Intermetallics 12 (2004) 1191–1195

www.elsevier.com/locate/intermet

Molecular dynamic simulation of glass formation in binary

liquid metal: Cu –Ag using EAM
L. Qi, H.F. Zhang*, Z.Q. Hu
Shenyang National Laboratory for Materials Science, Institute of Metal Research, Chinese Academy of Sciences,
No. 72 Wenhua Road, Shenyang 110016, China
Available online 28 May 2004

Abstract
Based on the embedded-atom method, a constant-pressure, constant-temperature molecular dynamics technique is applied to obtain an
atomic description of glass formation process in eutectic Cu40Ag60 alloy. By using radial distribution function and pair analysis methods, the
structure and glass forming ability of this alloy is studied by quenching from the liquid at different cooling rates (1 £ 1011, 1 £ 1012, 5 £ 1012,
1 £ 1013 and 1 £ 1014 K/s). It is observed that the retention of amorphous structure requires extremely high cooling rates. Structure analyses
of the alloys in the simulations reveal the evolvement of the different clusters at various quenching rate during the quenching process.
q 2004 Elsevier Ltd. All rights reserved.
Keywords: B. Glasses, metallic; C. Rapid solidification processing; E. Simulations, atomistic

1. Introduction Mei [19] applied the nearest-neighbor model of EAM

When a liquid metal is quenched through the supercooled
region, a phase transition from liquid to glass takes place.
Several techniques have been proposed to obtain a
disordered state [1 – 3]. Among them the rapid solidification
method is widely used for the amorphous phase. However,
due to the demand of a high cooling rate this method is
restricted in most experimental cases. The computer
simulation of molecular dynamics (MD) is therefore
applied.
Recent years have witnessed considerable progress in the
development of empirical or semi-empirical many-body
potentials for MDs simulations. The embedded-atom
method (EAM) of Daw and Baskes [4,5] (and further
developed by Johnson [6 –9]), N-body potentials proposed
by Finnis and Sinclair (FS) [10] (and further developed by
Ackland et al. [11 –13]) and the tight-binding model of
many-body potentials given by Rosato et al. [14] consist of
three main aspects of the present empirical or semi-
empirical many-body potentials [15]. The EAM has been
successfully applied to a wide range of aspects in the solid
state, such as in point defects, dislocations and surfaces
[16 – 18].
* Corresponding author. Tel./fax: þ 86-24-23891320.
E-mail address: hfzhang@imr.ac.cn (H.F. Zhang).
0966-9795/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.intermet.2004.04.003
proposed by Johnson [3] to the liquid state by using MDs
simulations; the results for structure factors and diffusive
coefficients of noble metals are in good agreement with
experiments. In 1999, Qi and Çagin [20] had simulated the
glass formation and crystallization of Cu –Ag alloy using
MDs in conjunction with the quantum Sutton –Chen (Q-SC)
force fields (FF), last year Sheng et al. [21] also studied
Cu – Ag alloy based on the same potential. In this paper,
a different potential—EAM developed by Mei is
used to discuss the structure and glass formation ability
of the eutectic Cu – Ag alloy (Cu40Ag60) with different
cooling rates (1 £ 1011, 1 £ 1012, 5 £ 1012, 1 £ 1013 and
1 £ 1014 K/s).
2. Simulation methods
In the EAM, the total energy of an N-atom system has the
form [4]
X X
Etot ¼ F i ð ri Þ þ fi;j ðri;j Þ
i i;j
i–j
where Fi;j ðri;j Þ is a two-body central potential between atom
i and j with the separation distance ri;j; ; Fi ðri Þ is
the embedding energy of atom i with the electron density
1192 L. Qi et al. / Intermetallics 12 (2004) 1191–1195

Table 1
Model parameters for the Cu and Ag metals

Cu Ag

EC 2.95 4.44
fe 1.0 2.4118
a 5.909 5.082
b 5.85a 5.96a
d 12.06a 12.77a
g 6.82054a 7.00515a
fe 0.1217a 0.0990a

fe is in unit of eV. Other parameters are dimensionless.

a
Reference [19].

ri due to all its neighbors:

X
ri ¼ fj ðrij Þ Fig. 1. Average volume of Cu40Ag60 during cooling at a series of cooling
jð–iÞ rates.

Three formulae to establish the EAM potential of Cu40Ag60
can be obtained as follows:
r ¼ re exp½2bðr1 =r1e 2 1Þ
fðrÞ ¼ 2fe ½1 þ dðr=r1e 2 1Þexp½2gðr=r1e 2 1Þ ð2Þ
   a=b
a r r
FðrÞ ¼ 2EC 1 2 ln
b re re
1 X
þ fe sm exp½2ðpm 2 1Þg
2 m
   pm ðg=bÞ
d r r cooling process at different cooling rates. The RDF shows
£ 1 þ ðpm 2 1Þd 2 pm ln ð3Þ
b re re
The parameters in these formulae are listed in Table 1. The
details can be searched in Ref. [19].
The simulations were performed with the system
consisting of 640 atoms in a cubic box with periodic
boundary conditions along all the three directions. The
simulations are based on constant-temperature, constant-
pressure method (NPT-MD). The typical time step is order
of a few femtoseconds when the system settles down in a
thermal equilibrium. Equations of motion are numerically
integrated using the Verlet algorithm. For each quenching
experiment, the starting liquid state is obtained by heating
the solid slowly through the liquidus temperature. The
system is melted and homogenized at a temperature well
above the liquidus in the range of 1300– 1500 K and
then rapidly cooled with the desired rate (e.g. 1 £ 1011,
1 £ 1012, 5 £ 1012, 1 £ 1013 and 1 £ 1014 K/s) down to
room temperature, noting the change of volume and
structure to decreasing of temperature, using pair analysis
(PA) method [22] to study the change of microstructure
during the cooling process, and radial distribution function
(RDF) to analyze the glass forming ability of the alloy.
3. Results and discussion
Fig. 1 shows the variation of the volume as Cu40Ag60 is
cooled at different cooling rates. A step downward jump in
volume for the cooling process is due to the crystallization
of the Cu40Ag60 alloy. Different cooling rate leads to
different crystallization temperature, the larger the cooling
ð1Þ
rate the lower the crystallization temperature. At 1 £ 1011
and 1 £ 1012 K/s cooling rate, the crystallization tempera-
ture is about 880 and 840 K; at 1 £ 1013 K/s cooling rate, the
crystallization temperature becomes 540 K. However, at
1 £ 1014 K/s cooling rate, the variation of the volume is
continuous. This indicates the formation of a metallic glass.
Fig. 2 shows the RDF of the model structure during the
the melting structure as the sample is cooled from 1100 to
900 K. In Fig. 2(a) and (b), it proves the presence of an fcc
crystal structure at 800 K. However, at 800 K, the RDF still
shows the melting structure in Fig. 2(c) which is of 1 £ 1013
cooling rate, the reason is that fast cooling rate leads to low
crystallization temperature, at 1 £ 1013 K/s the sample
crystallization temperature is about 530 K (Fig. 1) so the
RDF still shows the melting structure at 800 K. In Fig. 2(d),
from 1100 to 700 K, the RDF exhibits the melting structure,
at 600 K the RDF shows the splitting of second peak. This
splitting of the second peak is a well-known characteristic
feature in the RDF of the existence of a metallic glass. Thus,
quenching the Cu40Ag60 alloy from the liquid to 300 K at
the rate of 1 £ 1014 K/s forms a metallic glass. The
amorphous formation temperature is decided by Wendt –
Abraham parameter RA ; defined by RA ¼ gmin =gmax : Here,
gmin ðgmax Þ is the value of gðrÞ at the first minimum
(maximum) in the RDF, as the result, the amorphous
formation temperature is about 620 K.
One method that can be used to analyze the local
environments is the pair analysis (PA) [22]. The PA is a
method for analyzing structures by a decomposition of the
RDF according to the local environment of the bonded pairs.
The PA can be used to distinguish between various local
packing orders, in particular fcc, bcc, hcp, and liquid
environments. Specifically, most of bonded pairs in fcc are
of type 1421, bcc are of 1442 and 1661, liquid and
amorphous are of 1551, 1431 and 1541, as shown in Fig. 3.
 L. Qi et al. / Intermetallics 12 (2004) 1191–1195 1193

Fig. 2. Radial distribution function (RDF) of Cu40Ag60 during the cooling processes at various cooling rates. (a) 1 £ 1011, (b) 1 £ 1012 (c) 1 £ 1013 and
(d) 1 £ 1014 K/s.

Using the atomic positions stored for the Ag –Cu alloy the liquid structure is held till room temperature, so at this
during cooling, its RDF is decomposed into a set of RDF’s, cooling rate a metallic glass is formed.
each for a specified PA type. The four major types of pairs Fig. 4 shows the two-dimensional cross-sectional
(1421, 1551, 1431, 1541) for Cu40Ag60 are monitored projections of the structures obtained at the different
during quenching. Fig. 3 shows the relative fraction of the quenching rates. These projections straightly exhibit
various PA pair types, as a function of temperature during a
set of quenching rates from 1 £ 1012 to 1 £ 1014 K/s.
Fig. 3(a) clearly indicates that there is an abrupt decrease
of 1431, 1551 and 1541 pairs, at 1 £ 1012 K/s cooling rate,
and the population of 1421 pairs (representing fcc) shows
abrupt increase at the same temperature which indicates the
crystallization of the liquid alloy. In Fig. 3(b) the case is
almost the same as in Fig. 3(a), the only difference is that the
crystallization temperature is lower than (a), this result is in
coincidence with the RDF and volume curve. In Fig. 3(c),
there is no decrease of 1431, 1551, 1541 pairs, and 1421
pairs do not change abruptly also. This indicates that

Fig. 4. Two-dimensional cross-sectional projections of the structures

obtained at different quenching rates. (a) 1 £ 1011, (b) 1 £ 1013, and
Fig. 3. The schematic diagram for some bond pairs. (c) 1 £ 1014 K/s.
1194 L. Qi et al. / Intermetallics 12 (2004) 1191–1195

the structure of the Cu40Ag60 alloy. The open ball represents

Ag and the close ball represents Cu. Note that a clearer
visualization would require a three-dimensional view with
crystal rotations. However, some differences in terms of
homogeneity can already be seen in these two-dimensional
cross-sectional projections [21]. For the two slower cooling
rates Fig. 5(a) and (b) shows clustering. Moreover, the
distribution of atoms in Fig. 5(a) is more orderly than that in
Fig. 5(b), this is because lower cooling rate means more
diffuse time, that is, at 1 £ 1011 K/s cooling rate, the
atoms have more time to move than at 1 £ 1013 K/s
cooling rate. In Fig. 5(c), the open ball and the close ball
distribute unorderly and homogeneously. This means due to
slower atomic movements the alloy remains liquid structure
at the quench rate of 1 £ 1014 K/s, e.g. a metallic glass is
formed.

4. Summary and conclusion

This MD study is aimed at studying the glass forming

ability and uncovering the atomic-level structure of the
Ag –Cu alloy obtained through rapid quenching at different
extreme cooling rates. A series of simulations has been
performed and the following results are summarized.

(1) The results presented here shown that the EAM

function can correctly predict the glass transition and
crystallization of liquid alloys during rapid
solidification.
(2) The retention of amorphous structures requires extre-
mely high-cooling rate (1 £ 1014 K/s).
(3) The crystallization temperature depend on the quench-
ing rate, the larger the quenching rate the lower the
crystallization temperature.
(4) Structure analyses of the alloys in the simulations
revealed the evolvement of the different clusters at
different quenching rates during the quenching process.

Acknowledgements

This work was supported by the National High

Technical Research and Development Programme of
China (2001AA331010) and National Key Basic
Research and Development Programme of China
(G2000067201).

References

[1] Massobric C, Pontikis V, Martin G. Phys Rev 1990;B41:10486.

[2] Cardellini F, Contini V, Mazzone G. Scripta Metall Mater 1995;4:641.
[3] Jang JSC, Koch CC. J Mater Res 1990;5:498.
Fig. 5. Relative amounts of various atomic-bonded pairs in Cu40Ag60 [4] Daw MS, Baskes MI. Phys Rev Lett 1983;50:1285.
at different temperatures during different cooling rates. (a) 1 £ 1012, [5] Daw MS, Baskes MI. Phys Rev 1984;B29:6443.
(b) 1 £ 1013 and (c) 1 £ 1014 K/s. [6] Johnson RA. Phys Rev 1988;B37:3924.
 L. Qi et al. / Intermetallics 12 (2004) 1191–1195 1195

[7] Johnson RA. Phys Rev 1988;B37:6121. [15] Chen KY, Liu HB. J Phys Condens Matter 1995;7:2379.
[8] Johnson RA. Phys Rev 1989;B39:12554. [16] Norskov JK, Lang ND. Phys Rev 1980;B21:2131.
[9] Johnson RA. Phys Rev 1990;B41:717. [17] Stott MJ, Zaremba E. Phys Rev 1980;B22:1564.
[10] Finnis MW, Sinclair JE. Philos Mag 1984;A50:45. [18] Foiles SM, Baskes MI, Daw MS. Phys Rev 1986;B33:7983.
[11] Ackland GJ, Thetford R. Philos Mag 1987;A56:15. [19] Mei J. Phys Rev 1992;B43:4653.
[12] Ackland GJ, Tichy G, Vitek V, Finnis MW. Philos Mag 1987;A56:15. [20] Qi Y, Çagin T. Phys Rev 1999;B59:3527.
[13] Ackland GJ, Vitek V. Phys Rev 1990;1349:2681. [21] Sheng HW, He JH, Ma E. Phys Rev 2002;B65:184203.
[14] Rosato V, Guillope M, Legrand B. Philos Mag 1989;A59:321. [22] Honeycutt JD, Andersen HC. J Phys Chem 1987;91:4950.