Molecular Structure and Environment Dependence of Shear-Driven Chemical Reactions: Tribopolymerization of Methylcyclopentane, Cyclohexane and Cyclohexene on Stainless Steel

Tribochemistry, which is another name for mechanochemistry driven by shear, deals with complex and dynamic interfacial processes that can lead to surface wear or formation of beneficial tribofilms. For better mechanistic understanding of these processes, we investigated the reactivity of tribopolymerization of organic molecules with different internal ring strain (methylcyclopentane, cyclohexane, and cyclohexene) on a stainless steel (SS) surface in inert (N2), oxidizing (O2), and reducing (H2) environments at room temperature. On the clean stainless steel surface, precursor molecules were found to physisorb with a broad range of molecular orientations. In inert and reducing environments, the strain-free cyclohexane showed the lowest tribochemical activity among the three molecules tested. Compared to the N2 environment, the tribochemical activity in H2 was suppressed. In the O2 environment, only cyclohexene produced tribofilms and methylcyclopentane while cyclohexane did not. When tribofilms were analyzed with Raman spectroscopy, the spectral features of diamond-like carbon (DLC) or amorphous carbon (a-C) were observed due to photochemical degradation of triboproducts. Based on infrared spectroscopy, tribofilms were found to be organic polymers containing oxygenated groups. Whenever polymeric tribrofilms were produced, wear volume was suppressed by orders of magnitudes but not completely to zero. These results support previously suggested mechanisms which involve surface oxygen as a reactant species in the tribopolymerization process.


Introduction
Tribochemistry is the study of chemical reactions that are facilitated by friction at a sliding interface. Tribochemical reactions could result from frictional heat or the exposure of highly reactive sites and dangling bonds [1][2][3][4][5]. When 49 Page 2 of 14 frictional heating and surface wear are negligible, tribochemical reactions occur via direct mechanical activation of interfacial molecules or atomic sites in the sliding interface [6][7][8]. It has been shown that tribofilms formed from tribochemical reactions of lubricant additives prevent the direct contact of asperities on the mating surfaces, hence minimizing friction and wear [9][10][11][12]. The prototypical filmforming material is zinc dithiophosphate (ZDDP), a widelyused anti-wear additive that decomposes upon frictional shear to form a tribofilm that can protect surfaces and reduce wear [13][14][15]. However, thermal decomposition products of ZDDP in internal combustion engines of automobile can deteriorate the performance of the catalytic converter in the exhaust line [10,16]. Thus, developing alternative anti-wear additives is imperative, and this requires better understanding of tribochemical reaction pathways and the mechanisms that lead to the formation of functional tribofilms.
Shear-induced tribofilm formation can vary depending on applied stress, molecular structure of reactants, environmental gas, and surface chemistry of materials involved [13,[17][18][19][20][21]. Numerous tribochemistry investigations have demonstrated an exponential relationship between applied mechanical stress or force and reaction rate or yield [13,[22][23][24][25][26]. For example, the growth of tribopolymer from α-pinene on dehydroxylated silica in N 2 was found to have an exponential relationship with applied stress [23]. In a study of wear reduction by cyclopropylcarboxylic acid (CPCa), which has a highly strained three-membered ring, the zero-stress activation energy was comparable to the C-C bond dissociation energy within cyclopropane, indicating that opening of the 3-membered ring was involved in tribochemical activation [27,28]. In a study of the tribochemical polymerization of α-pinene, pinane, and n-decane, tribochemical activity was found to be the highest for α-pinene which has highest ring strain due to the presence of the 4-membered ring and a double bond in the 6-membered ring [11]. Reactive molecular dynamics (MD) simulations showed that the double bond in the ring can provide a reaction path that is not accessible in thermal reaction conditions [29,85]. This computational study also suggested that the structural deformation of reacting species by shear may contribute to the mechanical activation of tribochemical reactions [29].
In addition to quantifying the tribochemical reactivity of organic molecules, it is also important to chemically characterize the produced tribofilms. In the literature, Raman spectroscopy has been extensively employed to analyze tribofilms formed from organic precursors [14,28,[30][31][32][33][34][35][36]. Vibrational spectral features reported for such tribofilms often showed two broad bands at ~ 1370 cm −1 and ~ 1580 cm −1 , which are very similar to characteristic bands of amorphous carbon (a-C) or diamond-like-carbon (DLC) materials [14,28,[30][31][32][33][34][35][36]. Since these carbon materials are well known to reduce friction and wear, their formation could explain the good lubrication efficiency with substantial wear reduction in tribochemistry studies of organic precursors [30,36,37]. However, it should be noted that a-C and DLC are usually produced by high-energy physical or chemical vapor deposition processes or pyrolysis at extremely high temperatures [38,39]. Thus, their formation must be tested and confirmed (or disputed) via independent analysis [40]. For a better understanding of tribochemical mechanism, having knowledge of the chemical nature of tribofilms is critical.
In this work, we studied the molecular structure dependence of tribopolymerization reactions of adsorbed organic precursors and demonstrated several methods for characterizing the tribofilm products. Vapor phase lubrication (VPL) has been shown to effectively minimize friction and wear of various inorganic surfaces [16,37,[41][42][43][44][45][46][47]. In VPL, lubricating molecules are delivered to the sliding interface via adsorption from the gas phase. Since there is no liquid solvent, tribochemical reaction products with vapor pressure lower than the delivered precursor molecule will remain on the surface, which allows further characterization of triboproducts [2,4,22,[48][49][50]. Here, methylcyclopentane (CH 3 -c-C 5 H 9 ), cyclohexane (c-C 6 H 12 ), and cyclohexene (c-C 6 H 10 ) were chosen to evaluate the reactivities of cyclic compounds with a 5-membered ring versus a 6-membered ring, as well as the impact of C = C double bond in the 6-membered ring. The ring strain is smaller in the cyclohexane than the cyclopentane [51]. In addition to precursor molecular structures, the effect of surrounding gas-inert (dry N 2 ), reducing (10% H 2 in N 2 ), and oxidizing (O 2 ) environment-was also evaluated. To investigate the interaction between organic vapors and the surface, polarization-modulation reflection absorption infrared spectroscopy (PM-RAIRS) and MD simulations were used to measure the adsorption isotherms and geometry, respectively, of cyclic hydrocarbons on stainless steel surfaces at ambient temperature. The produced tribofilms were analyzed with energy-dispersive X-ray (EDX) mapping for elemental composition, as well as Raman and infrared spectroscopy for chemical structures.

Experimental Methods
A bidirectional ball-on-flat tribometer was used to study tribochemical reactions of methylcyclopentane (97%; Sigma-Aldrich), cyclohexane (≥ 99%; Sigma-Aldrich) and cyclohexene (99%; Thermo Scientific) in dry N 2 (moisture volume concentration ~ 18 ppm), dry O 2 (purity 99.994%) and dry H 2 (10% in N 2 ) gas environments. A schematic of the setup designed to conduct tribotests is shown in Fig.  S1. The bearing-grade ball (diameter = 3 mm) and substrate were AISI 440C stainless steel (SS; McMaster-Carr). The substrate was polished to RMS roughness of roughly 20 nm using various grits of sandpaper and a 0.03 µm alumina colloidal solution. The RMS roughness of the ball was ~ 6 nm [52]. The substrate and ball were both rinsed with ethanol followed by water and UV/Ozone for 30 min prior to tribotesting [47]. The applied normal load was kept at 0.5 N. At this load the maximum Hertzian contact pressure was calculated to be ~ 0.45 GPa. The sliding speed was kept at 3 mm/s, so the average increase of flash temperature was less than 5 °C [11,17]. The length of the sliding track was 2.3 mm. The saturated vapor stream of each precursor was produced by flowing dry N 2 , dry O 2 or dry H 2 (10% in N 2 ) into a flask containing the corresponding liquid, and a target partial pressure relative to its saturation (P/P sat ) was obtained by mixing the vapor-free gas stream with the saturated vapor stream at the corresponding ratio [25,52]. The adsorption isotherms of organic vapor on polished stainless steel (SS) surface were studied with PM-RAIRS. For PM-RAIRS experiment, a ThermoNicolet Nexus 670 spectrometer equipped with an MCT detector and a Hinds Instrument photoelastic modulator was used. The modulator generated the P-and S-polarizations, alternating at 50 kHz of the incident IR beam. The P-polarization has the electric field parallel to the incidence beam plane, while the S-polarization has the field perpendicular to the plane [53,54]. The P + S sum signal gives the vibrational spectrum of mostly gaseous species, while the (P − S)/(P + S) signal, difference normalized with total signal, solely provides the vibrational spectrum of the adsorbed species [53,55]. The spectral resolution was set at 4 cm −1 for the SS experiment. For comparison, another PM-RAIRS measurement was done for a gold (Au) surface with a 1 cm −1 resolution. All spectra shown in this study were normalized with the spectrum collected with a clean surface in dry N 2 or O 2 spectrum as the background.
Tribofilms accumulated on the sliding track were imaged with atomic force microscopy (AFM; Digital instrument, Multimode). The imaging was done in tapping-mode with Bruker TESPA-V2 probes. Tribopolymerization yield was then estimated by calculated the volume above the reference plane, which was the pristine SS substrate surface outside the sliding track [11]. 3D optical profilometry (Zygo NewView 7300) was used to image the topography of the sliding track to estimate the wear volume after solvent-rinsing of triofilms produced and accumulated from 600 reciprocating cycles of sliding. (Details of calculating tribofilm yield and wear volume can be found in the Supporting Information) A Bruker Hyperion 3000 IR Microscope coupled to an IFS 66/s spectrometer was used for IR analysis. A reflective objective lens 15 × (NA = 0.4) was used to analyze an area of 50 × 50 μm 2 . The IR spectra were collected with 4 cm −1 resolution and averaged over 400 scans. For Raman analysis, a Horiba LabRam HR Evolution Vis-NIR system with an objective lens 100 × (NA = 0.9) was used. In Raman data collection, the acquisition time was 1 s and the data were accumulated 10 times. The excitation laser wavelength and power were 532 nm and 2.05 mW, respectively. Elemental analysis of the tribofilm via energy-dispersive x-ray (EDX) mapping was carried out using an Apreo S system (Thermo Fisher Scientific) with an electron accelerating voltage of 15 kV.

Computational Methods
Molecular dynamics (MD) simulation was used to study the adsorption of cyclohexane, cyclohexene, and methylcyclopentane on amorphous Fe 2 O 3 and Fe 3 O 4 surfaces. The surface of a stainless-steel sample is known to be covered with a thin oxide layer that contains both chromium oxides and iron oxides. Analysis of such oxide layer on stainlesssteel using SEM/EDX and XPS in previous literature has shown the iron content to be at least three times higher than chromium content after mechanical polishing [56,57]. Also, ex-situ XPS and TEM analysis of wear tracks formed on bearing steel revealed the presence of FeO, Fe 2 O 3 , Fe 3 O 4 at the surface [58,59]. Therefore, given the limitations of reactive MD simulations (size scale and availability of potentials), the surface of steel is usually approximated as iron oxide (Fe 2 O 3 and Fe 3 O 4 ) [58,[60][61][62]. Two model systems were designed using QuantumATK [63], one for monolayer coverage and the other for sub-monolayer coverage on the iron oxide. The model for monolayer coverage contained 100 molecules of one of the three precursors, while the system for sub-monolayer coverage contained 30 molecules.
The amorphous iron oxide slabs were created using a heat and quench method that involved heating a crystalline Fe 2 O 3 or Fe 3 O 4 slab to 4000 K and then cooling it back to room temperature. Next, a vacuum region was inserted on top of the slabs to create surfaces. The amorphous surfaces were then passivated through reactions with water molecules until the number of reactions between the water and the surface reached steady state. The final dimensions of the amorphous slabs were 5 nm × 5 nm × 1.3 nm (length × width × thickness). The positions of atoms at the bottommost 0.4 nm of each slab were fixed. The organic molecules were placed randomly about 1 nm above the surface.
The simulation process started with energy minimization, followed by dynamic simulations of thermal equilibration at 300 K. Then the system temperature was increased to 450 K at 0.1 K/ps heating rate. This heating rate has been used previously to capture the adsorption process of tricresyl phosphate [64]. A repulsive wall was placed at the top of the simulation box to prevent atoms from leaving the simulation box. The canonical ensemble (NVT) was applied to all unconstrained atoms and charge equilibration was performed throughout the simulation. The initial simulation box dimensions were 5 × 5 × 6 nm 3 with periodic boundary conditions along the surface and fixed boundaries in the surface-normal direction.
The interactions between atoms were modeled using the ReaxFF force field [65,66] with a set of parameters previously developed for Fe/P/O/C/H [67]. This force field has previously been used to simulate phosphorus-containing organic molecules on Fe, Fe 2 O 3 , and Fe 3 O 4, where the adsorption and dissociation energies were reproduced accurately [62,64,67]. The time step for dynamics was 0.25 fs. All simulations were performed using the Large Atomic/ Molecular Massively Parallel Simulation (LAMMPS) software [68]. Postprocessing was carried out using in-house python scripts and OVITO software [69].

Adsorption of Precursors on SS Surface in VPL Condition
First, it is important to know how precursor molecules supplied through the vapor phase are adsorbed on the substrate surface. For that purpose, PM-RAIRS analysis was conducted since this technique could selectively detect the vibrational spectrum of molecular species adsorbed on metal surfaces that are in equilibrium with the same molecules in the gas phase [54]. Figure 1 shows the PM-RAIRS spectra of methylcyclopentane, cyclohexane, and cyclohexene adsorbed on the SS surface. For comparison, the vapor phase spectra of the precursors are also plotted in Fig. 1. In Fig. 1a, the gas phase spectrum of methylcyclopentane shows both symmetric and antisymmetric stretch modes of the CH 3 group at 2878 cm −1 and 2962 cm −1 , respectively, along with broad side bands or shoulders due to rovibrational excitations [70]. In contrast, the PM-RAIRS spectra of the methylcyclopentane adsorbed on the SS surface show only one sharp peak at 2962 cm −1 , which is the asymmetric stretch (ν a ) mode of CH 3 , and the symmetric stretch (ν s ) peak of CH 3 is negligible at 2878 cm −1 . The same trend was observed in the higher-resolution PM-RAIRS spectra of methylcyclopentane adsorbed on the Au surface (Fig. S2). On metallic surfaces with free electrons in the conduction band, only the surface-normal component can be detected in PM-RAIRS and the vibrational mode parallel to the surface is not detected due to the screening by the image charge in the substrate [71]. Thus, the absence of the CH 3 symmetric stretch peak means that the CH 3 side group of methylcyclopentane is parallel to the surface. In the PM-RAIRS spectra of the methylcyclopentane adsorbed on SS, side band tails with small intensities due to rovibrational broadening are still noticeable around the CH 3 asymmetric stretch peak. The CH 2 asymmetric vibration peak is not observed at ~ 2935 cm −1 , because its IR absorption cross-section is small (so its intensity is much smaller than the CH 3 asymmetric peak in the vapor phase spectrum).
In the PM-RAIRS spectra of cyclohexane adsorbed on the SS surface shown in Fig. 1b (and on Au in Fig. S2 as well), the CH 2 asymmetric peak (ν a ) is very strong at 2935 cm −1 . The CH 2 symmetric peak (ν s ) of the adsorbate spectrum is negligible at 2860 cm −1 , while it is quite strong in the gas phase spectrum (again with the broad rovibrational bands in both side of the peak). In the case of cyclohexene (Fig. 1c), the stretch mode of the sp 2 C-H group can be seen at ~ 3040 cm −1 in the vapor phase spectrum, but this peak is very weak in the PM-RAIRS Fig. 1 PM-RAIRS spectra of a methylcyclopentane, b cyclohexane, and c cyclohexene adsorbed on the stainless-steel surface at P/ P sat = 15% (black), 30% (red), 50% (blue) and then after purging out organic vapor with N 2 (magenta). For comparison, the vapor phase spectrum of each precursor (green) is also shown. Above 50% P/P sat , the vapor phase signal was saturated in our in-situ cell, so PM-RAIRS measurement could not be done. The PM-RAIRS spectra of the same precursors measured in O 2 as well as the PM-RAIRS spectra measured with a gold substrate are shown in Fig. S2. (Color figure online) adsorbate spectrum. This means that the C = C doublebond of cyclohexene is preferentially parallel to the surface. Similar to the methylcyclopentane case, the weak rovibrational side bands are evident around the CH 2 asymmetric stretch peak in the PM-RAIRS of cyclohexane and cyclohexene adsorbed on the SS surface.
In Fig. 2, the peak areas of the PM-RAIRS spectra are plotted as a function of relative partial pressure (P/P sat ) of each precursor molecule in the gas phase. The PM-RAIRS peak area is proportional to the surface coverage; thus, this plot is effectively equivalent to the adsorption isotherm [72]. The maximum P/P sat employed in this study was 50%; beyond this, the vapor phase signal of the PM-RAIRS measurement was saturated. In this P/P sat range, the adsorbate signal intensity (peak area) increases nonlinearly with a concave curvature [73][74][75]. This is characteristic of a type-III adsorption isotherm, which is observed when the adsorbate-substrate interaction is weaker than the adsorbate-adsorbate interaction [73,76]. When the precursor vapor was purged out, the PM-RAIRS signal decreased to the initial baseline, indicating that there was no residue or chemisorbed species on the SS substrate surface. In other words, the SS surface was inert, i.e., did not react with the precursor molecules studied here at room temperature. All molecules were physisorbed in the absence of associated shear force, suggesting that the interaction between cyclic hydrocarbon and SS is primarily governed by van der Waals force; thus, any species remaining on the surface after frictional shear must be formed via tribochemical reactions.
To better understand the molecular adsorption geometry, reactive MD simulations were carried out simulating the adsorption process of the three precursor molecules on iron oxide surfaces. No chemisorption of the molecules was  Fig. S3, were similar to the trends observed in the results for simulations with Fe 2 O 3 surface in sub-monolayer coverage, which is shown in Fig. 3.
In all simulations, the orientation angle distributions of the C1-C6 side group of methylcyclopentane and the C1 = C2 double bond of cyclohexene (Fig. 3, Fig. S3, orange bars) are centered around 0° and 180°, which is consistent with the experimental finding from the PM-RAIRS analysis (Fig. 1). A very weak band in the C = C double bond region in the PM-RAIRS spectrum of the 50% P/P sat case was observed (Fig. 1c). Although quantitative correlations between the PM-RAIRS signal intensity and the tilt angle distribution from MD cannot be made, the C = C peak intensity relative to the CH 2 asymmetric mode was quite small in the PM-RAIRS spectra of adsorbates compared to the gas phase spectrum (Fig. 1c). These results suggested that most of the C = C double bonds were laying parallel to the surface with a broad angular distribution. The distributions for the ring plane angles with respect to the surface in Fig. 3 and S3 (black bars) were averaged around 90°, suggesting that the precursor molecules were adsorbed on oxide surfaces mostly vertically. This is somewhat different from the cryogenic temperature condition in which cyclic organic molecules typically adsorb with their molecular planes parallel to the surface [77]. All distributions reported in Figs. 3 and S3 have standard deviations of at least 25°. Such broad distributions are probably consequences of dynamic interactions among adsorbates which are in equilibrium with the gas phase molecules at 300 K. This is because the interactions between adsorbates are stronger than interactions with indicates the data measured upon purging of the vapor after P/ P sat = 50% measurement. Error bars are the standard deviation from three different measurements  . 4 a Friction coefficient measured on SS in dry N 2 without and with 15%, 30%, and 60% P/P sat of cyclohexene. Optical microscope images of the sliding tracks formed in each vapor environment are also shown. b Wear volume per length of the sliding track measured with images taken by 3D profilometry after removing the tribopolymers with acetone. c Depth profiles along the sliding tracks formed in different vapor environments. The root mean square roughness (Sq) of each slide track is also reported the substrate, which is the characteristic of type-III adsorption isotherm [73,76]. The dynamic equilibrium with the gas phase molecules may explain the presence of weak rovibrational features in the PM-RAIRS spectra of adsorbate species (Fig. 1) which are absent in the spectra of adsorbates on metal surfaces at cryogenic temperatures [78][79][80].

VPL Efficiency and Tribopolymerization Yield
When the molecule of interest shows the type-II adsorption isotherm, the VPL study is typically carried out in the P/ P sat condition at which the formation of monolayer coverage is assured [72,81]. However, due to the type-III adsorption isotherm (Fig. 2), it could not be determined the P/P sat forming the monolayer coverage of precursors on the clean SS surface [72]. Therefore, we conducted VPL experiments of cyclohexene in N 2 at 15%, 30% and 60% P/P sat , and the results are shown in Fig. 4. The friction coefficient at 15% P/P sat was 0.35, which was higher than at 30% and 60% P/ P sat (around 0.23), but still lower than in the precursor-free environment. Although the wear volume per length was similar at 15%, 30%, and 60% P/P sat , the sliding track at 15% P/P sat (Sq = 0.18 µm) was rougher than that at higher P/ P sat (Sq = 0.04 µm), implying that the precursor partial pressure (or adsorbate surface coverage) was not high enough to readily form tribopolymers which protected the surface from mechanical wear. When P/P sat was increased from 30% to 60%, the reaction yields increased because there were more reactant molecules adsorbed on the surface (Fig. S4). Furthermore, it was empirically found that when P/P sat was 30% or above, all three precursors provided sufficient lubrication, as determined by the ability to suppress wear of the substrate in dry N 2 environment [72]. Since the focus of this study was to test the effect of precursor molecular structure on tribochemistry, we performed all tribochemical measurements at P/P sat = 30%. Since the adsorbate-adsorbate interaction is favored (as implied from the type-III adsorption isotherm in Fig. 2), it was expected that once tribopolymerization was initiated, leaving a small amount of polymer in the sliding track, the precursor vapor would readily absorb into the polymer, facilitating further reactions in subsequent sliding cycles [11]. Also, it is commonly known that organic vapors with molecular structures similar to a polymer can readily dissolve (or absorb) into the polymer and act as a plasticizer [77]. This effect was observed in our previous study of tribopolymerization of α-pinene [11]. Figure 5a compares the steady-state coefficient of friction of the self-mated 440C SS surface in different vapor conditions. Figure 5b depicts the wear volume per reciprocating sliding cycle (i.e., wear rate) calculated after 600 bidirectional sliding cycles at the applied contact stress of ~ 450 MPa. In the absence of precursor vapor, the coefficient of friction was around 0.7, and the surface exhibited severe wear [53,72]. When significant wear occurred, the sliding track appeared dark in the optical image due to light scattering from the rough surface, and particulate debris were accumulated around the sliding track (Fig. S5). In N 2 and H 2 environments, the presence of all three precursors tested at P/P sat = 30% reduced the friction coefficient to approximately 0.23, which was close to a typical value reported for VPL of organic molecules [29,81]. The friction coefficient gradually reached to a steady-state value within 50 cycles and stayed constant until stopping the measurement at 600 cycles (Fig. S6). The wear volume per length also decreased by about three orders of magnitude compared to the precursor-free environment ( Fig. 5b; details about the wear calculation in Fig. S7). In addition, triboproducts with iridescent color could be seen in optical microscopy images, which is typical of polymeric tribofilms with varying thickness [11]. When O 2 was used as the carrier gas, methylcylcopentane and cyclohexane did not show any lubrication effect; the friction coefficient was comparable to the case of the precursor-free environments (Fig. 5a) and wear volumes were about two orders of magnitude higher than in N 2 and H 2 environments (Fig. 5b). Only cyclohexene could provide sufficient VPL along with formation of lubricious tribofilms that could be seen with optical microscopy.
Since the pristine SS surface was chemically inert in the absence of any frictional shear in the vapor environments tested (Figs. 1 and 2), the tribofilms must be the product of tribopolymerization of physisorbed molecules taking place upon frictional shear by the counter-surface. The tribopolymerization yield was determined by estimating the volume of the tribofilm above the reference plane using tappingmode AFM (details about the yield calculation in Fig. S7). Figure 6a shows AFM images of the endpoints and middle of sliding tracks after tribopolymerization reactions of cyclohexene in N 2 , O 2 , and H 2 environments. The AFM images of tribofilms produced from the methylcyclopentane and cyclohexane precursors in N 2 and H 2 environments are shown in Figure S8. Figure 6b compares the tribopolymerization yields at the same applied load, sliding speed, and sliding distance for different precursor molecules in inert (N 2 ), reducing (10% H 2 in N 2 ), and O 2 gas environments. The amount of tribopolymer material adhered to SS ball was negligible compared to that on the substrate, so only the volume of tribopolymers on the substrate was calculated (Fig. S9). In N 2 , cyclohexane showed the lowest yield among the three precursors tested. In H 2 , the overall yields decreased by ~ 50% as compared to the inert gas environment, but the relative trend among three precursor molecules was the same as the N 2 case. In O 2 , methylcyclopentane and cyclohexane did not produce any measurable amount of tribopolymer, and only cyclohexene produced the tribopolymer. The tribopolymerization yield of cyclohexene in the O 2 environment was about ~ 50% higher than that in the N 2 environment.

Chemical Analysis of Tribopolymerization Products
The ends of the sliding tracks, where tribopolymerization products were accumulated, were analyzed with EDX mapping. Figure 7 compares the elemental compositions determined from EDX mapping of the track ends after VPL testing of the three precursors for 600 reciprocating cycles in inert (N 2 ), reducing (10% H 2 in N 2 ), and oxidizing (O 2 ) environments. As a reference, the chemical composition of the 440C SS measured outside the sliding track is also shown [82]. In the absence of precursor vapors, EDX analysis revealed the uptake of a large amount of oxygen without The error bars are standard deviations from three independent measurements an increase in carbon. This must be due to oxidation of wear debris formed during sliding [17,50]. It is possible that the wear debris was oxidized by a trace amount of oxygen in the test environment during the tribotesting or by air during the sample transfer for ex-situ analysis.
In the cases of methylcyclopentane and cyclohexane in O 2 , the chemical compositions detected with EDX were similar to the severe wear cases observed in the friction test without precursors. This is consistent with the inefficiency of VPL by these precursors in O 2 , as shown in Fig. 5. In the cases where tribofilms were formed (Fig. 6) and friction and wear were greatly reduced (Fig. 5), EDX indicated the accumulation of carbon-containing species at the ends of sliding tracks. The iron and chromium signals in these regions are primarily from the substrate beneath the tribofilms, as the probe depth of EDX is larger than the tribofilm thickness (less than 2 µm; Fig. 6) [83,84]. EDX also found the presence of oxygen in tribofilms formed by oxygen-free precursors in N 2 and H 2 environments. Tribochemical reaction pathways studied by ReaxFF-MD suggested that methylcyclopentane, cyclohexane and cyclohexene react with surface oxygen to form the final triboproducts [85]. Determining whether these oxygen species are due to wear debris or incorporated into the tribopolymers required molecular spectroscopy.
In the tribochemistry literature, Raman analysis has been frequently employed for vibrational spectroscopic analysis of tribofilms formed from organic precursors [9, 27, 30-32, 34-36, 86-89]. Many previous studies reported the formation of a-C or hydrogenated DLC due to frictional activation of organic precursors and hypothesized that these carbon films formed in situ are responsible for lubrication and wear prevention [9, 27, 34-36, 87, 89]. To compare with such previous reports, we performed micro-Raman analysis of the tribofilms (Fig. 8).
In the wear tracks formed in the absence of precursor molecules in the gas phase, the spectral features characteristic of hematite (Fe 2 O 3 ) and magnetite (Fe 3 O 4 ) were observed (Fig. S11) [90,91]. After VPL testing for methylcyclopentane and cyclohexane in O 2 , the sliding track also showed the same oxide bands, although the relative intensities of the peaks at 630 cm −1 and 1300 cm −1 were slightly different. These results confirmed that methylcylcopentane and cyclohexane could neither prevent the wear of SS surface (Fig. 5) nor produce tribopolymers (Fig. 6) in the O 2 environment.
As can be seen in Fig. 8, all tribofilms identified with AFM ( Fig. 6) exhibited Raman spectra with broad D-band and G-band characteristic of a-C or H-DLC. Although this might be viewed as 'consistent with previous literature', we concluded that the vibrational spectral features of a-C or H-DLC were due to photochemical degradation of tribopolymerization products based on the control experiments. Details of the control experiments conducted to test the hypothesis of tribochemical synthesis of a-C or H-DLC are given elsewhere [40].
To obtain molecular spectral features without the photochemical artifact, tribopolymers produced from methylcyclopentane, cyclohexane, and cyclohexene in the N 2 environment were analyzed with IR spectroscopy coupled with microscopy. As shown in Fig. 9, all tribopolymers formed from these three precursors exhibited similar vibrational spectral features: broad bands at ~ 1450 cm −1 which can be attributed to CH 2 bending modes and two bands at  Fig. S10 49 Page 10 of 14 2880 cm −1 and 2920 cm −1 that could be ascribed to symmetric and asymmetric stretches of alkyl groups, respectively. There was no discernable peak at ~ 3030 cm −1 in the IR spectrum of the tribopolymers formed from cyclohexene, indicating that the C = C double-bonds are all converted to single bonds or dissociated during the tribopolymerization process [29,85]. All tribopolymers showed a strong peak at ~ 1700 cm −1 and a broad band spanning from 3100 to 3700 cm −1 , indicating that they have carbonyl (C = O) and hydroxyl (O-H) groups. The precursor molecules do not have any oxygenated functional groups. Thus, they must come from the substrate. In fact, previous tribochemistry studies of -pinene and pinane have also reported that the tribopolymers contained oxygenated functional groups [11,20,29,92,93], and those groups must originate from the involvement of surface oxygen atoms in tribochemical reactions [22,23,29].

Molecular Structure Dependence and Environmental Effect of Tribopolymerization
In inert gas environment (N 2 ), the tribopolymerization yield was lower for cyclohexane as compared to methylcyclopentane and cyclohexene (Fig. 6). It is known that, among cycloalkanes, cyclohexane with the 6-membered ring has the lowest ring strain [94]. Thus, our result is consistent with the hypothesis that the ring strain of a precursor plays a role in tribochemical activation [51]. Reactive MD simulations have shown that, prior to tribochemical reaction, reactants undergo some physical deformation deviating from their equilibrium conformation [22,29], which gives a hint of the impact of molecular structure on reactivity.
In the reducing environment (H 2 ), the relative effect of molecular structure of precursors was similar to the trend found in the N 2 environment, but the overall yield was Fig. 8 Raman spectra collected at the end of the sliding tracks where tribofilms or wear debris were piled up after 600 reciprocating cycles in 30% P/ P sat of precursor vapors in inert (dry N 2 ), reducing (10% H 2 in N 2 ), and oxidizing (dry O 2 ) gas environments. Excitation wavelength = 532 nm; laser power = 2.05 mW; objective lens = 100x (NA = 0.9). The reference spectra of hematite (Fe 2 O 3 ) and magnetite (Fe 3 O 4 ) are shown in Fig. S11 lower than the N 2 case (Fig. 6). Reactive MD simulations suggested that the initial tribochemical activation step of cycloalkane involves dehydrogenation reaction by the surface oxygen [85]. In the presence of H 2 in the surrounding gas, it is likely that such dehydrogenation reactions may be suppressed because, even if surface oxygens are activated by frictional shear, they could react with H 2 impinging from the gas phase, forming stable hydroxyl groups. Our previous study showed that fully hydroxylated silica surfaces are not highly reactive, as compared to dehydroxylated surfaces, toward tribochemical reaction of another organic precursor (α-pinene) [23].
In the oxidative environment (O 2 ), all three molecules physisorb as readily as in N 2 (Fig. S2), but only the tribochemical reactions of cyclohexene can be activated effectively. In the case of cyclohexene, the oxidative chemisorption can occur at shear-activated surface sites through the C = C double bond [29,85]. Once the precursor molecule is chemisorbed, it can be associated with other molecules during sliding [24,28,72,81,85]. The fact that tribochemical reactions of cycloalkane are negligible in O 2 indicates that dehydrogenative chemisorption is suppressed or passivated if the reactive surface sites are reacted with O 2 impinging from the gas phase. The competition of reaction sites between cycloalkane and O 2 is supported by the results shown in Fig. S12. When the partial pressure of cyclohexane was increased from 30 to 45% or 80% in the O 2 environment, the friction coefficient, wear volume, and Sq of sliding track decreased, which was accompanied by formation of tribopolymers. But the tribopolymerization yield was still quite low, compared to the N 2 environment. Even if iron atoms are momentarily exposed at the worn surface, they are likely to be oxidized by O 2 impinging from the gas phase. Overall, the oxygen competed with cycloalkane for reaction sites on SS and, thus, tribochemical reactions of cycloalkane may not occur readily [33,50,[95][96][97].

Conclusion
The effects of molecular structure (methylcyclopentane, cyclohexane, and cyclohexene) and ambient gas (N 2 , O 2 , and H 2 ) on tribochemical reactivity on stainless steel were investigated. Under the experimental conditions of this study, it was found that the more strained molecules (methylcyclopentane and cyclohexene) exhibited higher reaction yield than the nearly-strain-free molecule (cyclohexane) in all environmental conditions. This supports the hypothesis that the internal ring strain of a molecule can facilitate tribochemical reactivity.
Moreover, tribochemical reactivities in reducing (H 2 ) and oxidizing (O 2 ) environments were different from those in an inert (N 2 ) environment. Hydrogen could impede dehydrogenative reactions of the cyclic hydrocarbons by surface O atoms, and thus the reactivities were suppressed in H 2 as compared to N 2 . In an oxidizing environment, dehydrogenative chemisorption of methylcyclohexane and cyclohexane were suppressed, thus preventing the formation of tribopolymers, and resulting in severe wear. In contrast, tribopolymerization of cyclohexene, which could be activated by the oxidative chemisorption of C = C, was facilitated in O 2 . Vibrational spectroscopic analysis of tribofilms showed that they are polymeric species containing C = O and OH groups. The presence of oxygenated groups in the formed tribopolymers supported the previously suggested mechanisms which involved the surface oxygen in the intermediate step.