Spontaneous Assembly of Carbon-Based Chains in Polymer Matrixesthrough Surface Charge TemplatesO. Gennari,† S. Grilli,*,† S. Coppola,† V. Pagliarulo,† V. Vespini,† G. Coppola,‡ S. Bhowmick,‡

M. A. Gioffre,‡ G. Gentile,§ V. Ambrogi,∥ P. Cerruti,§ C. Carfagna,§,∥ and P. Ferraro†

†National Institute of Optics, National Council of Research of Italy, Via Campi Flegrei 34, 80078 Pozzuoli, Italy‡Institute of Microelectronics and Microsystems, National Council of Research of Italy, Via Pietro Castellino, Napoli, Italy§Institute of Polymer Chemistry and Technology, National Council of Research of Italy, Via Campi Flegrei 34, 80078 Pozzuoli, Italy∥Department of Chemical, Materials and Production Engineering, University of Naples, Piazzale Tecchio, 80125 Naples, Italy

*S Supporting Information

ABSTRACT: Stable chains of carbon-based nanoparticleswere formed directly in polymer matrixes through anelectrode-free approach. Spontaneous surface charges weregenerated pyroelectrically onto functionalized ferroelectriccrystals, enabling the formation of electric field gradients thattriggered the dipole−dipole interactions responsible for thealignment of the particles, while embedded in the polymersolution. The phenomenon is similar to the dielectrophoreticalignment of carbon nanotubes reported in the literature.However, here the electric fields are generated spontaneouslyby a simple heat treatment that, simultaneously, aligns theparticles and provides the energy necessary for curing the hostpolymer. The result is a polymer sheet reinforced with well-aligned chains of carbon-based particles, avoiding the invasiveimplementation of appropriate electrodes and circuits. Because polymers with anisotropic features are of great interest forenhancing the thermal and/or the electrical conductivity, the electrode-free nature of this technique would improve the scalingdown and the versatility of those interconnections that find applications in many fields, such as electronics, sensors, andbiomedicine. Theoretical simulations of the interactions between the particles and the charge templates were implemented andappear in good agreement with the experimental results. The chain formation was characterized by controlling differentparameters, including surface charge configuration, particle concentration, and polymer viscosity, thus demonstrating thereliability of the technique. Moreover, micro-Raman spectroscopy and scanning electron microscopy were used for a thoroughinspection of the assembled chains.

■ INTRODUCTION

Among the wide variety of materials that are used in optics andbiomedicine, carbon-based nanoparticles (NPs) are the bestcandidates1 (e.g., graphene, single- and multiwalled carbonnanotubes SWCNT/MWCNT, and fullerenes). Their perfectstructure can enhance specific properties such as mechanical,thermal, and electrical conductivity.2 They are used in a widerange of fields including nanoelectronics and photonics,3−6

biocompatible materials,7 and biomedicine.1 Moreover, CNTsare widely used as additives to various structural materials, and itis largely demonstrated that they can modify drastically theproperties of the matrix in which they are included.8 The finalproperties of the resulting polymer composite depend stronglyon the CNT orientation.9 For this purpose, it is highly desirableto control their orientation for producing functionalized filmsand composites. In past years, it has been widely demonstratedthat although carbon NPs are difficult to disperse in liquidsuspensions, they can be oriented by various techniques.10 Thesetechniques make use of electrical fields,11,12 magnetic fields,13

shear flows,14 mechanical forces,15 virus and DNA templat-ing,16,17 dewetting,18 and blown bubble films.19 Recently,thermocapillary flows have been used for creating arrays ofpurely semiconducting SWCNTs.20 Alignment of CNTsthrough dielectrophoresis (DEP21) is very appealing because itcan be controlled easily by the application of electric fields.Dielectric NPs suspended in a liquid medium can be arranged inan orderly fashion under the influence of a nonuniform electricfield. Different DEP-based techniques have been reported in theliterature for manipulating NPs. An electric field has been usedfor aligning CNTs in glass fiber-reinforced thermosettingcomposites.22 CNT−FETs have been attained by DEP trappingof semiconductive CNTs.23 Selective deposition of SWCNTshas been proposed to align nanotubes in the form of thin films.24

Separation of CNTs and polystyrene microparticles has been

Received: September 17, 2013Revised: November 20, 2013Published: November 29, 2013

Article

pubs.acs.org/Langmuir

© 2013 American Chemical Society 15503 dx.doi.org/10.1021/la403603d | Langmuir 2013, 29, 15503−15510

achieved by DEP.25 MWCNTs have also been deposited bysurfactant-free DEP,26 and electric fields have been used foraligning functionalized MWCNTs into PMMA matrixes.27

Recently, self-assembly of NPs into chains has been achievedby DEP.28 The orientation mechanism is attributed to dipole−dipole interaction, and the attractive force between two polarizedNPs varies inversely to the fourth power of the interparticledistance. CNTs dispersed in polymers are reported to beassembled into quasi one-dimensional chainlike structures byDEP approaches.29 However, most of these mentionedtechniques often require external electrodes and voltages, andusually the chain formation varies with the electric potential andwith the electrode spacing.9,30 Recently, we published resultsconcerning the use of periodically poled ferroelectric crystals forDEP-based applications such as particle trapping,31 surfacecharge lithography,32 or liquid nanodispensing.33 Self-assemblingof liquid polymer by electro-hydro-dynamic effects was alsoachieved on periodical poled substrates.34 Nevertheless, it isimportant to note that recently periodically poled ferroelectriccrystals are emerging as smart functionalized substrates fornanoassembling particles by photochemical methods for realiz-ing plasmonic nanostructures.35 However, to the best of ourknowledge, the use of mere surface charges that arise from one-domain pyroelectric crystals (i.e., crystals without reverseddomains) has never been reported for DEP manipulation ofcarbon NPs in liquid suspensions as well as in polymer matrixes.In this paper, we present an electrode- and circuit-free method

for the alignment of carbon-based NPs along well-defined chains,embedded into polymer matrixes. Unlike the conventionaltechniques, the chains are formed through the electric fieldsarising spontaneously over the surface of a pyroelectric crystalunder thermal gradient conditions. The observed effect is veryrapid because the chains appear soon after warming the crystal,and they extend over relatively large areas with high uniformityand reproducibility. Thanks to the spontaneous nature of thedriving electric fields, the chains are formed in a very versatilefashion. In fact, the chains can be generated free from constraintsand even floating and reconfigurable, as well as anchored to somedesired geometrical configurations. In this way, the method doesnot require the pretreatment of samples, as usually done in thecase of traditional DEP techniques. In fact, lithographic and/oretching processes are required usually for fabricating theelectrodes that ensure the desired differential potential. More-over, appropriate wire systems are usually needed for assuring theelectrical connection of the electrodes with the external circuitand voltage generator. Therefore, a sort of on-demand carbonchain formation is provided by the proposed method that,compared to traditional electrode-based DEP techniques,28

would favor the development of a more compact interconnectiontechnology with a significant impact in many fields such aselectronics and sensors. We demonstrate the reliability of thetechnique first by showing that the chains are obtainable for awide variety of carbon-based NPs and second by presenting adeep characterization of the phenomenon in terms of differentprocess parameters that regulate the dynamics.

■ CHAINS OF CARBON-BASED PARTICLESBecause carbon-based building blocks are of great interest for awide variety of applications, a number of PMMA solutions (950000 molecular weight (MW) at 9% (p/V) in anisole; η = 1 cP, εr= 3.6, Microchem Corp., Newton, MA) were prepared withdifferent kinds of carbon NPs dispersed therein: MWCNTs(purity >90%, D × L 110−170 nm × 5−9 μm); SWCNTs

(purity: 50−70%, D × L 1.2−1.5 nm × 2−5 μm); graphite(platelet nanofibers, purity = 99%, D × L 50−250 nm × 0.5−5μm); fullerene (purity: 99.5%). Simple magnetic stirring at roomtemperature for 30 min mixed the NPs with a concentration of0.03% (p/V). All of the NPs were purchased from Sigma Aldrich,St. Louis, MO. The dispersion stability is shown in Figure 1a,

while Figure 1b shows the typical random state of MWCNTs inPMMA prior to DEP manipulation. The charge template,responsible for the DEP mechanism, was generated byappropriate thermal gradients applied to the z surface of z-cutlithium niobate (LiNbO3) or lithium tantalate (LiTaO3) crystals(both sides polished, 500 μm thick, from Crystal Technology,Inc.). In fact, such crystals, hereafter called “driving crystals”,exhibit pyroelectricity. Figure 1c shows the schematic view of theexperimental procedure. The metallic layer into the schemerepresents the conductive coupling edge to which the chains areanchored. Each of the suspension samples was casted onto the z-face of the driving crystal and sandwiched by a conventionalcoverslip to reduce the evaporation rate of the solvent. Thesandwich was heated onto a conventional hot plate for about 1min at 150 °C. After heating, the sample was allowed to cool atroom temperature on the stage of a conventional upright opticalmicroscope. Figure 1d−g shows the optical microscope images

Figure 1. (a) Dispersion stability of the suspensions of MWCNTs (A),SWCNTs (B), graphite (C), and fullerene (D). (b) Initial random stateof MWCNTs. (c) Schematic view of the process (not to scale) showingthe initial random state of the CNTs (left) that, after thermal treatmentonto a conventional hot plate, is converted into an ordered distributionof carbon-based chains (right); the metallic layer represents theconductive coupling edge where the chains are anchored. Opticalmicroscope view of typical chains of MWCNTs (d), SWCNTs (e),graphite (f), and fullerene (g).

Langmuir Article

dx.doi.org/10.1021/la403603d | Langmuir 2013, 29, 15503−1551015504

of the typical chains driven by the pyroelectric effect. The NPsare first randomly dispersed into the polymer matrix (see Figure1a), and, after a short thermal treatment, they appear wellchained. The images clearly show that all of the carbon allotropesinvestigated in this work chained after the application of thethermal gradient. Typically the longest chains were obtained inthe case of MWCNTs, while the shortest ones were obtained inthe case of fullerene (see Figure 1g). The phenomenon is due tothe nonuniform electric field generated onto the surface of thedriving crystal during the warming and cooling processes, whichare regulated simply by a heat exchange of the crystal underatmospheric conditions. In fact, one of the powerful features ofthe technique is the usage of very simple equipment. No externalatmospheric-controlled conditions are required, and thespontaneity of the surface charges is enough to form well-defined and stable chains of MWCNTs embedded into PMMA.Moreover, the free-constraint formation of the chains occurswithout the use of any electrode and/or external circuit. In fact,despite the conventional DEP-based techniques, the chainsassemble spontaneously along the electric field lines arising fromthe thermal gradient incurred by the crystal.According to DEP theory, the electric field of the driving

crystal induces electric dipoles to the CNTs, leading to aparticle−particle interaction. This interaction induces theparticles to move into the polymer solution with the main effectof forming particle clusters mostly in the shape of chains alignedalong the electric field lines.36,37 It is well-known that, underellipsoid approximation, the DEP force FDEP incurred by theparticle is proportional to the gradient of the square of theelectric field38−40

π ε= ∇Fabc

f E2

3Re{ } ( )DEP m CM

2

where a, b, c (a > b = c) are the half lengths of the major ellipsoidaxes, εm the permittivity of the medium, and E the electric field.The DEP force depends on the frequency ω of the applied fieldby means of the Clausius−Mossotti factor:39

ε ε

ε=

* − **fCM

p m

m

where εp* and εm* are the complex permittivities of the particle andmedium, respectively. A general complex permittivity is given byε* = ε − j(σ/ω). Figure 2 shows the typical dispersion curve ofthe particle polarizability (real part of the Clausius−Mossotti

function) in the case of MWCNTs and PMMA medium. Theconductivity and the dielectric constant of MWCNTs and ofPMMA are reported to be 104 S/m and 10−3 S/m, and 2.5 and 3,respectively (Sigma Aldrich and MicroChem datasheets). In thiscase, however, a steady electric field gradient occurs, leading to aDEP phenomenon in DC regime where the polarizability isgoverned by the conductivities:39

ωσ σ

σ→ →

−f0: Re( )CM

p m

m

Considering the conductivity values mentioned above, the CNTsare subject to a positive DEP force. A common feature of thevarious carbon allotropes is the electro-orientation along theelectric field lines.41 The oriented NPs interact with each other asdipoles,22 and, driven by the opposite charges of their extremities,NPs move closer gradually, forming a network of head-to-headconnections that leads to the aggregation of chainlike structures.The rotational force originates from the torque on the dipolemoment. In fact, the dipole moment P is proportional to theapplied electric field E as follows:42

α=P E

where α is the static polarizability tensor that, in the case of NPs,is highly anisotropic.42 The polarizability along the tube axis α|| ismuch higher than that perpendicular to the tube axis (α⊥).Therefore, for a nanotube oriented at an angle θ with respect toE, unless for θ close to 90°, the dipole moment of the nanotube isalong the tube axis with

α θ=P E cos

Due to the real part of the induced polarization, a torque isexerted on the dipole moment:

α θ θ= × =t EP E sin cos2

The resultant torque causes the CNTs to rotate and align in thedirection of the field.22,43,44 The spatial inhomogeneity of theelectric fields bring about nanotube migration toward the highfield regions, thus attracting CNTs to one another and enablingthe formation of chains. Because higher electric fields inducehigher field-induced forces on the CNTs, a corresponding higherdensity of oriented tubes is obtained.In the case of fullerene, the pentagonal sheet structure of the

lattice deviates from planar and thus prevents the delocalizationof a net charge. Therefore, the spherical fullerene NPs are lessinfluenced by the electric field distribution, compared to graphiteand CNTs. Hence, the fullerene exhibits the shorter chains asshown in Figure 1g.

■ CONTROLLING THE CHAIN FORMATIONThe self-assembling effect was investigated more fully in the caseof MWCNTs dispersed into PMMA, with particular attention tospecific features that appeared to control the chain formation: thecoupling edges; the CNT concentration; the polymer viscosity.

Coupling Edges. The results shown in Figure 1d−g refer tochains grown under constraint-free configurations; that is, theNPs aggregate under the simple action of the spontaneouselectric field generated by the thermal gradient of the drivingcrystal. However, the charge template can be configured easily tocontrol the spatial distribution of the chains. What we call“coupling edges” here can be configured appropriately toredistribute the charges and configure the chains as desired.Different configurations were investigated to demonstrate the

Figure 2. Typical dispersion curve of the particle polarizability in thecase of MWCNTs mixed with the PMMA solution. The values ofpermittivity and conductivity are reported in the main text.

Langmuir Article

dx.doi.org/10.1021/la403603d | Langmuir 2013, 29, 15503−1551015505

reliability of the procedure, i.e., electrical conductive layers,crystal grooves, domain walls, and confined liquid layers. Theconductive layers were of two different types. In oneconfiguration, aluminum foils cut by a conventional scalpel andmade adherent to the crystal surface by gentle pressure wereadopted. In a second configuration, titanium tips deposited by e-beam evaporation and patterned by a lift-off process were used.The grooves resulted from gentle scratching of the crystal surfaceby a conventional diamond tip or by spontaneous stress-relatedcracks. The domain walls were obtained by standard electric fieldpoling,32 and the confined liquid layer was deposited onto thecrystal surface by a conventional swab tip soaked by theappropriate suspension.Movies 1 and 2 in the Supporting Information show the typical

formation of chains in correspondence to the tip of an aluminumfoil (dark region) adherent to the crystal surface. These movieshave been recorded under a conventional inverted opticalmicroscope. The sandwich sample (CNT suspension squeezedbetween a glass coverslip and LN wafer) was placed on thetransparent stage of the microscope, while a hot glass slide (atabout 170 °C) was placed on the top, in contact with the LNwafer, thus inducing a sort of “heat pulse” responsible for thepyroelectric effect. At the beginning, the CNTs clearly appeardistributed randomly into the PMMA solution, and, immediatelyafter placing the hot slide (made visible by the slight turbulence),the CNTs assemble into well-defined chains that follow theelectric field lines emerging from the edges of the aluminum tip.The entire self-assembling process takes typically from about 5 to15 s. Stable chains remain embedded into the polymer matrixwhile it becomes rigid due to solvent evaporation. In fact, thegreat advantage of the technique is two-fold. On one hand, thepyroelectric effect is very quick (i.e., charges appear on the crystalsurface immediately after contact with the hot plate) in contrastto other works where tens of minutes are required for theformation of the chains. On the other hand, because one singleheat pulse simultaneously induces the appearance of the chargesand accelerates the solvent evaporation, a rigid polymer withembedded chains can be obtained easily. Hence, the operatingprinciple of the technique is very simple, and the self-assemblingof the chains is very rapid and repeatable. The only critical featureis the stability of the chains that depends significantly on theviscosity grade of the polymer solution. In fact, low viscosity isrequired at the very beginning of the self-assembling to haveCNTs sufficiently free to move under the action of the electricfield. However, viscosity must increase just after the self-assembling effect when stable and permanent chains are desired.The most stable chains were achieved typically by allowing thePMMA film to evaporate partially for about 5 min just aftercasting onto the crystal surface. Certainly, such preventiveevaporation must be moderate so that the successive heat pulsesdo not make the polymer too hard and thus prevent NPassembly.This evaporation-related method can be used for other

intriguing manipulations of the NP chains. For example, whenusing raw PMMA (without preventive evaporation), the chainsare equally well formed but they can fluctuate well under theaction of standard microfluidic instabilities, driven by pressuredifferences and/or air bubbles. Movie 3 in the SupportingInformation shows a magnified view of the typical chainformation under no preventive evaporation. The CNT chainsassemble well along the field lines, which are perpendicular to thestraight edge of the aluminum foil. After completion, they clearlyfluctuate and bend along the vertical direction that corresponds

to the pressure gradient applied to the sandwich sample.Moreover, this movie shows clearly the possibility of restoringchains partially broken by additional heat pulses, namely byinducing supplementary heat gradients after thermal stabiliza-tion.The variety of chain configurations is demonstrated by a series

of optical microscope images as shown in Figure 3, where the

CNT chains appear well anchored to the edges of conductivelayers under different geometries. Parallel chains can be obtainedby using straight edges (a, b), radial chains are generated bytiplike edges (c, d), and chains connecting two facing dots or tipsare also obtainable (e, f). The connection between the facinglayers was obtained by grounding one of them. The DEPinteraction of neighboring particles is regulated by both DEPattractive chaining force and repelling force.45 In particular, therepulsion force affects the neighboring particles with theirconnecting line perpendicular to the local field lines. Thisrepulsion force is a decreasing function of particle center-to-center distance, thus always inducing a gap between neighboringchains aligned along parallel field lines (see Figure 3a−f).The inset in panel f of Figure 3 shows the instant during the

heat pulse when the CNTs move frantically prior to attaining thefinal stable assessment. COMSOL simulations were performedfor investigating the distribution of the electric field across thepyroelectric crystal (see Supporting Information for furtherdetails). Figure 3g,h shows the results in the case of two facing

Figure 3.Microscope view of typicalMWCNT chains in PMMA formedacross different kinds of conductive layer edges: parallel straight edges oftitanium layers (a); far from a straight edge of a titanium layer (b); tip ofa titanium (c) and of an aluminum foil layer (d); dots of a titanium layer(e); facing tips of aluminum foil layers (f). (g) COMSOL simulation ofthe electric field lines corresponding to the geometry in panel c. (h)COMSOL simulation of the electric field lines corresponding to thegeometry in panel f.

Langmuir Article

dx.doi.org/10.1021/la403603d | Langmuir 2013, 29, 15503−1551015506

titanium tips and demonstrates a perfect agreement with theCNT alignment observed experimentally. The self-assemblyacross other coupling edges is shown in Figure 4, where thechains are generated across different kinds of crystal grooves (a),across a domain wall (b), and along a confined liquid layer (c).

A domain-reversed LN crystal46 was used for the secondconfiguration (Figure 4b). All of these results demonstrate that,because of the constraint-free nature of the driving charges, thistechnique could open the way to a revolutionary platform for thefabrication of high versatile, cost-effective, and highly mini-

aturized electrical networks, useful for the future generation offlexible electronics.

Concentration of Carbon Nanotubes. The tiplikeconfiguration in Figure 3c was used for investigating the behaviorof the chaining mechanism by varying the CNT concentration.Five different suspensions of MWCNTs in PMMA wereprepared with the following values of particle/weight content:0.0025%, 0.0049%, 0.0082%, 0.0098%, 0.049%. Figure 5a showsthe trend of the chain length with increasing concentration of theCNTs, up to a maximum value of about 4 mm. In fact, the higherdensity of CNTs favors the particle−particle interaction overlonger distances. The integrity of the MWCNTs after thepyroelectric treatment was examined bymicro-Raman inspectionbefore and after the chain formation (see SupportingInformation for details about the equipment). Figure 5b showsthe Raman spectra in the range 1200−1800 cm−1 of pristineCNTs (curve a), randomly dispersed in PMMA (curve b), andaligned in PMMA matrix (curve c). As an example, the result ofthe spectral deconvolution carried out on pristine CNTs was alsoreported (curves a′). Each spectrum collected on the areacontaining MWCNTs shows a band at ∼1345 cm−1 (D-band)and a complex band whose components are centered at ∼1575cm−1 (G-band) and ∼1605 cm−1 (D′-band). Moreover, thespectra of MWCNTs embedded in PMMA show the presence ofadditional bands attributed to the acrylic matrix.47 These bandsare centered at about 1448 cm−1 and 1729 cm−1. Concerning thebands characteristic of CNTs, the D band is usually attributed toamorphous or disordered carbon in the CNTs. The G bandoriginates from the in-plane tangential stretching of carbon−carbon bonds in graphenelike sheets. Finally, the D′ band,evidenced in the Raman spectra as a shoulder of the G-band athigher frequencies, is another feature induced by disorder anddefects in the CNTs. By performing the spectral deconvolution,the intensity ratio of the bands G and D (IG/ID) was calculatedand can be considered as a parameter that indicates the degree oforder in the nanotube structure. The results show that forpristine MWCNTs the value of IG/ID is about 13, thus indicatinga low amount of defective sites in their structure. A similar valueof IG/ID was recorded for the alignedMWCNTs, whereas a slightreduction of this ratio was observed for MWCNTs randomlydispersed in PMMA, for which IG/ID was∼9. This reduction canbe attributed to the insurgence of defects during the dispersion of

Figure 4.Microscope view of typicalMWCNT chains in PMMA formedacross different kinds of grooves (a), across a domain wall (b), and alonga confined strip (c).

Figure 5. (a) Variation of the chain maximum length by varying the CNT concentration; (b) Raman spectra in the range 1200−1800 cm−1 of pristineCNTs (spectrum a), randomly dispersed in PMMA (spectrum b) and aligned in the PMMA matrix (spectrum c).

Langmuir Article

dx.doi.org/10.1021/la403603d | Langmuir 2013, 29, 15503−1551015507

the CNTs within the polymer solution and the solvent casting. Inthe case of MWCNT in PMMA, either randomly dispersed oraligned, a slight upshift was observed for the G band, from 1573cm−1 (pristine MWCNT) to 1579 cm−1. This small up-shiftingphenomenon, already reported for MWCNTs embedded inPMMA, can be attributed to a certain degree of interfacialinteraction between the polymer and the CNTs that limits thenanotube−nanotube interactions.48,49Polymer Viscosity. It is well-known that the viscosity can

oppose particle motion,50 thus impacting the dynamics of chainassembly. However, in this case, the viscosity of the host polymervaries rapidly during chain formation because of the fast solventevaporation induced by the thermal treatment. Therefore, boththe early viscosity and the bulk fluid motion induced thermallyare of crucial importance for the fabrication of stable chains. Theparticle mobility has to be sufficiently high at the very beginningof the process for favoring particle interaction and has to decreaserapidly to freeze the final configuration. The initial high mobilityis ensured by the early low viscosity and by the increase of kineticenergy induced by the heat pulse. Successively, the viscosityincreases rapidly by solvent evaporation. Additional experimentswere performed for testing the technique in the case of variousdynamic viscosities η by dispersing 0.049% MWCNTs indifferent matrixes: raw PDMS (Sylgard 184 Silicone ElastomerKit, Dow Corning; η = 3900 cP, εp = 2.65 from datasheet);PDMS diluted (mix ratio 1:1) in toluene (purity ≥99.9%, SigmaAldrich) (η = 0.07 Pa s) and in dichloromethane (DCM, purity≥99.9%, Sigma Aldrich) (η = 0.12 Pa s); styrene (purity ≥99%,Sigma Aldrich); glycerine (purity ≥99.5%, Carlo Erba Reagents,Italy) (η = 1.57 Pa s). The viscosities were evaluated with therheological measurements reported in the Supporting Informa-tion.No chains were observed in the case of raw PDMS because of

the relatively high viscosity (see the inset in Figure 6a) thatcaused the friction to exceed the DEP force of the pyroelectricfield, thus preventing the CNTs to move and redistribute alongthe field lines. The significant role of viscosity in chain formationwas demonstrated by the results obtained in the case of PDMSdiluted in toluene (Figure 6a) and DCM (Figure 6b), where

relatively long CNT chains were generated, similarly to PMMA-based experiments. Figure 6c,d shows the chains obtained in thecase of styrene (purity ≥99%, Sigma Aldrich) and glycerin(purity ≥99.5%, Carlo Erba Reagents, Italy), where the slightlyhigher viscosities made the process more difficult, leading toshorter chains.Figure 7a,b shows the optical microscope images of CNT

chains embedded permanently into a PMMA sheet, after

complete solvent evaporation. The alignment of the CNTs inthe polymer matrix was investigated further by scanning electronmicroscopy (SEM), as shown in Figure 7c,d under two differentmagnifications. The SEM analysis was performed on a FEIQuanta 200 FEG SEM (Eindhoven, The Netherlands) at 10 kVacceleration voltage and with a secondary electron detector.Before the analysis, the samples were mounted onto SEM stubsby means of carbon adhesive disks and sputter-coated with agold−palladium alloy. The images show the typical head-to-headalignment.

■ THEORETICAL INTERPRETATIONThe magnified view of chain formation in movie 3 in SupportingInformation gives insight into what happens during the self-assembling effect. The development of chains can be brokendown into five fundamental stages: comb; orientation; early stagechains; addition; stretching and tightening. Images of samplecorresponding to these stages are shown in Figure 7e. First, theCNTs close to the conductive edge are strongly attracted andform a stable comblike structure with teeth along the electric field

Figure 6. Microscope images of MWCNT chains dispersed intomatrixes with different viscosities: PDMS diluted in toluene (a), PDMSdiluted in DCM (b), styrene (c), and glycerin (d). The inset in panel ashows the lack of chains in the case of undiluted PDMS.

Figure 7. (a, b) Microscope images of MWCNT chains embeddedpermanently into a PMMA sheet after complete solvent evaporation. (c,d) SEM images of typical CNT chains. (e) Sample images of the typicalstages leading to the formation of the CNT chains. The circles highlightthe regions where the CNTs assemble according to the correspondingstage.

Langmuir Article

dx.doi.org/10.1021/la403603d | Langmuir 2013, 29, 15503−1551015508

lines. The remaining CNTs, away from the coupling edge, moveturbulently under the action of thermal convection, electricalfield, and viscous drag. The DEP force polarizes and orients theparticles along the electric field lines. In this way, widespreadearly stage chains form through the particle−particle interaction.Because the DEP force is proportional to the square of theelectric field strength, the early stage chains appear denser in theregions closer to the aluminum edge, where the electric field isstronger. Moreover, the field E represents the field intensity atthe particle center, and the center of individual tubes is muchcloser to the conductive edges than the center of early stagechains. Therefore, the individual tubes feel stronger DEP forceover the viscous drag and, consequently, move faster. Indeed,individual CNTs, rather than in short chains, move most rapidlytoward broken connections in subchains and provide the mostcommonly observed means of restoring the broken connections.Additional unchained particles progressively interact with theearly stage chains that operate as catalysts for the formation oflonger chains. At the end, turbulence decreases significantly dueto thermal steadying and long chains stretch and tighten eachother to the comb, thus leading to the stable final configuration.Other interesting considerations can bemade in the case of the

sharp edges in movie 1 in Supporting Information, where theformation stages exhibit an additional amazing effect. The earlystage chains appear to form first across the two corners of thelayer where the electric field gradient is in fact stronger and alsobeing able to break relatively large bundles of CNTs. Even thebranching effect is clearly visible across the corners, againaccording to the field lines.

■ CONCLUSIONSWe developed here an electrode-free DEP approach forpromoting a spontaneous formation of carbon-based chainsembedded into polymer matrixes. The chaining effect isremarkably reliable and exhibits high repeatability. CNT chainsup to 4 mm long were produced, also embedded into polymerlayers. The usage of spontaneous pyroelectric charges providesthe approach with unprecedented simplicity and compactness.These features would be of great interest for developing highlyintegrated circuits in those applications where the scaling downof the interconnection technology is highly desired, includingelectronic devices, sensors, and thermally conductive polymers,as well as biomedicine, where CNTs may operate as nanovectorsfor drug delivery purposes.

■ ASSOCIATED CONTENT*S Supporting InformationRheological measurements. Simulations of the pyroelectric fields.Micro-Raman equipment. Movies of NPs during the sponta-neous assembly. This material is available free of charge via theInternet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*E-mail: simonetta.grilli@ino.it.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe authors acknowledge the Italian Ministry of Research forfinancial support, under the “Futuro in Ricerca 2010”

Programme (Protocol RBFR10FKZH) and EFOR-CABIRCNR project.

■ REFERENCES(1) Ryoo, S.-R.; Kim, Y.-K.; Kim, M.-H.; Min, D.-H. Behaviors of NIH-3T3 fibroblasts on graphene/carbon nanotubes: Proliferation, focaladhesion, and gene transfection studies. ACS Nano 2010, 4, 6587−98.(2) Sundaray, B.; Subramanian, V. Electrical conductivity of a singleelectrospun fiber of poly (methyl methacrylate) and multiwalled carbonnanotube nanocomposite. Appl. Phys. Lett. 2006, 88, 143114.(3) Huang, Y.; Duan, X.; Wei, Q.; Lieber, C. M. Directed assembly ofone-dimensional nanostructures into functional networks. Science 2001,291, 630−3.(4) Burg, B. R.; Poulikakos, D. Large-scale integration of single-walledcarbon nanotubes and graphene into sensors and devices usingdielectrophoresis: A review. J. Mater. Res. 2011, 26, 1561−1571.(5) Lipomi, D. J.; et al. Skin-like pressure and strain sensors based ontransparent elastic films of carbon nanotubes.Nat. Nanotechnol. 2011, 6,788−92.(6) Yamada, T.; et al. A stretchable carbon nanotube strain sensor forhuman-motion detection. Nat. Nanotechnol. 2011, 6, 296−301.(7) Pinto, A. M.; et al. Biocompatibility of poly(lactic acid) withincorporated graphene-based materials. Colloids Surf., B 2013, 104,229−38.(8) Monti, M.; Natali, M.; Petrucci, R.; Kenny, J. M.; Torre, L. Impactdamage sensing in glass fiber reinforced composites based on carbonnanotubes by electrical resistance measurements. J. Appl. Polym. Sci.2011, 122, 2829−2836.(9) Duchamp,M.; et al. Controlled positioning of carbon nanotubes bydielectrophoresis: Insights into the solvent and substrate role. ACSNano2010, 4, 279−84.(10) Yan, Y.; Chan-Park, M. B.; Zhang, Q. Advances in carbon-nanotube assembly. Small 2007, 3, 24−42.(11) Chen, X. Q.; Saito, T.; Yamada, H.; Matsushige, K. Aligningsingle-wall carbon nanotubes with an alternating-current electric field.Appl. Phys. Lett. 2001, 78, 3714.(12) Bubke, K.; Gnewuch, H.; Hempstead, M.; Hammer, J.; Green, M.L. H. Optical anisotropy of dispersed carbon nanotubes induced by anelectric field. Appl. Phys. Lett. 1997, 71, 1906.(13) Hone, J.; et al. Electrical and thermal transport properties ofmagnetically aligned single wall carbon nanotube films. Appl. Phys. Lett.2000, 77, 666.(14) Hobbie, E. K.; et al. Optical measurements of structure andorientation in sheared carbon-nanotube suspensions. Rev. Sci. Instrum.2003, 74, 1244.(15) Lim, J. K.; et al. Alignment strategies for the assembly ofnanochains with submicron diameters. Small 2010, 6, 1736−40.(16) Dang, X.; et al. Virus-templated self-assembled single-walledcarbon nanotubes for highly efficient electron collection in photovoltaicdevices. Nat. Nanotechnol. 2011, 6, 377−84.(17) Maune, H. T.; et al. Self-assembly of carbon nanotubes into two-dimensional geometries using DNA origami templates. Nat. Nano-technol. 2010, 5, 61−6.(18) Huang, J.; Kim, F.; Tao, A. R.; Connor, S.; Yang, P. Spontaneousformation of nanoparticle stripe patterns through dewetting.Nat. Mater.2005, 4, 896−900.(19) Yu, G.; Cao, A.; Lieber, C. M. Large-area blown bubble films ofaligned nanochains and carbon nanotubes. Nat. Nanotechnol. 2007, 2,372−7.(20) Jin, S. H.; et al. Using nanoscale thermocapillary flows to createarrays of purely semiconducting single-walled carbon nanotubes. Nat.Nanotechnol. 2013, 8, 347−55.(21) Mokry, P.; Marvan, M.; Fousek, J. Patterning of dielectricnanoparticles using dielectrophoretic forces generated by ferroelectricpolydomain films. J. Appl. Phys. 2010, 107, 094104.(22) Monti, M.; Natali, M.; Torre, L.; Kenny, J. M. The alignment ofsingle walled carbon nanotubes in an epoxy resin by applying a DCelectric field. Carbon 2012, 50, 2453−2464.

Langmuir Article

dx.doi.org/10.1021/la403603d | Langmuir 2013, 29, 15503−1551015509

(23) Cicoria, R.; Sun, Y. Dielectrophoretically trapping semiconductivecarbon nanotube networks. Nanotechnology 2008, 19, 485303.(24) Li, P.; Xue,W. Selective deposition and alignment of single-walledcarbon nanotubes assisted by dielectrophoresis: From thin films toindividual nanotubes. Nanoscale Res. Lett. 2010, 5, 1072−8.(25) Zhang, C.; et al. Dielectrophoretic separation of carbon nanotubesand polystyrenemicroparticles.Microfluid. Nanofluid. 2009, 7, 633−645.(26) Moscatello, J.; et al. Surfactant-free dielectrophoretic depositionof multi-walled carbon nanotubes with tunable deposition density.Carbon 2010, 48, 3559−3569.(27) Chen, M.; et al. Alignment and dispersion of functionalizedcarbon nanotubes in polymer composites induced by an electric field.Authors’ reply. Carbon 2008, 46, 706−710.(28) Slopek, R. P.; Gilchrist, J. F. Self-assembly of chains in acrylatemonomer via nanoparticle dielectrophoresis. J. Phys. D: Appl. Phys.2010, 43, 045402.(29) Dimaki, M.; Bøggild, P. Frequency dependence of the structureand electrical behaviour of carbon nanotube networks assembled bydielectrophoresis. Nanotechnology 2005, 16, 759−763.(30) Liu, H.; Takagi, D.; Chiashi, S.; Homma, Y. Transfer andalignment of random single-walled carbon nanotube films by contactprinting. ACS Nano 2010, 4, 933−8.(31) Grilli, S.; Ferraro, P. Dielectrophoretic trapping of suspendedparticles by selective pyroelectric effect in lithium niobate crystals. Appl.Phys. Lett. 2008, 92, 232902.(32) Grilli, S.; Vespini, V.; Ferraro, P. Surface-charge lithography fordirect PDMS micro-patterning. Langmuir 2008, 24, 13262−5.(33) Ferraro, P.; et al. Dispensing nano-pico droplets and liquidpatterning by pyroelectrohydrodynamic shooting. Nat. Nanotechnol.2010, 5, 429−435.(34) Xi, X.; Zhao, D.; Tong, F.; Cao, T. The self-assembly andpatterning of thin polymer films on pyroelectric substrates driven byelectrohydrodynamic instability. Soft Matter 2012, 8, 298.(35) Yraola, E.; Molina, P.; Plaza, J. L.; Ramírez, M. O.; Bausa, L. E.Spontaneous emission and nonlinear response enhancement by silvernanoparticles in a Nd3+-doped periodically poled LiNbO3 laser crystal.Adv. Mater. 2013, 25, 910−915.(36) Hermanson, K. D.; Lumsdon, S. O.; Williams, J. P.; Kaler, E. W.;Velev, O. D. Dielectrophoretic assembly of electrically functionalmicrochains from nanoparticle suspensions. Science 2001, 294, 1082−6.(37) Nicotra, O. E.; La Magna, A.; Coffa, S. Particle-chain formation ina dc dielectrophoretic trap; a reaction-diffusion approach. Appl. Phys.Lett. 2009, 95, 073702.(38) Jones, T. B. Electromechanics of Particles; Cambridge UniversityPress: New York, 2005; Ch. 2.(39) Morgan, H.; Green, N. G. Dielectrophoretic manipulation of rod-shaped viral particles. J. Electrostatics 1997, 42, 279−293.(40) Burg, B. R.; Bianco, V.; Schneider, J.; Poulikakos, D. Electrokineticframework of dielectrophoretic deposition devices. J. Appl. Phys. 2010,107, 124308-JAP107−124308−11.(41) Gimsa, J. A comprehensive approach to electro-orientation,electrodeformation, dielectrophoresis, and electrorotation of ellipsoidalparticles and biological cells. Bioelectrochemistry 2001, 54, 23−31.(42) Benedict, L. X.; Louie, S. G.; Cohen, M. L. Static polarizabilities ofsingle-wall carbon nanotubes. Phys. Rev. B 1995, 52, 8541−8549.(43) Liu, Y.; Chung, J.-H.; Liu, W. K.; Ruoff, R. S. Dielectrophoreticassembly of nanowires. J. Phys. Chem. B 2006, 110, 14098−14106.(44) Hsu, H.-Y.; Sharma, N.; Ruoff, R. S.; Patankar, N. A. Electro-orientation in particle light valves. Nanotechnology 2005, 16, 312−319.(45) Ding, H.; Liu, W.; Shao, J.; Ding, Y.; Zhang, L.; Niu, J. Influence ofinduced-charge electrokinetic phenomena on the dielectrophoreticassembly of gold nanoparticles in a conductive-island-based micro-electrode system. Langmuir 2013, 29, 12093−12103.(46) Grilli, S.; Paturzo, M.; Miccio, L.; Ferraro, P. In situ investigationof periodic poling in congruent LiNbO3 by quantitative interferencemicroscopy. Meas. Sci. Technol. 2008, 19, 074008.(47)Mueller, A.; Vigolo, B.; McRae, E.; Soldatov, A. V. Raman study ofinhomogeneities in carbon nanotube distribution in CNT−PMMAcomposites. Phys. Status Solidi B 2010, 247, 2810−13.

(48) Bokobza, L.; Zhang, J. Raman spectroscopic characterization ofmultiwall carbon nanotubes and of composites. eXPRESS Polym. Lett.2012, 6, 601−8.(49) McClory, C.; McNally, T.; Baxendale, M.; Potschke, P.; Blau, W.;Ruether, M. Electrical and rheological percolation of PMMA/MWCNTnanocomposites as a function of CNT geometry and functionality. Eur.Polym. J. 2010, 46, 854−68.(50) Li, Z.; Wang, H. Drag force, diffusion coefficient, and electricmobility of small particles. I. Theory applicable to the free-moleculeregime. Phys. Rev. B 2003, 68, 061206−9.

Langmuir Article

dx.doi.org/10.1021/la403603d | Langmuir 2013, 29, 15503−1551015510