multi-wall carbon nanotube aqueous dispersion monitoring by using a4f-uv-mals

ORIGINAL PAPER

Multi-wall carbon nanotube aqueous dispersion monitoringby using A4F-UV-MALS

Julien Gigault & Bruno Grassl & Gaëtane Lespes

Received: 22 March 2011 /Revised: 10 September 2011 /Accepted: 11 September 2011 /Published online: 27 September 2011# Springer-Verlag 2011

Abstract In this work, the potentiality of asymmetricalflow field-flow fractionation (A4F) hyphenated to UVdetector and multi-angle light scattering (MALS) wasinvestigated for accurately determining multi-walled carbonnanotube (MWCNT) length and its corresponding disper-sion state in aqueous medium. Fractionation key parameterswere studied to obtain a method robust enough forheterogeneous sample characterization. The main A4Fconditions were 10−5 mL min−1 NH4NO3, elution flow of1 mL min−1, and cross flow of 2 mL min−1. The recoverywas found to be (94±2)%. Online MALS analysis of elutedMWCNT suspension was performed to obtain lengthdistribution. The length measurements were performed witha 4% relative standard deviation, and the length values wereshown to be in accordance with expected ones. Thecapabilities of A4F-UV-MALS to size characterize variousMWCNT samples and differentiate them according to theirmanufacturing process were evaluated by monitoring ball-milled MWCNT and MWCNT dispersions. Thecorresponding length distributions were found to be over150–650 and 150–1,156 nm, respectively. A4F-UV-MALSwas also used to evaluate MWCNT dispersion state inaqueous medium according to the surfactant concentrationand sonication energy involved in the preparation of thedispersions. More especially, the presence or absence ofaggregates, number and size of different populations, as wellas size distributions were determined. A sodium dodecylsulfate concentration of 15 to 30 mmol L−1 and a sonication

energy ranged over 20–30 kJ allow obtaining an optimalMWCNT dispersion. It is especially valuable for studyingnanomaterials and checking their manufacturing processes,size characterization being always of high importance.

Keywords Field-flow fractionation .MWCNT. Lightscattering . Length characterization . Dispersion state

Introduction

Since their discovery in 1991, carbon nanotubes (CNTs)have given rise to a great deal of interest for their uniqueproperties [1]. They are generally formed as either single-walled (SW) or multi-walled (MW) CNTs. A SWCNT is ananotube with only one wrapped graphene sheet while aMWCNT is a tube with a collection of concentric SWCNT.Carbon nanotubes may have very different lengths anddiameters. Typically, their diameters range over 1–2 and 2–20 nm for SWCNTs and MWCNTs, respectively, whilenanotube lengths are from 100 nm up to some 10 μm.SWCNTs and MWCNTs are thus different not only in termsof structure and dimensions but also in terms of flexibilityand mass. These differences also lead to different mechan-ical, thermal, electrical, and hydrodynamic properties andbehaviors, especially at the nano-scale, and differentapplications [2, 3].

CNTs are generally produced in raw samples containingimpurities, bundles, and ropes that are constituted of severalnanotubes very strongly linked some to the others. Inindustrial applications needing to have monodisperse andpure nanotubes as much as possible, manufacturers gene-rally use a post-treatment step after synthesis in order toeliminate the non-desired objects and either separate thenano-sized from micro-sized nanotubes or directly obtain

J. Gigault :B. Grassl :G. Lespes (*)Laboratoire de Chimie Analytique BioInorganique et Environnement,UMR IPREM 5254 UPPA/CNRS- Technopôle Hélioparc,Université de Pau et des Pays de l’Adour (UPPA),Av. du Président Angot,64053 Pau cedex, Francee-mail: [email protected]

Anal Bioanal Chem (2011) 401:3345–3353DOI 10.1007/s00216-011-5413-5

nano-sized nanotubes [4]. There exist different post-treatments such as laser irradiation [4, 5], high-temperature treatment, or ball-milling treatment [6, 7].The ball-milling process is performed by using a cylindricaldevice for grinding materials like CNTs. This last treatmentis particularly useful in order to purify MWCNTs. Com-plementary to this treatment, MWCNT dispersion in liquidmedium is also usually performed since a lot of applicationsneed MWCNTs in such physical state. One way to disperseMWCNTs in an aqueous surfactant solution is to sonicate itafter introduction of MWCNTs. Indeed sonication providesa mechanical energy leading to the MWCNT exfoliationfrom the original bundles, while surfactant molecules, bysorbing on the surface of MWCNT, prevent them from anyaggregation [8]. Generally, in order to obtain a stableMWCNT aqueous dispersion, anionic surfactants such assodium dodecyl sulfate (SDS) and sonication are used [9,10]. Nevertheless, there is a considerable lack of knowledgeabout the joint influence of these operating parameters (i.e.,surfactant concentration and sonication energy) on theMWCNT structure integrity in aqueous media.

Moreover, some challenges have still to be taken up suchas their size characterization and the control of theiraggregation in liquid media according to their manufactur-ing process. Thus, there is a need for analytical method ableto provide a relevant size-based characterization fromnanotube dispersion, i.e., an accurate measurement of theirsize and length distribution as well as an evaluation of theirdispersion state. To obtain such information one relevantanalytical strategy consists in hyphenating an onlinefractionation to one or several complementary detectors.Fractionation of MWCNTs according to their size can beachieved through several techniques such as size exclusionchromatography (SEC), capillary electrophoresis (CE), orfield-flow fractionation (FFF) [3, 4, 11–14]. These techni-ques have been proposed and used to characterize CNTs.Thus, it was previously seen in the case of SWCNT that thecoupling between asymmetrical flow field-flow fraction-ation (usually noted As-Fl-FFF or A4F, this last abbrevia-tion being used later on) and multi-angle light scattering(MALS) is an analytical technique of choice for their sizecharacterization in a length range up to 2,000–3,000 nm[15, 16]. Indeed, A4F intrinsically has a fractionation powerhigher than SEC, and the size range that can be fractionatedis larger than SEC one. Additionally, the fractionation isgentle with shear forces, interactions with the separationsystem or risk of analyte damage, being less important thanin CE [3, 16, 17]. According to our knowledge, there is nopaper investigating the potential of A4F-UV-MALS forMWCNT length determination associated with an analyti-cal validation approach. Nevertheless, this analytical step iscrucial since the MWCNT hydrodynamic behavior in anA4F channel is expected to be totally different from

SWCNT one, as already mentioned. Indeed, the motionand the separation of SCWNTs in a cross-flow-driven FFFhave been previously found to be governed by theorientation and diffusion coefficient of carbon nanotubes.Additionally, their separation was shown to be dependenton their diameter and flexibility, and not on their length [3,18, 19]. In case of MWCNTs, their separation occursaccording to their length, diameter, and rigidity [3]. Theirelution in a conventional-flow FFF channel was found to beeither in normal mode or steric mode, following their size[3, 13]. In the corresponding studies, the diameters of thedifferent MWCNT samples were similar, around 15–30 nm,and the length ranges over 200–5,000 and 5,000–20,000 nm, respectively. Complementarily, these workswere especially devoted to the use of FFF separation as apurification technique. No information about online sizemeasurement and size distribution was given. Considering,on one hand, that MWCNTs and SWCNTs have to be takenas different analytical objects since they have differentphysico-chemical characteristics, and, on the other hand,that information about MWCNT FFF-based characteriza-tion from the literature is scarce, the study of MWCNTs byA4F-UV-MALS appears of interest, the analytical informa-tion about it still being needed.

Accordingly, the objective of this work was to investi-gate asymmetrical flow field-flow fractionation coupled toUV and multi-angle light scattering for accurately deter-mining and monitoring MWCNT length distribution inaqueous dispersions. Because carbon nanotube character-ization is of high interest, MWCNTs coming from industrialprocesses were considered. Regarding the informationavailable in the literature and our previous study performedon 200–2,000-nm-long SWCNTs, MWCNTs taken in thepresent work were in the same length range [15]. Thecapabilities of A4F-UV-MALS were evaluated with regardto (1) the size characterization of various MWCNT samplesand their differentiation according to their manufacturingpost-treatment (i.e., ball-milled or not), (2) the determina-tion of the MWCNT dispersion state in aqueous mediumaccording to the conditions of the dispersion preparation.As SDS remains the surfactant most in use, it wasspecifically used in this work [9, 10]. Finally, suchanalytical capabilities having never been investigatedbefore; this study provides relevant information that allowsthe field of applications of A4F-UV-MALS to be expanded.

Experimental

Chemical

Ammonium nitrate (NH4NO3, 99.5%) and SDS (98.5%)were purchased from Sigma-Aldrich. Sodium hydroxide

3346 J. Gigault et al.

(NaOH, 99%) used to adjust mobile phase pH waspurchased from Merck (Merck, Darmstadt, Germany). Thewater used was Milli-Q 18 MΩ cm (Millipore System,Bedford, MA, USA). Filters used for mobile phase wereDurapore 0.1 μm from Millipore.

High-purity MWCNTs were purchased from Nanocyl(Reference NC3100). They were considered as representa-tive samples as they are characteristic of industrialproduction. The first sample used contains MWCNTs withthe following size characteristics (manufacturer data):average diameter of 9.5±0.3 nm and maximum length ina range of 1,100–1,400 nm (obtained by TEM techniques).This length range was considered as an indicative value andnamed such later on. This sample was considered as crudesince no post-treatment was applied after its synthesis. Asecond sample was used. These nanotubes were originallythe same as the first sample, but they were ball-milled aftersynthesis (manufacturer process). Both samples have purityover 95% (TGA analysis).

Sample preparation

To determine the optimal operating conditions of A4F-UV-MALS and evaluate its potentialities for MWCNT character-ization, non-filtered aqueous dispersions of crude and ball-milled MWCNTs in the presence of SDS were used. An idealMWCNT dispersion can be defined by the maximum ofindividual nanotubes dispersed that are not physically modi-fied by dispersion treatment such as surfactant added andsonication applied. In the case of nanomaterials dispersed inSDS aqueous solution, it is generally admitted that a surfactantconcentration over its critical micellar concentration (CMC) isneeded (CMC=8×10−3 mol L−1 at room temperature [20].The crude and ball-milled MWCNT aqueous dispersionswere obtained with a 20×10−3 mol L−1 SDS concentrationand by applying total sonication energy of 20 kJ(corresponding to 250 W for 1 min and 20 s). The crudeMWCNT dispersion was considered as a test sample. In orderto study the capabilities of A4F-UV-MALS for MWCNTaqueous dispersion monitoring, different dispersions werealso prepared with the MWCNT (test sample) from variousSDS concentrations tested from 0 to 40 mmol L−1. All thesedispersions were prepared by adding MWCNT powder in aSDS aqueous solution, the suspension obtained being thensonicated (see more details on the used system later on) withenergy tested from 15 to 120 kJ (corresponding to 250 W for1 to 8 min). The MWCNT final concentration was 0.01 gL−1

in all the dispersions involved in this work. During thesonication step, MWCNT samples were placed in an ice bathat 1.2±0.2 °C in order to compensate the sample warmingand keep its temperature at 20.1±0.4 °C. Temperature is animportant parameter because surfactant critical micellarconcentration is determined at room temperature (20 °C).

Instruments

The sonication energy was applied by using a 500 WVibra Cell VC 505 sonicator equipped with a probe(Sonics and Materials Inc., Newtown, USA). The asym-metrical flow field-flow fractionation (A4F) system wasan Eclipse 3 (Wyatt Technology, Dernbach, Germany).The trapezoidal-shaped channel dimensions were 26.5 cmin length, and respectively 0.6 and 2.1 cm in width. Forthe experiments, different spacers were tested at 190, 250,and 350 μm thickness. Regenerated cellulose membranewith a 10-kDa cut-off was used (Nadir membrane). Flowrates were controlled with an Agilent Technologies 1100series isocratic pump equipped with a micro-vacuumdegasser. Detection chain consisted in a variable wave-length ultraviolet/visible spectrophotometer (UV) (AgilentTechnologies 1100 series from Agilent, Tokyo, Japan)tuned at 254 nm and a MALS detector DAWN HELEOSat 658 nm (Wyatt Technology, Santa Barbara, USA). TheUV wavelength was chosen as it allows the signalobtained to mainly depend on the light absorption by thedispersed carbon nanotubes, the diffusion being thennegligible [21]. All injections were performed with anautosampler (Agilent Technologies 1100 series). Data fromMALS detector were collected and treated with Astra5.3.1.5 software (Wyatt Technology).

Fractionation conditions

The mobile phase was an aqueous solution containingNH4NO3 or SDS. Various ionic strengths were tested from0 to 10−3 mol L−1. The injection flow rate during the focusstep was adjusted at 0.2 mL min−1 and the cross-flow was3 mL min−1. These values were preliminarily adjusted inorder to avoid any aggregation phenomenon. The runsequences first contained two short and consecutive stepsof elution and focusing without injection to equilibrate thesystem and follow the signal baseline. During these steps,all impurities and SDS micelle not adsorbed onto CNTsurface were removed with the cross-flow through themembrane. The complete removal of SDS was controlledby examining the baseline and by injecting blank samples.

Different cross-flow rates were tested, from 0.3 to2.5 mL min−1 in order to obtain convenient MWCNTfractionation. At the end of the fractionation process, a rinsestep without cross-flow was applied. Each run wasreplicated six times in order to calculate experimentaluncertainties.

Analytical setup

Different equations were used to evaluate the effectivenessand quality of analyte fractionation [22, 23]. First, the

Multi-wall carbon nanotube aqueous dispersion monitoring 3347

recovery from an A4F run, i.e., the ratio between recoveredmass after analysis and injected mass, was expressed as:

R %ð Þ ¼ SS0

$ 100 ð1Þ

with S and S0 the peak areas obtained with and withoutcross-flow, respectively. Recovery was calculated from bothUV and MALS signal. With similar values being obtained,only the recovery from MALS signal is given later on.

Fractionation in A4F normal mode was based on thediffusion coefficient of the analytes and thereby on theirhydrodynamic radius, Rh. Equation 2 links Rh to theretention parameter R according to [23]:

Rh ¼kTV0 1%Rð Þ1=3

phVcw2Rð2Þ

with V0 being the void volume, Vc, the cross flow, η, theviscosity of the mobile phase, R, the retention parameterdefined as t0/tR (ratio between void and retention times), w,the channel thickness, and k, the Boltzmann constant. WhenVc is kept constant and tR sufficiently long (meaning, t0/tR<<1), Eq. 2 could be simplified in a linear relationshipbetween Rh and tR:

Rh ¼ A$ tR ð3Þ

with A assumed to be constant under constant operatingconditions. In the present case, the retention parameter wasfound to be <<1 over tR=2 min.

The MWCNT length was obtained using Rodlikeparticle fit formalism as it appears as the best way ofcalculation because MWCNTs can be likened to relativelyrigid and full tubes and not as flexible tubes as SWCNT[24]. This choice has been previously validated [25]. Thedata (MALS signals from the 18 angles) were collected bythe DAWN system and the corresponding MWCNT lengthevaluated by using this fit formalism. The procedureleading to obtaining the length from MALS measurementis described elsewhere [25]. Briefly, length can be evaluatedstarting with Eq. 4 from the Zimm approach [26]:

Rq

K»c¼ M $ P qð Þ ð4Þ

where Rθ is the Rayleigh ratio (depending on the light-scattered intensity), c the mass concentration of the analytesin the solvent, M their weight-average molar mass, K* anoptical constant, and P(θ) the theoretically derived formfactor. P(θ) can be calculated for a variety of structures. Forrodlike particle fit, P(θ) is expressed as [26]:

P qð Þ ¼Z2u

0

sintt

dt% sinuð Þ2

u2ð5Þ

with u=(2π/λ)×L×sin2(θ/2) and L is the rod (MWCNT)length. Thus, from Eqs. 4 and 5, MWCNT length can bedirectly determined by plotting light-scattering signal atdifferent angles as a function of the angle value. In otherwords, this means that, from MALS measurements ofscattered light, length is directly estimated by this particularfit formalism from Astra software, without any calculationof the usually calculated and used gyration radius.

Moreover, for rodlike particle with diameter d<<L,hydrodynamic radius (Rh) is given by [24]:

Rh ¼L

sinln L=dð Þ½ ' ð6Þ

This equation is useful to have in an experimentalevaluation of the hydrodynamic radii from lengths and thusverify if the fractionation is effective and what mode offractionation is occuring. Practically, it was used later on toestimate Rh from L obtained from MALS measurementsand Astra software calculations.

Statistical tests were performed in a 95% confidenceinterval (i.e., α=0.05). Linear curves, plotted from nexperimental data (xi,yi) (i=1 to n, n=24 in the presentstudy), were validated by considering three criteria: preci-sion, significance, and lack of bias. Precision was evaluatedby the determination coefficient, R2. Significance waschecked by a Fisher–Snedecor test, which has to be higherthan the reference value Fα,1,n-1 in order to have a significantfitting. Lack of bias was controlled by plotting the n residuesei=yi–f(xi), which have to be non-auto-correlated (i.e., nopossible fitting between them).

Results and discussion

Fractionation conditions

In optimal conditions, it is expected that MWCNT lengths(1) are distributed over the largest range because of thesample polydispersity and (2) have the maximum lengthLmax, in agreement with the indicative length value.Fractionation is effective when the retention time is suchthat eluted analytes are separated from the void peak, with arecovery close to 100% (R(%)). Mobile phase and A4Fchannel spacer thickness were investigated by performingeach run with the channel main flow (V) at 1 mL min−1 andthe cross-flow value (Vc) at 1.2 mL min−1. Then, cross-flowwas studied with optimized mobile phase composition andchannel spacer thickness.

Mobile phase was first considered. Among the differentmonovalent salts, ammonium nitrate has been found to becompatible with various detectors and to have no interfer-ence with the channel components. So, it was considered in


this work [22, 23]. Additionally, sodium dodecyl sulfatewas also tested without using other salt compounds (such asammonium nitrate) because, in the literature, authors oftenused it in FFF mobile phase.

The ionic strength (I) was tested from 0 (i.e., Milli-Qwater only) to 10−3 mol L−1. The recoveries obtained arepresented in Fig. 1. It can be noticed that the highestrecovery is obtained for an ionic strength of 10−5 mol L−1

with ammonium nitrate.SDS appears to be inappropriate in the mobile phase.

Indeed, whatever the ionic strength (over 0), the recoveryobtained with SDS never reaches 100% (76±4% maximumwith I=10−5 mol L−1) as illustrated in Fig. 1. Thecorresponding typical fractogram of MWCNTs obtained inconditions corresponding to the maximum recovery showsa signal irregular, noisy, and not well-defined from the voidpeak (Fig. 2). In comparison with the fractogram obtainedwith NH4NO3, the particular peak shape seems to indicatethat SDS as mobile phase promotes interactions betweenMWCNTs and the membrane in the A4F channel.

According to these results, a 1.10−5 mol L−1 NH4NO3

solution was chosen as mobile phase.Channel spacer thickness in A4F separation could directly

influence carbon nanotube elution rate and thus interacts withthe membrane, as usually observed [15]. Different spacerthicknesses were tested from 190 to 350 μm. For 190 μm,the MWCNT signal observed was in the void volume. For250 and 350 μm spacer thicknesses, the results are presentedin Fig. 3. The 250-μm spacer allows the MWCNT peak tobe well separated from the void volume and satisfactoryanalysis duration, while, by using the 350-μm spacer, thepeak is very wide with a slight splitting signal. Thisparticular shape can be explained by significant interactionbetween the analytes and the membrane due to a too lowelution rate [23]. Consequently, the 250-μm spacer was usedlater on.

Then, different cross-flow rates were tested over therange 0.6–2.5 mL min−1. The cross-flow rates allowing thehighest recovery were from 1.8 to 2.5 mL min−1. All thesecross-flow values give fractionation peaks well separatedfrom the void volume. Additionally, Fig. 4 presents thecorresponding length ranges obtained. Over 2.0 mL min−1,the length range becomes narrower with the maximumlength significantly decreased. The cross-flow rate of2.0 mL min−1 gives the longest range of MWCNTs (from201±8 to 1,178±21 nm). From all of these observations, across-flow of 2.0 mL min−1 appears to give a satisfactoryfractionation especially with regard to any interaction thatcould occur between the analytes and the membrane. With

Fig. 1 MWCNT (test sample) recoveries R(%) calculated from Eq. 1for different ionic strengthes and two salt natures

Fig. 2 Typical fractograms of MWCNT (test sample) obtained for twodifferent mobile phase natures with a ionic strength of 1.10−5 mol L−1

(operating conditions, Vc=1.2 mL min−1, V=1.0 mL min−1, 10 kDamembrane cut-off, and 250 μm spacer thickness; injected volume,100 μL)

Fig. 3 Typical fractograms of MWCNT (test sample) obtained(evaluated from six replicated analyses, mean RSD=5%) according todifferent spacer thicknesses (operating conditions, Vc=1.2 mL min−1, V=1.0 mL min−1, regenerated cellulose membrane with 10 kDa cut-off;injected volume, 100 μL)


this cross-flow, a length range representative of theMWCNT sample can be obtained.

Applications

Fractionation and size characterization

The MWCNT fractionation method with the A4F con-ditions previously selected (summarized in Table 1) wasthen validated.

Complementarily, the capabilities to characterize inlength and differentiate the size-distributions of MWCNTsamples obtained from various manufacturing processeswere investigated. Figure 5 represents the typical fracto-gram obtained for the crude MWCNT sample (test sample).The fractionation quality and effectiveness were evaluatedtaking into account the criteria previously defined [15], i.e.,recovery, repeatability (from six replicated analyses), andlinearity of the relationship between hydrodynamic radiusand retention times (Eq. 3).

The average recovery was 94±2%, and the averagerelative standard deviation of the retention time at themaximum of the MALS peak was 1.3%. It was previously

demonstrated for carbon nanotubes that A4F fractionationselectivity is not shape-dependent and that the sizecalibration (Rh= f(tR)), by using carbon nanotubes them-selves, is a relevant procedure [15]. In the present study,MWCNT hydrodynamic radii (calculated from Eq. 6)versus retention times could be fitted by the followinglinear relation:

Rh ¼ 16:23$ tR % 3:1 ð7Þ

This relation was statistically validated as previouslydescribed (see “Experimental” section). Because Rh wereevaluated independently from tR (see “Analytical setup”),the validation of the Eq. 7 shows that the fractionation ofthe MWCNTs occurs in the normal mode.

Additionally, the maximum length after fractionation(1,156±23 nm, RSD=4%) was found to be in accordancewith the MWCNT length directly measured by MALS (directinjection into the instrument; 1,074±48 nm). A4F-MALS alsogives a maximum length in agreement with the indicativevalue (1,100–1,400 nm). All of these results show that anaccurate length measurement can be achieved.

Complementarily, UV can be used jointly to A4F andMALS because it is known as a concentration detector and canbe used like this. Indeed, concerning the light scattering(MALS), signal depends on both concentration and size ofthe analytes. So, in the case of polydisperse samples, it is notappropriate to use light scattering signal area to quantifyanalytes because their size distributions vary considerably. TheUV peak area (without fractionation, Vc=0 mL min−1) wasthen plotted according to different MWCNT quantitiesinjected. The corresponding linear fitting (n=24 correspondingto four points replicated six times) was statistically validatedas it was found to be precise (R2=0.99), significant, and

Fig. 4 MWCNT (test sample) length ranges obtained (evaluated fromsix replicated analyses, mean RSD=7%) according to different cross-flow rates (operating conditions, V=1.0 mL min−1, regeneratedcellulose membrane with 10 kDa cut-off, 250 μm spacer thickness,injected volume, 100 μL)

Table 1 AFlFFF fractionation conditions

AFlFFF parameters Optimal conditions

Mobile phase 1.10-5 mol L−1 of NH4NO3

Elution flow 1 mL min−1

Cross flow 2.0 mL min−1

Injection flow 0.2 mL min−1

Injection volume 100 μL

Membrane nature Regenerate cellulose

Membrane cut-off 10 kDa

Spacer 250 μm

Fig. 5 Typical AFlFFF fractogram of MWCNT (test sample) withcorresponding length distribution (from six replicated analyses)obtained with selected operating conditions (see Table 1) (operatingconditions, Vc=2.0 mL min−1, V=1.0 mL min−1, regenerated cellulosemembrane with 10 kDa cut-off, 250 μm spacer thickness; injectedvolume, 100 μL)


without bias (not auto-correlated residues). These results allowfor concluding that UV peak area can be used in order toestimate the quantity of MWCNT dispersed.

As a first application of A4F-UV-MALS, both samples(test and ball-milled) were fractionated. Figure 6 presents thetypical fractograms and the corresponding length variationswith retention times obtained by A4F-UV-MALS. Ball-milling effect on the MWCNT lengths can be clearlyevidenced by examining the length variation. Indeed, themaximum length is 650 nm for ball-milled sample whereas itreaches 1,156 nm for the test sample. Length distributionsfor both samples are similar for the shortest MWCNTranging from 150 to 300 nm. Additionally, a maximum ofUV signal for ball-milled sample appears for retention timeshorter than the maximum obtained for the test sample. Thismaximum is obtained for MWCNT length around 300 nm. Itcorresponds to the maximum number of MWCNTs dis-persed. All these observations indicate that the shortestMWCNTs are not affected by ball-milling process, while asignificant breaking occurs over 300 nm MWCNT length.As a consequence, the number of nanotubes and so too thedistribution in number differ between crude and ball-milledsamples. This difference in number distribution explains theparticular variations of lengths for both samples leading tohave MWCNTs with different lengths eluting at the sametime. Indeed, the analyte distribution in the channel thicknessdepends on the number of analytes in the injected bulksample volume. These results show that the influence ofpost-treatment such as ball-milling can be visualized byA4F-UV-MALS as this method not only gives a reliablelength determination but also provides relevant informationon MWCNT distribution.

These results can be compared with those previouslyobtained for SWCNTs with the same A4F channel [15]. Theoperating conditions are found to be similar, except thecross-flow rates, which were found to be 0.9 and 2 mL min−1

for SWCNTs and MWCNTs, respectively. The elution is innormal mode for both types of carbon nanotubes. Accordingto their diameter and length (diameters of 1.3±0.3 and 9.5±0.3 nm, and length ranges over 120–1,950 and 200–1,180 nm for SWCNTs and MWCNTs, respectively), andtaking into account their flexibility/rigidity, it can beassumed that carbon nanotubes are eluted according to theirlength mainly. This could explain why the optimal cross-flow is higher for MWCNTs than for SWCNTs.

Complementarily, the selectivity S calculated from therelationship Log tR versus log L is 0.68±0.02 for SWCNTs.It is similar to the value found for MWCNTs, with a mean of0.63±0.03. However, by examining these results in moredetail, S is equal to 0.53±0.02 and 0.72±0.02 for crude andballed-milled nanotubes, respectively, showing that theselectivity negatively depends on the length. In a range oflengths over 100–2,000 nm, the elution mode and selectivityremain similar whatever the type of nanotube, the lengthseeming then to be the main key parameter of thefractionation. It is probable that the behavior in a fraction-ation channel of the single and multi-wall carbon nanotubesconsidered in this work does not appear to depend on theflexibility/rigidity of the tubes since they are relatively short.Additionally, the use of MALS complementary to a particlefit formalism adapted to a particular type of nanotube (i.e.,SWCNTs or MWCNTs) can also allow implicitly taking intoconsideration their difference in terms of rigidity and massdistribution. So, the present results show the robustness ofA4F-MALS for nanotube characterization.

Dispersion state evaluation

The A4F-UV-MALS was also applied for the optimizationand quality evaluation of MWCNT dispersion in aqueousmedia. A monitoring of MWCNT length distribution ofvarious distributions obtained by varying the operatingparameters, i.e., SDS concentration and sonication energyduring dispersion preparation was performed. In order toestimate the quality of the MWCNT dispersions, severalcriteria were taken into account. Thus, dispersion wasconsidered to be optimal when the highest maximum length(i.e., the highest Lmax) and consequently, the largest lengthrange (i.e., the largest Lmin–Lmax) was obtained with noaggregate. The presence of aggregates can be evidenced bya split signal on the MWCNT fractogram and a non-continuous variation of the length with retention time.Indeed, when MWCNTs are aggregated, the correspondingsizes increase more rapidly than those of the individuallydispersed nanotubes. In the same time, their masses

Fig. 6 Typical A4F-UV-MALS fractograms for both test (crudeMWCNT, black dot and line) and ball-milled (red dot and line) samples(no filtered dispersions), with their corresponding length distributions(operating conditions, Vc=2.0 mL min−1, V=1.0 mL min−1, regeneratedcellulose membrane with 10 kDa cut-off, 250 μm spacer thickness;injected volume, 100 μL)


increase less as MWCNT density is relatively low.Accordingly, the lengths obtained from the formalismexpressed by Eqs. 4 to 6 vary more rapidly than expectedwith retention time, leading to a discontinuity of the curveplotting L as a function of tR. The elution is also disturbedsince the hydrodynamic diameters of the aggregates arelarger than those of individually dispersed nanotubes. As aconsequence, the MALS signal splits, as already noticedelsewhere [19]. When no aggregate is present, obtaining thehighest Lmax means that all the MWCNTs are individuallydispersed, and the longest ones can be measured.

MWCNT dispersion was then studied by A4F-UV-MALSby varying SDS concentration and sonication energy. Thus,different dispersions were prepared and corresponding fracto-grams recorded. Figure 7 presents the variations of themaximum MWCNT length value (Fig. 7a) and UV peak areasignal (Fig. 7b) measured from these fractograms, accordingto both considered operating parameters. It appears that thehighest Lmax are obtained with surfactant concentration

ranged from 15 to 30 mmol L−1 and sonication energy from20 to 30 kJ (Fig. 7a). The maximum MWCNT quantitiesdispersed individually are obtained for SDS concentrationand sonication energy taken at their maximum values(Fig. 7b). These results are in agreement with the literaturewhere it is indicated that a convenient surfactant concentra-tion must be over the CMC value (8 mmol L−1 for SDS) inorder to produce efficient coating on the MWCNT surface toprevent them to any aggregation by electrostatic repulsion[27].

Nevertheless, in the range of SDS concentration where theUV peak is found maximum, MWCNT suspension is onlycomposed of short MWCNTs as it can be clearly evidenced onFig. 7a. During the sonication process, a too-important energydamages the MWCNT structure and breaks the longestnanotubes to form several smaller ones. Consequently, theMWCNT quantities grow up considerably due to MWCNTsat once individually dispersed and broken. If sonicationenergy is lower than 20 kJ, some bundles are observed.

As an illustration, Fig. 8 presents the typical fractogramand corresponding length variation obtained with a SDSconcentration of 20 mmol L−1 and a sonication energy of18 kJ. A shoulder on the peak appears with a non-continuouslength variation. As explained previously, it indicates thepresence of MWCNT aggregates eluting after 9 min retentiontime.

According to these results, an optimal MWCNT disper-sion can be obtained using a SDS concentration between 15and 30 mmol L−1 with a sonication energy from 20 to30 kJ. Complementarily, it appears that the use of UVdetector only is not sufficient to check the efficiency ofMWCNT dispersion with respect to their original lengths[27]. Nevertheless, the joint use of UV and MALS detectorshyphenated to A4F appears to be more appropriate thaneither MALS or UVonly. Indeed, MWCNT size information

Fig. 7 MWCNT (crude) maximum length (a) and UV peak area (b)obtained after A4F, according to SDS concentration and sonication energyapplied (operating conditions, Vc=2.0 mL min−1, V=1.0 mL min−1,regenerated cellulose membrane with 10 kDa cut-off, 250 μm spacerthickness; injected volume, 100 μL)

Fig. 8 Typical A4F fractogram of a selected MWCNT (crude)dispersion (20 mM SDS concentration and a sonication energy of18 kJ) with corresponding length distribution (operating conditions,Vc=2.0 mL min−1, V=1.0 mL min−1, regenerated cellulose membranewith 10 kDa cut-off, 250 μm spacer thickness; injected volume,100 μL)


obtained by UV-MALS and relative to their state of dispersionand preservation is at once qualitative and quantitative.

Conclusion

In this work, the potential of A4F-MALS was investigated foraccurately measuring MWCNTs. Through the resultsobtained, the interest to couple and use jointly A4F with UVand MALS has been also highlighted. First, A4F-MALS hasbeen showed to provide size data on ball-milled MWCNTs,which is relevant for obtaining information on the influence ofthis process on manufactured carbon nanotubes. Additionally,the capability of A4F-UV-MALS analytical system forMWCNT aqueous dispersion monitoring has been demon-strated. All these results confirm that A4F-UV-MALS is apowerful analytical tool able to evaluate MWCNT dispersionstate in aqueous medium and, more especially, the presence orabsence of aggregates, the number and size of differentpopulations, as well as size distributions. It is valuable forstudying nanomaterials and controls their manufacturing andsynthesis processes as well as their physico-chemical behaviorin a given medium, the size characterization being always ofhigh importance.

Acknowledgment We kindly thank Dr. Julien Amadou fromNanocyl for providing us carbon nanotube samples and for histechnical support.

References

1. Iijima S (1991) Helical microtubules of graphitic carbon. Nature354:56–58

2. Hennrich F, Chan C, Moore V, Rolandi M, O’Connell MJ (2006) In:O’Connell MJ (ed) Carbon nanotubes: properties and applications,CRC, Boca Raton, p. 1

3. Bai X, Li D, Du D, Zhang H, Chen L, Liang J (2004) Laserirradiation for purification of aligned carbon nanotube films.Carbon 42:2125–2127

4. Chen B, Selegue JP (2002) Separation and characterization ofsingle-walled and miltiwalled carbon nanotubes by using flowfield-flow fractionation. Anal Chem 74:4774–4780

5. Li C-H, Liu H-C, Tseng S-C, Lin Y-P, Chen S-P, Li J-Y, Wu K-H,Juang J-Y (2006) Enhancement of the field emission properties oflow-temperature-growth multi-wall carbon nanotubes by KrF excimerlaser irradiation post-treatment. Diam Relat Mater 15:2010–2014

6. Zapata-Massot C, Le Bolay N (2008) Effect of ball milling in atumbling ball mill on the properties of multi-wall carbon nanotubes.Chem Eng Process 47:1356–1362

7. Barkauskas J, Stankevičiene I, Selskis A (2010) A novelpurification method of carbon nanotubes by high-temperaturetreatment with tetrachloromethane. Sep Purif Technol 71:331–336

8. Vaisman L, Wagner HD, Marom G (2006) The role of surfactantsin dispersion of carbon nanotubes. Adv Colloid Interface Sci 128–130:37–46

9. Yurekli K, Mitchell CA, Krishnamoorti R (2004) Small-angleneutron scattering from surfactant-assisted aqueous dispersions ofcarbon nanotubes. J Am Chem Soc 126:9902–9903

10. Grossiord N, Regev O, Loos J, Meuldijk J, Koning CE (2005)Time-dependent study of the exfoliation process of carbonnanotubes in aqueous dispersions by using UV-visible spectroscopy.Anal Chem 77:5135–5139

11. Duesberg GS (1998) Separation of carbon nanotubes by sizeexclusion chromatography. Chem Commun 3:435–436

12. Doorn SK, Fields RE, Hu H, Hamon MA, Haddon RC, SelegueJP, Majidi V (2002) High resolution capillary electrophoresis ofcarbon nanotubes. J Am Chem Soc 124:3169–3174

13. Tagmatarchis N, Zattoni A, Reschiglian P, Prato M (2005)Separation and purification of functionalised water-soluble multi-walled carbon nanotubes by flow field-flow fractionation. Carbon43:1984–1989

14. Suárez B, Simonet BM, Cárdenas S, Valcárcel M (2006)Separation of carbon nanotubes in aqueous medium by capillaryelectrophoresis. J Chromatogr A 1128:282–289

15. Gigault J, Le Hécho I, Dubascoux S, Potin-Gautier M, Lespes G(2010) Single walled carbon nanotube length determination byasymmetrical-flow field-flow fractionation hyphenated to multi-anglelaser-light scattering. J Chromatogr A 1217:7891–7897

16. Lespes G, Gigault J (2011) Hyphenated analytical techniques formultidimensional characterisation of submicron particles: a review.Anal Chim Acta 692:26–41

17. López-Lorente AI, Simonet BM, Valcárcel M (2011) Electrophoreticmethods for the analysis of nanoparticles. TrAC-Trend Anal Chem30:58–71

18. Chun J, Fagan JA, Hobbie EK, Bauer BJ (2008) Size separationof single-wall carbon nanotubes by flow-field flow fractionation.Anal Chem 80:2514–2523

19. Phelan FR Jr, Bauer BJ (2007) Simulation of nanotube separationin field-flow fractionation (FFF). Chem Eng Sci 62:4620–4635

20. Paillet S, Grassl B, Desbrières J (2009) Rapid and quantitativedetermination of critical micelle concentration by automaticcontinuous mixing and static light scattering. Anal Chim Acta636:236–241

21. Jiang L, Gao L, Sun J (2003) Production of aqueous colloidaldispersions of carbon nanotubes. J Colloid Interface Sci 260:89–94

22. Dubascoux S, VonDer Kammer F, Le Hécho I, GautierMP, Lespes G(2008) Optimisation of asymmetrical flow field flow fractionation forenvironmental nanoparticles separation. J Chromatogr A 1206:160–165

23. Schimpf ME, Caldwell K, Giddings JC (2000) Field flowfractionation handbook, Wiley-IEEE, New York

24. Mansfield ML, Douglas JF (2008) Transport properties of rodlikeparticles. Macromolecules 41:5422–5432

25. Gigault J, Grassl B, Le Hécho, Lespes G (2011) Accuratedetermination of the length of carbon nanotubes using multi-anglelight scattering. Micro Chim. Acta. (in press)

26. Wyatt PJ (1993) Light scattering and the absolute characterizationof macromolecules. Anal Chim Acta 272:1–40

27. Yu J, Grossiord N, Koning CE, Loos J (2007) Controlling thedispersion of multi-wall carbon nanotubes in aqueous surfactantsolution. Carbon 45:618–623


multi-wall carbon nanotube aqueous dispersion monitoring by using a4f-uv-mals

Documents