A dye that oxidizes only in the light
Doping an organic semiconductor means adding or removing electrons. Normally that takes a strong molecular dopant, and strong dopants tend to be reactive and unstable. Photocatalytic doping avoids them. A thin polymer film is dipped into an acetonitrile solution of a photocatalyst and lit with blue light (455 nm, 50 mW cm⁻²).
The photocatalyst here is 10-methylacridinium (Acr-Me+). It is harmless in the dark. Its excited state, Acr-Me+*, takes an electron from the polymer PBTTT, which leaves a positive polaron on the chain. Oxygen from the air then takes the electron back, so the catalyst returns to its starting state. Only the oxygen and the salt are used up. The Li+ salt supplies the anions that move into the film to balance the polarons' charge.
Schematic, after Fig. 1a. The film is ~30 nm thick; the solution holds 10 mM Acr-Me⁺ and 100 mM Li⁺ salt (Methods).
The counterion is not optional
Without any added salt the doped PBTTT film reaches only 1.31 ± 0.23 S cm⁻¹. With LiTFSI, the salt used in the group's earlier work, it reaches 659 ± 188 S cm⁻¹, more than two orders of magnitude higher.
A polaron left on its own is unstable. It needs a negative ion close by, and the only anions that can enter the film come from the salt. Up to now the salt was treated as passive, a supply of counterions that does not change how fast the electron transfer runs.
Values from the text (Fig. 1c, Fig. S1). Log scale.
The smaller the anion, the higher the conductivity
The authors tested four lithium salts of the same family of imide anions. Their side groups grow from F (FSI−) through CF3 (TFSI−) and C2F5 (PFSI−) to C4F9 (NFSI−). Conductivity falls as the anion grows: 970 ± 107 S cm⁻¹ with FSI−, 659 ± 188 with TFSI− and 187 ± 40 with NFSI−.
The trend is the same in thin and thick films, so it is not about how deep ions can diffuse. It also holds for other acridinium catalysts and for the glycolated polymer PgBTTT, which reaches 2005 ± 160 S cm⁻¹ with FSI−. Other doping methods, such as ion-exchange doping, usually work best with large counterions. That opposite trend hints at a different mechanism.
Values from the text (Fig. 1c–e, Fig. S6). The PFSI⁻ value is in Table S1 (not consulted) and is not drawn. Anion drawings are schematic.
The neutral band fades, the polaron band grows
Absorption spectra track the doping directly. Neutral PBTTT absorbs at 550 nm. Polarons absorb in the infrared, monitored here at 1600 nm. With FSI− the 550 nm peak bleaches fast and the infrared band rises strongly within the first 5 minutes of light. After 10 minutes the neutral peak is almost gone, so the film is fully doped.
With NFSI− the bleaching is much slower, and a residual neutral peak is still there after 15 minutes. Conductivity rises fastest with FSI− over the same first 5 minutes.
Schematic, after Fig. 2a,b: band shapes and time courses are drawn, not digitised.
Three methods give the same order
X-ray photoelectron spectroscopy separates the sulfur signal of neutral and oxidized thiophene units. The oxidized fraction, used as the doping level, is 41%, 37%, 33% and 16% for FSI−, TFSI−, PFSI− and NFSI−. The counterions' own O, N and F signals follow the same order, so fewer anions enter the film as they grow.
Electron paramagnetic resonance (EPR), which detects the unpaired spins of polarons, weakens from FSI− to NFSI−. Pristine PBTTT shows no signal. The Seebeck coefficient, which falls as carrier density rises, is lowest with FSI− (23.44 ± 0.80 µV K⁻¹) and highest with NFSI− (38.62 ± 0.45 µV K⁻¹).
Values from the text (Fig. 2c–e, Fig. S11). Changing the assumed XPS peak shift by ±0.1 eV changes the doping levels by only ~3–4%.
Anions settle among the side chains
X-ray scattering (GIWAXS) shows that PBTTT stays edge-on after doping, with its backbones standing upright on the substrate. Doping widens the lamellar spacing, the distance between backbones set apart by the side chains, and slightly shrinks the π–π stacking distance. Both changes fit counterions sitting mainly among the side chains.
From FSI− to NFSI− the π–π distance grows from 3.57 to 3.67 Å, and the lamellar spacing falls from 27.92 to 26.62 Å. These changes track the lower doping level. Fewer anions inside means less swelling and looser π-stacking.
End values from the text (Fig. 2f, Table S2). The drawing is schematic and not to scale.
Large anions leave the dye aggregated
The key evidence comes from the solution. Dilute Acr-Me+ (0.1 mM) absorbs at 358 and 416 nm, and the salts leave these peaks alone. A broad band between 500 and 700 nm, however, grows as the anion gets larger. It is stronger at 10 mM, the concentration used for doping. This band is charge-transfer absorption between stacked acridinium ions, a sign of aggregates. The 10 mM solution looks bright yellow with FSI− and dark green with NFSI−. The salts themselves are transparent and redox-inert.
Time-resolved fluorescence, measured in Huotian's part of the work, shows that the excited catalyst lives longer with small anions: 31.28 ns (FSI−), 26.99 ns (TFSI−), 24.08 ns (PFSI−) and 21.72 ns (NFSI−). A longer-lived excited state has more time to reach the polymer and take an electron.
Peak positions and lifetimes from the text (Fig. 3a–c, Fig. S21). Spectra and vial colours are schematic, after Fig. 3b,c. The PL experiments were done by H. Zhang and F. Gao (author contributions).
FSI⁻ wedges the stacks apart
Density functional theory (DFT) calculations reproduce the 450–700 nm band for stacks of two to four Acr-Me+ ions. The bulkier anions (TFSI−, PFSI−, NFSI−) stabilise the stacks through van der Waals contacts. Only the small FSI− pushes into or next to a dimer, increases the spacing between the ions and weakens the aggregate band. The more FSI− ions, the larger the distortion.
Breaking up the stacks also changes the electron affinity (EA) of the catalyst. Measured EA levels are −4.35 eV for pristine (aggregated) Acr-Me+, −4.37 eV with NFSI− and −4.51 eV with FSI−, close to the monomer. DFT gives a ~0.1 eV increase with FSI−. The shift is modest, but the group earlier saw several-fold conductivity gains from 0.1–0.2 eV EA differences between catalysts. Together with more free catalyst molecules, it can explain the better doping.
EA values from the text (Fig. S22); DFT results from Fig. 4 and the text. Molecule drawings are schematic.
Take the salt away, and the difference goes too
If the anion acted only inside the film, its size should matter at any concentration. The test: use only 0.1 equivalents of LiFSI or LiNFSI instead of 10. At this low dose both solutions show a similar 500–700 nm aggregate band, and the doped films reach similar conductivities.
So the counterion acts already in the solution, where electrons are transferred, by controlling how many catalyst molecules are free and active. Counterion size is a new design handle for photocatalytic doping. The authors note that the microscopic details may differ for other acridinium catalysts.
Schematic, after Fig. 3d,e. The paper reports the 0.1 eq conductivities as similar; numbers are not quoted in the text, so the bars are drawn equal.