"A STUDY OF SCAL E C RACKIN G A N D ITS E F FECTS ON OXIDATION AND HOT CORROSION"

For many high temperature applications, oxidation (or hot corrosion) is an important mode of degradation of metals and alloys. Degradation mechanisms may be divided into two categories: one dealing with the chemical and transport aspects of scale growth or dissolution, and the other dealing with mechanical aspects such as stresses and scale fracture. Some applications, such as corrosion/erosion, combine both aspects in a complicated manner. Much research has been concerned with relationships between alloy composition and scale growth rates, attempting to identify alloy compositions and growth mechanisms that form compact, slow-growing scales, such as Cr{sub 2}O{sub 3} or Al{sub 2}O{sub 3}. Nevertheless, in practice a very common mode of scale degradation is cracking and spalling, followed by re-oxidation. Efforts to understand scale stresses and ultimately scale fracture have been hampered by the simultaneous interaction of numerous variables in determining the state of stress. Thus complex issues are involved in both experimental measurements and theoretical models of stresses and fracture of oxide scales. In this study we have considered both chemical/transport issues (as applied to the oxidation and hot corrosion of SiC and Ni-Cr Alloys) and mechanical issues of oxidation, but the emphasis has been on mechanical issues. In the followingmore » sections we will briefly describe the highlights of each of several projects, and where appropriate, will attach preprints or reprints of papers that describe in more detail the results of a particular study.« less

For many high temperature applications, oxidation (or hot corrosion) is an important mode of dega'adation of metals and alloys. Degradation mechanisms may be divided into two categories: one dealing with the chemical and transport aspects of scale gTowth or dissolution, and the other dealing with mechanical aspects such as stresses and scale fracture. Some appliations, such as corrosion/erosion, combine both aspects in a complicated manner. Much research hasbeen Concerned with relationships between alloy composition and scalegrowth rates, attempting to identify alloy compositions and growth mechanisms that form compact, slow-growing scales, such as Cr203 or A1203.
Nevertheless, in practice a very conunon mode of scale degradation is cracking and spalling, followed by re-oxidation. Efforts to understand scale stresses and ultimately scale fracture have been hampered by the simultaneous interaction of numerous variables in determining the state of stress. Thus complex issues are involved in both experimental measurements and theoretical models of stresses and fracture of oxide scales.
In this study we have considered both chemical/transport issues (as applied to the oxidation and hot corrsosion of SiC and Ni-Cr Alloys) and mechanical issues of oxidation, but the emphasis has been on mechanical issues. In the following sections we will briefly describe the highlights of each of several projects, and where approprial:e, will attach preprints or reprints of papers that describe in more detail the results of a particular study. The r'_TtCAL STUDIES OF STRESSES II THEOR .... __ ' "" ' In the previous annual report, it was suggested that three types of input were essential to the application of theoretical models for oxide scale cracking.
i) an understanding of substrate deformation mechanisms which appear to provide stress relief during oxidation; ii) a determination of unrelaxed thermal and/or growth stresses as a function of temperature and time; iii) the application of these unrelaxed stresses to the bi-material interface which contains a crack.
Progress in ali three of these issues includes examination of some of the Ni/NiO results described previously (Stout, et al. 1990;Abury, 1988 (Stout, et al. 1990;Abury, 1988) or slightly compressive. Third, when oxidation was conducted on two grades of nickel, the grade that retaxed the fastest (smallest oxide stresses) 'also developed the smallest oxide layer after 30 minutes of transient oxidation. This is consistelat with the transient oxidation stage of a nickel-base, alumina forming alloy (see section VII, this report) where it was found that the transient oxidation stage was dependent upon the grain size. These three observations ali lead to metal substrate relaxation as being a possible rate controlling mechanism in the transient oxidation process.
First, it was assumed that as the oxide initially formed, large stresses developed in the oxide and metal but that some relaxation process in the metal relieves these stresses. At 900°C (1173K) or T/Tru~0.68, the dominant creep mechanism is a glide/climb process as originally described by Weertman (1968 Rhines, et ', (1970) and Ashby (1973). If stresses on the order of 10-3g are ! assumed during oxidation, the creep or relaxation rate is on the order of 5 x 10.3 s-1, This is on the order of the lattice parameter recovery rate (~5 x 10-4 s-1) which was observed to '= occur during the transient oxidation stage where large fluctuations in the lattice parameter were observed. In fact, however, the inferred stresses during this stage were u,_ the order of 10.2 la. These very large stresses would lead to glide/climb relaxation rates of 100 s'l, much faster than those observed. There are two possibilities' i) a slower creep rate controlling mechanism controls relaxation near an oxide/metal interface (not too likely); ii) the relaxation process is partly offset by a continued stress build up due to oxide formation in the transient stage.
We are inclined toward the latter view, with the total observed strain (lattice parameter change) being a combination of relaxation and continued oxide growth. In terms of residual strain these are competitive processes, so that the time dependent residual strain, eR(t), depends on both the creep kinetics, _c,,and the oxidation kinetics, giving where _:c(t)for Ni/NiO is glici_/climb dependent, Both the creep rate as affected by the dislocation substructure, and the rate of oxidation will be affected by stress, time and temperature, While it is premature to suggest a mechanism, it appears as though the stresses eventually diminished to a small value, giving _c~_ox with the lattice becoming fully relaxed. This is sigrlificant since it was originally thouglat that rapid diffusion down grain boundaries would assist the tnmsient oxidation process, This would lead to short time oxidation kinetics being increased by fine grain structurest', Equation 6 predicts just the opposite as was observed by the experimental data and illustrated in Fig. 1. The tentative conclus' _,nthen is that the transient oxidation stage is inversely proportional to the rate of substrate relaxation. Note that for the ste_dy-state oxidation process, no effect of grain size was observed. This is consistent with the subsu'ate being fully relaxed and the steady-state plateau in metal lattice parameter observed after a transient stage. MPa, this only gives a KI value of 0,78 MPa -m 1/2, insufficient for film cracking. We are in the process of analyzing such films tk)rresidual stress and the rate of relaxation in such ,!" The additional possibility is that in these alloys it is solute diffusion clown grain boundaries which is enhanced leading to protective oxides and thus enhanced oxidation resistance in fine-grain sm_ctures, In this "elastic-plastic" analysis, the sample material is an iron crystal with elastic constants of C ll 24,2, C12 = 14.65 and C44 -11,2 in units of 104 MPa and the crack .
length a = 0,01 m, All of the discretized superdislocations are assumed straight and parallel , to the z-axis, and also lie in a plane which cuts the xy plane in a slip trace at an angle 0 to the x-axis. The dislocation can move in this plane in a direction which makes an angle (1) " with the z-axis. Ali of the parameters used in the computation scheme for various KI are listed in Table 1, The elastic simulations from this model were compared with the analytical elasticity solutions (Hellan, 1984); as expected, no difference was found, lt was found that significant regions of calculated elastic strecs far exceeded I(X)GPa, which is unacceptable for equilibrium since this exceeds the theoretical sla'ength by a factor of three. The elasticplastic stress contours were also compared with corresponding elastic stress contours. No obvious differences were found at a distance about 1 mm away fromthe tip which is als() i consistent with the continuum model si_',tce the plastic zone size was .<i1 nam, It should be pointed out that a reasonable ageement with a continuum small-scale yielding solution was also obtained to within about 1 btm "'f the crack tip.
On the other hand, the stress state very near to the crack tip is at large variance with either the elasticity or the continuum elastic-plastic results in a number of ways. First, the present analysis removes the stress field singularity (at the crack tip). Instead of the highest tension point, the stress at the crack tip became sli_ ,dy compressive clue to dislocation • shielding. The result of a completely shielded crack tip is unexpected although anticipated, in the context of the Rice-Thomson model (Rice and Thompson, 1974). As is seen in

III. EXPERIMENTAL MEASUREMENT OF SCALE STRESSES
Progress in the first year of this grant was focussed on two objectives, The first was to evaluate the sign and magnitude of the residual elastic strain that develops across the interface between a metal substa'ate and an attached oxide, The oxide fomas at elevated temperature but upon cooling, very significant residual stresses can develop, We chose Cr/Cr203 Itsour model system and x-ray ditfi'action as ftle basic technique for evaluating the residual strain in both tlm oxide and the Cr subsmtte, Tlm results of this study 1, demonstrated in-plane compressive stresses iraCr203 of several thousand MPa, and in , excess of that expected from a calculatior_ of lattice mismatch between the oxide and the metal due to c(_ling from high temperature (Stout, et al, 1986), We concluded that growth stresses developed at elevated temperature during oxide formation could make a significant contribution to that observed residual stress at room temperature, The second objective was to design a procedure and ,'mexperimental apparatus by which the x-ray diffractiorl technique could be directly applied at elevated temperature such that the conclusion drawn above could be experimentally verified, To our knowledge, this experiment had never been done irathe context of the intencled application, Rapid progress had been made in that direction by the end of the first year, The naa, jor accomplishment in tlm second year of this grant wt_sto demonstrate, by direct experiment, the existence of significant growth stresses iraa metalA_xide scale system at high temperature, The experimental pr{×'edure which we designed allows high , + Fin:IiReport .... l-)ngeII tetnperntt.lre X.ray dil'fraction n',ettsurernerlts to be mnde during oxlde growth, Brlel!'ly,the experh'rtertts utilize the interplanm' lattice spncings of crystalline alloys arid their attached + oxides tis strain gauges, Ata unoxldized nlloy is initi,'llly hettted to elevated temperature in tin inert lttmosphere, and, of course, its l_ttticeexpands, The interplanar spacings of the lttttice tire measured by X-ray diffrttction its a function of temperature, After the alloy is fully equilibrated madits interplanar sp_tcings rerrmin unchanged, preheated oxygen ts introduced iiito the furnace at ten.tperature, The alloy oxidizes isothemaally arid the ehtstic response of II.tcalloy s+,,bstr+tte to the developing scitle is rnonttored by the diffraction technique, Figure 2 illustrates a typical experin.tenml run on a 25_m thick Ni foil, beginning +trroom temperature (RT) , The data points on the upper straight line represent the metal strain (irt percent) due to themml expansion to 900°C, in tin inert tttmosphere in order to minimize oxidation, Once therrrml equilibrium has been established at 900 C, oxygen is introduced and NiO begirts to grow on the foil, The response of the foil tc) that [sothemml oxidation process is shown by the strain dattt in the inset of Fig, 2, .
An initial compression is followed by _mappttrent relaxation and that pattern is repeated 3 more times in a little more than 2 hours after oxidation commenced, Therettfter, the strain stabilizes tit the themrally expanded value (about 1,36 % strain relative to RT) and does not ch+mge until cooling is initi+tted, In this pm+ticul_u • experiment, there was no measureable residu+tlstrain (other than them.tal strltin) tlfter the foil was cooled with its oxide scale intact to room ternperature, In other experiments, the metal strain due to oxid_ttion is gradual with no evidence of the oscillating behavior illustrated in Fig, 2, In these cases, signific+mt residuttl stress is evident at RT, The two types of behavior just described t'mvebeen experimentally confim.ted iri additional furls and the appropriate checks have been made to eliminate the possible effects of sttmple rotation and/or cwstalUte rotation during the oxidation process, We conclude th+ttthe metal behavior at high temperature is responding to the developing oxides. The details of that response are not as yet understood, but we suspect that the oscillating strain behttvior at high temperature rnay be due to successive relaxations of accumulanrtg metal strain, e,g,, by cracking, as the oxide grows, Such behavior seems to minimize the final residu+tl stress, The tnajor accon.tplishment irt the third and final year of this grant was to refine our procedures for measurirt ; high ternpernture growth stresses madto apply these pr(x:edures to different materials, in order to evaluate l'urther stress release, The corresponding strtxinhtst( ry rts recorded in the (2()()) lttl.ticespttcirtg of the alloy is sltown by tile;closed syrnbols, The linear expansiott of the lattice during hettting corresponds to previous, irtclei_endentresults, Upon oxidation, however, the alloy substrate gradually g(_s into compression relative to the thermally expanded but Otherwise tmsu'ained state, This strain accumulates over the oxidation period anti then is apparently preserved as the major component of the residual strain after cooling is complete, The thertnal conu'actton due to cooling is indistinguishable, except in sign, ft'ore the expansion due to heating, Thus the alloy residual swain is tit a state of slight compression relative to the,initially unstrained, room temperattlre lattice, The relative strain of the alloy throughout the thermal cycle ,just described can be determined because the strain-free state could be established prior to the experiment, This is not the case for the attached oxide scale, "I'hesw,fin-free state of the oxide can only be obtained after the experiment, once it is detached from its substrate. Because the amount of oxide is so small, we have developed a procedure in the last several months by which the oxide is separated from its substrate, thereby relaxing any residual stress that rnay be present. Briefly, the substrate and oxide together are placed in the top of a finely tapered glass capillary and the matal is dissolvedaway with a bromine solution, After dissolution, the solution is repeatedly dectmted rotd the oxide flakes are washed, Then the released oxide is allowed to settle into the finest part of the capillary. That end of the capillary is then broken of _tndplaced in a Debye-Scherrer X-ray camera for precise lattice parameter rheas urements, Our initial results from this procedure shows that the scale that develops on NiCrAIY at I(X)0C is pure o_-A1203,and its lattice parameters are indistinguishable from those of' an independent standard, This means that there can be no compositional effects during oxidation or cooling that could have misled us irt our interpretation of the observed changes,

IV, THERMAL CYCLING OXIDATION AND FRACTURE OF CR203 ON NI.30CR
Nickel-30Cr is a simple alloy that can serve as a model for oxidation and corrosion processes on complex, Ni-base superalloys, Nickel-30Cr reliably rortns a Cr203 scale during high temperature oxidation, but the scale cracks and/or spalls on cooling to ambient temperatures, Ttte oxidation of Ni-30Cr alloy was studied under thermal cycling conditions to ext+.minethe scale fracture process, In a typical experiment a sample was tieated rapidly to the oxidizing temperature (1000°C) in O2 and held isothem+mllyfor some time to grow a Cr203 scale, Then a series of three thermal cycles (isothermal temperature-, lqnal Report ... page 13 cool -return to isothermal temperature) with increasing ATs of 100, 200 and 3('_0°Cwere imposed, culminating in c(mling to room temperature. Using acoustic emis,': ,1(AE) we monitored the occurrence of scale cracking during these experim,mts. Typically, no significant cracking was observed during the isothermal regime, but extensive cracking and spalling were observed.during the cool down and occasionally during the thermal cycling. The results of our study showed that thick scales grown at 1000°C cracked at a smaller AT than thin scales, This suggests that either (a) the growth stress increases with scale thickness or (b) the stre,.;sto fracture a thick scale is less than that for a thin scale.
Inference "a" is more likely, For a given scale thickness grown at two temperatures (800 vs 1100°C), a smaller thermal stress was needed to fracture the scale grown at the lower temperature. This result is probably due to less stress relaxation, and hence a larger net growth stress, at the lower temperature. Clearly, other factors, such as the microstructures of the scales, could also vary in these tests and could affect the fracture stress. A draft L , paper on this work is attached.   Fig. 4. A paper describing these calculations has been published: Barnes, et al., 1989, and a preprint isattached.

VI. OXIDATION OF SIC IN THE PRESENCE OF ALKALI SALT , VAPORS
The purpose of this study is to characterize the oxidation response of SiC when the environment contains small concentrations of alkali salt vapors. The oxidation of SiC in "clean" environments, eg. pure 02, produces a SiO2 scale which is very protective at high temperatures. On the other hand, alka!i vapor species can react with the SiO2 and modify its properties. In our studies, potassium has been incorporated into the SiO2 scale formed on sintered o_-SiC,and the potassium has drastically lowered the scale viscosity and increased the rate of transport of oxidant across the scale to the SiC surface. As a consequence, we have observed oxidation rates as much as 700-800 times higher when the environment contained a few hundred ppm (up to --325ppm) of a potassium salt than that for "clean" oxidation. Furthermore, the oxidation kinetics changed from parabolic to linear with the addition of potassium. A plot of the oxidation/corrosion rate, as a function of activity of potassium salts in the gas phase, is shown in Fig. 5. A paper describing this work has been submitted for publication and a preprint is attached.
As a closing effort on this study, we are considering some schemes which are intended to reduce the rate of oxidation in alkali contaminated environments. period, regardless of alloy grain size. Nevertheless, some cracking or spalling in the initial stages, where the oxide is relatively thin, might have been below the threshold detection limit of the system. A substantial amount of acoustic activity was observed on cooling from the oxidation temperature to room temperature, but ,._ocorrelation with alloy grain size was found. This behavior suggests the stress-related behavior is established by the steady state oxidation process, which is controlled by transport through oxide grain boundaries.
A draft of a paper (prepared by the graduate student) describing this study is attached. _ ..f _ . Linear rate constant as a function of potassium salt vapor level for CO2-O2-H20,gas mixture at 1300°Cwith variouspH20, Figure 5 Corrosionrates for the oxidation of SlC as a function of potassium salt vapor level,