Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Understanding nanocellulose chirality and structure–properties relationship at the single fibril level

Understanding nanocellulose chirality and structure–properties relationship at the single fibril... ARTICLE Received 25 Nov 2014 | Accepted 19 May 2015 | Published 25 Jun 2015 DOI: 10.1038/ncomms8564 OPEN Understanding nanocellulose chirality and structure–properties relationship at the single fibril level 1 1 1 1 2,3 2 Ivan Usov , Gustav Nystro¨m , Jozef Adamcik , Stephan Handschin , Christina Schu¨tz , Andreas Fall , 2 1 Lennart Bergstro¨m & Raffaele Mezzenga Nanocellulose fibrils are ubiquitous in nature and nanotechnologies but their mesoscopic structural assembly is not yet fully understood. Here we study the structural features of rod-like cellulose nanoparticles on a single particle level, by applying statistical polymer physics concepts on electron and atomic force microscopy images, and we assess their physical properties via quantitative nanomechanical mapping. We show evidence of right-handed chirality, observed on both bundles and on single fibrils. Statistical analysis of contours from microscopy images shows a non-Gaussian kink angle distribution. This is inconsistent with a structure consisting of alternating amorphous and crystalline domains along the contour and supports process-induced kink formation. The intrinsic mechanical properties of nanocellulose are extracted from nanoindentation and persistence length method for transversal and longitudinal directions, respectively. The structural analysis is pushed to the level of single cellulose polymer chains, and their smallest associated unit with a proposed 2  2 chain-packing arrangement. 1 2 Department of Health Science and Technology, ETH Zurich, Schmelzbergstrasse 9, LFO E23, Zurich 8092, Switzerland. Department of Materials and Environmental Chemistry, Stockholm University, Svante Arrhenius vag 16C, Stockholm 10691, Sweden. Wallenberg Wood Science Center, KTH, Teknikringen 56, Stockholm 10044, Sweden. Correspondence and requests for materials should be addressed to R.M. (email: [email protected]). NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 1 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 n nature, cellulose, the major load-bearing structure in all of substructures, down to the single cellulose chain. From the living plants, shows complex superstructures, chirality and statistical analysis, the persistence length of the W-CNF is 1,2 Ichiral inversions over different length scales . This extracted, and indirectly, its intrinsic rigidity in the longitudinal fascinating structural behaviour, in combination with a high direction, while the direct measurements of the Young’s specific strength and stiffness, has inspired work to engineer moduli obtained using PF-QNM characterize mechanical proper- cellulose materials with tailored mechanical and optical ties of the nanocellulose samples in the transversal plane. 3,4 properties . Over the last decades, elongated rod-like cellulose Most importantly, the detailed statistical investigation of the nanoparticles, shorter cellulose nanocrystals (CNCs) and longer morphology of the W-CNF provides convincing evidence that cellulose nanofibrils (CNF), have been used to form chiral the commonly accepted model of CNF structure as built of 5 6 7 nematic liquid crystals , aerogels , photonic and inorganic alternating regions of crystalline and amorphous cellulose 8 9,30,31 hybrid materials . Even though these materials have impressive domains along the fibril length cannot explain the properties and show great potential for a broad range of presence of kinks. applications, the fine structure of their nanocellulose components has not yet been fully elucidated . Only through a detailed knowledge of the chirality and the structure of the Results smallest nanocellulose building blocks, new strategies for Overview of the nanocellulose systems. The low-magnification advanced bottom-up nanotechnologies and assembly of new images of the three different nanocellulose systems (W-CNF, helical metamaterials may become available. W-CNC and B-CNC) via three different microscopy techniques On the atomic length scale, high-resolution neutron and X-ray (AFM, Cryo-SEM and TEM) are shown in Fig. 1. On the basis of scattering data have revealed the crystalline structure of cellulose the initial visual inspection we provide a description of their down to the exact atomic positions of the polymers in the unit morphology and typical structural parameters. In the W-CNF cell . How CNC arrange in chiral nematic liquid crystalline (Fig. 1a–c), CNFs have a relatively constant cross-section with 11–13 14 15 phases and how CNF arrange in films , aerogels or average widths in the range of B2  3 nm and lengths foams is also well characterized. In between these two length B0.1  1mm. One particular feature of this sample is that almost scales, however, a gap remains, where the structure and assembly every nanofibril has sharp bends between otherwise straight of nanocellulose fibrils are not yet fully understood. Given the segments, a feature hereinafter referred to as kinks. Moreover, hierarchical nature of cellulose, an investigation of this these nanofibrils can sometimes still be observed to be assembled intermediate length scales may reveal important structural in fibril bundles, despite the strong mechanical treatment applied information, improve the fundamental understanding and pave during the homogenization process for sample preparation. The the way to new strategies in materials assembly. W-CNC sample (Fig. 1d–f) is represented by rather straight Traditional nanocellulose research relies on the disintegration particles that have a large variation in cross-section along their of a cellulosic raw material into its smallest possible components contours. They can be described as large multistranded bundles of (CNC and CNF) and the subsequent re-assembly of these laterally assembled CNFs. Their average widths are in the range of building blocks into new materials. The harsh chemical and B3  30 nm, but the lengths are similar to the W-CNF, that is, mechanical treatments used in the disintegration process may B0.1  1mm. In rare occasions it is possible to observe thin have a large influence on the final structural properties of the segments with kinks, a signature of the W-CNF precursor cellulose nanoparticles . This is often overlooked and only a few samples, indicating that the hydrolysis was not fully completed 18,19 detailed experimental investigations have been performed . (Supplementary Fig. 1). For the B-CNC (Fig. 1g–i), the observed Previous experimental microscopy and scattering work nanofibrils have typical lengths of B1–5mm and irregular widths have allowed resolving average particle dimensions , indirect in the range of B10  50 nm. The features of nanocellulose (from neutron scattering) and direct evidence (from fibril systems are consistent within all distinct microscopy techniques 22,23 bundles) of particle twist, as well as local mechanical used in this study, which confirms that any possible artefact properties with peak force quantitative nanomechanical mapping associated with either the microscopy technique or sample pre- 24,25 (PF-QNM) . Simulations have confirmed a twisted structure at paration procedure is below the interpretable resolution. 26,27 equilibrium , and also estimated single particle mechanical During the preparation of the nanocellulose samples, ionizable properties similar to the measured values . However, issues related groups are introduced (-COOH for W-CNF and W-CNC, and to the quantification and origin of the CNF conformations remain -SO H for B-CNC) that electrostatically improve the stability of elusive and the fundamental question of the (in)solubility of the particles in dispersion. The charge density of W-CNC and cellulose in water has recently become a matter of intense debate . B-CNC was measured with polyelectrolyte titration, yielding With the rapid development of experimental techniques, new and gravimetrically normalized values of 0.4 mmol g (W-CNC) more detailed structural information becomes available, allowing and 0.03 mmol g (B-CNC). Because the charge density is not these outstanding fundamental questions to be revisited and expected to be affected by the hydrolysis treatment , we assign a conclusively assessed. charge density of 0.4 mmol g also for the W-CNF sample. In this work, state-of-the-art atomic force, cryogenic scanning On the basis of the microscopy images, we find that there is a electron and transmission electron microscopy (AFM, Cryo-SEM larger fraction of individual particles of W-CNF compared with and TEM, respectively) are combined with advanced statistical W-CNC. The better colloidal stability of W-CNF may be related analysis and concepts from polymer physics to investigate three to a steric hindrance effect due to the kinks preventing different types of nanocellulose, TEMPO(2,2,6,6-tetramethylpi- neighbouring fibrils to come into close contact and form peridine-1-oxyl)-mediated oxidized wood cellulose nanofibrils bundles. The stronger tendency for B-CNC compared with (W-CNF), wood cellulose nanocrystals (W-CNC) and sulfuric W-CNC to form bundles is probably related to the relatively low acid hydrolyzed bacterial cellulose nanocrystals (B-CNC). All charge density of B-CNC. It should also be noted that the pH of samples show clear evidence of a right-handed chirality both at the W-CNC and B-CNC samples is drastically decreased during the level of bundles of fibrils and on the individual fibril level hydrolysis, causing a reduction in the charge density of the CNC (W-CNF). Deeper investigation of one of the nanocellulose particles, resulting in the formation of bundles and aggregates. It families (W-CNF) reveals detailed information on particle is possible that not all of these aggregates become fully dimensions and provides compelling evidence for multiple levels redispersed when the pH is readjusted. 2 NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 ARTICLE All types of nanocellulose possess a right-handed chirality. The helical cellulose thickenings . In contrast to these observations of magnified microscopy images reveal that all cellulose fibrils and higher-order structures, the results presented here establish that crystals with observable chirality are right-handed (Fig. 2). the single CNFs with observable chirality always bear a right- For the W-CNF (Fig. 2a–c), right-handed chirality is observed handed twisting. Because also the observed chirality of the fibril both on the single fibril level (Fig. 2a, left and Supplementary bundles is right-handed, this indicates that there is no chiral Fig. 2) and for bundles of fibrils (Fig. 2a, right). We note that a inversion on this length scale. twist in amplitude measurements is observable only if the local Recent studies based on density functional theory of right- microfibril cross-section has distinct corners—for cylindrical handed hard helices indicate that the way chirality transfers from fibrils this would be invisible. For the CNC particles, chirality the building block to the chiral nematic phase is non-trivial and observations were only possible on crystal bundles, since isolated depends at least on two different parameters, entropy (excluded CNCs with constant and uniform cross-section could not be volume of left-/right-handed particle pairs) and thermodynamics observed. (volume fraction of rods) . This suggests that a chirality The presence of the right-handed twist along cellulose inversion from a right-handed particle to a left-handed chiral nanoparticles and bundles of nanoparticles is not routinely nematic phase may be expected, provided that the concentration observable (5–10% from hundreds of visualized particles); of rods in the chiral nematic is in the right range . Thus, the however, when a twist is observable, it is always right-handed. results at the single fibril and small bundle level given here may This relatively low amount of twisted particles in W-CNF could provide the necessary information to bridge the gap between be explained by a finite AFM resolution, which is approaching the theoretical models and experimental results in nanocellulose limit to detect these features. For W-CNC and B-CNC, the cholesteric nematic phases. twisting normally occurs in bundles with high structural organization of single particles, but could be hidden in the case of loosely packed aggregates. PF-QNM of the nanocellulose samples. The mechanical 22,23 These observations are in line with previous experimental properties of cellulose in the transversal direction were probed 26,27 36 and theoretical studies, supporting a right-handed chirality. using PF-QNM . Previously, this method has been successfully 37,38 However, the experimental results presented to date, rely on used for systems of amyloid fibrils that have similar higher-order structures of fibril aggregates that are in mm-length dimensions to cellulose fibrils. Figure 3a shows an AFM image scale. This distinction is important since there are different with the composition of all three types of nanocellulose particles: examples in nature where chirality inversion occurs over different W-CNF, W-CNC and B-CNC, mixed in 1:1:1 proportion, with length scales and cellulose may follow the same trends. For the corresponding PF-QNM map depicted in Fig. 3b. The example, the observed cholesteric liquid crystal ordering of CNCs Young’s modulus values, E , were estimated according to the QNM is left-handed and the tracheary elements, responsible for Derjaguin–Mueller–Toporov (DMT) model, and were collected transporting water in most plants, are reinforced by left-handed along the contours of each cellulose type to obtain the a d –5 nm 5 nm –10 nm 10 nm –40 nm 40 nm 1,000 nm 1,000 nm 1,000 nm b e h 250 nm 500 nm 1,000 nm c f i 150 nm 300 nm 600 nm Figure 1 | Overview microscopy images of the nanocellulose samples. (a–c) TEMPO-mediated oxidized W-CNF, (d–f) W-CNC and (g–i) B-CNC samples via AFM (a,d,g) Cryo-SEM (b,e,h) and TEM (c,f,i). All AFM images have the same magnification in order to provide a direct comparison between nanocellulose particles. NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 3 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 41 42 corresponding modulus distributions (Fig. 3c–e). We find that the polysaccharide polymers or carbon nanotubes . The Young’s moduli lay in the range of 20–50 GPa for all three types coordinates of 2,380 fibrils were acquired from AFM images of nanocellulose particles, and are in good agreement to that covers 450-mm area of mica substrate, using a specially previously measured moduli from PF-QNM on nanocellulose designed in-house software written in MATLAB (see Methods) . (6–50 GPa; refs 24,25). The Gaussian fits of these distributions Objects that were considered to be bundles of fibrils with with parameters m and s (mean value and s.d.) result in distinctively larger diameters were, whenever possible, discarded ± ± the following elastic moduli (m s): 34.4 5.3 GPa for W-CNF from the analysis. The positions of the kinks were determined ± ± (Fig. 3c), 31.1 5.9 GPa for W-CNC (Fig. 3d) and 32.3 4.1 GPa manually using special masks (Fig. 4b, green boxes). Two for B-CNC (Fig. 3e). Moreover, we note that the PF-QNM segments of a contour that are crossing a mask element measurements yield remarkably close values of Young’s moduli (entering and exiting segments) define a contour angle for all three samples investigated. Since the AFM sample (Fig. 4c), which is also referred as a kink angle. One important preparation conditions allow for a qualitative side-by-side note to mention is that it is difficult to distinguish between very comparison of the samples, these results indicate that the small kink angles (below 20) and random thermal fluctuation in mechanical properties of the samples do not depend on the the AFM images. For example, Fig. 4d depicts a fibril with two nanocellulose-processing conditions. obvious kinks that are unambiguously determined and one particular place with a hint of a kink only. The threshold for observable kinks does affect the results, but in a predictive Tracking and statistical analysis of W-CNF. The TEMPO- manner. mediated oxidized W-CNF system is of particular interest for the It is well known that CNF consists of cellulose chains with application of statistical analysis since in comparison with the different degrees of order, from highly crystalline arrangements to other nanocellulose systems, it has rather uniform contour a slightly perturbed distribution of the chains. Two models have shapes. The statistical analysis of the nanofibril contours provides been proposed to describe the distribution of the disordered and unique insight on the nature of their kinks, and moreover, ordered regions in CNF . In the first model, the less-ordered it allows an estimation of the Young’s modulus value of rigid chains, often referred to as amorphous chains, are distributed between crystalline regions of chains along the fibril direction segments via the persistence length approach . Furthermore, the basic morphological parameters such as length and height (analogous to semicrystalline polymers) . In the second model, distributions allow elucidating the packing model of single the less-ordered chains are located towards the surface of fibrils cellulose polymer chains within the nanocellulose fibril. with a crystalline core of chains . The former model is We have used a tracking procedure to obtain coordinates of the commonly invoked as a rationale for CNC particle preparation, CNF contours (Fig. 4a, blue lines) similar to the one used in as hydrolysis is believed to primarily dissolve the amorphous 33,40 previous lines of work on amyloid fibrils , carrageenan regions of cellulose chains, leaving only the short, crystalline a d –15° 15° –30° 30° 50 nm 100 nm 150 nm 500 nm b e h 75 nm 150 nm 500 nm c f i 75 nm 150 nm 500 nm Figure 2 | All samples show right-handed twisting. (a–c) TEMPO-mediated oxidized W-CNF, (d–f) W-CNC and (g–i) B-CNC. The images are acquired from the AFM amplitude channel (a, left) and phase channel (a, right and d,g), Cryo-SEM (b,e,h) and TEM (c,f,i). The arrows point to regions along fibril or crystal contours, where it is possible to detect a right-handed twisting. The corresponding AFM height maps for the images a,d,g are shown in Supplementary Fig. 3. The colour bar in g also applies to a. 4 NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 ARTICLE 10 nm 100 GPa a b W-CNF –10 nm 0 GPa W-CNC B-CNC 1,500 nm c e 20 20 15 15 15 10 10 5 5 5 0 0 0 10 20 30 40 50 10 20 30 40 50 10 20 30 40 50 DMT modulus (GPa) DMT modulus (GPa) DMT modulus (GPa) g 80 100 GPa Kink1 Kink Kink2 Kink Kink2 0 GPa Kink Kink Kink1 0 0 100 200 300 400 500 100 nm Contour length (nm) Figure 3 | Peak force quantitative nanomechanical mapping. (a) AFM height channel visualizing the composition of all nanocellulose samples mixed in 1:1:1 proportion. The nanocellulose particles are distinguished and labelled according to their morphological features. (b) The corresponding PF-QNM map. The scale bar applies to both a,b panels. (c–e) DMT modulus distributions measured on particles representing (c) W-CNF, (d) W-CNC and (e) B-CNC. (f) PF-QNM images of W-CNFs and their profiles represented by red, blue and black lines. (g) Mechanical properties along the profiles of fibrils shown in f and the corresponding kink positions. particles. However, if the regions between the crystalline parts are angle distribution. In such an eventuality, by maintaining the randomly packed arrangements of moderately oriented chains— assumption of kink angles originating from amorphous domains, amorphous regions to a good approximation—we would expect a the final kink angle distribution would then be the weighted Gaussian distribution of the kink angles because of equal sum of individual (symmetric) Gaussian distributions reflecting probability to bend in both directions, and, thus, leading to a different angle probabilities for different fibril thicknesses, Gaussian distribution of the excess, that is, the observable kink resulting into a non-Gaussian, yet fully symmetric, final angle, as in general random walk statistics. We note that a distribution of kink angles. It is therefore the non-Gaussian and Gaussian distribution of bending angles is also expected on the non-symmetric nature of the distribution given in Fig. 4f, which basis of the worm-like chain formalism. It is clear that the allows ruling out amorphous domains as a cause of the presence observed peak in the kinks distribution at 60 does not support of kinks, independently of the cross-section distribution of the the first model, and the alternative model with less-ordered nanocellulose fibrils considered in the statistics. surface chains may possibly be more appropriate to describe the Additional support to the conclusions drawn from the kink cellulose polymer-chain arrangement in CNF. angle distribution analysis, ruling out the presence of ‘softer’ In order to estimate the influence of the contact with the amorphous zone regions alternating to stiffer crystalline ones, is substrate surface on the distribution of kink angles, the analogical given by the lack of significant changes in Young’s moduli analysis was conducted for images obtained using Cryo-SEM and measured using PF-QNM in correspondence of the kink regions, AFM on graphite substrate (Supplementary Fig. 4). The number which would be expected if these regions were amorphous (see of traced fibrils was 100 and 120 for Cryo-SEM and AFM on Fig. 3f,g). The crystalline core-amorphous shell model is also in 46–48 graphite correspondingly; hence, the statistics is 20 times less line with previous NMR studies , as well as the recent detailed accurate than the one reported in the main text, yet, the same experimental study that combined small-angle neutron scattering, conclusion is supported in both cases. wide-angle X-ray scattering, NMR and Fourier transform infrared We further note that the distribution of kink angles given in spectroscopy . Our results are therefore suggesting that the Fig. 4f accounts for kinks observed in fibrils of polydisperse observed kinks may result from the mechanical treatment during cross-sections. This may imply, in principle, that each individual the sample preparation and not from the presence of fibril can contribute to the final statistic with its own kink amorphous regions on the fibril contour. NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 5 & 2015 Macmillan Publishers Limited. All rights reserved. DMT modulus (GPa) % ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 a f 0 30 60 90 120 150 Kink angle,  (deg) 250 nm Kink area b c Angle of deviation (mask) (kink angle) Tracked contour 0 300 600 900 1,200 Contour length, L (nm) e Gaussian h × 10 distribution Obvious kinks Kink? Hypothetical amorphous 0 20° region 0123456 Height, h (nm) Figure 4 | Tracking procedure of W-CNF and results of statistical analysis on fibril contours. (a) AFM image with tracked contours (blue curves). Objects that are considered to be bundles of CNF were discarded from tracking (examples are pointed by white arrows). (b) Magnified region of the AFM image from a containing one tracked contour. The green frames are special masks, placed manually, allowing tracking of the contours with heterogeneous flexibility. In the statistical analysis these areas were considered to represent the vicinities of kinks. (c) Magnified region of the AFM image from b showing the particular mask area with the contour represented by its points. Two segments of the contour that enter or exit the kink area define an angle of contour deviation (kink angle) a.(d) Example of CNF with two obvious kinks and one uncertain case. It is non-trivial to state whether this is a kink or elastic bending due to thermal fluctuations. An example of the 20 angle is shown as a guide to eye. (e) The kink angle distribution should have a Gaussian shape in case of hypothetical amorphous regions corresponding to kinks because of equal probability to bend in both directions. (f) Kink angle distribution of 2,380 tracked CNFs. The bins in the region below 20 (violet line corresponds to this cutoff value) can lack some counts because of the manual threshold of kink angle assignments. (g) Contour length distribution fitted with normalized density function of the log-normal distribution with parameters m ¼ 5.94 and s ¼ 0.77. (h) Height distribution of all points along all tracked contours. The most probable height value is 2.35 nm. Many size measures in nature tend to have a log-normal fibrils. The broad height distribution of the W-CNF peak can be distribution, for example, the lengths of inert appendages (hair, attributed to the possibility of fibrils to split and thus have various 49 50 claws and nails) in biology , or the lengths of amyloid fibrils ; sizes and packing models. Some representative examples of here we show that the length of CNFs follows the same type of fibril-splitting are shown in Fig. 5. distribution. The log-normal distribution has the following probability density function f(L; m, s): Persistence length and second moment of area of the W-CNF. ðÞ lnðÞ L  m The persistence length l is an essential characteristic of polymers, pffiffiffiffiffi 2s fLðÞ ; m; s¼ e ð1Þ or in general fibrillar-like objects, and is directly related to the Ls 2p mechanical properties on a longitudinal inflection. It is formally where L is the total contour length of the fibrils, m and s are the defined as the length over which an angular correlation in the mean value and the s.d. of the length’s natural logarithm, tangent direction to a fibril contour is decreased on average by e respectively, and A is a distribution normalizing constant. The times in three-dimensional space . Following this definition, the length distribution of W-CNF is shown in Fig. 4g with a log- bond correlation function (BCF) is the most common way to normal distribution fit. Parameters of the fit are as follows: evaluate the persistence length and in two-dimensional space it 52  l/2l ± ± m ¼ 5.94 0.08 and s ¼ 0.77 0.06, which corresponds to the has the following form : hcosyi¼ e , where y is the angle average length hLi¼ 511 nm via the well-known relation between tangent directions of any two segments along a fibril m þ s =2 hi L ¼ e . contour. Another useful method of the persistence length The height distribution of W-CNF is shown in Fig. 4h. The estimation is to evaluate the mean-squared end-to-end distance height of the rigid segments does not differ from the height in the (MSED) between contour segments of a fibril, which for a vicinities of kinks (Supplementary Fig. 5) and is estimated to be in worm-like chain model has a following theoretical dependence the range h ¼ (1.9–2.7) nm, with the most probable height value on the internal contour length l in two-dimensional space : 2  l/2l hhi¼ 2.35 nm, which corresponds to the average diameter of hR i¼ 4l[l–2l(1–e )], where R is the direct distance between 6 NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. Number of fibers, n Number of kinks, n f k Number of points, n p NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 ARTICLE a b c 150 nm –5 nm 5 nm d e f 120 nm Figure 5 | Examples of splitting of W-CNF. The splitting event might occur at the end of W-CNF (a,c,d,f), as well as along the contour (b,e). The arrows point towards regions of interest. The images are obtained by (a–c) AFM and (d–f) Cryo-SEM. The scale bar in a and the colour bar in c apply to all AFM images (a–c). The scale bar in d applies to all Cryo-SEM images (d–f). any pair of segments along a fibril contour. A different method I (ref. 56); therefore, we neglect all other possible sources of errors that can be successfully applied to very stiff fibrillar-like objects, in this calculation. Because of the underestimation of the for which at internal contour length is smaller than the persistence length as a result of lost kink counts at low angles, corresponding persistence length (lol), is the mean-squared this value is also underestimated. Furthermore, nanocellulose, as midpoint displacement (MSMD). The equation, describing many other types of natural fibrils, is known to possess anisotropy the behaviour of an arc midpoint deviation, is derived with in the continuum mechanical property tensor and therefore, the the assumption that this deviation is small in comparison with the elastic behaviour in the axial and radial directions are expected corresponding internal contour length (|u |ool) and thus has the to differ. Thus, while in other natural fibrous systems, small 54 2 3 2 form u ¼ l =48l, where u is the midpoint mean-squared differences between the nanoindentation measurements and the x x displacement between any pair of segments along a fibril contour. results of statistical analysis of fibril bending may arise from Because the kinks of the W-CNF may originate from the slightly different elastic moduli on longitudinal bending E mechanical treatment, the deformation in these areas did go and on transversal bending E (refs 33,39), in the case of QNM beyond the elastic limit and, thus, cannot represent elastic nanocellulose, this deviation may become more significant . properties of the material on axial bending. We thus divide the W-CNFs using mask elements as split points and discard areas inside them to avoid kinks. Hence, in the resulting data, Observation of single cellulose polymer chains. In rare individual contours correspond only to rigid segments. The value occasions it is possible to observe very thin, possibly single of the persistence length obtained by the end-to-end distance cellulose polymer chains (Fig. 7a). This observation is in agree- versus internal contour length is l ¼ 2.84mm (Fig. 6). ment with previous observations of molecularly thin sections MSED 57,58 Supplementary Fig. 6 shows the analogue estimations of along CNF contours . In contrast to those findings, however, persistence length of W-CNF rigid segments via BCF and our data show fully flexible particles indicating that the observed MSMD. The resulting values are close: l ¼ 2.54mm and objects are single polymer chains rather than single crystalline BCF l ¼ 2.49mm. However, the relative errors and fitting quality layers of cellulose chains, as suggested in (ref. 58). A longitudinal MSMD provided by the end-to-end distance method is better than by the section is shown in Fig. 7b, while the height distribution of other two approaches (Supplementary Fig. 6b,e). We use only the 20 tracked fibrils is presented in Fig. 7c (average height data of the rigid segments in the Young’s modulus calculation, ± 0.44 0.15 nm). Owing to crossings with other thicker fibrils but because of the mentioned inability to identify all kinks, the and the polymer overlapping with itself, we set an upper limit for persistence length may be slightly underestimated. possible real height values at 1 nm. Similar objects could be The persistence length is related to the bending rigidity, EI,of detected also in the W-CNC sample (Supplementary Fig. 7a–c). the fibrillar-like object: EI ¼ lk T, where k is the Boltzmann The root mean-squared surface roughness of the (3-Aminopropyl) B B constant and T is the temperature . As we reported before, triethoxysilane-modified mica surface between cellulose particles the height and thus the diameter of the cellulose rigid is 0.19 nm, which explains the relatively high variation in the segments can be estimated to be h ¼ (1.9–2.7) nm. Here we longitudinal height profile. The obtained height value is in good assume a round cross-section for CNFs and, taking into account agreement with the dimensions of the cellulose chain within the errors on the height values, the area moment of inertia cellulose Ib crystal structure with a monoclinic unit cell and base 4 4 lays in the range I ¼ ph /64 ¼ (0.64–2.61) nm . Together with lattice parameters aD0.78 nm and bD0.82 nm (refs 10,26). 6 3 k T ¼ 4.14  10 Pa nm , the Young’s modulus is found to be Taking into account the most probable height value for the E ¼ lk T/I ¼ 4.5–18.4 GPa. The deviation from the average majority of the tracked W-CNFs (2.35 nm), this suggests that a l B height has the highest impact on the elastic modulus value E 4  4 cellulose chain-packing arrangement is the dominating due to the power 4 dependence of the area moment of inertia structure. However, the height distribution broadness of W-CNF NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 7 & 2015 Macmillan Publishers Limited. All rights reserved. Error, nm ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 × 10 a c 3,600 1.2 0.9 3,300 0.6 3,000 0.3 2,700 0 50 100 150 200 250 300 350 0100 200 300 Processing length, nm Contour length, L (nm) bd x 10 1.01 180 3 ≈ 2.84 μm MSED 1 120 2 0.99 60 1 0.98 0 0 45 90 135 180 50 100 150 200 250 300 350 Processing length, nm Internal contour length, l (nm) Figure 6 | Length distribution of W-CNF rigid segments and estimation of their persistence length. (a) Persistence length l calculated via the MSED MSED method versus the processing length. (b) Adjusted coefficient of determination (goodness of fit) R and fitting error versus the processing length. adj (c) Length distribution of W-CNF rigid segments. The red vertical lines in a–c correspond to the processing length 170 nm at which the fitting error is minimal. (d) MSED versus internal contour length fit at the distance with the minimal fitting error. The resulting persistence length is l ¼ 2.84mm. MSED –2 nm 2 nm 150 nm b e 1.5 1.5 1.2 1.2 0.9 0.9 0.6 0.6 0.3 0.3 0 0 0 300 600 900 1,200 0 100 200 300 Internal contour length, l (nm) Internal contour length, l (nm) c f 900 300 ⟨h⟩ = 0.44 nm ⟨h⟩ = 0.84 nm = 0.15 nm  = 0.17 nm h h 600 200 300 100 0 0 0 0.3 0.6 0.9 1.2 1.5 0 0.3 0.6 0.9 1.2 1.5 Height, h (nm) Height, h (nm) Figure 7 | Observation of single cellulose chain and nanofibril with a 2  2 chain-packing structure. (a–c) The single polymer chain and (d–f) nanofibril with polymer chains composed in a possible 2  2 chain-packing structure. (a,d) AFM images with the tracked contours represented by blue lines that are slightly shifted for better visualization. (b,e) Height profiles along the tracked contours from a,d, respectively. (c,f) Height distributions of points along 20 tracked single cellulose polymer chains with the average height h ¼ 0.44 nm and s.d. s ¼ 0.15 nm, and 2  2 cellulose nanofibril with the average height h ¼ 0.84 nm and s.d. s ¼ 0.17 nm. Cutoffs at 1 nm (b,c) and 1.5 nm (e,f) were introduced to discard height data from chain-fibril crossings. (Fig. 4h), together with evidence of splitting also supports the based on scattering measurements, was reported to contain presence of n  m structures, where n and m could possibly be 24 polymer chains and favoured a ‘rectangular’ model , which is equal to 3, 4, 5 and 6. The structure of cellulose fibrils in wood, close in numbers to what we have derived in this study. 8 NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. 2  , nm MSED adj Height, h (nm) Number of points, n Number of points, n Height, h (nm) MS end-to-end dist, Number of segments, n 2 2 ⟨R ⟩ (nm ) NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 ARTICLE Furthermore, in very rare occasions we detected thin W-CNF in of a Gaussian distribution of kink angles that the kinks present in which the cellulose polymer chains possibly packed in a 2  2 W-CNF are not a result of alternating amorphous and crystalline 64–66 chain-packing structure (Fig. 7d). The longitudinal section and domains, as previously proposed in literature . Rather the height distribution are shown in Fig. 7e,f. The average height they may originate from the processing conditions used in the for these objects is almost double in comparison with single preparation of the nanofibrils. PF-QNM was used to probe polymer chains and equals to 0.84 nm (Fig. 7f). By analogy, we set mechanical properties of the nanocellulose in the transversal up an upper limit at 1.5 nm for the possible height values. direction, showing values of Young’s moduli, E , in the QNM Supplementary Fig. 7d depicts both single and 2  2 cellulose range of 20–50 GPa. Using the persistence length l method, polymer chains forming a remarkable network with a well- Young’s moduli in the longitudinal direction were extracted defined entanglement centre. Such structures could possibly in the E ¼ 4.5–18.4 GPa range, although this range can be originate from peeling off single cellulose chains during the harsh underestimated because of the influence of undetected kinks with mechanical CNF preparation process. small angle of deviations. Finally, we have described the statistics Finally, we calculated the persistence length for single cellulose of single free cellulose polymer chains, detected in both W-CNF polymer chains adsorbed on mica via the BCF and the MSED and W-CNC and extracted a persistence length in the range methods. The MSMD method cannot be suitably used here since l ¼ 63–65 nm. The discrepancy between ‘bare’ persistence length the assumption that the persistence length should be much larger l ¼ 5–15 nm and the observable persistence length originates than the contour length—as we discussed before—is not justified from electrostatic repulsion and can be rationally explained by the in the present case. The resulting values are: l ¼ 65 nm OSF theory. BCF (Supplementary Fig. 8a–c) and l ¼ 63 nm (Supplementary MSED Fig. 8d–f), respectively. These values are much larger than the Methods theoretical and simulation predictions of the ‘bare’ persistence Preparation of TEMPO-mediated oxidized W-CNF. The TEMPO-mediated oxidation of wood pulp was performed according to the procedure introduced in length, reported to be in the range of 5–15 nm (ref. 59). We ref. 67. The soft-wood pulp was first treated in a phosphate buffer at pH 6.8 at explain this discrepancy with the electrostatic repulsion between 60 C. The desired amount of sodium chlorite, TEMPO and sodium hypochlorite charged monomers along polymer chains and employ the was added and the dispersion was stirred for 2 h and 20 min, after which the pulp 60,61 Odijk–Skolnick–Fixman (OSF) theory to rationalize this was washed with deionized water and collected with vacuum filtration. The TEMPO-mediated oxidized material was dispersed in deionized water and observation. Within the framework of the OSF theory the total disintegrated by homogenization with a Microfluidizer M-110 EH (Microfluidics, persistence length can be represented as a sum of ‘bare’ (l ) USA), in a sequence consisting of four passes through the microchannels with and electrostatic (l ) components: l ¼ l þ l , where OSF 0 OSF diameters of 400–200mm at a pressure of 900 bar and for four passes through 2 2 l ¼ l r =4A (ref. 62), r is the Debye length, l ¼ 0.71 nm OSF B D B the microchannels with diameters of 200–100mm at 1,500 bar. The resulting suspension of TEMPO-mediated W-CNFs was first sonicated for 10 min using a is the Bjerrum length and A is the distance between two 13-mm-wide titanium probe at an output power of 70% (Vibra-Cell VC 750, neighbouring charged units. The charge density of 0.4 mmol g Sonics, USA), followed by centrifugation for 60 min at 4000 g. The W-CNF dis- was measured at pH 10, but the final value, at which the AFM persion was diluted to 0.001 w/w% with MilliQ water and shaken thoroughly. samples were prepared, was pH 5.65. By taking this pH change into account as well as the contribution from the counterions Preparation of W-CNC. W-CNCs were prepared by hydrochloric acid hydrolysis (I ¼ 4  10 M), the ionic strength is estimated to cI of the TEMPO-mediated oxidized CNFs (W-CNF) according to the procedure 6 63 I ¼ 6  10 M. Using the previously established approach we described in (ref. 32). Hydrochloric acid to a final concentration of 2.5 M was added to a dispersion of 100 g of a 1 w/w% (dry weight basis) W-CNF gel diluted calculate the degree of dissociation b ¼ 0.08, which leads to an with 316 ml deionized water and heated to 105 C for 6 h. The reaction was effective charge density of 0.032 mmol g . Assuming the quenched by dilution with the fivefold amount of deionized water. The hydrolysed distribution of charges only on the surface of particles, with material was washed with deionized water, collected by centrifugation for 10 min at average dimensions obtained from the statistical analysis, an 4000 g and dialysed for 5 days against deionized water using Sigma-Aldrich dialysis average distance between charges of AD7.7 nm is found. Under membranes with a molecular weight cutoff of B14,000 Da. After the dialysis, the suspension was first sonicated for 10 min using a 13-mm-wide titanium probe with these conditions of ionic strength, the Debye length r can be an output of 70% (Vibra-Cell VC 750, Sonics), followed by a centrifugation for estimated to be r D122 nm, and the resulting l D44.7 nm, D OSF 10 min at 4000 g. The deagglomerated dispersion was diluted to 0.001 w/w% with leading to a total expected persistence length of B50–60 nm, MilliQ water and shaken thoroughly. which is in excellent agreement with the experimental results (l ¼ 65 nm; l ¼ 63 nm), providing a solid basis for the BC MSED Preparation of B-CNC. B-CNCs were prepared from commercially available understanding of the large observed persistence length of coconut gel cubes (Chaokoh, Thailand). The coconut cubes (with a size of 3 3 individual nanocellulose chains. In order to test whether B1  1  1cm ) were pretreated by first washing for three times with 2 dm of deionized water, followed by stirring in 2 dm of a 0.1 M sodium hydroxide counterion condensation plays a role in the final observable solution for 48 h and finally washing with deionized water until the pH stabilized at persistence length, we perform a quick Oosawa–Manning B7. The hydrolysis was performed by soaking 100 g of the pretreated coconut counterion condensation check. The condensation of cubes in sulfuric acid with a concentration of 40 w/w% at 80 C for 4 h. The counterions occurs when l r41, where r is the linear charge hydrolysed materials were washed twice with deionized water, collected by centrifugation and dialysed for 5 days against deionized water using Sigma-Aldrich density (r ¼ 1/A ¼ 0.13 nm ). In the present case l rE0.09, dialysis membranes with a molecular weight cutoff of B14,000 Da. After the suggesting that counterion condensation can be discarded. dialysis, the suspension was first sonicated for 10 min using a 13-mm-wide titanium probe (Vibra-Cell VC 750, Sonics), at an output power of 70%, followed by centrifugation for 60 min at 4000 g. The deagglomerated dispersion was diluted to 0.001 w/w% with MilliQ water and shaken thoroughly. Discussion In the present work we provide a comprehensive and consistent structural description over multiple length scales of nanocellulose Surface charge determination. The surface charge of W-CNC and B-CNC was determined with polyelectrolyte titration using a Stabino Particle Charge Mapping of different origin and pretreatment: TEMPO-mediated oxidized system (Microtrac Europe GmbH, Germany). The nanocellulose dispersions were W-CNF, W-CNC and B-CNC. We find that all types of diluted in MilliQ water and then titrated with a 0.001 N polydiallyl dimethyl nanocellulose fibrils and crystals with an observable twisting ammonium chloride solution. An average charge density was obtained from three along the contour possess a right-handed chirality. The right- or more measurements. handed chirality was detected on the level of fibril bundles in all systems and on the single fibril level in W-CNF. By a statistical AFM and PF-QNM measurements. A droplet of 0.001 w/w% solution of analysis of kink angle distribution, we conclude from the absence different types of cellulose was deposited on chemically modified mica with NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 9 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 (3-Aminopropyl)triethoxysilane following a protocol described in ref. 41. AFM and 5. Revol, J. F., Bradford, H., Giasson, J., Marchessault, R. H. & Gray, D. G. PF-QNM measurements were performed by using a MultiMode VIII Scanning Helicoidal self-ordering of cellulose microfibrils in aqueous suspension. Int. J. Probe Microscope (Bruker, USA) covered with an acoustic hood to minimize Biol. Macromol. 14, 170–172 (1992). vibrational noise. AFM images were acquired continuously in the tapping mode 6. Hamedi, M. et al. Nanocellulose aerogels functionalized by rapid layer-by-layer under ambient conditions using commercial cantilevers (Bruker). In order to assembly for high charge storage and beyond. Angew. Chem. Int. Ed. 52, perform PF-QNM measurements on all cellulose nanoparticles using the same 12038–12042 (2013). substrate and under identical environmental and AFM tip conditions, small 7. Shopsowitz, K. E., Qi, H., Hamad, W. Y. & Maclachlan, M. J. Free-standing aliquots of 0.001 w/w% dispersions of W-CNF, W-CNC and B-CNC were mixed in mesoporous silica films with tunable chiral nematic structures. Nature 468, proportion 1:1:1 before deposition on mica surfaces. The AFM cantilevers (Bruker) 422–425 (2010). were calibrated on the calibration samples (Bruker)—typically low-density 8. Olsson, R. T. et al. Making flexible magnetic aerogels and stiff magnetic polyethylene and polystyrene—covering the following ranges of Young’s moduli: nanopaper using cellulose nanofibrils as templates. Nat. Nanotechnol. 5, from 100 MPa to 2 GPa (for low-density polyethylene) and from 1 to 20 GPa (for 584–588 (2010). polystyrene). The measurements were performed on small indentation depths to 9. Moon, R. J., Martini, A., Nairn, J., Simonsen, J. & Youngblood, J. Cellulose avoid any artefacts of substrate. The analysis of the elastic modulus was performed nanomaterials review: structure, properties and nanocomposites. Chem. Soc. using the Nanoscope Analysis software and calculated according the DMT model. Rev. 40, 3941–3994 (2011). 10. Nishiyama, Y., Langan, P. & Chanzy, H. Crystal structure and hydrogen- Transmission electron microscopy. TEM imaging was carried out on carbon- bonding system in cellulose Ib from synchrotron X-ray and neutron fiber coated copper grids that were glow-discharged for 45 s (Emitech K100X, GB) diffraction. J. Am. Chem. Soc. 124, 9074–9082 (2002). directly before sample fixation. Sample grid preparation for negative staining was 11. Majoinen, J., Kontturi, E., Ikkala, O. & Gray, D. G. SEM imaging of chiral as follows: 5ml of sample dispersion for 1 min, 5ml of 2% uranyl acetate for 1 s and nematic films cast from cellulose nanocrystal suspensions. Cellulose 19, again 5ml of 2% uranyl acetate for 15 s to achieve a noncrystalline film of stain 1599–1605 (2012). embedding the fibres. Following each step, the excess moisture was drained along 12. Kelly, J. A. et al. Evaluation of form birefringence in chiral nematic mesoporous the periphery using a piece of filter paper. Dried grids were examined using TEM materials. J. Mater. Chem. C 2, 5093–5097 (2014). (FEI, model Morgagni, NL) operated at 100 kV. 13. Lagerwall, J. P. F. et al. Cellulose nanocrystal-based materials: from liquid crystal self-assembly and glass formation to multifunctional thin films. NPG Asia Mater. 6, e80 (2014). Cryo-SEM. Sample aliquots (3.5ml) were applied to glow-discharged, carbon- 14. Saito, T., Uematsu, T., Kimura, S., Enomaea, T. & Isogai, A. Self-aligned coated Cu-grids for 1 min, blotted with filter paper along the periphery and plunge- integration of native cellulose nanofibrils towards producing diverse bulk frozen in liquid ethane. Vitrified specimens were then transferred and mounted materials. Soft Matter 7, 8804–8809 (2011). under liquid nitrogen on a self-made grid holder and finally transferred under liquid nitrogen into a precooled (  120 C) freeze-fracturing system BAF 060 (Bal- 15. Kobayashi, Y., Saito, T. & Isogai, A. Aerogels with 3D ordered nanofiber Tec/Leica, Vienna). For freeze-drying the samples were warmed up in 5 C skeletons of liquid-crystalline nanocellulose derivatives as tough and increments every 15 min until  80 C was reached at 10 mbar. Coating was transparent insulators. Angew. Chem. Int. Ed. 53, 10394–10397 (2014). performed with 1.5 nm tungsten at 45 followed by 1.5 nm under continuous 16. Wicklein, B. et al. Thermally insulating and fire-retardant lightweight elevation angle changes from 45 to 90 and back to 45. Cryo-SEM was anisotropic foams based on nanocellulose and graphene oxide. Nat. performed in a field emission SEM Leo Gemini 1530 (Carl Zeiss, Germany) Nanotechnol. 10, 277–283 (2015). equipped with a cold stage to maintain the specimen temperature at  110 C 17. Beck-Candanedo, S., Roman, M. & Gray, D. G. Effect of reaction conditions on (VCT Cryostage, Bal-Tec/Leica). Signals from the SE-inlens detector (acceleration the properties and behavior of wood cellulose nanocrystal suspensions. voltage 5 kV) were used for image formation. Only the contrast and brightness Biomacromolecules 6, 1048–1054 (2005). of the pictures were adjusted. 18. Shinoda, R., Saito, T., Okita, Y. & Isogai, A. Relationship between length and degree of polymerization of TEMPO-oxidized cellulose nanofibrils. Biomacromolecules 13, 842–849 (2012). Tracking of the TEMPO-oxidized W-CNFs. The coordinates of the CNFs 19. Saito, T., Kuramae, R., Wohlert, J., Berglund, L. A. & Isogai, A. An ultrastrong (W-CNF) were obtained using an in-house programme written in MATLAB. Each nanofibrillar biomaterial: the strength of single cellulose nanofibrils revealed via tracked CNF can be represented by its contour—a sequence of points connected sonication-induced fragmentation. Biomacromolecules 14, 248–253 (2013). with straight segments that are positioned along the fibril bright ridge on an AFM 20. Fernandes, A. N. et al. Nanostructure of cellulose microfibrils in spruce wood. image. All contours acquired in this study have a constant distance between pro- jections of these points on the image plane, which is the step size s ¼ 2.9 nm. This Proc. Natl Acad. Sci. USA 108, E1195–E1203 (2011). 21. Orts, W. J., Godbout, L., Marchessault, R. H. & Revol, J.-F. Enhanced ordering way of tracking is similar to the procedure we previously applied for amyloid fibril 33,40 systems with one particular addition. Owing to significant directional variation of liquid crystalline suspensions of cellulose microfibrils: a small angle neutron of W-CNF contours (low curvature along rigid segments, high curvature in scattering study. Macromolecules 31, 5717–5725 (1998). vicinities of kinks) we employed a concept of masks that define kink areas. The 22. Hanley, S., Revol, J., Godbout, L. & Gray, D. Atomic force microscopy and affinity of a contour to bend is different depending on whether points are inside or transmission electron microscopy of cellulose from Micrasterias denticulata; outside the mask area; in other words, it allows contours to have heterogeneous evidence for a chiral helical microfibril twist. Cellulose 4, 209–220 (1997). stiffness (large stiffness along rigid segments and low stiffness in vicinities of kinks). 23. Khandelwal, M. & Windle, A. Origin of chiral interactions in cellulose This is essential for the tracking algorithm to correctly follow the actual features of supra-molecular microfibrils. Carbohydr. Polym. 106, 128–131 (2014). nanocellulose fibrils. The position of these masks for each object and the initial 24. Lahiji, R. R. et al. Atomic force microscopy characterization of cellulose contours were initialized manually, the latter with the auxiliary help of the A* nanocrystals. Langmuir 26, 4480–4488 (2010). pathfinding algorithm . To obtain subpixel accuracy, contours were deformed and 25. Postek, M. T. et al. Development of the metrology and imaging of cellulose fitted precisely to the cellulose fibrils middle lines using the slightly modified Open nanocrystals. Meas. Sci. Technol. 22, 024005 (2011). 69,70 Active Contours algorithm . All statistical information was acquired from high- 26. Matthews, J. F. et al. Computer simulation studies of microcrystalline cellulose resolution AFM images with spatial dimensions of 15  15mm and 5,120  5,120 Ib. Carbohydr. Res. 341, 138–152 (2006). pixels. In total, data on 2,380 fibrils from W-CNF were extracted to reach sufficient 27. Paavilainen, S., Rog, T. & Vattulainen, I. Analysis of twisting of cellulose statistical significance. All data-processing methods used in this work such as the nanofibrils in atomistic molecular dynamics simulations. J. Phys. Chem. B 115, calculation of the kink angle, length and height distributions, persistence length 3747–3755 (2011). evaluation and plotting fibril height, profiles were performed in the same software. 28. Tashiro, K. & Kobayashi, M. Theoretical evaluation of three-dimensional elastic constants of native and regenerated celluloses: role of hydrogen bonds. Polymer (Guildf) 32, 1516–1526 (1991). References 29. Lindman, B., Karlstro¨m, G. & Stigsson, L. On the mechanism of dissolution of 1. Godinho, M. H., Canejo, J. P., Pinto, L. F. V., Borges, J. P. & Teixeira, P. I. C. cellulose. J. Mol. Liq. 156, 76–81 (2010). How to mimic the shapes of plant tendrils on the nano and microscale: spirals 30. Klemm, D. et al. Nanocelluloses: a new family of nature-based materials. and helices of electrospun liquid crystalline cellulose derivatives. Soft Matter 5, 2772–2776 (2009). Angew. Chem. Int. Ed. 50, 5438–5466 (2011). 2. Gray, D. G. Isolation and handedness of helical coiled cellulosic thickenings 31. Siro, I. & Plackett, D. Microfibrillated cellulose and new nanocomposite materials: a review. Cellulose 17, 459–494 (2010). from plant petiole tracheary elements. Cellulose 21, 3181–3191 (2014). 32. Salajkova´, M., Berglund, L. A. & Zhou, Q. Hydrophobic cellulose nanocrystals 3. Håkansson, K. M. O. et al. Hydrodynamic alignment and assembly of nanofibrils resulting in strong cellulose filaments. Nat. Commun. 5, 4018 modified with quaternary ammonium salts. J. Mater. Chem. 22, 19798–19805 (2014). (2012). 4. Kelly, J. A., Giese, M., Shopsowitz, K. E., Hamad, W. Y. & MacLachlan, M. J. 33. Usov, I., Adamcik, J. & Mezzenga, R. Polymorphism complexity and The development of chiral nematic mesoporous materials. Acc. Chem. Res. 47, handedness inversion in serum albumin amyloid fibrils. ACS Nano 7, 1088–1096 (2014). 10465–10474 (2013). 10 NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 ARTICLE 34. Gray, D. G. Chiral nematic ordering of polysaccharides. Carbohydr. Polym. 25, 61. Skolnick, J. & Fixman, M. Electrostatic persistence length of a wormlike 277–284 (1994). polyelectrolyte. Macromolecules 10, 944–948 (1977). 35. Belli, S., Dussi, S., Dijkstra, M. & van Roij, R. Density functional theory for 62. Dobrynin, A. V. & Rubinstein, M. Theory of polyelectrolytes in solutions and at chiral nematic liquid crystals. Phys. Rev. E 90, 020503 (2014). surfaces. Prog. Polym. Sci. 30, 1049–1118 (2005). 36. Adamcik, J., Berquand, A. & Mezzenga, R. Single-step direct measurement of 63. Fall, A. B., Lindstro¨m, S. B., Sundman, O., Odberg, L. & Wågberg, L. Colloidal amyloid fibrils stiffness by peak force quantitative nanomechanical atomic force stability of aqueous nanofibrillated cellulose dispersions. Langmuir 27, microscopy. Appl. Phys. Lett. 98, 193701 (2011). 11332–11338 (2011). 37. Adamcik, J. et al. Measurement of intrinsic properties of amyloid fibrils by the 64. Montanari, S., Roumani, M., Heux, L. & Vignon, M. R. Topochemistry of peak force QNM method. Nanoscale 4, 4426–4429 (2012). carboxylated cellulose nanocrystals resulting from TEMPO-mediated oxidation. 38. Ling, S. et al. Modulating materials by orthogonally oriented b-strands: Macromolecules 38, 1665–1671 (2005). composites of amyloid and silk fibroin fibrils. Adv. Mater. 26, 4569–4574 (2014). 65. Siqueira, G., Tapin-Lingua, S., Bras, J., da Silva Perez, D. & Dufresne, A. 39. Lamour, G., Yip, C. K., Li, H. & Gsponer, J. High intrinsic mechanical flexibility Morphological investigation of nanoparticles obtained from combined of mouse prion nanofibrils revealed by measurements of axial and radial mechanical shearing, and enzymatic and acid hydrolysis of sisal fibers. Cellulose Young’s moduli. ACS Nano 8, 3851–3861 (2014). 17, 1147–1158 (2010). 40. Jordens, S., Isa, L., Usov, I. & Mezzenga, R. Non-equilibrium nature of 66. Hamedi, M. M. et al. Highly conducting, strong nanocomposites based on two-dimensional isotropic and nematic coexistence in amyloid fibrils at liquid nanocellulose-assisted aqueous dispersions of single-wall carbon nanotubes. interfaces. Nat. Commun. 4, 1917 (2013). ACS Nano 8, 2467–2476 (2014). 41. Schefer, L., Adamcik, J. & Mezzenga, R. Unravelling secondary structure 67. Saito, T. et al. Individualization of nano-sized plant cellulose fibrils by direct changes on individual anionic polysaccharide chains by atomic force surface carboxylation using TEMPO catalyst under neutral conditions. microscopy. Angew. Chem. Int. Ed. 53, 5376–5379 (2014). Biomacromolecules 10, 1992–1996 (2009). 42. Li, C. & Mezzenga, R. Functionalization of multiwalled carbon nanotubes and 68. Hart, P., Nilsson, N. & Raphael, B. A formal basis for the heuristic their pH-responsive hydrogels with amyloid fibrils. Langmuir 28, 10142–10146 determination of minimum cost paths. IEEE Trans. Syst. Sci. Cybern. 4, (2012). 100–107 (1968). 69. Kass, M., Witkin, A. & Terzopoulos, D. Snakes: active contour models. Int. J. 43. Usov, I. & Mezzenga, R. FiberApp: an open-source software for tracking and analyzing polymers, filaments, biomacromolecules, and fibrous objects. Comput. Vis. 1, 321–331 (1988). Macromolecules 48, 1269–1280 (2015). 70. Smith, M. B. et al. Segmentation and tracking of cytoskeletal filaments using 44. Nishiyama, Y. Structure and properties of the cellulose microfibril. J. Wood Sci. open active contours. Cytoskeleton (Hoboken) 67, 693–705 (2010). 55, 241–249 (2009). 45. Battista, O. Hydrolysis and crystallization of cellulose. Ind. Eng. Chem. 42, Acknowledgements 502–507 (1950). We acknowledge support from the Swiss National Science Foundation (SNF; 2-77002-11) 46. Wickholm, K., Larsson, P. & Iversen, T. Assignment of non-crystalline forms in and by the Scientific Center for Optical and Electron Microscoy of ETH Zurich cellulose I by CP/MAS 13 C NMR spectroscopy. Carbohydr. Res. 312, 123–129 (ScopeM). G.N. acknowledges funding from the Gunnar Sundblad Research Foundation. (1998). C.S. and L.B. acknowledge the Wallenberg Wood Science Center for funding. Korneliya 47. Sturcova, A., His, I., Apperley, D. C., Sugiyama, J. & Jarvis, M. C. Structural Gordeyeva is thanked for technical assistance. Lars Wågberg and Nicholas Tchang details of crystalline cellulose from higher plants. Biomacromolecules 5, Cervin are thanked for providing the TEMPO-mediated oxidized Domsjo¨ (60/40 1333–1339 (2004). spruce/pine) wood cellulose pulp. 48. Newman, R. H. Estimation of the lateral dimensions of cellulose crystallites using 13C NMR signal strengths. Solid State Nucl. Magn. Reson. 15, 21–29 (1999). Author contributions 49. McGeoch, C. C. A Guide to Experimental Algorithmics (Cambridge University I.U. designed the cellulose tracking and data-processing methods, performed Press, 2012). statistical analysis and wrote the manuscript. G.N. contributed to the data analysis 50. Usov, I., Adamcik, J. & Mezzenga, R. Polymorphism in bovine serum albumin and writing of the manuscript. J.A. performed AFM and PF-QNM experiments. S.H. fibrils: morphology and statistical analysis. Faraday Discuss. 166, 151–162 performed Cryo-SEM and TEM experiments. C.S. carried out the nanocellulose (2013). preparation. A.F. contributed to the data analysis. L.B. contributed to the design of 51. Rubinstein, M. & Colby, R. H. Polymer Physics (Oxford University Press, 2003). the study, interpretation of results and wrote the manuscript. R.M. contributed to the 52. Doi, M. & Edwards, S. F. The Theory of Polymer Dynamics (Oxford University data analysis, result interpretation, designed and directed the study and wrote the Press, 1986). manuscript. 53. Rivetti, C., Guthold, M. & Bustamante, C. Scanning force microscopy of DNA deposited onto mica: equilibration versus kinetic trapping studied by statistical Additional information polymer chain analysis. J. Mol. Biol. 264, 919–932 (1996). Supplementary Information accompanies this paper at http://www.nature.com/ 54. Smith, J. F., Knowles, T. P. J., Dobson, C. M., Macphee, C. E. & Welland, M. E. naturecommunications Characterization of the nanoscale properties of individual amyloid fibrils. Proc. Natl Acad. Sci. USA. 103, 15806–15811 (2006). Competing financial interests: The authors declare no competing financial interests. 55. Manning, G. S. Polymer persistence length characterized as a critical length for instability caused by a fluctuating twist. Phys. Rev. A 34, 668–670 (1986). Reprints and permission information is available online at http://npg.nature.com/ 56. Usov, I. & Mezzenga, R. Correlation between nanomechanics and polymorphic reprintsandpermissions/ conformations in amyloid fibrils. ACS Nano 8, 11035–11041 (2014). How to cite this article: Usov, I. et al. Understanding nanocellulose chirality and 57. Li, Q. & Renneckar, S. Molecularly thin nanoparticles from cellulose: isolation structure–properties relationship at the single fibril level. Nat. Commun. 6:7564 of sub-microfibrillar structures. Cellulose 16, 1025–1032 (2009). doi: 10.1038/ncomms8564 (2015). 58. Li, Q. & Renneckar, S. Supramolecular structure characterization of molecularly thin cellulose I nanoparticles. Biomacromolecules 12, 650–659 (2011). This work is licensed under a Creative Commons Attribution 4.0 59. Kroon-Batenburg, L. M. J., Kruiskamp, P. H., Vliegenthart, J. F. G. & Kroon, J. International License. The images or other third party material in this Estimation of the persistence length of polymers by MD simulations on small fragments in solution. application to cellulose. J. Phys. Chem. B 101, 8454–8459 article are included in the article’s Creative Commons license, unless indicated otherwise (1997). in the credit line; if the material is not included under the Creative Commons license, 60. Odijk, T. Polyelectrolytes near the rod limit. J. Polym. Sci. Polym. Phys. Ed. 15, users will need to obtain permission from the license holder to reproduce the material. 477–483 (1977). To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 11 & 2015 Macmillan Publishers Limited. All rights reserved. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nature Communications Springer Journals

Understanding nanocellulose chirality and structure–properties relationship at the single fibril level

Download PDF
11 pages

Loading next page...
 
/lp/springer-journals/understanding-nanocellulose-chirality-and-structure-properties-FpKfaAX9M0

References (138)

Publisher
Springer Journals
Copyright
Copyright © 2015 by The Author(s)
Subject
Science, Humanities and Social Sciences, multidisciplinary; Science, Humanities and Social Sciences, multidisciplinary; Science, multidisciplinary
eISSN
2041-1723
DOI
10.1038/ncomms8564
Publisher site
See Article on Publisher Site

Abstract

ARTICLE Received 25 Nov 2014 | Accepted 19 May 2015 | Published 25 Jun 2015 DOI: 10.1038/ncomms8564 OPEN Understanding nanocellulose chirality and structure–properties relationship at the single fibril level 1 1 1 1 2,3 2 Ivan Usov , Gustav Nystro¨m , Jozef Adamcik , Stephan Handschin , Christina Schu¨tz , Andreas Fall , 2 1 Lennart Bergstro¨m & Raffaele Mezzenga Nanocellulose fibrils are ubiquitous in nature and nanotechnologies but their mesoscopic structural assembly is not yet fully understood. Here we study the structural features of rod-like cellulose nanoparticles on a single particle level, by applying statistical polymer physics concepts on electron and atomic force microscopy images, and we assess their physical properties via quantitative nanomechanical mapping. We show evidence of right-handed chirality, observed on both bundles and on single fibrils. Statistical analysis of contours from microscopy images shows a non-Gaussian kink angle distribution. This is inconsistent with a structure consisting of alternating amorphous and crystalline domains along the contour and supports process-induced kink formation. The intrinsic mechanical properties of nanocellulose are extracted from nanoindentation and persistence length method for transversal and longitudinal directions, respectively. The structural analysis is pushed to the level of single cellulose polymer chains, and their smallest associated unit with a proposed 2  2 chain-packing arrangement. 1 2 Department of Health Science and Technology, ETH Zurich, Schmelzbergstrasse 9, LFO E23, Zurich 8092, Switzerland. Department of Materials and Environmental Chemistry, Stockholm University, Svante Arrhenius vag 16C, Stockholm 10691, Sweden. Wallenberg Wood Science Center, KTH, Teknikringen 56, Stockholm 10044, Sweden. Correspondence and requests for materials should be addressed to R.M. (email: [email protected]). NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 1 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 n nature, cellulose, the major load-bearing structure in all of substructures, down to the single cellulose chain. From the living plants, shows complex superstructures, chirality and statistical analysis, the persistence length of the W-CNF is 1,2 Ichiral inversions over different length scales . This extracted, and indirectly, its intrinsic rigidity in the longitudinal fascinating structural behaviour, in combination with a high direction, while the direct measurements of the Young’s specific strength and stiffness, has inspired work to engineer moduli obtained using PF-QNM characterize mechanical proper- cellulose materials with tailored mechanical and optical ties of the nanocellulose samples in the transversal plane. 3,4 properties . Over the last decades, elongated rod-like cellulose Most importantly, the detailed statistical investigation of the nanoparticles, shorter cellulose nanocrystals (CNCs) and longer morphology of the W-CNF provides convincing evidence that cellulose nanofibrils (CNF), have been used to form chiral the commonly accepted model of CNF structure as built of 5 6 7 nematic liquid crystals , aerogels , photonic and inorganic alternating regions of crystalline and amorphous cellulose 8 9,30,31 hybrid materials . Even though these materials have impressive domains along the fibril length cannot explain the properties and show great potential for a broad range of presence of kinks. applications, the fine structure of their nanocellulose components has not yet been fully elucidated . Only through a detailed knowledge of the chirality and the structure of the Results smallest nanocellulose building blocks, new strategies for Overview of the nanocellulose systems. The low-magnification advanced bottom-up nanotechnologies and assembly of new images of the three different nanocellulose systems (W-CNF, helical metamaterials may become available. W-CNC and B-CNC) via three different microscopy techniques On the atomic length scale, high-resolution neutron and X-ray (AFM, Cryo-SEM and TEM) are shown in Fig. 1. On the basis of scattering data have revealed the crystalline structure of cellulose the initial visual inspection we provide a description of their down to the exact atomic positions of the polymers in the unit morphology and typical structural parameters. In the W-CNF cell . How CNC arrange in chiral nematic liquid crystalline (Fig. 1a–c), CNFs have a relatively constant cross-section with 11–13 14 15 phases and how CNF arrange in films , aerogels or average widths in the range of B2  3 nm and lengths foams is also well characterized. In between these two length B0.1  1mm. One particular feature of this sample is that almost scales, however, a gap remains, where the structure and assembly every nanofibril has sharp bends between otherwise straight of nanocellulose fibrils are not yet fully understood. Given the segments, a feature hereinafter referred to as kinks. Moreover, hierarchical nature of cellulose, an investigation of this these nanofibrils can sometimes still be observed to be assembled intermediate length scales may reveal important structural in fibril bundles, despite the strong mechanical treatment applied information, improve the fundamental understanding and pave during the homogenization process for sample preparation. The the way to new strategies in materials assembly. W-CNC sample (Fig. 1d–f) is represented by rather straight Traditional nanocellulose research relies on the disintegration particles that have a large variation in cross-section along their of a cellulosic raw material into its smallest possible components contours. They can be described as large multistranded bundles of (CNC and CNF) and the subsequent re-assembly of these laterally assembled CNFs. Their average widths are in the range of building blocks into new materials. The harsh chemical and B3  30 nm, but the lengths are similar to the W-CNF, that is, mechanical treatments used in the disintegration process may B0.1  1mm. In rare occasions it is possible to observe thin have a large influence on the final structural properties of the segments with kinks, a signature of the W-CNF precursor cellulose nanoparticles . This is often overlooked and only a few samples, indicating that the hydrolysis was not fully completed 18,19 detailed experimental investigations have been performed . (Supplementary Fig. 1). For the B-CNC (Fig. 1g–i), the observed Previous experimental microscopy and scattering work nanofibrils have typical lengths of B1–5mm and irregular widths have allowed resolving average particle dimensions , indirect in the range of B10  50 nm. The features of nanocellulose (from neutron scattering) and direct evidence (from fibril systems are consistent within all distinct microscopy techniques 22,23 bundles) of particle twist, as well as local mechanical used in this study, which confirms that any possible artefact properties with peak force quantitative nanomechanical mapping associated with either the microscopy technique or sample pre- 24,25 (PF-QNM) . Simulations have confirmed a twisted structure at paration procedure is below the interpretable resolution. 26,27 equilibrium , and also estimated single particle mechanical During the preparation of the nanocellulose samples, ionizable properties similar to the measured values . However, issues related groups are introduced (-COOH for W-CNF and W-CNC, and to the quantification and origin of the CNF conformations remain -SO H for B-CNC) that electrostatically improve the stability of elusive and the fundamental question of the (in)solubility of the particles in dispersion. The charge density of W-CNC and cellulose in water has recently become a matter of intense debate . B-CNC was measured with polyelectrolyte titration, yielding With the rapid development of experimental techniques, new and gravimetrically normalized values of 0.4 mmol g (W-CNC) more detailed structural information becomes available, allowing and 0.03 mmol g (B-CNC). Because the charge density is not these outstanding fundamental questions to be revisited and expected to be affected by the hydrolysis treatment , we assign a conclusively assessed. charge density of 0.4 mmol g also for the W-CNF sample. In this work, state-of-the-art atomic force, cryogenic scanning On the basis of the microscopy images, we find that there is a electron and transmission electron microscopy (AFM, Cryo-SEM larger fraction of individual particles of W-CNF compared with and TEM, respectively) are combined with advanced statistical W-CNC. The better colloidal stability of W-CNF may be related analysis and concepts from polymer physics to investigate three to a steric hindrance effect due to the kinks preventing different types of nanocellulose, TEMPO(2,2,6,6-tetramethylpi- neighbouring fibrils to come into close contact and form peridine-1-oxyl)-mediated oxidized wood cellulose nanofibrils bundles. The stronger tendency for B-CNC compared with (W-CNF), wood cellulose nanocrystals (W-CNC) and sulfuric W-CNC to form bundles is probably related to the relatively low acid hydrolyzed bacterial cellulose nanocrystals (B-CNC). All charge density of B-CNC. It should also be noted that the pH of samples show clear evidence of a right-handed chirality both at the W-CNC and B-CNC samples is drastically decreased during the level of bundles of fibrils and on the individual fibril level hydrolysis, causing a reduction in the charge density of the CNC (W-CNF). Deeper investigation of one of the nanocellulose particles, resulting in the formation of bundles and aggregates. It families (W-CNF) reveals detailed information on particle is possible that not all of these aggregates become fully dimensions and provides compelling evidence for multiple levels redispersed when the pH is readjusted. 2 NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 ARTICLE All types of nanocellulose possess a right-handed chirality. The helical cellulose thickenings . In contrast to these observations of magnified microscopy images reveal that all cellulose fibrils and higher-order structures, the results presented here establish that crystals with observable chirality are right-handed (Fig. 2). the single CNFs with observable chirality always bear a right- For the W-CNF (Fig. 2a–c), right-handed chirality is observed handed twisting. Because also the observed chirality of the fibril both on the single fibril level (Fig. 2a, left and Supplementary bundles is right-handed, this indicates that there is no chiral Fig. 2) and for bundles of fibrils (Fig. 2a, right). We note that a inversion on this length scale. twist in amplitude measurements is observable only if the local Recent studies based on density functional theory of right- microfibril cross-section has distinct corners—for cylindrical handed hard helices indicate that the way chirality transfers from fibrils this would be invisible. For the CNC particles, chirality the building block to the chiral nematic phase is non-trivial and observations were only possible on crystal bundles, since isolated depends at least on two different parameters, entropy (excluded CNCs with constant and uniform cross-section could not be volume of left-/right-handed particle pairs) and thermodynamics observed. (volume fraction of rods) . This suggests that a chirality The presence of the right-handed twist along cellulose inversion from a right-handed particle to a left-handed chiral nanoparticles and bundles of nanoparticles is not routinely nematic phase may be expected, provided that the concentration observable (5–10% from hundreds of visualized particles); of rods in the chiral nematic is in the right range . Thus, the however, when a twist is observable, it is always right-handed. results at the single fibril and small bundle level given here may This relatively low amount of twisted particles in W-CNF could provide the necessary information to bridge the gap between be explained by a finite AFM resolution, which is approaching the theoretical models and experimental results in nanocellulose limit to detect these features. For W-CNC and B-CNC, the cholesteric nematic phases. twisting normally occurs in bundles with high structural organization of single particles, but could be hidden in the case of loosely packed aggregates. PF-QNM of the nanocellulose samples. The mechanical 22,23 These observations are in line with previous experimental properties of cellulose in the transversal direction were probed 26,27 36 and theoretical studies, supporting a right-handed chirality. using PF-QNM . Previously, this method has been successfully 37,38 However, the experimental results presented to date, rely on used for systems of amyloid fibrils that have similar higher-order structures of fibril aggregates that are in mm-length dimensions to cellulose fibrils. Figure 3a shows an AFM image scale. This distinction is important since there are different with the composition of all three types of nanocellulose particles: examples in nature where chirality inversion occurs over different W-CNF, W-CNC and B-CNC, mixed in 1:1:1 proportion, with length scales and cellulose may follow the same trends. For the corresponding PF-QNM map depicted in Fig. 3b. The example, the observed cholesteric liquid crystal ordering of CNCs Young’s modulus values, E , were estimated according to the QNM is left-handed and the tracheary elements, responsible for Derjaguin–Mueller–Toporov (DMT) model, and were collected transporting water in most plants, are reinforced by left-handed along the contours of each cellulose type to obtain the a d –5 nm 5 nm –10 nm 10 nm –40 nm 40 nm 1,000 nm 1,000 nm 1,000 nm b e h 250 nm 500 nm 1,000 nm c f i 150 nm 300 nm 600 nm Figure 1 | Overview microscopy images of the nanocellulose samples. (a–c) TEMPO-mediated oxidized W-CNF, (d–f) W-CNC and (g–i) B-CNC samples via AFM (a,d,g) Cryo-SEM (b,e,h) and TEM (c,f,i). All AFM images have the same magnification in order to provide a direct comparison between nanocellulose particles. NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 3 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 41 42 corresponding modulus distributions (Fig. 3c–e). We find that the polysaccharide polymers or carbon nanotubes . The Young’s moduli lay in the range of 20–50 GPa for all three types coordinates of 2,380 fibrils were acquired from AFM images of nanocellulose particles, and are in good agreement to that covers 450-mm area of mica substrate, using a specially previously measured moduli from PF-QNM on nanocellulose designed in-house software written in MATLAB (see Methods) . (6–50 GPa; refs 24,25). The Gaussian fits of these distributions Objects that were considered to be bundles of fibrils with with parameters m and s (mean value and s.d.) result in distinctively larger diameters were, whenever possible, discarded ± ± the following elastic moduli (m s): 34.4 5.3 GPa for W-CNF from the analysis. The positions of the kinks were determined ± ± (Fig. 3c), 31.1 5.9 GPa for W-CNC (Fig. 3d) and 32.3 4.1 GPa manually using special masks (Fig. 4b, green boxes). Two for B-CNC (Fig. 3e). Moreover, we note that the PF-QNM segments of a contour that are crossing a mask element measurements yield remarkably close values of Young’s moduli (entering and exiting segments) define a contour angle for all three samples investigated. Since the AFM sample (Fig. 4c), which is also referred as a kink angle. One important preparation conditions allow for a qualitative side-by-side note to mention is that it is difficult to distinguish between very comparison of the samples, these results indicate that the small kink angles (below 20) and random thermal fluctuation in mechanical properties of the samples do not depend on the the AFM images. For example, Fig. 4d depicts a fibril with two nanocellulose-processing conditions. obvious kinks that are unambiguously determined and one particular place with a hint of a kink only. The threshold for observable kinks does affect the results, but in a predictive Tracking and statistical analysis of W-CNF. The TEMPO- manner. mediated oxidized W-CNF system is of particular interest for the It is well known that CNF consists of cellulose chains with application of statistical analysis since in comparison with the different degrees of order, from highly crystalline arrangements to other nanocellulose systems, it has rather uniform contour a slightly perturbed distribution of the chains. Two models have shapes. The statistical analysis of the nanofibril contours provides been proposed to describe the distribution of the disordered and unique insight on the nature of their kinks, and moreover, ordered regions in CNF . In the first model, the less-ordered it allows an estimation of the Young’s modulus value of rigid chains, often referred to as amorphous chains, are distributed between crystalline regions of chains along the fibril direction segments via the persistence length approach . Furthermore, the basic morphological parameters such as length and height (analogous to semicrystalline polymers) . In the second model, distributions allow elucidating the packing model of single the less-ordered chains are located towards the surface of fibrils cellulose polymer chains within the nanocellulose fibril. with a crystalline core of chains . The former model is We have used a tracking procedure to obtain coordinates of the commonly invoked as a rationale for CNC particle preparation, CNF contours (Fig. 4a, blue lines) similar to the one used in as hydrolysis is believed to primarily dissolve the amorphous 33,40 previous lines of work on amyloid fibrils , carrageenan regions of cellulose chains, leaving only the short, crystalline a d –15° 15° –30° 30° 50 nm 100 nm 150 nm 500 nm b e h 75 nm 150 nm 500 nm c f i 75 nm 150 nm 500 nm Figure 2 | All samples show right-handed twisting. (a–c) TEMPO-mediated oxidized W-CNF, (d–f) W-CNC and (g–i) B-CNC. The images are acquired from the AFM amplitude channel (a, left) and phase channel (a, right and d,g), Cryo-SEM (b,e,h) and TEM (c,f,i). The arrows point to regions along fibril or crystal contours, where it is possible to detect a right-handed twisting. The corresponding AFM height maps for the images a,d,g are shown in Supplementary Fig. 3. The colour bar in g also applies to a. 4 NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 ARTICLE 10 nm 100 GPa a b W-CNF –10 nm 0 GPa W-CNC B-CNC 1,500 nm c e 20 20 15 15 15 10 10 5 5 5 0 0 0 10 20 30 40 50 10 20 30 40 50 10 20 30 40 50 DMT modulus (GPa) DMT modulus (GPa) DMT modulus (GPa) g 80 100 GPa Kink1 Kink Kink2 Kink Kink2 0 GPa Kink Kink Kink1 0 0 100 200 300 400 500 100 nm Contour length (nm) Figure 3 | Peak force quantitative nanomechanical mapping. (a) AFM height channel visualizing the composition of all nanocellulose samples mixed in 1:1:1 proportion. The nanocellulose particles are distinguished and labelled according to their morphological features. (b) The corresponding PF-QNM map. The scale bar applies to both a,b panels. (c–e) DMT modulus distributions measured on particles representing (c) W-CNF, (d) W-CNC and (e) B-CNC. (f) PF-QNM images of W-CNFs and their profiles represented by red, blue and black lines. (g) Mechanical properties along the profiles of fibrils shown in f and the corresponding kink positions. particles. However, if the regions between the crystalline parts are angle distribution. In such an eventuality, by maintaining the randomly packed arrangements of moderately oriented chains— assumption of kink angles originating from amorphous domains, amorphous regions to a good approximation—we would expect a the final kink angle distribution would then be the weighted Gaussian distribution of the kink angles because of equal sum of individual (symmetric) Gaussian distributions reflecting probability to bend in both directions, and, thus, leading to a different angle probabilities for different fibril thicknesses, Gaussian distribution of the excess, that is, the observable kink resulting into a non-Gaussian, yet fully symmetric, final angle, as in general random walk statistics. We note that a distribution of kink angles. It is therefore the non-Gaussian and Gaussian distribution of bending angles is also expected on the non-symmetric nature of the distribution given in Fig. 4f, which basis of the worm-like chain formalism. It is clear that the allows ruling out amorphous domains as a cause of the presence observed peak in the kinks distribution at 60 does not support of kinks, independently of the cross-section distribution of the the first model, and the alternative model with less-ordered nanocellulose fibrils considered in the statistics. surface chains may possibly be more appropriate to describe the Additional support to the conclusions drawn from the kink cellulose polymer-chain arrangement in CNF. angle distribution analysis, ruling out the presence of ‘softer’ In order to estimate the influence of the contact with the amorphous zone regions alternating to stiffer crystalline ones, is substrate surface on the distribution of kink angles, the analogical given by the lack of significant changes in Young’s moduli analysis was conducted for images obtained using Cryo-SEM and measured using PF-QNM in correspondence of the kink regions, AFM on graphite substrate (Supplementary Fig. 4). The number which would be expected if these regions were amorphous (see of traced fibrils was 100 and 120 for Cryo-SEM and AFM on Fig. 3f,g). The crystalline core-amorphous shell model is also in 46–48 graphite correspondingly; hence, the statistics is 20 times less line with previous NMR studies , as well as the recent detailed accurate than the one reported in the main text, yet, the same experimental study that combined small-angle neutron scattering, conclusion is supported in both cases. wide-angle X-ray scattering, NMR and Fourier transform infrared We further note that the distribution of kink angles given in spectroscopy . Our results are therefore suggesting that the Fig. 4f accounts for kinks observed in fibrils of polydisperse observed kinks may result from the mechanical treatment during cross-sections. This may imply, in principle, that each individual the sample preparation and not from the presence of fibril can contribute to the final statistic with its own kink amorphous regions on the fibril contour. NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 5 & 2015 Macmillan Publishers Limited. All rights reserved. DMT modulus (GPa) % ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 a f 0 30 60 90 120 150 Kink angle,  (deg) 250 nm Kink area b c Angle of deviation (mask) (kink angle) Tracked contour 0 300 600 900 1,200 Contour length, L (nm) e Gaussian h × 10 distribution Obvious kinks Kink? Hypothetical amorphous 0 20° region 0123456 Height, h (nm) Figure 4 | Tracking procedure of W-CNF and results of statistical analysis on fibril contours. (a) AFM image with tracked contours (blue curves). Objects that are considered to be bundles of CNF were discarded from tracking (examples are pointed by white arrows). (b) Magnified region of the AFM image from a containing one tracked contour. The green frames are special masks, placed manually, allowing tracking of the contours with heterogeneous flexibility. In the statistical analysis these areas were considered to represent the vicinities of kinks. (c) Magnified region of the AFM image from b showing the particular mask area with the contour represented by its points. Two segments of the contour that enter or exit the kink area define an angle of contour deviation (kink angle) a.(d) Example of CNF with two obvious kinks and one uncertain case. It is non-trivial to state whether this is a kink or elastic bending due to thermal fluctuations. An example of the 20 angle is shown as a guide to eye. (e) The kink angle distribution should have a Gaussian shape in case of hypothetical amorphous regions corresponding to kinks because of equal probability to bend in both directions. (f) Kink angle distribution of 2,380 tracked CNFs. The bins in the region below 20 (violet line corresponds to this cutoff value) can lack some counts because of the manual threshold of kink angle assignments. (g) Contour length distribution fitted with normalized density function of the log-normal distribution with parameters m ¼ 5.94 and s ¼ 0.77. (h) Height distribution of all points along all tracked contours. The most probable height value is 2.35 nm. Many size measures in nature tend to have a log-normal fibrils. The broad height distribution of the W-CNF peak can be distribution, for example, the lengths of inert appendages (hair, attributed to the possibility of fibrils to split and thus have various 49 50 claws and nails) in biology , or the lengths of amyloid fibrils ; sizes and packing models. Some representative examples of here we show that the length of CNFs follows the same type of fibril-splitting are shown in Fig. 5. distribution. The log-normal distribution has the following probability density function f(L; m, s): Persistence length and second moment of area of the W-CNF. ðÞ lnðÞ L  m The persistence length l is an essential characteristic of polymers, pffiffiffiffiffi 2s fLðÞ ; m; s¼ e ð1Þ or in general fibrillar-like objects, and is directly related to the Ls 2p mechanical properties on a longitudinal inflection. It is formally where L is the total contour length of the fibrils, m and s are the defined as the length over which an angular correlation in the mean value and the s.d. of the length’s natural logarithm, tangent direction to a fibril contour is decreased on average by e respectively, and A is a distribution normalizing constant. The times in three-dimensional space . Following this definition, the length distribution of W-CNF is shown in Fig. 4g with a log- bond correlation function (BCF) is the most common way to normal distribution fit. Parameters of the fit are as follows: evaluate the persistence length and in two-dimensional space it 52  l/2l ± ± m ¼ 5.94 0.08 and s ¼ 0.77 0.06, which corresponds to the has the following form : hcosyi¼ e , where y is the angle average length hLi¼ 511 nm via the well-known relation between tangent directions of any two segments along a fibril m þ s =2 hi L ¼ e . contour. Another useful method of the persistence length The height distribution of W-CNF is shown in Fig. 4h. The estimation is to evaluate the mean-squared end-to-end distance height of the rigid segments does not differ from the height in the (MSED) between contour segments of a fibril, which for a vicinities of kinks (Supplementary Fig. 5) and is estimated to be in worm-like chain model has a following theoretical dependence the range h ¼ (1.9–2.7) nm, with the most probable height value on the internal contour length l in two-dimensional space : 2  l/2l hhi¼ 2.35 nm, which corresponds to the average diameter of hR i¼ 4l[l–2l(1–e )], where R is the direct distance between 6 NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. Number of fibers, n Number of kinks, n f k Number of points, n p NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 ARTICLE a b c 150 nm –5 nm 5 nm d e f 120 nm Figure 5 | Examples of splitting of W-CNF. The splitting event might occur at the end of W-CNF (a,c,d,f), as well as along the contour (b,e). The arrows point towards regions of interest. The images are obtained by (a–c) AFM and (d–f) Cryo-SEM. The scale bar in a and the colour bar in c apply to all AFM images (a–c). The scale bar in d applies to all Cryo-SEM images (d–f). any pair of segments along a fibril contour. A different method I (ref. 56); therefore, we neglect all other possible sources of errors that can be successfully applied to very stiff fibrillar-like objects, in this calculation. Because of the underestimation of the for which at internal contour length is smaller than the persistence length as a result of lost kink counts at low angles, corresponding persistence length (lol), is the mean-squared this value is also underestimated. Furthermore, nanocellulose, as midpoint displacement (MSMD). The equation, describing many other types of natural fibrils, is known to possess anisotropy the behaviour of an arc midpoint deviation, is derived with in the continuum mechanical property tensor and therefore, the the assumption that this deviation is small in comparison with the elastic behaviour in the axial and radial directions are expected corresponding internal contour length (|u |ool) and thus has the to differ. Thus, while in other natural fibrous systems, small 54 2 3 2 form u ¼ l =48l, where u is the midpoint mean-squared differences between the nanoindentation measurements and the x x displacement between any pair of segments along a fibril contour. results of statistical analysis of fibril bending may arise from Because the kinks of the W-CNF may originate from the slightly different elastic moduli on longitudinal bending E mechanical treatment, the deformation in these areas did go and on transversal bending E (refs 33,39), in the case of QNM beyond the elastic limit and, thus, cannot represent elastic nanocellulose, this deviation may become more significant . properties of the material on axial bending. We thus divide the W-CNFs using mask elements as split points and discard areas inside them to avoid kinks. Hence, in the resulting data, Observation of single cellulose polymer chains. In rare individual contours correspond only to rigid segments. The value occasions it is possible to observe very thin, possibly single of the persistence length obtained by the end-to-end distance cellulose polymer chains (Fig. 7a). This observation is in agree- versus internal contour length is l ¼ 2.84mm (Fig. 6). ment with previous observations of molecularly thin sections MSED 57,58 Supplementary Fig. 6 shows the analogue estimations of along CNF contours . In contrast to those findings, however, persistence length of W-CNF rigid segments via BCF and our data show fully flexible particles indicating that the observed MSMD. The resulting values are close: l ¼ 2.54mm and objects are single polymer chains rather than single crystalline BCF l ¼ 2.49mm. However, the relative errors and fitting quality layers of cellulose chains, as suggested in (ref. 58). A longitudinal MSMD provided by the end-to-end distance method is better than by the section is shown in Fig. 7b, while the height distribution of other two approaches (Supplementary Fig. 6b,e). We use only the 20 tracked fibrils is presented in Fig. 7c (average height data of the rigid segments in the Young’s modulus calculation, ± 0.44 0.15 nm). Owing to crossings with other thicker fibrils but because of the mentioned inability to identify all kinks, the and the polymer overlapping with itself, we set an upper limit for persistence length may be slightly underestimated. possible real height values at 1 nm. Similar objects could be The persistence length is related to the bending rigidity, EI,of detected also in the W-CNC sample (Supplementary Fig. 7a–c). the fibrillar-like object: EI ¼ lk T, where k is the Boltzmann The root mean-squared surface roughness of the (3-Aminopropyl) B B constant and T is the temperature . As we reported before, triethoxysilane-modified mica surface between cellulose particles the height and thus the diameter of the cellulose rigid is 0.19 nm, which explains the relatively high variation in the segments can be estimated to be h ¼ (1.9–2.7) nm. Here we longitudinal height profile. The obtained height value is in good assume a round cross-section for CNFs and, taking into account agreement with the dimensions of the cellulose chain within the errors on the height values, the area moment of inertia cellulose Ib crystal structure with a monoclinic unit cell and base 4 4 lays in the range I ¼ ph /64 ¼ (0.64–2.61) nm . Together with lattice parameters aD0.78 nm and bD0.82 nm (refs 10,26). 6 3 k T ¼ 4.14  10 Pa nm , the Young’s modulus is found to be Taking into account the most probable height value for the E ¼ lk T/I ¼ 4.5–18.4 GPa. The deviation from the average majority of the tracked W-CNFs (2.35 nm), this suggests that a l B height has the highest impact on the elastic modulus value E 4  4 cellulose chain-packing arrangement is the dominating due to the power 4 dependence of the area moment of inertia structure. However, the height distribution broadness of W-CNF NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 7 & 2015 Macmillan Publishers Limited. All rights reserved. Error, nm ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 × 10 a c 3,600 1.2 0.9 3,300 0.6 3,000 0.3 2,700 0 50 100 150 200 250 300 350 0100 200 300 Processing length, nm Contour length, L (nm) bd x 10 1.01 180 3 ≈ 2.84 μm MSED 1 120 2 0.99 60 1 0.98 0 0 45 90 135 180 50 100 150 200 250 300 350 Processing length, nm Internal contour length, l (nm) Figure 6 | Length distribution of W-CNF rigid segments and estimation of their persistence length. (a) Persistence length l calculated via the MSED MSED method versus the processing length. (b) Adjusted coefficient of determination (goodness of fit) R and fitting error versus the processing length. adj (c) Length distribution of W-CNF rigid segments. The red vertical lines in a–c correspond to the processing length 170 nm at which the fitting error is minimal. (d) MSED versus internal contour length fit at the distance with the minimal fitting error. The resulting persistence length is l ¼ 2.84mm. MSED –2 nm 2 nm 150 nm b e 1.5 1.5 1.2 1.2 0.9 0.9 0.6 0.6 0.3 0.3 0 0 0 300 600 900 1,200 0 100 200 300 Internal contour length, l (nm) Internal contour length, l (nm) c f 900 300 ⟨h⟩ = 0.44 nm ⟨h⟩ = 0.84 nm = 0.15 nm  = 0.17 nm h h 600 200 300 100 0 0 0 0.3 0.6 0.9 1.2 1.5 0 0.3 0.6 0.9 1.2 1.5 Height, h (nm) Height, h (nm) Figure 7 | Observation of single cellulose chain and nanofibril with a 2  2 chain-packing structure. (a–c) The single polymer chain and (d–f) nanofibril with polymer chains composed in a possible 2  2 chain-packing structure. (a,d) AFM images with the tracked contours represented by blue lines that are slightly shifted for better visualization. (b,e) Height profiles along the tracked contours from a,d, respectively. (c,f) Height distributions of points along 20 tracked single cellulose polymer chains with the average height h ¼ 0.44 nm and s.d. s ¼ 0.15 nm, and 2  2 cellulose nanofibril with the average height h ¼ 0.84 nm and s.d. s ¼ 0.17 nm. Cutoffs at 1 nm (b,c) and 1.5 nm (e,f) were introduced to discard height data from chain-fibril crossings. (Fig. 4h), together with evidence of splitting also supports the based on scattering measurements, was reported to contain presence of n  m structures, where n and m could possibly be 24 polymer chains and favoured a ‘rectangular’ model , which is equal to 3, 4, 5 and 6. The structure of cellulose fibrils in wood, close in numbers to what we have derived in this study. 8 NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. 2  , nm MSED adj Height, h (nm) Number of points, n Number of points, n Height, h (nm) MS end-to-end dist, Number of segments, n 2 2 ⟨R ⟩ (nm ) NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 ARTICLE Furthermore, in very rare occasions we detected thin W-CNF in of a Gaussian distribution of kink angles that the kinks present in which the cellulose polymer chains possibly packed in a 2  2 W-CNF are not a result of alternating amorphous and crystalline 64–66 chain-packing structure (Fig. 7d). The longitudinal section and domains, as previously proposed in literature . Rather the height distribution are shown in Fig. 7e,f. The average height they may originate from the processing conditions used in the for these objects is almost double in comparison with single preparation of the nanofibrils. PF-QNM was used to probe polymer chains and equals to 0.84 nm (Fig. 7f). By analogy, we set mechanical properties of the nanocellulose in the transversal up an upper limit at 1.5 nm for the possible height values. direction, showing values of Young’s moduli, E , in the QNM Supplementary Fig. 7d depicts both single and 2  2 cellulose range of 20–50 GPa. Using the persistence length l method, polymer chains forming a remarkable network with a well- Young’s moduli in the longitudinal direction were extracted defined entanglement centre. Such structures could possibly in the E ¼ 4.5–18.4 GPa range, although this range can be originate from peeling off single cellulose chains during the harsh underestimated because of the influence of undetected kinks with mechanical CNF preparation process. small angle of deviations. Finally, we have described the statistics Finally, we calculated the persistence length for single cellulose of single free cellulose polymer chains, detected in both W-CNF polymer chains adsorbed on mica via the BCF and the MSED and W-CNC and extracted a persistence length in the range methods. The MSMD method cannot be suitably used here since l ¼ 63–65 nm. The discrepancy between ‘bare’ persistence length the assumption that the persistence length should be much larger l ¼ 5–15 nm and the observable persistence length originates than the contour length—as we discussed before—is not justified from electrostatic repulsion and can be rationally explained by the in the present case. The resulting values are: l ¼ 65 nm OSF theory. BCF (Supplementary Fig. 8a–c) and l ¼ 63 nm (Supplementary MSED Fig. 8d–f), respectively. These values are much larger than the Methods theoretical and simulation predictions of the ‘bare’ persistence Preparation of TEMPO-mediated oxidized W-CNF. The TEMPO-mediated oxidation of wood pulp was performed according to the procedure introduced in length, reported to be in the range of 5–15 nm (ref. 59). We ref. 67. The soft-wood pulp was first treated in a phosphate buffer at pH 6.8 at explain this discrepancy with the electrostatic repulsion between 60 C. The desired amount of sodium chlorite, TEMPO and sodium hypochlorite charged monomers along polymer chains and employ the was added and the dispersion was stirred for 2 h and 20 min, after which the pulp 60,61 Odijk–Skolnick–Fixman (OSF) theory to rationalize this was washed with deionized water and collected with vacuum filtration. The TEMPO-mediated oxidized material was dispersed in deionized water and observation. Within the framework of the OSF theory the total disintegrated by homogenization with a Microfluidizer M-110 EH (Microfluidics, persistence length can be represented as a sum of ‘bare’ (l ) USA), in a sequence consisting of four passes through the microchannels with and electrostatic (l ) components: l ¼ l þ l , where OSF 0 OSF diameters of 400–200mm at a pressure of 900 bar and for four passes through 2 2 l ¼ l r =4A (ref. 62), r is the Debye length, l ¼ 0.71 nm OSF B D B the microchannels with diameters of 200–100mm at 1,500 bar. The resulting suspension of TEMPO-mediated W-CNFs was first sonicated for 10 min using a is the Bjerrum length and A is the distance between two 13-mm-wide titanium probe at an output power of 70% (Vibra-Cell VC 750, neighbouring charged units. The charge density of 0.4 mmol g Sonics, USA), followed by centrifugation for 60 min at 4000 g. The W-CNF dis- was measured at pH 10, but the final value, at which the AFM persion was diluted to 0.001 w/w% with MilliQ water and shaken thoroughly. samples were prepared, was pH 5.65. By taking this pH change into account as well as the contribution from the counterions Preparation of W-CNC. W-CNCs were prepared by hydrochloric acid hydrolysis (I ¼ 4  10 M), the ionic strength is estimated to cI of the TEMPO-mediated oxidized CNFs (W-CNF) according to the procedure 6 63 I ¼ 6  10 M. Using the previously established approach we described in (ref. 32). Hydrochloric acid to a final concentration of 2.5 M was added to a dispersion of 100 g of a 1 w/w% (dry weight basis) W-CNF gel diluted calculate the degree of dissociation b ¼ 0.08, which leads to an with 316 ml deionized water and heated to 105 C for 6 h. The reaction was effective charge density of 0.032 mmol g . Assuming the quenched by dilution with the fivefold amount of deionized water. The hydrolysed distribution of charges only on the surface of particles, with material was washed with deionized water, collected by centrifugation for 10 min at average dimensions obtained from the statistical analysis, an 4000 g and dialysed for 5 days against deionized water using Sigma-Aldrich dialysis average distance between charges of AD7.7 nm is found. Under membranes with a molecular weight cutoff of B14,000 Da. After the dialysis, the suspension was first sonicated for 10 min using a 13-mm-wide titanium probe with these conditions of ionic strength, the Debye length r can be an output of 70% (Vibra-Cell VC 750, Sonics), followed by a centrifugation for estimated to be r D122 nm, and the resulting l D44.7 nm, D OSF 10 min at 4000 g. The deagglomerated dispersion was diluted to 0.001 w/w% with leading to a total expected persistence length of B50–60 nm, MilliQ water and shaken thoroughly. which is in excellent agreement with the experimental results (l ¼ 65 nm; l ¼ 63 nm), providing a solid basis for the BC MSED Preparation of B-CNC. B-CNCs were prepared from commercially available understanding of the large observed persistence length of coconut gel cubes (Chaokoh, Thailand). The coconut cubes (with a size of 3 3 individual nanocellulose chains. In order to test whether B1  1  1cm ) were pretreated by first washing for three times with 2 dm of deionized water, followed by stirring in 2 dm of a 0.1 M sodium hydroxide counterion condensation plays a role in the final observable solution for 48 h and finally washing with deionized water until the pH stabilized at persistence length, we perform a quick Oosawa–Manning B7. The hydrolysis was performed by soaking 100 g of the pretreated coconut counterion condensation check. The condensation of cubes in sulfuric acid with a concentration of 40 w/w% at 80 C for 4 h. The counterions occurs when l r41, where r is the linear charge hydrolysed materials were washed twice with deionized water, collected by centrifugation and dialysed for 5 days against deionized water using Sigma-Aldrich density (r ¼ 1/A ¼ 0.13 nm ). In the present case l rE0.09, dialysis membranes with a molecular weight cutoff of B14,000 Da. After the suggesting that counterion condensation can be discarded. dialysis, the suspension was first sonicated for 10 min using a 13-mm-wide titanium probe (Vibra-Cell VC 750, Sonics), at an output power of 70%, followed by centrifugation for 60 min at 4000 g. The deagglomerated dispersion was diluted to 0.001 w/w% with MilliQ water and shaken thoroughly. Discussion In the present work we provide a comprehensive and consistent structural description over multiple length scales of nanocellulose Surface charge determination. The surface charge of W-CNC and B-CNC was determined with polyelectrolyte titration using a Stabino Particle Charge Mapping of different origin and pretreatment: TEMPO-mediated oxidized system (Microtrac Europe GmbH, Germany). The nanocellulose dispersions were W-CNF, W-CNC and B-CNC. We find that all types of diluted in MilliQ water and then titrated with a 0.001 N polydiallyl dimethyl nanocellulose fibrils and crystals with an observable twisting ammonium chloride solution. An average charge density was obtained from three along the contour possess a right-handed chirality. The right- or more measurements. handed chirality was detected on the level of fibril bundles in all systems and on the single fibril level in W-CNF. By a statistical AFM and PF-QNM measurements. A droplet of 0.001 w/w% solution of analysis of kink angle distribution, we conclude from the absence different types of cellulose was deposited on chemically modified mica with NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 9 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 (3-Aminopropyl)triethoxysilane following a protocol described in ref. 41. AFM and 5. Revol, J. F., Bradford, H., Giasson, J., Marchessault, R. H. & Gray, D. G. PF-QNM measurements were performed by using a MultiMode VIII Scanning Helicoidal self-ordering of cellulose microfibrils in aqueous suspension. Int. J. Probe Microscope (Bruker, USA) covered with an acoustic hood to minimize Biol. Macromol. 14, 170–172 (1992). vibrational noise. AFM images were acquired continuously in the tapping mode 6. Hamedi, M. et al. Nanocellulose aerogels functionalized by rapid layer-by-layer under ambient conditions using commercial cantilevers (Bruker). In order to assembly for high charge storage and beyond. Angew. Chem. Int. Ed. 52, perform PF-QNM measurements on all cellulose nanoparticles using the same 12038–12042 (2013). substrate and under identical environmental and AFM tip conditions, small 7. Shopsowitz, K. E., Qi, H., Hamad, W. Y. & Maclachlan, M. J. Free-standing aliquots of 0.001 w/w% dispersions of W-CNF, W-CNC and B-CNC were mixed in mesoporous silica films with tunable chiral nematic structures. Nature 468, proportion 1:1:1 before deposition on mica surfaces. The AFM cantilevers (Bruker) 422–425 (2010). were calibrated on the calibration samples (Bruker)—typically low-density 8. Olsson, R. T. et al. Making flexible magnetic aerogels and stiff magnetic polyethylene and polystyrene—covering the following ranges of Young’s moduli: nanopaper using cellulose nanofibrils as templates. Nat. Nanotechnol. 5, from 100 MPa to 2 GPa (for low-density polyethylene) and from 1 to 20 GPa (for 584–588 (2010). polystyrene). The measurements were performed on small indentation depths to 9. Moon, R. J., Martini, A., Nairn, J., Simonsen, J. & Youngblood, J. Cellulose avoid any artefacts of substrate. The analysis of the elastic modulus was performed nanomaterials review: structure, properties and nanocomposites. Chem. Soc. using the Nanoscope Analysis software and calculated according the DMT model. Rev. 40, 3941–3994 (2011). 10. Nishiyama, Y., Langan, P. & Chanzy, H. Crystal structure and hydrogen- Transmission electron microscopy. TEM imaging was carried out on carbon- bonding system in cellulose Ib from synchrotron X-ray and neutron fiber coated copper grids that were glow-discharged for 45 s (Emitech K100X, GB) diffraction. J. Am. Chem. Soc. 124, 9074–9082 (2002). directly before sample fixation. Sample grid preparation for negative staining was 11. Majoinen, J., Kontturi, E., Ikkala, O. & Gray, D. G. SEM imaging of chiral as follows: 5ml of sample dispersion for 1 min, 5ml of 2% uranyl acetate for 1 s and nematic films cast from cellulose nanocrystal suspensions. Cellulose 19, again 5ml of 2% uranyl acetate for 15 s to achieve a noncrystalline film of stain 1599–1605 (2012). embedding the fibres. Following each step, the excess moisture was drained along 12. Kelly, J. A. et al. Evaluation of form birefringence in chiral nematic mesoporous the periphery using a piece of filter paper. Dried grids were examined using TEM materials. J. Mater. Chem. C 2, 5093–5097 (2014). (FEI, model Morgagni, NL) operated at 100 kV. 13. Lagerwall, J. P. F. et al. Cellulose nanocrystal-based materials: from liquid crystal self-assembly and glass formation to multifunctional thin films. NPG Asia Mater. 6, e80 (2014). Cryo-SEM. Sample aliquots (3.5ml) were applied to glow-discharged, carbon- 14. Saito, T., Uematsu, T., Kimura, S., Enomaea, T. & Isogai, A. Self-aligned coated Cu-grids for 1 min, blotted with filter paper along the periphery and plunge- integration of native cellulose nanofibrils towards producing diverse bulk frozen in liquid ethane. Vitrified specimens were then transferred and mounted materials. Soft Matter 7, 8804–8809 (2011). under liquid nitrogen on a self-made grid holder and finally transferred under liquid nitrogen into a precooled (  120 C) freeze-fracturing system BAF 060 (Bal- 15. Kobayashi, Y., Saito, T. & Isogai, A. Aerogels with 3D ordered nanofiber Tec/Leica, Vienna). For freeze-drying the samples were warmed up in 5 C skeletons of liquid-crystalline nanocellulose derivatives as tough and increments every 15 min until  80 C was reached at 10 mbar. Coating was transparent insulators. Angew. Chem. Int. Ed. 53, 10394–10397 (2014). performed with 1.5 nm tungsten at 45 followed by 1.5 nm under continuous 16. Wicklein, B. et al. Thermally insulating and fire-retardant lightweight elevation angle changes from 45 to 90 and back to 45. Cryo-SEM was anisotropic foams based on nanocellulose and graphene oxide. Nat. performed in a field emission SEM Leo Gemini 1530 (Carl Zeiss, Germany) Nanotechnol. 10, 277–283 (2015). equipped with a cold stage to maintain the specimen temperature at  110 C 17. Beck-Candanedo, S., Roman, M. & Gray, D. G. Effect of reaction conditions on (VCT Cryostage, Bal-Tec/Leica). Signals from the SE-inlens detector (acceleration the properties and behavior of wood cellulose nanocrystal suspensions. voltage 5 kV) were used for image formation. Only the contrast and brightness Biomacromolecules 6, 1048–1054 (2005). of the pictures were adjusted. 18. Shinoda, R., Saito, T., Okita, Y. & Isogai, A. Relationship between length and degree of polymerization of TEMPO-oxidized cellulose nanofibrils. Biomacromolecules 13, 842–849 (2012). Tracking of the TEMPO-oxidized W-CNFs. The coordinates of the CNFs 19. Saito, T., Kuramae, R., Wohlert, J., Berglund, L. A. & Isogai, A. An ultrastrong (W-CNF) were obtained using an in-house programme written in MATLAB. Each nanofibrillar biomaterial: the strength of single cellulose nanofibrils revealed via tracked CNF can be represented by its contour—a sequence of points connected sonication-induced fragmentation. Biomacromolecules 14, 248–253 (2013). with straight segments that are positioned along the fibril bright ridge on an AFM 20. Fernandes, A. N. et al. Nanostructure of cellulose microfibrils in spruce wood. image. All contours acquired in this study have a constant distance between pro- jections of these points on the image plane, which is the step size s ¼ 2.9 nm. This Proc. Natl Acad. Sci. USA 108, E1195–E1203 (2011). 21. Orts, W. J., Godbout, L., Marchessault, R. H. & Revol, J.-F. Enhanced ordering way of tracking is similar to the procedure we previously applied for amyloid fibril 33,40 systems with one particular addition. Owing to significant directional variation of liquid crystalline suspensions of cellulose microfibrils: a small angle neutron of W-CNF contours (low curvature along rigid segments, high curvature in scattering study. Macromolecules 31, 5717–5725 (1998). vicinities of kinks) we employed a concept of masks that define kink areas. The 22. Hanley, S., Revol, J., Godbout, L. & Gray, D. Atomic force microscopy and affinity of a contour to bend is different depending on whether points are inside or transmission electron microscopy of cellulose from Micrasterias denticulata; outside the mask area; in other words, it allows contours to have heterogeneous evidence for a chiral helical microfibril twist. Cellulose 4, 209–220 (1997). stiffness (large stiffness along rigid segments and low stiffness in vicinities of kinks). 23. Khandelwal, M. & Windle, A. Origin of chiral interactions in cellulose This is essential for the tracking algorithm to correctly follow the actual features of supra-molecular microfibrils. Carbohydr. Polym. 106, 128–131 (2014). nanocellulose fibrils. The position of these masks for each object and the initial 24. Lahiji, R. R. et al. Atomic force microscopy characterization of cellulose contours were initialized manually, the latter with the auxiliary help of the A* nanocrystals. Langmuir 26, 4480–4488 (2010). pathfinding algorithm . To obtain subpixel accuracy, contours were deformed and 25. Postek, M. T. et al. Development of the metrology and imaging of cellulose fitted precisely to the cellulose fibrils middle lines using the slightly modified Open nanocrystals. Meas. Sci. Technol. 22, 024005 (2011). 69,70 Active Contours algorithm . All statistical information was acquired from high- 26. Matthews, J. F. et al. Computer simulation studies of microcrystalline cellulose resolution AFM images with spatial dimensions of 15  15mm and 5,120  5,120 Ib. Carbohydr. Res. 341, 138–152 (2006). pixels. In total, data on 2,380 fibrils from W-CNF were extracted to reach sufficient 27. Paavilainen, S., Rog, T. & Vattulainen, I. Analysis of twisting of cellulose statistical significance. All data-processing methods used in this work such as the nanofibrils in atomistic molecular dynamics simulations. J. Phys. Chem. B 115, calculation of the kink angle, length and height distributions, persistence length 3747–3755 (2011). evaluation and plotting fibril height, profiles were performed in the same software. 28. Tashiro, K. & Kobayashi, M. Theoretical evaluation of three-dimensional elastic constants of native and regenerated celluloses: role of hydrogen bonds. Polymer (Guildf) 32, 1516–1526 (1991). References 29. Lindman, B., Karlstro¨m, G. & Stigsson, L. On the mechanism of dissolution of 1. Godinho, M. H., Canejo, J. P., Pinto, L. F. V., Borges, J. P. & Teixeira, P. I. C. cellulose. J. Mol. Liq. 156, 76–81 (2010). How to mimic the shapes of plant tendrils on the nano and microscale: spirals 30. Klemm, D. et al. Nanocelluloses: a new family of nature-based materials. and helices of electrospun liquid crystalline cellulose derivatives. Soft Matter 5, 2772–2776 (2009). Angew. Chem. Int. Ed. 50, 5438–5466 (2011). 2. Gray, D. G. Isolation and handedness of helical coiled cellulosic thickenings 31. Siro, I. & Plackett, D. Microfibrillated cellulose and new nanocomposite materials: a review. Cellulose 17, 459–494 (2010). from plant petiole tracheary elements. Cellulose 21, 3181–3191 (2014). 32. Salajkova´, M., Berglund, L. A. & Zhou, Q. Hydrophobic cellulose nanocrystals 3. Håkansson, K. M. O. et al. Hydrodynamic alignment and assembly of nanofibrils resulting in strong cellulose filaments. Nat. Commun. 5, 4018 modified with quaternary ammonium salts. J. Mater. Chem. 22, 19798–19805 (2014). (2012). 4. Kelly, J. A., Giese, M., Shopsowitz, K. E., Hamad, W. Y. & MacLachlan, M. J. 33. Usov, I., Adamcik, J. & Mezzenga, R. Polymorphism complexity and The development of chiral nematic mesoporous materials. Acc. Chem. Res. 47, handedness inversion in serum albumin amyloid fibrils. ACS Nano 7, 1088–1096 (2014). 10465–10474 (2013). 10 NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8564 ARTICLE 34. Gray, D. G. Chiral nematic ordering of polysaccharides. Carbohydr. Polym. 25, 61. Skolnick, J. & Fixman, M. Electrostatic persistence length of a wormlike 277–284 (1994). polyelectrolyte. Macromolecules 10, 944–948 (1977). 35. Belli, S., Dussi, S., Dijkstra, M. & van Roij, R. Density functional theory for 62. Dobrynin, A. V. & Rubinstein, M. Theory of polyelectrolytes in solutions and at chiral nematic liquid crystals. Phys. Rev. E 90, 020503 (2014). surfaces. Prog. Polym. Sci. 30, 1049–1118 (2005). 36. Adamcik, J., Berquand, A. & Mezzenga, R. Single-step direct measurement of 63. Fall, A. B., Lindstro¨m, S. B., Sundman, O., Odberg, L. & Wågberg, L. Colloidal amyloid fibrils stiffness by peak force quantitative nanomechanical atomic force stability of aqueous nanofibrillated cellulose dispersions. Langmuir 27, microscopy. Appl. Phys. Lett. 98, 193701 (2011). 11332–11338 (2011). 37. Adamcik, J. et al. Measurement of intrinsic properties of amyloid fibrils by the 64. Montanari, S., Roumani, M., Heux, L. & Vignon, M. R. Topochemistry of peak force QNM method. Nanoscale 4, 4426–4429 (2012). carboxylated cellulose nanocrystals resulting from TEMPO-mediated oxidation. 38. Ling, S. et al. Modulating materials by orthogonally oriented b-strands: Macromolecules 38, 1665–1671 (2005). composites of amyloid and silk fibroin fibrils. Adv. Mater. 26, 4569–4574 (2014). 65. Siqueira, G., Tapin-Lingua, S., Bras, J., da Silva Perez, D. & Dufresne, A. 39. Lamour, G., Yip, C. K., Li, H. & Gsponer, J. High intrinsic mechanical flexibility Morphological investigation of nanoparticles obtained from combined of mouse prion nanofibrils revealed by measurements of axial and radial mechanical shearing, and enzymatic and acid hydrolysis of sisal fibers. Cellulose Young’s moduli. ACS Nano 8, 3851–3861 (2014). 17, 1147–1158 (2010). 40. Jordens, S., Isa, L., Usov, I. & Mezzenga, R. Non-equilibrium nature of 66. Hamedi, M. M. et al. Highly conducting, strong nanocomposites based on two-dimensional isotropic and nematic coexistence in amyloid fibrils at liquid nanocellulose-assisted aqueous dispersions of single-wall carbon nanotubes. interfaces. Nat. Commun. 4, 1917 (2013). ACS Nano 8, 2467–2476 (2014). 41. Schefer, L., Adamcik, J. & Mezzenga, R. Unravelling secondary structure 67. Saito, T. et al. Individualization of nano-sized plant cellulose fibrils by direct changes on individual anionic polysaccharide chains by atomic force surface carboxylation using TEMPO catalyst under neutral conditions. microscopy. Angew. Chem. Int. Ed. 53, 5376–5379 (2014). Biomacromolecules 10, 1992–1996 (2009). 42. Li, C. & Mezzenga, R. Functionalization of multiwalled carbon nanotubes and 68. Hart, P., Nilsson, N. & Raphael, B. A formal basis for the heuristic their pH-responsive hydrogels with amyloid fibrils. Langmuir 28, 10142–10146 determination of minimum cost paths. IEEE Trans. Syst. Sci. Cybern. 4, (2012). 100–107 (1968). 69. Kass, M., Witkin, A. & Terzopoulos, D. Snakes: active contour models. Int. J. 43. Usov, I. & Mezzenga, R. FiberApp: an open-source software for tracking and analyzing polymers, filaments, biomacromolecules, and fibrous objects. Comput. Vis. 1, 321–331 (1988). Macromolecules 48, 1269–1280 (2015). 70. Smith, M. B. et al. Segmentation and tracking of cytoskeletal filaments using 44. Nishiyama, Y. Structure and properties of the cellulose microfibril. J. Wood Sci. open active contours. Cytoskeleton (Hoboken) 67, 693–705 (2010). 55, 241–249 (2009). 45. Battista, O. Hydrolysis and crystallization of cellulose. Ind. Eng. Chem. 42, Acknowledgements 502–507 (1950). We acknowledge support from the Swiss National Science Foundation (SNF; 2-77002-11) 46. Wickholm, K., Larsson, P. & Iversen, T. Assignment of non-crystalline forms in and by the Scientific Center for Optical and Electron Microscoy of ETH Zurich cellulose I by CP/MAS 13 C NMR spectroscopy. Carbohydr. Res. 312, 123–129 (ScopeM). G.N. acknowledges funding from the Gunnar Sundblad Research Foundation. (1998). C.S. and L.B. acknowledge the Wallenberg Wood Science Center for funding. Korneliya 47. Sturcova, A., His, I., Apperley, D. C., Sugiyama, J. & Jarvis, M. C. Structural Gordeyeva is thanked for technical assistance. Lars Wågberg and Nicholas Tchang details of crystalline cellulose from higher plants. Biomacromolecules 5, Cervin are thanked for providing the TEMPO-mediated oxidized Domsjo¨ (60/40 1333–1339 (2004). spruce/pine) wood cellulose pulp. 48. Newman, R. H. Estimation of the lateral dimensions of cellulose crystallites using 13C NMR signal strengths. Solid State Nucl. Magn. Reson. 15, 21–29 (1999). Author contributions 49. McGeoch, C. C. A Guide to Experimental Algorithmics (Cambridge University I.U. designed the cellulose tracking and data-processing methods, performed Press, 2012). statistical analysis and wrote the manuscript. G.N. contributed to the data analysis 50. Usov, I., Adamcik, J. & Mezzenga, R. Polymorphism in bovine serum albumin and writing of the manuscript. J.A. performed AFM and PF-QNM experiments. S.H. fibrils: morphology and statistical analysis. Faraday Discuss. 166, 151–162 performed Cryo-SEM and TEM experiments. C.S. carried out the nanocellulose (2013). preparation. A.F. contributed to the data analysis. L.B. contributed to the design of 51. Rubinstein, M. & Colby, R. H. Polymer Physics (Oxford University Press, 2003). the study, interpretation of results and wrote the manuscript. R.M. contributed to the 52. Doi, M. & Edwards, S. F. The Theory of Polymer Dynamics (Oxford University data analysis, result interpretation, designed and directed the study and wrote the Press, 1986). manuscript. 53. Rivetti, C., Guthold, M. & Bustamante, C. Scanning force microscopy of DNA deposited onto mica: equilibration versus kinetic trapping studied by statistical Additional information polymer chain analysis. J. Mol. Biol. 264, 919–932 (1996). Supplementary Information accompanies this paper at http://www.nature.com/ 54. Smith, J. F., Knowles, T. P. J., Dobson, C. M., Macphee, C. E. & Welland, M. E. naturecommunications Characterization of the nanoscale properties of individual amyloid fibrils. Proc. Natl Acad. Sci. USA. 103, 15806–15811 (2006). Competing financial interests: The authors declare no competing financial interests. 55. Manning, G. S. Polymer persistence length characterized as a critical length for instability caused by a fluctuating twist. Phys. Rev. A 34, 668–670 (1986). Reprints and permission information is available online at http://npg.nature.com/ 56. Usov, I. & Mezzenga, R. Correlation between nanomechanics and polymorphic reprintsandpermissions/ conformations in amyloid fibrils. ACS Nano 8, 11035–11041 (2014). How to cite this article: Usov, I. et al. Understanding nanocellulose chirality and 57. Li, Q. & Renneckar, S. Molecularly thin nanoparticles from cellulose: isolation structure–properties relationship at the single fibril level. Nat. Commun. 6:7564 of sub-microfibrillar structures. Cellulose 16, 1025–1032 (2009). doi: 10.1038/ncomms8564 (2015). 58. Li, Q. & Renneckar, S. Supramolecular structure characterization of molecularly thin cellulose I nanoparticles. Biomacromolecules 12, 650–659 (2011). This work is licensed under a Creative Commons Attribution 4.0 59. Kroon-Batenburg, L. M. J., Kruiskamp, P. H., Vliegenthart, J. F. G. & Kroon, J. International License. The images or other third party material in this Estimation of the persistence length of polymers by MD simulations on small fragments in solution. application to cellulose. J. Phys. Chem. B 101, 8454–8459 article are included in the article’s Creative Commons license, unless indicated otherwise (1997). in the credit line; if the material is not included under the Creative Commons license, 60. Odijk, T. Polyelectrolytes near the rod limit. J. Polym. Sci. Polym. Phys. Ed. 15, users will need to obtain permission from the license holder to reproduce the material. 477–483 (1977). To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ NATURE COMMUNICATIONS | 6:7564 | DOI: 10.1038/ncomms8564 | www.nature.com/naturecommunications 11 & 2015 Macmillan Publishers Limited. All rights reserved.

Journal

Nature Communications – Springer Journals

Published: Jun 25, 2015

There are no references for this article.