review h-r paper revised
TRANSCRIPT
-
8/12/2019 Review H-R Paper Revised
1/47
The prediction of solid solubility of alloys: developments and applications of
Hume-Rotherys rules
Y.M. Zhang
!. R.". #vans$
%. Yanga
1. Centre for Materials Research and School of Engineering and Materials Science
Queen Mary, University of London
Mile End Road
London E1 4NS
. !e"art#ent of Che#istry,
University College London,
$ %ordon Street,
London &C1' $(), U*
a(uthor to +ho# corres"ondence should e addressed. E#ail- S.yang/#ul.ac.u0,
1
mailto:[email protected]:[email protected] -
8/12/2019 Review H-R Paper Revised
2/47
&bstract:
n the 12$s, 'u#e3Rothery hel"ed to #a0e the art of #etallurgy into a science y
the discovery of rules for the "rediction of soluility in alloys. heir si#"licity and
generality #ade the# eco#e one of the #ost i#"ortant rules in #aterials science.n
the fe+ decades after 'u#e3Rothery5s discovery, #any researchers have tried to
#a0e 6corrections7 to '3R rules ai#ing to #a0e the# +or0 etter in general alloy
syste#s. hose researches included e8"lanations of the rules using /uantu# and
electron theories and ne+ co#inations of the factors to give etter #a""ing. n this
"a"er, +e revie+ #ost of these contriutions and introduce recent "rogress in
soluility "rediction using artificial neural net+or0s.
*ey+ords- soluility, alloys, 'u#e3Rothery rules, electronegativity, artificial neural
net+or0.
-
8/12/2019 Review H-R Paper Revised
3/47
.' (ntroduction
he #ost +idely acce"ted vie+ of scientific #ethod is ased on the creative
e#ergence of hy"otheses or con9ectures :1; +hich gradually eco#e +ell3trenched in
the for# of estalished theories as #ore su""orting e8"eri#ental evidence is sought
and found. ( contrasting #odel for scientific discovery is attriuted to 1?@131@@A in +hich large a#ounts of data are first collected, asse#led into tales,
surveyed and fro# +hich theories are devised :;. ( central deate in the history and
"hiloso"hy of science focuses on these contrasting e8"lanations of scientific #ethod
and is concisely articulated y %illies in his analysis of artificial intelligence :B;.
'u#e3Rothery5s rules, occu"ying a central s"ace at the heart of #etallurgy, are in the
=aconian tradition. n the 12$5s, after surveying the availale soluility data, 'u#e3
Rothery distinguished the factors that influence co#"ound for#ation and control
alloying ehaviour.here e8ists a connection et+een soluility, ato#ic sie, crystal
structure and a "articular concentration of valence electrons in an alloy :4, ?;. 'u#e3
Rothery added other ideas, develo"ing conce"ts +hich are no+ 0no+n collectively as
'u#e3Rothery5s rules :@3D;.
-
8/12/2019 Review H-R Paper Revised
4/47
Miutani :1G;, Massals0i :4, 1D; and as descried in iogra"hical s0etches aout
'u#e3Rothery :12;. Researches on '3R rules losso#ed in the 12B$s312D$s ut the
"rediction of soluility +as gradually su"erseded y calculation of "hase diagra#s
>C(LH'(!A.'o+ever, the si#"licity and generality of '3R rules still #a0e the#
one of the #ost i#"ortant cornerstones in #aterials science. &atson and &einert in
$$$ :$; #entioned #ost of these rules are still useful today as in 'u#e3Rothery5s
ti#e, and discussed the a""lications to the transition and nole alloys, oth +ith each
other and +ith #ain grou" ele#ents. HarthI :1; discussed a""lication of D3N and
valence electron rules for the intl "hases >+hich aredefined as 6se#i#etallic or even
#etallic co#"ounds +here the underlying ionic3covalent onds "lay such an
i#"ortant role that che#ically ased valence rules can e used to account for
stoichio#etry and oserved structural features7A and their e8tension. Recently these
rules have also een used in nanocrystal gro+th and co#"ound for#ing tendency :;
as +ell as ther#odyna#ically staility of ordered structures in FJ se#iconductor
alloys :B;. n this revie+, the authors ai# to address so#e recent "rogress on the
develo"#ent and a""lication of '3R rules.
(s cautioned y Hettifor :12;, ecause different rules +ere e8"ressed or stressed y
'u#e3Rothery at different ti#es, it is so#eti#es difficult to define +hat constitutes
K'u#e3Rothery Rules5 and this confusion is e8tant. here is general agree#ent that,
in order of i#"ortance, the ato#ic sie factor is first, follo+ed y the electro3
negativity effect. he i#"ortance of the electron concentration >eFa ratioA in
deter#ining solid soluility oundaries is recognised in so#e cases ut other factors
are rarely discussed in sufficient detail. Surveying #etallurgical and "hysical science
4
-
8/12/2019 Review H-R Paper Revised
5/47
"ulications in general, different sources e8"ress 'u#e3Rothery5s ideas using ter#s
such as- effects, "rinci"les, factors or "ara#eters :4;.
$.' #arly )ormation and Revision of Hume-Rotherys Rules
n 12?, after ta0ing his Hh! under Sir 'arold Car"enter at the Royal School of
Mines, 'u#e3Rothery returned to 8ford and +or0ed on inter#etallic co#"ounds,
#etallogra"hy and che#istry, e8tending his ideas on the for#ation of co#"ounds.
'e e8a#ined "hase diagra#s of the nole and related #etals >i.e. Cu, (g and (uA,
es"ecially those alloyed +ith the =3sugrou" ele#ents >these include Li, =e, =, C, N,
,
-
8/12/2019 Review H-R Paper Revised
6/47
n the 12B$s, 'u#e3Rothery shifted his attention to characterisation of ato#ic sie y
nearest neighour distance instead of volu#e :@;. +o of the 'u#e3Rothery rules
controlling solid soluility +ere discovered- 1A the first 'u#e3Rothery rule, the
ato#ic sie factor, said that if the ato#ic dia#eters of the solvent and solute differ y
#ore than aout 1431?O then the "ri#ary solid soluility +ill e very restricted A
the second rule e#"hasised the i#"ortance of the electron concentration >or electron
"er ato# ratioA in deter#ining the "hase oundary. =oth of these rules are "resented
in his classical "a"er in 12B4 :@;.
(lthough 'u#e3Rothery at that ti#e had found t+o i#"ortant guidelines +hich
decided the for#ation of "ri#ary solid solutions, he +as unclear ho+ to classify
inter#etallic co#"ounds. n 12BG, after studied the silver rich anti#ony3silver alloy
syste# +ith Reynolds :G;, 'u#e3Rothery eca#e a+are of a third factor restricting
solid soluility, that is, electroche#ical factor #a8i#u# solid soluility reduced as
the electronegativity difference et+een solute and solvent increased ecause of the
co#"etition to for# inter#etallic co#"ounds.
he relative valence rule +as #entioned in the 12B4 "a"er, and the i#"ortance of this
rule +as stressed y 'u#e3Rothery in early editions of his fa#ous oo0 The
Structure of Metals and Alloys:G;. 'o+ever, in his later versions of this oo0, it is
stated K#ore detailed e8a#ination has not confir#ed this and, in its general for#, the
su""osed "rinci"le #ust no+ e discarded5 :D, 2;.
*.' )urther development and application of Hume-Rotherys Rules
@
-
8/12/2019 Review H-R Paper Revised
7/47
he develo"#ent of 'u#e3Rothery5s Rules can e classified into t+o categories- he
first is the develo"#ent +ithin each rule, and the second is the develo"#ent of
ana#or"hoses or alternatives of these rules as a +hole in order to get #ore "o+erful
and "recise "redictions. n the first category, researchers, 1A "rovided e8"lanations of
s"ecific rule>sA fro# ele#entary electron theory, A "ointed out the +ea0ness and
deficiency of individual rules. n the second category, researchers have atte#"ted to
e8tend the rule>sA or itsFtheir alternatives to +ider a""lications. n +hat follo+s, the
discussion "rogresses along these t+o "aths.
*. +evelopment and application of each rule
*.. &tomic %i,e )actor
he ato#ic sie factor rule is usually "resented in the follo+ing +ay :@;- 6if the
ato#ic dia#eters of the solute and solvent differ y #ore than 14O, the soluility is
li0ely to e restricted ecause the lattice distortion is too great for sustitutional
soluility.7 &hen the sie factor is unfavourale, the "ri#ary solid soluility +ill e
restricted +hen the sie factor is favourale, other factors li#it the e8tent of solid
soluility and it is of secondary i#"ortance. &aer et al.:B$; a""lied the sie factor
alone to 14B ter#inal solid solutions and +ithin 2$.B1O >??2F@12A of the syste#s
+here lo+ solid soluility +as "redicted, lo+ solid soluility +as indeed oserved.
n the other hand, it +as less easy to "redict extensivesolid soluility +hen there +as
a s#all sie difference it achieved only a ?$ O >4$BFD$4A success rate. he sie factor
rule has een e8"lained y using ele#entary electron theory. t can e sho+n that :B13
B4;, if a #isfitting solute ato# is regarded as an elastic s"here +hich is then
co#"ressed or e8"anded into a hole of the +rong sie in the solute lattice, +hich can
G
-
8/12/2019 Review H-R Paper Revised
8/47
e treated as an isotro"ic elastic continuu#, the ensuing total strain energy E in oth
the #atri8 and solute can e esti#ated as
B
$D rE= , >E/uation 1A
+here P is the shear #odulus and r$and >1Ar$are the unstrained radii of the solvent
and solute ato#s. a0ing as $.14 >as the sie factor rule declaresA or $.1?, andB
$r
$.G eJ, this gives TkE B4 at 1$$$ *. !ar0en and %urry :1$; "roved that at
te#"erature , the "ri#ary solid soluility +ould e restricted to elo+ aout 1 at.O
+hen the energy of solution e8ceeds 40= "er ato#, +here 0= is =olt#ann5s
constant. (lthough elastic theory cannot e a""lied strictly at the ato#ic level, this
gives a si#"le e8"lanation of 'u#e3Rothery5s sie factor rule.
Mott :BB; "rovided a /uantu# #echanical asis for elastic theory ased the &igner3
Seit +ave function T$>rA, for an electron in the lo+est state in a &igner3Seit cell.
he +avefunction for an electron in the alloy can e e8"ressed a""ro8i#ately as T>rA
u >rA e8" >ik rA, +ith u >rA having the for#s of an ( ato# or a = ato# &igner3Seit
+avefunction T$>rA inside the &igner3Seit cells of ( or = ato#s >if ( and = are of
the sa#e valencyA in the alloy, +here ris the vectorial "osition of the electron and kis
the +ave3vector.
-
8/12/2019 Review H-R Paper Revised
9/47
he validity of the sie factor has een deated since the rule +as "ro"osed. 'u#e3
Rothery et al. the#selves "ointed out :@; that the e8act 6ato#ic dia#eter7 of an
ele#ent is al+ays difficult to define. hey defined the ato#ic dia#eter as given y
the nearest3neighour distance in a crystal of the "ure #etal. 'o+ever, this dia#eter
cannot necessarily e transferred to the alloy ecause 1A the Kradius5 of an ato# is
"roaly affected y coordination nu#er. E8ce"t for the heavy ele#ents, ele#ents
of the = su3grou"s tend to crystallie +ith coordination nu#er D3N, +here N is the
grou" to +hich the ele#ent elongs. his is due to the "artly covalent nature of the
forces in these crystals and, e8ce"t in %rou" J =, results in the ato#s having t+o
sets of neighours at different distances in the crystal. A n so#e structures there are
great variations in the closest distance et+een "airs of ato#s at their closest distance
of a""roach, de"ending on the "osition and direction in the lattice. BA n for#ing a
solid solution, the Ksies5 of individual ato#s #ay change according to the nature and
degree of local dis"lace#ents. n the case of anisotro"ic or co#"le8 structures or
+here the coordination nu#ers are lo+, the closest distance of a""roach does not
ade/uately e8"ress the sie of the ato# +hen in solid solution :1D;.
-
8/12/2019 Review H-R Paper Revised
10/47
he ato#ic di#ensions can e calculated y using "seudo"otential theory, such as the
+or0 done y 'ayes et al. :B@; on Li3Mg, nglesfield :BG3B2; on 'g, Cd and Mg
alloys, 'ayes and Woung :4$; on al0ali alloys, Stroud and (shcroft :41; for Cu3(l,
Li3Mg and Cu3n, Meyer et al.:4344; on analying the diffusion ther#o"o+ers of
dilute al0ali #etal alloys, on calculating the lattice s"acings and co#"ressiilities of
non3transition ele#ent solids and for analying residual resistivities in silver and gold
and Singh and Woung :4?; on heats of solution at infinite dilution. hey can also e
otained fro# the free3electron #odel develo"ed y =roo0s :4@; and have een used
y Magnaterra and Meetti :4G, 4D;.
he actual individual ato#ic sies can also e esti#ated fro# static distortions in a
solid solution y #odulation in diffuse X3ray scattering :42, ?$; or fro# +ea0ening of
the interference #a8i#a analogous to ther#al effects :?13?4;.
-
8/12/2019 Review H-R Paper Revised
11/47
investigation has not confir#ed the generality of this rule. (n e8a#"le is that
#onovalent silver can dissolve aout $O alu#iniu# ut trivalent alu#iniu#
dissolves aout 4O silver. 'o+ever, for high valence, covalently onded
co#"onents, the relative valence factor a""lies +ell. E/uation A
11
-
8/12/2019 Review H-R Paper Revised
12/47
+here I is the ioniation energy, A is electron affinity, and Y is Mulli0an
electronegativity. !ividing y .D, gives a""ro8i#ately Hauling5s e#"irical scale.
&atson and =ennett :@$; "oint out that in the case of transition #etals, the "artly
filled d states at energies near the
-
8/12/2019 Review H-R Paper Revised
13/47
n addition to these efforts to e8"lain each rule using /uantu# or electronic theory,
another category of develo"#ent has een to #a" the soluleFinsolule syste#s in a
t+o di#ensional diagra#.
*.$ Mapping and derivatives of Hume-Rotherys Rules
(lthough e8tensive researches have een done to +or0 out the theory ehind the '3R
rules, it +ould e very useful if the soluility of the #aterials can e #a""ed
diagra##atically on a Cartesian syste#. So researchers can si#"ly calculate the
coordinate of the ele#ent to "redict the soluility using such a diagra#. his
direction started fro# !ar0en and %urry :1$;, follo+ed y %schneidner :?G;,
Cheli0o+s0y :11;, (lonso et al.:1, 1B;, and hang :14, 1?; a#ong others.
*.$. +aren-"urry method
n 12?B, !ar0en and %urry :1$; "ro"osed a diagra##atic #ethod to descrie the
solid soluility of fifty alloy syste#s >!% #ethodA. hey used the sie factor as
ascissa and electroche#ical factor >electronegativityA as ordinate to "lot soluilities
of each alloy syste# and then dra+ an elli"se >rsolventZ 1?O as #inor a8is, YsolventZ $.4
as "rolate a8is, +here rsolvent and Ysolvent are radius and electronegativity of solvent
res"ectivelyA to se"arate the solule ele#ents fro# the insolule ele#ents. he result
is sho+n in
-
8/12/2019 Review H-R Paper Revised
14/47
unsuccessful classification in silver syste#s +as, at that ti#e, attriuted to the use of
una""roved electronegativity values of silver ut %schneidner :?G; argued that even if
a""roved electronegativity values for silver +ere used, the classification +ould still e
"oor.
&aer et al.:B$; e8a#ined the universality of the 'u#e3Rothery sie rule and the
!% #ethod for "redicting solid soluilities. (fter analyed 14?? inary alloy
syste#s, they confir#ed 'u#e3Rothery5s sie factor and sho+ed that the
electronegativity is an i#"ortant consideration in the for#ation of solid solutions.
>insert figure 1 hereA
n 12D$, %schneidner :?G; again a""lied the !% #ethod to create a soluleFinsolule
classification y introducing the effect of crystal structure >electronic3crystal structure
!ar0en3%urry #ethod, ECS!%A.
-
8/12/2019 Review H-R Paper Revised
15/47
e8"ected 4A. n cases +here the solvent is ansp3ele#ent and solute is a d3ele#ent, no
solid solutions are "redicted to for#, regardless of either the crystal structure or sie
factors. (""lying the ECS!% rules to ten solvents >Mg, (l, called electronic and crystal structure, sie #odel, ECSA, and
"ointed out that their #ethod should also e /uite good at "redicting the e8tent of
another ele#ent in a inary co#"ound. n co#"ounds, iA the co#"atiility of the
crystal structure of the solute +ith either or oth of the co#"onents of the
inter#etallic "hase is regarded as a critical factor iiA the valence of the solute
co#"ared +ith the co#"onents is also decisive iiiA if the aove t+o criteria are
favourale, then ato#ic sie factor +ould e the final decisive issue and ecause of
the less elastic nature of co#"ounds co#"ared +ith ele#ental #etals, a Z1$O
li#itation on ato#ic sie should e a""lied.
*.$.$ /helio0sys method
n the 12G$s, t+o events occurred that led :11; to the vie+ that so#e of the arriers to
understanding solid soluility in inter#etallic alloys could e re#oved. ne +as that
Miede#a and collaorators "redicted and classified heats of for#ation for regular
inter#etallic alloys +hich is "redo#inantly deter#ined y the electronegativity
difference > \ A and the difference in electron density at the oundary of the
&igner3Seit cell > WSn A of "ure #etals :@43@D;. he other +as that *auf#ann and
co3+or0ers develo"ed ion3i#"lantation techni/ues and conducted ion3i#"lantation to
1?
-
8/12/2019 Review H-R Paper Revised
16/47
"rovide a +ide range of ne+ and uni/ue #etastale alloy syste#s, unotainale y
conventional #etallurgical "rocedures :@23G1; +hich led to suse/uent efforts to
study solid soluility in alloys :G3G?;.
n 12G2, Cheli0o+s0y introduced a gra"hical "rocedure si#ilar to the !ar0en3%urry
#ethod to analyse solid soluility in the case of divalent hosts. n his +or0, a different
"air of coordinates +ere introduced- the electron density at the oundary of ul0
ato#ic cells, WSn , and the electronegativity\ . (s #entioned efore, these t+o
coordinates are the funda#ental "ara#eters in a successful se#i3e#"irical theory of
heats of alloy for#ation develo"ed y Miede#a and co3+or0ers :@43@D;. n his ne+
0ind of "lot, Cheli0o+s0y +as ale to give #ore reliale "redictions. (n e8a#"le of
Cheli0o+s0y "lots and a co#"arison +ith !ar0en3%urry "lots are sho+n in insert figure hereA
-
8/12/2019 Review H-R Paper Revised
17/47
Hauling5s scale, +hile in Cheli0o+s0y "lots the Miede#a scale is used. =oth scales
actually have a good correlation as sho+n y Miede#a et al. :@4;. he other co3
ordinate is different, ato#ic sie is used in !ar0en3%urry "lots and electron cell3
oundary density is used in Cheli0o+s0y5s #ethod. (lthough higher accuracy has
een estalished, so#e e8ce"tions re#ain. hese e8ce"tions suggest that
Cheli0o+s0y5s #ethod is still susce"tile to so#e i#"rove#ent :1;.
*.$.* &lonsos Method
n the 12D$s, fro# analysis of oth !ar0en3%urry and Cheli0o+s0y #ethods, (lonso
and Si#oar :1; "ro"osed a sche#e containing all three coordinates >ato#ic sie,
electronegativity, electron cell3oundary densityA. he suggestion +as also "ro"osed
y Miede#a and !e Chatel :2B;. =y incor"orating a sie factor in a ne+ gra"hical
#ethod, they i#"roved on the original Miede#a coordinate sche#e "ro"osed y
Cheli0o+s0y. n their analysis, each che#ical ele#ent is characteried y three
"ara#eters- the ato#ic volu#e V , the electronegativity \ , and the electron
density at the oundary of ul0 ato#ic cells bn . (s the result,\ and sn >the
difference of electronegativity and electron density at the oundary of ul0 ato#ic
cellsA co#ined into a ne+ "ara#eter, !" , the heat of for#ation of an e/ui3ato#ic
co#"ound, calculated using a se#i3e#"irical for#ula "ro"osed y Miede#a,
( ) ( ) #n$%" s! +=
BF1
\ , >E/uation BA
+here H and Q are universal constants and R is another constant +hich deviates fro#
ero only +hen one of the #etals is "olyvalent +ithpelectrons :@?, @D, 24;. hen, the
t+o "ara#eters !" and V >e8"ressed as &igner3Seit radius, R&A are used to
construct a t+o3di#ensional #a". n this #a", the che#ical ele#ents insolule in a
given host are neatly se"arated fro# the solule ones y a straight line. he e8a#"les
1G
-
8/12/2019 Review H-R Paper Revised
18/47
and the co#"arisons +ith Cheli0o+s0y5s "lot are sho+n in aA and B>A. n
these "a"er, the relative i#"ortance of the three coordinates >ato#ic sie,
electronegativity and electron cell3oundary densityA is de#onstrated also they
e8"lained the success of the sche#es of !ar0en and %urry and of Cheli0o+s0y y the
fact that ato#ic sie and electron cell3oundary density are strongly correlated in
given class of #etals. Later, (lonso et al.a""lied this #ethod to the "rediction of
solid soluility in nole #etal, transition #etal and sp#etal ased alloys :1B, 2?;.
)ones has a""lied this #ethod to the solid soluility of t+o light #etals, #agnesiu#
and alu#iniu# :2@;.
(fter the heat of for#ation calculations y Miede#a, others succeeded in "redicting
the heats of for#ation of different inary alloys fro# oth first3"rinci"les and se#i3
e#"irical #ethods :2G31$D;. !uring the last t+enty years and ased on calculated
heats of for#ation of alloys and se#i3e#"irical theories +ith "ara#eters such as
electronegativity difference, ato#ic dia#eter and nu#er of covalent onds, a large
nu#er of "redicted #a8i#u# solid soluilities of alloys, design of alloy syste#s or
for#ation of co#"ounds have a""eared :1$2312;. his indicates these #ethods still
have vitality in the "rediction of alloy for#ation even though near thirty years has
"assed.
>insert figure B hereA
*.$.1 Zhang 23 Method
hang and co3+or0ers used gra"hical #ethods +ith various "ara#etersFcoordinates to
"redict the for#ation of a#or"hous alloys and solid solutions. Several factors affect
the for#ation of a#or"hous alloys and solid solutions, so#e acting against each other
>e/uation GA. :1?, 1B$;.
1D
-
8/12/2019 Review H-R Paper Revised
19/47
n 12DB :1B1;, they a""lied Miede#a5s coordinates to the "rediction of inary
a#or"hous alloy for#ation and found that this #ethod +or0ed /uite +ell. More
s"ecifically, they found >1A the ranges of for#ale and non3for#ale inary
a#or"hous alloys can e, to a good e8tent, se"arated y a straight line, +hich is
$11
B21$BF1
\ =
sn , >E/uation 4A
+here \ and sn are the differences of electronegativity and electron density
res"ectivly at the oundary of ul0 ato#ic cells as #entioned efore >A of the D
a#or"hous alloys for#ed y #elt3/uenching, the "rediction accuracy +as 2.GO >BA
of the ?4 non3for#ale a#or"hous alloys, the "rediction accuracy +as @@.GO >4A of
the 1 a#or"hous alloys for#ed y non3#elt /uenching, the "rediction accuracy +as
4G.@O. f the total 1?G alloys studied, the overall accuracy +as GG.GO. hese results
co#"ared +ell +ith the +or0 done y Shi et al. :1B; using ond3"ara#etric
diagra#s, in +hich the ond "ara#eters &r' F and !(Vr' F >+here the ' is ato#ic
valence, &r is ato#ic 0ernel radius e/uals a""ro8i#ately the "ositive ionic radius
not including the valence electrons, and !(Vr is covalent radiusA +ere used to "redict
the for#ation of inary a#or"hous alloys, and the "rediction accuracy +as D$O.
a0ing the "ara#eters used in Miede#a5s coordinates +ithout the sie factor, hang
:1BB; co#ined the t+o che#ical coordinates \ andBF1
sn into one-
( ) 1F1B21$ BF1\ = sny , >E/uation ?A
and using the radius difference of ele#ents
12
-
8/12/2019 Review H-R Paper Revised
20/47
O
1
1
#
##x
= >E/uation @A
as the other coordinate to constructed a t+o3di#ensional #a". insert tale 1 hereA
hang and co3+or0ers "ro"osed another gra"hical #ethod, +hich co#ined &r' F
and !(Vr' F as an electron factor, then together +ith a sie factor to "rovide t+o
coordinates for the study of solid soluility, they descried the for#ation of
a#or"hous alloys :1B4;. he t+o coordinates are e8"ressed y >1A ond3"ara#etric
function %%&
xr
'x
r
'
+B
B
1
cov
and >A half of the e#"irical interato#ic distance
#res"ectively. &here &r' F is the ratio of ato#ic valence and ato#ic 0ernel radius,
and !(Vr' F the ratio of ato#ic valence and covalent radius. he *ernal radius
$
-
8/12/2019 Review H-R Paper Revised
21/47
e/uals a""ro8i#ately the "ositive ionic radius. %x is the Hauling electronegativity.
&hen the coordinate "oint of a host ele#ent is re"resented in the chart, the closer is
the "oint of a solute ele#ent to it, the s#aller the differences of the electron factors
and of the sie factors et+een the solute ele#ent and the host ele#ent. hang :1B$;
a""lied this #ethod to 1$D$ inary and got a "rediction accuracy of 2BO.
hang used a #odified electron factor
( )%B%AA!(V
%A
B&
))
r
')
r
'y
+
=
B
B
1 >E/uation GA
as ordinate and O1
1
#
##x
= as ascissa to search for solid soluilities at roo#
te#"erature in 4@$ inary alloys. n this +or0 the transition #etals of the fourth,
fifth and si8th long "eriods and 1D non3transition #etals are studied. E/uation DA
1
-
8/12/2019 Review H-R Paper Revised
22/47
-
8/12/2019 Review H-R Paper Revised
23/47
Hredictions of solid soluility of other alloy syste#s using first "rinci"les can e
found else+here :14?31@@;.
1.' %olubility prediction using artificial neural net0ors
(lthough a large nu#er of researches on soluility "rediction have een "ulished in
the last century, fe+ of the# have atte#"ted to "redict the soluility /uantitatively.
he sie factor has een regarded as the #ost i#"ortant factor ut has never een
/uantitatively "roved. Recently, the authors :1@G; have used one tool of artificial
intelligence, artificial neural net+or0s >(NNsA, to si#ulate the "rocess that 'u#e3
Rothery used to derive the 'u#e3Rothery5s Rules fro# e8"eri#ents. 1A ato#ic sie "ara#eter, >A
valence "ara#eter, >BA electroche#ical "ara#eter, i.e. electronegativity and >4A
structure "ara#eter of solvent and solute ato#s. hree different e8"ressions of these
"ara#eters +ere used to e8a#ine +hich gave the est "erfor#ance-
1. he ra+ data that 'u#e3Rothery used.
. he original collected values for each "ara#eter of solvent and solute ato#s.
B. he functionalied "ara#eters-
B
-
8/12/2019 Review H-R Paper Revised
24/47
iA -or the sie factor. he difference et+een the ato#ic dia#eters of solvent
and solute ato#s divided y the dia#eter of the solvent ato#s.
iiA -or the valence factor. riginal values are used, leaving the neural
net+or0 to decide the relations et+een valence of solvents and solutes.
iiiA -or the electroche,ical factor. he difference et+een that of the solvent
and solute ato#s.
ivA -or the structure para,eter. he structures are e8"ressed in three sets of
nu#ers re"resenting "ri#itive cell di#ensions, angles and syste#s. he
three sets are >1A unit cell length, >A a8es angles and >BA >si#"le ase3
centred face3centred ody3centredA.
&hen the original values of in"ut "ara#eters are used, the training "erfor#ance of the
(NN is /uite good >+ith regression coefficient R$.22@A ut the "rediction of the
testing set is "oor. &hen the functionalied values +ere used as in"ut "ara#eters,
oth the training and testing set of the (NN gave good "erfor#ance. hrough a
search of all co#inations of those "ara#eter for#ats, the est for#at of in"ut
"ara#eters +as deter#ined as the functionalied values.
(nother uncertainty in the '3R rule is the ato#ic sie difference- so#e researchers
elieve the threshold is 14O and others elieve it5s 1?O. he "erfor#ance of (NN
using the 1?O criterion is slightly higher than that for 14O +hich i#"lies 1?O is a
etter threshold in the sie factor.
Using the sa#e a""roach and including the 1?O criterion, the structure "ara#eter +as
introduced in ter#s of +hether the structure of solvents and solutes are the sa#e or
4
-
8/12/2019 Review H-R Paper Revised
25/47
not >i.e. 1 sa#e, $ not sa#eA. here +as no i#"rove#ent in correlation. his indicates
that the structure "ara#eter do not "lay a very i#"ortant role in soluility.
1.$ +etermination of the output parameters
at.OA fro# e8"eri#ent results and y3a8is are
"redicted soluility >at.OA "redicted fro# (NN. Most of the data "oints are located
close to y8 indicating an accurate "rediction.
1.* +etermination of the relative importance of each rule
t has een recognied that each rule has a different influence on soluility- the #ost
i#"ortant rule is the ato#ic sie factor, follo+ed ne8t y the electroche#ical factor
>electronegativityA. 'o+ever, these "ara#eters are not +holly inde"endent of each
other their inter"lay #a0es the deter#ination of soluility very difficult. So the
deter#ination of the relative i#"ortance of each rule is not easy. n this research, the
"erfor#ance of the (NN +as evaluated +hen so#e of the rules +ere delierately
o#itted. Using the #ean error of the testing set as the #ain criterion for accuracy of
?
-
8/12/2019 Review H-R Paper Revised
26/47
the "rediction, ato#ic sie has the strongest effect ecause, +hen it is o#itted, the
error is highest. Electronegativity has a stronger influence than valence. &hen "airs of
"ara#eters are o#itted, the "erfor#ance of the "rediction eca#e +orse. #ission of
the structure "ara#eter only had s#all effect on the "erfor#ance of the (NN, +hich
i#"lies that the structure "ara#eter does not "lay a very i#"ortant role, and indeed
'u#e3Rothery did not include it in 12B4.
6.' %ummary
!uring the decades after 'u#e3Rothery5s Rules +ere "ulished, the value and
reliaility of the# has een discussed e8tensively. hey have een deated, e8tended
and refined in different +ays, as #entioned in the te8t. (lso, as 'u#e3Rothery and
co3+or0ers stated at the ti#e +hen they first raised these rules- Kn general, the
soluility li#it is #ainly deter#ined y these factors, and it is their inter"lay that
#a0es the results so co#"le85 :@;. 'o+ever, even though one cannot "redict the solid
soluility li#its accurately using 'u#e3Rothery5s rules, they are still useful
guidelines for 9udging the soluility of alloy syste#s or for#ation of inter#etallic
co#"ounds.
Massals0i :4; suggests that although 'u#e3Rothery5s rules are i#"ortant, insufficient
docu#ented e8a#"les of "ractical a""lication and "redictive ca"aility have een
de#onstrated so far and this is a real challenge to the future of alloy develo"#ent and
of inter#etallic co#"ound for#ation. (t "resent, it is necessary to accu#ulate a large
a#ount of +or0 to investigate +hich rules have een deter#ined and +hich have not.
Ne+ investigation tools, such as ( >(rtificial ntelligenceA, can e used to validate
those rules and to e8"lore ne+ "ara#eters or rules in the future.
@
-
8/12/2019 Review H-R Paper Revised
27/47
&cno0ledgments
he authors are grateful to School of Engineering and Materials Science of Queen
Mary, University of London to su""ort this +or0 y "roviding research studentshi" to
W.M. hang.
G
-
8/12/2019 Review H-R Paper Revised
28/47
References
:1; *. R. Ho""er. Con9ectures and Refutations- he %ro+th of Scientific
*no+ledge. Routledge, (ingdon. U*. 12@B. "". BB3@?.
:;
-
8/12/2019 Review H-R Paper Revised
29/47
:1?; =. &. hang and S. . Liao. Hrogress on the theories of solid soluility of
alloys >"artA. Shang'ai Met. 1, 1222, B31$.
:1@; . =. Massals0i and '. &. *ing. (lloy "hases of the nole #etals. Hrog.
Mater Sci. 1$, 12@B, B3GD.
:1G; . =. Massals0i and U. Miutani. Electronic3structure of 'u#e3Rothery
"hases. Hrog. Mater. Sci. , 12GD, 1?13@.
:1D; . =. Massals0i. Structure and staility of alloys. n- R. &. Cahn and H.
'aasen eds. Hhysical Metallurgy. Ne+ Wor0- North 'olland, 122@. "". 1B?3
$4.
:12; !. %. Hettifor. &illia# 'u#e3Rothery- his life and science. n- E. (. urchi,
R. !. Shull and (. %onis eds. Science of alloys for the 1st century- a
'u#e3Rothery sy#"osiu# celeration. &arrendale- MS >he Minerals,
Metals ^ Materials SocietyA, $$$. "". 23B.
:$; R. E. &atson and M. &einert. he 'u#e3Rothery _"ara#eters_ and onding
in the 'u#e3Rothery and transition3#etal alloys. n- E. (. urchi, R. !. Shull
and (. %onis eds. Science of alloys for the 1st century- a 'u#e3Rothery
sy#"osiu# celeration. &arrendale- MS >he Minerals, Metals^ Materials
SocietyA, $$$. "". 1$?3112.
:1; E. Harthe.
-
8/12/2019 Review H-R Paper Revised
30/47
:2; &. 'u#e3Rothery, R. E. S#all#an and C. &. 'a+orth. he Structure of
Metals and (lloys. ?th ed. London- Metals and Metallurgy rust of the
nstitute of Metals and the nstitution of Metallurgists, 12@2.
:B$; ). . &aer, *. (. %schneidner, (. C. Larson and W. H. Margaret. Hrediction
of solid soluility in #etallic alloys. rans. Metal. Soc. (ME. G, 12@B,G1G3GB.
:B1; ).
-
8/12/2019 Review H-R Paper Revised
31/47
:4?; S. Singh and &. '. Woung. 'eats of solution for si#"le inary alloys. ). Hhys.
-
8/12/2019 Review H-R Paper Revised
32/47
:@1; &. =. Hearson. he Crystal Che#istry and Hhysics of Metals and (lloys. Ne+
Wor0 London- &iley3nterscience, 12G. ".@D.
:@; *. W. Li and !. ECSA #odel for "redicting inary solid solutions. Hrog. Mater. Sci. 42,
$$4, 41134D.
:@4; (. R. Miede#a. Electronegativity Hara#eter for ransition3Metals 3 'eat of
-
8/12/2019 Review H-R Paper Revised
33/47
-
8/12/2019 Review H-R Paper Revised
34/47
:2; X. C. %ui, S. . Liao, '. &. Xie and =. &. hang. Haraola #odel of
for#ation la+ of /uasicrystal ased on the fourth transition #etals. Rare Met.
Mater. and Eng. B?, $$@, 1$D$31$D4.
:2B; (. R. Miede#a and H.
-
8/12/2019 Review H-R Paper Revised
35/47
:1$@; %. =oolo, ).
-
8/12/2019 Review H-R Paper Revised
36/47
:1$; L. *. eles, ).
-
8/12/2019 Review H-R Paper Revised
37/47
:1BB; =. &. hang. !escri"tion of the for#ation of inary a#or"hous3alloys y
che#ical coordinates. Hhysica = ^ C. 1B, 12D?, B123B.
:1B4; =. &. hang. he effect of sie factor in the for#ation of a#or"hous alloys.
(ctaMetall Sin. 1G, 12D1, D?32.
:1B?; =. &. hang. Se#i3e#"irical theory of solid soluility in inary3alloys. .
Metal0d. G@, 12D?, @43G$.
:1B@; =. &. hang and S. . Liao. heory of solid soluility for rare earth #etal
ased alloys. . Hhys. =- Condens. Matter. 22, 122@, B?34B.
:1BG; R. &. lesins0i and %. ). (aschian. he Cu3Si >Co""er3SiliconA syste#. ).
Hhase E/uili. G, 12D@, 1G$31GD.
:1BD; (. unger, S. '. &ei, L. %.
-
8/12/2019 Review H-R Paper Revised
38/47
:14D; S. Wa#a#oto, . &a0aayashi and '. *oayashi. Calculation of solid
soluility and co#"ound for#aility of (l3alloys y e8tended 'uc0el3#ethod.
). )"n. nst. Met. ?G, 122B, 1B@G31BGB.
:142; &. . Luo and M. E. Schlesinger. her#odyna#ics of the iron3caron3inc
syste#. Metall. Mater. rans. =. ?, 1224, ?@23?GD.
:1?$; =. 'alle#ans H. &ollants and ). R. Roos. her#odyna#ic reassess#ent and
calculation of the
-
8/12/2019 Review H-R Paper Revised
39/47
:1@; M. '. 138ASi solid solution
structural staility. ). (lloys Co#"d. 4@?, $$D, 4@34@G.
:1@G; W. M. hang, S. Wang and ). R. %. Evans. Revisiting 'u#e3rothery`s rules
+ith artificial neural net+or0s. (cta Mater. ?@, $$D, 1$24311$?.
B2
-
8/12/2019 Review H-R Paper Revised
40/47
ale Ca"tions-
ale 1 Co#"arison of the for#ation of inary a#or"hous alloys using
different coordinates >redra+n fro# :12;A
ale Co#"arison of the "rediction of solid soluilities y different #ethods
>redra+n fro# :1?;A
4$
-
8/12/2019 Review H-R Paper Revised
41/47
ales-
ale 1
Hure Miede#a5s
coordinate criterion
Miede#a5s
coordinate "lus sie
factor
=ond3"ara#etric
diagra#
>1A (#or"hous alloys
y #elt /uenching
total D D B
+rong @ 1$ B
accuracy >OA 2.G DG.D 2$.@
>A Non3for#ale
a#or"hous alloys
total ?4 ?4 ?G
+rong 1D 2 G
accuracy >OA @@.G DB.B DG.G
>BA (#or"hous alloys
y non3#elt/uenching
total 1 1 BD
+rong 11 G 1$
accuracy >OA 4G.@ @@.G GB.G
otal accuracy >OA >1A >A D.4 D@ DD.D
>1A >A >BA GG.G DB.4 D4.B
41
-
8/12/2019 Review H-R Paper Revised
42/47
ale
No. of alloy syste#s and the "rediction accuracy
otal No. Hrediction accuracy O
'u#e3Rothery5s rule >sie factor onlyA 14B @G.@
!3% Method 14?? G@.@Cheli0o+s0y Method 12 D
(lonso Method B4 2$
hang =& Method >"araola se"arationA BD@4 DG.
hang =& Method >elli"se se"arationA BD@4 2$.B
4
-
8/12/2019 Review H-R Paper Revised
43/47
-
8/12/2019 Review H-R Paper Revised
44/47
-
8/12/2019 Review H-R Paper Revised
45/47
>aA
>A
-
8/12/2019 Review H-R Paper Revised
46/47
>aA
>A
-
8/12/2019 Review H-R Paper Revised
47/47
0 20 40 60 80 100-20
0
20
40
60
80
100
Experimental Solubility T (at.%)
PredictedSolubilityA(at.%)
Best Linear Fit: A = (0.977) T + (0.23)
R = 0.993
0 20 40 60 80 100-20
0
20
40
60
80
100
Experimental Solubility T (at.%)
PredictedSolubilityA(at.%)
Best Linear Fit: A = (0.962) T + (-1.19)
R = 0.992
0 20 40 60 80 100-20
0
20
40
60
80
100
Experimental Solubility T (at.%)
PredictedSolub
ilityA(at.%)
Best Linear Fit: A = (0.966) T + (0.0162)
R = 0.992
Data Points
Best Linear Fit
A = T
(a) Training set (b) Testing set
(c) Whole set