*2.1. Phosphorylation*

Phosphorylation of the two selected cellulosic source materials of this study—cotton linters (CL) and eucalyptus prehydrolysis kraft pulp (hwPHK)—was accomplished using concentrated phosphoric acid. Triethyl phosphate was used as solvent and phosphorus pentoxide served as a reactive binder for water released during esterification. Microwave-assisted pressure digestion in HNO3/H2O2 and subsequent ICP-OES analyses confirmed that the desired low degrees of phosphorylation were obtained. While the degree of substitution by phosphate moieties (DSP) was 0.18 for the cotton linters sample (corresponding to a phosphorous content of 33.4 g kg−1; Table 1), it was 0.24 for hwPHK (49.1 g kg−1). At this low degree of phosphorylation, the cellulose derivatives were proven not to be water-soluble yet. This is a prerequisite for cellulose coagulation from solution state, such as in NMMO/H2O or TBAF/DMSO, triggered by the addition of the anti-solvent water. Dissolution in aqueous media is also not desired for any use as cell sca ffolding material. 31P NMR as exemplarily conducted for the hwPHK sample confirmed the expected selective substitution of the primary alcohol groups in C6 position of the anhydroglucose units (data not shown).


**Table 1.** Phosphorous contents, DSP values (degree of phosphorylation), and selected properties of 3% cellulose phosphate lyogels and aerogels compared to the non-derivatized starting materials. The range of variation represents the 95% confidence interval.

Previously, we have shown that a wide range of cellulosic source materials can be dissolved up to 3 wt.% cellulose content within comparatively short dissolution time [36]. Optically clear solutions of su fficiently low viscosity, and hence, good workability, were obtained. Targeting solution casting, a solid content of 3 wt.% was envisaged for the cellulose phosphates of this study, too.

*Molecules* **2020**, *25*, 1695

Like their non-derivatized counterparts, the two types of cellulose phosphates could be easily dissolved at the target concentration of 3 wt.% in DMSO that contained 16.6 wt.% TBAF and 0.95 wt.% H2O. Visually and microscopically clear solutions were obtained within 4 h of dissolution at 60 ◦C. Solution casting of the comparatively low-viscous solutions and subsequent submersion of the molds in water a fforded self-standing transparent hydrogels. According to common practice, the hydrogels were subjected to a series of solvent exchange steps aiming to incrementally increase the ethanol content up to 100% to prepare the gels for supercritical carbon dioxide (scCO2) drying. Since dimensional stability is a key property for many applications, and particularly challenging to achieve for ultra-lightweight biopolymer-derived polysaccharide materials, possible swelling and shrinking throughout all steps of the aerogel preparation procedure were assessed. It was shown that coagulation and hydrogel formation, respectively, occurred across the entire cast solution as evident from the shape and dimensions of the free-standing hydrogels obtained. However, slight shrinkage by syneresis was observed when the gels were left in the PTFE molds, such as for 72 h. Interestingly, this was not the case when cardboard molds were used [40]. Immersion of the CL and hwPHK hydrogel samples in water of incrementally increasing ethanol content (50%, 75%, 96%, and 100%) caused the gels to shrink by up to 18 vol.% (Figure 1, Table 1). This is in agreemen<sup>t</sup> with previous work [36] and has been observed for other cellulose II gels too. Examples include gels formed by water-induced coagulation of cellulose from solvent systems like NMMO·H2O, [EMIm][OAc]/DMSO, or Ca(SCN)2·8H2O/LiCl [32].

**Figure 1.** Change in volume of the gels during regeneration in ethanol followed by supercritical carbon dioxide (scCO2) drying (-P indicates phosphorylated samples). Error bars indicate the 95% confidence interval.

Interestingly, both types of hydrogels prepared from the respective phosphorylated cellulosic materials su ffered from less shrinkage during the solvent exchange sequence than their P-free counterparts (Figure 1). This is presumably due to the presence of negatively charged phosphate groups on the surface of cellulose II nanoparticles that form and aggregate spontaneously during coagulation to a fford three-dimensional networks [7]. Repulsive forces and possibly steric hindrance can partially impede supramolecular arrangemen<sup>t</sup> of cellulose chains. This could explain why virtually no shrinkage—in case of hwPHK-P even minimal swelling—was observed after completing the solvent exchange (Table 1). Significant shrinkage of all gels, however, was caused by the final scCO2 drying step. In sum of the entire solvent exchange and scCO2 drying process, the volume loss for the resulting aerogels was about 39–46% for the non-derivatized cellulosic materials and 28–31% for their phosphorylated counterparts. A similar extent of shrinkage (29 vol.%)—occurring mostly during the final scCO2 drying step—has been recently reported for anisotropic cellulose II gels. The latter were obtained by self-assembly of cellulose (3 wt.%) in super-cooled 1,1,3,3-tetramethylguanidinium acetate under the impact of decelerated antisolvent infusion [28]. As discussed earlier [12], shrinkage is governed by solvent-polymer as well as polymer-polymer interactions. They can be quantified by cohesive energy *E*coh or cohesive energy density *e*coh when *E*coh is related to unit volume. The square root of *e*coh and Hildebrand solubility parameter δH, respectively, is frequently used to predict solvent–polymer interactions. Since *E*coh represents the sum of contributions by dispersion forces, permanent dipol-dipol forces, and hydrogen bonding, any variation of polymer-solvent composition potentially impacts these interactions. While the changes in composition throughout the lengthy solvent exchange procedure are rather small, scCO2 drying causes rapid changes in composition of the interstitial fluids. This is most pronounced in the initial stage of drying. Molecular dynamics simulations have recently shown that the solubility parameters change significantly with enrichment of scCO2 by co-solvents, such as ethanol. It changes in a reverse way when the volume fraction of co-solvent is reduced [41]. For pure scCO2 (and pure co-solvent), δH decreases with temperature and rises with pressure and density. Addition of co-solvents, such as ethanol, increases δH significantly, boosting with the amount of added co-solvent. Considering the di fferent stages of scCO2 drying, i.e., i) CO2 pressurization, ii) transition from liquid to supercritical state causing considerable changes in density, iii) formation and elution of a scCO2-expanded ethanol phase of rapidly decreasing ethanol content, and iv) final depressurization, it is evident that this part of the aerogel preparation is the most sensitive part with regard to shrinkage, specifically for ultra-lightweight aerogels. Beyond that it is assumed that small quantities of structural water—which acts as a softener for the cellulose II networks—are released towards the end of the drying procedure. This is caused by condensation of labile hydroxyl-groups which seems to start already when the gels are transferred to absolute ethanol since most of the gels sti ffened slightly during this stage, comparable to hornification of cellulose at elevated temperature.

According to the di fferent extent of shrinkage throughout the solvent exchange and scCO2 drying procedure, somewhat lower bulk densities were obtained for the cellulose phosphate aerogels. While the values of the latter were largely similar (47 vs. 50 mg cm<sup>−</sup>3), that of their counterparts from non-derivatized cellulose varied to a larger extent and ranged from 58 (CL) to 71 mg cm<sup>−</sup><sup>3</sup> (hwPHK, Table 1).

### *2.2. Chemical Integrity of Cellulose Phosphates during Dissolution*

In an attempt to verify whether the chemical integrity of the cellulose phosphates was preserved throughout dissolution, both CL-P and hwPHK-P were subjected to elemental analysis prior to and after dissolution in TBAF·H2O/DMSO (20 ◦C, 16 h). The results revealed a significant loss of phosphate groups for both of the samples. While for CL-P, the DSP value decreased from 0.18 to 0.13 (−28%) and was more pronounced for hwPHK-P (−55%; DSP 0.209 vs. 0.095). Kinetic studies, as exemplarily conducted for the hwPHK-P sample, showed that the losses occur mainly in the initial stage of dissolution since the DSP values remained largely constant after two hours of dissolution time (Table 2).


**Table 2.** DSP values of hwPHK-P (initial DSP = 0.209) after extraction with DMSO or a solution of 16.6 wt.% TBAF in DMSO having a water content of 0.95 wt.% (labeled TBAF/DMSO) at room temperature and for di fferent time periods. The range of variation indicates the 95% confidence interval (*n* = 3).

This suggests that for hwPHK about 55% of the presumably introduced phosphates were not covalently bonded but resisted being trapped in the phosphorylated cellulose. Even if released during dissolution in TBAF/DMSO, it was trapped again upon water-induced coagulation, but was then largely removed during the various solvent exchange steps. This conclusion is supported by the fact that identical final DSP values (0.096) were obtained after 8 h of dissolution time at both room temperature and 60 ◦C. Aiming to exclude interference of TBAF, the hwPHK-P dissolution experiment was repeated, however, using DMSO as solvent only (4 h, room temperature). Following repeated washings with ethanol and vacuum drying, a remaining P content of 71% was found. Altogether, this implies that the used washing procedure after phosphorylation (*n*-hexanol, ethanol, water) as proposed elsewhere [42] is not e fficient enough. The results were also a motivation to re-check similar hwPHK-P retention samples of a previous study for any possible reduction of DSP. These samples had been used for preparation of potential cell sca ffolding materials via the *Lyocell* route [39]. Cellulose was here dissolved in molten *N*-methylmorpholine-*N*-oxide monohydrate (NMMO·H2O) at 100 ◦C using propyl gallate and *N*-benzylmorpholine-*N*-oxide (NBnMO) as stabilizers. Repeating dissolution of the hwPHK-P retention samples (DSP = 0.25) in NMMO·H2O at 110 ◦C, subsequent coagulation by ethanol and final scCO2 drying revealed that also in this dissolution/coagulation approach the DSP decreased; however to a lesser extent (33%, DSP = 0.17). A DSP value of 0.22 was obtained (13% loss) when the hwPHK-P sample was extracted with ethanol only at room temperature.

### *2.3. TBAF Content of Cellulose Phosphate Aerogels*

Even though TBAF/DMSO is considered a direct, non-derivatizing cellulose solvent, selected aerogel samples were subjected to nitrogen and fluorine analysis by energy-dispersive X-ray spectroscopy (EDAX). While for CL and hwPHK aerogels low nitrogen (and fluorine) contents close to the detection limit were determined (0.05% N, data not shown), the latter were significantly higher for their phosphorylated counterparts (Figure 2A). Independent of the cellulosic source material and degree of phosphorylation (CL-P DSP = 0.18; hwPHK-P DSP = 0.24), respectively, significantly higher values of 0.74% N were obtained. Considering both the nominal DSP values of the two phosphorylated cellulosic source materials (DSP = 0.2 on rough estimate) and the real DSP ones (DSP = 0.09 = 45% of the nominal DSP), it can be concluded that about all covalently introduced phosphate moieties carry one ammonium counter ion. The over-proportional residual content of fluoride ions is di fficult to explain. Likely reasons could be remnants of phosphoric acid competing with fluorine for TBA cations or partial degradation of TBAF (triggering release of tributylamine, but-1-ene and the thermodynamically stable HF2 − ions; cf. above) and preferred removal of the nitrogenous compounds during solvent exchange and scCO2 drying.

Covalent immobilization of both nitrogen and fluorine, however, can be ruled out since repeated washing of the samples with deionized water (3×) a fforded EDAX spectra free of any N and F signals (Figure 2B).

**Figure 2.** EDAX spectra of aerogels obtained without (**A**, non-derivatized CL) and after (**B**, phosphorylated CL) implementing a H2O washing step prior solvent exchange and scCO2 drying.
