Controlled Dispersion and Setting of Cellulose Nanobril (CNF) - Carboxymethyl Cellulose (CMC) Pastes

This work investigated the redispersion and setting behavior of highly loaded (~18 wt.% solids in water) pastes of cellulose nanobrils (CNFs) with carboxymethyl cellulose (CMC). A single-screw extruder was used to continuously process CNF+CMC pastes into cord. The adsorption of CMC onto the CNF brils was assessed through zeta potential and titration which revealed a surface charge change of ~61 % from -36.8 mV and 0.094 mmol/g COOH for pure CNF to -58.1 mV and 0.166 mmol/g COOH for CNF+CMC with a CMC degree of substitution of 0.9. Dried CNF with adsorbed CMC was found to be fully redispersible in water and re-extruded back into a cord without any diculties. On the other hand, chemical treatment with hydrochloric acid, a carbodiimide crosslinker, or two wet strength enhancers (polyamide epichlorohydrin and polyamine epichlorohydrin) completely suppressed the dispersibility previously observed for dried-untreated CNF+CMC. Turbidity was used to quantify the level of redispersion or setting achieved by the untreated and chemically treated CNF+CMC in both water and a strong alkaline solution (0.1 M NaOH). Depending on the chemical treatment used, FTIR analysis revealed the presence of ester, N-acyl urea, and anhydride absorption bands which were attributed to newly formed linkages between CNF brils, possibly explaining the suppressed redispersion behavior. Water uptake of the differently treated and dried CNF+CMC materials agreed with both turbidity and FTIR results.

Introduction M NaOH). To investigate the possible setting of CNF + CMC, different chemical treatments were assessed including two popular waterborne wet strength enhancers (polyamide epichlorohydrin (PAE) and polyamine epichlorohydrin (PAmE)) and a waterborne carbodiimide crosslinker (CDI). Treatment with hydrochloric acid (HCl), the multivalent cationic salt calcium chloride (CaCl 2 ) and the polycationic polymer polyethyleneimine (PEI) was also investigated. Similarly, the effectiveness of the setting behavior of the different chemical treatments was assessed through turbidity. Lastly, FTIR was used to elucidate the possible mechanisms behind the suppressed redispersion behavior observed for each chemical treatment.
Experimental Section/methods Materials Hydrochloric acid (HCl) solution at 37 %, calcium chloride (CaCl 2 ) powder, and branched polyethyleneimine (PEI, M W ~800) were procured from Sigma-Aldrich. Standardized 0.1 M NaOH solution was purchased from Fisher Scienti c. Picassian® XL 702, a waterborne polycarbodiimide (CDI) crosslinker with a solids content of 40%, was kindly provided by Stahl Polymers. Two waterborne crosslinking resins of polyamide epichlorohydrin (Polycup™ 9200, Lot#0002475278, 20 % solids, PAE) and polyamine epichlorohydrin (Polycup™ 7360A, Lot# 2459303, 38 % solids, PAmE), were kindly supplied by Solenis LLC. Powders of carboxymethyl cellulose sodium salt (CMC-Na) with three different degrees of substitution (D.S.) of 0.7, 0.9, and 1.2 and with a molecular weight of Mw = 250,000 (Lot# MKCK7917, Lot# MKCF4819, and Lot# MKCF8509) were purchased from Sigma Aldrich. Mechanically brillated CNFs were procured from University of Maine, Orono, ME, USA at a concentration of ~ 23.5 wt.% solids in water (Batch #122, 90% nes retained). The process of isolating this speci c type of CNFs is explained in detail by C.A. de Assis et al (Assis et al. 2018). Two TEM images of the mechanically brillated CNF used are shown Figure S1. Puri ed water was produced using a Barnstead system and used for all experiments. Unless otherwise stated, solutions of HCl, CaCl 2 , PEI, PAE, and PAmE were prepared with water.

Preparation of Highly Loaded CNF + CMC Pastes
The preparation of highly loaded CNF + CMC pastes has been explained in detail in a previous publication (El Awad Azrak et al. 2020). In brief, 52 g of CNF (at ~ 23.5 wt.%) was loaded into a high shear torque mixer (Plasti-Corder PL 2100 Electronic Torque Rheometer, C. W. Brabender, South Hackensack, NJ) equipped with Banbury type mixing blades and mixed at 55°C and 120 rpm. While the CNF was being mixed, CMC powder was gradually added until reaching a dry weight CMC:CNF ratio of 0.1:1. Water was added to adjust the paste's solids concentration to ~ 18 wt.%. The paste was mixed until complete CMC incorporation, which was signaled by a constant toque reading/plateau. The entire mixing process took on average less than 40 min and yielded approximately 66 g of wet CNF + CMC paste. This mixing process was repeated for each of the different CMC degrees of substitution (0.7, 0.9, and 1.2). The prepared CNF + CMC pastes were kept in the refrigerator to avoid any microbiological growth and used as quickly as possible. It is important to note that the word "paste" is used throughout the report to refer to these highly loaded CNF + CMC mixtures. The word "suspension" is used to refer to dilute suspensions prepared from the wet or dried pastes, respectively.

CNF + CMC Filament/Cord Processing
The prepared CNF + CMC pastes were extruded using a Brabender torque rheometer (ATR) with an attached single-screw extruder unit (L/D = 25, barrel diameter of 1.9 cm, and a conventional 3:1 compression screw). A lament head attachment with a 2 mm nozzle was used. Extrusion was carried out at 25°C and at a screw speed of 10 rpm. The lament/cord was collected using a cardboard roll.

Chemical Treatment
The extruded wet CNF + CMC cords were cut into ~ 100 mm long segments with an approximate wet diameter of 2 mm (see Figure S2a). Treatment solutions of 0.1 M HCl, 0.1 M CaCl 2 , 10 % PEI, 10 % PAE, and 10 % PAmE were freshly prepared. The still wet cord segments were carefully submerged into the different treatment solutions, respectively. Treatment time, solution concentration, and drying/curing temperature were optimized through preliminary experimental trials with HCl and assumed to also apply for CaCl 2 (see Figure S3, Figure S4, and Figure S5). It was determined that a concentration of 0.1 M was ideal to retain the cord's dimensions while a soak time of 20 min allowed for the treatment to penetrate a distance of 2 mm (i.e., the cord's diameter). After the respective treatment, excess solution was drained off and the cord segments were oven dried at 70 ºC for 1 h, to avoid any oxidation or browning. For treatment with PEI, PAE, and PAmE, the cord segments were cured in the oven at 70°C for PEI, and 85°C or 25°C for PAE and PAmE (per supplier recommendations). Due to CDI's reactivity at room temperature conditions and the required acidic conditions for correct activation, the treatment process was slightly adjusted. CDI was mixed directly into the CNF + CMC paste at a dry weight ratio of CMC:CDI of 1:1. Subsequently, the CNF + CMC + CDI cord segments were submerged for 20 minutes in 0.1 M HCl and cured at 70°C or 25°C. Lastly, oven dried CNF + CMC cord segment without any treatment, termed "untreated" were also prepared and dried at 70°C for 1 h.

Pelletization
For all treatments and degrees of substitution, the dried and cured CNF + CMC cords were cut by hand using scissors into pellets with an approximate length of ~ 1 mm. Longer untreated CNF + CMC cords were pelletized using a commercial pelletizer (Davis Standard, Pawcatuck, CT, Model PK-102) running at 30 rpm for the rehydration and remixing analysis.

Zeta Potential
The prepared CNF + CMC pastes (D.S. = 0.7, 0.9, and 1.2 all at ~ 18 wt.%) were diluted to 1 wt.%. CNFs adsorbed with CMC were precipitated by centrifugation (10,000 RFC) with any unbound CMC remaining in the supernatant. The supernatant was discarded and the CNF brils were resuspended by shear mixing in puri ed water. This process was repeated 3 times. Subsequently, the washed suspensions were further diluted to a nal concentration of 0.025 wt.%. Zeta potential measurements were carried out using a Zetasizer Nano ZS (Malvern Panalytical) and disposable folded capillary cells (Model: DTS1070). The viscosity of the suspensions was not measured but was assumed to be that of water due to the low solids concentration. The pH of all the suspensions was measured to be 6.6 (HANNA® Instruments). Eight measurements were recorded, and the average was reported along with ± STD error bars. Quality criteria were met for all suspensions analyzed, yet the count rate varied for pure CNF suspensions.

Titration
From the ~ 1 wt.% CNF + CMC suspensions prepared for zeta potential analysis, 5 mL were added to 250 mL beakers and diluted with 50 mL of deionized water. 500 mL of 0.1 M HCl was added to fully protonate the carboxylic acid groups on the CNFs. The resulting mixture was titrated with 5 mM NaOH solution. The pH was monitored using an Oakton Acorn pH meter. Representative titration curves (pH verses titrant volume) can be seen in Figure S6. The volume between the two in ection points on the curve was used to calculate the amount of COOH present on the nano bers.
Turbidity 0.17 g of the treated or untreated CNF + CMC pellets were added into a scintillation vial followed by 17 mL of puri ed water or 17 mL of a 0.1 M NaOH solution (i.e., a target concentration of 1 wt.% solids). The pellets were stirrer in solution for 24 h at 1000 rpm and 25 ºC. This was repeated for each different degree of substitution and treatment type. Additionally, 1 wt.% never-dried CNF + CMC and pure CNF suspensions were prepared as control groups. Following stirring, 100 mL of the supernatant, with an unknown solids concentration of redispersed CNFs, was diluted with an additional 17 mL of puri ed water. Dilution was performed to be in the most accurate reading range of the turbidity meter. The dilute suspensions were analyzed using a nephelometric turbidimeter (Vernier®, Beaverton, OR) consisting of a 90º photodiode detector and an 890 nm infrared LED light source. The turbidity meter was calibrated using a 100 NTU Formazin standard and DI water (i.e., 0 NTU) in a glass cuvette. Before collection, a 10 mm stir bar was added to the sample glass cuvette, and the turbidity meter was placed on top of a stir plate. Turbidity was continuously collected every 0.5 s for 650 s while the suspension was being stirred at 400 rpm. Six measurements were carried out per sample group with a fresh diluted sample loaded for every measurement. The average between all 6 measurements is reported along with ± STD error bars as stated under the gures, respectively.

FTIR-ATR
Treated and untreated CNF + CMC pellets, CMC powder (D.S. = 0.7), and a pure cast CNF lm were conditioned in a vacuum oven for 24 h at room temperature. The samples were placed in direct contact with the ATR crystal and a 550 cm − 1 to 4000 cm − 1 spectra was captured using an FTIR-ATR system (PerkinElmer) with a resolution of 4 cm − 1 . 20 sample scans were collected, and background subtraction was performed, respectively. Baseline correction was applied to all the spectra for fair comparison.
Scanning Electron Microscopy (SEM) with carbon tape and then sputter coated (SPI sputter coater) with a platinum-gold target for 60 s. No polishing or sanding was used. The samples were imaged using a Quanta 650 FEG eld emission electron microscope at 3 KeV and a spot size of 5. The working distance varied from 39-41 mm to achieve the highest resolution.

CNF + CMC Bulk Processing and Pelletizing
Extrusion-based processes are one of the most versatile, commercially available, and easily scalable processing methods used for current commodity polymers. For these reasons, a single-screw extruder was employed to continuously extrude the highly loaded (~ 18 wt.%) CNF + CMC pastes. Results show that wet lament with a diameter of ~ 2 mm (~ 1 mm dry) can be successfully processed without any restrictions on its nal length. The extrusion and collection process are shown in Video S1. The extruded cord was collected and oven dried on a carboard roll as shown in Fig. 1a and Fig. 1b. Constrained drying was necessary for the CNF + CMC cord to retain its shape and avoid shrinking/warping (see Fig. 1c verses Figure S7). Following drying, the cord was easily pelletized by using a commercial pelletizer running at 30 rpm as shown in Fig. 1d and Video S2. The processed CNF + CMC pellets had an average length of 6.73 ± 2.5 mm. Roughly 10 g of dry cord were processed and presented in Fig. 1d.
With the current set up used, dry CNF + CMC extrusion output rates of up to 0.25 ± 0.072 kg/h and nozzle speeds of up to 7.68 ± 2.22 m/min were possible at a screw speed of 110 rpm. Although the nozzle speeds are lower than those typically reported in conventional spinning processes (20-150 m/min) (Lundahl et al. 2017), lament die extrusion can process larger diameter cord (> 1 mm) versus those typically seen for spinnerets (10-500 mm) (Mather and Wardman 2011), making it a more effective bulk processing methodology. Additionally, larger diameter cord allowed for easier handling and collection as the wet cord extrudate was relatively fragile and does not strain harden as conventional synthetic polymeric lament.
The mechanical properties of the processed CNF + CMC cords are not expected to be higher than those typically observed for CNF, (2,2,6,6-tetramethylpiperidin-1-yl)oxidanyl (TEMPO) oxidized CNF or composite ber laments prepared using wet-or dry-spinning methodologies (Håkansson et al. 2014;Kafy et al. 2017;Clarkson et al. 2019). This is due to the lower shear rates typically observed in larger nozzles and the lack of lament drawing, both of which facilitate CNF bril alignment in the axial direction and improve the mechanical properties. However, the intent of this report is to understand the redispersion and setting behavior of highly loaded CNF + CMC systems rather than improve the mechanical performance of the extruded lament/cord.
The surface morphology of the oven dried and untreated CNF + CMC cords with a varying CMC degree of substitution (D.S. = 0.7, 0.9, and 1.2) is shown in Fig. 2 below. It is evident that there is a lack of bril alignment when Fig. 2 is compared to other reported drawn CNF lament micrographs (Kim et al. 2019;Cai et al. 2020). Furthermore, there appears to be no difference between cords prepared with CMC with a D.S. of 0.7 and 0.9 ( Fig. 2a and Fig. 2b, respectively), while the cord prepared with a D.S. of 1.2 appeared to have a slightly rougher surface (Fig. 2c).

CMC Adsorption onto CNF brils
As mentioned in the introduction, CMC has been reported to adsorb irreversibly to the surface of CNF even under mild heating conditions (22°C) in low solids (~ 1 wt.%) suspensions in water (Butchosa and Zhou 2014). This adsorption occurs due to the attachment of unsubstituted CMC cellulose units onto the surface of exposed cellulose chains on CNF brils through hydrogen bonding and can be improved if cations (e.g., Ca 2+ ) are added to the CNF suspension (Duker et al. 2007;Liu et al. 2011). In this work, both the Banbury high-shear mixing process and the relatively high mixing temperatures (55°C) used to prepare CNF + CMC pastes also caused CMC to adsorb irreversibly to the surface of CNF even at high solids loadings (~ 18 wt.%). This was con rmed through zeta potential and titration measurements on dilute CNF + CMC suspensions prepared from the different pastes. As shown in Table 1, CMC adsorption caused an increase in surface charge from − 36.8 ± 2.04 mV for pure CNF to -59.3 ± 1.31 mV for CNF + CMC (D.S. = 0.7). Additionally, there was no statistical difference between the three CMC degrees of substitution with respect to zeta potential (one-way ANOVA: = 2.22, = 0.133, R 2 = 0.175). This could mean that the surfaces were completely saturated even with a more negative CMC with a D.S. of 1.2.
As shown in Table 1, titration of the different CNF + CMC suspensions with NaOH revealed the presence of weak acid functionalities from the adsorbed CMC's carboxylic acids. More speci cally, surface charge increased from 0.094 ± 0.005 mmol/g COOH for pure CNF to 0.166 ± 0.035 mmol/g COOH for CNF + CMC with a degree of substitution of 0.9. However, while zeta potential revealed a change in magnitude of surface charge (mV) for all D.S. in CNF + CMC when compared to pure CNF, titrated surface charges for CNF + CMC with a D.S. of 0.7 and 1.2 were not signi cantly different to pure CNF (0.104 ± 0.024 and 0.082 ± 0.012 mmol/g COOH, respectively). As expected, the titrated charges for pure CNF and CNF + CMC are lower than those typically observed for TEMPO oxidized CNF (up to 1.5 mmol/g) (Lasseuguette 2008;Jiang and Hsieh 2016) due to the lower level of carboxylic functionalities.  Table 1 Zeta potential and titration of dilute pure CNF and CNF + CMC suspensions. Each suspension was mixed, centrifuged, and resuspended in water 3 times to remove non-adsorbed CMC or other loose species. Eight measurements per sample were carried out for zeta potential analysis while 3 measurements per sample were performed for titration. The average and standard deviation of the samples are displayed. 5 mM NaOH was used for titration. Given that CMC adsorption was present, its plausible that the CNF bril's ability to form irreversible hydrogen bonds between themselves was suppressed. Hence, the pelletized and untreated cord was able to rehydrate itself rapidly (under 20 min) without any stirring/mixing as shown in Fig. 3(a-c). These rehydrated pellets were re-mixed using the same Banbury mixer and reformed into the original 18 wt.% paste as shown in Fig. 3d. Complete re-mixing was con rmed by an equivalent torque reading on the mixer when compared to a freshly prepared CNF + CMC paste (see Figure S8). It is important to note that the torque plateau for the rehydrated CNF + CMC pellets was reached in less than 2 min verses the original ~ 19 min (see Figure S8), hence complete re-mixing was easily and quickly achieved. Furthermore, this paste was re-extruded into a cord without any di culty and dried as shown in Fig. 3e and Fig. 3f and appeared identical to the original cord before pelletizing (Fig. 1c). Importantly, this shows that by adding CMC, the material can be extruded, dried, pelletized, rehydrated, and extruded to the same consistency to "close the loop", which conceptually allows dry shipment of material.

Dispersibility and Setting Behavior of CNF + CMC
To evaluate the dispersibility and setting behavior of highly loaded CNF + CMC pastes, different chemical treatments were selected and applied to the extruded wet cords. The treatments assessed were HCl, CaCl , PEI, CDI, PAE, and PAmE. Light scattering techniques like turbidity are often used to monitor the degree of brillation/quality during CNF manufacturing processes (Moser et al. 2015;Desmaisons et al. 2017) and, in this work, served to gauge the level of redispersion achieved by the treated and untreated dry CNF + CMC materials in water. Turbidity results in Fig. 4e show that complete redispersion is observed when the dried-untreated CNF + CMC samples (~ 29 NTU) are compared to the never-dried CNF + CMC dilute suspensions (~ 30 NTU), regardless of CMC's degree of substitution. On the other hand, chemical treatment with HCl, CDI, PAE and PAmE all signi cantly suppressed the level of redispersion achieving a value of less than 1 NTU when compared to the control groups of never-dried pure CNF (~ 29 NTU), neverdried CNF + CMC (~ 30 NTU), and dried-untreated CNF + CMC (~ 29 NTU). As shown in Fig. 4(a-d) and Figure S9, complete redispersion was assessed visually where no observable macroscopic pellets were left in the suspension after mixing versus clearly visible pellets that did not redisperse (i.e., semipermanent setting). Furthermore, pellets treated with HCl, CDI, PAE, or PAmE did not show any sign of redispersion even after 4 weeks of being immersed in water.
Unlike PAmE, PAE, CDI, and HCl, treatment with PEI and CaCl 2 did not seem to suppress the redispersion behavior of CNF + CMC possibly due to the inability of these treatments to form crosslinks between CNF brils which then permitted the adsorbed CMC to redisperse them as shown in Figure S9. It is important to note that the lower turbidity values observed for the PEI treatment (~ 19 NTU) are due to the strong cationic nature of the ammonium groups which caused the negatively charged brils to occulate while in suspension rather than the lack of pellet redispersion. This was con rmed through a solids contents measurement of the supernatant (~ 1 wt.%) after redispersion which matched with the original target concentration (1 wt.%) when water was added to the dry pellets. Additionally, for all treatments studied, the degree of carboxymethylation or substitution (D.S. = 0.7, 0.9, and 1.2) of CMC appears to have no signi cant effect on the level of redispersion.
For the best performing setting treatments (HCl, CDI, PAE, and PAmE) dispersion in a strong alkaline solution (0.1 M NaOH, pH > 10) was also performed. The different treatments were assessed at two curing temperatures. Dispersion in NaOH helped isolate the effect of the different treatments on crosslinking by promoting negative surface charges on CMC by neutralizing any acid present, thus improving the swelling and dispersion behavior of CNF + CMC brils (Uetani and Yano 2012). This then revealed differences between treatments which were not evident in water (pH ~ 7). The results in Fig. 5b show that HCl was effective regardless of the curing temperature achieving low turbidity values (6.7 NTU for 25°C and 6.0 NTU for 70°C, respectively) while PAmE and CDI were only effective at their higher curing temperatures of 85°C and 70°C achieving turbidity values of ~ 4 NTU and ~ 3 NTU, respectively. PAE was not effective regardless of the curing temperature hence high turbidity values were observed (23.9 NTU for 25°C and 17.7 NTU for 85°C), however there was a difference between the dispersion behavior observed for a cure temperature of 85°C and 25°C, (one-way ANOVA: = 74.3, = , R 2 = 0.881). The difference in dispersion behavior observed between the two different curing temperatures for CDI, PAE and PAmE suggests the formation of physical crosslinks rather than purely structural irreversibly changes that might have occurred during drying (i.e., horni cation), while at higher temperatures chemical crosslinking occurs (Fernandes Diniz et al. 2004). As shown in Figure S10 and Fig. 5a small pellet fragments were visible for all treatments hence some level of dispersion was still possible in the studied time period (24 h).
Optical micrographs, shown in Fig. 6, of the redispersed CNF + CMC brils in water and 0.1 M NaOH revealed that the bril's morphology remained intact before and after redispersion when compared to never-dried pure CNF (Fig. 6a) and dried-untreated CNF + CMC (Fig. 6b). It is important to note that the reduced presence of CNF brils displayed in the micrographs for CDI (Fig. 6c), HCl (Fig. 6d), PAE (Fig. 6e), and PAmE (Fig. 6f) match with the results obtained for turbidity. Perhaps only non-crosslinked and or smaller brils were able to be redispersed back into the solution. Furthermore, expanded optical micrographs show the same behavior (see Figure S11 and Figure S12). To understand the possible setting mechanism behind the chemical treatments (HCl, CDI, PAE, and PAmE) which signi cantly suppressed the dispersibility of CNF + CMC, FTIR was used. As shown in Figure S13 through Figure S16, complete spectra (550 cm − 1 to 4000 cm − 1 ) were captured for each treatment at two curing temperatures. As expected for all samples, common cellulose IR peaks at 3335 cm − 1 , 2899 cm − 1 , 1429 cm − 1 , 1368 cm − 1 , 1335 cm − 1 , 1024 cm − 1 , and 896 cm − 1 were observed. These peaks were assigned to the O-H hydroxyl stretching, C-H stretching, and bending and stretching of -CH 2 , -CH, -OH, and C-O bonds in cellulose, respectively (Aguayo et al. 2018;Hospodarova et al. 2018).

FTIR Analysis of Treated CNF + CMC
Further analysis of the 1000 cm − 1 to 1800 cm − 1 spectrum region, shown in Fig. 7, revealed key peak differences between the treated CNF + CMC samples and the control groups of pure CMC powder, pure CNF, untreated CNF + CMC, and cured pure compounds of CDI, PAE, and PAmE. As shown in Fig. 7a treatment with HCl caused the formation of a strong peak at 1731 cm − 1 which was assigned to the carbonyl stretching of newly formed ester linkages (i.e., COOC) or, equally likely, the carbonyl stretching of protonated carboxylates (i.e., COOH) (Müller et al. 2010 Protonation and esteri cation could not be separated due to the similarity of the peak locations for the C = O stretching vibration which typically lies between 1760 − 1700 cm − 1 . However, esteri cation crosslinking between CNF brils with adsorbed CMC could possibly explain the suppressed dispersibility behavior observed for both cure temperatures during turbidity measurements (Fig. 5b). A schematic of the possible ester linkages between CNF brils with adsorbed CMC is shown in Fig. 8a ). However, because these peaks were also present in the spectra for pure CDI cured at 70°C (shown by the dotted vertical lines in Fig. 7b), they could simply originate from the added presence of the resin in the CNF + CMC system. The second CDI-carboxylic acid reaction route leads to the formation of anhydride linkages and urea as byproduct from the reaction between O-acyl urea and a additionally available carboxylic acid (Posthumus et al. 2007). The formation of anhydride linkages can then possibly explain the observed strong 1033 cm − 1 peak (CO-O-CO stretching) and the 1754 cm − 1 peak (C = O stretching), which were not present in the pure CDI spectra (shown by the dashed lines in Fig. 7b), while the 1448 cm − 1 (N-C-N asymmetric stretching) and the 1247 cm − 1 peak (C-N stretching, amine) can be linked to the urea byproduct (Piasek Z. and Urbanski T. 1962;Lu et al. 2016). A schematic of the possible anhydride linkages formed between CNF brils with adsorbed CMC is shown in Fig. 8b (Posthumus et al. 2007).
There is also the possibility that ester linkages were formed between CMC units through a subsequent reaction between anhydride linkages and a nucleophile like that of a hydroxyl from either CMC or CNF leading to intra-or inter-molecular linkages which could also explain the appearance of the 1754 cm − 1 peak/shoulder (C = O stretching, COOC) which was not present for the pure CDI compound (see dashed line in Fig. 7b) (Posthumus et al. 2007;Mojarradi 2011). The peak observed at 1716 cm − 1 (C = O stretching) could belong to carboxylic acids which were not converted during the CDI reaction. It is important to note that both reaction routes can occur simultaneously and the ratio between N-acyl urea and anhydride/urea has been reported to be dependent on the mobility of the reactive groups and ratio of concentration between carbodiimide and carboxyl groups (Posthumus et al. 2007). Additionally, similar to treatment with HCl, the lack of the 1592 cm − 1 peak for either CDI 25°C and CDI 70°C points to a possible consumption of the (COO − ) groups for the formation of N-acyl urea, anhydride, and ester linkages as well as the possible protonation of the carboxylates due to the low pH at treatment (0.1 M HCl). Although both CDI 25°C and CDI 70°C appeared to contain the same major peaks in the 1000 cm − 1 to 1800 cm − 1 region, the disappearance of the 2117 cm − 1 peak (assigned to the -N = C = N-stretching vibration of carbodiimides (Pham and Winnik 2006)) shown in Figure S17, con rmed that a higher curing temperature (70°C) was necessary for complete CDI resin reaction/activation. Hence, the lack of complete CDI activation could possibly explain the much higher turbidity values observed for the CDI treatment cured at 25°C (27.6 NTU) when compared to the CDI treatment cured at 70°C (3.3 NTU) as shown in Fig. 5b.
Treatment with PAE, shown in Fig. 7c, revealed amide I and amide II peaks at 1641 cm − 1 (C = O stretching, CONH) and 1546 cm − 1 (N-H bending), respectively. These were also present in the spectra for pure PAE cured at 85°C (see vertical dotted lines in Fig. 7c and Fig. 8c Yang et al. 2017). The ester bonds are formed from the thermally induced reaction between (COO − ) and the azetidinium groups in the PAE resin (Siqueira et al. 2015). A schematic of the possible ester linkages between CNF brils with adsorbed CMC is shown in Fig. 8c (Yang et al. 2018). On the other hand, pure PAE has been reported to crosslink with itself forming ester linkages though its tail end carboxylates or, as a secondary crosslinking reaction, form 2-propanol linkages between the azetidinium groups on its backbone (Obokata and Isogai 2007;Chattopadhyay et al. 2013;Siqueira 2014). Hence, PAE esteri cation with CMC and self-crosslinking could possibly explain the shift or appearance of the two ester absorption bands as shown in Fig. 7c. It is important to note that while HCl and CDI treatment may have led to protonation of CMC's carboxylate groups, treatment with PAE or PAmE did not require a low pH and hence the peaks observed are likely ester carbonyl stretching vibrations. When the two cure temperatures (25°C and 85°C) are compared, the ester band (1742 cm − 1 ) is almost not present for the 25°C cured sample possibly due to a reduced number of linkages formed, hence it was much easier for CNF + CMC to redisperse as seen in the higher turbidity values (see Fig. 5b, 23.9 NTU for 25°C verses 17.7 NTU for 85°C).
As shown in Fig. 7d, a similar spectrum to the PAE treatment was observed for treatment with PAmE. Again, the amide peaks simply con rmed the presence of the PAmE resin in the CNF + CMC system as these were also present in the pure PAmE cured at 85°C spectra (dotted lines). Similar to PAE, two different absorptions bands for carbonyl stretching of esters were observed, one at 1742 cm − 1 (dashed line) for PAmE treated CNF + CMC and one at 1732 cm − 1 (dotted line) for pure PAmE cured at 85°C. This is possibly due to the fact that both resin systems (PAE and PAmE) can react with the same functional groups (carboxylates, hydroxyls, thiols, and amines) which led to identical crosslinking linkages. Perhaps the only difference between the two resin spectra (PAE verses PAmE) is the lower peak intensities observed for PAmE. However, the performance of PAmE was superior in terms of redispersion suppression to that of PAE, as observed in Fig. 5b. The reason for this is unknown but could be linked to the high cationic charge of PAmE which might have allowed for the resin to interact more strongly with the exposed (COO − ) groups in CMC or other slightly negative charges on the surface of CNF brils as well as go into the nanopores between brils (Yang et al. 2017). Furthermore, it is important to note that unlike all other treatments, the carboxylate peak (1592 cm − 1 ) was not consumed signi cantly for either cure temperature (25°C and 85°C). When the two cure temperatures are compared for treatment with PAmE, the intensity for the 1732 cm − 1 peak for a cure temperature of 25°C were of weaker intensity when compared to the spectra for 85°C, which might explain the much higher turbidity value of 21.7 NTU when compared to 3.9 NTU, respectively (see Fig. 5b).  Table 2 summarizes the FTIR results shown in Fig. 7 where the intensity of the main peaks of interest (1033 cm − 1 , 1549 cm − 1 , 1591 cm − 1 , 1641 cm − 1 , 1731 cm − 1 , and 1754 cm − 1 ) was normalized by that of the intensity of the invariant 1024 cm − 1 peak assigned to the C-O stretching in cellulose (Colom et al. 2003;Lionetto et al. 2012). Overall, due to the numerous different reaction routes, treatment with CDI could have led to the formation of ester, N-acyl urea, anhydride, or all of the above linkages which made it a very effective treatment for the setting of CNF + CMC. PAmE was an equally effective setting treatment possibly due to the formation of ester linkages and, unlike PAE, had a high cationic charge which might have improved its interaction with CMC's carboxylate groups. However, both of these treatments were only effective when cured at higher temperatures (70°C and 85°C, respectively). On the other hand, HCl was effective even when cured at room temperature (25°C). Lastly, PAE was the least effective setting treatment for CNF + CMC regardless of the curing temperature.
Water Uptake of Treated CNF + CMC Water uptake of the differently treated CNF + CMC with a D.S. of 0.7 agreed with both the observed turbidity behavior and the possibly newly formed linkages detected through FTIR. For example, the best performing setting treatments of HCl and CDI (both cured at 70°C) absorbed on average only 20.1 ± 0.19 %, and 17.5 ± 0.58 % of their weight in water, respectively. PAE and PAmE treated CNF + CMC (both cured at 85°C) absorbed 41.6 ± 0.32 % and 59.7 ± 1.1 % of their weight in water, respectively. The higher water uptake for the PAE and PAmE treated CNF + CMC led to a much more exible structure which could, in turn, be more susceptible to redispersion by mechanical stirring (i.e., softer structure). On the other hand, HCl and CDI treated materials were much more rigid and possibly less susceptible to redispersion by stirring. The exibility of the treated CNF + CMC materials is qualitatively shown in Figure S18. Unlike HCl and CDI, hydrated PAE and PAmE treated CNF + CMC cord segments could be folded onto itself without fracturing into two pieces.

Rami cations of efforts
More broadly, the results obtained show that the redispersion and setting behavior of CNF with adsorbed CMC can be controlled through different chemical treatments. Industrially, redispersion grants the ability to transport CNF in a dry state, thus providing a signi cant reduction in weight and transportation cost.
On the other hand, semi-permanent setting of CNF + CMC systems allows for CNF products (e.g., sheets, cups, lids, clamshells, etc.) to retain their shape after being processed while at the same time reducing their susceptibility to high humidity conditions, and hence extending the life of use and expanding their possible applications. Furthermore, bulk processing of CNF + CMC through conventional polymer processing equipment like a Banbury high-shear mixer, a single-screw extruder, and a commercial pelletizer takes advantage of the already developed polymer melt-processing industry and reduces the initial capital investment destined for new processing equipment.

Conclusion
Highly loaded CNF+CMC pastes (~18 wt.%) were prepared and continuously processed into cord using a single-screw extruder. Availability of data and material: Additional information regarding this manuscript can be found free of charge in the supporting information.
Code availability: Not applicable.
Authors' contributions: All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Sami M. El Awad Azrak and Jared A. Gohl. The rst draft of the manuscript was written by Sami M. El Awad Azrak and all authors commented on previous versions of the manuscript. All authors read and approved the nal manuscript.
Ethics approval: All authors have read and understood the "Ethical Responsibilities of Authors" in the journal's "Submission Guidelines", including the passage on screening for plagiarism with computer software.
Consent to participate: Not applicable.
Consent for publication: All authors have fully read, agreed, and approved the nal version of the manuscript which has been submitted for possible publication.

Supporting Information
The following information is provided free of charge: TEM images of CNF brils at two different magni cations (PDF) Video recording of the extrusion and collection process for CNF+CMC pastes (mp4) Video recording of the pelletization of dry CNF+CMC cord (mp4) Photograph of wet and dry CNF+CMC cord segments (PDF) Representative titration curves (PDF) Photograph of CNF+CMC cord treated with HCl and cured at different temperatures (PDF) Analysis of the penetration rate of HCl into CNF+CMC (PDF) Plastogram/torque curves for the mixing of rehydrated CNF+CMC pellets and a freshly prepared CNF+CMC paste (PDF) Photograph of redispersed CNF+CMC pellets after stirring in DI water and NaOH (PDF) Optical micrographs of redispersed CNF brils (PDF) FTIR spectra of HCl, CDI, PAE, and PAmE treated CNF+CMC (PDF) Figure 1 Single-screw lament/cord extrusion of a CNF+CMC paste (at ~18 wt.%) through a 2 mm nozzle (a), collected CNF+CMC wet cord extrudate on a cardboard roll (b), oven dried CNF+CMC cord (c), and pelletized cord (d).
Page 24/30 Figure 2 Surface morphology of the extruded and dried untreated CNF+CMC cords with a varying degree of CMC carboxymethylation or substitution of 0.7 (a), 0.9 (b), and 1.2 (c).

Figure 3
Pelletized untreated CNF+CMC pellets at different stages of the rehydration process starting completely dry (a), ~30 s after adding water (b), 20 min after adding water (c), after being re-mixed in the Banbury shear mixer (d), and re-extruded into a lament/cord (e) and oven dried at 35 °C on a cardboard roll (f). CaCl2, HCl, and CDI treated samples were dried at 70 °C while PAE, PAmE treated samples were dried at 85 °C. The CNF+CMC pellets were stirrer for 24 h in water at 25 °C before turbidity was collected and the pictures were taken. For each treatment type and D.S., a freshly prepared suspension was analyzed. Six measurements were carried out per sample. The mean and standard deviation error bars are displayed, respectively.

Figure 5
Redispersion behavior at different points in time for CNF+CMC treated with PAE cured at 85 °C in 0.1 M NaOH (a) and turbidity response of different treated and untreated CNF+CMC materials in 0.1 M NaOH at two curing temperatures(b). The suspensions were stirrer for 24 h before turbidity was collected. For each treatment type, two curing temperatures were evaluated while the CMC's D.S. remained constant at 0.7. A freshly prepared suspension was analyzed for each measurement. Six measurements were carried out per sample. The mean and standard deviation error bars are displayed, respectively. The black scale bars in the top right have a length of 400 µm. Expanded (i.e., lower magni cation) optical micrographs of the same dispersions are shown in Figure S11 and Figure S12. The control groups of pure CNF, pure CMC powder, and untreated CNF+CMC are also shown on the respective plots. The samples were conditioned under vacuum for 24 h at room temperature before testing. All the spectra shown were baseline corrected and 20 scans were collected. All intensities were all normalized by the invariant cellulose peak intensity at 1024 cm-1. Dashed vertical lines mark the peak locations found in the treated CNF+CMC spectra while the dotted vertical lines mark peaks found in the pure compound spectra, respectively.

Figure 8
Schematic of CNF with adsorbed CMC through hydrogen bonding with possible ester linkages (circled in green) formed due to treatment with HCl (a), anhydride linkages (circled in red) due to treatment with CDI (b), and ester linkages (circled in green) formed due to treatment with PAE and PAmE (c). Note that amide groups (circled in blue) are present in the PAE and PAmE backbone. All the schematics shows CMC with a D.S. of 1.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.