A New Electroaffinity Scale; Together with Data on Valence States and on

Valence Ionization Potentials and Electron Affinities

Robert S. Mulliken

Citation: J. Chem. Phys. 2, 782 (1934); doi: 10.1063/1.1749394

View online: http://dx.doi.org/10.1063/1.1749394
View Table of Contents: http://jcp.aip.org/resource/1/JCPSA6/v2/i11
Published by the American Institute of Physics.

Additional information on J. Chem. Phys.

Journal Homepage: http://jcp.aip.org/
Journal Information: http://jcp.aip.org/about/about_the_journal
Top downloads: http://jcp.aip.org/features/most_downloaded
Information for Authors: http://jcp.aip.org/authors

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
NOVEMBER, 1934 JOURNAL OF CHEMICAL PHYSICS VOLUME 2

ANew Electroaffinity Scale; Together with Data on Valence States and on Valence
Ionization Potentials and Electron Affinities
ROBERT S. MULLIKEN, Ryerson Physical Laboratory, University of Chicago
(Received May 14, 1934)

A new approximate "absolute" scale of electro negativity, scale (like the others) is rather largely empirical, especially
or electroaffinity, is set up. The absolute electroaffinity on as to its quantitative validity; and it remains to be seen
this scale is equal to the average of ionization potential and whether or not the latter will be more than very rough
electron affinity. These quantities must, however, in when tested for a wider range of cases. Nevertheless the
general, be calculated not in the ordinary way, but for new scale has a degree of theoretical background and
suitable "valence states" of the positive and negative ion.' foundation which throws some new light on the physical
Also, the electro affinity of an atom has different values for meaning of the concept of electronegativity (or electro-
different values of its valence; in general the electroaffinity affinity). The basis of the present scale, it should be men-
as here calculated (in agreement with chemical facts) is tioned, is simpler and more certain for univalent than for
larger for higher valences. Electroaffinity values have been polyvalent atoms.-The nature of valence states of atoms
calculated here for H, Li, B, C, N, 0, F, Cl, Br, I. They is briefly discussed. It is hoped that the tabulations of
show good agreement in known cases with Pauling's atomic valence state energies and valence state ionization
electronegativity scale based on thermal data, and with potentials and electron affinities given at the end of this
the dipole moment scale. The present electronegativity paper may be useful in problems of molecular structure.

A. UNIVALENT ATOMS the two atoms have the same degree of eiec-
tronegativity," then "the terms corresponding to
O NEchemical
of the most familiar and useful of
concepts is that of relative A+B- and A-B+ will occur with the same
coefficient"; that is, a= {3. For a highly polar
electronegativity. The physical basis of this has
remained obscure, although it has been evident molecule with A positive, a exceeds 'Y and {3.
that the electro negativity of an atom must be Starting with the above idea, Pauling has
related somehow to its electron affinity or its succeeded in setting up an empirical scale of
ionization potential, or both. Recently two approximate relative electronegativities on the
empirical scales of relative eiectronegativity have basis of thermal data interpreted with the help
been set up. In the present paper, a possible of a simple quantum-mechanical argument. 2 The
third, absolute, scale is discussed. consistency of observed thermal data with
In general, the wave function of a molecule relations demanded by the scale is rather good,
AB, where A and B are both univalent, can be indicating that a fairly definite electronegativity
approximated by forming a linear combination of value can be assigned to each element, as one
a wave function, which may be symbolized by would indeed expect from general chemical
1/;(A - B), corresponding to a state with a Pauling- experience. In connection with these and other
Slater electron-pair bond, with two wave func- electronegativity values, however, it should be
tions corresponding to ionic states A+B- and bome in mind that we do not know electronega-
A-B+. That is, approximately,!' 2 tivity to be a concept capable of exact quanti-
tative definition.
(1) We may now inquire whether there is any way
of determining theoretically the conditions under
If a= {3, we have a bond which, although it
which a={3 is to be expected in Eq. (1), other
contains polar terms, is almost as purely nonpolar
as any bond can be, and is a typical covalent than when A=B. Let us suppose we have a case
where the energy required to produce A++B-
bond. 2 If A and B are identical, a= {3 is obviously
from A + B is the same as that required to pro-
necessary; here 'Y is much larger than a or {3.I For
duce A-+B+. Further, let us imagine that A+
A not identical with B, Pauling states2 that "if
and B- approach each other without deforma-
1 Cf. J. C. Slater, Phys. Rev. 35, 514-5 (1930) for the tion, until the distance between their nuclei is
case of H 2 • Also S. Weinbaum, J. Chern. Phys. 1, 593 equal to the value re which occurs in the actual
(1933).
2 Cf. L. Pauling, J. Am. Chern. Soc. 54, 3570 (1932). molecule AB; and the same for A- and B+. If the
782

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
 NEW ELECTROAFFINITY SCALE 783

energy given out should be the same in the two [That the equality of the quantities (1 + E) /2 for
cases, then the net energy of the two imagined two atoms is an approximate criterion for their
forms A+B- and A-B+ would be the same. for equal participation in a chemical bond has
r= re (and nearly, if not exactly, so for other already been stated by Hund,3 presumably on
values of r). Then in the actual molecule, ac- the basis of an argument like that just given,
cording to quantum mechanics, both should although Hund does not go through this, nor
participate equally, i.e., a={3 in Eq. (1), pro- refer to the concept of electronegativity: d.
vided (see later paper in this journal) subsequent paper in later issue for further

f l/;(A-B)JIl/;(A+B-)dT
details.] According to Eq. (3), two univalent
atoms have equal electronegativity if the sum-
or average-of ionization potential plus electron
-E f l/;(A - B)1/;(A+B-)dT affinity is the same for each.
This suggests that we may define the absolute
= f l/;(A-B)JI1/;(A-B+)dT
electronegativity of an atom A approximately as
h + EA or, probably better for some purposes, as

-E f l/;(A-B)1/;(A-B+)dT,
(fA +EA)/2. Note that hand EA are really the
ionization potentials of A and A-, respectively.
The suggested definition carries with it the
a condition whose approximate fulfillment, for implication that the electro negativity of an atom
two atoms A and B of equal electronegativity, is approximately independent of its closeness of
seems plausible (E=actual energy of l/;AB). approach to another atom. For our definition is
Actually, we cannot expect the conditions based on the assumption that the net total
mentioned, in particular that of equal energy of energy, at r= re of AB, is nearly the same for
attraction of A++B- and A-+B+, to be met A++B- as for A-+B+; it is of course, certain
exactly. The equal energy condition would, of that this equality holds for large values of r
course, always be fulfilled if only the simple (Coulomb forces only).
inverse-square attraction of opposite charges Referring back to Eq. (1), one can readily see
were involved. To this is added, however, a qualitatively why EA and h are of equal im-
balancing repulsive force which may in general be portance for the electronegativity of an atom A.
expected to differ in the two cases. No simple The negativeness of atom A in 1f;AB is greater the
means of estimating the magnitudes of such larger the coefficient {3 and the smaller the
differences is evident, although crystal lattice coefficient a. (For atom B the relations are
data would have some bearing. These are, reversed.) Coefficient (3 is large if l/;(A-B+) is of
however, not available for the particular combi- low energy, and this is favored by large EA.
nations A+B- and A-B+ in which we are mainly Coefficient a is small if l/;(A+B-) is of high energy,
interested. and this is favored by large h. One sees then that
Let us, nevertheless, as a trial assumption to be large EA and large h both help to promote
tested empirically, suppose that this difference negativeness of atom A in AB. The fact that
between A+B- and A-B+ can be neglected, and precisely, the quantity (E A + fA) /2 is a good
that the other condition mentioned above is ful- measure of electronegativity is less obvious, but
filled. Then whenever we have will be shown empirically in the following, by
comparison of an (E A +h)/2 scale with other
(2)
empirically established scales of electronegativity.
we expect a= (3 in Eq. (1), so that A and B are of In a subsequent paper, the theoretical basis of
equal electro negativity according to Pauling's the various electronegativity scales will be gone
definition. Here f and E stand for ionization into more thoroughly.
potential and electron affinity, respectively; both The foregoing discussion indicates the neces-
may conveniently be expressed in volts. If (2) sity of considering both the 1f;(A+B-) and the
holds, then 3 F. Hund, Zeits. f. Physik 73, 1 (1931); d. especially
(3) pp. 18-19.

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
784 ROBERT S. MULLIKEN

TABLE I. Electroaffinities.

Atom Electroaff. Atom Electroaff.

and Abs. reI. to Elec. Mom. and Abs. reI. to Elec. Mom.
state electroaff. H(b-7.12) (e) +2.70 Thermal X 1018 state electroaff. H(b -7.12) (e) +2.70 Thermal X 10"
(a) (b) (e) (d) (e) (fJ (a) (b) (e) (d) (e) (fJ
H(s, V,) 7.12 0.00 0.00 0.00 0.00 B(s2P. VI) 3.75 ±0.3 -3.37 -1.25
F(P', VI) 12.66±0.2 5.54 2.05 2.00 - (sp', Va) 4.21±0.3 -2.91 -1.08
Cl(p', VI) 9.81±0.2 2.69 1.00 0.94 1.03 10.28±0.3s 3.16 1.17
BTtP'. VI) 9.06±0.2 1.94 0.72 0.7,<; 0.78 6.23±0.3q -0.89 -0.33
I(p'. VI) 8.23±O.3 1.11 0.41 0.40 0.38 -----~

Be(sp. V2) 3.06±0.2 -4.06 -1.51

10.02±0.5 2.90 1.07} 1.40 1.53 6.33±0.2s -0.79 -0.29
0lP', V,) 11.4 ±1:!I 4.3 1.6 (104° angle 4.69±0.21 -2.43 -0.90
inH,O)
Li(s. VI) 2.86±0.2s -4.26 -1.58
N(s'p a, Va) 7.40±0.4 0.28
0.10 }
[8.9?] :II [1.81] [0.661] s (0.43) 0.68 (if 92°
(sp', V,) 8.50±0.4 1.38 0.51 0.95 1.04 angle in
20.76±0.7s 13.64 5.05 (if 100° an- H2S)
12.58±0.5q 5.46 2.02 gle in P (0.10) 0.36 (if 98°
NH.) angle in
PH,)
C(s'p', V,) 5.38±0.3 -1.74 -0.64 As - 0.10 (if 96°
(sp', V.) 5.93±0.3 -1.19 -0.44 [ -0.28]p angle in
14.90±0.3s 7.78 2.88 [3.04]s AsH,)
8.17±0.3q 1.05 0,39 0.55 (0.5)

Notes for Table I. In column (b), the electroaffinities are column d agree with the thermal (presumably q) value in
all of the first stage type (one-half the energy difference column (e). It is thought that the column (e) values thus
between a singly-charged positive and negative ion each obtained are likely to be more trustworthy than the
in the same valence state VI_ I where I is as given in column column (d) values. The values in column (f) are from
(a», except those marked II. The latter are second-stage electric moment data and are in electric moment units,
electroaffinities (one-fourth the energy difference between based on compounds H m X;6 Malone's value of 1.32 for 0,
0++, Vo and 0-, Vo or N++, VI and N-, VI). The electron which was based on an assumed HOH angle of 90 0 in
added or removed is a p electron, except in the cases marked H 20, has here been corrected in accordance with Mecke's
s or g. For the s-p hybrid types g, the electroaffinities band spectrum determination of this angle as 1040 40'.
given-are obtained from a weighted mean of sand p values In column (d) the data of column (c) have been divided by
(1 : 3 weight for Sp3, 1 : 2 for Sp2 and sp" 1 : 1 for sp). 2.70 to assist comparison with columns (e) and (f). Values
This method is not quite right* for g in all cases, but in parentheses in columns (e) and (f) are uncertain; the
probably nearly enough so, since all values are rather value in column (e) for iodine also appears somewhat un-
rough. Necessary numerical data were taken from Table certain. All values for polyvalent elements in columns (b),
IV, except the electron affinity of H (0.715 volt), calcu- (c) and (d) are distinctly uncertain, mainly because of
lated quantum-mechanically by E. A. Hylleraas (Zeits. f. the uncertainty of the electron affinity data available for
Physik 65, 209 (1930». The values in column (e) are from these elements; rough estimates of uncertainty are given
Pauling's paper2 (units=vo!ts!), except the sand p values in column (b). Added in proof: If we use (e) +2.75 for
for carbon, given in brackets; these have been obtained by column (d), the values 2.01,0.98,0.70 and 0040 are obtained
applying to the sand p values in column (d) the same for F, Cl, Br, I, in close agreement with column (e).
correction (0.16) that is needed to make the g value in

f(A-B+) contributions in building up a good HCl is of the 1~;+ type; this is to be built partly
wave function for a moderately polar molecule, from H+Cl, partly from H++Cl- and from
e.g., HC!. Even in cases like NaC!, it may not H-+Cl+. For H++Cl-, there is no difficulty,
always be wise to neglect the f(Na-Cl+) term. Cl- being in a IS state which necessarily gives
The "absolute electronegativities" (1+ E) /2 12;+ on combination with H+. In the case of
for H, F, Cl, Br, I, and other atoms are given in H-+C!+, H- is ls2 , IS, but C!+, if we use its
Table I. Taking 1+ E for H, we get 14.25 volts. normal state, is S2p\ sp. Now SP(Cl+) +IS(H-)
For the other atoms, it is clear that the ordinary can give only triplet states' of HC!. [The rule that
minimum ionization potentials and electron 3A+IB cannot give lAB, but only 3AB, is no
affinities must be replaced by corrected values. longer strict for heavy atoms like I, but never-
This we see on considering Eq. (1) as applied, for theless still holds approximately there in the
example, to HC!. fAB for (the normal state of) sense that ""AB, which is lAB in the case we are
considering, should still contain only rather
* See reference 6. smaIl contributions from (3A+)+(lB-).]

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
 NEW ELECTROAFFINITY SCALE 785

With H-(IS) we must then use a singlet state different I and E values for the normal state
or states of CI+. There are two such states of low S2p5, 2Pl!; but the sum I + E is the same.J4a
energy, namely, ID and IS of S2p4. To calculate Besides the "absolute electronegativities" just
the proper I for CI, we should take a value obtained, the relative electronegativities referred
corresponding to some sort of average of the ID to hydrogen as zero are given in Table I, and are
and IS of CI+. seen to behave in a reasonable way. It may be
The nature of this average can be determined noted that if we had used the ordinary ionization
by observing that the four chlorine outer-shell p potentials of the halogens, this would not be the
electrons (3p) which are not used for valence case; for example, the value for iodine relative to
purposes in HCI are necessarily in the 3p7r hydrogen would be negative. Comparing the
condition (ml= ±l). The two electrons which relative electronegativities here obtained with
form the bond are in the u condition (ml= 0). those found by Pauling from thermal data, it is
These statements apply to all three of the terms· found (d. Table I, columns d and e) that the two
lft(H - Cl), lft(H+CI-), lft(H-C1+) which make the sets of values are proportional within the
principal contributions to lftHC I. The electronic uncertainties of the thermal-data scale and of the
structure of HCI may be described as I and E data. A similar agreement is found
1S22s22p63s23P7r4U2, all but the last two electrons (column f) with another scale of relative elec-
belonging definitely to the CI atom. In lft(H - Cl), tronegativities, based on electric moments. 5
one u electron is 3pu of CI, the other is 1s of H; in The agreements just found, even though
H+C1-, both u electrons are 3pu of CI-; in H-CI+, probably to some extent fortuitous, indicate that
both are 1s of H. We are now interested in the present "absolute" scale is at least roughly
H-CI+; we see that the proper state of CI+ for correct for univalent atoms. Granting this, the
union with H- is ........ 3P7r\ which is a fact that "absolute" electronegativies are at least
singlet state (1~+ in fact). Unlike ID and 15, roughly proportional to the quantity (I +E)
however, this is not a real state of uncombined gives new insight into the physical interpretation
C1+; nevertheless it has a meaning for combined of the concept of electro negativity.
CI+. If the present method of calculating absolute
It belongs to the category of "valence states" electronegativies is a good one, the scale can
(d. section C, below). It is a valence state of type readily be extended almost immediately to
S2p7r 4, Vo·(here Vo means zero valence in respect univalent electropositive atoms (Li, Na, K, ... ,
to homopolar bonding). Its energy will be found CUI, Ag, Au I, perhaps TIl, etc.). The necessary I
under the heading S27r 2 7r2 , VO in Table III and values are known, while the electron affinities are
S2P\ Va in Table IV. This has been estimated doubtless small enough so that even the un-
with the help of a method of Slater,4 and is found certain estimates which can be made of their
to be between the energies of ID and IS, as
4a At this Doint a question which might be raised will be
expected. Although there are some difficulties in answered. Why cannot we make use of excited triplet
applying the method (d. Table III), and states of H-, e.g., ls2s, 35, in combination with 3p of Cl+,
since 35 +3p is capable of giving 12;+ of HCI? The energy
although spectroscopic data on the ID and IS of 1.12.1, 35 and other triplet states should be only about
states are available only for F (not for CI, Br, I), 0.7 volt higher than that of 1.1 2 , IS, if we assume zero
electron affinity for nx, n> 1, in H-, and the total energy
it has been possible to estimate the desired I of 35 +3p would be considerably lower than that of (ls2, IS)
values for F, CI, Br, and I with an error which plus (3p', ID or IS). It is true that such terms should
contribute to 12;+ of HCI, but these contributions would
probably does not exceed ±O.S volt at most. probably be small compared with the others, for the
Using these corrected I values and the ordinary E reason that Cl+ (,P) would give rise to an electron con-
figuration 3p-rr33p(J" or 3P-rr 23p(J"2, instead of the required
values, we get the absolute electronegativities 3p-rr4 • The question just raised and answered for H- has
given in Table I. [Actually, I and E values for its analogue for Cl- and the others, although because of
the much larger electron affinity for formation of unexcited
the valence state S2p5, VI of the halogen, given in Cl-, etc., the (unstable) excited states are relatively much
Table IV, have been used, instead of the slightly higher in energy than for H-, and so need much less to
be considered.
6 H. M. Smallwood, Zeits. f. physik. Chemie B19, 242
• J. C. Slater, Phys. Rev. 34, 1293 (1929). J. H. Van (1932). J. G. Malone, J. Chern. Phys. 1, 197 (1933);
Vleck, Phys. Rev. 45, 405 (1934) has recently given a M. G. Malone and A. L. Ferguson, J. Chem. Phys. 2, 99
simpler method of !letting the energy formulas. (1934).

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
786 ROBERT S. MULLIKEN

values should give fairly satisfactory results (d. AO+A- as for AO-A+, and for A+O~A+ as for
Li in Table I, for which 1=5.37, E (est.)=0.34). A-O++A-, and if also
With this inclusion of electropositive atoms
(which give negative electronegativities relative f if;(A-O-A)JIif;(A-O+A-)dr
to H), the scale might perhaps better be described
as a scale of electroaffinity rather than of
electronegativity,-if we may return to a some-
"" f if;(A-O-A)Hif;(A-O-A+)dr,

what obsolescent word. Essentially correct results etc., then we might define !(Io+Eo) and IIo
might probably be expected for all univalent + EEo) /4, respectively, as the first-stage and
s-valent atoms if their electroaffinities are taken second-stage electroa.ffinities of the 0 atom.
proportional to their ionization potentials, with [Note that the energies of coming together of
the same proportionality constant as for Li. AO-A+, A+O~A+, etc., would be more compli-
. cated here than in the case earlier considered of a
B. POLYVALENT ATOMS molecule AB; there would be energies of inter-
~Te may next consider whether, and, if so, how, action between the two A + atoms, or the A and
the present scale of electroaffinities can be A+, for example; hence the likelihood of cor-
extended to polyvalent atoms. The two other rectness of the suppositions just made is harder
scales (thermal and electric-moment), which have to check than in the case of AB.]-A similar need
already been developed for a few polyvalent for considering both first- and second-stage
atoms (but only for the case of single bonds), can electroaffinities occurs, it may be noted, in
serve here as an empirical guide and check. double-bonded molecules, e.g. BeO.
As an example of a polyvalent atom, we may The values of 10 and 110 must, of course, be
consider oxygen. Instead of Eq. (1) we now have, chosen to correspond to the right "valence
for a compound AOB, or more simply AzO, the states" of 0+ and 0++, which are 0+(S2p3, VI) and
following (we neglect terms such as if;(A+OA-), 0++ (S2p2, Va), respectively. This is because in
if;(A-A, 0), etc.): O++H-+H, the 0 must be in a doublet state
(VI, intermediate between 2D and 2p of S2p3) ,
if;(A 20) =aif;(A-O-A)+bif;(A-O-A+)
since only in this way can the spins of 0+, H-,
+cif;(A - 0+A -) +dif;(A+O~A +) and H (respectively, !, 0, !) combine to give the
+eif;(A-O++A-). (4) zero resultant spin characteristic of H 20 in its
normal state. It may be noted that here the 0+
Suppose, for purposes of argument, that the
forms one homopolar bond (with the neutral H),
energy required to produce A+O++A- from
one ionic bond with the H+. In the case of 0++
A+O+A should be the same as to produce
+H-+H-, the 0 has to be in a singlet state
A+O-+A+, and to produce 2A++0~ the same
(Va, intermediate between ID and 15: d. the
as to produce 2A-+0++. [Actually, the co-
valence state of CI+ in H-Cl+, above), since both
existence of these two relations is not likely to be
H ions are in singlet states (zero spin). The 0-
fulfilled more than approximately, at best.]
and O~ ions in H+HO- and H+H+O~ are re-
Then we should have
spectively in the states S2p5, 2P= VI and S2p6,
Io-EA =IA -Eo. hence Io+Eo =IA+E A, 15= Va.
If one knew that band c in Eq. (4) were much
as in Eqs. (2), (3). Also we should have
larger than d and e, one might expect the first-
IIo-2EA =2h -EEo; and stage electronegativity to agree with Pauling's
(IIo+EEo)/2 =IA+E A, (5) scale value; or if d and e were much larger than c
and d, the second-stage electronegativity might
where 110 , EEo are the double ionization po- be expected to show such an agreement. If band
tential and electron affinity (O~O++, O~O~). If d should be larger than c and e, it is difficult to
the net energy of coming together, undeformed, say what might be expected. If b, c, d, e are all of
to give A - 0 distances equal to those in the since Sp4 has one s and two p valence electrons),
actual molecule A 20, should be the same for it would be possible to get agreement with the

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
 NEW ELECTROAFFI~ITY SCALE 787

comparable magnitude, some average of first- empirical values. On calculation, one finds that a
and second-stage electronegativities might be 56 percent participation of S2p3 and 44 percent
suitable. The following data suggest that the participation of Sp4 in the nitrogen valence would
actual situation may be of this nature. give a first-stage electronegativity agreeing with
Table I shows that the estimated first-stage the thermal value 0.95. Since such a relative
and second-stage electronegativities of 0 (whose participation of S2p3 and Sp4 appears not un-
values, it should be noted,-d. Table I,-are reasonable, we have some support for the
somewhat uncertain, because of uncertainty in attractive possibility that we may get nearly
the electron affinities) do not differ very greatly, correct absolute electronegativities by using just
and that their average agrees fairly well with the first-stage values.
electronegativities estimated from thermal and This idea gains further support, as we shall see,
electric moment data. This is encouraging. from the evidence on carbon. In the case of
For nitrogen we ought to consider a first-stage, oxygen, it is also not unreasonable. The calcu-
a second-stage, and a third-stage electronega- lated first-stage electronegativity is there lower
tivity, but it is impossible to estimate the than the empirical values, which might be
electron affinity for the process N -,;N= with any explained by a need of considering the second-
accuracy. The value for N -,;N~ is also a mere stage electronegativity too; but it seems quite
guess, and that for N -,;N- is decidedly uncertain. possible that a small participation of Sp5 instead
According to Table I, both first-stage and second- of exclusively S2p\ in the bivalence of oxygen,
stage electronegativities (although the value for might serve to account for the discrepancy on the
the latter is doubtful) give values which are basis of first-stage electronegativities alone.
definitely too low compared with thermal and Without denying that higher-stage electro-
electric moment data if, as is usual, we assume negatlvltles may be important and possibly
the N atom in its ordinary valence state (V 3 ) to needed for best results,-but noting that,
have the normal-state configuration S2p3. For the fortunately, they seem not to differ radically
ions N+ and N++, as is easily shown, we should from first-stage values,-we shall from now on
then use the valence states S2p2, V 2, and S2p, assume that approximately correct values of the
2P= VI, respectively. The valence state S2p2, V 2, true effective electronegativities of atoms can be
it may be noted (S2pXpy in Table II), is neither a obtained by using first-stage electronegativities.
triplet nor a singlet state, but really a mixture of Turning to carbon, we find that in its normal
both, with emphasis on the triplet, however.-It electron configuration S2p2 with valence two, this
appears, incidentally, that the third-stage elec- element should be decidedly electropositive, i.e.,
troT)egativity for S2p3, V 3, if we may extrapolate below hydrogen on the electronegativity scale.
from the first- and second-stage values, might Carbon with valence four, however (spa, V 4),
agree with the empirical data. according to our method, if we take the first-
TherE'lis, however, another possibility, namely, stage electronegativity averaged over the one s
that we ought to consider not the electronega- and three p valences, is electronegative relative
tivity of the S2p3 configuration, but something to hydrogen, the result (0.39±0.1) here obtained
intermediate between this and that of the Sp4 being in rather good agreement with the value
configuration. The sp4, like S2p3, gives only three (0.55) given by Pauling (d. Table I). The
valence bonds, but because of a partial mixing of process of averaging, with weight 1 : 3, the sand
S with p valence, can give stronger bonds. One p electroaffinities of the valence state Sp3, V 4
can calculate an s as well as a p electronegativity should probably be nearly equivalent to calcu-
for Sp4 by taking I and E; the first-stage sand p lating the first-stage electronegativity for the
electronegativities for Sp4 of nitrogen are given in Pauling-Slater tetrahedral or q type of orbital
Table I. It will be seen that by taking a suitably for the valence state q4, V4 (d. Van Vleck6
weighted mean of first-stage electronegativities
for S2p3 (here only p valence and electronegativity 6 J. H. Van Vleck, J. Chern. Phys. 2, 22 (1934). (Van

are possible) and for Sp4 (here we should take a Vleck considers a range of valence states of carbon from
9'. V, (complete s-p hybridization, !J. = t) to spa, V, (no
1 : 2 weighted mean of sand p electronegativities, hybridization, !J. = 1 in Van Vleck's Eq. (7).)

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
788 ROBERT S. MULLIKEN

for a discussion of the valence states Sp3, V4 and positive relative to hydrogen'; this would proba-
q4, V4). The good agreement just noted gives bly be expected in the radical CH 2 , whereas in
added support to the present method, and gives CH 4 the carbon is supposed to be relatively
considerable confidence that it may be extended negative.
to other cases, a few examples of which are given No attempt will be made here to make
in Table I. extensive predictions; the method by which this
In connection with the use of first-stage can be done for heats of formation of compounds
electronegativities for polyvalent atoms, a con- can be seen from Pauling's paper. 2 Very roughly
sideration of diatomic radicals may prove in- quantitative predictions could be made by using
structive. In the case of the single 0 - H bond the electronegativities in column (d) of Table I,
in the radical OH, for example, the proper which have been reduced to approximately the
electronegativity for the 0 atom is found to be same numerical values as Pauling's scale. Many
just the first-stage electronegativity, if we apply more predictions could be made by working out
the reasoning used above for the type AB; the the scale values for all the chemical elements. In
result is unaffected by the potential second many cases, however, this cannot yet be done
valence of O. If in H 20 each of the two single satisfactorily because of insufficient spectroscopic
bonds is very similar to the one bond in OH, as is data. Most of the scale values in Table I can be
generally supposed, then the first-stage elec- made more accurate when additional or more
tronegativity of 0 should be appropriate for 0 in reliable spectroscopic data and, in particular,
H 20 as well as in OH. A similar result should more accurate electron affinities become avail-
hold in other cases. (This would still be true even able. It must also be remembered that the scale is
if the actual A - B bond in a diatomic radical as yet largely empirical in character, and,
should differ from that in a saturated molecule. especially for polyvalent atoms, needs further
In CH, for instance, the actual C - H bond may investigation and testing from both theoretical
be formed almost wholly by a carbon p orbital and experimental standpoints.
whereas in CH 4 the C - H bonds may be con- A result which should hold quite generally in
structed with the help of q carbon orbitals. cases where valence can be varied in steps of two
Nevertheless, each C-H bond in CH 4 might by exciting inner electrons is that, for a given
closely resemble the hypothetical case of a C - H atom, electronegativity increases with increasing
single bond formed in C - H by a carbon q valence. In other words, the lowest valence is the
electron, justifying the use of a first-stage q most electropositive. This is perhaps nothing
electronegativity for carbon in CHd new; what is new here is the scale for measuring
A variety of applications of electronegativity these differences. For example, trivalent boron is
values calculated by the present method can be 0.33 unit electropositive relative to hydrogen in
made. Because of the intimate relation of Table I, column c, while monovalent boron, if
electroaffinity to thermal data and to electric it were stable, would be 1.25 units electropositive.
moments, as shown by the setting up of electro- A similar resul t is to be expected in other cases,
negativity scales on the basis of the latter, e.g., Tl, where both valences are actually
various predictions in regard to dipole moments, realized. Similarly divalent carbon is 0.64 unit
bond moments, and energies of formation might electropositive, tetravalent carbon 0.39 unit
be made from electroaffinities calculated by the electronegative; a similar difference is to be
present method. For instance, single molecules of . expected in Sn, where both valences are easily
LiI, BeO, and HF should have bonds of about realized. The difference here results chiefly from
equal polarity and also of about equal electric the higher electro negativity of s than of corre-
moment, if the parallelism of electric moments to sponding p valences. In more complex atoms, d
electronegativity differences continues to hold valences also enter. A calculation of electro-
with the more electropositive elements. LiH affinities for the various valence states of Cu,
molecules should be intermediate between HCl Ag, Au, Zn, Cd, Hg would be interesting.
and HF in polarity. If compounds with divalent although spectroscopic data are somewhat
carbon could be isolated, the carbon should be incomplete.

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
 NEW ELECTROAFFINITY SCALE 789
TABLE II. Interaction energies for atomic states belonging to configurations ls22sm2pn (d. also Table IIA).

State G, G2 State
I G,
I G2 State G,
---
G2
S2, '5= Vo 0 0 sp3 See Table IIA (s')p4, ap -4 -15
..
...
sp, ap -1 0 (s')pa, '5 -3 -15 'D -9
'p
V, -!
1 0
0
2D
2p
-6
0
'5
x2yz, V, .." 0
-13!
(S2)p, 2P= V, -1
---
0 xyz, Va .. -10! 11"'11"0", V2 . -13!
sP', .p -2
---
-5 11"11"0", V3 .. -9 q 2 7r7r, V2 " -12
'p 1 -5 x 2y, 0"211", } -3
11"211"2, x'y2, Vo " -6
2D -1 1 2
7r '1f, VI "
'5 -1 10 11"20", V, " -6 0"'11"', Vo " -9
---
sxy, S1I"0", Va -1 -3! sP'. .p -3 -15 Sp5, ap -3 -20
2p 0 -15 'P -1 "
S1I"1I", V3 -1 -2 2D -2 -9 sx2y2z, V,}
sx 2 , 2
$U , VI -1 4 25 " 0 S1I"211" 20", V 2 -2! "
S1I"', V, -1 1 sx'yz, V3 " -l3! S7r 27rU 2 , V2
--- ---
(S2)p',3P -2 -5 S1I"'1I"0", Va " -13! (S2)P5, 2P= V, -5 -20
---
'D " 1 SU 27r7r, Va " -12 Sp6, 25= V, -3 -30
---
'5 " 10 S1l"'1I"' } (S2)p', '5= Vo -6 -30
xy, 11"0", V, " -3! sx 2y'; V, " -6
11"11", V 2 " -2 S0"211"2, V, " -9
X2, 0"2, Vo " 4
11"', Vo " 1

Explanation of Table II: (a) The symbols s, p, x, y, z. 11". of the types which Slater calls J's and K's are given by
11"2, <Y, here respectively denote 2s, 2p, 2px, 2py, 2pz,
11"11", the following formula: F"(ls,1s)+aFO(2s,2s)+2mF"(ls,2s)
2P1l"{'2P+, or 2p_,), ip1l" +2PL, (2p1l" +)2 or (2PL)2, 2pO"(2po), + 2nF"(ls,2P) +mnFO(2s,2p) - mGO(1s,2s) - (n/3)G'(1s,2p)
while (S2) indicates either (2S)2 or nothing. The letter V +(3GO(2p,2p), where a=l if m=2, otherwise a=O, and
denotes a "valence state," whose valence is given by the where {3=0, 0, 1, 3, 6, 10, 15, for n=O, 1,2,3,4,5,6, re-
subscript. G, means (1/3)G'(2s, 2p), G2 means (1/2S)G' spectively.
Up. 2p) of Slater.' The numbers in the columns headed (e) The way in which {3, and the coefficients of G, and
G, and G2 give the numerical factor with which G, or G, G, in Table II, were obtained is illustrated by Table IIA.
appears in the energy expression for the state in question. The remaining F's and G's in the formula just given under
(b) Certain terms which are constam for anyone (b) correspond to J's and K's involving pairs of electrons
electron configuration have been omitted from Tables II not both of 2p type; in connection with them cf. Slater's
and IIA. The omitted terms which come from integrals tables, reference 4, p. 1312.

C. ApPENDIX: INTERACTION INTEGRALS AND by others,6a energy expressions are also tabulated
ENERGY DATA FOR CONFIGURATIONS Is22sm2pn for certain "valence states." The term valence
state, introduced by Van Vleck,6 describes a
In the following Table II are given the
certain hypothetical state of interaction of the
numerical factors with which certain interaction
electrons of an atomic electron configuration. In
integrals called J's and K's, or F's and G's, occur
formulating these, we shall here most often use
in Slater's method 4 of approximating the energies
the classification 2px, 2py, 2pz for the 2p orbitals
of the states of complex atoms. The total energy
of an atom in a given state is approximately (2p z== 2po, 2px = 2-!(2P+l + 2P_l),
equal to the sum of certain terms which are 2py = 2-!i-1 (2P+l- 2P_l».
constant for a given electron configuration, plus
others which vary from one state to another of Still other types of valence states, involving
the configuration. various hybrid orbitals formed by mixing 2s and
In addition to the actual atomic states associ- 2p, could be given, but are omitted from the
ated with the various electron configurations in 6a Slater, Condon and Shortley, in part, and recently
Table II, whose energies have already been given Van Vleck (Reference 4, Eqs. 22, 38).

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
 790 ROBERT S. MULLIKEN

TABLE IIA. Illustration of derivation of results in Table II. A "valence state" is an atom state chosen so as
to have as nearly as possible the same condition
State J qq J o" J"" K q" Kn G1 GO(2p, 2p) G2 of interaction of the atom's electrons with one
Sp3, 55 0 2 1 -2 -1 -3 3 -15 another as when the atom is part of a molecule.
35 -1 4 0 -1 0 1 " -15
3D 0 2 1 -1 0 -2 -6 The electrons in question may be classified as
'D 0 2 1 -1 0 0 -6 non-valence electrons and valence electrons. The
3p 1 0 2 0 -1 -2 0
Ip 1 0 2 0 -1 0 0 non-valence electrons are usually in pairs (e.g.,
sxyz, V 4 0 3 0 -H 0 -i! -10} 2px2 in the configuration ls22s22px22Py2pz) , in
S7r7rU, V 4 0 2 1 -1 -q -9
which case the ms values of the two electrons are
sx 2 y, SII27r, }
2 0 -1 0 -It -3
S7r 27r, V2 necessarily opposite in sign so that (d. Slater) the
S7r 2U, V2 0 2 1 -1 0 -1t -6
coefficient of the integral K(2px, 2px) is zero.
Explanation: Below each heading Jqq, etc., is given the
A valence state of an atom is one in which the
factor by which this integral is to be multiplied in getting latter's valence electrons behave toward one
the energy of the state in question; these numbers are de- another as if each were paired somehow with a
termined by Slater's method,4 modified in the case of the valence electron of a foreign atom, but not with
valence (unpaired) electrons in valence states in that - tK any valence electron of the given atom. Before
is put down for the K interaction of each two valence
going further, we note that this question of
electrons (see text). The symbols J qq , J"u, J .... , K"q, K .... ,
and G, ~ (1/3)Gl(2s, 2p) refer, respectively, to Slater's pairing of the electrons in a valence state of an
J(2po, 2po), J(2P±I, 2po), J(2P±1, 2P±1), K(2p±1, 2po), atom affects only the integrals K, and then only
K(2p±1, 2P±1), and K(2s, 2po, ±l), and respectively reduce of electrons not intra-atomically paired. The J's,
(see Slater's tables,' p. 1312) to F"(2p, 2p)+(4/25)F2 for a given detailed electron configuration
(2p, 2p); FO(2p, 2p) - (2/25)F'(2p, 2p); F"(2p, 2p)
+(1/25)F2(2P, 2p); (3/2S)G2(2p, 2p); (6/25)G2(2p, 2p);
specified in terms of ml values or of the x, y, z
(1/3)GI(2s, 2p), with F"(2p, 2p) = GO(2p, 2p) and F'(2p, 2p) classification (e.g., 2P 22py2pz or 2P+12p_l(2po)2),
X

=G2(2p, 2p) here (equivalem p electrons). By using these do not depend on the ms values of the electrons.
relations for the J's and K's, the results in the last three The correct contribution of' the K's t~ the
columns are obtained. The results in Table II were ob- energy for a valence state of a given atom is
tained in the same way. A simpler method for the ordinary
atomic states has recently been given by Van Vleck
obtained by writing -tK(i,j) for every combi-
(reference 4, Eqs. (22) and (38». Table IIA gives only nation (i,j) of the valence electrons of the given
those J's and K's (including G 1 =K(2s, 2p» which vary atom taken two at a time; this result, or essen-
for different states of Sp3. tially this, has been obtained by Slater and
others.b, 7 Any unpaired non-valence electron
which may be present can be treated for this
tables below; such states, often important for
purpose as a valence electron; this case of
atoms having both sand p valences, are discussed
unpaired non-valence electrons is of interest
by Van Vleck. 6
when we consider certain ionization processes,
It is evident that
e.g., removal of a non-valence electron from a
fluorine atom which belongs to an F 2 molecule.
The result just stated is based on the fact that
since in the x, y, z classification 2px, 2py, 2pz are the K interaction between two electrons i,j each
equivalent; these integrals may be denoted Jqq. in a definite orbital state (definite n, t, and ml or
I t is then easily shown 6b that J(2px, 2py) = J(2p,,;, n, t, and x, y, z, classification) is zero if the two
2pz) = J(2py, 2pz) = J(2po, 2P+l), which may be electrons have opposite ms values (+t, -t), but
denoted J"q; and that K(2px, 2py) =K(2px, 2pz) equal to - K(i,j) if both have the same ms
=K(2py, 2pz)=K(2po, 2P+l), which may be value. Now any unpaired electron in a valence
denoted K"o' Hence the integrals for electron state (including unpaired non-valence electrons)
configurations expressed in terms of 2px, 2py, 2pz is in a condition such that the probability that its
can be tabulated under the headings J qq , J"q, Ku
in Table II (d. Table IIA). 7 Cf. J. C. Slater, Phys. Rev. 38, 1109 (1931), discussion
of several electrons with only spin degeneracy. The results
given there can of course be used even if two or more
6b Cf. Slater, Reference 4, p. 1133. electrons are on the same atom.

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
 NEW ELECTROAFFINITY SCALE 791

TABLE III. Approximate energies of valence states in volts.

Energy Method oj estimation [N. B. In Energy Method oj estimation [N. B. In

(above many cases averages were used (above many cases averages were used
Stat, Atom or ion; normal (ef. e.g. "xy. V,) because State Atom or ion; normal (ef. e.g. "xy, V,) because
normal staie state Slater's relations (Table II) normal state state Slater's relations (Table II)
energy) do not hold well] energy) do not hold well]
All 0.00 "P', 'S: C- [1.83] Extr.
N 3.06 "P', 'D plus aver. of !('P-'D)
'p, V, [0.97] Extrapolated from Be, B+ and ('D-'S)/3
3.3.> 'p, 'P plus (lP-:-W)/4 0+ 4.29 Ditto
5.70 Dltto F++ 5.45
8.00
s'21r7tU, Va "P','S: N 1.69 "p', 'D minus aver. of ('D-'S)/3
xy, V, [2.52] Extr. from Be, B+ andWP-'D)
7.50 p', 'P plus ltG, (est. 6G,~6000) 2.35 Ditto
12.43 Same method (est. 6G2~8560)
1I'"2q, VI 0+ 3.32 Bame as 'D of "P'
x2, Vo [3.07] Extr. from Be, B+
(8.43) xy, V, plus 7tG, (see XU, V,) "P', 'S: C- [8.84] Extr.
(13.75) Ditto N (13.86) 'P', 'D-4!G" where HG,
~ (8922) est. from 0+, F++;
S2p,2P All 0.00 and 2D est. from 0+, F++
0+ 18.88 Ditto, with HG,~ (13,113) ~
,xy, V, [2.55] Extr. aver. of [tc'P-'P)-3('P
5.35 'D-4!G,; 9G2~ (8950) by extr. -'D)/4] andt('S-2D)
from C+ etc. F++ 23.84 Ditto, with 4!G,~(17,147) from
C+ 8.15 'D-4!G,; 9G,~17,780, ave. 'P, 'P, 'D and 0+
value from 4P, 2P, 2D, 28
N++ 10.88 Same method; 9G, ~ 25,790 "P', 'So C- [9.81] Extr.
0+++ 13.63 Same method; 9G2~32,990 N (15.69) 'P', 'D+3G" with G, as for
----------------------------
Be- [2.60] Extr. 0+ 21.57
8X'JyZ, Va
Ditto
B 6.28 'D+3G,; G, as for sxy. V, F++ 27.37
C+ 9.97 Ditto
N++ 13.52
0+++ 17.02
xy" V, "P, 'P: C+ (18.0) ,2p', 'P" N- [0.38] Extr.
(see note 0 0.67 Aver. of (ID-'P)/4 and (ID- 'P)
b) -H'S-ID)
F+ 0.95 Ditto
apo: B-
S'2p2, [0.28] Extr. Cl+ (0.74) From ("xyz, V,-"P','P,)
(see note C 0.49 Aver. of (lD-W)/4 and Br+ (0.78) =(V2- 3 Pcenter 01 oravitll)+
b) (lD-'P)/4-t('S-1D) I+ (0.92) ('P,.g.-'P,), with '1".0.
0.66 Ditto - 3P, ~ (3P, - 'Po)/3 ~ 0.04,
0.89 (0.17). (0.35) for Cl+, Br+, I+;
(1.07) and (V2-'P'.n.l est. roughly
by assuming ratio of this
[0.90] Extr. quantity to ionization po-
1.80 "p', lD plus aver. of (IS-1D)/3 tential of atom is same as for
and H'D-'P) F: e.g., forCI+, est. (V,-'P,.g.)
2.73 Ditto ~0.70~ (0.93/17.32)(12.96)
3.60
(4.43) ,'P', 'P,: 0 2.81 "p', ID plus aver. of HID-3P)
and (IS-ID)/3
[4.66] Extr. F+ 3.87 Ditto
8.16 'D+!G1-4!G,: !GI~ [10,460]; Cl+ (2.92) Rough estimates for Cl+, Br+, I+
6G2~ (11,200) from 'D, 'P and Br+ (2.79) made as for V2above
ext-r. I+ (2.69)
11.59 'D+!G1-4!G,; tGI ~ 13,644 and
6G,~ 16,040; ave. values from "p', 'p" N- [11.54] Extr.
'S 'D 'P 'S lD lp o 17.08 8P', 'P plus est. ('P-lP)/4 (est.
0++ 15.09 Sam~ ~eth~d; tal ~ 16,973 and ~15,120)
6G,~20,055 F+ 22.70 ,p', 'P plus ('P- 1 P)/4
F+++ 18.51 'D+!G1-4!G,; tGI~20,448 and (W-lP~75,574)
6G,~24,61O from'S, 'D, 'P, Ne++ 28.03 8P', 'P plus est. ('P- IP)/4
ID, lp and Extr. ("t.~22,470)

8x2y. V2 } [5.68] Extr.

=su2.w, VI? 9.89 3D+tGI+3G,; GI and G, as for
=811"2'11", VI? 8XYZ, V4
14.06 Ditto ,'x'y'z. VI~'P \ ,2p', 'Plj: F 0.02
18.19 = 8 2 7r 21r'2u. Vl=2P j CI 0.04
22.30 Br 0.15
I 0.31
x2yz, VI? "p', 'Po: N+ [27.4] Rough estimate
x'y', Vo
s2xyz, Va "P', 'S: C- [0.79] Extr.
N 1.33 Aver. of tc'D-'S) and ('D-'S)
-3('P-'D)/4
0+ 1.85 Ditto
F++ 2.35

Notes: (a) In the above table, values are given for most Zeits. f. Physik 84, 746 (1933) and I. S. Bowen, Phys.
but not all cases where the necessary data on atomic Rev. 45, 82 (1934).
energies are available either directly or with the help of (b) For S2p<, ap, corrections of H3P 2- 3P l) have been
fairly reliable interpolations or extrapolations. Data used included to take care of the fact that the component
in the calculations were taken from R. F. Bacher and S. 3P 2 is below the center of gravity of ap by that amount.
Goudsmit, Atomic Energy Slates; additional data on C+, For S2p2, ap, a correction of 2(ap 2-sP o)/3 has been applied
N+, N++, 0++, 0+++, and on F+, F++, F+++ from B. Edlen, to correct for the difference between 3P center of OTa01.ty and ap 0-.

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
 792 ROBERT S. MULLIKEN

TABLE IV. Some ionization potentials and electron affinities (volts). (Most values are only approximate. Valence states
given here are only xyz states: d. Table III; ionization potentials and electron affinities for removal of
electrons from valence states having CT, 71" quantization can easily be obtained, when needed,
with the help of Tables II, III. Removal of s electron is indicated by s in Ioniz.
pot. column; in all other cases a p electron is removed.)

State of atom State of pos. ion Ioniz. pot. State of atom State of pos. ion Ioniz. pot.
(or neg. ion) (or atom) (or el. aff.) (or neg. ion) (or atom) (or el. aff.)
Li-(s', '5 = Vol Li(s, '5 = V,) 0.34ts N-(s'p', V,) N(s'P'. V,) [ -7??]P'
Li(s, '5 = V,) Li+('5=Vo) 5.37s N(s'p', V,) N++(s'p, V,) 42.62*P'
Be-(s'p, 'P = V,) Be(s', '5) -0.57t O-(,'p', 'P =V,) O(s'p', 'Pt) 2.2a
(s'p, V,) (sP, V,) 2.78*s (s'P', V,) (s'P', V,) 2.87*
(sP', V,) (sp, V,) 0.18* O(s'p', 'P,) O+(s'p\ '5) 13.55
Be(s','5) Be+(s, '5=V,) 9.28s (S'P,': V,) (s'P', V,) 14.73*
(sP, V,) (s, V,) 5.93* (s'P', V,) 17.17*
(sp, V,) (P,'P=V,) 9.87*s (sP', V,) 31.76*s
(p', V,) (P, V,) 5.72*
O-(s2P",,'5 = Vol O(s'P', 'P,) [ _6.5]aap'
B-(s'p', 'P) B(s'p,'P=V,) 0.12t (S'P', V,) [-5.8*] P'
(s'P', Vol (s'P, V,) -0.78* O(s'P', 'P,) O++(s'P', 'P,) 48.48P'
(s'P', V,) (sp', V,) 5.19*s (S2p4, V2) (s'p', Vol 51.41*P'
(sp', V,) (sP', V,) -0.21*
B(s'p, 'P = V,) B+(s','5=Vo) 8.28 F-(s'P'.,,'5 = Vol F(s'p', ,p,!) 4.13b
(sP', V,) (sP, V,) 8.63* (s'p', 'P=V,) 4.15
(sP', V,) (P', V,) 15.36*s F(s'p', 'P,!) F+(S2p', 'P,) 17.32'
(s'p', 'f. = V,) (s'P', V,) 18.25'
C-(s'P'. '5) cts'P', 'P) 1.37t (s'p', Vol 21.17*
(s'P', V,) (s'P', V,) 0.03*
(s'p', V,) (sp', V,) 8.74*s Cl-(s'P", Vol Cl(s'p', 'p'!) 3.75 b
(sp', V,) (sP', V,) 0.69' (s'P', V,) 3.79'
C(s'P', 'P) C+(s'p, 'P = V,) 11.22 Cl(s'p', 'P,!) Cl+(s'p', 'P,) 12. 96.a
(s'p', V,) (s'p, V,) 10.73* (s'P', V,)
(s'P', V,) (sP', V,) 18.84*s
(S'P:: V,)
ts2p4, Yo) tt~~.a
(sp', V,) (sp', V,) 11.17'
(sP', V.) (pl, V,) (21.06)*s Br-(s't', Vol Br(s'p', 'p'!) 3.53b
(s'p', V,) 3.68'
N-(s'p', 'P) N(s'p','5) 0.04t Br(s'p', 'p'!) Br+(s'p', 'P,) 11.80 d
(s'p', V,) 0.99*
:U~:a
(S'P', V,) (s'P.'.. V,) (s'P', V,)
(s'p', V,) (sp', V,) 13.52*s (s'p',Vo)
(sP', V,) (sp', V,) 2.36'
N(s'p', '5) N+(s'p', 'P,) 14.48 I-(s'~:, Vol I(s'p', 'p'!) 3.22 b
(s'P', V,) (s'P', V,) 13.81' (s'P', V,) 3.53'
(s'p', V,) (sp', V.) 24.77*s I(s'p', 'P'l) I+(s'p', 'P,) 10.55' d
(sp', V,) (sp', V,) 12.24'
lU~:d
(S'P':, V,) (s'p', V,)
(sp', V,) (sp', V,) 14.63* (s'p', Vol
(sp', V,) (P', V,) [28.0*] s

Notes for Table IV. t indicates electron affinities taken electron affinities (t, a, aa, b) or ordinary ionization po-
from a paper by Glockler. 8 These are based on a method tentials (d. note c below) with valence state energy data
of extrapolation similar to one used by Bacher and given in Table III. In the case of the electron affinities,
Goudsmit (private communication), who consider the it should be noted that the data (all given in brackets [ ]
latter as reliable as any available. They point out, how- in Table III) were obtained by simple extrapolations
ever, that the method implies a power series approximation havin~ a similar basis to that used by Glockler for the
for ionization energy, which should become less convergent ordinary electron affinities. In case better values for
for negative ions (electron affinities). Unfortunately the the ordinary electron affinities should be determined in
available accurate electron affinity data are inadequate to future, the resulting correction needed for the ordinary
serve as a check on the accuracy of the above extrapola- electron affinity of any given atom in Table II I should
tion: for H, the extrapolated value agrees closely with also be approximately right for all the valence state
the accurate quantum-mechanically calculated value (0.715 electron affinities of the same atom.
volt); for 0 and F the extrapolated values are about 1 a is based on a fairly reasonable experimental value
volt lower than the empirical, but the empirical value 2.2±0.2 volts obtained by Lozier by electron impact
for 0 is uncertain, while the data used in the extrapolation methods.' This is probably preferable to the extrapolated
are uncertain in both cases. [In the preceding sentence, we value (d. bracketed comment in t note above).
are not using Glockler's extrapolated value +3.80 volts aa indicates a very rough electron affinity value (150
for the electron affinity of 0, but a value 1.24 volts. ±50 kcal.) based on crystal lattice constants.'O
Glockler's value is based on the dubious value 18.6 volts b indicates values based on crystal lattice data, and
for the ionization potential of F, while the value 1.24 is considered accurate to ±2 percent, given by Mayer and
obtained by Glockler's method if the value 17.32 volts Helmholtz.u It should, however, be noted that the value
estimated by the writer (d. note c below) is used for the for F may be less accurate, because it involves the heat of
ion. pot. of F.] It seems probable that the extrapolated dissociation D of F 2 ; if for instance the accepted value of
electron affinities of Glockler (except for 0 and perhaps a
few other cases where reliable data were not available) 8 G. Glockler, Phys. Rev. 46, 111 (1934).
are not in error by amounts greater than ±0.3 to ±1.0 9 W. W. Lozier, Phys. Rev. 46, 268 (1934).
volts. 10]. E. Mayer and Maltbie, Zeits. f. Physik 75, 748
(1932).
* indicates valence state electron affinities and ionization 11 ] . E. Mayer and L. Helmholtz, Zeits. f. Physik 75, 19
potentials which have been obtained by comi;Jining ordinary (1932).

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
 NEW ELECTROAFFINITY SCALE 793

D should be raised, the M. and H. value of the electron available data on X. and Am are used (d. e.g., Mulliken,
affinity of F would be raised by !D. The accepted value of Phys. Rev. 46, 549, 1934, Table lIB).
D is 2.80±0.03 volts, but in the wriler's opinion the • indicates for fluorine an estimate of the ionization
indicated probable error ±0.03 is far too small, and the potential made from data on the s'p43p, 4P, 'P states of
value 2.80 itself has no real foundation, ;although it may fluorine, using for the 3p term value an estimate based on
happen to be nearly correct. The value 2.80 volts (63.3 interpolation in the series N, 0, Ne, Na; the present
±0.07 kcal.) is based on a determination of the maximum estimate of the ionization potential is believed to be much
(X2900) of the ultraviolet continuous absorption of F 2 by better than Dingle's value 18.6 volts, which is obviously
von Wartenberg and Taylor, combined with an extrapola- too large. The estimated value 10.55 volts for iodine is
tion of the interval Xconv . -Xmax as observed in 12, Br2 based on interpolation and other comparisons between
and Cb, to F 2 (X conv . = wave-length of convergence of data on other atoms and is believed to be at least as well
vibrational levels, corresponding to dissociation; not ob- founded as other estimates in the literature. The ionization
served in F 2).12 The extrapolation used (linear variation of potentials given for all other atoms are spectroscopic
X.-Xm with atomic weight A) has no theoretical basis, values (d. Bacher and Goudsmit).
and might well be deceptive, since F occupies an excep- d indicates values for Cl+, Br+, 1+ which are based on a
tional position among the halogens; furthermore, the method of estimation (see Table III) which may give
supposed linear variation of X.-Xm is actually not even appreciable errors (probably not worse than ±0.5, how-
empirically fulfilled for 12, Br2, Ch when the most reliable ever). If we may judge by a comparison of F+, Cl+, Br+, 1+
with C, Si, Ge, Sn, the values 2.92, 2.79, 2.69 for s'x 2y'. Vo
in Table III, and so the ionization potentials V r -+- Vo in
12 H. von Wartenberg, G. Sprenger and J. Taylor,
Bodenstein-Festband, p. 61. (Erganzungsband of the Table IV, are about 0.35 volt too high for Cl+, Br+,
Zeits. f. physik. Chemie, Leipzig, 1931.) about 0.4 too low for 1+; but this is uncertain.

m. is parallel to that of any other electron in a by judicious averaging, etc. (d. Table III), it
similar condition in the same atom is just !. The seems probable that the final results given in
quantum-mechanical energy integral taken over a Table III should represent fairly good ap-
state in which parallel and antiparallel orienta- proximations to the desired valence state
tions of two ma's are equally probable is the energies.
average of the energy values 0 and - K for the In Table IV ionization potentials and electron
two cases considered separately. Hence -!K(i,j) affinities of a number of atoms are given, for
is the correct result. This argument, while both sand p electrons. Additional values can
perhaps not rigorous, is ~lluminating. readily be obtained, when needed, by making use
In Table III the energies of various valence of the excitation energies of valence states, given
states of a number of atoms and ions are esti- in Table III. Most of the ionization processes
mated numerically by the use of the energy given in Table IV correspond to removal of a
expressions in Table II taken in connection with valence electron so that if the atom is in state Vn.
available spectroscopic data. In many cases its positive ion is in state Vn - 1 • Similarly, many
the results obtained can be regarded as only of the electron affinities correspond to addition of
approximate for the well-known reason (among an electron in such a way as to go from an atom
others) that the observed states do not fit state Vn to a negative ion state Vn-l. This type of
Slater's formulas very well. For example, the ionization potentials and electron affinities is
observed ratio of the intervals (1S _1 D) / (lD _3P), what is needed for getting the electronegativity
which should be 3 : 2 for configuration S2p2 or values of Table I. The electron affinities, it
S2p4 according to Slater's formulas, is actually should be noted, are not accurate (probable
much nearer 1 : 1. In spite of this difficulty, errors ±O.3 up to ± 1 volt) except in the case
whose effects have, it is hoped, been minimized of the halogens: d. Table IV, notet.

Downloaded 03 Jun 2012 to 136.159.235.223. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions