9.2 Carburization

Whether an alloy is likely to be carburized or decarburized depends on the carbon activity (a_C) in the environment and that of the alloy. The thermodynamic condition dictates either carburization, a_{C environment} > a_{C metal}, or decarburization, a_{C environment} < a_{C metal}. Thus one needs to know the carbon activities of both the environment and the alloy.

Carburization can proceed by one of the following reactions when the environment contains CH₄, CO, or H₂ and CO:

9.1

9.2

9.3

Assuming that carburization follows Reaction 9.1, the carbon activity in the environment can be calculated by

9.4

Similarly, if carburization follows Reaction 9.2, known as Boudouard reaction, the carbon activity of the environment can also be calculated:

9.5

When carburization follows Reaction 9.3, the carbon activity in the environment is

9.6

Data for carbon activities of commercial alloys at temperatures below 1200 °C are relatively limited. Natesan (1976) reported that a_C for 2.25Cr–1Mo steel is in the range of 1 × 10⁻¹ to 10⁻² from 550 to 750 °C. Natesan and Kassner (1973) reported the carbon activities of Fe–18Cr–8Ni alloys.

For carbon steels, carbon activity can be estimated by assuming that it is in equilibrium with cementite (Fe₃C):

9.7

9.8

Environments containing gas mixtures such as CO/CO₂ or H₂/H₂O are very significant in terms of carburization/oxidation phenomena. Let us, for example, consider the following reaction:

9.9

The variance of the equilibrium is equal to 2 in this case of condensed phases without miscibility since there are three independent components and three phases in equilibrium. At equilibrium, we have

9.10

where and are the mole fractions of hydrogen and water vapor in the gas mixture, respectively. Assuming that the activities of the condensed phases are unity and that the gases show ideal behavior, ΔG° is independent of the total pressure p, and the gas composition is a function of temperature only. Under these conditions, the equilibrium is considered as pseudo‐invariant.

The stability domains of the condensed phases are represented in (or ) versus f(T) plots (Chaudron diagram) or as versus 1/T. In effect, Eq. 9.10 can be written as

9.11

Assuming that ΔH° and ΔS° are independent of temperature, is a linear function of 1/T, as represented in Figure 9.1. The straight line marks the boundary between the stability domains of each oxide, i.e. M₂O_b and M₂O_b′.

If several oxides of the same metal can be formed, it is necessary to proceed in the same way as described in Chapter for the case of equilibrium with a simple gas.

The stability domains of iron and iron oxides, i.e. FeO, Fe₂O₃, and Fe₃O₄, as functions of gas composition and temperature are represented in Figures 9.2 and 9.3. Figure 9.2 corresponds to a Chaudron diagram and Figure 9.3 to the plot of or versus 1/T.

Note that both diagrams exhibit a triple point, T₀ = 830 K, O^′, and O^′′, respectively, in Figure 9.3, corresponding to a pseudo‐invariant system, insofar as the pressure is not a factor in the equilibrium since there are three independent components and four phases. Also note that B^′ and B^′′ in Figure 9.2 (intersection of, respectively, the curves (2) and (2′) and (3) and (3′)) correspond to the inversion temperature of the following equilibrium:

9.12

that is, T₁ = 1100 K. Moreover, for T < 1100 K, CO is more reducing than hydrogen, whereas the reverse applies for T > 1100 K.

A more general treatment of high temperature equilibria for oxidation, and carburization or decarburization of iron or steel, requires the consideration of the equilibria H₂O/H₂, CO₂/CO, and CH₄/H₂. Therefore, the reactions chiefly responsible for this corrosion at high temperature are

9.13

9.14

9.15

9.16

The data presented in Figures 9.4–9.6 show the variation with temperature of the equilibrium constant for each of the four Reactions 9.13–9.16.

The constants are given in terms of the partial pressure of each gas, that is, the total pressure multiplied by the volume fraction (not weight percentage) of the gas in question. For example, in a gas mixture containing 45% CO, 15% CO₂, and 40% N₂ at a total pressure of 1 atm, the partial pressure of CO is 0.45, and that of CO₂ is 0.15.

The curves presented in the several figures enable one to predict the way in which a mixture containing CO and CO₂ or H₂ and H₂O tends to react with iron or iron oxide. (They do not indicate the rate at which this tendency may be followed.) As an illustration, for the gas mixture described above, the pCO/pCO₂ ratio is 0.45/0.15 or 3. Figure 9.4 shows that this ratio is equal to the equilibrium constant at 1130 °C; hence at this temperature this mixture is inert to iron or iron oxide, at higher temperature it cannot reduce iron oxide and tends to oxidize iron, and at a lower temperature it tends to reduce iron oxide and cannot oxidize iron.

To ascertain whether this mixture tends to carburize or to decarburize steel at 800 °C, the value of the equilibrium constant for Reaction 9.3 is computed; it is p²CO/pCO₂ = (0.45)²/0.15 = 1.35. Figure 9.5 shows that at 800 °C this gas is decarburizing toward steel of any carbon content. A ratio of 6 instead of 1.35 would indicate (see Figure 9.5) that at 800 °C the gas mixture tends to decarburize steel containing less than 0.8% carbon and to carburize steel containing a higher concentration of carbon.

The use of these charts for mixtures of hydrogen and methane (Figure 9.6) is analogous to that for carbon monoxide and carbon dioxide.

Note that at high temperatures the oxide that first forms is FeO, and if the partial pressure of oxygen is low, this is the only oxide that forms. At somewhat higher oxygen pressures, a layer of Fe₃O₄ forms on the top of FeO, and at still higher pressures, a layer of Fe₂O₃ forms over Fe₃O₄.

Below about 600 °C, FeO is unstable and decomposes into iron and Fe₃O₄ so that high temperature scales on iron examined at room temperature show only iron, Fe₃O₄, and possibly Fe₂O₃.

Even though in cases where the gas mixture may not reach an equilibrium condition, it will be of great benefit to better understand the gas–metal reaction in terms of thermodynamic equilibrium in multicomponent gas environments. The thermodynamic equilibrium gaseous composition along with its thermodynamic potentials, such as carbon activity, a_C, oxygen potential, , hydrogen potential, , methane potential, , and other potentials, can be determined using a commercial software program such as HSC Chemistry for Windows (HSC 2006, Finland), ChemSage/FactSage (GTT 2015, Germany), and Gemini (Thermodata 2002, France).

The environment is also being characterized in terms of a_C, and ,, , etc., and in complex gas mixtures such as H₂–CH₄–H₂S, H₂–CO–H₂O–H₂S, as well as reported in the open literature (Barralis and Maeder 1997; Christ 1998; Coltters 1985; Forseth and Kofstad 1998; Grabke 1998a, b, Jakobi and Gommans 2003; Nishiyama et al. 2003; Rocabois et al. 1996; Yazawa 1979).

As reported above, carburization depends on the carbon activity in the environment and that of the alloy. The phenomenon refers to ingress of carbon into the material in the presence of carbonaceous gases such as CO, CO₂, CH_4, and other hydrocarbons. Carbon is transferred to the material surface, diffuses through it, and forms various carbides with the alloying elements. Usually, it happens at T > 800 °C and a_C < 1. Apart from carbon inward diffusion and carbide precipitation, MD is also observed for lower temperatures and greater a_C (a_C > 1). But, during metallic carburization, other effects may occur, namely, discontinuous precipitation, microcrack formation, and volume expansion, as will be discussed next.

The growth of discontinuously precipitated carbides was described by Cahn (1959) and Hillert (1969) during the 1950–1960s. Many solid‐state reactions such as eutectoid transformations like pearlite formation and discontinuous precipitation result in a lamellar structure. In the case of discontinuously precipitated carbide in austenite, Cr diffusion occurs in the austenite/austenite grain boundary as the boundary of the growing cell moves through the supersaturated austenite. In the alloy 602CA (alloy type NiFeCrAl), a fine substructure of carbides (probably M₇C₃) and a new Ni‐enriched, Cr‐depleted austenite were formed from the carbon‐saturated austenite. The carbon supersaturation decreases discontinuously as the carburized austenite/newly formed austenite cell boundary advances. The carburization thus constitutes the driving force for the phase transformation, which occurs by a discontinuous cooperative growth of carbide and Cr‐ and C‐depleted austenite. This type of reaction represents a fast precipitation process at lower temperatures where bulk diffusion is slow. It has been suggested that the force, actually pulling the cellular boundary of discontinuous precipitation, is the difference in atomic size, which gives rise to strain energy in the concentration gradient ahead of the growing new austenite grain. This mechanism was originally proposed by Sulonen (1994).

Two different morphologies of the discontinuously precipitated carbides were also detected in AISI 304L as well as in nanosized carbon filament formation during MD of stainless steel (Lin and Tsai 2003; Szakálos et al. 2002).

In the case of the alloy 602CA, the cellular growth rate of discontinuous precipitated carbides means that the advancement of the cellular boundary is governed by the rate at which a sufficient amount of Cr can move to the carbide phase via the cellular boundary.

Every Ni atom has a short diffusion distance between the carburized bulk (old austenite) and the newly formed low‐alloyed austenite, while Cr has orders of magnitude longer diffusion distances toward the nearest Cr carbide. This means that the average Ni atom has a very short dwell time, and, as a consequence, the average Cr fraction in the boundary could be much higher than the bulk fraction of 0.25. The cellular growth in this case is characterized by a small fraction of discontinuous precipitates, and both the new phases formed are close to equilibrium, i.e. the Cr‐diffusion process, and, subsequently, the Cr‐carbide formation, is almost accomplished.

It is possible to estimate the cellular growth rate based on diffusion limitations. It is assumed that the cellular boundary can move forward in steps of the boundary thickness and before it moves forward, all metal transport needed to reach equilibrium must be accomplished. Consider a segment of the cellular boundary, with S the lamellar spacing in the cellular structure and S/2 the characteristic diffusion distance.

Using an approximation of Fick’s first law, we can write

9.17

where Δm is the amount of Cr metal in the segment that diffuses and forms carbide Δx(S/2) is the diffusion length, Δt is the time for one segment of cellular growth ΔC_Cr (2X_Cr) is the Cr content in the cellular boundary, which may be much higher than the bulk content (X_Cr) (and the difference in Cr concentration toward the carbide may therefore be estimated as ΔC_Cr = 2X_Cr) A is the diffusion cross‐sectional area or A = δ per unit length ρ is the density of the diffused metal, i.e. Cr δ D_GB is the effective grain boundary diffusion coefficient of the Cr atoms with LAρ as a characteristic size of the structure, in this case L = δ, which is the cellular boundary thickness or segment thickness.

The growth increment or movement of the cellular boundary is set to the grain boundary thickness, δ, and thus the diffusion‐controlled velocity of the cellular boundary v₁ may be written as

9.18

Using the model for cellular growth proposed by Hillert (1969), it gives the maximum growth rate:

9.19

The constant 16 in Eq. 9.19 is directly comparable with the constant 8 in Eq. 9.18. This means that Hillert’s model predicts two times faster growth rate of the discontinuously precipitated carbide structure. This implies that the MD corrosion rate could be much higher if the atmosphere was more aggressive, i.e. with higher carbon activity. The cellular structure is thus of vital importance and facilitates the relatively fast MD kinetics at lower temperatures.

Carburization during MD conditions may result in low ductility and microcracks, which form flat metal fragments on the surface in different sizes up to 200 µm × 60 µm × 10 µm, where 10 µm represents the typical crack depth. Crack configuration suggests that the carburized surface metal developed compressive stresses, which could result in shear cracks parallel to the surface, favored by a ductility drop. Distribution of the microcracks all over the carburized surface, revealing the crack propagation path through the carbide network, shows that general surface carburization can be simultaneously involved with oxidation and microcrack formation, as revealed by cross sections of 602CA at 540 °C.

Carburization gives rise to compressive stresses due to volume expansion (Schnaas and Grabke 1978), but at higher temperatures, such as 650 °C, stress relaxation may take place fast enough to avoid crack formation. An additional effect may be that microcracks form during cooling because of different thermal expansion in the carburized layer and the bulk. If the uncarburized bulk “shrinks” somewhat more during cooling than the carburized surface layer, the result would be cracks parallel to the surface.

The microcrack formation during cooling may explain the surprisingly low metal attack resistance for 602CA during cyclic conditions as pointed out by Toh et al. (2003).

Several phases such as oxides and carbides, which will form in an AISI 304L steel during metal carburization/dusting conditions, give rise to a substantial volume expansion. This may contribute to the rather rapid degradation process that characterizes the metal aggression phenomenon. Some possible phases and calculated volume expansions, when formed in the steel matrix, are given in Table 9.1. These extended Pilling–Bedworth ratios (EPBR) for different phases are based on density changes for the metal during phase transformation. The phase stability conditions are calculated with Thermo‐Calc.

The EPBR are calculated by virtual volume expansion or density change of the metal component during phase transformation, i.e. if C, O, or N had zero mass:

9.20