Substrate Effects on the Corrosion Behavior of Tungsten- and Titanium-Doped Diamond-like Carbon Films on 316L Stainless Steel and Ti6Al4V

Abstract

Diamond-like carbon (DLC) films are widely used for corrosion protection, but the same film may perform differently on different substrates. Tungsten-doped (W-DLC) and titanium-doped (Ti-DLC) DLC films were sputtered with the same deposition parameters on 316L stainless steel and Ti6Al4V. The films were characterized by SEM, EDX, water contact angle measurement, and Raman spectroscopy. Their corrosion behavior in 3.5 wt% NaCl was evaluated by potentiodynamic polarization and electrochemical impedance spectroscopy. The substrate affected the wettability of Ti-DLC but had little effect on that of W-DLC. It hardly changed the bonding structure of either film. The effective capacitances were too high for insulating barriers. The corrosion current density of each film was also largely independent of the substrate. Both observations indicate that the electrochemical response was governed mainly by the metal-doped film surfaces. As a result, the improvement in corrosion resistance depended on the bare substrate. Both films substantially improved the corrosion resistance of 316L. On Ti6Al4V, Ti-DLC gave only a small improvement, and W-DLC gave no clear benefit. The coatings were therefore most beneficial on 316L, whose passive film is less resistant to chloride attack.

1. Introduction

Thin films are a common way to protect metals from corrosion. The same film, however, can behave differently on different substrates. This substrate effect has received less attention than the film itself. It matters for 316L stainless steel and Ti6Al4V, two alloys widely used in load-bearing and blood-contacting medical devices. 316L has long been used for coronary stents [1,2], while Ti6Al4V is the most widely used titanium alloy for orthopedic implants [3]. Both alloys release metal ions in chloride-rich body fluids. These ions can trigger inflammatory responses. Metallic stents can also provoke a foreign-body inflammatory reaction that contributes to in-stent restenosis [4]. A protective surface layer is therefore needed.

Diamond-like carbon (DLC) is a leading candidate for this layer. It is an amorphous carbon network of sp² and sp³ bonds, which gives it high hardness, low friction, chemical inertness, and good biocompatibility [5,6,7,8,9,10]. Its properties can be tuned by the deposition method, the dopant, and the film architecture, such as interlayers and multilayers [9]. Our group has studied DLC thickness and interlayer design [11], internal and residual stress [12,13], substrate roughness [14], and precursor chemistry [15].

Doping is widely used to tailor DLC [16,17]. The corrosion resistance of DLC depends on its structure, porosity, and roughness [18,19,20]. It also depends on wettability, because a hydrophobic surface limits electrolyte contact [21]. Fluorine doping follows this route on 316 stainless steel and NiTi [21,22]. Copper doping improved the corrosion resistance of bronze and Ti6Al4V [23,24].

Titanium is a common dopant. It forms TiC nanocrystals, lowers residual stress, and improves adhesion to steel [25]. In co-doped films on 304 stainless steel, TiC crystallites lengthened the corrosion path and densified the surface [26]. On 316L, Ti-doped DLC with 5.33 at% Ti gave the lowest corrosion current and the highest polarization resistance, attributed to a TiO₂-rich surface [27]. Ti-doped DLC with a Ti interlayer also improved 316L [28]. On 304L, Ti-doped DLC (Ti-DLC) with a lower Ti content and a larger amorphous-carbon fraction gave the best wear and tribocorrosion resistance [29]. Titanium also has an optimal content range: too little or too much reduces protection [30].

The effect of tungsten doping is less clear. On nitrided steel in sulfuric acid, W-doped DLC (W-DLC) did not outperform plain DLC, likely because it grew columnar, more porous, and more hydrophilic [31]. In simulated seawater, however, W-DLC on bearing steel gave a more positive corrosion potential and a lower corrosion current than undoped DLC [32]. In artificial seawater, W-DLC with 0.9 at% W outperformed undoped DLC, whereas 3.1 and 15.1 at% W performed worse [19]. Gradient W-DLC also remained amorphous and protective on a WC-based cermet in 3.5 wt% NaCl under 20 MPa alternating hydrostatic pressure [33]. That coating, however, had an undoped DLC top layer, so its surface acted as a barrier. The benefit of tungsten therefore seems to depend on film structure and substrate, not on the dopant alone.

Across dopants, similar mechanisms are proposed: lower roughness, relieved stress, higher hydrophobicity, and passivating phases such as TiC, TiO₂, or SiO₂ [21,26,27,28,29,30,34]. Most treat the film as a barrier that keeps the electrolyte away from the substrate. In that case, the corrosion rate follows the substrate exposed at pores [20]. Metal doping, however, raises the conductivity of DLC by many orders of magnitude [35,36]. A conductive film may therefore act as an electrode rather than only a barrier. In that case, the electrochemical response should depend less on the substrate. Undoped DLC was reported to be more stable on Ti6Al4V than on 316L [37]. Whether this holds for conductive, metal-doped DLC and for different dopants is still unclear.

In this study, W-DLC and Ti-DLC films were sputtered with the same deposition parameters on 316L and Ti6Al4V. Because the resulting W and Ti contents differed, each film was compared with itself on the two substrates. The aim was to determine whether the substrate changes the wettability and structure of each film, its corrosion rate, and the mechanism that controls its corrosion.

2. Materials and Methods

2.1. Substrates and Film Deposition

Two substrates were used: 316L austenitic stainless steel and Ti6Al4V titanium alloy (Gredmann Group, Taipei, Taiwan) (Table 1 and Table 2). Plates were cut into 12 mm × 12 mm coupons and used as received to represent commercially supplied material. They were not ground or polished before deposition, so the films followed the as-received surface texture. Their average roughness (S_a) was 119 nm for 316L and 377 nm for Ti6Al4V. The influence of substrate topography on DLC surface roughness depends on film thickness [38]. It is treated here as part of the substrate effect. The coupons were ultrasonically cleaned in deionized water, ethanol, and isopropanol, then dried.

Table 1. Nominal chemical composition of 316L stainless steel (wt%).

C	Mo	Mn	Ni	Cr	Fe
≤0.03	2.5	2	10	18	Balance

Table 2. Nominal chemical composition of Ti6Al4V titanium alloy (wt%).

O	Fe	V	Al	Ti
≤0.2	≤0.25	4	6	Balance

The films were deposited by magnetron sputtering from W and Ti targets (99.95%). Argon (99.95%) was the sputtering gas, and C₂H₂ was the carbon precursor. Before deposition, the substrates were cleaned in an Ar plasma (20 sccm, −180 V bias, 10 min). A Cr adhesion interlayer was then deposited for both films from a Cr target (99.95%). The main parameters were a base pressure of 4.5 × 10⁻⁵ torr, a working pressure of 5 × 10⁻³ torr, a target-to-substrate distance of 85 mm, a rotation of 10 rpm, a DC power of 100 W, no substrate bias, and a C₂H₂/(Ar + C₂H₂) flow ratio of 18%. The substrates were not heated. The deposition time was adjusted to give a film thickness of about 1.5 µm. For each film, both substrates were coated in the same run.

Cross-sectional scanning electron microscopy (SEM) gave mean film thicknesses of 1.51 ± 0.08 µm (W-DLC/316L), 1.57 ± 0.03 µm (W-DLC/Ti6Al4V), 1.49 ± 0.11 µm (Ti-DLC/316L), and 1.39 ± 0.07 µm (Ti-DLC/Ti6Al4V) in the pre-test sections (Section 3.4). Six specimens were tested: the two bare substrates and each film on each substrate (W-DLC/316L, W-DLC/Ti6Al4V, Ti-DLC/316L, and Ti-DLC/Ti6Al4V). Figure 1 summarizes the workflow.

Figure 1. Experimental workflow for the preparation and evaluation of W-DLC and Ti-DLC films on 316L and Ti6Al4V.

2.2. Surface Characterization

Surface and cross-sectional morphology were examined by SEM before and after the corrosion test, and composition was measured by energy-dispersive X-ray spectroscopy (EDX). Pre-test cross-sections were taken from a second specimen coated in the same run. For cross-sections, each specimen was cut and hot-mounted. The mounted specimens were ground with SiC papers from 240 to 4000 grit and polished to a mirror finish with diamond suspension on a polishing cloth. Wettability was measured by the sessile-drop method with distilled water (five 5 µL droplets, 23 ± 1 °C, read after 5 s). Because the apparent contact angle depends on roughness [39], it is used only to compare specimens.

Raman spectra were recorded from 1000 to 2000 cm⁻¹. The broad DLC band was fitted with two Gaussians: the D band at about 1350–1400 cm⁻¹ and the G band at about 1540–1580 cm⁻¹ [40]. A cosmic-ray spike was removed before fitting. A linear baseline was fitted simultaneously with the two bands, and ID/IG was taken as the ratio of the integrated areas of the D and G bands. ID/IG does not give the sp³ fraction directly.

2.3. Electrochemical Corrosion Testing

Corrosion tests used 3.5 wt% NaCl, deaerated with N₂ for 30 min, in a three-electrode cell. This standard chloride medium was chosen to compare the films and substrates; the results are not intended to predict in vivo behavior. The specimen (1 cm² exposed area) was the working electrode, Ag/AgCl the reference electrode, and a graphite rod the counter electrode. After 30–60 min at open circuit, electrochemical impedance spectroscopy (EIS) was measured at the open-circuit potential with a 10 mV amplitude from 100 kHz to 1 Hz (51 points). Spectra were fitted in ZView (version 3.1, Scribner Associates Inc., Southern Pines, NC, USA) The arcs did not close at 1 Hz, so the polarization resistance (R_p) was obtained by extrapolating the fitted model and is used as supporting evidence.

Potentiodynamic polarization was performed on the same specimen, from −0.2 V versus the open-circuit potential to +1.0 V versus Ag/AgCl at 1 mV/s. The corrosion potential (E_corr) and corrosion current density (i_corr) were obtained by Tafel extrapolation over 50–150 mV from E_corr [41]. With Tafel slopes of about 200 mV/dec, this window covered only about half a decade of current. The i_corr values are therefore compared between specimens rather than used as absolute corrosion rates. All potentials are given versus Ag/AgCl. Each condition was measured once, so the uncertainties reported in the tables are fitting errors.

3. Results

3.1. Surface Composition

EDX confirmed the intended dopant in each film on both substrates (Figure 2, Table 3). No Fe or Ni was detected in the analyzed areas, so the films covered the substrates there. W-DLC contained about 9 at% W on 316L and 7 at% W on Ti6Al4V. Ti-DLC contained far more dopant, about 23 at% Ti on 316L and 26 at% Ti on Ti6Al4V. Both films used a Cr interlayer, so the Ti signal comes mainly from the film. On Ti6Al4V, the substrate may add a little, although no Al or V was detected. Cr from the interlayer was detected only under Ti-DLC (0.3–0.4 at%). In W-DLC, the high W content (54–59 wt%) raises the mean atomic number and density of the film, which probably shortened the electron range, so the interlayer signal was not detected under W-DLC.

Figure 2. EDX spectra of the doped DLC films: (a) W-DLC/316L; (b) W-DLC/Ti6Al4V; (c) Ti-DLC/316L; (d) Ti-DLC/Ti6Al4V.

Table 3. Elemental composition of the doped DLC films determined by EDX.

W-DLC/316L			W-DLC/Ti6Al4V
Element	Weight%	Atomic%	Element	Weight%	Atomic%
C K	39.79	90.46	C K	44.57	91.94
Ar K	1.11	0.76	Ar K	1.21	0.75
W M	59.10	8.78	W M	54.21	7.31
Total	100.00		Total	100.00
Ti-DLC/316L			Ti-DLC/Ti6Al4V
Element	Weight%	Atomic%	Element	Weight%	Atomic%
C K	44.02	75.75	C K	40.80	73.25
Ar K	1.42	0.74	Ar K	1.52	0.82
Ti K	53.60	23.13	Ti K	56.87	25.60
Cr K	0.97	0.38	Cr K	0.81	0.33
Total	100.00		Total	100.00

Both dopant levels are high compared with optimized films. The Ti content of Ti-DLC (23–26 at%) is far above the 3–6 at% at which Ti-DLC gave the best corrosion resistance [25,27,30]. At such levels, Ti-DLC usually forms a TiC/amorphous-carbon nanocomposite [26,29,30], which tends to protect less. A Ti-DLC with 26.8 at% Ti on 304L gave the highest corrosion current in its series [29], and one with 15.24 at% Ti corroded faster than bare 316L in simulated body fluid [27].

The W content of W-DLC (7–9 at%) is lower. Sputtered W-DLC often contains W₂C or WC_1−x nanocrystals, at least at 10–20 at% W [31,42]. Reported effects on corrosion differ. W-DLC with 3.1–15.1 at% W protected worse than undoped DLC in artificial seawater [19], whereas W-DLC with 25–50 wt% W protected 100Cr6 steel better in 3.5 wt% NaCl [32].

These phases were not confirmed in our films, so the reported structures are used only as a guide when interpreting their corrosion behavior. For each film, the dopant content was similar on the two substrates, so differences between substrates are not due to composition.

3.2. Surface Wettability

Figure 3 shows the water contact angles. Bare 316L had the highest contact angle (81°), and bare Ti6Al4V the lowest (65°). All surfaces were hydrophilic (below 90°). Because only water was used, the results show relative wettability, not surface free energy [43]. The low angle of Ti6Al4V reflects its polar TiO₂ surface, and its higher roughness may add to it [39]. All four films fell between these limits (68–79°). Each film therefore lowered the contact angle on 316L and raised it on Ti6Al4V.

Figure 3. Water contact angles of the bare substrates and of W-DLC and Ti-DLC on 316L and Ti6Al4V. Error bars show the standard deviation of five droplets.

The two films responded differently to the substrate. W-DLC had nearly the same contact angle on both substrates (75° and 73°), within the droplet scatter. Ti-DLC was clearly less wettable on Ti6Al4V (79°) than on 316L (68°).

3.3. Raman Spectroscopy

Both films showed broad D and G bands on both substrates (Figure 4, Table 4). The G band lay at 1552–1555 cm⁻¹ and was about 200 cm⁻¹ wide, indicating a highly disordered carbon network. ID/IG ranged from 1.47 to 1.56, with fitting uncertainties of about ±0.3. Because the broad D band overlaps the background, ID/IG depends on the baseline and is compared only between spectra fitted in the same way. For each film, the band positions on the two substrates differed by less than 6 cm⁻¹, and the bandwidths and ID/IG agreed within the fitting uncertainty. The substrate therefore had no significant effect on the bonding structure of either film.

Figure 4. Raman spectra with fitted D and G bands: (a) W-DLC/316L; (b) W-DLC/Ti6Al4V; (c) Ti-DLC/316L; (d) Ti-DLC/Ti6Al4V.

Table 4. Raman fitting parameters of W-DLC and Ti-DLC on 316L and Ti6Al4V.

Specimen	D Pos. (cm−1)	D FWHM (cm−1)	G Pos. (cm−1)	G FWHM (cm−1)	ID/IG
W-DLC/316L	1386.4 ± 11.9	336.3 ± 16.5	1554.2 ± 1.5	197.2 ± 6.1	1.52 ± 0.32
W-DLC/Ti6Al4V	1383.8 ± 11.4	342.2 ± 15.9	1552.7 ± 1.3	200.0 ± 5.6	1.47 ± 0.29
Ti-DLC/316L	1388.0 ± 13.8	354.3 ± 19.0	1555.0 ± 1.4	202.0 ± 7.4	1.56 ± 0.36
Ti-DLC/Ti6Al4V	1393.3 ± 10.4	351.9 ± 12.9	1552.0 ± 1.0	204.4 ± 5.4	1.49 ± 0.26

Note: pos., band position; FWHM, full width at half maximum. Uncertainties are standard errors of the peak fits.

3.4. Morphology Before and After Corrosion

Figure 5 and Figure 6 compare each film on the two substrates before and after the test. No pits or delamination were found on any film after testing. The bare substrates also showed no resolvable pits in the imaged areas. Bare 316L, however, reached several mA cm⁻² above about 0.8 V during polarization (Section 3.5). Its attack at high potential was therefore probably shallow and uniform or located outside the imaged areas.

Figure 5. Surface SEM images before (left) and after (right) the corrosion test: (a,b) bare 316L; (c,d) bare Ti6Al4V; (e,f) W-DLC/316L; (g,h) W-DLC/Ti6Al4V; (i,j) Ti-DLC/316L; (k,l) Ti-DLC/Ti6Al4V.

Figure 6. Cross-sectional SEM images before (left) and after (right) the corrosion test: (a,b) bare 316L; (c,d) bare Ti6Al4V; (e,f) W-DLC/316L; (g,h) W-DLC/Ti6Al4V; (i,j) Ti-DLC/316L; (k,l) Ti-DLC/Ti6Al4V.

The substrate changed the surface morphology of Ti-DLC the most. On 316L, the film followed the grain relief of the steel and showed fine crack-like lines already before testing (Figure 5i,j). On Ti6Al4V, it was covered with discrete bright particles of about 1 µm (Figure 5k,l). This particle-rich surface was found only for Ti-DLC/Ti6Al4V. It may come from Ti-rich clusters or from nodular growth on the rougher substrate. W-DLC followed the substrate texture on both substrates and was nearly unchanged after testing (Figure 5e–h).

In cross-section (Figure 6), Ti-DLC remained continuous on both substrates. The surface lines of Ti-DLC on 316L did not extend through the film (Figure 6j). W-DLC also covered both substrates, but gaps appeared along the film/resin boundary, most clearly on Ti6Al4V (Figure 6g,h). These were present before testing and most likely formed during mounting and polishing, so they are not corrosion damage. The gap at the edge of the bare 316L (Figure 6a,b) was also present before testing.

3.5. Potentiodynamic Polarization

Figure 7 shows the polarization curves, and Table 5 lists the fitted parameters. The two bare substrates differed greatly in corrosion resistance. Bare 316L had the lowest E_corr and the highest i_corr. Bare Ti6Al4V corroded much more slowly, consistent with its more protective native oxide. The coupons were not polished, and crevice attack at the cell seal cannot be excluded.

Figure 7. Potentiodynamic polarization curves in 3.5 wt% NaCl: (a) W-DLC and (b) Ti-DLC on 316L and Ti6Al4V, with the bare substrates for reference.

Table 5. Polarization parameters in 3.5 wt% NaCl.

Specimen	Ecorr (V)	βa (mV/dec)	βc (mV/dec)	icorr (A cm−2)	icorr Ratio (Substrate/Film)
316L	−0.483	194 ± 2	208 ± 3	1.87 × 10−5	—
Ti6Al4V	−0.206	226 ± 2	169 ± 2	8.89 × 10−7	—
W-DLC/316L	−0.159	229 ± 2	167 ± 1	9.43 × 10−7	20
W-DLC/Ti6Al4V	−0.112	203 ± 2	213 ± 2	1.14 × 10−6	0.8
Ti-DLC/316L	−0.108	208 ± 2	169 ± 2	3.34 × 10−7	56
Ti-DLC/Ti6Al4V	+0.036	220 ± 4	179 ± 2	3.16 × 10−7	2.8

Note: β_a and β_c, anodic and cathodic Tafel slopes; i_corr ratio, i_corr of the bare substrate divided by that of the coated specimen. Uncertainties of β_a and β_c are standard errors of the Tafel fits.

Each film raised E_corr on both substrates, and its polarization curves on the two substrates were similar. The substrate changed its i_corr by no more than about 20% and hardly affected its Tafel slopes (Table 5). All specimens, including the bare substrates, gave Tafel slopes well above those expected for simple charge-transfer control, so the slopes were not used to identify the controlling reaction. The corrosion current density of each film was therefore considered nearly independent of the substrate. The reduction in i_corr, however, depended strongly on the substrate. On 316L, both films lowered i_corr by more than an order of magnitude because the passive film of the steel is less stable in chloride. On Ti6Al4V, the reduction was small or absent because the substrate already has a protective oxide.

3.6. Electrochemical Impedance Spectroscopy

Table 6 lists the parameters obtained with the equivalent circuit in Figure 8, and Figure 9 shows the Nyquist plots. Both films gave larger arcs than the bare substrate. For each film, the arcs on the two substrates were of similar size.

Table 6. Equivalent-circuit parameters from EIS fitting in 3.5 wt% NaCl.

Specimen	Rs (Ω cm2)	Q (S sn cm−2)	n	Ceff (µF cm−2)	Rp (Ω cm2)	χ2
316L	48.88	2.42 × 10−4	0.548	6.2	7165 ± 2200	1.1 × 10−3
Ti6Al4V	49.42	1.47 × 10−5	0.879	5.4	15,818 ± 410	1.9 × 10−4
W-DLC/316L	31.46	1.45 × 10−5	0.786	1.8	30,014 ± 6300	5.3 × 10−3
W-DLC/Ti6Al4V	34.3	9.76 × 10−6	0.806	1.4	23,328 ± 3900	9.3 × 10−3
Ti-DLC/316L	36.01	1.28 × 10−5	0.803	1.9	37,455 ± 10,100	7.3 × 10−3
Ti-DLC/Ti6Al4V	57.5	1.27 × 10−5	0.667	0.35	46,216 ± 13,700	8.2 × 10−3

Note: Uncertainties of R_p are standard errors of the equivalent-circuit fit; χ², chi-squared value of the fit.

Figure 8. Equivalent circuit used to fit all EIS spectra.

Figure 9. Nyquist plots in 3.5 wt% NaCl: (a) W-DLC and (b) Ti-DLC on 316L and Ti6Al4V. Insets show the low-impedance region.

All spectra showed a single, incomplete, flattened arc, with no distinct second time constant in the measured range. They were therefore fitted with the single-time-constant circuit in Figure 8. In this circuit, R_s is the solution resistance, R_p the polarization resistance, and Q and n the coefficient and exponent of the constant-phase element (CPE). A two-CPE circuit did not improve the fit, and its extra parameters were not resolved. The simpler single-CPE circuit was therefore adopted. The same circuit was used for all specimens so that their parameters could be compared directly. For the bare substrates, the CPE and R_p describe the native passive film; for the coated specimens, they describe the film surface. Because R_p was obtained by extrapolation below 1 Hz, its value depends on the model. A second time constant below 1 Hz cannot be excluded. The effective capacitance (C_eff) was calculated from the CPE parameters with the Brug formula [44]:

$$C_{\mathrm{e}\mathrm{f}\mathrm{f}}=Q^{{\displaystyle \frac{1}{n}}}{\left({\displaystyle \frac{1}{R_{\mathrm{s}}}}+{\displaystyle \frac{1}{R_{\mathrm{p}}}}\right)}^{{\displaystyle \frac{n-1}{n}}}$$

(1)

where Q is the CPE coefficient and n is the CPE exponent. C_eff expresses the non-ideal CPE response as an equivalent ideal capacitance. The Brug formula assumes a distribution of time constants along the electrode surface. For a dielectric film, whose properties vary normal to the surface, the power-law model is more appropriate, but it requires the resistivity profile of the film [45]. C_eff from Equation (1) is therefore compared with the capacitance of a dielectric film only in order of magnitude. All fits gave a chi-squared value (χ²) below 10⁻².

The fitted R_p followed the polarization results, and bare 316L had the lowest R_p. Bare Ti6Al4V, however, had only about twice the R_p of bare 316L, although its i_corr was much lower. The arcs of the more resistive specimens were far from closed at 1 Hz, so their extrapolated R_p is likely underestimated. R_p was therefore used only to compare trends. For each film, R_p on the two substrates overlapped within the fitting uncertainty. Relative to the bare substrate, both films raised R_p more on 316L than on Ti6Al4V. On Ti6Al4V, the gain was moderate for Ti-DLC and small for W-DLC. The small gain for W-DLC/Ti6Al4V depended on the extrapolation, so this specimen showed no clear improvement. The surface lines of Ti-DLC/316L did not lower R_p below that of the bare substrate. There was therefore no electrochemical evidence that they formed continuous paths to the substrate during the test.

The effective capacitance helps to distinguish barrier behavior from conductive behavior. The coated specimens gave C_eff values in the µF cm⁻² range (Table 6). An insulating carbon layer of the same thickness would give only a few nF cm⁻² [46], two to three orders of magnitude less. The EIS response was therefore inconsistent with the expected insulating barrier layer. The high capacitance most likely reflects the electrical conductivity of the metal-containing films [35,36], although electrolyte-filled defects and surface oxides may also contribute. The capacitance therefore arises mainly at the film surface, for which the Brug formula is appropriate. In sputtered Ti-DLC, only about 3 at% Ti lowered the resistivity by more than five orders of magnitude [25]. TiC-rich Ti-DLC has been reported to be more conductive and less protective than lightly doped films [27]. A thin surface oxide could also give µF-range values. On Ti-DLC, such capacitances were attributed to a TiO₂-rich passive film only a few nanometers thick [27]. In both cases, the response differs from that of an insulating layer, and the film surface likely takes part in the charge-transfer reactions. This held even for W-DLC, which contained only 7–9 at% W.

The CPE exponent was well below 1 for all specimens, reflecting rough and heterogeneous surfaces (Table 6). It was lowest for bare 316L, consistent with a heterogeneous, partly active surface. For W-DLC, it was similar on both substrates. For Ti-DLC, it was lower on Ti6Al4V than on 316L, consistent with its particle-rich surface. A low n also lowers the C_eff obtained from Equation (1). With the n of Ti-DLC/316L, Ti-DLC/Ti6Al4V would give a C_eff similar to those of the other coated specimens. Its low C_eff therefore reflects its heterogeneous surface, not a lower conductivity. For this reason, C_eff values are compared here only in order of magnitude.

4. Discussion

Two limiting cases help to interpret the results. Barrier coatings, such as undoped hydrogenated DLC, are highly resistive and block charge transfer [5,47,48]. Their performance is limited by pores through which the electrolyte reaches the substrate [20]. At sufficient metal content, DLC becomes electrically conductive [5,35,36], and the film surface itself can then act as an electrode. Real films lie between these two limits, because pores and surface defects can still expose the substrate. The high C_eff values indicate that W-DLC and Ti-DLC lie close to the conductive limit. This behavior is discussed below in three respects.

First, the corrosion rate was governed mainly by the film. For each film, i_corr and R_p were nearly the same on both substrates, although the bare substrates differed greatly in corrosion rate. A barrier coating would instead follow the substrate exposed at its pores. Under such a model, coated Ti6Al4V would show a much lower corrosion current density than coated 316L, as the bare substrates do. This substrate dependence was not observed for either film. Even if all the current on 316L came from exposed steel, the ratio of the coated to bare corrosion current densities implies an exposed area of several percent, far larger than the defects observed by SEM.

Second, the substrate determines how much the film improves corrosion resistance. Because the film mainly governs the corrosion rate, the improvement depends mainly on how fast the bare substrate corrodes. 316L has a thin Cr₂O₃-based passive film that is prone to chloride breakdown [49], so it gained the most from the films. Ti6Al4V is already protected by a stable TiO₂-based film [50], so the films added little. Ti-DLC coatings on Ti6Al4V even showed a lower polarization resistance than the bare substrate in phosphate-buffered solution [51]. Other studies follow the same pattern. The improvement from a doped DLC film is largest on active magnesium alloys [52], smaller on stainless steel [19], and smallest on passive titanium alloys [51].

Third, the substrate can still change the surface properties of the film, such as its wettability, corrosion potential, and morphology. On Ti6Al4V, Ti-DLC had a more positive corrosion potential, a less wettable surface, and a more heterogeneous surface. Its corrosion rate, however, stayed the same. Because the corrosion potential is a mixed potential, a small exposed area can shift it without noticeably changing the corrosion rate. A shift to a more positive corrosion potential without a lower corrosion rate has also been reported for Ti-rich DLC [30]. W-DLC was hardly affected. The two films also differed in dopant content, so this difference cannot be assigned to the dopant type alone. The oxides on the film and at defects offer a likely explanation. Ti in the film can form TiO₂ at the surface [27,28,29]. Defects that penetrate both the film and the Cr interlayer would expose Ti6Al4V, which forms a similar TiO₂-based film. Exposed 316L instead carries a Cr-rich passive film that is less stable in chloride [49]. Such sites can act as anodes, as reported for Ti/Ti-DLC on magnesium alloy [52]. They shift the mixed potential most when the film itself is passive, as TiO₂-covered Ti-DLC is likely to be. Tungsten dissolves much faster in neutral and alkaline solutions than in acid, because its oxide is not stable there [53]. The W-DLC surface therefore probably stays more active. Its corrosion potential is then less sensitive to the exposed substrate. This interpretation needs to be verified by surface analysis.

Metal-doped DLC can behave as a conductive surface, as found here. Its corrosion rate then depends little on the substrate, so it protects most on substrates whose passive film breaks down in chloride, such as 316L stainless steel. On passive substrates such as Ti6Al4V, the film is better justified by other functions, such as the tribocorrosion resistance reported for Ti-DLC on this substrate [51]. Metal doping also supports such functions: it relieves stress [16,25] and improves wear resistance [30,32]. Long-term durability was not tested here. It depends strongly on the interlayer: DLC-coated implants have failed in vivo by slow crevice corrosion of the interlayer [10], and interlayer choice controls the stability of DLC on 316L [54].

5. Conclusions

W-DLC and Ti-DLC films were sputtered with the same deposition parameters on as-received 316L and Ti6Al4V, and each film was compared with itself on the two substrates. Because the W and Ti contents differed, the differences between the two films reflect both dopant type and dopant content. The main conclusions are as follows:

(1)

The corrosion rate of each film was nearly independent of the substrate. High effective capacitances indicated that both metal-doped films behaved mainly as conductive electrodes rather than insulating barriers.

(2)

The benefit of each film depended on the bare substrate. Both films strongly improved the corrosion resistance of 316L, whose passive film is vulnerable to chloride. On Ti6Al4V, whose passive film is stable in chloride, W-DLC gave no clear benefit and Ti-DLC only a small one.

(3)

The substrate changed the surface state of Ti-DLC but not its corrosion rate. On Ti6Al4V, Ti-DLC had a more positive corrosion potential and a less wettable surface. This may be because titanium oxide forms both on the film and on the substrate exposed at defects that penetrate the interlayer.