On the Degradation of Retained Austenite in Transformation Induced Plasticity Steel

University of Wollongong Research Online Australian Institute for Innovative Materials Papers Australian Institute for Innovative Materials 1-1-2020 On the Degradation of Retained Austenite in Transformation Induced Plasticity Steel Ilana Timokhina Azdiar Adil Gazder University of Wollongong, azdiar@uow.edu.au Jiangting Wang Ilias Bikmukhametov Peter Hodgson See next page for additional authors Follow this and additional works at: https://ro.uow.edu.au/aiimpapers Part of the Engineering Commons, and the Physical Sciences and Mathematics Commons Recommended Citation Timokhina, Ilana; Gazder, Azdiar Adil; Wang, Jiangting; Bikmukhametov, Ilias; Hodgson, Peter; Niessen, Frank; and Pereloma, Elena V., "On the Degradation of Retained Austenite in Transformation Induced Plasticity Steel" (2020). Australian Institute for Innovative Materials - Papers. 4174. https://ro.uow.edu.au/aiimpapers/4174 Research Online is the open access institutional repository for the University of Wollongong. For further information contact the UOW Library: research-pubs@uow.edu.au On the Degradation of Retained Austenite in Transformation Induced Plasticity Steel Abstract © 2020, The Minerals, Metals & Materials Society and ASM International. A transformation-induced plasticity steel was thermomechanically processed and then transformed to bainite at an isothermal transformation temperature of 723 K for 1800 seconds, which exceeds the time required for completion of the bainite transformation. The formation of lenticular-shaped carbides with a triclinic lattice and internal substructure was found after thermomechanical processing. After 16 years of storage at room temperature, the decomposition of retained austenite into pearlite was observed for the first time at this temperature. Disciplines Engineering | Physical Sciences and Mathematics Publication Details Timokhina, I., Gazder, A., Wang, J., Bikmukhametov, I., Hodgson, P., Niessen, F. & Pereloma, E. (2020). On the Degradation of Retained Austenite in Transformation Induced Plasticity Steel. Metallurgical and Materials Transactions A: Physical Metallurgy and Materials Science, Authors Ilana Timokhina, Azdiar Adil Gazder, Jiangting Wang, Ilias Bikmukhametov, Peter Hodgson, Frank Niessen, and Elena V. Pereloma This journal article is available at Research Online: https://ro.uow.edu.au/aiimpapers/4174 1 On the degradation of retained austenite in transformation induced plasticity steel. 2 3 Ilana B. Timokhinaa, Azdiar A. Gazderb, Jiangting Wanga, Ilias Bikmukhametovad, Peter D. 4 Hodgsona , Frank Niessenb, Elena V. Perelomab,c 5 6 7 aInstitute bElectron cSchool Microscopy Centre, University of Wollongong, Wollongong, NSW 2500, Australia of Mechanical, Materials, Mechatronic and Biomedical Engineering, University of Wollongong, Wollongong, NSW 2522, Australia 8 9 10 for Frontier Materials, Deakin University, Geelong, VIC 3220, Australia dDepartment of Metallurgical & Materials Engineering, The University of Alabama, Tuscaloosa, Alabama, 35487, USA 11 12 ilana.timokhina@deakin.edu.au, azdiar@uow.edu.au, jiangting.wang@deakin.edu.au, 13 ibikmukhametov@ua.edu, peter.hodgson@deakin.edu.au, contact@fniessen.com, 14 elenap@uow.edu.au 15 16 Corresponding author: ilana.timokhina@deakin.edu.au 17 18 Abstract 19 A transformation-induced plasticity steel was thermo-mechanically processed and then 20 transformed to bainite at an isothermal transformation temperature of 723 K for 1800s, which 21 exceeds the time required for completion of the bainite transformation. The formation of 22 lenticular-shape carbides with a triclinic lattice and internal substructure was found after 23 thermomechanical processing. After 16 years storage at room temperature the decomposition 24 of retained austenite into pearlite was observed for the first time at this temperature. 25 26 27 Keywords: transformation-induced plasticity steel; retained austenite decomposition; atom 28 probe tomography; pearlite formation at room temperature; transmission electron microscopy; 29 electron back-scattering diffraction. 30 31 32 1. Introduction 1 33 Retained austenite (RA) is an important phase in many modern advanced high strength steels, 34 such as Transformation-Induced Plasticity (TRIP) steels [1-5], nanobainitic steels [6-8], 35 quenched and partitioned (Q&P) steels [9-12] and medium-Mn steels [13-15]. In all of these 36 steel grades, RA is the product of an incomplete bainite transformation due to the high Si 37 concentration leading to the high carbon content of RA [16, 17]. To a large extent, the carbon 38 and manganese contents of the RA determine its chemical stability [1, 18, 19]. The 39 microstructure and stability of the RA is important for production and in-service performance 40 of components manufactured from the steel, as any changes in the microstructure (phase 41 constituents, morphology, coarseness, etc.) have a direct effect on the mechanical properties 42 (toughness, yield strength, crashworthiness, etc.). A high carbon concentration in the RA is 43 desirable as it promotes stability on cooling and against the deformation-induced martensite 44 transformation [18, 19]. However, this also makes the RA more susceptible to decomposition 45 into ferrite and carbides at elevated temperatures due to the increased driving force for 46 carbides precipitation from austenite with higher carbon content [20]. The precipitation of 47 carbides in bainite during early stages of the bainite transformation also leads to less stable RA 48 and its transformation to martensite on subsequent cooling [21-23]. The stability of the RA is 49 also somewhat related to its morphology, as it has been commonly reported that film-like RA 50 between neighbouring subunits of bainitic ferrite (BF) has a higher carbon content compared to 51 the blocky RA located between sheaves of bainite [20, 24]. However, it was recently shown that 52 blocky RA could have a high carbon content similar to film RA [25]. 53 With the progress of the incomplete bainite transformation, the carbon enrichment of the RA 54 increases until a critical level, at which the RA decomposition takes place. For example, this 55 occurs when the holding time at isothermal transformation temperature exceeds the time 56 required for the bainitic reaction [17]. Alternatively, for the same holding time, an increase in 57 the isothermal bainite transformation temperature/tempering temperature will have a similar 58 effect on the carbon content in the RA and result in accelerated RA decomposition [26, 27]. 2 59 To-date, there is limited information available with respect to the mechanism of RA 60 decomposition and on the types of carbides formed. Park et al. [26] reported that no 61 decomposition of the RA was observed in a Fe-0.2C-2.Mn-1Si-1Al (wt.%) TRIP steel after 62 tempering at temperatures below 473K. However, at higher temperatures, the first precipitate 63 to form in partially decomposed RA was -carbide followed by the precipitation of cementite. 64 The latter became globular and spheroidised at longer holding times at high temperatures. On 65 the other hand, in a medium Mn steel (Fe-0.1C-5Mn wt.%) the RA decomposition was delayed to 66 temperatures above 673 K; blocky RA decomposed into ferrite and rod-like cementite at 723 K. 67 Similarly, the formation of cementite was only detected after tempering for 30 min or longer at 68 723K for a Fe-0.2C-3Mn-2Si wt.% steel [23]. Full decomposition of the RA into ferrite and 69 cementite was observed in both TRIP and nanobainitic steels after reheating and holding for 3 70 hrs at 723K [28]. Atom probe study of hot-worked tool steel showed a laminar arrangement of 71 cementite after RA decomposition at 883K [22]. Another morphology of carbide formed on the 72 decomposition of RA has been reported by Sandvik [29] in a 0.9C-2Si-0.5Mn-0.42Cr wt.% steel 73 at 653 K , with ~100nm thick, lenticular plate-shaped carbides with a triclinic crystal lattice and 74 a midrib. 75 To the best of the authors’ knowledge, there have not been any studies to-date on RA stability 76 during long term service life at ambient temperatures. In this work, this deficiency is addressed 77 by studying the microstructure of thermo-mechanically processed TRIP steel [30] after more 78 than 16 years in storage using a combination of transmission electron microscopy (TEM), 79 electron back-scattering diffraction (EBSD), and atom probe tomography (APT). 80 81 2. Material and Methods 82 The nominal composition of the TRIP steel used in this study was Fe-0.21C-1.51Mn- 83 1.49Si- 0.004Mo - 0.01Al - 0.036Nb wt.%.Thermo-mechanical processing (TMP) was 3 84 undertaken on a laboratory rolling mill. The TMP schedule (Fig. 1) was constructed by utilising 85 continuous-cooling-transformation data [31]. The strip was reheated in a 15 kW muffle furnace 86 and soaked at 1523 K for 120 s. After soaking, the strip was 25% rough rolled at 1373 K 87 followed by a 120s hold to uniformly condition the recrystallised austenite. Subsequent 47% 88 finish rolling in the non-recrystallised region was undertaken at 1123 K followed by cooling at 1 89 90 K ⋅ s −1 to the accelerated cooling start temperature of 943 K to form ~50 vol.% of polygonal 91 Following this, two spray guns were used to cool the strip at ~20 K/s to 773 K to avoid pearlite 92 formation. At this point, coiling was simulated by lowering the strip into a fluid bed furnace, 93 covering it with aluminium oxide sand and isothermally holding at 723 K for 1800 s to form 94 bainite. Following this, the strip was quenched in an iced brine solution (Fig. 1). The final strip 95 thickness was ~6.5 mm. The strip was produced in 2002 and kept in storage at room 96 temperature (RT) for 16 years. This enables study of the effect of prolonged storage on the TRIP 97 steel microstructure. ferrite; which is the optimal amount required to stabilise the highest volume fraction of RA [32]. 98 99 100 101 102 103 104 105 106 Figure 1. Thermomechanical processing schedule. 107 108 The volume fraction of RA was measured using a Philips PW 1130 diffractometer equipped with 109 a Ni-monochromated Cu Kα radiation source operated at 40 keV and 25 mA in Bragg-Brentano 110 geometry. The relative intensities of the peaks were recorded by varying 2θ from 40-90 in 4 111 112 113 continuous scan mode at a rate of 0.5 min−1 with a step size of 0.05. The volume fraction of RA was estimated by the direct comparison method using the integrated intensities of the (200)𝛼 and (211)𝛼 peaks from bcc ferrite and the (200)𝛾 and (220)𝛾 peaks from fcc austenite [33]. 114 Since the original strip was produced in 2002, electron backscattering diffraction (EBSD) maps 115 of the microstructure after TMP was not carried out as this technique was not accessible at that 116 time. However, the strip subjected to prolonged room temperature storage for 16 years, was 117 mapped. 118 The samples for EBSD were cut from the centre of the aged strip width along the normal 119 direction–rolling direction (ND–RD) and mechanically ground to 0.3 mm thickness using up to 120 1200 grit silicon carbide paper. Ø3 mm discs were punched out, after ensuring that each disc 121 contained a short chord parallel to the RD, in order to identify the macroscopic sample 122 coordinates. The discs were manually ground to ∼70 μm thickness using 2400 and 4000 grit 123 silicon carbide papers, and then twin-jet electro-polished to produce electron transparent foils 124 using a solution of 90% methanol and 10% perchloric acid in a Struers Tenupol-5 operated at 125 30 V, ∼150 mA and 243 K. 126 EBSD orientation and Energy Dispersive Spectroscopy (EDS) elemental data were obtained 127 simultaneously from the centre of the ND–RD cross- section using a JEOL JSM-7001F field 128 emission gun scanning electron microscope operated at 15 kV accelerating voltage, ~5.1 nA 129 probe current, 12 mm working distance at ×1000 magnification. The microscope was fitted with 130 a Nordlys-II(S) EBSD detector and an 80 mm2 X-Max EDS detector interfacing with the Oxford 131 Instruments (OI) Aztec software suite. A step size of 0.06 µm was employed such that a map 132 comprising 2000 × 1500 pixels, which corresponds to an area of 20 × 90 µm2 , was collected 133 over ~61.5 hours. 134 The EBSD mapping conditions were optimised beforehand with a 16.33 ms camera exposure 135 time, 43 and 32 reflectors employed for the bcc and fcc phases, respectively, 4×4 binning, 3 136 background frames, a Hough resolution of 60. The speed of acquisition of the individual electron 137 backscattering patterns was 13.5 Hz (~73.9 ms) with up to 11 Kikuchi bands concurrently 5 138 indexed via the OI “Refined Accuracy” algorithm. The raw EBSD maps returned an overall 139 indexing rate of 97.48% such that most of the zero solutions were concentrated at boundary 140 interfaces. 141 EDS maps were obtained using a 0-20 keV energy range, 2048 channels, a process time of 3, a 142 detector dead time of 50–55% and a pixel dwell time of 24.8 ms. Over the full ‘TruMap’ area, the 143 Fe-K, Mn-K, Si-K, C-K, Nb-L, Cu-L, Al-K and Mo-L lines returned distributions of relative 144 frequency versus counts per second (cps) with maximum counts rounded-off to 1055, 232, 335, 145 225, 119, 57, 670 and 695, respectively. 146 Post-processing of the EBSD maps were undertaken using the Oxford Instruments HKL Channel- 147 5 software suite. In brief, it involved the removal of wild spikes and cyclic extrapolation of zero 148 solutions up to five neighbours followed by thresholding the band contrast histogram to 149 delineate unindexable regions (Fig. 2a). Since the EBSD maps were indexed as iron fcc 150 (austenite) and bcc (ferrite) during acquisition, a previously developed procedure using the OI 151 Channel-5 software suite [34] was modified to initially segment the phases into austenite, 152 polygonal ferrite, and an unsegmented fraction comprising bainite, a mixed region of 153 martensite/pearlite and carbides. While a detailed step-by-step guide to segment polygonal 154 ferrite from the unsegmented fraction is given in Ref. [34], a brief description of the procedure 155 to segment bainite and the mixed region of martensite/pearlite and carbides is as follows. 156 To segment the bainite and the mixed region of martensite/pearlite and carbides, the C-K EDS 157 158 159 160 161 162 163 map data of that subset was imported as an *.tiff image into the Ilastik v3.1 software. Ilastik is an open source software for image classification and segmentation [35]. Using the in-built, machine learning pixel classification algorithms, the subset was segmented into an image comprising three grayscale colour thresholds for bainite, the mixed region of martensite/pearlite and carbides signifying lower, medium and highest carbon contents, respectively (Fig. 2b last carbon- K map). A specifically written MATLAB script read and assigned the varying grayscale colours to individual pixels of the *.ctf file. Following this, the *.ctf file was re-imported back into OI 6 164 165 Channel-5 and the varying grayscale colours assigned as separate phases; in order to colour- code the phase map (Fig. 2c phase map). 166 Bright field (BF), dark field (DF) and selected area electron diffraction (SAED) studies were 167 undertaken on a Philips CM20, operated at 200 kV to characterise the initial microstructure 168 after TMP in 2002. Similar work was performed to study the strip after 16-year storage using a 169 JEOL 2100F operated at 200 kV. In both cases, electron transparent thin foils were prepared by 170 twin-jet electro-polishing. Orientation distributions along BF layers were studied by SAED 171 patterns using an aperture of 1.1 µm nominal diameter. 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 Figure 2. Representative (a) band contrast, (b) carbon – K and (c) phase maps. In (b), the 187 areas exclude polygonal ferrite and austenite. The area denoted by the white dashed rectangle 188 in (c) is magnified in (d). In (c) and (d) red = austenite; blue = polygonal ferrite; green = 189 ferrite in bainite; yellow = carbides; aqua = mixed region (see text for details). 190 7 191 Atom probe tomography (APT) was used to study the variation in local chemical composition of 192 the phases, solute redistribution and segregation across the phases [36]. APT needles were 193 prepared from bulk polished sample and TEM foils using focused ion beam milling (FIB) in a 194 dual beam FEG-SEM (FEI Quanta 3D). For example, pearlitic regions were identified and their 195 location marked on the foil using TEM (Fig. 3). Following this, Pt supporting layers were 196 deposited in the region of interest on both sides of the foil (Fig. 3b). The targeted specimens 197 were then cut from the region of interest and lifted out on to an APT sample holder. Finally, APT 198 needles were sharpened using the FIB operated at 30 kV for initial shaping and 8 kV for final 199 polishing. 200 201 202 203 204 205 206 207 Figure 3. TEM micrograph (a) and position of APT sample (b) taken from pearlite shown in (a). 208 F is ferrite, P is pearlite, B is bainite. Dash line outlines the pearlite region and APT sample taken 209 from this area. 210 211 APT was conducted on a Cameca LEAP 5000 XR operated in voltage mode with a pulse rate of 212 200 kHz and a sample temperature of 60 K. Data reconstruction and quantitative analysis was 213 undertaken within the IVAS 3.6.14 software suite [36]. Phase composition was determined from 214 the regions without visible coarse particles, boundaries and/or defects. 215 Thermodynamic modelling was applied to assess the phase equilibrium for the local 216 composition of RA at ambient temperature. The calculations were performed with Thermo-Calc 217 2020a [37] using the TCFE9 thermodynamics database [38]. 8 218 3. Results 219 The microstructure after TMP comprised ~50% of polygonal ferrite with layers of granular 220 bainite and acicular ferrite (Fig. 4a). RA islands and films were located in-between polygonal 221 ferrite grains, at interfaces between polygonal ferrite and BF, between BF laths and as 222 martensite/RA constituent (Fig. 4). XRD confirmed that the steel contained ~5% RA. In addition, 223 coarse, lenticular-shaped carbides (hereafter referred to as LSC) were detected in bainite 224 between the BF laths (Fig. 4b). The morphology of LSC differed significantly from those seen 225 previously in upper and lower bainite [17, 31] and they also contained a high density of internal 226 faults (Figs. 4b-4d). The average size of LSC was estimated using their projected lengths in the 227 (111)𝛼 matrix plane such that they averaged 370100 nm length and 10030 nm width. 228 Indexing their diffraction patterns showed that the LSC crystal structure was close to triclinic 229 (or alternatively, distorted orthorhombic) with approximate lattice parameters of a = 0.638 nm, 230 b = 0.505 nm, c = 0.459 nm, and α = 90°, β = 70.1° and γ = 84.7° [29]. In some cases, TEM also 231 revealed the presence of LSC near RA; an observation that could be ascribed to the partial 232 decomposition of RA that leads to the formation of LSC and ferrite (Figs. 4c and 4d). TEM 233 investigation showed that “recovered ferrite”, or ferrite grains with dislocation density lower 234 than that in BF but a higher than in PF (for example, the area marked by F in Fig. 4c), are present 235 in the areas in close proximity to LSC, presumably as a result of the decomposition of RA. 236 237 238 239 240 9 241 Figure 4. Optical micrograph (a), TEM bright field (b, c) and TEM dark field from (002) 242 diffraction spot (d) micrographs showing initial microstructure of TRIP steel after 1800 s 243 isothermal hold at 723K (a), lenticular carbides (b) and decomposed austenite (c, d) in the 244 samples after TMP. Zone axis for (b) is [21̅1̅], and for (c) is [110] // [010]c. 245 246 SEM on bulk samples (Fig. 5a) together with EBSD (Fig. 2) and TEM (Fig. 6) after 16 years room- 247 temperature storage confirmed the following microstructural features: (i) the formation of 248 lamella-like, coarse pearlite between ferrite grains (Figs. 5, 6c), (ii) lamella-like, fine or 249 degenerate pearlite formed at the ferrite/bainite interface, (Figs. 2c, 2d, 6a, 6b) and (iii) the 250 presence of LSC in bainite (Figs. 2d and 6d). 251 252 253 254 255 256 257 Figure 5. SEM micrograph showing position of APT sample (dash line) taken from pearlite (a), 258 corresponding Fe-C atom map (b) and composition profile across cementite in pearlite (along 259 dash line shown in atom map (b)) (c). 260 261 In the representative phase map shown in Fig. 2c, austenite, polygonal ferrite, ferrite in bainite 262 and carbides are denoted by red, blue, green and yellow colours which correspond to map areas 263 of 0.18%, 69.4%, 23.5% and 0.49%, respectively. The yellow areas (Figs. 2c and 2d) possess the 10 264 highest carbon content (Fig. 2b) and their morphology correlates with the LSC seen in TEM (Fig. 265 6d). The 6.43% area denoted by the aqua colour comprises the mixed region of martensite, 266 pearlite or secondary formed ferrite via RA decomposition. It is also noted that the aqua areas in 267 the immediate vicinity of a carbide could be a diffusion zone of carbon enrichment; in which 268 case these localised regions could also be carbon enriched BF. 269 270 271 272 273 274 275 276 277 278 279 280 Figure 6. Representative TEM micrographs of the microstructure after 16 years in storage: (a) 281 lamella-like ([1-10]a//[-100]c and (b) degenerate pearlite, (c) pearlite in ferrite and (d) 282 presence of LSP in bainite (zone axis is [-2-11]). PF is polygonal ferrite, B is bainite, P is pearlite 283 and LSC is lenticular shape carbides. Note that coarse pearlite is not edge-on in (c) in order to 284 better reveal the cementite layers. 285 11 286 Two types of pearlite with different inter-lamellar distances were observed: (i) coarse and (ii) 287 fine. The thickness of cementite layers in coarse pearlite was 12±3 nm, while the thickness of 288 ferrite was 100±10 nm (Fig. 5). In the case of fine pearlite (Figs. 6a, and 7), the thickness of 289 cementite was ~10 ±2 nm and the thickness of ferrite was ~ 50±5 nm. 290 291 292 293 294 295 296 Figure 7. Representative carbon atom map (a) and concentration profile (b) along the box 297 shown in (a) of pearlite from the site specific sample shown in Figure 3. 298 299 The carbon atom maps from the site-specific APT needles proved the formation of cementite 300 and ferrite lamellae (Figs. 5b, 5c and 7a). The average cementite and ferrite thicknesses were 301 comparable to that obtained from SEM and TEM data for both coarse and fine pearlite. The 302 average carbon content calculated from the centre of these cementite layers varied from 18 to 303 27 at.%; which correlates well with its expected ~25 at.% carbon content (Figs. 5c, 7b). The 1D 304 concentration profiles also showed: (i) the partitioning of Mn to cementite (increasing from 1.5 305 at.% in ferrite to ~6 at.% within the cementite layers) and, (ii) the depletion of Si in cementite 306 layers to ~0.01 at.% compared to ~3.8 at.% in ferrite (Figures 5c and 7b). 307 Based on APT, it was found that prolonged storage for 16 years affected the redistribution of 308 solutes, mainly carbon, between and within the phases. This occurred in BF and RA/martensite 12 309 due the high dislocation densities of these phases. Firstly, segregation of C and Mn at 310 ferrite/bainite interfaces was observed (Fig. 8a). 311 312 313 314 315 316 317 Figure 8. Representative carbon atom map (a, c) and concentration profile (along the box shown 318 in (a)) across ferrite/bainite interface (b); and segregation of C at dislocations in 319 martensite/retained austenite (5 at.%C iso-surface ) (c). F is ferrite and B is bainite. 320 321 The local concentration profile along the selected box perpendicular to a ferrite/bainite 322 interface shows an increase in C level from ~ 0.06 ± 0.01 at.% in the matrix of polygonal ferrite 323 to 1.2 at.% at the interface. At the same time, the BF continued to be saturated in carbon (~0.2 ± 324 0.02 at.%). The Mn level is 2.4 at.% at the interface and it gradually decreases to ~ 1.38 ± 0.1 325 at.% in the matrix of polygonal ferrite and to ~ 1.56 ± 0.08 at.% in BF (Figs. 8a, 8b). Secondly, a 326 carbon segregation at dislocations and the formation of carbon clusters in martensite/RA were 327 also found in this condition. (Fig. 8c). 328 329 4. Discussion 13 330 Since maintaining the required level of the RA stability is necessary for TRIP steel sustainability, 331 it is important to study the microstructural changes after prolonged room temperature 332 exposure. 333 In this regard, the addition of Si to this TRIP steel is important as it typically inhibits the 334 formation of cementite; a process that removes carbon from RA. It follows that inhibiting 335 cementite formation is crucial to the overall stability of RA which increases with the progress of 336 the bainite transformation [39]. However, if the coiling time exceeds the time of the bainite 337 reaction, then the RA can decompose with the formation of secondary ferrite and carbides [29]. 338 The TEM investigation revealed the presence of LSC within bainite after isothermal holding for 339 1800 s. The morphology of these LSC is dissimilar to ε-carbides or cementite that precipitate 340 rapidly in upper or lower bainite [17]. It also differs from the thin carbide needles formed in BF 341 laths observed after 40 min holding at 773 K in a 1.83Si wt.% steel [40] such that their 342 formation was ascribed to the secondary stage of the bainite reaction. As these LSC were not a 343 representative feature in the TMP steel with a shorter (600s) holding time at 723 K (please see 344 Refs. [31, 32] by the same co-authors), and bainite formation was completed in that condition, it 345 could be suggested that the LSC formed by decomposition of residual austenite that was highly- 346 enriched in carbon (4-7at.%). Previous APT studies of RA in TRIP steels after TMP [25, 31] 347 indicate a wide range of carbon contents depending on its size, location and neighbouring 348 phases as well as the inhomogeneous distribution of carbon within RA. Thus, prolonged 349 exposure at 723 K enables carbon to continue diffusing from supersaturated BF into RA. 350 Consequently, regions close to BF are enriched in carbon; which triggers the formation of LSC in 351 thin RA layers between BF laths (Fig. 4b) or in part of a coarser RA (Fig. 4c). The part of the RA 352 that is less enriched in C remains stable on cooling to room temperature. These LSC closely 353 resemble the carbides observed by Sandvik [29] where LSC demonstrated extensive faulting 354 and lenticularity and their crystal structure was identified as triclinic or distorted 355 orthorhombic. All of these factors support the hypothesis proposed by Sandvik [29] that the 14 356 formation of such carbides takes place during the decomposition of RA by the shear mechanism. 357 Here it should be noted that a phase forming by a diffusion-controlled mechanism is expected to 358 contain a comparatively lower density of faults [29]. Moreover, it was also suggested by Sandvik 359 [29] that the high density of internal defects in the LSC could be inherited from the stress 360 accumulation in RA that is associated with BF formation. 361 Based on the above considerations, a proposal for the kinetics of the transformation events 362 occurring during the isothermal holding time is proposed. The nucleation of BF immediately 363 followed by the initiation of its growth, begins the process of carbon partitioning into residual 364 austenite. Since the distribution of carbon is inhomogeneous within and between differently 365 sized morphologies of RA, this phase could possess low, intermediate and supersaturated levels 366 of carbon. In turn, this leads to the RA behaving differently with increased isothermal holding 367 time. Here, the RA with the lowest carbon content easily transforms to martensite during 368 quenching whereas the RA with an intermediate carbon content remains stable at room 369 temperature. It follows then that the RA supersaturated with carbon decomposes, probably by 370 the shear mechanism, to form LSC and secondary ferrite. 371 In this case, the effect of increasing the isothermal holding time during coiling can be 372 summarized as having two effects. (i) It leads to an increase in the fraction of supersaturated 373 RA, which decomposes easily during subsequent holding. (ii) Alternatively, it increases the 374 overall stability of low carbon RA, which, in turn, inhibits martensite formation during 375 quenching. 376 The hypothesis that RA contains varying carbon content is strengthened by TEM investigations 377 of the samples after prolonged room-temperature storage. Imaging revealed the presence of 378 pearlite at BF/polygonal ferrite interfaces or in-between polygonal ferrite grains (Figs. 5a and 6 379 a-c). However, the EBSD map (Fig. 3c) revealed that not all of the RA decomposed into pearlite, 380 which also indicates the inhomogeneity in carbon content between and within the RA crystals. It 381 is interesting to note that if the area percentages of the red (0.18%) and aqua (6.43%) regions 15 382 are summed, the resulting 6.61% area fraction is similar to the original RA fraction estimated by 383 XRD after TMP. The decomposition of RA following the initial rapid formation of BF has only 384 been studied at elevated temperatures [23, 26, 28, 29]. The possibility of RA decomposing into 385 thermodynamically stable ferrite and carbides at room temperature has not been reported to 386 date. 387 Due to the high rate of carbon diffusion, even at RT, carbon atoms will continue re-arranging 388 with the formation of clusters and Cottrell atmospheres at dislocations in BF (Fig. 8c) competing 389 with carbon diffusion to RA where carbon has higher solubility compared to BF. Since APT 390 studies after TMP showed that carbon content in RA typically exceeds 3 at.% and can be as high 391 as 7 at.%, and taking into account that these values could have been increased during several 392 years at RT, equilibrium calculations were conducted in Thermo-Calc to assess the 393 thermodynamic driving force for pearlite formation from RA, Δ𝐺𝑝𝑒𝑎𝑟𝑙𝑖𝑡𝑒 , which is defined as: 394 Δ𝐺𝑝𝑒𝑎𝑟𝑙𝑖𝑡𝑒 = 𝐺𝑓𝑒𝑟𝑟𝑖𝑡𝑒 + 𝐺𝑐𝑒𝑚𝑒𝑛𝑡𝑖𝑡𝑒 − 𝐺𝑎𝑢𝑠𝑡𝑒𝑛𝑖𝑡𝑒 (1) with 𝐺 being the Gibbs energy of the phases. The values for 𝐺𝑓𝑒𝑟𝑟𝑖𝑡𝑒 and 𝐺𝑐𝑒𝑚𝑒𝑛𝑡𝑖𝑡𝑒 are under the 395 assumption that cementite and ferrite partition in para-equilibrium, i.e. only by diffusion of the 396 interstitial element C. The calculation was conducted on the intermediate composition of RA Fe- 397 4.80C-1.06Mn-1.40Si at.%. Fig. 9 shows the molar driving force for pearlite formation from RA 398 as a function of temperature with a step size of 50 K. Driving force and temperature show an 399 inversely proportional relationship in which the driving force increases from 1.0 to 5.5 kJ/mol 400 during cooling in the 500 K temperature interval. The computed driving force supports that the 401 transformation at low temperature is definitely thermodynamically feasible. The pearlite 402 property model in Thermo-Calc [41] was used to attempt an assessment of the experimentally- 403 observed pearlite formation kinetics. However, since hardly any experimental data on C 404 boundary diffusivity and interfacial mobility is available below 800 K, the calibration of the 405 model was optimized for the temperature range 800 – 1000 K. Thus, predictions of the kinetics 16 406 of pearlite formation at room-temperature require too much of an extrapolation and are not 407 reliable. 408 409 410 411 412 413 414 Figure 9. Molar driving force for pearlite formation, Δ𝐺𝑝𝑒𝑎𝑟𝑙𝑖𝑡𝑒 , from the intermediate 415 composition of retained austenite under the assumption of para-equilibrium between ferrite 416 and cementite as a function of temperature. 417 418 5. Conclusions 419 Lenticular shape carbides with high density of internal defects were found after isothermal 420 holding for 1800 s and after 16 years storage at room temperature. It was suggested that the 421 carbides were formed as a result of decomposition of residual austenite. This austenite is 422 located in close proximity to BF and enriched in carbon. Partial RA decomposition with pearlite 423 formation was observed in TRIP steel after 16 years storage at room temperature. This supports 424 the inhomogeneity in the distribution of carbon within and between RA grains based on their 425 morphology and location in a TMP TRIP steel. Furthermore, it posed the question regarding the 426 longtime RA stability under service conditions and its effect on component properties. Thermo- 427 Calc calculations support the feasibility of pearlite formation from the RA due to a high 428 thermodynamic driving force for its formation. 17 429 Acknowledgements 430 The authors would like to acknowledge the financial support of Deakin University and the 431 University of Wollongong, Australia. Deakin University’s Advanced Characterisation Facility is 432 acknowledged for use of the FIB-SEM, TEM, and APT. The JEOL JSM-7001F at the UOW-EMC 433 was funded by the Australian Research Council – Linkage, Infrastructure, Equipment and 434 Facilities grant LE0882613. The Oxford Instruments 80 mm2 X-Max EDS detector was funded 435 via the 2012 UOW Major Equipment Grant scheme. A. Gazder acknowledges the 2019 AIIM for 436 Gold - Investigator grant. 437 438 References 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 [1] B.C. De Cooman, Curr. Opin. Solid State Mater. Sci., 2004, vol. 8(3–4), pp. 285-303. [2] P.J. Jacques, Curr. Opin. Solid State Mater. Sci., 2004, vol. 8, pp. 259-265. [3] O. Matsumura, Y. Sakuma, H. Takechi, Scripta Metall., 1987, vol. 21, pp. 1301-1306. [4] I.B. Timokhina, P.D. Hodgson, E.V. Pereloma, Metall. Mater. Trans. A, 2004, vol. 35A(8), pp. 2331-2341. [5] E. Pereloma, H. Beladi, L. Zhang, I. Timokhina, Metall. Mater. Trans. A, 2012, vol. 43A, pp. 3958-3971. [6] F.G. Caballero, H.K.D.H. Bhadeshia, K.J.A. Mawella, D.J. Jones, P. Brown, Mater. Sci. Technol., 2002, vol. 18(3), pp. 279-284. [7] F.G. Caballero, H.K.D.H. Bhadeshia, Curr. Opin. Solid State Mater. Sci., 2004, vol. 8(3), pp. 251257. [8] F. Caballero, C. Garcia-Mateo, M. Miller, Mater. Sci. Technol., 2015, vol. 31, pp. 764-772. [9] J. Speer, D.K. Matlock, B.C. De Cooman, J.G. Schroth, Acta Mater., 2003, vol. 51(9), pp. 26112622. [10] D.V. Edmonds, K. He, F.C. Rizzo, B.C. De Cooman, D.K. Matlock, J.G. Speer, Mater. Sci. Eng. A, 2006, vol. 438-440, pp. 25-34. [11] L. Liu, B. He, G.-J. Cheng, H.-W. Yen, M.X. Huang, Scr. Mater. 2018, vol. 150, pp. 1-6. [12] X. Tan, D. Ponge, W. Lu, Y. Xu, X. Yang, X. Rao, D. Wu, D. Raabe, Acta Mater., 2019, vol. 165, pp. 561-576. [13] T. Furukawa, H. Huang, O. Matsumura, Mater. Sci. Technol., 1994, vol. 10, pp. 964-970. [14] D.W. Suh, J.H. Ryu, M.S. Joo, H.S. Yang, K. Lee, H.K.D.H. Bhadeshia, Metall. Mater. Trans. A, 2013, vol. 44A, pp. 286-293. [15] Y.-K. Lee, J. Han, Mater. Sci. Technol., 2015, vol. 31(7), pp. 834-856. [16] H.K.D.H. Bhadeshia, D. Edmonds, Metall. Trans. A, 1979, vol. 10, pp. 895-907. [17] H.K.D.H. Bhadeshia, Bainite in Steels: Transformations, Microstructure and Properties, Second ed. ed., Institute of Materials, Minerals and Mining, London, 2001. [18] E. Pereloma, A.A. Gazder, I. Timokhina, Retained austenite: transformation-induced plasticity, in: R. Colas, G.E. Totten (Eds.) Encyclopedia of Iron, Steel, and Their Alloys, CRC Press, New York, 2016, pp. 3088-3103. [19] E. De Moor, D.K. Matlock, J.G. Speer, M.J. Merwin, Scr. Mater.,2011, vol. 64(2), pp. 185-188. 18 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 [20] A.S. Podder, I. Lonardelli, A. Molinari, H.K.D.H. Bhadeshia, Proc. R. Soc. A, 2011, vol. 467, pp. 3141-3156. [21] W. Shi, L. Li, B.C. De Cooman, P. Wollants, C.-X. Yang, J. Iron Steel Res. Int., 2008, vol. 15, pp. 61-64. [22] C. Lerchbacher, S. Zinner, H. Leitner, Mater. Sci. Eng. A, 2013, vol. 564, pp. 163 - 168. [23] A.S. Podder, H.K.D.H. Bhadeshia, Mater. Sci. Eng. A, 2010, vol. 527(7), pp. 2121 - 2128. [24] C. Garcia-Mateo, F.G. Caballero, M.K. Miller, J.A. Jimenez, J. Mater. Sci., 2012, vol. 47(2) pp. 1004-1010. [25] Z.P. Xiong, A.A. Saleh, R.K.W. Marceau, A.S. Taylor, N.E. Stanford, A.G. Kostryzhev, E.V. Pereloma, Acta Mater., 2017, vol. 134, pp. 1 - 15. [26] H.-S. Park, J.-B. Seol, N.-S. Lim, S.-I. Kim, C.-G. Park, Mater. Des., 2015, vol. 82, pp. 173 - 180. [27] Y.K. Lee, H.C. Shin, Y.C. Jang, S.H. Kim, C.S. Choi, Scr. Mater., 2002, vol. 47(12), pp. 805 - 809. [28] F.G. Caballero, C. Garcia-Mateo, C. Garcia de Andres, Metall. Trans. A, 2005, vol. 46(3), pp. 581-586. [29] B.P.J. Sandvik, MTA, 1982, vol. 13(5), pp. 777-787. [30] I.B. Timokhina, P.D. Hodgson, E.V. Pereloma, in: E.F. Palmiere, M. Mahfouf, C. Pinna (Eds.) International Conference on Thermomechanical Processing: Mechanics, Microstructure & Control, Sheffield University, UK, 2003, pp. 115-121. [31] E.V. Pereloma, I.B. Timokhina, M.K. Miller, P.D. Hodgson, Acta Mater., 2007, vol. 55(8), pp. 2587-2598. [32] I.B. Timokhina, E.V. Pereloma, P.D. Hodgson, Mater. Sci. Technol., 2001, vol. 17(2), pp. 135140. [33] B.D. Cullity, Elements of x-ray diffraction, Addison-Wesley Publishing Company, Inc., Reading, MA, 1978. [34] A.A. Gazder, F. Al-Harbi, H.T. Spanke, D.R.G. Mitchell, E.V. Pereloma, Ultramicroscopy, 2014, vol. 147, pp. 114-132. [35] C. Sommer, C. Straehle, U. Koethe, F.A. Hamprecht, Ilastik: Interactive Learning and Segmentation Toolkit, IEEE International Symposium on Biomedical Imaging, doi:10.1109/ISBI.2011.5872394. ISBN 978-1-4244-4127-3, 2011, pp. 230–33. [36] M.K. Miller, R.G. Forbes, Atom-Probe Tomography: The Local Electro Atom Probe, Springer, New York, 2014. [37] J.O. Andersson, T. Helander, L. Höglund, P. Shi, B. Sundman, Calphad, 2002, vol. 26(2), pp. 273-312. [38] Y. Hirotsu, S. Nagakura, Acta Metall., 1972, vol. 20(4), pp. 645-655. [39] H.K.D.H. Bhadeshia, Acta Metall., 1980, vol. 28(8), pp. 1103-1114. [40] Y.C. Jung, Y. Ohomori, K. Nakai, H. Ohtsubo, ISIJ Int., 1997, vol. 37(8), pp. 789-796. [41] J.-Y. Yan, J. Ågren, J. Jeppsson, Metall. Trans. A, 2020, accepted. 506 19