Accepted Manuscript

Title: Inverse Gas Chromatography for Natural FibreCharacterisation: Identification of the Critical Parameters toDetermine The Brunauer Emmett Teller Specific Surface Area

Author: A. Legras A. Kondor M.T. Heitzmann R.W. Truss

PII: S0021-9673(15)01661-1DOI: http://dx.doi.org/doi:10.1016/j.chroma.2015.11.033Reference: CHROMA 357047

To appear in: Journal of Chromatography A

Received date: 14-9-2015Revised date: 10-11-2015Accepted date: 10-11-2015

Please cite this article as: A. Legras, A. Kondor, M.T. Heitzmann, R.W. Truss, InverseGas Chromatography for Natural Fibre Characterisation: Identification of the CriticalParameters to Determine The Brunauer Emmett Teller Specific Surface Area, Journalof Chromatography A (2015), http://dx.doi.org/10.1016/j.chroma.2015.11.033

This is a PDF file of an unedited manuscript that has been accepted for publication.As a service to our customers we are providing this early version of the manuscript.The manuscript will undergo copyediting, typesetting, and review of the resulting proofbefore it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers thatapply to the journal pertain.

Page 1 of 28

Accep

ted

Man

uscr

ipt

BET specific surface areas were determined by Inverse Gas Chromatography for

samples of flax, kenaf and cellulose fibres.

The effect of experimental conditions on the BET surface area values were

investigated.

Bast fibres showed a large variability within a batch compared to synthesised

cellulose fibres.

An experimental procedure to determine the BET surface area values for natural

fibres is proposed.

LegrasHighlights.doc

Page 2 of 28

Accep

ted

Man

uscr

ipt

Inverse Gas Chromatography for Natural Fibre

Characterisation: Identification of the Critical

Parameters to Determine The Brunauer Emmett Teller

Specific Surface Area

A. Legrasa,b,∗, A. Kondorc, M.T. Heitzmanna, R.W. Trussa

aSchool of Mechanical and Mining Engineering, The University of Queensland, Brisbane,QLD 4072, Australia

bCooperative Research Centre for Advanced Composites Structures Australia Pty Ldt,1/320 Lorimer Street, Port Melbourne, VIC 3207, Australia

cSurface Measurement Systems LTD, 5 Wharfside, Rosemont Road, Alperton, MiddlesexHA0 4PE, UK

Abstract

Inverse Gas Chromatography (IGC) is an alternative technique to determinethe specific surface area of natural fibres. Natural fibres have a complexsurface chemistry and unique microstructure that challenge the current ca-pabilities to perform surface characterisation. This study investigated theinfluence of multiple parameters on the measured Brunauer Emmett Teller(BET) specific surface area for samples of flax, kenaf and BioMid R© cellulosefibres using IGC. The BET surface area of kenaf and flax differed with 0.51m2.g−1 and 1.35 m2.g−1 respectively, the former being similar to the cellu-lose fibres (0.54 m2.g−1). The data was calculated under conditions wherethe BET equation showed good linearity (R2 > 0.995). Repeatability wasexcellent so that two runs sufficed to obtain representative BET surface areavalues. The findings showed the choice of solvent was important for all spec-imens to avoid any misleading data comparison due to molecular orientationeffects that impact the adsorbent-adsorbate interactions. The higher sur-face area of the flax sample, and its higher variability, was correlated witha higher surface roughness observed under optical microscopy. Packing the

∗Corresponding author. Tel: +61 7 3346 9570, Fax: +61 733 654 799Email address: a.legras@uq.edu.au (A. Legras)

Preprint submitted to Journal of Chromatography A November 10, 2015

LegrasManuscript

Page 3 of 28

Accep

ted

Man

uscr

ipt

chromatography column with long or chopped fibres produced results thatwere statistically insignificant.

Keywords: Brunauer Emmett Teller theory, Inverse Gas Chromatography,Natural Fibres, Specific Surface Area

1. Introduction1

Bast fibres have been traditionally destined for the textile industry and2

this remains the primary application with constant innovations for techni-3

cal clothing, textiles etc. However, the last decades have seen a significant4

trend to utilise natural fibres in other sectors, particularly the automotive5

industry [1, 2, 3]. Natural fibres appeal to vehicle manufacturers with their6

excellent strength to weight ratio, low cost, low carbon footprint and avail-7

ability. They are integrated into polymer matrices as filler or reinforcement8

elements for interior components [4, 5]. Biocomposites have also emerged as9

an alternative to wood plastic composites for building materials.10

11

Diversification into novel applications places new demands on the fibre12

processing and properties. One of the major issues is that natural fibres are13

generally hydrophilic and consequently are inherently incompatible with hy-14

drophobic commodity polymers. Natural fibres also suffer from considerable15

batch-to-batch heterogeneity and particularly dimensional variability, which16

directly affect the tensile properties. Natural fibre moisture sensitivity is an-17

other issue for the biocomposite durability.18

19

The compatibility, dispersibility and reinforcement capability of natural20

fibres are related to the fibre surface energy and to its specific surface area.21

Bast fibres have a complex 3D microstructure with multiple membranes en-22

closing the lumen. The chemical composition and the molecular arrangement23

vary between the layers and depend on the fibre specie. The fibre location24

within the plant stem and the growth conditions also play a major role in the25

physico-chemical structure, creating fibres with unique and complex surfaces26

[6, 7]. Figure 1 illustrates the architecture of bast fibre bundles with cellulose27

microfibrils embedded into a matrix of hemicellulose and lignin. The outer28

layer also contains pectins, waxes and other extractives. The fibre surface29

chemistry, its adsorption capacity as well as its wettability and dispersability30

2

Page 4 of 28

Accep

ted

Man

uscr

ipt

in a matrix correlate with the fibre surface energy values and depend also31

on the accessible surface area. The surface energy provides information on32

the intermolecular forces that can occur at the fibre surface. These combine33

long range Van der Waals forces and short range chemical forces, also known34

as dispersive and polar forces respectively. The nature of these interactions35

and their intensity also depend on the accessible surface area.36

37

Figure 1: Structure of an elementary plant fibre showing the different layers and the ori-entation of the cellulose microfibrils (Adapted from [8] and reproduced with authorisationof the author).

Traditional methods used to determine the fibre surface energy involve38

the measurement of contact angles. Various approaches reviewed by Williams39

[9] and Heng et al. [10] provide methods for measuring droplet angles to cal-40

culate the fibre surface tension. The most common are the Young model,41

the Fowkes and extended Fowkes (Owens & Wendt) approaches and the Van42

Oss et al. model. Capillary rise, Wilhelmy plate and sessile drop are com-43

mon methods and numerous studies have been published on natural fibres44

[11, 12, 13]. Although these techniques are excellent on flat surfaces, nat-45

3

Page 5 of 28

Accep

ted

Man

uscr

ipt

ural fibre’s porous structure and heterogeneous surface properties challenge46

the accuracy and the validity of the experimental data. These methods are47

based on liquid-solid interactions where swelling and dissolution may occur48

and skew the data.49

50

Inverse Gas Chromatography is an alternative tool to study surface en-51

ergies. IGC is based on solid-gas interactions and the affinity between gas52

probe molecules and the fibre surface molecules is quantified. It is a versatile53

technique to characterise samples of any shape, as long as the specimen can54

be packed in the column and molecular interactions can occur. IGC exists55

since the early 50s and it has been applied mostly to pharmaceutical industry56

[14], for characterisation of zeolites [15, 16] and carbon nanostructures [17]57

but IGC has not been well established yet for natural fibre characterisation.58

Most of the studies on natural fibres have been performed with home built59

equipment, using different solvents and various calculation procedures. The60

lack of information and details of experimental procedures make it difficult61

to compare and validate data. Moreover, the large variability of natural fibre62

properties and their complexity compared to synthesised and well designed63

man made materials means that a systematic study of the common proce-64

dures and models implemented in IGC is required to assess its capabilities65

as a technique to characterise natural fibre surfaces.66

67

IGC requires the specific surface area to determine the surface energy, es-68

pecially as this can depend on surface coverages. Usually, the specific surface69

area of a solid is determined by BET method which uses nitrogen sorption70

[18, 19] or krypton [20, 21], for surfaces areas below 0.5 m2.g−1 where the71

nitrogen technique shows limitations [22]. These techniques involve extreme72

conditions of high vacuum at low temperature (77K) under which the fibre73

properties are likely to change and thus the BET surface area.74

75

An alternative and preferred technique for measuring the BET surface76

area of low surface area natural fibres would be to use IGC at room temper-77

ature. This paper systematically studies the influence of various parameters78

that may affect the measured specific surface area using BET theory with79

IGC. On the basis of the results of this study, a procedure applicable to nat-80

ural fibres is proposed.81

82

4

Page 6 of 28

Accep

ted

Man

uscr

ipt

2. BET theory83

The BET theory was developed with nitrogen but is applicable to other84

gases such as those used in IGC. Five types of isotherms can occur depend-85

ing on the adsorption scenario [23, 24]. The BET equation is applicable on86

isotherms type II and IV only, where there is a formation of a monolayer fol-87

lowed by multi-layers and further capillary condensation. The BET equation88

is given by:89

P

n(P0 − P )=

C − 1

nmC(P

P0

) +1

nmC(1)

90

where P is the solvent partial pressure in the gas phase (Torr), P0 the satu-91

rated solvent vapor pressure (Torr), n the amount of gas adsorbed (Mol.g−1),92

nm the monolayer capacity (Mol.g−1) and C the BET constant. The BET93

equation fits the isotherm (type II or IV) over a specific range of equilibrium94

pressure P/P0, usually for 0.05 < P/P0 < 0.35. The monolayer capacity95

nm can be determined from the slope and intercept of the linearised BET96

equation fitted to the isotherm. The BET specific surface area (m2.g−1) is97

expressed as:98

SBET = aNAnm (2)

99

with a the molecule cross section area, NA the Avogadro Number and nm100

the monolayer capacity. Since nm and a are known, the specific surface area101

SBET can be calculated.102

103

3. Experimental procedure104

3.1. Materials105

Two types of bast fibres and one type of cellulose fibre were used in the106

study. The Canadian linseed flax (variety unknown) specimen was supplied107

by Composites Innovation Centre (CIC), Winnipeg, Manitoba (Canada).108

The fibres were not retted and have been mechanically decorticated by a lab109

scale scutching machine [25]. Kenaf fibres (variety KK60) were provided by110

Engage Eco Products Co. Ltd. in Thailand. The fibres were locally ribbon111

retted, rinsed and dried before shipping. Both flax and kenaf samples were112

5

Page 7 of 28

Accep

ted

Man

uscr

ipt

characterised in the as-received conditions. BioMid R© cellulose fibres were113

supplied by ENC International (South Korea). BioMid R© is a 100% cellulose-114

based continuous filament produced from dry-jet-wet spinning process. The115

feedstock is a mixture of softwood and hardwood chips, a by-product from the116

wood pulp and paper industry. The cellulose is extracted from the biomass117

and then injected through a spinneret. Also, BioMid R© structure is different118

from the bast fibres that have a membrane structure. BioMid R© fibres are119

expected to be a pure cellulose sample and was considered as a reference for120

the study. The sample details are summarised in Table 1.121

Table 1: Industrial fibre grades used for the study

Sample Variety Fibre processing

Cellulose BioMid R© (ENC International,South Korea)

Dry-jet-wet spinningprocess

Kenaf KK60 (Thailand) Water retting

Flax Linseed flax (Canadian varietyunknown)

Mechanical decortica-tion by scutching

3.2. Methods122

The strategy to study the influence of experimental parameters on the123

output BET value involved four criteria. The experimental approach is de-124

tailed in the following paragraphs and summarised in Table 2.125

126

• Reproducibility within the column:127

Various diffusion processes can drive the elutant molecule flow in a column128

packed with porous material. These scenarios depend on the column dimen-129

sion, the sample porosity, the packing homogeneity and the flow rate [23].130

Natural fibre pore widths range from micrometers down to a few nanometers131

[26, 27, 28] and so these exceed in size the elutant molecules (cross sectional132

area ca. 10 A2 to 100 A2 [29]). As a consequence, the free diffusion pro-133

cess dominates the molecule flow into the column: the gas probe molecules134

travel both in axial and longitudinal directions. As they elute, these likely135

encounter cavities and asperities where they will adsorb before complete elu-136

tion. Whether this phenomenon occurs randomly and if it further affects the137

flow rate is questionable. Successive runs were performed on the same chro-138

matography column under identical experimental conditions to assess the139

6

Page 8 of 28

Accep

ted

Man

uscr

ipt

reproducibility of the BET experiment. Note that this experiment is pos-140

sible with inverse gas chromatography as the column can be used multiple141

times.142

143

• Gas probe:144

Various solvents can be used to run a BET experiment with IGC, the re-145

quirement being non-polar adsorbates where surface and no bulk sorption146

occurs. Octane, heptane and cyclohexane are common adsorbates. Whether147

the nature of the solvent affects the BET value of natural fibres or not has148

not been clarified so far and few authors specify which gas probe has been149

used to determine the specific surface area. For the sake of clarity and to150

know if data obtained from various solvents can be compared, the impact of151

the solvent choice on the output value was investigated. Among the common152

adsorbates used for the BET experiment, octane and cyclohexane were se-153

lected for two main reasons. The latter showed better retention peaks than154

other solvents, for instance, hexane and heptane had too low retention times.155

The second reason is that octane and cyclohexane differ in their molecular156

structure and chemical properties, which facilitates the observation of effects157

due to molecule geometry.158

159

• Variability within a batch:160

The variability of the BET specific surface area measured using octane was161

investigated within a batch of natural fibres. The specific surface area is162

expected to fluctuate as the diameter, porosity and the surface profile vary163

between fibres. Little information is currently available as to whether the164

specific surface area changes and to what extent. A chromatography column165

usually contains ca. a gram of fibres i.e. a relatively small amount of mate-166

rial. It is necessary to estimate how variable the BET value is for grading167

procedures.168

169

• Sample packing:170

Another variable rarely specified is the sample packing. Unlike powder par-171

ticles, natural fibres can be packed in the column in multiple ways. For172

instance, post-processed fibres are usually chopped and well separated com-173

pared with fibres as received. Short and long fibres may behave differently in174

the inverse gas chromatography column. In this experiment, the fibres were175

cut into ”short” fibres of 2 cm length and compared with ”long” fibres of176

ca. 10 cm length (usual fibre length in a column of 4 mm internal diameter).177

7

Page 9 of 28

Accep

ted

Man

uscr

ipt

178

Table 2: Strategy to investigate the critical parameters for BET experiment

Criterion Experimental Method

Reproducibility within the column Repeat 3 runs per columnGas probe Run with octane and cyclohexaneVariability within a batch Run 5 columns per batch

Sample packing Pack the column with long (10 cm)and short (2 cm) fibres

All experiments were conducted with an Inverse Gas Chromatograph Sur-179

face Energy Analyser (IGC SEA) from Surface Measurement Systems (Lon-180

don, UK). This commercial equipment is set-up for pulse chromatography;181

a precise amount of adsorbate is transported by the carrier gas through the182

column containing the fibres. Adsorption followed by desorption occur at the183

fibre surface and an elution peak results. The configuration of the IGC SEA184

is schematised in Figure 2. The retention time was determined by a Flame185

Ionization Detector (FID) to benefit a high sensitivity compared to thermal186

conductivity detector [14, 30]. Silane-treated glass columns were filled with187

ca. 0.7 g to 1 g of fibres. The 4 mm internal diameter column was preferable188

for packing the fibres. To insert the fibres in a column, wax-free dental floss189

was tied to the end of the fibres and then pulled through the column. Once190

the fibres were in place, the dental floss was removed and the column was191

then plugged with silanised glass wool to avoid any contamination in the in-192

jection system. For the experiment with chopped fibres, the short fibres were193

introduced in the column using a funnel and then packed with a column pack-194

ing device. The columns were then plugged as mentioned previously. The195

sample bed length was ca. 30 mm to minimise peak broadening due to free196

molecular diffusion in the column. Helium was the carrier gas and methane197

was the reference gas to determine the dead time, which represents the time198

necessary for a molecule to travel across the column without any interaction.199

Octane and cyclohexane were injected over a coverage range (n/nm) within200

0.01 to 0.44, the minima and maxima values depending on the sample mass.201

All experiments were carried out under the same conditions (30◦C, 0% RH)202

with column conditioning for 1 hour (40◦C, 0% RH) before the first injection203

only. The carrier gas flow rate was set up at 10 mL.min−1.204

205

8

Page 10 of 28

Accep

ted

Man

uscr

ipt

Figure 2: IGC Surface Energy Analyser set up (Modified image supplied by SMS).

The BET theory was applied to determine the specific surface area of the206

fibres according to the following procedure:207

208

- The retention time was determined as the time corresponding to the209

peak centre of mass (CoM) rather than the time of the maximum FID signal.210

The peak CoM was preferred as most of the elution peaks were asymmetric.211

212

- The solvent vapour pressure P0 was calculated with the modified Antoine213

equation [29] described as:214

P0 = exp[C1 +C2

T+ C3ln(T ) + C4T

C5 ] (3)

215

with C1, C2, C3, C4 and C5 constants specific to the solvent and T the216

temperature (K). P0 is expressed in Pa.217

218

- The linearised BET equation was fitted to the isotherms (amount ad-219

sorbed vs. relative pressure) in the range of 0.05 < P/P0 < 0.35. The range220

9

Page 11 of 28

Accep

ted

Man

uscr

ipt

of calculation was adjusted so that R2 ≥ 0.995 with P/P0 upper limit down221

to a minimum of 0.25 to get representative values.222

4. Results and Discussions223

4.1. Reproducibility within a column224

Table 3 shows the repeatability of the BET experiment with octane on225

both bast fibres and on the BioMid R© sample. In each case, all the runs226

were performed on the same column. An illustration of the data fitted to the227

linearised BET equation for the kenaf sample is given in Figure 3.228

229

Table 3: Reproducibility BET experiment

BET Specific Surface Area (Octane) (m2.g−1) at 30◦C and 0% RH

Specimen Run 1 Run 2 Run 3 Mean Std (%)

BioMid R© 0.546 0.545 0.543 0.545 0.1Kenaf 0.503 0.494 0.501 0.500 0.5Flax 1.373 1.423 1.440 1.412 3.5

With standard deviations less than 5% for all samples, the BET experi-230

ment showed excellent reproducibility. One run should suffice to determine231

the BET specific surface area of a fibre specimen but it is suggested that two232

runs be completed to avoid any possible outlier. The BET surface area of233

BioMid R© and kenaf fibres were similar (0.55 m2.g−1 and 0.50 m2.g−1) whilst234

the flax fibre showed tripled specific surface area (1.41 m2.g−1). This could235

be directly related to the surface roughness of the flax fibres. This batch has236

not been retted, which means that the microbial degradation process that237

helps separation of the fibres from non-fibrous tissues was omitted. Morri-238

son et al. [31] and Akin et al. [32] showed that insufficient retting leads239

to poor separation of the non-fibrous material (cuticle/epidermis and woody240

core) from the bast fibres. These remaining tissues tend to entangle with241

the fibres during next mechanical processing steps of scutching and hack-242

ling. In this study, the flax fibres were scutched without being retted and243

hence it was expected to observe numerous non-fibrous tissues spread on the244

surface. Optical microscope images (Figure 4) clearly showed that the flax245

fibres were covered with fragments of cuticle and epidermis tissues whereas246

10

Page 12 of 28

Accep

ted

Man

uscr

ipt

Figure 3: Plot of P/n(P0 − P ) versus P/P0 for successive runs on kenaf fibres.

the kenaf fibre surface appeared neat and clean, similar to the BioMid R fi-247

bres. The flax fibre surface roughness and heterogeneity accounted for a high248

BET surface area. BET surface areas of plant fibres were measured with IGC249

under similar conditions (30◦C, 0% RH); Ashori et al. [33] found to cotton250

fibres specific surface area of 0.75 m2.g−1 and Cordeiro et al. [34] obtained251

BET area values from bast fibres and other plant fibres that ranged from252

0.10 m2.g−1 to 2.79 m2.g−1. BET specific surface areas were also reported253

on grass fibres with values between 0.81 m2.g−1 and 1.19 m2.g−1 [35].254

255

11

Page 13 of 28

Accep

ted

Man

uscr

ipt

Figure 4: Optical microscope images of A) BioMid R©, B) Kenaf and C) Flax fibre samples(Images obtained with an Axio Zoom.V16 microscope by CIC).

12

Page 14 of 28

Accep

ted

Man

uscr

ipt

4.2. Solvent dependence256

BET values calculated with different solvents are shown in Table 4. For257

each specimen, both octane and cyclohexane BET experiments were per-258

formed on the same sample column with a minimum of 2 runs. Both BioMid R©259

and bast fibres showed a similar trend; BET values obtained with cyclohexane260

were lower than those obtained with octane. These results should be related261

to molecular orientation. Cyclohexane has predominantly a chair conforma-262

tion whilst octane is linear, so the effect of molecular orientation is stronger263

in the latter. The average cross sections are 6.3 × 10−19 m2 and 3.9 × 10−19264

m2 for octane and cyclohexane respectively, but octane width cross section265

is much smaller hence the molecules can access pores that are “invisible”for266

cyclohexane. This explains why BET values measured with octane are higher267

than those calculated with cyclohexane. The effect of molecular orientation268

and the consideration of uncertainties due to average “a”values for the cal-269

culation of surface energy have been previously discussed by Donnet et al.270

and Mukhopadhyay et al. [36, 37].271

272

Note that flax and kenaf fibres showed a large difference between octane273

and cyclohexane BET experiment compared to BioMid R© sample (ca. 50%274

and 15% difference respectively). The BioMid R© fibres were manufactured275

using a wet spinning process. These fibres are expected to be relatively276

homogeneous and to have relatively smooth surface even at the molecular277

level. Natural fibres on the other hand are known to be highly heterogeneous278

both physically and chemically (see Figure 1). Consequently, their surface279

might be expected to have asperities over a range of length scales down to280

the molecular level. Such features would amplify the effects of adsorbate281

orientation on the surface. The more linear octane molecule would have the282

ability to pack more densely on the surface giving a higher specific surface283

area while the cyclohexane molecule would pack less densely on the surface284

and generate lower measured specific surface areas. This is consistent with285

the observed results.286

287

4.3. Variability within a batch288

As expected, the variability within the bast fibre batches was more pro-289

nounced than for the BioMid R© fibres. As illustrated in Fig.5, the latter290

averaged 0.54 m2.g−1 and fell within 0.5 m2.g−1 and 0.58 m2.g−1 i.e. ± 7%291

13

Page 15 of 28

Accep

ted

Man

uscr

ipt

Table 4: Effect of adsorbateBET Specific Surface Area (m2.g−1)

Specimen Solvent Mean Std (%)

BioMid R© Octane 0.55 0.1Cyclohexane 0.47 0.2

Kenaf Octane 0.50 0.5Cyclohexane 0.27 0.2

Flax Octane 1.41 3.5Cyclohexane 0.75 0.3

variation. Bast fibre BET surface area values spread over wider range; be-292

tween 1.22 m2.g−1 and 1.49 m2.g−1 for flax and 0.38 m2.g−1 to 0.63 m2.g−1 for293

kenaf batches, hence a variation of ca. ± 10% and ± 25% respectively. This294

was to be expected as natural materials have quite variable structure due295

to growth conditions, position within the plant, and damage during harvest-296

ing and processing. The range of BET surface areas reflects this phenomenon.297

298

14

Page 16 of 28

Accep

ted

Man

uscr

ipt

Figure 5: Individual value plots of the BET Surface Area (m2.g−1) with 95% ConfidenceInterval (CI).

4.4. Sample packing299

Table 5 shows the effect of fibre length on the BET values. The ranges of300

variation for BioMid R© (ca. +10%) and kenaf (ca. +20%) fibres agree well301

with the previous findings on the variability within a batch and the effect of302

fibre length could be considered negligible in that case. However, short flax303

fibres BET surface area values stepped outside the confidence interval (95%304

CI) with a variation of 20% i.e. chopping the fibres induced significant effects.305

The non consistency of these results could be related to the physico-chemical306

differencies between the fibres. As seen in Figure 4, both BioMid R© and307

kenaf fibre surface were neat and homogeneous compared to the flax fibres308

that were unretted. Chopping the latter may have shredded the fibre ends,309

i.e. multiplied the accessible surface area. It may also have peeled off some310

of the cuticle/epidermis fragments and hence opened access to new surfaces311

that displayed different chemistries than the outer layer. Both phenomena312

likely generated new molecular interactions with the adsorbate.313

314

15

Page 17 of 28

Accep

ted

Man

uscr

ipt

Table 5: Effect of fibre length

BET Specific Surface Area (m2.g−1)

Specimen Packing Mean Std (%)

BioMid R© (Octane) long 0.55 0.1short 0.59 0.4

Kenaf (Octane) long 0.50 0.5short 0.42 0.1

Flax (Cyclohexane) long 0.75 0.3short 0.92 1.5

5. Conclusions315

The BET specific surface area of kenaf and flax fibres differed, with an316

average of 0.51 m2.g−1 vs. 1.35 m2.g−1 respectively and the kenaf fibres317

showed similar BET value to cellulose fibres (ca. 0.54 m2.g−1). The high318

specific area of flax, compared with kenaf and cellulose, was related to the fi-319

bre surface roughness. Bast fibres had larger batch-to-batch variability than320

synthesised cellulose fibres, which is a consequence of natural fibre structural321

irregularities and heterogeneous properties.322

323

The BET values obtained by IGC SEA showed a noticeable dependence324

on the elutant properties. For all specimens, the specific surface areas calcu-325

lated from octane measurements were higher than those from cyclohexane.326

This phenomenon is likely an effect of molecular orientation. Sample packing327

also affected the BET surface area values but no clear trend could be estab-328

lished. It is possible that chopping the unretted flax fibres either shredded329

the fibre ends or removed lightly adherent cuticle /epidermal material on the330

surface increasing the accessible surface area.331

332

Based on these findings, the following protocol for determining the BET333

specific surface area of natural fibres by IGC is proposed:334

335

• Pack the chromatography column with the sample as is. Chopping fi-336

bres may induce non negligible effects.337

338

• Consider BET values obtained with the same solvent only for direct339

16

Page 18 of 28

Accep

ted

Man

uscr

ipt

comparison.340

341

• For accurate calculation, consider the linearised BET equation over a342

coverage range (n/nm) so that the correlation factor R2 > 0.995.343

344

• Since repeatability is excellent (commercial equipment), two runs per345

column should be sufficient to ensure confident results, assuming none of the346

data is an outlier.347

348

These experimental data highlight the structural heterogeneity between349

different species of bast fibres, in term of both chemical and physical singu-350

larities. Further data acquisition on natural fibres is necessary to strengthen351

these models and extend the database to get consistent references. However,352

the current results have demonstrated the potential of IGC for characteri-353

sation of natural fibre surfaces. The authors encourage the development of354

inverse gas chromatography for fibre grading as a complementary technique355

to traditional methods.356

6. Acknowledgements357

This study was undertaken as part of a collaboration between the Com-358

posites Innovation Centre (CIC) and the Cooperative Research Centre for359

Advanced Composites Structures Australia (CRC-ACS) research project, es-360

tablished and supported under the Australian Government’s Cooperative361

Research Centre Program. The author acknowledge the CRC-ACS Project362

P1.1 team and the CIC for their access to equipment. The technical support363

from Surface Measurement Systems to develop the IGC experimental work364

is also gratefully acknowledged.365

7. References366

[1] J. Holbery, D. Houston, Natural-fiber-reinforced polymer composites in367

automotive applications, JOM 58 (11) (2006) 80–86.368

[2] A. Ashori, Wood-plastic composites as promising green-composites for369

automotive industries!, Bioresource Technology 99 (11) (2008) 4661 –370

4667, exploring Horizons in Biotechnology: A Global Venture.371

17

Page 19 of 28

Accep

ted

Man

uscr

ipt

[3] M. Karus, M. Kaup, Natural fibres in the european automotive industry,372

Journal of Industrial Hemp 7 (1) (2002) 119–131.373

[4] M. Huda, L. Drzal, D. Ray, A. Mohanty, M. Misra, Natural-fibre com-374

posites in the automotive sector, Properties and performance of natural-375

fibre composites (2008) 256–261.376

[5] G. Marsh, Next step for automotive materials, Materials Today 6 (4)377

(2003) 36 – 43.378

[6] K. Charlet, C. Baley, C. Morvan, J. Jernot, M. Gomina, J. Brard, Char-379

acteristics of herms flax fibres as a function of their location in the380

stem and properties of the derived unidirectional composites, Compos-381

ites Part A: Applied Science and Manufacturing 38 (8) (2007) 1912 –382

1921.383

[7] K. Satyanarayana, J. Guimaraes, F. Wypych, Studies on lignocellulosic384

fibers of brazil. part i: Source, production, morphology, properties and385

applications, Composites Part A: Applied Science and Manufacturing386

38 (7) (2007) 1694 – 1709.387

[8] C. Baillie, 4. Natural Fibre Sources, Green Composites - Polymer Com-388

posites and the Environment, Woodhead Publishing, 2004.389

[9] D. R Williams, Particle engineering in pharmaceutical solids processing:390

surface energy considerations, Current pharmaceutical design 21 (19)391

(2015) 2677–2694.392

[10] J. Y. Y. Heng, D. F. Pearse, F. Thielmann, T. Lampke, A. Bismarck,393

Methods to determine surface energies of natural fibres: A review, Com-394

posite Interfaces 14 (7-9) (2007) 581–604.395

[11] G. Collins, A surface tension method for measuring the perimeters of396

fibres and the contact angle of liquids against fibres, Journal of the397

Textile Institute Transactions 38 (2) (1947) T73–T77.398

[12] H. Shen, S. Nutt, D. Hull, Direct observation and measurement of fiber399

architecture in short fiber-polymer composite foam through micro-ct400

imaging, Composites Science and Technology 64 (13) (2004) 2113–2120.401

18

Page 20 of 28

Accep

ted

Man

uscr

ipt

[13] M. Z. Rong, M. Q. Zhang, Y. Liu, H. M. Yan, G. C. Yang, H. M.402

Zeng, Interfacial interaction in sisal/epoxy composites and its influence403

on impact performance, Polymer composites 23 (2) (2002) 182–192.404

[14] S. Mohammadi-Jam, K. E. Waters, Inverse gas chromatography appli-405

cations: A review, Advances in Colloid and Interface Science 212 (0)406

(2014) 21–44.407

[15] A. Askin, C. Bilgic, Thermodynamics of adsorption of hydrocarbons on408

molecular sieves nay and cay by inverse gas chromatography, Chemical409

Engineering Journal 112 (1) (2005) 159–165.410

[16] A. V. J. C. Eva Dıaz, Salvador Ordonez, Adsorption characterisation of411

different volatile organic compounds over alumina, zeolites and activated412

carbon using inverse gas chromatography, Journal of chromatography A413

1049 (1) (2004) 139–146.414

[17] E. Papirer, E. Brendle, F. Ozil, H. Balard, Comparison of the surface415

properties of graphite, carbon black and fullerene samples, measured by416

inverse gas chromatography, Carbon 37 (8) (1999) 1265–1274.417

[18] H. K. Livingston, The cross-sectional areas of molecules adsorbed on418

solid surfaces, Journal of Colloid Science 4 (5) (1949) 447–458.419

[19] A. Bismarck, I. Aranberri-Askargorta, J. Springer, T. Lampke,420

B. Wielage, A. Stamboulis, I. Shenderovich, H.-H. Limbach, Surface421

characterization of flax, hemp and cellulose fibers; surface properties422

and the water uptake behavior, Polymer Composites 23 (5) (2002) 872–423

894.424

[20] R. A. Beebe, J. B. Beckwith, J. M. Honig, The determination of small425

surface areas by krypton adsorption at low temperatures, Journal of the426

American Chemical Society 67 (9) (1945) 1554–1558.427

[21] A. J. Rosenberg, Rapid, precise measurements of krypton adsorption428

and the surface area of coarse particles1, Journal of the American Chem-429

ical Society 78 (13) (1956) 2929–2934.430

[22] K. Sing, The use of nitrogen adsorption for the characterisation of porous431

materials, Colloids and Surfaces A: Physicochemical and Engineering432

Aspects 187 (2001) 3–9.433

19

Page 21 of 28

Accep

ted

Man

uscr

ipt

[23] F. Thielmann, Introduction into the characterisation of porous materials434

by inverse gas chromatography, Journal of Chromatography A 1037 (12)435

(2004) 115–123.436

[24] B. Charmas, R. Leboda, Effect of surface heterogeneity on adsorption on437

solid surfaces: Application of inverse gas chromatography in the studies438

of energetic heterogeneity of adsorbents, Journal of Chromatography A439

886 (12) (2000) 133–152.440

[25] M. Alcock, L. Van Loon, C. Karunakaran, S. Huo, S. Potter, C. Ul-441

ven, High resolution characterisation of biofibre/resin interface bonds of442

thermoset composites through synchrotron illumination, ECCM15, 15th443

European Conference on Composite Materials, Italy, 24-28 June 2012.444

[26] A. K. Bledzki, M. Letman, A. Viksne, L. Rence, A comparison of com-445

pounding processes and wood type for wood fibre-pp composites, Com-446

posites Part A: Applied Science and Manufacturing 36 (6) (2005) 789–447

797.448

[27] J. E. Stone, A. M. Scallan, Effect of component removal upon the porous449

structure of the cell wall of wood, Journal of Polymer Science Part C:450

Polymer Symposia 11 (1) (1965) 13–25.451

[28] M. A. Arsene, K. Bilba, H. Savastano Junior, K. Ghavami, Treatments452

of non-wood plant fibres used as reinforcement in composite materials,453

Materials Research 16 (2013) 903–923.454

[29] R. H. Perry, D. W. Green, J. O. Maloney, Perry s chemical engineers455

handbook, McGraw Hill, New York, 1997.456

[30] W. Wang, Q. Hua, Y. Sha, D. Wu, S. Zheng, B. Liu, Surface properties457

of solid materials measured by modified inverse gas chromatography,458

Talanta 112 (2013) 69–72.459

[31] M. J. Morrison, A. G. Andre, Processing apparatus and methods, uS460

Patent App. 10/586,194 (Jan. 14 2005).461

[32] D. E. Akin, L. L. Rigsby, W. Perkins, Quality properties of flax fibers462

retted with enzymes, Textile research journal 69 (10) (1999) 747–753.463

20

Page 22 of 28

Accep

ted

Man

uscr

ipt

[33] A. Ashori, M. Ornelas, S. Sheshmani, N. Cordeiro, Influence of mild464

alkaline treatment on the cellulosic surfaces active sites, Carbohydrate465

Polymers 88 (4) (2012) 1293–1298.466

[34] N. Cordeiro, C. Gouveia, A. Moraes, S. Amico, Natural fibers character-467

ization by inverse gas chromatography, Carbohydrate Polymers 84 (1)468

(2011) 110 – 117.469

[35] N. Cordeiro, M. Ornelas, A. Ashori, S. Sheshmani, H. Norouzi, Inves-470

tigation on the surface properties of chemically modified natural fibers471

using inverse gas chromatography, Carbohydrate Polymers 87 (4) (2012)472

2367–2375.473

[36] J.-B. Donnet, R.-Y. Qin, M.-J. Wang, A new approach for estimating474

the molecular areas of linear hydrocarbons and their derivatives, Journal475

of colloid and interface science 153 (2) (1992) 572–577.476

[37] P. Mukhopadhyay, H. Schreiber, Aspects of acid-base interactions and477

use of inverse gas chromatography, Colloids and Surfaces A: Physico-478

chemical and Engineering Aspects 100 (1995) 47 – 71.479

21

Page 23 of 28

Accep

ted

Man

uscr

ipt

Figure 1: Structure of an elementary plant fibre showing the differentlayers and the orientation of the cellulose microfibrils (Adapted from [8] andreproduced with authorisation of the author).

Figure 2: IGC Surface Energy Analyser set up (Modified image suppliedby SMS).

Figure 3: Plot of P/n(P0 − P ) versus P/P0 for successive runs on kenaffibres.

Figure 4: Optical microscope images of A) BioMid R©, B) Kenaf and C)Flax fibre samples (Images obtained with an Axio Zoom.V16 microscope byCIC).

Figure 5: Individual value plots of the BET Surface Area (m2.g−1) with95% Confidence Interval (CI).

22

Page 24 of 28

Accep

ted

Man

uscr

ipt

LegrasFigure 1.pdf

Page 25 of 28

Accep

ted

Man

uscr

ipt

LegrasFigure 2.pdf

Page 26 of 28

Accep

ted

Man

uscr

ipt

LegrasFigure 3.pdf

Page 27 of 28

Accep

ted

Man

uscr

ipt

LegrasFigure 4.pdf

Page 28 of 28

Accep

ted

Man

uscr

ipt

LegrasFigure5.pdf