Effective Young's Modulus of Bacterial and Microfibrillated Cellulose

Mar 16, 2012 - The Northwest Composite Centre, University of Manchester, Paper Science Building, Sackville Street, Manchester, M13 9PL,. United Kingdo...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/Biomac

Effective Young’s Modulus of Bacterial and Microfibrillated Cellulose Fibrils in Fibrous Networks Supachok Tanpichai,†,‡ Franck Quero,†,‡ Masaya Nogi,§ Hiroyuki Yano,∥ Robert J. Young,†,‡ Tom Lindström,⊥ William W. Sampson,†,‡ and Stephen J. Eichhorn*,†,‡,# †

Materials Science Centre, School of Materials, University of Manchester, Grosvenor Street, Manchester, M13 9PL, United Kingdom The Northwest Composite Centre, University of Manchester, Paper Science Building, Sackville Street, Manchester, M13 9PL, United Kingdom § The Institute of Scientific and Industrial Research, Osaka University, Mihogaoka 8-1, Ibaraki, Osaka, 567-0047, Japan ∥ Research Institute for the Sustainable Humanosphere, Kyoto University, Uji, Kyoto 611-011, Japan ⊥ Innventia AB, Box 5604, 114 86 Stockholm, Sweden ‡

S Supporting Information *

ABSTRACT: The deformation micromechanics of bacterial cellulose (BC) and microfibrillated cellulose (MFC) networks have been investigated using Raman spectroscopy. The Raman spectra of both BC and MFC networks exhibit a band initially located at ∼1095 cm−1. We have used the intensity of this band as a function of rotation angle of the specimens to study the cellulose fibril orientation in BC and MFC networks. We have also used the change in this peak’s wavenumber position with applied tensile deformation to probe the stress-transfer behavior of these cellulosic materials. The intensity of this Raman band did not change significantly with rotation angle, indicating an inplane 2D network of fibrils with uniform random orientation; conversely, a highly oriented flax fiber exhibited a marked change in intensity with rotation angle. Experimental data and theoretical analysis shows that the Raman band shift rate arising from deformation of networks under tension is dependent on the angles between the axis of fibrils, the strain axis, the incident laser polarization direction, and the back scattered polarization configurations. From this analysis, the effective moduli of single fibrils of BC and MFC in the networks were estimated to be in the ranges of 79−88 and 29−36 GPa, respectively. It is shown also that for the model to fit the data it is necessary to use a negative Poisson’s ratio for MFC networks and BC networks. Discussion of this in-plane “auxetic” behavior is given.



INTRODUCTION Nanofibers, from sources such as bacterial cellulose (BC) and microfibrillated cellulose (MFC), have recently attracted interest due to their superior properties compared to micrometer-sized fibers, such as high strength, high aspect ratio, high specific surface area, low thermal expansion, and low density.1−6 It is often cited, with little justification, that these fibers have high moduli, close to the crystal modulus of cellulose (∼138 GPa).7 It is, therefore, imperative that reliable measurements of the moduli of cellulose nanofibers are made so that their properties can be fully realized. Such measurements are the subject of this study. To date, two methods have been used to estimate Young’s modulus of a single BC nanofibril. An atomic force microscopy tip bending method8 yielded a value of 78 ± 17 GPa. A value of 114 GPa was found using a Raman spectroscopy technique,4 which is closer to a generally accepted value of 138 GPa for the crystal modulus of cellulose I, found using an X-ray diffraction method;7,9 nevertheless, this is still only an estimate. Values between 50 and 143 GPa for Young’s modulus of tunicate and © 2012 American Chemical Society

plant derived cellulose nanowhiskers have also been reported, again using a Raman spectroscopic method.10,11 Single fibrils of microfibrillated cellulose (MFC), which form a dense network, are also often reported to have mechanical properties similar to cellulose crystals.12 However, the literature provides little direct evidence for this. MFC was pioneered by Herrick et al.13 and Turbak et al.14 To prepare MFC, a wood pulp suspension is typically passed repeatedly through a mechanical homogenizer to increase the degree of fibrillation; inevitably, the energy consumption increases with the number of passes and also damage to the fibrils.3 Young’s modulus of MFC networks has been reported to be ∼15 GPa, which is far from the value one might expect from a network of fibrils with the crystal modulus of cellulose I (138 GPa).7,9,15,16,17,18,21,22 If the fibrils themselves are thought to have a modulus close to the crystal modulus then a simple calculation, using the efficiency factor of 3/8 proposed by Received: January 12, 2012 Revised: March 11, 2012 Published: March 16, 2012 1340

dx.doi.org/10.1021/bm300042t | Biomacromolecules 2012, 13, 1340−1349

Biomacromolecules

Article

Krenchel,23 yields a value of ∼50 GPa for a network of fibers. Aulin et al. obtained values of 30 GPa for MFC sheets and used a model to predict the amorphous modulus of cellulose, which was close to other published values.24 It is clear that this value is much lower than that predicted for the idealized case of a random 2D network of stiff cellulose fibers. Reports of direct measurement of the modulus of single fibrils of MFC are scarce. One measurement, on fibrils generated from lyocell fibers (cellulose II, not cellulose I) reports a value of 98.6 GPa for the modulus obtained from AFM cantilever tip bending measurements.25 It should be pointed out that this value is close to the crystal modulus of cellulose II (∼90 GPa).26 No similar measurement on cellulose I fibrils has been reported to the authors’ knowledge. In this paper we report a measurement of the molecular deformation of cellulose fibrils in BC and MFC networks and derive the effective Young’s modulus of individual fibrils. We also show that the Poisson’s ratio of these nanofibril networks is negative. Such “auxetic” behavior of fibrous networks has been documented in the literature,27,28 but not for in-plane deformation of cellulose-based nanofibrous networks. Auxetic behavior in the out-of-plane direction for paper sheets has been reported previously.29,30 Raman spectroscopy is a useful technique to study the molecular deformation of many polymeric fibers, including cellulosic materials. This technique relies on the measurement of a change in the position of a Raman band on the application of external tensile deformation.31,32 The technique was first used to observe a change in the vibrational frequencies of Raman bands obtained from a monocrystalline polydiacetylene fiber subjected to tensile deformation.33 Subsequently, this method has been applied to monitor the molecular deformation in numerous natural and regenerated cellulose fibers,34,35 paper36 and wood.37 Recently, the technique has been applied to monitor the stress-transfer process in cellulose nanofiberbased composites.38,39 The present work shows how a better estimate of fibril modulus can be obtained taking into account off axis deformation.



(Microfluidizer M-110EH, Microfluidics Crop.), equipped with two differently sized Z-shaped chambers (each chamber connected in series). This slurry passed through the first chamber with a diameter of 200 μm and then the second chamber with a diameter of 400 μm, 3×. This was followed by a 5× pass through a chamber pair with diameters of 200 and 100 μm. After this process, MFC films were prepared as follows. A 1 g L−1 MFC suspension was prepared by diluting the homogenized MFC (2.5%), after which the dispersion was mixed using a propeller and an ultrasound bath (Branson 5510, Branson Ultrasonics, U.S.A.). The suspension was then filtered on a Durapore membrane (0.65 μm Durapore DVPP, Millipore, U.S.A.). The filtered sample was then dried under restraint using drying frames at 50 °C in an oven. These networks are referred as MFC-A. MFC-B networks were manufactured using the same commercial sulfite softwood dissolving pulp as above. MFC-B was manufactured using a similar process to MFC-A, but using a carboxymethylation pretreatment (degree of substitution = 0.1), previously described in details by Wågberg et al.20 The treated pulp was filtered and washed. After this pretreatment, the fibrils were passed through a high-pressure fluidizer (Microfluidizer M-110EH, Microfluidics Crop.) equipped with a chamber pair with a diameters of 200 and 100 μm. Finally, films were prepared as described above. Steam Exploded Flax Fiber. The steam exploded flax fibers were provided by FH Reutlingen, Germany. The fibers were extracted from bundles and bleached in hydrogen peroxide for 48 h to reduce fluorescence from the laser used by the Raman spectrometer. Field Emission Gun Scanning Electron Microscopy (FEGSEM). BC and MFC networks were sputter-coated with gold at 40 mA for 4 min, and fixed on metal stubs using carbon tabs. The network surfaces were examined using a scanning electron microscope (Philips XL-30 FEG-SEM) with an acceleration voltage of 4 kV. Polarized Raman Spectroscopy and Orientation of the Cellulose. The molecular orientation of individual flax fibers and BC and MFC networks was characterized using a Raman spectrometer (Renishaw system-1000, Wotton-under-Edge, U.K.) using a 785 nm near-infrared laser. An Olympus BH-1 microscope with a 50× objective lens was used to focus the laser beam on the surface of the samples to a spot size of approximately 1−2 μm. The laser power at the sample surface was 1 mW. Raman spectra were obtained in the range of 1050 to 1150 cm−1. A polarization configuration was used where both the incident laser beam and scattered light were polarized

EXPERIMENTAL SECTION

Bacterial Cellulose. Gluconacetobacter xylinum (No. 13693; National Institute of Technology and Evaluation, Tokyo, Japan) and Hestrin-Schramm (HS) medium were used to produce BC networks. The cells for the inoculum were cultured in test tubes statically at 27 °C for 2 weeks. The thick gel produced during culturing was then squeezed aseptically to remove the embedded cells. The cell suspension (25 mL) was then transferred as an inoculum for the main culture (500 mL of medium), which was incubated statically at 27 °C for 18 days to produce BC networks. BC networks (35 mm in diameter) were purified by boiling with 2% NaOH for 2 h and then by washing with distilled water, followed by hot pressing at 2 MPa and 120 °C for 4 min to completely remove the bulk water. Microfibrillated Cellulose Preparation. A commercial bleached sulfite softwood cellulose pulp (Domsjö ECO Bright, Domsjö Fabriker AB, Sweden) consisting of 40% pine (Pinus sylvestris) and 60% spruce (Picea abies) with a hemicellulose content of 14% and a lignin content of 1% was used as a source for MFC. These data were obtained from the pulp supplier. Before use, both pulps were thoroughly washed with deionized water. MFC networks were prepared using the following process, which has been described elsewhere.19 A 4 wt % cellulose suspension was first passed through a refiner (Angle Refiner R1L, Escher-Wyss). This was followed by an enzymatic treatment with monocomponent endoglucanase at 50 °C for 2 h. The treated pulp suspension was then washed with deionized water, and was refined again. This suspension was passed through a high-pressure fluidizer

Figure 1. Schematic of the polarization configurations. parallel to the principal axis of the spectrometer (Figure 1). This polarization configuration is denoted “VV”. Flax fibers as well as BC and MFC strips were mounted onto 20 mm gauge length paper testing cards and fixed using a two-part cold-curing epoxy resin (Araldite, Huntsman, U.K.). The samples were rotated from 0 to 90° using a rotation stage. A Raman spectrum was recorded at each 5° increment using an exposure time of 30 s and four accumulations for BC networks and flax fibers and 10 s and five accumulations for MFC networks; the different exposure times were required to achieve the same signal-to-noise ratios from the samples. The intensities of the Raman band initially located at ∼1095 cm−1 were determined using a least-squares fit of a mixed Gaussian/Lorentzian function. Intensities were normalized by dividing the data by the intensity at 0° for BC and MFC networks. For flax fibers the intensity was normalized by dividing the data by the maximum intensity, which was obtained at approximately 5°. Experiments were performed in triplicate for each 1341

dx.doi.org/10.1021/bm300042t | Biomacromolecules 2012, 13, 1340−1349

Biomacromolecules

Article

sample and means and standard deviations for each data set were determined. Polarized Raman Spectroscopy and Deformation. The relationship between the stress transfer within the networks and the polarization angle of the laser was investigated using the same Raman spectrometer described in the previous section. BC and MFC networks were deformed using a customized deformation rig (Deben Microtest, Deben, Bury St Edmunds, U.K.), incorporating a 2 kN load cell. The samples were prepared following the procedure described in the previous section on specimen preparation. A half-wave plate and a polarizer/analyzer were inserted to rotate the polarization direction for the incident and scattered light with respect to the axis of the samples, coincident with the deformation axis. Two polarization arrangements were selected: one where both the polarization of the incident laser beam and the analyzer are parallel to the principal axis of the spectrometer (denoted “VV”) and another where the polarization of the scattered light is rotated by 90° using the analyzer (“VH”) (see Figure 1). During tensile deformation, samples were rotated from 0 to 90° using 10° increments for BC, MFC-A and MFC-B networks at each 0.1% strain increment from 0 to 1%. Raman spectra were collected at each rotation angle increment using an exposure time of 30 s and four accumulations for BC networks, and 10 s and five accumulations for MFC networks. The peak positions of the Raman band initially located at ∼1095 cm−1 were determined in the same way as for Raman band intensities in the previous section. Experiments were performed in triplicate for each sample and means and standard deviations for each data set were determined.



RESULTS AND DISCUSSION

Figure 2 shows FEG-SEM images of the surface of BC, MFC-A, and MFC-B networks at the same magnification. We observe that BC networks are made of smaller fibrils than MFC networks. The latter have larger fibrils and fiber aggregates because the number of passes through the high pressure fluidizer was insufficient to convert all fibers into nanofibrils. In addition these images show that the networks consist of a planar structure of fibrils. BC and MFC networks have been previously shown to comprise randomly oriented nanofibrils4 and microfibrils.25 The physical properties and morphology of BC and MFC networks and flax fibers are reported and compared in Table 1. The diameter of BC fibrils measured from FEG-SEM images was found to be in the range of 50−120 nm. The density of networks of these fibrils was found to be 1.10 ± 0.04 g cm−3. As the culture time of a BC network increases, its thickness and density are found to increase.38 On the other hand, the densities of MFC-A and MFC-B networks, having thicknesses of 86 and 77 μm, respectively, were found to be 1.31 ± 0.03 and 1.32 ± 0.05 g cm−3, respectively, which is similar to that of MFC networks reported previously.22,40 Figure 3 reports typical Raman spectra for BC, MFC-A, and MFC-B networks using the “VV” polarization configuration. The Raman band initially located at ∼1095 cm−1 is highlighted. This band is thought to correspond to C−O stretching modes within the backbone structure of cellulose41 and possibly the glycosidic C−O−C stretch modes.37 As noted earlier, when the networks were deformed, the Raman band initially located at ∼1095 cm−1 shifted toward a lower wavenumber position. The positions of this Raman band as a function of strain, are plotted and fitted using a linear function; the fit gradient is called the Raman band shift rate, d(ΔνR)/dε, and its magnitude is used to

Figure 2. Field emission gun scanning electron microscope (FEGSEM) images of (a) BC, (b) MFC-A, and (c) MFC-B networks at the same magnification.

quantify the level of stress-transfer. The fibril modulus (Efibril) has been found to be proportional to this shift rate, that is34 E fibril ∝

d(Δν R ) dε

(1)

where ΔνR is the Raman band position and ε is the tensile strain. The Raman band shift rate of fibers in composites has also been found to be dependent on the angle between the laser polarization direction and both the fiber orientation and deformation axis.42,43 Consequently, we report the influence of 1342

dx.doi.org/10.1021/bm300042t | Biomacromolecules 2012, 13, 1340−1349

Biomacromolecules

Article

these factors on the Raman band shift rate obtained for cellulose nanofiber networks.

Figure 3. Typical Raman spectra obtained with the “VV” polarization configuration for BC, MFC-A, and MFC-B networks highlighting the band located at ∼1095 cm−1.

Figure 5. Typical Raman spectra obtained from (a) flax fibers and (b) BC, MFC-A, and MFC-B networks in the region of the Raman band initially located at ∼1095 cm−1 using the “VV” and “VH” polarization configurations.

Figure 4. Schematic of a single fibril at angles between the fibril direction and strain axis (θ), the strain direction and the laser polarization axis (ϕ), and the fibril direction and the laser axis (α) in a 2D plane of a cellulose network.

Figure 4 shows the geometrical relationship between all angles: namely, θ, the angle between the strain direction and the fibril axis; ϕ, the angle between the strain direction and the laser polarization; and α, the angle between the fibril orientation and the laser polarization. BC, MFC-A, and MFC-B networks were deformed at different angles, ϕ, in both the “VV” and “VH” polarization configurations. Figure 5a and b show typical Raman spectra obtained from flax fibers and BC and MFC networks, respectively, using both the “VV” and the “VH” polarization configurations. It can be seen that each polarization configuration significantly influences the intensity of the Raman band initially located at ∼1095 cm−1. This arises due to the directions and the magnitudes of the Raman tensor components of different molecular vibrations coincident with the laser polarization.44,45 Figure 6 shows the change in the normalized intensity of the Raman band initially located at ∼1095 cm−1 as a function of the angle, ϕ, between the axis of samples and the laser polarization direction. It can be seen that the intensity of this Raman band is at a maximum when the cellulose molecules within the fibrils of the flax fiber are parallel to the laser polarization direction. For flax fibers, the maximum intensity value is obtained at a specific ϕ angle, which is often called the microfibril angle. The intensity

Figure 6. Normalized intensity of the Raman band initially located at ∼1095 cm−1 as a function of the angle ϕ between the polarization direction of the laser and flax fiber and BC, MFC-A, and MFC-B network samples. Black lines represent fits to the data.

is minimized when the alignment of cellulose molecules within the microfibrils of the flax fiber are perpendicular to the laser polarization direction. In other words, as the angle ϕ increases 1343

dx.doi.org/10.1021/bm300042t | Biomacromolecules 2012, 13, 1340−1349

Biomacromolecules

Article

directly related to the molecular deformation of the cellulose backbone,31,32 and have been used to monitor the molecular deformation of cellulose-based fibers and composites.31,32,34−39 Figures 8 and 9 show the variation of the wavenumber position of the Raman band initially located at ∼1095 cm−1 as a

from 0° to 90° relative to the polarization direction of the laser, the intensity of the Raman band initially located at ∼1095 cm−1 decreases gradually. Table S1 (Supporting Information) reports the intensity ratio of 0° to 90°; for BC and MFC networks; this is approximately constant with unit value whereas for the flax fibers the ratio is 0.1. This is due to the high degree of alignment of cellulose chains along the flax fiber’s axis.10,36 The data for a flax fiber were fitted using the equation41 I = a + b cos4(ϕ + θ )

(2)

where a and b are constants and are related to the derivatives of the polarizability tensors as a function of normal coordinates. An optimized fitting of the experimental data was found for a = 0.11, b = 0.89, and θ = 5.7°, which is of the same order of magnitude as a reported value of the microfibril angle of a flax fiber,42 that is, 9.6 ± 2.5°, and therefore equal to the offset between the cellulose molecules and the polarization direction. Figure 6 also includes the typical normalized intensities of BC and MFC networks as a function of the rotation angle ϕ; these were insensitive to orientation angle, confirming a uniform orientation of fibrils within the networks.4 Figure 7a,b shows typical Raman band shifts for BC and MFC-A networks upon tensile deformation. Similar shifts of the Raman band initially located at ∼1095 cm−1 can also be observed for MFC-B networks. These shifts are thought to be

Figure 8. Shifts in the position of the Raman band initially located at ∼1095 cm−1 as a function of tensile deformation for BC networks at different angles ϕ (0, 30, 60, and 90°) to the laser axis with (a) “VV” and (b) “VH” polarization configurations. Black lines represent fits to the data.

function of tensile strain for BC and MFC-A networks with the “VV” and “VH” polarization configurations. Data are reported for rotation angles of 0, 30, 60, and 90°. From Figure 9a, some deviation from the linear fit to the data at ∼0.3% strain is noted, which may be due to a yielding of the sheet at this point. The data appear to “dip” below the line, which suggests perhaps that fiber−fiber bonds are disrupted in some way. It is clear that the Raman band shift rate for the “VV” polarization configuration is dependent on the rotation angle. In the “VH” polarization configuration, the Raman band shift rate for carbon nanotube networks has been found to be independent of the angle between the sample axis and the laser polarization direction.46 In contrast, our data suggest a weak dependence on this angle since significant differences are observed in the shift rate for BC networks from 0 and 90° at the higher strains investigated. This might be due to some BC fibril reorientation occurring when increasing the tensile deformation. This mechanism is supported by data reported in Figure 10a where a change

Figure 7. Typical shifts in the position of the Raman band initially located at ∼1095 cm−1 for (a) BC and (b) MFC-A networks subjected to tensile deformation. 1344

dx.doi.org/10.1021/bm300042t | Biomacromolecules 2012, 13, 1340−1349

Biomacromolecules

Article

Figure 9. Shifts in the position of the Raman band initially located at ∼1095 cm−1 as a function of tensile deformation for MFC-A networks at different angles ϕ (0, 30, 60, and 90°) to the laser axis with (a) “VV” and (b) “VH” polarization configurations. Black lines represent fits to the data.

Figure 10. Normalized intensity of the Raman band located at ∼1095 cm−1 as a function of the angle ϕ between the polarization direction of the laser and the tensile deformation direction. Percentages of 0 and 1.5% indicate the level of tensile deformation of (a) BC and (b) MFCA networks. Black arrows indicate the tensile deformation direction.

from an isotropic to a slightly anisotropic intensity distribution is noted; however, no orientation change can be observed for MFC networks, as shown in Figure 10b. MFC-B networks also exhibit similar behavior. This may be because MFC networks are densely packed, with a high degree of bonding, giving little rotational freedom to fibrils. Therefore, the shift rate was found to be constant from 0 to 90° for MFC-A and MFC-B network. We return to this finding in our subsequent analysis of plotting of the band shift rate as a function of the rotation angle in the “VH” polarization configuration for BC, MFC-A, and MFC-B networks. In single short-fiber composites the relationship between the Raman shift rate and the fiber orientation θ (ϕ = 0°), is given by43,44,46,47 s(θ ) = S0(cos2 θ − ν sin 2 θ)

Bernier, and Perez for a model describing the strains of fibers within networks under load.48 For two-dimensional (2D) and uniformly dispersed materials, the Raman intensity for the different polarization configurations (“VV” and “VH”) can be considered as an intensity weighted average of the contribution of single fibrils oriented at different directions in the network. The Raman intensities at an angle, ϕ, for the “VV” and “VH” polarization configurations can be determined using44,46 1 π

∫0

IVH =

1 π

∫0

(3)

where S(θ) is the Raman band shift rate at an angle θ, S0 is the band shift rate for a fiber aligned parallel to the strain direction, and ν is Poisson’s ratio of the matrix of the composite. We assume that the same relationship holds for networks of fibers or fibrils. Indeed, a similar relationship is derived by Kallmes,

π

IVV =

π

I0 cos4(θ + ϕ)dθ =

3 I0 8

I ′0 cos2(θ + ϕ) sin 2(θ + ϕ)dθ =

(4)

I ′0 8 (5)

where I0 and I′0 are the maximum intensities for the “VV” and “VH” polarization configurations, respectively. The Raman shift 1345

dx.doi.org/10.1021/bm300042t | Biomacromolecules 2012, 13, 1340−1349

Biomacromolecules

Article

band rates for the “VV” and “VH” polarization configurations are given by the following equations44 ⎡1 S VV (ϕ) = ⎢ ⎣π

π

I0(cos4(θ + ϕ)S0 ⎤ (cos2(θ) − ν sin 2(θ ))dθ ⎥/IVH ⎦

∫0

S S = 0 (1 − ν) + 0 (1 + ν) cos(2ϕ) 2 3

⎡1 S VH(ϕ) = ⎢ ⎣π

∫0

π

(6)

I ′0 (cos2(θ + ϕ)sin 2(θ + ϕ)S′0

⎤ (cos2(θ) − ν sin 2(θ ))dθ ⎥/IVH ⎦

=

S′0 (1 − ν) 2

(7)

Figure 11 shows the theoretical dependence of the Raman band shift rate with the “VV” polarization configuration on

Figure 11. Dependence of the band shift rates on orientation angle with the “VV” polarization configuration generated by using eq 6 with the values of Poisson’s ratio between 0.3 and −0.3.

orientation angle, as given by eq 6; note that eq 7 reveals no such dependence in the “VH” polarization configuration. A range of Poisson’s ratio (from −0.3 to +0.3) is used to demonstrate the dependence of the band shift on this parameter. For isotropic materials Poisson’s ratio is known to be bound from −1 to +0.5.49 The value of Poisson’s ratio of cellulose is known to be dependent on the source; for example ν = 0.38 for single ramie fibers and ν = 0.46 for flax fibers.50,51 For individual nanofibrils, Poisson’s ratio values have been reported51 to be between −0.26 and −1.17; although this auxetic behavior is only found to occur in the crystals forming the nanofibrils. It is found that values of −0.1, −0.2, and −0.1 for Poisson’s ratio provide the best fit to the Raman band shift rate experimental data of BC, MFC-A, and MFC-B networks, respectively, as shown in Figure 12a,b. The negative Poisson’s ratio implies that the networks expand in the transverse direction when they are stretched, demonstrating auxetic behavior.52 The mechanisms that give rise to this are unclear, but one possibility could be the presence of re-entrant networks of fibers or transverse strains

Figure 12. Shift rates of the Raman band initially located at ∼1095 cm−1 with the “VV” polarization configuration for (a) BC and (b) MFC-A and MFC-B networks as a function of rotation angle ϕ, and (c) with the “VH” polarization configuration for BC, MFC-A, and MFC-B networks. Black lines represent fits to the data and dotted lines indicate the upper and lower 95% confidence bands to the data.

arising in elements perpendicular to the deformation axis due to high levels of bonding within polygonal constructs of fibers. The angle dependence of the Raman band shift rate for BC networks was analyzed using the “VV” polarization configuration. The Raman band shift rate at 0° for BC networks was found to be −1.1 ± 0.1 cm−1 %−1. However, it is worth noting 1346

dx.doi.org/10.1021/bm300042t | Biomacromolecules 2012, 13, 1340−1349

Biomacromolecules

Article

nanotubes.44 The Raman band shift rate is, however, not constant for BC networks, probably due to a fibril reorientation effect, as shown earlier. The theory assumes a constant shift rate but does not take into account reorientation effects. As fibers reorient toward the tensile axis, as they are seen to do for BC networks (Figure 10a), then there is less requirement for a model that takes into account off-axis fiber orientation. The model could be adapted to take this reorientation into account, with a variable change in angle with deformation. The use of samples with highly oriented fibrils however, as have been reported recently, could be an alternative way of obtaining fibril stiffness’ without the need to adapt the model to a dynamic change in orientation.55 Raman band shift rates (S0) of BC, MFC-A, and MFC-B networks estimated by fitting the Raman shift data obtained at various angles ϕ using the “VV” polarization configuration with Mathematica were found to be −1.3, −0.5, and −0.5 cm−1 %−1, respectively. BC, MFC-A, and MFC-B networks, on the other hand, yielded shift rates S0 of −1.4, −0.6, and −0.6 cm−1 %−1, respectively, using the “VH” polarization configuration. Values of S0 for each BC and MFC network using VV and VH polarization configurations are found to be similar. Young’s modulus of BC and MFC networks are calculated using the equation

that the Raman band shift rate as a function of tensile strain for BC networks is found to be dependent on the thickness of the network.38 The Raman band shift rate with respect to strain obtained from BC networks cultured for 3 days, −1.4 ± 0.1 cm−1 %−1 has been reported to be higher than BC networks cultured for 6 days, −0.9 ± 0.1 cm−1 %−1.38 These experiments were performed at a strain angle of 0°, with the “VN” polarization configuration (the scattered light was not polarized38). From this study it was shown that the layered networks produced with longer culturing times have lower stress-transfer efficiency; the stress-transfer efficiency being reduced through the weak van der Waals bonding between multiple layers resulting in delamination.38 Consequently, to better estimate Young’s modulus of BC fibrils, it is better to use more planar networks of fibers, reducing this “layer effect”. The choice of BC networks having a low culturing time seems to be preferential for this purpose. Problems can however arise if BC networks are too thin, whereby the intensity of the Raman band initially located at ∼1095 cm−1 is too low to accurately assess its position with deformation, and subsequently the estimation of the band shift rate. For MFC-A and MFC-B networks, although these networks were prepared by different pretreatment methods (enzymatic and carboxymethylation pretreatments), little difference is observed for the Raman band shift rate with respect to the angle dependence using the “VV” polarization configuration. This may be due to the similar size of nanofibers in these networks.53 The effect of these two pretreatment methods on stress-transfer remains a topic for future work. It is clear that the magnitude of the Raman band shift rate for BC networks is greater than that for MFC networks. This is due to the fact that BC fibrils form finer networks compared to MFC, which increase fibril/fibril interactions.38 It is known that these fibril−fibril interactions are dominated by hydrogen bonds, which are formed during the wet pressing of the fibrous networks. BC networks differ in one very significant way to MFC networks in that they consist of fully jointed fibril bifurcations, formed when bacteria divide during processing. These bifurcations in BC networks are assumed to be stiffer than fibril/fibril interactions in MFC. Consequently, an increased amount of hydrogen bonding, and the presence of bifurcations, would yield a larger Raman band shift rate. Nakagaito et al.54 have studied the morphology of BC and MFC fibrils for preparing composite materials. They found BC pellicles have a network structure of extremely fine and interconnected, continuous, and dimensionally uniform ribbonlike elements. Young’s modulus of BC composites with epoxy resin was reported to be 28 GPa, which was higher than that of MFC-based composites (19 GPa54). One reason for this could be their method of preparation involving high levels of mechanical disintegration; BC fibrils on the other hand are produced using relatively benign processing, while harsher conditions might damage and shorten the MFC cellulose fibrils, reducing their reinforcement efficiency. It is important to add that, after homogenization, the degree of polymerization of MFC has been reported to be reduced by 27%, which has also been shown to compromise the mechanical properties of the fibrils.12 Figure 12c reports the angle dependence of the Raman band shift rate for BC, MFC-A, and MFC-B networks using the “VH” polarization configuration. For MFC-A and MFC-B networks, the Raman band shift rate is constant over all angles, as observed for isotropic PVA films reinforced with carbon

Enetwork ×

d(Δν R ) = S0 dσ

(8)

where σ is the fiber or network stress. A value of d(ΔνR)/dσ has been reported to be −4.3 cm−1 GPa−1 and has been previously used to determine the modulus of cellulose materials.4,10,11 This calibration value was obtained from the micromechanical deformations of a number of natural cellulose fibers and regenerated cellulose fibers using Raman spectroscopy31 and is assumed to hold for both BC and MFC nanofibrils. To calculate the effective modulus of single fibrils of BC and MFC, Krenchel’s analysis23 can be used (see Supporting Information and eq S1). The equation from this analysis has been used to estimate Young’s modulus of single fragments of microcrystalline cellulose,32 cellulose nanowhiskers,10,11 and bacterial cellulose.4 The efficiency factor for a two-dimensional in-plane random network of fibers is found to be 3/8.23 This result and eq 8 can be used to determine the effective modulus of a single filament using the Raman band shift rates. The effective modulus of a single filament of BC and MFC and Poisson’s ratios for each material obtained from the best fitting of experimental data are reported in Table S2 (see Supporting Information). It is worth noting that the values of the effective modulus of BC and MFC fibrils determined from the Raman shift rates with the “VV” polarization configuration are similar to that obtained with the “VH” polarization configuration. This suggests good consistency in the values of the effective modulus estimated from the Raman band shift rates. The Young’s moduli of the BC and MFC fibrils were found to have a range of values; respectively 79−88 and 29−36 GPa. The effective moduli of single fibrils within BC, MFC-A, and MFC-B networks are much lower than the generally accepted experimental value for the crystal modulus of cellulose I, 138 GPa.7 This could be because Raman spectroscopy records molecular deformation of both crystalline and amorphous regions. It is possible therefore that some averaging of the stiffness is recorded, resulting in a reduced value. The value of 1347

dx.doi.org/10.1021/bm300042t | Biomacromolecules 2012, 13, 1340−1349

Biomacromolecules

Article

78 ± 17 GPa reported by Guhados et al.8 is similar to BC networks obtained in both “VV” and “VH” polarization configurations, 79 ± 3 and 88 ± 10 GPa, respectively. Young’s modulus of a single filament of MFC is much lower than that of BC; this difference is most likely due to the lower degree of crystallinity. Indeed, Aulin et al.24 showed that when the crystallinity was accounted for the moduli should be around 30 GPa for MFC films prepared in a similar manner to this report. This is in reasonable agreement with values of the upper theoretical limit determined in this communication.

(4) Hsieh, Y. C.; Yano, H.; Nogi, M.; Eichhorn, S. J. Cellulose 2008, 15, 507−513. (5) Lu, J.; Wang, T.; Drzal, L. T. Composites, Part A 2008, 39, 738− 746. (6) Eichhorn, S. J.; Dufresne, A.; Aranguren, M.; Marcovich, N. E.; Capadona, J. R.; Rowan, S. J.; Weder, C.; Thielemans, W.; Roman, M.; Renneckar, S.; Gindl, W.; Veigel, S.; Keckes, J.; Yano, H.; Abe, K.; Nogi, M.; Nakagaito, A. N.; Mangalam, A.; Simonsen, J.; Benight, A. S.; Bismarck, A.; Berglund, L. A.; Peijs, T. J. Mater. Sci. 2010, 45, 1−33. (7) Sakurada, I.; Nukushina, Y.; Ito, T. J. Polym. Sci. 1962, 57, 651− 660. (8) Guhados, G.; Wan, W. K.; Hutter, J. L. Langmuir 2005, 21, 6642−6646. (9) Nishino, T.; Takano, K.; Nakamae, K. J. Polym. Sci., Part B: Polym. Phys. 1995, 33, 1647−1651. (10) Sturcova, A.; Davies, G. R.; Eichhorn, S. J. Biomacromolecules 2005, 6, 1055−1061. (11) Rusli, R.; Eichhorn, S. J. Appl. Phys. Lett. 2008, 93, 033111. (12) Henriksson, M.; Henriksson, G.; Berglund, L. A.; Lindstrom, T. Eur. Polym. J. 2007, 43, 3434−3441. (13) Herrick, F. W.; Casebier, R. L.; Hamilton, J. K.; Sandberg, K. R. J. Appl. Polym. Sci. 1983, 37, 797−813. (14) Turbak, A. F.; Snyder, F. W.; Sandberg, K. R. J. Appl. Polym. Sci. 1983, 37, 815−827. (15) Nakagaito, A. N.; Yano, H. Appl. Phys. A: Mater. Sci. Process. 2004, 78, 547−552. (16) Siro, I.; Plackett, D.; Hedenqvist, M.; Ankerfors, M.; Lindstrom, T. J. Appl. Polym. Sci. 2010, 119, 2652−2660. (17) Iwamoto, S.; Nakagaito, A. N.; Yano, H. Appl. Phys. A: Mater. Sci. Process. 2007, 89, 461−466. (18) Klemm, D.; Kramer, F.; Moritz, S.; Lindstrom, T.; Ankerfors, M.; Gray, D.; Dorris, A. Angew. Chem., Int. Ed. 2011, 50, 5438−5466. (19) Paakko, M.; Ankerfors, M.; Kosonen, H.; Nykanen, A.; Ahola, S.; Osterberg, M.; Ruokolainen, J.; Laine, J.; Larsson, P. T.; Ikkala, O.; Lindstrom, T. Biomacromolecules 2007, 8, 1934−1941. (20) Wågberg, L.; Decher, G.; Norgren, M.; Lindstrom, T.; Ankerfors, M.; Axnas, K. Langmuir 2008, 24, 784−795. (21) Saito, T.; Nishiyama, Y.; Putaux, J. L.; Vignon, M.; Isogai, A. Biomacromolecules 2006, 7, 1687−1691. (22) Abe, K.; Yano, H. Cellulose 2009, 16, 1017−1023. (23) Krenchel, H. Fibre Reinforcement; Akademisk Forlag: Copenhagen, 1964. (24) Aulin, C.; Gallstedt, M.; Lindstrom, T. Cellulose 2010, 17, 559− 574. (25) Cheng, Q. Z.; Wang, S. Q.; Harper, D. P. Composites, Part A 2009, 40, 583−588. (26) Eichhorn, S. J.; Young, R. J.; Davies, G. R. Biomacromolecules 2005, 6, 507−513. (27) Kuznetsov, A. A.; Fonseca, A. F.; Baughman, R. H.; Zakhidov, A. A. ACS Nano 2011, 5, 985−993. (28) Chen, L. Z.; Liu, C. H.; Wang, J. P.; Zhang, W.; Hu, C. H.; Fan, S. S. Appl. Phys. Lett. 2009, 94, 253111. (29) Stenberg, N.; Fellers, C. Nord. Pulp Paper Res. J. 2002, 17, 387− 394. (30) Baumgarten, H. L.; G€ottsching, L. Triaxial Deformation of Paper under Tensile Load, Fundamental Properties of Paper Related to Its Uses; Cambridge, England, 1973; The British Paper and Board Industry Federation: Plough Place, Fetter Lane, London EC4A 1AL, England, 1973; pp 227−253. (31) Eichhorn, S. J.; Young, R. J. Compos. Sci. Technol. 2003, 63, 1225−1230. (32) Eichhorn, S. J.; Young, R. J. Cellulose 2001, 8, 197−207. (33) Mitra, V. K.; Risen, W. M.; Baughman, R. H. J. Chem. Phys. 1977, 66, 2731−2736. (34) Eichhorn, S. J.; Sirichaisit, J.; Young, R. J. J. Mater. Sci. 2001, 36, 3129−3135. (35) Kong, K.; Eichhorn, S. J. Polymer 2005, 46, 6380−6390. (36) Bakri, B.; Eichhorn, S. J. Cellulose 2010, 17, 1−11.



CONCLUSIONS Raman spectroscopy has been successfully applied to study the micromechanical deformation of fibrous networks of BC and MFC. Shifts in the position of the Raman band initially located at ∼1095 cm−1, corresponding to C−O bond stretching along the cellulose molecular chains, have been used to estimate Young’s modulus of single filaments of BC and MFC. This is achieved by taking into account contributions from both onaxis and off-axis fibrils by analyzing the shifts in the position of the Raman band located at ∼1095 cm−1 as a function of rotation angle of the specimen with respect to the polarization direction of the laser. The effective moduli of single fibrils of BC and MFC were reported to be in the ranges of 79−88 and 29−36 GPa, respectively. These ranges of values are lower than the Young’s modulus reported for the crystal modulus (138 GPa) or of cellulose nanowhiskers (143 GPa). In addition to this, Poisson’s ratio for the networks has been found to be negative. A negative Poisson’s ratio is thought to arise due to re-entrant structures forming in thick and higher density fibrous networks, although this would require further research to confirm this.



ASSOCIATED CONTENT

S Supporting Information *

Tables S1 and S2 and the equation for the efficiency factor determination using Krenchel’s analysis. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel.: +44 (0)1392 72 5515. Fax: +44 (0)1392 21 7965. Present Address #

Physics, College of Engineering, Mathematics and Physical Sciences, Stocker Road, Exeter, U.K., EX4 4QL. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS One of the authors (S.T.) would like to thank the Royal Thai Government for funding a Ph.D. studentship. Two authors (F.Q. and S.J.E.) would like to thank the EPSRC for funding a Ph.D. studentship (to F.Q.) under Grant GR/F028946.



REFERENCES

(1) Eyholzer, C.; Bordeanu, N.; Lopez-Suevos, F.; Rentsch, D.; Zimmermann, T.; Oksman, K. Cellulose 2010, 17, 19−30. (2) Stelte, W.; Sanadi, A. R. Ind. Eng. Chem. Res. 2009, 48, 11211− 11219. (3) Siro, I.; Plackett, D. Cellulose 2010, 17, 459−494. 1348

dx.doi.org/10.1021/bm300042t | Biomacromolecules 2012, 13, 1340−1349

Biomacromolecules

Article

(37) Gierlinger, N.; Schwanninger, M.; Reinecke, A.; Burgert, I. Biomacromolecules 2006, 7, 2077−2081. (38) Quero, F.; Nogi, M.; Yano, H.; Abdulsalami, K.; Holmes, S. M.; Sakakini, B. H.; Eichhorn, S. J. ACS Appl. Mater. Interfaces 2010, 2, 321−330. (39) Rusli, R.; Shanmuganathan, K.; Rowan, S. J.; Weder, C.; Eichhorn, S. J. Biomacromolecules 2010, 11 (3), 762−768. (40) Henriksson, M.; Berglund, L. A.; Isaksson, P.; Lindstrom, T.; Nishino, T. Biomacromolecules 2008, 9 (6), 1579−1585. (41) Wiley, J. H.; Atalla, R. H. Carbohydr. Res. 1987, 160, 113−129. (42) Charlet, K.; Eve, S.; Jernot, J. P.; Gomina, M.; Breard, J. Mesomechanics 2009, 1, 233−236. (43) Andrews, M. C.; Day, R. J.; Hu, X.; Young, R. J. J. Mater. Sci. Lett. 1992, 11, 1344−1346. (44) Deng, L.; Eichhorn, S. J.; Kao, C.-C.; Young, R. J. ACS Appl. Mater. Interfaces 2011, 3, 433−440. (45) Tanaka, M.; Young, R. J. J. Mater. Sci. 2006, 41, 963−991. (46) Cooper, C. A.; Young, R. J.; Halsall, M. Composites, Part A 2001, 32, 401−411. (47) Wood, J. R.; Zhao, Q.; Wagner, H. D. Composites, Part A 2001, 32, 391−399. (48) Kallmes, O.; Bernier, G.; Perez, M. Paper Technol. Ind. 1977, 18, 222−228; 18, 243−225; 18, 328−331. (49) Greaves, G. N.; Greer, A. L.; Lakes, R. S.; Rouxel, T. Nat. Mater. 2011, 10, 823−837. (50) Nishiyama, Y. J. Wood Sci. 2009, 55 (4), 241−249. (51) Peura, M.; Grotkopp, I.; Lemke, H.; Vikkula, A.; Laine, J.; Muller, M.; Serimaa, R. Biomacromolecules 2006, 7, 1521−1528. (52) Evans, K. E. Endeavour 1991, 15, 170−174. (53) Plackett, D.; Anturi, H.; Hedenqvist, M.; Ankerfors, M.; Gallstedt, M.; Lindstrom, T.; Siro, I. J. Appl. Polym. Sci. 2010, 117, 3601−3609. (54) Nakagaito, A. N.; Yano, H. Appl. Phys. A: Mater. Sci. Process. 2005, 80, 155−159. (55) Sehaqui, H.; Ezekiel Mushi, N.; Morimune, S.; Salajkova, M.; Nishino, T.; Berglund, L. A. ACS Appl. Mater. Interfaces 2012, 4, 1043−1049.

1349

dx.doi.org/10.1021/bm300042t | Biomacromolecules 2012, 13, 1340−1349