Low alloy Cr-Mo-V steels are extensively used for high-temperature and high-pressure applications in petrochemical industries owing to their excellent elevated temperature performance and good resistance to oxidation and hydrogen embrittlement [1]. The conventional heat treatment of low alloy Cr-Mo-V steels consists of quenching after austenitization and high temperature tempering. Generally, the impact toughness of the quenched and tempered steels increases with increasing the cooling rate while quenching. However, low-alloy Cr-Mo-V steel components with heavy cross-section are necessary for larger pressure vessels construction in future petrochemical refining plants with higher process efficiencies. Because of the heavy cross-section, lath bainite and/or martensite are obtained at the surface of the component during quenching, while a large amount of coarse granular bainite (GB) may appear in the interior region. Previous research [2] has shown that an excellent combination of strength and toughness can be achieved by appropriate high temperature tempering of low alloy Cr-Mo-V steels with lath bainite or martensite, compared to the GB microstructure. This is due to the GB usually containing massive M-A constituents and equiaxed bainitic ferrite matrix. These massive M-A constituents decomposed into ferrite and coarse carbides and/or agglomerated carbides during the tempering process, which has an adverse effect on impact toughness [[3], [4], [5]]. Meanwhile, compared to those of lath bainite and martensite matrix, the equiaxed bainitic ferrite has a larger effective grain size, which has a relatively low ability to inhibit crack propagation [6,7]. Therefore, it seems that a remarkable improvement in toughness could not be achieved only through optimization of the high temperature tempering process. Modification of quenched bainitic microstructure is an essential method to obtain superior microstructure and mechanical properties after appropriate high temperature tempering. In other words, the impact toughness after high temperature tempering may be improved by decreasing the volume fraction of the M-A constituents, the proportion of large-sized M-A constituents, and the effective grain size of bainitic ferrite matrix in the quenched microstructure.

It is widely accepted that the M-A constituents within GB are transformed from carbon-rich austenite during the cooling [8,9]. Therefore, the formation of bainitic ferrite and M-A constituents in GB is closely related to the micro-segregation of carbon in the undercooled austenite. Some studies have shown that the morphology and volume fraction of M-A constituents and the effective grain size of bainitic ferrite could be adjusted by controlling the bainite transformation temperature, quenching rate, and amount of alloying elements. Shin et al. [10] have reported that the finish cooling temperatures and cooling rates affect the formation of GB, and found that the volume fraction and size of the M-A constituents, as well as the effective grain size of bainitic ferrite, tended to decrease with decreasing finish cooling temperature or increasing cooling rate. Bonnevie et al. [11] have investigated the formation of M-A constituents in coarse-grained heat-affected zones and intercritical reheated heat-affected zones of structural steels, found that silicon increased the density of M-A constituents and the proportion of massive M-A constituents, and favored their formation over a wider range of temperatures. Baker et al. [12] have found that the addition of a minor vanadium (0.05% V) reduced the size and areal fraction of the M-A constituents and improved toughness of the intercritically reheated coarse grained heat affected zone (IC CGHAZ) in low carbon microalloyed steel. However, although the addition of 0.11% V or 0.03% Nb refined the bainitic ferrite matrix to some extent, the larger mean size and areal fraction of the M-A constituents resulted in the deterioration of the IC CGHAZ toughness.

In recent years, as the cleanliness of steels has been greatly improved, RE have been realized as crucial micro-alloying elements. It has been reported [[13], [14], [15], [16]] that the addition of RE also has a remarkable effect on the bainitic transformation behavior, especially in low carbon low alloy steels. Liang et al. [17,18] have indicated that RE elements decrease the interfacial energy of the grain boundaries and hinder the diffusion of carbon atoms, which reduces the carbon micro-segregation in undercooled austenite, and inhibits the formation of equiaxed GB; the microstructure of bainitic ferrite and the substructures were also refined. In a previous work, we also found the addition of RE could decrease the onset temperature of diffusional phase transformation in low carbon plain steel and low carbon low alloy steel remarkably, and attributed the changes to the hinder effect of RE on the carbon diffusion according to the first principle calculation results [16]. Bai et al. [19] have reported that RE plays a vital role in retarding the micro-segregation of carbon during bainitic transformation in low carbon high-Nb steels, thus reducing the grain size of bainitic ferrite from 8.78 to 4.85 μm and refining the M-A constituents to obtain a more dispersed distribution. While RE solubility is very low in steels, as reported in Ref. [20,21] (in fact the accurate measurement of RE solubility in steel is an unsolved problem up to now), excess addition of RE leads to the formation of large number of large RE-containing inclusions, which are extremely detrimental to the impact toughness.

Accordingly, it is anticipated that the addition of a small amount of RE may enhance impact toughness of GB through its alloying effect. Therefore, in the present study, we explored the possibility of obtaining optimal impact toughness by using an appropriate amount of RE added to the low alloy 2.25Cr-1Mo-0.25 V steel. Five types of low-alloy 2.25Cr-1Mo-0.25 V steels with RE contents varying between 0 and 0.048 wt.% were fabricated using a vacuum smelting furnace. The effects of RE content on transformation behavior and microstructure were analyzed in detail, and their effects on impact toughness were systematically discussed.

Five 20 kg ingots of low alloy 2.25Cr-1Mo-0.25 V steels with the total RE content varying from 0 to 0.048 wt.% were produced by vacuum-induction melting. The chemical compositions of these steels are listed in Table 1. An advanced vacuum smelting technique was used to reduce the levels of tramp elements (P, S, O, etc.), which are essential to the micro alloying effects of RE [16]. The ingots were subsequently reheated to 1150 °C and hot forged into rods 40 mm × 40 mm. After that, 60 mm × 40 mm × 13 mm steel blocks were machined from the forged bars by wire electrical discharge machining for subsequent heat treatments. The heat treatment consists of normalizing at 940 °C for 2 h and tempering at 700 °C for 4 h, which is the procedure commonly adopted in manufacturing large forgings in the practical industry.

To evaluate the effects of RE addition on the tensile properties and impact toughness, tensile test samples were prepared along the forging direction and the standard Charpy V-notch impact test samples were prepared with the notch perpendicular to the forging direction, respectively. The tensile tests with sample ofφ5 mm × 30 mm in gauge size were conducted at ambient temperature in a Zwick Z050 tensile testing machine at a crosshead speed of 3 mm⋅min^-1. The Charpy V-notch impact test were conducted in the temperature range from -120 °C to 20 °C in a Zwick RKP-450 instruments drop weight impact tester. In order to reduce errors in the data interpretation, a regression analysis for absorbed impact energy vs. test temperature was conducted with a hyperbolic tangent curve fitting method [22]. The regression curves were used to measure the upper shelf energy (USE), ductile to brittle transition temperature (DBTT). The DBTT is regarded as the temperature corresponding to the average value of the USE and lower shelf energy.

In order to investigate the effects of different RE contents on bainitic transformation temperatures, dilatometric tests were perform on the five kind of steels to examine the temperatures of the onset and end points of bainite transformation by an induction heating dilatometric machine (Linseis RITA L78). Solid samples of φ3 mm × 10 mm were used for these tests. The samples were heated at a rate of 1 °C /s, austenized at 940 °C for 20 min, and then continuously cooled to room temperature at a rate of 3 °C/s. This cooling rate is about average cooling rate from 940 °C to 300 °C for the normalized samples to cool in air.

Metallographic samples were prepared using the conventional metallographic techniques. Quantitative analyses of inclusions and prior austenite average grain sizes of the steels were performed by using a ZEISS AXIOVERT optical microscope. Microstructural examinations were performed with a FEI Inspect F50 scanning electron microscopy (SEM). In order to identify the characteristics of M-A constituents of the steels, the metallographic samples were also etched using Lepera reagent (mixture of 4% Picral +1% sodium metabisulfite solution in 1:1 ratio) for 20 s A previous study [23] has indicated that the highlight in the M-A constituents is attributed to the enrichment of austenite stabilizers, which leads to lighter etching than lean alloying elements bainitic ferrite. Thus, the M-A constituents appear white, while bainitic ferrite is gray after etched by Lepera reagent. The observations were made using Olympus OLS4000 confocal laser scan microscope (CLSM). The characteristics of M-A constituents were quantitatively examined by using metallographic image analysis software Image-pro plus 6.0. In order to obtain information on the chemical composition of the M-A constituents, electron probe microanalysis (EPMA) was used to analyze the distribution of alloying elements in different M-A constituents of the samples with a JEOL JXA-8530 F system operated at 10 kV and 50 mA in spot mode. The effective grain size of the samples with different RE contents was analyzed by electron backscatter diffraction (EBSD) method. EBSD was performed on the FEI Inspect F50 SEM with a step length of 0.5 μm, and the data were interpreted using the Oxford Instruments Channel 5HKL program package. In this study, a crystallographic misorientation higher than 15° was defined as an effective grain boundary of the low alloy Cr-Mo-V bainite steels. Because it can effectively inhibit or arrest crack propagation [10]. The amount of retained austenite in the samples was estimated by X-ray diffraction (XRD) analysis; the details pertaining to this technique can be found in the work of Huda [24]. Thin foil specimens for a Tecnai F20 transmission electron microscope (TEM) observation were prepared with a twin-jet electropolisher at 18 V using a solution of 10 vol.% perchloric acid and 90 vol.% ethanol at -25 °C.

The mechanical properties tests were performed on the normalized and tempered samples with RE contents of 0˜0.048 wt.%. The results of the tensile test are listed in Table 2. The ultimate tensile strength (UTS), yield strength (YS) and elongation (A) are in the ranges 778-788 MPa, 676-684 MPa, and 20.5-21% for these steels, respectively. This reveals that the strengths and elongation of these samples remain approximately constant. In contrast to the relatively small influence of the RE additives on tensile properties, the impact behavior changes markedly with the addition of a small amount of RE into the reference (denoted as Ref. in the figures and tables). The upper shelf energy (USE) and DBTT values are also listed in Table 2 and Fig. 1. The USEs of all the samples are higher than 230 J; the USE of the 0.012RE steel sample is slightly higher than those of the other four samples. The DBTT decreases with increasing RE content, and reaches the lowest value in 0.012RE steel, and finally increases with RE content. The optimum impact properties were obtained in the 0.012RE steel. Notably, in 0.048RE steel, the value of the USE was the lowest at 230 J, and the value of the DBTT was the highest at -24 °C.

The volume fractions, average diameters, aspect ratios, and densities of inclusions for the investigated five steels are given in Table 3. The average aspect ratios of the inclusions range from 1.36 to 1.38 for five steels, suggesting that they are almost spherical in all the cases. When the RE content is lower than 0.012 wt.%, the volume fraction and density of the inclusions remain low level and have not an obvious change with increasing RE content. However, the addition of RE in excess of 0.012 wt.% leads to a marked increase in the volume fraction, density and average size of inclusions. Particularly, for the 0.048RE steel, the average size and density of the inclusions markedly increases from 1.70 to 2.56 μm, 68.2 to 143 mm^-2, respectively, compared to the reference steel. The characteristics of the typical inclusions observed by SEM and/or TEM and the corresponding results of EDS analysis are presented in Fig. 2, Fig. 3, Fig. 4. The inclusions in reference steel are primarily Al-Mn compound oxides and/or sulfides mainly containing Mn, Al, S, O and Fe, as shown in Fig. 2(a)-(d). As for the 0.012RE steel, some of the inclusions containing RE elements were detected by TEM-EDS analysis. A TEM image of such a particle is shown in Fig. 3(a), and the particle contains cerium, lanthanum, sulfur and oxygen, as determined from EDS (see Fig. 3(c)). Structural investigation was also carried out with the help of the selected area electron diffraction (SAED) technique (Fig. 3(b)), and the dark particle observed was identified as RE₂O₂S base on TEM-EDS and SAED analysis. When RE content is 0.048 wt.%, except for the RE oxysulfides, a few larger sized particles (3-5 μm) were also observed in the 0.048RE steel, as shown in Fig. 4(a) and (b). The results from the EDS analysis, as shown in Fig. 4(c), are typical of most of these particles. They indicated that the particle was a carbon-enriched inclusion.

Comparing the changes in the areal fraction and size of the inclusions with the variations in the Charpy impact properties for all the cases, it is evident that the obvious improvement in the Charpy impact properties upon adding 0.012 wt.% RE to the 2.25Cr-1Mo-0.25 V steel cannot be explained by the changes in the area fraction and size of the inclusions alone. The authors also attempted to measure the change in prior austenite grain size in all the normalized samples. However, the average prior austenite grain sizes for the five samples are almost the same, and range from 31.3 to 34.6 μm, as listed in Table 3. In order to explain the reasons for this observation, the effect of RE addition on the transformation and microstructure will be discussed as follows.

The ΔL-T curves of the investigated five steels obtained at the cooling rate of 3 °C/s are shown in Fig. 5. It can be seen that the ΔL values as a function of T are altered by the change in the amount of added RE. The influence of RE addition in term of the variation in the onset and end points of the bainitic transformation (marked as B_s and B_f, respectively) is also included in the inserted table of Fig. 5. It is obvious that the onset of transformation in the five samples occurred in the medium-temperature range (536-573 °C), above those of upper bainite. Therefore, the B_s of the five samples should be identified as the onset points of GB transformation instead of that of pearlite or upper bainite. It is found that the onset temperature of bainitic transformation decreases obviously with the increase in RE addition when the amount added is lower than 0.020 wt.%. For instance, the addition of 0.012 wt.% RE results in an obvious decrease (˜33 °C) compared to the reference steel. On the other hand, an abnormal phenomenon is also observed when the RE content is increased from 0.020 to 0.048 wt.%: the B_s almost keeps as a constant and even increases slightly from 540 °C to 546 °C. Meanwhile, it is noted that the RE additives in the amounts investigated have virtually no influence on the variation in the end points of the bainite transformation in the present experiment, which is well in agreement with the results of Gao et al. [9] that B_f of GB is unrelated to the quenching rates or microstructure types of GB.

Additionally, it needs to be stressed that the average prior austenite grain sizes in all five samples are similar, which allows us to compare properly the effect of RE alloying elements without the influence of austenite grain size on continuous cooling transformation kinetics in the present study [25]. This is apparently a result of the fact that not all of the RE added to the steels interact with the sulfur, oxygen, and phosphorus, resulting in the formation of non-metallic inclusions. Therefore, a certain amount of RE should exist in the solid solution in these RE-bearing steels [16].

3.4.1. Normalized microstructures

The SEM images of the steels with different RE contents are shown in Fig. 6(a)-(e). The normalized microstructures of the samples are GB, basically composed of bainitic ferrite and M-A constituents. Based on the characteristics of bainitic ferrite and M-A constituents, GB was classified into two types, namely GB₁ and GB₂. As shown in Fig. 6(a), in GB₁ region, equiaxed bainitic ferrite formed inside prior austenite grains and decorated randomly with massive M-A constituents. GB₂ can be classified differently from GB₁ by the presence or distribution of the M-A constituents and morphology of bainitic ferrite; GB₂ contains elongated M-A constituents and has well-developed lath substructures inside the bainitic ferrite matrix. These massive M-A constituents are mostly located at the prior austenite grain boundaries and randomly distributed in the regions of equiaxed bainitic ferrite, while the M-A constituents in stringer form are distributed parallel to each other, mainly at the lath bainitic ferrite boundaries. The massive M-A particles were typically 2-5 μm in size, whereas the elongated M-A particles were 1-5 μm in length and 0.3-1 μm in width.

In order to distinguish accurately bainitic ferrite and the M-A constituents, the samples were further etched in Lepera solution [26]. The CLSM images of the samples are shown in Fig. 7(a)-(e). Bainitic ferrite and the M-A constituents are colored gray and while, respectively. Based on the above-mentioned microstructural features observed by SEM and CLSM, the volume fractions of GB₁, GB₂, and M-A constituents were measured separately by using Image-Pro Plus software, and the statistical results are listed in Table 4. The results show that only a small amount of GB₂ existed in normalized microstructure of the reference steel. The volume fractions of GB₁, GB₂, and M-A constituents are about 81.2, 18.8 and 9.6%, respectively. The volume fraction of GB₂ in 0.012RE steel is much more than that in the reference steel, while the volume fractions of the M-A constituents and GB₁ are significantly lower than those of the reference steel. Furthermore, almost no massive M-A constituent along the prior austenite grain boundaries is observed, and the dot-shape and elongated M-A constituents relatively uniformly distributed in the bainitic microstructure of the 0.012RE steel, compared to that of the reference steel. Additionally, it should be noted that the volume fractions of the M-A constituents, GB₁, and GB₂ do not show obvious changes when the content of RE further increases from 0.020 wt.% to 0.048 wt.%. This corresponds well with the above observation that the volume fractions of GB₁ and M-A constituents decrease with decreasing onset temperature of GB transformation.

The two types of M-A constituents in GB were further characterized by EPMA. Fig. 8(a) shows a typical massive M-A constituent at a prior grain boundary in the GB₁ region of the normalized sample of the reference steel. The alloying elements redistribute among the M-A constituents and bainitic ferrite was measured qualitatively by EPMA line-scanning. The distribution of carbon along a horizontal red line is shown in Fig. 8(c). It can be obviously seen that the concentration of carbon is much higher in M-A constituents than in bainitic ferrite matrix. Some studies argue that Mn, Ni, Cr, etc. might facilitate the formation of M-A constituents [23]. However, enrichment of these elements in M-A constituents was not observed in this study. These results strongly support the fact that the partition of elemental carbon can occur alone since there was insufficient time for the diffusion of substitutional elements during bainitic transformation. The concentration distributions of carbon in two types of M-A constituents have been plotted in Fig. 8(c) and (d) to provide the quantitative information of carbon partition affected by the morphology and distribution of the M-A constituents. Compared with the elongated M-A constituents that exist in the GB₂ region (Fig. 8(b) and (d)), it seems that the massive M-A constituents in the GB₁ region contain a much higher concentration of carbon (about 1.5 wt.%) than the elongated M-A constituents in the GB₂ region (about 0.7 wt.%). These characteristics suggest a higher degree of carbon partition during the formation of GB₁ compared to GB₂. This is highly consistent with the previous research of Okada et al. [27], where massive M-A constituents tended to form at a lower quenching rate and/or higher temperature region of bainitic transformation. The massive M-A constituents contained higher carbon concentrations, and therefore, more twin-type martensite and retained austenite than the elongated M-A constituents.

In order to explain the changes in the microstructures due to RE addition, EBSD and XRD were also performed in the present study. The crystallographic characteristics of the normalized microstructures of the reference and 0.012RE steels were further analyzed by EBSD (Fig. 9(a) and (b), respectively). It is widely accepted that the threshold value of high misorientation boundary is usually defined at 15° because up to this value it can deflect and/or arrest cleavage crack propagation. In terms of the results of EBSD analysis, although the average prior austenite grain size did not change significantly due to the addition of 0.012 wt.% RE, it can be seen from Fig. 9(a) and (b) that the 0.012RE steel contained more lath substructures than the reference steel. Consequently, the effective grain sizes of the reference and 0.012RE steel measured are about 6.74 and 4.92 μm, respectively. Additionally, Fig. 9(c) and (d) show the distributions and morphologies of the retained austenite of the two steels, in which red color corresponds to the retained austenite phase with FCC structure. Within the limits of detection, the 0.012RE steel presented a more homogeneous and smaller sized retained austenite than that of the reference steel. It is also seen that the small dot-shaped retained austenite is uniformly distributed in the bainitic ferrite matrix (Fig. 9(d)). By contrast, the reference steel has relatively massive, as well as small-dot shape retained austenite. This is good agreement with the results of the SEM observation and a previous study on the morphology and distribution of the M-A constituents [28]. The explanation for this result is a higher level of carbon enrichment in the massive M-A constituents relative to that in the elongated M-A constituents, as the results of EPMA revealed earlier, resulting in more retained austenite remaining in the M-A constituents at ambient temperature due to high chemical stability. It can be reasonably speculated that most of the massive M-A constituents have twinned martensite structures based on our previous observations [28].

Some of the small-sized retained austenite is not able to be accurately detectable by EBSD using the scanning step of 0.5 μm because the scanning step may be much larger than the width of remained austenite. Therefore, the volume fractions of retained austenite were further quantitatively studied by using XRD analysis. Fig. 10 shows the XRD spectra for the normalized samples with different RE contents, and the volume fractions of retained austenite from XRD analysis are listed in Table 4. For the reference steel, the volume fraction of retained austenite was about 5.4 vol.%. By adding 0.007 and 0.012 wt.% RE to the 2.5Cr-1Mo-0.25 V steel, the volume fraction of retained austenite decreased dramatically to 4.8 and 3.1 vol.%, respectively. However, the volume fraction of retained austenite revealed only a very small difference between the 0.020RE and 0.048RE steels (2.6 and 2.5 vol.%, respectively).

3.4.2. Decomposition of M-A constituents after high-temperature tempering

Fig. 11 shows the TEM images of the GB microstructures before and after tempering at 700 °C for 4 h. Two types of M-A constituents distributed in bainitic ferrite matrix before tempering were detected in reference steel (Fig. 11(a) and (b)). Fig. 11(a) shows a typical massive M-A constituents distributed at the triple junction of the prior austenite grain boundaries in the GB₁ region. By tilting the TEM foil sample, twinned substructures within martensite can be clearly observed (marked A in Fig. 11(a)). It is widely accepted that the twinned martensite structure in the massive M-A constituents is associated with the high carbon content resulted from carbon partitioning during GB transformation [29]. On the other hand, the elongated M-A constituents, which are mainly distributed in lath bainitic ferrite from the GB₂ region (Fig. 11(b)), contain a high density of dislocation (marked B in Fig. 11(b)). After tempering, the massive M-A constituents were decomposed into ferrite and coarse agglomerated carbides in the GB₁ region. Those carbides mainly are striped M₇C₃ and ellipsoid M₂₃C₆ type carbides, which is consistent with the results of our previous research [30]. For the GB₂ region, the decomposition of the elongated M-A constituents resulted in the formation of discontinuous carbides along the lath ferrite interfaces (as shown in Fig. 11(d)) of the GB₂ region. Meanwhile, although lath bainitic ferrite with low-angle boundaries merged and recrystallized into polygonal ferrite to some extent under the driving force of the energy stored in the high density of dislocations, most of the lath bainitic ferrite interfaces can still be clearly observed in the GB₂ region. Compared with tempered GB₁, except for a large number of lath substructures existing in the bainitic ferrite of the GB₂ region, it is obvious that more homogeneous and finer carbides distributed in bainitic matrix. This is because a higher density of substructures and lower degree of carbon micro-segregation in the GB₂ region will provide more sites for carbide nucleation and lower the kinetics of carbide growth, respectively.

It is widely accepted that RE elements can exist in steels in three different forms: inclusions, solid solutions and intermetallic phases. Generally, the RE will be largely consumed by forming highly stable sulfides, oxi-sulfides, and oxides when the total amount of RE is very low or high amount of S and O exist in the steel [31]. It has been reported that modified non-metallic inclusions can trigger grain refinement and improve the deformation compatibility of inclusions and the surrounding matrix, therefore, they have been widely used for obtaining good mechanical properties over the past several decades. In our previous research, the improvement in the compatibility of inclusions and the surrounding matrix by adding RE has also been confirmed [16]. Except for the non-metallic inclusions, which are almost the same in 0.007RE and 0.012RE samples, the effect of RE addition on bainitic transformation and the mechanical properties will be discussed in the following section. Additionally, the solution of RE in ferrite steels is very limited. When the amount of RE added is in excess, apart from the small amount of RE solution that existed in the steels, it is assumed that the precipitation of RE-enriched carboxides and/or carbides with increasing RE content occur (RE enriched carboxide particles can be found in 0.048RE, as shown in Fig. 4).

As stated above, the microstructures of the normalized steels are typically GB composed of bainitic ferrite and M-A constituents. In order to determine the effects of RE on GB transformation, the formation of GB is introduced briefly as follows. The component Mo is an important element in low-alloy 2.25Cr-1Mo-0.25 V steels that, has an obvious effect in delaying the transformation of the undercooled austenite to pearlite in the high-temperature region. Therefore, the undercooled austenite of the low-alloy 2.25Cr-1Mo-0.25 V steels would not be directly decomposed into the pearlite in the relatively higher temperature range during the continuous cooling process. As the temperature decreases in air cooling condition, there is insufficient time for the diffusion of substitutional elements, mainly Ni, Mn, Cr, and Mo, whereas carbon diffusion in undercooled austenite is still relatively quick, resulting in the formation of carbon-depleted and carbon-rich regions in prior austenite before bainitic transformation occurs. When undercooled austenite is cooled below the onset temperature of GB transformation, the equiaxed and/or lath bainitic ferrite nucleate in those carbon-depleted regions and grow separately, accompanied by carbon further diffusing into the adjacent austenite along the migrating bainitic ferrite/austenite interface. With bainitic ferrite growing, meeting, and merging with each other, a fraction of lath bainitic ferrite is merged into coarse equiaxed ferrite and the volume fraction of the undercooled austenite reduces gradually, resulting in the formation of carbon-rich under-cooled austenite islands. During the cooling sequence, the carbon-rich austenite islands will be transformed into martensite and/or austenite (i.e. M-A) constituents depending on their mechanical and chemical stability. It is can easily concluded that the shape and characteristics of bainitic ferrite and the M-A constituents are closely linked to the continuous cooling rate, bainite transformation temperature and chemical composition, which mainly depend on the diffusion of Fe and C atoms. Based on the above comprehensive analysis, a schematic of the GB transformation is summarized, as shown in Fig. 12. In terms of the transformation features in GB, it can be further classified into GB₁ and GB₂ types, as proposed by Qiao et al. [9] and Fang et al. [32] and mentioned before (Fig. 6, Fig. 7). In general, GB₂ forms at lower temperatures and has finer bainitic ferrite and M-A constituents compared to GB₁.

As discussed above, the volume fractions of GB₁, GB₂, and M-A constituents vary with the changes in the amount of RE. This should be explained from the perspective of RE solid solution in low-alloy Cr-Mo-V steels. Theoretical calculations have demonstrated that RE elements usually segregated at prior austenite grain boundaries that decreased the interfacial energies of the grain boundaries in iron alloys [16,33]. Some experimental results also revealed the RE solution at grain boundaries and/or interfaces in other low-alloy steels [31,34]. These RE solution hinder the diffusion of carbon atoms and improve the stability of undercooled austenite [16,17]. According to the formation mechanism of GB mentioned above, the lower diffusion rate of carbon atoms delays the formation of carbon-depleted and carbon-rich regions in prior austenite before bainitic ferrite transformation occurs. Consequently, lower onset temperature of GB transformation and the higher volume fraction of GB₂ are obtained as the RE solution increased. In addition, the formation of GB is related to the movement of bainitic ferrite/carbon-rich undercooled austenite interfaces that depend on the ability of carbon diffusion in the near-interface regions. The segregation of RE solution produces a drag effect at the interfaces [15], retarding the diffusion of carbon so as to decrease the size of bainitic ferrite. Besides, some researches have reported that the RE solution can increase the solid solubility of alloying elements because the interaction coefficients between RE and C, V, Nb, etc. are negative [35,36]. Further evidence could be provided from the research of Liu et al. [37] that the addition of RE into low-alloy vanadium-bearing steels would suppress the precipitation of VC markedly in the steel, thus increasing the stability of the undercooled austenite and causing the CCT curves to shift to the right. Therefore, the solution of RE can inhibit the formation of GB₁ with coarser bainitic ferrite and higher degree of carbon-enriched M-A constituents in normalized samples, thus a higher volume fraction of GB₂ microstructure with both lath bainitic ferrite and fine elongated M-A constituents can be obtained. However, the solid solubility of RE in iron crystals is believed to be very low. Therefore, excess addition of RE cannot further affect bainite transformation in the present study.

It is obvious that the addition of appropriate amount of RE has an important effect in inhibiting the formation of GB₁, as well as increasing the volume fraction of GB₂. Consequently, the impact properties of the steel are enhanced. There are two possible factors that can explain the higher volume fraction of GB₂, the superior impact toughness is. On one hand, a certain amount of elongated and massive carbon-rich M-A constituents existed in all the normalized samples, these M-A constituents will be decomposed into carbides and/or carbides aggregates after 700 °C tempering. However, massive M-A constituent with higher carbon content existing in the GB₁ regions easily decomposed into coarser carbides and/or carbides aggregates, which can act as slip barriers and provide potential sites for fracture initiation [3]. An appropriate amount of RE decreases the bainitic transformation temperature and promotes the formation of GB₂, resulting in a lower volume fraction of carbon-rich massive M-A constituents in normalized samples, and thus, superior impact properties are obtained in the normalized and tempered samples. On the other hand, although the RE additives in the amounts investigated have virtually no influence on the variation in prior austenite grain size in the present experiments, a smaller average effective grain size can be obtained in the RE-treated steel with a higher volume fraction of GB₂ than in the reference steel because of the smaller effective grain size usually observed in the GB₂ region compared to that in the GB₁ region (Fig. 9). It is widely accepted that the effective grain size, rather than the prior austenite grain size in the bainitic microstructure, plays a decisive role in resisting crack propagation. That is, the small effective grain size not only effectively suppress crack propagation under an impact load but also may alleviate stress concentration at the boundaries and enhance the coordination to plastic deformation, which contributes to an increase in Charpy impact energy. Consequently, the value of DBTT is much lower for the 0.012RE steel than for the reference steel. This is the reason why the impact properties of the 2.25Cr-1Mo-0.25 V steel improved remarkably with increasing amount of RE from 0 to 0.012 wt.%.

Unfortunately, the addition of RE to those steels markedly increases the DBTT and lowers USE when the content of RE is more than 0.012 wt.%, as may be seen by comparing the 0.048RE steel with the reference and 0.012RE steels (Table 2). This is observed, in spite of the effectiveness of RE in improving the impact properties by altering bainite transformation, as mentioned above. The explanation for this result is given in terms of the high levels of inclusions present in the excessively RE-treated steels. Wang et al. [38] also noted that a similar loss in the impact properties in their RE-treated steels. They ascribed this to the very low RE solubility in low-alloy steels; the excessive addition of RE leads to the formation of a large number of large-sized RE inclusions that, are extremely detrimental to the impact properties. According to the analysis of the impact fracture surface, some large-sized particles acted as the initiation of cleavage fracture can easily found in the Charpy specimens fractured at -120 °C. As Fig. 13(a) shows, a decohered particle of 3 μm in size (pointed to by a white arrow) initiates cleavage microcracking, with river-pattern lines stemming from it toward the contiguous grains. The particle was identified as a RE-enriched inclusion by EDS analysis, as shown in Fig. 13(c). Therefore, the fact that large-sized inclusion forms a microcrack as Griffith’s crack and triggers the final global cleavage fracture of the Charpy impact specimen is verified. Meanwhile, many particles in the microvoids of the fracture surfaces of the Charpy specimens fractured at 20 °C are also found (Fig. 13(b) and (d)), which are large, shattered inclusions caused by the local high-stress zones that promote the formation of microvoids during ductile rupture. Therefore, the very poor Charpy impact performance of 0.048 RE steel appears to be due to the high volume fraction of large-sized inclusions.

In addition, there are other influencing factors not studied in the present work, including definite content and distribution of dissolved RE in the steels, which may give more useful information on why modified microstructure and improved impact properties were obtained by adding appropriate amounts of RE to 2.25Cr-1Mo-0.25 V steel. Therefore, further investigations will be carried out in the next stage.

(1) Addition of RE to low-alloy 2.25Cr-1Mo-0.25 V steel can decrease the start temperature of bainitic transformation that results in the formation of more lath bainitic ferrite having smaller effective grain sizes and the reduction of carbon-rich M-A constituents in the normalized microstructures of the steels. After being subjected to tempering at 700 °C for 4 h, the M-A constituents decomposed into ferrite and carbides and/or agglomerated carbides.

(2) When the steel contained 0.012 wt.% RE, an excellent combination of strengths, USE, and DBTT could be obtained due to the decrease in the effective grain size and the amount of coarse agglomerated carbides as a result of the presence of more lath bainitic ferrite and lower volume fraction of the massive M-A constituents in the normalized specimens. Unfortunately, in this work, the addition of RE in excess of 0.012 wt.% leads to a substantial increase in the volume fraction and density of large-sized inclusions, which is extremely detrimental to the impact properties.