- Email: [email protected]
Surface structure and corrosion resistance of short-time heat-treated NiTi shape memory alloy
Applied Surface Science 257 (2010) 1573–1582
Contents lists available at ScienceDirect
Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc
Surface structure and corrosion resistance of short-time heat-treated NiTi shape memory alloy D. Vojtˇech ∗ , M. Vodˇerová, J. Fojt, P. Novák, T. Kubásek Department of Mecotals and Corrosion Engineering, Institute of Chemical Technology, Prague, Technická 5, 166 28 Prague 6, Czech Republic
a r t i c l e
i n f o
Article history: Received 8 April 2010 Received in revised form 17 August 2010 Accepted 21 August 2010 Available online 20 September 2010 Keywords: NiTi alloy Heat treatment Oxidation Corrosion Shape memory alloy
a b s t r a c t NiTi alloys are attractive materials that are used for medicine, however, Ni-release may cause allergic reactions in an organism. The Ni-release rate is strongly affected by the surface state of the NiTi alloy that is mainly determined by its processing route. In this study, a NiTi shape memory alloy (50.9 at.% Ni) was heat-treated by several regimes simulating the shape setting procedure, the last step in the manufacture of implants. Heating temperatures were between 500 and 550 ◦ C and durations from 5 to 10 min. Heat treatments were performed in air at normal and low pressure and in a salt bath. The purpose of the treatments was to obtain and compare different surface states of the Ni–Ti alloy. The surface state and chemistry of heat-treated samples were investigated by electron microscopy, X-ray photoelectron spectroscopy and Raman spectrometry. The amount of nickel released into a model physiological solution of pH 2 and into concentrated HCl was taken as a measure of the corrosion rate. It was found that the heat treatments produced surface TiO2 layers measuring 15–50 nm in thickness that were depleted in nickel. The sample covered by the 15-nm thick oxide that was treated at 500 ◦ C/5 min in a low pressure air showed the best corrosion performance in terms of Ni-release. As the oxide thickness increased, due to either temperature or oxygen activity change, Ni-release into the physiological solution accelerated. This finding is discussed in relation to the internal structure of the oxide layers. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Nearly equi-atomic NiTi alloys (nitinol) have attracted much attention due to their shape memory effect, superelastic behavior, high tensile strength, good corrosion resistance and biocompatibility. These characteristics make NiTi alloys attractive for biomedical applications, such as stents, which are generally utilized to restore a damaged blood vessel or an oesophagus, among others. NiTi alloys are generally regarded as highly corrosion resistant, similar to stainless steel or titanium. The reason is the spontaneous formation of a native passive TiO2 layer on the surface. Even a weak oxidizing environment, such as water, air or humidity, in contact with nitinol at low temperature is sufficient to produce a few nanometers of a passive protective layer. However, in the processing of NiTi alloys they experience various forming, heat- or surface-treating steps, which strongly influence the surface structure, chemistry and, therefore, corrosion performance. In a human body, various fluids may come into contact with NiTi implants. In particular, serious problems may arise when a strongly acidic fluid, like gastric juice, attacks the surface of the alloy. In general, the corrosion of nitinol has two aspects. First, cor-
∗ Corresponding author. Tel.: +420 220 444290. E-mail address: [email protected] (D. Vojtˇech). 0169-4332/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2010.08.097
rosion is accompanied with nickel release from an implant into the surrounding body fluid and tissue, which may enhance an allergic reaction in a sensitive organism. Second, corrosion may cause pitting and a reduction of mechanical performance of an implant. In an extreme case, an implant broken due to corrosion may produce sharp fragments that are dangerous for surrounding tissue [1]. For these reasons, the corrosion of nitinol has become an important issue in biomedical engineering during the last decade. To slow down this corrosion process, a highly protective adherent layer with a low concentration of nickel and internal defects must cover the surface of nitinol. A number of oxidation treatment procedures have been reported so far to support the preferential Ti oxidation and increase the protective effect against corrosion. These include thermal oxidation either in air or under a low oxygen pressure, etching, passivation in nitric acid, anodization, electrochemical polishing, oxygen ion implantation, laser oxidation, hydrothermal oxidation in water, sol–gel techniques, cathodic deposition and others [2–9]. Most often, the last processing step in the manufacture of nitinol stents is the shape setting. The shape setting treatment means a short-time heating of nitinol at around 500 ◦ C and serves to achieve a desired shape and superelastic behavior of an implant. Heating can be performed in air, argon, vacuum or a salt bath. In oxidizing environments, titanium in nitinol generally oxidizes preferentially over nickel due to the higher thermodynamic stability of TiO2 as
1574
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
Table 1 Summary of heat-treatment (HT) regimes used in this study and their designation. Sample designation
Heat-treatment regime
CHE
No HT, chemical etching in a 1HF + 4HNO3 + 5H2 O (by volume) solution HT at 510 ◦ C/10 min in air at a normal pressure HT at 510 ◦ C/5 min in a molten salt bath HT at 500 ◦ C/5 min in air at a low pressure of 10 Pa HT at 550 ◦ C/5 min in air at a low pressure of 10 Pa
NP S 500LP 550LP
compared to NiO [3]. At sufficiently high temperatures, titanium reacts rapidly even with traces of oxygen present in a surrounding environment. As a result, a TiO2 -enriched and Ni-depleted surface oxide layer forms [10–13]. The corrosion performance of a nitinol implant thus depends on properties of such an oxide layer, particularly, on its thickness, chemistry, adherence and internal structure. In the literature, there are few studies on the influence of the shape setting treatment on corrosion behavior. For this reason, our study is concerned with short-time heat treatments of nitinol at around 500 ◦ C. Treatments were carried out in various environments in order to simulate different oxidation potentials and the purpose was to prepare and compare various surface states of the Ni–Ti alloy. Corrosion tests were performed in a relatively aggressive solution with pH 2. The solution simulates gastric fluid, so that information obtained by these tests is important for implants, such as oesophageal stents, for example. The fracture of a stent in a patient’s body due to corrosion may cause serious problems [14,15].
spectroscope (XPS) ESCA Probe P (a pressure in the analytical chamber of 2 × 10−8 Pa, monochromatic Al K␣ X-ray source, binding energy calibration with respect to the energy of Au 4f7/2 peak, Ar pressure during sputter depth profiling 3 × 10−3 Pa, energy of Ar+ ions during sputtering 5 keV, calibration of the sputtering rate with respect to titanium dioxide). Chemical compositions were calculated from areas of deconvoluted elemental peaks by analytical software CasaXPS version 2.3.14. The thickness of TiO2 layer was estimated as a thickness at which a Ti:O ratio is 1:2 or lower. A Ti:O ratio lower than 1:2 indicates oxygen adsorption on the surface TiO2 . Detailed spectra of Ti 2p and Ni 2p were recorded using a constant pass energy of 20 eV at a 0.1-eV step size. The corrosion behavior of both chemically etched and heattreated samples was determined by the immersion test. Three samples for each surface state were tested. Samples were exposed to a simulated physiological solution containing 9 g/l NaCl with pH 2 (adjusted by the addition of HCl) at a temperature of 37 ◦ C. The low pH value was used in this experiment because it simulates a gastric juice environment. The immersion period was 168 h. Simultaneously, the corrosion behavior was also determined in a concentrated HCl. Samples were exposed to this strongly aggressive environment for 6 h at 20 ◦ C. In both corrosion tests, the corrosion rate was taken as the concentration of nickel released into corrosive medium. Nickel concentrations in the simulated physiological solution and in the concentrated HCl were measured by ICP mass spectrometry (ELAN 6000) and by atomic absorption spectrometry (GBC 932 plus), respectively.
2. Experimental
3. Results and discussion
NiTi alloy (50.9% Ni; hereafter, all concentrations are in at.% unless otherwise stated) was used for heat treatments in our experiments. The 0.45-mm-thick wire with a tensile strength of 1580 MPa and Af of 25 ◦ C was provided by an industrial supplier. According to the supplier specifications, the wire was cold drawn (45% deformation) and annealed. Detailed specifications of the annealing conditions were not given. The NiTi wire was first subjected to chemical etching in a 1HF + 4HNO3 + 5H2 O (by volume) solution in order to remove traces of previous processing steps and to obtain a defined and unified surface state. Prior to heat treatments, the surfaces of all samples were carefully washed and dried. In this study, heat treatment regimes were selected to reflect real conditions of the shape setting procedures used by manufacturers of medical stents. Therefore, they were performed at different temperatures between 500 and 550 ◦ C in air both at ambient pressure and at a low pressure of 10 Pa and also in a molten salt bath containing a mixture of 60% KNO3 and 40% NaNO3 . Low pressures are sometimes used to prevent excessive oxidation of nitinol. The advantage of the shape setting in salt baths is the fast heat transfer from the bath to the treated wire and therefore the short time needed for the shape setting. After each heat treatment, the wire was immediately quenched into cold water. Detailed descriptions of heat treatment procedures applied in our experiment and their designation are given in Table 1. All heat-treated samples appeared in characteristic interference colors ranging from light golden to a dark blue color. The surface morphology of samples was observed by scanning electron microscopy (SEM, Hitachi S4700, an acceleration voltage of 10 kV, in the regime of secondary electrons). Due to the very low thickness of surface layers on samples, the phases present were not detected by X-ray diffraction, but a Raman spectrometer Labram HR (excitation source – laser with a wavelength of 532 nm operating at 40 mW) was used for this purpose. Additionally, the chemical composition of the surface and elemental profiling in the sub-surface regions of samples were characterized by an X-ray photoelectron
3.1. Structure, chemical and phase composition Fig. 1a presents a SEM view of the CHE surface. One can see than the surface is not flat. Instead, chemical etching causes the formation of small dimples of a few micrometers in size on the surface. There are also scratch marks resulting from the cold drawing process in which spherical or elongated non-metallic inclusions are observed. These inclusions consisting mainly of titanium carbides originate from the melting process of nitinol and from its contamination by the melting crucible. Oxidation in air at 510 ◦ C/10 min modifies the surface structure, as shown in Fig. 1b. In contrast to the CHE sample, the oxide layer formed at high temperature has a typical grainy morphology with a grain size of about 20 nm. It has been demonstrated in a number of studies [16] that surface treatments of nitinol performed at room or slightly higher temperatures generally produce amorphous surface oxides. The reason is that slow diffusivities of species prevent the formation of ordered crystalline arrangements of atoms. At about 500 ◦ C diffusion becomes sufficiently faster to produce crystalline oxide layers that are more thermodynamically stable. The surface morphologies of the S and 500LP samples presented in Fig. 1c and d are almost identical to the previous one because they consist of fine grains of about 20 nm in size. It appears that oxygen activity, which is different in the treatments illustrated by Fig. 1b–d, does not influence the oxide grain size significantly. The main factor affecting the grain size is heattreatment temperature, which is demonstrated in Fig. 1e where a view of the 550LP sample is shown. Although the oxygen partial pressure was identical to that of the 500LP, the grain size grew considerably from approximately 20 to50 nm as the temperature increased by 50 ◦ C. It was already suggested in our previous work [11] that nanometer-sized titania grains grow relatively fast above 500 ◦ C due to the high driving force of this growth. Survey XPS spectra of the investigated samples are almost identical, and they mainly contain characteristic peaks of Ti, O, and contamination C; see Fig. 2 for illustration. As far as nickel
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
1575
Fig. 1. Surface morphologies of etched and heat-treated samples: (a) CHE, (b) NP, (c) S, (d) 500LP, and (e) 550LP (SEM). For sample designation see Table 1.
release from nitinol is concerned, knowledge of surface chemistry is important. Surface chemical compositions (excluding contamination carbon) are given in Table 2. High concentrations of oxygen,
Table 2 Surface chemical compositions (in at.%) of the etched and heat-treated samples (XPS). For sample designation see Table 1. Sample
CHE NP S 500LP 550LP
Element concentration (at.%) Ni
Ti
O
0.5 2.4 1.5 0 0
22.0 16.2 10.7 15.5 11.2
77.5 81.4 87.8 84.5 88.8
much exceeding that in TiO2 , result from its adsorption on the surface from the surrounding atmosphere. In the case of the S sample treated in the salt bath, the oxygen probably originates mainly from a decomposition of nitrate to nitrite at the alloy surface, which may be written as NO3 − → NO2 − + O
(1)
The O-rich layer on the CHE sample results from two simultaneous chemical reactions. The first one includes the dissolution of the TiO2 native layer in hydrofluoric acid to form soluble fluoride complexes: TiO2 + 6F− + 4H+ → [TiF6 ]2− + 2H2 O
(2)
The second process is preferential oxidation of titanium by nitric acid back to titanium dioxide, which is accompanied by the inward
1576
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
diffusion of nickel: Ti + 4NO3 − + 4H+ → TiO2 + 4NO2 + 2H2 O
Fig. 2. Survey XPS spectrum of the NP sample showing Ti, O, Ni and contamination C peaks.
(3)
Regarding Ni concentrations, one can see in Table 2 that these concentrations are low in all samples and do not exceed a few atomic percents. The highest Ni amount is detected on the NP sample, whereas the lowest are on both 500LP and 550LP ones. The nickel depletion on the oxidized surface results from preferential oxidation of titanium due to the high thermodynamic stability of TiO2 [3]. Generally, the oxidation processes of Ti-rich intermetallics are controlled by oxygen inward diffusion through the surface oxide. At the oxide/metal interface, the preferential oxidation of Ti produces a new titania, which is accompanied by nickel inward diffusion. It was demonstrated in several studies that sufficiently long oxidation at high temperatures may even produce Ni-enriched zones beneath the oxide [10]. Such zones may consist of Ni3 Ti, Ni4 Ti and even pure Ni. Therefore, they represent Ni reservoirs that may enhance nickel release, particularly in the case of defect occurrence in the surface oxide. Fortunately, as will be shown in Fig. 3, significant nickel accumulation beneath the oxide is not observed in this study, probably due to short oxidation periods.
Fig. 3. Elemental profiles through thin sub-surface regions on the investigated samples (XPS): (a) CHE, (b) 500LP, (c) 550LP, (d) S, and (e) NP. Estimated thickness of TiO2 is indicated by dashed lines. The TiO2 layer corresponds to a Ti:O ratio of 1:2 or lower. A Ti:O ratio lower than 1:2 indicates oxygen adsorption on the surface TiO2 .
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
1577
Fig. 4. Detailed Ti 2p spectra acquired on the surface of the samples: (a) CHE, (b) NP, (c) S, (d) 500LP, and (e) 550LP (XPS). All spectra are dominated by the Ti4+ oxidation state.
Fig. 3 shows elemental profiles measured in thin sub-surface regions by XPS. As expected, the thinnest oxide layer covers the surface of the etched sample (CHE); see Fig. 3a. Although the layer is very thin (about 3 nm), it is strongly depleted in nickel; see also Table 2. The reason may be the Ni dissolution in the etchant: Ni + 2H+ → Ni2+ + H2
(4)
which occurs simultaneously with the oxidation described by Eq. (3). For samples heat-treated at high temperatures, oxygen enrichment is detected to much larger distances under the surface (Fig. 3b–e). Among these samples, the thinnest oxide is observed on the 500LP one (Fig. 3b). Here, a thickness of TiO2 layer appears to be about 15 nm. It is in accordance with the relatively short heat-treatment duration and low oxygen pressure; see Table 1. In contrast, on the sample denoted as 550LP, which was treated at a significantly higher temperature, there is a TiO2 zone exceeding 50 nm in thickness (Fig. 3c). It is evident that the temperature
increase by 50 ◦ C strongly accelerated the oxidation kinetics despite the low oxygen partial pressure. For the sample treated in the salt bath (S), the titania surface layer achieves about 25 nm, slightly above that on the NP sample treated in air (∼20 nm); see Fig. 3d and e. However, the heat-treatment duration of the latter was 10 min (see Table 1), which was double that of the former. Therefore, the salt bath appears to be a more effective oxygen source than air. In the salt bath, the probability of the contact between oxidizing particles, in this case the nitrate anion, see Eq. (1), and the material surface is much higher than in air. Fig. 4 shows detailed Ti 2p spectra of both etched and heattreated samples. They are almost identical because the Ti4+ oxidation state of titanium predominates (Eb2p3/2 (Ti4+ ) = 458.8 eV [17]) on all surfaces. The preferential oxidation of titanium is in accordance with the Gibbs energies of formation of oxides, which are −889.5 and −211.7 kJ/mol for TiO2 and NiO (at 298 K), respectively [3]. Therefore, as indicated before, titanium and oxygen
1578
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
Fig. 5. Detailed Ni 2p spectra acquired on the surface of the samples: (a) CHE (dominated by Ni0 state), (b) NP (dominated by Ni2+ state), and (c) S (dominated by Ni2+ state) (XPS).
on the surface originate from titanium dioxide. Regarding nickel release, nickel oxidation states are also important. Detailed Ni 2p spectra of the samples are presented in Fig. 5. For the CHE sample, the Ni peak is hardly observed due to the low surface Ni concentration (Table 2). However, it seems that metallic nickel Ni0 in the NiTi state (Eb2p3/2 (Ni0 in NiTi) = 853.3 eV [17]) is predominant on this sample. Perhaps the low temperature and duration of oxidation as well as preferential oxidation of Ti are the reasons for the non-oxidized state of Ni. In contrast, after heat treatments in air and salt, nickel becomes oxidized as Ni2+ (Eb2p3/2 = 854.6 eV). Fig. 6 shows the Ti–Ni–O phase diagram computed for 530 ◦ C. The observed behavior of Ni on the heat-treated samples can be explained by this diagram. The wide area of TiO2 and NiO mixtures at high oxygen partial pressures (activities), i.e., on the surface where the material is in contact with air or molten salt, corresponds to the heat-treated samples. On the other hand, metallic nickel occurs at very low oxygen activities in the surface interior where oxygen diffusing inward is present [16]. Due to the low thickness of titanium dioxide on the surface of heat-treated samples ranging from 15 to 50 nm (see Fig. 3), the crystallographic nature of this oxide had to be determined by Raman spectrometry. Results are presented in Fig. 7. It is known that rutile is the most thermodynamically stable titania modification at high temperatures. However, in this experiment rutile was identified to predominate only on the NP sample, as evidenced by the presence of characteristic peaks at 610 and 445 cm−1 . On the S sample, there was a mixture of rutile and anatase showing a characteristic peak at 145 cm−1 . The other samples treated under low pressure air (500LP and 550LP) contained mainly anatase on their surface despite the high heat-treatment temperatures. Anatase is generally regarded as a metastable titania modification, which readily
transforms to rutile at high temperatures. The following sequence of TiO2 modifications is usually observed as both oxidation temperature increases and oxidation time is prolonged: 1, amorphous; 2, anatase; 3, rutile [19]. Therefore, it is believed that the presence of anatase on the S, 500LP and 550LP samples is attributable to the short oxidation periods of these samples, which is half of that of the NP sample. On the NP sample, the diffusion time is sufficient to rearrange atoms into the more stable rutile lattice. Beside the oxidation time, the oxidation pressure also seems to affect the
Fig. 6. Ti–Ni–O phase diagram computed for 530 ◦ C [18]. There is a wide area of TiO2 and NiO mixtures at high oxygen partial pressures (activities) corresponding to the heat-treated samples.
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
1579
Fig. 8. Ni-release into both corrosive media versus thickness of the surface oxides. Dark symbols represent Ni-release into the simulated physiological solution (pH 2), while open symbols Ni-release into the concentrated HCl (indicated by arrows). The dashed line indicates an estimated dependence of Ni-release into the simulated physiological solution on the oxide thickness.
Fig. 7. Raman spectra of the heat-treated samples. Spectra of both rutile and anatase are also included for comparison.
resulting titania phase composition. It appears that lower oxidation pressures for the 500LP and 550LP samples support the formation of less dense anatase modification (densities of rutile and anatase are 4.2 and 3.8 g/cm3 , respectively [20]). However, to support this statement, more experiments under various pressures would be needed. 3.2. Corrosion Because the wire used in this experiment was very thin, it was hard to measure the potentiodynamic curves accurately. For this reason, corrosion behavior was only explored in terms of nickel release into the corrosive medium. Results for both corrosive environments are summarized in Table 3. In the case of physiological solution, one can see that the CHE sample shows the best corrosion performance due to a nickel release below 1 g/cm2 . This sample is followed by the 500LP sample heat-treated at low pressure, which released 4 g/cm2 Ni. In contrast, the fastest corrosion occurred for the 550LP sample treated at the highest temperature. It is important to notice that the increase of the heat-treatment temperature by only 50 ◦ C led to a threefold acceleration of nickel release from 4 to 12 g/cm2 . It will be discussed later how this increase is directly related to
the thickness of surface oxide. Table 3 also illustrates that the samples treated in salt and normal pressure air release nickel ions at rates of about 8 g/cm2 , i.e., between the extremes above. By testing the samples in concentrated hydrochloric acid, it was observed that the samples behave differently. The highest corrosion resistance was measured for the 500LP sample, while the others released nickel at rates two or three times higher. Perhaps the strongly acidic and chloride-rich environment changes the mechanism and rate controlling process of corrosion, as will be discussed later. The first parameter that should be taken into account in relation to the corrosion behavior is the real surface area of the samples. Due to surface treatments that modify the surface morphology (Fig. 1), the real surface area differs from the nominal surface area calculated from the wire length and diameter. One would expect that the larger the real surface, the faster the Ni-release. Although it is difficult to estimate the real surface from the morphologies presented in Fig. 1, this figure indicates that differences in the real surface area between NP, S and 500LP samples (Fig. 1b–d) are probably small due to similar grain size of titania on these samples. However, the corrosion behavior of these samples differs dramatically (Table 3). In addition, the 550LP sample had the coarsest grains on its surface, and thus the lowest real surface, and it showed the fastest
Table 3 Nickel release into corrosive media (related to the nominal sample surface). All values are arithmetical means of three measurements. Physiological solution (pH 2, exposition for 168 h) Sample
Ni-release (g/cm2 )
CHE NP S 500LP 550LP
0.5 8.0 8.4 4.0 12.3
± ± ± ± ±
0.2 1.9 2.6 1.8 3.1
Concentrated HCl (exposition for 6 h)
Sample
Ni-release (mg/cm2 )
CHE NP S 500LP 550LP
9.0 10.9 10.3 3.9 8.9
± ± ± ± ±
0.7 1.4 1.5 0.5 0.5
Fig. 9. Defects in oxide layer on the 550LP sample (marked by arrows). Defects may originate from external mechanical or thermal loading, differences between molar volumes of metals and oxides and other factors.
1580
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
Ni-release. For these reasons, it seems more reasonable to relate the corrosion behavior to the titania thickness than to the surface area. Another parameter affecting corrosion resistance is the thickness of the surface oxide layer. Fig. 8 relates the values of nickel release into both media to the thickness of TiO2 surface layers estimated from the elemental profiles in Fig. 3. Consider first Nirelease into the simulated physiological solution. In this case, there is an evident relation between the Ni-release rate and oxide thickness because thicker oxides produce faster Ni-release, which is
expressed by a dashed line in Fig. 8. This finding seems surprising at first sight because one would expect an inverse trend, i.e., the thicker the TiO2 surface layers, the better the protection against Nirelease and corrosion. However, similar behavior was reported, for example, by Zhu et al. [21], who measured breakdown potentials of nitinol oxidized at various temperatures to produce different oxide thicknesses. They found that the highest breakdown potential corresponded to oxide layers of about 10 nm in thickness. As the oxide became thicker, the breakdown potential was reduced considerably, i.e., corrosion resistance worsened. In our case, the
Fig. 10. Surface of samples corroded in concentrated HCl for 6 h: (a) 550LP, (b) NP, (c) CHE, (d) S, and (e) 500LP. (a–d) Almost all surfaces are attacked by the corrosion medium. (e) Areas with the original longitudinal scratches, i.e., without corrosion attack, are present.
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
best performance is observed for the CHE sample with an oxide with a thickness of about 3 nm. Among heat-treated samples, the lowest Ni-release rate corresponds to the 500LP sample covered by an oxide layer that is 15 nm in thickness. On the other hand, the 50-nm-thick oxide detected on the 550LP sample provides the worst protection against Ni-release from all of the heat-treatment regimes. Because the nickel surface concentrations on the samples are very small (see Table 2), it can be assumed that Ni-release is controlled by the outward penetration of nickel from the metallic substrate. By comparing results given in Table 3 and Fig. 7, no correlation between Ni-release and titania crystallographic modification is found. Both 500LP and 550LP samples, which show very different corrosion behaviors, are dominated by anatase on the surface. Moreover, the oxidation state of Ni does not seem to affect the Ni-release rate significantly, because the Ni0 state, being the most prone to dissolution, is only detected on the CHE sample (Fig. 5), which, however, shows the slowest Ni-release. Therefore, it is believed that the corrosion behavior is primarily influenced by defects present in oxides. Any defect that accelerates Ni transport through the oxide accelerates Ni-release as well. Defects can be regarded as an intrinsic characteristic of each oxide layer. Defects like micro- and nano-cracks (Fig. 9) may originate, for example, from external mechanical or thermal loading of a material. In addition, differences between molar volumes of metals and oxides also play a role. Large differences usually result in internal compressive stresses in oxides, which may be released by cracking. Another source of defects is cooling from heat-treatment temperatures, which is very fast in our experiment. Due to the differences in thermal expansion between NiTi alloy and titania, stress induced in the latter may eventually also lead to its cracking. Generally, the susceptibility of an oxide layer to internal stress and defects by mechanisms suggested above primarily depends on its thickness. The larger the thickness, the higher the number of defects. In other words, sufficiently thin and compact oxide layers provide more efficient barriers against the outward penetration of nickel from the substrate, as is observed in the corrosion test performed in the model physiological solution in this study. There is another factor that should be considered in this context: the grain size of the oxide. It is shown in Fig. 1 that the samples heat-treated at 500–510 ◦ C, i.e., NP, S and 500LP, are covered by grains of about 20 nm in size. In contrast, the treatment at 550 ◦ C (550LP) produces about 50-nm titania grains. It can be assumed that an oxide layer consisting of fine grains better releases the internal stress than one composed of coarse grains. Stress release occurs by processes similar to hightemperature creep, i.e., they include grain boundary slip, which is faster for a fine-grained structure. It can be seen in Fig. 8 that the corrosion behavior of the samples in the concentrated HCl significantly differs from that in the less aggressive simulated physiological solution. No correlation between Ni-release and oxide thickness is observed in this case. The amounts of Ni released into the acid are similar for the CHE, NP, S and 550LP samples. One exception is the 500LP sample, which showed significantly slower Ni-release. Therefore, it appears that the oxide thickness is not the primary factor affecting Ni-release in this case. The probable reason is that Cl− anions present in high concentration, together with strongly acidic conditions, rapidly break down the surface oxide layers, according to the following reaction: TiO2 + 4H+ + xCl− → Ti4+ + 2H2 O + xCl−
(5)
It is assumed that chloride anions only catalyze this reaction because they do not form special compounds. Afterwards, i.e., after a certain incubation period, nickel release occurs rapidly on the alloy surface and is controlled by the dissolution given by Eq. (4). For the CHE sample, the original titania layer is very thin (about 3 nm, see Fig. 3). Therefore, it is rapidly broken down by the concentrated acid, despite its low defect concentration. On the other
1581
hand, thick layers present on the S, NP and 550LP samples show a higher defect concentration (Fig. 9), so that the incubation period is also short. Perhaps the 500LP sample represents a certain optimum of low defect concentration and sufficient thickness. In this case, dissolution of the oxide is slower, i.e., incubation is longer. The difference between the 500LP sample and the others is also illustrated in Fig. 10, which presents SEM micrographs of surfaces corroded in the concentrated HCl for 6 h. One can see that almost all of the surface of the 550LP, NP, CHE and S samples is attacked by the corrosion medium because the original surface is not observed. On the other hand, on the 500LP sample there are areas showing the original morphology with longitudinal scratches due to cold drawing. It is likely that, in these areas, the incubation period needed for bulk dissolution has not yet been exceeded. 4. Conclusions In this work, we studied several heat treatments of a NiTi alloy at about 500 ◦ C for 5–10 min in various environments, i.e., under conditions close to the shape setting procedure generally used in NiTi implant manufacture. The chemically etched surface was investigated for comparison. The treated materials were studied with respect to surface structure and chemistry and their impact on corrosion resistance. The corrosion rate was measured as the rate of Ni-release in this study, which is an important characteristic for biodegradability. We found that the applied heat-treatment regimes produced surface oxides of a thickness ranging from 15 to 50 nm. The thinnest oxide of about 15 nm was detected on the sample heat-treated at 500 ◦ C under a low pressure air. This surface layer showed the best corrosion performance among all of the heat-treated samples, both in the model physiological solution and in concentrated HCl. It thus appears that the shape setting performed at low pressures is a way to achieve a good corrosion resistance of nitinol. Another important finding of our research is that thicker oxides are more susceptible to nickel release. For corrosion in a model physiological solution of pH 2, which is closer to a human body environment, it is important that all heat treatments negatively affect corrosion resistance. Therefore, additional surface treatments, such as chemical or electrochemical etching performed after shape setting, could be recommended to improve corrosion performance. Acknowledgements The research on NiTi alloys and on surface nanostructures is financially supported by the Czech Academy of Sciences (projects no. IAA200100902 and KAN300100801) and by the Ministry of Education, Youth and Sports of the Czech Republic (project no. MSM6046137302). References [1] C. Heintz, G. Riepe, L. Birken, E. Kaiser, N. Chakfé, M. Morlock, G. Delling, H. Imig, Corroded nitinol wires in explanted aortic endografts: an important mechanism of failure? J. Endovasc. Ther. 8 (2001) 248–253. [2] S.A. Shabalovskaya, J. Anderegg, F. Laab, P.A. Thiel, G. Rondelli, Surface conditions of nitinol wires, tubing, and as-cast alloys. The effect of chemical etching, aging in boiling water, and heat treatment, J. Biomed. Mater. Res. 65B (2003) 193–203. [3] S.A. Shabalovskaya, G.C. Rondelli, A.L. Undisz, J.A. Anderegg, T.D. Burleigh, M.E. Rettenmayr, The electrochemical characteristics of native Nitinol surfaces, Biomaterials 30 (2009) 3662–3671. [4] N. Munroe, Ch. Pulletikuthi, W. Haider, Enhanced biocompatibility of porous nitinol, J. Mater. Eng. Perform. 18 (2009) 765–767. [5] S. Shabalovskaya, J. Anderegg, J. Van Humbeeck, Critical overview of nitinol surfaces and their modifications for medical applications, Acta Biomater. 4 (2008) 447–467. [6] W. Simka, M. Kaczmarek, A. Baron-Wiechec, G. Nawrat, J. Marciniak, J. Zak, Electropolishing and passivation of NiTi shape memory alloy, Electrochim. Acta (2009), doi:10.1016/j.electacta.2009.11.097.
1582
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
[7] M.H Wong, F.T. Cheng, H.C. Man, Comparison of corrosion resistance and apatite-forming ability of NiTi treated by different low-temperature methods, J. Alloys Compd. 466 (2008) L5–L10. [8] M. Pohl, T. Glogowski, S. Kuhn, C. Hessing, F. Unterumsberger, Formation of titanium oxide coatings on NiTi shape memory alloys by selective oxidation, Mater. Sci. Eng. A 481–482 (2008) 123–126. [9] A. Michiardi, C. Aparicio, J.A. Plannell, F.J. Gil, Electrochemical behaviour of oxidized NiTi shape memory alloys for biomedical applications, Surf. Coat. Technol. 201 (2007) 6484–6488. [10] L. Zhu, J.M. Fino, A. Pelton, in: T.W. Duerig, A. Pelton (Eds.), Oxidation of Nitinol, Proceedings of the International Conference on Shape Memory and Superelastic Technologies (SMST-2003), Nitinol Devices Components, 2003, pp. 357–366. [11] D. Vojtˇech, P. Novák, M. Novák, L. Joska, T. Fabián, J. Maixner, V. Machoviˇc, Cyclic and isothermal oxidations of nitinol wire at moderate temperatures, Intermetallics 16 (2008) 424–431. [12] G.S. Firstov, R.G. Vitchev, H. Kumar, B. Blanpain, J. Van Humbeeck, Surface oxidation of NiTi shape memory alloy, Biomaterials 23 (2002) 4863–4871. [13] Y.W. Gu, B.Y. Tay, C.S. Lim, M.S. Yong, Characterization of bioactive surface oxidation layer on NiTi alloy, Appl. Surf. Sci. 252 (2005) 2038–2049. [14] M. Wiedmann, Self-expanding antireflux stent fracture, Endoscopy 41 (2009) E129–E130.
[15] A.H. Goenka, Esophageal migration of fractured SEMS fragment, Endoscopy 41 (2009) E204–E205. [16] D. Vojtˇech, L. Joska, J. Leitner, Influence of a controlled oxidation at moderate temperatures on the surface chemistry of nitinol wire, Appl. Surf. Sci. 254 (2008) 5664–5669. [17] R.M. Wang, C.L. Chu, T. Hu, C. Guo, X.B. Sheng, P.H. Lin, C.Y. Chung, P.K. Chu, Surface XPS characterization of NiTi shape memory alloy after advanced oxidation processes in UV/H2 O2 photocatalytic system, Appl. Surf. Sci. 253 (2007) 8507–8512. [18] C.W. Bale, P. Chartrand, S.A. Degterov, G. Eriksson, K. Hack, R. Ben Mahfoud, J. Melanc¸on, A.D. Pelton, S. Petersen, FactSage thermochemical software and databases, Calphad 26 (2002) 189–228. [19] K.S. Kim, K.K. Jee, W.C. Kim, W.Y. Jang, S.H. Han, Effect of heat treatment temperature on oxidation behaviour in Ni–Ti alloy, Mater. Sci. Eng. A 481–482 (2008) 658–661. [20] U. Diebold, The surface science of titanium dioxide, Surf. Sci. Rep. 48 (2003) 53–229. [21] L. Zhu, C. Trepanier, A.R. Pelton, Oxidation of Nitinol and its Effect on Corrosion Resistance, Proceedings of ASM Materials and Processes for Medical Devices Conference 2003, Nitinol Devices and Components, 2003, pp. 1–6.
Contents lists available at ScienceDirect
Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc
Surface structure and corrosion resistance of short-time heat-treated NiTi shape memory alloy D. Vojtˇech ∗ , M. Vodˇerová, J. Fojt, P. Novák, T. Kubásek Department of Mecotals and Corrosion Engineering, Institute of Chemical Technology, Prague, Technická 5, 166 28 Prague 6, Czech Republic
a r t i c l e
i n f o
Article history: Received 8 April 2010 Received in revised form 17 August 2010 Accepted 21 August 2010 Available online 20 September 2010 Keywords: NiTi alloy Heat treatment Oxidation Corrosion Shape memory alloy
a b s t r a c t NiTi alloys are attractive materials that are used for medicine, however, Ni-release may cause allergic reactions in an organism. The Ni-release rate is strongly affected by the surface state of the NiTi alloy that is mainly determined by its processing route. In this study, a NiTi shape memory alloy (50.9 at.% Ni) was heat-treated by several regimes simulating the shape setting procedure, the last step in the manufacture of implants. Heating temperatures were between 500 and 550 ◦ C and durations from 5 to 10 min. Heat treatments were performed in air at normal and low pressure and in a salt bath. The purpose of the treatments was to obtain and compare different surface states of the Ni–Ti alloy. The surface state and chemistry of heat-treated samples were investigated by electron microscopy, X-ray photoelectron spectroscopy and Raman spectrometry. The amount of nickel released into a model physiological solution of pH 2 and into concentrated HCl was taken as a measure of the corrosion rate. It was found that the heat treatments produced surface TiO2 layers measuring 15–50 nm in thickness that were depleted in nickel. The sample covered by the 15-nm thick oxide that was treated at 500 ◦ C/5 min in a low pressure air showed the best corrosion performance in terms of Ni-release. As the oxide thickness increased, due to either temperature or oxygen activity change, Ni-release into the physiological solution accelerated. This finding is discussed in relation to the internal structure of the oxide layers. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Nearly equi-atomic NiTi alloys (nitinol) have attracted much attention due to their shape memory effect, superelastic behavior, high tensile strength, good corrosion resistance and biocompatibility. These characteristics make NiTi alloys attractive for biomedical applications, such as stents, which are generally utilized to restore a damaged blood vessel or an oesophagus, among others. NiTi alloys are generally regarded as highly corrosion resistant, similar to stainless steel or titanium. The reason is the spontaneous formation of a native passive TiO2 layer on the surface. Even a weak oxidizing environment, such as water, air or humidity, in contact with nitinol at low temperature is sufficient to produce a few nanometers of a passive protective layer. However, in the processing of NiTi alloys they experience various forming, heat- or surface-treating steps, which strongly influence the surface structure, chemistry and, therefore, corrosion performance. In a human body, various fluids may come into contact with NiTi implants. In particular, serious problems may arise when a strongly acidic fluid, like gastric juice, attacks the surface of the alloy. In general, the corrosion of nitinol has two aspects. First, cor-
∗ Corresponding author. Tel.: +420 220 444290. E-mail address: [email protected] (D. Vojtˇech). 0169-4332/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2010.08.097
rosion is accompanied with nickel release from an implant into the surrounding body fluid and tissue, which may enhance an allergic reaction in a sensitive organism. Second, corrosion may cause pitting and a reduction of mechanical performance of an implant. In an extreme case, an implant broken due to corrosion may produce sharp fragments that are dangerous for surrounding tissue [1]. For these reasons, the corrosion of nitinol has become an important issue in biomedical engineering during the last decade. To slow down this corrosion process, a highly protective adherent layer with a low concentration of nickel and internal defects must cover the surface of nitinol. A number of oxidation treatment procedures have been reported so far to support the preferential Ti oxidation and increase the protective effect against corrosion. These include thermal oxidation either in air or under a low oxygen pressure, etching, passivation in nitric acid, anodization, electrochemical polishing, oxygen ion implantation, laser oxidation, hydrothermal oxidation in water, sol–gel techniques, cathodic deposition and others [2–9]. Most often, the last processing step in the manufacture of nitinol stents is the shape setting. The shape setting treatment means a short-time heating of nitinol at around 500 ◦ C and serves to achieve a desired shape and superelastic behavior of an implant. Heating can be performed in air, argon, vacuum or a salt bath. In oxidizing environments, titanium in nitinol generally oxidizes preferentially over nickel due to the higher thermodynamic stability of TiO2 as
1574
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
Table 1 Summary of heat-treatment (HT) regimes used in this study and their designation. Sample designation
Heat-treatment regime
CHE
No HT, chemical etching in a 1HF + 4HNO3 + 5H2 O (by volume) solution HT at 510 ◦ C/10 min in air at a normal pressure HT at 510 ◦ C/5 min in a molten salt bath HT at 500 ◦ C/5 min in air at a low pressure of 10 Pa HT at 550 ◦ C/5 min in air at a low pressure of 10 Pa
NP S 500LP 550LP
compared to NiO [3]. At sufficiently high temperatures, titanium reacts rapidly even with traces of oxygen present in a surrounding environment. As a result, a TiO2 -enriched and Ni-depleted surface oxide layer forms [10–13]. The corrosion performance of a nitinol implant thus depends on properties of such an oxide layer, particularly, on its thickness, chemistry, adherence and internal structure. In the literature, there are few studies on the influence of the shape setting treatment on corrosion behavior. For this reason, our study is concerned with short-time heat treatments of nitinol at around 500 ◦ C. Treatments were carried out in various environments in order to simulate different oxidation potentials and the purpose was to prepare and compare various surface states of the Ni–Ti alloy. Corrosion tests were performed in a relatively aggressive solution with pH 2. The solution simulates gastric fluid, so that information obtained by these tests is important for implants, such as oesophageal stents, for example. The fracture of a stent in a patient’s body due to corrosion may cause serious problems [14,15].
spectroscope (XPS) ESCA Probe P (a pressure in the analytical chamber of 2 × 10−8 Pa, monochromatic Al K␣ X-ray source, binding energy calibration with respect to the energy of Au 4f7/2 peak, Ar pressure during sputter depth profiling 3 × 10−3 Pa, energy of Ar+ ions during sputtering 5 keV, calibration of the sputtering rate with respect to titanium dioxide). Chemical compositions were calculated from areas of deconvoluted elemental peaks by analytical software CasaXPS version 2.3.14. The thickness of TiO2 layer was estimated as a thickness at which a Ti:O ratio is 1:2 or lower. A Ti:O ratio lower than 1:2 indicates oxygen adsorption on the surface TiO2 . Detailed spectra of Ti 2p and Ni 2p were recorded using a constant pass energy of 20 eV at a 0.1-eV step size. The corrosion behavior of both chemically etched and heattreated samples was determined by the immersion test. Three samples for each surface state were tested. Samples were exposed to a simulated physiological solution containing 9 g/l NaCl with pH 2 (adjusted by the addition of HCl) at a temperature of 37 ◦ C. The low pH value was used in this experiment because it simulates a gastric juice environment. The immersion period was 168 h. Simultaneously, the corrosion behavior was also determined in a concentrated HCl. Samples were exposed to this strongly aggressive environment for 6 h at 20 ◦ C. In both corrosion tests, the corrosion rate was taken as the concentration of nickel released into corrosive medium. Nickel concentrations in the simulated physiological solution and in the concentrated HCl were measured by ICP mass spectrometry (ELAN 6000) and by atomic absorption spectrometry (GBC 932 plus), respectively.
2. Experimental
3. Results and discussion
NiTi alloy (50.9% Ni; hereafter, all concentrations are in at.% unless otherwise stated) was used for heat treatments in our experiments. The 0.45-mm-thick wire with a tensile strength of 1580 MPa and Af of 25 ◦ C was provided by an industrial supplier. According to the supplier specifications, the wire was cold drawn (45% deformation) and annealed. Detailed specifications of the annealing conditions were not given. The NiTi wire was first subjected to chemical etching in a 1HF + 4HNO3 + 5H2 O (by volume) solution in order to remove traces of previous processing steps and to obtain a defined and unified surface state. Prior to heat treatments, the surfaces of all samples were carefully washed and dried. In this study, heat treatment regimes were selected to reflect real conditions of the shape setting procedures used by manufacturers of medical stents. Therefore, they were performed at different temperatures between 500 and 550 ◦ C in air both at ambient pressure and at a low pressure of 10 Pa and also in a molten salt bath containing a mixture of 60% KNO3 and 40% NaNO3 . Low pressures are sometimes used to prevent excessive oxidation of nitinol. The advantage of the shape setting in salt baths is the fast heat transfer from the bath to the treated wire and therefore the short time needed for the shape setting. After each heat treatment, the wire was immediately quenched into cold water. Detailed descriptions of heat treatment procedures applied in our experiment and their designation are given in Table 1. All heat-treated samples appeared in characteristic interference colors ranging from light golden to a dark blue color. The surface morphology of samples was observed by scanning electron microscopy (SEM, Hitachi S4700, an acceleration voltage of 10 kV, in the regime of secondary electrons). Due to the very low thickness of surface layers on samples, the phases present were not detected by X-ray diffraction, but a Raman spectrometer Labram HR (excitation source – laser with a wavelength of 532 nm operating at 40 mW) was used for this purpose. Additionally, the chemical composition of the surface and elemental profiling in the sub-surface regions of samples were characterized by an X-ray photoelectron
3.1. Structure, chemical and phase composition Fig. 1a presents a SEM view of the CHE surface. One can see than the surface is not flat. Instead, chemical etching causes the formation of small dimples of a few micrometers in size on the surface. There are also scratch marks resulting from the cold drawing process in which spherical or elongated non-metallic inclusions are observed. These inclusions consisting mainly of titanium carbides originate from the melting process of nitinol and from its contamination by the melting crucible. Oxidation in air at 510 ◦ C/10 min modifies the surface structure, as shown in Fig. 1b. In contrast to the CHE sample, the oxide layer formed at high temperature has a typical grainy morphology with a grain size of about 20 nm. It has been demonstrated in a number of studies [16] that surface treatments of nitinol performed at room or slightly higher temperatures generally produce amorphous surface oxides. The reason is that slow diffusivities of species prevent the formation of ordered crystalline arrangements of atoms. At about 500 ◦ C diffusion becomes sufficiently faster to produce crystalline oxide layers that are more thermodynamically stable. The surface morphologies of the S and 500LP samples presented in Fig. 1c and d are almost identical to the previous one because they consist of fine grains of about 20 nm in size. It appears that oxygen activity, which is different in the treatments illustrated by Fig. 1b–d, does not influence the oxide grain size significantly. The main factor affecting the grain size is heattreatment temperature, which is demonstrated in Fig. 1e where a view of the 550LP sample is shown. Although the oxygen partial pressure was identical to that of the 500LP, the grain size grew considerably from approximately 20 to50 nm as the temperature increased by 50 ◦ C. It was already suggested in our previous work [11] that nanometer-sized titania grains grow relatively fast above 500 ◦ C due to the high driving force of this growth. Survey XPS spectra of the investigated samples are almost identical, and they mainly contain characteristic peaks of Ti, O, and contamination C; see Fig. 2 for illustration. As far as nickel
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
1575
Fig. 1. Surface morphologies of etched and heat-treated samples: (a) CHE, (b) NP, (c) S, (d) 500LP, and (e) 550LP (SEM). For sample designation see Table 1.
release from nitinol is concerned, knowledge of surface chemistry is important. Surface chemical compositions (excluding contamination carbon) are given in Table 2. High concentrations of oxygen,
Table 2 Surface chemical compositions (in at.%) of the etched and heat-treated samples (XPS). For sample designation see Table 1. Sample
CHE NP S 500LP 550LP
Element concentration (at.%) Ni
Ti
O
0.5 2.4 1.5 0 0
22.0 16.2 10.7 15.5 11.2
77.5 81.4 87.8 84.5 88.8
much exceeding that in TiO2 , result from its adsorption on the surface from the surrounding atmosphere. In the case of the S sample treated in the salt bath, the oxygen probably originates mainly from a decomposition of nitrate to nitrite at the alloy surface, which may be written as NO3 − → NO2 − + O
(1)
The O-rich layer on the CHE sample results from two simultaneous chemical reactions. The first one includes the dissolution of the TiO2 native layer in hydrofluoric acid to form soluble fluoride complexes: TiO2 + 6F− + 4H+ → [TiF6 ]2− + 2H2 O
(2)
The second process is preferential oxidation of titanium by nitric acid back to titanium dioxide, which is accompanied by the inward
1576
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
diffusion of nickel: Ti + 4NO3 − + 4H+ → TiO2 + 4NO2 + 2H2 O
Fig. 2. Survey XPS spectrum of the NP sample showing Ti, O, Ni and contamination C peaks.
(3)
Regarding Ni concentrations, one can see in Table 2 that these concentrations are low in all samples and do not exceed a few atomic percents. The highest Ni amount is detected on the NP sample, whereas the lowest are on both 500LP and 550LP ones. The nickel depletion on the oxidized surface results from preferential oxidation of titanium due to the high thermodynamic stability of TiO2 [3]. Generally, the oxidation processes of Ti-rich intermetallics are controlled by oxygen inward diffusion through the surface oxide. At the oxide/metal interface, the preferential oxidation of Ti produces a new titania, which is accompanied by nickel inward diffusion. It was demonstrated in several studies that sufficiently long oxidation at high temperatures may even produce Ni-enriched zones beneath the oxide [10]. Such zones may consist of Ni3 Ti, Ni4 Ti and even pure Ni. Therefore, they represent Ni reservoirs that may enhance nickel release, particularly in the case of defect occurrence in the surface oxide. Fortunately, as will be shown in Fig. 3, significant nickel accumulation beneath the oxide is not observed in this study, probably due to short oxidation periods.
Fig. 3. Elemental profiles through thin sub-surface regions on the investigated samples (XPS): (a) CHE, (b) 500LP, (c) 550LP, (d) S, and (e) NP. Estimated thickness of TiO2 is indicated by dashed lines. The TiO2 layer corresponds to a Ti:O ratio of 1:2 or lower. A Ti:O ratio lower than 1:2 indicates oxygen adsorption on the surface TiO2 .
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
1577
Fig. 4. Detailed Ti 2p spectra acquired on the surface of the samples: (a) CHE, (b) NP, (c) S, (d) 500LP, and (e) 550LP (XPS). All spectra are dominated by the Ti4+ oxidation state.
Fig. 3 shows elemental profiles measured in thin sub-surface regions by XPS. As expected, the thinnest oxide layer covers the surface of the etched sample (CHE); see Fig. 3a. Although the layer is very thin (about 3 nm), it is strongly depleted in nickel; see also Table 2. The reason may be the Ni dissolution in the etchant: Ni + 2H+ → Ni2+ + H2
(4)
which occurs simultaneously with the oxidation described by Eq. (3). For samples heat-treated at high temperatures, oxygen enrichment is detected to much larger distances under the surface (Fig. 3b–e). Among these samples, the thinnest oxide is observed on the 500LP one (Fig. 3b). Here, a thickness of TiO2 layer appears to be about 15 nm. It is in accordance with the relatively short heat-treatment duration and low oxygen pressure; see Table 1. In contrast, on the sample denoted as 550LP, which was treated at a significantly higher temperature, there is a TiO2 zone exceeding 50 nm in thickness (Fig. 3c). It is evident that the temperature
increase by 50 ◦ C strongly accelerated the oxidation kinetics despite the low oxygen partial pressure. For the sample treated in the salt bath (S), the titania surface layer achieves about 25 nm, slightly above that on the NP sample treated in air (∼20 nm); see Fig. 3d and e. However, the heat-treatment duration of the latter was 10 min (see Table 1), which was double that of the former. Therefore, the salt bath appears to be a more effective oxygen source than air. In the salt bath, the probability of the contact between oxidizing particles, in this case the nitrate anion, see Eq. (1), and the material surface is much higher than in air. Fig. 4 shows detailed Ti 2p spectra of both etched and heattreated samples. They are almost identical because the Ti4+ oxidation state of titanium predominates (Eb2p3/2 (Ti4+ ) = 458.8 eV [17]) on all surfaces. The preferential oxidation of titanium is in accordance with the Gibbs energies of formation of oxides, which are −889.5 and −211.7 kJ/mol for TiO2 and NiO (at 298 K), respectively [3]. Therefore, as indicated before, titanium and oxygen
1578
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
Fig. 5. Detailed Ni 2p spectra acquired on the surface of the samples: (a) CHE (dominated by Ni0 state), (b) NP (dominated by Ni2+ state), and (c) S (dominated by Ni2+ state) (XPS).
on the surface originate from titanium dioxide. Regarding nickel release, nickel oxidation states are also important. Detailed Ni 2p spectra of the samples are presented in Fig. 5. For the CHE sample, the Ni peak is hardly observed due to the low surface Ni concentration (Table 2). However, it seems that metallic nickel Ni0 in the NiTi state (Eb2p3/2 (Ni0 in NiTi) = 853.3 eV [17]) is predominant on this sample. Perhaps the low temperature and duration of oxidation as well as preferential oxidation of Ti are the reasons for the non-oxidized state of Ni. In contrast, after heat treatments in air and salt, nickel becomes oxidized as Ni2+ (Eb2p3/2 = 854.6 eV). Fig. 6 shows the Ti–Ni–O phase diagram computed for 530 ◦ C. The observed behavior of Ni on the heat-treated samples can be explained by this diagram. The wide area of TiO2 and NiO mixtures at high oxygen partial pressures (activities), i.e., on the surface where the material is in contact with air or molten salt, corresponds to the heat-treated samples. On the other hand, metallic nickel occurs at very low oxygen activities in the surface interior where oxygen diffusing inward is present [16]. Due to the low thickness of titanium dioxide on the surface of heat-treated samples ranging from 15 to 50 nm (see Fig. 3), the crystallographic nature of this oxide had to be determined by Raman spectrometry. Results are presented in Fig. 7. It is known that rutile is the most thermodynamically stable titania modification at high temperatures. However, in this experiment rutile was identified to predominate only on the NP sample, as evidenced by the presence of characteristic peaks at 610 and 445 cm−1 . On the S sample, there was a mixture of rutile and anatase showing a characteristic peak at 145 cm−1 . The other samples treated under low pressure air (500LP and 550LP) contained mainly anatase on their surface despite the high heat-treatment temperatures. Anatase is generally regarded as a metastable titania modification, which readily
transforms to rutile at high temperatures. The following sequence of TiO2 modifications is usually observed as both oxidation temperature increases and oxidation time is prolonged: 1, amorphous; 2, anatase; 3, rutile [19]. Therefore, it is believed that the presence of anatase on the S, 500LP and 550LP samples is attributable to the short oxidation periods of these samples, which is half of that of the NP sample. On the NP sample, the diffusion time is sufficient to rearrange atoms into the more stable rutile lattice. Beside the oxidation time, the oxidation pressure also seems to affect the
Fig. 6. Ti–Ni–O phase diagram computed for 530 ◦ C [18]. There is a wide area of TiO2 and NiO mixtures at high oxygen partial pressures (activities) corresponding to the heat-treated samples.
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
1579
Fig. 8. Ni-release into both corrosive media versus thickness of the surface oxides. Dark symbols represent Ni-release into the simulated physiological solution (pH 2), while open symbols Ni-release into the concentrated HCl (indicated by arrows). The dashed line indicates an estimated dependence of Ni-release into the simulated physiological solution on the oxide thickness.
Fig. 7. Raman spectra of the heat-treated samples. Spectra of both rutile and anatase are also included for comparison.
resulting titania phase composition. It appears that lower oxidation pressures for the 500LP and 550LP samples support the formation of less dense anatase modification (densities of rutile and anatase are 4.2 and 3.8 g/cm3 , respectively [20]). However, to support this statement, more experiments under various pressures would be needed. 3.2. Corrosion Because the wire used in this experiment was very thin, it was hard to measure the potentiodynamic curves accurately. For this reason, corrosion behavior was only explored in terms of nickel release into the corrosive medium. Results for both corrosive environments are summarized in Table 3. In the case of physiological solution, one can see that the CHE sample shows the best corrosion performance due to a nickel release below 1 g/cm2 . This sample is followed by the 500LP sample heat-treated at low pressure, which released 4 g/cm2 Ni. In contrast, the fastest corrosion occurred for the 550LP sample treated at the highest temperature. It is important to notice that the increase of the heat-treatment temperature by only 50 ◦ C led to a threefold acceleration of nickel release from 4 to 12 g/cm2 . It will be discussed later how this increase is directly related to
the thickness of surface oxide. Table 3 also illustrates that the samples treated in salt and normal pressure air release nickel ions at rates of about 8 g/cm2 , i.e., between the extremes above. By testing the samples in concentrated hydrochloric acid, it was observed that the samples behave differently. The highest corrosion resistance was measured for the 500LP sample, while the others released nickel at rates two or three times higher. Perhaps the strongly acidic and chloride-rich environment changes the mechanism and rate controlling process of corrosion, as will be discussed later. The first parameter that should be taken into account in relation to the corrosion behavior is the real surface area of the samples. Due to surface treatments that modify the surface morphology (Fig. 1), the real surface area differs from the nominal surface area calculated from the wire length and diameter. One would expect that the larger the real surface, the faster the Ni-release. Although it is difficult to estimate the real surface from the morphologies presented in Fig. 1, this figure indicates that differences in the real surface area between NP, S and 500LP samples (Fig. 1b–d) are probably small due to similar grain size of titania on these samples. However, the corrosion behavior of these samples differs dramatically (Table 3). In addition, the 550LP sample had the coarsest grains on its surface, and thus the lowest real surface, and it showed the fastest
Table 3 Nickel release into corrosive media (related to the nominal sample surface). All values are arithmetical means of three measurements. Physiological solution (pH 2, exposition for 168 h) Sample
Ni-release (g/cm2 )
CHE NP S 500LP 550LP
0.5 8.0 8.4 4.0 12.3
± ± ± ± ±
0.2 1.9 2.6 1.8 3.1
Concentrated HCl (exposition for 6 h)
Sample
Ni-release (mg/cm2 )
CHE NP S 500LP 550LP
9.0 10.9 10.3 3.9 8.9
± ± ± ± ±
0.7 1.4 1.5 0.5 0.5
Fig. 9. Defects in oxide layer on the 550LP sample (marked by arrows). Defects may originate from external mechanical or thermal loading, differences between molar volumes of metals and oxides and other factors.
1580
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
Ni-release. For these reasons, it seems more reasonable to relate the corrosion behavior to the titania thickness than to the surface area. Another parameter affecting corrosion resistance is the thickness of the surface oxide layer. Fig. 8 relates the values of nickel release into both media to the thickness of TiO2 surface layers estimated from the elemental profiles in Fig. 3. Consider first Nirelease into the simulated physiological solution. In this case, there is an evident relation between the Ni-release rate and oxide thickness because thicker oxides produce faster Ni-release, which is
expressed by a dashed line in Fig. 8. This finding seems surprising at first sight because one would expect an inverse trend, i.e., the thicker the TiO2 surface layers, the better the protection against Nirelease and corrosion. However, similar behavior was reported, for example, by Zhu et al. [21], who measured breakdown potentials of nitinol oxidized at various temperatures to produce different oxide thicknesses. They found that the highest breakdown potential corresponded to oxide layers of about 10 nm in thickness. As the oxide became thicker, the breakdown potential was reduced considerably, i.e., corrosion resistance worsened. In our case, the
Fig. 10. Surface of samples corroded in concentrated HCl for 6 h: (a) 550LP, (b) NP, (c) CHE, (d) S, and (e) 500LP. (a–d) Almost all surfaces are attacked by the corrosion medium. (e) Areas with the original longitudinal scratches, i.e., without corrosion attack, are present.
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
best performance is observed for the CHE sample with an oxide with a thickness of about 3 nm. Among heat-treated samples, the lowest Ni-release rate corresponds to the 500LP sample covered by an oxide layer that is 15 nm in thickness. On the other hand, the 50-nm-thick oxide detected on the 550LP sample provides the worst protection against Ni-release from all of the heat-treatment regimes. Because the nickel surface concentrations on the samples are very small (see Table 2), it can be assumed that Ni-release is controlled by the outward penetration of nickel from the metallic substrate. By comparing results given in Table 3 and Fig. 7, no correlation between Ni-release and titania crystallographic modification is found. Both 500LP and 550LP samples, which show very different corrosion behaviors, are dominated by anatase on the surface. Moreover, the oxidation state of Ni does not seem to affect the Ni-release rate significantly, because the Ni0 state, being the most prone to dissolution, is only detected on the CHE sample (Fig. 5), which, however, shows the slowest Ni-release. Therefore, it is believed that the corrosion behavior is primarily influenced by defects present in oxides. Any defect that accelerates Ni transport through the oxide accelerates Ni-release as well. Defects can be regarded as an intrinsic characteristic of each oxide layer. Defects like micro- and nano-cracks (Fig. 9) may originate, for example, from external mechanical or thermal loading of a material. In addition, differences between molar volumes of metals and oxides also play a role. Large differences usually result in internal compressive stresses in oxides, which may be released by cracking. Another source of defects is cooling from heat-treatment temperatures, which is very fast in our experiment. Due to the differences in thermal expansion between NiTi alloy and titania, stress induced in the latter may eventually also lead to its cracking. Generally, the susceptibility of an oxide layer to internal stress and defects by mechanisms suggested above primarily depends on its thickness. The larger the thickness, the higher the number of defects. In other words, sufficiently thin and compact oxide layers provide more efficient barriers against the outward penetration of nickel from the substrate, as is observed in the corrosion test performed in the model physiological solution in this study. There is another factor that should be considered in this context: the grain size of the oxide. It is shown in Fig. 1 that the samples heat-treated at 500–510 ◦ C, i.e., NP, S and 500LP, are covered by grains of about 20 nm in size. In contrast, the treatment at 550 ◦ C (550LP) produces about 50-nm titania grains. It can be assumed that an oxide layer consisting of fine grains better releases the internal stress than one composed of coarse grains. Stress release occurs by processes similar to hightemperature creep, i.e., they include grain boundary slip, which is faster for a fine-grained structure. It can be seen in Fig. 8 that the corrosion behavior of the samples in the concentrated HCl significantly differs from that in the less aggressive simulated physiological solution. No correlation between Ni-release and oxide thickness is observed in this case. The amounts of Ni released into the acid are similar for the CHE, NP, S and 550LP samples. One exception is the 500LP sample, which showed significantly slower Ni-release. Therefore, it appears that the oxide thickness is not the primary factor affecting Ni-release in this case. The probable reason is that Cl− anions present in high concentration, together with strongly acidic conditions, rapidly break down the surface oxide layers, according to the following reaction: TiO2 + 4H+ + xCl− → Ti4+ + 2H2 O + xCl−
(5)
It is assumed that chloride anions only catalyze this reaction because they do not form special compounds. Afterwards, i.e., after a certain incubation period, nickel release occurs rapidly on the alloy surface and is controlled by the dissolution given by Eq. (4). For the CHE sample, the original titania layer is very thin (about 3 nm, see Fig. 3). Therefore, it is rapidly broken down by the concentrated acid, despite its low defect concentration. On the other
1581
hand, thick layers present on the S, NP and 550LP samples show a higher defect concentration (Fig. 9), so that the incubation period is also short. Perhaps the 500LP sample represents a certain optimum of low defect concentration and sufficient thickness. In this case, dissolution of the oxide is slower, i.e., incubation is longer. The difference between the 500LP sample and the others is also illustrated in Fig. 10, which presents SEM micrographs of surfaces corroded in the concentrated HCl for 6 h. One can see that almost all of the surface of the 550LP, NP, CHE and S samples is attacked by the corrosion medium because the original surface is not observed. On the other hand, on the 500LP sample there are areas showing the original morphology with longitudinal scratches due to cold drawing. It is likely that, in these areas, the incubation period needed for bulk dissolution has not yet been exceeded. 4. Conclusions In this work, we studied several heat treatments of a NiTi alloy at about 500 ◦ C for 5–10 min in various environments, i.e., under conditions close to the shape setting procedure generally used in NiTi implant manufacture. The chemically etched surface was investigated for comparison. The treated materials were studied with respect to surface structure and chemistry and their impact on corrosion resistance. The corrosion rate was measured as the rate of Ni-release in this study, which is an important characteristic for biodegradability. We found that the applied heat-treatment regimes produced surface oxides of a thickness ranging from 15 to 50 nm. The thinnest oxide of about 15 nm was detected on the sample heat-treated at 500 ◦ C under a low pressure air. This surface layer showed the best corrosion performance among all of the heat-treated samples, both in the model physiological solution and in concentrated HCl. It thus appears that the shape setting performed at low pressures is a way to achieve a good corrosion resistance of nitinol. Another important finding of our research is that thicker oxides are more susceptible to nickel release. For corrosion in a model physiological solution of pH 2, which is closer to a human body environment, it is important that all heat treatments negatively affect corrosion resistance. Therefore, additional surface treatments, such as chemical or electrochemical etching performed after shape setting, could be recommended to improve corrosion performance. Acknowledgements The research on NiTi alloys and on surface nanostructures is financially supported by the Czech Academy of Sciences (projects no. IAA200100902 and KAN300100801) and by the Ministry of Education, Youth and Sports of the Czech Republic (project no. MSM6046137302). References [1] C. Heintz, G. Riepe, L. Birken, E. Kaiser, N. Chakfé, M. Morlock, G. Delling, H. Imig, Corroded nitinol wires in explanted aortic endografts: an important mechanism of failure? J. Endovasc. Ther. 8 (2001) 248–253. [2] S.A. Shabalovskaya, J. Anderegg, F. Laab, P.A. Thiel, G. Rondelli, Surface conditions of nitinol wires, tubing, and as-cast alloys. The effect of chemical etching, aging in boiling water, and heat treatment, J. Biomed. Mater. Res. 65B (2003) 193–203. [3] S.A. Shabalovskaya, G.C. Rondelli, A.L. Undisz, J.A. Anderegg, T.D. Burleigh, M.E. Rettenmayr, The electrochemical characteristics of native Nitinol surfaces, Biomaterials 30 (2009) 3662–3671. [4] N. Munroe, Ch. Pulletikuthi, W. Haider, Enhanced biocompatibility of porous nitinol, J. Mater. Eng. Perform. 18 (2009) 765–767. [5] S. Shabalovskaya, J. Anderegg, J. Van Humbeeck, Critical overview of nitinol surfaces and their modifications for medical applications, Acta Biomater. 4 (2008) 447–467. [6] W. Simka, M. Kaczmarek, A. Baron-Wiechec, G. Nawrat, J. Marciniak, J. Zak, Electropolishing and passivation of NiTi shape memory alloy, Electrochim. Acta (2009), doi:10.1016/j.electacta.2009.11.097.
1582
D. Vojtˇech et al. / Applied Surface Science 257 (2010) 1573–1582
[7] M.H Wong, F.T. Cheng, H.C. Man, Comparison of corrosion resistance and apatite-forming ability of NiTi treated by different low-temperature methods, J. Alloys Compd. 466 (2008) L5–L10. [8] M. Pohl, T. Glogowski, S. Kuhn, C. Hessing, F. Unterumsberger, Formation of titanium oxide coatings on NiTi shape memory alloys by selective oxidation, Mater. Sci. Eng. A 481–482 (2008) 123–126. [9] A. Michiardi, C. Aparicio, J.A. Plannell, F.J. Gil, Electrochemical behaviour of oxidized NiTi shape memory alloys for biomedical applications, Surf. Coat. Technol. 201 (2007) 6484–6488. [10] L. Zhu, J.M. Fino, A. Pelton, in: T.W. Duerig, A. Pelton (Eds.), Oxidation of Nitinol, Proceedings of the International Conference on Shape Memory and Superelastic Technologies (SMST-2003), Nitinol Devices Components, 2003, pp. 357–366. [11] D. Vojtˇech, P. Novák, M. Novák, L. Joska, T. Fabián, J. Maixner, V. Machoviˇc, Cyclic and isothermal oxidations of nitinol wire at moderate temperatures, Intermetallics 16 (2008) 424–431. [12] G.S. Firstov, R.G. Vitchev, H. Kumar, B. Blanpain, J. Van Humbeeck, Surface oxidation of NiTi shape memory alloy, Biomaterials 23 (2002) 4863–4871. [13] Y.W. Gu, B.Y. Tay, C.S. Lim, M.S. Yong, Characterization of bioactive surface oxidation layer on NiTi alloy, Appl. Surf. Sci. 252 (2005) 2038–2049. [14] M. Wiedmann, Self-expanding antireflux stent fracture, Endoscopy 41 (2009) E129–E130.
[15] A.H. Goenka, Esophageal migration of fractured SEMS fragment, Endoscopy 41 (2009) E204–E205. [16] D. Vojtˇech, L. Joska, J. Leitner, Influence of a controlled oxidation at moderate temperatures on the surface chemistry of nitinol wire, Appl. Surf. Sci. 254 (2008) 5664–5669. [17] R.M. Wang, C.L. Chu, T. Hu, C. Guo, X.B. Sheng, P.H. Lin, C.Y. Chung, P.K. Chu, Surface XPS characterization of NiTi shape memory alloy after advanced oxidation processes in UV/H2 O2 photocatalytic system, Appl. Surf. Sci. 253 (2007) 8507–8512. [18] C.W. Bale, P. Chartrand, S.A. Degterov, G. Eriksson, K. Hack, R. Ben Mahfoud, J. Melanc¸on, A.D. Pelton, S. Petersen, FactSage thermochemical software and databases, Calphad 26 (2002) 189–228. [19] K.S. Kim, K.K. Jee, W.C. Kim, W.Y. Jang, S.H. Han, Effect of heat treatment temperature on oxidation behaviour in Ni–Ti alloy, Mater. Sci. Eng. A 481–482 (2008) 658–661. [20] U. Diebold, The surface science of titanium dioxide, Surf. Sci. Rep. 48 (2003) 53–229. [21] L. Zhu, C. Trepanier, A.R. Pelton, Oxidation of Nitinol and its Effect on Corrosion Resistance, Proceedings of ASM Materials and Processes for Medical Devices Conference 2003, Nitinol Devices and Components, 2003, pp. 1–6.