At high temperature, zirconium interacts with water as follow: Z r + 2 H 2 O → Z r O 2 + H 2

The mobility of zirconium ion, Zr⁴⁺, is expected to be extremely low, as evidenced by its immobility during the growth of anodic oxide film (Cox and Pemsler, 1968). This characteristic makes oxygen ions, O²⁻, the only mobile species in the corrosion system. However, Whitton, (1968) also proposed that, at a relatively high temperature, such as a loss-of-coolant accident (LOCA) condition, there might be some migration of small zirconium ions. To maintain the charge neutrality of the system, an electron flux moving in the opposite direction is required. As an ionic oxide without available free electrons, it is believed that the transformation of oxygen ions is achieved through vacancies ( V O ″ ) (Cox and Pemsler, 1968; Allen et al., 2012a) in the system, through either bulk diffusion or grain boundary diffusion (Cox and Pemsler, 1968). Cox and Pemsler, (1968) used radioactive oxygen measurements and found that bulk diffusion is over five orders of magnitude slower than grain boundary diffusion, which means that grain boundary is the dominant diffusion path. It should be noted that, as an element belonging to Group IV, zirconium accommodates a large amount of oxygen (Pieraggi et al., 1995). Therefore, the weight gain of zirconium corrosion comprises two parallel processes (Zino et al., 2021), the formation of oxide and the solid solution of oxygen (Pieraggi et al., 1995; Zino et al., 2021), as illustrated in Figure 1.

The Research on corrosion behavior of zirconium alloys has predominantly focused on temperatures ranging from 300°C to 1,400°C. Due to the dominant application of pressurized water reactors (PWRs), extensive investigation aims to the corrosion mechanism at the operating condition of PWRs, where the coolant-side temperature is 360°C. The occurrence of Fukushima Nuclear Disaster also attracts public attention to the corrosion behavior at higher temperature. When loss-of-coolant accident (LOCA) happens, the rapid depressurization in main loop caused by facture of pipes leads to evaporation of coolant, which generates steam as corrosion medium. At the same time, cladding tubes experience a dramatic temperature increase to over 1,000°C due to the loss of coolant. Therefore, the corrosion experiments exceeding 1,000°C is designed to simulate LOCA in PWRs. Moreover, research around 500°C, aimed for the nodular corrosion observed in the boiling water reactors (BWRs) (Charquet, 2001; Xie et al., 2017; Samanta et al., 2023).

Under operating condition, it is commonly observed that there are three stages of corrosion (Wei et al., 2013; Cui et al., 2022), as shown in Figure 2A. The pre-transition stage demonstrates cubic law kinetics governed by the Wagner rule (Wagner, 1933), followed by the transitory stage composed by a set of successive cubic curves. Finally, corrosion enters the post-transition stage where the weight gain curve increases linearly with time. As temperature increases, the occurrence of transition is much earlier, and the corrosion only contains two stages, as shown in Figure 2B. The initial stage obeying a cubic or parabolic law would transform to a quasi-linear subsequent stage.

Nagase et al. (2003) performed experiments in the temperature range of 500–1,300°C, and found a quasi-cubic law with an exponent n around 2.5–3 when temperature is lower than 900°C, as shown in Figure 2C. When temperature exceeds 1,000°C (1273K), the cubic law completely transforms into a parabolic law. The authors suggested that this transformation results from the stabilization of tetragonal phase of ZrO₂ (t-ZrO₂) at higher temperature. Apart from the oxidation exponent, the increased temperature can also significantly shorten the time required by breakaway phenomenon. For example, 180 days is needed at 360°C (Wei et al., 2013), while only 20 min is needed at 1,000°C (Cui et al., 2022). Meanwhile, as temperature increases, α-Zr with a close-packed hexagonal structure transform to β-Zr (cubic structure) (Wilhelm and Garcia, 1987; Lee et al., 2017). The oxygen, as a stabilizing α phase element, forces β-Zr near the oxide-metal interface to be converted to α-Zr and, thus, leads to a layer of α-Zr(O) phase underneath the oxide-metal interface (Kim et al., 2011).

The morphology of oxide grains has been comprehensively investigated using various electron microscopy techniques. In the early stage of corrosion, small equiaxed oxide grain form and grow anisotropically into columnar grains (Jeong et al., 1999; Yilmazbayhan et al., 2006), as depicted in Figure 3. The strong texture and compact arrangement of these columnar grains contribute to the protective nature of oxide film. However, at the coolant-side of oxide film, the dense columnar grains transform into the loose and porous equiaxed grains, as shown inFigure 3. This transformation leads to the degradation of the film’s protective function.

Apart from morphology changes, the structure of oxide also undergoes transformation over time. At the service temperature, the monoclinic phase of ZrO₂ is the only thermodynamically stable phase (Setiadinata, 2016), exhibiting over six different structures dependent on temperature and pressure (Whitney, 1965; BLOCK et al., 1985; Ohtaka et al., 1988; Ohtaka et al., 1990; Ohtaka et al., 1991; Ohtaka et al., 2001), as illustrated in Figure 4. Despite t-ZrO₂ being thermodynamically unstable at room temperature or even at the reactor core temperature, it is still observed alongside the monoclinic phase (m-ZrO₂) (Godlewski et al., 1991; Yilmazbayhan et al., 2006). Generally, it is suggested that t-ZrO₂ can be stabilized by the following factors.

The stress relaxation in oxide film caused by t-m transformation is the common reason for porosity (Qin et al., 2006; Park et al., 2010). The first compelling evidence of the porosity along the oxidation direction in oxide film corroded at 360°C was provided by Hudson et al. (2009), Hudson and Smith (2009), Ni et al. (2010), Ni et al. (2011a), Ni et al. (2011b), Ni et al. (2012) using Fresnel contrast of transmission electron microscopy (TEM), as shown in Figure 6A. Based on the observed porosity, they proposed a corrosion transition mechanism that these nanopores serve as a short path for oxygen diffusion, contributing to breakaway phenomenon. This mechanism was confirmed by their experiments using ¹⁸O to trace the diffusion of oxygen and characterization with Nano secondary ion mass spectrometry (NanoSIMS) (Ni et al., 2012). At higher temperature, Cui et al. (2024) observed nanovoids in an oxide film formed at 1,000°C, as depicted in Figure 6B, and attributed their formation of Zr₅Sn₃, secondary phase particles. At the same temperature, Park et al. (2010) observed micro-pores and micro-cracks at oxide grain boundaries as the grain size increases, as shown in Figure 6C. They indicated that these micro-pores and micro-cracks not only enhance oxygen diffusion but also increase the magnitude of local stress, leading to the propagation of cracks.

Cracks serve as another indicator for the degradation of oxide film. Regarding the direction of cracks relative to the O-M interface, they can be classified into two types: lateral cracks along the interface and the vertical cracks perpendicular to the interface. Isolated lateral cracks with relatively small size are frequently observed above the convex part of the O-M interface (Bossis et al., 2000; Ni et al., 2011a; Tejland and Andren, 2012), as shown in Figure 7A. For continuous lateral cracks, they are periodic separated in the oxide film (Yilmazbayhan et al., 2004), as shown in Figure 7B. Based on the oxide film thickness beneath the continuous lateral cracks, it is considered as the symbol of the onset of transition (Yilmazbayhan et al., 2006; Polatidis et al., 2013; Lee et al., 2017). The connection of small lateral cracks is believed to create the continuous ones (Wallwork et al., 1965; Ni et al., 2011a). The compressive stress in the oxide film, which may block the propagation of cracks, can be overcome as the toughened ceramic (Ni et al., 2011a). In contrast, Kim et al. (2011) suggested that the accumulation of hydrogen, caused by the reduction of α-Zr(O) grain boundaries, is responsible for the formation of lateral cracks, based on their experimental observation at 1,000°C. Even though the cracks severely degrade the protective function of oxide film, the exact impact of lateral cracks on the corrosion is disputed. For example, Lee et al. (2017) suggested that the continuous lateral cracks disturb the oxygen diffusion towards the matrix before transition, while no longer having influence on corrosion after transition.

Particular attention is paid to the vertical cracks, as they act as a direct diffusion path for corrosion medium to reach the matrix. Finite element simulation by Platt et al. (2014) demonstrated that the stress induced by t-m transformation can lead to the formation of cracks along grain boundaries. Clear evidence for continuous nano-pipes along the corrosion direction in the oxide film was provided by Hu et al. (2019) in the corrosion experiment of Zr-0.5Nb, Zr-1Nb and Zr-4 alloy at 360°C. They suggested that these nano-pipes, derived from the connection of nano-pores near the interface, can accelerate the diffusion of hydrogen. Similarly, Figure 7C shows the microchannels perpendicular to the O-M interface (Cui et al., 2022) in an oxide formed at 1,000°C. Cui et al. (2022) believed that these microchannels, located above the concave part of the interface, also result from the connection of nano-pores. At the outer side of oxide film, the tensile stress generated from the convex geometry of the tube leads to vertical cracks starting from the surface of oxide film and propagating toward metal (Lee et al., 2019; Cui et al., 2022). In the matrix adjacent to the O-M interface, Huang et al. (2015) found the vertical cracks next to the aligned β-Nb particles in the corrosion of Zr-0.93Sn-0.82Nb-0.17Fe-0.019S alloy at 360°C. They suggested that the stress concentration caused by delayed oxidation of β-Nb particles is the dominant reason for their formation. At 1,000°C, Lee et al. (2019) also observed vertical cracks in the metal matrix, as shown in Figure 7D, and attribute their formation to the brittle matrix with the concentration of oxygen. During sample preparation, external stress brought from cutting or thermal mounting can easily lead to brittle failure.

Numerous experimental results indicate that the O-M interface formed on zirconium alloy is undulated rather than a flat plane. Barberis et al. (1997) and Blank et al. (1992) described it as cauliflower-like structure, where the larger unit of undulation is composed by a range of second undulations with relatively small sizes. Platt et al. (2015a) and Parise et al. (1998) utilized an egg-box type structure to illustrate the geometry of the O-M interface, simplifying its shape by employing a sinusoidal function in their calculation. This undulated shape is believed to be caused by different local corrosion rate. Ni et al. (2012) used various techniques, including electron energy loss spectroscopy (EELS), atom probe tomography (APT) and NanoSIMS, to characterize the chemistry of O-M interface formed on ZIRLO and Zr-4 alloy corroded at 360°C using primary water chemistry (pure H₂O with addition of 2 ppm by wt. Li and 1,000 ppm by wt. B). They found that ZrO₂ at the advanced corrosion front (the convex part of O-M interface) and ZrO at the concave part, preserved by slower corrosion rate. Scanning electron microscopy (SEM) and TEM observations from Cui et al. (2022) also supported the conclusion that the convex part of the O-M interface formed on zirconium alloy corroded at 1,000°C is derived from the faster corrosion rate due to microchannels perpendicular to the interface.

As a reflection of local corrosion rate, the roughness of the O-M interface is expected to change with corrosion time, as demonstrated by many published works. Ni et al. (2011a), using focused ion beam sectioning, reconstructed the area around the O-M interface formed on stress-relieved ZIRLO alloy corroded at 360°C, showing the interface roughness as a function of corrosion time, consistent with the tendency observed at 1,000°C (Cui et al., 2022). The roughness reaches a maximum around the transition and gradually decrease afterward. A statistics analysis of the O-M interface amplitude formed in A-0%Sn, ZIRLO, and Zr-4 alloy corrosion experiments at 360°C obtained an identical trend (Platt et al., 2015b). They also observed a recovery after the first reduction, probably caused by a second transition. Cui et al. (2022) provided a detailed explanation of interface roughness evolution, suggesting that the relatively smooth O-M interface after transition represents the completed development of microchannels dominating the corrosion rate in the oxide film.

The undulated morphology of the O-M interface, caused by local differences in corrosion rate, has a significant influence on the oxide film, particularly on the stress state of oxide film. Both Parise et al. (1998) and Platt et al. (2015a), using finite element analysis, indicated that the direction of the vertical stress component above the convex part of the undulated interface is opposite to the concave part, as shown in Figure 8A. Parise et al. (1998) also reported that the stress value is positively proportional to the amplitude of undulation, as shown in Figure 8B. When the amplitude of the O-M interface reaches a certain level, the tensile stress in oxide above the convex part of interface is large enough to the form isolated cracks, as observed by numerous works (Bossis et al., 2000; Ni et al., 2011a; Tejland and Andren, 2012; Cui et al., 2022). Furthermore, compressive stress can efficiently stabilize t-ZrO₂, as confirmed by Raman spectroscopy (Cui et al., 2022), aligning with the simulation results of Platt et al. (2014). Building on the vacancy field, Cui et al. (2022) also established a crack formation mechanism for the cases with smooth O-M interface, suggesting that the morphology of cracks strongly depends on the configuration of O-M interface.

So far, the development of zirconium alloys has predominately focused on three systems: Zr-Sn alloy, Zr-Nb alloy, and Zr-Sn-Nb alloy. Initially, the Sn element was induced to eliminate the negative effects of nitrogen introduced by the Kroll process, an industrial method for preparing Zr metal. Consequently, the Zr-1 alloy, known as Zr-2.5Sn, has a similar corrosion behavior to Zr forming a spongy structure. The accidental introduction of transition metal elements, such as Cr, Fe and Ni, significantly improved the corrosion resistance, leading to the development of the Zr-2 alloy. However, Ni, which contributed to increased hydride formation, was eventually replaced by Fe to create the Zr-3 alloy. The early developed Zr-3 alloy, with unsatisfying mechanical properties, underwent further modifications and evolve into the Zr-4 alloy. The currently optimized Zr-4 alloy has been developed with precise control of N content and decreased Sn content, forming the original Zr-System.

Russia originally established the Zr-Nb system (Shebaldov et al., 2000), and was further developed by different countries. For example, France produced the M5 alloy (Zhao, 2001), Canada the Zr-2.5Nb alloy (Warr et al., 1996), and Korea the HANA series alloy (Chakravartty et al., 1995). Combining the advantages of Zr-Sn and Zr-Nb alloy systems, the third generation of zirconium alloy, namely, Zr-Sn-Nb, was developed. This includes the ZIRLO alloy in America (Sabol et al., 1989), the E365 alloy in Russia (Nikulina et al., 1996), and the N18 and N36 alloys in China (Aldeen et al., 2022). Table 1 summarizes the zirconium alloys developed by various countries.

Considering the development of zirconium alloys, several critical alloy elements, including Nb, Sn and Fe, are reviewed, with a specific focus on the second phase particles (SPPs) that they formed. The main reason is that both the formation and the dissolution of SPPs changes the electrochemistry properties of oxide film, which inevitably influences the oxygen ion diffusion in the oxide. Meanwhile, the delayed oxidation and different thermal behavior of SPPs can lead to concentrated stress and generate cracks.