Ecological, flexible and transparent cellulose-based substrates without post-production treatment for organic electronic devices

In the past few decades, technological advances have aroused the interest of industries and consumers for flexible electronic devices. However, the substrates currently used, such as glass and polyethylene terephthalate (PET), present problems regarding their performance and destination, since the first is difficult to handle and the second comes from non-renewable sources. Common properties required in substrates to provide their use in organic electronics are flexibility, stability and sufficient transparency. Therefore, as a sustainable and efficient alternative, the present study aimed to develop a totally cellulose-based substrate, a natural abundant polymer that presents thermal stability, mechanical strength, recyclability and is biodegradable. Different substrates were produced using microfibrils from Eucalyptus sp. A pure microfiber substrate weighing 25 g m−2 was obtained by the vacuum filtration method and paper-forming machine. The other four substrates were obtained by the casting method containing cellulose acetate matrix and freeze-dried microfibrils reinforcement at different concentrations. In addition, a substrate containing 1.0% of the suspended microfibrils as reinforcement in the cellulose acetate matrix was produced. A conductive thin film of poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS) was deposited by air-brush technique as an electrode to evaluate the electrical performance of the substrates. The obtained films were characterized by their optical, thermal and morphological properties, showing a great potential to be used as substrate in organic electronic devices, being applied for an ethanol gas sensor device.


Introduction
Flexible electronics have been developed in the past few years as an emerging technology. Gathering technology and sustainability is a challenge, since energy harvesting, gas sensing and other electronic devices are fabricated to try to reduce carbon emission using environmentally friendly materials from the fabrication process to the product afterlife [1,2] Thermoplastic substrates have been used in the fabrication of electronic devices because they have flexibility, reduced cost and enable the roll-to-roll processes. One of the most used materials is polyethylene terephthalate (PET), a well-known polymer due to its interesting mechanical and optical properties. Nonetheless, PET is made from petrochemical sources, and it may cause serious environmental problems, since it presents resistance to air, water, microorganisms and solar radiation, resulting in a slow degradation process, together with the difficult recycling process and the growing demand [3,4].
In order to reduce environmental impact and to cooperate with a sustainable carbon cycle, there is a necessity to develop alternatives based on vegetal and biodegradable resources to be used as substitutes for conventional substrates such as PET. Cellulose is one of the potential substitutes since it is a natural and abundant raw material that can be recycled, has almost zero toxicity and attractive production cost. Moreover, cellulose can be easily converted to membrane and films, which present morphological and mechanical properties compatible with substrates used in organic electronic devices [5][6][7][8]. However, one challenge is to produce cellulose membranes with transparency enough for application as substrates on electronic devices, enabling the light entrance in solar cells or light emission in organic light-emitting diodes.
An alternative for producing transparent substrates with the desired properties is a chemical reaction of cellulose acetylation. Cellulose acetate (CA) is a transparent thermoplastic that presents low toxicity, mechanical and electrical properties that attract its use in electronic development. Nevertheless, this material presents relatively high hydrophobicity that limits its use, since modern materials used in the device's fabrication, are waterbased, also in accordance with sustainability. One example is the poly (3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS), a conductive ink, which is a waterbased polymeric dispersion used to prepare supercapacitors [6], active layers in gas sensors [9,10], hole transport layer or electrode in organic solar cells [11][12][13], and as hole injection layer in organic lightemitting diodes [14,15]. Cheng et al [16] showed that it can also be combined with cellulose derivatives to produce porous systems to be used as a printable conductive ink.
Substrate hydrophobicity can be decreased by incorporating nanocellulose (NC) in the matrix [17]. NC is obtained from the cellulose fiber and is used in nanocomposites as a matrix or reinforcement [18,19]. Properties such as flexibility and low thermal expansion coefficient make it a potential substitute for non-renewable substrate sources [20]. Po-Chen Lin et al [21] reported a membrane composed by cellulose nanofibers and silver nanowires to be used as substrate for photovoltaic devices. The authors showed a transparent membrane with optical transmittance up to 78% and low sheet resistance. Other interesting result to observe is the low coefficient of thermal expansion (CTE), that is important since it can cause deformations under temperature variation, hindering some applications. Indeed, cellulose-based substrates have revealed coefficient of thermal expansion in the order of 10 -6 /K, being possible to decrease it as reinforcement matrix are added. It is interesting to note that in the present study it may not be a problem, since PEDOT:PSS presents values in the same order of magnitude of cellulose-based substrates [22]. Compared with other flexible substrates, common PET membranes used in organic devices present CTE around 10 -5 /K, one order of magnitude higher than the reported for cellulose.
Regarding the membrane transmittance, Legnani et al. [22] used a biocompatible membrane based on bacterial cellulose as substrates for organic lightemitting diodes. The authors obtained an opaque bacterial cellulose membrane that became transparent after depositing a system composed of boehmite nanoparticles and epoxy modified siloxane. Yagyu et al. [23] showed that it is possible to obtain a transparent cellulose substrate, up to 90% of optical transmittance, by acetylating a cellulose nanopaper using low-polarity solvents such as toluene. This procedure was carried out as a surface treatment after obtaining the cellulose nanopaper. Despite the success achieved by those studies, the methods may not be economically feasible, simple to execute, or eco-friendly processes.
Cebrian et al. [24] presented a method to enhance the optical transmittance of bacterial cellulose membranes. The procedure consists of coating a solution based on recycled petrochemical plastics, such as polystyrene (PS), and d-limonene, an ecological solvent extracted from orange. The optical transmittance, important for application as OLED's, drastically increased from the pristine to the treated, corresponding to 46% of enhancement at 550 nm. This enhancement was attributed to the incorporation of PS into the bacterial cellulose's three-dimensional matrix, resulting in the light scattering reduction and, due to the different refractive index, rinsed the optical transmittance in the visible range.
Though the discussed studies showed the production and application of cellulose-based membranes in organic devices, regarding the mechanical and optical properties, it has not been discussed the possibility of producing such membranes without the need for post-production treatment. In this scope, the present study aims to develop a transparent, flexible and sustainable substrate totally based on cellulose to be used in flexible organic electronics and that does not require surface treatments to achieve comparable transmittance values. The use as substrate paves the way in the electronic development since the device only requires that the material does not degrade when in contact with the device's deposited layers. As mentioned, one of the most used materials as the electrode in those devices is PEDOT:PSS, since it presents flexibility, processability, optical and electrical properties that enable its use in organic electronics in general. The conductive ink is also an ecofriendly material because it is a water-based polymer. Thereafter, in addition to the cellulose-based substrate development and characterization regarding the electrical, optical and morphological properties, its use and stability were also evaluated when a thin PEDOT:PSS layer is deposited. The properties of the substrates together with the conductive polymer were also evaluated to observe the stability and possibility to establish its use as a potential substitute to PET in organic electronics. Zhao et al [25] showed that the hydrogen bonds of cellulose and PEDOT:PSS improve the stability of the films, which can be used for applications such as sensing devices.
As explored, PEDOT:PSS can be used as active layer in organic sensing devices. Aiming an application, the conductive membrane was applied to monitoring ethanol in a confined environment using the conductive polymer to interact with the alcoholic vapor. Ethanol is a flammable gas that can be toxic depending on its concentration. The most challenge on currently research is producing devices that does not require high working temperatures that can vary from 150°C to 200°C [26]. So, the possibility to apply such devices using room temperature can help the energy economy and avoid high temperatures that might degrade the devices.
2 Experimental procedure

Materials
Cellulose microfibrils were obtained using the cardshaped industrial bleached pulp Kraft, using the Eucalyptus sp. wood, from Suzano Papel e Celulose company. PEDOT:PSS aqueous solution was acquired from Sigma-Aldrich (1.3% wt. of polymer in water, 500 lg L -1 of PEDOT and 800 lg L -1 of PSS). Cellulose acetate (CA-419028, number average molecular weight Mn * 50,000 determined by gel permeation chromatography and density of 1.3 g/ mL at 25°C) was acquired from Sigma-Aldrich and acetone (P.A, 99.5%) from Neon.

Cellulose microfibrils preparation
44.43 g of pulp Kraft was manually fragmented and added in a blender together with 2 L of distilled water. After 5 min, a homogeneous fiber suspension was obtained with a concentration of 2% of dried pulp in water, considering the pulp humidity (9.98%) determined by gravimetric analysis. Calculation details are presented in the Supporting Information (SI). The suspension was then treated using the mechanical defibrillation process in the microprocessor mill from Masuko Sangyo Co., model MKCA6-3. The mill set parameters were 0.1 lm and rotation of 1500 rpm with 5 cycles and consistency of 2%. Those parameters were established based on previous studies [27], wherewith this configuration it is found a viscosity of around 10 mPa S and a crystallinity degree of 75%. In the cited study, it was observed that as the number of cycles increases, both viscosity and crystallinity degree decrease, inferring that it also decreases the polymerization degree, and 5 cycles are part of an optimized configuration. Before, the material presented a liquid aspect and, after passing through the mill, the resulting material presents a gel aspect. Cellulose microfibrils suspension, named nanocellulose (NC), was stored in a plastic flask, covered and kept at 4°C to avoid degradation.

Preparation of the membranes (A) Nanocellulose films
Nanocellulose thin films were prepared with 2% of total solid content based on the literature [27]. The procedure comes from the protocol 2540 G from APHA [28], and it is described in the SI. The 2% of total solid content was determined, and nanocellulose films were obtained with a grammage of 25 g m -2 with a total area equal to 1.22 9 10 -2 m 2 . This grammage value is based on previous studies [27]. Details of wet mass determination to obtain this grammage value can be found in SI.
Briefly, 15.09 g of nanocellulose and 70 mL of deionized water were added to a beaker and manually homogenized with a glass stick. The dispersion was deposited on a nylon mesh on the top of a filter paper, both coupled to a Bü chner funnel to perform vacuum filtration. After the vacuum filtration, the nanocellulose film, paper filter and nylon mesh were put together in the paper maker machine with a Rapid-Kö ethen apparatus, plus one more nylon mesh, paper filter and metal grid on the top of previous setup to cover and protect the material. Nanocellulose films were prepared under constant pressure of 50 kPa at 90°C for 25 min. After the drying process, a translucid nanocellulose membrane with a grammage of 25 g m -2 was obtained. The membrane was named NC-25. The paper maker procedure was carried out following the International Organization for Standardization 5269-2 (ISO, 1980) and Tappi T 205 sp-02 (Tappi, 2004a) standards.
(B) Films based on cellulose acetate, cellulose acetate with reinforcement matrix of cellulose particles and cellulose suspension The films were prepared by casting method, one made of 100% cellulose acetate, named pure cellulose acetate (PCA) and the other three with different cellulose mass concentrations equal to 0.5%, 1.0% and 1.5% in a matrix of cellulose acetate. These films were named CA-0.5cp, CA-1.0cp and CA-1.5cp, where CA is cellulose acetate, cp is cellulose particle (freezedried microfibrils), and the numbers corresponds to the concentration of cellulose in the cellulose acetate matrix. In addition to these, a membrane composed of CA and 1.0% of nanocellulose (nc) that is the microfibrils suspension was also produced, named as CA-1.0nc.
The production consists of adding 1 g of commercial cellulose acetate in 12 mL of acetone in a beaker and cover to avoid solvent evaporation. Then, the mixture was added to a shaker (Tecnal, model TE-420), operating at 260 rpm for 20 min, resulting in a homogeneous and soluble mixture. The mixture in the beaker was sonicated by ultrasound bath (Ultra-Sonic Cleaner, model USC 1450) for 20 min to avoid morphology defects in the films, such as bubbles that may harm the mechanical resistance due to microfractures. The mixture was added to a 6-cm petri dish, and then, the resultant membrane was placed in a vacuum desiccator to control the solvent evaporation, resulting in transparent PCA membranes.
To obtain the composite membranes, the cellulosic suspension (NC) was freeze-dried to remove water presented in the microfibrils [29]; thereafter, it was put in the analytic mill (IKA, model A10), and the resulting powder sample was named as cellulose particle (CP). Then, 1 g of CA was mixed with 0.01 g of CP in 12 mL of acetone in a 120-mL beaker. This mixture was added to the shaker for 20 min at 260 rpm of constant rotation. The resultant homogeneous solution was sonicated in bath for 20 min to remove the bubbles and then placed in the vacuum desiccator. The resulted membrane is the CA-1.0cp. The same procedure was applied to obtain CA-0.5cp and CA-1.5cp, varying the CP mass, being 0.005 g for the first and 0.015 g for the latter composite. A membrane containing the cellulose microfibrils was prepared without passing through freeze-drying process to evaluate the water influence since vacuum desiccator methodology enables the films formation using microfibrils in suspension. It was produced membranes containing 1.0% of NC suspension, named CA-1.0nc, to compare both substrates composed of CP or NC in the same concentration.

Deposition of PEDOT:PSS layer on the substrates
Aqueous conductive ink of PEDOT:PSS was deposited by airbrush technique on the prepared substrates described before using Hobby V8-AE04 equipment, with nozzle aperture variation from 0.3 to 0.8 mm, air pressure regulation in a range of 15 to 50 Psi and using a 20-mL glass vial containing the conductive ink. The air pressure was regulated via a manometer with a filtration set. During the deposition procedure, the substrates were maintained in contact with a heating plate (Fisaton, model 503-1) accompanying a digital thermometer. The 1.5 cm 2 substrates were fixed on the hot plate, perpendicular to the nozzle aperture, distanced by 58 cm. Ten depositions of the conductive polymer were performed. Each deposition lasts 10 s, with 15-s interval between each deposition (each layer). This interval is enough to dry the solvent from the former deposited layer. Airbrush deposition was carried out with compressed air at 15 Psi, and the substrates were kept in thermal equilibrium with the hot plate at 70-75°C. It was also deposited on polyethylene terephthalate PET from Delta-Technologies and glass substrates to compare the thin film properties.

Samples characterization
Thermogravimetric analyses (TGA) were performed in an equipment from Shimadzu, model DTG-60, from 30 to 900°C with synthetic air atmosphere (50 mL min -1 ) and heating rate of 10°C min -1 . Fourprobe measurements were taken in a Jandel Universal probe equipment, using 10 nN of contact force, taking 10 measurements for each sample. Sheet resistance (R s ) values were acquired using direct and reversal electrical current for each measurement. Bend cycles were performed using a homemade tool that keeps the bend angle at 45°, where 1500 bend cycles were made to evaluate mechanical stability. The sheet resistance was measured after bend cycles steps. Electrical conductivity values (r) were determined using the relation r ¼ 1 R s t , where t is the thickness of the PEDOT:PSS thin films. The thickness values were obtained by profilometry using a Veeco Dektak 150 equipment taking 5 measurements for each sample. UV-Vis spectroscopy was carried out in a Shimadzu equipment, model UV-2450, between 300 and 900 nm and using air as transmittance blank. For samples containing thin films of PEDOT:PSS deposited on the substrates, the transmittance spectra were acquired using the respective substrates as blank. The wettability of thin films was evaluated by drop casting 10 lL of deionized water on the films deposited on the substrates, and then, the contact angle was measured using Inkscape program.
Atomic force microscopy (AFM) topographic images were acquired on a Shimadzu equipment, model SPM-9700, using doped silicon probes in semi-contact mode on the different substrates with and without PEDOT:PSS thin films. Scanning electron microscopy (SEM) images were acquired in a Tescan microscope, model VEGA 3 LMU, operating at 10 kV and with a work distance of 5 mm. Gold film was deposited on the samples for SEM analysis using a Balzers equipment, model SCD 030. The sputtering was performed for 1.5 s with an electrical current of 30 mA, resulting in a 90-nm-thick gold layer. Cross-section images were acquired using the same SEM microscope, through cryogenic fracture of samples using liquid nitrogen. Transmission electron microscopy (TEM) images were acquired in a Jeol microscope, model JEM 1200EX-II with high-resolution chamber from Orius, model SC1000B, operating at 120 kV. For TEM analysis, the nanocellulose suspension was diluted in deionized water, resulting in a low-concentration nanocellulose dispersion, and then drop casted on a 200-mesh metallic grid, dried at room temperature and atmospheric conditions on a flat surface, avoiding dust and other contaminations.
Gas sensing devices were fabricated by bonding electrical contacts on the cellulose-based membrane and PEDOT:PSS as active layer assembly with silver epoxy adhesive (model 8331S-14G fabricated by MG chemicals), followed by a drying process at 65°C in a vacuum oven for 180 min. The sensors were characterized by exposing them to ethanol vapor using a gas sensor characterization system, as presented in previous studies [30,31]. The electrical resistance of the thin films was monitored under ethanol vapor flowing through the system. Cycles of dry air were performed for 8 min followed by the mixture of dry air plus 330 9 10 3 ppm of ethanol vapor for 4 min. The percentual sensitivity is calculated by relating the initial resistance and the resistance during the exposure. To evaluate the behavior of the conductive membrane with humidity, the devices' electrical conductivity was measured under 80% and 10% of relative humidity environment using water vapor plus dry air mixture.

Results and discussion
Flexible organic electronic devices demand specific substrates properties as optical transmittance. In this study, we propose a method to obtain membranes based on cellulose acetate and cellulose particles that presents optical transmittance enough for its application. Figure 1 presents photographs of the obtained samples, showing the membranes' transparency. Cellulose nanofibrils presents optical transparency that can be enhanced using lower grammages [27]. It is desired to obtain samples with high transparency, but not a translucid film as observed in the nanocellulose paper NC-25, which may harm the optical device's performance. The obtained samples, named as PCA and CA-based, are comparable to those developed by Yang et al. [32], since the authors showed substrates with high transparency using cellulose nanowhiskers, or rod-like structures, which were modified by acetylation to become dispersible in acetone and to improve the tensile strength and Young's modulus. Despite the authors enhanced the mechanical properties, the optical transparency reduced but presented values between 60 and 80%. Figure S1, in the SI, presents SEM images of the bleached pulp Kraft before the mechanical defibrillation process, which reveals the initial dimensions, in the micrometrical range, and the interconnected fiber elements. Figure S2 shows TEM images for cellulose microfibrils obtained after the in natura fiber mechanical defibrillation process, and it is possible to observe the superficial grid organized microfibrils. After the mechanical defibrillation process, the fiber mean diameter reduces from tens of micrometers to tens of nanometers, evidencing the nano-structuring process. The observed values range is in accordance with previous reports, where the authors obtained microfibrils with diameters lower than 100 nm [27,33,34]. Figure S2 C) presents a TEM image of microfibrils with diameters varying from 6 to 30 nm, and micrometric lengths. Pääkkö et al. [35] discussed that the microfibrils present many fibrils grids with a higher number of hydrogen bonds between the chains that promote the microfibrils nano-grids use in the composite membranes, justifying its use as reinforcement in the present report.
Thermogravimetric analyses for the produced membranes and the commercial PET are shown in Fig. 2(A). There are three main thermal events, being Fig. 1 Photographs of the cellulose-based membranes produced. NC-25 is the nanocellulose film with grammage of 25 g m -2 . PCA refers to pure cellulose acetate, CA-1.0nc refers to cellulose acetate with 1.0% nanocellulose as matrix reinforcement. CA-1.5cp, CA-1.0cp and CA-0.5cp refer to cellulose acetate with cellulose particles with concentrations of 1.5%, 1.0% and 0.5%, respectively. NC-25 is translucid, while the other samples present high transparency the first occurring between 30°C and 130°C that can be related to the water loss or volatile compounds that can be adsorbed by CA and by cellulose reinforcement [36]. The maximum mass loss occurs in the second thermal event, between 210°C and 360°C, that can be associated with the degradation of the cellulose acetate polymer chains [37]; the third event occurs between 330 and 470°C and is related to the cellulose reinforcement polymer chains rupture [38]. After 470°C, the decomposition is slow and is related to the combustion of the remaining products [36]. PET sample has two main thermal events, it starts to degrade at 282°C up to 427°C, attributed to the presence of volatile impurities, as additives such as diethylene glycol, used during the synthesis. The second thermal event is around 427 to 525°C, related to the rupture of ester links presented along the polymer chain, resulting in the PET polymer degradation and both are in accordance with the literature [39,40].
For PCA and CA-based samples, the first thermal event is responsible for 3 to 8% of mass loss, being the second around 70% of mass loss, except the NC-25 sample, which corresponds to 48%. The third process is responsible for 23% of mass loss for PCA and CAbased substrates, and for NC-25 corresponds to 40% loss. For PET samples, the first stage corresponds to 75%, and the second one for 23% of mass loss. The remaining mass associated with the PCA is less than 1%. For the CA-based and NC-25 substrates, it is less than 5% of the remaining mass, and for PET, it corresponds to 2% of the remaining mass. Figure 2(B) presents thermogravimetric analysis comparing the cellulose acetate with cellulose particles and the pure cellulose acetate samples. The curves present the same aspect, but in the first thermal event, it is possible to notice that the CA-based substrates presented less mass loss than the PCA sample, being 3.6% for CA-0.5cp, 3.2% for CA-1.0cp, 2.3% for CA-1.5cp, and 5.9% for PCA. It can be related to the lower water content volatilized during the first stage for the samples with cellulose particles reinforcement. During the second degradation step, mass loss is intense for all samples, corresponding to 70% as cited. In the graph, it was shown a zoom that refers to the middle of the second step to highlight the shift among the curves. This shift for higher temperatures as the percentage of cellulose particles increases is related to the beginning of the degradation step, being 222°C for PCA and 238°C, 249°C and 254°C for CA-0.5cp, CA-1.0cp and CA-1.5cp, respectively. Those results suggest higher thermal stability for the substrates with cellulose particles reinforcement, as proposed by Ratanawilai et al. [41], where the authors discussed that the lower mass loss and the higher thermal stability are due to the higher interfacial adhesion between the matrix and the reinforcement. Regarding the remaining mass after the carbonization process, it can be related to the reinforcement addition in the CA matrix, because for PCA the remaining mass is 0.8%, while for CA-0.5cp, CA-1.0cp and CA-1.5cp it is 3.6, 4.1 and 4.9%, respectively.
Graphs in Fig. 2(C) relate the PCA, commercial PET and the sample composed of nanocellulose with a grammage of 25 g m -2 (NC-25). PCA and NC-25 presented higher water loss and volatile materials at lower temperatures, being 5.9 and 8.1%, respectively, while PET presented stability until 282°C. Furthermore, PET presented stability about 60 and 73°C higher than PCA and NC-25 samples, respectively. Related to the remaining mass after the carbonization to ashes, the commercial PET presented 2.4%, while the ecological samples PCA and NC-25 presented 0.8 and 4.6%, respectively. Those results indicate that PCA presented around three times less residue at the end of the burning process, in relation to PET substrate and NC-25 presented 2 times more residues than PET. Despite the results suggesting that PET samples are more stable for use, Miranda Vidales et al. [39] showed similar results obtained for a resinbased substrate that presents high thermal resistance and can be used in many applications.
Graphs in Fig. 2(D) present a thermogravimetric analysis comparison between the cellulose acetate with 1% of cellulose particles (CA-1.0cp) and 1% of cellulose microfibrils (CA-1.0nc). CA-1.0nc presented a 26% higher mass loss in the first degradation step when compared to CA-1.0cp. This result can be related to a higher water quantity adsorbed since the reinforcement used during the substrate making was in a water suspension. Moreover, the film containing cellulose particles presented higher stability, since CA-1.0cp started the second degradation step 30°C after the CA-1.0nc sample. During the second degradation step, both curves are overlapped and the third step starts at 354°C for CA-1.0nc and 357°C for CA-1.0cp. In the end, the nc-based substrate presented 43% less solid residue compared to the cpbased sample. In that scope, the nano-cellulosic suspension used as reinforcement instead of cellulose particles resulted in less solid residues. However, the thermal stability observed was lower than cp-based substrate.
Nogi et al. [42] discussed that most electronic devices are fabricated under 150°C. Therefore, based on the obtained results, it is possible to notice that the developed samples are good candidates to be used as substrates for flexible organic electronics due to the high thermal stability in the desired temperature range. Figure 3 presents the SEM images for pure cellulose acetate (A) and cellulose acetate with 0.5 (B), 1.0 (C) and 1.5% (D) of cellulose particles as reinforcement. Films with cellulose particles present similar morphology, with aspect of homogeneous and smooth surface, except for the presence of spaced points of cellulose particles, while the PCA sample does not present these points. PCA membrane presents some scratches that might be from the petri dish used during the fabrication process. The analyzed membranes do not present fissures, nor bubbles, indicating that solvent evaporation performed in the desiccator has produced films with high uniformity and low roughness. Figure S3 presents the crosssection SEM images which were used to estimate the film's thickness around 260 lm for PCA, and 290, 300 and 310 lm for CA-0.5cp, CA-1.0cp and CA 1.5cp, respectively. These values suggest that the substrate thickness is proportional to the concentration of cellulose particles used as matrix reinforcement.
NC-25, shown in Fig. 3(E), has a compact and nonporous structure with the aspect of a dense film. Nogi et al. [42] showed that when there is a dense packing of the cellulose microfibrils, the pores are relatively small, and the spaces between fibers are also small, which interferes with the light scattering and transmittance, leading to translucent membranes, as observed in the present study. Cross-section image from Figure S3 (E) reveals that this membrane is thinner than the other prepared membranes, corresponding to 26 nm of thickness, corroborating with the previous explanation. On the other hand, Zhu et al. [43] showed that densification can enhance the optical transmittance of wood samples. The authors developed films with 90% of transmittance by densification under high pressure, removing 99% of water inside the films, suggesting that despite the films being well packed and compact with small pores, this observed densification is not able to enhance the optical transmittance due to the remaining water in the membranes as shown by TGA results.
Agglomeration of cellulose particles on substrates' surface was also explained by Dufresne [44]. The author suggested that the reinforcement components agglomeration, or the self-association, is due to the high density of hydroxyl groups on the surface of the cellulose matrix, leading to a strong tendency to create hydrogen bonds with adjacent fibers, and reducing the interaction with the surrounding matrix. Figure 3(F) refers to the substrate with 1.0% of nanocellulose as reinforcement, CA-1.0nc. This film is homogeneous and does not present agglomeration on the surface, which can be associated with the fabrication method, because the CP-based substrates were prepared with powder obtained using cellulosic suspension freeze-dried and milled, while the NCbased substrate uses the suspension itself which might avoid the particles' agglomeration. The crosssection image from Figure S3 (F) shows that the sample presents a regular thickness, with a value around 270 lm, which is 30 lm thinner than the cellulose acetate with 1% of cellulose particles. Figure 4 shows the cross-section images with higher magnification; (A) to (D) are referent to PCA, CA-0.5cp, CA-1.0cp and CA-1.5cp, respectively. One can notice the homogeneity of the cellulose acetate matrix, as well as the similarity among the CA-based and the PCA substrates. The membranes are dense and present homogeneity over all the cross-section areas even with cellulose particles as reinforcement. Figure 4(E) shows the dense, compact and non-porous NC-25 substrate. Comparing image (E) with the other samples presented in Fig. 4, the differences in the bulk morphology can be seen, leading to the correlation between the optical transmittance and the morphology discussed before. Figure 4(F) shows the cross-section image for CA-1.0nc, which corresponds to the substrate with cellulose microfibrils in the CA matrix. The microfibrils can be seen enfolded by the polymeric matrix and that presents diameters higher than those shown in TEM images. The delamination process using a mechanic mill starts on the external shell of the cell wall on the cellulose fibers, and then goes to the internal shells [45]. Therefore, the microfibrils suspension resulting from the mechanical defibrillation process might present higher dimensions shown in TEM images because the milling process may not delaminate the innermost layers of the cell wall of some fibers.
The morphology of the substrate is important for application in electronic devices, especially regarding surface patterns and substrate roughness. Both fabricated and commercial samples were evaluated by AFM analysis before and after the PEDOT:PSS deposition. Figure 5 shows AFM topographic images for the fabricated substrates before PEDOT:PSS deposition. The images referent to commercial PET and glass are presented in Figures S4 (A)  The samples are homogeneous when compared to each other, while the pure cellulose acetate in (A) presents less homogeneity. It is important to note that AFM images were acquired in regions where the surfaces are smoother, and not in regions with agglomerates of cellulose particles, as seen in SEM images. For these regions, the roughness of the substrates was measured, since for some applications, surfaces with higher or lower roughness values may be more interesting. The measured and discussed values represent the results of regions evaluated in the presented images. The results obtained show that the roughness of these smooth regions on films is relatively low, with values smaller than a few tens of nanometers and vary with the change in the amount of cellulose. For the evaluated regions, it is as follows, for instance: PCA sample presents 5.8 nm of roughness, while CA-0.5cp, CA-1.0cp and CA-1.5cp present 7.4, 3.2 and 6.0 nm, respectively. Compared with images from CA-1.0-cp and CA-1.0-nc samples, it is observed that the roughness of the first sample is 25% lower than the roughness of the second, which can be associated with the scratches observed in SEM images. Moreover, the surface of CA-1.0nc is more heterogeneous when compared to the other cp-based samples.
Those differences can also be attributed to the fact that the NC-based sample is fabricated from the microfibrils in a suspension before the film formation, while cp refers to the powder sample obtained after freeze-drying the cellulosic suspension and then put in the mill, as detailed in the preparation of the membranes' section. The different procedures to obtain the reinforced samples lead to differences regarding the relationship between the constituents and the solvent evaporation, especially considering that there is water present in the nano-cellulosic suspension, which might result in differences in the membrane's surface patterns. The substrate NC-25, based on pure nanocellulose, presents the higher This membrane is about 40 times rougher than the others. This variation can be attributed to the paper formation procedure.
Compared to commercial glass and PET, the prepared films present differences in the surface pattern, with exception of NC-25, but the roughness values are similar. PET has a roughness of about 4.8 nm, while glass has about 3.5 nm. Choi et al. [46] discussed that an ideal surface to be used in organic electronics must present roughness lower than 20 nm. Regarding roughness, substrates containing pure cellulose acetate or with cellulose particles and nanocellulose suspension have presented promising morphological characteristics to be used in organic electronic devices.
PEDOT:PSS aqueous conductive ink was deposited by air-brush technique, and the resultant films were evaluated by AFM analysis. Figure 6 shows the images of the fabricated substrates with PEDOT:PSS. The images for commercial PET and glass substrates are presented in Figure S4 (C) and (D).
All the substrates present an increment in their roughness values, with the exception for the glass substrate, which was reduced from 4.5 to 2.5 nm. PEDOT:PSS deposited on glass substrates tends to be well settled due to the surface energy between glass and the ink, resulting in thin films with high homogeneity. This is the most used technique to produce electronic devices, in accordance with the literature [47]. After the PEDOT:PSS depositions, the roughness values increased for the evaluated samples but, excepting NC-25, remained lower than 20 nm as follows: 2.8X for PCA, 2.1X for CA-0.5cp, CA-1.5cp and NC-25. For CA-1.0nc the value increased 3.6X and for PET 2.5X. For the samples with NC or CP together with cellulose acetate, the roughness values are lower than for PCA substrate, and substrate CA-1.0cp presented the lower roughness variation after depositing PEDOT:PSS, corresponding to an increment of about 1.3X.
Optoelectronic devices require electrical and optical properties that allow and enhance their operation. For OPV's the substrate and electrode need to present optical transmittances that enable the light to achieve the active layer and promote the energy conversion. In addition, the electrode must present a relatively high electrical conductivity to collect the charge carriers and distribute to the external circuit.
Sheet resistance values obtained with four-probe measurements were conducted on the PEDOT:PSS  Table 1.
Substrates containing 0.5 and 1.0% of cellulose particles present Rs values lower than substrate containing only cellulose acetate. CA-1.5cp presents values close to the PCA. All the PEDOT:PSS thin films presented uniformity along with the substrates due to the low value variations.
Thin films deposited on the substrate containing 1.0% of nanocellulose suspension reinforcement in the cellulose acetate matrix (CA-1.0nc) present sheet resistance values 50% higher than the substrate containing 1.0% of cellulose particles (CA-1.0cp), showing the great potential of nanocellulose as reinforcement to flexible substrates. Concerning commercial substrates, conductive films deposited on glass and PET, have presented mean Rs values of (8.1 ± 0.8) and (2.0 ± 0.2) kX sq -1 , respectively. Substrate NC-25 presented the highest Rs among the obtained samples, but lower than glass substrate, which is commonly used for flexible devices  fabrication. The values are also comparable to the PET, with only the CA-1.0cp with Rs lower than PET. The literature commonly attributes better results for Rs values on glass than on PET substrates, but the obtained values can be associated with different thicknesses of the films. The calculated electrical conductivity, by relating Rs and thickness, of PED-OT:PSS deposited on different substrates is shown in Table 1. Values for the cellulose acetate with cellulose or nanocellulose particles are in the same range regarding the experimental errors, and all of the samples present higher values in relation to pure cellulose acetate, glass and the NC-25 substrates. Moreover, the thin film deposited on commercial PET presented the same values as CA-based substrates, indicating that the fabricated membranes are comparable to the current materials used in most reported electrodes and devices.
Das Neves et al. [48] reported a PEDOT aqueous ink without PSS, which was deposited via air-brush on glass substrates. The best value of electrical conductivity reported was 26 S cm -1 , which is very close to our values over cellulose-based substrates. The scope of the cited paper was to deposit an aqueous conductive ink produced from chemical routes without using the insulator polymer PSS nor other surfactants, keeping electrical properties and being semi-transparent substrates, with 60% of transmittance at 550 nm for the cited conductivity. Zhao et al. [49] presented a cellulose-based substrate with PED-OT:PSS polymerized in situ, which resulted in a high conductive paper with electrical conductivity up to 300 S cm -1 to be used as supercapacitor electrodes. Despite the promising results concerning electrical conductivity values, the substrates are totally opaque. Ko et al. [50] reported a highly conductive cellulosebased electrode containing PEDOT:PSS with electrical conductivity up to 124 S cm -1 , which was obtained after treatment with polar solvents as dimethyl sulfoxide (DMSO) and ethylene glycol (EG). As also reported by das Neves et al. [13] the treatment with those solvents can enhance the electrical conductivity by promoting changes in the PEDOT domains and in the morphology of the conductive films.
Nan et al. [51] presented a membrane based on cellulose acetate, PEDOT:PSS and graphene to be used as an ionic soft actuator. The polymer was deposited via dip casting the membrane in the aqueous conductive ink. The authors performed the same treatment of DMSO proposed in the previously reported studies, resulting in electrical conductivities up to 6 9 10 -4 S cm -1 , enough for its application. Jü rgensen et al. [52] printed a PEDOT:PSS ink treated with a secondary alcohol ethoxylate with nine ethylene oxide units as a surfactant to reduce its surface tension. The ink was inkjet printed on commercial cellulose diacetate substrates to be used as a transparent electrode for OLED's. In the cited study, the authors achieved electrical conductivities up to 97 S cm -1 on CA substrates after treating the PEDOT:PSS with DMSO. With this, the substrates fabricated in the present study, together with the conductive polymer PEDOT:PSS, presented electrical response and optical transmittance attractive to be used in organic electronic devices.
Contact angle values for water drop on the prepared membranes are shown in Figure S5. As the percentage of cellulose particles increases, the hydrophobicity of the substrate based on cellulose acetate decreases. For CA-0.5pc, CA-1.0 pc and CA-1.5cp, the values are 47.4°, 40.2°and 34.6°, respectively, while the pure cellulose acetate presents a contact angle of 80.5°and the NC-25 presents 17.4°. The values are intermediate when compared to commercial glass and PET, which present 10.4°and 81.9°, respectively. As expected, pure cellulose acetate presents a higher hydrophobicity among the fabricated samples, but when cellulose particles are added, the substrates become more hydrophilic due to the wettability properties of CP. In the paper written by Jü rgensen et al. [52], they presented the importance of the wettability property on the thin film deposition process, where the authors were concerned with modifying the conductive ink by adding surfactants. In the present study, the differences in the conductive thin film deposition on distinct substrates that do not present the same wettability properties, without treating the ink dispersion, were observed.
Those results are important to determine the deposition method to be applied because different surface energies lead to a different interaction between the conductive inks and the substrate. For instance, techniques such as spin coating and slot die coating are more difficult to be used on substrates with higher hydrophobicity due to the interaction between ink and substrate. For application, there is a need to balance those properties, since if the material presents high hygroscopic properties, it will hinder the application in environments that presents humidity. Moreover, in general, research in organic electronic fields is trying to be more ecological, meaning that it is currently being avoided solvents that are not eco-friendly. The most desired solvent to produce eco conductive inks is water. Here comes the opposite side. If the substrates present high hydrophobicity, the deposition techniques become a limitation, since the interaction between water-based inks and the substrates is not ideal for coating procedures. Furthermore, being hydrophobic can be good in some applications as a non-contaminating surface, where the main interest in those kinds of substrates is associated to self-cleaning surface. For organic electronic devices, during the deposition process or after the final device, chemical or biological contamination should be avoided using high hydrophobic substrates [53]. In this present study, it cannot be in the superhydrophobic limit, but in a half-term to promote both deposition and avoid contaminations or swelling effects. As a commercial issue, the slot die coating is the closest technique to large-scale application. With this, it is possible to enhance the optical transmittance, the mechanical properties, thermal stability and enable the use of deposition techniques used in large-scale approaches by adding cellulose particles in the cellulose acetate matrix, revealing again the potential of these films as substrates for flexible organic electronic devices. Figure 7(A) presents the transmittance spectra in the UV-Vis-NIR region for substrates without PED-OT:PSS thin films. All substrates presented transmittance values higher than 70% at 550 nm. Pure cellulose acetate presented a value of around 90%, comparable to commercial PET and glass substrates. It is observed that NC-25 presents a very low optical transmittance over the entire range due to its translucid properties, corroborating the SEM results, which have shown a compact and non-porous membrane, which harms the light trespassing the substrate.
It is possible to notice that as the concentration of cellulose particles increases, the optical transmittance decreases. At 550 nm, PCA presents 90% of transmittance, while the samples reinforced with cellulose particles with concentrations of 0.5, 1.0 and 1.5% present 85%, 82% and 80% of optical transmittance, respectively. Nevertheless, the addition does not substantially impact the transmittance values and even with 80%, the sample might be used as a transparent substrate as was also observed by Yang et al. [32].
Comparing the substrates with 1% of nanocellulose (AC-1.0nc) and 1% of cellulose particles (AC-1.0cp), it is observed that both present close transmittance values of about 85% at 550 nm. However, the values between 400 and 900 nm for the AC-1.0cp are more constant than those for AC-1.0nc. For the first sample, the values in the cited range varied by 5%, while for the second, it varied 10%.
After PEDOT:PSS deposition, the values of optical transmittance have reduced, as expected since the conductive polymer presents an inherent blue color that rises the absorption in the visible range and consequently reduces the transparency. The spectra can be seen in Fig. 7 (B). Despite the visual decrease in the transmittance values, with exception of NC-25, all samples present values higher than 50% at  [48] by similar films on glass. A similar study reported by Valtakari et al. [54] shows that it is possible to obtain a more transparent substrate with PEDOT:PSS by spin coating the ink, enhancing the transmittance but reducing the electrical conductivity. Therefore, it is possible to notice that there is a trade-off when trying to merge both high transparency and low sheet resistance, which also depends on the conductive ink deposition procedure and on the application of the substrates. The influence of temperature on properties such as optical transmittance and electrical conductivity can be seen in previous studies [13,[55][56][57], which basically reveal that the optical transmittance does not suffer influence on films annealed on temperatures up to 80°C. The electrical resistance also remains stable when evaluated in temperatures up to 120°C. With those values, it is possible to apply the fabricated samples together with PEDOT:PSS as an ecological, flexible, semitransparent and conductive substrate to be used in organic electronic devices.
To evaluate the mechanical properties of the conductive membrane the sheet resistance values in function of the number of bend cycles were measured. 1500 concave bends on CA-1.0cp were performed because it is the most conductive membrane. Graph shown in Fig. S6 shows that the average values of Rs remain in the same order of magnitude when comparing without any bends and after 1500 cycles, revealing that the thin film on the substrate presents mechanical resistance enough for keep the full electrical performance.
Other stability measurement consists of evaluating the electrical conductivity in different relative humidity environments. The CA-1.0cp membrane with PEDOT:PSS was exposed to 10% and 80% RH conditions. The electrical conductivity results are 13.2 S cm -1 at 10% RH and 12.8 S cm -1 for 80% RH, which shows a very subtle difference, representing only 3% from the lower to the highest percentage of water in the environment. This result reveals that the samples are stable under high RH percentages.
CA-1.0cp was exposed to ethanol using PED-OT:PSS as active layer to monitor the presence of this flammable alcoholic vapor. The device was exposed to a saturated environment as a concept proof of possible application of these membranes in further organic electronic devices studies. The characteristic curve is shown in Fig. S7, where one can notice a sensitivity around 6% on detecting ethanol. This is a stable curve in a confined space environment under room temperature, where the result for sensitivity is comparable to other recent study reported in the literature [58].

Conclusion
The present study reported a methodology for obtaining flexible and transparent cellulosic substrates, as well as relevant and interesting results for potential use in organic electronics. Starting with the mechanical defibrillation process, the microfibrils were used to produce an initial membrane with a weight of 25 g m -2 , which had a translucent appearance but was not transparent, compromising the transmittance. To overcome this limitation, a substrate of pure cellulose acetate and cellulose acetate reinforced with cellulose particles (previously dehydrated) at different concentrations of 0.5%, 1.0% and, 1.5% was then produced. To compare the methodology, a cellulose acetate substrate was produced with 1.0% nanocellulosic suspension as a reinforcement matrix, which resulted in a lower thickness when compared to the same substrate containing cellulose particles. In addition to thickness, another important factor for organic electronics is the roughness of the substrate, which was investigated by using the AFM technique, showing that the samples are smooth, even with cellulose particles as matrix reinforcement, while these values increase when the reinforcement is made by the nano-cellulosic suspension. Thermal stability investigated by thermogravimetric analysis showed that the fabricated samples present thermal stability up to 200°C, enough for usability as substrate for organic devices. The contact angle between a water drop and the samples showed that it is possible to improve the wettability of the substrates increasing the concentration of cellulose particles or by using nanocellulose as reinforcement matrix. Due to the fabrication method, from the mixture to controlling the solvent evaporation, the substrates presented optical transmittance higher than 70% at 550 nm, without needing post-production treatment, and the resulted substrates presented a homogeneous morphology, investigated by SEM technique. Cross-section images showed that the cellulose particles are well adhered to the cellulose acetate matrix, and the nc-based substrate presented microfibrils enfolded by the polymer, while the poor transmittance of NC-25 samples can be explained by its bulk morphology. Aqueous conductive ink PEDOT:PSS was deposited by airbrush technique and resulted in a conductive thin film with electrical conductivities up to 15 S cm -1 and after the deposition procedure, the optical transmittance values reduced to 50%, where further studies will indicate the optimization procedure to deposit conductive polymers with high transparency and electrical conductivity. The ecologic samples were compared to commercialized PET and glass substrates, being considered potential candidates to be utilized as substrates for organic electronic devices. The CA-1.0cp was applied to ethanol detection with PEDOT:PSS as active layer, which represented 6% of sensitivity, reassuring the prospect of this ecologic and conductive membrane in organic electronic devices.