Thermodynamics
of the nickel-t
A tight-binding-bond C. Colinet
and A. Pasture1
Anderscn
In recent ads
on
the
properties
explained cy
the phase
the
knowledge
atomic
the
For
the
numbers white
and data.
Hamiltonians meters possibtc
to
principles canonical
ref.
[I ] for
obtain
them
band structure basis
as
the
which
do
them
arc
the
of par;lit is
fitting
110~
first-
[2] or from
by Harrison
[ 31 OI
not
retical
only
tight-binding
the
Method diagram\
fol-
the
the
compound tat
phase of
system
a~-c
diagram: phase
it
ISC’C-hascd
the solid
the
wc
calculated have
phase
i\ worth
diagram
\,uluch of the energics dcterminc
but
k
verk
solution
the that
sensitilc
to
phases to
Ni’l‘i
cxpc~n~nc~l-
noting
propertic\ \ccins
that
found
boundaries
of
of the three
equilibrium
b\,
kkthe
\vetl \vith
\amc as in the
the
the
cncl-g!
the Cluster
decomposition
not
of
daci-iptioir
[ IO, I 11. The
that
\vxtern\
[O]. The
agr-ee reasonably
pcritectic
both
wc‘rc determined
using
cxpcrinicnt;itl)
Nikl‘i
toI- Thee)-
fr_aniewVrk
free-energv
(c’I3LM)
c,xhibit
also
of
mahc
generation
interest
((‘VM)
phases
Lattice
of
the
OlII
[5- SI.
that
iicw
hut
diagrams
liquid
phase
;I
great
calculations
ctctci-mind
for
c)C
method
f~~CllWd
systems
properfit
cntcring
and
have
;111op. 31.c
within
\.ariation
the
first-ordci
the
NiLTi
have
Pliasc
obtained
ctustcr
part
\\c
candidatc4
studv.
wcrc
10
and
which
cast:.
in ;I Hamiltonian
Hamiltonian.
high-p”rformancc
phenomen:i
last
the
an~ilogoti~
Ni-AI
promising
In
re\utts
studiex.
on
interaction5
details) by
previous
attention
of
[-I].
model
tnut.fin-tiil’c,rl,ital
of solid
tight-binding
calculations
proposed
directly
ix
linear
arc the sole
the choice
either
which
catcutntions.
parameters
more
pure
in
calculations.
If’. at the beginning, over
and
of
avaitabtc
tight-binding
potential
to
the sole
propertics
readily
stumbled
(see
from
first-principles
the
accura
possible
proper-tie5
of alloys
of the constituents
for
Slatcr-Kostcr required
reasonable
.lepwn
cl-band
In
can now bc
become
elemental arc
ol meth-
and
canonical
approximation.
and alloys with
stability
that
literature.
structure
thermodynamic
of
constituents
dcvclopmcnt
density
it has
particular.
examine
data.
local
and computed
In
[ I].
the
electronic
of metals
study
:I
with
years,
and accurate
based
many
sys tern
approach
I. Introduction
efficient
itanium
thl\ the
which
but bc too
the cs-
C. Colinet, A. Pusturei I Thermodynamics
othermic with respect to the other phases. It has been thought that it is due to the Bethe lattice approximation. Therefore, in this study, we decide to keep the Bethe lattice approximation only for the liquid phase and to use the coherentpotential approximation (CPA) [ 121, combined method with the generalized perturbation escribe energy-properties of (GPM) [13] to d BCC- and FCC-based solid solutions. The paper is organized as follows: in section 2 and 3, we present a brief review of the quantumand statistical-mechanical approaches used in our calculations. In section 4, we present the results of the calculations for the Ni-Ti system. A complete equilibrium phase diagram is calculated for the Ni-Ti system and compared to the experimental one.
2. Model 2.1.
The tight-binding-bond
model
The quantum mechanical model is based on a tight-binding (TB) approximation for the Hamiltonian of the alloy which includes both diagonal and off-diagonal disorder. The oneelectron TB Hamiltonian is described using a basis of five d orbitals per atom: H=
C E,~l&L)(&I ‘P
+ C P,,.,.Iir.L)(jvl ‘F.,U
.
- E”J~!,qL”
+ K&,“$q
in which the bandwidth is 254, d is the bond length and s is the Wigner-Seitz radius. The first term C - E,, corresponds to E in eq. (1) and is taken to be the fixed-energy zero and center of the d-band. Therefore, the volume and structure dependence enters only in the variation of the Slater-Koster parameters. Paxton et al. [15] have shown that it is a good approximation. In this study of the Ni-Ti system, we use Andersen and Jepsen’s most localized Slater-Koster and potential parameters for each element at their equilibrium atomic volume [ 141. Two difficulties are arising from alloying effects: (i) the Slater-Koster parameters for mixed pairs are assumed to be given by the geometric mean of the corresponding parameters for the elemental metals [7]. (ii) The shift between the on-site energies of the two metals is determined self-consistently by imposing local charge neutrality, a reasonable approximation for a transition metal alloy, where charge transfers are known to be negligible. Imposing the condition of local charge neutrality, the main contribution from the d-electrons to the energy of formation, AE, is given by (4)
(1)
In this expression, lip) is the ket for the orbital p on site i, the on-site and hopping energies E,, and pIfiL.,,, give respectively the effectively atomic energy of the p orbital on site i and its coupling to the orbital v on site j. It can be shown [4] that the TB Hamiltonian is directly analogous to the first-order LMTO Hamiltonian H ip.,r, = K,
239
of the Ni-Ti system
where the alloy (U~~~‘,‘Y) and (UL,,nd) bond energies are given
pure by
elements
E,.
u ~~~~~= c In
( (E - E,,)n,,(E) -x
dE ,
’ EF
U’h<~“d =c
f (E -
E:h)&(E)dE
T
(2)
where C,, - EYIp and & are standard potential parameters and Sllr,,” is the matrix of LMTO structure constants [4,14]. Matrix elements in the second term, in the usual tight-binding model, take the Slater-Koster form [14]:
where n(E) and n:(E) are the electronic density of states for the alloy and pure elements i, respectively. E,, (Ej:) are respectively the onside energy for species i and orbital (Y in the alloy (pure element).
The assumption that each atom is assu~~uxi to remain charge neutral by varying the on-site hamiltonian matrix in such ;I way that the cncrgh splittings between the different orbitals on the same atom arc prcscrved ensures that the tightbinding bond model is consistent with the force theorem [ 161.
The quantity which is of central interest in OUIcalculations (see eqs. (5)-(h)) is the electronic density of states. According to the state which characterizes the alloy at ;I given composition, i.e. liquid state, solid solution or compound, three different steps are required to calculate the electronic density of states of the alloy. 2.2. I. Stoichiometric~ cornpounds To calculate the electronic density of states of compounds, we have used the recursion method 1171,which constructs ~1 new orthogonal hasi { I
where ([I,,} and {/I,,) are the matrix elements. The coefficients N,, and hi are calculated up to ;I given step tz,, and the continued fraction is then terminated in the usual way: (I,, ,i,, = (I, and h2,I ,,,,= bf these calculations fol [ 171. W,t r epcat AI the nonequivalent sites of the compound studied and the electronic density of states of the compound is given by the sum of these local densities of statcs. _._.-. ’ 3 3 Solid solutions For solid solutions, the configuration-averaged Green function at arbitrary composition is required. The best approach is the zeroth-order mean field approximation. i.e. the coherent
potential approximation average Green function using the identit!
((‘PA). ‘fhe proJected of xi i atom is calculated
(i),(E)
CJ)[
L
7
(;,,,,():
I
1 + t,(;,;,,(b-
tr)]
where C; ’ (-_ fr) i4 the Circcn operator corresponds to the C’PA Hamiltonian. X CT,,1;~) (ipl ‘I’ and
(8)
,I
that
+ X CJ,,,, ,<, /i/l. i Xit11 /,, /I
is the Green function of the pure metal R (tfetincd hy EL = 0 and W,, 01. The potential CT is dctermincd self-conby the condition on the scitttcring
(; '(z)
refercncc ,!3,,,,,,,, )_
sistently operators
whet-c
( IO)
(OhlC
C;,;,,(z) = J c
(z)lOh)
(III
\ ‘l‘o arialyre the ordering effects in solid solutionx, it i\ nectzssa~-1’ to use the gc:neralked pcrturhation method (GPM) [ l.31. the ordering king cxpresscd in terms of pair and energy many-body interactions using ;I perturbation expansion of the random CPA energy in concentration fuctuations. Now ’I
where 7‘,, is the diagonal part of the scattering matrix and 11 is the order of the cluster cxpansion. It has heal shown that for binary transition metal alloys [ 11, the contributions of the triplet and larger cluster interactions to the ordering energy are small compared with that of the pairs. so we shall retain only the first-nearest-neighbor
C. Coliner, A. Pasture1 I Thermodynamics
interactions for FCC-based phases whereas and second-nearest-neighbor interactions been considered for BCC-based phases.
firsthave
2.2.3. Liquid alloys To calculate the density of states of the liquid phase, we have used the scalar version of the alloy Cluster Bethe Lattice Method (CBLM) [ 111 which is a very appropriate method for describing liquid systems. The Green’s function of atom I is given by G,(Z) = c %G,(z).
(13)
G,(Z) is scalar inside the invariant subspaces (Y(I) defined by the types of orbitals chosen (d in our case) [ll]; it is given by the following set of equations:
(144
of the Ni-Ti system
computed self-consistently for each composition and degree of SRO. The bond energy is also calculated as a function of SRO and the effective pair interactions of the liquid alloys are obtained from its dependence [7,19].
3. Phase diagram
calculations
As discussed previously, the internal energies can be considered as the starting point of the phenomenological treatment of alloy phase equilibria by means of statistical-mechanics techniques like the cluster variation method (CVM). In order to compute the phase diagram, we need to know the total free energy of the binary alloy in a given phase, then its energy and also its entropy of formation. For the alloy in its phase (Y, the total free energy may be written as F”=xF*,+(l-x)F”,+E~-TS;
Hou is the hamiltonian without hybridization and 4fi is the part of the mean second moment of the DOS on a state (Y due to the coupling with states p of the neighbors:
(16)
where x is the concentration of the A element, FP the free energy of pure element I in phase (Y, Ep and ST respectively the enthalpy and entropy of alloy formation. The first simplifying assumption regarding the evaluation of this total free energy is to assume that the free energy for the pure elements can be written as F; = E; - TS;
where n, is the degeneracy of the subspace LX and g,,(R) the pair correlation function between atoms of type I and J, provided by the reference system. To describe the liquid state in the metallic alloys, it is also possible to take a Bethe lattice with a coordination number roughly equal to 12, a value which is consistent with the local coordination found in the liquid metallic alloys [18]. In this case, to describe SRO the pair correlation functions are replaced by the pair probabilities. It has been shown that for thermodynamic data, i.e. AL? and hS, there two approaches give similar results [19]. The density of states is
241
(17)
in which the structural cohesive energy EP and entropy SF are both temperature independent. The second assumption assumes that Sy is a purely configurational term, which means that the remaining entropy depends only on the concentration via the first two terms on the right hand side of eq. (16). For the calculations of the electronic density of states and energies of formation, we have seen in the previous section that three families of phases have to be distinguished. In the present approach, we use also different configurational entropy approximations depending on the nature of the phase being considered.
this
In taken
cast.
equal
the
configurational
entropy
Liquid
i4
alloys
shown
been
to zero.
to take their
it into
For
the
entropy
solid
solutions,
is dcscribcd
maximum hedron
cluster
used
containing
the BCC
lattice:
entropy
configurational
of the CVM.
in our
first
study
is the
and second
in this
of a BCC
the
by means
approximation,
disordered
in
the molar
system
is given
of
hard
diamctcr
but
interact
tetra-
neighbors
[Xl
by
account
ill
through
display
SK0
I].
it i4 csscntial
and
the
data.
;I refcrencc
mixture
same
The
I?()- 7
thermodynamic
calculations. ;I
also
ma>
rcccntlh
diffcrcnt
[ 32.231 ix ;I good Garting method.
If WC arc only
dynamic
data.
ximilar
point
in
the
the
disordered
if the
of the
FCC‘
within
the
Bcthe
ticluid
tree but
trantctra-
cnergb with
now
interactions
lattice
for
conligura
by the
CVM
structure
first-nearest-neighbor
potential
in the therms-
is appr~~ximatcd
cncrgy
the
which
and
arc found
tional
approximation
I,!
have
foi- ;I var.iationaJ
intercstcd
allo\;\
fret
all
Coulomb
sition-metal-h~lsc~i hcdron
\uch
charactcrired
charges
results
()I
perform
which
;I hcrccncd
has
dctcrmination To
system
sphcra
24
calculated
a~~l~r~~ximation
[Xl.
( IS)
whcrc
it’,,/,,. l,,i.
rcspcctivcly
1
and
yj;’
the probability
triangles.
second-neighbor
pairs
and
points
their
subscripts
alloy).
R
For
in (i
and
pairs.
the
x,
A
or
In
cienotc
grand
first-neighbor B
given in
;I
structure
potential
WC have [S]
6 2: yi,') In j.j,' ' + 5 x X) In .t-, ) / // cluster
following
probabilities
consistency
arc
related
the
equilibrium
convenient
f1 given
phase
to minimi;lc
the
b!
by
whcrc
p
is
( 19)
by
the
relationships:
the
the present potential
The
to dctcrminc it is more
binary
gas constant. FCC
order
diagram.
tctrahcdra.
configuration
equals
is the perfect
the disordered
y{l’
of finding
cffectivc
work,
chcmicnt
is carried
Out with
independent
configurational
cast
the
tctrahcdron
configurational
variables
of
tetrahedron peraturc taking
7‘ and
the
respect
to 2 set
variables:
in
approximation.
chemical
01 the
these to
bc
w#,~, at constant
effective
account
In
of the grand
a~-e chosen
probabilities
into
potential.
the minimisation
the tcn-
potential
normalization
I-(.
constraint
(7’) iI,,
z 7
u’,,h/
(2Oa)
-
This (1) .‘,,
=
;
M’,,hi
’
minimisation
Iteration
(Xb)
(Nl)
[?a]. The !,A(?I = T
Wiii,
.
lz’,,i,
(Xc)
( Xd
equations
have been presented bc repcatcd The
Y/ = x ill
NI
is
)
done
method
using
the
Natural
developed
by
Kikuchi
used in the present elsewhere
[X]
and
model will
not
hcrc.
equilibrium
phases
I
and
II
scheme
as proposed
phase is
diagram
computed by Kikuchi
bctwcen
using
the
two same
and De Fontaine
C. Co&et,
A. Pasture1 I Thermodynarrks
[26]. For the same initial value of the effective chemical potential p, the grand potential of phases I (0,) and II (n,,) are calculated using the procedure described above. If 0, = fl,,, the equilibrium conditions are realised. but if not, the value of p is modified until 0, = .12,,.
4. Results for the Ni-Ti
system
The Ni-Ti phase diagram displayed in fig. 1 is characterized by a liquid phase, a FCC (A,) phase in the Ni-rich part, a BCC (AI) phase in the Ti-rich part for high temperatures; two stoichiometric compounds NiTi, and Ni,Ti and one intermediate phase with variable range of solubility NiTi [27]. 4.1.
Energies
of formation
of compounds
As a first step, we present the values of the energies of formation of the three intermediate phases observed in the equilibrium phase diagram. We have also calculated energies of formation of metastable phases, which are essentially superstructures of BCC or FCC lattices, to study if our TB Hamiltonian is able to predict the correct ground states. In table 1, we compare our values obtained by coupling TB-LMTO Hamiltonian and the recursion method with values obtained by performing self-consistent-
0.10
0.33
0.50
0.70
0.90
x NI Fig.
1. Experimental
phase
diagram
of the Ni-Ti
system
243
of the Ni-Ti system
Table I Calculated
energies
Structure
of formation
in the Ni-Ti
system.
AH, (kJ/mol) TBB
LMTO
Ref. 1291
NiTi,-EY,
-3s
-35
-29
NiTi-L IO NiTi-B2 NiTi-B32
-36.8 -46 -36.5
-30.6 -40.9 -38.5
-34
Ni,Ti&Ll~ Ni,Ti-DO, Ni,Ti-DO
-45.8 -44.4 -46.4
-49.4 -46.X -49.7
21
-43
LMTO total-energy calculations. This comparison is done to check our alloying model, i.e. TB Hamiltonian (with only d orbitals. Shiba approximation, . . .) and bond energy contribution to the total energy. Neglecting the repulsive part of the total energy is acceptable since all the studied compounds or solid solutions are based on closepacked structures and thus have similar local environment [28]. Retaining the d contribution to the TB-LMTO Hamiltonian is sufficient to explain the strong negative values of the alloying formation energies in the Ni-Ti system. The comparison between both sets of data shows that the different approximations used in our TB-bond (TBB) model are acceptable. More particularly, for x,, = 0.5 we have found that the B2 structure is more stable than the Ll,, or B32 structures. For xNI = 0.75 we have checked that the DO,, hexagonal structure is more stable than the Ll, or DO, cubic structures. For both compositions, self-consistent LMTO total-energy calculations and TBB energy calculations point out strong competitions between the different BCC or FCC superstructures. More particularly, for xN, = 0.75 it seems to be possible to obtain the metastable Ll, phase from the Al phase if rapid cooling is used for instance. Both theoretical approaches give values of formation energies which are roughly 20% larger than the experimental ones [29]. It is in fact necessary to perform full-potential calculations to obtain a better agreement [30,31].
The previous attempt to calculate the phase diagram of the Ni-Ti system is based on the CBLM approximation for BCC- and FCC-based solutions. In fig. 7 arc compared the conccntration dependence of the calculated energy of mixing E,~,,,,,(x) for th e completely random KY’ and BCC solid solutions using both the TBCBLM and TB-CPA approaches. We can set that the agreement between the two sets of calculations is better for the FCC-based structure than for the BCC-based structure. This results from the fact that the BCC structure is more sensitive to the Bethc lattice approximation. However, this difference will be sufficient to obtain the phase boundaries of the peritectic decomposition of the NiTi, compound in agrccment with the experimental phase diagram. as shown in the next section.
It is also essential to have a good description of the thermodynamic data of the liquid phase since a great part of the phase diagram is from equilibrium between the liquid
and solid phases. The scalar \,ersion of the alloy (‘BLM coupled with the TU-LMTO Hamiltonian gives calculated values of mixing cnergicx which are displayed in tig. 3. The agreement with the experimental data [.12-331 is very good. capccinlly for the Ni-rich part. Let us mention that the evolution as a function of the composition is \‘er!’ similar to those of B<‘(‘- and FCC-based solid soiutions (see fig. 2).
‘1‘0 obtain the phase diagram given by the competition between different structures (HC’P. BC‘C. FCC or liquid) it is essential to know the thermodynamic propertics of pure Ni and Ti OI-. in other words. the difference between the free energies of the liquid phase and the different crystalline phases. Two ways can be usetl to obtain such data. The first one i\ to use thermodynamic compilations [ 3.51. The second one airsumes that the structural energies between FCC‘. BCC and HCP structures for Ni and Ti elements arc fixed to the values obtained from the selfconsistent LMTO total-energy calculations. We obtain A,!?,,,., , (( = 2.9 (0.7) kJ/atom and for the Ni (Ti) AE I!< I’ , <‘C= 3.3 (-3.5) kJ/atom parameters. i.c. clement. The remaining A(;, ,.( ,.(Ni) and AC;, ,(( ( (Ti), are obtained from the experimental thermochemical data. melting temperatures and latent heats of melting. while AS H( ( ,,( ,,(Ti) is obtained from the experimental HCP-B(‘<’ transformation tcmperature. In fig. 4. we compare both approaches for the Ti elcmcnt. of the Ni-l‘i \ystcm. The phase diagram
C. Colinet.
A. Pasture1
I Thermodynamics
+Ref35 a LMTO fee-hcp
600
2000
1200 T(I<)
Fig. 4. Gibbs
energy
of phases
of the Ti element.
calculated by combining both FCC- and BCCbased free-energy curves with the liquid freeenergy and the free-energies of the three compounds is shown in fig. 5. Although the cubic B2 phase is experimentally reported to be stable over a small concentration range, it has been described as a stoichiometric compound. At high temperatures, the Ni-Ti phase diagram can be viewed as resulting from a BCC-FCC competition; FCC dominates on the Ni side with the FCC solid solution and the metastable Ll, (Ni,Ti) FCC superstructure. On the Ti side, BCC dominates with a BCC Ti terminal phase above 710 K. The solid A2 solution displays a range in good agreement with the experimental one, contrary to the previously reported calculations based on the CBLM [S]. The eutectic I ”
20001
0
Fig. 5. The calculated
1”
0.25
Ni-Ti
I,
0.50 XNi phase
”
r
0.75
diagram
“,
,
1
of the Ni-Ti
system
245
equilibrium is found now between the NiTi, compound and the A2 solid solution, the eutectic temperature being 60 K higher than the experimental one. The congruently melting temperatures of Ni,Ti and NiTi compounds are correctly predicted. The stability of the B2 phase can be understood using simple arguments similar to those used to explain the BCC-FCC competition for pure transition metals. Mixing Ni and Ti corresponds to an average number of d electrons more favourable for the BCC structure than for the FCC structure. Correspondingly, the experimental phase diagram does not exhibit a B2 ground state, although the B2 phase appears above about 900 K [27]. Experimental observations are not conclusive on which path leads to the most stable state. It is either a martensitic transformation to a monoclinic phase not studied here [36], or a decomposition at about 900 K into Ni,Ti-DO,, and NiTi,-E9, [37]. Such discussion is beyond the scope of our TBB analysis but preliminary results based on the self-consistent LMTO calculations indicate that the decomposition between DO,, and E9, structures is the most likely way [38].
5. Conclusion The tight-binding calculation of the Ni-Ti system coupled with CPM-CPA for solid solutions and scalar-CBLM for liquid alloys give results in good agreement with experimental informations. (i) The large interactions between the d states of Ni and the d states of Ti are the main factors governing the highly negative formation energies of the different compounds occurring in this system. We have found for xNI = 0.5 and 0.75 that the compounds crystallize in the B2 and DOzJ structures respectively. (ii) The formation energy of the A2 terminal solution on the Ti-rich side obtained by CPA calculations displays less exothermic values than those given by alloy-CBLM calculations. This leads to a eutectic equilibrium between the NiTi, compound and the A2 solution which is now in agreement with experimental data.
1IS/
References
\I
C\‘;Lwd:t.
Cd\.
b’ Ducastelle.
in: Order
and Phaac Stability
in Allo!\.
Structure.
cd\.
and
and
Pettifor.
3 (North-Holland. in: Papaconstantopoulo~.
D.
Vol.
Structure York. W.A.
ol
Elcmcntal
O.K.
(Freeman.
( 19X3)1571. <‘. (‘olinct. (‘ondcns.
A.
(lYY7)
Condens.
D.H.
I.e.
Phy.:
Condcns.
He\wud
and
I (IYXY)
Ph!\. A
Kc\.
Pa\turel.
Lctt. J.
171I
New
and the Propcrtio ii
I%\\:
Phys.
Kc\.
13 45
I’. 3
Rev.
and
and
A.
Paaturcl.
J
3 ( IYY I) 78’)s.
Matter
Phys.
X1
Hider
and
A.
I%wur.cl.
0.
( IYYI ) YYh5. ( 1951 ) YYX
L.M.
Nguyen
Phy\.
R. Vclicky.
Falicov.
Mnnh.
Rev.
b’. Ducastclle
1’71
0.
cds.
Paxton,
I’hb\.
Rev.
D.G.
Pettlfor.
Haydock.
Ehrenreich, York.
Pa\turcl
Jepsen
F.
Baaaanl.
M.
Methfcsel
B 41 (15’90)
of F.
Ph!\.
F 0 (lY7h)
and
in.
I’hyics.
Fun11 and
IY85)
in: Solid
State
Vol.
703’1
liil
Enrw (‘ourw
MI’.
‘I o\i
Pol;itoglou.
State
IYXO) p. 315.
Physics.
Vol.
I)uII~>I\
.~ntl :\
!
I’;~\lutc~l.
M
in: (‘cinstituti~~n
Ncu (’
I’or-h. IY (IYHh)
Notin
and
ot Bln.tr\
19X5).
.I.
2X5. Hcrty.
I hcc-mochlm
Ph)\.
Kc\.
Ii iS (IYXS)
A. Pastmel.
AT.
1537
Paxton
antI hl. \a11
ior publication in Ph!\. Kc\. 1%. l.i~chand I. Arpshofcn. Therm. Acta l-71 (lYS8) 171 suhmittcd
German
lihl
bin. M.S.
and (;.R
Saint
40, ccl\.
H
Pro\,
[iSI
35. 4s.
and F. Seitz (Academic
H
Prew.
M.Ci
(IYXI)
4’1.
A.T.
Dindalc.
Physical
Vall\he\.
l’etrushcv\ku. SGTL
Lahoratol-y
W. Fuehrer.
K
J. Phys.
PieI-re.
Metall
Iran\.
J.
Phc\.
;
OL
Cscl‘d
(‘hem.
5% ii
data tor pure clement~,
(btthardt.
A
E- I.3 (IYX.7)
Kulik.
Alloy
to hc puhli\hcti
0.
Mcrclcr
and b
1.77. Phahc Diagram\.
Internationa.
IYYO: 2nd ccl.) 13. 2x71. A Pa\turcl. C’olinet.
Nattonal
( 1988).
Keport
in: Blnaq (ASM
A 1.. F.rmak~>\. Ku\\.
(‘.
Schiltgnardc. Solid
I’ll\\
p. %7.
Manh.
1171 J .L. Murray.
141x7). in:
I’ (‘hlcu\.
( IYYI ) 301
3
_I. I’h!\.
Methfesscl. Nguyx
Stauh.
and F. Seitz (Academic
D. Turnhull
iinil
( 1OSl) 155
4x
Massaldd Phyxio.
J M.
and K. Antlcrh~~.
Pettltor.
1.311Yu.0.
X137
I>. Turnhull
Pa\turcl
Matter
(lY75)
Ciachon.
li’j Ii. 13.71R.M.
p. 5’). HM.
II!
( 1~~77) 2XIO.
Ehrcnreich.
and D. <;lotrcl.
School
P. (Ilieu\.
Schilfpaardc.
3381.
H.
illI
icwlt\. A
(Mc~ira~+Hill,
;\cta
t3 3S
;md F. <‘?I-ol-
(
(In\tllutL,
(1WO) l5US.
C‘cmdcna.
I’S] D.
1311 1). Kev
PIl)\
1076) 1,. 2.30
unpuhlishccl
Hancen
l301 M.
Phys.
.I. Php
Amsterdam.
A.‘[.
York.
and
and F. Gauticr.
(North-Holland.
thrcnrcich.
M
and
International
L.XXXIX.
Marct.
(Inst.
(ir-ccnboocl
12’1,A t’a\turel. J. Hatncr and I’. Hictcr. l’hp. Kc\ 15 i2 ( IYSS) 50(1Y I’.?\A Pa\turel and J, Hatncr. Phv\. Kcv 13 34 ( IUM) S.i511-11R. Kikuchi. J (‘hem. Ph>\. h(l (IY7-I) 11171 1751(‘ Sl$i nnd J.M Sanchez. :\cla Metall. .;? ( I’M 1 Illc~1701K. Kihuchi and I> dc Fontalnc. Nat Ifur. St,~nd Kcp
737.
Andersen.
berms
A.
L3 33 (IYXh)
S. Kirkpatrick
( IYhX)
I75
M
I).;\
I‘. Pommc.
I3 47
I%)\.:
J
1.331.
Lackmann.
Marct.
NOSP-4YO
Pasturel.
P. Hider
C‘. Colinet.
Mayou.
l’a\turel.
Kc\.
5837.
and A.
Matter
Rohhins
( IY8-l)
New
Pres\.
IYXO) pp. 476500.
Jepsen.
C’. C‘olinet.
Kikuchi,
117) R.
1901 ). the Band
dnd
Bristol.
Alloy\
Lc.
New
I I’)1 I\. (701\l.
I).(;.
1571.
1I.H.
O.K.
0.
P. Hider
Phy\.:
Kc\.
Structure
San Franciw),
anti
Matter
C‘. Colinet.
I).
Amsterdam. Handbook of
S~~htls (Plenum
Electronic
Andersen
M.D.
de Boer
lYX6). Harrison.
01 Solids
K.
F.R.
b+let;tl\
l.lqllld
r-.\am
I’h!\ics. C’ohesion
111:
Ii.
A
Materi:ll\ 1.
Paxton
ccl Park.
.ind
Iv.
I II OH. \;,I1