Microstructural Evolution and Carbon Partitioning in Interstitial Free Weld Simulated API 5L X60 Steel

Abstract The microstructural characterizations and partitioning of carbon element in the weld heat affected zones of a commercial API 5L X60 line pipe steel were studied by applying a high speed heating and cooling dilatometry technique in the present research work. The hollow cylindrical specimens were quickly heated to 1000°C, soaked for only 5 s followed by continuous cooling to ambient temperature. Besides the construction of CCT diagram of this high strength low alloy steel using the dilatation data, the hardening response, microstructural features and carbon partitioning of weld simulated specimens were investigated. The obtained results showed that the hardening response of samples increased from 142 to 261HV10kg with increasing cooling rates. These hardening variations were attributed to the changes in microstructural features and carbon partitioning that occurred between the microconstituents present in the microstructures of weld simulated samples.


Introduction
Microalloyed high-strength low-alloy (HSLA) weldable steels have been developed over the last half-century and are widely used in applications such as offshore platforms, shipbuilding and the construction of pipelines for oil and gas transportation [1,2]. Oil and gas pipelines are the highpressure welded vessels, so safety and operational reliability are most important for these applications, especially for the high-pressure gas pipelines [3,4]. The materials best suited for the requirements of oil and gas pipeline industry are interstitial free advanced high strength low alloy steels. Specifications regarding chemical composition, mechanical properties and other important issues like welding, cutting, manufacturing etc. for these materials are determined and standardized by American Petroleum Institute: API [5], International Organization for Standards: ISO [6] and other national agencies working in this field of research work. API standards are generally taken as the main references by many national agencies in order to set their own specifications for these materials and hence are accepted and popularly used worldwide. Some of the most important materials used in pipelines are various generations of X-grade microalloyed high-strength weldable steels [7][8][9], which are well specified according to API 5L standard [5]. The main features of these steels include high yield strength [10,11], high toughness [12], good weldability [13] and low ductile-to-brittle transition temperature [14]. In order to further improve the weldability properties of these high strength steels, carbon content is usually kept low (≤0.1 wt.%) and the effect of reduction in strength and hardenability due to carbon reduction is compensated by adding manganese (between 1.4 to 1.9 wt.%) in addition to some minor amounts of strong carbide forming microalloyed elements such as Nb [15], Ti [16], V [17] and Mo [18]. These microalloyed elements, which lead to fine-grained microstructures, are responsible for simultaneous increase of strength, toughness and welding properties [19].  Investigation of welded HSLA steel microstructures show various heat affected zones (HAZ) including coarsegrained heat affected zone (CGHAZ), fine grained heat affected zone (FGHAZ), and inter-critical heat affected zone (ICHAZ). These regions form adjacent to the weld deposit as a result of being heated to temperatures ≥1200, 900-1200 and 700-900 ∘ C, respectively [20]. In addition, a great deal of researches directed on the toughness of welded parts indicated that FGHAZ and ICHAZ are far more tougher than CGHAZ in welded parts [21,22]. This is due to the formation of prior coarse-grained austenite [23] in the CGHAZ of welded steels, which subsequently transform to plate-like bainitic ferrite matrix associated with some martensite/austenite islands during cooling. The martensite/austenite islands in plate-like bainitic ferrite microstructures decrease the toughness of welded HSLA steels. Accordingly, grain refinement is the conventional method for improving the toughness of HAZ of these steels [24,25]. Although the microstructure of pipeline microalloyed steels mainly consists of upper bainite in HAZ, the amounts of martensite/austenite islands in their microstructures are considered as the principal factor in controlling their toughness and weldability. On the other hand, the toughness of HAZ in weldable high strength steels is extremely dependent on the percentages of carbon and other alloying elements which are effectively present in the martensite/austenite islands.
The investigation of microstructures and mechanical properties of practical HAZ of welded parts, such as oil and gas steel pipelines, is very expensive and time-consuming. Therefore, it is necessary to similar weld situations by using various thermal simulations processes [26][27][28][29] and then evaluate the hardening variations in relation with the amounts of different microconstituents in order to anticipate the quality of welded steel pipelines. One of the most applied techniques used for weld simulation is rapid heating and cooling dilatation of steel samples [26][27][28].
In this research, the microstructural features in relation with the hardening variations and macrohardness property of API 5L X-60 steel extensively used as pipelines in the gas and petroleum industry was investigated under various cooling conditions by using a high-speed heating and cooling automatic dilatometer. Moreover, the volume fractions of different microconstituents and carbon redistribution in the weld simulated microstructures were evaluated.

Materials and Experimental Procedure
The chemical composition of API 5L X-60 pipeline steel used in this research work is presented in Table 1. A Theta-model high-speed heating and cooling dilatometer was used in order to investigate in detail the phase transformation and microstructural changes during continuous cooling of weld simulated samples. The hollow cylindrical specimens with the inner and outer diameters of 3.5 and 4.9 mm, respectively; and 13 mm height were used for dilatometry study. These samples were heated rapidly to 1000 ∘ C, soaked for 5s in the vacuum furnace at 1.32×10 −7 atm (10 −4 torr) and then cooled to ambient temperature at various cooling rates from 0.3 to 360 ∘ C/s by helium gas injection. The heat-treated samples under various cooling rates or times (t8/5) are listed in Table 2. Typical investigated samples were marked as CC0.3, CC1, CC3, CC10, CC50, CC136 and CC360 corresponding to the samples continuously cooled at 0.3, 1, 3, 10, 50, 136 and 360 ∘ C/s, respectively. The heat treatment cycles performed in this work are schematically shown in Figure 1. The mechanical grinding and polishing procedures were carried out according to ASTM E3 standard (http://www.astm.org/standard/E3.htm), and then the samples were etched with a 2% nital solution (2 ml

HNO
3 and 98 ml C 2 H 5 OH) to reveal various microstructural features. Light and electron microscopic observations were carried out using an Olympus-PMG3 optical microscope and a TESCAN-MIRA 3 field-emission scanning electron microscope (FE-SEM), respectively. The micrographs were analyzed by image analyzing MIP software and the related volume fractions of different microconstituents were measured using the manual point count method according to the ASTM E 562 standard. In order to qualitatively investigate the carbon partitioning between the microconstituents, the spot-line scanning for carbon analysis was measured at various locations of polygonal ferrite, quasi-polygonal ferrite, degenerate pearlite, platelike bainitic ferrite and martensite/austenite islands by using energy-dispersive X-ray spectroscopy (EDS) technique. The Vickers macrohardness measurements with a load of 10 kg (HV 10kg ) were conducted on the weld simulated samples based on the ASTM E92 standard using an Instron Wolpert Gmbh-DLA 722 universal hardness tester.

Optical Microscopy and Hardness
Typical light micrographs taken from various weld simulated API X-60 samples are shown in Figure 2. The micrograph of CC0.3 samples (Figure 2a) indicates that the microstructure is mainly consisting of polygonal ferrite (light grains) with some carbide/ferrite dark contrasting regions called degenerate pearlite. It is interesting to point out that pearlite colony in plain carbon steels is a laminate microconstituent composed of alternating layers of ferrite and cementite formed by cooperative growth of these layers from prior austenite grain boundary at a relatively low cooling rate [30]; this is why degenerate pearlite might have formed by cementite nucleation at ferrite/prior austenite interfaces and then trapped with layers of ferrite in samples with very low carbon content under moderate cooling rates. Therefore, the degenerate pearlite formation in this sample is due to relatively slow cooling rate and very low carbon content (0.07%C). Further investigation of Figure 2a shows clearly that the polygonal ferrite grain boundaries are geometrically curved in the micrograph of CC0.3 sample. This is because of the reduction in carbon and iron diffusions due to the presence of considerable amounts of manganese and molybdenum elements in the alloy com- position. In addition, the cooling rate of 0.3 ∘ C/s is so low that neither coarse nor fine laminate pearlite microphases can be formed and hence the degenerate pearlite forms instead. The micrograph of CC10 sample is presented in Figure 2b, indicating that the microstructure is mainly composed of the polygonal/quasi polygonal ferrite matrix and some dispersed degenerate pearlite. It is further observed that the ferrite grain boundaries are not completely continuum in the micrograph of CC10 sample in comparison with that of CC0.3 ones. Moreover, the amounts of grain boundary ferrite and degenerate pearlite have been increased while the matrix ferrite grains are also finer in micrograph of CC10 sample due to the higher cooling rate. The resultant microstructure of the samples cooled at 50 ∘ C/s (Figure 2c) is majorly composed of plate-like bainitic ferrite, associated with superimposed martensite/austenite islands and some quasi-polygonal ferrite. The ferrite matrix became finer and no evidence of degenerate pearlite formation was observed in these samples. With increasing cool-ing rate, the volume fraction of polygonal ferrite decreased, while the blocks of martensite/austenite islands became finer and more elongated in the plate-like bainitic ferrite matrix. These results indicate that plate-like bainitic ferrite matrix with dispersed martensite/austenite islands have formed as a result of the mixed diffusional and shear phase transformation mechanisms. It is proposed that the platelike bainitic ferrite forms at moderate cooling rates when the diffusion of carbon atoms from ferrite/prior austenite interface is relatively high but not sufficient enough for cementite nucleation and growth [31]. As a consequence of plate-like bainitic ferrite formation, the carbon content of prior austenite increases so that the austenite to bainitic ferrite phase transformation ceases and therefore the austenite is entrapped within the bainitic ferrite matrix. Such austenite entrapment results in the formation of islands consisting of martensite and/or retained austenite microconstituent in the plate-like bainitic ferrite microstructure. Increasing the cooling rate to 360 ∘ C/s (Fig-ure 2d) causes the martensite/austenite islands to lose their blocky shape and become extremely fine and more elongated. Moreover, the plate-like bainitic ferrite is extensively observed in the needle-like form because of firstly the high cooling rates and secondly the decrease in carbon and iron diffusion rates. For more comparison, the volume fractions of various microconstituents developed in the various weld simulated samples are listed in Table 2. The results show that the polygonal/quasi polygonal ferrite volume fraction decreased from 91.5 vol% in sample CC0.3 to 86 vol% in sample CC10 while the degenerate pearlite volume fraction increased from 8.5 to 14 vol%, respectively. The mentioned decrease in polygonal ferrite volume fraction in conjunction with increased degenerate pearlite formation are the reasons for the increase in the mean macrohardness from 142 to 191 HV 10kg in the samples cooled from 0.3 to 10 ∘ C/s ( Figure 3). Further increase in cooling rate from 10 to 50 ∘ C/s, resulted in the activation of mixed shear and diffusional phase transformation mechanisms and converted some polygonal and quasi-polygonal ferrite to plate-like bainitic ferrite that was richer in carbon. As a result, the microstructure of the samples cooled at 50 ∘ C/s consists of martensite/austenite islands in plate-like bainitic ferrite matrix and has the higher mean macrohardness value of 215HV 10kg . Taking into consideration Figures 2c and 2d, it is clear that although it is not possible to measure the volume fraction of individual microconstituents present in the microstructure quantitatively but it is evident that more martensite/austenite islands have formed in the elongated plate-like bainitic ferrite matrix of CC360 sample in comparison with that of CC50 sample. As a result, the mean macrohardness measurement value for the CC360 sample is the highest and equals 261HV 10kg (Figure 3).

Figure 3:
The macrohardness results of various weld simulated API 5L X-60 pipeline steel samples at different cooling rates. For information and better comparison, the macrohardness of raw API 5L X-60 steel samples in normalized condition is 173HV10kg. Figure 4 shows typical SEM micrographs in secondary electron mode taken from various weld simulated API X-60 samples. In comparison with the relevant light micrographs shown in Figure 2, the microconstituents including ferrite grains, degenerate pearlite, and plate-like bainitic ferrite associated with martensite/austenite islands are seen with a better contrasting resolution in the microstructures. These SEM micrographs confirm firstly that the platelike bainitic ferrite has formed in the CC50 and CC360 samples (Figures 4b, 4c and 4d), while a mixture of polygonal and quasi polygonal ferrites has mainly formed with some degenerate pearlite in the CC0.3 and CC10 marked samples (Figures 4a and 4b). Secondly, the shape of plate-like bainitic ferrite crystals becomes more elongated and develops needle-like shape morphology with increasing cooling rate from 10 to 360 ∘ C/s as can be seen from the smaller sizes of martensite/austenite islands.

Electron Micrographs and Carbon Partitioning
Since the austenite to ferrite phase transformation accompanies with redistribution of carbon in the prior austenite, the EDS spot-line scanning was extensively used within various microphases in order to investigate the level of carbon partitioning in the microstructures. In this regard, Figure 5 illustrates typical electron micrographs with superimposed EDS spot-line scanning for the carbon analysis, showing the variation of carbon concentration in terms of EDS numbers (EDSN S ) within the polygonal ferrite, degenerate pearlite, plate-like bainitic ferrite and martensite/austenite islands developed in different weld simulated samples. For more comparison, the EDS spotline scanning results are also shown graphically in Figures 5a ′ to 5d ′ . This is worth noting that although the EDS spot-line scanning technique for the carbon analysis cannot be used to measure the carbon concentration quantitatively and precisely, it has been used by several investigators to qualitatively compare the level of carbon contents of different microconstituents [32][33][34]. Careful investigation of Figure 5a Figure 5b ′ also shows the same situation in CC10 samples for degenerate pearlite region whose average carbon concentration has been significantly increased to 13.25 EDSNs (S.D. = 2.40) while that of polygonal ferrite region is 6.11  EDSNs (S.D. = 1.12). This is due to the fact that the degenerate pearlite area contains some iron carbide which is enriched in carbon [30,35]. Further investigation of both Figures 5a ′ and Figure 5b ′ shows that the EDS number of carbons decreased from the central region of polygonal ferrite grains to the vicinity of ferrite/austenite interfaces. This can be attributed to the low migration or diffusion rate of carbon from the more carbon-saturated central regions of polygonal ferrite formed at early stage of phase transformation in comparison to that of lateral ferrite area developed at the interfaces of ferrite/austenite areas. The low rate of carbon diffusion from central region toward the interfaces is because of the presence of considerable amounts of alloying elements including Si, Mn, and Mo in the steel composition. On the other hand, the carbon diffusion from the vicinity of polygonal ferrite grain into its ferrite/prior austenite interface is much faster because the polygonal ferrite crystals are less compact and have more lattice defects near the ferrite/prior austenite inter-faces, causing the higher level of carbon redistribution and so more saturation in the regions nearby the prior austenite [36,37]. Therefore, the degenerate pearlite is nucleated and grown from carbon-saturated ferrite/prior austenite in the faces. Another point can be extracted from further investigation of EDS spot-line scanning for carbon analysis is that the average carbon EDS number of polygonal ferrite has been increased from 4.33 EDSNs (S.D. = 1.66) of CC0.3 marked samples to 6.11 (S.D. = 1.12) for CC10 ones. This can be explained by significantly less amount of carbon rejection occurred from the polygonal ferrite in to the remaining prior austenite area during diffusional phase transformation at higher cooling rate of 10 ∘ C/s. Moreover, the carbon partitioning between plate-like bainitic ferrite and martensite/austenite islands formed in CC50 and CC360 samples are compared in Figures 5c ′ and 5d ′ . In the first sight, it is observed that the average carbon EDSNs for martensite/austenite islands in CC50 marked samples is 12.25 EDSNs (S.D. = 0.81) while that one for bainitic ferrite  is 5.55 EDSNs (S.D. = 1.81). This means that the martensite/austenite islands held a significantly higher carbon concentration than that of bainitic ferrite. As already explained for the samples cooled at 0.3 and 10 ∘ C/s, the same situation existed here as well; i.e. when plate-like bainitic ferrite nucleated and grown from the prior austenite grain boundaries in the CC50 and CC360 marked samples, the carbon rejection from the bainitic ferrite into the prior austenite resulted in enhanced carbon saturation within the prior austenite and consequently these carbon enriched austenite areas consequently transformed to the martensite/austenite islands. On the other hand, when the average carbon EDSNs of martensite/austenite islands in CC360 samples (Ave. = 8.75 EDSNs & S.D. = 4.49) are compared with that of CC50 ones (Ave. = 12.25 EDSNs & S.D. = 0.81), it is observed that the carbon content of martensite/austenite islands has decreased significantly with the increase in cooling rate from 50 to 360 ∘ C/s due to the less amount of rejection of carbon from bainitic ferrite into the prior austenite areas. In addition, when the average EDSNs for plate-like bainitic ferrite in CC50 and CC360 samples are compared with those of polygonal ferrite formed in the CC0.3 and CC10 samples, it is obviously seen that the carbon concentration of bainitic ferrite increased because of lesser times to diffuse from bainitic ferrite to the remaining austenite in the higher continuous cooled samples. This can be shown by the dilatation data ( Figure 6) provided in Table 3. As is seen, the carbon diffusion and partitioning time between bainitic ferrite and remaining prior austenite has been decreased from seven to one second when the cooling rate has been increased from 50 to 360 ∘ C/s. The experimented results show that the carbon concentration of martensite/austenite islands in CC50 samples are much higher than that of CC360 ones (Figures 5c ′  and d ′ ). This result shows that the carbon partitioning between different microconstituents has occurred which is quite variable during phase transformation even under rapid cooling rates of 50 and 360 ∘ C/s. On one hand, martensitic transformation start temperature depends on the concentration of carbon and other alloying elements present in the associated prior austenite areas. On the other hand, the carbon rejection to prior austenite has increased the carbon content of the austenite regions near to the prior austenite/bainitic ferrite interface in com-   .3  1000  780  685  490  370  -1365  CC1  300  778  642  480  365  -410  CC3  100  760  639  470  355  -135  CC10  30  745  595  460  345  -41  CC50  715  555  --375  7  CC136  703  494  --375  3  CC360  620  480  --375  1 parison with the interior region of prior austenite. This causes the central regions of prior austenite to transform to martensite during cooling at a higher temperature, due to lower carbon content and hence the higher martensite start temperatures. In addition, the regions nearby to austenite/bainitic ferrite interfaces which have higher carbon content may be transformed to martensite with a relatively low temperature. In fact, during the martensite transformation in central regions of unstable prior austenite, the regions near to austenite interface remains untransformed and needs lower temperatures for martensite formation due to higher carbon content. Moreover, the high dislocation densities in the semi-stable austenite interface layer which are formed due to the high volumetric and martensitic-phase-transformation shear strains from the austenite central regions, help the interfacial austenite to get stable and, therefore further interfacial martensitic formation is ceased [38]. Eventually, the mentioned happenings result in the formation and dispersion of martensite/austenite islands in plate-like bainitic ferrite matrix in high cooled samples.

Weld Simulated CCT-Diagram
Typical Expansion-Temperature and Temperature-Time curves of various weld simulated API X-60 pipeline steel samples taken from a high-speed heating and cooling dilatometer are shown in Figure 6. As seen, all the curves drawn are cooled from 1000 ∘ C austenitizing temperature and their contraction-time dependence is linear to room temperature. This indicates that the slope change of dilatometry curves (length variation or expansiontemperature curves) has been related to the phase transformation due to the phase volume change. The investigation of dilatometry cooling curves for CC0.3 and CC10 samples shown in Figures 6a and 6b reveals that both of the samples have undergone a double stage phase transformation because of double line slope changes in these curves. The first slope change in Figure 6a, for the samples CC0.3 starts and finishes at 780 and 685 ∘ C and the second one at 490 and 370 ∘ C, respectively. While the first slope change for CC10 marked samples in Figure 6b, begins and ends at 745 and 595 ∘ C, and the second one at 460 and 345 ∘ C, respectively. Integrating these results with microscopic explanation mentioned in section (3.2), it can be clearly understood that the first and second stage of phase changes on dilatometric curves are related to ferrite (either polygo- nal or quasi-polygonal) and degenerate pearlite formation, respectively. Moreover, the results of Figure 6a show the carbon-saturated polygonal ferrite formation from prior austenite in CC0.3 sample started at 780 ∘ C (time 617 s) and finished during 321 seconds at 685 ∘ C (time 938 s). From this point afterwards the carbon-saturated polygonal ferrite and remaining prior austenite coexist and during further cooling till 490 ∘ C (time 1600 s), the carbon is rejected from polygonal ferrite and diffuses into the prior austenite making it richer in carbon and ready for transformation to degenerate pearlite. Then, the second phase change i.e. the degenerate pearlite formation begins at 490 ∘ C (time 1600 s) and ends during 380 s at 370 ∘ C (time 1980 s). The same situation exists for CC10 samples (Figure 6b). The carbon-rich polygonal and quasi-polygonal ferrite formation started at 745 ∘ C (time 108 s) during 14 s and finishes at 595 ∘ C while the degenerate pearlite was formed at 460 ∘ C (time 136 s) and completed at 345 ∘ C (time 149 s) in 13 seconds. Comparing the phase transformation temperatures of CC0.3 and CC10 samples (Table 3) including the polyg-onal ferrite start temperature (Ar 3 ), the polygonal ferrite finish temperature (Ar 1 ), the degenerate pearlite start temperature T DP S and the degenerate pearlite finish temperature T DP f obviously shows they are shifted to lower temperatures due to increased cooling rate. In addition, considering the diffusion duration for carbon partitioning into prior austenite which starts at Ar 3 and finishes at T DP f indicates that the carbon diffusion time is decreased from 1365 s in CC0.3 to 41 s in CC10 samples (Table 3) resulting in higher carbon content of polygonal/quasi-polygonal ferrite formed in CC0.3 samples.
Two slop changes can also be seen on dilatometry cooling curves taken from 50 to 360 ∘ C/s weld simulated samples (Figures 6c and d) which are related to the plate-like bainitic ferrite and the martensite formations. Interpretation of these dilation results together with the microstructural features presented in part (3.2) demonstrates that the first slope change in Figure 6c curve is attributed to plate-like bainitic ferrite formation which starts at 715 ∘ C (time 4 s) and completes after 4 s at 555 ∘ C (time 8 s). The second slope change is attributed to martensite formation which starts at 375 ∘ C (time 11 s). Accordingly, the investigation of Figure 6d indicates that the plate-like bainitic ferrite has formed in CC360 samples at 620 ∘ C (time 0.5 s) and finished only in 0.5 s at 480 ∘ C (time 1 s). From this point afterwards, further cooling down to 375 ∘ C (time 1.5 s), caused the carbon rejection from plate-like bainitic ferrite to prior austenite and making the central regions of remaining prior austenite ready for martensitic phase transformation. This figure clearly shows that the martensite formation starts in nearly 1.5 s after cooling from 1000 ∘ C austenitizing temperature. Based on these results, the continuous cooling transformation (CCT) diagram for API 5L X60 steel was drawn and presented in Figure 7 by using the data obtained from dilatometry cooling curves and the microstructural observations.