Formation of Organic Alloys in Ternary-BlendSolar Cells with Two Acceptors Having Energy-Level Offsets Exceeding 0.4 eVPetr P. Khlyabich,† Melda Sezen-Edmonds,† Jenna B. Howard,‡ Barry C. Thompson,‡

and Yueh-Lin Loo*,†,§

†Department of Chemical and Biological Engineering, Princeton University, Princeton, New Jersey 08544, United States‡Department of Chemistry and Loker Hydrocarbon Research Institute, University of Southern California, Los Angeles, California90089-1661, United States§Andlinger Center for Energy and the Environment, Princeton University, Princeton, New Jersey 08544, United States

*S Supporting Information

ABSTRACT: Recent studies demonstrated that with proper selection of chemicallycompatible constituents the open-circuit voltage (Voc) of ternary-blend solar cellscan be tuned across the composition window of the active layer. In this study, weprobed the limit of the offset between the lowest unoccupied molecular orbital(LUMO) energy levels of the two acceptors in ternary blends containing one donorand two acceptors. We demonstrate, for the first time, that ternary-blend activelayers with two acceptors having an energy-level difference between their LUMOlevels exceeding 0.4 eV can still result in solar cells exhibiting composition-dependent open-circuit voltage (Voc). Our results suggest strong electronicinteractions between the acceptors, with the electron wave function delocalizedover multiple molecules. These findings have broadened the library of possiblecandidates for active layers of ternary-blend solar cells with tunable Voc andestablished guidelines for the design of next-generation materials for efficientperformance of such devices.

Binary bulk-heterojunction organic solar cells are quicklyapproaching the practical efficiency limit of 11%.1−8

Further increases in efficiency are possible through theintroduction of tandem solar cells in which subcells areconnected either in series or in parallel.9,10 Despite the increasein power conversion efficiencies, such improvements areachieved at the expense of fabrication complexity.11−14 Recently,the use of ternary-blend photoactive layers containing either twodonors and one acceptor (D1/D2/A) or one donor and twoacceptors (D/A1/A2) has become a comparatively simpleapproach to realizing single-junction organic solar cells withgreater efficiencies.15−20 This approach preserves the simplicityof solar cell fabrication while paving the way for accessingefficiencies beyond the limit set by single-junction solar cellscomprising binary blends of donor and acceptor pairs.21−26

Ternary-blend solar cells benefit from an increase in short-circuit current density (Jsc) upon the introduction of a thirdphotoabsorbing component in the active layer. The Jscenhancement stems from an increased number of capturedphotons through the introduction of a constituent having anabsorption profile that complements that of the parent donor−acceptor pair.16,27,28 Until recently, the open-circuit voltage (Voc)of ternary-blend solar cells was thought to be pinned by the

difference in the energy levels of the higher of the two highestoccupied molecular orbitals (HOMOs) of the donors and thelowest unoccupied molecular orbital (LUMO) of the acceptor inactive layers comprising two donors and an acceptor.29,30

However, recent studies demonstrated that with proper selectionof the constituents the Voc can be continuously tuned betweenthe largest and smallest differences of the HOMO energy levelsof the individual donors and the LUMO energy level of theacceptor for D1/D2/A systems or between the largest andsmallest differences of the LUMO energy levels of the individualacceptors and the HOMO energy level of the donor for D/A1/A2systems without negatively impacting the photocurrent or fillfactor (FF).15−17,19,31−35 This tunability in Voc in ternary-blendsolar cells is described by the organic alloy model17,31 in whichindividual constituents preserve their excitonic character due tothe localized nature of the exciton, yet dissociated holes andelectrons are delocalized over much larger distances.32,36

This composition-dependent Voc in ternary-blend solar cells iscorrelated with the chemical compatibility between the active-

Received: July 14, 2017Accepted: August 23, 2017Published: August 23, 2017

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

© 2017 American Chemical Society 2149 DOI: 10.1021/acsenergylett.7b00620ACS Energy Lett. 2017, 2, 2149−2156

layer constituents.36,37 When the two polymer donors in D1/D2/A systems are miscible or can cocrystallize, the correspondingternary-blend solar cells exhibit composition-dependent Voc. Onthe other hand, phase separation between the two donorconstituents limits the Voc of such ternary-blend solar cells to theenergy difference between the LUMO energy level of theacceptor and the higher-lying HOMO energy level of the twodonor polymers.Less explored is how the energy-level offsets between the

HOMOs of the two donors or the LUMOs of the two acceptorsfor D1/D2/A or D/A1/A2 systems, respectively, affect Voc. It hasbeen proposed that this tunability in Voc and simultaneous Jscenhancement in ternary-blend solar cells can only occur whenthese energy-level offsets are below 0.3 eV,38 and experiments todate have not been inconsistent with this assertion.15,16,35,39,40

While a recent study demonstrated Jsc enhancement in D1/D2/Asolar cells when the energy-level offset between the HOMOs ofthe two donors is 0.37 eV,28 the Voc values of these solar cellsappear to be pinned by the difference in the energy levels of the

LUMO of the acceptor and the higher of the two HOMOs of thedonors.28 With a series of D/A1/A2 ternary blends comprisingfullerene derivatives, we systematically evaluated how theLUMO−LUMO energy offset of the fullerenes affects thetunability of the Voc in the resulting solar cells. This elucidationshould shed light on the connection between the constituentHOMO or LUMO energy-level offsets and the ability tomaintain efficient photocurrent generation and charge transportin ternary-blend solar cells. Finally, establishing the energeticrequirements for tunable Voc in ternary-blend solar cells will alsoallow rapid screening of potential constituents for active layersand will simultaneously promote understanding of the opera-tional limits and device physics of ternary-blend solar cells.In our current study, we explored ternary-blend solar cells of

the type D/A1/A2 that exhibit efficient charge generation andtransport. We demonstrate, for the first time, that the Voc of thesesolar cells remains tunable across the composition window of theactive layer despite an energy-level difference between theLUMOs of the two acceptors that exceeds 0.4 eV and that is as

Figure 1. Chemical structures and corresponding energy levels of PBTTT-C14,41 C60, PC61BM, ICBA, bis-PC61BM, tris-PC61BM, and ICTA. TheLUMO energy levels of the electron acceptors used in the study were measured by solution CV and have been internally referenced to the LUMOenergy level of PC61BM for this study.

Table 1. Characterization of Ternary Blends and the Open-Circuit Voltage of Solar Cells Comprising Them

blend A1 A2 morphology ΔLUMOA1‑A2 (eV) Voc (V)

A PC61BM bis-PC61BM A2 does not intercalate 0.12 0.48−0.62B PC61BM ICBA A1 and A2 intercalate 0.17 0.48−0.63C PC61BM tris-PC61BM A2 does not intercalate 0.20 0.48−0.58D PC61BM ICTA A2 does not intercalate 0.42 0.48−0.72E C60 ICTA A2 does not intercalate 0.51 tunablea

aThe Voc for ternary-blend solar cells containing ternary blend E at 20% C60 loading of the total acceptor content is 0.6 V and is an intermediatevalue between what one would expect of PBTTT-C14:C60 (below 0.48 V) and PBTTT-C14:ICTA (0.72 V) binary-blend solar cells.

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00620ACS Energy Lett. 2017, 2, 2149−2156

2150

high as 0.51 eV. Voc tunability requires strong electronicinteractions between A1 and A2; therefore, our findings suggestan electron wave function that is delocalized over large distances.For this study, we investigated five ternary-blend systems

comprising a semicrystalline polymer, poly[2,5-bis(3-tetrade-cylthiophen-2-yl)thieno[3,2-b]thiophene] (PBTTT-C14),41 asthe polymer donor (D) and phenyl-C61-butyric acid methyl ester(PC61BM) as the parent acceptor (A1); a second fullerenederivative served as A2. The chemical structures of each of theconstituents are shown in Figure 1.We focused on ternary blendscontaining one donor and two acceptors, D/A1/A2, to elucidatehow the energy-level offset between the LUMOs of acceptors A1and A2 impacts device performance. All acceptors have similarsurface energies and thus satisfy the morphological criterion ofchemical compatibility in order for us to realize ternary-blendsolar cells with Vocs that are tunable across the compositionwindow of the active layer.36

Table 1 contains the ternary-blend compositions, whether theacceptors intercalate between the side chains of PBTTT-C14,and the energy-level offsets between the LUMOs of PC61BM andA2. Ternary blend A comprises bis-PC61BM as A2, providing anenergy difference between the LUMOs of the acceptors of 0.12eV. In ternary blend B, indene-C60 bisadduct, ICBA, is A2. It has aLUMO energy level that is 0.17 eV higher than that of PC61BM.In ternary blend C, tris-PC61BM is A2; the energy-level differencebetween the LUMOs of the acceptors is 0.20 eV. We alsoexamined ternary blend D, where indene-C60 trisadduct, ICTA,with a LUMO energy level 0.42 eV higher than that of PC61BM,is A2. Finally, to probe how the LUMO−LUMO offset in A1 and

A2 pairs limits Voc tunability, we studied ternary blend E, in whichC60 is A1 and ICTA is A2, to give an energy-level differencebetween the LUMOs of the acceptors of 0.51 eV. We kept theoverall polymer/fullerene weight ratio at 1:4 and only changedthe weight ratio between A1 and A2. The LUMO energy levels forall of the fullerene derivatives were measured using solutioncyclic voltammetry (CV; see Figure S1); the energy-leveldifferences between the LUMOs of PC61BM and A2 for all ofthe blends correlate well with previously reported values.42−44

We fabricated ternary-blend solar cells whose active-layercompositions are specified above. Binary-blend solar cellscontaining PC61BM demonstrate optimal device performanceat 1:4 donor to acceptor weight ratio. While it is known thatbinary-blend solar cells containing PBTTT-C14 and either bis-PC61BM, tris-PC61BM, or ICTA show optimized performancewhen the active layer comprises a 1:1 donor to acceptor weightratio, we kept the overall polymer/fullerene ratio at 1:4 becausePC61BM is present in all of the active layers and PC61BM readilyintercalates between the alkyl side chains of PBTTT-C14.45

Keeping the donor/acceptor ratio constant throughout allowedus to minimize the parameters that can affect solar-cell outputcharacteristics. All of the devices have low efficiencies because thephotocurrents are limited by the low absorption of the acceptorsin the active layers. However, the preservation of FFs above0.42and in many cases above 0.5in all of our devicessuggests that localized trap states are not limiting deviceoperation. Rather, an imbalance in charge-carrier transport inthe active layers results in space-charge limitation.46 This

Figure 2. Open-circuit voltage of optimized ternary-blend solar cells containing (a) PBTTT-C14, PC61BM, and bis-PC61BM; (b) PBTTT-C14,PC61BM, and ICBA; (c) PBTTT-C14, PC61BM, and tris-PC61BM; and (d) PBTTT-C14, PC61BM, and ICTA as a function of the fraction ofPC61BM of the two acceptors; the donor/acceptor weight ratio was kept constant at 1:4.

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00620ACS Energy Lett. 2017, 2, 2149−2156

2151

limitation, however, does not impact the conclusions drawnbased on the Voc trends that we observe herein.Figure 2 demonstrates the Voc behavior of ternary-blend solar

cells upon the introduction of different A2s as the thirdcomponent in these blends. As expected, the introduction ofbis-PC61BM (ternary blend A), ICBA (ternary blend B), and tris-PC61BM (ternary blend C) leads to tunable Voc in ternary-blendsolar cells as the acceptor pairs are mutually compatible and theirLUMO energy level differences are less than 0.2 eV. Solar cellswith ternary blend D (ICTA as A2) display similarly tunable Vocacross its blend composition window, even when the energydifference between the LUMOs of the two acceptors exceeds 0.4eV (Figure 2d).Inspired by the ability to tune the Voc in ternary-blend solar

cells with an energy difference between the LUMOs of the twoacceptors exceeding 0.4 eV (ternary blend D), we furtherfabricated ternary-blend solar cells containing PBTTT-C14, C60,and ICTA (ternary blend E) to test the energetic limits on tuningVoc. Substituting PC61BM with C60 resulted in an energy-leveldifference between the LUMOs of the two acceptors of 0.51 eV.C60, however, has poor solubility in organic solvents, whichlimited its loading to 20% of the total acceptor content in solarcells comprising ternary blend E. At 20% C60 loading, the Voc ofthese solar cells is at 0.60 V, an intermediate value between whatone would expect of PBTTT-C14:C60 (below 0.48 V) andPBTTT-C14:ICTA binary-blend solar cells. This result demon-strates that we can access Voc that is not pinned by thosedetermined by the energy-level difference of PBTTT-C14:C60and PBTTT-C14:ICTA binary-blend solar cells, even when the

LUMO energy-level difference of the two acceptors in ourternary-blend solar cell exceed 0.5 eV. If we invoke the organicalloy model,17,31 the composition-dependent Voc observed in allof the ternary-blend solar cells in this study implies strongelectronic interactions between A1 and A2; therefore, the electronwave function is substantially delocalized across many molecules,even when the LUMO energy levels between A1 and A2 differ bymore than 0.5 eV.15,17,31 It follows that the minority fullerene inD/A1/A2 blends is not creating localized traps. Rather, itsaddition favors delocalization of electronic states over largedistances across the composition window, and according to onestudy, this delocalization can span as many as 30 fullerenemolecules.32

Figure S2 contains representative J−V plots for binary- andternary-blend solar cells in this study. Reference solar cellscontaining PBTTT-C14:PC61BM, exhibit an average Jsc of 6.4mA/cm2, which is comparable to previous reports.47 Incorpo-ration of A2 in PBTTT-C14:PC61BM blends results in solar cellswith progressively decreasing Jsc. Because the fraction of totalacceptor relative to donor is invariant, the reduction in Jsc in allcases stems from the lower absorptivity of bis-PC61BM, ICBA,tris-PC61BM, and ICTA, relative to that of PC61BM.15,48 Despitea decrease in Jsc, the FFs of these devices remain at or around 0.50when PC61BM comprises the major fraction of the acceptor inthe blend. We interpret this observation to imply that chargegeneration and transport in these ternary-blend solar cells areefficient, even when the energy-level difference between theLUMOs of the two acceptors is above 0.4 eV and as high as 0.5eV. Nonetheless, the decrease in Jsc accompanying the increase in

Figure 3. (a) UV−vis absorbance spectrum of a PBTTT-C14 thin film. (b) UV−vis absorbance spectrum of a thin film of PBTTT-C14:ICTA at a1:4 weight ratio. (c) UV−vis absorbance spectra of ternary-blend films containing PBTTT-C14, PC61BM, and ICTA with increasing PC61BMcontent. (d) UV−vis absorbance spectra of ternary-blend films containing PBTTT-C14, PC61BM and ICBA with increasing PC61BM content.

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00620ACS Energy Lett. 2017, 2, 2149−2156

2152

A2 fraction in ternary-blend solar cells results in a gradualdecrease in power conversion efficiency, which stems fromreduced absorption of A2 compared to that of PC61BM. Morespecifically, the reduction in absorption of the active layer whenA2 is introduced accounts for more than two-thirds of thereduction in Jsc between ternary-blend and PBTTT-C14:PC61BM solar cells. This analysis assumes that onlysingle-pass photoabsorption contributes to the Jsc whenphotoabsorption on subsequent passes can account for up to25% of the Jsc.

49 We thus believe the reduction of absorption ofthe active layer to be the first-order reason for the observeddecrease in Jsc in ternary-blend solar cells compared to PBTTT-C14:PC61BM solar cells. It follows that we should be able toincrease the power conversion efficiency by optimizing thethickness of the ternary-blend active layers, but we have opted tomaintain identical processing conditions for all active layers toprovide a common basis with which we can compare across thesolar cells under study. Despite low current densities and powerconversion efficiencies, our model study demonstrates thatcharge generation and transport can be effectively maintaineddespite an energy-level difference between the LUMOs of thetwo acceptors that exceeds 0.4 eV. In binary-blend solar cells withbis-PC61BM, ICBA, tris-PC61BM, and ICTA, we observe FFs of0.42. The lower FF in these devices compared to those for binary-blend solar cells of PBTTT-C14:PC61BM solar cells originatesfrom an imbalance in charge-carrier transport46,50 as the electronmobilities of bis-PC61BM, ICBA, tris-PC61BM, and ICTA are lowrelative to the hole mobility of PBTTT-C14.48

The performance of ternary-blend solar cells in this studysupports the notion of an organic alloy, with strong electronicinteractions between A1 and A2 such that the electron wavefunction is delocalized across many molecules. In the absence ofsuch delocalization between A1 and A2, we would expect to seeevidence of localized trap states or defect states, such as thosepreviously observed when PC61BM is mixed with PC84BM,51

manifested as a precipitous drop in FF because of extensiverecombination and a sharp decrease in Jsc even with a minuteaddition of A2. Instead, our ternary-blend solar cells maintain FFsabove 0.42 throughout the composition window and the drop in

Jsc only comes when A2 with lower absorptivity is incorporatedsubstantially. We thus surmise that our ternary blends formorganic alloys with extensive electron wave function delocaliza-tion between A1 and A2 in order for us to observe composition-dependent Voc in the resulting solar cells. However, furtherstudies are necessary to estimate the extent of electron wavefunction delocalization.51

The fullerene acceptors explored herein are chemicallycompatible with each other and can be divided into twocategories: those that intercalate between the side chains of thePBTTT-C14 (PC61BM and ICBA) to form bimolecularcrystals45 and those that do not,45 presumably due to sterics(bis-PC61BM, tris-PC61BM, and ICTA). This intercalation canbe studied by X-ray diffraction but is more easily tracked viaabsorption measurements. Figure 3a shows the absorptionprofile of PBTTT-C14. We observe in Figure 3b that theintroduction of ICTA does not alter the absorption profile ofPBTTT-C14 significantly. ICTA absorbs in the near-UV; we seean increase in the absorption below 400 nm attributable to ICTA.The absorption at longer wavelengths that is associated withPBTTT-C14 remains unchanged. Consistent with literaturereports,45 the comparison between panels (a) and (b) in Figure 3indicates that ICTA does not intercalate between the side chainsof PBTTT-C14. Figure 3c shows the absorption profiles ofternary blends of PBTTT-C14:PC61BM:ICTA across thecomposition window. While the absorption profiles aresubstantially different from that of PBTTT-C14, they aremarginally different from each other. These absorption profilesare different from that of PBTTT-C14 because PC61BM readilyintercalates between the side chains of the polymer donor,resulting in bimolecular crystals whose absorption is distinctfrom that of PBTTT-C14.52 The observation that adding ICTAand increasing its concentration does not substantially alter theabsorption profiles in Figure 3c indicates the preservation ofbimolecular crystals at all compositions. The absorption profileswith the addition of A2 are qualitatively similar for ternary blendsA and C. Our data thus indicate that in ternary blends containingPC61BM and a nonintercalating fullerene derivative (ternaryblends A, C, and D) the presence of A2 does not disrupt the

Figure 4. Illustration of ternary-blendmorphologies where (a) both acceptors, A1 and A2, intercalate (ternary blend B) and (b) only one acceptor,A1, intercalates (ternary blends A, C, D, and E) between the side chains of PBTTT-C14. Dashed circles represent the extent of delocalization offree charges (electron).

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00620ACS Energy Lett. 2017, 2, 2149−2156

2153

formation of bimolecular crystals between PBTTT-C14 andPC61BM. Figure 3d shows the absorption profiles of ternaryblends of PBTTT-C14:PC61BM:ICBA across the compositionwindow. Because ICBA also readily intercalates between the sidechains of PBTTT-C14, the similarity of absorption profilesacross the composition window suggests that the bimolecularcrystals are structurally similar, whether they form betweenPBTTT-C14 and PC61BM, between PBTTT-C14 and ICBA, orat any other intermediate compositions.Complementary X-ray diffraction analyses are provided in

Supporting Information. Figure S3 contains the X-ray diffractionimages, along with the collapsed one-dimensional X-raydiffraction traces, of thin films containing ternary blends atdifferent A1/A2 ratios. The ternary blends are semicrystalline, asevidenced by the presence of a strong, out-of-plane reflectionalong qxy = 0. Figure S4 shows quantification of the positions ofthe primary reflections extracted from the GIXD patterns ofbinary and ternary blends. Consistent with the picture that binaryblends form bimolecular crystals because one or both of theacceptors is/are capable of intercalating between the side chainsof PBTTT-C14, the X-ray diffraction patterns reveal a primaryreflection associated with the characteristic spacing of thebimolecular crystals at 0.22 and at 0.24 Å−1 for PBTTT-C14:PC61BM and PBTTT-C14:ICBA, respectively. Binaryblends comprised of fullerene derivatives that are bulkier reveala primary reflection between 0.26 and 0.28 Å−1. This larger qspacing corresponds to a smaller characteristic distance that iscomparable to that of PBTTT-C14 alone; this observation thusindicates that bis-PC61BM, tris-PC61BM, and ICTA do notintercalate between the side chains of PBTTT-C14.45 The GIXDpatterns of all ternary-blend thin films in this study yieldedprimary reflections between 0.22 and 0.24 Å−1. This observationis consistent with our absorption studies and is evidence thatbimolecular crystals are formed in all of the ternary blends,independent of whether A2 intercalates because PC61BM does soreadily. Figure 4 provides a simple illustration of these twoscenarios: one in which both the acceptors intercalate betweenthe side chains of PBTTT-C14 (Figure 4a; for ternary blend B,with ICBA as A2) and one in which only PC61BM intercalatesbetween the side chains of PBTTT-C14 (Figure 4b). The latter isthe scenario for ternary blends A, C, and D; while A2 in theseternary blends does not readily form bimolecular crystals withPBTTT-C14, they do not preclude the formation of bimolecularcrystals between PBTTT-C14 and PC61BM.Earlier studies have demonstrated that the morphology of

ternary blends determines the Voc behavior in ternary-blend solarcells.36 As expected, the introduction of ICBA to PBTTT-C14:PC61BM (ternary blend B) leads to tunable Voc in ternary-blend solar cells as both acceptors can intercalate between sidechains of PBTTT-C14 and they are known to be miscible witheach other (Figure 2a).53 This observation is consistent with theorganic alloymodel previously used to describe D1/D2/A andD/A1/A2 systems, in which the electron wave function is delocalizedover large distances and is determined by the averagecomposition of D1/D2 or A1/A2.

15,17,31,54 In ternary blendswhere A2 does not intercalate between the side chains ofPBTTT-C14, the Voc of these solar cells still falls between thosecorresponding to binary-blend solar cells of PBTTT-C14:A1 andPBTTT-C14:A2. Still in the framework of the organic alloymodel, this observation implies strong electronic interactionsbetween constituents A1 and A2.The similarity in how the Voc varies across the composition

window in solar cells containing ternary blends A and B (Figure

2a,b) further suggests that A1 and A2 must maintain strongelectronic interactions with each other; therefore, the electronwave function is delocalized over large distances whether A2intercalates. We have schematized this distance over whichdelocalization takes place with circles in Figure 4b. We believeelectron delocalization to extend beyond the intercalatedacceptors; the Voc of devices comprising such ternary blends isthus determined by the average composition of A1/A2. A recentstudy demonstrated that the electron wave function can extendover 30 fullerene molecules for systems containing one donorand two fullerene acceptors.32 As a result, it is quite possible forchanges in the A1/A2 ratio to alter the overall electron wavefunction, leading to the observation of tunable Voc in ternary-blend solar cells.15−17,32

Figure 2 also shows that the Voc behavior extracted from solarcells comprising ternary blend C is qualitatively different fromthose comprising the other ternary blends. The Voc of solar cellscomprising ternary blends A, B, and D can be tuned continuouslyover 60−130 mV across the entire composition window. Thisbehavior is consistent with earlier observations of the Vocbehavior in ternary-blend solar cells containing one donor andtwo fullerene acceptors.15,39,55 Solar cells containing ternaryblend C exhibit Voc that can be tuned by only 20 mV as thePC61BM content is increased from 20 to 80%. Qualitatively, thisbehavior is consistent with those more commonly observed internary-blend solar cells containing D1/D2/A blends.16,33 Wespeculate that this subtle difference in the Voc behavior betweenthe ternary-blend solar cells stems from small differences in themorphology of the active layer. In the context of the organic alloymodel, the introduction of tris-PC61BM beyond 20% does notalter the electronic structure of the delocalized state. We surmisethat this invariance in delocalization stems from an invariance ofthe local composition spanning the distance over which thefullerenes are delocalized, which then leads to an invariance in thecomposition at the donor/acceptor interface. This hypothesisfurther suggests that tris-PC61BM is only partially misciblebeyond 20% in PC61BM, leading to phase separation of two solidsolutions comprising both fullerenes, but this phase separationmust occur sufficiently far away from the donor/acceptorinterface where charge separation takes place to maintain anear-constant Voc that is intermediate of those of the respectivebinary-blend solar cells. Clearly, further studies are necessary toinvestigate the origin of this difference.23

The ability to tune the Voc in ternary-blend solar cells when theenergy-level offset between the HOMOs of the two donors orLUMOs of the two acceptors exceeds 0.4 eV provides a pathwayto achieve power conversion efficiencies beyond the practicalefficiency limit for binary-blend solar cells, which has beenestimated to be approximately 11% for those containingfullerenes as acceptors.1−8 To exceed this limit, the donor andacceptor constituents of the ternary-blend active layer must bejudiciously selected so (a) their absorption profiles arecomplementary to maximize Jsc and (b) they readily form anorganic alloy so that the Voc is tunable across the compositionwindow. A case in point would be a ternary-blend active layer thatcontains two donors, a low-bandgap constituent with strongabsorption characteristics in the near-IR and a large-bandgapconstituent whose absorption starts in the UV and extends to thevisible range of the solar spectrum. Provided that the energylevels are appropriately matched with the acceptor constituent toenable efficient exciton dissociation and there is extensive wavefunction delocalization between the two donor constituents,36,37

we should be able to extend the practical efficiency limit of single-

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00620ACS Energy Lett. 2017, 2, 2149−2156

2154

junction solar cells. A recent estimate of the efficiency limit forternary-blend solar cells is 14%;38 this estimate assumes anenergy-level offset between the HOMOs of the two donors orLUMOs of the two acceptors of 0.3 eV. That these energy-leveloffsets can be greater than 0.3 eV affords greater tunability of theVoc, which should extend the efficiency limit beyond what hadbeen predicted.In summary, we have demonstrated ternary-blend solar cells

containing one donor and two fullerene acceptors (D/A1/A2)with tunable Voc when the LUMO−LUMO offsets of the twoacceptors exceed 0.4 eV and are as high as 0.5 eV. Independent ofwhether the second acceptor intercalates between the side chainsof PBTTT-C14, chemical compatibility between the acceptorsensures strong electronic interactions, which allows delocaliza-tion of the electron wave function over both A1 and A2.Accordingly, the Voc of such ternary-blend solar cells is tunable,with its magnitude determined by the overall acceptorcomposition. Our results have broadened the library of possiblecandidates for active layers of ternary-blend solar cells withtunable Voc, providing a pathway toward materials design formore efficient single-junction solar cells.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acsenergy-lett.7b00620.

Materials, materials characterization, synthetic proce-dures, UV−vis absorption, CV traces, device fabricationand characterization, J−V curves of ternary-blend solarcells, GIXD images and out-of-plane X-ray diffractiontraces, positions of the primary reflections extracted fromthe GIXD images, and tabulated solar cell data (PDF)

■ AUTHOR INFORMATIONCorresponding Author*E-mail: lloo@princeton.edu.ORCIDMelda Sezen-Edmonds: 0000-0003-0476-6815Yueh-Lin Loo: 0000-0002-4284-0847NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSGIXD experiments were conducted at the Cornell High EnergySynchrotron Source supported by the NSF & NIH/NIGMS viaNSF Award DMR-1332208. We acknowledge funding from theNational Science Foundation (ECCS-1549619 and CMMI-1537011) and Princeton Center for Complex Materials fundedunder NSF-MRSEC (DMR-1420541) for support of P.P.K.,M.S.E., and Y.L.L. J.B.H. and B.C.T. acknowledge support by theNational Science Foundation (CBET Energy for Sustainability)CBET-1436875.

■ REFERENCES(1) Thompson, B. C.; Khlyabich, P. P.; Burkhart, B.; Aviles, A. E.;Rudenko, A.; Shultz, G. V.; Ng, C. F.; Mangubat, L. B. Polymer-BasedSolar Cells: State-of-the-Art Principles for the Design of Active LayerComponents. Green 2011, 1, 29−54.(2) Janssen, R. A. J.; Nelson, J. Factors Limiting Device Efficiency inOrganic Photovoltaics. Adv. Mater. 2013, 25, 1847−1858.

(3) Scharber, M. C.; Sariciftci, N. S. Efficiency of Bulk-HeterojunctionOrganic Solar Cells. Prog. Polym. Sci. 2013, 38, 1929−1940.(4) Zhao, W.; Qian, D.; Zhang, S.; Li, S.; Inganas, O.; Gao, F.; Hou, J.Fullerene-Free Polymer Solar Cells with over 11% Efficiency andExcellent Thermal Stability. Adv. Mater. 2016, 28, 4734−4739.(5) Liu, C.; Yi, C.; Wang, K.; Yang, Y.; Bhatta, R. S.; Tsige, M.; Xiao, S.;Gong, X. Single-Junction Polymer Solar Cells with Over 10% Efficiencyby a Novel Two-Dimensional Donor−Acceptor Conjugated Copoly-mer. ACS Appl. Mater. Interfaces 2015, 7, 4928−4935.(6) Huang, J.; Li, C.-Z.; Chueh, C.-C.; Liu, S.-Q.; Yu, J.-S.; Jen, A. K.-Y.10.4% Power Conversion Efficiency of ITO-Free Organic PhotovoltaicsThrough Enhanced Light Trapping Configuration. Adv. Energy Mater.2015, 5, 1500406.(7) Zhang, S.; Ye, L.; Hou, J. Breaking the 10% Efficiency Barrier inOrganic Photovoltaics: Morphology and Device Optimization of Well-Known PBDTTT Polymers. Adv. Energy Mater. 2016, 6, 1502529.(8) Bin, H.; Gao, L.; Zhang, Z.-G.; Yang, Y.; Zhang, Y.; Zhang, C.;Chen, S.; Xue, L.; Yang, C.; Xiao, M.; et al. 11.4% Efficiency Non-Fullerene Polymer Solar Cells with Trialkylsilyl Substituted 2D-Conjugated Polymer as Donor. Nat. Commun. 2016, 7, 13651.(9) Ameri, T.; Dennler, G.; Lungenschmied, C.; Brabec, C. J. OrganicTandem Solar Cells: A Review. Energy Environ. Sci. 2009, 2, 347.(10) Ameri, T.; Li, N.; Brabec, C. J. Highly Efficient Organic TandemSolar Cells: A Follow up Review. Energy Environ. Sci. 2013, 6, 2390.(11) Zheng, Z.; Zhang, S.; Zhang, J.; Qin, Y.; Li, W.; Yu, R.; Wei, Z.;Hou, J. Over 11% Efficiency in Tandem Polymer Solar Cells Featured bya Low-Band-Gap Polymer with Fine-Tuned Properties. Adv. Mater.2016, 28, 5133−5138.(12) Zhou, H.; Zhang, Y.; Mai, C.-K.; Collins, S. D.; Bazan, G. C.;Nguyen, T.-Q.; Heeger, A. J. Polymer Homo-Tandem Solar Cells withBest Efficiency of 11.3%. Adv. Mater. 2015, 27, 1767−1773.(13) You, J.; Dou, L.; Yoshimura, K.; Kato, T.; Ohya, K.; Moriarty, T.;Emery, K.; Chen, C.-C.; Gao, J.; Li, G.; et al. A Polymer Tandem SolarCell with 10.6% Power Conversion Efficiency. Nat. Commun. 2013, 4,1446.(14) Zhang, K.; Gao, K.; Xia, R.; Wu, Z.; Sun, C.; Cao, J.; Qian, L.; Li,W.; Liu, S.; Huang, F.; et al. High-Performance Polymer Tandem SolarCells Employing a New n-Type Conjugated Polymer as anInterconnecting Layer. Adv. Mater. 2016, 28, 4817−4823.(15) Khlyabich, P. P.; Burkhart, B.; Thompson, B. C. Efficient TernaryBlend Bulk Heterojunction Solar Cells with Tunable Open-CircuitVoltage. J. Am. Chem. Soc. 2011, 133, 14534−14537.(16) Khlyabich, P. P.; Burkhart, B.; Thompson, B. C. CompositionalDependence of the Open-Circuit Voltage in Ternary Blend BulkHeterojunction Solar Cells Based on Two Donor Polymers. J. Am.Chem. Soc. 2012, 134, 9074−9077.(17) Khlyabich, P. P.; Burkhart, B.; Rudenko, A. E.; Thompson, B. C.Optimization and Simplification of Polymer−fullerene Solar Cellsthrough Polymer and Active Layer Design. Polymer 2013, 54, 5267−5298.(18) Ameri, T.; Khoram, P.; Min, J.; Brabec, C. J. Organic TernarySolar Cells: A Review. Adv. Mater. 2013, 25, 4245−4266.(19) An, Q.; Zhang, F.; Zhang, J.; Tang, W.; Deng, Z.; Hu, B. VersatileTernary Organic Solar Cells: A Critical Review. Energy Environ. Sci.2016, 9, 281−322.(20) Huang, H.; Yang, L.; Sharma, B. Recent Advances in OrganicTernary Solar Cells. J. Mater. Chem. A 2017, 5, 11501−11517.(21) Park, K. H.; An, Y.; Jung, S.; Park, H.; Yang, C. The Use of an N-Type Macromolecular Additive as a Simple yet Effective Tool forImproving and Stabilizing the Performance of Organic Solar Cells.Energy Environ. Sci. 2016, 9, 3464−3471.(22) Zhao, W.; Li, S.; Zhang, S.; Liu, X.; Hou, J. Ternary Polymer SolarCells Based on Two Acceptors and One Donor for Achieving 12.2%Efficiency. Adv. Mater. 2017, 29, 1604059.(23) Baran, D.; Ashraf, R. S.; Hanifi, D. A.; Abdelsamie, M.; Gasparini,N.; Rohr, J. A.; Holliday, S.; Wadsworth, A.; Lockett, S.; Neophytou, M.;et al. Reducing the Efficiency−stability−cost Gap of Organic Photo-voltaics with Highly Efficient and Stable Small Molecule AcceptorTernary Solar Cells. Nat. Mater. 2016, 16, 363−369.

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00620ACS Energy Lett. 2017, 2, 2149−2156

2155

(24) Kumari, T.; Lee, S. M.; Kang, S.-H.; Chen, S.; Yang, C. TernarySolar Cells with a Mixed Face-on and Edge-on Orientation Enable anUnprecedented Efficiency of 12.1%. Energy Environ. Sci. 2017, 10, 258−265.(25) Zhang, G.; Zhang, K.; Yin, Q.; Jiang, X.-F.; Wang, Z.; Xin, J.; Ma,W.; Yan, H.; Huang, F.; Cao, Y. High-Performance Ternary OrganicSolar Cell Enabled by a Thick Active Layer Containing a LiquidCrystalline Small Molecule Donor. J. Am. Chem. Soc. 2017, 139, 2387−2395.(26) Gasparini, N.; Lucera, L.; Salvador, M.; Prosa, M.; Spyropoulos,G. D.; Kubis, P.; Egelhaaf, H.-J.; Brabec, C. J.; Ameri, T. High-Performance Ternary Organic Solar Cells with Thick Active LayerExceeding 11% Efficiency. Energy Environ. Sci. 2017, 10, 885−892.(27) Cho, Y. J.; Lee, J. Y.; Chin, B. D.; Forrest, S. R. Polymer BulkHeterojunction Photovoltaics Employing a Squaraine Donor Additive.Org. Electron. 2013, 14, 1081−1085.(28) Lu, L.; Xu, T.; Chen, W.; Landry, E. S.; Yu, L. Ternary BlendPolymer Solar Cells with Enhanced Power Conversion Efficiency. Nat.Photonics 2014, 8, 716−722.(29) Koppe, M.; Egelhaaf, H.-J.; Dennler, G.; Scharber, M. C.; Brabec,C. J.; Schilinsky, P.; Hoth, C. N. Near IR Sensitization of Organic BulkHeterojunction Solar Cells: Towards Optimization of the SpectralResponse of Organic Solar Cells. Adv. Funct. Mater. 2010, 20, 338−346.(30) Ameri, T.; Min, J.; Li, N.; Machui, F.; Baran, D.; Forster, M.;Schottler, K. J.; Dolfen, D.; Scherf, U.; Brabec, C. J. PerformanceEnhancement of the P3HT/PCBM Solar Cells through NIRSensitization Using a Small-Bandgap Polymer. Adv. Energy Mater.2012, 2, 1198−1202.(31) Street, R. A.; Davies, D.; Khlyabich, P. P.; Burkhart, B.;Thompson, B. C. Origin of the Tunable Open-Circuit Voltage inTernary Blend Bulk Heterojunction Organic Solar Cells. J. Am. Chem.Soc. 2013, 135, 986−989.(32) Street, R. A.; Khlyabich, P. P.; Rudenko, A. E.; Thompson, B. C.Electronic States in Dilute Ternary Blend Organic Bulk HeterojunctionSolar Cells. J. Phys. Chem. C 2014, 118, 26569−26576.(33) Khlyabich, P. P.; Rudenko, A. E.; Burkhart, B.; Thompson, B. C.Contrasting Performance of Donor−Acceptor Copolymer Pairs inTernary Blend Solar Cells and Two-Acceptor Copolymers in BinaryBlend Solar Cells. ACS Appl. Mater. Interfaces 2015, 7, 2322−2330.(34) Lu, H.; Zhang, J.; Chen, J.; Liu, Q.; Gong, X.; Feng, S.; Xu, X.; Ma,W.; Bo, Z. Ternary-Blend Polymer Solar Cells Combining Fullerene andNonfullerene Acceptors to Synergistically Boost the PhotovoltaicPerformance. Adv. Mater. 2016, 28, 9559−9566.(35) Lee, T. H.; Uddin, M. A.; Zhong, C.; Ko, S.-J.; Walker, B.; Kim, T.;Yoon, Y. J.; Park, S. Y.; Heeger, A. J.; Woo, H. Y.; et al. Investigation ofCharge Carrier Behavior in High Performance Ternary Blend PolymerSolar Cells. Adv. Energy Mater. 2016, 6, 1600637.(36) Khlyabich, P. P.; Rudenko, A. E.; Thompson, B. C.; Loo, Y.-L.Structural Origins for Tunable Open-Circuit Voltage in Ternary-BlendOrganic Solar Cells. Adv. Funct. Mater. 2015, 25, 5557−5563.(37) Gobalasingham, N. S.; Noh, S.; Howard, J. B.; Thompson, B. C.Influence of Surface Energy on Organic Alloy Formation in TernaryBlend Solar Cells Based on Two Donor Polymers. ACS Appl. Mater.Interfaces 2016, 8, 27931−27941.(38) Savoie, B. M.; Dunaisky, S.; Marks, T. J.; Ratner, M. A. The Scopeand Limitations of Ternary Blend Organic Photovoltaics. Adv. EnergyMater. 2015, 5, 1400891.(39) Cheng, P.; Li, Y.; Zhan, X. Efficient Ternary Blend Polymer SolarCells with Indene-C60 Bisadduct as an Electron-Cascade Acceptor.Energy Environ. Sci. 2014, 7, 2005−2011.(40) Li, H.; Zhang, Z.-G.; Li, Y.; Wang, J. Tunable Open-CircuitVoltage in Ternary Organic Solar Cells. Appl. Phys. Lett. 2012, 101,163302.(41) McCulloch, I.; Heeney, M.; Bailey, C.; Genevicius, K.;MacDonald, I.; Shkunov, M.; Sparrowe, D.; Tierney, S.; Wagner, R.;Zhang, W.; et al. Liquid-Crystalline Semiconducting Polymers withHigh Charge-Carrier Mobility. Nat. Mater. 2006, 5, 328−333.(42) Faist, M. A.; Keivanidis, P. E.; Foster, S.; Wobkenberg, P. H.;Anthopoulos, T. D.; Bradley, D. D. C.; Durrant, J. R.; Nelson, J. Effect of

Multiple Adduct Fullerenes on Charge Generation and Transport inPhotovoltaic Blends with Poly(3-Hexylthiophene-2,5-Diyl). J. Polym.Sci., Part B: Polym. Phys. 2011, 49, 45−51.(43) Faist, M. A.; Kirchartz, T.; Gong, W.; Ashraf, R. S.; McCulloch, I.;de Mello, J. C.; Ekins-Daukes, N. J.; Bradley, D. D. C.; Nelson, J.Competition between the Charge Transfer State and the Singlet Statesof Donor or Acceptor Limiting the Efficiency in Polymer:FullereneSolar Cells. J. Am. Chem. Soc. 2012, 134, 685−692.(44) Nardes, A. M.; Ferguson, A. J.; Whitaker, J. B.; Larson, B. W.;Larsen, R. E.; Maturova, K.; Graf, P. A.; Boltalina, O. V.; Strauss, S. H.;Kopidakis, N. Beyond PCBM: Understanding the PhotovoltaicPerformance of Blends of Indene-C 60 Multiadducts with Poly(3-Hexylthiophene). Adv. Funct. Mater. 2012, 22, 4115−4127.(45) Miller, N. C.; Cho, E.; Gysel, R.; Risko, C.; Coropceanu, V.;Miller, C. E.; Sweetnam, S.; Sellinger, A.; Heeney, M.; McCulloch, I.;et al. Factors Governing Intercalation of Fullerenes and Other SmallMolecules Between the Side Chains of Semiconducting Polymers Usedin Solar Cells. Adv. Energy Mater. 2012, 2, 1208−1217.(46) Mihailetchi, V. D.; Wildeman, J.; Blom, P. W. M. Space-ChargeLimited Photocurrent. Phys. Rev. Lett. 2005, 94, 126602.(47) Miller, N. C.; Sweetnam, S.; Hoke, E. T.; Gysel, R.; Miller, C. E.;Bartelt, J. A.; Xie, X.; Toney, M. F.; McGehee, M. D. Molecular Packingand Solar Cell Performance in Blends of Polymers with a BisadductFullerene. Nano Lett. 2012, 12, 1566−1570.(48) Kang, H.; Cho, C.-H.; Cho, H.-H.; Kang, T. E.; Kim, H. J.; Kim,K.-H.; Yoon, S. C.; Kim, B. J. Controlling Number of Indene SolubilizingGroups inMultiadduct Fullerenes for TuningOptoelectronic Propertiesand Open-Circuit Voltage in Organic Solar Cells. ACS Appl. Mater.Interfaces 2012, 4, 110−116.(49) Song, Y.; Chang, S.; Gradecak, S.; Kong, J. Visibly-TransparentOrganic Solar Cells on Flexible Substrates with All-GrapheneElectrodes. Adv. Energy Mater. 2016, 6, 1600847.(50) Faist, M. A.; Shoaee, S.; Tuladhar, S.; Dibb, G. F. A.; Foster, S.;Gong, W.; Kirchartz, T.; Bradley, D. D. C.; Durrant, J. R.; Nelson, J.Understanding the Reduced Efficiencies of Organic Solar CellsEmploying Fullerene Multiadducts as Acceptors. Adv. Energy Mater.2013, 3, 744−752.(51) Street, R. A.; Krakaris, A.; Cowan, S. R. Recombination ThroughDifferent Types of Localized States in Organic Solar Cells. Adv. Funct.Mater. 2012, 22, 4608−4619.(52) Parmer, J. E.; Mayer, A. C.; Hardin, B. E.; Scully, S. R.; McGehee,M. D.; Heeney, M.; McCulloch, I. Organic Bulk Heterojunction SolarCells Using Poly(2,5-Bis(3-Tetradecyllthiophen-2-Yl)Thieno[3,2,-b]-Thiophene. Appl. Phys. Lett. 2008, 92, 113309.(53) Khlyabich, P. P.; Rudenko, A. E.; Street, R. A.; Thompson, B. C.Influence of Polymer Compatibility on the Open-Circuit Voltage inTernary Blend Bulk Heterojunction Solar Cells. ACS Appl. Mater.Interfaces 2014, 6, 9913−9919.(54) Mollinger, S. A.; Vandewal, K.; Salleo, A. Microstructural andElectronic Origins of Open-Circuit Voltage Tuning in Organic SolarCells Based on Ternary Blends. Adv. Energy Mater. 2015, 5, 1501335.(55) Kang, H.; Kim, K.-H.; Kang, T. E.; Cho, C.-H.; Park, S.; Yoon, S.C.; Kim, B. J. Effect of Fullerene Tris-Adducts on the PhotovoltaicPerformance of P3HT:Fullerene Ternary Blends. ACS Appl. Mater.Interfaces 2013, 5, 4401−4408.

ACS Energy Letters Letter

DOI: 10.1021/acsenergylett.7b00620ACS Energy Lett. 2017, 2, 2149−2156

2156