You are on page 1of 8

Thermodynamics, kinetics, and crystallization of

Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
bulk metallic glass
Benjamin A. Legg
a
, Jan Schroers
b,
*
, Ralf Busch
c
a
Department of Mechanical Engineering, Oregon State University, Corvallis, OR 97331, USA
b
Department of Mechanical Engineering, Yale University, 15 Prospect St., New Haven, CT 06511, USA
c
Saarland University, Department of Materials Science and Engineering FR 8.4, Saarbru cken 66041, Germany
Received 19 June 2006; received in revised form 21 September 2006; accepted 22 September 2006
Available online 13 December 2006
Abstract
We report on dierential scanning calorimetry (DSC) studies to characterize the thermodynamics, kinetics and crystallization pro-
cesses of the bulk metallic glass-forming alloy Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
. The heat capacity of the alloy is measured for the crystalline, glassy
and supercooled liquid phases. The heating rate dependence of the glass transition is used to calculate the kinetic fragility. Crystallization
kinetics are determined under isothermal conditions and used to construct a timetemperature-transformation (TTT) diagram. The
experimentally determined crystallization kinetics are t to calculate the activation energy for crystallization. Our results suggest that
Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
is neither a thermodynamically nor a kinetically stabilized glass former. Other contributions, including the activa-
tion energy for crystallization and the use of a ux are considered in the discussion to explain the good glass formability of this alloy.
2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Bulk metallic glass; Platinum; Crystallization; Thermodynamics; Kinetics
1. Introduction
In recent years, a number of new bulk metallic glass
(BMG) forming alloys have been developed. These alloys
show an extraordinary ability to resist crystallization and
may solidify as a glass when cooled at sucient rates.
One of the best metallic glass formers yet discovered is
the Pd
43
Cu
27
Ni
10
P
20
alloy, which under certain processing
conditions has demonstrated a remarkably low critical
cooling rate for glass formation of 0.005 K/s [1], resulting
in a critical casting thickness of >72 mm [2]. By compari-
son, a pure metal must be quenched at rates exceeding
10
9
K/s. Based on the similarities between platinum and
palladium, the Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
alloy has been devel-
oped, with a critical cooling rate on the order of 20 K/s
[3]. Its low liquidus temperature and its large supercooled
liquid region convey tremendous processing advantages
over conventional platinum alloys [4]. In subsequent inves-
tigations it was revealed that this alloy behaves rather
unusually for a BMG. It was found to display outstanding
superplastic formability [4], and even more surprisingly, it is
the rst monolithic BMG that exhibits a plasticity of 20% in
an unconned geometry and a fracture toughness of
K
Ic
= 80 MPa m
1/2
[5]. The authors explain this ductility
in terms of an unusually small ratio of shear modulus to
bulk modulus, which results in a Poisson ratio of 0.42.
Novikovs and Sokolovs nding of a correlation between
an alloys room temperature Poissons ratio and its liquid
fragility suggest a rather fragile liquid behavior of this alloy
[6]. Studies on Zr-based BMG have associated fragile liquid
behavior with poor GFA suggesting that low fragility is not
in line with good glass forming ability of an alloy [7,8].
In this work we are studying the correlation of various
factors with glass formability, with particular emphasis
on fragility. Therefore, the fragility, specic heat and crys-
tallization kinetics of the Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
alloy are
determined using dierential scanning calorimetry.
1359-6454/$30.00 2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2006.09.024
*
Corresponding author. Tel.: +1 203 432 4346; fax: +1 203 432 6775.
E-mail address: jan.schroers@yale.edu (J. Schroers).
www.actamat-journals.com
Acta Materialia 55 (2007) 11091116
2. Experimental methods
Experiments were carried out using a power-compen-
sated dierential scanning calorimeter (DSC) Perkin
Elmer DSC7. The alloy was prepared in a sealed quartz
tube from constituents with purity of 99.95% or better.
Details of alloy preparation are given in Ref. [3]. All mea-
surements were performed using argon purge gas. Alumi-
num sample pans were used when possible, but graphite
sample pans were used for procedures that exceeded the
liquidus temperature of 853 K.
The absolute value of the specic heat capacity was
obtained for the glassy, crystalline and liquid states. The
material was brought to a certain temperature at a rate
of 0.33 K/s and held isothermally for 300 s. This results
in a step of the heat ux, dQ/dt:
dQ
dt

oQ
ot
_ _
_
T60

oQ
ot
_ _
_
T0
c
dT
dt
: 1
The term oQ=ot
_
T60
corresponds to the power necessary
to heat or cool the sample and its sample pan at a constant
rate of 0.33 K/s, oQ=ot
_
T0
is the power needed to hold
the sample and its pan isothermally and c refers to the heat
capacity of the sample and the pan. In order to identify the
heat ow into the sample, identical measurements were
also performed on the empty sample pan and a sapphire
reference. The specic heat capacity of the sample is then
calculated as
c
p
T
metal

_
Q
metal

_
Q
pan
_
Q
sapphire

_
Q
pan

m
sapphire
l
metal
m
metal
l
sapphire
c
p
T
sapphire
; 2
where m
i
is the mass, l
i
is the molar mass, and c
p
(T)
sapphire
is the specic heat capacity of the sapphire, obtained from
ASTM charts [9]. The heat capacity measurements were re-
peated at 15 K intervals between 338 and 743 K for an ini-
tially glassy and initially crystalline sample, and also at
15 K intervals between 973 and 628 K for an initially liquid
sample.
The onset and completion of the glass transition tem-
perature were measured with DSC temperature scans,
by heating initially glassy samples at rates between 0.033
and 2 K/s.
Isothermal crystallization curves were obtained at tem-
peratures between 538 and 568 K by heating initially
glassy samples to the isothermal temperature at a rate
of 1 K/s and holding there until crystallization was
observed. Isothermal crystallization curves at tempera-
tures between 688 and 743 K were obtained by bringing
the samples to an annealing temperature >983 K for at
least 2 min and then cooling to the isothermal tempera-
ture at a rate of 2 K/s. The higher temperature samples
were encased in boron oxide ux to reduce oxidation
that might lead to heterogeneous nucleation.
3. Results and discussion
3.1. Thermodynamics
Fig. 1 shows a DSC heating scan obtained by heating
amorphous Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
at a rate of 0.5 K/s. The
glass transition appears as a change in heat capacity at
T
g
. This is followed by a series of exothermic peaks, which
begin at T
x
, the onset of crystallization. On further heating,
the material begins to melt at the solidus temperature, T
s
,
and completes melting at the liquidus temperature, T
l
.
The crystalline and glassy states have equivalent Gibbs free
energies at the fusion temperature T
f
, which is estimated
using the melting peak temperature. By integrating the area
of the melting event, the total heat of fusion, DH
f
, is calcu-
lated at 11.4 kJ/g atom. The entropy of fusion can then be
estimated as DS
f
= DH
f
/T
f
. With a melting peak tempera-
ture of 776 K, DS
f
is calculated at 14.7 J/g atom K. This
value for DS
f
is large for a BMG and will be seen to have
important consequences on the driving force for crystalli-
zation and the solidliquid interfacial energy.
The specic heat capacity measurements are displayed in
Fig. 2. The heat capacity of the amorphous phase is similar
to that of the crystalline phase at low temperatures, but
when heated above the glass transition temperature, the
amorphous phase relaxes into the supercooled liquid state
and its c
p
assumes the higher heat capacity of a liquid.
The large thermal stability of the alloy allows the heat
capacity of the supercooled liquid to be measured at a tem-
perature 300 K below the liquidus.
Well above the Debye temperature, the heat capacity of
the liquid may be described as
c
p liquid
3R cT dT
2
3
Fig. 1. DSC heat ow curve obtained with a heating rate of 0.5 K/s.
Indicated from left to right are the onset and completion of the glass
transition, the onset of crystallization, the solidus, the fusion peak and the
liquidus.
1110 B.A. Legg et al. / Acta Materialia 55 (2007) 11091116
and the heat capacity of the crystalline mixture may be de-
scribed as
c
p crystal
3R aT bT
2
; 4
where R = 8.314 J/g atom K and a, b, c and d are tting
parameters [10]. The tting parameters are found to be
a = 0.0044622 J/g atom K
2
, b = 0.000011977 J/g atom
K
3
, c = 0.0056268 J/g atom K
2
and d = 5764840 J K/
g atom.
From the experimentally determined heat capacity and
the enthalpy of fusion, the enthalpy dierence between
the liquid and crystalline states can be calculated as func-
tion of temperature, as
DHT
lx
DH
f

_
T
f
T
Dc
lx
p
T
0
dT
0
; 5
where Dc
lx
p
c
p-liquid
c
p-crystal
. The enthalpy dierence cal-
culated by Eq. (5) is shown in Fig. 3. This gure also de-
picts the directly measured heat released during
crystallization, which agrees well with the calculated
values.
There are small deviations of the crystallization enthalpy
from calculated values, which suggests that the crystalliza-
tion products might depend on temperature. At suciently
low temperatures, isothermal crystallization does not reach
completion, so these data are not included. Temperature
scans from low-temperatures display multiple crystalliza-
tion peaks and this is represented schematically by the
jogged crystallization path in Fig. 3.
Similarly to the enthalpy, the entropy dierence between
the liquid and the crystal can be calculated as
DST
lx
DS
f

_
T
f
T
Dc
lx
p
T
0

T
0
dT
0
: 6
This result is plotted in Fig. 4. The Kauzmann tempera-
ture, T
k
, is the theoretical point at which DS
lx
= 0, and
can be found from this plot to be approximately 396 K.
The Gibbs free energy dierence can be calculated as
DG
lx
DH
lx
TDS
lx
: 7
In Fig. 5 [1115], the calculated Gibbs free energy curve
for this alloy is shown and compared with other glass-
forming alloys. At any given undercooling, the value of
DG
lx
represents a lower limit on the driving force for
crystallization. The slope of DG
lx
for the Pt alloy ex-
ceeds that of any other alloy in Fig. 5. This is surprising,
as it is not in line with previous ndings where a corre-
lation between DG
lx
and critical cooling rate was ob-
served [15].
Fig. 3. Enthalpy dierence between the liquid and crystalline states, as
calculated from Eq. (5). The squares represent crystallization enthalpies
measured by DSC. When heating from low temperatures, crystallization
takes place via multiple exothermic peaks, shown schematically by the
jogged crystallization path.
Fig. 4. Calculated entropy dierence between liquid and crystal, as a
function of temperature.
Fig. 2. DSC heat capacity measurements were obtained from initially
liquid (j), crystalline (d) and glassy (n) samples, and used to obtain best
ts for liquid and crystalline heat capacity.
B.A. Legg et al. / Acta Materialia 55 (2007) 11091116 1111
3.2. Kinetics
Upon heating, the glass transition represents the relaxa-
tion of the glassy solid into a supercooled liquid state. By
observing this transition in DSC at dierent heating rates,
we can estimate how relaxation time varies with tempera-
ture. Selected observations, taken at a variety of heating
rates, are shown in Fig. 6.
When higher heating rates are used, the glass transition
clearly shifts to higher temperatures. However, the width of
the glass transition, DT
g
= T
g complete
T
g onset
, remains rel-
atively constant at 20 K. At each heating rate, R, the
relaxation time for the glass transition can be calculated as
s
g

DT
g
R
: 8
When the relaxation time is compared with onset tempera-
ture, the relationship is expected to follow a VogelFul-
cherTamann (VFT) equation, such that
s
g
s
0
exp
D

T
0
T
g
T
0
_ _
9
Ref. [16].
The value s
0
is the theoretical innite temperature relax-
ation time, with a value of 2 10
13
s, while D
*
and T
0
are
tting parameters. D
*
is commonly referred to as the fragil-
ity parameter and T
0
is the VFT temperature. At T
0
, the
relaxation time is predicted to approach innity.
Because DT
g
remains constant, Eq. (8) shows that the
relaxation time is inversely proportional to the heating
rate. In Fig. 7, the inverse heating rate is plotted vs. the
onset of the glass transition and the VFT tting parameters
are obtained by a nonlinear curve t. The best t values of
D
*
and T
0
are found to be 16.4 and 336 K, respectively.
These VFT parameters are then used to obtain an esti-
mate of the alloys viscosity. As a relaxation phenomenon,
the viscosity of a glass-forming alloy is also expected to fol-
low a VFT relationship with regards to temperature
g g
0
exp
D

T
0
T T
0
_ _
: 10
Eq. (10) is essentially identical to Eq. (9), but s
0
has been
replaced by g
0
, the innite temperature viscosity. The in-
nite temperature viscosity is a function of molar volume as
g
0

N
A
h
V
m
11
and has a value of approximately 4 10
5
Pa s [17]. N
A
and h are the Avogadro and Planck constants, respectively.
In other BMG systems, the VFT parameters obtained by
viscosity measurements have been seen to correlate well
with those obtained by observation of the glass transition
[12,18]. Having already obtained the VFT parameters,
Fig. 5. Gibbs free energy dierence between liquid and crystal phases on
undercooling, for selected metallic alloys. Critical cooling rates are noted
in parenthesis. See Refs. [4,15].
Fig. 6. Glass transition, observed in DSC at heating rates between 0.05
and 2 K/s. Fig. 7. Modied Angell plot, showing the VFT t to heating rate data.
1112 B.A. Legg et al. / Acta Materialia 55 (2007) 11091116
Eq. (10) can be used to estimate the temperature depen-
dence of the viscosity. When comparing the VFT viscosity
prediction with superplastic forming experiments per-
formed by Schroers at 543 K [4], the results correspond
within reason.
Within certain families of BMG, a positive correlation
between glass-forming ability and the fragility parameter
D
*
has been reported [7,8]. However, this criterion is not
always valid when comparing dierent families of BMG,
particularly metalmetal alloys with metalmetalloid
alloys. For instance, PdNiCuP with the highest glass
forming ability [1,2] shows relatively fragile liquid behavior
[19]. While the fragility of PtCuNiP is not atypical for a
BMG and is comparable with other Pt-based alloys [20],
it is rather low considering the alloys high GFA.
3.3. Crystallization
Isothermal crystallization studies were carried out at
temperatures between T
g
and T
l
. By integrating the heat
released during crystallization, the total crystallization
enthalpy and the timescales for 5, 50 and 95% crystalliza-
tion were recorded. These results where combined with
observations of the glass transition to generate a TTT dia-
gram for crystallization and relaxation, shown in Fig. 8.
The TTT diagram indicates a nose shape, with the min-
imum crystallization time occurring somewhere between
568 and 688 K. At intermediate temperatures in the vicinity
of the nose, crystallization occurred too rapidly to observe
isothermally. Data above the nose were obtained by
quenching from the melt while data below the nose were
obtained by heating a glassy sample. Crystallization pro-
ceeds by distinctly dierent processes in these two regimes.
Below the nose, crystallization time is seen to decrease
with rising temperature. This can be understood as a result
of increasing atomic mobility with increasing temperature.
Crystallization that occurs in this regime is incomplete on
laboratory timescales. The enthalpy released during iso-
thermal crystallization is just over 50% of the predicted
DH
lx
and more enthalpy release will occur if the sample
is later heated to higher temperatures. The alloy displays
a remarkable resistance to crystallization in this regime,
with extrapolations of the TTT diagram suggesting about
four orders of magnitude dierence between crystallization
time and relaxation time.
TEM micrographs taken after isothermal crystallization
at 565 and 583 K have displayed needle-like crystals, with
length on the order of 100 nm and width of 10 nm [22].
This is indicative of an interfacially limited growth mecha-
nism with high anisotropy that typically occurs at shallow
undercoolings where low driving forces and high kinetics
prevail.
At temperatures above the nose, the crystallization time
tends to decrease when temperature is reduced. This is a
consequence of an increase in driving force for crystalliza-
tion with larger supercooling. Based on the similarity
between DH
x
and DH
lx
, crystallization in the high-
temperature regime is believed to approach 100% comple-
tion. However, the exact nature of the crystallization is
highly temperature dependent. As shown in Fig. 9, the
shape of the crystallization curve varies greatly between
693 and 743 K.
For all crystallization above the nose, there appears to
be a relatively large statistical scatter in onset time. The
material appears unchanged for long periods of time, until
a rapid crystallization event occurs. In order to assess this
scatter, eight measurements were made of the 5% and 95%
crystallization times at 703 K. The average value for t
5%
was 180 s, with a large standard deviation of 78 s. How-
ever, once crystallization was initiated it proceeded rapidly
and consistently, with an average value for t
95%
t
5%
of
34 s and a standard deviation of just 0.7 s. This behavior
Fig. 8. Partial TTT diagram for crystallization and relaxation of
Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
.
Fig. 9. DSC crystallization isotherms, taken above the nose of the TTT
diagram. Curves have been shifted to allow for comparison.
B.A. Legg et al. / Acta Materialia 55 (2007) 11091116 1113
was previously reported in both Zr- and Pd-based BMG
systems and is explained as a nucleation controlled crystal-
lization process [21,23].
3.4. Activation energy
The lower temperature DSC crystallization isotherms
were modeled using JohnsonMehlAvramiKolomogrov
(JMAK) theory in an attempt to estimate the activation
threshold for nucleation. In order to conduct this analysis,
several assumptions were made about nucleation and
growth mechanisms.
The crystal growth rate, u, was determined under an
assumption of interfacial controlled growth, such that
u f
k
b
T
3pa
2
0
g
1 exp
N
A
DG
lx
k
b
T
_ _ _ _
: 12
Here, k
b
is the Boltzmann constant. The value f represents
an accommodation factor for acceptance of new atoms
from the liquid. The formation of highly elongated crystals
suggests a value of f < 1. In this study it is assumed that
f = 0.2(T
f
T
x
)/T
f
0.057, following Onorato and Ulh-
mann [24].
The rate of crystal nucleation, I
ss
, is approximated using
Turnbulls expression for steady state nucleation
I
ss

A
v
g
exp
16pc
3
sl
3k
b
TDG
2
v
_ _
; 13
where A
v
is a function of the atomic volume, c
sl
is the
solidliquid interfacial energy and DG
v
is the free energy
dierence per unit volume [25]. The prefactor A
v
can be
approximated as 1.5 10
36
Pa/m
3
, using the expression
A
v
N
0
v
k
b
T=3pa
3
0
, where N
0
v
is the number of atoms per
unit volume [26]. The nucleation rate is very sensitive to
interfacial energy, because the activation energy for forma-
tion of a crystal nucleus has a cubic dependence on c
sl
.
However, initial analysis of DSC crystallization curves
with JMAK theory indicates that the nucleation rate is
not constant, but increases with time. This observation
may be understood by a theory of transient nucleation.
When a material has been rapidly quenched to low temper-
atures, a certain period of time is needed to establish the
atomic cluster distribution to support steady state nucle-
ation. This transient period has a characteristic timescale,
which can be calculated to an order of magnitude as
s
tr

n
2
a
2
0
o

D
; 14
where n
*
is the number of atoms in a critically sized nu-
cleus, o
*
is the number of atoms in the surface of the critical
nucleus and D is the diusion coecient [26,27]. For a gi-
ven temperature and interfacial energy, the transient time
can be determined by using a critical nucleus radius of
r
*
= c
sl
/DG
v
and estimating the diusion coecient from
viscosity by the StokesEinstein relation. However, Eq.
(14) is only accurate to an order of magnitude and to
accommodate experimental deviations, an additional t-
ting coecient of C
tr
is introduced, as shown in Eq. (15).
s
tr
C
tr

n
2
a
2
0
o

D
: 15
Under these conditions of interfacial growth and steady
state nucleation, JMAK theory allows the crystallized vol-
ume fraction, x, to be obtained as a function of time as
x 1 exp
p
3
u
3
I
ss
t s
tr

4
_ _
16
Ref. [27].
Because crystallization enthalpy is roughly proportional
to crystallized volume fraction, the derivative of Eq. (16)
Fig. 10. Comparison of isothermal DSC heat ow curves (solid line) with
theoretical JMAK models (dashed line). The JMAK model was obtained
using an interfacial energy of c
sl
= 0.86 J/m
2
and a transient period
adjustment factor of C
tr
= 7.9. Measurements were performed at (a)
568 K, (b) 558 K, (c) 548 K and (d) 538 K.
Fig. 11. TTT diagram for crystallization of the PtCuNiP glass at low
temperatures, comparing individual DSC measurements with the JMAK
prediction.
1114 B.A. Legg et al. / Acta Materialia 55 (2007) 11091116
can be compared directly with DSC heat ow curves. The
tting parameters of c
sl
and C
tr
were determined by opti-
mizing crystallization peak height and peak time, respec-
tively. Best ts where found to when c
sl
= 0.086 J/m
2
and
C
tr
= 7.9. The actual and theoretical isothermal crystalliza-
tion curves are compared in Fig. 10, and a remarkable cor-
relation is seen. A dierent perspective is provided in
Fig. 11, where the results are shown on a TTT diagram.
Small deviations can likely be attributed to the occurrence
of a secondary crystallization peak at high temperatures,
simplication of the initial transient period as a step func-
tion, and neglecting heterogeneous nucleation.
4. General discussion
The Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
alloy possesses an excellent
GFA, which is reected in its critical casing thickness of
up to 20 mm when B
2
O
3
is used [3]. The three main factors
that have commonly been used to explain the GFA of an
alloy are the bulk thermodynamic driving force for crystal-
lization, the atomic mobility (described by kinetic fragility)
and the interfacial energy (which leads to an activation
energy for nucleation). In this work, all three contributions
were determined.
In terms of atomic mobility, the low T
g
, excellent form-
ability [4] and ductile behavior [5] of this alloy suggest a
fragile liquid behavior [6] and this was quantitatively con-
rmed through T
g
shift measurements. Within several alloy
families, the glass-forming ability of an alloy has been seen
to correlate with fragility, such that stronger liquids have
better GFA [7,8]. The excellent glass former Zr
41.2
Ti
13.8
-
Cu
12.5
Ni
10
Be
22.5
, with a D
*
exceeding 20, can be considered
a kinetically stabilized alloy [28]. However, with a D
*
of
just 16.4, the Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
alloy has less excep-
tional kinetic properties.
Thermodynamics appears to be of critical importance in
other systems. For example, Pd
43
Cu
27
Ni
10
P
20
is even more
fragile than Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
[29], but can be consid-
ered a thermodynamically stabilized alloy due to an excep-
tionally low DS
f
[19]. But with Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
, the
nding of a large DS
f
, and its correspondingly large DG
lx
would not suggest good GFA. The driving force for crys-
tallization of Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
is even larger than that
of the binary ZrNi glass former, which has a critical cool-
ing rate of 10
4
K/s, which obviously makes this criterion
insucient to explain the alloys GFA.
Having analyzed thermodynamics and kinetics, we pos-
tulate that the liquidcrystal interface plays a critical role
in determining the alloys GFA. It has been shown through
various examples that interfacial energy can control the
path toward equilibrium [30]. This becomes apparent in
Eq. (12), where the interfacial energy has the strongest inu-
ence on the nucleation rate. Taking into account the alloys
thermodynamic and kinetic properties and using a classical
nucleation and growth model, it appears that a solidliquid
interfacial energy of 0.086 J/m
2
is required to explain the
alloys crystallization behavior. This is signicantly larger
than for Pd
43
Cu
27
Ni
10
P
20
at 0.079 J/m
2
[23], Pd
40
Cu
30
-
Ni
10
P
20
at 0.067 J/m
2
[25] and Zr
41.2
Ti
13.8
Cu
12.5
Ni
10
Be
22.5
at 0.04 J/m
2
[18]. Thus, it seems likely the Pt alloy can be
considered an interface stabilized system. A high interfacial
energy is not unexpected, because it is in keeping with
models by Spaepen, which predict proportionality between
c
sl
and DH
f.
However, the conclusions should be treated with cau-
tion. The crystallization modeling focused on low-temper-
ature regimes where crystal growth is very slow. These
slow growth rates serve to isolate nuclei and limit the inu-
ence of heterogeneous nucleation. Conversely, at tempera-
tures above the nose, crystallization appears to be initiated
from a single source and heterogeneous nucleation is of
critical importance. When comparing results with Zr and
Pd alloys, it is possible that their lower reported values of
c
sl
actually reect the degree to which heterogeneous nucle-
ation aects crystallization kinetics.
Indeed, in all three alloy families, a large eect of heter-
ogeneous inuences on the crystallization kinetics has been
observed. The critical cooling rate of Pd-based BMG could
be improved signicantly by processing in B
2
O
3
, which
reduces the oxides of all the constituents in the alloy [25].
A dierent approach was taken in Zr-based BMG when
Sc and Y were used to reduce the oxides of the constituents
of the alloy [31]. The addition of Sc to Zr improves critical
cooling rate by about one order of magnitude. For the Pt
alloy, it was reported that when exposed to B
2
O
3
it can
be cast up to 20 mm. Without ux, the alloy can be only
be cast amorphous in rods with a diameter of 6 mm.
5. Conclusion
The thermodynamic functions Pt
57.3
Cu
14.6
Ni
5.3
P
22.8
were
measured and a large driving force for crystallization was
determined. The kinetic fragility was measured and the
material was found to have a fragility parameter of 16.4.
So, while the materials glass-forming ability remains high,
neither the driving force nor the fragility parameter alone
would indicate exceptional GFA.
Crystallization was seen to proceed by two dierent
mechanisms. At temperatures above the nose of the TTT
diagram, crystallization seems to be initiated by a random
heterogeneous nucleation event, followed by rapid crystal
growth. This results in a large statistical variation in onset
time, but a fairly consistent growth period. At low temper-
atures, both onset and growth rates were highly repeatable.
Crystallization microstructure suggests that low-tempera-
ture growth occurs by a relatively slow process of interfa-
cial controlled growth. Low-temperature crystallization
isotherms indicate excellent stability against crystallization,
with crystallization times roughly four orders of magnitude
longer than relaxation times. This results in an exception-
ally large processing window for superplastic forming.
The long crystallization times can be explained by classical
nucleation and growth models with a relatively large acti-
vation threshold for crystallization. However, from this
B.A. Legg et al. / Acta Materialia 55 (2007) 11091116 1115
analysis we cannot determine if the origin of the large acti-
vation energy is due to a reduction in heterogenous inu-
ences, which previous results suggest plays a crucial role
in crystallization kinetics.
Acknowledgement
This work is based upon support by the National
Science Foundation under Grant No. DMR-0205940.
References
[1] Schroers J, Johnson WL. Appl Phys Lett 2002;80:2069.
[2] Nishiyama N, Inoue A. Mater Trans JIM 1997;38:464.
[3] Schroers J, Johnson WL. Appl Phys Lett 2004;84:3666.
[4] Schroers J. JOM 2005;57(5):35.
[5] Schroers J, Johnson WL. Phys Rev Lett 2004;93:255506.
[6] Novikov VN, Sokolov AP. Nature 2004;431:961.
[7] Shadowspeaker L, Busch R. Appl Phys Lett 2004;85:2508.
[8] Mukherjee S, Schroers J, Johnson WL, Rhim WK. Phys Rev Lett
2005;94:245501.
[9] Glade SC, Busch R, Lee DS, Johnson WL, Wunderlich RK, Fecht H-
J. J Appl Phys 2000;87:7242.
[10] Kubaschewski O, Alcock CB, Spencer PJ. Materials thermochemis-
try. New York: Pergamon Press; 1993.
[11] Busch R. JOM 2000;52(7):39.
[12] Busch R, Liu W, Johnson WL. J Appl Phys 1998;83:4134.
[13] Kuno M, Shadowspeaker LA, Schroers J, Busch R. Mater Res Soc
Symp Proc 2004;806:MM5.2.
[14] Busch R, Kim YJ, Johnson WL. J Appl Phys 1995;77:4039.
[15] Gallino I, Shah M, Busch R. Acta Mater 2006.
doi:10.1016/j.actamat.2006.09.040.
[16] Ediger MD, Angell CA, Nagel SR. J Phys Chem 1996;100:
13200.
[17] Glasstone S, Laidler KJ, Eyring H. The theory of rate processes. New
York: McGraw-Hill; 1941.
[18] Masuhr A, Waniuk TA, Busch R, Johnson WL. Phys Rev Lett
1999;82:2290.
[19] Fan GJ, Lo er JF, Wunderlich RK, Fecht H-J. Acta Mater
2004;52:667.
[20] Kato H, Wada T, Hasegawa M, Saida J, Inoue A, Chen HS. Scripta
Mater 2006;54:2023.
[22] Schroers J, Johnson WL. Appl Phys Lett 2000;76:2343.
[21] Shadowspeaker L, Hono K, private communication, National
Institute of Materials Science, Tsukuba, Japan, 2005..
[23] Schroers J, Wu Y, Busch R, Johnson WL. Acta Mater 2001;49:
2773.
[24] Onorato PIK, Uhlmann DR. J Non-Cryst Solids 1976;22:367.
[25] Lo er JF, Schroers J, Johnson WL. Appl Phys Lett 2000;77:
681.
[26] Uhlmann DR. Crystallization and melting in glass-forming systems.
In: Gray TJ, Frechette VD, editors. Materials science research, vol.
4. New York: Plenum Press; 1969. p. 172.
[27] Christian JW. The theory of transformations in metals in alloys. Ox-
ford: Pergamon Press; 1965.
[28] Mukherjee S, Schroers J, Zhou Z, Johnson WL, Rhim W-K. Acta
Mater 2004;52:3689.
[29] Fan GJ, Fecht H-J, Lavernia EJ. Appl Phys Lett 2004;84:487.
[30] Schroers J, Volkmann T, Herlach D, Allen DR, Perepezko JH. Int J
Rapid Sol 1996;9:267.
[31] Ku ndig AA. Mater Trans 2002;43:3206.
1116 B.A. Legg et al. / Acta Materialia 55 (2007) 11091116

You might also like