doi.org/10.26434/chemrxiv.9255446.

v1

The Limited Predictive Power of the Pauling Rules

Janine George, David Waroquiers, Davide Di Stefano, Guido Petretto, Gian-Marco Rignanese, Geoffroy
Hautier

Submitted date: 06/08/2019 • Posted date: 06/08/2019

Licence: CC BY-NC-ND 4.0
Citation information: George, Janine; Waroquiers, David; Di Stefano, Davide; Petretto, Guido; Rignanese,
Gian-Marco; Hautier, Geoffroy (2019): The Limited Predictive Power of the Pauling Rules. ChemRxiv.
Preprint.

The Pauling rules have served to analyze and rationalize crystal structures for decades. So far, there is no
statistical answer on how these five rules perform. Here, we show the analysis of all five Pauling rules based
on the geometrical data of around 5000 oxides. To do this, we determined the coordination environments of all
5000 oxides and compared our geometrical analysis to the prediction of the Pauling rules. We arrived at the
conclusion that the Pauling rules - especially the electrostatic valence principle - perform poorly for most of the
oxides: only 13% of all oxides simultaneously satisfy rules from two to five. To arrive at such an intuitive
understanding of the stability of crystals that Pauling’s rules gave shortly after their development again, we
should develop new and improved rules.

File list (3)

Preprint_5_8_19.pdf (1.15 MiB) view on ChemRxiv download file

Supporting Information_5_8_19.pdf (768.20 KiB) view on ChemRxiv download file

CSV_Files_Supporting_Info_5_8_19.zip (1.45 MiB) view on ChemRxiv download file

The limited predictive power of the Pauling rules
Janine George,a David Waroquiers,a Davide Di Stefano,a Guido Petretto,a Gian-Marco
Rignanese,a and Geoffroy Hautier*a
a Institute of Condensed Matter and Nanosciences, Université Catholique de Louvain,
Chemin des étoiles 8, 1348 Louvain-la-Neuve, Belgium, E-mail:
geoffroy.hautier@uclouvain.be

Abstract

The Pauling rules have served to analyze and rationalize crystal structures for decades. So
far, there is no statistical answer on how these five rules perform. Here, we show the analysis
of all five Pauling rules based on the geometrical data of around 5000 oxides. To do this, we
determined the coordination environments of all 5000 oxides and compared our geometrical
analysis to the prediction of the Pauling rules. We arrived at the conclusion that the Pauling
rules – especially the electrostatic valence principle – perform poorly for most of the oxides:
only 13% of all oxides simultaneously satisfy rules from two to five. To arrive at such an intuitive
understanding of the stability of crystals that Pauling’s rules gave shortly after their
development again, we should therefore develop new and improved rules.

Introduction

In 1929, Pauling published the nowadays famous five Pauling rules.[1] They are described in
nearly every solid-state textbook, they serve to suggest reasonable crystal structures and to
rationalize crystal structures.
Within these rules, the structures are described with the help of coordination polyhedra that
can be connected via corners, edges, and faces. The rules predict stable coordination
polyhedra for certain cation-anion combinations with the help of their radius ratio. Furthermore,
they address local charge compensation within the crystal structures, i.e. they describe the
mathematical connections between charges of the cations and anions forming the coordination
polyhedra. Moreover, they predict the connections of the coordination polyhedra via corners,
edges, or faces depending on the oxidation states and coordination numbers of the crystal.
The rule of parsimony anticipates the number of constituents in a stable crystal.
Pauling based his rules on crystal structures known at the time of the publication and on the
simple consideration of electrostatics. Pauling himself already describes that there should be
exceptions to the rules: “They are not rigorous in their derivation or universal in their
application, but they have been found useful as a criterion for the probable correctness of
reported structures for complex crystals and as an aid in the X-ray investigation of crystals by
making possible the suggestion of reasonable structures for experimental test.” Furthermore,
Pauling has already suspected that his rule would only work for certain substances which are
largely ionic in character. In the following study, we will therefore concentrate on oxides which
are expected to be rather ionic in character due to the high electronegativity of oxygen. Also,
oxides are much more frequent than other chemistries such as nitrides, or sulfides; this
simplifies a statistical evaluation.

1
Due to the drawbacks of Pauling’s rules, there have been several additions to the original
Pauling rules. Baur showed how the distortion of the polyhedra is connected to Pauling’s
electrostatic valence principle.[2] The second rule was further extended within the bond valence
method.[3-4] Beck categorized structures and their fulfillment of the electrostatic valence rule in
oxides of the type AB2O6 with the help of the fact that the number of cation bonds should be
the same as the number of anion bonds if only cation-anion bonds exist.[5] Also, radii describing
the reality in crystal structures in a better way – the so called Shannon radii - were introduced.[6]
In contrast to Pauling’s univalent radii,[7] they depend on the valences and the coordination of
the cations.
Up to now, there is no study that answers the question how all Pauling rules actually perform
based on a statistical evaluation of the rules with the help of real crystal structures. We hope
such a statistical evaluation will be a first step towards the development of new and improved
rules describing the coordination polyhedra and their connections within stable crystal
structures.
In the following, we will test each of the rules in a statistical manner for oxides and will then
discuss trends among the structural exceptions to the rules which will shed light on the question
why the Pauling rules are violated.

Results and Discussion

1. Radius Ratio Rule: The first rule describes the determination of the coordination
number/coordination environment of a cation based on a hard sphere model of the atoms: "[...]
The coordination number of the cation [is determined] by the radius ratio [of cation and
anion].”[1] This is illustrated in Figure 1. A coordination environment is only stable if the radius
ratio of cation and anion arrives in the geometrically derived stability limit of this environment.

Figure 1. Illustration of the radius-ratio rule (Pauling’s first rule) that is based on purely
geometric considerations to decide whether a coordination polyhedron is stable. The relevant
ratio is the one of the radii of the cation and the anion building the coordination polyhedron.
The figure was inspired by reference [8].

We found that the prediction of this rule agreed with the geometrical assessment of the
structures in 66% of tested environments. Figure 2a shows an elementwise analysis of the
fulfillment of the rule. The radius ratio rule agrees well with the geometrical determination of
the Pauling rules for those cases where the cation only shows very few different coordination
environments and where the radius ratio predicts the most frequent of those coordination
environments. For instance, Si, P, and S show mostly tetrahedral coordination environments
and Pauling’s rule predicts this coordination.[9] In contrast, many of the alkali and alkaline earth
metals and some transition metals show many different environments and the radius ratio does
not even predict the most frequent one. For example, the most frequent environment for Sc,
Cd, and Ca is the octahedron. However, the radius ratio predicts face-capped octahedrons
2
and square antiprisms instead.[9] As one can already see for the Ca case, including the
valences into this assessment can clearly not resolve the issue since Ca typically shows only
the valence of 2+.

Figure 2. Assessment of the 1st rule. a) Ratio of correct predictions by Pauling’s rule to all
tested predictions. The prediction is correct, if it agrees with the geometrical assessment of the
structure. Only those environments were tested, where our geometrical assessment resulted
in an environment listed in Table 1. b) 1- Ratio of Shannon entropy of environments for certain
element to maximum Shannon entropy. For certain elements such as Si and P, only one
coordination environment dominates.

To assess the general limit of the first rule in more detail and to remove the strong dependency
on the radii used in our assessment, we calculated the Shannon entropy of the frequencies of
the coordination environments for each element as shown in Figure 2b. A high Shannon
entropy reflects that there are many coordination environments for this element with small
frequencies, a small Shannon entropy reflects that there are only a few coordination
environments with high frequencies. For the lighter elements in the main groups 3-6 only one
environment dominates so that a prediction of such an environment is easily possible. As
mentioned before, this is very different for alkali and alkaline earth metals. Many different
environments with a similar frequency exist.

3
To sum up, this rule works well for elements where only one coordination environment
dominates and this one is predicted by the radius ratio of the cation and the oxygen. Future
efforts should be devoted to understanding under which conditions (e.g., neighboring atoms)
each of the environments form. This might enable us to predict the coordination environment
correctly for more elements from the periodic table.

2. Electrostatic Valence Rule: The second rule deals with local charge compensation within
crystal structures: “In a stable coordination structure the electric charge of each anion tends to
compensate the strength of the electrostatic valence bonds reaching to it from the cations at
the centres of the polyhedra of which it forms a corner [...].” The application of this rule is
demonstrated on α-quartz. In this example, the charge of the oxygen atom (-2) is compensated
by the bond strengths of the two silicon neighbors (+1) (see Figure 3). The bond strength of
the bond from the cation is calculated by dividing the valence of the cation by its coordination
number (+4/4).

Figure 3. Illustration of the electrostatic valence rule (Pauling’s second rule).

We checked this rule for all oxygen atoms in our data set (Figure 4a, left). The rule is only
nearly exactly fulfilled for roughly 20% of all oxygen atoms. We find some extreme deviations
from the expected ideal valence of more than 50%, i.e. the sum of the bond strengths of the
neighbouring atoms results in a value of 1 or 3 instead of 2.
Figure 4b shows the fulfillment of the rule per element. There is no easy trend within the
periodic table but exceptions are notable. Many phosphate and vanadate compounds are
among the exceptions, e.g. InPO4 (Materials Project ID: mp-7566),[10] and CrVO4[11] (mp-
19418) which crystallize in the same structure type. Cr is octahedrally-coordinated, V is
tetrahedrally coordinated by oxygen. CrVO4 and the calculation of the bond strengths sums
are depicted in Figure 4c. Within the structure, there are two different oxygen types present.
For one, the sum of the bond strengths arrives at 2.25, for the other at 1.75, which would lead
to an absolute deviation from the ideal valence of 12.5 % in both cases. With the help of the
“co-ordination rule” that Beck used in his research on AB2O6 compounds and their fulfillment
of the second rule, this can also be easily understood:[5] the numbers of bonds of Cr and V to
O should be the same as the number of bonds from O to Cr and V. This can only be fulfilled if
there are at least two different types of oxygen environments because dividing all cation bonds
(6 for Cr and 4 for V) by the number of all oxygens (4) results in a non-integer number (2.5).
For the coordination numbers present in the structures, the rule can therefore not be fulfilled.
As before, many alkali, alkaline earth metals and lanthanides show exceptions to this rule.
Among these exceptions are many perovskite-derived structures such as Ba3MgRu2O9 (mp-
6520) or Ba3MgNb2O9 (mp-6474), pyrochlores such as Nd2Cd2O7 (mp-5472),[12] pyroxenes
such as MgSiO3 (mp-3470),[13] olivines such as NaCdPO4 (mp-6684),[14] and melilites such as
Ca2Mg(Si2O7).[15] For several of these compounds, the deviations from the second Pauling rule
have already been discussed in literature.[15-16]
There are also some notable binary exceptions: β-Ga2O3 (mp-886), θ-Al2O3 (mp-7048), S3O9[17]
(mp-2414), P2O5[18] (mp-2173), and VO2 (HT) (mp-561637)[19]. Oxygens surrounded by Al, Si,
4
and Sn agree well with this rule (~60%). Also, oxygens surrounded by some transition metals
(Sc, Rh, Pd, Ir) show good agreement. In contrast, oxygens surrounded by P exhibit many
exceptions to this rule. This underlines again that there is no clear trend throughout the periodic
table because oxygens surrounded by P which is located directly next to Si within the periodic
table display many more exceptions.

Figure 4. Assessment of the 2nd rule. a) (left) Share of the oxygen atoms that show at most a
certain percentage absolute deviation from the ideal valence -2. (right) Share of the oxygen
atoms in structures with only very symmetric coordination environments that show at most a
certain percentage absolute deviation from the ideal valence -2. Here, only structures with
environments are considered that show a maximum continuous symmetry measure of 0.1 and
are assigned with a high certainty (they have a coordination environment fraction of 0.95) and
are therefore very symmetric. b) Elementwise depiction of the fulfillment of the second rule. c)
Depiction of relevant connections of polyhedra for the calculation of the sum of bond strengths
within the crystal structure of CrVO4. The sum of bond strengths arrives in one case at 2.25
and in the other at 1.75.

5
According to the rule of Baur,[2] distortions of chemical environments are another local charge
compensation mechanism in crystal structures. To assess this in more detail, we repeated the
beforehand analysis for structures with only very symmetric chemical environments (see
Figure 4a, right) which are around 240 materials out of the 5000 we investigated before. This
time, the rule was nearly perfectly fulfilled for more than 90% of environments so that we also
see that the distortion of the environments has an influence on the fulfillment of the rule.
The second Pauling rule works only well if you concentrate on nearly perfect coordination
environments and/or specific elements such as Al, Si, or Sn. This is especially relevant for
efforts where this local charge compensation rule is used for crystal structure prediction, as
done in ref [20]. It would only work correctly for certain elements constituting the oxide. If one
would like to use this rule for structure prediction, the rule should be used in a more flexible
manner, i.e. allowing for a certain deviation from the perfect fulfillment of the rule. A statistical
testing of the rules of Baur is beyond the scope of our article.
3. The Sharing of Edges and Faces:
The third rule deals with the stability of crystals based on connections of the coordination
polyhedra which is illustrated in Figure 5: “The presence of shared edges, and particularly of
shared faces, in a coordinated structure decreases its stability; [...].”

Figure 5. Illustration of the Rule on “The Sharing of Edges and Faces” (Pauling’s third rule).
In stable crystals, corner connections of coordination polyhedra are preferred over edge and
especially face connections.

The shares of the connections of all connected pairs of polyhedra in our data set are depicted
in Figure 6a (left). The connections via corners dominate with 64%. Then, the edge
connections follow with 27%. The lowest percentage of connected polyhedra pairs are
connected via faces. This agrees well with Pauling’s rule.
We defined structures showing face connections of polyhedra as exceptions to the third rule.
We therefore investigated the pairs of polyhedra only showing corner and edge connections
and the ones showing face connections in more detail. Figure 6b shows which elements tend
to fulfil the rule. The higher the percentage in Figure 6b, the lower the numbers of face
connections of polyhedra of these cations. In general, smaller elements seem to fulfil the rule
better (low period and high group number). We tested this hypothesis, by comparing the
numbers of cations with coordination polyhedra that are only connected via corners or edges
to the ones connected via faces as a function of the atomic radius (see Supporting Information
Figure S7). Both distributions are clearly different, so that the size of the element seems to
have an influence on the type of connection. Again, Al and Si fulfil the rule very well. This is
also the case for some transition metals such as V, Cr, Mo, Cu, Ag, Au, Zn, Cd, and Hg.
Several alkali and alkaline earth elements and their polyhedra have an unsatisfying fulfillment
of the rule, i.e. many more face connections.
We also analysed the structures showing face connections (42% of all tested structures) in
more detail: among them were many perovskites, e.g. BaTiO3 (mp-5986)[21-22] as depicted in

6
Figure 6c[23-24], and fluorite-related phases with high coordination number such as La3NbO7
(mp-560349)[25]. Some binary structures including Ti, V, or Al also do not fulfil the rule: several
of the structures belong to the corundum structure type such as Al2O3 (mp-1143, see Figure
6c)[26-27], Ti2O3 (mp-458),[28] Fe2O3 (mp-24972)[27-28], V2O3 (mp-18937, HT)[28-29]. Others are
suboxides of vanadium and titanium: V3O5 (mp-622497),[30] V4O7 (mp-567080),[31] Ti5O9 (mp-
748)[32] belonging to the Magnéli phases.[33]
If we exclude cations with high coordination numbers from our analysis (only coordination
numbers ≤8 are allowed), we arrive at a result in even better agreement with Pauling’s third
rule (see Figure 6a, right): less face connections. Pauling already writes in the second part of
his rule that “this effect [of lowering the stability of the materials by edges and faces] is large
for cations with large valence and small coordination number” because of the different changes
of the Coulomb interactions when tetrahedra and octahedra share edges and faces. [1]
Exceptional structures with coordination numbers smaller or equal to 8 are for example
melilites such as Ca2Mg(Si2O7) where the coordination polyhedron of the eightfold coordinated
Ca2+ shows face connections.[15] Again, also structures of the corundum structure type such as
Al2O3 (mp-1143)[26-27], where polyhedra with sixfold coordination exhibit face connections, are
among the exceptions.

7
Figure 6. Assessment of the 3rd rule. a) Shares of connected pairs of polyhedra that are
connected via corners, edges and faces for all coordination polyhedra (left) and only
coordination polyhedra with a coordination number (CN) equal to 8 or smaller (right). b)
Elementwise fulfillment of the rule. The fulfillment of the rule seems to depend on the atomic
radius of the elements. c) Structure of BaTiO3 (mp-5986) [21-22] and Al2O3 (corundum, mp-1143)
showing connections of polyhedra via faces.

To conclude, many structures (42% of all tested structures) have at least one face connection
even at normal pressure or temperature conditions even though the overall number of face
connections in our data set is small. This should be accounted for in future rules on the stability
of crystals. The size of each atom influences the frequency of face connections of the
polyhedra.

4. The Nature of Contiguous Polyhedra: The fourth rule is an extension of the third rule. It
refers to crystals containing different cations and the connections of the polyhedra within them
(see Figure 7 for a depiction of the rule): “In a crystal containing different cations those with
large valence and small coordination number tend not to share polyhedron elements with each
other.”

Figure 7. Depiction of the 4th rule which deals with “The Nature of Contiguous Polyhedra”: a
stable crystal should not exhibit connections between coordination polyhedra with a small
coordination number (CN) and around a cation with a large valence.

8
Figure 8. Assessment of the 4th rule. a) The dependency of the share of non-connected pairs
of polyhedra of all considered pairs of polyhedra on the coordination numbers, and valences
of the cations. b) Elementwise depiction of the rule fulfillment. c) One of the exceptions, α-
NaFeO2, is shown. The octahedral coordination environments of Fe(+III) show connections.
Li(+I) is also octahedrally coordinated so that Fe(+III) is the cation with the lowest coordination
number and the highest valence.

Again, we analysed pairs of polyhedra but this time also non-connected pairs of polyhedra. We
plot the share of non-connected pairs of all considered pairs of polyhedra (selected by a
distance criterion) as a function of the coordination numbers and as a function of the valences
(see Figure 8a). We find a tendency following the rule for the coordination numbers. There are
also some exceptions visible. There is no such a clear tendency for the valences: the shares
of non-connected pairs of polyhedra are very similar for many combinations of valences.

9
Among the structural exceptions to this rule (structures in which the polyhedra of cations with
the highest valence and smallest coordination number show connections!) are many structures
that belong to the perovskite structure type such as BaTiO3 (mp-5986) [21-22], pyrochlores such
as Nb2Cd2O7 (mp-5472),[12] pyroxenes such as MgSiO3 (mp-3470),[13] fluorite-related phase
such as La3NbO7 (mp-560349)[25]. In all cases, the cations with the highest valence (and the
smallest coordination number) show connections of their coordination polyhedra.
Several structures belonging to the α-NaFeO2 structure type[34] (mp-19359, see Figure 8c for
a depiction of the structure) are also among the exceptions. Here, all cations have the same
coordination number and those cations with the largest valence share polyhedron elements
with each other. In the specific case of α-NaFeO2 this might be rationalized with the help of
antiferromagnetic interactions of the iron atoms[35-36] which might not be as strong as in this
structure if the Fe and Na were direct neighbors.
In contrast the aforementioned example, all cations in the orthorhombic perovskite TbVO3 (mp-
565331)[37] have the same valence (+III) and the polyhedra of those cations with low
coordination number (here 6) share corner connections. We are, however, aware that one
could disagree with our definition of a small coordination number in this specific case but
Pauling was not this precise in his original publication.
If we look at an elementwise exception to the rule (Figure 8b), several structures including
phosphorous are exceptions to this rule. In contrast, structures including silicon fulfil the rule
better and structures including sulfur fulfil this rule nearly perfectly. This correlates nicely with
the fact that condensed phosphate anions are formed more easily and are more frequent than
condensed anions of sulfates which need very harsh synthesis conditions.[38] Those
condensed phosphates exhibit many exceptions to this rule. We find 133 structures in our
dataset including P2O74- anions, and only 5 including S2O72- anions which are exceptions to
this fourth rule. Again, also many crystal structures including alkaline earth metals are
exceptions.
In sum, there is a dependency of the missing connections between coordination polyhedra on
the coordination numbers, but none on the valences. Again, no clear trend in the periodic table
for the fulfillment of this rule is visible. Notable is that structures including elements from group
6 and 11 and several main group elements fulfil the rule well.

5. The Rule of Parsimony: The fifth rule predicts the number of constituents in stable crystal
structures. According to Pauling, it should hold that “The number of essentially different kinds
of constituents in a crystal tends to be small.” The constituents he is referring to are
coordination polyhedra of chemical identical cations (i.e. cations of the same element and
valence) and their connections (see Figure 9 for a depiction of the rule).

Figure 9. Visualization of the Rule of Parsimony (Pauling’s 5th rule). Crystals with unit cells
that only include one kind of coordination environment of chemical identical cations are more
stable than others including more than one kind of coordination environment.

10
To analyse this rule, we analysed whether these chemical identical cations (i.e. cations of the
same element and valence) had the same coordination numbers; this is the loosest
interpretation of the rule. We only checked structures that had more than one chemical identical
cation within the primitive cell which excludes structures like α-quartz. The percentage of
structures having chemical identical cations with the same coordination numbers amounts to
around 70% (see Figure 10a). This already indicates that there are many exceptions to this
rule. When one excludes those elements that show mainly one coordination environment, the
percentage of structures fulfilling the rule drops by 1 %.
The elementwise analysis of the rule fulfillment is depicted in Fig 10c. Structures including Sn
fulfil this rule nearly perfectly. In contrast, structures including Ga and Ge show worse
fulfillment. Among the exceptional structures are also many including alkali metals. One of the
binary exceptions we found is β-Ga2O3 (mp-886) [39-40] in which Ga(+III) shows both fourfold
and sixfold coordination (see Figure 10c). We also found binary exceptions for Al, e.g.: θ-Al2O3
(mp-7048), which is isomorphous to β-Ga2O3,[41] and Κ-Al2O3 (mp-2254), which also shows a
four- and sixfold coordination of Al(+III).[42] As2O5 (mp-1788) contains As(+V) both with fourfold
and sixfold coordination (see Figure 10c).[43]

Figure 10. Assessment of the 5th rule. a) Pie chart depicting the structures fulfilling the rule of
parsimony (only coordination numbers are considered). b) Exceptions: Crystal structure of β-
Ga2O3 [39-40] and As2O5[43]. The crystal structure of β-Ga2O3 [39-40] shows both tetrahedral and
octahedral coordination of Ga(+III). β-Ga2O3 is more stable than α-Ga2O3 (energy above hull
and ref. [39]). The structure of As2O5 shows both fourfold and sixfold coordinations of As(+V).
c) Elementwise fulfillment of the fifth rule when only coordination numbers are considered.

11
Several other structure types are among the exceptions: one of them is the garnet structure
type X3Y5O12 (e.g., Er3Ga5O12 (mp-12236), Lu3Al5O12 (mp-14132)) [44] where Y shows both
tetrahedral and octahedral coordination. The Langbeinite K2Mg2(SO4)3 (mp-6299)[45] also
belongs to the exceptions – K shows both nine- and twelvefold coordination. Several borate
structures such as SmCd(BO2)5 (mp-14556)[46] where B is both coordinated by three and four
oxygen atoms and iodate structure such as Bi(IO3)3 (mp-31259)[47] where I shows both five-
and eightfold coordination are also among the exceptions. Also, several Ruddlesden-Popper
phases such as Sr3Ti2O7 (mp-3349),[48] Sr3Fe2O7 (mp-18820),[49] and Sr3Ru2O7 (mp-5868),[50]
are further exceptions to this rule. Here, Sr shows two different coordination environments.

In brief, the fulfillment of this rule amounts to around 70% if one only considers coordination
numbers. Again, many exceptions were found and no clear trend among the exceptions within
the periodic table was visible.

Combined Assessment of the Five Rules: To assess the overall performance of the Pauling
rules, we combine the assessment of rules 2 to 5 in this paragraph. We avoid assessing the
first rule, because the assessment will strongly dependent on the quality of the univalent radii.

Figure 11. Combined assessment of rules 2 to 5. a) Dependency of the rule fulfillment on each
of the rules and the influence of the coordination numbers (CN) numbers on the rule fulfillment.
b) Percentages of structures including this element which fulfil the rule.

Among the structures that fulfil the rules two to five are binary structures in the rutile structure
type such as GeO2[51] (mp-470), spinels like MgAl2O4[52-53] (mp-3536), structures in the scheelite
12
structure type (CaWO4) such as BiAsO4[54] (mp-561068), and many structures in the ZrSiO4
(mp-4820) structure type. Furthermore, some phosphates such as α-AlPO4 (mp-3955,
berlinite) which is isotypic to α-quartz,[55-56] several sulfates such as In2(SO4)3 (mp-541450),[57]
and silicates such as Mg2SiO4 (mp-2895)[58] are among the fulfilling structures. If one looks at
the elementwise fulfillment of these four rules in Figure 11b, one can see that structures
including Al, Si, and Sc tend to fulfil the rule better than structures including P, and the alkali
and alkaline earth metals.
In sum, 13% of all tested structures fulfil the rules two to five simultaneously. This indicates
that structures opposing Pauling’s rules are quite frequent. In Figure 11a, we show the
dependency of the fulfillment of the rule on each of the rules by only assessing rule three to
five, two to four and so on. The second rule is clearly the rule that is fulfilled the worst. Including
the third rule also has a drastic influence on the overall performance of the Pauling rules.
Furthermore, excluding structures with coordination numbers lager than 8 mainly influences
the performance of the third rule: for all assessments including the third rule, the change of the
number of crystals fulfilling those rules is more drastic than for the assessment of the rules
two, four, and five where rule three is excluded. The rule with the lowest importance for the
fulfilment of all rules is the fifth rule. However, recall that we used a very loose version of this
rule for the assessment of the structures; we only considered coordination numbers and not
coordination environments and the number of their connections.

Conclusion

This is the first statistical analysis of all five Pauling rules. We find many exceptions to all rules.
Only 13% of all tested structures fulfil the rules 2 to 5 simultaneously .
The rule that is fulfilled worst is the second rule: only 20% of all oxygen atoms fulfil this rule on
local charge compensation nicely. As already predicted by Baur, we find a strong dependency
of the rule fulfillment on the distortion of the polyhedra: if only perfect environments are
considered, the rule is nearly perfectly fulfilled.
The fifth rule also shows many exceptions: many structures show more than one coordination
number of chemical equivalent cations (same element and valence).
For the third and fourth rule, we also see exceptions. They especially occur, if the preferred
coordination number of a certain cation is very high. These exceptions also correlate with the
atomic radius of the atoms; the higher the radius, the more exceptions. Furthermore, we do
not find the dependency on the valence for cations to not show connections, that is predicted
by the fourth rule. The different electrostatic interactions between the polyhedra with different
valences might not have such a strong influence as assumed by Pauling. Furthermore, metal-
metal bonds or magnetic interactions might also have an influence on the fulfillment of these
rules and should be considered in future rules. The dependency on the coordination number
was correctly predicted by Pauling.
We also find fundamental problems with the first rule: there are cations of the same element
and valence that show very different coordination environments (Ca2+). The radius-ratio rule
which only considers cation and anion radius cannot predict such an environment. Including a
dependency of the radii on the coordination number as it is present in the Shannon radii[6] does
not solve the issue: this would not be a prediction of the coordination environment any more.
Instead of using the radius-ratio rule, one could use statistics of the coordination environments
to check if the found coordination environment is usual or not. Future efforts should be devoted
to describe additional influences on the coordination environments within rules predicting the
coordination environments.
13
Based on the number of exceptions that we found, new rules that predict the stability of crystals
should be derived. Several very common structure types such as perovskites are among the
exceptions; they should be included in future rules. Furthermore, one should be very careful
using such rules for structure prediction. They only work for very few cases. One could overlook
the majority of important stable structures.

Methods:

We analysed the ab initio relaxed structures of around 5300 oxides from the Materials Project
Database that originally stem from the experimental part of the Inorganic Crystal Structure
Database (ICSD)[59] to ensure that no predicted structures are included in our analysis. We
only focused on materials with an energy above hull smaller or equal to 0.025 eV/atom (status
on 10.4.2019). This ab initio relaxation corrects for uncertainties of the bond lengths and
especially the coordination environments within the original structure predictions via X-ray or
neutron diffraction. We arrive at quantitatively similar results if we use purely experimental data
of 5000 oxides from the Crystallography Open Database (Figures S1-S6 in the Supporting
Information).[60-64] We ensured that no duplicates were present in this data set and we worked
with the primitive cells. The oxidation states are determined with the help of a bond-valence
analysis that is implemented in pymatgen.[65] If this analysis failed, we used the oxidation states
as given in the ICSD (only for the Materials Project data). To analyse the coordination
environments, we used the ChemEnv package with the default
MultiWeigthsChemEnvStrategy, that was developed in the previous study on the statistics of
coordination environments in oxides.[9] We applied it to a primitive unit cell of the oxide. To
analyse the Pauling rules, we then only used the coordination environment that was assigned
with the highest confidence (i.e. highest “ce_fraction”).
To test the first rule, we compared the coordination environments predicted by Pauling’s rule
with the coordination environments as evaluated by the ChemEnv package. We only
considered those environments for which the algorithm implemented in the ChemEnv
packages resulted in one of the 7 environments that Pauling first used and that are listed in
Table 1 (~ 41800 environments). To calculate the radius ratio, we used the univalent radii that
Pauling developed.[7] Those radii only exist for a limited number of cations and they have the
advantage that they do not depend on the coordination number and the valences.
To assess how well the first rule can work in principle, we calculated the Shannon entropy of
the environments for each element and divided it by the maximal possible Shannon entropy
(there are 66 different environments implemented).

Table 1. Environments and corresponding radius ratio windows that were considered in our
analysis of the radius ratio rule.

Environment Radius Ratio Window

Tetrahedron ≥0.225, <0.414
Octahedron ≥0.414, <0.592
Face-Capped Octahedron ≥0.592, <0.645
Square Antiprism ≥0.645, <0.732
Tricapped Triangular Prism or Cube ≥0.732, <1.0
Cuboctahedron ≥1.0

14
To assess the second rule, we calculated bond valence sums for each oxygen atom present
in our data set and compared them to the ideal valence of oxygen in an oxide (-2). To focus
on only very symmetric environments, we used environments with a continuous symmetry
measure smaller than 0.1 and a coordination environment fraction higher than 0.95. To arrive
at an elementwise analysis, we analysed the ratio of the number of cations around the oxygens
that fulfil the rule to the number of all cations of around investigated oxygens for each element.
To analyse the third rule, we focused on connected pairs of polyhedra within the stable oxides,
which show a cation-cation distance smaller or equal to 8 Å, and analysed their connections.
We defined exceptional structures as only those that show face connections. To arrive at an
elementwise analysis of the exceptions, we calculated the ratio of the number of cations that
showed corner and edge connections to the number of all cations that showed connections for
each element.
To analyse the fourth rule, we again focused on pairs of polyhedra, which show a cation-cation
distance smaller or equal to 8 Å, within stable oxides that have different cations (i.e. they differ
in valence or coordination number, ~ 5099 compounds), and analysed their connections as a
function of the coordination numbers or the valences in the polyhedra pairs. We defined
exceptions to this rule as structures where those cations with the smallest coordination number
and largest valence that is present in the structure, show connected polyhedra. The latter
version is the least strict interpretation of the fourth rule. We stress that other definitions of the
exceptions would have been possible. To arrive at an elementwise analysis of the exceptions,
we calculated the ratio of the number of cations with the highest valence and smallest
coordination number that showed no connections to the number of cations with the highest
valence and smallest coordination number for each element.
To analyse the fifth rule, we only looked at crystals having at least two chemical identical
cations (i.e. the same element with the same valence, 5161 compounds) within the primitive
unit cell. We then tested whether chemical identical cations showed the same coordination
number. For the elementwise analysis, we calculated the ratio of the number of chemical
identical cations that showed the same coordination number and the number of all chemical
identical cations for each element.
To analyse rules two to five, we furthermore assumed that the fourth and fifth rule are fulfilled
if the precondition for testing these rules is not fulfilled (i.e. they don’t have cations differing in
their valences or coordination numbers, or they don’t have more than one cation of the same
element and valence in the primitive unit cell). This ensured that we could also test compounds
such as α-SiO2 in the overall analysis (5276 compounds were analysed).
To guarantee a significant result, we only showed elements in the elementwise depiction of
the fulfillment of the rules for which more than 50 coordination environments (rules 1-5) or 25
structures including this element (combined assessment of rule 2-5) were present. The number
of environments that were considered in the study are shown in the Supporting Information
(see Figure S8). It demonstrates that conclusions drawn for P containing compounds are very
reliable.
In the case of the experimental structures, we used 50 structures as a limit for the elementwise
analysis of the combined assessment of the rules 2 to 5.
CSV Files including the structures fulfilling and not fulfilling the rules 2 to 5 are included in the
Supporting Information.
All crystal structures have been depicted with VESTA.[66]

15
Acknowledgements

We would like to thank all crystallographers and inorganic chemists filling the crystal
structure databases with entries. We acknowledge the Consortium des Équipements de
Calcul Intensif en Fédération Wallonie Bruxelles (CÉCI) for computational resources.

Keywords: Solid-state structures • Chemoinformatics • Computational Chemistry • Oxides •

Pauling’s Rules

[1] L. Pauling, J. Am. Chem. Soc. 1929, 51, 1010.

[2] W. H. Baur, Trans Am. Crystallogr. Assoc. 1970, 6, 129-155.
[3] N. Brese, M. O'Keeffe, Acta Cryst. B 1991, 47, 192–197.
[4] M. O'Keefe, N. E. Brese, J. Am. Chem. Soc. 1991, 113, 3226-3229.
[5] P. Beck Horst, Z. Kristallogr. - Cryst. Mater. 2013, 228, 271.
[6] R. Shannon, Acta Cryst. A 1976, 32, 751.
[7] L. Pauling, The Nature of the Chemical Bond and the Structure of Molecules and
Crystals: An Introduction to Modern Structural Chemistry, Cornell University Press,
1960.
[8] W. Peng, G. Petretto, G.-M. Rignanese, G. Hautier, A. Zevalkink, Joule 2018, 2,
1879-1893.
[9] D. Waroquiers, X. Gonze, G.-M. Rignanese, C. Welker-Nieuwoudt, F. Rosowski, M.
Göbel, S. Schenk, P. Degelmann, R. André, R. Glaum, G. Hautier, Chem. Mat. 2017,
29, 8346.
[10] R. Mooney, Acta Cryst. 1956, 9, 113-117.
[11] B. C. Frazer, P. J. Brown, Phys. Rev. 1962, 125, 1283-1291.
[12] F. Brisse, D. J. Stewart, V. Seidl, O. Knop, Can. J. Chem. 1972, 50, 3648-3666.
[13] N. Morimoto, D. E. Appleman, H. T. Evans Jr., Z. Kristallog. 1960, 114, 120.
[14] Y. A. Ivanov, M. A. Simonov, N. V. Belov, Kristallografiya 1974, 19, 163-164.
[15] E. Warren Β, Z. Kristallog. 1930, 74, 131.
[16] M. Cameron, J. J. Papike, Am. Mineral. 1981, 66, 1-50.
[17] W. S. McDonald, D. W. J. Cruickshank, Acta Cryst. 1967, 22, 48-51.
[18] D. Stachel, I. Svoboda, H. Fuess, Acta Cryst. C 1995, 51, 1049-1050.
[19] F. Théobald, R. Cabala, J. Bernard, J. Solid State Chem. 1976, 17, 431-438.
[20] T. S. Bush, C. R. A. Catlow, P. D. Battle, J. Mater. Chem. 1995, 5, 1269-1272.
[21] H. T. Evans, Acta Cryst. 1951, 4, 377.
[22] H. Megaw, Acta Cryst. 1962, 15, 972-973.
[23] V. Goldschmidt, Norske Videnskaps-Akademi. Matematisk-Naturvidenskapelige
Klasse 1926, 117.
[24] D. B. Rogers, R. D. Shannon, A. W. Sleight, J. L. Gillson, Inorg. Chem. 1969, 8, 841-
849.
[25] R. Lam, F. Wiss, J. E. Greedan, J. Solid State Chem. 2002, 167, 182-187.
[26] W. H. Bragg, J. Chem. Soc., Trans. 1922, 121, 2766-2787.
[27] L. Pauling, S. B. Hendricks, J. Am. Chem. Soc. 1925, 47, 781-790.
[28] V. M. Goldschmidt, T. Barth, G. Lunde, Norske Vidensk. Selsk. Skrifter I Mat. Naturv.
Kl. 1925, 5, 1.
[29] G. Andersson, Acta chem. scand 1954, 8, 1599-1606.
[30] H. Horiuchi, N. Morimoto, M. Tokonami, J. Solid State Chem. 1976, 17, 407-424.
[31] J.-L. Hodeau, M. Marezio, J. Solid State Chem. 1978, 23, 253-263.
[32] S. Anderson, L. Jahnberg, Ark. Kemi 1963, 21, 413.
[33] M. Weller, T. Overton, J. Rourke, F. Armstrong, Inorganic Chemistry, OUP Oxford,
2014.
[34] S. Goldsztaub, Bull. Minéral. 1935, 6-76.
[35] Z. Tomkowicz, B. van Laar, Phys. Status Solidi A 1974, 23, 683-688.
16
[36] T. McQueen, Q. Huang, J. W. Lynn, R. F. Berger, T. Klimczuk, B. G. Ueland, P.
Schiffer, R. J. Cava, Phys. Rev. B 2007, 76, 024420.
[37] M. Reehuis, C. Ulrich, P. Pattison, B. Ouladdiaf, M. C. Rheinstädter, M. Ohl, L. P.
Regnault, M. Miyasaka, Y. Tokura, B. Keimer, Phys. Rev. B 2006, 73, 094440.
[38] J. Bruns, M. Eul, R. Pöttgen, M. S. Wickleder, Angew. Chem. Int. Ed. 2012, 51, 2204-
2207.
[39] S. Geller, J. Chem. Phys. 1960, 33, 676-684.
[40] J. Ahman, G. Svensson, J. Albertsson, Acta Cryst. C 1996, 52, 1336-1338.
[41] H. C. Stumpf, A. S. Russell, J. W. Newsome, C. M. Tucker, J. Ind. Eng. Chem. 1950,
42, 1398-1403.
[42] L. Smrcok, V. Langer, M. Halvarsson, S. Ruppi, Z. Kristallog. 2001, 216, 409-412.
[43] M. Jansen, Angew. Chem. Int. Ed. 1977, 16, 314-315.
[44] F. Euler, J. A. Bruce, Acta Cryst. 1965, 19, 971-978.
[45] A. Zemann, J. Zemann, Acta Cryst. 1957, 10, 409-413.
[46] A. Wiesch, K. Bluhm, Acta Cryst. 1997, 53, 1730-1733.
[47] B. Bentria, D. Benbertal, M. Bagieu-Beucher, R. Masse, A. Mosset, J. Chem.
Crystallogr. 2003, 33, 867-873.
[48] S. N. Ruddlesden, P. Popper, Acta Cryst. 1958, 11, 54-55.
[49] E. Lucchini, D. Minichelli, G. Sloccari, Acta Cryst. B 1973, 29, 2356-2357.
[50] H. Müller-Buschbaum, J. Wilkens, Z. Anorg. Allg. Chem. 1990, 591, 161-166.
[51] V. M. Goldschmidt, zpch 1932, 17B, 172.
[52] W. H. Bragg, Nature 1915, 95, 561-561.
[53] L. Passerini, Gazz. Chim. Ital. 1929, 59, 280-288.
[54] R. Mooney, Acta Cryst. 1948, 1, 163-165.
[55] Y. Muraoka, K. Kihara, Phys. Chem. Miner. 1997, 24, 243-253.
[56] D. Schwarzenbach, Naturwissenschaften 1965, 52, 343-344.
[57] M. Krause, R. Gruehn, Z. Kristallog. 1995, 210, 427.
[58] W. L. Bragg, G. B. Brown, Z. Kristallog. - Cryst. Mater. 1926, 63, 538.
[59] R. Allmann, R. Hinek, Acta Cryst. A 2007, 63, 412-417.
[60] R. T. Downs, M. Hall-Wallace, Am. Mineral. 2003, 88, 247-250.
[61] S. Grazulis, D. Chateigner, R. T. Downs, A. F. T. Yokochi, M. Quiros, L. Lutterotti, E.
Manakova, J. Butkus, P. Moeck, A. Le Bail, J. Appl. Crystallogr. 2009, 42, 726-729.
[62] S. Gražulis, A. Daškevič, A. Merkys, D. Chateigner, L. Lutterotti, M. Quirós, N. R.
Serebryanaya, P. Moeck, R. T. Downs, A. Le Bail, Nucleic Acids Res. 2012, 40,
D420-D427.
[63] S. Gražulis, A. Merkys, A. Vaitkus, M. Okulič-Kazarinas, J. Appl. Crystallogr. 2015,
48, 85-91.
[64] A. Merkys, A. Vaitkus, J. Butkus, M. Okulič-Kazarinas, V. Kairys, S. Gražulis, J. Appl.
Cryst. 2016, 49.
[65] S. P. Ong, W. D. Richards, A. Jain, G. Hautier, M. Kocher, S. Cholia, D. Gunter, V. L.
Chevrier, K. A. Persson, G. Ceder, Comp. Mater. Sci. 2013, 68, 314.
[66] K. Momma, F. Izumi, J. Appl. Crystallogr. 2011, 44, 1272–1276.

17
18
Preprint_5_8_19.pdf (1.15 MiB) view on ChemRxiv download file
 Supporting Information for

The limited predictive power of the Pauling rules

Janine George,a David Waroquiers,a Davide Di Stefano,a Guido Petretto,a Gian-Marco
Rignanese,a and Geoffroy Hautier*a
a Institute of Condensed Matter and Nanosciences, Université Catholique de Louvain,
Chemin des étoiles 8, 1348 Louvain-la-Neuve, Belgium,
E-mail: geoffroy.hautier@uclouvain.be

1
Figure S1. Assessment of the first rule. This time based on the experimental structures from
the COD database.

2
Figure S2. Assessment of the second rule. This time based on the experimental structures
from the COD.

Figure S3. Assessment of the third rule based on the experimental structures from the COD.
3
Figure S4. Assessment of the fourth rule based on the experimental structures from the
COD.

4
Figure S5. Assessment of the fifth rule based on the experimental structures from the COD.

Figure S6. Assessment of the rules 2 to 5 based on the experimental structures.

5
Figure S7. Connections via corners and edges, and connection via faces of the polyhedra
pairs as a function of the atomic radius. One can see two different distributions.

6
Figure S8. Number of environments considered in our dataset that is based on the Materials
Project. a) Elements with number of environments between 0 and 2000. c) Elements with
number of environments between 2000 and 8500.

7
Supporting Information_5_8_19.pdf (768.20 KiB) view on ChemRxiv download file
Other files

CSV_Files_Supporting_Info_5_8_19.zip (1.45 MiB) view on ChemRxiv download file