The Pennsylvania State University The Graduate School

MECHANOCHEMISTRY OF ADSORBED MOLECULES AT

TRIBOLOGICAL INTERFACES

A Dissertation in

Chemical Engineering

by

Xin He

 2019 Xin He

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

December 2019 ii

The dissertation of Xin He was reviewed and approved* by the following:

Seong H. Kim Professor of Chemical Engineering and Materials Science Dissertation Advisor Chair of Committee Graduate Program Coordinator

Adri C.T. van Duin Professor of Mechanical Engineering Professor of Chemical Engineering Professor of Engineering Science and Mechanics

Themis Matsoukas Professor of Chemical Engineering

Xueyi Zhang John J. and Jean M. Brennan Clean Energy Early Career Assistant Professor of Chemical Engineering

*Signatures are on file in the Graduate School iii

Abstract

Mechanochemistry refers to reactions activated by mechanical force or stress. It is ubiquitous in engineering systems as well as daily life; but poorly understood because the reactions are taken place at solid tribological interfaces. The main challenges in mechanochemistry research are to identify the reactants and products at sliding interfaces, which can be overcome via vapor phase lubrication (VPL). The main questions in this work are the role of shear force or stress in mechanochemical reaction and how surface chemistry properties affect the reaction rate.

This study investigated the molecular structure dependence of precursor molecules during mechanochemical reactions. Several monomers were studied to reveal the molecular structure dependence at tribological interfaces, including -pinene (C10H16), pinane (C10H18) and n-decane

(C10H18), which are all hydrocarbon precursors containing 10 carbon atoms. The friction coefficient of these molecules was around ~0.2 and they can be polymerized by mechanical shear.

The molecules with ring structure (-pinene and pinane) were found to produce more products compared with linear molecule n-decane. The modified Arrhenius-type equation is used to study the relationship between tribopolymerization yield and applied load; as well investigate how critical activation volume (V*) is affected with the structure of monomers. The estimated V* of

-pinene and pinane, which possess higher internal strain, showed the lowest activation volume, while the value is higher for n-decane (10%), which possesses low internal strain.

The experimental results were then compared with molecular dynamic (MD) simulations with a ReaxFF reactive force field to reveal the physical meaning of activation volume. The tribo- polymerizable model reactants, allyl alcohol and -pinene were studied. The results suggested that the precursor molecules first chemisorb on the surfaces through surface oxygen. During the sliding iv process, the precursors undergo partial distortion from its equilibrium conformation, corresponding to the critical activation volume for mechanochemical reactions. The activated intermediates polymerize through the formation of C-O-C ether bond, rather than direct C-C covalent bond.

The effect of surface chemistry and the surrounding gas environment on mechanochemical polymerization reactions are also revealed. Mechanochemical reactions of -pinene are studied on eight substrate materials – chemically reactive group (palladium, nickel, copper, stainless steel) and chemically inert group (gold, silicon oxide, aluminum oxide, DLC) The more reactive surfaces appear to be capable of chemisorbing -pinene molecules even without the mechanical shear process. In the case of reactive surfaces, the critical activation volumes are relatively lower compared to that only capable of physisorption. Such chemisorption process can be significantly enhanced through the oxidative gas environment, in turn increasing tribochemical reactivity. When the water is introduced to the allyl alcohol system at silicon oxide tribological interfaces, it enhances the tribo-polymerization activity, in turn increasing the yield rate of tribo-polymers. The estimated activation energy is found lower with the presence of water molecules.

The tribochemistry is widely-investigated in anti-wear tribofilm formation.

Environmentally-friendly ionic liquids (ILs) are being developed as ashless additives for hydraulic fluids. Candidate ILs, at a treat rate of 0.5 wt.%, were blended into a non-polar mineral base oil, a hydrophilic Polyalkylene glycol (PAG), and an oil-soluble PAG (OSP). Compared with a commercial primary zinc dithiophosphate (ZDDP), the ILs showed lower friction coefficient and wear volume. This attributes to the formation of a protective layer on the contact surface as revealed by characterization of wear scar morphology and composition. In addition to the superior v lubricating performance, these ILs have advantages of higher thermal stability and lower toxicity than commercial hydraulic fluid additives.

vi

Table of Contents

List of Figures ...... x

List of Tables ...... xxi

List of Equations ...... xxii

Acknowledgements ...... xxiv

Chapter 1 ...... 1

Introduction and Motivation ...... 1

Chapter 2 ...... 6

Experiment details and data processing ...... 6

Vapor phase lubrication tests ...... 6 Sample preparation ...... 11 Calculation of flash temperature ...... 12 Tribo-polymer yield calculation ...... 13 Mechanical behavior tests ...... 16 Adsorption isotherm measurement methods and principles ...... 19 Chemical Characterization ...... 21 Molecular Dynamics simulation ...... 22 Chapter 3 ...... 23

Tribochemical synthesis of nano-lubricant films from adsorbed molecules at sliding solid interface: Tribo-polymers from α-pinene, pinane, and n-decane ...... 23

Overview ...... 23 Introduction ...... 24 Experimental Details ...... 26 Result and Discussion ...... 28 Conclusion ...... 44 Supporting Information ...... 46 Chapter 4 ...... 51 vii

Mechanochemistry of physisorbed molecules at tribological interfaces: ...... 51

Molecular structure dependence of tribochemical polymerization ...... 51

Overview ...... 51 Introduction ...... 53 Experiment Details...... 55 Results and Discussion ...... 57 Conclusion ...... 69 Supporting Information ...... 70 Chapter 5 ...... 72

Mechanochemistry at solid surfaces: Polymerization of adsorbed molecules by mechanical shear at tribological interfaces ...... 72

Overview ...... 72 Introduction ...... 74 Experimental Details ...... 77 Result and Discussion ...... 80 Conclusion ...... 90 Supporting Information ...... 91 Chapter 6 ...... 99

Mechanochemical association reaction of interfacial molecules driven by shear ...... 99

Overview ...... 99 Introduction ...... 101 Methods...... 102 Results and Discussion ...... 105 Conclusion ...... 114 Supporting Information ...... 116 Chapter 7 ...... 131 viii

Surface chemistry dependence of mechanochemical reaction of adsorbed molecule – An experimental study on tribopolymerization of -pinene on metal, metal oxide, and carbon surfaces ...... 131

Overview ...... 131 Introduction ...... 133 Experiment Details...... 136 Results and Discussion ...... 139 Conclusion ...... 152 Supporting Information ...... 153 Chapter 8 ...... 156

Effect of gas environment on mechanochemical reaction: A model study with tribo- polymerization of -pinene in inert, oxidative, and reductive gases ...... 156

Overview ...... 156 Introduction ...... 157 Experiment Details...... 159 Results and Discussion ...... 162 Conclusion ...... 172 Supporting Information ...... 173 Chapter 9 ...... 175

Mechanochemical reactions of adsorbates at tribological interfaces: ...... 175

Tribo-polymerizations of allyl alcohol co-adsorbed with water on silicon oxide ...... 175

Overview ...... 175 Introduction ...... 176 Experiment Details...... 179 Results and Discussion ...... 182 Conclusion ...... 192 Supporting Information ...... 193 Chapter 10 ...... 194 ix

Low-Toxicity Ionic liquids as environmentally-friendly lubricant additives ...... 194

Overview ...... 194 Introduction ...... 196 Experiment details ...... 199 Results and Discussion ...... 203 Conclusion ...... 212 Supporting Information ...... 213 Appendix ...... 214

Contact Mechanics versus Amonton’s Law in Macroscale Friction ...... 214

– Experimental and Computational Study ...... 214

Overview ...... 214 Introduction ...... 216 Experiment Details...... 218 Results and Discussion ...... 220 Conclusion ...... 229 Reference ...... 230

x

List of Figures

Figure 1- 1 The reaction coordinate diagrams of the reaction without mechanical force (solid) and

the reaction under mechanical force (dashed) ...... 2

Figure 2- 1 Tribometer set and digital multimeter (DMM) ...... 7

Figure 2- 2 Excel output file of ‘avg’ file ...... 7

Figure 2- 3 Schematic of Vapor flow system for environmental control ...... 10

Figure 2- 4 (a) Schematic of vapor diffusion inside the bubble (b) relationship between diffusion time and radius of bubble ...... 10

Figure 2- 5 NanoScope Analysis interface for tribo-polymer yield calculation ...... 15

Figure 2- 6 Cross-section profile (left) and 3D image (right) obtained from optical profilometry software ...... 15

Figure 2- 7 Example for force-sample deformation (F-D) curve...... 18

Figure 2- 8 Example for JKR fitting ...... 18

Figure 2- 9 Schematic illustration of attenuate total reflection (ATR) and polarization modulation‐infrared reflection‐adsorption spectroscopy (PM‐IRRAS) spectroscopy ...... 19

Figure 3- 1 (a) Friction coefficient of stainless steel in dry N2 and VPL with 30% p/푝푠푎푡 α-pinene, pinane, and n-decane; (b) Cross-section line profile (top), optical profilometry image (middle) and optical microscope image (bottom) of wear tracks after friction tests in dry N2 as well as VPL with α-pinene, pinane, and n-decane. In (b), tribo-polymers were removed by ethanol to analyze the damage in the substrate surface. The unit of the y-axis of the cross-section line profile is m...... 30

Figure 3- 2 ToF-SIMS ion maps (a,c,e) and mass spectra (b,d,f) and of tribo-polymer products deposited at the end of the slide track produced during (a,b) -pinene VPL, (c,d) pinane - pinene, (e,f) n-decane VPL. In the ion maps, blue is the iron ion (m/e=55.93 amu), red is the chromium ion (m/e=51.94 amu), and green is the total intensity of ions with m/e = 164~300 amu. The mass of ions marked with * in the inset of (b) are 216, 228, and 244 amu...... 32 xi

Figure 3- 3 Infrared spectra of vapor, adsorbate, and tribo-polymer of (a) α-pinene and (b) pinane. The vapor spectra were the sum signal of PM-RAIRS measurements; the adsorbate spectra were the difference signal of PM-RAIRS normalized with the sum signal. The tribo- polymer spectra were obtained from the deposits within a 50  50 m2 area at the end of the sliding track...... 34

Figure 3- 4 Tapping-mode AFM images and line profiles of tribo-polymers piled at the ends and sides of the slide track after VPL tests in (a) α-pinene, (b) pinane, and (c) n-decane vapors. The line profile was taken along the sliding direction at the left end of the slide track. Each AFM image was taken over a 90 m  90m area. Indentation test positions were marked as green cross mark on the AFM images and arrow on the cross-section profiles...... 39

Figure 3- 5 Force-distance curve of (a) α-pinene tribo-polymers, (b) pinane tribo-polymers and (c) n-decane tribo-polymers accumulated at the end of slide track on stainless steel. Each tribo- polymer was probed with a new clean AFM tip. The piezo-scanner travel distance was converted to the tip-sample distance by subtracting the cantilever deflection distance. The tips was used for tapping-mode imaging before the force-distance measurements...... 40

Figure 3- 6 Lubrication effect of the tribo-polymer films in the absence of vapor supply. The tribo- polymer film was synthesized by rubbing stainless steel surfaces for 1200 cycles in the presence of 30% p/푝푠푎푡 α-pinene (black), pinane (red), or n-decane (blue) and then friction was measured continuously in dry N2 after stop supplying vapors. The 30% p/푝푠푎푡 n- pentanol (orange) is also tested for 1200 cycles and then measured in dry N2...... 44

Figure 3-S 1 PM-RAIRS spectra of (a) α-pinene, (b) pinane and (c) n-decane adsorbed on a metallic substrate. The insets show the adsorbate signal intensity as a function of relative partial pressure of the vapor...... 47

Figure 3-S 2 Raw indentation test data (force-distance curves) of (a) α-pinene tribo-polymer (b) pinane tribo-polymer (c) n-decane tribo-polymer (d) stainless steel surface with clean tip (e) stainless steel surface used tip. In (a), the black lines show the extrapolations of the linear portion of the loading curve and the free-standing line. The cross point of these extrapolated lines was assumed to be zero indentation position in the tip-sample distance calculation...... 48

Figure 3-S 3 JKR model fit of the unloading curves of (a,b) α-pinene tribo-polymer, (c,d) pinane tribo-polymer, and (e,f) n-decane tribo-polymer...... 50

Figure 4- 1 Coefficient of friction of 440C stainless steel measured in dry N2 and nitrogen containing -pinene, pinane, and n-decane vapor at marked contact loads. The partial pressure for the organic vapor was 30% of its saturation value (푝/푝푠푎푡). Error bars represents standard deviation calculated from more than 3 measurements. The inset shows the line profile across the wear track (left), optical profilometry image (center), optical microscope image (right) of the wear track made in dry N2 tribo-testing...... 58 xii

Figure 4- 2 Cross-section line profile (left), optical profilometry image (middle), and optical microscope image (right) of sliding tracks after friction tests at the lowest and highest Hertzian contact pressure ...... 60 Figure 4- 3 Tapping-mode AFM images of tribo-polymers accumulated at the ends and sides of the slide track after 600 cycles of reciprocating slide at the lowest and highest contact loads tested in (a) α-pinene, (b) pinane, and (c) n-decane vapors (30% p/psat). The area and height scales of the image are shown above each panel of images...... 60 Figure 4- 4 Comparison of adsorption isotherm (black) from PM-RAIRS measurements and total yield of tribo-polymer (red) after VPL tests at various vapor pressures of (a) α-pinene (b) pinane and (c) n-decane. The applied contact load was 0.5 GPa and the total number of sliding cycles was 600...... 62 Figure 4- 5 Load dependence of tribo-polymer yield for α-pinene (black), pinane (blue), and n- decane (red). The p/psat of each vapor was kept constant at 30%...... 63 Figure 4- 6 (a) Lubrication effect of the tribo-polymer films after the lubricant vapor supply is stopped. (b) Extension of the lubrication effect by dragging the tribo-polymer piled outside the slide track into the slide track. (c) Maintaining the lubrication effect by periodically replenishing tribo-polymers. The tribo-polymer film synthesis (marked with yellow background) was carried out by rubbing stainless steel surfaces for 600 cycles in the presence of 30% p/psat α-pinene. The applied load during the entire friction test was 0.5 GPa...... 67

Figure 4-S 1 AFM force-distance curve of 훼-pinene tribo-polymer accumulated at the end of the slide track in (a) dry nitrogen, (b) ambient air, (c) 30% p/psat of 훼-pinene vapor. The piezo displacement was converted to the sample indentation depth...... 70

Figure 5- 1 Snapshot of the MD simulation of allyl alcohol sliding between amorphous SiO2 slabs. The colors of the spheres represent atom types: tan = silicon, red = oxygen, green = carbon, and white = hydrogen. 78

Figure 5- 2 (a) Semi-log plot of the tribo-polymer yield (ry, normalized with the contact area and sliding time) versus applied contact pressure. The synthesis condition was as follows: sliding speed = 0.4 cm/s; sliding span and cycle = 2.5mm and 600 cycles; allyl alcohol vapor = 30% p/psat; temperature = 295 K. Insets are AFM images of the left, middle, and right regions of a slide track tested at a contact pressure of 0.45 GPa, the mass spectrum of tribo-products from ToF-SIMS analysis, and the selected ion map from ToF-SIMS imaging. (b) Semi-log plot of the tribo-polymer production rate (rp, calculated from the number of molecules containing more than 3 carbon atoms) from MD simulations versus the normal contact pressure. The simulation condition was as follows: sliding speed and time = 10 m/s and 1 ns; number of allyl alcohol molecules = 75; temperature = 300 K. Inset images show the structures of an allyl alcohol molecule and an intermediate forming a dimer...... 81

Figure 5- 3 Snapshot images from the ReaxFF simulation showing the association paths of (a) two and (b) three allyl alcohol molecules during the slide at the SiO2/SiO2 interface (contact xiii

pressure = 1 GPa; slide speed = 10 m/s; temperature = 300K). Only the atoms involved in mechanochemical reactions are highlighted and numbered for clarification purpose...... 84

Figure 5- 4 Radial distribution functions (RDFs) for distances between carbon atoms (C1, C2, C3; numbered from OH group, shown as inset to Figure 5-4a), alcoholic oxygen atoms (Oa), surface oxygen atoms (Os), and silicon atoms (Si) at the interface of silicon oxide. Red curves are at 300K without load or sliding. Black curves are during the slide (contact load = 1 GPa; slide speed = 10 m/s; temperature = 300K). Green curves are at 900K without load or sliding. Each RDF is for a different atom-atom distance: (a) C1-Oa, (b) C1-Oa, (c) C1-Oa, (d) C1-Os, (e) C2-Os, (f) C3-Os, (g) C2-C3, (h) Oa-Oa, and (i) Si-Oa...... 87

Figure 5-S 1 (a) Friction coefficients measured in dry N2 and allyl alcohol vapor (p/psat = 40 %). Optical profilometry line profile and optical image of (b) wear track after tribo-test in dry N2 and (c) slide track after tribo-test in allyl alcohol vapor and removal of tribo-polymers by rinsing with ethanol...... 91

Figure 5-S 2 XPS spectrum of tribo-polymer film piled at the end of sliding tracks. Tribo-polymers were synthesized by sliding a glass bead on a thermal oxide film on a silicon wafer at 0.4 GPa for 600 reciprocating cycles in 30% p/psat allyl alcohol vapor...... 94

Figure 5-S 3 Comparison of ReaxFF MD and DFT simulation results for reactions forming (a) the single bond between the carbon atoms of two C=C double bonds and (b) the covalent bond between hydroxyl oxygen (Oa) and one of the carbons in the C=C double bond with proton donation from Oa to the surface oxygen atom (Os)...... 96

Figure 5-S 4 Temporal profile of reactants, association products (x > 3), chemisorbed or partially dissociated species (x=3 or x<3) at a contact pressure of (a) 0.5 GPa, (b) 1 GPa, and (c) 2 GPa. The black line is the result from fitting with a (1 − 푒 − 푟푝푡) function. Here, x is the number of carbon atoms in the molecule or intermediate species...... 97

Figure 5-S 5 Snapshot images of (a) an intermediate with the C3 (dark green) and Oa (dark red) atoms at 1.47 Å from each other and (b) an intermediate with the Oa atom (purple) positioned at 2.1 Å from the under-coordinated Si atom (large gray)...... 97

Figure 6- 1 The configuration of the model system illustrating the three stages of the MD simulations. From left to right: energy minimization, compression, and sliding...... 105

Figure 6- 2 (a) Friction coefficients of hydroxylated and dehydroxylated surfaces in -pinene VPL condition at a 40 % partial pressure relative to saturation. A borosilicate ball was used as a counter-surface. The applied Hertzian contact pressure was 0.32 GPa and the sliding speed was 4 mm/s. Insets are AFM images of reaction products piled in and along the slide tracks. (b) C1s XPS spectra of the shear-induced polymers produced on silicon oxide surfaces with two different surface conditions. (c) Semi-log plot of the normalized tribo-polymer yield xiv

against the contact pressure for -pinene sheared on the dehydroxylated silicon oxide surface...... 106

Figure 6- 3 Histograms of the molecular weights of species with ten or more carbon atoms from simulations at 1 GPa for (a) hydroxylated surface without sliding, (b) non-hydroxylated surface without sliding, (c) hydroxylated surface after 2ns of sliding, and (d) non- hydroxylated surface after 2ns of sliding. Side view images of the MD simulations are shown as insets in (a) and (b), where reactive surface sites are highlighted by red spheres. Snapshots of representative shear-induced reaction products are shown as insets to (c) and (d)...... 108

Figure 6- 4 Snapshots at various times from ReaxFF-MD simulations following one -pinene molecule. The numbers on atoms are given following the IUPAC nomenclature. The 4- membered ring consists of C1, C2, C7, and C6. The 6-membered ring is made of C2, C3, C4, C5, C6, and C7. The C=C double bond exists between C3 and C4. The C8, C9, and C10 are methyl side groups. The oxygen atoms are from the silicon oxide substrate and numbered following the order of covalent bond formation to -pinene molecules. Atom colors correspond to: C-brown, O-blue, and H-yellow...... 110

Figure 6- 5 Physical deformation of the 4-membered ring during the oxidation of -pinene by reaction with surface oxygen atoms (steps shown in Figures 4b, 4c, and 4d). Here, 1, 2, 7, and 6 represent the C6-C1-C2, C1-C2-C7, C2-C7-C6, and C7-C7-C1 bond angles, respectively. The volume of a tetrahedral box (marked in dotted lines) is also shown. . 113

Figure 6-S 1 Activation of the hydroxylated surface by friction at a vapor pressure of -pinene insufficient for monolayer lubrication; when -pinene vapor pressure was subsequently increased to p/psat = 40% (sufficient enough for monolayer coverage), shear-induced mechanochemical polymerization occurred more readily compared to the fully hydroxylated original surface...... 116

Figure 6-S 2 Evidence of dehydroxylation of the UV/O3-cleaned hydroxylated silica surface upon thermal annealing at 450 oC for 12 hours. After thermal dehydroxylation, the water contact angle increased from <5o to 50~55o...... 117

Figure 6-S 3 (a) Cristobalite SiO2, (b) changes in temperature and potential energy during the heating and cooling cycle used to create the amorphous structure, and (c) final structure of the amorphous silica...... 118 Figure 6-S 4 (a) Model system for simulation of hydroxylation. (b) Density of silanol groups on the silica surface during the hydroxylation process...... 119 Figure 6-S 5 Density of different functional groups on the non-hydroxylated and hydroxylated model silica surfaces...... 120 Figure 6-S 6 (a) Average molecular weight of reaction products containing more than 10 carbon atoms in MD simulations for the hydroxylated and non-hydroxylated surfaces at contact xv

pressures of 1 GPa and 3 GPa. (b) Calculation of the critical activation volume based on the number of -pinene molecules associated in simulations at 1GPa and 3GPa conditions...... 121 Figure 6-S 7 Density of chemisorbed -pinene molecules on the hydroxylated and non- hydroxylated surfaces at the end of the energy equilibrium step in MD simulations..... 123 Figure 6-S 8 Snapshots of the hydrocarbon species formed in MD simulations with the hydroxylated and non-hydroxylated silica surfaces after 2 ns sliding. The atom numbers at the connection between molecules are provided adjacent to each snapshot. It is important to note that all association products are connected through the C-O-C ether bond; no products were found to be connected via direct C-C covalent bonding...... 124 Figure 6-S 9 Energetics of the oxidative activation (O1  C6) coordinate along which a slightly deformed -pinene molecule (red line) reacts with a dangling oxygen of the silica surface (as shown in Figures 4b and 4c in the main text). The surface oxygen atom was modeled as an isolated (HO)3Si-O species for simplification in the energy calculation. The activation barrier is lower compared to the same reaction of an undeformed -pinene in its equilibrium structure (blue line)...... 125 Figure 6-S 10 Comparison of the energetics for an oxidative activation step where a surface dangling oxygen (O1) is reacting with the one of the carbons in the double bond (C4) of -pinene. The red and blue lines correspond to a slightly deformed and undeformed - pinene molecule, respectively...... 127 Figure 6-S 11 Energetics for the reaction between an oxygen radical at the C4 position of one - pinene and the C3 carbon of another -pinene. In the case of the deformed -pinene intermediate (red curve), the initial state is so unstable and reactive that there is no activation barrier. In contrast, the fully relaxed intermediate (blue curve) shows an activation barrier for the same reaction...... 128 Figure 6-S 12 Energetics for the reaction between a hydroxyl group at the C4 position of one - pinene and the C3 carbon of another -pinene. In this scheme, the reaction proceeds first through donation of hydrogen from C3-O-H to a surface O-Si(OH)3 species followed by formation of C3-O-C4 ether bond formation. The activation energy is slightly lower (by ~8 kcal/mol) for the deformed intermediate (red line), compared to the fully relaxed intermediate (blue line)...... 129 Figure 6-S 13 Regio-selectivity in oxidative activation of carbon atoms in reaction products containing more than ten carbon atoms in MD simulations at a contact pressure of 1 GPa for (a) hydroxylated and (b) non-hydroxylated silica surfaces...... 130

Figure 7- 1 Friction coefficients measured during the sliding of a silicon nitride ball (diameter = 3/32 inch) against native oxide of 440C stainless steel (black), nickel (red), copper (blue), palladium (purple), gold (orange), silicon wafer (magenta), aluminum oxide (olive), and DLC (green) in (a) dry nitrogen, (b) 40% p/psat n-pentanol vapor and (c) 40% p/psat α- pinene vapor. The nominal Hertzian contact pressure calculated using the bulk mechanical property was ~0.5 GPa...... 139 xvi

Figure 7- 2 Tapping-mode AFM images of tribopolymers piled up at the ends and sides of the sliding track after 600 reciprocating cycles of tribo-test with a silicon nitride ball at a Hertzian contact pressure of 0.5 GPa in α-pinene vapor (p/psat = 40%) condition. The scan area was 70 μm × 70 μm; the height full scale is 1 m for the images in the left side and 0.2 m for the images in the right side...... 141 Figure 7- 3 Total yield of the tribopolymerization products calculated by integrating the positive volume in AFM images. The total product volume (m3) was normalized with the sliding track area (m2) and the total sliding time (s). Except the DLC sample, the average and standard deviation were calculated from the multiple tests for 4  5 times for each material. Due to the difficulty of finding the wearless slide track with very little tribopolymer products, only one successful measurement of the product yield was done for the DLC sample...... 142 Figure 7- 4 Semi-log plot of the contact pressure dependence of the tribopolymerization yield for α-pinene on copper (black), nickel (red), stainless steel (green), and silicon oxide (blue) substrates. The partial pressure of α-pinene vapor was 40% p/psat and the total reciprocating cycle of the silicon nitride ball was 600. All data points are the average from 3 repeats (some were 4 repeats); for some data points, the error bars are smaller than the symbol size...... 144 Figure 7- 5 Infrared spectra of α-pinene adsorbed on (a) palladium, (b) copper, (c) gold and (d) silicon substrates at p/psat = 30%, 60% and 90%. The chemical formula in bracket is the surface composition. Also shown is the spectra taken after purging the -pinene vapor with N2 after the 90% p/psat measurement. The spectra in (a-c) were taking with PM-RAIRS and the spectra in (d) were obtained with ATR-IR...... 147 Figure 7- 6 IR spectra of tribopolymers piled up at the ends of the track on selected substrates after 600 reciprocating cycles of tribo-test with a silicon nitride ball in α-pinene vapor (p/psat = 40%) condition. Also shown is the IR spectrum of -pinene vapor measured during the PM-RAIRS experiments of the data shown in Figure 7-5...... 149 Figure 7- 7 Schematic energy diagram along the reaction coordinate when (a) interfacial molecules are chemisorbed and then mechanically activated for reaction, (b) interfacial molecules undergo mechanically-assisted chemisorption on reactive surfaces, and (c) solid surfaces are not reactive and physisorbed molecules are activated by mechanical shear. Echemads is the activation energy for chemisorption. Echemtherm and Ephystherm are the activation energy of chemisorbed (or during chemisorption) and physisorbed species, respectively, in the absence of mechanical effects. The reaction pathways or mechanisms might be different in three cases. Echemmech and Ephysmech are the amount of decrease in the activation energy (associated with Em = σ ∙ ∆V ∗) of chemisorbed and physisorbed species, respectively, upon the application of mechanical shear. The Eaeff term is the net activation energy governing the reaction yield or rate of mechanochemical reactions. Note that it is still larger for the physisorption case in (c) compared to the chemisorption case in (a) and (b)...... 150

xvii

Figure 7-S 1 Cross-section line profile (top), optical profilometry image (middle), and optical microscope image (bottom) of sliding tracks after friction tests under 0.5 GPa in 40% p/psat n-pentanol...... 153 Figure 7-S 2 Cross-section line profile (top), optical profilometry image (middle), and optical microscope image (bottom) of sliding tracks after friction tests under 0.5 GPa in 40% p/psat α-pinene. The substrate were cleaned with ethanol after the friction test to remove the residue tribo-polymers to evaluate substrate deformation...... 154 Figure 7-S 3 Adsorption isotherm of α-pinene on metal or metal oxide from PM-RAIRS measurements and on silicon oxide from ATR-IR measurement ...... 155

Figure 8- 1 The friction coefficient between silicon nitride ball and native oxide of (a) diamond- like carbon (DLC), (b) silicon wafer, (c) palladium and (d) copper in 40% p/psat α-pinene vapor in inert (dry nitrogen; red color), oxidative (dry air; green color), and reductive (hydrogen in Ar; blue color) carrier gases. In dry case, the friction coefficient would be high and unstable (0.79±0.15)...... 162 Figure 8- 2 Tapping-mode AFM images of -pinene tribo-polymers accumulated at both ends and in the middle of the slide track after 400 cycles of reciprocating sliding with the silicon nitride ball at a Hertzian contact pressure of 0.5 GPa in oxidative (top), inert (middle) and reductive (bottom) gases containing 40% p/psat -pinene vapor. The image size is 70 μm × 70 μm and the height full scale is 2 m. The cross-section line profile (white line) of the right end is plotted in the right side of images...... 164 Figure 8- 3 Tribo-polymerization yield of α-pinene on highly-reactive (Pd and CuO) and relatively-inert (DLC and SiO2) in oxidative (green), inert (red), and reductive (blue) gas environments. The normalized (μm/s) was calculated by dividing the total volume of tribo- product (μm3) with the sliding track area (μm2) and total sliding time (s). The average and standard error were obtained from measurements in the same reaction condition for 3-5 times...... 166 Figure 8- 4 PM-RAIRS spectra of α-pinene adsorbed on (a,b) Pd and (c,d) CuO (native oxide on Cu) surfaces at p/psat = 15% (black), 30% (red), 50% (blue), and 75% (pink), and then after purging out -pinene (green) in (a,c) oxidative and (b,d) reductive gas environments. The insets plot the total peak area...... 167 Figure 8- 5. IR spectra of tribo-polymers of -pinene on (a) Pd and (b) CuO substrates produced in oxidative (green), inert (red), and reductive (blue) gas environments. The micro-IR analysis was carried out for tribo-polymers accumulated at the end of the slide track after 400 cycles of reciprocating sliding with silicon nitride balls in 40% p/psat of -pinene vapor. The spectral intensity is normalized with the C-H stretch band for comparison purpose...... 170 Figure 8- 6 Force-distance (F-D) curve of α-pinene tribopolymers accumulated at the end of the sliding track on (a,b) Pd and (c,d) CuO in (a,c) oxidative and (b,d) inert gases. The elastic modulus estimated from the JKR fit of the retraction curve is shown in each panel. .... 171

xviii

Figure 8-S 1 optical profilometry image (left) and cross-section profile (right) of sliding tracks after friction tests at the 0.5 GPa Hertzian contact pressures in 40% p/psat α-pinene in dry air, nitrogen and hydrogen (10% in Ar). The substrates were cleaned with ethanol after the friction test to evaluate substrate damage without interference from triboreaction products...... 173 Figure 8-S 2 JKR model fit of the unloading curves of (a,c) α-pinene tribo-polymer created in dry air carrier gas on Pd and CuO, respectively; (b,d) α-pinene tribo-polymer created in dry nitrogen carrier gas on Pd and CuO, respectively...... 174

Figure 9- 1 Water-allyl alcohol vapor/liquid equilibrium curve. The blue dotted line represented bubble point curve while the red dotted represented dew point curve. The mixed liquid compositions were based on the black dots on figure...... 180

Figure 9- 2 Infrared spectra of adsorbates on silicon oxide in (a) water rich mixture (yalcohol =0.02), (b) alcohol rich mixture (yalcohol =0.9) and (c) allyl alcohol vapor at p/psat =70%. The read spectra is taken after purging the ATR chamber with N2 after the water-allyl alcohol mixture measurement. The ratio of C-H area of adsorbates to the residue is displayed in the yellow box...... 183 Figure 9- 3 Loadings (component weights) of allyl alcohol and water obtained from TFA of allyl alcohol and water mixtures, and plotted as a function of allyl alcohol mole fraction. ... 184 Figure 9- 4 Friction coefficient measured during the sliding of borosilicate glass against silicon wafer in 70% p/psat of several vapor mixture. The liquid used to create the vapor contains allyl alcohol with mole fraction from zero to one, as showed on the right side of friction map. The Hertzian contact pressure keeps constant at 0.34 GPa...... 185 Figure 9- 5 Semilog plot of the contact pressure dependence of the tribopolymerization yield for water/allyl alcohol mixture at yalcohol = 0.02 (blue) and pure allyl alcohol vapor (black; obtained from previous work). The total reciprocating cycle of the borosilicate ball was 400...... 187 Figure 9- 6 Tapping-mode AFM images of tribo-polymers created at both ends and in the middle of the slide track after 400 cycles of reciprocating sliding under a Hertzian contract pressure of 0.34 GPa in allyl alcohol/water vapor created from several liquid mixtures. The image size is 70 μm × 70 μm and height size is -2 μm ~ 2 μm...... 187 Figure 9- 7 Tribo-polymerization yield (red) at wear volume (black) on silicon wafer in vapor environment created from liquid mixture with different allyl alcohol mole fraction. .... 190 Figure 9- 8 IR spectra of tribo-polymers piled up at the ends of the track on silicon wafer after 400 reciprocating cycles of tribo test under water/allyl alcohol vapor mixture (yalcohol= 0.9, blue; yalcohol= 0.02; red) and pure allyl alcohol vapor(black) conditions...... 191

Figure 9-S 1 Surface topography after tribo-test in allyl alcohol- water mixture vapor environment ...... 193 xix

Figure 9-S 2 Target (test) and predicted spectra of allyl alcohol and water used in TFA of liquid mixtures...... 193

Figure 10- 1 Molecular structure of base oils and ionic liquids (ILs) ...... 198 Figure 10- 2 TGA curves of base oils (PAG, OSP, mineral blends), and oil-miscible ILs ...... 204 Figure 10- 3 Friction behavior of PAG (a), OSP (b) and mineral blends (c) with ZDDP and ILs. The stabilized friction coefficient of last 200 m were plotted in (d). The wear rate on 52100 bearing ball surface is shown in (e) ...... 204 Figure 10- 4 SEM surface morphology of wear scar on 52100 steel ball surface ...... 206 Figure 10- 5 XPS core-level spectra of P 2p and O 1s on worn 52100 steel ball surface lubricated with 5 wt% ZDDP in PAG, OSP and Mineral blends base oils respectively. (g) ~ (i) show the depth profile ...... 209 Figure 10- 6 XPS core-level spectra of P 2p and O 1s on worn 52100 steel ball surface lubricated with 5 wt% IL-1 in PAG, OSP and Mineral blends base oils respectively. (g) ~ (i) show the depth profile ...... 210

Figure 10-S 1 EDS spectra of IL-1 and ZDDP additized tribo-films in selected base oil ...... 213

Figure A- 1 the schematic of ideal and non-ideal contact ...... 218 Figure A- 2 the average friction force with respect to surface roughness and normal load at 440C stainless steel interfaces. The data on perfectly smooth surface (c) & (d) were obtained from extrapolating the friction force measured on surfaces with different roughness (a) & (b)...... 221 Figure A- 3 Optical microscope image (left), profilometry image (middle) and cross-section line profile (right) inside and outside the slide track. The experiments were performed with 0.5 N normal load on surface with Rq = 0.054 µm and 0.151 µm respectively...... 222 Figure A- 4 The relationship between plastic deformation contributed real contact area and normal load at different roughness ...... 223 Figure A- 5 The shear strength (estimated from friction force in Figure A-2 and real contact area from Table A-2) versus normal load L1/3. The data perfectly smooth surface is calculated based on Hertzian contact mechanics in Table A-1 ...... 224 Figure A- 6 (a) the schematic of bi-directional sliding and uni-directional sliding (b) & (c) the friction coefficient of two sliding mode under 0.1 N (b) and 0.2 N (c) on Rq = 0.01 µm substrate...... 226 Figure A- 7 The frictional behavior under VPL during bii-direction scan, uni-direction scan and uni-direction scan at clean surface under 0.5 N normal load ...... 227 xx

Figure A- 8 The hardness (a) and reduced modulus (b) of substrate inside and outside the slide track after 100 cycles scan. The indentation depth of nanoindentaion test are selected as 50 nm, 100 nm and 150 nm...... 228

xxi

List of Tables

Table 2- 1 Antoine equation parameters of selected chemicals ...... 9 Table 2- 2 Substrate properties and vendors ...... 12 Table 2- 3 Parameters for Peclet number calculation ...... 13

Table 3- 1 Elastic modulus of tribo-polymers calculated from the AFM indentation data...... 42

Table 4- 1 Comparison of the critical activation volume (∆푉 ∗) calculated from Figure 4-5 and the molar volume calculated from the density of liquid...... 65

Table 5-S 1 Chemical composition of tribo-polymer film produced from allyl alcohol at the sliding interface of silicon oxide...... 94

Table 6-S 1 Properties of the amorphous silica glass created in this work compared to the results reported in previous MD simulations and experiments...... 118

Table 7- 1 The critical activation volume (∆푉 ∗) calculated from Figure 7-4 and the intercept of each semi-log scale load dependence plot. The error is standard error from linear regression. The molar volume of α-pinene 푉푚표푙 is 263.8 Å3...... 145

Table A- 1 The Hertzian contact mechanics on Rq = 0 substrate ...... 219

Table A- 2 The numerically calculated real contact area contributed by plastic deformation & elastic deformation (blue) and elastic deformation alone (white) ...... 223

xxii

List of Equations

Equation 1- 1 ...... 4

Equation 2- 1 ...... 8 Equation 2- 2 ...... 8 Equation 2- 3 ...... 8 Equation 2- 4 ...... 8 Equation 2- 5 ...... 12 Equation 2- 6 ...... 13 Equation 2- 7 ...... 16 Equation 2- 8 ...... 17 Equation 2- 9 ...... 17

Equation 3-S 1...... 46 Equation 3-S 2...... 46

Equation 4- 1 ...... 53 Equation 4- 2 ...... 63 Equation 4- 3 ...... 63 Equation 4- 4 ...... 64 Equation 4- 5 ...... 64

Equation 5- 1 ...... 75

Equation 5-S 1...... 92 Equation 5-S 2...... 92 Equation 5-S 3...... 92 Equation 5-S 4...... 93

Equation 6-S 1...... 120 Equation 6-S 2...... 121 Equation 6-S 3...... 121

Equation 7- 1 ...... 133 Equation 7- 2 ...... 144 Equation 7- 3 ...... 144

Equation 9- 1 ...... 177 xxiii

Equation 9- 2 ...... 188 Equation 9- 3 ...... 188

Equation 10- 1 ...... 199

Equation A- 1 ...... 216 Equation A- 2 ...... 216 Equation A- 3 ...... 216 Equation A- 4 ...... 220

xxiv

Acknowledgements

I would like to express my sincere gratitude to my advisors Dr. Seong H. Kim for helping me throughout my study and life at Penn State. He has taught me how to become an independent researcher. I appreciate all his contributions of ideas, enthusiasm and time to make my Ph.D. experience productive and meaningful. I am also thankful for the guidance he provided during group meetings, making me better at presentations and scientific thinking. The road to Ph.D. is not easy, but it becomes the true wealth for me.

I also thank my defense committee members, Dr. Adri van Duin, Dr. Themis Matsoukas,

Dr. Xueyi Zhang, for taking time to attend my presentation. Their insightful comments and encouragement broaden my research from various perspective, as well improving the quality of this thesis.

My sincere thanks also goes to the project funder (National Science Foundation, grant no.

CMMI-1435766) and collaborators. Dr. Ashlie Martini, Dr. Jejoon Yeon and Arash Khajeh from

UC Merced contributed to all the simulation works in Chapter 5 & 6; Dr. Jane Wang and Zhong

Liu from Northwestern University helped with the mathematical calculation in Appendix A. It’s my honor and pleasure to work with these great minds in tribology.

I would like to thank the entire Kim group. This amazing group has been a source of friendships as well as good advice and collaboration. Dr. Anthony Barthel was able to help me with every question I had and all technique tricks of the instruments. Without his patience, I could not finish my first manuscript by the first year. Dr. Jiawei Luo helped me with valuable discussion about research ideas as well as the life and fun in State College. It would be tired and exhausted without the weekly gathering Friday night with friends, which could be one of my best memories xxv in the happy valley. Two brilliant visiting scholar, Dr. Lei Chen and Chen Xiao are professional in tribology field. It has been great fortune to have these friends understand the project I was working on and exchange our thoughts.

I sincerely acknowledge my internship mentor Dr. Jun Qu and material processing group in Oak Ridge National Laboratory. I am grateful for the invaluable experience in the last year of my Ph.D. life. Dr. Qu’s education and lifestyle impressed me a lot; and also helped me build the foundation for my future career. The superior laboratory research facilities and great colleagues made every day more pleasant in this new place.

Last but not least, I would like to thank my family for their continuous encouragement and love. They always patiently listened to my complaints and supported me in all my pursuits. They are most important part of my life and made me who I am today.

1

Chapter 1

Introduction and Motivation

Mechanochemistry refers to the reactions initiated by mechanical force or stress. As one important category of chemical reaction, it is ubiquitous in nature and engineering systems.1–3 The mechanically induced or facilitated chemical reactions occur at the sliding interfaces of two solid substrates are also known as tribochemistry.4,5 One example is the organic synthesis through ball milling of solid reagents.2 Unlike thermal chemistry, photochemistry and electrochemistry, which occur upon heating, photon irradiation, or electrical bias, mechanochemistry is less well known.

A variety of processes may take place at buried sliding interfaces including:

1. The frictional heat is generated during sliding process due to fast sliding speed or low dissipation efficiency, the thermal induced reaction can easily occur.6

2. On poorly lubricated surfaces, the wear could occur. The highly energetic dangling bonds on worn surface may react easily with species in gas environment.7

3. At sliding interfaces, the shear force or stress may induce or facilitated the chemical reactions, as shown in Figure 1-1.8–10

In the study of mechanochemistry or tribochemistry, the key challenges are to identify the products and reactants trapped at tribological interfaces as well as the role of shear force or energy in reaction dynamics and steps. 2

Figure 1- 1 The reaction coordinate diagrams of the reaction without mechanical force (solid) and the reaction under mechanical force (dashed)

Several attempts have been made in the past to study tribochemistry. Ball milling and ultrasonic cavitation are two types of mechanochemical process in macroscale systems; however, the precise stresses applied to the molecules is difficult to acquire.11,12 It is hard to isolate or characterize the tribo-chemical products. Tribofilm formation via anti-wear additives in lubricant is one relatively well-studied and widely used field in tribochemistry. The additives could be activated under shear stress and react with contact surfaces and/or wear debris at the lubricating interface to grow a compliant and protective tribofilm.13,14 How much of these tribofilms originate from the shear stress at sliding interfaces or because of the dangling bonds exposing at the worn surface and initiating the chemisorption of additives cannot be determined. Atomic force microscope (AFM)15 nano-wear experiments can provide in-situ observation;16 the worn surface area is not sufficient for chemical characterization. Pulling on individual molecules with

AFM tips can rupture bonds with pN resolution, but the technique still shows restrictions. The 3 detected molecules need to be long enough to detect rupture; the bonds between tip and sample in such systems are stronger than the bond under investigation.17–20

The challenges of identifying the mechanochemical reaction with heating and triboemission induced reactions can be overcome by a vapor phase lubrication (VPL) system.21

Vapor phase lubrication refers to the tribological interfaces lubricated via continuously adsorbing vapor molecules.22 In controlled environment, the molecules would adsorb on material surfaces till its equilibrium thickness regardless of the sliding process, achieving the self-healing or replenished film.21,22 This film with mono- or multi-layer molecules can lubricate the interface and prevent wear. Using substrate material with high thermal conductivity or low sliding speed, the flash temperature increase is estimated to be less than 10 ̊C, meaning that the thermal induced reaction is negligible.23,24 To guarantee the effective lubrication process, the adsorbate film needs to keep above minimum thickness, which is typically a monolayer.25 In the case of n-pentanol, the wear could easily occur while with high load or insufficient vapor pressure (<10% p/psat). The wear may expose dangling bonds to the vapor environment, which may react with n-pentanol molecules to produce tribo-polymers at worn surface.25

In the case where frictional heat is negligible and interfaces are well protected by lubricants, the chemical reactions can be considered as mechanical force or stress assisted.26 The mechanically induced or assisted chemical reactions were discussed by Eyring and Prandtl-Tomlinson model respectively.27,28 Eyring’s reaction rate model was used by Zhurkov to study the fracture properties of different materials including metals, polymers and ceramic.29 From his theory, the fracture rate is dominated by mechanical stress and temperature, and also shows a reaction time dependence.

Later in 1978, Bell revealed the relationship between external forces and chemical bond breakage based on Zhurkov’s model.30 The so called Bell model shows that the reactions can be facilitate 4 with the influence of a mechanical energy term. This model also can explain the nanoscale wear measured under AFM and tribo-film formation,8,16,31 which is modified based on an Arrhenius- type kinetic equation. The reaction rate r can be expressed as:

Ea − σΔV∗ r ∝ A exp (− ) Equation 1- 1 kbT

Where A represents the preexponential factor, kb is the Boltzmann constant and T is the temperature at interfaces. The mechanical energy here can be expressed as the product of critical

∗ 8,10,31–36 activation volume at constant shear stress (σΔV ). The term Ea depends on the surface chemistry property, which can be considered as a constant for same system.37 The critical activation volume is assumed to be correlate with the volume change to activate the molecules, but the physical meaning still remain unclear.

Theoretically, the mechanically induced chemical reactions shows shear force or shear stress dependence. The activation volume ΔV* can be estimated experimentally via the relationship between reaction yield or rate and driving force.8,10,31,35,36,38,39 However, in order to obtain atomic view of mechanistic pathways and physical meaning of activation volume, the study integrated the experiments and computational simulations. Molecular dynamics (MD) simulations were performed with a reactive force field (ReaxFF), a versatile and widely used method for various systems.40–43

The mechanistic understanding of the mechanochemistry could help the development of ionic environmentally-friendly additives for hydraulic systems. The ashless ionic liquid additives have been designed and the tribological performance were tested in various fluids. The tribofilm characterization were carried out to interpret the wear protection mechanism. The toxicity and 5 environmental hazard of newly developed lubricants were assessed by Environment Protection

Agency (EPA) approved chronic tests.44

This dissertation will focus on the factors that may govern the shear-induced tribo-reaction rate at tribological interfaces. The main questions discussed in this dissertation are as follows:

1. The tribochemical reactions show molecular structure dependence, which correlates

with the term ‘critical activation volume ΔV*. The mechanical response of shear

created tribo-polymers depends on the monomer reactants. (Chapter 3, 4)

2. The physical meaning of activation volume is investigated experimentally and

computationally. The molecular details of interfacial reactions involving the role of

surface material as well as the structure of reactants have been revealed by ReaxFF-

MD simulations. (Chapter 5, 6)

3. The experiments study systematically designs and studies the critical role of solid

surface chemistry in mechanochemical polymerization of precursor molecules.

(Chapter 7)

4. The effect of gas environment on tribochemical polymerization of precursor molecules

and its influence on surface chemistry property have been studied. (Chapter 8, 9)

5. The physicochemical properties and tribological behaviors of eco-friendly ionic liquid

additives are investigated for hydraulic systems. (Chapter 10)

The key techniques and characterization methods on tribo-induced polymers have also been introduced in details for references (Chapter 2)

6

Chapter 2

Experiment details and data processing

Vapor phase lubrication tests

A home-made reciprocating ball-on-flat tribometer was employed to perform the friction and wear tests in this work. The instrument is presented in Figure 2-1. The linear variable displacement transducer (LVDT) is a Schaevitz DC-EC 250 with 0.25 inch stroke range, used to record the reciprocating movement. The sliding speed is controlled by the motor, which is typically

4 cm/s. The load cell is a Honeywell Model 31 low range precision miniature load cell sensing up to 150g. The forward and backward friction force is measured through the load cell. The LVDT and load cell send measured results to the computer software through digital multimeter (DMM).

The DMM is an Agilent 34410A high performance multimeter. Before the friction tests, the balance arm should be adjusted to be horizontal to get zero load condition. The fixed load is then added on one end of cantilever. A 3/8 inch diameter ball is used as counter surface sliding against the substrate material. The load varies to achieve the desired contact pressure based on material elastic modulus and Poisson’s ratio. The contact pressure is calculated via Hertzian contact mechanics. The friction tests are performed in the environment chamber with continuous vapor flow.

After the friction tests, two data files will be created. One contains all the raw data points for recorded cycles; another only includes the average data with ‘avg’ suffix. In the averaged file, the data recorded are shown as Figure 2-2. Friction coefficient is calculated by averaging the difference between forward and backward values. The Friction coefficient – sliding cycles can be plotted using column 1 & 3. It should be noted that the column one only show the recorded cycles. 7

The actual sliding cycles usually equal to No. ×2 (depending on the sliding speed). The last column represents the regression R2 of the LVDT position; data can be convincing with R2 > 0.9.

Figure 2- 1 Tribometer set and digital multimeter (DMM)

Figure 2- 2 Excel output file of ‘avg’ file

8

The environment control is conducted with continuous vapor flow system, which is crucial for the vapor phase lubrication. The desired partial pressure of vapor is created by mixing a dry carrier gas (eg. N2) and another stream flowing through a glass tower filled with glass beads and the liquid of interest. Figure 2-2 shows a schematic for the continuous flow system to dynamically control the vapor partial pressure inside the chamber. One ‘buffer cup’ is put upside down on the top of the environmental cell to prevent the back flow. The inert gas enters at the bottom of the tower and then the bubble flowing through the liquid will be saturated liquid vapor. The partial pressure of vapor from these bubbles can be calculated using Fick’s Law of diffusion. The Figure

2-3 shows the schematic of the saturated vapor at the boundary diffusing into the bubble till the center. The governing equation can be written as:

1 휕 2 휕푃푣 1 휕푃푣 2 (푟 ) = 푟 휕푟 휕푟 퐷푣 휕푡 Equation 2- 1

Where, r is the distance of saturated vapor to the bubble center, Pv is the partial pressure at

2 different position. Dv represents the diffusion coefficient, which is typically 0.5 cm /s. The original bubble diameter is can be estimated by the glass beads diameter, which is 3/32 inch in diameter.

The liquid-vapor boundary can be assumed as 100% p/psat. The overall diffusion rate (from all directions) to the center point is 0. So the boundary conditions can be expressed as blow

푃 (푟 , 푡) = 푃 푣 0 푠푎푡 Equation 2- 2

∂푃푣 = 0 ∂r 푟=0 Equation 2- 3

푃 (푟, 0) = 0 푣 Equation 2- 4

9

The time required to saturate one bubble with different diameter is shown as Figure 2-3 (b).

The saturated vapor of this stream can be achieved via changing the glass beads size (bubble size) and glass beads height (time of flow) in the glass tower. The time of the bubble staying in the liquid can be estimated by stopwatch. The saturated vapors may condense easily when flow to a glass tower with lower temperature. After vapor stream being determined as saturated, the partial pressure can be acquired by mixing it with a separate dry inert stream. The partial pressure can be estimated using the Antoine equation and Antoine parameters for the liquid of interest, as shown in Table 2-1.

Table 2- 1 Antoine equation parameters of selected chemicals

Chemical Temperature, K A B C

water45 273 ~ 303 5.40221 1838.675 -31.737

α-pinene46 292.59 ~ 428.91 3.92161 1411.869 -68.817

pinane47 330.27 ~ 415.01 6.0518 1520.85 −65.5155

n-decane48 243.49 ~ 310.59 0.21021 440.616 -156.896

allyl alcohol49 294 ~ 370.23 8.78252 4510.213 143.647

n-pentanol50 307.1 ~ 411 4.68277 1492.549 -91.621

10

Figure 2- 3 Schematic of Vapor flow system for environmental control

Figure 2- 4 (a) Schematic of vapor diffusion inside the bubble (b) relationship between diffusion time and radius of bubble

11

Sample preparation

The silicon (100) wafers are purchased from Wafer World, Inc. (West Palm Beach, FL,

USA) Cleaning procedure is important for lubrication experiments. The surfaces was cleaned for

15 min in a UV/ozone followed by an RCA first step procedure. The solution is composed with 5 parts of deionized water, 1 part of ammonia water (29% by weight of NH3) and 1 part of aqueous

H2O2 (hydrogen peroxide, 30%). The samples were cleaned in the solution at ~80 ̊C for 10- 15 minutes. In this case, the base-peroxide liquid is able to remove the organic residue as well as insoluble particles. After the treatment, a thin silicon oxide layer (1~2 nm) will form on the silicon substrate.51

For metal surfaces listed in Table 2-2, as received samples will show macroscopic roughness due to manufacturing process and cutting process. Hand polishing with sandpapers from

320 ~ 2000 grit is performed on the samples, and then followed by a micropolish with a 1 µm colloidal alumina slurry (Buehler). The material substrate can get a root-mean-square (rms) roughness around 10 ~ 20 nm of surface finishing, determined by optical white light profil ometer.

Organic solvents have been used for sample cleaning for sample preparation and characterization.

They can easily remove the organic contaminants on material surfaces. However, the previous research showed that cleaning with organic solvents may leave organic residual on surfaces, which plays a significant role in the friction and lubrication tests.52 The UV-ozone treatment is proven to be effective to remove the organic species. The more appropriate cleaning procedure should follow three steps: DI water rinsing to remove the inorganic salt or particles; ethanol or isopropanol rinsing to remove the organic species; 20-minute UV-ozone treatment to further cleaning the organic residue from step 2. 12

Table 2- 2 Substrate properties and vendors

Substrate Young’s modulus Poisson’s Ratio Shear modulus Vendor GPa Gpa 440 C 200 0.27 ~ 0.30 ~ 83.9 McMaster-Carr

Copper 121 ~ 133 0.34 ~ 0.35 44 ~ 49 McMaster-Carr

Nickel 190 ~ 220 0.305 ~ 0.315 72 ~ 86 McMaster-Carr

Palladium 118 ~ 124 0.385 ~ 0.395 42 ~ 46 Sigma-Aldrich

Gold 76 ~ 81 0.415 ~ 0.425 26 ~ 30 Sigma-Aldrich

Silicon 140 ~ 180 0.365 ~ 0.275 60 ~ 63 Wafer World, Inc

Sapphire 343 ~ 370 0.28 ~ 0.33 131 ~ 142 Argonne National Laboratory

Calculation of flash temperature

(Part of this section is reproduced from Chapter 3 Supporting Information)

The heat is generated by friction behavior at the Hertzian contact area during the sliding.

A numerical approximation by J.F. Archard et al.53 can be used to estimate the frictional temperature rise. First, the Peclet (Pe) number of the system was calculated:

푉 푎 Pe = 2 휅 Equation 2- 5 where a is the radius of contact area, 휅 (푚2/푠) is thermal diffusivity and V (m/s) is the sliding velocity. The thermal diffusivity 휅 is the thermal conductivity divided by density and specific heat capacity, where parameters are listed in Table 2-3. Under a steady state heat conduction approximation, the temperature increase (휃) due to frictional can be estimated through: 13

푄 휃 = 4 푎 퐾 Equation 2- 6 where Q (W) is the total rate of heat generation, which could be calculated from the work of friction, and K (W/(K·m)) is the thermal conductivity of stainless steel.

Table 2- 3 Parameters for Peclet number calculation

Substrate Thermal conductivity Density Specific heat capacity W/m×K Mg/m3 J/kg×K 440 C 24.2 ~7.65 ~460

Copper 147 ~370 8.93 ~ 8.94 372 ~ 388

Nickel 67 ~ 91 8.83 ~ 8.95 452 ~ 460

Palladium 70 ~ 77 11.95 ~ 12.1 240 ~ 250

Gold 305 ~ 319 19.25 ~ 19.35 125 ~ 135

Silicon 84 ~ 100 2.28 ~ 2.38 668 ~ 715

Sapphire 34.6 ~ 40 3.93 ~ 4.01 648 ~ 750

Tribo-polymer yield calculation

After the friction tests, the tribo-polymer yield is estimated via atomic force microscope

(AFM; Digital Instrument MultiMode scanning probe microscope) and surface topography is measured through profilometry (Zygo NewView 7300). The created tribo-polymer showed viscoelastic behavior during force distance response.23 The tapping mode AFM will have the tip touch the surface only for a short time during scanning, thus avoiding the concern of dragging 14 across the piled tribo-polymers. The z-scanning size is usually set as 2 µm (maximum 3.7 µm) and xy-scanning size is adjusted to image the each section of tribo-polymer on the track, which is typically in the range of 70 µm × 70 µm ~ 90 µm × 90 µm. Due to the reciprocating sliding of the ball, the in-situ created tribo-polymer will pile up at the ends and both sides of the slide track. The

AFM images usually include two ends of the slide track and ~3 sections from the middle of the track. The AFM image scan of one end of slide track is shown in Figure 2-4 from NanoScope

Analysis. The cross section line profile can be obtained along the wear track (blue), the area above the reference plane can be obtained through integration. The cross section lines perpendicular to the slide track (red) often show a plateau shape. The total yield can be easily estimated by the integrations from each sections. The simplest integration method is to assume a constant function passing through the point, known as rectangle rule. An alternative way is to export the 3D image from Nanoscope Analysis, and then calculate the volume using ImageJ software. The details of this method is explained by https://imagej.net/Volume_Calculator; but the artifacts from exported

3D image should be removed before calculations.

The friction tests on ductile material or with high load will lead to ineligible plastic deformation of substrate materials. The surface topography effect should be removed to minimize the error of tribo-polymer yield calculation. The profilometry is employed to obtain the 3D information and cross-section profile of substrate surface. The total volume above and below the reference plane (red line) should be taken into account for tribo-polymer yield estimation, as shown in Figure 2-5. The volume can be estimated by the multiplication of average cross section area times the length of slide track.

15

Figure 2- 5 NanoScope Analysis interface for tribo-polymer yield calculation

Figure 2- 6 Cross-section profile (left) and 3D image (right) obtained from optical profilometry software

16

Mechanical behavior tests

The mechanical properties of tribo-polymers are crucial to the lubrication performance. In atomic force microscope, force-sample displacement curves (F-D curves) are used to measure the force response between tip and surface in contact-AFM mode. This relationship can be used to analyze the elastic properties and viscoelastic behavior of the tribo-polymers. Practically, the F-D curves show the relationship of the deflection of cantilever versus the extension of the sample stage scanner (piezoelectric scanner). To obtain the force-sample displacement curve, the cantilever deformation during indentation process should be removed using following equation (2-7).

퐹 퐷 = 퐷 − 푠푎푚푝푙푒 푡표푡푎푙 푘 Equation 2- 7

As shown here, Dtotal is the number directly export from software, representing the sample deformation Dsample plus the cantilever deformation. Here the cantilever deformation can be estimated as force F dividing spring constant of cantilever k (nN/nm). The real F-D curves is shown as blow Figure (2-6). The zero indentation position is set as the cross point of the extrapolated line of loading part and the free-standing curves.

The elastic modulus of the tribopolymer can be estimated from the stiffness of the unloading curve using the Oliver-Pharr indentation model.54,55 This method would be more accurate while the sample is less viscoelastic. Since the tribo-polymer is compliant, the Johnson-

Kendall-Roberts (JKR) model might be more appropriate. The curve fitting method for force- sample displacement analysis is shown here:56 17

4 1 퐹 3 퐹 3 1 + √1 − 1 + √1 − 푟2 퐹 2푟2 퐹 퐷 − 퐷 = 푎푑ℎ − 푎푑ℎ 푐표푛푡푎푐푡 푅 2 3푅 2 Equation 2- 8 ( ) ( )

Where Dcontact represents the displacement at which the tip first contacts the sample. r is the contact radius when F=0 and Fadh is the maximum adhesion force. R is the radius of curvature of

1 AFM tip (or reduced radius, R = 1 1 ). + 푅1 푅2

1 Then the reduced modulus E’ (E’ = 1−휈12 1−휈22 , and υ is the Poisson’s ratio) can be + 퐸1 퐸2 expressed as

−3푅퐹푎푑ℎ E′ = 푟3 Equation 2- 9

One example of JKR fitting is presented in Figure 2-7. 18

Figure 2- 7 Example for force-sample deformation (F-D) curve

Figure 2- 8 Example for JKR fitting

19

Adsorption isotherm measurement methods and principles

In order to study the adsorption isotherm on material surfaces, the measurements need to be carried out with an attenuated total reflection (ATR) on silicon oxide surface or polarization modulation‐infrared reflection‐adsorption spectroscopy (PM‐IRRAS) on metal surfaces. Figure 2-

8 schematically compares the two techniques. For ATR-IR setting, the crystal possesses higher refractive index (n1) than that of sample of interest (n2) (vapor). The IR incidence angle (θi) should be higher than the critical angle (arcsin(n2/n1)), so that the total reflection can be achieved at crystal/vapor interface. In this case, the evanescent wave penetrates into the vapor phase and interacts with the absorption band of the samples. The spectra result is plotted as log (1/R) which represents the absorbance. Before experiments, a silicon crystal was prepared by the cleaning method discussed above, which will result in a native oxide surface with hydroxyl groups on top.

The ATR crystal is promptly mounted to the environment chamber and purged with dry inert gas

(e.g. N2). The background spectra will stabilize till the chamber is filled with inert gas. The adsorption of vapor of interests onto silicon oxide was accomplished by changing the vapor pressure inside the chamber, which is discussed in ‘vapor phase lubrication test’ section.

Figure 2- 9 Schematic illustration of attenuate total reflection (ATR) and polarization modulation‐infrared reflection‐adsorption spectroscopy (PM‐IRRAS) spectroscopy 20

The PM-RAIRS is used to characterize thin films or monolayer on metal substrate, due to the advantage of high surface sensitivity as well as the surface selection rule. A photo-elastic modulator generates alternating linear states of polarized beam as shown in Figure 2-8. Two polar electric vectors are perpendicular to each other, where p component is parallel to incident plane and s component is perpendicular to that. Only p-polarization interacts with dipole moments at the metal surface, while s-polarization is canceled due to surface selection rule.57,58 The sum (p+s) signal will carried the spectra of both adsorbate and gas phase molecules while the difference (p- s) signal contains only vibration spectra of adsorbates. As a result, the ratio (p-s)/(p+s) represents only the information of adsorbed molecules. In the experiments, the PM-RAIRS has been performed on Copper (CuO), Palladium (Pd) and Gold (Au) surfaces because of their stronger reflection signals. The scan number should be at least 1000 to give decent signal to noise ratio.

The environment control method is same as that for ATR-IR experiments.

21

Chemical Characterization

Time-of-Flight Secondary Ion Mass Spectrometry (TOF-SIMS) is employed to investigate the surface spectroscopy and surface imaging of the tribo-polymers. It works via using a pulsed

2+ ion beam (e.g. 퐵𝑖3 ) to remove the molecules from the very outermost surface layers. The removed particles are accelerated and their mass can be determined by measuring the time at which they get to the detector. In our experiment, ToF-SIMS measurement is conducted using a Physical

Electronic nanoToF II equipped with bismuth cluster liquid metal ion gun, triple focusing

2+ electrostatic analyzer and a multichannel plate (MCP) detector. A 퐵𝑖3 ion beam is used to sputter the area of interest (usually the end of slide track) and the mass-to-charge ratios of the ejected secondary ions are recorded.

X-ray photoelectron spectroscopy (XPS) is used to analyze the chemical states of tribo- polymers. It works via exciting a tribo-polymer surface with mono-energetic Al kα x-rays causing photoelectrons to be emitted from the surface. One electron energy analyzer is used to measure the energy of the emitted photoelectrons. The PHI, VersaProbe instrument is employed to measure the tribo-polymer accumulated in a 0.2 × 0.2 mm2 area.

Micro-InfraRed Spectroscopy can obtain infrared spectrum of absorption of the thin tribo- polymers. A Bruker Hyperion 3000 IR Microscope Coupled to IFS 66/s spectrometer is used to perform the IR measurement. The typical scan range for tribo-polymer is 1200 ~ 3800 cm-1, and the regimes of interest are usually hydroxyl group (-OH, 3200 ~ 3600 cm-1), alkene (=C-H, 3000

~ 3100 cm-1), alkane (C-H, 2800 ~ 3000 cm-1) and carbonyl group (C=0, 1650 ~ 1800 cm-1).

22

Molecular Dynamics simulation

Computational molecular dynamics (MD) simulations are performed through collaboration with

Prof. Ashlie Martini at University of California, Merced. MD simulation can mimic the atom physical movement assuming a given potential energy function. So that the energy function could calculate the force experienced by any atom given the positions of the other atoms. However, the conventional MD simulation is operated without reactive force field. Recently, many reactive force field have been developed to investigate the chemical reactions, including REBO, COMB and

ReaxFF. ReaxFF has been demonstrated in various systems and appears to be the most suitable tool studying molecular details in tribo-chemical reaction.

23

Chapter 3

Tribochemical synthesis of nano-lubricant films from adsorbed molecules at sliding solid

interface: Tribo-polymers from α-pinene, pinane, and n-decane

Part of this chapter is reproduced with permission from Elsevier: He, X.; Barthel, A. J.; Kim, S. H.

Tribochemical synthesis of nano-lubricant films from adsorbed molecules at sliding solid interface: Tribo- polymers from α-pinene, pinane, and n-decane. Surf. Sci. 2016, 648, 352-359

Overview

The mechanochemical reactions of adsorbed molecules at sliding interfaces were studied for α- pinene (C10H16), pinane (C10H18), and n-decane (C10H22) on a stainless steel substrate surface.

During vapor phase lubrication, molecules adsorbed at the sliding interface could be activated by mechanical shear. Under the equilibrium adsorption condition of these molecules, the friction coefficient of sliding steel surfaces was about 0.2 and a polymeric film was tribochemically produced. The synthesis yield of α-pinene tribo-polymers was about twice as much as pinane tribo-polymers. In contrast to these strained bicyclic hydrocarbons, n-decane showed much weaker activity for tribo-polymerization at the same mechanical shear condition. These results suggested that the mechanical shear at tribological interfaces can induce the opening of the strained ring structure of -pinene and pinane, which leads to polymerization of adsorbed molecules at the sliding track. On a stainless steel surface, such polymerization reactions of adsorbed molecules do not occur under typical surface reaction conditions. The mechanical properties and boundary lubrication efficiency of the produced tribo-polymer films are discussed. 24

Introduction

Lubrication is necessary for mechanical devices to prevent wear of sliding parts and increase the device operation reliability. Mechanical systems with sliding contacts often fail due to surface wear at the sliding interface.59 To avoid this problem, protective coatings have been applied, yet the durability of surface coatings often limits the reliability of such approaches employing one-time application of lubricant coatings.60,61 As an alternative, lubrication by molecular adsorbate films in equilibrium with the vapor phase has been developed; this method is called vapor phase lubrication (VPL).62–64

In VPL, the vapor molecules from the surrounding environment adsorb on the surface to form mono- or multilayers, which can prevent wear.21 Just like coatings, the adsorbed molecules might be removed (desorbed) during sliding, yet as long as the vapor pressure of the adsorbate is kept above a certain level, the lubricant molecules could adsorb to its equilibrium thickness to replenish the film; thus, lubrication with self-healing can be possible. Alcohol has been demonstrated to be an effective lubricant for various kinds of materials.21,65,66 VPL has also been demonstrated to be effective for lubrication of microelectromechanical systems (MEMS) without interfering with the device function or performance.22

One limitation of VPL is the fact that the vapor pressure of lubricating molecules must be kept above a certain level at all times so that the minimum thickness of adsorbate films is assured.

The minimum thickness is typically a monolayer.67 In the case of n-pentanol, when its vapor pressure is reduced to less than 10% of the saturation vapor pressure, the surface could wear due to insufficient supply of lubricant molecules.62 When this happens on a silicon substrate, the wear exposes dangling bonds of silicon at the damaged surface, which could induce chemical reactions of n-pentanol molecules, producing tribo-polymers at the wear track.7 These tribo-polymers 25 formed from n-pentanol could provide lubrication effects in the absence of n-pentanol vapor, but its lubrication effect diminished too quickly. Also, the formation of tribo-polymers from n- pentanol requires reactive surface sites which may not be available, especially on well-lubricated interfaces.

During VPL, molecules adsorbed at the sliding solid-solid interface might undergo mechanochemical or tribochemical reactions under high contact pressure and frictional shear force.10,68 In these cases, chemical reaction is activated by mechanical energy, rather than thermal or photochemical origins.11 Under mechanical compression or shear, the potential energy surface of molecules may be distorted and the energy barrier be lowered along a reaction coordinate, 69,70 expediting or allowing chemical reactions that would not occur under thermal or photochemical conditions.3,10,68 Recently, it has been shown that allyl alcohol adsorbed on stainless steel could be polymerized under interfacial shear conditions.71 Even though allyl alcohol has a double bond, it does not undergo typical radical polymerization in thermal reaction conditions.72 However, it was easily polymerized under the mechanical shear condition. More interestingly, the tribo-polymer produced from allyl alcohol at the sliding track acted as a boundary lubrication film for an extended period of time without continuous supply of lubricant vapor.71

In this paper, we report tribochemical reactions of α-pinene (C10H16) and pinane (C10H18).

These are bicyclic terpenes with highly-strained 4-membered rings. α-pinene is the major constituent of pine tree oil or turpentine and is known to undergo the ring opening of the strained

4-membered ring to generate a relatively stable tertiary pinane carbocationic species and produce polymers.73,74 The viscoelastic properties of poly-α-pinene is suitable for a tackifying agent of pressure-sensitive adhesives.75 The same property could be used for boundary lubrication of sliding interfaces. It was hypothesized that the mechanical shear at tribological interfaces could 26 assist the ring opening or rearrangement leading to the production of tribo-polymeric films at the sliding track during VPL. The mechanochemical or tribochemical reaction activity of these strained bicyclic molecules was compared with n-decane (C10H22), a linear alkane molecule without any internal strain. The viscoelastic property and boundary lubrication efficiency of the produced tribo-polymer films were studied.

Experimental Details

A custom-made reciprocating ball-on-flat tribometer with an environment control capability was employed to test friction and wear between an AISI 440C stainless steel substrate and an AISI 440C stainless steel ball (diameter=3 mm). The substrate was polished with fine grit sandpapers and a polishing solution with 1 µm colloidal alumina. Bearing-grade 440C balls with smooth surface finish were purchased from McMaster-Carr (Ohio, USA). All substrate and ball surfaces were cleaned with ethanol followed by UV/ozone. The friction and wear tests were performed with a contact load of 2 N, which corresponded to a Hertzian contact pressure of about

0.6 GPa. The sliding speed was 0.4 cm/s and the sliding span was 2.5 mm. After the friction test, the slide track was analyzed in ambient air by optical profilometry (Zygo NewView 7300).

The vapor control during the friction and wear test was done in a continuous flow mode.

Details of the vapor control system were previously described.76 The liquids used to create vapor environments were α-pinene (2,6,6-trimethylbicyclo[3.1.1]hept-2-ene) (CAS No. 7785-70-8), pinane (2,6,6-trimethylbicyclo[3.1.1]heptane) (CAS No. 473-55-2), and n-decane (CAS No. 124-

18-5). The saturated organic vapor flow was created by bubbling N2 carrier gas through a flask filled with the organic liquid and glass beads. The saturation vapor pressure (푝푠푎푡) of α-pinene, pinane, and n-decane is 4.75 Torr, 2.19 Torr, and 1.43 Torr, respectively, at room temperature. 27

The saturated vapor stream was mixed with the dry N2 carrier gas to obtain the desired partial pressure of the vapor. The partial pressures of the organic vapors are expressed as percent relative to saturation (p/푝푠푎푡×100%). In this experiment, we applied 30% p/푝푠푎푡 vapor pressure of α-pinene, pinane, and n-decane, which appears to be high enough to form a coverage close to the monolayer on stainless steel surface.21

The adsorption of organic vapor was monitored with polarization-modulation reflection absorption infrared spectroscopy (PM-RAIRS). A ThermoNicolet Nexus 670 spectrometer with a

Hinds Instrument photoelastic modulator was used for PM-RAIRS experiments. PM-RAIRS probes the adsorption layer through absorption of IR beam sequentially modulated with s- and p- polarizations.77 The metal selection rule of IR reflection allows detection of the adsorbed layer without the interference of the gas phase.57 The sum (p+s) signal of these two polarizations carries the signal of both adsorbate and gas phase molecules while the difference (p−s) signal contains only the vibration spectrum of the adsorbate molecules. Consequently, the ratio (p−s)/(p+s) represents only the vibration spectrum of adsorbate layer. In this experiment, the PM-RAIRS measurement was conducted on copper surface, instead of stainless steel surface, because copper gives stronger reflection signals.58,78 The PM-RAIRS data of near-saturation vapor pressure is shown instead of the 30% p/psat spectra for better clarity with a higher signal to noise ratio; the spectral features in those spectra were the same except the total signal intensity. The full adsorption isotherm behaviors of the molecules used in this experiment are shown in the Supplementary

Information.21,79–81

The tribo-polymers formed and accumulated at both ends of the slide tracks were analyzed with micro-IR spectroscopy and time-of-flight secondary ion mass spectrometry (ToF-SIMS).

Micro-IR analysis was carried out using a Bruker Hyperion 3000 IR Microscope coupled to IFS 28

66/s spectrometer with a 15X objective lens. The analyzed areas were 50 m2. ToF-SIMS analysis was conducted using a Physical Electronics nanoToF II equipped with a bismuth cluster liquid metal ion gun, a triple focusing electrostatic analyzer, and a multichannel plate (MCP) detector. A

2+ 12 2 2 Bi3 ion beam (30 keV; total dose = 6.95×10 ions/cm ) was used to sputter a 70 × 70 m area of interest. The mass to charge ratios of the ejected secondary ions were measured to investigate the molecular composition of tribo-polymers.

The tribo-polymers around the slide track were imaged with atomic force microscopy

(AFM) using a Digital Instrument MultiMode scanning probe microscope. Silicon AFM tips

(NANOSENSORS™ PPP-FM) were cleaned with UV/ozone to remove any residual contaminant species. The spring constants of the AFM cantilevers were calibrated using the Sader method; they were 2.5 N/m for the AFM cantilever used for α-pinene test, 2.8 N/m for pinane test, and 2.7 N/m for n-decane test.82 Topographic images of tribo-products were obtained with a tapping-mode scanning. The scan size of the AFM images was 90 μm × 90 μm. The viscoelastic properties of the tribo-polymer were investigated by recording the normal deflection signal of the cantilever during the tip approach and retraction with different set values of final indentation force (20 – 120 nN). The load and unload rate of the vertical piezo-scanner during the indentation test was 440 nm/s. The tip-surface distance or indentation depth of the sample was calculated by subtracting the cantilever defection distance from the total travel distance of the piezo-scanner.83

Result and Discussion

Friction and wear were measured between the AISI 440C stainless steel ball and substrate under 30% p/ 푝푠푎푡 α-pinene, pinane, and n-decane vapors. As shown in Figure 3-1a, the unlubricated friction was very high and noisy in dry nitrogen. Because there was no lubricant 29 between sliding interfaces, both ball and substrate showed severe wear (Figure 3-1b). In contrast, tribo-tests in 30% p/푝푠푎푡 of α-pinene, pinane, and n-decane vapors showed a friction coefficient of

~0.2 that remained stable for the duration of test. This friction coefficient value is in good agreement with the typical values observed for VPL with organic molecules.21,22 After 600 reciprocating cycles of friction tests, deposits with iridescent color could be seen at the ends of the sliding tracks in optical microscope images (Figure 3-1a). The iridescent color was produced due to the interference of the light reflected from the top and bottom surfaces of the deposit on the substrate surface. The surface of the deposit looked smooth, suggesting that they are not wear debris or particulates of substrate materials. These features in the optical images indicated that the deposits must be polymeric species with varying thickness and smooth surfaces formed during the friction test in α-pinene, pinane and n-decane vapors.

30

Figure 3- 1 (a) Friction coefficient of stainless steel in dry N2 and VPL with 30% p/푝푠푎푡 α- pinene, pinane, and n-decane; (b) Cross-section line profile (top), optical profilometry image

(middle) and optical microscope image (bottom) of wear tracks after friction tests in dry N2 as well as VPL with α-pinene, pinane, and n-decane. In (b), tribo-polymers were removed by ethanol to analyze the damage in the substrate surface. The unit of the y-axis of the cross-section line profile is m.

The lubrication and wear prevention effects of the adsorbed vapors under equilibrium adsorption conditions were confirmed by analyzing the wear track with optical profilometry.

Figure 3-1b presents the line profiles as well as three-dimensional images of the wear tracks. In 31

dry N2 environment, the wear marks on the substrate were ~8 μm deep and ~80 μm wide. In contrast, there were no discernible wear mark in the slide tracks tested in α-pinene and pinane vapors when polymeric deposits were removed by rinsing and rubbing with a cotton swap wet with ethanol. A slight depression (<10 nm deep, ~40 μm wide) could be identified in the line profile

(Figure 3-1b). The inner surface of this depressed track was very smooth. This depression was caused by the plastic deformation of the substrate due to the applied load, not due to mechanical wear of the slide track. In the case of n-decane, the accumulation of polymeric deposits was much smaller compared to the -pinene and pinane VPL cases. Although the wear track tested in the n- decane VPL had dark spots in optical microscope images, the topographic analysis indicated that the substrate wear was also negligible (Figure 3-1b).

32

Figure 3- 2 ToF-SIMS ion maps (a,c,e) and mass spectra (b,d,f) and of tribo-polymer products deposited at the end of the slide track produced during (a,b) -pinene VPL, (c,d) pinane -pinene,

(e,f) n-decane VPL. In the ion maps, blue is the iron ion (m/e=55.93 amu), red is the chromium ion (m/e=51.94 amu), and green is the total intensity of ions with m/e = 164~300 amu. The mass of ions marked with * in the inset of (b) are 216, 228, and 244 amu.

The polymeric nature of the reaction product was confirmed with ToF-SIMS. Figure 3-2 shows the selected ion maps of the reaction products piled at the ends of the sliding tracks as well

2+ as the mass spectra of ions desorbed upon sputtering with Bi3 ions. The images of the ToF-SIMS ion maps were very similar to the optical images shown in Figure 3-1a. Outside the polymeric deposit region, the main ions detected were iron (m/e = 55.93 amu) and chromium (m/e = 51.94 amu) which are the main constituents of the stainless steel surface. Within the deposit region, these inorganic ions from the substrate were not detected. Instead, organic ion peaks were detected with masses extending up to ~500 amu. Note that the molecular weights of the adsorbed α-pinene, pinane, and n-decane molecules are 136 amu, 138 amu, and 144 amu, respectively. Also, the ions with masses higher than the molecular weight of the adsorbed molecules were separated by 13

22,62 amu (CH) and 14 amu (CH2), consistent with fragmentation of polymeric hydrocarbons. These

ToF-SIMS data clearly indicated the presence of organic molecules whose molecular weights are 33 higher than 500 amu. Due to heavy fragmentation of organic species by high energy ion sputtering and low detection yield of the MCP detector for higher mass organic ions, the maximum molecular weights of the polymeric species could not be determined. It was reported that typical molecular weight of poly-α-pinene produced by cationic polymerization is 700-800 amu.73 There are no references for molecular weight comparison of the polymers produced from pinane and n-decane since these molecules cannot be polymerized via chemical initiators. The ToF-SIMS results confirmed that the reaction products detected around the sliding track are polymers. Since the stainless steel surface cannot induce such polymerization reactions of the adsorbed molecules at ambient conditions, they must be produced via tribochemical or mechanochemical mechanisms.

Thus, they will be called tribo-polymers henceforth.

In order to determine the exact molecular structure of tribo-polymers, nuclear magnetic resonance (NMR) analysis is required. But the amount of tribo-products synthesized during the friction test was too small for NMR analysis.73,84 This is one of the main challenges in studying tribochemical reactions. Although not as definitive as NMR, IR spectroscopy can provide qualitative information of the molecular structure of tribo-products. In the previous adsorption isotherm study of n-decane using PM-RAIRS, it was found that a complete monolayer may not form on a flat surface at ambient conditions with partial vapor pressure lower the saturation condition.21 However, capillary condensation around the periphery of the contact may allow formation of at least a monolayer of adsorbed molecules to lubricate the contacts.79,85,86 The organic reaction product can also be detected with using micro-IR spectroscopy.25,71

34

(a) (b )

Figure 3- 3 Infrared spectra of vapor, adsorbate, and tribo-polymer of (a) α-pinene and (b) pinane. The vapor spectra were the sum signal of PM-RAIRS measurements; the adsorbate spectra were the difference signal of PM-RAIRS normalized with the sum signal. The tribo- polymer spectra were obtained from the deposits within a 50  50 m2 area at the end of the sliding track.

Figure 3-3 compares the IR spectra of α-pinene and pinane adsorbates and their tribo- polymers with the vapor phase spectra of these molecules. In Figure 3-3a, the IR spectrum of - pinene vapor shows peaks at 1270 cm-1, 1370 cm-1, 1385 cm-1, 1450 cm-1, and 1470 cm-1 in the C-

H bending vibration region and 2840 cm-1, 2880 cm-1, 2910 cm-1, 2945 cm-1, 2985 cm-1, and 3030 cm-1 in the C-H stretch region.87 The small peak at 3030 cm-1 is characteristic to the =C-H vibrations of -pinene. The C=C stretch band at 1660 cm-1 were difficult to identify due to its weak intensity and the interference from gas-phase water signals (sharp spikes in 1400 – 1800 cm-

1 -1 ). The 2985 cm peak is the asymmetric stretch of the CH3 group at the strained 4-membered ring. 35

This peak position is higher than typical values of the CH3 asymmetric stretch modes; it is well known that the stretch vibration position of the CH3 groups attached to the strained ring is blue- shifted.88,89 The adsorbate spectrum of -pinene showed the peaks identical to the vapor phase spectrum, indicating that the adsorbed -pinene on the stainless steel surface remains intact without any dissociation through chemical reactions with the stainless steel surface.

The -pinene tribo-polymer exhibited the C-H bending peaks almost identical to the molecular -pinene, but the C-H stretch peaks were changed drastically. Especially, the 2985 cm-

1 -1 and 3030 cm peaks disappeared, indicating that the strained 4-membered ring to which two CH3 groups are attached and the C=C double bond in the heptene ring are converted during the tribo- polymerization reactions. The relative intensities of 2880 cm-1, 2920 cm-1, and 2948 cm-1 peaks were very similar to poly--pinene produced via cationic polymerization.90 The broad peaks at

3400 cm-1 and 1720 cm-1 were hydroxyl and carbonyl peaks. These groups were not present in the adsorbate spectrum. It is known that the unsaturated double bond of poly--pinene can be easily oxidized in air, producing hydroxyl and carbonyl groups.90 The hydroxyl and carbonyl groups in the -pinene tribo-polymer is believed to be formed during the air exposure of the sample inevitably happened during the sample transfer from the tribo-test cell to the IR spectrometer; thus, the presence of hydroxyl and carbonyl groups supported that the -pinene tribo-polymer has the molecular functional groups or structures similar to the chemically-synthesized poly--pinene. In

ToF-SIMS analyses (Figure 3-2), it was noted that ions with m/e= 216 amu, 228 amu, and 244 amu were prominent in the mass spectrum of the -pinene tribo-polymer (marked with * in the inset of Figure 3-2b), compared to those of the pinane and n-decane tribo-products. Since their mass differences are 12 amu and 16 amu, they must correspond to fragment ions containing carbonyl (C=O) groups, consistent with the presence of the 1720 cm-1 peak in IR (Figure 3-3a). 36

These results indicated that although initiation and propagation mechanisms are different, the tribochemical reactions of -pinene produce polymers having oxidation behaviors similar to the poly--pinene synthesized through cationic polymerization.

In Figure 3-3b, the IR spectrum of pinane vapor has the peaks at 2880 cm-1, 2910 cm-1, and

2940 cm-1 as well as a shoulder peak at 2985 cm-1 in the C-H stretch region and two peaks at 1380 cm-1 and 1470 cm-1 in the C-H bending region. Pinane has the strained 4-membered ring but does not have the C=C bond; thus, the presence of the 2985cm-1 peak but the absence of the 3030 cm-1 are consistent with the molecular structure of pinane. Again, the IR spectrum of the adsorbed pinane has the same peaks of the pinane vapor, indicating that the adsorbed species are molecularly intact. Since pinane is not readily polymerized chemically, no reference spectrum could be found in the literature. However, it was noted that all peaks of the pinane tribo-polymer were the same as those observed for the -pinene tribo-polymer, except that the peak intensities at 3400 cm-1 and

1720 cm-1 were much lower. The absence of the 2985 cm-1 peak in the pinane tribo-polymer spectrum indicated that the strained 4-membered ring of pinane must be dissociated and involved in polymerization reactions. These results suggested that the pinane tribo-polymer must have molecular structure similar to the -pinene tribo-polymer, except the C=C double bonds which are subject to oxidations in air.

The exact initiation mechanism for tribochemical polymerization (Figure 3-2) could not be determined unambiguously; but it must originate from mechanochemical processes since no chemical initiator was used and the vapor molecules were physisorbed without dissociation on the stainless steel surface (Figure 3-3). There was no wear, other than small plastic deformation, of the stainless steel substrate (Figure 3-1). Thus the exposure of reactive dangling bonds by substrate wear, which was the key requirement for tribo-polymer formation from n-pentanol adsorbed on a 37 silicon wafer,25 is not responsible for the tribo-polymerization of vapor molecules studied in this work. Since the sliding speed was very low in our test condition, frictional heating of the slide track was negligible. Based on the thermal diffusivity of stainless steel and the measured friction coefficient (~0.2), the temperature rise of the sliding contact was estimated to be less than one degree (see Supplementary Information).53 Ruling out these possibilities, one could attribute the activation and initiation of the adsorbed molecule for polymerization to a mechanochemical or tribochemical process.

Under mechanical stress or shear, the potential energy of chemical bonds could be distorted, facilitating chemical reactions that would not occur under normal conditions.11,34,69,91 Based on the

IR analysis (Figure 3-3), it could be proposed that the C-H or C-C bonds of the highly-strained 4- membered rings of -pinene and pinane are the sites most likely to be activated upon shear of molecules at the sliding solid interface. Note that even though n-decane does not have internal strain, it was also polymerized tribochemically at the sliding interface (Figure 3-2f). The mechanical energy at the shearing interface must be high enough to dissociate C-C and or C-H bonds. The presence of ring strain or C=C double bond could facilitate these dissociation processes.

This would release the strain of the 4-membered rings of these molecules, forming relatively stable tertiary ions or radicals.73,74,92

Scheme 1 illustrates the intermediate species involved in the cationic polymerization of - pinene. In the case of -pinene, the proton donation from the cationic initiator produces tertiary carbenium ion with the pinane structure.73,74,92 Then, the pinane type cation isomerizes to the p- menthane and bornane type ions, which undergo propagation reactions forming polymers. The mechanochemical ring-opening of -pinene would produce ions or radicals whose structures are similar to the p-menthane and bornane type cations shown in Scheme 1. Since these intermediates 38 are confined and mixed with monomers within the physical contact region between two solid surfaces, they could undergo polymerization reactions. In the case of pinane, the main intermediate would be the bornane type species; thus, the produced tribo-polymer would not contain double bonds, which makes it less susceptible to oxidation.

Scheme 1. Polymerization of -pinene via cationic intermediates. 92

The tribo-polymerization yield was estimated by measuring the volume of the tribo- polymer deposits produced during 600 reciprocating cycles under VPL conditions with AFM.

Figure 3-4 displays the 90 m  90 m AFM images of tribo-polymers piled at both ends and the middle of the slide tracks after 600 cycles of VPL tests. From the line profiles shown in the left of the AFM images, it can be seen that the tribo-polymer products were mostly pushed and piled at the turning points of the ball sliding action. In the middle of wear track, the tribo-polymers were squeezed to the side of the track. By integrating the line profiles over the entire sliding track, the total amounts of the tribo-polymer product produced during the 600 sliding cycles were estimated to be approximately 7.0×103 μm3, 3.2×103 μm3, and 0.5×103 μm3 for -pinene, pinane, and n- decane VPL, respectively. The large difference in polymerization yields between strained bicyclic molecules (-pinene and pinane) versus linear alkane molecule (n-decane) supported the 39 hypothesis that the strained ring can easily be activated mechanochemically. The higher polymerization yield for -pinene than pinane must be due to the activation of the C=C double bond or additional ring strain due to the presence of the C=C double bond in the heptene ring.

Figure 3- 4 Tapping-mode AFM images and line profiles of tribo-polymers piled at the ends and sides of the slide track after VPL tests in (a) α-pinene, (b) pinane, and (c) n-decane vapors. The line profile was taken along the sliding direction at the left end of the slide track. Each AFM image was taken over a 90 m  90m area. Indentation test positions were marked as green cross mark on the AFM images and arrow on the cross-section profiles. 40

The mechanical properties of tribo-polymers produced from -pinene, pinane, and n- decane were investigated with AFM by recording the force as a function of tip-surface distance during the indentation load and unload cycle (Figure 3-5).83,93–95 The indentation positions were chosen to be on the tribo-polymer piled up at the end of the wear track, which are marked in Figure

3-4. In that way, the indentation depth did not exceed a few percent of the tribo-polymer thickness to avoid any contribution from the solid surface.96 The zero indentation position was set as the cross-points of the extrapolated lines of the linear portions of the loading and free-standing curves.

It is noted that the force curve started increasing even before the tip surface distance became zero.

This was due to the polymer films transferred to the tip surface during the tapping mode imaging

(see Supplementary Information). In this plot, the slope of the load and unload curve corresponds to the stiffness of the material being probed with the AFM tip and the largest negative force measured is the adhesion force between the tip and the probed surface.

Figure 3- 5 Force-distance curve of (a) α-pinene tribo-polymers, (b) pinane tribo-polymers and (c) n-decane tribo-polymers accumulated at the end of slide track on stainless steel. Each tribo- polymer was probed with a new clean AFM tip. The piezo-scanner travel distance was converted 41 to the tip-sample distance by subtracting the cantilever deflection distance. The tips was used for tapping-mode imaging before the force-distance measurements.

The stiffness of the -pinene tribopolymer as well as the adhesion force increased as the indentation depth was increased (Figure 3-5a). This result implied that -pinene tribopolymer is viscoelastic.97 Since a constant loading rate was used, a larger indentation depth meant a longer contact time; the -pinene tribo-polymer had more time to flow and make greater contact with the probing tip surface. The strong adhesion to the tip surface resulted in a longer pulling of a meniscus before the tip was completely separated from the polymer surface and returned to the free standing position (zero force) during the unloading cycle.98 This behavior is consistent with the viscoelastic property of poly--pinene, which is widely used as a tackifying agent for pressure-sensitive adhesives.75 So, the viscoelastic response observed in Figure 3-5b confirmed the similarity of the mechanical properties of -pinene tribopolymer to the chemically-synthesized poly--pinene, which is supported by the IR analysis (Figure 3-3a).

The pinane tribo-polymer showed more elastic behavior than the -pinene tribo-polymer

(Figure 3-5b). The hysteresis between the loading and unloading curves was much smaller for the pinane tribo-polymer. This could be due to the difference in relative abundance of p-menthane and bornane moieties in the tribo-products of -pinene versus pinane. The n-decane tribo-polymer showed much stiffer behavior than the pinane tribo-polymer (Figure 3-5c). Again, these differences in mechanical properties must be due to the distinct chemical structures of precursor molecules involved in tribochemical polymerization at the sliding interface.

The elastic modulus of the material probed with the AFM tip can be estimated from the stiffness of the unloading curve using the Oliver-Pharr indentation model or fitting the sample 42

deformation with the Hertzian model.54, 99 In these calculations, the contact area was estimated

from the indentation depth and the nominal dimensions of the four-sided pyramidal AFM tip

provided from the manufacturer.99 The Poisson ratio of the tribo-polymer products is unknown;

thus, for a simple comparison purpose, it was assumed to be ~0.45, which is within the typical

range for rubber materials (0.4~0.49).100 Since the tribo-polymer is compliant, the Johnson-

Kendall-Roberts (JKR) model might be more appropriate (see the Supplementary Information).56

The elastic modulus values estimated from the force-distance curves shown in Figure 3-5 are listed

in Table 3-1. Although the absolute values could not be determined because the uncertainty in the

Poisson ratio and the zero-point contact position, these values should be taken as a qualitative trend.

The -pinene tribo-polymer still shows the lowest elastic modulus around 1 GPa, then the next is

the pinane tribo-polymer, and the hardest is the n-decane tribo-polymer. The modulus value

obtained for -pinene from the Oliver-Pharr method was comparable to the value reported for

hydrogenated poly--pinene (2.5 GPa).84 There is no literature value for the other two polymers

since they cannot be synthesized through conventional polymerization methods.

Table 3- 1 Elastic modulus of tribo-polymers calculated from the AFM indentation data.

Polymers From Oliver-Pharr modela) From Hertzian fitb) From JKR fitc)

-pinene tribo-polymer 1.4 – 1.7 GPa 0.5 GPa 1.0– 1.1 GPa

Pinane tribo-polymer 2.9 – 3.6 GPa 2.8 GPa 1.9 – 4.7GPa n-decane tribo-polymer 3.5 – 4.2 GPa 7.2 GPa 4.8 – 7.3 GPa

a) From the stiffness of the unloading curve of the 60 nN and 120 nN indentation data b) From the loading curve of the 120 nN indentation data c) From unloading curve fitting of the 60 nN and 120 nN indentation data

43

The lubrication effects of these tribo-polymers in the absence of continuous supply of organic vapors were tested. The 30% p/푝푠푎푡 α-pinene, pinane, or n-decane vapor was supplied for the first 1200 cycles to produce tribo-polymers and then the friction test was continued without vapor supply (Figure 3-6). For comparison, Figure 3-6 also shows the test result of n-pentanol VPL where the tribo-polymerization activity is negligible.21 In the case of n-pentanol VPL, the lubrication effect is lost immediately when the n-pentanol vapor supply was ceased. In the case of tribo-polymers, when the vapor supply was stopped, the friction coefficient was slightly higher than the VPL case; but it stayed around 0.22 and stable. The monomer dissolved in the tribo- polymer would act as a plasticizer,101–103 making the tribo-polymers more viscoelastic thus enhancing the lubrication effects of the tribo-polymer. The slight increase of the friction coefficient at the cessation of the vapor supply could be due to changes in the mechanical properties of tribo- polymer films when they lose plasticizing molecules from the bulk.

When the tribo-polymers were consumed completely, then the friction coefficient rose to

0.8, which was the same as that in dry N2 test, indicating the lack of lubricant inside the sliding track. The lubrication effects of α-pinene and pinane tribo-polymers were similar, although the duration was different. The α-pinene tribo-polymers worked for 8000 cycles and then failed, while the pinane tribo-polymers functioned only for about 4000 cycles. The longer duration of the lubrication effect for the -pinene triboproduct could be attributed to the greater amount of the lubricating polymer films (Figure 3-4) and the more viscoelastic properties of the -pinene polymer (Figure 3-5). The n-decane tribo-product was not very effective at providing or maintaining lubrication effects when the n-decane vapor supply was stopped. 44

Figure 3- 6 Lubrication effect of the tribo-polymer films in the absence of vapor supply. The tribo- polymer film was synthesized by rubbing stainless steel surfaces for 1200 cycles in the presence of 30% p/푝푠푎푡 α-pinene (black), pinane (red), or n-decane (blue) and then friction was measured continuously in dry N2 after stop supplying vapors. The 30% p/푝푠푎푡 n-pentanol (orange) is also tested for 1200 cycles and then measured in dry N2.

Conclusion

Vapor molecules physisorbed on solid surfaces can undergo mechanochemically-induced reactions during sliding in tribological processes. The vapor of bicyclic molecules, α-pinene and pinane, was shown to produce a polymer-like film during sliding between stainless steel surfaces 45 which worked as an effective lubricant film in the absence of the continuous supply of the organic vapor. The α-pinene and pinane tribo-polymer products seemed to have molecular structure and physical properties similar to polymer produced via ring-opening reactions. The highly-strained 4- membered ring of these terpene molecules seemed to be easily activated via a mechanochemical mechanism in the tribological interface and their reaction products worked as a boundary lubrication film.

46

Supporting Information

Estimation of temperature rise due to friction at the sliding interface

The heat is generated by friction behavior at the Hertzian contact area during the sliding.

A numerical approximation by J.F. Archard et al. 53 can be used to estimate the frictional temperature rise. First, the Peclet (Pe) number of the system was calculated:

푉 푎 Pe = 2 휅 Equation 3-S 1 where a is the radius of contact area, 휅 (푚2/푠) is thermal diffusivity and V (m/s) is the sliding velocity. The Peclet number here is 0.006, which means the frictional heat was dissipated quickly by thermal diffusion to the substrate. Under a steady state heat conduction approximation, the temperature increase (휃) due to frictional heat was estimated as

푄 휃 =  0.13 K 4 푎 퐾 Equation 3-S 2 where Q (W) is the total rate of heat generation, which could be calculated from the work of friction, and K (W/(K·m)) is the thermal conductivity of stainless steel.

Adsorption isotherm on metal surface

Figure 3-S1 shows the PM-RAIRS spectra of three vapors tested in this experiment. A polished copper was used, instead of stainless steel, for PM-RAIRS measurements since copper gives much stronger PM-RAIRS signals, due to higher IR refractivity, than stainless steel. All three vapors exhibited an isotherm behavior of type-II BET adsorption with small adsorption energy. The intermediate p/psat region (from ~15 to 70-80%) shows a relatively weak dependence of the signal intensity on the vapor pressure and at p/psat higher than 70~80%, the signal intensity increases rapidly due to condensation. These isotherm behaviors suggested that the adsorbate coverage in the intermediate p/psat region is at least a monolayer. 47

Figure 3-S 1 PM-RAIRS spectra of (a) α-pinene, (b) pinane and (c) n-decane adsorbed on a metallic substrate. The insets show the adsorbate signal intensity as a function of relative partial pressure of the vapor.

Elastic modulus calculation using Johnson-Kendall-Roberts (JKR) model

The JKR theory is widely applied to represent the interaction during AFM indentation for compliant sample. D.M. Ebenstein et al.56 gave detailed derivations of the equation used in our study for JKR model fitting. Representative raw indentation test data (force-distance curves) are shown in Figure 3-S2. For the JKR fitting, it is critical to find the zero indentation point. Since some tribo-polymers were transferred to the tip surface during the indentation and friction test, the snap-in behavior was not observed. In this case, the zero indentation point could be estimated 48 by the cross-point of the extrapolated lines of the linear portion of the loading curve and the free- standing portion, as shown in Figure 3-S2(a). The same method was used for the zero deflection point for the data in Figures 3-S2(b) and 3-S2(c). Figure 3-S2(d) and 3-S2(e) show the difference of the indentation tests on stainless steel surface with the clean tip and the used tip which just did indentation on tribo-polymers. The stronger adhesion force in 3-S2(e) results from the tribo- polymer adhered on the tip, which shows the transfer property of tribo-polymer. Figure 3-S3 displays the JKR fit results of the 60nN and 120nN indentation test results. The modulus estimated from the JKR fit was listed in Table 3-1 of the main paper.

Figure 3-S 2 Raw indentation test data (force-distance curves) of (a) α-pinene tribo-polymer (b) pinane tribo-polymer (c) n-decane tribo-polymer (d) stainless steel surface with clean tip (e) stainless steel surface used tip. In (a), the black lines show the extrapolations of the linear portion 49 of the loading curve and the free-standing line. The cross point of these extrapolated lines was assumed to be zero indentation position in the tip-sample distance calculation.

50

Figure 3-S 3 JKR model fit of the unloading curves of (a,b) α-pinene tribo-polymer, (c,d) pinane tribo-polymer, and (e,f) n-decane tribo-polymer. 51

Chapter 4

Mechanochemistry of physisorbed molecules at tribological interfaces:

Molecular structure dependence of tribochemical polymerization

Part of this chapter is reproduced with permission from American Chemistry Society: He, X.; Kim,

S. H. Mechanochemistry of physisorbed molecules at tribological interfaces: Molecular structure dependence of tribochemical polymerization. Langmuir 2017, 33, 2717-2724.

Overview

Physisorbed molecules at a sliding solid interface could be activated by mechanical shear and react each other forming polymeric products which are often called tribo-polymers. The dependence of tribo-polymerization yield on applied load and adsorbate molecular structure was studied to obtain mechanistic insights into mechanochemical reactions at a tribological interface of stainless steel.

Three hydrocarbon precursors containing 10 carbon atoms  α-pinene (C10H16), pinane (C10H18), and n-decane (C10H22)  were chosen for this study. -Pinene and pinnae are bicyclic compounds with different ring strains. N-decane was chosen as a reference molecule without any internal strain.

By comparing the adsorption isotherm of these molecules and tribo-polymer products, the reaction yield was found to be proportional to the amount of adsorbed molecules. An Arrhenius-type analysis of the applied load dependence of the tribo-polymerization yield revealed how the critical activation volume (V*) varies with the structure of adsorbed molecules. The experimentally- determined V* values of -pinene, pinane, and n-decane were 3%, 8%, and 10% of their molar 52 volumes, respectively. The molecule with the largest ring strain (-pinene) showed the smallest

V*, which implies the critical role of internal molecular strain in the mechanochemical initiation of polymerization reaction. The tribo-polymer film synthesized in situ at the sliding interface exhibited an excellent boundary lubrication effect in the absence of any external supply of lubricant molecules.

53

Introduction

Mechanochemical reactions initiated by a mechanical force or action are ubiquitous in nature and engineering systems. The most obvious example would be chemical reactions induced by frictional heat at sliding interfaces. There are numerous chemical reactions occurring at mechanical interfaces where frictional heat generation is negligible.9,26 In such cases, chemical reactions must be initiated or driven by directly channeling mechanical force or energy into chemical reaction coordinates which results in destabilization of reactants or reduction of the activation barrier for chemical reactions.1,11 Thus, chemical reactions that usually do not occur in thermal, photochemical, or electrochemical conditions can take place at mechanical interfaces.

However, fundamental understanding on how mechanical actions alter or control reaction dynamics or mechanisms is not well established.104,105 This is in part due to experimental difficulties in isolation and purification of reaction products as well as in-situ monitoring of reactant consumption and product formation.106,107

In tribology, the influence of mechanically applied forces on interfacial phenomena such as frictional dissipation, wear and tribochemical reactions has been modeled with a concept that an externally applied force facilitates the thermal transition of atoms or molecules across an energy barrier, thereby promoting slip or bond dissociation. Such mechanically-assisted thermal activation was first proposed by Prandtl and Eyring, and further developed and applied to a wide range of interfacial phenomena.10,26,29–31,34,108 One common feature of all proposed models is that they can be modified and expressed as an Arrhenius-type equation:

퐸 −퐹∙∆푥∗ 퐸 −휎∙∆푉∗ Reaction rate or yield = 퐴 ∙ 푒푥푝 (− 푎 ) = 퐴 ∙ 푒푥푝 (− 푎 ) Equation 4- 1 푘푇 푘푇 54

where A is the preexponential factor, 퐸푎 is the activation energy for typical thermal reactions, k is the ideal gas constant (1.38 × 10-23 J/K), and T is the system temperature; in this equation, the applied mechanical energy is written in two forms: force (F) times critical bond length (∆푥∗) or stress (휎) times critical activation volume (∆푉∗). Equation (1) implicates that the effect of applied mechanical energy is to lower the thermal activation barrier of interfacial slip or chemical reaction. Here, ∆푥∗ can be considered as the minimum amount of bond length change needed to reach the transition state. For example, when the Morse potential of a chemical bond is perturbed by the applied force producing a local maximum in the bond dissociation channel,2 then

∆푥∗ would be the distance between the equilibrium bond distance (at which the potential energy is minimum) and the length where the potential energy is maximum. In laboratory experiments, it would be difficult to control the force applied to individual atoms or molecules; thus, the contact pressure or mechanical shear (휎) would be a more appropriate term to consider. Then, ∆푉∗ must be the minimum amount of change in molecular volume to initiate chemical reactions of that specific molecule.

This paper reports how ∆푉∗ varies depending on the structure of precursor molecule sheared at the sliding interface. This study was carried out under a vapor-phase lubrication (VPL) condition which provide several advantages for experimental study of mechanochemical reaction mechanisms. In VPL, the supply of reactants to the sliding solid interface can be easily controlled through vapor adsorption isotherm.81,109 In typical liquid-based lubrications, reaction intermediates at the interface are readily dissolved into the lubrication oil, which make the detection of reaction products extremely difficult. In VPL, the reaction product with lower vapor pressure will remain on the substrate,23,71 allowing detection and quantification of the mechanochemical reaction products after VPL tests. Also, the mechanical wear of substrate can be fully suppressed in VPL 55 conditions,21,62 eliminating the need to consider the chemical reactions induced or caused by dangling bonds of the worn surface.20

In this work, we investigated the polymerization reaction of α-pinene (C10H16), pinane

(C10H18), and n-decane (C10H22) at the sliding interface of AISI 440C stainless steel. Their molecular structures are shown in Figure 4-1. These ten-carbon molecules have different internal strains. Both α-pinene and pinane have strained 4-membered ring.88,110,111 The presence of C=C double bond in the 6-membered ring significantly increases the internal ring strain of α-pinene, compared to pinane. N-decane is a linear alkane without such strains, so it can be a good reference for comparison. The critical activation volume of these molecules for polymerization reactions under VPL conditions was determined from load dependence of the polymer production yield. The tribochemically synthesized organic products, often called tribo-polymers, exhibited an excellent boundary lubrication with limited self-healing capability. The demonstrated in-situ synthesis of lubricating polymer films could be employed as a means to lubricate micro-mechanical devices.

Experiment Details

All friction and wear tests were performed using a custom-made reciprocating ball-on-flat tribometer with an environment control capability.21,78,112 The substrate was flat plates of AISI

440C stainless steel polished with sandpapers and then a micro-polish solution with 1 µm colloidal alumina slurry (Buehler). The root mean square (RMS) roughness of the polished stainless steel surfaces was around 20-25 nm. Commercially available 3mm diameter 440C stainless steel balls

(McMaster-Carr, Ohio, USA) were used as a counter surface to rub against the flat stainless steel surface. The RMS roughness of the ball surface was below 10 nm.21,113,114 The ball sliding span 56 and speed were 2.5 mm and 4 mm/s, respectively. Before tribo-tests, both substrate and ball surfaces were cleaned with ethanol and then UV/ozone treatment for 30 minutes to remove all chemical residues.52 The applied load was adjusted so that the Hertzian contact pressure varied in a range from ~0.4 GPa to ~1.5 GPa.

The tribotest were carried out in a continuous-flow cell with a desired relative partial pressure (푝/푝푠푎푡) of vapor in dry nitrogen which was produced by mixing a dry nitrogen stream

78 (푁2 ) and the saturated organic vapor stream of interest. In this study, (+)-α-pinene (2,6,6- trimethylbicyclo[3.1.1]hept-2-ene; CAS No. 7785–70-8), pinane (2,6,6- trimethylbicyclo[3.1.1]heptane; CAS No. 473-55-2), and n-decane (CAS No. 124-18-5) were chosen to investigate the effects of molecular structure (especially, internal ring strains) on mechanochemical reactions. All friction and wear tests were conducted at room temperature and ambient pressure. In our tribo-test conditions, the maximum flash temperature was estimated to be less than 20 oC due to slow sliding speed and low friction.115,116 Since thermal decomposition temperature of α-pinene and pinane are ~350 oC and 450 oC, respectively, thermally activation reactions of the adsorbed molecules due to frictional heat is unlikely.111

The adsorption isotherms of three precursor molecules were measured with polarization- modulation reflection-absorption infrared spectroscopy (PM-RAIRS) using a Thermo-Nicolet

Nexus 670 spectrometer equipped with a MCT detector. Since the 440C stainless steel has a poor reflectivity of IR, a polished copper substrate was used as a model surface to conduct PM-RAIRS experiments. The higher IR reflectivity of copper made it easier to detect the adsorbed species.117

In previous studies, it was already shown that the adsorption isotherm on stainless steel in ambient conditions is similar to that on copper.78,117 The details of the PM-RAIRS system setup and 57 experimental procedure for the adsorption isotherm measurement were discussed in a previous paper.23

The tribo-polymers created during the sliding process were spread over the wear track and piled up at the periphery of the wear track. Atomic force microscope (AFM; Digital Instrument

MultiMode scanning probe microscope) was used to measure the yield of tribo-polymer products.

Three-dimensional topographic images of tribo-polymers at different segments of the wear track were collected through tapping-mode scanning in ambient air; the volume of each segment was added to calculate the total amount of products. AFM tips (PPP-FM) were purchased from

NANOSENSORS푇푀 and cleaned with UV/ozone to remove organic contaminants before imaging.

The nominal tip radius was about 7~10 nm and the spring constant of the cantilever was calibrated to be 2.81 N/m using the Sader method.118 The effect of vapor environment on the viscoelastic property of tribo-polymer was tested by performing force-distance indentation measurements in a controlled vapor environment. The maximum indentation force was set to 60 nN for all measurements. The wear track on the substrate after removal of tribo-polymers via ethanol rinsing was analyzed with optical profilometry using a Zygo NewView 7300 system.

Results and Discussion

The effects of -pinene, pinane, and n-decane vapor adsorption on friction and wear behaviors of 440C stainless steel were tested in a contact load ranging from 0.39 GPa to 1.54 GPa.

Figure 4-1 displays the friction coefficients measured in α-pinene, pinane and n-decane vapor 58

Figure 4- 1 Coefficient of friction of 440C stainless steel measured in dry N2 and nitrogen containing -pinene, pinane, and n-decane vapor at marked contact loads. The partial pressure for the organic vapor was 30% of its saturation value (푝/푝푠푎푡). Error bars represents standard deviation calculated from more than 3 measurements. The inset shows the line profile across the wear track (left), optical profilometry image (center), optical microscope image (right) of the wear track made in dry N2 tribo-testing.

conditions at the lowest and highest normal contact pressures tested in this study. The results of tribo-testing in dry N2 are also shown for comparison. In the absence of any organic vapor in the surrounding gas phase, the friction coefficient is high and very unstable. In contrast, the friction coefficients measured in the presence of 30% p/psat of -pinene, pinane, and n-decane are around

0.17 and 0.2, regardless the adsorbate molecular structure and the contact pressure. The value is 59 slightly higher than the typical coefficients observed in VPL conditions.21,71 It might be due to the presence of tribo-polymers produced inside the slide track.23

It is noted that even at the contact pressure lower than the yield strength of 440C stainless steel (0.42~0.45 GPa),119 the substrate wears badly producing 3~4 m deep trenches after 600 cycles of reciprocating motion of the ball at a contact pressure of 0.39 GPa (Figure 4-1 inset). The roughness of the bottom of the wear track (~1.1 µm) is much larger than the roughness of the

78 pristine polished surface, implying that adhesive wear occurs in dry N2. When organic vapors adsorb at the sliding interface from the gas phase, then such adhesive wear processes can be prevented and the adsorbed molecules can provide lubrication effects. Thus, when the contact load is lower than the yield stress of the substrate, the wear mark in the substrate is almost invisible

(Figure 4-2; 0.39 GPa cases). When the contact pressure is increased above the substrate yield strength, plastic deformation take place and smooth compression mark is left on the substrate

(Figure 4-2). In the case of n-decane, the substrate wear was observed when the contact pressure was increased to >1 GPa; thus, the maximum load for the n-decane test was limited to 0.83 GPa. 60

Figure 4- 2 Cross-section line profile (left), optical profilometry image (middle), and optical microscope image (right) of sliding tracks after friction tests at the lowest and highest Hertzian contact pressure

Figure 4- 3 Tapping-mode AFM images of tribo-polymers accumulated at the ends and sides of the slide track after 600 cycles of reciprocating slide at the lowest and highest contact loads tested in (a) α-pinene, (b) pinane, and (c) n-decane vapors (30% p/psat). The area and height scales of the image are shown above each panel of images.

61

After tribo-tests for 600 reciprocating cycles, iridescent organic deposits could be seen with naked eyes. Time-of-flight secondary ion mass spectrometry (ToF-SIMS) of the organic deposits revealed typical fragmentation patterns of polymeric materials.23 Fourier transform infrared (FT-

IR) spectroscopic analysis of the organic deposits formed in the -pinene tribo-test showed the spectral features consistent with the chemically-synthesized poly--pinene.23,120 Based on these previous analysis results, the organic deposits will be called tribo-polymers hereafter. The total amounts of the tribo-polymer produced during 600 cycles of friction tests at various contact loads were measured with AFM. Figure 4-3 displays representative images of tribo-polymers piled up at both ends of sliding track and remained in the middle of track. By adding the volume above the pristine surface level, the total amount of the tribo-polymer produced could be obtained.

In order to find the relationship between the relative partial pressure of vapor molecules in the gas phase and the tribo-polymer production yield, the adsorption isotherm of each vapor was measured with PM-RAIRS and compared with the amount of tribo-polymers measured with AFM.

It should be noted that the absolute thickness of adsorbate layers cannot be calculated from the

PM-RAIRS intensity; so, only qualitative dependence on the vapor partial pressure can be compared. In Figure 4-4, the PM-RAIRS intensities of the adsorbate molecules measured at various vapor pressures are superimposed with the total tribo-polymer amount produced at each vapor pressure. Although the data scatters, it can be seen that the tribo-polymer yield increases as the amount of the vapor adsorbed on the substrate surface increases. Note that in the presence of tribo-polymer at the sliding interface, the actural amount of organic molecules involved in tribo- chemical reactoins could be different from the adsorption isotherm obtained for the clean surface.

62

Figure 4- 4 Comparison of adsorption isotherm (black) from PM-RAIRS measurements and total yield of tribo-polymer (red) after VPL tests at various vapor pressures of (a) α-pinene (b) pinane and (c) n-decane. The applied contact load was 0.5 GPa and the total number of sliding cycles was 600.

For all three vapors tested in this study, the PM-RAIRS intensities as a function of vapor pressure shown in Figure 4-4 appear to follow the type-II BET adsorption isotherm behavior.23 At a vapor pressure of ~30% p/psat, the physisorbed molecule coverage appears to be at least one monolayer thick and the vapor pressure dependence of the tribo-polymer yield is relatively weak at that pressure range, thus making the data collection less susceptible to small fluctuations in the vapor pressure. So, the load dependence of the tribo-polymer yield was studied while keeping p/psat at 30% for -pinene, pinane, and n-decane. In Figure 4-5, the normalized tribo-polymerization yield is calculated by dividing the total tribo-polymer volume measured with AFM with the slide track area and the total slide time and plotted as a function of applied normal load. 63

Figure 4- 5 Load dependence of tribo-polymer yield for α-pinene (black), pinane (blue), and n- decane (red). The p/psat of each vapor was kept constant at 30%.

It is considered that the friction force F is proportional to the contact area by definition:

퐹 ∝ 휎 × 퐴 Equation 4- 2

Where 휏 is the interfacial shear strength or the frictional force per unit area. Through mathematical transformation by dividing both sides of eq 4-2 with normal load N; one can get the result as shown blow:

휎 ∝ 훼 × 푃 Equation 4- 3

This relationship has been observed in various conditions including wearless single asperity contact experiments.121,122 The results suggested a linear dependence between shear stress and normal pressure: 64

휎 = 휎0 + 훼푃 Equation 4- 4

While α can be assumed to be the same as the friction coefficient and σ0 is a constant related to surface adhesion. In the case of this experiment set up, the surface roughness is negligible compared with Hertzian indentation depth; the surface adhesion term τ0 can be considered as 0.

The linear trend in the semi-log plot (Figure 4-5) clearly shows the exponential dependence of the tribo-polymerization on the applied load. Thus, eq. 4-1 can be used to determine the critical activation volume (∆푉∗).10,26,31,34 The shear stress term σ in eq. 4-1 could be expressed as a linear function of normal pressure P as eq. 4-4.123 Inserting eq. 4-2 into eq. 4-1 gives the following equation:

∗ ∗ 휎0 · ∆푉 퐸푎 ∆푉 · 훼 푙푛푘 = (푙푛퐴 + − ) + 푃 Equation 4- 5 푘푏푇 푘푏푇 푘푏푇

Since the sliding speed and friction in our experiment were low enough to keep the flash temperature less than ~20 oC and the average temperature increase less than 1 oC,124 the terms in the parentheses in the right-hand side would be constant. Thus, the slope of the semi-log plot in

Figure 4-5 can be used to calculate ∆푉∗ from eq. (3). In the absence of any wear of substrate

(Figure 4-2), the pressure dependence of shear stress (α) can be assumed to be the same as the friction coefficient (Figure 4-1).

Table 4-1 compares the ∆푉∗ value for -pinene, pinane, and n-decane calculated from the load dependence measurements and the molar volume (Vm) of each molecule. The molar volume of the physisorbed molecules would be the same or comparable with their molar volume in the

∗ liquid state. It is noted that the ∆푉 /푉푚 ratio is the smallest (~3%) for -pinene, larger (~8%) for pinane, and the largest (~10%) for n-decane. 65

Table 4- 1 Comparison of the critical activation volume (∆푉∗) calculated from Figure 4-5 and the molar volume calculated from the density of liquid.

Critical activation ∆푉∗ Adsorbed molecule ∗ Molar volume Vm × 100% volume ∆푉 푉푚 -pinene 8.2 ± 0.6 Å3 263.8 Å3 3.1 % Pinane 21.9 ± 0.7 Å3 272.1 Å3 8.0 % n-decane 33 ± 3 Å3 323.8 Å3 10 %

The magnitude of ∆푉∗ can be considered as how much the molecule should be distorted to initiate and undergo tribo-chemical polymerization reactions. Then, the trend observed for - pinene, pinane, and n-decane (Table 4-1) reveals the effect of molecular structure on tribo- chemical reactions. The most significant difference among three molecules tested in this study is the ring strain.88,110,111 The 2s and 2p orbitals of the carbon atom involved in a covalent single bond are hybridized into the tetrahedral sp3 orbitals where the ideal bond angle is 109.3o. However, the carbon atoms in the 4-membered ring of -pinene and pinane cannot accommodate this ideal bond angle for the tetrahedral geometry and thus involves considerable bending of the bond angle. This is known as the angle strain. The presence of the 4-membered ring attached to the 6-membered ring also twists the arm-chair conformation of the 6-membered ring which normally has very little strain in the bond angle. In the case of -pinene, the C=C double bond in the 6-membered ring produces additional ring strain. The sp2 hybridized carbon atoms in the C=C double bond imposes the 120o bond angle, which induces distortion of other bond angles in the 6-membered ring. Thus, it is expected that -pinene can easily undergo ring-opening upon small deformation or distortion from its equilibrium structure by external forces at the sliding interface, giving the smallest ∆푉∗ 66 among three molecules studied in this work. The exact mechanism of tribo-polymerization reactions after the initial activation is beyond the scope of this experimental study and may require computational simulations.

In contrast, n-decane has no ring strain. In addition, the C-C single bonds in n-decane are almost freely rotatable since the rotational energy barrier due to steric hindrance (about 4 kcal/mol) is comparable with the thermal energy at room temperature (~2.5 kcal/mol). Thus, n-decane can conform easily to accommodate external confinements imposed by the solid surface.125–128 Thus, it requires a large degree of deformation to dissociate or produce reactive intermediates for polymerization reactions. Note that liquid is considered incompressible. Thus, 10% change in the molecular volume is extremely large distortion from its equilibrium state.

67

This is the reason that n-decane is difficult to activate; but, it is also remarkable that even chemically-inert n-decane can be activated and involved in polymerization reactions at the tribological interface. This shows the highly non-equilibrium nature of the tribo-chemical reaction pathway.

Figure 4- 6 (a) Lubrication effect of the tribo-polymer films after the lubricant vapor supply is stopped. (b) Extension of the lubrication effect by dragging the tribo-polymer piled outside the slide track into the slide track. (c) Maintaining the lubrication effect by periodically replenishing tribo-polymers. The tribo-polymer film synthesis (marked with yellow background) was carried out by rubbing stainless steel surfaces for 600 cycles in the presence of 30% p/psat α-pinene. The applied load during the entire friction test was 0.5 GPa. 68

The tribo-polymer created in situ in the slide track in VPL conditions could provide boundary lubrication effects even in the absence of any external supply of lubricant molecules to the sliding interface. Figure 4-6 demonstrates the lubrication efficiency of tribo-polymers produced from α-pinene. In this test, tribo-polymers were produced by rubbing the surface for 600 cycles at a ~0.5 GPa load in 30% p/psat α -pinene vapor and then the vapor supplied was discontinued while the friction coefficient was measured continuously. The data in Figure 4-6 shows that when the α-pinene vapor supply is stopped, the friction coefficient increases slightly from ~0.17 to ~0.25 and remains steady for an extended period. The slight increase at the cessation of the vapor flow is due to the change in mechanical properties of the tribo-polymer film. The monomer molecules absorbed from the vapor phase into the tribo-film can act as a plasticizer, making the tribo-polymer film more viscoelastic. The viscoelastic nature of the tribo-film by absorption of α-pinene from the vapor phase can be seen in the indentation test of the tribo-film

(see Supporting Information). It is expected that the harder polymer film could give a higher friction than the viscoelastic polymer film.

In Figure 4-6a, the friction coefficient rises suddenly to ~0.7 and becomes chaotic at ~9000 cycles. At this point, the tribo-polymers are almost completely squeezed out of the slide track due to the reciprocating motion of the counter surface. When the sliding span is increased momentarily, then tribo-polymers piled outside the slide track (Figure 4-3) can be pulled or dragged into the slide track again. This can extend the lubrication effect for an additional ~6000 cycles (Figure 4-

6b). A more reliable way to extend and maintain the lubrication effect by the tribo-polymer films would be to replenish them periodically before they are fully consumed or squeezed out. This is demonstrated in Figure 4-6c where the α-pinene vapor supply is turned on for 600 cycles after every ~8000 cycles of sliding without any vapor supply. These results suggest that tribo-films 69 synthesized in situ within the slide track could be employed for selective lubrication of small mechanical systems such as microelectromechanical system (MEMS) devices without contaminating other areas where lubrication is not necessary.25,129,130

Conclusion

Mechanochemically-induced polymerization reactions occurring at a tribological interface were studied using a ball-on-flat sliding geometry in a vapor phase lubrication (VPL) condition.

In conventional lubrication systems utilizing liquid lubricants, the mechanistic study of mechanochemical reactions is difficult since reaction products are often removed from the production sites and dissolved into the liquid phase. This difficulty can be circumvented by choosing precursors that can be supplied via gas phase and produce polymeric products which have low vapor pressures and thus remain at the site of mechanochemical synthesis. The comparison of the tribo-polymerization yield with the structure of precursor molecules revealed the relationship between the mechanochemical activation at the sliding interface and the internal strain of the precursor molecule. It was demonstrated that tribo-polymer films synthesized in-situ at the sliding interface during the VPL condition can provide boundary lubrication effects for an extended period in the absence of any external supply of lubricant molecules to the sliding interface.

70

Supporting Information

Force-distance indentation curves of -pinene tribo-polymer in different vapor

environments

The plasticizer effect of vapor absorption for α -pinene tribo-polymer was tested by indentation measurements in dry N2, ambient air (relative humidity 40%), and 30% p/psat of α- pinene vapor. The indentation test was conducted at the end of the slide track where the tribo- polymer film was thick enough that the force-distance curve would not be affected by the

Figure 4-S 1 AFM force-distance curve of 훼-pinene tribo-polymer accumulated at the end of the slide track in (a) dry nitrogen, (b) ambient air, (c) 30% p/psat of 훼-pinene vapor. The piezo displacement was converted to the sample indentation depth.

131 substrate surface. The loading and unloading rate was 400 nm/s. In dry N2, the indentation and retraction slopes were largest and the hysteresis between these two curves was the smallest. After the indentation tests in dry N2, a significant amount of tribo-polymers was transferred to the AFM tip; so, subsequent measurements showed repulsive force at larger distances of the tip surface from the actual polymer surface. When 30% p/psat α-pinene vapor was introduced to the surrounding 71 gas phase, the slopes of loading and unloading curves decreased substantially and the loading- unloading hysteresis increased. This behavior indicated that the tribo-polymer film become more viscoelastic due to absorption of its monomer vapor. Similar plasticizer effect was observed upon introduction of humid air (relative humidity = 40%).

72

Chapter 5

Mechanochemistry at solid surfaces: Polymerization of adsorbed molecules by mechanical

shear at tribological interfaces

Part of this chapter is reproduced with permission from American Chemistry Society: Yeon, J.; He,

X.; Martini, A.; Kim, S. H. Mechanochemistry at solid surfaces: Polymerization of adsorbed molecules by mechanical shear at tribological interfaces. ACS Appl. Mater. Interfaces 2017, 9, 3142-3148.

The entire scope of this work was designed by S.H.K. and A.M. J.Y. carried out the ReaxFF simulations, and X.H. conducted experimental work. S.H.K., A.M., J.Y., and X.H. contributed to the manuscript writing. J.Y. and X.H. contributed equally to this work.

Overview

Polymerization of allyl alcohol adsorbed and sheared at a silicon oxide interface is studied using tribo-tests in vapor phase lubrication conditions and reactive molecular dynamics simulations. The load dependences of product formation obtained from experiments and simulations were consistent, indicating that the atomic-scale processes observable in the simulations were relevant to the experiments. Analysis of the experimental results in the context of mechanically-assisted thermal reaction theory, combined with the atomistic details available from the simulations, suggested that the association reaction pathway of allyl alcohol molecules induced by mechanical shear is quite different from chemically-induced polymerization reactions. Findings suggested that some degree of distortion of the molecule from its equilibrium state is necessary for mechanically- 73 induced chemical reactions to occur and such a distortion occurs during mechanical shear when molecules are covalently anchored to one of the sliding surfaces. 74

Introduction

Mechanochemical reactions are ubiquitous, but often unnoticed or considered atypical. An example of mechanochemistry is a chemical reaction that occurs at the sliding interface of two solid materials, often called tribochemical reactions. 4,5,132 Another example is synthesis of organic chemicals by collision of solid particles as in ball milling processes.104,133–135 These reactions are quite different from chemical reactions that occur upon heating, photon irradiation, or electrical bias. In thermochemistry, thermal excitation of molecules drives electrons in the ground state to go through a transition state along a specific reaction coordinate. The energy difference between the ground and transition states is called an activation barrier; this barrier could be lowered by using a proper catalyst to accelerate the chemical reaction rate. In photochemistry, the absorption of photons leads to excitation of electrons from the ground state to an excited state followed by propagation or relaxation of the excited electronic state to a lower energy state which is different from the initial state. In electrochemistry, transfer of electrons from an oxidizing species to a reducing species takes place when electrical contact is made between two reacting species with different electrochemical potentials; such redox reactions can be facilitated by application of electrical potential or bias. In contrast to these chemical reactions, where the transition or flow of electrons between electronic states leads to changes in atomic positions of molecules involved in reactions, a mechanistic understanding of mechanochemical reactions, in which a mechanical force alters reaction energies and pathways, is not well established. In this work, we are specifically interested in how mechanical force or stress can be transferred to molecules from external solid surfaces, displacing the molecular conformation from equilibrium states or positions, and how that can lead to chemical reactions involving changes in electronic states of reacting species. 75

At tribological interfaces, frictional energy can induce a significant increase in temperature, especially when the friction and sliding speed are high; such frictional heat can induce thermal reactions.6 When solid surfaces wear or fracture, high-energy particles such as electrons or photons can be emitted. These are called triboplasma and triboemission.136,137 Then, electrochemical or photochemical reactions can be accompanied. It is known that tribochemical reactions can take place even without substantial frictional heat or surface wear and fracture.9,10,23 In this case, external mechanical force or energy is assumed to be directly channeled into reaction coordinates, inducing distortion or dissociation of molecules. In such cases, the reaction rate or yield increases exponentially with the applied force or energy. 26,31 This is often modeled as a mechanically- assisted thermal reaction where the mechanical energy is effectively lowering the activation barrier; thus, the reaction rate (rp) or yield (ry) can be expressed in the following Arrhenius-type equation: 26,31

∗ ∗ 퐸푎−∆푥 ∙퐹 퐸푎−휎∙∆푉 ry= 퐴 ∙ 푒푥푝 (− ) = 퐴 ∙ 푒푥푝 (− ) Equation 5- 1 푘퐵푇 푘퐵푇

where A is the preexponential factor, kB the Boltzmann constant, and T the system temperature; here, the mechanical energy term is expressed as ∆푥∗ ∙ 퐹 at a given applied force or 휎 ∙ ∆푉∗ at a

9,10,26,30,31,34 given contact stress and counteracts the thermal activation barrier (퐸푎). In this context, the mechanical effect can be viewed as equivalent to the catalysis effect because it lowers the reaction barrier. In eq. (1), the proportionality constants, ∆푥∗ and ∆푉∗ , are called ‘critical activation distance’ and ‘critical activation volume’, respectively, based on the dimensional analysis. However, their exact physical meaning is still unclear.

Theoretically, it was shown that the dissociation channel of the Morse potential of a chemical bond can bend down under the influence of external force (F). 11,69,138 In that case, an

∗ 69 energy maximum point occurs at a distance ∆푥 away from the equilibrium bond distance (푥푒). 76

∗ The energy at 푥푒 + ∆푥 is much lower than the equilibrium bond energy (Eb); thus, bond dissociation can occur at a temperature much lower than typical thermal reaction temperatures.

Experimentally, ∆푉∗ can be obtained from the dependence of reaction yield or rate on contact pressure or shear stress ().10,31,34,139 Although, the magnitude of ∆푉∗ can be determined experimentally, its physical meaning related to mechanistic pathways is still ambiguous.

In order to obtain a mechanistic understanding of mechanical activation of chemical reactions, this work integrated experimental and computational studies of tribo-polymerization of allyl alcohol adsorbed at a sliding interface of silicon oxide. Experimental tribo-tests were conducted in a vapor phase lubrication (VPL) condition where frictional heat generation as well as surface wear are negligible (see Sections I and II of Supporting Information);21,23,71 thus, heat-, plasma-, and radiation-induced pathways are insignificant. Since silicon oxide is an insulator, electrochemical pathways can be ruled out. Computational molecular dynamics (MD) simulations were performed using a reactive force field called ReaxFF which can handle a sufficiently large number of atoms and molecules under dynamic and reactive environments. 140–144 We verified that

ReaxFF MD simulation results are relevant to the experimental data by comparing the slopes of the ln(ry) vs. contact pressure (P) plot from the experiment and the ln(rp) vs. P plot from the simulation results. Note that the magnitude of A in eq (1) varies depending on whether rp or ry is plotted since conversion of ry to rp requires multiplication by additional parameters; but, this does not alter the Arrhenius slope of the pressure dependence. MD simulations revealed the dominant reaction pathway for the association of allyl alcohol induced by mechanical shear, which is quite different from typical radical-based polymerization reactions. The physical meaning of the magnitude of ∆푉∗ is discussed based on the computation results.

77

Experimental Details

Tribo-polymerization of allyl alcohol at a sliding interface of silicon oxide was studied using a ball-on-flat tribometer capable of environmental control during friction testing.71 The substrate was thermally grown oxide layers on a silicon wafer and the counter surface was sodium borosilicate glass balls (Pyrex, thermal expansion coefficient = 3.3 ppm/K; diameter = 2.38 mm).

Borosilicates balls were used for the most of the experimental tests after confirming that the product yields obtained with silica balls and borosilicate balls were similar at one loading condition.

The use of borosilicate balls instead of silica balls was mainly because the former have much smoother surface, which eliminates or reduces asperity contact issues. Note that both borosilicate and silica balls show similar wear behaviors in dry and n-alcohol VPL conditions.21,145 The sliding track and ball surfaces were analyzed with optical profilometry (Zygo NewView 7300) after the tribo-test and no wear was observed. Due to surface charging problems, scanning electron microscopy (SEM) analysis for detection of micro-cracks smaller than the resolution of optical profilometry was not carried out. Even if there were a few micro-cracks detectable at the SEM resolution, their total surface area would be much smaller than the contact area (especially, in the absence of wear). Thus, the effect of micro-cracks on total reaction yields would be insignificant.

Tribo-polymer production yield was measured with atomic force microscopy (AFM)

(Digital Instrument, MultiMode). Tribo-polymer products were analyzed with X-ray photoelectron spectroscopy (XPS) (PHI, VersaProbe) (see section III of Supporting Information) and time-of- flight secondary ion mass spectrometry (ToF-SIMS) (PHI, TRIFT V nanoTOF equipped with a 30

2+ keV Bi3 ion gun). The average and maximum flash temperature increases were calculated to be

4 oC and 12 oC, respectively (see Section II of Supporting Information). The substrate wear was monitored with optical profilometry (Zygo, NewView 7300). In dry N2, the surface was always 78 worn, even at the lowest load tested in this experiment (0.25 GPa). However, there was negligible wear in the presence of allyl alcohol vapor with p/psat  10%, as evidenced by no discernable wear mark within the resolution limit of the optical profilometer after removing tribo-polymer products with ethanol rinse (see Section I of the Supporting Information).

Figure 5- 1 Snapshot of the MD simulation of allyl alcohol sliding between amorphous SiO2 slabs. The colors of the spheres represent atom types: tan = silicon, red = oxygen, green = carbon, and white = hydrogen.

The MD simulation configuration used to model mechanochemical reactions of allyl alcohol occurring at the sliding interface between two amorphous SiO2 slabs is shown in Figure 5-

1. The size of the periodic box was 31.9 × 31.9 × 70 Å3. Instead of the ball-on-flat geometry, a slab-on-slab geometry was used since the curvature of the counter-ball surface is effectively flat 79 at the nanometer length scale of the simulation and this geometry allowed more accurate control of number of molecules being sheared at the interface. The MD simulations consisted of three steps: (i) energy minimization and equilibration, (ii) vertical compression at a 1 m/s speed, and

(iii) lateral sliding at a 10 m/s speed for 1ns. The durations of the equilibration and vertical compression steps were carefully determined for each simulation based on the system potential energy and the position of slabs. To evaluate the effect of load on mechanochemical reactions, MD simulations were performed at six different contact pressure (P) conditions: 0.5, 1, 1.5, 2, 2.5 and

3 GPa. Also, to distinguish thermo-chemical reactions from mechanochemical reactions, additional MD simulations were carried out at temperatures of 300 K and 900 K, with no applied load or sliding. In all simulations, 75 molecules of allyl alcohol were included, corresponding to approximately one monolayer of allyl alcohol molecules on the amorphous SiO2 surfaces. The bottom rigid body region was fixed during all simulations.

All MD simulations used the recently developed ReaxFF force field.146 Verification of the force field parameters was performed by comparison to density functional theory (DFT) calculations (see Figure 5-S4 of the Supporting Information). Note that if MD simulations were run with the NVE ensemble (constant number of atoms, volume and energy; i.e., adiabatic condition), then the flash temperature at a sliding speed of 10 m/s would be extremely high.

However, the flash temperature at the experimental condition was estimated to be only a few degrees; with the weak temperature dependence of allyl alcohol reactions in the absence of active catalysts (such as cationic initiators), the experimental sliding contact could be assumed to be isothermal. For this reason, simulations were carried under the NVT (constant number of atoms, volume and temperature) ensemble at 300 K. In this case, the frictional heat is dissipated to the heat sink, keeping the simulation temperature at 300K even at a sliding speed of 10 m/s. For 80 comparison with thermally-activated reaction pathways, separate simulations were carried out for the 900 K case without any mechanical compression or shear.

Result and Discussion

Allyl alcohol molecules physisorbed at the sliding interface can not only provide lubrication effects, 21,62 but they also can react with each other, forming polymeric deposits in and around the slide track.71 These polymers are often called tribo-polymers. The AFM images shown as insets in Figure 5-2a show the tribo-polymers piled at both ends of and spread along the slide track. In the VPL experiment, the adsorbed or absorbed allyl alcohol evaporates from the tribo- polymer once the sample is removed from the environment-controlled tribotest cell. This allows chemical analysis of the tribo-polymer, which is one of the main advantages of studying tribochemistry in the VPL condition than the liquid-phase lubrication. ToF-SIMS analysis shows ions desorbing from the tribo-polymer areas whose masses are higher than the molecular weight of allyl alcohol (58 amu; Figure 5-2a). Due to severe fragmentation of organic molecules in ToF-

SIMS analysis, the molecular weight distribution of intact polymeric species cannot be determined;

+ + but, the presence of ion peaks separated by the mass of CH (13 amu) and CH2 (14 amu) are consistent with typical fragmentation patterns of polymeric species. XPS analysis of tribo-polymer products revealed that the atomic ratio of C/O is 3.7  0.2 (see Section III of Supporting

Information), which is higher than that of the monomer (3.0). In our previous paper on tribo- reactions of allyl alcohol on stainless steel surfaces, we analyzed the tribo-polymer products with

IR.23 These analyses confirmed the production of polymeric (or oligomeric) species; but, the exact chemical composition or molecular structure cannot be determined from XPS, ToF-SIMS, and IR analyses. It might require NMR analysis, which requires a relative large quantity of sample. 81

Figure 5- 2 (a) Semi-log plot of the tribo-polymer yield (ry, normalized with the contact area and sliding time) versus applied contact pressure. The synthesis condition was as follows: sliding speed

= 0.4 cm/s; sliding span and cycle = 2.5mm and 600 cycles; allyl alcohol vapor = 30% p/psat; temperature = 295 K. Insets are AFM images of the left, middle, and right regions of a slide track tested at a contact pressure of 0.45 GPa, the mass spectrum of tribo-products from ToF-SIMS analysis, and the selected ion map from ToF-SIMS imaging. (b) Semi-log plot of the tribo-polymer production rate (rp, calculated from the number of molecules containing more than 3 carbon atoms) from MD simulations versus the normal contact pressure. The simulation condition was as follows: sliding speed and time = 10 m/s and 1 ns; number of allyl alcohol molecules = 75; temperature =

300 K. Inset images show the structures of an allyl alcohol molecule and an intermediate forming a dimer.

82

Note that the contact pressure in the ball-on-flat geometry is not constant. It would be ideal to use the flat-on-flat test geometry in experiments; however, the truly constant pressure in the contact region cannot be obtain due to asperity contacts. For this reason, we did not compare the friction force or absolute yield at the same contact pressure condition; instead, the pressure dependence of the tribochemical reaction described by eq. (1) is physically more meaningful and can be compared with simulation results. If one wants to compare the effect of different contact geometry in experiments, then the pressure distribution within the contact region should be taken into account. But, as long as the contact geometry is kept constant (as in our experiment), the pressure dependence of the reaction yield is still valid for the purpose of this study.

The tribo-polymerization yield (푟푦 ) could be obtained from the total amount of tribo- polymers imaged with AFM along the slide track. When this yield is plotted on a logarithmic scale against the applied contact pressure (P); a linear relationship with a slope of 0.66  0.04 can be seen clearly (Figure 5-2a). Since the frictional temperature rise is negligible, the temperature term in eq.(1) can be assumed to be constant. If the shear stress () is proportional to P, i.e. 휎 = 휎표 +

∗ 123 휇∙∆푉 휇 ∙ 푃 where 휎표 and 휇 are constant, then, eq.(1) can be simplified as 푟푦 ∝ 푒푥푝 ( 푃). Then, 푘퐵푇

휇∙∆푉∗ the slope of the semi-log plot of 푟푦 vs. P corresponds to . In the absence of substrate wear, the 푘퐵푇 constant term  can be assumed to be the same as the friction coefficient, 21,22,62 which was measured to be ~0.3-0.35 (see Figure 5-S1 of the Supporting Information). Then, the ∆푉∗ value for the tribo-polymerization of allyl alcohol at the silicon oxide interface is determined to be 7.8 

0.6 Å3. This value is ~7% of the molar volume of allyl alcohol in the liquid state. Since the molecule is incompressible, the distortion of its volume by ~7% is a significant change. Likely, this is the reason that tribo-polymerization reactions cannot occur at normal conditions without 83 substantial mechanical stress. It should be noted that allyl alcohol is not polymerized readily via a conventional free-radical polymerization mechanism due to the resonant stabilization of allylic radicals.147,148 Thus, the polymerization of allyl alcohol upon mechanical shear at the tribological interface must occur via pathways that are different from conventional chemical polymerizations.

In order to shed light on the tribo-polymerization mechanisms, we carried out ReaxFF MD simulations of allyl alcohol being sheared at the interface of two amorphous silicon oxide surfaces.

Due to computational limitations, MD simulations cannot be performed at the same length and time scales as the experiments. Thus, it is important to verify the relevance between the experimental and computational results by comparing a variable that can be obtained from both approaches. Here, we used the Arrhenius slope of the pressure dependence of tribo-polymerization reaction described by eq 5-1.

MD simulations showed that allyl alcohol reacts during the slide; the reaction products include chemisorbed species, fragments, and larger molecules formed by association of two or three allyl alcohol molecules. Since the experimental work can detect only the association products with low vapor pressures, we focus on the simulation products with chain lengths longer than three carbons. By fitting the temporal profile of association products with an exponential function (see

Figure 5-S5a in the Supporting Information), we can calculate the reaction rate constant (rp) at a given contact pressure. When rp is plotted against P on a semi-log plot (Figure 5-2b), the slope is

0.92  0.13, which is reasonably close to the experimentally determine value (0.66  0.04). This confirms that ReaxFF MD simulation results are relevant to real reactions taking place in the experiment, although the contact load and sliding speed are quite different due to computational limitations.

84

Figure 5- 3 Snapshot images from the ReaxFF simulation showing the association paths of (a) two and (b) three allyl alcohol molecules during the slide at the SiO2/SiO2 interface (contact pressure

= 1 GPa; slide speed = 10 m/s; temperature = 300K). Only the atoms involved in mechanochemical reactions are highlighted and numbered for clarification purpose.

After confirming the pressure dependence in simulations is comparable with the experimentally-observed trend and ruling out micro-crack effects as well as frictional heat effects, we can seek into mechanistic information from simulation results to explain the experimental data observed for mechanochemical reactions. In order to compare with thermally-activated reactions, the simulation results for reactions at 900 K without sliding are also presented in this paper.

Figure 5-3a and 3b display snapshots of several molecules at various time steps during the sliding simulation, which provides critical information about how allyl alcohol molecules are associated. Figure 5-3a shows two molecules approaching and reacting with each other. At t = 5 ps, one molecule (C8-C7-C6-O5) is chemisorbed to the surface O9 atom at the C8 position. At t =

27 ps, the O5 atom of that molecule is covalently bonded to C4 of the other molecule. The C3 85 atom, which was involved in the C=C double bond with C4, is now attached the surface O11 atom.

Similarly, C7 is now bonded to the surface O10 atom. At t = 377 ps, the hydrogen attached to O5 is deprotonated and now moved to the surface O12 atom. Figure 5-3b shows the association of three molecules. Similar to Figure 5-3a, one molecule (O1-C1-C2-C3) is chemisorbed to the surface O13 atom before the inter-molecular reaction. At t = 16 ps, the O1 hydroxyl is covalently bonded to the C7 atom of the second molecule and its double-bond pair C8 is bonded to the surface

O16 atom. At t = 332 ps, the O5 hydroxyl group is bonded to the C11 atom of the third molecule and the entire molecule (trimer) is liberated from the surface.

When more molecules were monitored throughout the sliding period, two common features of mechanochemical association reactions were observed. First, one of the reacting molecules is initially chemisorbed at the moving solid surface. It appears that anchoring one molecule to the surface helps transfer the mechanical force or action from the solid surface to the other molecule being reacted. This may explain why it is rare to observe the same mechanochemical reactions in static high-pressure conditions without interfacial shear. Second, the majority of chemical bonds leading to association of allyl alcohol molecules are formed between the hydroxyl group of one molecule and one of the carbon atoms in the C=C double bond of another molecule, not between two C=C double bonds as in typical radical polymerization reactions. This is very different from the radical polymerization reactions of allyl alcohol initiated by Lewis acid catalysts where the carbon radicals are covalently bonded to the carbon atoms of other C=C double bonds.147,148 More details of the dynamics as well as statistical aspects of mechanochemical reactions can be seen in the radial distribution functions (RDFs) of various pairs of atoms. Figure 5-4 compares the RDFs of mechanochemical reactions from simulations of sliding at a 2GPa contact pressure and 300K 86 with the RDFs of thermal reactions from simulations run at 300K and 900 K without load or sliding.

Figure 5-4a shows the RDF of the distance between the first carbon atom from the hydroxyl group (C1) and the hydroxyl oxygen (Oa). The main peak at 1.45 Å corresponds to the equilibrium

149 bond length (xe) for C1-Oa of allyl alcohol. At 300 K without sliding, almost all C1-Oa bonds are populated at this equilibrium distance. As soon as the interface slides, a significant portion of these bonds are elongated to 1.55-1.64 Å; this must be an intermediate state. The critical

* dissociation length (xe+x ) would be at a longer distance than this and would not be observed in an RDF since it is the transition state existing only momentarily. A similar elongation of the C1-

Oa length can be seen for the thermal reaction at 900 K. 87

Figure 5- 4 Radial distribution functions (RDFs) for distances between carbon atoms (C1, C2, C3; numbered from OH group, shown as inset to Figure 5-4a), alcoholic oxygen atoms (Oa), surface oxygen atoms (Os), and silicon atoms (Si) at the interface of silicon oxide. Red curves are at 300K without load or sliding. Black curves are during the slide (contact load = 1 GPa; slide speed = 10 m/s; temperature = 300K). Green curves are at 900K without load or sliding. Each RDF is for a different atom-atom distance: (a) C1-Oa, (b) C1-Oa, (c) C1-Oa, (d) C1-Os, (e) C2-Os, (f) C3-Os,

(g) C2-C3, (h) Oa-Oa, and (i) Si-Oa.

In the RDFs of C2-Oa and C3-Oa (Figures 5-4b and 4c), similar intermediate distances can be seen at 1.56Å – 1.63 Å. The population of these intermediate states increases upon sliding at 2

GPa or heating to 900 K, compared to the 300 K. It is noted that the covalent bond formation 88 between C2 and Oa of two molecules (1.45 Å) can be seen only for the 2 GPa slide case (Figure

5-4b). The peak at 2.54 Å in the C2-Oa RDF corresponds to the intra-molecular distance between these atoms in one molecule. In the C3-Oa RDF (Figure 5-4c), the covalent bond formation between these atoms appears to be more likely at 900 K compared to the 2 GPa slide case. Note that at 300K without sliding, the sharp peak at 1.47 Å in the C3-Oa RDF is longer than typical C-

O single bond length (1.43-1.45 Å). This intermediate species is formed due to strong hydrogen bonding interactions, resulting donation of hydrogen from the hydroxyl group to a surrounding oxygen atom (see Figure 5-S5b in Supporting Information) and could be an intermediate state unique to the 300 K static state. Again, the two broad peaks centered at ~2.6 Å and ~3.7 Å are the intra-molecular distances between these atoms in the same molecule. Two values correspond to two different rotational conformers (gauche versus trans along the C1 sp3- C2 sp2 rotational axis).149

The surface oxygen (Os) atoms are involved in chemisorption of allyl alcohol molecules.

The peaks at 1.49-1.5 Å in the C2-Os RDF (Figure 5-4e) and 1.45 Å in the C3-Os RDF (Figure 5-

4f) correspond to the covalent bonds between the molecule and the surface oxygen. The C2-Os bond appears to be slightly longer than that of the C3-Os bond; this could be due to steric hindrance of C1 and C3 against the surface. In the 900K thermal reaction, the propensity of forming covalent bonds with Os seems to be higher at C3 than C2. The small peak at 1.45 Å in the C1-Os RDF

(Figure 5-4d) of the 2 GPa slide case might be formed when the C1 radical is formed by dissociation of the C1-Oa bond during the mechanochemical reactions and compensated by the surface oxygen (Os). The chemisorption probability through the C1-Os bond formation (Figure 5-

4d) appears to be lower than those through the C2-Os and C3-Os sites (Figures 5-4e and 4f). 89

In the C2-C3 RDF (Figure 5-4g), the sharp peak at 1.33 Å at 300 K is the C=C double bond length.149 At 300 K, a small fraction of the C=C double bond appears to be elongated to ~1.45 Å; this might be the consequence of the chemisorption of C2 and C3 to the Os site as indicated by a small peak at 1.45 Å in the C2-Os and C3-Os RDFs (Figures 5-4e and 4f) or intermolecular interactions between two allyl alcohols (leading to the 1.47 Å in the C3-Oa RDF in Figure 5-4c).

It is noted that, upon interfacial shear at 2 GPa and thermal activation at 900 K, the C=C double bond is almost completely disappeared and elongated to 1.4 – 1.5 Å. Again, this must be an intermediate state since its distance is shorter than typical length of C-C single bond (1.54 Å).149

In the Oa-Oa RDF (Figure 5-4h), the sharp peak at 2.62 Å corresponds to the O-HO hydrogen bond. This peak is prominent at 300K, reduced at the sliding condition, and completely suppressed at 900 K. Since the hydrogen bond is weak, it is not stable at 900 K. In the Si-Oa RDF

(Figure 5-4i), the peak at 1.57 Å must be due to the reaction of alcoholic hydroxyl with the Si dangling bond produced when the surface oxygen (Os) reacts with C2 and C3 of allyl alcohol molecules. This bond is formed only during the mechanical shear or at high temperature conditions and it is not formed in the static condition at 300 K. The peak at 2.1 Å, which is more abundant in the sliding case, is the distance between the hydroxyl group and the undercoordinated silicon (u-

Si) atom (see Figure S-5b in Supporting Information). Those u-Si species reacts with water and allyl alcohol molecules, forming Si-OHs or Si-ORs species during the sliding. The Si-OR species are important here since they are the intermediate species involved in tribo-polymerization reactions (Figures 5-3 and 5-4).

With the new insights obtained from ReaxFF MD simulation, we can attempt to interpret the physical meaning of the ~7.8 Å3 value for ∆푉∗ obtained from the experiments. When we compare the equilibrium C-O bond distance (xe) and the distances of C-O in the intermediate states, 90 it is clear that x* would be larger than 0.1-0.2 Å along the C-O bond axis. The exact value of x* at the transition state could not be obtained with ReaxFF MD simulations. Let’s assume that the

* upper limit of x would be comparable to xe (even though this is unlikely). Then, the critical volume change ∆V* would be only ~3 Å3 (i.e. cube of the length, 1.45Å, of the C-O bond that is dissociating or newly forming); obviously, this is much smaller than the experimentally determined value. Thus, a change over a longer distance must be involved. This leads to the premise that some degree of distortion of the entire molecule from its equilibrium state must occur to activate and initiate mechanochemical reactions. Such a distortion is possible during the mechanical shear of molecules covalently anchored to the solid surface by the counter-surface

(Figure 5-3). This is evidenced by the broadening of the intra-molecular distances in the C3-Oa

RDFs for the 2 GPa sliding case (Figure 4c).

Conclusion

This research explored the mechanistic details underlying tribo-polymerization of allyl alcohol molecules adsorbed and sheared at a silicon oxide interface. A comparison of the Arrhenius slopes of the pressure dependence of tribo-polymerization obtained from experimental measurements and ReaxFF MD simulations verified that atomistic details from the simulations can be used to explain reaction pathways and the physical meaning of the experimentally- determined critical activation volume (∆V*). It was found that mechanically-induced chemical reactions are initiated through distortion of the entire molecule from its equilibrium state, not just single bonds that are being dissociated or formed, and that such structural distortion of molecules is facilitated when molecules are covalently anchored to one of the sliding surfaces.

91

Supporting Information

I. Wear prevention by adsorption of allyl alcohol vapor and tribopolymer formation

The wear of the substrate after friction tests (Figure 5-S1a) was analyzed with optical profilometry. In dry N2, the surface was always worn severely, even at the lowest load (0.25 GPa) tested in this experiment (Figure 5-S1b). The friction coefficient was low, likely because wear debris particles made third-body rolling contact conditions.7,22,62 In the presence of allyl alcohol vapor with p/psat  10%, the substrate did not wear; after removing tribo-polymer products with ethanol rinse, there was no discernable wear mark within the resolution limit of the optical profilometer used in this study (Figure 5-S1c).

Figure 5-S 1 (a) Friction coefficients measured in dry N2 and allyl alcohol vapor (p/psat = 40 %).

Optical profilometry line profile and optical image of (b) wear track after tribo-test in dry N2 and

(c) slide track after tribo-test in allyl alcohol vapor and removal of tribo-polymers by rinsing with ethanol.

II. Calculation of flash temperature due to frictional energy dissipation 92

During the sliding process, the frictional energy could be dissipated as heat. The local temperature due to frictional heat is often called flash temperature.116 Various numerical models have been developed to estimate the average flash temperature rise, ∆푇푓. If the model proposed by

116 M.F. Ashby et al is used, then ∆푇푓 can be estimated as:

푙 ∆푇 = 0.5 µ푃푣 푓  3.9 oC 푓 푘 Equation 5-S 1 where µ is friction coefficient (see Figure 1a), P normal contact pressure (0.4 GPa), v sliding speed

푎√휋 (4 mm/s), lf linear diffusion length (= =14 m; where a is the radius of the Hertzian contact 2 area =16 m at a contact pressure of 0.4 GPa), and k thermal conductivity of silicon oxide (1.2

W/(m·K)).

Alternatively, the average temperature increase, ∆푇푎푣푔 , can be estimated using the model developed by J.F Archard et al.53 In this model, the Peclet (Pe) number is used to determine if the heat dissipation is dominated by advection or diffusion:

푣 푎 Pe = 2 휅 Equation 5-S 2 where a is the radius of the Hertzian contact area (16 m), 휅 the thermal diffusivity (8.710-7 m2/s), and v the sliding speed (4 mm/s). The Pe number calculated for our system is less than 0.04, meaning that the frictional heat will be dissipated quickly to the substrate through the diffusive transport. Then, assuming the steady state heat conduction, ∆푇푎푣푔 is estimated to be:

푄 ∆푇 =  4.6 ̊퐶 푎푣푔 4 푎 푘 Equation 5-S 3 93 where Q is the total rate of heat generation (420 W) calculated from the work of friction, and k is the thermal conductivity of silicon dioxide (1.2 W/(K·m)).

In the Archard model, the maximum flash temperature increase, ∆푇푓,푚푎푥 , can be estimated assuming that when the total applied load is absorbed by plastic deformation at a single area:53

휇∙푔∙휋∙푌 퐹0.5∙푣 ∆푇 = 0.25  12 oC 푓,푚푎푥 휌∙푐 휅(휋∙푌)0.5 Equation 5-S 4 where g is the gravitational acceleration (9.8 m/s), Y the yield pressure (155 MPa),  the density

(2.65 kg/m3), c the specific heat (730 J/(kgK)), and F the applied normal load (0.22 N). This value can be taken as the upper limit of the flash temperature increase.

III. XPS analysis of tribo-polymer product

Tribo-polymers piled up at the end of the slide track were analyzed with XPS. The XPS analysis spot size was less than 50 m in diameter and the tribo-polymer area was larger than that. Figure

S-2 shows the high-resolution XPS spectra of the O1s, C1s, and Si2p photoelectron regions. Four slide tracks were analyzed and the average atomic ratio of C/O was calculated (Table S-1). By fitting the C1s spectra with a Voigt line profile function, the relative concentrations of carbon species with different oxidation states were obtained (Table S-1). 94

Figure 5-S 2 XPS spectrum of tribo-polymer film piled at the end of sliding tracks. Tribo-polymers were synthesized by sliding a glass bead on a thermal oxide film on a silicon wafer at 0.4 GPa for

600 reciprocating cycles in 30% p/psat allyl alcohol vapor.

Table 5-S 1 Chemical composition of tribo-polymer film produced from allyl alcohol at the sliding interface of silicon oxide.

C 1s high resolution fit Chemical C/O atomic species ratio C-C/C=C C-O C=O O-C=O

Tribo-polymer* 3.7  0.2# 74.8  1.4 % 15.6  0.5 % 7.0  0.5 % 3.4  0.5 %

Allyl alcohol+ 3 66.7 % 33.3 % - -

* From XPS analysis; + From molecular structure; # The contribution from the substrate was removed based on the Si 2p intensity and the O:Si ratio of the substrate measured outside the tribo- tested region.

95

IV. Details of ReaxFF MD simulation process

The ReaxFF force field was chosen to simulate the tribochemical reactions under shearing conditions in atomic scale.140,150 ReaxFF is one of the reactive empirical force field methods, which is based on bond order – bond distance relationship. This allows the ReaxFF force field to calculate not only short distance covalent bonds, but also the transient state of bond formation or dissociation and long range interaction, with comparable accuracy to DFT. The ReaxFF potential parameters applied in this study were originally developed for PTFE and PDMS interactions.146 In this force field, parameters for Si/O/H/C/F interactions are combined together for hydrocarbon and fluoro- carbon systems. The original C/H/O and Si/C interaction parameters were developed by

Chenoweth et al.,151,152 and then the carbon parameters were updated by Srinivasan et al. using

DFT.143 Based on this force field, parameters to describe the C-O, C-C-O and C-O-C angles, and the C-O-C-O and C-C-O-C dihedrals were updated to realistically simulate the behavior of carbon- oxide chain and bulk system, producing a comparable match with DFT.

To check the validity of the ReaxFF force field used in this study, the energy curves for hypothetical reactions of allyl alcohol molecules were calculated and compared with the DFT calculation results for the same reactions. All DFT calculations were performed with B3LYP, 6-

311G++ G(d,p) basis sets. Figure 5-S3 shows the energy profile of two association pathways of allyl alcohol molecules. The first reaction energy curve, depicted in Figure 5-S3a, was simulated for the reaction between two C=C bond bonds. The second case, shown in Figure 5-S3b, was for the reaction between Oa and C2 atoms, and donating hydrogen from alcoholic hydroxyl (Oa) to the surface oxygen (Os). In both cases, the agreement between ReaxFF and DFT calculation results was good. These results indicate that the ReaxFF force field used in this study will be able to simulate the reactions of allyl alcohol with an accuracy comparable with DFT. 96

(a)

(b)

Figure 5-S 3 Comparison of ReaxFF MD and DFT simulation results for reactions forming (a) the single bond between the carbon atoms of two C=C double bonds and (b) the covalent bond between hydroxyl oxygen (Oa) and one of the carbons in the C=C double bond with proton donation from

Oa to the surface oxygen atom (Os).

V. Reaction rates and intermediate species from ReaxFF MD simulations

Figure 5-S4 displays the temporal profiles of all species during the sliding for 1 nm at various contact pressures. One critical caveat of VPL is that the intermediates with high vapor pressures would be lost to the vapor during the tribotest. For this reason, we ignored the molecules containing less than 3 carbons observed in MD simulations (Figure 5-S4) when we compared the results with the experimental data. By fitting the temporal profiles of products with the number of carbons (x) 97

larger than three with an exponential function approaching a plateau value (푦(푡) = 퐴(1 − 푒−푟푝푡)), the rate constant (푟푝) at a given contact pressure (P) can be obtained. Then, from the slope of the semi-log plot of 푟푝 versus P shown in Figure 5-1b, the critical activation volume can be estimated.

Figure 5-S 4 Temporal profile of reactants, association products (x > 3), chemisorbed or partially dissociated species (x=3 or x<3) at a contact pressure of (a) 0.5 GPa, (b) 1 GPa, and (c) 2 GPa.

The black line is the result from fitting with a (1 − 푒−푟푝푡) function. Here, x is the number of carbon atoms in the molecule or intermediate species.

Figure 5-S 5 Snapshot images of (a) an intermediate with the C3 (dark green) and Oa (dark red) atoms at 1.47 Å from each other and (b) an intermediate with the Oa atom (purple) positioned at

2.1 Å from the under-coordinated Si atom (large gray). 98

In the RDF between C3 and Oa (Figure 5-4c), there is a sharp peat at 1.47 Å in the case of 300K without sliding. Figure S-5a shows the snapshot of an intermediate with a distance of 1.47 Å between the C3 and Oa atoms of two molecules. This bond formation between C3 and Oa appears to follow the hydrogen donation from Oa to a surrounding atom. Thus, it is speculated that the C3-

Oa reaction at 300K is the consequence of the strong hydrogen bonding interaction between Oa-

Oa (Figure 5-4i) or Oa-Os pairs; this may facilitate the hydrogen abstraction from the Oa atom and the formation of the C3-Oa covalent bond. The population of the 1.47 Å peak in the C3-Oa RDF decreases when the system temperature is increased to 900 K where hydrogen bonding interactions are insignificant due to high thermal energy (Figure 5-4i). The C3-Oa bond formation path is relatively rare in the sliding simulations (Figure 5-4c). Instead, the Oa-C2 bond formation is dominant during the sliding (Figure 5-4b). In contrast, the Oa-C2 bond formation appears to be insignificant in thermal reaction conditions (Figure 5-4b).

Figure 5-S5b shows the Oa atom of an intact hydroxyl (OaH) interacting with the under- coordinated silicon (u-Si) atom, which is responsible for the 2.1 Å peak in the Si-Oa RDF in Figure

4j. Such long-range interactions between the OaH group and the u-Si atom were frequently observed during the sliding condition where the u-Si species are more abundant, compared to the thermal reaction conditions at 300 K and 900 K without sliding (Figure 4j). This state is an unstable intermediate state; the covalent bond between Oa and u-Si is not possible until the hydrogen atom of the OaH is donated to another oxygen atom nearby.

99

Chapter 6

Mechanochemical association reaction of interfacial molecules driven by shear

Part of this chapter is reproduced with permission from American Chemistry Society: Khajeh, A; He,

X.; Yeon, J.; Martini, A.; Kim, S. H. Mechanochemical association reaction of interfacial molecules driven by shear. Langmuir 2018, 34, 5971-5977.

The entire scope of this work was designed by S.H.K. and A.M. A.K. carried out the ReaxFF simulations, J.Y. performed the energy barrier calculations, and X.H. conducted experimental work. S.H.K.,

A.M., A.K., and X.H. contributed to the manuscript writing.

Overview

Shear-driven chemical reaction mechanisms are poorly understood because the relevant reactions are often hidden between two solid surfaces moving in relative motion. Here, this phenomenon is explored by characterizing shear-induced polymerization reactions that occur during vapor phase lubrication of α-pinene between sliding hydroxylated and dehydroxylated silica surfaces, complemented by reactive molecular dynamics simulations. The results suggest that oxidative chemisorption of the α-pinene molecules at reactive surface sites, which transfers oxygen atoms from the surface to the adsorbate molecule, is the critical activation step. Such activation takes place more readily on the dehydroxylated surface. During this activation, the most strained part of the α-pinene molecules undergoes a partial distortion from its equilibrium geometry, which appears to be related to the critical activation volume for mechanical activation. Once α-pinene molecules are activated, association reactions occur between the newly attached oxygen and one 100 of the carbon atoms in another molecule, forming ether bonds. These findings have general implications for mechanochemistry since they reveal that shear-driven reactions may occur through reaction pathways very different from their thermally induced counterparts and, specifically, the critical role of molecular distortion in such reactions.

101

Introduction

Although not as well-known as thermal, photochemical, and electrochemical reactions, mechanochemical reactions are ubiquitous and play critical roles in natural and engineering systems. Mechanochemistry describes chemical reactions that are induced or facilitated by mechanical force or energy.153 In biology, examples include the unfolding of protein molecules upon mechanical stretch,154,155 shear-induced gene expression and enzyme activation, and mechanically stimulated differentiation of stem cells.156–158 Engineering examples include organic synthesis via ball milling,11 rearrangement of macromolecular scaffolds via ultrasonication,159 transformation of anti-wear additives forming protective coatings at tribological interfaces,8,34,39,160 and wear of silica and silicate glass surfaces in humid environments.109,145 Among these, chemical reactions at sliding solid interfaces are least understood because of the complexity and dynamic nature of the interfacial conditions which are continuously evolving during the shear.104,105 The lack of fundamental understanding of chemical reactions induced by shear hampers chemists’ ability to design new compounds, such as novel lubricants that can mediate energy waste and material loss at the technically challenging tribological interfaces present in all moving mechanical systems. Herein, we report experimental and computational studies elucidating molecular mechanisms of shear-induced polymerization reactions for a model system: -pinene molecules adsorbed on sliding interfaces of silicon oxide.

When mechanical or interfacial shear is involved, it is hypothesized that the applied shear stress can facilitate chemical reactions by decreasing the activation energy or changing the reaction path.27–29,161 These theories also posit the existence of a shear-induced activation volume which is critical to understanding how much a reaction energy barrier is decreased by shear.162,163 However, due to the complexity of processes during sliding, it is difficult to relate this activation volume to 102 specific physical change.70 In addition to mechanical shear, the surface chemistry of the sliding solids play critical roles in determining dominant reaction pathways and kinetics.164–166 Intuitively, the molecules chemically anchored to the sliding surface would experience more mechanical shear than the molecules slipping within the interface.37 However, the exact roles of surface chemistry and mechanical shear need to be unraveled for a deeper understanding of mechanochemistry.

Here, the experimental design was inspired by the fact that vapor phase lubrication (VPL) can effectively suppress material loss for most solid surfaces.81 The model system chosen was - pinene,167 which has high internal strain due to the presence of the four-membered ring and C=C double bond in the six-membered ring and undergoes mechanochemical polymerization reactions when sheared on a silica surface.168,169 The hypothesis studied in this work is that chemical reactions between the -pinene and the silica will facilitate oligomerization during sliding. The experimental results were used to validate reactive molecular dynamics (MD) simulations with a

ReaxFF reactive force field, which then revealed mechanistic insights into the shear-induced polymerization of -pinene on silicon oxide surfaces – especially, the role of chemisorption and shear-induced deformation of molecules which can be generalized to other mechanochemical reactions at sliding interfaces.

Methods

Experimental Details

Friction tests were conducted with a custom-built reciprocating ball-on-flat tribometer with an environment control capacity.23 The hydroxylated silica substrate was prepared by cleaning a silicon wafer with the RCA-1 solution (5:1:1 mixture of DI water, 30% ammonium hydroxide and 103

30% hydrogen peroxide at 70 oC) followed by rinsing with DI water and then exposure to

UV/ozone. The dehydroxylated silicon substrate was obtained by heating the hydroxylated surface at 450 oC in dry nitrogen for 12 hours, and then cooling to room temperature in dry nitrogen. The counter-surface in friction tests was sodium borosilicate glass balls (Pyrex; thermal expansion coefficient = 3.3 ppm/K; diameter = 2.38 mm). The surface roughness was estimated via optical profilometry to be ~4 nm after removal of the ball curvature. The concentration of alpha-pinene

21 was 30~40% p/psat to ensure the formation of a monolayer on the surface. The sliding speed was kept at 4 mm/sec; at this condition, the average flash temperature increase due to friction was estimated to be 4-12 oC only.23 The polymeric products within and aside the slide track were imaged with atomic force microscopy (AFM; Digital instrument, MultiMode). The mechanochemical reaction yield was estimated by normalizing the total volume of products with the sliding area and time. X-ray photoelectron spectroscopy (XPS; PHI, VersaProbe) was used to analyze the chemical composition of the reaction products accumulated in sliding over a 0.2  0.2 mm2 area.

Computer Simulations

Shear-induced mechanochemical polymerization of -pinene molecules between two amorphous silica slabs was investigated using ReaxFF-MD simulations (see Figure 6-1). Two amorphous SiO2 slabs were created as described in the Supporting Information. A slab-on-slab geometry was chosen for the simulation to represent the elastic flattening of the solid surface in the ball-on-flat contact zone. This also resulted in confinement of reacting molecules within the sliding interface.

Note that the roughness of the surfaces in simulations (~0.2 nm) was smaller than that of the 104 substrate in experiments (~1 nm) due to the length scale limitations of the model. In both cases, the elastic deformation depth under the normal load was larger than the surface roughness. Initially,

31 α-pinene molecules were confined between the amorphous silica slabs, to form approximately a monolayer of -pinene on each surface. The size of the simulation box was 3.19×3.19×8.00 nm3.

To capture the effect of surface reactivity on the yield, simulations were performed both on non- hydroxylated and hydroxylated surfaces. The non-hydroxylated surface was modeled as silicon oxide produced in the absence of water molecules [see Figure 6-S3]. The hydroxylated surface was created by modeling reactions with water at 500 K [see Figure 6-S4]. The density of defects decreased upon hydroxylation: for example, from ~1.2 oxygen radicals per nm2 on the non- hydroxylated surface to ~0.5 per nm2 on the hydroxylated surface [see Figure 6-S5]. Each MD simulation consisted of three main steps: (i) energy minimization and equilibration at 300 K until the energy was stable, (ii) compression by moving the top slab toward the bottom slab in the z- direction at a speed of 5 m/s until the average pressure was approximately equal to the target value

(either 1 or 3 GPa), and (iii) sliding the top slab at 10 m/s in the x-direction for 2 ns. Since the flash temperature has been shown to be negligible during sliding in experiments, all simulations were carried out at 300 K by applying the NVT (constant number of atoms, volume and temperature) ensemble. The ReaxFF force field was employed with parameters to describe interactions between Si/O/C/H atoms.170,171 Performing the ReaxFF simulations rather than the quantum mechanics calculations enabled us to investigate the dynamics of the system at a more extensive physical length scale and a longer timescale. Although the ReaxFF can provide a reasonable approximation of reactivity, it should be noted that the force field requires substantial validation for a specific system. MD simulations were performed using the Large

Atomic/Molecular Massively Parallel Simulation (LAMMPS) software172 and the post-processing 105 was carried out using OVITO software.173 Additional details are provided in the Supplement

Information.

Figure 6- 1 The configuration of the model system illustrating the three stages of the MD simulations. From left to right: energy minimization, compression, and sliding.

Results and Discussion

Figure 6-2a shows the friction coefficients measured during VPL for the fully hydroxylated and thermally dehydroxylated silica surfaces. Thermal dehydroxylation was used as a means of making the surface more reactive. Another approach would be to pre-condition the surface under poor lubrication conditions [see Figure 6-S1 in Supporting Information].174 Surface wear under poor lubrication conditions would expose dangling bonds that can readily react with adsorbed molecules. Thermal dehydroxylation will occur through dehydration reaction among adjacent hydroxyl groups forming siloxane bridges.174 Such siloxane bonds at the surface would be subject to high strain (due to unfavorable bond angles) and thus have high reactivity. 106

Figure 6- 2 (a) Friction coefficients of hydroxylated and dehydroxylated surfaces in -pinene VPL condition at a 40 % partial pressure relative to saturation. A borosilicate ball was used as a counter- surface. The applied Hertzian contact pressure was 0.32 GPa and the sliding speed was 4 mm/s.

Insets are AFM images of reaction products piled in and along the slide tracks. (b) C1s XPS spectra of the shear-induced polymers produced on silicon oxide surfaces with two different surface conditions. (c) Semi-log plot of the normalized tribo-polymer yield against the contact pressure for -pinene sheared on the dehydroxylated silicon oxide surface.

For chemical analysis purposes, we used the thermally dehydroxylation surface since it allows better control of the surface chemistry. The dehydroxylated surface became partially hydrophobic [see Figure 6-S2]. Atomic force microscopy (AFM) imaging after the VPL tests revealed accumulations of reaction products in and along the slide track (insets in Figure 6-2a).

The dehydroxylated surface had a higher friction coefficient and significantly more products than 107 the hydroxylated surface. The higher friction coefficient (~0.3) is within the typical range for polymer surfaces.175 The C1s XPS analysis of the reaction products (Figure 6-2b) shows the presence of oxygen groups (C-O at ~286.5 eV, C=O at ~287.5 eV, and O-C=O at ~289 eV), even though -pinene has no oxygen. The dehydroxylated surface produced more oxygenated products than the hydroxylated surface. These results imply that the surface chemistry of the solid substrate plays a critical role in reaction yield and pathways.

For the dehydroxylated surface, the shear-induced polymerization yield was high enough to conduct contact pressure dependence measurements to determine the critical activation volume

(V*), as shown in Figure 6-2c.167 This could not be done reliably on the hydroxylated surface because the reaction yield was too low to be measured with enough precision. Using an Arrhenius- type activated reaction model,10,34,70 the V* of this reaction was found to be about 8.30.7 Å3 on the dehydroxylated surface [see Supporting Information for calculation details]. This value corresponds to ~3% of the molecular volume of -pinene in the liquid state (liquid molar volume divided by the Avogadro’s number). In equilibrium conditions, molecules are incompressible. This then raises a question about the physical meaning of the volume change between the reactant and transition states of the molecule.

The molecular origins of the higher yield of shear-induced polymerization on the dehydroxylated surface, the production of oxygenated products from oxygen-free precursor molecules on silicon oxide, and the critical activation volume calculated from the pressure- dependence data could not be understood from the experimental studies alone; so, ReaxFF-MD simulations were performed to address these questions [see Figure 6-1]. Figures 3a and 3b show the molecular weight (MW) distribution of molecular species remaining after applying 1 GPa pressure for 2 ns without any lateral shear action. On both non-hydroxylated and hydroxylated 108 surfaces, no association products of -pinene molecules were observed. The slight increase in the molecular weights is due to the uptake of oxygen atoms from the silicon oxide substrates.

Figure 6- 3 Histograms of the molecular weights of species with ten or more carbon atoms from simulations at 1 GPa for (a) hydroxylated surface without sliding, (b) non-hydroxylated surface without sliding, (c) hydroxylated surface after 2ns of sliding, and (d) non-hydroxylated surface after 2ns of sliding. Side view images of the MD simulations are shown as insets in (a) and (b), 109 where reactive surface sites are highlighted by red spheres. Snapshots of representative shear- induced reaction products are shown as insets to (c) and (d).

Figures 3c and 3d show the MW distribution of species with 10 or more carbon atoms after shear. Here, the formation of shear-induced reaction products is evidenced by the presence of products whose MWs are heavier than dimers. The comparison with the control case without shear

(Figures 3a and 3b) clearly shows that oligomerization does not occur without interfacial shear; in other words, chemical reactions are facilitated or initiated by interfacial shear. There is significantly more oligomerization on the non-hydroxylated surface, which is congruent with the experimental result (Figure 6-2). Note that, due to the limited sliding time and number of reactant molecules in MD simulations, high MW products could not be formed.

In MD simulations, it was found that the increase in applied contact pressure from 1 GPa to 3 GPa resulted in more products at the end of sliding [see Figure 6-S6a]. When these two data points were used for the Arrhenius-type activation model analysis, V* was estimated to be ~6 Å3 on the non-hydroxylated surface [see Figure 6-S6b]. Although accuracy of this value is limited due to insufficient data points, it is still intriguing to note that the simulation value is reasonably close to the experimentally determined V* (Figure 6-2c). These observed consistencies imply that the molecular details available in the ReaxFF-MD simulations are relevant to the experimental observations, despite the differences in size and time scales between the experimental and simulation conditions.

Figure 6-4 displays snapshots from the simulation following the trajectory of one -pinene molecule. At t = 114 ps after initiation of sliding (Figure 6-4b), the surface O1 atom is approaching 110 the C6 atom of the -pinene molecule. At t = 120 ps (Figure 6-4c), the O1 atom is covalently bonded to the C6 atom; at t = 124 ps (Figure 6-4d), the second oxygen is attached to the C1 atom and the covalent bond between C1 and C6 is broken. Following the trajectories of all molecules forming association products, it was found that -pinene molecules are first oxidatively activated by reactions with dangling oxygen groups at the surface before the sliding [see Figure 6-S7] and/or hydroxyl groups during the sliding before they are associated with another molecule. This result indicates that the oxidative chemisorption or activation of the interfacial molecule is necessary for shear-induced oligomerization reactions to occur.

Figure 6- 4 Snapshots at various times from ReaxFF-MD simulations following one -pinene molecule. The numbers on atoms are given following the IUPAC nomenclature. The 4-membered ring consists of C1, C2, C7, and C6. The 6-membered ring is made of C2, C3, C4, C5, C6, and C7. 111

The C=C double bond exists between C3 and C4. The C8, C9, and C10 are methyl side groups.

The oxygen atoms are from the silicon oxide substrate and numbered following the order of covalent bond formation to -pinene molecules. Atom colors correspond to: C-brown, O-blue, and

H-yellow.

The oxidized intermediate species in Figure 6-4c meets with another -pinene molecule at t = 886 ps (Figure 6-4e). Here, it is important to note that a covalent bond is formed between the

O1 atom attached to the C6 of the first molecule and the C10 atom of the second molecule. For all

10 association products formed within the 2 ns simulation duration, the covalent bonds associating two molecules are formed between oxygen atoms of the oxidized intermediate species and the carbon atoms of the neighboring molecule [see Figure 6-S8]. This is drastically different from typical catalytic polymerization reactions of -pinene under thermal conditions, where cationic intermediates are formed and the polymerization proceeds through C-C covalent bond formation.90

This difference implies that thermal reaction mechanisms cannot be applied or utilized to explain mechanochemical reactions.

When the 10 oligomers produced in MD simulations are analyzed [Figure 6-S8], it is found that the C=C double bonds in the 6-membered ring are more likely to be attacked by the oxygen atom of the intermediate species and activated by oxidation reactions with the surface. The next most likely association reaction appears to be with the 4-membered ring or methyl groups attached to the 4-membered ring. These are the highly strained regions within the -pinene molecule.

Although the number of products is not large enough to give reliable statistics, this trend suggests that the internal strain of molecules plays a critical role in mechanochemical reactions. 112

The role of internal strain in the oxidative activation of -pinene forming intermediate species is shown in Figure 6-5. At t = 98 ps, the C1-C6 bond length (dC1-C6) is slightly elongated from 1.5 Å to 1.75 Å, which is accompanied by an increase in the C2-C7-C6 bond angle (7) and a decrease in 1. This deformation appears to facilitate the association of C6 with O1 at t = 106 ps, at which time dO1-C6 decreases from 2.3 Å to 1.9 Å (still too far to covalently bond). Compared to the total energy for a small cluster of reacting species only, the physical deformation is found to substantially reduce the activation energy for the oxygen transfer from the surface to -pinene

[see Figures S9 and S10]. As O1 approaches C6 (t = 106 ps), dC1-C6 is elongated further to 1.9 Å,

6 decreases to 84o, and 7 increases further to 110o. This corresponds to Figure 6-4b. Finally, at t = 124 ps (which corresponds to Figure 6-4d), the covalent bond between O1 and C6 is formed

(decreasing dO1-C6 to 1.5 Å). At the same time, O2 is covalently bonded to C1 (dO2-C1 decreases from ~3 Å to 1.5 Å); this is accompanied with the ring opening, which is evidenced by the increase

o of dC1-C6 from 1.9 Å to 3.4 Å and the increase of 2 and 7 to ~140 . Also, dC1-C2, dC2-C7, and dC6-C7 get slightly shorter by 0.1 Å (data not shown). During this entire process of forming the oxidized intermediate species, the effective volume change is about 10 Å3. In the subsequent association between the activated intermediate and -pinene, the shear-induced deformation of the molecule is found to facilitate the C-O-C bond formation [see Figures S11 and S12].

113

Figure 6- 5 Physical deformation of the 4-membered ring during the oxidation of -pinene by reaction with surface oxygen atoms (steps shown in Figures 4b, 4c, and 4d). Here, 1, 2, 7, and

6 represent the C6-C1-C2, C1-C2-C7, C2-C7-C6, and C7-C7-C1 bond angles, respectively. The volume of a tetrahedral box (marked in dotted lines) is also shown.

The critical activation volume (8.3 Å3) obtained from experiment was based on the derivation of Arrhenius equation. In the case of simulation, the volume change was estimated from the space occupied by the four member ring in α-pinene molecule and two oxygen atoms 114

(chemisorption sites). The encompassed volume decreases around 10 Å3 during the opening of the four membered ring. While the critical activation barrier can be considered as the shear assisted activation during chemisorption step (Chapter 6: Supporting Information), the volume change here is consistent with the ΔV* value obtained from the experimental yield rate data.

MD simulation results also explain the XPS analysis data (Figure 6-2b) which indicate that the shear-induced polymerization products are highly oxidized. The polymerization occurs not through the formation of direct C-C covalent bond between two molecules, but through the formation of the C-O-C ether bond (responsible for the 286.5 eV peak in Figure 6-2b). Also, unreacted terminal oxygen atoms will be further oxidized, forming carbonyl or carboxyl groups

(287.5 eV, and 289 eV in Figure 6-2b) upon exposure to air. In simulations, the non-hydroxylated surface produces more oxidized products than the hydroxylated surface [see Figure 6-S13], which is congruent with the XPS result showing more oxidized C1s components for the reaction products on the dehydroxylated surface (Figure 6-2b).

Conclusion

Vapor phase lubrication experiments of -pinene sheared between hydroxylated and dehydroxylated silica surfaces provided an ideal test bed for exploring the effects of surface reactivity on mechanically driven chemical reactions. Measurements of friction during sliding and characterization of the resultant reaction products revealed higher yield on the dehydroxylated surface, the formation of oxygenated products from oxygen-free precursor molecules, and an activation volume corresponding to ~3% of the molar volume of -pinene. Reactive MD 115 simulations of hydroxylated and non-hydroxylated surfaces provided atom-scale explanations for these observations. First, non-hydroxylated surfaces were shown to react more with the -pinene and have more oligomerization during shear than the hydroxylated surfaces. Also, the activation volumes calculated from the pressure-dependence of the yield in experiments and simulations for the non-hydroxylated surface were comparable to each other. This consistency with experimental observations indicated the simulations were indeed relevant to the experiments and supported the hypothesis that more reactivity of the substrate surface corresponds to higher oligmerization yield during shear. Second, the simulations showed that oligomerization between two -pinene molecules occurs via oxygen atoms obtained from surface reactions, providing direct evidence of the connection between surface reactivity and yield. Lastly, detailed analysis of individual molecules during the simulation revealed that shear facilitates association reactions through deformation of strained parts of the molecule, which in turn enables the molecule to bond with surface oxygen as the first step towards oligomerization.

116

Supporting Information

Activation of the silica substrate surface by rubbing in poorly lubricated condition in experiment

Figure 6-S 1 Activation of the hydroxylated surface by friction at a vapor pressure of -pinene insufficient for monolayer lubrication; when -pinene vapor pressure was subsequently increased to p/psat = 40% (sufficient enough for monolayer coverage), shear-induced mechanochemical polymerization occurred more readily compared to the fully hydroxylated original surface.

Water contact angle on fully hydroxylated and thermally dehydroxylated silica surfaces 117

Figure 6-S 2 Evidence of dehydroxylation of the UV/O3-cleaned hydroxylated silica surface upon thermal annealing at 450 oC for 12 hours. After thermal dehydroxylation, the water contact angle increased from <5o to 50~55o.

Creation of amorphous SiO2 model structure

The amorphous silica slabs were derived from the cristobalite SiO2 structure. This crystalline material is a high-temperature form of silica glass that is metastable below 1740 K.176 The standard method of heating and cooling was used to create the amorphous structure of the two slabs in the sliding model.177 Figure 6-S3 illustrates the initial and final structures along with the temperature and potential energy during the heating and cooling cycles. The NVT (constant number of atoms, volume and temperature) ensemble was used during heating from room temperature to 4000 K and then cooling down again to 300 K. To minimize strain in the resultant amorphous structure, the heating and cooling rate was fixed at 0.02 K/fs which was the lowest rate possible within the timescale constraints of the simulation. 118

Figure 6-S 3 (a) Cristobalite SiO2, (b) changes in temperature and potential energy during the heating and cooling cycle used to create the amorphous structure, and (c) final structure of the amorphous silica.

The physical data extracted from characterization of the amorphous silica model were compared with those reported in previous studies, as shown in Table 6-S1. All properties calculated for the amorphous structure agreed with the previously reported data within 6% error.

Table 6-S 1 Properties of the amorphous silica glass created in this work compared to the results reported in previous MD simulations and experiments.

This study Previous MD Experimental (from MD simulation) simulations178 values179,180

3 Bulk density (g/cm ) 2.1 2.1 2.2 Si-O RDF first peak (Å) 1.58 1.56 1.62 Si-O-Si (degree) 152.0 151.0 144.0 O-Si-O (degree) 110.5 110.0 109.5 119

Hydroxylation of the model SiO2 surface

With the aim of creating two systems with different surface reactivities, hydroxylation of the amorphous surface was carried out for one model. Figure 6-S4a shows the model system in which 300 water molecules were added at the top of the amorphous silica (Figure 6-S4a; box size

= 3.19×3.19×3.00 nm3). Then, the temperature was increased from 300 K to 500 K and NVT simulations were performed for 800 ps. Figure 6-S4b shows the change in the density of silanol groups formed at the surface. The density of silanol groups increased and reached a saturation value at ~300 ps, after which there was negligible change. Lastly, the excess un-reacted water molecules were removed from the system. Figure 6-S5 shows the density of surface functional groups for the non-hydroxylated and hydroxylated models.

Figure 6-S 4 (a) Model system for simulation of hydroxylation. (b) Density of silanol groups on the silica surface during the hydroxylation process.

120

Figure 6-S 5 Density of different functional groups on the non-hydroxylated and hydroxylated model silica surfaces.

Pressure dependence of the oligomerization and critical activation volume calculation

Based on the models proposed for the mechanically assisted thermal activation phenomena,10,30,34,38,108 the rate of reaction (or yield at a constant reaction time) can be expressed as an Arrhenius-type function:

∗ Ea − σ ∆V r푦 = A exp (− ) kbT Equation 6-S 1

where ry is the reaction rate or normalized yield, A is a preexponential factor, Ea is the activation energy barrier for thermal reaction, σ is the applied shear stress, ΔV* is the critical activation volume, kb is the Boltzmann constant, and T is the substrate temperature. The shear stress term can be expressed with the applied contact pressure (P) and friction coefficient (): 121

σ = σ + αP 0 Equation 6-S 2

Here, σ0 and α are constants. Inserting equation 6-S 2 in equation 6-S 1 gives:

∗ ∗ σ0 ∆V Ea ∆V . α ln (r푦) = (ln 퐴 + − ) + 푃 kbT kbT kbT Equation 6-S 3

When the flash temperature increase due to friction is negligible, the terms in the parentheses on the right side of equation (3) would remain constant.167 Therefore, on a semi-log plot of the reaction yield against the applied pressure, the slope can be used to estimate ΔV*. In Figure 6-2c of the main text plotting the normalized yield of mechanochemically produced polymers versus the applied contact pressure, the slope of the linear fit corresponds to ΔV*  8.3 Å3.

Figure 6-S 6 (a) Average molecular weight of reaction products containing more than 10 carbon atoms in MD simulations for the hydroxylated and non-hydroxylated surfaces at contact pressures of 1 GPa and 3 GPa. (b) Calculation of the critical activation volume based on the number of - pinene molecules associated in simulations at 1GPa and 3GPa conditions.

122

Figure 6-S6a displays the MD simulation results obtained for two contact pressures: 1 GPa and 3

GPa. Although the statistics in MD simulations are poor due to the limited data set, it is still possible to approximate ΔV* in simulations using these two data points. Processing the simulation results with equation (3) and the friction coefficient found in the experiment (shown in Figure 6-

2a in the main text) gave the ΔV* value of ~6 Å3 for the non-hydroxylated surface, which is reasonable when compared with the experimentally determine value (~8.3 Å3). For the hydroxylated surface, it was difficult to get the experimental value because the amount of the reaction products was too small to be measured accurately.

Surface dependence of chemisorption before sliding

In MD simulations, the reaction yield appeared to correlate with the amount of chemisorbed - pinene species. Figure 6-S7 shows the surface density of -pinene species before the onset of interfacial shear. 123

Figure 6-S 7 Density of chemisorbed -pinene molecules on the hydroxylated and non- hydroxylated surfaces at the end of the energy equilibrium step in MD simulations

Association of two molecules via C-O-C bond formation 124

Figure 6-S 8 Snapshots of the hydrocarbon species formed in MD simulations with the hydroxylated and non-hydroxylated silica surfaces after 2 ns sliding. The atom numbers at the connection between molecules are provided adjacent to each snapshot. It is important to note that all association products are connected through the C-O-C ether bond; no products were found to be connected via direct C-C covalent bonding.

125

Comparison of total energy changes for selected reaction pathways

Figure 6-S 9 Energetics of the oxidative activation (O1  C6) coordinate along which a slightly deformed -pinene molecule (red line) reacts with a dangling oxygen of the silica surface (as shown in Figures 4b and 4c in the main text). The surface oxygen atom was modeled as an isolated

(HO)3Si-O species for simplification in the energy calculation. The activation barrier is lower compared to the same reaction of an undeformed -pinene in its equilibrium structure (blue line).

126

To check the effect of deformation of a-pinene molecule on the polymerization of a-pinene, we performed a series of ReaxFF MD simulations using simple geometries.

Figure 6-S9 represents the donation of O from (HO)3Si-O to C6 of normal & strained a- pinene. The reaction of C6-O formation for normal pinene is uphill reaction with 105.9 kcal/mol, while the energy barrier of C6-O bond formation for deformed pinene is 55.58 kcal/mol. This is because the O donation creates relatively larger bond & angle strain with overcooridnated C atom for normal a-pinene. On the other hand, thanks to distorted / broken structure, bond of dangling

Oxygen was more favorable for deformed a-pinene than a normal one. 127

Figure 6-S 10 Comparison of the energetics for an oxidative activation step where a surface dangling oxygen (O1) is reacting with the one of the carbons in the double bond (C4) of -pinene.

The red and blue lines correspond to a slightly deformed and undeformed -pinene molecule, respectively.

Similarly, as shown in Figure 6-S10, the energy barrier for C4-O bond formation reaction of normal a-pinene was larger than deformed a-pinene. Energy barrier for oxygen donation to C4 site for normal pinene is 84.89 kcal/mol, while the energy barrier for deformed one is 51.93 kcal/mol. The deformed a-pinene molecule contains more atomistic angle/bond strain energy for the atoms around C4, which caused the lower energy barrier for C-O bond formation. 128

Figure 6-S 11 Energetics for the reaction between an oxygen radical at the C4 position of one - pinene and the C3 carbon of another -pinene. In the case of the deformed -pinene intermediate

(red curve), the initial state is so unstable and reactive that there is no activation barrier. In contrast, the fully relaxed intermediate (blue curve) shows an activation barrier for the same reaction.

Figure 6-S11 depicts the oxygen transport and formation of C4-O-C3 connection between two a-pinene molecules. It is shown that the existence of a dangling oxygen on C4 for deformed a-pinene molecules are so unstable, that inducing a high initial relative energy at the beginning of 129 reaction. As a result, the C4-O-C3 formation for deformed a-pinene is the downhill reaction. On the other hand, energy barrier of C4-O-C3 formation with normal a-pinene is calculated as 52.15 kcal/mol. This shows that the deformed a-pinenes are more likely to accept the C-O-C connection with other a-pinene molecules.

Figure 6-S 12 Energetics for the reaction between a hydroxyl group at the C4 position of one - pinene and the C3 carbon of another -pinene. In this scheme, the reaction proceeds first through donation of hydrogen from C3-O-H to a surface O-Si(OH)3 species followed by formation of C3-

O-C4 ether bond formation. The activation energy is slightly lower (by ~8 kcal/mol) for the deformed intermediate (red line), compared to the fully relaxed intermediate (blue line). 130

In Figure 6-S12, on top of the a-pinene molecule and hydroxylated a-pinene, O-Si(OH)3 is introduced for the hydrogen donation. Here, we calculated the energy of H hopping from C3-O-H to O-Si(OH)3, which will bring the formation of C3-O-C4 connection and Si(OH)4. Our calculation shows that the deformed a-pinene case showed the 62.84 kcal/mol, while the normal a-pinene brings 70.76 kcal/mol. Like other examples, this calculation also proved that the energy barrier for the C-O-C formation reaction is lower for the deformed a-pinenes, than that of normal a-pinenes.

Distribution of oxygen atoms in the shear-induced association reaction products

Figure 6-S 13 Regio-selectivity in oxidative activation of carbon atoms in reaction products containing more than ten carbon atoms in MD simulations at a contact pressure of 1 GPa for (a) hydroxylated and (b) non-hydroxylated silica surfaces.

131

Chapter 7

Surface chemistry dependence of mechanochemical reaction of adsorbed molecule – An experimental study on tribopolymerization of -pinene on metal, metal oxide, and carbon

surfaces

Part of this chapter is reproduced with permission from American Chemistry Society: He, X.; Kim,

S. H. Surface chemistry dependence of mechanochemical reaction of adsorbed molecule – An experimental study on tribopolymerization of α-pinene on metal, metal oxide, and carbon surfaces. Langmuir 2018, 34,

2432-2440.

Overview

Mechanochemical reactions between adsorbate molecules sheared at tribological interfaces can induce association of adsorbed molecules, forming oligomeric and polymeric products (often called tribopolymers). This study revealed the role or effect of surface chemistry of the solid substrate in mechanochemical polymerization reactions. As a model reactant, -pinene was chosen since it was known to readily form tribopolymers at the sliding interface of stainless steel under vapor phase lubrication (VPL) conditions. Eight different substrate materials were tested – palladium, nickel, copper, stainless steel, gold, silicon oxide, aluminum oxide, and diamond-like carbon (DLC). All metal substrates and DLC were initially covered with surface oxide species formed naturally in air or during the oxidative sample cleaning. It was found that the tribopolymerization yield of -pinene is much higher on the substrates that can chemisorb - pinene, compared to the ones on which only physisorption occurs. From the load-dependence of 132 the tribopolymerization yield, it was found that the surfaces capable of chemisorption give a smaller critical activation volume for the mechanochemical reaction, compared to the ones capable of physisorption only. Based on these observations and infrared spectroscopy analyses of the adsorbed molecules and the produced polymers, it was concluded the mechanochemical reaction mechanisms might be different between chemically-reactive and inert surfaces and the chemical reactivity of the substrate surface greatly influences the tribochemical polymerization reactions of adsorbed molecules.

133

Introduction

Mechanochemistry refers to reactions activated by mechanical forces.1,104,181 Compared to conventional thermochemistry, electrochemistry, and photochemistry where reactions are induced by heat, potential bias, or light irradiation, the molecular mechanisms or key parameters governing yields or selectivity of mechanochemical reactions are much less studied and understood. When a substantial amount of frictional heat is generated at high sliding speeds and not dissipated fast enough through conduction to the solid substrate, then thermal activations of interfacial molecules can be considered to play important roles.182,183 When severe wear of solid materials occurs at high applied loads or in poorly-lubricated conditions, then reactive surface sites such as dangling bonds or emission of charged particles (electrons or ions) or photons could be involved in initiation of chemical reactions of molecules present at the wearing surface.136,184,185 At relatively low load and/or slow sliding speed conditions, frictional heating and tribo-emissions are insignificant; but, mechanochemical reactions can still take place.10,23,34,71,167,186

In the case where frictional heating and tribo-emissions are negligible, one can hypothesize that perturbation of the potential energy of a molecular system under the action of external mechanical force could substantially lower the activation energy of a chemical reaction, allowing reactions to take place even at low temperature where typical thermal reactions do not occur.10,23,166,167 In this scheme, the decrease in activation barrier by a mechanical action can be modeled with a modified Arrhenius-type equation:9,10,26,34,69,166,167

−(퐸 −퐸 ) Reaction rate or yield = 퐴 ∙ 퐸푥푝 푎 푚 푘푏푇 Equation 7- 1 where A is the pre-exponential factor which will vary depending on the unit of the left-hand side of the equation, 퐸푎 is the activation barrier of thermal reaction without any external mechanical 134

shear, 퐸푚 is the amount of decrease in the activation barrier due to the mechanical shear, 푘푏 is the

Boltzmann constant ( 1.38 × 10−23 퐽/퐾 ), and 푇 is the (average) temperature of the sliding interface. When the mechanical force is applied directly to the molecule along the reaction

∗ coordinate, then the mechanical term can be expressed as 퐸푚 = 퐹 ∙ ∆푥 where 퐹 is the force applied to the specific bond and ∆푥∗ is the difference in bond length at the equilibrium and transition states.3,11,26,69 However, it would be impossible to experimentally measure how much force is applied directly along the specific bond of a molecule. Practically, it is more feasible to control the applied pressure or estimate the shear stress in experiments. Alternatively, the

∗ mechanical term can be expressed as 퐸푚 = 휎 ∙ ∆푉 where 휎 is the applied stress and the proportionality constant (∆푉∗ ) has the dimension of volume; so, it is often called a critical activation volume.8,10,34,36,38,39,166,167,187

Recently, we have studied tribopolymerization reactions of three hydrocarbons  -pinene, pinane, and n-decane.23 These three molecules consists of 10 carbon atoms with different molecular structures. -Pinene is a bicyclic terpene with 4-membered ring and one C=C double- bond in the 6-membered ring; thus, it has a very high internal strain.73,74 Pinane is the hydrogenated form of pinene; without the C=C bond in the 6-membered ring, its internal strain is reduced compared to -pinene. N-decane is a linear alkane without any internal strain in its equilibrium state. When these molecules are adsorbed on the native oxide layer (mostly consisting of Cr2O3) of AISI 440C stainless steel (SS) and sheared at an applied load of 0.4 ~ 1.5 GPa at room temperature, all three molecules (even, n-decane) produced polymeric products.23,167 From the applied pressure dependence and the measured friction coefficient, the shear stress was estimated.

By analyzing the shear stress dependence of the tribopolymerization yield, it was found that the internal strain of the molecule has a great impact on the critical activation volume  -pinene has 135

∗ the smallest ∆푉 (~3% of its molar volume, 푉푚표푙), pinane has an intermediate value (~8% of 푉푚표푙), and n-decane has the largest value (~10% of 푉푚표푙).

In equilibrium state, molecules are not compressible; then, what is the physical meaning of the magnitude of ∆푉∗? In order to study this question, a computational approach was necessary.

In another mechanochemical system studied using molecular dynamics (MD) simulations with a reactive force field called ReaxFF, it was found that tribopolymerization reactions of allyl alcohol at a sliding interface of silicon oxide appear to be associated with deformation of adsorbed molecules. This result implies that the physical process governing ∆푉∗ must be shear-induced deformation of the adsorbed molecules.166 It was also found that such deformations are facilitated when molecules are chemisorbed on the solid surface. This finding suggests that the substrate and counter-surface are not just an inert boundary or wall that confines the molecules and delivers the mechanical energy from the external mechanical actuator to the molecule being sheared at the interface; they are involved in mechanochemical reactions as an active participant.

The involvement of the solid substrate is expected when mechanochemical reactions result in removal of substrate atoms such as wear of silicon, silicon oxide, silicate glass, and gallium arsenide in humid environments or wear of carbon materials upon rubbing against silicon in vacuum.188–193 However, it is not straightforward to see the involvement of the solid substrate in association reactions among adsorbed molecules by mechanical shear.8,10,23,71,167 So, this study designed and performed an experimental investigation to test if the solid surface chemistry plays a critical role in mechanochemical polymerization of adsorbate molecules. -Pinene was chosen as a model adsorbate system since it was found to have a good mechanochemical reactivity in previous studies.23,167 Eight different substrates were chosen based on their activity or inertness in various catalytic reactions – Pd, Ni, Co, SS, Au, SiO2, Al2O3, and diamond-like carbon (DLC). All 136 experimental works were carried out in ambient pressure conditions; thus, Ni, Co, and SS are

194–196 covered with native oxide layers (NiO, CuO, and Cr2O3, respectively). Pd is also covered with a monolayer of oxide.197,198 Even the Au surface could contain oxide species when cleaned

199 200 with UV/O3. The DLC surface is also partially oxidized in ambient air. The comparison of tribopolymer yield of -pinene on these surfaces under vapor phase lubrication (VPL) conditions clearly revealed the critical role of the surface chemistry in the activation of adsorbed molecules in mechanochemical reactions.

Experiment Details

A custom-built reciprocating ball-on-flat tribometer with an environment control capability was employed to conduct all tribo-tests. Details of the system was described previously elsewhere.21 Eight substrates were tested in this study. A silicon (100) wafer was purchased from

Wafer World, Inc. (West Palm Beach, FL, USA). The silicon wafer surface was covered with 1-2 nm thick native oxide.112,130 A hydrogenated diamond-like carbon (DLC) film (approximately 1

m thick) was deposited on a silicon wafer at Argonne National Laboratory.113 Details of the DLC deposition was described in a previous paper.201 An aluminum oxide wafer (with the (0001) c- plane) was purchased from University Wafer (Boston, MA, USA). Polycrystalline gold and palladium foils (~200 m thick) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Thick copper, nickel, and AISI 440C stainless steel (SS) substrates were purchased from McMaster-Carr

(Elmhurst, IL, USA). The Cu, Ni, and SS substrates were polished with fine grit sandpapers and then polishing slurry containing 1 μm colloidal alumina. Using optical profilometry (Zygo

NewView 7300), the polished surfaces were found to have a root-mean-square (rms) roughness 137

less than 30 nm. All the substrates were cleaned with ethanol followed by a UV/O3 treatment to make sure there is no organic residue before tribo-tests.52 The counter ball was made of silicon nitride (diameter = 3/32 inch; purchased from McMaster-Carr). Optical profilometry analysis showed that the rms roughness of the ball surface was around 10 nm after removal of the sphere curvature. Silicon nitride was chosen because its surface is relatively inert in the absence of water.202 It also has high elastic modulus so that its deformation would be small during tribo-tests.

In friction tests, the normal load was adjusted to keep the Hertzian contact pressure at around 0.5

GPa for all materials. The Hertzian deformation depth varied in the range of 20 ~ 140 nm, depending on the substrate. For the load dependence test, the normal load was varied from ~0.4

GPa up to ~0.9 GPa. The Hertzian deformation depth of each sample was larger than the surface roughness of the sample; thus, two solid surfaces within the nominal contact area are indeed in intimate and conformal contact.125 The sliding speed was 0.4 cm/s and the sliding span was 2.5 mm. At this low speed, the flash temperature rise was estimated to be less than 5 ℃.23

The sliding contact region during the tribo-test was continuously purged with a mixture of

o dry nitrogen stream (with a dew point of -80 C and <100 ppm of O2) and a nitrogen stream saturated with the organic vapor. The mixing ratio was fixed to give the partial pressure of organic vapor at ~40% of its saturation vapor pressure (p/psat = 40%). Based on our previous adsorption isotherm studies, a full coverage of physisorbed molecules is normally formed at p/psat higher than

~20% for most inorganic substrates.21 Thus, it was expected to form at least a monolayer thick adsorbate layer which is sufficient to give good VPL effects to prevent the wear of substrate. Two organic vapors were tested – n-pentanol (CAS no. 71-41-0) and (+)-α-pinene (2,6,6- trimethylbicyclo[3.1.1]hept-2-ene, CAS no. 7785-70-8). N-pentanol was used as a reference gas that provides vapor phase lubrication effects without producing tribopolymers.21,167 α-Pinene was 138 chosen since it was found to be very active in mechanochemical polymerization reactions in our previous study.23,167

The adsorption isotherm of α-pinene on palladium, gold, copper, SS, and aluminum (with alumina native oxide) substrates were investigated with polarization-modulation reflection- absorption infrared spectroscopy (PM-RAIRS). A ThermoNicolet Nexus 670 spectrometer equipped with a mercury cadmium telluride (MCT) detector was applied for PM-RAIRS experiment. The experimental detail of adsorption isotherm study with PM-RAIRS was described in a previous paper. Attenuated total reflection infrared (ATR-IR) spectroscopy was used to measure the adsorption isotherm of -pinene on the silicon oxide surface. A double-side polished,

~700 m thick silicon wafer was cut into a dimension that fit to the ATR cell and both ends of this sample was polished to a 45° bevel angle for IR input and output. This ATR substrate was cleaned in the same way used for the tribo-test substrate. The experimental detail of adsorption isotherm study with ATR-IR was described in a previous paper.76

The tribo-polymers piled up at the ends and sides of the sliding track were imaged with atomic force microscopy (AFM; Digital instrument MultiMode scanning probe microscope).

Tapping mode silicon AFM tips (TESPA-V2 Bruker AFM probes) were cleaned with UV/O3 before imaging. The spring constant of cantilevers were calibrated to be ~37 N/m using the Sader method. The scan size was kept at 70 μm × 70 μm for all AFM images. AFM images were taken along the sliding track and the volumes above the reference surface (flat region outside the slide track) were added to obtain the total amount of tribo-polymers produced by sliding in -pinene vapor environment (p/psat = 40%). A selected set of tribo-polymers was analyzed with a micro-IR system (Hyperion 3000 Microscope coupled to IFS 66/s spectrometer with an MCT-a detector). 139

After AFM imaging, the tribo-polymers were rinsed off with acetone and the slide tracks were analyzed again with optical profilometry.

Results and Discussion

Figure 7- 1 Friction coefficients measured during the sliding of a silicon nitride ball (diameter =

3/32 inch) against native oxide of 440C stainless steel (black), nickel (red), copper (blue), palladium (purple), gold (orange), silicon wafer (magenta), aluminum oxide (olive), and DLC

(green) in (a) dry nitrogen, (b) 40% p/psat n-pentanol vapor and (c) 40% p/psat α-pinene vapor. The nominal Hertzian contact pressure calculated using the bulk mechanical property was ~0.5 GPa.

Friction coefficients measured with a silicon nitride sliding on different surface materials in dry nitrogen, 40% p/psat n-pentanol vapor, and 40% p/psat α-pinene vapor are plotted in Figure

7-1. In the dry inert environment (Figure 7-1a), the measured friction coefficient varies widely  as low as ~0.01 on the DLC substrate and as high as ~1.4 and noisy on the copper substrate. These 140

21 friction behaviors in dry N2 are governed by the material properties. In the n-pentanol vapor condition (Figure 7-1b), the adsorbed molecules provide boundary lubrication effects; thus, the measured friction coefficient converses to ~0.15  0.2, which is a typical value observed for boundary lubrication by self-assembled organic monolayers or friction modifiers.21,203 There was no tribochemical reaction products detected or severe wear of the substrates noticed, other than some mild plastic deformation in the case of relatively soft substrates such as gold and copper [see

Figure 7-S1 in Supporting Information].

Figure 7-1c displays the friction coefficients measured for the same material pairs in the

-pinene vapor condition. Compared to the n-pentanol VPL case, the coefficient value scatters over a larger range, varying from 0.15 to 0.3. It is not clear what determines the magnitude of friction coefficient. The coefficient value does not seem to correlate with the amounts of tribopolymer produced (see Figures 2 and 3); for example, the stainless steel and gold substrates give a similar coefficient value (Figure 7-1c), although the tribopolymer production yield is different by a factor of almost one order of magnitude (Figure 7-3). However, it is noteworthy to point out that the friction coefficient values observed in the -pinene vapor condition fall in the typical range reported for plastic materials in the literature.204–206

Figure 7-2 compares the AFM images collected at the left and right ends as well as in the middle of the slide tracks on all material types tested in this study. Based on the topographic images, the substrates tested here can be divided into two groups – a high yield group which includes palladium, copper, nickel, and stainless steel (left side in Figure 7-2) and a low yield group to which gold, silicon oxide, aluminum oxide and DLC belong (right side in Figure 7-2). In most cases, the tribopolymers are accumulated at the both ends of the slide track and a small amount can also be seen at both sides along the track. On the palladium, nickel, and copper substrates, thin 141 continuous films of tribopolymers could be seen inside the slide track. This implies that the polymers are produced in the slide track and then pushed to the end by the sliding ball. On the aluminum oxide and DLC surfaces, the tribopolymer products are so little that they form discontinuous particulates.

Figure 7- 2 Tapping-mode AFM images of tribopolymers piled up at the ends and sides of the sliding track after 600 reciprocating cycles of tribo-test with a silicon nitride ball at a Hertzian contact pressure of 0.5 GPa in α-pinene vapor (p/psat = 40%) condition. The scan area was 70 μm

× 70 μm; the height full scale is 1 m for the images in the left side and 0.2 m for the images in the right side.

In order to calculate the tribopolymerization yield, the volumes of the polymeric products discernable above the reference plane (surface outside the slide track) in AFM images were 142 calculated and then added along the slide track. In the case of copper and gold, plastic deformation cause a small extrusion of the substrate material above the reference plane along the rim of the slide track; the degree of plastic deformation is similar to the n-pentanol VPL case [see Figure 7-

S2 in Supporting Information]. The extruded volume due to the plastic flow of the substrate was negligible compared to the triboproduct volume on copper, but it was comparable to that on gold.

The total volume was normalized with the slide track area (contact diameter  slide distance) and the total slide time (reciprocating time per cycle  600 cycles). Figure 7-3 compares the normalized tribopolymerization yield at a Hertzian contact pressure of 0.5 GPa in 40% p/psat -pinene vapor.

The palladium, copper, nickel and stainless steel substrates (left column in Figure 7-2) show reaction yields about one order of magnitude higher than the gold, silicon oxide, aluminum oxide, and DLC substrates (right column in Figure 7-2).

Figure 7- 3 Total yield of the tribopolymerization products calculated by integrating the positive volume in AFM images. The total product volume (m3) was normalized with the sliding track 143 area (m2) and the total sliding time (s). Except the DLC sample, the average and standard deviation were calculated from the multiple tests for 4  5 times for each material. Due to the difficulty of finding the wearless slide track with very little tribopolymer products, only one successful measurement of the product yield was done for the DLC sample.

Since these tribopolymers were produced under the condition where frictional heating and mechanical wear of the substrate material are negligible, the reactions must be through mechano- chemical mechanisms. If the effect of mechanical action is to lower the chemical reaction barrier, then one could deconvolute the chemical activation and mechanical activation components in the

Arrhenius-type analysis using equation (1). For that purpose, the load dependence of the tribopolymerization yield was measured for four substrates (three in the high-yield group and one in the low-yield group) and shown in Figure 7-4. The linearity of the tribochemical polymerization 144

yield in the semi-log plot against the applied Hertzian contact pressure indicates that the reactions

are indeed activated by mechanical actions.

Figure 7- 4 Semi-log plot of the contact pressure dependence of the tribopolymerization yield for

α-pinene on copper (black), nickel (red), stainless steel (green), and silicon oxide (blue) substrates.

The partial pressure of α-pinene vapor was 40% p/psat and the total reciprocating cycle of the silicon nitride ball was 600. All data points are the average from 3 repeats (some were 4 repeats); for some data points, the error bars are smaller than the symbol size.

The shear stress σ term in 퐸푚 of equation 7-1 can be expressed as a linear function of

normal contact pressure 푃 (GPa):121,123

휎 = σ0 + 훼푃 Equation 7- 2

where σ0 is the term determined by the adhesive interaction, which can be neglected in the

macroscopic tribo-testing.207 The proportionality constant 훼 can be replaced with the friction

∗ coefficient measured in the experiment. Replacing the 퐸푚 term in equation 7-1 with 휎 ∙ ∆푉 and

inserting 7-2 here, one can express the Arrhenius-type equation in the following form:

∗ ∗ 퐸푎 σ0∆푉 ∆푉 훼 ln(yield) = (푙푛퐴 − + ) + 푃 푘푏푇 푘푏푇 푘푏푇 Equation 7- 3

From linear regression of the data in Figure 7-4, the intercept and slope of equation (3) can be

obtained. Then, the slope gives the critical activation volume, ∆푉∗. Since 퐴 is constant, 푇 does not

∗ change substantially, σ0 is negligible, and ∆푉 varies only in a small range, the change in the 145

intercept must be governed mostly by the difference in the chemical activation energy, 퐸푎 , although its absolute value cannot be obtained without determining 퐴 and σ0.

The activation volume and the intercept value calculated from the data in Figure 7-4 are listed in Table 7-1. It is noted that ∆푉∗ is smaller for the substrates in the high-yield group (copper, stainless steel, and nickel) compared to the silicon oxide surface which is in the low-yield group.

Based on the previous finding about the physical meaning of ∆푉∗ from the ReaxFF MD study,166 it can be interpreted that the degree of shear-induced physical deformation needed for -pinene to undergo tribopolymerization is smaller on the copper, stainless steel, and nickel substrates compared the silicon oxide surface. Then, an important question is why and how the polymerization of the same reactant (-pinene) has a different ∆푉∗ depending on the substrate.

Table 7- 1 The critical activation volume (∆푉∗) calculated from Figure 7-4 and the intercept of each semi-log scale load dependence plot. The error is standard error from linear regression. The

3 molar volume of α-pinene 푉푚표푙 is 263.8 Å .

Substrate Critical activation volume Intercept (actual surface) ퟑ Å % of 푽풎풐풍 Copper (CuO) 5.8 ± 1.0 2.2 ± 0.4 -23.17 ± 0.04

Stainless steel (Cr2O3) 8.7 ± 0.9 3.3 ± 0.3 -23.34 ± 0.03

Nickel (NiO) 9.3 ± 1.7 3.5 ± 0.6 -23.43 ± 0.07

Silicon (SiO2) 12.4 ± 2.1 4.7 ± 0.8 -25.41 ± 0.06

146

The answer might be found by noticing that the copper, stainless steel, and nickel substrates show smaller absolute values in the intercept than the silicon oxide substrate, indicating that the chemical activation energy itself (퐸푎) is smaller on the high-yield substrates than the low-yield substrate. The previous ReaxFF MD simulation study hinted that the mechanical energy transfer from the substrate to the interfacial molecule is facilitated when the molecule is chemisorbed on

166 the solid surface. Based on this hypothesis, the lower 퐸푎 value (smaller absolute value of the intercept) could mean that -pinene molecule are chemisorbed on the surface of the high-yield substrate materials. This can be tested by analyzing the surface species formed and remaining on the solid surface after exposure to -pinene vapor.

147

Figure 7- 5 Infrared spectra of α-pinene adsorbed on (a) palladium, (b) copper, (c) gold and (d) silicon substrates at p/psat = 30%, 60% and 90%. The chemical formula in bracket is the surface composition. Also shown is the spectra taken after purging the -pinene vapor with N2 after the

90% p/psat measurement. The spectra in (a-c) were taking with PM-RAIRS and the spectra in (d) were obtained with ATR-IR.

Figure 7-5 compares the adsorbed species on palladium and copper of the high-yield group and gold and silicon oxide of the low-yield group. Due to the low reflectivity of IR, PM-RAIRS data of stainless steel (with Cr2O3 at the surface), nickel (with NiO layer), and aluminum (with

Al2O3 layer) were noisy. The DLC film on the silicon wafer was too thick to do ATR-IR analysis. 148

In all spectra shown in Figure 7-5, the peak at ~3060 cm-1 is clearly detected, implying that the

C=C double bond in the 6-membered ring is mostly intact in the adsorbed -pinene molecule. The overall peak shapes of the detected adsorbate species are also consistent with the spectral feature of the -pinene vapor (shown in Figure 7-6). Although the thickness of the adsorbed layer could not be calculated due to technical limitation of PM-RAIRS, the spectral intensity of the adsorbed species measured while increasing the -pinene vapor pressure generally follows the type-II adsorption isotherm behavior [see Figure 7-S3 in Supporting Information].

When -pinene vapor was purged after measuring the adsorbate spectrum at p/psat = 90%, residual organic species were detected on the palladium and copper substrates. The peak area of the residual species of palladium is almost the same as the total intensity of the adsorbed species at p/past = 60% of -pinene. In the case of copper, its peak area is about the same as the adsorbed species at p/psat = 30%. In contrast, very little amount of residue is seen on the gold surface. In the case of silicon oxide, the residual species is below the detection limit. These results clearly imply that -pinene is chemisorbed on the palladium and copper substrates (likely on other high-yield substrates), while it is only physisorbed on the gold and silicon oxide surface (and aluminum oxide surface as well). Here, it is interesting to note that palladium and copper oxide are known to have catalytic effects in various chemical reactions of organic species,208–210 while gold is inert in general (except some nanoparticle forms) and silicon oxide is usually used as an inert support in many catalytic reactions.211–213

In Figure 7-5a and 5b, the residual species detected on the palladium and copper substrates display different IR features. This means that the chemisorbed species on these two substrates have different chemical structures. If that is the case, it is expected that the tribopolymers produced from these intermediates would have different chemical structures. Figure 7-6 displays the IR spectra of 149 the tribopolymer products piled up at the end of the slide track on four substrates in the high-yield group. Although -pinene has no oxygen in it and the reaction was carried out in nitrogen environment, the IR spectra indicates that the products contain hydroxyl (OH) and carbonyl (C=O) groups. Previously, the time-of-flight secondary ion mass spectrometry analysis of the -pinene tribopolymer produced on stainless steel , also found that the tribopolymer product contains oxygenated species.23

Figure 7- 6 IR spectra of tribopolymers piled up at the ends of the track on selected substrates after 600 reciprocating cycles of tribo-test with a silicon nitride ball in α-pinene vapor (p/psat =

40%) condition. Also shown is the IR spectrum of -pinene vapor measured during the PM-

RAIRS experiments of the data shown in Figure 7-5.

150

It cannot be determined how much of these oxygenated species originate from the oxidative chemisorption on reactive oxide surfaces or due to post-synthesis oxidation of the product upon exposure to air.23 If the tribopolymerization has no surface dependence, the produced polymers would have the same (or similar) molecular structures at the same sliding condition; then, the post- synthesis oxidation would be the same for all tribopolymers of -pinene regardless of the surface chemistry of the substrate. But, the data in Figure 7-6 showed that the relative areas of OH and

C=O peaks compared to the alkyl peaks are different depending on the substrate. Also, the deconvoluted components of the CH2 and CH3 stretch peaks of the -pinene tribopolymer vary with the substrate. These results clearly support that chemisorption (probably oxidative chemisorption) on reactive sites of the high-yield substrates plays critical roles in the tribopolymerization yield and the structure of the produced polymer.

Figure 7- 7 Schematic energy diagram along the reaction coordinate when (a) interfacial molecules are chemisorbed and then mechanically activated for reaction, (b) interfacial molecules undergo mechanically-assisted chemisorption on reactive surfaces, and (c) solid surfaces are not 151

ads reactive and physisorbed molecules are activated by mechanical shear. Echem is the activation

therm therm energy for chemisorption. Echem and Ephys are the activation energy of chemisorbed (or during chemisorption) and physisorbed species, respectively, in the absence of mechanical effects. The

mech mech reaction pathways or mechanisms might be different in three cases. Echem and Ephys are the

∗ amount of decrease in the activation energy (associated with Em = σ ∙ ∆V ) of chemisorbed and

eff physisorbed species, respectively, upon the application of mechanical shear. The Ea term is the net activation energy governing the reaction yield or rate of mechanochemical reactions. Note that it is still larger for the physisorption case in (c) compared to the chemisorption case in (a) and (b).

Based on the surface chemistry dependence of the critical activation volume and the intercept (pre-exponential term) of the modified Arrhenius relationship (Table 7-1), the degree of

-pinene chemisorption (Figure 7-5), and differences in IR spectra of tribopolymers (Figure 7-6), the tribopolymerization of -pinene is believed to vary depending on the surface chemistry of the solid substrate. Figure 7-7 schematically illustrate energy diagrams for mechanochemical reactions on highly-reactive and relatively-inert surfaces. It appears that the chemically-reactive surface can activate the molecules through chemisorption and the chemisorbed molecules can more readily undergo mechanochemical reactions with a less degree of physical deformation; in other words, they are more susceptible to mechanical activations (Figure 7-7a). It is also possible that the mechanochemical effect may occur during the chemisorption step (Figure 7-7b). It is difficult to distinguish which step the mechanochemical effect is larger – the first step where a molecule is chemisorbed and activated to the intermediate step or the second step where the association reaction between the intermediate and other molecules occurs. In any case, the effective activation 152 energy for the tribopolymerization of the -pinene model system is lower for the substrate materials that can chemisorb -pinene molecules (left column in Figure 7-2), compared to the ones that simply physisorb -pinene (right column in Figure 7-2; Figure 7-7c). In addition to the chemisorption, the local strain and electronic structure perturbation of the substrate during the elastic deformation inside the contact area might play a role in chemical reactivity of the surface.214,215 However, the contribution from the electronic state change upon the substrate strain cannot be determined independently in the current study.

Conclusion

Mechanochemical reactions of -pinene forming tribopolymers were studied on eight solid substrates  palladium, nickel, copper, stainless steel, gold, silicon oxide, aluminum oxide, and diamond-like carbon (DLC). The substrates covered with transition metal oxides, which are often used as catalysts in various reactions of hydrocarbons, appear to be much more reactive for tribopolymerization reactions than the substrates that are known to be relatively inert or used as support materials in catalytic reactions. The more reactive surfaces are found to be capable of chemisorbing -pinene molecules even before mechanical shear is applied. On the surfaces capable of chemisorption, the critical activation volume change for the mechanical reaction to occur is much lower compared to the reactions occurring on the surfaces capable of physisorption only. Since the chemisorption path would vary depending on the substrate, the molecular structure of tribopolymers of -pinene seems also different. 153

Supporting Information

Surface topography after tribo-test in 40% p/psat n-pentanol vapor condition

Figure 7-S 1 Cross-section line profile (top), optical profilometry image (middle), and

optical microscope image (bottom) of sliding tracks after friction tests under 0.5 GPa in

40% p/psat n-pentanol.

154

Surface topography after tribo-test in 40% p/psat α-pinene vapor condition

Figure 7-S 2 Cross-section line profile (top), optical profilometry image (middle), and optical microscope image (bottom) of sliding tracks after friction tests under 0.5 GPa in 40% p/psat α- pinene. The substrate were cleaned with ethanol after the friction test to remove the residue tribo- polymers to evaluate substrate deformation.

155

Adsorption isotherm of α-pinene on various metal or metal oxide

Figure 7-S 3 Adsorption isotherm of α-pinene on metal or metal oxide from PM-RAIRS measurements and on silicon oxide from ATR-IR measurement

156

Chapter 8

Effect of gas environment on mechanochemical reaction: A model study with tribo-

polymerization of -pinene in inert, oxidative, and reductive gases

Part of this chapter is reproduced with permission from Springer: He, X.; Pollock, A; Kim, S. H.

Effect of gas environment on mechanochemical reaction: A model study with tribo-polymerization of α- pinene in inert, oxidative, and reductive gases. Tribol. Lett. 2019, 67, 25.

Overview

Mechanically-induced or assisted chemical reactions readily occur in many tribological interfaces; however, the reactants involved in such reactions are often not well understood. Using the tribo- polymerization of -pinene as a model system, this study investigated how the surround gas environment influences the surface chemistry of substrate controlling the tribochemical reactivity.

Based on the hypothesis that oxidative chemisorption of -pinene at the sliding solid surface plays critical role, inert (dry N2), oxidative (dry air), and reductive (10% H2 in Ar) gas environments were chosen to alter the degree of surface oxidation in the sliding contact. Comparing the tribo- polymerization yield of -pinene on two highly-reactive substrates (Pd and CuO) and two relatively-inert substrates (diamond-like carbon and silicon oxide), it was found that the oxidative gas significantly enhances the tribochemical reactivity of -pinene. Infrared spectroscopy analysis was employed to confirm that the chemisorption of -pinene on the surface in the oxidative gas environment plays a critical role. The gas environment was also found to affect the chemical composition and elastic modulus of the tribo-polymer products. 157

Introduction

Mechanochemistry is an important category of chemical transformation processes ubiquitous in engineering systems as well as daily life.3 It refers to chemical reactions initiated or facilitated by mechanical impact, compression, tension, or shear.1,104,181 An example is mechanochemical synthesis of organic and inorganic compounds in ball milling processes.104,133–

135 Chemical reactions occurring at sliding or shearing interfaces is specifically called tribochemistry.4,5,132 Compared with the well-studied thermal, photochemical, and electrochemical reactions, mechanistic understanding of tribochemical reactions is not well established. Key challenges in mechanistic study of tribochemistry is to identify the reactants and products involved in the contact region and their stoichiometric relationships as well as to understand how mechanical force or energy is channeled into the reaction coordinate.

Tribochemical reactions could be categorized into three types. The first is thermal reaction due to frictional heat. When the flash temperature at the contact increases substantially at high sliding speeds, thermal reactions can be initiated.183,216 The second case is the involvement of highly reactive surface sites or energetic particles. When a solid surface wears or mechanically damaged, then dangling bonds will be exposed at the surface and initiate chemisorption of molecules impinging from the lubricant phase.25 Also, bond dissociation during wear or fracture can be accompanied by emission of high energy photons or electrons, which can induce chemical reactions in the lubricant phase.136,184,217 The third case is chemical reactions occurring at tribological interfaces where frictional heat is negligible and wear is completely suppressed.9,23,34,218 In this case, chemical reactions are believed to induced or facilitated by mechanical shear of interfacial molecules. 158

In the third case, chemical reactions are often explained with mechanically-assisted

Arrhenius-type kinetics arguments.26,34,153 In many cases, the tribochemical reaction rate is observed to increase with an exponential function of applied stress.8,11,38,39,69,160,187,218 Because the product of stress and volume has the dimension of energy, this term can be incorporated into the

Arrhenius equation as a term counter-acting or reducing the activation energy (barrier) of reaction.166,219 Then, the slope of tribochemical reaction rate data in the semi-log plot against the applied stress can be converted to a critical activation volume (ΔV*).8,34,38,39,160,187,218 Although

ΔV* is readily calculated from this kind of stress-dependent reaction rate measurements, its physical meaning is not fully understood yet.

In an attempt to elucidate the stress dependence of tribochemical reactions, molecular dynamics (MD) simulation studies with ReaxFF reactive force fields were recently carried out.166,219 Computer simulation studies suggested that ΔV* may not be the actual change in molecular volume; instead, it would be just a physical constant related to how easily the interfacial molecule being sheared mechanically can deform from its equilibrium conformation.153,219

These simulation studies also suggested that such deformation would be highly facilitated if reacting molecules are chemically anchored to one of the sliding surfaces.166 In other words, chemisorbed molecules would be much more susceptible to shear-induced deformation than physisorbed molecules. The deformation of molecular structures such as chemical bond angles or lengths inevitably accompany destabilization of molecules, which will facilitate initiation of chemical reactions. It is also possible that the energetics of transition state is also altered. This has been experimentally confirmed by comparing tribochemical reaction yields of -pinene on chemically-reactive versus relatively inert substrates and allyl alcohol on partially- and fully- hydroxylated silica surfaces.166,219,220 159

In this paper, we report the effect of gas environment (carrier gas) in tribochemical polymerization of -pinene (C10H16) on four different substrates. One of the lingering questions in the previous experimental studies was why tribochemical polymerization products of -pinene contain oxygenated species although the reactant supplied to the surface does not have any oxygen.23 The hypothesis explaining this observation was oxidative chemisorption of -pinene on the substrate surface that already contains oxygen.219 In order to test this hypothesis, we carried out the tribochemical polymerization study of -pinene on highly-reactive (Pd and CuO) and relatively-inert (diamond-like carbon, DLC, and SiO2) substrates in inert (dry N2), oxidative (dry air), and reductive (H2 in Ar) gas environments. If the hypothesis is right, the reactions in oxidative gas would produce the largest amount of tribochemical products and those in reductive gas would produce the least amount. This hypothesis was also supported by analyzing the amount of chemisorbed species on the surface at different environments as well as characterizing the tribochemical reaction products.

Experiment Details

Tribochemical reactions were carried out with a custom-built ball-on-flat reciprocating tribometer equipped with a vapor-phase environment control capability.21 The counter-surface was silicon nitride balls with a 3/32 inch diameter. The root-mean-square (rms) roughness of the ball surface was less than 10 nm after removal of the sphere curvature. Silicon nitride balls were chosen for this study because they showed negligible wear during the sliding in our experimental condition.221 The reciprocating slide amplitude was 2.5 mm. The sliding speed was 0.4 cm/s; at 160 this speed, the flash temperature rise of the surface was estimated to be less than 5 ̊C at such a low sliding speed.23,124

Four substrate materials chosen for this study were native oxide on Si, highly-hydrogenated

DLC, palladium, and copper. A Si(100) wafer was purchased from Wafer World, Inc. (West Palm

Beach, FL, USA). The as-received wafer was covered with ~1 nm thick native oxide layer.7 The hydrogenated DLC film (~1 μm thick) was deposited on a silicon wafer using plasma-enhanced chemical vapor deposition (PECVD). Details of the deposition process were described in previous publications.201,222,223 Both Si wafer and DLC film had very smooth surfaces with a rms roughness of ~1 nm. A polycrystalline Pd foil (~200 m thick; Sigma-Aldrich, St. Louis, MO, USA) and a

Cu plate (1mm thick; McMaster-Carr, Elmhurst, IL, USA) were polished with a fine grit sandpaper and then micropolished with 1 μm colloidal alumina particle solution (Buehler) to get a rms roughness of ~20 nm. This rms roughness is smaller than the elastic deformation depth calculated from the Hertzian contact model (dcopper = 79 nm and dpalladium = 75 nm at a contact pressure 0.5

GPa); this assured the intimate contact between the ball and the substrate.125 Note that the Cu substrate surface is easily oxidized in air and covered with a thin layer of CuO.194 All substrate were cleaned with 200 proof ethanol and followed by 30 min UV/ozone treatment to remove the chemical residues.52

The environment control was achieved by continuously flowing the carrier gas containing

α-pinene (Sigma-Aldrich; 2,6,6-Trimethylbicyclo[3.1.1]hept-2-ene; CAS no. 7785-70-8) over the tribo-testing interface. The partial pressure of -pinene was 40% relative to the saturation (p/psat).

This vapor pressure condition was chosen because it is known from the previous adsorption isotherm study that a full monolayer coverage is attained at p/psat 15% and multilayers start

21 growing at p/psat  about 75%. Three carrier gases used in this study were dry N2 (with a dew 161

o point lower than -70 C), dry air (~20% O2 + 80% N2), and 10% H2 in Ar. Since both nitrogen and argon are chemically inert (do not react with the substrate and -pinene in our experimental conditions), they do not interfere with tribochemical reactions. The carrier gas was bubbled through a pure α-pinene liquid reservoir to produce the gas stream saturated with -pinene vapor.

This saturated gas stream was mixed with the vapor-free gas stream to obtain the desired partial pressure of -pinene in the gas delivered to the tribochemical test chamber.

The adsorption and desorption of α-pinene on highly-reactive substrates (Pd and CuO) were analyzed with polarization-modulation reflection-adsorption infra-red spectroscopy (PM-RAIRS) using a Thermo-Nicolet Nexus 670 spectrometer equipped with a mercury cadmium telluride

(MCT) detector. The experiment details and theory of PM-RAIRS were described in a previous paper.23 The tribo-polymers produced during the sliding and piled up at the ends of the sliding track were analyzed with micro-FTIR spectroscopy. A Bruker Hyperion 3000 IR microscope coupled to IFS 66/s spectrometer was employed for micro-FTIR experiments. The scan area was

50 µm × 50 µm containing a decent amount of tribo-polymers.

The volume of tribo-polymers accumulated around the sliding-track was measured with atomic force microscopy (AFM) using a Digital Instrument MultiMode scanning probe system. A tapping mode AFM scanning with a silicon AFM probe (TESPA-V2 Bruker) was used to collect the three-dimensional topographic images of tribo-polymers at different segments of the sliding track. All AFM probes were cleaned with UV/ozone to remove chemical contaminants before the experiments. The volume of the tribo-polymer of each segment (70 m  70 m) was calculated using the substrate surface outside the slide tract as a reference and added up to obtain the total amount of tribo-polymer in each test. The viscoelastic property of the α-pinene tribo-polymers produced in N2 and O2 containing carrier gases was construed from force-distance indentation 162 measurements in ambient air. The spring constant of AFM probes (PPP-FM, NANOSENSORS) used for indentation tests was calibrated with the Sader method to be ~3 N/m.82,118 The maximum indentation force was set at 60 nN; the indentation depth was less than 15 nm (< 10% of tribo-film thickness), so that the substrate effect was negligible.96

Results and Discussion

Environment dependence of tribo-polymerization of -pinene

Figure 8- 1 The friction coefficient between silicon nitride ball and native oxide of (a) diamond- like carbon (DLC), (b) silicon wafer, (c) palladium and (d) copper in 40% p/psat α-pinene vapor in inert (dry nitrogen; red color), oxidative (dry air; green color), and reductive (hydrogen in Ar; blue color) carrier gases. In dry case, the friction coefficient would be high and unstable (0.79±0.15).

163

The friction coefficients measured with the silicon nitride ball sliding on four substrate materials in inert (dry N2), oxidative (dry air), and reductive (10% H2 in Ar) gases containing 40% p/psat α-pinene vapor are shown in Figure 8-1. From the vapor phase lubrication (VPL) study of non-reactive organic molecules, it is known that the friction coefficient of sliding interfaces covered a monolayer of physisorbed molecules is typically 0.15 – 0.2.21 On the relatively-inert substrates (DLC and SiO2), the friction coefficients measured in the inert and reductive gases are within this typical value range; but, the coefficient in the oxidative gas is higher (around 0.25 –

0.3). In the previous study, it was speculated that the higher friction coefficient in tribo- polymerization conditions might correlate with the presence of polymeric products in the sliding track.220 On the highly-reactive substrates (Pd and CuO), the friction coefficients in the inert and reductive gases are still higher than the typical value range of the physisorbed molecules and the coefficient in the oxidative gas is even slightly higher than the values observed in the inert and reductive gases. If the correlation between the friction coefficient and tribo-polymerization activity speculated in the previous study holds here,220 then these results suggest that the tribopolymerization yield of -pinene would be higher in the oxidative gas environment than the inert and reductive gas environments.

The tribo-polymerization yield of -pinene was estimated by measuring the volume of the reaction products piled up around the sliding track with AFM. The AFM images of tribo-polymers at three locations (left and right ends as well as in the middle of the slide track) are displayed in

Figure 8-2. Note that the topographic features in the AFM images are not wear particulates. When the slide tracks are rinsed with alcohol, then they are dissolved and rinsed off and there is no wear damage on the substrate. In the case of Pd and CuO substrates, only slight plastic deformation could be seen (see Figure 8-S1 in the Supporting Information). The data in Figure 8-2 clearly show 164 that the amount of -pinene tribo-polymers is significantly larger for the oxidative gas case. On the relatively-inert substrates (DLC and SiO2), only minute amounts of tribo-polymers can be seen in the inert gas; in the reductive gas, the tribo-polymers are negligible. On the highly-reactive substrates (Pd and CuO), the amount of tribo-polymers seen in AFM images is significantly reduced in the reductive gas, compared to the oxidative and inert gas environments.

Figure 8- 2 Tapping-mode AFM images of -pinene tribo-polymers accumulated at both

ends and in the middle of the slide track after 400 cycles of reciprocating sliding with the

silicon nitride ball at a Hertzian contact pressure of 0.5 GPa in oxidative (top), inert

(middle) and reductive (bottom) gases containing 40% p/psat -pinene vapor. The image

size is 70 μm × 70 μm and the height full scale is 2 m. The cross-section line profile (white

line) of the right end is plotted in the right side of images.

165

The total volume of -pinene tribo-polymers was calculated by adding the volumes of topographic features above the substrate reference plane from a series of AFM images along the slide track. Dividing the total volume with the total slide area and the sliding time gives the tribo- polymerization yield. Figure 8-3 compares the tribo-polymerization yield of -pinene at a Hertzian contact pressure of 0.5 GPa in oxidative (green), inert (red), and reductive (blue) gas environments.

It clearly shows that the reductive gas environment significantly suppresses the tribo- polymerization yield on the highly-reactive substrates (Pd and CuO), compared to the oxidative and inert gas environments. On the relatively-inert substrates (DLC and SiO2), only the oxidative gas environment can assist tribo-polymerization of -pinene. The tribo-polymerization yield is reduced by a factor of ~10 in the inert gas and the tribo-polymerization does not occur readily in the reductive gas.

166

Figure 8- 3 Tribo-polymerization yield of α-pinene on highly-reactive (Pd and CuO) and relatively-inert (DLC and SiO2) in oxidative (green), inert (red), and reductive (blue) gas environments. The normalized (μm/s) was calculated by dividing the total volume of tribo-product

(μm3) with the sliding track area (μm2) and total sliding time (s). The average and standard error were obtained from measurements in the same reaction condition for 3-5 times.

The higher tribo-polymerization yield of -pinene in the oxidative gas and the suppression

in the reductive gas are consistent with the hypothesis that the oxidative chemisorption of -pinene

is the prerequisite for effective activation by mechanical shear at the sliding interface.166,219 To

further validate this hypothesis, the evidence of oxidative chemisorption of -pinene was sought

for the highly-reactive substrates (Pd and CuO) using PM-RAIRS. Figure 8-4 exhibits the PM-

RAIRS spectra of -pinene adsorbed on the metal substrate surface while the p/psat of -pinene

vapor is increased stepwise to 15%, 30%, 50%, and 75% and then after -pinene vapor is

completely purged out. It is difficult to distinguish if the adsorbed molecules are physisorbed or

chemisorbed from the spectral features during the adsorption process. This is because the

chemisorbed species still retain most of C-H and C-C bonds intact, except the one involved in the

chemisorption process. Because the physisorbed species are in equilibrium with the vapor

molecules in the gas phase, they will desorb from the surface upon purging the -pinene vapor

(decreasing p/psat to 0%). The PM-RAIRS spectra of Pd and CuO surfaces after purging the -

pinene vapor clearly show a significant amount of chemisorbed species remaining on the surface

in the oxidative gas. A similar observation was made in the I nert gas environment.220 However,

the spectral intensity of residual species is much lower in the reductive gas environment.

167

Figure 8- 4 PM-RAIRS spectra of α-pinene adsorbed on (a,b) Pd and (c,d) CuO (native oxide on

Cu) surfaces at p/psat = 15% (black), 30% (red), 50% (blue), and 75% (pink), and then after purging out -pinene (green) in (a,c) oxidative and (b,d) reductive gas environments. The insets plot the total peak area.

168

These results indicate that the chemisorption of -pinene on the highly-reactive Pd and CuO surfaces is influenced by the surrounding gas. In the presence of hydrogen, the chemisorption is significantly suppressed. Then, based on the previous MD simulation results,219 the shear-induced activation of interfacial molecules is expected to be suppressed or lower in the reductive gas environment. This is consistent with the trio-polymerization yield data shown in Figures 2 and 3.

The same experiment could not be done for the relatively-inert DLC and SiO2 surfaces.

However, it is already known that the DLC surface is readily oxidized in air.224 So, the oxygenated functional groups at the DLC surface are expected to be involved in chemisorption of -pinene.

In the oxidative gas (dry air), the oxygenated group at the DLC surface will be replenished as they are consumed in tribo-polymerization reactions of -pinene, sustaining the tribochemical reactivity. In the inert gas (dry N2), a trace amount of O2 and H2O impurity can still oxidize the

DLC surface;225 thus, a small amount of tribochemical activity is still possible. In contrast, the presence of H2 in the gas phase, the oxidation of DLC surface is almost completely suppressed (as

226 evidenced by the super-lubricity of DLC in H2 gas). This explains the complete suppression of tribo-polymerization activity of -pinene on DLC in the presence of H2 in the gas phase. A similar argument could be used to explain the environment gas dependence of tribochemical activity of

-pinene on SiO2, although we do not have independent spectroscopic or experimental data for the SiO2 surface.

169

Properties of tribo-polymers of -pinene produced in different gas environments

Not only the tribo-polymerization yield of adsorbed molecules, but the chemical

composition and physical properties of tribo-polymers can also vary depending on the gas

environment. The exact identity or stoichiometry of functional groups in the tribo-polymer of -

pinene could not be determined; but, qualitative information could be obtained from IR analysis.

Figure 8-5 presents the micro-IR spectra of -pinene tribo-polymers produced on the highly-

reactive Pd and CuO surfaces in oxidative, inert, and reductive gas environments. The same

experiment could not be reliably performed for the relatively-inert DLC and SiO2 surfaces because

the amount of tribo-polymers was insufficient to get the spectra with a good signal-to-noise ratio.

The IR spectra in Figure 8-5 shows that the signal intensities of the OH stretch mode (3050 – 3600

cm-1) and the C=O stretch mode (1650 – 1800 cm-1) are stronger for the tribo-polymers produced

in the oxidative environment than the ones produced in the reductive environment. This implies

that the oxygen in the surrounding gas are incorporated into the molecular structure of tribo-

polymers; it must be via dissociated adsorption of oxygen at the substrate surface which produces

surface sites for oxidative chemisorption of -pinene. Without knowing the exact reaction

stoichiometry and the molecular structure of reaction products, further discuss of reaction

mechanisms is not possible in this study. 170

Figure 8- 5. IR spectra of tribo-polymers of -pinene on (a) Pd and (b) CuO substrates produced in oxidative (green), inert (red), and reductive (blue) gas environments. The micro-IR analysis was carried out for tribo-polymers accumulated at the end of the slide track after 400 cycles of reciprocating sliding with silicon nitride balls in 40% p/psat of -pinene vapor. The spectral intensity is normalized with the C-H stretch band for comparison purpose.

171

Figure 8- 6 Force-distance (F-D) curve of α-pinene tribopolymers accumulated at the end of

the sliding track on (a,b) Pd and (c,d) CuO in (a,c) oxidative and (b,d) inert gases. The elastic

modulus estimated from the JKR fit of the retraction curve is shown in each panel.

Interestingly, the difference in relative abundance of hydroxyl and carbonyl species in - pinene tribopolymers (Figure 8-5) affects their mechanical properties. Figure 8-6 displays the 172 force-distance curve (F-D curve) measured with AFM on the -pinene tribo-polymers produced on Pd and CuO in oxidative and inert gas environments. The retraction part of the F-D curve was fitted with the Johnson–Kendall–Roberts (JKR) model to estimate the modulus of the tribo- polymer (Figure 8-S2).56,227 It is found that the α-pinene tribopolymer produced in the oxidative gas has a lower modulus (less than 0.5 GPa), while that produced in the inert gas has a higher modulus (2.31 GPa for the polymer on Pd and 1.65 GPa for the polymer on CuO).

Conclusion

The effect of ambient gas on the tribochemical polymerization of -pinene on highly- reactive and relatively-inert substrate surfaces was studied. The surrounding gas can affect the oxidation state of the substrate surface, which in turn determines the chemisorption capacity of the molecules being polymerized upon shear. The chemisorption of interfacial molecules is critical to effectively transfer the mechanical shear force or energy to the reaction coordinate. In oxidative environment, the solid surface can facilitate oxidative chemisorption of -pinene molecules; in contrast, such chemisorption is suppressed in reductive environment and so does tribochemical reaction yield. The incorporation of oxygen species from the ambient gas to tribo-products via surface chemisorption also alters mechanical properties of the tribo-product.

173

Supporting Information

Surface topography after tribo-test in 40% p/psat α-pinene vapor in oxygen, nitrogen and hydrogen gas condition

Figure 8-S 1 optical profilometry image (left) and cross-section profile (right) of sliding tracks after friction tests at the 0.5 GPa Hertzian contact pressures in 40% p/psat α-pinene in dry air, nitrogen and hydrogen (10% in Ar). The substrates were cleaned with ethanol after the friction test to evaluate substrate damage without interference from triboreaction products. 174

Elastic calculation through Johnson–Kendall–Roberts (JKR) fitting of unloading F-D curves

Figure 8-S 2 JKR model fit of the unloading curves of (a,c) α-pinene tribo-polymer created in dry air carrier gas on Pd and CuO, respectively; (b,d) α-pinene tribo-polymer created in dry nitrogen carrier gas on Pd and CuO, respectively.

175

Chapter 9

Mechanochemical reactions of adsorbates at tribological interfaces:

Tribo-polymerizations of allyl alcohol co-adsorbed with water on silicon oxide

Xin He, Dien Ngo, and Seong H. Kim*

Department of Chemical Engineering and Materials Research Institute, Pennsylvania State

University, University park, PA 16802, USA.

Overview

Mechanochemical reactions occurring at tribological interfaces can be significantly affected by co-adsorption and mechanochemical reactions were studied for water and allyl alcohol binary system at silicon oxide tribological interfaces. As a model reactant, allyl alcohol could be polymerized under shear stress during vapor phase lubrication. Friction measurements with different allyl alcohol- water ratios, ranging from 0 to 1, showed that the incorporation of water molecules could improve the tribopolymer yield rate of allyl alcohol. The investigation of co- adsorption via infra-red spectroscopy suggested that the presence of water may facilitate the oxidative chemisorption process of allyl alcohol molecules, particularly in water-rich regime. It was found that the wear behavior on silicon substrate showed dependence based on the proportion of water in vapor phase. In water-rich phase, the substrate is well-lubricated, while turns to alcohol- rich phase, the wear can be clearly observed. The chemical structure analysis revealed that the water molecules participated in the tribo-chemical polymerization as reactant rather than catalyst.

176

Introduction

Most mechanical systems with tribological sliding parts are operated in ambient environments where vapor molecules can adsorb on the surfaces.66,228 Water molecules are ubiquitous and easily adsorbing on most of material surbstrates, which play a critical role in adhesion and wear at sliding interfaces.229,230 To most tribology systems, water is known to be detrimental weather in the form of vapor or liquid.231–233 The adhesive wear rate can be increased when the water vapor is incorporated.234 For metal substrates, the galvanic corrosion occurs with liquid state water which is able to effectively conduct the electrons and ions between the sliding surfaces.235 To guarantee the reliability of mechanical devices, especially those for coastal and marine applications, water effect on lubrication performance is an inevitable topic.

Several lubrication methods have been proposed to solve the problem of tribology system failure, including coatings and self-assembly layers (SAMs).236,237 Lubrication by continuously adsorbing vapors, which is also known as vapor phase lubrication (VPL) has brought interests for the use in small-scale mechanic systems recently.62 At controlled environment with sufficient partial pressure of lubricant vapor, the molecules could adsorb to its equilibrium thickness regardless of interfacial sliding process, achieving the self-healing and replenishment lubricating film.81 The only limitation of VPL is that the vapor pressure of lubricants should keep at least monolayer thickness of adsorbate film. In the case of n-pentanol, the wear could occur on silicon substrate when the supply of lubricant molecules is not sufficient.7 The exposed dangling bond or highly reactive surface sites may initiate chemical reactions in the lubricant phase. Several organic molecules have been demonstrated as effective lubricants for various kinds of materials as long as the monolayer adsorption layer is assured.21

177

During VPL, molecules adsorbed at the sliding solid interfaces might undergo chemical reaction under high contact pressure or frictional shear stress.219,238 At well-lubricated interfaces where the frictional heat are negligible and triboemissions are ruled out due to lubrication layer, one can conclude that the reactions be classified as mechanochemistry. Thus the shear stress or pressure used to be interpreted as the driving force for these reactions.9,10,34,69,70,140 One common model to describe the load dependent mechanochemical reaction is the modified Arrhenius type equation expressed as:

∗ 퐸푎 − 휎∆푉 r = A exp (− ) 푘푏푇 Equation 9- 1

where A means the preexponential factor, kb the Boltzmann constant, Ea the thermal activation energy and T the reaction temperature. Here, the mechanical energy is represented by the product of shear stress σ and critical activation volume ΔV*, which reduces the activation barrier of the reaction. From semi-log plot of reaction rate r and shear stress σ, the critical activation volume

ΔV* can be obtained.9,10,34,69,70,140 The recent ReaxFF force field MD simulation results provide the physical meaning of activation volume.219 Molecules at the tribological interfaces undergo shear induced physical deformation from its equilibrium to the activated intermediate.219 The degree of deformation could be explained via critical activation volume (ΔV*).8–10,16,39,69,70,187

The simulation and experiment results also revealed that such mechanical energy transfer can be facilitated while the molecules are covalently bonded to one sliding solid through surface oxygen.219 Hence, the physisorbed molecules, like n-pentanol, hardly undergo shear-induced deformation probably due to stable molecular structure or ineffective mechanical energy transfer.

It suggests that the substrate did not only work as inert wall to transfer the mechanical energy, but also involves in mechanochemical reactions. The more chemically reactive substrate materials are 178 found to be able to chemisorb some lubricant before the shear stress is applied.37 On surfaces with chemisorption property, the critical activation volume and the activation energy (barrier) appear to be much lower than that on inert surfaces. Such tendency also could be altered by oxidative status of gas environment; for example, oxygen gas significantly increases the mechanochemical reaction rate of adsorbates even on chemically inert material.239

Water is a ubiquitous resource in nature, it has rarely been used as monomer for the construction of polymers. In mechanic systems, water also has been recognized as a contaminant in lubricants.231–233 The study of water effect on shear-induced reactions at tribological interfaces not only bring fundamental interests to the mechanochemistry study, but also provide understandings to lubrication performance. Therefore, this paper investigates the mechanically induced reactions at tribological interfaces of allyl alcohol and water mixture. Allyl alcohol is selected as the model lubricant which has been demonstrated to be reactive under frictional shear because of the unsaturated C=C double bond in the allylic alkyl.166 Previous study suggested that the ratio of water content would affect the thickness and structure of n-alcohol-water binary adsorbate.240 If the water is able to change the chemisorption capability of surface to lubricant monomers, the difference in tribopolymerization rate should be expected. Several mixing ratios of two component were tested based on the phase diagram. The mole fraction of allyl alcohol in the vapor phase was varied from one to zero; the total partial pressure in the environmental controlled chamber was maintained as 70% p/psat of the vapor mixture. The relative thickness and structure of adsorbed molecules at the silicon oxide surfaces is studied via infrared spectroscopy (IR) and target factor analysis (TFA). The comparison of tribopolymer yield rate and wear behavior of allyl alcohol- water vapor mixture clearly revealed the critical role of the water in mechanochemical reactions.

179

Experiment Details

Reciprocating tribometer and vapor flow set-up

A homemade ball-on-flat reciprocating tribometer with environmental control system was employed to measure the friction and wear between a borosilicate glass ball and silicon wafer. The borosilicate glass ball (Pyrex; thermal expansion coefficient = 3.3 ppm/K; diameter = 2.38) has a roughness ~4 nm after removal of the ball curvature. The silicon (100) surface (Wafer World, Inc.

West Palm Beach, FL, USA) was covered with 1− 2nm-thick native oxide. Before the tribotests, the wafer was cleaned with ethanol followed by a UV/ozone treatment to remove the organic residue. The sliding speed of tribometer was kept at 0.4 cm/s and the sliding track distance was

2.5 mm. The flash temperature rise was estimated to be around 5 ⁰C at such low speed.124,166 In friction test, the normal load was 0.1 N, which corresponded to a Hertzian pressure at 0.34 GPa.

The Hertzian deformation depth was 115 nm, much larger than the surface roughness; thus guarantee the intimate contact between the two solid surfaces.125

The mole fraction of the vapor phase for the tests was controlled by varying the composition of liquid according to the vapor/liquid equilibrium curve. In Figure 9-1, the solid red dots on the dew point curve (red dotted line) represent the vapor compositions which were used for friction test. Two pure chemicals (yalcohol = 0 & 1) and five allyl alcohol- water mixtures (yalcohol

= 0.009, 0.02, 0.37, 0.66 & 0.9) were selected. The corresponding liquid mixtures are prepared base on the solid blue dots on bubble point curve (blue dotted line). The environmental control was maintained through the continuous flow mode as nitrogen gas bubbled through these liquid mixtures. To prevent the condensation in saturated vapor environment, the partial pressures of the vapors were set as 70% p/psat by controlling the ratio of saturated liquid stream and dry nitrogen.

That means the vapor pressure (pwater+alcohol, from water rich to alcohol rich phase) of selected 180 mixtures are 2.2 kPa, 2.5 kPa, 2.7 kPa, 3.3 kPa, 3.1 kPa, 2.5 kPa and 2.2 kPa respectively. On the silicon surface, it is sufficient to maintain the monolayer coverage of adsorbates and measurable amount of tribopolymers for mechanochemical reactions.

Figure 9- 1 Water-allyl alcohol vapor/liquid equilibrium curve. The blue dotted line represented bubble point curve while the red dotted represented dew point curve. The mixed liquid compositions were based on the black dots on figure.

Attenuated total reflection infrared (ATR-IR) and Target factor analysis (TFA)

ATR-IR spectroscopy was employed to measure the adsorption spectra of water-allyl alcohol mixture on silicon oxide surfaces. A 725 µm-thick silicon crystal was double-side polished and cut into a dimension to tit to the ATR cell, where both ends of this wafer was polished to a 45˚ bevel angle for IR input and output. The cleaning method for silicon ATR crystal was same as that for tribotests. The infrared experiment details for measuring the adsorption spectra with ATR was discussed in our previous research study. 181

The adsorption of water and allyl alcohol from their mixtures to the solid/air interface was analyzed using target factor analysis (TFA). The theory and applications of TFA are well described in the literature241–243 and only the main steps are discussed here. The experimental ATR-IR spectra of water and allyl alcohol formed a data matrix in which each column is an ATR-IR spectrum. In

TFA analysis, principal factor analysis was first used to determine the number of main components contributing to the data matrix, and to decompose the matrix into abstract row and column matrices via singular value decomposition. The ATR-IR spectra of pure allyl alcohol and water were then used to transform the abstract matrices into physically meaningful ones. The transformation gave rise to row matrix containing predicted ATR-IR spectra of water and allyl alcohol, and column matrix consisting of loadings (component weights) that are the contributions of the predicted spectra in the data matrix. The TFA analysis was performed using Matlab code provided by

Malinowski.243 It was assumed in the analysis that there were no change of molecular conformation as well interactions between molecules in the mixtures.

Topographic analysis

Atomic force microscope (AFM; Digital instrument MultiMode scanning probe miscroscope) was used to image the tribo-polymers piled up at the ends and sides of the sliding track. Silicon AFM probes (TESPA-V2 Bruker AFM probes) were cleaned with UV/ozone to remove any residual contaminats. Te spring constants of the AFM cantilvers were calivrated using the Sader method; [ref] they were ~40 N/m for imaging tests. The scan size of the AFM images was 70 µm × 70 µm. After AFM imaging, the tribo-polymers were rinsed off with ethanol and the slding track was analyzed in ambient air by optical profilometry (Zygo NewView 7300). Through comparing several segments of the AFM images and topography of sliding tracks obtained from profilometry, the total yield of tribo-polymers can be estimated. The tribo-polymers created and 182 accumulated at both ends of the sliding track were analzed with micro-FTIR spectroscopy. It was carried out using a Bruker Hyperion 3000 IR Microscope coupled to IFS 66/s spectrometer with a

15× objective lens.

Results and Discussion

The adsorption of allyl alcohol-water mixtures on either sides of azeotrope composition (y=0.02

& 0.9) and pure allyl alcohol vapor were measured through ATR-IR on silicon oxide surface. As shown in Figure 9-2, the partial pressure p/psat is controlled at 70% to ensure at least one monolayer

-1 adsorbates. The peak at ~3080 cm is clearly detected in all spectra, indicating that the C=C double bond in allyl group is intact in adsorbed allyl alcohol moleclues. It is readily found that the incorpration of water molecules significantly changed the allyl alcohol adsorbate structure. The overall carbon hydride (C-H) peak shapes of the allyl alcohol are different with the presence of water in the vapor (Figure 9-2a & b). When the chamber was purged after measuring the adsorbate specturm at p/psat = 70%, residual organic species were still detected with different relative intensity, depending on the water content in the mixture. In the case of pure allyl alcohol (Figure

9-2c), the relative C-H peak intensity (compared with that before purging) is minimum. In contrast, larger amount of chemisorbed species are seen on the silicon oxides when there is water in the vapor phase (Figure 9-2.a & b). Especially in the case of vapor mixture in water-rich phase, the allyl alcohol only counts a small proportion in vapor phase, but the chemisobed species appear to be significant. This could imply that water may facilitate the chemisorption process of allyl alcohol molecules on silicon oxide surfaces.

183

Figure 9- 2 Infrared spectra of adsorbates on silicon oxide in (a) water rich mixture (yalcohol =0.02),

(b) alcohol rich mixture (yalcohol =0.9) and (c) allyl alcohol vapor at p/psat =70%. The read spectra is taken after purging the ATR chamber with N2 after the water-allyl alcohol mixture measurement.

The ratio of C-H area of adsorbates to the residue is displayed in the yellow box.

Principal factor analysis shows that there are two components mainly contributing to the data matrix containing ATR-IR spectra of water and allyl alcohol mixtures (see Supporting Information,

Figure 9-S2). The loadings (component weights) of allyl alcohol and water obtained from target transformation are plotted as a function of allyl alcohol mole fraction and given in Figure 9-5.

With low proportion of allyl alcohol molecules in the vapor phase (yalcohol < 0.37; azeotrope composition), the water molecules accumulation on silicon oxide is facilitated. However, when the mole fraction increases, allyl alcohol replaces water at the adsorption layer. It suggested that the low proportion of allyl alcohol molecules shows synergistic effect with water adsorption.

184

Figure 9- 3 Loadings (component weights) of allyl alcohol and water obtained from TFA of allyl alcohol and water mixtures, and plotted as a function of allyl alcohol mole fraction.

Figure 9-4 compares the friction coefficient measured with a ball-on-flat reciprocating sliding configuration in 70% p/psat water/ allyl alcohol mixture vapor. It is known that this relative partial pressure is high enough to form layered structure of coverage on silicon oxide.21 In pure water vapor yalcohol = 0, it shows high friction coefficient and large standard deviation, corresponding to

7 the previous study. While with allyl alcohol vapor (yalcohol = 1.0), the measured friction coefficient is low as 0.17 and stable, which is slightly higher than the typical value of boundary lubrication

21 by physisorbed molecules. When the water is introduced to the allyl alcohol vapor (0

1), the friction coefficient grows to 0.3 ~ 0.5. In this case, the friction behavior shows more fluctuation, corresponding to the larger error bar. What governs this friction behavior still remains 185 unclear. It is possible that the formation of tribopolymers might correlate with the friction coefficient. If so, the tribopolymer yield would be higher with the presence of water.

Figure 9- 4 Friction coefficient measured during the sliding of borosilicate glass against silicon wafer in 70% p/psat of several vapor mixture. The liquid used to create the vapor contains allyl alcohol with mole fraction from zero to one, as showed on the right side of friction map. The

Hertzian contact pressure keeps constant at 0.34 GPa.

Figure 9-5 compares the AFM images captured at the two ends as well as one part in the middle of the sliding track in all selected vapor mixtures tested in this study. The tribo-polymers pile mostly at the ends of the slide track as well as the sides of middle parts, which implies that the reactions were taken place at the interfaces and products were pushed to the ends by the reciprocating movement of borosilicate ball. The result from pure water vapor (yalcohol = 0) is not 186 listed here because no polymeric product but only wear can be found after friction tests. (Figure

9-S1) To calculate the tribopolymerization yield, the volumes above the pristine surface level was added up along the entire sliding track. The volumes could be readily obtained for tribopolymers created in pure allyl alcohol and water rich regime, since no wear is observed on reference plane.

(Figure 9-S1) In the case of allyl alcohol rich regime, the surface wear was considered during the tribopolymer yield calculation. The yield of tribopolymers created in allyl alcohol- water mixture vapors appear to be higher than that in pure allyl alcohol. It is already known that the chemisorption process could facilitate the shear-induced reactions. Then, the question is how the water molecules change the chemisorption status of allyl alcohol on silicon oxide surfaces. The result in Figure 9-

2 suggested that the allyl alcohol and water has synergistic effect during co-adsorption. In order to validate the effect of water molecules on shear-induced reaction, the load dependence of the tribopolymerization yields are investigated.

187

Figure 9- 6 Tapping-mode AFM images of tribo-polymers created at both ends and in the middle of the slide track after 400 cycles of reciprocating sliding under a Hertzian contract pressure of

0.34 GPa in allyl alcohol/water vapor created from several liquid mixtures. The image size is 70

μm × 70 μm and height size is -2 μm ~ 2 μm.

Figure 9- 5 Semilog plot of the contact pressure dependence of the tribopolymerization yield for

water/allyl alcohol mixture at yalcohol = 0.02 (blue) and pure allyl alcohol vapor (black; obtained

from previous work). The total reciprocating cycle of the borosilicate ball was 400. 188

When the wear was not observed during tribopolymer formation, such as that in pure allyl alcohol and water rich mixture vapor, where the friction heating and mechanical wear of silicon substrates are negligible, the reaction could be considered as pure mechanochemically induced.

Therefore, the load dependence of the tribopolymer yield was studied for the selected vapor compositions (yalcohol = 0.02 & 1.0). The yield rates show linear behavior versus normal contact pressure on semi-log plot (Figure 9-6), following the modified Arrhenius equation 9-1. Since the shear stress σ (GPa) can be expressed as a linear function of normal contact pressure P (GPa):

σ=σ0 + 훼푃 Equation 9- 2

121,122 Where σ0 is the term determined by the adhesive contacts for small scale contacts, which can be neglected in our macroscopic system. The constant 훼 is interpreted as friction coefficient in the calculation when there is no wear during the sliding. Substituting the σ in modified Arrhenius equation 9-1, it can be written as:

∗ ∗ σ0∆푉 ∆푉 훼 퐸푎 ln 푘 = (푙푛퐴 + ) + 푃 − Equation 9- 3 푘푏푇 푘푏푇 푘푏푇

The sliding speed of the tribo-meter in this experiment was slow, making the temperature raise less than 5 ℃.23 The yield rate k (m/s) is obtained through normalizing the total yield (m3) by the slide track area (m2) and the friction test time of 400 cycles (s).

The critical activation volume ΔV* of allyl alcohol on one surface can be obtained from the slope of each line in the semi-log plot (Figure 9-6). To investigate the role of water in mechanochemical reaction, the yield is normalized by the slide track area and the total time to get the normalized yield rate (with unit m/s). It is noted that the slope is similar for pure allyl alcohol vapor and water-rich mixture. On the basis of the finding about the interpretation of ∆푉∗ from 189 experiment/ simulation study, it represents the degree of shear-induced physical deformation for

219 adsorbed molecules to undergo tribopolymerization. The activation volume in yalcohol = 0.02 is only slightly smaller than that in pure allyl alcohol (yalcohol = 1.0), which can hardly be the explanation for the yield difference. Previous study demonstrated that the substrates with high chemical reactivity improve the tribochemical reaction rate through the oxidative chemisorption.37

This process could also be facilitated by oxidative environment gas.239 It is found that the vapor mixture in water-rich regime (yalcohol = 0.02) shows smaller absolute value of intercept than that in pure allyl alcohol vapor (yalcohol = 1.0), implying that the chemical activation energy (Ea) is smaller for water incorporated allyl alcohol adsorbates. The results support the hypothesis that the presence of water molecules is beneficial during the oxidative chemisorption of alcohol molecules.

Figure 9-7 compares the wear volume and tribo-polymer yield of allyl alcohol- water vapor mixture at different ratios. It clearly shows that there are two regimes of wear behavior on either side of azeotrope point – a wear free region (brown) and worn region (green). In the case of pure chemical vapor (yalcohol = 0 & 1.0), the surfaces show damage in water vapor while well lubricated via tribo-polymers created in allyl alcohol vapor, which agrees with the previous results.7,166 In water rich phase, no wear can be observed from profilometry image (Figure 9-S1). The surface seems to be well protected by adsorbed molecules and mechanically created tribo-polymer. With increasing allyl alcohol ratio in the mixture till azeotrope point (Rallyl = 37%), the yield of tribo- polymers keeps growing to the maximum. Here the yield is twice as much as that created in pure allyl alcohol vapor. Then turns to allyl alcohol rich phase, the trend of tribo-polymer yield versus alcohol mole fraction is decreasing, but still higher than that in pure allyl alcohol. Meanwhile, some mild damage is found from the cross-sectional topography images. What governs the wear 190 behavior cannot be determined from current study. It is possible that the oxidative chemisorption of allyl alcohol molecules at water-rich phase fully occupy the surface sites and provide lubrication effect. However, in alcohol-rich phase, some surface sites would be still occupied by water molecules due to relative less chemisorption (as shown in Figure 9-2). Under this circumstance, the adsorbed water on silicon oxide may weaken the Si-O bond and facilitate the formation of

7 Siball-O-Sisubstrate bridges, which lead to tribochemical wear during reciprocating sliding.

Figure 9- 7 Tribo-polymerization yield (red) at wear volume (black) on silicon wafer in vapor environment created from liquid mixture with different allyl alcohol mole fraction.

191

Figure 9- 8 IR spectra of tribo-polymers piled up at the ends of the track on silicon wafer after

400 reciprocating cycles of tribo test under water/allyl alcohol vapor mixture (yalcohol= 0.9, blue; yalcohol= 0.02; red) and pure allyl alcohol vapor(black) conditions.

It is difficult to determine the exact role of water adsorbates during tribo-polymer formation. If the adsorbed water only play the role as catalyst or intermidiate during tribopolymerization reaction, the chemical structure of produced polymers would be similar regardless of the water content.

However, the relative peak intensity of IR spectra in Figure 9-8 are different depending on the water content. The relative areas of hydroxyl group (-OH) to carbon hydride group (C-H) are listed in Figure 9-8 1. It is noted that ratio AOH/ACH is smaller for the tribopolymers in pure allyl alcohol 192

vapor (yalcohol = 1.0) compared to that in water- allyl alcohol mixture. As discussed above, it is possible that the water molecules facilitate the oxidative chemisorption process of allyl alchol or the C-O-C covalent bond formation, which in turn creating more hydroxyl tail in the tribopolymer chain. On the basis of this, it might also explains the higher yield rate for tribopolymers in water- allyl alcohol vapor mixture.

Conclusion

The effect of water on the tribochemical polymerization and wear behavior of allyl alcohol during vapor phase lubrication is studied. The incorporation of water molecules could facilitate the chemisorption capability of allyl alcohol to silicon oxide surfaces, which in turn improves the mechanochemical reaction rate. In water rich phase, the trend is esspecially obvious; where the substrate show highest chemisoption ratio even the water content in vapor phase is limited. It is suggested that the chemisorpiton of interfacial molecules helps transfer the mechanical energy to the reaction coordinate, which in turn lower the activation barrier of mechanochemial reaction.

The proportion of water is critial to the wear behavior. The siliocon substrates tend to wear more in alcohol-rich environment while be protected in water-rich mixture.

193

Supporting Information

Figure 9-S 1 Surface topography after tribo-test in allyl alcohol- water mixture vapor environment

Figure 9-S 2 Target (test) and predicted spectra of allyl alcohol and water used in TFA of liquid mixtures

194

Chapter 10

Low-Toxicity Ionic liquids as environmentally-friendly lubricant additives

Xin He1, Chanaka Kumara1, Huimin Luo2, Teresa J. Mathews3, Jun Qu1,*

1Materials Science & Technology Division, Oak Ridge National Laboratory, Oak Ridge TN,

USA

2Energy & Transportation Science Division, Oak Ridge National Laboratory, Oak Ridge TN,

USA

3Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge TN, USA

Overview

Hydraulic systems provide a reliable, efficient, and cost-effective means to transmit energy than electrical or other mechanical systems, leading to the widespread use of hydraulic fluids globally.

Because >50% of all lubricants sold globally end up in the environment by volatilization, spills or other loss applications, it is increasingly recognized that in addition to meeting the rheological and tribological requirements, hydraulic fluids should also be low or non-toxic and adequately biodegradable. Here, we present our latest development of oil-soluble, low-toxicity ionic liquids

(ILs) as candidate anti-wear additives for hydraulic fluids. The tribological behavior and environmental toxicity of selected ILs, at 0.5 wt.% concentration blends in a hydrophilic polyalkylene glycol (PAG), an oil soluble PAG (OSP), and a non-polar mineral base oil relative to that of a conventional anti-wear additive, primary zinc dithiophosphate (ZDDP). Tribological performance of the IL-containing lubricants were tested using a steel-steel contact under boundary 195 lubrication at 82 ⁰C. The candidate ILs showed lower friction coefficient and wear volume than

ZDDP. Worn surface characterization revealed the microstructures and compositions of the protective tribofilms formed on the contact areas. Acute and chronic toxicity tests which exposed

Ceriodaphnia dubia, a common aquatic bioindicator species demonstrated that most candidate ILs are more benign than ZDDP. Because of their higher thermal stability, improved lubricating functionality, and lower toxicity, these ILs may potentially be qualified for use in environmentally friendly lubricants (EALs) for hydraulics. 196

Introduction

World wildly the production, application and disposal of the lubricants may easily bring severe environmental pollution and health aspects.244 The use of Environmentally Acceptable

Lubricants (EALs) becomes a requirement as public issues to protect the nature and living beings in special.245,246 Particularly, hydraulic fluid plays a role as energy transfer medium in automobile automatic transmissions, brakes, and power steering.247 To address the current EALs in application for marine or coastal industry, attempts have been made on hydraulic fluid to meet the environmental friendly criteria, to be more specific, readily biodegradation and low aquatic toxicity. Four major types of lubricants have been determined by Environment Protection Agency

(EPA) to prevent the potential hazardous pollution from the lubricant discharge or accidental spills.

Water and water based lubricants are effective for limited light-duty components.248 Vegetable oils and synthetic ester are presents decent lubricating properties and biodegradability.249,250 However, the low tolerance to elevated temperatures may lead to varnish deposits in hydraulic pump.251

Vegetable oils even form acids and sludge when encountered with water, which are corrosive to equipment.252 Polyalkylene glycols (PAGs) not only possess outstanding load-bearing and lubricating property, but also are able to tolerate heat that readily create vanish in other fluids.253

As eco-environmental lubricant, PAG is biodegrable and water soluble.247 The recently developed oil soluble PAG (OSP) allow the formulators to incorporate the advantages of PAG in combination with conventional mineral based oil.254

A typical hydraulic system contains a number of additives including antioxidant, dispersant, anti-foam and anti-wear additive. Zinc dialkyldithiophosphate (ZDDP) is the most common anti- wear additive in automotive lubricants currently. It is known to be low-cost and effective in lubricant industry. ZDDP is toxic because of its zinc content and long-chain thiophosphate.254,255 197

The lower-toxic alternatives, ashless phosphorous-based (e.g., TCP) anti-wear additives, however have inferior wear protection functionality.

Ionic liquids (ILs) have been explored as the potential ashless additives possessing outstanding properties including non-volatility, high thermal stability, low flammability.13,256

Compared with common ZDDPs, ILs have stronger absorbance to positively charged metallic substrates due to ionic structure, hence present promising tribological properties. One concern for

IL additives used to be their limited solubility in non-polar hydrocarbon oils. It has been learned that the 3D quaternary structure with relatively long alkyl chains could effectively reduce the charge density to improve the solubility in non-polar base oils.257 On the basis of this theory, several achievements on oil miscible IL additives have been reported.258 The current literature on developing ILs as lubricant additives has been focused on the ILs’ physiochemical and tribological properties, but there is little info about the environmental impact of the oil-soluble ILs.

The common agreement among most of studies emphasized the toxicity of cations over the anions. There are limited reports for the toxicity of certain groups of halogen-containing ILs and results suggested that they raise serious concern for both environment and lubrication performance.

259,260 It is known that the long linear alkyl chains in head group of cations play a determined role in toxicity. The wildly used imidazolium cations were also proved to be hardly biodegradable due to amide functionality.261 While considered as the relative less impact on toxicity, the anion structure showed potential influence on toxicity and corrosion. It is reported that the side chain length are pronounced to the toxicity of ILs on living cells.262 Shorter or branched alkyl chain in ionic liquid could promise the biodegradability; but in such a case, the problem of solubility reoccurs. To develop the environmentally acceptable IL additives for hydraulic applications, the 198

extensive experimental comparison is required to evaluate the lubrication performance as well as

the toxicity.

This study intends to develop low-toxicity ILs as potential eco-friendly lubricant additives.

Three types of base oil, including PAG, OSP and mineral oil were selected. A primary ZDDP and

several potential ILs were brought into comparison for lubrication performance as well as aquatic

toxicity to the water flea Daphnia magna. The tribo-film chemistry properties and topography were

measured by various techniques. Daphnia are known to be the ideal microorganisms for toxicity

testing because of their short reproductive cycle, sensitivity to water quality, and widely cultured

in an aquarium. The acute toxicity of additives in water miscible PAG were determined using

standard U.S. Environmental Protection Agency (EPA) protocols.

Figure 10- 1 Molecular structure of base oils and ionic liquids (ILs)

199

Experiment details

Three base oils chosen for this study were Polyalkylene Glycol (PAG), Oil Soluble PAG

(OSP) and mineral blend oil. The PAG (UCONTM Lubricant 50-HB-170) and OSP (UCONTM

Lubricant OSP-32) were provided by Dow Chemical Company. The viscosity at 40 ⁰C of these two polar base oils measured on a PertroLab Minivis II viscometer were 33.0 and 30.4 cSt respectively as shown in Table 10-1. The mineral blend oil is mixed by Chevron Neutral Oil 100R and Chevron Neutral Oil 220R (Chevron Corporation) to reach the target viscosity (32.0 cSt) at 40

⁰C based on the Refutas equation. A simplified equation for mineral blends is expressed in the following form

휐푏푙푒푛푑 = 푒푥푝(푒푥푝(휒표𝑖푙−1 ln(ln(휐표𝑖푙−1 + 0.8)) + 휒표𝑖푙−2 ln(ln(휐표𝑖푙−2 + 0.8)))) −0.8 Equation 10- 1

Where υ is the kinematic viscosity in centistokes and χ is the mass fractions of oil in the mineral blend. A primary ZDDP was purchased from Lubrizol. Trioctylammonium bis(2-ethylhexyl) phosphate [N888H][DEHP] was synthesized in our lab. The details of the procedure was discussed in previous work.257,260,263,264 The design of newly developed ILs were based on Lubricant

Substance Classification (LuSC) list. The details are not shown here due to invention disclosure.

A TGA-2950 (TA Instruments) was used to conduct the thermogravimetric analysis (TGA) at a 10

⁰C/min heating rate in air. The selected additives were made by adding 0.5 wt% into PAG, OSP and mineral blend, respectively. The solubility of ILs in base oils were determined by direct observation after centrifuge test at 13000 rpm for 5min.

200

Table 10- 1 Viscosity of Selected Base Oils

Base oil Measured viscosity (cSt) 100 °C 82 °C 40 °C 23 °C PAG (50-HB-170) 7.1 10.2 33.0 67.0 OSP-32 5.8 8.4 30.4 66.1 Neutral Oil 100R 3.9 5.7 20.0 42.3 Neutral Oil 220R 5.9 9.1 39.9 97.6 Mineral blend oil 5.2 8.0 32.7 76.5 Mixed based on Refutas equation

A Plint TE 77 tribometer with a reciprocating ball-on-flat configuration was employed to conduct all boundary lubrication tribotests. The M2 tool steel substrates were purchased from

McMaster-Carr (Elmhurst, IL, USA). The substrates were polished with fine grit sandpaper to reach a roughness (Ra) in the range of 60 ~70 nm. An AISI 52100 bearing steel ball (10 mm diameter) was selected as the counter surface, which has a roughness around nm. Both surfaces were clean with isopropanol to make sure there is no residue. The friction and wear tests were performed with a constant load of 100 N, which corresponded to a maximum initial Hertzian

2 2 contact pressure of 1.68 GPa. Thus, the composite roughness σ (√푅푞,푏푎푙푙 + 푅푞,푠푢푏푠푡푟푎푡푒 ) is estimated to be nm at contact. The maximum ratio of lubrication film thickness (h) to the composite roughness σ, known as λ ratio (h/ σ), is <0.1at operating temperature, ensuring boundary lubrication (λ < 1). The oscillation frequency used was 10 Hz and the sliding span was 10 mm. All tribotests were conducted at a temperature of 180 ˚F (82˚C) to prevent the seal compounds damage and degradation of the oil. After the friction test, the residue oil was rinsed off with isopropanol and the slide track was analyzed by optical profilometry (Wyko NT9100). 201

After tribotests for 1000 m total distance, the wear scar was investigated using scanning electron microscopy (SEM) and energy-dispersive spectroscopy (EDS) on a JEOL 6500F SEM equipped with an EDAX system. To obtain the morphology and chemical composition of the outermost layer, EDS analysis was performed at a potential of 5.0 kV over a constant time of 100 s on wear scar on ball surfaces. A Thermal Scientific K-Alpha XPS system was used to conduct the X-ray photoelectron spectroscopy (XPS) analysis. A 400 µm diameter spot was focused during the acquisition of spectra. The depth profile measurements of the tribo-films were obtained by using an argon-ion sputter gun at 3.5 keV. The Shirley background was used to subtract the backgrounds during analysis and a symmetric Gauss-Lorentz product line shape was applied for curve fitting.

Ceriodaphnia dubia is known as one model organism for aqueous toxicity tests due to its fast reproduction rate and high sensitivity to environment change. (Designing ionic liquids: the chemical structure role in the toxicity. Ecotoxicology) The laboratory bred, less than 24 hours old

Daphnia neonates were used in this test, which were all born within 8 hours of each other. The acute and chronic toxicity tests were conducted with six PAG based lubricants (neat oil, 5 wt% of

ZDDP, [P8888][DEHP], [N888H][DEHP], IL-1 and IL-2 respectively). The stock solution used were diluted to 0.001% for acute and chronic tests with Dilute Mineral Water (DMW; laboratory control water.) and the water renewal period was 24 hours. The polystyrene microbeakers were used as test chamber with minimum 15 mL water for each. The temperature for all tests were controlled at 25.0 ⁰C. In acute toxicity tests, 4 neonates were put in each chamber and 3 replicates were performed per treatment. The 5-day period tests used survival rate as the criteria to assess the toxicity. In chronic toxicity tests, each chamber contained 1 neonate and 10 replicates were 202 performed per treatment. The toxicity is assessed with respect to the survival rate and reproduction rate in 7-day period experiment.

203

Results and Discussion

Table 10- 2 Experimental solubility of ILs in base oils

Solubility (wt. %) PAG (50-HB-170) OSP-32 Mineral blends ZDDP >5% >5% >5% [N888H][DEHP] >5% >5% >5% IL-1 >5% >5% <1% IL-2 >5% >5% <1%

Table 10-2 lists the experimental solubility of ILs in selected base oils. It is noted that the solubility of IL-2 and IL-1 are both <1% in mineral blend oil. It is hypothesized that the poor solubility might correlate with the short alkyl chains in head groups. The ZDDP and

[N888H][DEHP] showed solubility >5%, which were attribute to the long carbon tail in head group.

TGA curves of selected ILs and base oils are compared in Figure 10-2. All chemicals showed similar thermal stability except for IL-2, with decomposition temperature at ~200 ˚C. The IL-2 has higher thermal stability (~250 ˚C) probably due to the stronger dipole-dipole interactions between cation and anion groups. It is notice that, the ZDDP left ~20 % of weight after decomposition, which is higher than other ILs and base oils (<10% and 0% respectively). It is speculated the metal zinc contributes to the residue. The high content such ash solids could be detrimental to the environment, making ZDDP not priority choice for EALs.

204

\ Figure 10- 2 TGA curves of base oils (PAG, OSP, mineral blends), and oil-miscible ILs

Figure 10- 3 Friction behavior of PAG (a), OSP (b) and mineral blends (c) with ZDDP

and ILs. The stabilized friction coefficient of last 200 m were plotted in (d). The wear

rate on 52100 bearing ball surface is shown in (e) 205

The friction coefficients measured with 52100 bearing steel ball sliding on M2 steel substrate in neat PAG, OSP and Mineral oils as well as containing 5 wt% primary ZDDP and ionic liquids are summarized in Figure 10-3. On relative polar base oils (PAG and OSP), the friction coefficients measured in base oil are as low as 0.83 and 0.98, respectively; but, the coefficient in mineral blend oil is higher (around 0.12). It was speculated that the lower friction coefficient in polar base oils might attribute to the stronger adsorbance of molecules. PAG has higher O/C ratio and possesses more polarity to promote the interactions with the metal substrate, so that gave a lower friction coefficient than OSP. As a result, the wear rates of both polar oil are around 2×10-9 mm3/N·m, which is only half of that in nonpolar mineral blends. When the base oils were combined with primary ZDDP, the friction coefficients shows comparable values with neat base oil in OSP and mineral blends. The value is slightly higher for ZDDP additized PAG than neat base oil, which is probably due to the inevitable competition in surface adsorption. It is observed that a significantly reduced wear rate (30 ~ 50%) was achieved by adding 0.5 wt% ZDDP.

The ammonium-based ILs ([N888H][DEHP], green and IL-1, blue) show significant improvement on friction and wear, as presented in Figure 10-3 (a), (b) and (c). Owing to their inherent polarity (ions) for stronger adsorption, the friction coefficient decreased from the initial

0.09~0.10 and stabilized at 0.6 for PAG and 0.8 for OSP and mineral blends, respectively. The formed tribofilm help reduce the wear rate by one order of magnitude. The phosphonium IL IL-2 with shorter alkyl chain appeared to be less effective than ammonium based ILs. These IL-2 molecules with high polarity seem to be prone to form clusters in the base oils. It produced only slightly higher friction coefficient around 0.09 in base oils, due to the friction modifier features of 206 the clusters. In this case, the low surface coverage of IL may not create continuous tribofilm during reciprocating sliding, which in turn leads to a higher wear rate. This phenomenon is particularly notable in mineral blends, the wear rate (4 × 10-9 mm3/N·m) is as high as that in neat oil.

Figure 10- 4 SEM surface morphology of wear scar on 52100 steel ball surface

Figure 10-4 compares the SEM images of the scar on ball surfaces created in ZDDP and

IL-1 additized base oils. It is obvious that ZDDP + mineral blends generated a smoother and more uniform tribofilm. The high oxygen and phosphor contents in EDS spectrum (Figure S10-1) indicate that the thick film is created at worn surface. In contrast, the tribofilm formed in ZDDP additized polar base oils is thin with uneven distribution, as shown in Figure 10-4 (a) and (b). In

EDS, the oxygen and phosphor contents are decreasing as the base oil become more polar, 207 suggesting a possible competition between the primary ZDDP and base oil molecules for surface adsorption sites. This also explains the discontinuous coverage of tribofilms in PAG and OSP base oil. Interestingly, the tribofilm created in PAG showed a brittle structure, suggesting that the tribofilm is mainly composed of iron sulphide. The iron/zinc phosphate layer can hardly deposit on the top in PAG base oil. [ref Martin] Incorporated with IL-1, the same trend can be observed for these three base oils. The tribofilm is barely seen with naked eyes in PAG and OSP oil, corresponding to the negligible phosphor peak in EDS. Different from ZDDP, the IL-1 + mineral blends created thick but uneven distributed tribofilm. This could correlate with the poor solubility of the ILs (< 0.1%).

The XPS spectra of P 2p and O 1s on the 52100 steel ball surface lubricated via 5 wt%

ZDDP in PAG, OSP and mineral blends base oils are shown in Figure 10-5. The deconvolution of

P 2p presented 2 peaks due to spin orbit coupling; a 2p3/2 (green) peak with higher binding energy

(BE) and 2p1/2 (blue) peak with lower BE. Whereas the underlying area of 2p3/2 is twice as much as that of 2p1/2. The decomposed phosphate anion usually combined with metal ions from wear debris to create tribofilms. In Figure 10-5 (a)~(c), the P 2p peak is centered at 133.1 eV for ZDDP tribofilm created in PAG oil, while the peak gradually shifts to 133.9 eV for that created in non- polar mineral blends. It is noticed that the tribofilm grown in polar PAG oil showed fracture structure (Figure 10-4), corresponding to relatively higher S content. Under this circumstance, the cross-linking degree of polyphosphate networks is lower due to the impurity.265 In contrast, the polymerized phosphate quickly builds up on worn surface in mineral blend oil because the more polar ZDDP will dominate the surface adsorbate sites. The phosphate with tetrahedral structure could form on the less polymerized, sulfur rich phosphate layer, so that the red shift in BE peak position is observed in phosphorus component. In Figure 10-5 (d)~(f), the O 1s can be deconvolute 208 to two separate peaks, iron oxides and others possible component, including phosphate and carboxylate. The tribofilm created in OSP base oil showed highest relative phosphate ratio, corresponding to the low wear rate shown in Figure 10-2. The depth-composition profile of the tribo-films formed in ZDDP additized lubricants are presented in Figure 5 (g)~(i). On ZDDP tribo- film created in non-polar mineral oil, the phosphorus concentration started at ~10 at. % on outmost layer and slowly dropped to ~2 at. % after 800 s Ar ion beam sputtering; while that in polar PAG oil quickly decreased till less than 2 at. % (~200 s) with the depth of sputtering. Interestingly, the tribo-film formed in OSP showed lowest thickness. It is speculated that the relative phosphate content show more significant effect on wear reduction than the film thickness.266

Figure 10-6 show the binding energies of elements and depth composition profile of tribo- film on worn scar of 52100 steel ball surface lubricated through IL-1 additized base oils. Similarly, the signal of P 2p exhibited 2 deconvoluted component: 2p3/2 and 2p1/2. The ILs tribo-film show same P 2p peak position around 133.2 eV, meaning the polymerization of phosphate is consistent regardless of the polarity of base oil molecules. The relative ratio between oxygen in phosphate and iron oxide is negligible as presented in Figure 10-6 (d)~(f). The depth profile shows that the concentration of phosphorus started around 10 at. % and decreased to ~ 1 at% during the time of ion sputtering. The sputtering time for tribofilm formed in PAG, OSP, mineral blends are 50s,

100s, and 250s respectively, indicating the strong correlation between film thickness and base oil polarity. 209

Figure 10- 5 XPS core-level spectra of P 2p and O 1s on worn 52100 steel ball surface lubricated with 5 wt% ZDDP in PAG, OSP and Mineral blends base oils respectively. (g)

~ (i) show the depth profile

210

Figure 10- 6 XPS core-level spectra of P 2p and O 1s on worn 52100 steel ball surface lubricated with 5 wt% IL-1 in PAG, OSP and Mineral blends base oils respectively. (g) ~ (i) show the depth profile

211

Table 10- 3 Survival tests of daphnia with selected lubricants

Days No additive IL-1 IL-2 [M888H][DEHP] [P8888][DEHP] ZDDP 1 10 10 10 10 10 10 2 10 10 10 6 0 0 3 10 10 10 0 0 0 4 10 10 10 0 0 0 5 10 10 10 0 0 0 6 10 10 10 0 0 0 7 10 10 10 0 0 0

Table 10- 4 Reproduction tests of daphnia with selected lubricants

Days No additive IL-1 IL-2 [N888H][DEHP] [P8888][DEHP] ZDDP 1 0 0 0 0 0 0 2 0 0 0 0 0 0 3 32 36 45 0 0 0 4 2 0 13 0 0 0 5 91 98 82 0 0 0 6 128 159 141 0 0 0 7 175 201 172 0 0 0 Grand total 428 494 453 0 0 0

Chronic toxicity tests on Ceriodaphnia dubia were carried out to assess the aquatic toxicity

of six PAG based lubricants (neat oil, 5 wt% of ZDDP, [P8888][DEHP], [N888H][DEHP], IL-1

and IL-2). Table 10-3& 4 gives the survival rate of Daphnia in chronic toxicity tests, as well as

reproduction in chronic toxicity tests for each trail of lubricants. The PAG neat base oil shows 100%

survival rate in 7-day chronic toxicity tests. It is noticed that the 5 wt% IL-1 (blue) and IL-2

(magenta) additized PAG lubricants are as benign as neat base oil from these two tests. It was

speculated that the short alkyl chain length in head group might correlate with their less toxic

property, validated our environmentally-friendly ILs design criterion. For [P8888][DEHP] (cyan),

[N888H][DEHP] (green) and ZDDP (orange) with longer alkyl chain (C>8) length, they show 212 higher toxicity during short time of exposure (Daphnia neonates survival rates reduced to 0 less than 3 days). The reproduction rate in chronic tests reflect the long term effect of additives on

Daphnia neonates. In accordance with survival tests, the additives with long alkyl chain

([P8888][DEHP], [N888H][DEHP] and ZDDP) negatively affected reproductive output. All 10 replicate groups of Daphnia neonates gave no offspring per individual. In EPA approved PAG base oil, the neonates successfully reproduced and had 43 number of offspring per individual.

Interestingly, this number is slightly higher when PAG base oil is additized with 5 wt% IL-1 or

IL-2; they could each give 49 and 45 number of offspring, respectively. It appears that these two short chain ILs are able to positively stimulate the reproduction ability of Daphnia neonates or they are as benign as the PAG molecules.

Conclusion

Eco-friendly ILs were synthesized and characterized for their anti-wear properties in polar and non-polar base oils as well as the toxicity. The ILs showed the thermal decomposition temperature higher than that in hydraulic systems. The top candidate ILs (e.g. IL-1) demonstrated superior friction reduction and wear protection in all three base oils, in comparison with the conventional ZDDP. Through the characterization of tribo-films, it is noticed that the anti-wear performance showed no correlation with film thickness. The toxicity tests showed high variance for ILs; while some ILs used for engine oil additives are similar to ZDDP, the newly designed ILs

(IL-1 & IL-2) presented very low toxicity.

213

Supporting Information

Figure 10-S 1 EDS spectra of IL-1 and ZDDP additized tribo-films in selected base oil

214

Appendix

Contact Mechanics versus Amonton’s Law in Macroscale Friction

– Experimental and Computational Study

Xin He1, Zhong Liu2, Lei Chen3, Jane Wang2, and Seong Kim1,*

1 Department of Chemical Engineering and Materials Research Institute, Pennsylvania State

University, State College, PA, United States

2 Department of Mechanical Engineering, Northwestern University, Evanston, IL, United States

3 Tribology Research Institute, State Key Laboratory of Traction Power, Southwest Jiaotong

University, Chengdu, China

Overview

This study discusses what governs the interfacial friction in the absence of wear. In the nanoscale, various contact mechanics models have been successfully used to describe the friction responses in terms of contact areas with a proportionality constant term called shear strength. In these contact mechanics models, friction force typically varies in proportion to the 2/3 power of the total load.

However, in the macroscopic scale, friction force is observed to be linearly proportional to the applied load; this is called Amonton’s law. If we put the Hertzian contact mechanics equation the

Amonton’s law equal, then the shear strength of the macroscopic contact would vary in proportion to the 1/3 power of the applied load. This relationship is investigated experimentally at different scale of surface roughness and load in vapor phase lubrication. The results incorporated with 215 mathematical calculation suggested that the interfacial shear behavior is affected by surface roughness as well as shear hardening. 216

Introduction

The friction force between two sliding surfaces is observed to be proportional to the normal contact load over a wide scale range. This empirical relationship is known as Amonton’s Law.

Friction is defined by the Amonton’s Law which predicts a linear dependence of friction on applied load without any strong dependence on contact area A, as shown in equation

퐹 (푓푟𝑖푐푡𝑖표푛 푓표푟푐푒) = 휇 × 퐿 (푁표푟푛푎푙 푙표푎푑) Equation A- 1

This approximation is valid for various systems including dry and lubricated rubbing surfaces. The

Amonton’s Law describe the macroscopic phenomenon, while the contact area is assumed to be projected area rather than the real contact area. However, for Hertzian contact mechanics267, the model works reasonable well to predict the contact area at the macroscale as long as the surface roughness is much smaller than the elastic deformation of substrate at a given load, as shown in

Figure 1 (a). In contact mechanics, the friction is described as the product of shear strength and contact area, as equation A-2:

퐹 = 휏 (푠ℎ푒푎푟 푠푡푒푛푔푡ℎ) × 퐴 Equation A- 2

Equating these two equations (A-1 & A-2), one gets a peculiar relationship:

휏 ∝ 퐿1/3 Equation A- 3

In other words, the shear strength of contact increases in proportion to L1/3 in the macroscale, while it is usually constant in the nanoscale friction.

Several attempts have been done by researchers to mathematically explain the Amonton’s

Law in macroscale. A convincing mechanism was proposed by Bowden and Tabor in twentieth century.268 In real contact, only a tiny fraction of contact surfaces are able to form junctions due to 217 inevitable surface roughness, so called real contact area.269–271 The friction force is modeled as linear relationship with the contact area of the involved asperities. Greenwood and Williamson’s model is based on the assumption that the idealized asperities with an exponential height distribution.272 Here the normal load only works through increasing the ratio of asperity contacts, and there is no plastic deformation taken into account. The friction force can be expressed as an equation proportional to the number of contacts and hence the normal load.

This curious relationship between shear strength 휏 and normal load L is difficult to confirm experimentally because of the complexity of macroscale contact associated with surface roughness and wear. This complexity can be circumvented by using vapor phase lubrication (VPL) tests on surfaces with varying roughness.65,130 The VPL works through the vapor molecules from the environment adsorbing on the surface to form full-coverage layers, which in turn prevent wear. It takes advantage of self-healing characteristic regardless of sliding process. As long as the vapor pressure of the surrounding gas is maintained above a certain level, the lubricant molecules could adsorb to its equilibrium thickness.21 In the case of n-pentanol, the surface can be well-lubricated with vapor pressure larger than 10% relative to saturation.22,79 The data on perfectly smooth surface can then be obtained from extrapolating the friction force measured on surfaces with different roughness.

In this paper, we studied the effect of surface roughness and normal load on shear strength during vapor phase lubrication. The n-pentanol vapor phase lubrication system is applied to prevent wear. Incorporated with elastic-plastic contact model calculation, we focused on the application range of 휏 ∝ 퐿1/3 and how does surface roughness affect the relationship between τ and L. The hypothesis was the plastic deformation may dominate the interfacial shear behavior at high roughness. The friction tests monitored friction force change as a function of cycles at light 218 and heavy loads. The mechanical properties of the sliding track were compared with clean substrates.

Figure A- 1 the schematic of ideal and non-ideal contact

Experiment Details

The vapor phase lubrication (VPL) tests were performed with a home-made ball-on-flat reciprocating tribometer equipped with an environmental control chamber. The ball and substrate material are AISI 440C stainless steel (McMaster-Carr). The substrate was polished with sandpapers with different fine grits and a polishing solution with 1 µm colloidal alumina. The surface finish roughness Rq (root mean square) ranged from 0.022 ~ 0.151 µm. The ball material

(diameter = 3/32 inch) showed smooth surface with roughness less than 0.010 µm. A silicon wafer

(100) was purchased from Wafer World, Inc. The as-received wafer had native oxide layer around

1 nm thickness. The initial Hertzian contact area and indentation depth are provided in Table A-1 219 as a reference. The substrates were cleaned with organic solvent and followed by 20-min

UV/ozone treatment to remove the chemical contaminants.52 The sliding speed was 0.4 cm/s and the sliding track was 2.5 mm.

Table A- 1 The Hertzian contact mechanics on Rq = 0 substrate

Applied load, N 0.05 0.1 0.2 0.5 Max.Hertz contact pressure, 0.447 0.563 0.710 0.964 GPa Hertz indentation depth, µm 0.045 0.071 0.112 0.207 2 Hertz contact area, mm 1.67E-4 2.66E-4 4.23E-4 7.78E-4

The vapor control was maintained with a continuous flow mode during the friction tests.

N-pentanol is selected as the vapor lubricants due to superior lubricity. The details of the vapor control system were described elsewhere. The partial pressure (p/psat) of n-pentanol (Sigma-

Aldrich) was 40% relative the saturation, which ensures full monolayer coverage. After the friction test, the slide track was analyzed by optical profilometry (Zygo NewView 7300). Nanomechanical properties were studied by nanoindenter (Hysitron TI 950) with a Berkovich tip. The penetration depth was controlled in the range from 50 to 150 nm during nanoindentation tests. The average and standard deviation of elastic modulus and hardness were obtained from 15 repeat tests for samples inside and outside the slide tracks.

The contact area on rough surface is numerically calculated by elastic-plastic contact model.

The permanent deformation due to plastic strain is considered into the calculation. The gap between surfaces h can be expressed as 220

ℎ (푥, 푦) = ℎ0 + 푔 (푥, 푦) + 푣 (푥, 푦) + 푣푝 (푥, 푦) + 푠 (푥, 푦) − 훿 Equation A- 4

where h0 represents the initial gap, g is calculated based on geometry, 푣 and 푣p are elastic deformation and plastic deformation respectively. s is contributed from surface roughness and δ is nominal approach. The details of the calculation and flowchart is introduced in details elsewhere.273

Results and Discussion

Friction coefficients measured with different loads in 40% p/psat n-pentanol vapor on surfaces with adjusted roughness (0.022 ~0.151 µm) are plotted in Figure 2. The data represent the average and standard deviation of 100 reciprocating cycles of sliding for multiple tests at each condition. The vapor condition (p/psat = 40%) adequately allow the full coverage of adsorbed n- pentanol molecules without condensation. It is found that in Figure A-1 (a), the friction force shows higher value on smoother substrate under fixed normal load. In the case of vapor phase lubrication, the shear strength only depends on the physical properties of lubricant molecules; thus the shear strength τ is assumed to be a constant for all VPL tests. On the basis on equation A-2, it is speculated that real contact points (larger A) would be larger on smooth surfaces. This statement is verified by Table A-1 from computational study as well. From Figure A-2 (a), the theoretical friction coefficient at Rq = 0 can be obtained from intercept. The result is plotted in Figure A-2 (c), where the slope (0.144±0.0021) represents the friction coefficient at 0 roughness. While the friction force is plotted versus normal load, the surface roughness plays a role in linear fitting

(Figure A-2 (b)). The slope can be considered the friction coefficient at a given roughness Rq.

Figure A-2 (d) shows the correlation between friction coefficient and surface roughness, which 221

can be fitted by linear equation. Then the intercept would be the friction coefficient at Rq = 0. The value is reasonable close to that in Figure A-2 (c), reflecting the boundary lubrication coefficient of adsorbed n-pentanol layers.

Figure A- 2 the average friction force with respect to surface roughness and normal load at 440C stainless steel interfaces. The data on perfectly smooth surface (c) & (d) were obtained from extrapolating the friction force measured on surfaces with different roughness (a) & (b).

222

Figure A- 3 Optical microscope image (left), profilometry image (middle) and cross-section line profile (right) inside and outside the slide track. The experiments were performed with 0.5 N normal load on surface with Rq = 0.054 µm and 0.151 µm respectively.

One thing should be notice is that the topographic change after friction test is

indistinguishable in macroscopic, meaning that the real contact area remained same before and

after experiment. Figure 3 shows the line profiles inside and outside the slide track under 0.5 N

normal load (highest) for smooth and rough substrates. The line roughness is in the range of error,

so that the sliding process did not change the macroscopic asperity structure.

To reveal the relationship between shear strength and normal load, the real contact area is

estimated via plastic-elastic model numerically. Table A-2 presented the real contact area

(contributed from elastic + plastic deformation (blue) and only elastic deformation (white)

respectively) on rough substrate with certain normal load. The contact area change due to plastic

deformation then can be obtained, as shown in Figure 4. At high load or roughness, the area

difference from ideal condition become larger, meaning the plastic deformation plays a more 223

important role. Then, the numerical model base on purely elastic deformation would not work to

explain the relationship between shear strength and normal load.

Table A- 2 The numerically calculated real contact area contributed by plastic deformation &

elastic deformation (blue) and elastic deformation alone (white)

Rq, μm 0.05 N 0.1 N 0.2 N 0.5 N 0.022 1.58E-04 2.38E-04 3.92E-04 7.30E-04 0.022 1.59E-04 2.42E-04 4.05E-04 8.12E-04 0.055 4.96E-05 1.11E-04 2.25E-04 5.27E-04 0.055 6.28E-05 1.30E-04 2.60E-04 6.29E-04 0.068 6.98E-05 1.20E-04 2.37E-04 4.85E-04 0.068 7.56E-05 1.38E-04 2.70E-04 5.59E-04 0.092 4.55E-05 8.06E-05 1.50E-04 3.28E-04 0.092 5.44E-05 9.91E-05 1.91E-04 4.35E-04 0.151 4.35E-05 8.08E-05 1.22E-04 2.24E-04 0.151 5.07E-05 9.46E-05 1.48E-04 2.80E-04

Figure A- 4 The relationship between plastic deformations contributed real contact area and normal load at different roughness 224

On the basis of the real contact area result in Figure 4, the relationship between effective

shear stress and load1/3 is calculated in Figure 5 for different roughness substrates. In nanoscale

contact mechanics, τ is assumed to be constant. However for macroscopic friction, τ is not constant

even in the case of Rq = 0. The numerical derived relationship at Rq = 0 and experimental measured

data at Rq = 0.022 show linear behavior. As we calculated in equation A-3, when roughness is not

1/3 comparable with indentation depth (Rq= 0.022 µm case), the 휏 ∝ 퐿 still works. While turns to

higher roughness, the shear strength increases. Then, an important question is why the interfacial

shear strength does not follow the 휏 ∝ 퐿1/3 relationship when the roughness is not negligible.

Figure A- 5 The shear strength (estimated from friction force in Figure A-2 and real contact area from Table A-2) versus normal load L1/3. The data perfectly smooth surface is calculated based on

Hertzian contact mechanics in Table A-1

225

Several assumptions have been proposed to explain this phenomenon. The shear strength may be pressure-dependent; during sliding process, the interfacial ordering among the molecules may undergo structural deformation. However, the contribution from the adsorbed layer cannot be determined experimentally in current design. The tribochemical reactions were observed for some molecules at tribological interfaces. In the case where frictional heat or wear-exposed dangling bonds are negligible, the chemical reactions occurring at tribological interface is believed to be induced or facilitated by mechanical shear of interfacial molecules. The mechanical energy could transfer from sliding interfaces to reaction coordinate, resulting in higher shear strength. In the case of n-pentanol VPL, no tribo-product is observed from chemical characterization or the yield is lower than the detect resolution.

Although the macroscopic surface roughness remained same before and after friction tests, the plastic deformation of substrates may occur for asperities especially for running in period.

Since the plastic deformation is irreversible process, so it would be path dependent. The friction measured in bi-directional versus uni-directional reciprocating sliding are compared in Figure A-

6. The friction was first measured in both directions during the reciprocating cycles on stainless steel surface (Rq ≈ 0.01 µm); then at the same slide track, the frication was measured in only one direction and lift the ball during the return to the initial position (as shown in Figure A-6 (a)). It is noticed that the friction coefficient became smaller when turns into uni-direction sliding under low load, while the value remain same for high load. When the contact load is comparable with the shear strength of the substrate, the initial plastic deformation of asperities are small. During reciprocating sliding (bi-direction), the asperities deforms to the opposite direction up to the ball movement, corresponding to the large friction force. For uni-direction sliding process, the 226 asperities may toward to the same direction, so that additional deformation is less, which could explain the smaller friction force. While the contact load is too large, the austerities are fully deformed at running period (red box in Figure). The substantial deformation with the change in the sliding direction would play a negligible role on friction force. This means that, in the macroscopic testing, the friction force measured in the reciprocating cycle is the combination of

(i) shear strength of the boundary lubrication layer between asperity contact regions and (ii) plastic deformation of asperity contacts. The second component is dependent on the sliding direction and the degree of deformation in the sliding previous cycle.

Figure A- 6 (a) the schematic of bi-directional sliding and uni-directional sliding (b) & (c) the friction coefficient of two sliding mode under 0.1 N (b) and 0.2 N (c) on Rq = 0.01 µm substrate.

227

Figure A- 7 The frictional behavior under VPL during bii-direction scan, uni-direction scan and uni-direction scan at clean surface under 0.5 N normal load

The plastic deformation of asperities during running in period was observed in Figure A-

7. The normal load used (0.5 N) is high enough to create large deformation at first few cycles. It

can be observed from 3-dimesional map that the friction force quickly decreases with the bi-

direction sliding going on; while during uni-direction sliding, the friction force shows negligible

change because of the fully plastic deformation of asperities on substrate. However, in the case 228 when the ball was moved to a new location, the initial plastic deformation occurred again during running in period.

During the reciprocating sliding, the mechanical properties of substrate may change as well.

To answer the question, nanoindentation measurements were performed inside and outside the slide track. Figure A-8 shows the hardness and reduced modulus change with three different indentation depths: 50 nm, 100 nm and 150 nm. It is noted that both properties show higher values inside slide track for 50 nm indentation tests, while the value are similar for the measurements with larger indentation depth. This could correlate with the Hertzian indentation depth with ball surface (~100 nm). The result adequately reveals that the surface layer of the slide track is hardened due to the interfacial shear. Both effects, the plastic deformation of asperities and mechanical properties change during sliding, may contribute to the shear strength on rough substrates.

Figure A- 8 The hardness (a) and reduced modulus (b) of substrate inside and outside the slide track after 100 cycles scan. The indentation depth of nanoindentaion test are selected as 50 nm,

100 nm and 150 nm.

229

Conclusion

The relationship between shear strength and normal load at macroscale lubrication is discussed through experimental and computational study. The n-pentanol vapor phase lubrication system is employed to minimize the error from wear. The classical Amonton’s Law and contact mechanics indicate the 1/3 power exponential relationship for ideal elastic contact. The result showed that it is valid only when the surface roughness is negligible compared with Hertzian indentation depth.

With the increasing of normal load and roughness, the plastic deformation may dominate the interfacial contact behavior. The plastic deformation may change the shape of asperities to make it different from the theory governing equation. During continuous reciprocating sliding, the shear induced hardening can occur in the sliding track.

230

Reference

(1) Fernández-Bertran, J. F. Mechanochemistry : An Overview. Pure Appl. Chem. 1999, 71 (4), 581–586. https://doi.org/10.1351/pac199971040581. (2) Bowmaker, G. A. Solvent-Assisted Mechanochemistry. Chem Commun 2013, 49 (4), 334–348. https://doi.org/10.1039/c2cc35694e. (3) Black, A. L.; Lenhardt, J. M.; Craig, S. L. From Molecular Mechanochemistry to Stress- Responsive Materials. J. Mater. Chem. 2011, 21, 1655. https://doi.org/10.1039/c0jm02636k. (4) Fischer, T. E. Tribochemistry. Annu. Rev. Mater. Sci. 1988, 18, 303–323. https://doi.org/10.1146/annurev.ms.18.080188.001511. (5) Hsu, S. M.; Zhang, J.; Yin, Z. The Nature and Origin of Tribochemistry. Tribol. Lett. 2002, 13 (2), 131–139. https://doi.org/10.1023/A:1020112901674. (6) Wisniak, J. Matches-The Manufacture of Fire. Indian J. Chem. Technol. 2005, 12 (3), 369–380. (7) Barnette, A. L.; Asay, D. B.; Kim, D.; Guyer, B. D.; Lim, H.; Janik, M. J.; Kim, S. H. Experimental and Density Functional Theory Study of the Tribochemical Wear Behavior of SiO2 in Humid and Alcohol Vapor Environments. Langmuir 2009, 25 (12), 13052– 13061. https://doi.org/10.1021/la901919z. (8) Zhang, J.; Spikes, H. On the Mechanism of ZDDP Antiwear Film Formation. Tribol. Lett. 2016, 63 (2), 1–15. https://doi.org/10.1007/s11249-016-0706-7. (9) Adams, H. L.; Garvey, M. T.; Ramasamy, U. S.; Ye, Z.; Martini, A.; Tysoe, W. T. Shear Induced Mechanochemistry: Pushing Molecules Around. J. Phys. Chem. C 2015, 119 (13), 7115–7123. https://doi.org/10.1021/jp5121146. (10) Gosvami, N. N.; Bares, J. A.; Mangolini, F.; Konicek, A. R.; Yablon, D. G. Mechanisms of Antiwear Tribofilm Growth Revealed in Situ by Single-Asperity Sliding Contacts. Science (80-. ). 2015, 348 (6230), 102–106. (11) Beyer, M. K.; Clausen-Schaumann, H. Mechanochemistry: The Mechanical Activation of Covalent Bonds. Chem. Rev. 2005, 105 (8), 2921–2948. https://doi.org/10.1021/cr030697h. (12) Brantley, J. N.; Wiggins, K. M.; Bielawski, C. W. Unclicking the Click: Mechanically Facilitated 1,3-Dipolar Cycloreversions. Science (80-. ). 2011, 333 (6049), 1606–1609. https://doi.org/10.1126/science.1207934. (13) Zhou, Y.; Qu, J. Ionic Liquids as Lubricant Additives : A Review. ACS Appl. Mater. Interfaces 2017, 9 (4), 3209−3222. https://doi.org/10.1021/acsami.6b12489. (14) Spikes, H. The History and Mechanisms of ZDDP. Tribol. Lett. 2004, 17 (3), 469–489. https://doi.org/10.1023/B:TRIL.0000044495.26882.b5. (15) Binnig, G.; Quate, C. F.; Gerber, C. Atomic Force Microscope. Phys. Rev. Lett. 1986, 56 231

(9), 930–933. https://doi.org/10.1103/PhysRevLett.56.930. (16) Jacobs, T. D. B.; Carpick, R. W. Nanoscale Wear as a Stress-Assisted Chemical Reaction. Nat. Nanotechnol. 2013, 8 (2), 108–112. https://doi.org/10.1038/nnano.2012.255. (17) Duwez, A.-S.; Cuenot, S.; Jérôme, C.; Gabriel, S.; Jérôme, R.; Rapino, S.; Zerbetto, F. Mechanochemistry: Targeted Delivery of Single Molecules. Nat. Nanotechnol. 2006, 1 (2), 122–125. https://doi.org/10.1038/nnano.2006.92. (18) Merkel, R.; Nassoy, P.; Leung, A.; Ritchie, K.; Evans, E. Energy Landscapes of Receptor–Ligand Bonds Explored with Dynamic Force Spectroscopy. Nature 1999, 397 (6714), 50–53. https://doi.org/10.1038/16219. (19) Rief, M.; Gautel, M.; Oesterhelt, F.; Fernandez, J. M.; Gaub, H. E. Reversible Unfolding of Individual Titin Immunoglobulin Domains by AFM. Science (80-. ). 1997, 276 (5315), 1109–1112. https://doi.org/10.1126/science.276.5315.1109. (20) Rico, F.; Gonzalez, L.; Casuso, I.; Puig-Vidal, M.; Scheuring, S. High-Speed Force Spectroscopy Unfolds Titin at the Velocity of Molecular Dynamics Simulations. Science (80-. ). 2013, 342 (6159), 741–743. https://doi.org/10.1126/science.1239764. (21) Barthel, A. J.; Kim, S. H. Lubrication by Physisorbed Molecules in Equilibrium with Vapor at Ambient Condition: Effects of Molecular Structure and Substrate Chemistry. Langmuir 2014, 30 (22), 6489–6478. https://doi.org/10.1021/la501049z. (22) Asay, D. B.; Dugger, M. T.; Kim, S. H. In-Situ Vapor-Phase Lubrication of MEMS. Tribol. Lett. 2008, 29 (1), 67–74. https://doi.org/10.1007/s11249-007-9283-0. (23) He, X.; Barthel, A. J.; Kim, S. H. Tribochemical Synthesis of Nano-Lubricant Films from Adsorbed Molecules at Sliding Solid Interface : Tribo-Polymers from α-Pinene , Pinane , and n-Decane. Surf. Sci. 2016, 648, 352–359. https://doi.org/http://dx.doi.org/10.1016/j.susc.2016.01.005. (24) Archard, J. F. Contact and Rubbing of Flat Surfaces. J. Appl. Phys. 1953, 24 (8), 981–988. https://doi.org/10.1063/1.1721448. (25) Barnette, A. L.; Asay, D. B.; Ohlhausen, J. A.; Dugger, M. T.; Kim, S. H. Tribochemical Polymerization of Adsorbed N-Pentanol on SiO2 during Rubbing: When Does It Occur and Is It Responsible for Effective Vapor Phase Lubrication? Langmuir 2010, 26 (25), 16299–16304. https://doi.org/10.1021/la101481c. (26) Spikes, H.; Tysoe, W. On the Commonality Between Theoretical Models for Fluid and Solid Friction, Wear and Tribochemistry. Tribol. Lett. 2015, 59 (1), 21. https://doi.org/10.1007/s11249-015-0544-z. (27) Eyring, H. Viscosity, Plasticity, and Diffusion as Examples of Absolute Reaction Rates. J. Chem. Phys. 1936, 4 (4), 283–291. https://doi.org/10.1063/1.1749836. (28) Prandtl, L. Ein Gedankenmodell Zur Kinetischen Theorie Der Festen Körper. Zeitschrift für Angew. Math. und Mech. 1928, 8 (2), 85–106. (29) S.N. Zhurkov. Kinetic Concept of the Strength of Solids. Int. J. Fract. Mech. 1965, 1 (11), 232

311–322. (30) Bell, G. I. Models for the Specific Adhesion of Cells to Cells. Science 1978, 200 (4342), 618–627. https://doi.org/10.1126/science.347575. (31) Jacobs, T. D. B.; Gotsmann, B.; Lantz, M. A.; Carpick, R. W. On the Application of Transition State Theory to Atomic-Scale Wear. Tribol. Lett. 2010, 39 (3), 257–271. https://doi.org/10.1007/s11249-010-9635-z. (32) Park, N. S.; Kim, M. W.; Langford, S. C.; Dickinson, J. T. Atomic Layer Wear of Single- Crystal Calcite in Aqueous Solution Scanning Force Microscopy. J. Appl. Phys. 1996, 80 (5), 2680–2686. https://doi.org/10.1063/1.363185. (33) Sheehan, P. E. The Wear Kinetics of NaCl under Dry Nitrogen and at Low Humidities. Chem. Phys. Lett. 2005, 410 (1–3), 151–155. https://doi.org/10.1016/j.cplett.2005.05.060. (34) Jacobs, T. D. B.; Carpick, R. W. Nanoscale Wear as a Stress-Assisted Chemical Reaction. Nat. Nanotechnol. 2013, 8 (2), 108–112. https://doi.org/10.1038/nnano.2012.255. (35) Chen, L.; Wen, J.; Zhang, P.; Yu, B.; Chen, C.; Ma, T.; Lu, X.; Kim, S. H.; Qian, L. Nanomanufacturing of Silicon Surface with a Single Atomic Layer Precision via Mechanochemical Reactions. Nat. Commun. 2018, 9 (1), 1542. https://doi.org/10.1038/s41467-018-03930-5. (36) Chen, L.; Xiao, C.; He, X.; Yu, B.; Kim, S. H.; Qian, L. Friction and Tribochemical Wear Behaviors of Native Oxide Layer on Silicon at Nanoscale. Tribol. Lett. 2017, 65 (4), 1–8. https://doi.org/10.1007/s11249-017-0922-9. (37) He, X.; Kim, S. H. Surface Chemistry Dependence of Mechanochemical Reaction of Adsorbed Molecules - An Experimental Study on Tribopolymerization of α ‑ Pinene on Metal, Metal Oxide, and Carbon Surfaces. Langmuir 2018, 34 (7), 2432−2440. https://doi.org/10.1021/acs.langmuir.7b03763. (38) Jacobs, T. D. B.; Gotsmann, B.; Lantz, M. A.; Carpick, R. W. On the Application of Transition State Theory to Atomic-Scale Wear. Tribol. Lett. 2010, 39 (3), 257–271. https://doi.org/10.1007/s11249-010-9635-z. (39) Felts, J. R.; Oyer, A. J.; Hernández, S. C.; Whitener, K. E.; Robinson, J. T.; Walton, S. G.; Sheehan, P. E. Direct Mechanochemical Cleavage of Functional Groups from Graphene. Nat. Commun. 2015, 6, 6467. https://doi.org/10.1038/ncomms7467. (40) Verners, O.; Van Duin, A. C. T. Comparative Molecular Dynamics Study of Fcc-Ni Nanoplate Stress Corrosion in Water. Surf. Sci. 2015, 633, 94–101. https://doi.org/10.1016/j.susc.2014.10.017. (41) Liu, Y.; Szlufarska, I. Chemical Origins of Frictional Aging. Phys. Rev. Lett. 2012, 109 (18), 1–5. https://doi.org/10.1103/PhysRevLett.109.186102. (42) Li, Q.; Tullis, T. E.; Goldsby, D.; Carpick, R. W. Frictional Ageing from Interfacial Bonding and the Origins of Rate and State Friction. Nature 2011, 480 (7376), 233–236. https://doi.org/10.1038/nature10589. 233

(43) Buehler, M. J.; Van Duin, A. C. T.; Goddard, W. A. Multiparadigm Modeling of Dynamical Crack Propagation in Silicon Using a Reactive Force Field. Phys. Rev. Lett. 2006, 96 (9), 1–4. https://doi.org/10.1103/PhysRevLett.96.095505. (44) U.S. Environmental Protection Agency; USEPA. Short-Term Methods for Estimating the Chronic Toxicity of Effluents and Receiving Waters to Freshwater Organisms Fourth Edition October 2002; 2002. https://doi.org/http://www.dep.state.fl.us/water/wastewater/docs/ctf.pdf. (45) Bridgeman, O. C.; Aldrich, E. W. Vapor Pressure Tables for Water. J. Heat Transfer 1964, 86 (2), 279–286. (46) Hawkins, J. E.; Armstrong, G. T. Physical and Thermodynamic Properties of Terpenes.1 III. The Vapor Pressures of α-Pinene and β-Pinene2. J. Am. Chem. Soc. 1954, 76 (14), 3756–3758. https://doi.org/10.1021/ja01643a051. (47) Wang, Y.; Huang, N.; Xu, B.; Wang, B.; Bai, Z. Measurement and Correlation of the Vapor Pressure of a Series of α-Pinene Derivatives. J. Chem. Eng. Data 2014, 59 (2), 494–498. https://doi.org/10.1021/je400951k. (48) Carruth, G. F.; Kobayashi, R. Vapor Pressure of Normal Paraffins Ethane through N- Decane from Their Triple Points to about 10 Mm Mercury. J. Chem. Eng. Data 1973, 18 (2), 115–126. https://doi.org/10.1021/je60057a009. (49) Ewert, M. Recherches Sur La Theorie Des Solutions Concentrees. XIII. Les Solutions Aqueuses de Composes Organiques. Bull. Soc. Chim. Belg. 1936, 45, 493–515. (50) Kemme, H. R.; Kreps, S. I. Vapor Pressure of Primary N-Alkyl Chlorides and Alcohols. J. Chem. Eng. Data 1969, 14 (1), 98–102. https://doi.org/10.1021/je60040a011. (51) Graubner, V. M.; Jordan, R.; Nuyken, O.; Schnyder, B.; Lippert, T.; Kötz, R.; Wokaun, A. Photochemical Modification of Cross-Linked Poly(Dimethylsiloxane) by Irradiation at 172 Nm. Macromolecules 2004, 37 (16), 5936–5943. https://doi.org/10.1021/ma049747q. (52) Barthel, A. J.; Luo, J.; Hwang, K. S.; Lee, J. Y.; Kim, S. H. Boundary Lubrication Effect of Organic Residue Left on Surface after Evaporation of Organic Cleaning Solvent. Wear 2016, 350–351, 21–26. https://doi.org/10.1016/j.wear.2015.12.010. (53) Archard, J. F.; Rowntree, R. A. The Temperature of Rubbing Bodies; Part 2, the Distribution of Temperatures. Wear 1988, 128 (1), 1–17. https://doi.org/10.1016/0043- 1648(88)90249-9. (54) Pharr, G.; Oliver, W. Measurement of Thin Film Mechanical Properties Using Nanoindentation. Mrs Bull. 1992, 17 (7), 28–33. https://doi.org/10.1557/S0883769400041634. (55) Pharr, G. M.; Oliver, W. C.; Brotzen, F. R. On the Generality of the Relationship among Contact Stiffness, Contact Area, and Elastic Modulus during Indentation. J. Mater. Res. 1992, 7 (03), 613–617. https://doi.org/10.1557/JMR.1992.0613. (56) Ebenstein, D. M.; Wahl, K. J. A Comparison of JKR-Based Methods to Analyze Quasi- Static and Dynamic Indentation Force Curves. J. Colloid Interface Sci. 2006, 298 (2), 234

652–662. https://doi.org/10.1016/j.jcis.2005.12.062. (57) Greenler, R. G. Infrared Study of Adsorbed Molecules on Metal Surfaces by Reflection Techniques. J. Chem. Phys. 1966, 44 (1), 310. https://doi.org/10.1063/1.1726462. (58) Greenler, R. G. Reflection Method for Obtaining the Infrared Spectrum of a Thin Layer on a Metal Surface. J. Chem. Phys. 1969, 50 (5), 1963. https://doi.org/10.1063/1.1671315. (59) Williams, J. A.; Le, H. Tribology and MEMS. J. Phys. D Appl. Phys. 2006, 201, R201– R214. https://doi.org/10.1088/0022-3727/39/12/R01. (60) Maboudian, R.; Ashurst, W. R.; Carraro, C. Tribological Challenges in Micromechanical Systems. Tribol. Lett. 2002, 12 (2), 95–100. https://doi.org/10.1023/A:1014044207344. (61) Onclin, S.; Ravoo, B. J.; Reinhoudt, D. N. Engineering Silicon Oxide Surfaces Using Self- Assembled Monolayers. Angew. Chemie - Int. Ed. 2005, 44 (39), 6282–6304. https://doi.org/10.1002/anie.200500633. (62) Asay, D. B.; Dugger, M. T.; Ohlhausen, J. A.; Kim, S. H. Macro- To Nanoscale Wear Prevention via Molecular Adsorption. Langmuir 2008, 24 (10), 155–159. https://doi.org/10.1021/la702598g. (63) Gellman, A. J. Vapor Lubricant Transport in MEMS Devices. Tribol. Lett. 2004, 17 (3), 455–461. https://doi.org/10.1023/B:TRIL.0000044493.73585.e3. (64) Yoo, S.-S.; Kim, D.-E. Vapor Phase Lubrication Using High Molecular Weight Lubricant for Friction Reduction of Metals. Int. J. Precis. Eng. Manuf. 2014, 15 (5), 867–873. https://doi.org/10.1007/s12541-014-0410-1. (65) Strawhecker, K.; Asay, D. B.; McKinney, J.; Kim, S. H. Reduction of Adhesion and Friction of Silicon Oxide Surface in the Presence of N-Propanol Vapor in the Gas Phase. Tribol. Lett. 2005, 19 (1), 17–21. https://doi.org/10.1007/s11249-004-4261-2. (66) McFadden, C. F.; Gellman, A. J. Metallic Friction: The Effect of Molecular Adsorbates. Surf. Sci. 1998, 409, 171–182. https://doi.org/10.1016/S0039-6028(98)00178-2. (67) Allen, C. M.; Drauglis, E. Boundary Layer Lubrication: Monolayer or Multilayer. Wear 1969, 14 (5), 363–384. https://doi.org/10.1016/0043-1648(69)90017-9. (68) Wiggins, K. M.; Brantley, J. N.; Bielawski, C. W. Polymer Mechanochemistry: Force Enabled Transformations. ACS Macro Lett. 2012, 1, 623–626. https://doi.org/10.1021/mz300167y. (69) Beyer, M. K. The Mechanical Strength of a Covalent Bond Calculated by Density Functional Theory. J. Chem. Phys. 2000, 112 (17), 7307–7312. https://doi.org/10.1063/1.481330. (70) Spikes, H.; Tysoe, W. On the Commonality Between Theoretical Models for Fluid and Solid Friction, Wear and Tribochemistry. Tribol. Lett. 59. https://doi.org/10.1007/s11249- 015-0544-z. (71) Barthel, A. J.; Combs, D. R.; Kim, S. H. Synthesis of Polymeric Lubricating Films Directly at the Sliding Interface via Mechanochemical Reactions of Allyl Alcohols 235

Adsorbed from the Vapor Phase. RSC Adv. 2014, 4 (50), 26081–26086. https://doi.org/10.1039/c4ra02283a. (72) Rinsch, C. L.; Chen, X. L.; Panchalingam, V.; Eberhart, R. C.; Wang, J. H.; Timmons, R. B. Pulsed Radio Frequency Plasma Polymerization of Allyl Alcohol: Controlled Deposition of Surface Hydroxyl Groups. Langmuir 1996, 12 (12), 2995–3002. https://doi.org/10.1021/la950685u. (73) Liu, Z.; Zhang, T.; Zeng, W.; Zhu, H.; An, X. Cationic Polymerization of α-Pinene Using Keggin Silicotungstic Acid as a Homogeneous Catalyst. React. Kinet. Mech. Catal. 2011, 104 (1), 125–137. https://doi.org/10.1007/s11144-011-0333-0. (74) Higashimura, T.; Deng, Y. Cationic Polymerization of α-Pinene with the Binary Catalyst AIC13/SbC13. Makromol, Chem. 1992, 2321, 2311–2321. (75) Sherriff, A. I.; College, N. Mechanism for the Action of Tackieing Resins in Pressure- Sensitive Adhesives. J. Appl. Plymer Sci. 1973, 17, 3423–3438. (76) Asay, D. B.; Kim, S. H. Evolution of the Adsorbed Water Layer Structure on Silicon Oxide at Room Temperature. J. Phys. Chem. B 2005, 109, 16760–16763. https://doi.org/10.1021/jp053042o. (77) Barner, B. J.; Green, M. J.; Saez, E. I.; Corn, R. M. Polarization Modulation Fourier- Transform Infrared Reflectance Measurements of Thin-Films and Monolayers at Metal- Surfaces Utilizing Real-Time Sampling Electronics. Anal. Chem. 1991, 63 (1), 55–60. https://doi.org/Doi 10.1021/Ac00001a010. (78) Barthel, A. J.; Gregory, M. D.; Kim, S. H. Humidity Effects on Friction and Wear between Dissimilar Metals. Tribol. Lett. 2012, 48 (3), 305–313. https://doi.org/10.1007/s11249-012-0026-5. (79) Asay, D. B.; Kim, S. H. Molar Volume and Adsorption Isotherm Dependence of Capillary Forces in Nanoasperity Contacts. Langmuir 2007, 23 (24), 12174–12178. https://doi.org/10.1021/la701954k. (80) Asay, D. B.; Barnette, A. L.; Kim, S. H. Effects of Surface Chemistry on Structure and Thermodynamics of Water Layers at Solid-Vapor Interfaces. J. Phys. Chem. C 2009, 113, 2128–2133. https://doi.org/10.1021/jp806815p. (81) Barthel, A. J.; Al-Azizi, A.; Surdyka, N. D.; Kim, S. H. Effects of Gas or Vapor Adsorption on Adhesion, Friction, and Wear of Solid Interfaces. Langmuir 2014, 30 (11), 2977–2992. https://doi.org/10.1021/la402856j. (82) Sader, J. E. Frequency Response of Cantilever Beams Immersed in Viscous Fluids with Applications to the Atomic Force Microscope. J. Appl. Phys. 2002, 92 (10), 6262–6274. https://doi.org/10.1063/1.1512318. (83) Cappella, B.; Dietler, G. Force-Distance Curves by Atomic Force Microscopy. Surf. Sci. Rep. 1999, 34 (1–3), 1–104. https://doi.org/10.1016/S0167-5729(99)00003-5. (84) Satoh, K.; Nakahara, A.; Mukunoki, K.; Sugiyama, H.; Saito, H.; Kamigaito, M. Sustainable Cycloolefin Polymer from Pine Tree Oil for Optoelectronics Material: Living 236

Cationic Polymerization of β-Pinene and Catalytic Hydrogenation of High-Molecular- Weight Hydrogenated Poly(β-Pinene). Polym. Chem. 2014, 5 (9), 3222. https://doi.org/10.1039/c3py01320k. (85) Hsiao, E.; Marino, M. J.; Kim, S. H. Effects of Gas Adsorption Isotherm and Liquid Contact Angle on Capillary Force for Sphere-on-Flat and Cone-on-Flat Geometries. J. Colloid Interface Sci. 2010, 352 (2), 549–557. https://doi.org/10.1016/j.jcis.2010.09.005. (86) Asay, D. B.; de Boer, M. P.; Kim, S. H. Equilibrium Vapor Adsorption and Capillary Force: Exact Laplace–Young Equation Solution and Circular Approximation Approaches. J. Adhes. Sci. Technol. 2010, 24 (15–16), 2363–2382. https://doi.org/10.1163/016942410X508271. (87) NIST Standard Reference Database 69: NIST Chemistry WebBook. (88) Kinzel, D.; Stolle, A.; Ondruschka, B.; González, L. Quantum Chemical Investigation of the Thermal Rearrangement of Cis- and Trans-Pinane. Phys. Chem. Chem. Phys. 2010, 12 (33), 9884–9892. https://doi.org/10.1039/c001019g. (89) Shea, K. J.; Lease, T. J. The Type 2 Intramolecular Imino Diels-Alder Reaction. Synthesis and Structural Characterization of Bicyclo[n.3.1] Bridgehead Olefins. J. Am. Chem. Soc. 1993, 7 (c), 2248–2260. (90) Binet, M. L.; Commereuc, S.; Verney, V. Thermo-Oxidation of Polyterpenes : Infuence of the Physical State. 2000, 36, 2133–2142. (91) Furlong, O. J.; Miller, B. P.; Tysoe, W. T. Shear-Induced Surface-to-Bulk Transport at Room Temperature in a Sliding Metal-Metal Interface. Tribol. Lett. 2011, 41, 257–261. https://doi.org/10.1007/s11249-010-9711-4. (92) Silvestre, A. J. D.; Gandini, A. Terpenes: Major Sources, Properties and Applications. In Monomers, Polymers and Composites from Renewable Resources; 2008; pp 17–38. (93) Rezende, C. A.; Lee, L.-T.; Galembeck, F. Surface Mechanical Properties of Thin Polymer Films Investigated by AFM in Pulsed Force Mode. Langmuir 2009, 25 (17), 9938–9946. https://doi.org/10.1021/la9010949. (94) Cappella, B.; Silbernagl, D. Nanomechanical Properties of Polymer Thin Films Measured by Force-Distance Curves. Thin Solid Films 2008, 516 (8), 1952–1960. https://doi.org/10.1016/j.tsf.2007.09.042. (95) Cappella, B.; Kaliappan, S. K.; Sturm, H. Using AFM Force - Distance Curves To Study the Glass-to-Rubber Transition of Amorphous Polymers and Their Elastic - Plastic Properties as a Function of Temperature. Macromolecules 2005, 38, 1874–1881. https://doi.org/10.1021/ma040135f. (96) Cai, X.; Bangert, H. Hardness Measurements of Thin Films-Determining the Critical Ratio of Depth to Thickness Using FEM. Thin Solid Films 1995, 264 (1), 59–71. https://doi.org/10.1016/0040-6090(95)06569-5. (97) Opdahl, A.; Kim, S. H.; Koffas, T. S.; Marmo, C.; Somorjai, G. A. Surface Mechanical Properties of PHEMA Contact Lenses: Viscoelastic and Adhesive Property Changes on 237

Exposure to Controlled Humidity. J. Biomed. Mater. Res. A 2003, 67 (1), 350–356. https://doi.org/10.1002/jbm.a.10054. (98) Hsiao, E.; Veres, B. D.; Tudryn, G. J.; Kim, S. H. Identification of Mobile Species in Cationic Polymer Lubricant Layer on Silicon Oxide from AFM and XPS Analyses. Langmuir 2011, 27, 6808–6813. https://doi.org/10.1021/la2002699. (99) Xi, X.; Kim, S. H.; Tittmann, B. Atomic Force Microscopy Based Nanoindentation Study of Onion Abaxial Epidermis Walls in Aqueous Environment Atomic Force Microscopy Based Nanoindentation Study of Onion Abaxial Epidermis Walls in Aqueous Environment. J. Appl. Phys. 2015, 117. https://doi.org/10.1063/1.4906094. (100) Mott, P. H.; Roland, C. M. Limits to Poisson’s Ratio in Isotropic Materials. Phys. Rev. B - Condens. Matter Mater. Phys. 2009, 80 (13), 1–4. https://doi.org/10.1103/PhysRevB.80.132104. (101) Cadogan, D. F.; Howick, C. J. Plasticizers. In Kirk-Othmer Encyclopedia of Chemical Technology; 2010; pp 1–30. (102) Mekonnen, T.; Mussone, P.; Khalil, H.; Bressler, D. Progress in Bio-Based Plastics and Plasticizing Modifications. J. Mater. Chem. A 2013, 1 (43), 13379–13398. https://doi.org/10.1039/c3ta12555f. (103) Wypych, G. Plasticizer Motion and Diffusion. In Handbook of Plasticizers; 2012; pp 165– 185. https://doi.org/10.1016/B978-1-895198-50-8.50009-6. (104) James, S. L.; Adams, C. J.; Bolm, C.; Braga, D.; Collier, P.; Friscic, T.; Grepioni, F.; Harris, K. D. M.; Hyett, G.; Jones, W.; et al. Mechanochemistry: Opportunities for New and Cleaner Synthesis. Chem. Soc. Rev. 2012, 41 (1), 413–447. https://doi.org/10.1039/c1cs15171a. (105) Ito, K.; Martin, J. M.; Minfray, C.; Kato, K. Formation Mechanism of a Low Friction ZDDP Tribofilm on Iron Oxide. Tribol. Trans. 2007, 50 (2), 211–216. https://doi.org/10.1080/10402000701271010. (106) Costa, H. L.; Spikes, H. A. Impact of Ethanol on the Formation of Antiwear Tribofilms from Engine Lubricants. Tribol. Int. 2016, 93, 364–376. https://doi.org/10.1016/j.triboint.2015.09.021. (107) Pereira, G.; Lachenwitzer, A.; Kasrai, M.; Bancroft, G. M.; Norton, P. R.; Abrecht, M.; Gilbert, P. U. P. A.; Regier, T.; Blyth, R. I. R.; Thompson, J. Chemical and Mechanical Analysis of Tribofilms from Fully Formulated Oils Part 1 – Films on 52100 Steel. Tribol. - Mater. Surfaces Interfaces 2007, 1 (1), 48–61. https://doi.org/10.1179/175158407X189293. (108) Vahdat, V.; Ryan, K. E.; Keating, P. L.; Jiang, Y.; Shashishekar, P.; Schall, J. D.; Turner, K. T.; Harrison, J. A.; Carpick, R. W. Atomic-Scale Wear of Amorphous Hydrogenated Carbon during Intermittent Contact : A Combined Study Using Experiment , Simulation , and Theory. ACS Nano 2014, 8 (7), 7027–7040. https://doi.org/10.1021/nn501896e. (109) Alazizi, A.; Barthel, A. J.; Surdyka, N. D.; Luo, J.; Kim, S. H. Vapors in the Ambient - A Complication in Tribological Studies or an Engineering Solution of Tribological 238

Problems? Friction 2015, 3 (2), 85–114. https://doi.org/10.1007/s40544-015-0083-5. (110) Stolle, A.; Ondruschka, B.; Hopf, H. Thermal Rearrangements of Monoterpenes and Monoterpenoids. Helv. Chim. Acta 2009, 92 (9), 1673–1719. https://doi.org/10.1002/hlca.200900041. (111) Stolle, A.; Ondruschka, B. An Effort to Generalize the Thermal Isomerization of 6,6- Dimethylbicyclo[3.1.1]Heptanes and 6,6-Dimethylbicyclo[3.1.1]Heptenes: Comparative Pyrolysis of Pinane, α-Pinene, and β-Pinene. J. Anal. Appl. Pyrolysis 2009, 85 (1–2), 252– 259. https://doi.org/10.1016/j.jaap.2008.12.002. (112) Barnette, A. L.; Ohlhausen, J. A.; Dugger, M. T.; Kim, S. H. Humidity Effects on in Situ Vapor Phase Lubrication with N-Pentanol. Tribol. Lett. 2014, 55 (1), 177–186. https://doi.org/10.1007/s11249-014-0345-9. (113) Alazizi, A.; Draskovics, A.; Ramirez, G.; Erdemir, A.; Kim, S. H. Tribochemistry of Carbon Films in Oxygen and Humid Environments: Oxidative Wear and Galvanic Corrosion. Langmuir 2016, 32 (8), 1996–2004. https://doi.org/10.1021/acs.langmuir.5b04207. (114) Al-Azizi, A. a.; Eryilmaz, O.; Erdemir, A.; Kim, S. H. Nano-Texture for a Wear-Resistant and near-Frictionless Diamond-like Carbon. Carbon N. Y. 2014, 73, 403–412. https://doi.org/10.1016/j.carbon.2014.03.003. (115) Smith, E. H.; Arnell, R. D. A New Approach to the Calculation of Flash Temperatures in Dry, Sliding Contacts. Tribol. Lett. 2013, 52 (3), 407–414. https://doi.org/10.1007/s11249-013-0224-9. (116) Ashby, M. F.; Abulawi, J.; Kong, H. S. Temperature Maps for Frictional Heating in Dry Sliding. Tribol. Trans. 1991, 34 (4), 577–587. https://doi.org/10.1080/10402009108982074. (117) Greenler, R. G. Infrared Study of Adsorbed Molecules on Metal Surfaces by Reflection Techniques. J. Chem. Phys. 1966, 44, 310–315. https://doi.org/10.1063/1.1726462. (118) Sader, J. E.; Chon, J. W. M.; Mulvaney, P. Calibration of Rectangular Atomic Force Microscope Cantilevers. Rev. Sci. Instrum. 1999, 70 (10), 3967–3969. https://doi.org/10.1063/1.1150021. (119) Hosoi, Y. Introduction to Stainless Steels. In Journal of Japan Institute of Light Metals; 1987; Vol. 37, pp 624–635. https://doi.org/10.2464/jilm.37.624. (120) Liu, Z.; Zhang, T.; Zeng, W.; Zhu, H.; An, X. Cationic Polymerization of ??-Pinene Using Keggin Silicotungstic Acid as a Homogeneous Catalyst. React. Kinet. Mech. Catal. 2011, 104 (1), 125–137. https://doi.org/10.1007/s11144-011-0333-0. (121) Homola, A. M.; Israelachvili, J. N.; Mcguiggan, P. M.; Gee, M. L. FUNDENTAL EXPERIMENTAL STUDIES IN TRIBOLOGY: THE TRANSITION FROM “INTERFACIAL” FRICTION OF UNDAMAGED MOLECULARLY SMOOTH SURFACES TO “NORMAL" FRICTION WITH WEAR. Wear 1990, 136 (1), 65–83. (122) So/rensen, M.; Jacobsen, K.; Stoltze, P. Simulations of Atomic-Scale Sliding Friction. 239

Phys. Rev. B 1996, 53 (4), 2101–2113. https://doi.org/10.1103/PhysRevB.53.2101. (123) Carpick, R. W.; Salmeron, M. Scratching the Surface : Fundamental Investigations of Tribology with Atomic Force Microscopy. Chem. Rev. 1997, 97 (4), 1163–1194. https://doi.org/10.1021/cr960068q. (124) Archard, J. F. The Temperature of Rubbing Surfaces. Wear 1959, 2 (6), 438–455. https://doi.org/10.1016/0043-1648(59)90159-0. (125) Al-Azizi, A. A.; Eryilmaz, O.; Erdemir, A.; Kim, S. H. Effects of Nanoscale Surface Texture and Lubricant Molecular Structure on Boundary Lubrication in Liquid. Langmuir 2013, 29 (44), 13419–13426. https://doi.org/10.1021/la402574d. (126) Perkin, S. Ionic Liquids in Confined Geometries. Phys. Chem. Chem. Phys. 2012, 14 (15), 5052–5062. https://doi.org/10.1039/c2cp23814d. (127) Demirel, A.; Granick, S. Glasslike Transition of a Confined Simple Fluid. Phys. Rev. Lett. 1996, 77 (11), 2261–2264. https://doi.org/10.1103/PhysRevLett.77.2261. (128) Israelachvili, J. Solvation Forces and Liquid Structure, as Probed by Direct Force Measurements. Acc. Chem. Res. 1987, 20 (11), 415–421. https://doi.org/10.1021/ar00143a005. (129) Johnson, R. S.; Washburn, C. M.; Staton, A. W.; Moorman, M. W.; Manginell, R. P.; Dugger, M. T.; Dirk, S. M. Thermally-Activated Pentanol Delivery from Precursor Poly(p- Phenylenevinylene)s for MEMS Lubrication. Macromol. Rapid Commun. 2012, 33 (16), 1346–1350. https://doi.org/10.1002/marc.201200325. (130) Kim, S. H.; Asay, D. B.; Dugger, M. T. Nanotribology and MEMs. Nanotoday 2007, 2 (5), 22–29. (131) Cai, X.; Bangert, H. Hardness Measurements of Thin Films-Determining the Critical Ratio of Depth to Thickness Using FEM. Thin Solid Films 1995, 264 (1), 59–71. https://doi.org/10.1016/0040-6090(95)06569-5. (132) Donnet, C.; Belin, M.; Augé, J. C.; Martin, J. M.; Grill, A.; Patel, V. Tribochemistry of Diamond-like Carbon Coatings in Various Environments. Surf. Coatings Technol. 1994, 68–69 (C), 626–631. https://doi.org/10.1016/0257-8972(94)90228-3. (133) Baláž, P.; Achimovičová, M.; Baláž, M.; Billik, P.; Cherkezova-Zheleva, Z.; Criado, J. M.; Delogu, F.; Dutková, E.; Gaffet, E.; Gotor, F. J.; et al. Hallmarks of Mechanochemistry: From Nanoparticles to Technology. Chem. Soc. Rev. 2013, 42 (18), 7571–7637. https://doi.org/10.1039/c3cs35468g. (134) Wang, G.-W. Mechanochemical Organic Synthesis. Chem. Soc. Rev. 2013, 42, 7668– 7700. https://doi.org/10.1039/c3cs35526h. (135) Friščić, T.; Halasz, I.; Beldon, P. J.; Belenguer, A. M.; Adams, F.; Kimber, S. a J.; Honkimäki, V.; Dinnebier, R. E. Real-Time and in Situ Monitoring of Mechanochemical Milling Reactions. Nat. Chem. 2013, 5 (1), 66–73. https://doi.org/10.1038/nchem.1505. (136) Nakayama, K. Triboplasma Generation and Triboluminescence in the Inside and the Front 240

Outside of the Sliding Contact. Tribol. Lett. 2016, 63 (1), 12. https://doi.org/10.1007/s11249-016-0700-0. (137) Wasem, J. V.; LaMarche, B. L.; Langford, S. C.; Dickinson, J. T. Triboelectric Charging of a Perfluoropolyether Lubricant. J. Appl. Phys. 2003, 93 (4), 2202–2207. https://doi.org/10.1063/1.1536011. (138) Ribas-Arino, J.; Marx, D. Covalent Mechanochemistry: Theoretical Concepts and Computational Tools with Applications to Molecular Nanomechanics. Chem. Rev. 2012, 112 (10), 5412–5487. https://doi.org/10.1021/cr200399q. (139) He, X.; Kim, S. H. Mechanochemistry of Physisorbed Molecules at Tribological Interfaces: Molecular Structure Dependence of Tribochemical Polymerization (Submitted). Langmuir 2016. (140) Yeon, J.; van Duin, A. C. T.; Kim, S. H. Effects of Water on Tribochemical Wear of Silicon Oxide Interface: Molecular Dynamics (MD) Study with Reactive Force Field (ReaxFF). Langmuir 2016, 32 (4), 1018–1026. https://doi.org/10.1021/acs.langmuir.5b04062. (141) Wen, J.; Ma, T.; Zhang, W.; Psofogiannakis, G.; van Duin, A. C. T.; Chen, L.; Qian, L.; Hu, Y.; Lu, X. Atomic Insight into Tribochemical Wear Mechanism of Silicon at the Si/SiO2 Interface in Aqueous Environment: Molecular Dynamics Simulations Using ReaxFF Reactive Force Field. Appl. Surf. Sci. 2016, 390, 216–223. https://doi.org/10.1016/j.apsusc.2016.08.082. (142) Islam, M. M.; Bryantsev, V. S.; van Duin, a. C. T. ReaxFF Reactive Force Field Simulations on the Influence of Teflon on Electrolyte Decomposition during Li/SWCNT Anode Discharge in Lithium-Sulfur Batteries. J. Electrochem. Soc. 2014, 161 (8), E3009– E3014. https://doi.org/10.1149/2.005408jes. (143) Goverapet Srinivasan, S.; Van Duin, A. C. T.; Ganesh, P. Development of a ReaxFF Potential for Carbon Condensed Phases and Its Application to the Thermal Fragmentation of a Large Fullerene. J. Phys. Chem. A 2015, 119 (4), 571–580. https://doi.org/10.1021/jp510274e. (144) Kim, S.-Y.; van Duin, A. C. T.; Kubicki, J. D. Molecular Dynamics Simulations of the Interactions between TiO2 Nanoparticles and Water with Na+ and Cl−, Methanol, and Formic Acid Using a Reactive Force Field. J. Mater. Res. 2012, 28 (03), 513–520. https://doi.org/10.1557/jmr.2012.367. (145) He, H.; Qian, L.; Pantano, C. G.; Kim, S. H. Mechanochemical Wear of Soda Lime Silica Glass in Humid Environments. J. Am. Ceram. Soc. 2014, 97 (7), 2061–2068. https://doi.org/10.1111/jace.13014. (146) Chipara, A. C.; Tsafack, T.; Owuor, P.; Yeon, J.; Junkermeier, C. E.; Duin, A. C. T. Van; Bhowmick, S.; Asif, S. A. S.; Brunetto, G.; Douglas, S.; et al. Amphibious Adhesive Using Solid-Liquid Polymer Mixes. Science (80-. ). Under review. (147) Oh, S. J.; Kinney, D. R.; Wang, W.; Rinaldi, P. L. Studies of Allyl Alcohol Radical Polymerization by PFG-HMQC and HMBC NMR at 750 MHz. Macromolecules 2002, 35 241

(7), 2602–2607. https://doi.org/10.1021/ma012053m. (148) Jagur-Grodzinski, J. Nanostructured Polyolefins / Clay Composites : Role of the Molecular Interaction at the Interface. Polym. Adv. Technol. 2006, 17 (April), 395–418. https://doi.org/10.1002/pat. (149) Kao, J.; Katz, T. Conformational Analysis of Allyl Alcohol. An Ab Initio Molecular Orbital-Study. J. Mol. Struct. 1984, 108 (3–4), 229–239. (150) Yue, D.-C.; Ma, T.-B.; Hu, Y.-Z.; Yeon, J.; van Duin, A. C. T.; Wang, H.; Luo, J. Tribochemistry of Phosphoric Acid Sheared between Quartz Surfaces: A Reactive Molecular Dynamics Study. J. Phys. Chem. C 2013, 117 (48), 25604–25614. https://doi.org/10.1021/jp406360u. (151) Chenoweth, K.; Cheung, S.; Duin, A. C. T. Van; Goddard, W. a; Kober, E. M.; Field, F.; Iii, W. a G. Simulations on the Thermal Decomposition of a Poly ( Dimethylsiloxane ) Polymer Using the ReaxFF Reactive Force Field Simulations on the Thermal Decomposition of a Poly ( Dimethylsiloxane ) Polymer Using the ReaxFF Reactive. J. Am. Chem. Soc. 2005, 127 (19), 7192–7202. https://doi.org/10.1021/ja050980t. (152) Chenoweth, K.; van Duin, A. C. T.; Goddard, W. a. ReaxFF Reactive Force Field for Molecular Dynamics Simulations of Hydrocarbon Oxidation. J. Phys. Chem. A 2008, 112 (5), 1040–1053. https://doi.org/10.1021/jp709896w. (153) Spikes, H. Stress-Augmented Thermal Activation : Tribology Feels the Force. Friction 2018, 6 (1), 1–31. https://doi.org/10.1007/s40544-018-0201-2. (154) Do, J. L.; Friščić, T. Mechanochemistry: A Force of Synthesis. ACS Cent. Sci. 2017, 3 (1), 13–19. https://doi.org/10.1021/acscentsci.6b00277. (155) Makarov, D. E. Perspective: Mechanochemistry of Biological and Synthetic Molecules. J. Chem. Phys. 2016, 144 (3). https://doi.org/10.1063/1.4939791. (156) Li, Y. S.; Shyy, J. Y.; Li, S.; Lee, J.; Su, B.; Karin, M.; Chien, S. The Ras-JNK Pathway Is Involved in Shear-Induced Gene Expression. Mol. Cell. Biol. 2015, 16 (11), 5947–5954. https://doi.org/10.1128/mcb.16.11.5947. (157) Wang, H.; Riha, G. M.; Yan, S.; Li, M.; Chai, H.; Yang, H.; Yao, Q.; Chen, C. Shear Stress Induces Endothelial Differentiation from a Murine Embryonic Mesenchymal Progenitor Cell Line. Arterioscler. Thromb. Vasc. Biol. 2005, 25 (9), 1817–1823. https://doi.org/10.1161/01.ATV.0000175840.90510.a8. (158) Sriram, K.; Laughlin, J. G.; Rangamani, P.; Tartakovsky, D. M. Shear-Induced Nitric Oxide Production by Endothelial Cells. Biophys. J. 2016, 111 (1), 208–221. https://doi.org/10.1016/j.bpj.2016.05.034. (159) Chen, Z.; Mercer, J. A. M.; Zhu, X.; Romaniuk, J. A. H.; Cegelski, L.; Martinez, T. J.; Burns, N. Z.; Xia, Y. Mechanochemical Unzipping of Insulating Polyladderene To. Science (80-. ). 2017, 479 (6350), 475–479. https://doi.org/10.1126/aan2797. (160) Bhaskaran, H.; Gotsmann, B.; Sebastian, A.; Drechsler, U.; Lantz, M. A.; Despont, M.; Jaroenapibal, P.; Carpick, R. W.; Chen, Y.; Sridharan, K. Ultralow Nanoscale Wear 242

through Atom-by-Atom Attrition in Silicon-Containing Diamond-like Carbon. Nat. Nanotechnol. 2010, 5 (3), 181–185. https://doi.org/10.1038/nnano.2010.3. (161) Drummond, C.; Israelachvili, J.; Richetti, P. Friction between Two Weakly Adhering Boundary Lubricated Surfaces in Water. Phys. Rev. E. Stat. Nonlin. Soft Matter Phys. 2003, 67 (6 Pt 2), 066110. https://doi.org/10.1103/PhysRevE.67.066110. (162) Briscoe, B. J.; Evans, D. C. B. In The Shear Properties of Langmuir-Blodgett Layer. Proc. R. Soc. London A Math. Phys. Eng. Sci. R. Soc. 1982, 380, 389–407. (163) Kauzmann, W.; Eyring, H. The Viscous Flow of Large Molecules. J. Am. Chem. Soc. 1940, 62 (11), 3113–3125. https://doi.org/10.1021/ja01868a059. (164) Murphy, K. L.; Tysoe, W. T.; Bennett, D. W. A Comparative Investigation of Aryl Isocyanides Chemisorbed to Palladium and Gold: An ATR-IR Spectroscopic Study. Langmuir 2004, 20 (5), 1732–1738. https://doi.org/10.1021/la030293k. (165) Murphy, K.; Azad, S.; Bennett, D. W.; Tysoe, W. T. Adsorption, Decomposition and Isomerization of Methyl Isocyanide and Acetonitrile on Pd(111). Surf. Sci. 2000, 467 (1– 3), 1–9. https://doi.org/10.1016/S0039-6028(00)00807-4. (166) Yeon, J.; He, X.; Martini, A.; Kim, S. H. Mechanochemistry at Solid Surfaces: Polymerization of Adsorbed Molecules by Mechanical Shear at Tribological Interfaces. ACS Appl. Mater. Interfaces 2017, 9 (3), 3142–3148. https://doi.org/10.1021/acsami.6b14159. (167) He, X.; Kim, S. H. Mechanochemistry of Physisorbed Molecules at Tribological Interfaces: Molecular Structure Dependence of Tribochemical Polymerization. Langmuir 2017, 33 (11), 2717–2724. https://doi.org/10.1021/acs.langmuir.6b04028. (168) Hickenboth, C. R.; Moore, J. S.; White, S. R.; Sottos, N. R.; Baudry, J.; Wilson, S. R. Biasing Reaction Pathways with Mechanical Force. Nature 2007, 446 (7134), 423–427. https://doi.org/10.1038/nature05681. (169) Kean, Z. S.; Niu, Z.; Hewage, G. B.; Rheingold, A. L.; Craig, S. L. Stress-Responsive Polymers Containing Cyclobutane Core Mechanophores: Reactivity and Mechanistic Insights. J. Am. Chem. Soc. 2013, 135 (36), 13598–13604. https://doi.org/10.1021/ja4075997. (170) Senftle, T. P.; Hong, S.; Islam, M. M.; Kylasa, S. B.; Zheng, Y.; Shin, Y. K.; Junkermeier, C.; Engel-Herbert, R.; Janik, M. J.; Aktulga, H. M.; et al. The ReaxFF Reactive Force- Field: Development, Applications and Future Directions. npj Comput. Mater. 2016, 2, 15011. https://doi.org/10.1038/npjcompumats.2015.11. (171) Van Duin, A. C. T.; Dasgupta, S.; Lorant, F.; Goddard, W. A. ReaxFF: A Reactive Force Field for Hydrocarbons. J. Phys. Chem. A 2001, 105 (41), 9396–9409. https://doi.org/10.1021/jp004368u. (172) Plimpton, S. Plimpton1995.Pdf. Journal of Computational Physics. 1995, pp 1–19. https://doi.org/10.1006/jcph.1995.1039. (173) Stukowski, A. Visualization and Analysis of Atomistic Simulation Data with OVITO-the 243

Open Visualization Tool. Model. Simul. Mater. Sci. Eng. 2010, 18 (1). https://doi.org/10.1088/0965-0393/18/1/015012. (174) Banerjee, J.; Bojan, V.; Pantano, C. G.; Kim, S. H. Effect of Heat Treatment on the Surface Chemical Structure of Glass: Oxygen Speciation from in Situ XPS Analysis. J. Am. Ceram. Soc. 2018, 101 (2), 644–656. https://doi.org/10.1111/jace.15245. (175) Santner, E.; Czichos, H. Tribology of Polymers. Tribol. - Mater. Surfaces Interfaces 1989, 22 (2), 103–109. https://doi.org/10.1016/S1644-9665(12)60235-0. (176) Smith, D. K. Opal, Cristobalite, and Tridymite: Noncrystallinity versus Crystallinity, Nomenclature of the Silica Minerals and Bibliography. Powder Diffr. 1998, 13 (1), 2–19. https://doi.org/10.1017/S0885715600009696. (177) Stachurski, Z. H. On Structure and Properties of Amorphous Materials; 2011; Vol. 4. https://doi.org/10.3390/ma4091564. (178) Yeon, J. Development of a Reaxff Reactive Force Field for Silicon/Oxygen/Hydrogen/Fluoride Interactions and Applications to Hydroxylation and Friction, The Pennsylvania State University, 2016. (179) Mozzi, R. L.; Warren, B. E. The Structure of Vitreous Silica. J. Appl. Crystallogr. 1969, 2 (4), 164–172. https://doi.org/10.1107/S0021889869006868. (180) Zhuravlev, L. T. The Surface Chemistry of Amorphous Silica. Zhuravlev Model. Colloids Surfaces A Physicochem. Eng. Asp. 2000, 173 (1), 1–38. https://doi.org/https://doi.org/10.1016/S0927-7757(00)00556-2. (181) Takacs, L. The Historical Development of Mechanochemistry. Chem. Soc. Rev. 2013, 42, 7649–7659. https://doi.org/10.1039/c2cs35442j. (182) Gwak, M.; Jung, T.; Yoh, J. J. Friction-Induced Ignition Modeling of Energetic Materials. J. Mech. Sci. Technol. 2009, 23 (7), 1779–1787. https://doi.org/10.1007/s12206-009- 0603-1. (183) Sulem, J.; Famin, V.; Noda, H. Thermal Decomposition of Carbonates in Fault Zones: Slip-Weakening and Temperature-Limiting Effects. J. Geophys. Res. 2009, 114 (B6), 1– 14. https://doi.org/10.1029/2009JB006576. (184) Su, Y.; Yang, Y.; Zhang, H.; Xie, Y.; Wu, Z. Enhanced Photodegradation of Methyl Orange with TiO 2 Nanoparticles Using a Triboelectric Nanogenerator. Nanotechnology 2013, 24, 295401. https://doi.org/10.1088/0957-4484/24/29/295401. (185) Gates, R. S.; Hsu, M.; Klaus, E. E. Tribochemical Mechanism of Alumina With Water. Tribol. Trans. 1989, 32 (3), 357–363. https://doi.org/10.1080/10402008908981900. (186) Chen, L.; He, H.; Wang, X.; Kim, S. H.; Qian, L. Tribology of Silicon / Silicon Oxide in Humid Air : Transition from Severe Chemical Wear to Wearless Behavior at Nanoscale. 2015, 1–10. https://doi.org/10.1021/la504333j. (187) Gotsmann, B.; Lantz, M. A. Atomistic Wear in a Single Asperity Sliding Contact. Phys. Rev. Lett. 2008, 101 (12), 125501. https://doi.org/10.1103/PhysRevLett.101.125501. 244

(188) He, H.; Qian, L.; Pantano, C. G.; Kim, S. H. Effects of Humidity and Counter-Surface on Tribochemical Wear of Soda-Lime-Silica Glass. Wear 2015, 342–343, 100–106. https://doi.org/https://doi.org/10.1016/j.wear.2015.08.016. (189) Konicek, A. R.; Grierson, D. S.; Gilbert, P. U. P. A.; Sawyer, W. G.; Sumant, A. V; Carpick, R. W. Origin of Ultralow Friction and Wear in Ultrananocrystalline Diamond. Phys. Rev. Lett. 2008, 100 (23), 235502. https://doi.org/10.1103/PhysRevLett.100.235502. (190) Qi, Y.; Chen, L.; Jiang, S.; Yu, J.; Yu, B.; Xiao, C.; Qian, L. Investigation of Silicon Wear against Non-Porous and Micro-Porous SiO2 Spheres in Water and in Humid Air. RSC Adv. 2016, 6 (92), 89627–89634. https://doi.org/10.1039/C6RA18152J. (191) Wang, X.; Kim, S. H.; Chen, C.; Chen, L.; He, H.; Qian, L. Humidity Dependence of Tribochemical Wear of Monocrystalline Silicon. ACS Appl. Mater. Interfaces 2015, 7 (27), 14785–14792. https://doi.org/10.1021/acsami.5b03043. (192) Yu, B.; Gao, J.; Chen, L.; Qian, L. Effect of Sliding Velocity on Tribochemical Removal of Gallium Arsenide Surface. Wear 2015, 330–331, 59–63. https://doi.org/https://doi.org/10.1016/j.wear.2014.11.026. (193) Grierson, D. S.; Carpick, R. W. Nanotribology of Carbon-Based Materials. Nano Today 2007, 2 (5), 12–21. https://doi.org/https://doi.org/10.1016/S1748-0132(07)70139-1. (194) Platzman, I.; Brener, R.; Haick, H.; Tannenbaum, R. Oxidation of Polycrystalline Copper Thin Films at Ambient Conditions. J. Phys. Chem. C 2008, 112 (4), 1101–1108. https://doi.org/10.1021/jp076981k. (195) Tardio, S.; Abel, M.-L.; Carr, R. H.; Castle, J. E.; Watts, J. F. Comparative Study of the Native Oxide on 316L Stainless Steel by XPS and ToF-SIMS. J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 2015. https://doi.org/10.1116/1.4927319. (196) Unutulmazsoy, Y.; Merkle, R.; Fischer, D.; Mannhart, J.; Maier, J. The Oxidation Kinetics of Thin Nickel Films between 250 and 500 °C. Phys. Chem. Chem. Phys. 2017, 19 (13), 9045–9052. https://doi.org/10.1039/C7CP00476A. (197) Conrad, H.; Ertl, G.; Küppers, J.; Latta, E. E. Interaction of NO and O2 with Pd(111) Surfaces. I. Surf. Sci. 1977, 65 (1), 235–244. https://doi.org/https://doi.org/10.1016/0039- 6028(77)90304-1. (198) Orent, T. W.; Bader, S. D. LEED and ELS Study of the Initial Oxidation of Pd(100). Surf. Sci. 1982, 115 (2), 323–334. https://doi.org/https://doi.org/10.1016/0039-6028(82)90412- 5. (199) Krozer, A.; Rodahl, M. X-Ray Photoemission Spectroscopy Study of UV/Ozone Oxidation of Au under Ultrahigh Vacuum Conditions. J. Vac. Sci. Technol. A 1997, 15 (3), 1704–1709. https://doi.org/10.1116/1.580924. (200) Yang, M.; Marino, M. J.; Bojan, V. J.; Eryilmaz, O. L.; Erdemir, A.; Kim, S. H. Quantification of Oxygenated Species on a Diamond-like Carbon (DLC) Surface. Appl. Surf. Sci. 2011, 257 (17), 7633–7638. https://doi.org/https://doi.org/10.1016/j.apsusc.2011.03.152. 245

(201) Erdemir, A. U.; Eryilmaz, O. L.; Nilufer, I. B.; Fenske, G. R. Synthesis of Superlow- Friction Carbon Films from Highly Hydrogenated Methane Plasmas. Surf. Coatings Technol. 2000, 133–134, 448–454. https://doi.org/10.1016/S0257-8972(00)00968-3. (202) Dante, R. C.; Kajdas, C. K. A Review and a Fundamental Theory of Silicon Nitride Tribochemistry. Wear 2012, 288, 27–38. https://doi.org/10.1016/j.wear.2012.03.001. (203) Campen, S.; Green, J.; Lamb, G.; Atkinson, D.; Spikes, H. On the Increase in Boundary Friction with Sliding Speed. Tribol. Lett. 2012, 48 (2), 237–248. https://doi.org/10.1007/s11249-012-0019-4. (204) Kato, K. Wear in Relation to Friction — a Review. Wear 2000, 241 (2), 151–157. https://doi.org/https://doi.org/10.1016/S0043-1648(00)00382-3. (205) Nak-Ho, S.; Suh, N. P. Effect of Fiber Orientation on Friction and Wear of Fiber Reinforced Polymeric Composites. Wear 1979, 53 (1), 129–141. https://doi.org/https://doi.org/10.1016/0043-1648(79)90224-2. (206) Unal, H.; Mimaroglu, A.; Kadıoglu, U.; Ekiz, H. Sliding Friction and Wear Behaviour of Polytetrafluoroethylene and Its Composites under Dry Conditions. Mater. Des. 2004, 25 (3), 239–245. https://doi.org/https://doi.org/10.1016/j.matdes.2003.10.009. (207) Persson, B. N. J. Thermal Interface Resistance: Cross-over from Nanoscale to Macroscale. J. Phys. Condens. Matter 2013, 26 (1), 15009. https://doi.org/10.1088/0953- 8984/26/1/015009. (208) Ben-Moshe, T.; Dror, I.; Berkowitz, B. Oxidation of Organic Pollutants in Aqueous Solutions by Nanosized Copper Oxide Catalysts. Appl. Catal. B Environ. 2009, 85 (3), 207–211. https://doi.org/https://doi.org/10.1016/j.apcatb.2008.07.020. (209) Scott, W. J.; Stille, J. K. Palladium-Catalyzed Coupling of Vinyl Triflates with Organostannanes. Synthetic and Mechanistic Studies. J. Am. Chem. Soc. 1986, 108 (11), 3033–3040. https://doi.org/10.1021/ja00271a037. (210) Stille, J. K. The Palladium-Catalyzed Cross-Coupling Reactions of Organotin Reagents with Organic Electrophiles [New Synthetic Methods (58)]. Angew. Chemie Int. Ed. English 1986, 25 (6), 508–524. https://doi.org/10.1002/anie.198605081. (211) Choi, H. C.; Kundaria, S.; Wang, D.; Javey, A.; Wang, Q.; Rolandi, M.; Dai, H. Efficient Formation of Iron Nanoparticle Catalysts on Silicon Oxide by Hydroxylamine for Carbon Nanotube Synthesis and Electronics. Nano Lett. 2003, 3 (2), 157–161. https://doi.org/10.1021/nl025876d. (212) Haruta, M. When Gold Is Not Noble: Catalysis by Nanoparticles. Chem. Rec. 2003, 3 (2), 75–87. https://doi.org/10.1002/tcr.10053. (213) Wan, Y.; Wang, H.; Zhao, Q.; Klingstedt, M.; Terasaki, O.; Zhao, D. Ordered Mesoporous Pd/Silica−Carbon as a Highly Active Heterogeneous Catalyst for Coupling Reaction of Chlorobenzene in Aqueous Media. J. Am. Chem. Soc. 2009, 131 (12), 4541– 4550. https://doi.org/10.1021/ja808481g. (214) Mavrikakis, M.; Hammer, B.; N\orskov, J. K. Effect of Strain on the Reactivity of Metal 246

Surfaces. Phys. Rev. Lett. 1998, 81 (13), 2819–2822. https://doi.org/10.1103/PhysRevLett.81.2819. (215) Srivastava, D.; Brenner, D. W.; Schall, J. D.; Ausman, K. D.; Yu, M.; Ruoff, R. S. Predictions of Enhanced Chemical Reactivity at Regions of Local Conformational Strain on Carbon Nanotubes: Kinky Chemistry. J. Phys. Chem. B 1999, 103 (21), 4330–4337. https://doi.org/10.1021/jp990882s. (216) Kitamura, M.; Mukoyoshi, H.; Fulton, P. M.; Hirose, T. Coal Maturation by Frictional Heat during Rapid Fault Slip. Geophys. Res. Lett. 2012, 39 (August), 1–3. https://doi.org/10.1029/2012GL052316. (217) Hsu, S. M.; Gates, R. S. Effect of Materials on Tribochemical Reactions between Hydrocarbons and Surfaces. J. Phys. D. Appl. Phys. 2006, 39 (15), 3128–3137. https://doi.org/10.1088/0022-3727/39/15/S02. (218) Gosvami, N. N.; Bares, J. A.; Mangolini, F.; Konicek, A. R.; Yablon, D. G.; Carpick, R. W. Mechanisms of Antiwear Tribofilm Growth Revealed in Situ by Single-Asperity Sliding Contacts. Science (80-. ). 2015, 348 (6230), 102–106. https://doi.org/10.1126/science.1258788. (219) Khajeh, A.; He, X.; Yeon, J.; Kim, S. H.; Martini, A. Mechanochemical Association Reaction of Interfacial Molecules Driven by Shear. Langmuir 2018, 34, 5971–5977. https://doi.org/10.1021/acs.langmuir.8b00315. (220) He, X.; Kim, S. H. Surface Chemistry Dependence of Mechanochemical Reaction of Adsorbed Molecules - An Experimental Study on Tribopolymerization of α-Pinene on Metal, Metal Oxide, and Carbon Surfaces. Langmuir 2018, 34 (7), 2432–2440. https://doi.org/10.1021/acs.langmuir.7b03763. (221) Dante, R. C.; Kajdas, C. K. A Review and a Fundamental Theory of Silicon Nitride Tribochemistry. Wear 2012, 288, 27–38. https://doi.org/10.1016/j.wear.2012.03.001. (222) Eryilmaz, O. L.; Erdemir, A. Surface Analytical Investigation of Nearly-Frictionless Carbon Films after Tests in Dry and Humid Nitrogen. Surf. Coat. Technol. 2007, 201, 7401–7407. https://doi.org/10.1016/j.surfcoat.2007.02.005. (223) Marino, M. J.; Hsiao, E.; Bradley, L. C.; Eryilmaz, O. L.; Erdemir, A.; Kim, S. H. Is Ultra-Low Friction Needed to Prevent Wear of Diamond-like Carbon (DLC)? An Alcohol Vapor Lubrication Study for Stainless Steel/DLC Interface. Tribol. Lett. 2011, 42, 285– 291. https://doi.org/10.1007/s11249-011-9771-0. (224) Al-Azizi, A. A.; Eryilmaz, O.; Erdemir, A.; Kim, S. H. Surface Structure of Hydrogenated Diamond-like Carbon: Origin of Run-in Behavior Prior to Superlubricious Interfacial Shear. Langmuir 2015, 31 (5), 1711–1721. https://doi.org/10.1021/la504612c. (225) Marino, M. J.; Hsiao, E.; Chen, Y.; Eryilmaz, O. L.; Erdemir, A.; Kim, S. H. Understanding Run-in Behavior of Diamond-like Carbon Friction and Preventing Diamond-like Carbon Wear in Humid Air. Langmuir 2011, 27, 12702–12708. https://doi.org/10.1021/la202927v. (226) Erdemir, A.; Eryilmaz, O. L.; Kim, S. H. Effect of Tribochemistry on Lubricity of DLC Fi 247

Lms in Hydrogen ☆ . Surf. Coat. Technol. 2014, 257, 241–246. https://doi.org/10.1016/j.surfcoat.2014.08.002. (227) Johnson, K. L.; Kendall, K.; Roberts, A. D. Surface Energy and the Contact of Elastic Solids. Proc. R. Soc. A Math. Phys. Eng. Sci. 1971, 324 (1558), 301–313. https://doi.org/10.1098/rspa.1971.0141. (228) Gellman, A. J.; Ko, J. S. The Current Status of Tribological Surface Science. Tribol. Lett. 2001, 10 (1), 39–44. https://doi.org/10.1023/A:1009070127753. (229) Lancaster, J. K. A Review of the Influence of Environmental Humidity and Water on Friction, Lubrication and Wear. Tribol. Int. 1990, 23 (6), 371–389. https://doi.org/10.1016/0301-679X(90)90053-R. (230) Jones, R.; Pollock, H. M.; Cleaver, J. A. S.; Hodges, C. S. Adhesion Forces between Glass and Silicon Surfaces in Air Studied by AFM: Effects of Relative Humidity, Particle Size, Roughness, and Surface Treatment. Langmuir 2002, 18 (21), 8045–8055. https://doi.org/10.1021/la0259196. (231) Schatzberg, P.; Felsen, I. M. Effects of Water and Oxygen during Rolling Contact Lubrication. Wear 1968, 12 (5), 331–342. https://doi.org/10.1016/0043-1648(68)90536-X. (232) Schatzberg, P.; Felsen, I. M. Influence of Water on Fatigue-Failure Location and Surface Alteration During Rolling-Contact Lubrication. J. Lubr. Technol. 1969, 91 (2), 301. https://doi.org/10.1115/1.3554919. (233) L, G.; D, S. The Acceleration of Pitting Failure by Water in the Lubricant. J. Inst. Pet. 1958, 44, 406–410. (234) De Baets, P.; Kalacska, G.; Strijckmans, K.; Van De Velde, F.; Van Peteghem, A. P. Experimental Study by Means of Thin Layer Activation of the Humidity Influence on the Fretting Wear of Steel Surfaces. Wear 1998, 216 (2), 131–137. https://doi.org/10.1016/S0043-1648(97)00189-0. (235) Stratmann, M.; Streckel, H. On The Atmospheric Corrosion Of Metals Which Are Covered With Thin Electrolyte Layers - I. Verification Of The Experimental Technique. Corros. Sci. 1990, 30 (6/7), 681–696. https://doi.org/https://doi.org/10.1016/0010- 938X(90)90032-Z. (236) Teer, D. G. New Solid Lubricant Coatings. Wear 2002, 251 (1–12), 1068–1074. https://doi.org/10.1016/s0043-1648(01)00764-5. (237) Xiao, X.; Hu, J.; Charych, D. H.; Salmeron, M. Chain Length Dependence of the Frictional Properties of Alkylsilane Molecules Self-Assembled on Mica Studied by Atomic Force Microscopy. Langmuir 2002, 12 (2), 235–237. https://doi.org/10.1021/la950771u. (238) Yeon, J.; van Duin, A. C. T. ReaxFF Molecular Dynamics Simulations of Hydroxylation Kinetics for Amorphous and Nano-Silica Structure, and Its Relations with Atomic Strain Energy. J. Phys. Chem. C 2016, 120, 305–317. https://doi.org/10.1021/acs.jpcc.5b09784. (239) He, X.; Pollock, A.; Kim, S. H. Effect of Gas Environment on Mechanochemical 248

Reaction : A Model Study with Tribo-Polymerization of α-Pinene in Inert , Oxidative , and Reductive Gases. Tribol. Lett. 2019, 67 (1), 1–9. https://doi.org/10.1007/s11249-019- 1136-0. (240) Barnette, A. L.; Kim, S. H. Coadsorption of N-Propanol and Water on SiO 2: Study of Thickness, Composition, and Structure of Binary Adsorbate Layer Using Attenuated Total Reflection Infrared (ATR-IR) and Sum Frequency Generation (SFG) Vibration Spectroscopy. J. Phys. Chem. C 2012, 116 (18), 9909–9916. https://doi.org/10.1021/jp2099416. (241) Day, J. P. R.; Campbell, R. A.; Russell, O. P.; Bain, C. D. Adsorption Kinetics in Binary Surfactant Mixtures Studied with External Reflection FTIR Spectroscopy. J. Phys. Chem. C 2007, 111 (25), 8757–8774. https://doi.org/10.1021/jp067051o. (242) Ngo, D.; Baldelli, S. Adsorption of Dimethyldodecylamine Oxide and Its Mixtures with Triton X-100 at the Hydrophilic Silica/Water Interface Studied Using Total Internal Reflection Raman Spectroscopy. J. Phys. Chem. B 2016, 120 (48), 12346–12357. https://doi.org/10.1021/acs.jpcb.6b08853. (243) Malinowski, E. R. Factor Analysis in Chemistry (3rd Ed.); 2002. https://doi.org/10.1080/00401706.1994.10485411. (244) Madanhire, I.; Mbohwa, C. Mitigating Environmental Impact of Petroleum Lubricants; 2013. (245) Linko, Y. Y.; Lämsä, M.; Huhtala, A.; Rantanen, O. Lipase Biocatalysis in the Production of Esters. J. Am. Oil Chem. Soc. 1995, 72 (11), 1293–1299. https://doi.org/10.1007/BF02546202. (246) Borras, F.; de Rooij, M.; Schipper, D. Rheological and Wetting Properties of Environmentally Acceptable Lubricants (EALs) for Application in Stern Tube Seals. Lubricants 2018, 6 (4), 100. https://doi.org/10.3390/lubricants6040100. (247) Nagendramma, P.; Kaul, S. Development of Ecofriendly/Biodegradable Lubricants: An Overview. Renew. Sustain. Energy Rev. 2012, 16 (1), 764–774. https://doi.org/10.1016/j.rser.2011.09.002. (248) Wan, G. T. Y.; Kenny, P.; Spikes, H. A. Elastohydrodynamic Properties of Hydraulic Fluids. Tribol. Int. 1984, 17 (6), 309–315. (249) Eisentraeger, A.; Schmidt, M.; Murrenhoff, H.; Dott, W.; Hahn, S. Biodegradability Testing of Synthetic Ester Lubricants - Effects of Additives and Usage. Chemosphere 2002, 48 (1), 89–96. https://doi.org/10.1016/S0045-6535(02)00084-X. (250) Rickman, J. C.; Barrett, D. M.; Bruhn, C. M. Nutritional Comparison of Fresh, Frozen and Canned Fruits and Vegetables. Part 1. Vitamins C and B and Phenolic Compounds. J. Sci. Food Agric. 2007, 87 (2), 930–944. https://doi.org/10.1002/jsfa. (251) Erhan, S. Z.; Sharma, B. K.; Perez, J. M. Oxidation and Low Temperature Stability of Vegetable Oil-Based Lubricants. Ind. Crops Prod. 2006, 24 (3), 292–299. https://doi.org/10.1016/j.indcrop.2006.06.008. 249

(252) Holliday, R. L.; King, J. W.; List, G. R. Hydrolysis of Vegetable Oils in Sub- And Supercritical Water. Ind. Eng. Chem. Res. 1997, 36 (3), 932–935. https://doi.org/10.1021/ie960668f. (253) Millett, W. H. Polyalkylene Glycol Synthetic Lubricants. Ind. Eng. Chem. 2005, 42 (12), 2436–2441. https://doi.org/10.1021/ie50492a019. (254) Verma, S.; Kumar, V.; Gupta, K. D. Performance Analysis of Flexible Multirecess Hydrostatic Journal Bearing Operating with Micropolar Lubricant. Lubr. Sci. 2012, 24 (6), 273–290. https://doi.org/10.1002/ls. (255) Betton, C. I. Lubricants and Their Environmental Impact. In Chemistry and Technology of Lubricants; Mortier, R. M., Fox, M. F., Orszulik, S. T., Eds.; Springer Netherlands: Dordrecht, 2010; pp 435–457. https://doi.org/10.1023/b105569_15. (256) Zhou, F.; Liang, Y.; Liu, W. Ionic Liquid Lubricants: Designed Chemistry for Engineering Applications. Chem. Soc. Rev. 2009, 38 (9), 2590–2599. https://doi.org/10.1039/b817899m. (257) Yu, B.; Bansal, D. G.; Qu, J.; Sun, X.; Luo, H.; Dai, S.; Blau, P. J.; Bunting, B. G.; Mordukhovich, G.; Smolenski, D. J. Oil-Miscible and Non-Corrosive Phosphonium- Based Ionic Liquids as Candidate Lubricant Additives. Wear 2012, 289, 58–64. https://doi.org/10.1016/j.wear.2012.04.015. (258) Zhou, Y. Is More Always Better? Tribofilm Evolution and Tribological Behavior Impacted by the Concentration of ZDDP, Ionic Liquid, and ZDDP-Ionic Liquid Combination. ACS Appl. Mater. Interfaces. (259) Gusain, R.; Gupta, P.; Saran, S.; Khatri, O. P. Halogen-Free Bis(Imidazolium)/Bis(Ammonium)-Di[Bis(Salicylato)Borate] Ionic Liquids as Energy- Efficient and Environmentally Friendly Lubricant Additives. ACS Appl. Mater. Interfaces 2014, 6 (17), 15318–15328. https://doi.org/10.1021/am503811t. (260) Pisarova, L.; Gabler, C.; Dörr, N.; Pittenauer, E.; Allmaier, G. Thermo-Oxidative Stability and Corrosion Properties of Ammonium Based Ionic Liquids. Tribol. Int. 2012, 46 (1), 73–83. https://doi.org/10.1016/j.triboint.2011.03.014. (261) Garcia, M. T.; Gathergood, N.; Scammells, P. J. Biodegradable Ionic Liquids Part II. Effect of the Anion and Toxicology. Green Chem. 2005, 7 (1), 9–14. https://doi.org/10.1039/b411922c. (262) Studzińska, S.; Buszewski, B. Study of Toxicity of Imidazolium Ionic Liquids to Watercress (Lepidium Sativum L.). Anal. Bioanal. Chem. 2009, 393 (3), 983–990. https://doi.org/10.1007/s00216-008-2523-9. (263) Elsentriecy, H. H.; Qu, J.; Luo, H.; Meyer, H. M.; Ma, C.; Chi, M. Improving Corrosion Resistance of AZ31B Magnesium Alloy via a Conversion Coating Produced by a Protic Ammonium-Phosphate Ionic Liquid. Thin Solid Films 2014, 568 (1), 44–51. https://doi.org/10.1016/j.tsf.2014.08.010. (264) Zhou, Y.; Dyck, J.; Graham, T. W.; Luo, H.; Leonard, D. N.; Qu, J. Ionic Liquids Composed of Phosphonium Cations and Organophosphate, Carboxylate, and Sulfonate 250

Anions as Lubricant Antiwear Additives. Langmuir 2014, 30 (44), 13301–13311. https://doi.org/10.1021/la5032366. (265) Crobu, M.; Rossi, A.; Mangolini, F.; Spencer, N. D. Chain-Length-Identification Strategy in Zinc Polyphosphate Glasses by Means of XPS and ToF-SIMS. Anal. Bioanal. Chem. 2012, 403 (5), 1415–1432. https://doi.org/10.1007/s00216-012-5836-7. (266) Qu, J.; Luo, H.; Chi, M.; Ma, C.; Blau, P. J.; Dai, S.; Viola, M. B. Comparison of an Oil- Miscible Ionic Liquid and ZDDP as a Lubricant Anti-Wear Additive. Tribol. Int. 2014, 71, 88–97. https://doi.org/10.1016/j.triboint.2013.11.010. (267) Hertz, H. R. Ueber Die Berührung Fester Elastischer Körper. Journal für die reine und angewandte Mathematik (Crelle’s Journal). 1882, p 156. https://doi.org/10.1515/crll.1882.92.156. (268) Bowden, F. P.; Tabor, D. The Friction and Lubrication of Solids; Clarendon Press: Oxford, 1954. (269) Barber, J. R. Multiscale Surfaces and Amontons’ Law of Friction. Tribol. Lett. 2013, 49 (3), 539–543. https://doi.org/10.1007/s11249-012-0094-6. (270) Otsuki, M.; Matsukawa, H. Systematic Breakdown of Amontons’ Law of Friction for an Elastic Object Locally Obeying Amontons’ Law. Sci. Rep. 2013, 3, 3–8. https://doi.org/10.1038/srep01586. (271) Gao, J.; Luedtke, W. D.; Gourdon, D.; Ruths, M.; Israelachvili, J. N.; Landman, U. Frictional Forces and Amontons’ Law: From the Molecular to the Macroscopic Scale. J. Phys. Chem. B 2004, 108 (11), 3410–3425. https://doi.org/10.1021/jp036362l. (272) Greenwood, J. a.; Williamson, J. B. P. Contact of Nominally Flat Surfaces. Proc. R. Soc. A Math. Phys. Eng. Sci. 1966, 295 (1442), 300–319. https://doi.org/10.1098/rspa.1966.0242. (273) Wang, Z. J.; Wang, W. Z.; Hu, Y. Z.; Wang, H. A Numerical Elastic-Plastic Contact Model for Rough Surfaces. Tribol. Trans. 2010, 53 (2), 224–238. https://doi.org/10.1080/10402000903177908.

251

Curriculum Vitae XIN HE Department of Chemical Engineering, Pennsylvania State University, PA 16802, United States Phone: (814)321-5542, E-mail: [email protected]; [email protected] Education Ph.D., Chemical Engineering 2014-2019 The Pennsylvania State University Advisor: Prof. Seong Han Kim Project: Lubrication by Chemical Reaction Products at Sliding Interface (National Science Foundation) B.S., Chemical Engineering 201 0-2014 Tianjin University

Research Experiences Research assistant (Internship 2018-2019) Oak Ridge National Laboratory  Ionic Liquid Additized Environmentally-Friendly Hydraulic Fluids Graduate research assistant (2014-2019) Pennsylvania State University  Tribochemical Synthesis of Nano-Lubricant In-Situ  Fundamental Understanding to Mechanochemistry at Tribological Interfaces

Peer Reviewed Publications Google Scholar: https://scholar.google.com/citations?view_op=list_works&hl=en&user=MSokAmsAAAAJ Select publications (Full list in Google Scholar) 1. X He, D Ngo, SH Kim, Mechanochemical reactions of adsorbates at tribological interfaces: Tribo- polymerizations of allyl alcohol co-adsorbed with water on silicon oxide. Langmuir, 2019, (accepted) 2. X He, Z Liu, L Chen, J Wang, SH Kim, Investigating physical factors contributing to friction of rough surfaces in the absence of wear. 2019, (in preparation) 3. X He, A Pollock, SH Kim; Effect of Gas Environment on Mechanochemical Reaction: A Model Study with Tribo-Polymerization of α-Pinene in Inert, Oxidative, and Reductive Gases. Tribology Letters, 2019, 67:25 4. X He, SH Kim; Surface chemistry dependence of mechanochemical reaction of adsorbed molecule – An experimental study on tribopolymerization of α-pinene on metal, metal oxide, and carbon surfaces. Langmuir, 2018, 34 (7), 2432–2440 5. A Khajeh, X He, J Yeon, SH Kim, A Martini; Mechanochemical association reaction of interfacial molecules driven by shear. Langmuir, 2018, 34(21), 5971-5977 6. X He, SH Kim; Mechanochemistry of Physisorbed Molecules at Tribological Interfaces: Molecular Structure Dependence of Tribochemical Polymerization. Langmuir, 2017, 33(11), 2717-2724 7. J Yeon*, X He* (co-first author), A Martini, SH Kim; Mechanochemistry at Solid Surfaces: Polymerization of Adsorbed Molecules by Mechanical Shear at Tribological Interfaces. ACS Applied Materials & Interfaces, 2017, 9, 3142-3148 8. X He, AJ Barthel, SH Kim; Tribochemical synthesis of nano-lubricant films from adsorbed molecules at sliding solid interface: Tribo-polymers from α-pinene, pinane, and n-decane. Surface Science, 2016, 648, 352-359

Conference Presentations 1. Ionic Liquid Additized Environmentally-Friendly Hydraulic Fluids. 74th STLE Annual Meeting 2019, Nashville, Tennessee 2. Mechanochemical association reaction of interfacial molecules driven by shear. (Poster) 2018 Gordon Research Conference on Tribology 3. Mechanochemistry of Physisorbed Molecules at Tribological Interface. 72nd STLE Annual Meeting 2017, Atlanta, Georgia 4. Tribochemical Reactions of Adsorbed Organic Molecules and Formation of Highly-efficient Polymeric Lubricant Films at Sliding Solid Interfaces. STLE Tribology Frontiers Conference 2016, Chicago, Illinois 5. Tribochemical Synthesis of Nano-Lubricant Films from Adsorbed Molecules at Sliding Solid Interface. 71st STLE Annual Meeting 2016, Las Vegas, Nevada