The Role of Microstructure on Tensile Plastic Behavior of Ductile Iron GJS 400 Produced through Di ﬀ erent Cooling Rates, Part I: Microstructure

: A series of samples made of ductile iron GJS 400 was cast with di ﬀ erent cooling rates, and their microstructural features were investigated. Quantitative metallography analyses compliant with ASTM E2567-16a and ASTM E112-13 standards were performed in order to describe graphite nodules and ferritic grains. The occurrence of pearlite was associated to segregations described through Energy Dispersive X-ray Spectroscopy (EDS) analyses. Results were related to cooling rates, which were simulated through MAGMASOFT software. This microstructural characterization, which provides the basis for the description and modeling of the tensile properties of GJS 400 alloy, subject of a second part of this investigation, highlights that higher cooling rates reﬁnes microstructural features, such as graphite nodule count and average ferritic grain size.


Introduction
Ductile Irons (DIs) are ternary Fe-C-Si alloys in which graphite forms as spheroidal particles (nodules), allowing for a good compromise between mechanical properties and a low production cost [1,2]. The number of graphite nodules and their shape are the result of a various technological factors which influence cooling rate and physicochemical state of the liquid metal [1,3,4]. The cooling rate is mainly affected by the wall thickness, the thickness of the neighboring parts of the casting section, and the initial temperature of the metal and mold and the mold material to absorb heat. The physicochemical state of the liquid metal is in turn affected primarily by the chemical composition, charge materials, furnace atmosphere, holding time, liquid metal superheating, preconditioning, spheroidization, and inoculation processes used in the foundry practice [1][2][3][4][5][6][7][8][9][10][11]. The cooling conditions under which the eutectoid reaction takes place together with alloying elements influence the metallic matrix microstructure [12][13][14]. So, the production route to design and shape optimal ductile iron microstructures with proper mechanical properties is very complex, involving aforementioned different factors, as well as implemented heat treatment conditions [15,16].
Silicon is a graphitizer element which hinders the occurrence of iron carbide. Its effect is estimated via the CE (Carbon Equivalent) relationship, CE = %C + 1/3%Si. A CE value of 4.26 denotes the eutectic composition [1]. Silicon seems to play a negligible role in determining the ferrite grain size, and it can segregate around the graphite nodules, thus being a possible cracking site [17].
Copper is a common alloying element in DI because of its graphitizing effect. It promotes pearlite formation, in particular, when coupled with small Mn additions [18].
The chemical composition together with the cooling conditions after casting affect the microstructure of the alloy. A number of parameters, describing the cooling curve, can be found in literature (a list is provided in [1,2]) that may be related to the graphite shape. In this work, the transformation temperatures and the corresponding undercooling will be taken into account. The inoculation practice and the cooling rate cooperates to control the nodule count, while the conditions under which the eutectoid reaction takes place influence the matrix microstructure [12,13].
The tensile plastic behavior of ductile iron is very sensitive to microstructure and casting defects. In this connection, strain hardening analysis is a powerful tool to study the effect of microstructure on its tensile plastic behavior of ductile iron. Angella et al. [19] shows that the dislocation-density-related Voce equation describes properly the correlations between strain hardening and microstructure of metallic alloys. From published literature [19][20][21][22][23][24], there is limited information on the effect of microstructure on tensile plastic behavior of ductile iron in terms of the strain hardening effect and micro-mechanisms occurring during deformation of its microstructure. Hence, the tensile flow curves modeling associated with an explicit correlation between plastic behavior and some microstructure parameters have not yet been clearly disclosed. This work, which provides the microstructural basis for the description and modeling of the tensile behavior of GJS 400 alloy [25], will investigate the correlation between the cooling rates near eutectic and eutectoid transformations and the microstructural features of the alloy. Cooling rates are estimated through MAGMAsoft v.5.3 taking into account the solidification of actual samples.

Materials and Methods
The chemical composition of the GJS 400 produced by Zanardi Fonderie S.p.A. (Minerbe-VR, Italy) is reported in Table 1. Carbon and sulfur contents were measured through a combustion infrared detection technique with a LECO CS744 by LECO (St. Joseph, MI, USA), while the other elements were detected by optical emission spectrometer with a ARL3460 by Thermo Fisher Scientific (Waltham, MA, USA). The value of CE is 4.45%, which makes the alloy hypereutectic. The residual Mg is 0.046%, which allows for graphite spheroidization [2]. Nodularization treatment was performed in a tundish cover ladle, using a Fe-Si-Mg alloy (Si 45 wt%, Mg 6,5 wt%), together with the alloying elements needed to achieve the desired chemical composition. After alloying, the melt was gravity poured in horizontal green sand molds (silica sand with clay and sea-coal addition, plus 3.5% water to activate clay), shaped with a pattern plate and formed with a green sand molding plant, in order to obtain the following samples complying with EN 1563 standard [26], namely ( Figure 1): 1.
a Lynchburg sample with 25 mm diameter; and 2.
this simulation are the 3D geometry of the casting system, the chemical composition of the alloy, the thermophysical parameters of the materials involved and alloy-mold and mold-environment heat transfer coefficients. The thermophysical parameters of the green sand, in particular the thermal conductivity, used for the simulation were determined by Zanardi Fonderie S.p.A. through an extensive experimental campaign aimed at the fine tuning of the parameters governing the heat fluxes. The actual set up of the gravity casting process was taken into account.
(a) (b) The liquid metal was poured into the molds through the pouring basin and then, by mean of the gating system, it filled the cavity of all the samples. Specimen for metallographic analyses were taken in the lower part of the samples (see Figure 1). In particular, six specimens from each samples were investigated through Scanning Electron Microscopy (SEM) with a SU70 microscope by Hitachi (Tokyo, Japan) equipped with Energy Dispersive X-ray Spectroscopy (EDS) detector (Noran 6 system by Thermo Fisher Scientific (Waltham, MA, USA) for elemental microanalysis. The acceleration voltage was 20 kV and the working distance was about 15 mm. After conventional mechanical polishing, the samples were etched with Nital 10% for 5 s to highlight the grain boundaries of the ferritic matrix and the pearlitic islands. Nodule count, nodularity, average diameter of the graphitic nodules, and volumetric fractions of graphite and pearlite were determined through Digital Image Analysis, by means of ImageJ software [27], of SEM images complying with ASTM standard E2567-16a [28], whilst the determination of the average ferritic grain size was carried out through OM complying with ASTM standard E112-13 [29]. ASTM standards were chosen because to the authors' experience they are more commonly used.
ASTM standard E2567-16a requires that at least 500 graphite particles with a minimum MFD (Maximum Feret diameter) of 10 µm must be analyzed. A particle with a shape factor (ratio between the area of the particles and the area of the reference circle, this latter being related to the MFD) higher than 0.60 is defined as a nodule. Nodularity is then defined as the ratio between the total area of the nodules and the total area of the graphite particles. Nodule count is given by the ratio of the nodules and the test area, expressed in mm 2 .
Grain size measurement were performed through the Hilliard single-circle procedure described in the ASTM standard E112-13. A single circle was blindly applied on at least five fields. A minimum of 35 intercepts between the circle and the grain boundaries is required. The ASTM grains size G is calculated as a function of the mean intercept, i.e., the ratio of the test line and the number of intercept. The average grain size can be thus calculated.
Since no direct measurement was possible, simulations of temperatures during cooling were performed through the Iron Module of the commercial software MAGMASOFT v5.3 by MAGMA (Aachen, Germany) in order to correlate cooling conditions with the microstructure. The inputs for this simulation are the 3D geometry of the casting system, the chemical composition of the alloy, the thermophysical parameters of the materials involved and alloy-mold and mold-environment heat transfer coefficients. The thermophysical parameters of the green sand, in particular the thermal conductivity, used for the simulation were determined by Zanardi Fonderie S.p.A. through an extensive experimental campaign aimed at the fine tuning of the parameters governing the heat fluxes. The actual set up of the gravity casting process was taken into account.

Simulated Cooling Curves
The molten metal experienced significantly different solidification rates. Simulations of the casting system consisting of molten metal poured into sand molds were performed in Zanardi Fonderie S.p.A., and the cooling curves are reported in Figure 2. Data refer to the barycenter of the lower portion of the samples, where the specimens for metallographic analyses were taken. Eutectic (T s ) and eutectoid (T e ) equilibrium temperatures can be estimated on the basis of the chemical composition [14,30]: (1) T e = 739 • C + 18.4%Si + 2%Si 2 -14%Cu -45%Mn + 2%Mo -24%Cr -27.5%Ni + 7.1%Sb = 787.8, • C (2) where "%el" represents the weight content of the element in the alloy. These equations hold for Si content up to 3 wt%, Mn, Cu, Cr, Ni content up to 1 wt%, and Mo content up to 0.5 wt% [14].

Simulated Cooling Curves
The molten metal experienced significantly different solidification rates. Simulations of the casting system consisting of molten metal poured into sand molds were performed in Zanardi Fonderie S.p.A., and the cooling curves are reported in Figure 2. Data refer to the barycenter of the lower portion of the samples, where the specimens for metallographic analyses were taken. Eutectic (Ts) and eutectoid (Te) equilibrium temperatures can be estimated on the basis of the chemical composition [14,30]: (1) Te = 739°C + 18.4%Si + 2%Si 2 -14%Cu -45%Mn + 2%Mo -24%Cr -27.5%Ni + 7.1%Sb = 787.8, °C (2) where "%el" represents the weight content of the element in the alloy. These equations hold for Si content up to 3 wt%, Mn, Cu, Cr, Ni content up to 1 wt%, and Mo content up to 0.5 wt% [14].
It can be seen (Figures 2 and 4) that in the neighborhood of transformation temperatures the slope of the cooling curves varies abruptly because of the exothermic nature of eutectic and eutectoid transformation upon cooling. It can also be seen that for the Lynchburg sample at about 1000 °C the alloy experiences a reduction in cooling rate, which is an effect of the solidification occurring in the feeder. As shown in Figure 3, indeed, the temperature decreases slower when the metal in the feeder undergoes solidification, an effect that disappears once solidification is complete. This phenomenon is not apparent in other molds because of their different geometries, and it is thought that it does not affect significantly microstructural features because it occurs far from the transformation It can be seen ( Figure 2 and Figure 4) that in the neighborhood of transformation temperatures the slope of the cooling curves varies abruptly because of the exothermic nature of eutectic and eutectoid transformation upon cooling. It can also be seen that for the Lynchburg sample at about 1000 • C the alloy experiences a reduction in cooling rate, which is an effect of the solidification occurring in the feeder.
As shown in Figure 3, indeed, the temperature decreases slower when the metal in the feeder undergoes solidification, an effect that disappears once solidification is complete. This phenomenon is not apparent in other molds because of their different geometries, and it is thought that it does not affect significantly microstructural features because it occurs far from the transformation temperatures. Cooling rates near Ts and Te (eutectic and eutectoid equilibrium temperatures, respectively) are given in Figure 4a,b, respectively. Table 2 summarizes cooling rates at the transformation temperatures, together with the undercooling experienced by the four samples, calculated as the difference between the eutectic temperature according to Equation (1) and the minimum temperature at the beginning of solidification.   Figure 4 and Table 2 show that the Lynchburg sample provided the fastest solidification rate, while at the eutectoid temperature the cooling rate is the second highest. It is worth noting that the  Cooling rates near T s and T e (eutectic and eutectoid equilibrium temperatures, respectively) are given in Figure 4a,b, respectively. Table 2 summarizes cooling rates at the transformation temperatures, together with the undercooling experienced by the four samples, calculated as the difference between the eutectic temperature according to Equation (1) and the minimum temperature at the beginning of solidification.  Table 2 show that the Lynchburg sample provided the fastest solidification rate, while at the eutectoid temperature the cooling rate is the second highest. It is worth noting that the differences in cooling rates are much higher at the eutectic temperature (there is a factor of about 20 between the highest and the lowest cooling rate), while at the eutectoid temperature they are comparable (only a factor of about 3). Moreover, variations in cooling rates are much higher in the proximity of the eutectic temperature rather than around eutectoid temperature, as an effect of reduced heat transfer from metal to heated mold. Cooling rates near Ts and Te (eutectic and eutectoid equilibrium temperatures, respectively) are given in Figure 4a,b, respectively. Table 2 summarizes cooling rates at the transformation temperatures, together with the undercooling experienced by the four samples, calculated as the difference between the eutectic temperature according to Equation (1) Figure 4 and Table 2 show that the Lynchburg sample provided the fastest solidification rate, while at the eutectoid temperature the cooling rate is the second highest. It is worth noting that the differences in cooling rates are much higher at the eutectic temperature (there is a factor of about 20

Microstructure
In Figure 5, representative SEM micrographs from Secondary Electron Imaging (SEI) of GJS 400 produced from the four different samples are reported. With slower solidification rates ( Figure 4a) the microstructure became apparently coarser, with an evident increase of nodule size, while pearlite was present only in the specimens from Y-block samples (Figure 5b-d), and barely detectable in the specimens from Lynchburg sample (Figure 5a). This qualitative description can be supported through quantitative measurements according to ASTM standard E2567-16a [16]. Table 3  between the highest and the lowest cooling rate), while at the eutectoid temperature they are comparable (only a factor of about 3). Moreover, variations in cooling rates are much higher in the proximity of the eutectic temperature rather than around eutectoid temperature, as an effect of reduced heat transfer from metal to heated mold.

Microstructure
In Figure 5, representative SEM micrographs from Secondary Electron Imaging (SEI) of GJS 400 produced from the four different samples are reported. With slower solidification rates ( Figure 4a) the microstructure became apparently coarser, with an evident increase of nodule size, while pearlite was present only in the specimens from Y-block samples (Figure 5b-d), and barely detectable in the specimens from Lynchburg sample (Figure 5a). This qualitative description can be supported through quantitative measurements according to ASTM standard E2567-16a [16]. Table 3 presents the results of image analysis, showing measurements on graphite features, defined in Section 2, and calculations on the volume fractions of the constituents. Together with the mean values, individual values measured on each specimen from each of the four samples are given.    In Figure 6a,b, SEM micrographs of a pearlite island in GJS 400 from Y 25 mm sample are reported. The clear lamellar pattern, i.e., parallel lamellae at an almost uniform distance, that can be seen in Figure 6b is not frequent, since pearlite often shows a complex configuration, in which the lamellar structure is irregular. Therefore, the characteristic widths of ferritic channels in the pearlitic islands could not be measured and can only be estimated to span between 100 and 300 nm, independently of cooling rates. In Figure 6a,b, SEM micrographs of a pearlite island in GJS 400 from Y 25 mm sample are reported. The clear lamellar pattern, i.e., parallel lamellae at an almost uniform distance, that can be seen in Figure 6b is not frequent, since pearlite often shows a complex configuration, in which the lamellar structure is irregular. Therefore, the characteristic widths of ferritic channels in the pearlitic islands could not be measured and can only be estimated to span between 100 and 300 nm, independently of cooling rates.

EDS Analyses
The local chemical composition of GJS 400 specimens from the four different samples was investigated through EDS. In particular, the concentration gradient of Si and Mn between couples of graphitic nodules was considered. Results are significantly different whether or not pearlite is present. Figure 7 shows a typical example of Si and Mn content in the region between two nodules separated by a pearlitic island (Y 75 mm sample). The Mn enrichment (positive segregation) and Si depletion (negative segregation) throughout pearlite is a common feature shown by every specimen, independently of the mold geometry. When there is no pearlite, neither Si nor Mn shows composition gradient (Figure 8).
It has to be pointed out that the EDS probe overestimated the Mn content, which is about 0.1% (Table 1). This is thought to be an issue of EDS analysis itself, since it is difficult to determine the quantity of trace elements (concentration lower than about 1% wt). Mn content is indeed low and this could affect the absolute values given by the EDS measurements. Its gradient, though, can be considered significant.

EDS Analyses
The local chemical composition of GJS 400 specimens from the four different samples was investigated through EDS. In particular, the concentration gradient of Si and Mn between couples of graphitic nodules was considered. Results are significantly different whether or not pearlite is present. Figure 7 shows a typical example of Si and Mn content in the region between two nodules separated by a pearlitic island (Y 75 mm sample). The Mn enrichment (positive segregation) and Si depletion (negative segregation) throughout pearlite is a common feature shown by every specimen, independently of the mold geometry. When there is no pearlite, neither Si nor Mn shows composition gradient (Figure 8).

Discussion
The GJS 400 microstructures are consistent with the simulated solidification rates (Figure 4), so that microstructural features result finer when cooling rates are higher (Table 3), in agreement with what reported in literature [10,31]. Nodule count measurement as a function of cooling rate at Ts (Figure 9) is consistent with the relationship found by Górny et al. in ductile iron with no Cu addition [10]. The presence of Cu in the alloys investigated in this work could account for the increase of nodule count at the same cooling rate. It has to be pointed out that the EDS probe overestimated the Mn content, which is about 0.1% (Table 1). This is thought to be an issue of EDS analysis itself, since it is difficult to determine the quantity of trace elements (concentration lower than about 1% wt). Mn content is indeed low and this could affect the absolute values given by the EDS measurements. Its gradient, though, can be considered significant.

Discussion
The GJS 400 microstructures are consistent with the simulated solidification rates (Figure 4), so that microstructural features result finer when cooling rates are higher (Table 3), in agreement with what reported in literature [10,31]. Nodule count measurement as a function of cooling rate at T s (Figure 9) is consistent with the relationship found by Górny et al. in ductile iron with no Cu addition [10]. The presence of Cu in the alloys investigated in this work could account for the increase of nodule count at the same cooling rate. Higher cooling rates around eutectic temperature also lead to higher undercooling, which can be in turn fairly related to nodule count and nodule mean diameter ( Figure 10). Higher cooling rates around eutectic temperature also lead to higher undercooling, which can be in turn fairly related to nodule count and nodule mean diameter ( Figure 10).

Figure 9.
Nodule count (NA) as a function of cooling rate (C) at the eutectic temperature Ts (red dots). The black line represents the relationship between cooling rate and nodule count in [10].
Higher cooling rates around eutectic temperature also lead to higher undercooling, which can be in turn fairly related to nodule count and nodule mean diameter ( Figure 10). Volume fraction of graphite, nodule count, and the mean nodule diameter, listed in Table 3, can be used to calculate the mean distance λ between graphite nodules through the Fullman's equation [32]: Where is the volume fraction of graphite, is the nodule count, and d is the mean diameter of the nodules.
The mean values for the four molds calculated through Equation (3) are reported in Table 4.  Volume fraction of graphite, nodule count, and the mean nodule diameter, listed in Table 3, can be used to calculate the mean distance λ between graphite nodules through the Fullman's equation [32]: where V g is the volume fraction of graphite, N A is the nodule count, and d is the mean diameter of the nodules.
The mean values for the four molds calculated through Equation (3) are reported in Table 4. The mean value for the Y 50 specimens is slightly higher than the one for the Y 75, despite the higher cooling rate, mainly because of the higher graphite volume fraction ( Table 3).
The graphite content (Table 3, V g in Equation (3)) is consistent with the Wojnar estimation [33] based on the carbon content of the alloy: Being %C the weight content of the alloy (3.63%), Equation (4) predicts a graphite volume fraction of 11.7%.
As already found in literature [20], ferritic grain size decreases when solidification rate increases ( Figure 11).
Being %C the weight content of the alloy (3.63%), Equation (4) predicts a graphite volume fraction of 11.7%.
As already found in literature [20], ferritic grain size decreases when solidification rate increases ( Figure 11). While no apparent composition gradients can be seen in ferritic grains, pearlitic islands show positive Mn segregation and negative Si segregation (Figures 7 and 8), in agreement with literature [34][35][36]. These gradients are related to what happens during eutectic solidification. Mn is continuously rejected from the solidification front to the melt metal, making Mn content increase in the last to freeze zone, namely, the grain boundaries between nodules. Mn, as well as other carbide forming elements like Cr and V at the left side of Fe in the periodic table, promotes pearlite, which explains why it is found in pearlitic islands.
On the other hand, Si, which promotes graphite formation in the melt metal, tends to remain in the first to freeze zone, around the graphite nodules, promoting ferrite.
After solidification, solid state transformations take place. In particular, ferrite nucleates and grows in austenitic grains, which transforms into ferrite and graphite. If cooling is fast enough, thus allowing for larger undercooling, the eutectoid transformation occurs and pearlite forms [2,14]. Table 2 and Figure 4b show that cooling rates at the eutectoid temperature were low for all the four samples, and this is consistent with the very low pearlite volume fractions found. In the Lynchburg mold pearlite it is barely detectable, even if the cooling rate at the eutectoid transformation was higher than in Y 50 and Y 75 samples. This suggests that a major role was played by cooling rate at solidification which, at the eutectic temperature, was much higher in the Lynchburg mold. This may have reduced Mn segregations, thus lowering the pearlite content. So, pearlite may be the product of segregations during solidification rather than the result of different cooling rates through the intercritical interval Ar1-Ar3. While no apparent composition gradients can be seen in ferritic grains, pearlitic islands show positive Mn segregation and negative Si segregation (Figures 7 and 8), in agreement with literature [34][35][36]. These gradients are related to what happens during eutectic solidification. Mn is continuously rejected from the solidification front to the melt metal, making Mn content increase in the last to freeze zone, namely, the grain boundaries between nodules. Mn, as well as other carbide forming elements like Cr and V at the left side of Fe in the periodic table, promotes pearlite, which explains why it is found in pearlitic islands.
On the other hand, Si, which promotes graphite formation in the melt metal, tends to remain in the first to freeze zone, around the graphite nodules, promoting ferrite.
After solidification, solid state transformations take place. In particular, ferrite nucleates and grows in austenitic grains, which transforms into ferrite and graphite. If cooling is fast enough, thus allowing for larger undercooling, the eutectoid transformation occurs and pearlite forms [2,14]. Table 2 and Figure 4b show that cooling rates at the eutectoid temperature were low for all the four samples, and this is consistent with the very low pearlite volume fractions found. In the Lynchburg mold pearlite it is barely detectable, even if the cooling rate at the eutectoid transformation was higher than in Y 50 and Y 75 samples. This suggests that a major role was played by cooling rate at solidification which, at the eutectic temperature, was much higher in the Lynchburg mold. This may have reduced Mn segregations, thus lowering the pearlite content. So, pearlite may be the product of segregations during solidification rather than the result of different cooling rates through the intercritical interval Ar1-Ar3.

Conclusions
Different microstructures of GJS 400 were obtained through different geometries leading to different cooling rates, which were calculated through simulation of the actual gravity casting system. The microstructures were characterized in details, quantifying nodule count, nodularity, average diameter of the graphitic nodules, and volumetric fractions of graphite and pearlite compliant with the minimum requirements of statistics of the standard ASTM E2567-16a [28], and the average ferritic grain size complying with the standard ASTM E112-13 [29]. These features result finer as the solidification rate increases.
Positive segregation (enrichment) of Mn and negative segregation (depletion) of Si was observed in the pearlitic islands.
The cooling rates around the eutectoid temperature were very similar and very low, which prevented pearlite formation. Data suggest that the occurrence of pearlite is related to segregations during solidification, rather than to cooling rates at the eutectoid temperature.
This microstructural characterization provides the basis for the description and modeling of the tensile properties of GJS 400 alloy, the subject of a second part of this investigation [25].