Performance limits of plasmon-enhanced organic photovoltaics

Due to their strong scattering and large local-field enhancement properties, the use of plasmonic metal nanoparticles has been demonstrated as an efficient means for improving light management in organic photovoltaic devices (OPVs).^1–3 However, while a range of strategies have been presented for incorporating plasmonic particles into OPVs,^4–9 it remains an open question as to what represents the ideal placement of the nanoparticles within the device for optimum charge carrier enhancement and extraction.^10,11 Two main strategies have been pursued in the past, in which plasmonic particles have been either blended with the active layer or embedded inside charge-selecting interfacial layers. Embedding the particles in buffer layers adjacent to the active layer allows one to make use of far-field scattering while avoiding disruptions of the sensitive bulk heterojunction morphology.^5,12 At the same time, this approach may inhibit problems related to exciton quenching and charge recombination.^13,14 Furthermore, placing metal nanoparticles at the interface next to the active layer could be a potentially favorable route for enhanced forward and/or back scattering depending on whether nanoparticles are placed in the front or the back of the device, respectively.^7,15 Conversely, the close proximity of metal nanoparticles to the absorbing chromophores could ensure a more efficient use of the strong optical near-field enhancement effect.³ The latter can achieve several orders of magnitude enhancement in local field intensity but decays quickly with increasing distance between the absorber.¹⁶ As such, blending metal nanoparticles directly into the active layer, potentially using a thin tunnel barrier, can be advantageous with regards to maximizing absorption enhancement and, consequently, the photocurrent in photovoltaic polymer blends.^17,18

To date, most of the metal nanoparticles used to enhance OPV performance have been spherical nanoparticles of gold or silver. Those nanoparticles suffer from relatively high optical losses and rather low electric field (E-field) concentration effects,¹⁹ when compared with more anisotropic structures. Thus, one might imagine that placing anisotropic metal nanoparticles with sharp tips, featuring high scattering to absorption ratios and strong E-field concentration properties around particle asperities,²⁰ in the bulk polymer blend of OPV devices could potentially allow for better exploitation of the near-field scattering mechanism.

Recently, we have shown that plasmon-resonant silver nanoprisms can act as efficient optical nanoantennas capable of significantly enhancing the yield of charge carriers in bulk heterojunction blends.^12,21–23 We therefore chose silver nanoprisms as a model system for studying the effect of the vertical position of metal nanoparticles inside the active layer of an OPV device consisting of Glass/Indium-Tin-Oxide (ITO)/poly(3, 4-ethylenedioxythiophene):poly(4-styrenesulfonate) (PEDOT:PSS)/PIDT-PhanQ:PC₇₁BM/Ag. PIDT-PhanQ is a phenanthroquinoxaline-based low-bandgap polymer (E_g = 1.67 eV) with reported power conversion efficiencies above 6%.²⁴ We show that controlling the nanoparticle position plays a critical role for engineering plasmonic OPV devices featuring both high internal quantum efficiencies and large absorption coefficients at sub-100 nm active layer thicknesses with these materials.

We first measured the external quantum efficiencies (EQE) and extracted the short-circuit photocurrents (J_sc) for OPV devices without metal nanoprisms (as reference) while varying the active layer thickness (Figure S4(a)).²⁵ We then extracted the internal quantum efficiencies (IQE) as a function of the active layer thickness integrated over all wavelengths by comparing experimental J_sc values from EQE measurements with modeled J_sc values determined by computing absorption using finite-difference time-domain (FDTD) simulations (Figure 2(b)).^26,27 We further compared the experimental J_sc values of reference devices and the simulated J_sc values of devices containing metal nanoparticles at different relative positions inside the heterojunction blend using the FDTD formalism. Under these conditions, we find an optimum particle position close to the anode, based on which we are able to predict a maximum enhancement in J_sc of ∼31% for comparable active layer thicknesses and ∼13% compared to the overall best performing thickness without nanoparticles by additionally fine tuning the particle density.

Figure 1(a) schematically depicts the underlying device structure used in the experiment and simulations. In the simulations, silver nanoprisms (AgNPs) were incorporated into the active layer at different vertical positions. We extracted the size and shape of the AgNP from a transmission electron microscopy (TEM) image (Figure S2)²⁵ of a batch of colloidal AgNPs prepared in our lab by the photo-induced conversion of nanoparticle seeds using methods described previously.²⁸ We chose the particle dimensions such that the particle plasmon resonance would enhance the near infrared region of the absorption spectrum once overcoated with the polymer (Figure S3(a)).²⁵ The parameters extracted from the TEM micrograph are r = 7.5 nm, a = 20.8 nm, and a thickness of 12 nm (r: radius of curvature, a: edge length). In particular, we positioned silver nanoprisms randomly in a horizontal plane of the active layer. This should result in a more realistic assessment of the plasmonic enhancement effect in an OPV device compared to simulations where the placement of metal nanoparticles are assumed to be periodic. All FDTD simulations were performed using Lumerical FDTD Solutions 8.7.4.²⁹ Importantly, we excluded the absorption within the metal nanoparticles from generating photocurrent using a refractive index filter. We take this conservative approach to allow us to better account for parasitic losses in the metal nanoparticles. The absorption of the metal is included in the simulation, but is assumed not to generate photocurrent. This procedure and other simulation details are described in detail in the supplementary information.²⁵ 

Based on the EQE measurements and the simulated absorption of the active layer (Figures 2(a), 2(b), and S4),²⁵ we extracted the integrated IQE as a function of the layer thickness (Figure 2(c)). Experimentally, the IQE decreases with increasing layer thickness. Because of the interpenetrating nature of the bulk heterojunction morphology, and the finite carrier mobility, the recombination losses tend to increase with increasing active layer thickness.^26,30 As a result, the fill factor and J_sc tend to decrease with increasing film thickness in many OPVs. For simplicity, we focus primarily on J_sc here,³¹ noting that the total power conversion enhancements (when comparing thin, plasmonically enhanced devices relative to optically thick films) are, thus, more conservative estimates.

In order to overcome noise caused by sample-to-sample variation, we fit a linear curve to the integrated IQE vs. film thickness values as shown in Figure 2(c). We then used these wavelength-integrated IQE values to calculate J_sc for devices with AgNPs positioned in different locations. We note that using a single integrated IQE rather than a wavelength-dependent IQE is an approximation, but one that is justified by the relatively flat IQE vs. wavelength spectra that have been reported for many recent devices, including those using our model PIDT-PhanQ/fullerene blend.^32,33

First, we computed J_sc for simulated devices with the nanoprisms positioned in 3 vertical planes: (1) directly on top of PEDOT:PSS, (2) at the vertical center of the active layer, and (3) on the silver back contact. Figure 3(a) shows J_sc for devices without AgNPs (experimental) and J_sc for devices with AgNPs (simulation) at the different vertical positions corrected by their respective integrated IQEs. To verify that the J_sc enhancement is not affected by our choice of fitting, we also carried out calculations of J_sc using a thickness independent IQE derived from an average of all the experimental device thicknesses (Figure S5).²⁵ The initial particle density (2 × 10¹⁰ cm⁻²) for the simulated devices was set such that there would be no near-field coupling between the particles (average center-to-center particle distance is (2×10¹⁰ cm⁻²)^−½= 70 nm while the edge length of the AgNPs is ∼21 nm).

Both experimental and simulated short-circuit currents shown in Figure 3(a) show a typical thin-film interference pattern as a function of thickness with two maxima below 250 nm. As is commonly reported for OPVs, the first interference maximum occurs at ∼70 nm, whereas the second, broader maximum occurs at ∼250 nm film thickness.^34–36 

Figure 3(a) shows that J_sc for the present device structure can indeed be significantly enhanced for thicknesses around the first interference maximum by placing the particles in the center of the active layer, while placing the particles on the front interface or the rear interface leads to less enhancement. As expected, around the second interference maximum, the polymer film is thick enough that absorption is very efficient across all wavelengths, and nearly no enhancement can be achieved with the contribution of silver nanoprisms using the particle positions shown in Figure 3(a). Indeed, losses due to parasitic absorption in the plasmonic particles begin to decrease device performance for thick polymer films.

Since we observed the highest overall performance near the first interference maximum, we chose to study the device performance as a function of the z-position of AgNPs in the active layer with more detail at this thickness (77 nm). In Figure 3(b), we calculate J_sc as the AgNP layer is moved inside the active layer from the PEDOT interface towards the silver electrode. We found an optimum position of 13 nm from the PEDOT interface, which gives rise to an enhancement of 31% compared to the reference device with the same active layer thickness. We hypothesize that the shape of J_sc vs. AgNP position in Figure 3(b) is determined by complex optical interference of near-field enhancement as well as reflected and scattered light inside the device.

An additional advantage of using colloidal metal nanoparticles as optical antennas is that the absorption enhancement can be tuned by varying the particle density. In practice, the benefit of additional absorption enhancement needs to be carefully balanced against potential losses due to charge recombination, exciton quenching effects at the particle surface, and reduced active material volume.

In Figure 4(a), we explore the effect of the particle density on J_sc for the best performing plasmonic device with 77 nm thick active layer thickness. With increasing particle density J_sc increases, reaches a maximum and decreases again. We suspect that this result is achieved through a combination of effects. First, at low density, increasing the particle density provides for more absorption enhancement. However, as the particle density increases, interparticle interactions lead to a red-shift in the plasmon resonance peak, shifting some of the field enhancement below the band edge of the polymer (Figure 4(b)). In addition, as the metal particle density increases, there is a decrease in the amount of active semiconductor material and the ratio of backward scattering (reflection out of the device) to forward scattering (into the device) become less favorable (Figure S3(b)).²⁵ At the maximum of the plot in Figure 4(a), J_sc reaches a value of 14.06 mA/cm² for a density of 5 × 10¹⁰ cm⁻² (2.95% v/v), which represents an enhancement of 31% relative to the 77-nm-thick device with no nanoparticles, and still 13% compared to the overall best “optically thick” reference device with a 225 nm active layer thickness. Interestingly, this optimum density is consistent with previous reports,¹² though, based on our discussion, we would expect the exact optimum density to depend on particle size, shape, and the optical properties of the host matrix.

Based on the conditions we examined, we calculated the best plasmon-enhanced device performance for sub-100 nm bulk heterojunction films in the case of a 77 nm thick active layer with AgNPs positioned 13 nm away from the PEDOT layer and a particle density of ∼5 × 10¹⁰ cm⁻². In order to understand whether this maximum was caused by the particular randomization of positions that we chose initially, we tried 10 additional randomizations of the particle positions and orientations. We found an average J_sc of 14.06 ± 0.06 mA/cm² (Figure S7)²⁵ suggesting that our results are general and independent of the random seeding of the spatial distribution of our nanoprisms. We also simulated the same device using similar size Ag nanospheres (Figure S6)²⁵ and AgNPs that are fully coated with a 1 nm thick SiO₂ charge barrier. In the former case, J_sc is reduced to 12.22 mA/cm² while in the latter case J_sc is reduced to 12.63 mA/cm². These values represent enhancements of ∼14% and ∼18%, respectively, compared to a corresponding reference device with the same thickness without plasmonic particles. Assuming that V_oc = 0.87 V and Fill Factor = 0.64²⁴ are not changing with the incorporation of nanoparticles,¹⁵ the power conversion efficiency increases to 7.83% from 5.95% for a 77 nm thick device under optimum conditions.

Precise control of the position of a film of silver nanoprisms within the active layer can significantly improve plasmonic device performance of OPVs. When placed at a distance of ∼15 nm from the PEDOT anode with a density of ∼3% (v/v) the short-circuit current density of a low band-gap material with properties typical of today's OPV materials can be enhanced by over 30% compared to the reference device with the same active layer thickness and, perhaps surprisingly, by 13% even when compared to the best device thickness (225 nm). These simulated enhancements are consistent with, if somewhat larger than, those that have been achieved experimentally to date.^17,37–40 These results suggest that while current experiments are realizing a large fraction of the performance gains that are theoretically achievable, there is still some room for improvement using plasmonic enhancement schemes with the current generation of OPV materials.

This paper is based in part on work supported by the Office of Naval Research (Nos. N00014-14-1-0170 and N00014-11-1-0300), the State of Washington through the University of Washington Clean Energy Institute and the Asian Office of Aerospace R&D (No. FA2386-11-1-4072). M.S. acknowledges a fellowship from the Portuguese Foundation for Science and Technology (FCT) (No. SFRH/BPD/71816/2010).