A. Å-scale microstructure

We start by exploring reP3HT and its blends. As expected, pristine reP3HT forms ordered semicrystalline films illustrated by three orders of (h00) Bragg peaks in the out-of-plane direction, indicative of end-to-end lamellar packing with a spacing of d = 16.5 Å, consistent with previous reports. ^{Reference Verploegen, Mondal, Bettinger, Sok, Toney and Bao11,Reference DeLongchamp, Kline, Fischer, Richter and Toney13,Reference Treat, Brady, Smith, Toney, Kramer, Hawker and Chabinyc36–Reference O’Connor, Kline, Conrad, Richter, Gundlach, Toney and DeLongchamp38} This is shown in Fig. S1 of the Supplementary Material. Bragg reflections in reP3HT films are angularly dependent, with highest intensity occurring along the Q _z axis, indicating that lamellae are packed predominantly edge-on with respect to the substrate. Consistent with this orientation, the inter-chain backbone π-stacking is primarily in-plane, as evident by a high-intensity (010) reflection near Q _xy ∼1.67 Å⁻¹ with a d-spacing of 3.76 Å.

Figure 2 shows GIXD images for both reP3HT and raP3HT blended with the three small molecules at a fixed monomer/molecule mole ratio corresponding to a 1:1 w/w P3HT:Spiro-MeOTAD blend. All shown films were thermally annealed. Blending reP3HT with PCBM results in the addition of a broad halo centered at Q ∼1.35 Å⁻¹ due to scattering by amorphous PCBM aggregates. The polycrystalline features of reP3HT are retained with a bimodal lamellar crystallite orientation distribution indicated by the Q = 0.38 Å⁻¹ (100) reflection with intensity peaking near the Q _z and Q _xy axes. Such a distribution is consistent with the presence of a bimodal (010) Bragg reflection. GIXD images of reP3HT blends with PDI [Fig. 2(b)] follow a qualitatively similar trend compared to PCBM blends, with (h00) reflections with highest intensity along the Q _z and Q _xy axes, indicating preferred crystallite orientations in both edge-on and face-on fashions. Blends of reP3HT with Spiro-MeOTAD show a P3HT crystallite orientation distribution that differs qualitatively from both PCBM and PDI. Specifically, the orientation changes to primarily edge-on, similar to neat reP3HT films (Fig. 3).

FIG. 2.

2D GIXD patterns of thermally annealed thin film blends of reP3HT with (a) PCBM, (b) PDI, and (c) Spiro-MeOTAD, as well as raP3HT with (d) PCBM, (e) PDI, and (f) Spiro-MeOTAD. reP3HT blends show characteristic (h00) lamellar diffraction peaks of the semicrystalline polymer and no obvious Bragg peaks due to any of the small molecules. In contrast, raP3HT is completely amorphous, while only the raP3HT:PDI blend shows obvious diffraction peaks of PDI. The largely vertical Bragg rods of PDI indicate that its crystallites are highly oriented relative to the substrate.

FIG. 3.

Orientation distribution functions of the reP3HT lamellar (100) Bragg reflection as a function of the polar angle in neat reP3HT films as well as in blends with small molecules at the same monomer/small molecule mole ratio. The polar angle = 0° for the surface normal. PDI and PCBM blends show dual texture with a large face-on-oriented crystallite fraction, whereas Spiro-MeOTAD blends show primarily edge-on-oriented reP3HT crystallites.

In contrast to reP3HT, raPHT is amorphous in both neat films and blends, even after thermal annealing. This is due to the fact that, on average, raPHT has a shorter delocalization length of its electronic wavefunction compared to reP3HT, leading to weak inter-chain interactions and thus little propensity to crystallize. Since wavefunction delocalization tends to planarize the polymer backbone, reP3HT is characterized by a larger persistence length than raP3HT. This quantity is related to the mean chromophore size, which is a nontrivial quantity to estimate. The amorphous nature of raP3HT is evident in the lower panels of Fig. 2, which show a complete absence of P3HT Bragg peaks. PDI was the only small molecule that showed crystallite formation in raP3HT, which is consistent with its relatively planar core, leading to relatively strong π-stacking interactions. Interestingly, this suggests that in reP3HT blends, partial crystallization of the polymer may limit crystallization by the PDI. We found that changing the solvent from the rapidly evaporating chloroform to the lower vapor pressure chlorobenzene leads to the appearance of (isotropic) diffraction rings due to PDI as illustrated in Fig. S2. Thus, kinetic trapping of the initial film state may evidently lead to stunted PDI crystallization even after thermal annealing.

The GIXD images above clearly show that the distribution of reP3HT crystallite orientations relative to the substrate is a function of the small molecule chemical structure. To put this observation on a more quantitative footing, we have reduced 2D (100) reP3HT Bragg reflections to 1D crystallite orientation distribution functions (ODFs) by (i) first radially integrating the Bragg peak in polar coordinates over a small Q region centered about the peak, (ii) scaling the resulting intensity with a sin(polar angle) factor to account for the angular dependence of the number of crystallites contributing to diffraction at a given Q-space point, ^{Reference DeLongchamp, Kline, Fischer, Richter and Toney13,Reference Baker, Jimison, Mannsfeld, Volkman, Yin, Subramanian, Salleo, Alivisatos and Toney39} and (iii) by normalizing the corrected intensity by its integral over the accessible polar angle range.

The ODF for pristine reP3HT films indicates lamellar crystallites with a preferred edge-on orientation between approximately 10–45° of the surface normal with a peak centered about 25°. As mentioned previously, both PCBM and PDI blends with reP3HT show bimodal texture; closer inspection of the ODFs for both molecules shows that the predominant P3HT crystallite orientation is face-on (π-stacking out-of-plane), with a minor fraction displaying an edge-on orientation. Interestingly, the ODF for the reP3HT:Spiro-MeOTAD blend indicates that P3HT crystallites in this blend are not only primarily edge-on-oriented, in contrast to the other small molecules, but that they are more highly edge-on-oriented than neat, thermally annealed reP3HT films. We can operationally define the ratio of face-on- to edge-on-oriented crystallite populations as the ratio of ODF integrals from ∼0° to 45 and 45° to ∼90°. This calculation yields ratios of 0.52, 0.65, 1.04, and 1.23 for reP3HT:Spiro-MeOTAD, neat reP3HT, reP3HT:PDI, and reP3HT:PCBM, respectively.

B. Nanometer-scale morphology

We have used GISAXS link the Å-scale microstructure of our blend films with its nanometer-scale morphology responsible for long-distance charge transport through the interpenetrating polymer/molecule network. ^{Reference Guo, Ruderer, Rawolle, Korstgens, Birkenstock, Perlich and Muller-Buschbaum40–Reference Ungar, Liu, Zeng, Glettner, Prehm, Kieffer and Tschierske43} Fig. 4 shows in-plane line cuts obtained from 2D GISAXS images near the image horizon. The monotonically decreasing GISAXS curve for the reP3HT:PCBM blend [Fig. 4(a)] shows two shoulders (plateau-like regions) at Q _xy ∼0.03 Å⁻¹ and Q _xy ∼0.08 Å⁻¹, corresponding to two distinct characteristic length scales of roughly 20 and 8 nm, respectively. We interpret these features as arising due to the presence of a hierarchical blend network mesostructure, as no such features exist for pure reP3HT films. The 20 nm feature is possibly related to phase-separated P3HT crystallites—a size scale that has previously been observed for reP3HT blends using AFM. ^{Reference Dante, Peet and Nguyen44–Reference Ayzner, Tassone, Tolbert and Schwartz46} The second length scale of ∼8 nm is consistent with small, largely disordered PCBM clusters, which give rise to the broad scattering ring at wide angles at ∼1.4 Å⁻¹ [Fig. 2(a)]. We note in passing that, though tempting, using the Guinier approximation ^{Reference Guinier47} to calculate effective radii of gyration from shoulder-like features in SAXS curves is problematic for thin blend films due to the expected importance of inter-particle scattered wave interference. ^{Reference Weyerich, Brunner-Popela and Glatter48}

FIG. 4.

Horizontal 1D line cuts of GISAXS patterns taken near the image horizon for (a) reP3HT and (b) raP3HT blends with different small molecules. Only reP3HT:PCBM and raP3HT:Spiro-MeOTAD blends show the appearance of shoulder-like features in this Q range, which are associated with formation of a characteristic length scale of polymer/small molecule phase separation on the nanometer scale. This feature is substantially more pronounced in the reP3HT:PCBM blend.

In stark contrast, reP3HT:PDI films give rise to a featureless GISAXS curve in the accessible Q range, suggesting two possible explanations: The PDI is fully miscible in the polymer matrix and forms no phase-separated aggregates on these length scales, or PDI aggregates are large enough so as to fall outside the limits of detection. For reP3HT:Spiro-MeOTAD blends, the GISAXS profile is similarly featureless with no obvious characteristics of phase separation over this length scale range.

Figure 4(b) shows GISAXS curves for raP3HT blend films. raP3HT:PCBM and raP3HT:PDI display featureless monotonic curves with no well-defined length scales of phase separation. Though similar to reP3HT:PDI blends, for PCBM this represents a drastic change in the microstructure relative to its morphology in the reP3HT matrix. This is not altogether surprising. Early work on organic solar cell performance based on conjugated polymer:PCBM blends showed that for semicrystalline reP3HT, the polymer:fullerene weight ratio that led to the maximum power conversion efficiency was close to 1:1. ^{Reference Kim, Choulis, Nelson, Bradley, Cook and Durrant49} However, for PCBM blends with amorphous MEH-PPV polymer, the optimal polymer:fullerene weight ratio was in the vicinity of 1:4, corresponding to an increased fullerene mole fraction relative to reP3HT blends. ^{Reference Hoppe and Sariciftci50,Reference Shrotriya, Li, Yao, Moriarty, Emery and Yang51} Thus, if the electron current is largely limited by the microstructural state of the fullerene network, it may be hypothesized that the PCBM mole fraction leading to a near-optimal charge transport network within a conjugated polymer matrix depends on the polymer crystallite concentration. A detailed study of the microstructure evolution as a function of small molecule mole fraction is beyond the scope of this work.

Interestingly, the GISAXS curve for raP3HT:Spiro-MeOTAD blends shows a shoulder-like feature at Q _xy ∼0.09 Å⁻¹, corresponding to an approximate length scale of 7 nm. This is similar to PCBM in reP3HT films. Thus, it appears that Spiro-MeOTAD—a nonspherical small molecule with 3D frontier molecular orbital extent—preferentially results in relatively well-defined length scale of polymer/molecule phase separation in the (thermally annealed) amorphous polymer host. This behavior is diametrically opposed to what we observed in P3HT blends with PCBM—the quintessential quasispherical small molecule. However, in both cases the small molecule gave rise to no obvious Bragg reflections in either polymer host; this differs qualitatively in raP3HT blends for PDI with its 2D molecular orbital extent.

To gain a better understanding of the difference in small molecule behavior as a function of polymer chain microstructure, we have used GIXD and GISAXS to interrogate blends of the three small molecules with a different amorphous polymer host. For this purpose, we have used poly(isobutylmethacrylate) (PIB), which has a molecular weight that is comparable to P3HT but a chemical structure that differs qualitatively from P3HT. The chemical structure of PIB is shown as an inset in Fig. 5, which displays GISAXS curves for both as-cast and thermally annealed PIB:molecule blends. We note that, although both polymers are amorphous, we expect that the difference in monomer chemical structure of PIB and raP3HT would lead to different inter-monomer (and possibly monomer-small molecule) interactions for the two polymers. However, a systematic investigation of blend microstructure as a function of polymer interaction parameters is beyond the scope of this work.

FIG. 5.

(a) GISAXS curves for as-cast and thermally annealed blends of poly(isobutylmethacrylate) (PIB) with different small molecules. The chemical structure of PIB is shown as an inset. Upon annealing, only the PIB:Spiro-MeOTAD shows the appearance of a well-defined shoulder. A similar feature is evident in as-cast PIB:PCBM blends but largely disappears upon annealing. The lower panels show GIXD patterns for annealed blends of PIB with (b) PCBM, (c) PDI, (d) Spiro-MeOTAD. The images show that, similar to raP3HT blends, PDI forms highly oriented crystallites within the polymer matrix.

Figures 5(a) and 5(c) show that, similar to its behavior in raP3HT blends, PDI forms oriented crystallites and gives rise to a largely featureless scattering curves in both its as-cast and annealed states over the Q range accessible with GISAXS. The situation is entirely different for both PCBM and Spiro-MeOTAD. In PIB:PCBM films, the GISAXS curve for the as-cast film shows a shoulder-like feature at Q ∼0.02 Å⁻¹, corresponding to a characteristic phase segregation length scale of roughly 30 nm. This feature partially disappears upon thermal annealing. In PIB:Spiro-MeOTAD blends, we observe precisely the oppositely behavior. Namely, the as-cast film gives rise to a featureless GISAXS profile, but upon thermal annealing a shoulder at Q ∼0.01 Å⁻¹ (d ∼60 nm) becomes prominent. Thus, both molecules with little propensity for crystallization within the polymer matrix lead to phase separation on a length scale of the same order as common exciton diffusion lengths. However, for one molecule this corresponds to a kinetically frozen state (PCBM), whereas for the other (Spiro-MeOTAD), this state is located substantially closer in free energy to thermodynamic equilibrium.

The differences primarily between the morphology of polymer blends with PCBM and Spiro-OMeTAD—the two small molecules with a 3D molecular orbital extent—are intriguing yet puzzling. In particular, although Spiro-MeOTAD has demonstrated the ability to phase-segregate on the 10 s of nanometers length scale upon thermal annealing, GISAXS data implies that it does so only in amorphous polymer matrices. In an attempt to better understand these observations and to improve the comparison across the small molecule series, we have performed resonant elastic X-ray scattering (REXS) measurements in the vicinity of the carbon K-edge. ^{Reference Fink, Schierle, Weschke and Geck52–Reference Vettier54} At these energies, this technique is sometimes referred to as resonant soft X-ray scattering. ^{Reference Liu, Brady and Wang55–Reference Collins, Cochran, Yan, Gann, Hub, Fink, Wang, Schuettfort, McNeill, Chabinyc and Ade58} Because atomic scattering factors become complex in the vicinity of an absorption edge, the scattering contrast can be tuned (and enhanced) by tuning the X-ray energy across the transition threshold. This technique not only allows us to extend our scattering measurements to lower Q and thus larger length scales, but we can also compare scattering off- and on-resonance (in the near-edge 1 s → π* transition region) to potentially gain additional insight into the nature of the scattering inhomogeneities due to energy-dependent scattering contrast differences. X-ray absorption spectra of the probed molecules is shown in Fig. S6.

Figure 6 shows REXS curves of reP3HT (a, c) and raP3HT (b, d) blend films. reP3HT:PCBM films both off- and on-resonance look qualitatively similar: The scattered intensity appears to plateau at Q ∼0.01 Å⁻¹, which is very consistent with the position of the well-defined shoulder observed with GISAXS in the same blends. Further support for this is shown in Fig. S5 of the Supplementary Material, which shows that the on-resonance REXS curve for this blend is qualitatively similar to GISAXS data. For reP3HT:PDI blends, the off-resonant scattering intensity I is well represented by power law scaling over the measured Q range, that is, $I \propto {Q^p}$ , where p is negative. On resonance, however, the reP3HT:PDI curve displays a relatively well-defined shoulder roughly corresponding to a 90 nm length scale. This implies that the contribution of the so-called pure resonant scattering function to the total intensity becomes substantial, thus modifying the effective contrast responsible for low-Q scattering. This must be due to reasonably strong contrast between the π-electron density (when probing near the absorption edge) and the surrounding matrix. We may expect this to be the case for highly segregated PDI clusters.

FIG. 6.

REXS curves collected in the vicinity of the carbon K-edge for reP3HT (a, c) and raP3HT (b, d) blends with the three small molecules. The top panels (a, b) correspond to scattering of 250 eV photons, which is well below the carbon K-edge absorption onset. The bottom panels (c, d) correspond to 284.4 eV photons, which is just above the absorption onset. Higher energies did not give rise to curves that differed qualitatively from those collected at 284.4 eV. reP3HT:Spiro-MeOTAD and reP3HT:PDI show signatures of clustering on the 100 nm scale, which is too large for efficient charge generation. All raP3HT blends give rise to relatively featureless curves, though with power law exponents (over certain Q windows) that differ significantly among the three small molecules. The fitted power law exponents are shown next to the curves.

Interestingly, off-resonant REXS curves of reP3HT:Spiro-MeOTAD films contain a shoulder at a fairly low Q of ∼0.003 Å⁻¹, which corresponds to approximately 210 nm. The location of this feature moves slightly to lower Q when the X-ray energy is tuned on resonance while at the same time becoming more pronounced. Thus, together with GISAXS data, our results show that reP3HT:Spiro-MeOTAD blends are devoid of well-defined morphological features on the 10 nm scale but contain large scattering inhomogeneities well in excess of expected exciton diffusion lengths. We note that at larger Q, the intensity is well represented by a power law; in the figure we have indicated the values of p obtained from fitting the intensity to such a model.

REXS curves for raP3HT blends with all three small molecules display significant differences in slope among the different molecules but no obvious shoulder-like features commonly associated with relatively well-defined length scales of phase separation. Such behavior may result from fractal-like morphologies, though we are unable to verify this at the moment. However, as mentioned above, raP3HT:Spiro-MeOTAD does show evidence for clustering on the 10 nm scale, which stands in contrast to both PCBM and PDI in raP3HT.