Experimental assessment of the gray cast iron production by inoculant injection

An initial assessment of the gray cast irons production by injecting an inoculant with a conveying gas into a molten iron bath was evaluated at a laboratory scale. A numerical simulation was carried out to determine the hydrodynamic behavior between the inoculant particles injected into the molten iron. It was determined that an optimal interaction between the particles with the molten iron occurs at a lance depth of 7 cm and for the particle sizes fine (211 to 297 μ m) and medium (421 to 590 μ m), the residence time was of 0.38 and 0.4 s, respectively. The melting time was calculated at 0.0008 and 0.003 s for the particle sizes fine and medium, respectively. So, after injection, the FeSi of the inoculant melts quickly, releasing the elements of the inoculant which interact with the molten iron and forms oxides and sulfides creating nucleating sites during solidification. The injection technique allows obtaining a type-A graphite distribution for the fine and medium particle sizes. The number of eutectic cells was increased when the inoculant particle size was decreased despite of the low graphitisers elements, and manganese contents used in the gray cast iron manufacturing.


Introduction
Gray Cast iron has been produced mainly as a primary option for multiple automotive components during many decades because of its good properties as castability, machinability, heat conductivity, and vibration damping capacity, in combination with reasonable strength and low total cost of production [1,2]. A high liquid iron undercooling is required to form small size micro clusters that act as stable homogeneous nuclei for graphite particles; however, such high undercooling is very difficult to achieve in common foundry practices. In gray cast iron, the nucleation of graphite is mainly carried out by inoculation which improves the graphite dispersion, increasing the nucleation of eutectic grains by heterogeneous nucleation [3]. Inoculants are alloys added in small amounts to increase the number of active nuclei for the graphite nucleation and high effective grain refinement is achieved [4,5]. Inoculants are ferrosilicon alloys that may contain Al, Ca, Ba, Sr, Zr, Rare Earths, as well known as inoculant elements that promote and participate in the creation of micron-sized active compounds in the melt, to act as effective graphite nucleation sites [6]. So, inoculants are added to produce heterogeneous nucleation of these graphite flakes and obtain the desired distribution of them. Improving manufacturing processes for cast iron is always needed and many gray iron studies focused on theories of graphite lamellar nucleation [7,8], pre-inoculation treatments [8,9], microstructure and mechanical properties [10][11][12] and inoculation fading effect [13,14] have been widely investigated but, few studies have been focused on the inoculant addition method. There are two main methods of inoculation: ladle and late inoculation. In the former case, the inoculant is added either to the liquid iron flow pouring into the ladle or just afterward. Late inoculation refers to treatment after the metal has left the ladle, for example, placing the fine-grain inoculant in the pouring basin (stream inoculation) or by using an insert in the mold down gating system (in-mold inoculation) [15]. Another method barely explored consists of adding the inoculant powder by a carrier gas into the molten iron. Despite powder injection has become an important tool in metallurgical operations focused primarily on steelmaking processes. The injection efficiency depends mainly on the interaction (solid particle and conveying gas) with the liquid metal [16]. A. Derda and M. Soiński [17] added inoculants to hypo-eutectic gray cast iron by a pneumatic transportation device. They found high effectiveness and stability, based on the silicon composition changes after performing the inoculation through pneumatic injection of the powdered inoculant by argon as a carrier gas into the liquid. Other works related to the powder injection technique are focused on the lead refining process [18][19][20]. The injection efficiency depends mainly on the flow patterns. Particle concentration is often low; hence, their motion is determined by the fluid flow. However, particle trajectories are very important as they stablish the residence time, which represents the time required for the particles leaving the lance tip until they reach the liquid surface. As residence time increases, the injected particles will have more opportunity to react with the metal bath, and higher efficiencies are achieved [18,19]. The basic requirement for heterogeneous nucleation lies in the ability of the molten iron to wet the inoculant particles. Therefore, a highly effective inoculation is expected by injecting the inoculant by a conveying gas into the molten iron where more intimate contact between the particles and the melt will occur. This work deals with the manufacturing of gray iron through the injection of a powder commercial inoculant with argon as conveying gas. The operational injection parameters of the physical trials were set by a computational fluid dynamics (CFD) and particle tracing study, where the residence time as a function of the lance depth was determined while compared with the theoretical melting time. The gray cast iron produced was characterized by microscopy techniques to determine the graphite flake features based on the standard ASTM A247 as well as the eutectic cell characteristics. Figure 1 shows the experimental scheme used for the injection trials. A coreless induction furnace (50 kW and 9.6 kHz) with a capacity of 30 kg to melt the metallic charge was used. The temperature was measured with a B-type thermocouple dipped in the molten metal and not showed in figure 1. The injection system consisted of a sealed chamber where inoculant powder was charged and it was joined to a low carbon steel lance with an inner diameter of 1 cm. Argon of commercial purity was used as conveying gas, and the amount injected was controlled by a flowmeter. Two commercial inoculants base FeSi which contain the inoculating elements Zr, Al, Ca, Ba as strong graphitizing were used for the inoculation. Table 1 shows the chemical composition of the inoculants determined by the chemical absorption technique. The conventional inoculant "b" was used to obtain a reference heat by the ladle method, while the inoculant "a" was used to evaluate the inoculation method. The inoculant "a" as received was sieved to determine the particle size range. Figure 2 shows the particle size distribution of the commercial inoculant "a". As can be seen, the inoculant as received shows a wide particle size distribution in the range from 38 to 840 μm. Inoculant "a" powder particles were classified for the injection trials as fine and medium particles which contain a particle size in the range from 211 to 297 μm and from 421 to 590 μm, respectively. These two groups of particle sizes named fine and medium particles were considered as input for the numerical simulation as well as studied in the injection trials.

Numerical simulation
Numerical simulations were carried out to establish the hydrodynamic behavior in the lancefurnace crucible system to different lance depth. Velocity fields computed were used to determine the particle motion for particles in the presence of an external field and to find the residence time of the particles injected into the liquid iron. The numerical simulation was carried out on the CFD´s module of Comsol Multiphysics software [21]. The hydrodynamic behavior of the gas-liquid can be carried out with the continuity (Eq. 1) and Navier-Stokes equations (Eq. 2), whereas the tracking of the interface for the two-phase flow gas-liquid is obtained by using the phase-field method. In the solution domains, the continuity and momentum equations in the usual form were considered: Where σ1 is a deviatoric stress tensor and the field variables, velocity u and pressure p. The evolution of the interface is controlled by the phase-field variable () of the Cahn-Hilliard equation (Eq.3).
Where F/ = G is the chemical potential (Pa) and  is the mobility (m 3 s kg -1 ). Turbulence variables were obtained by the standard k ε  model, whereas the particle trajectories were described based on the Newton´s second law with a drag force estimated using the empirical equation proposed by Khan and Richardson [22]. Particles are considered to be spherical in shape. Detailed information about the mathematical model and numerical issues can be found in [18,22], where the same methodology was used to study the flow patterns (at lab scale) obtained during the lead and aluminum refining processes by silver and magnesium particle injection, respectively. The numerical simulation was carried out in two dimensions and the lance was located at the centerline of the furnace crucible, as can be seen in Figure 3 where the boundary conditions are also shown. The experimental conditions and the relevant physical data used for the CFD study are given in Table 2. The numerical modeling was carried out considering that the system is isothermal at the temperature of 1450 °C. The gas velocity was determined by preliminary injection trials considering enough gas flow rate. This allows the powder inoculant to be projected with enough kinetic energy, with the aim to overcome the liquid iron surface tension, and the particles penetrate into the molten bath. The lance deep was varied at 3, 7 and 9 cm from the liquid bath free-surface. The previous was done to study the effect of gas injection depth on the hydrodynamic behavior, and the residence time of fine and medium size particles, injected independently. Based on the simulation results, the better lance depth was chosen for the experimental injection trials.

Melting time determination
The residence time, determined in the numerical simulation, was complemented with the melting time calculation to understand the interaction between the inoculant particles and the molten iron bath. The melting rate is determined by the relative magnitude of the resistance to transfer heat from the surrounding to the solid metal. Considering an energy balance, the rate at which heat is transferred from the environment to the solid equals the rate of absorption of the latent heat of fusion [19]. The environment is considered as semi-infinite because the temperature gradient in the environment is negligible [25]. The time required for the full melting of a sphere in a semi-infinite medium can be calculated with equation (4) [25].
Where: R is the radius of the sphere, TO is the initial temperature of the medium; TM represents the metal melting temperature; ρs is the density of solid; ΔHs is the latent fusion heat; km represents the thermal conductivity of the medium; Cpm is the heat capacity of the medium: ρm is the density of the medium; tf represents the melting time and n is the form factor in heat conduction which is 2 for a sphere. For this analysis, the molten iron represents the medium and the inoculant particles, the solid. Table 3 shows the parameters considered for the melting time determination.

Injection trials
4 heats (designated as A, B, C, and D) of gray iron of 8 kg each were produced in an induction furnace. The base iron was prepared with gray iron scrap, low C steel, pig iron, FeSi, and high purity carbon riser to adjust the chemical composition. Table 4 shows the experimental conditions of the inoculants for the gray cast iron produced. 40 g of the commercial inoculant was used for inoculation (0.5 wt % of the mass charge) for the injection trials. The first heat (Heat A) was inoculated by the common ladle method by using a higher amount (1.0 wt % of the mass charge) of a conventional inoculant (FeSi alloy -Ca-Al-Ba), while the other three were inoculated by injecting a lower amount of a more effective inoculant (FeSi alloy -Ca-Al-Ba-Zr-Mn) through an argon flow as conveying gas, by using the experimental array depicted in Figure 1. Heats A and B were carried out by using the particle size of the inoculant as received which comprises a wide particle size range, while for heats C and D, the particle size range was narrowed. The injection trials were carried as follows: 40 g of the inoculant as received, which includes fine and medium particle size, were placed in the powder dispenser, after reach 1450 °C in the liquid bath, the induction furnace is shut down and then the lance is introduced into the iron bath. The valves of the injection system are opened and the inoculant is injected into the iron bath by an argon flow. After 7 s, the lance was taken away and immediately, the molten metal was transferred to a preheated ladle and then poured into green sand molds previously prepared.
The following experimental parameters were kept constant during injection trials: bath temperature 1450 °C, the distance between the lance and the liquid surface 7 cm, argon gas flow rate 10 L min -1 and injection time 7 s. Figure 4 shows the pattern used to the preparation of green sand molds, the dimensions of the plates are of 120 x 40 mm and a thickness ranging from 25.4 to 4.23 mm. The microstructural analysis was carried out to the cast plates of 12.7 and 25.4 mm of thickness for the four heats produced. microscope Olympus PMG-3 model. Unetched microstructure was used to evaluate the graphite flake morphology at 100x, according to the standard ASTM A247, to evaluate the size and distribution of the graphite flakes. The graphite length was determined by the software Carnoy 2.0, by selecting randomly thirty graphite flakes of three different regions for each heat and thickness. The eutectic cell size was measured according to the procedure reported by Fras and López [27] after etching the samples with Stead's reagent (2.5 g CuCl2, 10 g MgCl2, 5 ml HCl, 250 ml ethanol) during 2.5 hours. Etched microstructures were obtained with 3% nital, and phase volume fraction measurements including graphite, pearlite, ferrite, and carbides were carried out using the image-analyzer with the software Image J 4.1. Carbides were revealed by etching 2 min with a water solution of ammonium persulfate (10% vol) [28]. At least, four different fields of the samples were considered in the microstructural examination.    It is observed from figures 5 to 7 that at the beginning of the injection, the argon jet containing the inoculant particles expands rapidly and penetrates only a specific distance into the molten iron before rising vertically. As time proceeds, the bubble grows until it reaches a size where the buoyancy forces exceed the surface tension forces and detachment occurs. For longer times, when gas bubbles passage continuously through a free surface and then burst, this process is accompanied by jets shooting up, after jets reach their peak height and fall due to gravity, induce oscillations on the free surface causing its deformation by the wave propagation. Thus, the dynamics of the free surface in this stage produce the formation of droplets through way of jet drops and the bursting of the gas bubble. From figure 5 it is observed that when the lance deep is at 3 cm, the inoculant particles penetrate only a very short distance into the molten iron and the particles reach the molten iron surface very quickly, then most of the particles leave the molten bath without reacting, decreasing the inoculant efficiency. Then, a marked effect is not observed when the particle size is incremented from fine to medium size. When the lance depth is increased to 7 cm ( Figure 6), it is observed a better interaction of the particles with the molten iron regarding a lance depth of 3.0 cm. During injection, some particles are agglomerates; this effect was most clearly observed when the particle size was incremented, as was reported previously [29]. As the lance depth is increased until reach 9 cm, the turbulence is increased in the molten iron during the injection process, increasing the interaction between the inoculant particles and the liquid iron; however, the higher turbulence formed produces molten metal splashing that reaches the upper part of the induction furnace which causes loss of material and of course, a dangerous working condition during the gray iron fabrication. The residence time indicates the time when the particles leave the nozzle until a particle reaches the liquid surface. From figures 5 to 7, the residence time was determined in 0.16, 0.38, and 0.48 s for the fine particles (211 to 297 μm) for a lance depth of 3, 7, and 9 cm, respectively. When the particle size was increased to medium size (421 to 590 μm), the residence time was determined at 0.16, 0.40, and 0.50 s for a lance depth of 3, 7, and 9 cm, respectively. Therefore, the residence time is increased when the lance depth and the particle size are increased, which contributes to deeper penetration of the inoculant particles. Results from Figure 5 to 7 can be summarized as follows. When both the particle size and lance depth increase, for a fixed gas flow rate; the largest path followed by the particles within the lance in connection with its bigger particle size, allows for the particles to preserve their momentum when discharged. These promote the breakup of the gas-liquid interface, letting particles descend through the liquid more easily. Hence, particle trajectories inside the molten, during its rise to the liquid free surface are larger, giving more opportunity to the particles to interact with the molten. According to the hydrodynamic behavior, as well as the particle residence time showed in figures 5 to 7, the optimal interaction between the inoculant particles and the molten iron occurs at a lance depth of 7 cm. This lance depth was used during the experimental injection trials.

Melting time
Inoculants are added to the melt at a high superheat and they dissolve quickly after the addition. Wang et al [30] have shown that ferrosilicon particles will reach the temperature of the melt within a fraction of a second. The melting time was calculated by applying the equation (4) and data reported in Table 3. The melting point of the inoculant used at the injection trials is 1320 °C and it was added into molten iron at 1450 °C. For the 250 μm (R = 0.000125 m) in diameter particle size, equation ( The melting time is calculated by solving numerically the nonlinear equation (5). So, tf, the melting time for the 250 μm in diameter particles is 0.0008 s. For the 500 μm the melting time is 0.003 s. These calculated times represent the melting times for the fine (211 to 297 μm) and medium (421 to 590 μm) particle sizes, respectively. Figure 8 shows the relationship between the residence and melting times with the particle size. It is observed in figure 8 that the melting time is increased when the particle size increases. Two regions are observed in figure 8 (I and II), the first region represents the particle size range evaluated in this work, while the second region represents commercial inoculants with bigger particle size between 591 and 3000 μm that are used at industrial furnaces. Figure 8 also shows the residence time evaluated for the two particle sizes and three different injection depths. Thus, the inoculant is added to the iron bath and quickly melts; however, the inoculating elements contained in the inoculant interact with the oxygen and sulfur in the melt to form oxides and sulfides that allow the graphite nucleation during solidification. The interaction between the inoculating elements with the iron bath is longer as the residence time is increased. Experiments showed that the formation of oxides/sulfides at a certain temperature in the melt or during solidification depends on the concentration of the deoxidizers (Al, Sr, Ca, etc.) and the sulfide forming elements (Ca, Mo, Mn, etc.) contained in the inoculant [31]. Table 5 shows the nominal chemical composition of the inoculated gray irons produced. The results represent the average of five measurements. The carbon equivalent value (CE) is also reported showing that heats fabricated correspond to hypo-eutectic gray irons. The manganese and sulfur contents were kept low for the castings produced while the silicon content was also kept low for the castings B, C, and D. It is expected a low ability of graphitization of the produced castings due to the low amount of (Si, S, P), and manganese used in the chemical composition. Therefore, the obtained graphite features and microstructural characteristics are mainly attributed to the inoculation method and the inoculant used. Heat A was produced by a conventional ladle process with a higher inoculating amount (1.0 wt %), while the other heats were produced by the injection process with a lower inoculant (0.5 wt %) and graphitising elements to enhance the injection process on the graphite flake formation. However, the castings produced are in the range of chemical composition for typical unalloyed cast irons [32]. The metallurgical microstructures of the produced gray cast iron are shown in Figures 9 and 10 for the sample thickness 12.7 and 25.4 mm, respectively. It is observed from figure 9 that heat A shows a random distribution of uniform size flakes which is considered as a type-A graphite distribution based on the standard ASTM A247. This type of graphite is wanted in most of the cast iron applications. Heat B presents a type-E graphite distribution considered as an undesired distribution with preferred orientation and segregation of the graphite flakes. The heats C and D produced by injecting fine and medium inoculant particle size presented a type-A graphite distribution. It is observed that heat C contains a higher amount of graphite flakes than heat D and also the flakes are shorter in length.  The microstructural results are very similar when the sample thickness is increased from 12.7 to 25.4 mm for all heats, as can be observed in figure 10. The main difference between the microstructures obtained with a higher thickness is that the graphite flakes are longer and in a lower amount than those obtained for the thinnest thickness. Heat B showed a type-E graphite distribution for both thickness, this heat was manufactured by injecting the inoculant with the particle size as received. As it is shown in figure 2, the wide distribution of fine, medium and coarse particles reduced the effect of the inoculant, the coarsest particles present an agglomeration while the finest particles easily floated, both behaviors aids to have a low efficiency during inoculation. Table 6 summarizes the graphite flake features according to the standard ASTM A247. It is observed from table 6 that heat A considered as a reference presents the highest volume fraction of graphite for both thicknesses, followed by heat C, which presents 13.39 % of graphite in the plate of 12.7 mm. It has been reported [11] that the graphite flake volume varies from 11 to 16 % for hypo-eutectic gray cast irons with high carbon equivalent. Heat B shows the lowest volume fraction, followed by heat D. Heat B and D contained the biggest particle sizes and during the injection trials the particles are easily agglomerated aided by the hydrodynamics of the injection process; thus, an important amount of unreacted inoculant is removed before tapping the iron to the molds, decreasing the volume fraction of formed graphite. Therefore, the injection process requires a fine particle size to have an adequate interaction with the molten iron to obtain a high volume fraction of graphite and a type-A graphite distribution. Despite, heats C and D present the desired type-A graphite distribution with shorter flakes even considering the lowest silicon, manganese, and carbon contents used in the iron chemical composition. Stead´s reagent was used to reveal the eutectic cells in gray iron. The size and number of eutectic cells directly reflect the refinement of eutectic grains. Figures 11 and 12 show the eutectic cell features for the thickness of the samples of 12.7 and 25.4 mm, respectively. The number of eutectic cells mm -2 is summarized in Table 7.   The features of the eutectic cells are influenced by the chemical composition and the inoculation technique. Carbon and silicon have important effects on the number of eutectic cells, as the chemical composition of the iron approaches the eutectic point, the number of eutectic cells is increased [32]. The chemical composition of the produced gray cast irons is far from eutectic with an equivalent carbon of 3.44 in average for the heats produced by the injection technique. In such a way, the eutectic cell count is mainly aided by the inoculation technique. It is evident from figures 11 and 12, and table 7, that the number of eutectic cells is increased when the thickness is decreased due to a higher cooling rate than thicker samples. Heat C produced by injecting the fine size particles allows obtaining the highest number of eutectic cells. The inoculant injected as fine particle size provides more nucleation sites improving the graphitization of the melt. For the developed injection trials, the number of eutectic cells decreases as the particle size of the inoculant increases. A more refined microstructure is obtained when the number of eutectic cells is increased and the mechanical properties of the cast irons are enhanced [4,33].   The low amount of graphitiser elements (Si, P, S) and the manganese content in the heats allow obtaining a high amount of pearlite in the cast iron matrix. It is observed that the pearlite is finer for the 12.7 mm thickness plates because of a higher cooling rate than the 25.4 mm samples. The samples were etched with ammonium persulfate to reveal white regions attributed to carbides as cementite for the developed castings. In general, it was observed a low amount of cementite for the produced heats for both thickness and Figure 15 only presents the results for the C and D heats for both thickness evaluated. Also, it is shown that the amount of cementite is slightly increased when the thickness of the casting is decreased. Table 8 shows the volume fraction of phases formed in the heats produced. The inoculating by the injection technique allows obtaining an adequate graphite flake formation and together with the used inoculant, the free carbides formation was avoided. As a result, a lower amount of ferrite was obtained resulting in a fully pearlitic matrix. The inoculation technique promote high eutectic cell count, with favorable graphite features, and consequently, a high amount of pearlite was obtained for the low inoculant addition.

Conclusions
Gray cast irons were produced by injecting a powder inoculant with argon as a conveying gas into a molten iron bath. The results obtained are summarized as follows: 1. Despite the low graphitisers elements and manganese contents in the gray cast iron, it was possible to obtain a type-A graphite distribution for the particle sizes fine (211 to 297 μm) and medium (421 to 590 μm).
2. The number of eutectic cells was increased when the inoculant particle size and the casting thickness were decreased. 3. The hydrodynamic behavior and particle paths of the inoculant particles injected into molten iron with a conveying gas were determined by numerical simulation. 4. The particles residence time was determined as a function of different lance depth and particle sizes where an optimal interaction between the inoculant particles and the molten iron occurs at a lance depth of 7 cm. 5. The inoculant injected into the iron bath quickly melt based on the melting time determination and then, the inoculating elements (Ca, Al, Mn, Zr, Ba) are released to form oxides and sulfides that have adequate interaction with the iron bath based on the hydrodynamic behavior and the residence time. 6. The injection technique and the used inoculant promote adequate graphite features that lead to lower amounts of undercooled graphite, free cementite, and ferrite phase which could represent an attractive option for the production of gray cast iron.                Table 1. Chemical composition of the inoculant. Table 2. Parameters considered in the numerical simulation. Table 3. Parameters considered for the melting time calculation. Table 4. Experimental conditions of the inoculants for the gray cast irons produced. Table 5. Chemical composition of the fabricated heats. Table 6. Graphite features of the fabricated gray irons. Table 7. Eutectic cells for the heats produced. Table 8. Volume fraction of phases formed.