Professional Documents
Culture Documents
Abstract
Measurements are reported of the liquid and liquidsolid mixture (mush) densities of succinonitrileacetone (SCN
ACE) alloys for use in crystal growth and solidification studies. The measured data for the liquid is analyzed and
consolidated into a correlation that expresses the liquid density as a function of composition and temperature. This
correlation also allows for the calculation of the thermal and solutal coefficients of volume expansion for liquid SCN
ACE alloys. Using the measured results for the liquidsolid mixture densities, the solid densities of SCNACE alloys
during solidification in a mushy mode are estimated based on either the Lever or Scheil solidification model, and are
found to be close to solid densities for pure SCN. # 2001 Elsevier Science B.V. All rights reserved.
0022-0248/01/$ - see front matter # 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 2 - 0 2 4 8 ( 0 0 ) 0 0 9 0 5 - 2
D.L. Ceynar, C. Beckermann / Journal of Crystal Growth 222 (2001) 380391 381
The initially unknown volume of the flask, mixed in a sealed flask at a temperature approxi-
corresponding to one of the graduations on the mately 208C above the liquidus temperature.
capillary tube, was determined using the metho- The concentration of acetone in the SCNACE
dology described in an ASTM Standard for the alloy, Co , was determined by measuring the
calibration of volumetric flasks [11]. This standard liquidus temperature, TL . For the present range
chooses distilled water as a calibration material of concentrations, the liquidus line in the SCN
and 208C as the standard calibration temperature. ACE equilibrium phase diagram is accurately
The concept of the calibration is simple: knowing represented by a linear relationship [1,2], such that
the density of water, rw , at the temperature T [12]
Co (TL T0 )/m (3)
and measuring the mass of the water contained in
the flask, Mw , the volume of the flask at T, VT , can where T0 is the melting temperature of pure SCN
be calculated from VT Mw /rw . However, this and m is the liquidus slope. The values
expression cannot be used directly, because of the T0 58.081oC and m 2.8 0.08 K/wt% [1,2]
presence of buoyancy and thermal expansion of were used throughout the present study. The
the glass. Taking those effects into account, the uncertainty in the liquidus slope was estimated
volume at the standard calibration temperature of from the plots in Refs. [1,2].
208C is given by The liquidus temperature was measured accord-
ML ME ing to the following procedure. The flask contain-
V20 [1 a(T 20)] (1) ing the SCNACE alloy was submerged in a large,
rw ra
temperature-controlled bath, completely melted,
where ML is the balance indication of the loaded thoroughly stirred, and then cooled below the
(filled) flask in grams, ME is the balance indication liquidus temperature to nucleate small crystals.
of the empty (air filled) flask in grams, ra is the Next the temperature of the bath was slowly
density of the air in the flask in g/cm3, a is the increased every 30 min, in increments of 18C, to
coefficient of thermal expansion of the flask in melt the crystals. When the temperature ap-
8C1, and T is the temperature of the flask in oC. proached TL and the crystals became very small,
The coefficient of thermal expansion of the flask the increments were adjusted to 0.18C/h. The long
was measured as part of the present study. It was hold times, together with repeated stirring, allowed
found to be a=2.5 105 8C1, which is in the solidliquid mixture to reach equilibrium.
agreement with literature values for the type of Since thermal equilibrium is attained quickly,
glass used. Using distilled water, it was verified diffusion of solute in the liquid is the rate-limiting
that a varies by less than 0.1 105 oC1 over a process for reaching equilibrium. Considering the
temperature range from 158C to 508C. Temper- slow melting rates (less than 1 mm/s), the small size
atures were measured using calibrated thermo- of the crystals (less than 1 mm), and the mass
couples with an accuracy of 0.18C. diffusivity of ACE in SCN (1.3 109 m2/s), the
Using the calibrated flask to measure the characteristic solute diffusion time is less than
volume, V, the density of the SCNACE alloy, 1000 s. The temperature at which the last crystal
r, at any temperature can then be determined from melted was taken as the liquidus temperature, TL .
(ML ME )[1 a(T 20)] Combining the uncertainties in the temperature
r ra (2) measurement ( 0.18C) and the crystal melting
V
observation ( 0.18C) yields an overall uncer-
tainty of 0.148C in TL . Repeated measurements,
2.2. Preparation of the SCNACE alloy and using different stirring procedures and even longer
determination of concentration hold times, verified this uncertainty in the mea-
sured liquidus temperature. Together with the
The succinonitrile and acetone used to prepare estimated uncertainty in the liquidus slope m
the SCNACE alloy were as-received 99% and ( 0.08 K/wt%), the corresponding uncertainty
99.7% pure, respectively. The two substances were in the acetone concentration ranges from 0.15
D.L. Ceynar, C. Beckermann / Journal of Crystal Growth 222 (2001) 380391 383
wt% for Co 4.9 wt% to 0.52 wt% for Co = mixture region of the SCNACE phase diagram.
17.7 wt% (i.e., about 3% of Co ). The solid consisted of dendrites finely dispersed in
It should be noted that for the alloys prepared the liquid (i.e., a mush). The flask was held at each
from the as-received impure SCN and ACE, a temperature for at least 30 min, with periodic
concentration calculated from Eq. (3) actually shaking and mixing of the solution. The measure-
represents an equivalent acetone concentration ments were terminated at a solid fraction of no
which includes the effect of the impurities on the more than 30%, because at higher solid fractions
liquidus temperature. The as-received SCN was mixing of the solidliquid mixture becomes in-
found to have a liquidus temperature of 56.78C, effective and reading of the meniscus position is
which is about 1.48C below the melting point of unreliable.
pure SCN. Hence, the impurities in the SCN are
equivalent to 0.5 wt% ACE. Because most of the
present density measurements are carried out at 2.4. Uncertainty analysis
much higher ACE concentrations, the effect of the
impurities is small. More importantly, as is shown In view of Eq. (2), the uncertainty in the
below, the extrapolation of the present density density measurement, dr, arises from errors in
measurements to vanishing solute concentrations the measured masses, ML and ME , volume, V,
gives liquid densities that are in excellent agree- thermal expansion coefficient, a, and tempera-
ment with literature data for very pure SCN. ture, T. Errors in the air density, ra , are
negligible. Hence, the uncertainty can be estimated
2.3. Experimental procedure from
" 2 2
Before use, the volumetric flask was thoroughly @r @r
cleaned with several successive rinses of distilled dr dML dME
@ML @ME
water and ACE, dried with a heat gun, and 4
weighed to obtain its empty mass. The flask was 2 2 2 #1=2
@r @r @r
then filled with the SCNACE alloy and weighed da dT dV
@a @T @V
again to determine its loaded mass. The loaded
flask was immersed in a large, temperature-
controlled water bath capable of maintaining a where the partial derivatives are obtained from
temperature uniformity of 0.058C to perform Eq. (2). The estimated uncertainties in the
the liquidus temperature and volume measure- measured quantities are dML dME 0.4
ments. The liquidus temperature was measured 103 g, da 0.1 105oC1, dT 0.18C, and
before and after each experiment and found to be dV 7.6 103 cm3. The resulting uncertainty in
the same within the stated uncertainty. each single density measurement is dr 1.4
The variation in volume of the SCNACE alloy 104 g/cm3, which can be considered very low.
with temperature was measured by incrementally Virtually all of the uncertainty arises from the
decreasing the temperature of the bath and noting volume measurement and is related to identifying
the respective volume reading on the flask at each the meniscus position in the capillary tube.
temperature. Here, care was taken to determine Although a smaller inner diameter of the capillary
the position of the meniscus in the capillary tube in tube would reduce the uncertainty of the volume
a manner consistent with the volume calibration. measurement for the liquid, it prevents effective
The measurements were started at a temperature stirring of the solidliquid mixture for the density
approximately 20308C above TL , and ended at a measurements below the liquidus temperature.
temperature below TL when the solid fraction was Thus, the chosen inner diameter of 4 mm repre-
no more than 30% (as estimated from the Lever sents a compromise between accuracy and the
Rule; see below). Hence, densities were measured ability to perform density measurements for a
in both the single-phase liquid and the liquidsolid homogeneous solidliquid mixture.
384 D.L. Ceynar, C. Beckermann / Journal of Crystal Growth 222 (2001) 380391
Fig. 2. Liquid and liquidsolid mixture densities of SCNACE versus temperature for various alloy concentrations.
D.L. Ceynar, C. Beckermann / Journal of Crystal Growth 222 (2001) 380391 385
Table 1
Results of linear regression analysis of measured liquid densities, rl=alT+bl (g/cm3), at various alloy concentrations
where rl is in g/cm3, C is in wt%, and T is in 8C. each of the concentrations measured in the
This composite correlation describes the variation experiments (Co =0.5, 4.9, 9.7, 11.7, 17.7 wt%).
of the liquid density of SCNACE alloys with It can be seen that the correlation agrees very well
temperature and composition, and is valid for with the measured densities. The largest difference,
ACE concentrations ranging from 0 to 18 wt% although still within the confidence interval, exists
and temperatures ranging from the liquidus for Co 0.5 wt%, which can perhaps be attributed
temperature to at least 308C above liquidus. The to the effect of the impurities. Excluding those data
95% confidence interval between the correlation would not have resulted in significant changes in
and all measured liquid density data is the coefficients of the liquid density correlation.
1.67 104 g/cm3, which is close to the measure- Using Eq. (6), the liquid densities of pure SCN
ment uncertainty (1.4 104 g/cm3). Therefore, no (C 0) at various temperatures are calculated and
attempt was made to include higher-order terms in compared with literature values in Table 2. The
the correlation. calculated densities are found to be in excellent
The composite liquid density correlation, agreement with the reported data for pure liquid
Eq. (6), is plotted as dashed lines in Fig. 2 for SCN [1316].
386 D.L. Ceynar, C. Beckermann / Journal of Crystal Growth 222 (2001) 380391
Table 2
Comparison of liquid densities of pure SCN calculated from Eq. (6) with data available in the literature; the densities from the
International Critical Tables are obtained from a given linear correlation; no uncertainties are available for the data of Refs. [13,15,16]
Present correlation, Eq. (6) Int. Critical Tables [14] Other data
T (oC) rl (g/cm3) rl 0.2 102 (g/cm3) rl (g/cm3)
3.3.1. Correlation of measured liquidsolid where Cs and Cl are the average solute concentra-
mixture densities tions in the solid and liquid, respectively, fsV is the
The mixture densities, rm , are shown in Fig. 2 as solid volume fraction, and fsM is the solid mass
a function of temperature for each of the alloy fraction. The two solid fractions are related by
388 D.L. Ceynar, C. Beckermann / Journal of Crystal Growth 222 (2001) 380391
Table 3
Results of polynomial line fit of measured liquidsolid mixture densities, rm am T 2 bm T cm (g/cm3), at various alloy
concentrations
44.3 4.9 6.33 105 2.9650 103 0.98018 0.9997 8.1 105
30.8 9.7 1.79 105 0.5435 103 1.0203 0.9997 8.3 105
25.4 11.7 0.64 105 1.0769 103 1.0175 0.9994 9.8 105
8.6 17.7 1.11 105 1.0137 103 0.99400 0.9994 8.3 105
increases from about 3% to 8% as the solid fractions, the solid density was only calculated
fraction decreases from 15% to 5%. For Co 4.9 from Eq. (15) for fsM > 5%. At fsM 5%, Eq. (16)
wt%, dfsM /fs increases from about 7% to 28% as gives an uncertainty drs of about 4.2 103 g/cm3
the solid fraction decreases from 30% to 5%. Only for all four alloy concentrations. At the highest
at the highest solid fractions is the uncertainty in solid fractions (530%) for each of the alloy
fsM comparable to the difference caused by the concentrations, the uncertainty drs decreases to
Lever Rule and Scheil equation. The relatively about 1.3 103 g/cm3. All of these uncertainties
large uncertainties in fsM at the lowest alloy concen- in the solid density are large compared to the
tration are caused by both the steep increase in the uncertainty in the measured liquid and liquidsolid
solid fraction with decreasing temperature and the mixture densities (dr=1.4 104 g/cm3), indicat-
uncertainty in the phase diagram parameters. ing that most of the uncertainty arises from the
Differentiating Eq. (15), it can be seen that all solid fraction estimates. Nonetheless, the above
three partial derivatives in Eq. (16) go to infinity as uncertainties in the solid density, ranging from
the solid fraction, fsM , approaches zero. This about 0.1% to 0.4%, can still be considered low
means that for vanishing solid fractions, very for most practical purposes.
small uncertainties in r*l , rm , and fsM produce a
very large uncertainty in the solid density. This is 3.3.4. Results for the solid density variation
not surprising, since it is difficult to measure solid The solid densities during solidification calcu-
densities using a mixture that is mostly liquid. lated from Eq. (15) for fsM > 5% are plotted in
Because of the large uncertainties at very low solid Fig. 7 as a function of temperature for each of the
Fig. 7. Solid densities of SCNACE during solidification versus temperature for various alloy concentrations, and comparison with
solid densities of pure SCN of Wulff and Westrum [18] and of 99.7% pure SCN measured in the present study; for easy reference, liquid
and liquidsolid mixture densities are included in the lower portion of the graph.
390 D.L. Ceynar, C. Beckermann / Journal of Crystal Growth 222 (2001) 380391
alloy concentrations Co . It can be seen that the SCN sample. The purity of the SCN was estimated
Lever and Scheil estimates for the solid mass to exceed 99.7%. The solid density of the SCN
fraction give, as mentioned earlier, similar solid sample was measured following the procedure
density variations for each Co . At the highest solid described in an ASTM Standard for determination
fractions, the difference in the solid densities of glass density by buoyancy [19]. The experiments
caused by the two solid fraction estimates is of were repeated several times. The resulting average
the same order of magnitude as the total un- value of the solid density at 21oC, together with
certainty in the solid density, drs . At lower solid the estimated uncertainty, is plotted in Fig. 7. It
fractions, the two solid fraction estimates give can be seen that the present solid density
virtually the same solid density, indicating that the measurement for 99.7% pure SCN falls above
consideration of microscopic solute concentration the trend line of the pure SCN data of Wulff and
gradients in the solid is not important in analyzing Westrum [18], but is still within their uncertainty.
the present measurements. More importantly, the data point falls above the
Also included in Fig. 7 are measured solid solid density estimated for the SCN-11.7 wt%
densities for pure SCN from Wulff and Westrum ACE alloy at 21 oC, thus removing any apparent
[18], together with a trend line through their data contradiction.
assuming a linear variation of the solid density of The small range of ACE concentrations in the
pure SCN with temperature. It is meaningful to solid in the present experiments (52 wt%), and
compare the present calculated solid densities to the relatively large uncertainties in the solid
these data for pure SCN, because the low partition densities at low solid fractions, prevent a correla-
ratio, k, of the SCNACE system results in very tion of the present solid densities as a function of
low ACE concentrations in the solid during Cs and T [similar to the composite liquid density
solidification (at the highest solid fractions, Cs 50 correlation, Eq. (6)]. Nonetheless, the solid den-
.6 wt% for Co 4.9 wt% and Cs 52 wt% for sities presented in Fig. 7 should serve as reasonable
Co 17.7 wt%). Fig. 7 shows that the calculated estimates for use in a variety of solidification
solid densities for the SCNACE alloys are studies.
generally in the same range as the data of Wulff
and Westrum for pure SCN. The addition of ACE 4. Conclusions
can be expected to decrease the density of the solid,
as is observed for the liquid. The calculated solid Measurements were made of the liquid and
densities support this trend. For example at 258C, liquidsolid mixture densities of SCNACE alloys.
the solid density for Co 11.7 wt% is below that The results for the liquid are combined by
for Co 9.7 wt%. However, many of the calcu- regression analysis into a correlation that describes
lated solid densities for the SCNACE alloys fall the density as a function of temperature and com-
above the trend line for pure SCN in Fig. 7. position, up to about 20 wt% ACE. The composite
This apparent contradiction may be explained correlation, Eq. (6), gives the liquid density to
by the relatively large estimated uncertainty within a 95% confidence interval of 1.67 104 g/
in the density data of Wulff and Westrum [18] cm3 or about 0.017%. The thermal and solutal
(Fig. 7). expansion coefficients, bT and bC, are then derived
Because of the relatively large uncertainty in the from this correlation. In the limit of vanishing
Wulff and Westrum [18] data, independent experi- ACE concentration, the present results for the
ments for measuring the solid density of SCN were liquid density and bT are found to be in excellent
performed as part of the present study. These agreement with data available in the literature for
experiments are preliminary and limited to room pure SCN. This indicates that the trace impurities
temperature (218C). More detailed experiments in the starting SCN have little effect on the present
will be reported in a future publication. In brief, measurements. It is found that the density of the
condensate was removed from the condenser of a saturated liquid of SCNACE alloys is remarkably
SCN distillery and used to directionally solidify a constant.
D.L. Ceynar, C. Beckermann / Journal of Crystal Growth 222 (2001) 380391 391