Friction at the Atomic Scale

Friction at the Atomic Scale

Iused to dread the first week of De-cember. It wasn’t the darkness orBoston’s pre-snow drizzle that made

me gloomy, and it wasn’t the nonexis-tent parking at holiday-frenzied shop-ping malls. This was the week when ab-stracts were due for the annual Marchmeeting of the American Physical Soci-ety, the meeting of condensed-matterphysicists. In 1986 my colleague AllanWidom and I had developed an experi-mental technique that could measurethe frictional force of one-atom-thickfilms sliding along flat solid surfaces.The problem was, I could find nowhereto classify my atomic-scale friction ab-stract within a myriad of March meet-ing subject categories.

It was not that research on friction didnot exist. I had always been welcomedby the multidisciplinary American Vac-uum Society, in sessions on macroscop-ic-scale friction or nanometer-scale sci-ence. But mainstream physicists seemedto have no interest in the topic. Withnear unanimity, they would attribute theorigins of friction as something to dowith surface roughness. Given the every-day familiarity and economic impact offriction, one would have thought thatthey might have been more interested.(By most estimates, improved attentionto friction and wear would save devel-oped countries up to 1.6 percent of theirgross national product, a whopping$116 billion for the U.S. alone in 1995.)

In fact, I wasn’t really alone in my re-search interests. The late 1980s markedthe advent of many new techniques, in-cluding my own, that could study theforce of friction, either experimentally,by sliding atoms on crystalline sub-strates, or theoretically, using new com-puter models. I first referred to the fieldas “nanotribology”—friction, or tribol-ogy, studied in well-defined geometrieson the nanometer scale—in a January

Friction at the Atomic ScaleLong neglected by physicists, the study of friction’s

atomic-level origins, or nanotribology, indicates that the force stems from various unexpected sources, including sound energy

by Jacqueline Krim

74 Scientific American October 1996

GRINDING wears away sliding surfaces. Such instances offriction had always been associated with permanent damageto the surfaces. But new studies have shown that friction canpersist at high levels even in the absence of wear or damage.

Copyright 1996 Scientific American, Inc.

1991 publication, and others began us-ing the term as well. What was once agrassroots community of isolated re-searchers was progressively becoming anaccepted scientific field in its own right.

Since then, nanotribologists have beenregularly discovering that atomic-scalefriction can differ significantly from whatis observed at the macroscopic level.Friction has very little to do with micro-scopic surface roughness, and in someinstances, dry surfaces are actually slick-er than wet ones. The force is complexenough that, even if we can perfectly

characterize a sliding interface, we can-not accurately predict the friction thatwill occur at that interface. If the pre-cise nature between microscopic con-tacts and macroscopic materials couldbe determined, then better understand-ing of friction could lead to such indus-trial innovations as improved lubri-cants and wear-resistant machine parts.

Such technological considerations havedriven humans to attempt to understandfriction since prehistoric times. Morethan 400,000 years ago, our hominidancestors in Algeria, China and Java

were making use of friction when theychipped stone tools. By 200,000 B.C.E.,Neanderthals had achieved a clear mas-tery of friction, generating fire by therubbing of wood on wood and by thestriking of flint stones. Significant de-velopments also occurred 5,000 yearsago in Egypt, where the transportationof large stone statues and blocks for theconstruction of the pyramids demand-ed tribological advances in the form oflubricated wooden sledges.

Writing the Classics

Modern tribology began perhaps500 years ago, when Leonardo

da Vinci deduced the laws governingthe motion of a rectangular block slid-ing over a flat surface. (Da Vinci’s workhad no historical influence, however,because his notebooks remained un-published for hundreds of years.) In the17th century the French physicist Guil-laume Amontons rediscovered the lawsof friction after he studied dry slidingbetween two flat surfaces.

Amontons’s conclusions now help toconstitute the classic laws of friction.First, the friction force that resists slid-ing at an interface is proportional to the“normal load,” or the force that squeez-es the surfaces together. Second, andperhaps counterintuitively, the amountof friction force does not depend on theapparent area of contact. A small blocksliding on a surface experiences as muchfriction as does a large block of the sameweight. To these rules is sometimes add-ed a third law, attributed to the 18th-century French physicist Charles-Au-gustin de Coulomb (better known forhis work in electrostatics): the frictionforce is independent of velocity oncemotion starts. No matter how fast youpush a block, it will experience nearlythe same amount of resistance.

Amontons’s and Coulomb’s classicalfriction laws have far outlived a varietyof attempts to explain them on a funda-mental basis in terms of, say, surfaceroughness or molecular adhesion (at-traction between particles in the oppos-ing surfaces). By the mid-1950s, surfaceroughness had been ruled out as a vi-able mechanism for most everyday fric-tion. Automobile makers and othershad found, surprisingly, that the fric-tion between two surfaces is sometimesless if one of the surfaces is rougher thanthe other [see “Friction,” by FredericPalmer; Scientific American, Febru-ary 1951]. Furthermore, friction can in-

Friction at the Atomic Scale Scientific American October 1996 75

FPG

IN

TER

NATI

ON

AL


crease when two surfaces are madesmoother. In cold welding, for instance,highly polished metals stick togetherquite firmly.

Molecular adhesion, though, was astrong possibility, a conclusion reachedin large part because of the ingeniouswork of F. P. Bowden, David Tabor andtheir co-workers at the University ofCambridge. They also found that fric-tion, though independent of apparentmacroscopic contact area, as Amontonshad stated, is in fact proportional to thetrue contact area. That is, the microscop-ic irregularities of the surfaces touchand push into one another. The sum ofall these contact points constitutes thetrue contact area. Having establishedthat some kind of intimate link existedbetween friction and adhesion, the Cam-bridge group presumed that friction re-

sulted primarily from adhesive bondingat true contact points that was so strongthat tiny fragments were continuallybeing worn away.

But this explanation was wrong. Itsimply could not explain the fact thatsubstantial friction exists even in casesin which wear is negligible. Indeed, un-der Tabor’s own supervision in the1970s, a gifted Ph.D. candidate, JacobN. Israelachvili, developed a “surface-forces apparatus” for atomic-scale fric-tion measurements and found clear evi-dence of wear-free friction. The mea-surement left Tabor to puzzle over wherethat friction might be coming from.

Israelachvili’s apparatus explores thelubricated contacts between uniformmica surfaces. It takes advantage of thefact that mica is atomically smooth:cleaving a piece of mica leaves a surface

that has atomically flat areas spanningas much as one square centimeter, a dis-tance of more than 10 million atoms.(In contrast, typical surfaces might stayflat for a distance of 20 atoms, whereassmooth metals might go on for 300atoms.) When two mica surfaces touch,an interface free of atomic pits or moun-tains (“asperities”) is formed. In the de-vice the backs of the mica surfaces aregenerally glued onto crossed half-cylin-ders that can be moved in two directionsin the horizontal plane. To measure thecontact area and separation, research-ers shine a coherent light beam acrossthe gap and look at a resulting opticaleffect called an interference pattern, aseries of dark and light bands. Deflec-tions of springs connected to the half-cylinders indicate the frictional force.

Early on, the surface-forces appara-tus allowed atomic-scale verification ofthe macroscopic deduction that frictionis proportional to the true contact area.But it would be nearly two decades be-fore Israelachvili, now a full professorat the University of California at SantaBarbara, and his colleagues would es-tablish the elusive link between frictionand adhesion. They discovered that fric-tion did not correlate with the strengthof the adhesive bond itself. Rather fric-tion was connected to adhesive “irre-versibility,” or how differently surfacesbehave when they stick together as com-pared with when they are in the processof becoming unstuck. But in their tri-umph, the investigators could not ad-dress the explicit physical mechanismthat gave rise to the friction they weremeasuring.

James A. Greenwood of the Universi-ty of Cambridge, a world authority ontribological contact between rough sur-faces, summed up the situation in 1992when he wrote, “If some clever personwould explain why friction exists, andis proportional to the [true] area of con-tact, our problem would be solved.”

Good Vibrations

Aleading candidate for that clever person is Gary M. McClelland of

the IBM Almaden Research Center. Inthe 1980s he derived a very simple mod-el for wear-free friction based on vibra-tions of atomic lattices. Unknown toMcClelland, the model had been pub-lished by G. A. Tomlinson of the BritishNational Physical Laboratory in 1929,as had a far more sophisticated treat-ment by Jeffrey B. Sokoloff and his co-

Friction at the Atomic Scale76 Scientific American October 1996

EARLY STUDIES OF FRICTION, such as those done in the 18th century by theFrench physicist Charles-Augustin de Coulomb, helped to define the classical laws offriction and attempted to explain the force in terms of surface roughness, a feature thathas now been ruled out as a significant source.

UN

IVER

SIT

Y O

F LE

EDS L

IBR

AR

Y


workers at Northeastern University in1978. But these works had received lit-tle attention.

Friction arising from atomic-latticevibrations occurs when atoms close toone surface are set into motion by thesliding action of atoms in the opposingsurface. (The vibrations, which are reallyjust sound waves, are technically calledphonons.) In this way, some of the me-chanical energy needed to slide one sur-face over the other is converted to soundenergy, which is eventually transformedinto heat. To maintain the sliding, moremechanical energy must be added, and,hence, one has to push harder.

The amount of mechanical energytransformed into sound waves dependson the nature of the sliding substances.Solids are much like musical instrumentsin that they can vibrate only at certaindistinct frequencies, so the amount ofmechanical energy consumed will de-pend on the frequencies actually excited.If the “plucking” action of the atoms in

an opposing surface resonates with oneof the frequencies of the other, then fric-tion arises. But if it is not resonant withany of the other surface’s own frequen-cies, then sound waves are effectively notgenerated. This feature opens the excit-ing possibility that sufficiently smallsolids, which have relatively few reso-nant frequencies, might exhibit nearlyfrictionless sliding.

In any case, McClelland, excited bythe fact that not only wear-free but alsonearly friction-free sliding was a theo-retical possibility, proceeded to collabo-rate with his colleague C. Mathew Mateand others. To measure nanometer-scalefriction, they adapted a newly inventedinstrument: the atomic-force microscope.With it, they published their first obser-vations of friction, measured atom byatom, in a landmark 1987 paper.

An atomic-force microscope consistsof a sharp tip mounted at the end of acompliant cantilever. As the tip isscanned over a sample surface, forces

that act on the tip deflect the cantilever.Various electrical and optical means(such as capacitance and interference)quantify the horizontal and vertical de-flections. The microscope can detectfriction, adhesion and external loadingforces as small as a piconewton, or 10–12

newton. (Loosely speaking, a piconew-ton is to a fly’s weight as a fly’s weightis to an average person’s.) By the early1990s, the IBM researchers had set uptheir friction-force microscope in ultra-high vacuums, allowing them to studythe sliding of a diamond tip over a crys-talline diamond surface with a contactarea estimated to be less than 20 atomsin extent.

McClelland and his colleagues’ mea-surements yielded a friction force thatexhibited no dependence on normalload. According to the classical frictionlaws, this result would have implied zerofriction. But not only was friction evi-dent, the shear stress, or force per arearequired to maintain the sliding, was


SHEAR STRESS, the amount of force per unit of true contactarea needed to maintain the sliding of one object on another, isone measure of friction that has been explored with several in-

struments. Collectively, they have recorded a range of stressspanning 12 orders of magnitude, all in experimental geometriesfree of wear, surface damage and roughness.

10–2 10–1 1 101 102 103 104 105 106 107 108 109 1010

SHEAR STRESS (NEWTONS PER SQUARE METER)

QUARTZ

LIGHT

CANTILEVER

NEEDLE

ELECTRODE

MICA

LUBRICANT

SPRING

CLEAVED MICABETWEEN

HALF-CYLINDERS

SAMPLE

QUARTZ-CRYSTAL MICROBALANCEcan measure friction between its electrodeand a layer of material only one or twoatoms thick deposited on the electrode.Changes in the vibrational properties of thequartz indicate how much the depositedlayer slips on the underlying surface. Com-puter simulations of the sliding layers, suchas that of a liquid layer of krypton (white ininset) on top of gold (blue), are used toconfirm the findings of the microbalance.

SURFACE-FORCES APPARATUSmakes use of two cleaved mica surfaces,which are the smoothest surfaces known.Investigators can place lubricant films,which can be as thin as a few molecules,between the mica surfaces and slide themabout, to see how the films affect the slid-ing (insets).

LATERAL-FORCE MICROSCOPEis a variation of the atomic-force micro-scope. It employs a fine needle mounted ona cantilever. The tip deflects as it dragsalong the sample’s surface. Light reflect-ing off the tip indicates the degree of de-flection, thus providing a measure of thefriction between the tip and the surface.Researchers have used the microscope topush “islands” of carbon 60 (green crystalsin inset) across a salt surface.

JAR

ED S

CH

NEI

DM

AN

DES

IGN

MA

RK

O. R

OB

BIN

S

John

s H

opki

ns U

nive

rsit

y

RO

LAN

D L

ÜTH

IU

nive

rsit

y of

Bas

el


enormous: one billion new-tons per square meter, or150,000 pounds per squareinch. That force is largeenough to fracture top-gradesteel. It was becoming clearthat even if the atomic natureof the sliding contact wascompletely known, our abil-ity to predict the friction forceoccurring at that contact wasvirtually nonexistent.

To date, nanotribologistshave collectively observed aremarkable range of shearstresses, from 0.01 newton to10 billion newtons per squaremeter. For example, RolandLüthi, Ernst Meyer and co-workers at the Institute ofPhysics at the University of Basel havepushed “islands” of one-molecule-thickbuckminsterfullerene (“buckyballs,” orcarbon 60) along a crystalline salt sur-face with a modified atomic-force mi-croscope tip approaching single-atomdimensions. They found shear stressesof 10,000 to 100,000 newtons persquare meter, orders of magnitude low-er than those associated with typicalmacroscopic-scale solid lubricants, suchas graphite powder. (The shear stressappears high only because it is mea-sured over a square meter of true—notapparent—contact area, which in gen-eral is orders of magnitude smaller thanthe apparent contact area. When graph-ite is used, say, to lubricate a lock cylin-der, even the apparent contact area isquite small, so the actual friction en-countered can be rather low.) The re-searchers also measured the force need-ed to slide the tip over the top of thebuckyball island and found it to be“stickier” than the salt.

Shear stresses orders of magnitudelower have been observed in my ownlaboratory by means of a quartz crystalmicrobalance, a device that for decadeshas been used to weigh samples as lightas a few nanograms. It consists of a sin-gle crystal of quartz that stably oscillatesat high frequency (five to 10 milliontimes a second). We deposit metal-filmelectrodes onto its surfaces and thencondense single-atom-thick films of adifferent material onto the electrodes.The condensation onto the microbal-ance lowers the frequency, providing ameasure of how well the film particlescan track the shaking of the underlyingquartz substrate. The smaller the result-ing amplitude of vibration, the greater

the friction from the “rubbing” action ofthe film sliding about on the substrate.

The quartz microbalance is currentlythe only experimental apparatus oper-ating on a timescale short enough to seehow atomic-scale friction depends onvelocity. Although the third classic lawof friction states that friction is indepen-dent of velocity, researchers later foundthis rule to be untrue. (Coulomb himselfsuspected as much but could not proveit.) For example, to decelerate an auto-mobile uniformly and stop it without ajerk, the driver must ease up on thebrake in the final moments, demonstrat-ing that friction increases with slowerspeeds. Such macroscopic velocity de-pendencies are almost always attribut-ed to changes at the microscopic con-tact points (which can melt at high slid-ing speeds and can increase in area atlow speeds, where they “tear apart”

more slowly and hence have more timeto form bonds). But for a geometry inwhich the contact area remains fixed,such as that of our quartz microbalance,friction is in fact predicted to exhibitjust the opposite behavior, increasing indirect proportion to the sliding speed.We recently confirmed this observationfor one-atom-thick solid films slidingover crystalline silver and gold surfaces.

Slippery when Dry

But analytic theories did not predictour surprising discovery in 1989

that krypton films sliding on crystallinegold surfaces were slipperier when dry.We observed that friction forces for liq-uid films were about five times higherthan for solid films, with shear stressesfor the solid films being a minuscule 0.5newton per square meter for slidingspeeds of one centimeter per second. Theeffect was so counterintuitive to methat I held off publishing the result formore than a year after discovering it.

Why should a liquid layer cause morefriction on the atomic scale when inmost situations in everyday life it lubri-cates two surfaces? Computational stud-ies have supplied the crucial link, forthey open a rare window into molecu-


CHEMICAL REACTIONS can occur be-tween two sliding interfaces. In this set-up, an ethane molecule, composed of twocarbon atoms (green) and six hydrogenatoms (blue), is sandwiched between twodiamond surfaces, which are terminatedwith hydrogen atoms (1). As the surfacesslide, the ethane loses a hydrogen atom(2), becoming an ethyl radical. The free

CONTACT POINTS are the places wherefriction occurs between two rough sur-faces sliding past each other (top). If the“normal load”—the force that squeezesthe two together—rises, the total area ofcontact increases (bottom). That increase,and not the surface roughness, governsthe degree of friction.

JAR

ED S

CH

NEI

DM

AN

DES

IGN

1 2

DIAMOND ETHANE


lar behavior that is unattainable by anyother means. Several researchers havebroken new nanotribological groundwith the computer. They include UziLandman of the Georgia Institute ofTechnology, who pioneered simulationsof point contacts; Judith A. Harrison ofthe U.S. Naval Academy, who modeledinterfacial chemical effects; and JamesBelak of Lawrence Livermore NationalLaboratory, who analyzed machiningand wear.

It was Mark O. Robbins and his co-workers at Johns Hopkins University,though, who answered the question ofliquid friction when they simulated one-atom-thick krypton films sliding on crys-talline gold surfaces. They demonstrat-ed that liquid krypton atoms, beingmore mobile than solid krypton, could“get stuck” more easily in the gaps be-tween the solid gold atoms. Note thatthe shearing takes place between a solidand a liquid surface, a situation differ-ent from macroscopic cases of liquid lu-brication. In those instances, the shear-ing takes place within the bulk of theliquid (that is, at a liquid-liquid inter-face), which usually offers less resis-tance to shearing than does a solid-liq-uid interface.

The near-perfect agreement between

Robbins’s model and our experimentalresult is both surprising and revealing,because all the friction in his calcula-tions was attributed to lattice vibrations(sound waves). His model neglectedfriction from electrical effects. For insu-lating surfaces, such friction arises fromthe attraction of positive and negativecharges that have separated at the inter-face. (A similar attraction is achievedwhen a balloon is rubbed on hair andleft to cling to a wall.) When one orboth of the surfaces in contact are met-al, however, then the buildup of chargeshould not be significant. Rather anoth-er type of electronic friction may occur,as suggested by Mats Persson of Chal-mers University of Technology in Göte-borg, Sweden, and extensively investi-gated by theorist Bo N. J. Persson ofthe Jülich Research Center in Germany.That friction is related to resistance feltby mobile electrons within the metallicmaterial as they are dragged along bythe opposing surface.

Physicists know that such friction ex-ists but not how important it is (whichis why small solids might exhibit nearlyfrictionless, instead of completely fric-tionless, sliding). The success of themodel calculated by Robbins and hiscolleagues seemed to imply that elec-tronic effects play no significant role infriction.

To investigate this issue further, werecently measured the force needed toslide one- and two-atom-thick solidfilms of xenon along a crystalline silversurface, and we observed that frictionincreased by approximately 25 percentfor the two-atom-thick xenon film.

Did this 25 percent increase stem fromelectronic effects? Probably not. Bo Pers-

son, Robbins and Soko-loff have performed inde-pendent computer simula-tions of the xenon-silversystem, and their prelimi-nary computational re-sults indicate that the fric-tion associated with soundwaves is much greater fortwo layers than for one.Basically, two layers makefor a more elaborate “mu-sical instrument,” so thereare more resonant frequen-cies to excite and hencemore friction. Electronicfriction undoubtedly ex-ists, but its strength maybe determined in largepart by only those atoms

immediately adjacent to the interface.The parameters selected to representmetal surfaces in a simulation couldeasily mask it. But as theoretical andsimulational efforts become increasing-ly sophisticated, we should eventuallybe able to estimate with precision theproportion of energy loss that is associ-ated with electronic effects and latticevibration.

Rewriting the Rules

The recent progress in nanotribologyclearly demonstrates that the laws

of macroscopic friction are inapplicableat the atomic scale. We can now rewritethe laws of friction in a more generalway. First, the friction force depends onhow easily two surfaces become stuckrelative to becoming unstuck: it is pro-portional to the degree of irreversibilityof the force that squeezes the two sur-faces together, rather than the outrightstrength of the force. Second, the fric-tion force is proportional to the actual,rather than apparent, area of contact.Finally, the friction force is directly pro-portional to the sliding speed of the in-terface at the true contact points, solong as the surfaces are not allowed toheat up and the sliding speed remainswell below the speed of sound. (Nearthat speed, it levels off because the lat-tice vibrations cannot carry away thesound energy rapidly enough.)

The discrepancy between microscop-ic and macroscopic frictional phenome-na greatly diminishes if one notes thatthe true area of contact between macro-scopic objects is likely to be proportion-al to the squeezing force. The harder yousqueeze, the more area comes into con-


hydrogen subsequently removes a hydro-gen atom from the diamond and bondswith it to form a molecule of hydrogengas (3). The ethyl radical eventually be-comes chemically bound to one of the di-amond surfaces (4). The diagram is basedon computer simulations conducted byJudith A. Harrison and her co-workers atthe U.S. Naval Academy.

MIC

HA

EL G

OO

DM

AN

3 4


tact. So friction appears to be propor-tional to the normal load, as Amontonsstated.

And what ever became of surfaceroughness? Alas, its importance seemsto shrink. Physicists had presumed thatsurface irregularities played a role in

stick-slip friction, in which surfaces glid-ing past one another momentarily clingand then let go. Notable examples in-clude screeching train brakes and finger-nails on blackboards. Roughness wasthought to cause the random nature ofthe sticking and the slipping. But Steve

Granick and his colleagues at the Uni-versity of Illinois recently observed stick-slip friction in lubricated contacts be-tween nominally “perfect” mica surfac-es. They applied millions of repetitivecycles of a sinusoidal force to confinedliquids without wear and observed re-sults suggesting that randomness (spe-cifically, so-called 1/f noise) may be in-trinsic to the stick-slip friction itself.

Considering the current race to man-ufacture machine components with as-toundingly small dimensions, what istoday considered fundamental researchon the atomic scale may give way to-morrow to direct application. For in-stance, we now know why substancesmade of branched-chain molecules makebetter lubricants than straight-chain mol-ecules, even though the branched-chainones are, in bulk form, more viscous.(They remain as a liquid under greaterforces than do the straight-chained mol-ecules and thus are better able to keeptwo solid surfaces from touching.) Nano-tribologists working with known con-tact geometries may one day help chem-ists understand friction-induced reac-tions taking place on surfaces or aidmaterials scientists in designing sub-stances that resist wear. As the need toconserve both energy and raw materialsbecomes more urgent, physicists’ rushto understand basic frictional processescan be expected only to accelerate.

To obtain high-quality reprints of thisarticle, please see page 127.


The Author

JACQUELINE KRIM, a professor of physics at Northeast-ern University and a member of its Center for InterdisciplinaryResearch on Complex Systems, received her B.A. from the Uni-versity of Montana and her Ph.D. from the University of Wash-ington. She received a National Science Foundation Presiden-tial Young Investigator Award in 1987 and a creativity awardin materials research in 1992. She has chaired the nationalboard of trustees of the American Vacuum Society and current-ly serves on the executive board of its surface science division.She thanks the NSF for its continued research support.

Further Reading

History of Tribology. D. Dowson. Longman, London, 1979.Fundamentals of Friction: Macroscopic and Microscopic Pro-cesses. Edited by I. L. Singer and H. M. Pollock. Kluwer, 1992.

Handbook of Micro/Nanotribology. Edited by B. Bhushan. CRCPress, 1995.

Nanotribology: Friction, Wear and Lubrication at the AtomicScale. B. Bhushan, J. N. Israelachvili and U. Landman in Nature, Vol.374, pages 607–616; April 13, 1995.

Physics of Sliding Friction. Edited by B.N.J. Persson and E. Tosatti.Kluwer, 1996.

DIAMONDLIKE TIP made of carbon(blue) and hydrogen (yellow) slides acrossthe face of a similar material, a diamondsurface made of carbon (green) and termi-nated with hydrogen atoms (red). Suchcomputer simulations help in tribochem-istry, or the study of friction-induced reac-tions. In this particular computation, thetip and surface deformed, but no chemi-cal reactions occurred.

SA

JUD

ITH

A. H

AR

RIS

ON

U.S

. N

aval

Aca

dem

y


Documents

Friction at the Atomic Scale