ISIJ International, Vol. 47 (2007), No. 2, pp.

269–276

Eutectic Cell and Nodule Count in Cast Iron

Part II. Experimental Verification

Edward FRAS,1) Marcin GÓRNY1) and Hugo F. LÓPEZ2)

1) AGH University of Science and Technology, Reymonta 23, 30-059 Cracow, Poland.
2) Materials Department, University of Wisconsin-Milwaukee, 3200 N. Cramer Street, Milwaukee, WI 53211, U.S.A.
(Received on July 25, 2006; accepted on November 6, 2006 )

In this work the predictions of a proposed theoretical analysis aimed at elucidating eutectic cell counts, N
in gray iron cast and nodule counts, Nn in ductile iron were experimentally verified. The work was focused
on processing flake graphite and ductile cast iron under various inoculation conditions in order to achieve
various physicochemical states in the experimental melts. In addition, plates of various wall thicknesses
were cast and the resultant eutectic cell N and nodule Nn count were established. Moreover, thermal analy-
sis was employed for determinations of the maximum degree of graphite eutectic undercooling, DTm. A
comparison was made between the theoretical predictions and the experimental determinations of eutectic
cells, N, nodule counts, Nn and DTm. It was found that N and Nn can be related to the casting modulus, M
and as a consequence to the wall thickness, s in plate shaped castings. In particular, the outcome of this
work indicates that the predictions of the theoretical analysis are in good agreement with the experimental
outcome.
KEY WORDS: gray cast iron; nodular cast iron; eutectic cell count; nodule count.

known practical relationships between different metallurgi-

1. Introduction
cal parameters,6–8 casting wall thickness, s, and DTm, N and
There are various quality indicators commonly employed Nn.
in cast iron foundry practice, such as the carbon equivalent,
CE, which is a function of the chemical composition. From
2. Experimental Procedure
the theoretical analysis given in part I of this work, it is
apparent that the quality of the liquid cast iron can also be 2.1. Flake Graphite Cast Iron
determined from the nucleation coefficients Ns and b. The Experimental melts were made in an electric induction
magnitudes of the Ns and b coefficients can be estimated as furnace of intermediate frequency in a 15 kg capacity cru-
fitting parameters in Eq. (1), Part I. In this case, the maxi- cible. The raw materials were pig iron, steel scrap, commer-
mum degree of undercooling at the onset of the graphite cially pure silicon, sulphur and ferro-phosphorus. Melting
eutectic solidification, DTm has to be experimentally meas- was followed by liquid iron superheating to 1 420°C and in-
ured using thermal analysis, and the eutectic cell N, or nod- oculation using FOUNDRYSIL (73–78% Si, 0.75–1.25%
ule Nn count determined from quantitative metallography. Ca, 0.75–1.25% Ba, 0.75–1.25% Al) with a 0.2–0.5 cm
In general, the objective of this work is to show that it is granulation, and added as 0.5% of the total charge weight.
plausible to obtain correlations between Ns and b values, After various time intervals (1.5; 5; 10; 15; 20 and 25 min)
with (i) the chemical composition of the cast iron, (ii) type from the instant of inoculation, the cast iron was poured
and amounts of spheroidization and inoculation agent, (iii)
superheating temperature and bath holding times and (iv) Table 1. Data on material mold ability to absorb heat.1–3)
time after inoculation. Once this information is available
together with the initial metal temperature, Ti just after fill-
ing of mold, Eqs. (2), (10) and (11), (13) derived in part I
can be employed for N or Nn predictions. In addition, cast-
ings with different moduli can be poured in molds with
molding materials that have different abilities to absorb
heat (see Table 1).
A second aim of this paper is to show in some detail the
potential of the physical model proposed in Part I, by con-
sidering the experimental results of this work, as well as
from other authors.4,5 In turn, it is expected that the out-
come of this work will provide further insight into well

269 © 2007 ISIJ

 ISIJ International, Vol. 47 (2007), No. 2
into plate shaped molds of s
6; s
10; s
16; s
22 and
s
30 mm in thickness. In the plates with 6, 10 and 16 mm
in thickness, the length and height were 100 mm, while it
was 140 mm for all the remaining plates. In all of the cases,
the plates had a common gating system. Table 2 gives the
chemical composition of the experimental cast irons.
The foundry molds were prepared using conventional
molding sand. In addition, they were instrumented with
Pt/PtRh10 thermocouples enclosed in quartz sleeves of
1.6 mm in diameter for plates of 6 and 10 mm in thickness,
and of 3 mm in diameter for plates of other thicknesses.
The thermocouple tips were located in the geometrical cen-
ter of each mold cavity normal to the heat transfer flow to
improve the measurement accuracy. An Agilent 34970A
electronic module was employed for numerical temperature
recording. Figure 1(a) shows some typical cooling curves.
These curves were then used for determinations of the ini-
tial metal temperature, Ti just after filling of mold, and of
the minimal temperature, Tm at the onset of eutectic solidifi-
cation. The maximum undercooling at the onset of eutectic
Table 2. Time after inoculation and the chemical composition
of flake graphite cast iron.
Fig. 1. Cooling curves for (a) flake graphite cast iron, s
30 mm
and (b) nodular cast iron, s
22 mm.
Fig. 2. (a) Graphite eutectic cell boundaries for casting I/7 in a s
10 mm thick plate and (b) nodular cast iron structure
for melt 6 in a s
6 mm thick plate.
© 2007 ISIJ
solidification in the individual plates was established from
DTm
TsTm ................................(1)
where Ts is the equilibrium temperature for graphite eutec-
tic,
Ts
1 1545.25Si14.88P .....................(2)
After solidification, specimens for metallographic examina-
tion were taken from the geometrical centers of the plates.
Metallographic examinations were made on polished and
etched (Stead reagent) specimens to reveal the graphite
eutectic cell boundaries. Figure 2(a) shows a typical planar
microstructure (on the specimen cross-section), with dis-
tinct eutectic cells. The planar microstructure is character-
ized by the eutectic cell count, NF (cell density), which
gives the average number of graphite eutectic cells per unit
area. NF can be determined by means of the, so-called vari-
ant II of the Jeffries method, and applying the Saltykov for-
mula as an unbiased estimator for the rectangle, S of obser-
vation9)
N i 0.5N w 1
NF
.........................(3)
F
where Ni is the number of eutectic cells inside S, Nw is the
number of eutectic cells that intersect the sides of S but not
their corners and F is the surface area of S.
Assuming that the spatial grain configurations follow the
so-called Poisson–Voronoi model,10) a stereological formula
can be employed for N determinations (spatial cell density).
This yields the average number of eutectic cells per unit
volume as
N
0.568(NF)3/2 ..............................(4)
Table 3 gives the experimental outcome on NF, s, times
after inoculation and DTm for the eutectic transformation at
the onset of solidification.
2.2. Nodular Cast Iron
Test melts were made in an electric induction furnace of
8 000 kg capacity. The raw materials employed were pig
iron, nodular scrap and commercially pure silicon. After
melting and preheating at 1 485°C, cast iron was poured
into a casting ladle where it was spheroidized using the
270
 ISIJ International, Vol. 47 (2007), No. 2

cored wired injection method. Different inoculants in vari-
ous amounts were used (Table 4). The aim of using differ-
ent inoculants and inoculation processes was to induce dif-
ferent maximum undercoolings. The chemical composition
of the nodular iron is given in Table 5. Nodular cast iron
was poured into similar molds as for gray cast iron. Also,
the experimental methods used in determining the cooling
curves and in sample preparation for metallography were
Table 3. Wall thickness, cell count and maximum undercool-
ing.
similar as for flake graphite cast iron.
Metallographic evaluations of nodule count Nn,F (per unit
area) were made on samples cut from the plates geometrical
centers. Figure 2(b) shows a typical nodule structure found
in melt 6. The area nodule count (average value from 5
measurements), Nn,F was measured using a Leica QW quan-
titative analyser. In ductile iron the graphite nodules are
characterized by Raleigh distributions11) so the volumetric
nodule count, Nn can be related to the planar nodule count
Nn,F using the Wiencek equation12)
3
N n,F
Nn
...............................(5)
f gr

where fgr is the volume fraction of graphite at room temper-

ature ( fgr
0.11–0.14).
Table 6 shows the experimental results for DTm and of
Nn,F in melts 1–7.

3. Results and Discussion

3.1. Maximum Degree of Undercooling D Tm
The theoretical analysis of the solidification process in
cast iron is given in part I of this work. Accordingly, Eqs.
(2) and (11), part I, show that DTm at the onset of the eutec-
tic solidification is related to the eutectic cell count, N in
flake graphite cast iron, or to the graphite nodule count Nn
in nodular cast iron as
• Flake graphite cast iron
1/ 8
 2 5 Ts3a 6 
∆Tm
 6  ...........(6)
 π Le N (1 f γ )cef µ ϕ M
2 3 3 6


Table 5. Chemical composition of experimental melts.

Table 4. Metallurgical treatment including spheroidizers and inoculants.

271 © 2007 ISIJ

 ISIJ International, Vol. 47 (2007), No. 2

• Ductile cast iron Tl is the liquidus temperature at the onset of austenite solid-
ification (Table 1, Part I), a is mold material ability to ab-
1/ 2 1/ 4
1 a3   Ts3  sorb heat, Le is the latent heat of graphite eutectic, fg is
∆Tm
 .......(7)
c  2 zB2 N n Le M   5 3 3 austenite volume fraction, cef is effective specific heat of
π β D 
3
metal (Eq. (A9), Part I), m is growth coefficient of eutectic
for flake graphite cast iron (Table1, Part I), Fcast, Vcast are
where the surface area and volume of the casting respectively, c is
Vcast the specific heat of metal, b is a coefficient (Table 1, part I),
M
.................................(8) D is the diffusivity of carbon in austenite.
Fcast For plates with wall thickness, s lengths, l and heights, h
which easily exceed s, the casting modulus can be calcu-
f
cefBcB1 ................................(9)
lated from Eq. (8) as M
Vcast/Fcast(slh)/(2lhh). Thus, the
plate wall thickness can be related to M by
Ti T T
B
ln , B1
ln l , B2
ln i ..........(10) s
2M....................................(12)
Tl Ts Ts
Estimations were made using thermophysical data (Table 1,
z
0.410.93B2 ............................(11) part I), Eqs. (4)–(7) and (12) and the experimental outcome
of this work on NF and Nn,F (Tables 3 and 6), and the results
Table 6. Experimental values for s, DTm, and Nn,F.
plotted in Fig. 3. In particular, the solid line corresponds to
cast iron of average chemical composition (Tables 2 and 5),
and average Ts values. Moreover, it is worth mentioning
that these results agree with the work of Dopp et al.5) on
measured nodule counts and DTms. In particular, notice
from Fig. 3 that the experimental results are closely located
to the curves predicted using the theoretical analysis.
Hence, Eqs. (6) and (7) show that under a constant wall
thickness, s, as N or Nn, decrease, DTm increases.

3.2. Ns and b Nucleation Coefficients

The experimental outcome given in Tables 3 and 6 is
plotted in Fig. 4. Considering this data and taking into ac-
count the expression for heterogeneous nucleation (Eq. (1),
part I).
 b 
N nuc
N s exp   ....................(13)
 ∆Tm 

as well as using Eqs. (4) and (5) for determining the corre-
sponding Nnuc
N or Nn values, the nucleation coefficients,
Ns and b can be obtained from statistical methods (see
Table 7). Notice that in these calculations the correlation
coefficients are relatively high (Table 7). In addition, from
Table 7 it is apparent that the nucleation coefficients Ns and
b in Eq. (13) for inoculated cast iron (melts I/2 to I/7) are

Fig. 3. Maximum degree of undercooling for plates of various thicknesses (a) as a function of cell count, NF and (b) as a
function of the nodule count, Nn,

© 2007 ISIJ 272

 ISIJ International, Vol. 47 (2007), No. 2

Fig. 4. Eutectic cell count, N as a function of DTm (a) for experimental melts I/1–I/7 and (b) of the nodule count Nn as a
function of DTm for experimental melts 1–7. Points, experimental results; lines, results based on Eq. (13) using
data for Ns and b from Table 7.

Table 7. Nucleation coefficients.

Fig. 5. Effect of dimensionless time, td after inoculation on b and Ns.

Fig. 6. Eutectic cell count as a function of DTm in flake graphite cast iron.

time dependent. Accordingly, from the data given in Table
7, the inoculation effect can be inferred by the changes in
cell count NF as a function of the time ta from the instant in
which the inoculant was introduced into the melt. More-
over, it is found that after 25 min, the observed changes in
NF are negligible and this time tr can be considered as a ref-
erence point. Hence, the changes in cell count can be ex-
pressed as a function of a dimensionless time given by
t ent that prolonged bath holding times after inoculation lead
td
a ..................................(14)
tr
The dependence of Ns and b as a function of td is shown in
Fig. 5. A polynomial approximation indicates that the re-
gression curves for inoculated cast iron are of the form
b
96.9122.6td59.2td2 [°C]...............(15)
Ns
106(6.50.8td5.3td2) [cm3] ............(16)
From Eqs. (15) and (16), as well as from Fig. 6, it is appar-
to a reduction in Ns while increasing b. For comparison pur-
poses, the experimental outcome between N and DTm is

273 © 2007 ISIJ

 ISIJ International, Vol. 47 (2007), No. 2

Fig. 7. Influence of the wall thickness on (a) the eutectic cell and (b) nodule count. Points, experimental results; lines, re-
sults based on Eqs. (4), (5) and (17)–(20).

Fig. 8. Eutectic cell count NF and DTm as a function of td after inoculation.
, experimental results; —, results based on
Eqs. (4), (12) and (15)–(17).

compared with the published results of Mampaey4) (see Fig.

8). Notice once again a good agreement as inferred by the
high correlation coefficient of 0.98.
From Table 7, in all of the different melts investigated
(different chemical composition and metallurgical treat-
ment and in consequence liquid melts with different sub-
strate size distributions) it is not surprising the fact that the
exhibited Ns and b for a given melt are different. Yet, in
each case N or Nn are an exponential function of DTm that is
well described by Eq. (13), thus confirming that Eq. (13) is
in good agreement with the experimental outcome.
Fig. 9. Influence of the wall thickness, s and amount of inocu-
3.3. Eutectic Cell and Nodule Count
lant, mi on the nodule count.
, experimental results; —,
In part I of this work analytical expressions (Eqs. (10) results based on Eqs. (5), (12), (19), (20), (24) and (25).
and (13), part I) were derived that enable the determinations
of volumetric eutectic cell, N or nodule count, Nn. These Equations (17) and (19) can be employed for determina-
equations can be rewritten as tions of nodule count when DTm is not known. However, in
• For flake graphite cast iron this case the data on Ns and b for a given cast iron needs to
be available. Determinations based on Eqs. (4), (5) and
Ns (12), (17)–(20) after taking into account Ns and b (Table 7),
N
..................(17)
exp[8 ProductLog ( y )] average chemical composition (Tables 2 and 5), as well as
data from Table 1, part I (see Fig. 7) reveals good agree-
where ment between the experimental outcome and the predic-
 π 6 Le N s (1 f γ )cef2 µ 3φ 3 M 6  tions of the proposed theoretical analysis.
b
y
  ...........(18) Figure 8 shows some examples on the variation of NF
8  2 5 Ts3a 6  and DTm with td for flake graphite cast iron of s
16 mm
and a Ti
1 240°C. Notice from these figures that prolonged
• For ductile cast iron bath holding times lead to a reduction in N while increasing
DTm. An explanation for this behaviour can be found by
Ns considering Eqs. (17) and (18), which clearly show that N
Nn
................(19)
exp[2 ProductLog ( y n )] increases and DTm decreases as b decreases and Ns in-
creases. In addition, Fig. 6 shows that extended times after
where inoculation leads to a reduction in Ns, and an increase in b.
1/ 2 1/ 4
 zBN s Le M 3   π 5 D 3β 3  However, the effect on lowering Ns is dominant, in agree-
y n
bc     .........(20) ment with the experimental outcome.
 2a 3   Ts
3
 Figure 9 includes the average experimental results given

© 2007 ISIJ 274

 ISIJ International, Vol. 47 (2007), No. 2

Fig. 10. Influence of the amount of inoculant mi on Ns and b values.

Table 8. Nucleation coefficients.

Fig. 11. Nodule count as a function of mi for a wall thickness

s
15 mm.
, experimental results; —, results based on
Eqs. (5), (12), (19), (20), (24) and (25).
in the work of Dopp et al.5) In addition, cooling curves are
also given so it is possible to identify the temperatures
Ti
1 200°C and Ti
1 220°C for wall thicknesses of 10 mm oculant) the following relationships were found
and 15 mm, respectively. From this data, the nucleation pa-
rameters (Ns, b) can be estimated by considered the follow- Ns
9.26 108mi1.02 109mi21.98 107 ; cm3 ...(24)
ing:
1. From Fig. 9, two point coordinates were calculated, b
12.4739.40mi80.10mi2 ; °C ............(25)
(Nn,F,10, s10) and (Nn,F,15, s15) for each amount mi of inoculant,
where Nn,F,10, Nn,F,15 are the nodule counts for wall thickness These expressions are plotted in Fig. 10. Notice that the
s10 and s15 (that is, 10 and 15 mm). correlation coefficients for the Ns
f(mi) and b
f(mi) func-
2. Known Nn,F,10 and Nn,F,15 values were converted into tions were 0.99 and 0.94, respectively. Moreover, using Eqs.
Nn,10 and Nn,15 using Eq. (5). (4), (12), (19), (20), (24) and (25) as well as data from
3. Numerical methods were used to solve the system of Table 1, part I of this work, Nn,F was estimated and graphi-
equations given below: cally plotted in Fig. 9 (continuous lines), finding good
agreement with the experimental outcome.
Ns Figure 11 shows the experimental results of Fig. 9 as a
N n ,10
.............(21)
exp[2 ProductLog ( y n ,10 )] function of the amount of inoculant, mi for a plate with a
wall thickness, s
15 mm. Considering the nucleation pa-
Ns rameters Ns and b given by Eqs. (24) and (25), the chemical
N n ,15
.............(22) composition for the cast irons (Table 8), selected thermo-
exp[2 ProductLog ( y n ,15 )]
physical data (Table 1, part I of this work) and Eqs. (5),
(12), (19) and (20), the effect of mi on the nodule count,
where the yn,10 and yn,15 functions (see Eq. (19)) are given Nn,F was estimated. The results from these calculations are
by shown in Fig. 11 (continuous line) and they indicate once
1/ 2 1/ 4
again that that the experimental results are in general close
 zB N L s 3   π 5β 3 D 3  to the theoretical predictions.
y n ,10
bc  10 s3 e 10   3  Therefore, it is in general possible to develop a base with
 2a   Ts 
......(23) experimental data, which can then be used to find various
1/ 2 1/ 4
 zB N L s 3   π 5β 3 D 3  correlations. Among others, not only to correlate Ns and b
y n ,15
bc  15 s3 e 15   3  values, with the amount of inoculant and the time after in-
 2a   Ts  oculation, but also with the chemical composition of the
bath, the type of spheroidization and inoculation agent, su-
After taking into account the data from Table 1, part I of perheating temperature and bath holding times.
this work, a numerical† solution to the above equations,
yields the Ns and b magnitudes for each amount of inocu-
4. Conclusions
lant (see Table 8). Moreover, by considering the estimated
nucleation coefficients given in Table 8 and using statistical (1) Theoretical calculations were made and then com-
methods, Ns
f(mi) and b
f(mi), (where mi is amount of in- pared with the experimental results on eutectic cell count in

†
In this case, the set of instructions FindRoot in MathematicaTM programme was used with processing times of less than one second.

275 © 2007 ISIJ

 ISIJ International, Vol. 47 (2007), No. 2
flake graphite, or nodule count in nodular cast iron and with
the wall thickness of plate shaped castings. It was found
that the predictions of the theoretical analysis given in part
I of this work are in good agreement with the experimental
outcome.
(2) In typical nodular cast irons, nucleation coefficients
were estimated as Ns
3.1 107–2.9 108 cm3 and b
7.4–
100.0°C, and for typical uninoculated and inoculated flake
graphite cast irons, Ns
1.6 105–1. 106 cm3, b
78.0–
1 159.0°C and Ns
8.4 105–1.6 107 cm3, b
82.0–162.0°C
respectively.
Acknowledgments
This work was supported by Grant PBZ/KBN-100/T08/
2003.
© 2007 ISIJ
REFERENCES
1) R. E. Showman and R. C. Aufderheide: AFS Trans., 111 (2003), 567.
2) H. Fredriksson and H. Svensson: The Physical Metallurgy of Cast
Iron, ed. by H. Fredricsson and M. Hillert, North-Holland, New
York, (1985), 273.
3) T. Midea and J. V. Shah: AFS Trans., 110 (2002), 121.
4) F. Mampaey: 55th Int. Foundry Cong., Moscow, (1988), 1.
5) R. Doop, B. Prinz, K. J. Reifferscheid, E. Schurmann and Th.
Shulze: 60th World Foundry Cong., Hague, (1993), 234.
6) A. Javaid, J. Thompson and K. G. Davis: AFS Trans., 110 (2002),
889.
7) H. D. Merchant: Recent Research on Cast Iron, Gordon and Breach,
Science Publishers, New York, (1968), 1.
8) A. Javaid, J. Thompson, M. Sahoo and K. G. Davis: AFS Trans., 107
(1999), 441.
9) J. Ryś: Stereology of Materials, Fotobit, Cracow, (1995), 343.
10) J. Osher: Quantitative Gefuengenanalysie, DVG Leipzig-Stuttgard,
Germany, (1994), 35.
11) L. Wojnar: Acta Stereol., 5/2 (1986), 319.
12) K. Wiencek and J. Ryś: J. Mater. Eng., 3 (1998), 396.
276