Magnetron Sputtered Electron Blocking Layer as an Efficient Method to Improve Dye-Sensitized Solar Cell Performance

: The main efficiency loss is caused by an intensive recombination process at the interface of fluorine-doped tin oxide (FTO) and electrolyte in dye-sensitized solar cells. Electrons from the photoanode can be injected back to the redox electrolyte and, thus, can reduce the short circuit current. To avoid this, the effect of the electron blocking layer (EBL) was studied. An additional thin film of magnetron sputtered TiO 2 was deposited directly onto the FTO glass. The obtained EBL was characterized by atomic force microscopy, scanning electron microscopy, optical profilometry, energy dispersive spectroscopy, Raman spectroscopy and UV-VIS-NIR spectrophotometry. The results of the current-voltage characteristics showed that both the short circuit current (Isc) and fill factor (FF) increased. Compared to traditional dye-sensitized solar cell (DSSC) architecture, the power conversion efficiency (n) increased from 4.67% to 6.07% for samples with a 7 X 7 mm2 active area and from 2.62% to 3.06% for those with an area of 7 X 80 mm2.


Introduction
A fter approxim ately 30 years of research, dye-sensitized solar cells (D SSCs) have m aintained their strong importance among scientific and engineering groups across the world. Many approaches have tried to refine the Gratzel [1] architecture of a dye-sensitized m esoporous titanium oxide layer w ith an iodine-based electrolyte and platinum counter electrode. A significant increase in efficiency would not have been possible without new dyes, whose absorption spectra better match the spectral distribution of both synthetic [2,3] and natural [4] sunlight. Much work has been devoted to improving electrode perform ance. New w ide bandgap sem iconductors w ith different m orphologies (e.g., in nanowires, nanotubes, and nanospheres) have been tested [5]. Various electrolyte formulations have been developed and tested in order to improve DSSC performance, increase their lifetime, and reduce their toxicity [6,7].
Besides chem ical stability and technical durability (e.g., a tim e-dependent decrease in pow er conversion efficiency (PCE) [8,9] or the leakage of liquid electrolytes due to the unsealing of the cells [10]), some typical physical challenges [11,12] are yet to be solved. A frequently reported mechanism that leads to a decrease in DSSC efficiency is the electron-electrolyte recombination process.
In an ideal m odel, the photoexcited electrons in dyes are injected into a w ide bandgap n-type sem iconductor layer, typically TiO2 , ZnO or SnO 2 , and are then transferred to a fluorine-doped tin oxide (SnO2 :F, FTO) w orking electrode, w hile the holes created in the dye are; filled by electrons from the triiodide present in the redox electrolyte ( 2e-+ 1--3I-). Unfortunately, there are, several possible paths through which electrons recombine. The most significant of these are recombinations at the TiO2/electrolyte and FTO/electrolyte interfaces. To challenge these pathways, the concept of the electron blocking; layer (EBL) w as developed. The role ot this layer ie to block electrona from feeing injected bade into the electrolyte by providing more preferable transport to the FTO via an appropriate band structure.
The EBL can be set directly on a mesoporous n-type semiconductor [13] or on the FTO surface [14].
There are fwo main conditions that need to lie Culfilled. Firstly, the layer must be continuous to prevent FTO contact w ith the electrolyte, and sec: ondly, the layer has to be thin enough to avoid high ohmic resistance. The latter factor could lower tire open circuit voltage (Uoc) and short circuit curient (Isc) due to a greater probability of recombination from structural defects. The blocking layers deposited on mesoporous TiO2 aee generally represented by w ide bandgap isolators, such as A^O 3 or lTfO2 [15], Z^S n C fi [16], MgO [ 17] or doped ZnO [e8]. Indeed, -t is difficult: to produce a continuous layer on a mesoporous structure. CVD methods are the most promising because the reaction gas is able to overlay the whote surface exposed to the gas. How ever, toxic or hazardous preiursors and by-products are often a great complication of this technique. Additionally, the required high temperature (en the range of several hundred degrees Celsiut) for the CVD growth process cculd significantly influence the, other layers. Thus, it is m uch simpler tee prepare an eleciron blocking layer directly onto the planae FTO glass, as shown in Figure 1. As reported by Zhu [S9], thee idominant eecombination process occurs at the substratetelectrolyte interface rather than across thee TiO2 matrix. Numerous methods have been commonly used for EBL compact layer preparation, such as sol-gel [ea], chemical vapor deposition [21], spray-coating [22], and electrcdeposition [23]. Low-cost depositions are typically considered to be the most prcm iting, but they do not provide a process 1:hat is highly repeatable, while the p arameters, such as thickness, stoichiometry and purity of the layer, are crudal. Among the many potentia1 techmques, the m agnetron sputtering m ethod [24,25] seems to be the best choice because of its sim plicity of use and repeatability. It is necessary to choose an appropriate material as an EBL. For inttance, SnO2 [26], ZnO [at] or Nb2 O5 [28] were recently reported. However, SiO 2 is suggested to be the beet EBL due to its low cost, chemical inertnesa to elsctrolytes, easy availaedity, good n-type eonductivity and matching elecpronic band structure to the photoanode. Herein, a TiO2 RF m agnetron sputtered layer was investigated as an electron blocking layer in screen-printed dye-sensitized solar cells. The results, presented in the follow ing sections, were obtained by the current-voltage characteristics (U -I), scanning electron m icroscopy (SEM ), atom ic force m icroscopy (AFM), optical profilom etry (OP), energy dispersion spectroscopy (EDS), Raman spectroscopy (RS) and UV-VIS-NIR spectrophotometry. In this paper, two sizes of the active layer were studied-small-and large-scale solar cells w ith dim ensions 7 X 7 m m 2 and 7 X 80 mm 2. The pow er conversion efficiency, as a result of the additional electron blocking layer, was increased by 30% and 17%, respectively.

RF Magnetron Sputtered Electron Blocking Layer Preparation
The TiO2 com pact layer was prepared on the FTO glass (TEC A7, 6 -8 D/m). The substrates were ultrasonicated in acetone and isopropanol for 5 min and 15 m in, respectively. Then, they were transferred to a glove box immediately. The electron blocking layer was sputtered by a Mini SPECTROS™ m agnetron system from the TiO2 target (99.99% pure, K urt J. Lesker Company®, Jefferson Hills, USA). The cham ber w as pumped down to about 10-5 Torr. A time of 5 min of pre-sputtering w as applied to rem ove any possible contam inants from the target surface. The RF sputtering was carried out at 300 °C to enhance adhesion and, thus, electric contact between the two layers. The param eters of the process were set as a 3 m Torr argon w orking gas pressure under a power of 100 W. The deposition time was tuned to receive a layer of approximately 30 nm, which was reported as the most efficient thickness for the EBL [29].

Dye-Sensitized Solar Cell Fabrication
The FTO glass (cleaned as described above) w as used to fabricate the w orking and counter electrodes. M esoporous TiO2 film was deposited using the screen printing method [30] from titania paste (18NR-T, Greatcell Solar, Elanora, Australia) and sintered up to 565 °C. The obtained thickness of 11 pm was determ ined by a stylus profilometer (Bruker DektakXT, Billerica, USA). According to the manufacturer's datasheet, anatase nanoparticles dispersed in the paste have an average size of 20 nm. Sam ples were im m ersed for 24 h in a 10-4 M ethanolic solution of ruthenium dye (N719, Greatcell Solar, Elanora, Australia) at room temperature. The layers were then rinsed w ith ethanol to remove the un-adhered m olecules. A counter electrode w as prepared by the same printing m ethod using a platinum paste (PT1, Greatcell Solar, Elanora, Australia). A 60 pm lam ination foil (DuPont Surlyn®, W ilm ington, USA) was used to seal both electrodes. The DSSCs were filled w ith a commercially available iodine-based redox electrolyte (EL-HPE, DyeSol, Elanora, Australia).

Measurement ofTiO2 Electron Blocking Layer Thickness
The thickness of the m agnetron sputtered nanom etric TiO2 thin film was determ ined by an optical profilometer (Profilm3D, Film etrics, San Diego, USA) in the white light interferometry (WLI) mode. Before deposition, the edge of the FTO glass was covered by polyimide tape and then removed. Additionally, the taped area w as slightly cleaned by isopropyl alcohol to rem ove any glue residue.
Then, the step height was imaged on a profilometer. The received data were used to prepare the height histogram presented in Figure 2 . The cyan color represents the bare FTO surface and the olive indicates the FTO covered by TiO2 . The grey area was not included in the calculations because of the irregular layer surface corrupted by detaching the tape. Two Gauss functions were fitted to the obtained peaks.
The difference between the peak positions, calculated to be 32.7 nm, was considered as the thickness of the TiO2 layer.

Energy Dispersive Spectroscopy (EDS) Spectrum
The energy dispersive spectroscopy (Sw iftEDS 300, Oxford Instrum ents, H igh W ycom be, UK, beam energy 15 kV) outcome, shown in Figure 3, was obtained from a sample with a Ti02/FTO/glass stack. Tine signal from the carbon (C) elem ent is a result of the atm ospheric adsorption of carbon dioxide? because the sample was not heated before its; transition to the EDS chamber. "The? small peak from silicon (Si) cam e from the glass substrate, w hich, in fact, is a silica. The observed oxygen (O) signal came simultaneously from the CO2 adsorbate and the TiO2, SnO;>:F and SiOx layers.   Table 1.
The broad m ultipeak signal from tin (Sn) was caused by the FTO 's presence in the analyzed sample. The m anufacturer's datasheet suggests thaa the FTO film is approxim ately 250 nm. This significant thickness, comp ared to the electron penetration depth, explains w hy tire? Sin signal has the greatest im portance in the obtained spectrum. The m ost im portant piece of inform ation from the received spectrum is that the TiO2 thin film is present in the sample. This is observable in the three peaks that are tabulated in Table 1. Lines w ith double num eration are a superposition of two signals with similar energy. For instance, the peaks for O are a superposition of K a-1 and K a.2 transitions.

Scanning Electron Microscopy (SEM) and Atomic Force Microscopy (AFM) Images
The electron blocking layer's m orphology w as studied w ith a scanning electron m icroscope (Quanta™ 3D FEG, ThermoFisher Scientific, Hillsboro, USA) before the deposition of the mesoporous TiO2 . The topography, shown in Figure 4c, is similar to that of the bare FTO, which was also observed by Choi et al. [31]. The reason for this result is the FTO 's high roughness. Because of this roughness, covering the FTO with a thin film of material would not significantly influence the RIMS parameter. The roughness of the titanium oxide EBL was measured by an atomic force microscope in tap ping mode (Agilent 5500, Agilent Technologies, oanta Clara, USA). The ISMS? decreased shghtly from 26.9 nm for the bare FTO to 23.7 nm for a stack of the :30 nm TiO2 com pact layers on the FTO. The intersection, seen in Figure 4a, contains all of the photoanode layers (from tire bottom: glass, FTO, TiC>2 com pact and TiO2 m esoporous layers). In reference to the profilometer m easurement, the thickness of the mesoporous film was also confirmed to be approximedely 11 pm. The morphology of thie sintered TiOy nanoparticles with an average diameter of 20 nm c an be observed in Figure 4b.

Raman Spectrum of Electron Blocking Layer
The polym orphism of the titanium oxide electron blocking layer was investigated by Ram an spectroecppy. The m easurem e nts were made using an integrated Ram an spectrogcope (LabRAM , HORIBA, Kyoto, Japan). The source used for the measurements was a Fle-N e paser with a wavblength of 6312.8 nm at 17 mW power. "Tire whole system was complemented by a confocsl microscope coupled w ith an 800 nm focused lens and a two-dśmensional m ultichannel CCD detector. Additionally, the syste m was equi pped with an automatic Czmrny-Turner spectrograph with a slot width d 30 mm for e w ide-flat field of -view. The measurements were tauen at room temperature .
According to other studies [32], the obtainwd spectrum of TiO2, shown in

Transmittance and Absorbance Spectra o f the Photoanode
Because the light pathw ay to the dye leads through the w hole photoanode, UV-VIS-NIR spectrophotom etry (Jasco V-670 with deuterium and halogen lamps) w as used to m easure the transm ittance for each layer. The signal was averaged from a spot size? of a 3 X 3 m m 2 area. Figure 6 presents the transm ittance versus the w avelength specirum 0 f com positions from bare FTO up to FTC) covered by a dye-sensitized T i0 2 layer. Jus;!; a slight decrease in transmitted light was observed for" the additional 30 nm TiF>2 film in the dye absorption w avelength eange, w hich indicates that the EBL layer has no gignifinant influence on the am ount (of light transm itted to the N719 molecules. Measured absorption peaks near the 390 nm and 530 nm are consistent w ith those reported by other studies [33,34]. Tire dye absorbance "was m easured for a 10-4 Ml ethanolic solution (99.8% ethanol, Honeywell, Charlotte,. USA) in a quartz cuvette wtth 1 cm of optical thickness. (gray) and mesopotous dye-sensitized TiO2 (red), compared to the absorbance spectrum oi the N719 dye (orange).

3.(3. Performance of the DSSC Devices
The performance of the DSSC devices was investigated by a solar sim u k tor (CLASS-01, PV Test Solutions) under AM1.5 illumination with a light intensity of 100 mW/cm2. The received current-vnttage As recently shown, a thin layer of TiO2 on the FTO electrode significantly increases the onset of cathodic dark current, pointing to the increase in a potential barrier to the electron transfer from the FTO to the electrolyte [37]. In our studies the presence of the electron blocking layer significantly increased the short circuit current (Isc) and, as a consequence, also increased the fill factor (FF). For small samples, an increase from 5.33 mA to 6.44 mA and from 0.64 to 0.70 w as observed, respectively. For larger samples, the same parameters increased from 41.67 mA to 44.80 mA and from 0.55 to 0.58, respectively.
In both cases, the im provem ent of those param eters indicates that the recom bination process was considerably mitigated. As mentioned above, the pow er conversion efficiency (n) increased by 30% and 17%, depending on the sample size. Because of the larger surface, there is a higher probability of the electrons being scattered or recom bined, especially due to the crystal structure's defects. The longer path to the electrodes may also cause a relaxation, with incidentally met holes in the FTO. We attem pted to m itigate this problem via the preparation of a silver electrode along the active area's edge. The role of this electrode was to collect the photogenerated electrons, possibly close to the place where they were excited. The m easurement errors, presented in Table 2, were calculated by standard deviation. The parameters were obtained from eight samples for every type of DSSC device.

Conclusions
A thin film of TiO2 was studied as an efficient electron blocking layer. Due to its simplicity of use and high repeatability, RF m agnetron sputtering w as chosen as the deposition method. A 30 nm thick layer w as sputtered on bare FTO and then dye-sensitized solar cells were fabricated by a screen printing technique. The UV-VIS-N IR spectrophotom etry transm ittance spectra showed no significant influence on the am ount of light provided through the photoanode to the dye molecules in its absorbance w avelength range. In both cases of the active area (7 X 7 m m 2 and 7 X 80 m m 2), an increase in pow er conversion efficiency due to decreased recom bination (and thus an increased short circuit current) was reported. Because of the larger area, the cross section of recom bination for m echanism s other than the FTO/electrolyte interface has greater importance. Thus, the percentage growth of efficiency was higher for small samples (30%) than for larger ones (17%).
DSSCs, as third generation solar cells, provide a great opportunity to serve as low -cost and com m on photovoltaics and could displace silicon-based devices. Present and future works will certainly improve both the efficiency and stability of DSSCs. However, there are still m any physical, chemical and technical problems that need to be resolved to deploy these devices into line production and distribute them to private consumers.
Author Contributions: Conceptualization, D.A., J.D. and P.K.; writing-original draft preparation, data curation, formal analysis, investigation, resources-sample preparation and visualization, D.A.; methodology, J.D.; project administration, funding acquisition and validation, P.K. and J.D.; supervision and writing-review and editing, P.K. and J.R. All authors have read and agreed to the published version of this manuscript.