Friction mechanism of individual multilayerednanoparticlesOfer Teveta, Palle Von-Huthb, Ronit Popovitz-Birob, Rita Rosentsveiga, H. Daniel Wagnera, and Reshef Tennea,1

aDepartment of Materials and Interfaces, Weizmann Institute of Science, Rehovot 76100, Israel; and bChemical Research Support, Weizmann Institute ofScience, Rehovot 76100, Israel

Edited by Allan H. MacDonald, University of Texas at Austin, Austin, TX, and approved September 26, 2011 (received for review April 26, 2011)

Inorganic nanoparticles of layered [two-dimensional (2D)] com-pounds with hollow polyhedral structure, known as fullerene-like nanoparticles (IF), were found to have excellent lubricatingproperties. This behavior can be explained by superposition ofthree main mechanisms: rolling, sliding, and exfoliation-materialtransfer (third body). In order to elucidate the tribological mechan-ism of individual nanoparticles in different regimes, in situ axialnanocompression and shearing forces were applied to individualnanoparticles using a high resolution scanning electron micro-scope. Gold nanoparticles deposited onto the IF nanoparticlessurface served as markers, delineating the motion of individual IFnanoparticle. It can be concluded from these experiments thatrolling is an important lubrication mechanism for IF-WS2 in the re-latively low range of normal stress (0.96±0.38 GPa). Sliding isshown to be relevant under slightly higher normal stress, wherethe spacing between the two mating surfaces does not permit freerolling of the nanoparticles. Exfoliation of the IF nanoparticlesbecomes the dominant mechanism at the high end of normalstress; above 1.2 GPa and (slow) shear; i.e., boundary lubricationconditions. It is argued that the modus operandi of the nanoparti-cles depends on their degree of crystallinity (defects); sizes; shape,and their mechanical characteristics. This study suggests thatthe rolling mechanism, which leads to low friction and wear, couldbe attained by improving the sphericity of the IF nanoparticle,the dispersion (deagglomeration) of the nanoparticles, and thesmoothness of the mating surfaces.

fullerene-like nanoparticles ∣ tribology ∣ rolling friction ∣ nanotribology

Friction causes wear and energy dissipation, and is responsible(directly or indirectly) for about one-third of the world’s

energy resource consumption (1,2). Therefore, friction and wearreduction are key factors in any future energy conservation plan.Similar in many ways to carbon fullerenes and nanotubes; hollowpolyhedral nanoparticles (NP) of the inorganic compounds WS2,MoS2, were first reported in 1992 (3). These NP, known as inor-ganic fullerene-like (IF) structures and inorganic nanotubes, wereshown to reduce friction and wear either as a pure solid lubricant;as additives to various fluid lubricants, or as part of self-lubricat-ing coatings (4, 5). The seamless-hollow nanostructures wereshown to be inherent to the layered structure of the material.The IF nanoparticles are common to layered metal dichalcogen-ides, MX2 (M ¼ W , Mo, Ti, Nb, Hf; X ¼ S, Se) compounds aswell as to numerous other inorganic compounds with layeredtwo-dimensional (2D) structure. Each layer in the MX2 materialconsists of three covalently bonded sheets with the metal atomssandwiched between two chalcogenide sheets. In WS2 (MoS2)each tungsten (molybdenum) atom is bonded to six sulfur atomsin a trigonal prismatic coordination mode. Weak van der Waalsforces are responsible for stacking the planar X-M-X layers to-gether. Both graphite and WS2 (MoS2) structures belong to thesame space group, P63/mmc (6).

Fig. 1A shows the transmission electron microscope (TEM)image of a hollow multiwall and polyhedral IF-WS2 nanoparticle.This nanoparticle is highly faceted, and its dimensions are 78and 82 nm along the two main axes. The particle contains 32–36

closed WS2 layers. The nominal distance between each pair oflayers is 6.20 Å (center to center). The diameter of the hollowvoid in the center of the multilayered WS2 nanoparticle is abouthalf the overall diameter of the nanoparticle. Fig. 1B shows theTEM image of a typical IF-MoS2 nanoparticle. This nanoparticleis distinguishable from the IF-WS2 nanoparticle by a number ofsalient features. In contrast to the IF-WS2, which are quasi-sphe-rical in shape, the synthesized IF-MoS2 NP possess higher crystal-line order but their shape is quashed. The numerous layers ofthe IF-MoS2 NP are evenly folded and their hollow core is tiny(ca. 1 nm in size). Therefore, the IF-MoS2 NP are likely to showless rolling and be more prone to sliding than the IF-WS2 NP, aphenomena which is confirmed by the present experiments.

The mechanical properties of the nanotubes and the individualWS2 NP have been studied quite extensively (7,8). Frictioncoefficients as low as 0.03 were measured in a macroscopic fric-tion experiment conducted with IF NP added to oil under severecontact conditions (9). Very efficient lubrication was also ob-tained for dry metallic films impregnated with IF NP underboundary lubrication conditions (10). Nonetheless, the frictioncoefficients were found to be generally higher (ca. 0.1) in dry testsas compared to oil-lubricated interfaces (0.06 and below).

Three main mechanisms which lead to reduced friction andwear are discussed in the literature (11):

1. Rolling—the IF nanoparticle acts as a ball bearing betweenthe mating surfaces (11), as shown in Fig. 2A.

2. Sliding—Given the low surface energy of its basal (001) planeand its robustness, the IF nanoparticle acts as a separatorproviding low friction and facile shearing between the matingsurfaces (12), as shown in Fig. 2B.

3. Exfoliation and transfer of films (third body)—exfoliated layersfrom the IF NP are deposited on the asperities of the matingsurfaces providing easy shearing (13), as shown in Fig. 2C.

Exfoliation and transfer of WS2 nanosheets (designated thirdbody) (14) were considered to be the predominant mechanismfor reducing both friction and wear in the case of IF NP. It wasshown that under large mechanical loads and shear forces, theNP are deformed and compressed. This structural deformationgives rise to dislocation and dislodging of WS2 nanosheets fromthe NP surface. The external nanosheets of the IF NP, a fewmonolayers thick each (15), are gradually transferred onto thesubstrate, thus reducing the friction between the two matingsurfaces. Furthermore, this process reduces the local heating,and hence the plastic deformation of the underlying metal surfaceand its oxidation, thereby slowing down the wear (16). Extensiveexfoliation eventually damages the IF NP and causes degradation

Author contributions: O.T., H.D.W., and R.T. designed research; O.T., P.v.H., and R.P.-B.performed research; R.R. contributed new reagents/analytic tools; O.T., H.D.W., and R.T.analyzed data; and O.T., H.D.W., and R.T. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: reshef.tenne@weizmann.ac.il.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1106553108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1106553108 PNAS ∣ December 13, 2011 ∣ vol. 108 ∣ no. 50 ∣ 19901–19906

CHEM

ISTR

Y

of the tribological behavior. However, oxidation of the sulfide NPtransforms it into a tungsten oxide film, which by itself is a goodlubricant (17).

During the sliding motion the IF NP preserves its shape undernormal and shear stress without rolling or exfoliation. Sliding isregarded as a very low friction mechanism when it occurs betweenweakly interacting, smooth surfaces (18). Nonetheless if the slid-ing particles are made of a hard material, degradation of the mat-ing surfaces by abrasion, plowing, or indentation is inevitable(19). In this case the friction and wear may increase dramaticallywith time.

When the rolling mechanism is effective, the NP act as nano-scopic ball bearings between the contact surfaces and provideextremely low friction coefficients (0.01 and below). It was pos-

tulated that the quasi-spherical IF NP are ideally suited to pro-vide this kind of friction mechanism (20). However, no directevidence to that effect has been established, so far. In fact, inlatter works, the rolling mechanism was debated and consideredto be of minor relevance. Recently, in situ uniaxial nanocompres-sion tests of individual IF-WS2 (MoS2) NP using TEM (21) andhigh resolution scanning electron microscope (HRSEM) (8),were published. These works renewed the discussion on the sig-nificance of rolling friction on the tribological behavior of IF NP.

In the present study in situ HRSEM and a nanomechanicalmanipulator were used to apply normal and shear stresses, com-bined, on an individual IF. Gold NP (GNP) were deposited on thelarger IF NP. Earlier works have indicated that the well adheringGNP tend to decorate the sharp sites of the faceted surface ofthe IF-WS2 NP (22). The GNP were used as markers for measur-ing the rolling, sliding, and exfoliation of the IF nanoparticlewithin the HRSEM. This study clarifies dependence of the fric-tion mechanisms: exfoliation, rolling, and sliding on the normal(uniaxial) stress.

Results and DiscussionIn the present study the tribological mechanism of individualinorganic IF NP is evaluated under different stress regimes byapplying uniaxial (normal) stress followed by shear. The methodused to evaluate the normal stress applied by the axial nano-compression test on individual IF NP is described in Materialsand Methods. The failure stress (strength) of individual IFunder uniaxial compression was found to be 1.9� 0.8 GPa and2.5� 1 GPa for WS2 and MoS2 respectively, see Table 1. Thelower values of strength of the IF-WS2 NP can be attributed totheir highly faceted shape and the presence of defects in the NP.In some of the experiments the IF broke under the load evenbefore the shearing movement could be imposed, indicatingtheir low strength. Finite element analysis (23) shows that apply-ing uniaxial load on the IF nanoparticle leads to stress concen-tration at the inner corner which eventually causes failure.

The exfoliation mechanism was shown to be the primary expla-nation for the reduction of friction, probably due to the fact thatcompressed NP and nanosheets were present in the majority ofthe tribology experiments (16, 24). In contrast to the exfoliationmechanism, which can be analyzed postmortem, rolling and slid-ing friction can only be observed during operation and hence aremore difficult to investigate in engineering-type tribological tests.

The rolling mechanism was debated and considered to be oflesser significance due to the following: the WS2 NP are not fullyspherical; they are polyhedral with corners. Moreover, the NP tendto agglomerate, and the gap between the two mating surfaces inthe contact area is usually smaller than the size of the agglomerate(13). The metal finish is also not perfect, with defects and irregu-larities occurring on the surface. Such surface irregularities impedethe rolling friction mechanism. In particular, the first nanoparticleto be clogged by the asperity blocks the free motion of the NP be-hind; i.e., jamming occurs. The impededNPare prone to combinedloading and shearing forces, and are fractured or exfoliated (24).

In spite of these arguments, the very recent in situ TEM (21)and the current in situ HRSEM (8) experiments show that rollingis not only plausible, but a valid mechanism for reduction of thefriction by the IF NP. This revives a long-standing dispute con-cerning the role of rolling friction in the tribological action of

Fig. 1. TEM image of a typical hollow multilayered IF-WS2 nanoparticle(A) and IF-MoS2 NP (B). The line profile, of the boxed area, gives an interlayerspacing of 6.20 Å.

Fig. 2. The three main friction mechanisms of multilayered IF NP whichare discussed in the literature: rolling (A), sliding (B), and exfoliation (C). Thebottom surface is stationary while the upper surface is moving to the left. Thered mark is a point of reference- i.e., gold nanoparticle in the present experi-ments.

Table 1. The average values and the standad deviations of thenormal stress in the nanocompression test (fracture stress) and inthe rolling, sliding and exfoliation experiments

rolling Sliding Exfoliation Fracture

Average STD Average STD Average STD Average STD

IF-WS2 0.96 0.38 1.65 0.54 1.82 0.59 1.94 0.79IF-MoS2 0.50 0.19 1.89 0.78 2.54 0.98

19902 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1106553108 Tevet et al.

these NP. The rolling mechanism was also challenged by the con-jecture that the NP are sliding and not rolling during the tribo-logical test. It is well established that the coefficient realizedduring rolling friction is generally 102 to 103 times lower than thatof sliding friction for corresponding materials (25). This differ-ence was in fact the motivation for a deeper investigation of thefriction mechanism of the IF NP.

In this study GNP with 16� 4 nm diameter were depositedon the larger IF NP, 100� 20 nm. The first attempt to decoratethe IF-WS2 surfaces with GNP was carried out using GNP in to-luene solution. However, dispersion of the IF NP decorated withGNP from toluene solution was found to be ineffective and analternative approach; i.e., deposition from aqueous solution wassubsequently attempted. Preparation of colloidal gold NP in aqu-eous solution was based on the classical Turkevich-Frens method(26,27), using trisodium citrate as a reducing agent in aqueoussolution. The attachment of the GNP to the IF NP was charac-terized by TEM and high resolution TEM (HRTEM). Fig. 3Ashows the TEM image of an IF-WS2 nanoparticle decorated witha gold nanoparticle from the aqueous solution. Fig. 3B showsHRTEM images of a spherical gold nanoparticle attached to theIF-WS2 NP surface. In Fig. 3, a few outer layers of the IF-WS2and the gold nanocrystal lattice planes are clearly visible.

The nanosize dimension and the quasi-spherical shape of the IFnanoparticle presents a challenge for the design of experimentsintended to prove the friction mechanism. In order to prove therolling mechanism, GNP were used as a marker for the relativeposition of the individual IF nanoparticle with respect to the sub-strate and the atomic force microscope (AFM) tip. In particular,the GNP served as a reference point clearly deciphering betweensliding and rolling movements during the shearing experiments inthe HRSEM.

Here, an AFM probe with plateau tip was used to apply a com-bined pressure (z-axis) and shearing stress (x-axis) on individualIF NP decorated with GNP placed on a Si wafer surface. Thiscombination makes it possible to distinguish between the rolling,

sliding, and exfoliation mechanisms of the IF-WS2 NP under dif-ferent loads and shearing forces.

Fig. 4 shows the schematics of a set-up used for applying axialload and shear movement causing rolling. The same set-up wasalso used for demonstrating the exfoliation and sliding of the IFNP. The rolling mechanism of IF-WS2 NP decorated with GNPis shown in Fig. 5, Movie S1. The arrow points out the GNPattached to the rolling IF-WS2 nanoparticle. The blue dot pointsout the starting point on the AFM tip. In this experiment theuniaxial compression stress was 0.6±0.25 GPa before applyingthe shear stress. Rolling was not observed with inferior loads, pos-sibly because the IF nanoparticle was not sufficiently tightened tothe two mating surfaces.

Sliding friction was also observed in the experiments. Fig. 6shows a series of images of a sliding IF-WS2 NP. Here sliding ofthe IF nanoparticle was observed along a distance of 170 nm dur-ing a shearing experiment. The asymmetric nanoparticle waslocked in the gap of the oblique-angle constriction between theflat AFM tip and the Si surface. The nanoparticle was locked withits acute angle vertex touching the flat tip surface. This situationcaused the normal stress to further increase, and the IF NP wasforced to slide rather than roll. The effective stress was neverthe-less inferior to 1.2 GPa, which is the onset stress for exfoliation.

The exfoliation mechanism can be seen in Fig. 7 A and B,which show the IF-WS2 nanoparticle with GNP attached beforeuniaxial compression. Fig. 7 C and D show IF-WS2 nanoparticledecorated with GNP during the uniaxial compression. Fig. 7 E

Fig. 3. TEM image of an IF-WS2 nanoparticle decorated by a gold nanopar-ticle from the aqueous solution (A); HRTEM images of gold nanoparticleattached to an IF-WS2 nanoparticle (B).

Fig. 4. Schematics showing how the flat AFM tip rolls an IF nanoparticle

Fig. 5. Demonstration of the rolling friction mechanism. (A) to (F) show therolling of an IF-WS2 nanoparticle decorated with GNP. The arrow points outthe GNP attached to the rolling IF-WS2 nanoparticle. The blue dot marks thestarting point on the AFM tip.

Tevet et al. PNAS ∣ December 13, 2011 ∣ vol. 108 ∣ no. 50 ∣ 19903

CHEM

ISTR

Y

and F show the same IF-WS2 nanoparticle after shear movementand exfoliation. In this experiment the uniaxial compressionstress was about 1.2 GPa before applying the shear stress.

Fig. 8 summarizes the data for the fracture stress and the stressvalues recorded for the different friction mechanisms of IF-WS2and IF-MoS2 NP. Only sliding and exfoliation-material transfermechanisms were observed for the IF-MoS2 NP. Rolling was ex-cluded, probably due to the squashed shape of the synthesizedIF-MoS2 NP as shown also in Fig. 1B.

Table 1 summarizes the average values and the standard devia-tions of the normal stress measured, before application of theshear and the stress that caused fracture in IF-WS2 and IF-MoS2under uniaxial compression.

This study shows that the superior tribological behavior ofinorganic IF NP ofWS2 can be ascribed to a combination of threemain mechanisms; i.e., rolling, sliding, and exfoliation-materialtransfer. However, rolling could not be achieved in the presentlyavailable IF-MoS2 NP, most likely due to their highly anisotropic(squashed) morphology. This behavior can be contrasted withbulk 2H-MX2 platelets, in which the main lubrication mechanismis attributed to the facile shearing of one MX2 layer with respectto the underlying and overlying layers.

From the present experiments one can conclude that for quasi-spherical IF-WS2 NP rolling is a predominant mechanism undera normal stress of 0.96±0.38 GPa and low speeds (quasi-static;i.e., 10−7 m∕ sec). Exfoliation of the IF NP becomes the dominat-ing mechanism under high normal stress; i.e., >1.2 GPa. The ex-foliated films coat the shearing surfaces, thereby providing lowfriction. These results are compatible with the compression fail-

ure stress (strength) of individual IF under uniaxial pressure of1.9±0.8 GPa and 2.5±1 GPa forWS2 andMoS2 respectively. Thelow shearing speed in this work is not expected to vary the com-pression strength of the NP, as compared to the case of pure nor-mal loading of the IF NP.

The actual friction coefficient can be explained by a superposi-tion of those three mechanisms. Massive exfoliation of the IF NPeventually causes the degradation of the IF and supersedes therolling and the sliding mechanisms. However, as long as thereis a fresh supply of IF NP to the tribological interface, rolling fric-tion cannot be excluded altogether.

Notwithstanding several key differences, the present workgenerally concurs with the tribological measurements of the IFNP under realistic conditions. So far the main tribological mea-surements and the applications of the NP have been assumed un-der two entirely different conditions: the NP were either mixedwith fluid lubricants (4, 9) or impregnated in metallic films(5, 10). The IF NP were shown to endow excellent lubricationbehavior to interfaces in both environments. This observationsuggests that a common mechanism controls the excellent tribo-logical behavior of the IF NP under different circumstances. Theexperiments described in the present work could be related moreclosely to the second case where no fluid lubricant is used (10).

While the experimental work was performed on a modestnumber (>40) of individual NP, it is believed that the work bearsrelevance to tribological tests under realistic conditions. To estab-lish this connection a simple model calculation was devised.

In this model, the following assumptions were made: the NPhave monosize distribution and they are not agglomerated; i.e.,they form a dense monolayer of NP on the substrate surface. Thecontact area between the flat loading surface and the nanopar-ticle is taken as one quarter of its cross-section (projected area)(23). In order to avoid exfoliation under low shearing rates thevalues of normal stress applied on individual IF should be lessthan 0.5 GPa. Therefore, the highest permissible normal loadwhich can be applied on 1 cm2 area of a densely packed film with-out causing exfoliation under shear movement, could not exceed1.25 · 104 N. This value is quite high for “realistic” tribologicalconditions. Therefore, this work suggests that a smooth contactarea and improved size-dispersion of the quasi-spherical NP

Fig. 6. Demonstration of a sliding mechanism of the IF nanoparticle. (A) to (C) show the sliding of an IF-WS2 nanoparticle along a distance of 170 nm. Thewhite dot points out the sliding onset on the Si wafer.

Fig. 7. SEM images of an IF-WS2 nanoparticle decorated with GNP, demon-strating exfoliation. (A) and (B) show image of the flat AFM tip above theWS2 nanoparticle. (C) and (D) show uniaxial pressing of the WS2 nanoparticleby the AFM tip under loads of 0.7 and 1.2 GPa, respectively. (E) and (F) showthe exfoliated IF nanoparticle in two magnifications. The AFM tip is seen atthe right side of the NP in (E). In this experiment combined normal and shearstresses were applied to the NP by the AFM tip.

Fig. 8. The stress values in ascending order, for the three main mechanismsand the fracture stress of IF-WS2 and IF-MoS2 NP.

19904 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1106553108 Tevet et al.

could extend the validity of the rolling mechanism and postponethe degradation of IF NP; which will lead to reduced friction andwear for extended periods.

Materials and MethodsThe IF-WS2 NP were synthesized using the fluidized bed reactor (FBR) madeof quartz glass (28). WO3 NP were fed into the reactor from the top. The sizedistribution of the IF-WS2 NP prepared in the FBR is between 60 and 300 nm.This size range is dictated by the size distribution of the oxide powder used asa precursor for the synthesis. Mass-produced IF-WS2 were obtained throughoxide to sulfide conversion at 800 and 840 °Cwith optional annealing at high-er temperatures (up to 950 °C) to remove the residual oxide in the core andanneal out the dislocations (29).

TEM (Philips model CM120) equipped with energy dispersive X-ray analy-zer (EDS- EDAX model Phoenix) was used to analyze the NP powder. TecnaiF30 UTmanufactured by FEI and operating at 300 kVwas used for the HRTEMexperiments.

There were several experimental obstacles to overcome in order to exploitthe GNP as a point of reference for the rolling process, and a marker of theposition of the individual IF nanoparticle. The in situ axial nanocompressionof individual NP in the HRSEM was achieved with a dry powder. The GNPwhich served as position markers were deposited from solution which wasfollowed by a careful drying. Prior to the GNP deposition, the IF powderhad to be deagglomerated by a careful sonication process, which did not leadto any observable damage to the NP. The solution had to be dried and theIF-WS2 NP dispersed without contamination. Further details of colloidal goldNP in aqueous solution, appear in the SI Text.

In order to disperse the IF’s on the substrate surface with minimum ag-glomeration, an experimental set-up was built. There are two main drivingforces for the agglomeration of IF NP: the defect at the corners of the facetednanoparticle are presumed to form reactive sites (22) that react with the am-bient and facilitate agglomeration; e.g., through hydrogen bonds and vander Waals interactions (30). Measuring the properties of single inorganicIF NP, relies on the ability to disperse the IF NP on the surface with minimumagglomeration. For that purpose a simple set-up, based on a micro powderdispersing apparatus was built, see Fig. S1. Further details appear in theSI Text.

In-situ axial nanocompression of individual NP in the HRSEM (8) was usedto apply a load on individual IF-WS2 NP. The experiments were performed in aHRSEM (LEO model Supra, 7426) equipped with a Kleindiek nanomanipula-tor. The radial axis of the nanomanipulator (r-radial, θ-elevation, ϕ-azimuth)was used for applying the load. The NP were dispersed on the thin edge of apiece of n-doped Si wafer (111 oriented). The nanomanipulator and the Siwafer were placed against each other perpendicular to the beam, in orderto enable in situ imaging of the tip displacement and the loaded nanopar-ticle. The nanocompression is executed with an atomic force microscope(AFM) cantilever probe along the nanomanipulator’s r-axis (0.25 nm resolu-tion). The AFM probes used in the present experiments were fabricated witha plateau tip (Team-Nanotech). These tips have a plateau contact diameter of350 nm, and nominal spring constant of 100 N∕m. The conductive coating onthese probes reduces the charging and resultant vibrations during scanningin the HRSEM. The HRSEM imaging conditions during the tests were: accel-erating voltage of 3–5 kV and focus distance of 3–5 mm.

The spring constant k of the AFM tip cantilever was determined for eachAFM probe by the Sader noise method (31). The spring constant in the samelot of the probes was found to vary between 44–69 N∕m.

High resolution images were taken using the In-Lens detector which islocated above the edge of the Silicon wafer, see schematic set-up in Fig. 9.Note that the SE2 secondary electron detector is located opposite the nano-manipulator, and is therefore partially shadowed.

The stress σn is defined as the load applied to a specimen divided by thecontact area between the AFM tip and the specimen. The stress, σn wasestimated as follows: σn ¼ Fn

SFn is the applied normal force, S represents the contact area between the

AFM tip and the nanoparticle surface. This technique allows for measure-ment of the applied normal force and evaluation of normal stress duringthe experiment.

The contact area was estimated from the contact length between the NPand the probe in the image (see e.g., Fig. 5) from the HRSEM image. Assum-ing that the nanoparticle is perfectly spherical, the contact length can bemeasured from the image. Using the Hertzian contact model the circular con-tact area can be estimated as a function of the stress. However, because theNP are often heavily faceted (IF-WS2) or oblate shaped (IF-MoS2), a differentapproach was taken. Here it is assumed that: the contact area of the facetedWS2 (MoS2) nanoparticle is a square and the contact length, measured from

HRSEM image is the length of the square. Therefore the contact area is thesquare of the length and the contact area as a function of stress, in thefaceted WS2 nanoparticle, is almost constant and does not vary accordingto the flat-on-sphere Hertzian contact model.

These assumptions are based on the following facts: the initial contactbetween the AFM tip and the faceted WS2 nanoparticle is a flat-on-flat con-tact; uniaxial stress on a face of the WS2 nanoparticle causes stress concen-tration sites at the corners (25). The shape of the facets of the WS2nanoparticle can be a square, rhombus, or pentagon (for simplicity onlysquare facets were used in the calculations).

The actual contact area can be as much as 25% smaller than the square ofthemeasured length. The difference between the stresses calculated by thesetwo different approaches (the round contact area-Hertzian model and thesquare constant contact area model) is not larger than 25%, which reflectsthe accuracy of the measurement.

HRSEM monitoring allowed accurate manipulation of the AFM tip to belocated exactly above the nanoparticle, and accurate determination of thefirst contact of the AFM tip with theWS2 (MoS2) nanoparticle. HRSEM imageswere recorded during the nanocompression experiments.

The applied force (Fn) was evaluated from the position of the AFM tip viaFn ¼ k · ðΔh − ΔdÞ, where k is the spring constant of the AFM cantilever;Δh isthe displacement of the AFM cantilever (see Fig. 10) andΔd, the deformationof the nanoparticle under load. Δd was estimated from the HRSEM imagestaken during the compression test by comparing the height of the deformedshape to the height of the NP before the contact. For Δh estimation two dif-ferent methods were used. The first utilized the difference between the na-nomanipulator position during the process and the position at initial contact.Alternatively, Δh was estimated from the position of the AFM tip after it slidaway or rolled off the IF. The location of the tip after the force was releasedindicates the force that was acting on the cantilever during the measure-ment. This latter method is usually suitable for relatively low applied forces.

The main source of uncertainty in these measurements is associated withthe estimate of the contact area between the AFM tip and the compressedmultilayered polyhedral nanoparticle. The contact area is considered to be asquare but it can be 25% smaller, if the shape is oval or circular. The uncer-tainty in the contact length, measured from the HRSEM images is about 5%.

Fig. 9. Schematic drawing of the set-up used for the nanocompression test

Fig. 10. Schematics showing AFM’s tip pressing multilayered polyhedralnanoparticle

Tevet et al. PNAS ∣ December 13, 2011 ∣ vol. 108 ∣ no. 50 ∣ 19905

CHEM

ISTR

Y

Therefore, the overall uncertainty in the contact area measurement is −30%,þ5%. The uncertainty in the measured spring constant, of the AFM tip’scantilever is 5%. The uncertainty in determining the deformation of themultilayered polyhedral nanoparticle has an effect of less than 1% on thecalculated force. Consequently the normal stress can be 36% higher or11% smaller than the calculated value.

ConclusionsThe friction reduction provided by the IF NP of WS2 (MoS2) canbe explained by a superposition of three main mechanisms; i.e.,rolling, sliding, and exfoliation-material transfer (14). Rollingfriction, which is known to lead to very low friction and wear,requires that the NP be spherically shaped and mechanically verystable. For the sliding mechanism, which also provides low fric-tion, the NP should also be mechanically stable and exhibit lowadhesion and chemical affinity to the underlying surfaces. In theexfoliation mechanism, the multilayered IF NP act as reservoir oflow friction layers which cover the mating surfaces with low fric-tion thin layers, a few monolayers thick (15).

In order to differentiate between the different friction me-chanisms relevant for the IF NP, an additional experiment was

performed in which combined shear and normal forces were ap-plied to individual IF NP. GNP were attached to the IF NP andserved as a position marker relative to the substrate.

It is clear from the current experiments that rolling of IF-WS2is a dominant mechanism under low shear rates and normal stressof 0.96±0.38 GPa. Sliding occurs at slightly higher normal stress,and exfoliation predominates under high normal stress above1.2 GPa. For IF-MoS2 the rolling mechanism was not observed,likely due to the asymmetric (squashed) shape of the NP. Slidingwas the dominant mechanism under low shear rates and normalstress of 0.5±0.19 GPa. Exfoliation became predominant underhigh normal stress of above 1.34 GPa.

ACKNOWLEDGMENTS. This work was supported by the European ResearchCouncil (ERC) project INTIF 226639 and the Israel Science Foundation. Wewould also like to thank the Harold Perlman Foundation and the Irvingand Cherna Moskowitz Center for Nano and Bio-Nano Imaging. H.D.W. isthe recipient of the Livio Norzi Professorial Chair in Materials Science. R.T.holds the Drake Family Chair of Nanotechnology and is the director ofthe Helen and Martin Kimmel Center for Nanoscale Science.

1. Jost P (1976) Economic Impact of Tribology. ProcMechanical Failures Prevention Group(NBS Special Pub. 423, Gaithersburg, MD), pp 117–139.

2. Stachowiak GW, Batchelor AW (1993) Eng.Tribol. Tribology Series, 24 (Elsevier Science,Amsterdam), pp 1–9.

3. Tenne R, Margulis L, Genut M, Hodes G (1992) Polyhedral and cylindrical structures oftungsten disulphide. Nature 360:444–446.

4. Greenberg R, Halperin G, Etsion I, Tenne R (2004) The effect of WS2 nanoparticles onfriction reduction in various lubrication regimes. Tribol Lett 17:179–186.

5. Rapoport L, et al. (2003) Tribological properties of WS2 nanoparticles under mixedlubrication. Wear 255:785–793.

6. Schutte WJ, De Boer JL, Jellinek F (1987) Crystal structures of tungsten disulfide anddiselenide. J Sol State Chem 70:207–209.

7. Kaplan-Ashiri I, et al. (2006) On themechanical behavior ofWS2 nanotubes under axialtension and compression. Proc Natl Acad Sci USA 103:523–528.

8. Tevet O, et al. (2010) Nanocompression of individual multilayered polyhedral nano-particles. Nanotechnology 21:365705–365711.

9. Tannous J, et al. (2011) Understanding the tribochemical mechanisms of IF-MoS2nanoparticles under boundary lubrication. Tribol Lett 41:55–64.

10. Friedman H, et al. (2007) Fabrication of self-lubricating cobalt coatings on metalsurfaces. Nanotechnology 18:115703–115711.

11. Wu JF, Zhai WS, Jie GF (2009) Preparation and tribological properties of WS2 nano-particles modified by trioctylamine. Proc Inst Mech Eng 223:695–703.

12. Fischer A (1996) Well-founaded selection of materials for improving wear resistance.Wear 194:238–245.

13. Rapoport L, et al. (2005) Friction and wear of fullerene-like WS2 under severe contactconditions: friction of ceramic materials. Tribol Lett 19:143–149.

14. Rapoport L, et al. (1997) Hollow nanoparticles of WS2 as potential solid-state lubri-cants. Nature 387:791–793.

15. Golan Y, et al. (1999) Microtribology and direct force measurement of WS2 nestedfullerene-like nanostructures. Adv Mater 11:934–937.

16. Rapoport L, et al. (2003) Modification of contact surfaces by fullerene-like solid lubri-cant nanoparticles. Surf Coat Tech 163-164:405–412.

17. Erdemir A (2000) A crystal-chemical approach to lubrication by solid oxides. Tribol Lett8:97–102.

18. Joly-Pottuz L, Martin JM, De Barros BouchetMI, Belin M (2009) Anomalous low frictionunder boundary lubrication of steel surfaces by polyols. Tribol Lett 34:21–29.

19. Fischer A (1996) Well-founded selection of materials for improving wear resistance.Wear 194:238–245.

20. Margolin A, Popovitz-Biro R, Albu-Yaron A, Rapoport L, Tenne R (2005) Inorganic full-erene-like nanoparticles of TiS2 . Chem Phy Lett 411:162–166.

21. Lahouij I, Dassenoy F, de Knoop L, Martin JM, Vacher B (2011) In situ TEM observationof the behavior of an individual fullerene-like MoS2 nanoparticle in a dynamic Con-tact. Tribol Lett 42:133–140.

22. Shahar C, Levi R, Cohen SR, Tenne R (2010) Gold nanoparticles as surface defect probesfor WS2 nanostructures. J Phys Chem Lett 1:540–543.

23. Kalfon-Cohen E, Barlam D, Tevet O, Cohen SR (2012) Insights on nano-compression ofWS2 inorganic fullerenes: a finite element study. J Mater Res, Available CambridgeJournals Online, Aug. 15, 2011 doi: 10.1557/jmr.2011.212.

24. Stefanov M, Enyashin AN, Heine T, Seifert G (2008) Nanolubrication: How do MoS2-based nanostructures lubricate? J Phys Chem C 112:17764–17767.

25. Braun OM (2005) Simple model of microscopic rolling friction. Phys Rev Lett95:126104–126108.

26. Turkevich J, Stevenson PC, Hillier J (1951) A study of the nucleation and growth pro-cesses in the synthesis of colloidal gold. Discuss Faraday Soc 11:55–75.

27. Frens G (1973) Controlled nucleation for regulation of particle-size in monodispersegold suspensions. Nat-Phys Sci 241:20–22.

28. Feldman Y, Zak A, Popovitz-Biro R, Tenne R (2000) New reactor for production of tung-sten disulfide hollow onion-like (inorganic fullerene-like) nanoparticles. Solid State Sci2:663–672.

29. Feldman Y, et al. (1996) Bulk synthesis of inorganic fullerene-like MS2 (M ¼ Mo, W)from the respective trioxides and the reaction mechanism. J Am Chem Soc118:5362–5367.

30. Schwarz US, Komura S, Safran SA (2000) Deformation and tribology of multi-walledhollow nanoparticles. Europhys Lett 50:762–768.

31. Sader JE (2002) Encyclopedia of Surface and Colloid Science, ed A Hubbard (MarcelDekker, New York), pp 846–856.

19906 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1106553108 Tevet et al.