Biomacromolecules 2008, 9, 57–65 57

The Shape and Size Distribution of Crystalline Nanoparticles

Prepared by Acid Hydrolysis of Native Cellulose
Samira Elazzouzi-Hafraoui,† Yoshiharu Nishiyama,*,† Jean-Luc Putaux,† Laurent Heux,†
Frédéric Dubreuil,† and Cyrille Rochas‡
Centre de Recherches sur les Macromolécules Végétales (CERMAV-CNRS), BP 53,
F-38041 Grenoble cedex 9, France - affiliated with Université Joseph Fourier and member of the Institut
de Chimie Moléculaire de Grenoble, and Laboratoire de Spectrométrie Physique, UMR 5588, BP 87,
F-38402 Saint Martin d’Hères, France
Received July 11, 2007; Revised Manuscript Received September 25, 2007

The shape and size distribution of crystalline nanoparticles resulting from the sulfuric acid hydrolysis of cellulose
from cotton, Avicel, and tunicate were investigated using transmission electron microscopy (TEM) and atomic
force microscopy (AFM) as well as small- and wide-angle X-ray scattering (SAXS and WAXS). Images of
negatively stained and cryo-TEM specimens showed that the majority of cellulose particles were flat objects
constituted by elementary crystallites whose lateral adhesion was resistant against hydrolysis and sonication
treatments. Moreover, tunicin whiskers were described as twisted ribbons with an estimated pitch of 2.4–3.2 µm.
Length and width distributions of all samples were generally well described by log-normal functions, with the
exception of tunicin, which had less lateral aggregation. AFM observation confirmed that the thickness of the
nanocrystals was almost constant for a given origin and corresponded to the crystallite size measured from peak
broadening in WAXS spectra. Experimental SAXS profiles were numerically simulated, combining the dimensions
and size distribution functions determined by the various techniques.

Introduction high salinity. As higher concentrations could be achieved, the

shape and size of the particles would play a more significant
It has been known for more than half a century that stable
role on the self-organization properties.
nanoparticle suspensions could be prepared by submitting native
Nanoparticles with different morphologies can be prepared
cellulose to a harsh sulfuric acid hydrolysis often followed by
by varying the source of cellulose and the hydrolysis conditions.
ultrasound treatments.1 Such nanoparticles, prepared from
For instance, whiskerlike particles with a length of the order of
several types of higher plant cellulose with different hydrolysis
micrometers are obtained by hydrolyzing highly crystalline
conditions, have been characterized by using methods such as
cellulose samples from tunicates18 and green algae,19 whereas
transmission electron microscopy, sedimentation, flow birefrin-
higher plant cellulose from cotton and wood pulp yields shorter
gence, and viscometry.1,2 It was generally concluded that the
particles a few hundreds of nanometers long. For a given system,
particles were elongated and flat, a few hundreds of nanometers
the dimensions of the cellulose nanoparticles were generally
long, 10–20 nm wide, and a few nm thick.2,3
measured using imaging and/or scattering techniques.20 In some
Nanoparticles from acid-hydrolyzed cellulose and chitin
cases, size distribution histograms were determined from
attracted a renewed interest when it was shown that, in
transmission electron micrographs,2,7 but the nature of the
suspension, they could form chiral nematic phases4,5 resembling
distribution has rarely been studied in detail. To our knowledge,
the highly textured organizations of microfibrils found in native
the size distribution was only taken into account in one case,
systems.6,7 The effects of hydrolysis conditions and surface
reported by Marchessault et al., who simulated the birefringence
charge density,8,9 ionic strength,10–14 and pH10,15 on the phase
properties of Ramie crystallites.2
separation behavior and structures of the colloidal system
In the study presented here, we submitted three sources of
have been extensively studied over the last 15 years. However,
cellulose to sulfuric acid hydrolysis. The shape and size
despite the previous observation that the particles were rather
distribution of the resulting nanoparticles were determined from
flat objects, they have usually been treated as cylinders with
transmission electron microscopy (TEM) images, and height
mean diameters and lengths in models used to describe their
measurements were performed using atomic force microscopy
phase separation behavior in suspension.9 In fact, for aqueous
(AFM). Small- and wide-angle X-ray scattering (SAXS and
suspensions stabilized by electrostatic repulsion at low ionic
WAXS) experiments were used to characterize the whole
strength, the exact shape of the particles mattered little because
colloidal system.21
the effective diameter generated by the repulsive charges on
the surface was substantially larger than the particle width.9
Steric stabilization of the nanocrystals with either surfactants16 Experimental Section
or surface chemical modification17 recently allowed studying
Cellulose Sources. Three sources of cellulose were used. Cotton
cellulose whiskers suspensions in apolar solvents and media with
linters were provided by Rhône-Poulenc Tubize Plastics (Belgium) and
used without any further purification. Avicel, a commercial microc-
* Corresponding author. E-mail: yoshi@cermav.cnrs.fr. Fax: +33
476547203.
rystalline cellulose resulting from the hydrochloric acid hydrolysis of
†
Centre de Recherches sur les Macromolécules Végétales. wood pulp and containing 20 µm particles, was purchased from FMC
‡
Laboratoire de Spectrométrie Physique. Europe NV, Belgium. Tunicin, the cellulosic mantle of Halocynthia

10.1021/bm700769p CCC: $40.75  2008 American Chemical Society

Published on Web 12/04/2007
58 Biomacromolecules, Vol. 9, No. 1, 2008
Figure 1. TEM micrographs of negatively stained cellulose whiskers obtained by sulfuric acid hydrolysis of cotton (a), Avicel (b), and tunicate
(c-e) cellulose. Insets: enlarged views of some characteristic particles. The arrows in (d) indicate zones where the whiskers are seen edge-on.
roretzi, a sea animal, was purified by alternating treatments with KOH
and NaClO2.22
Preparation of Cellulose Whiskers Suspensions. Cotton linters
were hydrolyzed according to the method described by Revol et al.4
They were treated with 65% sulfuric acid during 30 min at four different
temperatures, namely 45, 54, 63, and 72 °C. In the following, the
resulting samples will be referred to as Cot45, Cot54, Cot63, and Cot72,
respectively. The suspensions were washed by centrifugation, dialyzed
to neutrality against distilled water, and ultrasonicated for 4 min with
a rod-type sonicator (Branson Sonifier B12). This treatment heated up
the suspension by about 20 °C. After these treatments, the suspensions
were filtered on 8 µm then 1 µm cellulose nitrate membranes (Sartorius)
and stocked with mixed bed resin (Sigma TMD-8) in order to eliminate
residual electrolytes. Avicel was hydrolyzed at 72 °C following the
same procedure. The resulting sample will be referred to as Avi72.
Tunicin was treated with 48% sulfuric acid during 13 h at 55 °C. Acid-
free microcrystalline cellulose suspensions were obtained after repeated
centrifugations, dialysis to neutrality against distilled water, and
ultrasonication for 1 min. This suspension, referred to as Tun55 in the
following, was stocked with mixed bed resin.
Transmission Electron Microscopy. Drops of 0.001 wt % cellulose
microcrystal suspensions were deposited on glow-discharged carbon-
coated TEM grids. The specimens were then negatively stained with
2% uranyl acetate, prior to complete drying, and observed using a
Philips CM200 electron microscope operating at 80 kV. Images were
recorded on Kodak SO163 films. Selected negatives were digitized using
a Kodak Megaplus CCD camera. For each preparation, the width and
length of about 1000 particles (except tunicin, for which 200 particles
were counted) were measured from the TEM images by using the
AnalySIS software. Thin vitrified films of 0.1 wt % cellulose micro-
crystal suspensions were also prepared by quench-freezing in liquid
ethane by using a procedure described elsewhere.23 They were observed
by cryo-TEM, at low temperature (-180 °C), in a Gatan 626 cryoholder.
Atomic Force Microscopy. After a short sonication to prevent the
formation of aggregates, drops of dilute cellulose microcrystals
suspensions were deposited onto freshly cleaved mica. After 30 min,
the excess liquid was removed and the remaining film allowed to dry.
AFM observations were carried out using a Molecular Imaging Pico
Elazzouzi-Hafraoui et al.
Plus microscope operating in air and intermittent contact mode with a
Micromash NC36 tip.
Wide-Angle X-ray Scattering. Concentrated cellulose suspensions
(3–4 wt %) were allowed to dry onto flat Teflon surfaces. At this
concentration, the suspension of Avicel particles was more viscous than
those containing cotton or tunicin microcrystals. The resulting films
were X-rayed with a Ni-filtered Cu KR radiation (λ ) 1.542 Å), using
a Philips PW3830 generator operating at 30 kV and 20 mA. The films
were positioned either parallel or perpendicular to the X-ray beam.
Diffraction patterns were recorded on Fujifilm imaging plates, read with
a Fujifilm BAS-1800II bioimaging analyzer. Diffraction profiles were
obtained by radially integrating the intensity over fan-shaped portions
of the spectra. The diffraction peaks were fitted with pseudo-voigt peak
functions, assuming a linear background. The dimension of the crystal
perpendicular to the diffracting planes with hkl Miller indices, Dhkl,
was evaluated by using Scherrer’s expression:
0.9 × λ
Dhkl ) (1)
β1/2 × cos θ
where θ is the diffraction angle, λ the X-ray wavelength and β1/2 the
peak width at half of maximum intensity.24
Small-Angle X-ray Scattering. SAXS experiments were performed
on the BM02 beamline of the European Synchrotron Radiation Facility
(Grenoble, France). Dilute cellulose suspensions (<0.5 wt %) were sealed
in circular sample holders equipped with 25 µm thick mica windows. The
path length was 1 mm. Fourteen keV (λ ) 0.089 nm) X-rays were used.
Scattered intensities were recorded during 20 s exposures on a CCD
detector (Ropper Scientific) placed about 1.60 m behind the sample. The
radial intensity distribution was obtained by integrating the whole diagram
after a correction of camera distortion. Scattering data recorded from a
cell filled with water was subtracted as background.
Results
Particle Morphology. Figure 1 shows TEM micrographs of
elongated nanoparticles prepared by sulfuric acid hydrolysis of
the three types of cellulose. Particles from cotton (Figure 1a)
Crystalline Nanoparticles from Native Cellulose Biomacromolecules, Vol. 9, No. 1, 2008 59

Figure 2. Cryo-TEM micrograph of cellulose particles embedded in

vitreous ice: (a) cotton (Cot45), (b) tunicin (Tun55). The arrows point
to regions of the whiskers that are seen edge-on and generate a
stronger diffraction contrast.

and Avicel (Figure 1b) have a length between 100 and 300 nm,
whereas those from tunicin are several micrometers long and
have a whiskerlike morphology (Figure 1c,d). A significant
fraction of tunicin whiskers exhibit a kinked aspect (Figure 1e),
the defects being likely due to localized damage created during
the sonication treatment. For all samples, the negative staining
clearly reveals that most particles are composed of a few parallel
subunits. To check if this association was due to an artifactual
aggregation during staining and/or drying of the specimens on
the carbon film, cryo-TEM experiments were performed on
suspensions Cot45 and Tun55. The micrographs in Figure 2
show cellulose particles embedded in thin films of vitreous ice.
Both cotton (Figure 2a) and tunicin (Figure 2b) nanoparticles
are distributed in the film with a rather well-defined interparticle
distance, which indicates a significant electrostatic repulsion due Figure 3. Width (left column) and length (right column) distribution
to the negative charge of sulfate groups present on the surface histograms of Cot45, Cot54, Cot63, Cot72, Avi72, and Tun55
of the microcrystals after the sulfuric acid hydrolysis. Moreover, suspensions obtained by measuring the width and length on TEM
the images confirm that both cotton and tunicin nanocrystals micrographs of negatively stained preparations. The continuous
are composed of a varying number of parallel subunits. curves correspond to a fit of the data with a log-normal function.
Regular variations in width can often be seen along tunicin
whiskers. For negatively stained preparations, the accumulation larly to its long axis. For tunicin, the existence of a geometrical
of stain is more important around the narrower parts of the twist observed on TEM images had to be taken into account.
whiskers (Figure 2e), which means that these regions are also Assuming a rectangular cross-section and a twisted ribbon
thicker. The alternation of narrow and wide parts along the geometry, the width was defined as the largest dimension
whiskers suggests that they have a ribbonlike shape and that a measured perpendicularly to the whisker axis, while the thick-
homogeneous twist occurs. The same effect was observed in ness e was defined as the smallest dimension measured along
cryo-TEM images of vitrified specimens, proving that the twist the same nanocrystal. The corresponding size distribution
existed in solution and was not a drying/staining artifact (Figure histograms are shown in Figure 3. All distributions are asym-
2b). Using the longer whiskers, we measured the distance metrical, extending toward the longer and wider particles. Most
between two narrow or two wide regions and deduced a half- distributions were fitted using a log-normal function D whose
helical pitch of 1.2–1.6 µm. This distance is larger than the expression is:

( )
average length of the whiskers, which explains why, in TEM
- (ln(x) - ln(m))2
images, only one narrow region is seen along most whiskers. D ) A exp (2)
Length and Width Distributions. The length L of about s2
1000 particles from the three sources of cellulose was measured where m is the mode and s2 is the variance. Log-normal
directly from the TEM images of negatively stained preparations. functions are classically used to describe the size-distribution
For cotton and Avicel particles, the width was defined as the of objects obtained by fragmentation.25–27 The parameters m
largest dimension measured along the nanoparticle, perpendicu- and s2 calculated for the distributions of all suspensions are
60 Biomacromolecules, Vol. 9, No. 1, 2008 Elazzouzi-Hafraoui et al.

Table 1. Dimensions of Particles Prepared by Acid Hydrolysis at

Different Temperatures from Various Sources of Cellulose and
Determined From TEM Imagesa
Ln (nm) Ll ln (nm) ll en (nm)
sample [σL %] (nm) PL [σl %] (nm) Pl [σe %]
Cot45 141 [39] 163 1.15 27 [52] 34 1.28 n.d.
Cot45 n.d.b n.d. n.d. 14 [57] 18 1.30 n.d.
(cryo-TEM)
Cot54 131 [39] 151 1.15 21 [52] 27 1.27 n.d.
Cot63 128 [43] 151 1.18 26 [46] 32 1.20 n.d.
Cot72 105 [47] 128 1.21 21 [52] 26 1.27 n.d.
Avi72 105 [35] 118 1.12 12 [42] 15 1.18 n.d.
Tun55 1073 [67] 1560 1.45 28 [46] 34 1.21 9.2 [23]
a
Number average length (Ln) and associated standard deviation (σL),
length-weighted average length (Ll), length polydispersity index (PL),
number average width (ln), and associated standard deviation (σl), width-
weighted average width (ll), width polydispersity index (Pl), and number
average thickness (en). b n.d.: not determined.

collected in Supporting Information Table S1. Only the width

distribution of tunicin whiskers did not seem to be properly fitted
with a log-normal function, the class corresponding to an
average width of 20 nm being more populated than what would
be expected for log-normal distribution.
Number average lengths and widths (Ln and 1n, respectively)
were calculated from the size distributions. Considering the
asymmetry of the histograms, length- and width-weighted
average values (LL and ll, respectively), similar to weight-
averaged molecular weight for polymer, were calculated as well.
Length and width polydispersity indexes can be defined as PL
) LL/Ln and Pl ) ll/ln. The values measured and calculated
from the particles of the three sources of cellulose are sum-
marized in Table 1.
Typically, cotton nanocrystals had a length ranging from 25
to 320 nm and a width from 6 to 70 nm. The peak position of
length distributions shifted to lower values with increasing
hydrolysis temperature (Ln ) 141, 131, 128, and 105 nm for
Thyd ) 45, 54, 63, and 72 °C, respectively), e.g., the fraction of
shorter particles increased with increasing hydrolysis temper-
ature. No clear trend was observed for width distributions (ln
) 27, 21, 26, and 21 nm for Thyd ) 45, 54, 63, and 72 °C,
respectively). With increasing hydrolysis temperature, the
standard deviation of the width distribution decreased while that
of the length distribution increased.
The length of Avicel particles ranged from 35 to 265 nm
and their width from 3 to 48 nm. Their polydispersity in length
and width was lower that that of cotton particles. While the
average lengths of Avicel and cotton nanocrystals were rather Figure 4. AFM images (topography mode, 2.1 × 2.1 µm2) of cellulose
particles spread onto a freshly cleaved mica surface: (a) cotton
similar, the width of Avicel particles was about twice lower (Cot45), (b) tunicin (Tun55), (c) transverse height profiles determined
than that of cotton particles. along lines 1 and 2 in image (b).
In the case of tunicin, both straight (i.e., defect-free) and
kinked whiskers were taken into account. The average length ments. Parts a and b of Figure 4 show examples of tapping
of a kinked whisker was measured by summing the length of mode AFM images from Cot45 and Tun55 cellulose particles,
its constitutive segments separating defects. Sixteen percent of respectively, deposited onto freshly cleaved mica surfaces.
the Tun55 particles were found to contain at least one defect. Despite a sonication of the suspensions before spreading on
The average length of the whole population (kinked and straight mica, some aggregates were observed on the surface. A
particles) was Ln ) 1073 nm. The fact that tunicin whiskers preliminary statistical analysis of the whole pictures in terms
exhibited a twisted ribbon morphology was also taken into of height distribution led to inconclusive results, most probably
account to determine their width distribution. For each whisker due to particle superimposition during drying. As is illustrated
where the twist was clearly visible, the width and thickness (the in Figure 4b,c, transverse line profiles were performed on a
largest and smallest dimensions along one whisker, respectively) number of isolated particles. We obtained an average thickness
were measured once. Only the width distribution is shown in of 7.3 ( 1.5 nm for Cot45 and 10.6 ( 2.3 nm for Tun55.
Figure 3. The average width and thickness were found to be ln Longitudinal line profiles on images of isolated tunicin whiskers
) 28 nm and en ) 9.2 nm, respectively (Table 1). showed that the height could typically vary from 6 to 14 nm,
Particle Thickness. Scanning probe microscopy was used which is consistent with the existence of a geometrical twist,
to determine the nanoparticle thickness by topography measure- as shown from TEM images.
Crystalline Nanoparticles from Native Cellulose Biomacromolecules, Vol. 9, No. 1, 2008 61

Orientation and Size of the Elementary Crystallites. The

WAXS diagrams recorded from films of hydrolyzed cellulose
particles are shown in Supporting Information Figure S1. For
all samples, five diffraction rings or arcs, characteristic of
j
cellulose I, namely (110), (110), (102/012), (200), and (040)
according to the monoclinic indexation by Sugiyama et al.,28
could be recognized. When the film of cotton nanocrystals was
perpendicular to the incident beam (Supporting Information
Figure S1a), the rings were rather homogeneous, which is
characteristic of a random planar orientation of the particle axes.
There is a small anisotropy in the diagram from tunicin
(Supporting Information Figure S1c), which shows that the
whiskers were slightly oriented. The anisotropy in the diffraction
diagram from Avicel is very clear (Supporting Information
Figure S1e) and is likely due to the high viscosity of the
suspension, promoting the formation of oriented domains during
drying. In all cases, when the films were oriented parallel to
the X-ray beam, the diffraction diagrams were anisotropic
because the chain axis of the microcrystals laid more or less
parallel to the film surface. By tracing equatorial profiles from
the WAXS diagrams shown in Supporting Information Figure
S1, we thus obtained information on the crystalline order
perpendicular to the chain axis (Figure 5). In all samples, the
j was stronger when the incident
relative intensity of spot (110)
beam was parallel to the film surface, the effect being reversed
when the film was perpendicular to the beam and not clearly
marked for Avicel films. This means that a significant unipla-
narity of the nanoparticles existed in cotton and tunicate films,
j planes tended to be oriented parallel to the film surface.
as 110
For tunicin, we could reasonably assume that this uniplanarity
was enhanced by the ribbon shape of the whiskers and that the
wider face of the ribbons corresponded to the 110 j planes.
The average cross-sectional dimensions of the elementary
crystallites were evaluated by applying Scherrer’s expression 1
to the half-width of the peaks present in the diffraction profiles
calculated from the WAXS diagrams. The values, given in
Supporting Information Table S2, constitute a lower limit of
Figure 5. Equatorial wide-angle X-ray diffraction profiles calculated
the particle width because the broadening from the beam size
from the diffraction diagrams shown in Supporting Information Figure
has not been taken into account. From these values, we proposed S1 and recorded from films of cotton (a), Avicel (b), and tunicate (c)
the cross-sectional geometries summarized in Figure 6. Cotton cellulose particles (Cot45, Avi72, and Tun55, respectively), with the
and Avicel crystallites have a square cross-section, with widths incident beam parallel (|) or perpendicular (⊥) to the film; +:
of 6.0 and 4.4 nm, respectively. To properly combine the experimental profiles; -: peaks resulting from the deconvolution
j (110), and (200)
dimensions estimated perpendicularly to (110), procedure.
planes, tunicate crystallites would have a section with the shape
of a truncated parallelogram, 12 nm wide and 10 nm thick. The
elementary crystallite dimensions of cotton and tunicin particles a more sophisticated modeling of the system was required that
are in good agreement with those obtained by height analysis will be described in the following discussion.
of AFM images.
Small-Angle X-ray Scattering of the Suspensions. For a Discussion
dilute system composed of rodlike particles with a reasonably
monodisperse diameter, the Guinier approximation shows that The results obtained using imaging and scattering techniques
there is a regime where were combined to provide a general description of the crystalline
cellulose particles prepared by acid hydrolysis. The particles
ln(IQ) ) ln(I0) - Q2R2g ⁄ 2 (3) are flat, polydisperse in length and width, but rather homoge-
neous in thickness. They are composed of a small number of
Q is the scattering vector, I the intensity, and Rg the radius laterally associated elementary crystallites (3–4 for cotton, 2–3
of gyration of the particle cross-section. The approximation for Avicel and tunicin) and their thickness is that of one
applies in the scattering vector range that satisfies QRg , 1. In crystallite. This result is in agreement with the first observations
this Q range, if a plot of ln(IQ) versus Q2 (cross-section plot) made in the 1950s–1960s.1-3
exhibits a linear region, it is possible to calculate Rg and thus Comparison of the Results with Previous Works. The
estimate the diameter of the particles. SAXS diagrams were width of the cotton particles (Table 1) is about three times larger
recorded for suspensions Cot45, Cot72, Avi72, and Tun55. None than that measured by Dong et al.8 (7 nm) and Araki et al.12
of the four corresponding Guinier plots exhibited linear regions (5–10 nm) from suspensions prepared using the same conditions
in the Q range satisfying QRg , 1, which would be the case it and whose TEM images look very similar to ours. One can
the particles were described as monodisperse cylinders. Thus, explain this difference by supposing that these authors have
62 Biomacromolecules, Vol. 9, No. 1, 2008
Figure 6. Cross-sections of elementary crystallites deduced from the
analysis of peak broadening in WAXS profiles from films of cotton,
Avicel, and tunicate cellulose particles. The short bars indicate the
projection of cellulose molecules. The indexation of corresponding
lattice planes is described in Supporting Information Figure S2.
measured the width of elementary crystallites without taking
into account the fact that one particle could be inherently
constituted of several crystallites. The same comment can be
made for Avicel particles whose width is about three time larger
that that of wood pulp reported in literature (about 3.5 nm).4 In
fact, the widths reported in literature are close to those we
measured for the subelements constituting the nanoparticles and
which probably correspond to single crystals.
Using TEM images of tunicin whiskers, Kimura et al. have
measured a length of 1–3 µm and a width of 15–30 nm, which
are in satisfactory agreement with our own values considering
the high polydispersity. The thickness estimated from TEM and
AFM images is in agreement with the value determined by
Terech et al. (8.8 nm) who used small-angle neutron scattering
data and assumed a rectangular section.30 The cross-sectional
dimensions of tunicin particles are consistent with those reported
by van Daele et al. from TEM images of ultrathin sections of
Halocyntia papillosa tunicate cell walls.31 Using high-resolution
TEM images of similar H. papillosa sections, Helbert et al.
showed that the acid treatment blunted the angular lines of the
crystallites.32
Effect of Negative Staining on the Size Distribution. In
some cases, negative staining is known to induce an aggregation
of nanoparticles on the supporting carbon film.23 To rule out
the possibility of an artifactual aggregation of cellulose crys-
tallites due to negative staining, we compared the width
distribution determined from TEM images of negatively stained
preparations with that obtained from fast-frozen specimens for
which aggregation was assumed not to occur.
We used the width distribution histogram of negatively
stained Cot45 nanoparticles (Figure 7a) and assumed a constant
thickness of 6 nm. For each class of width, a free rotation of
the particle around its long axis was simulated and the apparent
width of the particle projected in the image plane was calculated.
The final histogram was determined by summing all the
simulated distributions (Figure 7b). It was then compared to
that obtained by measuring the apparent width of a number of
particles embedded in vitreous ice by using cryo-TEM images
such as the one shown in Figure 3 (Figure 7c). Both simulated
Elazzouzi-Hafraoui et al.
Table 2. Cellulose Crystallite Width (in nm) Estimated from the
Width at Half-Height of Maximum of X-Ray Diffraction Peaks from
Films of Cotton (Cot45), Tunicate (Tun55), and Avicel (Avi72)
Cellulose Whiskersa
peak cotton | cotton ⊥ tunicate | tunicate ⊥ Avicel | Avicel ⊥
j
110 5.8 5.1 8.9 9.0 4.0 4.1
110 5.6 5.5 11.2 10.4 3.8 4.0
102/012 6.2 10.0 3.0 3.5
200 6.4 7.2 9.0 10.0 4.3 5.0
a
The diffraction patterns were recorded with a beam perpendicular (⊥)
or parallel (|) to the film plane.
Figure 7. (a) Width distribution of hydrolyzed cotton cellulose Cot45
determined from TEM images of negatively stained preparations; (b)
width distribution calculated from the projection of 6 nm thick objects
free-rotating around thir long axis and having the width distribution
shown in (a); (c) width distribution of Cot45 particles determined from
cryo-TEM images.
and experimental histograms were in good agreement, which
indicates that: (i) the width distribution of cotton nanocrystals
determined from images of negative stained preparations
satisfactorily represents the width distribution of free-rotating
flat particles in suspension; (ii) the composite aspect of the
particles is not an artifact from negative staining.
Particle Twist. The existence of a geometrical twist for
dispersed cellulose microfibrils has been reported by many
authors.33 The twist is especially clearly observed in well-
dispersed ribbon-shaped bacterial cellulose.31 However, it is not
clear whether the origin of the twist is biological, due to a
rotational movement of bacteria or enzyme complexes, or
physical, due to the chiral nature of cellulose, or a combination
of both. The existence of a twist has also been suggested for
cellulose microcrystals but rarely directly observed. AFM
topography images recorded by Hanley et al. have shown that
Micrasterias denticulata microfibrils were twisted ribbons with
a half-pitch of 600–800 nm.34 However, by analyzing electron
diffraction diagrams recorded on disincrusted Micrasterias cell
walls, Kim et al. had previously demonstrated that the bands
of parallel microfibrils were laid flat in the wall, exhibiting a
clear untwisted uniplanar-axial orientation.35 In addition, using
a similar approach, Sugiyama et al. have shown that microfibrils
Crystalline Nanoparticles from Native Cellulose
Figure 8. Two-dimensional histogram of Cot45 particle dimensions
(vertical bars) and distribution function obtained by fitting the histo-
gram with expression 4 (mesh net).
forming criss-crossed cellulose layers in the cell wall of Valonia
were not twisted along their axis.36 In fact, cellulose molecules
tend to twist because the most stable conformation around the
glycosidic linkage deviates slightly from the P21 symmetry
imposed by the crystal structure.37 The geometrical twist
observed along dispersed microfibrils could result from a
relaxation process occurring after disincrustation in support of
the physical origin. The hydrolysis of native cellulose would
promote the release of the topological constraints induced during
the deposition process.
In the present study, TEM and cryo-TEM images strongly
suggested that tunicin whiskers were twisted ribbons with a pitch
of 2.4–3.2 µm. Height variations detected in AFM longitudinal
profiles of single tunicin whiskers were also consistent with the
twisted ribbon morphology, although the analysis of the height
data was often complicated by the local superimposition of
whiskers. The existence of a twist could not be confirmed for
cotton and Avicel particles. Indeed, assuming a helical pitch
similar to that of tunicin whiskers, a regular twist would be
hardly detected in TEM images as most particles are signifi-
cantly shorter.
Influence of the Cellulose Origin and Purification Condi-
tions. Native cellulose crystals are considered to be formed as
the result of a concerted biosynthesis of several chains by a
complex of cellulose synthases located at the cell membrane:
the terminal complex (TC).38 Depending on the biological origin,
various arrangements of TCs are observed, producing cellulose
crystals with different geometries. For example, very large
rectangular arrangements of TCs were observed in algal cell
membranes in which 20 nm wide microfibrils with square cross-
sections are produced.39 In bacterial cellulose synthesized by
Acetobacter, a linear array of TCs results in ribbonlike mi-
crofibrils.40 For higher plants, the current model based on
microscopic observations suggests an arrangement of six-
membered rosettes, each of them producing six cellulose
chains.41 In this case, the elementary crystalline units, composed
of 36 chains, have a 3–4 nm diameter, corresponding to the
crystallite size found for wood microfibrils or particles from
acid-hydrolyzed Avicel. Although the larger size of cotton
crystallites cannot be explained by the rosette model (as is the
case of other bast fibers such as ramie or flax), the regular
particle thickness suggests a fairly strict regulation of the
synthesis.
The morphology of the microcrystals probably originates from
the two-dimensional nature of the microfibril formation, as
microfibrils are deposited in layers in plant cell walls. There
Biomacromolecules, Vol. 9, No. 1, 2008 63
is very little chance to find associations of segments of
microfibrils between each layer, whereas, inside each layer, the
microfibrils tend to align parallel due to geometric constraints,
which certainly promotes the lateral adhesion of crystallites.
The number of microfibrils laterally adhering to each other and
the length of bound regions along the microfibrils may statisti-
cally vary in the cell wall, which, after acid hydrolysis, would
result in a variety of particles with a large width distribution.
TEM images of ultrathin cross-sections of Halocynthia
papillosa tunicate mantle have shown that microfibrils were
fairly well individualized and that lateral associations were
unlikely.31,32 Similar TEM images of the Halocynthia roretzi
tunic would be helpful to characterize the in vivo distribution
of microfibrils and see if the lateral association of individual
crystallites may result from the disincrustation procedure.
Among the three samples, tunicin contained nanocrystals
with the lowest polydispersity in width but the highest
standard deviation for length (Table 1). In cellulose from
higher plants like ramie, periodical distributions of acid-
susceptible domains were shown to exist, using small-angle
neutron scattering42 or acid hydrolysis43 data resulting in a
stable level-off degree of polymerization. Tunicin probably
has no such regular distribution of less-organized domains
and is more similar to algal cellulose such as that from
Valonia for which the acid hydrolysis results in a higher
polydispersity in molecular weight, without any evidence of
a level-off degree of polymerization.44 The low polydispersity
in width probably comes from the sparse organization in the
starting material as mentioned above, which results in smaller
possibility of lateral cohesion.
Recently, Saito et al. prepared transparent microfibril suspen-
sions using never-dried native cellulose submitted to a selective
oxidation using the so-called TEMPO radical, followed by a
mechanical homogenizing treatment.45 The authors observed that
the lateral size of the microfibrils was very close to the crystallite
width measured by WAXS in the parent sample. This suggests
that the lateral aggregation of crystallites observed in our case
was most likely induced by the preparation treatments and might
be avoided. Indeed, microfibril suspensions prepared by oxida-
tion/homogenization may be used as starting material for acid
hydrolysis in order to decrease the polydispersity in width of
the resulting nanocrystals.
Influence of the Hydrolysis Conditions. For the cotton
sample, the variance s2 of the log-normal distribution increased
for length and decreased for width with increasing hydrolysis
temperature (Supporting Information Table S1). In a case where
each particle has a probability to split into two smaller units
that does not depend on its size, s2 should remain constant.25,26
For the length distribution, the increasing s2 comes from the
fact that hydrolysis proceeds from the ends of the particles that
can be more easily swollen by the acid. A higher temperature
would result in a faster hydrolysis at the end of the nanoparticles,
linearly shifting the histogram to smaller values, which would
increase s2.
For the width distribution, the decrease of s2 with higher
hydrolysis temperatures (resulting in smaller particles, Table
1) can be interpreted by assuming that the splitting of wider
particles is accelerated by harsher treatment. The hydrolysis does
not directly separate crystallites that are laterally bound by
secondary interaction. However, a hydrolysis carried out with
harsher conditions shortens the crystallites and forms more
charged sulfate half-ester groups at the surface, thus facilitating
the separation. Because the energy of adhesion of two crystallites
64 Biomacromolecules, Vol. 9, No. 1, 2008
Figure 9. Observed (red +) and simulated (blue –) small-angle X-ray scattering profiles (cross-section plots ln(QI) ) f(Q2)) from dilute suspensions
of cotton, Avicel, and tunicate cellulose particles.
would be proportional to the contact area, thinner or shorter
particles should be separated more easily.
Shape Distribution of the Particles. The width and length
measured from TEM micrographs of cotton nanocrystals
seemed to be somewhat correlated: in general, the longer
particles were also the wider. To verify this assumption, we
plotted two-dimensional histograms as a function of width
and length. The histograms were fitted with several empirical
functions, and the following expression was found to be the
most satisfactory:
Ω(L,I) )
(( ))
(ln (l sin (θ) + L cos (θ)) - ln b1 2 dimensional probability function Ω(L,l) calculated from the
a exp - × TEM images, with thickness e as the adjustable parameter.
m1
(( ))
l cos(θ) - L sin(θ) - b2 2
exp -
m2 ln(l sin(θ) + L cos(θ))
where a, θ, b1, b2, m1, m2 were the parameters to refine.
Moreover, we assumed that the particle length and width were
linearly related, according to the expression:
l ) L tan θ + k
where k is a constant. Ω(L,l) contains two terms: (i) a log-normal
function representing the size distribution of particles with an
aspect ratio following expression 5; (ii) a Gaussian function
that takes into account the deviation of the length and width of
the particles from expression 5. A small m2 means that the
relation between width and length is more strictly respected.
Using TEM images of negatively stained preparations, a
couple (L,l) was measured for each particle. As an example,
the two-dimensional histogram corresponding to Cot45 particles
is shown in Figure 8 along with the Ω(L,l) function fitting the
data. The values of θ, b1, b2, m1, m2 obtained for Cot45, Cot72,
Avi72, and Tun55 samples are listed in Supporting Information
Table S2.
Simulation of the SAXS Profiles. At low concentration,
assuming that there is no interparticle correlation, the intensity
scattered by the suspension is a sum of contributions from all
Elazzouzi-Hafraoui et al.
particles. The contribution of a rectangular particle can be
calculated by integrating its form factor F over all orientations:
F )
sin ( Lx2 )sin( ly2 )sin( ez2 ) (6)
Lx ly ez
2 2 2
where L, l, and e are the length, width, and thickness of the
particle, respectively. A series of SAXS profiles were calculated
for a range of width and length, assuming a constant thickness.
All contributions were finally sum-weighted using the two-
Figure 9 shows the experimental SAXS data collected from
Cot45, Cot72, Avi72, and Tun55 suspensions, plotted as ln(IQ2)
(4) ) f(Q2). As previously mentioned, all plots are convex and do
not show any linear region in the Q range, satisfying QRg , 1.
SAXS curves were simulated for different particle thicknesses,
using the two-dimensional function Ω(L,l) determined from
particle width and length distributions, and superimposed to the
experimental profiles. The thickness was adjusted to obtain
(5) the better fit between simulated and experimental curves. The
particle thickness was estimated to be about 6 nm for cotton,
3.8 nm for Avicel, and 10 nm for tunicin. These values are in
good agreement with the AFM height measurements and with
the crystallite size estimated from the peak broadening of WAXS
diagrams.
Conclusion
For the three sources of cellulose considered in this study,
the crystalline nanoparticles resulting from a hydrolysis/soni-
cation treatment were rather flat and constituted of a few laterally
bound elementary crystallites that were not separated by
conventional acid hydrolysis and sonication process. In par-
ticular, the particles from cotton and Avicel cellulose, which
are classic samples from secondary cell wall, showed width
distributions following log-normal functions. We have shown
Crystalline Nanoparticles from Native Cellulose
that the description of cellulose nanoparticles as monodisperse
cylinders was not satisfactory to account for the X-ray scattering
properties of the suspensions. A more sophisticated approach,
taking into account the data collected with a large panel of
imaging and scattering techniques, was developed, allowing the
statistical description of the assembly of cellulose crystalline
nanoparticles. However, the relation between the biosynthetic
pathways and the morphology remains an open question. Besides
the biological relevance, the precise knowledge of the morphol-
ogy and size distribution of the particles is crucial to analyze
quantitatively statistical phenomenon like phase separation in
particle suspensions or percolation of rodlike objects. Indeed,
the determination of the particle density strongly depends on
the precision of the geometrical description of the objects.
Together with polydispersity, density is a key parameter that
influences the phase separation and the formation of chiral
nematic organizations in cellulose nanocrystal suspensions.
Acknowledgment. We thank ESRF and the Innovation
Center for Nanobiotechnologies (Nanobio, Grenoble) for grant-
ing access to SAXS and AFM facilities, respectively. We are
also grateful to H. Chanzy and B. Jean (CERMAV) for
stimulating discussions and I. Pignot-Paintrand for assistance
with TEM observations.
Supporting Information Available. Mode (m) and squared
standard deviation (s2) of the log-normal functions describing
the width (l) and length (L) distributions of particles from acid-
hydrolyzed cotton, Avicel, and tunicate cellulose. Parameters
of the function defined in expression 4 and used to fit
experimental width and length distribution histograms shown
in Figure 3. Wide-angle X-ray diffraction diagrams recorded
on films of cotton, tunicin, and Avicel cellulose particles. Cross-
sectional view of the structural model of cellulose I. This
material is available free of charge via the Internet at http://
pubs.acs.org.
References and Notes
(1) Rånby, B. G. Discus. Faraday Soc. 1951, 11, 158–164.
(2) Marchessault, R. H.; Morehead, F. F.; Koch, M. J. J. Colloid Sci.
1961, 16, 327–344.
(3) Mukherjee, S. M.; Woods, H. J. Biochim. Biophys. Acta 1953, 10,
499–511.
(4) Revol, J.-F.; Bradford, H.; Giasson, J.; Marchessault, R. H.; Gray,
D. G. Int. J. Biol. Macromol. 1992, 14, 170–172.
(5) Revol, J.-F.; Marchessault, R. H. Int. J. Biol. Macromol 1993, 15,
329–335.
(6) Abeysekera, R. M.; Willison, J. H. M. Biol. Cell 1990, 68, 251–257.
(7) Roland, J. C.; Reis, D.; Vian, B.; Satiat-Jeunemaitre, B.; Mosiniak,
M. Protoplasma 1987, 140, 75–91.
(8) Dong, X. M.; Revol, J. -F.; Gray, D. G. Cellulose 1998, 5, 19–32.
(9) Li, J.; Revol, J.-F.; Marchessault, R. H. J. Colloid Interface Sci. 1997,
192, 447–457.
(10) Dong, X. M.; Kimura, T.; Revol, J.-F.; Gray, D. G. Langmuir 1996,
12, 2076–2082.
(11) Li, J.; Revol, J.-F.; Naranjo, E.; Marchessault, R. H. Int. J. Biol.
Macromol. 1996, 18, 177–187.
Biomacromolecules, Vol. 9, No. 1, 2008 65
(12) Araki, J.; Kuga, S. Langmuir 2001, 17, 4493–4496.
(13) Furuta, T.; Yamahara, E.; Konishi, T.; Ise, N. Macromolecules 1996,
29, 8994–8995.
(14) Orts, W. J.; Godbout, L.; Marchessault, R. H.; Revol, J.-F. Macro-
molecules 1998, 31, 5717–5725.
(15) Belamie, E.; Davidson, P.; Giraud-Guille, M. M. J. Phys. Chem. B
2004, 108, 14991–15000.
(16) Heux, L.; Chauve, G.; Bonini, C. Langmuir 2000, 16, 8210–8212.
(17) Araki, J.; Wada, M.; Kuga, S. Langmuir 2001, 17, 21–27.
(18) Favier, V.; Chanzy, H.; Cavaillé, J. Y. Macromolecules 1995, 28,
6365–6367.
(19) Nishiyama, Y.; Kuga, S.; Wada, M.; Okano, T. Macromolecules 1997,
30, 6395–6397.
(20) De Souza Lima, M. M.; Wong, J. T.; Paillet, M.; Borsali, R.; Pecora,
R. Langmuir 2003, 19, 24–29.
(21) Elazzouzi, S. Doctoral Thesis, Université Joseph Fourier, Grenoble,
2006.
(22) Sugiyama, J.; Persson, J.; Chanzy, H. Macromolecules 1991, 24, 2461–
2466.
(23) Harris, J. R. NegatiVe Staining and Cryoelectron Microscopy: The
Thin Film Techniques; BIOS Scientific Publishers: Oxford, 1997.
(24) Klug, H. P.; Alexander, L. E. In X-Ray Diffraction Procedures for
Polycrystalline and Amorphous Materials; Wiley Interscience: New
York, 1954, p 491.
(25) Epstein, B. Franklin Inst. 1947, 244, 471–477.
(26) Ishii, T.; Matsushita, M. J. Phys. Soc. Jpn. 1992, 61, 3474–3477.
(27) Cheng, Z.; Redner, S. J. Phys. A: Math. Gen. 1990, 23, 1233–1258.
(28) Sugiyama, J.; Vuong, R.; Chanzy, H. Macromolecules 1991, 24, 4168–
4175.
(29) Araki, J.; Wada, M.; Kuga, S.; Okano, T. Colloids Surf. A 1998, 142,
75–82.
(30) Terech, P.; Chazeau, L.; Cavaillé, J. Y. Macromolecules 1999, 32,
1872–1875.
(31) van Daele, Y.; Revol, J.-F.; Gaill, F.; Goffinet, G. Biol. Cell 1992,
76, 87–96.
(32) Helbert, W.; Nishiyama, Y.; Takeshi, O.; Sugiyama, J. J. Struct. Biol.
1998, 124, 42–50.
(33) Ruben, G. C.; Bokelman, G. H.; Krakow, W. In Plant Cell Wall
Polymers; Lewis, N. G., Paice, M. G., Eds.; ACS Symposium Series,
Vol. 399; American Chemical Society: Washington, DC, 1989; pp
278–298.
(34) Hanley, S. J.; Revol, J.-F.; Godbout, L.; Gray, D. G. Cellulose 1997,
4, 209–220.
(35) Kim, N.-H.; Herth, W.; Vuong, R.; Chanzy, H. J. Struct. Biol. 1996,
117, 195–203.
(36) Sugiyama, J.; Chanzy, H.; Revol, J. F. Planta 1994, 193, 260–265.
(37) French, A. D.; Johnson, G. P. Cellulose 2004, 11, 449–462.
(38) Brown, R. M., Jr.; Montezinos, D. Proc. Natl. Acad. Sci. U.S.A. 1976,
73, 143–147.
(39) Ito, T.; Brown, R. M., Jr. Protoplasma 1988, 144, 160–169.
(40) Brown, R. M., Jr.; Willison, J. H. M.; Richardson, C. L. Proc. Natl.
Acad. Sci. U.S.A 1976, 73, 4565–4569.
(41) Read, S. M.; Bacic, T. Science 2002, 295, 59–60.
(42) Nishiyama, N.; Kim, U.-J.; Kim, D.-Y.; Katsumata, K. S.; May, R. P.;
Langan, P. Biomacromolecules 2003, 4, 1013–1017.
(43) Battista, O. A. Ind. Eng. Chem. 1950, 42, 502–507.
(44) Kuga, S.; Mutoh, N.; Isogai, A.; Usuda, M.; Brown, R. M., Jr. In
Cellulose, Structure and Functional Aspects; Kennedy, J. F., Philips,
J. O., Williams, P. A., Eds.; Ellis Horwood: Chichester, UK, 1989;
pp 75–80.
(45) Saito, T.; Nishiyama, Y.; Putaux, J.-L.; Vignon, M.; Isogai, A.
Biomacromolecules 2006, 7, 1687–1691.
BM700769P