Effect of Direct Powder Additions on the Solidification Structure and Microsegregation of 42CrMo4 Steel

Abstract

Inoculation and its link with the solidification structure is a relatively new field for low alloy steels. In this study, a cold crucible setup is used to realize direct particle inoculation of 50 g steel ingots. Eight different inoculants powders (oxides, nitrides and carbides) were tried with a 0.3 mass% level addition. Solidification structure sizes and morphologies, presence of inoculant particles and microsegregation have been characterized for all the samples. The best grain refinements were obtained for Si₃N₄, TiN and CeO₂ additions whereas the lowest microsegregation intensities are achieved for Si₃N₄, HfC and W₂C additions. The properties of the inoculants – misfits, solubility products and terminal velocities – are used to discuss those changes. The grain refinement could be linked to the misfit in good agreement with the literature. Other morphological changes (secondary arm spacing and dendricity) were attributed to the presence of inert particles staying inside the liquid during the solidification. Last but not the least, the flattest microsegregation profiles were possibly due to inoculant dissolution leading to a change in the MnS precipitation sequence.

1. Introduction

During the solidification of massive industrial ingots, the main arising defects are porosities, inclusions and chemical segregations.¹⁾ The control and reduction of those defects is always aimed to produce cleaner ingots that will be transformed in better final semi-products. Chemical segregations happen at the microscopic scale because of solute redistribution during solidification. Solidification shrinkage, deformation of the mushy zone and/or relative motion between the solid and the liquid phase can, if coupled to microsegregation, induce segregations at a much larger scale, i.e. macrosegregation.¹⁾ The shape and number of equiaxed grains in the melt are believed to act on the relative motion between the solid grains and the liquid before and after packing.²⁾ Therefore, any modification of the equiaxed grain population should lead – as shown by numerical simulations²⁾ – to different macrosegregations. Different leverages have been used in the literature like modifications of casting design (hot topping, mold shapes) and/or casting process (pouring rates, electromagnetic stirring).^3,4,5,6)

Inoculation is defined as an addition made into the melt designed to increase the number of heterogeneous nuclei in the melt. It has been extensively studied in the cast iron and aluminum industry.^7,8) Previous studies for low alloyed steels have focused on the addition of alloy that will dissolve and create/release particles.^9,10,11) Inoculant efficiency then relies either on the ability to dissolve the inoculant and if needed to nucleate useful heterogeneous nucleation sites. On the other hand, direct particle additions without master alloy – where the inoculation efficiency relies on the ability to add and keep those particles inside the liquid until the solidification begins – are rarer in the literature. Recent investigations show that centerline segregation in continuously cast steel was improved after TiN, TiC or ZrN additions but no study was done about the solidification structure.¹²⁾ More tests need to be performed to control this kind of inoculation technique and more robust criterions have to be designed for relevant inoculant selection.

In this study, the main focus is to give a fast and precise methodology to select inoculant powders and to assess their effect on the solidification structure. Inoculant selection is done accordingly to their crystallographic and chemical properties. Microstructure characterization is then performed to evaluate grain sizes, grain morphologies and secondary arm spacing. The new proposed tools will be used to discuss the outcome of different direct powder inoculation trials. Finally, the link between inoculation and microsegregation will be investigated to assess if inoculation can be beneficial to the chemical homogeneity of the as-cast product.

2. Experimental

2.1. Base Alloy

The steel grade in this study is 42CrMo4 (SAE 4140) which is a through and surface hardening steel. It is industrially ingot cast and its nominal composition is given in Table 1. According to a Scheil calculation performed with Thermo-Calc (TCFE9 database), it solidifies in δ-ferrite up to a solid fraction of 15% and then in γ-austenite.

Table 1. Composition of 42CrMo4 steel.

Fe	C	Cr	Mo	Mn	Si
bal.	0.4	1	0.14	0.72	0.2

2.2. Cold Crucible Setup

An induction furnace allowing semi-levitation of the liquid phase, as seen in Fig. 1, was used to perform the inoculation trial under a controlled argon atmosphere after primary vacuum cleaning. The copper crucible is water cooled during the whole experiment.

Fig. 1.

Cold crucible setup. (Online version in color.)

The inoculant is added in a drilled hole in the base alloy prior to the beginning of the melting. The added amount is set at 0.3 mass% for 50 g ingots. To minimize the inoculant loss, the holding time at the fully liquid state was set to 5 s.

Pyrometric controls were made to ensure that thermal cycles were reproducible. From those measurements, the estimated cooling rate during the solidification is in the 30–35°C/s range.

A Comsol simulation of the thermal field in stationary regime has been done by setting the boundary conditions with the help of the pyrometric acquisitions.¹³⁾ It shows that the order of magnitude for the isotherm velocity is 1 × 10⁻² m/s and 5 × 10³ K/m for the thermal gradient. Also, this simulation is predicting a solidification time around 1 s which agrees with experimental observations.

For the estimation of the liquid velocity due to the electromagnetic stirring, a fast speed acquisition camera was used to estimate the velocity of oxide islands at the surface of the liquid. Another counter estimation was performed by doing an energy balance between the electromagnetic field, the gravitational field and the viscous dissipation following Herlach.¹⁴⁾ It gives a liquid velocity between 1 × 10⁻² m/s (energy balance) and 5 × 10⁻² m/s (camera). In that interval of velocity and for particles in the 1 μm range, calculated Reynolds numbers are inferior to 15 × 10⁻² which is widely below 1.

A reproducibility study was preliminary done to ensure that the main solidification features (cooling rate, grain size, secondary arm spacing and equiaxed area) were stable for the base alloy. Ten ingots were cast in the same conditions. All the error bars provided in this study come from this preliminary study.¹⁵⁾

2.3. Experimental Characterizations

As inoculation aims to modify the grain size, shape and morphology, it is necessary to clearly quantify those properties. The following measurement procedures are a proposition of those metrics. They will be used later to discuss the different inoculation trials.

2.3.1. Metallography and Optical Microscopy

Samples were ground up to 4000 SiC paper. Final polishing was performed with 3 μm followed by 1 μm diamond suspensions. The solidification structure was revealed by warm etching in Bechet-Beaujard reagent. Macrographs were taken with a ZEISS Axioplan 2/Axiocam MRc5 optical microscope along with a motorized platform.

2.3.2. Solidification Grain Size Measurement

The freeware Fiji was used for image analysis.¹⁶⁾ Each picture was converted in 8 bit gray and histogram equalization was performed.¹⁷⁾ To ensure that the average brightness was constant for all the pictures, the Stack Contrast Adjustment plugin was used.¹⁸⁾

Manual outlining of roughly 200 grains per sample on four pictures in the equiaxed zone was performed with the use of dendrite primary axis orientations and perpendicular secondary arms. All four pictures are located at the same position for all the experiments. A custom Fiji macro was used to store all the data about each 2D contour in the same way as described in another publication.¹⁹⁾ A colored image of all analyzed pictures is also produced. An example is presented in Fig. 2. For each grain, the Feret diameter is retrieved from the drawn contour.¹⁹⁾ The Feret diameter will be used directly as grain size.

Fig. 2.

Microstructure output after grain size measurement in Fiji. (Online version in color.)

2.3.3. Morphological Measurements

From a morphological point of view, a new indicator called intragranular fraction was introduced. It is defined as the area fraction in each grain after its binarization. The closer to 1 this ratio is, the more globular the grain will be; whereas the closer to 0 it is, the more dendritic the grain will be. Schematic examples of globular and dendritic grains selected from Fig. 2 are presented in Fig. 3. For dendritic grains, the intragranular fraction is close to 0.5 whereas it is higher for globular cases. This new parameter can be compared to the internal fraction sometimes used in numerical studies,²⁰⁾ to describe the morphology of dendritic grains.

Fig. 3.

Example of grains with different intragranular fractions (from left to right 0.82, 0.60, 0.52). (Online version in color.)

For secondary arm spacing measurements, automatic intercepts with random positioning of 300 lines on the binarized structures were performed with the help of the Python 3.5 scikit-image library.²¹⁾ Secondary arm spacing is then defined as the mean value of all the intercepted lengths similarly to Monroe.²²⁾ Although it is not strictly the secondary arm spacing value, it is a fast way to provide fineness measurement for a given structure. Nevertheless, contrary to both grain size and intragranular fraction, the mean intercepted length is not defined relatively to each grain. It characterizes the full population of equiaxed grains and does not allow good tracking of heterogeneities and outliers in the measurement population.

2.3.4. Presence of Inoculant Particles

Characterizations were made with a FEI Quanta 600 scanning electron microscope with a field electron gun at 20 kV in backscattering mode. Chemical compositions were retrieved by EDX with a SDD Bruker detector. To improve the statistical analysis, a software – MeTiS²³⁾ – controlling the SEM/EDX system and identifying particles by a thresholding method was used. Acquisition times are in the 12 h order of magnitude for roughly 800 analyzed particles.

2.3.5. Microsegregation Measurements

A JEOL JXA 8530-F electron microprobe was used to characterize the microsegregation on grids of points located inside the equiaxed area. Grid dimensions (100 μm spacing on a 15 × 15 points square grid) were selected according to the structure characteristic sizes.²⁴⁾ Ranking of points by increasing chromium content, dividing by the equilibrium partition coefficient of the relevant element (values extracted from each binary Fe-X system in Thermo-Calc) and attribution of solid fraction, fs, from 0 to 1 then allows retrieving a general picture of the enrichment of the liquid phase during the solidification.²⁴⁾

Microsegregation intensity score, MIS, for chemical element X ∈[Cr, Mn, Mo] is defined as:   

$$\text{MIS}={\displaystyle \prod }_{\text{X}\in [\text{Cr},\mathrm{\hspace{0.17em} }\text{Mn},\mathrm{\hspace{0.17em} }\text{Mo}]}{\displaystyle \frac{{\text{mass\%X(f}}_{\text{s}}=0.8\text{)}-{\text{mass\%X(f}}_{\text{s}}=0.05\text{)}}{<\text{mass\%X}{>}_{0.05<{\text{f}}_{\text{s}}<0.8}}}$$

First 5% solid fraction (f_s < 0.05) were not taken into account as the extremum points are very sensitive to the position of the measurement points. In the same way, last 20% (f_s > 0.8) were removed because microsegregation is perturbed by MnS precipitation events. The mean value, <mass%X>, is calculated on this shortened domain. The microsegregation intensity score is then giving the extent of microsegregation that takes place after the solid fraction reaches 5% and before MnS precipitation events begin).

3. Inoculant Selection

This study presents a methodology to select inoculant powders based on the materials properties of both the inoculant and the liquid steel. While those properties are widely known to be important in the literature,^25,26,27) a reevaluation was proposed to improve the prediction of inoculation success.

A good inoculant particle should be able to stay and survive inside the liquid phase as well as should own some crystallographic planes and directions to allow heterogeneous nucleation. The relevant properties of the powders used here (misfits with ferrite and austenite, solubility products and terminal velocities) are defined thereafter.

3.1. Misfit

Inoculant powders were primary selected according to their crystallographic match with the nucleating ferritic solid. Bramfitt defined the planar lattice disregistry (more simply called misfit), δ, to account for all possible orientations of the different crystallographic lattices as follows:²⁵⁾   

$${\delta }_{{(\text{hkl})}_{\text{m}}}^{{(\text{hkl})}_{\text{p}}}=100\cdot {\displaystyle \sum }_{\text{i}=1}^3{\displaystyle \frac{\left|{\text{d}}_{{[\text{uvw}]}_{\text{p}}^{\text{i}}}\cdot cos\theta -{\text{d}}_{{[\text{uvw}]}_m^{\text{i}}}\right|}{3\cdot {\text{d}}_{{[\text{uvw}]}_m^{\text{i}}}}}$$

where p is standing for particle, m is standing for metallic primary phase, [uvw] is a low-Miller index direction, (hkl) is a crystallographic dense plan, θ is an angle between two crystallographic directions and d is an interatomic distance. For cubic lattices, the considered crystallographic plans and directions are (100), (110), (111) and [100], [110], [110]. For hexagonal lattices, the considered crystallographic plan and directions are only the (0001) basal plate and [1010], [1010], [2110].

In addition to the misfit with ferrite, δ_δ, the misfit with austenite, δ_γ, is also provided. It could be of importance because the base alloy undergoes a peritectic transformation.

3.2. Solubility Product

Stability of the compounds at the temperature of the melt is needed for the powder particle to act as a heterogeneous nucleation site.

For all the inoculant compounds, the precipitation/dissolution reaction from the 1 mass% dissolved states²⁸⁾ in pure liquid iron at an homogeneous temperature, T, and atmospheric pressure was considered for a generic MX inoculant:   

$$\alpha \cdot [\text{M}]+\beta \cdot [\text{X}]\leftrightarrow <{\text{M}}_{\alpha }{\text{X}}_{\beta }>$$

with M, a metallic element (e.g., Ce, Al, Zr, Si, Ti, W or Hf) and X, a non metallic element (e.g., C, O or N). The notations [X] and <X> are respectively indicating dissolved in liquid iron or solid X component.

For an ideal solution that respects Henry’s law, if the thermodynamic equilibrium of the previous reaction is reached and supposing that the activities of all solid phase are equal to 1, the reaction constant, K(T), can be derived as:²⁸⁾   

$$log(\text{K(T)})=log\left({\text{mass\%X}}^{\alpha }\times {\text{mass\%M}}^{\beta }\right)=\frac{-\mathrm{\Delta }{\text{G}}_{\text{r}}^0(\text{T})}{\text{RTln(}10\text{)}}$$

The solubility product, p_s, is then defined as the intercept of the log(mass%M) = f(log(mass%X)) line:   

$${\text{p}}_{\text{s}}=\frac{-\mathrm{\Delta }{\text{G}}_{\text{r}}^0(\text{T})}{\text{RT}\alpha \mathrm{\hspace{0.17em} }\text{ln(}10\text{)}}$$

The calculation temperature, T, is set to 1600°C. The free energies of reaction from elements in their standard states and dissolution energies to move from the standard state to the 1 mass% dissolved state in liquid iron were gathered from different sources.^29,30,31) This data was used to calculate the Gibbs free energy of dissolution in pure iron of each MX compound, $\mathrm{\Delta }{\text{G}}_{\text{r}}^0$.

3.3. Terminal Velocities

The behavior of a particle deposited at the free liquid steel surface can be mechanically described by the sum of applied forces.³²⁾ Before going any further, it is worth saying that this sum depends on the surface energies at the liquid/gas and liquid/particle interfaces. Furthermore, chemical and physical factors are also known to modify interfacial tensions.³³⁾ This kind of data for liquid steels is quite difficult to obtain and can sometimes strongly depend on the oxygen content in the gas phase.³⁴⁾

Another phenomenon is the motion of inoculant in the melt resulting in the accumulation of particles at the top or bottom of the melt. The maximum velocity reached by an emerged spherical particle in a motionless liquid, v_T, is estimated at low Reynolds numbers according to the Stokes law (Re<<1):³⁵⁾   

$${\text{v}}_{\text{T}}=\frac{2\text{g}}{9\mu }({\rho }_{\text{L}}-{\rho }_{\text{S}})\cdot {\text{r}}_{50}^2$$

42CrMo4 liquid steel density at liquidus temperature, ρ_L, was estimated thanks to Miettinen.³⁶⁾ Bulk densities of the solid particles at room temperature, ρ_S, were taken from Roucan.³⁷⁾ Dynamic viscosity, μ, at 1600°C and gravity, g, were respectively set to 4 × 10⁻³ Pa/s and 10 m/s² (value for Fe-1mass%C³⁴⁾). Median diameter values, d₅₀ = 2r₅₀, were characterized by laser diffraction.

The terminal velocity is de facto a comparison between the drag force, the buoyancy force and the weight for each kind of inoculant. Its sign is indicating if the particle tends to float (negative terminal velocity) or sedimentate (positive terminal velocity). Its magnitude can consequently gives an indication about the particle removal rate. In the present case, the liquid is not motionless because of the electromagnetic stirring. Consequently, the terminal velocity can only tell if a particle tends to be ejected from the liquid pool or to follow the forced convection loops.

3.4. Powder Grain Size

Different powders were chosen based on their availability in the 1–10 μm powder grain size. This size is a good compromise between a low nucleation undercooling and a good toughness.³⁸⁾ This activation undercooling was calculated for both austenite and ferrite (properties extracted from Miettinen³⁶⁾) and was below 2°C for 1 μm particles. In addition, the inoculant weight (0.3 mass%) was chosen to ensure that the number of added nucleation sites would be much greater that the number of equiaxed grains inside a base alloy casting.

3.5. Expected Outcome of Inoculation Trials

The eight inoculants selected for this study give a large range of properties in terms of misfits, solubility product and terminal velocities, as depicted in Table 2. As some of them are not expected to work, it should provide a strong basis to derive some criterions concerning the success of inoculation tests.

Table 2. Properties of inoculant powders.

Particle	δF (%)	δγ (%)	ps	d50 (μm)	vT (mm/s)
CeO2	5.6	6.7	−11	0.7	0.0030
Al2O3	7.3	6.5	−13	1	−0.16
ZrO2	7	0.02	−10	16.5	−12.3
SiO2	12	0.97	−5	0.6	−0.094
TiN	4.6	3.1	−3	0.6	−0.030
Si3N4	6.2	0.45	0.8	0.35	−0.025
W2C	8.1	1.94	4.8	4.13	+9.6
HfC	9.6	8.84	−1.4	6.88	+14.7

Based on the following criterion a first guess was taken prior to the experiment to predict the inoculation success:

- Undercooling: Bramfitt proposed that grain refinement would happen only for misfit below 12%.²⁵⁾

- Dissolution: As titanium nitrides are known to be stable in liquid steel melt,³⁹⁾ its solubility product was taken as a limit case. Above this limit, dissolution problems may happen.

- Ejection of the convection loop: If the liquid is considered motionless, a critical velocity exists to remove any particle inside the liquid during the period of time where the liquid metal is solidifying. By considering the characteristic size of the cast (~2 cm diameter) and the solidification time (~1 s), this critical velocity would be around 1 × 10⁻² m/s.

Properties for each inoculant have been represented in two scatter plots in Fig. 4. Two areas – green and red – have been defined with the limit values defined before. The green zone is representing the expected successful grain refinement zone. Its validity will be discussed later and particular attention will be drawn on the effect of inoculant belonging to the red zone (ZrO₂, W₂C, Si₃N₄ and HfC).

Fig. 4.

Prediction scatter plots ((a) and (b)) for successful inoculation (in green) with respect to inoculant properties. (Online version in color.)

4. Solidification Structure Results

In this section, the main focus is to study the impact of inoculation on the solidification structure. The grain size which is the parameter of primary importance will be studied along with other morphological changes.

4.1. Microstructure

Microstructures in the equiaxed grain zone for each inoculation trials are presented in Fig. 5. Visual comparisons are difficult because each grain cannot be easily identified at the first sight. So, grain size trends cannot be retrieved at this step. Nevertheless, it is still possible to outcast samples that exhibit very different equiaxed grain shapes.

Fig. 5.

Microstructure of the base alloy (a) and TiN (b), CeO₂ (c), Si₃N₄ (d), SiO₂ (e), Al₂O₃ (f), ZrO₂ (g), HfC (h), W₂C (i) inoculated samples. (Online version in color.)

For ZrO₂ additions (Fig. 5(g)), big globular grain pockets (green ellipses) can be found enclosed by very large equiaxed grains (red polygon) whereas for Al₂O₃ additions (Fig. 5(f)), small pockets of very fine structures (blue ellipses) are found in between very globular grains (green rectangle). To a lesser extent, solidification structures of Si₃N₄ and SiO₂ inoculated samples (Figs. 5(d) and 5(e)) also seem to follow the same behavior than Al₂O₃ additions.

4.2. Equiaxed Grain Size

For quantitative purposes, it is necessary to move to more precise image analysis measurements. Manual measurements were performed on four images of the same size (0.33 cm × 0.33 cm) enclosed in the equiaxed area. In fact, Fig. 5 accounts for 25% of the measurement area for each trial. Grain sizes are plotted against the misfit value with ferrite in Fig. 6. Error bars were determined with a previous study by measuring the grain size on ten base alloy castings.¹⁵⁾

Fig. 6.

Dependence of inoculant misfit with ferrite on the final solidification grain size. (Online version in color.)

Inoculation seems to be the most efficient for high misfits candidates like Si₃N₄, TiN and CeO₂. For those experiments where the misfit values were below 7%, the grain size was decreased from 13% up to 21% in comparison to the base alloy. Inoculants with misfits between 7% and 12% show lower levels of refinement that cannot be strictly separated from the reference within the limit of uncertainty. Finally, SiO₂ with a 12% misfit did not achieve any grain refinement. This is in good agreement with the existing literature which draws attention at the misfit as the main parameter.²⁵⁾

It is interesting to comment that the Si₃N4 inoculated alloy is the finest alloy although Si₃N₄ solubility product is quite high (p_s = 0.8). First, the growth restriction factor needs to be assessed to investigate possible solutal grain refinement.⁴⁰⁾ After retrieving liquidus slopes and equilibrium partition coefficients from Thermo-Calc (TCFE9 database) for all elements in Table 1, it was found that 0.3 mass% dissolved Si₃N₄ would change the growth restriction factor from −30.7 to −31.5 K/mass% which is probably insufficient to explain the observed refinement. Tyas²⁷⁾ first suggested that a chemical reaction was happening at the Si₃N₄/liquid interface producing an intermediary phase onto which austenite nucleation occurs. They verified their hypothesis by consecutive remeltings to favor the dissolution of the Si₃N₄. After one remelting of a Si₃N₄ inoculated sample, no grain refinement was further observed. Low dissolution kinetic could also be at stake and additional information at the steelmaking temperature in 42CrMo4 liquid steel would be needed to conclude on that matter. The Si₃N₄ case means that successful inoculation can happen at high solubility product. More complex criterion on dissolution kinetics would need to be derived.

In a similar way, carbide particles were also found inside the bulk metal although their terminal velocities and more importantly their solubility products are high. Low dissolution rates coupled to sedimentation could then explain the columnar interdendritic area location of the inoculant particles. In contrast to Si₃N₄ particles, no grain refinement was observed because the particles were captured in the columnar area without being able to nucleate new grains in the remaining liquid.

Finally, unexpected grain growth was observed for ZrO₂ addition although its misfit is low and its terminal velocity is high. The reason could lie in the sign of the terminal velocity which is positive contrary to the case of carbides additions (HfC, W₂C). Entrapped ZrO₂ particles in the liquid could be ejected at the top of the sample. As the temperature is decreasing, remaining particles at the upper liquid free surface could act as nucleation sites and create new grains. A part of those columnar grains could even detach themselves and would then become very large equiaxed dendritic grains. Similarly to the solubility product criterion, the terminal velocity criterion could then be refined. It seems that flotation is not preventing grains from becoming nucleation heterogeneous sites.

4.3. Equiaxed Grain Morphology

Mean intragranular fractions are presented for each sample in Fig. 7. This parameter is giving an indication about the morphology (dendritic or globular) of the equiaxed grains. This indication is relevant because it could modify the settling of the dendritic grains or the permeability of the packed granular medium in a large ingot.

Fig. 7.

Ranking of mean intragranular fraction after inoculation. B.A. stands for base alloy. (Online version in color.)

Samples with ZrO₂ and SiO₂ additions stand out of the rest of the trials. Their intragranular fraction is much lower which means that their grains are much more dendritic. It agrees well with the experimental grain morphologies seen in Figs. 5(e) and 5(g).

As the terminal velocity is low for SiO₂ particles while their misfit is high, they may easily follow the convection loops without acting as heterogeneous nucleation sites. The presence of those inert particles inside the liquid could then disturb the dendritic growth.⁴¹⁾ An indirect proof is provided by the presence of SiO₂ particles close to MnS inclusions (see Table 3 for a summary on the SEM characterizations). According to a simple Scheil calculation with Thermo-Calc (with 0.03 mass%S), those MnS inclusions form at high solid fraction (f_s>0.94) in 42CrMo4 steels. It would then indicate that SiO₂ particles remain in the liquid phase until high solid fractions.

Table 3. SEM observations in inoculated samples.

Inoculant	SEM observation
TiN	Observed, sometimes in cluster
CeO2	Few CeO2 observed
W2C	In the interdendritic area, enclosed in MnS
HfC	In the interdendritic area, enclosed in MnS
Si3N4	Not observed
Al2O3	Not observed
SiO2	In the interdendritic area
ZrO2	Not observed

Mean intercepted lengths are provided in Fig. 8. As the measurement is done on at least 300 randomly oriented lines, standard deviation values are very low and error bars are not represented because they would be too small.

Fig. 8.

Inoculation effect on secondary arm spacing. (Online version in color.)

Contrary to the grain size measurement in Fig. 6, the higher the misfit is, the smaller the mean intercepted length is. The relationship between the two parameters seems to be linear. To draw more precise conclusions, it is also needed to look at the morphology of the microstructure input in the intercept technique. For very dendritic structures, the mean intercepted length is smaller because all the secondary arms can intercept the measurement lines. This is the case for SiO₂ and ZrO₂ as confirmed by experimental observations in Figs. 5(g) and 5(e).

On the other hand, for highly globular structures, the mean intercepted length is closer to the grain size. It is a complementary result from the intragranular fraction measurement summarized in Fig. 7. For low misfit values, no effect on the secondary arm spacing is observed whereas for high misfit value, i.e. less efficient inoculant, secondary arm spacing refinement is observed. As proposed earlier, high misfit values like SiO₂ remaining particles could disturb the equiaxed dendrite growth. For medium range misfit values, as some particles are used to nucleate grains, the density of remaining particles inside the liquid would not be sufficient enough to trigger a change in morphology but it could still be enough to trigger a change in the length of the interdendritic arm spacing. Finally, no change would be observed for low misfits because all particles would be used as nucleation sites.

Inoculation could also slightly modify the secondary arm spacing by acting on the secondary dendrite arm ripening velocity. The ripening velocity is proportional to the cubic root of time with a proportionality coefficient that depends on the growth restriction factor, Q, the Gibbs-Thomson coefficient, Γ_s/l, and the diffusion coefficient in the liquid, D_l, as illustrated below:¹⁾   

$${\lambda }_2={({\text{Mt}}_{\text{f}})}^{1/3}\mathrm{\hspace{0.17em} }\mathrm{w}\mathrm{i}\mathrm{t}\mathrm{h}\mathrm{\hspace{0.17em} }\text{M}\propto \frac{{\mathrm{\Gamma }}_{\text{s}/\text{l}}{\text{D}}_{\text{l}}}{\text{Q}}$$

Such solutal refinement could affect the secondary arm spacing for the high solubility product inoculants Si₃N₄, HfC and W₂C. Further information on their effects on the Gibbs-Thomson coefficient along with a known value at steelmaking temperature of the diffusion coefficient would be needed to draw a more precise conclusion.

5. Microsegregation Results

With the previous results in mind, it is of wonder if the inoculation effect on the solidification structure is beneficial to the microsegregation profile. Microsegregation measurements on a grid of points are thereafter discussed.

5.1. Microsegregation Profiles

For W₂C and Si₃N₄ additions, tungsten and silicon mean values over the grid were respectively 0.25 mass% and 0.5 mass%. Those composition levels could mean that the W₂C addition dissolved up to 86% whereas the Si₃N₄ addition dissolved up to 96%. Moreover, no hafnium was found in the case of the HfC additions.

Microsegregation intensity scores – which indicate the extent of the microsegregation in between 5% and 80% solid fraction – are given in Fig. 9. For all the inoculation experiments, the microsegregation scores are less intense than in the base alloy. The three best scores (in the red ellipse) are achieved for HfC, W₂C and Si₃N₄ additions.

Fig. 9.

Influence of inoculation on the microsegregation score. B.A. stands for base alloy. (Online version in color.)

For the other samples located in the blue valley, the decreased microsegregation score could be mainly attributed to the refinement of the secondary arm spacings. While the secondary arm spacings are also finer for Si₃N₄, W₂C and HfC, other explanations have to be found.

One first difference is brought by the manganese microsegregation profiles (ranked by increasing Cr content) as presented in Fig. 10. For readability purpose, a moving average filter was applied with a width of 0.02 f_s. At very low solid fraction, the Mn content is higher than the content expected from Table 1 (0.72 mass%Mn). It is partly due to the moving average filter that averages each value with the ten preceding points. As a consequence, values for the first 10 points cannot be plotted. It has to be noted that the composition of the liquid phase prior to equiaxed solidification is not well known. At the bottom of each sample, columnar solidification enriches the liquid above the composition of Table 1. At high solid fraction, the noise is due to the precipitation of MnS inclusions. A preliminary comment would be to say that the mean manganese value of the base alloy seems higher than the rest of the inoculated samples. The calculation of average value is somehow difficult because the high solid fraction area (right of the curve) is highly dependent of the presence of manganese sulfide inclusions in the measurement area.

Fig. 10.

Mn reconstructed segregation profile. B.A. stands for base alloy. (Online version in color.)

This noise related to MnS inclusions is not present in the low microsegregation score samples – W₂C and HfC – it probably means that the manganese sulfide precipitation is playing a role in the reduction of the microsegregation score. To investigate that lead, new precipitation characterizations were proposed.

5.2. Additional Precipitate Characterizations

While precipitates presence is not directly linked to the solidification structures, it remains a powerful indicator of the outcome of the inoculant additions. Therefore, SEM characterizations were performed on all the samples to see if the inoculant particles could be observed in the bulk metal as summarized in Table 3.

At first sight, it can be seen that the additions were successful for most of the inoculants. Nevertheless, 2D SEM observations are not sufficient to clearly indicate that Si₃N₄, Al₂O₃ and SiO₂ additions were unsuccessful. If the number of added particles is very low, the probability to cut the ingot on a plane containing at least one particle coupled to the probability to actually see that particle at a proper magnification could be very low. Dissolution could also lead to a decrease in the number and size of the high solubility product particles like Si₃N₄.

W₂C and HfC particles were found in the columnar area in between the secondary dendritic arms. Furthermore, HfC particles were found encapsulated inside the MnS interdendritic inclusions. In Fig. 11, two examples are given. The Hf enriched phase appears in white, the MnS inclusion appears in dark grey and the matrix in light grey. Contrary to Fig. 11(a) with an Hf-rich core, the Hf enclosed particle in Fig. 11(b) is more complex. An Al-rich phase is enclosed in what seems to be the remaining of a former HfC particle.

Fig. 11.

Modified inclusions ((a) and (b)) after Hf enrichment observed in back scattered electron SEMmode. (Online version in color.)

It is clearly a sign that the addition of carbides is modifying the way manganese sulfide inclusions are precipitating. Consequently, additional investigations were done to assess if the MnS population were different between the inoculated sample and the base alloy. For that purpose the in-SEM automatic characterization tool MeTiS²³⁾ was used for the Si₃N₄, HfC and W₂C inoculated samples and the base alloy. The different results are summarized in Table 4.

Table 4. Properties of manganese sulfide inclusions.

Inoculant	Nv (1014 1/m3)	√Area (μm)	Elongation
Base alloy	1.35	2.7	0.57
Si3N4	1.18	1.7	0.47
W2C	1.74	4.4	0.47
HfC	1.48	1.8	0.50

The number and sizes of those manganese sulfides are varying from one sample to another. The manganese sulfides are refined and less elongated in comparison to the base alloy for the Si₃N₄ and HfC addition while their volumetric densities remain almost unchanged. On the other hand, there are bigger in size (39% raise) and number (22%) for the W₂C addition.

If this information is coupled with the carbides particles being found side by side or even enclosed inside MnS inclusions, it could mean that the precipitation sequence of MnS is modified in those samples. One possible explanation is that carbide and Si₃N₄ particles act as nucleation sites for the manganese sulfides at an earlier solidification stage. Due to the higher density of W₂C and HfC, those particles would sediment in the columnar zone and progressively deplete the liquid in both manganese and sulfur.

Sulfur removal is well known to decrease the solidification range as well as changing the surface energies.^42,43,44) The properties of the liquid would then be totally different and could explain why the microsegregation intensity is lower for inoculated samples where the grain size is not affected. Further investigations need to be done to verify that hypothesis.

6. Conclusion

This article provides a guideline to select new inoculants for low alloyed steels. The developed methodology is mainly focused on the calculation of three inoculant properties (misfit, solubility product and terminal velocity). Knowing those indicators is important to assess if an inoculant has sufficient chances of success.

With the help of this methodology, different inoculants were added into 42CrMo4 low alloyed steel. The modified ingots were carefully characterized in terms of grain size, morphology, mean intercepted length, presence of inoculant particles and microsegregation. The following propositions were made by using inoculant properties:

- The misfit seems to be the main criterion to predict grain size refinement. The best refinements were obtained for Si₃N₄, TiN and CeO₂.

- Other criterions are necessary to fully predict the inoculation outcome. While the solubility product was not of primary importance, high terminal velocity inoculants were proved to be inefficient in terms of grain refinement.

- The prediction of morphological and secondary arm spacing changes is not possible with the use of the misfit alone:

○ Medium and high misfit inoculants with low terminal velocities could remain in the liquid and disturb the dendritic growth because they were not used as nucleation sites. Inoculated microstructures would then be more dendritic with lower dendrite secondary arm spacing.

○ High solubility product inoculants regardless of their misfits could dissolve and modify the liquid properties resulting in smaller secondary arm spacing.

- Inoculation can modify the microsegregation establishment:

○ For high solubility product inoculant, the changes were mainly attributed to the secondary arm spacing refinement.

○ The microsegregation scores of high solubility product additions (W₂C, HfC and Si₃N₄) were lower than for any other additions. A change in the MnS precipitation sequence and sulfur depletion in the liquid could explain this effect. Further research on that matter needs to be done to validate the proposed mechanism. A flatter microsegregation profile before the end of solidification could be of matter for large ingots where the equiaxed grains settle in the liquid phase before the last solidification time.

Acknowledgments

This work was supported by Ascométal and ANRT France under a CIFRE Ph. D. fellowship [grant number 2015-1286].

List of Symbols

mass%: Mass percentage

MIS: Microsegregation intensity score

f_s: Solid fraction

${\delta }_{{(\text{hkl})}_1}^{{(\text{hkl})}_2}$: Misfit between two dense cristallographic planes (hkl)

d_[uvw]: Interatomic distance along the direction [uvw]

θ: Angle between two crystallographic directions

δ_δ: Misfit between particles and ferrite

δ_γ: Misfit between particles and austenite

T: Temperature

K(T): Reaction constant at the temperature T

$\mathrm{\Delta }{\text{G}}_{\text{r}}^0(\text{T})$: Gibbs free energy of dissolution at the temperature T

R: Ideal gas constant

Re: Reynolds number

p_s: Solubility product

v_T: Terminal velocity

g: Gravity acceleration

ρ: Density

r₅₀: Particle median radius

μ: Dynamic viscosity of liquid steel

Q: Growth restriction factor

Γ_S/L: Gibbs-Thomson coefficient

λ₂: Dendrite secondary arm spacing

t_f: Local solidification time

D_l: Diffusion coefficient inside the liquid phase

References

• 1)   J.  Dantzig and  M.  Rappaz: Solidification, Engineering Sciences, EFPL Press, Lausanne, (2009), 327, 477, 567–569.
• 2)   N.  Leriche: Ph.D. thesis, Université de Lorraine, (2015), http://docnum.univ-lorraine.fr/public/DDOC_T_2015_0276_LERICHE.pdf, (accessed 2019-05-09) (in French).
• 3)   J.  Delorme,  M.  Laubin and  M.  Maas: Casting and Solidification of Steel, IPC Science and Technology Press, Guildford, (1977), 213.
• 4)   A.  Kumar,  M.  Založnik,  H.  Combeau,  J.  Demurger and  J.  Wendenbaum: Proc. 1st Int. Conf. on Ingot Casting, Rolling and Forging, Jernkontoret, Stockholm, (2012), 3.
• 5)   A.  Kermanpur,  M.  Eskandari,  H.  Purmohamad,  M.  Soltani and  R.  Shateri: Mater. Des., 31 (2010), 1096.
• 6)   K.  Ayata,  T.  Mori,  T.  Fujimoto,  T.  Ohnishi and  I.  Wakasugi: Trans. Iron Steel Inst. Jpn., 24 (1984), 931.
• 7)   T. E.  Quested and  A.  Greer: Acta Mater., 53 (2005), 2683.
• 8)   L.  Elmquist,  S.  Salera and  A.  Diószegi: Int. J. Cast Met. Res., 23 (2010), 124.
• 9)   M.  Guo and  H.  Suito: ISIJ Int., 39 (1999), 722.
• 10)   M.  Mizumoto,  S.  Sasaki,  T.  Ohgai and  A.  Kagawa: Int. J. Cast Met. Res., 21 (2008), 49.
• 11)   M.  Ohno and  K.  Matsuura: ISIJ Int., 48 (2008), 1373.
• 12)   C.  Hu,  K.  Wu,  A.  Trotsan,  Y.  Li,  O.  Isayev,  O.  Hress and  I.  Rodionova: Metallurgist, 60 (2016), 888.
• 13)   N.  Reilly: Ph.D. thesis, Université de Lorraine, (2016), https://tel.archives-ouvertes.fr/tel-01535830/file/DDOC_T_2016_0264_REILLY.pdf, (accessed 2019-05-09), (in French).
• 14)   D. M.  Herlach: Crystals, 5 (2015), 355.
• 15)   T.  Picot: MoS thesis, Université de Lorraine, (2017), (in French).
• 16)   J.  Schindelin,  I.  Arganda-Carreras,  E.  Frise,  V.  Kaynig,  M.  Longair,  T.  Pietzsch,  S.  Preibisch,  C.  Rueden,  S.  Saalfeld,  B.  Schmid,  J.-Y.  Tivenez,  D. J.  White,  V.  Hartenstein,  K.  Elicierie,  P.  Tomancak and  A.  Cardona: Nat. Methods, 9 (2012), 676.
• 17)   K.  Zuiderveld: Graphics Gems, Academic Press Professional, San Diego, (1994), 474.
• 18)   J.  Michálek,  M.  Čapek,  X.  Mao and  L.  Kubínová: Biosignal 2010, INSTICC, Setubal, (2010), 49.
• 19)   M.  Gennesson,  J.  Zollinger,  D.  Daloz,  B.  Rouat,  J.  Demurger and  H.  Combeau: Materials, 11 (2018), 1007.
• 20)   M.  Založnik and  H.  Combeau: Proc. 12th Int. Conf. on Modeling of Casting, Welding, and Advanced Solidification Processes (MCWASP XII), TMS, Warrendale, PA, (2009), 165.
• 21)   S.  Van der Walt,  J. L.  Schönberger,  J.  Nunez-Iglesias,  F.  Boulogne,  J. D.  Warner,  N.  Yager,  E.  Gouillart and  T.  Yu: PeerJ, (2014), 2:e453. http://doi.org/10.7717/peerj.453
• 22)   W. S.  Monroe,  C.  Monroe and  R.  Foley: Mater. Charact., 127 (2017), 88.
• 23)   E.  Hénault: JEOL News, 41 (2006), 32.
• 24)   J.  Zollinger and  D.  Daloz: Mater. Charact., 62 (2011), 1058.
• 25)   B. L.  Bramfitt: Metall. Trans., 1 (1970), 1987.
• 26)   M.  Andersson,  J.  Janis,  L.  Holappa,  M.  Kivio,  P.  Naveau,  M.  Bradnt,  D.  Espinosa,  L.  Bellavia,  X.  Vanden Eynde,  E.  de Courcy,  L.  Chapuis,  T.  Iung,  S.  Ekerot and  C.  Van der Eijk: Grain Size Control by Means of Dispersed Non-Metallic Inclusions–GRAINCONT, European Comission, Brussel, (2011), 16.
• 27)   N. H.  Tyas: Ph.D. thesis, University of Cambridge, (2001), https://www.phase-trans.msm.cam.ac.uk/2012/Tyas_Thesis.pdf, (accessed 2019-05-09).
• 28)   C.  Lupis: Chemical Thermodynamics of Materials, Elsevier Science Publishing, New York, (1983), 169.
• 29)   A.  Rist,  M. F.  Ancey-Moret,  C.  Gatellier and  P. V.  Riboud: Elaboration de la fonte et de l’acier données thermodynamiques, Techniques de l’ingénieur, Saint-Denis, (2016), (in French).
• 30)   P.  Perrot: Propriétés thermodynamiques des composés minéraux, Techniques de l’ingénieur, Saint-Denis, (2002), (in French).
• 31)   D.  Janke and  W. A.  Fischer: Arch. Eisenhüttenwes., 49 (1978), 425.
• 32)   B. G.  Thomas,  Q.  Yuan,  S.  Mahmood,  R.  Liu and  R.  Chaudhary: Metall. Mater. Trans. B, 45 (2014), 22.
• 33)   W.  Tiller and  T.  Takahashi: Acta Metall., 17 (1969), 483.
• 34)   F.  Stouvenot and  H.  Gaye: Masse volumique, viscosité, tension superficielle et interfaciale des métaux, alliages et oxydes liquides, Technical Report IRSID, IRSID, Maizières-Les-Metz, (1995), RI 95.056 (in French).
• 35)   H.  Lamb: Hydrodynamics, Cambridge University Press, Cambridge, (1932), 599.
• 36)   J.  Miettinen and  S.  Louhenkilpi: Metall. Mater. Trans. B, 25 (1994), 909.
• 37)   J. P.  Roucan and  M. C.  Noël-Dutriaux: Propriétés physiques des composés minéraux, Techniques de l’ingénieur, Saint-Denis, (1987), (in French).
• 38)   Ø.  Grong,  L.  Kolbeinsen,  C.  van der Eijk and  G.  Tranell: ISIJ Int., 46 (2006), 824.
• 39)   G.  Heintze and  R.  McPherson: Weld. J., 65 (1986), 71s.
• 40)   M.  Easton,  C.  Davidson and  D.  St John: Metall. Mater. Trans. A, 41 (2010), 1528.
• 41)   J.  Sekhar and  R.  Trivedi: Mater. Sci. Eng. A, 147 (1991), 9.
• 42)   Z.  Liu,  Y.  Kobayashi,  M.  Kuwabara and  K.  Nagai: Mater. Trans., 48 (2007), 3079.
• 43)   B.  Keene: Int. Mater. Rev., 38 (1993), 157.
• 44)   K.  Mills,  B.  Keene,  R.  Brooks and  A.  Shirali: Philos. Trans. R. Soc. Lond. A, 356 (1998), 911.