Figure a,b,c presents surface scanning electron microscopy (SEM) images of the Thin/NR cells at different stages of fabrication. Densely packed nanorods were obtained over the entire deposition area on bare ITO. The 3D conformal nature of the cell surface can be appreciated from the SEM surface images, where the structure of the array can still be observed both after the blend coating (Figure b), and Ag contacts were applied (Figure c).
Figure 2 SEM/STEM characterization. (a) Electrodeposited ZnO nanorod arrays, (b) arrays coated with a thin P3HT:PCBM highly conformal layer, (c) Ag contact evaporated on top of the P3HT:PCBM layer (Thin/NR cells) with arrows indicating a few spots where shadowing (more ...)
Figure d,e,f,g,h presents SEM and STEM cross-sectional images of the Thin/NR cells. Figure i shows a conventional Thick/NR hybrid cell. It is seen that the nanorods are approximately 800-nm long, being coated by a thin layer of P3HT:PCBM blend (<50 nm as observed from the leading edge of the blend adjacent to the nanorod in Figure g, although the exact value was difficult to elucidate and some gradient could be present from the top to the bottom of the nanorods), and <50 nm Ag. The high conformality of the blend coating is best exemplified by Figure d,e,f,g,h. Approximately 50 nm is well below the mean free path of both electrons and holes in a polymer-fullerene blend; thus the blend morphology most likely does not even have to be completely optimised [29
]. Although the Ag coating on the ZnO nanorods is less uniform than the blend coating, owing to the fact that Ag preferentially deposits on surfaces exposed to the vapour source (see left-hand side of Figure d), the large sample-boat distance in the evaporator (35 cm) ensures a relatively high Ag coverage of the NRs. This is most clearly seen in Figure c, where only some small spots in the sample (see arrows in the figure) are not coated by Ag due to shadowing from adjacent rods), and also in Figure g where Ag can be seen forming a quasi-conformal coating all over the surface of a ZnO rod. The quasi-conformal Ag coating is found to be important for improving charge extraction and contributing to light trapping in the cell, as will be discussed later.
Figure a,b shows the EQE and PV data for the best Thin/NR and Thick/NR cells obtained, respectively. Strikingly, despite the smaller amount of organic blend used in the Thin/NR cell, it has a higher efficiency (1.34%) than the Thick/NR cell (1.07%), while the EQE spectra are very similar for both cells. On average, a 30% higher power conversion efficiency (η
) was obtained for Thin/NR cells, as well as both higher fill factor (FF) and Jsc
than the Thick/NR architecture, as shown in the table in Figure , confirming the superior performance of the quasi-conformal design. The highest efficiency obtained for the Thin/NR cell (1.34%) is comparable to other results for conventional thick cells using nanorods of similar dimensions as ours, with reported efficiencies ranging from 1.02% to 1.59% [30
]. It is worth noting that in the case of the conformal cells, at least three times less volume of blend is used than in non-conformal cells (as estimated from SEM images). Taking this into account, the short-circuit current densities per unit volume of blend obtained are up to three times higher for the Thin/NR cells than for the Thick/NR ones. This requirement for a lower blend volume effectively means lower fabrication costs for hybrid cells implementing the Thin/NR architecture.
Figure 3 EQE, J-V curves, PVD data and transient charge of best cells plus average photovoltaic parameters. (a) EQE of best performing Thin/NR and Thick/NR cells (idealised cell designs in the inset). (b)J-V curves of best performing cells of both architectures (more ...)
The rather low average values of Voc and FF observed are due to the fact that no seed layer was used prior to electrodeposition of the ZnO NRA, which leaves some ITO exposed and in contact with the blend. This does not affect the evaluation of the conformal architecture since the reference thick/NR cells are made using the same type of NRAs; thus, the same effect takes place. Another related factor that may contribute to a lower average Voc in the conformal cell is that silver may pass through the extremely thin layer of organic blend, thus partially shorting the device.
Assuming a similar or higher absorption in the Thick/NR architecture, the increase in efficiency for the Thin/NR cell indicates a more efficient charge extraction owing to the thin layer of blend [23
]. The slightly higher EQE obtained for the Thick/NR cell can be explained by the fact that the EQE measurements were performed in the dark. Under low-intensity conditions charge carrier recombination only plays a minor role, which can lead to overestimated EQEs especially for devices with non-ideal charge percolation pathways.
The responses of both types of cells as a function of light intensity (Figure b inset) show that only the Thin/NR cells present a linear increase in current density with light intensity up to 1 sun, demonstrating that efficient charge extraction occurs in the conformal cells even at high light intensity and also highlighting the influence of the light intensity on charge recombination dynamics. The non-linear increase of the Jsc
with light intensity for Thick/NR cells [33
] reflects increased recombination due to slow charge collection, which is also likely to be responsible for the smaller FF obtained for the Thick/NR cells. It has been suggested that nanorods can negatively affect the organisation of the thick organic layer [22
] which is consistent with the results of Figure b, i.e. charge collection from the majority of the thick blend in the Thick/NR cells that is not directly adjacent to the collection electrodes is expected to be poor.
The improved charge extraction of Thin/NR cells (Figure b inset) is confirmed by PVD and PCD measurements. Figure c presents the PVD lifetimes (determined from the decay half-lives) of the cells under quasi-open-circuit conditions as a function of light intensity. In the mostly mono-exponential decay curves, we found systematically shorter PVD lifetimes for the Thin/NR architecture, suggesting that charge carrier recombination is quicker. We attribute this directly to the shorter distances that charges have to travel from the external electrodes into the active film before they recombine with charge carriers from the opposing electrode. Since extraction is the complementary process, we infer that charge extraction should also be quicker from thin films (Thin/NR). Interestingly, the differences in the PVD rates between the Thin/NR and Thick/NR architectures are not linearly correlated to the organic film thickness. This suggests that charges in the thick film (Thick/NR) cannot travel through the whole organic layer without recombining but instead have a higher probability of annihilation with other charges that are trapped in islands of donor or acceptor material which form in the film due to its non-ideal internal morphology. This is further supported by the fact that the factor of 2 between the PVD lifetimes is conserved over varying background illumination, suggesting that the active layer morphology, which is intensity independent, plays a crucial role in determining the mechanisms of charge carrier recombination. This is also confirmed by PCD measurements [34
]. Integrals of these current transients (the transient charge) are shown in Figure d. At low background light intensities a similar amount of charges can be collected from both geometries. However, at higher light intensity, where charge densities increase and charge recombination plays a more important role, up to 65% more charges are extracted from the blend in the Thin/NR cell.
The optical density of our conformal cells was evaluated using absorption and reflectance measurements on the Thin/NR design compared to the Thick/NR (Figure ). The absorption of a standard bulk heterojunction design, Thick/flat cell, (see the ‘Methods’ section) was also evaluated as a reference. Figure a shows absorption data for the different cells prior to Ag evaporation. The Thick/flat cell consists of 300 nm of blend on ITO (i.e. without ZnO) and shows an absorption peak at approximately 500 nm as expected. On the other hand, samples incorporating ZnO show higher optical density at wavelengths below approximately 475 nm as a result of both light absorption and light scattering from the ZnO nanorods. In the 480- to 620-nm range, the Thick/NR and Thick/flat blend designs show very close absorption characteristics, and it is clearly seen that the blend in the Thin/NR design absorbs less light than the thick blend cells. This is expected due to the lower volume of material available for light absorption in the Thin/NR cell compared to the thick blend cells.
Absorption and reflectance measurements for Thin/NR, Thick/NR and Thick/flat architectures. (a) Comparison of absorption data without Ag contacts. (b) Reflectance measurements with Ag contacts.
The EQE results of Figure a and absorption results of Figure a together show higher light absorption of the Thin/NR cell than what could be accounted for solely by the amount of blend in the cell. In fact, there are other mechanisms at play which could enhance light absorption in the Thin/NR architecture, namely light being scattered by the nanorods and light trapping due to reflection from the quasi-conformal Ag top contact. In the first case, light scattering by ZnO nanorods is highly possible since it has been shown previously that tailoring the nanorod dimensions (diameter and length) allows effective optical engineering to enhance light absorption [35
]. As for light trapping, it is also highly possible since this has also previously been shown in similar SiNR-P3HT core-shell nanostructures [23
]. We explored the light scattering and trapping effects further by performing reflectance measurements on the different samples with the Ag top contacts present.
The Thick/flat cell reflects a considerably higher proportion of the light than the other two cell designs as a result of the flat Ag contact acting as a mirror and the absence of light scattering. The Thick/NR cell, on the other hand, reflects less light back to the detector than the Thick/flat cell, which is consistent with scattering of the light between the nanorods [35
]. Remarkably, despite having a smaller optical density (from Figure a), the Thin/NR cell reflects the least light, giving weight to the idea of light trapping from the quasi-conformal Ag top contact.
The measurements presented in Figure do not take into account the light scattered outside the reflectometer capture radius. Therefore, for the Thin/NR cell, integrating sphere measurements were performed at several wavelengths where scattering from the rods was expected (as noted from the UV–vis measurements of bare ZnO nanorod arrays) and where P3HT shows prominent absorption. Absorption was found to be uniformly high (approximately 82%) for these wavelengths, confirming that most light is absorbed by the Thin/NR architecture and not scattered out of the cell at angles which cannot be detected by the reflectometer. The 82% absorption of the Thin/NR cell gives a lower estimation (taking parasitic absorptions as zero) of approximately 72% for internal quantum efficiency (IQE) at wavelengths where P3HT is strongly absorbing [24
]. Determining parasitic absorption for nanostructured cells is complicated. However, deviation of the lower bound IQE from 100% in our Thin/NR cells is in part likely due to incomplete Ag electrode coverage, since the tilting of the nanorods leads to some shadowing of the evaporated Ag, and results in areas of the architecture that are not covered by the back contact (as can be clearly seen in Figure c).
The absolute absorption of the Thin/NR cell (not shown) was the same (approximately 82%) for the four wavelengths investigated (457, 476, 488 and 515 nm), at which there are different amounts of scattering and different absorption coefficients of P3HT providing further evidence that the quasi-conformal, highly reflective Ag top contact has an important contribution to the high absorption of the Thin/NR cell [41
]. Thus, our results clearly show that periodic nanostructures are not necessary in order to have high light absorption by the thin active layer in the conformal design.
As in the case of conventional Thick/NR hybrid cells, where efficiencies have been increased by varying the characteristics of the nanorod arrays [25
] or by introducing a top blocking layer, [24
] the control experiment presented here is expected to yield even higher efficiencies in the future by applying similar optimizations. Some clear strategies would include the control of the surface of the nanorods, which has been shown to play an important role in hybrid cells[45
], the deposition of a highly conformal top blocking layer (such as PEDOT:PSS [50
] or WO3
]) and the improvement of the conformal top contact coverage. In addition, optimising the blend thickness and tailoring the spacing and dimensions of the nanorods will enable further improvements in the IQE and EQE [52
]. Electrodepositing the ZnO NRAs using ordered, nanoporous templates such as anodic aluminium oxide is a promising way towards controlling the array parameters (NR diameter, NR length and pitch) [53
]. The optimal architecture will vary depending on the properties of the organic materials employed, which could be either a blend, as presented here, or a single active material [23
]. In particular, the Thin/NR architecture is particularly well suited for systems where poor charge collection is the limiting factor, such as P3DDT, P3OT, F8BT and HBC-PDI, since the low mobility of such materials would not be a detrimental characteristic if a very thin layer of material is used with both electrodes near each other, as is the case for the conformal architecture presented. For the same reason, the conformal approach could be of great interest for non-fullerene electron acceptors, which typically allow higher and broader absorption but cannot compete with fullerenes due to morphological issues [55