Seven decades of time, four techniques
In a bulk-heterojunction solar cell a photon first creates an exciton in the donor or the acceptor. At a donor–acceptor interface the exciton becomes a charge-transfer (CT) state: the electron sits on the acceptor and the hole on the donor, only 0.5–1 nm apart and still bound by their Coulomb attraction. The pair must then split, drift to the electrodes and avoid meeting other charges on the way.
No single measurement covers all of this. Transient absorption (TA) and electroabsorption (EA) follow the first 2 ns with sub-picosecond resolution. Optical probing of the electric field (OEFP) tracks how far charges drift in the same window. Transient photocurrent (TPC) and time-delayed collection field (TDCF) continue to 1 µs.
Schematic; time scales from the text.
Chosen for small HOMO offsets
The donor polymer PBDB-T is paired with the fullerene PC71BM and with the Y-series acceptor Y1. The benchmark acceptor ITIC is paired with PBDB-T-2Cl and PDCBT-2F. The main criterion was the gap between donor and acceptor HOMO levels, the "HOMO offset". It is 0.63 eV with PCBM but only 0.12, 0.10 and 0.02 eV in the non-fullerene blends.
Y1 absorbs out to about 800 nm and gives the best cell, 13.2% with 22.2 mA cm−2. The two ITIC cells have the highest voltages (1.02 and 1.12 V), because their donor HOMO lies about 1.5 eV below the acceptor LUMO. The cell with the smallest offset, PDCBT-2F:ITIC, has the highest voltage and the lowest efficiency.
Energy levels from Fig. 1b; cell parameters from the table in Fig. 1d.
The electron leaves within 200 fs
A 515 nm pulse excites the donor in PBDB-T:Y1. The TA spectrum shows no trace of the excited donor, even right after the pulse. Bleaching at about 580 and 630 nm comes from donor molecules that have lost an electron (D+). The electron has moved to the acceptor faster than the ~200 fs time resolution.
The acceptor band at 800 nm hardly changes. Quantum-chemical calculations show why: the Y1 anion absorbs almost where neutral Y1 does, with about half the oscillator strength, so its bleaching is largely cancelled. A narrow positive band at 670 nm is the Stark shift caused by the field of the new CT dipole. The steady-state electroabsorption of the device at −2 V looks the same, shifted by about 10 nm.
The PCBM and PBDB-T-2Cl:ITIC blends behave alike, so nearly every donor exciton becomes a CT state. PDCBT-2F:ITIC is the exception: most of its donor excitations decay to the ground state over hundreds of picoseconds, and only a minor fraction forms CT states.
Spectra schematic, after Fig. 3a; band positions from the text. The decay between curves follows scene 5.
The hole must move fast, or not at all
A 730 nm pulse excites the acceptor instead. Now the hole has to move from the acceptor HOMO to the donor HOMO (channel II), and the HOMO offset is its only driving force.
Comparing donor and acceptor bleaching gives a rough yield. In PBDB-T:Y1 about 6/7 of the acceptor excitations already form CT states by 0.5 ps. About half of the rest follow within a few picoseconds, giving about 90% or more. PBDB-T-2Cl:ITIC reaches 60–70%, and PDCBT-2F:ITIC, with a near-zero offset, hardly 20–30%.
In the ITIC blends no further CT states form later. Hole transfer seems to work only from the fresh, unrelaxed exciton. Once the exciton has relaxed into an excitonic polaron, its energy is lower and the transfer is blocked.
Efficiencies from the text (rough estimates); the pathway drawing is schematic.
A loss that ignores the light intensity
After CT formation the whole TA spectrum drops almost uniformly within a few picoseconds. Within 3 ps the donor bleaching falls to 82% of its peak in PBDB-T-2Cl:ITIC, 75% in PBDB-T:Y1 and 50% in PDCBT-2F:ITIC. With PCBM the drop is marginal.
A uniform drop means charges are going back to the ground state. Recombination between charges from different photons (non-geminate) would speed up at higher excitation densities. Here the decay is the same from 0.05 to 3.3 µJ cm−2, so the electron and hole recombine with their own partner: geminate recombination.
Kinetics schematic, after Fig. 5a,c; the 3 ps values and the fluence range from the text and figure.
A small offset is a low wall
The authors propose a simple picture. To move from the donor back to the acceptor, a hole must climb the HOMO offset. If the offset is small, some holes make the jump and recreate an excited acceptor. Excited acceptors relax to the ground state in about 2 ps, so the pair is lost quickly.
The loss follows the offset: more than 50% at about 0.02 eV (PDCBT-2F:ITIC), about 20% at about 0.1 eV (PBDB-T:Y1 and PBDB-T-2Cl:ITIC) and marginal above 0.6 eV (PCBM).
Why does the loss stop after a few picoseconds? The pair stabilises: the polar surroundings polarise around the charges and form a CT polaron, which lowers its energy and blocks the way back. The EA signal supports this. It decays more strongly than the bleaching, as if the initially delocalised hole contracted and its field was screened. (A related explainer follows hole back transfer in ITIC-family blends.)
Level picture schematic, after Fig. 6, using the PBDB-T:Y1 levels of Fig. 1b. Loss values from the text.
Free in about 30 picoseconds
The EA signal also measures how far apart the charges are. A pair further apart has a bigger dipole and a stronger field. After its first drop, the EA signal rises weakly but reproducibly over tens of picoseconds in every sample: the pair is spreading out until it passes the Langevin radius of about 15 nm, beyond which the charges count as free.
This takes about 30 ps in PBDB-T:Y1 and a little longer with PCBM. Calculations offer a reason for the faster split. The extra electron on Y1 sits mainly on two side groups about 2 nm apart. An electron spread that far is bound less strongly to its hole.
Kinetics schematic, after Fig. 7a; 30 ps, 15 nm and 2 nm from the text. The pair drawing is schematic.
Mobility falls a hundredfold, in every blend alike
OEFP and TPC measure the mean distance that carriers drift along the field, here about 2.3 × 105 V cm−1 at −2 V. Several nanometres appear within tens of picoseconds. Since electrons and holes move in opposite directions, that means 5–10 nm of separation, which fits the 15 nm reached in about 30 ps once diffusion is added. After that the drift grows almost logarithmically with time, so the carriers slow down steadily.
The mobility follows from μ = (dx/dt)/E0. At first it is several times lower with PCBM than in the non-fullerene blends. It falls steeply during the first ~100 ps as carriers relax in the density of states, and by 500 ps it is similar in all blends. It levels off after a few hundred nanoseconds, one to two orders of magnitude below its starting value. In PBDB-T:Y1 the late value, about 5 × 10−4 cm2 V−1 s−1, matches space-charge-limited current measurements. Most carriers are out within 50 ns.
Curves schematic, after Fig. 8a,b; values from the text and axes. Fig. 8 labels the ITIC blend with PDCBT-2F as "PCBT-2F" and "PBDB-T-2F".
Slow recombination, high fill factor
TDCF separates extraction from recombination. When the internal field is cancelled after the pulse, carriers stay put and their number falls only by recombination with other carriers. The bimolecular recombination coefficients γ are typical for organic cells but differ almost threefold. The best cell, PBDB-T:Y1, recombines slowest, and the paper reports that γ anticorrelates with the fill factor. A strong enough internal field must pull carriers out before they meet.
The overall verdict: separation is slightly faster with the non-fullerene acceptors, and transport is much the same in all four cells. The real difference is at the start. A small offset raises VOC, but below about 100 meV the losses from poor hole transfer and fast geminate recombination outweigh that gain. The HOMO offset should be 100 meV or more.
γ and FF after Fig. 9c (γ read from the plot); FF values also in Fig. 1d.