ISIJ International
Online ISSN : 1347-5460
Print ISSN : 0915-1559
ISSN-L : 0915-1559
Welding and Joining
Quantitative Evaluation of Nucleation Potency of Ti-containing Inclusions for Acicular Ferrite
Kangmyung SeoHoisoo RyooHee Jin KimChanghee Lee
Author information

2019 Volume 59 Issue 6 Pages 1105-1112


The nucleation potency of different types of inclusions for acicular ferrite (AF) plates was quantitatively evaluated for Ti-containing welds. The objective of this investigation was to determine the relative efficiency of AF nucleation among the inclusions and to understand the effect of nucleation potency on the AF content. A series of bead-in-groove welds having Ti concentrations in the range of 0.002 to 0.091 wt.% were used in this study. The nucleation event of the inclusions was examined using scanning electron microscopy for a large number of inclusions of various sizes. Then, the nucleation probability curves were constructed as a function of inclusion size for each weld. The experimental results demonstrated that the nucleation probability increased with the inclusion size for all welds. However, this size effect varied with the Ti content of the welds substantially. By employing two different parameters to represent the difference of nucleation potency in a quantitative manner, it was found that the Ti effect on AF content resulted from the corresponding change in nucleation potency.

1. Introduction

In contrast to steel base plates, weld metals deposited by conventional arc welding processes are generally high in oxygen and therefore have a high number density of oxide inclusions. These inclusions can act as nucleation sites for acicular ferrite (AF) plates, and the nucleation ability is affected by many factors due to the competition with other phases such as grain boundary ferrite and bainite. Thereby, a considerable amount of work has been directed towards identifying the factors controlling the formation of AF microstructure in C–Mn and low-alloy steel welds, notably austenite grain size,1,2,3,4) alloying content,3,5,6,7,8,9) cooling rate2,6) and inclusion characteristics.3,10,11,12,13,14,15,16)

Among the factors, many researchers have focused on inclusion characteristics because large amounts of AF were found to be associated with specific types of inclusion phases like galaxite14) and inclusions coated with TiO11,13,17) or Ti(C,N).12,18) Recently, Sarma and others19) reviewed the inclusions published in the literature and classified them as either active or inert depending on their ability for ferrite nucleation. Among the active inclusions, however, there remain areas of doubt as to the relative effectiveness or potency for AF nucleation. Previously, many researchers3,16,20,21,22,23) estimated AF nucleation potency under the assumption that a high fraction of observed AF must have resulted from an increase in nucleation potency of the inclusions. However, because there are many other factors involved in AF formation, the AF content does not indicated the nucleation potency of the inclusions unless all other factors are equal.

Ito and Nakanishi24) were the first to report the beneficial role of Ti addition for AF formation. Since then, numerous investigators9,17,25,26) have observed that the AF content varies substantially with the Ti content of the welds. A series of systematic studies performed by Evans17,25) with C–Mn steel welds demonstrated that even a impurity level of Ti addition on the order of several tens of ppm dramatically modified the as-deposited microstructure of C–Mn steel welds from 5% to 70% AF. This result indicates that such a minimal content of Ti can change the nature of inclusions from inert to active. Evans also showed that further addition of Ti up to 150 ppm resulted in a decrease in AF content, but Ti content over 150 ppm tended to gradually increase AF content. Based on these results, the effectiveness of Ti-containing inclusions for AF formation may not be the same but may vary with inclusion characteristics. To verify the change in inclusion characteristics due to changes in Ti content of the welds, Blais and others27) investigated the same welds as those of Evans focusing on the inclusion chemistry and elemental distributions in the inclusions. They found that, in a weld containing 28 ppm Ti, a phase identified as MnTi2O4 developed on the surface of spherical inclusions of Mn-silicate. As this weld exhibited the maximum observed AF content, they concluded that MnTi2O4 efficiently formed AF. At 150 ppm Ti, the Ti-rich particles formed were largely surrounded by the Mn-silicate phase, and were therefore no longer in contact with the matrix. This fact was thought to be responsible for a decreased amount of AF in the intermediate range of Ti, implying that the inclusions in the 150 ppm Ti weld were less effective than those in the 28 ppm Ti weld. An increase in AF content at 410 ppm Ti was observed associating with the formation of two distinct Ti-rich phases with different manganese concentrations. One of these phases was believed to be TiO, which is well known to be effective for ferrite nucleation due to a low lattice misfit with ferrite. More recently, Seo and others28,29) also studied how inclusion characteristics varied with the Ti content of bainitic-type welds. In a 720 ppm Ti weld, which resulted in the maximum observed AF content, Ti2O3 constituted the large majority of the inclusions. These inclusions were further observed to be fully surrounded by a Mn-depleted zone (MDZ) known to be effective for AF nucleation in the wrought steels.30,31) Above results indicate that Ti-containing inclusions are active for AF nucleation regardless of the Ti content of the weld, but their nucleation potency appears to be different depending on the inclusion phases of Ti-containing welds.

The effect of inclusion size on the nucleation potency of inclusions is well established. According to the heterogeneous nucleation mechanism, the energy barrier for ferrite nucleation from inclusions decreases with increasing inclusion size.31) Therefore, larger inclusions have higher nucleation potency than smaller inclusions. To verify such a size effect, several investigators2,33,34,35) measured the nucleation probability of inclusions as a function of inclusion size. They showed that the nucleation probability increases with inclusion size following a sinusoidal trend from zero probability for inclusions smaller than 0.2 μm to 100% probability for inclusions of 1.1 μm or larger. Based on these results, we expect that the probability curves would shift to smaller sizes when the nucleation potency of the inclusion increases. In other words, the nucleation probabilities in the transition region would be different for the nucleant inclusions if their nucleation potencies were not the same.

Based on the above assumption, we proposed a method in this study to quantitatively evaluate the relative potency of inclusions and then determine the relationship between the nucleation potency and AF content in the welds. The bainitic welds previously studied26,35) were selected to eliminate the potential formation of grain boundary ferrite (GBF). These welds were also confirmed to be similar in chemistry including oxygen content. The only true variable in this study was the Ti content of the welds, which can significantly change the inclusion characteristics as well as the AF content.29)

2. Experimental Procedures

2.1. Weld Preparation

The five weld metal samples studied in this investigation are the same as those used for the previous studies.26,29) They are bainitic-type weld metals, all having the same chemistry except for Ti content, which was varied from 0.002 to 0.091 wt.% with the same level of oxygen content. As described in Fig. 1(a), each weld was made by a two-step welding technique. First, a five-pass buttering weld was deposited in a 10 mm deep groove to minimize the base metal dilution, and then a single-pass bead-in-groove weld was made in a 5 mm deep groove having a slot for inserting Ti fibers (0.3 mm in diameter). The number of Ti fibers was varied from zero to five to change the Ti concentration of the resulting weld. All the welds were deposited with a heat input of 25 kJ/cm using ER 100S grade wire and Ar+20%CO2 shielding gas.

Fig. 1.

Preparation of experiment welds: (a) joint geometry with the addition of Ti fibers and (b) a typical macro-structure of an experimental weld showing the location of microstructural analysis.

Figure 1(b) shows a typical bead-in-groove experimental weld deposited with an 8 mm bead height. Microstructural, chemical and inclusion analyses were all made at the center of the experimental weld bead, that is, within the circle indicated in Fig. 1(b). The chemical analysis of the experimental welds was conducted using an optical emission spectrometer (OES). LECO equipment was used to measure oxygen and nitrogen. The chemical uniformity of Ti was confirmed by EPMA analysis in the vertical direction, and no difference in Ti concentration was found.

2.2. Metallographic Examination and Inclusion Analysis

Microstructural and inclusion analyses were conducted on the experimental welds using optical microscopy (OM) and scanning electron microscopy (SEM). Etched specimens were prepared using a 2% Nital solution for microstructural analysis. A total of ten OM micrographs were taken at ×500 magnification with the measurement area being 0.5 mm2, and the proportion of AF was measured by a point-counting method. In addition, the width of columnar austenite grains was measured from SEM micrographs by the linear intercept method in a direction parallel to the plate surface.

Due to the importance of inclusion characteristics, particular attention was given to the inclusion size distribution, number density and the microstructure in the area adjacent to inclusions. Other characteristics such as chemical composition and constituent phases were obtained from the previous study.26,29) For inclusion analysis, the polished sample surfaces were observed using SEM at a magnification of ×10000. At this magnification, it was possible to count the fine inclusions down to 0.1 μm. A minimum of 300 particles from over 40 images taken from each weld were counted to obtain reliable data. After etching the specimens with 2% Nital solution, inclusion particles were re-examined in SEM at a higher magnification of ×20000 to determine if they acted as nucleation sites or not and the size of each inclusion was also measured. Inclusions were classified as nucleant only when the inclusion has more than one ferrite plate emanated. Using the results obtained from over 150 inclusions, the nucleation probability of inclusions was calculated and was plotted as a function of inclusion size.33,34)

3. Results and Discussion

3.1. Weld Metal Chemistry

The compositions of the five welds studied are given in Table 1 along with that of the welding wire. The welding wire contains a minimal amount of Ti, that is, 0.005 wt.%. The number in the weld designation is the number of Ti fibers inserted into the slot before depositing the experimental weld. As shown in this table, the chemical compositions of all welds were remarkably consistent with the exception of Ti content, indicating that they have identical hardenability. The weld made without Ti fiber (0Ti) had a Ti concentration of 0.002 wt.%, which is somewhat lower than that of the welding wire (0.005 wt.%). The Ti concentration of the weld metals increased linearly with the number of added fibers and reached 0.091 wt.% in the 4Ti weld, as shown in Fig. 2. Such a wide variation in Ti content had little effect on the oxygen concentration, which fell within a narrow range of 330 to 366 ppm for all the samples. The consistency in oxygen as well as in sulfur content suggests that the inclusion volume fraction would be identical for all welds.36)

Table 1. Chemical compositions of the experimental welds in wt.% excluding oxygen and nitrogen (ppm).
Fig. 2.

The variation in Ti and oxygen concentrations in the experimental welds as a function of the number of Ti fibers inserted. (Online version in color.)

3.2. Weld Metal Microstructure

All the welds contained a large amount of AF and its content varied with Ti content. Figure 3 shows the OM microstructures, which reflect a mixed microstructure of AF and bainite. These images were taken for the samples that showed two extremes in AF content, i.e., the 1Ti weld with the lowest AF content of 48% and the 3Ti weld with the highest of 92%. The AF contents measured for all welds are listed in Table 2, and these values also reported in the previous reports.26,29) Both 0Ti and 1Ti welds exhibits nearly equal proportion of AF and bainite. A further increase in Ti content results in a higher proportion of AF and reached in its maximum for the 3Ti welds and then decreased to 62% for 4Ti. The micrographs in Fig. 3 also show that GBF barely formed along the austenite grain boundaries, the absence of GBF formation being further demonstrated in SEM micrograph in Fig. 4. The average grain widths measured over 20 grains for each weld are also listed in Table 2, and these showed no appreciable change with the Ti content. This fact, along with the absence of GBF formation, suggests that the austenite grain size does not influence the AF content in the present welds.

Fig. 3.

Optical microstructures taken from (a) 1Ti (0.023 wt.% Ti) and (b) 3Ti (0.072 wt.% Ti) welds.

Table 2. Results of microstructural analysis.
ID (Ti content, wt.%)AF content (%)Austenite grain size (μm)
0Ti (0.002)5273.2
1Ti (0.023)4871.1
2Ti (0.049)6972.1
3Ti (0.072)9274.7
4Ti (0.091)6273.8
Fig. 4.

SEM micrograph of 4Ti weld showing the absence of GBF formation along the austenite grain boundaries.

3.3. Inclusion Size Distribution and Area Density

Figure 5 shows a low-magnification SEM micrograph of the 0Ti weld taken in backscattered electron mode. Inclusions were nearly spherical in shape, but the sizes of the inclusions varied over a wide range. The size distributions constructed with an interval of 0.1 μm for all welds are presented in Fig. 6. These distributions were remarkably similar for all the welds and had a peak frequency of 25 to 30% at a particle size of 0.3 to 0.4 μm. A quantitative evaluation of the average inclusion diameter and the number density is given in Table 3. As shown in this table, all welds were nearly identical in average size as well as in area density. Such consistency in average size (0.42 to 0.48 μm) and density (1.55 to 1.71×104/mm2) was attributed to the identical welding condition (i.e., the same heat input) and very similar oxygen and sulfur concentrations (Table 1). Since the geometric features such as shape, density and size distribution were similar in all welds, these were unlikely to be the reason for the variation in AF content shown in Table 2.

Fig. 5.

Backscattered electron SEM micrograph revealing the inclusions of 0Ti (0.002 wt.% Ti) weld in the as-polished condition.

Fig. 6.

The distribution of inclusion sizes measured for all welds. (Online version in color.)

Table 3. Results of inclusion analysis.
(Ti, wt.%)
Number of
size (μm)
Inclusion analysis
Mean diameter
Number density
0Ti (0.002)4711.70.481.55 × 104
1Ti (0.023)4901.60.461.61 × 104
2Ti (0.049)4942.10.461.62 × 104
3Ti (0.072)5192.00.451.71 × 104
4Ti (0.091)5021.50.421.65 × 104

The inclusion chemistry and the constituent phase of the inclusion are also known to affect AF content. The inclusion chemistry of the present welds have been analyzed using the extracted particles and was reported previously.29) As the Ti content of the weld increased, its content in the inclusions also increased with a corresponding decrease in Mn and Si. Due to these change in concentration, the major phases of the inclusions changed from Mn-silicate for 0Ti to a mixture of Mn-silicate and (Mn,Ti)-spinel oxide for 1Ti and 2Ti, and then to Ti2O3 for 3Ti and 4Ti welds. The formation of a manganese-depleted zone (MDZ) in the region surrounding the inclusions was also reported in 2Ti, 3Ti and 4Ti welds and was believed to be responsible for the high content of AF in those welds. Further details related to the inclusion characteristics and the proposed nucleation mechanisms for AF are summarized in Table 4.

Table 4. Summary of inclusion characteristics and nucleation mechanisms proposed.29)
IDConstituent phases in inclusionsMajor nucleation mechanism proposed
0TiMnSiO3, MnS, MnTi2O4Low misfit
1TiMn-silicate, (Mn,Ti)-spinel oxide, MnSMDZ
2Ti(Mn,Ti)-spinel oxide, Mn-silicate, MnSMDZ
3TiTi2O3, γ-Al2O3MDZ
4TiTi2O3, γ-Al2O3, TiNMDZ

3.4. Nucleation Probability

Figure 7 shows high magnification SEM micrographs presenting a number of inclusions and their adjacent microstructures. The inclusions of the 1Ti weld were located in an area of bainite-dominant, and those of the 3Ti weld were in AF-dominant. Depending on whether the AF plate(s) were nucleated or not, each inclusion can be classified into two groups like nucleant (N) or non-nucleant (NN).

Fig. 7.

SEM micrographs showing the nucleant (N) and non-nucleant (NN) inclusions, (a) 1Ti (0.023 wt.% Ti) and (b) 3Ti (0.072 wt.% Ti) welds.

In Fig. 7, the inclusions determined to be acting as nucleation sites for AF were marked as ‘N’, and those not acting as nucleation sites were labeled ‘NN’. In this figure, the size effect was apparent as the larger inclusions tended to be classified as N and smaller ones were NN. To substantiate and quantify this trend, we evaluated the nucleation event as a function of inclusion size using a parameter referred to as ‘nucleation probability’.33) The nucleation probability of the inclusions for a given size range, di, was calculated by dividing the number of N inclusions counted in di by the total number (N+NN) of inclusions in the same range. Nucleation probability can be expressed as:   

(Nucleation probability) di = (no. of nucleant inclusion) di              / (total no. of inclusion) di              = N di / (N+NN) di

Figure 8 shows the results of the nucleation probabilities plotted as a function of inclusion size in a 0.1 μm interval for all welds. As reported by previous workers,2,33,34) the nucleation probability increased with increasing inclusion size for all welds. The probability eventually reached 100% when the inclusions were larger than 0.85 μm, implying that all inclusions over 0.85 μm act as nucleation sites for AF regardless of Ti content. When the inclusion size was smaller than this, the nucleation probabilities of each weld became to be quite different. For example, the nucleation probability of inclusions in the 3Ti weld remained high down to 0.45 μm and then it decreased rather rapidly with decreasing inclusion size, reaching about 40% at di = 0.15 μm. In contrast, the nucleation probability of the 1Ti weld began to decrease from 100% at di = 0.85 μm reaching 0% at di = 0.15 μm. Other welds showed the nucleation probability falling between those of 1Ti and 3Ti welds.

Fig. 8.

Nucleation probability curves constructed as a function of inclusion size.

The overall change in nucleation probability shown in Fig. 8 indicated that there is a sort of transition from nucleant to non-nucleant behavior as the inclusion size decreases. This is similar to the ductile-to-brittle transition (DBT) behavior that occurs in Charpy impact tests of the steels. Because of this, we expected that the construction of a probability curve would be useful for comparing the nucleation potencies among the different types of inclusions formed in the different welds. Like the traditional evaluation of steel toughness with DBT curves, the nucleation potencies of the inclusions were enhanced as the probability curve shifts to the left. To evaluate such an enhancement, we introduced two parameters similar to those often employed for toughness assessment of steel materials. Those parameters well established in steel materials are either the impact value at a given temperature or the ductile-brittle transition temperature (DBTT) determined from the DBT curve.

First, the relative potency among the different types of inclusions of the present welds was estimated from the curves shown in Fig. 8 by comparing their probabilities at a given size; that is, similar to the comparison of toughness values at a given temperature. To employ this method, the probability at di = 0.45 μm, which was close to the average size of the present welds, was taken from each curve and was plotted as a function of Ti content. As shown in Fig. 9, Ti addition up to 0.07 wt.% enhanced nucleation potency of inclusions, but further addition deteriorated potency to some extent. The second parameter considered to quantify the probability curves was the inclusion size that corresponds to 50% probability. This parameter is like the ductile-brittle transition temperature (DBTT) determined from the results of Charpy impact tests. The transition sizes at 50% probability were taken from each curve and were plotted as a function of Ti content of the welds. As shown in Fig. 10, the transition size decreases with titanium content, reaching its minimum of ~0.3 μm at 0.07 wt.%, and it then increases to 0.5 μm at 0.09 wt.%. Accordingly, both parameters make it possible to compare the nucleation potency among the nucleant inclusions quantitatively, making it easier to determine which types of inclusions are more efficient for AF nucleation.

Fig. 9.

The effect of Ti content on the nucleation probability at an inclusion size (di) of 0.45 μm.

Fig. 10.

The effect of Ti content on the transition size of inclusions determined at 50% probability.

The above results demonstrated that the nucleation potency of inclusions for AF was strongly affected by the Ti content of the welds most likely due to the corresponding change in microstructural features of the inclusions shown in Table 4. When the Ti content of the welds was low, as in 0Ti and 1Ti welds, the inclusions were predominantly Mn-silicate partly covered with MnTi2O4.29) This means that the inclusions were less effective for AF nucleation, resulting in about a 40% probability at di = 0.45 μm and in a transition size of about 0.5 μm. Increasing the Ti content to 0.07 wt.% resulted in Ti2O3 inclusions surrounded with a MDZ. This enhanced the nucleation potency so that the nucleation probability at di = 0.45 μm increased to 90%, and the transition size decreased to about 0.3 μm, resulting in a larger number of inclusions acting as nucleation sites.

Even with the same mechanism like MDZ, the extent or the amount of Mn depletion can affect the nucleation potency of inclusions. However, making some proper quantification of its extent in terms of minimum Mn concentration and/or the MDZ width was not possible or nearly impossible due to the practical limit of experimental data. Under this limitation, previous investigation29) discussed that the MDZ extent may different depending on the formation mechanism, that is, the MDZ formation in 3Ti weld is due to the absorption of Mn atoms by Ti2O3 while that in 1Ti was suggested to be resulted from the slag/metal reaction between Mn-silicate and γ-matrix. Due to such difference in MDZ formation, the extent of MDZ in 3Ti weld was assumed to be different from and would be higher than that of 1Ti weld. In case of 4Ti weld having inclusion microstructure similar to 3Ti, the increased Ti content in steel matrix could result in a lower potency for 4Ti weld.

3.5. Relationship between the Nucleation Potency and AF Content

Following the results shown in Figs. 8 and 9, we concluded that the nucleation potencies of inclusions in 0Ti and 1Ti welds were similar and were the lowest observed in this study. The inclusions in the 3Ti weld showed the highest potency, and those in 2Ti and 4Ti welds showed intermediate potencies. We expected that these differences in nucleation potency would correlate to the AF content. In order to determine the nature of this relationship, the AF content listed in Table 2 was plotted as a function of each parameter, i.e., the nucleation probability at di = 0.45 μm and the transition size at 50% probability. The results are shown in Fig. 11, which illustrates that the AF content was linearly related to those two parameters. Because the constituent phase of the inclusions was the only variable in the present welds (Table 4), the results shown in Fig. 11 demonstrate that the change in AF content with Ti concentration was solely due to the variation in the nucleation potency of inclusions in the present welds. This conclusion further implies that the difference in nucleation potency is an important factor that determines the AF content of the weld microstructure. Therefore, the nucleation potency needs to be evaluated using either one of the parameters suggested above when the AF content varies during the addition of oxidizing elements such as Ti, Al and Mg.

Fig. 11.

Relationship between AF content and nucleation potency quantified as (a) the nucleation probability at di = 0.45 μm and (b) the transition size at 50% probability.

An investigation of the AF microstructure showed that there were many ferrite plates emanating from the sides of the primary ferrite that nucleated from the inclusions. Those ferrites were not associated with the inclusion and are often referred to in the literature as ‘sympathetically nucleated ferrite’ following the first report by Arronson and Wells.37) Because a large number of AF plates were nucleated sympathetically at the austenite/ferrite interfaces, the nucleation potency of the austenite/ferrite interfaces is likely also an important factor in AF formation.2) If the inclusions have high nucleation potency, they can provide a large interfacial area for sympathetic nucleation. This interfacial area was assumed to be roughly proportional to the nucleation potency of the inclusions. Thus, inclusions having a high nucleation potency not only increase the number of nucleant inclusions but may also increase the austenite/ferrite interface area available for further nucleation of secondary ferrites. As the AF content showed a linear relationship with the nucleation potency of inclusions (Fig. 10), the extent of sympathetic nucleation may also have a linear relationship with the nucleation potency of the inclusions. This would be the subject of further investigation.

4. Conclusion

(1) Bead-in-groove welding performed with Ti-fibers produced weld metals having the same composition except for the Ti content. Due to the small change in oxygen content, the inclusion size and density remained the same for all the welds. In addition, the welds all had the same columnar grain size. Consequently, the only remaining factor that can influence the nucleation potency of the inclusions was the constituent phase of inclusions, which varied with the Ti content of the welds.

(2) A new methodology was developed to quantify the nucleation potency of the weld metal inclusions, and this methodology was used to compare the relative AF nucleation potencies among the different types of inclusions. In this methodology, we first constructed probability curves showing the nucleation rate as a function of inclusion size. We then quantified the nucleation potency by employing two parameters: the nucleation probability recorded at the average inclusion size and the transition size of inclusions showing 50% probability of nucleation. The relative AF nucleation potency of the weld inclusions were evaluated for different welds based on these parameters.

(3) For the present welds, the two parameters introduced in 2) above varied significantly with the Ti content of the welds, indicating a substantial difference in nucleation potency among the Ti-containing inclusions. The Ti2O3-type inclusions in the 0.07 wt.% Ti weld were the most effective for AF nucleation. In addition, both parameters showed linear relationships with the AF content of the welds, indicating that the variation in AF content with Ti concentration was due to the difference in nucleation potency of inclusions among the welds. Based on our work, the relative potency for AF nucleation needs to be evaluated as one of the inclusion characteristics by employing the parameters suggested in this study.

© 2019 by The Iron and Steel Institute of Japan