Research: Science and Education

www.JCE.DivCHED.org Vol. 82 No. 2 February 2005 Journal of Chemical Education 325

The bond-type triangle is being used increasingly in un-
dergraduate textbooks to illustrate the central importance of
electronegativity and its influence on bond type (1a, 2). It
has long been recognized that chemical bonds can be divided
into three fundamental types: ionic, covalent, and metallic.
van Arkels (3) depiction of the three bond types as the verti-
ces of a triangle, with compounds or elements of intermedi-
ate nature being located along the edges, was the first to
attract widespread attention (Figure 1A).
1
Ketalaar (4) fur-
ther developed the bond triangle by placing compounds
within it as well as on the edges (Figure 1B).
Recent work has focused on the rationalization of the
bond triangle by using functions of electronegativity as the
basis for determining the positions of substances within it.
Jensen

(5), Allen et al. (6), and Sproul (7) all recognized the
essential role of electronegativity and electronegativity dif-
ference in determining the properties of substances. All of
them concluded that the position of any binary compound
in the bond triangle could be specified by two coordinates:
electronegativity difference (!") and average electronegativ-
ity ("
av
) (Figure 2). These authors noted that plotting a graph
of !" versus "
av
produces a triangular array with ionic com-
pounds (high !", moderate "
av
) near the top, metals (low
!", low "
av
) near the lower left, and covalent substances (low
!", high "
av
) near the lower right. Both Sproul and Allen
suggested that the bond triangle as defined above can be di-
vided into covalent, ionic, and metallic regions with straight
lines as boundaries.
In the triangles devised by each of these authors (Figure
2), the vertices are occupied by Cs (metallic), F
2
(covalent),
and CsF (ionic); the elements lie along the metalliccova-
lent axis (!" = 0), while cesium compounds and fluorides
are found along the metallicionic and covalentionic edges,
respectively. As observed by Jensen (5), the compounds of
any element other than Cs or F lie on two diagonal lines,
parallel to the metallicionic and covalentionic edges, which
meet at the position occupied by the element itself. The left
line (with negative slope) includes compounds of the element
with elements less electronegative than itself, while the right
line (with positive slope) includes the compounds of the ele-
ment with more electronegative elements. All compounds
formed by a given pair of elements occupy the same posi-
tion.
Sproul (7b) did the most extensive work on quantifica-
tion of the bond triangle, plotting values of !" versus "
av
for
311 binary compounds and observing excellent correlation
between position on the graph and bond type as specified by
Wells (8). He excluded compounds whose bond type was said
by Wells to be ambiguous or questionable, such as semi-
metallic compounds. Sproul examined 15 different elec-
tronegativity scales and found that the best agreement
between position and bond type was obtained when Allens
spectroscopic electronegativity (9)later called configuration
energy (10)was used,
2
only 9 of the 311 compounds be-
ing misplaced from their specified regions. He found that,
for his selected compounds, the triangle could be divided into
covalent, ionic, and metallic regions by straight lines, paral-
lel to the metallicionic and covalentionic edges of the tri-
angle (Figure 2 bottom). These boundaries were defined by
the equations (with " values converted to Pauling units):

av
0 5 1 60 ! . . = + " " CCovalent Ionic: #
Metal Nonmetal:
av
= # + # " " 0 5 2 28 . . !
Electronegativity and the Bond Triangle W
Terry L. Meek* and Leah D. Garner
Department of Biological and Chemical Sciences, University of the West Indies, Cave Hill, Barbados;
*tlmeek@uwichill.edu.bb
Figure 1. Early bond triangles: (A) van Arkel, ref 3, and (B) Ketalaar,
ref 4.
A
B
Research: Science and Education
326 Journal of Chemical Education Vol. 82 No. 2 February 2005 www.JCE.DivCHED.org
Figure 2. Quantified bond triangles:
(top) Jensen, ref 5
(center) Allen et al., ref 6
(bottom) Sproul, ref 7b
Research: Science and Education
www.JCE.DivCHED.org Vol. 82 No. 2 February 2005 Journal of Chemical Education 327
Comments and Limitations
These boundaries are not entirely satisfactory, probably
because some of the compounds used were inappropriate (see
below) and Wellss specifications of bond type are inaccurate
in some cases and ambiguous in others. The covalentionic
borderline is just below the line that includes all compounds
of aluminum (" = 9.53 eV = 1.61 PU) with elements of
higher electronegativity, while the metalnonmetal (metal
insulator) line lies between those that include compounds of
phosphorus (" = 13.3 eV = 2.25 PU) and of hydrogen (" =
13.6 eV = 2.30 PU) with metals and metalloids. Accordingly,
1. No element with electronegativity greater than that of
Al would exhibit cationic character in any compound,
even with fluorine; Hg, Ga, In, Tl, Sn, and Pb are cat-
ionic in some compounds.
2. All elements with electronegativity equal to or less than
that of Al would form only ionic compounds with el-
ements more electronegative than P; the amphoteric
metals Al, Be, Cd, and Zn form covalent (or interme-
diate) compounds with many nonmetals.
3. No element that has an electronegativity less than that
of H would display anionic character in any com-
pound, even with cesium; compounds of P and the
metalloids Te and Po with the group 1 and group 2
metals are believed to have considerable ionic charac-
ter, though the tellurides and polonides are not well
characterized.
4. Any compound formed between two elements less elec-
tronegative than H would be metallic; some of these
elements are metalloids, and form semiconductors that
are not well-defined and do not display metallic con-
ductivity. The IIIV compounds are the best known
of these.
It must be noted that the distinction among bond types,
especially for polymeric substances, is quite arbitrary and it
is seldom possible to categorize compounds definitively as
being 100% ionic, covalent, or metallic. This was recognized
by Allen (6), who introduced a small triangular metalloid
region of substances with low !" and moderate "
av
. Jensen
(11) observed that there are not (or should not be) sharp
boundaries separating substances containing metallic, cova-
lent, and ionic bonds. In fact, most compounds exhibit a
mixture of two types of bonding and display some charac-
teristics of bothor even of all threetypes, as exemplified
by the IIIV semiconductors and some Zintl phases. Thus
the categorization of compounds as ionic, covalent, or me-
tallic is a considerable oversimplification, and nearly all com-
pounds are really intermediate in character. It would be more
accurate to say that the elements and many compounds can
be categorized as predominantly ionic, covalent, or metallic,
but most of them have some characteristics of at least two
bond types.
Jensen (11) also points out that the van ArkelKetalaar
triangle is a bond triangle and the accurate specification of
compounds and elements would require another dimension
such as is found in Grimms tetrahedron (12), with van der
Waals forces as the fourth vertex and substances with inter-
mediate degrees of polymerization occupying levels above the
array of three-dimensional polymers represented by the van
ArkelKetalaar triangle. The error of attempting to correlate
a property of bulk matter with the charge distribution in an
isolated molecule is further emphasized by the fact that some
compounds, although predominantly covalent molecular sub-
stances in the gas phase, consist of polyatomic cations and
anions in condensed phases. Well-known examples are N
2
O
5
and PCl
5
.
Despite these limitations, the position of a binary com-
pound in the bond triangle still gives a good indication of
the type of bonding that occurs in it and the bond triangle
gives some very useful insights (Figure 3). First is the realiza-
tion that bond type depends on both !" and "
av
, which is
confirmed by the positive slope of the ioniccovalent bound-
E
l
e
c
t
r
o
n
e
g
a
t
i
v
i
t
y

D
i
f
f
e
r
e
n
c
e

/

P
U
Average Electronegativity / PU
I
C
M
5
4
3
2
1
0
0 1 2 3 4 5
I
C
MD
M
E
l
e
c
t
r
o
n
e
g
a
t
i
v
i
t
y

D
i
f
f
e
r
e
n
c
e

/

P
U
5
4
3
2
1
0
Average Electronegativity / PU
0 1 2 3 4
5
I
C
MD
C/I
M
E
l
e
c
t
r
o
n
e
g
a
t
i
v
i
t
y

D
i
f
f
e
r
e
n
c
e

/

P
U
5
4
3
2
1
0
Weighted Average Electronegativity / PU
0 1 2 3 4 5
Figure 3. Evolution of quantified bond triangles: (A) dependence
on one variable, (B) boundaries according to Sproul and Allen,
and (C) this work.
A
B
C
Research: Science and Education
328 Journal of Chemical Education Vol. 82 No. 2 February 2005 www.JCE.DivCHED.org
ary and the negative slope of the metalnonmetal boundary
in Sprouls and Allens representations (Figure 3B). Sproul (7d)
points out that, if ionic character were associated only with a
large !", there would be a horizontal line at some critical !"
value such as the frequently cited 1.7 Pauling units (13), sepa-
rating ionic species from others.
3
Furthermore, if metallic char-
acter were associated only with a small "
av
, there would be a
vertical line at some critical "
av
valueperhaps between those
of Pb, the most electronegative metal, and Si, the least elec-
tronegative nonmetalseparating metals from nonmetals.
Although the above authors have demonstrated that bond
type depends on both !" and "
av
, none of them has offered
an explanation for the manner in which ioniccovalent or
metallicnonmetallic character varies with both parameters.
We now address these matters. We will also show that there
are gaps between the regions occupied by predominantly ionic
and predominantly covalent substances, and between metals
and nonmetals, which are occupied by substances generally
regarded as being of intermediate nature.
Ionic versus Covalent Bonding: Partial Charge
The dependence of ionic character on "
av
as well as on
!" is shown in Table 1; for each group of compounds with
very similar !" values, ionic character decreases as "
av
in-
creases. This variation results from the fact that partial charge,
rather than !" alone, is the critical parameter in determin-
ing ionicity. Although the unequal distribution of charge has
its origin in !", the magnitude of the partial charge is also
dependent on "
av
. We will show that partial charge is in fact
closely related to the ratio !": "
av
.
For diatomic singly bonded molecules of general formula
AB, Wilmshurst (14) suggested that the partial charge on A
(q
A
) is given by,
A
B A
B A
q =
#
+
" "
" "
(1a)
where "
A
and "
B
are the electronegativities of elements A and
B. Thus
A
av
q =
"
"
!
2
(1b)
At least four other more recently derived equations for
evaluating partial charge can be shown to be equivalent to
eq 1b. Smith (15) used an electronegativity equilibration ap-
proach to show that, in the singly bonded molecule AB, the
partial charge on A is given by,
where
A
A
A
A B
q =
#
=
" "
"
"
" " 2
,
"" "
A B
+
(2a)
thus:
"
"
" "
" " "
" "
A
A
A B
A B A
B A
= # =
+
#
=
#
" "
A B
+
q 1
2
1
1
==
!"
" 2
av
(2b)
Identical expressions are obtained using the methods of
Bratsch

(16) and von Szentpaly (17), if one makes the ap-
proximation that the charge coefficient (16) or atomic hard-
ness

(17) of an atom is equal to its electronegativity.
In Allens modified LewisLangmuir method

(18), for
diatomic molecules with single bonds:
= # #
A
Group No.
No. of
Unpaired
Electrons
q 22
"
" "
A
A B
+
(3a)
Thus, if neither atom has a formal charge:

1
2
2
"
" "
" "
" "
"
"
A
A
A B
B A
A B
av
= #
+
=
#
+
= q
!
(3b)
According to all of these equations, the partial charges
of bonded atoms (hence ionic character of a bond) increases
as the ratio !": "
av
increases, and all compounds whose at-
oms have the same partial charges will have the same !": "
av
ratio. Thus in a bond triangle with axes !" and "
av
, all com-
pounds with the same degree of ionic character will lie on a
straight line, passing through the origin (!" = "
av
= 0) when
extrapolated, with slope equal to 2q (Figure 3C). One such
line will have a slope equal to the critical q value above which
compounds are predominantly ionic, and will correspond to
the ioniccovalent boundary.
In fact, for diatomic molecules with multiple bonds the
partial charge is equal to the !":2"
av
ratio multiplied by the
bond order; this is recognized by Wilmshurst (14) and Allen
(18) in their articles, but the equalization-based methods (15,
16) do not take bond order into account and von Szentpaly
(17) considers only singly bonded molecules. Still, the ratio
!": "
av
in molecules with multiple bonds is equal to twice
the partial charge per bond and gives a good indication of
the extent of ionic character in the bond.
Metals versus Nonmetals: Band Gap
The critical parameter in determining metallic behavior
is band gap (E
g
), since conductivity is proportional to
exp(E
g
kT). Thus substances with very small band gaps are
metallic conductors, while those with large band gaps are in-
sulators. For the elements, it has been established (19) that
n o r e t c a r a h C c i n o I f o e c n e d n e p e D . 1 e l b a T " """"
v a
e c n a t s b u S !" U P /
"
v a
U P /
!":"
v a
t n a n i m o d e r P
e p y T d n o B
l C i L 6 9 . 1 9 8 . 1 7 3 0 . 1 c i n o I
F H 9 8 . 1 5 2 . 3 2 8 5 . 0 t n e l a v o C
I i L 5 4 . 1 4 6 . 1 4 8 8 . 0 c i n o I
N l A 5 4 . 1 4 3 . 2 0 2 6 . 0 e t a i d e m r e t n I
F r B 0 5 . 1 4 4 . 3 6 3 4 . 0 t n e l a v o C
i B i L 0 1 . 1 6 4 . 1 3 5 7 . 0 c i l l a t e M
e S g M 3 1 . 1 6 8 . 1 8 0 6 . 0 c i n o I
O C 7 0 . 1 8 0 . 3 7 4 3 . 0 t n e l a v o C
NO : e t " f e r m o r f ) s t i n u g n i l u a P ( s e u l a v . 9 1
Research: Science and Education
www.JCE.DivCHED.org Vol. 82 No. 2 February 2005 Journal of Chemical Education 329
" alone can be used to determine metallic character, as the
width of an energy band is inversely proportional to the en-
ergy of the level from which it arises. All metals have " val-
ues less than 1.92 (Si), while all nonmetals have " values
greater than 2.21 (As); the metalloids occupy the narrow
range of " between these values.
For compounds, the extent of metallic character depends
on both !" and "
av
. Sets of compounds with very similar
average electronegativities display increasing band gaps and
thus decreasing metallic character as electronegativity differ-
ence increases, as seen in Table 2. This was noted by Allen
and Capitani (20) and is consistent with equations, devised
by various authors, that define the band gap.
Phillips and van Vechten (21) suggested that, for binary
compounds with a total of eight valence electrons per AB
unit, the band gap in crystalline substances is composed of
ionic (E
i
) and covalent (E
c
) contributions according to:
g
E
22 2 2
= + E E
c i
(4)
Adams (22) redefined these terms in his equation,
2 2 2
= + E E C
g h
(5)
where C, the charge transfer energy, is proportional to !",
and E
h
, the homopolar energy gap, (like E
c
in eq 4) is de-
fined as the energy difference between the center of the va-
lence band and that of the conduction band. E
h
is a measure
of the effectiveness of interaction between atoms, and thus
should depend on the same factors as " and "
av
.
For s-valent diatomic molecules, Pettifor (23) gives the
equation,
2 2
2
4 = + ( ) w h E
AB
! (6)
where w
AB
is the separation between bonding and antibond-
ing states, h is the bond integral and is related to the average
atomic potential, and E, the free atomic energy level, is ap-
proximately equal to ". Each of these equations is of the form
E
2
= a
2
+ b
2
, where a is related to the average energy of the
atoms or to the effectiveness of their interaction (hence to
"
av
) and b is related to an energy difference (hence to !").
This suggests that substances with the same band gap will lie
on a line with the same values of (c
1
!"
2
+ c
2
"
av
2
). Such a
line would be a curve a segment of a circle if !" and "
av
contribute equally to band gap, or of an ellipse if they do
not. Thus the metalnonmetal boundary would be the curve
for which the E value corresponds to the largest band gap
that allows metallic conduction (Figure 3C).
It must be realized that !" and "
av
are derived from
properties of isolated atoms whereas band gap is a property
of an aggregated solid, so a rigorous relationship between
band gap and these parameters is unlikely. Still, qualitative
agreement has been observedJensen (5) reported a quick-
and-dirty test with a crude probe-buzzer-battery device that
showed a curved boundary, in a graph of !" vs "
av
, between
substances that showed detectable conductivity and sub-
stances that did not.
Polyatomic Molecules: Weighted Average " "" ""
One anomaly that arises in this representation of bond
type is that all of the compounds formed by a given pair of
elements occupy the same position on the graph. In fact, bond
type can vary with the valence of the central atom. For ex-
ample, tin(IV) halides are predominantly covalent, as are the
few stable ones of lead(IV), with all but the fluorides nor-
mally existing as monomeric molecules. However, lead(II) ha-
lides are ionic and those of tin(II) have substantial ionic
character.
For polyatomic molecules, the relationship among par-
tial charge, !", and "
av
is more complex than for diatomics.
For compounds formed between the same two elements,
natural population analysis calculations (24) show that the
partial charge on each terminal atom decreases as the valence
of the central atom increases. This suggests that the simple
"
av
may not be the appropriate parameter to use. We pro-
pose a weighted average electronegativity, defined for A
m
B
n
as:
=
+ m n
w
av
A
"
" ""
B
m n +
(7)
It follows that, using this definition of ("
av
)
w
:
" "
" "
A B

m n
m n
=
+ ( )
+
! !
w
av
"
(8)
The ratio !":("
av
)
w
is not rigorously related to atomic
chargewe have shown elsewhere (25) that, for molecules
A
n
B, the charge on A is given by
"
" "
A
A
q
n
n
=
#
+
!
2 1
BB
(9)
but nevertheless !":("
av
)
w
serves to discriminate between
compounds of the same two elements with different stoichi-
ometries. From eq 8, such compounds will have the same
!" but different ("
av
)
w
values. (Using the fluorides of lead as
examples, the ("
av
)
w
values for PbF
4
and PbF
2
are 3.73 and
3.41 Pauling units.) Each compound thus occupies a unique
position in the triangle. Because of the positive slope of the
covalentionic boundary, this can result in two such com-
pounds being in different bonding regions.
h t i w r e t c a r a h C c i l l a t e M f o n o i t a i r a V . 2 e l b a T !" """"
e c n a t s b u S "
v a
U P /
a
!" U P /
a
d n a B
V e / p a G
e c n a t c u d n o C
e G 9 9 . 1 0 7 . 0
b
r o t c u d n o c i m e S
s A a G 8 9 . 1 6 4 . 0 5 . 1
b
r o t c u d n o c i m e S
e S n Z 1 0 . 2 3 8 . 0 8 . 2
b
r o t a l u s n I
n S 2 8 . 1 0 0 0 . 0 8 .
bb
l a t e M
b S n I 2 8 . 1 3 3 . 0 2 . 0
c
r o t c u d n o c i m e S
e T d C 4 8 . 1 4 6 . 0 5 . 1
c
r o t c u d n o c i m e S
e S g M 5 8 . 1 3 1 . 1 ? 8 .
b
r o t a l u s n I
r B i L 0 8 . 1 8 8 . 1 ? 8 .
b
r o t a l u s n I
a
f e r m o r f ) s t i n u g n i l u a P ( s e u l a V . 9 1
b
f e R . b 1
c
f e R . 0 3
Research: Science and Education
330 Journal of Chemical Education Vol. 82 No. 2 February 2005 www.JCE.DivCHED.org
Boundaries: Well-Defined Compounds and Elements
The selection of compounds for establishing boundaries
between predominant bond types should be restricted to bi-
nary compounds that have been isolated and purified, have
known structures, and contain only heteronuclear bonds. This
last criterion excludes many types of compounds, for example:
(a) Compounds with discrete dinuclear ions, which could
be described as both ionic (since they contain ions)
and covalent (since some of the ions are polyatomic).
(b) Molecular compounds, including clusters, which con-
tain both homonuclear and heteronuclear covalent
bonds.
(c) Zintl compounds, which display some degree of me-
tallic interaction as well as containing simple cations
and extended polyatomic networks of covalently
bonded atoms, each with a formal negative charge (1c).
(d) The group 1 compounds M
3
E of the group 15 ele-
ments As, Sb, and Bi, whose structures feature homo-
nuclear intermetallic bonding as well as heteronuclear
interactions (26a). The group 2 compounds M
3
E
2
of
these elements also display some metallic properties.
(e) The suboxides of rubidium and caesium, which con-
tain cationic clusters of identical atoms, such as M
9
4+
and M
11
6+
, associated through metallic interactions
(1c).
The above criteria were applied to 321 binary com-
pounds, formed by the elements of groups 1, 2, and 1218
in periods 16,
4
in which the bonds are described by Green-
wood and Earnshaw (26) as being predominantly of one type.
A plot of !" vs ("
av
)
w
(Figure 4) was constructed for these
compounds; the 37 well-defined elements of these groups
omitting the metalloidswere also included, to help es-
tablish the covalentmetallic boundary. All of these species
are listed in the Supplemental Material.
W
The electronega-
tivity values used were those of Allen et al. (19), expressed in
Pauling units. For polyatomic molecules the weighted aver-
age electronegativity defined by eq 7 was used rather than
the simple average.
In this graph a triangular array of points is still obtained,
but there are some significant differences from the triangles
devised by other workers:
1. The compounds of a given element no longer lie on a
single line (Cs and F) or a pair of converging lines (all
other elements). Instead a number of lines, converg-
ing at the position of the element, are obtained; for
example, all monofluorides lie on one line, all
difluorides on another line, and so forth.
2. Some of these lines (polyfluorides, caesium compounds
of multiply-charged anions, etc.) extend outside the
boundaries of the CsCsFF
2
triangle. This necessi-
tates a relocation of the vertices, to produce a triangle
that includes all compounds. The metallic and cova-
lent vertices can be defined as the points with (!" =
0, ("
av
)
w
= 0) and (!" = 0, ("
av
)
w
= 5); the ionic vertex
would then be at (!" = 5, ("
av
)
w
= 2.5). This enlarged
triangle includes all well-defined compounds of the
main-group elements. Figure 4 shows that there is good
discrimination among predominant bond types.
Figure 4. Bond triangle for well-defined substances: circles = mainly ionic, squares = mainly
covalent, and triangles = mainly metallic.
Weighted Average Electronegativity
/
PU
E
l
e
c
t
r
o
n
e
g
a
t
i
v
i
t
y

D
i
f
f
e
r
e
n
c
e

/

P
U
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Research: Science and Education
www.JCE.DivCHED.org Vol. 82 No. 2 February 2005 Journal of Chemical Education 331
CovalentIonic Boundary
Compounds specified as covalent and those specified as
ionic are well separated by a straight line with the equation
!" = 0.62("
av
)
w
, as shown in Figure 4. This suggests that bi-
nary compounds with partial atomic charges greater than
0.31 (per bond) are predominantly ionic, while those with
q
A
values less than 0.31 are predominantly covalent. This find-
ing contrasts with Paulings implication (13) that 50% ionic
character marks the dividing line.
There is a small triangular region adjacent to the cova-
lentionic and covalentmetalloid boundariesthe third
side being the dashed line in Figure 4which contains very
few compounds specified as ionic or covalent. We will ob-
serve later that most of the compounds generally regarded as
being intermediate between ionic and covalent occupy this
region. A few compounds do not appear in their expected
regions of the triangle; most of these can be categorized as
follows:
1. The 14 compounds InCl, InBr, InI, GaCl, GaBr, GaI,
TlCl, TlBr, TlI, PbCl
2
, PbBr
2
, PbI
2
, BiF
3
, and Bi
2
O
3
,
specified as ionic, are in or very near to the covalent
region. (Five are in the small triangular region de-
scribed above.) It is possible that the misplacement of
these subvalent compounds arises from the defini-
tion of " as the average one-electron energy ($
l
) of all
of the valence electrons of an atom (7); for an atom
with the configuration s
n
p
m
:
"
$ $
=
+
+
n m
n m
s p
(10)
The predominately ionic character of these compounds
suggests that the metal atoms have electronegativities
considerably lower than those obtained from eq 10.
This would be consistent with the s electrons of the
metal taking little (if any) part in bonding, so that in-
cluding $
s
gives erroneously high " values for these el-
ements, hence low !" and high ("
av
)
w
values for their
compounds with nonmetals. This would be especially
pertinent for compounds of Tl(I), Pb(II), and Bi(III),
whose 6s orbitals are considerably stabilized by rela-
tivistic effects (27). (Apparently all subvalent ionic
compounds are misplaced, except SnO, PbO, PbF
2
,
and TlF). Only two other compounds specified as
ionicMgI
2
and MgSeare in the covalent section.
2. The only compounds specified as covalent that appear
in the ionic region are the polymeric species HgO,
PbF
4
, SnF
4
, SnF
2
, and GeF
2
. This is another instance
of the qualitative nature of the agreement between pre-
dominant bond type and location in the bond triangle.
As noted above, the ratio !":("
av
)
w
is not directly pro-
portional to partial charge,
5
even in polyatomic mol-
ecules.
MetalNonmetal Boundary
It is very difficult to establish the border between met-
als and nonmetals, since very few stoichiometric binary com-
pounds are known to be metallic. There is a substantial gap
in Figure 4 between compounds specified as metals and those
specified as nonmetals. The borders of these two regions in
the triangle appear to be defined by curved lines, which could
well mark the boundaries of a metalloid or semiconductor
region. However, linear boundaries cannot be ruled out at
this stage
The alloys LiBi and NaBi are very close to the interme-
diate, metalloid region in spite of having typical alloy struc-
tures (26a). They would be in the middle of the metallic
region if the electronegativity of Bi were assessed only from
the energy of its 6p orbitals, 0.599 Rydberg = 1.38 PU (19).
Intermediate Substances
The boundaries between bonding regions can be defined
more closely with the aid of another graph of !" versus ("
av
)
w
values, plotted in Figure 5 for 62 substances (8 elements and
54 compounds) that are usually described as intermediate.
These are also listed in the Supplemental Material.
W
They
are further categorized as follows:
CovalentIonic (23 Compounds)
Eighteen of these compounds are within the small tri-
angular area described above, separated from the covalent
region by the dashed line in Figures 4 and 5. Of the other
five, AlN is on the covalentionic boundary, two (Cd
3
N
2
and Zn
3
N
2
) are in the ionic region although quite close to
the covalentionic boundary, and two (SnS and PbS) are
in the covalent region but very close to the triangular area
containing most of the intermediate compounds. ZnSe, BeSe,
CdSe, and Be
2
C are very close to the metalloid region; Shriver
and Atkins (28) describe Be
2
C (and Al
4
C
3
) as borderline
between saline and metalloid, and an ab initio calculation
(29) suggests that Be
2
C is largely ionic. CdSe is a semicon-
ductor with a band gap of 1.7 eV (30).
Figure 5. Portion of bond triangle showing intermediate substances:
circles = covalentmetallic, squares = covalentionic, and triangles
= ionicmetallic.
Weighted Average Electronegativity
/
PU
E
l
e
c
t
r
o
n
e
g
a
t
i
v
i
t
y

D
i
f
f
e
r
e
n
c
e

/

P
U
0.0
0.5 1.0 1.5 2.0 2.5 3.0
0.5
1.0
1.5
2.0
2.5
Research: Science and Education
332 Journal of Chemical Education Vol. 82 No. 2 February 2005 www.JCE.DivCHED.org
CovalentMetallic (34 Substances)
The eight metalloid elements and 17 of the 26 cova-
lentmetallic compounds are in the metalloid region defined
above. AlSb and InSb are just inside the metallic region,
and GaSb is on the border between metals and metalloids.
The location of the semiconducting sulfides and selenides of
As, Sb, and Bi well inside of the covalent region seems very
anomalous.
IonicMetallic (5 Compounds)
These are also in the metalloid region. The somewhat
saltlike (26b) intermetallic compound CsAu, containing a
transition metal with electronegativity 1.92 PU (31), would
be on the metalnonmetal boundary.
Thus the diagram for intermediate substances indicates
that:
1. The metalloidnonmetal boundary is curved. A
straight line, drawn from the metalloidnonmetal bor-
der of the elements along the edge of the region occu-
pied by compounds specified as covalent and ionic,
places most known semiconductors among the pre-
dominantly covalent compounds; a straight line be-
tween covalent compounds and semiconductors places
several predominantly ionic compoundsthe iodides,
hydrides, sulfides, and selenides of the group 1 met-
alsamong the semiconductors.
The curvature of the metalloidnonmetal line suggests
that the metalmetalloid boundary is probably also a
curve rather than a straight line. All of the covalent
metallic and ionicmetallic substances examined lie
between or very close to the region between two curves,
as noted above.
2. There is an intermediate covalentionic region in the
bond triangle. Most compounds of this type occupy a
small triangular area adjacent to the covalentionic and
metalnonmetal boundaries. Of the specified com-
pounds, only a few subvalent ones and BeCl
2
, BeBr
2
,
MgI
2
, and MgSe are in this area. The latter four com-
pounds could well be described as intermediate.
Summary
1. The dependence of bond type on the two parameters
!" and ("
av
)
w
is examined. It is shown that, for di-
atomic molecules,
(a) The extent of ionic character in a bond depends
on the partial charges on the atoms, which for diatomic
molecules is proportional to the ratio !":("
av
).
(b) Metallic character is governed by the gap between
the highest occupied molecular orbital and the lowest
unoccupied molecular orbital, which is related to
c
1
(!")
2
+ c
2
("
av
)
2
.
2. These relationships are responsible for the shapes of
the covalentionic and metalnonmetal boundaries in
the bond triangle. The former is a straight line, while
the latter is a curve.
3. The critical values of partial charge and band gap at
which the boundaries occur may be estimated empiri-
cally by optimizing the sorting of a set of elements and
binary compounds that have been qualitatively char-
acterized as being predominantly ionic, predominantly
covalent, or predominantly metallic in their bonding.
Such a sorting procedure is at best only
semiquantitative, since we have no independent quan-
titative measure of predominant bond character.
4. Compounds of intermediate bond type also occupy
specific sections of the triangle.
(a) Most metalloids and semiconductors occupy an
area bounded by two curves, one separating them from
metals and the other separating them from insulators.
(b) Most compounds generally regarded as interme-
diate between ionic and covalent occupy a small, nearly
triangular region adjacent to the covalentionic and
covalentmetallic boundaries.
5. In spite of the limitations mentioned above, the posi-
tion in the triangle of a binary compound of main
group elements gives a reasonable indication of its pre-
dominant bond type. Lecturers and textbook writers
should bear the limitations in mind and point out that
the great majority of compounds contain bonds that
are in fact intermediate between two (or more) types,
even if they are predominantly of one kind. The exist-
ence of well-defined intermediate ioniccovalent and
metalnonmetal regions of the triangle, both of which
contain very few species having one predominant bond
type, is also notable.
6. Applicability of the bond triangle appears to be lim-
ited to simple binary compounds. Even for those, the
classification of Zintl compounds and others contain-
ing polyatomic ions, and of suboxides and other clus-
ter compounds, presents some difficulties. Such
compounds clearly contain at least two distinct types
of bonding, and the bond triangle in its present form
cannot represent this.
7. The usefulness of the bond triangle for categorizing
compounds of the main-group elements may be ex-
tended by the use of weighted average electronegativi-
ties to allow distinction between compounds of the
same elements with different stoichiometries. In such
cases a higher valency for the central atom leads to
greater covalent character and the compounds may
have different bond types.
W
Supplemental Material
A list of the substances specified as being one predomi-
nate type or an intermediate type are available in this issue
of JCE Online.
Acknowledgments
The authors are grateful to L. C. Allen of Princeton
University for the insights obtained in many stimulating dis-
cussions and to the reviewers for a number of helpful com-
ments and suggestions.
Research: Science and Education
www.JCE.DivCHED.org Vol. 82 No. 2 February 2005 Journal of Chemical Education 333
Notes
1. Jensen (11) has pointed out that triangles of this type can
be traced back to the work of Grimm and Fernelius in the period
19281936.
2. In fact, most of the major electronegativity scales show ex-
cellent correlation with one another and with Sprouls data set. Al-
though we will use the Allen scale in this article, any of the
commonly used ones would give similar results.
3. Pauling (13) in fact gives this !" value as the one that re-
sults in 50% ionic and 50% covalent character.
4. The data set was restricted to main-group elements and
their compounds at this time; the d-block metals and their com-
pounds will be examined in a subsequent article.
5. In fact the atomic charges per bond in these compounds
have all been calculated by a modified LewisLangmuir-type equa-
tion (25) as being substantially less than the !":("
av
)
w
ratios calcu-
lated from their empirical formulas. In all cases the charge per bond
is less than 0.31 in the polymer.
Literature Cited
1. Alcock, N. W. Bonding and Structure; Ellis Horwood: Lon-
don, 1990; (a) pp 1822, (b) pp 290291, (c) pp 309313.
2. (a) Norman, N. C. Periodicity and the p-Block Elements; Ox-
ford University Press: Oxford, 1993; pp 5556. (b) Mackay,
K. M.; Mackay, R. A.; Henderson, W. Introduction to Modern
Inorganic Chemistry, 5th ed.; Stanley Thornes: New York, 1996;
p 84. (c) van der Put, P. J. The Inorganic Chemistry of Materi-
als; Plenum: New York, 1998; p 12. (d) Wulfsberg, G. Inor-
ganic Chemistry; University Science Books: Sausalito, CA,
2000; pp 775778.
3. van Arkel, A. E. Molecules and Crystals; Interscience: London,
1949; p 205.
4. Ketalaar, J. A. A. Chemical ConstitutionAn Introduction to
the Theory of the Chemical Bond; Elsevier: New York, 1953; p
21.
5. Jensen, W. B. J. Chem. Educ. 1995, 72, 395398.
6. Allen, L. C.; Capitani, J. F.; Kolks, G. A.; Sproul, G. D. J.
Mol. Struct. 1993, 300, 647655.
7. (a) Sproul, G. D. J. Chem. Educ. 1993, 70, 531534. (b)
Sproul, G. D. J. Phys. Chem. 1994, 98, 66996703. (c) Sproul,
G. D. J. Phys. Chem. 1994, 98, 1322113224. (d) Sproul, G.
D. J. Chem. Educ. 2001, 78, 387390.
8. Wells, A. F. Structural Inorganic Chemistry, 5th ed., Claren-
don Press: Oxford, 1984.
9. Allen, L. C. J. Am. Chem. Soc. 1989, 111, 90039014.
10. Allen, L. C. J. Am. Chem. Soc. 1992, 114, 15101511 and
other articles.
11. Jensen, W. B. J. Chem. Educ. 1998, 75, 817828.
12. (a) Dehlinger, U. Z. Metallkunde 1934, 26, 227. (b) Laing,
M. A. Educ. Chem. 1993, 30, 160163.
13. Pauling, L. The Nature of the Chemical Bond, 3rd ed.; Cornell
University Press: Ithaca, NY, 1960; p 100.
14. Wilmshurst, J. K. J. Chem. Phys. 1959, 30, 561565.
15. Smith, D. W. J. Chem. Educ. 1990, 67, 559562.
16. Bratsch, S. G. J. Chem. Educ. 1988, 65, 223226.
17. von Szentpaly, L. J. Mol. Struct. 1991, 223, 7181.
18. Allen, L. C. J. Am. Chem. Soc. 1989, 111, 91159116.
19. Mann, J. B.; Meek, T. L.; Allen, L. C. J. Am. Chem. Soc. 2000,
122, 27802783.
20. Allen, L. C.; Capitani, J. F. J. Am. Chem. Soc. 1994, 116, 8810.
21. Phillips, J. C.; van Vechten, J. A. Phys. Rev. Lett. 1969, 22,
705.
22. Adams, D. M. Inorganic Solids; Wiley: New York, 1974; p110.
23. Pettifor, D. G. Bonding and Structure of Molecules and Solids;
Oxford University Press: Oxford, 1995; p 54.
24. Reed, A. E.; Weinstock, R. B.; Weinhold, F. J. Chem. Phys.
1985, 83, 735746.
25. Garner, L. D.; Meek, T. L.; Patrick, B. G. J. Mol. Struct.
(THEOCHEM) 2003, 620, 4347.
26. Greenwood, N. N.; Earnshaw, A. Chemistry of the Elements,
2nd ed.; Butterworth: Oxford, 1997; (a) p 555, (b) p 1177.
27. (a) Pitzer, K. S. Acc. Chem. Res. 1979, 12, 271276. (b) Pykko,
P.; Desclaux, J.-P. Acc. Chem. Res. 1979, 12, 276281.
28. Shriver, D. F.; Atkins, P. W. Inorganic Chemistry, 3rd ed.; Ox-
ford University Press: Oxford, 1999; p 362.
29. Fowler, P. W.; Tole, P. J. Chem. Soc, Chem. Commun. 1989,
16521654.
30. Goryunova, N. A. Chemistry of Diamondlike Semiconductors;
ChapmanHall: London, 1965; pp 116, 122.
31. Mann, J. B.; Meek, T. L.; Knight, E. G.; Capitani, J. F.; Allen,
L. C. J. Am. Chem. Soc. 2000, 122, 51325137.