Sie sind auf Seite 1von 9

Alpha Case Thickness Modeling in Investment Castings

W.J. BOETTINGER, M.E. WILLIAMS, S.R. CORIELL, U.R. KATTNER, and B.A. MUELLER
The alpha case thickness at the surface of a Ti-6Al-4V (wt pct) step wedge investment casting has
been measured and successfully predicted. The prediction uses temperature-time results obtained from
a heat flow simulation of the casting. The temperature-time results were coupled to a simple model
for diffusion of oxygen into the beta phase during continuous cooling. Oxygen concentration and
microhardness profiles were measured from the surface in contact with the ZrO2 face coat of the
shell mold into the interior of the casting. The oxygen content in the metal at the shell mold interface
was between 5 and 9.5 wt pct in general agreement with a thermodynamic calculation for bcc Ti in
contact with ZrO2. At the limit of the alpha case region, as determined by standard metallographic
technique, the oxygen concentration was found to be no more than 0.02 wt pct above the level of
oxygen in the bulk alloy. Using this information and one particular literature value for the activation
energy for diffusion of oxygen, a nearly linear relationship was obtained between the measured and
predicted alpha case thicknesses at various positions on the casting surface. Reduction of the prefactor
of this diffusion coefficient by a factor of 7.5 produces excellent agreement between predicted and
measured alpha case thicknesses. Such a reduction is not inconsistent with the scatter of literature
values for the diffusion coefficient.

I. INTRODUCTION

IT is well known that the surface regions of investment


castings of Ti alloys are subject to contamination with oxygen due to the reactivity of the metal with the ceramic shell
mold during solidification and subsequent cooling.[1] The
local increase in oxygen content in the metal near the surface
promotes the formation of an oxygen-rich Ti hexagonal solid
solution (alpha phase) at temperatures where the bulk alloy
would be single-phase beta and also alters the alpha/beta
structure near the surface during cooling to room temperature. The oxygen contaminated layer ranges in thickness
from 50 to 2000 mm, leads to a deterioration of surface
mechanical properties, and must be removed by chemical
milling before use. The prediction of the thickness of this
layer as a function of local cooling conditions at the mold
wall is necessary to determine the milling time, as well as
the excess thickness designed into as-cast shapes.
The standard foundry practice for estimating alpha case
thickness has been to find correlations between section thickness (roughly related to cooling rate) and alpha case thickness. This approach has proved to be unsatisfactory,
especially in complex shapes. With the increased use of
thermal modeling of castings, more accurate methods are
possible. Finite element thermal analysis coupled to unspecified criterion functions has been used to estimate alpha
case thicknesses for investment castings.[2] The microstructure of the alpha case region in pure Ti cast with different
mold materials has been examined by Wictorin et al.[3] The
stability of various rare earth oxide face coats, as well as
ZrO2 and Y2O3, were evaluated by Saha et al.[4]
In the present research, a Ti-6Al-4V (wt pct) test casting
W.J. BOETTINGER, M.E. WILLIAMS, S.R. CORIELL, and U.R.
KATTNER, Metallurgists, are with the Metallurgy Division, National Institute of Standards and Technology, Gaithersburg, MD 20899. B.A.
MUELLER, Process Scientist, is with Howmet Corporation, Whitehall,
MI 49461.
Manuscript submitted January 5, 2000.
METALLURGICAL AND MATERIALS TRANSACTIONS B

was poured under industrial conditions and the alpha case


thickness measured by standard industrial metallographic
methods. Results from an experimental examination of the
interaction zone between the shell mold and the alpha case
region were used to define a simple diffusion model for
oxygen to predict the thickness of alpha case as a function
of local cooling conditions at the casting surface. The local
cooling conditions were obtained from a thermal simulation
of the casting using the numerical finite element code, ProCAST* (VES Software, Annapolis, MD). Comparison of
measured and predicted alpha case thicknesses showed a
*Trade names are included for completeness only and do not constitute
an endorsement by the NIST.

linear relation whose slope can be set to unity by a reasonable


adjustment of the materials parameters of the model.
II. EXPERIMENTAL PROCEDURE
Several Ti-6Al-4V (wt pct) castings were fabricated for
this research using standard industrial practice, viz., vacuum
arc skull melting of alloy from a consumable electrode followed by pouring into preheated shell molds. The castings
were instrumented with type B thermocouples coated with
alumina. Survivability of the thermocouples was poor and
those that survived failed to register temperatures above the
alloy solidus. We believe this was due to the formation of
voids near the thermocouple tip due to reaction between the
alumina and the alloy. Castings were not sand blasted after
knockout to preserve particles of shell mold in contact with
the alloy. Optical metallography was performed on samples
etched with 2 vol pct HF.
Most of the results in this article were obtained from a
step wedge sample shown in Figure 1. Only a quarter section
is shown. The casting had four arms arranged radially, two
being 3.8-cm wide and two being 7.5 cm wide. To obtain
a range of cooling conditions, each arm of the casting had
seven steps. Only the bottom six steps (widths from 0.64 to
3.8 cm) were analyzed because the top step was too close

U.S. GOVERNMENT WORK


NOT PROTECTED BY U.S. COPYRIGHT

VOLUME 31B, DECEMBER 20001419

alloy far from the casting surface. For ZrO2 particles in the
face coat, a composition of 70 wt pct Zr (30 at pct Zr) and
28 wt pct O (68 at pct O) was obtained with the other
elements totaling about 2 wt pct. For the bulk alloy, average
concentrations of 5.7 wt pct Al, 3.8 wt pct V, and 0.7 wt
pct O were obtained. The measurements have an estimated
standard uncertainty of 60.5 wt pct. The measured Al and
V contents are within specifications for this alloy. However,
the measured O content is too high. The actual level of
oxygen in the interior of the casting is 0.25 wt pct, as
measured by a combustion technique. The line scans also
showed the oxygen concentrations in the bulk to be systematically high by approximately 0.5 wt pct. Therefore, the profiles were corrected to give oxygen values close to 0.25 wt
pct in the bulk. The microhardness profiles across the alpha
case region were obtained using a Vickers indenter with a
100 g load.
III. EXPERIMENTAL RESULTS
A. Shell Mold Microstructure
The SEM and optical micrographs of the interface
between the shell mold material and the alloy in a gate
section (not shown in Figure 1) with a very thin alpha case
region are shown in Figure 2. In the shell mold, the bright
particles in close proximity to the alloy are ZrO2. The regions
between the ZrO2 particles appear to be porosity. No Si
signal from the binder was detected in this region.
B. Alpha Case Microstructure and Thickness
Fig. 1Geometry of step wedge test casting. Only a quarter section is
shown. The full widths (W ) of the arms are 7.5 cm (left) and 3.8 cm (right).
The bottom six steps were analyzed. Their widths (S) increase from 0.64
to 3.8 cm.

to the riser. Alpha case measurements were performed near


the center of each step face and on the inner and outer sides
of each arm, i.e., toward the center of the mold and toward
the exterior.
Concentration profiles across the alpha case region were
measured using energy dispersive X-ray analysis with a
scanning electron microscope (SEM) on repolished and
unetched samples. For the oxygen analysis in the alpha case
region, an averaging technique is required because most of
the alpha case region is two phase, consisting of fine plates
of alpha in a beta matrix. Line scans approximately 75
mm long and oriented parallel to the casting surface were
employed for each data point. The position of this line scan
was varied in a direction perpendicular to the casting surface
to give the oxygen profile.
Using elemental standards for Ti, Al, V, Zr, and Si, quantitative software was employed to obtain the oxygen content
by difference (to yield 100 pct total). This procedure requires
precise control over the beam current, which was checked
using a Faraday cup before and after each measurement.
The probe current was 580 pA with a standard error of 1
pA at 20 keV. Standard and sample spectra were acquired
for 300 and 100 seconds, respectively. The degree of reliability of this procedure was checked through measurement of
the concentration of a pure ZrO2 face coat particle and the
1420VOLUME 31B, DECEMBER 2000

Typical of many Ti alloys with narrow liquidus-solidus


separation, there is no metallographic indication of a dendritic cast structure. On the metal side of the interface, a
thin (20 mm) almost featureless region of alpha phase is
typically observed in contact with the shell mold. This region
is usually absent at positions where a pore exists between
the alloy and shell mold. There is no indication of any oxide
phase of Ti such as TiO or Ti2O3 on the interface between
the shell mold and the metal. Farther from the shell mold, the
typical alpha plate structure of Ti-6Al-4V alloy is observed
within large (approximately 250 mm) prior beta grains. With
increasing distance from the interface, the alpha plate size
decreases until it reaches the plate size of the bulk alloy.
Qualitative microprobe analysis (on unetched material) of
the regions between alpha plates near the shell mold interface
indicates the presence of high levels of the beta stabilizer
V. This confirms the presence of residual beta that separates
the alpha plates in this region. Table I summarizes the measured alpha case thicknesses at the various positions along
the casting surface. Industry practice defines the limit of the
alpha case by the positions where the lighter region with
the coarser alpha plate structure fades into the finer plate
structure of the bulk alloy. These measurements have an
estimated standard error of 60.05 mm. The designations
inside and outside refer to the orientation of the step with
respect to the center of the four-arm casting arrangement.
C. Concentration and Hardness Profiles
Figures 3(a), (c), and (e) show measured oxygen and
microhardness profiles of the alpha case for the 0.64, 1.3,
METALLURGICAL AND MATERIALS TRANSACTIONS B

of the alpha case, as determined in the micrographs at the


right (Figures 3(b), (d), and (f)). Clearly, the metallographic
indication of alpha case is quite reliable.
Zirconium was also found in the alpha case region, consistent with the reduction of ZrO2 by Ti. The penetration distance of Zr into the metal was approximately one-tenth of
that of oxygen. The Zr concentration in the metal at the
interface with the shell mold was approximately 4 wt pct.
The Si levels in the metal were found to be quite low (,0.5
wt pct). These facts suggest that the modeling effort should
be focused on the reaction of Ti with ZrO2.
For the model developed subsequently, we are particularly
interested in determining the values of the oxygen concentration at the shell/metal interface and at the position where
metallography defines the limit of the alpha case. Figure 3
shows that the measured interface oxygen concentration is
in the range of 5 to 8 wt pct. Within the scatter of the
measurements, the concentration of oxygen at the limit of
the alpha case is indistinguishable from the concentration
of the bulk alloy. Thus, a more detailed analysis of the
measured oxygen profiles was performed by fitting the measured profiles with two different functions. According to the
model described subsequently, the concentration profile C(x)
after cooling should be given by
C 5 C0 2 (C0 2 C`) erf (x/x0)

(a)

[1]

where C0 and C` are the concentrations at the shell mold/


metal interface and in the bulk alloy, respectively, and x0 is
a characteristic decay distance. Another function is
employed for fitting and given by
C 5 C` 1 (C0 2 C`) exp (2x/x0)

(b)
Fig. 2Micrograph of shell mold/metal interface. (a) Optical, vertical lines
indicate width of alpha case region; and (b) backscattered SEM. The arrow
indicates the thin single-phase alpha region found at the surface. The alpha
plate structure with residual beta phase (light) is seen further from the mold.

Table I. Alpha Case Thicknesses Measured


Metallographically on the Step Wedge Test Casting
3.8-cm-Wide Arm

7.5-cm-Wide Arm

Step Thickness
(cm)

Inside
(cm)

Outside
(cm)

Inside
(cm)

Outside
(cm)

0.64
1.3
1.9
2.5
3.2
3.8

0.038
0.056
0.074
0.084
0.081
0.086

0.023
0.066
0.066
0.081
0.076
0.089

0.028
0.058
0.081
0.091
0.107
0.104

0.03
0.064
0.081
0.091
0.107
0.102

and 3.2 cm thick steps of the inner side of the 7.5 cm


wide arm. It can be seen that the microhardness and the
concentration profiles are quite similar. The vertical lines in
Figures 3(a), (c), and (e) indicate the position of the limit
METALLURGICAL AND MATERIALS TRANSACTIONS B

[2]

where the parameters have the same meaning. Table II shows


the values of these parameters and their standard deviations
obtained by least-squares fitting of the concentration data.
The parameter, x*, is the position of the edge of alpha case
region determine by metallography (Table I) and C* is the
concentration of oxygen obtained from the fitting function
for the position x*. For the subsequent model, the information to be extracted from Table II is the value of C0 and the
value of (C* 2 C`). The former lies between 5 and 9.5 wt
pct. The latter is at most 0.02 wt pct. We also note the similar
values for x*/x0 of approximately 3.5 for the error function
fits to the data. The significance of these parameters will be
discussed subsequently.
IV. MODEL FORMULATION
We will develop a model that only considers diffusion of
oxygen into beta Ti. Clearly, the formation of the alpha case
microstructure involves many other events.
Immediately after pouring, liquid metal will be in contact
with the shell mold for a short time before a thin skin of
solid forms. Due to the low superheat inherent in skull
melting techniques, this time is short. A stagnant fluid
boundary layer of liquid Ti must exist at the mold wall and
oxygen pickup will be limited by diffusion through this
layer. Because liquid diffusion rates are approximately equal
to interstitial diffusion rates at the alloys solidus (typically
1025 cm2/s), we will make a small correction of the beta
diffusion process to include oxygen pickup of the molten
alloy.
VOLUME 31B, DECEMBER 20001421

Fig. 3Plots of measured oxygen content (circles and solid lines) and microhardness (crosses and dashed lines) vs distance and corresponding micrographs
for three different step widths: (a) and (b) 0.6 cm, (c) and (d ) 1.3 cm, and (e) and ( f ) 3.2 cm. The solid lines are fits to the data using an error function.
The vertical lines indicate the extent of alpha case, as determined by optical metallography.

1422VOLUME 31B, DECEMBER 2000

METALLURGICAL AND MATERIALS TRANSACTIONS B

1 2

C
5
D
t
x
x

Table II. Analysis of Measured Oxygen Profiles


Step Width
(Fit Function) C0 (Wt Pct) C` (Wt Pct) x0 (cm) x*/x0
0.64
1.3
3.2
0.64
1.3
3.2

cm
cm
cm
cm
cm
cm

(exp)
(exp)
(exp)
(erf)
(erf)
(erf)

9.5
7.7
5.3
8.5
7.3
5.0

6
6
6
6
6
6

0.4
0.3
0.5
0.4
0.3
0.4

0.22
0.20
0.19
0.29
0.30
0.29

6
6
6
6
6
6

0.08
0.12
0.17
0.09
0.12
0.16

0.0036
0.0090
0.0197
0.0065
0.0156
0.0322

6.4
6.4
5.4
3.5
3.7
3.3

C*
0.24
0.21
0.21
0.29
0.30
0.29

Next, solidification begins at the mold wall with the formation of a layer of the beta phase according to the TiAl-V phase diagram. This layer quickly thickens at a rate
controlled by heat flow. The much slower oxygen penetration
is controlled by solid diffusion in the beta phase. As solidification is completed and the casting cools, an oxygen concentration profile will develop in the beta phase. The calculations shown subsequently reveal that the oxygen profile in
the beta phase is essentially fixed by the time the beta transus
temperature (b/a 1 b phase diagram boundary) is reached.
This is due to the strong temperature dependence of the
diffusion coefficient.
With continued cooling, the metal temperature will drop
below the beta transus temperature. Because oxygen is an
alpha stabilizer, the beta transus temperature increases with
increasing oxygen content. Thus, the alpha phase will first
nucleate near the mold surface and grow inward to form the
thin single-phase alpha layer seen in Figure 2(b). This kind
of growth requires diffusion through an ever thickening alpha
layer and is quite slow.[5]
As the temperature continues to fall, interior points will
drop below their local beta transus temperatures (i.e., become
supersaturated), and a microstructure of alpha platelets in a
beta matrix are observed. This type of alpha transition
requires only a local redistribution of the O, Al, and V and
retains beta between the alpha plates. Near the mold surface,
the plates grow at a higher temperature and produce a coarse
structure. Farther from the interface, where the local beta
transus temperature is lower, the alpha plate size is finer.
Thus, the alpha case microstructure is primarily governed
by the oxygen profile existing in the beta phase prior to
alpha formation.
Rather than treating these events in detail, we propose a
simple model that computes only the oxygen concentration
profile that is established near the mold surface in the beta
phase. The short-range diffusion of oxygen and other elements that is required to form the platelet structure of the
alpha phase from the beta phase is not modeled. Beta grain
boundary diffusion has also not been included. It does not
seem reasonable that grain boundary diffusion should be
significant compared to bulk diffusion for the high-temperature diffusion of an interstitial element.
V. DIFFUSION MODEL
We consider the one-dimensional penetration of oxygen
into the beta phase with a fixed concentration of oxygen at
the metal surface. We define a coordinate system such that
x 5 0 is the outer edge of the metal and x 5 ` is the interior
of the sample. The concentration of oxygen in the beta phase
is denoted by C(x, t) and satisfies the diffusion equation
METALLURGICAL AND MATERIALS TRANSACTIONS B

[3]

The boundary conditions are


C(0, t) 5 C0

[4]

C(`, t) 5 C`

[5]

which correspond to the concentrations of oxygen at the


metal surface and in the bulk, respectively. The initial concentration is taken as that of the bulk:
C(x, 0) 5 C`

[6]

If the boundary concentrations and the diffusion coefficients are constant (e.g., isothermal diffusion), there is a
well-known similarity solution for which isoconcentrates
move with the square root of time. The solution satisfying
the boundary conditions is
C 5 C0 2 (C0 2 C`) erf (x/!4Dt)

[7]

If we consider the case where D is a function of time,


D(T(t)) (through the temperature variation, T(t)) and if the
values of C0 and C` can be assumed to remain constant, the
same solution is valid if we replace the product Dt by a
scaled time, t, given by
t

t5

e D(T(t8))dt8

[8]

Here, we have neglected the spatial variation of the temperature, which should be small over the thickness of the alpha
case. If we assume the usual temperature dependence for
the diffusion coefficient,
D 5 D0 exp (2Q/RT )

[9]

then
t

t 5 D0

e exp (2Q/RT(t8))dt8

[10]

We are interested in the position x*(t) of the isoconcentrate,


C*, at the limit of the alpha case region. Equation [7] gives
x* (t) 5 2lt1/2

[11]

where l is the root of the equation:


erf (l) 5 1 2

C* 2 C`
C0 2 C`

[12]

For a given T(t) curve, the final alpha case thickness


after cooling to room temperature depends on only three
parameters: the temperature (or time) at which the integration is started, the value of Q, and the product, l !D0.
Indeed the final alpha case thickness is directly proportional
to l !D0.
VI. IMPLEMENTATION OF THE MODEL
A. Estimation of Oxygen Composition at the Shell Mold/
Metal Interface
For an accurate model of diffusion of oxygen into the
alloy, a value for the composition of oxygen in the alloy at
the shell mold/metal interface is required. Following the
VOLUME 31B, DECEMBER 20001423

Table III. Diffusion Parameters (from Zwicker[8])


D0 (cm2/s)

Source
[W13]
[R61]
[C33]
[R40b] (Ti-6Al-4V/Air)

Fig. 4Calculated oxygen content at the interface between ZrO2 and Ti


as a function of temperature for liquid, bcc, and hcp phases. The solid
parts of the curves give the approximate temperature where the different
phases of Ti are stable.

experimental observation of significant levels of Zr in the


metal near the shell mold, we will assume that the oxygen
content in the metal at the interface is determined by the
decomposition of ZrO2.
For the simplest model, we will neglect the presence of
Al and V in the Ti-6Al-4V alloy and also the contamination
from Zr. Thus, we will consider an interface between alpha Ti
and ZrO2 and assume that this interface is at thermodynamic
equilibrium for the temperature of the interface. Thus, the
activity of oxygen in the ZrO2 must equal that in the beta
(Ti, O) solid solution. In practice, we equate the O2 partial
pressures to find the oxygen content in the Ti at the interface.
The partial pressure of O2 in equilibrium with ZrO2 at a
temperature T is given by the expression
p02 5 exp

1 RT 2
DGf

[13]

where the value for DGf is obtained from[6] as


DGf 5 21081.5 1 0.177 * T (kJ/mole)

[14]

The phase diagram of the Ti-O system has been analyzed


by Pajunen and Kivilahti[7] by optimizing thermodynamic
and phase diagram information. Using the free energy functions for liquid, bcc, and hcp phases obtained from their
work, expressions for the partial pressure of O2 over the
different phases of Ti-O alloys as functions of composition
and temperature were obtained. These expressions are quite
lengthy and will not be reproduced here. By equating the
pressure exerted by the ZrO2 (Eq. [13]) to the pressure over
the Ti-O alloy, one can numerically obtain the value of the
oxygen composition in the metal required for equilibrium
as a function of temperature.
The results of these numerical calculations are shown in
Figure 4. Oxygen contents are given for liquid, bcc, and hcp
phases over a broad temperature range. Because we have
not treated the full equilibrium, we must determine by other
means the transition temperatures where the alloy changes
from liquid to bcc to hcp during cooling. For this purpose,
we have chosen nominal values of 1900 K for the liquidus/
solidus and 1273 K for the beta transus of Ti-6Al-4V, as
1424VOLUME 31B, DECEMBER 2000

1.6
3.14 3 104
8.3 3 1022
388

Q (KJ/mole)
201.7
287.4
130
249

indicated on the graph. The range where each phase is stable


is shown as solid lines on the oxygen composition curves
of Figure 4. The oxygen composition jumps when the alloy
changes phase. We note that the oxygen content in bcc is
essentially independent of temperature with a value of 11.3
at pct. In a Ti-O alloy, this corresponds to a value of 4.1 wt
pct. At 1273 K, the oxygen content in hcp is 25 at pct, which
corresponds to 10 wt pct. Thus, the thermodynamically estimated range of oxygen content for bcc and hcp is in rough
agreement with the 5 to 9.5 wt pct composition range measured by energy dispersive X-ray analysis. We note in Table
II that the value of C0 exhibits a decreasing trend with
increasing step width. This fact may indicate a shortcoming
of our approach, which will assume a fixed value of C0.
Calculations were also conducted for bcc Ti in contact with
SiO2 and Al2O3. Surface compositions of 25 and 14 at pct
oxygen were obtained for contact with bcc Ti. Higher oxygen
levels are obtained because these oxides are less stable
than ZrO2.
B. Other Materials Data for the Model
For a given temperature-time curve and activation energy
for diffusion, the predicted alpha case thickness is proportional to the quantity l !D0. We conducted our initial prediction of alpha case thickness using l 5 1.57. This value was
obtained from Eq. [12] by taking reasonable values for the
concentrations, C0, C*, and C`, viz.,7.7, 0.4, and 0.2 wt
pct, respectively. Fortunately, the values for l are fairly
insensitive to the exact values of the three concentrations.
For a different set of concentrations, i.e., 8, 0.22, and 0.20
wt pct, respectively, in better agreement with the smaller
value of (C* 2 C`) in Table II, l 5 2.13. A third estimate,
l 5 3.5, can be obtained from the value of x*/x0 in Table
II for the error function fit. After rearrangment of Eqs. [1]
and [12], it can be seen that l 5 x*/x0.
Table III (and Figure 5) summarize literature information
for the diffusion of oxygen in beta Ti found in Reference
8. Three sets of data, labeled [W13], [C33], and [R61], are
for pure oxygen diffusing into pure beta Ti. (The notation
refers to references given in Reference 8.) The data set
labeled [R40b] is for the combined penetration of nitrogen
and oxygen from air into a beta Ti-6Al-4V alloy. Extrapolation of the data in Figure 5 shows that the value of the
diffusion coefficient at the solidus has an uncertainty of at
least a factor of 10. Subsequently, we make predictions using
the [W13], [R40b], and [C33] data.
C. Simulated Cooling Curves
For model validation, it would have been preferable
to measure directly the temperature at the casting/mold
interface; however, as described previously, such measurements were unreliable. Therefore, simulations were
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 5Diffusion data from various references summarized by Ref. 8 (solid


lines) and best agreement between alpha-case thickness measurements and
predictions (dashed line). The data are for pure oxygen in pure Ti except
for [R40b], which is for the combined penetration of oyygen and nitrogen
from air into Ti-6Al-4V.

performed to predict the temperatures at the metal/mold


interface over the entire component surface using a finite
element code, ProCAST. The accuracy of the simulation
(choice of materials parameters and heat transfer coefficients) was verified by two means. The first is through
a favorable comparison between measured and simulated
cooling curves at lower temperatures. The second is the
prediction of fill for several production titanium geometries. The low mold superheats and metal preheats inherent in titanium casting make the complete fill of thin
cast sections difficult. In fact, a principal application
of simulations for titanium castings is the prediction of
casting nonfill in response to gating and process
changes.[9] Accurate simulation of nonfill forces not only
the thermal conditions to be correct, but also the fluid
flow representations. Therefore, nonfill predictions are
considered to be a very stringent validation condition for
titanium castings.
From the finite element numerical heat flow analysis
of the casting, temperature vs time curves were extracted
for the locations where the alpha case was measured.
The curves were for positions in the metal at the casting
surface. The curves were used as input to another code
for the alpha case thickness prediction. The temperature
vs time curves in the metal at the center of each step of
the 7.5 cm wide section are shown in Figure 6(a). The
predictions of the model were obtained from Eq. [8] using
values of t from the integration of Eq. [10] for each
temperature vs time curve. The integration was started at
the time when each position reached the solidus temperature and finished at the end of the data file, which corresponded to approximately 400 seconds after pouring. (The
liquidus and solidus temperatures assumed in the thermal
modeling were 1898 and 1868 K, respectively.) The
METALLURGICAL AND MATERIALS TRANSACTIONS B

(a)

(b)
Fig. 6(a) Calculated temperature vs time curves in the metal at the
surface of the different steps from the 7.5-cm-wide section. The curves
were obtained from a ProCAST simulation. (b) Predicted alpha-case thicknesses vs time at the same positions using D0 5 388 cm2/s, Q 5 249 KJ/
mole, and l 5 1.567 and beginning the integration at the solidus temperature
of 1868 K. In (a) and (b), the curves for individual casting locations (nodes)
are in the same order from top to bottom.

growth of the alpha case thicknesses for the various positions from the 7.5 cm section using the diffusion value
for [R40b] and for l 5 1.57 are given in Figure 6(b). It
can be seen that the alpha case thickness changes little
after the temperature reaches about 1422 K. Thus, the
alpha case thickness is essentially determined at a temperature well above the beta transus temperature (,996 K)
of the bulk alloy composition.
VOLUME 31B, DECEMBER 20001425

(a)

Fig. 8Plot of measured vs predicted alpha-case thickness using l 5 1.567


and the [R40b] diffusion data but starting the integration at the liquidus
temperature of 1898 K. A linear fit is shown.

(b)
Fig. 7(a) Plots of measured vs predicted alpha-case thickness using l
5 1.567 and the [W13] and [R40b] diffusion data. Linear fits are shown.
(b) Plot of measured vs predicted alpha-case thickness using l 5 1.567
and the [C33] diffusion data. A linear fit is shown.

VII. DISCUSSION
A. Comparison of Measured and Predicted Alpha
Case Thicknesses
A comparison of measured and predicted alpha case thicknesses using two different values for the diffusion coefficient
is shown in Figure 7(a). Shown are linear fits to the relationship between measured vs predicted thicknesses, which suggest significant credibility of the chosen alpha case model.
While the fits have intercepts near zero, use of the [W13]
diffusion data gives predictions that are too small by about
a factor of 1.6, and the predictions using the [R40b] diffusion
data are too large by about a factor of 2. We note no systematic difference for the quality of the fit for the data from the
3.8-cm-thick section vs the 7.5 cm thick section.
The other two sets of diffusion data, [C33] and [R61],
give poorer results. Figure 7(b) shows the results using the
diffusion data of [C33]. The intercept is far from zero and
the use of this diffusion data with this simple model appears
invalid. The [R61] data gives predicted alpha case thicknesses 5 times larger than the measured values. It is interesting to note that the effect of decreasing the value of Q is to
shift the y intercept to more negative numbers. These two
diffusion data sets, [C33] and [R61], will not be considered further.
Fitting the data of Figure 7(a) with the function, measured
5 constant * (predicted)n, gave exponents of 1.05 and 1.14
for the [R40b] and [W13] diffusion data, respectively. Thus,
it appears that the use of the [R40b] activation energy gives
1426VOLUME 31B, DECEMBER 2000

Fig. 9Plot of measured vs predicted alpha-case thickness using adjusted


parameters to obtain a linear fit, which has a slope of unity and intercept
of zero. The parameters used are l 5 2.13, D0 5 51.2 cm2/s, and Q 5
249 KJ/mole, and the integration was begun at 1875 K, a temperature
slightly above the solidus temperature of 1868 K.

a slightly better approximation to linear behavior. We will


explore the use of this diffusion data set subsequently.

B. Effect of Starting Temperature for the Integration


For the preceding predictions, the starting point for the
integration was taken as the solidus temperature. If this
temperature is increased, the predicted alpha case thicknesses will increase slightly. Figure 8 shows the effect of
beginning the integration at the liquidus temperature for the
[R40b] diffusion data. Again, a linear fit to the data is shown.
Choosing a temperature between the liquidus and solidus
appears to be necessary to approximate the pickup of oxygen,
while liquid was in contact with the mold face coat, as
described earlier.
METALLURGICAL AND MATERIALS TRANSACTIONS B

C. Adjustment of Parameters to Obtain Agreement


between Measured vs Predicted Alpha Case
Thicknesses
Figure 9 shows the result of beginning the integration at
a temperature that achieves a zero intercept and altering the
product l!D0 so that the linear fit to the measured vs predicted graph has a slope of unity. Although such fine tuning
of the model may not be justified given the approximations
in the model, it is instructive. The temperature that achieves
zero intercept is 1875 K, a temperature only 7 K above
the solidus for this alloy. Beginning the integration at the
estimated liquidus of 1898 K leads to an unacceptably large
intercept. The value of the product l!D0 that gives a slope
of unity for the [R40b] activation energy (249 KJ/mole) is
15.2 cm s21/2 rather than the previously selected value of
30.9 cm s21/2 (1.57 !388). To get such a value for
l!D0, one could decrease the value of l by a factor of 2.03
from 1.57 to 0.77 by changing the estimates for the various
concentrations. However, unreasonably large changes in the
concentrations are necessary to obtain a change in l of this
size. Also as noted previously, the concentration profiles
suggest that l should be larger than 1.57.
A second possibility is to adjust the prefactor for diffusion.
Using the better estimate for value for l of 2.13, a decrease
of the prefactor, D0, from 388 to 51 cm2/s for Q 5 249 KJ/
mole ([R40b] value) is necessary to obtain a slope of unity.
As shown by the dashed line in Figure 5, such a change in
D0 is not unreasonable given the large variation in diffusion
data. The dashed line is reasonably close to the [W13] data,
whose activation energy gave only slightly poorer linearity,
as described previously. One cannot be certain whether the
adjustment of these parameters reflects errors in the literature
values or represents a shortcoming of this simple model of
alpha case formation.
Finally, it should be remembered that the diffusion model
is one-dimensional and was only tested for large planar
surfaces. However, in applications to complex casting shapes
with regions near interior and exterior corners, the use of
calculated thermal histories from three-dimensional thermal
simulations will naturally modify the predictions of alpha
case thickness compared to planar surfaces. The differences

METALLURGICAL AND MATERIALS TRANSACTIONS B

in cooling history would likely cause a larger effect than


the need to model the diffusion in three dimensions.
VIII. CONCLUSIONS
1. Oxygen and microhardness profiles were measured
through the alpha case region in a Ti-6Al-4V (wt pct)
step wedge investment casting from the surface in contact
with the ZrO2 face coat of the shell mold into the interior.
The oxygen content in the metal at the shell mold interface
was between 5 and 9.5 wt pct, in general agreement with
a thermodynamic calculation for bcc Ti in contact with
ZrO2. At the limit of the alpha case region, as determined
by standard metallographic technique, the oxygen content
was found no more than 0.02 wt pct above the level of
oxygen in the bulk alloy.
2. The alpha case thickness has been successfully predicted
by coupling the cooling histories obtained from a heat
flow simulation with a simple model for diffusion of
oxygen into the beta phase.
3. This successful prediction employed one particular literature value for the activation energy for diffusion of oxygen in pure Ti, but with a diffusion coefficient prefactor
reduced by a factor of 7.5. Such a reduction is not inconsistent with the scatter in literature values for the diffusion coefficient.
REFERENCES
1. R. Gunstra, J. Thorne, and R. Ziesloft: Vacuum Metallurgy, Science
Press, Princeton, NJ, 1977, p. 461.
2. J. Schaedlich-Stubenrauch, P.R. Sahm, and H. Linn: 6th World Conf.
on Titanium, P. Lacombe, R. Tricot, and G. Beranger, eds., Les Editions
de Physique, Cedex, France, 1989, pp. 649-54.
3. L. Wictorin, N. El-Mahallawy, M.A. Taha, and H. Fredriksson: Cast
Met., 1992, vol. 4, pp. 182-87.
4. R.L. Saha, T.K. Nandy, R.D.K. Misra, and K.T. Jacob: Metall. Trans.
B, 1990, vol. 21B, pp. 559-66.
5. S.R. Coriell: NIST, Gaithersburg, MD, unpublished research, 1994.
6. J. Phys. Chem. Ref. Data, 1985, vol. 14, Suppl., p. 1690.
7. M. Pajunen and J. Kivilahti: Z. Metallkd., 1992, vol. 83, p. 17.
8. U. Zwicker: Titan und Titanlegierungen, Springer-Verlag, Berlin, 1974,
p. 105.
9. R.K. Foran, T. Hansen, and B. Mueller: in Modeling of Casting,
Welding and Advanced Solidification Processes, VII, M. Cross and J.
Campbell, eds., TMS, Warrendale, PA, 1995, pp. 771-77.

VOLUME 31B, DECEMBER 20001427

Das könnte Ihnen auch gefallen