2.1. Diameter-Dependent Light Harvesting
Figure 1a schematically presents the preparation of nanoparticle containing P3HT nanofibers by electrospinning P3HT/PEO solutions. The digital photograph shows a dark purple mat of P3HT/PEO nanofibers, also in Fig. 1b, presenting their strong light harvesting ability and free-standing nature at this stage. The as-spun fibers are subsequently heated in isopropanol (IPA) to remove the insulating PEO. The resultant P3HT fibers in Fig. 1c-f, are highly textured, continuous for several microns, and fused at the junctions between fibers. After PEO removal, the lowest achieved dimensions of these fibers were found to be 55 nm in height and 95 nm wide.
Thinner fibers should be capable of delivering more effective exciton dissociation after [6, 6]-phenyl-C61-butyric-acid-methyl-ester (PCBM) infiltration, as excitons will be produced closer to the surface. Here, our fibers provide exciton diffusion pathways of ≤ 28 nm, even before interdiffusion is considered.
Due to strong light scattering by the nanoweb, optical absorptance was determined by measuring the total reflectance and transmittance using an integrating sphere. Combination of these measurements provides a near-zero absorptance beyond 700 nm, indicating successful consideration of diffuse scattering (see Figure S1).
In Fig. 1g, SEM studies in conjunction with statistical analyses revealed a time-dependent reduction in the diameters of P3HT/PEO fibers upon PEO-reduced AgNP introduction. The as-spun diameters drop from 240 nm to 100 nm, as the spinning solution was aged for up to 96 h prior to electrospinning. We later investigate the mechanism behind this reduction in depth using PEO nanofibers. Presented in Fig. 1h and S2, after the nanoparticle introduction, the absorptance of the nanofiber-based films increases with the solution aging time before spinning. The absorptance peaks after 72–96 h, which correlates well with the minimum diameter of the spun fibers. Fibers electrospun immediately after nanoparticle addition to the spinning solution (0 h aging), offer a maximum light harvesting at 560 nm of 45%. The same mass of thinner fibers spun after 72–96 h, delivers 83% absorptance at this wavelength. Beyond 96 h, we observe the onset of fiber beading, which coincides with the observed drop in absorption to 71% at 168 h.
The first reason for this time-dependent light harvesting improvement can be attributed to the PEO-reduced AgNPs forming in solution over time, contributing to LSPR enhancements. However, we also identify that the degree of electrospinning-induced polymer alignment is linked to the fiber diameter, which decreases with aging time of the nanoparticle-containing solution. The extent of polymer chain alignment was determined for fiber samples of varying diameters - pristine P3HT nanofibers, and PEO-reduced AgNP-containing P3HT nanofibers electrospun after the spinning solution was aged for 24 and 96 h. Prior to PEO removal, these fibers possess diameters of 230 ± 56, 162 ± 41, and 99 ± 27 nm, respectively. The fibers in each sample were orientated along a common axis using a rotating drum collector (see Figure S1e).
In Fig. 1i-j, the absorption by the samples was greater when the incident light was polarized along (parallel) versus polarized across (perpendicular) to the fiber axes. This anisotropy indicates the chains are generally orientated along the axis of the nanofiber. This is because excitation of the polymer is dependent upon the orientation of the transition dipole moment with respect to the polarization of light, with a higher probability of absorption when the polymer backbone and the electric field of light are similarly orientated [18]. In Fig. 1i, aligned pristine P3HT nanofibers possess a 0–1 transition (560 nm) absorptance (A) ratio with parallel (∥) and perpendicularly (⊥) polarized light of A∥/A⊥ = 0.88. The nanoparticle-containing fibers spun after 24 h of aging, with an intermediate diameter, gave a A∥/A⊥ = 0.79 (see Figure S1c). Meanwhile, in Fig. 1j, the thinner PEO-reduced AgNP-containing P3HT nanofibers spun after 96 h, showed greater polarization dependence, with a ratio of 0.62. In Figure S1a, a randomly orientated fiber mat spun onto a static collector, showed no dependence with polarized light (A∥/A⊥ = 0.99).
These observations cannot be explained by differences in the extent of fiber orientation in a common direction by the rotating collector, as we would expect the thicker fibers to be shorter for a given volume, and so better aligned under the same drum rotation speed. Instead, we reveal that polymer chain alignment along the fiber axis improves with a reduction in the fiber diameter. This makes sense as fibers with a smaller diameter will have undergone a greater degree of uniaxial elongation, and the polymer chains are similarly stretched out in this direction, as communicated by Fig. 1k.
Along-fiber alignment leads to a greater degree of polymer orientated parallel to the substrate upon fiber collection. This in-plane configuration delivers greater absorption, as more polymer is orientated so that the π-π* transition dipole moment (along the backbone) has an effective overlap with the electric-field vector of the incident light, approximately normal to the substrate [14, 15, 20] This, along with an increasing plasmonic intensity, and longer light pathlengths due to scattering by the larger fiber surface area, results in greater light harvesting by fibers spun from aged solutions. Further optical analysis is conducted to deconvolute plasmonic and alignment-associated absorption enhancements as described later.
2.2 Reducing Achievable Nanofiber Diameters
Although polymer chain entanglement is essential for effective electrospinning, viscosity provides resistance to the elongation forces acting on the forming fiber, reducing the stretching and thinning of the jet. Therefore, one must use the lowest possible polymer concentration when attempting to produce the thinnest achievable nanofibers. However, when the concentration is too low, surface tension forces will dominate, leading to beading to minimize the surface energy, or the jet breaks into droplets, i.e., electrospraying. From a chloroform solution, the concentration of 900 kDa PEO must be greater than 0.50 wt% to avoid beading, yielding approximately 500 nm fibers. In the presence of a polar co-solvent, the beading onset concentration is reduced to 0.35 wt% [19]. Kim et al. suggested 100 nm fibers are spinnable with 9.4 wt% DMF and 3.9 wt% acetic acid. However, in Fig. 2a, the minimum PEO fiber diameter spun within this work in the absence of nanoparticles was 140 nm, using 8 wt% DMF. Further additions of DMF, or acetic acid, were found to lead to beading (see Figure S3).
Stable additions of up to 10 wt% AgNO3 wrt the PEO concentration could be made to 0.35 wt% PEO solutions in chloroform: DMF (92: 8 w/w) before aggregation. The DMF addition was used to solubilize the AgNO3. After reduction, and the formation of PEO-reduced AgNPs, we obtain a nanoparticle-containing spinning solution via a ‘one-pot’ synthesis. Upon electrospinning, silver addition was demonstrated to reduce the diameter of PEO nanofibers by over 40%, with our champion nanoweb possessing a diameter of 75 ± 17 nm, as shown in Fig. 2b and c. To the best of our knowledge, this presents the lowest diameter of PEO nanofibers spun from a chloroform-based solution. Diameter reductions were found to occur even with very low nanoparticle additions, although the reduction slowed above 0.1 wt% AgNO3 wrt PEO.
In Figure S4, the incorporation of nanoparticles was not found to induce any change in the surface tension of the spinning solution. However, the electrical conductivity of the solution in Fig. 2c was found to rise linearly with AgNO3 addition, up to 10 µS cm-1. The contribution by the nanoparticles towards greater conductivity has not been elucidated, however, we expect it can be largely ascribed to residual Ag+ and NO3- ions as they possess higher mobility and charge density. Oleylamine-capped nanoparticles (OA-capped NPs) produced a significantly smaller enhancement in electrical conductivity due to the lower concentration of residual ions, the larger nanoparticle size and hence lower mobility, and the insulating capping layer (see Figure S5). However, there are similar diameter reductions in Fig. 3c produced by these particles, suggesting that the conductivity enhancement is not the sole or primary reason for the reduction.
Figure 2d also reveals the addition of nanoparticles also has a considerable influence on the viscosity of a 1.5 wt% PEO solution, which correlates well with the decreasing diameter. We initially considered whether this was due to PEO chains adsorbing to the nanoparticles, reducing the proportion of ‘free’ PEO contributing towards the viscosity. However, OA-capped NPs also delivered a viscosity reduction despite their dense capping layer, (see Figure S6) and upon isolating the nanoparticles from the spinning solution after 48 h by centrifugation, we found no evidence of PEO capping or displacement of the oleylamine ligands, as investigated by Fourier Transform Infrared Spectroscopy (FTIR) in Figure S7. Alternatively, there is evidence that sufficiently small gold nanoparticles can increase the reptation tube diameters of PEO melts, leading to bulk rheological changes [39].
Lower viscosity reduces the resistance to fiber stretching, whilst the higher charge density enhances the elongation forces from the applied electric field, [29, 36] allowing for the production of thinner nanofibers. Further, without nanoparticles, the tapered linear section of the electrospinning jet spans > 5 cm of the 20 cm needle-collector separation. After nanoparticle addition, the whipping onset could be observed, by eye, to occur closer to the needle, within 3 cm, seen in Fig. 2e and S8. A majority of fiber stretching occurs after this first instability, therefore nanoparticles can provide a longer time for elongation before the solvent evaporates, or the fibers are collected at the ground plate.
The fiber uniformity was also improved with fewer beads observed. Whilst in the absence of nanoparticles, a PEO solution at this reduced viscosity would produce highly beaded fibers, the nanoparticles appear to impede beading [36]. We theorize that this may be attributed to a higher density of repulsive charges from residual ions and charges generated on the nanoparticles when the voltage bias is applied. As conveyed in Fig. 2f, this would favor the production of uniform fiber morphologies, which possess a higher surface area to volume ratio to minimize repulsion between charged species during fiber flight.
This reduction in viscosity is in fact a dynamic process, with significant reductions measured beyond 24 h. In Fig. 3a, we present a continuous reduction in viscosity as the spinning solution ages. The spinning solution viscosity can reduce by > 90%, producing thinner fibers as the solution ages, and beyond a critical aging time, the solution does not possess the necessary entanglement for uniform fiber formation, and beading begins to occur, as demonstrated in Fig. 3b. Later still, the solution is electrosprayed. The aging time required for this onset (typically 24–48 h) is dependent upon the type and loading of the nanoparticle, and initial viscosity of the solution. Until 24 h, the viscosity of the PEO-reduced AgNP-containing solution was stable due to the time taken for the reduction of the silver salt. With separately synthesized OA-capped NPs, the viscosity has already decreased significantly by 24 h. Interestingly, we found OA-capped AuNP addition provided a lesser reduction, which plateaued beyond 24 h.
Electrical conductivity was also found to increase over time, aiding the production of thinner fibers. Figure 3c shows a 0.8 µS cm− 1 increase in the PEO-reduced AgNP-containing solution corresponding to a 27% change, while the relative increases in OA-capped NP solutions were 50% and 120% for AgNP and AuNP analogues, or 0.2 and 0.6 µS cm− 1. Conductivity likely increases due to the reducing viscosity, which allows for higher mobility of charged ions and nanoparticles in solution. However, it should be noted the pristine PEO solution also increased from 0.2 to 0.4 µS cm− 1 over 96 h, perhaps due to improved solvation, but fiber diameters were stable over this period. Furthermore, if conductivity were the major factor determining the fiber diameter, we would expect thinner fibers to be produced immediately after nanoparticle addition (hour 0), as the electrical conductivity has already significantly increased.
We conclude that the mechanism for diameter reduction in PEO-based nanofibers upon nanoparticle addition, is primarily due to a reduction in viscosity, and enhanced electrical conductivity may act as a secondary factor, although perhaps necessary to maintain uniform fiber morphology.
2.3 Towards Ideal P3HT Nanofibers
As PEO dominates the electrospinning behavior of P3HT/PEO fibers, we find analogous diameter reductions upon P3HT addition. As previously described, in Fig. 1g and 3d-l, the reduction in diameter is easily observed. At additions of 10 wt% AgNO3 wrt PEO, sub-100 nm nanofibers could be spun from a 0.35 wt% PEO/ 0.25 wt% P3HT spinning solution in chloroform: DMF (95: 5 w/w), when aged for 96 h. It was necessary to limit DMF addition to 5 wt% to prevent beading and precipitation of P3HT, yielding a lowest spinnable diameter of P3HT/PEO without nanoparticle addition of 220 nm.
The mechanism for diameter reduction is independent of DMF addition, yet it is still required to solubilize AgNO3. In contrast, when using separately synthesized OA-capped Ag and AuNPs, which have previously shown enhanced exciton generation in plasmonic OPV devices, [39] we deliver a similar and significant diameter reduction in the absence of any antisolvent addition, as in Fig. 3d.
Due to P3HT addition and therefore higher initial viscosity, the optimal aging time to produce the thinnest uniform fibers was extended until 96 h. Lowering the P3HT content reduced diameters further, and the optimal aging time decreased accordingly. At 0.15 wt%, 85 nm fibers were spun at 72 h, and at < 0.10 wt% P3HT, 80 nm fibers were produced after 48 h (see Figure S9).
Brunauer-Emmett-Teller (BET) analysis revealed that the specific surface area of the as-spun P3HT/PEO nanofibers doubled upon 10 wt% AgNO3 addition and 96 h aging, from 5.55 ± 0.08 m2/g to 11.79 ± 0.13 m2/g, therefore indicating the potential for increased heterojunction surface area, and thus, exciton dissociation, upon backfilling the nanoweb with an electron donor (see Figure S10).
Due to the insulating nature of PEO, if it is not effectively removed by selective dissolution in hot IPA, it would significantly hamper exciton and charge carrier transport within the sample. Akin to previous works, [16, 19] a Thermogravimetric Analysis (TGA) isotherm at 370°C under nitrogen was employed to determine that ~ 1.5% of PEO remains from the washing procedure (see Figure S11). At this temperature, PEO is almost completely degraded, however, P3HT has negligible weight loss. In the X-ray Diffraction (XRD) trace in Fig. 5a, the presence of PEO in the as-spun fibers is indicated by the peaks at 19.2° and 23.4° 2θ, indexed as the (120) and (112) planes respectively, [41] which disappear entirely upon washing.
The material cost of the nanoparticles is substantial, therefore if a large proportion is lost upon washing, this would be undesirable. Comparison of TGA residuals at 900°C of washed P3HT and PEO-reduced AgNP-containing P3HT fibers (solution contained 0.35% PEO and 0.25% P3HT) provides an estimate for the Ag content, from which we determine that ~ 80% of the nanoparticle mass was retained (see Figure S12 and Table S1). This analysis was conducted in air to promote further degradation of the polymers than is possible under nitrogen, producing smaller, consistent residuals. As 55% of the fiber by mass (PEO) is removed during washing, 80% retention is considered acceptable.
For facile removal of PEO, the phenomena first reported by Kim et al., in which the polymers phase separate to form a P3HT core and PEO sheath, attributed to a difference in viscosity, [19] is integral. The as-spun fibers herein also show a core-shell structure as in Fig. 4a. This favorable morphology avoids the production of isolated PEO domains which would be difficult to remove by solvent. Energy-dispersive X-ray spectroscopy (EDX) line scans across the P3HT/PEO fibers in Fig. 4b-c, show that nanoparticles do not prevent the migration of PEO towards the fiber surface, as indicated by the higher intensity of oxygen atoms for ~ 20 nm at the exterior of fiber. Meanwhile, the sulfur count, proportional to the P3HT content, is absent for 10 nm, beyond which the sulfur count increases to a maximum towards the fiber core, as expected. EDX scans in Fig. 4d-e of the fiber post washing support the selective dissolution of the auxiliary polymer.
An uninterrupted P3HT core is equally important as it allows fibers to retain continuity as PEO is dissolved. Prior to washing, the fibers were cylindrical, measuring 99 ± 33 nm in width and 98 ± 25 nm in height, as determined by SEM in Fig. 3j and AFM in Fig. 4f-g, respectively. During the washing procedure, the fibers ‘slump’, becoming 116 ± 33 nm wide and 73 ± 25 nm high, presented in Fig. 3l and 4h-i. Lowering the P3HT content from 0.25 wt% to 0.15 wt% produced our smallest, optimized nanofibers as presented in Fig. 1c-f, at 96 ± 27 nm wide and 55 ± 13 nm high. Below 0.15 wt%, fibers became increasingly discontinuous upon PEO removal.
In Fig. 4j-o, Transmission Electron Microscopy (TEM) analysis shows an effective dispersion of nanoparticles along the length of the fiber. This is achieved due to the stability of nanoparticles in the spinning solution, and the surface charge of these particles providing repulsion during fiber formation.
2.4 Photoactive Nanofiber Performance
The lamellar spacing and size of crystallites was extracted from XRD traces to investigate the influence of electrospinning and nanoparticle presence (full details in Table S3). In Fig. 5a, the (100) plane of P3HT manifests as a strong, broad peak at 5.51° 2θ in the diffractogram of as-spun PEO-reduced AgNP-containing P3HT/PEO nanofibers. From this, we can calculate a lamellar P3HT d-spacing of 1.70 nm and a crystallite size of 7 nm (estimated by applying the Scherrer equation). Upon washing, the crystallite size rises to 10 nm, due to the elevated temperature. Annealing at 120°C for 20 minutes leads to a further increase, to 12 nm. However, we see no evidence that the extent of crystallinity, proportional to the peak area, is increased by the post-processing steps. In contrast, the aromatic π-π stacking or (010) peak was only very weakly detected at ~ 24° 2θ.
Equivalent washed P3HT nanofibers without particles (Figure S13a) possess a smaller d-spacing (1.63 nm) and larger crystallite size (12 nm). This suggests that PEO-reduced AgNPs may disrupt the stacking and interdigitation of the alkyl side chains, resulting in larger d-spacing and smaller crystallites. However, OA-capped NP-containing nanofibers in Fig. 5b presented more encouraging results, with greater crystallite sizes (15 nm and 13 nm for the Ag and Au analogues respectively) and the expected d-spacing (1.62 nm). Therefore, we do not find that the presence of nanoparticles systematically affects the nature of the crystallites.
Films spin-coated from a P3HT solution (in chloroform) and the PEO-reduced AgNP-containing spinning solution (Figure S13b), produced similar d-spacings and crystallite sizes to the corresponding fibers. This is despite the greater size confinement and faster solvent evaporation rate during fiber formation over spin-coating. We envisage that polymer alignment prior to crystallization provides a templating effect for crystallite formation which can offset any disruption caused by the presence of the nanoparticles or the electrospinning process.
Comparison of the extent of crystallinity between each sample is not reliable, as the peak intensities will be affected by differences in crystallite orientations between samples. Thin-films will often adopt an ‘edge-on’ conformation, presenting the alkyl sidechains at the film-substrate and film-air interfaces, [42] whilst we anticipate that crystallites may be aligned along our fibers, giving rise to both ‘edge-on’ and ‘face-on’ configurations.
Revisiting the optical properties, we observe a change in the shape of the spectral absorptance upon electrospinning. Akin to previous reports with electrospun or self-assembled nanofibers of P3HT, [19, 43–45] the ratio between the first two vibronic peak intensities (0–0 /0–1 at ~ 615 nm and ~ 560 nm respectively) is enhanced relative to thin films, as in Fig. 5d. This is indicative of increased J-aggregate behavior as described by Spano et al [46, 47]. In thin films, P3HT usually possesses H-aggregate characteristics. The co-facial stacking favors interchain interactions, and excitons are delocalized between neighboring molecules. In this regime, the 0–0 transition is dipole forbidden, only allowed by thermal excitation or defects, and as a result, thin films have a 0–0/0–1 transition absorbance ratio of ~ 0.5–0.8. This aggregate type dominates due to amorphous chains reducing the polymer chain planarity and consequently, conjugation lengths. Electrospinning produces aggregates with greater J-type character, where the 0–0 transition is allowed, and in Fig. 5d, we observe a relative intensity of ~ 0.9 in absorptance between the two transitions. Here, the polymer chain backbones are more planarized with less perturbance from amorphous chains, leading to increased intrachain interactions.
Analysis of the spinning solution reveals that DMF addition, a non-solvent for P3HT, begins to induce the precipitation of aggregates, as the orange, translucent P3HT solution in chloroform, becomes an opaque, dark red/purple. In Figure S15, vibronic peaks, associated with solid phase P3HT, appear alongside the π-π* transition at 451 nm of the well-dissolved polymer. In fact, controlled use of an antisolvent is a known method for the production of self-assembled P3HT nanofibers [44]. Therefore, nanostructures have already begun to form even before electrospinning. Upon spin-coating these solutions, we notice the 0–0/0–1 peak ratio is similar for films cast with or without the non-solvent, retaining dominant H-aggregate behavior. This suggests the self-assembled nanostructures are either H-aggregates and electrospinning is required to introduce J-type features, or that the aggregate density is too high to detect them. Previous studies have shown J-aggregates to appear as H-type aggregates in concentrated solutions or dense films, as the sidechains can attenuate the intrachain order of surrounding aggregates [43].
Additionally, red-shift was seen (0–0: 607 nm → 613 nm, and 0–1: 556 nm → 563 nm) in the absorptance traces in Fig. 5d upon electrospinning compared to thin-films cast by spin-coating. This shift arises as the spinning process elongates and planarizes the polymer backbone, increasing the conjugation length. With films, this is only achieved by energy-intensive thermal annealing processes. Due to the differences between the absorptance trace shapes, light harvesting was determined by integration of the absorptance area between 350 and 800 nm. This red-shift and enhanced 0–0 transition results in a lower energy absorption onset and thus greater harvesting by nanofibers between 550 and 650 nm, in comparison to the equivalent film, shown in Fig. 5g. This, and a higher absorption ‘tail’ below 450 nm, led to a 1.14x improvement in light harvesting of the P3HT fibers over the thin-film, rising from 31.2–36.6% (details in Figure S16 and Table S4).
More significant enhancements were observed from the introduction of nanoparticles in Fig. 5e and h. Incorporated into electrospun P3HT fibers, the addition of PEO-reduced AgNPs and OA-capped AuNPs delivered a strong broadband enhancement across the entire visible light region versus pristine nanofibers, of 1.30x and 1.38x, respectively, however no evident enhancement was observed from OA-capped AgNPs (1.04x). Compared to the pristine P3HT films, the PEO-reduced AgNP and OA-capped AuNP-containing P3HT nanofibers provide an enhancement of 1.48x and 1.58x, respectively.
Diffuse reflectance spectroscopy of nanoparticle-containing PEO nanofibers revealed that the plasmonic absorbance for each nanoparticle type was retained upon electrospinning, indicating the potential to harness the LSPR effect. However, OA-capped NPs are more intensely plasmonic than PEO-reduced AgNPs (see Figure S17c). It was therefore initially surprising to see PEO-reduced AgNPs deliver strong enhancement, given that we see no influence with OA-capped AgNPs. Due to their size, the nanoparticles do not scatter efficiently, [48] whilst, as the smallest, PEO-reduced AgNPs scatter the least. Therefore, we attempt to explain these observations with the longer wavelength, broad plasmon resonances of PEO-reduced AgNPs and OA-capped AuNPs, in comparison to OA-capped AgNPs. This provides a greater spectral overlap with the donor polymer, resulting in a stronger LSPR effect and superior absorption enhancement [49].
The peak position of plasmonic absorption is strongly dependent upon the refractive environment as shown by Finite-Difference Time-Domain (FDTD) simulations, used to solve the absorption and scattering spectra of the nanoparticles within the P3HT matrix (refractive index approximated to 1.95 [50, 51]). In Figure S18, the plasmonic absorption of OA-capped AgNPs is expected at 480 nm, presenting a modest spectral overlap with the semiconductor. Meanwhile, the OA-capped AuNPs plasmon resonance is centered at 580 nm, overlapping well with the maximum absorption of the semiconductor. We also estimate that PEO-reduced AgNPs possess a greater spectral overlap with the surface plasmon resonance found at 520 nm within this refractive environment, calculated by blue shifting the OA-capped AgNP simulated spectra by 40 nm as observed when dispersed in chloroform.
Figure 5f and i shows that OA-capped AuNP infusion produced similar but lesser broadband enhancements within thin-films, whilst OA-AgNPs again had no influence. This disparity between the enhancement observed in fibers (1.38x) and in films (1.17x) (Figure S16b-c) allows us to deconvolute the contributions of LSPR and electrospinning. The enhancement seen in thin films is assumed to originate only from plasmonic effects, meaning the difference between the enhancement in fibers and films can be ascribed to the additional electrospinning-induced polymer alignment from the nanoparticle-driven diameter reductions. Therefore, we find that the photon absorption enhancement from the combination of these two approaches – electrospinning photoactive nanofibers and noble metal nanoparticle introduction, is synergistic, providing a greater joint enhancement than the sum of their individual parts.