3.2. XRD Analysis
Figure 3 and
Figure 4 present the X-ray diffraction (XRD) spectra of all samples from both methods. The reference spectrum consists of cohenite (Fe, Ni, Co) 3 C up to 3% and Iron (the most pronounced peaks), with a mean hardness value of 5.39 GPa (Mohs scale) and a relatively high FC = 0.52. The temperature treatment of boron-doped samples at 650 °C (see
Figure 3) yielded a rich spectrum consisting of nanostructured h-BN (◊) (reference code 98-016-8892) with a mean grain size of 41 nm deduced by reflections from plan (002) using the Scherer formula. A c-BN phase (♦) is detectable (reference code 00-015-05000); however, the mean grain size could not be estimated, as thickening of the spectral lines occurred because of peak superposition. The concentrations of h-BN and c-BN are 53% and 47%, excluding the other constituents. The other phases are iron boride (*) Fe
2B (reference code 98-016-0791) (83%) and Iron Carbide C
3Fe
7 (17%), excluding the other constituents. Therefore, B-based phases have a 36% concentration, and Fe-based phases have a 64% concentration.
The samples treated at 850 °C show a c-BN phase (reference code 00-015-0500) and an h-BN phase (ref code 00-026-0773) with a mean grain size of 34 nm, as well as iron-boride (*) Fe2B (ref. code 98-016-0791). The remaining reflections are from iron boride. The boron phases are 19%, of which 10% is c-BN and 9% h-BN.
X-ray radiation is very energetic, and the reflected signal depends mainly on the radiation length (cathode type—Cu, Cr, Co, W, etc.), as well as on the nature of the characterized material—nitrides, carbides, oxides, or intermetallic compounds—and not on the grain size of the matrix. If a copper cathode is used, the depth could reach up to 5 μm, but it is often at the scale of 1 to 2 μm. According to the (H-h) displacement graphs (see
Section 3.6. Growth Dynamics of the BN Phases ), the treated zone of the samples reached 70 nm. Thus, the method takes an integral signal from a certain depth, and the thin surface composition has less of an impact in the case of theta-2theta geometry. As discussed in the nanoindentation section, the modified thickness reached 500 nm, which provides a reasonable accuracy for this method. This method reveals that films treated at 650 °C for 1 h in an N
2 atmosphere consist of 36% boron-based phases, 17% iron carbide, and 47% iron boride, while the samples treated at 850 °C consist of 19% boron-based phases and 81% iron boride. The main peak <111> of the c-BN phase of the sample treated at 650 °C (reference code 00-015-0500), which usually appears at 43.254° (
a =
b =
c = 3.62 Å) with spacing of
d = 2.09 Å, is shifted to 42.61
o. Applying the Bragg rule for Cu Kα radiation yields interplanar spacing of
d = 2.12 Å, corresponding to
a = 3.67 Å, keeping in mind that
for the cubic symmetry. It appears that the cell parameter is greater with 0.05 Å, suggesting a deficit of N
2 atoms forming the cell. However, the same is not applicable for the h-BN phases that do not suffer any lack of nitrogen atoms. Following the same procedure for the c-BN phase (reference code 00-015-0500) for the samples annealed at 650 °C, we found
a = 3.65 Å, which is closer to the original state, 3.62 Å.
The RIE-treated samples present different spectra (
Figure 4) than those of the temperature-treated samples. The most pronounced peaks at the higher power (300 W) spectrum at 44.6°, 65.03°, and 82.43° belong to the cubic Fe (03-065-4899), and all the other (minor) peaks belong to a cubic Fe
3O
4 (01-089-2355) iron oxide. Apparently, B-containing phases are not detectable via this method (BN phase was confirmed by XPS measurements only for the A300 sample).
Figure 4 also shows that the Fe peak (211) for both 200 W and 300 W samples suffers a notable intensity increase—from 1480 to 1730 a.u.—increasing the grain size from 20 nm to 35 nm, respectively. The same trend is evident for Fe (200) planes, where the intensity ranges from 736 to 790 a.u., and the grain sizes range from 20.3 to 27 nm. The latter results in a 55% increase in grain size, which produces better organized and aligned grains in a less densely packed structure. The latter opens a larger possibility of interface diffusion. It is important to remember that the iron samples have simple cubic packing wherein the free space is proportional to the r
2 of the spheres.
3.4. XPS Analysis
Compared to other experimental methods used in the present study, XPS is sensitive to the uppermost atomic monolayers. Its depth of analysis (around 4–6 nm) is defined as three times the mean free path of the photoemitted electrons in the solid, depending on their kinetic energy.
Figure 7 presents a wide spectrum of polished AISI 1045 reference samples. Besides the Fe
2p peaks and Fe LMM structure due to Auger transitions, only C1s- and O1s- peaks are visible. Note that about 60% of the XPS signal comes from the uppermost atomic surface layer and attenuates exponentially at depths below the surface. Therefore, the latter peaks can be attributed to carbon- and oxygen-containing adsorbed particles.
XPS analysis of the annealed samples aimed to establish the quantitative and qualitative composition of the sample surface, as well as to identify the products of precursor conversion due to past chemical reactions.
Table 2 summarizes the elemental concentrations for the annealed samples according to the XPS analysis. The above-mentioned higher sensitivity of XPS to the uppermost surface atomic layer can explain the relatively higher concentrations of carbon and oxygen attributed to adsorbed species, which, to some extent, shield the signal from the B-containing film below.
Sample 1 (450 °C; see 1/450 °C in
Table 2) shows that the boron concentration is the lowest, at 9 at%, which indicates that the rate of chemical conversion is insignificant at this temperature. Here, the BN phase is below 10%. The proof of this is the small concentration of nitrogen, which is less than 2 at%. Here, the layer is thin. The concentration of Fe that is a constituent element of the pad is higher, at 4.2 at%, suggesting that the overlying B-containing layer is thin. For the 2/650 °C and 3/850 °C samples (see
Table 2), the higher temperatures decrease the percentage of oxygen and increase that of carbon, while the concentration of the remaining elements—iron and sodium—remains constant. The treatments at higher temperatures lead to a higher degree of oxygen combustion, which leads mainly to increased carbon content. The latter ends with the film surface masked with carbon and hydrocarbon combustion products, leading to further agglomeration of the particles. At the higher temperature, the interaction between the boron precursor and the nitrogen flow is activated, resulting in more nanocrystalline boron nitride.
Characterizing the chemical composition of the surface layers of the samples requires a detailed study of the shape of the various photoelectron peaks.
Figure 8a presents the C1s peaks for all samples. They are very broad because there are many different C-containing chemical bonds. All the spectra can be deconvoluted into four peaks. The first group of peak contributions, located in the range of 282.9–283.3 eV, is characteristic of metal carbides [
30,
31]. The second group of contributions at 284.4–284.9 eV could be associated with C-C bonds [
32]. The peak contributions at 285.7 eV, 286.3, and 286.9-287.5 eV could be attributed to C-H, C-O-C, and C=O bonds, respectively [
33]. Sample 1 (450 °C) exhibits an additional C 1s peak at 289.1 eV, associated with COOH groups [
34,
35].
Furthermore, the O1s photoelectron regions can be fitted with several peak contributions (
Figure 8b). The first group of peaks at about 530 eV can be associated with O in the frame of FeO [
36]. The second peak-contribution group at about 531.2 eV can be attributed to the presence of C = O or –OH¯ groups [
37]. The third group of peaks at about 532.4 eV is due to the presence of crystal-hydrated water in the composition of Na
2B
4O
7 · 10H
2O [
38]. The last peak, found only in sample 2, can be assigned to O bonded with B and Na in the compound Na
2B
4O
7 · 10H
2O [
37]. Again, in sample 2, another peak at 528.9 eV can be associated with some amount of partially reduced non-stoichiometric iron oxide (Fe
2O
x).
Figure 9a,b displays the XPS spectra of the B1s core level for the annealed samples. The final annealing temperatures are indicated in the figure. The spectrum corresponding to 450 °C annealing is broad and can be fitted with three peak contributions. None of them can be associated with BN phases. The first and second peak contributions at 186.6 and 188.4 eV are associated with B–B bonds of elemental boron [
39], BC, and Fe–B phases (Fe
2B, FeB) [
40]. The third peak at 191.1 eV is associated with the organic phases of B [
40].
The spectra of samples 2 and 3 (
Figure 9b,c) deserve more attention. The spectra can be deconvoluted into five peaks. The first and second photoelectron lines located at 186.1 and in the range of 188.4–188.8 eV are assigned to formed phases of B
xC, Fe
2B, and FeB. The fourth and fifth peaks are associated with the organic phases of B and B in Na
2B
4O
7 · 10H
2O [
39,
40,
41]. The third peak at 190.1–190.3 eV (samples 2 and 3) belongs to the formed phase of BN [
42,
43]. This peak of sample 2 is more intense than that of sample 3, and its integrated area is approximately 31% of the area of the whole spectrum. This value is close to the calculated value from XRD analysis. Here, the value is lower due to the screening effect of the surface carbon.
For sample 3, the area of the corresponding BN phase at 190.3 eV is 18% of the area of the whole spectrum.
Table 3 summarizes the percentages of the boron phases.
For sample 1, the N1s XPS spectrum is provided in
Figure 10a, and no peaks are associated with the BN phase. Three peak contributions correspond to N bonded with/in organic compounds, such as the imine groups (R = N − R groups (398.2 eV)) [
44], amine groups (C – NR2 groups (399.9 eV)) [
45], or alkyl ammonium groups (NR4+ (401.3 eV)) [
46].
Figure 10b,c shows N1s photoelectron spectra for samples 2 and 3. The peak position for the BN phase is 397.8 eV. Consistent with the data reported above, the peak intensity in sample 3 is lower than that in sample 2. The peaks at 396.4–396.6 eV are associated with N in the composition of FeN [
47].
The reported XPS results show that the BN phase is formed in annealed samples 2 and 3. The precise calculation of its stoichiometry can be done after area normalizations (to their ionization cross sections) of B 1s and N 1s peaks corresponding to the BN phase. For sample 2, these peaks are located at 190.1 eV and 396.4 eV. The ratio of their normalized areas gives a B:N concentration ratio close to 2. The shortage of nitrogen resulting from non-stoichiometric BN phase formation is also seen in the thinner XRD (111) peak shift towards the smaller theta-2theta. A significantly better result was obtained for sample 3, considering the B 1s and N 1s peaks of BN at 190.1 eV and 396.7 eV, respectively. The concentration ratio is 1.1, corresponding to a perfect BN stoichiometry. In conclusion, sample 1, annealed at 450 °C, is composed of a mixture of organic B. The temperature was insufficient to form a BN phase. For samples two and three, annealed at 650 °C and 850 °C, respectively, a BN phase was formed. According to the analyses, the sample annealed at a lower temperature shows a slightly higher BN phase concentration within the instrumental error (~1 at%), but to some extent, it is non-stoichiometric. A stoichiometric BN phase formed after annealing at 850 °C.
3.6. Growth Dynamics of the BN Phases
During the annealing and RIE treatment of the samples, interdiffusion phenomena occurred between both the substrate and film media, leading to extensive phase formation. Temperatures, in general, play an extremely important role in solid kinetic processes and define temperature-affected variables, such as diffusivity and the rate of the chemical reaction plotted logarithmically versus 1/T (K). During these treatments, surface or bulk-diffusion processes controlled by transport via grain boundaries, cracks, or vacancies are common. Bulk diffusion occurs at very high temperatures and is characteristic of single-atom paths. Processes controlled by transport via grain boundaries, cracks, or vacancies depend on the grain size and film density. Many works [
48,
49,
50] have studied the effect of grain size and grain boundaries on the mechanical properties of materials. Different crystallite sizes of similar materials result in differing hardness. Huang et al. studied the effect of crystallite size on hardness in AlCrNbSiTiV [
51] and found that grains of 50.6 nm and 15.5 nm result in a differing hardness of 7.5 GPA. Vacancies act as transition centers that promote material transport, and according to the gradient model of vacancy concentration [
52], their concentration is highest near the surface and decreases within the sample, predicting a depth-dependent diffusion coefficient. However, dynamic manner, reaction rate, and diffusion constants depend on the temperature range and the complex nature of the acting film constituents.
The load-discharge indentation curves illustrated in
Figure 12 were acquired for the treated samples at different temperatures, resulting in different shapes depending on the sample hardness if measured at identical loads. Gradually increasing the load force at certain depths of penetration decreases the hardness. According to the Oliver–Pharr method [
53], the adopted evaluation parameters are
hc =
hm −
ε(
hm −
hr) and
he =
ε(
hm −
hr). Here,
ε is the form factor (0.7268 for Berkovich-type indenter);
he, hm, and
hr are the elastic, maximum, and residual depths, respectively; whereas
hc is the contact depth of the indenter with the material at maximum load. The hardness is estimated as
H = Fmax/Ap, where
Ap = 24.5 hc2 is the projected area for a Berkovich-type indenter, and
Fmax is the maximal applied load. To estimate the physical constants controlling phase-formation processes, the contact depth as
hc at maximum load was considered to calculate both the hardness and the D [m
2/s] parameter (see
Table 4 and
Table 5), recognizing that such loads do not exceed the depth of the coatings.
To describe the formation kinetics, Arrhenius plots were made for both thermal-annealing processes (
Figure 13) and ion-milling processes (RIE) (
Figure 14). The diffusion phenomenon is similar to that used to describe the rates of chemical reactions. It uses mathematical models and accurately describes physical phenomena.
A chemical reaction occurs, surmounting an energy barrier, where the thermodynamic probability,
P, and the kinetic energy of the particles are described by Boltzmann-type formulas, such as
P =
exp (−Δ
G*/RT). The free-energy difference, Δ
G*(per mol), between normal and activated states is known as the activation free energy. Smaller Δ
G* values and higher temperatures increase P exponentially and thus strongly enhance the prospects for atomic motion [
54]. A proportional form of
P has the diffusion coefficient introduced earlier, namely:
where
Do is the temperature-independent pre-exponential factor and depends on how often (frequency) molecules collide when all concentrations are 1 mol/L and whether the molecules are properly oriented when they collide; and
EA is the activation energy of the process and is used to describe the energy required to reach a transition state, also called “migration energy”. A higher value of
EA means a “solid” or better barrier; or slower chemical reaction, including phase transformations. In the case of two adjacent media, a region
x ≥ 0 of one substance and a region
x < 0 of another substance exist, and the reaction/diffusion-penetration profile should conform to the solution of the diffusion equation for a pair of semi-infinite solids:
respectively, where
H(x,t) is the hardness (
H) at a depth,
x, after the diffusion interval,
t;
Ho is the initial value of the element/compound; and
D its diffusivity; or Δ
H/Ho is the phase-formation process through its actual hardness value.
Equation (4) refers to the error function, while
erfc(x) = 1 −
erf(x) is the complementary error function. The rewritten form of (2)—
erfc −1(2
H/Ho) = x/[2
(Dt)1/2]—is a linear function of the depth,
x, and allows for the assessment of
D (see
Table 4 and
Table 5). Hence, the process rates of BN phase formation evaluated considering the film-hardness dynamic for both annealed and RIE-treated samples are:
respectively, where
R, the gas constant = 8.314 × 10
−3 [kJ/mol °K], which is similar to the Boltzmann constant,
kB, which relates the average relative kinetic energy of particles in gas/solids with its thermodynamic temperature.
Figure 13 presents relation Equation (5) and refers to the fitted data shown in
Table 4. The pre-exponential coefficient (2.4 × 10
−18) [m
2/s] refers to the rate of phase formation, similar to diffusion rate. Grigorov et al. [
55] reported an analogous process rate in the temperature range of 400–900 °C for silicon diffusion in TiN films.
Do is called the frequency factor or the attempt frequency of the reaction and accounts for the total number of collisions (leading to a reaction or not) per second. Therefore, the denser the barrier/substrate (free of defects), the fewer collisions with inner atoms. In the same report, the authors alleged that TiN thin films with different microstructures show different diffusivity coefficients. The B+ and Bo TiN films deposited with and without the ion-assisted PVD process show diffusivity from 400 to 900 °C, respectively:
and
The obtained activation energies are typical for grain-boundary diffusion [
53]. According to Sarah Khalil et al., Ag atoms appeared to follow the Σ3 grain-boundary transport process in SiC substrate annealed up to 1300 °C, and the diffusivity of 4 × 10
−20 (m
2/s) is a result of a process partially controlled by grain-boundary diffusion [
56].
According to
Table 5, the RIE process occurred at lower temperatures. The N
2-induced RF plasma should play an important role in the transformation of the B-containing coating into BN film. Moreover, the samples were subjected to a negative bias from 200 to 400 eV. The greater difference in the activation energy for the RIE (124 kJ/mol vs. 25.3 kJ/mol) suggests pronounced surface diffusion in the voids arising during this process. A similar
EA value was obtained for the BN formation via CVD (115.1 kJ/mol), and the process was controlled by the surface chemical-reaction kinetics [
57]. The RIE process occurred with relatively denser plasma (Ar + N
2) of 13.3 Pa (100 mTorr), promoting a shallow zone of microcracks. The changes in the near-surface morphology of RIE-treated samples evokes, for this temperature region, a non-typical temperature-dependent process (see the slope of the fitted curve in
Figure 14), which does not ensure enough time (and sufficient temperature) for the desired phase formation. Moreover, ion bombardment could result in resputtering of the constituent species, where both consist of light elements.
To relate the obtained results and gain a better understanding of the impact of both the thermal and ion treatment, a multiple linear regression was considered using two independent variables of temperature: polarization during RIE and hardness. The equation from the type is
z =
a0 +
a1x +
a2y, where
z is associated with the hardness (
H), and
a0 =
Ho yields the coefficients of the independent variables:
a0 = 5.14;
a1 = 0.52;
a2 = 0.19; or
.
Table 6 summarizes these results. Apparently, the temperature factor appears to be almost three times more efficient than ion bombardment for c-BN phase formation. For the independent tribological data (FC), a probable reassessment needs to be undertaken or ion bombardment, in an indirect manner, provides conditions for lowering the friction coefficient. It appears that both techniques should act synergistically. A technique that supplies simultaneous action of both processes could be reactive magnetron sputtering (e.g., HiPIMS) at elevated temperatures.