See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/5373706

Covalent radii revisited

Article in Dalton Transactions · July 2008

DOI: 10.1039/b801115j · Source: PubMed

CITATIONS READS

3,293 7,523

8 authors, including:

Beatriz Cordero Verónica Gómez

University of Barcelona BADRINAS SAU
9 PUBLICATIONS 3,527 CITATIONS 27 PUBLICATIONS 3,836 CITATIONS

SEE PROFILE SEE PROFILE

Ana E Platero-Prats Jorge Echeverria

Argonne National Laboratory University of Barcelona
56 PUBLICATIONS 5,649 CITATIONS 50 PUBLICATIONS 4,748 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Verónica Gómez on 10 May 2014.

The user has requested enhancement of the downloaded file.

PAPER www.rsc.org/dalton | Dalton Transactions

Covalent radii revisited†‡

Beatriz Cordero, Verónica Gómez, Ana E. Platero-Prats, Marc Revés, Jorge Echeverrı́a, Eduard Cremades,
Flavia Barragán and Santiago Alvarez*
Received 21st January 2008, Accepted 6th March 2008
First published as an Advance Article on the web 7th April 2008
DOI: 10.1039/b801115j

A new set of covalent atomic radii has been deduced from crystallographic data for most of the
elements with atomic numbers up to 96. The proposed radii show a well behaved periodic dependence
that allows us to interpolate a few radii for elements for which structural data is lacking, notably the
noble gases. The proposed set of radii therefore fills most of the gaps and solves some inconsistencies in
currently used covalent radii. The transition metal and lanthanide contractions as well as the
differences in covalent atomic radii between low spin and high spin configurations in transition metals
are illustrated by the proposed radii set.

Introduction by Batsanov.12 A good illustration of the intrinsic uncertainty

associated with the empirical definition of an atomic radius for
The atomic radius is not—and cannot be—an intrinsic property of an element is provided by the large differences between ionic,
an element. At least for a chemist, who is interested in atoms within covalent and metallic radii for the same element. For instance, the
molecules rather than as isolated entities, because the electron Shannon and Prewitt cationic radius of an element7 is typically
density cloud associated with an atomic nucleus is modified in at least 0.5 Å shorter than its covalent radius, while the opposite
different ways, both in shape and in size, by the different atoms situation occurs with the anionic radii, showing how strongly the
bonded to it. Probably the most elegant and unequivocal definition effective radius of an atom depends on its electronegativity relative
of the shape of an atom is provided by Bader’s atoms in molecules to that of the atom to which it is bonded.
theory:1 the space enclosed within a zero flux surface of the electron What do we need atomic radii for? They should provide a rough
density around the nucleus. But in a molecule such a shape is hardly idea of the steric bulk of diverse molecular fragments and allow
spherical enough to allow us to define an atomic radius. Even if for a semiquantitative analysis of kinetic data in the search for
we approximate its shape to a sphere and define an average radius, steric effects. Different reactivity of the early and late lanthanides
it is clear that the size depends strongly on the bonds that the can be associated with the differences in atomic radii.13 Covalent
atom is forming to its neighbours. Therefore, there is little hope radii are also used in crystallographic software to indicate the
of defining a theoretically sound atomic radius that is valid for a existence of chemical bonds and the subsequent establishment
given element in a wide variety of molecules. Yet, the concept of of molecular structures, even if there are always cases in which
attaching an approximate size to the atoms of a given element is deciding whether there is or is not a chemical bond between two
deeply rooted in chemistry since Bragg’s proposal that interatomic atoms is ‘quicksand territory’. Finally, covalent radii are used
distances in crystal structures can be considered as sums of atomic in crystallographic software to provide reasonable ball and stick
radii.2 The idea was complemented by Pauling’s metallic radii,3 three-dimensional representations of molecules.
van der Waals radii proposed by Bondi,4 Slater’s covalent radii5 In spite of the widespread use of covalent radii in structural
and ionic radii proposed by Shannon and Prewitt.6,7 For more chemistry, there are many elements for which a covalent radius has
detailed discussions and bibliography on atomic radii the reader is not been established. Furthermore, the most commonly used radii,
referred to the classical book of Pauling,8 the more recent books of from the Cambridge Structural Database (CSD), do not show any
Mingos,9 Alcock10 and Butler and Harrod,11 and a recent review clear periodic trend and even present some inconsistencies within
a transition series or within a periodic group. In other instances,
Departament de Quı́mica Inorgànica and Institut de Quı́mica Teòrica i the radii are too short in comparison with experimental bond
Computacional, Universitat de Barcelona, Martı́ i Franqués 1-11, 08028, distances. Consider as an example the Ba–N bond distances: 98%
Barcelona, Spain of the more than 600 such bond distances collected in the CSD
† This contribution reports the result of a joint project of the graduate
course “Structural Databases in Chemistry” during the academic year are at least 0.6 Å longer than the sum of the radii for Ba and N
2006–2007. (Fig. 1). Similar results can be found for the Ba–O bond distances
‡ Electronic supplementary information (ESI) available: A data sheet with as well as for other alkaline and alkaline earth elements. But for
the proposed covalent atomic radii including e.s.d.s, atomic numbers, atomic radii to be a useful tool they must be able to give expected
atomic symbols, Group and period of the first 96 elements, together
with CSD covalent radii, Pauling metallic radii and van der Waals radii; bond distances within one or two tenths of an ångström of the
scatterplot showing the effect of the choice of different reference elements mean experimental values. The very definition of elemental radii
on the covalent radii; statistical data of our comparison of experimental in the CSD14 indicates that the radii adopted are considered as
bond distances with the sum of atomic radii for a sample of element provisional: “The covalent radii have been defined over the years
pairs and a comparison of the radii proposed for elements of Groups
14 to 17 with those obtained from the Mogul knowledge base. See DOI: to minimise the need for editorial changes for specific structures.
10.1039/b801115j Elements not yet encountered in the CSD are assigned a radius

2832 | Dalton Trans., 2008, 2832–2838 This journal is © The Royal Society of Chemistry 2008
 To obtain a covalent radius for carbon we searched for Csp3 –
Csp3 single bonds, adopting criteria similar to those used for
nitrogen, only we could in this case restrict the agreement factor to
5% and limit the search to purely organic compounds, resulting in a
radius of 0.762(11) Å for a sample of 10 000 structural data.16 Bond
distances to carbon are affected by differences in hybridization,
as discussed by Alcock,10 and covalent radii for Csp2 and Csp
atoms using the same criteria have been found to be 0.727(20) and
0.689(6) Å, respectively. Even if one might expect pathological
cases, such as cyclopropanes or cyclobutanes, to affect the deduced
carbon radius, a search restricted to acyclic C–C bonds gives
practically the same covalent radius, 0.764(10) Å. To test the effect
Fig. 1 Distribution of the difference between the experimental Ba–N
of choosing carbon or nitrogen to deduce covalent radii for other
bond distances and the sum of the CSD covalent atomic radii.
elements, we obtained the covalent radius of carbon also from C–
N bonds with similar criteria and found a value of 0.754(16) Å,
of 1.50 Å.” The CSD team is well aware of these limitations, and indicating that the choice of the reference atom affects the derived
makes use of the set of covalent radii with a wide tolerance margin, radii by only about one hundreth of an ångström, well within one
combined with a series of tests and internal software to sensibly standard deviation. The covalent radius for O was derived from a
decide on the existence of bonding between each pair of atoms in sample of 10 000 acyclic C–O bond distances, in a search limited to
a structure. purely organic compounds with two-coordinate oxygen and four-
In recent years we have addressed this problem for first row coordinate carbon atoms with R ≤ 5%, and the value obtained
transition metals by carrying out systematic searches in the CSD was 0.676(28) Å.
and establishing a new collection of empirical covalent radii for Before adopting the covalent radii proposed here for our
those elements.15 Those radii seemed to follow a sensible trend reference elements, C, N and O, we must compare them with those
and allowed us to better handle the delicate question of whether previously used. The radii deduced by us for these elements are
there is a chemical bond between two transition metal atoms nearly coincident with those proposed by Alcock10 (Table 1), and
in M2 X2 rhombuses in bridged dinuclear complexes with two show a linear dependence on the atomic number. The covalent
X bridging atoms. With these precedents, we planned a set of radius for carbon given by other authors is nearly coincident with
systematic studies of crystallographic data that could provide us the one deduced by us, but those for nitrogen and oxygen present
with a consistent set of empirical covalent radii, taking advantage a more pronounced variation, with a significantly larger radius for
of the exponential growth of the structural information available oxygen, while the CSD radii for these three elements have the same
in structural databases, whose results are presented in this paper. value (0.68 Å).
A large variety of amines act as Lewis bases toward a number
Methodology of acids, including transition metals, rare earths and main
group metallic elements, offering a large number of structurally
There is no single element that forms a significant number of characterized metal–nitrogen bonds. Besides, such bonds have
structurally characterized bonds to all elements of the periodic a higher degree of covalency than the corresponding metal–
table. For that reason we have defined covalent radii from oxygen or metal–halogen bonds. In addition, nitrogen also forms
experimental bond distances to N, C or O, according to criteria covalent bonds with practically all non-metallic elements. In
discussed below. We therefore start by defining the radii for these total, the version of the CSD used for this study holds 37 410
three elements. crystal structures with 131 837 crystallographically independent
The covalent radius for nitrogen was obtained from N–N bond N-element bonds for 80 elements, disregarding the ubiquitous
distances in substituted hydrazines (only H or C atoms bonded to elements carbon and hydrogen. For that reason, the covalent radii
N) with three-coordinated nitrogen atoms, excluding all structures for most elements were deduced from N–element bond distances,
with an R factor larger than 10% or presenting disorder or restricting the searches to three- or four-coordinate nitrogen
errors. The resulting value, 0.706(13) Å, is the average of 2200 atoms. For hydrogen, an additional constraint imposed was that
crystallographically independent data and was used throughout the structural data came from a neutron diffraction experiment.
with three decimal figures for calculating other radii from N– The exceptions are some main group elements (including the
element bond distances to minimize rounding-off errors. It must be metals of the Zn group), for which the N–element bond distances
kept in mind, though, that the third decimal figure is not significant
and the radii collected in Table 2 are shown with only two decimals. Table 1 Comparison of the reference covalent radii with those proposed
The use of a more restrictive criterion for the structural data (R ≤ by other authors (all values in Å, e.s.d.s in parentheses)
5%) did not result in a significant change in the atomic radius or in
the standard deviation of the sample. Conversely, by allowing any Author C N O Ref.
coordination number of the N atoms we obtained practically the This work 0.762(11) 0.706(13) 0.661(19)
same results, 0.708(12) Å, but now representing 3046 independent Alcock 0.767 0.702 0.659 10
bond lengths. From here on, all the estimated standard deviations Mingos 0.77 0.74 0.72 9
Butler and Harrod 0.77 0.75 0.73 11
(e.s.d.) given in parentheses together with the radii are e.s.d.s of
Wells 0.77 0.74 0.74 17
the sample.

This journal is © The Royal Society of Chemistry 2008 Dalton Trans., 2008, 2832–2838 | 2833
either did not provide a sufficiently large data set or showed too of O would result in somewhat larger radii for K and Rb. Taking
much dispersion. In those cases we deduced the corresponding C as reference would result in significantly larger radii for these
radii from E–C distances, where E is an element of Group 12, 13, elements. However, the periodic dependence of the radii would
14, 15 (except N), 16 or 17, or from E–O bond distances for E an not be qualitatively affected, since different reference elements
alkaline or alkaline earth metal, Xe, Pu or Am. It is worth noting change only the height of the maxima in Fig. 2, but not their
that the covalent radius for F deduced here, 0.57(3) Å, is practically positions. For transition metals, if we take carbon as the reference
coincident with that proposed by Gillespie and Robinson from the element, the resulting radii are in general slightly smaller than
analysis of a variety of covalent fluoro compounds (0.54 Å).18 those deduced from M–N bonds, with significant differences (of
Bond length data were obtained from version 5.28 of the CSD19 up to 0.3 Å) especially for the early transition metals (Groups 13–
(400 977 entries) plus two 2007 updates (25 175 entries). Except 15). For lanthanides, the use of C as the reference element gives also
when otherwise specified, searches were restricted to structures smaller radii, but the values obtained show an erratic dependence
with no disorder, no errors and an agreement factor R of at on the atomic number. The small increase in the covalent radii of
most 10%. For obtaining the covalent radii of those elements Hg and Cd relative to their preceding elements that appears when
for which the number of structures was too small, one or more we use C as the reference element (Fig. 2) is found also if N is used
of those constraints was removed. For elements insufficiently or as reference instead, even if more pronounced. Tl, Pb and Bi give
not represented at all in the CSD, some structural data could be much longer radii when deduced from M–N bonds than from M–C
found in Karlsruhe’s inorganic database (November 2006 version bonds, and also give a much larger e.s.d., indicating that those M–
of the ICSD20 with 93 720 entries). As a general criterion, we N bonds are rather ionic and have correspondingly shallower
searched for covalent radii that may represent a given element in potential energy minima. Finally, Th and Pa would give smaller
its typical oxidation states and coordination numbers. Therefore, radii if C was taken as a reference element for them instead of N.
we exchange accuracy for generality, accepting that the uncertainty A further test of the effect of the choice of N or C as
associated with these two factors is embedded into the estimated reference elements for deducing the covalent radius of a given
standard deviation. Nevertheless, several tests have been made element corresponds to the case of Ru. For this metal, the radius
to quantify the importance of both coordination number and deduced from Ru–N bond lengths is 1.45(7) while from Ru–
oxidation state, especially in transition metals. A more detailed Cl distances we obtain 1.40(5) Å, which differ by less than a
discussion of the effects that coordination number, oxidation state, standard deviation. If Ru–C bond distances are used instead, we
bond order and other factors have on bond distances (hence on obtain a bimodal distribution corresponding to radii of 1.10 and
derived radii) can be found in the review of Batsanov.12 1.30 Å, approximately. However, if the C atom is restricted to be
Our choice of one of our three reference elements (C, N and O) to sp3 hybridized, we obtain again a similar value, 1.40(7) Å, while
deduce each radius might have an influence on its numerical value the shorter Ru–C distances are seen to correspond to Ru–C≡X
that must also be evaluated. In general, the shape of the periodic groups, which are likely to incorporate significant mutiple Ru–C
dependence of the covalent radii (see Fig. 2) is not significantly bond character.
altered (see ESI‡), and we comment here on the main differences In the case of copper, to exclude the long distances associated
found between our chosen radii and the alternative choices. For Si, with the Jahn–Teller effect present in Cu(II) compounds, only
P and S, the alternative radii deduced from bond distances to N are Cu–N distances corresponding to Cu atoms with coordination
about one tenth of an ångström shorter. In the case of the alkaline numbers of four or less were considered. Similarly, to rule out
and alkaline earth metals, the choice of N as a reference instead long distances associated with hypervalent compounds, the search

Fig. 2 Proposed covalent radii as a function of the atomic number. The vertical bars correspond to the estimated standard deviation. The labels TM1,
TM2 and TM3 indicate the position of the first, second and third transition series, respectively; Ln indicates the lanthanides and An the actinides.

2834 | Dalton Trans., 2008, 2832–2838 This journal is © The Royal Society of Chemistry 2008
for all non-transition elements (including Zn, Ag and Au) was the corresponding halides,26 much shorter than those of its two
restricted to coordination numbers 2, 3 or 4. In the same vein, for neighboring elements Th and U, although it must be noted that
main group elements, we restricted coordination numbers to the those bond distances vary some 0.3 Å within the same compound.
typical values (i.e., 2 for S, 3 for P or 4 for Si), but allowed for higher The radius of Ra has been deduced from RaF2 ,27 RaS and RaSe.28
coordination numbers, up to four, to include those compounds in For Pu, the only Pu–N distances available come from the structure
which the corresponding atom is acting as a two-, four- or six- of the [Pu(NCMe)9 ]3+ cation,29 but the average of 1.87(1) Å fits
electron donor. nicely into the trend shown by other actinides, and can be retained
Specific problems we have to deal with appear for the first as representative, subject to revision when structural data on more
transition series metals. One is the variability of coordination compounds become available. A radius for Pu derived from Pu–
number. Hence, for consistency, we have considered only bond O distances gives a somewhat smaller value (1.76 Å) but with a
distances in six-coordinate complexes, except for the case of copper too large dispersion (e.s.d. = 0.11 Å). For a few elements (Pm,
commented above. As for the effect of the oxidation states, in most At and Fr) guess values of covalent radii have been obtained
cases the covalent radii are quite insensitive to the oxidation state. by interpolation or extrapolation from neighboring elements and
Thus, the radii for Fe(II) and Fe(III) in either the low (l.s.) or high from the general trends.
spin (h.s.) forms differ by at most 0.01 Å, and also changes of
at most 0.03 Å result for V in oxidation states between +2 and
Results and discussion
+5. In the case of Mo, oxidation states +3 and +4 have the same
radii within one e.s.d., while oxidation states +5 and +6 have radii The covalent atomic radii proposed are collected in Table 2 and the
0.1 Å longer, a fact that seems to be associated with the presence periodic trends can be best appreciated in Fig. 2. The first clear fea-
of oxo ligands with significant Mo=O multiple bond character in ture is the presence of peaks that indicate that the largest element
the highest oxidation state. We find also that the radius of Co(II) of each period is the alkaline metal, while the smallest ones are the
is significantly larger than that of Co(III), but this is due to the halogen and the noble gas. Within each period, there is a sharp
fact that the former is dominated by high spin complexes and the decrease of the covalent radii from the alkaline to the Group 13
latter by low spin ones (see below). In the case of Mn we find also element, while the changes from Group 13 to Group 18 are much
significant differences (0.2 Å) between oxidation states +2 and +5, less pronounced (Fig. 3). In fact, differences in size between these
attributable in part more to the presence of high- and low-spin elements are significant only for the second period (from B to Ne),
compounds than to differences in oxidation state. In this case, while in the next periods the atomic radii for the corresponding
however, it is not easy to perform an automatic search for each elements are roughly constant. The exception to this behavior
oxidation state independently, since the higher oxidation states corresponds to the elements of the sixth period, for which there is
appear most often in mixed valence compounds. an increase in radii from Hg (Z = 80) to Po (Z = 84). These are
For some elements for which structural data are scarce we were precisely elements for which the atomic radii could not be very well
forced to use specific approaches to deduce sensible covalent radii. established due to the scarce number of structural data and the
Hence, a challenging case is constituted by the family of noble variability of their coordination number. Hence, the uncertainty
gases. For Ar and Kr we used the noble gas–noble metal bond in their radii is transferred to our extrapolated values for At and
distances in the gas phase reported by Gerry and coworkers.21 Rn. This situation was already spotted by Slater when much less
Although the data available for Xe compounds are significantly structural data were available. He noticed that for elements from
larger, its radius is highly sensitive to its oxidation state. We mercury to polonium it is very difficult on the basis of empirical
therefore disregarded the high oxidation states for which highly evidence to assign a unique radius from the crystal structures.5
ionic bonds are expected and took only data from compounds For a comparison of the similarities and differences between the
in the formal oxidation state +2. From more than one hundred theoretical radii for isolated atoms and the empirical radii deduced
Xe–E distances (E = N, O, F) we deduced a radius of 1.40(9) from bond distances presented here, the reader is referred to a
Å (all data from the ICSD). As expected, data from compounds recent paper by Ghosh and Biswas30 and references therein.
with higher oxidation states result in smaller values of the Xe
radius. It must be stressed that the radii proposed here based
on structural data are nearly coincident with those based on
extrapolated electronegativity.22 For He, Ne and Rn we could only
extrapolate covalent radii values from the neighboring elements in
the periodic system.
The covalent radius proposed for Ac, deduced from Ac–H
distances in AcH2 (2.15 Å),23 fits nicely into the periodic trend
and is not significantly different from that obtained from Ac–O
bond distances in Ac2 O3 (1.94 Å).24 In AcCl3 and AcBr3 , the Ac
atoms present tricapped trigonal prismatic coordination spheres
with three short bonds that would lead to a radius of 1.58 or
1.55 Å, respectively, but the average of nine Ac–X bonds gives a
radius of 2.15 or 2.13 Å, consistent with that deduced from Ac–
H bond distances. A covalent radius of 2.00 Å for Pa has been
deduced from bond distances in PaH3 .25 A radius 0.1 to 0.2 Å Fig. 3 Evolution of the covalent radii along the periods for main group
shorter would result from Pa–Cl and Pa–Br bond distances in elements.

This journal is © The Royal Society of Chemistry 2008 Dalton Trans., 2008, 2832–2838 | 2835
Table 2 Proposed covalent radii (Å), estimated standard deviation and Table 2 (Contd.)
number of bond distances analyzed
Z Element r e.s.d. n
Z Element r e.s.d. n
65 Tb 1.94 5 55
1 H 0.31 5 129 66 Dy 1.92 7 59
2 He 0.28 67 Ho 1.92 7 48
3 Li 1.28 7 5789 68 Er 1.89 6 66
4 Be 0.96 3 310 69 Tm 1.90 10 15
5 B 0.84 3 1770 70 Yb 1.87 8 122
6 Csp3 0.76 1 10 000 71 Lu 1.87 8 61
Csp2 0.73 2 10 000 72 Hf 1.75 10 53
Csp 0.69 1 171 73 Ta 1.70 8 88
7 N 0.71 1 2200 74 W 1.62 7 219
8 O 0.66 2 10 000 75 Re 1.51 7 476
9 F 0.57 3 10 000 76 Os 1.44 4 99
10 Ne 0.58 77 Ir 1.41 6 131
11 Na 1.66 9 1629 78 Pt 1.36 5 1768
12 Mg 1.41 7 3234 79 Au 1.36 6 114
13 Al 1.21 4 3206 80 Hg 1.32 5 137
14 Si 1.11 2 10 000 81 Tl 1.45 7 291
15 P 1.07 3 10 000 82 Pb 1.46 5 112
16 S 1.05 3 10 000 83 Bi 1.48 4 51
17 Cl 1.02 4 1987 84 Po 1.40 4 4
18 Ar 1.06 10 9 85 At 1.50
19 K 2.03 12 435 86 Rn 1.50
20 Ca 1.76 10 2647 87 Fr 2.60
21 Sc 1.70 7 32 88 Ra 2.21 2 3
22 Ti 1.60 8 231 89 Ac 2.15 1
23 V 1.53 8 389 90 Th 2.06 6 11
24 Cr 1.39 5 916 91 Pa 2.00 1
25 Mn l.s. 1.39 5 321 92 U 1.96 7 57
h.s. 1.61 8 929 93 Np 1.90 1 22
26 Fe l.s. 1.32 3 336 94 Pu 1.87 1 9
h.s. 1.52 6 1540 95 Am 1.80 6 11
27 Co l.s. 1.26 3 5733 96 Cm 1.69 3 16
h.s. 1.50 7 780
28 Ni 1.24 4 1030
29 Cu 1.32 4 1149 For the transition metals, plotted independently in Fig. 4, we
30 Zn 1.22 4 443 observe a clear decrease in the radius along the first transition
31 Ga 1.22 3 1330
32 Ge 1.20 4 1013
series. The deviation of the copper radius from the general trend
33 As 1.19 4 2015 must be attributed to the plasticity of its coordination sphere that
34 Se 1.20 4 1717 gives several coordination numbers and may even incorporate
35 Br 1.20 3 2140 weakly bonded ligands. We must recall that the slightly large
36 Kr 1.16 4 5
37 Rb 2.20 9 23 radius of Cu cannot be attributed to the presence of Jahn–
38 Sr 1.95 10 1500 Teller distortions in Cu(II) compounds, since we have explicitly
39 Y 1.90 7 30 excluded five- and six-coordinate compounds to disregard such
40 Zr 1.75 7 93
weak bonds. We must notice also that changes in covalent radii
41 Nb 1.64 6 18
42 Mo 1.54 5 97 on going from Group 11 to Group 12 elements are biased by the
43 Tc 1.47 7 96 change of reference bonded atom from nitrogen to carbon and
44 Ru 1.46 7 1032 we will not attempt to discuss such variations, which might not be
45 Rh 1.42 7 458
46 Pd 1.39 6 1892 significant. The radii for the second transition series elements show
47 Ag 1.45 5 1728 an evolution along the period practically parallel to those of the
48 Cd 1.44 9 19 first transition series, with values roughly 0.1 Å longer than for the
49 In 1.42 5 546 corresponding first row metal. The third transition series elements
50 Sn 1.39 4 2999
51 Sb 1.39 5 609 present radii that are practically coincident with the second series
52 Te 1.38 4 692 ones, with the exception of La. Such a behavior is well known and
53 I 1.39 3 451 is attributed to the lanthanide contraction.9 It is worth mentioning
54 Xe 1.40 9 2
55 Cs 2.44 11 24
that for the later transition elements there seems to be a crossover
56 Ba 2.15 11 3076 of the two curves and the size of Au becomes smaller than that
57 La 2.07 8 190 of Ag, a fact that was already pointed out by Schmidbaur and
58 Ce 2.04 9 47 coworkers.31 A similar inversion of the atomic size is apparent for
59 Pr 2.03 7 58
60 Nd 2.01 6 96 Cd and Hg in Fig. 4.
61 Pm 1.99 A feature that must be considered for first row transition
62 Sm 1.98 8 53 elements is the significant difference in bond distances that the
63 Eu 1.98 6 167
same metal presents in its high- and low-spin compounds. Hence,
64 Gd 1.96 6 178
for such metals it is practical to associate a different radius for

2836 | Dalton Trans., 2008, 2832–2838 This journal is © The Royal Society of Chemistry 2008
 Fig. 6 Evolution of the covalent radii down the main groups. Group 16
elements and noble gases, not shown for clarity, are practically coincident
with the halogens.
Fig. 4 Evolution of the covalent radii along the transition series.
contraction,9 is similar to the lanthanide contraction responsible
each of the two spin states. This can be easily done from structural for the nearly identical radii of Zr and Hf.
data because the M–N bond distances for Mn, Fe and Co show In summary, the set of atomic covalent radii proposed has been
bimodal distributions (Fig. 5), from which we can associate the deduced for most elements with atomic numbers up to 96 from the
maximum at shorter distances to the low spin and that at longer analysis of more than 228 000 experimental bond distances to N, C
distances to the high spin complexes. In the case of Co, even if or O atoms. Such covalent radii offer a wider and more consistent
the distribution of bond distances is dominated by short ones view of the periodic trends than those currently in use. The average
attributable to the most abundant low spin Co(III) compounds, a standard deviation of the radii is 0.06 Å with a maximum of 0.12 Å
small peak at larger values, probably corresponding to high spin (for K). Only six elements have a covalent radius with a standard
Co(II) complexes, allowed us to also assign a high spin radius to deviation of 0.10 Å or larger (Ca, Hf, Tm, Ar, Sr, Cs, Ba and K).
this element. The corresponding radii deduced from M–C bond Taking again the example of Fig. 1, we compare now the Ba–
distances give only one maximum for each metal that is roughly N distances with the sum of the atomic radii proposed here and
coincident with the shorter one deduced from M–N distances, find the distribution shown in Fig. 7, with an average deviation of
thus confirming that the attribution of the lower peak to low spin 0.10 Å and a maximum deviation of 0.53 Å, even if the radius for
complexes is a sensible approach, since we expect C-donor ligands Ba has not been deduced from Ba–N but from Ba–O distances.
to favor a low spin ground state. A look at Fig. 4 shows that the
low spin radii nicely follow the same trend as other transition
metals with only one spin state. It is also remarkable that the high
spin radii are significantly larger than those of the third transition
series metals, and that they also decrease monotonically with the
atomic number.

Fig. 7 Distribution of the difference between the experimental Ba–N

bond distances and the sum of the covalent radii proposed here (absolute
values), to be compared with Fig. 1.

All the radii are at least 0.5 Å shorter than the van der
Waals radii4 and are very similar to Pauling’s metallic radii for
Fig. 5 Bimodal distribution of the Fe–N bond distances in six-coordinate those elements for which the latter are defined. As a test of
iron compounds. The maximum at short distances is assigned to low spin the general applicability of the present set of covalent radii, we
complexes and the one at longer distances to high spin ones, regardless of have randomly chosen 160 different bond types not involving
the oxidation state. the reference elements (C, N and O) and compared the resulting
172 000 experimental bond distances with the corresponding radii
While the elements of most main groups show a monotonic sums. The results (ESI‡) show that 96% of the bond types have
increase in their covalent radii down the Group (Fig. 6), elements average bond distances that fall within three standard deviations of
of Groups 13 and 14 present a discontinuity between the second the radii sum, while 84% are found within two standard deviations.
and third periods, i.e., from Al to Ga and from Si to Ge. Such As an additional check, radii have been deduced for elements of
discontinuities should be attributed to the increased nuclear charge periodic Groups 14 to 17 from element–element bond lengths in
of the heavier elements of these couples, not fully screened by Men E–EMen and related compounds (n = 3, 2, 1 and 0 for Groups
the valence 3d electrons. Such an effect, termed transition series 14, 15, 16 and 17, respectively) in the Mogul knowledge base.32

This journal is © The Royal Society of Chemistry 2008 Dalton Trans., 2008, 2832–2838 | 2837
The agreement is excellent (data and plot provided as ESI‡), but 13 W. J. Evans, Inorg. Chem., 2007, 46, 3435.
the radii proposed here are representative of a much wider set of 14 CSD radii: http://www.ccdc.cam.ac.uk/products/csd/radii/.
15 A. A. Palacios, G. Aullón, P. Alemany and S. Alvarez, Inorg. Chem.,
structural data corresponding to a variety of bonds to different 2000, 39, 3166.
elements, and still their standard deviations are in general smaller. 16 Large bond distance data sets are truncated by the Vista program used
for our numerical analysis to 10 000 data, which is considered to provide
good enough statistics.
Acknowledgements 17 A. F. Wells, in Structural Inorganic Chemistry, Clarendon Press, Oxford,
5th edn., 1984.
Support from the Dirección General de Investigación (MEC), 18 R. J. Gillespie and E. A. Robinson, Inorg. Chem., 1992, 31, 1960.
project CTQ2005-08123-C02-02BQU, and from Comissió Inter- 19 F. H. Allen, Acta Crystallogr., Sect. B: Struct. Sci., 2002, 58, 380.
departamental de Ciència i Tecnologia (CIRIT), grant 2005SGR- 20 A. Belsky, M. Hellenbrandt, V. L. Karen and P. Luksch, Acta
Crystallogr., Sect. B: Struct. Sci., 2002, 58, 364.
00036, is acknowledged.
21 J. M. Thomas, N. R. Walker, S. A. Cooke and M. C. L. Gerry, J. Am.
Chem. Soc., 2004, 126, 1235.
References 22 L. C. Allen and J. E. Huheey, J. Inorg. Nucl. Chem., 1980, 42, 1523.
23 J. D. Farr, A. L. Giorgi, M. G. Bowman and R. K. Money, J. Inorg.
1 R. F. W. Bader, in Atoms in Molecules. A Quantum Theory, Clarendon Nucl. Chem., 1961, 18, 42.
Press, Oxford, 1990. 24 W. H. Zachariasen, Acta Crystallogr., 1949, 2, 388.
2 W. L. Bragg, Philos. Mag., 1920, 40, 169. 25 P. A. Sellers, S. Fried, R. E. Elson and W. H. Zachariasen, J. Am. Chem.
3 L. Pauling, J. Am. Chem. Soc., 1947, 69, 542. Soc., 1954, 76, 5935.
4 A. Bondi, J. Phys. Chem., 1964, 68, 441. 26 D. Brown, C. T. Reynolds and P. T. Moseley, J. Chem. Soc., Dalton
5 J. C. Slater, J. Chem. Phys., 1964, 41, 3199. Trans., 1972, 857; D. Brown, T. J. Petcher and A. Smith, Acta
6 R. D. Shannon and C. T. Prewitt, Acta Crystallogr., Sect. B: Struct. Crystallogr., Sect. B: Struct. Crystallogr. Cryst. Chem., 1969, 25, 178; D.
Crystallogr. Cryst. Chem., 1969, 25, 925; R. D. Shannon and C. T. Brown, T. J. Petcher and A. J. Smith, J. Chem. Soc. A, 1971, 908; R. P.
Prewitt, Acta Crystallogr., Sect. B: Struct. Crystallogr. Cryst. Chem., Dodge, G. S. Smith, Q. Johnson and R. E. Elson, Acta Crystallogr.,
1970, 26, 417. 1967, 22, 85; D. Brown, T. L. Hall and P. T. Moseley, J. Chem. Soc.,
7 R. D. Shannon, Acta Crystallogr., Sect. A, 1976, 32, 751. Dalton Trans., 1973, 686.
8 L. Pauling, in The Nature of the Chemical Bond, Cornell University 27 G. E. R. Schulze, Z. Phys. Chem., Abt. B, 1936, 32, 430.
Press, Ithaca, 1980. 28 F. Weigel and A. Trinkl, Radiochim. Acta, 1969, 12, 29.
9 D. M. P. Mingos, in Essential Trends in Inorganic Chemistry, Oxford 29 A. E. Enriquez, J. H. Matonic, B. L. Scott and M. P. Neu, Chem.
University Press, Oxford, 1998. Commun., 2003, 1892.
10 N. W. Alcock, in Bonding and Structure: Structural Principles in 30 D. C. Ghosh and R. Biswas, Int. J. Mol. Sci., 2002, 3, 87.
Inorganic and Organic Chemistry, Ellis Horwood, New York, 1990. 31 A. Bayler, A. Schier, G. A. Bowmaker and H. Schmidbaur, J. Am.
11 I. S. Butler and J. F. Harrod, Inorganic Chemistry: Principles and Chem. Soc., 1996, 118, 7006.
Applications, Benjamin/Cummings, Redwood City, CA, 1989. 32 I. J. Bruno, J. C. Cole, M. Kessler, J. Luo, W. D. S. Motherwell, L. H.
12 S. S. Batsanov, Russ. J. Inorg. Chem. (Transl. of Zh. Neorg. Khim.), Purkis, B. R. Smith, R. Taylor, R. I. Cooper, S. E. Harris and A. G.
1991, 36, 1694. Orpen, J. Chem. Inf. Comput. Sci., 2004, 44, 2133.

2838 | Dalton Trans., 2008, 2832–2838 This journal is © The Royal Society of Chemistry 2008

View publication stats