The three synthesized through-space polymers—TSP1, TSP2 and TSP3—are shown in Fig. 1. TSP1 is a through-space polymer comprising a PDI unit connected to a BDT–BDT unit, whereas TSP2 and TSP3 have the PDI unit connected to a BDT–DPP unit. The PDI is attached face-on to the BDT unit for TSP2 and to the DPP for TSP3. The molecular weights of the polymers are as follows: TSP1 (Mn = 19.7 kDa, polydispersity index = 2.3), TSP2 (Mn = 11.6 kDa, polydispersity index = 1.7) and TSP3 (Mn = 72.5 kDa, polydispersity index = 2.3), with the slightly lower molecular weight of TSP2 reflecting the difficulty in purifying the TSM2 monomer. Furthermore, the reference polymers Ref-P1, a BDT–BDT polymer with no tethered PDI (Mn = 63.1 kDa, polydispersity index = 3.68), and Ref-P2, a BDT–DPP copolymer with no tethered PDI (Mn = 58.2 kDa, polydispersity index = 3.78), were made.Fig. 1: Synthetic routes to model-interface polymers TSP1, TSP2 and TSP3.Here we show the key synthetic macrocycle formation and polymerization steps to afford the through-space polymers. Top-down representations are also shown for clarity. dba, dibenzylideneacetone; Ts, tosyl; NBS, N-bromosuccinimide; DMF, dimethylformamide; tol, toluene; BDT, benzodithiopehene.Source dataExperimental characterization of excited statesThe absorption spectra of all three through-space polymers—along with the two reference polymers and a relevant PDI—in chloroform solution are shown in Fig. 2a,b. The maximum absorption peak of TSP1 (Fig. 2a) is slightly blue-shifted compared with Ref-P1, with a \({\lambda }_{\max }\) at 495 nm (2.5 eV), whereas the absorption onset is red-shifted by 0.1 eV. The absorption spectrum shows clear features of both the PDI and Ref-P1 polymer units, and strong photoluminescence quenching was observed compared with Ref-P1 (Supplementary Fig. 1 and Supplementary Table 1). For TSP2 and TSP3 (Fig. 2b), the spectrum can be divided into the 450–570 nm (2.2–2.7 eV) region, where the PDI chromophore absorbs, and the 570–800 nm (1.5–2.2 eV) region, which corresponds with absorption from the conjugated polymer backbone. Both Ref-P2 and TSP3 have absorption maxima at ~760 nm (~1.63 eV) with a vibronic shoulder at ~690 nm (~1.80 eV). The absorption spectrum of TSP2 shows an inversion of the 0–0 and 0–1 intensities, such that the maximum is found at ~690 nm with a smaller peak at ~760 nm.Fig. 2: Optoelectronic characterization.a,b, Absorption spectra in dilute chloroform solution of Ref-P1, PDI and TSP1 (a), as well as Ref-P2, PDI, TSP2 and TSP3 (b). c,d, Photothermal deflection spectroscopy spectra of Ref-P1 (cyan), PDI (gold) and TSP1 (purple) (c), as well as Ref-P2 (green), TSP2 (pink) and TSP3 (blue) thin films (d). e,f, Electroluminescence spectra of Ref-P1 (cyan), PDI (gold) and TSP1 (purple) (e) at an injection current of 100 mA cm–2, as well as TSP2, TSP3 and Ref-P2 (f) as a function of the injection current. The insets show energy level diagrams of PDI and Ref-P1 (c), and PDI and Ref-P2 (d). The HOMOs are obtained from ionization potentials of solid films using air photoemission spectroscopy, whereas the LUMOs are from those HOMO energies plus the corresponding optical bandgaps.Source dataWe performed sensitive photothermal deflection spectroscopy (PDS) (Fig. 2c) and electroluminescence (Fig. 2e) on TSP1 films to experimentally detect the presence of CT states. In PDS, a clear absorption signal (Fig. 2c, purple line) below the onset of the strong polymer absorption (cyan line) is observed, at an energy (1.4–1.8 eV) that is lower than the optical gap of either pure component and consistent with the difference between the measured HOMO and LUMO energies of the polymer and PDI, respectively (Fig. 2c inset diagrams). Furthermore, the electroluminescence spectrum of TSP1 shows emission from a state at a lower photon energy (Fig. 2e, purple) than that of any locally excited states in PDI or Ref-P1 (Fig. 2a). The narrow band-gap Ref-P2, TSP2 and TSP3 polymers do not exhibit any substantial differences when using PDS (Fig. 2d), possibly due to the fact that the formed CT states lie closer in energy to the locally excited states. However, injection-dependent electroluminescence measurements (Fig. 2f) show the presence of distinguishable low-energy states in TSP3 but not TSP2, demonstrating that the regioisomeric manipulation of the through-space donor–acceptor interaction has been successful.Modelling and analysis of excited statesWe first calculate and analyse the excited-state structures of these materials to aid in interpreting their response to photoexcitation. This is uniquely possible in this model system due to the constrained molecular conformation, allowing us to model a reduced parameter space compared with conventional (non-bonded) donor–acceptor systems, where the large number of possible donor–acceptor configurations29 makes modelling of excited states at interfaces extremely challenging. To represent the interface at room temperature, we sample a thermodynamic ensemble of conformers using a high-throughput tight-binding molecular dynamics method—which will include the stiffening effect of conjugation—rather than use an athermal (frozen) model30. Statistical analysis of the geometries of the PDI acceptor unit relative to the polymer backbone from these ensembles shows that the PDI is oriented almost co-facially (±20°) to the polymer backbone and is constrained to sit within a tight seperation range (3–3.5 Å away; Supplementary Fig. 7). Excited-state calculations using time-dependent density functional theory (TDDFT)31,32,33 were performed on samples from the thermal ensemble after a few steps of DFT optimization (to get consistent bond lengths) using the B3LYP/6-31G* functional and basis set. The ensemble was used to efficiently find the global minimum by further DFT geometry optimization of five sampled conformers from the ensemble, with the most energetically stable conformation then selected for detailed calculations and characterization of the excited state.Beginning with an analysis of TSP1, the absorption spectra of the first 28 excited singlet states of the optimized PDI, Ref-P1 and TSP1 structures are shown in Fig. 3a–c. In TSP1, additional states with CT character and low oscillator strengths are present at lower energies than the lowest states in either Ref-P1 or PDI. Together with the experimental results, this confirms the presence of an interfacial state of CT character lying below the local exciton of either the PDI or BDT polymer, with good agreement between the experimental and modelling results. This demonstrates the capability of our synthetic approach to build a donor–acceptor heterojunction.Fig. 3: Modelling of excited states.a–g, Calculated excited-state spectra of PDI (a,b) shown twice on different energy scales to aid in vertical comparison, and trimers of Ref-P1 (c), Ref-P2 (d), TSP1 (e), TSP2 (f) and TSP3 (g). The asterisks indicate the lowest excited local excitonic state in each molecule.We now turn to TSP2 and TSP3 to see whether changing the position of the polymer co-monomer (BDT or DPP) relative to the acceptor affects the excited-state properties of the interface. First, we note that Ref-P2, TSP2 and TSP3 (Fig. 3d,f,g) all have lower-energy locally excited states compared with TSP1 (Fig. 3e) and Ref-P1 (Fig. 3c). This is due to the push–pull nature of Ref-P2, which leads to a lowered LUMO energy that in turn reduces the energy offset between the polymer donor and PDI acceptor in TSP2 and TSP3 compared with TSP1. We believe that the close energetic proximity of the weakly absorbing CT states to the strongly absorbing excitonic states made them indistinguishable in the PDS spectra (Fig. 2d). The experimentally measured electroluminescence (Fig. 2f) is in agreement with the lower-energy CT states seen in TSP3 compared with those seen in TSP2 (Fig. 3f,g and Supplementary Fig. 9). Furthermore, we note that the oscillator strengths of CT states calculated for TSP2 are generally greater than those of TSP3, which is in agreement with the experimentally measured electroluminescence spectra (Supplementary Fig 10), in which emission from TSP2 is more than an order of magnitude brighter than TSP3 at the same injection current with the same device architecture. The brighter CT emission in TSP2 may result from intensity borrowing enabled by the small energy gap between the CT state and closest-lying excitonic state34.We acknowledge here that as the experimental validation of CT states (EL and PDS) was performed on thin-film samples, they therefore may include contributions from intermolecular CT states. However, given the strong agreement with the modelled intramolecular CT states—as well as the strong intramolecular coupling due the face-on orientation between the PDI and polymer moiety—we believe that the CT state emission and absorption are probably dominated by intramolecular interactions, even in thin films.The above findings demonstrate that the constrained relative conformation of the donor and acceptor units in the synthesized molecules allows for accurate modelling of excited states of both singlet and CT character, with theoretical state distributions reproduced well by experimental characterization. The results show that the different molecular structures of the three TS polymers have a substantial impact on the interfacial CT state energies and brightnesses. We now continue our analysis focusing on the effect of donor moieties (BDT and DPP) on the acceptor (PDI).Electron–hole distributionsTo understand how the differences in the arrangement of molecular components relate to the observed state energy and brightness, we analysed our TDDFT calculations using the TheoDORE package35 to extract the CT character and charge distributions of the states. Fragment-based analysis allows us to study the electron and hole distributions of the excited states on individual moieties of the through-space polymers.The calculated electron–hole distributions of the lowest excited state for all three TS polymers (trimers) are visualized through the natural transition orbital decomposition, as well as electron–hole correlation plots indicating the location of the hole (x-axis) and electron (y-axis) on the polymer backbone and PDI (Fig. 4a–c). Correlation plots and natural transition orbital decompositions for higher-energy states are shown in Supplementary Figs. 12–14. In all TS polymers, the lowest excited state showed CT character, with the electron strongly localized on a single PDI moiety. For TSP1, the hole is delocalized (participation ratio, \({\rm{P{R}}}_{\rm{h}}=3.5\)) along the BDT–BDT backbone while the electron lies on the adjacent PDI, suggesting strong interactions between the active PDI and BDT moieties. For both TSP2 and TSP3, the hole is more localized (\({\rm{P{R}}}_{\rm{h}}\approx \,2\)) on the DPP unit, leading to a larger average separation between electron and hole in TSP2 than in TSP3.Fig. 4: Modelling excited-state charge distribution.a–c, Charge distribution of lowest calculated excited state for trimers of TSP1 (a), TSP2 (b) and TSP3 (c). Their oscillator strengths (f) and participation ratios (PR), which indicate the delocalization of the hole, are shown. d,e, The excited-state energy versus the electron–hole separation distance of calculated states of optimized trimers TSP1 (d), and TSP2 and TSP3 (e), are shown in the middle row. Blue symbols indicate the hole is on DPP, whereas red symbols indicate the hole is on BDT. The electron is always on the PDI. f,g, The excited-state energy versus electron–hole separation distance of all calculated states of the thermodynamic ensemble of conformers for trimers of TSP1 (f), and TSP2 and TSP3 (g). Blue symbols indicate CT states where the hole is on DPP, whereas red symbols indicate CT states where the hole is on BDT. The black symbols represent the averages of the calculated locally excited excitonic states.To determine whether this picture is consistent, we plot the energy of all excited states in TSP1, TSP2 and TSP3—all modelled as trimers—as a function of the electron–hole separation (Fig. 4d,e)36. For TSP1 (Fig. 4d), the CT states generally increase in energy with increasing electron–hole separation, which can be explained by the reduction of the Coulombic interaction. We can distinguish two families of CT states in TSP2 and TSP3 (Fig. 4e): states with a stronger localization of the hole on the DPP moieties (blue symbols), and states with the hole more localized on the BDT moieties (red symbols). At the same average electron–hole separation distance, the states where the hole is localized on the DPP are lower in energy than those where it is localized on the BDT fragments, and this is true for either relative position of the PDI. The CT states in TSP3 therefore have generally lower energies than TSP2, as the electron–hole separation is generally shorter due to the proximity of the DPP moiety to PDI in TSP3. The observations made above for the optimized structures are also visible in the collective CT states generated by the full ensemble of conformers (Fig. 4f,g). In the TSP1 ensemble, the excited-state energies are spread across a wider range at the same electron–hole separation. This is consistent with the large delocalization of holes in Ref-P1 and the configuration space explored by the ensemble. Furthermore, TSP2 showed a narrower spread of CT energies than TSP3, despite having similar delocalization characteristics.Kinetics of excited statesThe predicted differences in the energies and electron–hole interactions of the first excited states of TSP2 and TSP3 can be expected to influence excited-state dynamics. To investigate this, we performed transient-absorption spectroscopy measurements on all three TS polymers in solution (Fig. 5a,c,e), as well as Ref-P1, Ref-P2 and PDI (Supplementary Figs. 18–20) with the extracted kinetics of the three TS polymers shown in (Fig. 5b,d,f).Fig. 5: Transient absorption spectroscopy.a–f, Transient absorption spectra (a,c,e) and extracted kinetics (b,d,f) of TSP1 (a,b), TSP2 (c,d) and TSP3 (e,f) dissolved in toluene. TSP1 was excited at 410 nm, whereas TSP2 and TSP3 were excited at 700 nm.Figure 5a,b shows the transient absorption spectrum and kinetics of TSP1 in toluene solution (excited at 410 nm), where the conjugated polymer backbone absorbs more strongly than the PDI. The polymer ground-state bleach (GSB) is observed between 520 and 560 nm, with several photoinduced absorption (PIA) bands also present. Although difficult to assign, the PIA bands are distinct from those of Ref-P1, probably indicating the formation of a CT state faster than the 200 fs instrument response of our transient absorption spectroscopy set-up. The CT state then decays rapidly (within approximately hundreds of picoseconds), in clear contrast to the approximately nanosecond lifetime of PDI and Ref-P1 excitons (Supplementary Figs. 18 and 19).Compared with TSP1, the distinct absorption bands of the polymer donor and PDI acceptor in TSP2 and TSP3 allow for the preferential excitation of either component (PDI excitation is shown in Supplementary Figs. 21 and 22). In the case of TSP3 (Fig. 5e,f), we selectively excite the polymer donor (700 nm pump) to induce electron transfer from the polymer to the PDI. We observe clear GSB signatures for the PDI acceptor (500–570 nm) and polymer donor (620–760 nm), with a new PIA band at 820 nm that is not present in the transient absorption spectra of the individual PDI or Ref-P2 components (Supplementary Figs. 19 and 20). Perylene diimide anions possess absorption bands between 650–850 nm and we thus assign the 820 nm PIA to a negatively charged PDI (referred to here as PIA anion) overlapping with the polymer GSB37. There is minimal spectral evolution within the first picosecond (Fig. 5f), and we conclude that, like TSP1, electron transfer has been largely completed within the 200 fs instrument response of our set-up. After a few picoseconds, the observed transient absorption spectrum decays uniformly, with excited-state recombination completed by around 200 ps. Turning to TSP2, with donor excitation at 700 nm (Fig. 5c,d), there are substantial spectral evolutions within the first 10 ps. At 0.2–0.3 ps, we observe the PDI GSB between 500–570 nm and the polymer GSB between 650–800 nm. However, we do not observe a clear PDI anion PIA at this early time, but instead the Ref-P2 singlet (S1) exciton PIA at 950 nm (Supplementary Fig. 20). Over picosecond timescales, the polymer S1 PIA is lost and the PDI anion PIA at 820 nm grows in. Thus, it seems that electron transfer is much slower in TSP2 than TSP3. Recombination to the ground state then occurs on similar timescales to TSP3.In TSP2, despite the growth of the PDI anion PIA in the first 10 ps, we note there is no change in the intensity of the PDI GSB. This is surprising because if the electronic ground states of the polymer chain and PDI were electronically decoupled, we would expect the PDI GSB to mirror the rise of the corresponding anion PIA. This suggests that there are sufficiently strong electronic interactions between the polymer electron donor and PDI acceptor for the ground state optical transitions to be directly coupled; pumping one transition directly bleaches the other, similar to how the higher energy local π–π* transitions in donor–acceptor co-polymers are bleached when exciting the lowest energy donor to acceptor electronic transition38. We propose that this is enabled by the proximity of CT state energies to the local exciton in TSP2, which may enable mixing of the different excited-state wavefunctions. By contrast, the ultrafast appearance of the PDI GSB in TSP3 is consistent with the much faster rate of CT in this material. Thus, it is not possible to separate the ultrafast CT process from any potential coupling of the PDI and polymer ground-state optical transitions for TSP3.The lifetime of the formed excited state is very short in all through-space polymers, with all excited state species recombining by a few hundred picoseconds. For TSP1, the excited states decay considerably faster than in Ref-P1 and the pristine PDI (Supplementary Figs. 18 and 19). The formed CT state in TSP1 is lower in energy than the excitons in Ref-P1 and PDI, and therefore its faster decay to ground is consistent with the energy gap law39. On the other hand, for TSP2 and TSP3, the states that formed after photoexcitation decay on a similar timescale to those of the reference polymer Ref-P2 (Supplementary Fig. 20), which, being a low-gap D–A copolymer, has a shorter excited state lifetime than Ref-P1 but is similar to other D–A polymers in literature40. We conclude that the fast excited state recombination in all three through-space polymers results from the inability of charges to separate in the isolated polymer chain and the strong donor–acceptor electronic interaction due to the face-on polymer-PDI orientation. Indeed, this may even represent the intrinsic recombination timescales for spin-singlet CT states strongly localized at the donor:acceptor interface in OPV blends which could previously not be determined as the CT state can obtain longer-range electron-hole separation in a real OPV blend.
