Tuning the Porosity, Water Interaction, and Redispersion of Nanocellulose Hydrogels by Osmotic Dehydration

Osmotic dehydration (OD) was introduced as a method to reproducibly tune the water content and porosity of cellulose nanofiber (CNF) hydrogels. The hierarchical porosity was followed by electron microscopy (pores with a >100 μm diameter) and thermoporosimetry (mesopores), together with mechanical testing, in hydrogels with solid contents ranging from 0.7 to 12 wt %. Furthermore, a reciprocal correlation between proton conductivity and the ratio of water bound to the nanocellulose network was established, suggesting the potential of these systems toward tunable energy materials.


Preparation and characterization of cellulose nanofibers
TEMPO-mediated oxidation and fibrillation. The nanofibers were produced from never-dried cellulose pulp of bleached softwood obtained from a Finnish pulp mill, consisting of a mixture between spruce and pine. The TEMPO-mediated oxidation of the pulp was carried out following the protocol describe by Saito et al., 1 using hypochlorite in alkaline conditions. After the reaction, the oxidized pulp was washed and redispersed in water to a final concentration of ca. 1 wt%. The fibrillation process was carried out twice at 1850 bar, using a microfluidizer S2 (Microfluidics Int., USA) equipped with two Z-type chambers with diameters of 400 and 100 µm. The final solids content of the CNF suspension was 0.89 wt% which was measured by evaporating the excess water overnight at 105 C in a ventilated oven.

Osmotic dehydration (OD)
PEG solutions were prepared simply by mixing the proper amount of solid PEG with water to obtain the desired wt%. The suspension was then magnetically stirred until the complete dissolution of PEG. The CNF hydrogels were prepared with a grammage of 0.639 mg cm -2 .
Firstly the 0.89 wt% CNF suspension was diluted to 0.3 wt% and homogenised at 12.4 rpm using Turrax. An aliquot of 12.6 g was taken and further diluted to 0.2 wt% and mixed by vortex. The air trapped in the suspension was removed by vacuum, after which the nanofiber suspension was gently poured into the top compartment of the OD set-up (including Spectra/por Dialysis Membranes from regenerated cellulose, molecular weight cut-off 6 to 8 kDa). The bottom compartment was filled with the same amount of PEG solutions (15.6 ml) S3 with the appropriate concentration (25, 20, 17, 15, 12.5 and 10 wt%) and stirred with a magnetic bar. The dehydration was carried out for 24 hours at room temperature. The 24 h dehydration time was chosen from a set of robust trial-and-error type of experiments where we noticed that the wt% was always stable after 24 h. Depending on the characterization technique the samples were coagulated by immersion in a 0.1 M HCl solution for 10 min, after which they were washed until neutral pH and stored in aqueous solution. HCl coagulation was used for the samples that underwent mechanical testing and proton conduction measurement, because exchanging the carboxylic counterion to proton provided better mechanical strength and enabled measuring the proton conduction. In fact, the gels with low solids contents were too weak to handle for, e.g., the compression tests. By contrast, the thermoporosimetry measurements and redispersion experiments (rheology) were done right after the osmotic dehydration, i.e. without HCl coagulation. In fact, redispersion of CNFs was not feasible when the carboxylic groups were protonated. Moreover, the excess water was difficult to remove from the HCl-treated gels, and this would have affected the amount of free water analyzed by thermoporosimetry. The solid content of the hydrogels was obtained by drying the samples right after the dehydration in the oven at 105 • C for 24 hours. The weight was recorded before and after with a balance with five significant digits. The measurement was repeated 3 times.

Characterization of the hydrogels
The hydrogels obtained by OD are homogenous and smooth; Figure S2 shows three of them. freezing induced by the container (e.g., metal). The cross-section was prepared by cooling the samples with liquid nitrogen and fracture it with a scalpel. Prior to image with SEM all the samples were coated with a 2 nm thick layer of Au/Pd. Figure S3 shows a complementary image to the ones in Figure 1C. Here, a higher resolution image from a 0.7 wt% hydrogel is shown, i.e., the same sample as in Figure 1C (left hand side).
It reveals the compact structure of a thin (~2-3 µm) single wall separating the large, ~100 µm pores. The masses of the sealed crucibles were recorded before and after the measurements to ensure that the Al pans remained sealed during the measurement. The temperature was first brought to -50 °C at 20 K min -1 to crystalize all the freezable water in the samples. The temperature was then increased to -0.2 °C and held constant until the melting transition was completed i.e., until all the confined water melted. This step is essential to prevent supercooling during the subsequent recrystallization step. The temperature was then decreased at 2 K min -1 to -50 °C.
The freezing bound water (FBW) was calculated by integrating the resulting exothermic peak.
In the next step, the temperature was increased to +60 °C at 10 K min -1 and the total freezable water (TFW), sum of bound and free water, was calculated by integrating the resulting endothermic. A typical DSC curve of the hydrogel with 12 wt% CNF is shown in Figure S4A. S5 The TFW is shown in Figure S4B and Figure S4C shows the cumulative pore volume of the hydrogels.  to the cross-section using a 10 mm blade to a thickness of 320 ± 0.9 µm. The thickness was measured using a Sony IMX219 camera (12.3 Megapixels), according to the procedure reported in literature. 3 Experimental data were fitted using Randles circuit (inset in Figure S5).
The ionic conductivity of the samples was calculated using Equation 1, σ= eq.1 in which σ is the ionic conductivity, L represents thickness of the samples, Rp and A are ionic resistance and contact area between the electrodes and the hydrogels. The impedance curves are shown inf Figure S6. Figure S6. Impedance curves of the CNF hydrogel from which the conductivity was calculated.
Inset: Randles Circuit is in which Rs represents the bulk resistance, Rp the ion diffusion resistance and CPE the constant phase element.

Redispersibility of the osmotically dried hydrogels
After the 24 hours of dehydration, the hydrogels (0.7 -12 wt%) had been directly redispersed in deionized water up to the initial concentration of 0.2 wt%, after which they were gently stirred at 128 rpm with a magnetic bar for 24 hours. and all samples were allowed to relax for 2 minutes under the measuring head prior to the measurements. G' and G'' were obtained from the linear viscoelastic region in the stress curves. Onset of nonlinear behavior (ONL) was defined as the first G' data point to deviate over 5% from the linear region. Critical stress (CS) was obtained from stress-strain curves, as the first point where the strain begins to exhibit nonlinear behavior (over 10% deviation from slope = 1). Both ONL and CS are shown in Figure S7A.
Contribution of nanofiber aggregation. The redispersed hydrogels were centrifuge at 4500 rpm for 1 hour. The supernatant (1 ml) was collected, and its solid content was compared to the one prior centrifugation. Both were measured by a thermogravimetric analyser (TGA), model Q500 from TA Instruments, using a heating rate of 10 °C min -1 up to 105 °C and isotherm of 10 minutes. The standard deviation of this method was measured on triplicates.
The samples were compared to the suspension that was not dried at all and to the one dried at 105 °C in the oven. The difference between the concentrations before the centrifuge and the top 1 ml supernatant gives a measure of the redispersion degree since the eventual presence of aggregates at the bottom of the centrifuge tube will deplete the nanofiber concentration. This difference is shown in Figure S7B. All values of hydrogels prepared by OD lay very close to the standard deviation of the measurements (red line), meaning that the difference is just above the method uncertainty.
Thus, all OD samples show very good redispersion and very small aggregation. Instead, the data scatter with the sample prepared by oven drying is significantly higher, indicating strong aggregation and worse redispersion.