In situ rheometry of concentrated cellulose fibre suspensions and relationships with enzymatic hydrolysis

This is an author-deposited version published in :

http://oatao.univ-toulouse.fr/

Eprints ID : 9235

To link to this article :DOI:1016/j.biortech.2013.01.110

URL : http://dx.doi.org/2010.1016/j.biortech.2013.01.110

Open

Archive

TOULOUSE

Archive

Ouverte (OATAO)

OATAO is an open access repository that collects the work of Toulouse researchers and

makes it freely available over the web where possible.

To cite this version : Nguyen, Tien-Cuong and Anne-Archard,

Dominique and Coma, Véronique and Cameleyre, Xavier and

Lombard, Eric and Binet, Cédric and Nouhen, Arthur and To, Kim

Anh and Fillaudeau, Luc In situ rheometry of concentrated

cellulose fibre suspensions and relationships with enzymatic

hydrolysis. (2013) Bioresource Technology, vol. 133 . pp.

563-572. ISSN 0960-8524

In situ rheometry of concentrated cellulose fibre suspensions

and relationships with enzymatic hydrolysis

Tien-Cuong Nguyen

a,⇑

_{, Dominique Anne-Archard}

b

_{, Véronique Coma}

c

_{, Xavier Cameleyre}

a

_,

Eric Lombard

a

_{, Cédric Binet}

b

_{, Arthur Nouhen}

b

_{, Kim Anh To}

d

_{, Luc Fillaudeau}

a

a_{Laboratoire d’Ingénierie des Systèmes Biologiques et des Procédés (Université de Toulouse, INSA, INRA UMR792, CNRS UMR5504), Toulouse, France} b_{Université de Toulouse, INPT, UPS, IMFT (Institut de Mécanique des Fluides de Toulouse), Toulouse, France}

c_{Laboratoire de Chimie des Polymères Organiques UMR 5629 CNRS/Université Bordeaux 1, IPB/ENSCPB, Pessac, France} d_{School of Biotechnology and Food Technology, Hanoi University of Sciences and Technology, Viet Nam}

"We explore the suspending and enzymatic hydrolysis of microcrystalline cellulose, Whatman paper and extruded paper-pulp. "A methodology to determine on-line viscosity is proposed and validated.

"A structured rheological model is established.

"Suspension viscosity and particle size decreased rapidly during the enzymatic hydrolysis.

Keywords: Lignocellulose Rheology Paper pulp Hydrolyse Viscosity

a b s t r a c t

This work combines physical and biochemical analyses to scrutinize liquefaction and saccharification of complex lignocellulose materials. A multilevel analysis (macroscopic: rheology, microscopic: particle size and morphology and molecular: sugar product) was conducted at the lab-scale with three matrices: microcrystalline cellulose (MCC), Whatman paper (WP) and extruded paper-pulp (PP). A methodology to determine on-line viscosity is proposed and validated using the concept of Metzner and Otto (1957) and Rieger and Novak’s (1973). The substrate suspensions exhibited a shear-thinning behaviour with respect to the power law. A structured rheological model was established to account for the suspension viscosity as a function of shear rate and substrate concentration. The critical volume fractions indicate the transition between diluted, semi-diluted and concentrated regimes. The enzymatic hydrolysis was per-formed with various solid contents: MCC 273.6 gdm/L, WP 56.0 gdm/L, PP 35.1 gdm/L. During hydrolysis, the suspension viscosity decreased rapidly. The fibre diameter decreased two fold within 2 h of starting hydrolysis whereas limited bioconversion was obtained (10–15%).

1. Introduction

Lignocellulose biomass is one of the most abundant renewable resources and certainly one of the least expensive. Its conversion into ethanol fuel is eventually expected to provide a significant portion of the world’s energy requirements. The substrates used

are varied. They include woody substrates (hardwood and soft-wood), products from agriculture (straw) or those of lignocellulosic waste industries (food processing, paper).

In order to achieve economic viability, the biorefining of ligno-cellulosic resources must be operated at very high feedstock dry matter content. Paper pulp is quite appropriate for modern bio-refining, because it displays a low lignin content, it is free of inhib-itory compounds that can perturb fermentations and devoid of microbial contaminants.

Nevertheless, the enzyme liquefaction and saccharification of paper-like pulps are subject to the same constraints as other pulps obtained via alternative methods such as steam explosion or dilute acid hydrolysis. Therefore, a better scientific understanding and, ultimately, good technical control of these critical biocatalytic

Abbreviations: N, Mixing rate (rpm); d, Impeller diameter (m); C, Torque (N.m); P, Power (W);q, Density (kg/m3_{); Np, Power number; Re, Reynolds number; Re}

g,

Generalized Reynolds number; Re⁄_{, Rieger& Novak Reynolds;l, Viscosity (Pa.s); [l],}

Intrinsic viscosity; Kp, Geometrical constant; Ks, Metzner-Otto constant; _c, Shear rate (sÿ1_{); n, Power-law index; k, Consistency index (Pa.n}n_{);U, Volume fraction;}

D[4,3], Mean diameter (lm); Cm, Mass concentration (g/L); dm, Dry matter (g). ⇑ Corresponding author. Tel.: +33 661970369.

reactions, which involve complex matrices at high solid contents, is currently a major challenge if biorefining operations are to become commonplace.

Amongst the main parameters to be studied, the rheological behaviour of the hydrolysis suspension and the fibre particle size of, stand out as a major determinants of process efficiency and determine equipment to be used and the strategies applied ( Wi-man et al., 2010). The choice of agitation system, fundamental to heat and/or mass transfer, and to disruption of agglomerated par-ticles, influences the bioconversion of cellulose into simple sugar (Um, 2007). It requires detailed knowledge of the rheological behaviour of the substrate suspensions. However, these suspen-sions present such complex and unique properties that there are no standard method for studying fibre network deformation and pulp flow behaviour (Blanco et al., 2006).

Fibre suspension flow is a key factor and extensive studies have been reported in the pulp and paper scientific literature. Cellulose fibres in suspension form three-dimensional networks that exhibit viscoelastic properties (Wahren et al., 1964; Kerekes et al., 1985 ci-ted byAntunes, 2009). Measuring the rheological properties of fi-bre suspensions is complex, owing to multiple factors: (i) fifi-bre physical and mechanical properties and concentration ranges, (ii) fibre contacts and surface forces and (iii) forces on fibres and flocculation. Rheological behaviour of fibre suspensions is usually described by an apparent yield stress, a shear viscosity (Hershel– Buckley or Bingham models) and elasticity. The physical properties of cellulose fibre are considered such as swelling, dissolution, structure and strength of network. The strength of the network of the coarsest fibres determines the rheology of these materials (Wiman et al., 2010). The rheology of lignocellulose suspensions is of special interest and studies are numerous at different temperatures and concentrations, from dilute solutions 0.2–3.0% (Agoda-Tandjawa et al., 2010; Ferreira et al., 2003) to concentrated solutions 10–20% (Um and Hanley, 2008; Zhang et al., 2009). Both of these studies conclude that a shear-thinning behaviour occurs for any lignocellulosic substrate suspension: microcrystalline cel-lulose (Agoda-Tandjawa et al., 2010; Chaussy et al., 2011; Tatsumi and Matsumoto, 2007; Um and Hanley, 2008); hardwood paper-pulp (Blanco et al., 2006; Zhang et al., 2009); softwood paper-pulp (Ferreira et al., 2003; Wiman et al., 2010); sugar cane bagasse (Pereira et al., 2011). The viscosity of the suspension depends not only on the temperature and concentration (Ferreira et al., 2003) but also on the average fibre length (Lapierre et al., 2006). A longer fibre has a higher degree of polymerisation and generates a higher viscosity. During biological hydrolysis, the apparent viscosity of suspensions decreases (Pereira et al., 2011; Um, 2007) in parallel with a decrease of particle size (Wiman et al., 2010).

Traditionally, rotating viscometers have been used (Duffy and Titchener, 1975; Chase et al., 1989; Bennington et al., 1990). How-ever, normal commercial viscometers do not provide enough mix-ing to maintain uniform fibre distribution, which causes viscosity values close to the viscosity of the pure water (Blanco et al., 1995 cited byAntunes, 2009). Therefore, to study the rheological properties of fibre suspensions there is no standardized method but several measuring devices have been reported in the literature (Cui and Grace, 2007; Blanco et al., 2006; Chaussy et al., 2011; Der-akhshandeh et al., 2011). Plate torque-based devices have the high-est resolution and can be used to determine the rheological behaviour of pulp suspensions (Blanco et al., 2006). One difficulty remains in the definition of criteria to ascribe a viscosity to a het-erogeneous suspension, originally defined for homogeneous fluids in laminar flow (Blanco et al., 2006). To attain fluidisation, appar-ent yield stress must be exceeded throughout the suspension. Although fluidisation generally occurs in a turbulent regime, fluid-like behaviour at the floc level can be attained under non-tur-bulent conditions. One example is the flow induced in a rotary

device at slow rotational speeds just above the apparent yield stress; another example was found in spouted beds ( Dera-khshandeh et al., 2011).

Then on-line measurement of torque or mixing power in biore-actors may highlight viscosity of concentrated cellulose suspen-sions and may constitute a way to follow enzymatic hydrolysis reactions. Particle size, rheology, and rate of enzymatic hydrolysis could be correlated to operating conditions for example: mixing rate and impeller speed (Pereira et al., 2011; Samaniuk et al., 2011).

The aim of the present report was to investigate the dynamics of transfer phenomena and limitation of biocatalytic reactions with lignocelluloses resources under high concentration conditions. This study focuses on the characterisation of cellulose suspensions at different concentrations and coupling with the enzymatic kinetics of hydrolysis using on-line viscosimetry. In the literature, rheome-ters are used to determine ex situ suspension viscosity. These ap-proaches are limited by the number of samples and the substrate properties, predominately decantation and flocculation of material. To solve these problems, a method allowing the suspension viscos-ity to be followed is proposed. Firstly, cellulose fibre suspensions at various concentrations are investigated through on-line measure-ments in purpose-built bioreactor. Three real and model matrices are characterised by fiber morphology, diameter and concentra-tion. Using Metzner and Otto concept (1957), rheograms were determined. Rheological behaviour was then described by struc-tured rheological models. Secondly, the complex relationships be-tween fibre structure, degradation, chemical composition and rheological behaviour was scrutinised. To do so, physical and bio-chemical on-line and off-line analyses were conducted during the bioreaction. A relationship between viscosity change and biocata-lytic degradation of fibre was observed.

2. Methods

2.1. Experimental device

The experimental set-up consists of a tank and an impeller sys-tem connected to a viscometer working at imposed speed (Visco-tester HaakeVT550, Thermo Fisher Scientific, Ref: 002-7026) (Fig. 1). This allows on-line torque measurements. The rotational speed ranged between 0.5 and 800 rpm and torque between 1 and 30 mN m. The bioreactor was a homemade glass tank with a flat bottom (diameter: 82 mm, Hmax: 76 mm, V: 0.4 L) fitted with a water jacket. The impeller was a four-pitched blade turbine (IKA A200, stainless steel, d: 50 mm, l: 21 mm, w: 8 mm, 45° angle 25 mm from the bottom of the tank to maintain axial and radial flows. Temperature was controlled by circulation (cryostat Haake DC30 and K20) through the water jacket. A bioreactor panel control (B. Braun Biotech International MCU200 + microDCU300) was used for pH control and regulation, dissolved oxygen and temperature measurements. The viscometer and the cryostat ware controlled by software from HaakeRheoWin Job Manager (Thermo Fisher Sci-entific) which also ensured data recording (temperature, torque and mixing rate).

2.2. Substrates and enzymes

Three cellulose matrices were studied in order to investigate different fibre morphologies and particle size distributions ( Ta-ble 1): microcrystalline cellulose (ACROS Organics, Ref: 382310010), a dried and milled (Bosch MKM6003 mill) Whatman paper (Whatman International Ltd., Maidstone, England, Cat No. 1001 090) and paper-pulp (Tembec Co., Saint-Gaudens, France, type FPP31) after extrusion (7/8 mixing, 1/8 shear stress, Prism

TSE24MC, 400 mm failure, Thermo Electron Corp.). The Tembec pa-per-pulp was made from coniferous wood and contained 26.1% dry matter (75.1% cellulose, 19.1% hemicellulose, 2.2% Klason lignin and ash). The three substrates are henceforth referred to as MCC, for microcrystalline cellulose, WP for Whatman paper and PP for extruded paper pulp. The density of the three substrates was deter-mined by the volume method (proportion of substrate volume and added water volume in a volumetric flask of 100 mL). This density corresponds to the suspended matrix, including its initial water content. It was used to calculate the volume fraction, even though other definitions can be proposed it characterizes raw matter and emanates directly from the industrial process.

An enzyme cocktail (Enzyme ACCELLERASEÒ

Genecor, Ref. 3015155108) containing exoglucanases, endoglucanases (2800 CMC U/g, i.e. 57 ± 2.8 FPU/mL cited byAlvira et al., 2011), hemicellulases and b-glucosidases (775 pNPG U/g) was used. Its optimal temperature and pH were 50 °C (range 50–65 °C) and pH 4.8 (range 4–5). An ACCELLERASEÒ

1500 dosage rate of 0.1– 0.5 mL per gram of cellulose or roughly 0.05–0.25 mL per gram of biomass (depending on biomass composition) is recommended by the manufacturers.

2.3. Physical and chemical analysis 2.3.1. Laser particle size determination

Particle size distribution was determined through laser diffrac-tion analyses (Mastersizer 2000 Hydro, Malvern Instruments Ltd., SN: 34205-69, range from 0.02 to 2000

lm). A suspension

(approx-imately 5 g/L) was added drop by drop to the circulation loop (150 mL). Analysis are conducted at room temperature (20 °C) with obscuration rates (red k = 632.8 nm and blue k = 470.0 nm lights) ranging between 10% and 40%. Particle volume distribution and the associated cumulative curve versus particle diameter were determined. Laser diffraction analysis converts the detected scat-tered light into a particle size distribution. Successful deconvolu-tion relies on an appropriate descripdeconvolu-tion of light behaviour: either Mie theory or the Fraunhofer approximation (of Mie theory). Historically, the use of Mie theory was limited by computing power, which was eliminated in the last decade by dramatic in-creases in processing power. This method was designed for parti-cles, so relative measurements were made in order take complex particle shape, refractive index and measurement repeatability into consideration.

2.3.2. Morpho-granulometry

Fibre morphology was observed using a mopho-granulometer (Mastersizer G3S, Malvern Instruments Ltd., SN: MAL1033756, software Morphologi v7.21). This optical device includes a lens (magnification: from 1 to 50, dimension min/max: 0.5/

3000

lm) and a camera (Nikon CFI60). Samples were analysed by

two methods: ‘‘dry’’ and ‘‘wet’’. For ‘‘dry’’ analysis, the powders were dispersed using a specific dispersion unit (with air). For ‘‘wet’’ analysis, the suspensions (approximately 5 g/L) were ob-served between cover glasses and slides. A 1.5 mm  1.5mm sur-face was observed under standardized conditions (light intensity:

Bioreactor Cryostat Tp pH DC30 Torque Mixing Sampling pH and antibiotic adds

Fig. 1. Experimental set-up.

Table 1

Substrate properties (MCC: microcrystalline cellulose, WP: Whatman paper and PP: extruded paper pulp).

MCC WP PP Dry matter (%) 99 99 26 Cellulose (%) 100 90 75 D[4, 3] (lm) 70 250 190 q(g/L) 1623 ± 28 1200 ± 2 1346 ± 2 Crystallinity (%) 79.0 88.6 64.5

80 ± 0.2; magnification: 2.5). The images were filtered and ana-lysed to determine the number of particles and their geometric properties (diameter, aspect ratio, etc.).

2.3.3. Glucose concentration (YSI)

Glucose concentration was checked in the supernatant along enzymatic hydrolysis (Analyser YSI model 27A; Yellow Springs Instruments, Yellow Springs, Ohio, range 0–2.5 g/L ± 2%, sample volume = 25

lL).

2.4. Generalised power consumption curve and on-line viscosimetry Power consumption is described by the dimensionless power number Npversus the mixing Reynold number, Re and was estab-lished for Newtonian fluids, with:

Np¼ P d5

q

N3; Re ¼

q

N
d2

l

ð1Þ P ¼ 2p
N
C:

This single master curve depends only on impeller/reactor shape and geometry. In the laminar regime (Re < 10–100), the product Np
Re is a constant, named Kp, which is then defined as follows:

Np
Re ¼ Kp ð2Þ

Kpis a function of impeller shape and geometry for any Newto-nian fluid. A deviation from Eq. (2) indicates the end of laminar re-gime. In fully turbulent flow (Re > 104_–105_{) and for Newtonian} fluids, the dimensionless power number Npis assumed to be inde-pendent of mixing Reynolds number and equal to a constant, Np0.

In this case, three Newtonian fluids (distilled water, Marcol 52 oil and glycerol) were used to cover a large range of mixing Reynolds numbers. Viscosity for these calibration fluids (and also non-New-tonian fluids below) was measured with a cone and plate system (60 mm diameter, angle 2°, Mars III rheometer, Thermo Scientific) and for shear rate varying from 10ÿ2_{to 10}3_(sÿ1_{) at two different} temperatures: 20 and 40 °C. The density of the fluids was also determined by a densimeter (Mettler Toledo DE40, 0–3 g/cm3_, ±0.0001 g/cm3_{). The torque and mixing rate (ascent/descent cycles,} 0.5/800/0.5 rpm) were measured for each fluid at 20 and 40 °C. Cal-culating B and Re, the power consumption curve was then established.

The Kpvalue obtained was 68.8 which is comparable to values from the literature (Rushton et al., 1950: for propeller Kp: 40–50, for flat-blade turbine Kp: 66–76). Experimental results confirm that the laminar regime prevailed up to Re  30 (Fig. 2).

A semi-empirical model including laminar and transition re-gions were considered for the reference curve with a one-to-one relationship between Npand Re:

Np¼ Kp ReAg  n þ

a

Reb Ag  n  1=n ð3Þ The parameters n,

a

and b stand for the transition regime and adjustments to the experimental results lead to: n = 2;

a

= 3.22; b= ÿ0.208.

In the non-Newtonian case, a generalised mixing Reynolds number has to be defined as the viscosity is not a constant. The well-known Metzner and Otto concept (1957) was used: a viscos-ity

l

is defined as the Newtonian viscosity leading to the same power number.Metzner and Otto (1957)showed that the equiva-lent shear rate _

c

eqassociated to this viscosity (through the rheolog-ical behaviour of the fluid) is proportional to the rotation frequency, then introducing the Metzner–Otto parameter Ks:

_

c

eq¼ Ks
N ð4Þ

This leads to the generalized Reynolds number: Reg¼

q

N2ÿn
d2 k
Knÿ1

s

ð5Þ Ksis a constant depending only on the geometry of the stirring system. Eq. (5) can be extended to the transition region using a power equation (Jahangiri et al., 2001). Xanthan solutions (0.04%; 0.1%; 0.4%) in glucose solution (650 g/L) and in sucrose solution (943 g/L) were used to determine the proportionality constant Ks. Using the power consumption curve established with Newtonian fluids, the apparent viscosity

l

was calculated from torque and mixing rate measurements. The corresponding value of the shear rate, _

c

eq, was extracted from the rheograms of the Xanthan solu-tions. Rieger and Novak’s approach (1973) was used to determine the value of Ks: Eq. (1) with the generalized Reynolds number Regis written in a similar form:

Np
Re¼ KpðnÞ ð6Þ

With Re

¼q
N2ÿnk
d2and Kp(n) = Kp
Ksnÿ1.

The value of Ks is directly deduced from the curve Kp (-n) = f(n ÿ 1)using the previously determined Kpvalue. This leads to Ks 28 ± 4. In the case studied, the extension to the transition region using a power equation (Jahangiri et al., 2001) is not rele-vant. Once the experimental set-up was characterized by its power consumption curve Np(Re) and the Ksvalue, on-line viscosimetry of the suspension was performed before and along the biocatalytic reaction.

2.5. Methodology 2.5.1. Mixing substrate

The first step consisted in suspending the substrates in 300 ml of water. Each cycle of suspension is composed of (i) a homogeni-zation phase (500 rpm for 300 s) with substrate addition and (ii) torque measurement based on 100 s phase with increasing and decreasing mixing rates (10, 50, 100, 155, 200, 300, 500, 650 and 800 rpm) within viscosimeter capacity (Nmax= 800 rpm, Cmax - 30 mN m). The concentration chosen for a given experiment was reached by successive additions of substrate: 8  20 g for MCC, 7  3 g for WP and 11  3 g for PP.

2.5.2. Enzymatic hydrolysis

Enzymatic hydrolysis was carried out at 40 °C due to energy saving and the microbiological step during the fermentation pro-cess considering a simultaneous saccharification and fermentation (SSF) operation. The pH of the medium was adjusted to 4.8 using a solution of 85% orthophosphoric acid. To avoid contamination, 5

lL

of a solution of chloramphenicol (5 g/L) was added. Then enzymes were added when the suspension reached homogeneity and the torque values were stable.

Hydrolysis was investigated over 25 h at a mixing rate of 450 rpm and using the selected concentrations: 273.8 gdm/L for MCC; 56.0 gdm/L for WP and 35.1 gdm/L for PP. These concentra-tions were established to obtain a significant initial torque (C P 1.7 mN m) and to ensure accurate monitoring of its derivation during hydrolysis. These concentrations ensure initial laminar re-gimes for WP and PP and transitional regime for MCC (Table 2). The quantities of enzyme used were in agreement with supplier’s recommendations.

Decantation affects the suspension homogeneity and can lead to deposition under low mixing rates. This problem is exacerbated with MCC due to its higher density and higher compactness. So during the reaction, periods of higher mixing rates (500/650 rpm

for 300–600 s, every 1–3 h) were imposed in order to keep the sus-pensions uniform.

Samples were taken manually by a 6 mm diameter flexible con-nected to a 20 mL syringe. Each sample was about 6 mL, sufficient to perform analyses on 5/7 sub-samples. The total volume of sam-ples removed ranged from 30 to 42 mL (10–14% of initial volume). This order of decrease of suspension volume causes negligible im-pact on the suspension viscosity (at the end, a difference of 1–7% may be observed). The samples were used for rheological, granulo-metric and biochemical analysis during enzyme degradation.

3. Results and discussions

3.1. Viscosimetry of substrate suspensions

The rheological behaviour of suspensions is complex and is af-fected by multiple parameters such as concentration, shape, den-sity and surface properties. The viscoden-sity of the suspension was quantified as a function of the type of substrate, its concentration and the mixing conditions. Using the power consumption curve and the associated Churchill model, the on-line viscosity was esti-mated at 40 °C as a function of substrate concentration and mixing rate (Fig. 3). These raw data covered laminar and transition regimes.

For a given mixing rate and substrate concentration, the viscos-ity of the WP suspension was higher than that of the PP suspen-sion, and the viscosity of MCC was the lowest. As an example, for 155 rpm and a substrate concentration close to 64 g/L, the viscos-ities observed were

l

WP= 4560 mPa s,

l

PP= 100 mPa s, and

l

MCC= 2 mPa s with a decreasing volume fractions, UWP= 0.055 (64.8 gdm/L), UPP= 0.047 (16.5 gdm/L) and UMCC= 0.039 (64.0 gdm/L) respectively. For identical mixing rates and a sub-strate concentration close to 16 gdm/L, interpolation of the previ-ous results gives an estimate of

l

WP= 194 mPa s,

l

PP= 90 mPa s, and

l

MCC= 8 mPa s with decreasing volume fractions of UPP= 0.047 (64 g/L), UWP= 0.016 (19.7 g/L) and UMCC= 0.01 (16.5 g/L). For MCC, the results are in agreement with reported data with average fibre length and diameter equal to 1.7 and 0.077

lm,

respectively exhibiting 0.01 <

l

< 10 Pa s for 0.5 < %dm < 5% (Tatsumi et al., 1999). For all the studied concentra-tion of the three suspensions, the viscosity decreased as the mixing rate increased. All the suspensions were found to act as shear-thin-ning fluids.

The on-line measurements were firstly used to establish rheo-grams (considering only results in laminar regime) and to deter-mine the rheological behaviour of the suspensions. In a second step the impact of particle volume fraction on relative viscosity was investigated. This approach contributed to establish a struc-tured rheological model including several factors such as shear-rate, volume fraction and particle dimension.

3.1.1. Rheogram

Based on the Metzner and Otto concept, rheograms are identi-fied under the laminar flow regime (Re 6 30). Data obtained with the microcrystalline cellulose suspension were outside the laminar regime, so rheograms were only obtained for WP and PP.

As the suspensions exhibited a shear-thinning behaviour, sev-eral approximations, such as power-law, Sisko, Cross, Powell–Eyr-ing, Carreau and ‘‘local’’ power-law models can be used. In the investigated conditions, a power-law model was retained. It is written:

l

¼ k
_

c

nÿ1 _ð7Þ

For substrates and WP and PP, the rheological behaviour was de-scribed as a function of concentration and modelled by linear and exponential relationships (Table 3). The patterns observed are sim-ilar to those reported byBayod et al. (2005)andLuukkonen et al. (2001). In the concentration range studied, power-law indexes

0.1 1 10 100 1000 0.1 1 10 100 1000 10000 100000 Np (/) Re (/) Water 20°C Water 40°C Glycerol 20°C Glycerol 40°C Marcol oil 20°C Marcol oil 40°C Laminar Transition

Power consumption curve

Fig. 2. Power consumption curve.

Table 2

Experimental conditions of enzymatic hydrolysis (MCC: microcrystalline cellulose, WP: Whatman paper and PP: extruded paper pulp).

MCC WP PP

Substrate concentration (gdm/L) 273.8 56.0 35.1 Cellulose content (%) 100 90 75.1 Initial flow regime, Re 218 20 32 Initial viscosity (Pa s) 0.104 0.976 0.656

ranged between 0.28 and 0.50 for WP and between 0.57 and 0.68 for PP. Consistencies ranged between 88.8 and 6.2 Pa sn _{for WP} and between 18.0 and 3.5 Pa sn_{for PP.}

Their rheological behaviour generally exhibited viscoelastic properties (Agoda-Tandjawa et al., 2010; Tatsumi et al., 2001; Paakko et al., 2007). At a concentration of 10%dm and shear rates ranging from 1 to 100 sÿ1_{, the viscosity of corn stover (maize} thresh and residue) and pre-treated softwood suspensions, de-creased from 1.87 to 0.03 and 9 to 0.20 Pa s, respectively ( Pimeno-va and Hanley, 2004; Wiman et al., 2010) (Table 4). Considering dimension criteria, these values are much higher than those for MCC found in the present work.

Surprisingly, the viscosity appears to have the same order of magnitude for dilute and concentrated MCC suspensions (Bayod et al., 2005; Luukkonen et al., 2001) (Table 4). For an MCC concen-tration of 40%dm and for shear rates ranging from 1 to 100 sÿ1_{, the} viscosity of the suspension decreased from 8.0 to 0.3 Pa s ( Luukko-nen et al., 2001). This is similar to the values measured.

3.1.2. Relative viscosity of suspensions

In dilute suspensions, the particles are hydrodynamically inde-pendent and a linear relationship between viscosity and volume fraction is observed. The relative viscosity can be modelled by the Einstein equation:

l

l

0

¼ 1 þ k1

U

¼ 1 þ ½lŠ
Cm ð8Þ

For semi-dilute suspensions, the interactions between the par-ticles begin to interfere and can at first be taken into account by introducing a quadratic term:

0.0001 0.001 0.01 0.1 1 10 100 10 100 1000 V is c o s it y ( P a .s ) Mixing rate (RPM)

Transition curve Water 16.3gdm/L 64.0gdm/L 202.0 gdm/L 273.8 gdm/L 296.1 gdm/L 338.6 gdm/L 378.4 gdm/L

A

Laminar regime Transition regime 0.0001 0.001 0.01 0.1 1 10 100 0 0 0 1 0 0 1 0 1 V is c o s it y ( P a .s ) Mixing rate (RPM)

Transition curve Water 9.7 gdm/L 19.3 gdm/L 28.7 gdm/L 37.9 gdm/L 47.0 gdm/L 64.8 gdm/L

B

0.0001 0.001 0.01 0.1 1 10 100 10 100 1000 V isc os it y (P a. s) Mixing rate (RPM)

Transition curve Water 12.5 gdm/L 16.5 gdm/L 20.4 gdm/L 27.9 gdm/L 31.5 gdm/L 35.1 gdm/L 42.0 gdm/L

C

Fig. 3. Viscosity versus mixing rate at different substrate concentrations. MCC (A), WP (B) and PP (C) (MCC: microcrystalline cellulose, WP: Whatman paper and PP: extruded paper pulp).

Table 3

Evolution of power-law (n) and consistency (k) indexes versus substrate concentra-tion (Cm gdm/L) – (WP: Whatman paper and PP: extruded paper pulp).

Substrate n k

WP: 28.7–64.8 gdm/L n = ÿ0.006 Cm + 0.701 k = 0.724e0.075Cm

PP: 27.9–42.0 gdm/L n = ÿ0.008 Cm + 0.895 k = 0.138e0.116Cm

Table 4

Overview of published results (MCC: microcrystalline cellulose).

Author Substrate D[4, 3] (lm) Cm (%) n k (Pa.sn₎

Pimenova and Hanley (2004) Corn stover 120 5–30 0.05–0.9 0.05–1684

Wiman et al. (2010) Dilute acid pre-treated softwood 109 4–12 0.15–0.4 1–16

Bayod et al. (2005) MCC 33 0–7 0.8–0.9 0.8–2.5

l

l

0

¼ 1 þ

a

U

þ b

U

2 ð9Þ

The third regime corresponds to concentrated suspensions with a lot of contacts between the particles. The viscosity of the suspen-sion increases rapidly with volume fraction. WhenUreaches a critical value, each particle is confined in a cage formed by its near-est neighbours. For volume fractions above this value, only a vibra-tion of the particles inside the cage remains possible, and disappears completely whenUreaches the value of dense packing. Covering all concentration ranges, the most commonly used relationship between relative viscosity and volume fraction was proposed byQuemada (2006). Eq. (10) is used for a Newtonian regime.

l

l

0 ¼ 1 ÿ

U

U

max  ÿn 1 6 n 6 2 ð10Þ

The relative viscosity

l=l

wateris plotted versus the volume frac-tion at the same mixing rate for three suspensions (Fig. 4). In the plot for PP and WP, two regions are observed corresponding to two concentrations: (i) a dilute/semi-dilute concentration range exhibiting a low relative apparent viscosity (l/l0< 100 under 300 rpm) and a quasi-Newtonian behaviour (low viscosity varia-tions with the rotation frequency) with a linear variation of viscos-ity versus volume fraction and (ii) a concentrated regime indicating higher relative viscosity (l/l0> 100), a shear-thinning behaviour (displayed by the decreasing values of the relative vis-cosity when the mixing rate increases) and a strong increase with volume fraction. A critical volume fraction,Ucmay be assumed at the transition between two concentration regimes for all suspensions.

With an identical substrate volume fraction and mixing rate, the relative viscosity decreased from WP, PP to MCC. This may be explained by the differences in particle size and morphology. The particle diameter of the WP fibre is the largest so the relative vis-cosity of this suspension is greater than that of PP and MCC (e.g. for Uc= 0.05,

l

MCC= 2 mPa s,

l

PP= 100 mPa s and

l

WP= 4000 mPa s). For all suspensions, a transition from semi-dilute to concentrated regime is observed. A linear variation was shown for MCC in dilute regime. For an identical mixing rate, one critical volume fraction was identified for each suspensionUc 0.03; 0.09 and >0.24 for

WP, PP and MCC, respectively (Table 5).Luukkonen et al. (2001) proposed a critical volume fraction Uc 0.3 (equivalent to 47%dm) for MCC.

These results show that the viscosity of suspensions is strongly dependent on physical fibre properties among which size and shape as appear to make the major contributions (Horvath and Lindstrom, 2007; Lapierre et al., 2006; Wiman et al., 2010). 3.2. Enzymatic hydrolysis: impact on viscosity and particle size distribution

3.2.1. On-line viscosity

The changes in the physical appearance of the slurry are associ-ated to the biochemical changes occurring in the fibres. Under the action of enzymes, the cellulose chains are cut giving simple prod-ucts such as glucose (ultimate monomer). The glucose concentra-tion increased with the time of hydrolysis (between 1 and 25 h) to reach a final value that was very different for the three sub-strates: roughly 42 g/L for MCC (i.e. 13% bioconversion), 7 g/L for WP (i.e. 12% bioconversion) and 3 g/L for PP (i.e. 10% sion). If amorphous cellulose is taken as reference, the bioconver-sions attain 66.4%, 100%, 30.8% for MCC, WP and PP respectively. Amorphous cellulose was totally or almost totally hydrolysed indi-cating the efficiency of enzymatic attack. The bioconversion into glucose of the matrices studied was comparable to the results re-ported in the literature which vary between 3.6% and 45% (Dasari and Berson, 2007; Pereira et al., 2011; Szijarto et al., 2011).

Considering the conditions investigated (substrate, concentra-tion, and mixing rate) the initial viscosities were coherent with val-ues observed during suspension viscosimetry.

Firstly, a sharp decrease of viscosity was observed with WP and PP during hydrolysis whereas with MCC the reduction was only

Fig. 4. Evolution of the relative viscosity (MCC: microcrystalline cellulose, WP: Whatman paper and PP: extruded paper pulp) versus substrate volume fraction at mixing rate of 300 rpm.

Table 5

Critical volume fractions and substrate concentrations (MCC: microcrystalline cellulose, WP: Whatman paper and PP: extruded paper pulp).

MCC WP PP

Uc >0.24 0.03 0.09

Cm(g/L) 390 36.0 121.1

moderate (Fig. 5A). Under 450 rpm, it was greater for WP, 0.976– 0.001 Pa s and PP, 0.656–0.002 Pa s than for MCC, 0.104– 0.029 Pa s. Viscosity decreased 100 times after 5 h hydrolysis for WP and PP with final values almost reaching that of water. Surpris-ingly, viscosities of WP and PP fell lower than that of MCC.

With WP and PP, the viscosity fell during the first 5 h to reach similar levels. These results are supported by the literature over a wide range of matrices, particle sizes and enzyme/cellulose ratios.

For acid-pretreated sugarcane bagasse, viscosity was reduced by 77% to 95% after 6 h (Geddes et al., 2010) and by 75% to 82% within 10 h (Pereira et al., 2011). This decrease and final plateau depended on the enzyme loading (Geddes et al., 2010).

A typical pseudo-plastic behaviour was confirmed both in the initial step and during hydrolysis (Pereira et al., 2011; Wiman et al., 2010).

For spruce pulp (diameter initial: 91

lm), initial and final

vis-cosities (linitial/lfinal) were 0.24/0.028, 0.4/0.058 and 0.84/ 0.087

lm for concentrations of 10, 15 and 20% (w/w), respectively.

These data were correlated to mean diameters: 44, 53 and 57.5

lm

and conversion yields: 40%, 32% and 25%, respectively (Um, 2007). As mentioned, the decreasing viscosity during enzymatic hydrolysis is reported in literature. In terms of kinetics and propen-sity this mechanism could be explained by several assumptions: (i) the initial biochemical structure and composition of matrices, (ii) the ability to dissolve lignocellulosic material, (iii) the reduction

0.001 0.01 0.1 1 0 5 10 15 20 25 V iscosity (Pa.s) Hydrolysis time (h) MCC WP PP

A

0.001 0.01 0.1 1 0 50 100 150 200 250 V iscosity (Pa.s) D[4,3] (µm) MCC WP PP

B

of particles size and, (iv) the efficiency of the enzyme cocktail (activity, concentration).

3.2.2. Distribution of particle size

The physical properties of each matrix were very different, con-sidering their dimension, shape and compactness. The dimension and shape depend on the morphometry and particle size distribu-tion; they are subject to wide dispersion as illustrated inTable 6.

MCC fibres were dense crystalline particles (1620 kg/m3) with an angular shape (rectangle, square) resembling crystals. WP oc-curred as dissociated long curved fibres. PP suspension included long fibres with ramification. Aspect ratios were 0.605 ± 0.027, 0.448 ± 0.026 and 0.598 ± 0.024 for MCC, WP and PP respectively. Initial mean volume diameters and diameters at 10% and 90% of distribution are given inTable 6. Diameter distributions indicate bimodal populations. Equivalent diameters for fine and coarse frac-tions (maxima) were 30 and 120

lm, 80 and 480

lm, 80 and

350

lm for MCC, WP and PP respectively. The ratio between fine

and coarse populations is determined by considering the minima of the distribution curves. Initially, with WP and PP the major pop-ulation was the fine poppop-ulation, 73.9% ± 1.9 and 70.0% ± 7.0, respectively, while for MCC, the fine population (<60

lm)

repre-sented only 34.1% ± 6.6. Specific surface area exhibited wide heter-ogeneity of mean diameter and associated dispersion.

During hydrolysis, as the fibres were degraded, their length and shape changed significantly (Nguyen et al., 2012). The large parti-cles were hydrolysed; their mean diameter decreased for all sub-strates (Table 6) (suspension heterogeneity is confirmed by D[4,3] variability). The mean diameters were approximately halved within 2 h of hydrolysis, 110.8 to 49.4

lm, 241.6 to 139.2

lm and

276.0 to 167.2

lm for MCC, WP, PP respectively. This led to the

reduction of suspension viscosity (Fig. 5B). However, this effect was observed only for WP and PP for which D[4, 3] > 100

lm while

for MCC (D[4, 3] < 100

lm), the viscosity was not significantly

dependent on fibre mean diameter.

The fine populations increased to reach 84%, 94% and 74% for MCC, WP and PP respectively. With MCC, the halving of the mean diameter of Solkafloc within 25 h has already been reported (Um, 2007). For the hydrolysis of dilute acid pre-treated softwood (D[4, 3] = 109

lm, concentration: 10%w/w): the coarse population

(>100

lm) decreased from 44.2% to 19.7% after 24 h (

Wiman et al., 2010). These tendencies are observed for all substrates no-matter the mixing rate is. The mean diameter decrease in this pres-ent work occurred faster than forWiman et al., 2010reporting that the fibre diameter was stable for 10 h and was then reduced by 20% at 24 h.

For MCC, the hydrolysis effect was mainly observed on coarse particles (Table 6). The initial population tended towards a log-normal distribution (D[4, 3] = 49

lm) after 2 h. For WP, coarse

and fine populations were degraded giving four populations whose average diameters were 3, 20, 75 and 350

lm after 25 h which

indicates a macroscopic cutting effect on fibres. For PP, several mechanisms seem occur. In the first step (Table 6, t = 0.25 h), the split between coarse and fine is strengthened. The fine population increases and translates to a smaller diameter. The reduction pro-cess was observed later for the coarse particles (Table 6, t = 1 h). Around 25 h, a smoothing between coarse and fine particles arose. D[4, 3] increased at 25 h of hydrolysis (from 167.2 to 177.5

lm) as

a result of swelling and unwinding of macro-fibres during the 100 h hydrolysis (Fillaudeau et al., 2011). These results are corre-lated to the decrease of viscosity within 5 h of hydrolysis (Fig. 5A).

4. Conclusion

This study focussing on the rheometry of lignocellulosic suspen-sions explored enzymatic hydrolysis based on physical parameters. The rheometry was dependent on the substrate concentration, the mixing rate imposed (related to shear rate) and the fibre particle size/shape. A method for following viscosity on-line was proposed and used to characterise the rheological behaviour of suspensions as a function of concentration. During enzymatic hydrolysis, the change in viscosity was found due to enzymatic actions and mod-ifications of fibre properties. The decrease of fibre mean diameter could lead to the decrease of suspension viscosity and the effect of enzymatic attack.

Acknowledgement

Authors are grateful to ‘‘Programme de Bourses d’Excellence 2011’’ from the French Embassy in Viet Nam.

References

Agoda-Tandjawa, G., D, S., Berot, S., Blassel, C., Gaillard, C., Garnier, C., Doublier, J.-L., 2010. Rheological characterization of microfibrillated cellulose suspensions after freezing. Carbohydrate Polymers.

Alvira, P., Negro, M.J., Ballesteros, M., 2011. Effect of endoxylanase and alpha-L

-arabinofuranosidase supplementation on the enzymatic hydrolysis of steam exploded wheat straw. Bioresource Technology 102 (6), 4552–4558. Antunes, E.S., 2009. Flocculation Studies in Papermaking. In: Chemical Engineering

Department, University of Coimbra, pp. 199.

Bayod, E., Bolmstedt, U., Innings, F., Tornberg, E., 2005. Rheological characterization of fiber suspensions prepared from vegetable pulp and dried fibers. A comparatible study. Annual Transactions of the Nordic Rheology Society 13, 249–253.

Bennington, C.P.J., Kerekes, R.J., Grace, J.R., 1990. The yield stress of fiber suspensions. Canadian Journal of Chemical Engineering 68 (5), 748–757. Blanco, A., N, C., Fuente, E., Tijero, J., 2006. Rotor selection for a Searle-type device to

study the rheology of paper pulp suspensions. Chemical Engineering and Processing 46, 37–44.

Chase, W.C., Donatelli, A.A., Walkinshaw, J.W., 1989. Effects of Freeness and Consistency on the Viscosity of Hardwood and Softwood Pulp Suspensions (J. Tappi, Trans. English). TAPPI, Norcross, GA, pp. 199.

Chaussy, D., Martin, C., Roux, J.C., 2011. Rheological behavior of cellulose fiber suspensions: application to paper-making processing. Industrial & Engineering Chemistry Research 50 (6), 3524–3533.

Cui, H.P., Grace, J.R., 2007. Flow of pulp fibre suspension and slurries: a review. International Journal of Multiphase Flow 33 (9), 921–934.

Dasari, R.K., Berson, R.E., 2007. The effect of particle size on hydrolysis reaction rates and rheological properties in cellulosic slurries. Applied Biochemistry and Biotechnology 137, 289–299.

Derakhshandeh, B., Kerekes, R.J., Hatzikiriakos, S.G., Bennington, C.P.J., 2011. Rheology of pulp fibre suspensions: a critical review. Chemical Engineering Science 66 (15), 3460–3470.

Duffy, G.G., Titchener, A.L., 1975. Disruptive shear-stress of pulp networks. Svensk Papperstidning-Nordisk Cellulosa 78 (13), 474–479.

Ferreira, A.G.M., Silveira, M.T., Lobo, L.Q., 2003. The viscosity of aqueous suspensions of cellulose fibres. Part 2: Influence of temperature and mix fibres. Silva Lusitana 11 (1), 61–66.

Table 6

Evolution of d(0.1), d(0.9) D[4, 3] (lm) and fine and coarse population (%) during the enzymatic hydrolysis (MCC: microcrystalline cellulose, WP: Whatman paper and PP: extruded paper pulp).

Substrate 0 h 0.25 h 1 h 2 h 25 h MCC d(0.1) 13.9 7.5 6.4 5.4 5.7 D[4, 3] 110.8 75.1 66.5 49.4 49.7 d(0.9) 248.1 197.0 178.8 123.5 126.2 Fine 40.7 61.4 68.5 76.6 76.8 Coarse 59.3 38.6 31.5 23.4 23.2 WP d(0.1) 20.2 18.3 16.8 16.8 11.6 D[4, 3] 241.7 181.6 148.5 139.2 76.7 d(0.9) 707.7 558.2 474.3 423.4 163.4 Fine 72.0 76.4 80.6 85.2 93.9 Coarse 28.0 23.6 19.4 14.8 6.1 PP d(0.1) 21.4 21.4 17.9 15.6 16.7 D[4, 3] 276.0 222.6 206.7 167.2 177.5 d(0.9) 782.8 615.6 600.5 498.4 507.3 Fine 62.9 66.8 71.4 75.6 73.0 Coarse 37.1 33.2 28.6 24.3 27.0

Fillaudeau, L., Babau, M., Cameleyre, X., Lombard, E., Anne-Archard, D., 2011. Libération de substrats fermentescibles à partir de matrices lignocellulosiques issues de l’industrie papetière. Récents Progrès en Génie des Procédés, 101. Geddes, C.C., Peterson, J.J., Mullinnix, M.T., Svoronos, S.A., Shanmugam, K.T., Ingram,

L.O., 2010. Optimizing cellulase usage for improved mixing and rheological properties of acid-pretreated sugarcane bagasse. Bioresource Technology 101 (23), 9128–9136.

Horvath, A.E., Lindstrom, T., 2007. The influence of colloidal interactions on fiber network strength. Journal of Colloid and Interface Science 309 (2), 511–517. Jahangiri, M., Golkar-Narenji, M.R., Montazerin, N., Savarmand, S., 2001.

Investigation of the viscoelastic effect on the Metzner and Otto coefficient through LDA velocity measurements. Chinese Journal of Chemical Engineering 9 (1), 77–83.

Lapierre, L., Bouchard, J., Berry, R., 2006. On the relationship between fibre length, cellulose chain length and pulp viscosity of a softwood sulfite pulp. Holzforschung 60 (4), 372–377.

Luukkonen, P., Newton, J.M., Podczeck, F., Yliruusi, J., 2001. Use of a capillary rheometer to evaluate the rheological properties of microcrystalline cellulose and silicified microcrystalline cellulose wet masses. International Journal of Pharmaceutics 216 (1–2), 147–157.

Metzner, A., Otto, R., 1957. Agitation of non-Newtonian Fluids. AIChE Journal 3 (1), 3–10.

Nguyen, T.C., Anne-Archard, D., Cameleyre, X., Lombard, E., Binet, C., Fillaudeau, L., 2012. Production of fermentescible sugar from paper-pulp: looking for a dynamique and multiscale integrated models based on physical parameters. In: 8th International Conference on Renewable Resources and Biorefineries Toulouse, France, p62.

Paakko, M., Ankerfors, M., Kosonen, H., Nykanen, A., Ahola, S., Osterberg, M., Ruokolainen, J., Laine, J., Larsson, P.T., Ikkala, O., Lindstrom, T., 2007. Enzymatic hydrolysis combined with mechanical shearing and high-pressure homogenization for nanoscale cellulose fibrils and strong gels. Biomacromolecules 8 (6), 1934–1941.

Pereira, L.T.C., Pereira, L.T.C., Teixeira, R.S.S., Bon, E.P.D., Freitas, S.P., 2011. Sugarcane bagasse enzymatic hydrolysis: rheological data as criteria for

impeller selection. Journal of Industrial Microbiology and Biotechnology 38 (8), 901–907.

Pimenova, N.V., Hanley, A.R., 2004. Effect of corn stover concentration on rheological characteristics. Applied Biochemistry and Biotechnology 113, 347–360. Quemada, D., 2006. Modélisation rhéologique structurelle. Dispersions concentrées

et fluides complexes, Lavoisier.

Rushton, J., Costich, W., Everett, J., 1950. Power Characteristics of Mixing Impeller I and II. Chemical Engineering Progess 46 (9), 467–476.

Samaniuk, J.R., Scott, C.T., Root, T.W., Klingenberg, D.J., 2011. The effect of high intensity mixing on the enzymatic hydrolysis of concentrated cellulose fiber suspensions. Bioresource Technology 102 (6), 4489–4494.

Szijarto, N., Siika-aho, M., Sontag-Strohm, T., Viikari, L., 2011. Liquefaction of hydrothermally pretreated wheat straw at high-solids content by purified Trichoderma enzymes. Bioresource Technology 102 (2), 1968–1974. Tatsumi, D., Ishioka, S., Matsumoto, T., 1999. Effect of particle and salt

concentrations on the rheological properties of cellulose fibrous suspensions. Nihon Reoroji Gakkaishi 27 (4), 243–248.

Tatsumi, D., Ishioka, S., Matsumoto, T., 2001. Effect of fiber concentration and axial ratio on the rheological properties of cellulose fiber suspensions. Journal of Society of Rheology 30 (1), 27–32.

Tatsumi, D., Matsumoto, T., 2007. Rheological properties of cellulose fiber wet webs. Journal of Central South University of Technology 14, 250–253.

Um, B.-H. 2007. Optimization of Ethanol Production from Concentrated Substrate, Auburn University, pp. 268.

Um, B.H., Hanley, T.R., 2008. A comparison of simple rheological parameters and simulation data for Zymomonas mobilis fermentation broths with high substrate loading in a 3-L bioreactor. Applied Biochemistry and Biotechnology 145 (1/3), 29–38.

Wiman, M., Palmqvist, B., Tornberg, E., Liden, G., 2010. Rheological characterization of dilute acid pretreated softwood. Biotechnology and Bioengineering 108 (5), 1031–1041.

Zhang, X., Qin, W., Paice, M.G., Saddler, J.N., 2009. High consistency enzymatic hydrolysis of hardwood substrates. Bioresource Technology 100 (23), 5890– 5897.