Chemical Physics Letters 480 (2009) 127–131

Contents lists available at ScienceDirect

Chemical Physics Letters

journal homepage: www.elsevier.com/locate/cplett

Is the size of an atom determined by its ionization energy?

Hugo J. Bohórquez *, Russell J. Boyd
Department of Chemistry, Dalhousie University, Halifax, Nova Scotia, Canada B3H 4J3

a r t i c l e i n f o a b s t r a c t

Article history: We propose a general definition of the radius of an atom in terms of its ionization energy. A relationship
Received 26 February 2009 between these two fundamental properties is derived from the radial distribution function and the expo-
In final form 17 August 2009 nential decay of the electron density. Strong correlations with well-known atomic radii lead us to con-
Available online 19 August 2009
clude that we have found a universally valid definition of the atomic radius, defined exclusively in
terms of an intrinsic property, the ionization energy of an atom, which is equally available from experi-
ments and theoretical computations.
Ó 2009 Elsevier B.V. All rights reserved.

1. Introduction where x denotes the collection of electronic space and spin coordi-
nates. The summation over all spins is followed by an integration
According to Richard Feynman, if all scientific knowledge had to over the spatial coordinates of all the electrons but one (fermions
be reduced to one important statement it would be: ‘all things are are indistinguishable, which implies that the wave function W is
made of atoms’ [1]. Many properties, including atomic polarizabil- antisymmetric). Thus qðrÞ dr is the probability of finding an electron
ity [2], electronegativity [3], chemical hardness [4], diamagnetic in a volume element dr at a point r in position space. For an isolated
susceptibility [5] and atomic capacitance [6], among others, are atom the positively charged nucleus is the only attractor seen by
associated with the atomic size. Whereas the mass of an atom is the electron density. This attractive field has spherical symmetry
clearly defined and readily measured, the volume of an atom is and therefore it is convenient to use spherical polar coordinates
not uniquely defined. There is no consensus about how to estimate centered at the nucleus as the system of coordinates. The physically
atomic sizes. Several quantum-based sets of atomic radii, based meaningful electron density is spherical because an ensemble aver-
mainly on wave-function averages rather than in terms of experi- age must be taken over degenerate states. Therefore we study the
mentally-accessible observables, have been reported [7–15]. The spherical average, q
 ðrÞ, of the electron density given by
vast majority of the electron cloud is confined to a region within Z Z 2p Z p
a few angstroms of the nucleus, which suggests that there should 1 1
q ðrÞ ¼ qðrÞ dXr ¼ d/r sin hr dhqðrÞ: ð2Þ
be an effective radius r associated with the atomic size that charac- 4p 4p 0 0
terizes the extensive properties of the electron distribution. Both qðrÞ and q
 ðrÞ are normalized to the number of electrons,
In this communication we study the atomic radius in terms of a i.e.
fundamental observable of the system. In particular, we propose a
Z Z 1 Z 1
theoretical relationship between the atomic radius and the first
N¼ qðrÞ dr ¼  ðrÞr 2 dr ¼
4p q DðrÞ dr ð3Þ
ionization energy using the asymptotic limit of the exponential de- 0 0
cay of the electron density. An absolute scale for atomic radii
emerges that is defined in terms of the intrinsic properties of the with q ðrÞ being the probability density of finding an electron at a ra-
isolated atoms. We find a strong correlation between our set of dius r from the nucleus. DðrÞ ¼ 4pq  ðrÞr2 is the radial distribution
atomic radii and several other definitions. function which represents the probability density of finding an elec-
tron at a distance between r and r þ dr from the origin of the coor-
dinate system.
2. Methodology
The size of an atom is related to the probability of finding an
electron at a given distance from the nucleus as measured, e.g.,
The electron density distribution of an isolated atom is defined
by the radial distribution function, DðrÞ [16]. The radial distribution
by
XZ function of the ground state of the hydrogen atom provides a stan-
qðrÞ ¼ N ds0 fW ðxÞWðxÞg; ð1Þ dard definition of the atomic length unit, a0 . We can consider a
spins
generalization of this approach for obtaining an atomic radius
definition valid for any neutral atom. The maximum value in the
* Corresponding author. Fax: +1 902 494 1310. radial distribution function measures the statistical spread in the
E-mail address: [email protected] (H.J. Bohórquez). radial distances from the nucleus. The extreme value of the radial

0009-2614/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.cplett.2009.08.034
128 H.J. Bohórquez, R.J. Boyd / Chemical Physics Letters 480 (2009) 127–131

distribution function is given by the condition D0 ðrÞ ¼ 0, which

leads to the equation
1 q 0 ðrÞ
þ ¼ 0: ð4Þ
r 2q ðrÞ

It is well known that within the valence region the electron density
decays exponentially according to the theorem due to Hoffmann-
Ostenhoff and Hoffmann-Ostenhoff [28] and therefore
2
pffiffiffiffiffiffiffi
2me Ir
q ðrÞ / eh : ð5Þ

After replacing this expression in Eq. (4), a set of atomic radii can be
defined, hereafter denoted by r o , given exclusively in terms of the
first ionization energies, I, by b ro
35
rffiffiffiffiffi

Radial distribution function

E0 N(r)
r o ¼ a0 : ð6Þ 30
2I
25
2
Considering that the atomic unit of energy, E0 ¼ h =me a20 , is twice
the absolute value of the ionization energy of the electron in the 20

ground state of the hydrogen atom, IH , Eq. (6) can be reduced to a 15 D(r)
more simple relationship between the atomic size and the ioniza-
10
tion energy
rffiffiffiffiffi 5
IH
r o ¼ a0 : ð7Þ 0
I 0.001 0.010 0.100 1.000 10.000

Consequently, the radius of the hydrogen atom is the Bohr radius, Radial distance (au)
a0 , as expected. qffiffiffiffi Fig. 1. Strontium (1 S0 ) local momentum and radial distribution function. (a) Local

We must remark that the factor a10 E2I0 controls the exponential momentum term p
0 ðrÞ
~ðrÞ ¼  2h qq ðrÞ for Sr, at the numerical Hartree-Fock level of
4
decay of the electron density and hence its inverse can be consid- theory (log–log plot), with r 2 ½7:4  10 ; 16:74 au. The atomic shells are given by
the inflection points of p ~ðrÞ; the color code is accompanied by the electronic
ered directly a characteristic length of the electronic distribution in
population obtained for each shell. Z ¼ 38 and I ¼ 5:695 eV for strontium [30]. (b)
an atom. In addition, it is the limiting value for the local quantity Radial distribution function of the strontium (solid line) and cumulative electron
0
involved in Eq. (4), 2qqðrÞ ðrÞ
, which is identified as a local wave-number population NðrÞ (dotted line), with the radial distance in logarithmic scale, as
by Nagy and March [17], and a local component of the electronic obtained at the same level of theory. Within the radial distance r o ¼ 1:545 au
momentum operator, as studied by Ghosh [19,20], Cohen [21], (purple vertical dashed line), about 87% of the electronic population is enclosed
ðNðro Þ ’ 33eÞ.
Luo [22,23], Bohórquez and Boyd [24], and by Hamilton et al.
0 ðrÞ
[25], and is given by p ~ðrÞ ¼  2h qq ðrÞ . This local quantity provides
several important atomic properties [18], and for example its
inflexion points provide the boundaries for the electron shells, and increases going down a group. An experimentally-based atom-
whose electronic populations correctly obey the Aufbau principle, ic radii table has the advantage of including all the possible elec-
as computed at the Hartree-Fock level [24]. An illustration of the tronic effects affecting the valence electrons, which are not
local momentum appears in Fig. 1a, for the strontium atom typically considered in theoretical radii computations, such as rel-
(Z ¼ 38). The colors show the five electron shells of this atom, as ativistic effects, for example. Relativistic effects, which become
provided by the inflection points in p ~, i.e. those points where the important for heavy elements, have a substantial influence on
second derivative changes sign, p ~00 ¼ 0. For the core electrons, it the electronic structure of lanthanoids and actiniods, in particular
is found that p ~core ¼ 37:995p0 , in accordance with Kato’s cusp con- [27]. Their respective atomic radii decrease from periods 6 to 7 (for
dition [26]. The outermost shell (in cyan) exhibits the asymptotic all the elements from Ce to Yb and Th to No), as can be seen in
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
limit p ~val ¼ 2me ISr ¼ 0:647p0 . Table 1.
Fig. 1b shows the radial distribution function DðrÞ, the cumula- The values provided by Eq. (6) are assigned unambiguously to
tive electronic population NðrÞ and the location of the experimen- each isolated atom and therefore it is convenient to compare them
tal atomic radius, ro ¼ 1:545 au, from Eq. (7), according to the with other atomic radii definitions of the same nature. If these val-
experimental ionization energy (ISr ¼ 0:2094) as reported by NIST ues and the one introduced here are consistently measuring the
[30]. The computations for these graphics were done at the numer- same property (i.e. the atomic size of isolated neutral atoms), lin-
ical Hartree-Fock level with the NUMOL package from Becke and ear relationships are expected. Two categories of non-covalent
Dickson [29]. quantum-based radii were selected that include the radii for the
first 54 elements of the periodic table: (a) those based on the max-
imum electron density in the outermost orbital: Froese-Fischer [8],
3. Results and discussion Clementi et al. [9], Waber and Cromer [10] and Ghanty and Ghosh
[15], and (b) those based on the total electron density: Zhang et al.
The main equation derived in the previous section, i.e. Eq. (6), [11] and Boyd [12].
readily provides experimentally-based atomic radii. The experi- It is apparent from Fig. 2 that all data sets show similar periodic
mentally-derived atomic radii r o for the first 104 elements, as de- trends. In order to facilitate a comparison of the data sets, all sets
rived from the respective experimental ionization energies I (NIST were scaled to map the range [0, 1]. Thus, the respective radii were
database [30]), are shown in Table 1. The atomic radii in Table 1 transformed according to the linear relation rstd ¼ ðr  r He Þ=
show the very well-known periodic trends for the first five peri- ðr Rb  rHe Þ, where r He and r Rb are the minimum and maximum radii,
ods: the atomic radius decreases from left to right across a period, respectively, for Z 6 54.
 H.J. Bohórquez, R.J. Boyd / Chemical Physics Letters 480 (2009) 127–131 129

Table 1
Experimentally-based atomic radii, r o (au), for the first 104 elements as obtained from the experimental ionization energies [30].

H He
1.00 0.74
Li Be B C N O F Ne
1.59 1.21 1.28 1.10 0.97 1.00 0.88 0.79
Na Mg Al Si P S Cl Ar
1.63 1.33 1.51 1.29 1.14 1.15 1.02 0.93
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
1.77 1.49 1.44 1.41 1.42 1.42 1.35 1.31 1.31 1.33 1.33 1.20 1.51 1.31 1.18 1.18 1.07 0.99
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
1.80 1.55 1.48 1.43 1.42 1.38 1.37 1.36 1.35 1.28 1.34 1.23 1.53 1.36 1.26 1.23 1.14 1.06
Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
1.87 1.62 1.56 1.41 1.34 1.31 1.32 1.27 1.23 1.23 1.21 1.14 1.49 1.35 1.37 1.27 1.21 1.12
Fr Ra Ac Rf
1.83 1.61 1.62 1.51

Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
1.57 1.58 1.57 1.56 1.55 1.55 1.49 1.52 1.51 1.50 1.49 1.48 1.47 1.58
Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr
1.47 1.52 1.48 1.47 1.50 1.51 1.51 1.48 1.47 1.46 1.45 1.44 1.43 1.67

Fig. 2. Scaled atomic radii versus the atomic number for the first five rows of the periodic table. Six different theoretical definitions are plotted (open symbols): Froese-Fischer
[8] (squares), Clementi et al. [9] (crosses), Waber and Cromer [10] (diamonds), Zhang et al. [11] (circles), Boyd [12] (triangles), Ghanty and Ghosh [15] (stars), and the present
work (solid circles connected by lines).

The six sets of theoretical radii are plotted against the experi- atom, Ghanty and Ghosh proposed a similar idea for assigning
mentally-based radii (r o ) (from Table 1) in Fig. 3. The least-squares atomic sizes from the Kohn-Sham orbitals. For the valence elec-
linear fits (see Table 2) are shown in each case. The Pd radius re- trons, they reported the HOMO radii (used in present study),
ported by Waber and Cromer [10] is clearly an outlier, and is omit- whose energy is related to the ionization energy by Koopmans’
ted from the fit. All sets of atomic radii exhibit a correlation R > 0:9 theorem.
with respect to ro . This means that these atomic radii can be The simultaneous linear correlations between the proposed
approximated by the linear fits shown in Table 2. atomic radii and several other sets of radii confirm the hypothesis
Interestingly, the r o values exhibit the highest correlations stated in this Letter, i.e. that Eq. (6) measures the atomic size as
with two radius definitions indirectly involving the ionization well as any of the other definitions, with the practical advantage
energy in their computations: Zhang et al. [11] (R ¼ 0:96) and of being easily obtained from experimental ionization energies.
Ghanty and Ghosh [15] (R ¼ 0:96). Both define the atomic radius In addition to the strong correlations noted above, we want to
as the distance corresponding to the classical turning point emphasize that the method introduced herein is universally valid
of the valence electron. While Zhang et al. solves the and can be applied to both experimental and theoretical ionization
equation VðrÞ ¼ I, where VðrÞ is the potential energy of the energies.
130 H.J. Bohórquez, R.J. Boyd / Chemical Physics Letters 480 (2009) 127–131

1.5
4. Conclusions
Ghanty & Ghosh

1.0

ro (Å)
The atomic radius introduced in this Letter (Eq. (6) and Table 1)
is a quantum definition that exhibits several clear advantages over

0.5
previous ones, such as:

0.0
5.0

4.0

3.0

2.0

1.0

0.0
r (Å) 1. It can be computed from ionization energies derived from dif-
ferent theoretical methods: Koopmans’ theorem, the energy dif-

1.5
ference (I ¼ Eion  Eatom ) or from the numerical asymptotic limit
 0 ðrÞ
of the ratio qq ðrÞ .
Boyd (NS)

1.0
ro (Å)
2. It can be computed from experimentally-derived ionization

0.5
energies, providing the atomic radius for the whole periodic
table (Table 1).

0.0
3. It exhibits the expected periodic trends (as shown in Fig. 2)),
5.0

4.0

3.0

2.0

1.0

0.0
r (Å) including the lanthanoids and actiniods shell contraction.
4. It provides an unambiguous relationship between atomic sizes

0.0 0.5 1.0 1.5

and ionization energies (Eq. (6)). Early reports of the relation-
ship between the ionization energy I and the atomic radius r
Zhang et al.

ro (Å)
are based on Bohr’s model of the hydrogen atom [31,32], which
is often invoked to explain experimental observations. In these
pre-quantum approaches the ionization energy is considered to
be proportional to the electron-nucleus point charge interaction
5.0

4.0

3.0

2.0

1.0

0.0

r (Å) and therefore I / 1=r. The validity of this model is limited to

single-electron systems, and its theoretical extension to other
neutral atoms is not clearly justified. Nevertheless, this classical
1.5
Waber & Cromer

approach underlies subsequent atomic radius-ionization energy

1.0

studies [33–35]. Application of the local quantum theory to

ro (Å)

atomic systems, however, supports the concise relationship

0.5

I=IH ¼ a20 =r2o , as given by Eq. (6).

0.0
5.0

4.0

3.0

2.0

1.0

0.0

r (Å) For the derivation of the atomic radius introduced here, no ex-
plicit assumptions about trajectories or orbital-like behavior of the
valence electrons are required. In this sense, the radius ro provided
1.5
Clementi et al.

by Eq. (6) is a characteristic length associated with the ground-

1.0

state electron density of an atom. Given that it was derived from

ro (Å)

a maximal condition on the radial distribution, it is an optimal sta-

0.5

tistical estimate of the atomic extensive properties, and hence a

0.0

universally valid definition of atomic radius.

5.0

4.0

3.0

2.0

1.0

0.0

r (Å)
Acknowledgements
1.5

Gavin Heverly-Coulson, Erin R. Johnson, Axel D. Becke, Chérif F.

1.0
Froese

ro (Å)

Matta, and Alfonso Leyva are gratefully acknowledged for helpful

0.5

discussions and suggestions. Thanks to the Natural Sciences and

Engineering Research Council of Canada for financial support.
0.0

Computational facilities are provided in part by ACEnet, the regio-

5.0

4.0

3.0

2.0

1.0

0.0

nal high performance computing consortium for universities in

r (Å)

Atlantic Canada. ACEnet is funded by the Canada Foundation for

Fig. 3. Plots of six sets of theoretical radii versus the experimentally-based radii Innovation (CFI), the Atlantic Canada Opportunities Agency
proposed herein, r o . From left to right: Froese-Fischer [8], Clementi et al. [9], Waber
and Cromer [10], Zhang et al. [11], Boyd (non-scaled) [12] and Ghanty and Ghosh
(ACOA), and the provinces of Newfoundland and Labrador, Nova
[15]. The corresponding linear fits are plotted as red lines; their respective Scotia, and New Brunswick.
parameters appear in Table 2.

References
Table 2
Linear fit (r theor ¼ ar o þ r b ) between six sets of theoretical radii and the experimen-
pffiffiffiffiffiffiffiffiffiffiffiffi [1] R.P. Feynman, R. Leighton, M. Sands, The Feynman Lectures on Physics,
tally-based atomic radius, r o ¼ a0 E0 =2I. The radii are given in angstroms (Å) and R Addison-Wesley, 1964 (Chapter I, p. 1).
denotes the correlation coefficient. [2] J.K. Nagle, J. Am. Chem. Soc. 112 (12) (1990) 4741.
[3] D.C. Ghosh, K. Gupta, J. Theor. Comput. Chem. 5 (4) (2006) 895.
Method a rb R [4] P. Politzer, P. Jin, J.S. Murray, J. Chem. Phys. 117 (18) (2002) 8197.
(Å) [5] M. Putz, N. Russo, E. Sicilia, J. Phys. Chem. A 107 (28) (2003) 5461.
[6] J.C. Ellenbogen, Phys. Rev. A 74 (3) (2006) 034501.
Froese-Fischer [8] 4.416 1.404 0.95
[7] J.C. Slater, J. Chem. Phys. 41 (1964) 3199.
Clementi et al. [9] 4.089 1.401 0.90 [8] C. Froese-Fischer, Atomic Data 4 (301) (1972) 87.
Waber and Cromer [10] 3.428 1.166 0.95a [9] E. Clementi, D.L. Raimondi, W. Reinhard, J. Chem. Phys. 47 (4) (1967) 1300.
Zhang et al. [11] 4.871 1.185 0.96 [10] J.T. Waber, D.T. Cromer, J. Chem. Phys. 42 (12) (1965) 4116.
Boyd [12] 4.424 0.206 0.92 [11] M.B. Zhang, D.X. Zhao, Z.Z. Yang, J. Theor. Comput. Chem. 4 (1) (2005) 281.
Ghanty and Ghosh [15] 5.418 1.742 0.97 [12] R.J. Boyd, J. Phys. B 10 (12) (1977) 2283.
a
[13] R.F.W. Bader, W.H. Henneker, P.E. Cade, J. Chem. Phys. 46 (9) (1967) 3341.
For this linear fit the value corresponding to Pd was excluded. [14] D.C. Ghosh, R. Biswas, Int. J. Mol. Sci. 3 (2) (2002) 87.
 H.J. Bohórquez, R.J. Boyd / Chemical Physics Letters 480 (2009) 127–131 131

[15] T.K. Ghanty, S.K. Ghosh, J. Phys. Chem. (43) (1996) 17429. [27] P. Pyykko, Chem. Rev. 88 (1988) 563.
[16] R.J. Boyd, Nature 250 (5467) (1974) 566. [28] M. Hoffmann-Ostenhof, T. Hoffmann-Ostenhof, Phys. Rev. A 16 (5) (1977)
[17] A. Nagy, N.H. March, Mol. Phys. 90 (2) (1997) 271. 1782.
[18] M. Kohout, A. Savin, H. Preuss, J. Chem. Phys. 95 (3) (1991) 1928. [29] A.D. Becke, R.M. Dickson, J. Chem. Phys. 92 (6) (1990) 3610.
[19] S.K. Ghosh, B.M. Deb, Phys. Rep. 92 (1) (1982) 1. [30] W. Martin, A. Musgrove, S. Kotochigova, J. Sansonetti, Ground Levels and
[20] S.K. Ghosh, Curr. Sci. 52 (16) (1983) 769. Ionization Energies for The Neutral Atoms, vol. Physical Reference Data, NIST,
[21] L. Cohen, Phys. Lett. A 212 (6) (1996) 315. 2003, <http://physics.nist.gov>.
[22] S.L. Luo, Int. J. Theor. Phys. 41 (9) (2002) 1713. [31] A.S. Eve, Nature 107 (1921) 552.
[23] S. Luo, J. Phys. A 35 (25) (2002) 5181. [32] M. Saha, Nature 107 (1921) 682.
[24] H.J. Bohórquez, R.J. Boyd, J. Chem. Phys. 129 (2) (2008) 024110. [33] L.H. Ahrens, Geochim. Cosmochim. Acta 2 (3) (1952) 155.
[25] I.P. Hamilton, R.A. Mosna, L.D. Site, Theor. Chem. Acc. 118 (2) (2007) 407. [34] R.S. Roy, Spectochim. Acta B 27 (8) (1972) 323.
[26] W.A. Kato, Commun. Pure. Appl. Math. 10 (2) (1957) 151. [35] S. Tamura, J. Mater. Sci. Lett. 15 (19) (1996) 1678.