Effect of surface oxides on tritium entrance and permeation in FeCrAl alloys for nuclear fuel cladding: a review

1 Introduction

More than 30 countries in the Globe are generating electricity using the heat released by the fission of the uranium-235 isotope (IAEA 2021). Nuclear energy is clean and safe and does not contribute to global climate change (Nuclear Energy Institute 2021). The construction of newer Generation III nuclear power plants (NPP) is continuing in countries like China but in other countries like the USA, the number of nuclear power stations generating electricity is slowly decreasing due to decommissioning of existing reactors and the absence of new constructions (IAEA 2021). Figure 1 shows that practically no new NPPs were connected to the US civilian grid in the last 30 years (IAEA 2021). That is, the age of the existing US based plants is increasing, and many times these NPP are not economically competitive compared to the lower operational cost and lower regulations of natural gas burning plants. Many countries have recently recategorized nuclear energy as clean or green energy in order to protect the NPP from decommissioning (World Nuclear News 2021). At the same time, for the last decade, the international materials community has been engaged in finding safer materials to be used in existing LWRs (Rebak 2020a). These innovative safer materials are generally known as accident tolerant fuels (ATF), and their retrofit into the existing NPP will allow them to operate for longer times, possibly up to 80 or 100 years.

Figure 1:

Annual reactors connected to the grid in the USA. Few new NPPs connected in the last 30 years.

1.1 Accident tolerant materials

Figure 2 shows the proposed newer ATF materials to be used in existing LWRs in order to make them safer and more economical to operate. These newer materials for the fuel rods can be classified as cladding material (the long slender tubes) and as fuel material (the ceramic pellets based containing the fissile uranium). One of the simplest and most elegant manners of removing the zirconium alloys from the reactor core is to replace them using FeCrAl alloys (Rebak 2020a). The newer cladding will have the same external geometry as the current cladding, but a thinner wall thickness to compensate for parasitic thermal neutron absorption. Ferritic type of bcc FeCrAl alloys such as APMT (Fe + 21Cr + 5Al + 3Mo) and C26M (Fe + 12Cr + 6Al + 2Mo) are excellent candidates due a suite of attractive physicochemical properties and low cost of fabrication. The tubes of these alloys will be fabricated using powder metallurgy and extrusion followed by pilgering on a mandrel (Huang et al. 2022). The caps can be welded to both ends of the tubes using pressure resistance welding. FeCrAl alloys have outstanding resistance to attack by superheated steam at temperatures at least up to 1000 °C. FeCrAl alloys are also resistant to corrosion and environmental cracking in ∼300 °C water and have high mechanical properties at temperatures of 300 °C and higher (Huang et al. 2022; Rebak 2017; Rebak et al. 2020; Rebak et al. 2022; Yin et al. 2021).

Figure 2:

Accident tolerant fuel (ATF) rods consist in changing the cladding and/or the fuel itself.

Since the FeCrAl alloys are ferritic and their elements do not react with hydrogen to form stable hydrides, there were concerns that this will result in a higher content of tritium in the reactor coolant. This tritium would migrate from the fuel cavity into the coolant across the cladding wall (Evans and Rebak 2022; Garud et al. 2022; Rebak 2020a). The purpose of this review is to gather literature data that supports the evidence that surface oxides hinder the hydrogen isotopes entrance into metallic structures which rely on a stable and passive surface film for corrosion protection. The main area of the review will be for Fe-Cr alloys with the focus on FeCrAl alloys and their oxides.

In a previous review Garud et al. (2022) described the characteristics of all isotopes of hydrogen permeating though non-oxidized Fe based alloys, both austenitic and ferritic, which are currently used or intended to be used for nuclear applications. They broadly described the basic principles of hydrogen diffusion and the factors controlling diffusion and flux. Their review of literature data demonstrated that the solubility of hydrogen in ferritic alloys is lower than in austenitic alloys, but hydrogen permeabilities higher through ferritic materials than through austenitic materials (Garud et al. 2022). Also, literature information showed that the tritium permeation rates in FeCrAl alloys were between those in austenitic stainless steels and in ferritic FeCr steels. The activation energy for hydrogen permeation was approximately 30% higher in the austenitic alloys compared with the ferritic alloys (typically ∼50 kJ/mol in ferritic vs. ∼65 kJ/mol in the austenitic). As mentioned above, Garud et al. (2022) did not address in detail the effects of surface oxide on hydrogen/tritium entrance and permeation through the iron based alloys. These effects are covered in this review.

2 Hydrogen entrance and permeation through oxidized steels

2.2 Some empirical observations for metal oxides

The permeation response of a gas (diffusant) through a solid, in general, is typically described in terms of the parameter called permeability, P, of the solid–diffusant pair at a given temperature in a given external environment, and it is a normalized measure of the rate of permeation. More specifically, the permeability (P) quantifies the permeation rate of flow of a diffusant crossing unit area and through unit thickness of a fixed layer of solid, per unit partial pressure on one interface (upstream) and nearly zero partial pressure of the diffusing species at the other interface (downstream). Furthermore, the temperature dependence of P has been well described by the Arrhenius relation:

(1)P=P0⋅exp(−Qp/RT)

where,

P ₀ : is permeability constant or permeability pre-exponent (mol/[m s √MPa])

Q _p : is activation energy of the permeation process (kJ/mol),

T: is temperature on the absolute scale (K),

R: is the universal gas constant (8.3145 × 10⁻³ kJ/mol-K).

For additional details and discussion on the above characterization of permeability the reader may consult Garud et al. (2022).

Empirically, at least since 1935, metal oxides have been shown to display very low permeability for hydrogen, for example, in aluminum (Smithhells and Ransley 1935), in steels (Flint 1951), in zirconium (Gulbransen and Andrew 1959), and in FeCrAl alloys (Huffin and Williams 1960, Kripyakevich et al. 1971b with particular reference to the spinels). Additionally, it was noted that the natural or passive oxide layer on steels and other alloys, in general, reduces the hydrogen permeation rate through/over that of the bulk unoxidized alloys, so that the surface oxides of Cr and Fe also provide a permeation barrier effect.

Strehlow and Savage (1974) reported generally a factor of 100 or more reduction in the permeation rate of hydrogen isotopes through several oxidized heat-resisting alloys, for the hydrogen pressure range of 1.33 × 10⁻⁷ to 0.1 MPa and the temperature range of 573–1073 K. (From the scientific and technical application points of view the primary unit of temperature used in the following is based on the absolute scale, i.e., in degrees Kelvin (K); for conversion: temperature expressed in °C = K + 273.15.) Also, another synopsis (Maroni 1978) provided permeability reduction factors (PRFs). (PRF is defined as the ratio of permeation rate or flux through a given area obtained with no surface modification such as any oxide formation, coating, or special treatment, to that with a modification of/on the original substrata, with other conditions of permeation kept the same. The latter implies that the PRF is defined and is applicable for a specific combination of the substrata, the environment, and the surface condition; it serves a useful but limited purpose.) The PRFs of up to five orders of magnitude, depending on the level and integrity of oxide layer, were reported in several austenitic steels and refractory alloys, and of 200–600 for tritium in nickel/iron base alloys under steam (Maroni 1978). While these values are not directly applicable to FeCrAl alloys in LWR conditions these observations are indicative of the impact of surface oxides on PRFs.

Data reviewed by Chandra et al. (1976) showed more than three orders of magnitude reduction in tritium release rate, compared with that in the bulk niobium, below about 800 K (down to 300 K) in the presence of oxidized surface film; the reduction factor was still about 25 even in a reducing environment adversely affecting the oxide film integrity. Van Deventer and Maroni (1980, 1983 also reported significantly reduced hydrogen permeability due to the presence of oxides on FeCrAl alloys in the temperature range of 423–1073 K and at pressures from 2 × 10⁻⁶ to 0.015 MPa, e.g., see Figure 3.

1. For consistency and ease of comparison of data from varied sources and different units used by the cited authors, all data and parameters were converted in this work to the common unit of mol/[m s √MPa] for permeability. Where parameters were not given in the original work these were estimated graphically to best-fit the data trend or charts shown in the cited sources.

2. For the same reason, and tritium evaluation being of primary interest, the permeabilities displayed in all Figures are normalized or scaled by the isotopic ratio (Garud et al. 2022) of 1/√3 for hydrogen and √2/√3 for deuterium data, identified by the first letter of each label – H/D/T for hydrogen/deuterium/tritium – in the included Figures. However, all parameters in the Tables retain the original values and the actual isotope used, without normalization, unless otherwise noted.

3. The context for text noted in the associated labels is the reference cited, if any, in the label entry of all Figures.

4. Also, for expediency, the data references embedded in the labels of all Figures use the shorthand notation with the first four letters of the first author followed by the four numbers for the year of publication.

5. In all Figures of comparison the Arrhenius tritium permeability line for ferritic type 406 stainless steel with 83 wt% Fe, 13 wt% Cr, and ∼4 wt% Al (Bell et al. 1979) is included as a common anchoring line to facilitate visualizing/positioning of data/plots between the Figures for comparisons.)

Figure 3:

Comparison of tritium permeability response of some Fe-Cr steel alloys with or without aluminum, mostly in as-received oxidized or unoxidized conditions. Also included is the curve for (bcc) alpha-iron from Nelson and Stein (1973).

2.3 FeCrAl alloys in terms of LWR applications

FeCrAl alloys are specially developed steels with high alloying content of aluminum and chromium; these (and more common stainless steels) are modified forms of carbon steels developed specifically to better resist high temperature oxidation (corrosion). This increased resistance is imparted by the product of oxidation itself, owing to the stability and integrity of the protective scale of oxides containing chromia and alumina, provided sufficient levels of aluminum and chromium remain to maintain these oxides on a continued basis. At relatively lower temperatures of normal LWR service, starting with clean base metals, iron oxides and spinels (mixed oxides) are expected to be in the outer scales. Over the entire range of temperatures, a combination of these oxides will be found, depending on the temperature, oxygen partial pressure (at the fluid–metal interface), and the relative alloy content of the elements Fe, Cr, Al. The ease with which selective oxidation versus internal oxidation can occur, which also is dependent on these same conditions, has significant influence on the type and characteristics of the oxide layer, including morphology and defects distribution, which in turn impact, if not govern, the permeability of hydrogen and its isotopes.

Under environmental conditions simulating the normal LWR service, corrosion response of FeCrAl alloys has been investigated by many, as discussed under “Oxides” section below, demonstrating fairly well the excellent general passivity due to the protective oxides formed on these alloys. In some respects, e.g., the relatively much lower weight loss and/or oxidative growth, the FeCrAl alloys show superior corrosion properties compared to those of the 300 series austenitic steels (without aluminum). It is clear that the outside diameter of the cladding (facing the aqueous environment) will have developed and maintained a protective oxide (Figure 4) that can act as a resistance layer to the permeation of tritium from the fuel side of the cladding.

Figure 4:

Representation of the cladding in the fuel bundle of a light water reactor. The ID of the cladding may develop Al₂O₃ while the OD of the cladding may develop Cr₂O₃ plus spinels (depending on the environment).

On the upstream side facing the uranium dioxide fuel pellets the environment is much different, with nearly dry conditions and somewhat higher temperature than on the coolant side, and relatively low oxygen partial pressure (near equilibrium with the uranium dioxide). The low partial pressure of oxygen is expected to favor the oxidation of aluminum (Figure 4) over other elements. (It is interesting to note here that for the typical service conditions of PWRs and BWRs noted earlier, with cladding temperatures of 630 ± 5 K and 590 ± 5 K, respectively, the oxygen partial pressures were sufficient to form in-situ formation of oxides on the inside surface of Zircaloy cladding facing the fuel elements/gap as reported by Cubicciotti and Jones (1978). The operating temperature is not high enough to form the highly protective α-alumina normally formed in oxidation above about 1150 K; however, other forms of the polymorphic aluminum oxide are expected at the lower temperatures, also discussed under “Oxides” section below. Also, due to its highly stable nature, any α-alumina developed by pre-oxidation will be maintained under non-aqueous conditions, provided the aluminum content of alloy is not too low. In either case, due to the great thermodynamic affinity of aluminum for oxygen, the polymorphic aluminum oxide is expected to be of the self-healing type. Also, with increase in the required or desired burn-up (longer residence time of the fuel in the reactor) with progression in a typical current and anticipated fuel cycle, the oxygen partial pressure is likely to stay higher than the dissociation pressure of the in-situ aluminum oxide.

Therefore, as a result of the LWR operating conditions, it is of interest to examine the role or influence of various oxides (of iron, chromium, and aluminum) in the tritium permeation. Furthermore, to a first order, this influence can also be examined independent of the base alloy and it is covered in the next Section.

3 Review and assessment of data on oxidized fecral and steels

Additional results on the reduction of hydrogen permeability due to the presence of oxide films on alloyed steels are presented and discussed briefly in this section. It includes data mainly on FeCrAl steels, with select data on austenitic and RAFM steels for comparison, in oxidized conditions. Since the oxides and their influence on the permeation are generally similar in nature in these steels, the presentation is more chronological than ordered or separated by alloys classification.

One of the earliest works with careful consideration of the presence of oxide films in determining the hydrogen permeation response was reported by Flint (1951). It showed that the oxide films on AISI 347 stainless steel resulted in almost 400-fold reduction in permeation rate over that through the clean (non-oxidized) steel under identical test conditions that were even conducive to transient permeation rate due to some reduction of the oxide with test time. In addition, in the case of stainless steels, Flint suggested that some form of oxide layer development—oxidizing the base alloy by annealing in appropriate environment, alloying a thin layer of aluminum into the surface and then oxidizing it, or developing a glassy coating on the surface—effectively, would reduce the hydrogen permeability by several orders of magnitude over a wide range of temperatures of practical interest.

Huffin and Williams (1960), whose work on FeCrAl alloy was mentioned above, noted that the three orders of magnitude reduction in hydrogen permeation rate on oxidized surface (relative to that of the clean surface), shown in Figure 3, was likely due to the relatively slow diffusion of hydrogen in the oxide layer, even at the high test temperatures, above 922 K to about 1450 K. Steady-state permeation rates under conditions maintaining the oxide layer were too low to be measurable below 1255 K.

The reduction in hydrogen permeation response of several oxidized FeCrAl alloys (0.1–19.5% Cr and 0–8.9% Al) in the temperature range of 573 K–973 K was reported by Kripyakevich et al. (1971a), including effects of long term exposure to hydrogen atmosphere at the high temperature of 973 K. Additional results on hydrogen permeability reduction in these materials were presented with particular reference to the structure and composition of oxide films (Kripyakevich et al. 1971b) which is discussed later. The influence of oxide layers in FeCrAl alloys on the diffusion and permeation of hydrogen was also discussed with particular reference to the effects of grain-boundary contribution and Fe₃Al phase formation above ∼813 K (Sidorak et al. 1974).

The tritium diffusion investigations by Elleman et al. (Austin et al. 1973; Chandra et al. 1976; Elleman and Verghese 1974) appear to be the only ones to measure tritium concentration profiles in detail including the near surface regions and relate these to the observed diffusion response. They estimated the diffusivity of surface region to be about 2–3 orders lower than that of the bulk region in Type 304 stainless steel (in the temperature range of 319 K–453 K) and about 7–8 orders lower in Zircaloy-2 (between 521 and 653 K). The reduced diffusivity was attributed to the likely effect of trapping of tritium in the surface region (about 5 μm thin in the steel) that was a layer of an oxide film. The later study (Chandra et al. 1976) focused more on oxide films on Niobium.

The earlier work on permeation of deuterium in Type 309S stainless steel (Swansiger et al. 1974) considered the role of surface oxide growth (10–250 nm thick) and oxide composition under several oxidizing conditions, in the low temperature range of 523 K–773 K at 0.0133 MPa. The original as-received samples also had about 20 nm oxide layer. That work showed only minor influence on the permeabilities—with values within a factor ∼2 on the estimated mean of 16 samples with a wide variety of oxidation and surface treatments—apparently due to reduction/alteration of the oxide layer during tests and/or noted lack of oxide integrity. As such, the mean linearized response estimated here from the 16 samples is viewed to represent a mildly oxidized condition of 309S steel and the resulting estimates of permeation parameters are included in Table 1 (the lower section) for comparison with other data as discussed below. Nevertheless, with additional testing, it was also reported that the thin iron oxides were effective in reducing the permeability by a factor of 10 prior to the oxide reduction, at T > 570 K, by hydrogen (Swansiger et al. 1974).

From the work of Louthan and Derrick (1975) it is to be noted that the significant effects on the lowering of apparent permeability of deuterium during the progression of tests on stainless steels were attributed to the formation of oxide films during the tests which were too thin to be identified by X-ray diffraction and interference color matching, indicating a typical thickness of less than 250 nm. This measurable dynamic or transient effect on hydrogen permeation due to the growth (or reduction) of oxide film was also noted above in Type 347 steel (Flint 1951) and in FeCrAl (Huffin and Williams 1960).

In a follow-up work (Swansiger and Basatz 1979), three different stainless steels (309S, A-286 and 21-6-9) were investigated for the effect of thin oxide films by chemical etching. The permeability results for deuterium (at 475 K–700 K) and tritium (at 325 K–475 K) were compared with those for sputter-cleaned, palladium coated bulk alloys, with the upstream gas pressure in the range of 0.0013 MPa–0.1 MPa. The presence of thin oxide films reduced the permeability by two–three orders of magnitude. It was also reported that the barrier effect was dominated by the upstream surface oxides and that the barrier effect of a sample oxidized on both sides could be negated by depositing palladium on top of the upstream oxide. These graphical data are used in this paper to estimate the resulting permeability parameters for the oxidized samples of 21-6-9 and modified A-286 stainless steels. (The pressure dependence showed deviation from the square-root response in the data for oxidized 21-6-9 steel, but the data were fitted in the original works of Swansiger and Basatz (1979) with the assumption of square-root dependence. This contributed and displayed some extra scatter around the Arrhenius relation over the covered temperature range of 325 K–700 K.) These parameters are included in Table 1, which are further compared with other oxidized alloys as shown in Figure 5 – note: the permeability scale in Figure 5 is extended relative to that of other figures. Also shown in Figure 5 are the estimated bounding permeability lines for unoxidized/clean austenitic steels (Garud et al. 2022). This comparison, as well as that of Figure 6 discussed below, confirms the effect of oxide layers on the reduced permeabilities for variety of steels, including several FeCrAl alloys, over the temperature and pressure ranges also listed in Table 1.

Figure 5:

Permeability response of representative (fcc) austenitic stainless steels in various oxidation conditions, and comparison with the estimated bounding/central trend curves for the class of austenitic stainless steels.

Figure 6:

Range and comparison of permeability response of representative (bcc) FeCrAl alloys in various oxidation conditions.

Similar to the above work (Swansiger et al. 1974) on Type 309S steel, Swansiger et al. (1984) also investigated in detail the effect of various surface conditions with oxidation treatments on the steady state permeation of deuterium in an FeCrAl alloy – Fe with 15.8% Cr, 4.8% Al, 0.3%Y – at low temperatures below 723 K, with fairly similar results. The Arrhenius temperature dependence was reported for the most part with a fixed pressure level of 0.013 MPa to avoid influence of possible deviation from the square-root pressure dependence with the surface/oxidation condition, although tests were done in the wider pressure range of 1.33 × 10⁻⁵ MPa to 0.1 MPa. Results were reported on 18 different oxidation conditions, with or without palladium over-coating, although the oxidation was performed in air at fairly high temperatures (1173 K–1423 K). The observed PRFs ranged from at least 10 to about 500, relative to the unoxidized, clean or palladium-coated, reference condition. These permeability results on the 18 surface-oxide conditions can be viewed with some simplicity in four parts—permeability parameters for these cases as estimated in this paper are given in Table 1; these are also shown by the four dotted lines in Figure 6, where Swansiger’s reference line of unoxidized, palladium-over-clean condition is added for comparison:

1. Unpolished, oxidized – displaying the lowest permeability, two to three orders of magnitude lower than the reference and with highest activation energy of 77 kJ/mol.

2. Also unpolished, oxidized but palladium coated on both sides – displaying the highest permeability, still an order magnitude lower than the reference but with similar low activation energy, 45.2 kJ/mol.

3. Polished before oxidizing – response midway to that of (a) and (b), activation energy of 66.7 kJ/mol.

4. Remaining 15 surface-oxide conditions – with response between that of (b) and (c), and within a relatively narrow band spanning a factor of four in the permeability. The mean linearized response for these 15 sample conditions is estimated here with the Arrhenius parameters of permeability: pre-exponent of 4.217 × 10⁻⁵ mol/[m s √MPa] and activation energy ∼63.5 kJ/mol.

Van Deventer and Maroni (1983) summarized their work on several FeCrAl steel compositions tested for hydrogen permeation response over a wider range of 423 K–1073 K and upstream pressure range of 2 × 10⁻⁶ MPa to 0.015 MPa. Based on their graphical summary of these data from Figure 3 of their paper the permeability parameters for central trends are estimated here for three alloy compositions which are included in Table 1 and shown in Figure 6. Figure 2 of their paper also had the permeation data on one sample of 16Cr-5Al alloy tested at randomly selected temperatures and pressures in the rage, over a course of several weeks. The resulting estimates of permeation parameters are also included in Table 1 and the Arrhenius plot for tritium permeation is shown in Figure 6 which is at the upper limit for all the oxidized FeCrAl alloys data—this sample likely represents a partially oxidized condition and a pseudo-steady state response resulting from the variable conditions used in testing this sample.

Results of an investigation on the effectiveness of oxidized Fecralloy – a ferritic stainless steel with 5% Al and 0.3% Y – in reducing the hydrogen isotope permeation were reported by Forcey et al. (1985). In that work also, similar to the earlier work by Huffin and Williams (1960) on FeCrAl steel with 20% Cr and 5% Al, the apparent tenacity and effectiveness of thin oxidized layer on the Fecralloy were confirmed, showing three or more orders of magnitude reduction in permeation (relative to the unoxidized condition) but in a sample oxidized in air for 24 h at 1073 K. The oxide characteristics were attributed to the formation of alumina by the prolonged high temperature oxidation. Also, the addition of yttrium in Fecralloy was considered to enhance the oxide scale adhesion (Wukusick and Collins 1964, Quadakkers et al. 1991). Approximate estimates of the permeation parameters for other lesser degrees of oxidation – labeled here as low, medium and high – are determined here from the graphical results using Figure 6 of Forcey et al. (1985) and listed in Table 1, and the resulting tritium permeability response curves are shown in Figure 6.

The tritium permeability through Type 304L stainless steel naturally oxidized in room air on the upstream surface, with sputter-cleaned Pd-coated downstream permeating surface, was reported to be a factor of about 34 lower than that of the sputter-cleaned Pd-coated on both surfaces, in the temperature range of 373 K–443 K (Maienschein et al. 1987). These results are included in Figure 5.

Maienschein was also the first to use aluminum ion implantation followed by selective oxidation to generate an alumina-rich layer integral with the substrata of Type 304L stainless steel, to examine the effectiveness of this oxide layer in reducing the permeability of tritium (Maienschein et al. 1987). As a result, a reasonably coherent, crack-free alumina-rich oxide layer, only about 30 nm thick, was grown on the steel, first reducing the oxides of Fe and Ni on the surface but not reducing any Cr₂O₃ on the aluminum-ion implanted specimens, and then oxidizing at 973 K in low oxygen activity atmosphere. The data taken at 373 K and 443 K showed that the resultant Al₂O₃ layer reduced the permeability by 600–700 fold compared with clean stainless steel case. However, this reduction was considered to be less than expected from the diffusivity value of alumina, and this was attributed to less than perfect coverage by the alumina layer. Also, the activation energy of tritium permeation for the oxidized samples, estimated here to be 60.9 kJ/mol, was only slightly higher compared to 57.7 kJ/mol for the samples with (unoxidized) sputter-cleaned Pd-coated surfaces, over the relatively short temperature range. The estimated permeation parameters for the oxidized case are included in Table 1 and its permeability response is shown in Figure 5.

It is important to note that the various studies done with palladium coated surfaces are expected to yield upper limits of permeabilities for the underlying substrata and often meant to give baseline data for comparison purposes; in practice, components or surfaces are not palladium coated and are typically subject to natural or in-situ grown oxides. An example of the difference in these conditions is illustrated here in Figure 5 that shows at least two orders of magnitude lower tritium permeability in the naturally oxidized versus palladium coated surface in a typical Type 316L stainless steel in the temperature range of ∼500–800 K (Shan et al. 1991; Yao et al. 2000), and in the Type 304L response (Maienschein et al. 1987) noted just above.

About 30 nm thin oxide layer on 6 mm MANET II base alloy was found to reduce the hydrogen diffusivity by a factor of about 100 at 573 K and five at 973 K (Peruo et al. 1997), where the inner oxide layer was reportedly a chromium spinel containing manganese.

Results of deuterium permeation on an oxidized FeCrAl steel, with 22 wt% Cr, 5 wt% Al, Y, and Zr, in the temperature range of 635–850 K (Xu et al. 2016) (included in Table 1 and in Figure 6) showed more than a factor of 10 lower rates compared to the data on APMT steel (Hu et al. 2015), with very high activation energy, or about 55 compared to the central or best estimate for the group of unoxidized FeCrAl alloys as illustrated by Garud et al. (2022). Interestingly, in this case although the oxidation was carried out at a moderate temperature of 1073 K for 90 h in air, a layer of mainly α-alumina about 150–450 nm was noted by Xu et al. (2016).

Another study (Huang et al. 2020) showed that the PRF of at least an order of magnitude (∼15–25 times that of clean/non-oxidized, palladium coated sample of the same ferritic stainless steel) resulted from the naturally grown oxide layer (∼200–250 nm) on a commercial 430 stainless steel. This study was on the permeation of deuterium in the temperature range of 623 K–873 K and upstream (driving side) pressure of about 0.04 MPa–0.10 MPa. Although the stainless steel (with 16 wt% Cr) did not contain aluminum, it is included here as a ferritic steel and since the XPS results showed the oxide to be spinel containing Cr and Mn.

Observations on the permeability reduction due to oxidation of FeCrAl steel with Ce-oxide dispersion reported in a recent work (Urabe et al. 2020), in the lower temperature range of 373–623 K, are discussed here in more detail. Prior to start of the permeation test, some of the specimens were subject to oxidation at the low temperature of 563 K for 30 days in an autoclave steam environment which produced an oxide layer (about 0.3 μm thick) that was mostly chromium oxide. It was noted that the oxide was removed from the upstream, high pressure, tritium loading side of the permeation cell but maintained on the downstream (permeating out) side. It was also noted that the reduction in permeation due to the oxide layer was much larger when the permeating (oxide) side was facing moisturized Ar (purge gas) as compared to Ar with 1% hydrogen (Table 1). Also, in the former case, the temperature dependence was more significant – steeper slope in the usual Arrhenius plot, with very high, 96 kJ/mol, activation energy of permeation – than with the apparently reducing type purge gas environment. However, no consistent or clear explanation of the reason for these observations was apparent. Although a definite permeation barrier effect of the oxide layer was observed, the effect was subject to some interaction between the surface layer and external environment’s oxidizing potential and/or composition. This is in addition to the expected influence of the state of integrity of the surface (oxide) layer. As such, further work in this area, especially with the pertinent range of expected environment on both sides of the cladding, should be useful, if not essential, to better assess the barrier response to permeation of tritium.

Some of the spread in the reviewed permeation data plots and permeability curves (Figure 6) is likely due to the common use of normalizing and expressing the pressure dependence with the fixed square-root relation when some undetermined/variable deviation from this relation is expected but neglected for the tested oxidized surface conditions. Alternative or additional contributor to the spread: the effective surface area of diffusion through the substrata or bulk alloy is reduced and unaccounted variably depending on the overall state/integrity of the oxide.

Notwithstanding the wide variety of oxide/surface preparation and conditions and other factors contributing to the additional spread in data as noted, based on the results shown in Figure 6 for oxidized FeCrAl alloys and discussed above, the following bounds and central trend are considered approximate but reasonably conservative for the tritium permeability response [see Equation (1)] in oxidized FeCrAl alloys:

• Upper bound (estimate): P₀ = 3.36 × 10⁻⁵ mol/[m s √MPa] and Q_p = 58.0 kJ/mol

• Central trend (estimate): P₀ = 1.60 × 10⁻⁵ mol/[m s √MPa] and Q_p = 60.0 kJ/mol

• Lower bound (estimate): P₀ = 7.62 × 10⁻⁶ mol/[m s √MPa] and Q_p = 62.0 kJ/mol

Compared to the similar estimated bounds for unoxidized/clean FeCrAl steels (Garud et al. 2022) it is seen that the activation energies for oxidized steels are about 23% higher. Figure 7 is a composite plot that summarizes and compares the estimated permeation response bounds of representative FeCrAl alloys in clean versus oxidized conditions.

Figure 7:

Estimated bounds/central trend curves for tritium permeability response of oxidized FeCrAl alloys, and comparison with similar estimates for unoxidized condition.

Compared in Figure 8 are the empirical central trends estimated for the tritium permeation in unoxidized/clean condition of four general classes of alloys: (ferritic) RAFM steels, modern or modified RAFM and typical 400-series steels (without aluminum), aluminum containing FeCrAl steels, and (austenitic) stainless steels. Also included in Figure 8 is the estimated central trend for oxidized FeCrAl class of steels which is seen to be about a factor of three lower than the clean austenitic steel response and with nearly the same activation energy.

Figure 8:

Comparison of best estimate (central trend) response for tritium permeability of class of oxidized FeCrAl alloys with the estimates for unoxidized FeCrAl, RAFM, modified RAFM, and austenitic stainless steels.

4 Oxides

4.1 The role of oxides in decreasing hydrogen flux through FeCrAl and related steels

The majority of the oxidation response and related information on oxides described in the literature on FeCrAl alloys has emphasized the high temperature regime, typically above 1100 K, as may be expected from the intended application of FeCrAl. Although some of this literature is mentioned here for the context or references, few studies at lower temperatures are covered in some detail here mainly in support of the expected species, morphology, and pertinent characteristics of the oxides on the FeCrAl alloys. Therefore, this limited summary and discussion on oxides in relation to these alloys is covered below in four parts:

1. Related but general observations

2. Oxidation or corrosion in aqueous environments typical of LWR coolant conditions

3. Low temperature oxidation in non-aqueous environments with relatively low oxygen partial pressure

4. Permeability of alumina (Al₂O₃), chromia (Cr₂O₃), and iron oxides/spinels.

4.1.1 Related/general observations on oxides

Compared to the recent interest and testing in the LWR aqueous environments discussed below, the characteristics of oxides of ternary FeCrAl alloys in other reactor systems or applications have been extensively investigated, at least since the early 1960s (e.g., Wukusick and Collins 1964; Weisenburger et al. 2013) and earlier, e.g., Kornilov’s work reported by Case and Van Horn (1953a, 1953b), Chub et al. (1958), and Gulbransen and Andrew (1959). For example, improvement in the quality of the Al₂O₃ surface oxide was reported for Fe-25Cr-4Al alloy by the addition of reactive element (1 wt% yttrium), which was attributed to the increased adherence (Wukusick and Collins 1964; General Electric 1966) and subscale Y₂O₃ oxide likely promoting its sintering effect with Al₂O₃ or physical locking effect for the surface oxide. (The order of “Standard Free Energies” of oxide formation (Ellingham diagram)—from lowest, most stable and most oxygen-active to highest, less stable and less oxygen-active—is Y₂O₃, Al₂O₃, Cr₂O₃, and Fe₃O₄, over the temperature range of 300 K–2000 K (Barin, Knacke, and Kubaschewski 1977). It was reported that the (dry) air oxidation of FeCrAlY alloys is relatively independent of chromium content, whereas the steam corrosion resistance is lowered as the chromium content is lowered, and that pre-oxidation is necessary to retain steam corrosion resistance and freedom from embrittlement at low temperature (General Electric 1966). Kornilov’s work showed that the oxide film on FeCrAl alloys below 973 K in dry conditions consists of a solid solution of Al₂O₃, Cr₂O_3, and Fe₂O₃ in approximately the same ratio as the alloying elements, that is, an isomorphic mix of the spinel type oxides, with the lattice constant of oxides decreasing (first) to that of the γ-alumina with increasing temperature. Generally, in dry gaseous environment, the formation of an outer layer of chromium oxide increases the activity of aluminum underneath and inward diffusion of oxygen with outward diffusion of aluminum lead to internal oxidation forming alumina layer even with low oxygen potential. In the aqueous environment any prior (transitional) alumina likely dissolves in time but protective chromia or chromium-rich layer remains or develops.

It is interesting that, in a limited area of Zircaloy cladding internal surface examined from a PWR reactor (Cubicciotti and Jones 1978), a particle presumed to be alumina was found embedded in the surface. Also, among the chemical species from the water touched components which were found in actual steam power systems included the chromia (Cr₂O₃) and α-alumina (Al₂O₃) oxides (Jonas 1981) (Based on the phase stability and other related studies of aluminum oxides/hydroxides, the alumina oxide in these systems very likely is of the low-temperature transitional type instead of the high-temperature corundum type.) Presence of boehmite or Böhmite, a monohydrate-alumina (Al₂O₃∙H₂O), was also serendipitously found on the fuel cladding in the early experimental BWR (EBWR) with deionized water (Bredden 1963). (Aluminum alloys were extensively investigated up to 623 K and used in early water-moderated reactors for fuel cladding application, e.g., see USAEC 1960, Bredden 1963, and Dickinson 1965; this cladding work is not discussed further.)

4.1.2 Oxides in low-temperature aqueous environments

In the limiting case of ferritic steels with only Fe and 11 to 29.9 wt% Cr as major elements, observations on the passive oxide films in oxygenated water — albeit at the low temperatures of 298 K and 343 K — were reported based on the electron spectroscopy for chemical analysis (ESCA) by Olefjord and Fischmeister (1975). In particular, it was reported that:

1. The passive films consisted of two layers: a very thin outer layer of Cr(OH)₃ with a possible small content of trivalent iron — the hydroxide should dehydrate to chromia at higher temperature, such as 600 K — and an inner layer of possible (Fe, Cr)₃O₄ spinel.

2. The inner layer was similar to that seen for the same alloys in pure, dry oxygenated environment.

3. The composition changes were seen to be in accordance with the greater thermodynamic stability of Cr-rich oxides and, in the case of oxygenated water, consistent with preferential dissolution of iron in water leaving behind a chromium-rich layer.

4. The composition of dry-formed oxide film on the same alloys, which consisted mainly of iron oxide, slowly changed reaching that of the water-formed film, when exposed to water.

With respect to the structural (amorphous versus crystalline) character of typical passive oxide films the following observation may be noted. In the case of both the austenitic stainless steels and the ferritic (Fe-Cr) steels, Olefjord and Fischmeister (1975) noted that the electron diffraction results showed no distinct crystalline diffraction pattern and that the film analysis was difficult due to their amorphous character, which would indicate that amorphous oxides on a wide range of steels support, if not enhance, the passivity afforded.

In a limited but long-term (500 days) testing of a Kanthal APM alloy, with 22 wt% Cr, 5.8 wt% Al, and balance Fe, in simulated PWR water at 633 K superior corrosion resistance was reported along with the observation that the primary oxide phase consisted of α-Fe₂O₃, but with some porosity at the metal–oxide interface likely due to differential rates of stronger outward diffusion of iron vis-à-vis the inward oxygen diffusion (Park et al. 2015). The α-Fe₂O₃ was apparently related to lack of control of oxygen activity with excessive dissolved oxygen in the test water (Terrani et al. 2016). Similarly, recent summaries (Rebak et al. 2017; Terrani et al. 2016) illustrating the oxide morphology and composition of austenitic alloys, along with a detail comparison with similar characterization of oxides on the proposed APMT FeCrAl cladding alloy (An advanced powder metallurgical iron-base alloy with 22 wt% Cr, 5 wt% Al, 3 wt% Mo, and yttrium.) have been provided based on their long-term (about one year) testing under simulated oxidizing and reducing water chemistries of BWR and PWR normal operation near 573 K.

These and related results (Rebak 2018, 2020b) confirmed that the FeCrAl APMT material always developed a protective chromium-rich oxide film; the film in oxygenated water was double-layered—containing iron and chromium as hematite and spinel in the outer layer, and a thinner chromia containing inner layer—while in hydrogenated water it was single layered oxide also rich in chromium, and aluminum was not detected in these oxides (Gupta et al. 2018; Rebak et al. 2017). Similar observations on the oxides, corrosion resistance, and oxide stability on several other experimental FeCrAl alloys have been reported under simulated radiation effects and LWR radiolysis environment (P. Wang et al. 2020). More recent work detailing the study of oxides developed on the surface of FeCrAl alloys (APTM and C26M) under immersion tests up to 12 months in exposed simulated LWR water at ∼300 °C showed the presence of Cr rich inner layer (Yin et al. 2021). Additionally, tests in the simulated BWR-NWC water showed presence of Al in the oxide–alloy interface region, and the oxide on C26M also confirmed the inner oxide layer to be rich in Al (Yin et al. 2021).

As such, the common appearance of single or multilayer oxides providing the general passivity in the case of aqueous environments under normal operating conditions in (austenitic) stainless steels is applicable to the ferritic Fe-Cr and FeCrAl alloys.

4.1.3 Oxides in low to intermediate temperature dry atmosphere

Generally, at lower temperatures, the slower diffusion of aluminum relative to that of Fe and Cr, and the relatively low content of aluminum in the FeCrAl alloys of interest, favor the formation of iron and chromium oxides provided the environment’s oxygen activity is higher than their dissociation pressure. The initial oxides of aluminum at lower temperatures (below about 1000 K) are different polymorphs of the high-temperature α-Al₂O₃ which may transition to the α-Al₂O₃ depending on the time and temperature. The amorphous forms of aluminum oxides have been reported on aluminum, for example, up to 773 K (Shimizu et al. 1991a, 1991b), transitioning first to crystalline γ-Al₂O₃, while the formation of α-Al₂O₃ on FeCrAl alloys after 168 h in dry oxygen at 973 K was reported with XRD results (Josefsson et al. 2005). This relatively low temperature of formation for the α-Al₂O₃ was attributed to the presence of chromia (in the initial oxide) with its similar corundum structure facilitating the nucleation of α-Al₂O₃. (Growth of α-Al₂O₃ at the low temperature of 673 K was demonstrated, for example by Jin et al. (2002), by sputtering on a Cr₂O₃ template layer, with a later work (Andersson et al. 2004) showing successful development of single-phase α-alumina thin films, with high degree of crystallinity, at even lower temperature of 553 K on Si and at 833 K on Haynes230.)

Oxide morphology of a commercially available Kanthal APMT alloy was investigated at low to intermediate temperatures in the range of 573 K–873 K after aging for 100–1000 h in stagnant air (Li et al. 2018a). At 673 K and below, the oxide was amorphous, mixed or spinel of Fe and Cr. Increasing evidence of crystalline α-Al₂O₃, located mainly at the inner layer of the oxide, was reported at 773 K with increasing time at 100 h and above. The multilayered oxide was formed at 873 K with Al enriched inner region and Cr-Fe enriched outer region of the oxide. It was noted that a continuous crystalline α-Al₂O₃ can be developed at temperatures as low as 873 K by several hours of aging in stagnant air. Similar observations were reported also on a specially designed FeCrAl alloy ‘C26M’ (Fe-12Cr-6Al-2Mo-0.2Si-0.03Y), with 100–2000 h aging (Li et al. 2018b; Li et al. 2019); the main difference noted was that of even lower transitioning temperature of 773 K for the appearance of a continuous crystalline α-Al₂O₃. For both the Kanthal APMT and C26M compositions and test conditions examined, the kinetics of oxide growth was reported to be too slow to form a thick enough barrier.

Formation of protective layer of alumina, thought to be κ-alumina, has been reported in the temperature range of 673 K–873 K (Ejenstam et al. 2017) on several FeCrAl alloys, with compositions varying from 10 to 13 wt% Cr and 4 to 6 wt% Al; although this report deals with the lead or lead-bismuth eutectic environments, the conditions represent very low oxygen partial pressures – controlled to 10⁻⁷ wt% dissolved oxygen in liquid lead in their own tests. The protective layer formation was noted to require additions of minor reactive elements (RE), Ti, Zr, and Y, to the base alloy optimized relative to the carbon content suppressing Cr-carbide precipitation. Similar earlier study on FeCrAl model alloys (Weisenburger et al. 2013) also showed that 12.5–17 wt% Cr and 6–7.5 wt% Al are high enough to obtain thin, stable and protective (transient) alumina scales when exposed to 10⁻⁷ wt% dissolved oxygen in liquid lead at 673, 773, and 873 K. Observations on forming protective alumina under similar conditions on FeCrAl steels for lead or lead–bismuth environments, with estimated dissolved oxygen of 2.6 × 10⁻⁶ wt% at 773 K and 1.3 × 10⁻⁵ wt% at 873 K, were reported in another work (Lim et al. 2013). In that work, Cr and Nb were seen to have beneficial effect in forming a continuous alumina oxide layer, and in optimizing the aluminum content of the FeCrAl alloys, where external or internal alumina layers could be formed depending on the alloy composition and aging time.

Some observations on the oxidation of FeCrAl alloys in moderately higher temperature range covering about 723–973 K are briefly noted. In one study on the oxidation of Kanthal AF in dry oxygen, in the temperature range of 773 K–1173, the formation of (metastable) Al₂O₃ was reported after isothermally oxidizing at 973 K for 168 h (Josefsson et al. 2005). Results of isothermal oxidation at 873 K in dry oxygen and in wet oxygen, with water vapor 10% by volume, on FeCrAl RE (Kanthal AF) alloy were reported (Canovic et al. 2010); while some accelerated oxidation due to water vapor was noted at higher temperature of 1073 K, no effect of water vapor was observed at 873 K. It was observed that the thin oxide formed at 873 K consisted of an outer (Fe, Cr)₂O₃ corundum type oxide containing some Al and an inner layer of most likely amorphous Al-rich oxide, and the composition of the oxide was unaffected by the water vapor. Also, in another study (Kvernes et al. 1977) no effect was observed of water vapor on the isothermal oxidation of FeCrAl alloys containing 4 wt% Al, in the range of 953 K–1253 K.

It is apparent from the above brief summary of observations that in dry or humid oxygen atmosphere the FeCrAl alloys show similar kinetics and oxide formations below about 900 K, with iron-chrome based spinel type outer oxide layer and an aluminum-rich inner layer likely amorphous and/or metastable alumina type (Engkvist et al. 2009). In addition, majority of FeCrAl alloys of interest incorporate one or more rare-earth elements (such as yttrium, hafnium and cerium), their influence on the oxide characteristics may also be noted—referring to the review (Hou 2011) where the relabeling of these elements as RE is used—that their effect is to enhance the selective oxidation of aluminum, to slow down the rate of oxide growth, to reduce the outward diffusion of trivalent aluminum cations, and to improve the adhesion of an alumina to the base alloy.

The above empirical observations and noted results on oxide formations under non-aqueous conditions lend support to the notion (Rebak 2020b) that, under normal LWR operation, the oxygen partial pressure in the fuel cavity in equilibrium with the thermodynamic dissociation of urania may be enough to allow for a layer of alumina to form on the inside surface of the FeCrAl cladding wall (Figure 4).

Thus, it is reasonable to expect that the oxides formed on both the fuel facing and the coolant exposed sides of the cladding tube will provide some resistance to tritium permeation through the cladding which would lower the effective tritium permeability compared to that of the clean, unoxidized FeCrAl alloy condition. As noted earlier, as in the case of commonly used austenitic stainless steels in the LWR aqueous environments, the ferritic stainless steels, including the FeCrAl variety, also have a low corrosion rate under normal coolant conditions that is attributed to the formation and maintenance of a protective/passive chromium oxide film on the surface of these steels (Rebak 2018).

4.1.4 Hydrogen permeability in (Al, Cr, Fe) oxides

From the point of view of more realistic modeling and quantitative assessment of tritium permeation through the expected oxide-covered fuel cladding or other components it is useful, if not essential, to characterize the hydrogen transport parameters of the oxides known or expected to form in the material–environment systems of interest. There is only limited data on diffusivity of hydrogen in oxides of interest and even less data on permeability of the oxides (for definition of the terms see Garud et al. (2022)). Since in many ways the oxide layer(s)—whether grown naturally or purposely, and in-situ, in service, or externally—may be viewed as a form of coating that affects the hydrogen diffusion and permeation response of the unoxidized substrate, some data on oxidized coating are included at the end of this section for illustrating the oxide response, although coatings are otherwise mostly excluded from this review.

4.1.4.1 Diffusion of hydrogen in alumina (Al₂O₃)

Data on tritium permeability of variety of alumina products were reported by Roberts et al. (1979), Serra et al. (2005), and Katayama et al. (2015). The measurements by Roberts et al. (1979) dealt with very high temperature response— between 1473 K and 1723 K and partial pressure of hydrogen between 0.002 MPa and 0.05 MPa—on commercial, sintered tubes described as gas-tight. The reported pressure was for hydrogen that was mixed with tritium (as tracer) and helium, although the actual permeation rate was measured for the (tracer) tritium which was then converted using the isotope factor (Garud et al. 2022) for hydrogen. The permeation was found to be consistent with diffusion control process with estimated pressure exponent of 0.43, suggesting diffusion in atomic form of hydrogen within the oxide. Extreme high activation energy of 318.2 kJ/mol for tritium permeation was reported—perhaps a reflection of very low observed solubility and/or the high-density and high-purity of the gas-tight sintered alumina tubing used in this work, or a different hydrogen source.

Hydrogen transport and solubility parameters including permeability data were reported by Serra et al. (2005) also on commercial high-purity, high-density alumina tubes in the temperature range 1273–1673 K with 0.01–0.1 MPa partial pressures. For these conditions they also observed the square-root pressure dependence of permeation indicating diffusion limited transport through the oxide.

Katayama et al. (2015) reported testing on the permeability, diffusivity, and solubility of hydrogen in alumina tubes in a moderately high but narrow temperature range of 973 K–1073 K, although the graphical, Arrhenius plots showed result down to 673 K. The hydrogen partial pressure for tests was also in a moderately high but narrow range of 0.02–0.1 MPa and similar to above work (Serra et al. 2005) the square-root dependence indicative of atomic diffusional transport in the aluminum oxide was noted.

The permeability results from the above works on alumina tubes are summarized and compared in Figure 9. For comparison, the reported permeability data or parameters were adjusted for isotope effect and for the half-power pressure exponent and extrapolated in the temperature range. The apparent spread between the permeation plots for alumina from these investigations should be viewed also considering the limited data and in much different temperature ranges covered in the separate works.

Figure 9:

(Normalized) tritium permeability response of alumina oxides compared with oxidized/unoxidized FeCrAl and austenitic steels responses.

The overall agreement in these permeation data may be considered reasonable if one were to fit the data with common activation energy as is shown in Figure 9 with the following estimated parameters representing the tritium permeability of commercial purity sintered alumina:

• Central trend (interim estimate): P₀ = 4 × 10⁻⁷ mol/[m s √MPa] and Q_p = 114.5 kJ/mol.

In any case, the oxide permeability is many orders of magnitude lower than in the clean or unoxidized alloys and the oxide activation energy is significantly higher, as can be seen from Figure 9 that includes the best estimate curves from Figure 8 for clean as well as oxidized FeCrAl alloys and for clean austenitic steels. The narrow range of conditions and large scatter in permeability response of the oxides are deemed inadequate to give the bounding permeability parameters in this case of oxides.

Aside from the above hydrogen isotope permeation data in alumina there are a few more reports that deal with the diffusivity in oxides which are briefly summarized below for reference. Fowler et al. (1976, 1977 reported the tritium diffusivity response by out-gassing from several oxides. The diffusion coefficients in these oxides are much smaller than in the base metals, and with relatively much higher activation energies. Of interest to note is the suggestion that the observed high activation energies are indicative of chemical bonding of hydrogen within the oxides (Fowler et al. 1976). Also, the observed significant difference in diffusivity of Lucalox — alumina containing 0.2% MgO — compared to the sintered alumina, containing less than 50 ppm MgO, was attributed to the oxide impurity of Lucalox. The out-gassing of tritium in alumina and yttria was apparently diffusion controlled, but the form of diffusing tritium was unknown.

Roberts et al. (1979) whose permeation work was noted above also observed the diffusion of tritium in sintered alumina to be consistent with atomic diffusion process where the diffusant was dissolved, and the grain-boundaries and small closed pores within the oxide appeared to be not significant contributors to the overall permeation rate process.

Belonoshko et al. (2004) gave results of first-principles calculations of hydrogen diffusivity down to about 523 K in solid alumina oxide, also showing reasonably good agreement with the above experimental data on diffusivity (Fowler et al. 1977; Roberts et al. 1979; Serra et al. 2005). The derived activation energy for diffusion was 119.6 kJ/mol which is comparable to the above noted values for permeation that include the heat of solution, except for the very low activation energy of 33.5 kJ/mol reported by Katayama et al. (2015). Belonoshko et al. (2004) also gave results for liquid or amorphous alumina which showed higher diffusivity and smaller activation energy of about 87.8 kJ/mol. It was concluded that the hydrogen mobility in corrosion (oxide) scales was likely due to the diffusion of neutral hydrogen through the bulk (solid) alumina.

Additional first-principles calculations were reported by Pan et al. (2021a) and the calculated hydrogen diffusivities of atomic hydrogen and molecular type hydrogen in pure alumina were compared with the above experimental/analytical values. The estimated activation energy for the molecular-type hydrogen diffusion was 219 kJ/mol, more than double that estimated for the atomic hydrogen, 97.5 kJ/mol. The resulting estimates of diffusivity for these two forms of hydrogen essentially covered the above noted spread in the data. It was noted that the heat of solution and migration barrier of the molecular hydrogen are so high that dissolution and diffusion of the H₂ molecule in α-Al₂O₃ are very difficult and thus correspond to the lowest diffusivity, while at the same time requiring very high temperatures for these processes to occur. The lower estimated diffusivity for molecular type hydrogen seems to be generally close to experimental data (Fowler et al. 1976, 1977; Roberts et al. 1979), although some data (Serra et al. 2005) show higher diffusivity closer to the first-principles estimate for the atomic hydrogen, which is likely attributable to greater intrinsic or inevitable forms of defects in the oxide.

The above review of the first-principles calculations of diffusivity appear to support the form of diffusing species within the oxide, at least in alumina, to be either atomic hydrogen or molecular hydrogen depending on temperature range and the impurity/defect levels of the oxide.

4.1.4.2 Diffusion of hydrogen in chromia (Cr₂O₃) and iron oxides/spinels

The role of hydrogen permeation characteristics of oxides on iron alloys containing chromium and aluminum as main alloying elements was made more explicit by examining the (changes in) structure and composition of oxides and relating these to the (changes in) permeation rates in the temperature range of 573 K–973 K; though the hydrogen pressure used was high, about 0.1 MPa (Kripyakevich et al. 1971b). The oxidation treatment was carried out for 4, 10, and 14 h at 1073 K in dry, non-CO₂ air at 0.1 MPa, on the iron alloys with varying compositions of chromium and aluminum. From the thermal stability and protection points of view of the formed oxides, which were later subjected to the hydrogen atmosphere, it was observed that greater than 12% chromium was preferred, while the addition of aluminum also increased the oxide adherence and enhanced the formation of spinel oxides above 2.2% aluminum. The oxide on iron alloys with ∼19% Cr and 2.2–3% Al was found to be a mix of FeCr₂O₄ + Cr₂O₃ + Al₂O₃. It was concluded that the oxide structure played a significant role in the observed permeation response with the high density spinels providing much greater resistance to permeability than ferric oxides. It was also observed that although the alloys had substantially differing chemical composition and permeabilities in their initial unoxidized state, the permeabilities approached closer together after oxidation as the spinel oxide films appeared (Kripyakevich et al. 1971b).

Hydrogen diffusion coefficients of iron and chromium oxides on Type 302, Type 347 stainless steels, pure iron, and pure chromium were estimated mainly as a function of oxide thickness (Piggott and Siarkowski 1972) demonstrating that thin (∼9–100 nm) oxide films of α-Fe₂O₃ on the steels provided an excellent barrier. The estimated oxide diffusion coefficients were in the range of 1.8–66 × 10⁻²¹ m²/s compared to ∼1 × 10⁻¹² m²/s for the steel substrata. Films of Cr₂O₃ on pure chromium and α-Fe₂O₃ on pure iron were reported to have the hydrogen diffusion coefficients of ∼9.2 × 10⁻²⁰ m²/s and 1 × 10⁻²² m²/s, respectively. It was noted that the chemical composition of the oxides on steel samples changed, increasing in chromia with increase in oxide thickness, and the diffusion coefficient increased significantly beyond ∼100 nm. The latter was suspected to be due to cracking or loss of integrity of the thicker oxides. The higher diffusivity of hydrogen in Cr₂O₃ was speculated to be due to its slightly larger interstitial spacing than that in α-Fe₂O₃, but more likely related to difference in effects of impurities on the lattice of the two oxides.

Two items of interest to note from the work on Croloy at extreme low pressure (Renner and Raue 1979) that was reviewed separately (Garud et al. 2022):

1. Iron oxides (Fe₂O₃ and Fe₃O₄) and their integrity/defects were primary determining factors for the observed tritium permeation response on three different tube samples, tested at extremely low partial pressures and temperature range of 673 K–773 K. All samples showed reduction in the permeability by factor of ∼150 compared to the clean condition of the Croloy. The reduction was due to protective Fe₃O₄ oxide, ∼5 μm thick.

2. One sample was 6 times thicker than the other two, but showed very similar results, implying that the permeability was determined by the oxide properties and was essentially independent of the metal thickness (Renner and Raue 1979). The latter has obvious practical implications as well.

Permeability of hydrogen through layered iron oxide on two ferritic steels (one mild steel and the other 2¼Cr-1Mo) was reported by Tomlinson and Cory (1989) at one temperature (774 K) and the estimated values were in the range of 0.7 × 10⁻¹⁰ mol/[m s √MPa] to 2.6 × 10⁻¹⁰ mol/[m s √MPa]; this is significantly lower than all the unoxidized steel groups examined (Garud et al. 2022), but four orders of magnitude higher than that of pure alumina (see Figure 9), although the duplex iron oxide was observed to be porous and ∼15 μm thick.

Hydrogen permeability of iron oxides scales, mainly Fe₃O₄ at 823 K and FeO at 1073 K, with some Fe₂O₃ formed on iron, was reported to be one to two order smaller than that of iron and with significantly higher activation energy, at least 96.2 kJ/mol versus 34.3 kJ/mol for the iron (Ueda and Maruyama 2005). Ueda et al. also compared the permeability data of Cr₂O₃ on Fe-16Cr Type 430 stainless steel from Kurokawa et al. (2004) indicating about two–three orders of magnitude lower permeability than their data on iron oxides.

Chromium containing oxide scales, mainly a thin inner layer, about 0.68 μm, of dense chromia (α-Cr₂O₃) and a small amount of (Fe/Mn/Cr) spinels in the outer layer, were reported to reduce the deuterium permeability of Type 316 stainless steel by a factor of ∼30–80 at temperatures from 973 K to 873 K (Yu et al. 2013); the oxide was grown in 6 h at 1100 K with oxygen partial pressure of ∼3 × 10⁻¹⁹ Pa. Although permeability of chromia by itself was not assessed, the diffusion/permeation was dominated by the chromia so that the reported permeability response gives an approximate estimate that still shows high activation energy of at least 130.8 kJ/mol for deuterium permeation in chromia.

For the case of layered double oxides of Al₂O₃ and Cr₂O₃, based on their first-principles calculations Zhang et al. (2017) suggested that the diffusion rate of hydrogen isotopes is likely governed by the interface region between the two oxides as it has the highest estimated activation energy (255 kJ/mol) among the potential diffusion processes, in ideal case, compared to that of either bulk oxide.

The above illustrative summary of hydrogen isotope permeability through the (Fe, Cr, Al) oxides confirms their high activation energies and low permeabilities; the permeability resistance increases from iron oxide to chromia to alumina. The permeability of oxides shows more variability than in the case clean/unoxidized (Fe, Cr, Al) alloys reviewed in earlier sections.

4.1.4.3 Diffusion of hydrogen through (Al, Cr) oxide coatings

As noted at the beginning of this section, there is limited data on diffusivity/permeability of hydrogen in oxides of interest; however, in many ways the oxide layer(s) may be viewed as a form of coating that affects the hydrogen diffusion and permeation response of the unoxidized substrate. Therefore, some data on oxidized coatings are included briefly below for illustration and in support of the oxide response, although coatings, per se, are mostly excluded from this review.

Work on the development and effectiveness of oxide coated steels and aluminum containing alloys for limiting hydrogen permeation dates back even before 1950s and Flint (1951) may be consulted for details. As in the case of coatings, the actual barrier effect of oxides depends on their defect structure and (structural) integrity, and its viability depends on the stability of oxides under stress, temperature, and environmental conditions.

More recent extensive work of Forcey et al. (1989, 1991 also made observations/conclusions similar to the above noted work (Flint 1951). In this case, the investigations were carried out on aluminized austenitic and RAFM steels, including oxidization at 873 K in air – forming a thin (∼1 μm) surface oxide layer with Al₂O₃ which was found to offer an effective barrier to hydrogen permeation: reducing the permeation rate by three–four orders of magnitude, around 523 K–873 K. The aluminum oxide and its cohesion were found to determine the permeation reduction, not the thickness or aluminum content of the aluminizing layer.

Song et al. (1997) investigated the effect of various oxide coatings on hydrogen permeability of a simple aluminum alloy. Although the alloy is aluminum-based (∼99.3 %Al) the work is included here because of the coatings formed at low temperatures and which contained various phases of hydrated aluminum oxides. The permeation tests were done in the moderate temperature range of 673 K–773 K and hydrogen pressure of 0.01 MPa–0.1 MPa. The observed PRF relative to the permeability of naturally oxidized alloy ranged from about 4 to 20. They estimated that the permeation resistance offered by the oxide itself was about 100–2000 times that of the base alloy, with anodized films showing the lower resistance and oxide pre-filmed in water at 473 K showing the highest resistance to hydrogen permeation. These results suggested that (hydrated) aluminum oxide/hydroxide, crystalline or amorphous, offered appreciable resistance to hydrogen permeation at the test temperatures.

The effectiveness of alumina oxide coated on Eurofer steel was investigated by Levchuk et al. (2004) for deuterium permeation response at relatively high temperature of 973 K–1073 K, due to the measurement capability limitations. The oxide was deposited at 973 K forming a dense crystalline α-phase alumina, about 1 μm thick. The reduction in permeability due to the oxide, relative to that of the clean samples, was reported to be of the order of three magnitudes.

5 Permeation reduction factor (PRF)

The effect of an oxide layer on the permeation response of an underlying base alloy/substrate has often been discussed in terms of the PRF. PRF is usually defined as the ratio of permeation rate, or flux through a given area, obtained without the oxide layer to that obtained with the oxide layer, all other conditions remaining the same. The latter requirement is not always easy to fully satisfy and it makes difficult to compare results from different investigations on one-to-one basis. Nevertheless, for a given set of conditions, the PRF provides a relative measure of effectiveness of the presence of an oxide layer in acting as a permeation barrier.

Therefore, it is instructive to examine the PRF values from the literature reviewed in this work. These values of observed or estimated PRFs for various combinations of oxides and the underlying substrata are summarized in Table 2.

Generally the PRF for a specific oxide–substrate combination increases with decreasing temperature of the permeation. As such, the range-based values in Table 2 reflect this dependence for the temperature range where specified. For the reasons noted above and the fact that the permeation through an oxide layer is influenced by several factors, unlike that of the base alloy/substrate, the PRF values from row-to-row are comparable only on a relative basis. These factors include: characteristics of oxide layer under which permeation is measured, the stability, defect-density/structure, uniformity, and integrity of the oxide under conditions of permeation. Nevertheless, from these PRF values it seems reasonable to expect about two–three orders of magnitude reduction in tritium permeation, around 550 K–650 K temperatures for most steels of interest, depending on type/characteristics of the oxide layer, its uniformity, integrity and stability.

6 Additional factors/considerations

Three factors likely to have more direct but complex impact on the tritium permeation in any structural metal/alloy in general are: surface condition (other than oxide), neutron irradiation in LWRs, and (change in) microstructure of the alloy in service. Their impact on permeation in the case of FeCrAl alloys needs to be considered more fully than the cursory coverage given below.

7 Suggestions for future studies to fill gaps in the available information

Depending on the application specific needs, the following items are suggested for future developments:

1. One consideration of practical significance, apparent from the review and assessment of data thus far, that needs further work is the quality and dispersion/scatter (either intrinsic or due to method of determination) of the underlying set of data available for quantitative work. Large scatter in data poses some difficulty in interpretation and application, also suggesting that uncertainty quantification would be needed and that a quantitative model should enable/address such quantification.

2. Unlike in the case of clean/non-oxidized alloys, the hydrogen transport in the more practical case of oxide covered alloys is sensitive to oxide characteristics such as the composition, types and distribution/concentration of defects, formation history, impurities on/at the surface. As such, for better accuracy and reliability of estimating permeation response, more and specific data are needed on permeation of hydrogen isotopes, especially for naturally and/or in-situ oxidized non-ODS and ODS FeCrAl alloys under simulated service conditions.

3. Full and proper characterization of surface conditions and oxides, both before and after the permeation tests, would be useful for better interpretation and extension of data to service conditions.

4. A comprehensive computational model for tritium diffusion and permeation through oxidized FeCrAl alloy cladding needs to be setup and an initial quantification of expected upper and lower bounds need to be established; the permeability relations/considerations given in this review should provide a guide in this quantification.

5. The quantitative model should preferably include at least two-layers (an oxide layer on the cladding inside surface and the bulk alloy) for assessing steady-state response to hydrogen isotope transport. The model should be validated with the specific and well characterized data as noted above.

6. It is important to ascertain the maintainability of oxide layer (on either side of the cladding) during normal operation over the intended fuel cycle and under irradiation. That is, diffusion characteristics, defect structures, and permeation (barrier) performance of oxidized alloys need to be verified under simulated/limiting operating conditions or accelerated conditions.

7. More generally, for better use of data and consistency of its application, it is highly recommended to have uniform and internationally accepted protocols/standards for test procedures, techniques, and reporting. This will also help reduce the data scatter and uncertainty.

8. For the case of FeCrAl cladding conditions the model simplification of excluding hydrogen trapping effects is likely to be adequate and conservatively biased. Taking into account the effects of hydrogen trapping in the oxide/bulk alloy and/or its recombination on the downstream (exit) side of the cladding facing the aqueous environment may be considered in the permeation model as useful enhancements if additional consideration of estimating retention of tritium in the cladding is of value for its long term cycle.

9. Effects of surface condition, irradiation, and microstructure (including effects of possible Cr-rich α′ phase precipitation, and micro-alloying) were examined only briefly without definitive conclusions and may need more careful consideration and assessment to confirm the expectation of non-significance of these effects vis-à-vis hydrogen permeation during the anticipated thermal and radiation exposures; this becomes necessary as the fuel cycle or service life of cladding is extended.

8 Summary and conclusions

1. FeCrAl alloys are attractive materials for ATF cladding because of their good mechanical properties and excellent resistance to attack by steam.

2. During the operation of the power reactors, tritium may form in the fuel cavity and migrate through the cladding wall into the coolant water.

3. Since the FeCrAl alloys are ferritic (bcc) and none of the elements in FeCrAl alloys react with hydrogen forming stable hydrides, there was concern that the tritium concentration in the coolant may increase when FeCrAl are used in replacement for zirconium alloys.

4. Under normal operation conditions, it is expected that Al₂O₃ will form on the inner side of the cladding (fuel cavity) and that a Cr₂O₃ rich oxide would grow on the external surface (water side).

5. The main focus of this review was the application of FeCrAl alloys as a barrier to the tritium diffusion/permeation, particularly in the presence of expected surface oxides. The related data, observations, and literature were reviewed with critical assessment of the role of surface oxides in this permeation through ferritic steels containing chromium and/or aluminum as major alloying elements.

6. It is observed that in the case of oxidized alloys the permeation response is controlled or dominated by the oxides so that the relative performance or reduction in permeability of a broader class of alloys with similar oxide characteristics can be considered useful in this assessment.

7. It is reasonable to expect about two–three orders of magnitude reduction in tritium permeation, relative to the permeation response in clean, unoxidized condition for most FeCrAl steels of interest, around 550 K–650 K temperatures, depending on type/characteristics of the oxide layers on either side of the fuel cladding, their uniformity, integrity and stability.

8. The apparent activation energy for hydrogen diffusion through protective oxides is significantly higher than in the base Fe–Cr–Al alloy matrix.

9. The assessment also suggests that, in the case where an oxide layer is expected to act as a reliable and significant permeation barrier, increasing the base alloy (cladding) thickness is likely to be less of a sensitive contributor to further reducing the permeation response.

10. The actual barrier effect of oxides depends on their defect structure and integrity, and its viability depends on the stability of oxides under stress, temperature, and environmental conditions, implying that the permeability is determined primarily by the oxide properties and is less sensitive to the base metal characteristics.