Marko
Jošt
*a,
Eike
Köhnen
a,
Anna Belen
Morales-Vilches
b,
Benjamin
Lipovšek
c,
Klaus
Jäger
d,
Bart
Macco
e,
Amran
Al-Ashouri
a,
Janez
Krč
c,
Lars
Korte
e,
Bernd
Rech
e,
Rutger
Schlatmann
b,
Marko
Topič
c,
Bernd
Stannowski
b and
Steve
Albrecht
*a
aYoung Investigator Group Perovskite Tandem Solar Cells, Helmholtz-Zentrum Berlin für Materialien und Energie GmbH, Kekuléstraße 5, 12489 Berlin, Germany. E-mail: marko.jost@helmholtz-berlin.de; steve.albrecht@helmholtz-berlin.de
bPVcomB, Helmholtz Zentrum Berlin für Materialien und Energie, Schwarzschildstr. 3, 12489 Berlin, Germany
cUniversity of Ljubljana, Faculty of Electrical Engineering, Tržaška 25, 1000 Ljubljana, Slovenia
dYoung Investigator Group Nano-SIPPE, Helmholtz-Zentrum Berlin für Materialien und Energie GmbH, Albert-Einstein-Straße 16, 12489 Berlin, Germany
eInstitute for Silicon Photovoltaics, Helmholtz-Zentrum Berlin für Materialien und Energie GmbH, Kekuléstraße 5, 12489 Berlin, Germany
First published on 25th October 2018
Efficient light management in monolithic perovskite/silicon tandem solar cells is one of the prerequisites for achieving high power conversion efficiencies (PCEs). Textured silicon wafers can be utilized for light management, however, this is typically not compatible with perovskite solution processing. Here, we instead employ a textured light management (LM) foil on the front-side of a tandem solar cell processed on a wafer with a planar front-side and textured back-side. This way the PCE of monolithic, 2-terminal perovskite/silicon-heterojunction tandem solar cells is significantly improved from 23.4% to 25.5%. Furthermore, we validate an advanced numerical model for our fabricated device and use it to optically optimize a number of device designs with textures at different interfaces with respect to the PCE and energy yield. These simulations predict a slightly lower optimal bandgap of the perovskite top cell in a textured device as compared to a flat one and demonstrate strong interdependency between the bandgap and the texture position in the monolithic stack. We estimate the PCE potential for the best performing both-side textured device to be 32.5% for a perovskite bandgap of 1.66 eV. Furthermore, the results show that under perpendicular illumination conditions, for optimized designs, the LM foil on top of the cell performs only slightly better than a flat anti-reflective coating. However, under diffuse illumination, the benefits of the LM foil are much greater. Finally, we calculate the energy yield for the different device designs, based on true weather data for three different locations throughout the year, taking direct as well as diffuse illumination fully into account. The results further confirm the benefits of front-side texture, even more for BIPV applications. Overall, devices built on a both-side textured silicon wafer perform best. However, we show that devices with textured LM foils on the cell's front-side are a highly efficient alternative.
Broader contextMetal halide perovskite materials are perfectly suitable for a tandem solar cell due to their excellent optoelectronic properties and tunable bandgap. An excellent combination as an add-on to existing fabrication routes for conventional, market dominating silicon solar cells, it has the potential to increase the power conversion efficiency significantly while only marginally increasing the production costs. To push the power conversion efficiency (PCE) of the monolithic perovskite/silicon tandem devices further, efficient light management, including textured interfaces, is of utmost importance. In our work, we present an efficient monolithic perovskite/silicon-heterojunction tandem solar cell with a 25.5% PCE, enabled by a textured light management (LM) foil, applied at the front-side. By building an experimentally verified optical model, we are able to determine the losses in the fabricated device and use it to optically optimize and estimate the PCE of a number of configurations with a texture at different positions. Our simulations show that a 32.5% PCE is realistically achievable with a 1000 nm thick perovskite with a bandgap of 1.66 eV on top of a both-side textured silicon wafer. By further extending our model to include diffuse light conditions, we are able to predict the energy yield of the configurations under investigation. Our results highlight the beneficial effect of a front-side texture under outdoor conditions, and even more when building integrated photovoltaics (BIPV) are considered. |
In monocrystalline silicon solar cells, randomly distributed pyramids with an angle of 54.7° and a few micrometers in pyramid height and base length are usually implemented for efficient light in-coupling and light trapping. These pyramids are wet-chemically etched into the front- and back-sides of the wafer.8,9 While the front-side texture reduces reflection over the whole wavelength range, the back-side texture enhances photocurrent generation in the near infrared wavelength region. Textured silicon cells have also been used for tandem architectures. The back-side texture does not affect wet-chemical processing of a perovskite cell on the polished front-side. However, until recently textured front-sides were not suitable for the deposition of perovskite layers with spin-coating that gives the highest efficiencies and process control to date. Despite first proofs of concept shown with smooth nano-textures,10 full11 or hybrid5 vacuum perovskite deposition is needed to implement conformal coatings of the perovskite top cell. Alternatively, planar anti-reflection (AR) coatings, such as LiF or MgF2, or textured light management (LM) foils can yield further gains in photocurrents. By introducing a texture on the planar front-side of the device,12–14 the LM foil has two effects: it reduces reflection and enables enhanced light trapping by backscattering the upward propagating light at the LM foil/air interface.14 Being compatible with solution processing, the LM foils are thus perfectly suited for improving light management for this tandem application. Additionally, use of the LM foil allows one to freely choose a texture of any shape, such as nanocavities/microlenses15 or periodic inverted pyramids.16 Having the front- and back-side of the tandem textured differently opens up new routes for further optical optimization.
In Fig. 1 we highlight five monolithic 2-terminal device architectures with different light management strategies [device designs (A, B, C, D and E)]. The main milestones in tandem development with the authors and PCE reached by implementing textures at different interfaces in the devices are also denoted. The simplest device design without any textured interfaces is design A. However, the first monolithic perovskite/silicon tandem was built in early 2015 on a back-side textured homojunction silicon substrate [device (B)] with a PCE of 13.7%.17 Soon, the community turned to SHJ bottom cells due to their superior performance and ease of tandem integration. The first monolithic perovskite/SHJ tandem solar cell was built soon after, utilizing a both-side polished silicon wafer [device (A)] resulting in a PCE of 18%.18 The PCE was soon improved to 21.2% by using a textured LM foil instead of a planar AR coating on top of the device [device (E)] by Werner et al.12 The following tandem cells utilized a back-side textured silicon wafer [device (B)]. With this design PCEs of 20.5% in 201619 and 23.6% in 201720 were achieved. Recently, a cell with a 25.2% PCE was certified; the device was fabricated in a collaboration between Helmholtz-Zentrum Berlin (HZB), University of Oxford and Oxford PV using design (B).4,21 Simultaneously, for the first time a both-side textured silicon bottom cell with a perovskite top cell processed directly on the texture [device (D)] was reported by Sahli et al. with a 25.2% certified PCE as well.5 To fabricate the top cell on a textured substrate, they utilized a hybrid process in which first the inorganic compounds were conformally evaporated and afterwards the organic compounds were spin-coated on top. Interestingly, the combination of a back-side textured device with a textured LM foil on top of the flat front side [device (C)] has not yet been experimentally tested despite promising results on a fully flat tandem device12 and silicon single-junction cells.22 A similar approach has already been investigated in detail for perovskite single junction cells by some of the authors of this paper.13
In parallel, many papers have already investigated a wide range of perovskite/SHJ tandem solar cell designs using numerical optical simulations.23–28 Indeed, improved optics of the tandem devices are necessary to achieve efficiencies above 30%.26,28 Simulations were mostly focused on the effect of different perovskite bandgaps on efficiency,25–27 even for multiple (>2) junctions,27 and the effect of texturing.23,24 The detailed balance limit for a 2-terminal perovskite/silicon tandem states the optimal perovskite bandgap to be 1.73 eV.29 However, this does not include the full device stack. The validation of device simulations for actual tandem solar cells – including realistic optical properties also of the contact layers – by experiments is rarely reported,30 since the experimental and simulation papers are typically not closely related. Therefore, studies on optimal perovskite bandgap and solar cell layer thicknesses can be contradictive; some suggest an optimal perovskite bandgap above 1.7 eV,25,26 while some suggest a lower bandgap around 1.65 eV is sufficient.28,31 Additionally, those simulations assumed planar interfaces and neglected possible optical changes brought about by introducing textured surface designs at the front or rear side. Therefore, it is not yet clear what the perfect bandgap for the perovskite top-cell in different device designs is.
Analysis is usually done for perpendicular, direct illumination, as is the standard in optical simulations and for standard testing conditions. In outdoor applications under realistic weather, spectral and sun position conditions, most of the light does not enter the cell perpendicularly, especially if no axial tracking is applied. Depending on location, the ratio between diffuse and direct illumination spectra also differs.32 This makes yearly energy yield a key parameter in the analysis of photovoltaic device performance. While the first reports are promising, it is not yet clear whether a 2-terminal design can outperform the best single junction in reality.28,33,34 Thus, a more relevant energy yield analysis is necessary that includes light trapping, especially for diffuse incident light or oblique illumination in the morning or late afternoon.
In our contribution, we therefore focus on textured interfaces in monolithic perovskite/SHJ solar cells in both simulation and experiment. First, we fabricate a back-side textured tandem device in a p–i–n top cell configuration and investigate the effect of applying a textured LM foil on top of the flat front of the tandem solar cell. This way we are able to use the high perovskite film quality of the spin-coated perovskite while at the same time still implement efficient light management in the device. To evaluate the potential of our device, we perform optical simulations based on experimentally relevant input parameters. We analyse the tandems with texture at different interfaces, as illustrated in Fig. 1. We optimize the layer stack and perovskite bandgap for each of the designs and estimate the realistically achievable efficiency. By extending our model to include diffuse light in calculations, we are able to estimate the energy yield of all the designs and their improvements due to the texture under realistic environmental conditions. Our results highlight the significance of the front-side texture. We show and confirm that by applying a LM foil on top of a fabricated device, we can keep the benefits of the best performing spin-coated perovskite devices while implementing efficient light trapping. With that, our results give indications of the highest efficiencies obtainable in perovskite/SHJ tandem devices with different device designs and present detailed guidelines on how to reach them.
Fig. 2 Tandem solar cell device schematics of the experimentally realized architecture and SEM cross section image of the top cell with the layers as indicated. |
To improve light management in the fabricated tandem device with a planar front side, we apply a textured light management (LM) foil [device (C) in Fig. 1]. As a master for the LM foil fabrication by an UV nanoimprint lithography process,13 we used a KOH-etched silicon wafer with similar texture as our silicon bottom cell. Consequently, the top surface of the LM foil and our device is the same as it would be for the tandem potentially built on a both-side textured silicon wafer. The 100 μm thick LM foil was processed on a glass substrate and attached on the front-side using an index matching liquid (Norland Products Inc., n = 1.5). This configuration already resembles the module application where the front-side flat device would be encapsulated by glass (albeit thicker – 200 μm vs. 3.2 mm) covered by the LM foil. The LM foil fabrication details can be found in our previous paper.13 The tandem device fabrication process is described in detail in the Methods section.
Fig. 3a shows the EQE and 1-reflectance (1-R) of the best fabricated monolithic perovskite/SHJ tandem devices without and with LM foil as well as their AM 1.5G integrated photocurrent densities JSC_EQE calculated from the EQE spectra. A significant increase in the photocurrent density due to the LM foil can be observed for both sub cells: by more than 2 mA cm−2 for the top cell and by 0.92 mA cm−2 for the bottom cell. This improvement is higher than expected from single-junction cells13 due to a more detrimental air/IZO (nair:nIZO = 1:∼2) interface in the tandem compared to the air/glass (nair:nglass = 1:1.5) interface in the single-junction device. The total reflection is reduced to below 5% in the 300–1000 nm range. Overall, the reflective losses are only 2.48 mA cm−2 in the wavelength range of interest (300–1200 nm), as compared to 6.13 mA cm−2 for the cell without the LM foil. The optimized optical design of our stack is further confirmed by the very low parasitic losses. Less than 8% is lost in other layers for wavelengths between 470 and 1000 nm (<5% between 500 and 900 nm). Only in the UV region the optical losses are higher due to absorption in IZO, SnO2 and C60 layers in the top contact, as well as above 1050 nm by absorption in the back contact. The fabricated solar cell without any light trapping foil is well current matched at ca. 18 mA cm−2. Including the LM foil induces a strong current mismatch of more than 1 mA cm−2 between the top and bottom cell, limited by the silicon bottom cell. Thus there is more efficiency to gain by further optical optimization, which will be discussed in the simulation section.
The corresponding JV characteristics (measured with a mask area ≤ active area) without and with the LM foil are shown in Fig. 3b. The inset in the figure illustrates the illumination conditions. The respective performance metrics are summarized in Table 1. Applying the LM foil leads to a significantly higher short-circuit current density JSC by more than 1 mA cm−2, an increase from 17.3 to 18.5 mA cm−2. The open-circuit voltage (VOC) depends logarithmically on the photocurrent and is thus not affected by the LM foil with a value of 1.76 V. The tandem cell with the LM foil is current mismatched and therefore the fill factor (FF) in reverse scan is potentially enhanced from 76.4% to above 78.5% with the LM foil due to the limiting influence of the Si subcell.39 As a result, the power conversion efficiency (PCE) of 23.4% for the tandem device without light management is increased to 25.5% with the LM foil, in the reverse scan. The measured PCE is among the highest reported values for 2-terminal perovskite/silicon tandem solar cells. Note that the measured JSC using a mask just slightly smaller than the active area (mask area ≤ active area) will underestimate the measured JSC, as can be seen in Fig. S2 in the ESI.† Compared to the value from integrated small spot EQE spectra, in which all of the incident light is captured by the active area, in JV measurements almost 0.4 mA cm−2 is lost by scattering out of the active area due to the textured LM foil. Utilizing a larger mask that enables us to compensate for this effect, as shown in Fig. S3 (ESI†), a higher photocurrent and thus PCE are measured; 19.4 mA cm−2 and 26.5%, respectively. This value serves as an upper limit for the potential tandem efficiency when being integrated into a module with glass encapsulation where light scattering into the active area from outside the cell is enhanced due to backsheet, frame and grid reflection and subsequent light trapping. As slight effects of hysteresis are present, especially for the current matched device without the LM foil, we measured maximum power point (MPP) tracking. The PCEMPP are stable without and with the LM foil for over 5 minutes (Fig. S4, ESI†), displaying a good short-term stability of the fabricated device. For the device with the LM foil, the stabilized power output is very close to the reverse scan.
Illumination condition | Without LM foil | With LM foil (design C) | |||
---|---|---|---|---|---|
Forward | Reverse | Forward | Reverse | ||
(1) mask area ≤ active area | J SC [mA cm−2] | 17.3 | 17.3 | 18.5 | 18.5 |
V OC [V] | 1.76 | 1.76 | 1.76 | 1.76 | |
FF [%] | 75.7 | 76.4 | 78.3 | 78.5 | |
PCE [%] | 23.1 | 23.4 | 25.4 | 25.5 | |
(2) Small beam spot EQE | J SC_EQE [mA cm−2] | Perovskite | 17.95 | Perovskite | 20.21 |
Silicon | 17.96 | Silicon | 18.81 | ||
(3) Mask area > active area | J SC [mA cm−2] | 17.1 | 17.1 | 19.4 | 19.4 |
V OC [V] | 1.75 | 1.76 | 1.76 | 1.76 | |
FF [%] | 77.8 | 78.6 | 76.7 | 77.0 | |
PCE [%] | 23.4 | 23.7 | 26.3 | 26.5 | |
PCEMPP [%] | 23.4 | 26.5 |
The matching between simulations and experiment is shown in Fig. 4. In both cases, without and with the LM foil, an excellent match is obtained. The perovskite bandgap shift matches well with experiment, and interference patterns are well aligned, validating the model and obtained (n,k) spectra. The thicknesses used in simulations for the best match and the corresponding integrated absorption spectra over the solar spectrum JSC_SIM are shown in Table S2 (ESI†). The JSC_SIM values match well with experimental JSC_EQE; the silicon photocurrent density differs by less than 4%, and the perovskite current density by less than 2%, while for reflection no difference is observed. We also simulate the possible improvement if our fabricated device would have been finished by evaporating a LiF layer as an AR coating instead of using the LM foil (see Fig. S6a, ESI†). For our not fully optically optimized device, the improvement with the LM foil is clear, especially in the silicon wavelength range.
We start by optimizing the device (A) based on our experimental layer stack and (n,k) spectra as presented in ref. 25 and 26. We fixed the thicknesses of the layers in the bottom cell and allowed only the layers above the silicon substrate to vary in thickness. In addition, we also altered the perovskite bandgap in 20 nm steps between 0 and 80 nm (0 nm corresponds to the original perovskite CH3NH3PbI3 data with the absorption onset at around 795 nm – 1.56 eV42). For the efficiency estimations we assume a FF of 80% for all the device designs to allow a fair comparison on the optical changes due to textures. For this the optoelectrical quality of the perovskite film and adjacent contact layers should be the same, regardless of whether they are deposited on a planar or textured surface. The following equation was used for the VOC:
Proceeding from these results, we made the following assumptions for the optimization of the devices with textures: even with textures, the layer thicknesses of all the contact layers would stay mostly the same and thus their thicknesses are fixed for the simulations of the different texture positions, only the perovskite bandgap or its thickness would change. A higher bandgap is beneficial due to higher voltage potential, according to the abovementioned equation, if current matching can be realized. A thicker perovskite is beneficial because we can utilize a higher bandgap, however, we restricted the simulation to 1000 nm, as thicker layers are at the moment experimentally very hard to realize with high PCE.
Based on the validated assumptions discussed above, we performed optimizations of textured devices. With the validation, we reduce the investigated parameter space to the perovskite layer only. Hence, we could optimize devices with textured surfaces, which require computationally expensive ray-tracing approaches. We fixed the perovskite thickness to 1000 nm and tuned the bandgap in 5 nm steps until current matching was obtained. In simulations, we consider sharp transitions between layers that show bulk absorption. However, promising results with high PCEs were presented with graded heterojunctions48 and carbon monolayers from graphene49,50 that both may offer possible further improvements. We considered the device designs (A), (B), (C) and (D) as illustrated in Fig. 1. If the device has no LM foil, a planar AR coating (LiF) is considered. Table 2 shows the simulated optimal bandgap, extracted matched JSC_SIM, assumed FF and VOC for all the devices. As expected, the best case is device (D), followed by device (C) and device (B). According to our simulations, a both-side textured silicon tandem device (D) has the potential to reach a PCE of 32.5% for a perovskite thickness of 1000 nm and bandgap of 1.66 eV that would imply to have 1.26 V VOC from the perovskite top cell. Note that reducing the FF of device D to 77.5% due to potential lower perovskite film quality on the texture would mean that devices B, C and D would perform almost the same. Thus developing excellent perovskite film on the texture is necessary to enable the full potential of the double sided textured device D.
Device design | E g opt. [eV] | J SC_SIM [mA cm−2] | V OC [V] | FF [%] | PCE [%] |
---|---|---|---|---|---|
Flat (A) | 1.69 | 19.07 | 2.00 | 80 | 30.5 |
Back-side texture (B) | 1.65 | 20.01 | 1.96 | 80 | 31.4 |
Back side texture + LM foil (C) | 1.66 | 19.97 | 1.97 | 80 | 31.5 |
Both-side texture (D) | 1.66 | 20.56 | 1.97 | 80 | 32.5 |
Interestingly, the exact optimal bandgap to enable current matching highly depends on the device architecture and layers in the stack: recently, for the flat device (A) we found an optimal bandgap of 1.73 eV26 with close to 1.5 μm thickness. However, when utilizing less transparent C60 and IZO top layer (n,k) spectra derived from the measurements of the implemented films, a thickness above 2.4 μm is necessary to obtain current matching, see Table S3 (ESI†). Thus, the optimal bandgap for restricted maximal perovskite thickness of below 1 μm is 1.69 eV for flat tandems [device (A)] as presented in Table 2. When utilizing a back-side texture, the value even reduces to 1.65 eV as significantly more light is absorbed in the NIR regime, which is in agreement with previous work.31 This then has to be compensated for by a smaller bandgap of the top cell in order to achieve current matching, see Table 2. Such bandgaps are already very close to reported compositions with high efficiency and stability.45,46
The effect of the optimal top cell bandgap can be seen in Fig. 5a in which the simulated absorption spectra of all devices are shown for direct incident light as typically available in lab conditions. The most obvious effect is the benefit of using a back-side texture in the bottom cell. Despite having the same morphology on the front side, device (D) enables higher absorption values through the complete spectrum in both sub cells as compared to device (C). This is due to the different refractive index of the first layer (LM foil versus LiF) and different incident angles into the device; the device (C) still has a planar front-side in the top cell stack. Surprisingly, device (C) with a textured LM foil front side is only marginally better than device (B) with a planar AR coating (LiF) on the front. This is mostly due to the nc-SiOx:H layer, which dampens the interferences due to reflections from the silicon35 and reduces the improvement that the LM foil brings. Additionally, a flat front-side leads to a constructive interference at 1050 nm, which is beneficial for the long wavelength response.
We also carried out simulations with perovskite thicknesses of 800 and 1200 nm, which serve as guidelines for a thickness already achievable in efficient devices and as a future possibility, respectively. The results and discussion can be found in the ESI† (Table S5 and Fig. S9). The most important finding is that increasing the perovskite thickness by 20%, only increases the PCE by 0.1% absolute. Thus, with a perovskite thickness of 1000 nm, we can already reach almost the maximal efficiency.
A detailed loss analysis, i.e. absorption spectra for all the layers in the stack, for the best performing device (D) with a 1000 nm thick perovskite is shown in Fig. S10 (ESI†) and the corresponding JSC_SIM values in Table S6 (ESI†). Assuming 100% carrier collection probability, from the 46.23 mA cm−2 available in the spectrum, 41.36 mA cm−2 is converted to charge carriers (20.56 mA cm−2 in the perovskite and 20.80 mA cm−2 in silicon), corresponding to an 89.5% yield of the available spectrum. This is very close to the best silicon single junction devices that yield a carrier generation current of 42.5 mA cm−2.51 The main losses are parasitic absorption in IZO and total reflection, both contributing to a 1.34 mA cm−2 loss. 0.6 mA cm−2 is absorbed by the C60 layer in the front contact and 0.35 mA cm−2 in the middle ITO layer. 1.18 mA cm−2 is absorbed in the back contact layers. Overall, very low parasitic losses are found for this device stack. By comparing the fabricated device and best simulated device (C) the biggest difference is in photocurrent utilization due to the non-optimized perovskite bandgap and thickness in the fabricated device (Fig. S10, S11 and Table S6, ESI†). The reflection is further reduced due to the both-side textured design, while IZO losses are lower due to a 40 nm thinner layer. Other losses are comparable. This confirms excellent optical performance of the monolithic perovskite/SHJ tandems utilizing contact layers from our experiment. To further reduce losses in the UV and blue region, the most viable options are developing a more transparent electron selective material with lower absorption than C60 (e.g. PCBM, however, its evaporation is not easy52) and using a less absorptive front TCO material such as IO:H.53,54 However, high temperature annealing of the IO:H at 150–200 °C to obtain optimal properties is an obstacle the perovskite film would have to survive. Additionally, the IO:H would also reduce free carrier reflection and parasitic free carrier absorption in the NIR. The losses in the back-contact could also be reduced by e.g. applying silicon nanoparticles as a back-side reflector.20
Until now, we have only considered direct, perpendicular illumination. However, to fully evaluate the effect of implementing textures in tandem cells, we need to move closer to outdoor conditions where diffuse light contributes a significant portion of the overall irradiation. To obtain absorption spectra under diffuse illumination, we repeated the simulations, but this time we assumed fully diffuse (Lambertian) incident illumination.14 The spectra with simulated diffuse illumination are shown in Fig. 5b. Compared to previous simulations under direct, perpendicular illumination, here a clear improvement with the LM foil can be observed in the full wavelength range of interest. This is a result of a more advantageous average incidence angle of diffuse light into device (C) as compared to device (B).
V OC and FF under different light intensities were measured on the fabricated tandem cell without the LM foil by altering the LED sun simulator intensity, starting at 100% AM1.5 (100 mW cm−2) and then sweeping from 10% AM1.5 to 120% AM1.5 in 10% steps. The results are shown in Fig. S12 (ESI†); in EY calculations they were then normalized at 100% light intensity to the predicted VOC estimated from the bandgap (see Table 2). As expected, the JSC increases linearly with increasing intensity and VOC logarithmically. The FF is slightly higher at lower intensities (lower current causes less series resistance induced power loss in the front contact), however, its value is relatively constant between 76 and 78%. The current for both 100% illumination measurements, first and eleventh, are the same, while VOC and FF improve slightly. This is not uncommon, a similar light soaking effect was recently reported albeit for a different hole transporting material – NiO.55 These results show that perovskite/SHJ devices can function well even at low irradiance levels.
We calculated the EY for three locations in the United States of America – Washington D.C., Golden and Phoenix. Their meteorological data is provided by NREL as a typical meteorological year (TMY).56 Averaged global horizontal irradiation (GHI) and direct normal irradiation (DNI) spectra over a year of each location are shown in Fig. S14 (ESI†), and the yearly incident irradiation for each location in Table S8 (ESI†). The locations were chosen in such a way that they receive different amounts of solar irradiance and the ratio between diffuse and direct illumination varies greatly. We investigated two tilts of the solar cells (modules): 30° and 90° facing south in the northern hemisphere. These tilts are relevant for rooftop or power plant, and building integrated photovoltaics (BIPV), respectively. The yearly EY for each of the discussed configurations for the two tilts are shown in Fig. 6 and all the corresponding values can be found in Table S7 (ESI†). The colored bars indicate the yearly EY in kW h m−2 and the dark blue bars the relative improvement compared to the flat device (A) without any texture. Phoenix, with the highest amount of solar irradiation, yields the highest amount of energy, almost 50% more than Washington D.C. for the 30° tilt. Interestingly, for the 90° tilt all the designs would produce more in Golden than in Phoenix. This is due to a higher part of diffuse illumination (see Table S8, ESI†), while the contribution from direct illumination is lower for the 90° tilt due to a greater direct incidence angle onto the device on the wall, contributing less than for the 30° tilt. Analyzing the yearly EY, a more pronounced improvement using LM foil compared to a planar AR coating is observed than in the basic simulation with only perpendicular illumination. This is even more apparent for the BIPV application (90°) where with the LM foil more than 4.5% extra energy can be gained. Overall, device (D) is still the most promising configuration with ∼9.5% and ∼12% relative improvement to the fully flat device (A), for the 30° and 90° tilt, respectively. Assuming again that the FF of device D is 77.5 due to slightly poorer perovskite film or contact quality on the texture, 15 to 20 kW h m−2 would be lost. This makes the relative difference between devices C and D less than 1%. In addition the glass encapsulation scheme as typically used in modules is already partially already included in device C, thus the herein calculated benefit of device D is even less pronounced. Nevertheless, these results show that strong improvements in EY are expected for textured front-sides, even when just applying a textured LM foil on top of a planar wafer.
The textured device optimization was done using CROWM.14,15 The simulator is based on combined ray and wave optics models that enable simultaneous simulations of all segments of the device; the thick textured LM foil, the silicon wafer (incoherent light propagation assumed in both cases), and the thin-film solar cell stack(s) (coherent light propagation assumed). As the input parameters, realistic thicknesses and experimentally determined refractive indices of the materials were employed. The imaginary part of the refractive index of the perovskite absorber was wavelength-shifted to obtain the different bandgaps. The main outputs of the simulator are total reflectance, transmittance and absorptance in each layer. Their solar-spectrum wavelength integration equals the generated JSC or the equivalent JSC loss in each individual layer. The simulations were carried out in the wavelength range from 350 to 1200 nm which is a sufficiently broad range for the analyzed tandem solar cells. To include a realistic texture in the simulations, the texture profile of the random pyramids was obtained using atomic force microscopy (AFM) and imported directly into the simulator.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c8ee02469c |
This journal is © The Royal Society of Chemistry 2018 |