Add a third ingredient
An organic solar cell absorbs light in a blend of two materials: a donor and an acceptor. Many of today's best cells add a small amount of a second acceptor, a "guest", to make a ternary blend.
Ternary blends work, but often for reasons that are hard to pin down. How strongly the guest talks to the host electronically, and how that sets the energy levels of the mixture, is poorly understood at the level of single molecules. That is the question this paper takes on.
A host and a look-alike guest
The host is L8-BO, a widely used acceptor from the Y-series family. Its backbone is a long, flat chain of fused rings with electron-pulling groups at both ends and in the centre.
The guest, Y18-C3, keeps the conjugated backbone of its parent Y18, with a benzotriazole at the centre. The only change is the alkyl chain on that benzotriazole, cut down to three carbons. Host and guest differ mainly in the central unit and the side chains. The donor polymer in the cells is D18.
Drawings are schematic, after Fig. 1a; they are not chemical structures.
Closer stacking, longer reach
Y-series acceptors are flat, so they stack face to face like plates. Electrons hop between plates through the overlap of their π orbitals; the closer the plates, the better the overlap.
Side chains keep the molecules soluble, but they take up room. In films, Y18-C3 stacks at a shorter π–π distance than L8-BO, with a slightly longer crystal coherence length (21.42 vs 20.27 Å). Its absorption also reaches further into the near infrared: the film optical gap is 1.32 eV, against 1.41 eV for L8-BO.
Schematic, after Figs. 1b and 3c: spacings and spectra are not drawn to scale.
Two molecules, one lattice
Mixing two similar molecules usually gives either separate domains of each or a disordered jumble. Here the two acceptors crystallise together. Single-crystal X-ray diffraction shows L8-BO and Y18-C3 in one common lattice, in a strict 1:1 ratio, π-stacked with each other: a bimolecular co-crystal.
The co-crystal is flatter than either parent: the twist between end groups and core drops to 1.34°, against 7.12° in L8-BO and 5.35° in Y18-C3. Its π–π distances sit between those of the two pure crystals, and the packing is denser, with a more extended 3D stacking network.
Schematic mixed stacking, after Fig. 2a; not the solved coordinates.
A dial, not a switch
Because the backbones are so alike, neighbouring host and guest molecules couple strongly electronically. The blend then behaves less like two materials side by side and more like one new material, an alloy, much as the bandgap of AlxGa1−xAs can be tuned with x.
Adding Y18-C3 shifts the optical gap of the mixed films continuously from 1.38 to 1.34 eV, and the frontier levels move gradually instead of showing two separate sets. The 1:1 blend even emits a new photoluminescence peak of its own. In transient absorption, there are no separate decay channels for the two acceptors: the co-crystal acts as one electronic entity.
Schematic: the curve shows continuous vs. stepwise behaviour, not measured energies.
More current, squarer curve, nearly the same voltage
The D18:L8-BO binary cell gives 19.68% (Voc 0.914 V, Jsc 26.94 mA cm−2, FF 79.94%). Adding a little Y18-C3 (L8-BO:Y18-C3 = 5:1) raises Jsc to 27.63 mA cm−2 and FF to 81.66%, while Voc only slips to 0.908 V.
The result: 20.50%. The small voltage drop follows the narrower optical gap, and is partly offset by lower non-radiative loss (0.257 → 0.234 eV).
Curves drawn from a diode model matched to the Voc, Jsc and FF of Table 1; not the measured traces.
Order, balance, fewer losses
The ternary films keep the host's packing but stack slightly tighter (π–π 3.60 Å) with a longer coherence length (19.5 Å) than the binary. Electrons and holes move faster and more evenly: µh / µe = 1.10 / 1.20 × 10−3 cm2 V−1 s−1, the highest of all the blends.
Because the acceptors behave as one coupled network, the guest does not form deep energy traps. Together, more efficient charge generation, suppressed recombination and balanced transport raise the current and the fill factor.
Schematic animation: the counts are a cartoon, not data.
Half and half still works
In many ternary blends the guest helps only within a narrow range of ratios. Here every ternary ratio tested gives more than 19%. Even at 1:1, with Y18-C3 as half of all the acceptor, the cells reach 19.69% with an FF close to 80% and the highest Jsc of the series.
That robustness is what you would expect if the guest joins the host's lattice instead of disturbing it. The design rule: give the guest the host's backbone, and tune the details that matter, here side chains and central unit.