Figure 1 showed the overall strategy of this work with the main stages from the manufacture of freeze-dried NFCt cryogels to the characterization of swollen NFCt-cryogels. After immersion in water (pH 6.5) or in 20 mM HCl (pH 2) solution, dynamics of water molecules were characterized in order to evaluate the impact of the protonation of carboxyl groups on the absorption capacities and the compression properties of such porous materials. Indeed, at pH 2, carboxylic acids are protonated and thus neutral. Therefore, the suppression of the electrostatic repulsion between CNFt fibrils into the walls occurs so that the formation of hydrogen bonds between individual fibrils are strengthened. (Saito et al, 2011)
TEMPO-oxidized CNF (CNFt) dispersion was prepared at 2% w/v CNFt (pH 6.8) to obtain a physical gel as previously showed by rheological measurements. (Mandin et al, 2021) Indeed, at a given critical concentration, entanglements of nanofibrils through hydrogen bond formation and /or steric or electrostatic repulsion resulting in a gel formation. (Benselfelt et al., 2023; Mendoza et al, 2018) Cryogels were then obtained by a freeze-casting in which the freezing process of CNFt gels was achieved into two different ways (Deville, 2013;2017; Ferreira et al, 2021; Jaafar et al, 2020; Mandin et al, 2021): (i) non-directional (ND) freezing for which the CNFt gels were placed in a freezer (at -20°C) so that the nucleation of ice crystals is initiated on the whole surface of the gel and the crystals grow from the outside to the inside of the gel in an isotropic way; and (ii) unidirectional (UD) freezing in which the gel is cooled at -20°C from its bottom so that ice crystals growth is oriented vertically in the sample. In all cases, freezing leads to the densification of CNFt particles between crystal monoliths to form the wall of the cryogels after the sublimation step to remove ice crystals. Thus, porous materials are obtained whose pore size is the replica of the ice crystal size. This procedure avoids the collapse of pores through hydrogen bonding generally observed from air-dried cellulose suspension. (Buchtova et al, 2016; Ferreira et al, 2021) Through visual observation (see photos in Fig. 1), homogeneous cylindric-shaped cryogels were obtained but different external appearance is clearly visible from UD and ND freezing process (Fig. 1). While CNFt-ND cryogels looks like an isotropic foam, oriented and partly shinny walls were observed for UD-CNFt cryogels resulting from the vertical freezing procedure. In both cases, lightweight cylindrical cryogels with an apparent density, ρapp, of 22.6 ± 1.6 kg m− 3, and a very high air content at 98.6 ± 0.1% were obtained from the freeze-drying procedure of the CNFt gel.
CNFt cryogels morphology by SEM and X-Ray tomography
SEM images of longitudinal and transverse cross-sections were produced for both the CNFt-ND and the CNFt-UD cryogels (Fig. 2). It is obvious that the cryogels display different pore morphologies depending on the freezing process. As observed from Fig. 2a and 2c, CNFt-ND presented an interconnected microporous structure (honeycomb-like structure) with no specific orientation; this was expected since the freezing procedure was meant to induce isotropic ice crystal growth. In contrast, the longitudinal and transverse cross-sections of UD-CNFt cryogels (Fig. 2b and 2d, respectively) exhibited a marked orientation of the pore walls in the freezing direction as a result of the directional growth of the ice crystals under the longitudinal temperature gradient.
Additionally, the CNFt-UD cryogels display pores with ellipsoidal shape ranging primarily from 135 to 295 µm in length and from 89 to 124 µm in width. In contrast, the shape pores for CNFt-ND cryogels present dimensions mainly ranging from 200 to 458 µm in length and from 98 to 182 µm in width (Fig. 2e).
X-ray tomography provides complementary results to SEM in that it helps getting three-dimensional data and long-range variations with respect to the mold geometry. It is also a non-destructive technique, with the advantage that no preparation artefact is expected. Regarding the microstructure of the NFCt cryogels from X-ray tomography analysis, it is obvious that the two preparation conditions resulted in very different morphologies (Fig. 3).
NFCt-ND samples have small pores that are essentially isotropic with dimensions (diameters) in the order of 100–400 µm. This microstructure was in strong contrast with that of the unidirectional samples since NFCt-UD cryogels exhibited what appeared as pores with essentially polygonal cross-sections in the transverse direction. The small axis of these pores was in the order of 100–200 µm and the elongated one was in the order of 100–700 µm. Herringbone patterns running lengthwise were also observed with angles in the order of 30, 60 and 120°, which could be reminiscent of the ice crystal structure. An autocorrelation analysis backed up the relative size differences between the samples and it revealed that the microstructure was more periodic in the longitudinal direction for the ND sample and in the transverse direction for the UD sample (Appendix B). This could be explained by the lower temperature gradients in the longitudinal (ND) and radial (UD) directions during the manufacturing step.
For both cryogels, the thickness of the walls appeared to lie in the 10–30 µm range corresponding to 2–4 assembled nanofibrils. In summary, although similar in density, the two NFCt cryogels have very different pore morphologies, with large pores for NFCt-ND and long narrow pores for NFCt-UD cryogels. CNFt-UD cryogels has lower tortuosity due to oriented lamellae as compared to honey-comb like structure of the CNFt-ND cryogels suggesting that an easier diffusion of liquid inside the 3D network of CNFt-ND could be possible as already observed for polyurethane foams (Cui et al., 2020).
Monitoring HCl diffusion and carboxylic acid protonation
Before studying the absorption capacity by immersing cryogels in H2O (pH ∼ 6.5) or 20 mM HCl solution (pH = 2), we checked the impact of the acidic solution on both ND and UD cryogels by visually tracking the diffusion of a coloured pH indicator, methyl orange (MO), inside the CNFt cryogels. Cryogels were initially soaked in aqueous MO solution at pH 6.6 (t = 0) at which a homogeneous yellow coloration is observed after few minutes since pKa value of MO is ∼ 3.5 (Fig. 4a). Then, the cryogel is soaked in HCl solution (20 mM) so the acidic solution is allowed to diffuse freely inside the porous material. As evidenced in the Fig. 4a, colour change from yellow to pink is gradually observed from the edge towards the centre of the material as function of time. The CNFt-UD cryogel is entirely and homogenously coloured pink after 2h soaking while more time is necessary (2h30 min) to fully colour the ND cryogel. As this colour change is indicative of an acid bulk pH, i.e. pH < pKa of MO solution, this indicated that (i) a minimum time of 2 hours is necessary to HCl solution to diffuse entirely inside the CNFt cryogels and (ii) this diffusion process is faster for CNFt-UD than for CNFt-ND showing the influence of the material structuration on the liquid absorption in accordance with suggestions based on microscopy analysis.
In order to verify the protonation of carboxylate groups (-COO−) to carboxyl groups (-COOH) present on C6 carbons of cellulose nanofibrils between the immersion of cryogels in water (pH 6.5) and 20 mM HCl (pH 2.0), FTIR spectra were registered in both conditions. Figure 4b showed FTIR spectra of NFCt-UD cryogels after 10 min immersion in water at pH 6.5 (black line in Fig. 4b), and after 3 h immersion in 20 mM HCl solution at pH 2.0 (red line in Fig. 4b).
Both spectra display the expected cellulose spectra (Li et al., 2011; Schwanninger et al., 2004): at 3400 cm− 1 (vibrations of O-H bonds), 2900 cm− 1 (CH2 stretching vibrations), 1639 cm− 1 (adsorbed water), 1450 − 1230 cm− 1 (O-H and C-H bending vibrations), 1200 − 900 cm− 1 (C-O stretching vibrations), and 650 cm− 1 (O-H bending vibrations). From freeze-dried cryogels after immersion in water (black curve), the band at around 1605 cm− 1 is attributed to the stretching vibration of the carbonyl (C = O) from the deprotonated carboxylate groups (-COO−). The HCl treatment results in the shift of this band to 1728 cm− 1 (red curve) corresponding to the carbonyl (C = O) stretching frequency of the carboxylic acid groups (-COOH), providing evidence of the successful protonation. Identical FTIR spectra were obtained from CNFt-ND cryogels (Fig. S2). These result supports the visual observation of the colour change due to MO absorption (Fig. 4a) after 2 hours.
Dynamic states of water molecules by T 2 relaxation measurements
In order to examine, in more detail, the wetting state of CNFt gels (i.e., before the freeze-casting process) and freeze-dried cryogels after immersion in H2O and HCl solution, the water dynamics were characterized by \({T}_{2}\) TD-NMR measurements. The \({T}_{2}\) relaxation time of water protons is sensitive to variations in the local environment of water molecules. \({T}_{2}\) gives insight on mobility of water molecules depending on their environment, the shorter \({T}_{2}\) the lower the mobility of water molecules. (Codoni et al., 2015; Mandy Nagel et al., 2021) The results are presented in Fig. 5 in which the distribution of the spin-spin relaxation time \({T}_{2i}\) of protons from water molecules in aqueous gel and swelled cryogel samples as function of the relative population (\({P}_{2i}\)) where ‘i’ indicates the mobility domain. \({T}_{2i}\)and \({P}_{2i}\) values are summarized in Table S1.
First, the CNFt gel displays a main \({T}_{2}\) value at 540 ± 6 ms (\({T}_{2b}\)) accounting for 97 ± 1% of the water proton population and a very small proportion (∼3%) of water molecules with \({T}_{2e}\) at 5 ± 2 ms. This preponderant \({T}_{2b}\)domain indicates that water molecules evolve in a relative “homogenous” medium. Guo et al found a 1H \({T}_{2}\)value of 1012 ms (\({P}_{2}\) > 95%) for pectin gel at low concentration (1%) and observed a gradually decrease of T2 value with the increase of the gel concentration indicating that mobility of water molecules is reduced due to the formation of denser network structure with greater chain interlocking and/or more interchain interactions. (Guo et al., 2021)
After the cryogel formation and swelling in aqueous solutions, different mobility of water molecules associated to four or five distinct domains of \({T}_{2}\)values can be clearly distinguished among the different conditions tested, i.e. the cryogel’s structuration and the nature of immersion solution (Table S1): (i) one domain, \({T}_{2a}\), with values ranging from 9000 to 1000 ms, higher than the main \({T}_{2b}\) (540 ms) observed for the gel; (ii) the second domain, \({T}_{2b}\), display values in the same range than the gel, i.e. between 400 and 500 ms; (iii) the third one, \({T}_{2c}\), with values between 140 and 165 ms; (iv) a domain with lower values and only observed for HCl-swollen cryogels, \(d\), between 30 and 70 ms, and finally (v) a domain, \({T}_{2e}\), between 1 and 6 ms. These last three domains, \({T}_{2c}\), \({T}_{2d}\) and \({T}_{2e}\) indicate that some proportions of water molecules are found in environments with more constrained mobility than in the gel phase.
The relative population of water molecules observed at \({T}_{2}\) ≥ 400 ms (\({T}_{2a}+ {T}_{2b})\)suggests that the \({T}_{2b}\)domain from gel is splitted into two water molecule populations, \({P}_{2a}\) and \({P}_{2b}\), due the structuration of the CNFt gel during the freeze-casting procedure. One corresponding to the pores that do not contain any fibril due to phase separation during ice crystallisation and the second corresponding to gel phase likely close to the cryogel walls. In the initial step of the liquid immersion, the bulk water fills the micro-pores without affecting the structure of both cryogels so that this bulk water could be represented by these two domains observed at \({T}_{2}\) ≥ 400 ms (Fig. 5) where some water molecules are free to diffuse inside the micropore volume (\({T}_{2a}\) = 900–1000 ms) while the others (\({T}_{2b}\) = 400–500 ms) are located in the same environment of the initial gel. In this domain range, the water molecule population is slightly higher for CNFt-UD cryogels in H2O (91% vs 85% for CNFt-ND) and in HCl solution (68% vs 62% for NFCt-ND) that could be linked to the pore morphology of cryogels with more elongated pores for UD-NFCt (Fig. 3) enabling water molecules to move easily in the longitudinal direction. Water molecules at \({T}_{2a}\) and \({T}_{2b}\) are thus seeing their population to decline as compared to gel for both ND and UD cryogels in favor of the emergence of new populations (P2c and P2d) of water molecules at lower values, \({T}_{2c}\) (around 150 ms) and \({T}_{2d}\) (between 28–66 ms). Interaction through hydrogen bonds with -OH or -COO− / COOH groups would correspond to water molecules for which their mobility is slowed down. It should be noted that \({P}_{2a}\) and \({P}_{2b}\) populations decrease more strongly (-37% and − 31%) after HCl immersion as compared to H2O-swollen cryogels (-14% and − 8%) for ND and UD cryogels, respectively. Indeed, the new domain (\({P}_{2d}\)) observed between 28 and 66 ms (\({T}_{2d}\)) is specific to the HCl-swollen cryogels and it could be attributed to water molecules in interaction through hydrogen-bonds with protonated carboxyl groups. For CNFt-UD cryogels, this domain at 43.1 ms \({T}_{2d}\) value represents 7% of the total water molecules. Differently, in this “d” domain, CNFt-ND cryogels display a splitting of water populations (at \({T}_{2d}\) 66 and 28 ms) corresponding to 13% of water molecules. Consequently, populations with reduced mobility were higher at acidic pH and this was more marked for ND cryogels than UD cryogels suggesting the presence of more available carboxylic groups in ND cryogels and underlining the presence of hydrogen bonds between H2O and carboxyl groups.
Finally, the presence of water molecules with a restricted mobility (\({T}_{2e}\)) appears at a trace level (3% at 1 ms for CNFt-ND and 2% at 3 ms for CNFt-UD) as for the CNFt gel (1% at 5ms). These domains of mobility are thus to be put in relation with water molecules interacting to a greater or lesser extent with hydroxyl or carboxyl groups present at CNFt nanofibrils surfaces through hydrogen bonds. It could be also attributed to water molecules located in a closed volume restricting their mobility or in closed contact to cellulose surface.
Liquid absorption capacity of NFCt-ND and -UD cryogels
In order to investigate the impact of the protonation of available carboxyl groups on the swelling properties of CNFt cryogels during the immersion in HCl solution, the absorption capacity of cryogels was thus determined by immersing the specimens in water or in 20 mM HCl solution during 120 min and by weighing them at regular time intervals. The Fig. 6 presents the absorption capacity, Qt, expressed in grams of absorbed solution per gram of freeze-dried CNFt-ND and CNFt-UD cryogels during 120 min.
Table 1
Absorption capacity (Qe) and swelling degree (SR) of CNFt-ND and CNFt-UD in swollen state after immersion in water and aqueous HCl (20 mM) solution after 20 min.
Cryogels
in swollen state
|
Qe
(g g− 1)
|
SR at 20 min
(%)
|
H2O
|
HCl
|
H2O
|
HCl
|
CNFt-ND
|
41.0
± 0.7
|
37.6
± 0.8
|
29
± 2
|
27
± 1
|
CNFt-UD
|
46.0
± 1.9
|
43.7
± 1.1
|
24
± 7
|
18
± 4
|
As it can be seen, both ND and UD cryogels absorbed both liquid rapidly (< 2 min) due to their high porosity and hydrophilic character of cellulose. Both cryogels absorbed around 42 g g− 1 in water and HCl solution after 20 min immersion (Table 1) which corresponds to absorption capacity values usually found in the literature for such biobased porous materials. (Darpentigny et al, 2020; Jiang et al, 2014; Mendoza et al., 2019) However, different absorption kinetics were clearly observed between cryogels after immersion in water or HCl solution.
In water, after the rapid initial absorption rate, the absorption capacity increases slowly up to 20–25 min immersion as commonly observed for such systems in water (Barajas-Ledesma et al., 2020; Mendoza et al, 2019), but beyond a plateau region was not reached and both cryogels continue absorbing water suggesting a slow breaking of fibril network. As showed in Figure S3, after 20 min of water absorption, the swelling ratio (SR) of CNFt-ND continues to rise (+ 12% at 120 min) until the cryogel completely disintegrates after 12h. Differently, CNFt-UD cryogels swelled less (+ 4% at 120 min) and it looks like a hydrogel with a collapse of the base after 24 h (photo in Fig. 6) but it cannot be manipulated without destroying it.
At pH 6.5 (in water), the carboxylic groups present on C6 carbons are all charged (carboxylate form) and contribute to the water molecule attraction on the cellulosic surface. In this pH condition, water molecules penetrate into the cryogel structure and interact with available –OH and –COO− groups up to a certain extend weakening the cohesion between CNFt fibrils into the walls by increasing the average distance between adjacent charges and thus reducing the resistance to swelling of the cryogels. This thermodynamically favoured the expansion of CNFt walls up to the disruption of the cryogel structure as observed after cryogels immersion in H2O during several hours.
When immersed in HCl solution, the absorption process took effect in 1 min and then, cryogels did not display further liquid absorption capacity as function of time. In this case, the maximum equilibrium capacity (Qe) was achieved within the first 2 min and remains constant over time up to 120 min and more. Qe values at 20 min are 44 ± 1 g g− 1 for CNFt-UD cryogels and 37.6 ± 0.8 g g− 1 for CNFt-ND cryogels, respectively, that are slightly lower than values obtained for the corresponding cryogels in water (Table 1). The maximum SR is thus reached within the first few minutes of immersion and CNFt-UD swelled less (SR = 18 ± 4%) than CNFt-ND (SR = 27 ± 1%). It should also be noted that in HCl solution, CNFt cryogels swelled much less than in H2O and that NFCt cryogels in HCl solution maintained their shape and are easy to handle (photos in Fig. 6). Any destructuration is observed after several months in HCl solution. Consequently, at pH 2 due the protonation of carboxyl groups, the penetration and interaction of water molecules inside the wall structure of the UD cryogels was limited, lowering their swelling capacity.
These results evidenced that pore morphology of the porous materials and the pH of the immersion solution play a key role on the absorption process. In all cases, although absorption capacity of UD cryogels is always higher than ND cryogels but surprisingly their swelling degree is lower than ND cryogels indicating that the pore morphology of CNFt cryogels is the key parameter and that lamellar structuration is more resistant to the swelling process than honeycomb structure.
Preliminary results from X-Ray tomography of capillary uptakes in these foams (not shown here) have demonstrated that water uptake was fast and complete in the UD samples when water uptake was permitted in the longitudinal direction. In contrast, water uptake was much slower in the ND samples, probably due to increased tortuosity and a subsequent permeability decrease. Transverse water uptake was not always possible in the UD samples, demonstrating the strong texture of the unidirectional manufacturing path.
Shape recovery ability of cryogels after mechanical compression
The ability of acid-swollen cryogels to absorb and release water repeatedly while maintaining their structure and shape was observed as illustrated on the video (see Video in the Supplementary Material). After several manual compression, the cryogel return back to its original shape in few seconds and this sponge-like behaviour was conserved even after several months of storage time.
To better characterized this shape recovery ability, the swollen CNFt cryogels were evaluated in mechanical compression of cylindrical samples in the longitudinal direction, i.e. the freezing direction for UD cryogel. Uniaxial compression cycles of the cylindrical CNFt cryogels were thus investigated after swelling in water or in 20 mM HCl solution and the stress-strain curves of wetted CNFt cryogels prepared by UD and ND freezing methods are presented in Fig. 7 and their elastic modulus E (kPa) and critical stress, σc (kPa) are showed in Table S2.
In a first observation of the Fig. 8, during cycle 1 (blue curves), cryogels exhibited linear elastic regimes at low strain, afterwards a plastic plateau is observed with a slow increasing of the stress up to about to 50% strain. In the subsequent compression cycles, the stress-strain curves display a decrease in strength and stress at plastic-plateau. This is particularly marked for specimens immersed in HCl. Also, cryogels after immersion in H2O (Fig. 8A, B) no longer maintained their structural integrity after 20 compression cycles while HCl-treated cryogels could support at least 50 compression cycles (Fig. 7c, d) with a less destructuration of their macrostructure.
After immersion in water, during the cycle 1 (Fig. 7a, b; blue curves), the linear domain corresponding to the elastic deformation of the foams is twice more pronounced for CNFt-UD cryogels with Young’s modulus values of 42.5 kPa than CNFt-ND cryogels (21 kPa). Also, after the linear elastic deformation, at a critical stress of 2.4 kPa, CNFt-UD cryogels (Fig. 7b) display a more marked plateau region as compared to stress-strain curves of ND cryogels for which σc is lower at 1.9 kPa (Table S2). This plateau region corresponds to the collapse of CNFt walls due to plastic yielding as commonly observed for polymeric foams (Gupta et al., 2018; Patterson et al, 2023; Sehaqui et al., 2010).
In the second cycle, the Young’s modulus values decrease by a factor of two for CNFt-ND cryogel to ∼12 kPa as compared to the first cycle while this decrease is a little more marked (factor of three) to ∼15 kPa for CNFt-UD, respectively. In the subsequent cycles, the Young modulus continue to decline at about 6 kPa at the 10th cycle and the elastic-plateau domain transition becomes difficult to determine indicating the de-structuring of the material as observed in water absorption experiments. The weakening of hydrogen bonds between individual nanofibrils thermodynamically favoured the expansion of CNFt walls up to the disruption of the cryogel structure as observed after cryogels immersion in H2O during several hours and supported by the cryogel macrostructure that cannot withstand compression cycles after of 15–20 cycles. However, before completely destroyed over 15–20 compression cycles, CNFt-ND and CNFt-UD cryogels display excellent shape-recovery performance and they can recover their original shape at 90–95% as showed in Fig. 9.
When cryogels are immersed in HCl solution, the difference in strength and plateau stress of cryogels between H2O and HCl were 4–6 times higher after the first cycle indicating the reinforcement of CNFt walls (Fig. 8c, d). Indeed, higher values of the Young’s modulus were obtained: 119 and 151 kPa for HCl-swollen CNFt-ND and CNFt-UD cryogels, respectively, as compared to their corresponding H2O-swollen cryogels (Table S2). This increase is much more important for CNFt-ND with an improvement of a factor of 6 than this one for CNFt-UD cryogel (factor of 3). In the same time, critical stress values were 4 times higher after HCl immersion (8–9 kPa) for both cryogels. This indicates a reinforcement of the structure under compression after protonation of carboxyl groups at the surface of CNFt due to the reduction of repulsion forces between individual NFCt and the formation of hydrogen bonds between carboxyl groups.
As observed in water-swollen cryogels, cyclic compression leads to a dramatic decrease in Young's modulus and critical stress from the second cycle onwards.: E = 31 kPa/σc = 1.8 kPa and E = 22 kPa/σc = 1.5 kPa for HCl-swollen CNFt-ND and CNFt-UD, respectively (Table S2). This reduction in E is observed up to the 5th cycle, after which the elastic modulus remains constant at around 25 kPa and 14 kPa for CNFt-ND and CNFt-UD, respectively, up to the 50th compression cycle. The critical stress also decreases and despite their lower values, it can be determined across the compressive cycles for both cryogels.
From Fig. 7e, HCl-swollen cryogels exhibit a slightly lower performance with ∼87% shape recovery after the first cycle but show ∼76% shape recovery performance after 50 cycles with preservation of its macrostructural integrity. When HCl-treated CNFt cryogels are soaked in water and then undergo cyclic compression tests in water (Fig. S4), they maintain their mechanical performances up to 50 cycles of compression (Table S2). Furthermore, their shape recovery performances were enhanced as compared to the corresponding HCl-swollen cryogels as 85% shape recovery was obtained after 50 compression cycles.
These results illustrate that, after 50 cycles of compression, HCl-swollen CNFt-ND and CNFt-UD cryogels display still high shape recovery performance demonstrating the efficiency of the physical crosslinking on their mechanical properties. The post-crosslinked cryogels display an increase of the compressive modulus up to a factor 6 (for CNFt-UD) and this mechanical improvement is then preserved in water. Finally, it should also be noted that NFCt-UD cryogels still have slightly better mechanical properties than CNFt-ND.