Diamond-like Carbon Tribological Endurance using an Energetic Approach

Abstract

Reputed for their low friction coefficient and wear protection effect, diamond-like carbon (DLC) materials are considered amongst the most important lubricant coatings for tribological applications. In this framework, this investigation aims to elucidate the effect of a few operating parameters, such as applied stress and sliding amplitude on the friction lifetime of DLC coatings. Fretting wear tests are conducted using a 12.7 mm radius counterpart of 52100 steel balls slid against a substrate of the same material coated with a 2 ㎛ thickness DLC. Approximately, 5 to 57 N force is applied, generating a maximum Hertzian contact pressure of 430 to 662 MPa, corresponding to the applied force. The coefficient of friction (CoF) generates three regimes, first a running-in period regime, followed by a steady-state evolution regime, and finally a progressive increase of the CoF reaching the steel CoF value, as an indicator of reaching the substrate. To track the wear scenario, interrupted tests are performed with analysis combining scanning electron microscopy (SEM), energy dispersive X-ray spectroscopy (EDX), 3D profilometer and micro-Raman spectroscopy. The results show two endurance values: one characterizing the coating failure (Nc₁), and the other (Nc₂) indicating the friction failure which is situated where the CoF reaches a threshold value of μ_th = 0.3 in the third regime. The Archard energy density factor is used to determine the two endurance values (Nc₁, Nc₂). Based on this approach, a master curve is established delimitating both the coating and the friction endurances.

1. Introduction

In the last two decades, researches had proved that diamond-like carbon (DLC) coatings are potentially applicable in many industrial and biomedical domains due to their very important mechanical, optical, biomedical and tribological properties, such as their high hardness, optical transparency, chemical inertness and their very low friction coefficient and wear rates [1-3].

DLC coating are for instance used as protective thin films for rolling bearing contact to limit the fretting wear damages. In this paper, a tribological characterization of a DLC layer subjected to fretting wear against 52100 steel ball is investigated, parameters such as the contact pressure (which will be varied from 430 MPa to 622 MPa), the sliding amplitude (from 25 µm to 100 µm) will be considered, with a temperature balancing around 20o C and a relative humidity maintained between 25% and 30%.

Coupling various analytical investigations like SEM, EDX, Micro-Raman and 3D surface profiles, a sequential analysis of fretting damage processes is developed.

Numerous investigations [4,5] underline that the tribological response is controlled by the coating endurance but also by coating failure, transfers and third body behavior which can maintain low friction values long time after reaching the substrate, then a specific attention will be considered to characterize either the coating failure (i.e. substrate reaching) so called Nc1 and the friction failure Nc2 which is presently defined by the situation when the interfacial friction overpasses a µth = 0.3 friction threshold value. The coating failure (substrate reaching) occurs before the friction failure which itself is controlled by the third body modification (Nc2 > Nc1). A wear scenario is introduced where both topographic and chemical evolutions in the plane and ball fretting scars are described.

In addition to this qualitative analysis, a quantitative approach is introduced to predict the Nc2 and Nc1 endurances, a local Archard density parameter () previously developed in [6] is transposed to predict the interface failure (coating and friction failures).

Owing to the simplicity of this approach, single master curves have been established allowing us to predict the interface endurance through a large spectrum of pressure and sliding conditions.

2. Experimental Procedure

2-1. Materials

The specimen used in this study is a plane of 52100 chromium steel, coated with a thin film of DLC named DCY. The contact geometry chosen was a sphere/plane configuration to avoid alignment issues. The sphere is a 52100 chromium steel with a radius of 12.7 mm.

This coating is applied using a PECVD process. It consists of a carbon-Hydrogen based multi-layers coatings with chromium and tungsten. Its structure involves a top dense amorphous carbon layer and nano-scale grains CrN. Nano-indentations were performed to estimate the hardness and the elastic young modulus which were found Hѵ = 3000 and E = 364 GPa respectively. The thickness of the coating is 2 µm. It was deposited on a 52100 plane bearing steel with intermediate carbide interlayer to accommodate the coating substrate interface discontinuities. The coated plane displays a low Ra = 0.5 µm roughness value. Table 1 summarizes the composition and properties of the materials used in this study.

Table 1. Materials data

2-2. Experimental Layout

The fretting wear device used is composed of a mobile part activated by an electromagnetic shaker (Fig. 1(a)), and a fixed part on which is clamped the plane specimen. During the test, the normal force P is kept constant while the tangential force Q and displacement δ are recorded, which enables online plotting of the fretting loop Q–δ. More details about the characteristics of this equipment at room temperature can be found in previous studies [2,4].

Fig. 1. Contact configuration: (a) Schematic of the wear system, (b) fretting loop.

Both tangential force amplitude (Q*) and displacement amplitude (δ*) were recorded during the tests. It is important to mention that a significant part of the tangential displacement is in fact accommodated by the test system and the specimen elastic deformation. In fact, the relative sliding (δg) operating in the interface is better approximated using the residual displacement value (δ₀) measured when the tangential force is equal to zero so that no system or specimen accommodation is occurring. Hence for elastic contact response this allows us to assure δg = δ₀.

In the present investigation, the test frequency was fixed at 25 Hz, the temperature around 20o C and the relative humidity maintained around 28%. Two sets of experiments were performed to investigate the contact pressure and sliding amplitude effects. A first series of test was achieved fixing the sliding amplitude at ±50 µm and varying the normal force from 5 to 57 N in a way to investigate a maximum Hertzian contact pressure range between 430 and 662 MPa. The second series of tests was addressed keeping constant the normal force at 5N (430 MPa) and varying the sliding amplitude from δg = ± 25 µm to δg = ± 100 µm.

3. Results

3-1. Friction Analysis

Fig. 2 compares the evolution of the friction coefficient related to the µ = Q*/P ratio for the reference key loading condition (δg= ±50µm; P = 5N; (pmax = 430 MPa)).

Fig. 2. Friction evolution of the DLC/52100 steel interface compared to the uncoated 52100/52100 interface for the key loading conditions (δg = ± 50 µm; P = 5 N (p_max = 430 MPa)), (a) linear representation, (b) logarithmic representation.

The combined linear (Fig. 2(a)) and logarithmic (Fig. 2(b)) representations allow us to define the following friction sequences.

(I) : During a first period, the friction value is maintained at a rather high 0.4 value from the beginning to 2.10⁵ fretting cycles.

(II) : At the end of the sequence (I), the friction coefficient shows a sharp drop until a second lower 0.2 friction plateau which is maintained up to 10⁶ fretting cycles.

(III) : A smooth linear increase of the friction coefficient is now observed which is characterized by a very discontinuous evolution when the friction response overpasses 0.4.

The friction converges progressively toward the high 52100 steel / 52100 steel interface which for the studied condition stabilizes at 0.7 after 4M fretting cycles (Fig. 2).

This analysis underlines the benefits of the studied DLC layer considering that the 0.3 friction value is reached in sequence III after more than 2.10⁶ cycles. Furthermore, even after 8.10⁶ cycles the mean friction value remains 20% smaller than the corresponding steel/ steel interface.

To better interpret this typical behavior, interrupted tests were performed to illustrate the different friction periods. The following sequences are considered:

1. (N ≈ 1.10⁵ cycles: stabilized response in period (I)

2. (N ≈ 1.9.10⁵ cycles): transition from period (I) to period (II)

3. (N ≈ 1.10⁶ cycles): stabilized sequence in period (II)

4. (N ≈ 2.3.10⁶ cycles): linear increase of the friction coefficient in period (III) µ = µ_th = 0.3. Up to this threshold value, the increase of friction coefficient is shown to be smooth, however, beyond it, many spikes are observed (unstable evolution of the friction coefficient [5]) referring to the abrasive wear behavior (three-body abrasive wear) at the interface, therefore, friction failure is set at this value.

5. (N ≈ 5.10⁶ cycles): period (III) µ = 0.6 > µ_th.

For each interrupted sequence, optical, SEM, EDX, 3D surface profiles and micro-Raman analysis have been performed to formalize the interface damage scenario.

3-2. Wear Analysis

Figure 3 compares the opticals related to plane and ball fretting scar for each sequence.

Fig. 3. Opticals related to plane and ball fretting scars observed just after the test, before the surface cleaning (key loading condition: δg = ± 50 µm; P = 5 N (p_max = 430 MPa).

The analysis shows that up to the sequence 3, although the interface wear is activated (i.e. period II), the quantity of debris ejected from the interface remains very small. By contrast, a large amount of brown and red color debris can be observed outside the interface in sequence 4 and 5.

From these optical observations, it can be intuited that the substrate is reached between sequence 2 and 3. In sequence 3, the bright central metal/metal area is equivalent to the black DLC corona. The inner bright metal area then tends to extend in the following sequence, until representing most of 95% of the interface in sequence 5. It is interesting to note that abrasion wear processes are exclusively observed from the beginning until sequence 4, whereas adhesive wear damages can be observed in sequence 5 in the middle of the fretting scar. A deeper observation of sequence 5 shows that the initial substrate machined surface roughness is still apparent. This evolution is confirmed by the 3D surface profiles given in Figure 4.

Fig. 4. 3D surface profiles and related 2D axial profiles of the coated plane fretting scars achieved at different fretting wear sequences (Fig. 2) after ultrasonic cleaning of specimen in ethanol to remove wear debris.

The 52100 substrate seems to be effectively reached at the sequence 3 (stabilized low friction plateau related to the friction period II, Fig. 2). Then the wear tends to extend laterally but not in depth which can explain the appearance of the initial machined surface. It is only when severe adhesive wear mechanisms are activated in sequence 5 that a significant wear extension in depth can be observed.

This suggests a very high wear resistance of the top 52100 steel substrate interface compared to the DLC layer which is probably due to the hard CrN and carbide interlayers generated at the substrate interface. The chemical evolution of the fretted interface is now developed, by comparing SEM, EDX, chemical mapping and micro-Raman of the plane and the sphere fretting scars (Fig. 5).

Fig. 5. SEM and EDX chemical mapping of the plane fretting scar obtained at different damage sequences (after ultrasonic cleaning of specimen) (1: N=1.10⁵, 2: N=1.9.10⁵, 3: N=1.10⁶, 4: N=2.3.10⁶, 5: N=5.10⁶).

The sequential analysis of the EDX mapping (Fig. 5) shows interesting conclusions: during the initial plateau (I) the interface involves only the DLC amorphous carbon layer, the transition from period (I) to period (II) which exhibits a sharp drop of the friction value, corresponds to the situation when the interface reaches the CrN interface layer. Indeed, the low second friction plateau (II) corresponds to the 52100 steel substrate reaching condition, an annular structure is observed at sequence 3, high concentration of iron and oxygen is observed in the center of the fretting scar which involves various iron oxides. This central zone is surrounded by a Cr coronal which itself is surrounded by a W corona. These concentric structures correspond in fact to the successive wear through CrN and WC interlayers which characterize the studied DLC structure.

The extension of the surface wear damages during the period III, characterized by sequences 4 and 5 reveals a radial extension of the concentric annular structure, the inner iron zone bordered respectively by a rather thick Cr annular and a thinner external W corona. The oxygen mapping is superposed to the area defined by the inner ferrous and chromium rich zone, this suggests that both iron and chromium rich zone are generated in the whole interface. The large implication of iron-chromium oxides interaction is explaining the progressive increase of the friction value during period III.

Micro-Raman investigations were performed to analyze the oxides products generated in both plane and ball fretting scars during the fretting wear process as shown in Fig. 6. This analysis confirms the former EDX investigation. During sequences 1 and 2 the plane fretting interface mainly involves the amorphous layer equivalent to the top DLC coating. The analysis on the ball shows the presence of iron oxides at sequence 1 combined with amorphous carbon and chromium oxides [8-11] and potentially chromium carbide from the 52100 steel ball.

Fig. 6. Qualitative Micro-Raman analysis of plane and sphere fretting scars at different stages of the fretting wear damage (Fig.2) (CrC [7], Cr₂O₃ [8]).

This analysis suggests that the drop of the friction coefficient from period (I) to period (II) should be activated by DLC carbon transfers on the 52100 ball.

The analysis of sequence 3 which corresponds to the lower friction plateau of period (II) confirms the presence of carbon transfers on the ball, but also the presence of various Fe₂O₃ and Fe₃O₄ iron oxides [12] and a significant quantity of amorphous chromium carbide (CrC (886 cm^-1)) which is assumed to be related to the chromium carbide interlayer of the coating.

The analysis of the plane fretting scar at sequence 3 confirms that the fretted interface reaches the substrate interface in the center of the fretting scar. While the external black corona is mainly consisting of DLC carbon layer, the inner bright part reveals a huge quantity of iron and chromium oxides.

The analysis of sequence 4 which corresponds to the progressive increase of the friction coefficient around µ = µ_th = 0.3, confirms that at this stage of the fretting degradation, the interface is mainly consisting of poor lubricant iron oxides. There is no more amorphous carbon on the ball and the plane fretting scar. The carbon element is only found in the ejected debris outside the contact. Chromium oxides and carbide are now reduced and replaced by iron oxides because the fretted interface is now directly in contact with the steel substrate.

After this analysis it could be intuited that the low friction plateau observed in period (II) is controlled by the presence of a lubricious transfer layers on the ball consisting of amorphous carbon and potentially amorphous chromium carbides [7,13]. The fact that the interface reached the 52100 substrate is not a determinant factor, the friction behavior is controlled by the nature of the third body generated in the interface and transfers on the ball. When most of the carbon transfers are eliminated and replaced first by chromium oxides then by iron oxides, the friction failure (µ = µ_th) is activated.

An alternative explanation of the friction transition from period (I) (µ = 0.4) to period (II) (µ = 0.2) could be related to a polishing process of the DLC layer. According to [7], the wear process by polishing the top surface could reduce the friction value. However, the 3D surface analysis of sequence 1 and 3 shows that the plane fretted area is relatively smoother at sequence 1 (period (I)) than at sequence 3 (period (II)). This is confirmed by the computation of the mean surface roughness values (Fig. 7): \({\overline{R}_{a1} = 0.95\mu m} < {\overline{R}_{a2} = 1.63 \mu m}\) whereas the friction value is significantly larger at sequence 1.

Fig. 7. Surface morphologies Comparison of DLC fretting scars at sequence 1(period (I)) and sequence 3(period (II)).

This comparison suggests that the wear polishing process of the top DLC layer is presently not the dominating factor. The friction transition from period (I) to (II) is clearly controlled by the nature of the transfers as previously discussed.

3-3. Prediction of the Interface Endurance

As illustrated by the schema given in Fig. 8, two endurance values can be considered. First we consider the coating endurance Nc1 which is related to the situation where the friction value drops from the 0.4 plateau (i.e. period I) to the following low 0.2 friction plateau (i.e. period II). Indeed, from the given analysis, it appears that the substrate reaching condition corresponds to the activation of the lubricious third body in the interface. The second endurance value Nc₂ is related to the friction failure which is presently defined by the situation where the interface friction overpasses the threshold µ_th = 0.3 value in period (III). Note that this threshold friction value was chosen in accordance with contact mechanisms considerations. Indeed, above µ_th = 0.3, the maximum contact stress is shifted to the surface. This friction value is also related to a contact seizure criterion in tribology. The given analysis underlines a significant difference between Nc₁ and Nc₂ which suggests that the friction responses are highly influenced by the third body behavior.

Fig. 8. Schematization of the friction interface evolution, Nc 1: coating failure (i.e. substrate reaching), Nc₂: friction failure (µ > µ_th = 0.3)

3-4. Quantification of the Friction Endurance

As developed in the introduction, two sets of experiments were performed to rationalize the DLC/52100 friction endurance behavior, one keeping constant the sliding amplitude at δg = ±50 µm and varying the normal force, and a second where the normal force was kept constant at 5N while varying the sliding amplitude. The Nc1 and Nc₂ endurance values are compiled in the Table 2.

Table 2. Experimental calculated parameters (Hertzian contact radius (a_H), endurances (Nc₁ and Nc₂) and Archard's work density parameter (\(\tilde{W}_{H}\) and \(\tilde{W}_{S}\))). 

Different investigations[2,4,6] show that the prediction of the interface endurance must be related to a local description where the maximum wear depth is related to a local friction energy or a local Archard work density description.

Indeed, a global wear description (Archard parameter) which allows us to quantify the wear volume can be used:

 \(V=K \cdot \sum W\)      (1)

Fig. 9. Evolution of friction endurance Nc₂ as a function of the applied sliding amplitude (a) (P = 5N, p_max = 430 MPa), and the contact pressure (b) (δg = ±50µm).

with: V : the wear volume; K: the Archard wear coefficient and ∑W: the Archard work expressed as the product of the normal force (P) and the total sliding distance (S).

\(\Sigma W=P . S=P .4 . \delta g . N\)       (2)

However, assuming that the friction failure is related to a local degradation in the interface, we need to compute the maximum Archard work dissipated in the interface. The Hertzian pressure profile is expressed by:

\(\begin{aligned} &\text { If } R \leq 1 \\ &\begin{aligned} p(X, Y)=p(R) &=p_{\max }\left(1-R^{2}\right)^{1 / 2} \\ &=p_{\max }\left(1-X^{2}-Y^{2}\right)^{1 / 2} \end{aligned} \\ &\begin{aligned} \text { If } R>1 & \\ p(X, Y) &=p(R)=0 \end{aligned} \\ \\ &\begin{aligned} \text { with: } R=\sqrt{X^{2}+Y^{2}} ; X=\frac{x}{a_{H}} ; Y=\frac{y}{a_{H}} \end{aligned} \end{aligned}\)       (3)

Considering a sliding direction collinear to the X axis (Fig. 10), it was shown in [6] that the general distribution of the local Archard work dissipated in the sphere/plane contact is provided by the following integral:

Fig. 10. (a) Pressure field Illustration, (b) determination of the dissipated local energy.

\(\begin{aligned} &\widetilde{W}(X, Y)=2 \cdot p_{\max } \cdot a_{H} \cdot \int_{X-e}^{X+e}\left[1-X^{2}-Y^{2}\right]^{\frac{1}{2}} d X \\ &\text { with } \mathrm{R}<1 \end{aligned}\)       (4)

With the sliding ratio and the maximum interfacial shear under full sliding:

\(e=\frac{\delta g}{a_{H}}\) and \(q_{\max }=\mu \cdot p_{\max }\)

Along the median X axis, Y=0 and the former expression is simplified to:

\(\widetilde{W}(X, Y)=2 \cdot p_{\max } \cdot a_{H} \cdot \int_{X-e}^{X+e}\left[1-X^{2}\right]^{\frac{1}{2}} d X\)       (5)

The maximum Archard work density dissipated at the center of the sphere/plane interface (x = y = 0) following the Hertzian hypothesis is expressed by:

\(\widetilde{W}_{H}=2 \cdot p_{\max } \cdot a_{H}\left(e \cdot \sqrt{1-e^{2}}+\operatorname{Arcsin}(e)\right)\)       (6)

When e = 1, the maximum Archard work density is

\(\widetilde{W}_{H}=2 \cdot a_{H} p_{\max } \cdot a_{H}\)       (7)

If the sliding amplitude is very small comparatively to the Hertzian contact radius (i.e. e → 0), the simplified formulation can be approximated by:

\(\widetilde{W}_{S}=4 \cdot p_{\max } \cdot \delta g\)       (8)

Fig. 11 compares the evolution of the studied DLC/ 52100 friction endurances as a function of the applied Archard work density considering respectively the simplified formulation (\(\tilde{W}_{S}\)) and the exact Hertzian formulation (\(\tilde{W}_{H}\)).

Fig. 11. Evolution of the Nc₂ friction endurance obtained as a function of varying pressure conditions and varying sliding amplitude conditions [Table 2]: (a) versus the simplified Archard work approximation (Eq.8); (b) versus the exact Hertzian Archard work approximation (Eq.6,7).

The application of the simplified formulation leads to a large dispersion particularly in the low endurance domain where large sliding amplitude conditions are imposed. Indeed, the applied sliding amplitudes are rather large compared to the Hertzian contact radius so that the simplified formulation cannot be considered. By contrast, the Hertzian formulation which considers an elliptical distribution of contact pressure provides a very good correlation between the friction endurance values (Fig. 11 (b)).

All the experiments are aligned along a single master curve, although a very large spectrum of contact pressure and sliding amplitude were investigated. The endurance can be formalized using a very basic polynomial formulation so that:

\(\begin{aligned} &\mathrm{Nc}_{2}=\mathrm{A}_{2} . \widetilde{W}_{H}^{2}+\mathrm{A}_{1} \widetilde{W}_{H}+\mathrm{A}_{0} . \\ &\text { with } \mathrm{A}_{2}=-5.84 \mathrm{E}^{+06} ; \quad \mathrm{A}_{1}=48.27 ; \quad \mathrm{A}_{0}=4.0 \mathrm{E}^{+06} . \end{aligned}\)       (9)

A reverse expression can be extracted to formulate the limit Archard work density dissipated during a fretting cycle for a given DLC friction endurance:

\(\begin{aligned} &\widetilde{W_{H}}=\mathrm{B}_{2} \cdot N_{c 2}^{2}+\mathrm{B}_{1 .} N_{c 2}+\mathrm{B}_{0} . \\ &\text { with } \mathrm{B}_{2}=-3.0 \mathrm{E}^{-099} ; \quad \mathrm{B}_{1}=0.032 ; \quad \mathrm{B}_{0}=1.17 \mathrm{E}^{+05} \end{aligned}\)       (10)

This approach was transposed to the Nc1 coating failure description (Fig. 12), As expected, the Nc1 are systematically smaller than the Nc2 friction endurances. Again, a single master curve is observed whatever the applied pressure and sliding conditions are (i.e. All the experiments are aligned along a single master curve).

Fig. 12. Evolution of the applied Archard work density as a function of the coating (Nc₁) and friction (Nc₂) DCY/52100 fretting endurances.

A vertical evolution is observed, which suggests that the so called coating failure criterion is not a function of the local Archard work, but seems to be dependent on the test duration Nc₁ = cst = 3.10⁵.

This typical evolution was also observed for TiN or TiC hard coatings fretting investigations[14]. Different hypothesis has been considered like for instance a fatigue cracking process[15-17]. Surprisingly the fretting scar analysis does not reveal any cracking processes.

Deeper investigations are now required to interpret this typical evolution of the Nc1 coating failure response which indirectly was related to the carbon and CrC transfer process.

4. Conclusions

This fretting wear investigations of a commercial DLC layer fretted against ball bearing steel 52100 underlined the following aspects:

- The presence of the DLC layer highly improves the friction response, and the failure is only observed after reaching millions of cycles compared to the naked contact. Thus, a significant improvement in the component’s lifetime is guaranteed.

- The friction evolution has been formalized in three periods: A first medium 0.4 friction plateau (period (I)), followed by a low 0.2 friction plateau (period (II)) characterized by the activation of a lubricious amorphous carbon and CrC third body on the steel counterbody. after which the friction converges linearly to the steel/ steel friction value (period (III)) due to iron and chromium oxides debris generation.

- Friction failure N_c2 can be formalized using a local Archard work density parameter \(\tilde{W}\). This analysis confirms that the exact. This analysis Hertzian formulation must be considered (\(\tilde{W}_{H}\)). The simplified approximation is not suitable when the sliding amplitude is large compared to the Hertzian contact radius.

Further researches are planned to better interpret the “coating failure” condition which is presently related to the activation of a lubricious interface.