Fully printed organic solar modules with bottom and top silver nanowire electrodes

One of the advantages of organic photovoltaics (OPV) over other contemporary technologies is its relative ease of processing. There are, however, very few works that have realized fully printed devices, including the bottom electrode, let alone with a scalable process in a reasonable device size (>1 cm2). In this work, design steps and optimization processes towards fully printed OPV modules with scalable processes are demonstrated for the first time. An overview on issues related to upscaling with printed electrodes is first provided. The various issues are then addressed by a rational design process supported by measurements and calculations. Finally, a set of fully printed OPV modules are fabricated using these optimized parameters that have over 3.5‐cm2 active area with 5% efficiency. For the first time, this work has also demonstrated the process compatibility of fully printed device structures with non‐fullerene acceptor systems, which enables more design opportunities for the current generation of high‐performance OPV materials.


| INTRODUCTION
Solution processed thin film opto-electronic devices are particularly attractive for photovoltaic applications due to the high throughput and potential cost savings associated with roll-to-roll manufacturing.
In the case of thin film solar cells, tremendous effort has been put into demonstrating that high power conversion efficiencies can be achieved for small lab scale devices (active area (1 cm 2 ). 1,2 However, the vast majority of reports considers only partially solution processable architectures, that is, one or several of the interlayers and electrodes are most often processed by physical vapour deposition, for example, thermal evaporation, e-beam evaporation, ALD and sputtering, among others. [3][4][5] These architectures utilize substrates that are precoated and prepatterned with indium-tin-oxide (ITO) or ITO/Ag/ITO (IMI) sandwich structures, a process that relies on energy intensive sputtering process and requires indium, which is listed as a rare element. 6 Thus, it becomes necessary to either rely on external providers or to set up both, solution process and vacuum deposition, in a single production line, which unavoidably raises the cost of production. In contrast, printing bottom electrodes allows for all layers to be deposited by solution processable roll-to-roll printing method, which will greatly simplify the production process.
In the case of organic photovoltaics (OPV) and thin film printed photovoltaic technologies in general, the production costs are a key criterion that determines market competitiveness. Given that transparent conductors usually constitute a large part of the bill of materials (BOM), replacing materials processed via costly physical vapour techniques with potentially inexpensive solution processed alternatives will constitute an important cost advantage. 7,8 Besides the inherent economic benefit of solution processing, the possibility of depositing all layers of a device from the liquid phase, including both electrodes, will allow devices to be printed also on discretionary, untreated surfaces in an in-line production process, precluding the need for transferring the devices between different deposition and/or patterning equipment.
Fully printed OPV cells have already been demonstrated by various works in literature. [9][10][11][12][13][14] The printed electrodes, at least one of which must be semitransparent in order to allow light to pass through, are realized by various materials, methods and technologies. A contemporary solution to this challenge is to print a combination of opaque silver grid and transparent poly (3,4-ethylenedioxythiophene):poly (styrenesulfonate) (PEDOT:PSS) film. 15,16 This method however requires two processing steps and can cause short circuits. 17 Furthermore, the optical disturbance that is introduced by the grids also affects the appearance of the devices. While there are already many alternatives proposed to address these shortcomings, 18,19 one of the most promising technology is based on mesh network electrodes formed by highly conductive nanowire dispersion, such as silver (AgNW). 9,12,14,20,21 There are already numerous works published by us or other groups on the optimization of this type of electrode for organic solar cell application but most are still limited to smaller device areas. 3,9,[12][13][14]22 Besides the well-reported surface roughness issue associated with AgNW electrodes, 11,21,23,24 one apparent problem as device area increases is the increase in resistive losses associated with the electrodes. This necessitates the patterning of a single large cell into a number of smaller but interconnected cells, that is, a solar module. 25,26 An organic solar module is fabricated by patterning a large solar cell into smaller stripes and then serially connect them monolithically. An illustration of the interconnection scheme is shown in Figure 1. Three distinct patterning steps are introduced during the fabrication process, namely, the P1, P2 and P3 steps. The first step, P1, separates the bottom electrode into discontinuous conductive strips. The P2 step is performed after PEDOT:PSS layer deposition.
This creates a gap in all the interlayers to allow the top electrode to connect to the bottom electrode during the deposition process.
Finally, the P3 step separates the top electrode into discontinuous conductive strips so that individual cells are finally connected in series.
More details regarding the modularization process in an OPV can be found in some of our previous works. [26][27][28]31,32 Despite all the progress in small-scale fully printed solar cells, fully  29,38 The top AgNW electrode has a sheet resistance of roughly 6 Ω/□. The transmittance data of all these electrodes are shown in Figure S2. ITO based devices are also included for comparison. Figure 3 shows the current-voltage characteristics for cells with different ZnO layer thicknesses, whereas Figure 4 lists their respective performance parameters. All semitransparent devices have lower j sc and FF due to the low reflectance and conductivity of AgNW compared with evaporated silver (EvapAg). The result confirms what has already been suggested by the SEM images shown in Figure 2, that is, that AgNW bottom electrodes without ZnO layers of sufficient thickness are in direct contact with the active layer. This results in shorts between both electrodes, which lowers the open circuit voltage (V oc ), as also illustrated by other works. 23,37 AgNW electrodes with ZnO layer thickness above 100 nm perform similarly to those with ITO bottom electrodes ( Figure 4). However, such a thick planarization layer has also slightly reduced the short circuit current density (j sc ).
This is due to the absorption by the thick ZnO layer, which reduces printed organic solar modules lies on the critical pulsed LASER ablation process to establish such an efficient interconnection. However, the dark field microscopy image after the P2 LASER patterning step in Figure 5B indicates that the bottom AgNW electrode is severely damaged, which has disconnected individual cells in the module. This also reveals that the LASER ablation threshold of AgNW bottom electrodes is significantly lower than that of vacuum-deposited materials, such as ITO. Further attempts with lower total fluency at 0.626 J/cm 2 and much wider interconnection still result in non-connected modules. The coloured AgNWs in Figure 5C indicate the interconnections are still covered by the stacked layers that are printed on top of the AgNW electrode (see Figure 1). This has drastically increased the interconnection resistance, which severely limits the FF and even V oc of the module as shown in Figure 5A for both cases. The series resistances extracted from the curves show that they are over an order of magnitude higher than our previously reported values. 31 From the results obtained, it becomes clear that the process window for such structure is narrower than with ITO based electrodes. 30 Thus, in order to minimize the electrical resistance of the interconnection between cells, the ablation threshold of the bottom electrode must first be determined and the LASER power must be adjusted accordingly.

| LASER power optimization
A test is first performed to determine the upper limit of the total patterning LASER energy per unit area (total fluency) before the electrode is damaged. Using a nanosecond pulse LASER, patterning lines are drawn perpendicularly across layered stacks as depicted in Figure 6 and the change in electrical resistances across the bottom AgNW electrodes is measured. Two different layer stacks are tested to determine the limit on the total fluencies for the P1 and P2 patterning step, respectively.
In the previous section, thicker ZnO layers have been demonstrated to reduce the leakage current of the fully printed cells. However, the impact of ZnO thickness variation on the LASER energy intensity has not been investigated. The P1 step ( Figure 6A) should ensure that the individual conductive stripes at the bottom electrode are no longer connected. Figure 7A shows the change in resistance against the total fluency applied with different ZnO thicknesses. The shaded areas represent the total fluencies that are applicable for this processing step. Notice that the area is continuous at the right edge as the applicable total fluencies can be above the values presented. . Evaporated silver as top electrode is also included for reference (thin grey line). The detailed layout of the cells structure is provided in Figure S3 However, the total fluencies chosen should be the minimum applicable energy to minimize ablation damage to the substrate. It is clear from the data that the total fluencies required to cut open the AgNW electrode are significantly higher for electrodes with thicker ZnO layers, which could be an issue for less resilient substrates, such as PET. The ZnO blended AgNW electrode, however, provides a critical processing advantage over bare AgNW electrode. Most printed electrodes such as AgNW cannot be washed or ultrasonicated after P1 patterning step to remove conductive debris, unlike ITO electrodes.
One of the ways to reduce these conductive debris from being generated is by intermixing the AgNW electrode with another nonconductive layer, which is the ZnO layer in this case.
The second LASER patterning step, P2 ( Figure 6B F I G U R E 6 Measurement device structure for determining the total laser fluency limit: (A) for patterning the AgNW bottom electrode, that is, the P1 step; (B) for creating an opening for the connection between both electrodes, that is, the P2 step electrodes shown in Figure 2, it is reasonable to conclude the ablation threshold of the AgNW electrode is indeed enhanced by such coverage.
While the experiments in Figure 6B have indicated the LASER intensity threshold for the P2 patterning step, the critical relation between the interconnection resistance and the total fluency is still missing. In our previous work, we have devised a measurement structure, the cross Kelvin bridge resistor (CKBR), to measure such interconnection resistances of a thin film solar module. 31 In this work, we have fabricated an array of CKBRs to measure the effect of total fluencies on the interconnection resistance (R c ). The layout of the measurement structure, together with the calculation of R c and of the specific interconnection resistances p c are provided in Figure S4. The interconnection resistances resulting from P2 ablation using different laser fluencies are tabulated in Table 1 with their respective openings before top electrode deposition shown in Figure 8. Within all conditions tested, the wires are still clearly visible after LASER patterning.
Samples that have sustained higher intensity irradiation appear to have more distorted and fused wires when compared with samples irradiated with pulses of lower energy. This indicates that all LASER intensities tested are indeed high enough to remove the active layer and the ETL but still low enough not to remove the AgNWs at the bottom. However, even when all total fluencies tested are close to but within the threshold limit discussed earlier, p c and R c still show a significant variation across the range tested. This indicates that there is only a narrow process window for P2 ablation, which is limited by insufficient removal of active layer and ETL on one hand ( Figure 8A) and serious damage to the AgNW bottom electrode on the other hand ( Figure 8C). The residues at insufficient LASER power consist mainly of ZnO. Nonetheless, an optimal total fluency can still be found at 0.8 J/cm 2 with roughly R c = 2 Ω. Thus, total fluencies of 3000 and 0.8 J/cm 2 are selected for the P1 and P2 steps, respectively, which translates to the experimental parameters given in the experimental section.

| Interconnection structure optimization
While the p c obtained is well within the acceptable range for OPV module fabrication, R c is still relatively large for the module geometry similar to the ones used in this work. As a general rule of thumb, in order to ensure minimal resistive losses at the interconnection, the R c F I G U R E 7 Plots of resistance change of the AgNW bottom electrode after LASER patterning for the measurement structure in (A) Figure 6A, which corresponds to the P1 patterning step and in (B) Figure 6A, which corresponds to the P2 patterning step. Two different ZnO layer thicknesses were tested, which are represented by two different colours. The shaded areas represent the total fluencies which can be utilized for the particular patterning step. steps are first measured in Figure 9. Table 2 then shows the relation of R c and p c with different applied current and different interconnection widths in Figure 9.
All images in Figure 9 have indicated the presence of debris along the edges of the interconnection line. This is due to the Gaussian energy distribution in a LASER spot. 27,28 The higher intensity at the centre causes the wires to distort slightly while the edges are still partially covered. Undoubtedly, this has reduced the effective interconnection area and thus increased p c . The partially covered areas are roughly the same in size for all samples. Thus, p c would be reduced by roughly 10% for the wider interconnections but by more than 30% for the 70 μm wide interconnection if they were not considered. which significantly reduces the current crowding and increases the transfer length to tens of micrometre already. Apparently though, with such a wide interconnection width (>100 μm) the area closest from the P3 line is still not being utilized for current transfer.
As evident from Table 1, p c also reduces as the current per unit width of the interconnection increases. Joule heating of the AgNW network increases the resistance of the wires, but this effect is overcompensated by the welding of the nanowire junctions. [43][44][45][46] However, as indicated in other works, the sustained application of high current through the interconnection will result in damage to AgNW junctions and the eventual destruction of the network.
In order to quantify the effects described above, the changes in potential difference in seconds across the AgNW/AgNW interconnection under a constant applied current were monitored in Figure 11.
Two levels of current are tested in here. At 38 mA/cm ( Figure 11A), it corresponds to the current applied to the interconnection when a solar module is at its short circuit condition, while at 377 mA/cm F I G U R E 8 Top down dark field microscope images of interconnection regions after LASER ablation of the P2 line, using the total fluencies of 0.74, 0.80, and 0.90 J/cm 2 in Table 1. Notice that the higher total fluency has also affected the width of the opening, that is, the interconnection width. p c is calculated using the opening widths according to these photographs; (A) ZnO residues on top of the AgNW layer appear yellow. These residues separate the top AgNW from making direct contacts to the bottom AgNW network, which greatly increases R c and p c ; (B) the white dots indicates the AgNW in that region is melted and turns into a particle while the white lines are where wires had resided previously. The AgNW networks that are slightly damaged appears to give the best result; (C) the extensive whitening at the opening indicates the extensive damages to the AgNW network, which greatly increase R c and p c ( Figure 11B), it corresponds to the current at high injection conditions for the fully printed cells in Figures 3 and 4 during measurement.
Thus, the interconnection must remain stable permanently under the loading condition postulated in Figure 11A while it must endure, at least for a brief period of time, the condition postulated in Figure 11B.
The results are qualitatively similar to what has been presented in other works. 46 A brief summary of their findings are provided here.
When high current density is applied, the thermal stress will weld individual wires together due to the higher junction resistance than individual wires. This reduces the overall resistance of the network.
However, when such stress is too high, nanowires will be fused into silver nanoparticles, destroying the junction and wires altogether. [46][47][48] This explains the hikes in resistance. These two effects counteract each other continuously and result in the jumps in Figure 11B. These results indicate that interconnection widths of 100 and 140 μm are feasible candidates for AgNW/AgNW interconnections. As the enhancement at 140 μm width when compared with 100 μm width is marginal and both are within the R c of 0.7 Ω prescribed, the interconnection width of 100 μm is thus selected for its higher geometrical fill factor (GFF; module layout is given in Figure S5). GFF is the ratio between the active area producing photocurrent and the total area in a solar module (see Figure 1). While a high GFF module designs minimize the loss due to inactive area, it also amplifies the current density at the interconnection region by a factor of active area over the contact area. The average current density J I at interconnections of different widths is calculated as T A B L E 2 Interconnection resistance R c and specific interconnection resistance p c of CBKRs for different interconnection widths measured from Figure 9 Interconnection width 70 μm 100 μm 140 μm Thus, J I of the module layout in Figure S5 that has an active cell width of 5 mm with the maximum current density in Figure 4 is 0.54 and 0.77 A/cm 2 for interconnection widths of 100 and 70 μm, respectively. Thus, the power density of the interconnection region is calculated as 82 mW/cm 2 and 1.05 W/cm 2 using the R c values from   Figure 11B the potential difference for the 70 μm interconnection falls to zero at around 2 s. This indicates the interconnection is broken and the digital multimeter cannot measure the signal anymore. The stepwise plateaus shown in Figure 11B coincide with other works that under constant joule heating the junctions are sintered procedurally until a more efficient percolation pathway is formed 46

| Impact of AgNW formulation on interconnection resistance
There are numerous AgNW formulations from the literature and in the commercial market. The scope of the discussion thus far has been limited to AgNW formulations without additives. However, there are also AgNW inks that contain fillers to enhance their adhesion to substrate surfaces, as revealed by our previous works. 9 The presence of such fillers on the AgNW network is first confirmed by the SEM images in Figure S1a-c, which also compares the surface morphology before and after coating ZnO layers at different thicknesses. The filler within the AgNW film is clearly visible when Figure S1a is compared with Figure S1c. The reduced contrast between wires as well as the appearance of bubbles in Figure S1a are good indications that the AgNWs are at least partially submerged in the filler matrix. This is best illustrated in Figure S1b and Figure 2B.
Notice that the AgNWs are well intermixed and in more direct contact with the ZnO nanoparticles for AgNW films without fillers as compared with those with fillers.
The AgNW electrodes with fillers are then used to fabricate the same CBKR structures that were presented before. The results presented in Table 3 indicate that both R c and p c are orders of magnitude higher than the results for the AgNW films without fillers presented in Note: Active module area is 3.512 cm 2 . NOTE: PCE and j sc are calculated from active device area. Overall transmittance is measured at 550 nm for an encapsulated fully printed organic solar module.
around the AgNWs in Figure 12 as compared with Figure 9, are indeed hampering the contact between both electrodes. Given the high current density at the interconnection, such a high R c will cause too much potential drop and render the OPV module unusable as described in Section 3.2.

| FULLY PRINTED OPV MODULE
To demonstrate the processability of the fully printed semitransparent solar modules with AgNW/AgNW interconnections, two types of OPV modules were fabricated by a combination of the process parameters developed in the last two sections. In order to demonstrate the industrial relevance of our approach, two active layer systems were chosen which are used in manufacturing large area OPV modules.
Besides P3HT:oIDTBR, PV2000:PCBM that has demonstrated high performance in some of our previous works, was also utilized as photoactive layer for module fabrication. 22,49,50 The characteristics of PV2000:PCBM based devices of 0.1 cm 2 cell area are provided in Figure S6 and Table S1. The same cell structure and the same LASER patterning parameters as for P3HT:oIDTBR were employed for the PV2000:PCBM system. This proves the general applicability of the cell architecture and the processing method to both fullerene and nonfullerene-based systems.
The performance data are tabulated in Table 4 with their respective current voltage characteristics shown in Figure 13. The mod-

| CONCLUSION
This work has provided a guideline on the way to design and optimize fully printed semitransparent OPV modules with scalable process technologies. We have demonstrated, from cell to modules, the route we have taken to achieve a fully printed OPV module with respectable performance with fully printed ITO-free electrodes. AgNW have proven to be an excellent electrode material for small scale lab cells. [9][10][11][12][13][14]  Second, the critical interconnection is further optimized to achieve less than 0.7 Ω in resistance by analysing its change with respect to the width of interconnection. We have also experimentally proven intercon-  ZnO was coated at 5 mm/s 30 C at 400-μm gap to achieve 40-nm film thickness. The process was repeated three times in opposite directions to achieve 120-nm film thickness. The film was then annealed again at 120 C for 5 min. The bottom electrodes were then patterned by LASER at 4 W 60 kHz at 500 mm/s. The same procedure is also applied as the P1 step in module fabrication.
The photoactive layer was printed from one of the two formulations described above. P3HT:oIDTBR solution was coated at 33 mm/s 65 C at 400-μm gap. PV2000:PCBM solution was coated at 30 mm/s 65 C at 400-μm gap. PEDOT:PSS layer was then coated at 5 mm/s 65 C at 400-μm gap. All cells were annealed at 140 C for 5 min under nitrogen atmosphere. For organic solar modules, the samples were then transferred back to ambient for P2 LASER structuring at 800 mW 60 kHz at 500 mm/s. One LASER structuring lines is roughly 70 μm. Wider interconnection is achieved by overlapping each patterning line at 5-μm distance.
Top AgNW layer was coated at 20 mm/s 30 C at 400-μm gap with formulation that has no fillers. The top electrodes were patterned by LASER at 800 mW 60 kHz at 500 mm/s. The same procedure is also applied as the P3 step in module fabrication. The film was then annealed again at 120 C for 5 min under nitrogen atmosphere.
EvapAg layer was coated in evaporator under pressure <6 * 10 À6 bar for 200 nm at a rate <0.05 nm/s.
To make efficient contacts with the bottom electrode after fabrication, Q-tips that are soaked in toluene are used to gently wipe the active layer off. Silver paste is then painted at the contact positions of the exposed bottom electrode and at the contact positions of the top electrodes.
After, fabrication cells were encapsulated by adding a barrier glass on top of the device with the UV curable epoxy Delo-Katiobond ® LP655 for 2 min.