Fracture Mechanical Properties of CO2 Corrosion Product Scales and Their Relation to Localized Corrosion

G. Schmitt Laboratory for Corrosion Protection Iserlohn Polytechnic Frauenstuhlweg 31 D-58644 Iserlohn Germany T. Gudde and E. Strobel-Effertz Institute for Technical Chemistry Aachen University of Technology Tempiergraben 55 D-52056 Aachen Germany ABSTRACT For the first time fracture mechanical properties of scales from wet corrosion were considered with respect to the initiation steps of flow induced localized corrosion (FILC) of steel under conditions of scale forming corrosion processes. Fracture mechanical data for iron carbonate scales from carbon dioxide corrosion, assessed by first principle considerations, were compared with own experimental data in a failure mode diagram which correlates the tendency of scale spalling with scale thickness related critical strains. The destruction and spalling of iron carbonate scales is primarily the effect of intri sic stresses. Extrinsic stresses like wall shear stresses in flowing media are generally too small to contribute much to the local destruction of scales. FILC is initiated at sites of local spalling if critical flow intensities prevent re-formation of protective scales. INTRODUCTION The work-over of oil and gas wells is generally not caused by uniform corrosion but by localized corrosion such as pitting and/or flow induced localized corrosion (FILC), very often also called erosion corrosion. Therefore, to decide on appropriate mitigation methods the understanding of the mechanisms of localized corrosion is indispensable. One mechanism which still needs considerable research is the mechanism of FILC, specifically the initiation of FILC. If hydrodynamic forces are responsible for the local destruction of corrosion product scales these forces must be in the same order of magnitude as those forces binding the scale and its particles onto the metal substrate. This raises the following questions: i) What is the magnitude of the hydrodynamic forces acting - in technical systems - on the scaled surface up to high flow intensities? ii) What are the mechanical, specifically fracture mechanical properties and stabilities of corrosion product scales? This paper addresses for the first time the question of fracture mechanical properties of scales from wet corrosion with special consideration of iron carbonate scales formed during CO2 corrosion of carbon steels. The data obtained by assessments based on

literature data and first principles are compared with own experimental data. The results furnish the basis of a new understanding of the initiation of FILC and the effect of inhibitors on its prevention during CO2 corrosion of steel. GENERAL CONSIDERATIONS Scales from wet corrosion are brittle and mechanical forces arising from various sources can lead to cracking or detachment (spalling) of the scale. If spalling occurs at the interface metal/scale, unprotected metal is exposed to the corrosion medium and flow effects could create FILC as outlined above. Forces causing cracking or spalling of corrosion product scales can be either intrinsic or extrinsic in origin. Intrinsic forces are stresses arising during scale growth due to the volume mismatch of the growing scale and the metal it replaces. Extrinsic stresses arise from strains transmitted to the scale from the metal substrate due to mechanical stressing or straining of the metal component or due to a mismatch in the thermal expansion coefficient of the metal and the scale. Extrinsic stresses could be also transmitted to the scale by hydrodynamic forces, e.g. wall shear stresses. Fracture mechanics can be employed to analyze the possible failure mechanisms. It has been shown for oxide scales' ' that the fracture occurs when elastic strain builds up in the scale until the strain energy density exceeds the surface fracture energy of the scale. The semi-empirical results were displayed in the form of failure mode maps, relating the failure strain to the scale thickness. This procedure will be adopted now for the first time for iron carbonate scales. For that it needs information on scale properties such as Young's modulus E, surface fracture energy y and yield stress o; and their dependence on the polycrystallinity and porosity. Intrinsic elastic stresses o in the scale due to scale growth can be calculated according to if the scale is assumed to be ideally brittle. ASSESSMENT OF THE MECHANICAL PROPERTIES OF IRON CARBONATE SCALES Young's Moduli The Young's moduli Eof carbonate scales were not found in the literature. For metal oxides (MgO, FeO, Fe,OCoO, NiO, MnO) they range' between 130 GN/m' (FeO) and 248 GN/m' (MgO) and for sulfides (FeS, NiS, MnS) between 50 and 70 GNlm'. As the Young's moduli depend on the type of interatomic bonding, metal carbonates (FeCO3, MnCO2) should have similar Young's moduli as oxides, because both types of compounds generally exhibit ionic bonding. For assessments a value of Ef~Qp~ ~ 150 GN/m'should be sufficiently accurate. The Young's modulus of polycrystalline materials is decreased by porosity the effect of which can be expressed by semiempirical equations' like

E = Eo (1-p)3 or E = Eo exp (-bp) where Eo = modulus of the fully compact solid, p = porosity and b 3. If the porosity consists of big as well as small pores, only the small pores reduce E in Eq. (2) or (3), while big pores act as flaws (discontinuities) and sites of stress concentration and weaken the material towards fracture. This poisson number vsc of carbonate scales can be assessed to be approximately 0.3. It is 0.32 to 0.36 for oxides (FeO, CoO, NiO, MnO) and 0.33 for sulfides (FeS, CoS, NiS, MnS)3. Critical Fracture Strain The critical fracture strain a(crit) causing through-scale cracks that initiate at preexisting flaws can be calculated by Eq. (4) and (5)'.
crit

= (2y/F2 E c)

where y is the surface fracture energy [J/m2 = N/m], E the Young's modulus [GN/m2l and F a numerical factor depending on the shape and disposition of the flaw that can take the values 1.12 for a surface notch of depth c and infinite length, F = 1 for a buried notch of width 2c and infinite length, and F = (2/v) for a semicircular surface notch of radius c. Although the flaw size c needs not to have any consistent relationship to the scale thickness h, one would be expected over a range of scale thicknesses. So c = f-h (where f is a constant between 0 and 1) may be taken as a reasonable approximation, giving
crit

= (2y/F2 f E h)

Bearing in mind that the failure occurs at the largest flaw first, the scale thickness h provides an upper bound to c. Eq. (5) shows that thinner scales crack at higher strains than thicker scales. Surface Fracture Energy The surface fracture energy y, i.e. the work of fracture, can be calculated for single crystals from first principles, using Coulombic potentials and allowing for some relaxation of atomic positions of the surface'.Typical y-values range between 2.7 J/m~ (MnO) and 6.0 J/m' (Fe,O,) for oxides {MnO, FeO, CoO, NiO, Fe,OCr,O,, Fe,O,) and between 1.2 J/m'(FeS, NiS, MnS) and 2.5 J/m' (Alp S 3 ~ Cr,S, ) for sulfides'. As there exists a correlation between the surface fracture energy y and the elastic modules' ' y, can be assessed to be approximately by 3.0 J/m' = 3.0 N/m. Porosity decreases y. Emperical equations similar to those used for elastic moduli can be applied to account for porosity'. = o. (1 - p )3or

= o exp(-bp), again with b 3 Failure Mode Diagram Thus, with = 3.0 N/m2 E = 150 GN/m', F = 1 and f = 0.1 the failure mode diagram for the spalling of iron carbonate scales can be calculated for different scale thicknesses and different spalling boundaries. The boundary lines between the brittle spalling and ductile adherence are limited by the yield y, of the iron carbonate which was estimated from data in lit.to be approx. 1.4 GN/m2. In this regime the scale remains attached to the substrate and protective. The limiting scale thickness h according to Eq. 8 h 2 E/2y F2 f is 1.46 m, if F = 1 and f = 0.1. The strain. where the carbonate scale reaches the yield strength is = y/Esc = 9.33 10 -3 If spalling of iron carbonate scales would be observed e.g. at a scale thickness of 20 m, this would correspond to a strain of approx. 2.6 10-3 if 100% is taken, yielding according to Eq. 1 - extrinsic stresses excluded - an intrinsic growth stress of 857 MN/m2 = 557 N/mm2. If only 20% is assumed, crit would be approx. 2.1 10-3 and the intrinsic growth stress would amount to 214 N/mm2, respectively. In CO2 corrosion of carbon steel spalling of carbonate scales is indeed observed. The thickness of the spalled scale depends on the environmental conditions (CO2 partial pressure, temperature, pH, salt concentration, salt composition, etc.) and on the period of scale formation during corrosion. Although systematic investigations are still lacking some first experimental investigations will be described in the following. In the first period of corrosion, i.e. when the uncorroded (metallic) surface is exposed to the CO2 corrodent, the carbonate scale growth is topotactic, i.e. the carbonate scale surface images the topography of the surface. At the first sight the surface looks scalefree after a given exposure time because the original roughness including grinding marks or scratches can still be observed. In reality, however, an iron carbonate scale was grown, only in a topotactic manner. Topotactic scales were observed to spall at thicknesses in the order 10 to 20 pm. Below this topotactic scale polycrystalline growth of iron carbonate scales occurs in the second phase with various crystallite sizes depending on the corrosion system. These observations clearly indicate that iron carbonate scales can exhibit high intrinsic growth stresses, which can cause cracking and spalling of the scale. If this occurs at the interface to the base metal and the flow regime hinders the deposition of new protecting carbonate scale FILC can initiate at the spalling site.

In view of the fact that hydrodynamic forces, expressed in terms of wall shear stresses, range in the order of 10 -5 to 10-4 N/mm2 in most technical flow systems' and may reach locally - even at flow disturbances under extreme turbulent flow conditions' only values in the order of 10 2 to 10 3 N/mm2, and, therefore, are very small compared to scale adherence strengths ranging in the order of 1 to 30 N/mm' or intrinsic growth stresses ranging in the order of 102 to 103 N/mm2, it must be concluded that hydrodynamic forces will generally only contribute to the destruction and spalling of carbonate scales in flowing systems. The primary reason for the scale destruction lies in the fracture mechanical behaviour of the scale. Once the scale is removed from the metal surface it will be the flow intensity which allows or hinders the re-formation of protective scales. It is know from literature data'" and own experimental results"" that much lower wall shear stresses are needed to prevent scale formation than to initiate scale destruction. MORPHOLOGY OF IRON CARBONATE SCALES FROM STEEL CORROSION The fracture mechanical properties of scales strongly depend on their homogeneity and growth structure. For iron carbonate - scales from CO2 corrosion of steel some information has been published in recent years"". lt has been shown that the growth structure of steels with ferritic-pearlitic microstructure is influenced by the texture of the cementite particles which is incorporated in the scale phase". It is also well known and clearly documented in quite a number of photos from cross sections of iron carbonate scales that these scales are generally not homogeneous and contain flaws, pores and other imperfections". Their formation has been already subject to theoretical considerations ". Although these information on carbonate scales grown on ferritic-pearlitic steels are very valuable from the fundamental point of view, this knowledge is only of limited applicability in cases of quenched-and-tempered steels typically used for downhole completions of sweet gas wells. Carbonate scales grown on steels with martensiticbainitic microstructure do not show any carbide oriented texture and, therefore, differ strongly in their morphology from the carbonate scales grown on ferritic-pearlitic steels. This obviously influences the fracture mechanical properties of the scales and explains the long known and well documented significant difference in FILC susceptibility of steels with as-rolled or quenched-and-tempered microstructure. The new fracture mechanical aspects of scale protectivity prompted us to revisit the morphology and homogeneity of carbonate scales on both types of carbon steels with respect to the formation conditions of the scales. The circumferencial velocity of the rotating cage was 2 m/s, the exposure time 96 h. The steel grade with ferritic-pear)itic microstructure (J 55) yielded at both temperatures (50 and 80 C) significantly higher scale thicknesses than the quenched-and-tempered grade with martensitic-bainitic microstructure (C 75). At 50 C the scale thickness of the C 75 grade ranged in the order of 50 to 60 pm, while the scale was approximately 10 ym thicker at 80 C. At this temperature the steady state scale thickness is approximately 60 ym. The J 55 grade showed scale thicknesses of 60 to 70 ym at 50 C and 80 to 90 ym at 80 C. The higher thicknesses of carbonate scales on the J 55 grade corresponds to a

higher concentration of flaws and other inhomogeneities in the scale compared to the appearance of the scale on the C 75 grade. The flaws in the scale on the J 55 grade are typically distributed over the entire thickness of the scale while flaws in the scale on the C 75 grade tend to be located at the interface between the base material and the scale. Therefore, spalling of carbonate scales from C 75 grade steel has a different effect on FILC than in the case of J 55 steel.. After cracking the scale will be spalled leaving an area of unprotected metal surface which could undergo FILC if local flow intensities prevent the re-formation of protecting scale. However, in the case of J 55 carbonate scale 'membrane' formation generally seems to occur within the scale some distance from the interface steel/scale. Due to the many flaws distributed in the scale, though scale cracking can occur anywhere in the 'membrane'. The difference to the spalling of C 75 scales is, that in the case of J 55 scale spalling, the steel surface still remains covered with scale and hydrodynamic forces are hindered to influence the process of scale formation at the interface steel/scale. This explains why J 55 grade steels are considerably less susceptible to FILC than C 75 grades"" although the inhomogeneity of J 55 grown carbonate scales is much higher than carbonate scales on C 75 steel. The inhomogeneity of J 55 scales is demonstrated in SEM photographs of crack surfaces of a carbonate scale grown during 48 h of CO2 corrosion of steel grade J 55 in brine at 80 C under 5 bar CO2 The cracks were obtained by deep cooling of a well dried corroded coupon with liquid nitrogen and susequent sudden bending at a site where the coupon had been thinned by a saw cut. Different layers of different crystallinities are clearly visible in these photographs together with flaws and caverns mostly oriented parallel to the metal surface. FAILURE MODE DIAGRAM OF IRON CARBONATE SCALES - EXPERIMENTAL RESULTS The failure mode diagram for spalling of iron carbonate scale was constructed based on theoretical considerations and reasonable assumptions. It was now interesting to estimate critical scale thicknesses for spalling of experimentally produced carbonate scales with respect to the environmental conditions during their growth, and to compare the results with the trends. Thus, the critical fracture strains of carbonate scales, formed during CO2 corrosion of the 38 Mn 6 grades under conditions described above, were estimated using Eq. 5. For the surface fracture energy the estimated value of 3.0 N/m was used. The numerical factor F was chosen as 1 assuming only buried defects in the scale. The Young's modulus E was calculated according to Eq. 2 and 3 with E, = 150 GN/m'. The porosity p was calculated from the ratio of the area A, of the sum of all small defects identified in a defined area A, on a photo of a cross section of a given carbonate scale (magnification: x500). P = Ad/At

In practice, each single defect was circumscribed by the required rectangle and the resulting areas of the rectangles were summed up to yield Ad. From the cross section also the scale thickness h and the weakening factor f were taken. The factor f was calculated from the ratio of the composite defect length t, (perpendicular to the metal surface) and the total scale thickness h. f = td/h The composite defect length was estimated according to a graphical procedure proposed by Hancock and Nicholls' using standard techniques generally applied to determine the interaction of defects in welds". In practice, each defect was circumscribed by the required rectangle and the rules of interaction were applied as detailed in literature. For the estimation of the critical fracture strain of a carbonate scale produced under a given set of environmental conditions the scale on the 20 x 2 mm cross section of a test coupon (50 x 20 x 2 mm) was completely photographed with a magnification of 100. Ten characteristic areas were then selected and again photographed with a magnification of 500. These photos were used to estimate the scale thickness h, the porosity p and the weakening factor f. For each set of experimental conditions the lowest of the resulting c, values was plotted against the corresponding scale thickness. The experimental data lie generally well in the range calculated from theoretical considerations. Iron carbonate scales grown in uninhibited solutions or in the presence of inhibitors with poor efficiency were generally found to exhibit rather low critical fracture strains, while scales formed in the presence of highly effective inhibitors yielded critical fracture strains ranging approximately one order of magnitude higher. At critical fracture strains in the 10-3 range the critical stresses calculated according to Eq. 12 yielded values in the range of 20 to 750 N/mm2. Eq. 12 yielded values in the range of 20 to 750 N/mm2.
crit

= E crit

As spalling of iron carbonate scales was observed to occur at scale thicknesses in the 10 to 20 m range (e.g. the topotactic scales formed at the beginning of corrosion) it must be assumed that (according to Eq. 1 with E = E, = 150 GN/m') intrinsic stresses in the order of 550 to 770 N/mm2 (assuming 100% y) can be built up in carbonate scales. This is in the same order of critical stresses for fracturing and spalling of iron carbonate scales from CO2 corrosion. CONCLUSIONS Hydrodynamic forces, e.g. expressed in terms of wall shear stresses, reach - even at sites of extreme flow intensities' - at best only values in the order of 10-2 to 10-1 N/mm2. Compared to scale adherence strengths in the order of 10o to 3 101 N/mm2 or critical

fracture stresses in the order of 102 to 103 N/mm2 (which for the first time were calculated in this work for iron carbonate scales) discussable wall shear stresses are generally several orders of magnitude too small to cause destruction and spalling of iron carbonate scales from CO2 corrosion of carbon steels and, hence, initiate FILC. Experimental investigations on the morphology and homogeneity of iron carbonate scales formed under different environmental conditions, and subsequent estimation of critical fracture strains and stresses yielded good consistence of the experimental data with the failure mode diagram for iron carbonate scales developed from theoretical fracture mechanical considerations. Comparison of experimentally estimated critical fracture stresses with calculated intrinsic scale growth stresses revealed that destruction of iron carbonate scales in flowing media is primarily the effect of intrinsic stresses. Extrinsic stresses (e.g. at production tubings in concentric completed wells due to cyclic mechanical loading by production related pressure pulses) can contribute to exceed critical strains and stresses at the carbonate scales, depending on the specific corrosion conditions. In case of spalling local flow intensities will decide if new protective scale can re-form or if FILC is initiated. This comparison of fracture mechanical scale properties with hydrodynamic forces considers on the hydrodynamic side only the action of wall shear stresses. It cannot be excluded that other hydrodynamic impacts on the scales, e.g.' cyclic mechanical loading due to turbulence related local pressure changes, can contribute to the destruction of corrosion product scales.