You are on page 1of 8

The thermodynamics of amorphous phases in immiscible systems: The example of

sputter-deposited Nb–Cu alloys


C. Michaelsen, C. Gente, and R. Bormann

Citation: Journal of Applied Physics 81, 6024 (1997); doi: 10.1063/1.364451


View online: https://doi.org/10.1063/1.364451
View Table of Contents: http://aip.scitation.org/toc/jap/81/9
Published by the American Institute of Physics

Articles you may be interested in


Nanoscale phase separation in amorphous immiscible copper-niobium alloy thin films
Applied Physics Letters 90, 021904 (2007); 10.1063/1.2429017

The enthalpy state of amorphous alloys in an immiscible system


Applied Physics Letters 78, 1343 (2001); 10.1063/1.1352040

The microstructure and electrical transport properties of immiscible copper-niobium alloy thin films
Journal of Applied Physics 103, 033511 (2008); 10.1063/1.2836970

Stability of amorphous Cu/Ta and Cu/W alloys


Journal of Applied Physics 58, 3052 (1985); 10.1063/1.335855

Design of radiation resistant metallic multilayers for advanced nuclear systems


Applied Physics Letters 104, 241906 (2014); 10.1063/1.4883481

Bulk Metallic Glasses


Physics Today 66, 32 (2013); 10.1063/PT.3.1885
The thermodynamics of amorphous phases in immiscible systems: The
example of sputter-deposited Nb–Cu alloys
C. Michaelsen,a) C. Gente, and R. Bormann
Institute of Materials Research, GKSS Research Center, 21502 Geesthacht, Germany
(Received 9 December 1996; accepted for publication 29 January 1997)
Amorphous metallic alloys, frequently observed to occur in systems with large negative heats of
mixing, are much less common in systems which are immiscible in the equilibrium solid state, such
as Nb–Cu. However, amorphous Nb–Cu alloys can be produced over a wide composition range by
sputtering. Using isothermal and nonisothermal differential scanning calorimetry, both the kinetics
and the thermodynamics of these amorphous Nb–Cu alloys were characterized quantitatively. It
was found that the formation enthalpies of the amorphous alloys amounted to only 4.5–7.6 kJ/g
atom. These data were combined with a modeling of the thermodynamic functions of the system.
The unexpected low enthalpies and Gibbs energies of the amorphous phase demonstrate the
thermodynamic stabilization of the liquid phase which develops with undercooling. This is
connected with a change of sign in the heat of mixing of the liquid phase, which is positive at high
temperatures and negative at low temperatures. © 1997 American Institute of Physics. [S0021-
8979(97)07609-3]

I. INTRODUCTION
include low-temperature annealing, the experimental phase
Over the recent decades an increasing variety of non- formation ranges may approach those which are expected
equilibrium methods have been developed which allow the when metastable equilibria between amorphous and crystal-
synthesis of amorphous metallic alloys to be performed. 1–3 line solid solutions develop. Exceptions from these rules
Such methods are vapor phase deposition, melt quenching, may be observed in cases such as Nb–Cr, where the Gibbs
and, invented more recently, a number of methods which energy differences between the phases are small.6
enable the preparation of amorphous metallic alloys in the In contrast to systems with negative heats of mixing, in
solid state at relatively low temperatures. Among these latter the great majority of systems with positive heats of mixing
methods are vitrification by loading with hydrogen, solid- nonequilibrium processing results in the formation of super-
state reactions in multilayer films, mechanical alloying or saturated solid solutions rather than amorphous phases. Ex-
attrition, and particle irradiation. The tendency for amor- amples are systems which have very limited equilibrium
phous phase formation has been observed to be most pro- solid solubility and which do not form intermetallic phases,
nounced in alloy systems where the heats of mixing are such as Cu–Co, Cu–Fe, or Ag–Cu, which can nevertheless
strongly negative, hence in systems with strong tendency for be obtained as thermodynamically unstable solid solutions
compound formation. Examples are the Ti–Ni, Zr–Ni, and over wide composition ranges by mechanical alloying 7–9 as
the Zr–Co systems, in which the heats of formation of vari- well as by vapor phase deposition.10–13 Formation of solid
ous stable and metastable compounds, including the amor- solutions rather than amorphous phases in such systems is
phous phase, typically amount to several tens of kJ/g atom, 4,5 understandable from a thermodynamic viewpoint, in particu-
and in which amorphous phases can be obtained over wide lar in systems which exhibit immiscibility even in the liquid
composition ranges by most of the above mentioned meth- state. In such cases, the Gibbs energy of the amorphous
ods. This explains why investigations on amorphous phase phase, which is considered to be the low-temperature con-
formation have been carried out in the past mostly on sys- tinuation of the liquid phase, will not intersect with the
tems with negative heats of mixing. Gibbs energies of the solid solutions. Rather, the amorphous
Although the details of amorphous phase formation cer- phase will have Gibbs energies higher than the solid
tainly depend on the preparation method used, there is a solutions for all concentrations in such systems. Exceptions
general trend to correlate the phases formed and the concen- may occur in cases where the heats of mixing of the solid
tration ranges in which they occur with the thermodynamics solutions and the amorphous phases are quite different in
of the system considered.4 As the phases which compete with magnitude or even different in sign. A more sophisticated
amorphous phases are mainly the crystalline solid solu- thermodynamic ap- proach of amorphous phases would have
tions, two approaches are proposed to describe the composi- to consider the fact that amorphous phases are usually more
tion ranges of amorphous phase formation. First, the phase stable than expected from linear extrapolations of the liquid
formation ranges in cases where partitioning is completely Gibbs energies to low temperatures. 14 This stabilization of
suppressed by processing are usually given by the T 0 curves the amorphous phase originates from short-range ordering
where the Gibbs energies of amorphous and crystalline solid which develops in the liquid phase upon undercooling.
solutions intersect. Second, when the preparation conditions However, until now, this has only been investigated in
systems with negative heats of mixing in the
a)
liquid/amorphous phase.
Corresponding author. Electronic mail: carsten.michaelsen@gkss.de
Although these arguments indicate that amorphous
phases are unlikely to occur in alloy systems with positive

6024 J. Appl. Phys. 81 (9), 1 May 1997 0021-8979/97/81(9)/6024/7/$10.00 © 1997 American Institute of Physics
+ 12 kJ/g atom at 1800 °C. The positive heat of mixing in
the liquid phase is further revealed by the (eutectic or peri-
tectic) solidification temperature which is accepted in all pre-
vious investigations to be at 1080 ± 10 °C, close to the melt-
ing point of pure Cu. If the heat of mixing in the liquid phase
would be zero, or even negative, this solidification tempera-
ture would be significantly lower, and the phase diagram
would display a much more pronounced eutectic behavior.
For example, the eutectic temperature would be depressed by
approximately 200 °C if the liquid heat of mixing would be
zero. Since this is not observed, it can be deduced unambigu-
ously by a simple consideration of the phase diagram that the
liquid heat of mixing is positive. In spite of this high positive
heat of mixing, we have found the formation of amorphous
FIG. 1. Phase diagram of the Nb–Cu system, determined by the CALPHAD phases over a wide composition range to occur by sputtering.
method. The data points are taken from Refs. 24 and 25. The relatively poor In the present article, isothermal and nonisothermal DSC is
agreement between experimental and calculated liquidus curves is discussed used to characterize both the kinetics and the thermodynam-
in the appendix.
ics of amorphous Nb–Cu phases quantitatively. These data
are combined with a modeling of the thermodynamic func-
heats of mixing, there is a minor number of investigations tions of the system, thus providing deeper insight into the
that report on amorphous phase formation is systems such as thermodynamics of amorphous phase formation in an appar-
Ag–Fe,15,16 Ag–Co,17 Ag–Ni,17 Ag–Cu,11 W–Cu,18,19 and ently immiscible system.
Ta–Cu,19 some of which are immiscible even in the liquid
state. Some of these publications appear not fully convincing II. EXPERIMENT
when only diffraction methods were used to determine Alloy thin films with concentrations in the range 32–77
whether a phase was amorphous. However, additional sup- at. % Cu were deposited by triode-magnetron sputtering
port for the presence of an amorphous phase was provided in from elemental Nb (4N5) and Cu (4N) targets onto sap-
some cases by investigations of the transformation behavior phire substrates that were cooled either by flowing water or
towards equilibrium, such as differential scanning calorim- by liquid nitrogen. The base pressure prior to the deposition
etry (DSC) measurements, which showed that this transfor- was 2 × 10—9 mbar, and the sputtering was carried out in a
mation takes place in a rather narrow temperature range, a 3.7 × 10—3 mbar argon ambient (7N) at a deposition rate of
feature that is characteristic of crystallization of an amor- 1 nm/s. Each film was about 4 µm in total thickness. A row
phous phase, or characteristic of usual nucleation-and- of 16 substrates, each of 7 mm width, was positioned be-
growth processes in general. 20–22 In contrast, much broader tween the sputter sources so that a concentration gradient
transformation curves are usually found upon decomposition occurred along them which amounted to about 3 at. % per
and subsequent grain growth of supersaturated crystalline sample (‘‘composition spread’’). The compositions of the
solid solutions.7,8,23 The formation of amorphous phases in films were determined by energy-dispersive x-ray spectrom-
systems which are immiscible even in the liquid state is ex- etry in a scanning electron microscope. The structural inves-
ceptional, and is in contrast to the current understanding of tigations were performed by Cu Ka x-ray diffraction (XRD)
the thermodynamics of liquid phases. It indicates that the in symmetrical Bragg–Brentano geometry using a Siemens
Gibbs energy of the amorphous phase is much lower than D5000 powder diffractometer.
expected from the equilibrium phase diagram, and suggests Free-standing samples for the DSC experiments were
that a change of sign in the enthalpy of mixing of the liquid obtained by peeling the films from the substrates. Subse-
state may take place during cooling. quently, the films were loaded into Cu pans. The DSC mea-
In the present article we report on our investigation of surements were performed under argon atmosphere (6N) in
amorphous phase formation in the Nb–Cu system. As can be a Perkin-Elmer DSC 7. The constant-heating-rate traces were
seen in Fig. 1, the mutual solubility in the solid terminal taken at 40 °C/min and the base lines were established by
phases is negligible. The true liquidus curve is not well de- repeating the measurements under identical conditions. An
termined, and previous investigations indicated that impuri- isothermal DSC measurement was performed by heating to
ties may have major influences on the liquidus curve, so that the desired temperature at a rate of 20 °C/min, and then hold-
either a very flat liquidus curve was found, as displayed in ing the temperature until there was no further change in the
Fig. 1, or a liquid miscibility gap may be observed. How- signal.
ever, from both types of phase diagrams it can be concluded
that the Gibbs energy of mixing of the liquid phase is small
III. RESULTS
in this temperature range. This implies that enthalpy and en-
tropy just compensate each other at these temperatures, As illustrated in Fig. 2, the XRD measurements of the
hence the liquid phase enthalpy must be positive. Assuming Nb–Cu alloy films which were deposited on liquid-nitrogen-
an entropy of an ideally random solution, this allows us to cooled substrates showed diffuse halos typical of amorphous
estimate the liquid enthalpy of mixing to be about phases for all concentrations investigated (32–77 at. % Cu),

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Michaelsen, Gente, and Bormann 602
FIG. 4. Isothermal DSC trace of a Nb54Cu46 sample at 410 °C. The solid
line was obtained by a fit of the JMA kinetic Eq. (1) to the data points.
FIG. 2. XRD patterns of Nb–Cu alloy films with various compositions,
deposited onto liquid-nitrogen-cooled substrates.

but significant exothermic heat release which starts at about


with a slight shift of the peak position upon composition 100 °C and extends to the beginning of crystallization. Since
variation. The samples deposited on water-cooled substrates XRD did not reveal any structural change associated with
were fully amorphous only in the range 35–74 at. % Cu, and this low-temperature DSC signal, we attribute this reaction to
exhibited additional crystalline diffraction peaks indicative a structural relaxation of the amorphous phase. This conclu-
of body-centered-cubic (bcc) and face-centered-cubic (fcc) sion is supported by observations of similar low-temperature
solid solutions for compositions of 32 and 77 at. % Cu, re- reactions to occur in a large variety of amorphous alloys
spectively. studied in our group.26
Figure 3 displays some typical DSC traces of Nb–Cu In order to characterize the kinetics of crystallization in
alloy films with various compositions which were deposited more detail, an isothermal DSC measurement was performed
at a substrate temperature of — 196 °C. Identical DSC traces which is shown in Fig. 4. The most important feature of this
were obtained from samples deposited at room temperature. measurement is the occurrence of a peak in the isothermal
All traces show the occurrence of a relatively sharp peak at DSC trace. The occurrence of a peak in the isothermal DSC
approximately 400–450 °C. XRD investigations performed trace is indicative of a transformation that takes place by
after the DSC measurements, as well as XRD patterns re- nucleation and growth. It is inconsistent with the kinetics of
corded in situ during heating in a hot-stage chamber (not grain growth of a nanocrystalline material, thus supporting
shown), indicate that this DSC peak is associated with the our conclusion based on XRD that the material transforming
crystallization of the amorphous phases into the pure ele- is truly amorphous.20,21 The measurement shown in Fig. 4
ments. Samples rich in Cu show a minor DSC peak at ap- was evaluated using the Johnson–Mehl–Avrami (JMA)
proximately 340 °C. In situ XRD measurements performed equation for nucleation and growth processes,27–29 in which
on these samples show that Cu precipitates prior to Nb for the transformed volume fraction is given by
these Cu-rich concentrations upon annealing. In addition to
these DSC peaks, all traces show the occurrence of a minor X( t ) =1—exp ( — ( kt ) n ) , (1)

where k is a temperature dependent rate constant, and n is


the Avrami exponent which depends on the nucleation and
growth mechanisms as well as the growth morphology. The
transformation rate, dX/dt, was assumed to be proportional
to the isothermal DSC signal. As can be seen in Fig. 4, Eq.
(1) provides an excellent fit to the data except for the initial
part of the trace, for an Avrami exponent n = 2.4. The initial
decaying portion of the trace is likely due to the aforemen-
tioned structural relaxation of the amorphous phase prior to
its crystallization. The Avrami exponent so obtained is close
to 2.5, a value which would be expected for a three-
dimensional diffusion-controlled growth of precipitates that
are formed at constant rates (although the JMA theory is
strictly valid only for early stages of diffusion-controlled
processes). The observation of crystallization to take place
FIG. 3. DSC traces of Nb–Cu alloy films with various compositions which
by nucleation and growth indicates that nucleation barriers to
were prepared by sputtering onto liquid-nitrogen-cooled substrates, taken at this transformation exist, and hence the amorphous phase is
a heating rate of 40 ° C/min. metastable rather than unstable.
FIG. 6. Gibbs energy curves of the Nb–Cu system at a temperature of
FIG. 5. Total heat release of Nb–Cu alloy films with various compositions,
25 °C, calculated by the CALPHAD method. Bcc Nb and fcc Cu are taken as
obtained from the DSC measurements. The error bars indicate an estimated
Gibbs energy reference states.
experimental uncertainty of ± 1 kJ/g atom. The solid lines are the
enthalpies of formation of the bcc, fcc, and amorphous phases calculated by
the CALPHAD method for the temperature of 425 °C.
rial, should lead to the formation of an amorphous phase in
this concentration range. This is in agreement with our ex-
The enthalpies of crystallization obtained by integration perimental results which show single amorphous phases in
of the DSC traces are shown in Fig. 5. For the integration, the range 35–74 at. % Cu when deposited onto water-cooled
the low-temperature heat flow which we had attributed to substrates, and over the whole concentration range investi-
structural relaxation was not taken into account, hence the gated (32–77 at. % Cu) when deposited onto liquid-nitrogen-
enthalpies plotted in Fig. 5 should correspond to a thermally cooled substrates. Due to the negative heat of mixing, the
relaxed amorphous state. The enthalpies of crystallization so amorphous phase should be metastable with respect to con-
obtained lie in the range 4.5–7.6 kJ/g atom. This result dem- centration fluctuations. Consequently, a nucleation barrier
onstrates that the amorphous phase enthalpy is only a few should exist to crystallization of the amorphous phase, and
kJ/g atom above that of the equilibrium mixture of the pure transformation of the amorphous phase towards thermody-
elements. It indicates a rather high thermodynamic stability namic equilibrium requires a nucleation process, in agree-
of the amorphous phase which is not expected from an in- ment with our DSC investigations which show a nucleation-
spection of the equilibrium phase diagram. and-growth-type crystallization behavior. This nucleation
process may not be necessary for the crystalline solid solu-
IV. DISCUSSION tions which may reduce their Gibbs energies continuously
via spinodal decomposition. Therefore, DSC traces of such
Together with various data concerned with the equilib- crystalline solid solutions are usually much broader than
rium phases, the enthalpy data shown in Fig. 5 were incor- those shown in Fig. 3.7,8,23 Such broad DSC traces often
porated into a regular-solution-type CALPHAD assessment of found with crystalline solid solutions are untypical of
the Nb–Cu system. As discussed in more detail in the appen- nucleation-and-growth transformations.20–23
dix, a relatively simple approach was used for the CALPHAD Figure 7 shows the heats of mixing of the liquid/
calculation, with a parabolic concentration dependence of
enthalpies and excess entropies for all phases. The enthalpies
so obtained are also displayed in Fig. 5. As can be seen,
there is relatively good agreement between experiment and
calcu- lation, although the measurements indicate an
approximately linear concentration dependence of the
amorphous phase en- thalpy which is not well reproduced by
the parabolic CALPHAD description. The linear concentration
dependence of the enthalpy suggests that the amorphous
phase may be phase separated, a feature which was not
considered in the present CALPHAD approach. Figure 5
demonstrates that the heats of mixing are positive in the solid
solution phases but negative in the amorphous phase.
Figure 6 shows the Gibbs energies of the bcc, fcc, and
amorphous phases at room temperature as obtained by the
present CALPHAD calculation. It can be seen that the amor-
phous phase has a lower Gibbs energy than the solid solu- FIG. 7. Heats of mixing of the liquid/amorphous phase at various tempera-
tures, calculated by the CALPHAD method. In the present CALPHAD approach,
tions in the composition range 35–78 at. % Cu. Therefore, the heat of mixing is not temperature dependent below the assumed glass
methods of nonequilibrium processing such as sputtering, transition temperature of 627 °C. Above this temperature, the heat of mixing
which tend to produce compositionally homogeneous mate- increases with a — 1/T dependence.
amorphous phase for different temperatures as a function of composition. It can be seen that the heat of mixing of the
liquid phase is positive at high temperatures and amounts to hibits positive heats of formation, and it indicates that Nb–
about 20 kJ/g atom at 2000 °C. In contrast, the heat of mix- Cu alloys have electron concentrations close to the value
ing in the amorphous phase is negative and amounts to about where the heats of formation change their sign. Further, since
— 15 kJ/g atom. Thus, unlike atoms in the liquid phase, they the binding energies strongly depend on the atomic spacings,
attract each other at low temperatures and repulse each other significant changes in the heats of formation or mixing are
at high temperatures. Liquid solubility at high temperature is induced when the interatomic distances change as a function
therefore solely caused by entropy contributions. A decrease of composition, or as a function of temperature due, for in-
of enthalpy and entropy of mixing as a result of the devel- stance, to thermal expansion. The occurrence of such a situ-
opment of chemical short-range order upon undercooling is ation is consistent with the unusual Nb–Cu liquidus curve
typical of liquid phases with negative heats of mixing. 14 which indicates a complicated Gibbs energy versus concen-
However, it is exceptional in the Nb–Cu system that the heat tration dependence (see appendix), and which may result in a
of mixing changes its sign with decreasing temperature. miscibility gap at rather Cu-rich concentrations (fairly repul-
It should be emphasized that this conclusion is not af- sive interaction) in contrast to full solubility at more Nb-rich
fected by details of the thermodynamic calculation. Rather, concentrations (fairly attractive interaction). To conclude,
the change of sign in the heat of mixing can be directly the Nb–Cu system appears to have a critical d-electron con-
deduced from the experiments. First, the positive heat of centration where the heat of mixing changes sign as a func-
mixing in the liquid phase at high temperatures can be con- tion of composition or temperature.
cluded directly from the equilibrium phase diagram, as dis-
cussed in the introduction. Second, the negative heat of mix- V. CONCLUSIONS
ing in the amorphous phase is required by the low
Amorphous Nb–Cu alloys were prepared by sputtering,
crystallization enthalpies. If we consider the amorphous and investigated by differential scanning calorimetry. Rather
phase as the low-temperature extrapolation of the liquid small crystallization enthalpies in the range 4.5–7.6
phase, this implies that the Gibbs energies of the amorphous kJ/g atom were obtained. This result demonstrates that the
pure elements Nb and Cu are reasonably approximated by heat of mixing in the amorphous phase is negative, whereas
the low-temperature extrapolation of the Gibbs energies of it is positive in the liquid phase at high temperatures as re-
liquid Nb and Cu. Since these latter values are clearly higher vealed by the equilibrium phase diagram, thus implying a
than the experimental enthalpies (see Fig. 6), we can con- change of sign in the heat of mixing of the liquid phase as a
clude that the heat of mixing in the amorphous phase must function of temperature which has not been observed before.
be negative. This demonstrates that the change of sign in the The results are discussed in terms of the electronic properties
heat of mixing of the liquid/amorphous phase can be de- of the system, and it is suggested that the Nb–Cu alloy sys-
duced directly by a comparison of the measured crystalliza- tem has a critical d-electron concentration where the heats of
tion enthalpies with the equilibrium phase diagram, without mixing can change sign as a function of composition or tem-
any thermodynamic modeling. perature. Correspondingly, the formation of amorphous
The change of sign in the liquid enthalpy of mixing may phases can be expected to occur in a larger number of alloy
be explained in terms of the tight-binding model of the elec- systems which have similar electronic properties and which
tron theory for equiatomic AB transition metal alloys.30 In are lacking in equilibrium solid solubility.
this model the heat of formation is determined by two con-
tributions. The first contribution is attractive, and it results ACKNOWLEDGMENT
from the formation of an average alloy band written in the
The authors thank H. Teichler for helpful discussions.
virtual crystal approximation. It becomes most pronounced
for average d-electron concentrations dN¯ = 5 when the
APPENDIX: THERMODYNAMIC MODELING OF THE
bond- ing orbitals are all filled and the anti-bonding ones are
NB–CU SYSTEM
all empty. The second term, which is repulsive, arises from
the loss of bonding due to the atomic energy level mismatch Gibbs-energy curves for the liquid, amorphous, bcc and
in the alloy, as well as from differences in the atomic sizes. fcc phases of the Nb–Cu system were calculated using the
As a result, the heat of formation (at equiatomic CALPHAD method. The present assessment was based on ex-
composition perimental data of Nb solubilities in Cu,25 vapor pressure
31
measurements of Bailey et al.,32 and the three-phase equilib-
and small differences in ¯
d-electron concentration) has a para- rium examined by Smith et al. These data were supple-
mented by the present data concerning enthalpies of the
bolic dependence upon N d , and (for 4d transition metal al- nega- tive heats of formation of various intermetallic
loys) is negative for dN¯ values in the range 3.5–6.5, equilibrium compounds. This may explain why the Nb–Cu
where system ex-
the band is around half full, and positive outside this electron
concentration range. For the Nb–Cu system we have
d

= 7, a value which is located just outside the attractived
N¯ range. For comparison, the neighboring Nb–Ni
system has
N¯d =6.5, and is an attractive system as revealed by the
amorphous phases which are shown in Fig. 4. Conflicting perature range, and it implies that different impurity levels,
experimental liquidus data exist in the literature which show which give rise to minor changes of the Gibbs energies, can
the presence33 or the absence24,34 of a miscibility gap in the result in major changes of the phase diagram such as the
liquid phase. This indicates that the Gibbs energy curve of presence or absence of a miscibility gap.34 The occurrence of
the liquid phase is rather flat over a composition and tem-
a liquid phase miscibility gap with a critical point at about 85 TABLE I. Coefficients of the thermodynamic functions of the Nb–Cu sys-
at % Cu as reported by Terekhov et al.33 is unusual, as dis- tem determined by the CALPHAD method.
cussed by Okamoto et al.,35 and a CALPHAD modeling of this A (J g atom—1) B (J g atom—1 K—1) C (J g atom—1 K)
miscibility gap would require a large number of coefficients.
Liquid 169 182.84 —45.278 56 —104 166 724
Since impurities are believed to stabilize this liquid miscibil- Amorphous —62 314.56 83.303 42 0.0
ity gap,34 we used the liquidus points of Allibert et al.24 for bcc 49 328.91 0.0 0.0
our assessment in which the miscibility gap does not appear. fcc 19 600.91 16.756 04 0.0
In view of these uncertainties concerned with the liquid
phase, we employed a low weighting factor for these liqui-
dus points. In addition, we used a rather simple model in
which, following the calculation of Kaufman, 36 the liquid
lation of Kauzmann’s criterion. 38 This is avoided when the
phase has a parabolic concentration dependence of the
excess terms. The use of this simple approach explains the glass transition temperature is assumed to be higher than the
poor agreement between calculated and experimental phase crystallization temperature of approximately 700 K (427 °C).
dia- gram shown in Fig. 1. Alternatively, our calculations showed that violation of
Phase stabilities, G 0 (T), for pure Nb and Cu liquid bcc Kauzmann’s criterion could also be avoided for Tg = 700 K
and fcc phases were taken from the SGTE compilation.37 For when a substantially larger number of coefficients than occur
in Eq. A2 were used for the liquid/amorphous phase. How-

( )
the amorphous phase the pure element liquid phases were
taken as reference states. The model used for all phases was ever, in view of a lack of unambiguous experimental data for
the liquid Gibbs energy, we kept the number of coefficients
G i ( x,T ) =Giref +G idi +G xs . (A1) as small as possible and fixed Tg at 900 K (627 °C) in the
In this equation, G = xNbG (T) + xCuG (T) is the
ref 0,Nb 0,Cu present calculation. The coefficients so determined are given
in Table I.
i i i
Gibbs energy of a mechanical mixture of the elemental i
phases. Gidi in Eq. (A1) is the energy contribution due to
configurational entropy in an ideally random solution, and
Gxsi is given by
1
W. L. Johnson, Prog. Mater. Sci. 30, 81 (1986).
G ( x,T )=x Nb
xs
Cu( A+BT+C/T ) . (A2)
2
Materials Science and Technology, edited by R. W. Cahn, P. Haasen, E. J.
i Kramer, and J. Zarzycki (VCH, Weinheim, 1991), Vol. 9.
x
3
Due to lack of solubility data for the bcc phase, only the Physical Metallurgy, 4th ed., edited by R. W. Cahn and P. Haasen
(Elsevier, Amsterdam, 1996).
coefficient A was used for this phase. The coefficient C was 4
R. Bormann and K. Zo¨ ltzer, Phys. Status Solidi A 131, 691 (1992).
used only for the liquid phase in order to describe the excess 5
F. Ga¨ rtner and R. Bormann, J. Phys. (France) (Colloq.) 51, 4 (1990).
specific heat in the undercooled liquid.4,14 This excess spe- 6
C. Michaelsen, W. Sinkler, Th. Pfullmann, and R. Bormann, J. Appl.
cific heat is then given as 7
Phys. 80, 2156 (1996).
C. Gente, M. Oehring, and R. Bormann, Phys. Rev. B 48, 13244 (1993).
8
T. Klassen, U. Herr, and R. S. Averback, Mater. Res. Soc. Symp. Proc.
C 400, 25 (1996).
Ac p =—2x Nb x Cu . (A3) 9
E. Ma and M. Atzmon, Mater. Chem. Phys. 39, 249 (1995).
T2 10
E. Kneller, J. Appl. Phys. 33, 1355 (1962).
11
The absence of terms higher than linear in temperature in the S. Mader, H. Widmer, F. M. d’Heurle, and A. S. Nowick, Appl. Phys.
fcc, bcc, and amorphous phase descriptions implies confor- Lett. 3, 201 (1963).
12
13 E. Kneller, J. Appl. Phys. 35, 2210 (1965).
mity with the rule of Neumann–Kopp, for these three- C. Michaelsen, Philos. Mag. A 72, 813 (1995).
14
Cu R. Bormann, Mater. Sci. Eng. A 178, 55 (1994).
phases,i.e.,c p,i = p,i+ x Cu • c p,i .
• c Nb 15
C. L. Chien and K. M. Unruh, Phys. Rev. B 28, 1214 (1983).
xNb
The amorphous phase was thermodynamically treated as both the entropy and enthalpy of the amorphous phase to that
the low-temperature extrapolation of the liquid phase. Heats of the liquid phase at the glass temperature, T g . In cases such
of crystallization provided enthalpies of the amorphous as the Nb–Cu system where the measurements do not reveal a
phase relative to the equilibrium phases. Since the coefficient glass transition temperature, we usually set T g equal to the
C was used for the liquid but not for the amorphous phase, crystallization temperature, as is valid for most transition-
the liquid and amorphous phases had to be treated as two metal alloys. However, when this procedure was applied to
separate phases for computer reasons. The connection be- the Nb–Cu system using the model of Eq. (A2), the result was
tween liquid and amorphous phases was performed by fixing that the entropy of the amorphous phase fell below that of the
crystalline pure elements, leading to a vio- 16
M. Takao and H. Senno, J. Magn. Magn. Mater. 31–34, 949 (1983).
17
J. J. Hauser, Phys. Rev. B 12, 5160 (1975).
18
H. F. Rizzo, T. B. Massalski, and M. Nastasi, Metall. Trans. A 24, 1027
(1993).
19
M. Nastasi, F. W. Saris, L. S. Hung, and J. W. Mayer, J. Appl. Phys. 58,
3052 (1985).
20
L. C. Chen and F. Spaepen, Nature (London) 336, 366 (1988).
21
L. C. Chen and F. Spaepen, J. Appl. Phys. 69, 679 (1991).
22
C. Michaelsen and M. Dahms, Thermochim. Acta 288, 9 (1996).
23
C. Michaelsen, C. Gente, and R. Bormann, J. Mater. Res. (submitted).
24
C. Allibert, J. Driole, and E. Bonnier, C. R. Acad. Sci. (Paris) 268C, 1579
(1969).
25
H. Schlu¨ ter, Universita¨ t Go¨ ttingen (unpublished).
26
C. Michaelsen (unpublished).
27
W. A. Johnson and R. F. Mehl. Trans. Am. Inst. Min. Metall. Pet. Eng.
135, 1 (1939).
28
M. Avrami, J. Chem. Phys. 7, 1103 (1939); 8, 212 (1940); 9, 177 (1941).
29
J. W. Christian, The Theory of Transformations in Metals and Alloys, Part
I Equilibrium and General Kinetic Theory (Pergamon, Oxford, 1975).
30
Electron Theory in Alloy Design, edited by D. G. Pettifor and A. H.
Cottrell (Institute of Materials, London, 1992).
31
D. M. Bailey, G. R. Luecke, A. V. Hariharan, and J. F. Smith, J. Less- 34
D. J. Chakrabarti and D. E. Laughlin, Bull. Alloy Phase Diag. 2, 455
Common Met. 78, 197 (1981). (1982).
32 35
J. F. Smith, K. J. Lee, and D. M. Bailey, Bull. Alloy. Phase Diag. 5, 133 H. Okamoto and T. B. Massalski, J. Phase Equil. 12, 148 (1991).
36
(1984). L. Kaufman, Calphad 2, 117 (1978).
37
33 A. T. Dinsdale, NPL-Report No. DMA (A) 195, 1989 (unpublished).
G. I. Terekhov and L. N. Aleksandrova, Russ. Metall. 4, 218 (1984). 38
W. Kauzmann, Chem. Rev. 43, 219 (1948).

You might also like