Comprehensive morphological and structural investigation of cellulose I and II nanocrystals prepared by sulphuric acid hydrolysis

Abstract

Cellulose nanocrystals (CNCs) were produced from eucalyptus wood pulp using three different methods: (i) classical sulphuric acid hydrolysis (CN-I), (ii) acid hydrolysis of cellulose previously mercerized by alkaline treatment (MCN-II), and (iii) solubilization of cellulose in sulphuric acid and subsequent recrystallization in water (RCN-II). The three types of CNCs exhibited different morphologies and crystalline structures that were characterized using complementary imaging, diffraction and spectroscopic techniques. CN-I corresponded to the type I allomorph of cellulose while MCN-II and RCN-II corresponded to cellulose II. CN-I and MCN-II CNCs were acicular particles composed of a few laterally-bound elementary crystallites. In both cases, the cellulose chains were oriented parallel to the long axis of the particle, although they were parallel in CN-I and antiparallel in MCN-II. RCN-II particles exhibited a slightly tortuous ribbon-like shape and it was shown that the chains lay perpendicular to the particle long axis and parallel to their basal plane. The unique molecular and crystal structure of the RCN-II particles implies that a higher number of reducing chain ends are located at the surface of the particles, which may be important for subsequent chemical modification. While other authors have described nanoparticles prepared by regeneration of short-chain cellulose solutions, no detailed description was proposed in terms of particle morphology, crystal structure and chain orientation. We provide such a description in the present paper.

---

Introduction

Cellulose is a linear biopolymer consisting of β-(1,4)-linked D-glucopyranose units, occurring in nature in the form of slender semicrystalline nanofibrils where highly ordered regions (the crystallites) alternate with less organized ones.¹ In this so-called cellulose I (Cel-I), the chains are parallel and can be packed into two allomorphs, namely I_α (triclinic) and I_β (monoclinic). Most native celluloses contain a mixture of both allomorphs in various ratios.² Several other allomorphs have been described, most of which result from chemical treatments of Cel-I. These allomorphs differ by the packing and polarity of the constituting chains, as well as the hydrogen bond patterns established between them.³

Among these allomorphs, cellulose II (Cel-II), in which the chains are antiparallel,^4–6 can be prepared from Cel-I by two distinct processes. Mercerization is an essentially solid-state process during which cellulose fibres are swollen in alkali and recrystallized into Cel-II upon washing.⁷ An interdigitation mechanism was proposed to account for the fibrous transition from Cel-I to Cel-II: upon swelling, the cellulose chains from neighbouring nanofibrils of opposite polarity would intermingle and recrystallize into the antiparallel form upon washing and drying.^8,9 In the second process, cellulose is first dissolved and then regenerated by recrystallizing the chains into Cel-II upon washing.¹⁰ The transition is irreversible, which suggests that Cel-II is thermodynamically more stable.

From an industrial point of view, Cel-II is used to prepare materials such as films (cellophane), textile fibres (rayon, lyocell), membranes for dialysis, filtration and chromatography (cuprophan), papers (for packaging and printing purpose), pharmaceutical pellets (for drug release) and cellulose derivatives (such as methylcellulose and carboxymethylcellulose).^11–13

Due to the combination of unique physicochemical properties and environmental appeal, cellulose nanocrystals (CNCs) have enjoyed a large interest in the recent years.^14,15 CNCs are usually prepared by strong sulphuric acid hydrolysis that preferentially degrades the disordered regions of the native cellulose nanofibrils. An additional mechanical or sonication treatment releases rod-like nanocrystals that form stable colloidal suspensions.^16,17 In addition, the shape and size of CNCs depend on the source of cellulose.^15,18,19 CNCs offer a wide range of potential applications in fields ranging from packaging to biomedicine. Several unique properties, such as high specific strength and modulus, high aspect ratio, low density, large specific surface area and reactive surfaces that can facilitate the grafting of chemical species to confer and tailor new properties, stimulate their use in new functional and advanced materials.^15,20–23

So far, investigations of Cel-II have mostly focused on fibres and only a few recent studies have been carried out on CNCs. Cel-II nanocrystals have been prepared either by acid hydrolysis of mercerized fibres,^24–26 mercerization of Cel-I CNCs,²⁷ or after recrystallization of fractions of short cellulose chains in solution.^24,28–30 However, while these studies have generally combined the data from several imaging, diffraction and spectroscopic techniques, a complete structural picture of the nanocrystals has not been reported so far. The purpose of our work was thus to produce and characterize cellulose nanocrystals prepared from native eucalyptus Cel-I fibres by (i) sulphuric acid hydrolysis, (ii) acid hydrolysis of fibrous Cel-II obtained by mercerization of Cel-I, and (iii) recrystallization (regeneration) of cellulose chains produced by stronger acid hydrolysis of native cellulose. We compared the morphology, crystal structure, crystallinity index, surface charge, and degree of polymerization of the three types of nanocrystals that were characterized by complementary techniques, namely elemental analysis, zetametry, viscometry, transmission electron microscopy (TEM), atomic force microscopy (AFM), X-ray diffraction (XRD), Fourier-transform infrared and solid-state nuclear magnetic resonance spectroscopies (FTIR and NMR, respectively).

Experimental

Materials

Sulphuric acid (95.0–98.0 wt%, Vetec, P.A.), bis(ethylenediamine) copper(II) hydroxide solution 1.0 M in H₂O (Panreac) and cellulose membrane (Spectra/Por 1 Dialysis Membrane) were used as received. Bleached Kraft wood pulp sheets of eucalyptus urograndis hybrid (Eucalyptus urophylla and grandis) were provided by Conpacel Company (Limeira, São Paulo, Brazil). This pulp was chosen as raw material because it has a high cellulose content. Its chemical composition was determined as described elsewhere:²² 92% cellulose, 6% hemicelluloses, 0% lignin and 0.3% minerals.

Preparation of cellulose nanocrystals

Nanocrystals from native cellulose. The bleached Kraft eucalyptus wood pulp (WP) was ground in a Fourplex mill and then hydrolysed with sulphuric acid. The hydrolysis was performed at 45 °C for 50 min under vigorous stirring. For each gram of WP, we used 20 mL of 60 wt% H₂SO₄. Immediately following the hydrolysis, the material (an ivory-white dispersion) was diluted 10-fold with cold deionized water to stop the hydrolysis reaction, and centrifuged twice for 10 min at 10 000 rpm to remove the excess acid. The precipitate was then dialyzed against deionized water to remove non-reactive sulphate groups, salts and soluble sugars, until a neutral pH was reached. The resulting dialyzed suspension was ultrasonicated for 15 min in an ice bath (Branson sonifier 250 operating at 50 duty cycle and 40% output) and stored in a refrigerator. Drops of sodium hypochlorite were added to avoid bacterial growth. The resulting nanocrystals will be referred to as “CN-I” in the following.

Nanocrystals from mercerized cellulose. The milled WP was treated with a 20 wt% sodium hydroxide aqueous solution under mechanical stirring for 5 h at 25 °C, using 30 mL of solution per gram of material. The material was washed several times with distilled water until the alkali was completely removed, and dried at 40 °C for 12 h in an air-circulating oven. The mercerized pulp (MWP) was milled again and submitted to sulphuric acid hydrolysis, using the same conditions as those applied for the preparation of CN-I. This sample will be referred to as “MCN-II”.

Nanocrystals from regenerated cellulose. The hydrolysis of ground eucalyptus WP was carried out with 64 wt% H₂SO₄ at 40 °C for 20 min. The subsequent steps of dilution with water, centrifugation, and dialysis were carried out using the same conditions as those used for the preparation of CN-I. Following hydrolysis, the solution was transparent and upon addition of water, it became an ivory-white dispersion. Under these hydrolysis conditions, the cellulose fibres were thus solubilized and the soluble chains precipitated by addition of water, which corresponds to the so-called regeneration process. The resulting nanocrystals will be referred to as “RCN-II”.

CNC powders

CNC powders of the three samples were prepared by freeze-drying 0.5 wt% suspensions.

Shear-oriented CNC fibres

Oriented fibres were prepared from the CNC suspensions using the method described by Nishiyama et al.³¹ Briefly, the suspensions were centrifuged and the pellets were mixed with polyvinyl alcohol (PVA). A saturated solution of sodium borate (borax) was added to cross-link PVA and yield a viscoelastic gel that was uniaxially stretched with tweezers to form a fibre and left to dry under load.

Characterization

Elemental analysis. The total sulphur content S% of the samples was determined by elemental analysis using an Element Analyser (Perkin-Elmer Series II, CHNS/O Analyzer 2400) from the freeze-dried CNC powders. The values were converted to the number n of –OSO₃H groups of the CNC samples per 100 anhydroglucose units, using the method described in details by Hu et al.²⁹

Zeta potential. The zeta-potential ζ of 0.01 wt% CNC suspensions was measured on a Malvern DTS0230 instrument. All suspensions were diluted in an aqueous standard solution at pH 7, 5 mmol ionic strength and 200 μS cm⁻¹ conductivity, prepared by the addition of diluted NaOH solution and solid NaCl into deionized water.

Viscometry. The degree of polymerization (DP) of CNC samples was calculated according to TAPPI T 230 OM-99 standard,³² from the value of intrinsic viscosity [η] of the samples. The [η] value was obtained by measuring the flow of solvent and solutions in an Ostwald's viscometer. The solvent used was a mixture of a solution of bis(ethylenediamine) copper(II) hydroxide and deionized water. The intrinsic viscosity [η] was determined from eqn (1).where η_rel is the relative viscosity given by the ratio of the flow time of the CNC suspension and the flow time of the solvent, η_sp is the specific viscosity given by (η_rel − 1), and C is the concentration of the CNC suspension in g mL⁻¹. The DP was calculated using the equation recommended by the standard method SCAN-C15:62:³³

DP^0.905 = 0.75[η] (2)

where the values of 0.905 and 0.75 are constants characteristic of the polymer–solvent system and [η] is expressed in mL g⁻¹.

Scanning electron microscopy (SEM). Fragments of wood pulp were fixed on metal stubs using double-sided carbon tape. Some fibres were cut with a razor blade in order to visualize their cross-section. All specimens were coated with Au/Pd and observed in secondary electron mode in a Quanta 250 FEI microscope operating at 10 kV.

Transmission electron microscopy (TEM). Drops of 0.001 wt% CNC suspensions were deposited onto glow-discharged copper grids coated with a carbon support film. The specimens were negatively stained with 2 wt% uranyl acetate. Other specimens were kept unstained for electron diffraction purpose and mounted in a Gatan 626 specimen holder. After introduction in the microscope, the holder was cooled with liquid nitrogen and the specimen was observed at low temperature (−180 °C) under low illumination. All specimens were observed with a Philips CM200 ‘Cryo’ transmission electron microscope (FEI) operating at 200 kV. Images and diffraction patterns were recorded on a TVIPS TemCam F216 camera. The average dimensions of CNC populations were determined from the TEM images by measuring the length and width of about 100 particles using the ImageJ software.

X-ray diffraction (XRD). Strips of CNC films prepared by casting were fixed on a 0.2 mm collimator in such a way that the films were oriented either perpendicular or parallel to the X-ray beam. Fragments of PVA/CNC fibres were directly taped on the collimator. All specimens were placed in a vacuum chamber and X-rayed in transmission mode using a Philips PW3830 generator producing a Ni-filtered Cu Kα radiation (λ = 0.1542 nm), and operating at 30 kV and 20 mA. Two-dimensional (2D) diffraction patterns were recorded on Fujifilm imaging plates read with a Fujifilm BAS-1800II bioimaging analyser. Diffraction profiles were obtained by radially averaging the intensity over fan-shaped portions of the 2D spectra. A crystallinity index (CI) was calculated after peak deconvolution of the diffraction profiles.³⁴ The peak fitting was carried out using pseudo-Voigt functions. CI was calculated using eqn (3):where A_c and A_a are the areas under the crystalline peaks and the amorphous halo, respectively, determined by deconvolution between 2θ = 10 and 30°.

The dimension D_hkl of the crystallites perpendicular to the hkl diffracting planes was estimated from the peak broadening in the XRD profiles after deconvolution, using Scherrer's equation:

where K is a correction factor usually taken as 0.9 for cellulose, λ is the X-ray wavelength, θ is the diffraction angle and β_1/2 is the peak width at half maximum intensity.³⁵

Fourier transform infrared spectroscopy (FTIR). CNC powders were mixed with KBr in the form of compressed pellets and absorption spectra were recorded with a Perkin-Elmer Spectrum One instrument, in the range of 600–4000 cm⁻¹, a resolution of 4 cm⁻¹ and a total of 16 scans for each sample.

Solid-state nuclear magnetic resonance (NMR). Solid-state ¹³C-NMR experiments were performed on a Bruker AVANCE 400 spectrometer using a combination of cross-polarization, high power proton decoupling and magic angle spinning (CP/MAS). ¹³C-NMR spectra were acquired at 298 K, with a 4 mm probe operating at 100.13 MHz. MAS rotation was 12 kHz, CP contact times 2 ms and repetition time 1 s. The CI of the samples was also estimated from ¹³C-NMR data by comparing the areas for the peaks of crystalline and amorphous C4. By analogy with eqn (4), the calculation was done by dividing the area of crystalline peak (86 to 94 ppm) by the total area assigned to the C4 peak (80 to 94 ppm).

Results

Characterization of native and mercerized cellulose fibres

The SEM images in Fig. 1 show the morphological differences between the native wood pulp (WP) fibres and the mercerized ones (MWP). The raw WP fibres are long tortuous and sometimes twisted flat tubes with a relatively smooth surface (Fig. 1a–c), whereas the MWP fibres are solid cylinders, the lumen having collapsed during the alkaline treatment (Fig. 1d–f). The surface of the MWP fibres exhibits a sub-fibrillar structure that most likely corresponds to bundles of nanofibrils, which is consistent with what was observed for mercerized pinewood pulp³⁶ and cotton fibres.³⁷ The mercerization process also affects the tortuosity and swelling of the fibres (i.e., the fibre untwists and swells). However, the fibrous structure is largely kept intact after mercerization. The yield of mercerization, calculated on the initial amount of dried cellulose fibres was 83% (Table 1). In addition, there was no significant change in the DP values estimated by viscosity measurements: 898 and 920 for WP and MWP, respectively.

Fig. 1 SEM images of initial (a–c) and mercerized (d–f) wood pulp fibres (WP and MWP, respectively). In (c) and (f), the fibres were cut with a razor blade revealing their cross-section.

Table 1 Yield (Y in %), sulphur content (S in %), surface charge (SC in mmol g⁻¹), number of –OSO₃H groups per 100 anhydroglucose units (N), zeta potential (ζ in mV cm⁻¹) and degree of polymerization (DP) of the fibres and CNCs. “na”: not applicable

Sample	Y	S	SC	N	ζ	DP
a Calculated with respect to the initial amount of dried wood pulp.b Based on the initial cellulose content.c Calculated with respect to the initial amount of mercerized fibres.
WP	na	0.94	na	na	na	898 ± 18
MWP	83^a	0.98	na	na	na	920 ± 14
CN-I	64^a/70^b	1.36	0.424	7.12	−33.5 ± 2.0	115 ± 3
MCN-II	55^a/60^b/66^c	1.43	0.464	7.49	−35.7 ± 1.5	54 ± 3
RCN-II	44^a/48^b	1.75	0.546	9.25	−30.0 ± 0.9	17 ± 1

---

The XRD patterns recorded from WP and MWP fibrillar bundles are shown in Fig. 2. The fibre pattern of WP is typical of native wood Cel-I. It contains 3 equatorial arcs assigned to (11̄0)_I, (110)_I, and (200)_I planes (indices of allomorph I_β), as well as the characteristic thin meridional arc corresponding to the (004)_I planes (ESI Fig. S3a†).³⁸ The CI calculated for this sample after peak deconvolution is 33% (Table 2).

Fig. 2 X-ray diffraction patterns of initial (a) and mercerized (b) wood pulp fibres (WP and MWP, respectively). The fibrillar bundles are vertical with respect to the patterns.

Table 2 Crystallinity index (CI) of the samples estimated from XRD data (CI^XRD) and ¹³C-NMR data (CI^NMR) after peak deconvolution

Sample	CI^XRD (%)	CI^NMR (%)
WP	33	—
MWP	53	—
CN-I	56	50
MCN-II	68	63
RCN-II	66	60

---

The XRD pattern of MWP is clearly different and typical of Cel-II, displaying three main arcs on the equator, corresponding to the (11̄0)_II, (110)_II, and (020)_II planes (ESI Fig. S3b†).³⁸ The mercerization process thus resulted in a total allomorphic conversion from Cel-I to Cel-II. The calculated CI was 53% but its comparison with that of WP is not straightforward since the crystal structure for both specimens is different.

Characterization of cellulose nanocrystals

Yield, chain length, sulphur content, surface charge and zeta potential. The yield of the three preparation protocols as well as DP, sulphur content, surface charge and zeta potential ζ of the resulting CNC samples are given in Table 1. Seeing that lower yields around 30% are often reported in the literature, our values can be considered as high and are consistent with other reports.^12,29,39,40 The DP of CNC samples were 115, 54 and 17 for CN-I, MCN-II and RCN-II, respectively. The formula used in this study to calculate the DP is recommended for characterization of unsulphated celluloses with a DP between 90 and 3000. Thus, the DP values assessed by means of viscosity-average molecular weight are only an estimate of the chain length of the samples.

The sulphur content (Table 1) is related to the presence of anionic sulphate ester groups on the CNC surface created by esterification of hydroxyl groups during the sulphuric acid treatment, thus leading to colloidal CNC dispersions in water electrostatically stabilized by repulsive forces, as indicated by the negative ζ values (Table 1). The differences between the sulphur content values can be ascribed to the acid treatment conditions, since stronger sulphuric hydrolysis conditions are expected to result in higher levels of esterification sulphation.^41,42 The ζ values of the three samples are rather close (−33.5 ± 2.0 mV cm⁻¹ for CN-I, −35.7 ± 1.5 mV cm⁻¹ for MCN-II and −30.0 ± 0.9 mV cm⁻¹ for RCN-II), and considering that the zeta potential may also depend on the shape, size and structure of the particles, the small difference may not be significant and directly related to the sulphur content.

Morphology and crystal structure. ESI Fig. S1† shows the formation of birefringent domains induced by shear alignment of CNCs in aqueous suspensions under agitation. This effect results from the intrinsic shape and optical anisotropy of rod-like CNCs.⁴³ Thus, this behaviour hints that well-dispersed of anisometric nanosized cellulose particles were obtained in all three isolation procedures. In general, the critical concentration at which the particles self-organize depends on their aspect ratio and on the electrostatic repulsion.⁴⁴ Consequently, for the same concentration, the weaker shear-induced birefringence exhibited by the RCN-II suspension may be explained by a smaller particle aspect ratio.

TEM micrographs of the different CNC samples are shown in Fig. 3 and the CNC dimensions measured from TEM (average length L̄ and width D̄) and AFM (average thickness H̄) images (ESI Fig. S2†) are summarized in Table 3. The shape and size of the three types of particles are clearly different. CN-I (Fig. 3a and b) contains rod-like particles typical of nanocrystals prepared by conventional acid hydrolysis of native higher plant cellulose.^15,18 They are generally composite objects made of a few laterally-bound crystallites, in agreement with what was shown for cotton CNCs.¹⁹ As illustrated for CN-I particles shown in Fig. 4a–c, the shape and number of constituting crystallite significantly vary so that it is rather difficult to clearly define a ‘typical’ particle. Their average length and width determined from TEM images of negatively stained preparations are L̄ = 240 ± 52 nm and D̄ = 15.1 ± 1.4 nm, respectively (Fig. 3a and b), while a thickness H̄ = 3.8 ± 0.9 nm was measured from height profiles on AFM images (ESI Fig. S2a†).

Fig. 3 TEM images of negatively stained CN-I (a and b), MCN-II (c and d) and RCN-II (e and f) nanocrystals. The arrows in (f) indicate twist regions.

Table 3 Average dimensions of the CNCs. The length L̄ and width D̄ have been calculated from TEM images of negatively stained preparations of CN-I, MCN-II and RCN-II while the thickness H̄ has been determined from height profiles in AFM images. The aspect ratio was calculated as L̄/D̄

Sample	L̄ (nm)	D̄ (nm)	H̄ (nm)	Aspect ratio
CN-I	240 ± 52	15.1 ± 1.4	3.8 ± 0.9	16
MCN-II	132 ± 19	18.9 ± 2.2	5.2 ± 1.0	7
RCN-II	102 ± 10	12.0 ± 1.2	2.7 ± 0.6	8.5

---

Fig. 4 Higher magnification TEM images of CN-I (a–c), MCN-II (d–f) and RCN-II (g–i) particles (negative staining). The arrows in (h) and (i) indicate twist regions. Scale bars: 20 nm.

MCN-II particles are also rod-shaped but shorter and wider: L̄ = 132 ± 19 nm, D̄ = 18.9 ± 2.2 nm, H̄ = 5.2 ± 1.0 (Fig. 3c and d and ESI S2b†). Like for CN-I, these particles are composite objects (Fig. 4d–f).

As seen in Fig. 3e, the RCN-II suspension contains dense aggregates coexisting with vermicular objects. As these aggregates resisted mild sonication, we could not study them in details and focused our analysis on the individual particles. These acicular particles are more tortuous, longer (L̄ = 102 ± 10 nm) and thinner (D̄ = 12.0 ± 1.2) than MCN-II nanocrystals (Fig. 3e). The thickness measured from AFM images (H̄ = 2.7 ± 0.6 nm) (ESI Fig. S2c†) and the distribution of negative stain in TEM images (Fig. 3f and 4h and i) suggests that the particles are ribbon-shaped. As indicated by arrows in Fig. 3f and 4h and i, white regions are locally observed that likely correspond to local twists of the ribbons. The ribbons would thus be seen edge-on and the width of these regions is about 3 nm, in good agreement with the thickness measured by AFM (Table 3). Although the resolution of these TEM images is mainly limited by the granularity of the dry stain, the particles seem to be composed of imperfectly stacked lamellar subunits (Fig. 4g).

The two-dimensional XRD patterns recorded from CNC films and shear-oriented PVA/CNC fibres are shown in Fig. 5. When the films were oriented perpendicular to the X-ray beam (Fig. 5a–c), the patterns were mostly of powder type, exhibiting a number of concentric diffraction rings. When the films were oriented parallel to the X-ray beam (Fig. 5d–f), all samples yielded fibre patterns with a distribution of diffraction arcs. This is consistent with the fact that rod-shaped particles tend to take an in-plane orientation during drying. Their long axis lies parallel to the plane of the film but the azimuth of this axis is randomly oriented. The rings in the pattern of the CN-I film (Fig. 5a) are not strictly uniform in intensity due to some orientation of the rod-like particles during drying, an effect previously reported for nanocrystals prepared from other cellulose sources.¹⁹ Like for the parent fibres, the XRD pattern of CN-I is typical of Cel-I, although the five main diffraction rings are more clearly defined. The CI value is 56% (Table 2). This increase with respect to the CI of the initial WP fibres (33%) is consistent with the fact that acid hydrolysis has preferentially degraded disorganized regions that contributed to the amorphous background in the XRD pattern for WP.

Fig. 5 Two-dimensional wide-angle X-ray diffraction patterns of CN-I (a, d and g), MCN-II (b, e and h) and RCN-II (c, f and i) specimens: (a–c) the films were oriented perpendicular to the X-ray beam; (d–f) the films were parallel to the X-ray beam and vertical with respect to the patterns; (g–i) PVA/CNC fibres oriented vertical with respect to the patterns. In (f), the thin ring at large diffraction angle was attributed to some contaminant.

The XRD patterns for MCN-II and RCN-II films (Fig. 5b and c, respectively) correspond to Cel-II, exhibiting the 11̄0_II and well-separated 110_II and 020_II rings (ESI Fig. S3b and c†). For MCN-II, the intensities of the latter two are similar whereas for RCN-II, the intensity of 110_II is higher, suggesting some uniplanarity of the particles in the film. The d-spacings, summarized in Table 4, are consistent with previously reported measurements.^29,30,45 The CI value for MCN-II is 68% (Table 2), higher than that of the mercerized fibres (53%), and the CI of RCN-II is 66%, which is lower than the value that Hu et al.²⁹ estimated for their regenerated Cel-II nanocrystals using a similar deconvolution method (93%).

Table 4 d-Spacings and associated crystal widths corresponding to the three main reflections in the XRD profiles of the wood pulp fibres and CNCs. The indices are those defined by French³⁸

hkl		11̄0I	110I	200I
	d-Spacing (nm)	0.60	0.53	0.39
CN-I	Crystallite size (nm)	2.98	3.01	4.35

---

hkl		11̄0II	110II	020II
	d-Spacing (nm)	0.73	0.44	0.40
MCN-II	Crystallite size (nm)	5.85	5.03	4.71
RCN-II	Crystallite size (nm)	5.03	5.82	5.58

---

The CN-I fibre pattern recorded from a film parallel to the X-ray beam is similar to that of the parent WP fibres but it is better resolved and contains equatorial arcs assigned to (11̄0)_I, (110)_I, and (200)_I planes, and a thin meridional arc corresponding to the (004)_I planes (Fig. 5d and ESI S3a†). The pattern of the MCN-II film contains the 11̄0_II, 110_II and 020_II arcs in equatorial position whereas a thin 002_II meridional reflection can be seen (Fig. 5e and ESI S3b†). The 11̄0_II reflection remains equatorial in the pattern of the RCN-II film but the strong 110_II arc is now oriented along the meridian. The 020_II reflection is expected to lie at a 65° angle with respect to the equator⁴⁵ but due to some degree of misorientation, this reflection and its mirror counterpart with respect to the meridian are merged into a unique larger arc (Fig. 5f and ESI S3c†).

Fig. 5g–i show the XRD patterns for shear-oriented PVA/CNC fibres. In the case of CN-I and MCN-II, the resulting fibre patterns are a proof that the acicular nanocrystals were oriented along the stretching direction. This did not seem to be the case for RCN-II whose pattern only exhibits a very weak orientation (Fig. 5i). The reason for this lack of orientation is unclear since the aspect ratio of RCN-II particles is close to that of MCN-II nanocrystals that aligned well along the shear direction (Table 3). Particle aggregation during the preparation of the PVA/CNC mixture may have occurred, or a large amount of the aggregates seen in Fig. 3e was present, thus limiting the orientation during shear. In addition, the ribbon shape and associated twisting effect may induce some flexibility in the particles which would also add to the lack of orientation. However, one can note that the relative intensity of the 11̄0_II reflection is significantly lower compared to that in Fig. 5f, and that the intensity of the 110_II ring is still stronger than that of 020_II.

Selected area electron diffraction patterns were recorded from unstained specimens at low temperature, as shown in ESI Fig. S4.† For CN-I and MCN-II, it was possible to find bundles of locally aligned nanocrystals so that fibre patterns were recorded. The patterns from the three samples are thus similar to the XRD diagrams of the corresponding shear-oriented PVA/CNC fibres (Fig. 5g–i). For CN-I, the thin 004_I meridional reflection is clearly marked, as well as the strong equatorial 200_I and the merged 11̄0_I and 110_I reflections (ESI Fig. S4a†). For MCN-II, the thin meridional 004_II reflection appears and the strong equatorial 110_II and 020_II arcs have similar intensities (ESI Fig. S4b†). The pattern of the RCN-II sample does not reveal any clear orientation, probably because the nanocrystals are shorter which does not promote the formation of locally oriented bundles upon drying. However, the fact that the 110_II reflection is more intense than 020_II indicates a strong uniplanarity of the nanocrystals on the supporting carbon film (ESI Fig. S4c†). These data recorded by electron diffraction on a small number of particles are thus in good agreement with those collected by XRD from larger populations of CNCs.

The size of the elementary crystallites constituting the three types of CNC was evaluated from the broadening of the three stronger peaks in the XRD profiles corresponding to the powder patterns in Fig. 5a–c, using Scherrer's equation (eqn (4)). The results are summarized in Table 4 and were used to propose shapes of the cross-sections of the crystallites (ESI Fig. S5†). The cross-section of CN-I crystallites is squarish (approximately 3 × 3 nm²), with surfaces corresponding to (11̄0)_I and (110)_I planes (ESI Fig. S5a†), which is consistent with the thickness determined from AFM images (Table 3) and the estimation of the width of the laterally associated crystallites that constitute CN-I CNCs (Fig. 4a–c). This shape and dimensions were also reported for CNC prepared from Avicel¹⁹ and wood pulp from various sources.¹⁸ By dividing the average width of CN-I particles by that of the elementary crystallites, one finds that the CNC contains an average of 5 laterally bound units, in good agreement with the TEM images in Fig. 3a and b and 4a–c. The elementary crystallites in MCN-II particles are wider than those in CN-I, which is consistent with the TEM observations (Fig. 4d–f), and would have the shape of a truncated parallelogram (ESI Fig. S5b†). MCN-II particles would thus be made of 3–4 crystallites.

From the XRD data, the cross-section of the RCN-II elementary crystallites would have a more or less hexagonal shape with an average diameter of 5.5 nm (ESI Fig. S5c†). The direct comparison with the TEM images is more difficult since, as suggested by diffraction analyses, the chains would be oriented perpendicular to the long axis of the particle. Consequently, the cross-section of the ribbons cannot be visualized, except, possibly, in the twist regions (Fig. 4h and i). However, the width of these regions, in agreement with the thickness measured by AFM, is close to 3 nm, which is about twice smaller than the size deduced form the XRD data. ESI Fig. S5d† describes the cross-section that was deduced from the analysis of TEM and AFM micrographs. The reason of the discrepancy between the images and the XRD data is unclear. One possibility is that the aggregates that coexist with the ribbon-like nanocrystals contain Cel-II particles with a different morphology and that the XRD data correspond to an average of both types of particles. In that stage, the TEM images cannot be used to quantitatively evaluate the relative amounts of aggregates and individual ribbons and further work is needed to clarify this point.

Fig. 6 shows the FTIR spectra for the three types of CNCs. All spectra exhibit the typical absorption peaks and bands assigned to cellulose, although small differences can be seen between them due to the different crystal structures. Considering the broad band for the –OH stretching vibration in the range of 3700–3000 cm⁻¹, the spectra for MCN-II and RCN-II display two distinctive peaks at 3491 and 3447 cm⁻¹ related to intramolecular hydrogen bonding in Cel-II. The intensity of the peak at 1430 cm⁻¹ associated to the –CH₂–(C6)– bending vibration decreases significantly in the spectra for MCN-II and RCN-II compared to CN-I, owing to the change in conformation of hydroxymethyl group from tg to gt form. The CO at C-6 stretching vibration peak at 1035 cm⁻¹ for CN-I was shifted to 1029 and 1027 cm⁻¹ for MCN-II and RCN-II, respectively. This could be explained by changes in the torsional angles of β-(1,4)-D-glycosidic linkages. Similarly, the COC vibration at β-glycosidic linkage was switched from 898 in CN-I to 895 cm⁻¹ for MCN-II and RCN-II. The absorption band at 711 cm⁻¹ assigned to I_β cellulose was detected for CN-I, while the absorption band at 750 cm⁻¹ ascribed to I_α was not clearly seen. It means that this sample (CN-I) is rich in the I_β allomorph. These FTIR results indicate that both mercerization and regeneration procedures transformed the cellulose crystal structure from I_β to II. Moreover, these results are consistent with other reports in the literature.^12,28,46,47

Fig. 6 Fourier-transform infrared absorption spectra of CN-I, MCN-II and RCN-II powders in KBr pellets.

The ¹³C-NMR spectra recorded from the three CNC samples are shown in Fig. 7. The peaks common to all spectra are those from the carbons of the carbohydrate moiety, which appear between 58 and 110 ppm. The signal from 58 to 67 ppm is attributed to the C6 of the primary alcohol group. The cluster between 69 and 80 ppm is associated to the C2, C3, and C5 carbons. The region between 80 and 94 ppm corresponds to C4 and that between 95 and 110 ppm with the anomeric carbon C1.^30,48 The CN-I spectrum displays the typical chemical shifts assigned to the crystal lattice of cellulose I_β. The signals at 104 and 106 ppm are well separated. The spectra for MCN-II and RCN-II are typical of Cel-II, as revealed by the presence of peaks at 76 and 107 ppm and the absence of a peak at 65 ppm.⁴⁷ The main difference was observed near 97 ppm, where a small peak only present for RCN-II has been attributed to C1 anomeric carbons of the end monomer units (C1 reducing end of cellulose), which can be detected when the DP of cellulose is sufficiently low.^30,49 Hence, this suggests that the cellulose chains are shorter in RCN-II than in CN-I and MCN-II. This is in agreement with the DP values determined from viscometry measurements (Table 1), and with the results of Hu et al.²⁹ that showed that the acid-soluble cellulose chains that recrystallize have a low DP.

Fig. 7 ¹³C CP/MAS NMR spectra of CN-I, MCN-II and RCN-II powders.

Distinct peaks corresponding to carbon atoms in the crystalline and disordered regions could be detected. The signal at around 89 ppm is assigned to the C4 of the highly ordered cellulose chains in the crystallite core, whereas the signal at 84 ppm corresponds to the C4 of disordered cellulose.⁴⁸ The peaks around 65 ppm were assigned to the less ordered cellulose C6 carbons. The CI was also calculated from the ¹³C-NMR spectra and the results are listed in Table 2. Similar results for CI^NMR can be found in the literature, usually around 60 ± 5% for cellulose nanocrystals.^34,49,50 Although the CI values calculated from XRD and NMR data are different, their relative variation are similar and the conclusions based upon these variations are consistent.

Discussion

Sulphur content and surface charge

Taking into account the number of –OSO₃H groups per 100 anhydroglucose units (Table 1), the number of cellulose chains exposed on the CNC surface according to the cross-section of the elementary crystallites for each type of CNC deduced from XRD data (ESI Fig. S5†), and assuming that only 3 –OH groups are accessible for each cellobiose repeating unit, the percentage of esterified hydroxyl groups relative to the amount of available hydroxyl groups at the CNC surface was estimated to be 9.1, 15.8 and 19.3% for CN-I, MCN-II and RCN-II, respectively. The surface area that is not accessible due to the association of crystallites in the particles was neglected for these estimates. These values are consistent with those reported for nanocrystals extracted from native cotton by hydrolysis at 45 °C using 65 wt% H₂SO₄.⁴²

Three-dimensional shape and crystal orientation of the CNCs

As shown from images and diffraction data, the elementary crystallites in CN-I particles have a square cross-section and the chain axis is parallel to the long axis of the nanocrystal (Fig. 8a). The CN-I particles are made of a few parallel crystallites whose lateral association by hydrogen bonds has not be broken by acid hydrolysis.¹⁹ As the particles are rather flat, some uniplanarity might have been promoted during drying as a film. However, no such preferential orientation was detected in XRD powder patterns (Fig. 5a), which means that in each particle, the associated elementary crystallites are randomly oriented with respect to one another around the common chain axis.

Fig. 8 Scheme describing the orientation of the Cel-I and Cel-II crystallites with respect to the CN-I (a), MCN-II (b) and RCN-II (c) composite particles.

Mercerization is considered to be a solid-state transition in which the cellulose chains from neighbouring nanofibrils with opposite polarity can mix upon swelling in alkali. As demonstrated for Valonia cellulose, after recrystallization into Cel-II, the fibrous texture is preserved and the chain axis is still parallel to the fibre axis.⁵¹ This orientation is the one that was determined for MCN-II particles (Fig. 8b) and is consistent with the conclusions of Kim et al.²⁵ from electron diffraction data collected on CNCs from mercerized ramie cellulose. Yue et al.²⁶ reported a similar morphology for CNCs prepared by acid hydrolysis of mercerized cotton cellulose. Such rod-like nano-elements were also described by Hirota et al.²⁴ after TEMPO-mediated oxidation of mercerized cotton cellulose. MCN-II CNCs are shorter than CN-I particles. Indeed, one can expect that the intermingling of the chains upon swelling and subsequent recrystallization generate a significant number of disorganized regions that were thus preferentially degraded during acid hydrolysis. This assumption is confirmed by the 66% yield which means that a significant amount of material was dissolved during acid hydrolysis. Like for CN-I, MCN-II particles are made of a few associated elementary crystallites with no specific orientation relative to each other since no in-plane uniplanarity was detected in the XRD pattern. However, the crystallization upon washing generates larger crystalline domains as shown by broader constituting crystallites (ESI Fig. S5b†), in agreement with the observations of mercerized ramie CNCs by Kim et al.²⁵

The coarse morphology of RCN-II crystals resembles that of the particles prepared in similar conditions by Sèbe et al.,³⁰ Hirota et al.,²⁴ Martins et al.⁵² and Henrique et al.⁵³ However, the images shown by these authors were recorded either by AFM and lack the lateral resolution of TEM images, or by TEM but often from crowded assemblies, which makes the comparison of finer details difficult. The crystal orientation that we propose and that is illustrated in Fig. 8c, was deduced from several results: (i) since the 110_II is in meridional position in the pattern of Fig. 5f, the chain axis is perpendicular to the long axis of the particle, in agreement with the conclusion of Atkins et al.⁴⁵ who characterized fibres extruded from solutions of cellulose in hydrazine; (ii) the 110_II reflection is meridional and very strong (Fig. 5f and ESI S4c†), hence the (110)_II planes are parallel to the observation direction and perpendicular to the base plane of the ribbons; (iii) the 11̄0_II reflection is weak in the XRD pattern of Fig. 5c and absent in the electron diffraction pattern in ESI Fig. S4c,† hence, the (11̄0)_II planes are parallel to the base plane of the ribbons and perpendicular to the observation axis. As a consequence, the c-axis, that is normal to the (110)_II and (11̄0)_II planes should lie parallel to the ribbon plane. In that case, the width of the ribbons should more or less correspond to the length of the cellulose chains. An extended cellulose chain with 17 would have a length of about 9 nm that is in rather good agreement with the average width measured from TEM images (12 nm), considering the irregular and rough aspect of the ribbons. Surprisingly, no 00l_II reflection was observed in the diffraction patterns of Fig. 5f and ESI S4c,† as opposed to the patterns of MCN-II particles (Fig. 5e and h and ESI S4b†) in which the 002 and/or 004_II reflections are visible. Since the number of (001)_II planes, which is directly related to the width of the RCN-II ribbon, is small, and if the degree of order is lower in the c-direction, it is possible that the 00l signals are very weak and thus not clearly visible. To verify this assumption, it would be necessary to record low-temperature electron diffraction patterns from individual ribbons, like those recorded by Kim et al.²⁵ from individual nanocrystals of mercerized ramie cellulose (thus equivalent to our MCN-II particles). This is a real challenge considering the small volume of material probed by the electrons, the irregular morphology of the ribbons and the very high beam sensitivity of cellulose.

Ribbon-like cellulose II crystals have also been prepared by Buléon and Chanzy⁵⁴ by deacetylating a fraction of short-chain cellulose triacetate in a water/methylamine solution at 90 °C. Helbert et al.⁵⁵ used fractions of short chains prepared by hydrolysis of Avicel cellulose with phosphoric acid.⁵⁶ When crystallized at 90 °C, long lamellar crystals were obtained, exhibiting the same shape as those previously described by Buléon and Chanzy.⁵⁴ The crystal orientation was confirmed by electron diffraction and high-resolution lattice imaging:⁵⁵ the chain axis lay perpendicular to both the long axis and the base plane of the crystal. In both methods, the lamellar crystals were formed at high temperature (90 °C), as opposed to RCN-II ribbons that were prepared by aqueous regeneration at room temperature. This may explain the apparent lower ‘perfection’ of RCN-II crystals but, in addition, they also differ in that the chain axis is oriented in-plane. Indeed, it would be interesting to test the crystallization conditions described by Buléon and Chanzy⁵⁴ with the fraction used to prepare the RCN-II sample and see if well-defined lamellar single crystals are obtained.

The ribbon-like RCN-II nanocrystals present a strong analogy with the particles prepared by precipitation of mannan, a linear homopolymer of β-(1,4)-D-mannosyl residues found as an energy reserve or a cell-wall structural component in plants and algae.⁵⁷ Heux et al.⁵⁸ precipitated solutions of mannan from ivory nut ( 30) and the seaweed Acetabularia crenulata ( 350). The resulting crystals were described as rod-like or ribbon-like depending on the source of mannan, with a width of 7 and 30 nm, respectively, and corresponding to allomorph II. As first shown by Bittiger and Husemann⁵⁹ and confirmed by Heux et al.,⁵⁸ the mannan chains lie perpendicular to the long axis of the particles. The crooked aspect of the ribbons of mannan from A. crenulata is very similar to that of our RCN-II nanocrystals. However, in the former case, chain folding was believed to be involved.

Considering the irregular aspect of the ribbons prepared by regenerating solutions of short-chain cellulose and mannan at room temperature, two different linear polysaccharides in which the sugar units are connected by β-(1,4) bonds, one is tempted to describe this crystallization as ‘imperfect’. However, nanocrystals with a specific molecular organization, a high crystallinity and a fairly well defined width are obtained. The ribbons axially grow along the [11̄0]_II direction by hydrophobic stacking of cellulose sheets containing antiparallel chains.

Due to the specific organization of cellulose chains in RCN-II nanocrystals, a large number of reducing ends are exposed at the surface which could be used for chemical modification and grafting of functional groups to impart original properties such as stimuli-responsiveness and develop new applications (e.g., biosensing, bioimaging agents, etc.).

Conclusions

The three types of CNC prepared by sulphuric acid hydrolysis exhibit different morphologies and crystalline structures. When the acid hydrolysis conditions are set-up in such a way that the crystalline domains in the initial wood pulp and mercerized celluloses (WP and MWP, respectively) are preserved (60 wt% H₂SO₄, 45 °C, 50 min), the resulting nanocrystals retain the fibrillar nature of the parent fibres (i.e., the chain axis is parallel to the long axis of the acicular particles) and their initial allomorphic type (I for WP and II for the MWP). In both cases, the particles are mostly made of a few laterally-bound elementary crystallites. The unit nanocrystals in CNCs from mercerized cellulose (MCN-II) are shorter but broader than those prepared from cellulose I fibres (CN-I). If harsher conditions are used (64 wt% H₂SO₄, 40 °C, 20 min), resulting in the depolymerisation and dissolution of native cellulose, the short chains ( 17) recrystallize into Cel-II ribbons upon regeneration in water at room temperature. In these somewhat tortuous ribbons, the chain axis would lie perpendicular to the long axis of the nanocrystal and parallel to its basal plane. In addition, these nanoribbons are very similar in shape and molecular orientation to mannan II nanocrystals prepared by recrystallization of mannan, a linear polymer of β-(1,4)-D-mannosyl residues, suggesting that this mode of crystallization may be a feature of short-chain linear β-(1,4)-linked polysaccharides.

Although similar ribbons of recrystallized cellulose II have been reported by other authors, to our knowledge, it is the first time that a morphological and structural description is proposed. By comparison with the fibrillar nanocrystals prepared by acid hydrolysis of native or mercerized cellulose fibres, the unique molecular and crystal structure of the nanoribbons implies that a higher number of reducing chain ends are located at the particle surface, which may be important for subsequent chemical modification. Further characterization of these regenerated cellulose nanoparticles is in progress to evaluate their applicative potential as, for instance, excipient in tablets for drug release, biosensing and bioimaging agents, or as a constituents in nanocomposite materials.

Acknowledgements

We gratefully acknowledge CAPES, FAPEMIG, CNPq, Federal University of Uberlândia, LGP2 and JSPS for financial support. We also thank the NanoBio-ICMG platform in Grenoble for granting access to the TEM facility, as well as H. Chanzy and Y. Nishiyama (CERMAV) for stimulating discussions. LGP2 and CERMAV are part of Institut Carnot PolyNat (Investissements d'Avenir – grant agreement no. ANR-11-CARN-030-01) and LGP2 is part of LabEx Tec 21 (Investissements d’Avenir – grant agreement no. ANR-11-LABX-0030).

References

1. K. H. Gardner and J. Blackwell, Biopolymers, 1974, 13, 1975.
2. R. H. Atalla and D. L. VanderHart, Science, 1984, 223, 283.
3. M. Wada, Y. Nishiyama, H. Chanzy, T. Forsyth and P. Langan, Powder Diffr., 2008, 23, 92.
4. F. J. Kolpak and J. Blackwell, Macromolecules, 1976, 9, 273.
5. P. Langan, Y. Nishiyama and H. Chanzy, Biomacromolecules, 2001, 2, 410.
6. P. Langan, N. Sukumar, Y. Nishiyama and H. Chanzy, Cellulose, 2005, 12, 551.
7. J. O. Warwicker, J. Polym. Sci., Part A-1: Polym. Chem., 1967, 5, 2579–2593.
8. Y. Nishiyama, S. Kuga and T. Okano, J. Wood Sci., 2000, 46, 452.
9. T. Okano and A. Sarko, J. Appl. Polym. Sci., 1985, 30, 325.
10. B. Medronho and B. Lindman, Adv. Colloid Interface Sci., 2015, 222, 502.
11. P. K. Gupta, V. Uniyal and S. Naithani, Carbohydr. Polym., 2013, 94, 843.
12. J. Han, C. Zhou, A. D. French, G. Han and Q. Wu, Carbohydr. Polym., 2013, 94, 773.
13. Y. P. Liu and H. Hu, Fibers Polym., 2008, 9, 735.
14. Y. Habibi, L. A. Lucia and O. J. Rojas, Chem. Rev., 2010, 110, 3479.
15. A. Dufresne, Nanocellulose: From nature to high performance tailored materials, Walter de Gruyter GmbH, Berlin and Boston, MA, 2012.
16. R. H. Marchessault, F. F. Morehead and N. M. Walter, Nature, 1959, 184, 632.
17. S. M. Mukherjee and H. J. Woods, Biochim. Biophys. Acta, 1953, 10, 499.
18. B. S. Brito, F. V. Pereira, J.-L. Putaux and B. Jean, Cellulose, 2012, 19, 1527.
19. S. Elazzouzi-Hafraoui, Y. Nishiyama, J.-L. Putaux, L. Heux, F. Dubreuil and C. Rochas, Biomacromolecules, 2008, 9, 57.
20. S. J. Eichhorn, Soft Matter, 2011, 7, 303.
21. R. J. Moon, A. Martini, J. Nairn, J. Simonsen and J. Youngblood, Chem. Soc. Rev., 2011, 40, 3941.
22. R. M. dos Santos, W. P. Flauzino Neto, H. A. Silvério, D. F. Martins, N. O. Dantas and D. Pasquini, Ind. Crops Prod., 2013, 50, 707.
23. M. Mariano, N. El Kissi and A. Dufresne, J. Polym. Sci., Part B: Polym. Phys., 2014, 52, 791.
24. M. Hirota, N. Tamura, T. Saito and A. Isogai, Cellulose, 2012, 19, 435.
25. N. Kim, T. Imai, M. Wada and J. Sugiyama, Biomacromolecules, 2006, 7, 274.
26. Y. Yue, C. Zhou, A. D. French, G. Xia, G. Han, Q. Wang and Q. Wu, Cellulose, 2012, 19, 1173.
27. E. Jin, J. Guo, F. Yang, Y. Zhu, J. Song, Y. Jin and O. J. Rojas, Carbohydr. Polym., 2016, 143, 327.
28. P. Dhar, D. Tarafder, A. Kumar and V. Katiyar, RSC Adv., 2015, 5, 60426.
29. T. Q. Hu, R. Hashaikeh and R. N. Berry, Cellulose, 2014, 21, 3217.
30. G. Sèbe, F. Ham-Pichavant, E. Ibarboure, A. L. C. Koffi and P. Tingaut, Biomacromolecules, 2012, 13, 570.
31. Y. Nishiyama, G. P. Johnson, A. D. French, V. T. Forsyth and P. Langan, Biomacromolecules, 2008, 9, 3133.
32. Technical Association of the Pulp and Paper Industry (TAPPI), Standard test viscosity of pulp (capillary viscometer method), Atlanta, 1999, NORMA TAPPI T 230 om-99.
33. Scandinavian Pulp Paper and Board (SCAN), Standard test viscosity of cellulose in cupriethylenediamine solution (CED), Stockholm, 1980, NORMA SCAN C15:62.
34. S. Park, J. O. Baker, M. E. Himmel, P. A. Parilla and D. K. Johnson, Biotechnol. Biofuels, 2010, 3, 1.
35. H. P. Klug and L. E. Alexander, X-ray diffraction procedures for polycrystalline and amorphous materials, John Wiley & Sons, New York, 1974.
36. S. Sharma, S. S. Nair, Z. Zhang, A. J. Ragauskas and Y. Deng, RSC Adv., 2015, 5, 63111.
37. Y. Yue, G. Han and Q. Wu, BioResources, 2013, 8, 6460.
38. A. D. French, Cellulose, 2014, 21, 885.
39. A. Bendahou, Y. Habibi, H. Kaddami and A. Dufresne, J. Biobased Mater. Bioenergy, 2009, 3, 81.
40. G. H. D. Tonoli, E. M. Teixeira, A. C. Corrêa, J. M. Marconcini, L. A. Caixeta, M. A. P. da Silva and L. H. C. Mattoso, Carbohydr. Polym., 2012, 89, 80.
41. M. Roman and W. T. Winter, Biomacromolecules, 2004, 5, 1671.
42. N. Lin and A. Dufresne, Nanoscale, 2014, 6, 5384.
43. E. D. Cranston and D. G. Gray, Colloids Surf., A, 2008, 325, 44.
44. L. Heux, G. Chauve and C. Bonini, Langmuir, 2000, 16, 8210.
45. E. D. T. Atkins, J. Blackwell and M. H. Litt, Polymer, 1979, 20, 145.
46. S. Y. Oh, D. Yoo, Y. Shin, H. C. Kim, H. Y. Kim, Y. S. Chung, W. H. Park and J. H. Youk, Carbohydr. Res., 2005, 340, 2376.
47. R. Zuluaga, J.-L. Putaux, J. Cruz, J. Vélez, I. Mondragon and P. Gañán, Carbohydr. Polym., 2009, 76, 51.
48. M. A. Martins, L. A. Forato, L. H. C. Mattoso and L. A. Colnago, Carbohydr. Polym., 2006, 64, 127.
49. S. Park, D. K. Johnson, C. I. Ishizawa, P. A. Parilla and M. F. Davis, Cellulose, 2009, 16, 641.
50. I. A. Sacui, R. C. Nieuwendaal, D. J. Burnett, S. J. Stranick, M. Jorfi, C. Weder, E. J. Foster, R. T. Olsson and J. W. Gilman, ACS Appl. Mater. Interfaces, 2014, 6, 6127.
51. H. D. Chanzy and E. J. Roche, Appl. Polym. Symp., 1976, 28, 701.
52. D. F. Martins, A. B. de Souza, M. A. Henrique, H. A. Silvério, W. P. Flauzino Neto and D. Pasquini, Ind. Crops Prod., 2015, 65, 496.
53. M. A. Henrique, W. P. Flauzino Neto, H. A. Silvério, D. F. Martins, L. V. A. Gurgel, H. S. Barud, L. C. de Morais and D. Pasquini, Ind. Crops Prod., 2015, 76, 128.
54. A. Buléon and H. Chanzy, J. Polym. Sci., 1978, 16, 833.
55. W. Helbert and J. Sugiyama, Cellulose, 1998, 5, 113.
56. A. Isogai and M. Usuda, Mokuzai Gakkaishi, 1991, 37, 339.
57. H. D. Chanzy, A. Grosrenaud, R. Vuong and W. Mackie, Planta, 1984, 161, 320.
58. L. Heux, P. Hägglund, J.-L. Putaux and H. Chanzy, Biomacromolecules, 2005, 6, 324.
59. H. Bittiger and E. Husemann, J. Polym. Sci., Part B: Polym. Lett., 1972, 10, 367.

---

Footnote

† Electronic supplementary information (ESI) available: Fig. S1: shear-induced birefringence for 0.15 wt% CNC aqueous suspensions; Fig. S2: tapping mode AFM images of CN-I, MCN-II and RCN-II nanoparticles; Fig. S3: indexed XRD patterns of the three CNC samples; Fig. S4: TEM images and electron diffraction patterns of unstained preparations for CN-I, MCN-II and RCN-II nanoparticles; Fig. S5: tentative models of the cross-sections of CN-I, MCN-II and RCN-II crystallites. See DOI: 10.1039/c6ra16295a

---