**4. Discussion**

#### *4.1. Nanoparticle Dispersions*

The extraction of nanoparticles from cotton linters and corn starch was performed by sulfuric acid hydrolysis followed by neutralization with sodium hydroxide. To reduce the amount of hydrolyzed cellulose residues in the CNC product, a comparably long hydrolysis time of 3 h was chosen. The achieved gross yield was still >40 wt %. Exemplarily, other studies addressing the extraction of CNC from cotton linters reported gross yields of 52.7 wt % after 45 min at 45 ◦C (64 wt % H2SO4, 1:17.5 g mL−1) [52] and 54.4 wt % after 5 min at 45 ◦C (60 wt % H2SO4, 1:20 g mL−1) [53]. Similar short hydrolysis times in combination with the here presented extraction method could facilitate a distinctly higher gross yield. To evaluate the degree of conversion of the raw cellulose to CNC and soluble residues, X-ray diffraction measurements could complement the process evaluation

by giving information about the product crystallinity. Analogously, the gross yield of SNP from corn starch of 32.2 ± 0.7 wt % is comparably high (15% after 120 h at 40 ◦C and 25 wt % H2SO4 [54]).

The desired high ionic strength during washing enables flocculation of the nanoparticles. Therefore, a separation from the reaction solution is possible. The precipitation-redispersion mechanism enabled the removal of more than 99.8% of the ionic residues. Consequently, the nanoparticle dispersions showed no macroscopic phase separation over several weeks at 23 ◦C. Nevertheless, the presence of ions in the dispersions is expected to promote the formation of agglomerates [55,56] effecting larger apparent particle sizes. The actual particle size could be detected by atomic force microscopy and transmission electron microscopy.

A scale-up scenario regarding process time and the consumption of chemicals (Table 4) is derived from the applied process parameters and compared to a scaling approach documented by Reiner et al. [57]. Both approaches are normalized to a CNC product mass of 1 kg based on the respective yield. The rate-determining step of our approach is the hydrolysis time of 3 h. Further steps, comprising neutralization, washing, and homogenization, require only 1 h with the used equipment. Thus, the net process time is 4 h per batch. Reiner et al. used kraft pulp as feed stock and stopped the hydrolysis after 1.5 h by dilution and subsequent neutralization with NaOH. The unit operation times are given as 8 h for hydrolysis and neutralization, 24–48 h for gravity settling and initial purification and 24 h for filtration. The CNC produced by the overall faster process presented in this work has a higher residual ion content, but a markedly lower overall water consumption (75%) and facilitates a more compact reactor volume.


**Table 4.** Comparison of the masses of chemicals required to extract 1 kg of CNC from raw cellulose, based on the neutralization of sulfuric acid.

Analogously, 1 kg of SNP is produced from 3.1 kg corn starch converted with 26.3 kg H2SO4 (18 wt %). Neutralization and washing require 81.2 kg H2O. Complete neutralization is achieved by adding 16.9 kg NaOH. Due to the protracted hydrolysis, the net process time is 121 h per batch.

#### *4.2. Nanoparticle Coatings*

PLA needed to be corona-treated prior to the application of aqueous nanoparticle dispersions onto the hydrophobic substrate by blade-coating of the coating medium without contraction. High nanoparticle concentrations and therefore increased viscosities further facilitated good spreadability of the nanoparticle dispersions on the substrates. In particular, the presence of ions induced the gel-like character of the CNC dispersion [58]. The dry coating thicknesses of both CNC and SNP coatings exceeded the targeted values on PLA. It was assumed that in both cases, non-dense layers formed on the substrate surface due to agglomeration and the presence of microparticles. Furthermore, the hydration of the salt residues as well as water absorption of the hygroscopic nanoparticles must be factored in. Same applies for the coatings on paper. The intrinsically less smooth surface of paper was assumed to additionally contribute to the deviating dry coating thicknesses. For both CNC and SNP coatings, hydrolyzed residues as well as agglomerates accumulated on the coated substrates and thereby reduced the optical transmittance of the films.

The observed effect of lower oxygen permeation of CNC and SNP coatings can be attributed to the size and the structural organization of the nanoparticles in the coating layer. The structural organization influences the diffusion path length of gas molecules in the film [59]. No improvement regarding the water vapor permeation was found. The intrinsic hydrophobicity of the nanoparticles in conjunction with the hygroscopic effect of the ionic residues is assumed to particularly impair the water vapor barrier properties. Microscopic cracks were found in starch coatings on PLA indicating embrittlement during solvent evaporation. Gentler drying conditions are not viewed as expedient. Instead, the addition of plasticizers may facilitate the prevention of cracks and lead to improved techno-functional properties [60].

Since the application of a double coating layer onto paper did not yield pinhole-free substrates, paper was excluded from further analyses. Alternatively to blade-coating, impregnating paper by dip-coating may lead to a pinhole-free substrate [5], however, accompanied by a higher expenditure of nanoparticles.

Results from other studies addressing the oxygen permeability of CNC coatings and cast films are shown in Table 5. A strong impact of the r.h., the substrate material and the nanoparticles themselves is recognizable. The *OP Q*<sup>100</sup> at 50% r.h. of coatings in the present study were in the same range as plasticized nanocellulose films. However, compared to other approaches, the *OP Q*<sup>100</sup> at 50% r.h. was up to 2 orders of magnitude higher. It was concluded that narrowing the particle size distributions of CNC and SNP by removing aggregates may be the decisive factor to further reduce the oxygen permeability of the nanoparticle coatings on polymer substrates.


**Table 5.** Comparison of oxygen permeabilities of coatings of CNC and SNP with results from other studies.
