The effect of temperature on the nucleation of corrosion pits on titanium in Ringer's physiological solution

G. T. Burstein ^{, , a}, C. Liu ^a and R. M. Souto <sup>b

a</sup> Department of Materials Science and Metallurgy, University of Cambridge, Pembroke Street, Cambridge CB2 3QZ, UK
^b Department of Physical Chemistry, University of La Laguna, La Laguna, Tenerife, Spain

Received 11 November 2003;  accepted 5 February 2004.  Available online 24 March 2004.
Biomaterials
Article in Press, Corrected Proof - Note to users

1. Abstract

This paper describes the effect of temperature on the nucleation of corrosion pits on titanium microelectrodes in Ringer's physiological solution. The results are shown for potentials far below the pitting potential, and describe breakdown of passivity with no permanent propagation of pits. Nucleation events could be observed at all the temperatures used, although they were very rare events at 20°C. The frequency of breakdown rises significantly with increase in temperature. Examples are shown of current transients due to both pit nucleation and to metastable pit propagation, the latter being rare events. Analysis shows that these events constitute a significant fraction of the passive corrosion rate of titanium.

Author Keywords: Author Keywords: Corrosion; Pitting corrosion; Titanium; Microelectrode; Passivation; Breakdown

2. Article Outline

1. Introduction

2. Experimental

3. Results

4. Discussion

5. Conclusion

Acknowledgements

References

3. 1. Introduction

Titanium-based implant materials show very high resistance to pitting corrosion in physiological solutions because of their state of passivity [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 and 13]. This state of passivity involves a slow corrosion rate, with a correspondingly slow release of corrosion product: the titanium released from the implant material enters the physiological environment. There is considerable discussion over the mechanism by which this occurs and the ultimate destination of the released titanium in vivo [1, 6, 8, 12 and 14]. It is clear however, that the corrosion rate is low: corrosion rates releasing titanium in vitro have been measured [7] in the range 0.01-0.1 
g cm⁻² d⁻¹, and this figure is in broad agreement in work from more than one laboratory [2]. The passive corrosion rate of titanium is affected strongly by the presence of some dissolved species, such as fluoride ions [2, 7 and 11] and hydrogen peroxide [9]: both enhance corrosion considerably. The presence of chloride ions in physiological solutions causes pitting of many metals, but titanium (particularly pure titanium) is relatively resistant to pitting corrosion in physiological solutions, as measured by the fact that the pitting potential is high. Nevertheless, in the presence of chloride ions, some microscopic breakdown events can be observed on titanium at potentials far below the pitting potential. These events do not lead to pitting failure, but repassivate, and the phenomenon is consistent with the behaviour of other metals. The present paper describes the results of experiments designed to measure these microscopic events on titanium in Ringer's physiological solution, and to interpret their occurrence and temperature dependence in terms of the passivity of the metal.

Pitting corrosion causes metal failure by creation of small holes in a surface that is otherwise passive. It is generally associated with specific species in solution, by far the most common of which is chloride. Conventionally, pitting is easy to identify in the laboratory by polarization measurements. When the pitting potential is exceeded, the measured current rises continuously with increasing potential because the pits propagate by anodic dissolution of the metal. However, pits can and do form below the pitting potential, and observations show that they not only nucleate below the pitting potential, but they can also propagate there, into the so-called metastable state [15, 16, 17 and 18]. The pitting potential really only describes that potential above which pits can propagate in a stable growth mode. Below the pitting potential, pit growth cannot stabilise and the metastably propagating pits formed there repassivate. It appears that there is no "pitting potential" for nucleation and metastable propagation; these states of pitting occur throughout the passive regime of the metal, as far as can be ascertained. Nevertheless, metastable pitting is the precursor stage for stable pitting [15, 16, 17 and 18] and certainly causes surface damage.

In an earlier paper [19], we showed a preliminary result reporting that no breakdown of passivity on titanium could be observed when microelectrodes were polarised in Ringer's physiological solution. Microelectrodes are employed in this type of experiment to allow the tiny current transients due to breakdown of passivity to be observed superimposed on the passive current [19, 20, 21, 22, 23, 24, 25 and 26]. These breakdown events are the initiation of destructive pitting corrosion, but they also take place under quite mild conditions where no general pitting is observed. For example, they take place on stainless steels throughout the passive region in chloride solutions, far below the pitting potential. In these conditions, even though the events occur, the metal remains in an overall passive state. Detected pitting events show nucleation, and some may show metastable propagation; the latter is distinguishable from the former by the form of the current transients they produce. The pitting potential does not dictate or control pit nucleation; neither does it characterise metastable propagation. The pitting potential can therefore only define the limiting state of stable propagation.

The issue of pit nucleation in physiological solutions at potentials below the pitting potential needs to be considered, partly because of the potential surface condition of the implanted metal, and partly to aid understanding of the mechanism of metal release into the adjacent tissue. The size of the pit nucleation event is very tiny, estimated in some systems to be of dimension around 10-100 nm [21, 22, 23 and 24]. These are depassivating events, which require loss of the passivating oxide film. It has been proposed as a film rupture event in order to describe the short time of activation [24]. In many practical circumstances, these pit nucleation events would be undamaging and probably unnoticed, provided they do not develop into stable pits. However, in circumstances where the surface integrity or morphology is important, or as in the case of an implanted material, where the corrosion product can enter the tissue system, the consequences of localised breakdown are not so clear [19, 25 and 26]. The long-term effects of titanium release has been queried by others [1, 6, 8 and 14]. We showed in earlier communications that titanium is much more resistant to depassivation than stainless steel, implying that stainless steel could be regarded as an inferior material for this potential damage. However, although the frequency of pit nucleation is very much lower for titanium, we also showed that for stainless steel, the frequency decays to zero after a time [24] for fixed experimental conditions set below the pitting potential; this arises because repetitive nucleation and metastable propagation slowly annihilate the sites of pit activity on the stainless steel surface. By contrast, the frequency of pit nucleation on titanium does not decay to zero with time [19, 22 and 26]. After a long time of exposure, the titanium surface still undergoes low-frequency breakdown, in circumstances where the stainless steel frequency has decayed to zero. The origin of this is explained in terms of pit sites. It appears that pits on stainless steels are nucleated only at specific sites, believed to be sulphide inclusions [17, 20, 21, 23, 24 and 26], at least for those stainless steels which are unsensitised. Once those sites have been removed from the surface, either by repetitive pit nucleation, or by metastable propagation, the sites have gone, and no further nucleation is possible [24]. The fact that the frequency on titanium does not decay to zero [19, 22 and 26] implies that the nucleation sites on this metal are not inclusions at all, and probably arise simply from properties of the passivating film itself. The model proposed invokes the notion that the passive film ruptures explosively on a microscopic scale because chloride accumulates at the metal/film interface; the model is a reasonable explanation for the observed events [24]. Each nucleation event is accompanied by (and indeed observed by) a sharp anodic current transient, demonstrating rapid localised metal oxidation.

The fact that titanium did not show pit nucleation at all in Ringer's physiological solution at room temperature [19] is important in the study of the behaviour of metals as implant materials, because it shows the excellent state of passivity of this material. The preliminary work [19 and 22] involved one run only however, and no statistical analysis of the data was available. Alloying titanium as Ti-6Al-4 V reduces the electrochemical stability induced by passivity, and the alloy showed some breakdown events in Ringer's solution at room temperature [19]. We have now examined the behaviour as a function of temperature, using a statistical number of runs to show that titanium does indeed break down in Ringer's solution at room temperature, although the events are rare. It turns out that raising the temperature to 37°C (and higher) gives rise to a higher pit nucleation frequency on titanium, which we show below.

Raising the temperature generally causes metals to pit more readily. The observation of a critical pitting temperature relates of course, to stable pitting [27, 28 and 29]. Pit nucleation and metastable propagation have not been much examined in terms of temperature dependence. The work of Park et al. [29] shows that a rise in temperature helps stabilise pit growth on stainless steel, and hinders repassivation. Since pitting on stainless steel is known to arise from sulphide inclusions, the temperature effect is easily ascribed to processes relating to that particular system. However, the role of temperature needs to be examined for other systems too, some of which are not known to give rise to pitting from second phase particles, as with the case of titanium in Ringer's solution, described below.

4. 2. Experimental

This work was carried out on circular surface titanium microelectrodes of 50 
m diameter. The electrodes were made by casting sleeves of epoxy resin around titanium wires to which an electrical lead had been soldered at one end. The other end, which comprised the working surface, was ground to a finish of 4000 grit with silicon carbide paper. This formed the electrode surface, of area 2×10⁻⁵ cm². Only by using such a small electrode could these pit nucleation events be observed. This is because the current that describes them is very small in amplitude, measured in the picoampere range, and the events can be observed only if the background current is very small. The background current is the passive current, and comes of course from the whole electrode. Despite the fact that the passive current density of titanium is small, a surface area which is large obscures the nucleation events completely because of the magnitude of the passive current. The processes described below could not be observed on a corrosion test specimen of the larger size as used more conventionally in corrosion testing.

The method of investigating pit nucleation on these microelectrodes involved two-electrode polarisation using a reference electrode as a simultaneous counter electrode, as described earlier [19, 20, 21, 22, 23, 24, 25 and 26]. The combined reference/counter electrode was a 22 cm² sheet of silver (44 cm² for both sides) anodised in HCl solution to produce a film of AgCl, and equilibrated with the working electrolyte composition. This in combination with a small working electrode functions as a potentiostatic circuit when polarised in a two-electrode circuit, because the small currents from the small working electrode do not polarise the Ag/AgCl counter/reference electrode. This in turn allows potentiostatic polarisation of the working electrode without using a potentiostat, thereby minimising extraneous electronic noise. Polarisation was carried out using a bank of dry cell batteries linked to a voltage divider. The electrochemical cell, together with the voltage divider and battery circuit, were housed in a concentric pair of earthed Faraday cages. These Faraday cages lay immediately alongside the current amplifier which was also housed in a separate earthed Faraday cage. The three Faraday cages were all short-circuited. The current amplifier functioned as a zero-resistance ammeter, and was operated typically at 10¹⁰ V A⁻¹. The output of the current amplifier was fed into a digital voltmeter and thence into a computer. The rate of data acquisition was a compromise between recording at a fast rate and acquiring data for a sufficient period of time in a single run. Mostly, the data were recorded at 13.5 Hz, giving a time resolution of 74 ms. The applied potential was 0.500 V against the (Ag/AgCl/0.16 
Cl⁻) reference electrode, equivalent to 0.769 V(SHE) or 0.524 V(SCE).

Ringer's solution was made to composition 0.147 
NaCl, 0.00432 
CaCl₂, and 0.00404 
KCl, giving a total chloride concentration of 0.16 
, in twice-distilled water.

Experiments were run at temperatures of 20°C, 37°C and 50°C. The lowest of these temperatures was simply room temperature, and uncontrolled; the accuracy here was ±1°C for the duration of a single experiment, but the variability over the days required for making multiple runs was ±2°C. For the two higher temperatures, the electrolyte was heated by an electric wire-wound immersion heater (home-built). The cell was not thermostatically controlled, because this would have involved electrical switching transients generated from the thermostat which could have induced noise onto the measured currents. These experiments involve very sensitive measurement of very small current transients in the picoampere range, and elimination of extraneous noise is vital to their success. Thus the required temperature was reached by simply controlling a constant d.c. current through the heater so that the system achieved a steady state at that desired temperature. With the heater on continuously, no artefacts from its constant current were introduced into the system, but temperature control was also not as accurate as might have been achieved by electronic control. By using this method, the temperature accuracy and variability were estimated to be ±1.5 at 37°C and ±2°C at 50°C. The power source for the heating element was a bank of dry-cell batteries rather than an electronic supply; the purpose of using batteries was again the minimisation of extraneous noise and a.c. interference. The combined reference/counter electrode was held in a sidearm to the main cell, linked by an electrolyte bridge. This reference compartment was not heated, and the temperature of the reference electrode was therefore ambient.

Because of the stochastic nature of the noise events being logged, there is much scatter in results as shown before [17, 19, 20, 21, 22, 23, 24, 25 and 26]. Experiments were therefore repeated many times under identical conditions; between 12 and 16 repeat experiments were carried out.

5. 3. Results

Two examples of the current train which arise from polarising titanium in Ringer's solution containing sharp anodic current spikes, are shown in Fig. 1, for temperatures of 20°C and 37°C. The overall form of the traces is a decay in current continuously with time as the metal passivates by oxide film growth. The decay of current is occasionally interrupted by brief transient surges in the anodic direction, indicating pit nucleation. These are much more prominent at the higher temperature shown in Fig. 1. Initially, the overall current is high and causes overload of the current amplifier; for the present experiments, this occurred when the current was greater than ca. 1.2 nA under the amplification used. The overload range is not plotted in Fig. 1. This initial brief period could not be further analysed. Because of the small sizes of the anodic current spikes, Fig. 1 is shown on a rather expanded current scale, with a maximum of 150 pA, enabling the transient current spikes to be discerned. The current spikes, which arise from pit nucleation, can be distinguished from the background noise by inspection through the fact that pit nucleation is a uniquely anodic event (when observed under potential control, as here). Thus the pit current spikes lie exclusively in the anodic direction, whereas the general background noise lies equally in the anodic and cathodic directions, i.e., above and below the mean of the data. The anodic current spikes are very small. It is easy to see from the data shown in Fig. 1 that there are more anodic current transients at higher temperature. Note that the higher temperature also imparts a higher passivating current; the entire current trace is at higher current at 37°C than at 20°C. The 50°C traces (not shown) were higher still. The clearest way to show the effect of temperature on the current spikes is to plot the frequency of occurrence of these events. Before the frequency can be measured, the current spike events which are due to pit nucleation must be separated from the background noise, as described below.

(7K)

Fig. 1. Current from 50 
m diameter titanium specimens at a potential of 0.524 V (SCE) in Ringer's solution showing passivation with time and superimposed current spikes due to pit nucleation. Examples measured at temperatures of 20°C (lower trace) and 37°C (upper trace).

Current spikes were counted using an imposed amplitude threshold: for a given current threshold, all current spikes above that value are counted, and those below it are ignored. The threshold was imposed by setting a counting programme to detect and record the condition when two adjacent points in the data set differed by an amount greater than the threshold. By imposing separately, sequentially increasing thresholds, and plotting the number of events counted against the imposed threshold, the anodic current spikes were separated from the background noise. The method has been described and illustrated before [17, 19, 21, 22, 24 and 26]. The procedure means that pit nucleation current spikes of amplitude comparable with or smaller than the background noise were not counted because they could not be incorporated unambiguously.

The frequency of events was then counted as a function of time. This was done by counting anodic current spikes (of amplitude above the threshold) over a specified time range, and then dividing the number by that time span. The recording time of 39.5 min was divided into five equal parts for this purpose. The first 250 s were eliminated because the current amplifier was in an overload condition for approximately this period as described above. The result is shown in Fig. 2, where each data point is presented in the midpoint of its time span. Each data point here represents the mean of 12 identical runs, and the error bars are the 95% confidence limits. The scatter in the data that occurs from run to run is large, reflecting chiefly the stochastic nature of pit nucleation. Small errors or changes in the temperature that result from the lack of direct thermostatic control may add to the scatter. Note that the left-hand ordinate of Fig. 2 shows the frequency as measured, and the right-hand ordinate shows the frequency density, i.e. the frequency per unit area. We see from Fig. 2 that although the frequency of occurrence of anodic current spikes in Ringer's solution is very low at room temperature (20°C), when the temperature is higher, the frequency of events is also higher. This result is quite unambiguous, despite the relatively large scatter bands. The preliminary result reported earlier for titanium in Ringer's solution [19 and 22] at 20°C is consistent with Fig. 2. Here however, we have carried out a sufficient number of experiments to show that the frequency at 20°C is close to, but not exactly zero. To emphasize this point, it is to be noted that there are two data points in the 20°C data which lie at zero frequency with no scatter band: these two points gave zero frequency in all 12 identical runs; doubtless, a much greater number of runs would have yielded a low finite value here.

(8K)

Fig. 2. Frequency of pit nucleation on titanium as a function of time of exposure in Ringer's solution at 0.524 V (SCE). Each data point is the mean of 12 runs measured identically, with the scatter band showing 95% confidence limits in the value. Data were counted over a period of 500 s per point and are presented at the midpoint of the time interval they represent. The left-hand ordinate is the frequency measured over the 50 
m diameter wires. The right-hand ordinate shows the same data as the frequency density, normalised with respect to the surface area of the specimen.

Fig. 2 also shows that the frequency of occurrence of pit nucleation spikes on titanium may decay a little initially with time of polarisation, but it does not carry on decaying indefinitely to zero; rather, it approaches a steady non-zero state. Only the initial region shows decay in frequency, to about 1000 s. Thus it must be concluded that depassivation events continue indefinitely, as the preliminary data also showed [19 and 22]. The fact contrasts the behaviour of stainless steels in chloride solutions, where the frequency does indeed decay with time, and reaches zero eventually [24] (for constant potential). In order to show this temperature effect more clearly, the mean frequency of current spikes due to pit nucleation was measured over the period of time greater than 1000 s, and this is plotted as a function of temperature in Fig. 3. This (approximately) steady-state frequency increases quite strongly with increasing temperature. (The reason for excluding the first 1000 s from the analysed data for Fig. 3 arises because of some decay in frequency that occurs here, as described above).

(6K)

Fig. 3. The mean frequency density of pit nucleation on titanium in Ringer's solution as a function of temperature. The data are those measured from the mean of the last four data points in Fig. 2, i.e. over the last 1500 s, together with the standard deviation of that mean.

Examples of current transients are shown in Fig. 4; all the results shown were measured at 37°C. The transient in Fig. 4(a) displays the classical form for pit nucleation: a sharp rise in current initiated by the event is followed by a continuous slower decay. This form of transient is similar to that shown earlier for pit nucleation on titanium and for stainless steel [19, 20, 21, 22, 23, 24 and 26]. Transients of this type are reckoned to be due to nucleation only, with no evidence of propagation because of the continuous decay in current back to steady-state passivity. However, some metal dissolution during repassivation cannot be ruled out.

(15K)

Fig. 4. (a) Current transient (left-hand ordinate, solid line) due to pit nucleation of titanium in Ringer's solution at 37°C. The pit is nucleated at 874.5 s and the current then shows decay only towards the background; the current decays through repassiavtion. Also plotted is the total anodic charge accumulated under the transient as a function of time (right-hand ordinate, broken line). The time scale is real time since the start of polarisation. (b) Current transient (left-hand ordinate, solid line) due to nucleation and metastable propagation of a pit on titanium in Ringer's solution at 37°C. The pit is nucleated at 2341 s. The current reaches a maximum of 204-205 pA, maintained approximately for 1.2 s, before decaying back to the background through repassivation. The background current is 182 pA, giving a plateau propagation current of 22-23 pA. Also plotted is the total anodic charge accumulated under the transient as a function of time (right-hand ordinate, broken line). The time scale is real time since the start of polarisation. (c) Current transient (left-hand ordinate, solid line) due to nucleation of a pit and its metastable propagation on titanium in Ringer's solution at 37°C. The pit is nucleated at 82.7 s and the spike rises to a peak at 82.8 s. Fast repassivation occurs only as far as 83 s. The pit then propagates briefly reaching another current peak at 83.1 s before decaying more slowly by repassivation. Also plotted is the total anodic charge accumulated under the transient as a function of time (right-hand ordinate, broken line). The time scale is real time since the start of polarisation.

Some current transients showed distinct evidence of propagation into the metastable pit growth state. These were rare, and displayed only by the higher temperature runs; two such transients generated at 37°C are shown in Figs. 4(b) and (c). The transient in Fig. 4(b) shows the sharp nucleation similar to that in Fig. 4(a), but in this case, the current remains constant for a period of ca. 1.5 s before decaying. The fact that the current does not decay in the flat region is evidence of the fact that the pit propagates by metal dissolution briefly. Anodic charge generated by the transient must either reform the passivating film, in which case the current decays, or it must represent dissolved metal: charge due to dissolution cannot cause significant decay in current. In Fig. 4(c) an alternative transient due to metastable pit propagation is plotted. In this case, the current transient again starts with a sharp nucleation event, following which repassivation commences, as in Fig. 4(a). However repassivation is interrupted by a small surge in anodic current, peaking at 83.1 s, and a succeeding much slower current decay. This is evidence that the pit nucleated here starts initially to repassivate, but then also propagates briefly, before it too finally succeeds in repassivating.

These data show that pit nucleation occurs on titanium in Ringer's solution, and these nucleation events can give rise to metastable pit propagation, even below the pitting potential. All pit nucleation transients observed resulted in repassivation, even those which showed metastable propagation, and none propagated to become a stably growing pit. (A stably growing pit would result in a current transient that does not return to the steady passive state, but shows a permanent rise with time as propagation enlarges the reacting surface area.)

Note that the true rise time representing the initiation of these current transients was not resolved by recording the data at the rate of 13.5 Hz. Each of the current transients shown in Fig. 4 had a fast rise time. In the present experiments we aimed to collect data over a sufficient time in each run to measure the frequencies of events. The set of data recorded by the equipment was limited by the total available memory. There is however another reason for tracking the pit nucleation events in this way. The background noise, which limits the resolution of the smallest transients, shows greater amplitude at faster data acquisition rates, simply because this too, becomes better resolved. Thus the threshold for identifying pit nucleation transients and distinguishing them from the background noise is raised if the data acquisition rate were raised. The actual data acquisition rate used in the present experiments has two consequences in terms of interpretation. First, the amplitude of the initiation event was truncated by not resolving the rise time fully. Second, nucleation events which were inherently of low amplitude or very fast, would not have been detected at all. For the purposes of the present work, the experimental conditions imposed provided the best compromise for logging the data.

To give an idea of the amount of reaction that occurs in the current transients, the transient current in each was integrated to produce the anodic charge passed during the event as a function of time. Thus the graphs in Figs. 4(a)-(c) also show the integrated charge passed. Integration of the data to provide measurement of the charge requires insertion of a suitable background current. The background was measured by taking the mean value of the noise level immediately before nucleation, and immediately after repassivation. Note that the charge measured is the total charge passed under the transient only; it does not include the charge that passes simultaneously by virtue of the passive current flowing over the rest of the surface. Thus the pit event has finished when its anodic charge reaches a constant value, independent of time. The charge due to nucleation only in Fig. 4(a) is very small indeed, reaching an observed maximum of only 1.1 pC. Bigger transients of this type were also observed, with charges running into tens of pC. Smaller ones were also observed, of charge <1 pC. The metastable pitting events of Figs. 4(b) and (c) show much greater anodic charge evolved, reflecting the amount of dissolution of titanium that occurs prior to repassivation.

Using the measured charge in Figs. 4(a)-(c), it is possible to calculate the volume of metal reacted, V, using Faraday's law, given the density of titanium of 4.5 g cm³, and assuming oxidation to be as Ti(IV). The results are shown in Figs. 5(a)-(c) respectively. For these graphs the time scales of Fig. 4 have been adjusted so that the new origins of time and current are at the start of the transient. The nucleation transient of Fig. 5(a) shows a volume of metal reacted of 3×10⁻¹⁷ cm³, which is very small. The shape of the reacting surface site is unknown. However, by assuming that the charge represents regrowth of a passive film to a thickness of, say, 10 nm, then the area of the site would be 3×10⁻¹¹ cm², and if it were circular and flat, the exposed site would be of radius 30 nm. If the oxide thickness were 2 nm, then the assumed flat, circular site would have radius 69 nm. Neither of these is an accurate measure, but they do give an idea of the dimension of the event. For the growth of the metastable pits, the reacted volumes are of course larger than that of the nucleation event only, as plotted in Figs. 5(b) and (c). Thus the maximum volumes shown, when the sites have repassivated fully are 4.5×10⁻¹⁵ from Fig. 5(b) and 7.3×10⁻¹⁶ cm³ from Fig. 5(c). By assuming that dissolution of metal from these minute metastably propagating pits involves generation of a hemispherical metastable pit morphology (as shown for stainless steel [17]), the radius of the pit can also be calculated. The radius is plotted in Fig. 5(b) as well, assuming a hemispherical geometry for the entire transient. It is also possible to calculate the surface area of the hemisphere as it grows, assuming the hemispherical geometry, and thence obtain the current density from the current. This is plotted for the metastable pit transient in Fig. 5(c). The maximum rate of reaction is large, around 0.8 A cm⁻², and is the sort of magnitude expected for the current density from a propagating pit, but it is clearly insufficient to allow propagation of the pit into the stable state. Note that by using the hemispherical assumption for the metastable pit morphology, the maximum radius of the hemispheres calculated when the repassivation is complete is 130 nm and 70 nm, respectively for Figs. 5(b) and (c). Although these metastable pits repassivate fully, the sizes of the corroded sites do provide surface damage.

(14K)

Fig. 5. (a) The volume of metal reacted, V, (right-hand ordinate, broken line) under the pit nucleation transient shown in Fig. 4a. Also shown is the current transient itself (left-hand ordinate, solid line). The time and current scales have been recalibrated to give the point of initiation as the origin. Note that the volume scale is 10¹⁷V cm³. (b) The volume of metal reacted, V, (left-hand ordinate, solid line) under the current transient in Fig. 4b. Also shown is the equivalent radius that this would represent if the reaction produced a hemispherical metastable pit (right-hand ordinate, broken line). (c) The volume of the pit, (right-hand ordinate, broken line) assuming it grows as a hemisphere, calculated from the charge passed under the metastable pit current transient shown in Fig. 4c. Also shown is the current density flowing from this metastable pit (left-hand ordinate, solid line) calculated by dividing the current transient from Fig. 4c by the surface area of the metastable pit, assumed hemispherical. The time and current scales have been recalibrated to give the point of initiation as the origin.

6. 4. Discussion

We report here that nucleation of corrosion pits on passive titanium in Ringer's solution occurs at potentials well below the pitting potential. The pitting potential at room temperature has been measured to be >2 V (SCE) in the much more aggressive 1.5 
HCl [22]. These nucleation and metastable propagation events are microscopic; they occur on the metal surface but do not cause major structural damage to the metal because of their small dimension. They do not propagate to become major pits, although some nucleations may propagate briefly in the metastable state, as shown; the majority show no evidence of propagation. They nevertheless represent a surface instability of the passive state of the metal. Although we reported in our preliminary communication [19] that none was observed on commercially pure titanium in Ringer's solution at room temperature, the present result shows that some small signs of breakdown can be observed if sufficient runs are done. Many of the individual runs made at 20°C in the present work also showed no evidence of breakdown, but a few did show some activity. Important however, is the fact that raising the temperature shows clear sign of transient breakdown in passivity, albeit that each event is very small, and none propagates into a stably growing pit. At 37°C there is a considerable frequency of breakdown; the small absolute numbers involved in Fig. 2 arise of course, from the very small surface area of the specimen electrodes. The quantitative effect of temperature might be expected to follow an Arrhenius form. Although the data are limited, we show an Arrhenius plot in Fig. 6, together with a linear regression of the data in Arrhenius form. The graph shown in Fig. 6 has been generated by measuring the mean of each set of constant temperature data given in Fig. 2, except for the first data point. The assumption is that there is some small decay in frequency with time initially, as shown in Fig. 2, but the frequency has reached an approximate steady state after ca. 1000 s. The importance of showing this is to demonstrate, at least approximately, the high sensitivity of the breakdown frequency to temperature. The data point for a temperature of 37°C shows a frequency of 89.6 Hz cm⁻². The same temperature from the linearly regressed line gives a frequency of 68.2 Hz cm⁻². By raising the temperature to say, 40°C, the frequency is increased to 91.9 Hz cm⁻²; thus this small rise of temperature of only 3°C can cause an increase in the frequency of breakdown events of 35%, which is significant.

(9K)

Fig. 6. Arrhenius plot of the temperature dependence of the pit nucleation frequency density shown in Fig. 3 (black points). Also shown is an Arrhenius plot of the background passivation current in pA (white points). The background current was measured as the mean of the background noise in the last 100 points of each of the 12 identically measured runs for each temperature. Error bars show the standard deviation in the passivating current, and the 95% confidence limits of the mean frequency. The broken lines in both graphs are the linear regressions.

We have observed that at 37°C, titanium in Ringer's solution shows both metastable pit events as well as transients believed to be nucleation only. The propagation of metastable pits is a titanium dissolution process. Although these pits do not propagate far, and the metastable events are rare, they do nonetheless give a finite local titanium ion concentration in solution and must create a surface indentation. Clearly, there is facility to develop a significant number of sub-micrometre surface pits over a long period of time. The nucleation event itself also involves ejection of some small amount of oxidised titanium into the local solution, since its nature is that of depassivation. At the least, the passivating oxide must have broken down, and the broken residue must lie within the adjacent solution, presumably as minute particles of the solid oxide.

Unlike stainless steel, titanium shows continuing breakdown, for which the frequency eventually appears to reach a non-zero steady state. Although titanium shows a much smaller frequency of breakdown than 304L grade stainless steel [19 and 22], and can therefore be regarded as much more resistant to pit nucleation, the fact that these events carry on over long periods implies a potential cumulative effect in circumstances where the metal is employed in a closed system. For a stainless steel which has not been sensitised through localised chromium depletion, the breakdown of passivity by these pit nucleation processes appears to come only from discrete sites, normally reckoned to be sulphide inclusions [17, 18, 20, 21, 22, 23, 24, 25, 26 and 27]. Once these sulphide inclusions have been removed, either by metastable pit propagation, or by repetitive nucleation, it might then be expected that the passivity of the stainless steel would no longer be interrupted; this has indeed been observed [24] (although experiments in that work were carried out only over a very limited range of conditions). We have ascribed the process of pit nucleation on stainless steel to the rupture of the passivating film, caused by chloride accumulation at the metal/passive film interface [19, 20, 21, 22, 23, 24, 25 and 26].

Whatever the microstructural and microcompositional origins of pit nucleation in stainless steels, the same type of origin cannot be ascribed to pit nucleation at titanium. In stainless steels, chromium is the passivating element, and its concentration in the bulk metal is relatively low. In the case of titanium, passivity is provided by the oxide film of titanium, (not by a relatively low-concentration alloying element) and breakdown must occur in this oxide. The notion that pits do not appear to be generated at specific sites of metallurgical heterogeneity arises from the fact that the nucleation events do not disappear with time of exposure, and therefore they cannot eliminate sites by repetitive nucleation. We propose here, that growth of the passivating titanium oxide film involves migration of oxide ions through it, presumably via oxide anion vacancies. On reaching the metal/film interface, new oxide is formed at that interface. In the presence of chloride ions, Cl⁻ can migrate in parallel across the passivating oxide. Thus, if Cl⁻ reaches the metal/film interface, it would form the metal chloride. If sufficient metal chloride were to accumulate there, the oxide would rupture explosively, and reveal a bare metal surface, and form a microscopically saturated chloride solution. The origin of the explosive film rupture is the fact that metal chlorides generally have greater molar volumes than the corresponding oxides, and formation of the chloride salt at the metal/passive film interface must engender stresses which would tend to cause blistering in the passivating film. This is certainly true of the titanium chlorides, for which the molar volume is calculated as 38, 58.4 and 110 cm³ mol⁻¹ respectively for TiCl₂, TiCl₃ and TiCl₄ from published density data [30]. This compares with 18.8 cm³ mol⁻¹ for TiO₂ (rutile form). In this model, the pit is nucleated when the blister ruptures.

If, after pit nucleation at the site, sufficient chloride remains at the metal surface without total dilution for a sufficient period of time, metal dissolution continues, and the current transient then shows a nucleation event followed by metastable propagation. Propagation is caused by the fact that once the oxide film has been ruptured, the chloride which caused it becomes exposed to the electrolyte, and produces locally a high chloride concentration, with hydrolysis allowing the solution to acidify locally. The metal dissolves in the locally high chloride concentration. Repassivation ensues when the localised saturated chloride solution has been diluted. This is the model for the transient in Fig. 4(b). If however, the localised chloride solution dilutes immediately on exposure following passive film rupture, the resulting current transient shows nucleation only, with subsequent continuous decay back to the steady state as the site repassivates. This type of current transient is that depicted in Fig. 4(a).

The model fits all the available information. Assuming that the oxide film does indeed grow by anion migration across it, then it is clear that provided chloride is present, it must be possible for this anion to migrate in parallel with oxide ions. For a given passive current density, one would expect the rate of chloride migration across the passivating oxide film to be related to the chloride concentration as well as to the concentration of anion vacancies in that oxide film. The fact that these aqueous solutions are overwhelmingly water means that oxide ions are overwhelmingly available, and chloride migration must be comparatively rare. The accumulation of chloride at the metal/oxide film interface is only possible when there is sufficient available. In dilute solution, chloride ions reaching the metal/film interface do not necessarily cause damage under this model, since as the oxide film continues to grow, slowly at the passive current density, those chloride ions which reach that interface can be forced to leave the interface as the oxide grows around them. Such ions would then become embedded in the oxide itself, ultimately reaching the solution again, and cause no damage; they could not initiate a pit in that state. Indeed, some chloride anions which become incorporated into the passivating film may never even reach the metal/film interface, but remain embedded in the oxide film until they are lost by steady-state dissolution. The model requires that sufficient chloride accumulates at one specific spot in order for film rupture to occur.

The temperature effect would then arise because chloride ions must migrate across the passivating oxide film faster at higher temperature, as indeed must the oxide ions too. It is certainly true that the passivating currents are higher at higher temperatures (see Fig. 1). Although the final passive current was not determined, an idea of the comparison can be obtained by plotting the current at the end of each train (after 2400 s) against temperature. Because of the noise in the data, this was measured by taking the mean of the last 100 data points in each trace, representing the last 7.4 s of time, and plotting this as a function of T. The graph is shown in thje Arrhenius form in Fig. 6, together with the standard deviation in the data. The rise in passivating current with increase in T is clear. As with the pit nucleation frequency, the data can be fitted to an Arrhenius plot, and interestingly, the gradients of the two lines of Fig. 6 are not dissimilar. Thus for the passive current, ∂log I/∂T⁻¹=−3.4 kK from Fig. 6 (where I is the passivating current and T is temperature). The gradient for the nucleation frequency from Fig. 6 gives ∂log f/∂log T⁻¹=−4.2 kK (where f is the frequency). This correlation between the two Arrhenius gradients of Fig. 6 must however, be taken with caution. The inaccuracies of the three-temperature measurement for the Arrhenius plots, and the fact that many processes which are quite unrelated can have similar activation energies, can both cause this apparent similarity to be fortuitous. In this particular case, although the higher temperature would clearly allow faster migration of Cl⁻ ions across the film, it would also allow correspondingly faster O²⁻ across the films as well. The notion that the nucleation frequency might itself be related to the passive current density cannot be dismissed. However, we note that the overall passivating current plotted in Fig. 1 has not reached a steady state, even after 39.5 min polarisation, and the current continues to decay. The frequency of pit nucleation does appear to have reached a steady state, within the same period (at least within the scatter bands of the data) as shown in Fig. 2. The scatter bands are also large. The suggested correlation between pit nucleation rate and the passive current density can be established only with further experimentation.

The results described above can be used to get a rough idea of the number of events that would occur on titanium over a long period of time. Although only an approximation, a frequency of 68.2 Hz cm⁻² (from the regressed line read at 37°C in Fig. 6) for nucleation events would represent 2×10⁹ events per cm² in 1 year. Assuming that the events occur randomly over the surface, and there are no specific sites involved, this would be likely to involve a significant fraction of the entire surface. For example, if the mean radius over all events (each assumed circular) were 100 nm, the area covered by nucleation events would be 31% of the surface, assuming no repetitive nucleations at the same site. The figures should also be considered in the context of the known passive corrosion rate of titanium in physiological solution. If we consider the mean charge evolved in each nucleation event to be, say, 10 pC (see Fig. 4a), then the total anodic charge evolved by pit nucleation in 1 day would be 59 
C cm⁻². This represents 7.3 ng cm⁻² d⁻¹. This estimate is a significant fraction of the 10-100 ng cm⁻² d⁻¹ overall titanium release rate (in the passive state) shown by Strietzel et al. [7] and others [2]. Bearing in mind that we are unable to detect the smallest events (<2 pA in amplitude), it may represent a major part of the overall passive current. We must deduce therefore, that a significant part of the passive current density of titanium in these physiological solutions may be due to the microscopic breakdown of passivity induced by the chloride ions. The notion is consistent with the similarity in the Arrhenius gradients (from the plots in Fig. 6) ∂log I/∂log T⁻¹ and ∂log f/∂log T⁻¹ described above. This then, provides a new insight into the mechanism by which Ti may be released into tissue from implant materials [1, 6, 7, 12 and 14], even in the absence of abrasion or wear.

Breakdown of passivity on this microscopic scale should be considered in relation to the behaviour of implant alloys. The experiments described above are designed to measure microscopic events. These events lie in the range of up to hundreds of picoamperes in amplitude, but also down to the detection limit here of 2 pA. There are likely to be smaller events as well, which are undetected here because of the background noise in the experimental system. The breakdown events are fast. Each event involves formation of oxidised titanium, and in vivo, this must enter the adjacent tissue, at least initially. Although the frequency of breakdown is low, the adjacent tissue could bear some sensitivity, or provide an extra route through which the corrosion product can be dispersed in vivo. The distribution of corrosion and wear debris after implant has been discussed, but the long-term effects of these are not known [14]. The work of Bianco et al. [6] suggests that the titanium release by corrosion only will preferentially accumulate in local tissue. Other work [8, 12 and 14] suggests its wider distribution. These breakdown events do not lead to permanent breakdown by pitting, but they do generate surface defects and provide a mechanism of slow release of oxidised metal. It is impossible to detect these processes by conventional corrosion testing, where large specimens provide a relatively large background passive current. The long-term effect of these microscopic events requires further investigation. We note too, that coating titanium with hydroxyapatite is unlikely to prevent this microscopic breakdown, since pores in these coatings still allow corrosion processes to proceed [31].

7. 5. Conclusion

Breakdown of the passivity of titanium by nucleation of corrosion pits occurs in Ringer's solution at quite modest electrode potentials, well below the pitting potential. The frequency of breakdown is very low at ambient temperature, but increases significantly with increase in temperature. The frequency does not decay to zero with time, but appears to approach a steady state. Below the pitting potential, none of these pit nucleation processes develops into a stably propagating pit; some of them are capable however, of developing into metastably growing pits, particularly at higher temperatures. The very slow overall release rate estimated from the data is consistent with previously measured release rates.

The data are consistent with a model in which pit nucleation is caused by migration of chloride ions across the film in parallel with oxide ions. Accumulation of metal chloride (or perhaps in the case of titanium, metal oxychloride) at the metal/film interface may cause oxide film rupture as the nucleation event. The ability to propagate a metastably growing pit then depends on the amount of chloride accumulated at the nucleated site. The process provides a mechanism of surface damage in titanium implant systems, in which surface integrity may be affected and corrosion product be released. It is suggested that a significant fraction of the passive current density of titanium in physiological solutions may be due to the microscopic breakdown process. As such they are of fundamental importance to the mechanism of passivity of titanium in physiological systems.

8. Acknowledgements

We are grateful to the EPSRC for the financial support of this research programme. Contributions from the British Council and the Acciones Integradas Programme are also gratefully acknowledged.

9. References

1. J.J. Jacobs, A.K. Skipor, J. Black, R.M. Urban and J.O. Galante, Release and excretion of metal in patients who have a total hip replacement component made of titanium-base alloy. J Bone Jt Surg 73A (1991), pp. 1475-1486. Abstract-MEDLINE | Abstract-EMBASE  

2. A. Wisbey, P.J. Gregson, L.M. Peter and M. Tuke, Effect of surface treatment on the dissolution of titanium-based implant materials. Biomaterials 12 (1991), pp. 470-473. Abstract

3. K.E. Healy and P. Ducheyne, Passive dissolution kinetics of titanium in vitro. J Mater Sci: Mater Med 4 (1993), pp. 117-126. Abstract-EMBASE | Abstract-Compendex  

4. J.L. Ong, L.C. Lucas, G.N. Raikar and J.C. Gregory, Electrochemical corrosion analyses and characterization of surface-modified titanium. Appl Surf Sci 72 (1993), pp. 7-13. Abstract

5. J. Pan, D. Thierry and C. Leygraf, Electrochemical impedance spectroscopy study of the passive oxide film on titanium for implant application. Electrochim Acta 41 (1996), pp. 1143-1153. Abstract | Abstract + References | PDF (896 K)

6. P.D. Bianco, P. Ducheyne and J.M. Cuckler, Titanium serum and urine levels in rabbits with titanium implants in the absence of wear. Biomaterials 17 (1996), pp. 1937-1942. SummaryPlus | Full Text + Links | PDF (795 K)

7. R. Strietzel, A. Hösch, H. Kalbfleisch and D. Buch, In vitro corrosion of titanium. Biomaterials 19 (1998), pp. 1495-1499. Abstract | PDF (308 K)

8. J.J. Jacobs, A.K. Skipor, L.M. Patterson, N.J. Hallab, W.G. Paprosky, J. Black and J.O. Galante, Metal ion release in patients who have had a primary total hip arthroplasty: a prospective controlled longitudinal study. J Bone Jt Surg Am 80 (1998), pp. 1447-1458. Abstract-EMBASE | Abstract-MEDLINE  

9. C. Fonseca and M.A. Barbosa, Corrosion behaviour of titanium in biofluids containing H₂O₂ studied by electrochemical impedance spectroscopy. Corros Sci 43 (2001), pp. 547-559. SummaryPlus | Full Text + Links | PDF (900 K)

10. A.W.E. Hodgson, Y. Mueller, D. Forster and S. Virtanen, Electrochemical characterisation of passive films on Ti alloys under simulated biological conditions. Electrochim Acta 47 (2002), pp. 1913-1923. SummaryPlus | Full Text + Links | PDF (237 K)

11. N. Schiff, B. Grosgogeat, M. Lissac and F. Dalard, Influence of fluoride content and pH on the corrosion resistance of titanium and its alloys. Biomaterials 23 (2002), pp. 1995-2002. SummaryPlus | Full Text + Links | PDF (505 K)

12. N.J. Hallab, A. Skipor and J.J. Jacobs, Interfacial kinetics of titanium- and cobalt-based implant alloys in human serum: metal release and biofilm formation. J Biomed Mater Res Part A 65A (2003), pp. 311-318. Abstract-Compendex | Abstract-MEDLINE   | Full Text via CrossRef

13. M. Koike, Z. Cai, H. Fujii, M. Brezner and T. Okabe, Corrosion behavior of cast titanium with reduced surface reaction layer made by a face-coating methods. Biomaterials 24 (2003), pp. 4541-4549. SummaryPlus | Full Text + Links | PDF (1031 K)

14. Urban RM, Jacobs JJ, Gavrilovic J, Tomlinson MJ, Black J, Turner TM, Galante JO. Dissemination of metal-alloy particles to the liver and spleen with total hip or knee replacement prostheses. Proceedings of the Fifth World Biomaterials Conference (May 29-June 2 1996), vol. 1. Toronto: University of Toronto Press; 1996. p. 660.

15. G.S. Frankel, L. Stockert, F. Hunkeler and H. Böhni, Metastable pitting of stainless steel. Corrosion 43 (1987), pp. 429-436. Abstract-INSPEC  

16. H. Böhni and L. Stockert, The importance of metastable pitting corrosion in the case of high-alloy steels. Werkst Korros 40 (1989), pp. 63-71.

17. P.C. Pistorius and G.T. Burstein, Metastable pitting corrosion of stainless steel and the transition to stability. Philos Trans Roy Soc Ser A 341 (1992), pp. 531-559. Abstract-INSPEC  

18. H.S. Isaacs, The localized breakdown and repair of passive surfaces during pitting. Corros Sci 29 (1989), pp. 313-323. Abstract

19. R.M. Souto and G.T. Burstein, A preliminary investigation into the microscopic depassivation of passive titanium implant materials in vitro. J Mater Sci: Mater Med 7 (1996), pp. 337-343. Abstract-EMBASE | Abstract-Compendex  

20. G.T. Burstein and S.P. Mattin, Nucleation of corrosion pits on stainless steel. Philos Mag Lett 66 (1992), pp. 127-131. Abstract-INSPEC  

21. G.T. Burstein, S.P. Mattin and P.C. Pistorius, The nucleation and growth of corrosion pits on stainless steel. Corros Sci 35 (1993), pp. 57-62. Abstract

22. G.T. Burstein and R.M. Souto, Microscopic instability in the passivity of titanium in chloride solution. Electrochim Acta 40 (1995), pp. 1881-1888. Abstract | Abstract + References | PDF (910 K)

23. G.T. Burstein and S.P. Mattin, Detailed resolution of microscopic depassivation events on stainless steel leading to pitting. Philos Mag Lett 76 (1997), pp. 341-347.

24. G.T. Burstein and S.P. Vines, Repetitive nucleation of corrosion pits on stainless steel and the effects of surface roughness. J Electrochem Soc 148 (2001), pp. B504-B516. Abstract-INSPEC   | Full Text via CrossRef

25. CLiu, GT Burstein. Insights into passive dissolution of titanium in vitro. Proceedings of the Fifth World Biomaterials Conference (May 29-June 2, 1996), vol 1. Toronto: University of Toronto Press; 1996. p. 555.

26. Burstein GT, Liu C, Souto RM, Vines SP. Origins of pitting corrosion. Corros Eng Sci Technol, submitted for publication.

27. N.J. Laycock and R.C. Newman, Temperature dependence of the pitting potential for austenitic stainless steels above the critical pitting temperature. Corros Sci 40 (1998), pp. 887-902. SummaryPlus | Full Text + Links | PDF (458 K)

28. M.H. Moayed, N.J. Laycock and R.C. Newman, Dependence of the critical pitting temperature on surface roughness. Corros Sci 45 (2003), pp. 1203-1216. SummaryPlus | Full Text + Links | PDF (387 K)

29. J.O. Park, S. Matsch and H. Böhni, Effects of temperature and chloride concentration on pit initiation and early pit growth of stainless steel. J Electrochem Soc 149 (2002), pp. B34-B39. Abstract-INSPEC | Abstract-Compendex   | Full Text via CrossRef

30. Weast. Handbook of chemistry and physics, 57th ed. Cleveland, OH: CRC Press; 1976. p. B171.

31. R.M. Souto, M.M. Laz and R.L. Reis, Degradation characteristics of hydroxyapatite coatings on orthopaedic TiAlV in simulated physiological media investigated by electrochemical impedance spectroscopy. Biomaterials 24 (2003), pp. 4213-4221. SummaryPlus | Full Text + Links | PDF (335 K)

Corresponding author. Tel.: [phone]; fax: [phone]

---