Journal Pre-proof Scratching resistance of SiC-rich nano-coatings produced by noble gas ion mixing

SiC-rich nano-layers were produced at room temperature by applying ion beam mixing of various C/Si multilayer structures using argon and xenon ions with energy in the range of 40-120 keV and fluences between 0.25-3 x 10 16 ions/cm 2 . The mechanical behavior of the layers was characterized by scratch test. The scratching resistance of the ion mixed samples has been measured by standard scratch test applying an atomic-force microscope with a diamond-coated tip (radius < 15 nm) and they were compared to that measured on Si single crystal. The applied load varied in the range of 4-18 µN. The scratching resistance of the samples correlated with the effective areal density of the SiC; with increasing effective areal density the scratch depth decreases. Above sufficiently high effective areal density of SiC the scratch resistance (hardness) of the produced layer was somewhat higher than that of single crystal silicon. Previously it has been shown that such layers have excellent corrosion resistive properties as well. These findings allow to tune and design the mechanical and chemical properties of the SiC protective coatings. effective areal the scratch depth decreased. By the results on single Si the hardness of the irradiated sample was estimated; it turned out that the hardness of the SiC-rich layers is similar to that of Si if their effective areal density is higher than 1000 Si-C bond/nm 2 . It has been also shown that the chemical resistance of the ion mixed layer correlates with its hardness; if we prepare a layer with high chemical resistance its wear resistance will be automatically high, as well. The results enable the tailoring of the mechanical and chemical properties of SiC nano-coatings which could be served as protective layers in harsh environments.


INTRODUCTION
Thin films serving as protective coatings in different environments are essentially important in various applications. Silicon carbide (SiC) is a material with many advantageous properties like biocompatibility, high heat resistance, high wear resistance, good thermal conductivity and inertness in corrosive environments. Due to these properties among others it is often used as a protective coating in harsh and also in biological environment [1][2][3][4][5]. A variety of techniques are available to produce SiC films, like physical vapor deposition (PVD) and atmospheric pressure chemical vapor deposition (CVD) [6,7]. Sha et al. [8] used RF plasma sputtering and a compound SiC target to deposit 160 nm thick SiC on porous Si substrate, after deposition the sample was annealed between 700-1000 °C. Zhao et al. [9] deposited β-SiC with various crystalline sizes on Si at 350 °C applying catalytic CVD. Daves et al. [10] applied plasma-enhanced CVD at 400 °C to deposit 250 nm thick amorphous SiC thin films and studied the deposition parameters on the film properties. Laboriante et al. [2] deposited 50 nm thick SiC on Si (100) using low pressure CVD at 780 °C. These above works applied elevated temperatures and the thickness of the produced SiC was mostly above 100 nm.
If high or medium temperature, because of the possible damage of the to be coated substrate, is excluded one should apply other (mainly non-equilibrium) methods for the production of the coating layer. Similarly the thickness of the coatings can be also important: for miniature applications e.g. M/NEMS systems, micromotors, bio-implants ultra-thin coatings are desirable due to reduced physical space between the substrate and the surface [11,12]. Recently it has been shown that ion bombardment induced mixing (IBM) could be a J o u r n a l P r e -p r o o f good alternative to produce coatings with thicknesses below 100 nm at room temperature [13,14]. It has been reported that by applying (IBM) on C/Si multilayer structures the produced coatings exhibited excellent corrosion resistance properties [15]. Additionally it has been shown that this IBM can be described by TRIDYN simulation and consequently the corrosion resistance could be tailored by applying fast and cheap simulation techniques [16].
From the viewpoint of protective coatings besides the corrosion resistance the good mechanical behavior of the layer might be also important. As the thickness of the produced SiC-rich coatings was only some tens of nanometers conventional hardness tests are generally not suitable for testing. Alternatively a scratch test with an atomic force microscope (AFM) in which a tip is indented into the substrate and then moved at a fixed depth along the surface is applied for studying extreme thin layers [17]. It allows the sensing of tribological properties of hard coatings down to a deformation depth of several Å, thus, substrate related effects can be excluded [18]. E.g. Wienss et al. examined the scratching resistance of 100 nm thick magnetron sputtered diamond-like carbon coatings using AFM. They found a correlation between the hydrogen content in the film and the scratching resistance [19]. Wu et al. focused on the properties of B 4 C films thinner than 20 nm deposited by magnetron sputtering. A correlation was established between the compressive residual stress present in the films and the scratching resistance [20]. Beake et al. performed scratch tests on 5-80 nm thick tetrahedral amorphous carbon (ta-C) films deposited on silicon where they determined the critical load for failure in different layer thicknesses. Thicker coatings provided higher critical load [21]. Tangpatjaroen et al. investigated 500 nm thick nanocrystalline (nc) SiC films produced by chemical vapor deposition (CVD). They have shown that the scratching resistance of SiC compared to Si depends on the size of the contact [22]. It has been also shown that the deformation during scratching is size dependent which makes the evaluation difficult [23,24]. The scratch depth provided by the scratch test experiment in many cases can J o u r n a l P r e -p r o o f 4 be directly used to determine the wear resistance of the material studied, and additionally this phenomenological parameter in many cases is connected to such basic material parameters like hardness and friction coefficient [22,25,26].
Herein, we report on the scratching tests of SiC-rich nano-coating layers produced by ion mixing at room temperature. The coating layers were produced by applying various noble gas irradiation parameters on various multilayer structures consisting of C (10 nm) and Si (20 nm) layers on a Si substrate resulting in SiC-rich layers of various distributions. The scratch resistance was measured by AFM in combination with a diamond-coated-tipped cantilever.
We will show that the previously introduced effective areal density of SiC [15], which is in good correlation with the corrosion resistance of the layer, also correlates with the scratching resistance of the samples. Sufficiently high effective areal density of the irradiated sample resulted in a scratch resistance somewhat higher than that of single crystal silicon.

Production of SiC-rich layer by means of IBM
It has been shown previously that ion bombardment induced mixing (IBM) of layer systems containing pure C and Si layers may result in formation of SiC rich layer [13,15].
The actual SiC concentration distribution along the depth depends on the conditions of the IBM and the initial layer structure.
For producing various SiC in-depth distributions two different multilayer structures were irradiated by argon and xenon ions at room temperature. The structures C (10nm) /Si (20nm) /C (10nm) /Si (20nm) /C (10nm) / single-crystal Si-substrate and Si (20nm) /C (10nm) /Si (20nm)/C (10nm) / single-crystal Si-substrate were produced by sputtering; the J o u r n a l P r e -p r o o f 5 deposited layers were amorphous [14]. For easier reference the first structure will be called as 10-20, the latter one as 20-10 sample.
The angle of incidence was 7° with respect to the surface normal. The ion beam current density was 0.4 μA/cm 2 . The Xe + irradiation was performed in the Heavy Ion Cascade Implanter of the Institute for Particle and Nuclear Physics of the Wigner Research Centre for Physics in Budapest. The ion beam current density was 0.75 μA/cm 2 . In both cases the ion beam with typically millimeter dimensions was x-y scanned across the full sample surface (1-2 cm 2 ) in order to achieve good irradiation homogeneity within the exposed area. In both cases the heating effects are negligible.

AES depth profiling
The concentration distribution along the depth of the target components and the projectiles before and after the ion irradiation has been determined by AES depth profiling.
The Auger spectra were recorded by a STAIB DESA 150 pre-retarded Cylindrical Mirror Analyzer (CMA) in direct current mode. 1 keV Ar + ions were used for the AES depth profiling with an angle of incidence of 80º with respect to the surface normal. To avoid heating effects a mild ion sputtering has been applied; the removal rate was about 0.2 nm/s. This was achieved by applying an ion beam current of about 4 μA/mm 2 . One etching step lasted only 20 s. The ion current was kept constant during sputtering. The sample was rotated (6 rev/min) during ion bombardment. These parameters were chosen to minimize the ion bombardment-induced surface and interface morphology changes [27].
The detailed description of the AES evaluation is described elsewhere [15].
Summarizing; the measured C (KLL) Auger peak could be decomposed into graphitic and J o u r n a l P r e -p r o o f 6 carbide components. The relative sensitivity factor method for the calculation of the atomic concentrations has been applied [28]. The sputtering time was transformed to removed thickness for getting depth profiles [29]. Hence, the AES analysis provided the in-depth distributions of Si, C, SiC and Ar or Xe.

Scratch test
The AES depth profiles have shown that in the majority of the cases not the whole upper carbon layer was consumed by the IBM-induced compound formation. Therefore the SiC-rich region was covered by a carbon layer of various thicknesses. Before the scratch test this layer was removed by oxidation in microwave plasma for 10 min, which did not affect the SiC present [30]. The surface topography and the scratch resistance was investigated by an AFM (MFP-3D, Asylum Research, CA) applying a diamond-coated tip (tip radius < 15 nm) mounted on a stainless steel cantilever. Due to the high sensitivity and small tip radius, the diamond-coated tip was able to obtain surface morphology and to create constant-force scratches in the surface. Each sample received parallel scratches with 9 different normal forces systematically increasing from 4.6 μN to 17.4 μN, the testing forces were equally spaced in the range. The applied loading force was derived from the bending of the actual cantilever and its spring constant measured by thermal noise. The following procedure for each measurement has been applied. a./ Imaging of the area, b./ Creation of 1 μm long scratches with the preset force, the sliding speed was 800 nm/sec c./ Imaging of the area of the scratches with the same tip. The images were evaluated by applying Gwyddion [31] software. The scratch depth was calculated over several scan lines and averaged.
J o u r n a l P r e -p r o o f 7

SiC concentration distribution of the samples
The effects of the noble gas irradiation were studied thoroughly previously [8,9] , it has been shown that the in-depth distribution of the formed SiC could be tailored by varying    The quantification of these SiC distributions is not straightforward, especially because the concentration varies along the depth. For the easier handling of the SiC in-depth distribution and quantifying the amount of SiC produced the term areal density can be used, which is frequently applied in the case of IBM. It is the integral of an element, compound along the depth which gives the total amount of the chosen element, compound in the sample [33]. In our previous works, which dealt with the chemical resistance of the ion mixed layer against harsh chemical effects [15] we have introduced the term effective areal density of SiC which was defined as the integral of the concentration of SiC from that depth where its amount is J o u r n a l P r e -p r o o f 9 higher than 20% until that depth where its concentration goes below 20% again [15]. This was based on the experimental observations that good chemical resistance was connected to a certain SiC concentration. Table 1 summarizes the SiC effective areal density values of the samples studied by scratch test. They were calculated from the AES depth profiles by addition of the removed atoms from that depth where the SiC concentration is higher than 20% until that depth where its concentration goes below 20% again. This means that for irradiation 40 keV, 0.25 x 10 16 Ar + /cm 2 (Fig. 1b) the profile was integrated only between 9 and 13 nm. At the same time for irradiation 40 keV, 3 x 10 16 Ar + /cm 2 (Fig. 2a) the SiC concentration was mainly above 20%, therefore the profile was integrated between 7 and 63 nm.

Scratch test
The irradiated specimens and a single-crystal silicon have been tested with 9 equally spaced forces in the range of 4.6-17.4 μN. A typical measurement is depicted in Fig.3. Fig. 3 a

DISCUSSION
All the results, the experimentally measured scratch depth vs applied load curves for all samples are summarized in Fig. 4; the parameters are the irradiation conditions. The trends are clear, the scratch depth measured on a given sample increases (after a threshold) up to a saturation value, which remains constant during further increase of the load. Our primary interest is, however, to reveal dependence of the scratch resistance on the amount of SiC produced; this dependence needs another representation. Our previously introduced effective areal density of SiC showed a good correlation with the chemical resistance of the samples. In the present case it will be checked whether the same quantity can be used for the mechanical characterization of the samples after irradiation.
Note that the scratch test (the additional carbon layer which covered the SiC-rich region had been removed before the test) because of its low penetration depth measures the mechanical behavior of the region determined by the effective areal density, that is, that region which is responsible for the good chemical resistance. The effective areal density values calculated in this experiment for the different irradiations were summarized in Table 1. increasing effective areal density, that is, with increasing amount of SiC (consequently increased thickness of the SiC-rich layer) the scratch depth generally decreases down to a saturation value. Further increase of the SiC amount does not lead to significant improvement of the scratch resistances. This general behavior is somewhat obscured by some additional parameters. We observe that at low applied loads (Fig. 5a) at high effective areal densities, above 1000 Si-C bond/nm 2 from the three measured values two (for 1100 and 1826 Si-C bond/nm 2 ) is 0, which means that our load is lower than the threshold value. It fits the trend since with increasing effective areal density the scratch resistance increases and accordingly the threshold is higher than the applied load. At 1727 Si-C bond/nm 2 effective areal density well measurable scratch depth was obtained; this behavior might be explained that the threshold has some scatter and being close to it, scratch forms or does not form. This explanation is supported by the measurements at higher loads of 12. The scratch test is a relatively easy measurement; one should simply scratch the material and measure the scratch depth, which in principle can be directly connected to the wear resistance of the material [22,26,34]. It turns out, however, that the scratching process is a rather difficult one and one must be cautious with interpreting the results. If the scratching device, tip, is not sharp and the applied load is not too low, then the majority of the work made by the load while creating the scratch is used for plastic deformation. In this case the wear resistance is proportional to the hardness of the given material, H [35]. It means that a simple equation, namely the Archard equation [36] connects the experimentally easily where, V is the volume of material removed, k is a dimensionless wear coefficient, W is the normal load, x is the sliding distance, and H is the hardness of the softer of the two contacting materials using a diamond tip generally the studied material. Eq. 1 provides easily measurable data for determining the hardness of the material. higher than that of the softer (Si) material. This behavior was explained by the higher weight of the friction which means that friction dominates plastic deformation.
In our experiment similar diamond tip with the same radius has been applied but the applied load range was somewhat higher, 5-17 µN, than that of Tangpatjaroen et al. [22] The question arises which process is valid for our case, whether the plastic deformation or the friction process is dominating? Considering Fig. 5 it can be seen that with the increase of the effective areal density, with the increase the amount of SiC in the surface close layer, the scratch depth decreases. This behavior can be easily explained by eq 1 which tells that the J o u r n a l P r e -p r o o f scratch depth is inversely proportional to the hardness, assuming that the hardness of the composite material scales with the amount of SiC. If the friction process determines the scratch depth then we should consider that based on Tangpatjaroen et al. [22] the interfacial strength of the Si is higher than that of SiC which caused the strange result. In our case the amount of unreacted Si decreases with increasing effective areal density and the increase of the scratch depth is expected. Since the experimental results are just contrary we conclude that at our higher loads the scratch depth is inversely proportional to the hardness.
The SiC has various polytypes; in bulk form all of them are hard with a hardness in the range of 33 ±5 GPa. It is also known that the hardness of the micro and nano crystalline SiC layer is even higher than that of the bulk SiC, being in the range of 40 GPa. The hardness of bulk Si is around 10 GPa, that is much lower. Though the hardness measured on thin films generally differ from that of bulk still the above relation prevails for the thin films made of Si and SiC. In the case of SiC and Si thin films the hardness strongly and weakly depends on the way of preparation, respectively. Namely, in the case of SiC the hardness varies (depending on the deposition parameters used for thin film production) in the range of 25-50 GPa while in the case of Si this range is much lower 8-12 GPa [37,38]. It should be emphasized, however, that we have not found, hardness data for real thin layers of either material. Thus, for qualitative comparison of the hardness values we will use the Si single crystal (exhibiting well defined hardness value) sample for reference.
Considering the change of the scratch depth vs effective areal density (Fig. 5) we conclude that up to about 1000 Si-C bond/nm 2 effective areal density the scratch depth decreases with increasing areal density, while above this value remains constant. The qualitative explanation of the first region of the curves is simple; the effective areal density increases, the ratio of SiC/Si increases that is, the relative weight of the harder SiC increases J o u r n a l P r e -p r o o f thus, the scratch depth should decrease. Afterwards above 1000 Si-C bond/nm 2 the hardness of the intermixed layer is not increasing any more, it reached a stationary value.
Comparing the scratch depths obtained from the ion irradiated sample with that of Si it can be seen that at the lowest areal density independently from the applied load the scratch depth is anomaly high, that is the hardness is extremely low; the scratching depth is larger than that value measured at pure Si, respectively. This low value can be explained by the fact that in these samples low argon fluence irradiation results in a very low SiC production (see Fig. 1b). Practically we are measuring the mechanical properties of the non-irradiated sample, that is, the behavior of the first amorphous Si layer on a soft graphitic carbon layer. Based on this result we conclude that the hardness of this system is rather low. On the other hand in the case of high SiC effective areal densities (> 1000 Si-C bond/nm 2 ) the scratching depth is lower or approximately the same than that value measured at pure Si, therefore the hardness of the ion mixed layers is equal or higher than 12 GPa.
Concluding applying proper irradiation parameters the hardness of the irradiated samples approximates the properties of the single-crystal silicon. Though these hardness values are lower than that of the SiC, they are still sufficient for certain applications and by varying the film structure and parameters of irradiation it can be tailored according to the demand.
Previously the corrosion behavior of the ion mixed layers has been measured, as well [15]. Potentiodynamic corrosion test has been performed in an electrochemical threeelectrode glass cell in 4M KOH solution. The corrosion current densities were obtained by Tafel extrapolation from the polarisation curves from which the corrosion rates were calculated by Faraday's law. It is interesting to compare the corrosion and the scratch resistance of the samples. Fig. 6 shows together the corrosion rates and the scratch depths in function of the SiC effective areal density values for all samples. The shapes of the two curves are surprisingly similar, even the points at low areal densities behaves similarly. It means that the two quantities, corrosion resistance and scratch resistance (hardness) depends on a similar way on the amount and distribution of the SiC.
Consequently if either of them is tuned according to some demand the other simultaneously changes as well. The advantageous consequence of this behavior is that if we prepare a layer with high chemical resistance its wear resistance will be high as well.

CONCLUSIONS
Various C/Si multilayers were irradiated by Ar + and Xe + ions with energy in the range crystal Si the hardness of the irradiated sample was estimated; it turned out that the hardness of the SiC-rich layers is similar to that of Si if their effective areal density is higher than 1000 Si-C bond/nm 2 . It has been also shown that the chemical resistance of the ion mixed layer correlates with its hardness; if we prepare a layer with high chemical resistance its wear resistance will be automatically high, as well. The results enable the tailoring of the mechanical and chemical properties of SiC nano-coatings which could be served as protective layers in harsh environments.