Understanding Run-In Behavior of Diamond-Like Carbon Friction and

ARTICLE pubs.acs.org/Langmuir

Understanding Run-In Behavior of Diamond-Like Carbon Friction and Preventing Diamond-Like Carbon Wear in Humid Air Matthew J. Marino,† Erik Hsiao,† Yongsheng Chen,‡ Osman L. Eryilmaz,§ Ali Erdemir,§ and Seong H. Kim*,† †

Department of Chemical Engineering and Materials Research Institute, Pennsylvania State University, University Park, Pennsylvania 16802, United States ‡ EMS Energy Institute and John and Willie Leone Family Department of Energy and Mineral Engineering, Pennsylvania State University, University Park, Pennsylvania 16802, United States § Energy Systems Division, Argonne National Laboratory, Argonne, Illinois 60439, United States ABSTRACT: The friction behavior of diamond-like carbon (DLC) is very sensitive to the test environment. For hydrogen-rich DLC tested in dry argon and hydrogen, there was always an induction period, so-called “runin” period, during which the friction coefficient was high and gradually decreased before DLC showed an ultralow friction coefficient (less than 0.01) behavior. Regardless of friction coefficients and hydrogen contents, small amounts of wear were observed in dry argon, hydrogen, oxygen, and humid argon environments. Surprisingly, there were no wear or rubbing scar on DLC surfaces tested in n-pentanol vapor conditions, although the friction coefficient was relatively high among the five test environments. Ex situ X-ray photoelectron and near-edge X-ray absorption fine-structure spectroscopy analyses failed to reveal any differences in chemical composition attributable to the environment dependence of DLC friction and wear. The failure of getting chemical information of oxygenated surface species from the ex situ analysis was found to be due to facile oxidation of the DLC surface upon exposure to air. The removal or wear of this surface oxide layer is responsible for the run-in behavior of DLC. It was discovered that the alcohol vapor can also prevent the oxidized DLC surface from wear in humid air conditions.

I. INTRODUCTION Diamond-like carbon (DLC) represents an interesting class of carbon materials composed of varying degrees of sp2- and sp3-type C C bonds and hydrogen contents.1 3 Unlike other organic films, DLC can provide a hardness as high as bulk ceramic materials such as SiC and Si3N4.4 Materials with high hardness and stiffness have, in general, high wear resistance. DLC films adhere well on substrates containing carbide-forming elements such as silicon and titanium. With dependence on the deposition method and the source gas used, near-frictionless DLC films can be synthesized.5 7 Because of these unusual attributes, DLC films are being used in everyday devices ranging from razor blades to magnetic storage media of personal computers as well as manufacturing and automobile industries.8 The friction coefficients of highly hydrogenated DLC films typically span a range of 0.003 0.02 in vacuum or dry nitrogen environments.9,10 The spread in the friction coefficient values are mainly due to variations in the structure and composition of the films. It was proposed that the superlubricity could be related to extremely low adhesive friction on a molecular scale between hydrogenated carbon surfaces.5,11 An ultralow friction coefficient is observed after an initial induction period (called a run-in period) during which the friction coefficient is high and gradually decreases to a low value as the friction test cycle repeats.12 During the run-in period, the DLC surface wears off and carbon transfer films are often formed on courter-sliding surfaces.13 16 r 2011 American Chemical Society

In general, the run-in is a transient state where the system adjusts before reaching steady state. Some of the parameters governing the run-in responses include contact pressure, surface roughness, and interface layer.17 19 In the literature, the causes for this run-in phase have been attributed to removal of surface contaminants, surface conformity, oxide film formation, material transfer, lubricant reaction, and subsurface microstructure reorientation.17,18,20 22 Following the run-in phase, the system reaches a steady state where the coefficient of friction and wear rate reach and level off.18 Another and even more serious challenge is the environmental sensitivity of the DLC friction, especially to water vapor. In humid air, the superlubricity of hydrogenated DLC films disappears and their friction coefficients are measured to be in the range of 0.05 0.2.9,10,23 29 When these samples are placed and tested in vacuum, the friction coefficient decreases back to a ultralow value after the run-in period.29 The high friction in humid conditions normally results in a high wear rate.30 In contrast, the friction coefficient of the hydrogen-free amorphous carbon film decreases from a high value in dry conditions to the same range as the hydrogenated DLC films in humid environments.28,31 33 Received: July 27, 2011 Revised: August 31, 2011 Published: September 02, 2011 12702

dx.doi.org/10.1021/la202927v | Langmuir 2011, 27, 12702–12708

Langmuir This paper reports the origin of the run-in period of the DLC friction and the prevention of the DLC wear in ambient conditions using alcohol vapor phase lubrication. Two DLC samples were tested: one is highly hydrogenated (NFC-6; atomic hydrogen content = ∼40%) and the other mildly hydrogenated (NFC-10; atomic hydrogen content = ∼25%). The former shows ultralow friction in hydrogen (reducing) and dry argon (inert) environments and high friction in oxygen and humid (oxidizing) environments.34,35 In contrast, the latter gives high and unsteady friction in hydrogen and dry Ar environments and low friction in oxygen and humid environments. In all cases, the ultralow friction behavior is observed only after the initial run-in period during which the friction coefficient is high and mild wear occurs. The surface analysis data of the friction-tested samples suggests that the run-in behavior is associated with wear of the oxidized surface region of DLC which is inevitably formed upon exposure to air.20 It is found that the wear of DLC surface can be prevented below the detection limit by introducing n-pentanol vapor into the ambient although the friction coefficient is not ultralow (∼0.12) in the alcohol vapor environment.36 39 These results indicate that the ultralow friction is not a necessary condition for wear prevention of DLC even though it provides a wear-free sliding once it is achieved.

ARTICLE

II. EXPERIMENTAL DETAILS

H2O < 10 ppm; based on the supplier’s specification) was used as a carrier gas. The gas flow rate was 3 L/min. The Ar gas streams with n-pentanol vapor with a partial pressure 40% relative to its saturation vapor pressure (p/psat; p = 0.9 Torr) and 40% relative humidity (RH; pwater = 9.6 Torr) were produced by the method described in previous literature.42,43 The wear of the substrate and transfer films on the ball were imaged with an optical microscope. X-ray photoelectron spectroscopy (XPS) analysis was carried out using a Kratos Analytical Axis Ultra X-ray system to investigate the chemical composition of the DLC film surface. The photon energy was 1486.7 eV from a monochromatic Al Kα source. The C 1s peak was used to compensate for sample charging by normalizing the main peak to 285 eV. The DLC samples tested in controlled environments were retrieved from the test-rig and immediately sealed in Ar-purged glass bottles until they were mounted on the XPS sample stage. The sample exposure to air was unavoidable, but the procedure minimized the air exposure as little as possible. Oxygen K-edge (543.1 eV) X-ray absorption near edge structure (XANES) spectroscopy was performed at beamline U4B of the National Synchrotron Light Source (NSLS) at Brookhaven National Laboratory. Detailed information about the experimental setup can be found elsewhere.44 XANES spectra were collected in both fluorescence yield and total electron yield modes simultaneously. A spectrum of reference material including V, O, and Cr was also collected in every scan for energy alignment. The scan energy range was from 510 to 590 eV. DLC samples were mounted on a sample holder using double-sided tape. The XANES spectra were processed using Athena.45

Two types of DLC films were used in this study: mildly hydrogenated sample (atomic H percent = ∼25%) and highly hydrogenated sample (atomic H percent = ∼40%). These films will be called NFC-10 and NFC-6, respectively, following the name assigned to these films in previous publications.5,29 The DLC films were prepared by PECVD using a self-bias voltage of 500 V to a substrate and a chamber pressure of ∼10 mTorr. The source gases used for NFC-10 and NFC-6 were C2H2 and a mixture of 25% CH4 and 75% H2, respectively. The deposition procedure has been described previously in refs 2 and 5 and will not be elaborated further here except to note that the deposited films consisted of approximately 1 μm thick DLC layer deposited onto a ∼100 nm thick Si bond layer which was previously deposited onto clean Si wafers and 440C stainless steel balls (diameter = 3 mm). The roughness of the DLC coatings on the SS440C ball was measured to be ∼38 nm for NFC-6 and ∼41 nm for NFC-10 with optical profilometry. On the Si wafer substrates, the roughness of NFC-6 was ∼3 nm and that of NFC-10 was ∼4.5 nm. The DLC films were cleaned with ethanol prior to friction and wear testing. An environment-controlled reciprocating linear-motion pin-on-disk tribometer was used to measure friction coefficient for sliding the ball on the DLC surface at an applied normal load of 1 N and a sliding speed of 2 mm/s. With the use of the extended Hertzian theory for coated samples,40,41 the maximum Hertzian contact pressure and diameter are calculated to be ∼0.83 GPa and ∼49 μm for NFC-6 (elastic modulus = ∼60 GPa) and ∼0.86 GPa and ∼47 μm for NFC-10 (elastic modulus = ∼200 GPa).5 The lateral force exerted to the pin during the slide was measured with a strain gauge sensor. The strain gauge sensor was calibrated by placing known weights on top and measuring the voltage signal corresponding to the given load. The friction coefficient was calculated using Amonton’s law (friction force = friction coefficient  normal load). The friction coefficient was measured for a bidirectional linear motion. The standard deviation of the friction coefficient reported in this article was ∼0.01 once a steady state friction state was reached. The sample was kept inside a continuous flow gas chamber. The ball was in contact with the substrate and slid through an opening in the upper cap. When the ball was engaged on the substrate surface, the opening for the vapor vent was ∼0.8 cm2. An ultrahigh purity Ar gas (O2 < 8 ppm,

III. RESULTS AND DISCUSSION Understanding the origin of the run-in behaviors and the effects of ambient gas composition on DLC friction and wear are important to design optimum operation conditions for DLCbased lubrication and to prevent potential failures of DLC coatings. These questions will be addressed in the following order. First, the friction coefficient of the DLC films in inert (dry Ar), reducing (dry H2), and oxidizing (dry O2, humid Ar, and alcohol-containing Ar) environments will be presented. These results will then be compared with ex situ XPS and XANES surface analysis data. Combined with the oxidation time dependence data, the ex situ analysis results provide an insight into the run-in behavior. Finally, the prevention of DLC wear in a humid ambient will be demonstrated using alcohol vapor-phase lubrication. The effects of different environments on the friction of NFC-6 (highly hydrogenated DLC film) are shown in Figure 1. In the dry Ar and H2 environments, the friction coefficient begins at ∼0.2 and decreases to ∼0.025 within the first ∼50 cycles (run-in period). After the friction coefficient of ∼0.02 is achieved, it remains constant over the duration of the experiment. The ultralow friction response of the highly hydrogenated DLC film might be due to hydrogen termination of surface carbon atoms, thus eliminating σ or covalent bond interactions at the sliding interfaces.5,34,46 After the friction tests, there are clearly noticeable contact slide marks on the ball surface. The wear track on the substrate is so little that it is hard to see but identifiable. The O2 environment has a starting friction coefficient of ∼0.10 but gradually increases to ∼0.16 without the run-in behavior during the test period. There are some debris particles on the ball, but no explicit demarcation of the contact area can be seen. In Ar with 40% RH, the friction is at ∼0.22 at the start but begins to decrease as seen in the dry and hydrogen environments. However, it shows an unsteady behavior rather than keeping a 12703

dx.doi.org/10.1021/la202927v |Langmuir 2011, 27, 12702–12708

Langmuir

ARTICLE

Figure 1. Friction coefficients measured for self-mated NFC-6 (highly hydrogenated DLC) sliding in dry Ar (black), dry H2 (green), dry O2 (purple), 40% RH Ar (red), and Ar with 40% p/psat n-pentanol (blue) environments and optical images of the ball surfaces taken after the friction tests. The scale bar in the image is 20 μm.

Figure 2. Friction coefficients measured for self-mated NFC-10 (mildly hydrogenated DLC) sliding in dry Ar (black), dry H2 (green), dry O2 (purple), 40% RH Ar (red), and Ar with 40% p/psat n-pentanol (blue) environments and optical images of the ball surfaces taken after the friction tests. The scale bar in the image is 20 μm.

steady state value over time. The data shown in Figure 1 (red line) is an example; similar unsteady behaviors were observed unpredictably in repetition tests. The transfer film on the ball looks iridescent, instead of black, which may imply that the transfer film might not be just carbon but some tribochemical reaction products. A big contrast in friction and wear behaviors is observed in 40% p/psat n-pentanol-containing Ar environment. The friction coefficient is observed to be ∼0.12 without any noticeable run-in behavior and, more importantly, there is no sign of wear or transfer film on the ball surface as well as no noticeable wear track on the substrate. The NFC-10 (mildly hydrogenated DLC) film shows apparent differences from the NFC-6 film at the same environments (Figure 2). In the dry Ar and H2 environments, the friction coefficient shows the run-in behavior as in the case of NFC-6; but after the initial run-in period, the friction coefficient varies erratically over a wide range from 0.03 to 0.15. The accumulation of wear debris is clearly visible outside the contact area. The instability of the friction coefficient after the run-in period might be due to higher fractions of carbon-to-carbon interactions in the mildly hydrogenated DLC film. In dry O2, the friction coefficient begins at ∼0.08 and decreases to ∼0.05. After a short run-in period, the friction coefficient remains constant throughout the duration of the test in O2. In Ar with 40% RH, NFC-10 gives a

friction coefficient starting from ∼0.15 and then decreasing to ∼0.03 over the course of ∼100 cycles, which is drastically different from the unsteady friction behavior of NFC-6. Black wear debris particles can be seen on the ball surface tested in the humid Ar gas. What is consistent with NFC-6 is that the 40% p/psat n-pentanol vapor in Ar completely suppresses the wear or transfer film formation although the friction coefficient is relatively high at ∼0.12. Again, no run-in behavior is noticeable in the 40% p/psat n-pentanol vapor environment. Ex situ XPS and XANES analyses were conducted to find any correlations between the friction and wear behaviors and the surface chemical composition of the wear track. The XPS survey scan found only carbon and oxygen present within the probe depth of XPS for both NFC-6 and NFC-10. The atomic content of oxygen is 10 14% and the rest is carbon. It is important to note that the small changes observed in the ex situ XPS atomic concentration analysis did not show any noticeable or meaningful correlation with the friction behavior observed for different test environments as shown in Figures 1 and 2. The difference in the O/C ratio among the samples tested in the same condition was as big as the deviation observed for the different test conditions. The lack of meaningful correlation between the environment dependence of the run-in behavior and friction coefficient and the ex situ XPS analysis is somewhat unexpected, 12704

dx.doi.org/10.1021/la202927v |Langmuir 2011, 27, 12702–12708

Langmuir

ARTICLE

Figure 4. Oxygen K-edge XANES spectra of NFC-6 taken from the control (gray) and scratched regions in dry Ar (black), humid Ar (red), and n-pentanol-containing Ar (blue) environments.

Figure 3. Carbon 1s XPS spectra of (a) NFC-6 and (b) NFC-10 tested in dry Ar (black), dry H2 (green), dry O2 (purple), 40% RH Ar (red), and Ar with 40% p/psat n-pentanol (blue) environments. The inset of each panel shows the deconvoluted relative area of the C O (286.5 eV), CdO (287.8 eV), and O CdO (289.3 eV) components. The rest, not shown, is the C C (285 eV) component.

but it provides an important clue to understanding the DLC surface chemistry which will be discussed later in this section. Figure 3 compares representative C 1s high-resolution XPS spectra taken from the slide contact regions for NFC-6 and NFC-10 after friction tests in dry Ar, dry H2, dry O2, humid Ar, and n-pentanol-containing Ar environments. The friction cycles were run 3 times longer than the initial run-in periods for the dry Ar, dry H2, and humid Ar tests to expose the DLC surface responsible for the characteristic friction behavior observed in each test condition. For the dry O2 and n-pentanol-containing Ar tests, the same cycles as the other tests were used. In all cases, the C 1s peak shapes are very similar to each other, regardless of the friction test environments. Four distinct carbon species can be recognized: C C/C H species at 285 eV, C OH or C O C species at 286.5 eV, CdO species at 287.8 eV, and O CdO species at 289.3 eV. The insets in Figure 3 plot the amount of each oxygenated carbon species deconvoluted from the ex situ XPS data of the contact region tested in different environments. The C O species are most abundant among three types of oxygenated species in all cases. In Figure 3a, a slight increase in oxygenated species can be seen on the NFC-6 samples tested in dry O2 and humid Ar environments; but their amounts observed for the sample tested in n-pentanol-containing Ar environments look very close to the ones tested in dry Ar and dry H2 environments. For example, Figure 3b shows one set of the NFC-10 sample data where the differences in the oxygenated carbon contents between the samples are even smaller. These results imply that the relative functional group distribution detected in ex situ XPS analysis does not vary with the gas

composition of the friction test environment although the gas composition drastically influences the friction and wear behaviors. Likewise, the ex situ analyses of the oxygen species in the friction-tested regions with XANES did not reveal any meaningful dependence on the test environment. The O K-edge XANES spectra of representative NFC-6 samples friction-tested in different environments are shown in Figure 4. An O K-edge spectrum of the region outside the friction-tested area is also shown for comparison. There are two main peaks in the O K-edge XANES spectra: a sharp peak at 530.1 eV and a broad peak centered at 538.3 eV. These features are consistent with previously reported XANES for DLC.47,48 The former can be attributed to the O 1s f π*CdO transition and the latter to O 1s f σ*C O or σ*CdO.49 51 Like the C 1s XPS analysis results, there is no clear dependence of the O K-edge spectra on the friction-test environment. In fact, all O K-edge spectra of the friction-tested regions look very similar to that of the outside region. The only discernible minor peaks are a small shoulder at 535 eV, corresponding to O 1s f σ*O H, for the humid Ar test and a broad shoulder at 543 eV, corresponding to σ*CdO, for the n-pentanol-containing Ar test. However, their intensities are not significantly higher than the noise level. It is very important to correctly interpret the fact that the composition and distribution of surface species on the DLC films determined with ex situ XPS and XANES do not show any strong environmental dependence, while the friction and wear behaviors are very sensitive to the test environment. The reason for the failure to see chemical composition change in ex situ XPS and XANES is not contamination from air. With the use of the inelastic mean free-path of X-ray photoelectrons and typical densities of the carbon materials, the oxidized layer on DLC is estimated to be 2 2.5 nm from the O/C ratio observed from the XPS analysis (Figure 3).52 This value is consistent with the thickness (2.5 3 nm) estimated from the depth profiling measurements with ion sputtering for as-deposited and airexposed DLC coatings.47 If the DLC surface is truly terminated with hydrogenated carbon species (at least, after showing steadystate ultralow friction behavior for the NFC-6 surface), the surface would be hydrophobic and deposition of organic contaminants on such hydrophobic surfaces from air cannot be 2 3 nm thick. Moreover, the physisorbed contaminants will almost completely desorb in ultrahigh vacuum (UHV) chambers before XPS or XANES data collection. It should be noted that the observed oxygen-to-carbon ratio is too high to be attributed 12705

dx.doi.org/10.1021/la202927v |Langmuir 2011, 27, 12702–12708

Langmuir to one monolayer or small organic molecules chemisorbed on carbon surfaces.52 With the physisorption of organic contaminants ruled out, the fact that the oxidized surface layer is more than 2 nm and the composition of this layer is similar in ex situ chemical analyses implies the oxidation of DLC surfaces by air during the sample transfer from the friction test unit to the analytical systems. Unless the DLC samples are deposited, friction-tested, and spectroscopically analyzed in situ in a chamber with UHV and atmospheric gas control capabilities, the samples are inevitably exposed to air between the deposition and the environmentcontrolled friction-test and between the friction-test and the chemical analysis with XPS and XANES. Surface oxidation of solid materials upon exposure to air is a well-known phenomenon for metals and semiconductor materials such as aluminum, stainless steel, and silicon.53 It has been assumed that the DLC surface would be chemically inert in the literature.54 It is true that the DLC surface is relatively inert to chemical reactions compared to metals and semiconductor materials, but that does not mean that the DLC surface is resistant to oxidation in air. The data shown here, along with other data found in the literature,55,56 imply that the air-exposed DLC surface is indeed covered with a oxidized layer. The oxidation of the DLC surface in ambient air must be thermodynamically favorable so that the oxygenated species in the surface layer are in equilibrium with the oxygen and water vapor in air,53 regardless of the initial surface composition before air exposure. This conclusion is fully supported by the lack of any dependence of ex situ XPS and XANES analysis results (Figures 3 and 4) on the chemical composition of the surface terminated after friction tests in controlled environments (Figures 1 and 2). With this insight into the DLC surface oxidation, the run-in period of the DLC samples pre-exposed to air can be attributed to the removal and wear of the surface oxidized layer. The

ARTICLE

oxygenated functional groups in the surface oxide layers can induce dipolar or hydrogen-bonding interactions which are much stronger than simple van der Waals interactions of the hydrogen-terminated carbon surface.35 These chemical interactions could lead to higher friction and wear during the run-in period. In order to check how readily the DLC surface gets oxidized, a control experiment was performed. First, the DLC surface, cleaned with ethanol in air, was completely dried and then rubbed in dry Ar until the steady-state ultralow friction coefficient (