Absorbent Filaments from Cellulose Nanofibril Hydrogels through Continuous Coaxial Wet Spinning

A continuous and scalable method for the wet spinning of cellulose nanofibrils (CNFs) is introduced in a core/shell configuration. Control on the interfacial interactions was possible by the choice of the shell material and coagulant, as demonstrated here with guar gum (GG) and cellulose acetate (CA). Upon coagulation in acetone, ethanol, or water, GG and CA formed supporting polymer shells that interacted to different degrees with the CNF core. Coagulation rate was shown to markedly influence the CNF orientation in the filament and, as a result, its mechanical strength. The fastest coagulation noted for the CNF/GG core/shell system in acetone led to an orientation index of ∼0.55 (Herman’s orientation parameter of 0.40), Young’s modulus of ∼2.1 GPa, a tensile strength of ∼70 MPa, and a tenacity of ∼8 cN/tex. The system that underwent the slowest coagulation rate (CNF/GG in ethanol) displayed a limited CNF orientation but achieved an intermediate level of mechanical resistance, owing to the strong core/shell interfacial affinity. By using CA as the supporting shell, it was possible to spin CNF into filaments with high water absorption capacity (43 g water/g dry filament). This was explained by the fact that water (used as the coagulant for CA) limited the densification of the CNF core structure, yielding filaments with high accessible area and pore density.


EXPERIMENTS
The spinning conditions varied between different sets of experiments for several reasons. For example, as the rheological nature of GG and CA differ, 1 different solids contents and extrusion speeds were required to obtain suitable flow behavior for core/shell wet-spinning (Table S1).
Moreover, when optimizing the production rate (Table S2) and water absorption (Table S3), achieving the respective properties were prioritized instead of comparability with the standard samples. Tables S1 and S2 identify the spinning conditions used for the sample preparation and the determination of the maximum production rate, respectively. Table S3 specifies the conditions employed to spin absorbent filaments of CNF and CA, which were also used to study the effect of drawing, which will be discussed in the following section. These samples were prepared with an earlier iteration of the core/shell wet-spinning line (smaller winder and different needle size). S-4

EFFECT OF DRAWING ON ABSORBENT FILAMENTS WITH CNF AND CA
The extension rate applied on the filament in the bath was controlled by varying the winding speed according to Table S4. The extension rate !""#$%& was quantified by Equation S1 : where v needle is the extrusion speed of the dope, v winder is the winding speed and L bath is the coagulation distance (i.e., the distance a filament element travels in the coagulation bath). Thus, v winder /v needle corresponds to the drawing ratio (DR) and (v winder +v needle )/2 to the average speed of a filament element during the coagulation. The core/shell spinning system enabled clearly higher DR and extension rates than applied in CNF wet-spinning previously. Previous efforts to stretch CNF are compiled in Table S5 along with the present study. Interestingly, in combination with the CNF core, CA could be drawn to an even larger extent than CA alone (Table S4). This observation suggests synergies between both components. The physical and mechanical properties of the CNF/CA filaments with varying DR are presented in Table S6.

S-8
For water-coagulated CNF/CA, cross-sectional areas of the filament components were estimated assuming filament cross-section comprising two concentric rectangles: outer one with the dimensions measured for CNF/CA core/shell filament and inner one the dimensions of the CNF filament after CA shell removal. The inner rectangle was expected to be filled with CNF and the space between inner and outer rectangles with CA. Weight fractions of CNF and CA in a bicomponent filament were determined by weighing a batch of 0.1-0.3 g of filaments before and after removal of the shell. This was not possible for CNF/GG filaments given that shell removal would not be selective in this case, as both components disperse in the same solvents.
Total filament thickness, width and coarseness for systems with or without CA shell generally decrease with increasing DR. As the only exception, the thickness of the neat CNF filaments appeared to plateau at approximately 8 µm at DR of 5 and 9. This is probably related to the micrometer gauge approaching the limits of its resolution at such small thicknesses. This could also cause the unexpected development of the mass and volumetric ratios of CNF and CA in their drawn bicomponent filaments. At a DR of 9, the CNF/CA mass ratio (40:60, Figure S1a) becomes lower than their volumetric ratio (60:40, Figure S1b). Also, at a DR of 5, the CNF/CA mass ratio becomes exceptionally high (50:50, Figure S1a) and volumetric ratio exceptionally low (30:70, Figure S1b Figure S1. Effect of drawing on the linear density (a, corresponding to weight fractions) and cross-sectional area (b, corresponding to volume fractions) of the water-coagulated CNF/CA filament and its components. Error bars are defined as the standard deviation divided by the square root of the sample size (10 specimens). Spinning conditions: Tables S3, S4.
The increased DR also seems to produce denser filaments, at least in the case of those with CA shell retained. However, this densification with drawing is unlikely as pronounced as suggested by Table S6. For example, the 9-fold drawn filament has similar apparent coarseness to the 5fold drawn (~4 tex), despite the clearly smaller dimensions. This leads to an artificially high apparent density of 1.67±0.26 g/cm 3 , even though both the components alone have lower densities: CA 1.58±0.25 g/cm 3 (Table S6) and cellulose 1.5-1.6g/cm 3 . 13 Moreover, manual measurement and weighing of filament pieces to obtain the apparent coarseness is of limited S-10 resolution for light filaments. In fact, based on the extrusion and winding speeds as well as the solids content of the CNF and CA, a DR of 9 is expected to yield a core/shell filament with a calculated coarseness of only 1.7, which would correspond to a density of 0.73 g/cm 3 . For the other filaments, the calculated coarseness matched closely the measured values (Table S6).
Stress-strain curves of the filaments with and without CA and drawing are presented in Figure   S2. By increasing the DR, the strength and stiffness of the CNF/CA filaments augmented ( Figure   S2, purple). However, no clear effect of drawing was evident for neat CNF filaments ( Figure S2 (Table S6, Figure   S2). This implies that the structure adopted by CNF enclosed by CA in a water bath is suboptimal for load-bearing but more useful for other purposes, such as water absorption, as discussed in the main article.
As the mechanical properties of the filament were calculated based on the physical properties, they are subject to error. For example, since the thinnest filament (DR 9) had a surprisingly high thickness of 7.7±0.3 µm (possibly due to limited resolution of the micrometer gauge), its real tensile strength may be higher than the reported 14.2±6.0 MPa. Considering this, the Young's modulus and tensile strength of CNF/CA filaments definitely seem to augment with drawing ( Figure S2, Table S6). However, a different trend is observed for the values of specific modulus and strength as well as tenacity (Table S6). In fact, when normalizing the filament stiffness and strength against its coarseness instead of cross-sectional area, the strongest core/shell filament (DR 9) seems the weakest. This deviation arises from the questionably high apparent coarseness measured for this sample (Table S6). The trend observed for the Young's modulus and tensile strength normalized against the filament cross-sectional area can, therefore, be considered more reliable.  Tables S3, S4.
Nevertheless, the value of tenacity can be compared rather reliably, since it is independent of the filament cross-section and its fluctuations. The tenacity at varying DR (Table S6) indicate that single component CNF filaments produced at a DR of 5 presented the best performance.
Also, the Young's modulus, specific modulus and specific strength peak at a DR of 5. This implies that, even if the drawing induced a slightly more load-bearing structure even in the CNF component, excess drawing may damage it.
Previously, CNF filaments have been subjected to drawing after drying while immersed in water 3 or acetone-water mixture, 6 capitalizing on the plasticizing effect of water. 15 By this "post- Even though the core/shell spinning approach increased the maximum DR possible for wetspun CNF, from the previously reported 1.3 3 to 9, no improvement in strength was measured upon drawing ( Figure S2, inset). This might imply that an optimum DR value may exist for optimal filament performance, as indicated in the case of the post-drawn CNF filaments: increasing the DR from 1.2 to 1.3 failed to improve the tensile strength further above 289 MPa. 3 Such possible performance plateau with increasing DR has also been studied through simulations and flow focusing experiments at different extension rates. 17 When the extension rate was elevated, fibril orientation in a flow-focusing channel improved, though not as much as predicted through simulation. This led to the conclusion that increasing the extension on a CNF suspension also increases the resistance of the fibrils to orient. 17 A similar effect might cause the plateau observed at high extents of post-drawing and possibly also during core/shell spinning with drawing.
The effect of drawing on CNF restructuring can also be influenced by the extension rate. In post-drawing experiments, CNF filaments were stretched at extension rates ranging from 7.6×10 -5 to 1.0×10 -3 s -1 . 3,6 These are at least three decades lower than the lowest extension rates applied in the current core/shell spinning system. Possibly, the relaxation of CNF proceeds too slowly for the fibrils to react to the fast drawing. The hypothesis of a slow relaxation is supported by the increased alignment of CNF when the residence time of the hydrogel flowing in a needle was increased up to one minute. 18 Also, in rheometry determinations, the velocity profile of CNF (probably connected to hydrogel structure) has been shown to develop over approximately one minute. 19 However, we note that in these studies the CNF hydrogels were subjected to shear instead of extensional flow fields. Under extensional flow, the timescale of CNF alignment has been estimated to be ~0.31 s, 2 which implies that the CNF may react faster to extension than to shear. The slow relaxation, though, may be an advantage in certain cases, such as applications not requiring optimal fibril alignment but rather high porosity optimized by randomly oriented fibrils.

FILAMENT CRYSTALLINITY
The radial integrals of the 2D-WAXD patterns ( Figure S3) show that all the CNF-containing filaments display the typical diffraction peaks for cellulose I crystallites at scattering vectors 12 and 15.8 nm -1 . When CNF was extruded in a GG shell, the peak at 15.8 nm -1 was accompanied by a shoulder at 14.4 nm -1 . This may originate from the contribution of crystallized GG, since the filament with GG only also has a peak at this position. As seen in the diffractogram of the neat CA filament, CA is mostly amorphous and thus invisible in the normalized radial integral of CNF and CA combined. Compared to the neat CNF filament, only a slight increase in the amorphous domain between the cellulose crystallite peaks can be observed for CNF/CA filament. Figure S3. Radial integrals of the X-ray diffraction patterns of filaments coagulated for 5 min in ethanol, apart from GG (5 min in acetone) and CA (fast in water). Spinning conditions: CA -Tables S3, S4; others -Table S1. S-14

PROPERTIES
Physical-mechanical properties of the core/shell filaments spun with different material combinations and coagulation systems are compiled in Table S7. The extrusion speeds and solids content of the precursor dopes mostly determine the filament weight and dimensions. The thinnest and lightest filaments were obtained by extruding only GG (solids content 1 wt.%) at a speed of 3 ml/min. Larger and heavier filaments were spun when changing to CNF (solids content 1.5 wt.%) and adding a GG shell at a speed of 1 ml/min. The maximum dimensions and coarseness resulted from a similar CNF flow combined with a three times slower but 15 times more concentrated CA flow. Surprisingly, when removing the CA from this sample to obtain neat CNF filaments, a large width results, deviating from the trend. This implies that the drying and shell removal procedures flatten the CNF core even more heavily than the whole core/shell filament. Thus, even though the filament has an apparent coarseness of 20.7 tex (approx. one third of its precursor filament with both CNF and CA), this material is spread over a heavily flattened cross-section. As the apparent coarseness and density for this filament are low but it still maintains a fairly high load-bearing capacity, it has a remarkably high tenacity of 9.37±0.6 cN/tex in comparison to the other samples.  (Figure 4b) and water contact angles on the filament materials ( Figure S4), as discussed in the main article. Porosity is expected to pay an essential role in filament-water interactions.
All filaments comprising CNF only had an apparent density below 0.8 g/cm 3 , even though the real density might have been slightly higher for the one drawn at a DR of 9, as described earlier.
For a cellulosic material, a density below 0.8 g/cm 3 corresponds to a porosity of more than 50% (based on the density of pure crystalline cellulose of 1.5-1.6 g/cm 3 ). 13 This suggests that the core/shell spinning produces more porous CNF filaments than the 10% 22 or up to 32% 23 reported in earlier studies using a single needle for spinning of pure CNF. The combination of CNF and GG also allowed spinning of long filaments without breaking.
However, such filaments could be spun continuously only when collected by a conveyor belt (Figure 1c), since the filament was not strong enough for pick up from the bath through air, unless several minutes of immersion were allowed. This implies that the coagulation proceeds more slowly for GG (order of minutes) than for CA (order of seconds). Moreover, the CNF/GG filament could not be consistently drawn while coagulating, apart from the small amount of drawing that naturally occurs during winding and drying. During the winding, the gravity pulls the filament downwards while the winder pulls it upwards; and during the drying, the filament

S-18
tends to contract because of the solvent depletion. The winder restricts the contraction induced by drying, leading to a drawing effect.
Large batches of filament were obtained by optimizing the production speed. Using a larger prototype spinning line (Figure 1b) allowed an increased production rate of CNF/CA filaments, up to 33 m/min, applying the spinning conditions specified in Table S2. Even for the CNF/GG system, adjusting the spinning line with a supportive conveyor belt (Figure 1c) enabled a production speed of up to 1.2 m/min. However, this setup only allows for approximately 3 s of coagulation for CNF and 16 s for GG, considering that a filament element in the core travels through a distance of 24 cm at an average speed of 4.7 m/min, while a shell element covers the same distance at an average speed of 0.9 m/min (Table S2).
The quality of the CNF/GG filaments was seemingly improved by spinning right after re-filling the coagulation bath as full as possible. Filaments spun in a fresh and full coagulation bath appeared both rounder (less flattened) and able to bear more load. Thereafter, the filaments became increasingly brittle when spun to a coagulant that had already been somewhat diluted with water after extended extrusion operation. This phenomenon can be explained by approximating the speed of the coagulation induced by the diffusion of the coagulant into the filament according to Equation S2 21 where W is the thickness of the coagulated layer below the surface of the filament; V is the volume of the coagulant in the bath; D in is the diffusion coefficient of the coagulant into the solvent in the dope; c is the volumetric concentration of the coagulant in the solvent right below the filament surface; L is the length of the studied filament section and R is the filament radius.

Equation S2
shows that increasing the total volume of the coagulant and/or its concentration below the filament surface increase the coagulation speed. During spinning, both of these effects are counteracted, as the coagulant keeps evaporating and diluting due to the incoming flux of the dope solvent. Thus, for optimal spinning conditions a large coagulation bath with purge and make up system would be ideal.