Stress corrosion of Ni-based superalloys

Abstract The development of gas turbines to increase fuel efficiency is resulting in progressively higher operating temperatures in the under platform regions of the blades. These regions have traditionally been considered low risk areas. However, higher metal temperatures combined with stresses and the deposition of contaminants from the cooling air system may result in complex degradation mechanisms. Static stress corrosion testing has been conducted on C-ring specimens at a range of stresses in a hot corrosion environment. Cracks were observed in C-rings after exposure times greater than 100 h. Scanning electron microscopy (SEM) systems were used to image cracks and characterise deposits to improve understanding of the mechanism. Finite element analysis (FEA) has been used to model the stress intensity under test conditions. CMSX-4 specimens subject to static stresses combined with hot corrosion demonstrated significant material degradation (crack initiation and propagation) suggesting a combined stress corrosion mechanism resulting in cracking.


Introduction
Gas turbines are widely used in power generation systems. Developments to improve their efficiencies have led to increased operating temperatures of regions of some components, such as the under platform areas of turbine blades. The high stress state of the root pocket due to the high rotational speeds, combined with cooling air derived deposits and temperatures approaching the conditions associated with type II corrosion, can lead to cracking [1]. CMSX-4 (Table 1) is a single crystal Ni-based superalloy commonly used for 1st stage gas turbine blades as a result of its good high temperature creep-strength properties combined with production affordability [2]. However due to its composition (lower Cr content than other commonly used 1st stage turbine blade materials), CMSX-4 is susceptible to type II hot corrosion. This can result in damage that has the morphology of either pitting or broad fronted attack. Sumner et al. [3] have reported investigations of type II hot corrosion of CMSX-4, using statistical analysis of large data sets to generate models for specific conditions. They observed broad fronted attack and more rapid depletion of Cr in CMSX-4, when compared with IN738LC.
Reaserch conducted on hot corrosion mechanisms in 1970s-80s is summarised by Luthra & LeBlanc [4]. They concluded that hot corrosion could occur through a combination of three mechanisms: sulphidation-oxidation, formation of volatile compounds beneath the protective oxide layer or scale fluxing. Fluxing models have since gained the widest acceptance for deposit induced hot corrosion [5,6].
The process of type II hot corrosion of Ni-based superalloys requires the formation of a liquid eutectic film [5,6]. Type II hot corrosion occurs in the temperature range of 650-800 °C through the formation of minimum melting point mixtures of Na 2 SO 4 , NiSO 4 and CoSO 4 [4,5,8]. NiSO 4 and CoSO 4 compounds form as a result of the reaction of SO 3 with nickel and cobalt from the superalloy. A widely accepted mechanism for hot corrosion was proposed by Goebel & Pettit [9]. Their mechanism outlines two stages, firstly the incubation stage, where a liquid eutectic of Na 2 SO 4 , NiSO 4 and/or CoSO 4 forms on the component surface as a result of deposition coupled with a reaction between sulphur oxides and nickel and/or cobalt from the superalloy. The second stage is the propagation stage, where the fluxing of the surface oxide by a liquid deposit on the surface allows inward access, and outward Co/Ni transport. This form of attack often results in pitting damage with an outer NiO/CoO layer being formed, although sometimes a form of broad fronted attack develops [5,6].
For type II hot corrosion many researchers have noted the importance of a constant SO x supply for OPEN ACCESS sustained corrosion to occur [3,7,9,10]; this variation of the damage mechanism is known as gas induced acidic fluxing [8,11]. Without both gaseous SO x and a regular sulphate deposition flux, the corrosion reaction would cease to occur when all the reactants have been consumed.
Type II hot corrosion combined with static stress in Ni-based superalloys has not been extensively studied. However, stress corrosion cracking (SCC) is a well-documented failure mechanism especially in aqueous systems [12,13].
Studies have been conducted on the effects of stress on corrosion pitting growth in aluminium alloys [14]. It was found that corrosion pit growth could be affected by time, stress amplitude and frequency in a fatigue environment. The methodology of Ishihara et al. [14] was applied to Ni-based superalloys by Chan et al. [15]. They considered the point at which fatigue crack growth exceeds corrosion pit growth. However, neither of these studies consider the effect of hot corrosion on the materials stress intensity threshold (k th ), the threshold below which cracking does not occur.
Finite element analysis (FEA) is a commonly used method to calculate stresses within complex geometries or multiaxial loading states. This is done by meshing the geometry as a net of elements and nodes. The elements can deform as constrained by the material model, whereas the load is transferred from element to element through the node connections. FEA has been widely used to assess the stress in statically and cyclically loaded conditions.

C-ring test method
C-ring specimens were manufactured from CMSX-4 bars. Guidelines for the dimensions were taken from ISO 7539-5 [16]. The final dimensions for the specimens used in this testing are given in Figure 1. C-ring specimens where manufactured with a < 001 > crystallographic orientation aligned with the cylinder axis.
For target stress levels at a constant stain, the required displacement of the C-rings were calculated by first calculating the change in diameter (ΔD) required to achieve a given stress (Equation (1)).
FEA modelling was used to verify the stress calculations. Surrogate data from Siebörger et al. [17] for CMSX-4 provided the Young's modulus (E) for Equation (2) and the monotonic material properties used in FEA modelling. The final stressed diameters (D f ) were calculated using Equation (2): The C-rings were clamped to the final diameter (D f ) using A2 grade stainless steel M5 nuts, bolts and washers, and measured using a digital micrometre with a resolution of 1 μm (and accuracy of 2 μm). An average of five readings was used to determine the initial external diameter (D AV ) from which the final stressed diameter is calculated. These are given in Table 2.
Prior to testing, C-ring samples were cleaned in an ultrasonic bath with IPA (isopropyl alcohol). Corrosion exposures were carried out in a horizontal, controlled atmosphere furnace. The corrosion environment, deposit composition and deposit flux were controlled using the well-established deposit recoat methodology (e.g. Sumner et al. [3]). Specimens were coated with an 80/20 M mixture of Na 2 SO 4 /K 2 SO 4 . The mass of deposited salt was measured per unit area and specimens were recoated every 100 h in order to control the deposition flux. A gaseous environment of air -300 vppm SO 2 was used, and all testing was conducted at 550 °C. C-rings were stressed to 800, 700 and 500 MPa, and exposed for 100, 300 or 500 h exposure times with a target deposition flux of 5 μg/cm 2 /h. In addition, one C-ring at each target stress level was exposed for 300 h without any deposit.

Microscopy and analytical methods
Samples were mounted in a 50:50 mixture of MetPrep's epoxy set resin and ballotini (40-70 μm diameter glass spheres). Samples were then sectioned, using oil lubricant to prevent dissolution of corrosion products and deposits, before being ground and then polished to a 1 μm diamond paste finish (again using oil lubricant).
Both optical and SEM examinations of the samples were carried out. Optical microscopy was used to determine if cracking was present in the C-rings after each exposure period. SEM was used to characterise the results of the degradation mechanism's interaction with the alloy microstructure. FEI XL-30 and a JEOL 7800F field electron gun (FEG) SEMs equipped with backscatter energy-dispersive X-ray (EDX) detectors were used for characterisation and SEM imaging. SEM images were post-processed using Image J software to enable accurate measurement of features.

FEA analytical methods
FEA modelling was conducted using ANSYS Workbench 15 [18]. The material model used for this analysis was an isotropic model generated using monotonic surrogate material data for CMSX-4 from Siebörger et al. [17]. The C-ring was constrained between two plates ( Figure 2). A frictionless sliding contact was utilised between one of the two blocks and the C-ring in order to allow a small amount of relative movement. The boundary conditions were applied through a displacement equivalent to those calculated using Equation (1) (and listed in Table 3). The C-ring was modelled as three separate sections to allow more accurate refinement of the mesh in the central section of the C-ring. This was important as it was in this central region that it was anticipated that cracking could occur. A hex dominant mesh is used wherever possible; however meshing around the crack tips required the use of a tetrahedral mesh due to the size and complexity of the crack tip geometry.
When clamped, a multiaxial stress condition was predicted in the C-ring. As such the von Mises criterion was used to obtain local stress values. However it was also useful to consider the principal or normal stresses in relation to a mode I crack opening, as shown in Figure 3, as these stresses could be higher for the C-ring geometry in the principal x-axis plane when compared with the von Mises stress.
Linear elastic fracture mechanics (Equation (3)) were used to assess the local stress intensity range (Δk) for micro semi-elliptical cracks within the C-ring. The stress intensity was assessed using the ANSYS R15 [18]   no signs of cracking after 500 h of exposure to the test conditions. C-rings normally experienced cracking within the most highly stressed central region. However when cracks initiated off centre, cracking would occur either side of the centre line due to the shifted stress distribution around the C-Ring ( Figure 6).
The corrosion mechanism varied from attacking the gamma-prime ( ′ ) to attacking the gamma-matrix ( ) This is visible in SEM backscattered imaging as a shift in the contrast between the two microstructural features (Figure 7). This reduction in back scattered electrons is attributable to the lower atomic number of the S and O present in the corrosion products.
SEM imaging suggests that the initial combined presence of stress and hot corrosion results in the reaction of the ′ precipitates. Cracks then initiate from features similar to corrosion pit features ( Figure  7) and propagate through the ′ where corrosion is present (Figure 8). Using the results of EDX analysis ( Figure 5) it is hypothesised that this is because of the lower Cr and Co content of the ′ precipitates. stress intensity solver code. ANSYS uses T-stress evaluation [19] in order calculate the local crack tip stress intensity.
FEA generated local crack tip stress intensity predictions were then used to calculate the geometry factor (Υ) for finite surface cracks in the C-ring geometry. Δk can then be compared to k th to assess the likelihood of cracking. The k th of CMSX-4 has been reported to be 15 MPa.m 1/2 in air at 750 °C [21]. Stress intensities _calculated through FEA modelling can be compared to this in order to determine the likelihood of cracking and the effect of hot corrosion on k th .

C-ring results & discussion
Unstressed sections of CMSX-4 C-rings were corroded at 550 °C with a target deposition flux of 5 μg/ cm 2 /h and a gaseous environment of air -300 vppm SO 2 . Samples were removed after exposure times of 100, 300 and 500 h. Inspection showed formation of an oxide scale containing Co, Ni, S and O ( Figure  4 and 5). Sulphidation had occurred beneath the oxide scale, consistent with type II hot corrosion [7,22].
For stressed C-rings, cracking of corroded samples was observed at 800 and 700 MPa after exposures for 100 h; with visible cracking still occurring at 500 MPa for exposure times longer than 100 h ( Table  2). By contrast C-ring tests without deposit showed  This stress state would suggest cracks would firstly initiate and then propagate in the z-axis where the maximum principal is acting at a normal in mode I crack opening. However as cracks propagate and Δk exceeds k th then secondary cracks could propagate in all three principle directions. A summary of the stress conditions for various ΔD values is given in Table 3. FEA was further used to predict the stress intensity and concentration around a crack tip ( Figure 10) within the C-ring geometry. These micro-cracks were modelled in the central region of the C-ring using a refined tetrahedral mesh; the results are presented in Table 4.
FEA stress intensity modelling suggests that cracks or pits need to be greater than 100 μm for cracking to occur given a k th of 15 MPa.m 1/2 as the reported fatigue The corrosion attack shifts to the , and it is hypothesised that this happens when the protective NiO/ CoO rich oxide scale is formed, as this depletes Co from the alloy which is mainly concentrated in the matrix.

FEA principal and von Mises stress state modelling in C-rings
FEA modelling predicted that maximum stress occurred in the central region of the C-ring as shown in Figure 9. FEA also predicted the presence of a multi-axial stress state within the C-ring, where the largest resolved principal stress plane, referred to as maximum principal, occurs along the x-axis, and the second largest resolved stress plane, referred to as middle principal, occurs in along the z-axis.  threshold for CMSX-4 [21]. Therefore the presence of hot corrosion may have a significant effect on reducing the material's k th as well as concentrating stress through corrosion pitting. Analysis of a corrosion pit's size in cracked C-ring specimen implies that a 10 μm diameter pit has initiated cracking during these exposures (Figure 7). Using    A Kitagawa diagram [23] has been plotted to demonstrate the stress and crack or defect size needed to exceed the material's k th (Figure 11). This is done both for the calculated theoretical k th in hot corrosive conditions, and using the reported k th for air. the FEA calculated geometry factor Υ of 0·836, this gives a theoretical reduced k th of 3·748 MPa.m 1/2 when hot corrosion is simultaneously acting with a stress of 800 MPa; a 75% reduction. This means that cracking can occur at considerably lower applied stresses.  von Mises criterion, FEA calculated equivalent stress concurs with that from ISO 7539-5.
FEA stress intensity modelling around crack tips estimates that fatigue/fracture (k th ) can be reduced by up to 75% with the combined effect of hot corrosion in CMSX-4.
SEM imaging suggests the combined hot corrosion stress mechanism initially attacks ′ precipitates, cracks then propagate through precipitates as the mechanism attacks features ahead of its propagation path. A switch to attack of the matrix is observed. It is hypothesised that this occurs in order to form the NiO/CoO oxide scale which depletes Co from the matrix.

Conclusions
SEM/EDX characterisation of the corrosion product produced by stress corrosion in CMSX-4 C-rings at 550 °C is consistent with type II hot corrosion. Hot corrosion conditions at 550 °C combined with static stresses of greater than 500 MPa can cause a significant hot corrosion stress cracking mechanism. A lower limit seems to exist around 500 MPa. However at exposures greater than 100 h with a flux of 5 μg/cm 2 /h cracking is still visibly present.
FEA modelling predicts the multiaxial nature of the stress state within a clamped C-ring and the observed cracking in experimental testing supports the modelling results. By determining the effective stress through the