Abstract
A scalable procedure for nano-patterning the bulk heterojunction layer in organic photovoltaic (OPV) devices is reported. Nano-patterning is shown to increase light absorption in poly(3-hexylthiophene) : [6,6]-phenyl-C61-butyric acid methyl ester (P3HT : PCBM) devices (ITO\WO3\P3HT : PCBM\Ca\Al). Nano-patterning also modifies electric fields in OPV devices, thus affecting charge harvesting. Nano-patterned OPV devices with a power conversion efficiency of 4% are presented. Comparable efficiencies are also obtained by optimization of thicknesses in a flat-layer device. Trade-offs between absorption enhancement and charge harvesting deterioration induced by nano-patterning are discussed as well as possible optimization strategies.
Export citation and abstract BibTeX RIS
1. Introduction
Over the past few years, organic photovoltaics (OPV) have been gaining momentum with the steady arrival of new organic materials that have pushed the boundary of OPV power conversion efficiencies (PCE) above 8% [1, 2]. Because of the ease of manufacturing OPV with scalable roll-to-roll processes, their inherent flexibility, and potential low cost [3, 4], OPV could become an important technique in the energy landscape in the near future. The performance of OPV devices can be increased not only by improved photovoltaic materials, but also by favourable geometric configurations. For instance, the development of the bulk heterojunction (BHJ) morphology was a seminal advancement that remedied, to a large extent, the difficulties arising from the intrinsic short exciton diffusion length in donor polymers (∼10 nm). With that nano phase-separated morphology, the excitons created by photons in the donor polymers are able to reach acceptor molecules (typically fullerene derivatives) thereby efficiently completing the crucial charge separation step that must occur in any photovoltaic device. Concomitantly, the pursuit of the ideal nanostructure for perfect exciton separation [5, 6] and carrier collection [7] has also been the subject of multiple studies.
In addition to charge separation itself, transport of those charges remains a limitation of OPV preventing such systems from reaching their full potential. This is because the transport of separated charges across the convoluted BHJ morphology limits the thickness of the active layer and the total light absorption in many cases. Photonic structures have been intensely studied as a means to counteract charge transport losses by increases in light absorption [2, 4, 8–17]. Simple improvements that used geometrical optics were suggested early on [18], followed by light-trapping schemes based on periodic patterning by applying diffraction gratings [9] or nano- to micro-scale patterned substrates [9, 14, 15]. More recently, nanoimprint lithography (NIL), a low-cost method to fabricate large areas with submicrometre features, has been experimentally explored to create nanostructures in OPV devices [2, 4, 8, 11, 17, 19]. Others have reported the use of laser interference lithography to pursue light-trapping structures in OPV devices [12, 14].
OPV device nano-patterning can lead to notable performance improvements. Li et al recently reported an increase from 7.59% to 8.79% in PCE of a low band gap material system using dual plasmonic nanostructures and an enhancement to 8.38% was achieved with a simple grating structure [2]. However, most of the previous results were reported by comparing nanostructured devices with non-optimized, low-efficiency, regular, planar cells. For instance, OPV devices based on poly(3-hexylthiophene) : [6,6]-phenyl-C61-butyric acid methyl ester (P3HT : PCBM) were one of the most intensely studied BHJ systems in the last ten years. This P3HT : PCBM device had an average reported PCE of 3.0% through 2010, with over 80 reported PCEs between 3.6% and 4.0% and more than 100 PCEs above 4% [20]. Despite these findings, all previous patterned OPVs were compared with flat-layer devices with efficiencies of only 2–3.5% [5–8, 14, 15, 18]. Very few researchers describe efforts to optimize flat control devices nor do they report results from a statistically significant number of samples in order to account for performance variability. Although the comparisons are validated using a normalized fabrication (e.g. spin coating speed or the same volume) process for both patterned and flat cells, it is very important for the patterned devices to outperform the most efficient flat cell in order to claim success, especially when one takes into account the complexity and additional potential manufacturing cost involved in the patterning process.
Moreover, in addition to the inadequate comparisons, most of the previous work has focused on only light trapping while overlooking changes in charge transport that might result from patterning OPV devices. There is an unjustified assumption that pure light-trapping structures will lead to OPV device enhancement. Only a few reports briefly mentioned the effect of patterning on the electrical properties of the devices [2, 7], but the underlying causes were not studied in depth. Recently, Vervisch et al suggested that there is a trade-off between light trapping and uniformity of electric field distribution, which could be a key problem limiting the ultimate enhancement of patterning OPVs [21].
In this paper, we present a complete investigation of one nanostructure, which shows light-trapping effects. We fabricated planar and nanostructured OPV devices (ITO\WO3\P3HT : PCBM\Ca\Al) in a multi-parameter space, resulting in 18 distinct types of devices and over a hundred individually studied samples. Efficiencies of P3HT : PCBM-based OPVs exceeding 4% are demonstrated for patterned and flat devices. Although patterned cells are more efficient across some parameter sets, we have found that the absolute best efficiency device in the whole group remains a flat cell among these samples. We quantify the significance of pure, light-trapping photonic structures in OPVs. We have used a simplified model to open the discussion of the possible reasons limiting all of the additional trapped light from being able to contribute to more electrical current and an overall better performance of our devices. We also advocate a thorough reconsideration of the effects of nanostructures in OPV efficiency in order to adequately evaluate this enhancement strategy.
2. Experimental
To realize nanostructured OPVs, we chose NIL to pattern a commonly used negative transparent photoresist SU-8 (MicroChemicals) on glass substrates. OPV materials were casted or evaporated on top of these nanostructured substrates instead of being moulded directly. By this approach, we avoided the previously reported drawbacks of direct patterning of active layers, such as degradation, contamination and non-optimal configurations [14]. As SU-8's optical properties are very close to those of glass (see figure S1 (stacks.iop.org/JPhysD/45/024008/mmedia)), the SU-8/glass combination appears as an optically continuous nanostructured substrate. Furthermore, this technique is suitable to other photovoltaic systems without modifying the moulding conditions. The overall fabrication process of the devices is schematically represented step by step in figure 1. To pattern SU-8, we used perfluoropolyether (PFPE, by Liquidia Technologies) moulds, because it can be easily released from the patterned replicas without complex surface functionalization [4], thus allowing for easy, quick and high-quality nano-replication over 6 cm2. After moulding, SU-8 is cured to become a robust layer that can be baked at 200 °C without any degradation. We used this process to replicate a two-dimensional hexagonal nano-scale array of sub-visible wavelength posts. The posts are ∼180 nm wide and ∼180 nm tall with centre to centre spacing of ∼300 nm. On top of SU-8, an indium tin oxide (ITO) film (100 nm) with resistivity of ∼3 × 10−4 Ω cm was deposited by pulsed laser deposition (PLD) at 200 °C under O2 pressure of 20 mTorr. Tungsten trioxide (WO3) was thermally evaporated to form a buffer hole injector [22] and it allows good subsequent spin casting of the active polymer blend. Figure 2 shows the scanning electron microscopy (SEM) image of the patterned substrate prior to spin-coating the blend. To run a complete systematical study, three different thicknesses (12, 22, 32 nm) of WO3 were evaporated onto both patterned and flat devices prior to spin-coating the blend. A 1,2-dichlorobenzene solution of P3HT : PCBM (1 : 0.8 wt ratio with 25 mg ml−1 P3HT, P3HT by Rieke Metals and PCBM by Nano-C) was spin-coated at three different speeds as 800 RPM, 1000 RPM and 1200 RPM for 90 s with an acceleration time of 6 s, resulting in three different thicknesses (for the non-patterned samples) of ∼230 nm, ∼180 nm and ∼140 nm, respectively. These values were measured by spectroscopic ellipsometry from three different spots of the samples and later confirmed by SEM observations. The flat and patterned films' surface roughnesses were characterized by a tapping mode atomic force microscope with root mean square ∼2 nm (see figures S4 and S5 (stacks.iop.org/JPhysD/45/024008/mmedia)). The films were annealed at 140 °C on a hot plate for 10 min. Calcium and aluminum were subsequently evaporated on top of the active layers to form the top electrode. On each sample, six solar cells with an active area of 0.136 cm2 (measured by microscope) and distance of >0.25 cm between each other were fabricated. Figure 3 shows the cross sectional SEM image of one patterned OPV device. All device measurements were carried out under purified nitrogen (<0.1 ppm water and oxygen), except for optical reflection and angular IPCE measurements, which were made in air after all other tests were completed. AM 1.5 1 Sun measurements (with a solar simulator and an AM 1.5 Global Filter from Newport, light intensity was calibrated to be 100 mW cm−2 by a reference solar cell from Newport before and after each measurement) and space charge limited current (SCLC) hole mobility measurements were conducted with a Keithley 2400 source-measure unit. IPCE measurements were conducted under the chopped light from a monochromator (from Newport). The device was connected to a lock-in amplifier to measure the ac voltage signal via a current-to-voltage preamplifier. Normal incidence reflections were measured using an upright microscope, while angular reflection/IPCE was conducted using a homemade system with calibrated Si detectors. IPCEs were also measured to confirm the exact short-circuit current related to the I–V curves from the solar simulator.
Figure 1. Schematic procedure of nanoimprinted OPV device fabrication: 1, spin-coating SU-8 on glass substrate; 2, emboss SU-8 by PFPE mould to form nanostructures; 3, deposit ITO thin film on top of nanoimprinted SU-8 by PLD; 4, evaporate WO3 layer; 5, spin-coat active materials and evaporate Ca/Al as the electrode to complete the device.
Download figure:
Standard image
Figure 2. SEM image of patterned SU-8 coated with ITO/WO3 prior to spin-coating active materials.
Download figure:
Standard image
Figure 3. Cross-sectional SEM image of the nanostructured OPV device. (a) Whole view of the cross-section. (b) Detailed image of device configuration.
Download figure:
Standard image3. Results and discussion
3.1. Device performance
Both patterned and flat devices have an open-circuit voltage (Voc) of ∼0.6 V and fill factor (FF) of ∼0.6, with remarkably little difference between samples, and which are comparable to previous reported values in the literature [6, 8, 22]. We found that given the commonality of all of these metrics, the key factor that determined a device's PCE was the short-circuit current (Jsc) only. After selecting out the devices with obvious fabrication quality issues, we report all of the cells' Jsc and efficiencies in figure 4. Among all the observations, a few trends can be distinguished. In aggregate, we observed that the Jsc of our flat devices decreased with the thickness of the active layer, as did the efficiencies [20]. However, this is not true for the patterned devices. Jsc peaks at 1000 RPM in all the three groups with different thicknesses of WO3. While comparing the average result of each group with different thicknesses of WO3, we observed performance variations with the WO3 thickness. But, there is no general trend for all the groups and the differences are in general small.
Figure 4. All OPV cell's Jsc(a)–(c) (solid squares are flat cells and open squares are patterned cells) and efficiencies (d)–(f) (solid triangles are flat cells and open triangles are patterned cells). Under certain fabrication conditions, the patterned device showed better performance, but no global enhancement.
Download figure:
Standard imageFocusing on the best performers, the best flat cell's PCE is 4.52% and the highest Jsc is 11.74 mA cm−2, while the best patterned cell's PCE is 4.25% and the highest Jsc is 11.30 mA cm−2. In addition, taking an average of all the cells in one group (i.e. the same WO3 and blend spinning conditions), the best PCE of the flat samples is 4.27% and the highest Jsc is 11.22 mA cm−2, while the best average PCE of a similar group of the patterned samples is 3.98% and the highest Jsc is 10.69 mA cm−2. All these figures are well above the average of reported P3HT : PCBM devices [20], especially when compared with poor control samples. Furthermore, our nanostructured devices are overall better in performance than the vast majority of previously reported patterned devices, which were portrayed as successes in overcoming planar counterparts [5–7, 14, 15, 18]. However, as can be seen from this complete data set, no significant enhancement is actually found relative to optimized flat devices. Very significantly, the best efficiencies of the planar and patterned devices arise with different fabrication parameters (e.g. spin coating speed and WO3 thickness). This shows explicitly that holding fabrication parameters constant as the normalizing factor on comparing patterned and flat cells is misleading.
3.2. Angular and polarization characterization
In order to understand why the regular flat device remains the most efficient among all the devices fabricated, we focused first on the optical properties of the devices. As SEM micrographs sample only small areas, in order to judge the quality of the photonic structure and further explore the effects of nanostructures in OPV devices, we studied the devices' angle-dependent reflection and incident photon-to-electron conversion efficiency (IPCE). For non-normal illumination on nanostructured devices, resonant mode splitting occurs, which generates absorption enhancements that show the signature photonic crystal behaviour. Simulations of the photonic devices based on the materials and geometrical parameters (figure 5) predict the presence of quasi-guided modes [23], where strong light confinement should give rise to localized reflection minima. This prediction confirms that the quality and parameters of the experimental devices are close to the idealized structure presented above in figure 1. Figure 6 shows that the experimentally observed reflection minima at the blend absorption tail move towards λ = 600 nm with incident angle changing from 60° to 30° eventually disappearing. These reflection minima results in local, small IPCE enhancements. The local angular IPCE enhancements were found in both polarizations, because the nanostructure geometry is periodic in both lateral dimensions. As expected, this photonic crystal behaviour was found in all patterned devices while it was absent in the flat ones. The demonstration of this unmistakable photonic crystal phenomenon confirms faithful replication of the design, and that the optical quality and the manufacturing defects are not the reasons for the fact that patterned devices fail to outperform the best flat devices.
Figure 5. Experimental and simulated angular reflections of the patterned device with 12 nm WO3 and spun at 800 RPM. Solid lines are experimental measurements. Dashed lines are simulated reflections. Mismatch between simulations and experiments beyond λ > 650 nm is due to the larger uncertainty in the P3HT absorption constants' tail beyond this point [30].
Download figure:
Standard image
Figure 6. Angular reflections and IPCEs in both p and s polarization of the flat and patterned devices with 12 nm WO3 and spun at 800 RPM. (a) and (b) are the patterned devices. (c) and (d) are the flat devices. Local IPCE enhancements are indicated by arrows.
Download figure:
Standard image3.3. Normal incidence reflection and IPCE
Given the WO3 and ITO transparencies, the first-order, lower device reflection should correspond to higher absorption in the active layer of patterned devices. We performed normal incidence reflection measurements, and IPCE characterizations for all the devices. In figure 7, we show both experimental and simulated normal incidence reflections of the device group with 12 nm WO3. The other two groups with 22 and 32 nm WO3 have similar results (see figure S2 (stacks.iop.org/JPhysD/45/024008/mmedia)). All of the patterned devices exhibit relatively low reflection, especially at wavelengths ranging from λ ∼ 450 nm to λ ∼ 550 nm where P3HT : PCBM absorbs strongly. However, the 800 RPM flat devices reached similar levels of low reflection, comparable to all of the patterned ones. When narrowing the comparison to the 800 RPM flat and patterned devices, the flat devices exhibited lower reflection from ∼550 to ∼600 nm. Taking all these observations together showed that the devices from the 1000 and 1200 RPM group did benefit optically from nanostructures, but the optimized flat devices at 800 RPM had similar optical performance. From this perspective, while a nanostructured sample could have a lower reflection than a flat sample under certain control parameters, we could not conclude that the nanostructured devices have significant optical enhancement over all possible flat devices. Photonic geometry optimization is needed to obtain an absolute best optical performance. As will be shown below, this is possible but unfortunately, such photonic optimization is not enough. As can be appreciated in the 1000 RPM group, the patterned devices have the lowest group reflections but this does not translate into additional electrical current gains (see figure 8). This discordance also happened in other patterned and flat groups (see figure S3). The exciton generation profile [24] or other carrier transport issues related to geometrical factors may not be favourable for the charge carriers to be swept out despite good FFs and Vocs in the devices.
Figure 7. Experimental (a)–(c) and simulated (d)–(e) reflections from the glass sides of both flat and patterned devices with 12 nm WO3. Solid lines are flat samples' reflections. Dashed lines are patterned samples' reflections.
Download figure:
Standard image
Figure 8. IPCEs from glass sides of both flat and patterned devices with 12 nm WO3. Solid lines are flat samples' IPCEs. Dashed lines are patterned samples' IPCEs.
Download figure:
Standard image3.4. SCLC hole mobility measurement
We now direct our attention to the electronic integrity of the devices. In the P3HT : PCBM system, the holes are believed to be the limiting carriers [24]. Therefore, we evaluated hole transport both in flat and patterned films using hole-dominated diodes [2]. The current–voltage characteristics of the hole-only diodes have shown a square dependence of current on voltage (see figure 9), from which we can estimate the hole mobility using [2, 25]
where J is the current density, ε0 is the permittivity of free space, ε (= 4) is the dielectric constant of the P3HT : PCBM blend [26], μ0 is the mobility, V is the internal built-in voltage and L is the thickness of the active layer. The flat and patterned devices are found to be characterized by very similar hole mobilities, 6.66 × 10−4 cm2 V−1 s−1 and 8.83 × 10−4 cm2 V−1 s−1, respectively. (For the patterned device, an average thickness was used as an approximation in this simple formula as the patterning intrinsically induces blend thickness differences.) Although mobilities of the holes and electrons are reported to be enhanced by patterning, the reason for that is not clear [2]. The mobility difference between patterned and flat devices indeed favours the former but only slightly. This together with the good FFs and Vocs leads us to believe that device construction, including morphology, is equally good in the flat and patterned devices. So, the reason behind the unfulfilled overall enhancement must be sought elsewhere.
Figure 9. SCLC measurement for hole mobility measurement. Solid lines are fittings to equations to obtain the estimation of hole mobility. The averaged patterned device thickness is estimated from the pattern geometry.
Download figure:
Standard image3.5. Electric field strength in patterned devices and impact on device performance
Numerical simulation of the charge transport in organic BHJ solar cells may shed light on the loss mechanisms in patterned devices and give indications as to why they do not outperform the best flat cells. We use a mathematical model based on an effective medium approximation of the BHJ material [27]; the electron and hole number densities n, p (m−3) and the electric potential ψ (V) in the photoactive material are related by the classical semiconductor equations [28]. The semiconductor equations form a nonlinear system of three partial differential equations involving charge carrier drift in the electric field, −∇ψ (V m−1), as well as diffusion, which yields expressions for the charge carrier fluxes, Jn, Jp (m−2 s−1). Poisson's equation for electrostatics relates the electric potential to the charge density–q(n–p) (C m−3), where q denotes the elementary charge. The source term in this model is given by the net charge carrier generation rate density, which incorporates exciton generation due to light absorption, exciton dissociation into free charge carriers, as well as charge carrier recombination to excitons and exciton decay. The boundary values of n, p and ψ are prescribed; they depend on the bias voltage. A combined electro-photonic simulation, where the exciton generation rate density is computed from an auxiliary optics simulation [23], may be used to evaluate various patterned organic BHJ solar cells. Full details on this approach have been reported elsewhere [29]; here we only present simulation results for two 2D patterned BHJ solar cell devices with ridge-patterned front electrodes. These are chosen as representatives of different light trapping and carrier harvesting characteristics.
In figure 10 we show the computed electric field at maximum power in a cross section of the two devices; this figure also illustrates the device geometries, and in table 1 we summarize the computed performance factors. The two patterned solar cell devices have similar amounts of P3HT : PCBM BHJ material. While the light absorption in the active material is lower in device (b) than in device (a), device (b) apparently has better charge transport properties than device (a), which ultimately results in higher PCE. This indicates that it is not sufficient to focus only on improvements of the optical device properties by employing a photonic crystal structure which enhances the light absorption. These structures may, on the other hand, cause the electrical properties of the device to deteriorate [21]. Both the exciton dissociation and charge carrier collection efficiencies are similar in devices (a) and (b), whereas the main difference between the two devices is in the collection-to-output efficiency (table 1). This efficiency factor is computed by comparing the maximum power (W m−2) with the charge carrier collection rate (m−2 s−1) multiplied by the energy qVmp (J), where Vmp (V) denotes the bias voltage at maximum power. The collection-to-output efficiency is smaller than 1 if charge carriers are collected at the 'wrong' electrode, i.e. electrons at the anode or holes at the cathode. The spurious charge carrier flux will decrease the total current through the electrodes and therefore the output power, compared with the value expected from the charge carrier collection rate.
Figure 10. Logarithm to base 10 of the electric field strength (V m−1) at maximum power in two BHJ solar cell devices with ridge-patterned front electrodes. The devices are closed by a glass layer in the front and by an aluminum layer at the back. Length units are nanometres. The electric field is weaker in the ridge part because of the larger distance between the electrodes. The induced losses may outweigh the gain in light absorption achieved by the photonic crystal structure.
Download figure:
Standard imageTable 1. Computed performance factors for the two patterned solar cell devices (a) and (b).
| Unit | (a) | (b) | |
|---|---|---|---|
| Active material volume | 10−9 m3 m−2 | 101 | 96 |
| Light absorbed in P3HT : PCBM | W m−2 | 283 | 263 |
| Short-circuit current | A m−2 | 78.2 | 93.6 |
| Open-circuit voltage | V | 0.578 | 0.578 |
| Fill factor | — | 0.690 | 0.627 |
| Power conversion efficiency | % | 3.27 | 3.55 |
| Exciton dissociation efficiency at maximum power | % | 82.0 | 81.0 |
| Charge carrier collection efficiency at maximum power | % | 85.3 | 85.1 |
| Collection-to-output efficiency at maximum power | % | 77.6 | 96.2 |
In figure 11 we show the Euclidean norm of the current density j = −q(Jn − Jp) (A m−2) computed from the steady-state charge carrier fluxes at maximum power for both devices. The current density is low near the electrode in the ridge as well as everywhere in the deeper parts of the ridge region; the current through the electrode boundary will therefore be low in that region. Charge carrier pairs generated in these regions will not be swept out from there, which in the case of device (a) is supposed to lead to a larger spurious flux of electrons through the anode, thus explaining the lower performance despite its better optical features. Even in cases where devices are overall enhanced, the percentage of current increase may not be as much as that of absorption [2].
Figure 11. Current density (A m−2) at maximum power in the two patterned solar cell devices (a) and (b). The current density is low in regions of deeper ridge pattern, which deteriorates the charge transport properties of device (a).
Download figure:
Standard image4. Conclusion
We demonstrated a simple, low-cost, scalable scheme for fabricating nanostructured solar cells by NIL. Compared with imprinting the active PV layer, this method appears easier to adapt to most OPV systems. Clear photonic crystal effects were shown and agreed with simulations. We also compared the flat and patterned devices from both optical and electrical perspectives. Highly efficient P3HT : PCBM-based OPV devices were obtained in both flat and patterned devices, although no significant enhancement from the nanostructured devices was noted by us. By using a simplified optoelectrical model we suggested possible reasons for this such as a non-uniform electric field strength which could limit the performance of nanostructured OPVs in some cases, even when a clear light-trapping effect is found. Considering the promising results from emerging low band gap materials and the huge interest in this field, we conclude that future investigations on nanostructuring OPV should not only focus on light trapping but also on electrical management issues. A combined electro-photonic simulation [29] is required to assess the trade-off in light absorption enhancement and charge transport deterioration in patterned solar cell devices. Only a comprehensive design for each OPV material that combines both optics and electronics will result in nanostructured OPVs outperforming flat devices.
Acknowledgments
Support for this work from NSF (Solar:DMR-0934433) is gratefully acknowledged.