Corrosion of titanium alloys in pressurized water at 300 °C and 15 MPa

Commercially pure tanium (Ti CP) and two tanium alloys (Ti 64 and Ti10-2-3) representa ve of different metallurgical classes were exposed to pressurized water at 300 °C, 15 MPa in a dedicated stainless steel corrosion loop up to 1632 h. Mass uptake measurements as well as extensive characteriza on of the oxides formed on the polished samples using scanning and transmission electron microscopy, Raman spectroscopy, X-ray diffrac on, X-ray photoelectron spectroscopy, photoelectrochemistry and glow discharge op cal emission spectroscopy led to the following main conclusions: The oxide scale was made of a thin and con nuous TiO2 layer topped by small TiO2 anatase crystallites and large FeTiO3 ilmenite crystallites. The interface tortuosity increased with exposure dura on, as well as the amount of crystallites. From 217 h exposure, the thickness of the thin con nuous TiO2 layer remained constant with exposure dura on. The mass uptake increased approxima vely linearly with exposure dura on. These observa ons were consistent with a corrosion mechanism involving simultaneous oxide growth, dissolu on and precipita on. The corrosion rate of the three studied materials was es mated to be around 3 μm/year.


Introduc on
Titanium alloys could be good candidates for nuclear Pressurized Water Reactor (PWR) primary circuit structure components because of their low neutron ac va on and their good mechanical proper es. However, their corrosion resistance in such condi ons (15 MPa, 300 °C) is poorly known.
A few studies inves gated the corrosion behaviour of tanium alloys in liquid water environment at such temperature [1][2][3][4][5][6]. For example, Kaneda et al. [1] studied the corrosion resistance of different α/β and β tanium alloys exposed 500 h to supercri cal water (25 MPa, 290-550 °C). Thin oxide films (less than 1 μm) covered by dispersed par cles were observed at 290 °C for both α/β and β alloys but the composi on of the oxides was not inves gated at this temperature. Corrosion mechanisms and corrosion rate of tanium alloys in pressurized water at 300 °C are therefore s ll undetermined.
The present study aims at quan fying the corrosion resistance of commercially pure tanium and of two tanium alloys representa ve of different metallurgical classes in pressurized water at 300 °C and 15 MPa. Flat polished specimens were first exposed to this environment in a dedicated corrosion loop. The oxides formed were then characterized using complementary techniques. Finally, a corrosion mechanism consistent with all the obtained results was proposed and the corrosion rate was es mated.

Materials
Commercially pure tanium (Ti CP) and two tanium alloys (Ti 64 and Ti10-2-3) were studied in their mill-annealed state. Ti CP is a single phase a commercially pure tanium (ASTM grade 2), Ti 64 or Ti 6Al-4V (ASTM grade 5) is a two-phase a/b tanium alloy containing 4 vol% of b phase and Ti10-2-3 or Ti 10V 2Fe-3Al is a two-phase b metastable alloy containing 38 vol% of b phase. Their chemical composi on was in accordance with the ASTM and AMS standards [7,8]. More informa on about the microstructure and the crystallographic texture of the materials can be found in [9]. 1.2-mm thick rectangular specimens (1 cm × 2 cm) were cut from the billets, then ground with SiC paper and polished with diamond paste and finally with a colloidal silica suspension at pH = 9.

Exposure of the specimens to pressurized water
The polished specimens were exposed to pressurized water at 300 °C and 15 MPa in a stainless steel corrosion loop. Lithium hydroxide (3.1 mg LiOH /kg H2O ) was added to the water to set the pH 25 °C value at 10.1, which corresponded to a pH 300 °C of 7.5 [10]. A dissolved hydrogen concentra on of 2.25 mg H2 /kg H2O was imposed by the hydrogen gas pressure (0.14 MPa) in the vessel of the cold part of the loop. The water flow was 5 × 10 −3 m 3 .h −1 leading to a water speed inside the reactor of 7.2 m.h −1 during tests. A�er having been cooled down the whole water volume went through an ion-exchange resin tank in order to capture a part of the corrosion products dissolved in the water. The ion exchange resin was saturated with lithium hydroxide before the test campaign in order to avoid any varia on of lithium concentra on during tests. New samples were used for every studied exposure dura on.

Post-exposure characteriza on techniques
The nature and morphology of the oxide scales formed on the specimens were determined by scanning and transmission electron microscopy (SEM and TEM), Raman spectroscopy, Xray diffrac on (XRD), X-ray photoelectron spectroscopy (XPS) and photoelectrochemistry. The evolu on of the samples with increasing exposure dura on was quan fied by SEM image analysis of cross sec ons, glow discharge op cal emission spectroscopy (GD-OES) and mass uptake. In order to prevent oxide degrada on during prepara on of the cross sec ons for observa on by SEM, the specimens were previously electrochemically plated with nickel. Mass uptake measurements were done by weighing of each specimen before and a�er exposure on a Me�ler Toledo XP205 scale. More details regarding the characteriza on techniques were reported in [9].

Results
Nature and morphology of the oxides Figure 1 shows a SEM image and energy dispersive spectroscopy (EDS) mapping of the surface of a Ti CP sample exposed during 434 h. Two types of crystallites were discriminated. The first one is cons tuted by rela vely large crystallites (3-10 µm) containing tanium, iron and oxygen and the second one is cons tuted by rela vely small crystallites (0.3-1 µm) containing only tanium and oxygen. These two types of crystallites were observed on the three studied materials a�er all exposure dura ons. Local Raman spectroscopy presented in [9] allowed the iden fica on of the crystallographic phase of these two types of crystallites: FeTiO 3 ilmenite for the larger crystallites and TiO 2 anatase for the smaller ones.
The X-ray diffractograms showed in Figure 2 confirmed the presence of these two oxide phases along with the metallic tanium phases: a for all materials and b for Ti 64 and Ti10-2-3. No ceable was also the presence of TiO 2 ru le on Ti CP sample exclusively.
Complementary analysis by ASTAR-TEM presented in Figure 3 enabled to iden fy this phase as a thin con nuous oxide layer located on Ti CP at the interface between the metal and the environment. ASTAR-TEM analysis also confirmed the absence of TiO 2 ru le on Ti 64 and Ti10-2-3 alloys. The thin con nuous oxide layer that was observed by TEM and SEM on these alloys was made of TiO 2 anatase, so that the fron er between small crystallites and the con nuous layer did not appear as clearly as on Ti CP. Evolu on of the samples with increasing exposure dura ons Figure 4 shows SEM images of the cross sec ons of Ti CP samples exposed respec vely 217 h, 434 h and 834 h to pressurized water at 300 °C and 15 MPa. The cross sec ons of the other studied materials (not shown here) presented the same features. The con nuous oxide layer thickness was found to remain constant from 217 h up to 1632 h. The mean thickness values measured by SEM were approximately 30, 24 and 22 nm for Ti CP, Ti 64 and Ti10-2-3, respec vely. As the exposure dura on increased, the tortuosity of the interface between the materials and the environment increased. The mean corroded metal depth was defined as the quo ent of the volume of corroded metal by the ini al surface of the sample. It was es mated following the method illustrated in Figure 5. For each exposure dura on and each studied material, 20 SEM images were analysed. Figure 6a shows that the mean corroded metal depth increased as func on of the exposure dura on in an approximate linear way. Figure 6b indicates that the ra o of the maximum corrosion penetra on depth to the mean corroded metal depth was near 3.5:1 for all studied materials and all exposure dura ons.
GD-OES was used to quan fy the ra o of the amount of FeTiO 3 ilmenite to the total quan ty of oxide (both FeTiO 3 and TiO 2 ). Figure 7 indicates that this ra o (referred to as q in the following sec on) never exceeded 0.22. The main oxide formed on the samples was therefore TiO 2 . Figure 7 also indicates that this ra o was lower for Ti 64 than for the two other studied materials and that it increased with the exposure dura on.
Mass uptake values as func on of exposure dura on are presented in Figure 8. They increased almost linearly with exposure dura on. Mass uptake of Ti 64 was slightly lower than Ti CP and Ti10-2-3 ones.

Corrosion mechanism
The increase of interface's tortuosity with exposure dura on along with the steadiness of the con nuous oxide layer's thickness indicate that this TiO 2 layer underwent simultaneous dissolu on and growth. This type of mixed corrosion mechanism was modelled by Haycock [11]. In the present case, the con nuous oxide layer growth was assumed to be controlled by the diffusion of oxygen vacancies through the layer because it was found to grow inwards and to be a n-type semi-conductor [9]. The growth rate therefore decreased as the thickness increased during the first transitory period. During this period, the oxide layer growth is significantly faster than the dissolu on, leading to an increase of the oxide thickness from ~4 nm (na ve con nuous oxide thickness) up to ~25 nm (sta onary con nuous oxide thickness). At such thickness, growth rate and dissolu on rate are equal so that the thickness remained constant.
Such a mechanism is consistent with the significant value of the TiO 2 solubility in water in the condi ons of the study (around 1.2 × 10 −8 mol. kg −1 H 2 O [12]). Titanium species poten ally formed by the dissolu on of TiO 2 in water at pH = 7.5 and 300 °C are Ti(OH) 4 and Ti(OH) 5 − hydroxides [13]. The presence of TiO 2 anatase crystallites on the con nuous oxide layer surface of all studied material for all exposure dura ons indicates that dissolved tanium hydroxides have then reprecipitated. Part of these hydroxides co-precipitated with iron hydroxides (most probably originated from the corrosion of the stainless steel pressure vessel) forming FeTiO 3 ilmenite.
No ceable is the fact that no aluminium nor vanadium was detected in the oxides formed on Ti 64 and Ti10-2-3. The present authors s ll believe that these elements were oxidized and incorporated as ca ons in the thin con nuous TiO 2 oxide and finally dissolved in the water. Aluminium oxides solubility is indeed very high at 300 °C and pH = 7.5 (around 10 -3 mol.kg -1 H 2 O [14]). As the contribu on of the con nuous oxide layer to the overall oxide formed on the samples was very low, poten al presence of aluminium and vanadium in this layer was not detectable by GD-OES nor XPS.
Characteriza on by ASTAR-TEM, SEM and XRD revealed that the con nuous TiO 2 layer was made of ru le on Ti CP sample and made of anatase on Ti 64 and Ti10-2-3 samples. This result is consistent with the bandgap values measured by photoelectrochemistry presented in [9]. The TiO 2 crystallites were made of anatase on all materials. Ru le is for all temperature and pressure condi ons the thermodynamic stable phase of TiO 2 . However, due to kine cs, low temperatures such as 300 °C favor anatase forma on, which is consistent with the anatase crystallites observa on. Impuri es and dopants favor the forma on of either ru le (like hydrogen, lithium and vanadium) or anatase (like aluminium) [15][16][17]. In the present case, we believe that a part of the hydrogen produced by the reduc on of water balancing tanium oxida on is incorporated in the con nuous oxide layer. When the material does not contain aluminium, this hydrogen doping appears to have the dominant effect (leading to ru le forma on on Ti CP); otherwise, the effect of aluminium doping appears to dominate (leading to the forma on of anatase on the other studied materials).

Corrosion rate
The quan fica on of the mean corroded metal depth as func on of exposure dura on based on cross sec ons observa on by SEM (Figure 6a) led to the following es ma on of the corrosion rate: ~1.2 µm/year for the three studied materials. As the reference line used in this method might not correspond to the real ini al surface posi on, this es ma on is obviously a lower bound. Figure 6b indicates that the corrosion rate could reach locally a 4 mes higher value. Nevertheless, as the ra o of the maximum corrosion penetra on depth to the mean corroded metal depth remained constant while exposure dura on increased (Figure 6b), the corrosion appeared rela vely homogeneous. In par cular, no preferen al corrosion of alpha phase or beta phase was observed for Ti 64 or Ti10-2-3.
Following the mechanism proposed in the previous sec on, mass uptake per unit area Dm/S (g.cm -2 ) results from three contribu ons. The first one is the mass gain due to oxygen involved in the growth of the con nuous oxide layer; the second one is the mass gain due to oxygen and iron precipitated on the surface in either TiO 2 anatase or FeTiO 3 ilmenite crystallites; the last one is the poten al mass loss due to the tanium (and alloying elements) that was dissolved in the water and that did not precipitate as crystallites on the sample.
Eq. 1 is an approxima on of this mass uptake, where De represents the increase in the con nuous oxide layer thickness (cm), M i and r i the molar mass (g.mol -1 ) and the density (g.cm -3 ) of the i specie, r the corrosion rate (cm.s -1 ), t the me (s) from the beginning of the exposure, p the ra o of precipitated tanium to dissolved tanium and q the ra o of the amount of FeTiO 3 to the total amount of oxides. In the present experimental condi ons, the first term was much lower than the second one. In other words, the mass uptake resulted almost exclusively from the micrometric crystallites rather than from the nanometric con nuous layer. This equa on also emphasizes the fact that, with such a mixed oxide dissolu on and precipita on corrosion mechanism, mass uptake could be nega ve for p values inferior to 0.6. In the present case, the mass uptake experimental values were posi ve (Figure 8). They were used along with the q ra o es mated by GD-OES (Figure 7), to es mate the corrosion rate according to Eq.1. Values of 2.2, 2.8 and 3.8 µm/year were found for p values of 1, 0.9 and 0.8, respec vely, which is consistent with the lower bound determined previously.
As illustrated in Figure 8, this model suggests that the lower mass uptake of Ti 64 in comparison to the other studied materials originated from a lower ilmenite precipita on (lower q ra o) rather than from a lower corrosion rate.

Conclusion
In pressurized water at 300 °C, 15 MPa and pH 300 °C = 7.5, the corrosion rate of commercially pure tanium and of two tanium alloys representa ve of different metallurgical classes was found to be around 3 µm/year. The corrosion rate appeared to be controlled by the dissolu on of the con nuous nanometric TiO 2 oxide layer that formed on the surface of the studied materials. Growth of this oxide layer occurred at the metal-oxide interface and was assumed to be controlled by the diffusion of oxygen vacancies.
In the present experimental condi ons, the major part of the dissolved TiO 2 con nuous oxide re-precipitated on the specimens as both TiO 2 anatase and FeTiO 3 ilmenite crystallites. Iron involved in ilmenite precipita on was believed to originate from the corrosion of the pressure vessel made of stainless steel.