Nanocellulose and Nanochitin Cryogels Improve the Efficiency of Dye Solar Cells

: Biobased cryogel membranes were applied as electrolyte holders in dye solar cells (DSC) while facilitating carrier transport during operation. They also improved device performance and stability. For this purpose, cellulose nano ﬁ bers (CNF), TEMPO-oxidized CNF (TOCNF), bacterial cellulose (BC), and chitin nano ﬁ bers (ChNF) were investigated. The proposed materials and protocols for incorporating the electrolyte, via simple casting, avoided the typical problems associated with injection of the electrolyte through ﬁ lling holes, a major di ﬃ culty especially in manufacturing large area cells. Owing to the fact that cryogel membranes did not require any ori ﬁ ce for injection, they were e ﬀ ective in minimizing leakage and in retaining liquid electrolyte. The results indicated the reduction of performance losses compared to conventional electrolyte ﬁ lling, likely due to the better spatial distribution of electrolyte. DSCs based on BC cryogels had an initially higher performance and similar stability compared to those of the reference cells. When compared to reference cells, CNF and ChNF cryogels produced higher initial performance, but they underwent a faster degradation. The di ﬀ erence in stability was attributed to the e ﬀ ect of residual components, including lignin in CNF and proteins in ChNF, as demonstrated in bleaching experiments. TOCNF indicated a relatively poor performance, most likely because of residual aldehydes. Overall, we o ﬀ er a comprehensive evaluation based on current − voltage (IV) pro ﬁ les under simulated sunlight, incident photon-to-charge carrier e ﬃ ciency (IPCE), electrochemical impedance spectroscopy (EIS), and color image processing, together with accelerated DSC stability tests, to unveil the e ﬀ ects of new membrane-based assembly. Our results give guidelines for future developments related in particular to the e ﬀ ects of the tested biomaterials on device stability.


■ INTRODUCTION
Since their introduction in 1991 by Graẗzel and O'Regan, 1 dye solar cells (DSCs) have been developed steadily and have achieved a power conversion efficiency up to 13%. 2 Aiming of improving their efficiency and manufacture (cost and simplicity), DSCs have been assembled from a variety of materials and methods.However, DSC technology has not yet reached a large manufacturing scale.−5 Among these, cellulose is an intriguing option to reduce the environmental impacts of DSCs, both in terms of component life cycle and direct factors, such as toxicity, renewability, disposal and degradability.Even though long lifetimes, as long as 25 000 h, have been reported for DSCs, 6 the key challenge is to maintain such performance while using low-cost materials that are roll-to-roll compatible. 7,8 an attempt to meet the demands discussed above, various synthesized electrolyte fillers and gelators that ideally could render the electrolyte printable have been proposed. 9,10In contrast, herein, we investigated four biomaterials that can serve as an electrolyte-holding membrane to simplify the manufacturing of these devices and, simultaneously, to eliminate possible leakage pathways (mainly related to the electrolyte filling holes).This latter aspect was considered important in achieving long-term stability, as discussed in our earlier contribution where we demonstrated the concept of utilizing nanocellulose aerogel for electrolyte filling. 11Moreover, when injecting the liquid electrolyte into the cell assembly, the nanoporous photoelectrode may act as a filter preventing even distribution of electrolyte components. 8This results in significant efficiency losses, which cannot be addressed by simply changing the electrolyte composition. 12e propose biobased cryogel membranes in DSCs through which the electrolyte can be directly applied before sealing the cell.In turn, this prevents the "filtering" or segregation effects that occur when the electrolyte is injected or forced through the cell, as illustrated in Figure 1.
Four different biobased cryogel membranes were tested: bacterial cellulose (BC), cellulose nanofibers (CNF), TEMPOoxidized CNF (TOCNF), and chitin nanofibers (ChNF).Similar materials were tested for quantum dot solar cells in our earlier work. 13In practice, these materials comprise nanofibrils that form strong networks resulting in robust membranes, the properties of which are only affected by the source material and method used in the isolation of nanofibrils.CNFs were sourced from wood fibers by using top-down mechanical deconstruction.TOCNF was created similarly after oxidation of the fibers using a catalytic process (TEMPO oxidation) that introduces negative charges in the nanofibers and enhances their colloidal stability (increased electrostatic repulsion and dispersion).In contrast to CNF and TOCNF, BC was obtained by a bottom-up biogenesis via a microorganism that aerobically produces cellulose nanofibers.The membrane grown by the bacteria was freeze-dried to obtain a cryogel. 14inally, ChNF is similar in its fibrillar structure to CNF but consists of chitin, a different polysaccharide, that was extracted from marine organisms (shrimp shells, specifically).
In this study, whole DSC cells were prepared using biomaterial membranes to hold the electrolyte.Their performance and stability were compared to reference DSCs, namely, those filled by using the conventional injection method.Particular attention was paid to how the different materials affected the DSC long-term stability−for instance, the cell configuration was selected to allow thorough investigation of the electrolyte membrane.Beyond the selected application, stability data is relevant since it gives an indication of the suitability of these materials in DSCs.The application range of the type of materials used here is very wide.They can be used as DSC substrates 3−5 or as gelators of the electrolyte, 15 demonstrating the possibility of achieving functions normally assigned to synthetic polymers.Chitin has also been used in several applications in DSCs, including gel electrolyte compositions. 16Furthermore, chitin imparts the red color typical of the exoskeleton of lobsters and shrimps, and its derivative, chitosan, has been used as a dye in DSCs. 17ecently, chitosan has also been used as a gelating agent for water-based DSC electrolytes. 18egradation is one of the major hurdles for emerging photovoltaic technologies.A critical deficiency in the field of emerging photovoltaics is that information about cell degradation pathways is very limited due to the usual practice of publishing only good stabilities. 19,20Pushing the solar cells beyond their stable performance would, however, give more insight about failure mechanisms and is thus encouraged. 19,20o enable effective testing of the proposed new materials for DSC assembly, including assessment of their stability, we aimed to age the solar cells in conditions that would reveal the difference between materials.The challenge is that, in usual testing with visible light or full spectrum with UV filters, the testing is expected to take several thousand hours (i.e., several months). 21One option to accelerate the testing would be to utilize extremely high overall light intensity (≫1 Sun), 22 which does work for individual devices but does not scale easily for large series of devices.Instead, we opted to test the cells in harsh UV light to accelerate the degradation.As demonstrated in our previous study, the degradation rate appears proportional to the amount of UV light, 23 which in the current case helps to differentiate the performance in a shorter time.The primary performance loss mechanism in these type of studies was found to be the degradation of the charge carriers in the electrolyte 21,24 due to the photocatalytic reactions at the TiO 2 interface.

■ EXPERIMENTAL SECTION
Isolating Nanofibers and Preparing Cryogel Membranes.Bacterial cellulose (BC) was grown in Erlenmeyer flasks using a BCproducing bacterium, Komagataeibacter medellinensis.The BC was harvested from the air−liquid interface after 2 days of incubation at 28 °C.The BC pellicles were removed and purified using 0.1 M NaOH (60 °C) for 2 h, followed by washing with water.CNF was obtained from never-dried bleached birch pulp.First, an aqueous dispersion of 2% solid content was prepared followed by fibrillation through a highpressure M-110 P fluidizer (Microfluidics corp.Newton, MA, U.S.A.) for 6 passes.The method for tempo-oxidized cellulose nanofibril (TOCNF) preparation is detailed in our previous report. 25Purified chitin flakes from shrimp shells, with the degree of acetylation (DA) of ∼72%, were purchased from Sigma-Aldrich.For preparation of chitin nanofibers (ChNF), chitin flakes were dispersed in HCl solution to reach 1% at pH 4, then they were fibrillated in a microfluidizer for six passes.
For cryogel membrane preparation, 40 mL of 0.1% dispersions of CNF, ChNF, and TOCNF were dispersed with a tip sonicator (Branson 450 EC sonicator) for 10 min at 10% amplitude.The dispersions were vacuum-filtered through a hydrophilic polyvinylidene fluoride filter (0.45 μm, GVWP, Millipore) until a hydrogel cake was formed.The hydrogel cakes were transferred to a Petri dish, where the filter papers were removed with a previously described method. 26ubsequently, the free-standing gel-cakes were subjected to solvent exchange with ethanol and tert-butanol (both for three times).Next, the gel-cakes were cooled at −80 °C and freeze-dried with Labconco system (74200 series).BC cryogels were prepared directly from the wet BC membranes using solvent exchange and freeze-drying.
All of the tested cryogel membranes were dried overnight in a vacuum oven at 80 °C.Pieces of membranes were cut for use within the cells with an Epilog 35W Zing-laser scriber.The thickness of all the membranes was in the range of 50 μm.
Cell Structure and Assembly.Five groups of DSCs were assembled: the reference cells were assembled in the absence of the cryogel membranes (similar to our earlier work 27 ), and the other four groups included the respective cryogel membranes in the DSC.
The cells were sealed between glass substrates, which were made from fluorine-doped tin oxide (FTO) glass (TEC-15, Pilkington).In the reference cells, the counter electrode glass substrates had two small holes to allow liquid electrolyte injection into the cells.The cells with cryogels did not require holes.The cleaning of the glass substrates had four steps: it started with a rinsing with water, then washing with dishwashing detergent, followed by final rinsing with ethanol and then acetone.The substrates were also UV-treated for 15 min (UV/ozone ProCleaner, Bioforce nanosciences) before use.
The photoelectrode substrates were TiCl 4 -treated by immersing the glass plates for 30 min in a solution of titanium(IV) chloride tetrahydrofuran complex (1 wt %) in deionized water at 70 °C.The purpose of the treatment was to decrease the edge potential of the photoelectrode conduction band and to decrease the rate of recombination between the electrolyte and electrons in the photoelectrode. 28The TiO 2 photoelectrode itself was deposited on the treated photoelectrode substrate by screen printing (AT-60PD, ATMA).The TiO 2 layer consisted of three layers: the first two layers which form the absorbing part of the photoelectrode were printed with a paste which had TiO 2 of small particle size (Dyesol, DSL 18NR-T).The last layer had large, light scattering particles (Dyesol, DSL WER2−0).The complete photoelectrode was roughly 12 μm thick and with an area of 0.4 cm 2 .To finalise the TiO 2 layers, they were first sintered in 450 °C for 30 min, next they were TiCl 4treated again, followed by another sintering.The sensitization of the TiO 2 layers was performed by immersion in a dye solution overnight.The dye solution consisted of 0.2 mM of Z907 (Dyesol) in 1:1 solvent of acetonitrile (ACN) and tert-butylalcohol (TBA).
A platinum catalyst layer was formed on the counter electrode by pipetting 4 μL of 5 mM H 2 PtCl 6 in 2-propanol on the substrate.The counter electrodes were thermally treated after the catalyst deposition at 390 °C for 20 min.The electrolyte consisted of 0.5 M 1-methylbenzimidazole (NMBI), 0.5 M 1-propyl-3-methylimidazolium iodide (PMII), 0.1 M guanidinium thiocyanate (GuSCN), and 0.1 M I 2 in 3methoxypropionitrile, prepared and purified similarly to an earlier work. 23o seal the reference cells, the two electrodes were pressed together, separated by a frame of Surlyn 1702 ionomer resin film (20 μm thickness, DuPont).The cell was then heated, which melted the Surlyn and sealed the cell.Then, in the case of reference devices, liquid electrolyte was injected into the space between the substrates through the holes in the counter electrode, which were then sealed with a microscope glass by melting another layer of Surlyn.
The cells with biomembranes were assembled by first attaching the Surlyn frame foil to the counter electrode by heating it on a hot plate pressed against a Teflon piece.The cryogel membrane was placed on top of the dyed photoelectrode.Then, 3.5 μL of electrolyte was dropped on the photoelectrode with a micropipette (the reference cells were assembled with the same 3.5 μL amount of injected electrolyte to keep the cell groups comparable).The electrolyte spread evenly in the cryogels and there was only minimal leakage.Finally, the cells were sealed by pressing the counter electrode with the attached Surlyn foil against the photoelectrode on a hot plate.Unlike the reference cells, no holes were needed for electrolyte injection in the counter electrode substrates.To have enough space to fit the cryogel membranes inside the cells, two Surlyn frame foils were stacked to increase the distance between the electrodes to about 50 μm.
Contacts were made by attaching copper tapes to the electrodes; the electrical contact was improved with a layer of silver ink (SCP, Electrolube) between the substrates and the copper tapes.Finally, the whole contact was covered with a layer of epoxy glue.The cell geometry was selected so that it enabled detailed investigation of degradation, for example, by tracking electrolyte color.Note that the geometry (for example, via placement of contacts) does affect the cell efficiency, but we did not attempt to optimize such aspects.
DSC Performance.The cells were characterized using multiple techniques that revealed details of the DSCs operation.To investigate the reproducibility, 4−6 replicates with each different cell type were made (N indicated in Table 1), and the values reported in the figures are averages, the error bars representing standard deviation.A xenon lamp solar simulator (Peccell PEC-01) was used to measure the current−voltage (IV) characteristics of the DSCs.The simulator lamp produced a good match to the standard AM1.5G-spectrum.The light was calibrated to 1000 W/m 2 with a PV Measurements Inc. Si KG5 photodiode.The cells were measured at room temperature while covered with tape masks to prevent reflected light from affecting the The values represent the averages and errors represent standard deviations.
measurement.The potentiostat was a Keithley 2420 3A Sourcemeter.The set voltage range for the DSC measurements was from −0.3 to +0.8 V with a step size of 0.01 V and a delay of 0.1 s before logging each data point.Incident photon-to-charge carrier efficiency (IPCE) was measured with QE/IPCE Measurement system QEX7 (PV Measurements Inc.).The wavelength range used was from 300 to 1000 nm, without bias light.
Electrochemical impedance spectroscopy (EIS) measurements were done with a Zahner Zennium potentiostat.The AC frequency range used was from 4 MHz to 100 mHz.The measurements under illumination were performed in open circuit conditions.The amplitude of the voltage modulation was 10 mV.Equivalent circuits were fitted to the data with ZView2 (Scribner Associates, Inc.).The equivalent circuit model used in this study is given in literature. 29n Olympus E-620 camera was used to photograph the cells weekly with the purpose of monitoring chemical changes within the cells optically. 24The photographing setup was protected from ambient light with black canvas, and 4 LEDs attached to a movable tray with the camera provided a constant illumination conditions within the chamber.The white balance and color profile of the pictures was defined with a color checker passport (X-Rite) using Adobe Lightroom3.The average RGB values, which indicate changes in the cell, were calculated from the photographs with MatLab by selecting a small area from the cells for analysis.
DSC Stability.Electrolyte bleaching is a major cause for degradation of dye solar cells, as reported by numerous groups. 30−33 Note that even the best UV filters will leak some UV light.To efficiently compare the different biomaterials, we adopted accelerated tests at very harsh conditions (>1 Sun UV light level without filter).
The UV aging was performed in a weather chamber Arctest Arc-150.The temperature of the chamber was set to 12 °C, which rendered cell temperatures of 50 °C, given the heating effect of the UV lamps (Osram Ultra-Vitalux 300 W).The cells in the UV tests were aged in open circuit conditions.The spectrum of the employed lamp had 1.5 times the amount of UV light compared to the 1 Sun standard, but equivalent to 34% of one Sun intensity in the visible region.

■ RESULTS AND DISCUSSION
Cryogel Membranes for Electrolyte Holding. Figure 2 shows that the morphology of the biobased cryogels comprised randomly oriented, interconnected fibrils.BC cryogels formed distinctive three-dimensional networks of highly individualized fibrils (Figure 2a).As compared to TOCNF (Figure 2c), CNF (Figure 2b) consisted of individual nanofibers as well as bundles with lower degree of fibrillation.Due to TEMPO oxidization treatment, TOCNF were subjected to more extensive fibrillation and finer fibril sizes were observed.In case of the ChNF (Figure 2d), intermediate fibril sizes were observed after protonation of aminoacetyl groups of chitin in acidic medium.The porous network of the cryogels benefited from the sequential solvent exchange that lowered the surface tension (water to ethanol to tert-butanol), which significantly reduced the capillary forces prevalent during drying. 34hotovoltaic Performance upon Aging.Initially, the majority of the DSCs tested reached an efficiency η that was higher than that of the respective reference cell (Figure 3 and Table 1).Only TOCNF resulted in lower initial η through a low photocurrent I SC and fill factor FF (Figure 3).When the TOCNF membrane came into contact with the electrolyte, small black areas immediately appeared on its surface.This is explained by a reaction between the electrolyte and the membrane, involving residual aldehydes that were formed during TEMPO oxidation.Thus, TOCNF was omitted from further analyses and the focus was placed on the other biomaterials that produced efficiencies exceeding that of the reference cells.It should be noted that the cell geometry was selected to allow detailed investigation of cell degradation.Specifically, we monitored changes in the electrolyte color.For this purpose, an area inside the cell was left inactive, thereby reducing the relative cell efficiency.For the purpose of understanding the relative differences in performance and to gain insights on how the tested biomaterials affected DSC stability, this was a justified approach though the absolute efficiency value was underscored.
The higher initial efficiency of the biomaterial cells compared to the reference was due to a higher I SC : on average, I SC of ∼12 mA/cm 2 was measured for the BC and ChNF cells, while for the reference cells only ∼9 mA/cm 2 (Table 1).The difference in I SC is most likely not related to the cryogels themselves, but it is rather a result of the electrolyte distribution upon filling.In the case of the liquid reference cells, the electrolyte is injected into the space between the two electrodes.As the cell is filled, the liquid electrolyte flows on  the nanoporous TiO 2 layer, which produces component segregation or gradients (a effect).This causes accumulation of electrolyte additives to the side of the photoelectrode nearer to the electrolyte injection hole resulting in an uneven spatial performance distribution and consequent decrease of the overall performance. 35Especially, the accumulation of the voltage-increasing agent (in this case NMBI) on the photoelectrode is expected to reduce the current of the cell by restricting electron injection. 12This is supported by the fact that the current of the reference cells was initially low, but their open circuit voltage V OC was higher than that of the other cells (Figure 4).In contrast, the electrolyte was inserted into the cells with cryogel membranes by placing it directly on top of the photoelectrode followed by sealing.This method reduced the electrolyte transport distance by 3 orders of magnitude given the difference in flow direction (inplane versus out-of-plane).More specifically, in the case of cryogel membranes 50 μm was the characteristic distance for penetration in the out-of-plane direction while in the conventional injection it is 10−20 mm in the in-plane direction (Figure 1).As a consequence of the smaller characteristic diffusion length in the cells with cryogel membranes, the electrolyte segregation of filtering effect is eliminated, resulting in a higher initial efficiency. 12he performance over time shown in Figure 4 indicates that the BC and reference cells increased their I SC at the beginning of the aging protocol, whereas the ChNF and CNF cells showed a decreased I SC already from the start (Figure 4a).I SC of the BC and reference cells started to decrease at similar or slightly lower rates compared to that of ChNF and CNF cells.The fact that the reference cells improved the most in the first hours may be related to a time-dependent diffusion of diverse electrolyte components at the surface of the photoelectrode leading to a more homogeneous spatial distribution.Note that this effect, however, may not apply to larger cells, since previous studies have shown that the uneven performance persisted over time. 12,35ring the aging, the bacterial cellulose (BC) cells slightly improved initially, and then closely followed the aging process of the reference cells.In fact, all of the IV parameters of BC cells were very close to the reference cells, which is very promising for the viability of BC as an electrolyte membrane for DSCs.
V OC remained relatively stable during the aging in all of the cell groups, in particular after the first 50 h, decreasing from 0.66−0.72 to ∼0.64 V in the first two measurements and plateauing for the rest of the aging (Figure 4b).FF, likewise, performed similarly in all of the cell groups: it increased for all of the cells from ∼0.5 and stabilized in the range 0.7 to 0.8.The FF was not smaller due to smaller resistive losses−in contrast, the overall resistance comprised of all seriesconnected resistive components increased in all the cells (Table 1).Thus, the increased FF was likely due to the smaller photocurrent; with the decreased current, the effect of resistive losses was more limited.The changes to the resistive losses were investigated via EIS measurements, as described in later sections.Even though FF increased, the loss of I SC was so substantial, by comparison, that it had a dominant role in η.As a result, the aging behavior of the cell efficiencies was similar to the behavior of the short circuit currents (Figure 4).
The aging results demonstrate interesting differences between the biomaterials used, which acted as electrolyteabsorbing sponges in the cells.As indicated in Figure 4, when taking into account cell-to-cell variation, the BC cells follow the degradation trend of the reference cells, whereas the differences between the three different biomembranes are statistically significant.The long-term performance increases in the order CNF < ChNF < BC.The lower performance of the CNF cryogels relative to the reference cells may be explained by the contribution of the precursor fiber type, composition and processing method, which all play a crucial role for the long-term stability of the solar cells.One essential factor is the presence of residual lignin.To investigate the effect of lignin further, we tested various kraft and organosolv lignin nanoparticles (of diameter from 200 nm to 1 μm).It was found that lignin caused the electrolyte to turn brown in color, and the cell displayed a poorer performance compared to reference cells.Also, they degraded much faster compared to the reference glass cells (data not shown).Thus, it appears that lignin forms a complex with the iodine electrolyte and thus blocks the charge transfer agents in the cell.These observations support the fact that BC, which is lignin-free, performed the best and therefore is the most appealing biomaterial out of the studied options.
When the biomaterial cells had been aged under UV light for 216 h, the efficiency of all of the cells had decreased to <1% and the aging study was ended at that point.As expected, the UV light degraded the cells very quickly, revealing the differences between the cells in a short period of time.In a conventional light soaking experiment with a UV filter, the reference cells used here exceeded a lifetime of 1000 h and had an expected lifetime of several thousand hours. 36itial DSC Photocurrent.The initial photovoltaic measurements showed significant variability of I SC for the different cell types (Table 1).The variation in the photocurrent was investigated with IPCE measurements (note: the end results are missing due to device malfunction).The IPCE spectra of the initial state (Figure 5) showed similar profiles for all cell groups while the maximum IPCE value varied.The maximum IPCE at a wavelength of ∼550 nm was 70% for BC and ChNF cells while for the CNF and reference cells it was 10%-point lower.This is in good correspondence with the initial IV results: the initial I SC was highest for BC and ChNF cells and lowest for CNF and reference cells.In the case of the reference cells, this is, once again, likely related to the heterogeneous electrolyte filling in the cell.As mentioned before, the injection filling method caused voltage-increasing electrolyte additives to adsorb on the photoelectrode.These additives increase the voltage of the cell, which simultaneously  Curiously, the IPCE of CNF cells was lower compared to the other cryogel cells.Given that the filling method was the same for all cryogel cells, the most likely explanation is that the CNF itself and residual components such as lignin were detrimental to the photocurrent production.
Resistive Components.The photovoltaic analyses initially showed a variance in FF between cells; it also increased significantly during aging.EIS measurements can be used to distinguish changes in individual resistive components of the cell and thus it can help to understand the origin of the FF variations.Analysis of the resistances during aging can shed some light on whether the increasing FF was merely an effect of decreasing photocurrent or functional changes in resistive components.The EIS was measured three times during the aging of the cryogel-based DSCs, in the beginning, at the midpoint (around 100 h), and at the end (220 h).
Ohmic series resistance Rs, which is mainly caused by the resistance of the conducting substrates and the current collector contacts, remained close to the initial value or even decreased during the aging for all of the cell groups (Figure 6).Initially R S was close to 20 Ω, except for the reference cells, which had a slightly higher value.Eventually R S stabilized between 16 and 19 Ω in all cells.
Similarly, the aging caused only small changes to the charge transfer resistance (R CE ) of the electrolyte and counter electrode interface (Figure 6).Initially, the reference cells had only a slightly lower R CE compared to the biomaterial cells−this indicates that the cryogels were not physically blocking the catalyst layer of the counter electrode.R CE increased slightly in all the cells from 4−6 to 6−8 Ω (Figure 6).The increase was largest for the CNF and ChNF cells.While some changes occurred in the R CE , they were not very substantial and would only cause a marginal degradation of the cell performance.Since R CE is related to both the counter electrode and the electrolyte, changes in either of those could cause R CE to increase.For instance, in previous studies charge carrier losses have been shown to reduce R CE 23,31,33,36 − which is likely also the case here, which will be discussed later.
Among the series connected resistances, the electrolyte diffusion resistance R D increased the most during the aging.For all of the cells R D was initially only in the range of few ohms, but it reached ∼50 Ω for the reference and BC cells by the end of the aging.These values were 70 Ω for the ChNF cells and 130 Ω for CNF cells (Figure 6).R D is proportional to the amount of charge carriers in the electrolyte 29 and thus the 10-fold or larger increase in R D suggests a > 90% loss of charge carriers.During aging R D becomes the largest resistance component in these cells and causes the final overall resistive losses to be larger than initially.Thus, the increased FF during the aging in Figure 4 is a consequence of decreased I SC , instead of improved internal resistance.Overall, the decrease in I SC was large enough to compensate the losses caused by the increased series connected resistances.
Bleaching Effects in Charge Carriers.The photovoltaic data showed I SC reduction in all the cells.By visual inspection of the cells, a clear loss of electrolyte color was evident.The electrolyte color originates from the tri-iodide charge carriers and the loss of color is related to the loss of tri-iodide.The loss of charge carriers is a major loss mechanism in DSCs and it has been a dominant cause for degradation.The color change can be quantified from digital photos whereby the loss of the yellow color is directly proportional to the increase of the blue value in the RGB color space.Figure 7 shows that the bleaching was fast and severe for all the cells: the color of the electrolytes had reached their maximum value after only 96 h and, in the case of the reference cells, the magnitude of the blue pixel value more than tripled.The bleaching is very fast due to the high UV light intensity, which was used in the accelerated tests.The charge carriers continued to degrade after the values had stabilized, but the imaging technique was not reliable enough in detecting color differences at low concentrations of iodine. 24Likewise, the cells of the different groups (i.e., reference, BC, ChNF, and CNF) were not fully comparable against each other since the cryogels alter the color of the electrolyte they hold (Figure 7a).
Despite the different kinds of electrolytes in all of the cells, they bleached at similar rates, the difference in the blue pixel values are noted (Figure 7b) and all the cells displayed overlapping data for the first 100 h.The final levels of blue pixel values reflected the optical differences caused by the presence of the different biomaterials.The similarity in color degradation shown in Figure 7b demonstrates that in the presence of the cryogel membranes the UV light did not have significant or unexpected effects on the aging of the iodine electrolyte.When looking at R D values with EIS (Figure 6), the CNF was found to degrade more quickly than the other electrolytes, which is not as clear in Figure 7b.However, the initial cell color for CNF differed the most compared to the other cells.

■ SUMMARY
The DSC assembly process was enhanced by introducing electrolyte-soaking biobased cryogels between the electrodes of the cell.The biomaterial membrane absorbed the liquid electrolyte and held it in place during the assembly.Compared to a standard reference cell, an efficiency improvement of 44% was achieved with membranes prepared from bacterial cellulose, which maintained its performance in the UV light stability tests.ChNF and nanofibrillated cellulose, likewise, absorbed the electrolyte and increased the initial efficiency.However, the onset of degradation for these cells occurred earlier compared to the bacterial cellulose and reference cells.TEMPO-oxidized cellulose nanofibers reacted immediately with the electrolyte and resulted in a significantly decreased current and efficiency.
The EIS and photoimaging analysis indicated that there was a severe loss of charge carriers in all of the cells.The loss rates were slightly different, with the BC and the reference cell performing the best.The main reason why the other membranes did not give as high performance and stability as BC is likely due to residual components in the biomaterials, including lignin in the case of CNF and proteins in the case of ChNF.In particular, lignin was shown to negatively affect the operation of DSCs.
Overall, this study demonstrates that the use of membranes based on bacterial cellulose cryogels is a viable method to improve DSC fabrication.This would be especially useful for large-area modules, where the injection of the electrolyte is a challenge. 35The results gained here indicate the initial suitability of the different biomaterials for DSC preparation, as well as their long-term stability.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acssuschemeng.8b06501.
IV curves before and after aging; equivalent circuit used in the EIS analysis and examples of measured and fitted data; example photographes used for the bleaching analysis (PDF)

Figure 1 .
Figure 1.Conventional electrolyte filling method makes the electrolyte flow through the entire cell horizontally (10−20 mm distance) (left).By using an electrolyte membrane (right), the electrolyte can wet the photoelectrode vertically, reducing the diffusion length to 50 μm−the distance electrolyte travels decreases 3 orders of magnitude.

Figure 3 .
Figure 3. Initial average IV curves of each of the studied DSC type.The colored bands around the averages signify the standard deviation between different samples.ToCNF has only one sample and therefore no deviation.

Figure 4 .
Figure 4. Time-evolution of DSC parameters for each cell type in the aging test.The measurements were performed three times a week.The values represent the averages and errorbars indicate the standard deviation.

Figure 5 .
Figure 5.Initial IPCE results of the DSCs.The figure on the left shows the measured IPCE values.The error bars represent the deviation between different cells prepared with the same cryogel.The figure on the right displays the same data normalized by setting the maximum of each IPCE curve to 1 (in this comparison ref and CNF overlapped as well as BC and ChNF).

Figure 6 .
Figure 6.Changes in the internal resistances of the cells assembled with the cryogels.The resistance values were obtained by fitting to the Nyquist curves obtained from EIS measurements.

Figure 7 .
Figure 7. Electrolyte color was tracked with photographs and the bleaching of the electrolyte shows as increase in the blue pixel value.(a) The absolute blue pixel values of the cryogel cells.Since the cryogels themselves affect the coloration of the electrolyte, the values can not be directly compared with the reference cells.(b) The difference in the blue pixel value as compared to the original state of the device.

Table 1 .
List of Photovoltaic Parameters for Cells with the Different Cryogel Membranes before (Initial) and after the Aging Test (216 h) a