and quintet (S = 2) pairings of the two particles. Due to its sin-glet precursor, fission selectively populates the S = 0 triplet-pairconfiguration, which is energetically separated from the opticallyinactive triplet or quintet states due to the exchange interaction.Application of a magnetic field enables these triplet or quintetstates to be tuned into resonance with the optically active singlet-pair state when the Zeeman energy matches the singlet–tripletor singlet–quintet exchange splitting. At these field positions,bright singlet-pair states become hybridized with a dark triplet-or quintet-pair state, manifesting as a resonant reduction in therelevant PL spectral window (Fig. 1C) (16–18).

Crucially, the crossings directly address pairs with a specificexchange coupling. For an exchange interaction JS1 ·S2, whereS1,2 are the spin operators for the two triplets, the resonancesoccur at |J | (singlet-triplet crossing) and 3|J |/2, 3|J | (singlet–quintet crossings), giving a direct measurement of the exchange.(Here we take J > 0; SI Appendix, VI. Spin Hamiltonian andSign of J.) Furthermore, only the emission linked to the reso-nant triplet pair will be diminished at each level crossing. Themagnetic-field resonances will therefore selectively burn spectralholes linked to pairs with a given exchange coupling (Fig. 1D).From these resonant spectral changes, both the spin and opticalproperties of pair sites are therefore reconstructed. Importantly,since triplet pairs with different exchange interactions will haveseparated resonant fields, their associated spectra can be indi-vidually measured. Specific spin pairs with distinct spectral andspin properties can therefore be disentangled in an ensemblemeasurement and their local environment and microscopic prop-erties probed. This is the key principle of our approach to providea spin- and site-selective measurement of organic spin pairs.

TIPS-Tetracene. Of the expanding class of singlet fission materialsfor photovoltaic application, solution-processable systems witha triplet energy close to the bandgap of silicon are particularlyimportant since they could be integrated directly with establishedhigh-efficiency silicon technologies. One such material is TIPS-

tetracene [5,12-Bis((triisopropylsilyl)ethynyl)-tetracene] (Fig. 2A and B), a solution-processable derivative of the archetypalfission material tetracene (19, 20), which has been shown toundergo effective fission and generate exchange-coupled tripletpairs (13, 21, 22). Furthermore, singlet- and triplet-pair statesare nearly iso-energetic in TIPS-tetracene, and so PL can beused to interrogate the fission products (21–23). Here we useTIPS-tetracene to study the spin and electronic structure of cou-pled triplet excitons. To achieve both high spectral and high

A

C

B

Fig. 2. Level anticrossings of spin-1 pairs. (A) Chemical schematic and PLspectrum of TIPS-tetracene at 1.4 K (sample 1). (B) TIPS-tetracene unit celldisplaying four inequivalent molecules. (C) Magneto-PL at 1.4 K integratedacross all wavelengths showing a series of resonances (sample 1, pulsedfields).

5078 | www.pnas.org/cgi/doi/10.1073/pnas.1718868115 Bayliss et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

0

APP

LIED

PHYS

ICA

LSC

IEN

CES

field resolution, we perform measurements using both pulsed(<60 T) and continuous wave (cw) (<33 T) magnetic fields onthree identically prepared samples (Materials and Methods): sam-ple 1 under pulsed field at 1.4 K and samples 2 and 3 under cwfields at 2 K and 1.1 K, respectively. Samples are crystallites ofapproximately millimeter linear dimensions, containing multipledomains, prepared by evaporation from saturated solution, andwere not specifically oriented with respect to the magnetic field.We first identify triplet-pair level crossings in TIPS-tetraceneand then use these to spectrally characterize multiple distincttriplet pairs.

Triplet-Pair Level Crossings. Fig. 2C shows the changes in inte-grated PL up to 60 T for a TIPS-tetracene crystallite at1.4 K (sample 1, pulsed fields; Materials and Methods), where∆PL/PL = [PL(B) − PL(0)]/PL(0). Below 1 T, the conven-tional singlet fission magnetic-field effect is observed, indicativeof weakly coupled triplet pairs (19), while at & 1 T a very dif-ferent behavior arises. On top of the monotonic PL reductionwith field, which we discuss later, multiple PL resonances areapparent: a series below 15 T and additional resonances above30 T, indicating triplet-pair level anticrossings. As shown in Fig.1C, for a given triplet pair there are three possible resonanceswith the fission-generated singlet state, occurring with field ratios1:3/2:3. The number of resonances in Fig. 2C therefore indi-cates multiple triplet pairs with different exchange interactions.While the resonances at <15 T can be separated into two pro-gressions with 1:3/2:3 field ratios (red/blue labels, Fig. 2C) thisdoes not clearly assign them or associate them with particularoptical properties: We now show how the identified triplet-pair level crossings can be used to unambiguously decouple andspectrally characterize multiple interacting triplets in the samematerial.

Spectrally Resolving Interacting Triplets. Fig. 3 B–D showsmagneto-PL traces at three different wavelengths λa,b,c whichcorrespond to high-energy regions of the TIPS-tetracene PLspectrum (Fig. 3A). In contrast to the integrated measurements,the magneto-PL at λa and λb shows a clear progression ofthree resonances following the 1:3/2:3 field ratios expected forlevel anticrossings with the singlet state (Fig. 3C, Inset), givingexchange interactions of 0.44 meV and 0.34 meV, respectively;i.e., J/gµB = 3.79 T, 2.96 T, where g ' 2 is the exciton g factor

(13, 23) and µB the Bohr magneton. (Note that due to their spec-tral proximity, the 56-T λa resonance is also present in the λb

trace.) In contrast to the λa,b spectral positions, at λc, resonancesare present only at much higher fields of 33.4 T and 42.0 T. Sincethese do not occur at the expected 1:3/2 field ratios, we assignthem to the lowest–field—i.e., singlet–triplet—anticrossings ofdistinct triplet pairs with exchange interactions of 3.87 meV and4.87 meV, respectively (J/gµB = 33.4 T, 42.0 T). This is furthersupported by their distinct temperature dependences which wedescribe later.

As outlined in Fig. 1D, since the PL resonances for tripletpairs with different exchange interactions are readily separablein field, we can determine their emission characteristics fromthe spectral components that are diminished at each resonantfield position, i.e., the difference in PL (∆PLres) when off res-onance vs. on resonance: ∆PLres = |PL(Bres.)−PL(Boff res.)|.[We note that for an accurate off-resonance subtraction in thepresence of more slowly changing nonresonant field effects, wetake PL(Boff res.) as the average of the spectra either side ofthe resonance.] The spectra associated with each set of reso-nances (i.e., triplet pairs) are shown in Fig. 3 E–H and we labelthe associated triplet pairs TTa,b,c. The resulting PL spectrashow similar vibronic progressions, yet shifted peak emissionenergies with peaks centered at λa,b,c (Fig. 3E). The fact thatthe three spectra exhibit near-identical vibrational progressionsbut with an overall shift relative to each other shows that thestates differ predominantly in their electronic rather than vibra-tional coupling. The relative shift indicates a difference in thelocal environment between the triplet pairs which results indistinct electronic interactions with the surrounding molecules.The question arises as to why a single material supports mul-tiple triplet-pair sites with distinguishable electronic and spinenergy levels. A natural explanation is the different molecularconfigurations accessible in TIPS-tetracene in which there arefour rotationally inequivalent molecules in the crystal unit cell(Fig. 2B) (24). Multiple triplet pairs may therefore be supportedand due to their differing interaction strengths and electronicenvironments be associated with different exchange couplingsand optical emission spectra. (We note that as an alternativeapproach to species extraction, we find that independent spec-tral decomposition algorithms show good agreement with thespectra/lineshapes in Fig. 3; SI Appendix, IV. Numerical SpectralDecomposition.)

A B D

F

C

E G H

Fig. 3. Magneto-optical spectroscopy of triplet pairs formed by singlet fission. (A) A 1.4 K PL spectrum with features at λa = 562.5 nm, λb = 564.9 nm,λc = 567.1 nm highlighted (sample 1). (B–D) Magneto-PL traces for the three spectral positions λa−c (sample 2). (B and C) Magnetic-field resonances at λa

and λb corresponding to triplet pairs with exchange interactions of 0.44 meV and 0.34 meV (J/gµB = 3.79 T, 2.96 T). (C, Inset) Resonant fields appear withratios 1:3/2:3, as expected for the possible level crossings with the singlet state. Error bars are taken as 10% of the resonant linewidths. Dashed lines areguides to the eye. (D, Inset) Spectrally resolved PL measurements (marked field points) and integrated PL for reference (solid line) at λc. (E–H) Extractedspectra for the triplet pairs associated with the resonances (sample 1): overlaid (E) and shown individually (F–H).

Bayliss et al. PNAS | May 15, 2018 | vol. 115 | no. 20 | 5079

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

0

Vibrational Structure in TT Spectra. In keeping with previousassignments, the first spectral peaks at &560 nm are attributedto 0–0, i.e., zero-phonon, transitions (22). This is also con-sistent with the greater overlap of low-energy modes on thehigher-order vibrational transitions described in detail below(Fig. 4B). Sample 3 (measured at the lowest temperature) exhib-ited pronounced TTa signatures (Fig. 4A), with linewidths ofthe extracted spectra reaching as low as 0.5 nm (15 cm−1),significantly narrower than the ∼10-nm linewidth of the 0–0 peak in the steady-state PL spectrum. This allows us toidentify the vibronic transitions shown in Fig. 3 with greateraccuracy (Fig. 4B). (Note that sample 2 spectra—Fig. 3 F–H—were measured using a spectrometer with lower spectralresolution, limiting the minimum linewidths.) We use this spec-trum to extract four ground-state vibrational modes involvedin the emission process. Fig. 4B shows a stick spectrum ofthe progression of one lower-energy mode with wavenumberν1 = 310 cm−1 and three higher-energy modes (ν2,ν3, ν4 =1,160 cm−1, 1,270 cm−1, and 1,370 cm−1), showing good agree-ment with the measured spectra. These frequencies are inagreement with modes found in the ground-state Raman ofTIPS-Tetracene films (22) with ν1 similar to typical C-C-Cout-of-plane bending modes and ν2−4 similar to typical C-Cstretching/C-C-H bending modes (25).

To our knowledge these are unique measurements of nar-row optical spectra which can be associated with triplet pairs.The subnanometer optical linewidths obtained here are com-parable to those obtained in fluorescence line-narrowing exper-iments of tetracene (26), highlighting the sensitivity of thisapproach. In contrast to all-optical measurements, the spin sen-sitivity afforded here allows clear assignment to triplet pairs.In addition, we note that spectral extraction of triplet-pair sig-natures does not require clearly visible peaks in the bare PLspectrum. For example, the longer-wavelength peaks associatedwith TTa−c are unclear in the bare PL, and spectral decom-position of TT signatures is possible even in a sample withbarely visible λa,b peaks (SI Appendix, IV. Numerical SpectralDecomposition).

B

A

Fig. 4. Vibrational structure in subnanometer resonant PL. (A) Zero-fieldPL spectrum, TTa spectrum extracted from the 5.6-T resonance, and PLresonances for sample 3. (B) Resonant PL spectrum of TTa with idealizedvibrational progression (red lines) consisting of the 0–0 transition (ν00); adominant low-energy mode with wavenumber ν1 = 310 cm−1; and threehigher-energy modes ν2,ν3, ν4 = 1,160 cm1, 1, 270 cm1, and 1, 370 cm−1.

Fig. 5. Temperature dependence of triplet-pair signatures. (A and B)Temperature-dependent integrated PL traces and spectra (sample 1). (C)Low-temperature behavior of the λa and λb spectral features which cor-respond to the triplet pairs with J/gµB = 3.79 T and 2.96 T, respectively(sample 1).

Temperature-Dependent TT Signatures. The identified triplet-pair species are further distinguishable through their tempera-ture dependences. Fig. 5A shows the temperature dependence ofthe resonances in the integrated PL and the corresponding evo-lution of the emission spectrum. By 10 K, the resonances below15 T are lost, concurrent with the loss of the λa and λb spectralfeatures (Fig. 5B). The fact that the resonances at '33 T and 42T have distinct temperature dependences supports their assign-ment to the first crossing of different triplet pairs [rather than asingle species with a more structured exchange interaction (27)].By 30 K, no PL resonances are observed, with no magnetic-fieldeffect beyond ∼1 T. Measurement of PL spectra between 4.2 Kand 1.4 K (Fig. 5C) shows that the λa,b spectral features evolvesignificantly over this temperature range, indicating that escapefrom the associated emission sites has an activation temperatureon the order of a few kelvins (∼0.1 meV). Interestingly, this isapproximately the exchange coupling for TTa,b. However, wenote that this energy scale may alternatively be (i) a reorganiza-tion energy due to molecular reconfiguration or (ii) an electronicbarrier between different excited states.

High-Field Spin Mixing. While resonant spectral analysis pro-vides a window into the electronic structure associated withtriplet pairs, the magnetic lineshapes provide insight into spin-mixing mechanisms and the emissive species. The magneto-PLshows a monotonic decrease with field, up to nearly 50% at 60 T(Fig. 2C), a drastically higher field than the <0.5-T scale usuallyseen in organic systems. This unanticipated high-field effect canbe explained due to g-factor anisotropy which can nonresonantlymix the singlet |S〉 and m = 0 triplet state |T0〉, when triplets areorientationally inequivalent, analogous to ∆g effects observedin spin-1/2 pairs due to differences in isotropic g values (4,28, 29). The competition between spin-mixing ∆g Hamiltonianterms and total-spin–conserving exchange terms sets a charac-teristic saturation field for the effect ∝J/∆geff , where ∆geff

is the relevant effective g-factor difference (SI Appendix, VII.Spin Mixings). Triplet pairs with a larger exchange interaction

5080 | www.pnas.org/cgi/doi/10.1073/pnas.1718868115 Bayliss et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

0

APP

LIED

PHYS

ICA

LSC

IEN

CES

Fig. 6. Triplet-pair spin mixing. (A) Spectrally resolved high-field effect(sample 1) showing ∆PL/PL at spectral positions λa−c. (B) Simulation ofthe role of g anisotropy. Inclusion of an anisotropic g factor enhancesthe singlet-triplet level crossing (at field J/gµB = 3.8 T) and produces amonotonic reduction in PL with field.

should therefore have a larger characteristic field scale for thiseffect and hence also be distinguishable through their non-resonant spin mixing. Fig. 6C shows ∆PL/PL for the threedifferent spectral regions λa−c up to 68 T. The λa,b traces,which correspond to triplet pairs with similar exchange inter-actions (0.44 meV and 0.34 meV), show a similar nonresonantlineshape which saturates around 30 T, while the λc trace, asso-ciated with an order of magnitude larger exchange interaction,shows a much higher characteristic field scale for PL reduc-tion, consistent with this mechanism. We note that high-fieldeffects have rarely been observed in organic materials in general,and our observations show the relevance of spin-orbit coupling(responsible for g anisotropy), which is usually assumed to benegligible.

Singlet–Triplet Level Crossings. A difference in g matrices alsoprovides a mixing mechanism for the singlet–triplet crossings.Since the pure S = 1 triplet-pair states are antisymmetric withrespect to particle exchange, while the S = 0, 2 states are sym-metric (19), different mixing mechanisms are required forsinglet–triplet vs. singlet–quintet hybridization. Singlet–quintetmixing can be mediated by the intratriplet zero-field split-ting interaction (18) which characterizes the dipolar interactionbetween electron and hole and has strength D/gµB = 50 mT inTIPS-tetracene (13, 23, 24, 30). However, to first order this cou-pling leaves the singlet–triplet crossing forbidden (SI Appendix,VII. Spin Mixings). Clear singlet–triplet crossings seen for TTa,b

therefore indicate an additional mixing mechanism. As with thehigh-field effect, this can be provided by a ∆g Hamiltonian termwhich mixes singlet and triplets to first order (Fig. 6B and SIAppendix, VII. Spin Mixings) with strength ∼∆g ′effB ∼ 10−3Bfor an expected ∆g ′eff ∼ 10−3 (31). Additionally, this cross-ing can be mediated by hyperfine interactions, with typicalstrengths of approximately milliteslas in organic semiconductors(32, 33).

The Role of Kinetics in Magnetic-Field Effect. Interestingly, themagnetic linewidths of the PL resonances (Fig. 4A) are largerthan expected based purely on the mixing matrix elements for thecrossings, which would give linewidths of .50 mT. We obtainedsimilar linewidths in a single crystal sample (SI Appendix, V.Magneto-PL of TIPS-Tetracene Single Crystal), and thereforea distribution in J can be ruled out as the dominant line-broadening mechanism. Instead, as detailed in SI Appendix, VIII.Kinetic Modelling, this indicates the broadening is predominantlydue to the kinetics of the fission/fusion process.

For both resonant and nonresonant PL reductions, mixingis predominantly between the singlet and one other (triplet orquintet) pair state, and this sets a maximum ∆PL/PL of '50 %(neglecting annihilation to a single triplet). This maximum is

based on the distribution of the S = 0 character across one stateat zero field vs. two states at resonant positions/high field (18).The fact that the PL can be reduced by nearly 50% by a magneticfield (Figs. 2C and 4A) therefore indicates that strongly cou-pled triplet pairs can dominate the steady-state emission prop-erties of singlet-fission systems. For identifying singlet fission,the observation that exchange-coupled triplets can dominatesteady-state magnetic field effects is highly significant. Often,a low-field effect (.100 mT) characteristic of weakly coupledtriplets (19) is taken to be a signature of the fission process (6).In contrast, our results show that singlet-fission magnetic-fieldeffects can be drastically different between strongly and weaklycoupled triplets and that high-field effects (&1 T) can insteaddominate.

We note that for fission-generated triplet pairs the emissivespecies may be either a distinct singlet exciton or, as proposedrecently (22, 34), the triplet pairs themselves. While typicallychallenging to distinguish these scenarios, the combination ofkinetically broadened linewidths and near 50% resonant PLreductions naturally arises only when triplet pairs emit via a sep-arate singlet state, rather than directly themselves, showing theadditional utility of these measurements in distinguishing thesekinetic scenarios (SI Appendix, VIII. Kinetic Modelling).

Outlook. The magneto-optic resolution of organic triplet pairsopens up the possibility to correlate their exchange and elec-tronic structure with their chemical environment and physicalconformation. Since the mixing matrix elements relevant for thePL resonances depend on the relative orientation between theexternal field and the triplet pair (18), measuring orientation-ally dependent PL resonances should allow triplet pairs to beassigned to specific molecular configurations. Identification ofunambiguous spectral signatures of triplet pairs further meansthat these states can now be studied through purely opticalmeans. For example, triplet-pair microscopy could be used toobtain information on the spatial distribution of pair sites acrossmicrocrystalline domains and map their diffusion (35–37), andresonant excitation could be used to address specific triplet pairsthrough site-specific fluorescence (38, 39).

While here we spectrally resolve triplet pairs in a singlet-fissionmaterial, these results are applicable to a range of other organicspin-pair systems. For example, triplet–triplet encounters arepivotal in photovoltaic upconversion systems (40) and organiclight-emitting diodes (9, 41), and triplet-pair level anticrossingsshould also be observable in photovoltaic device architectures,where resonances could be measured through solar-cell pho-tocurrent or quantum-dot emission (11). In spin-1/2 pairs, anal-ogous spectrally resolved level crossings should help to clarifythe spin and electronic structure of the emissive species centralto thermally activated delayed fluorescence in next-generationorganic light-emitting diode materials, and extracting optical sig-natures from level crossings observed in synthetic and biologicalradical pairs should provide further insights into these key inter-mediates (4, 33, 42, 43). Finally, the nanoscale sensitivity ofexchange-coupled spins opens up the possibility to deliberatelyengineer them as joint spin-optical probes of complex molecularsystems.

Materials and MethodsSamples were excited by 532-nm, 514-nm, or 485-nm laser illumination(similar results were obtained across this wavelength range). A long-passfilter was used to remove the laser line, and the collected PL was senteither to an avalanche photodiode for the integrated measurements orthrough a monochromator to a nitrogen-cooled CCD for the spectrallyresolved measurements. Three different TIPS-tetracene crystallites preparedby evaporation from saturated solution were used, which we refer to assamples 1–3. X-ray diffraction confirmed that all samples indexed to thesame unit cell previously determined for TIPS-tetracene (CCDC database,dx.doi.org/10.5517/cc119qsv), demonstrating that they had the same

Bayliss et al. PNAS | May 15, 2018 | vol. 115 | no. 20 | 5081

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

0

underlying solid-state structure. Integrated and spectrally resolved experi-ments to 68 T were performed using sample 1 under pulsed magnetic field atLaboratoire national des champs magnetiques intenses Toulouse. Spectrallyresolved measurements up to 33 T were performed using samples 2 and 3under steady-state fields at the High Field Magnet Laboratory, Nijmegen.For low-temperature measurements samples were either immersed in liquidhelium (samples 1 and 3) or cooled via exchange gas with a surroundinghelium bath (sample 2), giving base temperatures of '1.4 K, 2 K, and 1.1 Kfor samples 1–3, respectively. PL spectra in Fig. 5C were taken with sample 1in helium under continuous pumping. Further details and comparison of the

samples are contained in SI Appendix, I. Samples, II. PL and ∆PLres Spectrain Samples 1–3, and III. Extraction of ∆PLres. The data underlying this paperare available at https://doi.org/10.17863/CAM.22102.

ACKNOWLEDGMENTS. This work was supported by HFML-RU/FOM andLNCMI-CNRS, members of the European Magnetic Field Laboratory (EMFL),and by EPSRC (United Kingdom) via its membership in the EMFL (GrantsEP/N01085X/1 and NS/A000060/1) and through Grant EP/M005143/1. L.R.W.acknowledges support from the Gates-Cambridge and Winton Scholar-ships. We acknowledge support from Labex ANR-10-LABX-0039-PALM, ANRSPINEX, and DFG SPP-1601 (Bi-464/10-2).

1. Petta JR, et al. (2005) Coherent manipulation of coupled electron spins insemiconductor quantum dots. Science 309:2180–2184.

2. Lubitz W, Lendzian F, Bittl R (2002) Radicals, radical pairs and triplet states inphotosynthesis. Acc Chem Res 35:313–320.

3. Clarke TM, Durrant JR (2010) Charge photogeneration in organic solar cells. ChemRev 110:6736–6767.

4. Steiner U, Ulrich T (1989) Magnetic field effects in chemical kinetics and relatedphenomena. Chem Rev 89:51–147.

5. McCamey D, et al. (2008) Spin rabi flopping in the photocurrent of a polymer light-emitting diode. Nat Mat 7:723–728.

6. Congreve DN, et al. (2013) External quantum efficiency above 100% in asinglet-exciton-fission-based organic photovoltaic cell. Science 340:334–337.

7. Mezyk J, Tubino R, Monguzzi A, Mech A, Meinardi F (2009) Effect of an externalmagnetic field on the up-conversion photoluminescence of organic films: The role ofdisorder in triplet-triplet annihilation. Phys Rev Lett 102:087404.

8. Liu R, Zhang Y, Lei Y, Chen P, Xiong Z (2009) Magnetic field dependent triplet-tripletannihilation in Alq3-based organic light emitting diodes at different temperatures. JAppl Phys 105:093719.

9. van Eersel H, Bobbert P, Coehoorn R (2015) Kinetic Monte Carlo study of triplet-tripletannihilation in organic phosphorescent emitters. J Appl Phys 117:115502.

10. Smith MB, Michl J (2010) Singlet fission. Chem Rev 110:6891–6936.11. Thompson NJ, et al. (2014) Energy harvesting of non-emissive triplet excitons in

tetracene by emissive PbS nanocrystals. Nat Mater 13:1039–1043.12. Tabachnyk M, et al. (2014) Resonant energy transfer of triplet excitons from

pentacene to PbSe nanocrystals. Nat Mater 13:1033–1038.13. Weiss LR, et al. (2017) Strongly exchange-coupled triplet pairs in an organic

semiconductor. Nat Phys 13:176–181.14. Tayebjee MJ, et al. (2017) Quintet multiexciton dynamics in singlet fission. Nat Phys

13:182–188.15. Basel BS, et al. (2017) Unified model for singlet fission within a non-conjugated

covalent pentacene dimer. Nat Commun 8:15171.16. Wakasa M, et al. (2015) What can be learned from magnetic field effects on singlet

fission: Role of exchange interaction in excited triplet pairs. J Phys Chem C 119:25840–25844.

17. Yago T, Ishikawa K, Katoh R, Wakasa M (2016) Magnetic field effects on triplet pairgenerated by singlet fission in an organic crystal: Application of radical pair model totriplet pair. J Phys Chem C 120:27858–27870.

18. Bayliss SL, et al. (2016) Spin signatures of exchange-coupled triplet pairs formed bysinglet fission. Phys Rev B 94:045204.

19. Merrifield RE (1971) Magnetic effects on triplet exciton interactions. Pure Appl Chem27:481–498.

20. Burdett JJ, Piland GB, Bardeen CJ (2013) Magnetic field effects and the role of spinstates in singlet fission. Chem Phys Lett 585:1–10.

21. Stern HL, et al. (2015) Identification of a triplet pair intermediate in singlet excitonfission in solution. Proc Natl Acad Sci USA 112:7656–7661.

22. Stern HL, et al. (2017) Ultrafast triplet pair formation and subsequent thermally acti-vated dissociation control efficient endothermic singlet exciton fission. Nat Chem9:1205–1212.

23. Bayliss SL, et al. (2014) Geminate and nongeminate recombination of triplet excitonsformed by singlet fission. Phys Rev Lett 112:238701.

24. Bayliss SL, et al. (2015) Localization length scales of triplet excitons in singlet fissionmaterials. Phys Rev B 92:115432.

25. Alajtal A, Edwards H, Elbagerma M, Scowen I (2010) The effect of laser wavelength onthe Raman spectra of phenanthrene, chrysene, and tetracene: Implications for extra-terrestrial detection of polyaromatic hydrocarbons. Spectrochim Acta A Mol BiomolSpectrosc 76:1–5.

26. Hofstraat J, Schenkeveld A, Engelsma M, Gooijer C, Velthorst N (1989) Temper-ature effects on fluorescence line narrowing spectra of tetracene in amorphoussolid solutions. Analytical implications. Spectrochim Acta A Mol Biomol Spectrosc A45:139–146.

27. Kollmar C, Sixl H, Benk H, Denner V, Mahler G (1982) Theory of two coupled tripletstates-electrostatic energy splittings. Chem Phys Lett 87:266–270.

28. Devir-Wolfman AH, et al. (2014) Short-lived charge-transfer excitons in organicphotovoltaic cells studied by high-field magneto-photocurrent. Nat Commun 5:4529.

29. Wang Y, Sahin-Tiras K, Harmon NJ, Wohlgenannt M, Flatte ME (2016) Immense mag-netic response of exciplex light emission due to correlated spin-charge dynamics. PhysRev X 6:011011.

30. Weil JA, Bolton JR (2007) Electron Paramagnetic Resonance (Wiley, New York), pp162–187.

31. Schott S, et al. (2017) Tuning the effective spin-orbit coupling in molecularsemiconductors. Nat Commun 8:15200.

32. McCamey D, et al. (2010) Hyperfine-field-mediated spin beating in electrostaticallybound charge carrier pairs. Phys Rev Lett 104:017601.

33. Zarea M, Carmieli R, Ratner MA, Wasielewski MR (2014) Spin dynamics of radicalpairs with restricted geometries and strong exchange coupling: The role of hyperfinecoupling. J Phys Chem A 118:4249–4255.

34. Yong CK, et al. (2017) The entangled triplet pair state in acene and heteroacenematerials. Nat Commun 8:15953.

35. Irkhin P, Biaggio I (2011) Direct imaging of anisotropic exciton diffusion and tripletdiffusion length in rubrene single crystals. Phys Rev Lett 107:017402.

36. Akselrod GM, et al. (2014) Visualization of exciton transport in ordered anddisordered molecular solids. Nat Commun 5:3646.

37. Wan Y, et al. (2015) Cooperative singlet and triplet exciton transport in tetracenecrystals visualized by ultrafast microscopy. Nat Chem 7:785–792.

38. Bassler H, Schweitzer B (1999) Site-selective fluorescence spectroscopy of conjugatedpolymers and oligomers. Acc Chem Res 32:173–182.

39. Orrit M, Bernard J, Personov R (1993) High-resolution spectroscopy of organicmolecules in solids: From fluorescence line narrowing and hole burning to singlemolecule spectroscopy. J Phys Chem 97:10256–10268.

40. Singh-Rachford TN, Castellano FN (2010) Photon upconversion based on sensitizedtriplet-triplet annihilation. Coord Chem Rev 254:2560–2573.

41. Baldo MA, Adachi C, Forrest SR (2000) Transient analysis of organic electrophos-phorescence. II. Transient analysis of triplet-triplet annihilation. Phys Rev B62:10967–10977.

42. Weiss EA, Ratner MA, Wasielewski MR (2003) Direct measurement of singlet- tripletsplitting within rodlike photogenerated radical ion pairs using magnetic field effects:Estimation of the electronic coupling for charge recombination. J Phys Chem A107:3639–3647.

43. Hiscock HG, et al. (2016) The quantum needle of the avian magnetic compass. ProcNatl Acad Sci USA 113:4634–4639.

5082 | www.pnas.org/cgi/doi/10.1073/pnas.1718868115 Bayliss et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 5

, 202

0