A luminescent end group for a lower voltage loss

Give the acceptor a light-emitting end

The best organic solar cells now pass 20%, but they still lose more voltage than inorganic or perovskite cells. The largest avoidable part is non-radiative recombination: an electron and a hole meet and their energy becomes heat. The less light a cell emits when it runs as an LED, the larger this loss, ΔE_nr.

Bright emitters and good solar-cell acceptors rarely come in one molecule. The authors take an A-DA′D-A acceptor like L8-BO and attach a triphenylamine, a strongly luminescent unit, to its end groups. The result, Z-Tri, dissolves well (12 mg mL⁻¹ in chloroform and chlorobenzene) and has a higher glass-transition temperature, 139 °C against 106 °C for L8-BO.

Schematic, after Fig. 1a: block drawing of the unit sequence, not the real structure. T_g from the text (Fig. S2).

Same colour in solution, bluer in the film

In solution the two acceptors look almost the same: absorption peaks at 728 nm (L8-BO) and 726 nm (Z-Tri). So the triphenylamine barely changes the single molecule. In films, Z-Tri peaks at 778 nm, 27 nm bluer than L8-BO at 805 nm, and its optical gap is 1.45 eV against 1.40 eV. The end group changes how the molecules pack.

Cyclic voltammetry puts both frontier levels of Z-Tri higher: HOMO/LUMO at −5.45/−3.52 eV, against −5.69/−3.88 eV for L8-BO. The triphenylamine weakens the pull of the end group on electrons. Z-Tri's levels lie between those of PM6 and L8-BO, a stepwise alignment, and a higher LUMO is what lets a cell reach a higher voltage.

Spectra schematic, after Fig. 1b; peak positions and gaps from the text. Levels after Fig. 1c; PM6 drawn without values.

A poor cell on its own, a record voltage as a guest

With the donor PM6, Z-Tri alone makes a weak cell: 3.97%, with 9.22 mA cm⁻² and a fill factor of 43.0%. Its voltage, though, is 1.001 V. The authors blame poor phase separation; the photocurrent does not saturate even at 3 V of effective bias.

So Z-Tri is used as a guest. Adding it to PM6:L8-BO raises the voltage step by step with content, and the best cell, at 20% Z-Tri, reaches 20.32% (20.11 ± 0.19% over 30 cells) with V_OC = 0.927 V, J_SC = 27.30 mA cm⁻² and FF = 80.3%. The PM6:L8-BO reference gives 19.03% at 0.894 V. The EQE peaks at 92.92% in the ternary cell, against 90.00% and 33.95%. Among reported cells above 20%, the authors find no higher V_OC.

J–V curves from a simple diode model matched to V_OC, J_SC and FF of Table 1 (ideality set to the light-intensity slopes), not the measured curves. EQE shapes schematic, after Fig. 2d; peak values from the text.

Faster, more balanced transport

Several device tests agree that the guest helps charges out. The probability that a photo-generated charge is collected rises from 0.975 to 0.989. The exponent α in J_SC ∝ P_light^α is 0.990 in the ternary cell, against 0.971 and 0.911 in the two binary cells, so fewer charges recombine bimolecularly. V_OC grows with light intensity with a slope of 1.07 kT/q, closer to 1 than the 1.13 and 1.22 kT/q of the binaries, which points to less trap-assisted recombination.

Hole and electron mobilities (space-charge-limited current) both rise, and their ratio approaches one. The ternary cell is also more stable: after 465 h of 1-sun light it keeps 80% of its efficiency, against 70% for PM6:L8-BO, and it keeps 80% after 550 h at 70 °C.

Values from the text (Figs. 2g–i, S5–S7, Table S3). Mobilities in 10⁻⁴ cm² V⁻¹ s⁻¹.

Where the missing voltage goes

The total loss E_loss = E_g − qV_OC splits into three parts. ΔE₁ is unavoidable: any material at room temperature emits black-body radiation above its gap. ΔE₂ is extra radiative loss from absorption below the gap. ΔE₃ = ΔE_nr is the non-radiative part.

ΔE₁ is the same for all three cells, 0.268–0.271 eV. The ternary cell has a slightly larger ΔE₂ than PM6:L8-BO (0.102 against 0.095 eV) but a clearly smaller ΔE_nr (0.196 against 0.218 eV), and its total loss falls from 0.592 to 0.571 eV. PM6:Z-Tri has the smallest total loss, 0.545 eV, because its ΔE_nr is only 0.137 eV. Its Urbach energy, a measure of disorder at the absorption edge, also drops: 21.2 meV for the ternary blend against 22.4 meV.

Values from Table 2; E_g from the crossing of absorption and emission. The three reported parts add up to within 0.011 eV of E_loss; the bars show them as reported.

Emitted light as a gauge of loss

The non-radiative loss is read from the cell's electroluminescence efficiency:

ΔE_nr = −kT · ln(EQE_EL)

PM6:L8-BO emits 2.1 × 10⁻⁴ photons per injected electron, the ternary cell 5.0 × 10⁻⁴ and PM6:Z-Tri 4.1 × 10⁻³. Each factor of ten is worth about 59 meV. The authors call 0.137 eV the lowest ΔE_nr reported for an organic solar cell so far.

Why is PM6:Z-Tri so bright? Its charge-transfer (CT) state sits closer to the acceptor's own excited state (LE). The measured LE–CT offset is 0.362 eV against 0.519 eV in PM6:L8-BO, and the emission lives longer. A CT state that mixes more with a bright LE state borrows its brightness. (The role of a bright guest is also the subject of the ternary-voltage explainer.)

EQE_EL and ΔE_nr from Table 2; LE–CT offsets from the text (Fig. S10). The line is drawn at 300 K.

The mixture outshines both acceptors

For narrow-gap acceptors, the photoluminescence quantum yield (PLQY) of the neat film tracks the radiative losses of the cell. Films of L8-BO and Z-Tri give 1.39% and 1.60%. The surprise is the acceptor mixture: L8-BO:Z-Tri reaches 1.87%, more than either component. A simple mixture would land between the two, so the two molecules must interact.

Electroluminescence says the same. The spectrum of the ternary cell lies between those of the two binary cells, as expected. But no linear combination of the two binary spectra reproduces it.

PLQY from the text (Fig. 3c). EL spectra schematic, after Fig. 3d and Fig. S15 (positions and widths illustrative).

Transient absorption finds a mixed acceptor phase

Transient absorption (pump at 650 nm) watches the bleach of the ground state near 800 nm. For the L8-BO:Z-Tri film the bleach peak sits between those of the two neat films, as their absorption and emission do.

The edge is different. Neat L8-BO has a sharp bleach onset near 840 nm, cut by an induced absorption around 850 nm that the authors attribute to polarons from interacting end groups. With only 20% Z-Tri this feature disappears and the bleach onset moves about 40 nm to the red. Pumping at 900 nm, where Z-Tri does not absorb, gives the same spectra, which still differ from neat L8-BO.

The authors conclude that L8-BO and Z-Tri form a new mixed phase, with fewer polarons and less aggregation-caused quenching. That explains the brighter mixture. In the full blends with PM6, charge separation runs as in PM6:L8-BO.

Spectra schematic, after Fig. 3e and Figs. S16–S19; only the positions named in the text are used. Packing sketch after Fig. 3f.

Two acceptors that like each other

How well two materials mix can be estimated from their surface tensions γ, with a Flory–Huggins parameter χ ∝ (√γ_A − √γ_B)². PM6, L8-BO and Z-Tri give 17.2, 26.1 and 24.2 mN m⁻¹. The donor–acceptor pairs give χ = 0.93 (PM6/L8-BO) and 0.60 (PM6/Z-Tri); the two acceptors give only 0.04, so they mix very well.

X-ray scattering (GIWAXS) shows face-on stacking in all films. Neat Z-Tri stacks more loosely than L8-BO (π–π distance 4.104 against 3.890 Å) and with a shorter coherence length (12.93 against 14.63 Å). Yet in the ternary blend the π–π coherence length grows from 15.66 to 16.07 Å. Infrared nanoscopy (PiFM) shows a uniform fibril network. The authors conclude that a luminescent guest which forms a well-mixed phase with the host is a route to lower ΔE_nr.

Surface tensions and χ from the text (Table S4); χ values recomputed here agree. GIWAXS from the text (Table S5). Packing sketch illustrative.