Tuning Thermally Activated DelayedFluorescence Emitter Photophysics throughSolvation in the Solid StateBenjamin L. Cotts,† Dannielle G. McCarthy,† Rodrigo Noriega,†,∇ Samuel B. Penwell,†,○ Milan Delor,†

David D. Devore,§ Sukrit Mukhopadhyay,§ Timothy S. De Vries,§ and Naomi S. Ginsberg*,†,‡,∥,⊥,#

†Department of Chemistry and ‡Department of Physics, University of California Berkeley, Berkeley, California 94720, United States§The Dow Chemical Company, Midland, Michigan 48674, United States∥Materials Sciences Division and ⊥Molecular Biophysics and Integrative Bioimaging Division, Lawrence Berkeley NationalLaboratory, Berkeley, California 94720, United States#Kavli Energy NanoSciences Institute, Berkeley, California 94720, United States

*S Supporting Information

ABSTRACT: Solid-state solvation (SSS) is analogous to liquid-phase solvation but occurs withinglassy matrices. Organic solutes with singlet charge transfer (1CT) excited states are especiallysusceptible to solvatochromism. Their 1CT states and photon emission energies decrease whensurrounding molecules with sterically unhindered polar moieties reorient to stabilize them.Thermally activated delayed fluorescence (TADF) organic light-emitting diodes feature suchsolutes as emitters in the solid state, employing efficient reverse intersystem crossing to harvestthe majority of electrogenerated triplets. Here we explore the potential of SSS to manipulate notonly these emitters’ 1CT states but also, concurrently, their singlet−triplet energy gaps (ΔEST)that control TADF. By solvating the TADF emitter 2PXZ-OXD with progressively increasingconcentrations of camphoric anhydride (CA) in a polystyrene film, we find that it is possible tofinely tune the emitter’s photophysics. We observe a maximum increase in prompt lifetime andcorresponding decrease in delayed lifetime of ∼60%. By contrast, the photoluminescencequantum yield peaks at an intermediate CA concentration, reflecting competition betweenincreasing reverse intersystem crossing yield and decreasing singlet oscillator strength. Ourfindings demonstrate technologically relevant fine control of emitter photophysical properties, as varying the extent of SSSreveals the convolved evolution of different kinetic rates as a function of the 1CT state energy and ΔEST.

The solvatochromism effect in the solution phase is well-known.1−3 An organic solute with a polar excitedsinglet state displays a red shift of its emission when

placed in a polar solvent. After excitation to the polar excitedstate, the solvent molecules reorganize, moving in space tostabilize the composite system. This response lowers theemitter state’s energy and results in the observed red shift of theemission (Figure 1c inset). The magnitude of this red shift isdirectly proportional to the solvent polarizability.1

It is less widely known that under certain conditionssolvatochromism can also take place in the solid state, albeitwith smaller magnitude. This phenomenon can occur in solidglassy matrices and is referred to as solid-state solvation (SSS).SSS was first formally observed and categorized by Bulovic andco-workers,4,5 who studied the shift in emission of a laser dye(DCM2) as a function of matrix polarizability. This effect isdirectly analogous to solvatochromism. After excitation of anorganic solute to a polar excited state, other polar smallmolecules within a glassy matrix rotate in space to locally orienttheir dipole moments and stabilize the solute excited state.

Increasing the polarizability of the matrix can increase themagnitude of the stabilization and hence the magnitude of thered shift of the emission. This effect will be especiallypronounced for solutes with charge-transfer (CT) excitedstates, which intrinsically have a dipole moment in the excitedstate that is much larger than that in the ground state.1,2 SSS,however, has not been utilized frequently,5−8 nor to its fullpotential. Here we use SSS to both systematically study andcontrol the photophysical properties of emitters with CTexcited states.Thermally activated delayed fluorescence (TADF) organic

light-emitting diode (OLED) emitters are a promising newclass of OLED emitters9−17 that have particularly interestingphotophysics to study using SSS. TADF OLED emitters featurea CT singlet excited state that minimizes the exchange

Received: March 27, 2017Accepted: June 5, 2017

Letterhttp://pubs.acs.org/journal/aelccp

© XXXX American Chemical Society 1526 DOI: 10.1021/acsenergylett.7b00268ACS Energy Lett. 2017, 2, 1526−1533

interaction to produce an electronic structure with a smallsinglet−triplet energy gap (ΔEST).

9−11,13−19 The lowest-energy1CT state is generally achieved through a molecular design thatincorporates donor−acceptor complexes that, as a byproduct oftheir structure, can also feature a locally excited singlet (1LE)localized on the donor unit that can couple with the 1CTstate.11,18,20,21 The small ΔEST enables efficient harvesting ofelectrogenerated triplets through thermally activated reverseintersystem crossing (RISC), allowing full TADF OLEDdevices to reach internal quantum efficiencies of up to100%10,11,18,20 (Scheme 1).By combining SSS-tuning with TADF OLED emitters, we

modulate the small ΔEST in order to untangle the complexphotophysics of these emitters. The ΔEST energy gap of TADF

OLED emitters (∼0−150 meV) and the practical tuning rangeof SSS (∼100 meV) are on the same order of magnitude,making for a potent pairing. Nevertheless, care must be takento closely follow the coupled changes to the singlet radiativerate and RISC rate. Because the two rates are both affected bystabilizing the CT singlet excited state, it is important to ensurethat the effect of increasing RISC from the triplet state is notnegated by decreased oscillator strength of the singlet state.The fine-tuning of the 1CT state energy, and hence of ΔEST,that we achieve with SSS elucidates the intricate details ofTADF photophysics and reveals the balance of effects that mustbe struck to optimize and control the performance of TADFemitters.Here we demonstrate that the pairing of SSS and TADF

yields an experimental platform to, for the first time, showdirectly how host codoping can change the emitter excited stateto drastically alter its photophysics. This manipulation isachieved by simultaneously altering the emitter’s RISC andradiative rates. To employ SSS to study the effect of modulatingthe amount of singlet excited-state stabilization on TADFphotophysics we use a ternary system and vary the ratio of apolar small molecule codopant and a nonpolar host. Whenincrementing the codopant concentration to increase hostmatrix polarizability, we observe monotonic changes in thephotoluminescence (PL) spectral peak, the prompt lifetime,and the delayed lifetime. Additionally, over this range, weobserve a rise and a subsequent fall in PL quantum yield (QY).This approach reveals the ideal magnitude of SSS-stabilizationthat optimizes the yield. It results from competition betweenthe improved efficiency of returning the triplet population tothe emissive singlet state and the concurrent decrease inoscillator strength of this singlet state with increasing

Figure 1. (a) Cartoon of ternary blend samples; (b) molecular structures of host polymer polystyrene, polar small molecule codopantcamphoric anhydride, and emitter molecule 2PXZ-OXD; (c) PL spectra of 0, 10, 20, and 40 wt % CA in PS with 6 wt % 2PXZ-OXD.Increasing the concentration of CA leads to growing red shifts of the emission energy as the magnitude of SSS increases. Inset in the upperright is a diagram showing the solvation mechanism of the system. Upon photoexciation to the 1CT excited state of 2PXZ-OXD, the CAmolecules rearrange their dipole moments to stabilize the polarized state and lower its energy. This causes the observed red shift.

Scheme 1. TADF Energy Level Diagram with Kinetic Systemof Ratesa

aLabeled rates are the radiative rate of the singlet state to ground state(kr), nonradiative rate of the singlet to ground state (knr

S ), intersystemcrossing rate from singlet to triplet (kisc), reverse intersystem crossingrate from triplet to singlet (krisc), and nonradiative rate of the triplet toground state (knr

T ). The energy gap between the first singlet state andtriplet state is also labeled (ΔEST).

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00268ACS Energy Lett. 2017, 2, 1526−1533

1527

polarization of the singlet excited state. This first systematicstudy of TADF using 1CT excited-state solvation to controlΔEST reveals a pathway to optimize emissive yields, derivedfrom better understanding the interplay of the various rateconstants within a TADF kinetic system.To elucidate the progressive effect of SSS on TADF emitter

photophysics, we measured the PL emission peak shift, theprompt and delayed PL lifetime changes, and the PL QYvariation in a series of ternary blend samples. Modeled after thework done by Madigan and Bulovic 5 studying laser dyes, eachfilm consisted of a fixed concentration (6 wt %) of the TADFemitter 2,5-bis(4-(10H-phenoxazin-10-yl)phenyl)-1,3,4-oxadia-zole (2PXZ-OXD) doped within a polystyrene (PS) hostmatrix, to which varying amounts of the small molecule polarcodopant, camphoric anhydride (CA), were added (Figure 1).The magnitude of the SSS effect was then varied within a seriesby making each film with a different ratio of CA to PS. 2PXZ-OXD was selected as a good model TADF emitter. Lee et al.achieved 14.9% external quantum efficiency in a device using 6wt % 2PXZ-OXD doped into a bis[2-(diphenylphosphino)-phenyl] ether oxide (DPEPO) host,12 an efficiency highenough to require TADF collection of triplet excitations. Assuch, 2PXZ-OXD is known to have a 1CT state, which isconfirmed by observing strong solvatochromism in the liquidphase (Figure S1).20,22 Although not optimized for deviceperformance, we used PS as an optically transparent andsolution-processable model host with a high triplet energy.23

Finally, CA was selected to be the SSS-active codopant becauseit is small, relatively sterically unhindered, and has a strongground-state dipole.5 Neither evidence of exciplex formationnor a dramatic change in spectral shape, apart from peak shift,was observed in optical absorption (Figure S2) or emissionspectra (Figure 1c). We measured these model samples with PLtechniques to isolate the influence of SSS on emitterperformance from any device optimization considerationssuch as charge injection. First, steady-state spectra weremeasured (Figures 1 and 2a). Next, prompt and delayedlifetimes were measured separately. The time-correlated single-photon counting (TCSPC) method was used to measure theprompt lifetime (Figure 2b, Table S1), and multichannelscaling (MCS) was used to measure the delayed lifetime(Figure 2c, Table S2). Last, PL QY was measured in anintegrating sphere following a published method24 (Figure 2d).Variation in the amount of CA in the PS:CA:2PXZ-OXD

system leads to a monotonic red shift of the PL emission peak(Figure 2a). Increasing the wt % CA from 0 to 20% red-shifts2PXZ-OXD peak emission from 494.25 to 525 nm, a change of147 meV (Figure S3). The trend of PL peak emission versus[CA] is relatively linear at lower concentrations but begins tolevel off at higher concentrations. This potential saturation mayoccur as the ability to more densely pack CA molecules aroundeach emitter diminishes at high [CA].The prompt and delayed lifetime also display monotonic

dependencies on [CA]. The full time-resolved PL traces for theprompt lifetime (Figure S4) are fit to a biexponential α1e

(−t/τ1)

+ α2e(−t/τ2), and the average lifetime τ =

α τ α τα τ α τ

++

1 12

2 22

1 1 2 2is reported

in Figure 2b. Figure 2b1 shows that increasing [CA] increasesthe average prompt lifetime from ∼9 to 14.5 ns over the rangeof 0−20 wt % CA. This effect follows approximately the sametrend as the PL peak shift in Figure 2a, beginning roughlylinearly with slight saturation at higher [CA] (Figure 2b).Conversely, increasing [CA] correspondingly decreases average

delayed lifetime from ∼617 to ∼256 μs over 0−20 wt % CA.Full traces for delayed lifetime (Figure S5) are fit to atriexponential decay, and the average lifetime is reported inFigure 2c. Nonmonoexponential decays of both lifetimes likelyarise from heterogeneity in the local environment of the 2PXZ-OXD emitter molecules in these films caused by the variablenumber of CA molecules around each emitter.25 Finally,increasing [CA] initially increases the PL QY at lowconcentrations over the range of ∼2.5−10 wt % but thenproceeds to decrease the PL QY at higher CA concentrations(Figure 2d). This variation is a complex convolution of thetrends in Figure 2a−c and will be discussed later in the text.Overall, the data in Figure 2 illustrate marked effects of

progressive increases in [CA] on the photophysical propertiesof the TADF OLED emitter in PS. We observe monotonicchanges in PL peak wavelength, prompt lifetime, and delayedlifetime, with all of these trends slightly saturating at higher

Figure 2. Comparison of the SSS-enabled change to the photo-physical observables of 2PXZ-OXD as a function of [CA]. (a) Thechange in PL peak, (b) average prompt lifetime, (c) averagedelayed lifetime, and (d) PL QY as a function of [CA] are shown.For reference, the corresponding number of CA per emitter is alsoindicated along the top horizontal axis. Monotonic trends areobserved for PL peak, average prompt lifetime, and delayedlifetime. A more complex relationship exists between PL QY and[CA]. The data suggest that an optimal ΔEST is reached around 5wt % CA, where the TADF rate is increased without diminishingthe oscillator strength of the S1 → S0 transition too much.

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00268ACS Energy Lett. 2017, 2, 1526−1533

1528

[CA]. This saturation is probably due to the limited volumeavailable to pack more CA around the emitter molecules at highcodoping levels. The PL QY, however, does not monotonicallychange. There is an optimal range of [CA] that maximizes thePL QY and therefore an optimal range for ΔEST around 5−7 wt% CA.Having described the multiple observations of the PL of

2PXZ-OXD as a function of codopant concentration, we nowdiscuss the possible kinetic origin of the observed trend in PLQY as the amount of SSS stabilization is increased. Tuning thekinetics of TADF emitters through SSS to maximize theemissive yield is not straightforward. The rates underlying theprompt and delayed fluorescence are highly interdependent,changing simultaneously with increased SSS-stabilization of theexcited singlet state (Scheme 1). There are five relevant rates inthe kinetic scheme that describes TADF: the radiative rate ofthe singlet (kr

S), the nonradiative rate of the singlet (knrS ), the

intersystem crossing (ISC) rate to the triplet state (kisc), theRISC rate from the triplet to the singlet (krisc), and thenonradiative rate of the triplet (knr

T ). Both prompt and delayedfluorescence are phenomenological combinations of these morefundamental rates. Prompt fluorescence lifetime is dependenton the radiative rate of the singlet (kr

S), nonradiative rate of thesinglet (knr

S ), and ISC to the triplet state (kisc). It can beapproximated as17,26,27

= + +k k k kprompt rS

nrS

isc

while the delayed fluorescence rate is mainly determined by theRISC rate, because it is a relatively slow thermally activatedprocess. It is, however, also codependent on the emissiveproperties of the singlet state. The delayed fluorescence ratecan be expressed as17,26,27

= + −+ +

⎛⎝⎜

⎞⎠⎟k k

kk k k

k1TADF nrT isc

rS

nrS

iscrisc

which is directly proportional to krisc. Therefore, as krisc isthermally activated, both krisc and kTADF will increase as thebarrier to activation (ΔEST) decreases. Using the kinetic

scheme described above and in Scheme 1, we now discuss howthe three monotonic trends observed in Figure 2a−c give riseto the nonmonotonic behavior of the PL QY (Figure 2d) as afunction of the increasing CA-enabled SSS.The SSS stabilization of the 1CT excited state causes the

redshift in the PL spectra as a function of matrix polarizabilityas shown in Figure 1. The mechanism of the SSS stabilization isanalogous to solvatochromism.5 The electron density of the2PXZ-OXD molecule instantaneously redistributes uponphotoexcitation to its 1CT excited state. The CA moleculesthen respond to the increased local polarizability by rotating inspace to align their permanent dipole moments to stabilize thepolarized charge distribution on the 2PXZ-OXD molecule(Figure 1 inset). This lowers the energy of the 1CT state andcauses the observed red shift in emission. Further evidence thatsupports SSS as the underlying mechanism for the observed redshift is provided by the observation of red-edge effects.1,28 Weprepared a series of films at codoping concentrations of 0, 2.5,5, and 10 wt % CA with a lower 2PXZ-OXD doping level (0.25wt %) to minimize emitter self-interaction and to clearlydemonstrate CA’s active role in SSS. This set of films was thenexcited at both 400 nm (center of absorption band) and at 371nm (blue edge of 1CT absorption band), and the resulting PLspectra were recorded. As can be seen in Figure S6, the PLspectra for the two excitation wavelengths are identical for allfilms in which CA is present, whereas in the 0.25 wt % 2PXZ-OXD film that has no CA codopant a blue shoulder appears forthe 371 nm excitation. This shoulder could arise from eitheremission from the 1LE state or emission from an uppervibrational energy state before relaxation down to the bandedge of the 1CT state. In either case, the fact that the 2PXZ-OXD molecule does not relax to the band edge before emissionwithout the presence of CA molecules confirms their active rolein solvation, as well as their potential to modify the electronicstructure and performance of TADF emitters.In addition to the clear spectral redshifts observed, SSS also

has a significant impact on the time-resolved PL of 2PXZ-OXD.Coincident with the lowering of the energy of the 1CT excitedstate caused by SSS we observed an increase in the prompt

Scheme 2. TADF Energy Level Diagrams and Change in PL QY as a Function of [CA] along with Projected Associated Changesin Singlet Yield (ΦS), Prompt Yield (Φprompt), and EL QYa

aTADF energy level diagram and relevant rates on the left-hand side depict a situation where there is no SSS stabilization of the 1CT singlet state,whereas for the energy level diagram on the right-hand side, the 1CT state energy has been lowered from the addition of a SSS stabilizationinteraction, modifying some of the rate constants’ magnitudes. In particular, as the strength of the SSS interaction increases and lowers the 1CT stateenergy, we expect krisc to increase and kr

S to decrease, affecting the overall radiative efficiency (PL QY) of the emitter. In the center a sketch of thechange in PL QY as a function of [CA] is pictured along with projected associated changes in singlet yield (ΦS), prompt yield (Φprompt), and EL QY.The sketched ΦS and Φprompt curves illustrate that the krisc change dominates at low [CA] to increase PL QY before saturating at higher [CA],causing further increases in [CA] to lower the PL QY. Finally, the projected EL QY curve reflects that any increases to PL efficiency from thesefactors will only be accentuated for electroluminescence efficiency.

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00268ACS Energy Lett. 2017, 2, 1526−1533

1529

lifetime (Figure 2a,b). We hypothesize that SSS stabilizationpotentially also further increases the polarized CT character ofthe state. This further polarization of the state should alsodecrease the oscillator strength of the radiative transition forthe 1CT state, thereby lowering its radiative rate (kr

S), resultingin the observed increase in prompt fluorescence lifetime(Scheme 2). In addition, any mixing of the 1LE and 1CT stateswould decrease with the lowering of the 1CT energy similarlyincreasing the prompt lifetime by decreasing the oscillatorstrength of the radiative transition.29−33 As such, increased[CA] and increased SSS stabilization of the 1CT state lead tolonger prompt PL lifetimes.Furthermore, we hypothesize that ΔEST decreases with

increasing local host polarizability, as indicated by the decreasein delayed lifetime and increase in prompt lifetime as a functionof increasing CA concentration, presumably because the tripletstate is less stabilized by CA-enabled SSS than the 1CT state.Density functional theory (DFT) calculations of naturaltransition orbitals (NTOs) (Figure S7) show that the T1state has much more localized π−π* character than the S1state. [Note that DFT calculations refer to generic first excitedsinglet (S1) and triplet (T1). The S1 state corresponds mostly to1CT, though it could also include mixing with a small amountof a 1LE, and the T1 state corresponds to a mixture of 3LE and3CT.] Further calculations confirm that the excited-statepolarity of the S1 state (20.6 D) is considerably higher thanthat of the T1 state (9.3 D). This would cause the more polar S1state to be stabilized more by CA than the T1 state, therebycausing a decrease in ΔEST with increased [CA]. Geometricconstraints imposed by the host matrix could change the singletand triplet energies or polarities so that they might differ fromthe results of our DFT calculations, but we do not anticipatethat they would alter our overall conclusions because theexperimental data is consistent with a decrease in ΔEST.Additionally, the literature suggests that the 3LE state couldplay the dominant role in RISC, and a 3LE state would be evenless affected by the presence of CA than a pure 3CTstate.11,18,34,35 Therefore, as ΔEST decreases with increasing[CA], the delayed lifetime decreases because the equilibriumpopulation distribution between triplet and singlet excitedstates shifts, increasing the net singlet excited-state population(Scheme 2). Viewed from a kinetic perspective, the delayedlifetime decreases as the rate of ISC/RISC cycling increases. Amore rapid thermally activated RISC leads to more exposure ofthe excited-state population to the emissive singlet states, whichdecreases the delayed lifetime.Stabilization of the excited singlet state leads to competing

contributions to the efficiency of the fluorescent emission,namely, reduced emission efficiency of the singlet state and theincrease in efficiency of transferring population from the excitedtriplet to the excited singlet state. Although the PL emissionpeak shift, the prompt lifetime, and the delayed lifetime allmonotonically vary as a function of [CA], the PL QY does not.This difference suggests a complex (nonlinear) competitionbetween different effects on the TADF kinetics thatconcurrently occur upon modulation of the magnitude ofSSS. To explain how the PL QY can increase at low [CA] andthen decrease at higher [CA] we must next consider thedifferent contributions to PL QY and their dependencies on[CA].In order to understand the underpinnings of PL QY in

TADF emitters it is useful to define the equations that state theefficiencies of the various pathways available to the excited-state

population in a TADF emitter, both nonradiative andradiative.17,26,27 This specificity allows us to parse out howthe observed variations in the delayed fluorescence and promptfluorescence manifest in the variation in PL QY as a function of[CA]. For a TADF emitter the major contribution to PL QYcomes from the prompt fluorescence due to the population thatradiatively decays from the singlet state without ever crossinginto the triplet state. The efficiency of this pathway is

Φ =+ +

kk k kprompt

rS

rS

nrS

isc

The PL QY that we measure, however, does not solely dependon the prompt PL QY. The portion of the population thatundergoes ISC can be recovered via TADF. The total radiativeefficiency, ΦPLQY, can be quantified through a consideration ofthe ISC/RISC cycling efficiency.26 The amount of excitedsinglet population that crosses over to the triplet state can bedescribed by

Φ =+ +

kk k kT

isc

rS

nrS

isc

Similarly, the efficiency of the reverse process of tripletpopulation returning to the excited singlet state is

Φ =+k

k kSrisc

nrT

risc

The portion of population that crosses over to the triplet stateand then back to the singlet is then, again, subject to the sameset of rates as the original singlet state population and can cycleback and forth repeatedly. Nevertheless, we can express theoverall PL QY in a closed form:26

Φ = Φ + Φ Φ + Φ Φ +

=Φ− Φ Φ

[1 ( ) ...]

1

PLQY prompt T S T S2

prompt

T S

In this closed form expression, ΦS, ΦT, and Φprompt all vary as afunction of [CA]. We expect Φprompt to decrease with increased[CA] as kr

S decreases, as indicated by the prompt lifetimeincrease. Conversely, the delayed lifetime data indicate that theΦS should increase with increasing [CA] as krisc increases withthe drop in ΔEST. Finally, we note that in the limit of krisc ≫ knr

T ,the singlet yield ΦS will approach unity, and any furtherincrease in [CA] will not significantly improve the harvesting oftriplet states. This nonlinear saturation in the relationshipbetween ΦS and [CA] helps to explain the nonmonotonicvariation in PL QY versus [CA]. At low [CA], ΦS is far fromsaturating and its increase with increasing [CA] dominates thetrend in PL QY versus [CA]. However, at high [CA], the ΦShas saturated and there is no longer significant additional tripletharvesting from increased SSS. The variation in PL QY as afunction of [CA] will then be dominated by the decrease in theoscillator strength of the singlet reflected in Φprompt. Thiscompetition leads to the local maximum at ∼5−10 wt % CA inFigure 2d.This argument is also sketched visually in Scheme 2, where

we explore the effects on the TADF kinetic scheme at low andhigh degrees of SSS. The sketch of the monotonicallydecreasing prompt yield (blue) and increasing, but saturating,singlet yield (red) illustrates the competing effects that controlthe relationship between overall PL QY and [CA] (green). For

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00268ACS Energy Lett. 2017, 2, 1526−1533

1530

lower [CA] (left scheme), the ΔEST and krS are relatively large,

while RISC efficiency is relatively low. By contrast, at high[CA] (right scheme), where the ΔEST is reduced and ISC/RISC cycling favors increased singlet population, the oscillatorstrength and kr

S will decrease because of the increased polarityof the solvated emitter 1CT state, thereby decreasing PL QY aswell. This effect is present at all levels of [CA] but dominatesonly at high [CA] when singlet yield has saturated near unity.At intermediate [CA], around 5−10% [CA], the singlet yieldhas increased without significantly decreasing the prompt yield;thus, the highest possible PL QY is achieved.We have assumed above that the variation in PL QY can be

explained by focusing on the dominant changes to krisc and krS

after stabilization of the 1CT state by SSS without invoking anychange to the other rates. As justification of this assumption, wedo not expect there to be dramatic changes to the nonradiativerates as a function of [CA]. Any slight variation of the matrixmorphology with the PS:CA ratio would likely affect knr

S and knrT

similarly. This would lead to similar changes to the prompt anddelayed lifetime rather than the opposing changes that weobserve. Furthermore, we do not expect that there is significantquenching at high CA from aggregation because promptlifetime would then be expected to decrease at high CA, and itdoes not.From all of the above measurements and analysis, we have

found that SSS is able to substantially change the photophysicsof TADF emitters with very fine control. For example, weobserve both an increase in prompt lifetime and a decrease indelayed lifetime of ∼60% using a codoping level between 0 and20 wt % CA, as well as a peak increase of 20% for the PL QYover the range of 0−6.25 wt % CA. Additionally, for eachweight percent of CA added between 0 and 20% CA, the PLemission peak shifts by ∼1.55 nm, which should allow for colorcoordinate tuning with nanometer wavelength precision over a30 nm range in devices. The peak emission is potentially easierto tune using small reproducible increments of [CA] ratherthan by resynthesizing the molecular host or emitter materialdifferently.34,36 From the standpoint of solvation at themolecular level, note that the range of codopant concentrationsthat we explored corresponds to average separation distances of∼1−2 nm in the polystyrene matrix or to codopant-to-emitterratios ranging between 0 and 11 (Figure 2), with the PL QYpeaking at ∼2.7. It is interesting that marked variations inphotophysical properties are observed as the number of CAmolecules per emitter molecule is incremented by a single CA.This observation illustrates the fine control achieved by themere addition of a very small, countable number of solid-statesolvent molecules.We have demonstrated an ability to fine-tune the

concentration of a polar codopant capable of SSS in order totease out the intricacies of TADF kinetics and efficiencies.Moreover, this new capability and the accompanying under-standing enables an opportunity to finely control TADF inlight-emitting devices such as OLEDs. We first considerdesirable properties of the ternary system that would be mostsuited to fine-tuning photophysics with SSS. We project thatSSS will be most potent for optimization of the performance ofTADF OLED emitters in such blends when a relativelynonpolar but conductive host material and a small, stericallyunhindered, conductive, polar small molecule codopant areused. We hypothesize that introducing this codoping platforminto a conductive host would produce qualitatively similarresults beyond some reduction of the effective range of

solvation due to screening of the codopant molecule. Emittermolecules most suited to SSS-tuning might have singlet−tripletenergy gaps that are small but nonzero, as an ideal startingpoint for finding the optimal degree of SSS. Even in deviceswithout codopants, where SSS is not deliberately employed, itmight still be at play because of host choice or residual solvent,so it should not be overlooked in order to optimally controlperformance.Second, we relate our PL findings back to electroluminescent

devices. Changes observed in PL QY as a function of SSS willbe translated into accentuated changes in the EL QY. For agiven system, EL QY is always smaller than PL QY. Althoughboth are governed by the same overall kinetics, in the ELpicture the initial population of the nonemissive triplet state is 3times greater than that of singlet state and leads to additionallosses due to increased exposure to the nonradiative pathway(knr

T ) (Figure S8). The increase in PL QY at low [CA] in Figure2d will translate into a steeper increase in the EL QY case, asthere is more triplet state population spared from nonradiativeloss to knr

T when krisc is increased (Scheme 2). Then at higher[CA], where the ΦS saturates with low ΔEST, the EL and PLQY converge to the same value. Therefore, the positive impactof SSS on efficiency in an EL device will be even greater thanthe value observed in our model system with PL. Furthermore,in an EL device, the brightness of the TADF OLED could alsobe increased without a loss of performance due to the use ofSSS. At high current, the amount of triplet−triplet annihilation(TTA) increases as more charges interact with each other. Thisleads to a decrease in efficiency at high current, which isreferred to as rollover. Decreasing delayed lifetime through SSSshould increase the refresh rate of each emitter on average andlead to less TTA for a given current. This adjustment shouldallow for stable operation at higher current and thereforeshould improve brightness.In sum, we have demonstrated the powerful effect that small

changes in SSS can have on the photophysics of TADF OLEDemitters, which have CT excited states. We discovered thatperturbing the photophysics to control luminescence yieldrequires a careful balance between minimization of the singlet−triplet energy gap and maximization of the singlet radiative rate.From a fundamental perspective, employing SSS provides a wayto systematically modify an emitter’s electronic structure inorder to elucidate the kinetic mechanisms behind TADF andcould be used to study many other CT-based emitters in thesolid state. It grants experimental control to tune the electronicstructure of one emitter to study its ability to conduct TADFwithout coarsely changing its overall molecular design. It is alsomore precise and tunable than the fullscale redesign of anemitter molecule. Furthermore, SSS offers a compelling newpathway to tune TADF device design. It could be implementedby including codopant molecules with proper chargeconduction properties or by including polar side chains onpolymer hosts. Nevertheless, an increase in the dielectricconstant of the host without the freedom for permanent dipolemoments to reorient is unlikely to have as potent an effect asSSS. More generally, besides implementing SSS as anoptimization strategy for TADF OLEDs, our work suggeststhat TADF device design should account for the effect of host−guest interactions on emitter photophysics in order tooptimally control device efficiency, color purity, and refreshtime.

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00268ACS Energy Lett. 2017, 2, 1526−1533

1531

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acsenergy-lett.7b00268.

Methods; liquid-phase solvent polarity steady-state PLdata; optical absorbance spectra for PS/CA/2PXZ-OXDfilms; PL steady-state emission in elecrtronvolt units; fullTR-PL traces and fit constants; “red-edge effect” steady-state PL spectra; natural transition orbitals for 2PXZ-OXD S1 and T1 orbitals; EL QY vs PL QY as a functionof ΦS simulation (PDF)

■ AUTHOR INFORMATIONCorresponding Author*E-mail: nsginsberg@berkeley.edu.ORCIDTimothy S. De Vries: 0000-0003-2403-7265Naomi S. Ginsberg: 0000-0002-5660-3586Present Addresses∇R.N.: Department of Chemistry, University of Utah, Salt LakeCity, UT 84112, United States.○S.B.P.: Department of Chemistry, The University of Chicago,Chicago, IL 60637, United States.NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis work has been supported by The Dow ChemicalCompany under Contract No. 244699. B.L.C. acknowledgesa National Science Foundation Graduate Research Fellowship(DGE 1106400). R.N. acknowledges financial support througha Philomathia Foundation Postdoctoral Fellowship. S.B.P.acknowledges a Department of Energy Graduate ResearchFellowship (Contract No. DE-AC05-060R23100). N.S.G.acknowledges an Alfred P. Sloan Research Fellowship, aDavid and Lucile Packard Foundation Fellowship for Scienceand Engineering, and a Camille and Henry Dreyfus Teacher-Scholar Award. Time-resolved fluorescence work at theMolecular Foundry was performed as part of the MolecularFoundry user program, supported by the Office of Science,Office of Basic Energy Sciences, of the U.S. Department ofEnergy under Contract No. DE-AC02-05CH11231. Specialthanks to Teresa Chen at the Molecular Foundry. The MichealD. McGehee group at Stanford University is gratefullyacknowledged for assistance running PL QY experiments.Special thanks to Dan Slotcavage and George Burkhard fromthe McGehee lab.

■ REFERENCES(1) Lakowicz, J. R. Principles of Fluorescence Spectroscopy, 3rd ed.;Springer Science and Business Media: New York, 2006.(2) Rettig, W. Charge Separation in Excited States of DecoupledSystemsTICT Compounds and Implications Regarding theDevelopment of New Laser Dyes and the Primary Process of Visionand Photosynthesis. Angew. Chem., Int. Ed. Engl. 1986, 25 (11), 971−988.(3) Rettig, W. Application of a Simplified Microstructural SolventInteraction Model to the Solvatochromism of Twisted IntramolecularCharge Transfer (TICT) States. J. Mol. Struct. 1982, 84 (3−4), 303−327.

(4) Bulovic, V.; Deshpande, R.; Thompson, M. E.; Forrest, S. R.Tuning the Color Emission of Thin Film Molecular Organic LightEmitting Devices by the Solid State Solvation Effect. Chem. Phys. Lett.1999, 308 (3−4), 317−322.(5) Madigan, C. F.; Bulovic, V. Solid State Solvation in AmorphousOrganic Thin Films. Phys. Rev. Lett. 2003, 91 (24), 247403.(6) Leblebici, S. Y.; Chen, T. L.; Olalde-Velasco, P.; Yang, W.; Ma, B.Reducing Exciton Binding Energy by Increasing Thin FilmPermittivity: An Effective Approach To Enhance Exciton SeparationEfficiency in Organic Solar Cells. ACS Appl. Mater. Interfaces 2013, 5(20), 10105−10110.(7) Green, A. P.; Butler, K. T.; Buckley, A. R. Tuning of the EmissionEnergy of Fluorophores Using Solid State Solvation for EfficientLuminescent Solar Concentrators. Appl. Phys. Lett. 2013, 102 (13),133501.(8) Chang, W.; Akselrod, G. M.; Bulovic, V. Solid-State Solvation andEnhanced Exciton Diffusion in Doped Organic Thin Films underMechanical Pressure. ACS Nano 2015, 9 (4), 4412−4418.(9) Endo, A.; Sato, K.; Yoshimura, K.; Kai, T.; Kawada, A.; Miyazaki,H.; Adachi, C. Efficient up-Conversion of Triplet Excitons into aSinglet State and Its Application for Organic Light Emitting Diodes.Appl. Phys. Lett. 2011, 98 (8), 083302.(10) Uoyama, H.; Goushi, K.; Shizu, K.; Nomura, H.; Adachi, C.Highly Efficient Organic Light-Emitting Diodes from DelayedFluorescence. Nature 2012, 492 (7428), 234−238.(11) Dias, F. B.; Bourdakos, K. N.; Jankus, V.; Moss, K. C.; Kamtekar,K. T.; Bhalla, V.; Santos, J.; Bryce, M. R.; Monkman, A. P. TripletHarvesting with 100% Efficiency by Way of Thermally ActivatedDelayed Fluorescence in Charge Transfer OLED Emitters. Adv. Mater.2013, 25 (27), 3707−3714.(12) Lee, J.; Shizu, K.; Tanaka, H.; Nomura, H.; Yasuda, T.; Adachi,C. Oxadiazole- and Triazole-Based Highly-Efficient ThermallyActivated Delayed Fluorescence Emitters for Organic Light-EmittingDiodes. J. Mater. Chem. C 2013, 1 (30), 4599−4604.(13) Hirata, S.; Sakai, Y.; Masui, K.; Tanaka, H.; Lee, S. Y.; Nomura,H.; Nakamura, N.; Yasumatsu, M.; Nakanotani, H.; Zhang, Q.; et al.Highly Efficient Blue Electroluminescence Based on ThermallyActivated Delayed Fluorescence. Nat. Mater. 2015, 14 (3), 330−336.(14) Adachi, C. Third-Generation Organic ElectroluminescenceMaterials. Jpn. J. Appl. Phys. 2014, 53 (6), 060101.(15) Tao, Y.; Yuan, K.; Chen, T.; Xu, P.; Li, H.; Chen, R.; Zheng, C.;Zhang, L.; Huang, W. Thermally Activated Delayed FluorescenceMaterials Towards the Breakthrough of Organoelectronics. Adv. Mater.2014, 26 (47), 7931−7958.(16) Bergmann, L.; Zink, D. M.; Brase, S.; Baumann, T.; Volz, D.Metal−Organic and Organic TADF-Materials: Status, Challenges andCharacterization. Top. Curr. Chem. 2016, 374 (3), 22.(17) Goushi, K.; Yoshida, K.; Sato, K.; Adachi, C. Organic Light-Emitting Diodes Employing Efficient Reverse Intersystem Crossing forTriplet-to-Singlet State Conversion. Nat. Photonics 2012, 6 (4), 253−258.(18) Zhang, Q.; Li, B.; Huang, S.; Nomura, H.; Tanaka, H.; Adachi,C. Efficient Blue Organic Light-Emitting Diodes Employing ThermallyActivated Delayed Fluorescence. Nat. Photonics 2014, 8 (4), 326−332.(19) Turro, N. J.; Scaiano, J. C.; Ramamurthy, V. Modern MolecularPhotochemistry of Organic Molecules, 1st ed.; University Science Books:Sausalito, CA, 2010.(20) Serevicius, T.; Nakagawa, T.; Kuo, M.-C.; Cheng, S.-H.; Wong,K.-T.; Chang, C.-H.; Kwong, R. C.; Xia, S.; Adachi, C. EnhancedElectroluminescence Based on Thermally Activated Delayed Fluo-rescence from a Carbazole−Triazine Derivative. Phys. Chem. Chem.Phys. 2013, 15 (38), 15850.(21) Dias, F. B.; Santos, J.; Graves, D. R.; Data, P.; Nobuyasu, R. S.;Fox, M. A.; Batsanov, A. S.; Palmeira, T.; Berberan-Santos, M. N.;Bryce, M. R.; et al. The Role of Local Triplet Excited States and D-ARelative Orientation in Thermally Activated Delayed Fluorescence:Photophysics and Devices. Adv. Sci. 2016, 3 (12), 1600080.(22) Ishimatsu, R.; Matsunami, S.; Shizu, K.; Adachi, C.; Nakano, K.;Imato, T. Solvent Effect on Thermally Activated Delayed Fluorescence

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00268ACS Energy Lett. 2017, 2, 1526−1533

1532

by 1,2,3,5-Tetrakis(carbazol-9-yl)-4,6-Dicyanobenzene. J. Phys. Chem.A 2013, 117 (27), 5607−5612.(23) Reineke, S.; Baldo, M. A. Room Temperature Triplet StateSpectroscopy of Organic Semiconductors. Sci. Rep. 2015, 4, 3797.(24) de Mello, J. C.; Wittmann, H. F.; Friend, R. H. An ImprovedExperimental Determination of External Photoluminescence QuantumEfficiency. Adv. Mater. 1997, 9 (3), 230−232.(25) Noriega, R.; Barnard, E. S.; Ursprung, B.; Cotts, B. L.; Penwell,S. B.; Schuck, P. J.; Ginsberg, N. S. Uncovering Single-MoleculePhotophysical Heterogeneity of Bright, Thermally Activated DelayedFluorescence Emitters Dispersed in Glassy Hosts. J. Am. Chem. Soc.2016, 138 (41), 13551−13560.(26) Baleizao, C.; Berberan-Santos, M. N. Thermally ActivatedDelayed Fluorescence as a Cycling Process between Excited Singletand Triplet States: Application to the Fullerenes. J. Chem. Phys. 2007,126 (20), 204510.(27) Baldo, M. A.; Forrest, S. R. Transient Analysis of OrganicElectrophosphorescence: I. Transient Analysis of Triplet EnergyTransfer. Phys. Rev. B: Condens. Matter Mater. Phys. 2000, 62 (16),10958−10966.(28) Demchenko, A. P. The Red-Edge Effects: 30 Years ofExploration. Luminescence 2002, 17 (1), 19−42.(29) Tominaga, K.; Walker, G. C.; Jarzeba, W.; Barbara, P. F.Ultrafast Charge Separation in ADMA: Experiment, Simulation, andTheoretical Issues. J. Phys. Chem. 1991, 95 (25), 10475−10485.(30) Bixon, M.; Jortner, J.; Verhoeven, J. W. Lifetimes for RadiativeCharge Recombination in Donor-Acceptor Molecules. J. Am. Chem.Soc. 1994, 116 (16), 7349−7355.(31) Herbich, J.; Kapturkiewicz, A. Electronic Structure andMolecular Conformation in the Excited Charge Transfer SingletStates of 9-Acridyl and Other Aryl Derivatives of Aromatic Amines. J.Am. Chem. Soc. 1998, 120 (5), 1014−1029.(32) Li, W.; Pan, Y.; Yao, L.; Liu, H.; Zhang, S.; Wang, C.; Shen, F.;Lu, P.; Yang, B.; Ma, Y. A Hybridized Local and Charge-TransferExcited State for Highly Efficient Fluorescent OLEDs: MolecularDesign, Spectral Character, and Full Exciton Utilization. Adv. Opt.Mater. 2014, 2 (9), 892−901.(33) Wang, S.; Cai, J.; Sadygov, R.; Lim, E. C. Intramolecular ChargeTransfer and Solvent-Polarity Dependence of Radiative Decay Rate inPhotoexcited Dinaphthylamines. J. Phys. Chem. 1995, 99 (19), 7416−7420.(34) Santos, P. L.; Ward, J. S.; Data, P.; Batsanov, A. S.; Bryce, M. R.;Dias, F. B.; Monkman, A. P. Engineering the Singlet−triplet EnergySplitting in a TADF Molecule. J. Mater. Chem. C 2016, 4 (17), 3815−3824.(35) Gibson, J.; Monkman, A. P.; Penfold, T. J. The Importance ofVibronic Coupling for Efficient Reverse Intersystem Crossing inThermally Activated Delayed Fluorescence Molecules. ChemPhysChem2016, 17 (19), 2956−2961.(36) Xie, G.; Chen, D.; Li, X.; Cai, X.; Li, Y.; Chen, D.; Liu, K.;Zhang, Q.; Cao, Y.; Su, S.-J. Polarity-Tunable Host Materials andTheir Applications in Thermally Activated Delayed FluorescenceOrganic Light-Emitting Diodes. ACS Appl. Mater. Interfaces 2016, 8(41), 27920−27930.

■ NOTE ADDED IN PROOFAfter submission of the manuscript, a related work, Haseyamaet al., Nanoscale Res. Lett. 2017, 12, 268, was published (April11, 2017) that included a similar codoping concept targetingmanipulation of the singlet−triplet energy gap of a TADFOLED emitter.

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00268ACS Energy Lett. 2017, 2, 1526−1533

1533