This paper presents the results of thorough evaluations of factorial experiments on the effects of varying boron and titanium contents, plus ‘plain’ thermal analysis (TA) cup type (‘round’ vs. ‘square’ cavity) on the properties of hypoeutectic gray iron. It quantifies the Occam’s razor effects of digital image analysis metallography metrics of gray iron microstructure and ultrasonic velocity to other more quotidian properties. 

The factorial experiments were comprised of nine distinct sample groups for three distinct levels of both ferroboron and ferrotitanium additions, with both ‘round’ and ‘square’ cavity TA cups poured from a single sample spoon of ~25 lb. capacity for each of the nine groups. Two sets of the nine sample groups were also cast consecutively for ample replication of each group to attain adequate statistical significance. 

The response metrics in addition to those of digital image analysis metallography and ultrasonic velocity were iron chemistries, Brinell hardnesses of the ultrasonic coupons, and mechanical properties of miniature test bars machined from the ultrasonic coupons. One original aspect of this research was the host foundry’s use of digital image analysis (IA) metallography for quantification of the graphite morphology of gray iron which is geometrically more complex than the graphite morphology of ductile iron.

Introduction

Prior research published by the host foundry reported on the elemental effects of titanium, nitrogen and manganese on the visible microstructure of gray iron coupons analyzed by digital image analysis metallography of which representative mold and cast coupon are shown in Figure 1. That original research correlated those Occam’s razor microstructural effects to the aging responses and metallurgical response metrics of gray iron. We then follow that with this current research, which quantifies the main effects of boron and titanium, plus ‘plain’ thermal analysis (TA) cup type (“round” versus “square” cavity), and their interaction effects on gray iron metallurgy. Furthermore, although boron and titanium are present in small concentrations in gray iron, they are nonetheless known to have significant effects on the physical and mechanical properties of gray iron. However, no published research prior to this paper has reported on interaction effects between those two elements. 

An earlier paper by the host foundry reported on the correlations of visible and quantified digital image analysis metallography metrics of flake graphite and matrix morphologies to other metrics of gray iron physical and mechanical properties whereby the independent (main) and interaction effects among titanium, nitrogen and manganese were also quantified. Thus, as logical progression to that prior paper, both boron and titanium additions were two independent variables considered in this current research for quantifying those elements’ effects on numerous properties of gray iron. A review of the literature on the effects of boron on the physical and mechanical properties of cast irons and mild hypoeutectoid steels indicates differing effects on those different ferrous materials. One particularly salient difference as reported in the literature is that boron strengthens mild hypoeutectoid steel that is devoid of free graphite particles, but it is reputed by many others to soften cast irons having the free graphite. 

Since the major difference between the cast irons and hypoeutectoid steel is that the cast irons contain a large volume of free graphite particles while the steel does not, it then follows that interaction between the cast iron matrix and graphite particles at elevated solid-state temperatures could be a major factor in boron’s effects on cast irons, which are not applicable for steel. Regarding that, one might surmise that the eutectoid ferritization––which occurs on the periphery of graphite particles enveloped by the matrix structure in cast irons––could be a significant metallurgical factor that is not applicable to steel. Accordingly, it must be appreciated that every graphite particle is a nucleation site for eutectoid ferritization.

In consideration of the fact that the graphite flakes of gray iron are geometrically more complex than the graphite nodules of ductile iron, several points to consider when selecting image analysis parameters for gray iron are as follows:

The three methods for measuring flake length are defined as follows:
1. “Serpentine” measures the tortuous path length along the main branch of the flake.
2. “Total” measures the total of the tortuous path lengths for all branches of a continuous flake.
3. “Tip-to-tip” measures the straight-line length from highest to lowest extreme points on the main branch of the flake. 

Although the current ASTM A247-24 does not specifically prescribe any of the three image analysis methods for measuring metrics of flake geometry, its original ASTM A247-67 version from the year 1967 cites the serpentine method in its section regarding the eight size classifications. We cite the obsoleted ASTM A247-67 version here because all subsequent revisions from the ensuing 57 years omit any mention, whatsoever, of any of the three geometric methods for quantifying flake graphite morphology. It is also noteworthy that the current ASTM A247-24 capitulates in acknowledging that it cannot be applied to digital image analysis metallography per its title change to Standard Test Method for Visual Evaluation of Graphite in Iron Castings. (The word “visual” in that title indicates that the standard does not apply to digital image analysis.)

For any individual flake particle, the measured length is always greatest for the ‘total’ method and least for the “tip-to-tip” method. 

The ‘serpentine’ method measures values between those of the ‘total’ and ‘tip-to-tip’ methods, but serpentine values are typically just slightly less than those for the ‘total’ method. That is because both the serpentine and total methods measure perimeters for length rather than a straight-line distance as is used for the tip-to-tip method. 
Regarding the minimum flake length criterion for the filtering/exclusion of the smallest particles in gray iron, note that the smallest Size Class 8 per ASTM A247-67 has a maximum (serpentine) length of 10 microns, so using a minimum length of 10 microns for any such analysis would exclude all Size Class 8 flakes from consideration. Since experience has shown that the lack of any size filter tends to inflate the number of Size Class 8 flakes present, a good minimum flake length might be 5 microns since that is half the maximum Size Class 8 length of 10 microns. 
The real minimum flake size per the minimum threshold size parameter will vary somewhat between the three flake measurement methods since the serpentine and total lengths will always be greater than tip-to-tip length for any individual flake. When a minimum particle size filter is used, a greater number of flakes are disproportionately filtered out for inadequate size by the tip-to-tip method and, to a lesser extent, by the serpentine method. Therefore, it is quite possible that the average and median flake sizes for both length and area could be greater for the tip-to-tip method than for the other two methods when the same value for minimum size is employed for all three methods. 

Since ASTM E2567 for the image analysis metallography of ductile iron prescribes a minimum “roundness” shape factor of 0.60 based on Feret diameters, the same may be an appropriate value in applying the ductile iron method for quantification of the flake graphite morphology of gray iron. (Feret diameter is the perpendicular distance between parallel tangents touching opposite sides of the profile, to microscopically measure the size of irregularly shaped particles.)

The gage variations for the three methods of quantifying gray iron flake morphology do not significantly differ, so we will report only on the serpentine method since (1) it can be grandfathered in from the original version of ASTM A247-67; ( 2) its numerical metrics are typically between the numbers for the tip-to-tip and total methods; and (3) the home foundry also reported on the serpentine method for its prior study on the effects of gray iron aging.

Procedure and Data Acquisition

The molten iron from the pressure-pour holding furnace for this testing had previously received tundish ladle magnesium treatments for making ductile iron by which most of the dissolved nitrogen in the base iron was purged. This iron was subsequently exposed to a shroud of gaseous nitrogen at low pressure over the course of almost two days in the holding furnace by which the magnesium loss and sulfur reversion from refractory lining back into melt resulted in a molten iron composition that was a close approximation of that of a typical hypoeutectic gray iron. 
Regarding any concern about the “dead” iron nuclei after the extended holding time, note that Grede–New Castle’s original method for casting coupons with iron from pressure-pour furnaces uses a miniature paper cup containing 1 gram of foundry-grade inoculant having composition of approximately 75% silicon, 1% calcium, 1% aluminum, and the balance iron, with particle sizing of 30 x 70 mesh. We then have instantaneous inoculation in the coupon mold by which carbide formation is prevented and desired graphite morphology is attained, even in iron that would otherwise be prone to carbides. Furthermore, there is nothing in the comprehensive results of photomicrographs, quantified digital image analysis metallography, and integral test bar mechanical properties that follow in this paper that demonstrate any evidence of the deleterious effects that are suspected of being associated with long held iron versus. freshly melted iron. Likewise, the results of the related prior original research are also bereft of any evidence of such effects.

The sampling of the molten from the holding furnace receiver trough was done using a pre-heated (pre-rinsed) “small” ceramic fiber spoon of ~4 lb. capacity, of which five spoonfuls were poured into a pre-heated “large” spoon of >20 lb. capacity in which alloy additions were made after the pre-heat rinse. The total elapsed time from the start of sampling from holding furnace until pouring into metallurgical sample receptacles was less than one minute for each sample set. 

The holding furnace temperature range in the receiver trough of Pressure-Pour No. 1 furnace having a mature refractory lining for the campaign of gray iron UV coupons poured on Sunday, June 25, 2023, was 2705F to 2761F. Those holding temperatures along with their corresponding “TP(°F)” peak temperatures in plain thermal cups upon subsequent pouring in sample receptacles are plotted in Figure 2 per the left vertical axis, and the corresponding spectrometer carbon contents are plotted under the temperature plots per the right vertical axis of Figure 2. The comparative temperature plots indicate a temperature loss of approximately 300F between holding furnace and cast samples, so the gray iron having an approximate carbon equivalent of 4.0% provided at least 200F of superheat in casting all samples in a little more than two hours. The corresponding gradual downward trend in carbon contents over the course of the elapsed time is as anticipated. The average carbon loss (burn-off) among the nine replicated groups is ~0.07% from the first set to second set of which the pouring sequences for the nine groups of each set were identical. Although the carbon burn-off is significant, it is proportionate among the nine replicated groups, and the replication was necessary to do proper statistical analysis that could not have otherwise been done.
Nine combination sample groups of ladle/melt treatments in two replicate sets were used for the pouring of all sample receptacles needed to attain adequate statistical significance. The sequence of pouring the nine groups of combinations for each of the two replicate sets of the ferroalloy additions is illustrated in the matrix of Figure 3, whereby the sequential numbers of the nine casting groups are entered in the matrix’s internal squares.
Here are the ladle spoon ferroalloy adds for the gray iron sampling campaign of Sunday, 06/25/23:
ferromanganese (FeMn); 10 grams for all nine (9) groups.

ferroboron (FeB); none for “None,” 3 grams for ‘Low’ & 6 grams for “High.”. 
ferrotitanium (FeTi); none for “None,” 5 grams for ‘Low’ & 10 grams for “High.”

The large spoon weight of molten iron was estimated to be 20 pounds for each of the nine (9) groups.
All 72 coupons were then allowed to cool to black heat in their molds prior to “pitch & catch” ultrasonic testing within an hour after pouring the last group of coupons. Ultrasonic testing was then periodically done over the span of 56 days to reach endpoints on the velocity rises. That non-destructive testing then concluded at that point to proceed with the destructive sectioning of coupons necessary for the correlative image analysis metallography work. 

Two of the four coupons from each treatment group of varying ladle treatments were then sectioned at a distance of .5 in. from the coupon bottoms where the sonic pulse had passed for the ultrasonic testing, with that followed by further preparation of the samples in metallographic mounts for digital image analysis metallography of their unetched graphite morphologies and etched matrix structures for ferrite/pearlite proportions totaling 100% (graphite dismissed from matrix total) and any anomalous phases such as primary carbides. The digital image analysis metallography was conducted at 100X magnification using a microscope having resolution of 1.17 pixels per micron (0.85 microns per pixel) at the 100X magnification.

The other two of the four coupons from each treatment group were machined for integral test bars. Those machined integral bars were then allowed to set another nine days at room temperature before being pulled for testing of mechanical properties, which was then followed by further metallographic analysis at the host foundry where the samples were originally cast. Chemistry analyses of physical samples of both spectrometer slugs and ultrasonic coupons are presented in Table 1.

Two general methodologies were employed for the statistical analysis of results; they are (1) DOE analysis for the three addition levels of both boron and titanium, plus the two geometric types of “plain” thermal analysis cups, and (2) Pearson statistical analysis for quantification of correlations for dual combinations of continuous variables.

Conclusions

We conclude with the following salient summary points for the results:

• Boron content is positively correlated to graphite flakes density per mm, the smaller Size Classes 6 & 7 flakes, ultrasonic velocity and hardness, and is inversely correlated to ferrite (positively correlated to pearlite).
• Titanium content is positively correlated to mean flake area, graphite percentage, and the larger Size Classes 1 & 2 flakes, and is inversely correlated to ultrasonic velocity, strengths and strength:hardness ratios.
• Boron suppresses ferrite formation by suppressing graphite undercooling per its negatory interaction with titanium whereby the effect of titanium in promoting graphite undercooling is minimized. 
• Ultrasonic velocity of the test coupons is positively correlated to hardness and strengths, and is inversely correlated to mean flake area, the larger Size Classes 1, 2 & 3 flakes, and ferrite content. Ultrasonic velocity is perhaps the most correlative metric to all other metrics considered in this study when evaluated in its entirety, so Figure 4 presents the Minitab gage repeatability and reproducibility (GR&R) results for ultrasonic velocity. For the span of 363 m/s encompassing the range of all ultrasonic velocities, the total GR&R study variation for ultrasonic velocity is 19.29%.
• The plain square cavity TA cups are more highly correlative to ultrasonic velocity than are the plain round cavity cups, so we ascribe that to the larger volume/weight of the square cups relative to ‘round’ cups which are the closer geometric match to the ultrasonic coupons.
• Since the plain TA cup geometry and size of round versus square cavity has profound effects on the resultant cooling curve metrics, one valid goal of plain cup TA testing would be minimization of those differences. That would then enable more accurate predictivity of any shrinkage which can result from the gating and risering of the many different part numbers of varying size and geometry.
• Ferrite content is positively correlated to the UTS/HBW ratio, and is inversely correlated to flake density, the smaller Size Classes 6 and 7 flakes, ultrasonic velocity and hardness.
• UTS is positively correlated to ultrasonic velocity, YS and strength:hardness ratios, and is inversely correlated to flake density and graphite percentage.
• YS is positively correlated to ultrasonic velocity, UTS and YS/HBW ratio, and is inversely correlated to nodule count (per DI image analysis method).
• Strength:hardness ratios are positively correlated to ferrite content and strengths, but are inversely correlated to graphite percentage, nodule count and hardness.
• Boron’s effects in increasing velocity and hardness while reducing ferrite are highly correlated to its effects on increasing graphite flake count/density and quantity of the smaller Size Classes 6 and 7 flakes.
• Titanium’s effects in reducing velocity and strengths are highly correlated to its effects on increasing mean flake size, graphite percentage and the larger Size Classes 1 and 2 flakes.
• One plausible reason for the greater complexity of the effects of boron and titanium on gray iron relative to ductile iron is that the solid-state graphitization, which occurs during the eutectoid transformation ,coarsens (or rounds) graphite flakes of gray iron, but not for graphite nodules of ductile iron which are already quite round before the transformation. 
• One particular benefit of the Pearson statistical analysis relative to the DOE analysis is its greater discrimination by accounting for the boron and titanium residuals as “continuous” rather than “categorical” variables. That greater discrimination results in greater than 90% statistical significance for correlations of boron and titanium contents to many of the other variables/metrics. 
• The spectrometer analyses for titanium of all the individual coupons positively trend well with their ferrite contents, and that is borne-out by linear regression analysis more so than by the DOE analysis.
• Many of the B & Ti elemental effects per the DOE analyses are “complementarily divergent” from each other rather than synergistic, so those divergences add to the complexity of some of the effects. Furthermore, the divergent rather than synergistic effects on some response metrics by the varying contents of boron and titanium might serve to explain some of low correlations per the Pearson statistical analysis by linear regressions. That is, the non-linear plots of some effects by DOE analyses might help explain why some results for the related continuous variables do not reflect results of the DOE analyses for categorical variables. DOEs are better for detecting nonlinearity when the three levels of categorical variables are employed, and the linear regression analysis which employs continuous variables serves to corroborate the DOE analyses when effects for the two extreme levels strongly differ.
• The strong softening/ferritizing effect by titanium may be at least in part the result of its combining with the nitrogen in preferentially forming titanium nitrides rather than boron nitrides. Free nitrogen is known to be a strong pearlite promoter in gray iron, but that effect may be diminished by its combining with the titanium additions. In turn, the boron may be renitent to combining with nitrogen such that the “free” boron that remains may also promote ferritization. Such an interaction between the titanium and boron also supportively explains some of the statistically significant interaction effects quantified by the DOE analysis. 
• The metallographic graphite morphology metrics reflective of the coarsening of flakes by boron might be the result of some thermal analysis metrics reported on by other research groups.
• The fact that every graphite particle is a nucleation site for solid-state eutectoid ferritization and concomitant rounding of the graphite particles might help explain the main and interaction effects of boron and titanium.
• There are no easy explanations for the maximum and minimum values of the effects from either of the middling “low” B or Ti adds. However, one can easily comprehend from graphs of the DOE results in the original unabridged IJMC paper that those effects that are maximized and minimized at extremes for the “none” or “high” alloy adds.