Crystallization pressure and volume variation during rust development in marine and urban-continental environments: Critical factors influencing exfoliation

David M. Bastidas^a,*, Jacob Ress^a, Ulises Martin^a, Juan Bosch^a, Angel La Iglesia^b, José M. Bastidas^c

^a National Center for Education and Research on Corrosion and Materials Performance, NCERCAMP-UA, Dept. Chemical, Biomolecular, and Corrosion Engineering. The University of Akron, 302 E Buchtel Ave, Akron, OH 44325-3906, United States

^b Institute of Geosciences (IGEO), CSIC, UCM, José Antonio Novais 2, 28040 Madrid, Spain

^c National Centre for Metallurgical Research (CENIM, CSIC), Av. Gregorio del Amo 8, 28040 Madrid, Spain

(*Corresponding Author: dbastidas@uakron.edu)

1. INTRODUCTIONTOP

Three stages have been reported in the formation of rust on steel exposed to natural environments (Hoerlé et al., 2004). In the first stage, a thin oxyhydroxide layer is formed, in the second stage, this layer is transformed in near-neutral aqueous environments into transitory green rust compounds, and in the third stage, these green rusts evolve into a fragile brownish layer (Misawa et al., 1974; Graedel and Frankenthal, 1990).

The mechanical stress is a source of alterations of the rust, such as swelling, cracking, spalling, and exfoliation. For instance, exfoliation corrosion is a stratified form of surface stress corrosion that results in layer-by-layer detachment of a corroding metal (Morcillo et al., 2017). The mechanical breaking process due to rust development of the steel rebar in concrete exposed to the action of marine environments is a common phenomenon (Chatterji, 2005). The volume expansion caused by the rust layer on concrete surrounding to a steel rebar causes spalling and cracking (Sánchez-Deza et al., 2017).

Depending on the atmospheric conditions, particularly the environmental humidity and chloride salinity levels in a marine environment, rust eventually becomes detached from the steel surface due to exfoliation processes. It has been reported that the rust layers generated in chloride-rich environments present a heterogeneous and porous structure, containing five crystalline phases, lepidocrocite (γ-FeOOH), and goethite (α-FeOOH) at the outermost surface (lepidocrocite upper and goethite lower), and akaganeite (β-FeOOH), magnetite (Fe₃O₄), and maghemite (γ-Fe₂O₃) in the inner part of the rust adhering to the steel substrate (Morcillo et al., 2017). Lepidocrocite and akaganeite are formed in dry periods of atmospheric corrosion and both are reduced to magnetite during wet periods (Schwarz, 1972; Lair et al., 2006; Tanaka et al., 2014). Akaganeite is formed in marine environments with a high chloride deposition rate and high relative humidity (RH) and is the least dense of the five compounds due to its open structure containing chloride ions (Mackay, 1960; Remazeilles and Refait, 2008).

The aim of this paper is to study the influence of volume variation which may produce mechanical stress on the rust and thus give rise to rust breaking. The pressure generated during the crystallization of mineral phases of the rust, lepidocrocite (γ-FeOOH), goethite (α-FeOOH), akaganeite (β-FeOOH), magnetite (Fe₃O₄), and maghemite (γ-Fe₂O₃) is also analyzed as another parameter influencing the mechanical stress.

2. MATERIALS AND METHODSTOP

Rusted carbon steel specimens exposed to different natural environments were studied, 13 from marine environments, and 4 from urban-continental environments. At least, three to four specimens were tested from each place localization. MADRID 1, and MADRID 2 specimens were withdrawn, after 25 years of exposure in an atmospheric corrosion test station located at the terrace roof of CENIM, the urban-continental environment of the City of Madrid, Spain. GRA 1 specimen was collected from a warehouse of construction materials in the city of Granada, Spain, the steel rebars have never been used and were stored outdoor for at least 10 years. The climate of Granada is an urban-continental environment, and it is located ~60 km far from the Mediterranean Sea coast, so the influence of marine aerosols is not expected. The morphology of the specimens presented a thick porous rust layer. GRA specimen was also collected from the city of Granada, Spain, and belonging to a collapsed reinforced concrete building 50 years old.

Deteriorated steel gratings were collected from a marine environment, LAGUARDIA specimens, belonging to the infrastructure of the seaport of A Guarda, Spain, at the Atlantic Ocean coast, receiving directly the marine aerosols spray. The steel grating was 30 years old, abundant corrosion products were observed. TUERCA, NORAY, and ESCOTILLA specimens were collected from a scrapped merchant ship from the port of Alicante, Spain, at the Mediterranean Sea coast. The specimens were at least 25 years old. These specimens showed a morphology with a conglomerate of particles containing cavities showing swift stratification. CORUÑA 3 and CORUÑA 2 specimens were taken from a dock of the port of A Coruña, Spain, at the Atlantic Ocean coast, the specimens were 25 years old. The BARNA specimen was collected from a dock of the port of Barcelona, Spain, at the Mediterranean Sea coast, the specimens were 30 years old. ALM PO, and ALM CA specimens were collected from the posts that support a thick chain located at the seashore of Almuñecar, Spain, at the Mediterranean Sea coast. This promenade was built in 1997 and the specimens were 22 years old. ALM 4, and ALM 3 specimens were collected from the fish market reinforced concrete building also from Almuñecar, Spain, at the Mediterranean Sea coast, 32 years old. At 200 m from the seacoast, the waters were loaded with chlorides which accelerated corrosion on the reinforcements, large cracks and detachment of corroded rebars were observed. Finally, 153C and 253C specimens were withdrawn after two years exposure at Cabo Vilano, Spain, a marine environment at the Atlantic Ocean coast located at 150 and 730 m from the shoreline, respectively, with high RH, mild temperature, high precipitation, and strong marine winds. The rust layer showed a laminar morphology with abundant exfoliation of thin sheets.

Table 1 includes the chemical composition of the 17 carbon steel specimens utilized. In general, the elemental composition is very similar for all the specimens studied, with a carbon content ranging from 0.13–0.40 wt.%.

The morphology of the rusted specimens was analyzed using a JEOL JSM-6400 scanning electron microscope (SEM), equipped with a LINK System energy-dispersive X-ray (EDX) micro-analyzer, operating in secondary electron (SE) and backscattered electron (BSE) imaging modes.

The X-ray diffraction (XRD) study of powdered specimens was performed using a Phillips PW 1730 diffractometer provided with a graphite monochromator and using Cu K_α radiation. The XRD patterns were registered from 2 to 65 degrees (2θ). The quantitative analysis was carried out using the matrix-flushing method (Chung, 1974), which can calculate the amorphous phase content by addition of silicon powder as standard reference material (SRM), which does not interfere with the lines of the other crystalline phases (Degen et al., 2014). The program used for phase identification was Highscore Plus Suite from PANalytical with the ICDD database (ICDD, 2018). The Rietveld refinement method was also applied for quantitative phase analysis (Rietveld, 1969).

3. RESULTSTOP

Figure 1 shows XRD patterns of the 17 carbon steel specimens studied. These patterns highlight the small height and the scarcity of diffraction peaks and the low peak intensity/background intensity ratio. Both factors are typical of specimens with very low crystallinity. The crystalline phases identified were: akaganeite, goethite, lepidocrocite, and magnetite. Amorphous phases were also identified.

Table 2 includes the results of XRD analysis using the matrix-flushing method Chung (1974), and Rietveld (1969) method. Analysis by matrix-flushing method found that the amorphous phases are the main components in all specimens, with a composition ranging from 27 to 62 wt.%, with an average content value of 41 wt.%. The second-most abundant phase is magnetite with a range between 8–58 wt.% in composition, and an average value of 25 wt.%. The amount of goethite is in the range between 7–39 wt.%, with an average content value of 21 wt.%. Akaganeite and lepidocrocite are the minor components with average content values of 9.3 and 4 wt.%, respectively. It should be noted that akaganeite phase is not present on the specimens collected from urban-continental environment (MADRID 1, MADRID 2 and GRA 1), except for the specimen labelled GRA which presented an akaganeite content of 12 wt.%. In the same way, ESCOTILLA and CORUÑA 3 specimens collected from marine environments do not contain akaganeite.

Figure 2 shows XRD patterns and Rietveld analysis for the rust layer formed on carbon steel specimen 153C after exposure to marine environment (150 m from the shoreline) for 2 years. The sharp Si peaks used as standard reference can be observed on a complex diagram corresponding to the mixture of akaganeite (ak), goethite (go), lepidocrocite (le), and magnetite (ma). Figure 3 shows XRD patterns and Rietveld analysis for the rust layer formed on carbon steel specimen 253C, after exposure to marine environment for (730 m from the shoreline) 2 years. Quantitative analysis of crystalline and amorphous phases measured by XRD using the Rietveld refinement method (Rietveld, 1969) is shown in Table 2. The phase contents determined by Rietveld refinement method were 32.6 and 46.2 wt.% amorphous, 21.0 and 23.1 wt.% magnetite, 20.5 and 27.2 wt.% goethite, 8.1 and 14.1 wt.% akaganeite, and finally 3.0 and 4.2 wt.% lepidocrocite. Similar results were obtained by both Rietveld and matrix-flushing quantitative analysis methods. Weighted profile values (R_wp) of 6.17% and 6.5% were obtained for 153C and 253C specimens, respectively. The silicon peak comes from the SRM material used. Table 2 indicates that, (i) in all the specimens the content of amorphous phases is high (27 to 62 wt.%), which is independent of the type of environment, marine or urban-continental; (ii) there is not a relationship between the percentage of amorphous phases and the crystalline phases; and (iii) the akaganeite phase appears on carbon steel specimens exposed to marine environments, except for the GRA specimen (urban-continental environment).

Figure 4 shows a SEM micrograph of carbon steel (153C specimen) of exfoliated flakes of ~1200 μm thickness and an intense fracture producing cleavage facets of ~300 μm thickness. A typical dissolution process can be observed (see the arrows). In general, the texture of the rust is typical of an amorphous phase. EDX microanalysis of the spots showed an atomic O/Fe ratio between 1.05 and 2.23 and an atomic Cl/Fe ratio between 0.03 and 0.17.

Figure 5 shows a SEM micrograph of carbon steel (153C specimen), a rust with great porosity can be observed. The rust layer shows the growth of lepidocrocite and/or goethite mineral phases of ~3 μm particle size (right), and amorphous phases (left) separated by a crack, along with new growing of lepidocrocite and/or goethite phases (see the arrow). The atomic O/Fe ratio measured using EDX microanalysis varied from 1.92 to 2.15, and the atomic Cl/Fe ratio varied from 0.00 to 0.07.

Figure 6 shows a SEM micrograph of carbon steel (253C specimen) of the inner surface of exfoliated flakes. Different arrays of lepidocrocite and/or goethite mineral phases of ~4 μm in size can be observed. Many cracks, pores and particles containing a chloride atomic percentage of ~5% can be seen, which may be attributed to the formation of akaganeite (see the arrows).

Figure 7 shows a SEM micrograph of carbon steel (253C specimen) presenting a laminar array of edge-edge joined lepidocrocite and/or goethite mineral phases with a size of up to ~20 μm, giving rise to high porosity. EDX microanalysis showed an atomic O/Fe ratio from 2.20 to 2.25, and the atomic Cl/Fe ratio varied from 0.02 to 0.07.

Figure 8 shows a SEM micrograph of carbon steel (CORUÑA 3 specimen), the formation of goethite-lepidocrocite crystals on a matrix of amorphous phases can be observed, as indicated by its brotoidal morphology. Large cracks can be observed, possibly due to the volume variation of the amorphous phases. EDX microanalysis showed an atomic Cl/Fe ratio between 0 and 0.08.

Figure 9 shows a SEM micrograph of carbon steel (GRA specimen) in which the presence of a rust layer between the steel bar (lower left area) and the concrete layer (upper right area) can be observed. The rust layer has a thickness of 285 μm, and it is very heterogeneous. EDX microanalysis showed a variation of O/Fe ratio from 2.00 to 2.80, and Cl/Fe ratio from 0 to 0.12, depending on the area where the measurement was made. An extensive network of cracks can be observed.

Figure 10 shows a SEM micrograph of carbon steel (ALM 3 specimen). A rounded particle of 18 μm in diameter can be observed, gelatinous in appearance, deposited on a base of the same material. A large number of shrinkage cracks appear. Rounded surfaces and their gelatinous appearance indicate a newly formed material. Small crystals appear on the gelatinous masses, some with cubic morphology, their chemical analysis proves to be sodium chloride. EDX microanalysis showed that O/Fe ratio measured in different areas varied between 2.10 and 2.40.

3.1. Mechanical stress generated during rust developmentTOP

The rust formed on steel in an atmospheric corrosion process generates stress, both on the rust and on the steel substrate (Chatterji, 2005). As the density of the rust is lower than the steel, the stress generated is compressive on the rust and tensile (σ) on the steel. Assuming that the growth in volume of goethite, as an example, with respect to the steel substrate is isotropic, the steel and goethite molar volume (V_m) (cm³ mol⁻¹) can be calculated: V_m(steel)=x³, and V_m(goethite)=y³, where x=1.990 cm, and y=2.751 cm are the conventional lattice parameters. Thus, V_m(steel)=7.88 cm³ mol⁻¹, and V_m(goethite)=20.82 cm³ mol⁻¹, see Table 3. The V_m of a crystalline phase is defined as the ratio between the molar mass and its density.

Table 3. Molar volume (V_m) (cm³ mol⁻¹), molar volume expansion ratio coefficient (α) (α=V_m(oxide)/V_m(Fe)), and standard free energy of formation () (kJ mol⁻¹), for the studied crystalline phases (Robie et al., 1979; Navrotsky et al., 2008).

Crystalline Phase	Vm, cm3 mol−1	α	, kJ mol−1
Iron	7.09	-	-
Carbon Steel	7.88	-	-
Akaganeite (β-FeOOH)	23.32	3.29	−481.70
Goethite (α-FeOOH)	20.82	2.94	−490.60
Lepidocrocite (γ-FeOOH)	22.42	3.16	−482.70
Maghemite (γ-Fe2O3)	32.89	2.32	−731.40
Magnetite (Fe3O4)	44.52	2.09	−1012.57

In order to calculate the mechanical stress (σ), the strain (ε) is first calculated, taking into account the plasticity criterion of Treska: σ₁–σ₃ = σ_y/2, where σ_y is the yield strength of steel (Chen, 1975). The ε value for carbon steel can be calculated in any direction using its conventional parameters: ε = [(2.75−1.99)/1.99]×100=38%. This ε value of 38% overcomes the breakage strain of a low-alloyed construction steel, such as the low carbon steel C15 (BS EN 10277-2, 2008), or the carbon steel C25 (BS EN 10250-2, 2000). The conventional ε value until breakage of C25 carbon steel is approximately of 24%, and the yield strength is of 400 MPa. This means that if the adhesion between goethite and the steel substrate is supposed to be perfect, the stresses to which they are subjected, both steel and goethite would be quite large and would also affect the steel producing small cracks. However, given that the resistance of the goethite is much lower than steel, since it is a highly oxidized crystalline compound, goethite would break as is formed.

3.2. Volume variation caused by structural transformationsTOP

Structural transformations between different rust phases have been reported in the literature (Schwertmann and Taylor, 1972; Tanaka et al., 2014), they induce variations in the V_m of a crystalline phase in terms of expansion or contraction. For instance, the transformation of magnetite into maghemite produces expansion, while the transformations of lepidocrocite into goethite, and akaganeite into magnetite or maghemite generate a contraction. In all cases, these transformations lead to the generation of stresses in the rust layer (Chatterji, 2005).

Table 3 summarizes the V_m of the five crystalline phases, iron and steel analyzed in the present work, calculated using data from Navrotsky et al. (2008), and Robie et al. (1979). Considering Table 3 results, it is simple to estimate the variation in volume (ΔV_m) caused by the transformation of a given crystalline phase, for instance in the transformation of magnetite into lepidocrocite (Fe₃O₄+3/2H₂O+1/4O₂→3(γ-FeOOH)). ΔV_m = 3V_m(γ-FeOOH) −V_m(Fe3O4); ΔV_m=(3×22.42)−44.52=22.74 cm³ mol⁻¹, ΔV_m is positive, i.e. expansion. Only solid phases should be considered to calculate the ΔV_m.

The percentage in ΔV_m (%) is calculated using the expression: ΔV_m (%) = [(V_mi/V_mf)−1]×100, where V_mf and V_mi are the molar volumes of the final and initial phases, respectively.

Table 4 sets out the possible phase transformations between the five crystalline phases present in the rust, the ΔV_m caused by the transformation, and the ΔV_m (%). Table 4 indicates that the highest ΔV_m (%) was yielded by the transformation of magnetite into any of the three oxyhydroxide phases considered (lepidocrocite, akaganeite, and goethite), with a ΔV_m (%) (expansion) of between 40.30% and 57.14%. Obviously, if the transformation is reversed (i.e. the three oxyhydroxide phases into magnetite), a similar ΔV_m (%) is produced, but in this case as a contraction. The transformation of magnetite into maghemite causes a ΔV_m (%) of 10.82% (expansion). With regard to the other crystalline phases, the transformation of maghemite into lepidocrocite, akaganeite or goethite causes a ΔV_m (%) (expansion) of 26.60% to 41.82%. The transformation of goethite into akaganeite or lepidocrocite produces a ΔV_m of 12.01% or 7.68% (expansion), respectively. Finally, the transformation of lepidocrocite into akaganeite causes a small ΔV_m of 4.01% (expansion). These changes in the V_m account for the mechanical stresses produced when weathering conditions vary (Suda et al., 1993).

The V_m parameter can also be used to calculate the molar volume expansion ratio coefficient (α) of an oxide to iron metal dissolved in the corrosion process defined as: α=V_m(oxide)/V_m(Fe).

Table 3 shows the calculated α parameter. It is found that akaganeite presents the highest value of α (3.29), followed by lepidocrocite (3.16), goethite (2.94), maghemite (2.32), and magnetite (2.09). These results may be interpreted in terms of the capacity to generate stresses within the rust layers producing an irreversible damage of the rust layer in the following order from high lo low α value: akaganeite > lepidocrocite > goethite > maghemite > magnetite.

3.3. Crystallization pressure generated during rust developmentTOP

It is known that the crystallization process of a solid phase from a solution in a confined space generates a pressure known as crystallization pressure, which can originate stress in the confined space (Steiger, 2005; Espinosa et al., 2008; Sánchez-Deza et al., 2017). Crystallization pressure is generated during the growth of a crystalline phase, and not in the formation of an amorphous phase (Sánchez-Deza et al., 2017).

According to Correns (1949), the pressure caused by the crystallization of a solid compound can be expressed by the super-saturation of a solution using the equation: Δp=p_c−p_e=(RT/V_m)lnS, where Δp is the variation of crystallization pressure (Pa), p_c is the pressure on the growing crystal, p_e is the environmental pressure, R is the molar gas constant (8.3145 MPa cm³ K⁻¹ mol⁻¹), T is the temperature (K), V_m was defined above as the molar volume of the crystalline phase, and S is the super-saturation, which is defined as the ratio of the solute concentration (c) to the saturated solution concentration (c₀) (S=c/c₀). In accordance with Neugebauer (1973), in the current study the activity of the solutions was used instead of their concentrations. In this way, S=a/a₀ where a is the activity of the solute for the saturated solution, and a₀ is the activity of the solute for the super-saturated solution.

The amorphous phases are probably the most soluble phases in a rust and can lead to a super-saturated solution with respect to the crystalline phases (higher density, higher melting point and largest crystalline lattice energy). The solubility of the rust compounds is unknown and cannot be calculated because of the lack of thermodynamic data in literature. The only thermodynamic data available for similar compounds are for ferric hydroxide (Fe(OH)_3(pp)) and ferrous hydroxide precipitates (Fe(OH)_2(pp)) (Wagman et al., 1982). If the activity value for the dissolution of ferric oxide is similar to the amorphous phase, it makes possible to calculate the chemical equilibrium constants (K_eq) of the goethite, lepidocrocite, akaganeite, magnetite, maghemite crystalline phases, Eqs. (1–5), and ferric hydroxide precipitates, Eq. 6. The log K_eq values included in Table 5 (Sanchez-Deza et al., 2017), were obtained by calculation of the standard free energy of reaction (), using the standard free energy of formation () from the literature, see Table 3 (Wagman et al., 1982; Chatterji, 2005; BS EN 10277-2, 2008).

If the a_Fe³⁺ values from Eq. (1) to Eq. (5) (see Table 5) are considered as the saturated solution, a_Fe³⁺ value from Eq. (6) (see Table 5) as the super-saturated solution, and the V_m values (see Table 3), it is possible to estimate the Δp caused by the crystallization process using the Correns’ equation. Table 5 includes the calculated values of Δp for the five crystalline phases studied. The Δp of goethite presents the highest value (374.99 MPa), followed by lepidocrocite (260.14 MPa), akaganeite (239.38 MPa), maghemite (189.22 MPa), and magnetite (26.45 MPa), ordered from high to low: goethite > lepidocrocite > akaganeite > maghemite > magnetite.

4. DISCUSSIONTOP

The model based on the α parameter is useful to explain the mechanical stress generated on the rust layer (Morcillo et al., 2017). However, the exclusive use of α for the particular case of akaganeite as the main parameter may be risky, because it is only 3.29 times higher than iron (see Table 3), and it is closed to the lepidocrocite (3.16), and goethite (2.94) values. Thus, even though being the α parameter very important, it is necessary to discuss the influence of other parameters that generate mechanical stress, such as the Δp produced during the growth of a crystalline phase, and the mechanical stress generated during the formation of a rust layer.

As described above, if the adhesion between goethite and the steel substrate is perfect, the steel substrate and goethite are subjected to high mechanical stress. The ε values for conventional carbon steel and C25 steel are 38% and 24%, respectively. From a mechanical point of view, the disruptive effect of crystalline phases on the rust layers is due to the generation of a mechanical stress that is greater than the cohesive forces of the rust (Tomlison, 1927; Chatterji, 2005).

In all the carbon steel specimens studied in this work, important quantities of amorphous phases have been identified along with smaller quantities of crystalline phases. In this way, rust specimens should be considered as a heterogeneous system formed by different phases whose mechanical and chemical properties depend on the content of each of the constituents, with the amorphous phases having a predominant role. The use of XRD technique aided by the matrix-flushing method (Chung, 1974), and the Rietveld refinement quantitative analysis method (Rietveld, 1969), allowed to highlight these high amorphous contents.

The current study also considered α and Δp parameters to assess the breaking of the rust. The crystalline phase transformations are favored by the atmospheric conditions found in marine environments, 80% RH or higher, a high level of chloride ions, and alternating wet/dry cycles. Especially in wet periods, the amorphous iron oxyhydroxides may be solubilized to start a dissolution-crystallization process which leads to the transformation of the amorphous phases into crystalline materials. In dry periods these processes cannot be taken place.

The hypothesis is that akaganeite, goethite, and lepidocrocite polymorphs show a correlation between the calculated Δp and their α parameter, in the sense that a high Δp value for a crystalline phase is accompanied by a high α value. Figure 11 shows the relationship between the calculated Dp value and the α parameter due to structural transformations for the Alicante, Barcelona, and Cabo Vilano specimens. It can be observed that Dp and α parameters are linearly related: Δp=189α−291, with a correlation coefficient of R²=0.7029. This linear relationship would support the hypothesis that α parameter may control the stress generated on the rust.

The extraordinarily high value of Δp calculated using the Correns’ equation, between 374.99 and 26.45 MPa (see Table 5), indicates that the crystallization process can easily crack the rust layer. For comparative purposes, it is interesting to note that Δp values obtained in Table 5 greatly exceed the uniaxial tensile strength (σ) of concrete, which is of the order of 0.2−10 MPa, leading to failure of the concrete cover surrounding a rust layer generated on a steel reinforcement (Noiriel et al., 2010; Castorena-González et al., 2019).

The presence of akaganeite at different depths of the rust layer, exerting a high crystallization pressure (239.38 MPa), and with a high α value (3.29), may produce a mechanical fracture along the inner rust layer at different depths (see Fig. 4). Thus, cracks appearing on the stratified rust bulk can be associated with changes in the compactness of the rust, caused by the presence of loose rust interlayers (Sánchez-Deza et al., 2018). The low compactness of the rust hinders the SEM analysis of chloride profile and the correlation with the formation of akaganeite to explain the exfoliation process on 153C and 253C specimens.

5. CONCLUSIONSTOP

ACKNOWLEDGEMENTSTOP

D.M. Bastidas, J. Ress, U. Martin and J. Bosch acknowledge funding from The University of Akron, Fellowship Program FRC-207367.

REFERENCESTOP