DLC-Based Coatings Obtained by Low-Frequency Plasma-Enhanced Chemical Vapor Deposition (LFPECVD) in Cyclohexane, Principle and Examples

: The LFPECVD (Low-Frequency Plasma-Enhanced Chemical Vapor Deposition) technique is now used on an industrial scale for the deposition of carbon-based coatings for several applications. This short review recalled the main principles of LFPECVD and provided examples of DLC-based ﬁlms. The main differences between low-frequency (LF) and radio-frequency (RF) discharges were also recalled here and examples of deposition and characterization of carbon-based ﬁlms were proposed. The inﬂuence of the bias voltage or the temperature of the active electrode on the deposition rate and the structure of a-C: H ﬁlms obtained in cyclohexane/hydrogen mixtures was ﬁrst discussed. Next, the properties of carbon-based ﬁlms doped with silicon were described and, ﬁnally, it was shown that multilayer architectures make it possible to reduce the stresses without altering their tribological properties. ﬁlms deposited with different bias voltages (530 V, 630 V, and 670 V). These spectra are a combination of two main peaks located around 50 and 140 ppm and are associated respectively to sp 3 and sp 2 -hybridized carbons. No signiﬁcant change was observed, indicating that the proportions of Csp 2 and Csp 3 were similar for the three samples. This observation therefore allows us to validate the second hypothesis on G band displacement (Table 2), which would come from the variation of the internal stresses rather than from an increase of sp 2 hybridization since this one was not observed in NMR. From NMR analyses, the sp 2 /sp 3 atio is about three in these ﬁlms, which seems quite consistent with that of the G band position observed on the corresponding Raman spectra. and Young’s modulus, tribological The periods were between 14 and 400 for a-C:H/a-SiC:H a stacks, between for a-C:H/a-SiC:H and only was for case, the thickness multilayer was varying, schematically


Introduction
Carbon-based films are innovative materials that have been merged as an alternative solution of silicon due to their electrical and mechanical properties [1,2]. Up to now, various production processes including wet and dry methods have been conducted for fabricating carbon-based films. Rathanasamy et al. grouped the most used fabrication techniques [3]. Carbon-based films are composed of structures ranging from hybridized sp 3 diamond to sp 2 bonded graphite. Among the different carbon-based films, Diamond-Like Carbon (DLC) coatings are widely used in various industries such as automotive [4], electronics [5], biomedical [6], and aerospace [7]. The name DLC refers to a large field of carbon-based materials with original properties, some of which even appear relatively far from that of diamonds. Two main characteristics allowing for the classification these different coatings are: the ratio of Csp 3 /Csp 2 carbon hybridizations and the hydrogen content in the layer. DLC deposited by Low-Frequency Plasma-Enhanced Chemical Vapor Deposition (LFPECVD) is classified in the category of films of hydrogenated amorphous carbon a-C:H.
The first study referring to DLC films dates back to 1953 [8]. The development of research and deposition of these films using different techniques increased in 1970s [9,10]. These coatings found their first application on magnetic recording media in the early The temperature of the gas of "cold" plasmas, such as low-pressure (alternative) selfsustaining discharges, is close to room temperature while the electrons have enough kinetic energy to allow the breaking of interatomic bonds as well as the ionization of atoms and/or molecules during electron-molecule collisions.
A classification of plasmas is made according to the value of the frequency f of the electrical excitation. On the one hand, there is the continuous discharge (f is zero) and, on the other hand, the alternative discharges. If a plasma is subjected to a brief electrical disturbance, a momentary separation of charges appears, creating restoring forces which tend to bring it back to its equilibrium position. This phenomenon causes the oscillation of the charged species and in particular of the electrons which, being lighter, are the first to respond to the disturbance: in a plasma, the electrons and the ions oscillate respectively at characteristic frequencies called electronic (Fep) and ionic (Fip) plasma frequencies. Thus, the operation of the discharges in alternating mode depends on the value of the frequency f of the electrical excitation with respect to these two values of plasma frequencies Fep and Fip.
The low-frequency discharge (LF) corresponds to the case where f < Fip < Fep, Fip, and Fep are about 1 MHz and 1 GHz respectively. This defines a range of frequencies varying from a few hundred Hz to a few hundred kHz. In this type of discharge, ions and electrons instantly follow the variations of the electric field. The maintenance of the discharge is carried out by the emission of secondary electrons from the cathode.
The Radio-Frequency (RF) or high-frequencies (HF) discharge correspond to the case where Fip < f < Fep. This defines a range from about 1 MHz to around 100 MHz. The frequency commonly used in industry is 13.56 MHz, and it is set by international telecommunications regulations. At such frequencies, ions cannot follow the evolution of the applied electric field, unlike electrons which oscillate at the excitation frequency of the discharge. The ions are sensitive only to the time average value of the cladding field. However, it must be considered that Fip is inversely proportional to the mass of the ion. Thus, the "light" ions will be quite sensitive to the RF field while the "heavy" ions will be practically frozen. Such discharges require the use of an impedance adapter between the generator and the plasma. This adapter transmits maximum power to the discharge and protects the generator by reducing the impedance seen by it to 50 Ω, corresponding to the output impedance of an RF generator.

1.
The plasma potential V p : near a surface, the plasma interacts with it. Because of their mass, the speed of electrons is much greater than that of ions. Thus, at the initial instant, the flow of electrons on a wall is much greater than that of ions. The plasma then acquires a concentration of positive residual charges due to the excess of positive ions; this is the plasma potential, V p . It is therefore positive with respect to the walls and is often used as a reference.

2.
The floating potential V f : it corresponds to the potential to which an isolated sample, immersed in a plasma, is fixed. It is due to the fact that this sample first charges negatively thanks to high mobility of electrons. The sample stabilizes at this potential V f when the balance between electronic and ionic flux is achieved.

3.
The self-bias voltage, V dc , appears on the active electrode with the introduction of a blocking capacitor between the generator and the cathode. During the first positive half-period of a cycle, electrons are attracted to the cathode, causing negative charges to build up due to the existence of the blocking capacity. An identical phenomenon is obtained, but in the opposite direction for the ions during the next negative halfperiod. However, due to the difference in mobilities of ions and electrons, this variation in the potential of the electrode is greater during positive alternation. Thus, after a few periods, a balance will be reached between the electronic and ionic currents and the voltage at the level of the cathode stabilizes at the continuous value V dc . The value of this voltage depends in particular on the area ratio between the grounded electrode (or the reactor) and the active electrode. It is given by Turban [35] as follows: with, A a : area of the anode; A c : area of the cathode; V p : mean plasma potential; n: pressure-dependent factor. The influence of the ratio of the areas of the active electrode to that of the mass, on the shapes of the plasma potential V p , and of the excitation signal V(t) (Figure 1) for direct and capacitive coupling is discussed by [36].

Example of Deposition
Deposition was carried out in an industrial-size Low-Frequency PECVD reactor with a capacitively coupled electrode configuration. The lower electrode serving as substrate holder and having a diameter of 250 mm was powered via a 40-kHz LF transmitter. The samples were positioned directly on this substrate holder. A shower, with a diameter of 250 mm, ensured the gases inlet homogeneously into the chamber. The shower to substrate holder distance was kept constant at 200 mm. The pumping unit consisted of a primary pump and a roots and a turbomolecular pump. The based vacuum achieved was of the order of 2-4 × 10 −6 mbar. The pressure was regulated using a rolling valve. The temperature of the substrates is floating and measured using a thermocouple. For the series of deposits carried out with higher substrate temperatures, an additional plate provided with heating resistors was added to the device.

Example of Deposition.
Deposition was carried out in an industrial-size Low-Frequency PECVD rea a capacitively coupled electrode configuration. The lower electrode serving as holder and having a diameter of 250 mm was powered via a 40-kHz LF transm samples were positioned directly on this substrate holder. A shower, with a di 250 mm, ensured the gases inlet homogeneously into the chamber. The show strate holder distance was kept constant at 200 mm. The pumping unit consiste mary pump and a roots and a turbomolecular pump. The based vacuum achiev the order of 2-4 × 10 −6 mbar. The pressure was regulated using a rolling valve. perature of the substrates is floating and measured using a thermocouple. For of deposits carried out with higher substrate temperatures, an additional plate with heating resistors was added to the device.
The discharge power supply was an Advanced Energy PE2500 (Advanced Metzingen, Germany), associatedwith an integrated tuning box.
A gas mixture of cyclohexane (C6H12) and hydrogen was used to synthes C:H based layers. The choice of the precursor is very important for this type o because it has a direct effect on the growth rate. Cyclohexane is advantageous o precursors, as shown by J. Robertson [19], because it shows a good compromise a low ionization potential (below 10 eV) and a relatively high deposition rate ( in their deposition conditions. Electrical measurements of the plasma were performed using a Tektronix h age probe (P6015A, Tektronicx, Les Ullis, France) directly connected to the polar and connected to an oscilloscope. These measurements made it possible to see t cal signal on the plate during the deposition process. Figure 1 shows two exa signals obtained in the case of a cyclohexane/hydrogen gas mixture for two plasma powers. These signals highlight the different electrical parameters defin previous paragraph, namely, the applied voltage VLF(t), its average value VM, th voltage ̅ , and the self-bias voltage Vdc (Figure 1). The discharge power supply was an Advanced Energy PE2500 (Advanced energies, Metzingen, Germany), associated with an integrated tuning box.
A gas mixture of cyclohexane (C 6 H 12 ) and hydrogen was used to synthesize the a-C:H based layers. The choice of the precursor is very important for this type of coating because it has a direct effect on the growth rate. Cyclohexane is advantageous over other precursors, as shown by J. Robertson [19], because it shows a good compromise between a low ionization potential (below 10 eV) and a relatively high deposition rate (2.5 µm/h) in their deposition conditions. Electrical measurements of the plasma were performed using a Tektronix high-voltage probe (P6015A, Tektronicx, Les Ullis, France) directly connected to the polarized plate and connected to an oscilloscope. These measurements made it possible to see the electrical signal on the plate during the deposition process. Figure 1 shows two examples of signals obtained in the case of a cyclohexane/hydrogen gas mixture for two different plasma powers. These signals highlight the different electrical parameters defined in the previous paragraph, namely, the applied voltage V LF (t), its average value V M , the plasma voltage V p , and the self-bias voltage V dc (Figure 1).
It thus appears that the geometric and electrical configurations of the reactor used in this study led to a relatively low self-bias voltage (even with a high power, as illustrated in Figure 1b), and always much lower at the average plasma potential. This result was also observed with other C 6 H 12 /H 2 ratios and other working pressures.
In this study, the main parameter was, therefore, not this self-bias voltage since it is not directly representative of ionic energy here, unlike the case of an RF power supply in capacitive coupling. In the present case, the monitored parameter corresponded to the value of the voltage V M of the signal V LF (t), which represents the total bias voltage applied to the substrate holder. This voltage, in the present study, was defined by: More precisely, the polarization voltage values given in this study was V M √ 2 (if we consider a perfect sinusoidal signal) because the value given by the power supply is a Root-Mean-Squared (RMS) value of the signal and not an average value.

Characterization Methods
13 C Nuclear Magnetic Resonance (NMR) spectra were recorded with a 4-mm Bruker CPMAS (Mannheim, Germany) probe head on a Bruker AVANCE 400 spectrometer operating at 100.62 MHz for 13 C frequency. The experimental details are given in [33]. Raman spectra were recorded at room temperature with a Jobin-Yvon Dilor (HORIBA, Longjumeau, France) spectrophotometer. Two different wavelengths were used in this study: 514 nm and 633 nm. The power was kept below 2.5 mW to avoid "graphitization" phenomena. The silicon calibration was performed using the lines of silicon (521 cm −1 ) and mercury (1122 cm −1 ). The broadband spectra were performed in several spectral ranges, each range being performed in 2 × 120 s, i.e., 4 min per range. The peak characteristics were then obtained by fitting the spectra with Gaussian functions.
Residual stress was determined using a bending method. Curvature of stainless-steel foils was measured by optical profilometer (Altisurf 520, Altimet, Marin, France) and the stress level was deduced from the Stoney's equation [37]. Hardness and Young's modulus of coatings were evaluated by means of a Nano Hardness Tester (CSM Instruments, Anton Paar, Buchs, Switzerland) using a Berkovich diamond tip. Values were calculated using the Oliver and Pharr method [38]. A ball-on-disk tribometer (CSM Instruments) was used for friction coefficient measurements. Six-mm Al 2 O 3 balls were used as the mating material and a 3 N load was applied to the system, corresponding to an initial Hertzian contact pressure of about 830 MPa. Sliding speed was kept at 0.2 m·s −1 for a fixed number of 50,000 cycles. All tests were performed in an ambient environment (20 • C and 40% RH).

Deposition Rate
Deposition rate clearly decreases with the bias voltage, which can be explained by an improved precursor dissociation rate (samples 1 to 4, Table 1 and Figure 2) favored by higher plasma power. Moreover, it was shown that the deposition rate decreased slightly at 300 • C (sample 6, Table 1 and Figure 2) compared with that deposited below 150 • C. This can be due to an adsorption-desorption equilibrium, which is the rate limiting process in the investigated temperature range [39]. The deposited atoms and radicals would diffuse over the surface in a weakly bound, adsorbed layer before being incorporated into the film. Table 1. Deposition conditions of samples of a-C:H related to Figure 2; the film's thickness is 2 µm for a deposition duration between 60-100 min. Reproduced and modified from [33]

Structure of a-C:H Films
Various information can be obtained from Raman spectra. The two main par of interest are the position of the G band and the ID/IG ratio of two bands intensi It is well known that a displacement of the G band position towards the large numbers is associated with an increase in the sp 2 hybridization [40], while the inc the ID/IG ratio is associated with an increase in the size of sp 2 -type aggregates in bon matrix [41]. Concerning the influence of the bias voltage, two exciting wavele = 514 and λ = 633 nm) were tested while the influence of the substrate temperat analyzed only with a single wavelength of 514 nm. Figure 3 shows Raman spectra of the different characterized a-C:H films. The on each of these spectra confirmed the presence of sp 2 carbon in aromatic form as a disordered character of the films, which is the case for an a-C:H coating [19].

Structure of a-C:H Films
Various information can be obtained from Raman spectra. The two main parameters of interest are the position of the G band and the ID/IG ratio of two bands intensities.
It is well known that a displacement of the G band position towards the largest wave numbers is associated with an increase in the sp 2 hybridization [40], while the increase in the ID/IG ratio is associated with an increase in the size of sp 2 -type aggregates in the carbon matrix [41]. Concerning the influence of the bias voltage, two exciting wavelengths (λ = 514 and λ = 633 nm) were tested while the influence of the substrate temperature was analyzed only with a single wavelength of 514 nm. Figure 3 shows Raman spectra of the different characterized a-C:H films. The D band on each of these spectra confirmed the presence of sp 2 carbon in aromatic form as well as a disordered character of the films, which is the case for an a-C:H coating [19].
The changes in the position of G band and I D /I G ratio, as a function of the deposition and analysis parameters, are reported in Table 2. Table 2. Values of the G band positions and on the I D /I G ratio for different films [42].  The changes in the position of G band and ID/IG ratio, as a function of the deposit and analysis parameters, are reported in Table 2.  First, these values highlight the influence of the exciting wavelength on the spectra. It can be seen that even if the evolutions were similar, an increase in wavelength from 514 nm to 633 nm shifted the G band towards the lowest wavenumbers and increased the value of the ID/IG ratio. The shift in the position of peak G, called dispersion of peak G, actually reflects the degree of disorder in the material. The study of this dispersion over a wide range of wavelengths (from IR to UV) makes it possible to separate the materials into two classes depending on the nature of the sp 2 bonds (aromatic chains or rings). The different positions of G peak obtained here for the wavelengths of 514 nm and 633 nm indicated that the film also contained aliphatic sp 2 bonds [43]. Regarding the evolution of the spectra, whatever the wavelength used, the values indicated a displacement of the G band towards the highest wavenumbers when the bias voltage increased. The same phenomenon occurred when the substrate temperature increased from 150 • C to 300 • C.

Deposition Conditions Raman Parameters
Two hypotheses can explain this evolution. The first comes from the model developed by Beeman [40] and corresponds to an increase of the sp 2 hybridization. The second is based on the intrinsic stress evolution. Indeed, the modification of the stress level leads to an evolution of the inter-atomic strengths within the films and therefore an evolution of the atomic vibration frequencies, which are the origin of Raman peaks positions [44].
This hypothesis will be verified in the following paragraph by determining the internal stresses of the films and by estimating the Csp 2 /Csp 3 ratios from the NMR analyses.
For I D /I G ratio, it can be observed that it remained quite constant when the bias voltage increased while a slight variation was observed when the substrate temperature changed. This variation may lead to an increase in the size of the sp 2 carbon aggregates in the material.
These Raman analyses also allow the comparison of these films with other existing DLC coatings. Figure 4 shows the evolution of the position of G band and I D /I G ratio as a function of the percentage of hydrogen for different DLC coatings [45]. Green points correspond to the values obtained in this study; their positions appear clearly in the range of a-C:H cotings obtained by PECVD.
Two hypotheses can explain this evolution. The first comes from the model developed by Beeman [40] and correspond increase of the sp 2 hybridization. The second is based on the intrinsic stress ev Indeed, the modification of the stress level leads to an evolution of the inter strengths within the films and therefore an evolution of the atomic vibration freq which are the origin of Raman peaks positions [44].
This hypothesis will be verified in the following paragraph by determining th nal stresses of the films and by estimating the Csp 2 /Csp 3 ratios from the NMR ana For ID/IG ratio, it can be observed that it remained quite constant when the b age increased while a slight variation was observed when the substrate temp changed. This variation may lead to an increase in the size of the sp 2 carbon aggre the material.
These Raman analyses also allow the comparison of these films with other DLC coatings. Figure 4 shows the evolution of the position of G band and ID/IG ra function of the percentage of hydrogen for different DLC coatings [45]. Green poi respond to the values obtained in this study; their positions appear clearly in the r a-C:H cotings obtained by PECVD.   Figure 5 shows the evolution of the residual stresses values as a function of the bias voltage ( Figure 5a) and the substrate temperature (Figure 5b). All films had compressive stresses and the values were between −400 MPa and −2200 MPa depending on the deposition parameters. These results, which show an overall increase in residual stresses with bias voltage and with substrate temperature, allow us to confirm the second hypothesis mentioned previously about the evolution of Raman spectra. In other words, an increase of bias voltage or of substrate temperature causes a modification of the carbon network, more precisely a reorganization of the sp 2 carbons in aromatic form. This could lead to G band displacement in Raman spectra ( Table 2). The increase in C-C bonds thus is accompanied by an increase in the intrinsic stresses of the films. This phenomenon is particularly visible above 600 V, which is a threshold in these deposition conditions. Figure 5 shows the evolution of the residual stresses values as a function of the bias voltage ( Figure 5a) and the substrate temperature (Figure 5b). All films had compressive stresses and the values were between −400 MPa and −2200 MPa depending on the deposition parameters. These results, which show an overall increase in residual stresses with bias voltage and with substrate temperature, allow us to confirm the second hypothesis mentioned previously about the evolution of Raman spectra. In other words, an increase of bias voltage or of substrate temperature causes a modification of the carbon network, more precisely a reorganization of the sp 2 carbons in aromatic form. This could lead to G band displacement in Raman spectra ( Table 2). The increase in C-C bonds thus is accompanied by an increase in the intrinsic stresses of the films. This phenomenon is particularly visible above 600 V, which is a threshold in these deposition conditions.  Nuclear Magnetic Resonance (NMR) appears to be the most reliable concerning the quantitative analysis of sp 2 /sp 3 ratio of carbon hybridization states [19]. Figure 6 shows the NMR spectra (HPDEC: High Power Proton Decoupling) of 13 C corresponding to three films deposited with different bias voltages (530 V, 630 V, and 670 V). These spectra are a combination of two main peaks located around 50 and 140 ppm and are associated respectively to sp 3 and sp 2 -hybridized carbons. No significant change was observed, indicating that the proportions of Csp 2 and Csp 3 were similar for the three samples. This observation therefore allows us to validate the second hypothesis on G band displacement (Table 2), which would come from the variation of the internal stresses rather than from an increase of sp 2 hybridization since this one was not observed in NMR. From NMR analyses, the sp 2 /sp 3 atio is about three in these films, which seems quite consistent with that of the G band position observed on the corresponding Raman spectra.

Mechanical Properties of a-C:H Films
Nuclear Magnetic Resonance (NMR) appears to be the most reliable concerning the quantitative analysis of sp 2 /sp 3 ratio of carbon hybridization states [19]. Figure 6 shows the NMR spectra (HPDEC: High Power Proton Decoupling) of 13 C corresponding to three films deposited with different bias voltages (530 V, 630 V, and 670 V). These spectra are a combination of two main peaks located around 50 and 140 ppm and are associated respectively to sp 3 and sp 2 -hybridized carbons. No significant change was observed, indicating that the proportions of Csp 2 and Csp 3 were similar for the three samples. This observation therefore allows us to validate the second hypothesis on G band displacement (Table 2), which would come from the variation of the internal stresses rather than from an increase of sp 2 hybridization since this one was not observed in NMR. From NMR analyses, the sp 2 /sp 3 atio is about three in these films, which seems quite consistent with that of the G band position observed on the corresponding Raman spectra. Hardness and Young modulus increased with bias voltage ( Figure 7) and substrate temperature (Table 3). Contrary to the results from other studies [46], no optimum of hardness and Young modulus was observed with plasma power. A complex precursor was used in the present case; during a-C:H growth, species as CmHn + molecular ion hit the surface and broke up, sharing their kinetic energy between the individual carbon atoms.
The key parameter, then, controlling the a-C:H film properties corresponds to the effective energy per C atom. The energy per C atom necessary to obtain an optimum on hardness and Young modulus values of a-C: H films will require an incident ion energy six times higher in case of cyclohexane compared with methane. This could explain the lack of optimum in this case [33]. In this investigation, plasma power and substrate temperature induce a structural and mechanical reinforcement of a-C:H films which is associated with a decrease in friction coefficient [33]. As far as the sp 2 /sp 3 ratio remains constant (about 3), this reinforcement can probably be attributed to a carbon network rearrangement, which is associated with a shift of G band peak toward higher wavenumbers (Table 2).  Hardness and Young modulus increased with bias voltage ( Figure 7) and substrate temperature (Table 3). Contrary to the results from other studies [46], no optimum of hardness and Young modulus was observed with plasma power. A complex precursor was used in the present case; during a-C:H growth, species as C m H n + molecular ion hit the surface and broke up, sharing their kinetic energy between the individual carbon atoms.
The key parameter, then, controlling the a-C:H film properties corresponds to the effective energy per C atom. The energy per C atom necessary to obtain an optimum on hardness and Young modulus values of a-C: H films will require an incident ion energy six times higher in case of cyclohexane compared with methane. This could explain the lack of optimum in this case [33]. In this investigation, plasma power and substrate temperature induce a structural and mechanical reinforcement of a-C:H films which is associated with a decrease in friction coefficient [33]. As far as the sp 2 /sp 3 ratio remains constant (about 3), this reinforcement can probably be attributed to a carbon network rearrangement, which is associated with a shift of G band peak toward higher wavenumbers ( Table 2).

Si Doped DLC
One of the different methods to reduce the stress of a-C:H films is based on the use of various film compositions, for instance, with silicon addition [33] by means of Low-Frequency enhanced chemical vapor deposition co-deposition. Recently, metal-containing hydrogenated diamond-like carbon coatings (Me-DLC) have attracted great interest since their initial appearance 3 decades ago [47][48][49]. The Me-C:H coating has high elasticity and a low friction coefficient, which can be beneficial for many engineering applications [50][51][52][53][54]. In addition, the hardness can be adapted to find a compromise between excellent behavior and low friction coefficient [27]. Many transition metals have been tested, for instance, Ti [55][56][57][58][59], Cr [60][61][62], W [63,64], Al [65][66][67], and Cu [68,69]. These metals are often introduced in DLC by coupling a magnetron sputtering discharge to the Low-Frequency enhanced chemical vapor deposition discharge [26].
TMS (tetramethylsilane, Si(CH 3 ) 4 ) can be added to the cyclohexane and hydrogen mixture as a second precursor for co-deposition of a-SiC:H (Si doped DLC) [33]. R is defined as TMS flow rate/(TMS + Cyclohexane) total flow rate. The deposition rate increased from 2.7 to 3.5 m/h when R rose to 0.5 and stabilized when 0.5 < R < 1. This increase can be attributed to a lower ionization potential of TMS compared with other hydrocarbon precursors. The result is consistent with that of J.W. Yi et al. [70] and A. Bendavid et al. [71]. Table 4 gathers results on these a-SiC:H films: composition, hardness, Young modulus, and residual stress according to various R values.

Si Doped DLC
One of the different methods to reduce the stress of a-C:H films is based on the use of various film compositions, for instance, with silicon addition [33] by means of Low-Frequency enhanced chemical vapor deposition co-deposition. Recently, metal-containing hydrogenated diamond-like carbon coatings (Me-DLC) have attracted great interest since their initial appearance 3 decades ago [47][48][49]. The Me-C:H coating has high elasticity and a low friction coefficient, which can be beneficial for many engineering applications [50][51][52][53][54]. In addition, the hardness can be adapted to find a compromise between excellent behavior and low friction coefficient [27]. Many transition metals have been tested, for instance, Ti [55][56][57][58][59], Cr [60][61][62], W [63,64], Al [65][66][67], and Cu [68,69]. These metals are often introduced in DLC by coupling a magnetron sputtering discharge to the Low-Frequency enhanced chemical vapor deposition discharge [26].
TMS (tetramethylsilane, Si(CH 3 ) 4 ) can be added to the cyclohexane and hydrogen mixture as a second precursor for co-deposition of a-SiC:H (Si doped DLC) [33]. R is defined as TMS flow rate/(TMS + Cyclohexane) total flow rate. The deposition rate increased from 2.7 to 3.5 µm/h when R rose to 0.5 and stabilized when 0.5 < R < 1. This increase can be attributed to a lower ionization potential of TMS compared with other hydrocarbon precursors. The result is consistent with that of J.W. Yi et al. [70] and A. Bendavid et al. [71]. Table 4 gathers results on these a-SiC:H films: composition, hardness, Young modulus, and residual stress according to various R values.  Figure 8 shows Raman spectra in function of Si content in the films, which depend on TMS content in the gas mixture (R). When silicon is added in a-C:H films, there was a clear shift of G peak position toward lower wavenumbers and the D peak disappeared. These evolutions correspond to those already highlighted by Papakonstantinou et al. [72] for Si-doped DLC films obtained from TMS/C 2 H 2 mixtures. G peak downshift can be attributed to Si-C bonds formation (3.10 eV) which has a lower energy than C-C bonds (3.70 eV). Moreover, compressive stress, which decreases with increasing Si content, can also explain this Raman result. The evolution of the D peak could be due to the opening of Csp 2 rings when silicon is added in the structure.  Figure 8 shows Raman spectra in function of Si content in the films, which depend on TMS content in the gas mixture (R). When silicon is added in a-C:H films, there was a clear shift of G peak position toward lower wavenumbers and the D peak disappeared. These evolutions correspond to those already highlighted by Papakonstantinou et al. [72] for Si-doped DLC films obtained from TMS/C2H2 mixtures. G peak downshift can be attributed to Si-C bonds formation (3.10 eV) which has a lower energy than C-C bonds (3.70 eV). Moreover, compressive stress, which decreases with increasing Si content, can also explain this Raman result. The evolution of the D peak could be due to the opening of Csp 2 rings when silicon is added in the structure. Addition of Si is interesting because it promotes a significative reduction of the residual compressive stress of a-C:H films without degrading other properties (Table 4). A slight decrease of hardness and Young modulus occurred as Si content increased, essentially due to the change of bonds; for instance, Si-H and Si-C with lower energy than that of C-C. However, the friction coefficient decreased significantly when Si content increased. This result can be explained by SiO2 particles synthesis during the friction process. Anyway, friction coefficient of silicon-doped a-C:H films can be measured below 0.05 for R < 0.5 (Si content below 14.5 at.%).

Multilayering Approach
The period (λ) of a multilayer system corresponds to the sum of the thicknesses of two consecutive layers forming the stack (one a-C:H layer and one a-SiC:H or Si-DLC layer). Different periods were studied here, but the total number of layers was always adjusted to maintain a final thickness of about 2 μm. The objective of this study was to highlight the influence of the nature and the thickness of the layers on the mechanical properties of the multilayer systems (residual stresses, hardness, and Young's modulus, tribological properties). For that, four elementary layers (one a-C:H layer, two different a-SiC:H layers, and one Si-DLC layer) were selected. Addition of Si is interesting because it promotes a significative reduction of the residual compressive stress of a-C:H films without degrading other properties (Table 4). A slight decrease of hardness and Young modulus occurred as Si content increased, essentially due to the change of bonds; for instance, Si-H and Si-C with lower energy than that of C-C. However, the friction coefficient decreased significantly when Si content increased. This result can be explained by SiO 2 particles synthesis during the friction process. Anyway, friction coefficient of silicon-doped a-C:H films can be measured below 0.05 for R < 0.5 (Si content below 14.5 at.%).

Multilayering Approach
The period (λ) of a multilayer system corresponds to the sum of the thicknesses of two consecutive layers forming the stack (one a-C:H layer and one a-SiC:H or Si-DLC layer). Different periods were studied here, but the total number of layers was always adjusted to maintain a final thickness of about 2 µm. The objective of this study was to highlight the influence of the nature and the thickness of the layers on the mechanical properties of the multilayer systems (residual stresses, hardness, and Young's modulus, tribological properties). For that, four elementary layers (one a-C:H layer, two different a-SiC:H layers, and one Si-DLC layer) were selected. Table 5 summarizes the main deposition parameters as well as the hardness, Young's modulus, and residual stresses values of these four selected monolayers. The multilayer systems correspond to the association of a-C:H coating with one of the other three layers. These three configurations are interesting because they allow us to study different cases: • an a-C:H/a-SiC:H a system for which the two elementary layers have similar hardness and Young's modulus but different residual stress values, • an a-C:H/a-SiC:H b system for which the two elementary layers have very different hardnesses, Young's modulus and levels of residual stresses, • an a-C:H/Si-DLC system for which the configuration is identical to that of the first stack, namely hardness and Young's modulus similar for the two layers but with lower residual stress values for the Si-DLC layer. The difference arises, in this case, from the tribological properties of the Si-DLC layer (with R = 0.2), which are clearly superior to those of the two a-SiC:H layers. The tested periods were between 14 nm and 400 nm for a-C:H/a-SiC:H a stacks, between 25 nm and 400 nm for a-C:H/a-SiC:H b stacks, and only a only 50 nm was studied for a-C:H/Si-DLC stacking. In each case, the thickness of the multilayer was set at 2 µm, varying, then, the total number of layers between 286 for a period of 14 nm and 10 for a period of 400 nm. These systems are schematically shown in Figure 9. • an a-C:H/Si-DLC system for which the configuration is identical to that of the first stack, namely hardness and Young's modulus similar for the two layers but with lower residual stress values for the Si-DLC layer. The difference arises, in this case from the tribological properties of the Si-DLC layer (with R = 0.2), which are clearly superior to those of the two a-SiC:H layers. The tested periods were between 14 nm and 400 nm for a-C:H/a-SiC:H a stacks, between 25 nm and 400 nm for a-C:H/a-SiC:H b stacks, and only a only 50 nm was studied for a-C:H/Si-DLC stacking. In each case, the thickness of the multilayer was set at 2 μm varying, then, the total number of layers between 286 for a period of 14 nm and 10 for a period of 400 nm. These systems are schematically shown in Figure 9.

Stress
Compressive residual stress was between −850 and −750 MPa for a-C:H/a-SiC:H a films, whereas it was between −600 and −500 MPa for a-C:H/a-SiC:H b coatings with different period thickness. For both a-C:H/a-SiC:H a and a-C:H/a-SiC:H b systems, whatever the thickness of the layers constituting the stacks, the values of the residual stresses of the multilayers ranged between those of two reference monolayers. It was the case for a period of 50 nm (Figure 10).  (Table 6) were measured as a function of the stacking period. The total th ness of these monolayer and multilayer films remained unchanged, about 2 µ m.
The results highlight two points: • Whatever the nature of stacking a-C:H/a-SiC:H a or a-C:H/a-SiC:H b , evolution hardness and Young's modulus as a function of the period are not significant. In ticular, no increase in hardness could be observed even for the smallest period o nm thick; • For the two multilayer systems, hardness and Young's modulus values were betw those of the two monolayers constituting the stack and even appear very close to t average value. This appeared more clearly in the second case for which the prope of the two monolayers are very different. Each multilayer stack being compose an identical proportion of a-C:H and a-SiC:H film, the obtained values therefore s to respect the "law of mixtures".
These mechanical behaviors, observed here for these stacks of amorphous layers, The results showed that this third system achieved, for a period of 50 nm, a level of residual stresses lower than the average of the stress values of the two monolayers.

Hardness and Young's Modulus
The hardness and Young's modulus of a-C:H/a-SiC:H a and a-C:H/a-SiC:H b multilayer systems (Table 6) were measured as a function of the stacking period. The total thickness of these monolayer and multilayer films remained unchanged, about 2 µm. The results highlight two points: • Whatever the nature of stacking a-C:H/a-SiC:H a or a-C:H/a-SiC:H b , evolutions of hardness and Young's modulus as a function of the period are not significant. In particular, no increase in hardness could be observed even for the smallest period of 14 nm thick; • For the two multilayer systems, hardness and Young's modulus values were between those of the two monolayers constituting the stack and even appear very close to their average value. This appeared more clearly in the second case for which the properties of the two monolayers are very different. Each multilayer stack being composed of an identical proportion of a-C:H and a-SiC:H film, the obtained values therefore seem to respect the "law of mixtures".
These mechanical behaviors, observed here for these stacks of amorphous layers, differ from those frequently encountered for crystalline multilayer systems. These systems are very often characterized by structural reinforcement for the thicknesses of weakest periods [73,74].
This result, however, does not agree with the work carried out by Logothedis et al. [75] on stacked layers of amorphous a-C carbon, one rich in Csp 3 and the other rich in Csp 2 . Indeed, they observed, despite the amorphous nature of their layer, similar changes to those measured for crystalline systems, namely an increase in the hardness of the multilayer system when the thickness of the period changes from 45 nm to 14 nm. In the present case, this behavior could be explained by the fact that, unlike the case studied by Logothetidis et al., the structures of a-C:H and a-SiC:H layers are different and that the interactions, which can exist when two associated layers are of similar a-C nature, do not occur when the structures of the layers are different (Raman analyses). Table 6 gathers the values of hardness and Young's modulus obtained for a-C:H/Si-DLC stack using a period of 50 nm compared with those of other two multilayer systems (a-C:H/a-SiC:H a and a-C:H/a-SiC:H b ). Values of four monolayers are also presented. The four-monolayer stack has a behavior similar to that of other two multilayer systems. In other words, it does not show any structural reinforcement, for the period of 50 nm, compared with two monolayers (a-C:H and Si-DLC). Due to the properties of these monolayers, its hardness and Young's modulus are equivalent to those of a-C:H/a-SiC:H a system.

Friction Behavior-Influence of the Period in a-C:H/a-SiC:H a Stack
In this section, only the tribological properties of a-C:H/a-SiC:H a multilayer system, with the highest hardness and Young's modulus values, were reported as a function of the stacking period. Figure 11 shows evolution of the friction coefficients as a function of the number of cycles of four a-C:H/a-SiC:H a stacks with periods of 14 nm, 50 nm, 280 nm, and 400 nm. The curves showed values between 0.09 and 0.25 for these four multilayer films. Under the same conditions, a-C:H single-layer coating has a friction coefficient of~0.15 with strong instabilities while a-SiC:H a coating has a coefficient friction of~0.25 with a lower resistance to wear. Moreover, the stacks with the smallest periods (14 nm and 50 nm) exhibited lower and more stable friction coefficients than those with the largest periods (280 nm and 400 nm).
It is important to specify that the last layer of the stack was an a-SiC:H a layer, which has, under these same test conditions, a higher friction coefficient and a greater wear rate than those of the a-C:H layer.
Thus, for the highest periods, the thicker a-SiC:H a layers at the surface (140 nm for the 280 nm period and 200 nm for the 400 nm period) could explain the higher friction coefficients at the beginning of the test (Figure 11). The instabilities occuring then as a function of the number of cycles could come from the differences in behavior of the two layers of the stack. cycles of four a-C:H/a-SiC:H a stacks with periods of 14 nm, 50 nm, 280 nm, and 400 nm. The curves showed values between 0.09 and 0.25 for these four multilayer films. Under the same conditions, a-C:H single-layer coating has a friction coefficient of ~0.15 with strong instabilities while a-SiC:H a coating has a coefficient friction of ~0.25 with a lower resistance to wear. Moreover, the stacks with the smallest periods (14 nm and 50 nm) exhibited lower and more stable friction coefficients than those with the largest periods (280 nm and 400 nm). For the first two stacks with periods of 14 nm and 50 nm, the smaller thicknesses of these a-SiC:H a layers (respectively 7 nm and 25 nm) have less significant impact on the evolution of the friction coefficient. This latter also has lower values compared to those of a-C:H single-layer film as well as greater stability as a function of the number of cycles.

Conclusions
The low-frequency plasma-enhanced chemical vapor deposition in cyclohexane atmospheres is an easy-to-implement technology for carbon-based coatings deposition. LFPECVD (like RF PECVD), which has the capacity to be implemented in any scale, is widely used in industry.
This work aimed at gathering the main principles of LF discharges as well as giving examples of its versatility for depositing doped diamond-like carbon and multilayered coatings. In the case of a-C:H, increasing the bias voltage or the substrate temperature led to a reorganization of sp 2 carbons in aromatic form, causing a shift of G band peak toward higher wavenumbers. This reorganization in aromatic form results of an increasing in C-C bonds, which is associated with a growth of the intrinsic stresses and a decrease of the friction coefficient of the films.
Si-doped DLC films were deposited by LFPECVD co-deposition from cyclohexane and tetramethylsilane mixtures. Increasing Si content, G peak position shifts toward lower wavenumbers, which is attributed to Si-C bonds, because of their lower energy compared with that of C-C bonds. The addition of silicon reduced the residual compressive stress of a-C:H films without significant change of the hardness and Young modulus, however the friction coefficient decreased significantly when Si content increased.
Regarding the multilayering approach, whatever the thickness of the layers forming the stacks, the values of their residual stresses were all between those of two reference monolayers. Hardness and Young's modulus values were also between those of the two monolayers constituting the stack, and they even appear very close to their average value. The friction coefficient seems to depend essentially on the sliding property of the top layer.