This article was downloaded by: [Stanford University Libraries]

On: 18 September 2012, At: 09:02

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Philosophical Magazine Letters

Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/tphl20

Long-range Finnis–Sinclair potentials

a a
A. P. Sutton & J. Chen
a
Department of Metallurgy and Science of Materials, Oxford
University, Parks Road, Oxford, OX1 3PH, England

Version of record first published: 20 Aug 2006.

To cite this article: A. P. Sutton & J. Chen (1990): Long-range Finnis–Sinclair potentials,
Philosophical Magazine Letters, 61:3, 139-146

To link to this article: http://dx.doi.org/10.1080/09500839008206493

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-

conditions

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to date. The
accuracy of any instructions, formulae, and drug doses should be independently
verified with primary sources. The publisher shall not be liable for any loss, actions,
claims, proceedings, demand, or costs or damages whatsoever or howsoever caused
arising directly or indirectly in connection with or arising out of the use of this material.
 MAGAZINELETTERS,1990, VOL.61, No. 3,139-146
PHILOSOPHICAL

Long-range Finnis-Sinclair potentials

By A. P. SUTTONand J. CHEN
Department of Metallurgy and Science of Materials, Oxford University, Parks Road,
Oxford OX1 3PH, England
[Received in present form 6 December 1989 and accepted 1 I December 19891

ABSTRACT
Downloaded by [Stanford University Libraries] at 09:02 18 September 2012

Finnis-Sinclair potentials are developed for computer simulationsin which van

der Waals type interactionsbetween well separated atomicclustersare as important
as the description of metallic bonding at short range. The potentials always favour
f.c.c. and h.c.p. structures over the b.c.c.structure. They display convenient scaling
properties for both length and energy, and a number of properties of the perfect
crystal may be derived analytically.

6 1. INTRODUCTION
N-body Finnis-Sinclair potentials (Finnis and Sinclair 1984) and embedded atom
potentials (Daw and Baskes 1984) represent the current state of the art in the
description of cohesion in metals for computer simulations involving up to a million
atoms. In both approaches the total internal energy is represented by a cohesive
functional of pair interactions and a predominantly repulsive pair potential. Using
arguments based on the second-moment approximation in tight-binding theory,
Finnis and Sinclair (1984) argued that the functional should be the square root. The
functional gives rise to N-body interactions in the sense that the force exerted by one
atom on another depends on the disposition of all neighbours to both atoms concerned.
By contrast, a pair potential alone gives rise to a force that depends only on the
separation of the two atoms concerned. This feature produces a much more physically
correct description of relaxation at free surfaces of metals than can be achieved with a
pair potential. For example, it generally leads to an inward relaxation of the surface
layer, in agreement with experimental observations of many transition metal surfaces,
and in contrast to non-oscillatorypair potentials (see for example Rosato, Guillope and
Legrand (1989)).
The long-range Finnis-Sinclair potentials that we describe in this Letter were
developed to enable us to perform certain computer simulations. We are interested in
modelling the mechanical interactions between clusters of atoms. Using a Lennard-
Jones pair potential it has already been shown by Pethica and Sutton (1988), and
confirmed by Smith, Bozzolo, Banerjea and Ferrante (1989), that there exists a
mechanical instability when two slabs of material are brought into contact. The
instability arises when the second derivative of the energy of interaction between the
slabs, with respect to the separation of the two surfaces, equals zero. The instability is
characterized by the two surfacesjumping together to form a single uniaxially strained
crystal. Clearly, the interaction between the slabs is critically dependent on the range of
the potential. Pethica and Sutton chose a Lennard-Jones pair potential because the
potential correctly described the long-range van der Waals interactions between the
slabs. This may be contrasted with existing Finnis-Sinclair and embedded atom
potentials which extend typically to third neighbours in f.c.c. and b.c.c. crystals and do
09504839/90 S3QO 0 1990 Taylor & Francis Ltd.
 140 A. P.Sutton and J. Chen

not have a l/r6 van der Waals tail. On the other hand, Lennard-Jones potential gives a
poor description of surface relaxation in metals. For these reasons we have developed
Finni-Sinclair potentials that have a long range of interaction. The idea is that we will
combine the superior description of short-range interactions afforded by an N-body
potential, to obtain a good description of surface relaxation, with a van der Waals tail
to give a correct description of the long-range interactions.

$2. THEPOTENTIALS
Van der Waals interactions are normally represented by an attractive l/r6 pair
potential. The obvious method of introducing them into a Finnis-Sinclair description
is therefore to add a negative l/r6 tail to the pair-potential component. However, we
Downloaded by [Stanford University Libraries] at 09:02 18 September 2012

prefer to keep the attractive contributions to the total energy separated from the
repulsive contributions and therefore we write the total energy in the following Finnis-
Sinclair form:

where

Here rijis the separation between atoms i andj, cis a positive dimensionlessparameter,
E is a parameter with the dimensions of energy, Q is a parameter with the dimensions of
length, and m and n are positive integers. For reasons that are described below, n must
be greater than m. The pair potential, is purely repulsive and the N-body term is
purely cohesive. Let p' denote the value of p, given by the sum in eqn. (3), for an atom
belonging to a free surface. The cohesive contribution this atom makes to the total
energy of the surface is then e" = -ecJps. If an additional atom is now placed above the
surface, and at a separation R from our surface atom, the value of e" changes to
-EC~+(Q/R)"]"~.
For large values of R, compared with a, we may expand the square root:

Equation (4) shows that when an atom approaches the surface it interacts at large
separations in a pairwise fashion, although the magnitude of this pair potential is
affected by the number of neighbours to each surface atom. As the separation, R,
decreases, the expansion in eqn. (4) becomes increasingly invalid and the interaction
changes smoothly to an N-body form. Thus, by choosing m=6 in eqn. (3), we can
achieve our objective of the long-range interactions between two clusters of atoms
being described by an attractive l/r6 pair potential, whereas at small separations the
interaction is N-body in nature. Furthermore, the transition between these two limits is
smooth and continuous.
There are further advantages in choosing power laws in eqns. (2) and (3), which are
directly analogous to the scaling properties of Lennard-Jones potentials (see for
 Long-range Finnis-Sinclair potentials 141

example Allen and Tildesley (1987), p. 327). We may define E and a in eqn. (1) as our
units of energy and length. For a given crystal structure the Finnis-Sinclair potential of
eqn. (1) is defined completely by the exponents m and n. That is because the equilibrium
condition for a particular crystal structure fixes the parameter c. If two metals, with the
same crystal structures, may be represented by the same values of the exponents, m and
n, then the results obtained for one metal may be directly converted into results for the
other simply by rescaling the units of energy and length. Moreover, the relative
stabilities that our potential predicts for the f.c.c., b.c.c. and h.c.p. crystal structures are
also determined entirely by m and n.
Let s', denote the following lattice sum in a perfect f.c.c. crystal:
Downloaded by [Stanford University Libraries] at 09:02 18 September 2012

where the sum is taken over all atomic separations, ri. from an arbitrary atom. Setting a'
equal to the f.c.c. lattice parameter then fixes our unit of length. The equilibrium
condition is that the total energy of the crystal does not change to first order when the
lattice parameter is varied. This leads to

c=- nS',
4%'
The cohesive energy per atom is then given by

E', =-ES', (2n -m), (7)

2m
and the bulk modulus, B', is given by

where d is the atomic volume, (a')'/4. Using eqns. (7) and (8) we obtain the following
relation between the bulk modulus and the cohesive energy:
SZ'B' nm
-18' (9)

The pressure-volume relation for the f.c.c. crystal is also readily obtained from eqns.
(1H3), (5), (6) and (8):

where u is the volume of a sample of the crystal, the volume of which at equilibrium is ug.
B' is the bulk modulus at the equilibrium volume and is given by eqn. (8). Simple
expressions for the elastic constants cll, c12 and c44 may also be derived. Elastic
stability of the crystal requires c44 >0, cl >0 and c1 >c12.It may be shown that these
conditions are satisfied for a cubic crystal provided that n >m.
Table 1 contains the parameters E, c, m and n that have been fitted to a variety of f.c.c.
metals. For each metal the experimental cohesive energy and crystal lattice parameter
were fitted exactly. The indices m and n were restricted to integer values, such that mn is
the nearest integer to 18SZ'B'/E', (see eqn. (9)).This meant that the bulk modulus was not
 Downloaded by [Stanford University Libraries] at 09:02 18 September 2012

Table 1. Potential parameters for f.c.c. metals. The cohesive energy per atom, l$,in electronvolts and the lattice parameter, a‘, in Angstrom
units, are fitted exactly; experimental values are from Kittel(l971). B‘ is the calculated bulk modulus (eV A- ’) and the experimental value
is in brackets. The calculated elastic constants c1 c I zand c44 are in eV A-’, and the experimental values are in brackets. Experimental
elastic constants for Ni, Cy Pd, Ag, Ir, Pt and Au are from Simmons and Wang (1971); for Rh from Walker, Ashkenazi and Dacarogna
(1981); for Pd and A1 from Hirth and Lothe (1982). ?
a
m n E (ev) C l$ Uf B‘ C11 c12 c44 FI)
E
Ni 6 9 1 - 5 7 0 7 ~ 1 0 - ~ 39.432 4.44 352 1-22(1.17) 1.44 (1.63) 1.11 (0.94) 050 (0.82)
8
D
Cu 6 9 1.2382~lo-’ 39.432 3-50 3.61 0.89 (0.89) 1.06 (1.10) 0-81 (0.78) 036 (051) 0,
D
Rh 6 12 4.9371~10-’ 144-41 5-75 3-80 1-68(1.68) 2.12 (263) 1.45 (1.20) 089 (1.21) a
Pd 7 12 4.1790~10-’ 10827 3-94 3.89 1.25 (1.22) 1.55 (1.46) 1.10 (1.10) 0.58 (044) 9
Ag 6 12 25415xlO-’ 144-41 296 4.09 069 (068) 088 (082) 060 (061) 037 (032)
Ir 6 14 2.4489~lo-’ 334.94 693 3.84 2.28 (2.31) 2-97 (3.74) 1.94 (1.60) 1.32 (1.68) a
Pt 8 10 1.9833 x lo-’ 34,408 5.86 3.92 1.73 (1.80) 1-96(223) 1.61 (1.59) 046 (048)
Au 8 10 1.2793~ 34.408 3.78 4.08 099 (1.03) 1.12 (1.17) 0.92 (0.97) 026 (0-28)
Pb 7 10 5.5765~lo-’ 45.778 2.04 4.95 0.26 (026) 0.31 (0.29) 0.24 (0-24) 0.10 (009)
A1 6 7 3.3147~lo-’ 16.399 3.34 4.05 0.47 (048) 0.51 (067) 0.45 (0.38) 010 (018)
 Long-range Finnis-Sinclair potentials 143

fitted exactly, but by restricting m and n to integer values we are more likely to find the
same values of m and n for different metals. In that case we will be able to take
advantage of the scaling properties of the potential. We further restricted m to be
greater than 6 because otherwise the potentials would have too large a range for
computer simulations. The fitting procedure was as follows. For each value of m greater
than 6, n was set equal to the nearest integer to 18s1‘Bf/c. Provided that n was greater
than m,E was then fitted to the cohesive energy, eqn. (7),and c was obtained from the
equilibrium condition, eqn. (6). The bulk modulus, eqn. (8), was always found to be
within a few per cent of the experimentalvalue. Several potentials with different values
of m and/or n were sometimesfound for each metal, although no potentials were found
with m > 10. The selections appearing in table 1 were made on the grounds that they
Downloaded by [Stanford University Libraries] at 09:02 18 September 2012

gave the closest agreement with the experimental elastic constants cI1,c12and c4&
Better agreement with the elastic constants could be achieved for Ni, Cu, Rh and A1 by
allowing m to be less than 6. For those metals the best agreement was obtained with
(m,n)=(4,13),(5, ll), (4,18) and (5,9) respectively.
We have also considered the relative stabilities of the f.c.c., b.c.c. and h.c.p. crystal
structures. It is well known (Ducastelle and Cyrot-Lackmann 1970) that the relative
stabilities of these crystal structures are determined by higher moments of the local
density of states than the second. Therefore, if our potential predicts that the stable
crystal structure for a known f.c.c. metal is indeed f.c.c., then this may only be
considered as fortuitous. Nevertheless, for the purposes of computer modelling using
these potentials it is important to ascertain the relative stabilities that the potentials
predict. Taking the f.c.c. structure as the reference, the parameter c is given by eqn. (6).
Let rb denote the ratio of the equilibrium b.c.c. lattice parameter to the equilibrium f.c.c.
lattice parameter that is predicted by the potential. Then r b is given by

where S,” denotes the following lattice sum in the b.c.c. crystal

In eqn. (12)ub is the b.c.c. lattice parameter and r j is the length of a b.c.c. lattice vector
from an arbitrarily chosen lattice site. The ratio of the cohesive energy of the b.c.c.
c,
structure,Et, to that of the f.c.c. structure, at their equilibrium lattice parameters, is
given by
m/(Zn-m) S b n/(2n-m)

Et/c=(g) [if] *

For all potentials with 4 <m < 10 and m + 1 <n <20 it is found that the ratio of the
lattice parameters rb, eqn. (1 l),is less than (2/3)’12,indicatingthat the nearest neighbour
spacing is always predicted to be smaller in the b.c.c. structure than in the f.c.c.
structure. It is also found that the cohesive energy of the b.c.c. structure is always
predicted to be less than the f.c.c. structure. Table 2 illustrates these results for the case
of m = 6. If the b.c.c. structure is taken as the reference then the parameter c is given by
eqn. (6), but with the f.c.c. lattice sums s‘, and sfmreplaced by the b.c.c. lattice sums S,”
and SL. Thus the potential for the b.c.c. reference structure is different from that for the
f.c.c. Nevertheless, it is found that the ratio of the equilibrium lattice parameters, rbin
 144 A. P.Sutton and J. Chen

Table 2. Calculated equilibrium ratio, r‘, of the b.c.c. and f.c.c. lattice parameters eqn. (11),
normalised to (2/3)’12.EE/E‘, is the ratio of the equilibrium cohesive energies in the b.c.c.
and f.c.c. structures, given by eqn. (13). Note that the f.c.c. structure is always favoured
energetically and that the nearest-neighbour spacing in b.c.c. is always less than that in
f.c.c.

m n r6 EEIE‘,
6 7 0974 0.998
6 8 0974 0995
6 9 0.975 0993
6 10 0976 0.990
6 11 0-977 0988
Downloaded by [Stanford University Libraries] at 09:02 18 September 2012

6 12 0978 0.985
6 13 0.978 0983
6 14 0979 0-981
6 15 0.980 0978
6 16 0981 0976
6 17 0-982 0.974
6 18 0982 0.972
6 19 0.983 0.970
6 20 0.984 0968

eqn. (ll), and the ratio of the equilibrium cohesive energies, eqn (13), remain
unchanged. We conclude that our potentials always favour the f.c.c. structure over b.c.c.
This is why we have fitted only f.c.c. metals in table 1.
For an h.c.p. crystal it is necessary to minimize the cohesive energy with respect to
the atomic spacing, a”, in the basal plane and the c/a ratio, which we call 1.
Minimization of the cohesive energy, E> of the h.c.p. structure with respect to 1leads to
the condition that

-
rn an n an ’
where Sk = Sk(I) is the h.c.p. lattice sum,

In eqn. (15) the sum is taken over all lattice vectors, from an arbitrarily chosen atomic
site in the h.c.p. structure, and over both atoms associated with each lattice vector. The
ratio, rb of the equilibrium value of a” to the equilibrium value a‘ is given by

Once eqn. (14) is solved for the equilibrium c/a ratio, for a given choice of rn and n, the
ratio of the cohesive energies in the h.c.p. and f.c.c. structures is given by

(17)

We find, for all potentials with 6 < rn < 10 and rn + 1<n <20, that Et/E‘, differs from
unity by less than 0-001, that a”/u‘ differs from 2- ‘1’ by less than 0.001, and that the c/a
 Long-range Finnis-Sinclair potentials 145

ratio deviates from the ideal value, (8/3)”*, by less than 0.001. Not surprisingly,
therefore, the (111) intrinsic stacking fault energies in f.c.c. are practically zero.

$3. DISCUSSION
By using inverse powers for both the repulsive pair potential, eqn. (2), and the
‘squares of the hopping integrals’, in eqn. (3), we have obtained a set of long-ranged
Finnis-Sinclair potentials. Each member of the set is determined essentially by the
indices, n and m, of the inverse powers. The parameters E and a merely define the scales
of energy and length and do not enter into the equilibrium condition for the crystal.
Downloaded by [Stanford University Libraries] at 09:02 18 September 2012

Thus we find that Cu and Ni are both described by the (m,n) =(6,9) potential, while Pt
and Au are both described by the (m,n) =(8,lO) potential. Therefore a simulation
performed with the Cu potential may be translated directly into a simulationfor the Ni
potential simply by rescaling the units of length and energy. The same holds for Pt and
Au. The choice of inverse powers was motivated by the observation in eqn. (4) that the
long-range interaction between two clusters is then described by an inverse-powerpair
potential. For m =6 this gives rise, effectively, to a van der Waals interaction at long
range. The importance of van der Waals interactionsin noble metals has been discussed
recently by Maggs and Ashcroft (1987). We note that the physical origin of this
interaction is quite distinct from the unsaturated covalent bonding that gives rise to the
N-body form of the energy at short range. In our potentials these two interactions have
been combined into one term. We regard this as a convenient mathematicalproperty of
the potential, which should not be interpreted physically as meaning that van der
Waals interactions become covalent bond interactions at short range. Although
potentials with m=6 could be found for all the metals listed in table 1, considerably
better fits to the elastic constants were obtained with higher values of m for some of the
metals. Those better fits are the ones included in the table.
As shown by Finnis and Sinclair (1984), effective pair potentials may be defined
from eqn. (1) by expanding the N-body term about a reference density. Our expansion
in eqn. (4) stems from the same observation.Taking the reference density to be an ideal
f.c.c. crystal, and substituting eqn. (6) for the parameter c, we obtain the following
effective pair potential:

The effective pair potential is therefore a Lennard-Jones m n potential. We believe this

is the reason why the stacking fault energy is practically zero. As noted by Ackland,
Tichy, Vitek and Finnis (1987) it is necessary to make the effective pair potential
positive in the vicinity of the third neighbours in order to get a positive stacking fault
energy. Ackland et al. (1987) achieved this by making suitable adjustments to the pair
potential component of their f.c.c. Finnis-Sinclair potentials. However, we note that,
although for our potentials the stackingfault energy is virtually zero, and the energy of
the h.c.p. phase is virtually the same as the energy of the f.c.c. phase, there is still a very
significant energy barrier to shearing one crystal structure into the other on the close-
packed planes. For many simulations, though not all, this is the only requirement for
meaningful results. In reality, the stacking fault energy is determined by higher
 146 brig-range Finnis-Sinclair potentials

moments of the local density of states than the second, just like the f.c.c.-h.c.p. energy
difference, and we should not be surprised that it is virtually zero in a second-moment
model.
Finally, it is clear from eqns. (1H3)and (17) that the range of our Finnis-Sinclair
potentials is the same as a Lennard-Jones rn - n potential. In a computer simulation the
potential will be truncated at some cut-off radius. The errors introduced by the cut-off
will be of the same order as those for the corresponding Lennard-Jones potential.

ACKNOWLEDGMENTS
Usefuf discussions were held with Dr J. B. Pethica. We are grateful to the Science
Downloaded by [Stanford University Libraries] at 09:02 18 September 2012

and Engineering Research Council and the Royal Society for financial support.

REFERENCES
ACKLAND, G. J., TICHY,G., VITJX, V., and FINNLS, M. W., 1987, Phil. Mag. A, 56, 735.
ALLEN,M. P., and TILDESLEY,D. J., 1987, Computer Simulation ofLiquids (Oxford University
Press).
DAW,M. S., and BAS- M. I., 1984, Phys. Reo. B, 29, 1285.
DUCASTLE, F., and CYROT-LACKMANN, F., 1970, J . Phys. Chern. Solids, 31, 1295.
FINNIS,M. W., and SINCLAIR, J. E., 1984, Phil. Mag.A, 50,45.
HIRTH,J. P., and LOTHE,J., 1982, Theory ofDislocation (New York John Wiley).
KIITEL,C., 1971, Introduction to Solid State Physics, fourth edition (New York John Wiley).
MAWS,A. C., and ASHCROFT,N. W., 1987, Phys. Rev. Lett., 58, 113.
PETHICA, J. B., and SUITON,A. P., 1988, J . oac. Sci. Technol. A, 6, 2490.
ROSATO, V., GUILLOPE,M., and LEGRAND, B., 1989, Phil. Mag. A, 59, 321.
SIMMONS, G., and WANG, H., 1971, Single Crystal Elastic Constants and Calculated Aggregated
Properties (Cambridge, Massachusetts: MIT Press).
SMITH,J. R., BOZZOLO, G., BANERIEA, A., and FERRANTE, J., 1989, Phys. Rev. Lett., 63, 1269.
WALKER, E., ASHKENAZI, J., and DACOROGNA, M., 1981, Phys. Reo. B, 24,2254.