Where a side chain splits
Good acceptors need two things that tend to fight. Ordered, crystalline packing moves charges well. A high photoluminescence quantum yield (PLQY) means few excitations end as heat, and so a smaller non-radiative voltage loss. But stronger order often quenches the emission.
The authors use a small chemical knob: where a branched alkyl chain splits. In L8-BO, both outer thiophenes carry 2-butyloctyl, which branches at the second carbon. The new series L8-BO-Cn+1 keeps 2-butyloctyl on one side and moves the branch on the other side to the first, third, fourth or fifth carbon. L8-BO-C4C4 moves it to the fourth carbon on both sides (4-butyldecyl).
Schematic, after Fig. 1a: block drawing of the backbone and a zigzag sketch of the outer side chains (arm lengths illustrative). The inner chains on the nitrogens are not drawn.
Colour and energy levels barely move
The side chain does not touch the conjugated backbone, and it shows. In dilute solution all acceptors absorb almost identically. Cyclic voltammetry gives the same frontier levels within 0.03 eV: LUMOs from −3.92 to −3.94 eV and HOMOs from −5.69 to −5.72 eV.
Only in films does a small difference appear. L8-BO-C1 and L8-BO-C3 keep the optical gap of L8-BO, 1.41 eV, while the gap narrows to 1.39 eV (C4), 1.37 eV (C5) and 1.36 eV (C4C4). The difference must come from how the molecules pack.
Levels after Fig. 1c (values as printed there); film gaps from the text. "≈1.41" marks the two gaps the text calls similar to that of L8-BO.
Positions three and four win
With the donor PM6, moving the branch inward to the first carbon hurts: 18.26% against 18.91% for L8-BO. Moving it outward to the third or fourth carbon raises VOC, JSC and FF together, to 19.63% (C3) and 19.78% (C4). Going further, to C5 or to both sides (C4C4), lowers the voltage again, to 0.857 and 0.853 V, and the efficiency to 18.69% and 18.54%.
Adding 20 wt% of a brominated relative, L8-BO-C4-Br, as a guest gives 20.42% (VOC 0.894 V, FF 81.7%), certified at 20.1%. The ternary blend still gives 18.51% at 300 nm thickness, 18.63% on 1 cm2 and 16.50% in a 16.8 cm2 module. A C4 cell keeps 80% of its efficiency for over 600 h of light, longer than an L8-BO cell.
J–V curves from a simple diode model matched to VOC, JSC and FF of Table 1 (ideality 1.2 assumed), not the measured curves. PCE values from Table 1.
The gain is in the non-radiative loss
The total loss Eloss = Egap − qVOC splits into three parts. ΔE1 is the unavoidable radiative loss of any absorber at room temperature; it is 0.263–0.264 eV for all six cells. ΔE2 is extra radiative loss from absorption below the gap, 0.056–0.068 eV. ΔE3 is the non-radiative part.
ΔE3 is what separates the molecules. It drops from 0.224 eV (L8-BO) to 0.209 eV (C3) and 0.210 eV (C4), and rises to 0.219 eV (C5) and 0.231 eV (C4C4). So the total loss is smallest for C4, 0.541 eV.
All values from Table 2; Egap from the derivative of the EQE spectrum. The three parts add up to within 0.006 eV of Eloss; the bars show them as reported.
Brighter films, brighter cells
ΔE3 is read from the light the cell gives off when driven as an LED, ΔE3 = −kT ln(EQEEL). The C3 and C4 cells emit 2.90 × 10−4 and 2.87 × 10−4 photons per electron, against 1.65 × 10−4 for L8-BO and 1.27 × 10−4 for C4C4.
The same order shows in the neat acceptor films. Their PLQY, measured in an integrating sphere, is highest for C4 (about 2.2%) and C3, and lower than L8-BO for C1, C5 and C4C4. Mixing with PM6 does not change the acceptors' crystallinity or aggregation noticeably, so the PLQY of the neat acceptor is a fair gauge of the cell's non-radiative loss.
EQEEL from Table 2. PLQY values read off Fig. 1i by eye (≈, means of n ≥ 3 films); the text gives only their order. Emission beyond 1,000 nm was added with an InGaAs detector.
Less static disorder, fewer dark low-energy states
Why does the branch position change the emission? In solution all acceptors have similar PLQY; only films differ. So it is about packing. A film with more energetic disorder has a broader spread of energies, with more low-lying states. By the energy-gap law, low-energy states lose their energy to vibrations faster, so more excitations end as heat.
Temperature-dependent PL separates static from dynamic disorder. The static disorder follows the PLQY trend, except for C1, whose crystal packing differs. Moving the branch outward reduces the crowding near the backbone and the disorder. At C5, or with 4-butyldecyl on both sides, the longer chains bring disorder back. C4 balances the two, with the lowest static disorder and the highest PLQY. Calculations on the crystals add a related hint: the share of J-type aggregates, which emit well, grows from C1 to C4.
Schematic, illustrating the argument of the text; the disorder values (Supplementary Table 10) are not shown. No numbers are drawn.
Bigger networks for electrons
Molecular dynamics simulations of the neat films show the packing side. When the branch sits at the first carbon, the chain splits right next to the flat backbone and sticks out on both sides of it: it blocks 7.3 Å in the stacking direction, against about 5.3–5.9 Å for the others. C1 molecules therefore stack poorly.
Counting molecules that touch face to face (more than 16 atom pairs closer than 4 Å), the largest connected network in the simulated films grows from 33 (C1) to 44 (L8-BO), 97 (C4) and 187 (C4C4) molecules, and the number of stacked pairs from 170 to 265. The donor parts of neighbouring molecules stack more closely, while acceptor–acceptor and donor–acceptor contacts barely change.
Values from Fig. 2 and the text. Side-view and network sketches schematic, after Fig. 2a–h.
Flatter molecules, closer pairs
Single crystals of C1, L8-BO, C3 and C4 show the packing directly. The backbone flattens as the branch moves out: the twist between end group and core falls from 4.77° (L8-BO) to 1.69° (C3) and 1.28° (C4), while C1 stays at 4.70°.
All four form a three-dimensional network, but C1 leaves much larger voids (21.0/24.5 Å against about 17–18/20–21 Å). Its pairs also stack differently, with a Y-type instead of a Y′-type dimer. For the S-type pair, C3 and C4 stack at about 3.4 Å against 3.5 Å for L8-BO and C1, and C4 has the shortest S′ distance and the smallest average π–π distance.
Distances, angles and void sizes from Fig. 3a,b and the text. Network sketch schematic, after Fig. 3a.
Order keeps rising; brightness peaks at four
X-ray scattering (GIWAXS) of the films agrees with the crystals. All molecules lie face-on. The π–π peak moves from 1.70 Å−1 (3.69 Å; C1 and L8-BO) to 1.71 Å−1 (3.67 Å) for the farther branch points. The coherence lengths grow steadily as the branch moves out and are largest for C4C4, which also has the highest electron mobility, 7.70 × 10−4 cm2 V−1 s−1.
So crystallinity rises all the way, but PLQY peaks at C4. In blends with PM6, the fibrils of both materials form an interwoven network, and they grow from C1 to C4C4; symmetric C4C4 goes too far. The asymmetric shift from the second to the fourth carbon is the best compromise, and the authors find the same strategy works for other A–DA′D–A acceptors.
PCE from Table 1; PLQY read off Fig. 1i (≈). π–π peaks, mobility and the coherence-length trend from the text. Trend arrows are qualitative.