CVD-graphene/graphene flakes dual-films as advanced DSSC counter electrodes

The use of graphene-based electrodes is burgeoning in a wide range of applications, including solar cells, light emitting diodes, touch screens, field-effect transistors, photodetectors, sensors and energy storage systems. The success of such electrodes strongly depends on the implementation of effective production and processing methods for graphene. In this work, we take advantage of two different graphene production methods to design an advanced, conductive oxide- and platinum-free, graphene-based counter electrode for dye-sensitized solar cells (DSSCs). In particular, we exploit the combination of a graphene film, produced by chemical vapor deposition (CVD) (CVD-graphene), with few-layer graphene (FLG) flakes, produced by liquid phase exfoliation. The CVD-graphene is used as charge collector, while the FLG flakes, deposited atop by spray coating, act as catalyst for the reduction of the electrolyte redox couple (i.e., I3-/I-- and Co+2/+3). The as-produced counter electrodes are tested in both I3-/I-- and Co+2/+3-based semitransparent DSSCs, showing power conversion efficiencies of 2.1% and 5.09%, respectively, under 1 SUN illumination. At 0.1 SUN, Co+2/+3-based DSSCs achieve a power conversion efficiency as high as 6.87%. Our results demonstrate that the electrical, optical, chemical and catalytic properties of graphene-based dual films, designed by combining CVD-graphene and FLG flakes, are effective alternatives to FTO/Pt counter electrodes for DSSCs for both outdoor and indoor applications.


Production of materials
Chemical vapor deposition and transfer of graphene. Continuous films of graphene were synthesized on Cu foil (Sigma Aldrich, thickness 20 m, 99.999%) by CVD by using a cold-wall reactor with methane as a carbon precursor [200]. The Cu foil was loaded into the CVD reactor, which was then heated to 1060°C in Ar atmosphere to anneal the foil for 10 min. After this annealing step, the graphene growth was performed at 1060 °C for 10 min (pressure of 25 mbar) by flowing CH4 (2 sccm), H2 (20 sccm) and Ar (1000 sccm). The CVD reactor was then cooled down to 120 °C before removing the samples to prevent substrate oxidation. Samples of CVD-graphene of 1×1 cm 2 size were transferred to glass or SiO2 substrates using wet transfer technique with poly(methyl methacrylate) (PMMA) as support medium [201]. Briefly, a thin layer of PMMA (2% solution in acetyl lactate, All resist GmbH) was deposited onto Cu/graphene by spin coating, and dried for 1 h at ambient conditions. The as-obtained samples were immersed in a 0.05 M solution of iron(III) chloride (FeCl3, Sigma-Aldrich) for 16 h to etch the Cu and release the graphene/PMMA film. Once the Cu was completely etched away, the graphene/PMMA membrane was removed from the FeCl3 solution using a glass slide and transferred to deionized water several times to remove the etchant residue. Subsequently, the membrane was removed from the water using glass or SiO2 substrates, and dried at ambient conditions. Finally, the PMMA support film was removed by immersing the sample in acetone (Sigma Aldrich) for 4 h and then rinsed in 2propanol (Sigma Aldrich).

Characterization of materials
Scanning electron microscopy (SEM) images of CVDgraphene were taken with a FE-SEM (Jeol JSM-7500 FA). The acceleration voltage was set to 5 kV.
Transmittance spectra of the CVD-graphene on glass were taken with a Cary Varian 6000i UVvis-NIR spectrometer, using a 1 mm pinhole holder. The pristine glass substrate was used as baseline. Each sample was measured 5 times and the averaged values were reported.
Optical absorption spectroscopy (OAS) measurements of the LPE-produced graphene flake dispersion in NMP were performed with a Cary Varian 6000i UVvis-NIR spectrometer. The absorption spectra were acquired using a 1 mL quartz glass cuvette. The inks were diluted to 1:100 in NMP, to avoid scattering losses at higher concentrations. The corresponding solvent baseline was subtracted to each spectrum. The concentration of graphitic flakes is determined from the optical absorption coefficient at 660 nm, using A = αlc where l is the light path length, c is the concentration of dispersed graphitic material, and α is the absorption coefficient, with α ~ 1390 L g −1 m −1 at 660 nm [180,204].
Raman spectroscopy measurements on CVDgraphene (transferred on glass and Si/SiO2) and LPEproduced graphene flakes were carried out by using a Renishaw inVia confocal Raman microscope using an excitation line of 532 nm (2.33 eV) with a 50× objective lens, and an incident power of ~ 1 mW on the samples. The LPE-produced flakes were obtained by drop-casting their dispersion in NMP onto a Si wafer with 300 nm of thermally grown SiO2 (LDB Technologies Ltd.). The samples were then dried under vacuum before the measurement. The spectra were fitted with Lorentzian functions. Statistical analysis was carried out by means of OriginPro 9.1 software.
Transmission electron microscopy (TEM) measurements of the LPE-produced graphene flakes were carried out with a JEM 1011 (JEOL) transmission electron microscope operating at an acceleration voltage of 100 kV. The samples were obtained by depositing 1:100-diluted graphene flake dispersion in NMP onto holey carbon (200 mesh grids). Subsequently, the samples were dried under vacuum overnight.
Atomic force microscopy (AFM) images of the LPEproduced flakes were taken using an Innova AFM (Bruker, Santa Barbara, CA). The measurements were taken in tapping mode with a NTESPA 3.75 mm cantilever (Bruker, 300 kHz k: 40 N m −1 ), in air at room temperature, with a relative humidity less than 30%. The software used for image analysis was Gwyddion version 2.43. Statistical analysis was carried out by means of Origin 9.1 software, using different AFM images of the sample. The sample for the measurements was prepared by drop-casting 1:30 diluted graphene flake dispersion in NMP onto mica sheets (G250-1, Agar Scientific Ltd., Essex, U.K.) and drying them under vacuum overnight.

Fabrication of graphene-based CEs
Glass/CVD-graphene were produced according to the protocols described in Section 2.1. Subsequently, the LPE-produced graphene flakes dispersions (see Section 2.1 for the details regarding the dispersion production) were deposited onto glass/CVD-graphene samples by spray coating. The dispersions were sprayed using a flux of N2 at 1 bar, and keeping the substrates at a temperature of 150 °C. By controlling the amount of sprayed dispersion, three different mass loadings of the graphene flakes (0.16, 0.32 and 0.48 mg cm -2 ) were deposited onto the CVD-graphene. After the deposition of the graphene flakes film, the substrates were annealed in glove box at 350°C for 2.5 h in order to remove solvent residuals.

Characterization of graphene-based CEs
Scanning electron microscopy (SEM) images were acquired with a FE-SEM (Jeol JSM-7500 FA). The acceleration voltage was set to 5 kV.
Atomic force microscopy and Raman spectroscopy measurements were acquired with the same instrumentations and parameters reported in Section 2.2.
Transmittance spectroscopy measurements of the graphene-based CEs were acquired with an integrating sphere-supported UV-Vis 2550 Shimadzu Spectrophotometer. A glass/CVD-graphene substrate was used as baseline. Each sample was measured 5 times and the average values were reported.
Sheet resistance measurements were performed with a Keithley Model 2612A Dual-channel System Source Meter in four-point probe configuration, using in line gold-plated probes of constant spacing (2 mm) contacting the surface of the films.
Specific surface area measurements of electrodes were carried out in Autosorb-iQ (Quantachrome) by Kr physisorption at temperatures of 77 K. The specific surface areas were calculated using the multipoint Brunauer-EmmettTeller (BET) model [205], considering equally spaced points in the P/P0 range from 0.009 -0.075 to. P0 is the vapour pressure of Kr at 77 K, corresponding to 2.63 Torr [206][207][208][209]. Before the measurements, the samples were degassed for 1 h at 60 °C under vacuum conditions to eliminate weakly adsorbed species.

Fabrication of DSSCs and symmetrical dummy cells
Fluorine-doped SnO2-coated glass (soda lime) substrates (8 Ω□ -1 , Pilkington Tec) were cleaned using successive ultrasonic baths in acetone (Sigma Aldrich) and ethanol (Sigma Aldrich). Two types of DSSCs using C106 and Y123 as dye sensitizer, respectively, were fabricated using specific protocols. For C106-based DSSCs, films of nanocrystalline TiO2 (0.5×0.5 cm 2 ) were deposited onto FTO-glass via screen-printing, using a TiO2 paste (Dyesol 18NR-T). Subsequently, the substrates were dried in an oven for 20 min at 120 °C in order to evaporate the solvent. The thickness of TiO2 layers was 6 m thick, as measured by Dektak Veeco 150 profilometer. The as-obtained samples were then exposed to a sintering procedure at 525 °C for 30 min. Then, the samples were soaked in the C106 solutions for 16 h, washed with ethanol and blown with compressed air, obtaining the photoanodes. Graphenebased CEs were prepared as described in Section 2.3. Reference counter electrodes based on FTO-Pt were also fabricated. Briefly, a Pt layer was deposited onto FTO by screen-printing an as-purchased paste containing Pt precursor (Platisol, Solaronix). Subsequently, the substrates were dried in an oven at 120 °C for 20 min, and sintered for at 480 °C 30 min. The devices were laminated with a 25 m-thick thermoplastic resin (Surlyn, Solaronix). After a hot-melting step, the distance between the two electrodes was measured to be about 20 m. A commercial I3 -/I --based electrolyte (High Performance Electrolyte -HPE-, Dyesol) was injected by vacuum assistance. Lastly, the cells were closed with glue. In the case of DSSC based on Y123 conductive Pilkington TEC glassy plates (4×15 cm 2 ) were immersed in 100 ml of TiCl4/water solution (40 mM) at 70 °C for 30 min, washed with water and ethanol and dried in an oven at 80 °C for 30 min. The TiO2 layers were deposited on the FTO glassy plates by screen-printing (frame with polyester fibers having 77.48 mesh cm -2 ). This procedure, involving two steps (coating and drying at 125 °C), was repeated twice. The TiO2-coated plates were gradually heated up to 325 °C. Then, the temperature was increased to 375 °C in 5 min, and afterwards to 500 °C. The plates were sintered at this temperature for 30 min, finally cooled down to room temperature. After the TiO2 film was treated with 40 mM TiCl4 solution, following the procedure previously described above, rinsed with water and ethanol. Lastly, a coating of TiO2 nanoparticles  nm in size, Dyesol) was deposited as scattering layer onto the samples by screen-printing and sintered at 500°C. Each anode was cut into rectangular pieces (area: 2×1.5 cm 2 ) having a masked spot area of 0.181 cm 2 with a total thickness of ~8 μm. The anode were soaked for 16 h in a dye solution composed by Y123 (0.1 mM) and chenodeoxy acid (5mM) in alcohol ter-butylic:acetonitrile (1:1). The assembly of the whole DSSCs followed the same protocols adopted for C106-based DSSCs, except for the use of Co(bpy-pz)2 (2+/3+) (bpy-pz= 6-(1H-pyrazol-1-yl)-2,2'-bipyridine)-based electrolyte (DN C09 and DN C10, Dyenamo).
Symmetrical dummy cells were produced by assembling two identical CEs of FTO, or FTO/graphene flakes, or FTO/Pt, or CVD-graphene/graphene flakes. The electrodes were sealed with 25 m-thick thermoplastic resin (Surlyn, Solaronix) filled with I3 -/Ielectrolyte (HSE, Dyesol). An active area of 1.44 cm 2 was made with a thermoplastic mask.

Characterization of DSSCs and dummy cells
The Electrochemical impedance spectroscopy (EIS) measurements of dummy cells were taken in dark conditions at room temperature using an Autolab 302N Modular Potentiostat (Metrohm) in two-electrode configuration under short circuit condition (0 V of electrical bias). The ac perturbation was set at 10 mV with frequencies ranging from 100 kHz to 0.1 Hz.

Results and discussion
The CVD-graphene and the LPE-produced graphene flakes (see Experimental, Section 2.1, for the material production details) were first characterized separately before the fabrication of the graphene-based CE. The morphology of the materials was evaluated by electron microscopy and AFM measurements.

Figure 2a
shows a representative SEM image of the CVD-graphene grown on Cu foil by using a cold-wall reactor (see additional detail in Experimental, Section 2.1) [200]. The sample is overall a continuous polycrystalline film without any tears or holes, in agreement with previous studies [200]. Such morphology resulted in a Rsheet of ~1.2 kΩ□ -1 . Partial coverage experiments with a growth time of 5 min revealed a high nucleation density (25000 grains mm -2 ) with an average grain size of a ~10 m (Figure 2b). The nucleation of graphene on Cu occurs primarily on the surface irregularities such as the rolling grooves and crystal terraces acting as energetically favorable spots for the nucleation of graphene [200,211]. Figure 2c,d show representative bright field TEM and AFM images, respectively, of the LPE-produced sample, consisting of irregularly shaped (Figure 2c) and nm-thick flakes (Figure 2d). The corresponding electron diffraction pattern of the TEM image is also shown (inset in Figure  2c), proving that the flakes are crystalline [212]. Statistical analysis (Figure 2e,f) indicate that the lateral size and the thickness of the flakes follow a lognormal distribution, peaked at ~190 nm and ~2 nm, respectively. These data indicate that the sample obtained by LPE of graphite is mostly composed by FLG flakes (experimental SLG thickness measured by AFM is typically between 0.4 and 1 nm[104, [213][214][215] depending on tip-surface interactions and image feedback settings [214,215]).Raman spectroscopy analysis was carried out to evaluate the structural properties of the asproduced materials. A typical Raman spectrum of graphene shows, as fingerprints, G, D and 2D peaks (see Supplementary material for more details) [192,[216][217][218]. For SLG the 2D band is roughly four times more intense than the G peak [216,217]. Multi-layer graphene (> 5 layers) exhibits a 2D peak, which is almost identical, in term of intensity and lineshape, to the graphite case (intensity of the 2D2 band is twice the 2D1 band) [217][218][219]. Instead, FLG (< 5 layers), has a 2D1 peak more intense than the 2D2 [220]. Taking into account the intensity ratios of the 2D1 and 2D2, it is possible to estimate the flake thickness [145,175,192,193,220]. Figure 3a shows the Raman spectrum of the CVDgraphene transferred onto Si/SiO2. The absence of the defect-related D peak and the ratio between the intensity of 2D and IG peak -I(2D)/I(G)of ~ 3 indicate that a high-quality SLG has been obtained [221], in agreement with previous studies [200]. The analysis of the 2D peak, a single and sharp Lorentzian band centered at ∼2683 cm -1 , also confirm that the sample is SLG [192,216,218]. Figure 3b displays a representative Raman spectrum of LPE-produced graphene flakes. The Raman spectrum of the native bulk graphite is also shown for comparison. The LPE-produced flakes exhibit an enhancement of the D and D' bands compared to those of the pristine graphite, in agreement with previous discussion [192,216,218,[222][223][224][225]. For graphite, the G peak is more intense of D, while the Raman statistical analysis for LPE-produced graphene flakes ( Figure S1) shows that the ratio between the intensities of the D and G peaks -I(D)/I(G)ranges between 0.35 and 0.75, with an average values of ~0.55 ( Figure S1a). However, the plot of I(D)/I(G) vs. FWHM(G) (Figure 3c) does not show a linear correlation, which means that the defects mainly originate from the flake edges without altering the structure of the basal plane [226,227]. The analysis of I(2D1)/I(2D2) (Figure 3d) demonstrates that the LPEproduced sample has a few-layer flakes enriched composition [210,217,220].
After the preparation and characterization of the CVDgraphene and LPE-produced graphene flakes, the CEs were fabricated by spraying the LPE-produced graphene flakes dispersion onto the CVD-graphene previously transferred onto the glass substrate, as detailed in Section 2.3. The concentration of graphene flakes in the LPE-produced dispersion (~0.36 gL -1 ) was estimated by OAS measurements (Figure S2) [204], and served to adjust the amount of deposited graphene flakes. Three different batch of graphene-based CE (named as graphene-A, graphene-B, graphene-C) were prepared with graphene flakes mass loading of 0.16, 0.32, 0.48 mg cm -2 , respectively. The morphology of the CEs was analyzed by SEM and AFM measurements. Figure 4a reports a SEM image of a representative graphenebased CE (graphene-B). The electrode surface is completely covered by a film of graphene flakes. Figure  4b shows the AFM images of the surface of the same electrode. The root means square roughness (Rrms) of the sample is 42 ± 2 nm and the average roughness (Ra) is 15 ± 2 nm. The SSA of the electrode was estimated by Brunauer, Emmett and Teller (BET) analysis [205] of physisorption measurement with Kr at 77 K [206,208,209], obtaining a value of ~123 ± 25 m 2 g -1 . Figure 4c shows the transmittance spectra of the CVDgraphene/FLG flakes dual film CEs, in comparison to those measured for both the CVD-graphene and conventional FTO/Pt CE (see fabrication detail in Experimental section). As previously reported, CVDgraphene shows an excellent optical transparency (Tr > 97% for visible wavelengths ranging between 400 and 800 nm). After graphene flakes deposition, graphene-A and graphene-B retain Tr ~25% and ~14%, respectively, at 700 nm, while graphene-C exhibit low Tr (< 5% for all the visible wavelength) due to the high mass loading of the graphene flakes (i.e., 0.48 mg cm -2 ). Figure 4d reports the Raman spectroscopy analysis of the graphene-based CE. The Raman spectrum of the graphene-based electrode resembles those obtained for the LPE-produced flakes, thus indicating that the spray deposition process did not affect the structural properties of the flakes.
The graphene-based CEs were used to fabricate DSSCs by using commercial I3 -/I --based electrolyte, and C106 dye (see detail of the DSSCs' fabrication in the Experimental section), which are simply named the corresponding CEs (i.e., graphene-A, graphene-B and graphene-C). Figure 5a shows the IV measurements of the devices under 1 SUN and in dark (inset to panel), respectively. Table 1 summarizes the main PV parameters of the fabricated DSSCs (as estimated by the IV curves). DSSCs with CE based on FTO and CVDgraphene are also shown for comparison. The PCE of such cells is less than 0.01%, indicating the need of a catalytic layer for obtaining an efficient electrolyte regeneration process. It is worth noticing that DSSCs based on only graphene flakes have been preliminary discarded because of the high Rsheet of their films (> 100 kΩ□ -1 for a mass loading of 0.16 mg cm -2 and > 10 kΩ□ -1 for both mass loadings of 0.32, 0.48 mg cm -2 ), which resulted in poor PV performances.Graphene-A exhibits a short circuit current -Jscof 5.67 mA cm -2 , an open circuit voltage -Vocof 740 mV, and a fill factor -FFof 37.2%, leading to a PCE of 1.62%. By increasing the mass loading of the graphene flakes, the PCE of graphene-B (2.13%) improves by 51% compared to graphene-A. This is mainly attributed to the higher Jsc and FF of graphene B (5.97 mA cm -2 and 54.9%, respectively) compared to those of graphene-A. However, a further increase of the mass loading of graphene flakes (graphene-C) decreases the PCE to 1.2%. The IV results can be tentatively explained by considering that the catalytic activity depends both by the amount of the catalytic materials and its electrical resistance. In absence of resistive effects, the increase of the mass loading of the catalytic materials enhances the catalytic activity of the CEs for the electrolyte regeneration reaction (I3 − reduction in our case), as typically observed for conventional FTO/Pt CE [228,229]. However, an excessive mass loading of graphene flakes can increase the electrical resistance of the corresponding layer, negatively affecting the catalytic activity of the superficial (i.e., more exposed) flakes, which interact more effectively with the electrolytes than the inner flakes. The IV curves in dark (inset in Figure  5a) show that current density at forward bias polarity for graphene-A and graphene-C (maximum current density of 2.1 mAcm -2 and 1.7 mAcm -2 , respectively) is lower than that of graphene-B (maximum current density of 12.0 mA cm -2 ). Differently from the reference cell adopting FTO/Pt CE, the current density for the graphene-based DSSCs does not saturate, which means that the diffusion limiting current density is not reached in the investigated voltage window (where corrosion effects can be excluded) [230], and resistive losses occur.  In order to elucidate the origin of such losses in presence of graphene-based CEs, EIS measurements were carried out on dummy cells consisting of two identical CEs sandwiching the I3 -/I --based electrolyte used for DSSCs (see detail of dummy cell fabrication in Experimental, Section 2.5). Electrochemical impedance spectroscopy is a perturbative techniques, which measures the current response of an electrochemical system when AC voltages are applied with difference frequency, thus computing the electrochemical impedance (Z) [231]. The analysis of Z allows the kinetics of the (photo)electrochemical processes [231], including the electronic and ionic ones occurring in the DSSCs, to be studied [230,[232][233][234]. In particular, by performing EIS on symmetrical dummy cells it is possible to elucidate both the electrochemical activity and the electrical resistance of CEs in simulated DSSC operating conditions, eliminating the photoanode contribution[101, 235,236]. The Z data can be expressed graphically in a Nyquist plot (imaginary part of Z -Im[Z]vs. real part of Z -Re[Z]-), which is typically composed of two semicircles for symmetrical dummy cells [232,237]: the first one (at frequency in the kHz-MHz range) is associated to the catalyst/electrolyte interface [232,237], the second one (at lower frequency, < 100 Hz) is connected with ionic diffusion processes occurring in the electrolyte [232,237]. Table 2. Photovoltaic parameters of our DSSCs (data extrapolated by IV curves in Figure 6) in comparison to literature data about cells based on Y123 dye and Co(bpy-pz)2(2+/3+) (bpy-pz= 6-(1H-pyrazol-1-yl)-2,2'-bipyridine)based electrolyte. Our DSSCs with all-graphene CEs were fabricated using CVD-graphene and graphene flakes (graphene-B), as described in the text. PEDOT:PSS stands for poly(3,4-ethylenedioxythiophene)poly(styrenesulfonate), GNPs for graphene nanoplatelets, and PProDOT for poly(3,4-propylenedioxythiophene. The intercept with the x-axis of the first semicircle represents the ohmic series resistance (Rs) given by both the electrolyte resistance and the resistance of external circuit, including the electrical resistance of the CEs [232,[237][238][239][240]. The diameter of the first semicircle is proportional to the electrode-electrolyte charge transfer resistance (Rct) [232,[237][238][239][240], which is related to catalytic activity of the CE towards the electrolyte redox reactions (low Rct corresponds to an high catalytic activity) [96,231]. Figure S3 reports the equivalent electrical circuit used to model the symmetrical dummy cells impedance, together with its description, in agreement with previous literature [231,240]. Figure 5b reports the Nyquist plots obtained for the symmetric dummy cells adopting different electrodes, i.e., graphene-B and FTO/Pt. Fluorine-doped tin oxide and graphene flakes-covered FTO (FTO/graphene flakes) were also characterized to ascertain the impedance contribution of each layer, i.e., the current collectors (FTO or CVD-graphene) and the catalytic films (Pt or graphene flakes). The FTO-based dummy cell has a low Rs ~20 Ωcm 2 , mainly ascribable to the Rsheet of FTO (8 Ω□ -1 ), and a high Rct (> 10 kΩcm 2 ), which indicates the absence of catalytic activity [241]. The FTO/graphene flakes-based dummy cell exhibits the same Rs of pristine FTO, while Rct (~650 Ωcm 2 ) decreases by one order of magnitude compared to that of FTO. For graphene-Bbased dummy cell, Rs is higher than 500 Ωcm 2 due to high Rsheet of CVD-graphene (~1.2 kΩ□ -1 ), as measured by four-point probe. Interestingly, the Rct of graphene-Bbased dummy cell (~250 Ωcm 2 ) decrease by ~62% compared to that of FTO/graphene flakes-based dummy cell. This indicates that the coupling between CVDgraphene and graphene flakes is effective for increasing the overall CE catalytic activity of the same graphene flakes. However, the Rct measured for graphene-Bbased dummy cell is still significantly higher than that of FTO/Pt (< 2 Ωcm 2 ), which indicate that graphene-based CEs have an insufficient catalytic activity with I3 -/I --based electrolyte for practical DSSCs operating under solar illumination, in agreement with previous studies [236,242].

CE
In order to exploit CVD-graphene/graphene flakes as CE for efficient DSSC under simulated 1 SUN for outdoor application, we tested them into DSSCs based on Y123 dye and Co(bpy-pz)2 (2+/3+) (bpy-pz= 6-(1Hpyrazol-1-yl)-2,2'-bipyridine)-based electrolyte. It has been demonstrated that graphene-based materials exhibit a catalytic activity for reducing the redox mediator of polypyridine complexes of Co 2+ /Co 3+ comparable to that of Pt [55,56,236,238,243]. Figure 6a shows the comparison between the IV curves obtained for a Co 2+ /Co 3+ -based DSSC adopting the optimized graphene-B (i.e., the best graphene-based CE in I3 -/I --based DSSCs) and FTO/Pt. In particular, the PCE of graphene-based DSSC achieved a PCE of 5.09%, which is much higher than the one reached with I3 -/I -based electrolyte (2.1%). We further extended the study and validation of our graphene-based DSSCs for indoor applications, by testing them under low illumination intensity conditions (< 1 SUN). In this case, the photocurrent densities are lower than those generated under 1 SUN, and consequently the PV performances are less affected by the CE series resistance. Under 0.1 SUN illumination, the graphene-based DSSC reached a PCE of 6.87% (Figure 6b), i.e., an increase of ~35% in comparison to the PCE measured at 1 SUN. This PCE increase at 0.1 SUN is mainly attributed to a higher FF (up to 0.62), which is instead limited at 1 SUN illumination (up to 0.47) due to the high Rs of the CVDgraphene current collector. Table 2 summarizes the PV parameters of innovative DSSCs using Y123 dye and Co 2+ /Co 3+ -based electrolytes. At 1 SUN illumination, our graphene-based DSSCs have shown PV parameters approaching those of cells based on Ag/PEDOT:PSS CEs, showing an overall PCE of ~5.1% vs 7% [244]. The main difference here is related to the Ag electrode, which clearly provide lower Rs than CVD-graphene, as demonstrated by the higher FF (66% vs 47%). This effect is further confirmed by the PV performances of the cells that used GNPs or PProDOT catalysts in combination with FTO. In these cases, the low Rs of the TCO contributes in bringing the FF up to 70% and 77%, respectively, leading to PCE beyond 9%. Overall, our graphene-based CEs can be used in high-performance TCO/Pt-free DSSCs, which can compete with state-of-the-art DSSCs made with glass/FTO in combination with various kinds of carbonand polymer-based catalysts [245]. This indicates the potential of the CVD-graphene/graphene flakes dualfilms as CEs in DSSCs, providing a viable alternative to be experimented also in the context of flexible devices.

Conclusions
In this work, we demonstrated advanced Pt-and transparent conductive oxide (TCO)-free counter electrodes (CEs) for dye-sensitized solar cells (DSSCs) based on dual films of graphene materials. In particular, single-layer graphene, produced by chemical vapor deposition (CVD) using a cold-wall reactor, has been used as current collector, while graphene flakes produced by liquid phase exfoliation (LPE) effectively acted as catalyst for reducing the electrolyte. The graphene-based DSSCs have been tested in configuration using both I3 -/I --and Co 2+ /Co 3+ -based electrolytes. In the first case, the optimized DSSCs using ruthenium complex dye (C106) reached a power conversion efficiency (PCE) of 2.1% at AM 1.5 G illumination (1 SUN. In the second case, [Co(bpypz)2] 2+/3+ (bpy-pz = 6-(1H-pyrazol-1-yl)-2,2'-bipyridine) was selected as redox couple, and cyclopentadithiophene-bridged donor-acceptor dye (Y123) as organic dye [71]. The graphene-based DSSCs achieved a PCE of 5.09% at 1 SUN. Noteworthy, the DSSCs have shown even better performance under low illumination intensity condition, reaching a PCE of 6.87% at 0.1 SUN. To the best of our knowledge, these PCE values, compete with record-high values reported for DSSCs using Pt-and TCO-free CEs [238,243]. In perspective, the replacement of conventional FTO/Pt CE with all-graphene-based CEs is promising for lowering the manufacturing costs of DSSCs, as well as for developing flexible solar cells designs. Our results, coupled with the recent progress on the synthesis of metal-free organic sensitizer[14, [42][43][44][45], quantum-dot sensitizer [195,196], perovskite-based sensitizer [197][198][199], and natural dyes [14,45], confirm DSSCs as an important branch within the current PV panorama.