10.1007-s10570-013-9972-9-Printed copy

download 10.1007-s10570-013-9972-9-Printed copy

of 12

Transcript of 10.1007-s10570-013-9972-9-Printed copy

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    1/12

    O R I G I N A L P A P E R

    Estimation of cellulose nanofibre aspect ratio

    from measurements of fibre suspension gel point

    Swambabu Varanasi Rongliang He

    Warren Batchelor

    Received: 6 February 2013 / Accepted: 9 June 2013 / Published online: 26 June 2013

    Springer Science+Business Media Dordrecht 2013

    Abstract Cellulose nanofibre aspect ratio controls

    the properties of sheets made from nanofibres and

    processing conditions, but aspect ratio is very difficult

    to measure. In this paper, aspect ratio was estimated

    from the gel point of a cellulose nanofibre suspension,

    the solids concentration at which the transition from a

    dilute to a semi-dilute suspension occurs. Four batches

    of cellulose nanofibres were tested. Two were pro-

    duced from softwood fibres using ball milling. Com-

    mercially produced microfibrillated cellulose material

    was also used, both in as supplied form and afterremoval of the larger fibres by filtering. The average

    diameter measured from SEM images of fibres ranged

    from 33 to 73 nm. One sample was too heavily treated

    and an average dimension could not be measured. The

    gel-point was measured both from the height of a layer

    of cellulose nanofibres sedimented from a dilute

    suspension or from the lowest solids concentration at

    which a yield stress could be measured using a vane

    rheometer. The two methods were closely in agree-

    ment for all samples. Aspect ratio was then calculated

    using either the effective medium (EMT) or crowding

    number (CN) theories. Aspect ratio calculated with an

    assumed fibre density of 1,500 kg/m3, using the CN

    theory ranged from 155 to 60. Use of the EMT theory

    reduced the calculated aspect ratio by between 11 and

    23 %. Reducing the assumed density in suspension

    from 1,500 to 1,166 kg/m3 reduced the calculated

    aspect ratio by 1214 %. The heavily treated sample

    had by far the lowest aspect ratio.

    Keywords Cellulose nanofibres Gel point Aspect ratio Yield stress Sedimentation

    Introduction

    Cellulose is the most common organic polymer in the

    world, representing 1.59 1012 tons of total annualbiomass growth, mostly in plants(Malcolm Brown et al.

    1996). Since it is formed from CO2 gas in the

    atmosphere via photosynthesis in plants, it is a sustain-

    able and renewable source of material. Plant fibres

    consist of cellulose nanofibrilsof 24 nm in diameteras a

    structural unit which is made up of cellulose crystallites

    of*4 nm in diameter (Malcolm Brown et al. 1996;

    Jakob et al.1995). Green properties such as biocom-

    patibility and biodegradability are also expected from

    S. Varanasi W. Batchelor (&)Department of Chemical Engineering, Australian Pulp

    and Paper Institute, Monash University,

    Melbourne, Australia

    e-mail: [email protected]

    S. Varanasi

    e-mail: [email protected]

    R. He

    Institute for Frontier Materials, Deakin University,

    Geelong, VIC 3217, Australia

    e-mail: [email protected]

    1 3

    Cellulose (2013) 20:18851896

    DOI 10.1007/s10570-013-9972-9

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    2/12

    these naturally-occurring raw materials, which are

    significantly advantageous for biomedical and environ-

    mental applications over synthetic organic and inor-

    ganic nanomaterials (Czaja et al. 2006). Over the last

    two decades, the synthesis of cellulose nanofibres and

    their applications in various fields have gained increas-

    ing attention due to their high strength and stiffness,combined with low weight, biodegradability and

    renewability (Siroand Plackett2010). Cellulose nano-

    fibres can be used as reinforcing fibres in bio compos-

    ites, strength additives (Ahola et al. 2008), and to

    produce hydrogels, anti-microbial films (Czaja et al.

    2006), and high-technology devices (Nogi et al.2009).

    Although there are many methods available for

    production of cellulose nanofibres (Eichhorn et al.

    2010), one significant area still being developed is the

    characterization of nanofibre dimensions (Zhang et al.

    2012). Whilst a combination of microscopic techniqueswith image analysis can provide information on cellu-

    lose nanofibre widths (Dufresne et al. 2000; Taniguchi

    and Okamura1998; Janarthanan and Sain 2006), it is

    more difficult to determine nanofibre lengths due to

    entanglements and difficulties in identifying both ends

    of individual nanofibres (Henriksson et al.2008; Ishii

    et al. 2011). Recently, Zhang et al. used a sedimentation

    method (Zhang et al. 2012) for measuring the aspect

    ratio of nanofibres based on the method proposed by

    Martinez et al. (Martinez et al. 2001) for wood pulp

    fibres. Using this method, the aspect ratio is estimatedbased on the gel point concentration, the threshold

    consistency at which a continuous network of fibres in

    suspension forms. Below the gel point concentration,

    the suspension does not contribute to the mechanical

    strength (Derakhshandeh et al. 2011). The measured

    aspect ratio has been shown to correlate with the

    concentration range over which cellulose nanofibre

    sheets could be formed using filtration (Zhang et al.

    2012; Varanasi and Batchelor 2013). One criticism that

    can be made of this method is that in order to observe a

    visible sedimented layer the fibres must form structuresthat can scatter light in the visible range, thus requiring

    complexes several hundred nanometres in size. An

    alternative method to probing fibres in suspension that

    avoids this problem is to study the suspension rheology.

    The types of rheological measurements that can be

    applied to fibre suspensions will depend on the solids

    content of the suspension, i.e. whether the suspension

    can be classified as dilute, semi-dilute or concentrated.

    One previous report (Ishii et al. 2011) of cellulose

    nanofibre length estimation used visco-elastic mea-

    surements of dilute suspensions and estimated the fibre

    length using the theory of linear visco-elasticity of

    semi-flexible rod-like polymers in dilute suspension,

    estimating the fibre length as 2.2 lm for 4 nm

    diameter fibres, giving a very high aspect ratio of 550.

    Other measurements in the concentrated and semi-dilute ranges include suspension yield stress measure-

    ments (Mosse et al. 2012b) to investigate the interac-

    tion of cationic polymers with cellulose fibres and

    shear viscosity measurements for characterizing the

    dispersion of carbon nanofibre suspensions colloidal

    systems (Xu et al.2005) or TEMPO oxidised cellulose

    nanofibre suspensions (Lasseuguette et al. 2008).

    These are a particularly sensitive probe of fibre

    suspensions because in the semi-dilute and concen-

    trated regimes the rheological properties are often

    dominated by the interaction and entanglement of thefibres, which in turn depend upon of the structural

    properties such as diameter, length and aspect ratio.

    In this paper we present a new method for

    estimating the gel-point of cellulose nanofibre sus-

    pensions by measuring suspension yield stress as a

    function of concentration. We compare the results to

    that obtained by sedimentation and estimate aspect

    ratio by both methods.

    Theory

    There are three concentration regimes for particles

    dispersed within a medium. Particles in suspension can

    be in dilute, semi-dilute or concentrated states. Of

    interest here is the boundary between these states. If the

    volume concentration is denoted by U then the connec-

    tivity threshold,Uc, is the boundary between the dilute

    and semi-dilute regions, the lowest volume fraction

    where the particles first form a continuous network and

    the rigidity threshold,Ur, is the boundary between the

    semi-dilute and concentrated regions, which is thelowest volume fraction at which the particle suspension

    will exhibit mechanical stability under load (Celzard

    etal. 2008). The connectivity threshold is also called the

    gel point (Martinez et al.2001) .

    In the work here, we shall consider the application

    of the semi-empirical Crowding Number (CN) and of

    the more rigorously derived Effective Medium theory

    (EMT) to characterise the transition between states of

    fibres in suspension.

    1886 Cellulose (2013) 20:18851896

    1 3

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    3/12

    EMT was originally developed to solve problems

    such as the conductivity of a material consisting of

    particles dispersed in a medium. Of most relevance is

    the extension of the theory to consider spheroids as the

    dispersed particles (Celzard et al. 2000) and the

    successful application to predicting the conductivity

    of spheroid graphene dispersed in a matrix of air. Forprolate or oblate ellipsoids, Ucis given (Celzard et al.

    2000) by

    Uc 9Lc 1 Lc 2 Lc 15 9Lc 1

    whereLcis the depolarization factor of the particles. If

    the particle is modelled as a prolate ellipsoid with

    equal minor axes, a and b then (Celzard et al.2000)Lcis given by

    Lc 1 e2

    2e3 ln

    1 e1 e

    2e

    2

    where e is the eccentricity which ise ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

    1 a=c2q

    where c is the major axis of the ellipsoid and when

    expressed in terms of the fibre aspect ratio, A(= c/a),

    yields e ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

    1 A2p

    . The rigidity threshold is then

    related to the connectivity threshold byUr= 4Uc.

    The Crowding Number (CN) theory has been

    developed by Kerekes and Schell, based on earlier

    work by Mason. The crowding number, N, is given byKerekes and Schell (1992) as

    N 2=3 U l=d 2 2=3 UA2 3where l and d are the fibre length and diameter,

    respectively, of the fibres in suspension. Based on the

    number of contacts developed in a suspension, the

    rigidity threshold has been identified as occurring at

    N = 60 (Celzard et al. 2009) from which

    Ur 90=A2 4

    The connectivity threshold was experimentally deter-mined to be 16 4 (Martinez et al. 2001) from

    analysis of PET measurements made on dilute fibre

    sedimentation experiments, from which

    Uc 24=A2 5The relationship between aspect ratio, A, and Urand

    Ucis shown in Fig.1.

    The aspect ratio can be calculated from either the

    connectivity or rigidity thresholds by the following

    equations as given in Table 1, where the equations for

    the EMT theory were determined by fitting the data in

    Fig.1.

    Finally, we are most readily able to measure the

    solid fraction and not the volume fraction. If we wish

    to relate the volume fraction to the solids fraction, C

    (kg fibre/kg of suspension), then we may write

    C qfU= qfU ql 1 U

    where qf and ql are

    the density of the fibres and liquid, respectively, which

    if U 1 can be simplified as C qf=ql

    U or

    U ql=qf

    C. The solids fraction forthe connectivity

    and rigidity thresholds are then denoted Cc and Cr,

    respectively. The appropriate fibre density to useremains an open question. The density of the cellulose

    nanofibres in the dry state is approximately 1,500 kg/m3,

    however it is not clear that this is applicable to a

    suspension of cellulose nanofibres in water, where a

    reduction of density due to an uptake of water into the

    structure is likely. Recent measurements of moisture

    diffusion of softwood, hardwood and cellulose nano-

    fibre samples have suggested that two phases of water

    exist, a slow diffusing phase tightly bound with the

    cellulose fibrils and a faster diffusing phase more

    loosely bound (Perkins and Batchelor 2012). Inparticular the measurements suggested that the tightly

    bound phase had a maximum capacity of approxi-

    mately 0.5 g water/g fibre (Perkins and Batchelor

    2012). The addition of this water to the fibres would

    reduce the density of the cellulose nanofibres in

    suspension to (2/3)9 1,500 ? (1/3) 9 1,000 kg/m3

    = 1,333 kg/m3. There exists a further complication

    that the samples tested here are not fully separated. If

    this is the case then they will form connected

    0

    20

    40

    60

    80

    100

    120

    140

    160

    180

    0 0.01 0.02 0.03 0.04 0.05 0.06

    AspectRaio

    Phi

    PhiC (EMT)

    PhiC(CN)

    PhiR(EMT)

    PhiR(CN)

    Fig. 1 Aspect ratio predicted from the connectivity (PhiC (Uc))

    and Rigidity (PhiR (Ur)) thresholds using the EMT and CN

    theories

    Cellulose (2013) 20:18851896 1887

    1 3

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    4/12

    assemblies with reduced density due to the water

    between the connected fibres. Given the uncertainties

    in the density we will calculate aspect ratios estimated

    from threshold solids concentration using three differ-

    ent densities using the equations given in Table2

    below.

    Experimental

    Material

    Cellulose nanofibres were prepared using two different

    methods. In the first method, nanofibres were prepared

    as reported previously (Varanasi et al. 2012; Varanasi

    and Batchelor 2013). Micro fibrillated cellulose (MFC)

    supplied from DAICEL Chemical Industries Limited

    (grade Celish KY-100G) was used as the starting

    material for preparing cellulose nanofibres as well as

    being tested. The material contains a mixture of

    nanofibres and larger fibres and particles. The large

    fibres and particles were filtered from the MFC sample

    using a solids concentration of 0.5 kg/m3 and two

    fabric filters with 100 lm openings to retain the larger

    fibres on the fabric. The filtrate was centrifuged at

    5,000 rpm for 20 min. After centrifuging, the super-

    natant was discarded and only the nanofibres at the

    bottom of tubes were collected. The yields of the

    filtration process before and after centrifugation were

    21.6 and 20 %, respectively. More details are given in

    Zhang et al. (2012). The starting material was also used

    for measurements. In this paper, the original sample

    and the sample prepared by filtration have been

    labelled as MFC and NF, respectively.

    In the second method, cellulose nanofibres were

    prepared using ball milling with a SPEX 8000 shaker

    mill. Firstly, cellulose pulp sheet (NIST reference

    material 8495) was cut into 5 9 5 cm pieces and

    soaked in deionised water overnight in the refrigerator

    to prepare 1 wt% solid suspension. The wet cellulose

    pieces were then shredded using a conventionalkitchen blender and then stirred at 70 C overnight.

    Then 20 g of 1 wt% cellulose pulp suspension, 45 g of

    cerium-doped Zirconium balls (0.5 mm in diameter)

    and 20 mL of deionised water were then placed in a

    70 ml polypropylene container and milled using Spex

    8000 ball mill for 60 and 90 min, for the NIST60 and

    NIST90 grades, respectively. The final suspension was

    filtered using a polyester mesh (opening size 125

    micron) to remove the zirconium balls and any larger

    remaining fibres. The filtrate was then stored in the

    refrigerator and used as required. Results from earlierexperiments to manufacture cellulose nanofibres from

    this starting material have been reported in Zhang et al.

    (2012).

    Diameter distribution

    SEM Images of MFC and Nanofibre samples were

    taken using a JEOL SEM (JSM-7001F FEGSEM) and

    NIST60 and NIST90 were taken using Supra 55 V

    SEM. Samples were highly diluted and a drop of

    suspension was cast on a metal plate, air dried and thenplatinum coated. Coated samples were imaged using

    SEM. The magnification ranged from 3,000 to 80,000.

    Diameter distribution of MFC, NF and NIST60

    samples were measured from SEM images. NIST90

    samples only contained a few fibres and square

    particles so the diameter distribution was not mea-

    sured for this sample.

    From each image, the width of each nanofibre

    observable in the image was measured using ImageJ.

    The fibre diameters measured in each image were then

    sorted into bins of 10 nm size and normalized tocounts/m2, by dividing by the total area of the image.

    Table 1 Equations to calculate aspect ratio using the EMT

    and CN theories

    Uc (EMT) Ur(EMT) Uc (CN) Ur(CN)

    A 2:52U0:58c A 5:64U0:58r A 4:90U0:5c A 9:49U0:5r

    Table 2 Equations used

    for calculating aspect ratio

    from solids fraction, C

    Assumed density,

    qf(kg/m3

    )

    EMT, Cc EMT, Cr CN, Cc CN, Cr

    1,500 A 3:19C0:58c A 7:13C0:58r A 6:0C0:5c A 11:61C0:5r1,333 A 2:98C0:58c A 6:66C0:58r A 5:66C0:5c A 10:95C0:5r1,166 A 2:76C0:58c A 6:17C0:58r A 5:29C0:5c A 10:24C0:5r

    1888 Cellulose (2013) 20:18851896

    1 3

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    5/12

    Fractions were calculated by dividing the individual

    bin counts/m2 with total counts/m2. In the event that a

    bin had fibres measured at different magnifications,

    the highest value of counts/m2 was used. Further

    details of the method are given in Zhang et al. (2012).

    Yield stress measurement

    Yield stress was measured on fibre suspensions using

    the Haake Rotovisco RV20 system with four vanes of

    38 mm (diameter) by 76 mm (length). This method

    was described in more detail in Mosse et al. (2012a, b).

    The suspension with known concentration was placed

    in a beaker and the vane fully immersed. The maximum

    torque, quoted from the user manual, measurable by

    the instrument was 0.049 Nm at 100 % full scale.

    Rheometer can be operated at three different scales, 1,

    10 and 100 % torque. Measurements were carried out

    using the 1 % scale. Maximum measurable torque

    value at 1 % scale is 4.9 9 10-4 Nm. The instrument

    output is in volts. The smallest subdivision is 1/1,000 of

    the maximum measureable torque and thus the lowest

    possible torque value that can be measured is

    4.9 9 10-7 Nm. All the values discussed in the results

    are within the 1 % scale range. Beaker dimensions are

    72 mm diameter and 92 cm length, and the beaker

    diameter is much larger than the vane diameter,

    minimising the chance of slip at the wall. Prior to each

    measurement series at a given solids concentration, the

    pulp suspension was stirred thoroughly using the vane

    rotating at 500 rpm (the highest speed setting),

    removing any air bubbles in the pulp and ensuring that

    the pulp was evenly distributed throughout the beaker.

    Each sample was mixed for a minimum of 2 min until

    the suspension was completely clear. Five runs were

    made for each sample at a given solids concentration.

    Before each run, the sample was mixed for 10 s at

    500 rpm, before being allowed to stand for 30 s.

    Following this, a measurement was taken by initiating

    shear flow at a rotation rate of 0.65 rpm, and recording

    the torque measured as a function of time.

    Yield stress is the maximum stress sustained by the

    suspension at the onset of continuous motion. The

    peak torque in each start-up of shear flow experiment

    was converted to yield stress using following equation

    (Dzuy and Boger1985):

    Tm pD3

    2

    H

    D 1

    3

    sy 6

    where Tm is the maximum torque measured, D is the

    diameter of the vane,His the height of the vane, and syis the yield stress.

    Sedimentation

    Estimates of nanofibre suspension gel point were madeby sedimentation experiments. This method was

    described in more detail in Zhang et al. (2012). The

    method was adapted for nanofibre suspensions from

    calibration curves published for wood pulps by Marti-

    nez et al. (Martinez et al. 2001). 250 ml of nanofibre

    suspensions with solids concentration (=solids frac-

    tion 9 suspension density) ranging from 0.5 to 5 kg/m3

    were decanted into measuring cylinders. The suspen-

    sion was agitated to suspend the fibres completely in the

    cylinder, andthen the fibres were allowed to settle. Once

    the fibres settled down completely, the height ofsediment in the cylinder was measured. A minimum

    of 48 h was allowed for the sedimentation process.

    Results and discussion

    Fibre cross-section measurements

    Figures2and3show the SEM images and distribution

    of fibre diameter of the MFC sample, while the

    corresponding data for the NF sample derived fromthe MFC sample are given in Figs. 4 and5. Figures2

    and 3 show that the MFC sample contains nanofibres as

    well as a few large fibres. The mean fibre diameter was

    73 nm. These largefibres were mostly removed through

    filtration, which reduced the mean diameter to 53 nm.

    Figures6and7show that NIST60 sample contains

    uniform nanofibres with average diameter of 33 nm.

    The milling time of 60 min has produced a very

    uniform sample with low fibre diameter. Figure8

    shows an image of the NIST90 sample. This was milled

    for 90 min. The increased milling time has clearlybroken down the nanofibres as observed in Fig. 8, as

    the sample consists mostly of irregularly shaped

    particles of low aspect ratio with only a few fibres.

    Sedimentation data to calculate suspension gel

    point

    The sedimentation data obtained for all four samples is

    shown in Fig. 9. For each sample, the sedimentation

    Cellulose (2013) 20:18851896 1889

    1 3

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    6/12

    behaviour was measured of fibre suspensions with

    different dilute concentrations. A graph of initial solid

    concentration (C0) versus the ratio of sediment height

    (hs) to initial suspension height (H0) was then plotted

    Fig. 2 SEM image of MFC sample

    Fig. 3 Diameter distribution of MFC sample

    Fig. 4 SEM image of Nanofibres sample

    Fig. 5 Distribution of fibre diameters for the nanofibre sample

    Fig. 6 SEM Image of NIST60 sample

    Fig. 7 Distribution of fibre diameters for the NIST60 sample

    1890 Cellulose (2013) 20:18851896

    1 3

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    7/12

    and fitted with a quadratic equation. The linear term of

    the fit then gives the gel concentration, gc. The resultsof the fits, fitting statistics and confidence intervals are

    given in Table3.

    Finally, it is interesting to consider whether the

    rigidity threshold can be applied to this data. As

    discussed previously, the rigidity threshold occurs at a

    volume concentration that is 3.75 times (Crowding

    number) or 4 times (EMF theory) (Celzard et al.2009)

    the connectivity threshold. The fits to the data show

    that these concentrations are only achieved at solids

    concentrations well above what would be required to

    produce Hs/H0 = 1, i.e. the lowest concentration

    where the suspension does not sediment, suggesting

    that the rigidity threshold is not applicable to sedi-

    mentation data.

    Yield stress measurements

    Yield stress is one of the most important rheological

    properties of semi-dilute or concentrated fibre suspen-

    sions. Unless it is exceeded, flow does not take place.

    Yield stress is the maximum stress reached when strain

    is increased to initiate flow, after which stress

    decreases. The maximum stress is called the ultimate

    shear strength and used as a measure of apparent yield

    stress (Liddel and Boger1996). A yield stress occurs

    because the fibres form a continuous network, wherebyfibres are restrained from moving because of contacts

    with other fibres (Derakhshandeh et al. 2011). The

    corollary is that a dilute fibre suspension will display no

    yield stress. In the work here yield stress is measured as

    a function of solids concentration in order to determine

    the transition from a dilute fibre suspension to a semi-

    dilute fibre suspension, i.e. the solids concentration at

    which a continuous network first forms.

    There are many methods available to measure yield

    stress. We used a vane rheometer, whereby the torque

    was measured as a function of time. If the fibresuspension has a yield stress at that concentration, then

    the torque value reaches maximum and then decreases

    (Derakhshandeh et al. 2011). If the fibre suspension

    does not have a yield stress at that concentration,

    torque profile of fibre suspension will be like the

    torque profile of water. This maximum torque value is

    called the peak torque in the rest of discussion.

    Torque values as function of time at different

    concentrations are shown in Figs.10, 11, 12 and 13

    for MFC, NF, NIST90 and NIST60 suspensions,

    respectively. In Fig.10, three replicates of torque

    Fig. 8 SEM Image of NIST90 sample

    Fig. 9 Sedimentation data

    Table 3 Gel point solids

    concentrations from

    sedimentation method and

    the standard errors of the fit

    given in brackets

    Sample Fitting equation Gel point solids

    concentration (kg/m3

    )

    Solids mass

    fraction

    MFC y = 2.03x2? 1.78x, R

    2= 0.98 1.78 (0.54) 0.00178

    NF y = 1.98x2? 2.30x, R

    2= 0.99 2.30 (0.27) 0.00230

    NIST90 y = 5.65x2? 9.35x, R

    2= 0.99 9.35 (0.50) 0.00935

    NIST60 y = 1.59x, R2= 0.99 1.59 (0.02) 0.00159

    Cellulose (2013) 20:18851896 1891

    1 3

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    8/12

    values as function of times for solids concentration of

    2.5 kg/m3 are shown. These show that results were

    consistent and reproducible. The deviation between

    runs for a given sample was at most 10 %. There were

    several interesting points from Figs. 10,11,12and13.

    Figures10 and11contain measurements made at 0 %

    solids concentration (distilled water only)and show thatthe vane rheometer has a small level of intrinsic drift.

    All measurements showed a small initial increase

    associated with commencement of rotation of the vanes

    followed by a drift downwardsas the rotation continued.

    The maximum peak torque observable when measuring

    water alone was approximately 7.2 9 10-6 N m. The

    effect of testing below and above the gel point can be

    seen very clearly in Fig.11. The test at a solids

    concentration of 2 kg/m3 was essentially indistinguish-

    able from the water curve indicating that the fibre

    suspension was dilute and the fibres did not form acontinuous network. Increasing the solids concentration

    to 2.5 kg/m3 however produced an observable yield

    stress with torque increasing to a yield stress of

    2.5 9 10-5 N m after 10 s before flattening out. Some

    other cases are less clear cut. For example the curve for

    the MFC sample at 1.0 kg/m3 solids concentration

    showed a similar shape to other solids concentrations

    which displayed a yield stress but the maximum torque

    reached was similar to that produced by water alone.

    The first unambiguous evidence of a yield stress was

    only observed when the solids concentration wasincreased to 1.5 kg/m3. An observable peak torque

    was observed at minimum concentration of 2.5 kg/m3

    for NF suspension, 10 kg/m3 for NIST90 suspension

    and 1.5 kg/m3 for NIST60 suspension.

    For the MFC and NF samples, Cc was determined

    by fitting the data of yield stress versus concentration,

    as described below. However, there was insufficient

    sample for the NIST90 and 60 samples, so for these

    two samples, Cc was determined as the lowest solids

    fraction at which a yield stress could be measured.

    For the MFC and NF samples, yield stress mea-

    surements were carried for the samples at a range of

    0

    0.00005

    0.0001

    0.00015

    0.0002

    0.00025

    0.0003

    0.00035

    0.0004

    0.00045

    0 10 20 30 40

    Torque

    (N.m

    )

    Time (sec)

    2.5

    2.5

    2.5

    1.5

    1.0

    0.5

    0

    Solids

    Concentration

    (kg/m3)

    Fig. 10 Torque measured from instrument as function of time

    for MFC sample and three individual measurement for solids

    concentration 2.5 kg/m3

    0

    0.000005

    0.00001

    0.000015

    0.00002

    0.000025

    0.00003

    0 5 10 15 20 25

    Torque(N.m

    )

    Time (sec)

    2.5

    2.0

    0

    Solids

    Concentration

    (Kg/m3)

    Fig. 11 Torque measured from instrument as function of time

    for NF sample

    0.000000

    0.000005

    0.000010

    0.000015

    0.000020

    0.000025

    0.000030

    0.000035

    0.000040

    0 5 10 15 20 25

    Torque(N.m

    )

    Time (sec)

    10

    8.0

    6.0

    SolidsConcentration

    (Kg/m3

    )

    Fig. 12 Torque measured from instrument as function of time

    for NIST90 Sample

    0.000000

    0.000005

    0.000010

    0.000015

    0.000020

    0.000025

    0 2 4 6 8

    Torque(N.m

    )

    Time (sec)

    2.5

    1.5

    Solids

    Concentration(Kg/m3)

    Fig. 13 Torque measured from instrument as function of time

    for NIST60 sample

    1892 Cellulose (2013) 20:18851896

    1 3

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    9/12

    different concentrations at which peak torque was

    observed. Yield stress was calculated from peak

    torques at different concentrations higher than the

    minimum concentration as mentioned in the experi-

    mental procedure for MFC and NF suspensions and

    plotted against concentration in Figs.14 and 15.

    Increasing suspension concentration increases yieldstress because the increase in the number of fibre

    contacts and entanglements, increases the strength of

    the fibre network formed. The dependency of the yield

    stress values on fibre weight percentage (=solids

    concentration 9 100/suspension density) has been

    correlated using a power-law model sy aCbm, whereCmis the fibre weight percentage. The resulting fitted

    constants are shown in the Table 4. These values are

    consistent with literature values for wood pulp fibre

    suspensions, which for a ranged from 1.8\ a\ 24.5

    while values of b ranged from 1.69\ b\ 3.02

    (Kerekes et al. 1985). Many key variables contributing

    to these differences were not measured or reported

    (Derakhshandeh et al. 2011). For example, Dalpke and

    Kerekes (2005) reported that a and b values depend

    upon aspect ratio (Dalpke and Kerekes 2005). The

    fitting constants for MFC and NF suspensions are

    within the range of previously reported results. Yield

    stress value reported for MFC suspension at concen-

    tration of 5 kg/m3 is 0.7 Pa which is close to the values

    reported in Karppinen et al. (2011).

    It can be inferred from Figs. 10,11,12and13that

    minimum yield stress values for MFC sample is

    between the solids concentrations of 11.5 kg/m3, for

    NF suspension between 2.0 and 2.5 kg/m3

    , forNIST90 suspension between 8.0 and 10 kg/m3, for

    NIST60 suspension 1.251.5 kg/m3. Assuming lowest

    possible yield stress can be measured as

    1.84 9 10-5 N m torque or a yield stress of 0.11 Pa,

    then the corresponding concentration values were

    found from the power law fitting equation.

    The gel points in Table 5calculated from the yield

    stress measurements are in excellent agreement with

    those obtained from the sedimentation data and given

    in Table3. The NF sample has lower aspect ratio

    compared to MFC although the NF sample has a loweraverage diameter than the MFC sample. There are

    likely to be two factors contributing to this. Firstly, the

    highest aspect ratio nanomaterial may have been

    removed with the supernatant after centrifugation. In

    addition, the higher aspect ratio nanofibres may have

    been retained preferentially on the fibres retained on

    the filter mat. The gel-points measured for the MFC,

    NF and NIST60 samples are also in good agreement

    with previous measurements from the literature. Gel

    0 0.5 1 1.50

    2

    4

    6

    8

    10

    12

    14

    16

    18

    Fibre weight percentage

    Yields

    tress(Pa)

    Experimental

    Kerekes et.al.(1985)

    Fig. 14 Effect of fibre weight percentage on yield stress for

    MFC suspension

    0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20

    5

    10

    15

    20

    25

    30

    35

    40

    Fibre weight percentage

    Yield

    stress

    (Pa)

    Experimental

    Kerekes et.al.(1985)

    Fig. 15 Effect of fibre weight percentage on yield stress for NF

    suspensions

    Table 4 Power-law fitting constants for MFC and NF samples

    Sample Power law fitting constants

    a b

    MFC 6.70 2.34

    NF 6.34 2.57

    Cellulose (2013) 20:18851896 1893

    1 3

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    10/12

    points estimated by sedimentation for three nanofibre

    samples derived from the same precursor material as

    the NIST90 and 60 samples, ranged from 1.76 to

    5.37 kg/m3 (Zhang et al. 2012) while the gel-point

    estimated from the change of slope of viscosity as a

    function of solids concentration of a MFC sample

    prepared by TEMPO oxidation of sulphite pine pulpwas approximately 2 kg/m3 (Lasseuguette et al. 2008).

    The data of Ishii et al. (2011) showed a boundary

    between dilute and semi-dilute suspensions of nano-

    fibre prepared by TEMPO oxidation of bleached

    hardwood kraft pulp was between 0.1 and 0.2 kg/m3.

    They estimated the boundary based on the terminal

    relaxation time of dilute suspensions. They have done

    measurements at 0.1 kg/m3 and 0.2 kg/m3 but inter-

    mediate concentrations were not tested.

    Calculated aspect ratios

    Table6 shows the aspect ratios calculated from all

    four samples using the values ofCc calculated from the

    sedimentation and yield stress measurements, using

    the EMT and CN theories at several different densities

    of the fibres in suspension, as given in Table 2.

    The Table6 shows firstly that estimated aspect

    ratio using either sedimentation or yield stress mea-

    surements still has considerable sources of uncer-

    tainty. We do not know the correct density of the fibresin suspension to use, and neither the model of a

    straight cylindrical fibre used in the CN theory or of a

    spheroid used in the EMT theory completely repre-

    sents the reality of the fibres in suspension. The basic

    Crowding Number theory is based on single averages

    only. Kropholler and Sampson (2001) reported that

    mean Crowding Number of fibre with a distribution of

    lengths is higher than that for fibres of uniform lengths

    equal to mean by a factor that is dependent only on the

    coefficient of variation of fibre length. The crowding

    number does not depend too strongly on coefficient of

    variation of the length distribution. EMT is actually

    based on inclusions with same aspect ratio and shape

    while allowing an arbitrary distribution of lengths. No

    method is available to calculate length distribution

    from average aspect ratio measurement.However it is encouraging that the data set is quite

    consistent between samples and with the different

    theories and fibre densities used. From the data the

    ranking order of aspect ratio is NIST60 having the

    largest aspect ratio followed by the MFC, the NF and

    the NIST90 samples, independent of the fibre density

    or the theory used. There is also excellent qualitative

    agreement between the SEM images of the different

    samples and the calculated aspect ratios. The exces-

    sive treatment applied to produce fibre fragmentation

    in the NIST90 samples is observable both in the SEMimage shown in Fig. 8as well as producing sedimen-

    tation curves and yield stress measurements that are

    substantially different to the other three samples. The

    aspect ratio estimated for the nanofibre prepared by

    Ishii et al. (2011) based on the method reported here

    assuming gel point concentration as 0.2 kg/m3 is 440,

    when the fibre density is assumed to be 1,500 kg/m3.

    This compares well with the aspect ratio of 550

    estimated by Ishii et al. (2011) from terminal relax-

    ation times. They reported that nanofibres are dis-

    persed without entanglement.It is also encouraging that the two independent

    methods of measuring Cc are also in excellent

    agreement, despite sedimentation relying on the

    formation of structures large enough to scatter visible

    light, while the yield stress measurement relies on the

    establishment of a continuous network of fibres. The

    results suggest that either method could be chosen

    depending on equipment availability and the rapidity

    with which the measurements are required.

    There still remains the question as to the meaning of

    the aspect ratio estimated here. As is observable in theSEM images of the fibres in Figs. 2, 4 and 6, the

    nanofibres are often not fully separated and often form

    network type structures. It is open question as to

    whether fibre lengths calculated by dividing aspect

    ratio by average fibre diameter are physically

    reasonable.

    However, as discussed in the theory, the statistics of

    fibre contacts in suspension depend only on aspect

    ratio. The aspect ratio controls both the connectivity

    Table 5 Gel points for MFC,NF, NIST90 and NIST60 sam-

    ples from yield stress measurements

    Suspension Gel point concentration

    of suspension (kg/m3

    )

    Solids mass

    fraction (kg/kg)

    MFC 1.76 1.76E-03

    NF 2.4 2.4E-03

    NIST90 10 10E-03

    NIST60 1.5 1.5E-03

    1894 Cellulose (2013) 20:18851896

    1 3

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    11/12

    (gel-point) and rigidity thresholds. We have demon-

    strated previously a method to rapidly form cellulose

    nanofibre sheets (Varanasi and Batchelor2013) using

    commercial woven filters. We have shown that almost

    100 % retention is possible if the sheets are formed at

    solids concentrations above the gel-point. The contin-

    uous network of nanofibres, whether individual orpresent as networks, prevents the nanofibres from

    being pulled through the 125 lm filter openings under

    the applied vacuum. Sheets were unable to be formed

    through filtration when the initial solids content was

    less than the gel-point (Varanasi and Batchelor2013).

    It should be noted that, when using this method, we

    were readily able to form sheets using filtration from

    the MFC, NF and NIST60 samples, which all have

    high aspect ratios, but that it proved impossible to

    form a sheet using the NIST90 sample, which had a

    very low aspect ratio.

    Conclusion

    Two simple methods were developed for measuring

    suspension gel point for cellulose nanofibres based on

    either yield stress measurements or sedimentation.

    Aspect ratios were calculated using effective medium

    theory (EMT) and Crowding number (CN) theory.

    Both theories showed reasonable agreement with each

    other. Four different types of cellulose nanofibres were

    prepared and tested. Aspect ratios were calculated for

    assumed density of fibre in suspension ranging from

    1,166 to 1,500 kg/m3. Aspect ratios of MFC, NF,

    NIST90 and NIST60 are 142, 125, 62 and 150 at

    density of 1,500 kg/m3. Suspension gel point andAspect ratio calculated from either the sedimentation

    or suspension yield stress methods are quite consistent

    between samples and with the different theories and

    fibre densities used.

    Acknowledgments The authors would like to acknowledge

    the facilities used with the Monash Center for Electron

    Microscopy. The authors would like to thank Wade Mosse

    and Mae-Gene Yew for assisting us with rheological

    measurements and Liyuan Zhang and A.Prof. Takuya Tsuzuki

    for helping us to develop the method for separating nanofibres

    from MFC sample. We acknowledge the financial support of theAustralian Research Council, Australian Paper, Nopco

    Australasia, Norske Skog, SCA Hygiene Australasia and Visy

    through Linkage Project Grants LP0989823 and LP0990526.

    Swambabu Varanasi thanks Monash University for a MGS and

    FEIPRS Scholarship.

    References

    Ahola S, Osterberg M, Laine J (2008) Cellulose nanofibrils

    adsorption with poly(amideamine) epichlorohydrin

    Table 6 Average aspect ratio values for calculated from gel point measurements from sedimentation and yield stress measurements

    and standard errors are reported as range in brackets

    Assumed

    density

    (kg/m3

    )

    Average aspect ratio

    MFC Nanofibre NIST90 NIST60

    Sedimentation

    CN 1,500 142 (28/18) 125 (8/7) 62(2/2) 150(1/1)

    1,333 134 (27/17) 118 (7/6) 58 (2/2) 142 (1/1)

    1,166 125 (25/16) 110 (7/6) 55 (2/1) 133 (1/1)

    EMT 1,500 125 (29/18) 108 (8/7) 48 (2/1) 134 (1/1)

    1,333 117 (27/17) 101 (7/6) 45 (1/1) 125 (1/1)

    1,166 108 (25/15) 93 (7/6) 41 (1/1) 116 (1/1)

    Yield stress

    CN 1,500 143 121 60 155

    1,333 135 115 57 146

    1,166 126 107 53 137

    EMT 1,500 126 105 46 1391,333 118 98 43 129

    1,166 109 90 40 120

    Cellulose (2013) 20:18851896 1895

    1 3

  • 8/10/2019 10.1007-s10570-013-9972-9-Printed copy

    12/12

    studied by QCM-D and application as a paper strength

    additive. Cellulose 15(2):303314. doi:10.1007/s10570-

    007-9167-3

    Celzard A, Mareche JF, Payot F (2000) Simple method for

    characterizing synthetic graphite powders. J Phys D Appl

    Phys 33(12):1556

    Celzard A, Fierro V, Pizzi A (2008) Flocculation of cellulose

    fibre suspensions: the contribution of percolation and

    effective-medium theories. Cellulose 15(6):803814. doi:

    10.1007/s10570-008-9229-1

    Celzard A, Fierro V, Kerekes R (2009) Flocculation of cellulose

    fibres: new comparison of crowding factor with percolation

    and effective-medium theories. Cellulose 16(6):983987.

    doi:10.1007/s10570-009-9314-0

    Czaja WK, Young DJ, Kawecki M, Brown RM (2006) The

    future prospects of microbial cellulose in biomedical

    applications. Biomacromolecules 8(1):112. doi:10.1021/

    bm060620d

    Dalpke B, Kerekes RJ (2005) The influence of fibre properties

    on the apparent yield stress of flocculated pulp suspensions.

    J Pulp Pap Sci 31(1):3943

    Derakhshandeh B, Kerekes RJ, Hatzikiriakos SG, BenningtonCPJ (2011) Rheology of pulp fibre suspensions: a critical

    review. Chem Eng Sci 66(15):34603470. doi:10.1016/

    j.ces.2011.04.017

    Dufresne A, Dupeyre D, Vignon MR (2000) Cellulose micro-

    fibrils from potato tuber cells: processing and character-

    ization of starch-cellulose microfibril composites. J Appl

    Polym Sci 76(14):20802092

    Dzuy NQ, Boger DV (1985) Direct yield stress measurement

    with the vane method. J Rheol 29(3):335347

    Eichhorn S, Dufresne A, Aranguren M, Marcovich N, Capadona

    J, Rowan S, Weder C, Thielemans W, Roman M, Ren-

    neckar S, Gindl W, Veigel S, Keckes J, Yano H, Abe K,

    Nogi M, Nakagaito A, Mangalam A, Simonsen J, Benight

    A, Bismarck A, Berglund L, Peijs T (2010)Review: currentinternational research into cellulose nanofibres and nano-

    composites. J Mater Sci 45(1):133. doi:10.1007/s10853-

    009-3874-0

    Henriksson M, Berglund LA, Isaksson P, Lindstrom T, Nishino

    T (2008) Cellulose nanopaper structures of high toughness.

    Biomacromolecules 9(6):15791585. doi:10.1021/bm800

    038n

    Ishii D, Saito T, Isogai A (2011) Viscoelastic evaluation of

    average length of cellulose nanofibers prepared by

    TEMPO-mediated oxidation. Biomacromolecules 12(3):

    548550. doi:10.1021/bm1013876

    Jakob HF, Fengel D, Tschegg SE, Fratzl P (1995) The ele-

    mentary cellulose fibril in picea abies: comparison of

    transmission electron microscopy, small-angle X-ray

    scattering, and wide-angle X-ray scattering results. Mac-

    romolecules 28(26):87828787. doi:10.1021/ma00130

    a010

    Janarthanan S, Sain M (2006) Isolation of cellulose microfibrils

    an enzymathic approach. Bioresources 1(2):176188

    Karppinen A, Vesterinen A-H, Saarinen T, Pietikainen P, Sep-

    pala J (2011) Effect of cationic polymethacrylates on the

    rheology and flocculation of microfibrillated cellulose.

    Cellulose 18(6):13811390. doi:10.1007/s10570-011-

    9597-9

    Kerekes RJ, Schell CJ (1992) Characterization of fibre floccu-

    lation regimes by a crowding factor. J Pulp Pap Sci 18(1):

    J32J38

    Kerekes RJ, Soszynski RM, Tam Doo PA (1985) The floccu-

    lation of pulp fibres.In: Proceedings of the 8th fundamental

    research symposium, Oxford, England. Mechanical Engi-

    neering Publications, pp 265310

    Kropholler HW, Sampson WW (2001) The effect of fibre length

    distribution on suspension crowding. J Pulp Pap Sci

    27(09):301305

    Lasseuguette E, Roux D, Nishiyama Y (2008) Rheological

    properties of microfibrillar suspension of TEMPO-oxi-

    dized pulp. Cellulose 15(3):425433

    Liddel PV, Boger DV (1996) Yield stress measurements with

    the vane. J Nonnewton Fluid Mech 63(23):235261. doi:

    10.1016/0377-0257(95)01421-7

    Malcolm Brown R Jr, Saxena IM, Kudlicka K (1996) Cellulose

    biosynthesis in higher plants. Trends Plant Sci

    1(5):149156. doi:10.1016/S1360-1385(96)80050-1

    Martinez DM, Buckley K, Jivan S, Lindstrom A, Thiruvenga-

    daswamy R, Olson JA, Ruth TJ, Kerekes RJ (2001)

    Characterizing the mobility of papermaking fibres duringsedimentation. In: Baker CF (ed) The science of paper-

    making, transactions of the 12th fundamental research

    symposium, Oxford. The Pulp and Paper Fundamental

    Research Society, Bury, Lancaster, UK, pp 225254

    Mosse WKJ, Boger DV, Garnier G (2012a) Avoiding slip in

    pulp suspension rheometry. J Rheol 56(6):15171533

    Mosse WKJ, Boger DV, Simon GP, Garnier G (2012b) Effect of

    cationic polyacrylamides on the interactions between cel-

    lulose fibers. Langmuir 28(7):36413649. doi:10.1021/

    la2049579

    Nogi M, Iwamoto S, Nakagaito AN, Yano H (2009) Optically

    transparent nanofiber paper. Adv Mater 21(16):

    15951598. doi:10.1002/adma.200803174

    Perkins EL, Batchelor WJ (2012) Water interaction in papercellulose fibres as investigated by NMR pulsed field gradi-

    ent. Carbohydr Polym 87:361367. doi:10.1016/j.carbpol.

    2011.07.065

    Siro I, Plackett D (2010) Microfibrillated cellulose and new

    nanocomposite materials: a review. Cellulose 17(3):

    459494. doi:10.1007/s10570-010-9405-y

    Taniguchi T, Okamura K (1998) New films produced from mi-

    crofibrillated natural fibres. Polym Int 47(3):291294. doi:

    10.1002/(sici)1097-0126(199811)47:3\291:aid-pi11[3.0.

    co;2-1

    Varanasi S, Batchelor W (2013) Rapid preparation of cellulose

    nanofibre sheet. Cellulose 20(1):211215. doi:10.1007/

    s10570-012-9794-1

    Varanasi S, Chiam HH, Batchelor WJ (2012) Application and

    interpretation of zero and short-span testing on nanofibre

    sheet materials. Nord Pulp Pap Res J 27(2):343352

    Xu J, Chatterjee S, Koelling KW, Wang Y, Bechtel SE (2005)

    Shear and extensional rheology of carbon nanofiber sus-

    pensions. Rheol Acta 44(6):537562. doi:10.1007/

    s00397-005-0436-5

    Zhang L, Batchelor W, Varanasi S, Tsuzuki T, Wang X (2012)

    Effect of cellulose nanofiber dimensions on sheet forming

    through filtration. Cellulose 19(2):561574. doi:10.1007/

    s10570-011-9641-9

    1896 Cellulose (2013) 20:18851896

    1 3

    http://dx.doi.org/10.1007/s10570-007-9167-3http://dx.doi.org/10.1007/s10570-007-9167-3http://dx.doi.org/10.1007/s10570-008-9229-1http://dx.doi.org/10.1007/s10570-009-9314-0http://dx.doi.org/10.1021/bm060620dhttp://dx.doi.org/10.1021/bm060620dhttp://dx.doi.org/10.1016/j.ces.2011.04.017http://dx.doi.org/10.1016/j.ces.2011.04.017http://dx.doi.org/10.1007/s10853-009-3874-0http://dx.doi.org/10.1007/s10853-009-3874-0http://dx.doi.org/10.1021/bm800038nhttp://dx.doi.org/10.1021/bm800038nhttp://dx.doi.org/10.1021/bm1013876http://dx.doi.org/10.1021/ma00130a010http://dx.doi.org/10.1021/ma00130a010http://dx.doi.org/10.1007/s10570-011-9597-9http://dx.doi.org/10.1007/s10570-011-9597-9http://dx.doi.org/10.1016/0377-0257(95)01421-7http://dx.doi.org/10.1016/S1360-1385(96)80050-1http://dx.doi.org/10.1021/la2049579http://dx.doi.org/10.1021/la2049579http://dx.doi.org/10.1002/adma.200803174http://dx.doi.org/10.1016/j.carbpol.2011.07.065http://dx.doi.org/10.1016/j.carbpol.2011.07.065http://dx.doi.org/10.1007/s10570-010-9405-yhttp://dx.doi.org/10.1002/(sici)1097-0126(199811)47:3%3c291:aid-pi11%3e3.0.co;2-1http://dx.doi.org/10.1002/(sici)1097-0126(199811)47:3%3c291:aid-pi11%3e3.0.co;2-1http://dx.doi.org/10.1002/(sici)1097-0126(199811)47:3%3c291:aid-pi11%3e3.0.co;2-1http://dx.doi.org/10.1002/(sici)1097-0126(199811)47:3%3c291:aid-pi11%3e3.0.co;2-1http://dx.doi.org/10.1002/(sici)1097-0126(199811)47:3%3c291:aid-pi11%3e3.0.co;2-1http://dx.doi.org/10.1002/(sici)1097-0126(199811)47:3%3c291:aid-pi11%3e3.0.co;2-1http://dx.doi.org/10.1007/s10570-012-9794-1http://dx.doi.org/10.1007/s10570-012-9794-1http://dx.doi.org/10.1007/s00397-005-0436-5http://dx.doi.org/10.1007/s00397-005-0436-5http://dx.doi.org/10.1007/s10570-011-9641-9http://dx.doi.org/10.1007/s10570-011-9641-9http://dx.doi.org/10.1007/s10570-011-9641-9http://dx.doi.org/10.1007/s10570-011-9641-9http://dx.doi.org/10.1007/s00397-005-0436-5http://dx.doi.org/10.1007/s00397-005-0436-5http://dx.doi.org/10.1007/s10570-012-9794-1http://dx.doi.org/10.1007/s10570-012-9794-1http://dx.doi.org/10.1002/(sici)1097-0126(199811)47:3%3c291:aid-pi11%3e3.0.co;2-1http://dx.doi.org/10.1002/(sici)1097-0126(199811)47:3%3c291:aid-pi11%3e3.0.co;2-1http://dx.doi.org/10.1007/s10570-010-9405-yhttp://dx.doi.org/10.1016/j.carbpol.2011.07.065http://dx.doi.org/10.1016/j.carbpol.2011.07.065http://dx.doi.org/10.1002/adma.200803174http://dx.doi.org/10.1021/la2049579http://dx.doi.org/10.1021/la2049579http://dx.doi.org/10.1016/S1360-1385(96)80050-1http://dx.doi.org/10.1016/0377-0257(95)01421-7http://dx.doi.org/10.1007/s10570-011-9597-9http://dx.doi.org/10.1007/s10570-011-9597-9http://dx.doi.org/10.1021/ma00130a010http://dx.doi.org/10.1021/ma00130a010http://dx.doi.org/10.1021/bm1013876http://dx.doi.org/10.1021/bm800038nhttp://dx.doi.org/10.1021/bm800038nhttp://dx.doi.org/10.1007/s10853-009-3874-0http://dx.doi.org/10.1007/s10853-009-3874-0http://dx.doi.org/10.1016/j.ces.2011.04.017http://dx.doi.org/10.1016/j.ces.2011.04.017http://dx.doi.org/10.1021/bm060620dhttp://dx.doi.org/10.1021/bm060620dhttp://dx.doi.org/10.1007/s10570-009-9314-0http://dx.doi.org/10.1007/s10570-008-9229-1http://dx.doi.org/10.1007/s10570-007-9167-3http://dx.doi.org/10.1007/s10570-007-9167-3