Carbohydrate Polymers 123 (2015) 416–423
Contents lists available at ScienceDirect
Carbohydrate Polymers journal homepage: www.elsevier.com/locate/carbpol
Rheology of semi-dilute suspensions of carboxylated cellulose nanofibrils Leila Jowkarderis a , Theo G.M. van de Ven b,∗ a b
Department of Chemical Engineering, McGill University, Montreal, Quebec H3A 0C5, Canada Pulp & Paper Research Center and Department of Chemistry, McGill University, Montreal, Quebec H3A 2A7, Canada
a r t i c l e
i n f o
Article history: Received 27 November 2014 Received in revised form 28 January 2015 Accepted 30 January 2015 Available online 9 February 2015 Keywords: Cellulose nanofibrils Rheology Creep–recovery Critical strain Ionic strength Polyelectrolyte
a b s t r a c t Cellulose nanofibrils (CNF) in water make entangled networks and stiff gels, which have a number of promising applications. In this work, the rheology of semi-dilute TEMPO-mediated oxidized CNF hydrogels, and the effects of cationic polyacrylamide and calcium ions on their viscoelastic properties are investigated. The elastic modulus varies with CNF volume fraction with a power law exponent of 4.52. Creep–recovery results show that suspensions with higher mass fractions exert a higher resistance against deformation, and a higher degree of recovery. Low ionic strengths and polyelectrolyte concentrations increase the creep deformation because of screening the surface charge. Higher ionic strengths and polyelectrolyte concentrations lead to fibril aggregation, which stiffens the network structure, decreasing the creep deformation. However, the recovery response is not significantly affected by additives. The critical strain at the onset of non-linear viscoelasticity is independent of mass fraction in two different concentration regimes, with a transition at 0.35% w/w CNF. © 2015 Elsevier Ltd. All rights reserved.
1. Introduction Increasing environmental concerns have led to efforts to increase the use of renewable and biodegradable materials in many industrial applications. Wood pulp is one of the major sources of raw materials for bioproducts. Disintegration of wood fibers into cellulose nanofibrils (CNF) has recently attracted significant interest. Cellulose is a linear polysaccharide with a large number of hydroxyl groups. CNF consists of aligned extended molecules that are adhered together by hydrogen bonds. Due to their high aspect ratio and high degree of crystallinity, cellulose nanofibrils form stiff hydrogels with high elasticity. They have been recommended as rheology modifier for food, paint and cosmetics (Herrick, Casebier, Hamilton, & Sandberg, 1983; Turbak, Snyder, & Sandberg, 1983). CNF films and CNF coated papers have been suggested for packaging applications (Henriksson, Berglund, Isaksson, Lindström, & Nishino, 2008; Spence, Venditti, Rojas, Habibi, & Pawlak, 2010; Syverud & Stenius, 2009). CNF has been also proposed to be used in the production of functional materials such as nanocomposites (Lopez-Rubio et al., 2007; Nakagaito & Yano, 2005) and antimicrobial films (Andresen et al., 2007).
∗ Corresponding author. Tel.: +1 5143986177. E-mail address: [email protected] (T.G.M. van de Ven). http://dx.doi.org/10.1016/j.carbpol.2015.01.067 0144-8617/© 2015 Elsevier Ltd. All rights reserved.
Since 1983, several methods have been developed for the preparation of cellulose nanofibrils. Fibril dimensions and the number of surface charges vary depending on the preparation method. CNF is mechanically produced by disintegration of wood fibers in a homogenizer under high shearing forces. The resulting nanofibrils are 5–20 nm in diameter and several micrometers in length (Herrick et al., 1983; Turbak et al., 1983). To avoid the high energy consumption for mechanical fibrillation of wood fibers, chemical pretreatment of wood pulp has been suggested. Chemical methods are mainly based on inducing strong electrostatic repulsions among the nanofibrils in water, by introducing negatively charged functional groups onto their surface. Nanofibrils can then be disintegrated by gentle mechanical treatments such as sonication. It has been shown that mild enzymatic hydrolysis of softwood pulp decreases the homogenization energy, and produces high aspect ratio nanofibrils with high mechanical strength (Henriksson & Berglund, 2007; Paakko et al., 2007). Paakko et al. (2007) added monocomponent endoglucanase to the softwood pulp suspension and passed it through a homogenizer after refining. The achieved fibril diameter was 5–20 nm, providing strong aqueous gels with large elastic modulus, practical for reinforcing multicomponent mixtures. TEMPO-mediated oxidation with sodium hypochlorite followed by gentle mechanical sonication has been introduced as another preparation method, yielding cellulose nanofibrils with 3–4 nm diameter, several micrometer length, and a large
L. Jowkarderis, T.G.M. van de Ven / Carbohydrate Polymers 123 (2015) 416–423
number of surface charges in the form of COO− groups (Isogai, Saito, & Fukuzumi, 2011; Saito, Nishiyama, Putaux, Vignon, & Isogai, 2006). Water dispersions of TEMPO-mediated oxidized CNF showed pseudo-plastic behavior, and a strongly consistency-dependent shear viscosity at concentrations >0.2% w/w (Lasseuguette, Roux, & Nishiyama, 2008; Saito, Kimura, Nishiyama, & Isogai, 2007). Fibril dimensions and its charge density directly affect the suspension rheology. It has been reported that the elastic modulus of CNF suspensions produced by TEMPO-mediated oxidation of palm tree wood pulp is larger than previously reported by Paakko et al. (2007) for CNF prepared by enzymatic hydrolysis, in spite of the similar fibril dimensions (Benhamou, Dufresne, Magnin, Mortha, & Kaddami, 2014). The authors ascribe it to the higher charge density of their oxidized fibrils compared to the non-oxidized CNF. The rheology of colloidal suspensions can be also affected by interparticle bridging (Abbas Zaman & Delorme, 2002; Otsubo, 1999; Swerin, 1998). This phenomenon is used in industry to control the product viscosity and stiffness. Bridging can be induced by addition of polyelectrolytes or multivalent ions. Oxidized cellulose nanofibrils are negatively charged, therefore cations such as Ca2+ , and polymers such as cationic polyacrylamide can flocculate them. Cationic polyacrylamide is widely used in papermaking as a retention aid (Hubbe, 2007; Vanerek, Alince, & van de Ven, 2000a, 2000b). Retention of fine particles necessitates their deposition on the fibers and fiber flocculation. Fine particles could be mineral fillers which are very small and might be lost during filtration in the papermaking process (van de Ven, 1984), or they could be fiber fines or hydrophobizers used for internal sizing (Hubbe, 2007). Paper machine conditions can significantly decrease the chemical efficiency of these agents if they are not retained in the sheet. Calcium ions are also present in papermaking suspensions, especially when calcium carbonate CaCO3 is used in paper as a filler. This paper aims to study the rheological behavior of TEMPOmediated oxidized CNF hydrogels in the semi-dilute regime. We have studied the rheology of dilute CNF suspensions previously (Jowkarderis & van de Ven, 2014). Moreover, we investigate the effect of interfibrillar bridging on the mechanical strength of the suspensions using cationic polyacrylamide and CaCl2 at various concentrations. Experimental results on the creep–recovery response, and the variations of the elastic modulus as a function of CNF volume fraction are compared to available theoretical models.
417
Polymer bridging was induced using cationic polyacrylamide (CPAM) (PERCOL292), with mass average molecular weight ∼5 ×106 , and degree of substitution ∼20%. A 5 g L−1 polymer solution was prepared by adding dry polymer beads to deionized water and subjecting it to magnetic stirring for 18 h. The solution was then used to make 0.3% w/w CNF suspensions with polymer concentrations in the range 10–200 mg g−1 dry CNF. Since the amount of polymer used in the suspension is very small, maximum ≈600 ppm, its effect on the viscosity of the suspending medium is negligible (less than 1%). CaCl2 , purchased from Sigma–Aldrich, was used for physical cross-linking of cellulose nanofibrils. A 1 M salt solution was added to 0.3% w/w CNF suspensions to make samples with salt concentrations ranging from 1 × 10−5 to 5 × 10−3 M. All samples were tested the second day after preparation. All the measurements were performed at pH ≈6.9. 2.2. Rheometry Rheology measurements were carried out using DHR (TA Instrument) and MCR 302 (Anton-Paar) rheometers. Parallel plates geometry (40 mm diameter) was used to study CNF suspensions with mass fractions m ≥ 0.3 %. Consistent results were obtained for all suspensions using either sand blasted, roughened, or smooth surface plates. The gap was set to 1 mm. Changing the gap from 1 to 2 mm did not affect the results. Suspensions with m < 0.3 % were tested using double-wall concentric cylinders (internal gap: 0.41 mm, external gap: 0.47 mm, effective height: 40 mm). The linear viscoelastic (LVE) region was determined at frequencies 0.5, 6.28 and 50 rad s−1 , over a strain range of 0.1–1000%. Subsequently, frequency sweep tests were conducted at 3% strain, which was within the linear region at all CNF volume fractions. The creep tests were conducted at stress = 0.2 Pa, for 10 min. The stress was then suddenly released and the recovery response was recorded for 10 min. Solvent traps were used to prevent water evaporation in all the tests. The temperature was set to 25 ◦ C. The samples were allowed to relax for 5 min before each measurement. All the tests were performed in triplicate with fresh samples, and their average is reported here. 3. Results and discussion 3.1. Elastic and viscous moduli
2. Experimental 2.1. Materials A 0.67% w/w aqueous suspension of CNF produced by TEMPOmediated oxidation of spruce wood pulp was received from Forest Products Laboratory (FPL) (Madison, WI, USA). The COONa density was reported as 0.65 mmol g−1 dry CNF. The dimensions of the nanofibrils were measured previously as ≈4.7 nm in diameter and ≈550 nm in length, and average aspect ratio r ≈ 110 (Jowkarderis & van de Ven, 2014). Due to the polydispersity in cellulose nanofibrils dimensions, it is not easy to directly determine the volume fraction of CNF suspensions. In this paper, the volume fraction is considered as ≈ m /r , where m is the mass fraction and r ≈ 1.5 is the relative density of cellulose (Hermans, Hermans, & Vermaas, 1945). CNF suspensions with mass fractions m ranging from 0.1% to 0.67% w/w were prepared by addition of deionized water, corresponding to volume fractions in the range 7 × 10−4 to 4.5 × 10−3 , within the semidilute regime. Recall that semi-dilute suspensions are obtained when r−2 < < r−1 (Mewis & Wagner, 2012).
The elastic G and viscous G moduli of TEMPO-mediated oxidized cellulose as a function of frequency are shown in Fig. 1, at various mass fractions. Our results are consistent with the data previously reported by Benhamou et al. (2014). G and G increase with increasing m , indicating the formation of stiffer networks. However, the G and G values are more than one order of magnitude larger than the values obtained by Lasseuguette et al. (2008) at similar mass fractions. The reason can be incomplete disintegration of wood fibers in their system, after mechanical treatments. Non-disintegrated fibers do not form strong network structures, and their dispersions are weaker than suspensions of nanofibrils with equal consistency. The charge density of the oxidized CNF is not mentioned by the authors. At m > 0.3 %, the elastic and viscous moduli are comparable to the observations of Paakko et al. (2007) for enzymatic hydrolyzed cellulose, and of Agoda-Tandjawa et al. (2010) for cellulose microfibrils prepared by strong mechanical treatments, with fibril dimensions larger than the oxidized fibrils used here. Therefore at high mass fractions, the higher surface charge compensates for the effect of smaller fibril dimensions. At m ≤ 0.3 %, the increase of G and G with frequency is noticeable, while Paakko et al. (2007)
418
L. Jowkarderis, T.G.M. van de Ven / Carbohydrate Polymers 123 (2015) 416–423
Fig. 2. G vs. at ω = 6.28 rad s−1 . Fig. 1. G (solid symbols) and G (open symbols) of CNF suspensions as a function of frequency, at various mass fractions. Error bars (smaller than the size of the symbols) show the standard deviation from the average of three measurements.
reported only a slight increase of the elastic modulus with frequency, for m as low as 0.125% w/w, indicating stiffer structures. This shows that at low mass fractions, the effect of fibril dimensions on the mechanical strength of the suspension dominates over the effect of charge content. Fig. 1 also shows that at m < 0.15 %, G < G in the entire range of applied frequency, indicating liquid like behavior. At m = 0.15%, G G , implying that the elastic properties are comparable to the viscous properties. Therefore, m ≈ 0.15 % is the gelation point of TEMPO-mediated oxidized cellulose with a surface charge ≈ 0.65 meq g−1 . In enzymatic hydrolyzed CNF suspensions, at m ≈ 0.12 %, G is about one order of magnitude larger than G (Paakko et al., 2007). Therefore, gelation starts at lower mass fractions of non-oxidized CNF, again highlighting the stronger effect of fibril dimensions over the surface charge at low mass fractions. A significant increase in the elastic modulus upon increasing the concentration has been reported in the literature for cellulosic gels (Iotti, Gregersen, Moe, & Lenes, 2011; Naderi, Lindström, & Pettersson, 2014; Naderi, Lindström, & Sundström, 2014). A power law equation, G ∝ cn , is suggested in most cases, where c is the fibril concentration. The elastic modulus reported by Paakko et al. (2007) scales with concentration with n ≈ 3, and Agoda-Tandjawa et al. (2010) reported n = 2.58, at ω = 1 Hz. Naderi, Lindström, and Sundström (2014) found n = 2.4 for carboxymethylated CNF at mass fractions m ranging from ≈0.3% to 6.2%, at ω=1 Hz. Naderi, Lindström, and Pettersson (2014) later reported that G scales with n ) with a power law exponent of 5.2 at lower mass fraction (G ∝ m m in the range 0.04–0.18%, at ω = 0.5 Hz. Models developed for the elastic modulus of fibrous gels as a function of concentration predict a power-law exponent of around 2 (Jones & Marques, 1990; Kroy & Frey, 1996; Satcher & Dewey, 1996), which is low compared to the values obtained for CNF suspensions. More recently, Hill (2008) has offered a scaling theory for the elastic modulus of CNF gels which yields a transition from a power law exponent of 11/3 to 7 depending on the fibril volume fraction . In a random and entangled microstructure he predicts G ≈ ˇE11/3 (1 + ˛) >
10/3
(1)
for ∼ 2.5a/L, where a is the fibril radius, L is the fibril length, ˇ is an order-one coefficient depending on the fibril characteristics, e.g. its crystallinity, E is the Young’s modulus, and ˛ is an index of twobody interactions depending on the electrolyte concentration and
fibrils surface charge. When is smaller than the volume fraction required for entanglement, which applies to this work, Hill’s theory predicts
G ≈ ˇE
11/3 (1 − )8/3
1+
˛ (1 − )
10/3
.
(2)
0.99). Recall from Fig. 1, G is significantly frequency dependent at low concentrations, therefore, at different frequencies, G relates to by different power law exponents. In this work, by increasing ω from 0.5 to 50 rad s−1 , n decreases from 4.78 to 4.02, indicating that at lower frequencies, the elastic modulus is more sensitive to volume fraction. The effect of frequency is neglected in available models for G as a function of fibrous gels consistency, mainly because the elastic modulus is almost frequency independent at high concentrations. However, the noticeable decrease of the power law exponent with increasing ω in the semi-dilute regime suggests that the frequency dependence of the elastic modulus should be considered in developing future theories. Hill’s model is in agreement with our experimental results at ω = 1 Hz, yielding k1 = ˇE/(1 − )8/3 = 18 GPa and k2 = ˛/(1 − ) = 93 (R2 > 0.99). The degree of crystallinity of TEMPO-mediated oxidized CNF has been reported to be high ≈75%, because oxidation with sodium hypochlorite does not change the original crystallinity of cellulose (Isogai et al., 2011; Rodionova et al., 2013). The Young’s modulus for the CNF with ∼75% crystallinity has been shown to be E ≈ 50 GPa (Eichhorn & Young, 2001), however, the magnitudes of ˇ, ˛ and are unknown. Hill (2008) calculated ˇ ≈ 1.8 and ˛ ≈ 31.5, by fitting Eq. (1) to the experimental data of Paakko et al. (2007) for enzymatic hydrolyzed CNF at ω = 1 Hz. These values may not be valid for our CNF suspensions because of the higher degree of crystallinity and the higher surface charge. The experimental data obtained at frequencies other than 1 Hz fit to Hill’s model with different values of k1 and k2 . We presume that ˇ and ˛ should not depend on frequency, therefore, should be the parameter which causes the variations in k1 and k2 .
L. Jowkarderis, T.G.M. van de Ven / Carbohydrate Polymers 123 (2015) 416–423
Fig. 3. Creep–recovery response of CNF suspensions at various mass fractions, = 0.2 Pa.
3.2. Creep–recovery tests 3.2.1. Creep–recovery response of CNF hydrogels The creep–recovery test is important in industrial applications because it shows how the material responds to an applied load. The measurements involve two steps. First, the creep test: upon the application of a constant stress for a certain time, the material deforms to a maximum strain ( max ). Second, the recovery test: the stress is released, and the material starts reforming its initial structure. In principle, as long as the applied stress is not higher than the yield stress, an elastic solid recovers all the strain. A viscoelastic material, however, cannot recover its initial structure, and some part of the deformation permanently remains in the sample. The unrecovered strain (v ) represents the viscous part of the viscoelasticity, and the recovered deformation (e = max − v ) represents the elastic part. The recovery percent is defined as Rec%=100 × e / max (Mezger, 2006). Fig. 3 shows the creep–recovery results for CNF suspensions with various concentrations at stress = 0.2 Pa. The yield stress for the suspensions shown in this figure varies from y ≈ 0.8 Pa at m = 0.3%, to y ≈ 10 Pa at m = 0.67%. As expected, deformation ( max ) decreases with increasing mass fraction, indicating an increase in stiffness. However, the strain is not completely recovered after releasing the stress, showing that CNF hydrogels have some viscous character in the semi-dilute regime, even when their response is predominantly elastic (Mezger, 2006). 3.2.2. Effect of polyelectrolyte The effect of polyelectrolyte on the mechanical strength of 0.3% CNF suspensions at = 0.2 Pa is shown in Fig. 4. The maximum deformation max increases with increasing the amount of C-PAM up to 50 mg polymer g−1 CNF, indicating a decrease in suspension stiffness. Adding more polymer, however, increases the sample stiffness, as max decreases. The transition happens at a polymer concentration between 50 and 100 mg polymer g−1 CNF, corresponding to about 0.1 to 0.2 mg polymer m−2 fibril surface. The reason is that C-PAM affects the suspension rheology in two ways. One is screening the surface charge, which causes a decrease in mechanical strength and viscosity (Jowkarderis & van de Ven, 2014). The other way is bridging between neighboring fibrils, which strengthens the network structure and leads to a higher resistance against deformation. At low polyelectrolyte concentrations, the effective surface charge controls the rheological properties, therefore the suspension stiffness decreases by adding C-PAM, and a large deformation is observed. At high concentrations of C-PAM, on
419
Fig. 4. Creep–recovery response of 0.3% CNF suspensions at various concentrations of C-PAM in units of mg g−1 CNF, = 0.2 Pa.
the other hand, interparticle bridging compensates for the decrease of electrostatic forces, and stiffens the material structure, therefore the deformation decreases. Variations of the elastic modulus G of 0.3% CNF suspensions as a function of C-PAM concentration also show lower stiffness at low, and higher stiffness at high C-PAM concentrations (data available, not shown). 3.2.3. Effect of CaCl2 Fig. 5 shows the effect of calcium ions on the creep–recovery response of 0.3% CNF suspensions at = 0.2 Pa. Similar to the effect of polyelectrolyte, network deformation increases at small CaCl2 concentrations, [Ca2+ ] 0.99 was obtained in all cases. Increasing the CNF mass fraction and the amount of CaCl2 (>0.2 mM) and >
C-PAM (∼100 mg g−1 ) increase the Burger model constants, indicating an increase in the suspension stiffness and rigidity (Laguna et al., 2013). Table 1 also shows the permanent strain ∞ calculated by Eq. (5) for all the suspensions, compared to the experimentally measured value v .
L. Jowkarderis, T.G.M. van de Ven / Carbohydrate Polymers 123 (2015) 416–423
421
Fig. 6. (a) G (solid lines) and G (dashed lines) versus strain for two CNF mass fractions, ω = 1 Hz. Solid arrow shows the maximum deformation in creep–recovery test, and dashed arrow shows the critical strain. (b) c versus CNF mass fraction, ω = 1 Hz.
3.3. Critical strain A linear viscoelastic (LVE) region, over which the elastic and viscous moduli are independent of the oscillation amplitude, was observed for all the suspensions. Fig. 6a shows the strain sweep results for two CNF mass fractions. The end of the LVE region where G or G begins to change upon increasing the strain is called the critical strain c , indicating the network breakage point. Exceeding the critical strain causes irreversible deformation in the network structure (Mezger, 2006). In this study, c is evaluated as the strain at which G has decreased to 90% of its linear value. It has been reported that in suspensions of softwood kraft pulp fibers, there is no variations in c by varying the mass fraction in the range 3–8% (Swerin, Powell, & Ödberg, 1992). Swerin (1998) also reported almost equal values of c for suspensions of softwood and hardwood pulp fibers with similar fiber diameters. He explains that at the critical strain, the applied shear disconnects the fiber–fiber contact points, introducing gaps between the connecting fibers, and breaks the network structure. The gap required to break a contact point might be a function of the fiber diameter, therefore, suspensions with different fiber types, but similar dimensions, exhibit similar critical strains (Swerin, 1998). Benhamou et al. (2014) on the other hand, have reported a decrease in c when the mass fraction of TEMPO-mediated oxidized CNF increases from 0.1% to 1%. Fig. 6b shows c for CNF suspensions with mass fractions m ≥ 0.2 % which exhibit G > G . There are two m regions in which
c is almost independent of mass fraction. The critical strain in suspensions with m ≥ 0.4 % is smaller than at m ≤ 0.3 %. High density network structure starts to break at smaller oscillation amplitudes. At lower concentrations, however, fibril networks do not exist, or they exist with low densities, therefore the elastic modulus starts decreasing at larger amplitudes (Amaria, Uesugi, & Suzukib, 1997). The transition happens at m ≈ 0.35 %. As previously shown in Fig. 1, the frequency dependence of G and G becomes less pronounced >
at m ∼0.35%. This mass fraction, therefore, can be identified with the onset of high density network formation. Comparing the strain sweep results to the creep–recovery responses shows that irreversible network deformation happens at strains < c . For example, the observed c at m = 0.5% is ≈10.1%, as shown with the dashed arrow in Fig. 6a; however the suspension does not recover more than ≈70%, after deforming to max ≈ 2.3 % (
Contents lists available at ScienceDirect
Carbohydrate Polymers journal homepage: www.elsevier.com/locate/carbpol
Rheology of semi-dilute suspensions of carboxylated cellulose nanofibrils Leila Jowkarderis a , Theo G.M. van de Ven b,∗ a b
Department of Chemical Engineering, McGill University, Montreal, Quebec H3A 0C5, Canada Pulp & Paper Research Center and Department of Chemistry, McGill University, Montreal, Quebec H3A 2A7, Canada
a r t i c l e
i n f o
Article history: Received 27 November 2014 Received in revised form 28 January 2015 Accepted 30 January 2015 Available online 9 February 2015 Keywords: Cellulose nanofibrils Rheology Creep–recovery Critical strain Ionic strength Polyelectrolyte
a b s t r a c t Cellulose nanofibrils (CNF) in water make entangled networks and stiff gels, which have a number of promising applications. In this work, the rheology of semi-dilute TEMPO-mediated oxidized CNF hydrogels, and the effects of cationic polyacrylamide and calcium ions on their viscoelastic properties are investigated. The elastic modulus varies with CNF volume fraction with a power law exponent of 4.52. Creep–recovery results show that suspensions with higher mass fractions exert a higher resistance against deformation, and a higher degree of recovery. Low ionic strengths and polyelectrolyte concentrations increase the creep deformation because of screening the surface charge. Higher ionic strengths and polyelectrolyte concentrations lead to fibril aggregation, which stiffens the network structure, decreasing the creep deformation. However, the recovery response is not significantly affected by additives. The critical strain at the onset of non-linear viscoelasticity is independent of mass fraction in two different concentration regimes, with a transition at 0.35% w/w CNF. © 2015 Elsevier Ltd. All rights reserved.
1. Introduction Increasing environmental concerns have led to efforts to increase the use of renewable and biodegradable materials in many industrial applications. Wood pulp is one of the major sources of raw materials for bioproducts. Disintegration of wood fibers into cellulose nanofibrils (CNF) has recently attracted significant interest. Cellulose is a linear polysaccharide with a large number of hydroxyl groups. CNF consists of aligned extended molecules that are adhered together by hydrogen bonds. Due to their high aspect ratio and high degree of crystallinity, cellulose nanofibrils form stiff hydrogels with high elasticity. They have been recommended as rheology modifier for food, paint and cosmetics (Herrick, Casebier, Hamilton, & Sandberg, 1983; Turbak, Snyder, & Sandberg, 1983). CNF films and CNF coated papers have been suggested for packaging applications (Henriksson, Berglund, Isaksson, Lindström, & Nishino, 2008; Spence, Venditti, Rojas, Habibi, & Pawlak, 2010; Syverud & Stenius, 2009). CNF has been also proposed to be used in the production of functional materials such as nanocomposites (Lopez-Rubio et al., 2007; Nakagaito & Yano, 2005) and antimicrobial films (Andresen et al., 2007).
∗ Corresponding author. Tel.: +1 5143986177. E-mail address: [email protected] (T.G.M. van de Ven). http://dx.doi.org/10.1016/j.carbpol.2015.01.067 0144-8617/© 2015 Elsevier Ltd. All rights reserved.
Since 1983, several methods have been developed for the preparation of cellulose nanofibrils. Fibril dimensions and the number of surface charges vary depending on the preparation method. CNF is mechanically produced by disintegration of wood fibers in a homogenizer under high shearing forces. The resulting nanofibrils are 5–20 nm in diameter and several micrometers in length (Herrick et al., 1983; Turbak et al., 1983). To avoid the high energy consumption for mechanical fibrillation of wood fibers, chemical pretreatment of wood pulp has been suggested. Chemical methods are mainly based on inducing strong electrostatic repulsions among the nanofibrils in water, by introducing negatively charged functional groups onto their surface. Nanofibrils can then be disintegrated by gentle mechanical treatments such as sonication. It has been shown that mild enzymatic hydrolysis of softwood pulp decreases the homogenization energy, and produces high aspect ratio nanofibrils with high mechanical strength (Henriksson & Berglund, 2007; Paakko et al., 2007). Paakko et al. (2007) added monocomponent endoglucanase to the softwood pulp suspension and passed it through a homogenizer after refining. The achieved fibril diameter was 5–20 nm, providing strong aqueous gels with large elastic modulus, practical for reinforcing multicomponent mixtures. TEMPO-mediated oxidation with sodium hypochlorite followed by gentle mechanical sonication has been introduced as another preparation method, yielding cellulose nanofibrils with 3–4 nm diameter, several micrometer length, and a large
L. Jowkarderis, T.G.M. van de Ven / Carbohydrate Polymers 123 (2015) 416–423
number of surface charges in the form of COO− groups (Isogai, Saito, & Fukuzumi, 2011; Saito, Nishiyama, Putaux, Vignon, & Isogai, 2006). Water dispersions of TEMPO-mediated oxidized CNF showed pseudo-plastic behavior, and a strongly consistency-dependent shear viscosity at concentrations >0.2% w/w (Lasseuguette, Roux, & Nishiyama, 2008; Saito, Kimura, Nishiyama, & Isogai, 2007). Fibril dimensions and its charge density directly affect the suspension rheology. It has been reported that the elastic modulus of CNF suspensions produced by TEMPO-mediated oxidation of palm tree wood pulp is larger than previously reported by Paakko et al. (2007) for CNF prepared by enzymatic hydrolysis, in spite of the similar fibril dimensions (Benhamou, Dufresne, Magnin, Mortha, & Kaddami, 2014). The authors ascribe it to the higher charge density of their oxidized fibrils compared to the non-oxidized CNF. The rheology of colloidal suspensions can be also affected by interparticle bridging (Abbas Zaman & Delorme, 2002; Otsubo, 1999; Swerin, 1998). This phenomenon is used in industry to control the product viscosity and stiffness. Bridging can be induced by addition of polyelectrolytes or multivalent ions. Oxidized cellulose nanofibrils are negatively charged, therefore cations such as Ca2+ , and polymers such as cationic polyacrylamide can flocculate them. Cationic polyacrylamide is widely used in papermaking as a retention aid (Hubbe, 2007; Vanerek, Alince, & van de Ven, 2000a, 2000b). Retention of fine particles necessitates their deposition on the fibers and fiber flocculation. Fine particles could be mineral fillers which are very small and might be lost during filtration in the papermaking process (van de Ven, 1984), or they could be fiber fines or hydrophobizers used for internal sizing (Hubbe, 2007). Paper machine conditions can significantly decrease the chemical efficiency of these agents if they are not retained in the sheet. Calcium ions are also present in papermaking suspensions, especially when calcium carbonate CaCO3 is used in paper as a filler. This paper aims to study the rheological behavior of TEMPOmediated oxidized CNF hydrogels in the semi-dilute regime. We have studied the rheology of dilute CNF suspensions previously (Jowkarderis & van de Ven, 2014). Moreover, we investigate the effect of interfibrillar bridging on the mechanical strength of the suspensions using cationic polyacrylamide and CaCl2 at various concentrations. Experimental results on the creep–recovery response, and the variations of the elastic modulus as a function of CNF volume fraction are compared to available theoretical models.
417
Polymer bridging was induced using cationic polyacrylamide (CPAM) (PERCOL292), with mass average molecular weight ∼5 ×106 , and degree of substitution ∼20%. A 5 g L−1 polymer solution was prepared by adding dry polymer beads to deionized water and subjecting it to magnetic stirring for 18 h. The solution was then used to make 0.3% w/w CNF suspensions with polymer concentrations in the range 10–200 mg g−1 dry CNF. Since the amount of polymer used in the suspension is very small, maximum ≈600 ppm, its effect on the viscosity of the suspending medium is negligible (less than 1%). CaCl2 , purchased from Sigma–Aldrich, was used for physical cross-linking of cellulose nanofibrils. A 1 M salt solution was added to 0.3% w/w CNF suspensions to make samples with salt concentrations ranging from 1 × 10−5 to 5 × 10−3 M. All samples were tested the second day after preparation. All the measurements were performed at pH ≈6.9. 2.2. Rheometry Rheology measurements were carried out using DHR (TA Instrument) and MCR 302 (Anton-Paar) rheometers. Parallel plates geometry (40 mm diameter) was used to study CNF suspensions with mass fractions m ≥ 0.3 %. Consistent results were obtained for all suspensions using either sand blasted, roughened, or smooth surface plates. The gap was set to 1 mm. Changing the gap from 1 to 2 mm did not affect the results. Suspensions with m < 0.3 % were tested using double-wall concentric cylinders (internal gap: 0.41 mm, external gap: 0.47 mm, effective height: 40 mm). The linear viscoelastic (LVE) region was determined at frequencies 0.5, 6.28 and 50 rad s−1 , over a strain range of 0.1–1000%. Subsequently, frequency sweep tests were conducted at 3% strain, which was within the linear region at all CNF volume fractions. The creep tests were conducted at stress = 0.2 Pa, for 10 min. The stress was then suddenly released and the recovery response was recorded for 10 min. Solvent traps were used to prevent water evaporation in all the tests. The temperature was set to 25 ◦ C. The samples were allowed to relax for 5 min before each measurement. All the tests were performed in triplicate with fresh samples, and their average is reported here. 3. Results and discussion 3.1. Elastic and viscous moduli
2. Experimental 2.1. Materials A 0.67% w/w aqueous suspension of CNF produced by TEMPOmediated oxidation of spruce wood pulp was received from Forest Products Laboratory (FPL) (Madison, WI, USA). The COONa density was reported as 0.65 mmol g−1 dry CNF. The dimensions of the nanofibrils were measured previously as ≈4.7 nm in diameter and ≈550 nm in length, and average aspect ratio r ≈ 110 (Jowkarderis & van de Ven, 2014). Due to the polydispersity in cellulose nanofibrils dimensions, it is not easy to directly determine the volume fraction of CNF suspensions. In this paper, the volume fraction is considered as ≈ m /r , where m is the mass fraction and r ≈ 1.5 is the relative density of cellulose (Hermans, Hermans, & Vermaas, 1945). CNF suspensions with mass fractions m ranging from 0.1% to 0.67% w/w were prepared by addition of deionized water, corresponding to volume fractions in the range 7 × 10−4 to 4.5 × 10−3 , within the semidilute regime. Recall that semi-dilute suspensions are obtained when r−2 < < r−1 (Mewis & Wagner, 2012).
The elastic G and viscous G moduli of TEMPO-mediated oxidized cellulose as a function of frequency are shown in Fig. 1, at various mass fractions. Our results are consistent with the data previously reported by Benhamou et al. (2014). G and G increase with increasing m , indicating the formation of stiffer networks. However, the G and G values are more than one order of magnitude larger than the values obtained by Lasseuguette et al. (2008) at similar mass fractions. The reason can be incomplete disintegration of wood fibers in their system, after mechanical treatments. Non-disintegrated fibers do not form strong network structures, and their dispersions are weaker than suspensions of nanofibrils with equal consistency. The charge density of the oxidized CNF is not mentioned by the authors. At m > 0.3 %, the elastic and viscous moduli are comparable to the observations of Paakko et al. (2007) for enzymatic hydrolyzed cellulose, and of Agoda-Tandjawa et al. (2010) for cellulose microfibrils prepared by strong mechanical treatments, with fibril dimensions larger than the oxidized fibrils used here. Therefore at high mass fractions, the higher surface charge compensates for the effect of smaller fibril dimensions. At m ≤ 0.3 %, the increase of G and G with frequency is noticeable, while Paakko et al. (2007)
418
L. Jowkarderis, T.G.M. van de Ven / Carbohydrate Polymers 123 (2015) 416–423
Fig. 2. G vs. at ω = 6.28 rad s−1 . Fig. 1. G (solid symbols) and G (open symbols) of CNF suspensions as a function of frequency, at various mass fractions. Error bars (smaller than the size of the symbols) show the standard deviation from the average of three measurements.
reported only a slight increase of the elastic modulus with frequency, for m as low as 0.125% w/w, indicating stiffer structures. This shows that at low mass fractions, the effect of fibril dimensions on the mechanical strength of the suspension dominates over the effect of charge content. Fig. 1 also shows that at m < 0.15 %, G < G in the entire range of applied frequency, indicating liquid like behavior. At m = 0.15%, G G , implying that the elastic properties are comparable to the viscous properties. Therefore, m ≈ 0.15 % is the gelation point of TEMPO-mediated oxidized cellulose with a surface charge ≈ 0.65 meq g−1 . In enzymatic hydrolyzed CNF suspensions, at m ≈ 0.12 %, G is about one order of magnitude larger than G (Paakko et al., 2007). Therefore, gelation starts at lower mass fractions of non-oxidized CNF, again highlighting the stronger effect of fibril dimensions over the surface charge at low mass fractions. A significant increase in the elastic modulus upon increasing the concentration has been reported in the literature for cellulosic gels (Iotti, Gregersen, Moe, & Lenes, 2011; Naderi, Lindström, & Pettersson, 2014; Naderi, Lindström, & Sundström, 2014). A power law equation, G ∝ cn , is suggested in most cases, where c is the fibril concentration. The elastic modulus reported by Paakko et al. (2007) scales with concentration with n ≈ 3, and Agoda-Tandjawa et al. (2010) reported n = 2.58, at ω = 1 Hz. Naderi, Lindström, and Sundström (2014) found n = 2.4 for carboxymethylated CNF at mass fractions m ranging from ≈0.3% to 6.2%, at ω=1 Hz. Naderi, Lindström, and Pettersson (2014) later reported that G scales with n ) with a power law exponent of 5.2 at lower mass fraction (G ∝ m m in the range 0.04–0.18%, at ω = 0.5 Hz. Models developed for the elastic modulus of fibrous gels as a function of concentration predict a power-law exponent of around 2 (Jones & Marques, 1990; Kroy & Frey, 1996; Satcher & Dewey, 1996), which is low compared to the values obtained for CNF suspensions. More recently, Hill (2008) has offered a scaling theory for the elastic modulus of CNF gels which yields a transition from a power law exponent of 11/3 to 7 depending on the fibril volume fraction . In a random and entangled microstructure he predicts G ≈ ˇE11/3 (1 + ˛) >
10/3
(1)
for ∼ 2.5a/L, where a is the fibril radius, L is the fibril length, ˇ is an order-one coefficient depending on the fibril characteristics, e.g. its crystallinity, E is the Young’s modulus, and ˛ is an index of twobody interactions depending on the electrolyte concentration and
fibrils surface charge. When is smaller than the volume fraction required for entanglement, which applies to this work, Hill’s theory predicts
G ≈ ˇE
11/3 (1 − )8/3
1+
˛ (1 − )
10/3
.
(2)
0.99). Recall from Fig. 1, G is significantly frequency dependent at low concentrations, therefore, at different frequencies, G relates to by different power law exponents. In this work, by increasing ω from 0.5 to 50 rad s−1 , n decreases from 4.78 to 4.02, indicating that at lower frequencies, the elastic modulus is more sensitive to volume fraction. The effect of frequency is neglected in available models for G as a function of fibrous gels consistency, mainly because the elastic modulus is almost frequency independent at high concentrations. However, the noticeable decrease of the power law exponent with increasing ω in the semi-dilute regime suggests that the frequency dependence of the elastic modulus should be considered in developing future theories. Hill’s model is in agreement with our experimental results at ω = 1 Hz, yielding k1 = ˇE/(1 − )8/3 = 18 GPa and k2 = ˛/(1 − ) = 93 (R2 > 0.99). The degree of crystallinity of TEMPO-mediated oxidized CNF has been reported to be high ≈75%, because oxidation with sodium hypochlorite does not change the original crystallinity of cellulose (Isogai et al., 2011; Rodionova et al., 2013). The Young’s modulus for the CNF with ∼75% crystallinity has been shown to be E ≈ 50 GPa (Eichhorn & Young, 2001), however, the magnitudes of ˇ, ˛ and are unknown. Hill (2008) calculated ˇ ≈ 1.8 and ˛ ≈ 31.5, by fitting Eq. (1) to the experimental data of Paakko et al. (2007) for enzymatic hydrolyzed CNF at ω = 1 Hz. These values may not be valid for our CNF suspensions because of the higher degree of crystallinity and the higher surface charge. The experimental data obtained at frequencies other than 1 Hz fit to Hill’s model with different values of k1 and k2 . We presume that ˇ and ˛ should not depend on frequency, therefore, should be the parameter which causes the variations in k1 and k2 .
L. Jowkarderis, T.G.M. van de Ven / Carbohydrate Polymers 123 (2015) 416–423
Fig. 3. Creep–recovery response of CNF suspensions at various mass fractions, = 0.2 Pa.
3.2. Creep–recovery tests 3.2.1. Creep–recovery response of CNF hydrogels The creep–recovery test is important in industrial applications because it shows how the material responds to an applied load. The measurements involve two steps. First, the creep test: upon the application of a constant stress for a certain time, the material deforms to a maximum strain ( max ). Second, the recovery test: the stress is released, and the material starts reforming its initial structure. In principle, as long as the applied stress is not higher than the yield stress, an elastic solid recovers all the strain. A viscoelastic material, however, cannot recover its initial structure, and some part of the deformation permanently remains in the sample. The unrecovered strain (v ) represents the viscous part of the viscoelasticity, and the recovered deformation (e = max − v ) represents the elastic part. The recovery percent is defined as Rec%=100 × e / max (Mezger, 2006). Fig. 3 shows the creep–recovery results for CNF suspensions with various concentrations at stress = 0.2 Pa. The yield stress for the suspensions shown in this figure varies from y ≈ 0.8 Pa at m = 0.3%, to y ≈ 10 Pa at m = 0.67%. As expected, deformation ( max ) decreases with increasing mass fraction, indicating an increase in stiffness. However, the strain is not completely recovered after releasing the stress, showing that CNF hydrogels have some viscous character in the semi-dilute regime, even when their response is predominantly elastic (Mezger, 2006). 3.2.2. Effect of polyelectrolyte The effect of polyelectrolyte on the mechanical strength of 0.3% CNF suspensions at = 0.2 Pa is shown in Fig. 4. The maximum deformation max increases with increasing the amount of C-PAM up to 50 mg polymer g−1 CNF, indicating a decrease in suspension stiffness. Adding more polymer, however, increases the sample stiffness, as max decreases. The transition happens at a polymer concentration between 50 and 100 mg polymer g−1 CNF, corresponding to about 0.1 to 0.2 mg polymer m−2 fibril surface. The reason is that C-PAM affects the suspension rheology in two ways. One is screening the surface charge, which causes a decrease in mechanical strength and viscosity (Jowkarderis & van de Ven, 2014). The other way is bridging between neighboring fibrils, which strengthens the network structure and leads to a higher resistance against deformation. At low polyelectrolyte concentrations, the effective surface charge controls the rheological properties, therefore the suspension stiffness decreases by adding C-PAM, and a large deformation is observed. At high concentrations of C-PAM, on
419
Fig. 4. Creep–recovery response of 0.3% CNF suspensions at various concentrations of C-PAM in units of mg g−1 CNF, = 0.2 Pa.
the other hand, interparticle bridging compensates for the decrease of electrostatic forces, and stiffens the material structure, therefore the deformation decreases. Variations of the elastic modulus G of 0.3% CNF suspensions as a function of C-PAM concentration also show lower stiffness at low, and higher stiffness at high C-PAM concentrations (data available, not shown). 3.2.3. Effect of CaCl2 Fig. 5 shows the effect of calcium ions on the creep–recovery response of 0.3% CNF suspensions at = 0.2 Pa. Similar to the effect of polyelectrolyte, network deformation increases at small CaCl2 concentrations, [Ca2+ ] 0.99 was obtained in all cases. Increasing the CNF mass fraction and the amount of CaCl2 (>0.2 mM) and >
C-PAM (∼100 mg g−1 ) increase the Burger model constants, indicating an increase in the suspension stiffness and rigidity (Laguna et al., 2013). Table 1 also shows the permanent strain ∞ calculated by Eq. (5) for all the suspensions, compared to the experimentally measured value v .
L. Jowkarderis, T.G.M. van de Ven / Carbohydrate Polymers 123 (2015) 416–423
421
Fig. 6. (a) G (solid lines) and G (dashed lines) versus strain for two CNF mass fractions, ω = 1 Hz. Solid arrow shows the maximum deformation in creep–recovery test, and dashed arrow shows the critical strain. (b) c versus CNF mass fraction, ω = 1 Hz.
3.3. Critical strain A linear viscoelastic (LVE) region, over which the elastic and viscous moduli are independent of the oscillation amplitude, was observed for all the suspensions. Fig. 6a shows the strain sweep results for two CNF mass fractions. The end of the LVE region where G or G begins to change upon increasing the strain is called the critical strain c , indicating the network breakage point. Exceeding the critical strain causes irreversible deformation in the network structure (Mezger, 2006). In this study, c is evaluated as the strain at which G has decreased to 90% of its linear value. It has been reported that in suspensions of softwood kraft pulp fibers, there is no variations in c by varying the mass fraction in the range 3–8% (Swerin, Powell, & Ödberg, 1992). Swerin (1998) also reported almost equal values of c for suspensions of softwood and hardwood pulp fibers with similar fiber diameters. He explains that at the critical strain, the applied shear disconnects the fiber–fiber contact points, introducing gaps between the connecting fibers, and breaks the network structure. The gap required to break a contact point might be a function of the fiber diameter, therefore, suspensions with different fiber types, but similar dimensions, exhibit similar critical strains (Swerin, 1998). Benhamou et al. (2014) on the other hand, have reported a decrease in c when the mass fraction of TEMPO-mediated oxidized CNF increases from 0.1% to 1%. Fig. 6b shows c for CNF suspensions with mass fractions m ≥ 0.2 % which exhibit G > G . There are two m regions in which
c is almost independent of mass fraction. The critical strain in suspensions with m ≥ 0.4 % is smaller than at m ≤ 0.3 %. High density network structure starts to break at smaller oscillation amplitudes. At lower concentrations, however, fibril networks do not exist, or they exist with low densities, therefore the elastic modulus starts decreasing at larger amplitudes (Amaria, Uesugi, & Suzukib, 1997). The transition happens at m ≈ 0.35 %. As previously shown in Fig. 1, the frequency dependence of G and G becomes less pronounced >
at m ∼0.35%. This mass fraction, therefore, can be identified with the onset of high density network formation. Comparing the strain sweep results to the creep–recovery responses shows that irreversible network deformation happens at strains < c . For example, the observed c at m = 0.5% is ≈10.1%, as shown with the dashed arrow in Fig. 6a; however the suspension does not recover more than ≈70%, after deforming to max ≈ 2.3 % (
