has a small, yet nonzero, quantum defect. Thecorresponding lifetime of Cs(36F) is t36F ≈ 18 ms(25). The fact that the decay of the molecularstate is different from the purely radiative decayof the Cs(nS) Rydberg state, and closer to thenearby (n – 4)F state, is another indication of thedegenerate mixing involved in the formation ofthe hybrid trilobite molecules.The class of ultralong-range molecule observed

here is an advantageous cross between a tradi-tional “trilobite” molecule and a low-l ultralong-range Rydberg molecule (7–12). The discoveryof the trilobite ultralong-range Rydberg mole-cule could open opportunities in ultracold chem-istry and strongly correlated many-body physics,as these exotic states require engineered meso-scopic localization and kilo-Debye permanentelectric dipole moments.

REFERENCES AND NOTES

1. S. Ospelkaus et al., Science 327, 853–857 (2010).2. L. Carr, D. DeMille, R. V. Krems, J. Ye, New J. Phys. 11, 055049

(2009).

3. H. Weimer, M. Müller, I. Lesanovsky, P. Zoller, H. P. Büchler,Nat. Phys. 6, 382–388 (2010).

4. M. D. Lukin et al., Phys. Rev. Lett. 87, 037901 (2001).5. W. Klemperer, K. K. Lehmann, J. K. G. Watson, S. C. Wofsy,

J. Phys. Chem. 97, 2413–2416 (1993).6. W. Li et al., Science 334, 1110 (2011).7. C. H. Greene, E. L. Hamilton, H. Crowell, C. Vadla, K. Niemax,

Phys. Rev. Lett. 97, 233002 (2006).8. V. Bendkowsky et al., Nature 458, 1005–1008 (2009).9. J. Tallant, S. T. Rittenhouse, D. Booth, H. R. Sadeghpour,

J. P. Shaffer, Phys. Rev. Lett. 109, 173202 (2012).10. M. A. Bellos et al., Phys. Rev. Lett. 111, 053001 (2013).11. A. T. Krupp et al., Phys. Rev. Lett. 112, 143008 (2014).12. D. A. Anderson, S. A. Miller, G. Raithel, Phys. Rev. Lett. 112,

163201 (2014).13. H. Saßmannshausen, F. Merkt, J. Deiglmayr, http://arxiv.org/

abs/1412.0846 (2014).14. C. H. Greene, A. S. Dickinson, H. R. Sadeghpour, Phys. Rev.

Lett. 85, 2458–2461 (2000).15. J. B. Balewski et al., Nature 502, 664–667 (2013).16. M. D. Lukin, P. R. Hemmer, Phys. Rev. Lett. 84, 2818–2821

(2000).17. P. W. Anderson, Phys. Rev. 109, 1492–1505 (1958).18. E. Fermi, Nuovo Cim. 11, 157–166 (1934).19. A. Omont, J. Phys. 38, 1343–1359 (1977).20. See supplementary materials on Science Online.

21. C. Bahrim, U. Thumm, I. I. Fabrikant, J. Phys. B 34, L195–L201(2001).

22. E. L. Hamilton, C. H. Greene, H. R. Sadeghpour, J. Phys. B 35,L199 (2002).

23. J. W. Cooper, Phys. Rev. 128, 681–693 (1962).24. I. I. Beterov, I. I. Ryabtsev, D. B. Tretyakov, V. M. Entin, Phys.

Rev. A 79, 052504 (2009).25. T. F. Gallagher, Rydberg Atoms (Cambridge Univ. Press,

Cambridge, 1994).

ACKNOWLEDGMENTS

Supported by NSF grant PHY-1205392 (D.B., J.Y., and J.P.S.)and by an NSF grant through ITAMP at the Harvard-SmithsonianCenter for Astrophysics (H.R.S.). S.T.R. thanks B. M. Peden andB. L. Johnson for insightful discussions. All data are available uponrequest.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/348/6230/99/suppl/DC1Materials and MethodsSupplementary TextReferences (26–28)

2 September 2014; accepted 6 February 201510.1126/science.1260722

TRIBOLOGY

Mechanisms of antiwear tribofilmgrowth revealed in situ bysingle-asperity sliding contactsN. N. Gosvami,1 J. A. Bares,1* F. Mangolini,2 A. R. Konicek,3

D. G. Yablon,3† R. W. Carpick1‡

Zinc dialkyldithiophosphates (ZDDPs) form antiwear tribofilms at sliding interfaces andare widely used as additives in automotive lubricants. The mechanisms governing thetribofilm growth are not well understood, which limits the development of replacementsthat offer better performance and are less likely to degrade automobile catalyticconverters over time. Using atomic force microscopy in ZDDP-containing lubricant basestock at elevated temperatures, we monitored the growth and properties of the tribofilmsin situ in well-defined single-asperity sliding nanocontacts. Surface-based nucleation,growth, and thickness saturation of patchy tribofilms were observed. The growth rateincreased exponentially with either applied compressive stress or temperature,consistent with a thermally activated, stress-assisted reaction rate model. Althoughsome models rely on the presence of iron to catalyze tribofilm growth, the films grewregardless of the presence of iron on either the tip or substrate, highlighting thecritical role of stress and thermal activation.

Additives are crucial components of lubri-cants used in a wide range of tribologicalapplications, including vehicles, turbines,and manufacturing equipment (1). Anti-wear additives and friction modifiers can

extend many industrial and automotive applica-tion lifetimes by orders of magnitude, resultingin considerable energy andmaterial savings. Oneof the most crucial modern antiwear additives iszinc dialkyldithiophosphate (ZDDP), chemicalformula Zn[S2P(OR)2]2, where R is an alkyl or arylgroup (2, 3) (fig. S2). Extensivemacroscopic studieshave shown that ZDDP molecules decomposeat rubbing interfaces (4, 5) and form protective

surface-bonded tribofilms that reduce wear byminimizing metal-to-metal contact of steel andiron (3) and other material pairs (6, 7). ZDDP-derived tribofilms consist of rough, patchy, pad-likefeatures that are composed of pyro- or orthophos-phate glasses in the bulk, with an outer nanoscalelayer of zinc polyphosphates and a sulfur-richlayer near the metal surface (3). However, thetribochemical film growth pathways are not es-tablished, and the factors that determine the filmmorphology and thickness (generally 50 to 150nm) are unknown (3). Furthermore, ZDDP’s effec-tiveness as an antiwear additive for advancedengine materials is not yet clear. For low-weight

materials such as Al- and Mg-based alloys, ZDDPforms robust tribofilms primarily on load-bearinginclusions but not on surrounding softer ma-trices (6, 8). Although ZDDP tribofilms can beformed between other nonferrous material pairs[e.g., low-frictiondiamond-like carbon (DLC) films],they are often less durable than those formedwhen steel or iron is present, for reasons not yetunderstood (9, 10). ZDDP often increases frictionallosses (3) and produces Zn-, P-, and S-containingcompounds in automobile engine exhaust, thusreducing catalytic converter efficiency and life-time (1, 3, 11, 12). However, despite decades ofresearch, no suitable substitute for ZDDP hasyet been found (12), and research efforts havesought to identify the beneficial mechanisms un-derlying the growth and antiwear properties ofZDDP-derived tribofilms.Various macroscopic methods have been de-

veloped to produce ZDDP tribofilms (13, 14), andthe resulting films have been studied by many exsitu mechanical and chemical approaches (3, 15)and atomistic simulations (8). It is widely as-sumed that the tribofilm acts as a protective layerthat is continually replenished, reducing metal-to-metal contact (3). Some studies indicate thatantiwear properties result from ZDDP’s ability toreduce peroxides in the base stock, thereby pre-venting corrosion (16, 17). One model explainingZDDP tribofilm formation on steel is based onhard and soft acid-base (HSAB) reactions (18),which require the exchange of Zn2+ and Fe3+

cations between the ZDDP and iron oxide wear

102 3 APRIL 2015 • VOL 348 ISSUE 6230 sciencemag.org SCIENCE

1Department of Mechanical Engineering and AppliedMechanics, University of Pennsylvania, Philadelphia, PA19104, USA. 2Department of Materials Science andEngineering, University of Pennsylvania, Philadelphia, PA19104, USA. 3Corporate Strategic Research, ExxonMobilResearch and Engineering, Annandale, NJ 08801, USA.*Present address: BorgWarner Powertrain Technical Center,Auburn Hills, MI 48326, USA. †Present address: SurfaceChar LLC,Sharon, MA 02067, USA. ‡Corresponding author. E-mail:carpick@seas.upenn.edu

RESEARCH | REPORTSon O

ctober 27, 2020

http://science.sciencemag.org/

Dow

nloaded from

particles, respectively, where the latter are di-gested within the tribofilm (19). Direct experi-mental evidence for this model is lacking (3), nordoes the model explain tribofilm formation onnonferrous surfaces (6, 7). In contrast,Mosey et al.’sfirst-principles atomistic simulations proposedthat tribofilm formation results from contactpressure–induced cross-linking of zinc phosphatemolecules (8), which are a thermal or catalyticdecomposition product of ZDDP (11, 15).There is no general consensus on the growth

mechanism, and no models conclusively explaineither the tribofilm patchiness or why the filmthickness is limited. All prior experiments havebeen conducted for macroscopic, multi-asperitycontacts (specific asperity contact areas and pres-sures thus being unknown) that are then analyzedpostmortemand ex situ, often after extracting thesample from base stock, which may alter thetribofilm (20). Althoughmacroscopic in situ studiesof zinc polyphosphates under static compression(21, 22) have shown irreversible loss of crystal-linity and little increase in polymerization withincreased pressure, these studies do not involvedynamic sliding. In situ single-asperity slidingstudies have several advantages: Contact loadsand geometries can be controlled and quantified;local tribofilm properties such as morphologicalevolution with nanometer resolution, tribofilmvolume, friction, adhesion, and wear can bemea-sured concurrently; and results can be comparedwith atomistic simulations (8).We conducted in situ single-asperity studieswith

an atomic force microscope (AFM). The AFM tipwas slid against an Fe-coated or uncoated Sisubstrate at temperatures up to 140°C whileimmersed in ZDDP-containing base stock (23)to dynamically generate the tribofilm (fig. S3).Low-load (10 to 20 nN) contact-mode imagingrevealed a soft, weakly bound thermal film, formedwithout prior sliding, that was easily removed bysliding with a load of 100 nN (fig. S4). This well-known “thermal film” is formed from adsorbeddecomposition products of ZDDP (15, 24). Typi-cal thermal film thicknesses of ~10 nm were ob-tained after ~1 hour of heating the base stockbath, but the thickness can vary with the age ofthe oil and heating time (24). After removing thethermal film with the tip, sliding was continuedwithin the same region with a higher normalload to induce the growth of the tribofilm.The morphological evolution of the tribo-

chemical products with increasing sliding cycles(one sliding cycle = one 1 mm × 1 mm image) re-vealed randomly located nucleation sites andsubsequent growth of the sliding-induced tribo-film (Fig. 1). The tribofilm grew vertically and lat-erally (only within the region scanned at highernormal load) with further sliding, leading to arough surface (movie S1). The total film volumeincreased linearly with sliding time during thefirst ~1200 cycles (Fig. 1, inset), indicating a zero-order reaction (23). The growth rate then increasedrapidly, fitting well to a power-law function cor-responding to an nth-order reaction with n =0.22 (fig. S7), which indicates a complex reactionpathway. The observed growth of a patchy film

SCIENCE sciencemag.org 3 APRIL 2015 • VOL 348 ISSUE 6230 103

6500 5500 4500 3500

3000

2400

2000

16001200900500

mn 38.53

0 nm

36 nm

1.0 µm

Slow Fast

Sliding cycles

Fig. 1. Morphology and volumetric growth of tribofilm. Tribofilm volume (mean T SD) versus slidingcycles, with linear and power-law fits to the initial and subsequent growth regimes, respectively. Inset showsa zoom-in of the initial growth period. Around the perimeter, clockwise from upper left: periodically acquired2 mm × 2 mm AFM images of an iron oxide surface using a DLC-coated silicon AFM tip immersed in ZDDP-containing base stock, acquired at a nonperturbative load of 20.0 T 0.1 nN. Below each image is thenumber of previously acquired 1 mm × 1 mm scans (“sliding cycles”) at a load of 340 T 2 nN (4.2 T 0.5 GPa).The images demonstrate progressive tribofilm growth where the higher load was applied.

Fig. 2.Tribofilm volu-metric growth ratedependence on con-tact pressure.Tribofilmgrowth rate is exponentialat low contact pressures(data are means T SD).Further growth isinhibited above ~5 GPaas the tip wears awaynewly deposited material.The 2 mm × 2 mm topo-graphic contact-modeAFM images shownwere acquired at a non-perturbative load aftergenerating tribofilms inthe central 1.0 mm ×0.5 mm regions at variouscontact pressures.

RESEARCH | REPORTSon O

ctober 27, 2020

http://science.sciencemag.org/

Dow

nloaded from

matches well with macroscopic results (25, 26).Such macroscopic studies cannot make clearwhether the patchiness resulted from multipleasperities applying a range of pressures at dif-ferent contact points, or by othermeans. Becausethe loads and contact geometry are well con-trolled in our single-asperity experiments, the het-erogeneity is apparently intrinsic to the growthmechanism. This may indicate that nucleation issensitively dependent on surface heterogeneitiessuch as defects or roughness and/or that thereare instabilities in the growth (perhaps affected byatomic-level stress variations within the single-asperity contact) immediately after randomly oc-curring nucleation events.At these stresses (~4 GPa), the tribofilm growth

rate was low, and the volume rarely reached alimiting value within the time frame of our ex-periments (~10 hours), whereas growth typicallysaturates within a few hours in macroscopic ex-periments (27). This discrepancy may be due todifferences in sliding speeds (~80 mm/s for theseAFM experiments versus millimeters to metersper second for macroscopic tests) or contactareas (on the order of 10 to 100 nm2 in AFMversus ~109 nm2 for macroscopic tests), both ofwhich reduce the area per unit time covered byAFM. The far larger amount of fluid exchangeand the multi-asperity nature of the macro-scopic contacts will also affect growth. Fortu-nately, AFM experiments performed at highernormal stresses (~6.5 GPa) enhanced the growthrate, and films reached a limiting height afterprolonged sliding. We observed tribofilm wearonce it reached a thickness of ~30 to 40 nm,preventing further growth (fig. S5). At this thick-ness, there was no observable contrast in fric-tion between the tribofilm and the surroundingsubstrate. However, before the tribofilm growthhad saturated, a transient increase in frictionwas observed (fig. S10). Further study is requiredto determine whether this effect is due to changesin tribofilm adhesion, modulus, roughness, orinterfacial shear strength. However, the increaseseen is consistent with macroscopic studies thatreport transient increases in friction for ZDDP-infused base stocks (28).Within the subnanometer vertical resolution

limits of our instrument, the tribofilms formedwithout any observable wear of the iron oxidesubstrate. The proposed HSAB mechanism re-quires substantial plastic deformation and wearof the substrate (18). Considering the nanoscaledimensions of the nucleation centers observed inour experiments, the possibilities of cation ex-change and digestion of atomic-scale debris viamolecular-level mechanical mixing cannot be ex-cluded. However, such a mechanism does not ex-plain observations of similar macroscopic ZDDPtribofilms on other substrates such as DLC andsilicon (6, 7, 10, 29). We also observed formationof tribofilms in AFM experiments using Si sub-strates with no Fe present (fig. S6); the growth rateand morphology of these films were indistinguish-able from those we formed on Fe substrates.Our results also provide direct evidence that

the tribofilm is not a product of sliding-induced

transformation of the adsorbed thermal film,because growth occurred in regions where thethermal film was completely removed (fig. S4).Rather, these results indicate that tribofilmgrowth is fed by molecular species from solu-tion into the contact zone, where tribochemicalreactions occur.Tribofilm growth rate and morphology were

investigated as a function of normal load, whichis directly related to the initial contact pressure(contact pressure at a fixed load will decrease asthe compliant tribofilm’s thickness increases).Multiple tribofilms were generated by slidingthe AFM probe for 2000 sliding cycles at 100°Cfor a range of fixed loads (i.e., different initialcontact pressures) (Fig. 2). Tribofilm morphol-ogies and volumes clearly reveal that growth isstrongly affected by contact pressure. Beyond5.2 T 0.6 GPa, tribofilm deformation and pile-up was observed and the growth rate was stabi-lized, indicating concurrent tribofilm generationand removal. This agrees with macroscopic ob-servations and directly demonstrates the sac-rificial property of ZDDP tribofilms beyond acritical thickness and contact pressure at thenanoscale (30).The stress-dependent growth rate Ggrowth rate

(nm3/s) fits well to a stress-activated Arrheniusmodel (Fig. 2):

Ggrowth rate ¼ G0 exp −DGact

kBT

� �ð1Þ

where the prefactor G0 depends on the effectiveattempt frequency and the molar volume of thegrowth species (23), DGact is the free activationenergy of the rate-limiting reaction in the growthprocess, kB is Boltzmann’s constant, and T isabsolute temperature. The fit assumes that DGact

is influenced by stress according to

DGact ¼ DUact − sDVact ð2Þ

where DUact is the internal activation energy(i.e., the energy barrier in the absence of stress),s is the mean value of the stress componentaffecting the activation barrier (assumed to bethe compressive contact pressure), and DVact isthe activation volume (31). The good fit sug-gests that tribofilm formation is an activatedprocess (31). We find DUact = 0.8 T 0.2 eV andDVact = 3.8 T 1.2 Å3, consistent with parame-ters for single atomic bond breaking or forma-tion processes. The stress dependence suggeststhat the observed heterogeneous nucleation(Fig. 1) could result from atomic-scale surfaceroughness, which would lead to varying contactareas and stresses for a given normal load; thus,the energy barrier for the relevant tribochemicalreaction would be lower where the local stressis higher.Experiments performed as a function of tem-

perature provide further support for an activatedtribochemical reaction mechanism (Fig. 3). Thevolumetric growth rate of tribofilms generatedby 5000 sliding cycles at ~4.4 GPa depended ex-ponentially upon temperature. From fitting Eq. 1,we obtain DGact = 0.62 T 0.10 eV. Using the initialcontact pressure determined from AFM force-distance data and using DVact from data in Fig. 2,we obtain DUact = 0.74 T 0.10 eV using Eq. 2, inexcellent agreement with the value obtainedfrom the stress-dependent data. This confirmsthe applicability of reaction rate theory by usingindependent stress- and temperature-dependentmeasurements. Our results provide a robust basisto support the idea that tribofilm growth occursvia stress-activated and thermally activated tribo-chemical reactions, in contrast to previous empir-ical approaches (25). Our data do not provide anydirect support for the HSAB model (18), whichasserts that tribofilms can form even at contactpressures as low as 1 MPa, where the entropy ofmixing, not stress and temperature, drives the

104 3 APRIL 2015 • VOL 348 ISSUE 6230 sciencemag.org SCIENCE

Fig. 3.Tribofilm volu-metric growth ratedependence on tem-perature.Growth rate(mean T SD) versustemperature data fittedwith an exponentialfunction (Eq. 1). The2 mm × 2 mm topo-graphic contact-modeAFM images shownwere acquired at anonperturbative loadafter generating tribo-films in the central1.0 mm × 0.5 mmregions at 80°, 100°,120°, and 140°C at aninitial contact pressureof ~4.4 GPa.

RESEARCH | REPORTSon O

ctober 27, 2020

http://science.sciencemag.org/

Dow

nloaded from

reaction (19). The data are consistent with mole-cular dynamics simulations showing that tribofilmformation can be driven by contact pressure (8).However, we note that the simulation studies wereperformed on simpler zinc phosphate systems(with no sulfur), and effects of sliding were not in-vestigated. Here, we used sliding experiments toshow directly the role of pressure and temper-ature in forming tribofilms from ZDDP itself.Ex situ chemical analysisusingenergy-dispersive

spectroscopy (EDS) and Auger electron spec-troscopy (AES) identified the tribofilms’ elemen-tal composition. Point spectroscopy and elementalmapping by EDS (Fig. 4A) revealed clear sig-natures of Zn, S, and P inside the tribofilm, asexpected from ZDDP-derived products (32). Muchsmaller peaks corresponding to P, S, and Zn wereobserved outside the tribofilm region; these areattributable to the thin (~10 nm), weakly boundthermal film (a large fraction of which is likelydissolved during solvent rinsing before the EDSmeasurements). Elemental maps (Fig. 4A) revealuniform distributions of P, S, and Zn inside thetribofilm. The distribution of Fe was uniformand indistinguishable between regions insideand outside the tribofilm, further showing that

no measurable wear or displacement of Fe wasinvolved in tribofilm formation. AES, more surface-sensitive than EDS, revealed Zn, S, and P in thetribofilm region only (Fig. 4B). Far more Fe wasseen outside the tribofilm, indicating that littleor no Fe is mixed into the outermost regions ofthe tribofilm.The observed reduction of tribofilm robust-

ness with increased thickness is consistent withreports that the modulus and hardness of mac-roscopic ZDDP tribofilms decrease with thickness(20). Furthermore, the contact pressure depen-dence of tribofilm formation reported here (Fig. 2)can explain the reported gradient in composition,structure, and mechanical properties of ZDDPtribofilms. Specifically, because the tribofilm hasa lower modulus than the substrate, the contactstress at constant load decreases as the tribofilmthickens. This in turn reduces the amount of stress-induced polymerization and other reactions thatproduce the tribofilm, resulting in a weaker, morecompliant, graded structure and a further re-duction in contact pressure. This feedback-drivenself-limiting growth mechanism hinges on thestress dependence of the thermally activatedgrowth that we have uncovered (Fig. 2).

Our results show that theZDDP tribofilmgrowthrate increases exponentially with applied pressureand temperature under single-asperity contact, invery good agreement with stress-assisted reac-tion rate theory; the kinetic parameters are con-sistent with a covalent bond reaction pathway.Repeated sliding at sufficiently high loads leadsto abundant tribochemical reactions and the as-sociated nucleation and growth of robust tribofilmswith a pad-like structure similar to macroscopi-cally generated films. The tribofilm is not aproduct of theweakly adsorbed thermal film, butinstead is generated from molecular species fedcontinuously into the contact zone.We confirmedthe sacrificial nature of the tribofilm beyond athreshold thickness, indicating that layers grownat lower applied pressures are weaker. The ob-servations imply that ZDDP’s antiwear behaviorderives from mechanical protection provided bythe tribofilm, as opposed to corrosion inhibition.We suggest that this in situ approach can be di-rectly applied to understand further molecular-level tribochemical phenomena and functionality,such as the behavior of other important lubricantadditives (e.g., friction modifiers) or films formedin vapor-phase lubrication (33).

SCIENCE sciencemag.org 3 APRIL 2015 • VOL 348 ISSUE 6230 105

Fig. 4. Ex situ chemical characterization. (A) EDS point spectra (estimated sampling depth of ~1 mm) acquired for regions (a) inside and (b) outsidethe tribofilm (i.e., for the portion of the substrate covered with the thermal film). (c) Secondary electron image of the 10 mm × 5.0 mm tribofilm.Corresponding elemental maps are shown for (d) Fe, (e) Zn, (f) P, and (g) S. (B) (h) Optical and (i) secondary electron image of a 10 mm × 5.0 mm tribofilmobtained by scanning AES. ( j) AES spectra for the tribofilm and the substrate (estimated sampling depth ~3 nm).

RESEARCH | REPORTSon O

ctober 27, 2020

http://science.sciencemag.org/

Dow

nloaded from

REFERENCES AND NOTES

1. R. I. Taylor, Faraday Discuss. 156, 361–382 (2012).2. A. M. Barnes, K. D. Bartle, V. R. A. Thibon, Tribol. Int. 34,

389–395 (2001).3. H. Spikes, Tribol. Lett. 17, 469–489 (2004).4. E. S. Ferrari, K. J. Roberts, M. Sansone, D. Adams, Wear 236,

259–275 (1999).5. M. L. S. Fuller, M. Kasrai, G. M. Bancroft, K. Fyfe, K. H. Tan,

Tribol. Int. 31, 627–644 (1998).6. M. A. Nicholls et al., Tribol. Lett. 18, 261–278 (2005).7. B. Vengudusamy, J. H. Green, G. D. Lamb, H. A. Spikes, Tribol.

Int. 44, 165–174 (2011).8. N. J. Mosey, M. H. Müser, T. K. Woo, Science 307, 1612–1615

(2005).9. B. Vengudusamy, J. H. Green, G. D. Lamb, H. A. Spikes, Tribol.

Lett. 51, 469–478 (2013).10. S. Equey et al., Tribol. Int. 41, 1090–1096 (2008).11. G. C. Smith, J. Phys. D 33, R187–R197 (2000).12. H. Spikes, Lubr. Sci. 20, 103–136 (2008).13. K. T. Miklozic, J. Graham, H. Spikes, Tribol. Lett. 11, 71–81 (2001).14. K. T. Miklozic, T. R. Forbus, H. A. Spikes, Tribol. Trans. 50,

328–335 (2007).15. M. A. Nicholls, T. Do, P. R. Norton, M. Kasrai, G. M. Bancroft,

Tribol. Int. 38, 15–39 (2005).16. J. J. Habeeb, W. H. Stover, ASLE Trans. 30, 419–426 (1986).17. F. Rounds, Tribol. Trans. 36, 297–303 (1993).18. J. M. Martin, Tribol. Lett. 6, 1–8 (1999).19. J. M. Martin, T. Onodera, C. Minfray, F. Dassenoy, A. Miyamoto,

Faraday Discuss. 156, 311–323 (2012).20. S. Bec et al., Proc. R. Soc. London Ser. A 455, 4181–4203

(1999).21. S. Berkani et al., Tribol. Lett. 51, 489–498 (2013).22. D. Shakhvorostov et al., J. Chem. Phys. 128, 074706

(2008).23. See supplementary materials on Science Online.24. M. Aktary, M. T. McDermott, J. Torkelson, Wear 247, 172–179

(2001).25. H. Fujita, H. A. Spikes, Tribol. Trans. 48, 567–575

(2005).26. M. Aktary, M. T. McDermott, G. A. McAlpine, Tribol. Lett. 12,

155–162 (2002).27. H. Fujita, R. P. Glovnea, H. A. Spikes, Tribol. Trans. 48,

558–566 (2005).28. B. Kim, R. Mourhatch, P. B. Aswath, Wear 268, 579–591

(2010).29. M. Burkinshaw, A. Neville, A. Morina, M. Sutton, Tribol. Int. 46,

41–51 (2012).30. Y. R. Li, G. Pereira, M. Kasrai, P. R. Norton, Tribol. Lett. 29,

201–211 (2008).31. T. D. B. Jacobs, R. W. Carpick, Nat. Nanotechnol. 8, 108–112

(2013).32. A. Morina, A. Neville, J. Phys. D 40, 5476–5487 (2007).33. A. L. Barnette, D. B. Asay, J. A. Ohlhausen, M. T. Dugger,

S. H. Kim, Langmuir 26, 16299–16304 (2010).

ACKNOWLEDGMENTS

Supported by the Marie Curie International Outgoing Fellowshipfor Career Development within the 7th European CommunityFramework Programme under contract PIOF-GA-2012-328776(F.M.), the Nanotechnology Institute through the Ben FranklinTechnology Development Authority, the University of PennsylvaniaSchool of Engineering and Applied Sciences, and NSF grantCMMI-1200019. We thank the University of PennsylvaniaNano/Bio Interface Center Facilities and the NanoscaleCharacterization Facility in the Singh Center for Nanotechnologyfor use of facilities, Evans Analytical Group (East Windsor, NJ)for AES measurements, ExxonMobil’s Corporate StrategicResearch laboratory for materials and financial support, Q. Tamfor Matlab analysis, T. D. B. Jacobs for transmission electronmicroscopy analysis of AFM probes, and A. Jackson for helpfuldiscussions.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/348/6230/102/suppl/DC1Materials and MethodsSupplementary TextFigs. S1 to S10Movie S1References (34–42)

15 July 2014; accepted 27 February 2015Published online 12 March 2015;10.1126/science.1258788

FRUSTRATED MAGNETISM

Large thermal Hall conductivity ofneutral spin excitations in afrustrated quantum magnetMax Hirschberger,1 Jason W. Krizan,2 R. J. Cava,2 N. P. Ong1*

In frustrated quantum magnets, long-range magnetic order fails to develop despite a largeexchange coupling between the spins. In contrast to the magnons in conventional magnets,their spin excitations are poorly understood. Here, we show that the thermal Hallconductivity kxy provides a powerful probe of spin excitations in the “quantum spin ice”pyrochlore Tb2Ti2O7. The thermal Hall response is large, even though the material istransparent. The Hall response arises from spin excitations with specific characteristicsthat distinguish them from magnons. At low temperature (<1 kelvin), the thermalconductivity resembles that of a dirty metal. Using the Hall angle, we construct a phasediagram showing how the excitations are suppressed by a magnetic field.

Spin waves or magnons, the elementary ex-citations in a magnet, can transport heatwhen a thermal gradient −∇T is applied.Becausemagnons are not charged, the ther-mal current is expected to be symmetric

with respect to the sign-reversal of a magneticfield H, resulting in a zero thermal Hall con-ductivity kxy. However, recent findings based onthe Berry curvature have overturned this semi-classical result. Following a prediction by Katsura,Nagaosa, and Lee (1), Onose et al. recently ob-served a weak kxy signal in the ferromagnetic in-sulator Lu2V2O7 (2). Subsequently, it was pointedout (3) that kxy should include a contributionfrom the magnetization current (4–6).In frustrated magnets, long-range magnetic

order fails to develop even at millikelvin tem-peratures T because of strong quantum fluctua-tions. The low-lying excitations are currently ofgreat interest (7–9). The pyrochlore Tb2Ti2O7 isan insulator in which Tb and Ti define interpen-etrating networks of tetrahedra (10). Each Tb3+

ionhas a large localmoment [9.4 bohrmagnetons(mB)], but the lowest lying level is a crystal-fieldinduced spin-1/2 doublet. The spin-spin inter-action is of the antiferromagnetic Ising-type,with easy axis along the local (111) axis. Suscep-tibility experiments report a Curie-Weiss tem-perature q = –19 K (10, 11). However, long-rangeorder is not detected down to 50 mK (11, 12).Neutron scattering observed diffuse scatteringin zero H, which condensed to a Bragg peak at(002) when H = 2 T (13). Recently, neutron dif-fraction at 50 mK and H = 0 has detected elasticdiffusive peaks at (1/2, 1/2, 1/2), with short cor-relation length (x ∼ 8 Å) (14) and “pinch points”(15), which is suggestive of incipient spin-ice ordersubject to strong quantum fluctuations. Distinctfromclassical spin-icepyrochlores (suchasDy2Ti2O7)(16), the ground state of Tb2Ti2O7 and Yb2Ti2O7—

broadly termed “quantum spin ice” (7, 8, 17)—ispredicted to host the quantum spin liquid (at T =0) and the thermal spin-liquid (when entropy dom-inates at finite T), as well as a still unidentifiedCoulomb ferromagnetic state (17–19). All threestates harbor exotic excitations.In spite of the extensive experimental litera-

ture, nearly nothing is known about the trans-port properties of the spin excitations. Here, weshow that kxy provides a powerful way to detectthe excitations and determine their properties.We investigated two crystals, cut with the x-y

plane (largest face) normal to (110) in sample1 and normal to (111) in sample 2. With H || zand thermal current density Jq || x (Fig. 1), wemeasured the longitudinal gradient −∂xT andthe transverse gradient −∂yT to obtain the ther-mal resistivity tensorWij (defined by −∂iT =WijJqj)and the thermal conductivity tensor kij = Wij

−1.Throughout, we plot kij divided by T in order toremove the entropy factor. The thermal Hallangle is defined by tanqH = kxy/kxx =Wyx/Wxx. Formeasurements of −∂yT, three types of thermom-eterswereused: rutheniumoxide (T<15K), Cernox(10≤ T ≤ 40 K), and chromel-constantan thermo-couples (T ≥ 25 K). The magnetoresistances ofthe thermometers were extensively calibrated.At 1 K, we resolve 0.5 mK in the Hall signal for alongitudinal dxT ∼ 100 mK. The fragility of thecrystals above 5 T and the unusually largemagneto-caloric effect in Tb2Ti2O7 below 5 Kposed challenges that required specially designedmounts and measurement protocols (20).The observed Hall response in Tb2Ti2O7 be-

comes quite large below 15 K. Seeing a largethermal Hall signal in an orange, transparentcrystal is strongly counterintuitive (because trans-parency implies the complete absence of chargecarriers, the usual source of aHall current). Hence,we have performed several tests in order to verifythat it is intrinsic (20). The Hall signal (always“holelike”) should be independently antisym-metric inH and in Jq. To reverse Jq, we warmedup the sample and reconfigured the Au wires. The

106 3 APRIL 2015 • VOL 348 ISSUE 6230 sciencemag.org SCIENCE

1Department of Physics, Princeton University, Princeton, NJ08544, USA. 2Department of Chemistry, PrincetonUniversity, Princeton, NJ 08544, USA.*Corresponding author. E-mail: npo@princeton.edu

RESEARCH | REPORTSon O

ctober 27, 2020

http://science.sciencemag.org/

Dow

nloaded from

Mechanisms of antiwear tribofilm growth revealed in situ by single-asperity sliding contactsN. N. Gosvami, J. A. Bares, F. Mangolini, A. R. Konicek, D. G. Yablon and R. W. Carpick

originally published online March 12, 2015DOI: 10.1126/science.1258788 (6230), 102-106.348Science 

, this issue p. 102; see also p. 40Scienceadditives.Schwarz). Understanding these mechanisms will help in the development of higher-performance and more effectiveexactly how ZDDP produces an anti-wear film under high stress or elevated temperature (see the Perspective by

examinedet al.finding a replacement for ZDDP: Its breakdown products shorten catalytic converter lifetime. Gosvami dialkydithiophosphate (ZDDP) has been used for decades to reduce engine wear. Now there is a strong incentive for

Additives in oil are vital for protecting engines from wear by forming films at sliding interfaces. ZincAdditive explanation for anti-wear

ARTICLE TOOLS http://science.sciencemag.org/content/348/6230/102

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2015/03/11/science.1258788.DC1

CONTENTRELATED http://science.sciencemag.org/content/sci/348/6230/40.full

REFERENCES

http://science.sciencemag.org/content/348/6230/102#BIBLThis article cites 39 articles, 1 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Copyright © 2015, American Association for the Advancement of Science

on October 27, 2020

http://science.sciencem

ag.org/D

ownloaded from