A Comparative Study of Acicular Ferrite Transformation Behavior between Surface and Interior in a Low C–Mn Steel by HT-LSCM

: In this study, the acicular ferrite transformation behavior of a Ti–Ca deoxidized low carbon steel was studied using a high-temperature laser scanning confocal microscopy (HT-LSCM). The in situ observation of the transformation behavior on the sample surface with different cooling rates was achieved by HT-LSCM. The microstructure between the surface and interior of the HT-LSCM sample was compared. The results showed that Ti–Ca oxide particles were effective sites for acicular ferrite (AF) nucleation. The start transformation temperature at grain boundaries and intragranular particles decreased with an increase in cooling rate, but the AF nucleation rate increased and the surface microstructure was more interlocked. The sample surface microstructure obtained at 3 ◦ C/s was dominated by ferrite side plates, while the ferrite nucleating sites transferred from grain boundaries to intragranular particles when the cooling rate was 15 ◦ C/s. Moreover, it was interesting that the microstructure and microhardness of the sample surface and interior were different. The AF dominating microstructure, obtained in the sample interior, was much ﬁner than the sample surface, and the microhardness of the sample surface was much lower than the sample interior. The combined factors led to a coarse size of AF on the sample surface. AF formed at a higher temperature resulted in the coarse size. The available particles for AF nucleation on the sample surface were quite limited, such that hard impingement between AF plates was much weaker than that in the sample interior. In addition, the transformation stress in austenite on the sample surface could be largely released, which contributed to a coarser AF plate size. The coarse grain size, low dislocation concentration and low carbon content led to lower hardness on the sample surface.


Introduction
The acicular ferrite formed in steel can be used in the technology called oxides metallurgy, where the acicular ferrite can nucleate at the surface of oxide inclusion during the transformation from austenite to ferrite. It is well known that the formation of acicular ferrite is strictly influenced by inclusion type, prior austenite grain size and cooling rate [1,2]. Intragranular acicular ferrite (AF) is known to provide an optimal combination of high strength and good toughness, because of its refined structure with high-angle grain boundaries and high-density of dislocations [3,4]. Thus, the fine interlocking microstructure is known to improve toughness in both the heat-affected zone (HAZ) or weld metal, and the deoxidized high-strength low-alloy (HSLA) steel [3,[5][6][7].
High-temperature laser scanning confocal microscopy (HT-LSCM) has been widely used to study the phase transformation from austenite to ferrite and grain growth in steels [8][9][10][11]. HT-LSCM is also an effective method to study inclusion-induced AF. Hanamura et al. [12] used this method to observe the nucleation of AF around inclusions and

Materials and Methods
Ti oxide showed distinct AF nucleating ability and received extensive research. In addition, strong deoxidants such as Ca, Mg and Zr are usually added together with Ti to achieve a dispersed distribution of fine-sized oxide. In the present research, low C-Mn steel with Ti-Ca complex deoxidation was chosen for the inclusion-induced AF transformation study. The experimental steel was melted in a 10 kg laboratory vacuum induction furnace. Ti-Ca was added as FeTi30 and SiCa30 alloy into the steel liquid before casting. The steel composition was shown in Table 1. Samples were taken from the cast ingot for microstructure observation and inclusion analysis by a DM2500 M optical microscope (OM, Leica Microsystems, Wetzlar, Germany) and an Ultra-55 scanning electron microscope (SEM, Carl Zeiss, Jena, Germany). The cast ingot was hot rolled into the steel plate and used for the subsequent experiment. Round samples with a diameter of 6 mm and thickness of 3 mm were machined and the top flat surface was mirror polished. The in situ observation was carried out on a LSM 510 Zeiss high-temperature laser scanning confocal microscope (LSCM, Carl Zeiss, Jena, Germany). The samples were heated to 1300 • C and held for 1 min, then cooled to room temperature with cooling rates of 3 • C/s, 6 • C/s and 15 • C/s, respectively. Transformation behaviors on the top surface were recorded by the microscope. Afterwards, the sample surface microstructures were observed before and after 4% nital etching. Furthermore, the samples were cut in half as shown in Figure 1 and the matrix microstructures inside the samples were also characterized by OM and electron backscattered diffraction (EBSD, Hikari, EDAX-TSL, Draper, UT, USA). The characterization position was at the center as circled in Figure 1. The macro Vickers hardness values for the top surface and inside the sample were measured by a KB hardness testing machine (KB Prüftechnik, Hochdorf-Assenheim, Germany). The microhardness was tested by an FM-700 microhardness tester (FUTURE-TECH CORP., Kawasaki, Japan). For comparison, the transformation-dilation experiment was carried out on Formastor-FII full-automatic transformation equipment (Fuji Electronic, Tokyo, Japan) using cylinder samples of dimensions Φ3 mm × 10 mm. The thermal cycles were similar to that for LSCM. center as circled in Figure 1. The macro Vickers hardness values for the top surface and inside the sample were measured by a KB hardness testing machine (KB Prüftechnik, Hochdorf-Assenheim, Germany). The microhardness was tested by an FM-700 microhardness tester (FUTURE-TECH CORP., Kawasaki, Japan). For comparison, the transformation-dilation experiment was carried out on Formastor-FII full-automatic transformation equipment (Fuji Electronic, Tokyo, Japan) using cylinder samples of dimensions Φ3 mm × 10 mm. The thermal cycles were similar to that for LSCM.

As-Cast Microstructure and Inclusion
The liquid steel was solidified in a copper mold and then taken out for cooling in air. A sample was prepared from the upper part of the ingot and the optical microstructure was shown in Figure 2. The morphology was of an AF-dominating type. The prior austenite grain size was extremely large in the cast condition and the whole field of view in Figure 2 was even within one parent austenite grain. Coarse ferrite plates forming at higher temperatures extended to a long distance and divided the austenite grain into parts. Finer ferrite plates formed within the local regions during cooling. The effective inclusion-inducing AF nucleation was characterized by SEM as shown in Figure 3. The Ti-Ca oxide inclusion also contained Al-Ca-Mn-S elements. Several mechanisms for inclusion-induced ferrite nucleation have been proposed including inert substrate mechanism, thermal strain mechanism, element depletion mechanism and lattice matching mechanism [23,24]. For Ti-oxide, most research believed the Mn depletion zone (MDZ) mechanism was reasonable, which was formed by Mn absorption into Ti-oxide particles or MnS adhering precipitation [1]. The formation of MDZ caused a decrease in Mn content and an increase in transformation temperature, which promoted intragranular AF nucleation around the inclusion [25,26]. In addition, the driving force for ferrite formation, i.e., the Gibbs free energy change for γ→α, increased with a decrease in Mn

As-Cast Microstructure and Inclusion
The liquid steel was solidified in a copper mold and then taken out for cooling in air. A sample was prepared from the upper part of the ingot and the optical microstructure was shown in Figure 2. The morphology was of an AF-dominating type. The prior austenite grain size was extremely large in the cast condition and the whole field of view in Figure 2 was even within one parent austenite grain. Coarse ferrite plates forming at higher temperatures extended to a long distance and divided the austenite grain into parts. Finer ferrite plates formed within the local regions during cooling.
center as circled in Figure 1. The macro Vickers hardness values for the top surface and inside the sample were measured by a KB hardness testing machine (KB Prüftechnik, Hochdorf-Assenheim, Germany). The microhardness was tested by an FM-700 microhardness tester (FUTURE-TECH CORP., Kawasaki, Japan). For comparison, the transformation-dilation experiment was carried out on Formastor-FII full-automatic transformation equipment (Fuji Electronic, Tokyo, Japan) using cylinder samples of dimensions Φ3 mm × 10 mm. The thermal cycles were similar to that for LSCM.

As-Cast Microstructure and Inclusion
The liquid steel was solidified in a copper mold and then taken out for cooling in air. A sample was prepared from the upper part of the ingot and the optical microstructure was shown in Figure 2. The morphology was of an AF-dominating type. The prior austenite grain size was extremely large in the cast condition and the whole field of view in Figure 2 was even within one parent austenite grain. Coarse ferrite plates forming at higher temperatures extended to a long distance and divided the austenite grain into parts. Finer ferrite plates formed within the local regions during cooling. The effective inclusion-inducing AF nucleation was characterized by SEM as shown in Figure 3. The Ti-Ca oxide inclusion also contained Al-Ca-Mn-S elements. Several mechanisms for inclusion-induced ferrite nucleation have been proposed including inert substrate mechanism, thermal strain mechanism, element depletion mechanism and lattice matching mechanism [23,24]. For Ti-oxide, most research believed the Mn depletion zone (MDZ) mechanism was reasonable, which was formed by Mn absorption into Ti-oxide particles or MnS adhering precipitation [1]. The formation of MDZ caused a decrease in Mn content and an increase in transformation temperature, which promoted intragranular AF nucleation around the inclusion [25,26]. In addition, the driving force for ferrite formation, i.e., the Gibbs free energy change for γ→α, increased with a decrease in Mn The effective inclusion-inducing AF nucleation was characterized by SEM as shown in Figure 3. The Ti-Ca oxide inclusion also contained Al-Ca-Mn-S elements. Several mechanisms for inclusion-induced ferrite nucleation have been proposed including inert substrate mechanism, thermal strain mechanism, element depletion mechanism and lattice matching mechanism [23,24]. For Ti-oxide, most research believed the Mn depletion zone (MDZ) mechanism was reasonable, which was formed by Mn absorption into Ti-oxide particles or MnS adhering precipitation [1]. The formation of MDZ caused a decrease in Mn content and an increase in transformation temperature, which promoted intragranular AF nucleation around the inclusion [25,26]. In addition, the driving force for ferrite formation, i.e., the Gibbs free energy change for γ→α, increased with a decrease in Mn content, which contributed to promoting AF nucleation [27]. In the present experimental steel, Ti oxide was the main inclusion type and was considered to be effective through the MDZ mechanism. content, which contributed to promoting AF nucleation [27]. In the present experimental steel, Ti oxide was the main inclusion type and was considered to be effective through the MDZ mechanism.

In situ Transformation Behaviors by LSCM
The whole transformation processes during cooling at different cooling rates were recorded by LSCM and the typical micrographs were presented in Figures 4-6. LSCM presented the sample surface change and temperature simultaneously so that we could identify the start temperatures of transformation at grain boundaries or at intragranular particles. Austenitizing temperature was chosen as 1300 °C to acquire a large enough austenite grain size for promoting intragranular ferrite formation. At 3 °C/s, transformation started at 680 °C and ferrite nucleation first happened on austenite grain boundaries, as indicated in Figure 4a. The initial growth of the grain boundary ferrite (GBF) was relatively slow. As the GBF grew along the boundaries and then towards the interior, intragranular ferrite plates were found to nucleate at 632 °C, as shown in Figure 4b. The subsequent transformation was accelerated through two nucleation ways, i.e., from intragranular sites and on boundaries. However, the final microstructure was dominated by the ferrite side plates (FSP) forming from boundaries. Figure 4c showed the intermediate stage of the transforming process. Figure 4d showed the end of transformation at about 575 °C. The coarse ferrite sheaves growing across the whole austenite grain were obvious.

In Situ Transformation Behaviors by LSCM
The whole transformation processes during cooling at different cooling rates were recorded by LSCM and the typical micrographs were presented in Figures 4-6. LSCM presented the sample surface change and temperature simultaneously so that we could identify the start temperatures of transformation at grain boundaries or at intragranular particles. Austenitizing temperature was chosen as 1300 • C to acquire a large enough austenite grain size for promoting intragranular ferrite formation. At 3 • C/s, transformation started at 680 • C and ferrite nucleation first happened on austenite grain boundaries, as indicated in Figure 4a. The initial growth of the grain boundary ferrite (GBF) was relatively slow. As the GBF grew along the boundaries and then towards the interior, intragranular ferrite plates were found to nucleate at 632 • C, as shown in Figure 4b. The subsequent transformation was accelerated through two nucleation ways, i.e., from intragranular sites and on boundaries. However, the final microstructure was dominated by the ferrite side plates (FSP) forming from boundaries. Figure 4c showed the intermediate stage of the transforming process. Figure 4d showed the end of transformation at about 575 • C. The coarse ferrite sheaves growing across the whole austenite grain were obvious. For the cooling rate of 6 °C/s, microstructural evolution was shown in Figure 5. Ferrite nucleation still started from austenite grain boundaries while the start temperature decreased to 670 °C. FSP had grown to a short distance at 640 °C while no intragranular nucleation was observed, as shown in Figure 5a. With the temperature decreasing to 625 °C, a number of intragranular acicular ferrite plates were observed to form, mostly at the  The transformation behavior for the cooling rate of 15 °C/s showed remarkable difference from that of 3 °C/s and 6 °C/s. Ferrite nucleation was observed to start at 601 °C at grain boundaries and intragranular particles simultaneously, as shown in Figure 6a. With the decrease in temperature, in Figure 6b, more intragranular primary and secondary AF plates nucleated and lengthened without the obvious formation of grain boundary FSP. In this condition, the parent austenite grain was segmented into smaller parts by intragranular ferrite plates, see Figure 6c. The AF plates grew in some particular orientations as a result of the K-S orientation relationship with parent austenite and formed a weaving configuration [28,29]. However, this trend did not continue to the end of transformation of all the retained austenite. It seemed that the available intragranular nucleation sites were limited. The AF formation stopped after all the available nucleation sites were used up, which was followed by the formation of parallel ferrite packets between the prior AF plates. Figure 6d showed the finish point of transformation by in situ observation. Even so, the surface microstructure evolution exhibited significant difference from the interior of the sample, which will be revealed in the following section.

Microstructural Comparison Between the Surface and Interior
The optical microstructures of different sample surfaces without etching after LSCM thermal cycles were presented in Figure 7, where two samples with lower austenitizing temperatures of 1000 °C and 1200 °C were also included. The nital etched surface micro- For the cooling rate of 6 • C/s, microstructural evolution was shown in Figure 5. Ferrite nucleation still started from austenite grain boundaries while the start temperature decreased to 670 • C. FSP had grown to a short distance at 640 • C while no intragranular nucleation was observed, as shown in Figure 5a. With the temperature decreasing to 625 • C, a number of intragranular acicular ferrite plates were observed to form, mostly at the intragranular particles. In comparison with Figure 4, the intragranular nucleating rate became higher and the AF fraction increased. AF plates were encountered during the growth of FSP, such that the chance for the formation of ferrite sheaves across the whole austenite grain decreased. In addition, the ferrite plates could also originate from the first-formed ferrite plate but with a different orientation, as shown in Figure 5b, where the long red arrow indicated the primary ferrite plate as well as its growth direction, while the short red arrows indicated the secondary ferrite plates. The transformation continued at 615 • C in Figure 5c. The transformation nearly finished at 565 • C, as shown in Figure 5d, and the morphology seemed more interlocked than that for 3 • C/s. The transformation behavior for the cooling rate of 15 • C/s showed remarkable difference from that of 3 • C/s and 6 • C/s. Ferrite nucleation was observed to start at 601 • C at grain boundaries and intragranular particles simultaneously, as shown in Figure 6a. With the decrease in temperature, in Figure 6b, more intragranular primary and secondary AF plates nucleated and lengthened without the obvious formation of grain boundary FSP. In this condition, the parent austenite grain was segmented into smaller parts by intragranular ferrite plates, see Figure 6c. The AF plates grew in some particular orientations as a result of the K-S orientation relationship with parent austenite and formed a weaving configuration [28,29]. However, this trend did not continue to the end of transformation of all the retained austenite. It seemed that the available intragranular nucleation sites were limited. The AF formation stopped after all the available nucleation sites were used up, which was followed by the formation of parallel ferrite packets between the prior AF plates. Figure 6d showed the finish point of transformation by in situ observation. Even so, the surface microstructure evolution exhibited significant difference from the interior of the sample, which will be revealed in the following section.

Microstructural Comparison between the Surface and Interior
The optical microstructures of different sample surfaces without etching after LSCM thermal cycles were presented in Figure 7, where two samples with lower austenitizing temperatures of 1000 • C and 1200 • C were also included. The nital etched surface microstructures and corresponding interior microstructures were shown in Figure 8. With a low austenitizing temperature of 1000 • C, the austenite grain size was fine and polygonal ferrite was formed via the reconstructive transformation mechanism, as shown in Figure 7a. With the increase in austenitizing temperature and austenite grain size, the amount of parallel ferrite plates increased and the surface relief phenomenon appeared as a result of the displacive transformation, as shown in Figure 7b,c. For the cooling rate of 15 • C/s (Figure 7d), with the massive formation of intragranular AF, the surface relief also became much clearer. Figure 8 showed the great difference between the surface and interior microstructures. In general, the interior microstructures for the experimental conditions were significantly finer than the surface. In Figure 8a-c, with the increase in cooling rate, the amount of intragranular AF plates increased and segmented the parent austenite grain into smaller regions. It corresponded to the result of in situ observation by LSCM. For the experimental cooling rates, AF dominating microstructures were always obtained in the interiors of different samples. AF plates obtained at 3 • C/s were coarse (Figure 8a') and they became much finer as the cooling rate increased to 6 • C/s and 15 • C/s. It was shown from the in situ transformation behavior that the intragranular AF nucleation rate increased at a higher cooling rate and the lengthening of the AF plates also speeded up. On the sample surface, the number of nucleating sites, i.e., effective oxide particles, was limited and the hard impingement probability between AF plates was low such that the AF plates on the sample surface could grow to a large size, but with a small plate number. On the contrary, the AF nucleation rate and hard impingement were remarkably improved in the sample interior and there was less space for AF plates to coarsen, which resulted in much finer microstructure.
The surface and interior microstructures were further analyzed by EBSD with the results shown in Figure 9. For the surface microstructures, the coarse plates and sheaves were characterized by high-angled boundaries. In addition, the low-angled boundary concentration increased with the increase of cooling rate. The interior AF microstructures exhibited dense high-angled boundaries. The AF plates within one parent austenite grain showed some particular orientations as a result of the orientation relationship. In the sample interior, besides the dominating AF plates, there also existed a certain amount of small ferrite grains with granular or quasi-polygonal morphology. It could be distinguished in optical micrographs in Figure 8 and also in EBSD maps. Some research indicated that polygonal ferrite could form from the remaining austenite between acicular ferrite grains, by a reconstructive mechanism after the reaction stasis of acicular ferrite [30]. However, according to the present in situ observation of the continuous cooling transformation behavior, especially for the cooling rates of 6 • C/s and 15 • C/s, austenite could almost completely transform into ferrite plates and packets via a displacive transformation regime accompanied with surface relief. In addition, in the EBSD maps of sample interior, the small granular ferrite grains within one parent austenite grain mostly have one similar or a few particular crystallographic orientations, whereas the orientations of ferrite grains formed by reconstructive mechanism should distribute randomly. This means the granular ferrite grains were actually the transverse sections of AF laths. Their nucleation sites were at inclusions above or beneath the viewing plane. Their formation belonged to the same transforming stage as the acicular ferrite plates.    The surface and interior microstructures were further analyzed by EBSD w results shown in Figure 9. For the surface microstructures, the coarse plates and s were characterized by high-angled boundaries. In addition, the low-angled bo concentration increased with the increase of cooling rate. The interior AF microstr exhibited dense high-angled boundaries. The AF plates within one parent austeni showed some particular orientations as a result of the orientation relationship. In t ple interior, besides the dominating AF plates, there also existed a certain amount o ferrite grains with granular or quasi-polygonal morphology. It could be distingui optical micrographs in Figure 8 and also in EBSD maps. Some research indicated t lygonal ferrite could form from the remaining austenite between acicular ferrite by a reconstructive mechanism after the reaction stasis of acicular ferrite [30]. Ho according to the present in situ observation of the continuous cooling transforma havior, especially for the cooling rates of 6 °C/s and 15 °C/s, austenite could almo pletely transform into ferrite plates and packets via a displacive transformation accompanied with surface relief. In addition, in the EBSD maps of sample inter small granular ferrite grains within one parent austenite grain mostly have one sim a few particular crystallographic orientations, whereas the orientations of ferrite

Micro and Macrohardness
The micro and macro Vickers hardness values of different microstructures, on the surface and inside the sample obtained under experimental cooling rates, were tested. The results were shown in Figure 10. In Figure 10a, microhardness for both the inside and surface of the sample increased with the increase in cooling rate while the surface microhardness was much lower than the sample interior. In Figure 10b, the macrohardness also increased with the cooling rate, but the difference between the surface and the inside was not obvious. In addition, the micro and macrohardness values tested from the sample interior were consistent with each other. However, the micro and macrohardness values tested from the sample surface varied significantly, where the surface microhardness was much lower than its macrohardness. This result was analyzed combined with microstructures. formed by reconstructive mechanism should distribute randomly. This means the granular ferrite grains were actually the transverse sections of AF laths. Their nucleation sites were at inclusions above or beneath the viewing plane. Their formation belonged to the same transforming stage as the acicular ferrite plates.

Micro and Macrohardness
The micro and macro Vickers hardness values of different microstructures, on the surface and inside the sample obtained under experimental cooling rates, were tested. The results were shown in Figure 10. In Figure 10a, microhardness for both the inside and surface of the sample increased with the increase in cooling rate while the surface microhardness was much lower than the sample interior. In Figure 10b, the macrohardness also increased with the cooling rate, but the difference between the surface and the inside was  Figures 11 and 12 showed examples of the indentation shape. The Vickers hardness tester indentation was a square. In Figure 11, the micro HV indentation side length was 12.6 µm and 9.5 µm for the sample surface and interior, where the indentation depth could be calculated as 2.5 µm and 1.9 µm, respectively. In Figure 12, the macro HV indentation side length for sample surface and interior was both about 253 µm and the indentation depth was calculated to be 51 µm. The thickness of surface microstructure could be measured from the cross-section, as shown in Figure 13. The thickness of the surface layer in samples with cooling rates of 3 • C/s, 6 • C/s and 15 • C/s was 9.3 µm, 7.0 µm and 5.6 µm, respectively. It was also found that carbide tended to precipitate under the surface ferrite layer as a result of carbon diffusion inwards. In the micro HV test carried out on the sample surface, the indenter could not penetrate through the surface layer and the indentation depth was, in fact, even less than half the layer thickness. However, the macro HV indentation depth was far more than the surface layer thickness, where the indentation depth was nearly determined by the interior microstructure hardness rather than the soft surface microstructure. Therefore, the micro HV test reflected the hardness of surface microstructure and interior microstructure factually. The macro HV test reflected the sample matrix hardness for both testing positions. not obvious. In addition, the micro and macrohardness values tested from the sample interior were consistent with each other. However, the micro and macrohardness values tested from the sample surface varied significantly, where the surface microhardness was much lower than its macrohardness. This result was analyzed combined with microstructures.  Figures 11 and 12 showed examples of the indentation shape. The Vickers hardness tester indentation was a square. In Figure 11, the micro HV indentation side length was 12.6 μm and 9.5 μm for the sample surface and interior, where the indentation depth could be calculated as 2.5 μm and 1.9 μm, respectively. In Figure 12, the macro HV indentation side length for sample surface and interior was both about 253 μm and the indentation depth was calculated to be 51 μm. The thickness of surface microstructure could be measured from the cross-section, as shown in Figure 13. The thickness of the surface layer in samples with cooling rates of 3 °C/s, 6 °C/s and 15 °C/s was 9.3 μm, 7.0 μm and 5.6 μm, respectively. It was also found that carbide tended to precipitate under the surface ferrite layer as a result of carbon diffusion inwards. In the micro HV test carried out on the sample surface, the indenter could not penetrate through the surface layer and the indentation depth was, in fact, even less than half the layer thickness. However, the macro HV indentation depth was far more than the surface layer thickness, where the indentation depth was nearly determined by the interior microstructure hardness rather than the soft surface microstructure. Therefore, the micro HV test reflected the hardness of surface microstructure and interior microstructure factually. The macro HV test reflected the sample matrix hardness for both testing positions.    Figures 11 and 12 showed examples of the indentation shape. The Vickers hardness tester indentation was a square. In Figure 11, the micro HV indentation side length was 12.6 μm and 9.5 μm for the sample surface and interior, where the indentation depth could be calculated as 2.5 μm and 1.9 μm, respectively. In Figure 12, the macro HV indentation side length for sample surface and interior was both about 253 μm and the indentation depth was calculated to be 51 μm. The thickness of surface microstructure could be measured from the cross-section, as shown in Figure 13. The thickness of the surface layer in samples with cooling rates of 3 °C/s, 6 °C/s and 15 °C/s was 9.3 μm, 7.0 μm and 5.6 μm, respectively. It was also found that carbide tended to precipitate under the surface ferrite layer as a result of carbon diffusion inwards. In the micro HV test carried out on the sample surface, the indenter could not penetrate through the surface layer and the indentation depth was, in fact, even less than half the layer thickness. However, the macro HV indentation depth was far more than the surface layer thickness, where the indentation depth was nearly determined by the interior microstructure hardness rather than the soft surface microstructure. Therefore, the micro HV test reflected the hardness of surface microstructure and interior microstructure factually. The macro HV test reflected the sample matrix hardness for both testing positions.      The above results indicated that there exists a significant difference in morphology and hardness between the surface microstructure and actual intragranular acicular ferrite. This could bring about deviation when in situ observation was used to study AF transformation behavior. Here, a transformation-dilation experiment was carried out to further analyze the cause and degree of the discrepancy. The result was shown in Figure 14. According to the temperature-dilation curves, the sample matrix transformation intervals moved towards a lower temperature with the increase in cooling rate, which was consistent with in situ observation. However, the critical transformation points viewed by LSCM differed from those indicated by temperature-dilation curves. The start and finish transformation temperatures of surface layers revealed by LSCM were both higher than that of the sample matrix revealed by dilatometry. It was also evidently indicated in Figure 14 that for the surface microstructure, start temperatures of nucleation at grain boundaries and intragranular particles decreased with the increase in cooling rate, where the two nucleation manners happened at the same time when the cooling rate was 15 °C/s. However, according to the microstructure morphology of sample interior, the matrix transformation took place only by intragranular AF nucleation. Thus the temperature-dilation curves reflected almost pure intragranular AF transformation behavior. Additionally, the deviation of the curve from a straight line at the early stage should be partly related to the sample surface transformation. The above results indicated that there exists a significant difference in morphology and hardness between the surface microstructure and actual intragranular acicular ferrite. This could bring about deviation when in situ observation was used to study AF transformation behavior. Here, a transformation-dilation experiment was carried out to further analyze the cause and degree of the discrepancy. The result was shown in Figure 14. According to the temperature-dilation curves, the sample matrix transformation intervals moved towards a lower temperature with the increase in cooling rate, which was consistent with in situ observation. However, the critical transformation points viewed by LSCM differed from those indicated by temperature-dilation curves. The start and finish transformation temperatures of surface layers revealed by LSCM were both higher than that of the sample matrix revealed by dilatometry. It was also evidently indicated in Figure 14 that for the surface microstructure, start temperatures of nucleation at grain boundaries and intragranular particles decreased with the increase in cooling rate, where the two nucleation manners happened at the same time when the cooling rate was 15 • C/s. However, according to the microstructure morphology of sample interior, the matrix transformation took place only by intragranular AF nucleation. Thus the temperature-dilation curves reflected almost pure intragranular AF transformation behavior. Additionally, the deviation of the curve from a straight line at the early stage should be partly related to the sample surface transformation. In the driving force for nucleation of a new phase, elastic strain energy ΔGε is a component of resistance [31]. ΔGε for sample surface layer could be negligible due to its free expansion during γ→α transformation. Thus, the driving force for transformation of the sample surface was higher than that of the sample interior and accordingly the transformation temperature was higher for the sample surface. AF on the sample surface could only nucleate at inclusion particles located in the surface layer of several microns, which were quite limited in number. AF plates could extend along the surface or towards the interior after nucleation, where K-S orientation relationship should be kept [28,32]. For In the driving force for nucleation of a new phase, elastic strain energy ∆Gε is a component of resistance [31]. ∆Gε for sample surface layer could be negligible due to its free expansion during γ→α transformation. Thus, the driving force for transformation of the sample surface was higher than that of the sample interior and accordingly the transformation temperature was higher for the sample surface. AF on the sample surface could only nucleate at inclusion particles located in the surface layer of several microns, which were quite limited in number. AF plates could extend along the surface or towards the interior after nucleation, where K-S orientation relationship should be kept [28,32]. For the sample interior, AF plates grew in any proper spatial directions but did not reach the sample surface because of its lower transforming temperature.
The coarse size of AF on the sample surface was due to several reasons. Firstly, AF formed at a higher temperature had a coarse size in itself, which was similar to the interior AF. Secondly, the available nucleation sites were quite limited such that hard impingement between AF plates was much weaker than that in the sample interior. In addition, ferrite plates formed via a displacive mechanism were affected by the austenite stress state. The transformation stress in austenite on the sample surface could be released largely and this contributed to a coarser AF plate size. With a coarse grain size, a lower dislocation concentration under high transformation temperature, and carbon diffusion away from the surface ferrite into the sample interior, the surface microstructure hardness was much lower than the whole AF microstructure in the same sample.

Conclusions
In situ observation of the transformation behavior of a Ti-Ca deoxidized low carbon steel was conducted using HT-LSCM. The microstructure between the surface and interior of the HT-LSCM sample was compared. The main conclusions are as follows: (1) Cooling rates have significant influence on the AF transformation behavior. With an increase in cooling rate, the nucleation rate of AF increased and the surface microstructure was more interlocked. Sample surface microstructure formed at 3 • C/s was dominated by ferrite side plates, while the ferrite nucleating sites transferred from grain boundaries to intragranular particles when the cooling rate was 15 • C/s. (2) The microstructure between the surface and interior of the HT-LSCM sample was obviously different. AF dominating microstructure was always obtained in the sample interior and was much finer than the sample surface. The micro and macrohardness values tested on the sample surface varied significantly, and the microhardness of the sample surface was much lower than that of the sample interior. (3) The start and finish transformation temperatures of surface layers were both higher than the sample matrix. For the sample surface, the start temperatures of nucleation at grain boundaries and intragranular particles decreased with the increase in cooling rate, where the two nucleation manners happened simultaneously when the cooling rate was 15 • C/s. However, the temperature-dilation curves reflected almost pure intragranular AF transformation behavior. (4) The combined factors led to the coarse size of AF on the sample surface. AF formed at a higher temperature resulted in a coarse size. The available particles for AF nucleation on the sample surface were quite limited, such that hard impingement between AF plates was much weaker than that in the sample interior. In addition, the transformation stress in austenite on the sample surface could be largely released, which contributed to a coarse AF plate size.  Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Conflicts of Interest:
The authors declare no conflict of interest.