Improved Light Utilization Efficiency for an ITO‐Free Semitransparent Organic Solar Cell Using a Multilayer Silver Back Electrode as Infrared Mirror

Semitransparent organic solar cells (STOSCs) exhibit promising application as power‐generating windows in buildings and agricultural greenhouses. Due to unique optical properties of organic semiconductors, they can efficiently absorb near‐infrared light while maintaining a high degree semitransparency in the visible range. Since power conversion efficiency (PCE) and average visible transmission (AVT) frequently stand in a trade‐off relationship, a major challenge in improving the overall performance of STOSCs is maximizing the product of both, called light utilization efficiency (AVT × PCE = LUE). Herein, using multiple layers of aluminium‐doped ZnO (AZO) and silver as an infrared reflecting back electrode, in order to increase current generation while maintaining high visible transparency, is proposed. Using optical modeling, the optimal layer thickness of the AZO layer sandwiched between two Ag layers is determined, leading to an increased photocurrent generation of up to 10%. Simultaneously, experimental findings show that the fill factor decreases with an increasing AZO layer thickness. By adjusting the thickness of the photoactive layer, the blend concentration, and improving the top electrode material the thus‐far highest reported LUE for indium tin oxide‐free STOSCs is attained, reaching 4.0% with a PCE of 8.7% and an AVT of 46.3%.


Introduction
With the rise of climate change, implementing a wide range of sources for sustainable energy generation becomes of increasing importance.Industrialization and the subsequent growth in the world´s population have prompted the search for practical, sustainable energy sources to replace fossil fuels. [1,2]As of 2020, still more than 75% of all global greenhouse gas emissions result from combustion of fossil fuels for energy. [3]Substituting for more low-carbon-emission energy sources is essential to mitigate anthropogenic climate change and its consequences. [4,5]8] However, in 2021, the installed photovoltaic (PV) capacity accounted for only 3.4% of the world´s total electricity generation. [9][12][13] This approach enables the utilization of additional space for solar energy generation, enhancing the potential for widespread adoption of renewable energy.
Promising applications for STOSCs other than incorporated into buildings are self-powered greenhouses in agriculture. [12,14,15]Modern agricultural practices employ a wide range of plastic products including films that are used to give shelter to crop from harsh weather conditions and insects.However, most agricultural plastic products are single use. [16]Through integrating semitransparent solar cells into these films, big areas of land can be utilized for renewable energy collection offering an additional source of income to the farmer while simultaneously protecting the plants.Furthermore, waste production is reduced, since semitransparent PV films would typically rather be engineered to have a longer lifespan.
Using organic semiconductors as active materials for these semitransparent device applications brings many advantages due to their optical properties.Contrary to inorganic materials like, for example, crystalline silicon, the absorption coefficient, starting from the bandgap, does not increase toward higher photon energies.By modifying the molecular structure, the optical DOI: 10.1002/solr.202300561Semitransparent organic solar cells (STOSCs) exhibit promising application as power-generating windows in buildings and agricultural greenhouses.Due to unique optical properties of organic semiconductors, they can efficiently absorb near-infrared light while maintaining a high degree semitransparency in the visible range.Since power conversion efficiency (PCE) and average visible transmission (AVT) frequently stand in a trade-off relationship, a major challenge in improving the overall performance of STOSCs is maximizing the product of both, called light utilization efficiency (AVT Â PCE = LUE).Herein, using multiple layers of aluminium-doped ZnO (AZO) and silver as an infrared reflecting back electrode, in order to increase current generation while maintaining high visible transparency, is proposed.Using optical modeling, the optimal layer thickness of the AZO layer sandwiched between two Ag layers is determined, leading to an increased photocurrent generation of up to 10%.Simultaneously, experimental findings show that the fill factor decreases with an increasing AZO layer thickness.By adjusting the thickness of the photoactive layer, the blend concentration, and improving the top electrode material the thus-far highest reported LUE for indium tin oxide-free STOSCs is attained, reaching 4.0% with a PCE of 8.7% and an AVT of 46.3%.
properties of the material can be adjusted in a way that nearinfrared light is utilized efficiently while maintaining semitransparency in the visible light range. [13,17]Furthermore, organic materials allow for solution processability of lightweight and flexible solar cells.With light utilization efficiencies (LUE) reaching up to 5.35%, [18] STOSC have already achieved very promising values.
The indole-2,10-diyl)bis(methanylylidene))bis(5,6-difluoro-3-oxo-2,3dihydro-1 H-indene-2,1-diylidene)) dimalononitrile). [18]The fabricated cell included a new design for a transparent electrode by integrating a thermally evaporated aperiodic bandpass filter of lithium fluoride (LiF)/tellurium dioxide (TeO 2 ) /LiF.Alternative approaches for STOSCs in literature followed by Guan   3 , improving donor:acceptor blend ratio, layer thicknesses, and using antireflection layers in the solar cell to maintain high efficiencies up to 12.95% at AVT values of 38.67%. [19]When looking for semitransparent cells with AVT values of over 50% without complex optical engineering, Huang et al. reported PCE of 10.01% and a remarkable AVT of 50.05%, achieving a LUE of 5.01%. [20]Through working with four different batches of PCE10-2F with different molecular weights, crystallinity was optimized to create favorable interfacial contact, thereby improving charge dynamics and reducing energy loss.In combination with sequential deposition of the acceptor Y6 film, selective absorption of the active layer was achieved, and the overall performance compared to that of one-step blend casting was improved.
][23] Despite its recognition for high conductivity and transparency in the visible region, this material is known to be brittle and its production is energy inefficient, which makes it inconvenient for applications in flexible STOSC.Additionally, indium is a metal that is rarely found in nature, which leads to a less sustainable solar cell production.Besides using ITO as an electrode, another fabrication step that is commonly seen in STOSC construction is utilizing halogenated solvents, such as chloroform (CF) and chlorobenzene (CB).Despite their strong ability to dissolve organic semiconductors, these are not environmentally friendly solvents and are unsuitable for large-scale production due to their toxicity. [5]To take semitransparent organic solar cell (STOSC) production to a sustainable and competitive level, working on replacements for both the use of ITO and halogenated solvents is advised.
Herein, we propose an alternative semitransparent device architecture that is ITO free and processable without the use of halogenated solvents using a multilayer Al-doped ZnO (AZO) and silver (Ag) back electrode (Figure 1).This back electrode efficiently reflects light in the infrared region back into the photoactive layer (thus increasing absorption therein), while remaining transparent in the visible range of light.This way the resulting PCE can be improved while simultaneously maintaining a high AVT.
Using optical simulations, we investigated the effect of changing the respective distance of two Ag electrode layers on the AVT and photogenerated current of STOSC based on PV-X Plus as photoactive layer.The results show that by increasing this distance from 50 to 180 nm, the product of AVT and photogenerated current is improved.While experimental findings confirm this through an increase of short-circuit current density J SC by up to 10%, it was also found that the fill factor would decrease simultaneously for larger distances between the two Ag layers.By optimizing the photoactive layer thickness, we were able to identify a cell stack that fulfills both the optical and electrical requirements for high AVT and current generation as well as improved fill factor, leading to an overall increase in LUE.In a last step, the AVT was optimized further by substituting for a highly conductive hole transport layer (HTL) as top electrode, allowing for a gridless cell architecture.This way an AVT of 46.3% and a PCE of 8.7% were reached resulting, to the best of our knowledge, in highest reported LUE for ITO-free STOSC of 4%.

Results and Discussion
Using an ITO-free infrared-reflecting back electrode, a series of STOSC were fabricated with the device structure of AZO|Ag| AZO|Ag|AZO|PV-X Plus|HTL-X|HTL-5|Ag-Grid.We used the optical simulation software CODE/SCOUT in order to get a better understanding of how the interference from coherent transmitted and reflected waves at the internal interfaces is determined by the thicknesses of the individual layers forming the back electrode.In a first step, we focused on simulating the AVT of the cell stack and the photogenerated current in the 140 nm-thick photoactive layer upon a variation of the respective distance of the two Ag layers.Since the thickness of the second AZO layer thickness (i.e., the one in between both Ag layers) is equivalent to the distance of the Ag layers, the corresponding simulated cell stack AVT and simulated photogenerated current is plotted as a function of a variation of this AZO thickness (Figure 2a).To find a point where the LUE is highest, the product of PCE and AVT should be maximized.Since the simulation program is based on optical modeling, electrical properties like, for example, the fill factor cannot be determined.Therefore, the PCE itself cannot be simulated.Under the assumption that fill factor and open-circuit voltage V OC remain rather unchanged for different AZO thicknesses, the simulated photogenerated current behaves proportionally to the PCE and therefore to the LUE.By multiplying the simulated photogenerated current with the simulated AVT for different AZO layer thicknesses, a value is established with which the optical favorability of the cell architecture can be assessed similarly to the LUE (Figure 2b).
When looking at the simulated photogenerated current of the full cell stack for different Ag distances, the maximum can be found at an AZO layer thickness of 0 nm, which is equivalent to having a single Ag layer that is twice as thick (14 nm) as the Ag layers in the double Ag electrode stack (each 7 nm).This is due to an increased reflection in the infrared region which consequently increases the absorption (in the photoactive layer) and therefore photogenerated current.The simulated cell stack AVT however has its maximum at a layer distance of 80 nm, close to the simulated photogenerated current minimum which lies at 70-80 nm.This trade-off between AVT and photogenerated current or rather PCE is known from literature and therefore expected. [22,24,25]The product of simulated photogenerated current and AVT has its maximum at thicker AZO layers of around 180-200 nm.
To test how meaningful the simulated results are, we compared them to an experimental series of two different STOSCs with the same device architecture as used for the simulated model (AZO|Ag|AZO|Ag|AZO|PV-X Plus|HTL-X|HTL-5|Aggrid, as depicted in Figure 3d.The layer thicknesses were identical except for the second AZO layer, which were 50-180 nm, respectively. To compare both cells, J-V measurements were performed under illumination of "1 sun" (AM1.5G,corrected for spectral mismatch) to investigate J SC, fill factor, open-circuit voltage V OC , and PCE of which the latter two are plotted in Figure S1, Supporting Information.As shown in Figure 3a, the medianmeasured J SC for the 50 nm layer thickness is at 18.6 mA cm À2 , whereas the value for the 180 nm layer thickness is at 20.0 mA cm À2 .In accordance with the simulations, an increase in J SC is indeed observed (Figure 3c).While the measured J SC for the cell stack with 50 nm AZO layer thickness is also in good agreement with the simulated result of 18.7 mA cm À2 , we observe that for the 180 nm-thick layer the measured one is higher than the simulated one.This discrepancy is likely due to an increased film roughness that has an influence on both the optical and electrical properties of the electrode and therefore on the whole cell stack.Measuring the surface roughness of the electrodes via atomic force microscopy (AFM) shows an average roughness for the electrode stack with 180 nm AZO of 2.7 nm, while the electrode with 50 nm AZO has an average roughness of 2.0 nm.Although the average roughness for both electrodes is relatively low to claim an impact on the morphology of the films above, the maximum measured roughness goes up to 31.3 nm for the electrode with 180 nm AZO while the one with 50 nm AZO has a maximum value of 25.7 nm.The AFM images can be seen in Figure S2, Supporting Information.
The influence of this difference in roughness on the cell performance can furthermore be seen when comparing the fill factor of both cells.We observe that with an increase in layer thickness, the fill factor decreases simultaneously, as shown in Figure 3b.To gain more insight into the influences of potential resistive effects, Suns-V OC measurements were carried out (Figure 4).
Since the Suns-V OC technique is used to determine pseudocurrent-voltage ( J-V ) characteristics of solar cells which are unaffected by their series resistance R s and therefore, only reflecting generation and recombination. [24]Generally, the pseudo-J-V curves of all cells show an increased fill factor and PCE compared to the normal J-V curve, demonstrating that the (real) fill factor is influenced by R s .Both real and pseudovalues of fill factor and PCE of the cell with the 50 nm-thick AZO   layer are in a comparable range (Figure 4a,c).The cells with 180 nm AZO however show a discrepancy among themselves in both real and pseudovalues of fill factor and PCE.This is likely due to the influences of a low parallel resistance R p particularly present in one of the measured cells (Figure 4d).The pseudo-J-V data is converted from a linear fit of the quasi-steady-state V OC as a function of light intensity.Using simulations based on a onediode model with shunt resistance, it was observed that the slope and therefore the area of this linear quasi-steady-state region is greatly influenced when a significant proportion of all charge carriers recombine across the R p (Figure 4e, see Supporting Information for the parameters).Since the cells with 50 nm AZO as well as cell one of the measured 180 nm AZO have a comparably higher R p , the pseudomaximum power point is derived from a V OC at a light intensity that lies within the linear quasi-steady-state area, meaning it can be assumed that the influences on the pseudo-fill factor for these cells are not dominated by R p .The cell two with 180 nm AZO with the lowest R p , however, has a maximum power point that correlates to a V OC barely outside of the linear range, meaning that pseudo-fill factor is not free of the influence of R p and therefore the aforementioned discrepancy to the other 180 nm AZO cell is observed.Even without the effect of R p and R s on the pseudo-fill factor for the other cells, the 180 nm-thick AZO cell number one (Figure 4b) shows a lower fill factor compared to the cells with 50 nm AZO, indicating that there are other additional resistance effects that stem from a change in morphology likely influencing surface and interface recombination.We also derived the ideality factor from the Suns-V OC data (see Figure S6, Supporting Information) and found a higher value of 1.2 for the 180 nmthick AZO layer compared to 1.0 for the 50 nm AZO thickness.Also, the V OC of the former is about 10-20 mV reduced compared to the latter.This points strongly toward a higher density of traps and enhanced trap-assisted nonradiative recombination.
An example from the literature for a correlation between increased surface roughness and a lower shunt resistance was previously reported in the master thesis of Mohammadzadeh, where a severe shunt problem affecting the device performance was linked to the roughness of the underlying silicon substrate. [25]To examine if the surface and interface roughness differs with increased AZO layer thickness, the electrode cross section was investigated with scanning electron microscopy (SEM), depicted in Figure 5.
The bottom electrode stack (AZO|Ag|AZO|Ag|AZO) with a layer thickness of 50 nm AZO in between the Ag layers is seen in Figure 5a, while the electrode with a layer thickness of 180 nm in between is depicted in Figure 5b. Figure 5c,d shows the full cell stacks respectively.Each of the AZO layers has been marked and the approximated thicknesses were determined and labelled based on the scale of the image.Although the interfaces of each AZO layer are identifiable, the resolution of the image and the thickness of Ag layers prevent a clear distinction of the latter from the rest of the stack; hence, they were not labeled.When comparing the electrodes, there is an observable difference between the 50 and 180 nm AZO layer thickness.The morphology of the layers themselves differs and interface as well as surface roughness are more pronounced for the layers on top of the thicker AZO layer.Whether this has a clear effect on active layer morphology could not be determined within the scope of this study.
Further measurements such as charge extraction by linearly increasing voltage (CELIV), space-charge limited current (SCLC) of single-carrier devices, and transient photocurrent were carried out (see Figure S3-S5, Supporting Information).However, no clear evidence for example reduced hole mobility was found.Only the transient photocurrent showed a slower response in the case of 180 AZO thickness, which is in accordance with the larger density of trap states as discussed above.
To investigate this effect on the fill factor to a broader extent, we looked at a wider variation of AZO layer thicknesses, shown in Figure 3e.Throughout all samples, a consistent decrease in fill factor for thicker AZO layers was observed.For the single Ag electrode (AZO|Ag|AZO), the fill factor is the highest, which would agree with the assumption that number and thickness of layers influence film morphology that influences optical and electrical behavior.Not only does the single Ag electrode show the best fill factor, it also shows excellent agreement with the simulated optical properties, that is, reflection, transmission, and absorption with the measured data, as shown in Figure 3f.
To overcome this fill factor issue, optical simulations were used to optimize the thicknesses of the layers in the cell stack for the single Ag electrode, keeping each thickness in reasonable ranges that accounts for electrical functionality of the cell.While doing so, we observed that by changing the photoactive layer thickness to 100 nm, the maximum simulated AVT for the cell stack and the minimum simulated photogenerated current of the photoactive layer would shift (Figure 6a) compared to the maxima and minima of the previous cell stack with 140 nm layer thickness (Figure 2a).When looking at the product of the cell stack AVT and current generation for different AZO layer thicknesses, we would observe that the best relation between those values lies either at a thickness of 0 or 200 nm.Using a single Ag electrode, that is, a 0 nm AZO layer thickness in combination with a thinner photoactive layer (100 nm), we were able to overcome the fill factor problem while simultaneously having an optically optimized solar cell stack.
This again was investigated experimentally.An analogue series of solar cells based on PV-X Plus as absorber material on a single Ag electrode (AZO 45 nm|Ag 14 nm|AZO 40 nm) was built with two different photoactive layer thicknesses of 100 and 140 nm.As shown in Figure 6c), the J SC decreases for a thinner photoactive layer of 100 nm compared to the 140 nm thick layer from an average 21.2-17.0mA cm À2 , which is due to reduced absorption.When looking at the fill factor, however, we notice an increase for the cell with a 100 nm photoactive layer thickness (Figure 6d).Nevertheless, the overall PCE still decreases for a thinner photoactive layer (Figure 6e).However, the higher AVT of 33% for the 100 nm-thick photoactive layer compared to 25.9% for the STOSC with 140 nm overcompensates this effect.Therefore, the overall LUE increases from 2.96% to 3.37%.
After optimization of the back electrode and photoactive layer, the top side of the cell stack was improved to further increase the LUE.In order to do so, the top HTL-5 layer was exchanged with a new high-conductivity PEDOT formulation called Clevios F HC Solar (R&D grade SCA 2003).This allowed omitting the Ag grid on top of the HTL, leading to a higher (visible) transmission while maintaining PCE.It should be noted that maintaining PCE is possible for the active cell area used (0.0925 cm 2 ), since the loss in fill factor through leaving out the grid completely is comparable to the increase in J SC due to a bigger illuminated area.For future plans of building cells or possibly modules with bigger cell areas, using Ag grid lines on top of the cell stack with a distance and width that provide the same trade-off between fill factor and J SC is therefore crucial to achieve comparable results in PCE.At the same time, we varied blend concentrations and spin-coating speeds for the photoactive layer to fine tune the optimal morphology and the trade-off between optical and electrical performance in the solar cell stack.The results are shown in Table 1 and Figure 7.The best LUE was calculated for a PV-X Plus concentration of 20 mg mL À1 at a spin coating speed of 2000 rpm, which is equivalent to a photoactive layer thickness of about 65 nm.For this cell, an average LUE of 4.0% was reached with an average PCE of 8.7% and an AVT of 46.3%.The external quantum efficiency (EQE) data as well as the corresponding J-V curves are plotted in Figure S8 and S9, Supporting Information.When comparing similar layer thicknesses of different blend concentrations, a trend was observed where the efficiencies increase with lower concentrations.One possible explanation could be a change in solubility at lower concentrations which could possibly increase the molecular weight distribution of the blend and therefore improve fill factor and current generation.This phenomenon of the effect of molecular weight on efficiency was previously reported by Karki et al.They observed that an increase in low-molecular-weight fraction (Mn = 2.5 kDa) of the donor material affected the overall PCE due to a decrease in fill factor and J SC . [26]This effect of decreasing PCE was observed consistently for increasing low-molecularweight fractions of up to 52%.
We also compared the impact of the direction of the incoming light.Figure S7, Supporting Information shows performance data comparing illumination through the top (PEDOT:PSS) and the bottom electrode (Ag/AZO).Due to the pronounced NIR-reflecting properties of the Ag electrode, the current is reduced almost by a factor of 2 when illuminated through that electrode.This is in perfect agreement with our optical simulations which are also shown in Figure S7, Supporting Information.The fill factor remains rather unchanged and V OC is reduced by about 20 mV.This can perfectly be explained by an almost 50% reduction of the photogeneration in the absorber layer.

Conclusion
We demonstrated an increased LUE for a solar cell stack based on PV-X Plus as absorber using an ITO-free AZO/Ag multilayer back electrode as infrared mirror.Both optical simulations and experimental results were used to strategically improve electrode and photoactive layer thickness to optimize optical interference and fill factor of the device.With this strategy, we achieved an STOSC cell with a PCE of 10.2%, AVT of 33.0%, and LUE of 3.37%.A newly developed PEDOT:PSS formulation called  CLEVIOS F HC SOLAR (R&D GRADE SCA 2003) was used to allow for a grid-less device architecture for cells with an active area of 0.0925 cm 2 .For bigger modules, an adjusted Ag grid structure will have to be introduced to achieve comparable results.In combination with optimization of the blend concentration and fine tuning of the photoactive layer thickness, the so far highest LUE of 4.0% for ITO-free STOSCs processed from nonhalogenated solvents was achieved, reaching a PCE of 8.7% with an AVT of 46.3%.Overall, this work provides new access to ITO-free semitransparent solar cells with excellent LUEs that offer new incentives for solar cell production without the use of halogenated solvents.

Experimental Section
Device Fabrication and Characterization: In a first step in the fabrication of semi-transparent devices, 25 Â 25 mm 2 glass substrates purchased from Glasmanufaktur Pfähler were cleaned in ultrasonic baths of acetone, isopropanol, and water for 10 min each and dry blown with nitrogen.Alternating layers of AZO and Ag were sputtered onto the glass substrates (double Ag: AZO 30 nm |Ag 7 nm |AZO 50 nm; 100 nm; 180 nm |Ag 7 nm | AZO 41.5 nm and single Ag: AZO 45 nm |Ag 14 nm |AZO 40 nm) structured into active areas for six individual solar cells and electrode contacts with a shadow mask.A PV-X Plus film (PV2300:PV-A-3:N1100 1:1:0.222 mg mL À1 from Raynergy Tek) was spin cast from o-xylene (1200 rpm for 140 nm, 3000 rpm for 100 nm) at 60 °C in a nitrogen-filled glove box (GB) onto heated substrates (110 °C) and annealed for 10 min at 110 °C.For the record device sequence (Figure 5), substrate heating before spin casting was left out.A 40 nm HTL-X layer (purchased from Raynergy Tek) followed by 60 nm HTL-5 layer (purchased from Raynergy Tek) was spin cast (GB) on top of the photoactive layer and annealed for 2 and 5 min at 110 °C, respectively.The HTL layers were structured with ethanol wiping followed by a photoactive layer structuring with o-xylene.For cells with a grid on the active area, a 100 nm-thick Ag grid structure was deposited via thermal evaporation at a pressure below 10 À5 mbar.Finally, a support structure for current collection of Cr (4 nm)/Ag (100 nm) was evaporated at a pressure below 10 À5 mbar on top of the PEDOT:PSS outside the active area on both sides at a distance of 0.2 cm.
For the record device stack (Figure 5), we used CLEVIOS F HC SOLAR (R&D GRADE SCA 2003, supplied by Heraeus) instead of HTL-5 and no grid structure was evaporated on top of the active area.Clevios F HC SOLAR is a water-based PEDOT:PSS (poly(3,4-ethylenedioxythiophene): polystyrene-sulfonate) ready-to-use formulation comprising 5% DMSO as high-boiling-point solvent, which was thoroughly filtered over a 0.2 μm hydrophilic filter.Clevios F HC SOLAR (SCA 2003) was characterized by the following properties: solids 1.11%, viscosity 26 mPas, pH 2.2, conductivity 819 S cm À1 .
Upon measurement with a Byk HazeGard, a film applied with a 12 μm wire bar on a Melinex 506D PET substrate and dried at 120 °C for 10 min showed a surface resistivity of 175 Ω sq À1 , a transmission of 88.3%, and a Haze 0.47%.
Current-voltage measurements under simulated AM1.5G light by a class A solar simulator (Newport SP94063A-SR1-167, corrected for spectral mismatch) were carried out with a computer-programmed Keithley 2400 source meter.
UV/vis measurements were carried out from 280 to 1200 nm by a Perkin Elmer Lambda 950 UV/vis/NIR spectrophotometer.Layer thicknesses were determined with a Veeco Dektak 150 measuring samples of single layers on glass.
Analysis: Optical modeling was performed by transfer matrix method using the software package CODE/SCOUT from W. Theiss.Optical coefficients n (refractive index) and k (extinction coefficient) for all materials were generated from a kim oscillator model, which was fit to measure reflection and transmission data of single layers with a given thickness on a glass substrate.The optical coefficients where then used to model the solar cell stack with defined layer thicknesses in order to calculate theoretical current generation and AVT of the cell considering the optical interference effects.Using the fit parameter setting layer, thickness optimization was performed to maximize AVT and current generation.
AVT is used to characterize semitransparent solar cells by incorporating the transmission of the solar cell stack T(λ), the intensity of the solar AM1.5G spectrum S(λ), and the photopic response of the human eye P(λ).It is defined as follows [27] AVT ¼

Figure 1 .
Figure 1.Layer stack of the semitransparent solar cell with a double Ag electrode on the backside used as a near-infrared mirror.

Figure 2 .
Figure 2. a) Simulated cell stack (140 nm photoactive layer thickness) AVT and simulated photogenerated current for AZO layer thicknesses from 0 to 200 nm.b) Product of simulated photogenerated current and simulated AVT at respective AZO layer thicknesses from 0 to 200 nm.

Figure 3 .
Figure 3. a) Short-circuit current density J SC , b) fill factor, and c) J-V curve for a series of solar cells with 50 and 180 nm AZO layer thickness, respectively.d) Device architecture of the STOSCs with the varied AZO layer thickness marked in red.e) Fill factor for a series of solar cells with 0-180 nm AZO layer thickness.0 nm AZO layer thickness corresponds to a single Ag electrode (AZO 45 nm|Ag 14 nm|AZO 40 nm).f ) Simulated reflection, transmission, and absorption spectra of a single Ag electrode (AZO 45 nm|Ag 14 nm|AZO 40 nm).

Figure 4 .
Figure 4. Pseudo J-V curve compared to measured J-V curve of cells with a a,c) 50 nm-thick AZO layer and b,d) 180 nm-thick AZO layer.e) V OC as a function of light intensity for different values of parallel resistance.

Figure 5 .
Figure 5. SEM image of the cross section of the bottom electrode stack with a) 50 nm AZO, b) 180 nm AZO, as well as the full cell stack with c) 50 nm AZO and d) 180 nm AZO.

Figure 6 .
Figure 6.a) Cell stack (100 nm-thick photoactive layer)-simulated AVT and simulated current generation for AZO layer thicknesses from 0 to 200 nm.d) Product of simulated current generation and simulated AVT at respective AZO layer thicknesses from 0 to 200 nm.b) J SC , c) fill factor, e) efficiency, and f ) reflection, transmission, and absorption spectra for a series of solar cell stacks with 140 and 100 nm photoactive layer thickness, respectively.

Figure 7 .
Figure 7. a) Efficiency and b) transmission of an AZO|Ag|AZO|PV-X Plus|HTL-X|CLEVIOS F HC SOLAR (R&D GRADE SCA 2003) cell stack with different blend spin-coating speeds [rpm] and blend concentrations.The gray bell curve represents the photopic response of the human eye.

Table 1 .
Characterization of an AZO|Ag|AZO|PV-X plus|HTL-X|CLEVIOS F HC SOLAR (R&D GRADE SCA 2003) cell stack with different blend spincoating speeds and blend concentrations.